summaryrefslogtreecommitdiffstats
path: root/module/zfs
diff options
context:
space:
mode:
Diffstat (limited to 'module/zfs')
-rw-r--r--module/zfs/arc.c4479
-rw-r--r--module/zfs/bplist.c349
-rw-r--r--module/zfs/dbuf.c2356
-rw-r--r--module/zfs/dmu.c1227
-rw-r--r--module/zfs/dmu_object.c162
-rw-r--r--module/zfs/dmu_objset.c1228
-rw-r--r--module/zfs/dmu_send.c1181
-rw-r--r--module/zfs/dmu_traverse.c406
-rw-r--r--module/zfs/dmu_tx.c1068
-rw-r--r--module/zfs/dmu_zfetch.c651
-rw-r--r--module/zfs/dnode.c1443
-rw-r--r--module/zfs/dnode_sync.c623
-rw-r--r--module/zfs/dsl_dataset.c3103
-rw-r--r--module/zfs/dsl_deleg.c735
-rw-r--r--module/zfs/dsl_dir.c1331
-rw-r--r--module/zfs/dsl_pool.c613
-rw-r--r--module/zfs/dsl_prop.c602
-rw-r--r--module/zfs/dsl_scrub.c1014
-rw-r--r--module/zfs/dsl_synctask.c225
-rw-r--r--module/zfs/fletcher.c145
-rw-r--r--module/zfs/gzip.c69
-rw-r--r--module/zfs/include/sys/arc.h138
-rw-r--r--module/zfs/include/sys/bplist.h89
-rw-r--r--module/zfs/include/sys/dbuf.h347
-rw-r--r--module/zfs/include/sys/dmu.h638
-rw-r--r--module/zfs/include/sys/dmu_impl.h239
-rw-r--r--module/zfs/include/sys/dmu_objset.h136
-rw-r--r--module/zfs/include/sys/dmu_traverse.h57
-rw-r--r--module/zfs/include/sys/dmu_tx.h139
-rw-r--r--module/zfs/include/sys/dmu_zfetch.h75
-rw-r--r--module/zfs/include/sys/dnode.h275
-rw-r--r--module/zfs/include/sys/dsl_dataset.h239
-rw-r--r--module/zfs/include/sys/dsl_deleg.h73
-rw-r--r--module/zfs/include/sys/dsl_dir.h160
-rw-r--r--module/zfs/include/sys/dsl_pool.h150
-rw-r--r--module/zfs/include/sys/dsl_prop.h82
-rw-r--r--module/zfs/include/sys/dsl_synctask.h83
-rw-r--r--module/zfs/include/sys/metaslab.h71
-rw-r--r--module/zfs/include/sys/metaslab_impl.h81
-rw-r--r--module/zfs/include/sys/refcount.h104
-rw-r--r--module/zfs/include/sys/rrwlock.h80
-rw-r--r--module/zfs/include/sys/spa.h554
-rw-r--r--module/zfs/include/sys/spa_boot.h45
-rw-r--r--module/zfs/include/sys/spa_impl.h196
-rw-r--r--module/zfs/include/sys/space_map.h162
-rw-r--r--module/zfs/include/sys/txg.h130
-rw-r--r--module/zfs/include/sys/txg_impl.h73
-rw-r--r--module/zfs/include/sys/uberblock.h50
-rw-r--r--module/zfs/include/sys/uberblock_impl.h63
-rw-r--r--module/zfs/include/sys/unique.h59
-rw-r--r--module/zfs/include/sys/vdev.h135
-rw-r--r--module/zfs/include/sys/vdev_file.h46
-rw-r--r--module/zfs/include/sys/vdev_impl.h305
-rw-r--r--module/zfs/include/sys/zap.h425
-rw-r--r--module/zfs/include/sys/zap_impl.h218
-rw-r--r--module/zfs/include/sys/zap_leaf.h244
-rw-r--r--module/zfs/include/sys/zfs_acl.h214
-rw-r--r--module/zfs/include/sys/zfs_context.h73
-rw-r--r--module/zfs/include/sys/zfs_ctldir.h74
-rw-r--r--module/zfs/include/sys/zfs_debug.h75
-rw-r--r--module/zfs/include/sys/zfs_dir.h76
-rw-r--r--module/zfs/include/sys/zfs_fuid.h125
-rw-r--r--module/zfs/include/sys/zfs_ioctl.h196
-rw-r--r--module/zfs/include/sys/zfs_rlock.h89
-rw-r--r--module/zfs/include/sys/zfs_vfsops.h140
-rw-r--r--module/zfs/include/sys/zfs_znode.h356
-rw-r--r--module/zfs/include/sys/zil.h382
-rw-r--r--module/zfs/include/sys/zil_impl.h109
-rw-r--r--module/zfs/include/sys/zio.h424
-rw-r--r--module/zfs/include/sys/zio_checksum.h73
-rw-r--r--module/zfs/include/sys/zio_compress.h82
-rw-r--r--module/zfs/include/sys/zio_impl.h143
-rw-r--r--module/zfs/include/sys/zvol.h70
-rw-r--r--module/zfs/lzjb.c128
-rw-r--r--module/zfs/metaslab.c1049
-rw-r--r--module/zfs/refcount.c195
-rw-r--r--module/zfs/rrwlock.c249
-rw-r--r--module/zfs/sha256.c129
-rw-r--r--module/zfs/spa.c4301
-rw-r--r--module/zfs/spa_boot.c47
-rw-r--r--module/zfs/spa_config.c444
-rw-r--r--module/zfs/spa_errlog.c437
-rw-r--r--module/zfs/spa_history.c428
-rw-r--r--module/zfs/spa_misc.c1410
-rw-r--r--module/zfs/space_map.c506
-rw-r--r--module/zfs/txg.c627
-rw-r--r--module/zfs/uberblock.c63
-rw-r--r--module/zfs/unique.c116
-rw-r--r--module/zfs/vdev.c2425
-rw-r--r--module/zfs/vdev_cache.c425
-rw-r--r--module/zfs/vdev_file.c188
-rw-r--r--module/zfs/vdev_label.c1078
-rw-r--r--module/zfs/vdev_mirror.c480
-rw-r--r--module/zfs/vdev_missing.c85
-rw-r--r--module/zfs/vdev_queue.c308
-rw-r--r--module/zfs/vdev_raidz.c1209
-rw-r--r--module/zfs/vdev_root.c118
-rw-r--r--module/zfs/zap.c1136
-rw-r--r--module/zfs/zap_leaf.c853
-rw-r--r--module/zfs/zap_micro.c1069
-rw-r--r--module/zfs/zfs_acl.c2680
-rw-r--r--module/zfs/zfs_byteswap.c175
-rw-r--r--module/zfs/zfs_ctldir.c1159
-rw-r--r--module/zfs/zfs_dir.c977
-rw-r--r--module/zfs/zfs_fm.c362
-rw-r--r--module/zfs/zfs_fuid.c704
-rw-r--r--module/zfs/zfs_ioctl.c3175
-rw-r--r--module/zfs/zfs_log.c694
-rw-r--r--module/zfs/zfs_replay.c878
-rw-r--r--module/zfs/zfs_rlock.c602
-rw-r--r--module/zfs/zfs_vfsops.c1652
-rw-r--r--module/zfs/zfs_vnops.c4561
-rw-r--r--module/zfs/zfs_znode.c1672
-rw-r--r--module/zfs/zil.c1735
-rw-r--r--module/zfs/zio.c2273
-rw-r--r--module/zfs/zio_checksum.c206
-rw-r--r--module/zfs/zio_compress.c148
-rw-r--r--module/zfs/zio_inject.c370
-rw-r--r--module/zfs/zvol.c1722
119 files changed, 76920 insertions, 0 deletions
diff --git a/module/zfs/arc.c b/module/zfs/arc.c
new file mode 100644
index 000000000..73aecb285
--- /dev/null
+++ b/module/zfs/arc.c
@@ -0,0 +1,4479 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * DVA-based Adjustable Replacement Cache
+ *
+ * While much of the theory of operation used here is
+ * based on the self-tuning, low overhead replacement cache
+ * presented by Megiddo and Modha at FAST 2003, there are some
+ * significant differences:
+ *
+ * 1. The Megiddo and Modha model assumes any page is evictable.
+ * Pages in its cache cannot be "locked" into memory. This makes
+ * the eviction algorithm simple: evict the last page in the list.
+ * This also make the performance characteristics easy to reason
+ * about. Our cache is not so simple. At any given moment, some
+ * subset of the blocks in the cache are un-evictable because we
+ * have handed out a reference to them. Blocks are only evictable
+ * when there are no external references active. This makes
+ * eviction far more problematic: we choose to evict the evictable
+ * blocks that are the "lowest" in the list.
+ *
+ * There are times when it is not possible to evict the requested
+ * space. In these circumstances we are unable to adjust the cache
+ * size. To prevent the cache growing unbounded at these times we
+ * implement a "cache throttle" that slows the flow of new data
+ * into the cache until we can make space available.
+ *
+ * 2. The Megiddo and Modha model assumes a fixed cache size.
+ * Pages are evicted when the cache is full and there is a cache
+ * miss. Our model has a variable sized cache. It grows with
+ * high use, but also tries to react to memory pressure from the
+ * operating system: decreasing its size when system memory is
+ * tight.
+ *
+ * 3. The Megiddo and Modha model assumes a fixed page size. All
+ * elements of the cache are therefor exactly the same size. So
+ * when adjusting the cache size following a cache miss, its simply
+ * a matter of choosing a single page to evict. In our model, we
+ * have variable sized cache blocks (rangeing from 512 bytes to
+ * 128K bytes). We therefor choose a set of blocks to evict to make
+ * space for a cache miss that approximates as closely as possible
+ * the space used by the new block.
+ *
+ * See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache"
+ * by N. Megiddo & D. Modha, FAST 2003
+ */
+
+/*
+ * The locking model:
+ *
+ * A new reference to a cache buffer can be obtained in two
+ * ways: 1) via a hash table lookup using the DVA as a key,
+ * or 2) via one of the ARC lists. The arc_read() interface
+ * uses method 1, while the internal arc algorithms for
+ * adjusting the cache use method 2. We therefor provide two
+ * types of locks: 1) the hash table lock array, and 2) the
+ * arc list locks.
+ *
+ * Buffers do not have their own mutexs, rather they rely on the
+ * hash table mutexs for the bulk of their protection (i.e. most
+ * fields in the arc_buf_hdr_t are protected by these mutexs).
+ *
+ * buf_hash_find() returns the appropriate mutex (held) when it
+ * locates the requested buffer in the hash table. It returns
+ * NULL for the mutex if the buffer was not in the table.
+ *
+ * buf_hash_remove() expects the appropriate hash mutex to be
+ * already held before it is invoked.
+ *
+ * Each arc state also has a mutex which is used to protect the
+ * buffer list associated with the state. When attempting to
+ * obtain a hash table lock while holding an arc list lock you
+ * must use: mutex_tryenter() to avoid deadlock. Also note that
+ * the active state mutex must be held before the ghost state mutex.
+ *
+ * Arc buffers may have an associated eviction callback function.
+ * This function will be invoked prior to removing the buffer (e.g.
+ * in arc_do_user_evicts()). Note however that the data associated
+ * with the buffer may be evicted prior to the callback. The callback
+ * must be made with *no locks held* (to prevent deadlock). Additionally,
+ * the users of callbacks must ensure that their private data is
+ * protected from simultaneous callbacks from arc_buf_evict()
+ * and arc_do_user_evicts().
+ *
+ * Note that the majority of the performance stats are manipulated
+ * with atomic operations.
+ *
+ * The L2ARC uses the l2arc_buflist_mtx global mutex for the following:
+ *
+ * - L2ARC buflist creation
+ * - L2ARC buflist eviction
+ * - L2ARC write completion, which walks L2ARC buflists
+ * - ARC header destruction, as it removes from L2ARC buflists
+ * - ARC header release, as it removes from L2ARC buflists
+ */
+
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+#include <sys/zfs_context.h>
+#include <sys/arc.h>
+#include <sys/refcount.h>
+#include <sys/vdev.h>
+#ifdef _KERNEL
+#include <sys/vmsystm.h>
+#include <vm/anon.h>
+#include <sys/fs/swapnode.h>
+#include <sys/dnlc.h>
+#endif
+#include <sys/callb.h>
+#include <sys/kstat.h>
+
+static kmutex_t arc_reclaim_thr_lock;
+static kcondvar_t arc_reclaim_thr_cv; /* used to signal reclaim thr */
+static uint8_t arc_thread_exit;
+
+extern int zfs_write_limit_shift;
+extern uint64_t zfs_write_limit_max;
+extern kmutex_t zfs_write_limit_lock;
+
+#define ARC_REDUCE_DNLC_PERCENT 3
+uint_t arc_reduce_dnlc_percent = ARC_REDUCE_DNLC_PERCENT;
+
+typedef enum arc_reclaim_strategy {
+ ARC_RECLAIM_AGGR, /* Aggressive reclaim strategy */
+ ARC_RECLAIM_CONS /* Conservative reclaim strategy */
+} arc_reclaim_strategy_t;
+
+/* number of seconds before growing cache again */
+static int arc_grow_retry = 60;
+
+/*
+ * minimum lifespan of a prefetch block in clock ticks
+ * (initialized in arc_init())
+ */
+static int arc_min_prefetch_lifespan;
+
+static int arc_dead;
+
+/*
+ * The arc has filled available memory and has now warmed up.
+ */
+static boolean_t arc_warm;
+
+/*
+ * These tunables are for performance analysis.
+ */
+uint64_t zfs_arc_max;
+uint64_t zfs_arc_min;
+uint64_t zfs_arc_meta_limit = 0;
+int zfs_mdcomp_disable = 0;
+
+/*
+ * Note that buffers can be in one of 6 states:
+ * ARC_anon - anonymous (discussed below)
+ * ARC_mru - recently used, currently cached
+ * ARC_mru_ghost - recentely used, no longer in cache
+ * ARC_mfu - frequently used, currently cached
+ * ARC_mfu_ghost - frequently used, no longer in cache
+ * ARC_l2c_only - exists in L2ARC but not other states
+ * When there are no active references to the buffer, they are
+ * are linked onto a list in one of these arc states. These are
+ * the only buffers that can be evicted or deleted. Within each
+ * state there are multiple lists, one for meta-data and one for
+ * non-meta-data. Meta-data (indirect blocks, blocks of dnodes,
+ * etc.) is tracked separately so that it can be managed more
+ * explicitly: favored over data, limited explicitly.
+ *
+ * Anonymous buffers are buffers that are not associated with
+ * a DVA. These are buffers that hold dirty block copies
+ * before they are written to stable storage. By definition,
+ * they are "ref'd" and are considered part of arc_mru
+ * that cannot be freed. Generally, they will aquire a DVA
+ * as they are written and migrate onto the arc_mru list.
+ *
+ * The ARC_l2c_only state is for buffers that are in the second
+ * level ARC but no longer in any of the ARC_m* lists. The second
+ * level ARC itself may also contain buffers that are in any of
+ * the ARC_m* states - meaning that a buffer can exist in two
+ * places. The reason for the ARC_l2c_only state is to keep the
+ * buffer header in the hash table, so that reads that hit the
+ * second level ARC benefit from these fast lookups.
+ */
+
+typedef struct arc_state {
+ list_t arcs_list[ARC_BUFC_NUMTYPES]; /* list of evictable buffers */
+ uint64_t arcs_lsize[ARC_BUFC_NUMTYPES]; /* amount of evictable data */
+ uint64_t arcs_size; /* total amount of data in this state */
+ kmutex_t arcs_mtx;
+} arc_state_t;
+
+/* The 6 states: */
+static arc_state_t ARC_anon;
+static arc_state_t ARC_mru;
+static arc_state_t ARC_mru_ghost;
+static arc_state_t ARC_mfu;
+static arc_state_t ARC_mfu_ghost;
+static arc_state_t ARC_l2c_only;
+
+typedef struct arc_stats {
+ kstat_named_t arcstat_hits;
+ kstat_named_t arcstat_misses;
+ kstat_named_t arcstat_demand_data_hits;
+ kstat_named_t arcstat_demand_data_misses;
+ kstat_named_t arcstat_demand_metadata_hits;
+ kstat_named_t arcstat_demand_metadata_misses;
+ kstat_named_t arcstat_prefetch_data_hits;
+ kstat_named_t arcstat_prefetch_data_misses;
+ kstat_named_t arcstat_prefetch_metadata_hits;
+ kstat_named_t arcstat_prefetch_metadata_misses;
+ kstat_named_t arcstat_mru_hits;
+ kstat_named_t arcstat_mru_ghost_hits;
+ kstat_named_t arcstat_mfu_hits;
+ kstat_named_t arcstat_mfu_ghost_hits;
+ kstat_named_t arcstat_deleted;
+ kstat_named_t arcstat_recycle_miss;
+ kstat_named_t arcstat_mutex_miss;
+ kstat_named_t arcstat_evict_skip;
+ kstat_named_t arcstat_hash_elements;
+ kstat_named_t arcstat_hash_elements_max;
+ kstat_named_t arcstat_hash_collisions;
+ kstat_named_t arcstat_hash_chains;
+ kstat_named_t arcstat_hash_chain_max;
+ kstat_named_t arcstat_p;
+ kstat_named_t arcstat_c;
+ kstat_named_t arcstat_c_min;
+ kstat_named_t arcstat_c_max;
+ kstat_named_t arcstat_size;
+ kstat_named_t arcstat_hdr_size;
+ kstat_named_t arcstat_l2_hits;
+ kstat_named_t arcstat_l2_misses;
+ kstat_named_t arcstat_l2_feeds;
+ kstat_named_t arcstat_l2_rw_clash;
+ kstat_named_t arcstat_l2_writes_sent;
+ kstat_named_t arcstat_l2_writes_done;
+ kstat_named_t arcstat_l2_writes_error;
+ kstat_named_t arcstat_l2_writes_hdr_miss;
+ kstat_named_t arcstat_l2_evict_lock_retry;
+ kstat_named_t arcstat_l2_evict_reading;
+ kstat_named_t arcstat_l2_free_on_write;
+ kstat_named_t arcstat_l2_abort_lowmem;
+ kstat_named_t arcstat_l2_cksum_bad;
+ kstat_named_t arcstat_l2_io_error;
+ kstat_named_t arcstat_l2_size;
+ kstat_named_t arcstat_l2_hdr_size;
+ kstat_named_t arcstat_memory_throttle_count;
+} arc_stats_t;
+
+static arc_stats_t arc_stats = {
+ { "hits", KSTAT_DATA_UINT64 },
+ { "misses", KSTAT_DATA_UINT64 },
+ { "demand_data_hits", KSTAT_DATA_UINT64 },
+ { "demand_data_misses", KSTAT_DATA_UINT64 },
+ { "demand_metadata_hits", KSTAT_DATA_UINT64 },
+ { "demand_metadata_misses", KSTAT_DATA_UINT64 },
+ { "prefetch_data_hits", KSTAT_DATA_UINT64 },
+ { "prefetch_data_misses", KSTAT_DATA_UINT64 },
+ { "prefetch_metadata_hits", KSTAT_DATA_UINT64 },
+ { "prefetch_metadata_misses", KSTAT_DATA_UINT64 },
+ { "mru_hits", KSTAT_DATA_UINT64 },
+ { "mru_ghost_hits", KSTAT_DATA_UINT64 },
+ { "mfu_hits", KSTAT_DATA_UINT64 },
+ { "mfu_ghost_hits", KSTAT_DATA_UINT64 },
+ { "deleted", KSTAT_DATA_UINT64 },
+ { "recycle_miss", KSTAT_DATA_UINT64 },
+ { "mutex_miss", KSTAT_DATA_UINT64 },
+ { "evict_skip", KSTAT_DATA_UINT64 },
+ { "hash_elements", KSTAT_DATA_UINT64 },
+ { "hash_elements_max", KSTAT_DATA_UINT64 },
+ { "hash_collisions", KSTAT_DATA_UINT64 },
+ { "hash_chains", KSTAT_DATA_UINT64 },
+ { "hash_chain_max", KSTAT_DATA_UINT64 },
+ { "p", KSTAT_DATA_UINT64 },
+ { "c", KSTAT_DATA_UINT64 },
+ { "c_min", KSTAT_DATA_UINT64 },
+ { "c_max", KSTAT_DATA_UINT64 },
+ { "size", KSTAT_DATA_UINT64 },
+ { "hdr_size", KSTAT_DATA_UINT64 },
+ { "l2_hits", KSTAT_DATA_UINT64 },
+ { "l2_misses", KSTAT_DATA_UINT64 },
+ { "l2_feeds", KSTAT_DATA_UINT64 },
+ { "l2_rw_clash", KSTAT_DATA_UINT64 },
+ { "l2_writes_sent", KSTAT_DATA_UINT64 },
+ { "l2_writes_done", KSTAT_DATA_UINT64 },
+ { "l2_writes_error", KSTAT_DATA_UINT64 },
+ { "l2_writes_hdr_miss", KSTAT_DATA_UINT64 },
+ { "l2_evict_lock_retry", KSTAT_DATA_UINT64 },
+ { "l2_evict_reading", KSTAT_DATA_UINT64 },
+ { "l2_free_on_write", KSTAT_DATA_UINT64 },
+ { "l2_abort_lowmem", KSTAT_DATA_UINT64 },
+ { "l2_cksum_bad", KSTAT_DATA_UINT64 },
+ { "l2_io_error", KSTAT_DATA_UINT64 },
+ { "l2_size", KSTAT_DATA_UINT64 },
+ { "l2_hdr_size", KSTAT_DATA_UINT64 },
+ { "memory_throttle_count", KSTAT_DATA_UINT64 }
+};
+
+#define ARCSTAT(stat) (arc_stats.stat.value.ui64)
+
+#define ARCSTAT_INCR(stat, val) \
+ atomic_add_64(&arc_stats.stat.value.ui64, (val));
+
+#define ARCSTAT_BUMP(stat) ARCSTAT_INCR(stat, 1)
+#define ARCSTAT_BUMPDOWN(stat) ARCSTAT_INCR(stat, -1)
+
+#define ARCSTAT_MAX(stat, val) { \
+ uint64_t m; \
+ while ((val) > (m = arc_stats.stat.value.ui64) && \
+ (m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \
+ continue; \
+}
+
+#define ARCSTAT_MAXSTAT(stat) \
+ ARCSTAT_MAX(stat##_max, arc_stats.stat.value.ui64)
+
+/*
+ * We define a macro to allow ARC hits/misses to be easily broken down by
+ * two separate conditions, giving a total of four different subtypes for
+ * each of hits and misses (so eight statistics total).
+ */
+#define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \
+ if (cond1) { \
+ if (cond2) { \
+ ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \
+ } else { \
+ ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \
+ } \
+ } else { \
+ if (cond2) { \
+ ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \
+ } else { \
+ ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\
+ } \
+ }
+
+kstat_t *arc_ksp;
+static arc_state_t *arc_anon;
+static arc_state_t *arc_mru;
+static arc_state_t *arc_mru_ghost;
+static arc_state_t *arc_mfu;
+static arc_state_t *arc_mfu_ghost;
+static arc_state_t *arc_l2c_only;
+
+/*
+ * There are several ARC variables that are critical to export as kstats --
+ * but we don't want to have to grovel around in the kstat whenever we wish to
+ * manipulate them. For these variables, we therefore define them to be in
+ * terms of the statistic variable. This assures that we are not introducing
+ * the possibility of inconsistency by having shadow copies of the variables,
+ * while still allowing the code to be readable.
+ */
+#define arc_size ARCSTAT(arcstat_size) /* actual total arc size */
+#define arc_p ARCSTAT(arcstat_p) /* target size of MRU */
+#define arc_c ARCSTAT(arcstat_c) /* target size of cache */
+#define arc_c_min ARCSTAT(arcstat_c_min) /* min target cache size */
+#define arc_c_max ARCSTAT(arcstat_c_max) /* max target cache size */
+
+static int arc_no_grow; /* Don't try to grow cache size */
+static uint64_t arc_tempreserve;
+static uint64_t arc_meta_used;
+static uint64_t arc_meta_limit;
+static uint64_t arc_meta_max = 0;
+
+typedef struct l2arc_buf_hdr l2arc_buf_hdr_t;
+
+typedef struct arc_callback arc_callback_t;
+
+struct arc_callback {
+ void *acb_private;
+ arc_done_func_t *acb_done;
+ arc_buf_t *acb_buf;
+ zio_t *acb_zio_dummy;
+ arc_callback_t *acb_next;
+};
+
+typedef struct arc_write_callback arc_write_callback_t;
+
+struct arc_write_callback {
+ void *awcb_private;
+ arc_done_func_t *awcb_ready;
+ arc_done_func_t *awcb_done;
+ arc_buf_t *awcb_buf;
+};
+
+struct arc_buf_hdr {
+ /* protected by hash lock */
+ dva_t b_dva;
+ uint64_t b_birth;
+ uint64_t b_cksum0;
+
+ kmutex_t b_freeze_lock;
+ zio_cksum_t *b_freeze_cksum;
+
+ arc_buf_hdr_t *b_hash_next;
+ arc_buf_t *b_buf;
+ uint32_t b_flags;
+ uint32_t b_datacnt;
+
+ arc_callback_t *b_acb;
+ kcondvar_t b_cv;
+
+ /* immutable */
+ arc_buf_contents_t b_type;
+ uint64_t b_size;
+ spa_t *b_spa;
+
+ /* protected by arc state mutex */
+ arc_state_t *b_state;
+ list_node_t b_arc_node;
+
+ /* updated atomically */
+ clock_t b_arc_access;
+
+ /* self protecting */
+ refcount_t b_refcnt;
+
+ l2arc_buf_hdr_t *b_l2hdr;
+ list_node_t b_l2node;
+};
+
+static arc_buf_t *arc_eviction_list;
+static kmutex_t arc_eviction_mtx;
+static arc_buf_hdr_t arc_eviction_hdr;
+static void arc_get_data_buf(arc_buf_t *buf);
+static void arc_access(arc_buf_hdr_t *buf, kmutex_t *hash_lock);
+static int arc_evict_needed(arc_buf_contents_t type);
+static void arc_evict_ghost(arc_state_t *state, spa_t *spa, int64_t bytes);
+
+#define GHOST_STATE(state) \
+ ((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \
+ (state) == arc_l2c_only)
+
+/*
+ * Private ARC flags. These flags are private ARC only flags that will show up
+ * in b_flags in the arc_hdr_buf_t. Some flags are publicly declared, and can
+ * be passed in as arc_flags in things like arc_read. However, these flags
+ * should never be passed and should only be set by ARC code. When adding new
+ * public flags, make sure not to smash the private ones.
+ */
+
+#define ARC_IN_HASH_TABLE (1 << 9) /* this buffer is hashed */
+#define ARC_IO_IN_PROGRESS (1 << 10) /* I/O in progress for buf */
+#define ARC_IO_ERROR (1 << 11) /* I/O failed for buf */
+#define ARC_FREED_IN_READ (1 << 12) /* buf freed while in read */
+#define ARC_BUF_AVAILABLE (1 << 13) /* block not in active use */
+#define ARC_INDIRECT (1 << 14) /* this is an indirect block */
+#define ARC_FREE_IN_PROGRESS (1 << 15) /* hdr about to be freed */
+#define ARC_L2_WRITING (1 << 16) /* L2ARC write in progress */
+#define ARC_L2_EVICTED (1 << 17) /* evicted during I/O */
+#define ARC_L2_WRITE_HEAD (1 << 18) /* head of write list */
+#define ARC_STORED (1 << 19) /* has been store()d to */
+
+#define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_IN_HASH_TABLE)
+#define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_IO_IN_PROGRESS)
+#define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_IO_ERROR)
+#define HDR_FREED_IN_READ(hdr) ((hdr)->b_flags & ARC_FREED_IN_READ)
+#define HDR_BUF_AVAILABLE(hdr) ((hdr)->b_flags & ARC_BUF_AVAILABLE)
+#define HDR_FREE_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FREE_IN_PROGRESS)
+#define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_L2CACHE)
+#define HDR_L2_READING(hdr) ((hdr)->b_flags & ARC_IO_IN_PROGRESS && \
+ (hdr)->b_l2hdr != NULL)
+#define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_L2_WRITING)
+#define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_L2_EVICTED)
+#define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_L2_WRITE_HEAD)
+
+/*
+ * Other sizes
+ */
+
+#define HDR_SIZE ((int64_t)sizeof (arc_buf_hdr_t))
+#define L2HDR_SIZE ((int64_t)sizeof (l2arc_buf_hdr_t))
+
+/*
+ * Hash table routines
+ */
+
+#define HT_LOCK_PAD 64
+
+struct ht_lock {
+ kmutex_t ht_lock;
+#ifdef _KERNEL
+ unsigned char pad[(HT_LOCK_PAD - sizeof (kmutex_t))];
+#endif
+};
+
+#define BUF_LOCKS 256
+typedef struct buf_hash_table {
+ uint64_t ht_mask;
+ arc_buf_hdr_t **ht_table;
+ struct ht_lock ht_locks[BUF_LOCKS];
+} buf_hash_table_t;
+
+static buf_hash_table_t buf_hash_table;
+
+#define BUF_HASH_INDEX(spa, dva, birth) \
+ (buf_hash(spa, dva, birth) & buf_hash_table.ht_mask)
+#define BUF_HASH_LOCK_NTRY(idx) (buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)])
+#define BUF_HASH_LOCK(idx) (&(BUF_HASH_LOCK_NTRY(idx).ht_lock))
+#define HDR_LOCK(buf) \
+ (BUF_HASH_LOCK(BUF_HASH_INDEX(buf->b_spa, &buf->b_dva, buf->b_birth)))
+
+uint64_t zfs_crc64_table[256];
+
+/*
+ * Level 2 ARC
+ */
+
+#define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */
+#define L2ARC_HEADROOM 4 /* num of writes */
+#define L2ARC_FEED_SECS 1 /* caching interval */
+
+#define l2arc_writes_sent ARCSTAT(arcstat_l2_writes_sent)
+#define l2arc_writes_done ARCSTAT(arcstat_l2_writes_done)
+
+/*
+ * L2ARC Performance Tunables
+ */
+uint64_t l2arc_write_max = L2ARC_WRITE_SIZE; /* default max write size */
+uint64_t l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra write during warmup */
+uint64_t l2arc_headroom = L2ARC_HEADROOM; /* number of dev writes */
+uint64_t l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */
+boolean_t l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */
+
+/*
+ * L2ARC Internals
+ */
+typedef struct l2arc_dev {
+ vdev_t *l2ad_vdev; /* vdev */
+ spa_t *l2ad_spa; /* spa */
+ uint64_t l2ad_hand; /* next write location */
+ uint64_t l2ad_write; /* desired write size, bytes */
+ uint64_t l2ad_boost; /* warmup write boost, bytes */
+ uint64_t l2ad_start; /* first addr on device */
+ uint64_t l2ad_end; /* last addr on device */
+ uint64_t l2ad_evict; /* last addr eviction reached */
+ boolean_t l2ad_first; /* first sweep through */
+ list_t *l2ad_buflist; /* buffer list */
+ list_node_t l2ad_node; /* device list node */
+} l2arc_dev_t;
+
+static list_t L2ARC_dev_list; /* device list */
+static list_t *l2arc_dev_list; /* device list pointer */
+static kmutex_t l2arc_dev_mtx; /* device list mutex */
+static l2arc_dev_t *l2arc_dev_last; /* last device used */
+static kmutex_t l2arc_buflist_mtx; /* mutex for all buflists */
+static list_t L2ARC_free_on_write; /* free after write buf list */
+static list_t *l2arc_free_on_write; /* free after write list ptr */
+static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */
+static uint64_t l2arc_ndev; /* number of devices */
+
+typedef struct l2arc_read_callback {
+ arc_buf_t *l2rcb_buf; /* read buffer */
+ spa_t *l2rcb_spa; /* spa */
+ blkptr_t l2rcb_bp; /* original blkptr */
+ zbookmark_t l2rcb_zb; /* original bookmark */
+ int l2rcb_flags; /* original flags */
+} l2arc_read_callback_t;
+
+typedef struct l2arc_write_callback {
+ l2arc_dev_t *l2wcb_dev; /* device info */
+ arc_buf_hdr_t *l2wcb_head; /* head of write buflist */
+} l2arc_write_callback_t;
+
+struct l2arc_buf_hdr {
+ /* protected by arc_buf_hdr mutex */
+ l2arc_dev_t *b_dev; /* L2ARC device */
+ daddr_t b_daddr; /* disk address, offset byte */
+};
+
+typedef struct l2arc_data_free {
+ /* protected by l2arc_free_on_write_mtx */
+ void *l2df_data;
+ size_t l2df_size;
+ void (*l2df_func)(void *, size_t);
+ list_node_t l2df_list_node;
+} l2arc_data_free_t;
+
+static kmutex_t l2arc_feed_thr_lock;
+static kcondvar_t l2arc_feed_thr_cv;
+static uint8_t l2arc_thread_exit;
+
+static void l2arc_read_done(zio_t *zio);
+static void l2arc_hdr_stat_add(void);
+static void l2arc_hdr_stat_remove(void);
+
+static uint64_t
+buf_hash(spa_t *spa, const dva_t *dva, uint64_t birth)
+{
+ uintptr_t spav = (uintptr_t)spa;
+ uint8_t *vdva = (uint8_t *)dva;
+ uint64_t crc = -1ULL;
+ int i;
+
+ ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
+
+ for (i = 0; i < sizeof (dva_t); i++)
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ vdva[i]) & 0xFF];
+
+ crc ^= (spav>>8) ^ birth;
+
+ return (crc);
+}
+
+#define BUF_EMPTY(buf) \
+ ((buf)->b_dva.dva_word[0] == 0 && \
+ (buf)->b_dva.dva_word[1] == 0 && \
+ (buf)->b_birth == 0)
+
+#define BUF_EQUAL(spa, dva, birth, buf) \
+ ((buf)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
+ ((buf)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
+ ((buf)->b_birth == birth) && ((buf)->b_spa == spa)
+
+static arc_buf_hdr_t *
+buf_hash_find(spa_t *spa, const dva_t *dva, uint64_t birth, kmutex_t **lockp)
+{
+ uint64_t idx = BUF_HASH_INDEX(spa, dva, birth);
+ kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
+ arc_buf_hdr_t *buf;
+
+ mutex_enter(hash_lock);
+ for (buf = buf_hash_table.ht_table[idx]; buf != NULL;
+ buf = buf->b_hash_next) {
+ if (BUF_EQUAL(spa, dva, birth, buf)) {
+ *lockp = hash_lock;
+ return (buf);
+ }
+ }
+ mutex_exit(hash_lock);
+ *lockp = NULL;
+ return (NULL);
+}
+
+/*
+ * Insert an entry into the hash table. If there is already an element
+ * equal to elem in the hash table, then the already existing element
+ * will be returned and the new element will not be inserted.
+ * Otherwise returns NULL.
+ */
+static arc_buf_hdr_t *
+buf_hash_insert(arc_buf_hdr_t *buf, kmutex_t **lockp)
+{
+ uint64_t idx = BUF_HASH_INDEX(buf->b_spa, &buf->b_dva, buf->b_birth);
+ kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
+ arc_buf_hdr_t *fbuf;
+ uint32_t i;
+
+ ASSERT(!HDR_IN_HASH_TABLE(buf));
+ *lockp = hash_lock;
+ mutex_enter(hash_lock);
+ for (fbuf = buf_hash_table.ht_table[idx], i = 0; fbuf != NULL;
+ fbuf = fbuf->b_hash_next, i++) {
+ if (BUF_EQUAL(buf->b_spa, &buf->b_dva, buf->b_birth, fbuf))
+ return (fbuf);
+ }
+
+ buf->b_hash_next = buf_hash_table.ht_table[idx];
+ buf_hash_table.ht_table[idx] = buf;
+ buf->b_flags |= ARC_IN_HASH_TABLE;
+
+ /* collect some hash table performance data */
+ if (i > 0) {
+ ARCSTAT_BUMP(arcstat_hash_collisions);
+ if (i == 1)
+ ARCSTAT_BUMP(arcstat_hash_chains);
+
+ ARCSTAT_MAX(arcstat_hash_chain_max, i);
+ }
+
+ ARCSTAT_BUMP(arcstat_hash_elements);
+ ARCSTAT_MAXSTAT(arcstat_hash_elements);
+
+ return (NULL);
+}
+
+static void
+buf_hash_remove(arc_buf_hdr_t *buf)
+{
+ arc_buf_hdr_t *fbuf, **bufp;
+ uint64_t idx = BUF_HASH_INDEX(buf->b_spa, &buf->b_dva, buf->b_birth);
+
+ ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx)));
+ ASSERT(HDR_IN_HASH_TABLE(buf));
+
+ bufp = &buf_hash_table.ht_table[idx];
+ while ((fbuf = *bufp) != buf) {
+ ASSERT(fbuf != NULL);
+ bufp = &fbuf->b_hash_next;
+ }
+ *bufp = buf->b_hash_next;
+ buf->b_hash_next = NULL;
+ buf->b_flags &= ~ARC_IN_HASH_TABLE;
+
+ /* collect some hash table performance data */
+ ARCSTAT_BUMPDOWN(arcstat_hash_elements);
+
+ if (buf_hash_table.ht_table[idx] &&
+ buf_hash_table.ht_table[idx]->b_hash_next == NULL)
+ ARCSTAT_BUMPDOWN(arcstat_hash_chains);
+}
+
+/*
+ * Global data structures and functions for the buf kmem cache.
+ */
+static kmem_cache_t *hdr_cache;
+static kmem_cache_t *buf_cache;
+
+static void
+buf_fini(void)
+{
+ int i;
+
+ kmem_free(buf_hash_table.ht_table,
+ (buf_hash_table.ht_mask + 1) * sizeof (void *));
+ for (i = 0; i < BUF_LOCKS; i++)
+ mutex_destroy(&buf_hash_table.ht_locks[i].ht_lock);
+ kmem_cache_destroy(hdr_cache);
+ kmem_cache_destroy(buf_cache);
+}
+
+/*
+ * Constructor callback - called when the cache is empty
+ * and a new buf is requested.
+ */
+/* ARGSUSED */
+static int
+hdr_cons(void *vbuf, void *unused, int kmflag)
+{
+ arc_buf_hdr_t *buf = vbuf;
+
+ bzero(buf, sizeof (arc_buf_hdr_t));
+ refcount_create(&buf->b_refcnt);
+ cv_init(&buf->b_cv, NULL, CV_DEFAULT, NULL);
+ mutex_init(&buf->b_freeze_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ ARCSTAT_INCR(arcstat_hdr_size, HDR_SIZE);
+ return (0);
+}
+
+/* ARGSUSED */
+static int
+buf_cons(void *vbuf, void *unused, int kmflag)
+{
+ arc_buf_t *buf = vbuf;
+
+ bzero(buf, sizeof (arc_buf_t));
+ rw_init(&buf->b_lock, NULL, RW_DEFAULT, NULL);
+ return (0);
+}
+
+/*
+ * Destructor callback - called when a cached buf is
+ * no longer required.
+ */
+/* ARGSUSED */
+static void
+hdr_dest(void *vbuf, void *unused)
+{
+ arc_buf_hdr_t *buf = vbuf;
+
+ refcount_destroy(&buf->b_refcnt);
+ cv_destroy(&buf->b_cv);
+ mutex_destroy(&buf->b_freeze_lock);
+
+ ARCSTAT_INCR(arcstat_hdr_size, -HDR_SIZE);
+}
+
+/* ARGSUSED */
+static void
+buf_dest(void *vbuf, void *unused)
+{
+ arc_buf_t *buf = vbuf;
+
+ rw_destroy(&buf->b_lock);
+}
+
+/*
+ * Reclaim callback -- invoked when memory is low.
+ */
+/* ARGSUSED */
+static void
+hdr_recl(void *unused)
+{
+ dprintf("hdr_recl called\n");
+ /*
+ * umem calls the reclaim func when we destroy the buf cache,
+ * which is after we do arc_fini().
+ */
+ if (!arc_dead)
+ cv_signal(&arc_reclaim_thr_cv);
+}
+
+static void
+buf_init(void)
+{
+ uint64_t *ct;
+ uint64_t hsize = 1ULL << 12;
+ int i, j;
+
+ /*
+ * The hash table is big enough to fill all of physical memory
+ * with an average 64K block size. The table will take up
+ * totalmem*sizeof(void*)/64K (eg. 128KB/GB with 8-byte pointers).
+ */
+ while (hsize * 65536 < physmem * PAGESIZE)
+ hsize <<= 1;
+retry:
+ buf_hash_table.ht_mask = hsize - 1;
+ buf_hash_table.ht_table =
+ kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP);
+ if (buf_hash_table.ht_table == NULL) {
+ ASSERT(hsize > (1ULL << 8));
+ hsize >>= 1;
+ goto retry;
+ }
+
+ hdr_cache = kmem_cache_create("arc_buf_hdr_t", sizeof (arc_buf_hdr_t),
+ 0, hdr_cons, hdr_dest, hdr_recl, NULL, NULL, 0);
+ buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t),
+ 0, buf_cons, buf_dest, NULL, NULL, NULL, 0);
+
+ for (i = 0; i < 256; i++)
+ for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--)
+ *ct = (*ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY);
+
+ for (i = 0; i < BUF_LOCKS; i++) {
+ mutex_init(&buf_hash_table.ht_locks[i].ht_lock,
+ NULL, MUTEX_DEFAULT, NULL);
+ }
+}
+
+#define ARC_MINTIME (hz>>4) /* 62 ms */
+
+static void
+arc_cksum_verify(arc_buf_t *buf)
+{
+ zio_cksum_t zc;
+
+ if (!(zfs_flags & ZFS_DEBUG_MODIFY))
+ return;
+
+ mutex_enter(&buf->b_hdr->b_freeze_lock);
+ if (buf->b_hdr->b_freeze_cksum == NULL ||
+ (buf->b_hdr->b_flags & ARC_IO_ERROR)) {
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+ return;
+ }
+ fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc);
+ if (!ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc))
+ panic("buffer modified while frozen!");
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+}
+
+static int
+arc_cksum_equal(arc_buf_t *buf)
+{
+ zio_cksum_t zc;
+ int equal;
+
+ mutex_enter(&buf->b_hdr->b_freeze_lock);
+ fletcher_2_native(buf->b_data, buf->b_hdr->b_size, &zc);
+ equal = ZIO_CHECKSUM_EQUAL(*buf->b_hdr->b_freeze_cksum, zc);
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+
+ return (equal);
+}
+
+static void
+arc_cksum_compute(arc_buf_t *buf, boolean_t force)
+{
+ if (!force && !(zfs_flags & ZFS_DEBUG_MODIFY))
+ return;
+
+ mutex_enter(&buf->b_hdr->b_freeze_lock);
+ if (buf->b_hdr->b_freeze_cksum != NULL) {
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+ return;
+ }
+ buf->b_hdr->b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t), KM_SLEEP);
+ fletcher_2_native(buf->b_data, buf->b_hdr->b_size,
+ buf->b_hdr->b_freeze_cksum);
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+}
+
+void
+arc_buf_thaw(arc_buf_t *buf)
+{
+ if (zfs_flags & ZFS_DEBUG_MODIFY) {
+ if (buf->b_hdr->b_state != arc_anon)
+ panic("modifying non-anon buffer!");
+ if (buf->b_hdr->b_flags & ARC_IO_IN_PROGRESS)
+ panic("modifying buffer while i/o in progress!");
+ arc_cksum_verify(buf);
+ }
+
+ mutex_enter(&buf->b_hdr->b_freeze_lock);
+ if (buf->b_hdr->b_freeze_cksum != NULL) {
+ kmem_free(buf->b_hdr->b_freeze_cksum, sizeof (zio_cksum_t));
+ buf->b_hdr->b_freeze_cksum = NULL;
+ }
+ mutex_exit(&buf->b_hdr->b_freeze_lock);
+}
+
+void
+arc_buf_freeze(arc_buf_t *buf)
+{
+ if (!(zfs_flags & ZFS_DEBUG_MODIFY))
+ return;
+
+ ASSERT(buf->b_hdr->b_freeze_cksum != NULL ||
+ buf->b_hdr->b_state == arc_anon);
+ arc_cksum_compute(buf, B_FALSE);
+}
+
+static void
+add_reference(arc_buf_hdr_t *ab, kmutex_t *hash_lock, void *tag)
+{
+ ASSERT(MUTEX_HELD(hash_lock));
+
+ if ((refcount_add(&ab->b_refcnt, tag) == 1) &&
+ (ab->b_state != arc_anon)) {
+ uint64_t delta = ab->b_size * ab->b_datacnt;
+ list_t *list = &ab->b_state->arcs_list[ab->b_type];
+ uint64_t *size = &ab->b_state->arcs_lsize[ab->b_type];
+
+ ASSERT(!MUTEX_HELD(&ab->b_state->arcs_mtx));
+ mutex_enter(&ab->b_state->arcs_mtx);
+ ASSERT(list_link_active(&ab->b_arc_node));
+ list_remove(list, ab);
+ if (GHOST_STATE(ab->b_state)) {
+ ASSERT3U(ab->b_datacnt, ==, 0);
+ ASSERT3P(ab->b_buf, ==, NULL);
+ delta = ab->b_size;
+ }
+ ASSERT(delta > 0);
+ ASSERT3U(*size, >=, delta);
+ atomic_add_64(size, -delta);
+ mutex_exit(&ab->b_state->arcs_mtx);
+ /* remove the prefetch flag if we get a reference */
+ if (ab->b_flags & ARC_PREFETCH)
+ ab->b_flags &= ~ARC_PREFETCH;
+ }
+}
+
+static int
+remove_reference(arc_buf_hdr_t *ab, kmutex_t *hash_lock, void *tag)
+{
+ int cnt;
+ arc_state_t *state = ab->b_state;
+
+ ASSERT(state == arc_anon || MUTEX_HELD(hash_lock));
+ ASSERT(!GHOST_STATE(state));
+
+ if (((cnt = refcount_remove(&ab->b_refcnt, tag)) == 0) &&
+ (state != arc_anon)) {
+ uint64_t *size = &state->arcs_lsize[ab->b_type];
+
+ ASSERT(!MUTEX_HELD(&state->arcs_mtx));
+ mutex_enter(&state->arcs_mtx);
+ ASSERT(!list_link_active(&ab->b_arc_node));
+ list_insert_head(&state->arcs_list[ab->b_type], ab);
+ ASSERT(ab->b_datacnt > 0);
+ atomic_add_64(size, ab->b_size * ab->b_datacnt);
+ mutex_exit(&state->arcs_mtx);
+ }
+ return (cnt);
+}
+
+/*
+ * Move the supplied buffer to the indicated state. The mutex
+ * for the buffer must be held by the caller.
+ */
+static void
+arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *ab, kmutex_t *hash_lock)
+{
+ arc_state_t *old_state = ab->b_state;
+ int64_t refcnt = refcount_count(&ab->b_refcnt);
+ uint64_t from_delta, to_delta;
+
+ ASSERT(MUTEX_HELD(hash_lock));
+ ASSERT(new_state != old_state);
+ ASSERT(refcnt == 0 || ab->b_datacnt > 0);
+ ASSERT(ab->b_datacnt == 0 || !GHOST_STATE(new_state));
+
+ from_delta = to_delta = ab->b_datacnt * ab->b_size;
+
+ /*
+ * If this buffer is evictable, transfer it from the
+ * old state list to the new state list.
+ */
+ if (refcnt == 0) {
+ if (old_state != arc_anon) {
+ int use_mutex = !MUTEX_HELD(&old_state->arcs_mtx);
+ uint64_t *size = &old_state->arcs_lsize[ab->b_type];
+
+ if (use_mutex)
+ mutex_enter(&old_state->arcs_mtx);
+
+ ASSERT(list_link_active(&ab->b_arc_node));
+ list_remove(&old_state->arcs_list[ab->b_type], ab);
+
+ /*
+ * If prefetching out of the ghost cache,
+ * we will have a non-null datacnt.
+ */
+ if (GHOST_STATE(old_state) && ab->b_datacnt == 0) {
+ /* ghost elements have a ghost size */
+ ASSERT(ab->b_buf == NULL);
+ from_delta = ab->b_size;
+ }
+ ASSERT3U(*size, >=, from_delta);
+ atomic_add_64(size, -from_delta);
+
+ if (use_mutex)
+ mutex_exit(&old_state->arcs_mtx);
+ }
+ if (new_state != arc_anon) {
+ int use_mutex = !MUTEX_HELD(&new_state->arcs_mtx);
+ uint64_t *size = &new_state->arcs_lsize[ab->b_type];
+
+ if (use_mutex)
+ mutex_enter(&new_state->arcs_mtx);
+
+ list_insert_head(&new_state->arcs_list[ab->b_type], ab);
+
+ /* ghost elements have a ghost size */
+ if (GHOST_STATE(new_state)) {
+ ASSERT(ab->b_datacnt == 0);
+ ASSERT(ab->b_buf == NULL);
+ to_delta = ab->b_size;
+ }
+ atomic_add_64(size, to_delta);
+
+ if (use_mutex)
+ mutex_exit(&new_state->arcs_mtx);
+ }
+ }
+
+ ASSERT(!BUF_EMPTY(ab));
+ if (new_state == arc_anon) {
+ buf_hash_remove(ab);
+ }
+
+ /* adjust state sizes */
+ if (to_delta)
+ atomic_add_64(&new_state->arcs_size, to_delta);
+ if (from_delta) {
+ ASSERT3U(old_state->arcs_size, >=, from_delta);
+ atomic_add_64(&old_state->arcs_size, -from_delta);
+ }
+ ab->b_state = new_state;
+
+ /* adjust l2arc hdr stats */
+ if (new_state == arc_l2c_only)
+ l2arc_hdr_stat_add();
+ else if (old_state == arc_l2c_only)
+ l2arc_hdr_stat_remove();
+}
+
+void
+arc_space_consume(uint64_t space)
+{
+ atomic_add_64(&arc_meta_used, space);
+ atomic_add_64(&arc_size, space);
+}
+
+void
+arc_space_return(uint64_t space)
+{
+ ASSERT(arc_meta_used >= space);
+ if (arc_meta_max < arc_meta_used)
+ arc_meta_max = arc_meta_used;
+ atomic_add_64(&arc_meta_used, -space);
+ ASSERT(arc_size >= space);
+ atomic_add_64(&arc_size, -space);
+}
+
+void *
+arc_data_buf_alloc(uint64_t size)
+{
+ if (arc_evict_needed(ARC_BUFC_DATA))
+ cv_signal(&arc_reclaim_thr_cv);
+ atomic_add_64(&arc_size, size);
+ return (zio_data_buf_alloc(size));
+}
+
+void
+arc_data_buf_free(void *buf, uint64_t size)
+{
+ zio_data_buf_free(buf, size);
+ ASSERT(arc_size >= size);
+ atomic_add_64(&arc_size, -size);
+}
+
+arc_buf_t *
+arc_buf_alloc(spa_t *spa, int size, void *tag, arc_buf_contents_t type)
+{
+ arc_buf_hdr_t *hdr;
+ arc_buf_t *buf;
+
+ ASSERT3U(size, >, 0);
+ hdr = kmem_cache_alloc(hdr_cache, KM_PUSHPAGE);
+ ASSERT(BUF_EMPTY(hdr));
+ hdr->b_size = size;
+ hdr->b_type = type;
+ hdr->b_spa = spa;
+ hdr->b_state = arc_anon;
+ hdr->b_arc_access = 0;
+ buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
+ buf->b_hdr = hdr;
+ buf->b_data = NULL;
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+ buf->b_next = NULL;
+ hdr->b_buf = buf;
+ arc_get_data_buf(buf);
+ hdr->b_datacnt = 1;
+ hdr->b_flags = 0;
+ ASSERT(refcount_is_zero(&hdr->b_refcnt));
+ (void) refcount_add(&hdr->b_refcnt, tag);
+
+ return (buf);
+}
+
+static arc_buf_t *
+arc_buf_clone(arc_buf_t *from)
+{
+ arc_buf_t *buf;
+ arc_buf_hdr_t *hdr = from->b_hdr;
+ uint64_t size = hdr->b_size;
+
+ buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
+ buf->b_hdr = hdr;
+ buf->b_data = NULL;
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+ buf->b_next = hdr->b_buf;
+ hdr->b_buf = buf;
+ arc_get_data_buf(buf);
+ bcopy(from->b_data, buf->b_data, size);
+ hdr->b_datacnt += 1;
+ return (buf);
+}
+
+void
+arc_buf_add_ref(arc_buf_t *buf, void* tag)
+{
+ arc_buf_hdr_t *hdr;
+ kmutex_t *hash_lock;
+
+ /*
+ * Check to see if this buffer is evicted. Callers
+ * must verify b_data != NULL to know if the add_ref
+ * was successful.
+ */
+ rw_enter(&buf->b_lock, RW_READER);
+ if (buf->b_data == NULL) {
+ rw_exit(&buf->b_lock);
+ return;
+ }
+ hdr = buf->b_hdr;
+ ASSERT(hdr != NULL);
+ hash_lock = HDR_LOCK(hdr);
+ mutex_enter(hash_lock);
+ rw_exit(&buf->b_lock);
+
+ ASSERT(hdr->b_state == arc_mru || hdr->b_state == arc_mfu);
+ add_reference(hdr, hash_lock, tag);
+ arc_access(hdr, hash_lock);
+ mutex_exit(hash_lock);
+ ARCSTAT_BUMP(arcstat_hits);
+ ARCSTAT_CONDSTAT(!(hdr->b_flags & ARC_PREFETCH),
+ demand, prefetch, hdr->b_type != ARC_BUFC_METADATA,
+ data, metadata, hits);
+}
+
+/*
+ * Free the arc data buffer. If it is an l2arc write in progress,
+ * the buffer is placed on l2arc_free_on_write to be freed later.
+ */
+static void
+arc_buf_data_free(arc_buf_hdr_t *hdr, void (*free_func)(void *, size_t),
+ void *data, size_t size)
+{
+ if (HDR_L2_WRITING(hdr)) {
+ l2arc_data_free_t *df;
+ df = kmem_alloc(sizeof (l2arc_data_free_t), KM_SLEEP);
+ df->l2df_data = data;
+ df->l2df_size = size;
+ df->l2df_func = free_func;
+ mutex_enter(&l2arc_free_on_write_mtx);
+ list_insert_head(l2arc_free_on_write, df);
+ mutex_exit(&l2arc_free_on_write_mtx);
+ ARCSTAT_BUMP(arcstat_l2_free_on_write);
+ } else {
+ free_func(data, size);
+ }
+}
+
+static void
+arc_buf_destroy(arc_buf_t *buf, boolean_t recycle, boolean_t all)
+{
+ arc_buf_t **bufp;
+
+ /* free up data associated with the buf */
+ if (buf->b_data) {
+ arc_state_t *state = buf->b_hdr->b_state;
+ uint64_t size = buf->b_hdr->b_size;
+ arc_buf_contents_t type = buf->b_hdr->b_type;
+
+ arc_cksum_verify(buf);
+ if (!recycle) {
+ if (type == ARC_BUFC_METADATA) {
+ arc_buf_data_free(buf->b_hdr, zio_buf_free,
+ buf->b_data, size);
+ arc_space_return(size);
+ } else {
+ ASSERT(type == ARC_BUFC_DATA);
+ arc_buf_data_free(buf->b_hdr,
+ zio_data_buf_free, buf->b_data, size);
+ atomic_add_64(&arc_size, -size);
+ }
+ }
+ if (list_link_active(&buf->b_hdr->b_arc_node)) {
+ uint64_t *cnt = &state->arcs_lsize[type];
+
+ ASSERT(refcount_is_zero(&buf->b_hdr->b_refcnt));
+ ASSERT(state != arc_anon);
+
+ ASSERT3U(*cnt, >=, size);
+ atomic_add_64(cnt, -size);
+ }
+ ASSERT3U(state->arcs_size, >=, size);
+ atomic_add_64(&state->arcs_size, -size);
+ buf->b_data = NULL;
+ ASSERT(buf->b_hdr->b_datacnt > 0);
+ buf->b_hdr->b_datacnt -= 1;
+ }
+
+ /* only remove the buf if requested */
+ if (!all)
+ return;
+
+ /* remove the buf from the hdr list */
+ for (bufp = &buf->b_hdr->b_buf; *bufp != buf; bufp = &(*bufp)->b_next)
+ continue;
+ *bufp = buf->b_next;
+
+ ASSERT(buf->b_efunc == NULL);
+
+ /* clean up the buf */
+ buf->b_hdr = NULL;
+ kmem_cache_free(buf_cache, buf);
+}
+
+static void
+arc_hdr_destroy(arc_buf_hdr_t *hdr)
+{
+ ASSERT(refcount_is_zero(&hdr->b_refcnt));
+ ASSERT3P(hdr->b_state, ==, arc_anon);
+ ASSERT(!HDR_IO_IN_PROGRESS(hdr));
+ ASSERT(!(hdr->b_flags & ARC_STORED));
+
+ if (hdr->b_l2hdr != NULL) {
+ if (!MUTEX_HELD(&l2arc_buflist_mtx)) {
+ /*
+ * To prevent arc_free() and l2arc_evict() from
+ * attempting to free the same buffer at the same time,
+ * a FREE_IN_PROGRESS flag is given to arc_free() to
+ * give it priority. l2arc_evict() can't destroy this
+ * header while we are waiting on l2arc_buflist_mtx.
+ *
+ * The hdr may be removed from l2ad_buflist before we
+ * grab l2arc_buflist_mtx, so b_l2hdr is rechecked.
+ */
+ mutex_enter(&l2arc_buflist_mtx);
+ if (hdr->b_l2hdr != NULL) {
+ list_remove(hdr->b_l2hdr->b_dev->l2ad_buflist,
+ hdr);
+ }
+ mutex_exit(&l2arc_buflist_mtx);
+ } else {
+ list_remove(hdr->b_l2hdr->b_dev->l2ad_buflist, hdr);
+ }
+ ARCSTAT_INCR(arcstat_l2_size, -hdr->b_size);
+ kmem_free(hdr->b_l2hdr, sizeof (l2arc_buf_hdr_t));
+ if (hdr->b_state == arc_l2c_only)
+ l2arc_hdr_stat_remove();
+ hdr->b_l2hdr = NULL;
+ }
+
+ if (!BUF_EMPTY(hdr)) {
+ ASSERT(!HDR_IN_HASH_TABLE(hdr));
+ bzero(&hdr->b_dva, sizeof (dva_t));
+ hdr->b_birth = 0;
+ hdr->b_cksum0 = 0;
+ }
+ while (hdr->b_buf) {
+ arc_buf_t *buf = hdr->b_buf;
+
+ if (buf->b_efunc) {
+ mutex_enter(&arc_eviction_mtx);
+ rw_enter(&buf->b_lock, RW_WRITER);
+ ASSERT(buf->b_hdr != NULL);
+ arc_buf_destroy(hdr->b_buf, FALSE, FALSE);
+ hdr->b_buf = buf->b_next;
+ buf->b_hdr = &arc_eviction_hdr;
+ buf->b_next = arc_eviction_list;
+ arc_eviction_list = buf;
+ rw_exit(&buf->b_lock);
+ mutex_exit(&arc_eviction_mtx);
+ } else {
+ arc_buf_destroy(hdr->b_buf, FALSE, TRUE);
+ }
+ }
+ if (hdr->b_freeze_cksum != NULL) {
+ kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t));
+ hdr->b_freeze_cksum = NULL;
+ }
+
+ ASSERT(!list_link_active(&hdr->b_arc_node));
+ ASSERT3P(hdr->b_hash_next, ==, NULL);
+ ASSERT3P(hdr->b_acb, ==, NULL);
+ kmem_cache_free(hdr_cache, hdr);
+}
+
+void
+arc_buf_free(arc_buf_t *buf, void *tag)
+{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+ int hashed = hdr->b_state != arc_anon;
+
+ ASSERT(buf->b_efunc == NULL);
+ ASSERT(buf->b_data != NULL);
+
+ if (hashed) {
+ kmutex_t *hash_lock = HDR_LOCK(hdr);
+
+ mutex_enter(hash_lock);
+ (void) remove_reference(hdr, hash_lock, tag);
+ if (hdr->b_datacnt > 1)
+ arc_buf_destroy(buf, FALSE, TRUE);
+ else
+ hdr->b_flags |= ARC_BUF_AVAILABLE;
+ mutex_exit(hash_lock);
+ } else if (HDR_IO_IN_PROGRESS(hdr)) {
+ int destroy_hdr;
+ /*
+ * We are in the middle of an async write. Don't destroy
+ * this buffer unless the write completes before we finish
+ * decrementing the reference count.
+ */
+ mutex_enter(&arc_eviction_mtx);
+ (void) remove_reference(hdr, NULL, tag);
+ ASSERT(refcount_is_zero(&hdr->b_refcnt));
+ destroy_hdr = !HDR_IO_IN_PROGRESS(hdr);
+ mutex_exit(&arc_eviction_mtx);
+ if (destroy_hdr)
+ arc_hdr_destroy(hdr);
+ } else {
+ if (remove_reference(hdr, NULL, tag) > 0) {
+ ASSERT(HDR_IO_ERROR(hdr));
+ arc_buf_destroy(buf, FALSE, TRUE);
+ } else {
+ arc_hdr_destroy(hdr);
+ }
+ }
+}
+
+int
+arc_buf_remove_ref(arc_buf_t *buf, void* tag)
+{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+ kmutex_t *hash_lock = HDR_LOCK(hdr);
+ int no_callback = (buf->b_efunc == NULL);
+
+ if (hdr->b_state == arc_anon) {
+ arc_buf_free(buf, tag);
+ return (no_callback);
+ }
+
+ mutex_enter(hash_lock);
+ ASSERT(hdr->b_state != arc_anon);
+ ASSERT(buf->b_data != NULL);
+
+ (void) remove_reference(hdr, hash_lock, tag);
+ if (hdr->b_datacnt > 1) {
+ if (no_callback)
+ arc_buf_destroy(buf, FALSE, TRUE);
+ } else if (no_callback) {
+ ASSERT(hdr->b_buf == buf && buf->b_next == NULL);
+ hdr->b_flags |= ARC_BUF_AVAILABLE;
+ }
+ ASSERT(no_callback || hdr->b_datacnt > 1 ||
+ refcount_is_zero(&hdr->b_refcnt));
+ mutex_exit(hash_lock);
+ return (no_callback);
+}
+
+int
+arc_buf_size(arc_buf_t *buf)
+{
+ return (buf->b_hdr->b_size);
+}
+
+/*
+ * Evict buffers from list until we've removed the specified number of
+ * bytes. Move the removed buffers to the appropriate evict state.
+ * If the recycle flag is set, then attempt to "recycle" a buffer:
+ * - look for a buffer to evict that is `bytes' long.
+ * - return the data block from this buffer rather than freeing it.
+ * This flag is used by callers that are trying to make space for a
+ * new buffer in a full arc cache.
+ *
+ * This function makes a "best effort". It skips over any buffers
+ * it can't get a hash_lock on, and so may not catch all candidates.
+ * It may also return without evicting as much space as requested.
+ */
+static void *
+arc_evict(arc_state_t *state, spa_t *spa, int64_t bytes, boolean_t recycle,
+ arc_buf_contents_t type)
+{
+ arc_state_t *evicted_state;
+ uint64_t bytes_evicted = 0, skipped = 0, missed = 0;
+ arc_buf_hdr_t *ab, *ab_prev = NULL;
+ list_t *list = &state->arcs_list[type];
+ kmutex_t *hash_lock;
+ boolean_t have_lock;
+ void *stolen = NULL;
+
+ ASSERT(state == arc_mru || state == arc_mfu);
+
+ evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost;
+
+ mutex_enter(&state->arcs_mtx);
+ mutex_enter(&evicted_state->arcs_mtx);
+
+ for (ab = list_tail(list); ab; ab = ab_prev) {
+ ab_prev = list_prev(list, ab);
+ /* prefetch buffers have a minimum lifespan */
+ if (HDR_IO_IN_PROGRESS(ab) ||
+ (spa && ab->b_spa != spa) ||
+ (ab->b_flags & (ARC_PREFETCH|ARC_INDIRECT) &&
+ lbolt - ab->b_arc_access < arc_min_prefetch_lifespan)) {
+ skipped++;
+ continue;
+ }
+ /* "lookahead" for better eviction candidate */
+ if (recycle && ab->b_size != bytes &&
+ ab_prev && ab_prev->b_size == bytes)
+ continue;
+ hash_lock = HDR_LOCK(ab);
+ have_lock = MUTEX_HELD(hash_lock);
+ if (have_lock || mutex_tryenter(hash_lock)) {
+ ASSERT3U(refcount_count(&ab->b_refcnt), ==, 0);
+ ASSERT(ab->b_datacnt > 0);
+ while (ab->b_buf) {
+ arc_buf_t *buf = ab->b_buf;
+ if (!rw_tryenter(&buf->b_lock, RW_WRITER)) {
+ missed += 1;
+ break;
+ }
+ if (buf->b_data) {
+ bytes_evicted += ab->b_size;
+ if (recycle && ab->b_type == type &&
+ ab->b_size == bytes &&
+ !HDR_L2_WRITING(ab)) {
+ stolen = buf->b_data;
+ recycle = FALSE;
+ }
+ }
+ if (buf->b_efunc) {
+ mutex_enter(&arc_eviction_mtx);
+ arc_buf_destroy(buf,
+ buf->b_data == stolen, FALSE);
+ ab->b_buf = buf->b_next;
+ buf->b_hdr = &arc_eviction_hdr;
+ buf->b_next = arc_eviction_list;
+ arc_eviction_list = buf;
+ mutex_exit(&arc_eviction_mtx);
+ rw_exit(&buf->b_lock);
+ } else {
+ rw_exit(&buf->b_lock);
+ arc_buf_destroy(buf,
+ buf->b_data == stolen, TRUE);
+ }
+ }
+ if (ab->b_datacnt == 0) {
+ arc_change_state(evicted_state, ab, hash_lock);
+ ASSERT(HDR_IN_HASH_TABLE(ab));
+ ab->b_flags |= ARC_IN_HASH_TABLE;
+ ab->b_flags &= ~ARC_BUF_AVAILABLE;
+ DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, ab);
+ }
+ if (!have_lock)
+ mutex_exit(hash_lock);
+ if (bytes >= 0 && bytes_evicted >= bytes)
+ break;
+ } else {
+ missed += 1;
+ }
+ }
+
+ mutex_exit(&evicted_state->arcs_mtx);
+ mutex_exit(&state->arcs_mtx);
+
+ if (bytes_evicted < bytes)
+ dprintf("only evicted %lld bytes from %x",
+ (longlong_t)bytes_evicted, state);
+
+ if (skipped)
+ ARCSTAT_INCR(arcstat_evict_skip, skipped);
+
+ if (missed)
+ ARCSTAT_INCR(arcstat_mutex_miss, missed);
+
+ /*
+ * We have just evicted some date into the ghost state, make
+ * sure we also adjust the ghost state size if necessary.
+ */
+ if (arc_no_grow &&
+ arc_mru_ghost->arcs_size + arc_mfu_ghost->arcs_size > arc_c) {
+ int64_t mru_over = arc_anon->arcs_size + arc_mru->arcs_size +
+ arc_mru_ghost->arcs_size - arc_c;
+
+ if (mru_over > 0 && arc_mru_ghost->arcs_lsize[type] > 0) {
+ int64_t todelete =
+ MIN(arc_mru_ghost->arcs_lsize[type], mru_over);
+ arc_evict_ghost(arc_mru_ghost, NULL, todelete);
+ } else if (arc_mfu_ghost->arcs_lsize[type] > 0) {
+ int64_t todelete = MIN(arc_mfu_ghost->arcs_lsize[type],
+ arc_mru_ghost->arcs_size +
+ arc_mfu_ghost->arcs_size - arc_c);
+ arc_evict_ghost(arc_mfu_ghost, NULL, todelete);
+ }
+ }
+
+ return (stolen);
+}
+
+/*
+ * Remove buffers from list until we've removed the specified number of
+ * bytes. Destroy the buffers that are removed.
+ */
+static void
+arc_evict_ghost(arc_state_t *state, spa_t *spa, int64_t bytes)
+{
+ arc_buf_hdr_t *ab, *ab_prev;
+ list_t *list = &state->arcs_list[ARC_BUFC_DATA];
+ kmutex_t *hash_lock;
+ uint64_t bytes_deleted = 0;
+ uint64_t bufs_skipped = 0;
+
+ ASSERT(GHOST_STATE(state));
+top:
+ mutex_enter(&state->arcs_mtx);
+ for (ab = list_tail(list); ab; ab = ab_prev) {
+ ab_prev = list_prev(list, ab);
+ if (spa && ab->b_spa != spa)
+ continue;
+ hash_lock = HDR_LOCK(ab);
+ if (mutex_tryenter(hash_lock)) {
+ ASSERT(!HDR_IO_IN_PROGRESS(ab));
+ ASSERT(ab->b_buf == NULL);
+ ARCSTAT_BUMP(arcstat_deleted);
+ bytes_deleted += ab->b_size;
+
+ if (ab->b_l2hdr != NULL) {
+ /*
+ * This buffer is cached on the 2nd Level ARC;
+ * don't destroy the header.
+ */
+ arc_change_state(arc_l2c_only, ab, hash_lock);
+ mutex_exit(hash_lock);
+ } else {
+ arc_change_state(arc_anon, ab, hash_lock);
+ mutex_exit(hash_lock);
+ arc_hdr_destroy(ab);
+ }
+
+ DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, ab);
+ if (bytes >= 0 && bytes_deleted >= bytes)
+ break;
+ } else {
+ if (bytes < 0) {
+ mutex_exit(&state->arcs_mtx);
+ mutex_enter(hash_lock);
+ mutex_exit(hash_lock);
+ goto top;
+ }
+ bufs_skipped += 1;
+ }
+ }
+ mutex_exit(&state->arcs_mtx);
+
+ if (list == &state->arcs_list[ARC_BUFC_DATA] &&
+ (bytes < 0 || bytes_deleted < bytes)) {
+ list = &state->arcs_list[ARC_BUFC_METADATA];
+ goto top;
+ }
+
+ if (bufs_skipped) {
+ ARCSTAT_INCR(arcstat_mutex_miss, bufs_skipped);
+ ASSERT(bytes >= 0);
+ }
+
+ if (bytes_deleted < bytes)
+ dprintf("only deleted %lld bytes from %p",
+ (longlong_t)bytes_deleted, state);
+}
+
+static void
+arc_adjust(void)
+{
+ int64_t top_sz, mru_over, arc_over, todelete;
+
+ top_sz = arc_anon->arcs_size + arc_mru->arcs_size + arc_meta_used;
+
+ if (top_sz > arc_p && arc_mru->arcs_lsize[ARC_BUFC_DATA] > 0) {
+ int64_t toevict =
+ MIN(arc_mru->arcs_lsize[ARC_BUFC_DATA], top_sz - arc_p);
+ (void) arc_evict(arc_mru, NULL, toevict, FALSE, ARC_BUFC_DATA);
+ top_sz = arc_anon->arcs_size + arc_mru->arcs_size;
+ }
+
+ if (top_sz > arc_p && arc_mru->arcs_lsize[ARC_BUFC_METADATA] > 0) {
+ int64_t toevict =
+ MIN(arc_mru->arcs_lsize[ARC_BUFC_METADATA], top_sz - arc_p);
+ (void) arc_evict(arc_mru, NULL, toevict, FALSE,
+ ARC_BUFC_METADATA);
+ top_sz = arc_anon->arcs_size + arc_mru->arcs_size;
+ }
+
+ mru_over = top_sz + arc_mru_ghost->arcs_size - arc_c;
+
+ if (mru_over > 0) {
+ if (arc_mru_ghost->arcs_size > 0) {
+ todelete = MIN(arc_mru_ghost->arcs_size, mru_over);
+ arc_evict_ghost(arc_mru_ghost, NULL, todelete);
+ }
+ }
+
+ if ((arc_over = arc_size - arc_c) > 0) {
+ int64_t tbl_over;
+
+ if (arc_mfu->arcs_lsize[ARC_BUFC_DATA] > 0) {
+ int64_t toevict =
+ MIN(arc_mfu->arcs_lsize[ARC_BUFC_DATA], arc_over);
+ (void) arc_evict(arc_mfu, NULL, toevict, FALSE,
+ ARC_BUFC_DATA);
+ arc_over = arc_size - arc_c;
+ }
+
+ if (arc_over > 0 &&
+ arc_mfu->arcs_lsize[ARC_BUFC_METADATA] > 0) {
+ int64_t toevict =
+ MIN(arc_mfu->arcs_lsize[ARC_BUFC_METADATA],
+ arc_over);
+ (void) arc_evict(arc_mfu, NULL, toevict, FALSE,
+ ARC_BUFC_METADATA);
+ }
+
+ tbl_over = arc_size + arc_mru_ghost->arcs_size +
+ arc_mfu_ghost->arcs_size - arc_c * 2;
+
+ if (tbl_over > 0 && arc_mfu_ghost->arcs_size > 0) {
+ todelete = MIN(arc_mfu_ghost->arcs_size, tbl_over);
+ arc_evict_ghost(arc_mfu_ghost, NULL, todelete);
+ }
+ }
+}
+
+static void
+arc_do_user_evicts(void)
+{
+ mutex_enter(&arc_eviction_mtx);
+ while (arc_eviction_list != NULL) {
+ arc_buf_t *buf = arc_eviction_list;
+ arc_eviction_list = buf->b_next;
+ rw_enter(&buf->b_lock, RW_WRITER);
+ buf->b_hdr = NULL;
+ rw_exit(&buf->b_lock);
+ mutex_exit(&arc_eviction_mtx);
+
+ if (buf->b_efunc != NULL)
+ VERIFY(buf->b_efunc(buf) == 0);
+
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+ kmem_cache_free(buf_cache, buf);
+ mutex_enter(&arc_eviction_mtx);
+ }
+ mutex_exit(&arc_eviction_mtx);
+}
+
+/*
+ * Flush all *evictable* data from the cache for the given spa.
+ * NOTE: this will not touch "active" (i.e. referenced) data.
+ */
+void
+arc_flush(spa_t *spa)
+{
+ while (list_head(&arc_mru->arcs_list[ARC_BUFC_DATA])) {
+ (void) arc_evict(arc_mru, spa, -1, FALSE, ARC_BUFC_DATA);
+ if (spa)
+ break;
+ }
+ while (list_head(&arc_mru->arcs_list[ARC_BUFC_METADATA])) {
+ (void) arc_evict(arc_mru, spa, -1, FALSE, ARC_BUFC_METADATA);
+ if (spa)
+ break;
+ }
+ while (list_head(&arc_mfu->arcs_list[ARC_BUFC_DATA])) {
+ (void) arc_evict(arc_mfu, spa, -1, FALSE, ARC_BUFC_DATA);
+ if (spa)
+ break;
+ }
+ while (list_head(&arc_mfu->arcs_list[ARC_BUFC_METADATA])) {
+ (void) arc_evict(arc_mfu, spa, -1, FALSE, ARC_BUFC_METADATA);
+ if (spa)
+ break;
+ }
+
+ arc_evict_ghost(arc_mru_ghost, spa, -1);
+ arc_evict_ghost(arc_mfu_ghost, spa, -1);
+
+ mutex_enter(&arc_reclaim_thr_lock);
+ arc_do_user_evicts();
+ mutex_exit(&arc_reclaim_thr_lock);
+ ASSERT(spa || arc_eviction_list == NULL);
+}
+
+int arc_shrink_shift = 5; /* log2(fraction of arc to reclaim) */
+
+void
+arc_shrink(void)
+{
+ if (arc_c > arc_c_min) {
+ uint64_t to_free;
+
+#ifdef _KERNEL
+ to_free = MAX(arc_c >> arc_shrink_shift, ptob(needfree));
+#else
+ to_free = arc_c >> arc_shrink_shift;
+#endif
+ if (arc_c > arc_c_min + to_free)
+ atomic_add_64(&arc_c, -to_free);
+ else
+ arc_c = arc_c_min;
+
+ atomic_add_64(&arc_p, -(arc_p >> arc_shrink_shift));
+ if (arc_c > arc_size)
+ arc_c = MAX(arc_size, arc_c_min);
+ if (arc_p > arc_c)
+ arc_p = (arc_c >> 1);
+ ASSERT(arc_c >= arc_c_min);
+ ASSERT((int64_t)arc_p >= 0);
+ }
+
+ if (arc_size > arc_c)
+ arc_adjust();
+}
+
+static int
+arc_reclaim_needed(void)
+{
+ uint64_t extra;
+
+#ifdef _KERNEL
+
+ if (needfree)
+ return (1);
+
+ /*
+ * take 'desfree' extra pages, so we reclaim sooner, rather than later
+ */
+ extra = desfree;
+
+ /*
+ * check that we're out of range of the pageout scanner. It starts to
+ * schedule paging if freemem is less than lotsfree and needfree.
+ * lotsfree is the high-water mark for pageout, and needfree is the
+ * number of needed free pages. We add extra pages here to make sure
+ * the scanner doesn't start up while we're freeing memory.
+ */
+ if (freemem < lotsfree + needfree + extra)
+ return (1);
+
+ /*
+ * check to make sure that swapfs has enough space so that anon
+ * reservations can still succeed. anon_resvmem() checks that the
+ * availrmem is greater than swapfs_minfree, and the number of reserved
+ * swap pages. We also add a bit of extra here just to prevent
+ * circumstances from getting really dire.
+ */
+ if (availrmem < swapfs_minfree + swapfs_reserve + extra)
+ return (1);
+
+#if defined(__i386)
+ /*
+ * If we're on an i386 platform, it's possible that we'll exhaust the
+ * kernel heap space before we ever run out of available physical
+ * memory. Most checks of the size of the heap_area compare against
+ * tune.t_minarmem, which is the minimum available real memory that we
+ * can have in the system. However, this is generally fixed at 25 pages
+ * which is so low that it's useless. In this comparison, we seek to
+ * calculate the total heap-size, and reclaim if more than 3/4ths of the
+ * heap is allocated. (Or, in the calculation, if less than 1/4th is
+ * free)
+ */
+ if (btop(vmem_size(heap_arena, VMEM_FREE)) <
+ (btop(vmem_size(heap_arena, VMEM_FREE | VMEM_ALLOC)) >> 2))
+ return (1);
+#endif
+
+#else
+ if (spa_get_random(100) == 0)
+ return (1);
+#endif
+ return (0);
+}
+
+static void
+arc_kmem_reap_now(arc_reclaim_strategy_t strat)
+{
+ size_t i;
+ kmem_cache_t *prev_cache = NULL;
+ kmem_cache_t *prev_data_cache = NULL;
+ extern kmem_cache_t *zio_buf_cache[];
+ extern kmem_cache_t *zio_data_buf_cache[];
+
+#ifdef _KERNEL
+ if (arc_meta_used >= arc_meta_limit) {
+ /*
+ * We are exceeding our meta-data cache limit.
+ * Purge some DNLC entries to release holds on meta-data.
+ */
+ dnlc_reduce_cache((void *)(uintptr_t)arc_reduce_dnlc_percent);
+ }
+#if defined(__i386)
+ /*
+ * Reclaim unused memory from all kmem caches.
+ */
+ kmem_reap();
+#endif
+#endif
+
+ /*
+ * An aggressive reclamation will shrink the cache size as well as
+ * reap free buffers from the arc kmem caches.
+ */
+ if (strat == ARC_RECLAIM_AGGR)
+ arc_shrink();
+
+ for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) {
+ if (zio_buf_cache[i] != prev_cache) {
+ prev_cache = zio_buf_cache[i];
+ kmem_cache_reap_now(zio_buf_cache[i]);
+ }
+ if (zio_data_buf_cache[i] != prev_data_cache) {
+ prev_data_cache = zio_data_buf_cache[i];
+ kmem_cache_reap_now(zio_data_buf_cache[i]);
+ }
+ }
+ kmem_cache_reap_now(buf_cache);
+ kmem_cache_reap_now(hdr_cache);
+}
+
+static void
+arc_reclaim_thread(void)
+{
+ clock_t growtime = 0;
+ arc_reclaim_strategy_t last_reclaim = ARC_RECLAIM_CONS;
+ callb_cpr_t cpr;
+
+ CALLB_CPR_INIT(&cpr, &arc_reclaim_thr_lock, callb_generic_cpr, FTAG);
+
+ mutex_enter(&arc_reclaim_thr_lock);
+ while (arc_thread_exit == 0) {
+ if (arc_reclaim_needed()) {
+
+ if (arc_no_grow) {
+ if (last_reclaim == ARC_RECLAIM_CONS) {
+ last_reclaim = ARC_RECLAIM_AGGR;
+ } else {
+ last_reclaim = ARC_RECLAIM_CONS;
+ }
+ } else {
+ arc_no_grow = TRUE;
+ last_reclaim = ARC_RECLAIM_AGGR;
+ membar_producer();
+ }
+
+ /* reset the growth delay for every reclaim */
+ growtime = lbolt + (arc_grow_retry * hz);
+
+ arc_kmem_reap_now(last_reclaim);
+ arc_warm = B_TRUE;
+
+ } else if (arc_no_grow && lbolt >= growtime) {
+ arc_no_grow = FALSE;
+ }
+
+ if (2 * arc_c < arc_size +
+ arc_mru_ghost->arcs_size + arc_mfu_ghost->arcs_size)
+ arc_adjust();
+
+ if (arc_eviction_list != NULL)
+ arc_do_user_evicts();
+
+ /* block until needed, or one second, whichever is shorter */
+ CALLB_CPR_SAFE_BEGIN(&cpr);
+ (void) cv_timedwait(&arc_reclaim_thr_cv,
+ &arc_reclaim_thr_lock, (lbolt + hz));
+ CALLB_CPR_SAFE_END(&cpr, &arc_reclaim_thr_lock);
+ }
+
+ arc_thread_exit = 0;
+ cv_broadcast(&arc_reclaim_thr_cv);
+ CALLB_CPR_EXIT(&cpr); /* drops arc_reclaim_thr_lock */
+ thread_exit();
+}
+
+/*
+ * Adapt arc info given the number of bytes we are trying to add and
+ * the state that we are comming from. This function is only called
+ * when we are adding new content to the cache.
+ */
+static void
+arc_adapt(int bytes, arc_state_t *state)
+{
+ int mult;
+
+ if (state == arc_l2c_only)
+ return;
+
+ ASSERT(bytes > 0);
+ /*
+ * Adapt the target size of the MRU list:
+ * - if we just hit in the MRU ghost list, then increase
+ * the target size of the MRU list.
+ * - if we just hit in the MFU ghost list, then increase
+ * the target size of the MFU list by decreasing the
+ * target size of the MRU list.
+ */
+ if (state == arc_mru_ghost) {
+ mult = ((arc_mru_ghost->arcs_size >= arc_mfu_ghost->arcs_size) ?
+ 1 : (arc_mfu_ghost->arcs_size/arc_mru_ghost->arcs_size));
+
+ arc_p = MIN(arc_c, arc_p + bytes * mult);
+ } else if (state == arc_mfu_ghost) {
+ mult = ((arc_mfu_ghost->arcs_size >= arc_mru_ghost->arcs_size) ?
+ 1 : (arc_mru_ghost->arcs_size/arc_mfu_ghost->arcs_size));
+
+ arc_p = MAX(0, (int64_t)arc_p - bytes * mult);
+ }
+ ASSERT((int64_t)arc_p >= 0);
+
+ if (arc_reclaim_needed()) {
+ cv_signal(&arc_reclaim_thr_cv);
+ return;
+ }
+
+ if (arc_no_grow)
+ return;
+
+ if (arc_c >= arc_c_max)
+ return;
+
+ /*
+ * If we're within (2 * maxblocksize) bytes of the target
+ * cache size, increment the target cache size
+ */
+ if (arc_size > arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) {
+ atomic_add_64(&arc_c, (int64_t)bytes);
+ if (arc_c > arc_c_max)
+ arc_c = arc_c_max;
+ else if (state == arc_anon)
+ atomic_add_64(&arc_p, (int64_t)bytes);
+ if (arc_p > arc_c)
+ arc_p = arc_c;
+ }
+ ASSERT((int64_t)arc_p >= 0);
+}
+
+/*
+ * Check if the cache has reached its limits and eviction is required
+ * prior to insert.
+ */
+static int
+arc_evict_needed(arc_buf_contents_t type)
+{
+ if (type == ARC_BUFC_METADATA && arc_meta_used >= arc_meta_limit)
+ return (1);
+
+#ifdef _KERNEL
+ /*
+ * If zio data pages are being allocated out of a separate heap segment,
+ * then enforce that the size of available vmem for this area remains
+ * above about 1/32nd free.
+ */
+ if (type == ARC_BUFC_DATA && zio_arena != NULL &&
+ vmem_size(zio_arena, VMEM_FREE) <
+ (vmem_size(zio_arena, VMEM_ALLOC) >> 5))
+ return (1);
+#endif
+
+ if (arc_reclaim_needed())
+ return (1);
+
+ return (arc_size > arc_c);
+}
+
+/*
+ * The buffer, supplied as the first argument, needs a data block.
+ * So, if we are at cache max, determine which cache should be victimized.
+ * We have the following cases:
+ *
+ * 1. Insert for MRU, p > sizeof(arc_anon + arc_mru) ->
+ * In this situation if we're out of space, but the resident size of the MFU is
+ * under the limit, victimize the MFU cache to satisfy this insertion request.
+ *
+ * 2. Insert for MRU, p <= sizeof(arc_anon + arc_mru) ->
+ * Here, we've used up all of the available space for the MRU, so we need to
+ * evict from our own cache instead. Evict from the set of resident MRU
+ * entries.
+ *
+ * 3. Insert for MFU (c - p) > sizeof(arc_mfu) ->
+ * c minus p represents the MFU space in the cache, since p is the size of the
+ * cache that is dedicated to the MRU. In this situation there's still space on
+ * the MFU side, so the MRU side needs to be victimized.
+ *
+ * 4. Insert for MFU (c - p) < sizeof(arc_mfu) ->
+ * MFU's resident set is consuming more space than it has been allotted. In
+ * this situation, we must victimize our own cache, the MFU, for this insertion.
+ */
+static void
+arc_get_data_buf(arc_buf_t *buf)
+{
+ arc_state_t *state = buf->b_hdr->b_state;
+ uint64_t size = buf->b_hdr->b_size;
+ arc_buf_contents_t type = buf->b_hdr->b_type;
+
+ arc_adapt(size, state);
+
+ /*
+ * We have not yet reached cache maximum size,
+ * just allocate a new buffer.
+ */
+ if (!arc_evict_needed(type)) {
+ if (type == ARC_BUFC_METADATA) {
+ buf->b_data = zio_buf_alloc(size);
+ arc_space_consume(size);
+ } else {
+ ASSERT(type == ARC_BUFC_DATA);
+ buf->b_data = zio_data_buf_alloc(size);
+ atomic_add_64(&arc_size, size);
+ }
+ goto out;
+ }
+
+ /*
+ * If we are prefetching from the mfu ghost list, this buffer
+ * will end up on the mru list; so steal space from there.
+ */
+ if (state == arc_mfu_ghost)
+ state = buf->b_hdr->b_flags & ARC_PREFETCH ? arc_mru : arc_mfu;
+ else if (state == arc_mru_ghost)
+ state = arc_mru;
+
+ if (state == arc_mru || state == arc_anon) {
+ uint64_t mru_used = arc_anon->arcs_size + arc_mru->arcs_size;
+ state = (arc_mfu->arcs_lsize[type] > 0 &&
+ arc_p > mru_used) ? arc_mfu : arc_mru;
+ } else {
+ /* MFU cases */
+ uint64_t mfu_space = arc_c - arc_p;
+ state = (arc_mru->arcs_lsize[type] > 0 &&
+ mfu_space > arc_mfu->arcs_size) ? arc_mru : arc_mfu;
+ }
+ if ((buf->b_data = arc_evict(state, NULL, size, TRUE, type)) == NULL) {
+ if (type == ARC_BUFC_METADATA) {
+ buf->b_data = zio_buf_alloc(size);
+ arc_space_consume(size);
+ } else {
+ ASSERT(type == ARC_BUFC_DATA);
+ buf->b_data = zio_data_buf_alloc(size);
+ atomic_add_64(&arc_size, size);
+ }
+ ARCSTAT_BUMP(arcstat_recycle_miss);
+ }
+ ASSERT(buf->b_data != NULL);
+out:
+ /*
+ * Update the state size. Note that ghost states have a
+ * "ghost size" and so don't need to be updated.
+ */
+ if (!GHOST_STATE(buf->b_hdr->b_state)) {
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+
+ atomic_add_64(&hdr->b_state->arcs_size, size);
+ if (list_link_active(&hdr->b_arc_node)) {
+ ASSERT(refcount_is_zero(&hdr->b_refcnt));
+ atomic_add_64(&hdr->b_state->arcs_lsize[type], size);
+ }
+ /*
+ * If we are growing the cache, and we are adding anonymous
+ * data, and we have outgrown arc_p, update arc_p
+ */
+ if (arc_size < arc_c && hdr->b_state == arc_anon &&
+ arc_anon->arcs_size + arc_mru->arcs_size > arc_p)
+ arc_p = MIN(arc_c, arc_p + size);
+ }
+}
+
+/*
+ * This routine is called whenever a buffer is accessed.
+ * NOTE: the hash lock is dropped in this function.
+ */
+static void
+arc_access(arc_buf_hdr_t *buf, kmutex_t *hash_lock)
+{
+ ASSERT(MUTEX_HELD(hash_lock));
+
+ if (buf->b_state == arc_anon) {
+ /*
+ * This buffer is not in the cache, and does not
+ * appear in our "ghost" list. Add the new buffer
+ * to the MRU state.
+ */
+
+ ASSERT(buf->b_arc_access == 0);
+ buf->b_arc_access = lbolt;
+ DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, buf);
+ arc_change_state(arc_mru, buf, hash_lock);
+
+ } else if (buf->b_state == arc_mru) {
+ /*
+ * If this buffer is here because of a prefetch, then either:
+ * - clear the flag if this is a "referencing" read
+ * (any subsequent access will bump this into the MFU state).
+ * or
+ * - move the buffer to the head of the list if this is
+ * another prefetch (to make it less likely to be evicted).
+ */
+ if ((buf->b_flags & ARC_PREFETCH) != 0) {
+ if (refcount_count(&buf->b_refcnt) == 0) {
+ ASSERT(list_link_active(&buf->b_arc_node));
+ } else {
+ buf->b_flags &= ~ARC_PREFETCH;
+ ARCSTAT_BUMP(arcstat_mru_hits);
+ }
+ buf->b_arc_access = lbolt;
+ return;
+ }
+
+ /*
+ * This buffer has been "accessed" only once so far,
+ * but it is still in the cache. Move it to the MFU
+ * state.
+ */
+ if (lbolt > buf->b_arc_access + ARC_MINTIME) {
+ /*
+ * More than 125ms have passed since we
+ * instantiated this buffer. Move it to the
+ * most frequently used state.
+ */
+ buf->b_arc_access = lbolt;
+ DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, buf);
+ arc_change_state(arc_mfu, buf, hash_lock);
+ }
+ ARCSTAT_BUMP(arcstat_mru_hits);
+ } else if (buf->b_state == arc_mru_ghost) {
+ arc_state_t *new_state;
+ /*
+ * This buffer has been "accessed" recently, but
+ * was evicted from the cache. Move it to the
+ * MFU state.
+ */
+
+ if (buf->b_flags & ARC_PREFETCH) {
+ new_state = arc_mru;
+ if (refcount_count(&buf->b_refcnt) > 0)
+ buf->b_flags &= ~ARC_PREFETCH;
+ DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, buf);
+ } else {
+ new_state = arc_mfu;
+ DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, buf);
+ }
+
+ buf->b_arc_access = lbolt;
+ arc_change_state(new_state, buf, hash_lock);
+
+ ARCSTAT_BUMP(arcstat_mru_ghost_hits);
+ } else if (buf->b_state == arc_mfu) {
+ /*
+ * This buffer has been accessed more than once and is
+ * still in the cache. Keep it in the MFU state.
+ *
+ * NOTE: an add_reference() that occurred when we did
+ * the arc_read() will have kicked this off the list.
+ * If it was a prefetch, we will explicitly move it to
+ * the head of the list now.
+ */
+ if ((buf->b_flags & ARC_PREFETCH) != 0) {
+ ASSERT(refcount_count(&buf->b_refcnt) == 0);
+ ASSERT(list_link_active(&buf->b_arc_node));
+ }
+ ARCSTAT_BUMP(arcstat_mfu_hits);
+ buf->b_arc_access = lbolt;
+ } else if (buf->b_state == arc_mfu_ghost) {
+ arc_state_t *new_state = arc_mfu;
+ /*
+ * This buffer has been accessed more than once but has
+ * been evicted from the cache. Move it back to the
+ * MFU state.
+ */
+
+ if (buf->b_flags & ARC_PREFETCH) {
+ /*
+ * This is a prefetch access...
+ * move this block back to the MRU state.
+ */
+ ASSERT3U(refcount_count(&buf->b_refcnt), ==, 0);
+ new_state = arc_mru;
+ }
+
+ buf->b_arc_access = lbolt;
+ DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, buf);
+ arc_change_state(new_state, buf, hash_lock);
+
+ ARCSTAT_BUMP(arcstat_mfu_ghost_hits);
+ } else if (buf->b_state == arc_l2c_only) {
+ /*
+ * This buffer is on the 2nd Level ARC.
+ */
+
+ buf->b_arc_access = lbolt;
+ DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, buf);
+ arc_change_state(arc_mfu, buf, hash_lock);
+ } else {
+ ASSERT(!"invalid arc state");
+ }
+}
+
+/* a generic arc_done_func_t which you can use */
+/* ARGSUSED */
+void
+arc_bcopy_func(zio_t *zio, arc_buf_t *buf, void *arg)
+{
+ bcopy(buf->b_data, arg, buf->b_hdr->b_size);
+ VERIFY(arc_buf_remove_ref(buf, arg) == 1);
+}
+
+/* a generic arc_done_func_t */
+void
+arc_getbuf_func(zio_t *zio, arc_buf_t *buf, void *arg)
+{
+ arc_buf_t **bufp = arg;
+ if (zio && zio->io_error) {
+ VERIFY(arc_buf_remove_ref(buf, arg) == 1);
+ *bufp = NULL;
+ } else {
+ *bufp = buf;
+ }
+}
+
+static void
+arc_read_done(zio_t *zio)
+{
+ arc_buf_hdr_t *hdr, *found;
+ arc_buf_t *buf;
+ arc_buf_t *abuf; /* buffer we're assigning to callback */
+ kmutex_t *hash_lock;
+ arc_callback_t *callback_list, *acb;
+ int freeable = FALSE;
+
+ buf = zio->io_private;
+ hdr = buf->b_hdr;
+
+ /*
+ * The hdr was inserted into hash-table and removed from lists
+ * prior to starting I/O. We should find this header, since
+ * it's in the hash table, and it should be legit since it's
+ * not possible to evict it during the I/O. The only possible
+ * reason for it not to be found is if we were freed during the
+ * read.
+ */
+ found = buf_hash_find(zio->io_spa, &hdr->b_dva, hdr->b_birth,
+ &hash_lock);
+
+ ASSERT((found == NULL && HDR_FREED_IN_READ(hdr) && hash_lock == NULL) ||
+ (found == hdr && DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) ||
+ (found == hdr && HDR_L2_READING(hdr)));
+
+ hdr->b_flags &= ~ARC_L2_EVICTED;
+ if (l2arc_noprefetch && (hdr->b_flags & ARC_PREFETCH))
+ hdr->b_flags &= ~ARC_L2CACHE;
+
+ /* byteswap if necessary */
+ callback_list = hdr->b_acb;
+ ASSERT(callback_list != NULL);
+ if (BP_SHOULD_BYTESWAP(zio->io_bp)) {
+ arc_byteswap_func_t *func = BP_GET_LEVEL(zio->io_bp) > 0 ?
+ byteswap_uint64_array :
+ dmu_ot[BP_GET_TYPE(zio->io_bp)].ot_byteswap;
+ func(buf->b_data, hdr->b_size);
+ }
+
+ arc_cksum_compute(buf, B_FALSE);
+
+ /* create copies of the data buffer for the callers */
+ abuf = buf;
+ for (acb = callback_list; acb; acb = acb->acb_next) {
+ if (acb->acb_done) {
+ if (abuf == NULL)
+ abuf = arc_buf_clone(buf);
+ acb->acb_buf = abuf;
+ abuf = NULL;
+ }
+ }
+ hdr->b_acb = NULL;
+ hdr->b_flags &= ~ARC_IO_IN_PROGRESS;
+ ASSERT(!HDR_BUF_AVAILABLE(hdr));
+ if (abuf == buf)
+ hdr->b_flags |= ARC_BUF_AVAILABLE;
+
+ ASSERT(refcount_is_zero(&hdr->b_refcnt) || callback_list != NULL);
+
+ if (zio->io_error != 0) {
+ hdr->b_flags |= ARC_IO_ERROR;
+ if (hdr->b_state != arc_anon)
+ arc_change_state(arc_anon, hdr, hash_lock);
+ if (HDR_IN_HASH_TABLE(hdr))
+ buf_hash_remove(hdr);
+ freeable = refcount_is_zero(&hdr->b_refcnt);
+ }
+
+ /*
+ * Broadcast before we drop the hash_lock to avoid the possibility
+ * that the hdr (and hence the cv) might be freed before we get to
+ * the cv_broadcast().
+ */
+ cv_broadcast(&hdr->b_cv);
+
+ if (hash_lock) {
+ /*
+ * Only call arc_access on anonymous buffers. This is because
+ * if we've issued an I/O for an evicted buffer, we've already
+ * called arc_access (to prevent any simultaneous readers from
+ * getting confused).
+ */
+ if (zio->io_error == 0 && hdr->b_state == arc_anon)
+ arc_access(hdr, hash_lock);
+ mutex_exit(hash_lock);
+ } else {
+ /*
+ * This block was freed while we waited for the read to
+ * complete. It has been removed from the hash table and
+ * moved to the anonymous state (so that it won't show up
+ * in the cache).
+ */
+ ASSERT3P(hdr->b_state, ==, arc_anon);
+ freeable = refcount_is_zero(&hdr->b_refcnt);
+ }
+
+ /* execute each callback and free its structure */
+ while ((acb = callback_list) != NULL) {
+ if (acb->acb_done)
+ acb->acb_done(zio, acb->acb_buf, acb->acb_private);
+
+ if (acb->acb_zio_dummy != NULL) {
+ acb->acb_zio_dummy->io_error = zio->io_error;
+ zio_nowait(acb->acb_zio_dummy);
+ }
+
+ callback_list = acb->acb_next;
+ kmem_free(acb, sizeof (arc_callback_t));
+ }
+
+ if (freeable)
+ arc_hdr_destroy(hdr);
+}
+
+/*
+ * "Read" the block block at the specified DVA (in bp) via the
+ * cache. If the block is found in the cache, invoke the provided
+ * callback immediately and return. Note that the `zio' parameter
+ * in the callback will be NULL in this case, since no IO was
+ * required. If the block is not in the cache pass the read request
+ * on to the spa with a substitute callback function, so that the
+ * requested block will be added to the cache.
+ *
+ * If a read request arrives for a block that has a read in-progress,
+ * either wait for the in-progress read to complete (and return the
+ * results); or, if this is a read with a "done" func, add a record
+ * to the read to invoke the "done" func when the read completes,
+ * and return; or just return.
+ *
+ * arc_read_done() will invoke all the requested "done" functions
+ * for readers of this block.
+ *
+ * Normal callers should use arc_read and pass the arc buffer and offset
+ * for the bp. But if you know you don't need locking, you can use
+ * arc_read_bp.
+ */
+int
+arc_read(zio_t *pio, spa_t *spa, blkptr_t *bp, arc_buf_t *pbuf,
+ arc_done_func_t *done, void *private, int priority, int zio_flags,
+ uint32_t *arc_flags, const zbookmark_t *zb)
+{
+ int err;
+ arc_buf_hdr_t *hdr = pbuf->b_hdr;
+
+ ASSERT(!refcount_is_zero(&pbuf->b_hdr->b_refcnt));
+ ASSERT3U((char *)bp - (char *)pbuf->b_data, <, pbuf->b_hdr->b_size);
+ rw_enter(&pbuf->b_lock, RW_READER);
+
+ err = arc_read_nolock(pio, spa, bp, done, private, priority,
+ zio_flags, arc_flags, zb);
+
+ ASSERT3P(hdr, ==, pbuf->b_hdr);
+ rw_exit(&pbuf->b_lock);
+ return (err);
+}
+
+int
+arc_read_nolock(zio_t *pio, spa_t *spa, blkptr_t *bp,
+ arc_done_func_t *done, void *private, int priority, int zio_flags,
+ uint32_t *arc_flags, const zbookmark_t *zb)
+{
+ arc_buf_hdr_t *hdr;
+ arc_buf_t *buf;
+ kmutex_t *hash_lock;
+ zio_t *rzio;
+
+top:
+ hdr = buf_hash_find(spa, BP_IDENTITY(bp), bp->blk_birth, &hash_lock);
+ if (hdr && hdr->b_datacnt > 0) {
+
+ *arc_flags |= ARC_CACHED;
+
+ if (HDR_IO_IN_PROGRESS(hdr)) {
+
+ if (*arc_flags & ARC_WAIT) {
+ cv_wait(&hdr->b_cv, hash_lock);
+ mutex_exit(hash_lock);
+ goto top;
+ }
+ ASSERT(*arc_flags & ARC_NOWAIT);
+
+ if (done) {
+ arc_callback_t *acb = NULL;
+
+ acb = kmem_zalloc(sizeof (arc_callback_t),
+ KM_SLEEP);
+ acb->acb_done = done;
+ acb->acb_private = private;
+ if (pio != NULL)
+ acb->acb_zio_dummy = zio_null(pio,
+ spa, NULL, NULL, zio_flags);
+
+ ASSERT(acb->acb_done != NULL);
+ acb->acb_next = hdr->b_acb;
+ hdr->b_acb = acb;
+ add_reference(hdr, hash_lock, private);
+ mutex_exit(hash_lock);
+ return (0);
+ }
+ mutex_exit(hash_lock);
+ return (0);
+ }
+
+ ASSERT(hdr->b_state == arc_mru || hdr->b_state == arc_mfu);
+
+ if (done) {
+ add_reference(hdr, hash_lock, private);
+ /*
+ * If this block is already in use, create a new
+ * copy of the data so that we will be guaranteed
+ * that arc_release() will always succeed.
+ */
+ buf = hdr->b_buf;
+ ASSERT(buf);
+ ASSERT(buf->b_data);
+ if (HDR_BUF_AVAILABLE(hdr)) {
+ ASSERT(buf->b_efunc == NULL);
+ hdr->b_flags &= ~ARC_BUF_AVAILABLE;
+ } else {
+ buf = arc_buf_clone(buf);
+ }
+ } else if (*arc_flags & ARC_PREFETCH &&
+ refcount_count(&hdr->b_refcnt) == 0) {
+ hdr->b_flags |= ARC_PREFETCH;
+ }
+ DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
+ arc_access(hdr, hash_lock);
+ if (*arc_flags & ARC_L2CACHE)
+ hdr->b_flags |= ARC_L2CACHE;
+ mutex_exit(hash_lock);
+ ARCSTAT_BUMP(arcstat_hits);
+ ARCSTAT_CONDSTAT(!(hdr->b_flags & ARC_PREFETCH),
+ demand, prefetch, hdr->b_type != ARC_BUFC_METADATA,
+ data, metadata, hits);
+
+ if (done)
+ done(NULL, buf, private);
+ } else {
+ uint64_t size = BP_GET_LSIZE(bp);
+ arc_callback_t *acb;
+ vdev_t *vd = NULL;
+ daddr_t addr;
+
+ if (hdr == NULL) {
+ /* this block is not in the cache */
+ arc_buf_hdr_t *exists;
+ arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);
+ buf = arc_buf_alloc(spa, size, private, type);
+ hdr = buf->b_hdr;
+ hdr->b_dva = *BP_IDENTITY(bp);
+ hdr->b_birth = bp->blk_birth;
+ hdr->b_cksum0 = bp->blk_cksum.zc_word[0];
+ exists = buf_hash_insert(hdr, &hash_lock);
+ if (exists) {
+ /* somebody beat us to the hash insert */
+ mutex_exit(hash_lock);
+ bzero(&hdr->b_dva, sizeof (dva_t));
+ hdr->b_birth = 0;
+ hdr->b_cksum0 = 0;
+ (void) arc_buf_remove_ref(buf, private);
+ goto top; /* restart the IO request */
+ }
+ /* if this is a prefetch, we don't have a reference */
+ if (*arc_flags & ARC_PREFETCH) {
+ (void) remove_reference(hdr, hash_lock,
+ private);
+ hdr->b_flags |= ARC_PREFETCH;
+ }
+ if (*arc_flags & ARC_L2CACHE)
+ hdr->b_flags |= ARC_L2CACHE;
+ if (BP_GET_LEVEL(bp) > 0)
+ hdr->b_flags |= ARC_INDIRECT;
+ } else {
+ /* this block is in the ghost cache */
+ ASSERT(GHOST_STATE(hdr->b_state));
+ ASSERT(!HDR_IO_IN_PROGRESS(hdr));
+ ASSERT3U(refcount_count(&hdr->b_refcnt), ==, 0);
+ ASSERT(hdr->b_buf == NULL);
+
+ /* if this is a prefetch, we don't have a reference */
+ if (*arc_flags & ARC_PREFETCH)
+ hdr->b_flags |= ARC_PREFETCH;
+ else
+ add_reference(hdr, hash_lock, private);
+ if (*arc_flags & ARC_L2CACHE)
+ hdr->b_flags |= ARC_L2CACHE;
+ buf = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
+ buf->b_hdr = hdr;
+ buf->b_data = NULL;
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+ buf->b_next = NULL;
+ hdr->b_buf = buf;
+ arc_get_data_buf(buf);
+ ASSERT(hdr->b_datacnt == 0);
+ hdr->b_datacnt = 1;
+
+ }
+
+ acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP);
+ acb->acb_done = done;
+ acb->acb_private = private;
+
+ ASSERT(hdr->b_acb == NULL);
+ hdr->b_acb = acb;
+ hdr->b_flags |= ARC_IO_IN_PROGRESS;
+
+ /*
+ * If the buffer has been evicted, migrate it to a present state
+ * before issuing the I/O. Once we drop the hash-table lock,
+ * the header will be marked as I/O in progress and have an
+ * attached buffer. At this point, anybody who finds this
+ * buffer ought to notice that it's legit but has a pending I/O.
+ */
+
+ if (GHOST_STATE(hdr->b_state))
+ arc_access(hdr, hash_lock);
+
+ if (HDR_L2CACHE(hdr) && hdr->b_l2hdr != NULL &&
+ (vd = hdr->b_l2hdr->b_dev->l2ad_vdev) != NULL) {
+ addr = hdr->b_l2hdr->b_daddr;
+ /*
+ * Lock out device removal.
+ */
+ if (vdev_is_dead(vd) ||
+ !spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER))
+ vd = NULL;
+ }
+
+ mutex_exit(hash_lock);
+
+ ASSERT3U(hdr->b_size, ==, size);
+ DTRACE_PROBE3(arc__miss, blkptr_t *, bp, uint64_t, size,
+ zbookmark_t *, zb);
+ ARCSTAT_BUMP(arcstat_misses);
+ ARCSTAT_CONDSTAT(!(hdr->b_flags & ARC_PREFETCH),
+ demand, prefetch, hdr->b_type != ARC_BUFC_METADATA,
+ data, metadata, misses);
+
+ if (vd != NULL) {
+ /*
+ * Read from the L2ARC if the following are true:
+ * 1. The L2ARC vdev was previously cached.
+ * 2. This buffer still has L2ARC metadata.
+ * 3. This buffer isn't currently writing to the L2ARC.
+ * 4. The L2ARC entry wasn't evicted, which may
+ * also have invalidated the vdev.
+ */
+ if (hdr->b_l2hdr != NULL &&
+ !HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr)) {
+ l2arc_read_callback_t *cb;
+
+ DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr);
+ ARCSTAT_BUMP(arcstat_l2_hits);
+
+ cb = kmem_zalloc(sizeof (l2arc_read_callback_t),
+ KM_SLEEP);
+ cb->l2rcb_buf = buf;
+ cb->l2rcb_spa = spa;
+ cb->l2rcb_bp = *bp;
+ cb->l2rcb_zb = *zb;
+ cb->l2rcb_flags = zio_flags;
+
+ /*
+ * l2arc read. The SCL_L2ARC lock will be
+ * released by l2arc_read_done().
+ */
+ rzio = zio_read_phys(pio, vd, addr, size,
+ buf->b_data, ZIO_CHECKSUM_OFF,
+ l2arc_read_done, cb, priority, zio_flags |
+ ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL |
+ ZIO_FLAG_DONT_PROPAGATE |
+ ZIO_FLAG_DONT_RETRY, B_FALSE);
+ DTRACE_PROBE2(l2arc__read, vdev_t *, vd,
+ zio_t *, rzio);
+
+ if (*arc_flags & ARC_NOWAIT) {
+ zio_nowait(rzio);
+ return (0);
+ }
+
+ ASSERT(*arc_flags & ARC_WAIT);
+ if (zio_wait(rzio) == 0)
+ return (0);
+
+ /* l2arc read error; goto zio_read() */
+ } else {
+ DTRACE_PROBE1(l2arc__miss,
+ arc_buf_hdr_t *, hdr);
+ ARCSTAT_BUMP(arcstat_l2_misses);
+ if (HDR_L2_WRITING(hdr))
+ ARCSTAT_BUMP(arcstat_l2_rw_clash);
+ spa_config_exit(spa, SCL_L2ARC, vd);
+ }
+ }
+
+ rzio = zio_read(pio, spa, bp, buf->b_data, size,
+ arc_read_done, buf, priority, zio_flags, zb);
+
+ if (*arc_flags & ARC_WAIT)
+ return (zio_wait(rzio));
+
+ ASSERT(*arc_flags & ARC_NOWAIT);
+ zio_nowait(rzio);
+ }
+ return (0);
+}
+
+/*
+ * arc_read() variant to support pool traversal. If the block is already
+ * in the ARC, make a copy of it; otherwise, the caller will do the I/O.
+ * The idea is that we don't want pool traversal filling up memory, but
+ * if the ARC already has the data anyway, we shouldn't pay for the I/O.
+ */
+int
+arc_tryread(spa_t *spa, blkptr_t *bp, void *data)
+{
+ arc_buf_hdr_t *hdr;
+ kmutex_t *hash_mtx;
+ int rc = 0;
+
+ hdr = buf_hash_find(spa, BP_IDENTITY(bp), bp->blk_birth, &hash_mtx);
+
+ if (hdr && hdr->b_datacnt > 0 && !HDR_IO_IN_PROGRESS(hdr)) {
+ arc_buf_t *buf = hdr->b_buf;
+
+ ASSERT(buf);
+ while (buf->b_data == NULL) {
+ buf = buf->b_next;
+ ASSERT(buf);
+ }
+ bcopy(buf->b_data, data, hdr->b_size);
+ } else {
+ rc = ENOENT;
+ }
+
+ if (hash_mtx)
+ mutex_exit(hash_mtx);
+
+ return (rc);
+}
+
+void
+arc_set_callback(arc_buf_t *buf, arc_evict_func_t *func, void *private)
+{
+ ASSERT(buf->b_hdr != NULL);
+ ASSERT(buf->b_hdr->b_state != arc_anon);
+ ASSERT(!refcount_is_zero(&buf->b_hdr->b_refcnt) || func == NULL);
+ buf->b_efunc = func;
+ buf->b_private = private;
+}
+
+/*
+ * This is used by the DMU to let the ARC know that a buffer is
+ * being evicted, so the ARC should clean up. If this arc buf
+ * is not yet in the evicted state, it will be put there.
+ */
+int
+arc_buf_evict(arc_buf_t *buf)
+{
+ arc_buf_hdr_t *hdr;
+ kmutex_t *hash_lock;
+ arc_buf_t **bufp;
+
+ rw_enter(&buf->b_lock, RW_WRITER);
+ hdr = buf->b_hdr;
+ if (hdr == NULL) {
+ /*
+ * We are in arc_do_user_evicts().
+ */
+ ASSERT(buf->b_data == NULL);
+ rw_exit(&buf->b_lock);
+ return (0);
+ } else if (buf->b_data == NULL) {
+ arc_buf_t copy = *buf; /* structure assignment */
+ /*
+ * We are on the eviction list; process this buffer now
+ * but let arc_do_user_evicts() do the reaping.
+ */
+ buf->b_efunc = NULL;
+ rw_exit(&buf->b_lock);
+ VERIFY(copy.b_efunc(&copy) == 0);
+ return (1);
+ }
+ hash_lock = HDR_LOCK(hdr);
+ mutex_enter(hash_lock);
+
+ ASSERT(buf->b_hdr == hdr);
+ ASSERT3U(refcount_count(&hdr->b_refcnt), <, hdr->b_datacnt);
+ ASSERT(hdr->b_state == arc_mru || hdr->b_state == arc_mfu);
+
+ /*
+ * Pull this buffer off of the hdr
+ */
+ bufp = &hdr->b_buf;
+ while (*bufp != buf)
+ bufp = &(*bufp)->b_next;
+ *bufp = buf->b_next;
+
+ ASSERT(buf->b_data != NULL);
+ arc_buf_destroy(buf, FALSE, FALSE);
+
+ if (hdr->b_datacnt == 0) {
+ arc_state_t *old_state = hdr->b_state;
+ arc_state_t *evicted_state;
+
+ ASSERT(refcount_is_zero(&hdr->b_refcnt));
+
+ evicted_state =
+ (old_state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost;
+
+ mutex_enter(&old_state->arcs_mtx);
+ mutex_enter(&evicted_state->arcs_mtx);
+
+ arc_change_state(evicted_state, hdr, hash_lock);
+ ASSERT(HDR_IN_HASH_TABLE(hdr));
+ hdr->b_flags |= ARC_IN_HASH_TABLE;
+ hdr->b_flags &= ~ARC_BUF_AVAILABLE;
+
+ mutex_exit(&evicted_state->arcs_mtx);
+ mutex_exit(&old_state->arcs_mtx);
+ }
+ mutex_exit(hash_lock);
+ rw_exit(&buf->b_lock);
+
+ VERIFY(buf->b_efunc(buf) == 0);
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+ buf->b_hdr = NULL;
+ kmem_cache_free(buf_cache, buf);
+ return (1);
+}
+
+/*
+ * Release this buffer from the cache. This must be done
+ * after a read and prior to modifying the buffer contents.
+ * If the buffer has more than one reference, we must make
+ * a new hdr for the buffer.
+ */
+void
+arc_release(arc_buf_t *buf, void *tag)
+{
+ arc_buf_hdr_t *hdr;
+ kmutex_t *hash_lock;
+ l2arc_buf_hdr_t *l2hdr;
+ uint64_t buf_size;
+
+ rw_enter(&buf->b_lock, RW_WRITER);
+ hdr = buf->b_hdr;
+
+ /* this buffer is not on any list */
+ ASSERT(refcount_count(&hdr->b_refcnt) > 0);
+ ASSERT(!(hdr->b_flags & ARC_STORED));
+
+ if (hdr->b_state == arc_anon) {
+ /* this buffer is already released */
+ ASSERT3U(refcount_count(&hdr->b_refcnt), ==, 1);
+ ASSERT(BUF_EMPTY(hdr));
+ ASSERT(buf->b_efunc == NULL);
+ arc_buf_thaw(buf);
+ rw_exit(&buf->b_lock);
+ return;
+ }
+
+ hash_lock = HDR_LOCK(hdr);
+ mutex_enter(hash_lock);
+
+ l2hdr = hdr->b_l2hdr;
+ if (l2hdr) {
+ mutex_enter(&l2arc_buflist_mtx);
+ hdr->b_l2hdr = NULL;
+ buf_size = hdr->b_size;
+ }
+
+ /*
+ * Do we have more than one buf?
+ */
+ if (hdr->b_datacnt > 1) {
+ arc_buf_hdr_t *nhdr;
+ arc_buf_t **bufp;
+ uint64_t blksz = hdr->b_size;
+ spa_t *spa = hdr->b_spa;
+ arc_buf_contents_t type = hdr->b_type;
+ uint32_t flags = hdr->b_flags;
+
+ ASSERT(hdr->b_buf != buf || buf->b_next != NULL);
+ /*
+ * Pull the data off of this buf and attach it to
+ * a new anonymous buf.
+ */
+ (void) remove_reference(hdr, hash_lock, tag);
+ bufp = &hdr->b_buf;
+ while (*bufp != buf)
+ bufp = &(*bufp)->b_next;
+ *bufp = (*bufp)->b_next;
+ buf->b_next = NULL;
+
+ ASSERT3U(hdr->b_state->arcs_size, >=, hdr->b_size);
+ atomic_add_64(&hdr->b_state->arcs_size, -hdr->b_size);
+ if (refcount_is_zero(&hdr->b_refcnt)) {
+ uint64_t *size = &hdr->b_state->arcs_lsize[hdr->b_type];
+ ASSERT3U(*size, >=, hdr->b_size);
+ atomic_add_64(size, -hdr->b_size);
+ }
+ hdr->b_datacnt -= 1;
+ arc_cksum_verify(buf);
+
+ mutex_exit(hash_lock);
+
+ nhdr = kmem_cache_alloc(hdr_cache, KM_PUSHPAGE);
+ nhdr->b_size = blksz;
+ nhdr->b_spa = spa;
+ nhdr->b_type = type;
+ nhdr->b_buf = buf;
+ nhdr->b_state = arc_anon;
+ nhdr->b_arc_access = 0;
+ nhdr->b_flags = flags & ARC_L2_WRITING;
+ nhdr->b_l2hdr = NULL;
+ nhdr->b_datacnt = 1;
+ nhdr->b_freeze_cksum = NULL;
+ (void) refcount_add(&nhdr->b_refcnt, tag);
+ buf->b_hdr = nhdr;
+ rw_exit(&buf->b_lock);
+ atomic_add_64(&arc_anon->arcs_size, blksz);
+ } else {
+ rw_exit(&buf->b_lock);
+ ASSERT(refcount_count(&hdr->b_refcnt) == 1);
+ ASSERT(!list_link_active(&hdr->b_arc_node));
+ ASSERT(!HDR_IO_IN_PROGRESS(hdr));
+ arc_change_state(arc_anon, hdr, hash_lock);
+ hdr->b_arc_access = 0;
+ mutex_exit(hash_lock);
+
+ bzero(&hdr->b_dva, sizeof (dva_t));
+ hdr->b_birth = 0;
+ hdr->b_cksum0 = 0;
+ arc_buf_thaw(buf);
+ }
+ buf->b_efunc = NULL;
+ buf->b_private = NULL;
+
+ if (l2hdr) {
+ list_remove(l2hdr->b_dev->l2ad_buflist, hdr);
+ kmem_free(l2hdr, sizeof (l2arc_buf_hdr_t));
+ ARCSTAT_INCR(arcstat_l2_size, -buf_size);
+ mutex_exit(&l2arc_buflist_mtx);
+ }
+}
+
+int
+arc_released(arc_buf_t *buf)
+{
+ int released;
+
+ rw_enter(&buf->b_lock, RW_READER);
+ released = (buf->b_data != NULL && buf->b_hdr->b_state == arc_anon);
+ rw_exit(&buf->b_lock);
+ return (released);
+}
+
+int
+arc_has_callback(arc_buf_t *buf)
+{
+ int callback;
+
+ rw_enter(&buf->b_lock, RW_READER);
+ callback = (buf->b_efunc != NULL);
+ rw_exit(&buf->b_lock);
+ return (callback);
+}
+
+#ifdef ZFS_DEBUG
+int
+arc_referenced(arc_buf_t *buf)
+{
+ int referenced;
+
+ rw_enter(&buf->b_lock, RW_READER);
+ referenced = (refcount_count(&buf->b_hdr->b_refcnt));
+ rw_exit(&buf->b_lock);
+ return (referenced);
+}
+#endif
+
+static void
+arc_write_ready(zio_t *zio)
+{
+ arc_write_callback_t *callback = zio->io_private;
+ arc_buf_t *buf = callback->awcb_buf;
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+
+ ASSERT(!refcount_is_zero(&buf->b_hdr->b_refcnt));
+ callback->awcb_ready(zio, buf, callback->awcb_private);
+
+ /*
+ * If the IO is already in progress, then this is a re-write
+ * attempt, so we need to thaw and re-compute the cksum.
+ * It is the responsibility of the callback to handle the
+ * accounting for any re-write attempt.
+ */
+ if (HDR_IO_IN_PROGRESS(hdr)) {
+ mutex_enter(&hdr->b_freeze_lock);
+ if (hdr->b_freeze_cksum != NULL) {
+ kmem_free(hdr->b_freeze_cksum, sizeof (zio_cksum_t));
+ hdr->b_freeze_cksum = NULL;
+ }
+ mutex_exit(&hdr->b_freeze_lock);
+ }
+ arc_cksum_compute(buf, B_FALSE);
+ hdr->b_flags |= ARC_IO_IN_PROGRESS;
+}
+
+static void
+arc_write_done(zio_t *zio)
+{
+ arc_write_callback_t *callback = zio->io_private;
+ arc_buf_t *buf = callback->awcb_buf;
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+
+ hdr->b_acb = NULL;
+
+ hdr->b_dva = *BP_IDENTITY(zio->io_bp);
+ hdr->b_birth = zio->io_bp->blk_birth;
+ hdr->b_cksum0 = zio->io_bp->blk_cksum.zc_word[0];
+ /*
+ * If the block to be written was all-zero, we may have
+ * compressed it away. In this case no write was performed
+ * so there will be no dva/birth-date/checksum. The buffer
+ * must therefor remain anonymous (and uncached).
+ */
+ if (!BUF_EMPTY(hdr)) {
+ arc_buf_hdr_t *exists;
+ kmutex_t *hash_lock;
+
+ arc_cksum_verify(buf);
+
+ exists = buf_hash_insert(hdr, &hash_lock);
+ if (exists) {
+ /*
+ * This can only happen if we overwrite for
+ * sync-to-convergence, because we remove
+ * buffers from the hash table when we arc_free().
+ */
+ ASSERT(zio->io_flags & ZIO_FLAG_IO_REWRITE);
+ ASSERT(DVA_EQUAL(BP_IDENTITY(&zio->io_bp_orig),
+ BP_IDENTITY(zio->io_bp)));
+ ASSERT3U(zio->io_bp_orig.blk_birth, ==,
+ zio->io_bp->blk_birth);
+
+ ASSERT(refcount_is_zero(&exists->b_refcnt));
+ arc_change_state(arc_anon, exists, hash_lock);
+ mutex_exit(hash_lock);
+ arc_hdr_destroy(exists);
+ exists = buf_hash_insert(hdr, &hash_lock);
+ ASSERT3P(exists, ==, NULL);
+ }
+ hdr->b_flags &= ~ARC_IO_IN_PROGRESS;
+ /* if it's not anon, we are doing a scrub */
+ if (hdr->b_state == arc_anon)
+ arc_access(hdr, hash_lock);
+ mutex_exit(hash_lock);
+ } else if (callback->awcb_done == NULL) {
+ int destroy_hdr;
+ /*
+ * This is an anonymous buffer with no user callback,
+ * destroy it if there are no active references.
+ */
+ mutex_enter(&arc_eviction_mtx);
+ destroy_hdr = refcount_is_zero(&hdr->b_refcnt);
+ hdr->b_flags &= ~ARC_IO_IN_PROGRESS;
+ mutex_exit(&arc_eviction_mtx);
+ if (destroy_hdr)
+ arc_hdr_destroy(hdr);
+ } else {
+ hdr->b_flags &= ~ARC_IO_IN_PROGRESS;
+ }
+ hdr->b_flags &= ~ARC_STORED;
+
+ if (callback->awcb_done) {
+ ASSERT(!refcount_is_zero(&hdr->b_refcnt));
+ callback->awcb_done(zio, buf, callback->awcb_private);
+ }
+
+ kmem_free(callback, sizeof (arc_write_callback_t));
+}
+
+void
+write_policy(spa_t *spa, const writeprops_t *wp, zio_prop_t *zp)
+{
+ boolean_t ismd = (wp->wp_level > 0 || dmu_ot[wp->wp_type].ot_metadata);
+
+ /* Determine checksum setting */
+ if (ismd) {
+ /*
+ * Metadata always gets checksummed. If the data
+ * checksum is multi-bit correctable, and it's not a
+ * ZBT-style checksum, then it's suitable for metadata
+ * as well. Otherwise, the metadata checksum defaults
+ * to fletcher4.
+ */
+ if (zio_checksum_table[wp->wp_oschecksum].ci_correctable &&
+ !zio_checksum_table[wp->wp_oschecksum].ci_zbt)
+ zp->zp_checksum = wp->wp_oschecksum;
+ else
+ zp->zp_checksum = ZIO_CHECKSUM_FLETCHER_4;
+ } else {
+ zp->zp_checksum = zio_checksum_select(wp->wp_dnchecksum,
+ wp->wp_oschecksum);
+ }
+
+ /* Determine compression setting */
+ if (ismd) {
+ /*
+ * XXX -- we should design a compression algorithm
+ * that specializes in arrays of bps.
+ */
+ zp->zp_compress = zfs_mdcomp_disable ? ZIO_COMPRESS_EMPTY :
+ ZIO_COMPRESS_LZJB;
+ } else {
+ zp->zp_compress = zio_compress_select(wp->wp_dncompress,
+ wp->wp_oscompress);
+ }
+
+ zp->zp_type = wp->wp_type;
+ zp->zp_level = wp->wp_level;
+ zp->zp_ndvas = MIN(wp->wp_copies + ismd, spa_max_replication(spa));
+}
+
+zio_t *
+arc_write(zio_t *pio, spa_t *spa, const writeprops_t *wp,
+ boolean_t l2arc, uint64_t txg, blkptr_t *bp, arc_buf_t *buf,
+ arc_done_func_t *ready, arc_done_func_t *done, void *private, int priority,
+ int zio_flags, const zbookmark_t *zb)
+{
+ arc_buf_hdr_t *hdr = buf->b_hdr;
+ arc_write_callback_t *callback;
+ zio_t *zio;
+ zio_prop_t zp;
+
+ ASSERT(ready != NULL);
+ ASSERT(!HDR_IO_ERROR(hdr));
+ ASSERT((hdr->b_flags & ARC_IO_IN_PROGRESS) == 0);
+ ASSERT(hdr->b_acb == 0);
+ if (l2arc)
+ hdr->b_flags |= ARC_L2CACHE;
+ callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
+ callback->awcb_ready = ready;
+ callback->awcb_done = done;
+ callback->awcb_private = private;
+ callback->awcb_buf = buf;
+
+ write_policy(spa, wp, &zp);
+ zio = zio_write(pio, spa, txg, bp, buf->b_data, hdr->b_size, &zp,
+ arc_write_ready, arc_write_done, callback, priority, zio_flags, zb);
+
+ return (zio);
+}
+
+int
+arc_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, uint32_t arc_flags)
+{
+ arc_buf_hdr_t *ab;
+ kmutex_t *hash_lock;
+ zio_t *zio;
+
+ /*
+ * If this buffer is in the cache, release it, so it
+ * can be re-used.
+ */
+ ab = buf_hash_find(spa, BP_IDENTITY(bp), bp->blk_birth, &hash_lock);
+ if (ab != NULL) {
+ /*
+ * The checksum of blocks to free is not always
+ * preserved (eg. on the deadlist). However, if it is
+ * nonzero, it should match what we have in the cache.
+ */
+ ASSERT(bp->blk_cksum.zc_word[0] == 0 ||
+ bp->blk_cksum.zc_word[0] == ab->b_cksum0 ||
+ bp->blk_fill == BLK_FILL_ALREADY_FREED);
+
+ if (ab->b_state != arc_anon)
+ arc_change_state(arc_anon, ab, hash_lock);
+ if (HDR_IO_IN_PROGRESS(ab)) {
+ /*
+ * This should only happen when we prefetch.
+ */
+ ASSERT(ab->b_flags & ARC_PREFETCH);
+ ASSERT3U(ab->b_datacnt, ==, 1);
+ ab->b_flags |= ARC_FREED_IN_READ;
+ if (HDR_IN_HASH_TABLE(ab))
+ buf_hash_remove(ab);
+ ab->b_arc_access = 0;
+ bzero(&ab->b_dva, sizeof (dva_t));
+ ab->b_birth = 0;
+ ab->b_cksum0 = 0;
+ ab->b_buf->b_efunc = NULL;
+ ab->b_buf->b_private = NULL;
+ mutex_exit(hash_lock);
+ } else if (refcount_is_zero(&ab->b_refcnt)) {
+ ab->b_flags |= ARC_FREE_IN_PROGRESS;
+ mutex_exit(hash_lock);
+ arc_hdr_destroy(ab);
+ ARCSTAT_BUMP(arcstat_deleted);
+ } else {
+ /*
+ * We still have an active reference on this
+ * buffer. This can happen, e.g., from
+ * dbuf_unoverride().
+ */
+ ASSERT(!HDR_IN_HASH_TABLE(ab));
+ ab->b_arc_access = 0;
+ bzero(&ab->b_dva, sizeof (dva_t));
+ ab->b_birth = 0;
+ ab->b_cksum0 = 0;
+ ab->b_buf->b_efunc = NULL;
+ ab->b_buf->b_private = NULL;
+ mutex_exit(hash_lock);
+ }
+ }
+
+ zio = zio_free(pio, spa, txg, bp, done, private, ZIO_FLAG_MUSTSUCCEED);
+
+ if (arc_flags & ARC_WAIT)
+ return (zio_wait(zio));
+
+ ASSERT(arc_flags & ARC_NOWAIT);
+ zio_nowait(zio);
+
+ return (0);
+}
+
+static int
+arc_memory_throttle(uint64_t reserve, uint64_t txg)
+{
+#ifdef _KERNEL
+ uint64_t inflight_data = arc_anon->arcs_size;
+ uint64_t available_memory = ptob(freemem);
+ static uint64_t page_load = 0;
+ static uint64_t last_txg = 0;
+
+#if defined(__i386)
+ available_memory =
+ MIN(available_memory, vmem_size(heap_arena, VMEM_FREE));
+#endif
+ if (available_memory >= zfs_write_limit_max)
+ return (0);
+
+ if (txg > last_txg) {
+ last_txg = txg;
+ page_load = 0;
+ }
+ /*
+ * If we are in pageout, we know that memory is already tight,
+ * the arc is already going to be evicting, so we just want to
+ * continue to let page writes occur as quickly as possible.
+ */
+ if (curproc == proc_pageout) {
+ if (page_load > MAX(ptob(minfree), available_memory) / 4)
+ return (ERESTART);
+ /* Note: reserve is inflated, so we deflate */
+ page_load += reserve / 8;
+ return (0);
+ } else if (page_load > 0 && arc_reclaim_needed()) {
+ /* memory is low, delay before restarting */
+ ARCSTAT_INCR(arcstat_memory_throttle_count, 1);
+ return (EAGAIN);
+ }
+ page_load = 0;
+
+ if (arc_size > arc_c_min) {
+ uint64_t evictable_memory =
+ arc_mru->arcs_lsize[ARC_BUFC_DATA] +
+ arc_mru->arcs_lsize[ARC_BUFC_METADATA] +
+ arc_mfu->arcs_lsize[ARC_BUFC_DATA] +
+ arc_mfu->arcs_lsize[ARC_BUFC_METADATA];
+ available_memory += MIN(evictable_memory, arc_size - arc_c_min);
+ }
+
+ if (inflight_data > available_memory / 4) {
+ ARCSTAT_INCR(arcstat_memory_throttle_count, 1);
+ return (ERESTART);
+ }
+#endif
+ return (0);
+}
+
+void
+arc_tempreserve_clear(uint64_t reserve)
+{
+ atomic_add_64(&arc_tempreserve, -reserve);
+ ASSERT((int64_t)arc_tempreserve >= 0);
+}
+
+int
+arc_tempreserve_space(uint64_t reserve, uint64_t txg)
+{
+ int error;
+
+#ifdef ZFS_DEBUG
+ /*
+ * Once in a while, fail for no reason. Everything should cope.
+ */
+ if (spa_get_random(10000) == 0) {
+ dprintf("forcing random failure\n");
+ return (ERESTART);
+ }
+#endif
+ if (reserve > arc_c/4 && !arc_no_grow)
+ arc_c = MIN(arc_c_max, reserve * 4);
+ if (reserve > arc_c)
+ return (ENOMEM);
+
+ /*
+ * Writes will, almost always, require additional memory allocations
+ * in order to compress/encrypt/etc the data. We therefor need to
+ * make sure that there is sufficient available memory for this.
+ */
+ if (error = arc_memory_throttle(reserve, txg))
+ return (error);
+
+ /*
+ * Throttle writes when the amount of dirty data in the cache
+ * gets too large. We try to keep the cache less than half full
+ * of dirty blocks so that our sync times don't grow too large.
+ * Note: if two requests come in concurrently, we might let them
+ * both succeed, when one of them should fail. Not a huge deal.
+ */
+ if (reserve + arc_tempreserve + arc_anon->arcs_size > arc_c / 2 &&
+ arc_anon->arcs_size > arc_c / 4) {
+ dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
+ "anon_data=%lluK tempreserve=%lluK arc_c=%lluK\n",
+ arc_tempreserve>>10,
+ arc_anon->arcs_lsize[ARC_BUFC_METADATA]>>10,
+ arc_anon->arcs_lsize[ARC_BUFC_DATA]>>10,
+ reserve>>10, arc_c>>10);
+ return (ERESTART);
+ }
+ atomic_add_64(&arc_tempreserve, reserve);
+ return (0);
+}
+
+void
+arc_init(void)
+{
+ mutex_init(&arc_reclaim_thr_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&arc_reclaim_thr_cv, NULL, CV_DEFAULT, NULL);
+
+ /* Convert seconds to clock ticks */
+ arc_min_prefetch_lifespan = 1 * hz;
+
+ /* Start out with 1/8 of all memory */
+ arc_c = physmem * PAGESIZE / 8;
+
+#ifdef _KERNEL
+ /*
+ * On architectures where the physical memory can be larger
+ * than the addressable space (intel in 32-bit mode), we may
+ * need to limit the cache to 1/8 of VM size.
+ */
+ arc_c = MIN(arc_c, vmem_size(heap_arena, VMEM_ALLOC | VMEM_FREE) / 8);
+#endif
+
+ /* set min cache to 1/32 of all memory, or 64MB, whichever is more */
+ arc_c_min = MAX(arc_c / 4, 64<<20);
+ /* set max to 3/4 of all memory, or all but 1GB, whichever is more */
+ if (arc_c * 8 >= 1<<30)
+ arc_c_max = (arc_c * 8) - (1<<30);
+ else
+ arc_c_max = arc_c_min;
+ arc_c_max = MAX(arc_c * 6, arc_c_max);
+
+ /*
+ * Allow the tunables to override our calculations if they are
+ * reasonable (ie. over 64MB)
+ */
+ if (zfs_arc_max > 64<<20 && zfs_arc_max < physmem * PAGESIZE)
+ arc_c_max = zfs_arc_max;
+ if (zfs_arc_min > 64<<20 && zfs_arc_min <= arc_c_max)
+ arc_c_min = zfs_arc_min;
+
+ arc_c = arc_c_max;
+ arc_p = (arc_c >> 1);
+
+ /* limit meta-data to 1/4 of the arc capacity */
+ arc_meta_limit = arc_c_max / 4;
+
+ /* Allow the tunable to override if it is reasonable */
+ if (zfs_arc_meta_limit > 0 && zfs_arc_meta_limit <= arc_c_max)
+ arc_meta_limit = zfs_arc_meta_limit;
+
+ if (arc_c_min < arc_meta_limit / 2 && zfs_arc_min == 0)
+ arc_c_min = arc_meta_limit / 2;
+
+ /* if kmem_flags are set, lets try to use less memory */
+ if (kmem_debugging())
+ arc_c = arc_c / 2;
+ if (arc_c < arc_c_min)
+ arc_c = arc_c_min;
+
+ arc_anon = &ARC_anon;
+ arc_mru = &ARC_mru;
+ arc_mru_ghost = &ARC_mru_ghost;
+ arc_mfu = &ARC_mfu;
+ arc_mfu_ghost = &ARC_mfu_ghost;
+ arc_l2c_only = &ARC_l2c_only;
+ arc_size = 0;
+
+ mutex_init(&arc_anon->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&arc_mru->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&arc_mru_ghost->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&arc_mfu->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&arc_mfu_ghost->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&arc_l2c_only->arcs_mtx, NULL, MUTEX_DEFAULT, NULL);
+
+ list_create(&arc_mru->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mru->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mfu->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+ list_create(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
+ sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_arc_node));
+
+ buf_init();
+
+ arc_thread_exit = 0;
+ arc_eviction_list = NULL;
+ mutex_init(&arc_eviction_mtx, NULL, MUTEX_DEFAULT, NULL);
+ bzero(&arc_eviction_hdr, sizeof (arc_buf_hdr_t));
+
+ arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
+ sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
+
+ if (arc_ksp != NULL) {
+ arc_ksp->ks_data = &arc_stats;
+ kstat_install(arc_ksp);
+ }
+
+ (void) thread_create(NULL, 0, arc_reclaim_thread, NULL, 0, &p0,
+ TS_RUN, minclsyspri);
+
+ arc_dead = FALSE;
+ arc_warm = B_FALSE;
+
+ if (zfs_write_limit_max == 0)
+ zfs_write_limit_max = ptob(physmem) >> zfs_write_limit_shift;
+ else
+ zfs_write_limit_shift = 0;
+ mutex_init(&zfs_write_limit_lock, NULL, MUTEX_DEFAULT, NULL);
+}
+
+void
+arc_fini(void)
+{
+ mutex_enter(&arc_reclaim_thr_lock);
+ arc_thread_exit = 1;
+ while (arc_thread_exit != 0)
+ cv_wait(&arc_reclaim_thr_cv, &arc_reclaim_thr_lock);
+ mutex_exit(&arc_reclaim_thr_lock);
+
+ arc_flush(NULL);
+
+ arc_dead = TRUE;
+
+ if (arc_ksp != NULL) {
+ kstat_delete(arc_ksp);
+ arc_ksp = NULL;
+ }
+
+ mutex_destroy(&arc_eviction_mtx);
+ mutex_destroy(&arc_reclaim_thr_lock);
+ cv_destroy(&arc_reclaim_thr_cv);
+
+ list_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
+ list_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
+ list_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
+ list_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
+ list_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
+ list_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
+ list_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
+ list_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
+
+ mutex_destroy(&arc_anon->arcs_mtx);
+ mutex_destroy(&arc_mru->arcs_mtx);
+ mutex_destroy(&arc_mru_ghost->arcs_mtx);
+ mutex_destroy(&arc_mfu->arcs_mtx);
+ mutex_destroy(&arc_mfu_ghost->arcs_mtx);
+
+ mutex_destroy(&zfs_write_limit_lock);
+
+ buf_fini();
+}
+
+/*
+ * Level 2 ARC
+ *
+ * The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk.
+ * It uses dedicated storage devices to hold cached data, which are populated
+ * using large infrequent writes. The main role of this cache is to boost
+ * the performance of random read workloads. The intended L2ARC devices
+ * include short-stroked disks, solid state disks, and other media with
+ * substantially faster read latency than disk.
+ *
+ * +-----------------------+
+ * | ARC |
+ * +-----------------------+
+ * | ^ ^
+ * | | |
+ * l2arc_feed_thread() arc_read()
+ * | | |
+ * | l2arc read |
+ * V | |
+ * +---------------+ |
+ * | L2ARC | |
+ * +---------------+ |
+ * | ^ |
+ * l2arc_write() | |
+ * | | |
+ * V | |
+ * +-------+ +-------+
+ * | vdev | | vdev |
+ * | cache | | cache |
+ * +-------+ +-------+
+ * +=========+ .-----.
+ * : L2ARC : |-_____-|
+ * : devices : | Disks |
+ * +=========+ `-_____-'
+ *
+ * Read requests are satisfied from the following sources, in order:
+ *
+ * 1) ARC
+ * 2) vdev cache of L2ARC devices
+ * 3) L2ARC devices
+ * 4) vdev cache of disks
+ * 5) disks
+ *
+ * Some L2ARC device types exhibit extremely slow write performance.
+ * To accommodate for this there are some significant differences between
+ * the L2ARC and traditional cache design:
+ *
+ * 1. There is no eviction path from the ARC to the L2ARC. Evictions from
+ * the ARC behave as usual, freeing buffers and placing headers on ghost
+ * lists. The ARC does not send buffers to the L2ARC during eviction as
+ * this would add inflated write latencies for all ARC memory pressure.
+ *
+ * 2. The L2ARC attempts to cache data from the ARC before it is evicted.
+ * It does this by periodically scanning buffers from the eviction-end of
+ * the MFU and MRU ARC lists, copying them to the L2ARC devices if they are
+ * not already there. It scans until a headroom of buffers is satisfied,
+ * which itself is a buffer for ARC eviction. The thread that does this is
+ * l2arc_feed_thread(), illustrated below; example sizes are included to
+ * provide a better sense of ratio than this diagram:
+ *
+ * head --> tail
+ * +---------------------+----------+
+ * ARC_mfu |:::::#:::::::::::::::|o#o###o###|-->. # already on L2ARC
+ * +---------------------+----------+ | o L2ARC eligible
+ * ARC_mru |:#:::::::::::::::::::|#o#ooo####|-->| : ARC buffer
+ * +---------------------+----------+ |
+ * 15.9 Gbytes ^ 32 Mbytes |
+ * headroom |
+ * l2arc_feed_thread()
+ * |
+ * l2arc write hand <--[oooo]--'
+ * | 8 Mbyte
+ * | write max
+ * V
+ * +==============================+
+ * L2ARC dev |####|#|###|###| |####| ... |
+ * +==============================+
+ * 32 Gbytes
+ *
+ * 3. If an ARC buffer is copied to the L2ARC but then hit instead of
+ * evicted, then the L2ARC has cached a buffer much sooner than it probably
+ * needed to, potentially wasting L2ARC device bandwidth and storage. It is
+ * safe to say that this is an uncommon case, since buffers at the end of
+ * the ARC lists have moved there due to inactivity.
+ *
+ * 4. If the ARC evicts faster than the L2ARC can maintain a headroom,
+ * then the L2ARC simply misses copying some buffers. This serves as a
+ * pressure valve to prevent heavy read workloads from both stalling the ARC
+ * with waits and clogging the L2ARC with writes. This also helps prevent
+ * the potential for the L2ARC to churn if it attempts to cache content too
+ * quickly, such as during backups of the entire pool.
+ *
+ * 5. After system boot and before the ARC has filled main memory, there are
+ * no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru
+ * lists can remain mostly static. Instead of searching from tail of these
+ * lists as pictured, the l2arc_feed_thread() will search from the list heads
+ * for eligible buffers, greatly increasing its chance of finding them.
+ *
+ * The L2ARC device write speed is also boosted during this time so that
+ * the L2ARC warms up faster. Since there have been no ARC evictions yet,
+ * there are no L2ARC reads, and no fear of degrading read performance
+ * through increased writes.
+ *
+ * 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that
+ * the vdev queue can aggregate them into larger and fewer writes. Each
+ * device is written to in a rotor fashion, sweeping writes through
+ * available space then repeating.
+ *
+ * 7. The L2ARC does not store dirty content. It never needs to flush
+ * write buffers back to disk based storage.
+ *
+ * 8. If an ARC buffer is written (and dirtied) which also exists in the
+ * L2ARC, the now stale L2ARC buffer is immediately dropped.
+ *
+ * The performance of the L2ARC can be tweaked by a number of tunables, which
+ * may be necessary for different workloads:
+ *
+ * l2arc_write_max max write bytes per interval
+ * l2arc_write_boost extra write bytes during device warmup
+ * l2arc_noprefetch skip caching prefetched buffers
+ * l2arc_headroom number of max device writes to precache
+ * l2arc_feed_secs seconds between L2ARC writing
+ *
+ * Tunables may be removed or added as future performance improvements are
+ * integrated, and also may become zpool properties.
+ */
+
+static void
+l2arc_hdr_stat_add(void)
+{
+ ARCSTAT_INCR(arcstat_l2_hdr_size, HDR_SIZE + L2HDR_SIZE);
+ ARCSTAT_INCR(arcstat_hdr_size, -HDR_SIZE);
+}
+
+static void
+l2arc_hdr_stat_remove(void)
+{
+ ARCSTAT_INCR(arcstat_l2_hdr_size, -(HDR_SIZE + L2HDR_SIZE));
+ ARCSTAT_INCR(arcstat_hdr_size, HDR_SIZE);
+}
+
+/*
+ * Cycle through L2ARC devices. This is how L2ARC load balances.
+ * If a device is returned, this also returns holding the spa config lock.
+ */
+static l2arc_dev_t *
+l2arc_dev_get_next(void)
+{
+ l2arc_dev_t *first, *next = NULL;
+
+ /*
+ * Lock out the removal of spas (spa_namespace_lock), then removal
+ * of cache devices (l2arc_dev_mtx). Once a device has been selected,
+ * both locks will be dropped and a spa config lock held instead.
+ */
+ mutex_enter(&spa_namespace_lock);
+ mutex_enter(&l2arc_dev_mtx);
+
+ /* if there are no vdevs, there is nothing to do */
+ if (l2arc_ndev == 0)
+ goto out;
+
+ first = NULL;
+ next = l2arc_dev_last;
+ do {
+ /* loop around the list looking for a non-faulted vdev */
+ if (next == NULL) {
+ next = list_head(l2arc_dev_list);
+ } else {
+ next = list_next(l2arc_dev_list, next);
+ if (next == NULL)
+ next = list_head(l2arc_dev_list);
+ }
+
+ /* if we have come back to the start, bail out */
+ if (first == NULL)
+ first = next;
+ else if (next == first)
+ break;
+
+ } while (vdev_is_dead(next->l2ad_vdev));
+
+ /* if we were unable to find any usable vdevs, return NULL */
+ if (vdev_is_dead(next->l2ad_vdev))
+ next = NULL;
+
+ l2arc_dev_last = next;
+
+out:
+ mutex_exit(&l2arc_dev_mtx);
+
+ /*
+ * Grab the config lock to prevent the 'next' device from being
+ * removed while we are writing to it.
+ */
+ if (next != NULL)
+ spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER);
+ mutex_exit(&spa_namespace_lock);
+
+ return (next);
+}
+
+/*
+ * Free buffers that were tagged for destruction.
+ */
+static void
+l2arc_do_free_on_write()
+{
+ list_t *buflist;
+ l2arc_data_free_t *df, *df_prev;
+
+ mutex_enter(&l2arc_free_on_write_mtx);
+ buflist = l2arc_free_on_write;
+
+ for (df = list_tail(buflist); df; df = df_prev) {
+ df_prev = list_prev(buflist, df);
+ ASSERT(df->l2df_data != NULL);
+ ASSERT(df->l2df_func != NULL);
+ df->l2df_func(df->l2df_data, df->l2df_size);
+ list_remove(buflist, df);
+ kmem_free(df, sizeof (l2arc_data_free_t));
+ }
+
+ mutex_exit(&l2arc_free_on_write_mtx);
+}
+
+/*
+ * A write to a cache device has completed. Update all headers to allow
+ * reads from these buffers to begin.
+ */
+static void
+l2arc_write_done(zio_t *zio)
+{
+ l2arc_write_callback_t *cb;
+ l2arc_dev_t *dev;
+ list_t *buflist;
+ arc_buf_hdr_t *head, *ab, *ab_prev;
+ l2arc_buf_hdr_t *abl2;
+ kmutex_t *hash_lock;
+
+ cb = zio->io_private;
+ ASSERT(cb != NULL);
+ dev = cb->l2wcb_dev;
+ ASSERT(dev != NULL);
+ head = cb->l2wcb_head;
+ ASSERT(head != NULL);
+ buflist = dev->l2ad_buflist;
+ ASSERT(buflist != NULL);
+ DTRACE_PROBE2(l2arc__iodone, zio_t *, zio,
+ l2arc_write_callback_t *, cb);
+
+ if (zio->io_error != 0)
+ ARCSTAT_BUMP(arcstat_l2_writes_error);
+
+ mutex_enter(&l2arc_buflist_mtx);
+
+ /*
+ * All writes completed, or an error was hit.
+ */
+ for (ab = list_prev(buflist, head); ab; ab = ab_prev) {
+ ab_prev = list_prev(buflist, ab);
+
+ hash_lock = HDR_LOCK(ab);
+ if (!mutex_tryenter(hash_lock)) {
+ /*
+ * This buffer misses out. It may be in a stage
+ * of eviction. Its ARC_L2_WRITING flag will be
+ * left set, denying reads to this buffer.
+ */
+ ARCSTAT_BUMP(arcstat_l2_writes_hdr_miss);
+ continue;
+ }
+
+ if (zio->io_error != 0) {
+ /*
+ * Error - drop L2ARC entry.
+ */
+ list_remove(buflist, ab);
+ abl2 = ab->b_l2hdr;
+ ab->b_l2hdr = NULL;
+ kmem_free(abl2, sizeof (l2arc_buf_hdr_t));
+ ARCSTAT_INCR(arcstat_l2_size, -ab->b_size);
+ }
+
+ /*
+ * Allow ARC to begin reads to this L2ARC entry.
+ */
+ ab->b_flags &= ~ARC_L2_WRITING;
+
+ mutex_exit(hash_lock);
+ }
+
+ atomic_inc_64(&l2arc_writes_done);
+ list_remove(buflist, head);
+ kmem_cache_free(hdr_cache, head);
+ mutex_exit(&l2arc_buflist_mtx);
+
+ l2arc_do_free_on_write();
+
+ kmem_free(cb, sizeof (l2arc_write_callback_t));
+}
+
+/*
+ * A read to a cache device completed. Validate buffer contents before
+ * handing over to the regular ARC routines.
+ */
+static void
+l2arc_read_done(zio_t *zio)
+{
+ l2arc_read_callback_t *cb;
+ arc_buf_hdr_t *hdr;
+ arc_buf_t *buf;
+ kmutex_t *hash_lock;
+ int equal;
+
+ ASSERT(zio->io_vd != NULL);
+ ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE);
+
+ spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd);
+
+ cb = zio->io_private;
+ ASSERT(cb != NULL);
+ buf = cb->l2rcb_buf;
+ ASSERT(buf != NULL);
+ hdr = buf->b_hdr;
+ ASSERT(hdr != NULL);
+
+ hash_lock = HDR_LOCK(hdr);
+ mutex_enter(hash_lock);
+
+ /*
+ * Check this survived the L2ARC journey.
+ */
+ equal = arc_cksum_equal(buf);
+ if (equal && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) {
+ mutex_exit(hash_lock);
+ zio->io_private = buf;
+ zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
+ zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
+ arc_read_done(zio);
+ } else {
+ mutex_exit(hash_lock);
+ /*
+ * Buffer didn't survive caching. Increment stats and
+ * reissue to the original storage device.
+ */
+ if (zio->io_error != 0) {
+ ARCSTAT_BUMP(arcstat_l2_io_error);
+ } else {
+ zio->io_error = EIO;
+ }
+ if (!equal)
+ ARCSTAT_BUMP(arcstat_l2_cksum_bad);
+
+ /*
+ * If there's no waiter, issue an async i/o to the primary
+ * storage now. If there *is* a waiter, the caller must
+ * issue the i/o in a context where it's OK to block.
+ */
+ if (zio->io_waiter == NULL)
+ zio_nowait(zio_read(zio->io_parent,
+ cb->l2rcb_spa, &cb->l2rcb_bp,
+ buf->b_data, zio->io_size, arc_read_done, buf,
+ zio->io_priority, cb->l2rcb_flags, &cb->l2rcb_zb));
+ }
+
+ kmem_free(cb, sizeof (l2arc_read_callback_t));
+}
+
+/*
+ * This is the list priority from which the L2ARC will search for pages to
+ * cache. This is used within loops (0..3) to cycle through lists in the
+ * desired order. This order can have a significant effect on cache
+ * performance.
+ *
+ * Currently the metadata lists are hit first, MFU then MRU, followed by
+ * the data lists. This function returns a locked list, and also returns
+ * the lock pointer.
+ */
+static list_t *
+l2arc_list_locked(int list_num, kmutex_t **lock)
+{
+ list_t *list;
+
+ ASSERT(list_num >= 0 && list_num <= 3);
+
+ switch (list_num) {
+ case 0:
+ list = &arc_mfu->arcs_list[ARC_BUFC_METADATA];
+ *lock = &arc_mfu->arcs_mtx;
+ break;
+ case 1:
+ list = &arc_mru->arcs_list[ARC_BUFC_METADATA];
+ *lock = &arc_mru->arcs_mtx;
+ break;
+ case 2:
+ list = &arc_mfu->arcs_list[ARC_BUFC_DATA];
+ *lock = &arc_mfu->arcs_mtx;
+ break;
+ case 3:
+ list = &arc_mru->arcs_list[ARC_BUFC_DATA];
+ *lock = &arc_mru->arcs_mtx;
+ break;
+ }
+
+ ASSERT(!(MUTEX_HELD(*lock)));
+ mutex_enter(*lock);
+ return (list);
+}
+
+/*
+ * Evict buffers from the device write hand to the distance specified in
+ * bytes. This distance may span populated buffers, it may span nothing.
+ * This is clearing a region on the L2ARC device ready for writing.
+ * If the 'all' boolean is set, every buffer is evicted.
+ */
+static void
+l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all)
+{
+ list_t *buflist;
+ l2arc_buf_hdr_t *abl2;
+ arc_buf_hdr_t *ab, *ab_prev;
+ kmutex_t *hash_lock;
+ uint64_t taddr;
+
+ buflist = dev->l2ad_buflist;
+
+ if (buflist == NULL)
+ return;
+
+ if (!all && dev->l2ad_first) {
+ /*
+ * This is the first sweep through the device. There is
+ * nothing to evict.
+ */
+ return;
+ }
+
+ if (dev->l2ad_hand >= (dev->l2ad_end - (2 * distance))) {
+ /*
+ * When nearing the end of the device, evict to the end
+ * before the device write hand jumps to the start.
+ */
+ taddr = dev->l2ad_end;
+ } else {
+ taddr = dev->l2ad_hand + distance;
+ }
+ DTRACE_PROBE4(l2arc__evict, l2arc_dev_t *, dev, list_t *, buflist,
+ uint64_t, taddr, boolean_t, all);
+
+top:
+ mutex_enter(&l2arc_buflist_mtx);
+ for (ab = list_tail(buflist); ab; ab = ab_prev) {
+ ab_prev = list_prev(buflist, ab);
+
+ hash_lock = HDR_LOCK(ab);
+ if (!mutex_tryenter(hash_lock)) {
+ /*
+ * Missed the hash lock. Retry.
+ */
+ ARCSTAT_BUMP(arcstat_l2_evict_lock_retry);
+ mutex_exit(&l2arc_buflist_mtx);
+ mutex_enter(hash_lock);
+ mutex_exit(hash_lock);
+ goto top;
+ }
+
+ if (HDR_L2_WRITE_HEAD(ab)) {
+ /*
+ * We hit a write head node. Leave it for
+ * l2arc_write_done().
+ */
+ list_remove(buflist, ab);
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if (!all && ab->b_l2hdr != NULL &&
+ (ab->b_l2hdr->b_daddr > taddr ||
+ ab->b_l2hdr->b_daddr < dev->l2ad_hand)) {
+ /*
+ * We've evicted to the target address,
+ * or the end of the device.
+ */
+ mutex_exit(hash_lock);
+ break;
+ }
+
+ if (HDR_FREE_IN_PROGRESS(ab)) {
+ /*
+ * Already on the path to destruction.
+ */
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if (ab->b_state == arc_l2c_only) {
+ ASSERT(!HDR_L2_READING(ab));
+ /*
+ * This doesn't exist in the ARC. Destroy.
+ * arc_hdr_destroy() will call list_remove()
+ * and decrement arcstat_l2_size.
+ */
+ arc_change_state(arc_anon, ab, hash_lock);
+ arc_hdr_destroy(ab);
+ } else {
+ /*
+ * Invalidate issued or about to be issued
+ * reads, since we may be about to write
+ * over this location.
+ */
+ if (HDR_L2_READING(ab)) {
+ ARCSTAT_BUMP(arcstat_l2_evict_reading);
+ ab->b_flags |= ARC_L2_EVICTED;
+ }
+
+ /*
+ * Tell ARC this no longer exists in L2ARC.
+ */
+ if (ab->b_l2hdr != NULL) {
+ abl2 = ab->b_l2hdr;
+ ab->b_l2hdr = NULL;
+ kmem_free(abl2, sizeof (l2arc_buf_hdr_t));
+ ARCSTAT_INCR(arcstat_l2_size, -ab->b_size);
+ }
+ list_remove(buflist, ab);
+
+ /*
+ * This may have been leftover after a
+ * failed write.
+ */
+ ab->b_flags &= ~ARC_L2_WRITING;
+ }
+ mutex_exit(hash_lock);
+ }
+ mutex_exit(&l2arc_buflist_mtx);
+
+ spa_l2cache_space_update(dev->l2ad_vdev, 0, -(taddr - dev->l2ad_evict));
+ dev->l2ad_evict = taddr;
+}
+
+/*
+ * Find and write ARC buffers to the L2ARC device.
+ *
+ * An ARC_L2_WRITING flag is set so that the L2ARC buffers are not valid
+ * for reading until they have completed writing.
+ */
+static void
+l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz)
+{
+ arc_buf_hdr_t *ab, *ab_prev, *head;
+ l2arc_buf_hdr_t *hdrl2;
+ list_t *list;
+ uint64_t passed_sz, write_sz, buf_sz, headroom;
+ void *buf_data;
+ kmutex_t *hash_lock, *list_lock;
+ boolean_t have_lock, full;
+ l2arc_write_callback_t *cb;
+ zio_t *pio, *wzio;
+
+ ASSERT(dev->l2ad_vdev != NULL);
+
+ pio = NULL;
+ write_sz = 0;
+ full = B_FALSE;
+ head = kmem_cache_alloc(hdr_cache, KM_PUSHPAGE);
+ head->b_flags |= ARC_L2_WRITE_HEAD;
+
+ /*
+ * Copy buffers for L2ARC writing.
+ */
+ mutex_enter(&l2arc_buflist_mtx);
+ for (int try = 0; try <= 3; try++) {
+ list = l2arc_list_locked(try, &list_lock);
+ passed_sz = 0;
+
+ /*
+ * L2ARC fast warmup.
+ *
+ * Until the ARC is warm and starts to evict, read from the
+ * head of the ARC lists rather than the tail.
+ */
+ headroom = target_sz * l2arc_headroom;
+ if (arc_warm == B_FALSE)
+ ab = list_head(list);
+ else
+ ab = list_tail(list);
+
+ for (; ab; ab = ab_prev) {
+ if (arc_warm == B_FALSE)
+ ab_prev = list_next(list, ab);
+ else
+ ab_prev = list_prev(list, ab);
+
+ hash_lock = HDR_LOCK(ab);
+ have_lock = MUTEX_HELD(hash_lock);
+ if (!have_lock && !mutex_tryenter(hash_lock)) {
+ /*
+ * Skip this buffer rather than waiting.
+ */
+ continue;
+ }
+
+ passed_sz += ab->b_size;
+ if (passed_sz > headroom) {
+ /*
+ * Searched too far.
+ */
+ mutex_exit(hash_lock);
+ break;
+ }
+
+ if (ab->b_spa != spa) {
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if (ab->b_l2hdr != NULL) {
+ /*
+ * Already in L2ARC.
+ */
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if (HDR_IO_IN_PROGRESS(ab) || !HDR_L2CACHE(ab)) {
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if ((write_sz + ab->b_size) > target_sz) {
+ full = B_TRUE;
+ mutex_exit(hash_lock);
+ break;
+ }
+
+ if (ab->b_buf == NULL) {
+ DTRACE_PROBE1(l2arc__buf__null, void *, ab);
+ mutex_exit(hash_lock);
+ continue;
+ }
+
+ if (pio == NULL) {
+ /*
+ * Insert a dummy header on the buflist so
+ * l2arc_write_done() can find where the
+ * write buffers begin without searching.
+ */
+ list_insert_head(dev->l2ad_buflist, head);
+
+ cb = kmem_alloc(
+ sizeof (l2arc_write_callback_t), KM_SLEEP);
+ cb->l2wcb_dev = dev;
+ cb->l2wcb_head = head;
+ pio = zio_root(spa, l2arc_write_done, cb,
+ ZIO_FLAG_CANFAIL);
+ }
+
+ /*
+ * Create and add a new L2ARC header.
+ */
+ hdrl2 = kmem_zalloc(sizeof (l2arc_buf_hdr_t), KM_SLEEP);
+ hdrl2->b_dev = dev;
+ hdrl2->b_daddr = dev->l2ad_hand;
+
+ ab->b_flags |= ARC_L2_WRITING;
+ ab->b_l2hdr = hdrl2;
+ list_insert_head(dev->l2ad_buflist, ab);
+ buf_data = ab->b_buf->b_data;
+ buf_sz = ab->b_size;
+
+ /*
+ * Compute and store the buffer cksum before
+ * writing. On debug the cksum is verified first.
+ */
+ arc_cksum_verify(ab->b_buf);
+ arc_cksum_compute(ab->b_buf, B_TRUE);
+
+ mutex_exit(hash_lock);
+
+ wzio = zio_write_phys(pio, dev->l2ad_vdev,
+ dev->l2ad_hand, buf_sz, buf_data, ZIO_CHECKSUM_OFF,
+ NULL, NULL, ZIO_PRIORITY_ASYNC_WRITE,
+ ZIO_FLAG_CANFAIL, B_FALSE);
+
+ DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev,
+ zio_t *, wzio);
+ (void) zio_nowait(wzio);
+
+ /*
+ * Keep the clock hand suitably device-aligned.
+ */
+ buf_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz);
+
+ write_sz += buf_sz;
+ dev->l2ad_hand += buf_sz;
+ }
+
+ mutex_exit(list_lock);
+
+ if (full == B_TRUE)
+ break;
+ }
+ mutex_exit(&l2arc_buflist_mtx);
+
+ if (pio == NULL) {
+ ASSERT3U(write_sz, ==, 0);
+ kmem_cache_free(hdr_cache, head);
+ return;
+ }
+
+ ASSERT3U(write_sz, <=, target_sz);
+ ARCSTAT_BUMP(arcstat_l2_writes_sent);
+ ARCSTAT_INCR(arcstat_l2_size, write_sz);
+ spa_l2cache_space_update(dev->l2ad_vdev, 0, write_sz);
+
+ /*
+ * Bump device hand to the device start if it is approaching the end.
+ * l2arc_evict() will already have evicted ahead for this case.
+ */
+ if (dev->l2ad_hand >= (dev->l2ad_end - target_sz)) {
+ spa_l2cache_space_update(dev->l2ad_vdev, 0,
+ dev->l2ad_end - dev->l2ad_hand);
+ dev->l2ad_hand = dev->l2ad_start;
+ dev->l2ad_evict = dev->l2ad_start;
+ dev->l2ad_first = B_FALSE;
+ }
+
+ (void) zio_wait(pio);
+}
+
+/*
+ * This thread feeds the L2ARC at regular intervals. This is the beating
+ * heart of the L2ARC.
+ */
+static void
+l2arc_feed_thread(void)
+{
+ callb_cpr_t cpr;
+ l2arc_dev_t *dev;
+ spa_t *spa;
+ uint64_t size;
+
+ CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG);
+
+ mutex_enter(&l2arc_feed_thr_lock);
+
+ while (l2arc_thread_exit == 0) {
+ /*
+ * Pause for l2arc_feed_secs seconds between writes.
+ */
+ CALLB_CPR_SAFE_BEGIN(&cpr);
+ (void) cv_timedwait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock,
+ lbolt + (hz * l2arc_feed_secs));
+ CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock);
+
+ /*
+ * Quick check for L2ARC devices.
+ */
+ mutex_enter(&l2arc_dev_mtx);
+ if (l2arc_ndev == 0) {
+ mutex_exit(&l2arc_dev_mtx);
+ continue;
+ }
+ mutex_exit(&l2arc_dev_mtx);
+
+ /*
+ * This selects the next l2arc device to write to, and in
+ * doing so the next spa to feed from: dev->l2ad_spa. This
+ * will return NULL if there are now no l2arc devices or if
+ * they are all faulted.
+ *
+ * If a device is returned, its spa's config lock is also
+ * held to prevent device removal. l2arc_dev_get_next()
+ * will grab and release l2arc_dev_mtx.
+ */
+ if ((dev = l2arc_dev_get_next()) == NULL)
+ continue;
+
+ spa = dev->l2ad_spa;
+ ASSERT(spa != NULL);
+
+ /*
+ * Avoid contributing to memory pressure.
+ */
+ if (arc_reclaim_needed()) {
+ ARCSTAT_BUMP(arcstat_l2_abort_lowmem);
+ spa_config_exit(spa, SCL_L2ARC, dev);
+ continue;
+ }
+
+ ARCSTAT_BUMP(arcstat_l2_feeds);
+
+ size = dev->l2ad_write;
+ if (arc_warm == B_FALSE)
+ size += dev->l2ad_boost;
+
+ /*
+ * Evict L2ARC buffers that will be overwritten.
+ */
+ l2arc_evict(dev, size, B_FALSE);
+
+ /*
+ * Write ARC buffers.
+ */
+ l2arc_write_buffers(spa, dev, size);
+ spa_config_exit(spa, SCL_L2ARC, dev);
+ }
+
+ l2arc_thread_exit = 0;
+ cv_broadcast(&l2arc_feed_thr_cv);
+ CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */
+ thread_exit();
+}
+
+boolean_t
+l2arc_vdev_present(vdev_t *vd)
+{
+ l2arc_dev_t *dev;
+
+ mutex_enter(&l2arc_dev_mtx);
+ for (dev = list_head(l2arc_dev_list); dev != NULL;
+ dev = list_next(l2arc_dev_list, dev)) {
+ if (dev->l2ad_vdev == vd)
+ break;
+ }
+ mutex_exit(&l2arc_dev_mtx);
+
+ return (dev != NULL);
+}
+
+/*
+ * Add a vdev for use by the L2ARC. By this point the spa has already
+ * validated the vdev and opened it.
+ */
+void
+l2arc_add_vdev(spa_t *spa, vdev_t *vd, uint64_t start, uint64_t end)
+{
+ l2arc_dev_t *adddev;
+
+ ASSERT(!l2arc_vdev_present(vd));
+
+ /*
+ * Create a new l2arc device entry.
+ */
+ adddev = kmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP);
+ adddev->l2ad_spa = spa;
+ adddev->l2ad_vdev = vd;
+ adddev->l2ad_write = l2arc_write_max;
+ adddev->l2ad_boost = l2arc_write_boost;
+ adddev->l2ad_start = start;
+ adddev->l2ad_end = end;
+ adddev->l2ad_hand = adddev->l2ad_start;
+ adddev->l2ad_evict = adddev->l2ad_start;
+ adddev->l2ad_first = B_TRUE;
+ ASSERT3U(adddev->l2ad_write, >, 0);
+
+ /*
+ * This is a list of all ARC buffers that are still valid on the
+ * device.
+ */
+ adddev->l2ad_buflist = kmem_zalloc(sizeof (list_t), KM_SLEEP);
+ list_create(adddev->l2ad_buflist, sizeof (arc_buf_hdr_t),
+ offsetof(arc_buf_hdr_t, b_l2node));
+
+ spa_l2cache_space_update(vd, adddev->l2ad_end - adddev->l2ad_hand, 0);
+
+ /*
+ * Add device to global list
+ */
+ mutex_enter(&l2arc_dev_mtx);
+ list_insert_head(l2arc_dev_list, adddev);
+ atomic_inc_64(&l2arc_ndev);
+ mutex_exit(&l2arc_dev_mtx);
+}
+
+/*
+ * Remove a vdev from the L2ARC.
+ */
+void
+l2arc_remove_vdev(vdev_t *vd)
+{
+ l2arc_dev_t *dev, *nextdev, *remdev = NULL;
+
+ /*
+ * Find the device by vdev
+ */
+ mutex_enter(&l2arc_dev_mtx);
+ for (dev = list_head(l2arc_dev_list); dev; dev = nextdev) {
+ nextdev = list_next(l2arc_dev_list, dev);
+ if (vd == dev->l2ad_vdev) {
+ remdev = dev;
+ break;
+ }
+ }
+ ASSERT(remdev != NULL);
+
+ /*
+ * Remove device from global list
+ */
+ list_remove(l2arc_dev_list, remdev);
+ l2arc_dev_last = NULL; /* may have been invalidated */
+ atomic_dec_64(&l2arc_ndev);
+ mutex_exit(&l2arc_dev_mtx);
+
+ /*
+ * Clear all buflists and ARC references. L2ARC device flush.
+ */
+ l2arc_evict(remdev, 0, B_TRUE);
+ list_destroy(remdev->l2ad_buflist);
+ kmem_free(remdev->l2ad_buflist, sizeof (list_t));
+ kmem_free(remdev, sizeof (l2arc_dev_t));
+}
+
+void
+l2arc_init(void)
+{
+ l2arc_thread_exit = 0;
+ l2arc_ndev = 0;
+ l2arc_writes_sent = 0;
+ l2arc_writes_done = 0;
+
+ mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL);
+ mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&l2arc_buflist_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL);
+
+ l2arc_dev_list = &L2ARC_dev_list;
+ l2arc_free_on_write = &L2ARC_free_on_write;
+ list_create(l2arc_dev_list, sizeof (l2arc_dev_t),
+ offsetof(l2arc_dev_t, l2ad_node));
+ list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t),
+ offsetof(l2arc_data_free_t, l2df_list_node));
+}
+
+void
+l2arc_fini(void)
+{
+ /*
+ * This is called from dmu_fini(), which is called from spa_fini();
+ * Because of this, we can assume that all l2arc devices have
+ * already been removed when the pools themselves were removed.
+ */
+
+ l2arc_do_free_on_write();
+
+ mutex_destroy(&l2arc_feed_thr_lock);
+ cv_destroy(&l2arc_feed_thr_cv);
+ mutex_destroy(&l2arc_dev_mtx);
+ mutex_destroy(&l2arc_buflist_mtx);
+ mutex_destroy(&l2arc_free_on_write_mtx);
+
+ list_destroy(l2arc_dev_list);
+ list_destroy(l2arc_free_on_write);
+}
+
+void
+l2arc_start(void)
+{
+ if (!(spa_mode & FWRITE))
+ return;
+
+ (void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0,
+ TS_RUN, minclsyspri);
+}
+
+void
+l2arc_stop(void)
+{
+ if (!(spa_mode & FWRITE))
+ return;
+
+ mutex_enter(&l2arc_feed_thr_lock);
+ cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */
+ l2arc_thread_exit = 1;
+ while (l2arc_thread_exit != 0)
+ cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock);
+ mutex_exit(&l2arc_feed_thr_lock);
+}
diff --git a/module/zfs/bplist.c b/module/zfs/bplist.c
new file mode 100644
index 000000000..93b7741d7
--- /dev/null
+++ b/module/zfs/bplist.c
@@ -0,0 +1,349 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/bplist.h>
+#include <sys/zfs_context.h>
+
+static int
+bplist_hold(bplist_t *bpl)
+{
+ ASSERT(MUTEX_HELD(&bpl->bpl_lock));
+ if (bpl->bpl_dbuf == NULL) {
+ int err = dmu_bonus_hold(bpl->bpl_mos,
+ bpl->bpl_object, bpl, &bpl->bpl_dbuf);
+ if (err)
+ return (err);
+ bpl->bpl_phys = bpl->bpl_dbuf->db_data;
+ }
+ return (0);
+}
+
+uint64_t
+bplist_create(objset_t *mos, int blocksize, dmu_tx_t *tx)
+{
+ int size;
+
+ size = spa_version(dmu_objset_spa(mos)) < SPA_VERSION_BPLIST_ACCOUNT ?
+ BPLIST_SIZE_V0 : sizeof (bplist_phys_t);
+
+ return (dmu_object_alloc(mos, DMU_OT_BPLIST, blocksize,
+ DMU_OT_BPLIST_HDR, size, tx));
+}
+
+void
+bplist_destroy(objset_t *mos, uint64_t object, dmu_tx_t *tx)
+{
+ VERIFY(dmu_object_free(mos, object, tx) == 0);
+}
+
+int
+bplist_open(bplist_t *bpl, objset_t *mos, uint64_t object)
+{
+ dmu_object_info_t doi;
+ int err;
+
+ err = dmu_object_info(mos, object, &doi);
+ if (err)
+ return (err);
+
+ mutex_enter(&bpl->bpl_lock);
+
+ ASSERT(bpl->bpl_dbuf == NULL);
+ ASSERT(bpl->bpl_phys == NULL);
+ ASSERT(bpl->bpl_cached_dbuf == NULL);
+ ASSERT(bpl->bpl_queue == NULL);
+ ASSERT(object != 0);
+ ASSERT3U(doi.doi_type, ==, DMU_OT_BPLIST);
+ ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPLIST_HDR);
+
+ bpl->bpl_mos = mos;
+ bpl->bpl_object = object;
+ bpl->bpl_blockshift = highbit(doi.doi_data_block_size - 1);
+ bpl->bpl_bpshift = bpl->bpl_blockshift - SPA_BLKPTRSHIFT;
+ bpl->bpl_havecomp = (doi.doi_bonus_size == sizeof (bplist_phys_t));
+
+ mutex_exit(&bpl->bpl_lock);
+ return (0);
+}
+
+void
+bplist_close(bplist_t *bpl)
+{
+ mutex_enter(&bpl->bpl_lock);
+
+ ASSERT(bpl->bpl_queue == NULL);
+
+ if (bpl->bpl_cached_dbuf) {
+ dmu_buf_rele(bpl->bpl_cached_dbuf, bpl);
+ bpl->bpl_cached_dbuf = NULL;
+ }
+ if (bpl->bpl_dbuf) {
+ dmu_buf_rele(bpl->bpl_dbuf, bpl);
+ bpl->bpl_dbuf = NULL;
+ bpl->bpl_phys = NULL;
+ }
+
+ mutex_exit(&bpl->bpl_lock);
+}
+
+boolean_t
+bplist_empty(bplist_t *bpl)
+{
+ boolean_t rv;
+
+ if (bpl->bpl_object == 0)
+ return (B_TRUE);
+
+ mutex_enter(&bpl->bpl_lock);
+ VERIFY(0 == bplist_hold(bpl)); /* XXX */
+ rv = (bpl->bpl_phys->bpl_entries == 0);
+ mutex_exit(&bpl->bpl_lock);
+
+ return (rv);
+}
+
+static int
+bplist_cache(bplist_t *bpl, uint64_t blkid)
+{
+ int err = 0;
+
+ if (bpl->bpl_cached_dbuf == NULL ||
+ bpl->bpl_cached_dbuf->db_offset != (blkid << bpl->bpl_blockshift)) {
+ if (bpl->bpl_cached_dbuf != NULL)
+ dmu_buf_rele(bpl->bpl_cached_dbuf, bpl);
+ err = dmu_buf_hold(bpl->bpl_mos,
+ bpl->bpl_object, blkid << bpl->bpl_blockshift,
+ bpl, &bpl->bpl_cached_dbuf);
+ ASSERT(err || bpl->bpl_cached_dbuf->db_size ==
+ 1ULL << bpl->bpl_blockshift);
+ }
+ return (err);
+}
+
+int
+bplist_iterate(bplist_t *bpl, uint64_t *itorp, blkptr_t *bp)
+{
+ uint64_t blk, off;
+ blkptr_t *bparray;
+ int err;
+
+ mutex_enter(&bpl->bpl_lock);
+
+ err = bplist_hold(bpl);
+ if (err) {
+ mutex_exit(&bpl->bpl_lock);
+ return (err);
+ }
+
+ if (*itorp >= bpl->bpl_phys->bpl_entries) {
+ mutex_exit(&bpl->bpl_lock);
+ return (ENOENT);
+ }
+
+ blk = *itorp >> bpl->bpl_bpshift;
+ off = P2PHASE(*itorp, 1ULL << bpl->bpl_bpshift);
+
+ err = bplist_cache(bpl, blk);
+ if (err) {
+ mutex_exit(&bpl->bpl_lock);
+ return (err);
+ }
+
+ bparray = bpl->bpl_cached_dbuf->db_data;
+ *bp = bparray[off];
+ (*itorp)++;
+ mutex_exit(&bpl->bpl_lock);
+ return (0);
+}
+
+int
+bplist_enqueue(bplist_t *bpl, const blkptr_t *bp, dmu_tx_t *tx)
+{
+ uint64_t blk, off;
+ blkptr_t *bparray;
+ int err;
+
+ ASSERT(!BP_IS_HOLE(bp));
+ mutex_enter(&bpl->bpl_lock);
+ err = bplist_hold(bpl);
+ if (err)
+ return (err);
+
+ blk = bpl->bpl_phys->bpl_entries >> bpl->bpl_bpshift;
+ off = P2PHASE(bpl->bpl_phys->bpl_entries, 1ULL << bpl->bpl_bpshift);
+
+ err = bplist_cache(bpl, blk);
+ if (err) {
+ mutex_exit(&bpl->bpl_lock);
+ return (err);
+ }
+
+ dmu_buf_will_dirty(bpl->bpl_cached_dbuf, tx);
+ bparray = bpl->bpl_cached_dbuf->db_data;
+ bparray[off] = *bp;
+
+ /* We never need the fill count. */
+ bparray[off].blk_fill = 0;
+
+ /* The bplist will compress better if we can leave off the checksum */
+ bzero(&bparray[off].blk_cksum, sizeof (bparray[off].blk_cksum));
+
+ dmu_buf_will_dirty(bpl->bpl_dbuf, tx);
+ bpl->bpl_phys->bpl_entries++;
+ bpl->bpl_phys->bpl_bytes +=
+ bp_get_dasize(dmu_objset_spa(bpl->bpl_mos), bp);
+ if (bpl->bpl_havecomp) {
+ bpl->bpl_phys->bpl_comp += BP_GET_PSIZE(bp);
+ bpl->bpl_phys->bpl_uncomp += BP_GET_UCSIZE(bp);
+ }
+ mutex_exit(&bpl->bpl_lock);
+
+ return (0);
+}
+
+/*
+ * Deferred entry; will be written later by bplist_sync().
+ */
+void
+bplist_enqueue_deferred(bplist_t *bpl, const blkptr_t *bp)
+{
+ bplist_q_t *bpq = kmem_alloc(sizeof (*bpq), KM_SLEEP);
+
+ ASSERT(!BP_IS_HOLE(bp));
+ mutex_enter(&bpl->bpl_lock);
+ bpq->bpq_blk = *bp;
+ bpq->bpq_next = bpl->bpl_queue;
+ bpl->bpl_queue = bpq;
+ mutex_exit(&bpl->bpl_lock);
+}
+
+void
+bplist_sync(bplist_t *bpl, dmu_tx_t *tx)
+{
+ bplist_q_t *bpq;
+
+ mutex_enter(&bpl->bpl_lock);
+ while ((bpq = bpl->bpl_queue) != NULL) {
+ bpl->bpl_queue = bpq->bpq_next;
+ mutex_exit(&bpl->bpl_lock);
+ VERIFY(0 == bplist_enqueue(bpl, &bpq->bpq_blk, tx));
+ kmem_free(bpq, sizeof (*bpq));
+ mutex_enter(&bpl->bpl_lock);
+ }
+ mutex_exit(&bpl->bpl_lock);
+}
+
+void
+bplist_vacate(bplist_t *bpl, dmu_tx_t *tx)
+{
+ mutex_enter(&bpl->bpl_lock);
+ ASSERT3P(bpl->bpl_queue, ==, NULL);
+ VERIFY(0 == bplist_hold(bpl));
+ dmu_buf_will_dirty(bpl->bpl_dbuf, tx);
+ VERIFY(0 == dmu_free_range(bpl->bpl_mos,
+ bpl->bpl_object, 0, -1ULL, tx));
+ bpl->bpl_phys->bpl_entries = 0;
+ bpl->bpl_phys->bpl_bytes = 0;
+ if (bpl->bpl_havecomp) {
+ bpl->bpl_phys->bpl_comp = 0;
+ bpl->bpl_phys->bpl_uncomp = 0;
+ }
+ mutex_exit(&bpl->bpl_lock);
+}
+
+int
+bplist_space(bplist_t *bpl, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
+{
+ int err;
+
+ mutex_enter(&bpl->bpl_lock);
+
+ err = bplist_hold(bpl);
+ if (err) {
+ mutex_exit(&bpl->bpl_lock);
+ return (err);
+ }
+
+ *usedp = bpl->bpl_phys->bpl_bytes;
+ if (bpl->bpl_havecomp) {
+ *compp = bpl->bpl_phys->bpl_comp;
+ *uncompp = bpl->bpl_phys->bpl_uncomp;
+ }
+ mutex_exit(&bpl->bpl_lock);
+
+ if (!bpl->bpl_havecomp) {
+ uint64_t itor = 0, comp = 0, uncomp = 0;
+ blkptr_t bp;
+
+ while ((err = bplist_iterate(bpl, &itor, &bp)) == 0) {
+ comp += BP_GET_PSIZE(&bp);
+ uncomp += BP_GET_UCSIZE(&bp);
+ }
+ if (err == ENOENT)
+ err = 0;
+ *compp = comp;
+ *uncompp = uncomp;
+ }
+
+ return (err);
+}
+
+/*
+ * Return (in *dasizep) the amount of space on the deadlist which is:
+ * mintxg < blk_birth <= maxtxg
+ */
+int
+bplist_space_birthrange(bplist_t *bpl, uint64_t mintxg, uint64_t maxtxg,
+ uint64_t *dasizep)
+{
+ uint64_t size = 0;
+ uint64_t itor = 0;
+ blkptr_t bp;
+ int err;
+
+ /*
+ * As an optimization, if they want the whole txg range, just
+ * get bpl_bytes rather than iterating over the bps.
+ */
+ if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX) {
+ mutex_enter(&bpl->bpl_lock);
+ err = bplist_hold(bpl);
+ if (err == 0)
+ *dasizep = bpl->bpl_phys->bpl_bytes;
+ mutex_exit(&bpl->bpl_lock);
+ return (err);
+ }
+
+ while ((err = bplist_iterate(bpl, &itor, &bp)) == 0) {
+ if (bp.blk_birth > mintxg && bp.blk_birth <= maxtxg) {
+ size +=
+ bp_get_dasize(dmu_objset_spa(bpl->bpl_mos), &bp);
+ }
+ }
+ if (err == ENOENT)
+ err = 0;
+ *dasizep = size;
+ return (err);
+}
diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c
new file mode 100644
index 000000000..d04610317
--- /dev/null
+++ b/module/zfs/dbuf.c
@@ -0,0 +1,2356 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/dmu.h>
+#include <sys/dmu_impl.h>
+#include <sys/dbuf.h>
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dmu_tx.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu_zfetch.h>
+
+static void dbuf_destroy(dmu_buf_impl_t *db);
+static int dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
+static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx);
+static arc_done_func_t dbuf_write_ready;
+static arc_done_func_t dbuf_write_done;
+static zio_done_func_t dbuf_skip_write_ready;
+static zio_done_func_t dbuf_skip_write_done;
+
+/*
+ * Global data structures and functions for the dbuf cache.
+ */
+static kmem_cache_t *dbuf_cache;
+
+/* ARGSUSED */
+static int
+dbuf_cons(void *vdb, void *unused, int kmflag)
+{
+ dmu_buf_impl_t *db = vdb;
+ bzero(db, sizeof (dmu_buf_impl_t));
+
+ mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL);
+ refcount_create(&db->db_holds);
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dbuf_dest(void *vdb, void *unused)
+{
+ dmu_buf_impl_t *db = vdb;
+ mutex_destroy(&db->db_mtx);
+ cv_destroy(&db->db_changed);
+ refcount_destroy(&db->db_holds);
+}
+
+/*
+ * dbuf hash table routines
+ */
+static dbuf_hash_table_t dbuf_hash_table;
+
+static uint64_t dbuf_hash_count;
+
+static uint64_t
+dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid)
+{
+ uintptr_t osv = (uintptr_t)os;
+ uint64_t crc = -1ULL;
+
+ ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (lvl)) & 0xFF];
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF];
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF];
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF];
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (blkid >> 0)) & 0xFF];
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (blkid >> 8)) & 0xFF];
+
+ crc ^= (osv>>14) ^ (obj>>16) ^ (blkid>>16);
+
+ return (crc);
+}
+
+#define DBUF_HASH(os, obj, level, blkid) dbuf_hash(os, obj, level, blkid);
+
+#define DBUF_EQUAL(dbuf, os, obj, level, blkid) \
+ ((dbuf)->db.db_object == (obj) && \
+ (dbuf)->db_objset == (os) && \
+ (dbuf)->db_level == (level) && \
+ (dbuf)->db_blkid == (blkid))
+
+dmu_buf_impl_t *
+dbuf_find(dnode_t *dn, uint8_t level, uint64_t blkid)
+{
+ dbuf_hash_table_t *h = &dbuf_hash_table;
+ objset_impl_t *os = dn->dn_objset;
+ uint64_t obj = dn->dn_object;
+ uint64_t hv = DBUF_HASH(os, obj, level, blkid);
+ uint64_t idx = hv & h->hash_table_mask;
+ dmu_buf_impl_t *db;
+
+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
+ for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) {
+ if (DBUF_EQUAL(db, os, obj, level, blkid)) {
+ mutex_enter(&db->db_mtx);
+ if (db->db_state != DB_EVICTING) {
+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
+ return (db);
+ }
+ mutex_exit(&db->db_mtx);
+ }
+ }
+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
+ return (NULL);
+}
+
+/*
+ * Insert an entry into the hash table. If there is already an element
+ * equal to elem in the hash table, then the already existing element
+ * will be returned and the new element will not be inserted.
+ * Otherwise returns NULL.
+ */
+static dmu_buf_impl_t *
+dbuf_hash_insert(dmu_buf_impl_t *db)
+{
+ dbuf_hash_table_t *h = &dbuf_hash_table;
+ objset_impl_t *os = db->db_objset;
+ uint64_t obj = db->db.db_object;
+ int level = db->db_level;
+ uint64_t blkid = db->db_blkid;
+ uint64_t hv = DBUF_HASH(os, obj, level, blkid);
+ uint64_t idx = hv & h->hash_table_mask;
+ dmu_buf_impl_t *dbf;
+
+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
+ for (dbf = h->hash_table[idx]; dbf != NULL; dbf = dbf->db_hash_next) {
+ if (DBUF_EQUAL(dbf, os, obj, level, blkid)) {
+ mutex_enter(&dbf->db_mtx);
+ if (dbf->db_state != DB_EVICTING) {
+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
+ return (dbf);
+ }
+ mutex_exit(&dbf->db_mtx);
+ }
+ }
+
+ mutex_enter(&db->db_mtx);
+ db->db_hash_next = h->hash_table[idx];
+ h->hash_table[idx] = db;
+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
+ atomic_add_64(&dbuf_hash_count, 1);
+
+ return (NULL);
+}
+
+/*
+ * Remove an entry from the hash table. This operation will
+ * fail if there are any existing holds on the db.
+ */
+static void
+dbuf_hash_remove(dmu_buf_impl_t *db)
+{
+ dbuf_hash_table_t *h = &dbuf_hash_table;
+ uint64_t hv = DBUF_HASH(db->db_objset, db->db.db_object,
+ db->db_level, db->db_blkid);
+ uint64_t idx = hv & h->hash_table_mask;
+ dmu_buf_impl_t *dbf, **dbp;
+
+ /*
+ * We musn't hold db_mtx to maintin lock ordering:
+ * DBUF_HASH_MUTEX > db_mtx.
+ */
+ ASSERT(refcount_is_zero(&db->db_holds));
+ ASSERT(db->db_state == DB_EVICTING);
+ ASSERT(!MUTEX_HELD(&db->db_mtx));
+
+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
+ dbp = &h->hash_table[idx];
+ while ((dbf = *dbp) != db) {
+ dbp = &dbf->db_hash_next;
+ ASSERT(dbf != NULL);
+ }
+ *dbp = db->db_hash_next;
+ db->db_hash_next = NULL;
+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
+ atomic_add_64(&dbuf_hash_count, -1);
+}
+
+static arc_evict_func_t dbuf_do_evict;
+
+static void
+dbuf_evict_user(dmu_buf_impl_t *db)
+{
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+
+ if (db->db_level != 0 || db->db_evict_func == NULL)
+ return;
+
+ if (db->db_user_data_ptr_ptr)
+ *db->db_user_data_ptr_ptr = db->db.db_data;
+ db->db_evict_func(&db->db, db->db_user_ptr);
+ db->db_user_ptr = NULL;
+ db->db_user_data_ptr_ptr = NULL;
+ db->db_evict_func = NULL;
+}
+
+void
+dbuf_evict(dmu_buf_impl_t *db)
+{
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(db->db_buf == NULL);
+ ASSERT(db->db_data_pending == NULL);
+
+ dbuf_clear(db);
+ dbuf_destroy(db);
+}
+
+void
+dbuf_init(void)
+{
+ uint64_t hsize = 1ULL << 16;
+ dbuf_hash_table_t *h = &dbuf_hash_table;
+ int i;
+
+ /*
+ * The hash table is big enough to fill all of physical memory
+ * with an average 4K block size. The table will take up
+ * totalmem*sizeof(void*)/4K (i.e. 2MB/GB with 8-byte pointers).
+ */
+ while (hsize * 4096 < physmem * PAGESIZE)
+ hsize <<= 1;
+
+retry:
+ h->hash_table_mask = hsize - 1;
+ h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP);
+ if (h->hash_table == NULL) {
+ /* XXX - we should really return an error instead of assert */
+ ASSERT(hsize > (1ULL << 10));
+ hsize >>= 1;
+ goto retry;
+ }
+
+ dbuf_cache = kmem_cache_create("dmu_buf_impl_t",
+ sizeof (dmu_buf_impl_t),
+ 0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0);
+
+ for (i = 0; i < DBUF_MUTEXES; i++)
+ mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL);
+}
+
+void
+dbuf_fini(void)
+{
+ dbuf_hash_table_t *h = &dbuf_hash_table;
+ int i;
+
+ for (i = 0; i < DBUF_MUTEXES; i++)
+ mutex_destroy(&h->hash_mutexes[i]);
+ kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
+ kmem_cache_destroy(dbuf_cache);
+}
+
+/*
+ * Other stuff.
+ */
+
+#ifdef ZFS_DEBUG
+static void
+dbuf_verify(dmu_buf_impl_t *db)
+{
+ dnode_t *dn = db->db_dnode;
+
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+
+ if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY))
+ return;
+
+ ASSERT(db->db_objset != NULL);
+ if (dn == NULL) {
+ ASSERT(db->db_parent == NULL);
+ ASSERT(db->db_blkptr == NULL);
+ } else {
+ ASSERT3U(db->db.db_object, ==, dn->dn_object);
+ ASSERT3P(db->db_objset, ==, dn->dn_objset);
+ ASSERT3U(db->db_level, <, dn->dn_nlevels);
+ ASSERT(db->db_blkid == DB_BONUS_BLKID ||
+ list_head(&dn->dn_dbufs));
+ }
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ ASSERT(dn != NULL);
+ ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
+ ASSERT3U(db->db.db_offset, ==, DB_BONUS_BLKID);
+ } else {
+ ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size);
+ }
+
+ /*
+ * We can't assert that db_size matches dn_datablksz because it
+ * can be momentarily different when another thread is doing
+ * dnode_set_blksz().
+ */
+ if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) {
+ dbuf_dirty_record_t *dr = db->db_data_pending;
+ /*
+ * It should only be modified in syncing context, so
+ * make sure we only have one copy of the data.
+ */
+ ASSERT(dr == NULL || dr->dt.dl.dr_data == db->db_buf);
+ }
+
+ /* verify db->db_blkptr */
+ if (db->db_blkptr) {
+ if (db->db_parent == dn->dn_dbuf) {
+ /* db is pointed to by the dnode */
+ /* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */
+ if (db->db.db_object == DMU_META_DNODE_OBJECT)
+ ASSERT(db->db_parent == NULL);
+ else
+ ASSERT(db->db_parent != NULL);
+ ASSERT3P(db->db_blkptr, ==,
+ &dn->dn_phys->dn_blkptr[db->db_blkid]);
+ } else {
+ /* db is pointed to by an indirect block */
+ int epb = db->db_parent->db.db_size >> SPA_BLKPTRSHIFT;
+ ASSERT3U(db->db_parent->db_level, ==, db->db_level+1);
+ ASSERT3U(db->db_parent->db.db_object, ==,
+ db->db.db_object);
+ /*
+ * dnode_grow_indblksz() can make this fail if we don't
+ * have the struct_rwlock. XXX indblksz no longer
+ * grows. safe to do this now?
+ */
+ if (RW_WRITE_HELD(&db->db_dnode->dn_struct_rwlock)) {
+ ASSERT3P(db->db_blkptr, ==,
+ ((blkptr_t *)db->db_parent->db.db_data +
+ db->db_blkid % epb));
+ }
+ }
+ }
+ if ((db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr)) &&
+ db->db.db_data && db->db_blkid != DB_BONUS_BLKID &&
+ db->db_state != DB_FILL && !dn->dn_free_txg) {
+ /*
+ * If the blkptr isn't set but they have nonzero data,
+ * it had better be dirty, otherwise we'll lose that
+ * data when we evict this buffer.
+ */
+ if (db->db_dirtycnt == 0) {
+ uint64_t *buf = db->db.db_data;
+ int i;
+
+ for (i = 0; i < db->db.db_size >> 3; i++) {
+ ASSERT(buf[i] == 0);
+ }
+ }
+ }
+}
+#endif
+
+static void
+dbuf_update_data(dmu_buf_impl_t *db)
+{
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ if (db->db_level == 0 && db->db_user_data_ptr_ptr) {
+ ASSERT(!refcount_is_zero(&db->db_holds));
+ *db->db_user_data_ptr_ptr = db->db.db_data;
+ }
+}
+
+static void
+dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf)
+{
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(db->db_buf == NULL || !arc_has_callback(db->db_buf));
+ db->db_buf = buf;
+ if (buf != NULL) {
+ ASSERT(buf->b_data != NULL);
+ db->db.db_data = buf->b_data;
+ if (!arc_released(buf))
+ arc_set_callback(buf, dbuf_do_evict, db);
+ dbuf_update_data(db);
+ } else {
+ dbuf_evict_user(db);
+ db->db.db_data = NULL;
+ if (db->db_state != DB_NOFILL)
+ db->db_state = DB_UNCACHED;
+ }
+}
+
+uint64_t
+dbuf_whichblock(dnode_t *dn, uint64_t offset)
+{
+ if (dn->dn_datablkshift) {
+ return (offset >> dn->dn_datablkshift);
+ } else {
+ ASSERT3U(offset, <, dn->dn_datablksz);
+ return (0);
+ }
+}
+
+static void
+dbuf_read_done(zio_t *zio, arc_buf_t *buf, void *vdb)
+{
+ dmu_buf_impl_t *db = vdb;
+
+ mutex_enter(&db->db_mtx);
+ ASSERT3U(db->db_state, ==, DB_READ);
+ /*
+ * All reads are synchronous, so we must have a hold on the dbuf
+ */
+ ASSERT(refcount_count(&db->db_holds) > 0);
+ ASSERT(db->db_buf == NULL);
+ ASSERT(db->db.db_data == NULL);
+ if (db->db_level == 0 && db->db_freed_in_flight) {
+ /* we were freed in flight; disregard any error */
+ arc_release(buf, db);
+ bzero(buf->b_data, db->db.db_size);
+ arc_buf_freeze(buf);
+ db->db_freed_in_flight = FALSE;
+ dbuf_set_data(db, buf);
+ db->db_state = DB_CACHED;
+ } else if (zio == NULL || zio->io_error == 0) {
+ dbuf_set_data(db, buf);
+ db->db_state = DB_CACHED;
+ } else {
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+ ASSERT3P(db->db_buf, ==, NULL);
+ VERIFY(arc_buf_remove_ref(buf, db) == 1);
+ db->db_state = DB_UNCACHED;
+ }
+ cv_broadcast(&db->db_changed);
+ mutex_exit(&db->db_mtx);
+ dbuf_rele(db, NULL);
+}
+
+static void
+dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t *flags)
+{
+ dnode_t *dn = db->db_dnode;
+ zbookmark_t zb;
+ uint32_t aflags = ARC_NOWAIT;
+ arc_buf_t *pbuf;
+
+ ASSERT(!refcount_is_zero(&db->db_holds));
+ /* We need the struct_rwlock to prevent db_blkptr from changing. */
+ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(db->db_state == DB_UNCACHED);
+ ASSERT(db->db_buf == NULL);
+
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ int bonuslen = dn->dn_bonuslen;
+
+ ASSERT3U(bonuslen, <=, db->db.db_size);
+ db->db.db_data = zio_buf_alloc(DN_MAX_BONUSLEN);
+ arc_space_consume(DN_MAX_BONUSLEN);
+ if (bonuslen < DN_MAX_BONUSLEN)
+ bzero(db->db.db_data, DN_MAX_BONUSLEN);
+ bcopy(DN_BONUS(dn->dn_phys), db->db.db_data,
+ bonuslen);
+ dbuf_update_data(db);
+ db->db_state = DB_CACHED;
+ mutex_exit(&db->db_mtx);
+ return;
+ }
+
+ /*
+ * Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync()
+ * processes the delete record and clears the bp while we are waiting
+ * for the dn_mtx (resulting in a "no" from block_freed).
+ */
+ if (db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr) ||
+ (db->db_level == 0 && (dnode_block_freed(dn, db->db_blkid) ||
+ BP_IS_HOLE(db->db_blkptr)))) {
+ arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+
+ dbuf_set_data(db, arc_buf_alloc(dn->dn_objset->os_spa,
+ db->db.db_size, db, type));
+ bzero(db->db.db_data, db->db.db_size);
+ db->db_state = DB_CACHED;
+ *flags |= DB_RF_CACHED;
+ mutex_exit(&db->db_mtx);
+ return;
+ }
+
+ db->db_state = DB_READ;
+ mutex_exit(&db->db_mtx);
+
+ if (DBUF_IS_L2CACHEABLE(db))
+ aflags |= ARC_L2CACHE;
+
+ zb.zb_objset = db->db_objset->os_dsl_dataset ?
+ db->db_objset->os_dsl_dataset->ds_object : 0;
+ zb.zb_object = db->db.db_object;
+ zb.zb_level = db->db_level;
+ zb.zb_blkid = db->db_blkid;
+
+ dbuf_add_ref(db, NULL);
+ /* ZIO_FLAG_CANFAIL callers have to check the parent zio's error */
+
+ if (db->db_parent)
+ pbuf = db->db_parent->db_buf;
+ else
+ pbuf = db->db_objset->os_phys_buf;
+
+ (void) arc_read(zio, dn->dn_objset->os_spa, db->db_blkptr, pbuf,
+ dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ,
+ (*flags & DB_RF_CANFAIL) ? ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED,
+ &aflags, &zb);
+ if (aflags & ARC_CACHED)
+ *flags |= DB_RF_CACHED;
+}
+
+int
+dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags)
+{
+ int err = 0;
+ int havepzio = (zio != NULL);
+ int prefetch;
+
+ /*
+ * We don't have to hold the mutex to check db_state because it
+ * can't be freed while we have a hold on the buffer.
+ */
+ ASSERT(!refcount_is_zero(&db->db_holds));
+
+ if (db->db_state == DB_NOFILL)
+ return (EIO);
+
+ if ((flags & DB_RF_HAVESTRUCT) == 0)
+ rw_enter(&db->db_dnode->dn_struct_rwlock, RW_READER);
+
+ prefetch = db->db_level == 0 && db->db_blkid != DB_BONUS_BLKID &&
+ (flags & DB_RF_NOPREFETCH) == 0 && db->db_dnode != NULL &&
+ DBUF_IS_CACHEABLE(db);
+
+ mutex_enter(&db->db_mtx);
+ if (db->db_state == DB_CACHED) {
+ mutex_exit(&db->db_mtx);
+ if (prefetch)
+ dmu_zfetch(&db->db_dnode->dn_zfetch, db->db.db_offset,
+ db->db.db_size, TRUE);
+ if ((flags & DB_RF_HAVESTRUCT) == 0)
+ rw_exit(&db->db_dnode->dn_struct_rwlock);
+ } else if (db->db_state == DB_UNCACHED) {
+ if (zio == NULL) {
+ zio = zio_root(db->db_dnode->dn_objset->os_spa,
+ NULL, NULL, ZIO_FLAG_CANFAIL);
+ }
+ dbuf_read_impl(db, zio, &flags);
+
+ /* dbuf_read_impl has dropped db_mtx for us */
+
+ if (prefetch)
+ dmu_zfetch(&db->db_dnode->dn_zfetch, db->db.db_offset,
+ db->db.db_size, flags & DB_RF_CACHED);
+
+ if ((flags & DB_RF_HAVESTRUCT) == 0)
+ rw_exit(&db->db_dnode->dn_struct_rwlock);
+
+ if (!havepzio)
+ err = zio_wait(zio);
+ } else {
+ mutex_exit(&db->db_mtx);
+ if (prefetch)
+ dmu_zfetch(&db->db_dnode->dn_zfetch, db->db.db_offset,
+ db->db.db_size, TRUE);
+ if ((flags & DB_RF_HAVESTRUCT) == 0)
+ rw_exit(&db->db_dnode->dn_struct_rwlock);
+
+ mutex_enter(&db->db_mtx);
+ if ((flags & DB_RF_NEVERWAIT) == 0) {
+ while (db->db_state == DB_READ ||
+ db->db_state == DB_FILL) {
+ ASSERT(db->db_state == DB_READ ||
+ (flags & DB_RF_HAVESTRUCT) == 0);
+ cv_wait(&db->db_changed, &db->db_mtx);
+ }
+ if (db->db_state == DB_UNCACHED)
+ err = EIO;
+ }
+ mutex_exit(&db->db_mtx);
+ }
+
+ ASSERT(err || havepzio || db->db_state == DB_CACHED);
+ return (err);
+}
+
+static void
+dbuf_noread(dmu_buf_impl_t *db)
+{
+ ASSERT(!refcount_is_zero(&db->db_holds));
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+ mutex_enter(&db->db_mtx);
+ while (db->db_state == DB_READ || db->db_state == DB_FILL)
+ cv_wait(&db->db_changed, &db->db_mtx);
+ if (db->db_state == DB_UNCACHED) {
+ arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+
+ ASSERT(db->db_buf == NULL);
+ ASSERT(db->db.db_data == NULL);
+ dbuf_set_data(db, arc_buf_alloc(db->db_dnode->dn_objset->os_spa,
+ db->db.db_size, db, type));
+ db->db_state = DB_FILL;
+ } else if (db->db_state == DB_NOFILL) {
+ dbuf_set_data(db, NULL);
+ } else {
+ ASSERT3U(db->db_state, ==, DB_CACHED);
+ }
+ mutex_exit(&db->db_mtx);
+}
+
+/*
+ * This is our just-in-time copy function. It makes a copy of
+ * buffers, that have been modified in a previous transaction
+ * group, before we modify them in the current active group.
+ *
+ * This function is used in two places: when we are dirtying a
+ * buffer for the first time in a txg, and when we are freeing
+ * a range in a dnode that includes this buffer.
+ *
+ * Note that when we are called from dbuf_free_range() we do
+ * not put a hold on the buffer, we just traverse the active
+ * dbuf list for the dnode.
+ */
+static void
+dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg)
+{
+ dbuf_dirty_record_t *dr = db->db_last_dirty;
+
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(db->db.db_data != NULL);
+ ASSERT(db->db_level == 0);
+ ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT);
+
+ if (dr == NULL ||
+ (dr->dt.dl.dr_data !=
+ ((db->db_blkid == DB_BONUS_BLKID) ? db->db.db_data : db->db_buf)))
+ return;
+
+ /*
+ * If the last dirty record for this dbuf has not yet synced
+ * and its referencing the dbuf data, either:
+ * reset the reference to point to a new copy,
+ * or (if there a no active holders)
+ * just null out the current db_data pointer.
+ */
+ ASSERT(dr->dr_txg >= txg - 2);
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ /* Note that the data bufs here are zio_bufs */
+ dr->dt.dl.dr_data = zio_buf_alloc(DN_MAX_BONUSLEN);
+ arc_space_consume(DN_MAX_BONUSLEN);
+ bcopy(db->db.db_data, dr->dt.dl.dr_data, DN_MAX_BONUSLEN);
+ } else if (refcount_count(&db->db_holds) > db->db_dirtycnt) {
+ int size = db->db.db_size;
+ arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+ dr->dt.dl.dr_data = arc_buf_alloc(
+ db->db_dnode->dn_objset->os_spa, size, db, type);
+ bcopy(db->db.db_data, dr->dt.dl.dr_data->b_data, size);
+ } else {
+ dbuf_set_data(db, NULL);
+ }
+}
+
+void
+dbuf_unoverride(dbuf_dirty_record_t *dr)
+{
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ uint64_t txg = dr->dr_txg;
+
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC);
+ ASSERT(db->db_level == 0);
+
+ if (db->db_blkid == DB_BONUS_BLKID ||
+ dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN)
+ return;
+
+ /* free this block */
+ if (!BP_IS_HOLE(&dr->dt.dl.dr_overridden_by)) {
+ /* XXX can get silent EIO here */
+ (void) dsl_free(NULL,
+ spa_get_dsl(db->db_dnode->dn_objset->os_spa),
+ txg, &dr->dt.dl.dr_overridden_by, NULL, NULL, ARC_WAIT);
+ }
+ dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
+ /*
+ * Release the already-written buffer, so we leave it in
+ * a consistent dirty state. Note that all callers are
+ * modifying the buffer, so they will immediately do
+ * another (redundant) arc_release(). Therefore, leave
+ * the buf thawed to save the effort of freezing &
+ * immediately re-thawing it.
+ */
+ arc_release(dr->dt.dl.dr_data, db);
+}
+
+/*
+ * Evict (if its unreferenced) or clear (if its referenced) any level-0
+ * data blocks in the free range, so that any future readers will find
+ * empty blocks. Also, if we happen accross any level-1 dbufs in the
+ * range that have not already been marked dirty, mark them dirty so
+ * they stay in memory.
+ */
+void
+dbuf_free_range(dnode_t *dn, uint64_t start, uint64_t end, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db, *db_next;
+ uint64_t txg = tx->tx_txg;
+ int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ uint64_t first_l1 = start >> epbs;
+ uint64_t last_l1 = end >> epbs;
+
+ if (end > dn->dn_maxblkid) {
+ end = dn->dn_maxblkid;
+ last_l1 = end >> epbs;
+ }
+ dprintf_dnode(dn, "start=%llu end=%llu\n", start, end);
+ mutex_enter(&dn->dn_dbufs_mtx);
+ for (db = list_head(&dn->dn_dbufs); db; db = db_next) {
+ db_next = list_next(&dn->dn_dbufs, db);
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+
+ if (db->db_level == 1 &&
+ db->db_blkid >= first_l1 && db->db_blkid <= last_l1) {
+ mutex_enter(&db->db_mtx);
+ if (db->db_last_dirty &&
+ db->db_last_dirty->dr_txg < txg) {
+ dbuf_add_ref(db, FTAG);
+ mutex_exit(&db->db_mtx);
+ dbuf_will_dirty(db, tx);
+ dbuf_rele(db, FTAG);
+ } else {
+ mutex_exit(&db->db_mtx);
+ }
+ }
+
+ if (db->db_level != 0)
+ continue;
+ dprintf_dbuf(db, "found buf %s\n", "");
+ if (db->db_blkid < start || db->db_blkid > end)
+ continue;
+
+ /* found a level 0 buffer in the range */
+ if (dbuf_undirty(db, tx))
+ continue;
+
+ mutex_enter(&db->db_mtx);
+ if (db->db_state == DB_UNCACHED ||
+ db->db_state == DB_NOFILL ||
+ db->db_state == DB_EVICTING) {
+ ASSERT(db->db.db_data == NULL);
+ mutex_exit(&db->db_mtx);
+ continue;
+ }
+ if (db->db_state == DB_READ || db->db_state == DB_FILL) {
+ /* will be handled in dbuf_read_done or dbuf_rele */
+ db->db_freed_in_flight = TRUE;
+ mutex_exit(&db->db_mtx);
+ continue;
+ }
+ if (refcount_count(&db->db_holds) == 0) {
+ ASSERT(db->db_buf);
+ dbuf_clear(db);
+ continue;
+ }
+ /* The dbuf is referenced */
+
+ if (db->db_last_dirty != NULL) {
+ dbuf_dirty_record_t *dr = db->db_last_dirty;
+
+ if (dr->dr_txg == txg) {
+ /*
+ * This buffer is "in-use", re-adjust the file
+ * size to reflect that this buffer may
+ * contain new data when we sync.
+ */
+ if (db->db_blkid > dn->dn_maxblkid)
+ dn->dn_maxblkid = db->db_blkid;
+ dbuf_unoverride(dr);
+ } else {
+ /*
+ * This dbuf is not dirty in the open context.
+ * Either uncache it (if its not referenced in
+ * the open context) or reset its contents to
+ * empty.
+ */
+ dbuf_fix_old_data(db, txg);
+ }
+ }
+ /* clear the contents if its cached */
+ if (db->db_state == DB_CACHED) {
+ ASSERT(db->db.db_data != NULL);
+ arc_release(db->db_buf, db);
+ bzero(db->db.db_data, db->db.db_size);
+ arc_buf_freeze(db->db_buf);
+ }
+
+ mutex_exit(&db->db_mtx);
+ }
+ mutex_exit(&dn->dn_dbufs_mtx);
+}
+
+static int
+dbuf_block_freeable(dmu_buf_impl_t *db)
+{
+ dsl_dataset_t *ds = db->db_objset->os_dsl_dataset;
+ uint64_t birth_txg = 0;
+
+ /*
+ * We don't need any locking to protect db_blkptr:
+ * If it's syncing, then db_last_dirty will be set
+ * so we'll ignore db_blkptr.
+ */
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ if (db->db_last_dirty)
+ birth_txg = db->db_last_dirty->dr_txg;
+ else if (db->db_blkptr)
+ birth_txg = db->db_blkptr->blk_birth;
+
+ /* If we don't exist or are in a snapshot, we can't be freed */
+ if (birth_txg)
+ return (ds == NULL ||
+ dsl_dataset_block_freeable(ds, birth_txg));
+ else
+ return (FALSE);
+}
+
+void
+dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx)
+{
+ arc_buf_t *buf, *obuf;
+ int osize = db->db.db_size;
+ arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+
+ /* XXX does *this* func really need the lock? */
+ ASSERT(RW_WRITE_HELD(&db->db_dnode->dn_struct_rwlock));
+
+ /*
+ * This call to dbuf_will_dirty() with the dn_struct_rwlock held
+ * is OK, because there can be no other references to the db
+ * when we are changing its size, so no concurrent DB_FILL can
+ * be happening.
+ */
+ /*
+ * XXX we should be doing a dbuf_read, checking the return
+ * value and returning that up to our callers
+ */
+ dbuf_will_dirty(db, tx);
+
+ /* create the data buffer for the new block */
+ buf = arc_buf_alloc(db->db_dnode->dn_objset->os_spa, size, db, type);
+
+ /* copy old block data to the new block */
+ obuf = db->db_buf;
+ bcopy(obuf->b_data, buf->b_data, MIN(osize, size));
+ /* zero the remainder */
+ if (size > osize)
+ bzero((uint8_t *)buf->b_data + osize, size - osize);
+
+ mutex_enter(&db->db_mtx);
+ dbuf_set_data(db, buf);
+ VERIFY(arc_buf_remove_ref(obuf, db) == 1);
+ db->db.db_size = size;
+
+ if (db->db_level == 0) {
+ ASSERT3U(db->db_last_dirty->dr_txg, ==, tx->tx_txg);
+ db->db_last_dirty->dt.dl.dr_data = buf;
+ }
+ mutex_exit(&db->db_mtx);
+
+ dnode_willuse_space(db->db_dnode, size-osize, tx);
+}
+
+dbuf_dirty_record_t *
+dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
+{
+ dnode_t *dn = db->db_dnode;
+ objset_impl_t *os = dn->dn_objset;
+ dbuf_dirty_record_t **drp, *dr;
+ int drop_struct_lock = FALSE;
+ boolean_t do_free_accounting = B_FALSE;
+ int txgoff = tx->tx_txg & TXG_MASK;
+
+ ASSERT(tx->tx_txg != 0);
+ ASSERT(!refcount_is_zero(&db->db_holds));
+ DMU_TX_DIRTY_BUF(tx, db);
+
+ /*
+ * Shouldn't dirty a regular buffer in syncing context. Private
+ * objects may be dirtied in syncing context, but only if they
+ * were already pre-dirtied in open context.
+ * XXX We may want to prohibit dirtying in syncing context even
+ * if they did pre-dirty.
+ */
+ ASSERT(!dmu_tx_is_syncing(tx) ||
+ BP_IS_HOLE(dn->dn_objset->os_rootbp) ||
+ dn->dn_object == DMU_META_DNODE_OBJECT ||
+ dn->dn_objset->os_dsl_dataset == NULL ||
+ dsl_dir_is_private(dn->dn_objset->os_dsl_dataset->ds_dir));
+
+ /*
+ * We make this assert for private objects as well, but after we
+ * check if we're already dirty. They are allowed to re-dirty
+ * in syncing context.
+ */
+ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT ||
+ dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
+ (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));
+
+ mutex_enter(&db->db_mtx);
+ /*
+ * XXX make this true for indirects too? The problem is that
+ * transactions created with dmu_tx_create_assigned() from
+ * syncing context don't bother holding ahead.
+ */
+ ASSERT(db->db_level != 0 ||
+ db->db_state == DB_CACHED || db->db_state == DB_FILL ||
+ db->db_state == DB_NOFILL);
+
+ mutex_enter(&dn->dn_mtx);
+ /*
+ * Don't set dirtyctx to SYNC if we're just modifying this as we
+ * initialize the objset.
+ */
+ if (dn->dn_dirtyctx == DN_UNDIRTIED &&
+ !BP_IS_HOLE(dn->dn_objset->os_rootbp)) {
+ dn->dn_dirtyctx =
+ (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN);
+ ASSERT(dn->dn_dirtyctx_firstset == NULL);
+ dn->dn_dirtyctx_firstset = kmem_alloc(1, KM_SLEEP);
+ }
+ mutex_exit(&dn->dn_mtx);
+
+ /*
+ * If this buffer is already dirty, we're done.
+ */
+ drp = &db->db_last_dirty;
+ ASSERT(*drp == NULL || (*drp)->dr_txg <= tx->tx_txg ||
+ db->db.db_object == DMU_META_DNODE_OBJECT);
+ while ((dr = *drp) != NULL && dr->dr_txg > tx->tx_txg)
+ drp = &dr->dr_next;
+ if (dr && dr->dr_txg == tx->tx_txg) {
+ if (db->db_level == 0 && db->db_blkid != DB_BONUS_BLKID) {
+ /*
+ * If this buffer has already been written out,
+ * we now need to reset its state.
+ */
+ dbuf_unoverride(dr);
+ if (db->db.db_object != DMU_META_DNODE_OBJECT)
+ arc_buf_thaw(db->db_buf);
+ }
+ mutex_exit(&db->db_mtx);
+ return (dr);
+ }
+
+ /*
+ * Only valid if not already dirty.
+ */
+ ASSERT(dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
+ (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));
+
+ ASSERT3U(dn->dn_nlevels, >, db->db_level);
+ ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) ||
+ dn->dn_phys->dn_nlevels > db->db_level ||
+ dn->dn_next_nlevels[txgoff] > db->db_level ||
+ dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level ||
+ dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level);
+
+ /*
+ * We should only be dirtying in syncing context if it's the
+ * mos, a spa os, or we're initializing the os. However, we are
+ * allowed to dirty in syncing context provided we already
+ * dirtied it in open context. Hence we must make this
+ * assertion only if we're not already dirty.
+ */
+ ASSERT(!dmu_tx_is_syncing(tx) ||
+ os->os_dsl_dataset == NULL ||
+ !dsl_dir_is_private(os->os_dsl_dataset->ds_dir) ||
+ !BP_IS_HOLE(os->os_rootbp));
+ ASSERT(db->db.db_size != 0);
+
+ dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);
+
+ if (db->db_blkid != DB_BONUS_BLKID) {
+ /*
+ * Update the accounting.
+ * Note: we delay "free accounting" until after we drop
+ * the db_mtx. This keeps us from grabbing other locks
+ * (and possibly deadlocking) in bp_get_dasize() while
+ * also holding the db_mtx.
+ */
+ dnode_willuse_space(dn, db->db.db_size, tx);
+ do_free_accounting = dbuf_block_freeable(db);
+ }
+
+ /*
+ * If this buffer is dirty in an old transaction group we need
+ * to make a copy of it so that the changes we make in this
+ * transaction group won't leak out when we sync the older txg.
+ */
+ dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP);
+ if (db->db_level == 0) {
+ void *data_old = db->db_buf;
+
+ if (db->db_state != DB_NOFILL) {
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ dbuf_fix_old_data(db, tx->tx_txg);
+ data_old = db->db.db_data;
+ } else if (db->db.db_object != DMU_META_DNODE_OBJECT) {
+ /*
+ * Release the data buffer from the cache so
+ * that we can modify it without impacting
+ * possible other users of this cached data
+ * block. Note that indirect blocks and
+ * private objects are not released until the
+ * syncing state (since they are only modified
+ * then).
+ */
+ arc_release(db->db_buf, db);
+ dbuf_fix_old_data(db, tx->tx_txg);
+ data_old = db->db_buf;
+ }
+ ASSERT(data_old != NULL);
+ }
+ dr->dt.dl.dr_data = data_old;
+ } else {
+ mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&dr->dt.di.dr_children,
+ sizeof (dbuf_dirty_record_t),
+ offsetof(dbuf_dirty_record_t, dr_dirty_node));
+ }
+ dr->dr_dbuf = db;
+ dr->dr_txg = tx->tx_txg;
+ dr->dr_next = *drp;
+ *drp = dr;
+
+ /*
+ * We could have been freed_in_flight between the dbuf_noread
+ * and dbuf_dirty. We win, as though the dbuf_noread() had
+ * happened after the free.
+ */
+ if (db->db_level == 0 && db->db_blkid != DB_BONUS_BLKID) {
+ mutex_enter(&dn->dn_mtx);
+ dnode_clear_range(dn, db->db_blkid, 1, tx);
+ mutex_exit(&dn->dn_mtx);
+ db->db_freed_in_flight = FALSE;
+ }
+
+ /*
+ * This buffer is now part of this txg
+ */
+ dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg);
+ db->db_dirtycnt += 1;
+ ASSERT3U(db->db_dirtycnt, <=, 3);
+
+ mutex_exit(&db->db_mtx);
+
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ mutex_enter(&dn->dn_mtx);
+ ASSERT(!list_link_active(&dr->dr_dirty_node));
+ list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
+ mutex_exit(&dn->dn_mtx);
+ dnode_setdirty(dn, tx);
+ return (dr);
+ } else if (do_free_accounting) {
+ blkptr_t *bp = db->db_blkptr;
+ int64_t willfree = (bp && !BP_IS_HOLE(bp)) ?
+ bp_get_dasize(os->os_spa, bp) : db->db.db_size;
+ /*
+ * This is only a guess -- if the dbuf is dirty
+ * in a previous txg, we don't know how much
+ * space it will use on disk yet. We should
+ * really have the struct_rwlock to access
+ * db_blkptr, but since this is just a guess,
+ * it's OK if we get an odd answer.
+ */
+ dnode_willuse_space(dn, -willfree, tx);
+ }
+
+ if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) {
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ drop_struct_lock = TRUE;
+ }
+
+ if (db->db_level == 0) {
+ dnode_new_blkid(dn, db->db_blkid, tx, drop_struct_lock);
+ ASSERT(dn->dn_maxblkid >= db->db_blkid);
+ }
+
+ if (db->db_level+1 < dn->dn_nlevels) {
+ dmu_buf_impl_t *parent = db->db_parent;
+ dbuf_dirty_record_t *di;
+ int parent_held = FALSE;
+
+ if (db->db_parent == NULL || db->db_parent == dn->dn_dbuf) {
+ int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+
+ parent = dbuf_hold_level(dn, db->db_level+1,
+ db->db_blkid >> epbs, FTAG);
+ parent_held = TRUE;
+ }
+ if (drop_struct_lock)
+ rw_exit(&dn->dn_struct_rwlock);
+ ASSERT3U(db->db_level+1, ==, parent->db_level);
+ di = dbuf_dirty(parent, tx);
+ if (parent_held)
+ dbuf_rele(parent, FTAG);
+
+ mutex_enter(&db->db_mtx);
+ /* possible race with dbuf_undirty() */
+ if (db->db_last_dirty == dr ||
+ dn->dn_object == DMU_META_DNODE_OBJECT) {
+ mutex_enter(&di->dt.di.dr_mtx);
+ ASSERT3U(di->dr_txg, ==, tx->tx_txg);
+ ASSERT(!list_link_active(&dr->dr_dirty_node));
+ list_insert_tail(&di->dt.di.dr_children, dr);
+ mutex_exit(&di->dt.di.dr_mtx);
+ dr->dr_parent = di;
+ }
+ mutex_exit(&db->db_mtx);
+ } else {
+ ASSERT(db->db_level+1 == dn->dn_nlevels);
+ ASSERT(db->db_blkid < dn->dn_nblkptr);
+ ASSERT(db->db_parent == NULL ||
+ db->db_parent == db->db_dnode->dn_dbuf);
+ mutex_enter(&dn->dn_mtx);
+ ASSERT(!list_link_active(&dr->dr_dirty_node));
+ list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
+ mutex_exit(&dn->dn_mtx);
+ if (drop_struct_lock)
+ rw_exit(&dn->dn_struct_rwlock);
+ }
+
+ dnode_setdirty(dn, tx);
+ return (dr);
+}
+
+static int
+dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
+{
+ dnode_t *dn = db->db_dnode;
+ uint64_t txg = tx->tx_txg;
+ dbuf_dirty_record_t *dr, **drp;
+
+ ASSERT(txg != 0);
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+
+ mutex_enter(&db->db_mtx);
+
+ /*
+ * If this buffer is not dirty, we're done.
+ */
+ for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next)
+ if (dr->dr_txg <= txg)
+ break;
+ if (dr == NULL || dr->dr_txg < txg) {
+ mutex_exit(&db->db_mtx);
+ return (0);
+ }
+ ASSERT(dr->dr_txg == txg);
+
+ /*
+ * If this buffer is currently held, we cannot undirty
+ * it, since one of the current holders may be in the
+ * middle of an update. Note that users of dbuf_undirty()
+ * should not place a hold on the dbuf before the call.
+ */
+ if (refcount_count(&db->db_holds) > db->db_dirtycnt) {
+ mutex_exit(&db->db_mtx);
+ /* Make sure we don't toss this buffer at sync phase */
+ mutex_enter(&dn->dn_mtx);
+ dnode_clear_range(dn, db->db_blkid, 1, tx);
+ mutex_exit(&dn->dn_mtx);
+ return (0);
+ }
+
+ dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);
+
+ ASSERT(db->db.db_size != 0);
+
+ /* XXX would be nice to fix up dn_towrite_space[] */
+
+ *drp = dr->dr_next;
+
+ if (dr->dr_parent) {
+ mutex_enter(&dr->dr_parent->dt.di.dr_mtx);
+ list_remove(&dr->dr_parent->dt.di.dr_children, dr);
+ mutex_exit(&dr->dr_parent->dt.di.dr_mtx);
+ } else if (db->db_level+1 == dn->dn_nlevels) {
+ ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf);
+ mutex_enter(&dn->dn_mtx);
+ list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr);
+ mutex_exit(&dn->dn_mtx);
+ }
+
+ if (db->db_level == 0) {
+ if (db->db_state != DB_NOFILL) {
+ dbuf_unoverride(dr);
+
+ ASSERT(db->db_buf != NULL);
+ ASSERT(dr->dt.dl.dr_data != NULL);
+ if (dr->dt.dl.dr_data != db->db_buf)
+ VERIFY(arc_buf_remove_ref(dr->dt.dl.dr_data,
+ db) == 1);
+ }
+ } else {
+ ASSERT(db->db_buf != NULL);
+ ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
+ mutex_destroy(&dr->dt.di.dr_mtx);
+ list_destroy(&dr->dt.di.dr_children);
+ }
+ kmem_free(dr, sizeof (dbuf_dirty_record_t));
+
+ ASSERT(db->db_dirtycnt > 0);
+ db->db_dirtycnt -= 1;
+
+ if (refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) {
+ arc_buf_t *buf = db->db_buf;
+
+ ASSERT(arc_released(buf));
+ dbuf_set_data(db, NULL);
+ VERIFY(arc_buf_remove_ref(buf, db) == 1);
+ dbuf_evict(db);
+ return (1);
+ }
+
+ mutex_exit(&db->db_mtx);
+ return (0);
+}
+
+#pragma weak dmu_buf_will_dirty = dbuf_will_dirty
+void
+dbuf_will_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
+{
+ int rf = DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH;
+
+ ASSERT(tx->tx_txg != 0);
+ ASSERT(!refcount_is_zero(&db->db_holds));
+
+ if (RW_WRITE_HELD(&db->db_dnode->dn_struct_rwlock))
+ rf |= DB_RF_HAVESTRUCT;
+ (void) dbuf_read(db, NULL, rf);
+ (void) dbuf_dirty(db, tx);
+}
+
+void
+dmu_buf_will_not_fill(dmu_buf_t *db_fake, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+
+ db->db_state = DB_NOFILL;
+
+ dmu_buf_will_fill(db_fake, tx);
+}
+
+void
+dmu_buf_will_fill(dmu_buf_t *db_fake, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+ ASSERT(tx->tx_txg != 0);
+ ASSERT(db->db_level == 0);
+ ASSERT(!refcount_is_zero(&db->db_holds));
+
+ ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT ||
+ dmu_tx_private_ok(tx));
+
+ dbuf_noread(db);
+ (void) dbuf_dirty(db, tx);
+}
+
+#pragma weak dmu_buf_fill_done = dbuf_fill_done
+/* ARGSUSED */
+void
+dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx)
+{
+ mutex_enter(&db->db_mtx);
+ DBUF_VERIFY(db);
+
+ if (db->db_state == DB_FILL) {
+ if (db->db_level == 0 && db->db_freed_in_flight) {
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+ /* we were freed while filling */
+ /* XXX dbuf_undirty? */
+ bzero(db->db.db_data, db->db.db_size);
+ db->db_freed_in_flight = FALSE;
+ }
+ db->db_state = DB_CACHED;
+ cv_broadcast(&db->db_changed);
+ }
+ mutex_exit(&db->db_mtx);
+}
+
+/*
+ * "Clear" the contents of this dbuf. This will mark the dbuf
+ * EVICTING and clear *most* of its references. Unfortunetely,
+ * when we are not holding the dn_dbufs_mtx, we can't clear the
+ * entry in the dn_dbufs list. We have to wait until dbuf_destroy()
+ * in this case. For callers from the DMU we will usually see:
+ * dbuf_clear()->arc_buf_evict()->dbuf_do_evict()->dbuf_destroy()
+ * For the arc callback, we will usually see:
+ * dbuf_do_evict()->dbuf_clear();dbuf_destroy()
+ * Sometimes, though, we will get a mix of these two:
+ * DMU: dbuf_clear()->arc_buf_evict()
+ * ARC: dbuf_do_evict()->dbuf_destroy()
+ */
+void
+dbuf_clear(dmu_buf_impl_t *db)
+{
+ dnode_t *dn = db->db_dnode;
+ dmu_buf_impl_t *parent = db->db_parent;
+ dmu_buf_impl_t *dndb = dn->dn_dbuf;
+ int dbuf_gone = FALSE;
+
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+ ASSERT(refcount_is_zero(&db->db_holds));
+
+ dbuf_evict_user(db);
+
+ if (db->db_state == DB_CACHED) {
+ ASSERT(db->db.db_data != NULL);
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ zio_buf_free(db->db.db_data, DN_MAX_BONUSLEN);
+ arc_space_return(DN_MAX_BONUSLEN);
+ }
+ db->db.db_data = NULL;
+ db->db_state = DB_UNCACHED;
+ }
+
+ ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL);
+ ASSERT(db->db_data_pending == NULL);
+
+ db->db_state = DB_EVICTING;
+ db->db_blkptr = NULL;
+
+ if (db->db_blkid != DB_BONUS_BLKID && MUTEX_HELD(&dn->dn_dbufs_mtx)) {
+ list_remove(&dn->dn_dbufs, db);
+ dnode_rele(dn, db);
+ db->db_dnode = NULL;
+ }
+
+ if (db->db_buf)
+ dbuf_gone = arc_buf_evict(db->db_buf);
+
+ if (!dbuf_gone)
+ mutex_exit(&db->db_mtx);
+
+ /*
+ * If this dbuf is referened from an indirect dbuf,
+ * decrement the ref count on the indirect dbuf.
+ */
+ if (parent && parent != dndb)
+ dbuf_rele(parent, db);
+}
+
+static int
+dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse,
+ dmu_buf_impl_t **parentp, blkptr_t **bpp)
+{
+ int nlevels, epbs;
+
+ *parentp = NULL;
+ *bpp = NULL;
+
+ ASSERT(blkid != DB_BONUS_BLKID);
+
+ if (dn->dn_phys->dn_nlevels == 0)
+ nlevels = 1;
+ else
+ nlevels = dn->dn_phys->dn_nlevels;
+
+ epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+
+ ASSERT3U(level * epbs, <, 64);
+ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
+ if (level >= nlevels ||
+ (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) {
+ /* the buffer has no parent yet */
+ return (ENOENT);
+ } else if (level < nlevels-1) {
+ /* this block is referenced from an indirect block */
+ int err = dbuf_hold_impl(dn, level+1,
+ blkid >> epbs, fail_sparse, NULL, parentp);
+ if (err)
+ return (err);
+ err = dbuf_read(*parentp, NULL,
+ (DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH | DB_RF_CANFAIL));
+ if (err) {
+ dbuf_rele(*parentp, NULL);
+ *parentp = NULL;
+ return (err);
+ }
+ *bpp = ((blkptr_t *)(*parentp)->db.db_data) +
+ (blkid & ((1ULL << epbs) - 1));
+ return (0);
+ } else {
+ /* the block is referenced from the dnode */
+ ASSERT3U(level, ==, nlevels-1);
+ ASSERT(dn->dn_phys->dn_nblkptr == 0 ||
+ blkid < dn->dn_phys->dn_nblkptr);
+ if (dn->dn_dbuf) {
+ dbuf_add_ref(dn->dn_dbuf, NULL);
+ *parentp = dn->dn_dbuf;
+ }
+ *bpp = &dn->dn_phys->dn_blkptr[blkid];
+ return (0);
+ }
+}
+
+static dmu_buf_impl_t *
+dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid,
+ dmu_buf_impl_t *parent, blkptr_t *blkptr)
+{
+ objset_impl_t *os = dn->dn_objset;
+ dmu_buf_impl_t *db, *odb;
+
+ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
+ ASSERT(dn->dn_type != DMU_OT_NONE);
+
+ db = kmem_cache_alloc(dbuf_cache, KM_SLEEP);
+
+ db->db_objset = os;
+ db->db.db_object = dn->dn_object;
+ db->db_level = level;
+ db->db_blkid = blkid;
+ db->db_last_dirty = NULL;
+ db->db_dirtycnt = 0;
+ db->db_dnode = dn;
+ db->db_parent = parent;
+ db->db_blkptr = blkptr;
+
+ db->db_user_ptr = NULL;
+ db->db_user_data_ptr_ptr = NULL;
+ db->db_evict_func = NULL;
+ db->db_immediate_evict = 0;
+ db->db_freed_in_flight = 0;
+
+ if (blkid == DB_BONUS_BLKID) {
+ ASSERT3P(parent, ==, dn->dn_dbuf);
+ db->db.db_size = DN_MAX_BONUSLEN -
+ (dn->dn_nblkptr-1) * sizeof (blkptr_t);
+ ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
+ db->db.db_offset = DB_BONUS_BLKID;
+ db->db_state = DB_UNCACHED;
+ /* the bonus dbuf is not placed in the hash table */
+ arc_space_consume(sizeof (dmu_buf_impl_t));
+ return (db);
+ } else {
+ int blocksize =
+ db->db_level ? 1<<dn->dn_indblkshift : dn->dn_datablksz;
+ db->db.db_size = blocksize;
+ db->db.db_offset = db->db_blkid * blocksize;
+ }
+
+ /*
+ * Hold the dn_dbufs_mtx while we get the new dbuf
+ * in the hash table *and* added to the dbufs list.
+ * This prevents a possible deadlock with someone
+ * trying to look up this dbuf before its added to the
+ * dn_dbufs list.
+ */
+ mutex_enter(&dn->dn_dbufs_mtx);
+ db->db_state = DB_EVICTING;
+ if ((odb = dbuf_hash_insert(db)) != NULL) {
+ /* someone else inserted it first */
+ kmem_cache_free(dbuf_cache, db);
+ mutex_exit(&dn->dn_dbufs_mtx);
+ return (odb);
+ }
+ list_insert_head(&dn->dn_dbufs, db);
+ db->db_state = DB_UNCACHED;
+ mutex_exit(&dn->dn_dbufs_mtx);
+ arc_space_consume(sizeof (dmu_buf_impl_t));
+
+ if (parent && parent != dn->dn_dbuf)
+ dbuf_add_ref(parent, db);
+
+ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT ||
+ refcount_count(&dn->dn_holds) > 0);
+ (void) refcount_add(&dn->dn_holds, db);
+
+ dprintf_dbuf(db, "db=%p\n", db);
+
+ return (db);
+}
+
+static int
+dbuf_do_evict(void *private)
+{
+ arc_buf_t *buf = private;
+ dmu_buf_impl_t *db = buf->b_private;
+
+ if (!MUTEX_HELD(&db->db_mtx))
+ mutex_enter(&db->db_mtx);
+
+ ASSERT(refcount_is_zero(&db->db_holds));
+
+ if (db->db_state != DB_EVICTING) {
+ ASSERT(db->db_state == DB_CACHED);
+ DBUF_VERIFY(db);
+ db->db_buf = NULL;
+ dbuf_evict(db);
+ } else {
+ mutex_exit(&db->db_mtx);
+ dbuf_destroy(db);
+ }
+ return (0);
+}
+
+static void
+dbuf_destroy(dmu_buf_impl_t *db)
+{
+ ASSERT(refcount_is_zero(&db->db_holds));
+
+ if (db->db_blkid != DB_BONUS_BLKID) {
+ /*
+ * If this dbuf is still on the dn_dbufs list,
+ * remove it from that list.
+ */
+ if (db->db_dnode) {
+ dnode_t *dn = db->db_dnode;
+
+ mutex_enter(&dn->dn_dbufs_mtx);
+ list_remove(&dn->dn_dbufs, db);
+ mutex_exit(&dn->dn_dbufs_mtx);
+
+ dnode_rele(dn, db);
+ db->db_dnode = NULL;
+ }
+ dbuf_hash_remove(db);
+ }
+ db->db_parent = NULL;
+ db->db_buf = NULL;
+
+ ASSERT(!list_link_active(&db->db_link));
+ ASSERT(db->db.db_data == NULL);
+ ASSERT(db->db_hash_next == NULL);
+ ASSERT(db->db_blkptr == NULL);
+ ASSERT(db->db_data_pending == NULL);
+
+ kmem_cache_free(dbuf_cache, db);
+ arc_space_return(sizeof (dmu_buf_impl_t));
+}
+
+void
+dbuf_prefetch(dnode_t *dn, uint64_t blkid)
+{
+ dmu_buf_impl_t *db = NULL;
+ blkptr_t *bp = NULL;
+
+ ASSERT(blkid != DB_BONUS_BLKID);
+ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
+
+ if (dnode_block_freed(dn, blkid))
+ return;
+
+ /* dbuf_find() returns with db_mtx held */
+ if (db = dbuf_find(dn, 0, blkid)) {
+ if (refcount_count(&db->db_holds) > 0) {
+ /*
+ * This dbuf is active. We assume that it is
+ * already CACHED, or else about to be either
+ * read or filled.
+ */
+ mutex_exit(&db->db_mtx);
+ return;
+ }
+ mutex_exit(&db->db_mtx);
+ db = NULL;
+ }
+
+ if (dbuf_findbp(dn, 0, blkid, TRUE, &db, &bp) == 0) {
+ if (bp && !BP_IS_HOLE(bp)) {
+ arc_buf_t *pbuf;
+ uint32_t aflags = ARC_NOWAIT | ARC_PREFETCH;
+ zbookmark_t zb;
+ zb.zb_objset = dn->dn_objset->os_dsl_dataset ?
+ dn->dn_objset->os_dsl_dataset->ds_object : 0;
+ zb.zb_object = dn->dn_object;
+ zb.zb_level = 0;
+ zb.zb_blkid = blkid;
+
+ if (db)
+ pbuf = db->db_buf;
+ else
+ pbuf = dn->dn_objset->os_phys_buf;
+
+ (void) arc_read(NULL, dn->dn_objset->os_spa,
+ bp, pbuf, NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
+ &aflags, &zb);
+ }
+ if (db)
+ dbuf_rele(db, NULL);
+ }
+}
+
+/*
+ * Returns with db_holds incremented, and db_mtx not held.
+ * Note: dn_struct_rwlock must be held.
+ */
+int
+dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, int fail_sparse,
+ void *tag, dmu_buf_impl_t **dbp)
+{
+ dmu_buf_impl_t *db, *parent = NULL;
+
+ ASSERT(blkid != DB_BONUS_BLKID);
+ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
+ ASSERT3U(dn->dn_nlevels, >, level);
+
+ *dbp = NULL;
+top:
+ /* dbuf_find() returns with db_mtx held */
+ db = dbuf_find(dn, level, blkid);
+
+ if (db == NULL) {
+ blkptr_t *bp = NULL;
+ int err;
+
+ ASSERT3P(parent, ==, NULL);
+ err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp);
+ if (fail_sparse) {
+ if (err == 0 && bp && BP_IS_HOLE(bp))
+ err = ENOENT;
+ if (err) {
+ if (parent)
+ dbuf_rele(parent, NULL);
+ return (err);
+ }
+ }
+ if (err && err != ENOENT)
+ return (err);
+ db = dbuf_create(dn, level, blkid, parent, bp);
+ }
+
+ if (db->db_buf && refcount_is_zero(&db->db_holds)) {
+ arc_buf_add_ref(db->db_buf, db);
+ if (db->db_buf->b_data == NULL) {
+ dbuf_clear(db);
+ if (parent) {
+ dbuf_rele(parent, NULL);
+ parent = NULL;
+ }
+ goto top;
+ }
+ ASSERT3P(db->db.db_data, ==, db->db_buf->b_data);
+ }
+
+ ASSERT(db->db_buf == NULL || arc_referenced(db->db_buf));
+
+ /*
+ * If this buffer is currently syncing out, and we are are
+ * still referencing it from db_data, we need to make a copy
+ * of it in case we decide we want to dirty it again in this txg.
+ */
+ if (db->db_level == 0 && db->db_blkid != DB_BONUS_BLKID &&
+ dn->dn_object != DMU_META_DNODE_OBJECT &&
+ db->db_state == DB_CACHED && db->db_data_pending) {
+ dbuf_dirty_record_t *dr = db->db_data_pending;
+
+ if (dr->dt.dl.dr_data == db->db_buf) {
+ arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
+
+ dbuf_set_data(db,
+ arc_buf_alloc(db->db_dnode->dn_objset->os_spa,
+ db->db.db_size, db, type));
+ bcopy(dr->dt.dl.dr_data->b_data, db->db.db_data,
+ db->db.db_size);
+ }
+ }
+
+ (void) refcount_add(&db->db_holds, tag);
+ dbuf_update_data(db);
+ DBUF_VERIFY(db);
+ mutex_exit(&db->db_mtx);
+
+ /* NOTE: we can't rele the parent until after we drop the db_mtx */
+ if (parent)
+ dbuf_rele(parent, NULL);
+
+ ASSERT3P(db->db_dnode, ==, dn);
+ ASSERT3U(db->db_blkid, ==, blkid);
+ ASSERT3U(db->db_level, ==, level);
+ *dbp = db;
+
+ return (0);
+}
+
+dmu_buf_impl_t *
+dbuf_hold(dnode_t *dn, uint64_t blkid, void *tag)
+{
+ dmu_buf_impl_t *db;
+ int err = dbuf_hold_impl(dn, 0, blkid, FALSE, tag, &db);
+ return (err ? NULL : db);
+}
+
+dmu_buf_impl_t *
+dbuf_hold_level(dnode_t *dn, int level, uint64_t blkid, void *tag)
+{
+ dmu_buf_impl_t *db;
+ int err = dbuf_hold_impl(dn, level, blkid, FALSE, tag, &db);
+ return (err ? NULL : db);
+}
+
+void
+dbuf_create_bonus(dnode_t *dn)
+{
+ ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock));
+
+ ASSERT(dn->dn_bonus == NULL);
+ dn->dn_bonus = dbuf_create(dn, 0, DB_BONUS_BLKID, dn->dn_dbuf, NULL);
+}
+
+#pragma weak dmu_buf_add_ref = dbuf_add_ref
+void
+dbuf_add_ref(dmu_buf_impl_t *db, void *tag)
+{
+ int64_t holds = refcount_add(&db->db_holds, tag);
+ ASSERT(holds > 1);
+}
+
+#pragma weak dmu_buf_rele = dbuf_rele
+void
+dbuf_rele(dmu_buf_impl_t *db, void *tag)
+{
+ int64_t holds;
+
+ mutex_enter(&db->db_mtx);
+ DBUF_VERIFY(db);
+
+ holds = refcount_remove(&db->db_holds, tag);
+ ASSERT(holds >= 0);
+
+ /*
+ * We can't freeze indirects if there is a possibility that they
+ * may be modified in the current syncing context.
+ */
+ if (db->db_buf && holds == (db->db_level == 0 ? db->db_dirtycnt : 0))
+ arc_buf_freeze(db->db_buf);
+
+ if (holds == db->db_dirtycnt &&
+ db->db_level == 0 && db->db_immediate_evict)
+ dbuf_evict_user(db);
+
+ if (holds == 0) {
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ mutex_exit(&db->db_mtx);
+ dnode_rele(db->db_dnode, db);
+ } else if (db->db_buf == NULL) {
+ /*
+ * This is a special case: we never associated this
+ * dbuf with any data allocated from the ARC.
+ */
+ ASSERT(db->db_state == DB_UNCACHED ||
+ db->db_state == DB_NOFILL);
+ dbuf_evict(db);
+ } else if (arc_released(db->db_buf)) {
+ arc_buf_t *buf = db->db_buf;
+ /*
+ * This dbuf has anonymous data associated with it.
+ */
+ dbuf_set_data(db, NULL);
+ VERIFY(arc_buf_remove_ref(buf, db) == 1);
+ dbuf_evict(db);
+ } else {
+ VERIFY(arc_buf_remove_ref(db->db_buf, db) == 0);
+ if (!DBUF_IS_CACHEABLE(db))
+ dbuf_clear(db);
+ else
+ mutex_exit(&db->db_mtx);
+ }
+ } else {
+ mutex_exit(&db->db_mtx);
+ }
+}
+
+#pragma weak dmu_buf_refcount = dbuf_refcount
+uint64_t
+dbuf_refcount(dmu_buf_impl_t *db)
+{
+ return (refcount_count(&db->db_holds));
+}
+
+void *
+dmu_buf_set_user(dmu_buf_t *db_fake, void *user_ptr, void *user_data_ptr_ptr,
+ dmu_buf_evict_func_t *evict_func)
+{
+ return (dmu_buf_update_user(db_fake, NULL, user_ptr,
+ user_data_ptr_ptr, evict_func));
+}
+
+void *
+dmu_buf_set_user_ie(dmu_buf_t *db_fake, void *user_ptr, void *user_data_ptr_ptr,
+ dmu_buf_evict_func_t *evict_func)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+
+ db->db_immediate_evict = TRUE;
+ return (dmu_buf_update_user(db_fake, NULL, user_ptr,
+ user_data_ptr_ptr, evict_func));
+}
+
+void *
+dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr, void *user_ptr,
+ void *user_data_ptr_ptr, dmu_buf_evict_func_t *evict_func)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+ ASSERT(db->db_level == 0);
+
+ ASSERT((user_ptr == NULL) == (evict_func == NULL));
+
+ mutex_enter(&db->db_mtx);
+
+ if (db->db_user_ptr == old_user_ptr) {
+ db->db_user_ptr = user_ptr;
+ db->db_user_data_ptr_ptr = user_data_ptr_ptr;
+ db->db_evict_func = evict_func;
+
+ dbuf_update_data(db);
+ } else {
+ old_user_ptr = db->db_user_ptr;
+ }
+
+ mutex_exit(&db->db_mtx);
+ return (old_user_ptr);
+}
+
+void *
+dmu_buf_get_user(dmu_buf_t *db_fake)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+ ASSERT(!refcount_is_zero(&db->db_holds));
+
+ return (db->db_user_ptr);
+}
+
+static void
+dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db)
+{
+ /* ASSERT(dmu_tx_is_syncing(tx) */
+ ASSERT(MUTEX_HELD(&db->db_mtx));
+
+ if (db->db_blkptr != NULL)
+ return;
+
+ if (db->db_level == dn->dn_phys->dn_nlevels-1) {
+ /*
+ * This buffer was allocated at a time when there was
+ * no available blkptrs from the dnode, or it was
+ * inappropriate to hook it in (i.e., nlevels mis-match).
+ */
+ ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr);
+ ASSERT(db->db_parent == NULL);
+ db->db_parent = dn->dn_dbuf;
+ db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid];
+ DBUF_VERIFY(db);
+ } else {
+ dmu_buf_impl_t *parent = db->db_parent;
+ int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
+
+ ASSERT(dn->dn_phys->dn_nlevels > 1);
+ if (parent == NULL) {
+ mutex_exit(&db->db_mtx);
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ (void) dbuf_hold_impl(dn, db->db_level+1,
+ db->db_blkid >> epbs, FALSE, db, &parent);
+ rw_exit(&dn->dn_struct_rwlock);
+ mutex_enter(&db->db_mtx);
+ db->db_parent = parent;
+ }
+ db->db_blkptr = (blkptr_t *)parent->db.db_data +
+ (db->db_blkid & ((1ULL << epbs) - 1));
+ DBUF_VERIFY(db);
+ }
+}
+
+static void
+dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ dnode_t *dn = db->db_dnode;
+ zio_t *zio;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);
+
+ mutex_enter(&db->db_mtx);
+
+ ASSERT(db->db_level > 0);
+ DBUF_VERIFY(db);
+
+ if (db->db_buf == NULL) {
+ mutex_exit(&db->db_mtx);
+ (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED);
+ mutex_enter(&db->db_mtx);
+ }
+ ASSERT3U(db->db_state, ==, DB_CACHED);
+ ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
+ ASSERT(db->db_buf != NULL);
+
+ dbuf_check_blkptr(dn, db);
+
+ db->db_data_pending = dr;
+
+ mutex_exit(&db->db_mtx);
+ dbuf_write(dr, db->db_buf, tx);
+
+ zio = dr->dr_zio;
+ mutex_enter(&dr->dt.di.dr_mtx);
+ dbuf_sync_list(&dr->dt.di.dr_children, tx);
+ ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
+ mutex_exit(&dr->dt.di.dr_mtx);
+ zio_nowait(zio);
+}
+
+static void
+dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
+{
+ arc_buf_t **datap = &dr->dt.dl.dr_data;
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ dnode_t *dn = db->db_dnode;
+ objset_impl_t *os = dn->dn_objset;
+ uint64_t txg = tx->tx_txg;
+ int blksz;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);
+
+ mutex_enter(&db->db_mtx);
+ /*
+ * To be synced, we must be dirtied. But we
+ * might have been freed after the dirty.
+ */
+ if (db->db_state == DB_UNCACHED) {
+ /* This buffer has been freed since it was dirtied */
+ ASSERT(db->db.db_data == NULL);
+ } else if (db->db_state == DB_FILL) {
+ /* This buffer was freed and is now being re-filled */
+ ASSERT(db->db.db_data != dr->dt.dl.dr_data);
+ } else {
+ ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL);
+ }
+ DBUF_VERIFY(db);
+
+ /*
+ * If this is a bonus buffer, simply copy the bonus data into the
+ * dnode. It will be written out when the dnode is synced (and it
+ * will be synced, since it must have been dirty for dbuf_sync to
+ * be called).
+ */
+ if (db->db_blkid == DB_BONUS_BLKID) {
+ dbuf_dirty_record_t **drp;
+
+ ASSERT(*datap != NULL);
+ ASSERT3U(db->db_level, ==, 0);
+ ASSERT3U(dn->dn_phys->dn_bonuslen, <=, DN_MAX_BONUSLEN);
+ bcopy(*datap, DN_BONUS(dn->dn_phys), dn->dn_phys->dn_bonuslen);
+ if (*datap != db->db.db_data) {
+ zio_buf_free(*datap, DN_MAX_BONUSLEN);
+ arc_space_return(DN_MAX_BONUSLEN);
+ }
+ db->db_data_pending = NULL;
+ drp = &db->db_last_dirty;
+ while (*drp != dr)
+ drp = &(*drp)->dr_next;
+ ASSERT(dr->dr_next == NULL);
+ *drp = dr->dr_next;
+ kmem_free(dr, sizeof (dbuf_dirty_record_t));
+ ASSERT(db->db_dirtycnt > 0);
+ db->db_dirtycnt -= 1;
+ mutex_exit(&db->db_mtx);
+ dbuf_rele(db, (void *)(uintptr_t)txg);
+ return;
+ }
+
+ /*
+ * This function may have dropped the db_mtx lock allowing a dmu_sync
+ * operation to sneak in. As a result, we need to ensure that we
+ * don't check the dr_override_state until we have returned from
+ * dbuf_check_blkptr.
+ */
+ dbuf_check_blkptr(dn, db);
+
+ /*
+ * If this buffer is in the middle of an immdiate write,
+ * wait for the synchronous IO to complete.
+ */
+ while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
+ ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
+ cv_wait(&db->db_changed, &db->db_mtx);
+ ASSERT(dr->dt.dl.dr_override_state != DR_NOT_OVERRIDDEN);
+ }
+
+ /*
+ * If this dbuf has already been written out via an immediate write,
+ * just complete the write by copying over the new block pointer and
+ * updating the accounting via the write-completion functions.
+ */
+ if (dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
+ zio_t zio_fake;
+
+ zio_fake.io_private = &db;
+ zio_fake.io_error = 0;
+ zio_fake.io_bp = db->db_blkptr;
+ zio_fake.io_bp_orig = *db->db_blkptr;
+ zio_fake.io_txg = txg;
+ zio_fake.io_flags = 0;
+
+ *db->db_blkptr = dr->dt.dl.dr_overridden_by;
+ dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
+ db->db_data_pending = dr;
+ dr->dr_zio = &zio_fake;
+ mutex_exit(&db->db_mtx);
+
+ ASSERT(!DVA_EQUAL(BP_IDENTITY(zio_fake.io_bp),
+ BP_IDENTITY(&zio_fake.io_bp_orig)) ||
+ BP_IS_HOLE(zio_fake.io_bp));
+
+ if (BP_IS_OLDER(&zio_fake.io_bp_orig, txg))
+ (void) dsl_dataset_block_kill(os->os_dsl_dataset,
+ &zio_fake.io_bp_orig, dn->dn_zio, tx);
+
+ dbuf_write_ready(&zio_fake, db->db_buf, db);
+ dbuf_write_done(&zio_fake, db->db_buf, db);
+
+ return;
+ }
+
+ if (db->db_state != DB_NOFILL) {
+ blksz = arc_buf_size(*datap);
+
+ if (dn->dn_object != DMU_META_DNODE_OBJECT) {
+ /*
+ * If this buffer is currently "in use" (i.e., there
+ * are active holds and db_data still references it),
+ * then make a copy before we start the write so that
+ * any modifications from the open txg will not leak
+ * into this write.
+ *
+ * NOTE: this copy does not need to be made for
+ * objects only modified in the syncing context (e.g.
+ * DNONE_DNODE blocks).
+ */
+ if (refcount_count(&db->db_holds) > 1 &&
+ *datap == db->db_buf) {
+ arc_buf_contents_t type =
+ DBUF_GET_BUFC_TYPE(db);
+ *datap =
+ arc_buf_alloc(os->os_spa, blksz, db, type);
+ bcopy(db->db.db_data, (*datap)->b_data, blksz);
+ }
+ }
+
+ ASSERT(*datap != NULL);
+ }
+ db->db_data_pending = dr;
+
+ mutex_exit(&db->db_mtx);
+
+ dbuf_write(dr, *datap, tx);
+
+ ASSERT(!list_link_active(&dr->dr_dirty_node));
+ if (dn->dn_object == DMU_META_DNODE_OBJECT)
+ list_insert_tail(&dn->dn_dirty_records[txg&TXG_MASK], dr);
+ else
+ zio_nowait(dr->dr_zio);
+}
+
+void
+dbuf_sync_list(list_t *list, dmu_tx_t *tx)
+{
+ dbuf_dirty_record_t *dr;
+
+ while (dr = list_head(list)) {
+ if (dr->dr_zio != NULL) {
+ /*
+ * If we find an already initialized zio then we
+ * are processing the meta-dnode, and we have finished.
+ * The dbufs for all dnodes are put back on the list
+ * during processing, so that we can zio_wait()
+ * these IOs after initiating all child IOs.
+ */
+ ASSERT3U(dr->dr_dbuf->db.db_object, ==,
+ DMU_META_DNODE_OBJECT);
+ break;
+ }
+ list_remove(list, dr);
+ if (dr->dr_dbuf->db_level > 0)
+ dbuf_sync_indirect(dr, tx);
+ else
+ dbuf_sync_leaf(dr, tx);
+ }
+}
+
+static void
+dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ dnode_t *dn = db->db_dnode;
+ objset_impl_t *os = dn->dn_objset;
+ dmu_buf_impl_t *parent = db->db_parent;
+ uint64_t txg = tx->tx_txg;
+ zbookmark_t zb;
+ writeprops_t wp = { 0 };
+ zio_t *zio;
+
+ if (!BP_IS_HOLE(db->db_blkptr) &&
+ (db->db_level > 0 || dn->dn_type == DMU_OT_DNODE)) {
+ /*
+ * Private object buffers are released here rather
+ * than in dbuf_dirty() since they are only modified
+ * in the syncing context and we don't want the
+ * overhead of making multiple copies of the data.
+ */
+ arc_release(data, db);
+ } else if (db->db_state != DB_NOFILL) {
+ ASSERT(arc_released(data));
+ /* XXX why do we need to thaw here? */
+ arc_buf_thaw(data);
+ }
+
+ if (parent != dn->dn_dbuf) {
+ ASSERT(parent && parent->db_data_pending);
+ ASSERT(db->db_level == parent->db_level-1);
+ ASSERT(arc_released(parent->db_buf));
+ zio = parent->db_data_pending->dr_zio;
+ } else {
+ ASSERT(db->db_level == dn->dn_phys->dn_nlevels-1);
+ ASSERT3P(db->db_blkptr, ==,
+ &dn->dn_phys->dn_blkptr[db->db_blkid]);
+ zio = dn->dn_zio;
+ }
+
+ ASSERT(db->db_level == 0 || data == db->db_buf);
+ ASSERT3U(db->db_blkptr->blk_birth, <=, txg);
+ ASSERT(zio);
+
+ zb.zb_objset = os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : 0;
+ zb.zb_object = db->db.db_object;
+ zb.zb_level = db->db_level;
+ zb.zb_blkid = db->db_blkid;
+
+ wp.wp_type = dn->dn_type;
+ wp.wp_level = db->db_level;
+ wp.wp_copies = os->os_copies;
+ wp.wp_dncompress = dn->dn_compress;
+ wp.wp_oscompress = os->os_compress;
+ wp.wp_dnchecksum = dn->dn_checksum;
+ wp.wp_oschecksum = os->os_checksum;
+
+ if (BP_IS_OLDER(db->db_blkptr, txg))
+ (void) dsl_dataset_block_kill(
+ os->os_dsl_dataset, db->db_blkptr, zio, tx);
+
+ if (db->db_state == DB_NOFILL) {
+ zio_prop_t zp = { 0 };
+
+ write_policy(os->os_spa, &wp, &zp);
+ dr->dr_zio = zio_write(zio, os->os_spa,
+ txg, db->db_blkptr, NULL,
+ db->db.db_size, &zp, dbuf_skip_write_ready,
+ dbuf_skip_write_done, db, ZIO_PRIORITY_ASYNC_WRITE,
+ ZIO_FLAG_MUSTSUCCEED, &zb);
+ } else {
+ dr->dr_zio = arc_write(zio, os->os_spa, &wp,
+ DBUF_IS_L2CACHEABLE(db), txg, db->db_blkptr,
+ data, dbuf_write_ready, dbuf_write_done, db,
+ ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
+ }
+}
+
+/* wrapper function for dbuf_write_ready bypassing ARC */
+static void
+dbuf_skip_write_ready(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+
+ if (!BP_IS_GANG(bp))
+ zio_skip_write(zio);
+
+ dbuf_write_ready(zio, NULL, zio->io_private);
+}
+
+/* wrapper function for dbuf_write_done bypassing ARC */
+static void
+dbuf_skip_write_done(zio_t *zio)
+{
+ dbuf_write_done(zio, NULL, zio->io_private);
+}
+
+/* ARGSUSED */
+static void
+dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb)
+{
+ dmu_buf_impl_t *db = vdb;
+ dnode_t *dn = db->db_dnode;
+ objset_impl_t *os = dn->dn_objset;
+ blkptr_t *bp = zio->io_bp;
+ blkptr_t *bp_orig = &zio->io_bp_orig;
+ uint64_t fill = 0;
+ int old_size, new_size, i;
+
+ ASSERT(db->db_blkptr == bp);
+
+ dprintf_dbuf_bp(db, bp_orig, "bp_orig: %s", "");
+
+ old_size = bp_get_dasize(os->os_spa, bp_orig);
+ new_size = bp_get_dasize(os->os_spa, bp);
+
+ dnode_diduse_space(dn, new_size - old_size);
+
+ if (BP_IS_HOLE(bp)) {
+ dsl_dataset_t *ds = os->os_dsl_dataset;
+ dmu_tx_t *tx = os->os_synctx;
+
+ if (bp_orig->blk_birth == tx->tx_txg)
+ (void) dsl_dataset_block_kill(ds, bp_orig, zio, tx);
+ ASSERT3U(bp->blk_fill, ==, 0);
+ return;
+ }
+
+ ASSERT(BP_GET_TYPE(bp) == dn->dn_type);
+ ASSERT(BP_GET_LEVEL(bp) == db->db_level);
+
+ mutex_enter(&db->db_mtx);
+
+ if (db->db_level == 0) {
+ mutex_enter(&dn->dn_mtx);
+ if (db->db_blkid > dn->dn_phys->dn_maxblkid)
+ dn->dn_phys->dn_maxblkid = db->db_blkid;
+ mutex_exit(&dn->dn_mtx);
+
+ if (dn->dn_type == DMU_OT_DNODE) {
+ dnode_phys_t *dnp = db->db.db_data;
+ for (i = db->db.db_size >> DNODE_SHIFT; i > 0;
+ i--, dnp++) {
+ if (dnp->dn_type != DMU_OT_NONE)
+ fill++;
+ }
+ } else {
+ fill = 1;
+ }
+ } else {
+ blkptr_t *ibp = db->db.db_data;
+ ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
+ for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) {
+ if (BP_IS_HOLE(ibp))
+ continue;
+ ASSERT3U(BP_GET_LSIZE(ibp), ==,
+ db->db_level == 1 ? dn->dn_datablksz :
+ (1<<dn->dn_phys->dn_indblkshift));
+ fill += ibp->blk_fill;
+ }
+ }
+
+ bp->blk_fill = fill;
+
+ mutex_exit(&db->db_mtx);
+
+ if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
+ ASSERT(DVA_EQUAL(BP_IDENTITY(bp), BP_IDENTITY(bp_orig)));
+ } else {
+ dsl_dataset_t *ds = os->os_dsl_dataset;
+ dmu_tx_t *tx = os->os_synctx;
+
+ if (bp_orig->blk_birth == tx->tx_txg)
+ (void) dsl_dataset_block_kill(ds, bp_orig, zio, tx);
+ dsl_dataset_block_born(ds, bp, tx);
+ }
+}
+
+/* ARGSUSED */
+static void
+dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb)
+{
+ dmu_buf_impl_t *db = vdb;
+ uint64_t txg = zio->io_txg;
+ dbuf_dirty_record_t **drp, *dr;
+
+ ASSERT3U(zio->io_error, ==, 0);
+
+ mutex_enter(&db->db_mtx);
+
+ drp = &db->db_last_dirty;
+ while ((dr = *drp) != db->db_data_pending)
+ drp = &dr->dr_next;
+ ASSERT(!list_link_active(&dr->dr_dirty_node));
+ ASSERT(dr->dr_txg == txg);
+ ASSERT(dr->dr_next == NULL);
+ *drp = dr->dr_next;
+
+ if (db->db_level == 0) {
+ ASSERT(db->db_blkid != DB_BONUS_BLKID);
+ ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
+
+ if (db->db_state != DB_NOFILL) {
+ if (dr->dt.dl.dr_data != db->db_buf)
+ VERIFY(arc_buf_remove_ref(dr->dt.dl.dr_data,
+ db) == 1);
+ else if (!BP_IS_HOLE(db->db_blkptr))
+ arc_set_callback(db->db_buf, dbuf_do_evict, db);
+ else
+ ASSERT(arc_released(db->db_buf));
+ }
+ } else {
+ dnode_t *dn = db->db_dnode;
+
+ ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
+ ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
+ if (!BP_IS_HOLE(db->db_blkptr)) {
+ int epbs =
+ dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
+ ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==,
+ db->db.db_size);
+ ASSERT3U(dn->dn_phys->dn_maxblkid
+ >> (db->db_level * epbs), >=, db->db_blkid);
+ arc_set_callback(db->db_buf, dbuf_do_evict, db);
+ }
+ mutex_destroy(&dr->dt.di.dr_mtx);
+ list_destroy(&dr->dt.di.dr_children);
+ }
+ kmem_free(dr, sizeof (dbuf_dirty_record_t));
+
+ cv_broadcast(&db->db_changed);
+ ASSERT(db->db_dirtycnt > 0);
+ db->db_dirtycnt -= 1;
+ db->db_data_pending = NULL;
+ mutex_exit(&db->db_mtx);
+
+ dprintf_dbuf_bp(db, zio->io_bp, "bp: %s", "");
+
+ dbuf_rele(db, (void *)(uintptr_t)txg);
+}
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
new file mode 100644
index 000000000..b6205bd50
--- /dev/null
+++ b/module/zfs/dmu.c
@@ -0,0 +1,1227 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dmu.h>
+#include <sys/dmu_impl.h>
+#include <sys/dmu_tx.h>
+#include <sys/dbuf.h>
+#include <sys/dnode.h>
+#include <sys/zfs_context.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dsl_prop.h>
+#include <sys/dmu_zfetch.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zap.h>
+#include <sys/zio_checksum.h>
+#ifdef _KERNEL
+#include <sys/vmsystm.h>
+#include <sys/zfs_znode.h>
+#endif
+
+const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
+ { byteswap_uint8_array, TRUE, "unallocated" },
+ { zap_byteswap, TRUE, "object directory" },
+ { byteswap_uint64_array, TRUE, "object array" },
+ { byteswap_uint8_array, TRUE, "packed nvlist" },
+ { byteswap_uint64_array, TRUE, "packed nvlist size" },
+ { byteswap_uint64_array, TRUE, "bplist" },
+ { byteswap_uint64_array, TRUE, "bplist header" },
+ { byteswap_uint64_array, TRUE, "SPA space map header" },
+ { byteswap_uint64_array, TRUE, "SPA space map" },
+ { byteswap_uint64_array, TRUE, "ZIL intent log" },
+ { dnode_buf_byteswap, TRUE, "DMU dnode" },
+ { dmu_objset_byteswap, TRUE, "DMU objset" },
+ { byteswap_uint64_array, TRUE, "DSL directory" },
+ { zap_byteswap, TRUE, "DSL directory child map"},
+ { zap_byteswap, TRUE, "DSL dataset snap map" },
+ { zap_byteswap, TRUE, "DSL props" },
+ { byteswap_uint64_array, TRUE, "DSL dataset" },
+ { zfs_znode_byteswap, TRUE, "ZFS znode" },
+ { zfs_oldacl_byteswap, TRUE, "ZFS V0 ACL" },
+ { byteswap_uint8_array, FALSE, "ZFS plain file" },
+ { zap_byteswap, TRUE, "ZFS directory" },
+ { zap_byteswap, TRUE, "ZFS master node" },
+ { zap_byteswap, TRUE, "ZFS delete queue" },
+ { byteswap_uint8_array, FALSE, "zvol object" },
+ { zap_byteswap, TRUE, "zvol prop" },
+ { byteswap_uint8_array, FALSE, "other uint8[]" },
+ { byteswap_uint64_array, FALSE, "other uint64[]" },
+ { zap_byteswap, TRUE, "other ZAP" },
+ { zap_byteswap, TRUE, "persistent error log" },
+ { byteswap_uint8_array, TRUE, "SPA history" },
+ { byteswap_uint64_array, TRUE, "SPA history offsets" },
+ { zap_byteswap, TRUE, "Pool properties" },
+ { zap_byteswap, TRUE, "DSL permissions" },
+ { zfs_acl_byteswap, TRUE, "ZFS ACL" },
+ { byteswap_uint8_array, TRUE, "ZFS SYSACL" },
+ { byteswap_uint8_array, TRUE, "FUID table" },
+ { byteswap_uint64_array, TRUE, "FUID table size" },
+ { zap_byteswap, TRUE, "DSL dataset next clones"},
+ { zap_byteswap, TRUE, "scrub work queue" },
+};
+
+int
+dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
+ void *tag, dmu_buf_t **dbp)
+{
+ dnode_t *dn;
+ uint64_t blkid;
+ dmu_buf_impl_t *db;
+ int err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ blkid = dbuf_whichblock(dn, offset);
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ db = dbuf_hold(dn, blkid, tag);
+ rw_exit(&dn->dn_struct_rwlock);
+ if (db == NULL) {
+ err = EIO;
+ } else {
+ err = dbuf_read(db, NULL, DB_RF_CANFAIL);
+ if (err) {
+ dbuf_rele(db, tag);
+ db = NULL;
+ }
+ }
+
+ dnode_rele(dn, FTAG);
+ *dbp = &db->db;
+ return (err);
+}
+
+int
+dmu_bonus_max(void)
+{
+ return (DN_MAX_BONUSLEN);
+}
+
+int
+dmu_set_bonus(dmu_buf_t *db, int newsize, dmu_tx_t *tx)
+{
+ dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
+
+ if (dn->dn_bonus != (dmu_buf_impl_t *)db)
+ return (EINVAL);
+ if (newsize < 0 || newsize > db->db_size)
+ return (EINVAL);
+ dnode_setbonuslen(dn, newsize, tx);
+ return (0);
+}
+
+/*
+ * returns ENOENT, EIO, or 0.
+ */
+int
+dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
+{
+ dnode_t *dn;
+ dmu_buf_impl_t *db;
+ int error;
+
+ error = dnode_hold(os->os, object, FTAG, &dn);
+ if (error)
+ return (error);
+
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ if (dn->dn_bonus == NULL) {
+ rw_exit(&dn->dn_struct_rwlock);
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ if (dn->dn_bonus == NULL)
+ dbuf_create_bonus(dn);
+ }
+ db = dn->dn_bonus;
+ rw_exit(&dn->dn_struct_rwlock);
+
+ /* as long as the bonus buf is held, the dnode will be held */
+ if (refcount_add(&db->db_holds, tag) == 1)
+ VERIFY(dnode_add_ref(dn, db));
+
+ dnode_rele(dn, FTAG);
+
+ VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED));
+
+ *dbp = &db->db;
+ return (0);
+}
+
+/*
+ * Note: longer-term, we should modify all of the dmu_buf_*() interfaces
+ * to take a held dnode rather than <os, object> -- the lookup is wasteful,
+ * and can induce severe lock contention when writing to several files
+ * whose dnodes are in the same block.
+ */
+static int
+dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
+ uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
+{
+ dsl_pool_t *dp = NULL;
+ dmu_buf_t **dbp;
+ uint64_t blkid, nblks, i;
+ uint32_t flags;
+ int err;
+ zio_t *zio;
+ hrtime_t start;
+
+ ASSERT(length <= DMU_MAX_ACCESS);
+
+ flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT;
+ if (length > zfetch_array_rd_sz)
+ flags |= DB_RF_NOPREFETCH;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ if (dn->dn_datablkshift) {
+ int blkshift = dn->dn_datablkshift;
+ nblks = (P2ROUNDUP(offset+length, 1ULL<<blkshift) -
+ P2ALIGN(offset, 1ULL<<blkshift)) >> blkshift;
+ } else {
+ if (offset + length > dn->dn_datablksz) {
+ zfs_panic_recover("zfs: accessing past end of object "
+ "%llx/%llx (size=%u access=%llu+%llu)",
+ (longlong_t)dn->dn_objset->
+ os_dsl_dataset->ds_object,
+ (longlong_t)dn->dn_object, dn->dn_datablksz,
+ (longlong_t)offset, (longlong_t)length);
+ return (EIO);
+ }
+ nblks = 1;
+ }
+ dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP);
+
+ if (dn->dn_objset->os_dsl_dataset)
+ dp = dn->dn_objset->os_dsl_dataset->ds_dir->dd_pool;
+ if (dp && dsl_pool_sync_context(dp))
+ start = gethrtime();
+ zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL);
+ blkid = dbuf_whichblock(dn, offset);
+ for (i = 0; i < nblks; i++) {
+ dmu_buf_impl_t *db = dbuf_hold(dn, blkid+i, tag);
+ if (db == NULL) {
+ rw_exit(&dn->dn_struct_rwlock);
+ dmu_buf_rele_array(dbp, nblks, tag);
+ zio_nowait(zio);
+ return (EIO);
+ }
+ /* initiate async i/o */
+ if (read) {
+ rw_exit(&dn->dn_struct_rwlock);
+ (void) dbuf_read(db, zio, flags);
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ }
+ dbp[i] = &db->db;
+ }
+ rw_exit(&dn->dn_struct_rwlock);
+
+ /* wait for async i/o */
+ err = zio_wait(zio);
+ /* track read overhead when we are in sync context */
+ if (dp && dsl_pool_sync_context(dp))
+ dp->dp_read_overhead += gethrtime() - start;
+ if (err) {
+ dmu_buf_rele_array(dbp, nblks, tag);
+ return (err);
+ }
+
+ /* wait for other io to complete */
+ if (read) {
+ for (i = 0; i < nblks; i++) {
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i];
+ mutex_enter(&db->db_mtx);
+ while (db->db_state == DB_READ ||
+ db->db_state == DB_FILL)
+ cv_wait(&db->db_changed, &db->db_mtx);
+ if (db->db_state == DB_UNCACHED)
+ err = EIO;
+ mutex_exit(&db->db_mtx);
+ if (err) {
+ dmu_buf_rele_array(dbp, nblks, tag);
+ return (err);
+ }
+ }
+ }
+
+ *numbufsp = nblks;
+ *dbpp = dbp;
+ return (0);
+}
+
+static int
+dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
+{
+ dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+
+ err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
+ numbufsp, dbpp);
+
+ dnode_rele(dn, FTAG);
+
+ return (err);
+}
+
+int
+dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
+ uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
+{
+ dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
+ int err;
+
+ err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
+ numbufsp, dbpp);
+
+ return (err);
+}
+
+void
+dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag)
+{
+ int i;
+ dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake;
+
+ if (numbufs == 0)
+ return;
+
+ for (i = 0; i < numbufs; i++) {
+ if (dbp[i])
+ dbuf_rele(dbp[i], tag);
+ }
+
+ kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs);
+}
+
+void
+dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, uint64_t len)
+{
+ dnode_t *dn;
+ uint64_t blkid;
+ int nblks, i, err;
+
+ if (zfs_prefetch_disable)
+ return;
+
+ if (len == 0) { /* they're interested in the bonus buffer */
+ dn = os->os->os_meta_dnode;
+
+ if (object == 0 || object >= DN_MAX_OBJECT)
+ return;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ blkid = dbuf_whichblock(dn, object * sizeof (dnode_phys_t));
+ dbuf_prefetch(dn, blkid);
+ rw_exit(&dn->dn_struct_rwlock);
+ return;
+ }
+
+ /*
+ * XXX - Note, if the dnode for the requested object is not
+ * already cached, we will do a *synchronous* read in the
+ * dnode_hold() call. The same is true for any indirects.
+ */
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err != 0)
+ return;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ if (dn->dn_datablkshift) {
+ int blkshift = dn->dn_datablkshift;
+ nblks = (P2ROUNDUP(offset+len, 1<<blkshift) -
+ P2ALIGN(offset, 1<<blkshift)) >> blkshift;
+ } else {
+ nblks = (offset < dn->dn_datablksz);
+ }
+
+ if (nblks != 0) {
+ blkid = dbuf_whichblock(dn, offset);
+ for (i = 0; i < nblks; i++)
+ dbuf_prefetch(dn, blkid+i);
+ }
+
+ rw_exit(&dn->dn_struct_rwlock);
+
+ dnode_rele(dn, FTAG);
+}
+
+static int
+get_next_chunk(dnode_t *dn, uint64_t *offset, uint64_t limit)
+{
+ uint64_t len = *offset - limit;
+ uint64_t chunk_len = dn->dn_datablksz * DMU_MAX_DELETEBLKCNT;
+ uint64_t subchunk =
+ dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);
+
+ ASSERT(limit <= *offset);
+
+ if (len <= chunk_len) {
+ *offset = limit;
+ return (0);
+ }
+
+ ASSERT(ISP2(subchunk));
+
+ while (*offset > limit) {
+ uint64_t initial_offset = P2ROUNDUP(*offset, subchunk);
+ uint64_t delta;
+ int err;
+
+ /* skip over allocated data */
+ err = dnode_next_offset(dn,
+ DNODE_FIND_HOLE|DNODE_FIND_BACKWARDS, offset, 1, 1, 0);
+ if (err == ESRCH)
+ *offset = limit;
+ else if (err)
+ return (err);
+
+ ASSERT3U(*offset, <=, initial_offset);
+ *offset = P2ALIGN(*offset, subchunk);
+ delta = initial_offset - *offset;
+ if (delta >= chunk_len) {
+ *offset += delta - chunk_len;
+ return (0);
+ }
+ chunk_len -= delta;
+
+ /* skip over unallocated data */
+ err = dnode_next_offset(dn,
+ DNODE_FIND_BACKWARDS, offset, 1, 1, 0);
+ if (err == ESRCH)
+ *offset = limit;
+ else if (err)
+ return (err);
+
+ if (*offset < limit)
+ *offset = limit;
+ ASSERT3U(*offset, <, initial_offset);
+ }
+ return (0);
+}
+
+static int
+dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
+ uint64_t length, boolean_t free_dnode)
+{
+ dmu_tx_t *tx;
+ uint64_t object_size, start, end, len;
+ boolean_t trunc = (length == DMU_OBJECT_END);
+ int align, err;
+
+ align = 1 << dn->dn_datablkshift;
+ ASSERT(align > 0);
+ object_size = align == 1 ? dn->dn_datablksz :
+ (dn->dn_maxblkid + 1) << dn->dn_datablkshift;
+
+ if (trunc || (end = offset + length) > object_size)
+ end = object_size;
+ if (end <= offset)
+ return (0);
+ length = end - offset;
+
+ while (length) {
+ start = end;
+ err = get_next_chunk(dn, &start, offset);
+ if (err)
+ return (err);
+ len = trunc ? DMU_OBJECT_END : end - start;
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_free(tx, dn->dn_object, start, len);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+
+ dnode_free_range(dn, start, trunc ? -1 : len, tx);
+
+ if (start == 0 && free_dnode) {
+ ASSERT(trunc);
+ dnode_free(dn, tx);
+ }
+
+ length -= end - start;
+
+ dmu_tx_commit(tx);
+ end = start;
+ }
+ return (0);
+}
+
+int
+dmu_free_long_range(objset_t *os, uint64_t object,
+ uint64_t offset, uint64_t length)
+{
+ dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err != 0)
+ return (err);
+ err = dmu_free_long_range_impl(os, dn, offset, length, FALSE);
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+int
+dmu_free_object(objset_t *os, uint64_t object)
+{
+ dnode_t *dn;
+ dmu_tx_t *tx;
+ int err;
+
+ err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED,
+ FTAG, &dn);
+ if (err != 0)
+ return (err);
+ if (dn->dn_nlevels == 1) {
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_bonus(tx, object);
+ dmu_tx_hold_free(tx, dn->dn_object, 0, DMU_OBJECT_END);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err == 0) {
+ dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
+ dnode_free(dn, tx);
+ dmu_tx_commit(tx);
+ } else {
+ dmu_tx_abort(tx);
+ }
+ } else {
+ err = dmu_free_long_range_impl(os, dn, 0, DMU_OBJECT_END, TRUE);
+ }
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+int
+dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t size, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ ASSERT(offset < UINT64_MAX);
+ ASSERT(size == -1ULL || size <= UINT64_MAX - offset);
+ dnode_free_range(dn, offset, size, tx);
+ dnode_rele(dn, FTAG);
+ return (0);
+}
+
+int
+dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ void *buf)
+{
+ dnode_t *dn;
+ dmu_buf_t **dbp;
+ int numbufs, i, err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+
+ /*
+ * Deal with odd block sizes, where there can't be data past the first
+ * block. If we ever do the tail block optimization, we will need to
+ * handle that here as well.
+ */
+ if (dn->dn_datablkshift == 0) {
+ int newsz = offset > dn->dn_datablksz ? 0 :
+ MIN(size, dn->dn_datablksz - offset);
+ bzero((char *)buf + newsz, size - newsz);
+ size = newsz;
+ }
+
+ while (size > 0) {
+ uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2);
+
+ /*
+ * NB: we could do this block-at-a-time, but it's nice
+ * to be reading in parallel.
+ */
+ err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
+ TRUE, FTAG, &numbufs, &dbp);
+ if (err)
+ break;
+
+ for (i = 0; i < numbufs; i++) {
+ int tocpy;
+ int bufoff;
+ dmu_buf_t *db = dbp[i];
+
+ ASSERT(size > 0);
+
+ bufoff = offset - db->db_offset;
+ tocpy = (int)MIN(db->db_size - bufoff, size);
+
+ bcopy((char *)db->db_data + bufoff, buf, tocpy);
+
+ offset += tocpy;
+ size -= tocpy;
+ buf = (char *)buf + tocpy;
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+ }
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+void
+dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ const void *buf, dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs, i;
+
+ if (size == 0)
+ return;
+
+ VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
+ FALSE, FTAG, &numbufs, &dbp));
+
+ for (i = 0; i < numbufs; i++) {
+ int tocpy;
+ int bufoff;
+ dmu_buf_t *db = dbp[i];
+
+ ASSERT(size > 0);
+
+ bufoff = offset - db->db_offset;
+ tocpy = (int)MIN(db->db_size - bufoff, size);
+
+ ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
+
+ if (tocpy == db->db_size)
+ dmu_buf_will_fill(db, tx);
+ else
+ dmu_buf_will_dirty(db, tx);
+
+ bcopy(buf, (char *)db->db_data + bufoff, tocpy);
+
+ if (tocpy == db->db_size)
+ dmu_buf_fill_done(db, tx);
+
+ offset += tocpy;
+ size -= tocpy;
+ buf = (char *)buf + tocpy;
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+}
+
+void
+dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs, i;
+
+ if (size == 0)
+ return;
+
+ VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
+ FALSE, FTAG, &numbufs, &dbp));
+
+ for (i = 0; i < numbufs; i++) {
+ dmu_buf_t *db = dbp[i];
+
+ dmu_buf_will_not_fill(db, tx);
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+}
+
+#ifdef _KERNEL
+int
+dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
+{
+ dmu_buf_t **dbp;
+ int numbufs, i, err;
+
+ /*
+ * NB: we could do this block-at-a-time, but it's nice
+ * to be reading in parallel.
+ */
+ err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, TRUE, FTAG,
+ &numbufs, &dbp);
+ if (err)
+ return (err);
+
+ for (i = 0; i < numbufs; i++) {
+ int tocpy;
+ int bufoff;
+ dmu_buf_t *db = dbp[i];
+
+ ASSERT(size > 0);
+
+ bufoff = uio->uio_loffset - db->db_offset;
+ tocpy = (int)MIN(db->db_size - bufoff, size);
+
+ err = uiomove((char *)db->db_data + bufoff, tocpy,
+ UIO_READ, uio);
+ if (err)
+ break;
+
+ size -= tocpy;
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+
+ return (err);
+}
+
+int
+dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size,
+ dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs, i;
+ int err = 0;
+
+ if (size == 0)
+ return (0);
+
+ err = dmu_buf_hold_array(os, object, uio->uio_loffset, size,
+ FALSE, FTAG, &numbufs, &dbp);
+ if (err)
+ return (err);
+
+ for (i = 0; i < numbufs; i++) {
+ int tocpy;
+ int bufoff;
+ dmu_buf_t *db = dbp[i];
+
+ ASSERT(size > 0);
+
+ bufoff = uio->uio_loffset - db->db_offset;
+ tocpy = (int)MIN(db->db_size - bufoff, size);
+
+ ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
+
+ if (tocpy == db->db_size)
+ dmu_buf_will_fill(db, tx);
+ else
+ dmu_buf_will_dirty(db, tx);
+
+ /*
+ * XXX uiomove could block forever (eg. nfs-backed
+ * pages). There needs to be a uiolockdown() function
+ * to lock the pages in memory, so that uiomove won't
+ * block.
+ */
+ err = uiomove((char *)db->db_data + bufoff, tocpy,
+ UIO_WRITE, uio);
+
+ if (tocpy == db->db_size)
+ dmu_buf_fill_done(db, tx);
+
+ if (err)
+ break;
+
+ size -= tocpy;
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+ return (err);
+}
+
+int
+dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ page_t *pp, dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs, i;
+ int err;
+
+ if (size == 0)
+ return (0);
+
+ err = dmu_buf_hold_array(os, object, offset, size,
+ FALSE, FTAG, &numbufs, &dbp);
+ if (err)
+ return (err);
+
+ for (i = 0; i < numbufs; i++) {
+ int tocpy, copied, thiscpy;
+ int bufoff;
+ dmu_buf_t *db = dbp[i];
+ caddr_t va;
+
+ ASSERT(size > 0);
+ ASSERT3U(db->db_size, >=, PAGESIZE);
+
+ bufoff = offset - db->db_offset;
+ tocpy = (int)MIN(db->db_size - bufoff, size);
+
+ ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
+
+ if (tocpy == db->db_size)
+ dmu_buf_will_fill(db, tx);
+ else
+ dmu_buf_will_dirty(db, tx);
+
+ for (copied = 0; copied < tocpy; copied += PAGESIZE) {
+ ASSERT3U(pp->p_offset, ==, db->db_offset + bufoff);
+ thiscpy = MIN(PAGESIZE, tocpy - copied);
+ va = zfs_map_page(pp, S_READ);
+ bcopy(va, (char *)db->db_data + bufoff, thiscpy);
+ zfs_unmap_page(pp, va);
+ pp = pp->p_next;
+ bufoff += PAGESIZE;
+ }
+
+ if (tocpy == db->db_size)
+ dmu_buf_fill_done(db, tx);
+
+ if (err)
+ break;
+
+ offset += tocpy;
+ size -= tocpy;
+ }
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+ return (err);
+}
+#endif
+
+typedef struct {
+ dbuf_dirty_record_t *dr;
+ dmu_sync_cb_t *done;
+ void *arg;
+} dmu_sync_arg_t;
+
+/* ARGSUSED */
+static void
+dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg)
+{
+ blkptr_t *bp = zio->io_bp;
+
+ if (!BP_IS_HOLE(bp)) {
+ dmu_sync_arg_t *in = varg;
+ dbuf_dirty_record_t *dr = in->dr;
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ ASSERT(BP_GET_TYPE(bp) == db->db_dnode->dn_type);
+ ASSERT(BP_GET_LEVEL(bp) == 0);
+ bp->blk_fill = 1;
+ }
+}
+
+/* ARGSUSED */
+static void
+dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg)
+{
+ dmu_sync_arg_t *in = varg;
+ dbuf_dirty_record_t *dr = in->dr;
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ dmu_sync_cb_t *done = in->done;
+
+ mutex_enter(&db->db_mtx);
+ ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
+ dr->dt.dl.dr_overridden_by = *zio->io_bp; /* structure assignment */
+ dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
+ cv_broadcast(&db->db_changed);
+ mutex_exit(&db->db_mtx);
+
+ if (done)
+ done(&(db->db), in->arg);
+
+ kmem_free(in, sizeof (dmu_sync_arg_t));
+}
+
+/*
+ * Intent log support: sync the block associated with db to disk.
+ * N.B. and XXX: the caller is responsible for making sure that the
+ * data isn't changing while dmu_sync() is writing it.
+ *
+ * Return values:
+ *
+ * EEXIST: this txg has already been synced, so there's nothing to to.
+ * The caller should not log the write.
+ *
+ * ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
+ * The caller should not log the write.
+ *
+ * EALREADY: this block is already in the process of being synced.
+ * The caller should track its progress (somehow).
+ *
+ * EINPROGRESS: the IO has been initiated.
+ * The caller should log this blkptr in the callback.
+ *
+ * 0: completed. Sets *bp to the blkptr just written.
+ * The caller should log this blkptr immediately.
+ */
+int
+dmu_sync(zio_t *pio, dmu_buf_t *db_fake,
+ blkptr_t *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+ objset_impl_t *os = db->db_objset;
+ dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
+ tx_state_t *tx = &dp->dp_tx;
+ dbuf_dirty_record_t *dr;
+ dmu_sync_arg_t *in;
+ zbookmark_t zb;
+ writeprops_t wp = { 0 };
+ zio_t *zio;
+ int err;
+
+ ASSERT(BP_IS_HOLE(bp));
+ ASSERT(txg != 0);
+
+ dprintf("dmu_sync txg=%llu, s,o,q %llu %llu %llu\n",
+ txg, tx->tx_synced_txg, tx->tx_open_txg, tx->tx_quiesced_txg);
+
+ /*
+ * XXX - would be nice if we could do this without suspending...
+ */
+ txg_suspend(dp);
+
+ /*
+ * If this txg already synced, there's nothing to do.
+ */
+ if (txg <= tx->tx_synced_txg) {
+ txg_resume(dp);
+ /*
+ * If we're running ziltest, we need the blkptr regardless.
+ */
+ if (txg > spa_freeze_txg(dp->dp_spa)) {
+ /* if db_blkptr == NULL, this was an empty write */
+ if (db->db_blkptr)
+ *bp = *db->db_blkptr; /* structure assignment */
+ return (0);
+ }
+ return (EEXIST);
+ }
+
+ mutex_enter(&db->db_mtx);
+
+ if (txg == tx->tx_syncing_txg) {
+ while (db->db_data_pending) {
+ /*
+ * IO is in-progress. Wait for it to finish.
+ * XXX - would be nice to be able to somehow "attach"
+ * this zio to the parent zio passed in.
+ */
+ cv_wait(&db->db_changed, &db->db_mtx);
+ if (!db->db_data_pending &&
+ db->db_blkptr && BP_IS_HOLE(db->db_blkptr)) {
+ /*
+ * IO was compressed away
+ */
+ *bp = *db->db_blkptr; /* structure assignment */
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+ return (0);
+ }
+ ASSERT(db->db_data_pending ||
+ (db->db_blkptr && db->db_blkptr->blk_birth == txg));
+ }
+
+ if (db->db_blkptr && db->db_blkptr->blk_birth == txg) {
+ /*
+ * IO is already completed.
+ */
+ *bp = *db->db_blkptr; /* structure assignment */
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+ return (0);
+ }
+ }
+
+ dr = db->db_last_dirty;
+ while (dr && dr->dr_txg > txg)
+ dr = dr->dr_next;
+ if (dr == NULL || dr->dr_txg < txg) {
+ /*
+ * This dbuf isn't dirty, must have been free_range'd.
+ * There's no need to log writes to freed blocks, so we're done.
+ */
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+ return (ENOENT);
+ }
+
+ ASSERT(dr->dr_txg == txg);
+ if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
+ /*
+ * We have already issued a sync write for this buffer.
+ */
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+ return (EALREADY);
+ } else if (dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
+ /*
+ * This buffer has already been synced. It could not
+ * have been dirtied since, or we would have cleared the state.
+ */
+ *bp = dr->dt.dl.dr_overridden_by; /* structure assignment */
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+ return (0);
+ }
+
+ dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
+ in = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
+ in->dr = dr;
+ in->done = done;
+ in->arg = arg;
+ mutex_exit(&db->db_mtx);
+ txg_resume(dp);
+
+ zb.zb_objset = os->os_dsl_dataset->ds_object;
+ zb.zb_object = db->db.db_object;
+ zb.zb_level = db->db_level;
+ zb.zb_blkid = db->db_blkid;
+
+ wp.wp_type = db->db_dnode->dn_type;
+ wp.wp_level = db->db_level;
+ wp.wp_copies = os->os_copies;
+ wp.wp_dnchecksum = db->db_dnode->dn_checksum;
+ wp.wp_oschecksum = os->os_checksum;
+ wp.wp_dncompress = db->db_dnode->dn_compress;
+ wp.wp_oscompress = os->os_compress;
+
+ ASSERT(BP_IS_HOLE(bp));
+
+ zio = arc_write(pio, os->os_spa, &wp, DBUF_IS_L2CACHEABLE(db),
+ txg, bp, dr->dt.dl.dr_data, dmu_sync_ready, dmu_sync_done, in,
+ ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
+ if (pio) {
+ zio_nowait(zio);
+ err = EINPROGRESS;
+ } else {
+ err = zio_wait(zio);
+ ASSERT(err == 0);
+ }
+ return (err);
+}
+
+int
+dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
+ dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ err = dnode_set_blksz(dn, size, ibs, tx);
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+void
+dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
+ dmu_tx_t *tx)
+{
+ dnode_t *dn;
+
+ /* XXX assumes dnode_hold will not get an i/o error */
+ (void) dnode_hold(os->os, object, FTAG, &dn);
+ ASSERT(checksum < ZIO_CHECKSUM_FUNCTIONS);
+ dn->dn_checksum = checksum;
+ dnode_setdirty(dn, tx);
+ dnode_rele(dn, FTAG);
+}
+
+void
+dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
+ dmu_tx_t *tx)
+{
+ dnode_t *dn;
+
+ /* XXX assumes dnode_hold will not get an i/o error */
+ (void) dnode_hold(os->os, object, FTAG, &dn);
+ ASSERT(compress < ZIO_COMPRESS_FUNCTIONS);
+ dn->dn_compress = compress;
+ dnode_setdirty(dn, tx);
+ dnode_rele(dn, FTAG);
+}
+
+int
+dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
+{
+ dnode_t *dn;
+ int i, err;
+
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ /*
+ * Sync any current changes before
+ * we go trundling through the block pointers.
+ */
+ for (i = 0; i < TXG_SIZE; i++) {
+ if (list_link_active(&dn->dn_dirty_link[i]))
+ break;
+ }
+ if (i != TXG_SIZE) {
+ dnode_rele(dn, FTAG);
+ txg_wait_synced(dmu_objset_pool(os), 0);
+ err = dnode_hold(os->os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ }
+
+ err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
+ dnode_rele(dn, FTAG);
+
+ return (err);
+}
+
+void
+dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
+{
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ mutex_enter(&dn->dn_mtx);
+
+ doi->doi_data_block_size = dn->dn_datablksz;
+ doi->doi_metadata_block_size = dn->dn_indblkshift ?
+ 1ULL << dn->dn_indblkshift : 0;
+ doi->doi_indirection = dn->dn_nlevels;
+ doi->doi_checksum = dn->dn_checksum;
+ doi->doi_compress = dn->dn_compress;
+ doi->doi_physical_blks = (DN_USED_BYTES(dn->dn_phys) +
+ SPA_MINBLOCKSIZE/2) >> SPA_MINBLOCKSHIFT;
+ doi->doi_max_block_offset = dn->dn_phys->dn_maxblkid;
+ doi->doi_type = dn->dn_type;
+ doi->doi_bonus_size = dn->dn_bonuslen;
+ doi->doi_bonus_type = dn->dn_bonustype;
+
+ mutex_exit(&dn->dn_mtx);
+ rw_exit(&dn->dn_struct_rwlock);
+}
+
+/*
+ * Get information on a DMU object.
+ * If doi is NULL, just indicates whether the object exists.
+ */
+int
+dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi)
+{
+ dnode_t *dn;
+ int err = dnode_hold(os->os, object, FTAG, &dn);
+
+ if (err)
+ return (err);
+
+ if (doi != NULL)
+ dmu_object_info_from_dnode(dn, doi);
+
+ dnode_rele(dn, FTAG);
+ return (0);
+}
+
+/*
+ * As above, but faster; can be used when you have a held dbuf in hand.
+ */
+void
+dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi)
+{
+ dmu_object_info_from_dnode(((dmu_buf_impl_t *)db)->db_dnode, doi);
+}
+
+/*
+ * Faster still when you only care about the size.
+ * This is specifically optimized for zfs_getattr().
+ */
+void
+dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512)
+{
+ dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
+
+ *blksize = dn->dn_datablksz;
+ /* add 1 for dnode space */
+ *nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
+ SPA_MINBLOCKSHIFT) + 1;
+}
+
+void
+byteswap_uint64_array(void *vbuf, size_t size)
+{
+ uint64_t *buf = vbuf;
+ size_t count = size >> 3;
+ int i;
+
+ ASSERT((size & 7) == 0);
+
+ for (i = 0; i < count; i++)
+ buf[i] = BSWAP_64(buf[i]);
+}
+
+void
+byteswap_uint32_array(void *vbuf, size_t size)
+{
+ uint32_t *buf = vbuf;
+ size_t count = size >> 2;
+ int i;
+
+ ASSERT((size & 3) == 0);
+
+ for (i = 0; i < count; i++)
+ buf[i] = BSWAP_32(buf[i]);
+}
+
+void
+byteswap_uint16_array(void *vbuf, size_t size)
+{
+ uint16_t *buf = vbuf;
+ size_t count = size >> 1;
+ int i;
+
+ ASSERT((size & 1) == 0);
+
+ for (i = 0; i < count; i++)
+ buf[i] = BSWAP_16(buf[i]);
+}
+
+/* ARGSUSED */
+void
+byteswap_uint8_array(void *vbuf, size_t size)
+{
+}
+
+void
+dmu_init(void)
+{
+ dbuf_init();
+ dnode_init();
+ arc_init();
+ l2arc_init();
+}
+
+void
+dmu_fini(void)
+{
+ arc_fini();
+ dnode_fini();
+ dbuf_fini();
+ l2arc_fini();
+}
diff --git a/module/zfs/dmu_object.c b/module/zfs/dmu_object.c
new file mode 100644
index 000000000..1b9247d66
--- /dev/null
+++ b/module/zfs/dmu_object.c
@@ -0,0 +1,162 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_tx.h>
+#include <sys/dnode.h>
+
+uint64_t
+dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ objset_impl_t *osi = os->os;
+ uint64_t object;
+ uint64_t L2_dnode_count = DNODES_PER_BLOCK <<
+ (osi->os_meta_dnode->dn_indblkshift - SPA_BLKPTRSHIFT);
+ dnode_t *dn = NULL;
+ int restarted = B_FALSE;
+
+ mutex_enter(&osi->os_obj_lock);
+ for (;;) {
+ object = osi->os_obj_next;
+ /*
+ * Each time we polish off an L2 bp worth of dnodes
+ * (2^13 objects), move to another L2 bp that's still
+ * reasonably sparse (at most 1/4 full). Look from the
+ * beginning once, but after that keep looking from here.
+ * If we can't find one, just keep going from here.
+ */
+ if (P2PHASE(object, L2_dnode_count) == 0) {
+ uint64_t offset = restarted ? object << DNODE_SHIFT : 0;
+ int error = dnode_next_offset(osi->os_meta_dnode,
+ DNODE_FIND_HOLE,
+ &offset, 2, DNODES_PER_BLOCK >> 2, 0);
+ restarted = B_TRUE;
+ if (error == 0)
+ object = offset >> DNODE_SHIFT;
+ }
+ osi->os_obj_next = ++object;
+
+ /*
+ * XXX We should check for an i/o error here and return
+ * up to our caller. Actually we should pre-read it in
+ * dmu_tx_assign(), but there is currently no mechanism
+ * to do so.
+ */
+ (void) dnode_hold_impl(os->os, object, DNODE_MUST_BE_FREE,
+ FTAG, &dn);
+ if (dn)
+ break;
+
+ if (dmu_object_next(os, &object, B_TRUE, 0) == 0)
+ osi->os_obj_next = object - 1;
+ }
+
+ dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, tx);
+ dnode_rele(dn, FTAG);
+
+ mutex_exit(&osi->os_obj_lock);
+
+ dmu_tx_add_new_object(tx, os, object);
+ return (object);
+}
+
+int
+dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
+ int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
+ return (EBADF);
+
+ err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_FREE, FTAG, &dn);
+ if (err)
+ return (err);
+ dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, tx);
+ dnode_rele(dn, FTAG);
+
+ dmu_tx_add_new_object(tx, os, object);
+ return (0);
+}
+
+int
+dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
+ int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
+ return (EBADF);
+
+ err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED,
+ FTAG, &dn);
+ if (err)
+ return (err);
+ dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, tx);
+ dnode_rele(dn, FTAG);
+
+ return (0);
+}
+
+int
+dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ ASSERT(object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));
+
+ err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED,
+ FTAG, &dn);
+ if (err)
+ return (err);
+
+ ASSERT(dn->dn_type != DMU_OT_NONE);
+ dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
+ dnode_free(dn, tx);
+ dnode_rele(dn, FTAG);
+
+ return (0);
+}
+
+int
+dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg)
+{
+ uint64_t offset = (*objectp + 1) << DNODE_SHIFT;
+ int error;
+
+ error = dnode_next_offset(os->os->os_meta_dnode,
+ (hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg);
+
+ *objectp = offset >> DNODE_SHIFT;
+
+ return (error);
+}
diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
new file mode 100644
index 000000000..7981e0682
--- /dev/null
+++ b/module/zfs/dmu_objset.c
@@ -0,0 +1,1228 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/cred.h>
+#include <sys/zfs_context.h>
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dsl_deleg.h>
+#include <sys/dnode.h>
+#include <sys/dbuf.h>
+#include <sys/zvol.h>
+#include <sys/dmu_tx.h>
+#include <sys/zio_checksum.h>
+#include <sys/zap.h>
+#include <sys/zil.h>
+#include <sys/dmu_impl.h>
+#include <sys/zfs_ioctl.h>
+
+spa_t *
+dmu_objset_spa(objset_t *os)
+{
+ return (os->os->os_spa);
+}
+
+zilog_t *
+dmu_objset_zil(objset_t *os)
+{
+ return (os->os->os_zil);
+}
+
+dsl_pool_t *
+dmu_objset_pool(objset_t *os)
+{
+ dsl_dataset_t *ds;
+
+ if ((ds = os->os->os_dsl_dataset) != NULL && ds->ds_dir)
+ return (ds->ds_dir->dd_pool);
+ else
+ return (spa_get_dsl(os->os->os_spa));
+}
+
+dsl_dataset_t *
+dmu_objset_ds(objset_t *os)
+{
+ return (os->os->os_dsl_dataset);
+}
+
+dmu_objset_type_t
+dmu_objset_type(objset_t *os)
+{
+ return (os->os->os_phys->os_type);
+}
+
+void
+dmu_objset_name(objset_t *os, char *buf)
+{
+ dsl_dataset_name(os->os->os_dsl_dataset, buf);
+}
+
+uint64_t
+dmu_objset_id(objset_t *os)
+{
+ dsl_dataset_t *ds = os->os->os_dsl_dataset;
+
+ return (ds ? ds->ds_object : 0);
+}
+
+static void
+checksum_changed_cb(void *arg, uint64_t newval)
+{
+ objset_impl_t *osi = arg;
+
+ /*
+ * Inheritance should have been done by now.
+ */
+ ASSERT(newval != ZIO_CHECKSUM_INHERIT);
+
+ osi->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE);
+}
+
+static void
+compression_changed_cb(void *arg, uint64_t newval)
+{
+ objset_impl_t *osi = arg;
+
+ /*
+ * Inheritance and range checking should have been done by now.
+ */
+ ASSERT(newval != ZIO_COMPRESS_INHERIT);
+
+ osi->os_compress = zio_compress_select(newval, ZIO_COMPRESS_ON_VALUE);
+}
+
+static void
+copies_changed_cb(void *arg, uint64_t newval)
+{
+ objset_impl_t *osi = arg;
+
+ /*
+ * Inheritance and range checking should have been done by now.
+ */
+ ASSERT(newval > 0);
+ ASSERT(newval <= spa_max_replication(osi->os_spa));
+
+ osi->os_copies = newval;
+}
+
+static void
+primary_cache_changed_cb(void *arg, uint64_t newval)
+{
+ objset_impl_t *osi = arg;
+
+ /*
+ * Inheritance and range checking should have been done by now.
+ */
+ ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE ||
+ newval == ZFS_CACHE_METADATA);
+
+ osi->os_primary_cache = newval;
+}
+
+static void
+secondary_cache_changed_cb(void *arg, uint64_t newval)
+{
+ objset_impl_t *osi = arg;
+
+ /*
+ * Inheritance and range checking should have been done by now.
+ */
+ ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE ||
+ newval == ZFS_CACHE_METADATA);
+
+ osi->os_secondary_cache = newval;
+}
+
+void
+dmu_objset_byteswap(void *buf, size_t size)
+{
+ objset_phys_t *osp = buf;
+
+ ASSERT(size == sizeof (objset_phys_t));
+ dnode_byteswap(&osp->os_meta_dnode);
+ byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t));
+ osp->os_type = BSWAP_64(osp->os_type);
+}
+
+int
+dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
+ objset_impl_t **osip)
+{
+ objset_impl_t *osi;
+ int i, err;
+
+ ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
+
+ osi = kmem_zalloc(sizeof (objset_impl_t), KM_SLEEP);
+ osi->os.os = osi;
+ osi->os_dsl_dataset = ds;
+ osi->os_spa = spa;
+ osi->os_rootbp = bp;
+ if (!BP_IS_HOLE(osi->os_rootbp)) {
+ uint32_t aflags = ARC_WAIT;
+ zbookmark_t zb;
+ zb.zb_objset = ds ? ds->ds_object : 0;
+ zb.zb_object = 0;
+ zb.zb_level = -1;
+ zb.zb_blkid = 0;
+ if (DMU_OS_IS_L2CACHEABLE(osi))
+ aflags |= ARC_L2CACHE;
+
+ dprintf_bp(osi->os_rootbp, "reading %s", "");
+ /*
+ * NB: when bprewrite scrub can change the bp,
+ * and this is called from dmu_objset_open_ds_os, the bp
+ * could change, and we'll need a lock.
+ */
+ err = arc_read_nolock(NULL, spa, osi->os_rootbp,
+ arc_getbuf_func, &osi->os_phys_buf,
+ ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb);
+ if (err) {
+ kmem_free(osi, sizeof (objset_impl_t));
+ /* convert checksum errors into IO errors */
+ if (err == ECKSUM)
+ err = EIO;
+ return (err);
+ }
+ osi->os_phys = osi->os_phys_buf->b_data;
+ } else {
+ osi->os_phys_buf = arc_buf_alloc(spa, sizeof (objset_phys_t),
+ &osi->os_phys_buf, ARC_BUFC_METADATA);
+ osi->os_phys = osi->os_phys_buf->b_data;
+ bzero(osi->os_phys, sizeof (objset_phys_t));
+ }
+
+ /*
+ * Note: the changed_cb will be called once before the register
+ * func returns, thus changing the checksum/compression from the
+ * default (fletcher2/off). Snapshots don't need to know about
+ * checksum/compression/copies.
+ */
+ if (ds) {
+ err = dsl_prop_register(ds, "primarycache",
+ primary_cache_changed_cb, osi);
+ if (err == 0)
+ err = dsl_prop_register(ds, "secondarycache",
+ secondary_cache_changed_cb, osi);
+ if (!dsl_dataset_is_snapshot(ds)) {
+ if (err == 0)
+ err = dsl_prop_register(ds, "checksum",
+ checksum_changed_cb, osi);
+ if (err == 0)
+ err = dsl_prop_register(ds, "compression",
+ compression_changed_cb, osi);
+ if (err == 0)
+ err = dsl_prop_register(ds, "copies",
+ copies_changed_cb, osi);
+ }
+ if (err) {
+ VERIFY(arc_buf_remove_ref(osi->os_phys_buf,
+ &osi->os_phys_buf) == 1);
+ kmem_free(osi, sizeof (objset_impl_t));
+ return (err);
+ }
+ } else if (ds == NULL) {
+ /* It's the meta-objset. */
+ osi->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
+ osi->os_compress = ZIO_COMPRESS_LZJB;
+ osi->os_copies = spa_max_replication(spa);
+ osi->os_primary_cache = ZFS_CACHE_ALL;
+ osi->os_secondary_cache = ZFS_CACHE_ALL;
+ }
+
+ osi->os_zil_header = osi->os_phys->os_zil_header;
+ osi->os_zil = zil_alloc(&osi->os, &osi->os_zil_header);
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ list_create(&osi->os_dirty_dnodes[i], sizeof (dnode_t),
+ offsetof(dnode_t, dn_dirty_link[i]));
+ list_create(&osi->os_free_dnodes[i], sizeof (dnode_t),
+ offsetof(dnode_t, dn_dirty_link[i]));
+ }
+ list_create(&osi->os_dnodes, sizeof (dnode_t),
+ offsetof(dnode_t, dn_link));
+ list_create(&osi->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
+ offsetof(dmu_buf_impl_t, db_link));
+
+ mutex_init(&osi->os_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&osi->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&osi->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ osi->os_meta_dnode = dnode_special_open(osi,
+ &osi->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT);
+
+ /*
+ * We should be the only thread trying to do this because we
+ * have ds_opening_lock
+ */
+ if (ds) {
+ VERIFY(NULL == dsl_dataset_set_user_ptr(ds, osi,
+ dmu_objset_evict));
+ }
+
+ *osip = osi;
+ return (0);
+}
+
+static int
+dmu_objset_open_ds_os(dsl_dataset_t *ds, objset_t *os, dmu_objset_type_t type)
+{
+ objset_impl_t *osi;
+
+ mutex_enter(&ds->ds_opening_lock);
+ osi = dsl_dataset_get_user_ptr(ds);
+ if (osi == NULL) {
+ int err;
+
+ err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
+ ds, &ds->ds_phys->ds_bp, &osi);
+ if (err) {
+ mutex_exit(&ds->ds_opening_lock);
+ return (err);
+ }
+ }
+ mutex_exit(&ds->ds_opening_lock);
+
+ os->os = osi;
+ os->os_mode = DS_MODE_NOHOLD;
+
+ if (type != DMU_OST_ANY && type != os->os->os_phys->os_type)
+ return (EINVAL);
+ return (0);
+}
+
+int
+dmu_objset_open_ds(dsl_dataset_t *ds, dmu_objset_type_t type, objset_t **osp)
+{
+ objset_t *os;
+ int err;
+
+ os = kmem_alloc(sizeof (objset_t), KM_SLEEP);
+ err = dmu_objset_open_ds_os(ds, os, type);
+ if (err)
+ kmem_free(os, sizeof (objset_t));
+ else
+ *osp = os;
+ return (err);
+}
+
+/* called from zpl */
+int
+dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
+ objset_t **osp)
+{
+ objset_t *os;
+ dsl_dataset_t *ds;
+ int err;
+
+ ASSERT(DS_MODE_TYPE(mode) == DS_MODE_USER ||
+ DS_MODE_TYPE(mode) == DS_MODE_OWNER);
+
+ os = kmem_alloc(sizeof (objset_t), KM_SLEEP);
+ if (DS_MODE_TYPE(mode) == DS_MODE_USER)
+ err = dsl_dataset_hold(name, os, &ds);
+ else
+ err = dsl_dataset_own(name, mode, os, &ds);
+ if (err) {
+ kmem_free(os, sizeof (objset_t));
+ return (err);
+ }
+
+ err = dmu_objset_open_ds_os(ds, os, type);
+ if (err) {
+ if (DS_MODE_TYPE(mode) == DS_MODE_USER)
+ dsl_dataset_rele(ds, os);
+ else
+ dsl_dataset_disown(ds, os);
+ kmem_free(os, sizeof (objset_t));
+ } else {
+ os->os_mode = mode;
+ *osp = os;
+ }
+ return (err);
+}
+
+void
+dmu_objset_close(objset_t *os)
+{
+ ASSERT(DS_MODE_TYPE(os->os_mode) == DS_MODE_USER ||
+ DS_MODE_TYPE(os->os_mode) == DS_MODE_OWNER ||
+ DS_MODE_TYPE(os->os_mode) == DS_MODE_NOHOLD);
+
+ if (DS_MODE_TYPE(os->os_mode) == DS_MODE_USER)
+ dsl_dataset_rele(os->os->os_dsl_dataset, os);
+ else if (DS_MODE_TYPE(os->os_mode) == DS_MODE_OWNER)
+ dsl_dataset_disown(os->os->os_dsl_dataset, os);
+ kmem_free(os, sizeof (objset_t));
+}
+
+int
+dmu_objset_evict_dbufs(objset_t *os)
+{
+ objset_impl_t *osi = os->os;
+ dnode_t *dn;
+
+ mutex_enter(&osi->os_lock);
+
+ /* process the mdn last, since the other dnodes have holds on it */
+ list_remove(&osi->os_dnodes, osi->os_meta_dnode);
+ list_insert_tail(&osi->os_dnodes, osi->os_meta_dnode);
+
+ /*
+ * Find the first dnode with holds. We have to do this dance
+ * because dnode_add_ref() only works if you already have a
+ * hold. If there are no holds then it has no dbufs so OK to
+ * skip.
+ */
+ for (dn = list_head(&osi->os_dnodes);
+ dn && !dnode_add_ref(dn, FTAG);
+ dn = list_next(&osi->os_dnodes, dn))
+ continue;
+
+ while (dn) {
+ dnode_t *next_dn = dn;
+
+ do {
+ next_dn = list_next(&osi->os_dnodes, next_dn);
+ } while (next_dn && !dnode_add_ref(next_dn, FTAG));
+
+ mutex_exit(&osi->os_lock);
+ dnode_evict_dbufs(dn);
+ dnode_rele(dn, FTAG);
+ mutex_enter(&osi->os_lock);
+ dn = next_dn;
+ }
+ mutex_exit(&osi->os_lock);
+ return (list_head(&osi->os_dnodes) != osi->os_meta_dnode);
+}
+
+void
+dmu_objset_evict(dsl_dataset_t *ds, void *arg)
+{
+ objset_impl_t *osi = arg;
+ objset_t os;
+ int i;
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ ASSERT(list_head(&osi->os_dirty_dnodes[i]) == NULL);
+ ASSERT(list_head(&osi->os_free_dnodes[i]) == NULL);
+ }
+
+ if (ds) {
+ if (!dsl_dataset_is_snapshot(ds)) {
+ VERIFY(0 == dsl_prop_unregister(ds, "checksum",
+ checksum_changed_cb, osi));
+ VERIFY(0 == dsl_prop_unregister(ds, "compression",
+ compression_changed_cb, osi));
+ VERIFY(0 == dsl_prop_unregister(ds, "copies",
+ copies_changed_cb, osi));
+ }
+ VERIFY(0 == dsl_prop_unregister(ds, "primarycache",
+ primary_cache_changed_cb, osi));
+ VERIFY(0 == dsl_prop_unregister(ds, "secondarycache",
+ secondary_cache_changed_cb, osi));
+ }
+
+ /*
+ * We should need only a single pass over the dnode list, since
+ * nothing can be added to the list at this point.
+ */
+ os.os = osi;
+ (void) dmu_objset_evict_dbufs(&os);
+
+ ASSERT3P(list_head(&osi->os_dnodes), ==, osi->os_meta_dnode);
+ ASSERT3P(list_tail(&osi->os_dnodes), ==, osi->os_meta_dnode);
+ ASSERT3P(list_head(&osi->os_meta_dnode->dn_dbufs), ==, NULL);
+
+ dnode_special_close(osi->os_meta_dnode);
+ zil_free(osi->os_zil);
+
+ VERIFY(arc_buf_remove_ref(osi->os_phys_buf, &osi->os_phys_buf) == 1);
+ mutex_destroy(&osi->os_lock);
+ mutex_destroy(&osi->os_obj_lock);
+ mutex_destroy(&osi->os_user_ptr_lock);
+ kmem_free(osi, sizeof (objset_impl_t));
+}
+
+/* called from dsl for meta-objset */
+objset_impl_t *
+dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
+ dmu_objset_type_t type, dmu_tx_t *tx)
+{
+ objset_impl_t *osi;
+ dnode_t *mdn;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ if (ds)
+ mutex_enter(&ds->ds_opening_lock);
+ VERIFY(0 == dmu_objset_open_impl(spa, ds, bp, &osi));
+ if (ds)
+ mutex_exit(&ds->ds_opening_lock);
+ mdn = osi->os_meta_dnode;
+
+ dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT,
+ DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx);
+
+ /*
+ * We don't want to have to increase the meta-dnode's nlevels
+ * later, because then we could do it in quescing context while
+ * we are also accessing it in open context.
+ *
+ * This precaution is not necessary for the MOS (ds == NULL),
+ * because the MOS is only updated in syncing context.
+ * This is most fortunate: the MOS is the only objset that
+ * needs to be synced multiple times as spa_sync() iterates
+ * to convergence, so minimizing its dn_nlevels matters.
+ */
+ if (ds != NULL) {
+ int levels = 1;
+
+ /*
+ * Determine the number of levels necessary for the meta-dnode
+ * to contain DN_MAX_OBJECT dnodes.
+ */
+ while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift +
+ (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
+ DN_MAX_OBJECT * sizeof (dnode_phys_t))
+ levels++;
+
+ mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =
+ mdn->dn_nlevels = levels;
+ }
+
+ ASSERT(type != DMU_OST_NONE);
+ ASSERT(type != DMU_OST_ANY);
+ ASSERT(type < DMU_OST_NUMTYPES);
+ osi->os_phys->os_type = type;
+
+ dsl_dataset_dirty(ds, tx);
+
+ return (osi);
+}
+
+struct oscarg {
+ void (*userfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
+ void *userarg;
+ dsl_dataset_t *clone_parent;
+ const char *lastname;
+ dmu_objset_type_t type;
+ uint64_t flags;
+};
+
+/*ARGSUSED*/
+static int
+dmu_objset_create_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct oscarg *oa = arg2;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ int err;
+ uint64_t ddobj;
+
+ err = zap_lookup(mos, dd->dd_phys->dd_child_dir_zapobj,
+ oa->lastname, sizeof (uint64_t), 1, &ddobj);
+ if (err != ENOENT)
+ return (err ? err : EEXIST);
+
+ if (oa->clone_parent != NULL) {
+ /*
+ * You can't clone across pools.
+ */
+ if (oa->clone_parent->ds_dir->dd_pool != dd->dd_pool)
+ return (EXDEV);
+
+ /*
+ * You can only clone snapshots, not the head datasets.
+ */
+ if (oa->clone_parent->ds_phys->ds_num_children == 0)
+ return (EINVAL);
+ }
+
+ return (0);
+}
+
+static void
+dmu_objset_create_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct oscarg *oa = arg2;
+ dsl_dataset_t *ds;
+ blkptr_t *bp;
+ uint64_t dsobj;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ dsobj = dsl_dataset_create_sync(dd, oa->lastname,
+ oa->clone_parent, oa->flags, cr, tx);
+
+ VERIFY(0 == dsl_dataset_hold_obj(dd->dd_pool, dsobj, FTAG, &ds));
+ bp = dsl_dataset_get_blkptr(ds);
+ if (BP_IS_HOLE(bp)) {
+ objset_impl_t *osi;
+
+ /* This is an empty dmu_objset; not a clone. */
+ osi = dmu_objset_create_impl(dsl_dataset_get_spa(ds),
+ ds, bp, oa->type, tx);
+
+ if (oa->userfunc)
+ oa->userfunc(&osi->os, oa->userarg, cr, tx);
+ }
+
+ spa_history_internal_log(LOG_DS_CREATE, dd->dd_pool->dp_spa,
+ tx, cr, "dataset = %llu", dsobj);
+
+ dsl_dataset_rele(ds, FTAG);
+}
+
+int
+dmu_objset_create(const char *name, dmu_objset_type_t type,
+ objset_t *clone_parent, uint64_t flags,
+ void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg)
+{
+ dsl_dir_t *pdd;
+ const char *tail;
+ int err = 0;
+ struct oscarg oa = { 0 };
+
+ ASSERT(strchr(name, '@') == NULL);
+ err = dsl_dir_open(name, FTAG, &pdd, &tail);
+ if (err)
+ return (err);
+ if (tail == NULL) {
+ dsl_dir_close(pdd, FTAG);
+ return (EEXIST);
+ }
+
+ dprintf("name=%s\n", name);
+
+ oa.userfunc = func;
+ oa.userarg = arg;
+ oa.lastname = tail;
+ oa.type = type;
+ oa.flags = flags;
+
+ if (clone_parent != NULL) {
+ /*
+ * You can't clone to a different type.
+ */
+ if (clone_parent->os->os_phys->os_type != type) {
+ dsl_dir_close(pdd, FTAG);
+ return (EINVAL);
+ }
+ oa.clone_parent = clone_parent->os->os_dsl_dataset;
+ }
+ err = dsl_sync_task_do(pdd->dd_pool, dmu_objset_create_check,
+ dmu_objset_create_sync, pdd, &oa, 5);
+ dsl_dir_close(pdd, FTAG);
+ return (err);
+}
+
+int
+dmu_objset_destroy(const char *name)
+{
+ objset_t *os;
+ int error;
+
+ /*
+ * If it looks like we'll be able to destroy it, and there's
+ * an unplayed replay log sitting around, destroy the log.
+ * It would be nicer to do this in dsl_dataset_destroy_sync(),
+ * but the replay log objset is modified in open context.
+ */
+ error = dmu_objset_open(name, DMU_OST_ANY,
+ DS_MODE_OWNER|DS_MODE_READONLY|DS_MODE_INCONSISTENT, &os);
+ if (error == 0) {
+ dsl_dataset_t *ds = os->os->os_dsl_dataset;
+ zil_destroy(dmu_objset_zil(os), B_FALSE);
+
+ error = dsl_dataset_destroy(ds, os);
+ /*
+ * dsl_dataset_destroy() closes the ds.
+ */
+ kmem_free(os, sizeof (objset_t));
+ }
+
+ return (error);
+}
+
+/*
+ * This will close the objset.
+ */
+int
+dmu_objset_rollback(objset_t *os)
+{
+ int err;
+ dsl_dataset_t *ds;
+
+ ds = os->os->os_dsl_dataset;
+
+ if (!dsl_dataset_tryown(ds, TRUE, os)) {
+ dmu_objset_close(os);
+ return (EBUSY);
+ }
+
+ err = dsl_dataset_rollback(ds, os->os->os_phys->os_type);
+
+ /*
+ * NB: we close the objset manually because the rollback
+ * actually implicitly called dmu_objset_evict(), thus freeing
+ * the objset_impl_t.
+ */
+ dsl_dataset_disown(ds, os);
+ kmem_free(os, sizeof (objset_t));
+ return (err);
+}
+
+struct snaparg {
+ dsl_sync_task_group_t *dstg;
+ char *snapname;
+ char failed[MAXPATHLEN];
+ boolean_t checkperms;
+ list_t objsets;
+};
+
+struct osnode {
+ list_node_t node;
+ objset_t *os;
+};
+
+static int
+dmu_objset_snapshot_one(char *name, void *arg)
+{
+ struct snaparg *sn = arg;
+ objset_t *os;
+ int err;
+
+ (void) strcpy(sn->failed, name);
+
+ /*
+ * Check permissions only when requested. This only applies when
+ * doing a recursive snapshot. The permission checks for the starting
+ * dataset have already been performed in zfs_secpolicy_snapshot()
+ */
+ if (sn->checkperms == B_TRUE &&
+ (err = zfs_secpolicy_snapshot_perms(name, CRED())))
+ return (err);
+
+ err = dmu_objset_open(name, DMU_OST_ANY, DS_MODE_USER, &os);
+ if (err != 0)
+ return (err);
+
+ /* If the objset is in an inconsistent state, return busy */
+ if (os->os->os_dsl_dataset->ds_phys->ds_flags & DS_FLAG_INCONSISTENT) {
+ dmu_objset_close(os);
+ return (EBUSY);
+ }
+
+ /*
+ * NB: we need to wait for all in-flight changes to get to disk,
+ * so that we snapshot those changes. zil_suspend does this as
+ * a side effect.
+ */
+ err = zil_suspend(dmu_objset_zil(os));
+ if (err == 0) {
+ struct osnode *osn;
+ dsl_sync_task_create(sn->dstg, dsl_dataset_snapshot_check,
+ dsl_dataset_snapshot_sync, os->os->os_dsl_dataset,
+ sn->snapname, 3);
+ osn = kmem_alloc(sizeof (struct osnode), KM_SLEEP);
+ osn->os = os;
+ list_insert_tail(&sn->objsets, osn);
+ } else {
+ dmu_objset_close(os);
+ }
+
+ return (err);
+}
+
+int
+dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive)
+{
+ dsl_sync_task_t *dst;
+ struct osnode *osn;
+ struct snaparg sn = { 0 };
+ spa_t *spa;
+ int err;
+
+ (void) strcpy(sn.failed, fsname);
+
+ err = spa_open(fsname, &spa, FTAG);
+ if (err)
+ return (err);
+
+ sn.dstg = dsl_sync_task_group_create(spa_get_dsl(spa));
+ sn.snapname = snapname;
+ list_create(&sn.objsets, sizeof (struct osnode),
+ offsetof(struct osnode, node));
+
+ if (recursive) {
+ sn.checkperms = B_TRUE;
+ err = dmu_objset_find(fsname,
+ dmu_objset_snapshot_one, &sn, DS_FIND_CHILDREN);
+ } else {
+ sn.checkperms = B_FALSE;
+ err = dmu_objset_snapshot_one(fsname, &sn);
+ }
+
+ if (err)
+ goto out;
+
+ err = dsl_sync_task_group_wait(sn.dstg);
+
+ for (dst = list_head(&sn.dstg->dstg_tasks); dst;
+ dst = list_next(&sn.dstg->dstg_tasks, dst)) {
+ dsl_dataset_t *ds = dst->dst_arg1;
+ if (dst->dst_err)
+ dsl_dataset_name(ds, sn.failed);
+ }
+
+out:
+ while (osn = list_head(&sn.objsets)) {
+ list_remove(&sn.objsets, osn);
+ zil_resume(dmu_objset_zil(osn->os));
+ dmu_objset_close(osn->os);
+ kmem_free(osn, sizeof (struct osnode));
+ }
+ list_destroy(&sn.objsets);
+
+ if (err)
+ (void) strcpy(fsname, sn.failed);
+ dsl_sync_task_group_destroy(sn.dstg);
+ spa_close(spa, FTAG);
+ return (err);
+}
+
+static void
+dmu_objset_sync_dnodes(list_t *list, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+
+ while (dn = list_head(list)) {
+ ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
+ ASSERT(dn->dn_dbuf->db_data_pending);
+ /*
+ * Initialize dn_zio outside dnode_sync()
+ * to accomodate meta-dnode
+ */
+ dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio;
+ ASSERT(dn->dn_zio);
+
+ ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS);
+ list_remove(list, dn);
+ dnode_sync(dn, tx);
+ }
+}
+
+/* ARGSUSED */
+static void
+ready(zio_t *zio, arc_buf_t *abuf, void *arg)
+{
+ blkptr_t *bp = zio->io_bp;
+ blkptr_t *bp_orig = &zio->io_bp_orig;
+ objset_impl_t *os = arg;
+ dnode_phys_t *dnp = &os->os_phys->os_meta_dnode;
+
+ ASSERT(bp == os->os_rootbp);
+ ASSERT(BP_GET_TYPE(bp) == DMU_OT_OBJSET);
+ ASSERT(BP_GET_LEVEL(bp) == 0);
+
+ /*
+ * Update rootbp fill count.
+ */
+ bp->blk_fill = 1; /* count the meta-dnode */
+ for (int i = 0; i < dnp->dn_nblkptr; i++)
+ bp->blk_fill += dnp->dn_blkptr[i].blk_fill;
+
+ if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
+ ASSERT(DVA_EQUAL(BP_IDENTITY(bp), BP_IDENTITY(bp_orig)));
+ } else {
+ if (zio->io_bp_orig.blk_birth == os->os_synctx->tx_txg)
+ (void) dsl_dataset_block_kill(os->os_dsl_dataset,
+ &zio->io_bp_orig, zio, os->os_synctx);
+ dsl_dataset_block_born(os->os_dsl_dataset, bp, os->os_synctx);
+ }
+}
+
+/* called from dsl */
+void
+dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
+{
+ int txgoff;
+ zbookmark_t zb;
+ writeprops_t wp = { 0 };
+ zio_t *zio;
+ list_t *list;
+ dbuf_dirty_record_t *dr;
+
+ dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ /* XXX the write_done callback should really give us the tx... */
+ os->os_synctx = tx;
+
+ if (os->os_dsl_dataset == NULL) {
+ /*
+ * This is the MOS. If we have upgraded,
+ * spa_max_replication() could change, so reset
+ * os_copies here.
+ */
+ os->os_copies = spa_max_replication(os->os_spa);
+ }
+
+ /*
+ * Create the root block IO
+ */
+ zb.zb_objset = os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : 0;
+ zb.zb_object = 0;
+ zb.zb_level = -1; /* for block ordering; it's level 0 on disk */
+ zb.zb_blkid = 0;
+
+ wp.wp_type = DMU_OT_OBJSET;
+ wp.wp_level = 0; /* on-disk BP level; see above */
+ wp.wp_copies = os->os_copies;
+ wp.wp_oschecksum = os->os_checksum;
+ wp.wp_oscompress = os->os_compress;
+
+ if (BP_IS_OLDER(os->os_rootbp, tx->tx_txg)) {
+ (void) dsl_dataset_block_kill(os->os_dsl_dataset,
+ os->os_rootbp, pio, tx);
+ }
+
+ arc_release(os->os_phys_buf, &os->os_phys_buf);
+ zio = arc_write(pio, os->os_spa, &wp, DMU_OS_IS_L2CACHEABLE(os),
+ tx->tx_txg, os->os_rootbp, os->os_phys_buf, ready, NULL, os,
+ ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
+
+ /*
+ * Sync meta-dnode - the parent IO for the sync is the root block
+ */
+ os->os_meta_dnode->dn_zio = zio;
+ dnode_sync(os->os_meta_dnode, tx);
+
+ txgoff = tx->tx_txg & TXG_MASK;
+
+ dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], tx);
+ dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], tx);
+
+ list = &os->os_meta_dnode->dn_dirty_records[txgoff];
+ while (dr = list_head(list)) {
+ ASSERT(dr->dr_dbuf->db_level == 0);
+ list_remove(list, dr);
+ if (dr->dr_zio)
+ zio_nowait(dr->dr_zio);
+ }
+ /*
+ * Free intent log blocks up to this tx.
+ */
+ zil_sync(os->os_zil, tx);
+ os->os_phys->os_zil_header = os->os_zil_header;
+ zio_nowait(zio);
+}
+
+void
+dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
+ uint64_t *usedobjsp, uint64_t *availobjsp)
+{
+ dsl_dataset_space(os->os->os_dsl_dataset, refdbytesp, availbytesp,
+ usedobjsp, availobjsp);
+}
+
+uint64_t
+dmu_objset_fsid_guid(objset_t *os)
+{
+ return (dsl_dataset_fsid_guid(os->os->os_dsl_dataset));
+}
+
+void
+dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat)
+{
+ stat->dds_type = os->os->os_phys->os_type;
+ if (os->os->os_dsl_dataset)
+ dsl_dataset_fast_stat(os->os->os_dsl_dataset, stat);
+}
+
+void
+dmu_objset_stats(objset_t *os, nvlist_t *nv)
+{
+ ASSERT(os->os->os_dsl_dataset ||
+ os->os->os_phys->os_type == DMU_OST_META);
+
+ if (os->os->os_dsl_dataset != NULL)
+ dsl_dataset_stats(os->os->os_dsl_dataset, nv);
+
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE,
+ os->os->os_phys->os_type);
+}
+
+int
+dmu_objset_is_snapshot(objset_t *os)
+{
+ if (os->os->os_dsl_dataset != NULL)
+ return (dsl_dataset_is_snapshot(os->os->os_dsl_dataset));
+ else
+ return (B_FALSE);
+}
+
+int
+dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen,
+ boolean_t *conflict)
+{
+ dsl_dataset_t *ds = os->os->os_dsl_dataset;
+ uint64_t ignored;
+
+ if (ds->ds_phys->ds_snapnames_zapobj == 0)
+ return (ENOENT);
+
+ return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset,
+ ds->ds_phys->ds_snapnames_zapobj, name, 8, 1, &ignored, MT_FIRST,
+ real, maxlen, conflict));
+}
+
+int
+dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
+ uint64_t *idp, uint64_t *offp, boolean_t *case_conflict)
+{
+ dsl_dataset_t *ds = os->os->os_dsl_dataset;
+ zap_cursor_t cursor;
+ zap_attribute_t attr;
+
+ if (ds->ds_phys->ds_snapnames_zapobj == 0)
+ return (ENOENT);
+
+ zap_cursor_init_serialized(&cursor,
+ ds->ds_dir->dd_pool->dp_meta_objset,
+ ds->ds_phys->ds_snapnames_zapobj, *offp);
+
+ if (zap_cursor_retrieve(&cursor, &attr) != 0) {
+ zap_cursor_fini(&cursor);
+ return (ENOENT);
+ }
+
+ if (strlen(attr.za_name) + 1 > namelen) {
+ zap_cursor_fini(&cursor);
+ return (ENAMETOOLONG);
+ }
+
+ (void) strcpy(name, attr.za_name);
+ if (idp)
+ *idp = attr.za_first_integer;
+ if (case_conflict)
+ *case_conflict = attr.za_normalization_conflict;
+ zap_cursor_advance(&cursor);
+ *offp = zap_cursor_serialize(&cursor);
+ zap_cursor_fini(&cursor);
+
+ return (0);
+}
+
+int
+dmu_dir_list_next(objset_t *os, int namelen, char *name,
+ uint64_t *idp, uint64_t *offp)
+{
+ dsl_dir_t *dd = os->os->os_dsl_dataset->ds_dir;
+ zap_cursor_t cursor;
+ zap_attribute_t attr;
+
+ /* there is no next dir on a snapshot! */
+ if (os->os->os_dsl_dataset->ds_object !=
+ dd->dd_phys->dd_head_dataset_obj)
+ return (ENOENT);
+
+ zap_cursor_init_serialized(&cursor,
+ dd->dd_pool->dp_meta_objset,
+ dd->dd_phys->dd_child_dir_zapobj, *offp);
+
+ if (zap_cursor_retrieve(&cursor, &attr) != 0) {
+ zap_cursor_fini(&cursor);
+ return (ENOENT);
+ }
+
+ if (strlen(attr.za_name) + 1 > namelen) {
+ zap_cursor_fini(&cursor);
+ return (ENAMETOOLONG);
+ }
+
+ (void) strcpy(name, attr.za_name);
+ if (idp)
+ *idp = attr.za_first_integer;
+ zap_cursor_advance(&cursor);
+ *offp = zap_cursor_serialize(&cursor);
+ zap_cursor_fini(&cursor);
+
+ return (0);
+}
+
+struct findarg {
+ int (*func)(char *, void *);
+ void *arg;
+};
+
+/* ARGSUSED */
+static int
+findfunc(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg)
+{
+ struct findarg *fa = arg;
+ return (fa->func((char *)dsname, fa->arg));
+}
+
+/*
+ * Find all objsets under name, and for each, call 'func(child_name, arg)'.
+ * Perhaps change all callers to use dmu_objset_find_spa()?
+ */
+int
+dmu_objset_find(char *name, int func(char *, void *), void *arg, int flags)
+{
+ struct findarg fa;
+ fa.func = func;
+ fa.arg = arg;
+ return (dmu_objset_find_spa(NULL, name, findfunc, &fa, flags));
+}
+
+/*
+ * Find all objsets under name, call func on each
+ */
+int
+dmu_objset_find_spa(spa_t *spa, const char *name,
+ int func(spa_t *, uint64_t, const char *, void *), void *arg, int flags)
+{
+ dsl_dir_t *dd;
+ dsl_pool_t *dp;
+ dsl_dataset_t *ds;
+ zap_cursor_t zc;
+ zap_attribute_t *attr;
+ char *child;
+ uint64_t thisobj;
+ int err;
+
+ if (name == NULL)
+ name = spa_name(spa);
+ err = dsl_dir_open_spa(spa, name, FTAG, &dd, NULL);
+ if (err)
+ return (err);
+
+ /* Don't visit hidden ($MOS & $ORIGIN) objsets. */
+ if (dd->dd_myname[0] == '$') {
+ dsl_dir_close(dd, FTAG);
+ return (0);
+ }
+
+ thisobj = dd->dd_phys->dd_head_dataset_obj;
+ attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
+ dp = dd->dd_pool;
+
+ /*
+ * Iterate over all children.
+ */
+ if (flags & DS_FIND_CHILDREN) {
+ for (zap_cursor_init(&zc, dp->dp_meta_objset,
+ dd->dd_phys->dd_child_dir_zapobj);
+ zap_cursor_retrieve(&zc, attr) == 0;
+ (void) zap_cursor_advance(&zc)) {
+ ASSERT(attr->za_integer_length == sizeof (uint64_t));
+ ASSERT(attr->za_num_integers == 1);
+
+ child = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+ (void) strcpy(child, name);
+ (void) strcat(child, "/");
+ (void) strcat(child, attr->za_name);
+ err = dmu_objset_find_spa(spa, child, func, arg, flags);
+ kmem_free(child, MAXPATHLEN);
+ if (err)
+ break;
+ }
+ zap_cursor_fini(&zc);
+
+ if (err) {
+ dsl_dir_close(dd, FTAG);
+ kmem_free(attr, sizeof (zap_attribute_t));
+ return (err);
+ }
+ }
+
+ /*
+ * Iterate over all snapshots.
+ */
+ if (flags & DS_FIND_SNAPSHOTS) {
+ if (!dsl_pool_sync_context(dp))
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);
+ if (!dsl_pool_sync_context(dp))
+ rw_exit(&dp->dp_config_rwlock);
+
+ if (err == 0) {
+ uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
+ dsl_dataset_rele(ds, FTAG);
+
+ for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
+ zap_cursor_retrieve(&zc, attr) == 0;
+ (void) zap_cursor_advance(&zc)) {
+ ASSERT(attr->za_integer_length ==
+ sizeof (uint64_t));
+ ASSERT(attr->za_num_integers == 1);
+
+ child = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+ (void) strcpy(child, name);
+ (void) strcat(child, "@");
+ (void) strcat(child, attr->za_name);
+ err = func(spa, attr->za_first_integer,
+ child, arg);
+ kmem_free(child, MAXPATHLEN);
+ if (err)
+ break;
+ }
+ zap_cursor_fini(&zc);
+ }
+ }
+
+ dsl_dir_close(dd, FTAG);
+ kmem_free(attr, sizeof (zap_attribute_t));
+
+ if (err)
+ return (err);
+
+ /*
+ * Apply to self if appropriate.
+ */
+ err = func(spa, thisobj, name, arg);
+ return (err);
+}
+
+void
+dmu_objset_set_user(objset_t *os, void *user_ptr)
+{
+ ASSERT(MUTEX_HELD(&os->os->os_user_ptr_lock));
+ os->os->os_user_ptr = user_ptr;
+}
+
+void *
+dmu_objset_get_user(objset_t *os)
+{
+ ASSERT(MUTEX_HELD(&os->os->os_user_ptr_lock));
+ return (os->os->os_user_ptr);
+}
diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c
new file mode 100644
index 000000000..857b9a343
--- /dev/null
+++ b/module/zfs/dmu_send.c
@@ -0,0 +1,1181 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dmu.h>
+#include <sys/dmu_impl.h>
+#include <sys/dmu_tx.h>
+#include <sys/dbuf.h>
+#include <sys/dnode.h>
+#include <sys/zfs_context.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_synctask.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zap.h>
+#include <sys/zio_checksum.h>
+
+static char *dmu_recv_tag = "dmu_recv_tag";
+
+struct backuparg {
+ dmu_replay_record_t *drr;
+ vnode_t *vp;
+ offset_t *off;
+ objset_t *os;
+ zio_cksum_t zc;
+ int err;
+};
+
+static int
+dump_bytes(struct backuparg *ba, void *buf, int len)
+{
+ ssize_t resid; /* have to get resid to get detailed errno */
+ ASSERT3U(len % 8, ==, 0);
+
+ fletcher_4_incremental_native(buf, len, &ba->zc);
+ ba->err = vn_rdwr(UIO_WRITE, ba->vp,
+ (caddr_t)buf, len,
+ 0, UIO_SYSSPACE, FAPPEND, RLIM64_INFINITY, CRED(), &resid);
+ *ba->off += len;
+ return (ba->err);
+}
+
+static int
+dump_free(struct backuparg *ba, uint64_t object, uint64_t offset,
+ uint64_t length)
+{
+ /* write a FREE record */
+ bzero(ba->drr, sizeof (dmu_replay_record_t));
+ ba->drr->drr_type = DRR_FREE;
+ ba->drr->drr_u.drr_free.drr_object = object;
+ ba->drr->drr_u.drr_free.drr_offset = offset;
+ ba->drr->drr_u.drr_free.drr_length = length;
+
+ if (dump_bytes(ba, ba->drr, sizeof (dmu_replay_record_t)))
+ return (EINTR);
+ return (0);
+}
+
+static int
+dump_data(struct backuparg *ba, dmu_object_type_t type,
+ uint64_t object, uint64_t offset, int blksz, void *data)
+{
+ /* write a DATA record */
+ bzero(ba->drr, sizeof (dmu_replay_record_t));
+ ba->drr->drr_type = DRR_WRITE;
+ ba->drr->drr_u.drr_write.drr_object = object;
+ ba->drr->drr_u.drr_write.drr_type = type;
+ ba->drr->drr_u.drr_write.drr_offset = offset;
+ ba->drr->drr_u.drr_write.drr_length = blksz;
+
+ if (dump_bytes(ba, ba->drr, sizeof (dmu_replay_record_t)))
+ return (EINTR);
+ if (dump_bytes(ba, data, blksz))
+ return (EINTR);
+ return (0);
+}
+
+static int
+dump_freeobjects(struct backuparg *ba, uint64_t firstobj, uint64_t numobjs)
+{
+ /* write a FREEOBJECTS record */
+ bzero(ba->drr, sizeof (dmu_replay_record_t));
+ ba->drr->drr_type = DRR_FREEOBJECTS;
+ ba->drr->drr_u.drr_freeobjects.drr_firstobj = firstobj;
+ ba->drr->drr_u.drr_freeobjects.drr_numobjs = numobjs;
+
+ if (dump_bytes(ba, ba->drr, sizeof (dmu_replay_record_t)))
+ return (EINTR);
+ return (0);
+}
+
+static int
+dump_dnode(struct backuparg *ba, uint64_t object, dnode_phys_t *dnp)
+{
+ if (dnp == NULL || dnp->dn_type == DMU_OT_NONE)
+ return (dump_freeobjects(ba, object, 1));
+
+ /* write an OBJECT record */
+ bzero(ba->drr, sizeof (dmu_replay_record_t));
+ ba->drr->drr_type = DRR_OBJECT;
+ ba->drr->drr_u.drr_object.drr_object = object;
+ ba->drr->drr_u.drr_object.drr_type = dnp->dn_type;
+ ba->drr->drr_u.drr_object.drr_bonustype = dnp->dn_bonustype;
+ ba->drr->drr_u.drr_object.drr_blksz =
+ dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
+ ba->drr->drr_u.drr_object.drr_bonuslen = dnp->dn_bonuslen;
+ ba->drr->drr_u.drr_object.drr_checksum = dnp->dn_checksum;
+ ba->drr->drr_u.drr_object.drr_compress = dnp->dn_compress;
+
+ if (dump_bytes(ba, ba->drr, sizeof (dmu_replay_record_t)))
+ return (EINTR);
+
+ if (dump_bytes(ba, DN_BONUS(dnp), P2ROUNDUP(dnp->dn_bonuslen, 8)))
+ return (EINTR);
+
+ /* free anything past the end of the file */
+ if (dump_free(ba, object, (dnp->dn_maxblkid + 1) *
+ (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), -1ULL))
+ return (EINTR);
+ if (ba->err)
+ return (EINTR);
+ return (0);
+}
+
+#define BP_SPAN(dnp, level) \
+ (((uint64_t)dnp->dn_datablkszsec) << (SPA_MINBLOCKSHIFT + \
+ (level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT)))
+
+static int
+backup_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
+ const dnode_phys_t *dnp, void *arg)
+{
+ struct backuparg *ba = arg;
+ dmu_object_type_t type = bp ? BP_GET_TYPE(bp) : DMU_OT_NONE;
+ int err = 0;
+
+ if (issig(JUSTLOOKING) && issig(FORREAL))
+ return (EINTR);
+
+ if (bp == NULL && zb->zb_object == 0) {
+ uint64_t span = BP_SPAN(dnp, zb->zb_level);
+ uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT;
+ err = dump_freeobjects(ba, dnobj, span >> DNODE_SHIFT);
+ } else if (bp == NULL) {
+ uint64_t span = BP_SPAN(dnp, zb->zb_level);
+ err = dump_free(ba, zb->zb_object, zb->zb_blkid * span, span);
+ } else if (zb->zb_level > 0 || type == DMU_OT_OBJSET) {
+ return (0);
+ } else if (type == DMU_OT_DNODE) {
+ dnode_phys_t *blk;
+ int i;
+ int blksz = BP_GET_LSIZE(bp);
+ uint32_t aflags = ARC_WAIT;
+ arc_buf_t *abuf;
+
+ if (arc_read_nolock(NULL, spa, bp,
+ arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
+ ZIO_FLAG_CANFAIL, &aflags, zb) != 0)
+ return (EIO);
+
+ blk = abuf->b_data;
+ for (i = 0; i < blksz >> DNODE_SHIFT; i++) {
+ uint64_t dnobj = (zb->zb_blkid <<
+ (DNODE_BLOCK_SHIFT - DNODE_SHIFT)) + i;
+ err = dump_dnode(ba, dnobj, blk+i);
+ if (err)
+ break;
+ }
+ (void) arc_buf_remove_ref(abuf, &abuf);
+ } else { /* it's a level-0 block of a regular object */
+ uint32_t aflags = ARC_WAIT;
+ arc_buf_t *abuf;
+ int blksz = BP_GET_LSIZE(bp);
+
+ if (arc_read_nolock(NULL, spa, bp,
+ arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
+ ZIO_FLAG_CANFAIL, &aflags, zb) != 0)
+ return (EIO);
+
+ err = dump_data(ba, type, zb->zb_object, zb->zb_blkid * blksz,
+ blksz, abuf->b_data);
+ (void) arc_buf_remove_ref(abuf, &abuf);
+ }
+
+ ASSERT(err == 0 || err == EINTR);
+ return (err);
+}
+
+int
+dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
+ vnode_t *vp, offset_t *off)
+{
+ dsl_dataset_t *ds = tosnap->os->os_dsl_dataset;
+ dsl_dataset_t *fromds = fromsnap ? fromsnap->os->os_dsl_dataset : NULL;
+ dmu_replay_record_t *drr;
+ struct backuparg ba;
+ int err;
+ uint64_t fromtxg = 0;
+
+ /* tosnap must be a snapshot */
+ if (ds->ds_phys->ds_next_snap_obj == 0)
+ return (EINVAL);
+
+ /* fromsnap must be an earlier snapshot from the same fs as tosnap */
+ if (fromds && (ds->ds_dir != fromds->ds_dir ||
+ fromds->ds_phys->ds_creation_txg >= ds->ds_phys->ds_creation_txg))
+ return (EXDEV);
+
+ if (fromorigin) {
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ if (fromsnap)
+ return (EINVAL);
+
+ if (dsl_dir_is_clone(ds->ds_dir)) {
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ err = dsl_dataset_hold_obj(dp,
+ ds->ds_dir->dd_phys->dd_origin_obj, FTAG, &fromds);
+ rw_exit(&dp->dp_config_rwlock);
+ if (err)
+ return (err);
+ } else {
+ fromorigin = B_FALSE;
+ }
+ }
+
+
+ drr = kmem_zalloc(sizeof (dmu_replay_record_t), KM_SLEEP);
+ drr->drr_type = DRR_BEGIN;
+ drr->drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC;
+ drr->drr_u.drr_begin.drr_version = DMU_BACKUP_STREAM_VERSION;
+ drr->drr_u.drr_begin.drr_creation_time =
+ ds->ds_phys->ds_creation_time;
+ drr->drr_u.drr_begin.drr_type = tosnap->os->os_phys->os_type;
+ if (fromorigin)
+ drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CLONE;
+ drr->drr_u.drr_begin.drr_toguid = ds->ds_phys->ds_guid;
+ if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
+ drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CI_DATA;
+
+ if (fromds)
+ drr->drr_u.drr_begin.drr_fromguid = fromds->ds_phys->ds_guid;
+ dsl_dataset_name(ds, drr->drr_u.drr_begin.drr_toname);
+
+ if (fromds)
+ fromtxg = fromds->ds_phys->ds_creation_txg;
+ if (fromorigin)
+ dsl_dataset_rele(fromds, FTAG);
+
+ ba.drr = drr;
+ ba.vp = vp;
+ ba.os = tosnap;
+ ba.off = off;
+ ZIO_SET_CHECKSUM(&ba.zc, 0, 0, 0, 0);
+
+ if (dump_bytes(&ba, drr, sizeof (dmu_replay_record_t))) {
+ kmem_free(drr, sizeof (dmu_replay_record_t));
+ return (ba.err);
+ }
+
+ err = traverse_dataset(ds, fromtxg, TRAVERSE_PRE | TRAVERSE_PREFETCH,
+ backup_cb, &ba);
+
+ if (err) {
+ if (err == EINTR && ba.err)
+ err = ba.err;
+ kmem_free(drr, sizeof (dmu_replay_record_t));
+ return (err);
+ }
+
+ bzero(drr, sizeof (dmu_replay_record_t));
+ drr->drr_type = DRR_END;
+ drr->drr_u.drr_end.drr_checksum = ba.zc;
+
+ if (dump_bytes(&ba, drr, sizeof (dmu_replay_record_t))) {
+ kmem_free(drr, sizeof (dmu_replay_record_t));
+ return (ba.err);
+ }
+
+ kmem_free(drr, sizeof (dmu_replay_record_t));
+
+ return (0);
+}
+
+struct recvbeginsyncarg {
+ const char *tofs;
+ const char *tosnap;
+ dsl_dataset_t *origin;
+ uint64_t fromguid;
+ dmu_objset_type_t type;
+ void *tag;
+ boolean_t force;
+ uint64_t dsflags;
+ char clonelastname[MAXNAMELEN];
+ dsl_dataset_t *ds; /* the ds to recv into; returned from the syncfunc */
+};
+
+static dsl_dataset_t *
+recv_full_sync_impl(dsl_pool_t *dp, uint64_t dsobj, dmu_objset_type_t type,
+ cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds;
+
+ /* This should always work, since we just created it */
+ /* XXX - create should return an owned ds */
+ VERIFY(0 == dsl_dataset_own_obj(dp, dsobj,
+ DS_MODE_INCONSISTENT, dmu_recv_tag, &ds));
+
+ if (type != DMU_OST_NONE) {
+ (void) dmu_objset_create_impl(dp->dp_spa,
+ ds, &ds->ds_phys->ds_bp, type, tx);
+ }
+
+ spa_history_internal_log(LOG_DS_REPLAY_FULL_SYNC,
+ dp->dp_spa, tx, cr, "dataset = %lld", dsobj);
+
+ return (ds);
+}
+
+/* ARGSUSED */
+static int
+recv_full_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ uint64_t val;
+ int err;
+
+ err = zap_lookup(mos, dd->dd_phys->dd_child_dir_zapobj,
+ strrchr(rbsa->tofs, '/') + 1, sizeof (uint64_t), 1, &val);
+
+ if (err != ENOENT)
+ return (err ? err : EEXIST);
+
+ if (rbsa->origin) {
+ /* make sure it's a snap in the same pool */
+ if (rbsa->origin->ds_dir->dd_pool != dd->dd_pool)
+ return (EXDEV);
+ if (rbsa->origin->ds_phys->ds_num_children == 0)
+ return (EINVAL);
+ if (rbsa->origin->ds_phys->ds_guid != rbsa->fromguid)
+ return (ENODEV);
+ }
+
+ return (0);
+}
+
+static void
+recv_full_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ uint64_t flags = DS_FLAG_INCONSISTENT | rbsa->dsflags;
+ uint64_t dsobj;
+
+ dsobj = dsl_dataset_create_sync(dd, strrchr(rbsa->tofs, '/') + 1,
+ rbsa->origin, flags, cr, tx);
+
+ rbsa->ds = recv_full_sync_impl(dd->dd_pool, dsobj,
+ rbsa->origin ? DMU_OST_NONE : rbsa->type, cr, tx);
+}
+
+static int
+recv_full_existing_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ int err;
+
+ /* must be a head ds */
+ if (ds->ds_phys->ds_next_snap_obj != 0)
+ return (EINVAL);
+
+ /* must not be a clone ds */
+ if (dsl_dir_is_clone(ds->ds_dir))
+ return (EINVAL);
+
+ err = dsl_dataset_destroy_check(ds, rbsa->tag, tx);
+ if (err)
+ return (err);
+
+ if (rbsa->origin) {
+ /* make sure it's a snap in the same pool */
+ if (rbsa->origin->ds_dir->dd_pool != ds->ds_dir->dd_pool)
+ return (EXDEV);
+ if (rbsa->origin->ds_phys->ds_num_children == 0)
+ return (EINVAL);
+ if (rbsa->origin->ds_phys->ds_guid != rbsa->fromguid)
+ return (ENODEV);
+ }
+
+ return (0);
+}
+
+static void
+recv_full_existing_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ dsl_dir_t *dd = ds->ds_dir;
+ uint64_t flags = DS_FLAG_INCONSISTENT | rbsa->dsflags;
+ uint64_t dsobj;
+
+ /*
+ * NB: caller must provide an extra hold on the dsl_dir_t, so it
+ * won't go away when dsl_dataset_destroy_sync() closes the
+ * dataset.
+ */
+ dsl_dataset_destroy_sync(ds, rbsa->tag, cr, tx);
+
+ dsobj = dsl_dataset_create_sync_dd(dd, rbsa->origin, flags, tx);
+
+ rbsa->ds = recv_full_sync_impl(dd->dd_pool, dsobj,
+ rbsa->origin ? DMU_OST_NONE : rbsa->type, cr, tx);
+}
+
+/* ARGSUSED */
+static int
+recv_incremental_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ int err;
+ uint64_t val;
+
+ /* must not have any changes since most recent snapshot */
+ if (!rbsa->force && dsl_dataset_modified_since_lastsnap(ds))
+ return (ETXTBSY);
+
+ /* must already be a snapshot of this fs */
+ if (ds->ds_phys->ds_prev_snap_obj == 0)
+ return (ENODEV);
+
+ /* most recent snapshot must match fromguid */
+ if (ds->ds_prev->ds_phys->ds_guid != rbsa->fromguid)
+ return (ENODEV);
+
+ /* temporary clone name must not exist */
+ err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
+ ds->ds_dir->dd_phys->dd_child_dir_zapobj,
+ rbsa->clonelastname, 8, 1, &val);
+ if (err == 0)
+ return (EEXIST);
+ if (err != ENOENT)
+ return (err);
+
+ /* new snapshot name must not exist */
+ err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
+ ds->ds_phys->ds_snapnames_zapobj, rbsa->tosnap, 8, 1, &val);
+ if (err == 0)
+ return (EEXIST);
+ if (err != ENOENT)
+ return (err);
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+recv_online_incremental_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ohds = arg1;
+ struct recvbeginsyncarg *rbsa = arg2;
+ dsl_pool_t *dp = ohds->ds_dir->dd_pool;
+ dsl_dataset_t *ods, *cds;
+ uint64_t flags = DS_FLAG_INCONSISTENT | rbsa->dsflags;
+ uint64_t dsobj;
+
+ /* create the temporary clone */
+ VERIFY(0 == dsl_dataset_hold_obj(dp, ohds->ds_phys->ds_prev_snap_obj,
+ FTAG, &ods));
+ dsobj = dsl_dataset_create_sync(ohds->ds_dir,
+ rbsa->clonelastname, ods, flags, cr, tx);
+ dsl_dataset_rele(ods, FTAG);
+
+ /* open the temporary clone */
+ VERIFY(0 == dsl_dataset_own_obj(dp, dsobj,
+ DS_MODE_INCONSISTENT, dmu_recv_tag, &cds));
+
+ /* copy the refquota from the target fs to the clone */
+ if (ohds->ds_quota > 0)
+ dsl_dataset_set_quota_sync(cds, &ohds->ds_quota, cr, tx);
+
+ rbsa->ds = cds;
+
+ spa_history_internal_log(LOG_DS_REPLAY_INC_SYNC,
+ dp->dp_spa, tx, cr, "dataset = %lld", dsobj);
+}
+
+/* ARGSUSED */
+static void
+recv_offline_incremental_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_flags |= DS_FLAG_INCONSISTENT;
+
+ spa_history_internal_log(LOG_DS_REPLAY_INC_SYNC,
+ ds->ds_dir->dd_pool->dp_spa, tx, cr, "dataset = %lld",
+ ds->ds_object);
+}
+
+/*
+ * NB: callers *MUST* call dmu_recv_stream() if dmu_recv_begin()
+ * succeeds; otherwise we will leak the holds on the datasets.
+ */
+int
+dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *drrb,
+ boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *drc)
+{
+ int err = 0;
+ boolean_t byteswap;
+ struct recvbeginsyncarg rbsa;
+ uint64_t version;
+ int flags;
+ dsl_dataset_t *ds;
+
+ if (drrb->drr_magic == DMU_BACKUP_MAGIC)
+ byteswap = FALSE;
+ else if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC))
+ byteswap = TRUE;
+ else
+ return (EINVAL);
+
+ rbsa.tofs = tofs;
+ rbsa.tosnap = tosnap;
+ rbsa.origin = origin ? origin->os->os_dsl_dataset : NULL;
+ rbsa.fromguid = drrb->drr_fromguid;
+ rbsa.type = drrb->drr_type;
+ rbsa.tag = FTAG;
+ rbsa.dsflags = 0;
+ version = drrb->drr_version;
+ flags = drrb->drr_flags;
+
+ if (byteswap) {
+ rbsa.type = BSWAP_32(rbsa.type);
+ rbsa.fromguid = BSWAP_64(rbsa.fromguid);
+ version = BSWAP_64(version);
+ flags = BSWAP_32(flags);
+ }
+
+ if (version != DMU_BACKUP_STREAM_VERSION ||
+ rbsa.type >= DMU_OST_NUMTYPES ||
+ ((flags & DRR_FLAG_CLONE) && origin == NULL))
+ return (EINVAL);
+
+ if (flags & DRR_FLAG_CI_DATA)
+ rbsa.dsflags = DS_FLAG_CI_DATASET;
+
+ bzero(drc, sizeof (dmu_recv_cookie_t));
+ drc->drc_drrb = drrb;
+ drc->drc_tosnap = tosnap;
+ drc->drc_force = force;
+
+ /*
+ * Process the begin in syncing context.
+ */
+ if (rbsa.fromguid && !(flags & DRR_FLAG_CLONE) && !online) {
+ /* offline incremental receive */
+ err = dsl_dataset_own(tofs, 0, dmu_recv_tag, &ds);
+ if (err)
+ return (err);
+
+ /*
+ * Only do the rollback if the most recent snapshot
+ * matches the incremental source
+ */
+ if (force) {
+ if (ds->ds_prev == NULL ||
+ ds->ds_prev->ds_phys->ds_guid !=
+ rbsa.fromguid) {
+ dsl_dataset_disown(ds, dmu_recv_tag);
+ return (ENODEV);
+ }
+ (void) dsl_dataset_rollback(ds, DMU_OST_NONE);
+ }
+ rbsa.force = B_FALSE;
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ recv_incremental_check,
+ recv_offline_incremental_sync, ds, &rbsa, 1);
+ if (err) {
+ dsl_dataset_disown(ds, dmu_recv_tag);
+ return (err);
+ }
+ drc->drc_logical_ds = drc->drc_real_ds = ds;
+ } else if (rbsa.fromguid && !(flags & DRR_FLAG_CLONE)) {
+ /* online incremental receive */
+
+ /* tmp clone name is: tofs/%tosnap" */
+ (void) snprintf(rbsa.clonelastname, sizeof (rbsa.clonelastname),
+ "%%%s", tosnap);
+
+ /* open the dataset we are logically receiving into */
+ err = dsl_dataset_hold(tofs, dmu_recv_tag, &ds);
+ if (err)
+ return (err);
+
+ rbsa.force = force;
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ recv_incremental_check,
+ recv_online_incremental_sync, ds, &rbsa, 5);
+ if (err) {
+ dsl_dataset_rele(ds, dmu_recv_tag);
+ return (err);
+ }
+ drc->drc_logical_ds = ds;
+ drc->drc_real_ds = rbsa.ds;
+ } else {
+ /* create new fs -- full backup or clone */
+ dsl_dir_t *dd = NULL;
+ const char *tail;
+
+ err = dsl_dir_open(tofs, FTAG, &dd, &tail);
+ if (err)
+ return (err);
+ if (tail == NULL) {
+ if (!force) {
+ dsl_dir_close(dd, FTAG);
+ return (EEXIST);
+ }
+
+ rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
+ err = dsl_dataset_own_obj(dd->dd_pool,
+ dd->dd_phys->dd_head_dataset_obj,
+ DS_MODE_INCONSISTENT, FTAG, &ds);
+ rw_exit(&dd->dd_pool->dp_config_rwlock);
+ if (err) {
+ dsl_dir_close(dd, FTAG);
+ return (err);
+ }
+
+ dsl_dataset_make_exclusive(ds, FTAG);
+ err = dsl_sync_task_do(dd->dd_pool,
+ recv_full_existing_check,
+ recv_full_existing_sync, ds, &rbsa, 5);
+ dsl_dataset_disown(ds, FTAG);
+ } else {
+ err = dsl_sync_task_do(dd->dd_pool, recv_full_check,
+ recv_full_sync, dd, &rbsa, 5);
+ }
+ dsl_dir_close(dd, FTAG);
+ if (err)
+ return (err);
+ drc->drc_logical_ds = drc->drc_real_ds = rbsa.ds;
+ drc->drc_newfs = B_TRUE;
+ }
+
+ return (0);
+}
+
+struct restorearg {
+ int err;
+ int byteswap;
+ vnode_t *vp;
+ char *buf;
+ uint64_t voff;
+ int bufsize; /* amount of memory allocated for buf */
+ zio_cksum_t cksum;
+};
+
+static void *
+restore_read(struct restorearg *ra, int len)
+{
+ void *rv;
+ int done = 0;
+
+ /* some things will require 8-byte alignment, so everything must */
+ ASSERT3U(len % 8, ==, 0);
+
+ while (done < len) {
+ ssize_t resid;
+
+ ra->err = vn_rdwr(UIO_READ, ra->vp,
+ (caddr_t)ra->buf + done, len - done,
+ ra->voff, UIO_SYSSPACE, FAPPEND,
+ RLIM64_INFINITY, CRED(), &resid);
+
+ if (resid == len - done)
+ ra->err = EINVAL;
+ ra->voff += len - done - resid;
+ done = len - resid;
+ if (ra->err)
+ return (NULL);
+ }
+
+ ASSERT3U(done, ==, len);
+ rv = ra->buf;
+ if (ra->byteswap)
+ fletcher_4_incremental_byteswap(rv, len, &ra->cksum);
+ else
+ fletcher_4_incremental_native(rv, len, &ra->cksum);
+ return (rv);
+}
+
+static void
+backup_byteswap(dmu_replay_record_t *drr)
+{
+#define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X))
+#define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X))
+ drr->drr_type = BSWAP_32(drr->drr_type);
+ drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen);
+ switch (drr->drr_type) {
+ case DRR_BEGIN:
+ DO64(drr_begin.drr_magic);
+ DO64(drr_begin.drr_version);
+ DO64(drr_begin.drr_creation_time);
+ DO32(drr_begin.drr_type);
+ DO32(drr_begin.drr_flags);
+ DO64(drr_begin.drr_toguid);
+ DO64(drr_begin.drr_fromguid);
+ break;
+ case DRR_OBJECT:
+ DO64(drr_object.drr_object);
+ /* DO64(drr_object.drr_allocation_txg); */
+ DO32(drr_object.drr_type);
+ DO32(drr_object.drr_bonustype);
+ DO32(drr_object.drr_blksz);
+ DO32(drr_object.drr_bonuslen);
+ break;
+ case DRR_FREEOBJECTS:
+ DO64(drr_freeobjects.drr_firstobj);
+ DO64(drr_freeobjects.drr_numobjs);
+ break;
+ case DRR_WRITE:
+ DO64(drr_write.drr_object);
+ DO32(drr_write.drr_type);
+ DO64(drr_write.drr_offset);
+ DO64(drr_write.drr_length);
+ break;
+ case DRR_FREE:
+ DO64(drr_free.drr_object);
+ DO64(drr_free.drr_offset);
+ DO64(drr_free.drr_length);
+ break;
+ case DRR_END:
+ DO64(drr_end.drr_checksum.zc_word[0]);
+ DO64(drr_end.drr_checksum.zc_word[1]);
+ DO64(drr_end.drr_checksum.zc_word[2]);
+ DO64(drr_end.drr_checksum.zc_word[3]);
+ break;
+ }
+#undef DO64
+#undef DO32
+}
+
+static int
+restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
+{
+ int err;
+ dmu_tx_t *tx;
+ void *data = NULL;
+
+ err = dmu_object_info(os, drro->drr_object, NULL);
+
+ if (err != 0 && err != ENOENT)
+ return (EINVAL);
+
+ if (drro->drr_type == DMU_OT_NONE ||
+ drro->drr_type >= DMU_OT_NUMTYPES ||
+ drro->drr_bonustype >= DMU_OT_NUMTYPES ||
+ drro->drr_checksum >= ZIO_CHECKSUM_FUNCTIONS ||
+ drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS ||
+ P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) ||
+ drro->drr_blksz < SPA_MINBLOCKSIZE ||
+ drro->drr_blksz > SPA_MAXBLOCKSIZE ||
+ drro->drr_bonuslen > DN_MAX_BONUSLEN) {
+ return (EINVAL);
+ }
+
+ if (drro->drr_bonuslen) {
+ data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8));
+ if (ra->err)
+ return (ra->err);
+ }
+
+ tx = dmu_tx_create(os);
+
+ if (err == ENOENT) {
+ /* currently free, want to be allocated */
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, 1);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+ err = dmu_object_claim(os, drro->drr_object,
+ drro->drr_type, drro->drr_blksz,
+ drro->drr_bonustype, drro->drr_bonuslen, tx);
+ } else {
+ /* currently allocated, want to be allocated */
+ dmu_tx_hold_bonus(tx, drro->drr_object);
+ /*
+ * We may change blocksize, so need to
+ * hold_write
+ */
+ dmu_tx_hold_write(tx, drro->drr_object, 0, 1);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+
+ err = dmu_object_reclaim(os, drro->drr_object,
+ drro->drr_type, drro->drr_blksz,
+ drro->drr_bonustype, drro->drr_bonuslen, tx);
+ }
+ if (err) {
+ dmu_tx_commit(tx);
+ return (EINVAL);
+ }
+
+ dmu_object_set_checksum(os, drro->drr_object, drro->drr_checksum, tx);
+ dmu_object_set_compress(os, drro->drr_object, drro->drr_compress, tx);
+
+ if (data != NULL) {
+ dmu_buf_t *db;
+
+ VERIFY(0 == dmu_bonus_hold(os, drro->drr_object, FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+
+ ASSERT3U(db->db_size, >=, drro->drr_bonuslen);
+ bcopy(data, db->db_data, drro->drr_bonuslen);
+ if (ra->byteswap) {
+ dmu_ot[drro->drr_bonustype].ot_byteswap(db->db_data,
+ drro->drr_bonuslen);
+ }
+ dmu_buf_rele(db, FTAG);
+ }
+ dmu_tx_commit(tx);
+ return (0);
+}
+
+/* ARGSUSED */
+static int
+restore_freeobjects(struct restorearg *ra, objset_t *os,
+ struct drr_freeobjects *drrfo)
+{
+ uint64_t obj;
+
+ if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj)
+ return (EINVAL);
+
+ for (obj = drrfo->drr_firstobj;
+ obj < drrfo->drr_firstobj + drrfo->drr_numobjs;
+ (void) dmu_object_next(os, &obj, FALSE, 0)) {
+ int err;
+
+ if (dmu_object_info(os, obj, NULL) != 0)
+ continue;
+
+ err = dmu_free_object(os, obj);
+ if (err)
+ return (err);
+ }
+ return (0);
+}
+
+static int
+restore_write(struct restorearg *ra, objset_t *os,
+ struct drr_write *drrw)
+{
+ dmu_tx_t *tx;
+ void *data;
+ int err;
+
+ if (drrw->drr_offset + drrw->drr_length < drrw->drr_offset ||
+ drrw->drr_type >= DMU_OT_NUMTYPES)
+ return (EINVAL);
+
+ data = restore_read(ra, drrw->drr_length);
+ if (data == NULL)
+ return (ra->err);
+
+ if (dmu_object_info(os, drrw->drr_object, NULL) != 0)
+ return (EINVAL);
+
+ tx = dmu_tx_create(os);
+
+ dmu_tx_hold_write(tx, drrw->drr_object,
+ drrw->drr_offset, drrw->drr_length);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+ if (ra->byteswap)
+ dmu_ot[drrw->drr_type].ot_byteswap(data, drrw->drr_length);
+ dmu_write(os, drrw->drr_object,
+ drrw->drr_offset, drrw->drr_length, data, tx);
+ dmu_tx_commit(tx);
+ return (0);
+}
+
+/* ARGSUSED */
+static int
+restore_free(struct restorearg *ra, objset_t *os,
+ struct drr_free *drrf)
+{
+ int err;
+
+ if (drrf->drr_length != -1ULL &&
+ drrf->drr_offset + drrf->drr_length < drrf->drr_offset)
+ return (EINVAL);
+
+ if (dmu_object_info(os, drrf->drr_object, NULL) != 0)
+ return (EINVAL);
+
+ err = dmu_free_long_range(os, drrf->drr_object,
+ drrf->drr_offset, drrf->drr_length);
+ return (err);
+}
+
+void
+dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc)
+{
+ if (drc->drc_newfs || drc->drc_real_ds != drc->drc_logical_ds) {
+ /*
+ * online incremental or new fs: destroy the fs (which
+ * may be a clone) that we created
+ */
+ (void) dsl_dataset_destroy(drc->drc_real_ds, dmu_recv_tag);
+ if (drc->drc_real_ds != drc->drc_logical_ds)
+ dsl_dataset_rele(drc->drc_logical_ds, dmu_recv_tag);
+ } else {
+ /*
+ * offline incremental: rollback to most recent snapshot.
+ */
+ (void) dsl_dataset_rollback(drc->drc_real_ds, DMU_OST_NONE);
+ dsl_dataset_disown(drc->drc_real_ds, dmu_recv_tag);
+ }
+}
+
+/*
+ * NB: callers *must* call dmu_recv_end() if this succeeds.
+ */
+int
+dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp)
+{
+ struct restorearg ra = { 0 };
+ dmu_replay_record_t *drr;
+ objset_t *os;
+ zio_cksum_t pcksum;
+
+ if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC))
+ ra.byteswap = TRUE;
+
+ {
+ /* compute checksum of drr_begin record */
+ dmu_replay_record_t *drr;
+ drr = kmem_zalloc(sizeof (dmu_replay_record_t), KM_SLEEP);
+
+ drr->drr_type = DRR_BEGIN;
+ drr->drr_u.drr_begin = *drc->drc_drrb;
+ if (ra.byteswap) {
+ fletcher_4_incremental_byteswap(drr,
+ sizeof (dmu_replay_record_t), &ra.cksum);
+ } else {
+ fletcher_4_incremental_native(drr,
+ sizeof (dmu_replay_record_t), &ra.cksum);
+ }
+ kmem_free(drr, sizeof (dmu_replay_record_t));
+ }
+
+ if (ra.byteswap) {
+ struct drr_begin *drrb = drc->drc_drrb;
+ drrb->drr_magic = BSWAP_64(drrb->drr_magic);
+ drrb->drr_version = BSWAP_64(drrb->drr_version);
+ drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time);
+ drrb->drr_type = BSWAP_32(drrb->drr_type);
+ drrb->drr_toguid = BSWAP_64(drrb->drr_toguid);
+ drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid);
+ }
+
+ ra.vp = vp;
+ ra.voff = *voffp;
+ ra.bufsize = 1<<20;
+ ra.buf = kmem_alloc(ra.bufsize, KM_SLEEP);
+
+ /* these were verified in dmu_recv_begin */
+ ASSERT(drc->drc_drrb->drr_version == DMU_BACKUP_STREAM_VERSION);
+ ASSERT(drc->drc_drrb->drr_type < DMU_OST_NUMTYPES);
+
+ /*
+ * Open the objset we are modifying.
+ */
+ VERIFY(dmu_objset_open_ds(drc->drc_real_ds, DMU_OST_ANY, &os) == 0);
+
+ ASSERT(drc->drc_real_ds->ds_phys->ds_flags & DS_FLAG_INCONSISTENT);
+
+ /*
+ * Read records and process them.
+ */
+ pcksum = ra.cksum;
+ while (ra.err == 0 &&
+ NULL != (drr = restore_read(&ra, sizeof (*drr)))) {
+ if (issig(JUSTLOOKING) && issig(FORREAL)) {
+ ra.err = EINTR;
+ goto out;
+ }
+
+ if (ra.byteswap)
+ backup_byteswap(drr);
+
+ switch (drr->drr_type) {
+ case DRR_OBJECT:
+ {
+ /*
+ * We need to make a copy of the record header,
+ * because restore_{object,write} may need to
+ * restore_read(), which will invalidate drr.
+ */
+ struct drr_object drro = drr->drr_u.drr_object;
+ ra.err = restore_object(&ra, os, &drro);
+ break;
+ }
+ case DRR_FREEOBJECTS:
+ {
+ struct drr_freeobjects drrfo =
+ drr->drr_u.drr_freeobjects;
+ ra.err = restore_freeobjects(&ra, os, &drrfo);
+ break;
+ }
+ case DRR_WRITE:
+ {
+ struct drr_write drrw = drr->drr_u.drr_write;
+ ra.err = restore_write(&ra, os, &drrw);
+ break;
+ }
+ case DRR_FREE:
+ {
+ struct drr_free drrf = drr->drr_u.drr_free;
+ ra.err = restore_free(&ra, os, &drrf);
+ break;
+ }
+ case DRR_END:
+ {
+ struct drr_end drre = drr->drr_u.drr_end;
+ /*
+ * We compare against the *previous* checksum
+ * value, because the stored checksum is of
+ * everything before the DRR_END record.
+ */
+ if (!ZIO_CHECKSUM_EQUAL(drre.drr_checksum, pcksum))
+ ra.err = ECKSUM;
+ goto out;
+ }
+ default:
+ ra.err = EINVAL;
+ goto out;
+ }
+ pcksum = ra.cksum;
+ }
+ ASSERT(ra.err != 0);
+
+out:
+ dmu_objset_close(os);
+
+ if (ra.err != 0) {
+ /*
+ * rollback or destroy what we created, so we don't
+ * leave it in the restoring state.
+ */
+ txg_wait_synced(drc->drc_real_ds->ds_dir->dd_pool, 0);
+ dmu_recv_abort_cleanup(drc);
+ }
+
+ kmem_free(ra.buf, ra.bufsize);
+ *voffp = ra.voff;
+ return (ra.err);
+}
+
+struct recvendsyncarg {
+ char *tosnap;
+ uint64_t creation_time;
+ uint64_t toguid;
+};
+
+static int
+recv_end_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct recvendsyncarg *resa = arg2;
+
+ return (dsl_dataset_snapshot_check(ds, resa->tosnap, tx));
+}
+
+static void
+recv_end_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct recvendsyncarg *resa = arg2;
+
+ dsl_dataset_snapshot_sync(ds, resa->tosnap, cr, tx);
+
+ /* set snapshot's creation time and guid */
+ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
+ ds->ds_prev->ds_phys->ds_creation_time = resa->creation_time;
+ ds->ds_prev->ds_phys->ds_guid = resa->toguid;
+ ds->ds_prev->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
+}
+
+int
+dmu_recv_end(dmu_recv_cookie_t *drc)
+{
+ struct recvendsyncarg resa;
+ dsl_dataset_t *ds = drc->drc_logical_ds;
+ int err;
+
+ /*
+ * XXX hack; seems the ds is still dirty and
+ * dsl_pool_zil_clean() expects it to have a ds_user_ptr
+ * (and zil), but clone_swap() can close it.
+ */
+ txg_wait_synced(ds->ds_dir->dd_pool, 0);
+
+ if (ds != drc->drc_real_ds) {
+ /* we are doing an online recv */
+ if (dsl_dataset_tryown(ds, FALSE, dmu_recv_tag)) {
+ err = dsl_dataset_clone_swap(drc->drc_real_ds, ds,
+ drc->drc_force);
+ if (err)
+ dsl_dataset_disown(ds, dmu_recv_tag);
+ } else {
+ err = EBUSY;
+ dsl_dataset_rele(ds, dmu_recv_tag);
+ }
+ /* dsl_dataset_destroy() will disown the ds */
+ (void) dsl_dataset_destroy(drc->drc_real_ds, dmu_recv_tag);
+ if (err)
+ return (err);
+ }
+
+ resa.creation_time = drc->drc_drrb->drr_creation_time;
+ resa.toguid = drc->drc_drrb->drr_toguid;
+ resa.tosnap = drc->drc_tosnap;
+
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ recv_end_check, recv_end_sync, ds, &resa, 3);
+ if (err) {
+ if (drc->drc_newfs) {
+ ASSERT(ds == drc->drc_real_ds);
+ (void) dsl_dataset_destroy(ds, dmu_recv_tag);
+ return (err);
+ } else {
+ (void) dsl_dataset_rollback(ds, DMU_OST_NONE);
+ }
+ }
+
+ /* release the hold from dmu_recv_begin */
+ dsl_dataset_disown(ds, dmu_recv_tag);
+ return (err);
+}
diff --git a/module/zfs/dmu_traverse.c b/module/zfs/dmu_traverse.c
new file mode 100644
index 000000000..512401470
--- /dev/null
+++ b/module/zfs/dmu_traverse.c
@@ -0,0 +1,406 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_pool.h>
+#include <sys/dnode.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu_impl.h>
+#include <sys/callb.h>
+
+#define SET_BOOKMARK(zb, objset, object, level, blkid) \
+{ \
+ (zb)->zb_objset = objset; \
+ (zb)->zb_object = object; \
+ (zb)->zb_level = level; \
+ (zb)->zb_blkid = blkid; \
+}
+
+struct prefetch_data {
+ kmutex_t pd_mtx;
+ kcondvar_t pd_cv;
+ int pd_blks_max;
+ int pd_blks_fetched;
+ int pd_flags;
+ boolean_t pd_cancel;
+ boolean_t pd_exited;
+};
+
+struct traverse_data {
+ spa_t *td_spa;
+ uint64_t td_objset;
+ blkptr_t *td_rootbp;
+ uint64_t td_min_txg;
+ int td_flags;
+ struct prefetch_data *td_pfd;
+ blkptr_cb_t *td_func;
+ void *td_arg;
+};
+
+/* ARGSUSED */
+static void
+traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
+{
+ struct traverse_data *td = arg;
+ zbookmark_t zb;
+
+ if (bp->blk_birth == 0)
+ return;
+
+ if (claim_txg == 0 && bp->blk_birth >= spa_first_txg(td->td_spa))
+ return;
+
+ zb.zb_objset = td->td_objset;
+ zb.zb_object = 0;
+ zb.zb_level = -1;
+ zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ];
+ VERIFY(0 == td->td_func(td->td_spa, bp, &zb, NULL, td->td_arg));
+}
+
+/* ARGSUSED */
+static void
+traverse_zil_record(zilog_t *zilog, lr_t *lrc, void *arg, uint64_t claim_txg)
+{
+ struct traverse_data *td = arg;
+
+ if (lrc->lrc_txtype == TX_WRITE) {
+ lr_write_t *lr = (lr_write_t *)lrc;
+ blkptr_t *bp = &lr->lr_blkptr;
+ zbookmark_t zb;
+
+ if (bp->blk_birth == 0)
+ return;
+
+ if (claim_txg == 0 || bp->blk_birth < claim_txg)
+ return;
+
+ zb.zb_objset = td->td_objset;
+ zb.zb_object = lr->lr_foid;
+ zb.zb_level = BP_GET_LEVEL(bp);
+ zb.zb_blkid = lr->lr_offset / BP_GET_LSIZE(bp);
+ VERIFY(0 == td->td_func(td->td_spa, bp, &zb, NULL, td->td_arg));
+ }
+}
+
+static void
+traverse_zil(struct traverse_data *td, zil_header_t *zh)
+{
+ uint64_t claim_txg = zh->zh_claim_txg;
+ zilog_t *zilog;
+
+ /*
+ * We only want to visit blocks that have been claimed but not yet
+ * replayed (or, in read-only mode, blocks that *would* be claimed).
+ */
+ if (claim_txg == 0 && (spa_mode & FWRITE))
+ return;
+
+ zilog = zil_alloc(spa_get_dsl(td->td_spa)->dp_meta_objset, zh);
+
+ (void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, td,
+ claim_txg);
+
+ zil_free(zilog);
+}
+
+static int
+traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
+ arc_buf_t *pbuf, blkptr_t *bp, const zbookmark_t *zb)
+{
+ zbookmark_t czb;
+ int err = 0;
+ arc_buf_t *buf = NULL;
+ struct prefetch_data *pd = td->td_pfd;
+
+ if (bp->blk_birth == 0) {
+ err = td->td_func(td->td_spa, NULL, zb, dnp, td->td_arg);
+ return (err);
+ }
+
+ if (bp->blk_birth <= td->td_min_txg)
+ return (0);
+
+ if (pd && !pd->pd_exited &&
+ ((pd->pd_flags & TRAVERSE_PREFETCH_DATA) ||
+ BP_GET_TYPE(bp) == DMU_OT_DNODE || BP_GET_LEVEL(bp) > 0)) {
+ mutex_enter(&pd->pd_mtx);
+ ASSERT(pd->pd_blks_fetched >= 0);
+ while (pd->pd_blks_fetched == 0 && !pd->pd_exited)
+ cv_wait(&pd->pd_cv, &pd->pd_mtx);
+ pd->pd_blks_fetched--;
+ cv_broadcast(&pd->pd_cv);
+ mutex_exit(&pd->pd_mtx);
+ }
+
+ if (td->td_flags & TRAVERSE_PRE) {
+ err = td->td_func(td->td_spa, bp, zb, dnp, td->td_arg);
+ if (err)
+ return (err);
+ }
+
+ if (BP_GET_LEVEL(bp) > 0) {
+ uint32_t flags = ARC_WAIT;
+ int i;
+ blkptr_t *cbp;
+ int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
+
+ err = arc_read(NULL, td->td_spa, bp, pbuf,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err)
+ return (err);
+
+ /* recursively visitbp() blocks below this */
+ cbp = buf->b_data;
+ for (i = 0; i < epb; i++, cbp++) {
+ SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
+ zb->zb_level - 1,
+ zb->zb_blkid * epb + i);
+ err = traverse_visitbp(td, dnp, buf, cbp, &czb);
+ if (err)
+ break;
+ }
+ } else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
+ uint32_t flags = ARC_WAIT;
+ int i, j;
+ int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
+
+ err = arc_read(NULL, td->td_spa, bp, pbuf,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err)
+ return (err);
+
+ /* recursively visitbp() blocks below this */
+ dnp = buf->b_data;
+ for (i = 0; i < epb && err == 0; i++, dnp++) {
+ for (j = 0; j < dnp->dn_nblkptr; j++) {
+ SET_BOOKMARK(&czb, zb->zb_objset,
+ zb->zb_blkid * epb + i,
+ dnp->dn_nlevels - 1, j);
+ err = traverse_visitbp(td, dnp, buf,
+ (blkptr_t *)&dnp->dn_blkptr[j], &czb);
+ if (err)
+ break;
+ }
+ }
+ } else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
+ uint32_t flags = ARC_WAIT;
+ objset_phys_t *osp;
+ int j;
+
+ err = arc_read_nolock(NULL, td->td_spa, bp,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err)
+ return (err);
+
+ osp = buf->b_data;
+ /*
+ * traverse_zil is just here for zdb's leak checking.
+ * For other consumers, there will be no ZIL blocks.
+ */
+ traverse_zil(td, &osp->os_zil_header);
+
+ for (j = 0; j < osp->os_meta_dnode.dn_nblkptr; j++) {
+ SET_BOOKMARK(&czb, zb->zb_objset, 0,
+ osp->os_meta_dnode.dn_nlevels - 1, j);
+ err = traverse_visitbp(td, &osp->os_meta_dnode, buf,
+ (blkptr_t *)&osp->os_meta_dnode.dn_blkptr[j],
+ &czb);
+ if (err)
+ break;
+ }
+ }
+
+ if (buf)
+ (void) arc_buf_remove_ref(buf, &buf);
+
+ if (err == 0 && (td->td_flags & TRAVERSE_POST))
+ err = td->td_func(td->td_spa, bp, zb, dnp, td->td_arg);
+
+ return (err);
+}
+
+/* ARGSUSED */
+static int
+traverse_prefetcher(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
+ const dnode_phys_t *dnp, void *arg)
+{
+ struct prefetch_data *pfd = arg;
+ uint32_t aflags = ARC_NOWAIT | ARC_PREFETCH;
+
+ ASSERT(pfd->pd_blks_fetched >= 0);
+ if (pfd->pd_cancel)
+ return (EINTR);
+
+ if (bp == NULL || !((pfd->pd_flags & TRAVERSE_PREFETCH_DATA) ||
+ BP_GET_TYPE(bp) == DMU_OT_DNODE || BP_GET_LEVEL(bp) > 0))
+ return (0);
+
+ mutex_enter(&pfd->pd_mtx);
+ while (!pfd->pd_cancel && pfd->pd_blks_fetched >= pfd->pd_blks_max)
+ cv_wait(&pfd->pd_cv, &pfd->pd_mtx);
+ pfd->pd_blks_fetched++;
+ cv_broadcast(&pfd->pd_cv);
+ mutex_exit(&pfd->pd_mtx);
+
+ (void) arc_read_nolock(NULL, spa, bp, NULL, NULL,
+ ZIO_PRIORITY_ASYNC_READ,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
+ &aflags, zb);
+
+ return (0);
+}
+
+static void
+traverse_prefetch_thread(void *arg)
+{
+ struct traverse_data *td_main = arg;
+ struct traverse_data td = *td_main;
+ zbookmark_t czb;
+
+ td.td_func = traverse_prefetcher;
+ td.td_arg = td_main->td_pfd;
+ td.td_pfd = NULL;
+
+ SET_BOOKMARK(&czb, td.td_objset, 0, -1, 0);
+ (void) traverse_visitbp(&td, NULL, NULL, td.td_rootbp, &czb);
+
+ mutex_enter(&td_main->td_pfd->pd_mtx);
+ td_main->td_pfd->pd_exited = B_TRUE;
+ cv_broadcast(&td_main->td_pfd->pd_cv);
+ mutex_exit(&td_main->td_pfd->pd_mtx);
+}
+
+/*
+ * NB: dataset must not be changing on-disk (eg, is a snapshot or we are
+ * in syncing context).
+ */
+static int
+traverse_impl(spa_t *spa, uint64_t objset, blkptr_t *rootbp,
+ uint64_t txg_start, int flags, blkptr_cb_t func, void *arg)
+{
+ struct traverse_data td;
+ struct prefetch_data pd = { 0 };
+ zbookmark_t czb;
+ int err;
+
+ td.td_spa = spa;
+ td.td_objset = objset;
+ td.td_rootbp = rootbp;
+ td.td_min_txg = txg_start;
+ td.td_func = func;
+ td.td_arg = arg;
+ td.td_pfd = &pd;
+ td.td_flags = flags;
+
+ pd.pd_blks_max = 100;
+ pd.pd_flags = flags;
+ mutex_init(&pd.pd_mtx, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&pd.pd_cv, NULL, CV_DEFAULT, NULL);
+
+ if (!(flags & TRAVERSE_PREFETCH) ||
+ 0 == taskq_dispatch(system_taskq, traverse_prefetch_thread,
+ &td, TQ_NOQUEUE))
+ pd.pd_exited = B_TRUE;
+
+ SET_BOOKMARK(&czb, objset, 0, -1, 0);
+ err = traverse_visitbp(&td, NULL, NULL, rootbp, &czb);
+
+ mutex_enter(&pd.pd_mtx);
+ pd.pd_cancel = B_TRUE;
+ cv_broadcast(&pd.pd_cv);
+ while (!pd.pd_exited)
+ cv_wait(&pd.pd_cv, &pd.pd_mtx);
+ mutex_exit(&pd.pd_mtx);
+
+ mutex_destroy(&pd.pd_mtx);
+ cv_destroy(&pd.pd_cv);
+
+ return (err);
+}
+
+/*
+ * NB: dataset must not be changing on-disk (eg, is a snapshot or we are
+ * in syncing context).
+ */
+int
+traverse_dataset(dsl_dataset_t *ds, uint64_t txg_start, int flags,
+ blkptr_cb_t func, void *arg)
+{
+ return (traverse_impl(ds->ds_dir->dd_pool->dp_spa, ds->ds_object,
+ &ds->ds_phys->ds_bp, txg_start, flags, func, arg));
+}
+
+/*
+ * NB: pool must not be changing on-disk (eg, from zdb or sync context).
+ */
+int
+traverse_pool(spa_t *spa, blkptr_cb_t func, void *arg)
+{
+ int err;
+ uint64_t obj;
+ dsl_pool_t *dp = spa_get_dsl(spa);
+ objset_t *mos = dp->dp_meta_objset;
+
+ /* visit the MOS */
+ err = traverse_impl(spa, 0, spa_get_rootblkptr(spa),
+ 0, TRAVERSE_PRE, func, arg);
+ if (err)
+ return (err);
+
+ /* visit each dataset */
+ for (obj = 1; err == 0; err = dmu_object_next(mos, &obj, FALSE, 0)) {
+ dmu_object_info_t doi;
+
+ err = dmu_object_info(mos, obj, &doi);
+ if (err)
+ return (err);
+
+ if (doi.doi_type == DMU_OT_DSL_DATASET) {
+ dsl_dataset_t *ds;
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ err = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
+ rw_exit(&dp->dp_config_rwlock);
+ if (err)
+ return (err);
+ err = traverse_dataset(ds,
+ ds->ds_phys->ds_prev_snap_txg, TRAVERSE_PRE,
+ func, arg);
+ dsl_dataset_rele(ds, FTAG);
+ if (err)
+ return (err);
+ }
+ }
+ if (err == ESRCH)
+ err = 0;
+ return (err);
+}
diff --git a/module/zfs/dmu_tx.c b/module/zfs/dmu_tx.c
new file mode 100644
index 000000000..bf560e565
--- /dev/null
+++ b/module/zfs/dmu_tx.c
@@ -0,0 +1,1068 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dmu.h>
+#include <sys/dmu_impl.h>
+#include <sys/dbuf.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dataset.h> /* for dsl_dataset_block_freeable() */
+#include <sys/dsl_dir.h> /* for dsl_dir_tempreserve_*() */
+#include <sys/dsl_pool.h>
+#include <sys/zap_impl.h> /* for fzap_default_block_shift */
+#include <sys/spa.h>
+#include <sys/zfs_context.h>
+
+typedef void (*dmu_tx_hold_func_t)(dmu_tx_t *tx, struct dnode *dn,
+ uint64_t arg1, uint64_t arg2);
+
+
+dmu_tx_t *
+dmu_tx_create_dd(dsl_dir_t *dd)
+{
+ dmu_tx_t *tx = kmem_zalloc(sizeof (dmu_tx_t), KM_SLEEP);
+ tx->tx_dir = dd;
+ if (dd)
+ tx->tx_pool = dd->dd_pool;
+ list_create(&tx->tx_holds, sizeof (dmu_tx_hold_t),
+ offsetof(dmu_tx_hold_t, txh_node));
+#ifdef ZFS_DEBUG
+ refcount_create(&tx->tx_space_written);
+ refcount_create(&tx->tx_space_freed);
+#endif
+ return (tx);
+}
+
+dmu_tx_t *
+dmu_tx_create(objset_t *os)
+{
+ dmu_tx_t *tx = dmu_tx_create_dd(os->os->os_dsl_dataset->ds_dir);
+ tx->tx_objset = os;
+ tx->tx_lastsnap_txg = dsl_dataset_prev_snap_txg(os->os->os_dsl_dataset);
+ return (tx);
+}
+
+dmu_tx_t *
+dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg)
+{
+ dmu_tx_t *tx = dmu_tx_create_dd(NULL);
+
+ ASSERT3U(txg, <=, dp->dp_tx.tx_open_txg);
+ tx->tx_pool = dp;
+ tx->tx_txg = txg;
+ tx->tx_anyobj = TRUE;
+
+ return (tx);
+}
+
+int
+dmu_tx_is_syncing(dmu_tx_t *tx)
+{
+ return (tx->tx_anyobj);
+}
+
+int
+dmu_tx_private_ok(dmu_tx_t *tx)
+{
+ return (tx->tx_anyobj);
+}
+
+static dmu_tx_hold_t *
+dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object,
+ enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2)
+{
+ dmu_tx_hold_t *txh;
+ dnode_t *dn = NULL;
+ int err;
+
+ if (object != DMU_NEW_OBJECT) {
+ err = dnode_hold(os->os, object, tx, &dn);
+ if (err) {
+ tx->tx_err = err;
+ return (NULL);
+ }
+
+ if (err == 0 && tx->tx_txg != 0) {
+ mutex_enter(&dn->dn_mtx);
+ /*
+ * dn->dn_assigned_txg == tx->tx_txg doesn't pose a
+ * problem, but there's no way for it to happen (for
+ * now, at least).
+ */
+ ASSERT(dn->dn_assigned_txg == 0);
+ dn->dn_assigned_txg = tx->tx_txg;
+ (void) refcount_add(&dn->dn_tx_holds, tx);
+ mutex_exit(&dn->dn_mtx);
+ }
+ }
+
+ txh = kmem_zalloc(sizeof (dmu_tx_hold_t), KM_SLEEP);
+ txh->txh_tx = tx;
+ txh->txh_dnode = dn;
+#ifdef ZFS_DEBUG
+ txh->txh_type = type;
+ txh->txh_arg1 = arg1;
+ txh->txh_arg2 = arg2;
+#endif
+ list_insert_tail(&tx->tx_holds, txh);
+
+ return (txh);
+}
+
+void
+dmu_tx_add_new_object(dmu_tx_t *tx, objset_t *os, uint64_t object)
+{
+ /*
+ * If we're syncing, they can manipulate any object anyhow, and
+ * the hold on the dnode_t can cause problems.
+ */
+ if (!dmu_tx_is_syncing(tx)) {
+ (void) dmu_tx_hold_object_impl(tx, os,
+ object, THT_NEWOBJECT, 0, 0);
+ }
+}
+
+static int
+dmu_tx_check_ioerr(zio_t *zio, dnode_t *dn, int level, uint64_t blkid)
+{
+ int err;
+ dmu_buf_impl_t *db;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ db = dbuf_hold_level(dn, level, blkid, FTAG);
+ rw_exit(&dn->dn_struct_rwlock);
+ if (db == NULL)
+ return (EIO);
+ err = dbuf_read(db, zio, DB_RF_CANFAIL | DB_RF_NOPREFETCH);
+ dbuf_rele(db, FTAG);
+ return (err);
+}
+
+/* ARGSUSED */
+static void
+dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
+{
+ dnode_t *dn = txh->txh_dnode;
+ uint64_t start, end, i;
+ int min_bs, max_bs, min_ibs, max_ibs, epbs, bits;
+ int err = 0;
+
+ if (len == 0)
+ return;
+
+ min_bs = SPA_MINBLOCKSHIFT;
+ max_bs = SPA_MAXBLOCKSHIFT;
+ min_ibs = DN_MIN_INDBLKSHIFT;
+ max_ibs = DN_MAX_INDBLKSHIFT;
+
+ /*
+ * For i/o error checking, read the first and last level-0
+ * blocks (if they are not aligned), and all the level-1 blocks.
+ */
+
+ if (dn) {
+ if (dn->dn_maxblkid == 0) {
+ if ((off > 0 || len < dn->dn_datablksz) &&
+ off < dn->dn_datablksz) {
+ err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
+ if (err)
+ goto out;
+ }
+ } else {
+ zio_t *zio = zio_root(dn->dn_objset->os_spa,
+ NULL, NULL, ZIO_FLAG_CANFAIL);
+
+ /* first level-0 block */
+ start = off >> dn->dn_datablkshift;
+ if (P2PHASE(off, dn->dn_datablksz) ||
+ len < dn->dn_datablksz) {
+ err = dmu_tx_check_ioerr(zio, dn, 0, start);
+ if (err)
+ goto out;
+ }
+
+ /* last level-0 block */
+ end = (off+len-1) >> dn->dn_datablkshift;
+ if (end != start && end <= dn->dn_maxblkid &&
+ P2PHASE(off+len, dn->dn_datablksz)) {
+ err = dmu_tx_check_ioerr(zio, dn, 0, end);
+ if (err)
+ goto out;
+ }
+
+ /* level-1 blocks */
+ if (dn->dn_nlevels > 1) {
+ start >>= dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ end >>= dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ for (i = start+1; i < end; i++) {
+ err = dmu_tx_check_ioerr(zio, dn, 1, i);
+ if (err)
+ goto out;
+ }
+ }
+
+ err = zio_wait(zio);
+ if (err)
+ goto out;
+ }
+ }
+
+ /*
+ * If there's more than one block, the blocksize can't change,
+ * so we can make a more precise estimate. Alternatively,
+ * if the dnode's ibs is larger than max_ibs, always use that.
+ * This ensures that if we reduce DN_MAX_INDBLKSHIFT,
+ * the code will still work correctly on existing pools.
+ */
+ if (dn && (dn->dn_maxblkid != 0 || dn->dn_indblkshift > max_ibs)) {
+ min_ibs = max_ibs = dn->dn_indblkshift;
+ if (dn->dn_datablkshift != 0)
+ min_bs = max_bs = dn->dn_datablkshift;
+ }
+
+ /*
+ * 'end' is the last thing we will access, not one past.
+ * This way we won't overflow when accessing the last byte.
+ */
+ start = P2ALIGN(off, 1ULL << max_bs);
+ end = P2ROUNDUP(off + len, 1ULL << max_bs) - 1;
+ txh->txh_space_towrite += end - start + 1;
+
+ start >>= min_bs;
+ end >>= min_bs;
+
+ epbs = min_ibs - SPA_BLKPTRSHIFT;
+
+ /*
+ * The object contains at most 2^(64 - min_bs) blocks,
+ * and each indirect level maps 2^epbs.
+ */
+ for (bits = 64 - min_bs; bits >= 0; bits -= epbs) {
+ start >>= epbs;
+ end >>= epbs;
+ /*
+ * If we increase the number of levels of indirection,
+ * we'll need new blkid=0 indirect blocks. If start == 0,
+ * we're already accounting for that blocks; and if end == 0,
+ * we can't increase the number of levels beyond that.
+ */
+ if (start != 0 && end != 0)
+ txh->txh_space_towrite += 1ULL << max_ibs;
+ txh->txh_space_towrite += (end - start + 1) << max_ibs;
+ }
+
+ ASSERT(txh->txh_space_towrite < 2 * DMU_MAX_ACCESS);
+
+out:
+ if (err)
+ txh->txh_tx->tx_err = err;
+}
+
+static void
+dmu_tx_count_dnode(dmu_tx_hold_t *txh)
+{
+ dnode_t *dn = txh->txh_dnode;
+ dnode_t *mdn = txh->txh_tx->tx_objset->os->os_meta_dnode;
+ uint64_t space = mdn->dn_datablksz +
+ ((mdn->dn_nlevels-1) << mdn->dn_indblkshift);
+
+ if (dn && dn->dn_dbuf->db_blkptr &&
+ dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,
+ dn->dn_dbuf->db_blkptr->blk_birth)) {
+ txh->txh_space_tooverwrite += space;
+ } else {
+ txh->txh_space_towrite += space;
+ if (dn && dn->dn_dbuf->db_blkptr)
+ txh->txh_space_tounref += space;
+ }
+}
+
+void
+dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len)
+{
+ dmu_tx_hold_t *txh;
+
+ ASSERT(tx->tx_txg == 0);
+ ASSERT(len < DMU_MAX_ACCESS);
+ ASSERT(len == 0 || UINT64_MAX - off >= len - 1);
+
+ txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+ object, THT_WRITE, off, len);
+ if (txh == NULL)
+ return;
+
+ dmu_tx_count_write(txh, off, len);
+ dmu_tx_count_dnode(txh);
+}
+
+static void
+dmu_tx_count_free(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
+{
+ uint64_t blkid, nblks, lastblk;
+ uint64_t space = 0, unref = 0, skipped = 0;
+ dnode_t *dn = txh->txh_dnode;
+ dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
+ spa_t *spa = txh->txh_tx->tx_pool->dp_spa;
+ int epbs;
+
+ if (dn->dn_nlevels == 0)
+ return;
+
+ /*
+ * The struct_rwlock protects us against dn_nlevels
+ * changing, in case (against all odds) we manage to dirty &
+ * sync out the changes after we check for being dirty.
+ * Also, dbuf_hold_level() wants us to have the struct_rwlock.
+ */
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ if (dn->dn_maxblkid == 0) {
+ if (off == 0 && len >= dn->dn_datablksz) {
+ blkid = 0;
+ nblks = 1;
+ } else {
+ rw_exit(&dn->dn_struct_rwlock);
+ return;
+ }
+ } else {
+ blkid = off >> dn->dn_datablkshift;
+ nblks = (len + dn->dn_datablksz - 1) >> dn->dn_datablkshift;
+
+ if (blkid >= dn->dn_maxblkid) {
+ rw_exit(&dn->dn_struct_rwlock);
+ return;
+ }
+ if (blkid + nblks > dn->dn_maxblkid)
+ nblks = dn->dn_maxblkid - blkid;
+
+ }
+ if (dn->dn_nlevels == 1) {
+ int i;
+ for (i = 0; i < nblks; i++) {
+ blkptr_t *bp = dn->dn_phys->dn_blkptr;
+ ASSERT3U(blkid + i, <, dn->dn_nblkptr);
+ bp += blkid + i;
+ if (dsl_dataset_block_freeable(ds, bp->blk_birth)) {
+ dprintf_bp(bp, "can free old%s", "");
+ space += bp_get_dasize(spa, bp);
+ }
+ unref += BP_GET_ASIZE(bp);
+ }
+ nblks = 0;
+ }
+
+ /*
+ * Add in memory requirements of higher-level indirects.
+ * This assumes a worst-possible scenario for dn_nlevels.
+ */
+ {
+ uint64_t blkcnt = 1 + ((nblks >> epbs) >> epbs);
+ int level = (dn->dn_nlevels > 1) ? 2 : 1;
+
+ while (level++ < DN_MAX_LEVELS) {
+ txh->txh_memory_tohold += blkcnt << dn->dn_indblkshift;
+ blkcnt = 1 + (blkcnt >> epbs);
+ }
+ ASSERT(blkcnt <= dn->dn_nblkptr);
+ }
+
+ lastblk = blkid + nblks - 1;
+ while (nblks) {
+ dmu_buf_impl_t *dbuf;
+ uint64_t ibyte, new_blkid;
+ int epb = 1 << epbs;
+ int err, i, blkoff, tochk;
+ blkptr_t *bp;
+
+ ibyte = blkid << dn->dn_datablkshift;
+ err = dnode_next_offset(dn,
+ DNODE_FIND_HAVELOCK, &ibyte, 2, 1, 0);
+ new_blkid = ibyte >> dn->dn_datablkshift;
+ if (err == ESRCH) {
+ skipped += (lastblk >> epbs) - (blkid >> epbs) + 1;
+ break;
+ }
+ if (err) {
+ txh->txh_tx->tx_err = err;
+ break;
+ }
+ if (new_blkid > lastblk) {
+ skipped += (lastblk >> epbs) - (blkid >> epbs) + 1;
+ break;
+ }
+
+ if (new_blkid > blkid) {
+ ASSERT((new_blkid >> epbs) > (blkid >> epbs));
+ skipped += (new_blkid >> epbs) - (blkid >> epbs) - 1;
+ nblks -= new_blkid - blkid;
+ blkid = new_blkid;
+ }
+ blkoff = P2PHASE(blkid, epb);
+ tochk = MIN(epb - blkoff, nblks);
+
+ dbuf = dbuf_hold_level(dn, 1, blkid >> epbs, FTAG);
+
+ txh->txh_memory_tohold += dbuf->db.db_size;
+ if (txh->txh_memory_tohold > DMU_MAX_ACCESS) {
+ txh->txh_tx->tx_err = E2BIG;
+ dbuf_rele(dbuf, FTAG);
+ break;
+ }
+ err = dbuf_read(dbuf, NULL, DB_RF_HAVESTRUCT | DB_RF_CANFAIL);
+ if (err != 0) {
+ txh->txh_tx->tx_err = err;
+ dbuf_rele(dbuf, FTAG);
+ break;
+ }
+
+ bp = dbuf->db.db_data;
+ bp += blkoff;
+
+ for (i = 0; i < tochk; i++) {
+ if (dsl_dataset_block_freeable(ds, bp[i].blk_birth)) {
+ dprintf_bp(&bp[i], "can free old%s", "");
+ space += bp_get_dasize(spa, &bp[i]);
+ }
+ unref += BP_GET_ASIZE(bp);
+ }
+ dbuf_rele(dbuf, FTAG);
+
+ blkid += tochk;
+ nblks -= tochk;
+ }
+ rw_exit(&dn->dn_struct_rwlock);
+
+ /* account for new level 1 indirect blocks that might show up */
+ if (skipped > 0) {
+ txh->txh_fudge += skipped << dn->dn_indblkshift;
+ skipped = MIN(skipped, DMU_MAX_DELETEBLKCNT >> epbs);
+ txh->txh_memory_tohold += skipped << dn->dn_indblkshift;
+ }
+ txh->txh_space_tofree += space;
+ txh->txh_space_tounref += unref;
+}
+
+void
+dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len)
+{
+ dmu_tx_hold_t *txh;
+ dnode_t *dn;
+ uint64_t start, end, i;
+ int err, shift;
+ zio_t *zio;
+
+ ASSERT(tx->tx_txg == 0);
+
+ txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+ object, THT_FREE, off, len);
+ if (txh == NULL)
+ return;
+ dn = txh->txh_dnode;
+
+ /* first block */
+ if (off != 0)
+ dmu_tx_count_write(txh, off, 1);
+ /* last block */
+ if (len != DMU_OBJECT_END)
+ dmu_tx_count_write(txh, off+len, 1);
+
+ if (off >= (dn->dn_maxblkid+1) * dn->dn_datablksz)
+ return;
+ if (len == DMU_OBJECT_END)
+ len = (dn->dn_maxblkid+1) * dn->dn_datablksz - off;
+
+ /*
+ * For i/o error checking, read the first and last level-0
+ * blocks, and all the level-1 blocks. The above count_write's
+ * have already taken care of the level-0 blocks.
+ */
+ if (dn->dn_nlevels > 1) {
+ shift = dn->dn_datablkshift + dn->dn_indblkshift -
+ SPA_BLKPTRSHIFT;
+ start = off >> shift;
+ end = dn->dn_datablkshift ? ((off+len) >> shift) : 0;
+
+ zio = zio_root(tx->tx_pool->dp_spa,
+ NULL, NULL, ZIO_FLAG_CANFAIL);
+ for (i = start; i <= end; i++) {
+ uint64_t ibyte = i << shift;
+ err = dnode_next_offset(dn, 0, &ibyte, 2, 1, 0);
+ i = ibyte >> shift;
+ if (err == ESRCH)
+ break;
+ if (err) {
+ tx->tx_err = err;
+ return;
+ }
+
+ err = dmu_tx_check_ioerr(zio, dn, 1, i);
+ if (err) {
+ tx->tx_err = err;
+ return;
+ }
+ }
+ err = zio_wait(zio);
+ if (err) {
+ tx->tx_err = err;
+ return;
+ }
+ }
+
+ dmu_tx_count_dnode(txh);
+ dmu_tx_count_free(txh, off, len);
+}
+
+void
+dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name)
+{
+ dmu_tx_hold_t *txh;
+ dnode_t *dn;
+ uint64_t nblocks;
+ int epbs, err;
+
+ ASSERT(tx->tx_txg == 0);
+
+ txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+ object, THT_ZAP, add, (uintptr_t)name);
+ if (txh == NULL)
+ return;
+ dn = txh->txh_dnode;
+
+ dmu_tx_count_dnode(txh);
+
+ if (dn == NULL) {
+ /*
+ * We will be able to fit a new object's entries into one leaf
+ * block. So there will be at most 2 blocks total,
+ * including the header block.
+ */
+ dmu_tx_count_write(txh, 0, 2 << fzap_default_block_shift);
+ return;
+ }
+
+ ASSERT3P(dmu_ot[dn->dn_type].ot_byteswap, ==, zap_byteswap);
+
+ if (dn->dn_maxblkid == 0 && !add) {
+ /*
+ * If there is only one block (i.e. this is a micro-zap)
+ * and we are not adding anything, the accounting is simple.
+ */
+ err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
+ if (err) {
+ tx->tx_err = err;
+ return;
+ }
+
+ /*
+ * Use max block size here, since we don't know how much
+ * the size will change between now and the dbuf dirty call.
+ */
+ if (dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,
+ dn->dn_phys->dn_blkptr[0].blk_birth)) {
+ txh->txh_space_tooverwrite += SPA_MAXBLOCKSIZE;
+ } else {
+ txh->txh_space_towrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_tounref +=
+ BP_GET_ASIZE(dn->dn_phys->dn_blkptr);
+ }
+ return;
+ }
+
+ if (dn->dn_maxblkid > 0 && name) {
+ /*
+ * access the name in this fat-zap so that we'll check
+ * for i/o errors to the leaf blocks, etc.
+ */
+ err = zap_lookup(&dn->dn_objset->os, dn->dn_object, name,
+ 8, 0, NULL);
+ if (err == EIO) {
+ tx->tx_err = err;
+ return;
+ }
+ }
+
+ /*
+ * 3 blocks overwritten: target leaf, ptrtbl block, header block
+ * 3 new blocks written if adding: new split leaf, 2 grown ptrtbl blocks
+ */
+ dmu_tx_count_write(txh, dn->dn_maxblkid * dn->dn_datablksz,
+ (3 + (add ? 3 : 0)) << dn->dn_datablkshift);
+
+ /*
+ * If the modified blocks are scattered to the four winds,
+ * we'll have to modify an indirect twig for each.
+ */
+ epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ for (nblocks = dn->dn_maxblkid >> epbs; nblocks != 0; nblocks >>= epbs)
+ txh->txh_space_towrite += 3 << dn->dn_indblkshift;
+}
+
+void
+dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object)
+{
+ dmu_tx_hold_t *txh;
+
+ ASSERT(tx->tx_txg == 0);
+
+ txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+ object, THT_BONUS, 0, 0);
+ if (txh)
+ dmu_tx_count_dnode(txh);
+}
+
+void
+dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space)
+{
+ dmu_tx_hold_t *txh;
+ ASSERT(tx->tx_txg == 0);
+
+ txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+ DMU_NEW_OBJECT, THT_SPACE, space, 0);
+
+ txh->txh_space_towrite += space;
+}
+
+int
+dmu_tx_holds(dmu_tx_t *tx, uint64_t object)
+{
+ dmu_tx_hold_t *txh;
+ int holds = 0;
+
+ /*
+ * By asserting that the tx is assigned, we're counting the
+ * number of dn_tx_holds, which is the same as the number of
+ * dn_holds. Otherwise, we'd be counting dn_holds, but
+ * dn_tx_holds could be 0.
+ */
+ ASSERT(tx->tx_txg != 0);
+
+ /* if (tx->tx_anyobj == TRUE) */
+ /* return (0); */
+
+ for (txh = list_head(&tx->tx_holds); txh;
+ txh = list_next(&tx->tx_holds, txh)) {
+ if (txh->txh_dnode && txh->txh_dnode->dn_object == object)
+ holds++;
+ }
+
+ return (holds);
+}
+
+#ifdef ZFS_DEBUG
+void
+dmu_tx_dirty_buf(dmu_tx_t *tx, dmu_buf_impl_t *db)
+{
+ dmu_tx_hold_t *txh;
+ int match_object = FALSE, match_offset = FALSE;
+ dnode_t *dn = db->db_dnode;
+
+ ASSERT(tx->tx_txg != 0);
+ ASSERT(tx->tx_objset == NULL || dn->dn_objset == tx->tx_objset->os);
+ ASSERT3U(dn->dn_object, ==, db->db.db_object);
+
+ if (tx->tx_anyobj)
+ return;
+
+ /* XXX No checking on the meta dnode for now */
+ if (db->db.db_object == DMU_META_DNODE_OBJECT)
+ return;
+
+ for (txh = list_head(&tx->tx_holds); txh;
+ txh = list_next(&tx->tx_holds, txh)) {
+ ASSERT(dn == NULL || dn->dn_assigned_txg == tx->tx_txg);
+ if (txh->txh_dnode == dn && txh->txh_type != THT_NEWOBJECT)
+ match_object = TRUE;
+ if (txh->txh_dnode == NULL || txh->txh_dnode == dn) {
+ int datablkshift = dn->dn_datablkshift ?
+ dn->dn_datablkshift : SPA_MAXBLOCKSHIFT;
+ int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ int shift = datablkshift + epbs * db->db_level;
+ uint64_t beginblk = shift >= 64 ? 0 :
+ (txh->txh_arg1 >> shift);
+ uint64_t endblk = shift >= 64 ? 0 :
+ ((txh->txh_arg1 + txh->txh_arg2 - 1) >> shift);
+ uint64_t blkid = db->db_blkid;
+
+ /* XXX txh_arg2 better not be zero... */
+
+ dprintf("found txh type %x beginblk=%llx endblk=%llx\n",
+ txh->txh_type, beginblk, endblk);
+
+ switch (txh->txh_type) {
+ case THT_WRITE:
+ if (blkid >= beginblk && blkid <= endblk)
+ match_offset = TRUE;
+ /*
+ * We will let this hold work for the bonus
+ * buffer so that we don't need to hold it
+ * when creating a new object.
+ */
+ if (blkid == DB_BONUS_BLKID)
+ match_offset = TRUE;
+ /*
+ * They might have to increase nlevels,
+ * thus dirtying the new TLIBs. Or the
+ * might have to change the block size,
+ * thus dirying the new lvl=0 blk=0.
+ */
+ if (blkid == 0)
+ match_offset = TRUE;
+ break;
+ case THT_FREE:
+ /*
+ * We will dirty all the level 1 blocks in
+ * the free range and perhaps the first and
+ * last level 0 block.
+ */
+ if (blkid >= beginblk && (blkid <= endblk ||
+ txh->txh_arg2 == DMU_OBJECT_END))
+ match_offset = TRUE;
+ break;
+ case THT_BONUS:
+ if (blkid == DB_BONUS_BLKID)
+ match_offset = TRUE;
+ break;
+ case THT_ZAP:
+ match_offset = TRUE;
+ break;
+ case THT_NEWOBJECT:
+ match_object = TRUE;
+ break;
+ default:
+ ASSERT(!"bad txh_type");
+ }
+ }
+ if (match_object && match_offset)
+ return;
+ }
+ panic("dirtying dbuf obj=%llx lvl=%u blkid=%llx but not tx_held\n",
+ (u_longlong_t)db->db.db_object, db->db_level,
+ (u_longlong_t)db->db_blkid);
+}
+#endif
+
+static int
+dmu_tx_try_assign(dmu_tx_t *tx, uint64_t txg_how)
+{
+ dmu_tx_hold_t *txh;
+ spa_t *spa = tx->tx_pool->dp_spa;
+ uint64_t memory, asize, fsize, usize;
+ uint64_t towrite, tofree, tooverwrite, tounref, tohold, fudge;
+
+ ASSERT3U(tx->tx_txg, ==, 0);
+
+ if (tx->tx_err)
+ return (tx->tx_err);
+
+ if (spa_suspended(spa)) {
+ /*
+ * If the user has indicated a blocking failure mode
+ * then return ERESTART which will block in dmu_tx_wait().
+ * Otherwise, return EIO so that an error can get
+ * propagated back to the VOP calls.
+ *
+ * Note that we always honor the txg_how flag regardless
+ * of the failuremode setting.
+ */
+ if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE &&
+ txg_how != TXG_WAIT)
+ return (EIO);
+
+ return (ERESTART);
+ }
+
+ tx->tx_txg = txg_hold_open(tx->tx_pool, &tx->tx_txgh);
+ tx->tx_needassign_txh = NULL;
+
+ /*
+ * NB: No error returns are allowed after txg_hold_open, but
+ * before processing the dnode holds, due to the
+ * dmu_tx_unassign() logic.
+ */
+
+ towrite = tofree = tooverwrite = tounref = tohold = fudge = 0;
+ for (txh = list_head(&tx->tx_holds); txh;
+ txh = list_next(&tx->tx_holds, txh)) {
+ dnode_t *dn = txh->txh_dnode;
+ if (dn != NULL) {
+ mutex_enter(&dn->dn_mtx);
+ if (dn->dn_assigned_txg == tx->tx_txg - 1) {
+ mutex_exit(&dn->dn_mtx);
+ tx->tx_needassign_txh = txh;
+ return (ERESTART);
+ }
+ if (dn->dn_assigned_txg == 0)
+ dn->dn_assigned_txg = tx->tx_txg;
+ ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
+ (void) refcount_add(&dn->dn_tx_holds, tx);
+ mutex_exit(&dn->dn_mtx);
+ }
+ towrite += txh->txh_space_towrite;
+ tofree += txh->txh_space_tofree;
+ tooverwrite += txh->txh_space_tooverwrite;
+ tounref += txh->txh_space_tounref;
+ tohold += txh->txh_memory_tohold;
+ fudge += txh->txh_fudge;
+ }
+
+ /*
+ * NB: This check must be after we've held the dnodes, so that
+ * the dmu_tx_unassign() logic will work properly
+ */
+ if (txg_how >= TXG_INITIAL && txg_how != tx->tx_txg)
+ return (ERESTART);
+
+ /*
+ * If a snapshot has been taken since we made our estimates,
+ * assume that we won't be able to free or overwrite anything.
+ */
+ if (tx->tx_objset &&
+ dsl_dataset_prev_snap_txg(tx->tx_objset->os->os_dsl_dataset) >
+ tx->tx_lastsnap_txg) {
+ towrite += tooverwrite;
+ tooverwrite = tofree = 0;
+ }
+
+ /* needed allocation: worst-case estimate of write space */
+ asize = spa_get_asize(tx->tx_pool->dp_spa, towrite + tooverwrite);
+ /* freed space estimate: worst-case overwrite + free estimate */
+ fsize = spa_get_asize(tx->tx_pool->dp_spa, tooverwrite) + tofree;
+ /* convert unrefd space to worst-case estimate */
+ usize = spa_get_asize(tx->tx_pool->dp_spa, tounref);
+ /* calculate memory footprint estimate */
+ memory = towrite + tooverwrite + tohold;
+
+#ifdef ZFS_DEBUG
+ /*
+ * Add in 'tohold' to account for our dirty holds on this memory
+ * XXX - the "fudge" factor is to account for skipped blocks that
+ * we missed because dnode_next_offset() misses in-core-only blocks.
+ */
+ tx->tx_space_towrite = asize +
+ spa_get_asize(tx->tx_pool->dp_spa, tohold + fudge);
+ tx->tx_space_tofree = tofree;
+ tx->tx_space_tooverwrite = tooverwrite;
+ tx->tx_space_tounref = tounref;
+#endif
+
+ if (tx->tx_dir && asize != 0) {
+ int err = dsl_dir_tempreserve_space(tx->tx_dir, memory,
+ asize, fsize, usize, &tx->tx_tempreserve_cookie, tx);
+ if (err)
+ return (err);
+ }
+
+ return (0);
+}
+
+static void
+dmu_tx_unassign(dmu_tx_t *tx)
+{
+ dmu_tx_hold_t *txh;
+
+ if (tx->tx_txg == 0)
+ return;
+
+ txg_rele_to_quiesce(&tx->tx_txgh);
+
+ for (txh = list_head(&tx->tx_holds); txh != tx->tx_needassign_txh;
+ txh = list_next(&tx->tx_holds, txh)) {
+ dnode_t *dn = txh->txh_dnode;
+
+ if (dn == NULL)
+ continue;
+ mutex_enter(&dn->dn_mtx);
+ ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
+
+ if (refcount_remove(&dn->dn_tx_holds, tx) == 0) {
+ dn->dn_assigned_txg = 0;
+ cv_broadcast(&dn->dn_notxholds);
+ }
+ mutex_exit(&dn->dn_mtx);
+ }
+
+ txg_rele_to_sync(&tx->tx_txgh);
+
+ tx->tx_lasttried_txg = tx->tx_txg;
+ tx->tx_txg = 0;
+}
+
+/*
+ * Assign tx to a transaction group. txg_how can be one of:
+ *
+ * (1) TXG_WAIT. If the current open txg is full, waits until there's
+ * a new one. This should be used when you're not holding locks.
+ * If will only fail if we're truly out of space (or over quota).
+ *
+ * (2) TXG_NOWAIT. If we can't assign into the current open txg without
+ * blocking, returns immediately with ERESTART. This should be used
+ * whenever you're holding locks. On an ERESTART error, the caller
+ * should drop locks, do a dmu_tx_wait(tx), and try again.
+ *
+ * (3) A specific txg. Use this if you need to ensure that multiple
+ * transactions all sync in the same txg. Like TXG_NOWAIT, it
+ * returns ERESTART if it can't assign you into the requested txg.
+ */
+int
+dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how)
+{
+ int err;
+
+ ASSERT(tx->tx_txg == 0);
+ ASSERT(txg_how != 0);
+ ASSERT(!dsl_pool_sync_context(tx->tx_pool));
+
+ while ((err = dmu_tx_try_assign(tx, txg_how)) != 0) {
+ dmu_tx_unassign(tx);
+
+ if (err != ERESTART || txg_how != TXG_WAIT)
+ return (err);
+
+ dmu_tx_wait(tx);
+ }
+
+ txg_rele_to_quiesce(&tx->tx_txgh);
+
+ return (0);
+}
+
+void
+dmu_tx_wait(dmu_tx_t *tx)
+{
+ spa_t *spa = tx->tx_pool->dp_spa;
+
+ ASSERT(tx->tx_txg == 0);
+
+ /*
+ * It's possible that the pool has become active after this thread
+ * has tried to obtain a tx. If that's the case then his
+ * tx_lasttried_txg would not have been assigned.
+ */
+ if (spa_suspended(spa) || tx->tx_lasttried_txg == 0) {
+ txg_wait_synced(tx->tx_pool, spa_last_synced_txg(spa) + 1);
+ } else if (tx->tx_needassign_txh) {
+ dnode_t *dn = tx->tx_needassign_txh->txh_dnode;
+
+ mutex_enter(&dn->dn_mtx);
+ while (dn->dn_assigned_txg == tx->tx_lasttried_txg - 1)
+ cv_wait(&dn->dn_notxholds, &dn->dn_mtx);
+ mutex_exit(&dn->dn_mtx);
+ tx->tx_needassign_txh = NULL;
+ } else {
+ txg_wait_open(tx->tx_pool, tx->tx_lasttried_txg + 1);
+ }
+}
+
+void
+dmu_tx_willuse_space(dmu_tx_t *tx, int64_t delta)
+{
+#ifdef ZFS_DEBUG
+ if (tx->tx_dir == NULL || delta == 0)
+ return;
+
+ if (delta > 0) {
+ ASSERT3U(refcount_count(&tx->tx_space_written) + delta, <=,
+ tx->tx_space_towrite);
+ (void) refcount_add_many(&tx->tx_space_written, delta, NULL);
+ } else {
+ (void) refcount_add_many(&tx->tx_space_freed, -delta, NULL);
+ }
+#endif
+}
+
+void
+dmu_tx_commit(dmu_tx_t *tx)
+{
+ dmu_tx_hold_t *txh;
+
+ ASSERT(tx->tx_txg != 0);
+
+ while (txh = list_head(&tx->tx_holds)) {
+ dnode_t *dn = txh->txh_dnode;
+
+ list_remove(&tx->tx_holds, txh);
+ kmem_free(txh, sizeof (dmu_tx_hold_t));
+ if (dn == NULL)
+ continue;
+ mutex_enter(&dn->dn_mtx);
+ ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
+
+ if (refcount_remove(&dn->dn_tx_holds, tx) == 0) {
+ dn->dn_assigned_txg = 0;
+ cv_broadcast(&dn->dn_notxholds);
+ }
+ mutex_exit(&dn->dn_mtx);
+ dnode_rele(dn, tx);
+ }
+
+ if (tx->tx_tempreserve_cookie)
+ dsl_dir_tempreserve_clear(tx->tx_tempreserve_cookie, tx);
+
+ if (tx->tx_anyobj == FALSE)
+ txg_rele_to_sync(&tx->tx_txgh);
+ list_destroy(&tx->tx_holds);
+#ifdef ZFS_DEBUG
+ dprintf("towrite=%llu written=%llu tofree=%llu freed=%llu\n",
+ tx->tx_space_towrite, refcount_count(&tx->tx_space_written),
+ tx->tx_space_tofree, refcount_count(&tx->tx_space_freed));
+ refcount_destroy_many(&tx->tx_space_written,
+ refcount_count(&tx->tx_space_written));
+ refcount_destroy_many(&tx->tx_space_freed,
+ refcount_count(&tx->tx_space_freed));
+#endif
+ kmem_free(tx, sizeof (dmu_tx_t));
+}
+
+void
+dmu_tx_abort(dmu_tx_t *tx)
+{
+ dmu_tx_hold_t *txh;
+
+ ASSERT(tx->tx_txg == 0);
+
+ while (txh = list_head(&tx->tx_holds)) {
+ dnode_t *dn = txh->txh_dnode;
+
+ list_remove(&tx->tx_holds, txh);
+ kmem_free(txh, sizeof (dmu_tx_hold_t));
+ if (dn != NULL)
+ dnode_rele(dn, tx);
+ }
+ list_destroy(&tx->tx_holds);
+#ifdef ZFS_DEBUG
+ refcount_destroy_many(&tx->tx_space_written,
+ refcount_count(&tx->tx_space_written));
+ refcount_destroy_many(&tx->tx_space_freed,
+ refcount_count(&tx->tx_space_freed));
+#endif
+ kmem_free(tx, sizeof (dmu_tx_t));
+}
+
+uint64_t
+dmu_tx_get_txg(dmu_tx_t *tx)
+{
+ ASSERT(tx->tx_txg != 0);
+ return (tx->tx_txg);
+}
diff --git a/module/zfs/dmu_zfetch.c b/module/zfs/dmu_zfetch.c
new file mode 100644
index 000000000..4d79fe98e
--- /dev/null
+++ b/module/zfs/dmu_zfetch.c
@@ -0,0 +1,651 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/dnode.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_zfetch.h>
+#include <sys/dmu.h>
+#include <sys/dbuf.h>
+
+/*
+ * I'm against tune-ables, but these should probably exist as tweakable globals
+ * until we can get this working the way we want it to.
+ */
+
+int zfs_prefetch_disable = 0;
+
+/* max # of streams per zfetch */
+uint32_t zfetch_max_streams = 8;
+/* min time before stream reclaim */
+uint32_t zfetch_min_sec_reap = 2;
+/* max number of blocks to fetch at a time */
+uint32_t zfetch_block_cap = 256;
+/* number of bytes in a array_read at which we stop prefetching (1Mb) */
+uint64_t zfetch_array_rd_sz = 1024 * 1024;
+
+/* forward decls for static routines */
+static int dmu_zfetch_colinear(zfetch_t *, zstream_t *);
+static void dmu_zfetch_dofetch(zfetch_t *, zstream_t *);
+static uint64_t dmu_zfetch_fetch(dnode_t *, uint64_t, uint64_t);
+static uint64_t dmu_zfetch_fetchsz(dnode_t *, uint64_t, uint64_t);
+static int dmu_zfetch_find(zfetch_t *, zstream_t *, int);
+static int dmu_zfetch_stream_insert(zfetch_t *, zstream_t *);
+static zstream_t *dmu_zfetch_stream_reclaim(zfetch_t *);
+static void dmu_zfetch_stream_remove(zfetch_t *, zstream_t *);
+static int dmu_zfetch_streams_equal(zstream_t *, zstream_t *);
+
+/*
+ * Given a zfetch structure and a zstream structure, determine whether the
+ * blocks to be read are part of a co-linear pair of existing prefetch
+ * streams. If a set is found, coalesce the streams, removing one, and
+ * configure the prefetch so it looks for a strided access pattern.
+ *
+ * In other words: if we find two sequential access streams that are
+ * the same length and distance N appart, and this read is N from the
+ * last stream, then we are probably in a strided access pattern. So
+ * combine the two sequential streams into a single strided stream.
+ *
+ * If no co-linear streams are found, return NULL.
+ */
+static int
+dmu_zfetch_colinear(zfetch_t *zf, zstream_t *zh)
+{
+ zstream_t *z_walk;
+ zstream_t *z_comp;
+
+ if (! rw_tryenter(&zf->zf_rwlock, RW_WRITER))
+ return (0);
+
+ if (zh == NULL) {
+ rw_exit(&zf->zf_rwlock);
+ return (0);
+ }
+
+ for (z_walk = list_head(&zf->zf_stream); z_walk;
+ z_walk = list_next(&zf->zf_stream, z_walk)) {
+ for (z_comp = list_next(&zf->zf_stream, z_walk); z_comp;
+ z_comp = list_next(&zf->zf_stream, z_comp)) {
+ int64_t diff;
+
+ if (z_walk->zst_len != z_walk->zst_stride ||
+ z_comp->zst_len != z_comp->zst_stride) {
+ continue;
+ }
+
+ diff = z_comp->zst_offset - z_walk->zst_offset;
+ if (z_comp->zst_offset + diff == zh->zst_offset) {
+ z_walk->zst_offset = zh->zst_offset;
+ z_walk->zst_direction = diff < 0 ? -1 : 1;
+ z_walk->zst_stride =
+ diff * z_walk->zst_direction;
+ z_walk->zst_ph_offset =
+ zh->zst_offset + z_walk->zst_stride;
+ dmu_zfetch_stream_remove(zf, z_comp);
+ mutex_destroy(&z_comp->zst_lock);
+ kmem_free(z_comp, sizeof (zstream_t));
+
+ dmu_zfetch_dofetch(zf, z_walk);
+
+ rw_exit(&zf->zf_rwlock);
+ return (1);
+ }
+
+ diff = z_walk->zst_offset - z_comp->zst_offset;
+ if (z_walk->zst_offset + diff == zh->zst_offset) {
+ z_walk->zst_offset = zh->zst_offset;
+ z_walk->zst_direction = diff < 0 ? -1 : 1;
+ z_walk->zst_stride =
+ diff * z_walk->zst_direction;
+ z_walk->zst_ph_offset =
+ zh->zst_offset + z_walk->zst_stride;
+ dmu_zfetch_stream_remove(zf, z_comp);
+ mutex_destroy(&z_comp->zst_lock);
+ kmem_free(z_comp, sizeof (zstream_t));
+
+ dmu_zfetch_dofetch(zf, z_walk);
+
+ rw_exit(&zf->zf_rwlock);
+ return (1);
+ }
+ }
+ }
+
+ rw_exit(&zf->zf_rwlock);
+ return (0);
+}
+
+/*
+ * Given a zstream_t, determine the bounds of the prefetch. Then call the
+ * routine that actually prefetches the individual blocks.
+ */
+static void
+dmu_zfetch_dofetch(zfetch_t *zf, zstream_t *zs)
+{
+ uint64_t prefetch_tail;
+ uint64_t prefetch_limit;
+ uint64_t prefetch_ofst;
+ uint64_t prefetch_len;
+ uint64_t blocks_fetched;
+
+ zs->zst_stride = MAX((int64_t)zs->zst_stride, zs->zst_len);
+ zs->zst_cap = MIN(zfetch_block_cap, 2 * zs->zst_cap);
+
+ prefetch_tail = MAX((int64_t)zs->zst_ph_offset,
+ (int64_t)(zs->zst_offset + zs->zst_stride));
+ /*
+ * XXX: use a faster division method?
+ */
+ prefetch_limit = zs->zst_offset + zs->zst_len +
+ (zs->zst_cap * zs->zst_stride) / zs->zst_len;
+
+ while (prefetch_tail < prefetch_limit) {
+ prefetch_ofst = zs->zst_offset + zs->zst_direction *
+ (prefetch_tail - zs->zst_offset);
+
+ prefetch_len = zs->zst_len;
+
+ /*
+ * Don't prefetch beyond the end of the file, if working
+ * backwards.
+ */
+ if ((zs->zst_direction == ZFETCH_BACKWARD) &&
+ (prefetch_ofst > prefetch_tail)) {
+ prefetch_len += prefetch_ofst;
+ prefetch_ofst = 0;
+ }
+
+ /* don't prefetch more than we're supposed to */
+ if (prefetch_len > zs->zst_len)
+ break;
+
+ blocks_fetched = dmu_zfetch_fetch(zf->zf_dnode,
+ prefetch_ofst, zs->zst_len);
+
+ prefetch_tail += zs->zst_stride;
+ /* stop if we've run out of stuff to prefetch */
+ if (blocks_fetched < zs->zst_len)
+ break;
+ }
+ zs->zst_ph_offset = prefetch_tail;
+ zs->zst_last = lbolt;
+}
+
+/*
+ * This takes a pointer to a zfetch structure and a dnode. It performs the
+ * necessary setup for the zfetch structure, grokking data from the
+ * associated dnode.
+ */
+void
+dmu_zfetch_init(zfetch_t *zf, dnode_t *dno)
+{
+ if (zf == NULL) {
+ return;
+ }
+
+ zf->zf_dnode = dno;
+ zf->zf_stream_cnt = 0;
+ zf->zf_alloc_fail = 0;
+
+ list_create(&zf->zf_stream, sizeof (zstream_t),
+ offsetof(zstream_t, zst_node));
+
+ rw_init(&zf->zf_rwlock, NULL, RW_DEFAULT, NULL);
+}
+
+/*
+ * This function computes the actual size, in blocks, that can be prefetched,
+ * and fetches it.
+ */
+static uint64_t
+dmu_zfetch_fetch(dnode_t *dn, uint64_t blkid, uint64_t nblks)
+{
+ uint64_t fetchsz;
+ uint64_t i;
+
+ fetchsz = dmu_zfetch_fetchsz(dn, blkid, nblks);
+
+ for (i = 0; i < fetchsz; i++) {
+ dbuf_prefetch(dn, blkid + i);
+ }
+
+ return (fetchsz);
+}
+
+/*
+ * this function returns the number of blocks that would be prefetched, based
+ * upon the supplied dnode, blockid, and nblks. This is used so that we can
+ * update streams in place, and then prefetch with their old value after the
+ * fact. This way, we can delay the prefetch, but subsequent accesses to the
+ * stream won't result in the same data being prefetched multiple times.
+ */
+static uint64_t
+dmu_zfetch_fetchsz(dnode_t *dn, uint64_t blkid, uint64_t nblks)
+{
+ uint64_t fetchsz;
+
+ if (blkid > dn->dn_maxblkid) {
+ return (0);
+ }
+
+ /* compute fetch size */
+ if (blkid + nblks + 1 > dn->dn_maxblkid) {
+ fetchsz = (dn->dn_maxblkid - blkid) + 1;
+ ASSERT(blkid + fetchsz - 1 <= dn->dn_maxblkid);
+ } else {
+ fetchsz = nblks;
+ }
+
+
+ return (fetchsz);
+}
+
+/*
+ * given a zfetch and a zsearch structure, see if there is an associated zstream
+ * for this block read. If so, it starts a prefetch for the stream it
+ * located and returns true, otherwise it returns false
+ */
+static int
+dmu_zfetch_find(zfetch_t *zf, zstream_t *zh, int prefetched)
+{
+ zstream_t *zs;
+ int64_t diff;
+ int reset = !prefetched;
+ int rc = 0;
+
+ if (zh == NULL)
+ return (0);
+
+ /*
+ * XXX: This locking strategy is a bit coarse; however, it's impact has
+ * yet to be tested. If this turns out to be an issue, it can be
+ * modified in a number of different ways.
+ */
+
+ rw_enter(&zf->zf_rwlock, RW_READER);
+top:
+
+ for (zs = list_head(&zf->zf_stream); zs;
+ zs = list_next(&zf->zf_stream, zs)) {
+
+ /*
+ * XXX - should this be an assert?
+ */
+ if (zs->zst_len == 0) {
+ /* bogus stream */
+ continue;
+ }
+
+ /*
+ * We hit this case when we are in a strided prefetch stream:
+ * we will read "len" blocks before "striding".
+ */
+ if (zh->zst_offset >= zs->zst_offset &&
+ zh->zst_offset < zs->zst_offset + zs->zst_len) {
+ /* already fetched */
+ rc = 1;
+ goto out;
+ }
+
+ /*
+ * This is the forward sequential read case: we increment
+ * len by one each time we hit here, so we will enter this
+ * case on every read.
+ */
+ if (zh->zst_offset == zs->zst_offset + zs->zst_len) {
+
+ reset = !prefetched && zs->zst_len > 1;
+
+ mutex_enter(&zs->zst_lock);
+
+ if (zh->zst_offset != zs->zst_offset + zs->zst_len) {
+ mutex_exit(&zs->zst_lock);
+ goto top;
+ }
+ zs->zst_len += zh->zst_len;
+ diff = zs->zst_len - zfetch_block_cap;
+ if (diff > 0) {
+ zs->zst_offset += diff;
+ zs->zst_len = zs->zst_len > diff ?
+ zs->zst_len - diff : 0;
+ }
+ zs->zst_direction = ZFETCH_FORWARD;
+
+ break;
+
+ /*
+ * Same as above, but reading backwards through the file.
+ */
+ } else if (zh->zst_offset == zs->zst_offset - zh->zst_len) {
+ /* backwards sequential access */
+
+ reset = !prefetched && zs->zst_len > 1;
+
+ mutex_enter(&zs->zst_lock);
+
+ if (zh->zst_offset != zs->zst_offset - zh->zst_len) {
+ mutex_exit(&zs->zst_lock);
+ goto top;
+ }
+
+ zs->zst_offset = zs->zst_offset > zh->zst_len ?
+ zs->zst_offset - zh->zst_len : 0;
+ zs->zst_ph_offset = zs->zst_ph_offset > zh->zst_len ?
+ zs->zst_ph_offset - zh->zst_len : 0;
+ zs->zst_len += zh->zst_len;
+
+ diff = zs->zst_len - zfetch_block_cap;
+ if (diff > 0) {
+ zs->zst_ph_offset = zs->zst_ph_offset > diff ?
+ zs->zst_ph_offset - diff : 0;
+ zs->zst_len = zs->zst_len > diff ?
+ zs->zst_len - diff : zs->zst_len;
+ }
+ zs->zst_direction = ZFETCH_BACKWARD;
+
+ break;
+
+ } else if ((zh->zst_offset - zs->zst_offset - zs->zst_stride <
+ zs->zst_len) && (zs->zst_len != zs->zst_stride)) {
+ /* strided forward access */
+
+ mutex_enter(&zs->zst_lock);
+
+ if ((zh->zst_offset - zs->zst_offset - zs->zst_stride >=
+ zs->zst_len) || (zs->zst_len == zs->zst_stride)) {
+ mutex_exit(&zs->zst_lock);
+ goto top;
+ }
+
+ zs->zst_offset += zs->zst_stride;
+ zs->zst_direction = ZFETCH_FORWARD;
+
+ break;
+
+ } else if ((zh->zst_offset - zs->zst_offset + zs->zst_stride <
+ zs->zst_len) && (zs->zst_len != zs->zst_stride)) {
+ /* strided reverse access */
+
+ mutex_enter(&zs->zst_lock);
+
+ if ((zh->zst_offset - zs->zst_offset + zs->zst_stride >=
+ zs->zst_len) || (zs->zst_len == zs->zst_stride)) {
+ mutex_exit(&zs->zst_lock);
+ goto top;
+ }
+
+ zs->zst_offset = zs->zst_offset > zs->zst_stride ?
+ zs->zst_offset - zs->zst_stride : 0;
+ zs->zst_ph_offset = (zs->zst_ph_offset >
+ (2 * zs->zst_stride)) ?
+ (zs->zst_ph_offset - (2 * zs->zst_stride)) : 0;
+ zs->zst_direction = ZFETCH_BACKWARD;
+
+ break;
+ }
+ }
+
+ if (zs) {
+ if (reset) {
+ zstream_t *remove = zs;
+
+ rc = 0;
+ mutex_exit(&zs->zst_lock);
+ rw_exit(&zf->zf_rwlock);
+ rw_enter(&zf->zf_rwlock, RW_WRITER);
+ /*
+ * Relocate the stream, in case someone removes
+ * it while we were acquiring the WRITER lock.
+ */
+ for (zs = list_head(&zf->zf_stream); zs;
+ zs = list_next(&zf->zf_stream, zs)) {
+ if (zs == remove) {
+ dmu_zfetch_stream_remove(zf, zs);
+ mutex_destroy(&zs->zst_lock);
+ kmem_free(zs, sizeof (zstream_t));
+ break;
+ }
+ }
+ } else {
+ rc = 1;
+ dmu_zfetch_dofetch(zf, zs);
+ mutex_exit(&zs->zst_lock);
+ }
+ }
+out:
+ rw_exit(&zf->zf_rwlock);
+ return (rc);
+}
+
+/*
+ * Clean-up state associated with a zfetch structure. This frees allocated
+ * structure members, empties the zf_stream tree, and generally makes things
+ * nice. This doesn't free the zfetch_t itself, that's left to the caller.
+ */
+void
+dmu_zfetch_rele(zfetch_t *zf)
+{
+ zstream_t *zs;
+ zstream_t *zs_next;
+
+ ASSERT(!RW_LOCK_HELD(&zf->zf_rwlock));
+
+ for (zs = list_head(&zf->zf_stream); zs; zs = zs_next) {
+ zs_next = list_next(&zf->zf_stream, zs);
+
+ list_remove(&zf->zf_stream, zs);
+ mutex_destroy(&zs->zst_lock);
+ kmem_free(zs, sizeof (zstream_t));
+ }
+ list_destroy(&zf->zf_stream);
+ rw_destroy(&zf->zf_rwlock);
+
+ zf->zf_dnode = NULL;
+}
+
+/*
+ * Given a zfetch and zstream structure, insert the zstream structure into the
+ * AVL tree contained within the zfetch structure. Peform the appropriate
+ * book-keeping. It is possible that another thread has inserted a stream which
+ * matches one that we are about to insert, so we must be sure to check for this
+ * case. If one is found, return failure, and let the caller cleanup the
+ * duplicates.
+ */
+static int
+dmu_zfetch_stream_insert(zfetch_t *zf, zstream_t *zs)
+{
+ zstream_t *zs_walk;
+ zstream_t *zs_next;
+
+ ASSERT(RW_WRITE_HELD(&zf->zf_rwlock));
+
+ for (zs_walk = list_head(&zf->zf_stream); zs_walk; zs_walk = zs_next) {
+ zs_next = list_next(&zf->zf_stream, zs_walk);
+
+ if (dmu_zfetch_streams_equal(zs_walk, zs)) {
+ return (0);
+ }
+ }
+
+ list_insert_head(&zf->zf_stream, zs);
+ zf->zf_stream_cnt++;
+
+ return (1);
+}
+
+
+/*
+ * Walk the list of zstreams in the given zfetch, find an old one (by time), and
+ * reclaim it for use by the caller.
+ */
+static zstream_t *
+dmu_zfetch_stream_reclaim(zfetch_t *zf)
+{
+ zstream_t *zs;
+
+ if (! rw_tryenter(&zf->zf_rwlock, RW_WRITER))
+ return (0);
+
+ for (zs = list_head(&zf->zf_stream); zs;
+ zs = list_next(&zf->zf_stream, zs)) {
+
+ if (((lbolt - zs->zst_last) / hz) > zfetch_min_sec_reap)
+ break;
+ }
+
+ if (zs) {
+ dmu_zfetch_stream_remove(zf, zs);
+ mutex_destroy(&zs->zst_lock);
+ bzero(zs, sizeof (zstream_t));
+ } else {
+ zf->zf_alloc_fail++;
+ }
+ rw_exit(&zf->zf_rwlock);
+
+ return (zs);
+}
+
+/*
+ * Given a zfetch and zstream structure, remove the zstream structure from its
+ * container in the zfetch structure. Perform the appropriate book-keeping.
+ */
+static void
+dmu_zfetch_stream_remove(zfetch_t *zf, zstream_t *zs)
+{
+ ASSERT(RW_WRITE_HELD(&zf->zf_rwlock));
+
+ list_remove(&zf->zf_stream, zs);
+ zf->zf_stream_cnt--;
+}
+
+static int
+dmu_zfetch_streams_equal(zstream_t *zs1, zstream_t *zs2)
+{
+ if (zs1->zst_offset != zs2->zst_offset)
+ return (0);
+
+ if (zs1->zst_len != zs2->zst_len)
+ return (0);
+
+ if (zs1->zst_stride != zs2->zst_stride)
+ return (0);
+
+ if (zs1->zst_ph_offset != zs2->zst_ph_offset)
+ return (0);
+
+ if (zs1->zst_cap != zs2->zst_cap)
+ return (0);
+
+ if (zs1->zst_direction != zs2->zst_direction)
+ return (0);
+
+ return (1);
+}
+
+/*
+ * This is the prefetch entry point. It calls all of the other dmu_zfetch
+ * routines to create, delete, find, or operate upon prefetch streams.
+ */
+void
+dmu_zfetch(zfetch_t *zf, uint64_t offset, uint64_t size, int prefetched)
+{
+ zstream_t zst;
+ zstream_t *newstream;
+ int fetched;
+ int inserted;
+ unsigned int blkshft;
+ uint64_t blksz;
+
+ if (zfs_prefetch_disable)
+ return;
+
+ /* files that aren't ln2 blocksz are only one block -- nothing to do */
+ if (!zf->zf_dnode->dn_datablkshift)
+ return;
+
+ /* convert offset and size, into blockid and nblocks */
+ blkshft = zf->zf_dnode->dn_datablkshift;
+ blksz = (1 << blkshft);
+
+ bzero(&zst, sizeof (zstream_t));
+ zst.zst_offset = offset >> blkshft;
+ zst.zst_len = (P2ROUNDUP(offset + size, blksz) -
+ P2ALIGN(offset, blksz)) >> blkshft;
+
+ fetched = dmu_zfetch_find(zf, &zst, prefetched);
+ if (!fetched) {
+ fetched = dmu_zfetch_colinear(zf, &zst);
+ }
+
+ if (!fetched) {
+ newstream = dmu_zfetch_stream_reclaim(zf);
+
+ /*
+ * we still couldn't find a stream, drop the lock, and allocate
+ * one if possible. Otherwise, give up and go home.
+ */
+ if (newstream == NULL) {
+ uint64_t maxblocks;
+ uint32_t max_streams;
+ uint32_t cur_streams;
+
+ cur_streams = zf->zf_stream_cnt;
+ maxblocks = zf->zf_dnode->dn_maxblkid;
+
+ max_streams = MIN(zfetch_max_streams,
+ (maxblocks / zfetch_block_cap));
+ if (max_streams == 0) {
+ max_streams++;
+ }
+
+ if (cur_streams >= max_streams) {
+ return;
+ }
+
+ newstream = kmem_zalloc(sizeof (zstream_t), KM_SLEEP);
+ }
+
+ newstream->zst_offset = zst.zst_offset;
+ newstream->zst_len = zst.zst_len;
+ newstream->zst_stride = zst.zst_len;
+ newstream->zst_ph_offset = zst.zst_len + zst.zst_offset;
+ newstream->zst_cap = zst.zst_len;
+ newstream->zst_direction = ZFETCH_FORWARD;
+ newstream->zst_last = lbolt;
+
+ mutex_init(&newstream->zst_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ rw_enter(&zf->zf_rwlock, RW_WRITER);
+ inserted = dmu_zfetch_stream_insert(zf, newstream);
+ rw_exit(&zf->zf_rwlock);
+
+ if (!inserted) {
+ mutex_destroy(&newstream->zst_lock);
+ kmem_free(newstream, sizeof (zstream_t));
+ }
+ }
+}
diff --git a/module/zfs/dnode.c b/module/zfs/dnode.c
new file mode 100644
index 000000000..e77834d60
--- /dev/null
+++ b/module/zfs/dnode.c
@@ -0,0 +1,1443 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/dbuf.h>
+#include <sys/dnode.h>
+#include <sys/dmu.h>
+#include <sys/dmu_impl.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_dataset.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu_zfetch.h>
+
+static int free_range_compar(const void *node1, const void *node2);
+
+static kmem_cache_t *dnode_cache;
+
+static dnode_phys_t dnode_phys_zero;
+
+int zfs_default_bs = SPA_MINBLOCKSHIFT;
+int zfs_default_ibs = DN_MAX_INDBLKSHIFT;
+
+/* ARGSUSED */
+static int
+dnode_cons(void *arg, void *unused, int kmflag)
+{
+ int i;
+ dnode_t *dn = arg;
+ bzero(dn, sizeof (dnode_t));
+
+ rw_init(&dn->dn_struct_rwlock, NULL, RW_DEFAULT, NULL);
+ mutex_init(&dn->dn_mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&dn->dn_dbufs_mtx, NULL, MUTEX_DEFAULT, NULL);
+ refcount_create(&dn->dn_holds);
+ refcount_create(&dn->dn_tx_holds);
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ avl_create(&dn->dn_ranges[i], free_range_compar,
+ sizeof (free_range_t),
+ offsetof(struct free_range, fr_node));
+ list_create(&dn->dn_dirty_records[i],
+ sizeof (dbuf_dirty_record_t),
+ offsetof(dbuf_dirty_record_t, dr_dirty_node));
+ }
+
+ list_create(&dn->dn_dbufs, sizeof (dmu_buf_impl_t),
+ offsetof(dmu_buf_impl_t, db_link));
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dnode_dest(void *arg, void *unused)
+{
+ int i;
+ dnode_t *dn = arg;
+
+ rw_destroy(&dn->dn_struct_rwlock);
+ mutex_destroy(&dn->dn_mtx);
+ mutex_destroy(&dn->dn_dbufs_mtx);
+ refcount_destroy(&dn->dn_holds);
+ refcount_destroy(&dn->dn_tx_holds);
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ avl_destroy(&dn->dn_ranges[i]);
+ list_destroy(&dn->dn_dirty_records[i]);
+ }
+
+ list_destroy(&dn->dn_dbufs);
+}
+
+void
+dnode_init(void)
+{
+ dnode_cache = kmem_cache_create("dnode_t",
+ sizeof (dnode_t),
+ 0, dnode_cons, dnode_dest, NULL, NULL, NULL, 0);
+}
+
+void
+dnode_fini(void)
+{
+ kmem_cache_destroy(dnode_cache);
+}
+
+
+#ifdef ZFS_DEBUG
+void
+dnode_verify(dnode_t *dn)
+{
+ int drop_struct_lock = FALSE;
+
+ ASSERT(dn->dn_phys);
+ ASSERT(dn->dn_objset);
+
+ ASSERT(dn->dn_phys->dn_type < DMU_OT_NUMTYPES);
+
+ if (!(zfs_flags & ZFS_DEBUG_DNODE_VERIFY))
+ return;
+
+ if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) {
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ drop_struct_lock = TRUE;
+ }
+ if (dn->dn_phys->dn_type != DMU_OT_NONE || dn->dn_allocated_txg != 0) {
+ int i;
+ ASSERT3U(dn->dn_indblkshift, >=, 0);
+ ASSERT3U(dn->dn_indblkshift, <=, SPA_MAXBLOCKSHIFT);
+ if (dn->dn_datablkshift) {
+ ASSERT3U(dn->dn_datablkshift, >=, SPA_MINBLOCKSHIFT);
+ ASSERT3U(dn->dn_datablkshift, <=, SPA_MAXBLOCKSHIFT);
+ ASSERT3U(1<<dn->dn_datablkshift, ==, dn->dn_datablksz);
+ }
+ ASSERT3U(dn->dn_nlevels, <=, 30);
+ ASSERT3U(dn->dn_type, <=, DMU_OT_NUMTYPES);
+ ASSERT3U(dn->dn_nblkptr, >=, 1);
+ ASSERT3U(dn->dn_nblkptr, <=, DN_MAX_NBLKPTR);
+ ASSERT3U(dn->dn_bonuslen, <=, DN_MAX_BONUSLEN);
+ ASSERT3U(dn->dn_datablksz, ==,
+ dn->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ ASSERT3U(ISP2(dn->dn_datablksz), ==, dn->dn_datablkshift != 0);
+ ASSERT3U((dn->dn_nblkptr - 1) * sizeof (blkptr_t) +
+ dn->dn_bonuslen, <=, DN_MAX_BONUSLEN);
+ for (i = 0; i < TXG_SIZE; i++) {
+ ASSERT3U(dn->dn_next_nlevels[i], <=, dn->dn_nlevels);
+ }
+ }
+ if (dn->dn_phys->dn_type != DMU_OT_NONE)
+ ASSERT3U(dn->dn_phys->dn_nlevels, <=, dn->dn_nlevels);
+ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || dn->dn_dbuf != NULL);
+ if (dn->dn_dbuf != NULL) {
+ ASSERT3P(dn->dn_phys, ==,
+ (dnode_phys_t *)dn->dn_dbuf->db.db_data +
+ (dn->dn_object % (dn->dn_dbuf->db.db_size >> DNODE_SHIFT)));
+ }
+ if (drop_struct_lock)
+ rw_exit(&dn->dn_struct_rwlock);
+}
+#endif
+
+void
+dnode_byteswap(dnode_phys_t *dnp)
+{
+ uint64_t *buf64 = (void*)&dnp->dn_blkptr;
+ int i;
+
+ if (dnp->dn_type == DMU_OT_NONE) {
+ bzero(dnp, sizeof (dnode_phys_t));
+ return;
+ }
+
+ dnp->dn_datablkszsec = BSWAP_16(dnp->dn_datablkszsec);
+ dnp->dn_bonuslen = BSWAP_16(dnp->dn_bonuslen);
+ dnp->dn_maxblkid = BSWAP_64(dnp->dn_maxblkid);
+ dnp->dn_used = BSWAP_64(dnp->dn_used);
+
+ /*
+ * dn_nblkptr is only one byte, so it's OK to read it in either
+ * byte order. We can't read dn_bouslen.
+ */
+ ASSERT(dnp->dn_indblkshift <= SPA_MAXBLOCKSHIFT);
+ ASSERT(dnp->dn_nblkptr <= DN_MAX_NBLKPTR);
+ for (i = 0; i < dnp->dn_nblkptr * sizeof (blkptr_t)/8; i++)
+ buf64[i] = BSWAP_64(buf64[i]);
+
+ /*
+ * OK to check dn_bonuslen for zero, because it won't matter if
+ * we have the wrong byte order. This is necessary because the
+ * dnode dnode is smaller than a regular dnode.
+ */
+ if (dnp->dn_bonuslen != 0) {
+ /*
+ * Note that the bonus length calculated here may be
+ * longer than the actual bonus buffer. This is because
+ * we always put the bonus buffer after the last block
+ * pointer (instead of packing it against the end of the
+ * dnode buffer).
+ */
+ int off = (dnp->dn_nblkptr-1) * sizeof (blkptr_t);
+ size_t len = DN_MAX_BONUSLEN - off;
+ ASSERT3U(dnp->dn_bonustype, <, DMU_OT_NUMTYPES);
+ dmu_ot[dnp->dn_bonustype].ot_byteswap(dnp->dn_bonus + off, len);
+ }
+}
+
+void
+dnode_buf_byteswap(void *vbuf, size_t size)
+{
+ dnode_phys_t *buf = vbuf;
+ int i;
+
+ ASSERT3U(sizeof (dnode_phys_t), ==, (1<<DNODE_SHIFT));
+ ASSERT((size & (sizeof (dnode_phys_t)-1)) == 0);
+
+ size >>= DNODE_SHIFT;
+ for (i = 0; i < size; i++) {
+ dnode_byteswap(buf);
+ buf++;
+ }
+}
+
+static int
+free_range_compar(const void *node1, const void *node2)
+{
+ const free_range_t *rp1 = node1;
+ const free_range_t *rp2 = node2;
+
+ if (rp1->fr_blkid < rp2->fr_blkid)
+ return (-1);
+ else if (rp1->fr_blkid > rp2->fr_blkid)
+ return (1);
+ else return (0);
+}
+
+void
+dnode_setbonuslen(dnode_t *dn, int newsize, dmu_tx_t *tx)
+{
+ ASSERT3U(refcount_count(&dn->dn_holds), >=, 1);
+
+ dnode_setdirty(dn, tx);
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ ASSERT3U(newsize, <=, DN_MAX_BONUSLEN -
+ (dn->dn_nblkptr-1) * sizeof (blkptr_t));
+ dn->dn_bonuslen = newsize;
+ if (newsize == 0)
+ dn->dn_next_bonuslen[tx->tx_txg & TXG_MASK] = DN_ZERO_BONUSLEN;
+ else
+ dn->dn_next_bonuslen[tx->tx_txg & TXG_MASK] = dn->dn_bonuslen;
+ rw_exit(&dn->dn_struct_rwlock);
+}
+
+static void
+dnode_setdblksz(dnode_t *dn, int size)
+{
+ ASSERT3U(P2PHASE(size, SPA_MINBLOCKSIZE), ==, 0);
+ ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
+ ASSERT3U(size, >=, SPA_MINBLOCKSIZE);
+ ASSERT3U(size >> SPA_MINBLOCKSHIFT, <,
+ 1<<(sizeof (dn->dn_phys->dn_datablkszsec) * 8));
+ dn->dn_datablksz = size;
+ dn->dn_datablkszsec = size >> SPA_MINBLOCKSHIFT;
+ dn->dn_datablkshift = ISP2(size) ? highbit(size - 1) : 0;
+}
+
+static dnode_t *
+dnode_create(objset_impl_t *os, dnode_phys_t *dnp, dmu_buf_impl_t *db,
+ uint64_t object)
+{
+ dnode_t *dn = kmem_cache_alloc(dnode_cache, KM_SLEEP);
+ (void) dnode_cons(dn, NULL, 0); /* XXX */
+
+ dn->dn_objset = os;
+ dn->dn_object = object;
+ dn->dn_dbuf = db;
+ dn->dn_phys = dnp;
+
+ if (dnp->dn_datablkszsec)
+ dnode_setdblksz(dn, dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ dn->dn_indblkshift = dnp->dn_indblkshift;
+ dn->dn_nlevels = dnp->dn_nlevels;
+ dn->dn_type = dnp->dn_type;
+ dn->dn_nblkptr = dnp->dn_nblkptr;
+ dn->dn_checksum = dnp->dn_checksum;
+ dn->dn_compress = dnp->dn_compress;
+ dn->dn_bonustype = dnp->dn_bonustype;
+ dn->dn_bonuslen = dnp->dn_bonuslen;
+ dn->dn_maxblkid = dnp->dn_maxblkid;
+
+ dmu_zfetch_init(&dn->dn_zfetch, dn);
+
+ ASSERT(dn->dn_phys->dn_type < DMU_OT_NUMTYPES);
+ mutex_enter(&os->os_lock);
+ list_insert_head(&os->os_dnodes, dn);
+ mutex_exit(&os->os_lock);
+
+ arc_space_consume(sizeof (dnode_t));
+ return (dn);
+}
+
+static void
+dnode_destroy(dnode_t *dn)
+{
+ objset_impl_t *os = dn->dn_objset;
+
+#ifdef ZFS_DEBUG
+ int i;
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
+ ASSERT(NULL == list_head(&dn->dn_dirty_records[i]));
+ ASSERT(0 == avl_numnodes(&dn->dn_ranges[i]));
+ }
+ ASSERT(NULL == list_head(&dn->dn_dbufs));
+#endif
+
+ mutex_enter(&os->os_lock);
+ list_remove(&os->os_dnodes, dn);
+ mutex_exit(&os->os_lock);
+
+ if (dn->dn_dirtyctx_firstset) {
+ kmem_free(dn->dn_dirtyctx_firstset, 1);
+ dn->dn_dirtyctx_firstset = NULL;
+ }
+ dmu_zfetch_rele(&dn->dn_zfetch);
+ if (dn->dn_bonus) {
+ mutex_enter(&dn->dn_bonus->db_mtx);
+ dbuf_evict(dn->dn_bonus);
+ dn->dn_bonus = NULL;
+ }
+ kmem_cache_free(dnode_cache, dn);
+ arc_space_return(sizeof (dnode_t));
+}
+
+void
+dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ int i;
+
+ if (blocksize == 0)
+ blocksize = 1 << zfs_default_bs;
+ else if (blocksize > SPA_MAXBLOCKSIZE)
+ blocksize = SPA_MAXBLOCKSIZE;
+ else
+ blocksize = P2ROUNDUP(blocksize, SPA_MINBLOCKSIZE);
+
+ if (ibs == 0)
+ ibs = zfs_default_ibs;
+
+ ibs = MIN(MAX(ibs, DN_MIN_INDBLKSHIFT), DN_MAX_INDBLKSHIFT);
+
+ dprintf("os=%p obj=%llu txg=%llu blocksize=%d ibs=%d\n", dn->dn_objset,
+ dn->dn_object, tx->tx_txg, blocksize, ibs);
+
+ ASSERT(dn->dn_type == DMU_OT_NONE);
+ ASSERT(bcmp(dn->dn_phys, &dnode_phys_zero, sizeof (dnode_phys_t)) == 0);
+ ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE);
+ ASSERT(ot != DMU_OT_NONE);
+ ASSERT3U(ot, <, DMU_OT_NUMTYPES);
+ ASSERT((bonustype == DMU_OT_NONE && bonuslen == 0) ||
+ (bonustype != DMU_OT_NONE && bonuslen != 0));
+ ASSERT3U(bonustype, <, DMU_OT_NUMTYPES);
+ ASSERT3U(bonuslen, <=, DN_MAX_BONUSLEN);
+ ASSERT(dn->dn_type == DMU_OT_NONE);
+ ASSERT3U(dn->dn_maxblkid, ==, 0);
+ ASSERT3U(dn->dn_allocated_txg, ==, 0);
+ ASSERT3U(dn->dn_assigned_txg, ==, 0);
+ ASSERT(refcount_is_zero(&dn->dn_tx_holds));
+ ASSERT3U(refcount_count(&dn->dn_holds), <=, 1);
+ ASSERT3P(list_head(&dn->dn_dbufs), ==, NULL);
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ ASSERT3U(dn->dn_next_nlevels[i], ==, 0);
+ ASSERT3U(dn->dn_next_indblkshift[i], ==, 0);
+ ASSERT3U(dn->dn_next_bonuslen[i], ==, 0);
+ ASSERT3U(dn->dn_next_blksz[i], ==, 0);
+ ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
+ ASSERT3P(list_head(&dn->dn_dirty_records[i]), ==, NULL);
+ ASSERT3U(avl_numnodes(&dn->dn_ranges[i]), ==, 0);
+ }
+
+ dn->dn_type = ot;
+ dnode_setdblksz(dn, blocksize);
+ dn->dn_indblkshift = ibs;
+ dn->dn_nlevels = 1;
+ dn->dn_nblkptr = 1 + ((DN_MAX_BONUSLEN - bonuslen) >> SPA_BLKPTRSHIFT);
+ dn->dn_bonustype = bonustype;
+ dn->dn_bonuslen = bonuslen;
+ dn->dn_checksum = ZIO_CHECKSUM_INHERIT;
+ dn->dn_compress = ZIO_COMPRESS_INHERIT;
+ dn->dn_dirtyctx = 0;
+
+ dn->dn_free_txg = 0;
+ if (dn->dn_dirtyctx_firstset) {
+ kmem_free(dn->dn_dirtyctx_firstset, 1);
+ dn->dn_dirtyctx_firstset = NULL;
+ }
+
+ dn->dn_allocated_txg = tx->tx_txg;
+
+ dnode_setdirty(dn, tx);
+ dn->dn_next_indblkshift[tx->tx_txg & TXG_MASK] = ibs;
+ dn->dn_next_bonuslen[tx->tx_txg & TXG_MASK] = dn->dn_bonuslen;
+ dn->dn_next_blksz[tx->tx_txg & TXG_MASK] = dn->dn_datablksz;
+}
+
+void
+dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ int i, old_nblkptr;
+ dmu_buf_impl_t *db = NULL;
+
+ ASSERT3U(blocksize, >=, SPA_MINBLOCKSIZE);
+ ASSERT3U(blocksize, <=, SPA_MAXBLOCKSIZE);
+ ASSERT3U(blocksize % SPA_MINBLOCKSIZE, ==, 0);
+ ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));
+ ASSERT(tx->tx_txg != 0);
+ ASSERT((bonustype == DMU_OT_NONE && bonuslen == 0) ||
+ (bonustype != DMU_OT_NONE && bonuslen != 0));
+ ASSERT3U(bonustype, <, DMU_OT_NUMTYPES);
+ ASSERT3U(bonuslen, <=, DN_MAX_BONUSLEN);
+
+ for (i = 0; i < TXG_SIZE; i++)
+ ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
+
+ /* clean up any unreferenced dbufs */
+ dnode_evict_dbufs(dn);
+ ASSERT3P(list_head(&dn->dn_dbufs), ==, NULL);
+
+ /*
+ * XXX I should really have a generation number to tell if we
+ * need to do this...
+ */
+ if (blocksize != dn->dn_datablksz ||
+ dn->dn_bonustype != bonustype || dn->dn_bonuslen != bonuslen) {
+ /* free all old data */
+ dnode_free_range(dn, 0, -1ULL, tx);
+ }
+
+ /* change blocksize */
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ if (blocksize != dn->dn_datablksz &&
+ (!BP_IS_HOLE(&dn->dn_phys->dn_blkptr[0]) ||
+ list_head(&dn->dn_dbufs) != NULL)) {
+ db = dbuf_hold(dn, 0, FTAG);
+ dbuf_new_size(db, blocksize, tx);
+ }
+ dnode_setdblksz(dn, blocksize);
+ dnode_setdirty(dn, tx);
+ dn->dn_next_bonuslen[tx->tx_txg&TXG_MASK] = bonuslen;
+ dn->dn_next_blksz[tx->tx_txg&TXG_MASK] = blocksize;
+ rw_exit(&dn->dn_struct_rwlock);
+ if (db)
+ dbuf_rele(db, FTAG);
+
+ /* change type */
+ dn->dn_type = ot;
+
+ /* change bonus size and type */
+ mutex_enter(&dn->dn_mtx);
+ old_nblkptr = dn->dn_nblkptr;
+ dn->dn_bonustype = bonustype;
+ dn->dn_bonuslen = bonuslen;
+ dn->dn_nblkptr = 1 + ((DN_MAX_BONUSLEN - bonuslen) >> SPA_BLKPTRSHIFT);
+ dn->dn_checksum = ZIO_CHECKSUM_INHERIT;
+ dn->dn_compress = ZIO_COMPRESS_INHERIT;
+ ASSERT3U(dn->dn_nblkptr, <=, DN_MAX_NBLKPTR);
+
+ /* XXX - for now, we can't make nblkptr smaller */
+ ASSERT3U(dn->dn_nblkptr, >=, old_nblkptr);
+
+ /* fix up the bonus db_size if dn_nblkptr has changed */
+ if (dn->dn_bonus && dn->dn_bonuslen != old_nblkptr) {
+ dn->dn_bonus->db.db_size =
+ DN_MAX_BONUSLEN - (dn->dn_nblkptr-1) * sizeof (blkptr_t);
+ ASSERT(dn->dn_bonuslen <= dn->dn_bonus->db.db_size);
+ }
+
+ dn->dn_allocated_txg = tx->tx_txg;
+ mutex_exit(&dn->dn_mtx);
+}
+
+void
+dnode_special_close(dnode_t *dn)
+{
+ /*
+ * Wait for final references to the dnode to clear. This can
+ * only happen if the arc is asyncronously evicting state that
+ * has a hold on this dnode while we are trying to evict this
+ * dnode.
+ */
+ while (refcount_count(&dn->dn_holds) > 0)
+ delay(1);
+ dnode_destroy(dn);
+}
+
+dnode_t *
+dnode_special_open(objset_impl_t *os, dnode_phys_t *dnp, uint64_t object)
+{
+ dnode_t *dn = dnode_create(os, dnp, NULL, object);
+ DNODE_VERIFY(dn);
+ return (dn);
+}
+
+static void
+dnode_buf_pageout(dmu_buf_t *db, void *arg)
+{
+ dnode_t **children_dnodes = arg;
+ int i;
+ int epb = db->db_size >> DNODE_SHIFT;
+
+ for (i = 0; i < epb; i++) {
+ dnode_t *dn = children_dnodes[i];
+ int n;
+
+ if (dn == NULL)
+ continue;
+#ifdef ZFS_DEBUG
+ /*
+ * If there are holds on this dnode, then there should
+ * be holds on the dnode's containing dbuf as well; thus
+ * it wouldn't be eligable for eviction and this function
+ * would not have been called.
+ */
+ ASSERT(refcount_is_zero(&dn->dn_holds));
+ ASSERT(list_head(&dn->dn_dbufs) == NULL);
+ ASSERT(refcount_is_zero(&dn->dn_tx_holds));
+
+ for (n = 0; n < TXG_SIZE; n++)
+ ASSERT(!list_link_active(&dn->dn_dirty_link[n]));
+#endif
+ children_dnodes[i] = NULL;
+ dnode_destroy(dn);
+ }
+ kmem_free(children_dnodes, epb * sizeof (dnode_t *));
+}
+
+/*
+ * errors:
+ * EINVAL - invalid object number.
+ * EIO - i/o error.
+ * succeeds even for free dnodes.
+ */
+int
+dnode_hold_impl(objset_impl_t *os, uint64_t object, int flag,
+ void *tag, dnode_t **dnp)
+{
+ int epb, idx, err;
+ int drop_struct_lock = FALSE;
+ int type;
+ uint64_t blk;
+ dnode_t *mdn, *dn;
+ dmu_buf_impl_t *db;
+ dnode_t **children_dnodes;
+
+ /*
+ * If you are holding the spa config lock as writer, you shouldn't
+ * be asking the DMU to do *anything*.
+ */
+ ASSERT(spa_config_held(os->os_spa, SCL_ALL, RW_WRITER) == 0);
+
+ if (object == 0 || object >= DN_MAX_OBJECT)
+ return (EINVAL);
+
+ mdn = os->os_meta_dnode;
+
+ DNODE_VERIFY(mdn);
+
+ if (!RW_WRITE_HELD(&mdn->dn_struct_rwlock)) {
+ rw_enter(&mdn->dn_struct_rwlock, RW_READER);
+ drop_struct_lock = TRUE;
+ }
+
+ blk = dbuf_whichblock(mdn, object * sizeof (dnode_phys_t));
+
+ db = dbuf_hold(mdn, blk, FTAG);
+ if (drop_struct_lock)
+ rw_exit(&mdn->dn_struct_rwlock);
+ if (db == NULL)
+ return (EIO);
+ err = dbuf_read(db, NULL, DB_RF_CANFAIL);
+ if (err) {
+ dbuf_rele(db, FTAG);
+ return (err);
+ }
+
+ ASSERT3U(db->db.db_size, >=, 1<<DNODE_SHIFT);
+ epb = db->db.db_size >> DNODE_SHIFT;
+
+ idx = object & (epb-1);
+
+ children_dnodes = dmu_buf_get_user(&db->db);
+ if (children_dnodes == NULL) {
+ dnode_t **winner;
+ children_dnodes = kmem_zalloc(epb * sizeof (dnode_t *),
+ KM_SLEEP);
+ if (winner = dmu_buf_set_user(&db->db, children_dnodes, NULL,
+ dnode_buf_pageout)) {
+ kmem_free(children_dnodes, epb * sizeof (dnode_t *));
+ children_dnodes = winner;
+ }
+ }
+
+ if ((dn = children_dnodes[idx]) == NULL) {
+ dnode_phys_t *dnp = (dnode_phys_t *)db->db.db_data+idx;
+ dnode_t *winner;
+
+ dn = dnode_create(os, dnp, db, object);
+ winner = atomic_cas_ptr(&children_dnodes[idx], NULL, dn);
+ if (winner != NULL) {
+ dnode_destroy(dn);
+ dn = winner;
+ }
+ }
+
+ mutex_enter(&dn->dn_mtx);
+ type = dn->dn_type;
+ if (dn->dn_free_txg ||
+ ((flag & DNODE_MUST_BE_ALLOCATED) && type == DMU_OT_NONE) ||
+ ((flag & DNODE_MUST_BE_FREE) && type != DMU_OT_NONE)) {
+ mutex_exit(&dn->dn_mtx);
+ dbuf_rele(db, FTAG);
+ return (type == DMU_OT_NONE ? ENOENT : EEXIST);
+ }
+ mutex_exit(&dn->dn_mtx);
+
+ if (refcount_add(&dn->dn_holds, tag) == 1)
+ dbuf_add_ref(db, dn);
+
+ DNODE_VERIFY(dn);
+ ASSERT3P(dn->dn_dbuf, ==, db);
+ ASSERT3U(dn->dn_object, ==, object);
+ dbuf_rele(db, FTAG);
+
+ *dnp = dn;
+ return (0);
+}
+
+/*
+ * Return held dnode if the object is allocated, NULL if not.
+ */
+int
+dnode_hold(objset_impl_t *os, uint64_t object, void *tag, dnode_t **dnp)
+{
+ return (dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, tag, dnp));
+}
+
+/*
+ * Can only add a reference if there is already at least one
+ * reference on the dnode. Returns FALSE if unable to add a
+ * new reference.
+ */
+boolean_t
+dnode_add_ref(dnode_t *dn, void *tag)
+{
+ mutex_enter(&dn->dn_mtx);
+ if (refcount_is_zero(&dn->dn_holds)) {
+ mutex_exit(&dn->dn_mtx);
+ return (FALSE);
+ }
+ VERIFY(1 < refcount_add(&dn->dn_holds, tag));
+ mutex_exit(&dn->dn_mtx);
+ return (TRUE);
+}
+
+void
+dnode_rele(dnode_t *dn, void *tag)
+{
+ uint64_t refs;
+
+ mutex_enter(&dn->dn_mtx);
+ refs = refcount_remove(&dn->dn_holds, tag);
+ mutex_exit(&dn->dn_mtx);
+ /* NOTE: the DNODE_DNODE does not have a dn_dbuf */
+ if (refs == 0 && dn->dn_dbuf)
+ dbuf_rele(dn->dn_dbuf, dn);
+}
+
+void
+dnode_setdirty(dnode_t *dn, dmu_tx_t *tx)
+{
+ objset_impl_t *os = dn->dn_objset;
+ uint64_t txg = tx->tx_txg;
+
+ if (dn->dn_object == DMU_META_DNODE_OBJECT)
+ return;
+
+ DNODE_VERIFY(dn);
+
+#ifdef ZFS_DEBUG
+ mutex_enter(&dn->dn_mtx);
+ ASSERT(dn->dn_phys->dn_type || dn->dn_allocated_txg);
+ /* ASSERT(dn->dn_free_txg == 0 || dn->dn_free_txg >= txg); */
+ mutex_exit(&dn->dn_mtx);
+#endif
+
+ mutex_enter(&os->os_lock);
+
+ /*
+ * If we are already marked dirty, we're done.
+ */
+ if (list_link_active(&dn->dn_dirty_link[txg & TXG_MASK])) {
+ mutex_exit(&os->os_lock);
+ return;
+ }
+
+ ASSERT(!refcount_is_zero(&dn->dn_holds) || list_head(&dn->dn_dbufs));
+ ASSERT(dn->dn_datablksz != 0);
+ ASSERT3U(dn->dn_next_bonuslen[txg&TXG_MASK], ==, 0);
+ ASSERT3U(dn->dn_next_blksz[txg&TXG_MASK], ==, 0);
+
+ dprintf_ds(os->os_dsl_dataset, "obj=%llu txg=%llu\n",
+ dn->dn_object, txg);
+
+ if (dn->dn_free_txg > 0 && dn->dn_free_txg <= txg) {
+ list_insert_tail(&os->os_free_dnodes[txg&TXG_MASK], dn);
+ } else {
+ list_insert_tail(&os->os_dirty_dnodes[txg&TXG_MASK], dn);
+ }
+
+ mutex_exit(&os->os_lock);
+
+ /*
+ * The dnode maintains a hold on its containing dbuf as
+ * long as there are holds on it. Each instantiated child
+ * dbuf maintaines a hold on the dnode. When the last child
+ * drops its hold, the dnode will drop its hold on the
+ * containing dbuf. We add a "dirty hold" here so that the
+ * dnode will hang around after we finish processing its
+ * children.
+ */
+ VERIFY(dnode_add_ref(dn, (void *)(uintptr_t)tx->tx_txg));
+
+ (void) dbuf_dirty(dn->dn_dbuf, tx);
+
+ dsl_dataset_dirty(os->os_dsl_dataset, tx);
+}
+
+void
+dnode_free(dnode_t *dn, dmu_tx_t *tx)
+{
+ int txgoff = tx->tx_txg & TXG_MASK;
+
+ dprintf("dn=%p txg=%llu\n", dn, tx->tx_txg);
+
+ /* we should be the only holder... hopefully */
+ /* ASSERT3U(refcount_count(&dn->dn_holds), ==, 1); */
+
+ mutex_enter(&dn->dn_mtx);
+ if (dn->dn_type == DMU_OT_NONE || dn->dn_free_txg) {
+ mutex_exit(&dn->dn_mtx);
+ return;
+ }
+ dn->dn_free_txg = tx->tx_txg;
+ mutex_exit(&dn->dn_mtx);
+
+ /*
+ * If the dnode is already dirty, it needs to be moved from
+ * the dirty list to the free list.
+ */
+ mutex_enter(&dn->dn_objset->os_lock);
+ if (list_link_active(&dn->dn_dirty_link[txgoff])) {
+ list_remove(&dn->dn_objset->os_dirty_dnodes[txgoff], dn);
+ list_insert_tail(&dn->dn_objset->os_free_dnodes[txgoff], dn);
+ mutex_exit(&dn->dn_objset->os_lock);
+ } else {
+ mutex_exit(&dn->dn_objset->os_lock);
+ dnode_setdirty(dn, tx);
+ }
+}
+
+/*
+ * Try to change the block size for the indicated dnode. This can only
+ * succeed if there are no blocks allocated or dirty beyond first block
+ */
+int
+dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db, *db_next;
+ int err;
+
+ if (size == 0)
+ size = SPA_MINBLOCKSIZE;
+ if (size > SPA_MAXBLOCKSIZE)
+ size = SPA_MAXBLOCKSIZE;
+ else
+ size = P2ROUNDUP(size, SPA_MINBLOCKSIZE);
+
+ if (ibs == dn->dn_indblkshift)
+ ibs = 0;
+
+ if (size >> SPA_MINBLOCKSHIFT == dn->dn_datablkszsec && ibs == 0)
+ return (0);
+
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+
+ /* Check for any allocated blocks beyond the first */
+ if (dn->dn_phys->dn_maxblkid != 0)
+ goto fail;
+
+ mutex_enter(&dn->dn_dbufs_mtx);
+ for (db = list_head(&dn->dn_dbufs); db; db = db_next) {
+ db_next = list_next(&dn->dn_dbufs, db);
+
+ if (db->db_blkid != 0 && db->db_blkid != DB_BONUS_BLKID) {
+ mutex_exit(&dn->dn_dbufs_mtx);
+ goto fail;
+ }
+ }
+ mutex_exit(&dn->dn_dbufs_mtx);
+
+ if (ibs && dn->dn_nlevels != 1)
+ goto fail;
+
+ /* resize the old block */
+ err = dbuf_hold_impl(dn, 0, 0, TRUE, FTAG, &db);
+ if (err == 0)
+ dbuf_new_size(db, size, tx);
+ else if (err != ENOENT)
+ goto fail;
+
+ dnode_setdblksz(dn, size);
+ dnode_setdirty(dn, tx);
+ dn->dn_next_blksz[tx->tx_txg&TXG_MASK] = size;
+ if (ibs) {
+ dn->dn_indblkshift = ibs;
+ dn->dn_next_indblkshift[tx->tx_txg&TXG_MASK] = ibs;
+ }
+ /* rele after we have fixed the blocksize in the dnode */
+ if (db)
+ dbuf_rele(db, FTAG);
+
+ rw_exit(&dn->dn_struct_rwlock);
+ return (0);
+
+fail:
+ rw_exit(&dn->dn_struct_rwlock);
+ return (ENOTSUP);
+}
+
+/* read-holding callers must not rely on the lock being continuously held */
+void
+dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t have_read)
+{
+ uint64_t txgoff = tx->tx_txg & TXG_MASK;
+ int epbs, new_nlevels;
+ uint64_t sz;
+
+ ASSERT(blkid != DB_BONUS_BLKID);
+
+ ASSERT(have_read ?
+ RW_READ_HELD(&dn->dn_struct_rwlock) :
+ RW_WRITE_HELD(&dn->dn_struct_rwlock));
+
+ /*
+ * if we have a read-lock, check to see if we need to do any work
+ * before upgrading to a write-lock.
+ */
+ if (have_read) {
+ if (blkid <= dn->dn_maxblkid)
+ return;
+
+ if (!rw_tryupgrade(&dn->dn_struct_rwlock)) {
+ rw_exit(&dn->dn_struct_rwlock);
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ }
+ }
+
+ if (blkid <= dn->dn_maxblkid)
+ goto out;
+
+ dn->dn_maxblkid = blkid;
+
+ /*
+ * Compute the number of levels necessary to support the new maxblkid.
+ */
+ new_nlevels = 1;
+ epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+ for (sz = dn->dn_nblkptr;
+ sz <= blkid && sz >= dn->dn_nblkptr; sz <<= epbs)
+ new_nlevels++;
+
+ if (new_nlevels > dn->dn_nlevels) {
+ int old_nlevels = dn->dn_nlevels;
+ dmu_buf_impl_t *db;
+ list_t *list;
+ dbuf_dirty_record_t *new, *dr, *dr_next;
+
+ dn->dn_nlevels = new_nlevels;
+
+ ASSERT3U(new_nlevels, >, dn->dn_next_nlevels[txgoff]);
+ dn->dn_next_nlevels[txgoff] = new_nlevels;
+
+ /* dirty the left indirects */
+ db = dbuf_hold_level(dn, old_nlevels, 0, FTAG);
+ new = dbuf_dirty(db, tx);
+ dbuf_rele(db, FTAG);
+
+ /* transfer the dirty records to the new indirect */
+ mutex_enter(&dn->dn_mtx);
+ mutex_enter(&new->dt.di.dr_mtx);
+ list = &dn->dn_dirty_records[txgoff];
+ for (dr = list_head(list); dr; dr = dr_next) {
+ dr_next = list_next(&dn->dn_dirty_records[txgoff], dr);
+ if (dr->dr_dbuf->db_level != new_nlevels-1 &&
+ dr->dr_dbuf->db_blkid != DB_BONUS_BLKID) {
+ ASSERT(dr->dr_dbuf->db_level == old_nlevels-1);
+ list_remove(&dn->dn_dirty_records[txgoff], dr);
+ list_insert_tail(&new->dt.di.dr_children, dr);
+ dr->dr_parent = new;
+ }
+ }
+ mutex_exit(&new->dt.di.dr_mtx);
+ mutex_exit(&dn->dn_mtx);
+ }
+
+out:
+ if (have_read)
+ rw_downgrade(&dn->dn_struct_rwlock);
+}
+
+void
+dnode_clear_range(dnode_t *dn, uint64_t blkid, uint64_t nblks, dmu_tx_t *tx)
+{
+ avl_tree_t *tree = &dn->dn_ranges[tx->tx_txg&TXG_MASK];
+ avl_index_t where;
+ free_range_t *rp;
+ free_range_t rp_tofind;
+ uint64_t endblk = blkid + nblks;
+
+ ASSERT(MUTEX_HELD(&dn->dn_mtx));
+ ASSERT(nblks <= UINT64_MAX - blkid); /* no overflow */
+
+ dprintf_dnode(dn, "blkid=%llu nblks=%llu txg=%llu\n",
+ blkid, nblks, tx->tx_txg);
+ rp_tofind.fr_blkid = blkid;
+ rp = avl_find(tree, &rp_tofind, &where);
+ if (rp == NULL)
+ rp = avl_nearest(tree, where, AVL_BEFORE);
+ if (rp == NULL)
+ rp = avl_nearest(tree, where, AVL_AFTER);
+
+ while (rp && (rp->fr_blkid <= blkid + nblks)) {
+ uint64_t fr_endblk = rp->fr_blkid + rp->fr_nblks;
+ free_range_t *nrp = AVL_NEXT(tree, rp);
+
+ if (blkid <= rp->fr_blkid && endblk >= fr_endblk) {
+ /* clear this entire range */
+ avl_remove(tree, rp);
+ kmem_free(rp, sizeof (free_range_t));
+ } else if (blkid <= rp->fr_blkid &&
+ endblk > rp->fr_blkid && endblk < fr_endblk) {
+ /* clear the beginning of this range */
+ rp->fr_blkid = endblk;
+ rp->fr_nblks = fr_endblk - endblk;
+ } else if (blkid > rp->fr_blkid && blkid < fr_endblk &&
+ endblk >= fr_endblk) {
+ /* clear the end of this range */
+ rp->fr_nblks = blkid - rp->fr_blkid;
+ } else if (blkid > rp->fr_blkid && endblk < fr_endblk) {
+ /* clear a chunk out of this range */
+ free_range_t *new_rp =
+ kmem_alloc(sizeof (free_range_t), KM_SLEEP);
+
+ new_rp->fr_blkid = endblk;
+ new_rp->fr_nblks = fr_endblk - endblk;
+ avl_insert_here(tree, new_rp, rp, AVL_AFTER);
+ rp->fr_nblks = blkid - rp->fr_blkid;
+ }
+ /* there may be no overlap */
+ rp = nrp;
+ }
+}
+
+void
+dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db;
+ uint64_t blkoff, blkid, nblks;
+ int blksz, blkshift, head, tail;
+ int trunc = FALSE;
+ int epbs;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ blksz = dn->dn_datablksz;
+ blkshift = dn->dn_datablkshift;
+ epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+
+ if (len == -1ULL) {
+ len = UINT64_MAX - off;
+ trunc = TRUE;
+ }
+
+ /*
+ * First, block align the region to free:
+ */
+ if (ISP2(blksz)) {
+ head = P2NPHASE(off, blksz);
+ blkoff = P2PHASE(off, blksz);
+ if ((off >> blkshift) > dn->dn_maxblkid)
+ goto out;
+ } else {
+ ASSERT(dn->dn_maxblkid == 0);
+ if (off == 0 && len >= blksz) {
+ /* Freeing the whole block; fast-track this request */
+ blkid = 0;
+ nblks = 1;
+ goto done;
+ } else if (off >= blksz) {
+ /* Freeing past end-of-data */
+ goto out;
+ } else {
+ /* Freeing part of the block. */
+ head = blksz - off;
+ ASSERT3U(head, >, 0);
+ }
+ blkoff = off;
+ }
+ /* zero out any partial block data at the start of the range */
+ if (head) {
+ ASSERT3U(blkoff + head, ==, blksz);
+ if (len < head)
+ head = len;
+ if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, off), TRUE,
+ FTAG, &db) == 0) {
+ caddr_t data;
+
+ /* don't dirty if it isn't on disk and isn't dirty */
+ if (db->db_last_dirty ||
+ (db->db_blkptr && !BP_IS_HOLE(db->db_blkptr))) {
+ rw_exit(&dn->dn_struct_rwlock);
+ dbuf_will_dirty(db, tx);
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ data = db->db.db_data;
+ bzero(data + blkoff, head);
+ }
+ dbuf_rele(db, FTAG);
+ }
+ off += head;
+ len -= head;
+ }
+
+ /* If the range was less than one block, we're done */
+ if (len == 0)
+ goto out;
+
+ /* If the remaining range is past end of file, we're done */
+ if ((off >> blkshift) > dn->dn_maxblkid)
+ goto out;
+
+ ASSERT(ISP2(blksz));
+ if (trunc)
+ tail = 0;
+ else
+ tail = P2PHASE(len, blksz);
+
+ ASSERT3U(P2PHASE(off, blksz), ==, 0);
+ /* zero out any partial block data at the end of the range */
+ if (tail) {
+ if (len < tail)
+ tail = len;
+ if (dbuf_hold_impl(dn, 0, dbuf_whichblock(dn, off+len),
+ TRUE, FTAG, &db) == 0) {
+ /* don't dirty if not on disk and not dirty */
+ if (db->db_last_dirty ||
+ (db->db_blkptr && !BP_IS_HOLE(db->db_blkptr))) {
+ rw_exit(&dn->dn_struct_rwlock);
+ dbuf_will_dirty(db, tx);
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ bzero(db->db.db_data, tail);
+ }
+ dbuf_rele(db, FTAG);
+ }
+ len -= tail;
+ }
+
+ /* If the range did not include a full block, we are done */
+ if (len == 0)
+ goto out;
+
+ ASSERT(IS_P2ALIGNED(off, blksz));
+ ASSERT(trunc || IS_P2ALIGNED(len, blksz));
+ blkid = off >> blkshift;
+ nblks = len >> blkshift;
+ if (trunc)
+ nblks += 1;
+
+ /*
+ * Read in and mark all the level-1 indirects dirty,
+ * so that they will stay in memory until syncing phase.
+ * Always dirty the first and last indirect to make sure
+ * we dirty all the partial indirects.
+ */
+ if (dn->dn_nlevels > 1) {
+ uint64_t i, first, last;
+ int shift = epbs + dn->dn_datablkshift;
+
+ first = blkid >> epbs;
+ if (db = dbuf_hold_level(dn, 1, first, FTAG)) {
+ dbuf_will_dirty(db, tx);
+ dbuf_rele(db, FTAG);
+ }
+ if (trunc)
+ last = dn->dn_maxblkid >> epbs;
+ else
+ last = (blkid + nblks - 1) >> epbs;
+ if (last > first && (db = dbuf_hold_level(dn, 1, last, FTAG))) {
+ dbuf_will_dirty(db, tx);
+ dbuf_rele(db, FTAG);
+ }
+ for (i = first + 1; i < last; i++) {
+ uint64_t ibyte = i << shift;
+ int err;
+
+ err = dnode_next_offset(dn,
+ DNODE_FIND_HAVELOCK, &ibyte, 1, 1, 0);
+ i = ibyte >> shift;
+ if (err == ESRCH || i >= last)
+ break;
+ ASSERT(err == 0);
+ db = dbuf_hold_level(dn, 1, i, FTAG);
+ if (db) {
+ dbuf_will_dirty(db, tx);
+ dbuf_rele(db, FTAG);
+ }
+ }
+ }
+done:
+ /*
+ * Add this range to the dnode range list.
+ * We will finish up this free operation in the syncing phase.
+ */
+ mutex_enter(&dn->dn_mtx);
+ dnode_clear_range(dn, blkid, nblks, tx);
+ {
+ free_range_t *rp, *found;
+ avl_index_t where;
+ avl_tree_t *tree = &dn->dn_ranges[tx->tx_txg&TXG_MASK];
+
+ /* Add new range to dn_ranges */
+ rp = kmem_alloc(sizeof (free_range_t), KM_SLEEP);
+ rp->fr_blkid = blkid;
+ rp->fr_nblks = nblks;
+ found = avl_find(tree, rp, &where);
+ ASSERT(found == NULL);
+ avl_insert(tree, rp, where);
+ dprintf_dnode(dn, "blkid=%llu nblks=%llu txg=%llu\n",
+ blkid, nblks, tx->tx_txg);
+ }
+ mutex_exit(&dn->dn_mtx);
+
+ dbuf_free_range(dn, blkid, blkid + nblks - 1, tx);
+ dnode_setdirty(dn, tx);
+out:
+ if (trunc && dn->dn_maxblkid >= (off >> blkshift))
+ dn->dn_maxblkid = (off >> blkshift ? (off >> blkshift) - 1 : 0);
+
+ rw_exit(&dn->dn_struct_rwlock);
+}
+
+/* return TRUE if this blkid was freed in a recent txg, or FALSE if it wasn't */
+uint64_t
+dnode_block_freed(dnode_t *dn, uint64_t blkid)
+{
+ free_range_t range_tofind;
+ void *dp = spa_get_dsl(dn->dn_objset->os_spa);
+ int i;
+
+ if (blkid == DB_BONUS_BLKID)
+ return (FALSE);
+
+ /*
+ * If we're in the process of opening the pool, dp will not be
+ * set yet, but there shouldn't be anything dirty.
+ */
+ if (dp == NULL)
+ return (FALSE);
+
+ if (dn->dn_free_txg)
+ return (TRUE);
+
+ /*
+ * If dn_datablkshift is not set, then there's only a single
+ * block, in which case there will never be a free range so it
+ * won't matter.
+ */
+ range_tofind.fr_blkid = blkid;
+ mutex_enter(&dn->dn_mtx);
+ for (i = 0; i < TXG_SIZE; i++) {
+ free_range_t *range_found;
+ avl_index_t idx;
+
+ range_found = avl_find(&dn->dn_ranges[i], &range_tofind, &idx);
+ if (range_found) {
+ ASSERT(range_found->fr_nblks > 0);
+ break;
+ }
+ range_found = avl_nearest(&dn->dn_ranges[i], idx, AVL_BEFORE);
+ if (range_found &&
+ range_found->fr_blkid + range_found->fr_nblks > blkid)
+ break;
+ }
+ mutex_exit(&dn->dn_mtx);
+ return (i < TXG_SIZE);
+}
+
+/* call from syncing context when we actually write/free space for this dnode */
+void
+dnode_diduse_space(dnode_t *dn, int64_t delta)
+{
+ uint64_t space;
+ dprintf_dnode(dn, "dn=%p dnp=%p used=%llu delta=%lld\n",
+ dn, dn->dn_phys,
+ (u_longlong_t)dn->dn_phys->dn_used,
+ (longlong_t)delta);
+
+ mutex_enter(&dn->dn_mtx);
+ space = DN_USED_BYTES(dn->dn_phys);
+ if (delta > 0) {
+ ASSERT3U(space + delta, >=, space); /* no overflow */
+ } else {
+ ASSERT3U(space, >=, -delta); /* no underflow */
+ }
+ space += delta;
+ if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_DNODE_BYTES) {
+ ASSERT((dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) == 0);
+ ASSERT3U(P2PHASE(space, 1<<DEV_BSHIFT), ==, 0);
+ dn->dn_phys->dn_used = space >> DEV_BSHIFT;
+ } else {
+ dn->dn_phys->dn_used = space;
+ dn->dn_phys->dn_flags |= DNODE_FLAG_USED_BYTES;
+ }
+ mutex_exit(&dn->dn_mtx);
+}
+
+/*
+ * Call when we think we're going to write/free space in open context.
+ * Be conservative (ie. OK to write less than this or free more than
+ * this, but don't write more or free less).
+ */
+void
+dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx)
+{
+ objset_impl_t *os = dn->dn_objset;
+ dsl_dataset_t *ds = os->os_dsl_dataset;
+
+ if (space > 0)
+ space = spa_get_asize(os->os_spa, space);
+
+ if (ds)
+ dsl_dir_willuse_space(ds->ds_dir, space, tx);
+
+ dmu_tx_willuse_space(tx, space);
+}
+
+static int
+dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
+ int lvl, uint64_t blkfill, uint64_t txg)
+{
+ dmu_buf_impl_t *db = NULL;
+ void *data = NULL;
+ uint64_t epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
+ uint64_t epb = 1ULL << epbs;
+ uint64_t minfill, maxfill;
+ boolean_t hole;
+ int i, inc, error, span;
+
+ dprintf("probing object %llu offset %llx level %d of %u\n",
+ dn->dn_object, *offset, lvl, dn->dn_phys->dn_nlevels);
+
+ hole = flags & DNODE_FIND_HOLE;
+ inc = (flags & DNODE_FIND_BACKWARDS) ? -1 : 1;
+ ASSERT(txg == 0 || !hole);
+
+ if (lvl == dn->dn_phys->dn_nlevels) {
+ error = 0;
+ epb = dn->dn_phys->dn_nblkptr;
+ data = dn->dn_phys->dn_blkptr;
+ } else {
+ uint64_t blkid = dbuf_whichblock(dn, *offset) >> (epbs * lvl);
+ error = dbuf_hold_impl(dn, lvl, blkid, TRUE, FTAG, &db);
+ if (error) {
+ if (error != ENOENT)
+ return (error);
+ if (hole)
+ return (0);
+ /*
+ * This can only happen when we are searching up
+ * the block tree for data. We don't really need to
+ * adjust the offset, as we will just end up looking
+ * at the pointer to this block in its parent, and its
+ * going to be unallocated, so we will skip over it.
+ */
+ return (ESRCH);
+ }
+ error = dbuf_read(db, NULL, DB_RF_CANFAIL | DB_RF_HAVESTRUCT);
+ if (error) {
+ dbuf_rele(db, FTAG);
+ return (error);
+ }
+ data = db->db.db_data;
+ }
+
+ if (db && txg &&
+ (db->db_blkptr == NULL || db->db_blkptr->blk_birth <= txg)) {
+ /*
+ * This can only happen when we are searching up the tree
+ * and these conditions mean that we need to keep climbing.
+ */
+ error = ESRCH;
+ } else if (lvl == 0) {
+ dnode_phys_t *dnp = data;
+ span = DNODE_SHIFT;
+ ASSERT(dn->dn_type == DMU_OT_DNODE);
+
+ for (i = (*offset >> span) & (blkfill - 1);
+ i >= 0 && i < blkfill; i += inc) {
+ boolean_t newcontents = B_TRUE;
+ if (txg) {
+ int j;
+ newcontents = B_FALSE;
+ for (j = 0; j < dnp[i].dn_nblkptr; j++) {
+ if (dnp[i].dn_blkptr[j].blk_birth > txg)
+ newcontents = B_TRUE;
+ }
+ }
+ if (!dnp[i].dn_type == hole && newcontents)
+ break;
+ *offset += (1ULL << span) * inc;
+ }
+ if (i < 0 || i == blkfill)
+ error = ESRCH;
+ } else {
+ blkptr_t *bp = data;
+ span = (lvl - 1) * epbs + dn->dn_datablkshift;
+ minfill = 0;
+ maxfill = blkfill << ((lvl - 1) * epbs);
+
+ if (hole)
+ maxfill--;
+ else
+ minfill++;
+
+ for (i = (*offset >> span) & ((1ULL << epbs) - 1);
+ i >= 0 && i < epb; i += inc) {
+ if (bp[i].blk_fill >= minfill &&
+ bp[i].blk_fill <= maxfill &&
+ (hole || bp[i].blk_birth > txg))
+ break;
+ if (inc < 0 && *offset < (1ULL << span))
+ *offset = 0;
+ else
+ *offset += (1ULL << span) * inc;
+ }
+ if (i < 0 || i == epb)
+ error = ESRCH;
+ }
+
+ if (db)
+ dbuf_rele(db, FTAG);
+
+ return (error);
+}
+
+/*
+ * Find the next hole, data, or sparse region at or after *offset.
+ * The value 'blkfill' tells us how many items we expect to find
+ * in an L0 data block; this value is 1 for normal objects,
+ * DNODES_PER_BLOCK for the meta dnode, and some fraction of
+ * DNODES_PER_BLOCK when searching for sparse regions thereof.
+ *
+ * Examples:
+ *
+ * dnode_next_offset(dn, flags, offset, 1, 1, 0);
+ * Finds the next/previous hole/data in a file.
+ * Used in dmu_offset_next().
+ *
+ * dnode_next_offset(mdn, flags, offset, 0, DNODES_PER_BLOCK, txg);
+ * Finds the next free/allocated dnode an objset's meta-dnode.
+ * Only finds objects that have new contents since txg (ie.
+ * bonus buffer changes and content removal are ignored).
+ * Used in dmu_object_next().
+ *
+ * dnode_next_offset(mdn, DNODE_FIND_HOLE, offset, 2, DNODES_PER_BLOCK >> 2, 0);
+ * Finds the next L2 meta-dnode bp that's at most 1/4 full.
+ * Used in dmu_object_alloc().
+ */
+int
+dnode_next_offset(dnode_t *dn, int flags, uint64_t *offset,
+ int minlvl, uint64_t blkfill, uint64_t txg)
+{
+ uint64_t initial_offset = *offset;
+ int lvl, maxlvl;
+ int error = 0;
+
+ if (!(flags & DNODE_FIND_HAVELOCK))
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+
+ if (dn->dn_phys->dn_nlevels == 0) {
+ error = ESRCH;
+ goto out;
+ }
+
+ if (dn->dn_datablkshift == 0) {
+ if (*offset < dn->dn_datablksz) {
+ if (flags & DNODE_FIND_HOLE)
+ *offset = dn->dn_datablksz;
+ } else {
+ error = ESRCH;
+ }
+ goto out;
+ }
+
+ maxlvl = dn->dn_phys->dn_nlevels;
+
+ for (lvl = minlvl; lvl <= maxlvl; lvl++) {
+ error = dnode_next_offset_level(dn,
+ flags, offset, lvl, blkfill, txg);
+ if (error != ESRCH)
+ break;
+ }
+
+ while (error == 0 && --lvl >= minlvl) {
+ error = dnode_next_offset_level(dn,
+ flags, offset, lvl, blkfill, txg);
+ }
+
+ if (error == 0 && (flags & DNODE_FIND_BACKWARDS ?
+ initial_offset < *offset : initial_offset > *offset))
+ error = ESRCH;
+out:
+ if (!(flags & DNODE_FIND_HAVELOCK))
+ rw_exit(&dn->dn_struct_rwlock);
+
+ return (error);
+}
diff --git a/module/zfs/dnode_sync.c b/module/zfs/dnode_sync.c
new file mode 100644
index 000000000..779cfc96f
--- /dev/null
+++ b/module/zfs/dnode_sync.c
@@ -0,0 +1,623 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/dbuf.h>
+#include <sys/dnode.h>
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dataset.h>
+#include <sys/spa.h>
+
+static void
+dnode_increase_indirection(dnode_t *dn, dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *db;
+ int txgoff = tx->tx_txg & TXG_MASK;
+ int nblkptr = dn->dn_phys->dn_nblkptr;
+ int old_toplvl = dn->dn_phys->dn_nlevels - 1;
+ int new_level = dn->dn_next_nlevels[txgoff];
+ int i;
+
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+
+ /* this dnode can't be paged out because it's dirty */
+ ASSERT(dn->dn_phys->dn_type != DMU_OT_NONE);
+ ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock));
+ ASSERT(new_level > 1 && dn->dn_phys->dn_nlevels > 0);
+
+ db = dbuf_hold_level(dn, dn->dn_phys->dn_nlevels, 0, FTAG);
+ ASSERT(db != NULL);
+
+ dn->dn_phys->dn_nlevels = new_level;
+ dprintf("os=%p obj=%llu, increase to %d\n", dn->dn_objset,
+ dn->dn_object, dn->dn_phys->dn_nlevels);
+
+ /* check for existing blkptrs in the dnode */
+ for (i = 0; i < nblkptr; i++)
+ if (!BP_IS_HOLE(&dn->dn_phys->dn_blkptr[i]))
+ break;
+ if (i != nblkptr) {
+ /* transfer dnode's block pointers to new indirect block */
+ (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED|DB_RF_HAVESTRUCT);
+ ASSERT(db->db.db_data);
+ ASSERT(arc_released(db->db_buf));
+ ASSERT3U(sizeof (blkptr_t) * nblkptr, <=, db->db.db_size);
+ bcopy(dn->dn_phys->dn_blkptr, db->db.db_data,
+ sizeof (blkptr_t) * nblkptr);
+ arc_buf_freeze(db->db_buf);
+ }
+
+ /* set dbuf's parent pointers to new indirect buf */
+ for (i = 0; i < nblkptr; i++) {
+ dmu_buf_impl_t *child = dbuf_find(dn, old_toplvl, i);
+
+ if (child == NULL)
+ continue;
+ ASSERT3P(child->db_dnode, ==, dn);
+ if (child->db_parent && child->db_parent != dn->dn_dbuf) {
+ ASSERT(child->db_parent->db_level == db->db_level);
+ ASSERT(child->db_blkptr !=
+ &dn->dn_phys->dn_blkptr[child->db_blkid]);
+ mutex_exit(&child->db_mtx);
+ continue;
+ }
+ ASSERT(child->db_parent == NULL ||
+ child->db_parent == dn->dn_dbuf);
+
+ child->db_parent = db;
+ dbuf_add_ref(db, child);
+ if (db->db.db_data)
+ child->db_blkptr = (blkptr_t *)db->db.db_data + i;
+ else
+ child->db_blkptr = NULL;
+ dprintf_dbuf_bp(child, child->db_blkptr,
+ "changed db_blkptr to new indirect %s", "");
+
+ mutex_exit(&child->db_mtx);
+ }
+
+ bzero(dn->dn_phys->dn_blkptr, sizeof (blkptr_t) * nblkptr);
+
+ dbuf_rele(db, FTAG);
+
+ rw_exit(&dn->dn_struct_rwlock);
+}
+
+static int
+free_blocks(dnode_t *dn, blkptr_t *bp, int num, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
+ uint64_t bytesfreed = 0;
+ int i, blocks_freed = 0;
+
+ dprintf("ds=%p obj=%llx num=%d\n", ds, dn->dn_object, num);
+
+ for (i = 0; i < num; i++, bp++) {
+ if (BP_IS_HOLE(bp))
+ continue;
+
+ bytesfreed += dsl_dataset_block_kill(ds, bp, dn->dn_zio, tx);
+ ASSERT3U(bytesfreed, <=, DN_USED_BYTES(dn->dn_phys));
+ bzero(bp, sizeof (blkptr_t));
+ blocks_freed += 1;
+ }
+ dnode_diduse_space(dn, -bytesfreed);
+ return (blocks_freed);
+}
+
+#ifdef ZFS_DEBUG
+static void
+free_verify(dmu_buf_impl_t *db, uint64_t start, uint64_t end, dmu_tx_t *tx)
+{
+ int off, num;
+ int i, err, epbs;
+ uint64_t txg = tx->tx_txg;
+
+ epbs = db->db_dnode->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
+ off = start - (db->db_blkid * 1<<epbs);
+ num = end - start + 1;
+
+ ASSERT3U(off, >=, 0);
+ ASSERT3U(num, >=, 0);
+ ASSERT3U(db->db_level, >, 0);
+ ASSERT3U(db->db.db_size, ==, 1<<db->db_dnode->dn_phys->dn_indblkshift);
+ ASSERT3U(off+num, <=, db->db.db_size >> SPA_BLKPTRSHIFT);
+ ASSERT(db->db_blkptr != NULL);
+
+ for (i = off; i < off+num; i++) {
+ uint64_t *buf;
+ dmu_buf_impl_t *child;
+ dbuf_dirty_record_t *dr;
+ int j;
+
+ ASSERT(db->db_level == 1);
+
+ rw_enter(&db->db_dnode->dn_struct_rwlock, RW_READER);
+ err = dbuf_hold_impl(db->db_dnode, db->db_level-1,
+ (db->db_blkid << epbs) + i, TRUE, FTAG, &child);
+ rw_exit(&db->db_dnode->dn_struct_rwlock);
+ if (err == ENOENT)
+ continue;
+ ASSERT(err == 0);
+ ASSERT(child->db_level == 0);
+ dr = child->db_last_dirty;
+ while (dr && dr->dr_txg > txg)
+ dr = dr->dr_next;
+ ASSERT(dr == NULL || dr->dr_txg == txg);
+
+ /* data_old better be zeroed */
+ if (dr) {
+ buf = dr->dt.dl.dr_data->b_data;
+ for (j = 0; j < child->db.db_size >> 3; j++) {
+ if (buf[j] != 0) {
+ panic("freed data not zero: "
+ "child=%p i=%d off=%d num=%d\n",
+ (void *)child, i, off, num);
+ }
+ }
+ }
+
+ /*
+ * db_data better be zeroed unless it's dirty in a
+ * future txg.
+ */
+ mutex_enter(&child->db_mtx);
+ buf = child->db.db_data;
+ if (buf != NULL && child->db_state != DB_FILL &&
+ child->db_last_dirty == NULL) {
+ for (j = 0; j < child->db.db_size >> 3; j++) {
+ if (buf[j] != 0) {
+ panic("freed data not zero: "
+ "child=%p i=%d off=%d num=%d\n",
+ (void *)child, i, off, num);
+ }
+ }
+ }
+ mutex_exit(&child->db_mtx);
+
+ dbuf_rele(child, FTAG);
+ }
+}
+#endif
+
+#define ALL -1
+
+static int
+free_children(dmu_buf_impl_t *db, uint64_t blkid, uint64_t nblks, int trunc,
+ dmu_tx_t *tx)
+{
+ dnode_t *dn = db->db_dnode;
+ blkptr_t *bp;
+ dmu_buf_impl_t *subdb;
+ uint64_t start, end, dbstart, dbend, i;
+ int epbs, shift, err;
+ int all = TRUE;
+ int blocks_freed = 0;
+
+ /*
+ * There is a small possibility that this block will not be cached:
+ * 1 - if level > 1 and there are no children with level <= 1
+ * 2 - if we didn't get a dirty hold (because this block had just
+ * finished being written -- and so had no holds), and then this
+ * block got evicted before we got here.
+ */
+ if (db->db_state != DB_CACHED)
+ (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED);
+
+ arc_release(db->db_buf, db);
+ bp = (blkptr_t *)db->db.db_data;
+
+ epbs = db->db_dnode->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
+ shift = (db->db_level - 1) * epbs;
+ dbstart = db->db_blkid << epbs;
+ start = blkid >> shift;
+ if (dbstart < start) {
+ bp += start - dbstart;
+ all = FALSE;
+ } else {
+ start = dbstart;
+ }
+ dbend = ((db->db_blkid + 1) << epbs) - 1;
+ end = (blkid + nblks - 1) >> shift;
+ if (dbend <= end)
+ end = dbend;
+ else if (all)
+ all = trunc;
+ ASSERT3U(start, <=, end);
+
+ if (db->db_level == 1) {
+ FREE_VERIFY(db, start, end, tx);
+ blocks_freed = free_blocks(dn, bp, end-start+1, tx);
+ arc_buf_freeze(db->db_buf);
+ ASSERT(all || blocks_freed == 0 || db->db_last_dirty);
+ return (all ? ALL : blocks_freed);
+ }
+
+ for (i = start; i <= end; i++, bp++) {
+ if (BP_IS_HOLE(bp))
+ continue;
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ err = dbuf_hold_impl(dn, db->db_level-1, i, TRUE, FTAG, &subdb);
+ ASSERT3U(err, ==, 0);
+ rw_exit(&dn->dn_struct_rwlock);
+
+ if (free_children(subdb, blkid, nblks, trunc, tx) == ALL) {
+ ASSERT3P(subdb->db_blkptr, ==, bp);
+ blocks_freed += free_blocks(dn, bp, 1, tx);
+ } else {
+ all = FALSE;
+ }
+ dbuf_rele(subdb, FTAG);
+ }
+ arc_buf_freeze(db->db_buf);
+#ifdef ZFS_DEBUG
+ bp -= (end-start)+1;
+ for (i = start; i <= end; i++, bp++) {
+ if (i == start && blkid != 0)
+ continue;
+ else if (i == end && !trunc)
+ continue;
+ ASSERT3U(bp->blk_birth, ==, 0);
+ }
+#endif
+ ASSERT(all || blocks_freed == 0 || db->db_last_dirty);
+ return (all ? ALL : blocks_freed);
+}
+
+/*
+ * free_range: Traverse the indicated range of the provided file
+ * and "free" all the blocks contained there.
+ */
+static void
+dnode_sync_free_range(dnode_t *dn, uint64_t blkid, uint64_t nblks, dmu_tx_t *tx)
+{
+ blkptr_t *bp = dn->dn_phys->dn_blkptr;
+ dmu_buf_impl_t *db;
+ int trunc, start, end, shift, i, err;
+ int dnlevel = dn->dn_phys->dn_nlevels;
+
+ if (blkid > dn->dn_phys->dn_maxblkid)
+ return;
+
+ ASSERT(dn->dn_phys->dn_maxblkid < UINT64_MAX);
+ trunc = blkid + nblks > dn->dn_phys->dn_maxblkid;
+ if (trunc)
+ nblks = dn->dn_phys->dn_maxblkid - blkid + 1;
+
+ /* There are no indirect blocks in the object */
+ if (dnlevel == 1) {
+ if (blkid >= dn->dn_phys->dn_nblkptr) {
+ /* this range was never made persistent */
+ return;
+ }
+ ASSERT3U(blkid + nblks, <=, dn->dn_phys->dn_nblkptr);
+ (void) free_blocks(dn, bp + blkid, nblks, tx);
+ if (trunc) {
+ uint64_t off = (dn->dn_phys->dn_maxblkid + 1) *
+ (dn->dn_phys->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ dn->dn_phys->dn_maxblkid = (blkid ? blkid - 1 : 0);
+ ASSERT(off < dn->dn_phys->dn_maxblkid ||
+ dn->dn_phys->dn_maxblkid == 0 ||
+ dnode_next_offset(dn, 0, &off, 1, 1, 0) != 0);
+ }
+ return;
+ }
+
+ shift = (dnlevel - 1) * (dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT);
+ start = blkid >> shift;
+ ASSERT(start < dn->dn_phys->dn_nblkptr);
+ end = (blkid + nblks - 1) >> shift;
+ bp += start;
+ for (i = start; i <= end; i++, bp++) {
+ if (BP_IS_HOLE(bp))
+ continue;
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ err = dbuf_hold_impl(dn, dnlevel-1, i, TRUE, FTAG, &db);
+ ASSERT3U(err, ==, 0);
+ rw_exit(&dn->dn_struct_rwlock);
+
+ if (free_children(db, blkid, nblks, trunc, tx) == ALL) {
+ ASSERT3P(db->db_blkptr, ==, bp);
+ (void) free_blocks(dn, bp, 1, tx);
+ }
+ dbuf_rele(db, FTAG);
+ }
+ if (trunc) {
+ uint64_t off = (dn->dn_phys->dn_maxblkid + 1) *
+ (dn->dn_phys->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ dn->dn_phys->dn_maxblkid = (blkid ? blkid - 1 : 0);
+ ASSERT(off < dn->dn_phys->dn_maxblkid ||
+ dn->dn_phys->dn_maxblkid == 0 ||
+ dnode_next_offset(dn, 0, &off, 1, 1, 0) != 0);
+ }
+}
+
+/*
+ * Try to kick all the dnodes dbufs out of the cache...
+ */
+void
+dnode_evict_dbufs(dnode_t *dn)
+{
+ int progress;
+ int pass = 0;
+
+ do {
+ dmu_buf_impl_t *db, marker;
+ int evicting = FALSE;
+
+ progress = FALSE;
+ mutex_enter(&dn->dn_dbufs_mtx);
+ list_insert_tail(&dn->dn_dbufs, &marker);
+ db = list_head(&dn->dn_dbufs);
+ for (; db != &marker; db = list_head(&dn->dn_dbufs)) {
+ list_remove(&dn->dn_dbufs, db);
+ list_insert_tail(&dn->dn_dbufs, db);
+ ASSERT3P(db->db_dnode, ==, dn);
+
+ mutex_enter(&db->db_mtx);
+ if (db->db_state == DB_EVICTING) {
+ progress = TRUE;
+ evicting = TRUE;
+ mutex_exit(&db->db_mtx);
+ } else if (refcount_is_zero(&db->db_holds)) {
+ progress = TRUE;
+ dbuf_clear(db); /* exits db_mtx for us */
+ } else {
+ mutex_exit(&db->db_mtx);
+ }
+
+ }
+ list_remove(&dn->dn_dbufs, &marker);
+ /*
+ * NB: we need to drop dn_dbufs_mtx between passes so
+ * that any DB_EVICTING dbufs can make progress.
+ * Ideally, we would have some cv we could wait on, but
+ * since we don't, just wait a bit to give the other
+ * thread a chance to run.
+ */
+ mutex_exit(&dn->dn_dbufs_mtx);
+ if (evicting)
+ delay(1);
+ pass++;
+ ASSERT(pass < 100); /* sanity check */
+ } while (progress);
+
+ rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
+ if (dn->dn_bonus && refcount_is_zero(&dn->dn_bonus->db_holds)) {
+ mutex_enter(&dn->dn_bonus->db_mtx);
+ dbuf_evict(dn->dn_bonus);
+ dn->dn_bonus = NULL;
+ }
+ rw_exit(&dn->dn_struct_rwlock);
+}
+
+static void
+dnode_undirty_dbufs(list_t *list)
+{
+ dbuf_dirty_record_t *dr;
+
+ while (dr = list_head(list)) {
+ dmu_buf_impl_t *db = dr->dr_dbuf;
+ uint64_t txg = dr->dr_txg;
+
+ mutex_enter(&db->db_mtx);
+ /* XXX - use dbuf_undirty()? */
+ list_remove(list, dr);
+ ASSERT(db->db_last_dirty == dr);
+ db->db_last_dirty = NULL;
+ db->db_dirtycnt -= 1;
+ if (db->db_level == 0) {
+ ASSERT(db->db_blkid == DB_BONUS_BLKID ||
+ dr->dt.dl.dr_data == db->db_buf);
+ dbuf_unoverride(dr);
+ mutex_exit(&db->db_mtx);
+ } else {
+ mutex_exit(&db->db_mtx);
+ dnode_undirty_dbufs(&dr->dt.di.dr_children);
+ }
+ kmem_free(dr, sizeof (dbuf_dirty_record_t));
+ dbuf_rele(db, (void *)(uintptr_t)txg);
+ }
+}
+
+static void
+dnode_sync_free(dnode_t *dn, dmu_tx_t *tx)
+{
+ int txgoff = tx->tx_txg & TXG_MASK;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ /*
+ * Our contents should have been freed in dnode_sync() by the
+ * free range record inserted by the caller of dnode_free().
+ */
+ ASSERT3U(DN_USED_BYTES(dn->dn_phys), ==, 0);
+ ASSERT(BP_IS_HOLE(dn->dn_phys->dn_blkptr));
+
+ dnode_undirty_dbufs(&dn->dn_dirty_records[txgoff]);
+ dnode_evict_dbufs(dn);
+ ASSERT3P(list_head(&dn->dn_dbufs), ==, NULL);
+
+ /*
+ * XXX - It would be nice to assert this, but we may still
+ * have residual holds from async evictions from the arc...
+ *
+ * zfs_obj_to_path() also depends on this being
+ * commented out.
+ *
+ * ASSERT3U(refcount_count(&dn->dn_holds), ==, 1);
+ */
+
+ /* Undirty next bits */
+ dn->dn_next_nlevels[txgoff] = 0;
+ dn->dn_next_indblkshift[txgoff] = 0;
+ dn->dn_next_blksz[txgoff] = 0;
+
+ /* ASSERT(blkptrs are zero); */
+ ASSERT(dn->dn_phys->dn_type != DMU_OT_NONE);
+ ASSERT(dn->dn_type != DMU_OT_NONE);
+
+ ASSERT(dn->dn_free_txg > 0);
+ if (dn->dn_allocated_txg != dn->dn_free_txg)
+ dbuf_will_dirty(dn->dn_dbuf, tx);
+ bzero(dn->dn_phys, sizeof (dnode_phys_t));
+
+ mutex_enter(&dn->dn_mtx);
+ dn->dn_type = DMU_OT_NONE;
+ dn->dn_maxblkid = 0;
+ dn->dn_allocated_txg = 0;
+ dn->dn_free_txg = 0;
+ mutex_exit(&dn->dn_mtx);
+
+ ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
+
+ dnode_rele(dn, (void *)(uintptr_t)tx->tx_txg);
+ /*
+ * Now that we've released our hold, the dnode may
+ * be evicted, so we musn't access it.
+ */
+}
+
+/*
+ * Write out the dnode's dirty buffers.
+ *
+ * NOTE: The dnode is kept in memory by being dirty. Once the
+ * dirty bit is cleared, it may be evicted. Beware of this!
+ */
+void
+dnode_sync(dnode_t *dn, dmu_tx_t *tx)
+{
+ free_range_t *rp;
+ dnode_phys_t *dnp = dn->dn_phys;
+ int txgoff = tx->tx_txg & TXG_MASK;
+ list_t *list = &dn->dn_dirty_records[txgoff];
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(dnp->dn_type != DMU_OT_NONE || dn->dn_allocated_txg);
+ DNODE_VERIFY(dn);
+
+ ASSERT(dn->dn_dbuf == NULL || arc_released(dn->dn_dbuf->db_buf));
+
+ mutex_enter(&dn->dn_mtx);
+ if (dn->dn_allocated_txg == tx->tx_txg) {
+ /* The dnode is newly allocated or reallocated */
+ if (dnp->dn_type == DMU_OT_NONE) {
+ /* this is a first alloc, not a realloc */
+ /* XXX shouldn't the phys already be zeroed? */
+ bzero(dnp, DNODE_CORE_SIZE);
+ dnp->dn_nlevels = 1;
+ }
+
+ if (dn->dn_nblkptr > dnp->dn_nblkptr) {
+ /* zero the new blkptrs we are gaining */
+ bzero(dnp->dn_blkptr + dnp->dn_nblkptr,
+ sizeof (blkptr_t) *
+ (dn->dn_nblkptr - dnp->dn_nblkptr));
+ }
+ dnp->dn_type = dn->dn_type;
+ dnp->dn_bonustype = dn->dn_bonustype;
+ dnp->dn_bonuslen = dn->dn_bonuslen;
+ dnp->dn_nblkptr = dn->dn_nblkptr;
+ }
+
+ ASSERT(dnp->dn_nlevels > 1 ||
+ BP_IS_HOLE(&dnp->dn_blkptr[0]) ||
+ BP_GET_LSIZE(&dnp->dn_blkptr[0]) ==
+ dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+
+ if (dn->dn_next_blksz[txgoff]) {
+ ASSERT(P2PHASE(dn->dn_next_blksz[txgoff],
+ SPA_MINBLOCKSIZE) == 0);
+ ASSERT(BP_IS_HOLE(&dnp->dn_blkptr[0]) ||
+ dn->dn_maxblkid == 0 || list_head(list) != NULL ||
+ dn->dn_next_blksz[txgoff] >> SPA_MINBLOCKSHIFT ==
+ dnp->dn_datablkszsec);
+ dnp->dn_datablkszsec =
+ dn->dn_next_blksz[txgoff] >> SPA_MINBLOCKSHIFT;
+ dn->dn_next_blksz[txgoff] = 0;
+ }
+
+ if (dn->dn_next_bonuslen[txgoff]) {
+ if (dn->dn_next_bonuslen[txgoff] == DN_ZERO_BONUSLEN)
+ dnp->dn_bonuslen = 0;
+ else
+ dnp->dn_bonuslen = dn->dn_next_bonuslen[txgoff];
+ ASSERT(dnp->dn_bonuslen <= DN_MAX_BONUSLEN);
+ dn->dn_next_bonuslen[txgoff] = 0;
+ }
+
+ if (dn->dn_next_indblkshift[txgoff]) {
+ ASSERT(dnp->dn_nlevels == 1);
+ dnp->dn_indblkshift = dn->dn_next_indblkshift[txgoff];
+ dn->dn_next_indblkshift[txgoff] = 0;
+ }
+
+ /*
+ * Just take the live (open-context) values for checksum and compress.
+ * Strictly speaking it's a future leak, but nothing bad happens if we
+ * start using the new checksum or compress algorithm a little early.
+ */
+ dnp->dn_checksum = dn->dn_checksum;
+ dnp->dn_compress = dn->dn_compress;
+
+ mutex_exit(&dn->dn_mtx);
+
+ /* process all the "freed" ranges in the file */
+ while (rp = avl_last(&dn->dn_ranges[txgoff])) {
+ dnode_sync_free_range(dn, rp->fr_blkid, rp->fr_nblks, tx);
+ /* grab the mutex so we don't race with dnode_block_freed() */
+ mutex_enter(&dn->dn_mtx);
+ avl_remove(&dn->dn_ranges[txgoff], rp);
+ mutex_exit(&dn->dn_mtx);
+ kmem_free(rp, sizeof (free_range_t));
+ }
+
+ if (dn->dn_free_txg > 0 && dn->dn_free_txg <= tx->tx_txg) {
+ dnode_sync_free(dn, tx);
+ return;
+ }
+
+ if (dn->dn_next_nlevels[txgoff]) {
+ dnode_increase_indirection(dn, tx);
+ dn->dn_next_nlevels[txgoff] = 0;
+ }
+
+ dbuf_sync_list(list, tx);
+
+ if (dn->dn_object != DMU_META_DNODE_OBJECT) {
+ ASSERT3P(list_head(list), ==, NULL);
+ dnode_rele(dn, (void *)(uintptr_t)tx->tx_txg);
+ }
+
+ /*
+ * Although we have dropped our reference to the dnode, it
+ * can't be evicted until its written, and we haven't yet
+ * initiated the IO for the dnode's dbuf.
+ */
+}
diff --git a/module/zfs/dsl_dataset.c b/module/zfs/dsl_dataset.c
new file mode 100644
index 000000000..93ea8aa11
--- /dev/null
+++ b/module/zfs/dsl_dataset.c
@@ -0,0 +1,3103 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dmu_objset.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dmu_tx.h>
+#include <sys/arc.h>
+#include <sys/zio.h>
+#include <sys/zap.h>
+#include <sys/unique.h>
+#include <sys/zfs_context.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/spa.h>
+#include <sys/zfs_znode.h>
+#include <sys/sunddi.h>
+
+static char *dsl_reaper = "the grim reaper";
+
+static dsl_checkfunc_t dsl_dataset_destroy_begin_check;
+static dsl_syncfunc_t dsl_dataset_destroy_begin_sync;
+static dsl_checkfunc_t dsl_dataset_rollback_check;
+static dsl_syncfunc_t dsl_dataset_rollback_sync;
+static dsl_syncfunc_t dsl_dataset_set_reservation_sync;
+
+#define DS_REF_MAX (1ULL << 62)
+
+#define DSL_DEADLIST_BLOCKSIZE SPA_MAXBLOCKSIZE
+
+#define DSL_DATASET_IS_DESTROYED(ds) ((ds)->ds_owner == dsl_reaper)
+
+
+/*
+ * Figure out how much of this delta should be propogated to the dsl_dir
+ * layer. If there's a refreservation, that space has already been
+ * partially accounted for in our ancestors.
+ */
+static int64_t
+parent_delta(dsl_dataset_t *ds, int64_t delta)
+{
+ uint64_t old_bytes, new_bytes;
+
+ if (ds->ds_reserved == 0)
+ return (delta);
+
+ old_bytes = MAX(ds->ds_phys->ds_unique_bytes, ds->ds_reserved);
+ new_bytes = MAX(ds->ds_phys->ds_unique_bytes + delta, ds->ds_reserved);
+
+ ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta));
+ return (new_bytes - old_bytes);
+}
+
+void
+dsl_dataset_block_born(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx)
+{
+ int used = bp_get_dasize(tx->tx_pool->dp_spa, bp);
+ int compressed = BP_GET_PSIZE(bp);
+ int uncompressed = BP_GET_UCSIZE(bp);
+ int64_t delta;
+
+ dprintf_bp(bp, "born, ds=%p\n", ds);
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ /* It could have been compressed away to nothing */
+ if (BP_IS_HOLE(bp))
+ return;
+ ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE);
+ ASSERT3U(BP_GET_TYPE(bp), <, DMU_OT_NUMTYPES);
+ if (ds == NULL) {
+ /*
+ * Account for the meta-objset space in its placeholder
+ * dsl_dir.
+ */
+ ASSERT3U(compressed, ==, uncompressed); /* it's all metadata */
+ dsl_dir_diduse_space(tx->tx_pool->dp_mos_dir, DD_USED_HEAD,
+ used, compressed, uncompressed, tx);
+ dsl_dir_dirty(tx->tx_pool->dp_mos_dir, tx);
+ return;
+ }
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ mutex_enter(&ds->ds_dir->dd_lock);
+ mutex_enter(&ds->ds_lock);
+ delta = parent_delta(ds, used);
+ ds->ds_phys->ds_used_bytes += used;
+ ds->ds_phys->ds_compressed_bytes += compressed;
+ ds->ds_phys->ds_uncompressed_bytes += uncompressed;
+ ds->ds_phys->ds_unique_bytes += used;
+ mutex_exit(&ds->ds_lock);
+ dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta,
+ compressed, uncompressed, tx);
+ dsl_dir_transfer_space(ds->ds_dir, used - delta,
+ DD_USED_REFRSRV, DD_USED_HEAD, tx);
+ mutex_exit(&ds->ds_dir->dd_lock);
+}
+
+int
+dsl_dataset_block_kill(dsl_dataset_t *ds, blkptr_t *bp, zio_t *pio,
+ dmu_tx_t *tx)
+{
+ int used = bp_get_dasize(tx->tx_pool->dp_spa, bp);
+ int compressed = BP_GET_PSIZE(bp);
+ int uncompressed = BP_GET_UCSIZE(bp);
+
+ ASSERT(pio != NULL);
+ ASSERT(dmu_tx_is_syncing(tx));
+ /* No block pointer => nothing to free */
+ if (BP_IS_HOLE(bp))
+ return (0);
+
+ ASSERT(used > 0);
+ if (ds == NULL) {
+ int err;
+ /*
+ * Account for the meta-objset space in its placeholder
+ * dataset.
+ */
+ err = dsl_free(pio, tx->tx_pool,
+ tx->tx_txg, bp, NULL, NULL, ARC_NOWAIT);
+ ASSERT(err == 0);
+
+ dsl_dir_diduse_space(tx->tx_pool->dp_mos_dir, DD_USED_HEAD,
+ -used, -compressed, -uncompressed, tx);
+ dsl_dir_dirty(tx->tx_pool->dp_mos_dir, tx);
+ return (used);
+ }
+ ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool);
+
+ ASSERT(!dsl_dataset_is_snapshot(ds));
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+
+ if (bp->blk_birth > ds->ds_phys->ds_prev_snap_txg) {
+ int err;
+ int64_t delta;
+
+ dprintf_bp(bp, "freeing: %s", "");
+ err = dsl_free(pio, tx->tx_pool,
+ tx->tx_txg, bp, NULL, NULL, ARC_NOWAIT);
+ ASSERT(err == 0);
+
+ mutex_enter(&ds->ds_dir->dd_lock);
+ mutex_enter(&ds->ds_lock);
+ ASSERT(ds->ds_phys->ds_unique_bytes >= used ||
+ !DS_UNIQUE_IS_ACCURATE(ds));
+ delta = parent_delta(ds, -used);
+ ds->ds_phys->ds_unique_bytes -= used;
+ mutex_exit(&ds->ds_lock);
+ dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD,
+ delta, -compressed, -uncompressed, tx);
+ dsl_dir_transfer_space(ds->ds_dir, -used - delta,
+ DD_USED_REFRSRV, DD_USED_HEAD, tx);
+ mutex_exit(&ds->ds_dir->dd_lock);
+ } else {
+ dprintf_bp(bp, "putting on dead list: %s", "");
+ VERIFY(0 == bplist_enqueue(&ds->ds_deadlist, bp, tx));
+ ASSERT3U(ds->ds_prev->ds_object, ==,
+ ds->ds_phys->ds_prev_snap_obj);
+ ASSERT(ds->ds_prev->ds_phys->ds_num_children > 0);
+ /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
+ if (ds->ds_prev->ds_phys->ds_next_snap_obj ==
+ ds->ds_object && bp->blk_birth >
+ ds->ds_prev->ds_phys->ds_prev_snap_txg) {
+ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
+ mutex_enter(&ds->ds_prev->ds_lock);
+ ds->ds_prev->ds_phys->ds_unique_bytes += used;
+ mutex_exit(&ds->ds_prev->ds_lock);
+ }
+ if (bp->blk_birth > ds->ds_origin_txg) {
+ dsl_dir_transfer_space(ds->ds_dir, used,
+ DD_USED_HEAD, DD_USED_SNAP, tx);
+ }
+ }
+ mutex_enter(&ds->ds_lock);
+ ASSERT3U(ds->ds_phys->ds_used_bytes, >=, used);
+ ds->ds_phys->ds_used_bytes -= used;
+ ASSERT3U(ds->ds_phys->ds_compressed_bytes, >=, compressed);
+ ds->ds_phys->ds_compressed_bytes -= compressed;
+ ASSERT3U(ds->ds_phys->ds_uncompressed_bytes, >=, uncompressed);
+ ds->ds_phys->ds_uncompressed_bytes -= uncompressed;
+ mutex_exit(&ds->ds_lock);
+
+ return (used);
+}
+
+uint64_t
+dsl_dataset_prev_snap_txg(dsl_dataset_t *ds)
+{
+ uint64_t trysnap = 0;
+
+ if (ds == NULL)
+ return (0);
+ /*
+ * The snapshot creation could fail, but that would cause an
+ * incorrect FALSE return, which would only result in an
+ * overestimation of the amount of space that an operation would
+ * consume, which is OK.
+ *
+ * There's also a small window where we could miss a pending
+ * snapshot, because we could set the sync task in the quiescing
+ * phase. So this should only be used as a guess.
+ */
+ if (ds->ds_trysnap_txg >
+ spa_last_synced_txg(ds->ds_dir->dd_pool->dp_spa))
+ trysnap = ds->ds_trysnap_txg;
+ return (MAX(ds->ds_phys->ds_prev_snap_txg, trysnap));
+}
+
+int
+dsl_dataset_block_freeable(dsl_dataset_t *ds, uint64_t blk_birth)
+{
+ return (blk_birth > dsl_dataset_prev_snap_txg(ds));
+}
+
+/* ARGSUSED */
+static void
+dsl_dataset_evict(dmu_buf_t *db, void *dsv)
+{
+ dsl_dataset_t *ds = dsv;
+
+ ASSERT(ds->ds_owner == NULL || DSL_DATASET_IS_DESTROYED(ds));
+
+ dprintf_ds(ds, "evicting %s\n", "");
+
+ unique_remove(ds->ds_fsid_guid);
+
+ if (ds->ds_user_ptr != NULL)
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+
+ if (ds->ds_prev) {
+ dsl_dataset_drop_ref(ds->ds_prev, ds);
+ ds->ds_prev = NULL;
+ }
+
+ bplist_close(&ds->ds_deadlist);
+ if (ds->ds_dir)
+ dsl_dir_close(ds->ds_dir, ds);
+
+ ASSERT(!list_link_active(&ds->ds_synced_link));
+
+ mutex_destroy(&ds->ds_lock);
+ mutex_destroy(&ds->ds_opening_lock);
+ mutex_destroy(&ds->ds_deadlist.bpl_lock);
+ rw_destroy(&ds->ds_rwlock);
+ cv_destroy(&ds->ds_exclusive_cv);
+
+ kmem_free(ds, sizeof (dsl_dataset_t));
+}
+
+static int
+dsl_dataset_get_snapname(dsl_dataset_t *ds)
+{
+ dsl_dataset_phys_t *headphys;
+ int err;
+ dmu_buf_t *headdbuf;
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+
+ if (ds->ds_snapname[0])
+ return (0);
+ if (ds->ds_phys->ds_next_snap_obj == 0)
+ return (0);
+
+ err = dmu_bonus_hold(mos, ds->ds_dir->dd_phys->dd_head_dataset_obj,
+ FTAG, &headdbuf);
+ if (err)
+ return (err);
+ headphys = headdbuf->db_data;
+ err = zap_value_search(dp->dp_meta_objset,
+ headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname);
+ dmu_buf_rele(headdbuf, FTAG);
+ return (err);
+}
+
+static int
+dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value)
+{
+ objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
+ uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
+ matchtype_t mt;
+ int err;
+
+ if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
+ mt = MT_FIRST;
+ else
+ mt = MT_EXACT;
+
+ err = zap_lookup_norm(mos, snapobj, name, 8, 1,
+ value, mt, NULL, 0, NULL);
+ if (err == ENOTSUP && mt == MT_FIRST)
+ err = zap_lookup(mos, snapobj, name, 8, 1, value);
+ return (err);
+}
+
+static int
+dsl_dataset_snap_remove(dsl_dataset_t *ds, char *name, dmu_tx_t *tx)
+{
+ objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
+ uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
+ matchtype_t mt;
+ int err;
+
+ if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
+ mt = MT_FIRST;
+ else
+ mt = MT_EXACT;
+
+ err = zap_remove_norm(mos, snapobj, name, mt, tx);
+ if (err == ENOTSUP && mt == MT_FIRST)
+ err = zap_remove(mos, snapobj, name, tx);
+ return (err);
+}
+
+static int
+dsl_dataset_get_ref(dsl_pool_t *dp, uint64_t dsobj, void *tag,
+ dsl_dataset_t **dsp)
+{
+ objset_t *mos = dp->dp_meta_objset;
+ dmu_buf_t *dbuf;
+ dsl_dataset_t *ds;
+ int err;
+
+ ASSERT(RW_LOCK_HELD(&dp->dp_config_rwlock) ||
+ dsl_pool_sync_context(dp));
+
+ err = dmu_bonus_hold(mos, dsobj, tag, &dbuf);
+ if (err)
+ return (err);
+ ds = dmu_buf_get_user(dbuf);
+ if (ds == NULL) {
+ dsl_dataset_t *winner;
+
+ ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP);
+ ds->ds_dbuf = dbuf;
+ ds->ds_object = dsobj;
+ ds->ds_phys = dbuf->db_data;
+
+ mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&ds->ds_deadlist.bpl_lock, NULL, MUTEX_DEFAULT,
+ NULL);
+ rw_init(&ds->ds_rwlock, 0, 0, 0);
+ cv_init(&ds->ds_exclusive_cv, NULL, CV_DEFAULT, NULL);
+
+ err = bplist_open(&ds->ds_deadlist,
+ mos, ds->ds_phys->ds_deadlist_obj);
+ if (err == 0) {
+ err = dsl_dir_open_obj(dp,
+ ds->ds_phys->ds_dir_obj, NULL, ds, &ds->ds_dir);
+ }
+ if (err) {
+ /*
+ * we don't really need to close the blist if we
+ * just opened it.
+ */
+ mutex_destroy(&ds->ds_lock);
+ mutex_destroy(&ds->ds_opening_lock);
+ mutex_destroy(&ds->ds_deadlist.bpl_lock);
+ rw_destroy(&ds->ds_rwlock);
+ cv_destroy(&ds->ds_exclusive_cv);
+ kmem_free(ds, sizeof (dsl_dataset_t));
+ dmu_buf_rele(dbuf, tag);
+ return (err);
+ }
+
+ if (!dsl_dataset_is_snapshot(ds)) {
+ ds->ds_snapname[0] = '\0';
+ if (ds->ds_phys->ds_prev_snap_obj) {
+ err = dsl_dataset_get_ref(dp,
+ ds->ds_phys->ds_prev_snap_obj,
+ ds, &ds->ds_prev);
+ }
+
+ if (err == 0 && dsl_dir_is_clone(ds->ds_dir)) {
+ dsl_dataset_t *origin;
+
+ err = dsl_dataset_hold_obj(dp,
+ ds->ds_dir->dd_phys->dd_origin_obj,
+ FTAG, &origin);
+ if (err == 0) {
+ ds->ds_origin_txg =
+ origin->ds_phys->ds_creation_txg;
+ dsl_dataset_rele(origin, FTAG);
+ }
+ }
+ } else if (zfs_flags & ZFS_DEBUG_SNAPNAMES) {
+ err = dsl_dataset_get_snapname(ds);
+ }
+
+ if (err == 0 && !dsl_dataset_is_snapshot(ds)) {
+ /*
+ * In sync context, we're called with either no lock
+ * or with the write lock. If we're not syncing,
+ * we're always called with the read lock held.
+ */
+ boolean_t need_lock =
+ !RW_WRITE_HELD(&dp->dp_config_rwlock) &&
+ dsl_pool_sync_context(dp);
+
+ if (need_lock)
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+
+ err = dsl_prop_get_ds(ds,
+ "refreservation", sizeof (uint64_t), 1,
+ &ds->ds_reserved, NULL);
+ if (err == 0) {
+ err = dsl_prop_get_ds(ds,
+ "refquota", sizeof (uint64_t), 1,
+ &ds->ds_quota, NULL);
+ }
+
+ if (need_lock)
+ rw_exit(&dp->dp_config_rwlock);
+ } else {
+ ds->ds_reserved = ds->ds_quota = 0;
+ }
+
+ if (err == 0) {
+ winner = dmu_buf_set_user_ie(dbuf, ds, &ds->ds_phys,
+ dsl_dataset_evict);
+ }
+ if (err || winner) {
+ bplist_close(&ds->ds_deadlist);
+ if (ds->ds_prev)
+ dsl_dataset_drop_ref(ds->ds_prev, ds);
+ dsl_dir_close(ds->ds_dir, ds);
+ mutex_destroy(&ds->ds_lock);
+ mutex_destroy(&ds->ds_opening_lock);
+ mutex_destroy(&ds->ds_deadlist.bpl_lock);
+ rw_destroy(&ds->ds_rwlock);
+ cv_destroy(&ds->ds_exclusive_cv);
+ kmem_free(ds, sizeof (dsl_dataset_t));
+ if (err) {
+ dmu_buf_rele(dbuf, tag);
+ return (err);
+ }
+ ds = winner;
+ } else {
+ ds->ds_fsid_guid =
+ unique_insert(ds->ds_phys->ds_fsid_guid);
+ }
+ }
+ ASSERT3P(ds->ds_dbuf, ==, dbuf);
+ ASSERT3P(ds->ds_phys, ==, dbuf->db_data);
+ ASSERT(ds->ds_phys->ds_prev_snap_obj != 0 ||
+ spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN ||
+ dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap);
+ mutex_enter(&ds->ds_lock);
+ if (!dsl_pool_sync_context(dp) && DSL_DATASET_IS_DESTROYED(ds)) {
+ mutex_exit(&ds->ds_lock);
+ dmu_buf_rele(ds->ds_dbuf, tag);
+ return (ENOENT);
+ }
+ mutex_exit(&ds->ds_lock);
+ *dsp = ds;
+ return (0);
+}
+
+static int
+dsl_dataset_hold_ref(dsl_dataset_t *ds, void *tag)
+{
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ /*
+ * In syncing context we don't want the rwlock lock: there
+ * may be an existing writer waiting for sync phase to
+ * finish. We don't need to worry about such writers, since
+ * sync phase is single-threaded, so the writer can't be
+ * doing anything while we are active.
+ */
+ if (dsl_pool_sync_context(dp)) {
+ ASSERT(!DSL_DATASET_IS_DESTROYED(ds));
+ return (0);
+ }
+
+ /*
+ * Normal users will hold the ds_rwlock as a READER until they
+ * are finished (i.e., call dsl_dataset_rele()). "Owners" will
+ * drop their READER lock after they set the ds_owner field.
+ *
+ * If the dataset is being destroyed, the destroy thread will
+ * obtain a WRITER lock for exclusive access after it's done its
+ * open-context work and then change the ds_owner to
+ * dsl_reaper once destruction is assured. So threads
+ * may block here temporarily, until the "destructability" of
+ * the dataset is determined.
+ */
+ ASSERT(!RW_WRITE_HELD(&dp->dp_config_rwlock));
+ mutex_enter(&ds->ds_lock);
+ while (!rw_tryenter(&ds->ds_rwlock, RW_READER)) {
+ rw_exit(&dp->dp_config_rwlock);
+ cv_wait(&ds->ds_exclusive_cv, &ds->ds_lock);
+ if (DSL_DATASET_IS_DESTROYED(ds)) {
+ mutex_exit(&ds->ds_lock);
+ dsl_dataset_drop_ref(ds, tag);
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ return (ENOENT);
+ }
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ }
+ mutex_exit(&ds->ds_lock);
+ return (0);
+}
+
+int
+dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag,
+ dsl_dataset_t **dsp)
+{
+ int err = dsl_dataset_get_ref(dp, dsobj, tag, dsp);
+
+ if (err)
+ return (err);
+ return (dsl_dataset_hold_ref(*dsp, tag));
+}
+
+int
+dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, int flags, void *owner,
+ dsl_dataset_t **dsp)
+{
+ int err = dsl_dataset_hold_obj(dp, dsobj, owner, dsp);
+
+ ASSERT(DS_MODE_TYPE(flags) != DS_MODE_USER);
+
+ if (err)
+ return (err);
+ if (!dsl_dataset_tryown(*dsp, DS_MODE_IS_INCONSISTENT(flags), owner)) {
+ dsl_dataset_rele(*dsp, owner);
+ return (EBUSY);
+ }
+ return (0);
+}
+
+int
+dsl_dataset_hold(const char *name, void *tag, dsl_dataset_t **dsp)
+{
+ dsl_dir_t *dd;
+ dsl_pool_t *dp;
+ const char *snapname;
+ uint64_t obj;
+ int err = 0;
+
+ err = dsl_dir_open_spa(NULL, name, FTAG, &dd, &snapname);
+ if (err)
+ return (err);
+
+ dp = dd->dd_pool;
+ obj = dd->dd_phys->dd_head_dataset_obj;
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ if (obj)
+ err = dsl_dataset_get_ref(dp, obj, tag, dsp);
+ else
+ err = ENOENT;
+ if (err)
+ goto out;
+
+ err = dsl_dataset_hold_ref(*dsp, tag);
+
+ /* we may be looking for a snapshot */
+ if (err == 0 && snapname != NULL) {
+ dsl_dataset_t *ds = NULL;
+
+ if (*snapname++ != '@') {
+ dsl_dataset_rele(*dsp, tag);
+ err = ENOENT;
+ goto out;
+ }
+
+ dprintf("looking for snapshot '%s'\n", snapname);
+ err = dsl_dataset_snap_lookup(*dsp, snapname, &obj);
+ if (err == 0)
+ err = dsl_dataset_get_ref(dp, obj, tag, &ds);
+ dsl_dataset_rele(*dsp, tag);
+
+ ASSERT3U((err == 0), ==, (ds != NULL));
+
+ if (ds) {
+ mutex_enter(&ds->ds_lock);
+ if (ds->ds_snapname[0] == 0)
+ (void) strlcpy(ds->ds_snapname, snapname,
+ sizeof (ds->ds_snapname));
+ mutex_exit(&ds->ds_lock);
+ err = dsl_dataset_hold_ref(ds, tag);
+ *dsp = err ? NULL : ds;
+ }
+ }
+out:
+ rw_exit(&dp->dp_config_rwlock);
+ dsl_dir_close(dd, FTAG);
+ return (err);
+}
+
+int
+dsl_dataset_own(const char *name, int flags, void *owner, dsl_dataset_t **dsp)
+{
+ int err = dsl_dataset_hold(name, owner, dsp);
+ if (err)
+ return (err);
+ if ((*dsp)->ds_phys->ds_num_children > 0 &&
+ !DS_MODE_IS_READONLY(flags)) {
+ dsl_dataset_rele(*dsp, owner);
+ return (EROFS);
+ }
+ if (!dsl_dataset_tryown(*dsp, DS_MODE_IS_INCONSISTENT(flags), owner)) {
+ dsl_dataset_rele(*dsp, owner);
+ return (EBUSY);
+ }
+ return (0);
+}
+
+void
+dsl_dataset_name(dsl_dataset_t *ds, char *name)
+{
+ if (ds == NULL) {
+ (void) strcpy(name, "mos");
+ } else {
+ dsl_dir_name(ds->ds_dir, name);
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+ if (ds->ds_snapname[0]) {
+ (void) strcat(name, "@");
+ /*
+ * We use a "recursive" mutex so that we
+ * can call dprintf_ds() with ds_lock held.
+ */
+ if (!MUTEX_HELD(&ds->ds_lock)) {
+ mutex_enter(&ds->ds_lock);
+ (void) strcat(name, ds->ds_snapname);
+ mutex_exit(&ds->ds_lock);
+ } else {
+ (void) strcat(name, ds->ds_snapname);
+ }
+ }
+ }
+}
+
+static int
+dsl_dataset_namelen(dsl_dataset_t *ds)
+{
+ int result;
+
+ if (ds == NULL) {
+ result = 3; /* "mos" */
+ } else {
+ result = dsl_dir_namelen(ds->ds_dir);
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+ if (ds->ds_snapname[0]) {
+ ++result; /* adding one for the @-sign */
+ if (!MUTEX_HELD(&ds->ds_lock)) {
+ mutex_enter(&ds->ds_lock);
+ result += strlen(ds->ds_snapname);
+ mutex_exit(&ds->ds_lock);
+ } else {
+ result += strlen(ds->ds_snapname);
+ }
+ }
+ }
+
+ return (result);
+}
+
+void
+dsl_dataset_drop_ref(dsl_dataset_t *ds, void *tag)
+{
+ dmu_buf_rele(ds->ds_dbuf, tag);
+}
+
+void
+dsl_dataset_rele(dsl_dataset_t *ds, void *tag)
+{
+ if (!dsl_pool_sync_context(ds->ds_dir->dd_pool)) {
+ rw_exit(&ds->ds_rwlock);
+ }
+ dsl_dataset_drop_ref(ds, tag);
+}
+
+void
+dsl_dataset_disown(dsl_dataset_t *ds, void *owner)
+{
+ ASSERT((ds->ds_owner == owner && ds->ds_dbuf) ||
+ (DSL_DATASET_IS_DESTROYED(ds) && ds->ds_dbuf == NULL));
+
+ mutex_enter(&ds->ds_lock);
+ ds->ds_owner = NULL;
+ if (RW_WRITE_HELD(&ds->ds_rwlock)) {
+ rw_exit(&ds->ds_rwlock);
+ cv_broadcast(&ds->ds_exclusive_cv);
+ }
+ mutex_exit(&ds->ds_lock);
+ if (ds->ds_dbuf)
+ dsl_dataset_drop_ref(ds, owner);
+ else
+ dsl_dataset_evict(ds->ds_dbuf, ds);
+}
+
+boolean_t
+dsl_dataset_tryown(dsl_dataset_t *ds, boolean_t inconsistentok, void *owner)
+{
+ boolean_t gotit = FALSE;
+
+ mutex_enter(&ds->ds_lock);
+ if (ds->ds_owner == NULL &&
+ (!DS_IS_INCONSISTENT(ds) || inconsistentok)) {
+ ds->ds_owner = owner;
+ if (!dsl_pool_sync_context(ds->ds_dir->dd_pool))
+ rw_exit(&ds->ds_rwlock);
+ gotit = TRUE;
+ }
+ mutex_exit(&ds->ds_lock);
+ return (gotit);
+}
+
+void
+dsl_dataset_make_exclusive(dsl_dataset_t *ds, void *owner)
+{
+ ASSERT3P(owner, ==, ds->ds_owner);
+ if (!RW_WRITE_HELD(&ds->ds_rwlock))
+ rw_enter(&ds->ds_rwlock, RW_WRITER);
+}
+
+uint64_t
+dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
+ uint64_t flags, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = dd->dd_pool;
+ dmu_buf_t *dbuf;
+ dsl_dataset_phys_t *dsphys;
+ uint64_t dsobj;
+ objset_t *mos = dp->dp_meta_objset;
+
+ if (origin == NULL)
+ origin = dp->dp_origin_snap;
+
+ ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp);
+ ASSERT(origin == NULL || origin->ds_phys->ds_num_children > 0);
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
+
+ dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
+ DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
+ VERIFY(0 == dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
+ dmu_buf_will_dirty(dbuf, tx);
+ dsphys = dbuf->db_data;
+ bzero(dsphys, sizeof (dsl_dataset_phys_t));
+ dsphys->ds_dir_obj = dd->dd_object;
+ dsphys->ds_flags = flags;
+ dsphys->ds_fsid_guid = unique_create();
+ (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
+ sizeof (dsphys->ds_guid));
+ dsphys->ds_snapnames_zapobj =
+ zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP,
+ DMU_OT_NONE, 0, tx);
+ dsphys->ds_creation_time = gethrestime_sec();
+ dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg;
+ dsphys->ds_deadlist_obj =
+ bplist_create(mos, DSL_DEADLIST_BLOCKSIZE, tx);
+
+ if (origin) {
+ dsphys->ds_prev_snap_obj = origin->ds_object;
+ dsphys->ds_prev_snap_txg =
+ origin->ds_phys->ds_creation_txg;
+ dsphys->ds_used_bytes =
+ origin->ds_phys->ds_used_bytes;
+ dsphys->ds_compressed_bytes =
+ origin->ds_phys->ds_compressed_bytes;
+ dsphys->ds_uncompressed_bytes =
+ origin->ds_phys->ds_uncompressed_bytes;
+ dsphys->ds_bp = origin->ds_phys->ds_bp;
+ dsphys->ds_flags |= origin->ds_phys->ds_flags;
+
+ dmu_buf_will_dirty(origin->ds_dbuf, tx);
+ origin->ds_phys->ds_num_children++;
+
+ if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) {
+ if (origin->ds_phys->ds_next_clones_obj == 0) {
+ origin->ds_phys->ds_next_clones_obj =
+ zap_create(mos,
+ DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
+ }
+ VERIFY(0 == zap_add_int(mos,
+ origin->ds_phys->ds_next_clones_obj,
+ dsobj, tx));
+ }
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+ dd->dd_phys->dd_origin_obj = origin->ds_object;
+ }
+
+ if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
+ dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
+
+ dmu_buf_rele(dbuf, FTAG);
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+ dd->dd_phys->dd_head_dataset_obj = dsobj;
+
+ return (dsobj);
+}
+
+uint64_t
+dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname,
+ dsl_dataset_t *origin, uint64_t flags, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = pdd->dd_pool;
+ uint64_t dsobj, ddobj;
+ dsl_dir_t *dd;
+
+ ASSERT(lastname[0] != '@');
+
+ ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx);
+ VERIFY(0 == dsl_dir_open_obj(dp, ddobj, lastname, FTAG, &dd));
+
+ dsobj = dsl_dataset_create_sync_dd(dd, origin, flags, tx);
+
+ dsl_deleg_set_create_perms(dd, tx, cr);
+
+ dsl_dir_close(dd, FTAG);
+
+ return (dsobj);
+}
+
+struct destroyarg {
+ dsl_sync_task_group_t *dstg;
+ char *snapname;
+ char *failed;
+};
+
+static int
+dsl_snapshot_destroy_one(char *name, void *arg)
+{
+ struct destroyarg *da = arg;
+ dsl_dataset_t *ds;
+ char *cp;
+ int err;
+
+ (void) strcat(name, "@");
+ (void) strcat(name, da->snapname);
+ err = dsl_dataset_own(name, DS_MODE_READONLY | DS_MODE_INCONSISTENT,
+ da->dstg, &ds);
+ cp = strchr(name, '@');
+ *cp = '\0';
+ if (err == 0) {
+ dsl_dataset_make_exclusive(ds, da->dstg);
+ if (ds->ds_user_ptr) {
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+ ds->ds_user_ptr = NULL;
+ }
+ dsl_sync_task_create(da->dstg, dsl_dataset_destroy_check,
+ dsl_dataset_destroy_sync, ds, da->dstg, 0);
+ } else if (err == ENOENT) {
+ err = 0;
+ } else {
+ (void) strcpy(da->failed, name);
+ }
+ return (err);
+}
+
+/*
+ * Destroy 'snapname' in all descendants of 'fsname'.
+ */
+#pragma weak dmu_snapshots_destroy = dsl_snapshots_destroy
+int
+dsl_snapshots_destroy(char *fsname, char *snapname)
+{
+ int err;
+ struct destroyarg da;
+ dsl_sync_task_t *dst;
+ spa_t *spa;
+
+ err = spa_open(fsname, &spa, FTAG);
+ if (err)
+ return (err);
+ da.dstg = dsl_sync_task_group_create(spa_get_dsl(spa));
+ da.snapname = snapname;
+ da.failed = fsname;
+
+ err = dmu_objset_find(fsname,
+ dsl_snapshot_destroy_one, &da, DS_FIND_CHILDREN);
+
+ if (err == 0)
+ err = dsl_sync_task_group_wait(da.dstg);
+
+ for (dst = list_head(&da.dstg->dstg_tasks); dst;
+ dst = list_next(&da.dstg->dstg_tasks, dst)) {
+ dsl_dataset_t *ds = dst->dst_arg1;
+ /*
+ * Return the file system name that triggered the error
+ */
+ if (dst->dst_err) {
+ dsl_dataset_name(ds, fsname);
+ *strchr(fsname, '@') = '\0';
+ }
+ dsl_dataset_disown(ds, da.dstg);
+ }
+
+ dsl_sync_task_group_destroy(da.dstg);
+ spa_close(spa, FTAG);
+ return (err);
+}
+
+/*
+ * ds must be opened as OWNER. On return (whether successful or not),
+ * ds will be closed and caller can no longer dereference it.
+ */
+int
+dsl_dataset_destroy(dsl_dataset_t *ds, void *tag)
+{
+ int err;
+ dsl_sync_task_group_t *dstg;
+ objset_t *os;
+ dsl_dir_t *dd;
+ uint64_t obj;
+
+ if (dsl_dataset_is_snapshot(ds)) {
+ /* Destroying a snapshot is simpler */
+ dsl_dataset_make_exclusive(ds, tag);
+
+ if (ds->ds_user_ptr) {
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+ ds->ds_user_ptr = NULL;
+ }
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ dsl_dataset_destroy_check, dsl_dataset_destroy_sync,
+ ds, tag, 0);
+ goto out;
+ }
+
+ dd = ds->ds_dir;
+
+ /*
+ * Check for errors and mark this ds as inconsistent, in
+ * case we crash while freeing the objects.
+ */
+ err = dsl_sync_task_do(dd->dd_pool, dsl_dataset_destroy_begin_check,
+ dsl_dataset_destroy_begin_sync, ds, NULL, 0);
+ if (err)
+ goto out;
+
+ err = dmu_objset_open_ds(ds, DMU_OST_ANY, &os);
+ if (err)
+ goto out;
+
+ /*
+ * remove the objects in open context, so that we won't
+ * have too much to do in syncing context.
+ */
+ for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE,
+ ds->ds_phys->ds_prev_snap_txg)) {
+ /*
+ * Ignore errors, if there is not enough disk space
+ * we will deal with it in dsl_dataset_destroy_sync().
+ */
+ (void) dmu_free_object(os, obj);
+ }
+
+ dmu_objset_close(os);
+ if (err != ESRCH)
+ goto out;
+
+ rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
+ err = dsl_dir_open_obj(dd->dd_pool, dd->dd_object, NULL, FTAG, &dd);
+ rw_exit(&dd->dd_pool->dp_config_rwlock);
+
+ if (err)
+ goto out;
+
+ if (ds->ds_user_ptr) {
+ /*
+ * We need to sync out all in-flight IO before we try
+ * to evict (the dataset evict func is trying to clear
+ * the cached entries for this dataset in the ARC).
+ */
+ txg_wait_synced(dd->dd_pool, 0);
+ }
+
+ /*
+ * Blow away the dsl_dir + head dataset.
+ */
+ dsl_dataset_make_exclusive(ds, tag);
+ if (ds->ds_user_ptr) {
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+ ds->ds_user_ptr = NULL;
+ }
+ dstg = dsl_sync_task_group_create(ds->ds_dir->dd_pool);
+ dsl_sync_task_create(dstg, dsl_dataset_destroy_check,
+ dsl_dataset_destroy_sync, ds, tag, 0);
+ dsl_sync_task_create(dstg, dsl_dir_destroy_check,
+ dsl_dir_destroy_sync, dd, FTAG, 0);
+ err = dsl_sync_task_group_wait(dstg);
+ dsl_sync_task_group_destroy(dstg);
+ /* if it is successful, dsl_dir_destroy_sync will close the dd */
+ if (err)
+ dsl_dir_close(dd, FTAG);
+out:
+ dsl_dataset_disown(ds, tag);
+ return (err);
+}
+
+int
+dsl_dataset_rollback(dsl_dataset_t *ds, dmu_objset_type_t ost)
+{
+ int err;
+
+ ASSERT(ds->ds_owner);
+
+ dsl_dataset_make_exclusive(ds, ds->ds_owner);
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ dsl_dataset_rollback_check, dsl_dataset_rollback_sync,
+ ds, &ost, 0);
+ /* drop exclusive access */
+ mutex_enter(&ds->ds_lock);
+ rw_exit(&ds->ds_rwlock);
+ cv_broadcast(&ds->ds_exclusive_cv);
+ mutex_exit(&ds->ds_lock);
+ return (err);
+}
+
+void *
+dsl_dataset_set_user_ptr(dsl_dataset_t *ds,
+ void *p, dsl_dataset_evict_func_t func)
+{
+ void *old;
+
+ mutex_enter(&ds->ds_lock);
+ old = ds->ds_user_ptr;
+ if (old == NULL) {
+ ds->ds_user_ptr = p;
+ ds->ds_user_evict_func = func;
+ }
+ mutex_exit(&ds->ds_lock);
+ return (old);
+}
+
+void *
+dsl_dataset_get_user_ptr(dsl_dataset_t *ds)
+{
+ return (ds->ds_user_ptr);
+}
+
+
+blkptr_t *
+dsl_dataset_get_blkptr(dsl_dataset_t *ds)
+{
+ return (&ds->ds_phys->ds_bp);
+}
+
+void
+dsl_dataset_set_blkptr(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx)
+{
+ ASSERT(dmu_tx_is_syncing(tx));
+ /* If it's the meta-objset, set dp_meta_rootbp */
+ if (ds == NULL) {
+ tx->tx_pool->dp_meta_rootbp = *bp;
+ } else {
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_bp = *bp;
+ }
+}
+
+spa_t *
+dsl_dataset_get_spa(dsl_dataset_t *ds)
+{
+ return (ds->ds_dir->dd_pool->dp_spa);
+}
+
+void
+dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp;
+
+ if (ds == NULL) /* this is the meta-objset */
+ return;
+
+ ASSERT(ds->ds_user_ptr != NULL);
+
+ if (ds->ds_phys->ds_next_snap_obj != 0)
+ panic("dirtying snapshot!");
+
+ dp = ds->ds_dir->dd_pool;
+
+ if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg) == 0) {
+ /* up the hold count until we can be written out */
+ dmu_buf_add_ref(ds->ds_dbuf, ds);
+ }
+}
+
+/*
+ * The unique space in the head dataset can be calculated by subtracting
+ * the space used in the most recent snapshot, that is still being used
+ * in this file system, from the space currently in use. To figure out
+ * the space in the most recent snapshot still in use, we need to take
+ * the total space used in the snapshot and subtract out the space that
+ * has been freed up since the snapshot was taken.
+ */
+static void
+dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds)
+{
+ uint64_t mrs_used;
+ uint64_t dlused, dlcomp, dluncomp;
+
+ ASSERT(ds->ds_object == ds->ds_dir->dd_phys->dd_head_dataset_obj);
+
+ if (ds->ds_phys->ds_prev_snap_obj != 0)
+ mrs_used = ds->ds_prev->ds_phys->ds_used_bytes;
+ else
+ mrs_used = 0;
+
+ VERIFY(0 == bplist_space(&ds->ds_deadlist, &dlused, &dlcomp,
+ &dluncomp));
+
+ ASSERT3U(dlused, <=, mrs_used);
+ ds->ds_phys->ds_unique_bytes =
+ ds->ds_phys->ds_used_bytes - (mrs_used - dlused);
+
+ if (!DS_UNIQUE_IS_ACCURATE(ds) &&
+ spa_version(ds->ds_dir->dd_pool->dp_spa) >=
+ SPA_VERSION_UNIQUE_ACCURATE)
+ ds->ds_phys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
+}
+
+static uint64_t
+dsl_dataset_unique(dsl_dataset_t *ds)
+{
+ if (!DS_UNIQUE_IS_ACCURATE(ds) && !dsl_dataset_is_snapshot(ds))
+ dsl_dataset_recalc_head_uniq(ds);
+
+ return (ds->ds_phys->ds_unique_bytes);
+}
+
+struct killarg {
+ dsl_dataset_t *ds;
+ zio_t *zio;
+ dmu_tx_t *tx;
+};
+
+/* ARGSUSED */
+static int
+kill_blkptr(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
+ const dnode_phys_t *dnp, void *arg)
+{
+ struct killarg *ka = arg;
+
+ if (bp == NULL)
+ return (0);
+
+ ASSERT3U(bp->blk_birth, >, ka->ds->ds_phys->ds_prev_snap_txg);
+ (void) dsl_dataset_block_kill(ka->ds, bp, ka->zio, ka->tx);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static int
+dsl_dataset_rollback_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ dmu_objset_type_t *ost = arg2;
+
+ /*
+ * We can only roll back to emptyness if it is a ZPL objset.
+ */
+ if (*ost != DMU_OST_ZFS && ds->ds_phys->ds_prev_snap_txg == 0)
+ return (EINVAL);
+
+ /*
+ * This must not be a snapshot.
+ */
+ if (ds->ds_phys->ds_next_snap_obj != 0)
+ return (EINVAL);
+
+ /*
+ * If we made changes this txg, traverse_dataset won't find
+ * them. Try again.
+ */
+ if (ds->ds_phys->ds_bp.blk_birth >= tx->tx_txg)
+ return (EAGAIN);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dsl_dataset_rollback_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ dmu_objset_type_t *ost = arg2;
+ objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+
+ /*
+ * Before the roll back destroy the zil.
+ */
+ if (ds->ds_user_ptr != NULL) {
+ zil_rollback_destroy(
+ ((objset_impl_t *)ds->ds_user_ptr)->os_zil, tx);
+
+ /*
+ * We need to make sure that the objset_impl_t is reopened after
+ * we do the rollback, otherwise it will have the wrong
+ * objset_phys_t. Normally this would happen when this
+ * dataset-open is closed, thus causing the
+ * dataset to be immediately evicted. But when doing "zfs recv
+ * -F", we reopen the objset before that, so that there is no
+ * window where the dataset is closed and inconsistent.
+ */
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+ ds->ds_user_ptr = NULL;
+ }
+
+ /* Transfer space that was freed since last snap back to the head. */
+ {
+ uint64_t used;
+
+ VERIFY(0 == bplist_space_birthrange(&ds->ds_deadlist,
+ ds->ds_origin_txg, UINT64_MAX, &used));
+ dsl_dir_transfer_space(ds->ds_dir, used,
+ DD_USED_SNAP, DD_USED_HEAD, tx);
+ }
+
+ /* Zero out the deadlist. */
+ bplist_close(&ds->ds_deadlist);
+ bplist_destroy(mos, ds->ds_phys->ds_deadlist_obj, tx);
+ ds->ds_phys->ds_deadlist_obj =
+ bplist_create(mos, DSL_DEADLIST_BLOCKSIZE, tx);
+ VERIFY(0 == bplist_open(&ds->ds_deadlist, mos,
+ ds->ds_phys->ds_deadlist_obj));
+
+ {
+ /* Free blkptrs that we gave birth to */
+ zio_t *zio;
+ struct killarg ka;
+
+ zio = zio_root(tx->tx_pool->dp_spa, NULL, NULL,
+ ZIO_FLAG_MUSTSUCCEED);
+ ka.ds = ds;
+ ka.zio = zio;
+ ka.tx = tx;
+ (void) traverse_dataset(ds, ds->ds_phys->ds_prev_snap_txg,
+ TRAVERSE_POST, kill_blkptr, &ka);
+ (void) zio_wait(zio);
+ }
+
+ ASSERT(!(ds->ds_phys->ds_flags & DS_FLAG_UNIQUE_ACCURATE) ||
+ ds->ds_phys->ds_unique_bytes == 0);
+
+ if (ds->ds_prev && ds->ds_prev != ds->ds_dir->dd_pool->dp_origin_snap) {
+ /* Change our contents to that of the prev snapshot */
+
+ ASSERT3U(ds->ds_prev->ds_object, ==,
+ ds->ds_phys->ds_prev_snap_obj);
+ ASSERT3U(ds->ds_phys->ds_used_bytes, <=,
+ ds->ds_prev->ds_phys->ds_used_bytes);
+
+ ds->ds_phys->ds_bp = ds->ds_prev->ds_phys->ds_bp;
+ ds->ds_phys->ds_used_bytes =
+ ds->ds_prev->ds_phys->ds_used_bytes;
+ ds->ds_phys->ds_compressed_bytes =
+ ds->ds_prev->ds_phys->ds_compressed_bytes;
+ ds->ds_phys->ds_uncompressed_bytes =
+ ds->ds_prev->ds_phys->ds_uncompressed_bytes;
+ ds->ds_phys->ds_flags = ds->ds_prev->ds_phys->ds_flags;
+
+ if (ds->ds_prev->ds_phys->ds_next_snap_obj == ds->ds_object) {
+ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
+ ds->ds_prev->ds_phys->ds_unique_bytes = 0;
+ }
+ } else {
+ objset_impl_t *osi;
+
+ ASSERT3U(ds->ds_phys->ds_used_bytes, ==, 0);
+ ASSERT3U(ds->ds_phys->ds_compressed_bytes, ==, 0);
+ ASSERT3U(ds->ds_phys->ds_uncompressed_bytes, ==, 0);
+
+ bzero(&ds->ds_phys->ds_bp, sizeof (blkptr_t));
+ ds->ds_phys->ds_flags = 0;
+ ds->ds_phys->ds_unique_bytes = 0;
+ if (spa_version(ds->ds_dir->dd_pool->dp_spa) >=
+ SPA_VERSION_UNIQUE_ACCURATE)
+ ds->ds_phys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
+
+ osi = dmu_objset_create_impl(ds->ds_dir->dd_pool->dp_spa, ds,
+ &ds->ds_phys->ds_bp, *ost, tx);
+#ifdef _KERNEL
+ zfs_create_fs(&osi->os, kcred, NULL, tx);
+#endif
+ }
+
+ spa_history_internal_log(LOG_DS_ROLLBACK, ds->ds_dir->dd_pool->dp_spa,
+ tx, cr, "dataset = %llu", ds->ds_object);
+}
+
+/* ARGSUSED */
+static int
+dsl_dataset_destroy_begin_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
+ uint64_t count;
+ int err;
+
+ /*
+ * Can't delete a head dataset if there are snapshots of it.
+ * (Except if the only snapshots are from the branch we cloned
+ * from.)
+ */
+ if (ds->ds_prev != NULL &&
+ ds->ds_prev->ds_phys->ds_next_snap_obj == ds->ds_object)
+ return (EINVAL);
+
+ /*
+ * This is really a dsl_dir thing, but check it here so that
+ * we'll be less likely to leave this dataset inconsistent &
+ * nearly destroyed.
+ */
+ err = zap_count(mos, ds->ds_dir->dd_phys->dd_child_dir_zapobj, &count);
+ if (err)
+ return (err);
+ if (count != 0)
+ return (EEXIST);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dsl_dataset_destroy_begin_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ /* Mark it as inconsistent on-disk, in case we crash */
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_flags |= DS_FLAG_INCONSISTENT;
+
+ spa_history_internal_log(LOG_DS_DESTROY_BEGIN, dp->dp_spa, tx,
+ cr, "dataset = %llu", ds->ds_object);
+}
+
+/* ARGSUSED */
+int
+dsl_dataset_destroy_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+
+ /* we have an owner hold, so noone else can destroy us */
+ ASSERT(!DSL_DATASET_IS_DESTROYED(ds));
+
+ /* Can't delete a branch point. */
+ if (ds->ds_phys->ds_num_children > 1)
+ return (EEXIST);
+
+ /*
+ * Can't delete a head dataset if there are snapshots of it.
+ * (Except if the only snapshots are from the branch we cloned
+ * from.)
+ */
+ if (ds->ds_prev != NULL &&
+ ds->ds_prev->ds_phys->ds_next_snap_obj == ds->ds_object)
+ return (EINVAL);
+
+ /*
+ * If we made changes this txg, traverse_dsl_dataset won't find
+ * them. Try again.
+ */
+ if (ds->ds_phys->ds_bp.blk_birth >= tx->tx_txg)
+ return (EAGAIN);
+
+ /* XXX we should do some i/o error checking... */
+ return (0);
+}
+
+struct refsarg {
+ kmutex_t lock;
+ boolean_t gone;
+ kcondvar_t cv;
+};
+
+/* ARGSUSED */
+static void
+dsl_dataset_refs_gone(dmu_buf_t *db, void *argv)
+{
+ struct refsarg *arg = argv;
+
+ mutex_enter(&arg->lock);
+ arg->gone = TRUE;
+ cv_signal(&arg->cv);
+ mutex_exit(&arg->lock);
+}
+
+static void
+dsl_dataset_drain_refs(dsl_dataset_t *ds, void *tag)
+{
+ struct refsarg arg;
+
+ mutex_init(&arg.lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&arg.cv, NULL, CV_DEFAULT, NULL);
+ arg.gone = FALSE;
+ (void) dmu_buf_update_user(ds->ds_dbuf, ds, &arg, &ds->ds_phys,
+ dsl_dataset_refs_gone);
+ dmu_buf_rele(ds->ds_dbuf, tag);
+ mutex_enter(&arg.lock);
+ while (!arg.gone)
+ cv_wait(&arg.cv, &arg.lock);
+ ASSERT(arg.gone);
+ mutex_exit(&arg.lock);
+ ds->ds_dbuf = NULL;
+ ds->ds_phys = NULL;
+ mutex_destroy(&arg.lock);
+ cv_destroy(&arg.cv);
+}
+
+void
+dsl_dataset_destroy_sync(void *arg1, void *tag, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ zio_t *zio;
+ int err;
+ int after_branch_point = FALSE;
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+ dsl_dataset_t *ds_prev = NULL;
+ uint64_t obj;
+
+ ASSERT(ds->ds_owner);
+ ASSERT3U(ds->ds_phys->ds_num_children, <=, 1);
+ ASSERT(ds->ds_prev == NULL ||
+ ds->ds_prev->ds_phys->ds_next_snap_obj != ds->ds_object);
+ ASSERT3U(ds->ds_phys->ds_bp.blk_birth, <=, tx->tx_txg);
+
+ /* signal any waiters that this dataset is going away */
+ mutex_enter(&ds->ds_lock);
+ ds->ds_owner = dsl_reaper;
+ cv_broadcast(&ds->ds_exclusive_cv);
+ mutex_exit(&ds->ds_lock);
+
+ /* Remove our reservation */
+ if (ds->ds_reserved != 0) {
+ uint64_t val = 0;
+ dsl_dataset_set_reservation_sync(ds, &val, cr, tx);
+ ASSERT3U(ds->ds_reserved, ==, 0);
+ }
+
+ ASSERT(RW_WRITE_HELD(&dp->dp_config_rwlock));
+
+ dsl_pool_ds_destroyed(ds, tx);
+
+ obj = ds->ds_object;
+
+ if (ds->ds_phys->ds_prev_snap_obj != 0) {
+ if (ds->ds_prev) {
+ ds_prev = ds->ds_prev;
+ } else {
+ VERIFY(0 == dsl_dataset_hold_obj(dp,
+ ds->ds_phys->ds_prev_snap_obj, FTAG, &ds_prev));
+ }
+ after_branch_point =
+ (ds_prev->ds_phys->ds_next_snap_obj != obj);
+
+ dmu_buf_will_dirty(ds_prev->ds_dbuf, tx);
+ if (after_branch_point &&
+ ds_prev->ds_phys->ds_next_clones_obj != 0) {
+ VERIFY(0 == zap_remove_int(mos,
+ ds_prev->ds_phys->ds_next_clones_obj, obj, tx));
+ if (ds->ds_phys->ds_next_snap_obj != 0) {
+ VERIFY(0 == zap_add_int(mos,
+ ds_prev->ds_phys->ds_next_clones_obj,
+ ds->ds_phys->ds_next_snap_obj, tx));
+ }
+ }
+ if (after_branch_point &&
+ ds->ds_phys->ds_next_snap_obj == 0) {
+ /* This clone is toast. */
+ ASSERT(ds_prev->ds_phys->ds_num_children > 1);
+ ds_prev->ds_phys->ds_num_children--;
+ } else if (!after_branch_point) {
+ ds_prev->ds_phys->ds_next_snap_obj =
+ ds->ds_phys->ds_next_snap_obj;
+ }
+ }
+
+ zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
+
+ if (ds->ds_phys->ds_next_snap_obj != 0) {
+ blkptr_t bp;
+ dsl_dataset_t *ds_next;
+ uint64_t itor = 0;
+ uint64_t old_unique;
+ int64_t used = 0, compressed = 0, uncompressed = 0;
+
+ VERIFY(0 == dsl_dataset_hold_obj(dp,
+ ds->ds_phys->ds_next_snap_obj, FTAG, &ds_next));
+ ASSERT3U(ds_next->ds_phys->ds_prev_snap_obj, ==, obj);
+
+ old_unique = dsl_dataset_unique(ds_next);
+
+ dmu_buf_will_dirty(ds_next->ds_dbuf, tx);
+ ds_next->ds_phys->ds_prev_snap_obj =
+ ds->ds_phys->ds_prev_snap_obj;
+ ds_next->ds_phys->ds_prev_snap_txg =
+ ds->ds_phys->ds_prev_snap_txg;
+ ASSERT3U(ds->ds_phys->ds_prev_snap_txg, ==,
+ ds_prev ? ds_prev->ds_phys->ds_creation_txg : 0);
+
+ /*
+ * Transfer to our deadlist (which will become next's
+ * new deadlist) any entries from next's current
+ * deadlist which were born before prev, and free the
+ * other entries.
+ *
+ * XXX we're doing this long task with the config lock held
+ */
+ while (bplist_iterate(&ds_next->ds_deadlist, &itor, &bp) == 0) {
+ if (bp.blk_birth <= ds->ds_phys->ds_prev_snap_txg) {
+ VERIFY(0 == bplist_enqueue(&ds->ds_deadlist,
+ &bp, tx));
+ if (ds_prev && !after_branch_point &&
+ bp.blk_birth >
+ ds_prev->ds_phys->ds_prev_snap_txg) {
+ ds_prev->ds_phys->ds_unique_bytes +=
+ bp_get_dasize(dp->dp_spa, &bp);
+ }
+ } else {
+ used += bp_get_dasize(dp->dp_spa, &bp);
+ compressed += BP_GET_PSIZE(&bp);
+ uncompressed += BP_GET_UCSIZE(&bp);
+ /* XXX check return value? */
+ (void) dsl_free(zio, dp, tx->tx_txg,
+ &bp, NULL, NULL, ARC_NOWAIT);
+ }
+ }
+
+ ASSERT3U(used, ==, ds->ds_phys->ds_unique_bytes);
+
+ /* change snapused */
+ dsl_dir_diduse_space(ds->ds_dir, DD_USED_SNAP,
+ -used, -compressed, -uncompressed, tx);
+
+ /* free next's deadlist */
+ bplist_close(&ds_next->ds_deadlist);
+ bplist_destroy(mos, ds_next->ds_phys->ds_deadlist_obj, tx);
+
+ /* set next's deadlist to our deadlist */
+ bplist_close(&ds->ds_deadlist);
+ ds_next->ds_phys->ds_deadlist_obj =
+ ds->ds_phys->ds_deadlist_obj;
+ VERIFY(0 == bplist_open(&ds_next->ds_deadlist, mos,
+ ds_next->ds_phys->ds_deadlist_obj));
+ ds->ds_phys->ds_deadlist_obj = 0;
+
+ if (ds_next->ds_phys->ds_next_snap_obj != 0) {
+ /*
+ * Update next's unique to include blocks which
+ * were previously shared by only this snapshot
+ * and it. Those blocks will be born after the
+ * prev snap and before this snap, and will have
+ * died after the next snap and before the one
+ * after that (ie. be on the snap after next's
+ * deadlist).
+ *
+ * XXX we're doing this long task with the
+ * config lock held
+ */
+ dsl_dataset_t *ds_after_next;
+ uint64_t space;
+
+ VERIFY(0 == dsl_dataset_hold_obj(dp,
+ ds_next->ds_phys->ds_next_snap_obj,
+ FTAG, &ds_after_next));
+
+ VERIFY(0 ==
+ bplist_space_birthrange(&ds_after_next->ds_deadlist,
+ ds->ds_phys->ds_prev_snap_txg,
+ ds->ds_phys->ds_creation_txg, &space));
+ ds_next->ds_phys->ds_unique_bytes += space;
+
+ dsl_dataset_rele(ds_after_next, FTAG);
+ ASSERT3P(ds_next->ds_prev, ==, NULL);
+ } else {
+ ASSERT3P(ds_next->ds_prev, ==, ds);
+ dsl_dataset_drop_ref(ds_next->ds_prev, ds_next);
+ ds_next->ds_prev = NULL;
+ if (ds_prev) {
+ VERIFY(0 == dsl_dataset_get_ref(dp,
+ ds->ds_phys->ds_prev_snap_obj,
+ ds_next, &ds_next->ds_prev));
+ }
+
+ dsl_dataset_recalc_head_uniq(ds_next);
+
+ /*
+ * Reduce the amount of our unconsmed refreservation
+ * being charged to our parent by the amount of
+ * new unique data we have gained.
+ */
+ if (old_unique < ds_next->ds_reserved) {
+ int64_t mrsdelta;
+ uint64_t new_unique =
+ ds_next->ds_phys->ds_unique_bytes;
+
+ ASSERT(old_unique <= new_unique);
+ mrsdelta = MIN(new_unique - old_unique,
+ ds_next->ds_reserved - old_unique);
+ dsl_dir_diduse_space(ds->ds_dir,
+ DD_USED_REFRSRV, -mrsdelta, 0, 0, tx);
+ }
+ }
+ dsl_dataset_rele(ds_next, FTAG);
+ } else {
+ /*
+ * There's no next snapshot, so this is a head dataset.
+ * Destroy the deadlist. Unless it's a clone, the
+ * deadlist should be empty. (If it's a clone, it's
+ * safe to ignore the deadlist contents.)
+ */
+ struct killarg ka;
+
+ ASSERT(after_branch_point || bplist_empty(&ds->ds_deadlist));
+ bplist_close(&ds->ds_deadlist);
+ bplist_destroy(mos, ds->ds_phys->ds_deadlist_obj, tx);
+ ds->ds_phys->ds_deadlist_obj = 0;
+
+ /*
+ * Free everything that we point to (that's born after
+ * the previous snapshot, if we are a clone)
+ *
+ * NB: this should be very quick, because we already
+ * freed all the objects in open context.
+ */
+ ka.ds = ds;
+ ka.zio = zio;
+ ka.tx = tx;
+ err = traverse_dataset(ds, ds->ds_phys->ds_prev_snap_txg,
+ TRAVERSE_POST, kill_blkptr, &ka);
+ ASSERT3U(err, ==, 0);
+ ASSERT(spa_version(dp->dp_spa) < SPA_VERSION_UNIQUE_ACCURATE ||
+ ds->ds_phys->ds_unique_bytes == 0);
+ }
+
+ err = zio_wait(zio);
+ ASSERT3U(err, ==, 0);
+
+ if (ds->ds_dir->dd_phys->dd_head_dataset_obj == ds->ds_object) {
+ /* Erase the link in the dir */
+ dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
+ ds->ds_dir->dd_phys->dd_head_dataset_obj = 0;
+ ASSERT(ds->ds_phys->ds_snapnames_zapobj != 0);
+ err = zap_destroy(mos, ds->ds_phys->ds_snapnames_zapobj, tx);
+ ASSERT(err == 0);
+ } else {
+ /* remove from snapshot namespace */
+ dsl_dataset_t *ds_head;
+ ASSERT(ds->ds_phys->ds_snapnames_zapobj == 0);
+ VERIFY(0 == dsl_dataset_hold_obj(dp,
+ ds->ds_dir->dd_phys->dd_head_dataset_obj, FTAG, &ds_head));
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+#ifdef ZFS_DEBUG
+ {
+ uint64_t val;
+
+ err = dsl_dataset_snap_lookup(ds_head,
+ ds->ds_snapname, &val);
+ ASSERT3U(err, ==, 0);
+ ASSERT3U(val, ==, obj);
+ }
+#endif
+ err = dsl_dataset_snap_remove(ds_head, ds->ds_snapname, tx);
+ ASSERT(err == 0);
+ dsl_dataset_rele(ds_head, FTAG);
+ }
+
+ if (ds_prev && ds->ds_prev != ds_prev)
+ dsl_dataset_rele(ds_prev, FTAG);
+
+ spa_prop_clear_bootfs(dp->dp_spa, ds->ds_object, tx);
+ spa_history_internal_log(LOG_DS_DESTROY, dp->dp_spa, tx,
+ cr, "dataset = %llu", ds->ds_object);
+
+ if (ds->ds_phys->ds_next_clones_obj != 0) {
+ uint64_t count;
+ ASSERT(0 == zap_count(mos,
+ ds->ds_phys->ds_next_clones_obj, &count) && count == 0);
+ VERIFY(0 == dmu_object_free(mos,
+ ds->ds_phys->ds_next_clones_obj, tx));
+ }
+ if (ds->ds_phys->ds_props_obj != 0)
+ VERIFY(0 == zap_destroy(mos, ds->ds_phys->ds_props_obj, tx));
+ dsl_dir_close(ds->ds_dir, ds);
+ ds->ds_dir = NULL;
+ dsl_dataset_drain_refs(ds, tag);
+ VERIFY(0 == dmu_object_free(mos, obj, tx));
+}
+
+static int
+dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx)
+{
+ uint64_t asize;
+
+ if (!dmu_tx_is_syncing(tx))
+ return (0);
+
+ /*
+ * If there's an fs-only reservation, any blocks that might become
+ * owned by the snapshot dataset must be accommodated by space
+ * outside of the reservation.
+ */
+ asize = MIN(dsl_dataset_unique(ds), ds->ds_reserved);
+ if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, FALSE))
+ return (ENOSPC);
+
+ /*
+ * Propogate any reserved space for this snapshot to other
+ * snapshot checks in this sync group.
+ */
+ if (asize > 0)
+ dsl_dir_willuse_space(ds->ds_dir, asize, tx);
+
+ return (0);
+}
+
+/* ARGSUSED */
+int
+dsl_dataset_snapshot_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ const char *snapname = arg2;
+ int err;
+ uint64_t value;
+
+ /*
+ * We don't allow multiple snapshots of the same txg. If there
+ * is already one, try again.
+ */
+ if (ds->ds_phys->ds_prev_snap_txg >= tx->tx_txg)
+ return (EAGAIN);
+
+ /*
+ * Check for conflicting name snapshot name.
+ */
+ err = dsl_dataset_snap_lookup(ds, snapname, &value);
+ if (err == 0)
+ return (EEXIST);
+ if (err != ENOENT)
+ return (err);
+
+ /*
+ * Check that the dataset's name is not too long. Name consists
+ * of the dataset's length + 1 for the @-sign + snapshot name's length
+ */
+ if (dsl_dataset_namelen(ds) + 1 + strlen(snapname) >= MAXNAMELEN)
+ return (ENAMETOOLONG);
+
+ err = dsl_dataset_snapshot_reserve_space(ds, tx);
+ if (err)
+ return (err);
+
+ ds->ds_trysnap_txg = tx->tx_txg;
+ return (0);
+}
+
+void
+dsl_dataset_snapshot_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ const char *snapname = arg2;
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+ dmu_buf_t *dbuf;
+ dsl_dataset_phys_t *dsphys;
+ uint64_t dsobj, crtxg;
+ objset_t *mos = dp->dp_meta_objset;
+ int err;
+
+ ASSERT(RW_WRITE_HELD(&dp->dp_config_rwlock));
+
+ /*
+ * The origin's ds_creation_txg has to be < TXG_INITIAL
+ */
+ if (strcmp(snapname, ORIGIN_DIR_NAME) == 0)
+ crtxg = 1;
+ else
+ crtxg = tx->tx_txg;
+
+ dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
+ DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
+ VERIFY(0 == dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
+ dmu_buf_will_dirty(dbuf, tx);
+ dsphys = dbuf->db_data;
+ bzero(dsphys, sizeof (dsl_dataset_phys_t));
+ dsphys->ds_dir_obj = ds->ds_dir->dd_object;
+ dsphys->ds_fsid_guid = unique_create();
+ (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
+ sizeof (dsphys->ds_guid));
+ dsphys->ds_prev_snap_obj = ds->ds_phys->ds_prev_snap_obj;
+ dsphys->ds_prev_snap_txg = ds->ds_phys->ds_prev_snap_txg;
+ dsphys->ds_next_snap_obj = ds->ds_object;
+ dsphys->ds_num_children = 1;
+ dsphys->ds_creation_time = gethrestime_sec();
+ dsphys->ds_creation_txg = crtxg;
+ dsphys->ds_deadlist_obj = ds->ds_phys->ds_deadlist_obj;
+ dsphys->ds_used_bytes = ds->ds_phys->ds_used_bytes;
+ dsphys->ds_compressed_bytes = ds->ds_phys->ds_compressed_bytes;
+ dsphys->ds_uncompressed_bytes = ds->ds_phys->ds_uncompressed_bytes;
+ dsphys->ds_flags = ds->ds_phys->ds_flags;
+ dsphys->ds_bp = ds->ds_phys->ds_bp;
+ dmu_buf_rele(dbuf, FTAG);
+
+ ASSERT3U(ds->ds_prev != 0, ==, ds->ds_phys->ds_prev_snap_obj != 0);
+ if (ds->ds_prev) {
+ uint64_t next_clones_obj =
+ ds->ds_prev->ds_phys->ds_next_clones_obj;
+ ASSERT(ds->ds_prev->ds_phys->ds_next_snap_obj ==
+ ds->ds_object ||
+ ds->ds_prev->ds_phys->ds_num_children > 1);
+ if (ds->ds_prev->ds_phys->ds_next_snap_obj == ds->ds_object) {
+ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
+ ASSERT3U(ds->ds_phys->ds_prev_snap_txg, ==,
+ ds->ds_prev->ds_phys->ds_creation_txg);
+ ds->ds_prev->ds_phys->ds_next_snap_obj = dsobj;
+ } else if (next_clones_obj != 0) {
+ VERIFY3U(0, ==, zap_remove_int(mos,
+ next_clones_obj, dsphys->ds_next_snap_obj, tx));
+ VERIFY3U(0, ==, zap_add_int(mos,
+ next_clones_obj, dsobj, tx));
+ }
+ }
+
+ /*
+ * If we have a reference-reservation on this dataset, we will
+ * need to increase the amount of refreservation being charged
+ * since our unique space is going to zero.
+ */
+ if (ds->ds_reserved) {
+ int64_t add = MIN(dsl_dataset_unique(ds), ds->ds_reserved);
+ dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV,
+ add, 0, 0, tx);
+ }
+
+ bplist_close(&ds->ds_deadlist);
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ASSERT3U(ds->ds_phys->ds_prev_snap_txg, <, tx->tx_txg);
+ ds->ds_phys->ds_prev_snap_obj = dsobj;
+ ds->ds_phys->ds_prev_snap_txg = crtxg;
+ ds->ds_phys->ds_unique_bytes = 0;
+ if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
+ ds->ds_phys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
+ ds->ds_phys->ds_deadlist_obj =
+ bplist_create(mos, DSL_DEADLIST_BLOCKSIZE, tx);
+ VERIFY(0 == bplist_open(&ds->ds_deadlist, mos,
+ ds->ds_phys->ds_deadlist_obj));
+
+ dprintf("snap '%s' -> obj %llu\n", snapname, dsobj);
+ err = zap_add(mos, ds->ds_phys->ds_snapnames_zapobj,
+ snapname, 8, 1, &dsobj, tx);
+ ASSERT(err == 0);
+
+ if (ds->ds_prev)
+ dsl_dataset_drop_ref(ds->ds_prev, ds);
+ VERIFY(0 == dsl_dataset_get_ref(dp,
+ ds->ds_phys->ds_prev_snap_obj, ds, &ds->ds_prev));
+
+ dsl_pool_ds_snapshotted(ds, tx);
+
+ spa_history_internal_log(LOG_DS_SNAPSHOT, dp->dp_spa, tx, cr,
+ "dataset = %llu", dsobj);
+}
+
+void
+dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx)
+{
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(ds->ds_user_ptr != NULL);
+ ASSERT(ds->ds_phys->ds_next_snap_obj == 0);
+
+ /*
+ * in case we had to change ds_fsid_guid when we opened it,
+ * sync it out now.
+ */
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_fsid_guid = ds->ds_fsid_guid;
+
+ dsl_dir_dirty(ds->ds_dir, tx);
+ dmu_objset_sync(ds->ds_user_ptr, zio, tx);
+}
+
+void
+dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv)
+{
+ uint64_t refd, avail, uobjs, aobjs;
+
+ dsl_dir_stats(ds->ds_dir, nv);
+
+ dsl_dataset_space(ds, &refd, &avail, &uobjs, &aobjs);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE, avail);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED, refd);
+
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION,
+ ds->ds_phys->ds_creation_time);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG,
+ ds->ds_phys->ds_creation_txg);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA,
+ ds->ds_quota);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION,
+ ds->ds_reserved);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID,
+ ds->ds_phys->ds_guid);
+
+ if (ds->ds_phys->ds_next_snap_obj) {
+ /*
+ * This is a snapshot; override the dd's space used with
+ * our unique space and compression ratio.
+ */
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED,
+ ds->ds_phys->ds_unique_bytes);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO,
+ ds->ds_phys->ds_compressed_bytes == 0 ? 100 :
+ (ds->ds_phys->ds_uncompressed_bytes * 100 /
+ ds->ds_phys->ds_compressed_bytes));
+ }
+}
+
+void
+dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat)
+{
+ stat->dds_creation_txg = ds->ds_phys->ds_creation_txg;
+ stat->dds_inconsistent = ds->ds_phys->ds_flags & DS_FLAG_INCONSISTENT;
+ stat->dds_guid = ds->ds_phys->ds_guid;
+ if (ds->ds_phys->ds_next_snap_obj) {
+ stat->dds_is_snapshot = B_TRUE;
+ stat->dds_num_clones = ds->ds_phys->ds_num_children - 1;
+ }
+
+ /* clone origin is really a dsl_dir thing... */
+ rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
+ if (dsl_dir_is_clone(ds->ds_dir)) {
+ dsl_dataset_t *ods;
+
+ VERIFY(0 == dsl_dataset_get_ref(ds->ds_dir->dd_pool,
+ ds->ds_dir->dd_phys->dd_origin_obj, FTAG, &ods));
+ dsl_dataset_name(ods, stat->dds_origin);
+ dsl_dataset_drop_ref(ods, FTAG);
+ }
+ rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
+}
+
+uint64_t
+dsl_dataset_fsid_guid(dsl_dataset_t *ds)
+{
+ return (ds->ds_fsid_guid);
+}
+
+void
+dsl_dataset_space(dsl_dataset_t *ds,
+ uint64_t *refdbytesp, uint64_t *availbytesp,
+ uint64_t *usedobjsp, uint64_t *availobjsp)
+{
+ *refdbytesp = ds->ds_phys->ds_used_bytes;
+ *availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE);
+ if (ds->ds_reserved > ds->ds_phys->ds_unique_bytes)
+ *availbytesp += ds->ds_reserved - ds->ds_phys->ds_unique_bytes;
+ if (ds->ds_quota != 0) {
+ /*
+ * Adjust available bytes according to refquota
+ */
+ if (*refdbytesp < ds->ds_quota)
+ *availbytesp = MIN(*availbytesp,
+ ds->ds_quota - *refdbytesp);
+ else
+ *availbytesp = 0;
+ }
+ *usedobjsp = ds->ds_phys->ds_bp.blk_fill;
+ *availobjsp = DN_MAX_OBJECT - *usedobjsp;
+}
+
+boolean_t
+dsl_dataset_modified_since_lastsnap(dsl_dataset_t *ds)
+{
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ ASSERT(RW_LOCK_HELD(&dp->dp_config_rwlock) ||
+ dsl_pool_sync_context(dp));
+ if (ds->ds_prev == NULL)
+ return (B_FALSE);
+ if (ds->ds_phys->ds_bp.blk_birth >
+ ds->ds_prev->ds_phys->ds_creation_txg)
+ return (B_TRUE);
+ return (B_FALSE);
+}
+
+/* ARGSUSED */
+static int
+dsl_dataset_snapshot_rename_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ char *newsnapname = arg2;
+ dsl_dir_t *dd = ds->ds_dir;
+ dsl_dataset_t *hds;
+ uint64_t val;
+ int err;
+
+ err = dsl_dataset_hold_obj(dd->dd_pool,
+ dd->dd_phys->dd_head_dataset_obj, FTAG, &hds);
+ if (err)
+ return (err);
+
+ /* new name better not be in use */
+ err = dsl_dataset_snap_lookup(hds, newsnapname, &val);
+ dsl_dataset_rele(hds, FTAG);
+
+ if (err == 0)
+ err = EEXIST;
+ else if (err == ENOENT)
+ err = 0;
+
+ /* dataset name + 1 for the "@" + the new snapshot name must fit */
+ if (dsl_dir_namelen(ds->ds_dir) + 1 + strlen(newsnapname) >= MAXNAMELEN)
+ err = ENAMETOOLONG;
+
+ return (err);
+}
+
+static void
+dsl_dataset_snapshot_rename_sync(void *arg1, void *arg2,
+ cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ const char *newsnapname = arg2;
+ dsl_dir_t *dd = ds->ds_dir;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ dsl_dataset_t *hds;
+ int err;
+
+ ASSERT(ds->ds_phys->ds_next_snap_obj != 0);
+
+ VERIFY(0 == dsl_dataset_hold_obj(dd->dd_pool,
+ dd->dd_phys->dd_head_dataset_obj, FTAG, &hds));
+
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+ err = dsl_dataset_snap_remove(hds, ds->ds_snapname, tx);
+ ASSERT3U(err, ==, 0);
+ mutex_enter(&ds->ds_lock);
+ (void) strcpy(ds->ds_snapname, newsnapname);
+ mutex_exit(&ds->ds_lock);
+ err = zap_add(mos, hds->ds_phys->ds_snapnames_zapobj,
+ ds->ds_snapname, 8, 1, &ds->ds_object, tx);
+ ASSERT3U(err, ==, 0);
+
+ spa_history_internal_log(LOG_DS_RENAME, dd->dd_pool->dp_spa, tx,
+ cr, "dataset = %llu", ds->ds_object);
+ dsl_dataset_rele(hds, FTAG);
+}
+
+struct renamesnaparg {
+ dsl_sync_task_group_t *dstg;
+ char failed[MAXPATHLEN];
+ char *oldsnap;
+ char *newsnap;
+};
+
+static int
+dsl_snapshot_rename_one(char *name, void *arg)
+{
+ struct renamesnaparg *ra = arg;
+ dsl_dataset_t *ds = NULL;
+ char *cp;
+ int err;
+
+ cp = name + strlen(name);
+ *cp = '@';
+ (void) strcpy(cp + 1, ra->oldsnap);
+
+ /*
+ * For recursive snapshot renames the parent won't be changing
+ * so we just pass name for both the to/from argument.
+ */
+ err = zfs_secpolicy_rename_perms(name, name, CRED());
+ if (err == ENOENT) {
+ return (0);
+ } else if (err) {
+ (void) strcpy(ra->failed, name);
+ return (err);
+ }
+
+#ifdef _KERNEL
+ /*
+ * For all filesystems undergoing rename, we'll need to unmount it.
+ */
+ (void) zfs_unmount_snap(name, NULL);
+#endif
+ err = dsl_dataset_hold(name, ra->dstg, &ds);
+ *cp = '\0';
+ if (err == ENOENT) {
+ return (0);
+ } else if (err) {
+ (void) strcpy(ra->failed, name);
+ return (err);
+ }
+
+ dsl_sync_task_create(ra->dstg, dsl_dataset_snapshot_rename_check,
+ dsl_dataset_snapshot_rename_sync, ds, ra->newsnap, 0);
+
+ return (0);
+}
+
+static int
+dsl_recursive_rename(char *oldname, const char *newname)
+{
+ int err;
+ struct renamesnaparg *ra;
+ dsl_sync_task_t *dst;
+ spa_t *spa;
+ char *cp, *fsname = spa_strdup(oldname);
+ int len = strlen(oldname);
+
+ /* truncate the snapshot name to get the fsname */
+ cp = strchr(fsname, '@');
+ *cp = '\0';
+
+ err = spa_open(fsname, &spa, FTAG);
+ if (err) {
+ kmem_free(fsname, len + 1);
+ return (err);
+ }
+ ra = kmem_alloc(sizeof (struct renamesnaparg), KM_SLEEP);
+ ra->dstg = dsl_sync_task_group_create(spa_get_dsl(spa));
+
+ ra->oldsnap = strchr(oldname, '@') + 1;
+ ra->newsnap = strchr(newname, '@') + 1;
+ *ra->failed = '\0';
+
+ err = dmu_objset_find(fsname, dsl_snapshot_rename_one, ra,
+ DS_FIND_CHILDREN);
+ kmem_free(fsname, len + 1);
+
+ if (err == 0) {
+ err = dsl_sync_task_group_wait(ra->dstg);
+ }
+
+ for (dst = list_head(&ra->dstg->dstg_tasks); dst;
+ dst = list_next(&ra->dstg->dstg_tasks, dst)) {
+ dsl_dataset_t *ds = dst->dst_arg1;
+ if (dst->dst_err) {
+ dsl_dir_name(ds->ds_dir, ra->failed);
+ (void) strcat(ra->failed, "@");
+ (void) strcat(ra->failed, ra->newsnap);
+ }
+ dsl_dataset_rele(ds, ra->dstg);
+ }
+
+ if (err)
+ (void) strcpy(oldname, ra->failed);
+
+ dsl_sync_task_group_destroy(ra->dstg);
+ kmem_free(ra, sizeof (struct renamesnaparg));
+ spa_close(spa, FTAG);
+ return (err);
+}
+
+static int
+dsl_valid_rename(char *oldname, void *arg)
+{
+ int delta = *(int *)arg;
+
+ if (strlen(oldname) + delta >= MAXNAMELEN)
+ return (ENAMETOOLONG);
+
+ return (0);
+}
+
+#pragma weak dmu_objset_rename = dsl_dataset_rename
+int
+dsl_dataset_rename(char *oldname, const char *newname, boolean_t recursive)
+{
+ dsl_dir_t *dd;
+ dsl_dataset_t *ds;
+ const char *tail;
+ int err;
+
+ err = dsl_dir_open(oldname, FTAG, &dd, &tail);
+ if (err)
+ return (err);
+ if (tail == NULL) {
+ int delta = strlen(newname) - strlen(oldname);
+
+ /* if we're growing, validate child name lengths */
+ if (delta > 0)
+ err = dmu_objset_find(oldname, dsl_valid_rename,
+ &delta, DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS);
+
+ if (!err)
+ err = dsl_dir_rename(dd, newname);
+ dsl_dir_close(dd, FTAG);
+ return (err);
+ }
+ if (tail[0] != '@') {
+ /* the name ended in a nonexistant component */
+ dsl_dir_close(dd, FTAG);
+ return (ENOENT);
+ }
+
+ dsl_dir_close(dd, FTAG);
+
+ /* new name must be snapshot in same filesystem */
+ tail = strchr(newname, '@');
+ if (tail == NULL)
+ return (EINVAL);
+ tail++;
+ if (strncmp(oldname, newname, tail - newname) != 0)
+ return (EXDEV);
+
+ if (recursive) {
+ err = dsl_recursive_rename(oldname, newname);
+ } else {
+ err = dsl_dataset_hold(oldname, FTAG, &ds);
+ if (err)
+ return (err);
+
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ dsl_dataset_snapshot_rename_check,
+ dsl_dataset_snapshot_rename_sync, ds, (char *)tail, 1);
+
+ dsl_dataset_rele(ds, FTAG);
+ }
+
+ return (err);
+}
+
+struct promotenode {
+ list_node_t link;
+ dsl_dataset_t *ds;
+};
+
+struct promotearg {
+ list_t shared_snaps, origin_snaps, clone_snaps;
+ dsl_dataset_t *origin_origin, *origin_head;
+ uint64_t used, comp, uncomp, unique, cloneusedsnap, originusedsnap;
+};
+
+static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep);
+
+/* ARGSUSED */
+static int
+dsl_dataset_promote_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *hds = arg1;
+ struct promotearg *pa = arg2;
+ struct promotenode *snap = list_head(&pa->shared_snaps);
+ dsl_dataset_t *origin_ds = snap->ds;
+ int err;
+
+ /* Check that it is a real clone */
+ if (!dsl_dir_is_clone(hds->ds_dir))
+ return (EINVAL);
+
+ /* Since this is so expensive, don't do the preliminary check */
+ if (!dmu_tx_is_syncing(tx))
+ return (0);
+
+ if (hds->ds_phys->ds_flags & DS_FLAG_NOPROMOTE)
+ return (EXDEV);
+
+ /* compute origin's new unique space */
+ snap = list_tail(&pa->clone_snaps);
+ ASSERT3U(snap->ds->ds_phys->ds_prev_snap_obj, ==, origin_ds->ds_object);
+ err = bplist_space_birthrange(&snap->ds->ds_deadlist,
+ origin_ds->ds_phys->ds_prev_snap_txg, UINT64_MAX, &pa->unique);
+ if (err)
+ return (err);
+
+ /*
+ * Walk the snapshots that we are moving
+ *
+ * Compute space to transfer. Consider the incremental changes
+ * to used for each snapshot:
+ * (my used) = (prev's used) + (blocks born) - (blocks killed)
+ * So each snapshot gave birth to:
+ * (blocks born) = (my used) - (prev's used) + (blocks killed)
+ * So a sequence would look like:
+ * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
+ * Which simplifies to:
+ * uN + kN + kN-1 + ... + k1 + k0
+ * Note however, if we stop before we reach the ORIGIN we get:
+ * uN + kN + kN-1 + ... + kM - uM-1
+ */
+ pa->used = origin_ds->ds_phys->ds_used_bytes;
+ pa->comp = origin_ds->ds_phys->ds_compressed_bytes;
+ pa->uncomp = origin_ds->ds_phys->ds_uncompressed_bytes;
+ for (snap = list_head(&pa->shared_snaps); snap;
+ snap = list_next(&pa->shared_snaps, snap)) {
+ uint64_t val, dlused, dlcomp, dluncomp;
+ dsl_dataset_t *ds = snap->ds;
+
+ /* Check that the snapshot name does not conflict */
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+ err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val);
+ if (err == 0)
+ return (EEXIST);
+ if (err != ENOENT)
+ return (err);
+
+ /* The very first snapshot does not have a deadlist */
+ if (ds->ds_phys->ds_prev_snap_obj == 0)
+ continue;
+
+ if (err = bplist_space(&ds->ds_deadlist,
+ &dlused, &dlcomp, &dluncomp))
+ return (err);
+ pa->used += dlused;
+ pa->comp += dlcomp;
+ pa->uncomp += dluncomp;
+ }
+
+ /*
+ * If we are a clone of a clone then we never reached ORIGIN,
+ * so we need to subtract out the clone origin's used space.
+ */
+ if (pa->origin_origin) {
+ pa->used -= pa->origin_origin->ds_phys->ds_used_bytes;
+ pa->comp -= pa->origin_origin->ds_phys->ds_compressed_bytes;
+ pa->uncomp -= pa->origin_origin->ds_phys->ds_uncompressed_bytes;
+ }
+
+ /* Check that there is enough space here */
+ err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir,
+ pa->used);
+ if (err)
+ return (err);
+
+ /*
+ * Compute the amounts of space that will be used by snapshots
+ * after the promotion (for both origin and clone). For each,
+ * it is the amount of space that will be on all of their
+ * deadlists (that was not born before their new origin).
+ */
+ if (hds->ds_dir->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
+ uint64_t space;
+
+ /*
+ * Note, typically this will not be a clone of a clone,
+ * so snap->ds->ds_origin_txg will be < TXG_INITIAL, so
+ * these snaplist_space() -> bplist_space_birthrange()
+ * calls will be fast because they do not have to
+ * iterate over all bps.
+ */
+ snap = list_head(&pa->origin_snaps);
+ err = snaplist_space(&pa->shared_snaps,
+ snap->ds->ds_origin_txg, &pa->cloneusedsnap);
+ if (err)
+ return (err);
+
+ err = snaplist_space(&pa->clone_snaps,
+ snap->ds->ds_origin_txg, &space);
+ if (err)
+ return (err);
+ pa->cloneusedsnap += space;
+ }
+ if (origin_ds->ds_dir->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
+ err = snaplist_space(&pa->origin_snaps,
+ origin_ds->ds_phys->ds_creation_txg, &pa->originusedsnap);
+ if (err)
+ return (err);
+ }
+
+ return (0);
+}
+
+static void
+dsl_dataset_promote_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *hds = arg1;
+ struct promotearg *pa = arg2;
+ struct promotenode *snap = list_head(&pa->shared_snaps);
+ dsl_dataset_t *origin_ds = snap->ds;
+ dsl_dataset_t *origin_head;
+ dsl_dir_t *dd = hds->ds_dir;
+ dsl_pool_t *dp = hds->ds_dir->dd_pool;
+ dsl_dir_t *odd = NULL;
+ uint64_t oldnext_obj;
+ int64_t delta;
+
+ ASSERT(0 == (hds->ds_phys->ds_flags & DS_FLAG_NOPROMOTE));
+
+ snap = list_head(&pa->origin_snaps);
+ origin_head = snap->ds;
+
+ /*
+ * We need to explicitly open odd, since origin_ds's dd will be
+ * changing.
+ */
+ VERIFY(0 == dsl_dir_open_obj(dp, origin_ds->ds_dir->dd_object,
+ NULL, FTAG, &odd));
+
+ /* change origin's next snap */
+ dmu_buf_will_dirty(origin_ds->ds_dbuf, tx);
+ oldnext_obj = origin_ds->ds_phys->ds_next_snap_obj;
+ snap = list_tail(&pa->clone_snaps);
+ ASSERT3U(snap->ds->ds_phys->ds_prev_snap_obj, ==, origin_ds->ds_object);
+ origin_ds->ds_phys->ds_next_snap_obj = snap->ds->ds_object;
+
+ /* change the origin's next clone */
+ if (origin_ds->ds_phys->ds_next_clones_obj) {
+ VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
+ origin_ds->ds_phys->ds_next_clones_obj,
+ origin_ds->ds_phys->ds_next_snap_obj, tx));
+ VERIFY3U(0, ==, zap_add_int(dp->dp_meta_objset,
+ origin_ds->ds_phys->ds_next_clones_obj,
+ oldnext_obj, tx));
+ }
+
+ /* change origin */
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+ ASSERT3U(dd->dd_phys->dd_origin_obj, ==, origin_ds->ds_object);
+ dd->dd_phys->dd_origin_obj = odd->dd_phys->dd_origin_obj;
+ hds->ds_origin_txg = origin_head->ds_origin_txg;
+ dmu_buf_will_dirty(odd->dd_dbuf, tx);
+ odd->dd_phys->dd_origin_obj = origin_ds->ds_object;
+ origin_head->ds_origin_txg = origin_ds->ds_phys->ds_creation_txg;
+
+ /* move snapshots to this dir */
+ for (snap = list_head(&pa->shared_snaps); snap;
+ snap = list_next(&pa->shared_snaps, snap)) {
+ dsl_dataset_t *ds = snap->ds;
+
+ /* unregister props as dsl_dir is changing */
+ if (ds->ds_user_ptr) {
+ ds->ds_user_evict_func(ds, ds->ds_user_ptr);
+ ds->ds_user_ptr = NULL;
+ }
+ /* move snap name entry */
+ VERIFY(0 == dsl_dataset_get_snapname(ds));
+ VERIFY(0 == dsl_dataset_snap_remove(origin_head,
+ ds->ds_snapname, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset,
+ hds->ds_phys->ds_snapnames_zapobj, ds->ds_snapname,
+ 8, 1, &ds->ds_object, tx));
+ /* change containing dsl_dir */
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ASSERT3U(ds->ds_phys->ds_dir_obj, ==, odd->dd_object);
+ ds->ds_phys->ds_dir_obj = dd->dd_object;
+ ASSERT3P(ds->ds_dir, ==, odd);
+ dsl_dir_close(ds->ds_dir, ds);
+ VERIFY(0 == dsl_dir_open_obj(dp, dd->dd_object,
+ NULL, ds, &ds->ds_dir));
+
+ ASSERT3U(dsl_prop_numcb(ds), ==, 0);
+ }
+
+ /*
+ * Change space accounting.
+ * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
+ * both be valid, or both be 0 (resulting in delta == 0). This
+ * is true for each of {clone,origin} independently.
+ */
+
+ delta = pa->cloneusedsnap -
+ dd->dd_phys->dd_used_breakdown[DD_USED_SNAP];
+ ASSERT3S(delta, >=, 0);
+ ASSERT3U(pa->used, >=, delta);
+ dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx);
+ dsl_dir_diduse_space(dd, DD_USED_HEAD,
+ pa->used - delta, pa->comp, pa->uncomp, tx);
+
+ delta = pa->originusedsnap -
+ odd->dd_phys->dd_used_breakdown[DD_USED_SNAP];
+ ASSERT3S(delta, <=, 0);
+ ASSERT3U(pa->used, >=, -delta);
+ dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx);
+ dsl_dir_diduse_space(odd, DD_USED_HEAD,
+ -pa->used - delta, -pa->comp, -pa->uncomp, tx);
+
+ origin_ds->ds_phys->ds_unique_bytes = pa->unique;
+
+ /* log history record */
+ spa_history_internal_log(LOG_DS_PROMOTE, dd->dd_pool->dp_spa, tx,
+ cr, "dataset = %llu", hds->ds_object);
+
+ dsl_dir_close(odd, FTAG);
+}
+
+static char *snaplist_tag = "snaplist";
+/*
+ * Make a list of dsl_dataset_t's for the snapshots between first_obj
+ * (exclusive) and last_obj (inclusive). The list will be in reverse
+ * order (last_obj will be the list_head()). If first_obj == 0, do all
+ * snapshots back to this dataset's origin.
+ */
+static int
+snaplist_make(dsl_pool_t *dp, boolean_t own,
+ uint64_t first_obj, uint64_t last_obj, list_t *l)
+{
+ uint64_t obj = last_obj;
+
+ ASSERT(RW_LOCK_HELD(&dp->dp_config_rwlock));
+
+ list_create(l, sizeof (struct promotenode),
+ offsetof(struct promotenode, link));
+
+ while (obj != first_obj) {
+ dsl_dataset_t *ds;
+ struct promotenode *snap;
+ int err;
+
+ if (own) {
+ err = dsl_dataset_own_obj(dp, obj,
+ 0, snaplist_tag, &ds);
+ if (err == 0)
+ dsl_dataset_make_exclusive(ds, snaplist_tag);
+ } else {
+ err = dsl_dataset_hold_obj(dp, obj, snaplist_tag, &ds);
+ }
+ if (err == ENOENT) {
+ /* lost race with snapshot destroy */
+ struct promotenode *last = list_tail(l);
+ ASSERT(obj != last->ds->ds_phys->ds_prev_snap_obj);
+ obj = last->ds->ds_phys->ds_prev_snap_obj;
+ continue;
+ } else if (err) {
+ return (err);
+ }
+
+ if (first_obj == 0)
+ first_obj = ds->ds_dir->dd_phys->dd_origin_obj;
+
+ snap = kmem_alloc(sizeof (struct promotenode), KM_SLEEP);
+ snap->ds = ds;
+ list_insert_tail(l, snap);
+ obj = ds->ds_phys->ds_prev_snap_obj;
+ }
+
+ return (0);
+}
+
+static int
+snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep)
+{
+ struct promotenode *snap;
+
+ *spacep = 0;
+ for (snap = list_head(l); snap; snap = list_next(l, snap)) {
+ uint64_t used;
+ int err = bplist_space_birthrange(&snap->ds->ds_deadlist,
+ mintxg, UINT64_MAX, &used);
+ if (err)
+ return (err);
+ *spacep += used;
+ }
+ return (0);
+}
+
+static void
+snaplist_destroy(list_t *l, boolean_t own)
+{
+ struct promotenode *snap;
+
+ if (!list_link_active(&l->list_head))
+ return;
+
+ while ((snap = list_tail(l)) != NULL) {
+ list_remove(l, snap);
+ if (own)
+ dsl_dataset_disown(snap->ds, snaplist_tag);
+ else
+ dsl_dataset_rele(snap->ds, snaplist_tag);
+ kmem_free(snap, sizeof (struct promotenode));
+ }
+ list_destroy(l);
+}
+
+/*
+ * Promote a clone. Nomenclature note:
+ * "clone" or "cds": the original clone which is being promoted
+ * "origin" or "ods": the snapshot which is originally clone's origin
+ * "origin head" or "ohds": the dataset which is the head
+ * (filesystem/volume) for the origin
+ * "origin origin": the origin of the origin's filesystem (typically
+ * NULL, indicating that the clone is not a clone of a clone).
+ */
+int
+dsl_dataset_promote(const char *name)
+{
+ dsl_dataset_t *ds;
+ dsl_dir_t *dd;
+ dsl_pool_t *dp;
+ dmu_object_info_t doi;
+ struct promotearg pa = { 0 };
+ struct promotenode *snap;
+ int err;
+
+ err = dsl_dataset_hold(name, FTAG, &ds);
+ if (err)
+ return (err);
+ dd = ds->ds_dir;
+ dp = dd->dd_pool;
+
+ err = dmu_object_info(dp->dp_meta_objset,
+ ds->ds_phys->ds_snapnames_zapobj, &doi);
+ if (err) {
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+ }
+
+ if (dsl_dataset_is_snapshot(ds) || dd->dd_phys->dd_origin_obj == 0) {
+ dsl_dataset_rele(ds, FTAG);
+ return (EINVAL);
+ }
+
+ /*
+ * We are going to inherit all the snapshots taken before our
+ * origin (i.e., our new origin will be our parent's origin).
+ * Take ownership of them so that we can rename them into our
+ * namespace.
+ */
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+
+ err = snaplist_make(dp, B_TRUE, 0, dd->dd_phys->dd_origin_obj,
+ &pa.shared_snaps);
+ if (err != 0)
+ goto out;
+
+ err = snaplist_make(dp, B_FALSE, 0, ds->ds_object, &pa.clone_snaps);
+ if (err != 0)
+ goto out;
+
+ snap = list_head(&pa.shared_snaps);
+ ASSERT3U(snap->ds->ds_object, ==, dd->dd_phys->dd_origin_obj);
+ err = snaplist_make(dp, B_FALSE, dd->dd_phys->dd_origin_obj,
+ snap->ds->ds_dir->dd_phys->dd_head_dataset_obj, &pa.origin_snaps);
+ if (err != 0)
+ goto out;
+
+ if (dsl_dir_is_clone(snap->ds->ds_dir)) {
+ err = dsl_dataset_own_obj(dp,
+ snap->ds->ds_dir->dd_phys->dd_origin_obj,
+ 0, FTAG, &pa.origin_origin);
+ if (err != 0)
+ goto out;
+ }
+
+out:
+ rw_exit(&dp->dp_config_rwlock);
+
+ /*
+ * Add in 128x the snapnames zapobj size, since we will be moving
+ * a bunch of snapnames to the promoted ds, and dirtying their
+ * bonus buffers.
+ */
+ if (err == 0) {
+ err = dsl_sync_task_do(dp, dsl_dataset_promote_check,
+ dsl_dataset_promote_sync, ds, &pa,
+ 2 + 2 * doi.doi_physical_blks);
+ }
+
+ snaplist_destroy(&pa.shared_snaps, B_TRUE);
+ snaplist_destroy(&pa.clone_snaps, B_FALSE);
+ snaplist_destroy(&pa.origin_snaps, B_FALSE);
+ if (pa.origin_origin)
+ dsl_dataset_disown(pa.origin_origin, FTAG);
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+}
+
+struct cloneswaparg {
+ dsl_dataset_t *cds; /* clone dataset */
+ dsl_dataset_t *ohds; /* origin's head dataset */
+ boolean_t force;
+ int64_t unused_refres_delta; /* change in unconsumed refreservation */
+};
+
+/* ARGSUSED */
+static int
+dsl_dataset_clone_swap_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ struct cloneswaparg *csa = arg1;
+
+ /* they should both be heads */
+ if (dsl_dataset_is_snapshot(csa->cds) ||
+ dsl_dataset_is_snapshot(csa->ohds))
+ return (EINVAL);
+
+ /* the branch point should be just before them */
+ if (csa->cds->ds_prev != csa->ohds->ds_prev)
+ return (EINVAL);
+
+ /* cds should be the clone */
+ if (csa->cds->ds_prev->ds_phys->ds_next_snap_obj !=
+ csa->ohds->ds_object)
+ return (EINVAL);
+
+ /* the clone should be a child of the origin */
+ if (csa->cds->ds_dir->dd_parent != csa->ohds->ds_dir)
+ return (EINVAL);
+
+ /* ohds shouldn't be modified unless 'force' */
+ if (!csa->force && dsl_dataset_modified_since_lastsnap(csa->ohds))
+ return (ETXTBSY);
+
+ /* adjust amount of any unconsumed refreservation */
+ csa->unused_refres_delta =
+ (int64_t)MIN(csa->ohds->ds_reserved,
+ csa->ohds->ds_phys->ds_unique_bytes) -
+ (int64_t)MIN(csa->ohds->ds_reserved,
+ csa->cds->ds_phys->ds_unique_bytes);
+
+ if (csa->unused_refres_delta > 0 &&
+ csa->unused_refres_delta >
+ dsl_dir_space_available(csa->ohds->ds_dir, NULL, 0, TRUE))
+ return (ENOSPC);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dsl_dataset_clone_swap_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ struct cloneswaparg *csa = arg1;
+ dsl_pool_t *dp = csa->cds->ds_dir->dd_pool;
+
+ ASSERT(csa->cds->ds_reserved == 0);
+ ASSERT(csa->cds->ds_quota == csa->ohds->ds_quota);
+
+ dmu_buf_will_dirty(csa->cds->ds_dbuf, tx);
+ dmu_buf_will_dirty(csa->ohds->ds_dbuf, tx);
+ dmu_buf_will_dirty(csa->cds->ds_prev->ds_dbuf, tx);
+
+ if (csa->cds->ds_user_ptr != NULL) {
+ csa->cds->ds_user_evict_func(csa->cds, csa->cds->ds_user_ptr);
+ csa->cds->ds_user_ptr = NULL;
+ }
+
+ if (csa->ohds->ds_user_ptr != NULL) {
+ csa->ohds->ds_user_evict_func(csa->ohds,
+ csa->ohds->ds_user_ptr);
+ csa->ohds->ds_user_ptr = NULL;
+ }
+
+ /* reset origin's unique bytes */
+ VERIFY(0 == bplist_space_birthrange(&csa->cds->ds_deadlist,
+ csa->cds->ds_prev->ds_phys->ds_prev_snap_txg, UINT64_MAX,
+ &csa->cds->ds_prev->ds_phys->ds_unique_bytes));
+
+ /* swap blkptrs */
+ {
+ blkptr_t tmp;
+ tmp = csa->ohds->ds_phys->ds_bp;
+ csa->ohds->ds_phys->ds_bp = csa->cds->ds_phys->ds_bp;
+ csa->cds->ds_phys->ds_bp = tmp;
+ }
+
+ /* set dd_*_bytes */
+ {
+ int64_t dused, dcomp, duncomp;
+ uint64_t cdl_used, cdl_comp, cdl_uncomp;
+ uint64_t odl_used, odl_comp, odl_uncomp;
+
+ ASSERT3U(csa->cds->ds_dir->dd_phys->
+ dd_used_breakdown[DD_USED_SNAP], ==, 0);
+
+ VERIFY(0 == bplist_space(&csa->cds->ds_deadlist, &cdl_used,
+ &cdl_comp, &cdl_uncomp));
+ VERIFY(0 == bplist_space(&csa->ohds->ds_deadlist, &odl_used,
+ &odl_comp, &odl_uncomp));
+
+ dused = csa->cds->ds_phys->ds_used_bytes + cdl_used -
+ (csa->ohds->ds_phys->ds_used_bytes + odl_used);
+ dcomp = csa->cds->ds_phys->ds_compressed_bytes + cdl_comp -
+ (csa->ohds->ds_phys->ds_compressed_bytes + odl_comp);
+ duncomp = csa->cds->ds_phys->ds_uncompressed_bytes +
+ cdl_uncomp -
+ (csa->ohds->ds_phys->ds_uncompressed_bytes + odl_uncomp);
+
+ dsl_dir_diduse_space(csa->ohds->ds_dir, DD_USED_HEAD,
+ dused, dcomp, duncomp, tx);
+ dsl_dir_diduse_space(csa->cds->ds_dir, DD_USED_HEAD,
+ -dused, -dcomp, -duncomp, tx);
+
+ /*
+ * The difference in the space used by snapshots is the
+ * difference in snapshot space due to the head's
+ * deadlist (since that's the only thing that's
+ * changing that affects the snapused).
+ */
+ VERIFY(0 == bplist_space_birthrange(&csa->cds->ds_deadlist,
+ csa->ohds->ds_origin_txg, UINT64_MAX, &cdl_used));
+ VERIFY(0 == bplist_space_birthrange(&csa->ohds->ds_deadlist,
+ csa->ohds->ds_origin_txg, UINT64_MAX, &odl_used));
+ dsl_dir_transfer_space(csa->ohds->ds_dir, cdl_used - odl_used,
+ DD_USED_HEAD, DD_USED_SNAP, tx);
+ }
+
+#define SWITCH64(x, y) \
+ { \
+ uint64_t __tmp = (x); \
+ (x) = (y); \
+ (y) = __tmp; \
+ }
+
+ /* swap ds_*_bytes */
+ SWITCH64(csa->ohds->ds_phys->ds_used_bytes,
+ csa->cds->ds_phys->ds_used_bytes);
+ SWITCH64(csa->ohds->ds_phys->ds_compressed_bytes,
+ csa->cds->ds_phys->ds_compressed_bytes);
+ SWITCH64(csa->ohds->ds_phys->ds_uncompressed_bytes,
+ csa->cds->ds_phys->ds_uncompressed_bytes);
+ SWITCH64(csa->ohds->ds_phys->ds_unique_bytes,
+ csa->cds->ds_phys->ds_unique_bytes);
+
+ /* apply any parent delta for change in unconsumed refreservation */
+ dsl_dir_diduse_space(csa->ohds->ds_dir, DD_USED_REFRSRV,
+ csa->unused_refres_delta, 0, 0, tx);
+
+ /* swap deadlists */
+ bplist_close(&csa->cds->ds_deadlist);
+ bplist_close(&csa->ohds->ds_deadlist);
+ SWITCH64(csa->ohds->ds_phys->ds_deadlist_obj,
+ csa->cds->ds_phys->ds_deadlist_obj);
+ VERIFY(0 == bplist_open(&csa->cds->ds_deadlist, dp->dp_meta_objset,
+ csa->cds->ds_phys->ds_deadlist_obj));
+ VERIFY(0 == bplist_open(&csa->ohds->ds_deadlist, dp->dp_meta_objset,
+ csa->ohds->ds_phys->ds_deadlist_obj));
+
+ dsl_pool_ds_clone_swapped(csa->ohds, csa->cds, tx);
+}
+
+/*
+ * Swap 'clone' with its origin head file system. Used at the end
+ * of "online recv" to swizzle the file system to the new version.
+ */
+int
+dsl_dataset_clone_swap(dsl_dataset_t *clone, dsl_dataset_t *origin_head,
+ boolean_t force)
+{
+ struct cloneswaparg csa;
+ int error;
+
+ ASSERT(clone->ds_owner);
+ ASSERT(origin_head->ds_owner);
+retry:
+ /* Need exclusive access for the swap */
+ rw_enter(&clone->ds_rwlock, RW_WRITER);
+ if (!rw_tryenter(&origin_head->ds_rwlock, RW_WRITER)) {
+ rw_exit(&clone->ds_rwlock);
+ rw_enter(&origin_head->ds_rwlock, RW_WRITER);
+ if (!rw_tryenter(&clone->ds_rwlock, RW_WRITER)) {
+ rw_exit(&origin_head->ds_rwlock);
+ goto retry;
+ }
+ }
+ csa.cds = clone;
+ csa.ohds = origin_head;
+ csa.force = force;
+ error = dsl_sync_task_do(clone->ds_dir->dd_pool,
+ dsl_dataset_clone_swap_check,
+ dsl_dataset_clone_swap_sync, &csa, NULL, 9);
+ return (error);
+}
+
+/*
+ * Given a pool name and a dataset object number in that pool,
+ * return the name of that dataset.
+ */
+int
+dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf)
+{
+ spa_t *spa;
+ dsl_pool_t *dp;
+ dsl_dataset_t *ds;
+ int error;
+
+ if ((error = spa_open(pname, &spa, FTAG)) != 0)
+ return (error);
+ dp = spa_get_dsl(spa);
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ if ((error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds)) == 0) {
+ dsl_dataset_name(ds, buf);
+ dsl_dataset_rele(ds, FTAG);
+ }
+ rw_exit(&dp->dp_config_rwlock);
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+int
+dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
+ uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv)
+{
+ int error = 0;
+
+ ASSERT3S(asize, >, 0);
+
+ /*
+ * *ref_rsrv is the portion of asize that will come from any
+ * unconsumed refreservation space.
+ */
+ *ref_rsrv = 0;
+
+ mutex_enter(&ds->ds_lock);
+ /*
+ * Make a space adjustment for reserved bytes.
+ */
+ if (ds->ds_reserved > ds->ds_phys->ds_unique_bytes) {
+ ASSERT3U(*used, >=,
+ ds->ds_reserved - ds->ds_phys->ds_unique_bytes);
+ *used -= (ds->ds_reserved - ds->ds_phys->ds_unique_bytes);
+ *ref_rsrv =
+ asize - MIN(asize, parent_delta(ds, asize + inflight));
+ }
+
+ if (!check_quota || ds->ds_quota == 0) {
+ mutex_exit(&ds->ds_lock);
+ return (0);
+ }
+ /*
+ * If they are requesting more space, and our current estimate
+ * is over quota, they get to try again unless the actual
+ * on-disk is over quota and there are no pending changes (which
+ * may free up space for us).
+ */
+ if (ds->ds_phys->ds_used_bytes + inflight >= ds->ds_quota) {
+ if (inflight > 0 || ds->ds_phys->ds_used_bytes < ds->ds_quota)
+ error = ERESTART;
+ else
+ error = EDQUOT;
+ }
+ mutex_exit(&ds->ds_lock);
+
+ return (error);
+}
+
+/* ARGSUSED */
+static int
+dsl_dataset_set_quota_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ uint64_t *quotap = arg2;
+ uint64_t new_quota = *quotap;
+
+ if (spa_version(ds->ds_dir->dd_pool->dp_spa) < SPA_VERSION_REFQUOTA)
+ return (ENOTSUP);
+
+ if (new_quota == 0)
+ return (0);
+
+ if (new_quota < ds->ds_phys->ds_used_bytes ||
+ new_quota < ds->ds_reserved)
+ return (ENOSPC);
+
+ return (0);
+}
+
+/* ARGSUSED */
+void
+dsl_dataset_set_quota_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ uint64_t *quotap = arg2;
+ uint64_t new_quota = *quotap;
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+
+ ds->ds_quota = new_quota;
+
+ dsl_prop_set_uint64_sync(ds->ds_dir, "refquota", new_quota, cr, tx);
+
+ spa_history_internal_log(LOG_DS_REFQUOTA, ds->ds_dir->dd_pool->dp_spa,
+ tx, cr, "%lld dataset = %llu ",
+ (longlong_t)new_quota, ds->ds_object);
+}
+
+int
+dsl_dataset_set_quota(const char *dsname, uint64_t quota)
+{
+ dsl_dataset_t *ds;
+ int err;
+
+ err = dsl_dataset_hold(dsname, FTAG, &ds);
+ if (err)
+ return (err);
+
+ if (quota != ds->ds_quota) {
+ /*
+ * If someone removes a file, then tries to set the quota, we
+ * want to make sure the file freeing takes effect.
+ */
+ txg_wait_open(ds->ds_dir->dd_pool, 0);
+
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ dsl_dataset_set_quota_check, dsl_dataset_set_quota_sync,
+ ds, &quota, 0);
+ }
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+}
+
+static int
+dsl_dataset_set_reservation_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ uint64_t *reservationp = arg2;
+ uint64_t new_reservation = *reservationp;
+ int64_t delta;
+ uint64_t unique;
+
+ if (new_reservation > INT64_MAX)
+ return (EOVERFLOW);
+
+ if (spa_version(ds->ds_dir->dd_pool->dp_spa) <
+ SPA_VERSION_REFRESERVATION)
+ return (ENOTSUP);
+
+ if (dsl_dataset_is_snapshot(ds))
+ return (EINVAL);
+
+ /*
+ * If we are doing the preliminary check in open context, the
+ * space estimates may be inaccurate.
+ */
+ if (!dmu_tx_is_syncing(tx))
+ return (0);
+
+ mutex_enter(&ds->ds_lock);
+ unique = dsl_dataset_unique(ds);
+ delta = MAX(unique, new_reservation) - MAX(unique, ds->ds_reserved);
+ mutex_exit(&ds->ds_lock);
+
+ if (delta > 0 &&
+ delta > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE))
+ return (ENOSPC);
+ if (delta > 0 && ds->ds_quota > 0 &&
+ new_reservation > ds->ds_quota)
+ return (ENOSPC);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dsl_dataset_set_reservation_sync(void *arg1, void *arg2, cred_t *cr,
+ dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ uint64_t *reservationp = arg2;
+ uint64_t new_reservation = *reservationp;
+ uint64_t unique;
+ int64_t delta;
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+
+ mutex_enter(&ds->ds_dir->dd_lock);
+ mutex_enter(&ds->ds_lock);
+ unique = dsl_dataset_unique(ds);
+ delta = MAX(0, (int64_t)(new_reservation - unique)) -
+ MAX(0, (int64_t)(ds->ds_reserved - unique));
+ ds->ds_reserved = new_reservation;
+ mutex_exit(&ds->ds_lock);
+
+ dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx);
+ mutex_exit(&ds->ds_dir->dd_lock);
+ dsl_prop_set_uint64_sync(ds->ds_dir, "refreservation",
+ new_reservation, cr, tx);
+
+ spa_history_internal_log(LOG_DS_REFRESERV,
+ ds->ds_dir->dd_pool->dp_spa, tx, cr, "%lld dataset = %llu",
+ (longlong_t)new_reservation, ds->ds_object);
+}
+
+int
+dsl_dataset_set_reservation(const char *dsname, uint64_t reservation)
+{
+ dsl_dataset_t *ds;
+ int err;
+
+ err = dsl_dataset_hold(dsname, FTAG, &ds);
+ if (err)
+ return (err);
+
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ dsl_dataset_set_reservation_check,
+ dsl_dataset_set_reservation_sync, ds, &reservation, 0);
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+}
diff --git a/module/zfs/dsl_deleg.c b/module/zfs/dsl_deleg.c
new file mode 100644
index 000000000..da5d15787
--- /dev/null
+++ b/module/zfs/dsl_deleg.c
@@ -0,0 +1,735 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * DSL permissions are stored in a two level zap attribute
+ * mechanism. The first level identifies the "class" of
+ * entry. The class is identified by the first 2 letters of
+ * the attribute. The second letter "l" or "d" identifies whether
+ * it is a local or descendent permission. The first letter
+ * identifies the type of entry.
+ *
+ * ul$<id> identifies permissions granted locally for this userid.
+ * ud$<id> identifies permissions granted on descendent datasets for
+ * this userid.
+ * Ul$<id> identifies permission sets granted locally for this userid.
+ * Ud$<id> identifies permission sets granted on descendent datasets for
+ * this userid.
+ * gl$<id> identifies permissions granted locally for this groupid.
+ * gd$<id> identifies permissions granted on descendent datasets for
+ * this groupid.
+ * Gl$<id> identifies permission sets granted locally for this groupid.
+ * Gd$<id> identifies permission sets granted on descendent datasets for
+ * this groupid.
+ * el$ identifies permissions granted locally for everyone.
+ * ed$ identifies permissions granted on descendent datasets
+ * for everyone.
+ * El$ identifies permission sets granted locally for everyone.
+ * Ed$ identifies permission sets granted to descendent datasets for
+ * everyone.
+ * c-$ identifies permission to create at dataset creation time.
+ * C-$ identifies permission sets to grant locally at dataset creation
+ * time.
+ * s-$@<name> permissions defined in specified set @<name>
+ * S-$@<name> Sets defined in named set @<name>
+ *
+ * Each of the above entities points to another zap attribute that contains one
+ * attribute for each allowed permission, such as create, destroy,...
+ * All of the "upper" case class types will specify permission set names
+ * rather than permissions.
+ *
+ * Basically it looks something like this:
+ * ul$12 -> ZAP OBJ -> permissions...
+ *
+ * The ZAP OBJ is referred to as the jump object.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_tx.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dsl_deleg.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/zio_checksum.h> /* for the default checksum value */
+#include <sys/zap.h>
+#include <sys/fs/zfs.h>
+#include <sys/cred.h>
+#include <sys/sunddi.h>
+
+#include "zfs_deleg.h"
+
+/*
+ * Validate that user is allowed to delegate specified permissions.
+ *
+ * In order to delegate "create" you must have "create"
+ * and "allow".
+ */
+int
+dsl_deleg_can_allow(char *ddname, nvlist_t *nvp, cred_t *cr)
+{
+ nvpair_t *whopair = NULL;
+ int error;
+
+ if ((error = dsl_deleg_access(ddname, ZFS_DELEG_PERM_ALLOW, cr)) != 0)
+ return (error);
+
+ while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+ nvlist_t *perms;
+ nvpair_t *permpair = NULL;
+
+ VERIFY(nvpair_value_nvlist(whopair, &perms) == 0);
+
+ while (permpair = nvlist_next_nvpair(perms, permpair)) {
+ const char *perm = nvpair_name(permpair);
+
+ if (strcmp(perm, ZFS_DELEG_PERM_ALLOW) == 0)
+ return (EPERM);
+
+ if ((error = dsl_deleg_access(ddname, perm, cr)) != 0)
+ return (error);
+ }
+ }
+ return (0);
+}
+
+/*
+ * Validate that user is allowed to unallow specified permissions. They
+ * must have the 'allow' permission, and even then can only unallow
+ * perms for their uid.
+ */
+int
+dsl_deleg_can_unallow(char *ddname, nvlist_t *nvp, cred_t *cr)
+{
+ nvpair_t *whopair = NULL;
+ int error;
+ char idstr[32];
+
+ if ((error = dsl_deleg_access(ddname, ZFS_DELEG_PERM_ALLOW, cr)) != 0)
+ return (error);
+
+ (void) snprintf(idstr, sizeof (idstr), "%lld",
+ (longlong_t)crgetuid(cr));
+
+ while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+ zfs_deleg_who_type_t type = nvpair_name(whopair)[0];
+
+ if (type != ZFS_DELEG_USER &&
+ type != ZFS_DELEG_USER_SETS)
+ return (EPERM);
+
+ if (strcmp(idstr, &nvpair_name(whopair)[3]) != 0)
+ return (EPERM);
+ }
+ return (0);
+}
+
+static void
+dsl_deleg_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ nvlist_t *nvp = arg2;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ nvpair_t *whopair = NULL;
+ uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+ if (zapobj == 0) {
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+ zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
+ DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+ }
+
+ while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+ const char *whokey = nvpair_name(whopair);
+ nvlist_t *perms;
+ nvpair_t *permpair = NULL;
+ uint64_t jumpobj;
+
+ VERIFY(nvpair_value_nvlist(whopair, &perms) == 0);
+
+ if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0) {
+ jumpobj = zap_create(mos, DMU_OT_DSL_PERMS,
+ DMU_OT_NONE, 0, tx);
+ VERIFY(zap_update(mos, zapobj,
+ whokey, 8, 1, &jumpobj, tx) == 0);
+ }
+
+ while (permpair = nvlist_next_nvpair(perms, permpair)) {
+ const char *perm = nvpair_name(permpair);
+ uint64_t n = 0;
+
+ VERIFY(zap_update(mos, jumpobj,
+ perm, 8, 1, &n, tx) == 0);
+ spa_history_internal_log(LOG_DS_PERM_UPDATE,
+ dd->dd_pool->dp_spa, tx, cr,
+ "%s %s dataset = %llu", whokey, perm,
+ dd->dd_phys->dd_head_dataset_obj);
+ }
+ }
+}
+
+static void
+dsl_deleg_unset_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ nvlist_t *nvp = arg2;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ nvpair_t *whopair = NULL;
+ uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+ if (zapobj == 0)
+ return;
+
+ while (whopair = nvlist_next_nvpair(nvp, whopair)) {
+ const char *whokey = nvpair_name(whopair);
+ nvlist_t *perms;
+ nvpair_t *permpair = NULL;
+ uint64_t jumpobj;
+
+ if (nvpair_value_nvlist(whopair, &perms) != 0) {
+ if (zap_lookup(mos, zapobj, whokey, 8,
+ 1, &jumpobj) == 0) {
+ (void) zap_remove(mos, zapobj, whokey, tx);
+ VERIFY(0 == zap_destroy(mos, jumpobj, tx));
+ }
+ spa_history_internal_log(LOG_DS_PERM_WHO_REMOVE,
+ dd->dd_pool->dp_spa, tx, cr,
+ "%s dataset = %llu", whokey,
+ dd->dd_phys->dd_head_dataset_obj);
+ continue;
+ }
+
+ if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0)
+ continue;
+
+ while (permpair = nvlist_next_nvpair(perms, permpair)) {
+ const char *perm = nvpair_name(permpair);
+ uint64_t n = 0;
+
+ (void) zap_remove(mos, jumpobj, perm, tx);
+ if (zap_count(mos, jumpobj, &n) == 0 && n == 0) {
+ (void) zap_remove(mos, zapobj,
+ whokey, tx);
+ VERIFY(0 == zap_destroy(mos,
+ jumpobj, tx));
+ }
+ spa_history_internal_log(LOG_DS_PERM_REMOVE,
+ dd->dd_pool->dp_spa, tx, cr,
+ "%s %s dataset = %llu", whokey, perm,
+ dd->dd_phys->dd_head_dataset_obj);
+ }
+ }
+}
+
+int
+dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset)
+{
+ dsl_dir_t *dd;
+ int error;
+ nvpair_t *whopair = NULL;
+ int blocks_modified = 0;
+
+ error = dsl_dir_open(ddname, FTAG, &dd, NULL);
+ if (error)
+ return (error);
+
+ if (spa_version(dmu_objset_spa(dd->dd_pool->dp_meta_objset)) <
+ SPA_VERSION_DELEGATED_PERMS) {
+ dsl_dir_close(dd, FTAG);
+ return (ENOTSUP);
+ }
+
+ while (whopair = nvlist_next_nvpair(nvp, whopair))
+ blocks_modified++;
+
+ error = dsl_sync_task_do(dd->dd_pool, NULL,
+ unset ? dsl_deleg_unset_sync : dsl_deleg_set_sync,
+ dd, nvp, blocks_modified);
+ dsl_dir_close(dd, FTAG);
+
+ return (error);
+}
+
+/*
+ * Find all 'allow' permissions from a given point and then continue
+ * traversing up to the root.
+ *
+ * This function constructs an nvlist of nvlists.
+ * each setpoint is an nvlist composed of an nvlist of an nvlist
+ * of the individual * users/groups/everyone/create
+ * permissions.
+ *
+ * The nvlist will look like this.
+ *
+ * { source fsname -> { whokeys { permissions,...}, ...}}
+ *
+ * The fsname nvpairs will be arranged in a bottom up order. For example,
+ * if we have the following structure a/b/c then the nvpairs for the fsnames
+ * will be ordered a/b/c, a/b, a.
+ */
+int
+dsl_deleg_get(const char *ddname, nvlist_t **nvp)
+{
+ dsl_dir_t *dd, *startdd;
+ dsl_pool_t *dp;
+ int error;
+ objset_t *mos;
+
+ error = dsl_dir_open(ddname, FTAG, &startdd, NULL);
+ if (error)
+ return (error);
+
+ dp = startdd->dd_pool;
+ mos = dp->dp_meta_objset;
+
+ VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ for (dd = startdd; dd != NULL; dd = dd->dd_parent) {
+ zap_cursor_t basezc;
+ zap_attribute_t baseza;
+ nvlist_t *sp_nvp;
+ uint64_t n;
+ char source[MAXNAMELEN];
+
+ if (dd->dd_phys->dd_deleg_zapobj &&
+ (zap_count(mos, dd->dd_phys->dd_deleg_zapobj,
+ &n) == 0) && n) {
+ VERIFY(nvlist_alloc(&sp_nvp,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ } else {
+ continue;
+ }
+
+ for (zap_cursor_init(&basezc, mos,
+ dd->dd_phys->dd_deleg_zapobj);
+ zap_cursor_retrieve(&basezc, &baseza) == 0;
+ zap_cursor_advance(&basezc)) {
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ nvlist_t *perms_nvp;
+
+ ASSERT(baseza.za_integer_length == 8);
+ ASSERT(baseza.za_num_integers == 1);
+
+ VERIFY(nvlist_alloc(&perms_nvp,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ for (zap_cursor_init(&zc, mos, baseza.za_first_integer);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ VERIFY(nvlist_add_boolean(perms_nvp,
+ za.za_name) == 0);
+ }
+ zap_cursor_fini(&zc);
+ VERIFY(nvlist_add_nvlist(sp_nvp, baseza.za_name,
+ perms_nvp) == 0);
+ nvlist_free(perms_nvp);
+ }
+
+ zap_cursor_fini(&basezc);
+
+ dsl_dir_name(dd, source);
+ VERIFY(nvlist_add_nvlist(*nvp, source, sp_nvp) == 0);
+ nvlist_free(sp_nvp);
+ }
+ rw_exit(&dp->dp_config_rwlock);
+
+ dsl_dir_close(startdd, FTAG);
+ return (0);
+}
+
+/*
+ * Routines for dsl_deleg_access() -- access checking.
+ */
+typedef struct perm_set {
+ avl_node_t p_node;
+ boolean_t p_matched;
+ char p_setname[ZFS_MAX_DELEG_NAME];
+} perm_set_t;
+
+static int
+perm_set_compare(const void *arg1, const void *arg2)
+{
+ const perm_set_t *node1 = arg1;
+ const perm_set_t *node2 = arg2;
+ int val;
+
+ val = strcmp(node1->p_setname, node2->p_setname);
+ if (val == 0)
+ return (0);
+ return (val > 0 ? 1 : -1);
+}
+
+/*
+ * Determine whether a specified permission exists.
+ *
+ * First the base attribute has to be retrieved. i.e. ul$12
+ * Once the base object has been retrieved the actual permission
+ * is lookup up in the zap object the base object points to.
+ *
+ * Return 0 if permission exists, ENOENT if there is no whokey, EPERM if
+ * there is no perm in that jumpobj.
+ */
+static int
+dsl_check_access(objset_t *mos, uint64_t zapobj,
+ char type, char checkflag, void *valp, const char *perm)
+{
+ int error;
+ uint64_t jumpobj, zero;
+ char whokey[ZFS_MAX_DELEG_NAME];
+
+ zfs_deleg_whokey(whokey, type, checkflag, valp);
+ error = zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj);
+ if (error == 0) {
+ error = zap_lookup(mos, jumpobj, perm, 8, 1, &zero);
+ if (error == ENOENT)
+ error = EPERM;
+ }
+ return (error);
+}
+
+/*
+ * check a specified user/group for a requested permission
+ */
+static int
+dsl_check_user_access(objset_t *mos, uint64_t zapobj, const char *perm,
+ int checkflag, cred_t *cr)
+{
+ const gid_t *gids;
+ int ngids;
+ int i;
+ uint64_t id;
+
+ /* check for user */
+ id = crgetuid(cr);
+ if (dsl_check_access(mos, zapobj,
+ ZFS_DELEG_USER, checkflag, &id, perm) == 0)
+ return (0);
+
+ /* check for users primary group */
+ id = crgetgid(cr);
+ if (dsl_check_access(mos, zapobj,
+ ZFS_DELEG_GROUP, checkflag, &id, perm) == 0)
+ return (0);
+
+ /* check for everyone entry */
+ id = -1;
+ if (dsl_check_access(mos, zapobj,
+ ZFS_DELEG_EVERYONE, checkflag, &id, perm) == 0)
+ return (0);
+
+ /* check each supplemental group user is a member of */
+ ngids = crgetngroups(cr);
+ gids = crgetgroups(cr);
+ for (i = 0; i != ngids; i++) {
+ id = gids[i];
+ if (dsl_check_access(mos, zapobj,
+ ZFS_DELEG_GROUP, checkflag, &id, perm) == 0)
+ return (0);
+ }
+
+ return (EPERM);
+}
+
+/*
+ * Iterate over the sets specified in the specified zapobj
+ * and load them into the permsets avl tree.
+ */
+static int
+dsl_load_sets(objset_t *mos, uint64_t zapobj,
+ char type, char checkflag, void *valp, avl_tree_t *avl)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ perm_set_t *permnode;
+ avl_index_t idx;
+ uint64_t jumpobj;
+ int error;
+ char whokey[ZFS_MAX_DELEG_NAME];
+
+ zfs_deleg_whokey(whokey, type, checkflag, valp);
+
+ error = zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj);
+ if (error != 0)
+ return (error);
+
+ for (zap_cursor_init(&zc, mos, jumpobj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ permnode = kmem_alloc(sizeof (perm_set_t), KM_SLEEP);
+ (void) strlcpy(permnode->p_setname, za.za_name,
+ sizeof (permnode->p_setname));
+ permnode->p_matched = B_FALSE;
+
+ if (avl_find(avl, permnode, &idx) == NULL) {
+ avl_insert(avl, permnode, idx);
+ } else {
+ kmem_free(permnode, sizeof (perm_set_t));
+ }
+ }
+ zap_cursor_fini(&zc);
+ return (0);
+}
+
+/*
+ * Load all permissions user based on cred belongs to.
+ */
+static void
+dsl_load_user_sets(objset_t *mos, uint64_t zapobj, avl_tree_t *avl,
+ char checkflag, cred_t *cr)
+{
+ const gid_t *gids;
+ int ngids, i;
+ uint64_t id;
+
+ id = crgetuid(cr);
+ (void) dsl_load_sets(mos, zapobj,
+ ZFS_DELEG_USER_SETS, checkflag, &id, avl);
+
+ id = crgetgid(cr);
+ (void) dsl_load_sets(mos, zapobj,
+ ZFS_DELEG_GROUP_SETS, checkflag, &id, avl);
+
+ (void) dsl_load_sets(mos, zapobj,
+ ZFS_DELEG_EVERYONE_SETS, checkflag, NULL, avl);
+
+ ngids = crgetngroups(cr);
+ gids = crgetgroups(cr);
+ for (i = 0; i != ngids; i++) {
+ id = gids[i];
+ (void) dsl_load_sets(mos, zapobj,
+ ZFS_DELEG_GROUP_SETS, checkflag, &id, avl);
+ }
+}
+
+/*
+ * Check if user has requested permission.
+ */
+int
+dsl_deleg_access(const char *dsname, const char *perm, cred_t *cr)
+{
+ dsl_dataset_t *ds;
+ dsl_dir_t *dd;
+ dsl_pool_t *dp;
+ void *cookie;
+ int error;
+ char checkflag = ZFS_DELEG_LOCAL;
+ objset_t *mos;
+ avl_tree_t permsets;
+ perm_set_t *setnode;
+
+ error = dsl_dataset_hold(dsname, FTAG, &ds);
+ if (error)
+ return (error);
+
+ dp = ds->ds_dir->dd_pool;
+ mos = dp->dp_meta_objset;
+
+ if (dsl_delegation_on(mos) == B_FALSE) {
+ dsl_dataset_rele(ds, FTAG);
+ return (ECANCELED);
+ }
+
+ if (spa_version(dmu_objset_spa(dp->dp_meta_objset)) <
+ SPA_VERSION_DELEGATED_PERMS) {
+ dsl_dataset_rele(ds, FTAG);
+ return (EPERM);
+ }
+
+ avl_create(&permsets, perm_set_compare, sizeof (perm_set_t),
+ offsetof(perm_set_t, p_node));
+
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ for (dd = ds->ds_dir; dd != NULL; dd = dd->dd_parent,
+ checkflag = ZFS_DELEG_DESCENDENT) {
+ uint64_t zapobj;
+ boolean_t expanded;
+
+ /*
+ * If not in global zone then make sure
+ * the zoned property is set
+ */
+ if (!INGLOBALZONE(curproc)) {
+ uint64_t zoned;
+
+ if (dsl_prop_get_dd(dd,
+ zfs_prop_to_name(ZFS_PROP_ZONED),
+ 8, 1, &zoned, NULL) != 0)
+ break;
+ if (!zoned)
+ break;
+ }
+ zapobj = dd->dd_phys->dd_deleg_zapobj;
+
+ if (zapobj == 0)
+ continue;
+
+ dsl_load_user_sets(mos, zapobj, &permsets, checkflag, cr);
+again:
+ expanded = B_FALSE;
+ for (setnode = avl_first(&permsets); setnode;
+ setnode = AVL_NEXT(&permsets, setnode)) {
+ if (setnode->p_matched == B_TRUE)
+ continue;
+
+ /* See if this set directly grants this permission */
+ error = dsl_check_access(mos, zapobj,
+ ZFS_DELEG_NAMED_SET, 0, setnode->p_setname, perm);
+ if (error == 0)
+ goto success;
+ if (error == EPERM)
+ setnode->p_matched = B_TRUE;
+
+ /* See if this set includes other sets */
+ error = dsl_load_sets(mos, zapobj,
+ ZFS_DELEG_NAMED_SET_SETS, 0,
+ setnode->p_setname, &permsets);
+ if (error == 0)
+ setnode->p_matched = expanded = B_TRUE;
+ }
+ /*
+ * If we expanded any sets, that will define more sets,
+ * which we need to check.
+ */
+ if (expanded)
+ goto again;
+
+ error = dsl_check_user_access(mos, zapobj, perm, checkflag, cr);
+ if (error == 0)
+ goto success;
+ }
+ error = EPERM;
+success:
+ rw_exit(&dp->dp_config_rwlock);
+ dsl_dataset_rele(ds, FTAG);
+
+ cookie = NULL;
+ while ((setnode = avl_destroy_nodes(&permsets, &cookie)) != NULL)
+ kmem_free(setnode, sizeof (perm_set_t));
+
+ return (error);
+}
+
+/*
+ * Other routines.
+ */
+
+static void
+copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj,
+ boolean_t dosets, uint64_t uid, dmu_tx_t *tx)
+{
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ uint64_t jumpobj, pjumpobj;
+ uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ char whokey[ZFS_MAX_DELEG_NAME];
+
+ zfs_deleg_whokey(whokey,
+ dosets ? ZFS_DELEG_CREATE_SETS : ZFS_DELEG_CREATE,
+ ZFS_DELEG_LOCAL, NULL);
+ if (zap_lookup(mos, pzapobj, whokey, 8, 1, &pjumpobj) != 0)
+ return;
+
+ if (zapobj == 0) {
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+ zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
+ DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+ }
+
+ zfs_deleg_whokey(whokey,
+ dosets ? ZFS_DELEG_USER_SETS : ZFS_DELEG_USER,
+ ZFS_DELEG_LOCAL, &uid);
+ if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == ENOENT) {
+ jumpobj = zap_create(mos, DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
+ VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0);
+ }
+
+ for (zap_cursor_init(&zc, mos, pjumpobj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ uint64_t zero = 0;
+ ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
+
+ VERIFY(zap_update(mos, jumpobj, za.za_name,
+ 8, 1, &zero, tx) == 0);
+ }
+ zap_cursor_fini(&zc);
+}
+
+/*
+ * set all create time permission on new dataset.
+ */
+void
+dsl_deleg_set_create_perms(dsl_dir_t *sdd, dmu_tx_t *tx, cred_t *cr)
+{
+ dsl_dir_t *dd;
+ uint64_t uid = crgetuid(cr);
+
+ if (spa_version(dmu_objset_spa(sdd->dd_pool->dp_meta_objset)) <
+ SPA_VERSION_DELEGATED_PERMS)
+ return;
+
+ for (dd = sdd->dd_parent; dd != NULL; dd = dd->dd_parent) {
+ uint64_t pzapobj = dd->dd_phys->dd_deleg_zapobj;
+
+ if (pzapobj == 0)
+ continue;
+
+ copy_create_perms(sdd, pzapobj, B_FALSE, uid, tx);
+ copy_create_perms(sdd, pzapobj, B_TRUE, uid, tx);
+ }
+}
+
+int
+dsl_deleg_destroy(objset_t *mos, uint64_t zapobj, dmu_tx_t *tx)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+
+ if (zapobj == 0)
+ return (0);
+
+ for (zap_cursor_init(&zc, mos, zapobj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
+ VERIFY(0 == zap_destroy(mos, za.za_first_integer, tx));
+ }
+ zap_cursor_fini(&zc);
+ VERIFY(0 == zap_destroy(mos, zapobj, tx));
+ return (0);
+}
+
+boolean_t
+dsl_delegation_on(objset_t *os)
+{
+ return (os->os->os_spa->spa_delegation);
+}
diff --git a/module/zfs/dsl_dir.c b/module/zfs/dsl_dir.c
new file mode 100644
index 000000000..48d87f97f
--- /dev/null
+++ b/module/zfs/dsl_dir.c
@@ -0,0 +1,1331 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_tx.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dsl_deleg.h>
+#include <sys/spa.h>
+#include <sys/zap.h>
+#include <sys/zio.h>
+#include <sys/arc.h>
+#include <sys/sunddi.h>
+#include "zfs_namecheck.h"
+
+static uint64_t dsl_dir_space_towrite(dsl_dir_t *dd);
+static void dsl_dir_set_reservation_sync(void *arg1, void *arg2,
+ cred_t *cr, dmu_tx_t *tx);
+
+
+/* ARGSUSED */
+static void
+dsl_dir_evict(dmu_buf_t *db, void *arg)
+{
+ dsl_dir_t *dd = arg;
+ dsl_pool_t *dp = dd->dd_pool;
+ int t;
+
+ for (t = 0; t < TXG_SIZE; t++) {
+ ASSERT(!txg_list_member(&dp->dp_dirty_dirs, dd, t));
+ ASSERT(dd->dd_tempreserved[t] == 0);
+ ASSERT(dd->dd_space_towrite[t] == 0);
+ }
+
+ if (dd->dd_parent)
+ dsl_dir_close(dd->dd_parent, dd);
+
+ spa_close(dd->dd_pool->dp_spa, dd);
+
+ /*
+ * The props callback list should be empty since they hold the
+ * dir open.
+ */
+ list_destroy(&dd->dd_prop_cbs);
+ mutex_destroy(&dd->dd_lock);
+ kmem_free(dd, sizeof (dsl_dir_t));
+}
+
+int
+dsl_dir_open_obj(dsl_pool_t *dp, uint64_t ddobj,
+ const char *tail, void *tag, dsl_dir_t **ddp)
+{
+ dmu_buf_t *dbuf;
+ dsl_dir_t *dd;
+ int err;
+
+ ASSERT(RW_LOCK_HELD(&dp->dp_config_rwlock) ||
+ dsl_pool_sync_context(dp));
+
+ err = dmu_bonus_hold(dp->dp_meta_objset, ddobj, tag, &dbuf);
+ if (err)
+ return (err);
+ dd = dmu_buf_get_user(dbuf);
+#ifdef ZFS_DEBUG
+ {
+ dmu_object_info_t doi;
+ dmu_object_info_from_db(dbuf, &doi);
+ ASSERT3U(doi.doi_type, ==, DMU_OT_DSL_DIR);
+ ASSERT3U(doi.doi_bonus_size, >=, sizeof (dsl_dir_phys_t));
+ }
+#endif
+ if (dd == NULL) {
+ dsl_dir_t *winner;
+ int err;
+
+ dd = kmem_zalloc(sizeof (dsl_dir_t), KM_SLEEP);
+ dd->dd_object = ddobj;
+ dd->dd_dbuf = dbuf;
+ dd->dd_pool = dp;
+ dd->dd_phys = dbuf->db_data;
+ mutex_init(&dd->dd_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ list_create(&dd->dd_prop_cbs, sizeof (dsl_prop_cb_record_t),
+ offsetof(dsl_prop_cb_record_t, cbr_node));
+
+ if (dd->dd_phys->dd_parent_obj) {
+ err = dsl_dir_open_obj(dp, dd->dd_phys->dd_parent_obj,
+ NULL, dd, &dd->dd_parent);
+ if (err)
+ goto errout;
+ if (tail) {
+#ifdef ZFS_DEBUG
+ uint64_t foundobj;
+
+ err = zap_lookup(dp->dp_meta_objset,
+ dd->dd_parent->dd_phys->dd_child_dir_zapobj,
+ tail, sizeof (foundobj), 1, &foundobj);
+ ASSERT(err || foundobj == ddobj);
+#endif
+ (void) strcpy(dd->dd_myname, tail);
+ } else {
+ err = zap_value_search(dp->dp_meta_objset,
+ dd->dd_parent->dd_phys->dd_child_dir_zapobj,
+ ddobj, 0, dd->dd_myname);
+ }
+ if (err)
+ goto errout;
+ } else {
+ (void) strcpy(dd->dd_myname, spa_name(dp->dp_spa));
+ }
+
+ winner = dmu_buf_set_user_ie(dbuf, dd, &dd->dd_phys,
+ dsl_dir_evict);
+ if (winner) {
+ if (dd->dd_parent)
+ dsl_dir_close(dd->dd_parent, dd);
+ mutex_destroy(&dd->dd_lock);
+ kmem_free(dd, sizeof (dsl_dir_t));
+ dd = winner;
+ } else {
+ spa_open_ref(dp->dp_spa, dd);
+ }
+ }
+
+ /*
+ * The dsl_dir_t has both open-to-close and instantiate-to-evict
+ * holds on the spa. We need the open-to-close holds because
+ * otherwise the spa_refcnt wouldn't change when we open a
+ * dir which the spa also has open, so we could incorrectly
+ * think it was OK to unload/export/destroy the pool. We need
+ * the instantiate-to-evict hold because the dsl_dir_t has a
+ * pointer to the dd_pool, which has a pointer to the spa_t.
+ */
+ spa_open_ref(dp->dp_spa, tag);
+ ASSERT3P(dd->dd_pool, ==, dp);
+ ASSERT3U(dd->dd_object, ==, ddobj);
+ ASSERT3P(dd->dd_dbuf, ==, dbuf);
+ *ddp = dd;
+ return (0);
+
+errout:
+ if (dd->dd_parent)
+ dsl_dir_close(dd->dd_parent, dd);
+ mutex_destroy(&dd->dd_lock);
+ kmem_free(dd, sizeof (dsl_dir_t));
+ dmu_buf_rele(dbuf, tag);
+ return (err);
+
+}
+
+void
+dsl_dir_close(dsl_dir_t *dd, void *tag)
+{
+ dprintf_dd(dd, "%s\n", "");
+ spa_close(dd->dd_pool->dp_spa, tag);
+ dmu_buf_rele(dd->dd_dbuf, tag);
+}
+
+/* buf must be long enough (MAXNAMELEN + strlen(MOS_DIR_NAME) + 1 should do) */
+void
+dsl_dir_name(dsl_dir_t *dd, char *buf)
+{
+ if (dd->dd_parent) {
+ dsl_dir_name(dd->dd_parent, buf);
+ (void) strcat(buf, "/");
+ } else {
+ buf[0] = '\0';
+ }
+ if (!MUTEX_HELD(&dd->dd_lock)) {
+ /*
+ * recursive mutex so that we can use
+ * dprintf_dd() with dd_lock held
+ */
+ mutex_enter(&dd->dd_lock);
+ (void) strcat(buf, dd->dd_myname);
+ mutex_exit(&dd->dd_lock);
+ } else {
+ (void) strcat(buf, dd->dd_myname);
+ }
+}
+
+/* Calculate name legnth, avoiding all the strcat calls of dsl_dir_name */
+int
+dsl_dir_namelen(dsl_dir_t *dd)
+{
+ int result = 0;
+
+ if (dd->dd_parent) {
+ /* parent's name + 1 for the "/" */
+ result = dsl_dir_namelen(dd->dd_parent) + 1;
+ }
+
+ if (!MUTEX_HELD(&dd->dd_lock)) {
+ /* see dsl_dir_name */
+ mutex_enter(&dd->dd_lock);
+ result += strlen(dd->dd_myname);
+ mutex_exit(&dd->dd_lock);
+ } else {
+ result += strlen(dd->dd_myname);
+ }
+
+ return (result);
+}
+
+int
+dsl_dir_is_private(dsl_dir_t *dd)
+{
+ int rv = FALSE;
+
+ if (dd->dd_parent && dsl_dir_is_private(dd->dd_parent))
+ rv = TRUE;
+ if (dataset_name_hidden(dd->dd_myname))
+ rv = TRUE;
+ return (rv);
+}
+
+
+static int
+getcomponent(const char *path, char *component, const char **nextp)
+{
+ char *p;
+ if (path == NULL)
+ return (ENOENT);
+ /* This would be a good place to reserve some namespace... */
+ p = strpbrk(path, "/@");
+ if (p && (p[1] == '/' || p[1] == '@')) {
+ /* two separators in a row */
+ return (EINVAL);
+ }
+ if (p == NULL || p == path) {
+ /*
+ * if the first thing is an @ or /, it had better be an
+ * @ and it had better not have any more ats or slashes,
+ * and it had better have something after the @.
+ */
+ if (p != NULL &&
+ (p[0] != '@' || strpbrk(path+1, "/@") || p[1] == '\0'))
+ return (EINVAL);
+ if (strlen(path) >= MAXNAMELEN)
+ return (ENAMETOOLONG);
+ (void) strcpy(component, path);
+ p = NULL;
+ } else if (p[0] == '/') {
+ if (p-path >= MAXNAMELEN)
+ return (ENAMETOOLONG);
+ (void) strncpy(component, path, p - path);
+ component[p-path] = '\0';
+ p++;
+ } else if (p[0] == '@') {
+ /*
+ * if the next separator is an @, there better not be
+ * any more slashes.
+ */
+ if (strchr(path, '/'))
+ return (EINVAL);
+ if (p-path >= MAXNAMELEN)
+ return (ENAMETOOLONG);
+ (void) strncpy(component, path, p - path);
+ component[p-path] = '\0';
+ } else {
+ ASSERT(!"invalid p");
+ }
+ *nextp = p;
+ return (0);
+}
+
+/*
+ * same as dsl_open_dir, ignore the first component of name and use the
+ * spa instead
+ */
+int
+dsl_dir_open_spa(spa_t *spa, const char *name, void *tag,
+ dsl_dir_t **ddp, const char **tailp)
+{
+ char buf[MAXNAMELEN];
+ const char *next, *nextnext = NULL;
+ int err;
+ dsl_dir_t *dd;
+ dsl_pool_t *dp;
+ uint64_t ddobj;
+ int openedspa = FALSE;
+
+ dprintf("%s\n", name);
+
+ err = getcomponent(name, buf, &next);
+ if (err)
+ return (err);
+ if (spa == NULL) {
+ err = spa_open(buf, &spa, FTAG);
+ if (err) {
+ dprintf("spa_open(%s) failed\n", buf);
+ return (err);
+ }
+ openedspa = TRUE;
+
+ /* XXX this assertion belongs in spa_open */
+ ASSERT(!dsl_pool_sync_context(spa_get_dsl(spa)));
+ }
+
+ dp = spa_get_dsl(spa);
+
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ err = dsl_dir_open_obj(dp, dp->dp_root_dir_obj, NULL, tag, &dd);
+ if (err) {
+ rw_exit(&dp->dp_config_rwlock);
+ if (openedspa)
+ spa_close(spa, FTAG);
+ return (err);
+ }
+
+ while (next != NULL) {
+ dsl_dir_t *child_ds;
+ err = getcomponent(next, buf, &nextnext);
+ if (err)
+ break;
+ ASSERT(next[0] != '\0');
+ if (next[0] == '@')
+ break;
+ dprintf("looking up %s in obj%lld\n",
+ buf, dd->dd_phys->dd_child_dir_zapobj);
+
+ err = zap_lookup(dp->dp_meta_objset,
+ dd->dd_phys->dd_child_dir_zapobj,
+ buf, sizeof (ddobj), 1, &ddobj);
+ if (err) {
+ if (err == ENOENT)
+ err = 0;
+ break;
+ }
+
+ err = dsl_dir_open_obj(dp, ddobj, buf, tag, &child_ds);
+ if (err)
+ break;
+ dsl_dir_close(dd, tag);
+ dd = child_ds;
+ next = nextnext;
+ }
+ rw_exit(&dp->dp_config_rwlock);
+
+ if (err) {
+ dsl_dir_close(dd, tag);
+ if (openedspa)
+ spa_close(spa, FTAG);
+ return (err);
+ }
+
+ /*
+ * It's an error if there's more than one component left, or
+ * tailp==NULL and there's any component left.
+ */
+ if (next != NULL &&
+ (tailp == NULL || (nextnext && nextnext[0] != '\0'))) {
+ /* bad path name */
+ dsl_dir_close(dd, tag);
+ dprintf("next=%p (%s) tail=%p\n", next, next?next:"", tailp);
+ err = ENOENT;
+ }
+ if (tailp)
+ *tailp = next;
+ if (openedspa)
+ spa_close(spa, FTAG);
+ *ddp = dd;
+ return (err);
+}
+
+/*
+ * Return the dsl_dir_t, and possibly the last component which couldn't
+ * be found in *tail. Return NULL if the path is bogus, or if
+ * tail==NULL and we couldn't parse the whole name. (*tail)[0] == '@'
+ * means that the last component is a snapshot.
+ */
+int
+dsl_dir_open(const char *name, void *tag, dsl_dir_t **ddp, const char **tailp)
+{
+ return (dsl_dir_open_spa(NULL, name, tag, ddp, tailp));
+}
+
+uint64_t
+dsl_dir_create_sync(dsl_pool_t *dp, dsl_dir_t *pds, const char *name,
+ dmu_tx_t *tx)
+{
+ objset_t *mos = dp->dp_meta_objset;
+ uint64_t ddobj;
+ dsl_dir_phys_t *dsphys;
+ dmu_buf_t *dbuf;
+
+ ddobj = dmu_object_alloc(mos, DMU_OT_DSL_DIR, 0,
+ DMU_OT_DSL_DIR, sizeof (dsl_dir_phys_t), tx);
+ if (pds) {
+ VERIFY(0 == zap_add(mos, pds->dd_phys->dd_child_dir_zapobj,
+ name, sizeof (uint64_t), 1, &ddobj, tx));
+ } else {
+ /* it's the root dir */
+ VERIFY(0 == zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1, &ddobj, tx));
+ }
+ VERIFY(0 == dmu_bonus_hold(mos, ddobj, FTAG, &dbuf));
+ dmu_buf_will_dirty(dbuf, tx);
+ dsphys = dbuf->db_data;
+
+ dsphys->dd_creation_time = gethrestime_sec();
+ if (pds)
+ dsphys->dd_parent_obj = pds->dd_object;
+ dsphys->dd_props_zapobj = zap_create(mos,
+ DMU_OT_DSL_PROPS, DMU_OT_NONE, 0, tx);
+ dsphys->dd_child_dir_zapobj = zap_create(mos,
+ DMU_OT_DSL_DIR_CHILD_MAP, DMU_OT_NONE, 0, tx);
+ if (spa_version(dp->dp_spa) >= SPA_VERSION_USED_BREAKDOWN)
+ dsphys->dd_flags |= DD_FLAG_USED_BREAKDOWN;
+ dmu_buf_rele(dbuf, FTAG);
+
+ return (ddobj);
+}
+
+/* ARGSUSED */
+int
+dsl_dir_destroy_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ dsl_pool_t *dp = dd->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+ int err;
+ uint64_t count;
+
+ /*
+ * There should be exactly two holds, both from
+ * dsl_dataset_destroy: one on the dd directory, and one on its
+ * head ds. Otherwise, someone is trying to lookup something
+ * inside this dir while we want to destroy it. The
+ * config_rwlock ensures that nobody else opens it after we
+ * check.
+ */
+ if (dmu_buf_refcount(dd->dd_dbuf) > 2)
+ return (EBUSY);
+
+ err = zap_count(mos, dd->dd_phys->dd_child_dir_zapobj, &count);
+ if (err)
+ return (err);
+ if (count != 0)
+ return (EEXIST);
+
+ return (0);
+}
+
+void
+dsl_dir_destroy_sync(void *arg1, void *tag, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ uint64_t val, obj;
+ dd_used_t t;
+
+ ASSERT(RW_WRITE_HELD(&dd->dd_pool->dp_config_rwlock));
+ ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
+
+ /* Remove our reservation. */
+ val = 0;
+ dsl_dir_set_reservation_sync(dd, &val, cr, tx);
+ ASSERT3U(dd->dd_phys->dd_used_bytes, ==, 0);
+ ASSERT3U(dd->dd_phys->dd_reserved, ==, 0);
+ for (t = 0; t < DD_USED_NUM; t++)
+ ASSERT3U(dd->dd_phys->dd_used_breakdown[t], ==, 0);
+
+ VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_child_dir_zapobj, tx));
+ VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_props_zapobj, tx));
+ VERIFY(0 == dsl_deleg_destroy(mos, dd->dd_phys->dd_deleg_zapobj, tx));
+ VERIFY(0 == zap_remove(mos,
+ dd->dd_parent->dd_phys->dd_child_dir_zapobj, dd->dd_myname, tx));
+
+ obj = dd->dd_object;
+ dsl_dir_close(dd, tag);
+ VERIFY(0 == dmu_object_free(mos, obj, tx));
+}
+
+boolean_t
+dsl_dir_is_clone(dsl_dir_t *dd)
+{
+ return (dd->dd_phys->dd_origin_obj &&
+ (dd->dd_pool->dp_origin_snap == NULL ||
+ dd->dd_phys->dd_origin_obj !=
+ dd->dd_pool->dp_origin_snap->ds_object));
+}
+
+void
+dsl_dir_stats(dsl_dir_t *dd, nvlist_t *nv)
+{
+ mutex_enter(&dd->dd_lock);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED,
+ dd->dd_phys->dd_used_bytes);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_QUOTA, dd->dd_phys->dd_quota);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_RESERVATION,
+ dd->dd_phys->dd_reserved);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO,
+ dd->dd_phys->dd_compressed_bytes == 0 ? 100 :
+ (dd->dd_phys->dd_uncompressed_bytes * 100 /
+ dd->dd_phys->dd_compressed_bytes));
+ if (dd->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDSNAP,
+ dd->dd_phys->dd_used_breakdown[DD_USED_SNAP]);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDDS,
+ dd->dd_phys->dd_used_breakdown[DD_USED_HEAD]);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDREFRESERV,
+ dd->dd_phys->dd_used_breakdown[DD_USED_REFRSRV]);
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDCHILD,
+ dd->dd_phys->dd_used_breakdown[DD_USED_CHILD] +
+ dd->dd_phys->dd_used_breakdown[DD_USED_CHILD_RSRV]);
+ }
+ mutex_exit(&dd->dd_lock);
+
+ rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
+ if (dsl_dir_is_clone(dd)) {
+ dsl_dataset_t *ds;
+ char buf[MAXNAMELEN];
+
+ VERIFY(0 == dsl_dataset_hold_obj(dd->dd_pool,
+ dd->dd_phys->dd_origin_obj, FTAG, &ds));
+ dsl_dataset_name(ds, buf);
+ dsl_dataset_rele(ds, FTAG);
+ dsl_prop_nvlist_add_string(nv, ZFS_PROP_ORIGIN, buf);
+ }
+ rw_exit(&dd->dd_pool->dp_config_rwlock);
+}
+
+void
+dsl_dir_dirty(dsl_dir_t *dd, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = dd->dd_pool;
+
+ ASSERT(dd->dd_phys);
+
+ if (txg_list_add(&dp->dp_dirty_dirs, dd, tx->tx_txg) == 0) {
+ /* up the hold count until we can be written out */
+ dmu_buf_add_ref(dd->dd_dbuf, dd);
+ }
+}
+
+static int64_t
+parent_delta(dsl_dir_t *dd, uint64_t used, int64_t delta)
+{
+ uint64_t old_accounted = MAX(used, dd->dd_phys->dd_reserved);
+ uint64_t new_accounted = MAX(used + delta, dd->dd_phys->dd_reserved);
+ return (new_accounted - old_accounted);
+}
+
+void
+dsl_dir_sync(dsl_dir_t *dd, dmu_tx_t *tx)
+{
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+
+ mutex_enter(&dd->dd_lock);
+ ASSERT3U(dd->dd_tempreserved[tx->tx_txg&TXG_MASK], ==, 0);
+ dprintf_dd(dd, "txg=%llu towrite=%lluK\n", tx->tx_txg,
+ dd->dd_space_towrite[tx->tx_txg&TXG_MASK] / 1024);
+ dd->dd_space_towrite[tx->tx_txg&TXG_MASK] = 0;
+ mutex_exit(&dd->dd_lock);
+
+ /* release the hold from dsl_dir_dirty */
+ dmu_buf_rele(dd->dd_dbuf, dd);
+}
+
+static uint64_t
+dsl_dir_space_towrite(dsl_dir_t *dd)
+{
+ uint64_t space = 0;
+ int i;
+
+ ASSERT(MUTEX_HELD(&dd->dd_lock));
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ space += dd->dd_space_towrite[i&TXG_MASK];
+ ASSERT3U(dd->dd_space_towrite[i&TXG_MASK], >=, 0);
+ }
+ return (space);
+}
+
+/*
+ * How much space would dd have available if ancestor had delta applied
+ * to it? If ondiskonly is set, we're only interested in what's
+ * on-disk, not estimated pending changes.
+ */
+uint64_t
+dsl_dir_space_available(dsl_dir_t *dd,
+ dsl_dir_t *ancestor, int64_t delta, int ondiskonly)
+{
+ uint64_t parentspace, myspace, quota, used;
+
+ /*
+ * If there are no restrictions otherwise, assume we have
+ * unlimited space available.
+ */
+ quota = UINT64_MAX;
+ parentspace = UINT64_MAX;
+
+ if (dd->dd_parent != NULL) {
+ parentspace = dsl_dir_space_available(dd->dd_parent,
+ ancestor, delta, ondiskonly);
+ }
+
+ mutex_enter(&dd->dd_lock);
+ if (dd->dd_phys->dd_quota != 0)
+ quota = dd->dd_phys->dd_quota;
+ used = dd->dd_phys->dd_used_bytes;
+ if (!ondiskonly)
+ used += dsl_dir_space_towrite(dd);
+
+ if (dd->dd_parent == NULL) {
+ uint64_t poolsize = dsl_pool_adjustedsize(dd->dd_pool, FALSE);
+ quota = MIN(quota, poolsize);
+ }
+
+ if (dd->dd_phys->dd_reserved > used && parentspace != UINT64_MAX) {
+ /*
+ * We have some space reserved, in addition to what our
+ * parent gave us.
+ */
+ parentspace += dd->dd_phys->dd_reserved - used;
+ }
+
+ if (dd == ancestor) {
+ ASSERT(delta <= 0);
+ ASSERT(used >= -delta);
+ used += delta;
+ if (parentspace != UINT64_MAX)
+ parentspace -= delta;
+ }
+
+ if (used > quota) {
+ /* over quota */
+ myspace = 0;
+
+ /*
+ * While it's OK to be a little over quota, if
+ * we think we are using more space than there
+ * is in the pool (which is already 1.6% more than
+ * dsl_pool_adjustedsize()), something is very
+ * wrong.
+ */
+ ASSERT3U(used, <=, spa_get_space(dd->dd_pool->dp_spa));
+ } else {
+ /*
+ * the lesser of the space provided by our parent and
+ * the space left in our quota
+ */
+ myspace = MIN(parentspace, quota - used);
+ }
+
+ mutex_exit(&dd->dd_lock);
+
+ return (myspace);
+}
+
+struct tempreserve {
+ list_node_t tr_node;
+ dsl_pool_t *tr_dp;
+ dsl_dir_t *tr_ds;
+ uint64_t tr_size;
+};
+
+static int
+dsl_dir_tempreserve_impl(dsl_dir_t *dd, uint64_t asize, boolean_t netfree,
+ boolean_t ignorequota, boolean_t checkrefquota, list_t *tr_list,
+ dmu_tx_t *tx, boolean_t first)
+{
+ uint64_t txg = tx->tx_txg;
+ uint64_t est_inflight, used_on_disk, quota, parent_rsrv;
+ struct tempreserve *tr;
+ int enospc = EDQUOT;
+ int txgidx = txg & TXG_MASK;
+ int i;
+ uint64_t ref_rsrv = 0;
+
+ ASSERT3U(txg, !=, 0);
+ ASSERT3S(asize, >, 0);
+
+ mutex_enter(&dd->dd_lock);
+
+ /*
+ * Check against the dsl_dir's quota. We don't add in the delta
+ * when checking for over-quota because they get one free hit.
+ */
+ est_inflight = dsl_dir_space_towrite(dd);
+ for (i = 0; i < TXG_SIZE; i++)
+ est_inflight += dd->dd_tempreserved[i];
+ used_on_disk = dd->dd_phys->dd_used_bytes;
+
+ /*
+ * On the first iteration, fetch the dataset's used-on-disk and
+ * refreservation values. Also, if checkrefquota is set, test if
+ * allocating this space would exceed the dataset's refquota.
+ */
+ if (first && tx->tx_objset) {
+ int error;
+ dsl_dataset_t *ds = tx->tx_objset->os->os_dsl_dataset;
+
+ error = dsl_dataset_check_quota(ds, checkrefquota,
+ asize, est_inflight, &used_on_disk, &ref_rsrv);
+ if (error) {
+ mutex_exit(&dd->dd_lock);
+ return (error);
+ }
+ }
+
+ /*
+ * If this transaction will result in a net free of space,
+ * we want to let it through.
+ */
+ if (ignorequota || netfree || dd->dd_phys->dd_quota == 0)
+ quota = UINT64_MAX;
+ else
+ quota = dd->dd_phys->dd_quota;
+
+ /*
+ * Adjust the quota against the actual pool size at the root.
+ * To ensure that it's possible to remove files from a full
+ * pool without inducing transient overcommits, we throttle
+ * netfree transactions against a quota that is slightly larger,
+ * but still within the pool's allocation slop. In cases where
+ * we're very close to full, this will allow a steady trickle of
+ * removes to get through.
+ */
+ if (dd->dd_parent == NULL) {
+ uint64_t poolsize = dsl_pool_adjustedsize(dd->dd_pool, netfree);
+ if (poolsize < quota) {
+ quota = poolsize;
+ enospc = ENOSPC;
+ }
+ }
+
+ /*
+ * If they are requesting more space, and our current estimate
+ * is over quota, they get to try again unless the actual
+ * on-disk is over quota and there are no pending changes (which
+ * may free up space for us).
+ */
+ if (used_on_disk + est_inflight > quota) {
+ if (est_inflight > 0 || used_on_disk < quota)
+ enospc = ERESTART;
+ dprintf_dd(dd, "failing: used=%lluK inflight = %lluK "
+ "quota=%lluK tr=%lluK err=%d\n",
+ used_on_disk>>10, est_inflight>>10,
+ quota>>10, asize>>10, enospc);
+ mutex_exit(&dd->dd_lock);
+ return (enospc);
+ }
+
+ /* We need to up our estimated delta before dropping dd_lock */
+ dd->dd_tempreserved[txgidx] += asize;
+
+ parent_rsrv = parent_delta(dd, used_on_disk + est_inflight,
+ asize - ref_rsrv);
+ mutex_exit(&dd->dd_lock);
+
+ tr = kmem_zalloc(sizeof (struct tempreserve), KM_SLEEP);
+ tr->tr_ds = dd;
+ tr->tr_size = asize;
+ list_insert_tail(tr_list, tr);
+
+ /* see if it's OK with our parent */
+ if (dd->dd_parent && parent_rsrv) {
+ boolean_t ismos = (dd->dd_phys->dd_head_dataset_obj == 0);
+
+ return (dsl_dir_tempreserve_impl(dd->dd_parent,
+ parent_rsrv, netfree, ismos, TRUE, tr_list, tx, FALSE));
+ } else {
+ return (0);
+ }
+}
+
+/*
+ * Reserve space in this dsl_dir, to be used in this tx's txg.
+ * After the space has been dirtied (and dsl_dir_willuse_space()
+ * has been called), the reservation should be canceled, using
+ * dsl_dir_tempreserve_clear().
+ */
+int
+dsl_dir_tempreserve_space(dsl_dir_t *dd, uint64_t lsize, uint64_t asize,
+ uint64_t fsize, uint64_t usize, void **tr_cookiep, dmu_tx_t *tx)
+{
+ int err;
+ list_t *tr_list;
+
+ if (asize == 0) {
+ *tr_cookiep = NULL;
+ return (0);
+ }
+
+ tr_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
+ list_create(tr_list, sizeof (struct tempreserve),
+ offsetof(struct tempreserve, tr_node));
+ ASSERT3S(asize, >, 0);
+ ASSERT3S(fsize, >=, 0);
+
+ err = arc_tempreserve_space(lsize, tx->tx_txg);
+ if (err == 0) {
+ struct tempreserve *tr;
+
+ tr = kmem_zalloc(sizeof (struct tempreserve), KM_SLEEP);
+ tr->tr_size = lsize;
+ list_insert_tail(tr_list, tr);
+
+ err = dsl_pool_tempreserve_space(dd->dd_pool, asize, tx);
+ } else {
+ if (err == EAGAIN) {
+ txg_delay(dd->dd_pool, tx->tx_txg, 1);
+ err = ERESTART;
+ }
+ dsl_pool_memory_pressure(dd->dd_pool);
+ }
+
+ if (err == 0) {
+ struct tempreserve *tr;
+
+ tr = kmem_zalloc(sizeof (struct tempreserve), KM_SLEEP);
+ tr->tr_dp = dd->dd_pool;
+ tr->tr_size = asize;
+ list_insert_tail(tr_list, tr);
+
+ err = dsl_dir_tempreserve_impl(dd, asize, fsize >= asize,
+ FALSE, asize > usize, tr_list, tx, TRUE);
+ }
+
+ if (err)
+ dsl_dir_tempreserve_clear(tr_list, tx);
+ else
+ *tr_cookiep = tr_list;
+
+ return (err);
+}
+
+/*
+ * Clear a temporary reservation that we previously made with
+ * dsl_dir_tempreserve_space().
+ */
+void
+dsl_dir_tempreserve_clear(void *tr_cookie, dmu_tx_t *tx)
+{
+ int txgidx = tx->tx_txg & TXG_MASK;
+ list_t *tr_list = tr_cookie;
+ struct tempreserve *tr;
+
+ ASSERT3U(tx->tx_txg, !=, 0);
+
+ if (tr_cookie == NULL)
+ return;
+
+ while (tr = list_head(tr_list)) {
+ if (tr->tr_dp) {
+ dsl_pool_tempreserve_clear(tr->tr_dp, tr->tr_size, tx);
+ } else if (tr->tr_ds) {
+ mutex_enter(&tr->tr_ds->dd_lock);
+ ASSERT3U(tr->tr_ds->dd_tempreserved[txgidx], >=,
+ tr->tr_size);
+ tr->tr_ds->dd_tempreserved[txgidx] -= tr->tr_size;
+ mutex_exit(&tr->tr_ds->dd_lock);
+ } else {
+ arc_tempreserve_clear(tr->tr_size);
+ }
+ list_remove(tr_list, tr);
+ kmem_free(tr, sizeof (struct tempreserve));
+ }
+
+ kmem_free(tr_list, sizeof (list_t));
+}
+
+static void
+dsl_dir_willuse_space_impl(dsl_dir_t *dd, int64_t space, dmu_tx_t *tx)
+{
+ int64_t parent_space;
+ uint64_t est_used;
+
+ mutex_enter(&dd->dd_lock);
+ if (space > 0)
+ dd->dd_space_towrite[tx->tx_txg & TXG_MASK] += space;
+
+ est_used = dsl_dir_space_towrite(dd) + dd->dd_phys->dd_used_bytes;
+ parent_space = parent_delta(dd, est_used, space);
+ mutex_exit(&dd->dd_lock);
+
+ /* Make sure that we clean up dd_space_to* */
+ dsl_dir_dirty(dd, tx);
+
+ /* XXX this is potentially expensive and unnecessary... */
+ if (parent_space && dd->dd_parent)
+ dsl_dir_willuse_space_impl(dd->dd_parent, parent_space, tx);
+}
+
+/*
+ * Call in open context when we think we're going to write/free space,
+ * eg. when dirtying data. Be conservative (ie. OK to write less than
+ * this or free more than this, but don't write more or free less).
+ */
+void
+dsl_dir_willuse_space(dsl_dir_t *dd, int64_t space, dmu_tx_t *tx)
+{
+ dsl_pool_willuse_space(dd->dd_pool, space, tx);
+ dsl_dir_willuse_space_impl(dd, space, tx);
+}
+
+/* call from syncing context when we actually write/free space for this dd */
+void
+dsl_dir_diduse_space(dsl_dir_t *dd, dd_used_t type,
+ int64_t used, int64_t compressed, int64_t uncompressed, dmu_tx_t *tx)
+{
+ int64_t accounted_delta;
+ boolean_t needlock = !MUTEX_HELD(&dd->dd_lock);
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(type < DD_USED_NUM);
+
+ dsl_dir_dirty(dd, tx);
+
+ if (needlock)
+ mutex_enter(&dd->dd_lock);
+ accounted_delta = parent_delta(dd, dd->dd_phys->dd_used_bytes, used);
+ ASSERT(used >= 0 || dd->dd_phys->dd_used_bytes >= -used);
+ ASSERT(compressed >= 0 ||
+ dd->dd_phys->dd_compressed_bytes >= -compressed);
+ ASSERT(uncompressed >= 0 ||
+ dd->dd_phys->dd_uncompressed_bytes >= -uncompressed);
+ dd->dd_phys->dd_used_bytes += used;
+ dd->dd_phys->dd_uncompressed_bytes += uncompressed;
+ dd->dd_phys->dd_compressed_bytes += compressed;
+
+ if (dd->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
+ ASSERT(used > 0 ||
+ dd->dd_phys->dd_used_breakdown[type] >= -used);
+ dd->dd_phys->dd_used_breakdown[type] += used;
+#ifdef DEBUG
+ dd_used_t t;
+ uint64_t u = 0;
+ for (t = 0; t < DD_USED_NUM; t++)
+ u += dd->dd_phys->dd_used_breakdown[t];
+ ASSERT3U(u, ==, dd->dd_phys->dd_used_bytes);
+#endif
+ }
+ if (needlock)
+ mutex_exit(&dd->dd_lock);
+
+ if (dd->dd_parent != NULL) {
+ dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD,
+ accounted_delta, compressed, uncompressed, tx);
+ dsl_dir_transfer_space(dd->dd_parent,
+ used - accounted_delta,
+ DD_USED_CHILD_RSRV, DD_USED_CHILD, tx);
+ }
+}
+
+void
+dsl_dir_transfer_space(dsl_dir_t *dd, int64_t delta,
+ dd_used_t oldtype, dd_used_t newtype, dmu_tx_t *tx)
+{
+ boolean_t needlock = !MUTEX_HELD(&dd->dd_lock);
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(oldtype < DD_USED_NUM);
+ ASSERT(newtype < DD_USED_NUM);
+
+ if (delta == 0 || !(dd->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN))
+ return;
+
+ dsl_dir_dirty(dd, tx);
+ if (needlock)
+ mutex_enter(&dd->dd_lock);
+ ASSERT(delta > 0 ?
+ dd->dd_phys->dd_used_breakdown[oldtype] >= delta :
+ dd->dd_phys->dd_used_breakdown[newtype] >= -delta);
+ ASSERT(dd->dd_phys->dd_used_bytes >= ABS(delta));
+ dd->dd_phys->dd_used_breakdown[oldtype] -= delta;
+ dd->dd_phys->dd_used_breakdown[newtype] += delta;
+ if (needlock)
+ mutex_exit(&dd->dd_lock);
+}
+
+static int
+dsl_dir_set_quota_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ uint64_t *quotap = arg2;
+ uint64_t new_quota = *quotap;
+ int err = 0;
+ uint64_t towrite;
+
+ if (new_quota == 0)
+ return (0);
+
+ mutex_enter(&dd->dd_lock);
+ /*
+ * If we are doing the preliminary check in open context, and
+ * there are pending changes, then don't fail it, since the
+ * pending changes could under-estimate the amount of space to be
+ * freed up.
+ */
+ towrite = dsl_dir_space_towrite(dd);
+ if ((dmu_tx_is_syncing(tx) || towrite == 0) &&
+ (new_quota < dd->dd_phys->dd_reserved ||
+ new_quota < dd->dd_phys->dd_used_bytes + towrite)) {
+ err = ENOSPC;
+ }
+ mutex_exit(&dd->dd_lock);
+ return (err);
+}
+
+/* ARGSUSED */
+static void
+dsl_dir_set_quota_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ uint64_t *quotap = arg2;
+ uint64_t new_quota = *quotap;
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+
+ mutex_enter(&dd->dd_lock);
+ dd->dd_phys->dd_quota = new_quota;
+ mutex_exit(&dd->dd_lock);
+
+ spa_history_internal_log(LOG_DS_QUOTA, dd->dd_pool->dp_spa,
+ tx, cr, "%lld dataset = %llu ",
+ (longlong_t)new_quota, dd->dd_phys->dd_head_dataset_obj);
+}
+
+int
+dsl_dir_set_quota(const char *ddname, uint64_t quota)
+{
+ dsl_dir_t *dd;
+ int err;
+
+ err = dsl_dir_open(ddname, FTAG, &dd, NULL);
+ if (err)
+ return (err);
+
+ if (quota != dd->dd_phys->dd_quota) {
+ /*
+ * If someone removes a file, then tries to set the quota, we
+ * want to make sure the file freeing takes effect.
+ */
+ txg_wait_open(dd->dd_pool, 0);
+
+ err = dsl_sync_task_do(dd->dd_pool, dsl_dir_set_quota_check,
+ dsl_dir_set_quota_sync, dd, &quota, 0);
+ }
+ dsl_dir_close(dd, FTAG);
+ return (err);
+}
+
+int
+dsl_dir_set_reservation_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ uint64_t *reservationp = arg2;
+ uint64_t new_reservation = *reservationp;
+ uint64_t used, avail;
+ int64_t delta;
+
+ if (new_reservation > INT64_MAX)
+ return (EOVERFLOW);
+
+ /*
+ * If we are doing the preliminary check in open context, the
+ * space estimates may be inaccurate.
+ */
+ if (!dmu_tx_is_syncing(tx))
+ return (0);
+
+ mutex_enter(&dd->dd_lock);
+ used = dd->dd_phys->dd_used_bytes;
+ delta = MAX(used, new_reservation) -
+ MAX(used, dd->dd_phys->dd_reserved);
+ mutex_exit(&dd->dd_lock);
+
+ if (dd->dd_parent) {
+ avail = dsl_dir_space_available(dd->dd_parent,
+ NULL, 0, FALSE);
+ } else {
+ avail = dsl_pool_adjustedsize(dd->dd_pool, B_FALSE) - used;
+ }
+
+ if (delta > 0 && delta > avail)
+ return (ENOSPC);
+ if (delta > 0 && dd->dd_phys->dd_quota > 0 &&
+ new_reservation > dd->dd_phys->dd_quota)
+ return (ENOSPC);
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+dsl_dir_set_reservation_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ uint64_t *reservationp = arg2;
+ uint64_t new_reservation = *reservationp;
+ uint64_t used;
+ int64_t delta;
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+
+ mutex_enter(&dd->dd_lock);
+ used = dd->dd_phys->dd_used_bytes;
+ delta = MAX(used, new_reservation) -
+ MAX(used, dd->dd_phys->dd_reserved);
+ dd->dd_phys->dd_reserved = new_reservation;
+
+ if (dd->dd_parent != NULL) {
+ /* Roll up this additional usage into our ancestors */
+ dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD_RSRV,
+ delta, 0, 0, tx);
+ }
+ mutex_exit(&dd->dd_lock);
+
+ spa_history_internal_log(LOG_DS_RESERVATION, dd->dd_pool->dp_spa,
+ tx, cr, "%lld dataset = %llu",
+ (longlong_t)new_reservation, dd->dd_phys->dd_head_dataset_obj);
+}
+
+int
+dsl_dir_set_reservation(const char *ddname, uint64_t reservation)
+{
+ dsl_dir_t *dd;
+ int err;
+
+ err = dsl_dir_open(ddname, FTAG, &dd, NULL);
+ if (err)
+ return (err);
+ err = dsl_sync_task_do(dd->dd_pool, dsl_dir_set_reservation_check,
+ dsl_dir_set_reservation_sync, dd, &reservation, 0);
+ dsl_dir_close(dd, FTAG);
+ return (err);
+}
+
+static dsl_dir_t *
+closest_common_ancestor(dsl_dir_t *ds1, dsl_dir_t *ds2)
+{
+ for (; ds1; ds1 = ds1->dd_parent) {
+ dsl_dir_t *dd;
+ for (dd = ds2; dd; dd = dd->dd_parent) {
+ if (ds1 == dd)
+ return (dd);
+ }
+ }
+ return (NULL);
+}
+
+/*
+ * If delta is applied to dd, how much of that delta would be applied to
+ * ancestor? Syncing context only.
+ */
+static int64_t
+would_change(dsl_dir_t *dd, int64_t delta, dsl_dir_t *ancestor)
+{
+ if (dd == ancestor)
+ return (delta);
+
+ mutex_enter(&dd->dd_lock);
+ delta = parent_delta(dd, dd->dd_phys->dd_used_bytes, delta);
+ mutex_exit(&dd->dd_lock);
+ return (would_change(dd->dd_parent, delta, ancestor));
+}
+
+struct renamearg {
+ dsl_dir_t *newparent;
+ const char *mynewname;
+};
+
+/*ARGSUSED*/
+static int
+dsl_dir_rename_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct renamearg *ra = arg2;
+ dsl_pool_t *dp = dd->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+ int err;
+ uint64_t val;
+
+ /* There should be 2 references: the open and the dirty */
+ if (dmu_buf_refcount(dd->dd_dbuf) > 2)
+ return (EBUSY);
+
+ /* check for existing name */
+ err = zap_lookup(mos, ra->newparent->dd_phys->dd_child_dir_zapobj,
+ ra->mynewname, 8, 1, &val);
+ if (err == 0)
+ return (EEXIST);
+ if (err != ENOENT)
+ return (err);
+
+ if (ra->newparent != dd->dd_parent) {
+ /* is there enough space? */
+ uint64_t myspace =
+ MAX(dd->dd_phys->dd_used_bytes, dd->dd_phys->dd_reserved);
+
+ /* no rename into our descendant */
+ if (closest_common_ancestor(dd, ra->newparent) == dd)
+ return (EINVAL);
+
+ if (err = dsl_dir_transfer_possible(dd->dd_parent,
+ ra->newparent, myspace))
+ return (err);
+ }
+
+ return (0);
+}
+
+static void
+dsl_dir_rename_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dir_t *dd = arg1;
+ struct renamearg *ra = arg2;
+ dsl_pool_t *dp = dd->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+ int err;
+
+ ASSERT(dmu_buf_refcount(dd->dd_dbuf) <= 2);
+
+ if (ra->newparent != dd->dd_parent) {
+ dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD,
+ -dd->dd_phys->dd_used_bytes,
+ -dd->dd_phys->dd_compressed_bytes,
+ -dd->dd_phys->dd_uncompressed_bytes, tx);
+ dsl_dir_diduse_space(ra->newparent, DD_USED_CHILD,
+ dd->dd_phys->dd_used_bytes,
+ dd->dd_phys->dd_compressed_bytes,
+ dd->dd_phys->dd_uncompressed_bytes, tx);
+
+ if (dd->dd_phys->dd_reserved > dd->dd_phys->dd_used_bytes) {
+ uint64_t unused_rsrv = dd->dd_phys->dd_reserved -
+ dd->dd_phys->dd_used_bytes;
+
+ dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD_RSRV,
+ -unused_rsrv, 0, 0, tx);
+ dsl_dir_diduse_space(ra->newparent, DD_USED_CHILD_RSRV,
+ unused_rsrv, 0, 0, tx);
+ }
+ }
+
+ dmu_buf_will_dirty(dd->dd_dbuf, tx);
+
+ /* remove from old parent zapobj */
+ err = zap_remove(mos, dd->dd_parent->dd_phys->dd_child_dir_zapobj,
+ dd->dd_myname, tx);
+ ASSERT3U(err, ==, 0);
+
+ (void) strcpy(dd->dd_myname, ra->mynewname);
+ dsl_dir_close(dd->dd_parent, dd);
+ dd->dd_phys->dd_parent_obj = ra->newparent->dd_object;
+ VERIFY(0 == dsl_dir_open_obj(dd->dd_pool,
+ ra->newparent->dd_object, NULL, dd, &dd->dd_parent));
+
+ /* add to new parent zapobj */
+ err = zap_add(mos, ra->newparent->dd_phys->dd_child_dir_zapobj,
+ dd->dd_myname, 8, 1, &dd->dd_object, tx);
+ ASSERT3U(err, ==, 0);
+
+ spa_history_internal_log(LOG_DS_RENAME, dd->dd_pool->dp_spa,
+ tx, cr, "dataset = %llu", dd->dd_phys->dd_head_dataset_obj);
+}
+
+int
+dsl_dir_rename(dsl_dir_t *dd, const char *newname)
+{
+ struct renamearg ra;
+ int err;
+
+ /* new parent should exist */
+ err = dsl_dir_open(newname, FTAG, &ra.newparent, &ra.mynewname);
+ if (err)
+ return (err);
+
+ /* can't rename to different pool */
+ if (dd->dd_pool != ra.newparent->dd_pool) {
+ err = ENXIO;
+ goto out;
+ }
+
+ /* new name should not already exist */
+ if (ra.mynewname == NULL) {
+ err = EEXIST;
+ goto out;
+ }
+
+ err = dsl_sync_task_do(dd->dd_pool,
+ dsl_dir_rename_check, dsl_dir_rename_sync, dd, &ra, 3);
+
+out:
+ dsl_dir_close(ra.newparent, FTAG);
+ return (err);
+}
+
+int
+dsl_dir_transfer_possible(dsl_dir_t *sdd, dsl_dir_t *tdd, uint64_t space)
+{
+ dsl_dir_t *ancestor;
+ int64_t adelta;
+ uint64_t avail;
+
+ ancestor = closest_common_ancestor(sdd, tdd);
+ adelta = would_change(sdd, -space, ancestor);
+ avail = dsl_dir_space_available(tdd, ancestor, adelta, FALSE);
+ if (avail < space)
+ return (ENOSPC);
+
+ return (0);
+}
diff --git a/module/zfs/dsl_pool.c b/module/zfs/dsl_pool.c
new file mode 100644
index 000000000..dacc57c81
--- /dev/null
+++ b/module/zfs/dsl_pool.c
@@ -0,0 +1,613 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dsl_pool.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/arc.h>
+#include <sys/zap.h>
+#include <sys/zio.h>
+#include <sys/zfs_context.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_znode.h>
+#include <sys/spa_impl.h>
+
+int zfs_no_write_throttle = 0;
+int zfs_write_limit_shift = 3; /* 1/8th of physical memory */
+int zfs_txg_synctime = 5; /* target secs to sync a txg */
+
+uint64_t zfs_write_limit_min = 32 << 20; /* min write limit is 32MB */
+uint64_t zfs_write_limit_max = 0; /* max data payload per txg */
+uint64_t zfs_write_limit_inflated = 0;
+uint64_t zfs_write_limit_override = 0;
+
+kmutex_t zfs_write_limit_lock;
+
+static pgcnt_t old_physmem = 0;
+
+static int
+dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **ddp)
+{
+ uint64_t obj;
+ int err;
+
+ err = zap_lookup(dp->dp_meta_objset,
+ dp->dp_root_dir->dd_phys->dd_child_dir_zapobj,
+ name, sizeof (obj), 1, &obj);
+ if (err)
+ return (err);
+
+ return (dsl_dir_open_obj(dp, obj, name, dp, ddp));
+}
+
+static dsl_pool_t *
+dsl_pool_open_impl(spa_t *spa, uint64_t txg)
+{
+ dsl_pool_t *dp;
+ blkptr_t *bp = spa_get_rootblkptr(spa);
+
+ dp = kmem_zalloc(sizeof (dsl_pool_t), KM_SLEEP);
+ dp->dp_spa = spa;
+ dp->dp_meta_rootbp = *bp;
+ rw_init(&dp->dp_config_rwlock, NULL, RW_DEFAULT, NULL);
+ dp->dp_write_limit = zfs_write_limit_min;
+ txg_init(dp, txg);
+
+ txg_list_create(&dp->dp_dirty_datasets,
+ offsetof(dsl_dataset_t, ds_dirty_link));
+ txg_list_create(&dp->dp_dirty_dirs,
+ offsetof(dsl_dir_t, dd_dirty_link));
+ txg_list_create(&dp->dp_sync_tasks,
+ offsetof(dsl_sync_task_group_t, dstg_node));
+ list_create(&dp->dp_synced_datasets, sizeof (dsl_dataset_t),
+ offsetof(dsl_dataset_t, ds_synced_link));
+
+ mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&dp->dp_scrub_cancel_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ return (dp);
+}
+
+int
+dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
+{
+ int err;
+ dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
+ dsl_dir_t *dd;
+ dsl_dataset_t *ds;
+ objset_impl_t *osi;
+
+ rw_enter(&dp->dp_config_rwlock, RW_WRITER);
+ err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp, &osi);
+ if (err)
+ goto out;
+ dp->dp_meta_objset = &osi->os;
+
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
+ &dp->dp_root_dir_obj);
+ if (err)
+ goto out;
+
+ err = dsl_dir_open_obj(dp, dp->dp_root_dir_obj,
+ NULL, dp, &dp->dp_root_dir);
+ if (err)
+ goto out;
+
+ err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
+ if (err)
+ goto out;
+
+ if (spa_version(spa) >= SPA_VERSION_ORIGIN) {
+ err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
+ if (err)
+ goto out;
+ err = dsl_dataset_hold_obj(dp, dd->dd_phys->dd_head_dataset_obj,
+ FTAG, &ds);
+ if (err)
+ goto out;
+ err = dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
+ dp, &dp->dp_origin_snap);
+ if (err)
+ goto out;
+ dsl_dataset_rele(ds, FTAG);
+ dsl_dir_close(dd, dp);
+ }
+
+ /* get scrub status */
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_FUNC, sizeof (uint32_t), 1,
+ &dp->dp_scrub_func);
+ if (err == 0) {
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_QUEUE, sizeof (uint64_t), 1,
+ &dp->dp_scrub_queue_obj);
+ if (err)
+ goto out;
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MIN_TXG, sizeof (uint64_t), 1,
+ &dp->dp_scrub_min_txg);
+ if (err)
+ goto out;
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MAX_TXG, sizeof (uint64_t), 1,
+ &dp->dp_scrub_max_txg);
+ if (err)
+ goto out;
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_BOOKMARK, sizeof (uint64_t), 4,
+ &dp->dp_scrub_bookmark);
+ if (err)
+ goto out;
+ err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_ERRORS, sizeof (uint64_t), 1,
+ &spa->spa_scrub_errors);
+ if (err)
+ goto out;
+ if (spa_version(spa) < SPA_VERSION_DSL_SCRUB) {
+ /*
+ * A new-type scrub was in progress on an old
+ * pool. Restart from the beginning, since the
+ * old software may have changed the pool in the
+ * meantime.
+ */
+ dsl_pool_scrub_restart(dp);
+ }
+ } else {
+ /*
+ * It's OK if there is no scrub in progress (and if
+ * there was an I/O error, ignore it).
+ */
+ err = 0;
+ }
+
+out:
+ rw_exit(&dp->dp_config_rwlock);
+ if (err)
+ dsl_pool_close(dp);
+ else
+ *dpp = dp;
+
+ return (err);
+}
+
+void
+dsl_pool_close(dsl_pool_t *dp)
+{
+ /* drop our references from dsl_pool_open() */
+
+ /*
+ * Since we held the origin_snap from "syncing" context (which
+ * includes pool-opening context), it actually only got a "ref"
+ * and not a hold, so just drop that here.
+ */
+ if (dp->dp_origin_snap)
+ dsl_dataset_drop_ref(dp->dp_origin_snap, dp);
+ if (dp->dp_mos_dir)
+ dsl_dir_close(dp->dp_mos_dir, dp);
+ if (dp->dp_root_dir)
+ dsl_dir_close(dp->dp_root_dir, dp);
+
+ /* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
+ if (dp->dp_meta_objset)
+ dmu_objset_evict(NULL, dp->dp_meta_objset->os);
+
+ txg_list_destroy(&dp->dp_dirty_datasets);
+ txg_list_destroy(&dp->dp_dirty_dirs);
+ list_destroy(&dp->dp_synced_datasets);
+
+ arc_flush(dp->dp_spa);
+ txg_fini(dp);
+ rw_destroy(&dp->dp_config_rwlock);
+ mutex_destroy(&dp->dp_lock);
+ mutex_destroy(&dp->dp_scrub_cancel_lock);
+ if (dp->dp_blkstats)
+ kmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
+ kmem_free(dp, sizeof (dsl_pool_t));
+}
+
+dsl_pool_t *
+dsl_pool_create(spa_t *spa, nvlist_t *zplprops, uint64_t txg)
+{
+ int err;
+ dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
+ dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
+ objset_impl_t *osip;
+ dsl_dataset_t *ds;
+ uint64_t dsobj;
+
+ /* create and open the MOS (meta-objset) */
+ dp->dp_meta_objset = &dmu_objset_create_impl(spa,
+ NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx)->os;
+
+ /* create the pool directory */
+ err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
+ ASSERT3U(err, ==, 0);
+
+ /* create and open the root dir */
+ dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
+ VERIFY(0 == dsl_dir_open_obj(dp, dp->dp_root_dir_obj,
+ NULL, dp, &dp->dp_root_dir));
+
+ /* create and open the meta-objset dir */
+ (void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
+ VERIFY(0 == dsl_pool_open_special_dir(dp,
+ MOS_DIR_NAME, &dp->dp_mos_dir));
+
+ if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
+ dsl_pool_create_origin(dp, tx);
+
+ /* create the root dataset */
+ dsobj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, 0, tx);
+
+ /* create the root objset */
+ VERIFY(0 == dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
+ osip = dmu_objset_create_impl(dp->dp_spa, ds,
+ dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
+#ifdef _KERNEL
+ zfs_create_fs(&osip->os, kcred, zplprops, tx);
+#endif
+ dsl_dataset_rele(ds, FTAG);
+
+ dmu_tx_commit(tx);
+
+ return (dp);
+}
+
+void
+dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
+{
+ zio_t *zio;
+ dmu_tx_t *tx;
+ dsl_dir_t *dd;
+ dsl_dataset_t *ds;
+ dsl_sync_task_group_t *dstg;
+ objset_impl_t *mosi = dp->dp_meta_objset->os;
+ hrtime_t start, write_time;
+ uint64_t data_written;
+ int err;
+
+ tx = dmu_tx_create_assigned(dp, txg);
+
+ dp->dp_read_overhead = 0;
+ zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
+ while (ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) {
+ if (!list_link_active(&ds->ds_synced_link))
+ list_insert_tail(&dp->dp_synced_datasets, ds);
+ else
+ dmu_buf_rele(ds->ds_dbuf, ds);
+ dsl_dataset_sync(ds, zio, tx);
+ }
+ DTRACE_PROBE(pool_sync__1setup);
+
+ start = gethrtime();
+ err = zio_wait(zio);
+ write_time = gethrtime() - start;
+ ASSERT(err == 0);
+ DTRACE_PROBE(pool_sync__2rootzio);
+
+ while (dstg = txg_list_remove(&dp->dp_sync_tasks, txg))
+ dsl_sync_task_group_sync(dstg, tx);
+ DTRACE_PROBE(pool_sync__3task);
+
+ start = gethrtime();
+ while (dd = txg_list_remove(&dp->dp_dirty_dirs, txg))
+ dsl_dir_sync(dd, tx);
+ write_time += gethrtime() - start;
+
+ if (spa_sync_pass(dp->dp_spa) == 1)
+ dsl_pool_scrub_sync(dp, tx);
+
+ start = gethrtime();
+ if (list_head(&mosi->os_dirty_dnodes[txg & TXG_MASK]) != NULL ||
+ list_head(&mosi->os_free_dnodes[txg & TXG_MASK]) != NULL) {
+ zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
+ dmu_objset_sync(mosi, zio, tx);
+ err = zio_wait(zio);
+ ASSERT(err == 0);
+ dprintf_bp(&dp->dp_meta_rootbp, "meta objset rootbp is %s", "");
+ spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
+ }
+ write_time += gethrtime() - start;
+ DTRACE_PROBE2(pool_sync__4io, hrtime_t, write_time,
+ hrtime_t, dp->dp_read_overhead);
+ write_time -= dp->dp_read_overhead;
+
+ dmu_tx_commit(tx);
+
+ data_written = dp->dp_space_towrite[txg & TXG_MASK];
+ dp->dp_space_towrite[txg & TXG_MASK] = 0;
+ ASSERT(dp->dp_tempreserved[txg & TXG_MASK] == 0);
+
+ /*
+ * If the write limit max has not been explicitly set, set it
+ * to a fraction of available physical memory (default 1/8th).
+ * Note that we must inflate the limit because the spa
+ * inflates write sizes to account for data replication.
+ * Check this each sync phase to catch changing memory size.
+ */
+ if (physmem != old_physmem && zfs_write_limit_shift) {
+ mutex_enter(&zfs_write_limit_lock);
+ old_physmem = physmem;
+ zfs_write_limit_max = ptob(physmem) >> zfs_write_limit_shift;
+ zfs_write_limit_inflated = MAX(zfs_write_limit_min,
+ spa_get_asize(dp->dp_spa, zfs_write_limit_max));
+ mutex_exit(&zfs_write_limit_lock);
+ }
+
+ /*
+ * Attempt to keep the sync time consistent by adjusting the
+ * amount of write traffic allowed into each transaction group.
+ * Weight the throughput calculation towards the current value:
+ * thru = 3/4 old_thru + 1/4 new_thru
+ */
+ ASSERT(zfs_write_limit_min > 0);
+ if (data_written > zfs_write_limit_min / 8 && write_time > 0) {
+ uint64_t throughput = (data_written * NANOSEC) / write_time;
+ if (dp->dp_throughput)
+ dp->dp_throughput = throughput / 4 +
+ 3 * dp->dp_throughput / 4;
+ else
+ dp->dp_throughput = throughput;
+ dp->dp_write_limit = MIN(zfs_write_limit_inflated,
+ MAX(zfs_write_limit_min,
+ dp->dp_throughput * zfs_txg_synctime));
+ }
+}
+
+void
+dsl_pool_zil_clean(dsl_pool_t *dp)
+{
+ dsl_dataset_t *ds;
+
+ while (ds = list_head(&dp->dp_synced_datasets)) {
+ list_remove(&dp->dp_synced_datasets, ds);
+ ASSERT(ds->ds_user_ptr != NULL);
+ zil_clean(((objset_impl_t *)ds->ds_user_ptr)->os_zil);
+ dmu_buf_rele(ds->ds_dbuf, ds);
+ }
+}
+
+/*
+ * TRUE if the current thread is the tx_sync_thread or if we
+ * are being called from SPA context during pool initialization.
+ */
+int
+dsl_pool_sync_context(dsl_pool_t *dp)
+{
+ return (curthread == dp->dp_tx.tx_sync_thread ||
+ spa_get_dsl(dp->dp_spa) == NULL);
+}
+
+uint64_t
+dsl_pool_adjustedsize(dsl_pool_t *dp, boolean_t netfree)
+{
+ uint64_t space, resv;
+
+ /*
+ * Reserve about 1.6% (1/64), or at least 32MB, for allocation
+ * efficiency.
+ * XXX The intent log is not accounted for, so it must fit
+ * within this slop.
+ *
+ * If we're trying to assess whether it's OK to do a free,
+ * cut the reservation in half to allow forward progress
+ * (e.g. make it possible to rm(1) files from a full pool).
+ */
+ space = spa_get_dspace(dp->dp_spa);
+ resv = MAX(space >> 6, SPA_MINDEVSIZE >> 1);
+ if (netfree)
+ resv >>= 1;
+
+ return (space - resv);
+}
+
+int
+dsl_pool_tempreserve_space(dsl_pool_t *dp, uint64_t space, dmu_tx_t *tx)
+{
+ uint64_t reserved = 0;
+ uint64_t write_limit = (zfs_write_limit_override ?
+ zfs_write_limit_override : dp->dp_write_limit);
+
+ if (zfs_no_write_throttle) {
+ atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK],
+ space);
+ return (0);
+ }
+
+ /*
+ * Check to see if we have exceeded the maximum allowed IO for
+ * this transaction group. We can do this without locks since
+ * a little slop here is ok. Note that we do the reserved check
+ * with only half the requested reserve: this is because the
+ * reserve requests are worst-case, and we really don't want to
+ * throttle based off of worst-case estimates.
+ */
+ if (write_limit > 0) {
+ reserved = dp->dp_space_towrite[tx->tx_txg & TXG_MASK]
+ + dp->dp_tempreserved[tx->tx_txg & TXG_MASK] / 2;
+
+ if (reserved && reserved > write_limit)
+ return (ERESTART);
+ }
+
+ atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK], space);
+
+ /*
+ * If this transaction group is over 7/8ths capacity, delay
+ * the caller 1 clock tick. This will slow down the "fill"
+ * rate until the sync process can catch up with us.
+ */
+ if (reserved && reserved > (write_limit - (write_limit >> 3)))
+ txg_delay(dp, tx->tx_txg, 1);
+
+ return (0);
+}
+
+void
+dsl_pool_tempreserve_clear(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
+{
+ ASSERT(dp->dp_tempreserved[tx->tx_txg & TXG_MASK] >= space);
+ atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK], -space);
+}
+
+void
+dsl_pool_memory_pressure(dsl_pool_t *dp)
+{
+ uint64_t space_inuse = 0;
+ int i;
+
+ if (dp->dp_write_limit == zfs_write_limit_min)
+ return;
+
+ for (i = 0; i < TXG_SIZE; i++) {
+ space_inuse += dp->dp_space_towrite[i];
+ space_inuse += dp->dp_tempreserved[i];
+ }
+ dp->dp_write_limit = MAX(zfs_write_limit_min,
+ MIN(dp->dp_write_limit, space_inuse / 4));
+}
+
+void
+dsl_pool_willuse_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
+{
+ if (space > 0) {
+ mutex_enter(&dp->dp_lock);
+ dp->dp_space_towrite[tx->tx_txg & TXG_MASK] += space;
+ mutex_exit(&dp->dp_lock);
+ }
+}
+
+/* ARGSUSED */
+static int
+upgrade_clones_cb(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg)
+{
+ dmu_tx_t *tx = arg;
+ dsl_dataset_t *ds, *prev = NULL;
+ int err;
+ dsl_pool_t *dp = spa_get_dsl(spa);
+
+ err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
+ if (err)
+ return (err);
+
+ while (ds->ds_phys->ds_prev_snap_obj != 0) {
+ err = dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
+ FTAG, &prev);
+ if (err) {
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+ }
+
+ if (prev->ds_phys->ds_next_snap_obj != ds->ds_object)
+ break;
+ dsl_dataset_rele(ds, FTAG);
+ ds = prev;
+ prev = NULL;
+ }
+
+ if (prev == NULL) {
+ prev = dp->dp_origin_snap;
+
+ /*
+ * The $ORIGIN can't have any data, or the accounting
+ * will be wrong.
+ */
+ ASSERT(prev->ds_phys->ds_bp.blk_birth == 0);
+
+ /* The origin doesn't get attached to itself */
+ if (ds->ds_object == prev->ds_object) {
+ dsl_dataset_rele(ds, FTAG);
+ return (0);
+ }
+
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_prev_snap_obj = prev->ds_object;
+ ds->ds_phys->ds_prev_snap_txg = prev->ds_phys->ds_creation_txg;
+
+ dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
+ ds->ds_dir->dd_phys->dd_origin_obj = prev->ds_object;
+
+ dmu_buf_will_dirty(prev->ds_dbuf, tx);
+ prev->ds_phys->ds_num_children++;
+
+ if (ds->ds_phys->ds_next_snap_obj == 0) {
+ ASSERT(ds->ds_prev == NULL);
+ VERIFY(0 == dsl_dataset_hold_obj(dp,
+ ds->ds_phys->ds_prev_snap_obj, ds, &ds->ds_prev));
+ }
+ }
+
+ ASSERT(ds->ds_dir->dd_phys->dd_origin_obj == prev->ds_object);
+ ASSERT(ds->ds_phys->ds_prev_snap_obj == prev->ds_object);
+
+ if (prev->ds_phys->ds_next_clones_obj == 0) {
+ prev->ds_phys->ds_next_clones_obj =
+ zap_create(dp->dp_meta_objset,
+ DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
+ }
+ VERIFY(0 == zap_add_int(dp->dp_meta_objset,
+ prev->ds_phys->ds_next_clones_obj, ds->ds_object, tx));
+
+ dsl_dataset_rele(ds, FTAG);
+ if (prev != dp->dp_origin_snap)
+ dsl_dataset_rele(prev, FTAG);
+ return (0);
+}
+
+void
+dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx)
+{
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(dp->dp_origin_snap != NULL);
+
+ (void) dmu_objset_find_spa(dp->dp_spa, NULL, upgrade_clones_cb,
+ tx, DS_FIND_CHILDREN);
+}
+
+void
+dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx)
+{
+ uint64_t dsobj;
+ dsl_dataset_t *ds;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+ ASSERT(dp->dp_origin_snap == NULL);
+
+ /* create the origin dir, ds, & snap-ds */
+ rw_enter(&dp->dp_config_rwlock, RW_WRITER);
+ dsobj = dsl_dataset_create_sync(dp->dp_root_dir, ORIGIN_DIR_NAME,
+ NULL, 0, kcred, tx);
+ VERIFY(0 == dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
+ dsl_dataset_snapshot_sync(ds, ORIGIN_DIR_NAME, kcred, tx);
+ VERIFY(0 == dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
+ dp, &dp->dp_origin_snap));
+ dsl_dataset_rele(ds, FTAG);
+ rw_exit(&dp->dp_config_rwlock);
+}
diff --git a/module/zfs/dsl_prop.c b/module/zfs/dsl_prop.c
new file mode 100644
index 000000000..212acbbc5
--- /dev/null
+++ b/module/zfs/dsl_prop.c
@@ -0,0 +1,602 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
+#include <sys/dmu_tx.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/spa.h>
+#include <sys/zio_checksum.h> /* for the default checksum value */
+#include <sys/zap.h>
+#include <sys/fs/zfs.h>
+
+#include "zfs_prop.h"
+
+static int
+dodefault(const char *propname, int intsz, int numint, void *buf)
+{
+ zfs_prop_t prop;
+
+ /*
+ * The setonce properties are read-only, BUT they still
+ * have a default value that can be used as the initial
+ * value.
+ */
+ if ((prop = zfs_name_to_prop(propname)) == ZPROP_INVAL ||
+ (zfs_prop_readonly(prop) && !zfs_prop_setonce(prop)))
+ return (ENOENT);
+
+ if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
+ if (intsz != 1)
+ return (EOVERFLOW);
+ (void) strncpy(buf, zfs_prop_default_string(prop),
+ numint);
+ } else {
+ if (intsz != 8 || numint < 1)
+ return (EOVERFLOW);
+
+ *(uint64_t *)buf = zfs_prop_default_numeric(prop);
+ }
+
+ return (0);
+}
+
+int
+dsl_prop_get_dd(dsl_dir_t *dd, const char *propname,
+ int intsz, int numint, void *buf, char *setpoint)
+{
+ int err = ENOENT;
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ zfs_prop_t prop;
+
+ ASSERT(RW_LOCK_HELD(&dd->dd_pool->dp_config_rwlock));
+
+ if (setpoint)
+ setpoint[0] = '\0';
+
+ prop = zfs_name_to_prop(propname);
+
+ /*
+ * Note: dd may be NULL, therefore we shouldn't dereference it
+ * ouside this loop.
+ */
+ for (; dd != NULL; dd = dd->dd_parent) {
+ ASSERT(RW_LOCK_HELD(&dd->dd_pool->dp_config_rwlock));
+ err = zap_lookup(mos, dd->dd_phys->dd_props_zapobj,
+ propname, intsz, numint, buf);
+ if (err != ENOENT) {
+ if (setpoint)
+ dsl_dir_name(dd, setpoint);
+ break;
+ }
+
+ /*
+ * Break out of this loop for non-inheritable properties.
+ */
+ if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
+ break;
+ }
+ if (err == ENOENT)
+ err = dodefault(propname, intsz, numint, buf);
+
+ return (err);
+}
+
+int
+dsl_prop_get_ds(dsl_dataset_t *ds, const char *propname,
+ int intsz, int numint, void *buf, char *setpoint)
+{
+ ASSERT(RW_LOCK_HELD(&ds->ds_dir->dd_pool->dp_config_rwlock));
+
+ if (ds->ds_phys->ds_props_obj) {
+ int err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
+ ds->ds_phys->ds_props_obj, propname, intsz, numint, buf);
+ if (err != ENOENT) {
+ if (setpoint)
+ dsl_dataset_name(ds, setpoint);
+ return (err);
+ }
+ }
+
+ return (dsl_prop_get_dd(ds->ds_dir, propname,
+ intsz, numint, buf, setpoint));
+}
+
+/*
+ * Register interest in the named property. We'll call the callback
+ * once to notify it of the current property value, and again each time
+ * the property changes, until this callback is unregistered.
+ *
+ * Return 0 on success, errno if the prop is not an integer value.
+ */
+int
+dsl_prop_register(dsl_dataset_t *ds, const char *propname,
+ dsl_prop_changed_cb_t *callback, void *cbarg)
+{
+ dsl_dir_t *dd = ds->ds_dir;
+ dsl_pool_t *dp = dd->dd_pool;
+ uint64_t value;
+ dsl_prop_cb_record_t *cbr;
+ int err;
+ int need_rwlock;
+
+ need_rwlock = !RW_WRITE_HELD(&dp->dp_config_rwlock);
+ if (need_rwlock)
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+
+ err = dsl_prop_get_ds(ds, propname, 8, 1, &value, NULL);
+ if (err != 0) {
+ if (need_rwlock)
+ rw_exit(&dp->dp_config_rwlock);
+ return (err);
+ }
+
+ cbr = kmem_alloc(sizeof (dsl_prop_cb_record_t), KM_SLEEP);
+ cbr->cbr_ds = ds;
+ cbr->cbr_propname = kmem_alloc(strlen(propname)+1, KM_SLEEP);
+ (void) strcpy((char *)cbr->cbr_propname, propname);
+ cbr->cbr_func = callback;
+ cbr->cbr_arg = cbarg;
+ mutex_enter(&dd->dd_lock);
+ list_insert_head(&dd->dd_prop_cbs, cbr);
+ mutex_exit(&dd->dd_lock);
+
+ cbr->cbr_func(cbr->cbr_arg, value);
+
+ VERIFY(0 == dsl_dir_open_obj(dp, dd->dd_object,
+ NULL, cbr, &dd));
+ if (need_rwlock)
+ rw_exit(&dp->dp_config_rwlock);
+ /* Leave dir open until this callback is unregistered */
+ return (0);
+}
+
+int
+dsl_prop_get(const char *dsname, const char *propname,
+ int intsz, int numints, void *buf, char *setpoint)
+{
+ dsl_dataset_t *ds;
+ int err;
+
+ err = dsl_dataset_hold(dsname, FTAG, &ds);
+ if (err)
+ return (err);
+
+ rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
+ err = dsl_prop_get_ds(ds, propname, intsz, numints, buf, setpoint);
+ rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
+
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+}
+
+/*
+ * Get the current property value. It may have changed by the time this
+ * function returns, so it is NOT safe to follow up with
+ * dsl_prop_register() and assume that the value has not changed in
+ * between.
+ *
+ * Return 0 on success, ENOENT if ddname is invalid.
+ */
+int
+dsl_prop_get_integer(const char *ddname, const char *propname,
+ uint64_t *valuep, char *setpoint)
+{
+ return (dsl_prop_get(ddname, propname, 8, 1, valuep, setpoint));
+}
+
+/*
+ * Unregister this callback. Return 0 on success, ENOENT if ddname is
+ * invalid, ENOMSG if no matching callback registered.
+ */
+int
+dsl_prop_unregister(dsl_dataset_t *ds, const char *propname,
+ dsl_prop_changed_cb_t *callback, void *cbarg)
+{
+ dsl_dir_t *dd = ds->ds_dir;
+ dsl_prop_cb_record_t *cbr;
+
+ mutex_enter(&dd->dd_lock);
+ for (cbr = list_head(&dd->dd_prop_cbs);
+ cbr; cbr = list_next(&dd->dd_prop_cbs, cbr)) {
+ if (cbr->cbr_ds == ds &&
+ cbr->cbr_func == callback &&
+ cbr->cbr_arg == cbarg &&
+ strcmp(cbr->cbr_propname, propname) == 0)
+ break;
+ }
+
+ if (cbr == NULL) {
+ mutex_exit(&dd->dd_lock);
+ return (ENOMSG);
+ }
+
+ list_remove(&dd->dd_prop_cbs, cbr);
+ mutex_exit(&dd->dd_lock);
+ kmem_free((void*)cbr->cbr_propname, strlen(cbr->cbr_propname)+1);
+ kmem_free(cbr, sizeof (dsl_prop_cb_record_t));
+
+ /* Clean up from dsl_prop_register */
+ dsl_dir_close(dd, cbr);
+ return (0);
+}
+
+/*
+ * Return the number of callbacks that are registered for this dataset.
+ */
+int
+dsl_prop_numcb(dsl_dataset_t *ds)
+{
+ dsl_dir_t *dd = ds->ds_dir;
+ dsl_prop_cb_record_t *cbr;
+ int num = 0;
+
+ mutex_enter(&dd->dd_lock);
+ for (cbr = list_head(&dd->dd_prop_cbs);
+ cbr; cbr = list_next(&dd->dd_prop_cbs, cbr)) {
+ if (cbr->cbr_ds == ds)
+ num++;
+ }
+ mutex_exit(&dd->dd_lock);
+
+ return (num);
+}
+
+static void
+dsl_prop_changed_notify(dsl_pool_t *dp, uint64_t ddobj,
+ const char *propname, uint64_t value, int first)
+{
+ dsl_dir_t *dd;
+ dsl_prop_cb_record_t *cbr;
+ objset_t *mos = dp->dp_meta_objset;
+ zap_cursor_t zc;
+ zap_attribute_t *za;
+ int err;
+ uint64_t dummyval;
+
+ ASSERT(RW_WRITE_HELD(&dp->dp_config_rwlock));
+ err = dsl_dir_open_obj(dp, ddobj, NULL, FTAG, &dd);
+ if (err)
+ return;
+
+ if (!first) {
+ /*
+ * If the prop is set here, then this change is not
+ * being inherited here or below; stop the recursion.
+ */
+ err = zap_lookup(mos, dd->dd_phys->dd_props_zapobj, propname,
+ 8, 1, &dummyval);
+ if (err == 0) {
+ dsl_dir_close(dd, FTAG);
+ return;
+ }
+ ASSERT3U(err, ==, ENOENT);
+ }
+
+ mutex_enter(&dd->dd_lock);
+ for (cbr = list_head(&dd->dd_prop_cbs); cbr;
+ cbr = list_next(&dd->dd_prop_cbs, cbr)) {
+ uint64_t propobj = cbr->cbr_ds->ds_phys->ds_props_obj;
+
+ if (strcmp(cbr->cbr_propname, propname) != 0)
+ continue;
+
+ /*
+ * If the property is set on this ds, then it is not
+ * inherited here; don't call the callback.
+ */
+ if (propobj && 0 == zap_lookup(mos, propobj, propname,
+ 8, 1, &dummyval))
+ continue;
+
+ cbr->cbr_func(cbr->cbr_arg, value);
+ }
+ mutex_exit(&dd->dd_lock);
+
+ za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
+ for (zap_cursor_init(&zc, mos,
+ dd->dd_phys->dd_child_dir_zapobj);
+ zap_cursor_retrieve(&zc, za) == 0;
+ zap_cursor_advance(&zc)) {
+ dsl_prop_changed_notify(dp, za->za_first_integer,
+ propname, value, FALSE);
+ }
+ kmem_free(za, sizeof (zap_attribute_t));
+ zap_cursor_fini(&zc);
+ dsl_dir_close(dd, FTAG);
+}
+
+struct prop_set_arg {
+ const char *name;
+ int intsz;
+ int numints;
+ const void *buf;
+};
+
+
+static void
+dsl_prop_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds = arg1;
+ struct prop_set_arg *psa = arg2;
+ objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
+ uint64_t zapobj, intval;
+ int isint;
+ char valbuf[32];
+ char *valstr;
+
+ isint = (dodefault(psa->name, 8, 1, &intval) == 0);
+
+ if (dsl_dataset_is_snapshot(ds)) {
+ ASSERT(spa_version(ds->ds_dir->dd_pool->dp_spa) >=
+ SPA_VERSION_SNAP_PROPS);
+ if (ds->ds_phys->ds_props_obj == 0) {
+ dmu_buf_will_dirty(ds->ds_dbuf, tx);
+ ds->ds_phys->ds_props_obj =
+ zap_create(mos,
+ DMU_OT_DSL_PROPS, DMU_OT_NONE, 0, tx);
+ }
+ zapobj = ds->ds_phys->ds_props_obj;
+ } else {
+ zapobj = ds->ds_dir->dd_phys->dd_props_zapobj;
+ }
+
+ if (psa->numints == 0) {
+ int err = zap_remove(mos, zapobj, psa->name, tx);
+ ASSERT(err == 0 || err == ENOENT);
+ if (isint) {
+ VERIFY(0 == dsl_prop_get_ds(ds,
+ psa->name, 8, 1, &intval, NULL));
+ }
+ } else {
+ VERIFY(0 == zap_update(mos, zapobj, psa->name,
+ psa->intsz, psa->numints, psa->buf, tx));
+ if (isint)
+ intval = *(uint64_t *)psa->buf;
+ }
+
+ if (isint) {
+ if (dsl_dataset_is_snapshot(ds)) {
+ dsl_prop_cb_record_t *cbr;
+ /*
+ * It's a snapshot; nothing can inherit this
+ * property, so just look for callbacks on this
+ * ds here.
+ */
+ mutex_enter(&ds->ds_dir->dd_lock);
+ for (cbr = list_head(&ds->ds_dir->dd_prop_cbs); cbr;
+ cbr = list_next(&ds->ds_dir->dd_prop_cbs, cbr)) {
+ if (cbr->cbr_ds == ds &&
+ strcmp(cbr->cbr_propname, psa->name) == 0)
+ cbr->cbr_func(cbr->cbr_arg, intval);
+ }
+ mutex_exit(&ds->ds_dir->dd_lock);
+ } else {
+ dsl_prop_changed_notify(ds->ds_dir->dd_pool,
+ ds->ds_dir->dd_object, psa->name, intval, TRUE);
+ }
+ }
+ if (isint) {
+ (void) snprintf(valbuf, sizeof (valbuf),
+ "%lld", (longlong_t)intval);
+ valstr = valbuf;
+ } else {
+ valstr = (char *)psa->buf;
+ }
+ spa_history_internal_log((psa->numints == 0) ? LOG_DS_INHERIT :
+ LOG_DS_PROPSET, ds->ds_dir->dd_pool->dp_spa, tx, cr,
+ "%s=%s dataset = %llu", psa->name, valstr, ds->ds_object);
+}
+
+void
+dsl_prop_set_uint64_sync(dsl_dir_t *dd, const char *name, uint64_t val,
+ cred_t *cr, dmu_tx_t *tx)
+{
+ objset_t *mos = dd->dd_pool->dp_meta_objset;
+ uint64_t zapobj = dd->dd_phys->dd_props_zapobj;
+
+ ASSERT(dmu_tx_is_syncing(tx));
+
+ VERIFY(0 == zap_update(mos, zapobj, name, sizeof (val), 1, &val, tx));
+
+ dsl_prop_changed_notify(dd->dd_pool, dd->dd_object, name, val, TRUE);
+
+ spa_history_internal_log(LOG_DS_PROPSET, dd->dd_pool->dp_spa, tx, cr,
+ "%s=%llu dataset = %llu", name, (u_longlong_t)val,
+ dd->dd_phys->dd_head_dataset_obj);
+}
+
+int
+dsl_prop_set(const char *dsname, const char *propname,
+ int intsz, int numints, const void *buf)
+{
+ dsl_dataset_t *ds;
+ int err;
+ struct prop_set_arg psa;
+
+ /*
+ * We must do these checks before we get to the syncfunc, since
+ * it can't fail.
+ */
+ if (strlen(propname) >= ZAP_MAXNAMELEN)
+ return (ENAMETOOLONG);
+ if (intsz * numints >= ZAP_MAXVALUELEN)
+ return (E2BIG);
+
+ err = dsl_dataset_hold(dsname, FTAG, &ds);
+ if (err)
+ return (err);
+
+ if (dsl_dataset_is_snapshot(ds) &&
+ spa_version(ds->ds_dir->dd_pool->dp_spa) < SPA_VERSION_SNAP_PROPS) {
+ dsl_dataset_rele(ds, FTAG);
+ return (ENOTSUP);
+ }
+
+ psa.name = propname;
+ psa.intsz = intsz;
+ psa.numints = numints;
+ psa.buf = buf;
+ err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+ NULL, dsl_prop_set_sync, ds, &psa, 2);
+
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+}
+
+/*
+ * Iterate over all properties for this dataset and return them in an nvlist.
+ */
+int
+dsl_prop_get_all(objset_t *os, nvlist_t **nvp, boolean_t local)
+{
+ dsl_dataset_t *ds = os->os->os_dsl_dataset;
+ dsl_dir_t *dd = ds->ds_dir;
+ boolean_t snapshot = dsl_dataset_is_snapshot(ds);
+ int err = 0;
+ dsl_pool_t *dp = dd->dd_pool;
+ objset_t *mos = dp->dp_meta_objset;
+ uint64_t propobj = ds->ds_phys->ds_props_obj;
+
+ VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ if (local && snapshot && !propobj)
+ return (0);
+
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ while (dd != NULL) {
+ char setpoint[MAXNAMELEN];
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ dsl_dir_t *dd_next;
+
+ if (propobj) {
+ dsl_dataset_name(ds, setpoint);
+ dd_next = dd;
+ } else {
+ dsl_dir_name(dd, setpoint);
+ propobj = dd->dd_phys->dd_props_zapobj;
+ dd_next = dd->dd_parent;
+ }
+
+ for (zap_cursor_init(&zc, mos, propobj);
+ (err = zap_cursor_retrieve(&zc, &za)) == 0;
+ zap_cursor_advance(&zc)) {
+ nvlist_t *propval;
+ zfs_prop_t prop = zfs_name_to_prop(za.za_name);
+
+ /* Skip non-inheritable properties. */
+ if (prop != ZPROP_INVAL &&
+ !zfs_prop_inheritable(prop) &&
+ (dd != ds->ds_dir || (snapshot && dd != dd_next)))
+ continue;
+
+ /* Skip properties not valid for this type. */
+ if (snapshot && prop != ZPROP_INVAL &&
+ !zfs_prop_valid_for_type(prop, ZFS_TYPE_SNAPSHOT))
+ continue;
+
+ /* Skip properties already defined */
+ if (nvlist_lookup_nvlist(*nvp, za.za_name,
+ &propval) == 0)
+ continue;
+
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME,
+ KM_SLEEP) == 0);
+ if (za.za_integer_length == 1) {
+ /*
+ * String property
+ */
+ char *tmp = kmem_alloc(za.za_num_integers,
+ KM_SLEEP);
+ err = zap_lookup(mos, propobj,
+ za.za_name, 1, za.za_num_integers, tmp);
+ if (err != 0) {
+ kmem_free(tmp, za.za_num_integers);
+ break;
+ }
+ VERIFY(nvlist_add_string(propval, ZPROP_VALUE,
+ tmp) == 0);
+ kmem_free(tmp, za.za_num_integers);
+ } else {
+ /*
+ * Integer property
+ */
+ ASSERT(za.za_integer_length == 8);
+ (void) nvlist_add_uint64(propval, ZPROP_VALUE,
+ za.za_first_integer);
+ }
+
+ VERIFY(nvlist_add_string(propval, ZPROP_SOURCE,
+ setpoint) == 0);
+ VERIFY(nvlist_add_nvlist(*nvp, za.za_name,
+ propval) == 0);
+ nvlist_free(propval);
+ }
+ zap_cursor_fini(&zc);
+
+ if (err != ENOENT)
+ break;
+ err = 0;
+ /*
+ * If we are just after the props that have been set
+ * locally, then we are done after the first iteration.
+ */
+ if (local)
+ break;
+ dd = dd_next;
+ propobj = 0;
+ }
+ rw_exit(&dp->dp_config_rwlock);
+
+ return (err);
+}
+
+void
+dsl_prop_nvlist_add_uint64(nvlist_t *nv, zfs_prop_t prop, uint64_t value)
+{
+ nvlist_t *propval;
+
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, value) == 0);
+ VERIFY(nvlist_add_nvlist(nv, zfs_prop_to_name(prop), propval) == 0);
+ nvlist_free(propval);
+}
+
+void
+dsl_prop_nvlist_add_string(nvlist_t *nv, zfs_prop_t prop, const char *value)
+{
+ nvlist_t *propval;
+
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_string(propval, ZPROP_VALUE, value) == 0);
+ VERIFY(nvlist_add_nvlist(nv, zfs_prop_to_name(prop), propval) == 0);
+ nvlist_free(propval);
+}
diff --git a/module/zfs/dsl_scrub.c b/module/zfs/dsl_scrub.c
new file mode 100644
index 000000000..950a91f78
--- /dev/null
+++ b/module/zfs/dsl_scrub.c
@@ -0,0 +1,1014 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/dsl_pool.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dnode.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/arc.h>
+#include <sys/zap.h>
+#include <sys/zio.h>
+#include <sys/zfs_context.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_znode.h>
+#include <sys/spa_impl.h>
+#include <sys/vdev_impl.h>
+#include <sys/zil_impl.h>
+
+typedef int (scrub_cb_t)(dsl_pool_t *, const blkptr_t *, const zbookmark_t *);
+
+static scrub_cb_t dsl_pool_scrub_clean_cb;
+static dsl_syncfunc_t dsl_pool_scrub_cancel_sync;
+
+int zfs_scrub_min_time = 1; /* scrub for at least 1 sec each txg */
+int zfs_resilver_min_time = 3; /* resilver for at least 3 sec each txg */
+boolean_t zfs_no_scrub_io = B_FALSE; /* set to disable scrub i/o */
+
+extern int zfs_txg_timeout;
+
+static scrub_cb_t *scrub_funcs[SCRUB_FUNC_NUMFUNCS] = {
+ NULL,
+ dsl_pool_scrub_clean_cb
+};
+
+#define SET_BOOKMARK(zb, objset, object, level, blkid) \
+{ \
+ (zb)->zb_objset = objset; \
+ (zb)->zb_object = object; \
+ (zb)->zb_level = level; \
+ (zb)->zb_blkid = blkid; \
+}
+
+/* ARGSUSED */
+static void
+dsl_pool_scrub_setup_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = arg1;
+ enum scrub_func *funcp = arg2;
+ dmu_object_type_t ot = 0;
+ boolean_t complete = B_FALSE;
+
+ dsl_pool_scrub_cancel_sync(dp, &complete, cr, tx);
+
+ ASSERT(dp->dp_scrub_func == SCRUB_FUNC_NONE);
+ ASSERT(*funcp > SCRUB_FUNC_NONE);
+ ASSERT(*funcp < SCRUB_FUNC_NUMFUNCS);
+
+ dp->dp_scrub_min_txg = 0;
+ dp->dp_scrub_max_txg = tx->tx_txg;
+
+ if (*funcp == SCRUB_FUNC_CLEAN) {
+ vdev_t *rvd = dp->dp_spa->spa_root_vdev;
+
+ /* rewrite all disk labels */
+ vdev_config_dirty(rvd);
+
+ if (vdev_resilver_needed(rvd,
+ &dp->dp_scrub_min_txg, &dp->dp_scrub_max_txg)) {
+ spa_event_notify(dp->dp_spa, NULL,
+ ESC_ZFS_RESILVER_START);
+ dp->dp_scrub_max_txg = MIN(dp->dp_scrub_max_txg,
+ tx->tx_txg);
+ }
+
+ /* zero out the scrub stats in all vdev_stat_t's */
+ vdev_scrub_stat_update(rvd,
+ dp->dp_scrub_min_txg ? POOL_SCRUB_RESILVER :
+ POOL_SCRUB_EVERYTHING, B_FALSE);
+
+ dp->dp_spa->spa_scrub_started = B_TRUE;
+ }
+
+ /* back to the generic stuff */
+
+ if (dp->dp_blkstats == NULL) {
+ dp->dp_blkstats =
+ kmem_alloc(sizeof (zfs_all_blkstats_t), KM_SLEEP);
+ }
+ bzero(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
+
+ if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB)
+ ot = DMU_OT_ZAP_OTHER;
+
+ dp->dp_scrub_func = *funcp;
+ dp->dp_scrub_queue_obj = zap_create(dp->dp_meta_objset,
+ ot ? ot : DMU_OT_SCRUB_QUEUE, DMU_OT_NONE, 0, tx);
+ bzero(&dp->dp_scrub_bookmark, sizeof (zbookmark_t));
+ dp->dp_scrub_restart = B_FALSE;
+ dp->dp_spa->spa_scrub_errors = 0;
+
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_FUNC, sizeof (uint32_t), 1,
+ &dp->dp_scrub_func, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_QUEUE, sizeof (uint64_t), 1,
+ &dp->dp_scrub_queue_obj, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MIN_TXG, sizeof (uint64_t), 1,
+ &dp->dp_scrub_min_txg, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MAX_TXG, sizeof (uint64_t), 1,
+ &dp->dp_scrub_max_txg, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_BOOKMARK, sizeof (uint64_t), 4,
+ &dp->dp_scrub_bookmark, tx));
+ VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_ERRORS, sizeof (uint64_t), 1,
+ &dp->dp_spa->spa_scrub_errors, tx));
+
+ spa_history_internal_log(LOG_POOL_SCRUB, dp->dp_spa, tx, cr,
+ "func=%u mintxg=%llu maxtxg=%llu",
+ *funcp, dp->dp_scrub_min_txg, dp->dp_scrub_max_txg);
+}
+
+int
+dsl_pool_scrub_setup(dsl_pool_t *dp, enum scrub_func func)
+{
+ return (dsl_sync_task_do(dp, NULL,
+ dsl_pool_scrub_setup_sync, dp, &func, 0));
+}
+
+/* ARGSUSED */
+static void
+dsl_pool_scrub_cancel_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = arg1;
+ boolean_t *completep = arg2;
+
+ if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
+ return;
+
+ mutex_enter(&dp->dp_scrub_cancel_lock);
+
+ if (dp->dp_scrub_restart) {
+ dp->dp_scrub_restart = B_FALSE;
+ *completep = B_FALSE;
+ }
+
+ /* XXX this is scrub-clean specific */
+ mutex_enter(&dp->dp_spa->spa_scrub_lock);
+ while (dp->dp_spa->spa_scrub_inflight > 0) {
+ cv_wait(&dp->dp_spa->spa_scrub_io_cv,
+ &dp->dp_spa->spa_scrub_lock);
+ }
+ mutex_exit(&dp->dp_spa->spa_scrub_lock);
+ dp->dp_spa->spa_scrub_started = B_FALSE;
+ dp->dp_spa->spa_scrub_active = B_FALSE;
+
+ dp->dp_scrub_func = SCRUB_FUNC_NONE;
+ VERIFY(0 == dmu_object_free(dp->dp_meta_objset,
+ dp->dp_scrub_queue_obj, tx));
+ dp->dp_scrub_queue_obj = 0;
+ bzero(&dp->dp_scrub_bookmark, sizeof (zbookmark_t));
+
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_QUEUE, tx));
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MIN_TXG, tx));
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_MAX_TXG, tx));
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_BOOKMARK, tx));
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_FUNC, tx));
+ VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_ERRORS, tx));
+
+ spa_history_internal_log(LOG_POOL_SCRUB_DONE, dp->dp_spa, tx, cr,
+ "complete=%u", *completep);
+
+ /* below is scrub-clean specific */
+ vdev_scrub_stat_update(dp->dp_spa->spa_root_vdev, POOL_SCRUB_NONE,
+ *completep);
+ /*
+ * If the scrub/resilver completed, update all DTLs to reflect this.
+ * Whether it succeeded or not, vacate all temporary scrub DTLs.
+ */
+ vdev_dtl_reassess(dp->dp_spa->spa_root_vdev, tx->tx_txg,
+ *completep ? dp->dp_scrub_max_txg : 0, B_TRUE);
+ if (dp->dp_scrub_min_txg && *completep)
+ spa_event_notify(dp->dp_spa, NULL, ESC_ZFS_RESILVER_FINISH);
+ spa_errlog_rotate(dp->dp_spa);
+
+ /*
+ * We may have finished replacing a device.
+ * Let the async thread assess this and handle the detach.
+ */
+ spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER_DONE);
+
+ dp->dp_scrub_min_txg = dp->dp_scrub_max_txg = 0;
+ mutex_exit(&dp->dp_scrub_cancel_lock);
+}
+
+int
+dsl_pool_scrub_cancel(dsl_pool_t *dp)
+{
+ boolean_t complete = B_FALSE;
+
+ return (dsl_sync_task_do(dp, NULL,
+ dsl_pool_scrub_cancel_sync, dp, &complete, 3));
+}
+
+int
+dsl_free(zio_t *pio, dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp,
+ zio_done_func_t *done, void *private, uint32_t arc_flags)
+{
+ /*
+ * This function will be used by bp-rewrite wad to intercept frees.
+ */
+ return (arc_free(pio, dp->dp_spa, txg, (blkptr_t *)bpp,
+ done, private, arc_flags));
+}
+
+static boolean_t
+bookmark_is_zero(const zbookmark_t *zb)
+{
+ return (zb->zb_objset == 0 && zb->zb_object == 0 &&
+ zb->zb_level == 0 && zb->zb_blkid == 0);
+}
+
+/* dnp is the dnode for zb1->zb_object */
+static boolean_t
+bookmark_is_before(dnode_phys_t *dnp, const zbookmark_t *zb1,
+ const zbookmark_t *zb2)
+{
+ uint64_t zb1nextL0, zb2thisobj;
+
+ ASSERT(zb1->zb_objset == zb2->zb_objset);
+ ASSERT(zb1->zb_object != -1ULL);
+ ASSERT(zb2->zb_level == 0);
+
+ /*
+ * A bookmark in the deadlist is considered to be after
+ * everything else.
+ */
+ if (zb2->zb_object == -1ULL)
+ return (B_TRUE);
+
+ /* The objset_phys_t isn't before anything. */
+ if (dnp == NULL)
+ return (B_FALSE);
+
+ zb1nextL0 = (zb1->zb_blkid + 1) <<
+ ((zb1->zb_level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT));
+
+ zb2thisobj = zb2->zb_object ? zb2->zb_object :
+ zb2->zb_blkid << (DNODE_BLOCK_SHIFT - DNODE_SHIFT);
+
+ if (zb1->zb_object == 0) {
+ uint64_t nextobj = zb1nextL0 *
+ (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT) >> DNODE_SHIFT;
+ return (nextobj <= zb2thisobj);
+ }
+
+ if (zb1->zb_object < zb2thisobj)
+ return (B_TRUE);
+ if (zb1->zb_object > zb2thisobj)
+ return (B_FALSE);
+ if (zb2->zb_object == 0)
+ return (B_FALSE);
+ return (zb1nextL0 <= zb2->zb_blkid);
+}
+
+static boolean_t
+scrub_pause(dsl_pool_t *dp, const zbookmark_t *zb)
+{
+ int elapsed_ticks;
+ int mintime;
+
+ if (dp->dp_scrub_pausing)
+ return (B_TRUE); /* we're already pausing */
+
+ if (!bookmark_is_zero(&dp->dp_scrub_bookmark))
+ return (B_FALSE); /* we're resuming */
+
+ /* We only know how to resume from level-0 blocks. */
+ if (zb->zb_level != 0)
+ return (B_FALSE);
+
+ mintime = dp->dp_scrub_isresilver ? zfs_resilver_min_time :
+ zfs_scrub_min_time;
+ elapsed_ticks = lbolt64 - dp->dp_scrub_start_time;
+ if (elapsed_ticks > hz * zfs_txg_timeout ||
+ (elapsed_ticks > hz * mintime && txg_sync_waiting(dp))) {
+ dprintf("pausing at %llx/%llx/%llx/%llx\n",
+ (longlong_t)zb->zb_objset, (longlong_t)zb->zb_object,
+ (longlong_t)zb->zb_level, (longlong_t)zb->zb_blkid);
+ dp->dp_scrub_pausing = B_TRUE;
+ dp->dp_scrub_bookmark = *zb;
+ return (B_TRUE);
+ }
+ return (B_FALSE);
+}
+
+typedef struct zil_traverse_arg {
+ dsl_pool_t *zta_dp;
+ zil_header_t *zta_zh;
+} zil_traverse_arg_t;
+
+/* ARGSUSED */
+static void
+traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
+{
+ zil_traverse_arg_t *zta = arg;
+ dsl_pool_t *dp = zta->zta_dp;
+ zil_header_t *zh = zta->zta_zh;
+ zbookmark_t zb;
+
+ if (bp->blk_birth <= dp->dp_scrub_min_txg)
+ return;
+
+ if (claim_txg == 0 && bp->blk_birth >= spa_first_txg(dp->dp_spa))
+ return;
+
+ zb.zb_objset = zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET];
+ zb.zb_object = 0;
+ zb.zb_level = -1;
+ zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ];
+ VERIFY(0 == scrub_funcs[dp->dp_scrub_func](dp, bp, &zb));
+}
+
+/* ARGSUSED */
+static void
+traverse_zil_record(zilog_t *zilog, lr_t *lrc, void *arg, uint64_t claim_txg)
+{
+ if (lrc->lrc_txtype == TX_WRITE) {
+ zil_traverse_arg_t *zta = arg;
+ dsl_pool_t *dp = zta->zta_dp;
+ zil_header_t *zh = zta->zta_zh;
+ lr_write_t *lr = (lr_write_t *)lrc;
+ blkptr_t *bp = &lr->lr_blkptr;
+ zbookmark_t zb;
+
+ if (bp->blk_birth <= dp->dp_scrub_min_txg)
+ return;
+
+ if (claim_txg == 0 || bp->blk_birth < claim_txg)
+ return;
+
+ zb.zb_objset = zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET];
+ zb.zb_object = lr->lr_foid;
+ zb.zb_level = BP_GET_LEVEL(bp);
+ zb.zb_blkid = lr->lr_offset / BP_GET_LSIZE(bp);
+ VERIFY(0 == scrub_funcs[dp->dp_scrub_func](dp, bp, &zb));
+ }
+}
+
+static void
+traverse_zil(dsl_pool_t *dp, zil_header_t *zh)
+{
+ uint64_t claim_txg = zh->zh_claim_txg;
+ zil_traverse_arg_t zta = { dp, zh };
+ zilog_t *zilog;
+
+ /*
+ * We only want to visit blocks that have been claimed but not yet
+ * replayed (or, in read-only mode, blocks that *would* be claimed).
+ */
+ if (claim_txg == 0 && (spa_mode & FWRITE))
+ return;
+
+ zilog = zil_alloc(dp->dp_meta_objset, zh);
+
+ (void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, &zta,
+ claim_txg);
+
+ zil_free(zilog);
+}
+
+static void
+scrub_visitbp(dsl_pool_t *dp, dnode_phys_t *dnp,
+ arc_buf_t *pbuf, blkptr_t *bp, const zbookmark_t *zb)
+{
+ int err;
+ arc_buf_t *buf = NULL;
+
+ if (bp->blk_birth == 0)
+ return;
+
+ if (bp->blk_birth <= dp->dp_scrub_min_txg)
+ return;
+
+ if (scrub_pause(dp, zb))
+ return;
+
+ if (!bookmark_is_zero(&dp->dp_scrub_bookmark)) {
+ /*
+ * If we already visited this bp & everything below (in
+ * a prior txg), don't bother doing it again.
+ */
+ if (bookmark_is_before(dnp, zb, &dp->dp_scrub_bookmark))
+ return;
+
+ /*
+ * If we found the block we're trying to resume from, or
+ * we went past it to a different object, zero it out to
+ * indicate that it's OK to start checking for pausing
+ * again.
+ */
+ if (bcmp(zb, &dp->dp_scrub_bookmark, sizeof (*zb)) == 0 ||
+ zb->zb_object > dp->dp_scrub_bookmark.zb_object) {
+ dprintf("resuming at %llx/%llx/%llx/%llx\n",
+ (longlong_t)zb->zb_objset,
+ (longlong_t)zb->zb_object,
+ (longlong_t)zb->zb_level,
+ (longlong_t)zb->zb_blkid);
+ bzero(&dp->dp_scrub_bookmark, sizeof (*zb));
+ }
+ }
+
+ if (BP_GET_LEVEL(bp) > 0) {
+ uint32_t flags = ARC_WAIT;
+ int i;
+ blkptr_t *cbp;
+ int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
+
+ err = arc_read(NULL, dp->dp_spa, bp, pbuf,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err) {
+ mutex_enter(&dp->dp_spa->spa_scrub_lock);
+ dp->dp_spa->spa_scrub_errors++;
+ mutex_exit(&dp->dp_spa->spa_scrub_lock);
+ return;
+ }
+ cbp = buf->b_data;
+
+ for (i = 0; i < epb; i++, cbp++) {
+ zbookmark_t czb;
+
+ SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
+ zb->zb_level - 1,
+ zb->zb_blkid * epb + i);
+ scrub_visitbp(dp, dnp, buf, cbp, &czb);
+ }
+ } else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
+ uint32_t flags = ARC_WAIT;
+ dnode_phys_t *child_dnp;
+ int i, j;
+ int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
+
+ err = arc_read(NULL, dp->dp_spa, bp, pbuf,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err) {
+ mutex_enter(&dp->dp_spa->spa_scrub_lock);
+ dp->dp_spa->spa_scrub_errors++;
+ mutex_exit(&dp->dp_spa->spa_scrub_lock);
+ return;
+ }
+ child_dnp = buf->b_data;
+
+ for (i = 0; i < epb; i++, child_dnp++) {
+ for (j = 0; j < child_dnp->dn_nblkptr; j++) {
+ zbookmark_t czb;
+
+ SET_BOOKMARK(&czb, zb->zb_objset,
+ zb->zb_blkid * epb + i,
+ child_dnp->dn_nlevels - 1, j);
+ scrub_visitbp(dp, child_dnp, buf,
+ &child_dnp->dn_blkptr[j], &czb);
+ }
+ }
+ } else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
+ uint32_t flags = ARC_WAIT;
+ objset_phys_t *osp;
+ int j;
+
+ err = arc_read_nolock(NULL, dp->dp_spa, bp,
+ arc_getbuf_func, &buf,
+ ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
+ if (err) {
+ mutex_enter(&dp->dp_spa->spa_scrub_lock);
+ dp->dp_spa->spa_scrub_errors++;
+ mutex_exit(&dp->dp_spa->spa_scrub_lock);
+ return;
+ }
+
+ osp = buf->b_data;
+
+ traverse_zil(dp, &osp->os_zil_header);
+
+ for (j = 0; j < osp->os_meta_dnode.dn_nblkptr; j++) {
+ zbookmark_t czb;
+
+ SET_BOOKMARK(&czb, zb->zb_objset, 0,
+ osp->os_meta_dnode.dn_nlevels - 1, j);
+ scrub_visitbp(dp, &osp->os_meta_dnode, buf,
+ &osp->os_meta_dnode.dn_blkptr[j], &czb);
+ }
+ }
+
+ (void) scrub_funcs[dp->dp_scrub_func](dp, bp, zb);
+ if (buf)
+ (void) arc_buf_remove_ref(buf, &buf);
+}
+
+static void
+scrub_visit_rootbp(dsl_pool_t *dp, dsl_dataset_t *ds, blkptr_t *bp)
+{
+ zbookmark_t zb;
+
+ SET_BOOKMARK(&zb, ds ? ds->ds_object : 0, 0, -1, 0);
+ scrub_visitbp(dp, NULL, NULL, bp, &zb);
+}
+
+void
+dsl_pool_ds_destroyed(dsl_dataset_t *ds, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
+ return;
+
+ if (dp->dp_scrub_bookmark.zb_objset == ds->ds_object) {
+ SET_BOOKMARK(&dp->dp_scrub_bookmark, -1, 0, 0, 0);
+ } else if (zap_remove_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_object, tx) != 0) {
+ return;
+ }
+
+ if (ds->ds_phys->ds_next_snap_obj != 0) {
+ VERIFY(zap_add_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_phys->ds_next_snap_obj, tx) == 0);
+ }
+ ASSERT3U(ds->ds_phys->ds_num_children, <=, 1);
+}
+
+void
+dsl_pool_ds_snapshotted(dsl_dataset_t *ds, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = ds->ds_dir->dd_pool;
+
+ if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
+ return;
+
+ ASSERT(ds->ds_phys->ds_prev_snap_obj != 0);
+
+ if (dp->dp_scrub_bookmark.zb_objset == ds->ds_object) {
+ dp->dp_scrub_bookmark.zb_objset =
+ ds->ds_phys->ds_prev_snap_obj;
+ } else if (zap_remove_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_object, tx) == 0) {
+ VERIFY(zap_add_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_phys->ds_prev_snap_obj, tx) == 0);
+ }
+}
+
+void
+dsl_pool_ds_clone_swapped(dsl_dataset_t *ds1, dsl_dataset_t *ds2, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = ds1->ds_dir->dd_pool;
+
+ if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
+ return;
+
+ if (dp->dp_scrub_bookmark.zb_objset == ds1->ds_object) {
+ dp->dp_scrub_bookmark.zb_objset = ds2->ds_object;
+ } else if (dp->dp_scrub_bookmark.zb_objset == ds2->ds_object) {
+ dp->dp_scrub_bookmark.zb_objset = ds1->ds_object;
+ }
+
+ if (zap_remove_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds1->ds_object, tx) == 0) {
+ int err = zap_add_int(dp->dp_meta_objset,
+ dp->dp_scrub_queue_obj, ds2->ds_object, tx);
+ VERIFY(err == 0 || err == EEXIST);
+ if (err == EEXIST) {
+ /* Both were there to begin with */
+ VERIFY(0 == zap_add_int(dp->dp_meta_objset,
+ dp->dp_scrub_queue_obj, ds1->ds_object, tx));
+ }
+ } else if (zap_remove_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds2->ds_object, tx) == 0) {
+ VERIFY(0 == zap_add_int(dp->dp_meta_objset,
+ dp->dp_scrub_queue_obj, ds1->ds_object, tx));
+ }
+}
+
+struct enqueue_clones_arg {
+ dmu_tx_t *tx;
+ uint64_t originobj;
+};
+
+/* ARGSUSED */
+static int
+enqueue_clones_cb(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg)
+{
+ struct enqueue_clones_arg *eca = arg;
+ dsl_dataset_t *ds;
+ int err;
+ dsl_pool_t *dp;
+
+ err = dsl_dataset_hold_obj(spa->spa_dsl_pool, dsobj, FTAG, &ds);
+ if (err)
+ return (err);
+ dp = ds->ds_dir->dd_pool;
+
+ if (ds->ds_dir->dd_phys->dd_origin_obj == eca->originobj) {
+ while (ds->ds_phys->ds_prev_snap_obj != eca->originobj) {
+ dsl_dataset_t *prev;
+ err = dsl_dataset_hold_obj(dp,
+ ds->ds_phys->ds_prev_snap_obj, FTAG, &prev);
+
+ dsl_dataset_rele(ds, FTAG);
+ if (err)
+ return (err);
+ ds = prev;
+ }
+ VERIFY(zap_add_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_object, eca->tx) == 0);
+ }
+ dsl_dataset_rele(ds, FTAG);
+ return (0);
+}
+
+static void
+scrub_visitds(dsl_pool_t *dp, uint64_t dsobj, dmu_tx_t *tx)
+{
+ dsl_dataset_t *ds;
+ uint64_t min_txg_save;
+
+ VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
+
+ /*
+ * Iterate over the bps in this ds.
+ */
+ min_txg_save = dp->dp_scrub_min_txg;
+ dp->dp_scrub_min_txg =
+ MAX(dp->dp_scrub_min_txg, ds->ds_phys->ds_prev_snap_txg);
+ scrub_visit_rootbp(dp, ds, &ds->ds_phys->ds_bp);
+ dp->dp_scrub_min_txg = min_txg_save;
+
+ if (dp->dp_scrub_pausing)
+ goto out;
+
+ /*
+ * Add descendent datasets to work queue.
+ */
+ if (ds->ds_phys->ds_next_snap_obj != 0) {
+ VERIFY(zap_add_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_phys->ds_next_snap_obj, tx) == 0);
+ }
+ if (ds->ds_phys->ds_num_children > 1) {
+ if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB) {
+ struct enqueue_clones_arg eca;
+ eca.tx = tx;
+ eca.originobj = ds->ds_object;
+
+ (void) dmu_objset_find_spa(ds->ds_dir->dd_pool->dp_spa,
+ NULL, enqueue_clones_cb, &eca, DS_FIND_CHILDREN);
+ } else {
+ VERIFY(zap_join(dp->dp_meta_objset,
+ ds->ds_phys->ds_next_clones_obj,
+ dp->dp_scrub_queue_obj, tx) == 0);
+ }
+ }
+
+out:
+ dsl_dataset_rele(ds, FTAG);
+}
+
+/* ARGSUSED */
+static int
+enqueue_cb(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg)
+{
+ dmu_tx_t *tx = arg;
+ dsl_dataset_t *ds;
+ int err;
+ dsl_pool_t *dp;
+
+ err = dsl_dataset_hold_obj(spa->spa_dsl_pool, dsobj, FTAG, &ds);
+ if (err)
+ return (err);
+
+ dp = ds->ds_dir->dd_pool;
+
+ while (ds->ds_phys->ds_prev_snap_obj != 0) {
+ dsl_dataset_t *prev;
+ err = dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
+ FTAG, &prev);
+ if (err) {
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
+ }
+
+ /*
+ * If this is a clone, we don't need to worry about it for now.
+ */
+ if (prev->ds_phys->ds_next_snap_obj != ds->ds_object) {
+ dsl_dataset_rele(ds, FTAG);
+ dsl_dataset_rele(prev, FTAG);
+ return (0);
+ }
+ dsl_dataset_rele(ds, FTAG);
+ ds = prev;
+ }
+
+ VERIFY(zap_add_int(dp->dp_meta_objset, dp->dp_scrub_queue_obj,
+ ds->ds_object, tx) == 0);
+ dsl_dataset_rele(ds, FTAG);
+ return (0);
+}
+
+void
+dsl_pool_scrub_sync(dsl_pool_t *dp, dmu_tx_t *tx)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ boolean_t complete = B_TRUE;
+
+ if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
+ return;
+
+ /* If the spa is not fully loaded, don't bother. */
+ if (dp->dp_spa->spa_load_state != SPA_LOAD_NONE)
+ return;
+
+ if (dp->dp_scrub_restart) {
+ enum scrub_func func = dp->dp_scrub_func;
+ dp->dp_scrub_restart = B_FALSE;
+ dsl_pool_scrub_setup_sync(dp, &func, kcred, tx);
+ }
+
+ if (dp->dp_spa->spa_root_vdev->vdev_stat.vs_scrub_type == 0) {
+ /*
+ * We must have resumed after rebooting; reset the vdev
+ * stats to know that we're doing a scrub (although it
+ * will think we're just starting now).
+ */
+ vdev_scrub_stat_update(dp->dp_spa->spa_root_vdev,
+ dp->dp_scrub_min_txg ? POOL_SCRUB_RESILVER :
+ POOL_SCRUB_EVERYTHING, B_FALSE);
+ }
+
+ dp->dp_scrub_pausing = B_FALSE;
+ dp->dp_scrub_start_time = lbolt64;
+ dp->dp_scrub_isresilver = (dp->dp_scrub_min_txg != 0);
+ dp->dp_spa->spa_scrub_active = B_TRUE;
+
+ if (dp->dp_scrub_bookmark.zb_objset == 0) {
+ /* First do the MOS & ORIGIN */
+ scrub_visit_rootbp(dp, NULL, &dp->dp_meta_rootbp);
+ if (dp->dp_scrub_pausing)
+ goto out;
+
+ if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB) {
+ VERIFY(0 == dmu_objset_find_spa(dp->dp_spa,
+ NULL, enqueue_cb, tx, DS_FIND_CHILDREN));
+ } else {
+ scrub_visitds(dp, dp->dp_origin_snap->ds_object, tx);
+ }
+ ASSERT(!dp->dp_scrub_pausing);
+ } else if (dp->dp_scrub_bookmark.zb_objset != -1ULL) {
+ /*
+ * If we were paused, continue from here. Note if the
+ * ds we were paused on was deleted, the zb_objset will
+ * be -1, so we will skip this and find a new objset
+ * below.
+ */
+ scrub_visitds(dp, dp->dp_scrub_bookmark.zb_objset, tx);
+ if (dp->dp_scrub_pausing)
+ goto out;
+ }
+
+ /*
+ * In case we were paused right at the end of the ds, zero the
+ * bookmark so we don't think that we're still trying to resume.
+ */
+ bzero(&dp->dp_scrub_bookmark, sizeof (zbookmark_t));
+
+ /* keep pulling things out of the zap-object-as-queue */
+ while (zap_cursor_init(&zc, dp->dp_meta_objset, dp->dp_scrub_queue_obj),
+ zap_cursor_retrieve(&zc, &za) == 0) {
+ VERIFY(0 == zap_remove(dp->dp_meta_objset,
+ dp->dp_scrub_queue_obj, za.za_name, tx));
+ scrub_visitds(dp, za.za_first_integer, tx);
+ if (dp->dp_scrub_pausing)
+ break;
+ zap_cursor_fini(&zc);
+ }
+ zap_cursor_fini(&zc);
+ if (dp->dp_scrub_pausing)
+ goto out;
+
+ /* done. */
+
+ dsl_pool_scrub_cancel_sync(dp, &complete, kcred, tx);
+ return;
+out:
+ VERIFY(0 == zap_update(dp->dp_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_BOOKMARK, sizeof (uint64_t), 4,
+ &dp->dp_scrub_bookmark, tx));
+ VERIFY(0 == zap_update(dp->dp_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SCRUB_ERRORS, sizeof (uint64_t), 1,
+ &dp->dp_spa->spa_scrub_errors, tx));
+
+ /* XXX this is scrub-clean specific */
+ mutex_enter(&dp->dp_spa->spa_scrub_lock);
+ while (dp->dp_spa->spa_scrub_inflight > 0) {
+ cv_wait(&dp->dp_spa->spa_scrub_io_cv,
+ &dp->dp_spa->spa_scrub_lock);
+ }
+ mutex_exit(&dp->dp_spa->spa_scrub_lock);
+}
+
+void
+dsl_pool_scrub_restart(dsl_pool_t *dp)
+{
+ mutex_enter(&dp->dp_scrub_cancel_lock);
+ dp->dp_scrub_restart = B_TRUE;
+ mutex_exit(&dp->dp_scrub_cancel_lock);
+}
+
+/*
+ * scrub consumers
+ */
+
+static void
+count_block(zfs_all_blkstats_t *zab, const blkptr_t *bp)
+{
+ int i;
+
+ /*
+ * If we resume after a reboot, zab will be NULL; don't record
+ * incomplete stats in that case.
+ */
+ if (zab == NULL)
+ return;
+
+ for (i = 0; i < 4; i++) {
+ int l = (i < 2) ? BP_GET_LEVEL(bp) : DN_MAX_LEVELS;
+ int t = (i & 1) ? BP_GET_TYPE(bp) : DMU_OT_TOTAL;
+ zfs_blkstat_t *zb = &zab->zab_type[l][t];
+ int equal;
+
+ zb->zb_count++;
+ zb->zb_asize += BP_GET_ASIZE(bp);
+ zb->zb_lsize += BP_GET_LSIZE(bp);
+ zb->zb_psize += BP_GET_PSIZE(bp);
+ zb->zb_gangs += BP_COUNT_GANG(bp);
+
+ switch (BP_GET_NDVAS(bp)) {
+ case 2:
+ if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
+ DVA_GET_VDEV(&bp->blk_dva[1]))
+ zb->zb_ditto_2_of_2_samevdev++;
+ break;
+ case 3:
+ equal = (DVA_GET_VDEV(&bp->blk_dva[0]) ==
+ DVA_GET_VDEV(&bp->blk_dva[1])) +
+ (DVA_GET_VDEV(&bp->blk_dva[0]) ==
+ DVA_GET_VDEV(&bp->blk_dva[2])) +
+ (DVA_GET_VDEV(&bp->blk_dva[1]) ==
+ DVA_GET_VDEV(&bp->blk_dva[2]));
+ if (equal == 1)
+ zb->zb_ditto_2_of_3_samevdev++;
+ else if (equal == 3)
+ zb->zb_ditto_3_of_3_samevdev++;
+ break;
+ }
+ }
+}
+
+static void
+dsl_pool_scrub_clean_done(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+
+ zio_data_buf_free(zio->io_data, zio->io_size);
+
+ mutex_enter(&spa->spa_scrub_lock);
+ spa->spa_scrub_inflight--;
+ cv_broadcast(&spa->spa_scrub_io_cv);
+
+ if (zio->io_error && (zio->io_error != ECKSUM ||
+ !(zio->io_flags & ZIO_FLAG_SPECULATIVE)))
+ spa->spa_scrub_errors++;
+ mutex_exit(&spa->spa_scrub_lock);
+}
+
+static int
+dsl_pool_scrub_clean_cb(dsl_pool_t *dp,
+ const blkptr_t *bp, const zbookmark_t *zb)
+{
+ size_t size = BP_GET_LSIZE(bp);
+ int d;
+ spa_t *spa = dp->dp_spa;
+ boolean_t needs_io;
+ int zio_flags = ZIO_FLAG_SCRUB_THREAD | ZIO_FLAG_CANFAIL;
+ int zio_priority;
+
+ count_block(dp->dp_blkstats, bp);
+
+ if (dp->dp_scrub_isresilver == 0) {
+ /* It's a scrub */
+ zio_flags |= ZIO_FLAG_SCRUB;
+ zio_priority = ZIO_PRIORITY_SCRUB;
+ needs_io = B_TRUE;
+ } else {
+ /* It's a resilver */
+ zio_flags |= ZIO_FLAG_RESILVER;
+ zio_priority = ZIO_PRIORITY_RESILVER;
+ needs_io = B_FALSE;
+ }
+
+ /* If it's an intent log block, failure is expected. */
+ if (zb->zb_level == -1 && BP_GET_TYPE(bp) != DMU_OT_OBJSET)
+ zio_flags |= ZIO_FLAG_SPECULATIVE;
+
+ for (d = 0; d < BP_GET_NDVAS(bp); d++) {
+ vdev_t *vd = vdev_lookup_top(spa,
+ DVA_GET_VDEV(&bp->blk_dva[d]));
+
+ /*
+ * Keep track of how much data we've examined so that
+ * zpool(1M) status can make useful progress reports.
+ */
+ mutex_enter(&vd->vdev_stat_lock);
+ vd->vdev_stat.vs_scrub_examined +=
+ DVA_GET_ASIZE(&bp->blk_dva[d]);
+ mutex_exit(&vd->vdev_stat_lock);
+
+ /* if it's a resilver, this may not be in the target range */
+ if (!needs_io) {
+ if (DVA_GET_GANG(&bp->blk_dva[d])) {
+ /*
+ * Gang members may be spread across multiple
+ * vdevs, so the best we can do is look at the
+ * pool-wide DTL.
+ * XXX -- it would be better to change our
+ * allocation policy to ensure that this can't
+ * happen.
+ */
+ vd = spa->spa_root_vdev;
+ }
+ needs_io = vdev_dtl_contains(&vd->vdev_dtl_map,
+ bp->blk_birth, 1);
+ }
+ }
+
+ if (needs_io && !zfs_no_scrub_io) {
+ void *data = zio_data_buf_alloc(size);
+
+ mutex_enter(&spa->spa_scrub_lock);
+ while (spa->spa_scrub_inflight >= spa->spa_scrub_maxinflight)
+ cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
+ spa->spa_scrub_inflight++;
+ mutex_exit(&spa->spa_scrub_lock);
+
+ zio_nowait(zio_read(NULL, spa, bp, data, size,
+ dsl_pool_scrub_clean_done, NULL, zio_priority,
+ zio_flags, zb));
+ }
+
+ /* do not relocate this block */
+ return (0);
+}
+
+int
+dsl_pool_scrub_clean(dsl_pool_t *dp)
+{
+ /*
+ * Purge all vdev caches. We do this here rather than in sync
+ * context because this requires a writer lock on the spa_config
+ * lock, which we can't do from sync context. The
+ * spa_scrub_reopen flag indicates that vdev_open() should not
+ * attempt to start another scrub.
+ */
+ spa_config_enter(dp->dp_spa, SCL_ALL, FTAG, RW_WRITER);
+ dp->dp_spa->spa_scrub_reopen = B_TRUE;
+ vdev_reopen(dp->dp_spa->spa_root_vdev);
+ dp->dp_spa->spa_scrub_reopen = B_FALSE;
+ spa_config_exit(dp->dp_spa, SCL_ALL, FTAG);
+
+ return (dsl_pool_scrub_setup(dp, SCRUB_FUNC_CLEAN));
+}
diff --git a/module/zfs/dsl_synctask.c b/module/zfs/dsl_synctask.c
new file mode 100644
index 000000000..21100225a
--- /dev/null
+++ b/module/zfs/dsl_synctask.c
@@ -0,0 +1,225 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_synctask.h>
+#include <sys/cred.h>
+
+#define DST_AVG_BLKSHIFT 14
+
+/* ARGSUSED */
+static int
+dsl_null_checkfunc(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ return (0);
+}
+
+dsl_sync_task_group_t *
+dsl_sync_task_group_create(dsl_pool_t *dp)
+{
+ dsl_sync_task_group_t *dstg;
+
+ dstg = kmem_zalloc(sizeof (dsl_sync_task_group_t), KM_SLEEP);
+ list_create(&dstg->dstg_tasks, sizeof (dsl_sync_task_t),
+ offsetof(dsl_sync_task_t, dst_node));
+ dstg->dstg_pool = dp;
+ dstg->dstg_cr = CRED();
+
+ return (dstg);
+}
+
+void
+dsl_sync_task_create(dsl_sync_task_group_t *dstg,
+ dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
+ void *arg1, void *arg2, int blocks_modified)
+{
+ dsl_sync_task_t *dst;
+
+ if (checkfunc == NULL)
+ checkfunc = dsl_null_checkfunc;
+ dst = kmem_zalloc(sizeof (dsl_sync_task_t), KM_SLEEP);
+ dst->dst_checkfunc = checkfunc;
+ dst->dst_syncfunc = syncfunc;
+ dst->dst_arg1 = arg1;
+ dst->dst_arg2 = arg2;
+ list_insert_tail(&dstg->dstg_tasks, dst);
+
+ dstg->dstg_space += blocks_modified << DST_AVG_BLKSHIFT;
+}
+
+int
+dsl_sync_task_group_wait(dsl_sync_task_group_t *dstg)
+{
+ dmu_tx_t *tx;
+ uint64_t txg;
+ dsl_sync_task_t *dst;
+
+top:
+ tx = dmu_tx_create_dd(dstg->dstg_pool->dp_mos_dir);
+ VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));
+
+ txg = dmu_tx_get_txg(tx);
+
+ /* Do a preliminary error check. */
+ dstg->dstg_err = 0;
+ rw_enter(&dstg->dstg_pool->dp_config_rwlock, RW_READER);
+ for (dst = list_head(&dstg->dstg_tasks); dst;
+ dst = list_next(&dstg->dstg_tasks, dst)) {
+#ifdef ZFS_DEBUG
+ /*
+ * Only check half the time, otherwise, the sync-context
+ * check will almost never fail.
+ */
+ if (spa_get_random(2) == 0)
+ continue;
+#endif
+ dst->dst_err =
+ dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
+ if (dst->dst_err)
+ dstg->dstg_err = dst->dst_err;
+ }
+ rw_exit(&dstg->dstg_pool->dp_config_rwlock);
+
+ if (dstg->dstg_err) {
+ dmu_tx_commit(tx);
+ return (dstg->dstg_err);
+ }
+
+ VERIFY(0 == txg_list_add(&dstg->dstg_pool->dp_sync_tasks, dstg, txg));
+
+ dmu_tx_commit(tx);
+
+ txg_wait_synced(dstg->dstg_pool, txg);
+
+ if (dstg->dstg_err == EAGAIN)
+ goto top;
+
+ return (dstg->dstg_err);
+}
+
+void
+dsl_sync_task_group_nowait(dsl_sync_task_group_t *dstg, dmu_tx_t *tx)
+{
+ uint64_t txg;
+
+ dstg->dstg_nowaiter = B_TRUE;
+ txg = dmu_tx_get_txg(tx);
+ VERIFY(0 == txg_list_add(&dstg->dstg_pool->dp_sync_tasks, dstg, txg));
+}
+
+void
+dsl_sync_task_group_destroy(dsl_sync_task_group_t *dstg)
+{
+ dsl_sync_task_t *dst;
+
+ while (dst = list_head(&dstg->dstg_tasks)) {
+ list_remove(&dstg->dstg_tasks, dst);
+ kmem_free(dst, sizeof (dsl_sync_task_t));
+ }
+ kmem_free(dstg, sizeof (dsl_sync_task_group_t));
+}
+
+void
+dsl_sync_task_group_sync(dsl_sync_task_group_t *dstg, dmu_tx_t *tx)
+{
+ dsl_sync_task_t *dst;
+ void *tr_cookie;
+
+ ASSERT3U(dstg->dstg_err, ==, 0);
+
+ /*
+ * Check for sufficient space.
+ */
+ dstg->dstg_err = dsl_dir_tempreserve_space(dstg->dstg_pool->dp_mos_dir,
+ dstg->dstg_space, dstg->dstg_space * 3, 0, 0, &tr_cookie, tx);
+ /* don't bother trying again */
+ if (dstg->dstg_err == ERESTART)
+ dstg->dstg_err = EAGAIN;
+ if (dstg->dstg_err)
+ return;
+
+ /*
+ * Check for errors by calling checkfuncs.
+ */
+ rw_enter(&dstg->dstg_pool->dp_config_rwlock, RW_WRITER);
+ for (dst = list_head(&dstg->dstg_tasks); dst;
+ dst = list_next(&dstg->dstg_tasks, dst)) {
+ dst->dst_err =
+ dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
+ if (dst->dst_err)
+ dstg->dstg_err = dst->dst_err;
+ }
+
+ if (dstg->dstg_err == 0) {
+ /*
+ * Execute sync tasks.
+ */
+ for (dst = list_head(&dstg->dstg_tasks); dst;
+ dst = list_next(&dstg->dstg_tasks, dst)) {
+ dst->dst_syncfunc(dst->dst_arg1, dst->dst_arg2,
+ dstg->dstg_cr, tx);
+ }
+ }
+ rw_exit(&dstg->dstg_pool->dp_config_rwlock);
+
+ dsl_dir_tempreserve_clear(tr_cookie, tx);
+
+ if (dstg->dstg_nowaiter)
+ dsl_sync_task_group_destroy(dstg);
+}
+
+int
+dsl_sync_task_do(dsl_pool_t *dp,
+ dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
+ void *arg1, void *arg2, int blocks_modified)
+{
+ dsl_sync_task_group_t *dstg;
+ int err;
+
+ dstg = dsl_sync_task_group_create(dp);
+ dsl_sync_task_create(dstg, checkfunc, syncfunc,
+ arg1, arg2, blocks_modified);
+ err = dsl_sync_task_group_wait(dstg);
+ dsl_sync_task_group_destroy(dstg);
+ return (err);
+}
+
+void
+dsl_sync_task_do_nowait(dsl_pool_t *dp,
+ dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
+ void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx)
+{
+ dsl_sync_task_group_t *dstg;
+
+ dstg = dsl_sync_task_group_create(dp);
+ dsl_sync_task_create(dstg, checkfunc, syncfunc,
+ arg1, arg2, blocks_modified);
+ dsl_sync_task_group_nowait(dstg, tx);
+}
diff --git a/module/zfs/fletcher.c b/module/zfs/fletcher.c
new file mode 100644
index 000000000..edda3c9a9
--- /dev/null
+++ b/module/zfs/fletcher.c
@@ -0,0 +1,145 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/types.h>
+#include <sys/sysmacros.h>
+#include <sys/byteorder.h>
+#include <sys/spa.h>
+
+void
+fletcher_2_native(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ const uint64_t *ip = buf;
+ const uint64_t *ipend = ip + (size / sizeof (uint64_t));
+ uint64_t a0, b0, a1, b1;
+
+ for (a0 = b0 = a1 = b1 = 0; ip < ipend; ip += 2) {
+ a0 += ip[0];
+ a1 += ip[1];
+ b0 += a0;
+ b1 += a1;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1);
+}
+
+void
+fletcher_2_byteswap(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ const uint64_t *ip = buf;
+ const uint64_t *ipend = ip + (size / sizeof (uint64_t));
+ uint64_t a0, b0, a1, b1;
+
+ for (a0 = b0 = a1 = b1 = 0; ip < ipend; ip += 2) {
+ a0 += BSWAP_64(ip[0]);
+ a1 += BSWAP_64(ip[1]);
+ b0 += a0;
+ b1 += a1;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1);
+}
+
+void
+fletcher_4_native(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ const uint32_t *ip = buf;
+ const uint32_t *ipend = ip + (size / sizeof (uint32_t));
+ uint64_t a, b, c, d;
+
+ for (a = b = c = d = 0; ip < ipend; ip++) {
+ a += ip[0];
+ b += a;
+ c += b;
+ d += c;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a, b, c, d);
+}
+
+void
+fletcher_4_byteswap(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ const uint32_t *ip = buf;
+ const uint32_t *ipend = ip + (size / sizeof (uint32_t));
+ uint64_t a, b, c, d;
+
+ for (a = b = c = d = 0; ip < ipend; ip++) {
+ a += BSWAP_32(ip[0]);
+ b += a;
+ c += b;
+ d += c;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a, b, c, d);
+}
+
+void
+fletcher_4_incremental_native(const void *buf, uint64_t size,
+ zio_cksum_t *zcp)
+{
+ const uint32_t *ip = buf;
+ const uint32_t *ipend = ip + (size / sizeof (uint32_t));
+ uint64_t a, b, c, d;
+
+ a = zcp->zc_word[0];
+ b = zcp->zc_word[1];
+ c = zcp->zc_word[2];
+ d = zcp->zc_word[3];
+
+ for (; ip < ipend; ip++) {
+ a += ip[0];
+ b += a;
+ c += b;
+ d += c;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a, b, c, d);
+}
+
+void
+fletcher_4_incremental_byteswap(const void *buf, uint64_t size,
+ zio_cksum_t *zcp)
+{
+ const uint32_t *ip = buf;
+ const uint32_t *ipend = ip + (size / sizeof (uint32_t));
+ uint64_t a, b, c, d;
+
+ a = zcp->zc_word[0];
+ b = zcp->zc_word[1];
+ c = zcp->zc_word[2];
+ d = zcp->zc_word[3];
+
+ for (; ip < ipend; ip++) {
+ a += BSWAP_32(ip[0]);
+ b += a;
+ c += b;
+ d += c;
+ }
+
+ ZIO_SET_CHECKSUM(zcp, a, b, c, d);
+}
diff --git a/module/zfs/gzip.c b/module/zfs/gzip.c
new file mode 100644
index 000000000..b257d4af7
--- /dev/null
+++ b/module/zfs/gzip.c
@@ -0,0 +1,69 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/debug.h>
+#include <sys/types.h>
+#include <sys/zmod.h>
+
+#ifdef _KERNEL
+#include <sys/systm.h>
+#else
+#include <strings.h>
+#endif
+
+size_t
+gzip_compress(void *s_start, void *d_start, size_t s_len, size_t d_len, int n)
+{
+ size_t dstlen = d_len;
+
+ ASSERT(d_len <= s_len);
+
+ if (z_compress_level(d_start, &dstlen, s_start, s_len, n) != Z_OK) {
+ if (d_len != s_len)
+ return (s_len);
+
+ bcopy(s_start, d_start, s_len);
+ return (s_len);
+ }
+
+ return (dstlen);
+}
+
+/*ARGSUSED*/
+int
+gzip_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len, int n)
+{
+ size_t dstlen = d_len;
+
+ ASSERT(d_len >= s_len);
+
+ if (z_uncompress(d_start, &dstlen, s_start, s_len) != Z_OK)
+ return (-1);
+
+ return (0);
+}
diff --git a/module/zfs/include/sys/arc.h b/module/zfs/include/sys/arc.h
new file mode 100644
index 000000000..749bf53e5
--- /dev/null
+++ b/module/zfs/include/sys/arc.h
@@ -0,0 +1,138 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ARC_H
+#define _SYS_ARC_H
+
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <sys/zio.h>
+#include <sys/dmu.h>
+#include <sys/spa.h>
+
+typedef struct arc_buf_hdr arc_buf_hdr_t;
+typedef struct arc_buf arc_buf_t;
+typedef void arc_done_func_t(zio_t *zio, arc_buf_t *buf, void *private);
+typedef int arc_evict_func_t(void *private);
+
+/* generic arc_done_func_t's which you can use */
+arc_done_func_t arc_bcopy_func;
+arc_done_func_t arc_getbuf_func;
+
+struct arc_buf {
+ arc_buf_hdr_t *b_hdr;
+ arc_buf_t *b_next;
+ krwlock_t b_lock;
+ void *b_data;
+ arc_evict_func_t *b_efunc;
+ void *b_private;
+};
+
+typedef enum arc_buf_contents {
+ ARC_BUFC_DATA, /* buffer contains data */
+ ARC_BUFC_METADATA, /* buffer contains metadata */
+ ARC_BUFC_NUMTYPES
+} arc_buf_contents_t;
+/*
+ * These are the flags we pass into calls to the arc
+ */
+#define ARC_WAIT (1 << 1) /* perform I/O synchronously */
+#define ARC_NOWAIT (1 << 2) /* perform I/O asynchronously */
+#define ARC_PREFETCH (1 << 3) /* I/O is a prefetch */
+#define ARC_CACHED (1 << 4) /* I/O was already in cache */
+#define ARC_L2CACHE (1 << 5) /* cache in L2ARC */
+
+void arc_space_consume(uint64_t space);
+void arc_space_return(uint64_t space);
+void *arc_data_buf_alloc(uint64_t space);
+void arc_data_buf_free(void *buf, uint64_t space);
+arc_buf_t *arc_buf_alloc(spa_t *spa, int size, void *tag,
+ arc_buf_contents_t type);
+void arc_buf_add_ref(arc_buf_t *buf, void *tag);
+int arc_buf_remove_ref(arc_buf_t *buf, void *tag);
+int arc_buf_size(arc_buf_t *buf);
+void arc_release(arc_buf_t *buf, void *tag);
+int arc_released(arc_buf_t *buf);
+int arc_has_callback(arc_buf_t *buf);
+void arc_buf_freeze(arc_buf_t *buf);
+void arc_buf_thaw(arc_buf_t *buf);
+#ifdef ZFS_DEBUG
+int arc_referenced(arc_buf_t *buf);
+#endif
+
+typedef struct writeprops {
+ dmu_object_type_t wp_type;
+ uint8_t wp_level;
+ uint8_t wp_copies;
+ uint8_t wp_dncompress, wp_oscompress;
+ uint8_t wp_dnchecksum, wp_oschecksum;
+} writeprops_t;
+
+void write_policy(spa_t *spa, const writeprops_t *wp, zio_prop_t *zp);
+int arc_read(zio_t *pio, spa_t *spa, blkptr_t *bp, arc_buf_t *pbuf,
+ arc_done_func_t *done, void *private, int priority, int zio_flags,
+ uint32_t *arc_flags, const zbookmark_t *zb);
+int arc_read_nolock(zio_t *pio, spa_t *spa, blkptr_t *bp,
+ arc_done_func_t *done, void *private, int priority, int flags,
+ uint32_t *arc_flags, const zbookmark_t *zb);
+zio_t *arc_write(zio_t *pio, spa_t *spa, const writeprops_t *wp,
+ boolean_t l2arc, uint64_t txg, blkptr_t *bp, arc_buf_t *buf,
+ arc_done_func_t *ready, arc_done_func_t *done, void *private, int priority,
+ int zio_flags, const zbookmark_t *zb);
+int arc_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, uint32_t arc_flags);
+int arc_tryread(spa_t *spa, blkptr_t *bp, void *data);
+
+void arc_set_callback(arc_buf_t *buf, arc_evict_func_t *func, void *private);
+int arc_buf_evict(arc_buf_t *buf);
+
+void arc_flush(spa_t *spa);
+void arc_tempreserve_clear(uint64_t reserve);
+int arc_tempreserve_space(uint64_t reserve, uint64_t txg);
+
+void arc_init(void);
+void arc_fini(void);
+
+/*
+ * Level 2 ARC
+ */
+
+void l2arc_add_vdev(spa_t *spa, vdev_t *vd, uint64_t start, uint64_t end);
+void l2arc_remove_vdev(vdev_t *vd);
+boolean_t l2arc_vdev_present(vdev_t *vd);
+void l2arc_init(void);
+void l2arc_fini(void);
+void l2arc_start(void);
+void l2arc_stop(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ARC_H */
diff --git a/module/zfs/include/sys/bplist.h b/module/zfs/include/sys/bplist.h
new file mode 100644
index 000000000..cdb93a6c3
--- /dev/null
+++ b/module/zfs/include/sys/bplist.h
@@ -0,0 +1,89 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_BPLIST_H
+#define _SYS_BPLIST_H
+
+#include <sys/dmu.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct bplist_phys {
+ /*
+ * This is the bonus buffer for the dead lists. The object's
+ * contents is an array of bpl_entries blkptr_t's, representing
+ * a total of bpl_bytes physical space.
+ */
+ uint64_t bpl_entries;
+ uint64_t bpl_bytes;
+ uint64_t bpl_comp;
+ uint64_t bpl_uncomp;
+} bplist_phys_t;
+
+#define BPLIST_SIZE_V0 (2 * sizeof (uint64_t))
+
+typedef struct bplist_q {
+ blkptr_t bpq_blk;
+ void *bpq_next;
+} bplist_q_t;
+
+typedef struct bplist {
+ kmutex_t bpl_lock;
+ objset_t *bpl_mos;
+ uint64_t bpl_object;
+ uint8_t bpl_blockshift;
+ uint8_t bpl_bpshift;
+ uint8_t bpl_havecomp;
+ bplist_q_t *bpl_queue;
+ bplist_phys_t *bpl_phys;
+ dmu_buf_t *bpl_dbuf;
+ dmu_buf_t *bpl_cached_dbuf;
+} bplist_t;
+
+extern uint64_t bplist_create(objset_t *mos, int blocksize, dmu_tx_t *tx);
+extern void bplist_destroy(objset_t *mos, uint64_t object, dmu_tx_t *tx);
+extern int bplist_open(bplist_t *bpl, objset_t *mos, uint64_t object);
+extern void bplist_close(bplist_t *bpl);
+extern boolean_t bplist_empty(bplist_t *bpl);
+extern int bplist_iterate(bplist_t *bpl, uint64_t *itorp, blkptr_t *bp);
+extern int bplist_enqueue(bplist_t *bpl, const blkptr_t *bp, dmu_tx_t *tx);
+extern void bplist_enqueue_deferred(bplist_t *bpl, const blkptr_t *bp);
+extern void bplist_sync(bplist_t *bpl, dmu_tx_t *tx);
+extern void bplist_vacate(bplist_t *bpl, dmu_tx_t *tx);
+extern int bplist_space(bplist_t *bpl,
+ uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
+extern int bplist_space_birthrange(bplist_t *bpl,
+ uint64_t mintxg, uint64_t maxtxg, uint64_t *dasizep);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_BPLIST_H */
diff --git a/module/zfs/include/sys/dbuf.h b/module/zfs/include/sys/dbuf.h
new file mode 100644
index 000000000..75ce27264
--- /dev/null
+++ b/module/zfs/include/sys/dbuf.h
@@ -0,0 +1,347 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DBUF_H
+#define _SYS_DBUF_H
+
+#include <sys/dmu.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/zio.h>
+#include <sys/arc.h>
+#include <sys/zfs_context.h>
+#include <sys/refcount.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define DB_BONUS_BLKID (-1ULL)
+#define IN_DMU_SYNC 2
+
+/*
+ * define flags for dbuf_read
+ */
+
+#define DB_RF_MUST_SUCCEED (1 << 0)
+#define DB_RF_CANFAIL (1 << 1)
+#define DB_RF_HAVESTRUCT (1 << 2)
+#define DB_RF_NOPREFETCH (1 << 3)
+#define DB_RF_NEVERWAIT (1 << 4)
+#define DB_RF_CACHED (1 << 5)
+
+/*
+ * The simplified state transition diagram for dbufs looks like:
+ *
+ * +----> READ ----+
+ * | |
+ * | V
+ * (alloc)-->UNCACHED CACHED-->EVICTING-->(free)
+ * | ^ ^
+ * | | |
+ * +----> FILL ----+ |
+ * | |
+ * | |
+ * +--------> NOFILL -------+
+ */
+typedef enum dbuf_states {
+ DB_UNCACHED,
+ DB_FILL,
+ DB_NOFILL,
+ DB_READ,
+ DB_CACHED,
+ DB_EVICTING
+} dbuf_states_t;
+
+struct objset_impl;
+struct dnode;
+struct dmu_tx;
+
+/*
+ * level = 0 means the user data
+ * level = 1 means the single indirect block
+ * etc.
+ */
+
+#define LIST_LINK_INACTIVE(link) \
+ ((link)->list_next == NULL && (link)->list_prev == NULL)
+
+struct dmu_buf_impl;
+
+typedef enum override_states {
+ DR_NOT_OVERRIDDEN,
+ DR_IN_DMU_SYNC,
+ DR_OVERRIDDEN
+} override_states_t;
+
+typedef struct dbuf_dirty_record {
+ /* link on our parents dirty list */
+ list_node_t dr_dirty_node;
+
+ /* transaction group this data will sync in */
+ uint64_t dr_txg;
+
+ /* zio of outstanding write IO */
+ zio_t *dr_zio;
+
+ /* pointer back to our dbuf */
+ struct dmu_buf_impl *dr_dbuf;
+
+ /* pointer to next dirty record */
+ struct dbuf_dirty_record *dr_next;
+
+ /* pointer to parent dirty record */
+ struct dbuf_dirty_record *dr_parent;
+
+ union dirty_types {
+ struct dirty_indirect {
+
+ /* protect access to list */
+ kmutex_t dr_mtx;
+
+ /* Our list of dirty children */
+ list_t dr_children;
+ } di;
+ struct dirty_leaf {
+
+ /*
+ * dr_data is set when we dirty the buffer
+ * so that we can retain the pointer even if it
+ * gets COW'd in a subsequent transaction group.
+ */
+ arc_buf_t *dr_data;
+ blkptr_t dr_overridden_by;
+ override_states_t dr_override_state;
+ } dl;
+ } dt;
+} dbuf_dirty_record_t;
+
+typedef struct dmu_buf_impl {
+ /*
+ * The following members are immutable, with the exception of
+ * db.db_data, which is protected by db_mtx.
+ */
+
+ /* the publicly visible structure */
+ dmu_buf_t db;
+
+ /* the objset we belong to */
+ struct objset_impl *db_objset;
+
+ /*
+ * the dnode we belong to (NULL when evicted)
+ */
+ struct dnode *db_dnode;
+
+ /*
+ * our parent buffer; if the dnode points to us directly,
+ * db_parent == db_dnode->dn_dbuf
+ * only accessed by sync thread ???
+ * (NULL when evicted)
+ */
+ struct dmu_buf_impl *db_parent;
+
+ /*
+ * link for hash table of all dmu_buf_impl_t's
+ */
+ struct dmu_buf_impl *db_hash_next;
+
+ /* our block number */
+ uint64_t db_blkid;
+
+ /*
+ * Pointer to the blkptr_t which points to us. May be NULL if we
+ * don't have one yet. (NULL when evicted)
+ */
+ blkptr_t *db_blkptr;
+
+ /*
+ * Our indirection level. Data buffers have db_level==0.
+ * Indirect buffers which point to data buffers have
+ * db_level==1. etc. Buffers which contain dnodes have
+ * db_level==0, since the dnodes are stored in a file.
+ */
+ uint8_t db_level;
+
+ /* db_mtx protects the members below */
+ kmutex_t db_mtx;
+
+ /*
+ * Current state of the buffer
+ */
+ dbuf_states_t db_state;
+
+ /*
+ * Refcount accessed by dmu_buf_{hold,rele}.
+ * If nonzero, the buffer can't be destroyed.
+ * Protected by db_mtx.
+ */
+ refcount_t db_holds;
+
+ /* buffer holding our data */
+ arc_buf_t *db_buf;
+
+ kcondvar_t db_changed;
+ dbuf_dirty_record_t *db_data_pending;
+
+ /* pointer to most recent dirty record for this buffer */
+ dbuf_dirty_record_t *db_last_dirty;
+
+ /*
+ * Our link on the owner dnodes's dn_dbufs list.
+ * Protected by its dn_dbufs_mtx.
+ */
+ list_node_t db_link;
+
+ /* Data which is unique to data (leaf) blocks: */
+
+ /* stuff we store for the user (see dmu_buf_set_user) */
+ void *db_user_ptr;
+ void **db_user_data_ptr_ptr;
+ dmu_buf_evict_func_t *db_evict_func;
+
+ uint8_t db_immediate_evict;
+ uint8_t db_freed_in_flight;
+
+ uint8_t db_dirtycnt;
+} dmu_buf_impl_t;
+
+/* Note: the dbuf hash table is exposed only for the mdb module */
+#define DBUF_MUTEXES 256
+#define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)])
+typedef struct dbuf_hash_table {
+ uint64_t hash_table_mask;
+ dmu_buf_impl_t **hash_table;
+ kmutex_t hash_mutexes[DBUF_MUTEXES];
+} dbuf_hash_table_t;
+
+
+uint64_t dbuf_whichblock(struct dnode *di, uint64_t offset);
+
+dmu_buf_impl_t *dbuf_create_tlib(struct dnode *dn, char *data);
+void dbuf_create_bonus(struct dnode *dn);
+
+dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid, void *tag);
+dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid,
+ void *tag);
+int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid, int create,
+ void *tag, dmu_buf_impl_t **dbp);
+
+void dbuf_prefetch(struct dnode *dn, uint64_t blkid);
+
+void dbuf_add_ref(dmu_buf_impl_t *db, void *tag);
+uint64_t dbuf_refcount(dmu_buf_impl_t *db);
+
+void dbuf_rele(dmu_buf_impl_t *db, void *tag);
+
+dmu_buf_impl_t *dbuf_find(struct dnode *dn, uint8_t level, uint64_t blkid);
+
+int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags);
+void dbuf_will_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
+void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx);
+void dmu_buf_will_not_fill(dmu_buf_t *db, dmu_tx_t *tx);
+void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx);
+void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx);
+dbuf_dirty_record_t *dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
+
+void dbuf_clear(dmu_buf_impl_t *db);
+void dbuf_evict(dmu_buf_impl_t *db);
+
+void dbuf_setdirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
+void dbuf_unoverride(dbuf_dirty_record_t *dr);
+void dbuf_sync_list(list_t *list, dmu_tx_t *tx);
+
+void dbuf_free_range(struct dnode *dn, uint64_t start, uint64_t end,
+ struct dmu_tx *);
+
+void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx);
+
+void dbuf_init(void);
+void dbuf_fini(void);
+
+#define DBUF_IS_METADATA(db) \
+ ((db)->db_level > 0 || dmu_ot[(db)->db_dnode->dn_type].ot_metadata)
+
+#define DBUF_GET_BUFC_TYPE(db) \
+ (DBUF_IS_METADATA(db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
+
+#define DBUF_IS_CACHEABLE(db) \
+ ((db)->db_objset->os_primary_cache == ZFS_CACHE_ALL || \
+ (DBUF_IS_METADATA(db) && \
+ ((db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
+
+#define DBUF_IS_L2CACHEABLE(db) \
+ ((db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL || \
+ (DBUF_IS_METADATA(db) && \
+ ((db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
+
+#ifdef ZFS_DEBUG
+
+/*
+ * There should be a ## between the string literal and fmt, to make it
+ * clear that we're joining two strings together, but gcc does not
+ * support that preprocessor token.
+ */
+#define dprintf_dbuf(dbuf, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char __db_buf[32]; \
+ uint64_t __db_obj = (dbuf)->db.db_object; \
+ if (__db_obj == DMU_META_DNODE_OBJECT) \
+ (void) strcpy(__db_buf, "mdn"); \
+ else \
+ (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
+ (u_longlong_t)__db_obj); \
+ dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
+ "obj=%s lvl=%u blkid=%lld " fmt, \
+ __db_buf, (dbuf)->db_level, \
+ (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+
+#define dprintf_dbuf_bp(db, bp, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
+ sprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \
+ dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \
+ kmem_free(__blkbuf, BP_SPRINTF_LEN); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+
+#define DBUF_VERIFY(db) dbuf_verify(db)
+
+#else
+
+#define dprintf_dbuf(db, fmt, ...)
+#define dprintf_dbuf_bp(db, bp, fmt, ...)
+#define DBUF_VERIFY(db)
+
+#endif
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DBUF_H */
diff --git a/module/zfs/include/sys/dmu.h b/module/zfs/include/sys/dmu.h
new file mode 100644
index 000000000..3b1e5c8fb
--- /dev/null
+++ b/module/zfs/include/sys/dmu.h
@@ -0,0 +1,638 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DMU_H
+#define _SYS_DMU_H
+
+/*
+ * This file describes the interface that the DMU provides for its
+ * consumers.
+ *
+ * The DMU also interacts with the SPA. That interface is described in
+ * dmu_spa.h.
+ */
+
+#include <sys/inttypes.h>
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/cred.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct uio;
+struct page;
+struct vnode;
+struct spa;
+struct zilog;
+struct zio;
+struct blkptr;
+struct zap_cursor;
+struct dsl_dataset;
+struct dsl_pool;
+struct dnode;
+struct drr_begin;
+struct drr_end;
+struct zbookmark;
+struct spa;
+struct nvlist;
+struct objset_impl;
+
+typedef struct objset objset_t;
+typedef struct dmu_tx dmu_tx_t;
+typedef struct dsl_dir dsl_dir_t;
+
+typedef enum dmu_object_type {
+ DMU_OT_NONE,
+ /* general: */
+ DMU_OT_OBJECT_DIRECTORY, /* ZAP */
+ DMU_OT_OBJECT_ARRAY, /* UINT64 */
+ DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
+ DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
+ DMU_OT_BPLIST, /* UINT64 */
+ DMU_OT_BPLIST_HDR, /* UINT64 */
+ /* spa: */
+ DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
+ DMU_OT_SPACE_MAP, /* UINT64 */
+ /* zil: */
+ DMU_OT_INTENT_LOG, /* UINT64 */
+ /* dmu: */
+ DMU_OT_DNODE, /* DNODE */
+ DMU_OT_OBJSET, /* OBJSET */
+ /* dsl: */
+ DMU_OT_DSL_DIR, /* UINT64 */
+ DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
+ DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
+ DMU_OT_DSL_PROPS, /* ZAP */
+ DMU_OT_DSL_DATASET, /* UINT64 */
+ /* zpl: */
+ DMU_OT_ZNODE, /* ZNODE */
+ DMU_OT_OLDACL, /* Old ACL */
+ DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
+ DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
+ DMU_OT_MASTER_NODE, /* ZAP */
+ DMU_OT_UNLINKED_SET, /* ZAP */
+ /* zvol: */
+ DMU_OT_ZVOL, /* UINT8 */
+ DMU_OT_ZVOL_PROP, /* ZAP */
+ /* other; for testing only! */
+ DMU_OT_PLAIN_OTHER, /* UINT8 */
+ DMU_OT_UINT64_OTHER, /* UINT64 */
+ DMU_OT_ZAP_OTHER, /* ZAP */
+ /* new object types: */
+ DMU_OT_ERROR_LOG, /* ZAP */
+ DMU_OT_SPA_HISTORY, /* UINT8 */
+ DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
+ DMU_OT_POOL_PROPS, /* ZAP */
+ DMU_OT_DSL_PERMS, /* ZAP */
+ DMU_OT_ACL, /* ACL */
+ DMU_OT_SYSACL, /* SYSACL */
+ DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
+ DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
+ DMU_OT_NEXT_CLONES, /* ZAP */
+ DMU_OT_SCRUB_QUEUE, /* ZAP */
+ DMU_OT_NUMTYPES
+} dmu_object_type_t;
+
+typedef enum dmu_objset_type {
+ DMU_OST_NONE,
+ DMU_OST_META,
+ DMU_OST_ZFS,
+ DMU_OST_ZVOL,
+ DMU_OST_OTHER, /* For testing only! */
+ DMU_OST_ANY, /* Be careful! */
+ DMU_OST_NUMTYPES
+} dmu_objset_type_t;
+
+void byteswap_uint64_array(void *buf, size_t size);
+void byteswap_uint32_array(void *buf, size_t size);
+void byteswap_uint16_array(void *buf, size_t size);
+void byteswap_uint8_array(void *buf, size_t size);
+void zap_byteswap(void *buf, size_t size);
+void zfs_oldacl_byteswap(void *buf, size_t size);
+void zfs_acl_byteswap(void *buf, size_t size);
+void zfs_znode_byteswap(void *buf, size_t size);
+
+#define DS_MODE_NOHOLD 0 /* internal use only */
+#define DS_MODE_USER 1 /* simple access, no special needs */
+#define DS_MODE_OWNER 2 /* the "main" access, e.g. a mount */
+#define DS_MODE_TYPE_MASK 0x3
+#define DS_MODE_TYPE(x) ((x) & DS_MODE_TYPE_MASK)
+#define DS_MODE_READONLY 0x8
+#define DS_MODE_IS_READONLY(x) ((x) & DS_MODE_READONLY)
+#define DS_MODE_INCONSISTENT 0x10
+#define DS_MODE_IS_INCONSISTENT(x) ((x) & DS_MODE_INCONSISTENT)
+
+#define DS_FIND_SNAPSHOTS (1<<0)
+#define DS_FIND_CHILDREN (1<<1)
+
+/*
+ * The maximum number of bytes that can be accessed as part of one
+ * operation, including metadata.
+ */
+#define DMU_MAX_ACCESS (10<<20) /* 10MB */
+#define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
+
+/*
+ * Public routines to create, destroy, open, and close objsets.
+ */
+int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
+ objset_t **osp);
+int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
+ objset_t **osp);
+void dmu_objset_close(objset_t *os);
+int dmu_objset_evict_dbufs(objset_t *os);
+int dmu_objset_create(const char *name, dmu_objset_type_t type,
+ objset_t *clone_parent, uint64_t flags,
+ void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
+int dmu_objset_destroy(const char *name);
+int dmu_snapshots_destroy(char *fsname, char *snapname);
+int dmu_objset_rollback(objset_t *os);
+int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
+int dmu_objset_rename(const char *name, const char *newname,
+ boolean_t recursive);
+int dmu_objset_find(char *name, int func(char *, void *), void *arg,
+ int flags);
+void dmu_objset_byteswap(void *buf, size_t size);
+
+typedef struct dmu_buf {
+ uint64_t db_object; /* object that this buffer is part of */
+ uint64_t db_offset; /* byte offset in this object */
+ uint64_t db_size; /* size of buffer in bytes */
+ void *db_data; /* data in buffer */
+} dmu_buf_t;
+
+typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
+
+/*
+ * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
+ */
+#define DMU_POOL_DIRECTORY_OBJECT 1
+#define DMU_POOL_CONFIG "config"
+#define DMU_POOL_ROOT_DATASET "root_dataset"
+#define DMU_POOL_SYNC_BPLIST "sync_bplist"
+#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
+#define DMU_POOL_ERRLOG_LAST "errlog_last"
+#define DMU_POOL_SPARES "spares"
+#define DMU_POOL_DEFLATE "deflate"
+#define DMU_POOL_HISTORY "history"
+#define DMU_POOL_PROPS "pool_props"
+#define DMU_POOL_L2CACHE "l2cache"
+
+/* 4x8 zbookmark_t */
+#define DMU_POOL_SCRUB_BOOKMARK "scrub_bookmark"
+/* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
+#define DMU_POOL_SCRUB_QUEUE "scrub_queue"
+/* 1x8 txg */
+#define DMU_POOL_SCRUB_MIN_TXG "scrub_min_txg"
+/* 1x8 txg */
+#define DMU_POOL_SCRUB_MAX_TXG "scrub_max_txg"
+/* 1x4 enum scrub_func */
+#define DMU_POOL_SCRUB_FUNC "scrub_func"
+/* 1x8 count */
+#define DMU_POOL_SCRUB_ERRORS "scrub_errors"
+
+/*
+ * Allocate an object from this objset. The range of object numbers
+ * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
+ *
+ * The transaction must be assigned to a txg. The newly allocated
+ * object will be "held" in the transaction (ie. you can modify the
+ * newly allocated object in this transaction).
+ *
+ * dmu_object_alloc() chooses an object and returns it in *objectp.
+ *
+ * dmu_object_claim() allocates a specific object number. If that
+ * number is already allocated, it fails and returns EEXIST.
+ *
+ * Return 0 on success, or ENOSPC or EEXIST as specified above.
+ */
+uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
+ int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
+int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
+ int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
+int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
+ int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+
+/*
+ * Free an object from this objset.
+ *
+ * The object's data will be freed as well (ie. you don't need to call
+ * dmu_free(object, 0, -1, tx)).
+ *
+ * The object need not be held in the transaction.
+ *
+ * If there are any holds on this object's buffers (via dmu_buf_hold()),
+ * or tx holds on the object (via dmu_tx_hold_object()), you can not
+ * free it; it fails and returns EBUSY.
+ *
+ * If the object is not allocated, it fails and returns ENOENT.
+ *
+ * Return 0 on success, or EBUSY or ENOENT as specified above.
+ */
+int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
+
+/*
+ * Find the next allocated or free object.
+ *
+ * The objectp parameter is in-out. It will be updated to be the next
+ * object which is allocated. Ignore objects which have not been
+ * modified since txg.
+ *
+ * XXX Can only be called on a objset with no dirty data.
+ *
+ * Returns 0 on success, or ENOENT if there are no more objects.
+ */
+int dmu_object_next(objset_t *os, uint64_t *objectp,
+ boolean_t hole, uint64_t txg);
+
+/*
+ * Set the data blocksize for an object.
+ *
+ * The object cannot have any blocks allcated beyond the first. If
+ * the first block is allocated already, the new size must be greater
+ * than the current block size. If these conditions are not met,
+ * ENOTSUP will be returned.
+ *
+ * Returns 0 on success, or EBUSY if there are any holds on the object
+ * contents, or ENOTSUP as described above.
+ */
+int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
+ int ibs, dmu_tx_t *tx);
+
+/*
+ * Set the checksum property on a dnode. The new checksum algorithm will
+ * apply to all newly written blocks; existing blocks will not be affected.
+ */
+void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
+ dmu_tx_t *tx);
+
+/*
+ * Set the compress property on a dnode. The new compression algorithm will
+ * apply to all newly written blocks; existing blocks will not be affected.
+ */
+void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
+ dmu_tx_t *tx);
+
+/*
+ * Decide how many copies of a given block we should make. Can be from
+ * 1 to SPA_DVAS_PER_BP.
+ */
+int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
+ dmu_object_type_t ot);
+/*
+ * The bonus data is accessed more or less like a regular buffer.
+ * You must dmu_bonus_hold() to get the buffer, which will give you a
+ * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
+ * data. As with any normal buffer, you must call dmu_buf_read() to
+ * read db_data, dmu_buf_will_dirty() before modifying it, and the
+ * object must be held in an assigned transaction before calling
+ * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
+ * buffer as well. You must release your hold with dmu_buf_rele().
+ */
+int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
+int dmu_bonus_max(void);
+int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
+
+/*
+ * Obtain the DMU buffer from the specified object which contains the
+ * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
+ * that it will remain in memory. You must release the hold with
+ * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
+ * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
+ *
+ * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
+ * on the returned buffer before reading or writing the buffer's
+ * db_data. The comments for those routines describe what particular
+ * operations are valid after calling them.
+ *
+ * The object number must be a valid, allocated object number.
+ */
+int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
+ void *tag, dmu_buf_t **);
+void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
+void dmu_buf_rele(dmu_buf_t *db, void *tag);
+uint64_t dmu_buf_refcount(dmu_buf_t *db);
+
+/*
+ * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
+ * range of an object. A pointer to an array of dmu_buf_t*'s is
+ * returned (in *dbpp).
+ *
+ * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
+ * frees the array. The hold on the array of buffers MUST be released
+ * with dmu_buf_rele_array. You can NOT release the hold on each buffer
+ * individually with dmu_buf_rele.
+ */
+int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
+ uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
+void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
+
+/*
+ * Returns NULL on success, or the existing user ptr if it's already
+ * been set.
+ *
+ * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
+ *
+ * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
+ * will be set to db->db_data when you are allowed to access it. Note
+ * that db->db_data (the pointer) can change when you do dmu_buf_read(),
+ * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
+ * *user_data_ptr_ptr will be set to the new value when it changes.
+ *
+ * If non-NULL, pageout func will be called when this buffer is being
+ * excised from the cache, so that you can clean up the data structure
+ * pointed to by user_ptr.
+ *
+ * dmu_evict_user() will call the pageout func for all buffers in a
+ * objset with a given pageout func.
+ */
+void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
+ dmu_buf_evict_func_t *pageout_func);
+/*
+ * set_user_ie is the same as set_user, but request immediate eviction
+ * when hold count goes to zero.
+ */
+void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
+ void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
+void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
+ void *user_ptr, void *user_data_ptr_ptr,
+ dmu_buf_evict_func_t *pageout_func);
+void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
+
+/*
+ * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
+ */
+void *dmu_buf_get_user(dmu_buf_t *db);
+
+/*
+ * Indicate that you are going to modify the buffer's data (db_data).
+ *
+ * The transaction (tx) must be assigned to a txg (ie. you've called
+ * dmu_tx_assign()). The buffer's object must be held in the tx
+ * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
+ */
+void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
+
+/*
+ * You must create a transaction, then hold the objects which you will
+ * (or might) modify as part of this transaction. Then you must assign
+ * the transaction to a transaction group. Once the transaction has
+ * been assigned, you can modify buffers which belong to held objects as
+ * part of this transaction. You can't modify buffers before the
+ * transaction has been assigned; you can't modify buffers which don't
+ * belong to objects which this transaction holds; you can't hold
+ * objects once the transaction has been assigned. You may hold an
+ * object which you are going to free (with dmu_object_free()), but you
+ * don't have to.
+ *
+ * You can abort the transaction before it has been assigned.
+ *
+ * Note that you may hold buffers (with dmu_buf_hold) at any time,
+ * regardless of transaction state.
+ */
+
+#define DMU_NEW_OBJECT (-1ULL)
+#define DMU_OBJECT_END (-1ULL)
+
+dmu_tx_t *dmu_tx_create(objset_t *os);
+void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
+void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
+ uint64_t len);
+void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
+void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
+void dmu_tx_abort(dmu_tx_t *tx);
+int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
+void dmu_tx_wait(dmu_tx_t *tx);
+void dmu_tx_commit(dmu_tx_t *tx);
+
+/*
+ * Free up the data blocks for a defined range of a file. If size is
+ * zero, the range from offset to end-of-file is freed.
+ */
+int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t size, dmu_tx_t *tx);
+int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t size);
+int dmu_free_object(objset_t *os, uint64_t object);
+
+/*
+ * Convenience functions.
+ *
+ * Canfail routines will return 0 on success, or an errno if there is a
+ * nonrecoverable I/O error.
+ */
+int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ void *buf);
+void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ const void *buf, dmu_tx_t *tx);
+void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ dmu_tx_t *tx);
+int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
+int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
+ dmu_tx_t *tx);
+int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t size, struct page *pp, dmu_tx_t *tx);
+
+extern int zfs_prefetch_disable;
+
+/*
+ * Asynchronously try to read in the data.
+ */
+void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
+ uint64_t len);
+
+typedef struct dmu_object_info {
+ /* All sizes are in bytes. */
+ uint32_t doi_data_block_size;
+ uint32_t doi_metadata_block_size;
+ uint64_t doi_bonus_size;
+ dmu_object_type_t doi_type;
+ dmu_object_type_t doi_bonus_type;
+ uint8_t doi_indirection; /* 2 = dnode->indirect->data */
+ uint8_t doi_checksum;
+ uint8_t doi_compress;
+ uint8_t doi_pad[5];
+ /* Values below are number of 512-byte blocks. */
+ uint64_t doi_physical_blks; /* data + metadata */
+ uint64_t doi_max_block_offset;
+} dmu_object_info_t;
+
+typedef void arc_byteswap_func_t(void *buf, size_t size);
+
+typedef struct dmu_object_type_info {
+ arc_byteswap_func_t *ot_byteswap;
+ boolean_t ot_metadata;
+ char *ot_name;
+} dmu_object_type_info_t;
+
+extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
+
+/*
+ * Get information on a DMU object.
+ *
+ * Return 0 on success or ENOENT if object is not allocated.
+ *
+ * If doi is NULL, just indicates whether the object exists.
+ */
+int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
+void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
+void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
+void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
+ u_longlong_t *nblk512);
+
+typedef struct dmu_objset_stats {
+ uint64_t dds_num_clones; /* number of clones of this */
+ uint64_t dds_creation_txg;
+ uint64_t dds_guid;
+ dmu_objset_type_t dds_type;
+ uint8_t dds_is_snapshot;
+ uint8_t dds_inconsistent;
+ char dds_origin[MAXNAMELEN];
+} dmu_objset_stats_t;
+
+/*
+ * Get stats on a dataset.
+ */
+void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
+
+/*
+ * Add entries to the nvlist for all the objset's properties. See
+ * zfs_prop_table[] and zfs(1m) for details on the properties.
+ */
+void dmu_objset_stats(objset_t *os, struct nvlist *nv);
+
+/*
+ * Get the space usage statistics for statvfs().
+ *
+ * refdbytes is the amount of space "referenced" by this objset.
+ * availbytes is the amount of space available to this objset, taking
+ * into account quotas & reservations, assuming that no other objsets
+ * use the space first. These values correspond to the 'referenced' and
+ * 'available' properties, described in the zfs(1m) manpage.
+ *
+ * usedobjs and availobjs are the number of objects currently allocated,
+ * and available.
+ */
+void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
+ uint64_t *usedobjsp, uint64_t *availobjsp);
+
+/*
+ * The fsid_guid is a 56-bit ID that can change to avoid collisions.
+ * (Contrast with the ds_guid which is a 64-bit ID that will never
+ * change, so there is a small probability that it will collide.)
+ */
+uint64_t dmu_objset_fsid_guid(objset_t *os);
+
+int dmu_objset_is_snapshot(objset_t *os);
+
+extern struct spa *dmu_objset_spa(objset_t *os);
+extern struct zilog *dmu_objset_zil(objset_t *os);
+extern struct dsl_pool *dmu_objset_pool(objset_t *os);
+extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
+extern void dmu_objset_name(objset_t *os, char *buf);
+extern dmu_objset_type_t dmu_objset_type(objset_t *os);
+extern uint64_t dmu_objset_id(objset_t *os);
+extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
+ uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
+extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
+ int maxlen, boolean_t *conflict);
+extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
+ uint64_t *idp, uint64_t *offp);
+extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
+extern void *dmu_objset_get_user(objset_t *os);
+
+/*
+ * Return the txg number for the given assigned transaction.
+ */
+uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
+
+/*
+ * Synchronous write.
+ * If a parent zio is provided this function initiates a write on the
+ * provided buffer as a child of the parent zio.
+ * In the absence of a parent zio, the write is completed synchronously.
+ * At write completion, blk is filled with the bp of the written block.
+ * Note that while the data covered by this function will be on stable
+ * storage when the write completes this new data does not become a
+ * permanent part of the file until the associated transaction commits.
+ */
+typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
+int dmu_sync(struct zio *zio, dmu_buf_t *db,
+ struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
+
+/*
+ * Find the next hole or data block in file starting at *off
+ * Return found offset in *off. Return ESRCH for end of file.
+ */
+int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
+ uint64_t *off);
+
+/*
+ * Initial setup and final teardown.
+ */
+extern void dmu_init(void);
+extern void dmu_fini(void);
+
+typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
+ uint64_t object, uint64_t offset, int len);
+void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
+ dmu_traverse_cb_t cb, void *arg);
+
+int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
+ struct vnode *vp, offset_t *off);
+
+typedef struct dmu_recv_cookie {
+ /*
+ * This structure is opaque!
+ *
+ * If logical and real are different, we are recving the stream
+ * into the "real" temporary clone, and then switching it with
+ * the "logical" target.
+ */
+ struct dsl_dataset *drc_logical_ds;
+ struct dsl_dataset *drc_real_ds;
+ struct drr_begin *drc_drrb;
+ char *drc_tosnap;
+ boolean_t drc_newfs;
+ boolean_t drc_force;
+} dmu_recv_cookie_t;
+
+int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
+ boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
+int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
+int dmu_recv_end(dmu_recv_cookie_t *drc);
+void dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
+
+/* CRC64 table */
+#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
+extern uint64_t zfs_crc64_table[256];
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DMU_H */
diff --git a/module/zfs/include/sys/dmu_impl.h b/module/zfs/include/sys/dmu_impl.h
new file mode 100644
index 000000000..96ce688e1
--- /dev/null
+++ b/module/zfs/include/sys/dmu_impl.h
@@ -0,0 +1,239 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DMU_IMPL_H
+#define _SYS_DMU_IMPL_H
+
+#include <sys/txg_impl.h>
+#include <sys/zio.h>
+#include <sys/dnode.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * This is the locking strategy for the DMU. Numbers in parenthesis are
+ * cases that use that lock order, referenced below:
+ *
+ * ARC is self-contained
+ * bplist is self-contained
+ * refcount is self-contained
+ * txg is self-contained (hopefully!)
+ * zst_lock
+ * zf_rwlock
+ *
+ * XXX try to improve evicting path?
+ *
+ * dp_config_rwlock > os_obj_lock > dn_struct_rwlock >
+ * dn_dbufs_mtx > hash_mutexes > db_mtx > dd_lock > leafs
+ *
+ * dp_config_rwlock
+ * must be held before: everything
+ * protects dd namespace changes
+ * protects property changes globally
+ * held from:
+ * dsl_dir_open/r:
+ * dsl_dir_create_sync/w:
+ * dsl_dir_sync_destroy/w:
+ * dsl_dir_rename_sync/w:
+ * dsl_prop_changed_notify/r:
+ *
+ * os_obj_lock
+ * must be held before:
+ * everything except dp_config_rwlock
+ * protects os_obj_next
+ * held from:
+ * dmu_object_alloc: dn_dbufs_mtx, db_mtx, hash_mutexes, dn_struct_rwlock
+ *
+ * dn_struct_rwlock
+ * must be held before:
+ * everything except dp_config_rwlock and os_obj_lock
+ * protects structure of dnode (eg. nlevels)
+ * db_blkptr can change when syncing out change to nlevels
+ * dn_maxblkid
+ * dn_nlevels
+ * dn_*blksz*
+ * phys nlevels, maxblkid, physical blkptr_t's (?)
+ * held from:
+ * callers of dbuf_read_impl, dbuf_hold[_impl], dbuf_prefetch
+ * dmu_object_info_from_dnode: dn_dirty_mtx (dn_datablksz)
+ * dmu_tx_count_free:
+ * dbuf_read_impl: db_mtx, dmu_zfetch()
+ * dmu_zfetch: zf_rwlock/r, zst_lock, dbuf_prefetch()
+ * dbuf_new_size: db_mtx
+ * dbuf_dirty: db_mtx
+ * dbuf_findbp: (callers, phys? - the real need)
+ * dbuf_create: dn_dbufs_mtx, hash_mutexes, db_mtx (phys?)
+ * dbuf_prefetch: dn_dirty_mtx, hash_mutexes, db_mtx, dn_dbufs_mtx
+ * dbuf_hold_impl: hash_mutexes, db_mtx, dn_dbufs_mtx, dbuf_findbp()
+ * dnode_sync/w (increase_indirection): db_mtx (phys)
+ * dnode_set_blksz/w: dn_dbufs_mtx (dn_*blksz*)
+ * dnode_new_blkid/w: (dn_maxblkid)
+ * dnode_free_range/w: dn_dirty_mtx (dn_maxblkid)
+ * dnode_next_offset: (phys)
+ *
+ * dn_dbufs_mtx
+ * must be held before:
+ * db_mtx, hash_mutexes
+ * protects:
+ * dn_dbufs
+ * dn_evicted
+ * held from:
+ * dmu_evict_user: db_mtx (dn_dbufs)
+ * dbuf_free_range: db_mtx (dn_dbufs)
+ * dbuf_remove_ref: db_mtx, callees:
+ * dbuf_hash_remove: hash_mutexes, db_mtx
+ * dbuf_create: hash_mutexes, db_mtx (dn_dbufs)
+ * dnode_set_blksz: (dn_dbufs)
+ *
+ * hash_mutexes (global)
+ * must be held before:
+ * db_mtx
+ * protects dbuf_hash_table (global) and db_hash_next
+ * held from:
+ * dbuf_find: db_mtx
+ * dbuf_hash_insert: db_mtx
+ * dbuf_hash_remove: db_mtx
+ *
+ * db_mtx (meta-leaf)
+ * must be held before:
+ * dn_mtx, dn_dirty_mtx, dd_lock (leaf mutexes)
+ * protects:
+ * db_state
+ * db_holds
+ * db_buf
+ * db_changed
+ * db_data_pending
+ * db_dirtied
+ * db_link
+ * db_dirty_node (??)
+ * db_dirtycnt
+ * db_d.*
+ * db.*
+ * held from:
+ * dbuf_dirty: dn_mtx, dn_dirty_mtx
+ * dbuf_dirty->dsl_dir_willuse_space: dd_lock
+ * dbuf_dirty->dbuf_new_block->dsl_dataset_block_freeable: dd_lock
+ * dbuf_undirty: dn_dirty_mtx (db_d)
+ * dbuf_write_done: dn_dirty_mtx (db_state)
+ * dbuf_*
+ * dmu_buf_update_user: none (db_d)
+ * dmu_evict_user: none (db_d) (maybe can eliminate)
+ * dbuf_find: none (db_holds)
+ * dbuf_hash_insert: none (db_holds)
+ * dmu_buf_read_array_impl: none (db_state, db_changed)
+ * dmu_sync: none (db_dirty_node, db_d)
+ * dnode_reallocate: none (db)
+ *
+ * dn_mtx (leaf)
+ * protects:
+ * dn_dirty_dbufs
+ * dn_ranges
+ * phys accounting
+ * dn_allocated_txg
+ * dn_free_txg
+ * dn_assigned_txg
+ * dd_assigned_tx
+ * dn_notxholds
+ * dn_dirtyctx
+ * dn_dirtyctx_firstset
+ * (dn_phys copy fields?)
+ * (dn_phys contents?)
+ * held from:
+ * dnode_*
+ * dbuf_dirty: none
+ * dbuf_sync: none (phys accounting)
+ * dbuf_undirty: none (dn_ranges, dn_dirty_dbufs)
+ * dbuf_write_done: none (phys accounting)
+ * dmu_object_info_from_dnode: none (accounting)
+ * dmu_tx_commit: none
+ * dmu_tx_hold_object_impl: none
+ * dmu_tx_try_assign: dn_notxholds(cv)
+ * dmu_tx_unassign: none
+ *
+ * dd_lock
+ * must be held before:
+ * ds_lock
+ * ancestors' dd_lock
+ * protects:
+ * dd_prop_cbs
+ * dd_sync_*
+ * dd_used_bytes
+ * dd_tempreserved
+ * dd_space_towrite
+ * dd_myname
+ * dd_phys accounting?
+ * held from:
+ * dsl_dir_*
+ * dsl_prop_changed_notify: none (dd_prop_cbs)
+ * dsl_prop_register: none (dd_prop_cbs)
+ * dsl_prop_unregister: none (dd_prop_cbs)
+ * dsl_dataset_block_freeable: none (dd_sync_*)
+ *
+ * os_lock (leaf)
+ * protects:
+ * os_dirty_dnodes
+ * os_free_dnodes
+ * os_dnodes
+ * os_downgraded_dbufs
+ * dn_dirtyblksz
+ * dn_dirty_link
+ * held from:
+ * dnode_create: none (os_dnodes)
+ * dnode_destroy: none (os_dnodes)
+ * dnode_setdirty: none (dn_dirtyblksz, os_*_dnodes)
+ * dnode_free: none (dn_dirtyblksz, os_*_dnodes)
+ *
+ * ds_lock
+ * protects:
+ * ds_user_ptr
+ * ds_user_evice_func
+ * ds_open_refcount
+ * ds_snapname
+ * ds_phys accounting
+ * ds_reserved
+ * held from:
+ * dsl_dataset_*
+ *
+ * dr_mtx (leaf)
+ * protects:
+ * dr_children
+ * held from:
+ * dbuf_dirty
+ * dbuf_undirty
+ * dbuf_sync_indirect
+ * dnode_new_blkid
+ */
+
+struct objset;
+struct dmu_pool;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DMU_IMPL_H */
diff --git a/module/zfs/include/sys/dmu_objset.h b/module/zfs/include/sys/dmu_objset.h
new file mode 100644
index 000000000..15df29a17
--- /dev/null
+++ b/module/zfs/include/sys/dmu_objset.h
@@ -0,0 +1,136 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DMU_OBJSET_H
+#define _SYS_DMU_OBJSET_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/arc.h>
+#include <sys/txg.h>
+#include <sys/zfs_context.h>
+#include <sys/dnode.h>
+#include <sys/zio.h>
+#include <sys/zil.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dsl_dataset;
+struct dmu_tx;
+struct objset_impl;
+
+typedef struct objset_phys {
+ dnode_phys_t os_meta_dnode;
+ zil_header_t os_zil_header;
+ uint64_t os_type;
+ char os_pad[1024 - sizeof (dnode_phys_t) - sizeof (zil_header_t) -
+ sizeof (uint64_t)];
+} objset_phys_t;
+
+struct objset {
+ struct objset_impl *os;
+ int os_mode;
+};
+
+typedef struct objset_impl {
+ /* Immutable: */
+ struct dsl_dataset *os_dsl_dataset;
+ spa_t *os_spa;
+ arc_buf_t *os_phys_buf;
+ objset_phys_t *os_phys;
+ dnode_t *os_meta_dnode;
+ zilog_t *os_zil;
+ objset_t os;
+ uint8_t os_checksum; /* can change, under dsl_dir's locks */
+ uint8_t os_compress; /* can change, under dsl_dir's locks */
+ uint8_t os_copies; /* can change, under dsl_dir's locks */
+ uint8_t os_primary_cache; /* can change, under dsl_dir's locks */
+ uint8_t os_secondary_cache; /* can change, under dsl_dir's locks */
+
+ /* no lock needed: */
+ struct dmu_tx *os_synctx; /* XXX sketchy */
+ blkptr_t *os_rootbp;
+ zil_header_t os_zil_header;
+
+ /* Protected by os_obj_lock */
+ kmutex_t os_obj_lock;
+ uint64_t os_obj_next;
+
+ /* Protected by os_lock */
+ kmutex_t os_lock;
+ list_t os_dirty_dnodes[TXG_SIZE];
+ list_t os_free_dnodes[TXG_SIZE];
+ list_t os_dnodes;
+ list_t os_downgraded_dbufs;
+
+ /* stuff we store for the user */
+ kmutex_t os_user_ptr_lock;
+ void *os_user_ptr;
+} objset_impl_t;
+
+#define DMU_META_DNODE_OBJECT 0
+
+#define DMU_OS_IS_L2CACHEABLE(os) \
+ ((os)->os_secondary_cache == ZFS_CACHE_ALL || \
+ (os)->os_secondary_cache == ZFS_CACHE_METADATA)
+
+/* called from zpl */
+int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
+ objset_t **osp);
+void dmu_objset_close(objset_t *os);
+int dmu_objset_create(const char *name, dmu_objset_type_t type,
+ objset_t *clone_parent, uint64_t flags,
+ void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
+int dmu_objset_destroy(const char *name);
+int dmu_objset_rollback(objset_t *os);
+int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
+void dmu_objset_stats(objset_t *os, nvlist_t *nv);
+void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
+void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
+ uint64_t *usedobjsp, uint64_t *availobjsp);
+uint64_t dmu_objset_fsid_guid(objset_t *os);
+int dmu_objset_find(char *name, int func(char *, void *), void *arg,
+ int flags);
+int dmu_objset_find_spa(spa_t *spa, const char *name,
+ int func(spa_t *, uint64_t, const char *, void *), void *arg, int flags);
+void dmu_objset_byteswap(void *buf, size_t size);
+int dmu_objset_evict_dbufs(objset_t *os);
+
+/* called from dsl */
+void dmu_objset_sync(objset_impl_t *os, zio_t *zio, dmu_tx_t *tx);
+objset_impl_t *dmu_objset_create_impl(spa_t *spa, struct dsl_dataset *ds,
+ blkptr_t *bp, dmu_objset_type_t type, dmu_tx_t *tx);
+int dmu_objset_open_impl(spa_t *spa, struct dsl_dataset *ds, blkptr_t *bp,
+ objset_impl_t **osip);
+void dmu_objset_evict(struct dsl_dataset *ds, void *arg);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DMU_OBJSET_H */
diff --git a/module/zfs/include/sys/dmu_traverse.h b/module/zfs/include/sys/dmu_traverse.h
new file mode 100644
index 000000000..3e0268911
--- /dev/null
+++ b/module/zfs/include/sys/dmu_traverse.h
@@ -0,0 +1,57 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DMU_TRAVERSE_H
+#define _SYS_DMU_TRAVERSE_H
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dnode_phys;
+struct dsl_dataset;
+
+typedef int (blkptr_cb_t)(spa_t *spa, blkptr_t *bp,
+ const zbookmark_t *zb, const struct dnode_phys *dnp, void *arg);
+
+#define TRAVERSE_PRE (1<<0)
+#define TRAVERSE_POST (1<<1)
+#define TRAVERSE_PREFETCH_METADATA (1<<2)
+#define TRAVERSE_PREFETCH_DATA (1<<3)
+#define TRAVERSE_PREFETCH (TRAVERSE_PREFETCH_METADATA | TRAVERSE_PREFETCH_DATA)
+
+int traverse_dataset(struct dsl_dataset *ds, uint64_t txg_start,
+ int flags, blkptr_cb_t func, void *arg);
+int traverse_pool(spa_t *spa, blkptr_cb_t func, void *arg);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DMU_TRAVERSE_H */
diff --git a/module/zfs/include/sys/dmu_tx.h b/module/zfs/include/sys/dmu_tx.h
new file mode 100644
index 000000000..2727daaaa
--- /dev/null
+++ b/module/zfs/include/sys/dmu_tx.h
@@ -0,0 +1,139 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DMU_TX_H
+#define _SYS_DMU_TX_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/inttypes.h>
+#include <sys/dmu.h>
+#include <sys/txg.h>
+#include <sys/refcount.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dmu_buf_impl;
+struct dmu_tx_hold;
+struct dnode_link;
+struct dsl_pool;
+struct dnode;
+struct dsl_dir;
+
+struct dmu_tx {
+ /*
+ * No synchronization is needed because a tx can only be handled
+ * by one thread.
+ */
+ list_t tx_holds; /* list of dmu_tx_hold_t */
+ objset_t *tx_objset;
+ struct dsl_dir *tx_dir;
+ struct dsl_pool *tx_pool;
+ uint64_t tx_txg;
+ uint64_t tx_lastsnap_txg;
+ uint64_t tx_lasttried_txg;
+ txg_handle_t tx_txgh;
+ void *tx_tempreserve_cookie;
+ struct dmu_tx_hold *tx_needassign_txh;
+ uint8_t tx_anyobj;
+ int tx_err;
+#ifdef ZFS_DEBUG
+ uint64_t tx_space_towrite;
+ uint64_t tx_space_tofree;
+ uint64_t tx_space_tooverwrite;
+ uint64_t tx_space_tounref;
+ refcount_t tx_space_written;
+ refcount_t tx_space_freed;
+#endif
+};
+
+enum dmu_tx_hold_type {
+ THT_NEWOBJECT,
+ THT_WRITE,
+ THT_BONUS,
+ THT_FREE,
+ THT_ZAP,
+ THT_SPACE,
+ THT_NUMTYPES
+};
+
+typedef struct dmu_tx_hold {
+ dmu_tx_t *txh_tx;
+ list_node_t txh_node;
+ struct dnode *txh_dnode;
+ uint64_t txh_space_towrite;
+ uint64_t txh_space_tofree;
+ uint64_t txh_space_tooverwrite;
+ uint64_t txh_space_tounref;
+ uint64_t txh_memory_tohold;
+ uint64_t txh_fudge;
+#ifdef ZFS_DEBUG
+ enum dmu_tx_hold_type txh_type;
+ uint64_t txh_arg1;
+ uint64_t txh_arg2;
+#endif
+} dmu_tx_hold_t;
+
+
+/*
+ * These routines are defined in dmu.h, and are called by the user.
+ */
+dmu_tx_t *dmu_tx_create(objset_t *dd);
+int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
+void dmu_tx_commit(dmu_tx_t *tx);
+void dmu_tx_abort(dmu_tx_t *tx);
+uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
+void dmu_tx_wait(dmu_tx_t *tx);
+
+/*
+ * These routines are defined in dmu_spa.h, and are called by the SPA.
+ */
+extern dmu_tx_t *dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg);
+
+/*
+ * These routines are only called by the DMU.
+ */
+dmu_tx_t *dmu_tx_create_dd(dsl_dir_t *dd);
+int dmu_tx_is_syncing(dmu_tx_t *tx);
+int dmu_tx_private_ok(dmu_tx_t *tx);
+void dmu_tx_add_new_object(dmu_tx_t *tx, objset_t *os, uint64_t object);
+void dmu_tx_willuse_space(dmu_tx_t *tx, int64_t delta);
+void dmu_tx_dirty_buf(dmu_tx_t *tx, struct dmu_buf_impl *db);
+int dmu_tx_holds(dmu_tx_t *tx, uint64_t object);
+void dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space);
+
+#ifdef ZFS_DEBUG
+#define DMU_TX_DIRTY_BUF(tx, db) dmu_tx_dirty_buf(tx, db)
+#else
+#define DMU_TX_DIRTY_BUF(tx, db)
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DMU_TX_H */
diff --git a/module/zfs/include/sys/dmu_zfetch.h b/module/zfs/include/sys/dmu_zfetch.h
new file mode 100644
index 000000000..c94bced93
--- /dev/null
+++ b/module/zfs/include/sys/dmu_zfetch.h
@@ -0,0 +1,75 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _DFETCH_H
+#define _DFETCH_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern uint64_t zfetch_array_rd_sz;
+
+struct dnode; /* so we can reference dnode */
+
+typedef enum zfetch_dirn {
+ ZFETCH_FORWARD = 1, /* prefetch increasing block numbers */
+ ZFETCH_BACKWARD = -1 /* prefetch decreasing block numbers */
+} zfetch_dirn_t;
+
+typedef struct zstream {
+ uint64_t zst_offset; /* offset of starting block in range */
+ uint64_t zst_len; /* length of range, in blocks */
+ zfetch_dirn_t zst_direction; /* direction of prefetch */
+ uint64_t zst_stride; /* length of stride, in blocks */
+ uint64_t zst_ph_offset; /* prefetch offset, in blocks */
+ uint64_t zst_cap; /* prefetch limit (cap), in blocks */
+ kmutex_t zst_lock; /* protects stream */
+ clock_t zst_last; /* lbolt of last prefetch */
+ avl_node_t zst_node; /* embed avl node here */
+} zstream_t;
+
+typedef struct zfetch {
+ krwlock_t zf_rwlock; /* protects zfetch structure */
+ list_t zf_stream; /* AVL tree of zstream_t's */
+ struct dnode *zf_dnode; /* dnode that owns this zfetch */
+ uint32_t zf_stream_cnt; /* # of active streams */
+ uint64_t zf_alloc_fail; /* # of failed attempts to alloc strm */
+} zfetch_t;
+
+void dmu_zfetch_init(zfetch_t *, struct dnode *);
+void dmu_zfetch_rele(zfetch_t *);
+void dmu_zfetch(zfetch_t *, uint64_t, uint64_t, int);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _DFETCH_H */
diff --git a/module/zfs/include/sys/dnode.h b/module/zfs/include/sys/dnode.h
new file mode 100644
index 000000000..c79ff48a6
--- /dev/null
+++ b/module/zfs/include/sys/dnode.h
@@ -0,0 +1,275 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DNODE_H
+#define _SYS_DNODE_H
+
+#include <sys/zfs_context.h>
+#include <sys/avl.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/zio.h>
+#include <sys/refcount.h>
+#include <sys/dmu_zfetch.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * dnode_hold() flags.
+ */
+#define DNODE_MUST_BE_ALLOCATED 1
+#define DNODE_MUST_BE_FREE 2
+
+/*
+ * dnode_next_offset() flags.
+ */
+#define DNODE_FIND_HOLE 1
+#define DNODE_FIND_BACKWARDS 2
+#define DNODE_FIND_HAVELOCK 4
+
+/*
+ * Fixed constants.
+ */
+#define DNODE_SHIFT 9 /* 512 bytes */
+#define DN_MIN_INDBLKSHIFT 10 /* 1k */
+#define DN_MAX_INDBLKSHIFT 14 /* 16k */
+#define DNODE_BLOCK_SHIFT 14 /* 16k */
+#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
+#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
+#define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
+
+/*
+ * Derived constants.
+ */
+#define DNODE_SIZE (1 << DNODE_SHIFT)
+#define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
+#define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
+#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
+#define DN_ZERO_BONUSLEN (DN_MAX_BONUSLEN + 1)
+
+#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
+#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
+#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
+
+/* The +2 here is a cheesy way to round up */
+#define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
+ (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
+
+#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
+ (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
+
+#define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
+ (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
+
+#define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
+
+struct dmu_buf_impl;
+struct objset_impl;
+struct zio;
+
+enum dnode_dirtycontext {
+ DN_UNDIRTIED,
+ DN_DIRTY_OPEN,
+ DN_DIRTY_SYNC
+};
+
+/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
+#define DNODE_FLAG_USED_BYTES (1<<0)
+
+typedef struct dnode_phys {
+ uint8_t dn_type; /* dmu_object_type_t */
+ uint8_t dn_indblkshift; /* ln2(indirect block size) */
+ uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
+ uint8_t dn_nblkptr; /* length of dn_blkptr */
+ uint8_t dn_bonustype; /* type of data in bonus buffer */
+ uint8_t dn_checksum; /* ZIO_CHECKSUM type */
+ uint8_t dn_compress; /* ZIO_COMPRESS type */
+ uint8_t dn_flags; /* DNODE_FLAG_* */
+ uint16_t dn_datablkszsec; /* data block size in 512b sectors */
+ uint16_t dn_bonuslen; /* length of dn_bonus */
+ uint8_t dn_pad2[4];
+
+ /* accounting is protected by dn_dirty_mtx */
+ uint64_t dn_maxblkid; /* largest allocated block ID */
+ uint64_t dn_used; /* bytes (or sectors) of disk space */
+
+ uint64_t dn_pad3[4];
+
+ blkptr_t dn_blkptr[1];
+ uint8_t dn_bonus[DN_MAX_BONUSLEN];
+} dnode_phys_t;
+
+typedef struct dnode {
+ /*
+ * dn_struct_rwlock protects the structure of the dnode,
+ * including the number of levels of indirection (dn_nlevels),
+ * dn_maxblkid, and dn_next_*
+ */
+ krwlock_t dn_struct_rwlock;
+
+ /*
+ * Our link on dataset's dd_dnodes list.
+ * Protected by dd_accounting_mtx.
+ */
+ list_node_t dn_link;
+
+ /* immutable: */
+ struct objset_impl *dn_objset;
+ uint64_t dn_object;
+ struct dmu_buf_impl *dn_dbuf;
+ dnode_phys_t *dn_phys; /* pointer into dn->dn_dbuf->db.db_data */
+
+ /*
+ * Copies of stuff in dn_phys. They're valid in the open
+ * context (eg. even before the dnode is first synced).
+ * Where necessary, these are protected by dn_struct_rwlock.
+ */
+ dmu_object_type_t dn_type; /* object type */
+ uint16_t dn_bonuslen; /* bonus length */
+ uint8_t dn_bonustype; /* bonus type */
+ uint8_t dn_nblkptr; /* number of blkptrs (immutable) */
+ uint8_t dn_checksum; /* ZIO_CHECKSUM type */
+ uint8_t dn_compress; /* ZIO_COMPRESS type */
+ uint8_t dn_nlevels;
+ uint8_t dn_indblkshift;
+ uint8_t dn_datablkshift; /* zero if blksz not power of 2! */
+ uint16_t dn_datablkszsec; /* in 512b sectors */
+ uint32_t dn_datablksz; /* in bytes */
+ uint64_t dn_maxblkid;
+ uint8_t dn_next_nlevels[TXG_SIZE];
+ uint8_t dn_next_indblkshift[TXG_SIZE];
+ uint16_t dn_next_bonuslen[TXG_SIZE];
+ uint32_t dn_next_blksz[TXG_SIZE]; /* next block size in bytes */
+
+ /* protected by os_lock: */
+ list_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */
+
+ /* protected by dn_mtx: */
+ kmutex_t dn_mtx;
+ list_t dn_dirty_records[TXG_SIZE];
+ avl_tree_t dn_ranges[TXG_SIZE];
+ uint64_t dn_allocated_txg;
+ uint64_t dn_free_txg;
+ uint64_t dn_assigned_txg;
+ kcondvar_t dn_notxholds;
+ enum dnode_dirtycontext dn_dirtyctx;
+ uint8_t *dn_dirtyctx_firstset; /* dbg: contents meaningless */
+
+ /* protected by own devices */
+ refcount_t dn_tx_holds;
+ refcount_t dn_holds;
+
+ kmutex_t dn_dbufs_mtx;
+ list_t dn_dbufs; /* linked list of descendent dbuf_t's */
+ struct dmu_buf_impl *dn_bonus; /* bonus buffer dbuf */
+
+ /* parent IO for current sync write */
+ zio_t *dn_zio;
+
+ /* holds prefetch structure */
+ struct zfetch dn_zfetch;
+} dnode_t;
+
+typedef struct free_range {
+ avl_node_t fr_node;
+ uint64_t fr_blkid;
+ uint64_t fr_nblks;
+} free_range_t;
+
+dnode_t *dnode_special_open(struct objset_impl *dd, dnode_phys_t *dnp,
+ uint64_t object);
+void dnode_special_close(dnode_t *dn);
+
+void dnode_setbonuslen(dnode_t *dn, int newsize, dmu_tx_t *tx);
+int dnode_hold(struct objset_impl *dd, uint64_t object,
+ void *ref, dnode_t **dnp);
+int dnode_hold_impl(struct objset_impl *dd, uint64_t object, int flag,
+ void *ref, dnode_t **dnp);
+boolean_t dnode_add_ref(dnode_t *dn, void *ref);
+void dnode_rele(dnode_t *dn, void *ref);
+void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx);
+void dnode_sync(dnode_t *dn, dmu_tx_t *tx);
+void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+void dnode_free(dnode_t *dn, dmu_tx_t *tx);
+void dnode_byteswap(dnode_phys_t *dnp);
+void dnode_buf_byteswap(void *buf, size_t size);
+void dnode_verify(dnode_t *dn);
+int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx);
+uint64_t dnode_current_max_length(dnode_t *dn);
+void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx);
+void dnode_clear_range(dnode_t *dn, uint64_t blkid,
+ uint64_t nblks, dmu_tx_t *tx);
+void dnode_diduse_space(dnode_t *dn, int64_t space);
+void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx);
+void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t);
+uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid);
+void dnode_init(void);
+void dnode_fini(void);
+int dnode_next_offset(dnode_t *dn, int flags, uint64_t *off,
+ int minlvl, uint64_t blkfill, uint64_t txg);
+void dnode_evict_dbufs(dnode_t *dn);
+
+#ifdef ZFS_DEBUG
+
+/*
+ * There should be a ## between the string literal and fmt, to make it
+ * clear that we're joining two strings together, but that piece of shit
+ * gcc doesn't support that preprocessor token.
+ */
+#define dprintf_dnode(dn, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char __db_buf[32]; \
+ uint64_t __db_obj = (dn)->dn_object; \
+ if (__db_obj == DMU_META_DNODE_OBJECT) \
+ (void) strcpy(__db_buf, "mdn"); \
+ else \
+ (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
+ (u_longlong_t)__db_obj);\
+ dprintf_ds((dn)->dn_objset->os_dsl_dataset, "obj=%s " fmt, \
+ __db_buf, __VA_ARGS__); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+
+#define DNODE_VERIFY(dn) dnode_verify(dn)
+#define FREE_VERIFY(db, start, end, tx) free_verify(db, start, end, tx)
+
+#else
+
+#define dprintf_dnode(db, fmt, ...)
+#define DNODE_VERIFY(dn)
+#define FREE_VERIFY(db, start, end, tx)
+
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DNODE_H */
diff --git a/module/zfs/include/sys/dsl_dataset.h b/module/zfs/include/sys/dsl_dataset.h
new file mode 100644
index 000000000..8665aec2d
--- /dev/null
+++ b/module/zfs/include/sys/dsl_dataset.h
@@ -0,0 +1,239 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_DATASET_H
+#define _SYS_DSL_DATASET_H
+
+#include <sys/dmu.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/zio.h>
+#include <sys/bplist.h>
+#include <sys/dsl_synctask.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dsl_dataset;
+struct dsl_dir;
+struct dsl_pool;
+
+typedef void dsl_dataset_evict_func_t(struct dsl_dataset *, void *);
+
+#define DS_FLAG_INCONSISTENT (1ULL<<0)
+#define DS_IS_INCONSISTENT(ds) \
+ ((ds)->ds_phys->ds_flags & DS_FLAG_INCONSISTENT)
+/*
+ * NB: nopromote can not yet be set, but we want support for it in this
+ * on-disk version, so that we don't need to upgrade for it later. It
+ * will be needed when we implement 'zfs split' (where the split off
+ * clone should not be promoted).
+ */
+#define DS_FLAG_NOPROMOTE (1ULL<<1)
+
+/*
+ * DS_FLAG_UNIQUE_ACCURATE is set if ds_unique_bytes has been correctly
+ * calculated for head datasets (starting with SPA_VERSION_UNIQUE_ACCURATE,
+ * refquota/refreservations).
+ */
+#define DS_FLAG_UNIQUE_ACCURATE (1ULL<<2)
+
+/*
+ * DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
+ * name lookups should be performed case-insensitively.
+ */
+#define DS_FLAG_CI_DATASET (1ULL<<16)
+
+typedef struct dsl_dataset_phys {
+ uint64_t ds_dir_obj; /* DMU_OT_DSL_DIR */
+ uint64_t ds_prev_snap_obj; /* DMU_OT_DSL_DATASET */
+ uint64_t ds_prev_snap_txg;
+ uint64_t ds_next_snap_obj; /* DMU_OT_DSL_DATASET */
+ uint64_t ds_snapnames_zapobj; /* DMU_OT_DSL_DS_SNAP_MAP 0 for snaps */
+ uint64_t ds_num_children; /* clone/snap children; ==0 for head */
+ uint64_t ds_creation_time; /* seconds since 1970 */
+ uint64_t ds_creation_txg;
+ uint64_t ds_deadlist_obj; /* DMU_OT_BPLIST */
+ uint64_t ds_used_bytes;
+ uint64_t ds_compressed_bytes;
+ uint64_t ds_uncompressed_bytes;
+ uint64_t ds_unique_bytes; /* only relevant to snapshots */
+ /*
+ * The ds_fsid_guid is a 56-bit ID that can change to avoid
+ * collisions. The ds_guid is a 64-bit ID that will never
+ * change, so there is a small probability that it will collide.
+ */
+ uint64_t ds_fsid_guid;
+ uint64_t ds_guid;
+ uint64_t ds_flags; /* DS_FLAG_* */
+ blkptr_t ds_bp;
+ uint64_t ds_next_clones_obj; /* DMU_OT_DSL_CLONES */
+ uint64_t ds_props_obj; /* DMU_OT_DSL_PROPS for snaps */
+ uint64_t ds_pad[6]; /* pad out to 320 bytes for good measure */
+} dsl_dataset_phys_t;
+
+typedef struct dsl_dataset {
+ /* Immutable: */
+ struct dsl_dir *ds_dir;
+ dsl_dataset_phys_t *ds_phys;
+ dmu_buf_t *ds_dbuf;
+ uint64_t ds_object;
+ uint64_t ds_fsid_guid;
+
+ /* only used in syncing context, only valid for non-snapshots: */
+ struct dsl_dataset *ds_prev;
+ uint64_t ds_origin_txg;
+
+ /* has internal locking: */
+ bplist_t ds_deadlist;
+
+ /* protected by lock on pool's dp_dirty_datasets list */
+ txg_node_t ds_dirty_link;
+ list_node_t ds_synced_link;
+
+ /*
+ * ds_phys->ds_<accounting> is also protected by ds_lock.
+ * Protected by ds_lock:
+ */
+ kmutex_t ds_lock;
+ void *ds_user_ptr;
+ dsl_dataset_evict_func_t *ds_user_evict_func;
+
+ /*
+ * ds_owner is protected by the ds_rwlock and the ds_lock
+ */
+ krwlock_t ds_rwlock;
+ kcondvar_t ds_exclusive_cv;
+ void *ds_owner;
+
+ /* no locking; only for making guesses */
+ uint64_t ds_trysnap_txg;
+
+ /* for objset_open() */
+ kmutex_t ds_opening_lock;
+
+ uint64_t ds_reserved; /* cached refreservation */
+ uint64_t ds_quota; /* cached refquota */
+
+ /* Protected by ds_lock; keep at end of struct for better locality */
+ char ds_snapname[MAXNAMELEN];
+} dsl_dataset_t;
+
+#define dsl_dataset_is_snapshot(ds) \
+ ((ds)->ds_phys->ds_num_children != 0)
+
+#define DS_UNIQUE_IS_ACCURATE(ds) \
+ (((ds)->ds_phys->ds_flags & DS_FLAG_UNIQUE_ACCURATE) != 0)
+
+int dsl_dataset_hold(const char *name, void *tag, dsl_dataset_t **dsp);
+int dsl_dataset_hold_obj(struct dsl_pool *dp, uint64_t dsobj,
+ void *tag, dsl_dataset_t **);
+int dsl_dataset_own(const char *name, int flags, void *owner,
+ dsl_dataset_t **dsp);
+int dsl_dataset_own_obj(struct dsl_pool *dp, uint64_t dsobj,
+ int flags, void *owner, dsl_dataset_t **);
+void dsl_dataset_name(dsl_dataset_t *ds, char *name);
+void dsl_dataset_rele(dsl_dataset_t *ds, void *tag);
+void dsl_dataset_disown(dsl_dataset_t *ds, void *owner);
+void dsl_dataset_drop_ref(dsl_dataset_t *ds, void *tag);
+boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, boolean_t inconsistentok,
+ void *owner);
+void dsl_dataset_make_exclusive(dsl_dataset_t *ds, void *owner);
+uint64_t dsl_dataset_create_sync(dsl_dir_t *pds, const char *lastname,
+ dsl_dataset_t *origin, uint64_t flags, cred_t *, dmu_tx_t *);
+uint64_t dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
+ uint64_t flags, dmu_tx_t *tx);
+int dsl_dataset_destroy(dsl_dataset_t *ds, void *tag);
+int dsl_snapshots_destroy(char *fsname, char *snapname);
+dsl_checkfunc_t dsl_dataset_destroy_check;
+dsl_syncfunc_t dsl_dataset_destroy_sync;
+dsl_checkfunc_t dsl_dataset_snapshot_check;
+dsl_syncfunc_t dsl_dataset_snapshot_sync;
+int dsl_dataset_rollback(dsl_dataset_t *ds, dmu_objset_type_t ost);
+int dsl_dataset_rename(char *name, const char *newname, boolean_t recursive);
+int dsl_dataset_promote(const char *name);
+int dsl_dataset_clone_swap(dsl_dataset_t *clone, dsl_dataset_t *origin_head,
+ boolean_t force);
+
+void *dsl_dataset_set_user_ptr(dsl_dataset_t *ds,
+ void *p, dsl_dataset_evict_func_t func);
+void *dsl_dataset_get_user_ptr(dsl_dataset_t *ds);
+
+blkptr_t *dsl_dataset_get_blkptr(dsl_dataset_t *ds);
+void dsl_dataset_set_blkptr(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx);
+
+spa_t *dsl_dataset_get_spa(dsl_dataset_t *ds);
+
+boolean_t dsl_dataset_modified_since_lastsnap(dsl_dataset_t *ds);
+
+void dsl_dataset_sync(dsl_dataset_t *os, zio_t *zio, dmu_tx_t *tx);
+
+void dsl_dataset_block_born(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx);
+int dsl_dataset_block_kill(dsl_dataset_t *ds, blkptr_t *bp, zio_t *pio,
+ dmu_tx_t *tx);
+int dsl_dataset_block_freeable(dsl_dataset_t *ds, uint64_t blk_birth);
+uint64_t dsl_dataset_prev_snap_txg(dsl_dataset_t *ds);
+
+void dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx);
+void dsl_dataset_stats(dsl_dataset_t *os, nvlist_t *nv);
+void dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat);
+void dsl_dataset_space(dsl_dataset_t *ds,
+ uint64_t *refdbytesp, uint64_t *availbytesp,
+ uint64_t *usedobjsp, uint64_t *availobjsp);
+uint64_t dsl_dataset_fsid_guid(dsl_dataset_t *ds);
+
+int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf);
+
+int dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
+ uint64_t asize, uint64_t inflight, uint64_t *used,
+ uint64_t *ref_rsrv);
+int dsl_dataset_set_quota(const char *dsname, uint64_t quota);
+void dsl_dataset_set_quota_sync(void *arg1, void *arg2, cred_t *cr,
+ dmu_tx_t *tx);
+int dsl_dataset_set_reservation(const char *dsname, uint64_t reservation);
+void dsl_dataset_set_flags(dsl_dataset_t *ds, uint64_t flags);
+int64_t dsl_dataset_new_refreservation(dsl_dataset_t *ds, uint64_t reservation,
+ dmu_tx_t *tx);
+
+#ifdef ZFS_DEBUG
+#define dprintf_ds(ds, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char *__ds_name = kmem_alloc(MAXNAMELEN, KM_SLEEP); \
+ dsl_dataset_name(ds, __ds_name); \
+ dprintf("ds=%s " fmt, __ds_name, __VA_ARGS__); \
+ kmem_free(__ds_name, MAXNAMELEN); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+#else
+#define dprintf_ds(dd, fmt, ...)
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_DATASET_H */
diff --git a/module/zfs/include/sys/dsl_deleg.h b/module/zfs/include/sys/dsl_deleg.h
new file mode 100644
index 000000000..a29e44e67
--- /dev/null
+++ b/module/zfs/include/sys/dsl_deleg.h
@@ -0,0 +1,73 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_DELEG_H
+#define _SYS_DSL_DELEG_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dsl_pool.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZFS_DELEG_PERM_NONE ""
+#define ZFS_DELEG_PERM_CREATE "create"
+#define ZFS_DELEG_PERM_DESTROY "destroy"
+#define ZFS_DELEG_PERM_SNAPSHOT "snapshot"
+#define ZFS_DELEG_PERM_ROLLBACK "rollback"
+#define ZFS_DELEG_PERM_CLONE "clone"
+#define ZFS_DELEG_PERM_PROMOTE "promote"
+#define ZFS_DELEG_PERM_RENAME "rename"
+#define ZFS_DELEG_PERM_MOUNT "mount"
+#define ZFS_DELEG_PERM_SHARE "share"
+#define ZFS_DELEG_PERM_SEND "send"
+#define ZFS_DELEG_PERM_RECEIVE "receive"
+#define ZFS_DELEG_PERM_ALLOW "allow"
+#define ZFS_DELEG_PERM_USERPROP "userprop"
+#define ZFS_DELEG_PERM_VSCAN "vscan"
+
+/*
+ * Note: the names of properties that are marked delegatable are also
+ * valid delegated permissions
+ */
+
+int dsl_deleg_get(const char *ddname, nvlist_t **nvp);
+int dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset);
+int dsl_deleg_access(const char *ddname, const char *perm, cred_t *cr);
+void dsl_deleg_set_create_perms(dsl_dir_t *dd, dmu_tx_t *tx, cred_t *cr);
+int dsl_deleg_can_allow(char *ddname, nvlist_t *nvp, cred_t *cr);
+int dsl_deleg_can_unallow(char *ddname, nvlist_t *nvp, cred_t *cr);
+int dsl_deleg_destroy(objset_t *os, uint64_t zapobj, dmu_tx_t *tx);
+boolean_t dsl_delegation_on(objset_t *os);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_DELEG_H */
diff --git a/module/zfs/include/sys/dsl_dir.h b/module/zfs/include/sys/dsl_dir.h
new file mode 100644
index 000000000..86b9636ce
--- /dev/null
+++ b/module/zfs/include/sys/dsl_dir.h
@@ -0,0 +1,160 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_DIR_H
+#define _SYS_DSL_DIR_H
+
+#include <sys/dmu.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_synctask.h>
+#include <sys/refcount.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dsl_dataset;
+
+typedef enum dd_used {
+ DD_USED_HEAD,
+ DD_USED_SNAP,
+ DD_USED_CHILD,
+ DD_USED_CHILD_RSRV,
+ DD_USED_REFRSRV,
+ DD_USED_NUM
+} dd_used_t;
+
+#define DD_FLAG_USED_BREAKDOWN (1<<0)
+
+typedef struct dsl_dir_phys {
+ uint64_t dd_creation_time; /* not actually used */
+ uint64_t dd_head_dataset_obj;
+ uint64_t dd_parent_obj;
+ uint64_t dd_origin_obj;
+ uint64_t dd_child_dir_zapobj;
+ /*
+ * how much space our children are accounting for; for leaf
+ * datasets, == physical space used by fs + snaps
+ */
+ uint64_t dd_used_bytes;
+ uint64_t dd_compressed_bytes;
+ uint64_t dd_uncompressed_bytes;
+ /* Administrative quota setting */
+ uint64_t dd_quota;
+ /* Administrative reservation setting */
+ uint64_t dd_reserved;
+ uint64_t dd_props_zapobj;
+ uint64_t dd_deleg_zapobj; /* dataset delegation permissions */
+ uint64_t dd_flags;
+ uint64_t dd_used_breakdown[DD_USED_NUM];
+ uint64_t dd_pad[14]; /* pad out to 256 bytes for good measure */
+} dsl_dir_phys_t;
+
+struct dsl_dir {
+ /* These are immutable; no lock needed: */
+ uint64_t dd_object;
+ dsl_dir_phys_t *dd_phys;
+ dmu_buf_t *dd_dbuf;
+ dsl_pool_t *dd_pool;
+
+ /* protected by lock on pool's dp_dirty_dirs list */
+ txg_node_t dd_dirty_link;
+
+ /* protected by dp_config_rwlock */
+ dsl_dir_t *dd_parent;
+
+ /* Protected by dd_lock */
+ kmutex_t dd_lock;
+ list_t dd_prop_cbs; /* list of dsl_prop_cb_record_t's */
+
+ /* gross estimate of space used by in-flight tx's */
+ uint64_t dd_tempreserved[TXG_SIZE];
+ /* amount of space we expect to write; == amount of dirty data */
+ int64_t dd_space_towrite[TXG_SIZE];
+
+ /* protected by dd_lock; keep at end of struct for better locality */
+ char dd_myname[MAXNAMELEN];
+};
+
+void dsl_dir_close(dsl_dir_t *dd, void *tag);
+int dsl_dir_open(const char *name, void *tag, dsl_dir_t **, const char **tail);
+int dsl_dir_open_spa(spa_t *spa, const char *name, void *tag, dsl_dir_t **,
+ const char **tailp);
+int dsl_dir_open_obj(dsl_pool_t *dp, uint64_t ddobj,
+ const char *tail, void *tag, dsl_dir_t **);
+void dsl_dir_name(dsl_dir_t *dd, char *buf);
+int dsl_dir_namelen(dsl_dir_t *dd);
+int dsl_dir_is_private(dsl_dir_t *dd);
+uint64_t dsl_dir_create_sync(dsl_pool_t *dp, dsl_dir_t *pds,
+ const char *name, dmu_tx_t *tx);
+dsl_checkfunc_t dsl_dir_destroy_check;
+dsl_syncfunc_t dsl_dir_destroy_sync;
+void dsl_dir_stats(dsl_dir_t *dd, nvlist_t *nv);
+uint64_t dsl_dir_space_available(dsl_dir_t *dd,
+ dsl_dir_t *ancestor, int64_t delta, int ondiskonly);
+void dsl_dir_dirty(dsl_dir_t *dd, dmu_tx_t *tx);
+void dsl_dir_sync(dsl_dir_t *dd, dmu_tx_t *tx);
+int dsl_dir_tempreserve_space(dsl_dir_t *dd, uint64_t mem,
+ uint64_t asize, uint64_t fsize, uint64_t usize, void **tr_cookiep,
+ dmu_tx_t *tx);
+void dsl_dir_tempreserve_clear(void *tr_cookie, dmu_tx_t *tx);
+void dsl_dir_willuse_space(dsl_dir_t *dd, int64_t space, dmu_tx_t *tx);
+void dsl_dir_diduse_space(dsl_dir_t *dd, dd_used_t type,
+ int64_t used, int64_t compressed, int64_t uncompressed, dmu_tx_t *tx);
+void dsl_dir_transfer_space(dsl_dir_t *dd, int64_t delta,
+ dd_used_t oldtype, dd_used_t newtype, dmu_tx_t *tx);
+int dsl_dir_set_quota(const char *ddname, uint64_t quota);
+int dsl_dir_set_reservation(const char *ddname, uint64_t reservation);
+int dsl_dir_rename(dsl_dir_t *dd, const char *newname);
+int dsl_dir_transfer_possible(dsl_dir_t *sdd, dsl_dir_t *tdd, uint64_t space);
+int dsl_dir_set_reservation_check(void *arg1, void *arg2, dmu_tx_t *tx);
+boolean_t dsl_dir_is_clone(dsl_dir_t *dd);
+void dsl_dir_new_refreservation(dsl_dir_t *dd, struct dsl_dataset *ds,
+ uint64_t reservation, cred_t *cr, dmu_tx_t *tx);
+
+/* internal reserved dir name */
+#define MOS_DIR_NAME "$MOS"
+#define ORIGIN_DIR_NAME "$ORIGIN"
+
+#ifdef ZFS_DEBUG
+#define dprintf_dd(dd, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char *__ds_name = kmem_alloc(MAXNAMELEN + strlen(MOS_DIR_NAME) + 1, \
+ KM_SLEEP); \
+ dsl_dir_name(dd, __ds_name); \
+ dprintf("dd=%s " fmt, __ds_name, __VA_ARGS__); \
+ kmem_free(__ds_name, MAXNAMELEN + strlen(MOS_DIR_NAME) + 1); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+#else
+#define dprintf_dd(dd, fmt, ...)
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_DIR_H */
diff --git a/module/zfs/include/sys/dsl_pool.h b/module/zfs/include/sys/dsl_pool.h
new file mode 100644
index 000000000..3bb4ad4ef
--- /dev/null
+++ b/module/zfs/include/sys/dsl_pool.h
@@ -0,0 +1,150 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_POOL_H
+#define _SYS_DSL_POOL_H
+
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/txg_impl.h>
+#include <sys/zfs_context.h>
+#include <sys/zio.h>
+#include <sys/dnode.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct objset;
+struct dsl_dir;
+struct dsl_dataset;
+struct dsl_pool;
+struct dmu_tx;
+
+enum scrub_func {
+ SCRUB_FUNC_NONE,
+ SCRUB_FUNC_CLEAN,
+ SCRUB_FUNC_NUMFUNCS
+};
+
+/* These macros are for indexing into the zfs_all_blkstats_t. */
+#define DMU_OT_DEFERRED DMU_OT_NONE
+#define DMU_OT_TOTAL DMU_OT_NUMTYPES
+
+typedef struct zfs_blkstat {
+ uint64_t zb_count;
+ uint64_t zb_asize;
+ uint64_t zb_lsize;
+ uint64_t zb_psize;
+ uint64_t zb_gangs;
+ uint64_t zb_ditto_2_of_2_samevdev;
+ uint64_t zb_ditto_2_of_3_samevdev;
+ uint64_t zb_ditto_3_of_3_samevdev;
+} zfs_blkstat_t;
+
+typedef struct zfs_all_blkstats {
+ zfs_blkstat_t zab_type[DN_MAX_LEVELS + 1][DMU_OT_TOTAL + 1];
+} zfs_all_blkstats_t;
+
+
+typedef struct dsl_pool {
+ /* Immutable */
+ spa_t *dp_spa;
+ struct objset *dp_meta_objset;
+ struct dsl_dir *dp_root_dir;
+ struct dsl_dir *dp_mos_dir;
+ struct dsl_dataset *dp_origin_snap;
+ uint64_t dp_root_dir_obj;
+
+ /* No lock needed - sync context only */
+ blkptr_t dp_meta_rootbp;
+ list_t dp_synced_datasets;
+ hrtime_t dp_read_overhead;
+ uint64_t dp_throughput;
+ uint64_t dp_write_limit;
+
+ /* Uses dp_lock */
+ kmutex_t dp_lock;
+ uint64_t dp_space_towrite[TXG_SIZE];
+ uint64_t dp_tempreserved[TXG_SIZE];
+
+ enum scrub_func dp_scrub_func;
+ uint64_t dp_scrub_queue_obj;
+ uint64_t dp_scrub_min_txg;
+ uint64_t dp_scrub_max_txg;
+ zbookmark_t dp_scrub_bookmark;
+ boolean_t dp_scrub_pausing;
+ boolean_t dp_scrub_isresilver;
+ uint64_t dp_scrub_start_time;
+ kmutex_t dp_scrub_cancel_lock; /* protects dp_scrub_restart */
+ boolean_t dp_scrub_restart;
+
+ /* Has its own locking */
+ tx_state_t dp_tx;
+ txg_list_t dp_dirty_datasets;
+ txg_list_t dp_dirty_dirs;
+ txg_list_t dp_sync_tasks;
+
+ /*
+ * Protects administrative changes (properties, namespace)
+ * It is only held for write in syncing context. Therefore
+ * syncing context does not need to ever have it for read, since
+ * nobody else could possibly have it for write.
+ */
+ krwlock_t dp_config_rwlock;
+
+ zfs_all_blkstats_t *dp_blkstats;
+} dsl_pool_t;
+
+int dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp);
+void dsl_pool_close(dsl_pool_t *dp);
+dsl_pool_t *dsl_pool_create(spa_t *spa, nvlist_t *zplprops, uint64_t txg);
+void dsl_pool_sync(dsl_pool_t *dp, uint64_t txg);
+void dsl_pool_zil_clean(dsl_pool_t *dp);
+int dsl_pool_sync_context(dsl_pool_t *dp);
+uint64_t dsl_pool_adjustedsize(dsl_pool_t *dp, boolean_t netfree);
+int dsl_pool_tempreserve_space(dsl_pool_t *dp, uint64_t space, dmu_tx_t *tx);
+void dsl_pool_tempreserve_clear(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx);
+void dsl_pool_memory_pressure(dsl_pool_t *dp);
+void dsl_pool_willuse_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx);
+int dsl_free(zio_t *pio, dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp,
+ zio_done_func_t *done, void *private, uint32_t arc_flags);
+void dsl_pool_ds_destroyed(struct dsl_dataset *ds, struct dmu_tx *tx);
+void dsl_pool_ds_snapshotted(struct dsl_dataset *ds, struct dmu_tx *tx);
+void dsl_pool_ds_clone_swapped(struct dsl_dataset *ds1, struct dsl_dataset *ds2,
+ struct dmu_tx *tx);
+void dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx);
+void dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx);
+
+int dsl_pool_scrub_cancel(dsl_pool_t *dp);
+int dsl_pool_scrub_clean(dsl_pool_t *dp);
+void dsl_pool_scrub_sync(dsl_pool_t *dp, dmu_tx_t *tx);
+void dsl_pool_scrub_restart(dsl_pool_t *dp);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_POOL_H */
diff --git a/module/zfs/include/sys/dsl_prop.h b/module/zfs/include/sys/dsl_prop.h
new file mode 100644
index 000000000..d66caa86c
--- /dev/null
+++ b/module/zfs/include/sys/dsl_prop.h
@@ -0,0 +1,82 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_PROP_H
+#define _SYS_DSL_PROP_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/dmu.h>
+#include <sys/dsl_pool.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dsl_dataset;
+struct dsl_dir;
+
+/* The callback func may not call into the DMU or DSL! */
+typedef void (dsl_prop_changed_cb_t)(void *arg, uint64_t newval);
+
+typedef struct dsl_prop_cb_record {
+ list_node_t cbr_node; /* link on dd_prop_cbs */
+ struct dsl_dataset *cbr_ds;
+ const char *cbr_propname;
+ dsl_prop_changed_cb_t *cbr_func;
+ void *cbr_arg;
+} dsl_prop_cb_record_t;
+
+int dsl_prop_register(struct dsl_dataset *ds, const char *propname,
+ dsl_prop_changed_cb_t *callback, void *cbarg);
+int dsl_prop_unregister(struct dsl_dataset *ds, const char *propname,
+ dsl_prop_changed_cb_t *callback, void *cbarg);
+int dsl_prop_numcb(struct dsl_dataset *ds);
+
+int dsl_prop_get(const char *ddname, const char *propname,
+ int intsz, int numints, void *buf, char *setpoint);
+int dsl_prop_get_integer(const char *ddname, const char *propname,
+ uint64_t *valuep, char *setpoint);
+int dsl_prop_get_all(objset_t *os, nvlist_t **nvp, boolean_t local);
+int dsl_prop_get_ds(struct dsl_dataset *ds, const char *propname,
+ int intsz, int numints, void *buf, char *setpoint);
+int dsl_prop_get_dd(struct dsl_dir *dd, const char *propname,
+ int intsz, int numints, void *buf, char *setpoint);
+
+int dsl_prop_set(const char *ddname, const char *propname,
+ int intsz, int numints, const void *buf);
+void dsl_prop_set_uint64_sync(dsl_dir_t *dd, const char *name, uint64_t val,
+ cred_t *cr, dmu_tx_t *tx);
+
+void dsl_prop_nvlist_add_uint64(nvlist_t *nv, zfs_prop_t prop, uint64_t value);
+void dsl_prop_nvlist_add_string(nvlist_t *nv,
+ zfs_prop_t prop, const char *value);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_PROP_H */
diff --git a/module/zfs/include/sys/dsl_synctask.h b/module/zfs/include/sys/dsl_synctask.h
new file mode 100644
index 000000000..4995bfe5a
--- /dev/null
+++ b/module/zfs/include/sys/dsl_synctask.h
@@ -0,0 +1,83 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_DSL_SYNCTASK_H
+#define _SYS_DSL_SYNCTASK_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/txg.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct dsl_pool;
+
+typedef int (dsl_checkfunc_t)(void *, void *, dmu_tx_t *);
+typedef void (dsl_syncfunc_t)(void *, void *, cred_t *, dmu_tx_t *);
+
+typedef struct dsl_sync_task {
+ list_node_t dst_node;
+ dsl_checkfunc_t *dst_checkfunc;
+ dsl_syncfunc_t *dst_syncfunc;
+ void *dst_arg1;
+ void *dst_arg2;
+ int dst_err;
+} dsl_sync_task_t;
+
+typedef struct dsl_sync_task_group {
+ txg_node_t dstg_node;
+ list_t dstg_tasks;
+ struct dsl_pool *dstg_pool;
+ cred_t *dstg_cr;
+ uint64_t dstg_txg;
+ int dstg_err;
+ int dstg_space;
+ boolean_t dstg_nowaiter;
+} dsl_sync_task_group_t;
+
+dsl_sync_task_group_t *dsl_sync_task_group_create(struct dsl_pool *dp);
+void dsl_sync_task_create(dsl_sync_task_group_t *dstg,
+ dsl_checkfunc_t *, dsl_syncfunc_t *,
+ void *arg1, void *arg2, int blocks_modified);
+int dsl_sync_task_group_wait(dsl_sync_task_group_t *dstg);
+void dsl_sync_task_group_nowait(dsl_sync_task_group_t *dstg, dmu_tx_t *tx);
+void dsl_sync_task_group_destroy(dsl_sync_task_group_t *dstg);
+void dsl_sync_task_group_sync(dsl_sync_task_group_t *dstg, dmu_tx_t *tx);
+
+int dsl_sync_task_do(struct dsl_pool *dp,
+ dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
+ void *arg1, void *arg2, int blocks_modified);
+void dsl_sync_task_do_nowait(struct dsl_pool *dp,
+ dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
+ void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_DSL_SYNCTASK_H */
diff --git a/module/zfs/include/sys/metaslab.h b/module/zfs/include/sys/metaslab.h
new file mode 100644
index 000000000..1c9d89e8f
--- /dev/null
+++ b/module/zfs/include/sys/metaslab.h
@@ -0,0 +1,71 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_METASLAB_H
+#define _SYS_METASLAB_H
+
+#include <sys/spa.h>
+#include <sys/space_map.h>
+#include <sys/txg.h>
+#include <sys/zio.h>
+#include <sys/avl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct metaslab_class metaslab_class_t;
+typedef struct metaslab_group metaslab_group_t;
+
+extern metaslab_t *metaslab_init(metaslab_group_t *mg, space_map_obj_t *smo,
+ uint64_t start, uint64_t size, uint64_t txg);
+extern void metaslab_fini(metaslab_t *msp);
+extern void metaslab_sync(metaslab_t *msp, uint64_t txg);
+extern void metaslab_sync_done(metaslab_t *msp, uint64_t txg);
+
+#define METASLAB_HINTBP_FAVOR 0x0
+#define METASLAB_HINTBP_AVOID 0x1
+#define METASLAB_GANG_HEADER 0x2
+
+extern int metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
+ blkptr_t *bp, int ncopies, uint64_t txg, blkptr_t *hintbp, int flags);
+extern void metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg,
+ boolean_t now);
+extern int metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg);
+
+extern metaslab_class_t *metaslab_class_create(void);
+extern void metaslab_class_destroy(metaslab_class_t *mc);
+extern void metaslab_class_add(metaslab_class_t *mc, metaslab_group_t *mg);
+extern void metaslab_class_remove(metaslab_class_t *mc, metaslab_group_t *mg);
+
+extern metaslab_group_t *metaslab_group_create(metaslab_class_t *mc,
+ vdev_t *vd);
+extern void metaslab_group_destroy(metaslab_group_t *mg);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_METASLAB_H */
diff --git a/module/zfs/include/sys/metaslab_impl.h b/module/zfs/include/sys/metaslab_impl.h
new file mode 100644
index 000000000..5980cbc84
--- /dev/null
+++ b/module/zfs/include/sys/metaslab_impl.h
@@ -0,0 +1,81 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_METASLAB_IMPL_H
+#define _SYS_METASLAB_IMPL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/metaslab.h>
+#include <sys/space_map.h>
+#include <sys/vdev.h>
+#include <sys/txg.h>
+#include <sys/avl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct metaslab_class {
+ metaslab_group_t *mc_rotor;
+ uint64_t mc_allocated;
+};
+
+struct metaslab_group {
+ kmutex_t mg_lock;
+ avl_tree_t mg_metaslab_tree;
+ uint64_t mg_aliquot;
+ int64_t mg_bias;
+ metaslab_class_t *mg_class;
+ vdev_t *mg_vd;
+ metaslab_group_t *mg_prev;
+ metaslab_group_t *mg_next;
+};
+
+/*
+ * Each metaslab's free space is tracked in space map object in the MOS,
+ * which is only updated in syncing context. Each time we sync a txg,
+ * we append the allocs and frees from that txg to the space map object.
+ * When the txg is done syncing, metaslab_sync_done() updates ms_smo
+ * to ms_smo_syncing. Everything in ms_smo is always safe to allocate.
+ */
+struct metaslab {
+ kmutex_t ms_lock; /* metaslab lock */
+ space_map_obj_t ms_smo; /* synced space map object */
+ space_map_obj_t ms_smo_syncing; /* syncing space map object */
+ space_map_t ms_allocmap[TXG_SIZE]; /* allocated this txg */
+ space_map_t ms_freemap[TXG_SIZE]; /* freed this txg */
+ space_map_t ms_map; /* in-core free space map */
+ uint64_t ms_weight; /* weight vs. others in group */
+ metaslab_group_t *ms_group; /* metaslab group */
+ avl_node_t ms_group_node; /* node in metaslab group tree */
+ txg_node_t ms_txg_node; /* per-txg dirty metaslab links */
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_METASLAB_IMPL_H */
diff --git a/module/zfs/include/sys/refcount.h b/module/zfs/include/sys/refcount.h
new file mode 100644
index 000000000..d3fe7b1f8
--- /dev/null
+++ b/module/zfs/include/sys/refcount.h
@@ -0,0 +1,104 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_REFCOUNT_H
+#define _SYS_REFCOUNT_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/inttypes.h>
+#include <sys/list.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * If the reference is held only by the calling function and not any
+ * particular object, use FTAG (which is a string) for the holder_tag.
+ * Otherwise, use the object that holds the reference.
+ */
+#define FTAG ((char *)__func__)
+
+#if defined(DEBUG) || !defined(_KERNEL)
+typedef struct reference {
+ list_node_t ref_link;
+ void *ref_holder;
+ uint64_t ref_number;
+ uint8_t *ref_removed;
+} reference_t;
+
+typedef struct refcount {
+ kmutex_t rc_mtx;
+ list_t rc_list;
+ list_t rc_removed;
+ int64_t rc_count;
+ int64_t rc_removed_count;
+} refcount_t;
+
+/* Note: refcount_t must be initialized with refcount_create() */
+
+void refcount_create(refcount_t *rc);
+void refcount_destroy(refcount_t *rc);
+void refcount_destroy_many(refcount_t *rc, uint64_t number);
+int refcount_is_zero(refcount_t *rc);
+int64_t refcount_count(refcount_t *rc);
+int64_t refcount_add(refcount_t *rc, void *holder_tag);
+int64_t refcount_remove(refcount_t *rc, void *holder_tag);
+int64_t refcount_add_many(refcount_t *rc, uint64_t number, void *holder_tag);
+int64_t refcount_remove_many(refcount_t *rc, uint64_t number, void *holder_tag);
+
+void refcount_init(void);
+void refcount_fini(void);
+
+#else /* DEBUG */
+
+typedef struct refcount {
+ uint64_t rc_count;
+} refcount_t;
+
+#define refcount_create(rc) ((rc)->rc_count = 0)
+#define refcount_destroy(rc) ((rc)->rc_count = 0)
+#define refcount_destroy_many(rc, number) ((rc)->rc_count = 0)
+#define refcount_is_zero(rc) ((rc)->rc_count == 0)
+#define refcount_count(rc) ((rc)->rc_count)
+#define refcount_add(rc, holder) atomic_add_64_nv(&(rc)->rc_count, 1)
+#define refcount_remove(rc, holder) atomic_add_64_nv(&(rc)->rc_count, -1)
+#define refcount_add_many(rc, number, holder) \
+ atomic_add_64_nv(&(rc)->rc_count, number)
+#define refcount_remove_many(rc, number, holder) \
+ atomic_add_64_nv(&(rc)->rc_count, -number)
+
+#define refcount_init()
+#define refcount_fini()
+
+#endif /* DEBUG */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_REFCOUNT_H */
diff --git a/module/zfs/include/sys/rrwlock.h b/module/zfs/include/sys/rrwlock.h
new file mode 100644
index 000000000..19a43c97f
--- /dev/null
+++ b/module/zfs/include/sys/rrwlock.h
@@ -0,0 +1,80 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_RR_RW_LOCK_H
+#define _SYS_RR_RW_LOCK_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <sys/inttypes.h>
+#include <sys/zfs_context.h>
+#include <sys/refcount.h>
+
+/*
+ * A reader-writer lock implementation that allows re-entrant reads, but
+ * still gives writers priority on "new" reads.
+ *
+ * See rrwlock.c for more details about the implementation.
+ *
+ * Fields of the rrwlock_t structure:
+ * - rr_lock: protects modification and reading of rrwlock_t fields
+ * - rr_cv: cv for waking up readers or waiting writers
+ * - rr_writer: thread id of the current writer
+ * - rr_anon_rount: number of active anonymous readers
+ * - rr_linked_rcount: total number of non-anonymous active readers
+ * - rr_writer_wanted: a writer wants the lock
+ */
+typedef struct rrwlock {
+ kmutex_t rr_lock;
+ kcondvar_t rr_cv;
+ kthread_t *rr_writer;
+ refcount_t rr_anon_rcount;
+ refcount_t rr_linked_rcount;
+ boolean_t rr_writer_wanted;
+} rrwlock_t;
+
+/*
+ * 'tag' is used in reference counting tracking. The
+ * 'tag' must be the same in a rrw_enter() as in its
+ * corresponding rrw_exit().
+ */
+void rrw_init(rrwlock_t *rrl);
+void rrw_destroy(rrwlock_t *rrl);
+void rrw_enter(rrwlock_t *rrl, krw_t rw, void *tag);
+void rrw_exit(rrwlock_t *rrl, void *tag);
+boolean_t rrw_held(rrwlock_t *rrl, krw_t rw);
+
+#define RRW_READ_HELD(x) rrw_held(x, RW_READER)
+#define RRW_WRITE_HELD(x) rrw_held(x, RW_WRITER)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_RR_RW_LOCK_H */
diff --git a/module/zfs/include/sys/spa.h b/module/zfs/include/sys/spa.h
new file mode 100644
index 000000000..24b3ca447
--- /dev/null
+++ b/module/zfs/include/sys/spa.h
@@ -0,0 +1,554 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_SPA_H
+#define _SYS_SPA_H
+
+#include <sys/avl.h>
+#include <sys/zfs_context.h>
+#include <sys/nvpair.h>
+#include <sys/sysmacros.h>
+#include <sys/types.h>
+#include <sys/fs/zfs.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Forward references that lots of things need.
+ */
+typedef struct spa spa_t;
+typedef struct vdev vdev_t;
+typedef struct metaslab metaslab_t;
+typedef struct zilog zilog_t;
+typedef struct spa_aux_vdev spa_aux_vdev_t;
+struct dsl_pool;
+
+/*
+ * General-purpose 32-bit and 64-bit bitfield encodings.
+ */
+#define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len))
+#define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len))
+#define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low))
+#define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low))
+
+#define BF32_GET(x, low, len) BF32_DECODE(x, low, len)
+#define BF64_GET(x, low, len) BF64_DECODE(x, low, len)
+
+#define BF32_SET(x, low, len, val) \
+ ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
+#define BF64_SET(x, low, len, val) \
+ ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
+
+#define BF32_GET_SB(x, low, len, shift, bias) \
+ ((BF32_GET(x, low, len) + (bias)) << (shift))
+#define BF64_GET_SB(x, low, len, shift, bias) \
+ ((BF64_GET(x, low, len) + (bias)) << (shift))
+
+#define BF32_SET_SB(x, low, len, shift, bias, val) \
+ BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
+#define BF64_SET_SB(x, low, len, shift, bias, val) \
+ BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
+
+/*
+ * We currently support nine block sizes, from 512 bytes to 128K.
+ * We could go higher, but the benefits are near-zero and the cost
+ * of COWing a giant block to modify one byte would become excessive.
+ */
+#define SPA_MINBLOCKSHIFT 9
+#define SPA_MAXBLOCKSHIFT 17
+#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
+#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
+
+#define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
+
+/*
+ * Size of block to hold the configuration data (a packed nvlist)
+ */
+#define SPA_CONFIG_BLOCKSIZE (1 << 14)
+
+/*
+ * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
+ * The ASIZE encoding should be at least 64 times larger (6 more bits)
+ * to support up to 4-way RAID-Z mirror mode with worst-case gang block
+ * overhead, three DVAs per bp, plus one more bit in case we do anything
+ * else that expands the ASIZE.
+ */
+#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
+#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
+#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
+
+/*
+ * All SPA data is represented by 128-bit data virtual addresses (DVAs).
+ * The members of the dva_t should be considered opaque outside the SPA.
+ */
+typedef struct dva {
+ uint64_t dva_word[2];
+} dva_t;
+
+/*
+ * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
+ */
+typedef struct zio_cksum {
+ uint64_t zc_word[4];
+} zio_cksum_t;
+
+/*
+ * Each block is described by its DVAs, time of birth, checksum, etc.
+ * The word-by-word, bit-by-bit layout of the blkptr is as follows:
+ *
+ * 64 56 48 40 32 24 16 8 0
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 0 | vdev1 | GRID | ASIZE |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 1 |G| offset1 |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 2 | vdev2 | GRID | ASIZE |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 3 |G| offset2 |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 4 | vdev3 | GRID | ASIZE |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 5 |G| offset3 |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 6 |E| lvl | type | cksum | comp | PSIZE | LSIZE |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 7 | padding |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 8 | padding |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * 9 | padding |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * a | birth txg |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * b | fill count |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * c | checksum[0] |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * d | checksum[1] |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * e | checksum[2] |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ * f | checksum[3] |
+ * +-------+-------+-------+-------+-------+-------+-------+-------+
+ *
+ * Legend:
+ *
+ * vdev virtual device ID
+ * offset offset into virtual device
+ * LSIZE logical size
+ * PSIZE physical size (after compression)
+ * ASIZE allocated size (including RAID-Z parity and gang block headers)
+ * GRID RAID-Z layout information (reserved for future use)
+ * cksum checksum function
+ * comp compression function
+ * G gang block indicator
+ * E endianness
+ * type DMU object type
+ * lvl level of indirection
+ * birth txg transaction group in which the block was born
+ * fill count number of non-zero blocks under this bp
+ * checksum[4] 256-bit checksum of the data this bp describes
+ */
+typedef struct blkptr {
+ dva_t blk_dva[3]; /* 128-bit Data Virtual Address */
+ uint64_t blk_prop; /* size, compression, type, etc */
+ uint64_t blk_pad[3]; /* Extra space for the future */
+ uint64_t blk_birth; /* transaction group at birth */
+ uint64_t blk_fill; /* fill count */
+ zio_cksum_t blk_cksum; /* 256-bit checksum */
+} blkptr_t;
+
+#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
+#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
+
+/*
+ * Macros to get and set fields in a bp or DVA.
+ */
+#define DVA_GET_ASIZE(dva) \
+ BF64_GET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0)
+#define DVA_SET_ASIZE(dva, x) \
+ BF64_SET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0, x)
+
+#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
+#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
+
+#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32)
+#define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x)
+
+#define DVA_GET_OFFSET(dva) \
+ BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
+#define DVA_SET_OFFSET(dva, x) \
+ BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
+
+#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
+#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
+
+#define BP_GET_LSIZE(bp) \
+ (BP_IS_HOLE(bp) ? 0 : \
+ BF64_GET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1))
+#define BP_SET_LSIZE(bp, x) \
+ BF64_SET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
+
+#define BP_GET_PSIZE(bp) \
+ BF64_GET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1)
+#define BP_SET_PSIZE(bp, x) \
+ BF64_SET_SB((bp)->blk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x)
+
+#define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 8)
+#define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 8, x)
+
+#define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8)
+#define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x)
+
+#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
+#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
+
+#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
+#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
+
+#define BP_GET_BYTEORDER(bp) (0 - BF64_GET((bp)->blk_prop, 63, 1))
+#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
+
+#define BP_GET_ASIZE(bp) \
+ (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
+ DVA_GET_ASIZE(&(bp)->blk_dva[2]))
+
+#define BP_GET_UCSIZE(bp) \
+ ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \
+ BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
+
+#define BP_GET_NDVAS(bp) \
+ (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
+ !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
+ !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
+
+#define BP_COUNT_GANG(bp) \
+ (DVA_GET_GANG(&(bp)->blk_dva[0]) + \
+ DVA_GET_GANG(&(bp)->blk_dva[1]) + \
+ DVA_GET_GANG(&(bp)->blk_dva[2]))
+
+#define DVA_EQUAL(dva1, dva2) \
+ ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
+ (dva1)->dva_word[0] == (dva2)->dva_word[0])
+
+#define ZIO_CHECKSUM_EQUAL(zc1, zc2) \
+ (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \
+ ((zc1).zc_word[1] - (zc2).zc_word[1]) | \
+ ((zc1).zc_word[2] - (zc2).zc_word[2]) | \
+ ((zc1).zc_word[3] - (zc2).zc_word[3])))
+
+#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
+
+#define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \
+{ \
+ (zcp)->zc_word[0] = w0; \
+ (zcp)->zc_word[1] = w1; \
+ (zcp)->zc_word[2] = w2; \
+ (zcp)->zc_word[3] = w3; \
+}
+
+#define BP_IDENTITY(bp) (&(bp)->blk_dva[0])
+#define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp))
+#define BP_IS_HOLE(bp) ((bp)->blk_birth == 0)
+#define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg))
+
+#define BP_ZERO(bp) \
+{ \
+ (bp)->blk_dva[0].dva_word[0] = 0; \
+ (bp)->blk_dva[0].dva_word[1] = 0; \
+ (bp)->blk_dva[1].dva_word[0] = 0; \
+ (bp)->blk_dva[1].dva_word[1] = 0; \
+ (bp)->blk_dva[2].dva_word[0] = 0; \
+ (bp)->blk_dva[2].dva_word[1] = 0; \
+ (bp)->blk_prop = 0; \
+ (bp)->blk_pad[0] = 0; \
+ (bp)->blk_pad[1] = 0; \
+ (bp)->blk_pad[2] = 0; \
+ (bp)->blk_birth = 0; \
+ (bp)->blk_fill = 0; \
+ ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
+}
+
+#define BLK_FILL_ALREADY_FREED (-1ULL)
+
+/*
+ * Note: the byteorder is either 0 or -1, both of which are palindromes.
+ * This simplifies the endianness handling a bit.
+ */
+#ifdef _BIG_ENDIAN
+#define ZFS_HOST_BYTEORDER (0ULL)
+#else
+#define ZFS_HOST_BYTEORDER (-1ULL)
+#endif
+
+#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
+
+#define BP_SPRINTF_LEN 320
+
+#include <sys/dmu.h>
+
+#define BP_GET_BUFC_TYPE(bp) \
+ (((BP_GET_LEVEL(bp) > 0) || (dmu_ot[BP_GET_TYPE(bp)].ot_metadata)) ? \
+ ARC_BUFC_METADATA : ARC_BUFC_DATA);
+/*
+ * Routines found in spa.c
+ */
+
+/* state manipulation functions */
+extern int spa_open(const char *pool, spa_t **, void *tag);
+extern int spa_get_stats(const char *pool, nvlist_t **config,
+ char *altroot, size_t buflen);
+extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
+ const char *history_str, nvlist_t *zplprops);
+extern int spa_check_rootconf(char *devpath, char *devid,
+ nvlist_t **bestconf, uint64_t *besttxg);
+extern boolean_t spa_rootdev_validate(nvlist_t *nv);
+extern int spa_import_rootpool(char *devpath, char *devid);
+extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props);
+extern int spa_import_faulted(const char *, nvlist_t *, nvlist_t *);
+extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
+extern int spa_destroy(char *pool);
+extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force);
+extern int spa_reset(char *pool);
+extern void spa_async_request(spa_t *spa, int flag);
+extern void spa_async_unrequest(spa_t *spa, int flag);
+extern void spa_async_suspend(spa_t *spa);
+extern void spa_async_resume(spa_t *spa);
+extern spa_t *spa_inject_addref(char *pool);
+extern void spa_inject_delref(spa_t *spa);
+
+#define SPA_ASYNC_CONFIG_UPDATE 0x01
+#define SPA_ASYNC_REMOVE 0x02
+#define SPA_ASYNC_PROBE 0x04
+#define SPA_ASYNC_RESILVER_DONE 0x08
+#define SPA_ASYNC_RESILVER 0x10
+
+/* device manipulation */
+extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
+extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
+ int replacing);
+extern int spa_vdev_detach(spa_t *spa, uint64_t guid, int replace_done);
+extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
+extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
+
+/* spare state (which is global across all pools) */
+extern void spa_spare_add(vdev_t *vd);
+extern void spa_spare_remove(vdev_t *vd);
+extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
+extern void spa_spare_activate(vdev_t *vd);
+
+/* L2ARC state (which is global across all pools) */
+extern void spa_l2cache_add(vdev_t *vd);
+extern void spa_l2cache_remove(vdev_t *vd);
+extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
+extern void spa_l2cache_activate(vdev_t *vd);
+extern void spa_l2cache_drop(spa_t *spa);
+extern void spa_l2cache_space_update(vdev_t *vd, int64_t space, int64_t alloc);
+
+/* scrubbing */
+extern int spa_scrub(spa_t *spa, pool_scrub_type_t type);
+
+/* spa syncing */
+extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
+extern void spa_sync_allpools(void);
+
+/* spa namespace global mutex */
+extern kmutex_t spa_namespace_lock;
+
+/*
+ * SPA configuration functions in spa_config.c
+ */
+
+#define SPA_CONFIG_UPDATE_POOL 0
+#define SPA_CONFIG_UPDATE_VDEVS 1
+
+extern void spa_config_sync(spa_t *, boolean_t, boolean_t);
+extern void spa_config_load(void);
+extern nvlist_t *spa_all_configs(uint64_t *);
+extern void spa_config_set(spa_t *spa, nvlist_t *config);
+extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
+ int getstats);
+extern void spa_config_update(spa_t *spa, int what);
+extern void spa_config_update_common(spa_t *spa, int what, boolean_t isroot);
+
+/*
+ * Miscellaneous SPA routines in spa_misc.c
+ */
+
+/* Namespace manipulation */
+extern spa_t *spa_lookup(const char *name);
+extern spa_t *spa_add(const char *name, const char *altroot);
+extern void spa_remove(spa_t *spa);
+extern spa_t *spa_next(spa_t *prev);
+
+/* Refcount functions */
+extern void spa_open_ref(spa_t *spa, void *tag);
+extern void spa_close(spa_t *spa, void *tag);
+extern boolean_t spa_refcount_zero(spa_t *spa);
+
+#define SCL_CONFIG 0x01
+#define SCL_STATE 0x02
+#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
+#define SCL_ALLOC 0x08
+#define SCL_ZIO 0x10
+#define SCL_FREE 0x20
+#define SCL_VDEV 0x40
+#define SCL_LOCKS 7
+#define SCL_ALL ((1 << SCL_LOCKS) - 1)
+#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
+
+/* Pool configuration locks */
+extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
+extern void spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw);
+extern void spa_config_exit(spa_t *spa, int locks, void *tag);
+extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
+
+/* Pool vdev add/remove lock */
+extern uint64_t spa_vdev_enter(spa_t *spa);
+extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
+
+/* Pool vdev state change lock */
+extern void spa_vdev_state_enter(spa_t *spa);
+extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
+
+/* Accessor functions */
+extern boolean_t spa_shutting_down(spa_t *spa);
+extern struct dsl_pool *spa_get_dsl(spa_t *spa);
+extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
+extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
+extern void spa_altroot(spa_t *, char *, size_t);
+extern int spa_sync_pass(spa_t *spa);
+extern char *spa_name(spa_t *spa);
+extern uint64_t spa_guid(spa_t *spa);
+extern uint64_t spa_last_synced_txg(spa_t *spa);
+extern uint64_t spa_first_txg(spa_t *spa);
+extern uint64_t spa_version(spa_t *spa);
+extern pool_state_t spa_state(spa_t *spa);
+extern uint64_t spa_freeze_txg(spa_t *spa);
+extern uint64_t spa_get_alloc(spa_t *spa);
+extern uint64_t spa_get_space(spa_t *spa);
+extern uint64_t spa_get_dspace(spa_t *spa);
+extern uint64_t spa_get_asize(spa_t *spa, uint64_t lsize);
+extern uint64_t spa_version(spa_t *spa);
+extern int spa_max_replication(spa_t *spa);
+extern int spa_busy(void);
+extern uint8_t spa_get_failmode(spa_t *spa);
+extern boolean_t spa_suspended(spa_t *spa);
+
+/* Miscellaneous support routines */
+extern int spa_rename(const char *oldname, const char *newname);
+extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
+extern char *spa_strdup(const char *);
+extern void spa_strfree(char *);
+extern uint64_t spa_get_random(uint64_t range);
+extern void sprintf_blkptr(char *buf, int len, const blkptr_t *bp);
+extern void spa_freeze(spa_t *spa);
+extern void spa_upgrade(spa_t *spa, uint64_t version);
+extern void spa_evict_all(void);
+extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
+ boolean_t l2cache);
+extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
+extern uint64_t bp_get_dasize(spa_t *spa, const blkptr_t *bp);
+extern boolean_t spa_has_slogs(spa_t *spa);
+extern boolean_t spa_is_root(spa_t *spa);
+
+/* history logging */
+typedef enum history_log_type {
+ LOG_CMD_POOL_CREATE,
+ LOG_CMD_NORMAL,
+ LOG_INTERNAL
+} history_log_type_t;
+
+typedef struct history_arg {
+ const char *ha_history_str;
+ history_log_type_t ha_log_type;
+ history_internal_events_t ha_event;
+ char ha_zone[MAXPATHLEN];
+} history_arg_t;
+
+extern char *spa_his_ievent_table[];
+
+extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
+extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
+ char *his_buf);
+extern int spa_history_log(spa_t *spa, const char *his_buf,
+ history_log_type_t what);
+void spa_history_internal_log(history_internal_events_t event, spa_t *spa,
+ dmu_tx_t *tx, cred_t *cr, const char *fmt, ...);
+
+/* error handling */
+struct zbookmark;
+struct zio;
+extern void spa_log_error(spa_t *spa, struct zio *zio);
+extern void zfs_ereport_post(const char *class, spa_t *spa, vdev_t *vd,
+ struct zio *zio, uint64_t stateoroffset, uint64_t length);
+extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
+extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
+extern uint64_t spa_get_errlog_size(spa_t *spa);
+extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
+extern void spa_errlog_rotate(spa_t *spa);
+extern void spa_errlog_drain(spa_t *spa);
+extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
+extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
+
+/* vdev cache */
+extern void vdev_cache_stat_init(void);
+extern void vdev_cache_stat_fini(void);
+
+/* Initialization and termination */
+extern void spa_init(int flags);
+extern void spa_fini(void);
+extern void spa_boot_init();
+
+/* properties */
+extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
+extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
+extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
+
+/* asynchronous event notification */
+extern void spa_event_notify(spa_t *spa, vdev_t *vdev, const char *name);
+
+#ifdef ZFS_DEBUG
+#define dprintf_bp(bp, fmt, ...) do { \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
+ char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
+ sprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
+ dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
+ kmem_free(__blkbuf, BP_SPRINTF_LEN); \
+ } \
+_NOTE(CONSTCOND) } while (0)
+#else
+#define dprintf_bp(bp, fmt, ...)
+#endif
+
+extern int spa_mode; /* mode, e.g. FREAD | FWRITE */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_SPA_H */
diff --git a/module/zfs/include/sys/spa_boot.h b/module/zfs/include/sys/spa_boot.h
new file mode 100644
index 000000000..b56073b97
--- /dev/null
+++ b/module/zfs/include/sys/spa_boot.h
@@ -0,0 +1,45 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_SPA_BOOT_H
+#define _SYS_SPA_BOOT_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/nvpair.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern char *spa_get_bootprop(char *prop);
+extern void spa_free_bootprop(char *prop);
+extern int spa_get_rootconf(char *devpath, char *devid, nvlist_t **bestconf_p);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_SPA_BOOT_H */
diff --git a/module/zfs/include/sys/spa_impl.h b/module/zfs/include/sys/spa_impl.h
new file mode 100644
index 000000000..8aeb414fe
--- /dev/null
+++ b/module/zfs/include/sys/spa_impl.h
@@ -0,0 +1,196 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_SPA_IMPL_H
+#define _SYS_SPA_IMPL_H
+
+#include <sys/spa.h>
+#include <sys/vdev.h>
+#include <sys/metaslab.h>
+#include <sys/dmu.h>
+#include <sys/dsl_pool.h>
+#include <sys/uberblock_impl.h>
+#include <sys/zfs_context.h>
+#include <sys/avl.h>
+#include <sys/refcount.h>
+#include <sys/bplist.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct spa_error_entry {
+ zbookmark_t se_bookmark;
+ char *se_name;
+ avl_node_t se_avl;
+} spa_error_entry_t;
+
+typedef struct spa_history_phys {
+ uint64_t sh_pool_create_len; /* ending offset of zpool create */
+ uint64_t sh_phys_max_off; /* physical EOF */
+ uint64_t sh_bof; /* logical BOF */
+ uint64_t sh_eof; /* logical EOF */
+ uint64_t sh_records_lost; /* num of records overwritten */
+} spa_history_phys_t;
+
+struct spa_aux_vdev {
+ uint64_t sav_object; /* MOS object for device list */
+ nvlist_t *sav_config; /* cached device config */
+ vdev_t **sav_vdevs; /* devices */
+ int sav_count; /* number devices */
+ boolean_t sav_sync; /* sync the device list */
+ nvlist_t **sav_pending; /* pending device additions */
+ uint_t sav_npending; /* # pending devices */
+};
+
+typedef struct spa_config_lock {
+ kmutex_t scl_lock;
+ kthread_t *scl_writer;
+ int scl_write_wanted;
+ kcondvar_t scl_cv;
+ refcount_t scl_count;
+} spa_config_lock_t;
+
+typedef struct spa_config_dirent {
+ list_node_t scd_link;
+ char *scd_path;
+} spa_config_dirent_t;
+
+typedef enum spa_log_state {
+ SPA_LOG_UNKNOWN = 0, /* unknown log state */
+ SPA_LOG_MISSING, /* missing log(s) */
+ SPA_LOG_CLEAR, /* clear the log(s) */
+ SPA_LOG_GOOD, /* log(s) are good */
+} spa_log_state_t;
+
+enum zio_taskq_type {
+ ZIO_TASKQ_ISSUE = 0,
+ ZIO_TASKQ_INTERRUPT,
+ ZIO_TASKQ_TYPES
+};
+
+struct spa {
+ /*
+ * Fields protected by spa_namespace_lock.
+ */
+ char spa_name[MAXNAMELEN]; /* pool name */
+ avl_node_t spa_avl; /* node in spa_namespace_avl */
+ nvlist_t *spa_config; /* last synced config */
+ nvlist_t *spa_config_syncing; /* currently syncing config */
+ uint64_t spa_config_txg; /* txg of last config change */
+ int spa_sync_pass; /* iterate-to-convergence */
+ pool_state_t spa_state; /* pool state */
+ int spa_inject_ref; /* injection references */
+ uint8_t spa_sync_on; /* sync threads are running */
+ spa_load_state_t spa_load_state; /* current load operation */
+ taskq_t *spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
+ dsl_pool_t *spa_dsl_pool;
+ metaslab_class_t *spa_normal_class; /* normal data class */
+ metaslab_class_t *spa_log_class; /* intent log data class */
+ uint64_t spa_first_txg; /* first txg after spa_open() */
+ uint64_t spa_final_txg; /* txg of export/destroy */
+ uint64_t spa_freeze_txg; /* freeze pool at this txg */
+ objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */
+ txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
+ vdev_t *spa_root_vdev; /* top-level vdev container */
+ uint64_t spa_load_guid; /* initial guid for spa_load */
+ list_t spa_config_dirty_list; /* vdevs with dirty config */
+ list_t spa_state_dirty_list; /* vdevs with dirty state */
+ spa_aux_vdev_t spa_spares; /* hot spares */
+ spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
+ uint64_t spa_config_object; /* MOS object for pool config */
+ uint64_t spa_syncing_txg; /* txg currently syncing */
+ uint64_t spa_sync_bplist_obj; /* object for deferred frees */
+ bplist_t spa_sync_bplist; /* deferred-free bplist */
+ uberblock_t spa_ubsync; /* last synced uberblock */
+ uberblock_t spa_uberblock; /* current uberblock */
+ kmutex_t spa_scrub_lock; /* resilver/scrub lock */
+ uint64_t spa_scrub_inflight; /* in-flight scrub I/Os */
+ uint64_t spa_scrub_maxinflight; /* max in-flight scrub I/Os */
+ uint64_t spa_scrub_errors; /* scrub I/O error count */
+ kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
+ uint8_t spa_scrub_active; /* active or suspended? */
+ uint8_t spa_scrub_type; /* type of scrub we're doing */
+ uint8_t spa_scrub_finished; /* indicator to rotate logs */
+ uint8_t spa_scrub_started; /* started since last boot */
+ uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
+ kmutex_t spa_async_lock; /* protect async state */
+ kthread_t *spa_async_thread; /* thread doing async task */
+ int spa_async_suspended; /* async tasks suspended */
+ kcondvar_t spa_async_cv; /* wait for thread_exit() */
+ uint16_t spa_async_tasks; /* async task mask */
+ kmutex_t spa_async_root_lock; /* protects async root count */
+ uint64_t spa_async_root_count; /* number of async root zios */
+ kcondvar_t spa_async_root_cv; /* notify when count == 0 */
+ char *spa_root; /* alternate root directory */
+ uint64_t spa_ena; /* spa-wide ereport ENA */
+ boolean_t spa_last_open_failed; /* true if last open faled */
+ kmutex_t spa_errlog_lock; /* error log lock */
+ uint64_t spa_errlog_last; /* last error log object */
+ uint64_t spa_errlog_scrub; /* scrub error log object */
+ kmutex_t spa_errlist_lock; /* error list/ereport lock */
+ avl_tree_t spa_errlist_last; /* last error list */
+ avl_tree_t spa_errlist_scrub; /* scrub error list */
+ uint64_t spa_deflate; /* should we deflate? */
+ uint64_t spa_history; /* history object */
+ kmutex_t spa_history_lock; /* history lock */
+ vdev_t *spa_pending_vdev; /* pending vdev additions */
+ kmutex_t spa_props_lock; /* property lock */
+ uint64_t spa_pool_props_object; /* object for properties */
+ uint64_t spa_bootfs; /* default boot filesystem */
+ uint64_t spa_failmode; /* failure mode for the pool */
+ uint64_t spa_delegation; /* delegation on/off */
+ list_t spa_config_list; /* previous cache file(s) */
+ zio_t *spa_suspend_zio_root; /* root of all suspended I/O */
+ kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
+ kcondvar_t spa_suspend_cv; /* notification of resume */
+ uint8_t spa_suspended; /* pool is suspended */
+ boolean_t spa_import_faulted; /* allow faulted vdevs */
+ boolean_t spa_is_root; /* pool is root */
+ int spa_minref; /* num refs when first opened */
+ spa_log_state_t spa_log_state; /* log state */
+ /*
+ * spa_refcnt & spa_config_lock must be the last elements
+ * because refcount_t changes size based on compilation options.
+ * In order for the MDB module to function correctly, the other
+ * fields must remain in the same location.
+ */
+ spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
+ refcount_t spa_refcount; /* number of opens */
+};
+
+extern const char *spa_config_path;
+
+#define BOOTFS_COMPRESS_VALID(compress) \
+ ((compress) == ZIO_COMPRESS_LZJB || \
+ ((compress) == ZIO_COMPRESS_ON && \
+ ZIO_COMPRESS_ON_VALUE == ZIO_COMPRESS_LZJB) || \
+ (compress) == ZIO_COMPRESS_OFF)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_SPA_IMPL_H */
diff --git a/module/zfs/include/sys/space_map.h b/module/zfs/include/sys/space_map.h
new file mode 100644
index 000000000..db9daef1f
--- /dev/null
+++ b/module/zfs/include/sys/space_map.h
@@ -0,0 +1,162 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_SPACE_MAP_H
+#define _SYS_SPACE_MAP_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/avl.h>
+#include <sys/dmu.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct space_map_ops space_map_ops_t;
+
+typedef struct space_map {
+ avl_tree_t sm_root; /* AVL tree of map segments */
+ uint64_t sm_space; /* sum of all segments in the map */
+ uint64_t sm_start; /* start of map */
+ uint64_t sm_size; /* size of map */
+ uint8_t sm_shift; /* unit shift */
+ uint8_t sm_pad[3]; /* unused */
+ uint8_t sm_loaded; /* map loaded? */
+ uint8_t sm_loading; /* map loading? */
+ kcondvar_t sm_load_cv; /* map load completion */
+ space_map_ops_t *sm_ops; /* space map block picker ops vector */
+ void *sm_ppd; /* picker-private data */
+ kmutex_t *sm_lock; /* pointer to lock that protects map */
+} space_map_t;
+
+typedef struct space_seg {
+ avl_node_t ss_node; /* AVL node */
+ uint64_t ss_start; /* starting offset of this segment */
+ uint64_t ss_end; /* ending offset (non-inclusive) */
+} space_seg_t;
+
+typedef struct space_map_obj {
+ uint64_t smo_object; /* on-disk space map object */
+ uint64_t smo_objsize; /* size of the object */
+ uint64_t smo_alloc; /* space allocated from the map */
+} space_map_obj_t;
+
+struct space_map_ops {
+ void (*smop_load)(space_map_t *sm);
+ void (*smop_unload)(space_map_t *sm);
+ uint64_t (*smop_alloc)(space_map_t *sm, uint64_t size);
+ void (*smop_claim)(space_map_t *sm, uint64_t start, uint64_t size);
+ void (*smop_free)(space_map_t *sm, uint64_t start, uint64_t size);
+};
+
+/*
+ * debug entry
+ *
+ * 1 3 10 50
+ * ,---+--------+------------+---------------------------------.
+ * | 1 | action | syncpass | txg (lower bits) |
+ * `---+--------+------------+---------------------------------'
+ * 63 62 60 59 50 49 0
+ *
+ *
+ *
+ * non-debug entry
+ *
+ * 1 47 1 15
+ * ,-----------------------------------------------------------.
+ * | 0 | offset (sm_shift units) | type | run |
+ * `-----------------------------------------------------------'
+ * 63 62 17 16 15 0
+ */
+
+/* All this stuff takes and returns bytes */
+#define SM_RUN_DECODE(x) (BF64_DECODE(x, 0, 15) + 1)
+#define SM_RUN_ENCODE(x) BF64_ENCODE((x) - 1, 0, 15)
+#define SM_TYPE_DECODE(x) BF64_DECODE(x, 15, 1)
+#define SM_TYPE_ENCODE(x) BF64_ENCODE(x, 15, 1)
+#define SM_OFFSET_DECODE(x) BF64_DECODE(x, 16, 47)
+#define SM_OFFSET_ENCODE(x) BF64_ENCODE(x, 16, 47)
+#define SM_DEBUG_DECODE(x) BF64_DECODE(x, 63, 1)
+#define SM_DEBUG_ENCODE(x) BF64_ENCODE(x, 63, 1)
+
+#define SM_DEBUG_ACTION_DECODE(x) BF64_DECODE(x, 60, 3)
+#define SM_DEBUG_ACTION_ENCODE(x) BF64_ENCODE(x, 60, 3)
+
+#define SM_DEBUG_SYNCPASS_DECODE(x) BF64_DECODE(x, 50, 10)
+#define SM_DEBUG_SYNCPASS_ENCODE(x) BF64_ENCODE(x, 50, 10)
+
+#define SM_DEBUG_TXG_DECODE(x) BF64_DECODE(x, 0, 50)
+#define SM_DEBUG_TXG_ENCODE(x) BF64_ENCODE(x, 0, 50)
+
+#define SM_RUN_MAX SM_RUN_DECODE(~0ULL)
+
+#define SM_ALLOC 0x0
+#define SM_FREE 0x1
+
+/*
+ * The data for a given space map can be kept on blocks of any size.
+ * Larger blocks entail fewer i/o operations, but they also cause the
+ * DMU to keep more data in-core, and also to waste more i/o bandwidth
+ * when only a few blocks have changed since the last transaction group.
+ * This could use a lot more research, but for now, set the freelist
+ * block size to 4k (2^12).
+ */
+#define SPACE_MAP_BLOCKSHIFT 12
+
+typedef void space_map_func_t(space_map_t *sm, uint64_t start, uint64_t size);
+
+extern void space_map_create(space_map_t *sm, uint64_t start, uint64_t size,
+ uint8_t shift, kmutex_t *lp);
+extern void space_map_destroy(space_map_t *sm);
+extern void space_map_add(space_map_t *sm, uint64_t start, uint64_t size);
+extern void space_map_remove(space_map_t *sm, uint64_t start, uint64_t size);
+extern int space_map_contains(space_map_t *sm, uint64_t start, uint64_t size);
+extern void space_map_vacate(space_map_t *sm,
+ space_map_func_t *func, space_map_t *mdest);
+extern void space_map_walk(space_map_t *sm,
+ space_map_func_t *func, space_map_t *mdest);
+extern void space_map_excise(space_map_t *sm, uint64_t start, uint64_t size);
+extern void space_map_union(space_map_t *smd, space_map_t *sms);
+
+extern void space_map_load_wait(space_map_t *sm);
+extern int space_map_load(space_map_t *sm, space_map_ops_t *ops,
+ uint8_t maptype, space_map_obj_t *smo, objset_t *os);
+extern void space_map_unload(space_map_t *sm);
+
+extern uint64_t space_map_alloc(space_map_t *sm, uint64_t size);
+extern void space_map_claim(space_map_t *sm, uint64_t start, uint64_t size);
+extern void space_map_free(space_map_t *sm, uint64_t start, uint64_t size);
+
+extern void space_map_sync(space_map_t *sm, uint8_t maptype,
+ space_map_obj_t *smo, objset_t *os, dmu_tx_t *tx);
+extern void space_map_truncate(space_map_obj_t *smo,
+ objset_t *os, dmu_tx_t *tx);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_SPACE_MAP_H */
diff --git a/module/zfs/include/sys/txg.h b/module/zfs/include/sys/txg.h
new file mode 100644
index 000000000..23bdff211
--- /dev/null
+++ b/module/zfs/include/sys/txg.h
@@ -0,0 +1,130 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_TXG_H
+#define _SYS_TXG_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define TXG_CONCURRENT_STATES 3 /* open, quiescing, syncing */
+#define TXG_SIZE 4 /* next power of 2 */
+#define TXG_MASK (TXG_SIZE - 1) /* mask for size */
+#define TXG_INITIAL TXG_SIZE /* initial txg */
+#define TXG_IDX (txg & TXG_MASK)
+
+#define TXG_WAIT 1ULL
+#define TXG_NOWAIT 2ULL
+
+typedef struct tx_cpu tx_cpu_t;
+
+typedef struct txg_handle {
+ tx_cpu_t *th_cpu;
+ uint64_t th_txg;
+} txg_handle_t;
+
+typedef struct txg_node {
+ struct txg_node *tn_next[TXG_SIZE];
+ uint8_t tn_member[TXG_SIZE];
+} txg_node_t;
+
+typedef struct txg_list {
+ kmutex_t tl_lock;
+ size_t tl_offset;
+ txg_node_t *tl_head[TXG_SIZE];
+} txg_list_t;
+
+struct dsl_pool;
+
+extern void txg_init(struct dsl_pool *dp, uint64_t txg);
+extern void txg_fini(struct dsl_pool *dp);
+extern void txg_sync_start(struct dsl_pool *dp);
+extern void txg_sync_stop(struct dsl_pool *dp);
+extern uint64_t txg_hold_open(struct dsl_pool *dp, txg_handle_t *txghp);
+extern void txg_rele_to_quiesce(txg_handle_t *txghp);
+extern void txg_rele_to_sync(txg_handle_t *txghp);
+extern void txg_suspend(struct dsl_pool *dp);
+extern void txg_resume(struct dsl_pool *dp);
+
+/*
+ * Delay the caller by the specified number of ticks or until
+ * the txg closes (whichever comes first). This is intended
+ * to be used to throttle writers when the system nears its
+ * capacity.
+ */
+extern void txg_delay(struct dsl_pool *dp, uint64_t txg, int ticks);
+
+/*
+ * Wait until the given transaction group has finished syncing.
+ * Try to make this happen as soon as possible (eg. kick off any
+ * necessary syncs immediately). If txg==0, wait for the currently open
+ * txg to finish syncing.
+ */
+extern void txg_wait_synced(struct dsl_pool *dp, uint64_t txg);
+
+/*
+ * Wait until the given transaction group, or one after it, is
+ * the open transaction group. Try to make this happen as soon
+ * as possible (eg. kick off any necessary syncs immediately).
+ * If txg == 0, wait for the next open txg.
+ */
+extern void txg_wait_open(struct dsl_pool *dp, uint64_t txg);
+
+/*
+ * Returns TRUE if we are "backed up" waiting for the syncing
+ * transaction to complete; otherwise returns FALSE.
+ */
+extern boolean_t txg_stalled(struct dsl_pool *dp);
+
+/* returns TRUE if someone is waiting for the next txg to sync */
+extern boolean_t txg_sync_waiting(struct dsl_pool *dp);
+
+/*
+ * Per-txg object lists.
+ */
+
+#define TXG_CLEAN(txg) ((txg) - 1)
+
+extern void txg_list_create(txg_list_t *tl, size_t offset);
+extern void txg_list_destroy(txg_list_t *tl);
+extern int txg_list_empty(txg_list_t *tl, uint64_t txg);
+extern int txg_list_add(txg_list_t *tl, void *p, uint64_t txg);
+extern void *txg_list_remove(txg_list_t *tl, uint64_t txg);
+extern void *txg_list_remove_this(txg_list_t *tl, void *p, uint64_t txg);
+extern int txg_list_member(txg_list_t *tl, void *p, uint64_t txg);
+extern void *txg_list_head(txg_list_t *tl, uint64_t txg);
+extern void *txg_list_next(txg_list_t *tl, void *p, uint64_t txg);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_TXG_H */
diff --git a/module/zfs/include/sys/txg_impl.h b/module/zfs/include/sys/txg_impl.h
new file mode 100644
index 000000000..7413c662b
--- /dev/null
+++ b/module/zfs/include/sys/txg_impl.h
@@ -0,0 +1,73 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_TXG_IMPL_H
+#define _SYS_TXG_IMPL_H
+
+#include <sys/spa.h>
+#include <sys/txg.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct tx_cpu {
+ kmutex_t tc_lock;
+ kcondvar_t tc_cv[TXG_SIZE];
+ uint64_t tc_count[TXG_SIZE];
+ char tc_pad[16];
+};
+
+typedef struct tx_state {
+ tx_cpu_t *tx_cpu; /* protects right to enter txg */
+ kmutex_t tx_sync_lock; /* protects tx_state_t */
+ krwlock_t tx_suspend;
+ uint64_t tx_open_txg; /* currently open txg id */
+ uint64_t tx_quiesced_txg; /* quiesced txg waiting for sync */
+ uint64_t tx_syncing_txg; /* currently syncing txg id */
+ uint64_t tx_synced_txg; /* last synced txg id */
+
+ uint64_t tx_sync_txg_waiting; /* txg we're waiting to sync */
+ uint64_t tx_quiesce_txg_waiting; /* txg we're waiting to open */
+
+ kcondvar_t tx_sync_more_cv;
+ kcondvar_t tx_sync_done_cv;
+ kcondvar_t tx_quiesce_more_cv;
+ kcondvar_t tx_quiesce_done_cv;
+ kcondvar_t tx_timeout_cv;
+ kcondvar_t tx_exit_cv; /* wait for all threads to exit */
+
+ uint8_t tx_threads; /* number of threads */
+ uint8_t tx_exiting; /* set when we're exiting */
+
+ kthread_t *tx_sync_thread;
+ kthread_t *tx_quiesce_thread;
+} tx_state_t;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_TXG_IMPL_H */
diff --git a/module/zfs/include/sys/uberblock.h b/module/zfs/include/sys/uberblock.h
new file mode 100644
index 000000000..93d936ae4
--- /dev/null
+++ b/module/zfs/include/sys/uberblock.h
@@ -0,0 +1,50 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License"). You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_UBERBLOCK_H
+#define _SYS_UBERBLOCK_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/vdev.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct uberblock uberblock_t;
+
+extern int uberblock_verify(uberblock_t *ub);
+extern int uberblock_update(uberblock_t *ub, vdev_t *rvd, uint64_t txg);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_UBERBLOCK_H */
diff --git a/module/zfs/include/sys/uberblock_impl.h b/module/zfs/include/sys/uberblock_impl.h
new file mode 100644
index 000000000..55a0dd5ae
--- /dev/null
+++ b/module/zfs/include/sys/uberblock_impl.h
@@ -0,0 +1,63 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_UBERBLOCK_IMPL_H
+#define _SYS_UBERBLOCK_IMPL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/uberblock.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * The uberblock version is incremented whenever an incompatible on-disk
+ * format change is made to the SPA, DMU, or ZAP.
+ *
+ * Note: the first two fields should never be moved. When a storage pool
+ * is opened, the uberblock must be read off the disk before the version
+ * can be checked. If the ub_version field is moved, we may not detect
+ * version mismatch. If the ub_magic field is moved, applications that
+ * expect the magic number in the first word won't work.
+ */
+#define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */
+#define UBERBLOCK_SHIFT 10 /* up to 1K */
+
+struct uberblock {
+ uint64_t ub_magic; /* UBERBLOCK_MAGIC */
+ uint64_t ub_version; /* SPA_VERSION */
+ uint64_t ub_txg; /* txg of last sync */
+ uint64_t ub_guid_sum; /* sum of all vdev guids */
+ uint64_t ub_timestamp; /* UTC time of last sync */
+ blkptr_t ub_rootbp; /* MOS objset_phys_t */
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_UBERBLOCK_IMPL_H */
diff --git a/module/zfs/include/sys/unique.h b/module/zfs/include/sys/unique.h
new file mode 100644
index 000000000..2ef3093ed
--- /dev/null
+++ b/module/zfs/include/sys/unique.h
@@ -0,0 +1,59 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_UNIQUE_H
+#define _SYS_UNIQUE_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* The number of significant bits in each unique value. */
+#define UNIQUE_BITS 56
+
+void unique_init(void);
+void unique_fini(void);
+
+/*
+ * Return a new unique value (which will not be uniquified against until
+ * it is unique_insert()-ed.
+ */
+uint64_t unique_create(void);
+
+/* Return a unique value, which equals the one passed in if possible. */
+uint64_t unique_insert(uint64_t value);
+
+/* Indicate that this value no longer needs to be uniquified against. */
+void unique_remove(uint64_t value);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_UNIQUE_H */
diff --git a/module/zfs/include/sys/vdev.h b/module/zfs/include/sys/vdev.h
new file mode 100644
index 000000000..c070d6f3d
--- /dev/null
+++ b/module/zfs/include/sys/vdev.h
@@ -0,0 +1,135 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_VDEV_H
+#define _SYS_VDEV_H
+
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu.h>
+#include <sys/space_map.h>
+#include <sys/fs/zfs.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern boolean_t zfs_nocacheflush;
+
+extern int vdev_open(vdev_t *);
+extern int vdev_validate(vdev_t *);
+extern void vdev_close(vdev_t *);
+extern int vdev_create(vdev_t *, uint64_t txg, boolean_t isreplace);
+extern void vdev_init(vdev_t *, uint64_t txg);
+extern void vdev_reopen(vdev_t *);
+extern int vdev_validate_aux(vdev_t *vd);
+extern zio_t *vdev_probe(vdev_t *vd, zio_t *pio);
+
+extern boolean_t vdev_is_bootable(vdev_t *vd);
+extern vdev_t *vdev_lookup_top(spa_t *spa, uint64_t vdev);
+extern vdev_t *vdev_lookup_by_guid(vdev_t *vd, uint64_t guid);
+extern void vdev_dtl_dirty(space_map_t *sm, uint64_t txg, uint64_t size);
+extern int vdev_dtl_contains(space_map_t *sm, uint64_t txg, uint64_t size);
+extern void vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
+ int scrub_done);
+extern boolean_t vdev_resilver_needed(vdev_t *vd,
+ uint64_t *minp, uint64_t *maxp);
+
+extern int vdev_metaslab_init(vdev_t *vd, uint64_t txg);
+extern void vdev_metaslab_fini(vdev_t *vd);
+
+extern void vdev_get_stats(vdev_t *vd, vdev_stat_t *vs);
+extern void vdev_clear_stats(vdev_t *vd);
+extern void vdev_stat_update(zio_t *zio, uint64_t psize);
+extern void vdev_scrub_stat_update(vdev_t *vd, pool_scrub_type_t type,
+ boolean_t complete);
+extern int vdev_getspec(spa_t *spa, uint64_t vdev, char **vdev_spec);
+extern void vdev_propagate_state(vdev_t *vd);
+extern void vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state,
+ vdev_aux_t aux);
+
+extern void vdev_space_update(vdev_t *vd, int64_t space_delta,
+ int64_t alloc_delta, boolean_t update_root);
+
+extern uint64_t vdev_psize_to_asize(vdev_t *vd, uint64_t psize);
+
+extern int vdev_fault(spa_t *spa, uint64_t guid);
+extern int vdev_degrade(spa_t *spa, uint64_t guid);
+extern int vdev_online(spa_t *spa, uint64_t guid, uint64_t flags,
+ vdev_state_t *);
+extern int vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags);
+extern void vdev_clear(spa_t *spa, vdev_t *vd);
+
+extern boolean_t vdev_is_dead(vdev_t *vd);
+extern boolean_t vdev_readable(vdev_t *vd);
+extern boolean_t vdev_writeable(vdev_t *vd);
+extern boolean_t vdev_allocatable(vdev_t *vd);
+extern boolean_t vdev_accessible(vdev_t *vd, zio_t *zio);
+
+extern void vdev_cache_init(vdev_t *vd);
+extern void vdev_cache_fini(vdev_t *vd);
+extern int vdev_cache_read(zio_t *zio);
+extern void vdev_cache_write(zio_t *zio);
+extern void vdev_cache_purge(vdev_t *vd);
+
+extern void vdev_queue_init(vdev_t *vd);
+extern void vdev_queue_fini(vdev_t *vd);
+extern zio_t *vdev_queue_io(zio_t *zio);
+extern void vdev_queue_io_done(zio_t *zio);
+
+extern void vdev_config_dirty(vdev_t *vd);
+extern void vdev_config_clean(vdev_t *vd);
+extern int vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg);
+
+extern void vdev_state_dirty(vdev_t *vd);
+extern void vdev_state_clean(vdev_t *vd);
+
+extern nvlist_t *vdev_config_generate(spa_t *spa, vdev_t *vd,
+ boolean_t getstats, boolean_t isspare, boolean_t isl2cache);
+
+/*
+ * Label routines
+ */
+struct uberblock;
+extern uint64_t vdev_label_offset(uint64_t psize, int l, uint64_t offset);
+extern int vdev_label_number(uint64_t psise, uint64_t offset);
+extern nvlist_t *vdev_label_read_config(vdev_t *vd);
+extern void vdev_uberblock_load(zio_t *zio, vdev_t *vd, struct uberblock *ub);
+
+typedef enum {
+ VDEV_LABEL_CREATE, /* create/add a new device */
+ VDEV_LABEL_REPLACE, /* replace an existing device */
+ VDEV_LABEL_SPARE, /* add a new hot spare */
+ VDEV_LABEL_REMOVE, /* remove an existing device */
+ VDEV_LABEL_L2CACHE /* add an L2ARC cache device */
+} vdev_labeltype_t;
+
+extern int vdev_label_init(vdev_t *vd, uint64_t txg, vdev_labeltype_t reason);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_VDEV_H */
diff --git a/module/zfs/include/sys/vdev_file.h b/module/zfs/include/sys/vdev_file.h
new file mode 100644
index 000000000..cd4967357
--- /dev/null
+++ b/module/zfs/include/sys/vdev_file.h
@@ -0,0 +1,46 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License"). You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_VDEV_FILE_H
+#define _SYS_VDEV_FILE_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/vdev.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct vdev_file {
+ vnode_t *vf_vnode;
+} vdev_file_t;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_VDEV_FILE_H */
diff --git a/module/zfs/include/sys/vdev_impl.h b/module/zfs/include/sys/vdev_impl.h
new file mode 100644
index 000000000..26904d089
--- /dev/null
+++ b/module/zfs/include/sys/vdev_impl.h
@@ -0,0 +1,305 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_VDEV_IMPL_H
+#define _SYS_VDEV_IMPL_H
+
+#include <sys/avl.h>
+#include <sys/dmu.h>
+#include <sys/metaslab.h>
+#include <sys/nvpair.h>
+#include <sys/space_map.h>
+#include <sys/vdev.h>
+#include <sys/dkio.h>
+#include <sys/uberblock_impl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Virtual device descriptors.
+ *
+ * All storage pool operations go through the virtual device framework,
+ * which provides data replication and I/O scheduling.
+ */
+
+/*
+ * Forward declarations that lots of things need.
+ */
+typedef struct vdev_queue vdev_queue_t;
+typedef struct vdev_cache vdev_cache_t;
+typedef struct vdev_cache_entry vdev_cache_entry_t;
+
+/*
+ * Virtual device operations
+ */
+typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *ashift);
+typedef void vdev_close_func_t(vdev_t *vd);
+typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
+typedef int vdev_io_start_func_t(zio_t *zio);
+typedef void vdev_io_done_func_t(zio_t *zio);
+typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
+
+typedef struct vdev_ops {
+ vdev_open_func_t *vdev_op_open;
+ vdev_close_func_t *vdev_op_close;
+ vdev_asize_func_t *vdev_op_asize;
+ vdev_io_start_func_t *vdev_op_io_start;
+ vdev_io_done_func_t *vdev_op_io_done;
+ vdev_state_change_func_t *vdev_op_state_change;
+ char vdev_op_type[16];
+ boolean_t vdev_op_leaf;
+} vdev_ops_t;
+
+/*
+ * Virtual device properties
+ */
+struct vdev_cache_entry {
+ char *ve_data;
+ uint64_t ve_offset;
+ uint64_t ve_lastused;
+ avl_node_t ve_offset_node;
+ avl_node_t ve_lastused_node;
+ uint32_t ve_hits;
+ uint16_t ve_missed_update;
+ zio_t *ve_fill_io;
+};
+
+struct vdev_cache {
+ avl_tree_t vc_offset_tree;
+ avl_tree_t vc_lastused_tree;
+ kmutex_t vc_lock;
+};
+
+struct vdev_queue {
+ avl_tree_t vq_deadline_tree;
+ avl_tree_t vq_read_tree;
+ avl_tree_t vq_write_tree;
+ avl_tree_t vq_pending_tree;
+ kmutex_t vq_lock;
+};
+
+/*
+ * Virtual device descriptor
+ */
+struct vdev {
+ /*
+ * Common to all vdev types.
+ */
+ uint64_t vdev_id; /* child number in vdev parent */
+ uint64_t vdev_guid; /* unique ID for this vdev */
+ uint64_t vdev_guid_sum; /* self guid + all child guids */
+ uint64_t vdev_asize; /* allocatable device capacity */
+ uint64_t vdev_ashift; /* block alignment shift */
+ uint64_t vdev_state; /* see VDEV_STATE_* #defines */
+ uint64_t vdev_prevstate; /* used when reopening a vdev */
+ vdev_ops_t *vdev_ops; /* vdev operations */
+ spa_t *vdev_spa; /* spa for this vdev */
+ void *vdev_tsd; /* type-specific data */
+ vdev_t *vdev_top; /* top-level vdev */
+ vdev_t *vdev_parent; /* parent vdev */
+ vdev_t **vdev_child; /* array of children */
+ uint64_t vdev_children; /* number of children */
+ space_map_t vdev_dtl_map; /* dirty time log in-core state */
+ space_map_t vdev_dtl_scrub; /* DTL for scrub repair writes */
+ vdev_stat_t vdev_stat; /* virtual device statistics */
+
+ /*
+ * Top-level vdev state.
+ */
+ uint64_t vdev_ms_array; /* metaslab array object */
+ uint64_t vdev_ms_shift; /* metaslab size shift */
+ uint64_t vdev_ms_count; /* number of metaslabs */
+ metaslab_group_t *vdev_mg; /* metaslab group */
+ metaslab_t **vdev_ms; /* metaslab array */
+ txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
+ txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
+ txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
+ boolean_t vdev_remove_wanted; /* async remove wanted? */
+ boolean_t vdev_probe_wanted; /* async probe wanted? */
+ list_node_t vdev_config_dirty_node; /* config dirty list */
+ list_node_t vdev_state_dirty_node; /* state dirty list */
+ uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
+ uint64_t vdev_islog; /* is an intent log device */
+
+ /*
+ * Leaf vdev state.
+ */
+ uint64_t vdev_psize; /* physical device capacity */
+ space_map_obj_t vdev_dtl; /* dirty time log on-disk state */
+ txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
+ uint64_t vdev_wholedisk; /* true if this is a whole disk */
+ uint64_t vdev_offline; /* persistent offline state */
+ uint64_t vdev_faulted; /* persistent faulted state */
+ uint64_t vdev_degraded; /* persistent degraded state */
+ uint64_t vdev_removed; /* persistent removed state */
+ uint64_t vdev_nparity; /* number of parity devices for raidz */
+ char *vdev_path; /* vdev path (if any) */
+ char *vdev_devid; /* vdev devid (if any) */
+ char *vdev_physpath; /* vdev device path (if any) */
+ uint64_t vdev_not_present; /* not present during import */
+ uint64_t vdev_unspare; /* unspare when resilvering done */
+ hrtime_t vdev_last_try; /* last reopen time */
+ boolean_t vdev_nowritecache; /* true if flushwritecache failed */
+ boolean_t vdev_checkremove; /* temporary online test */
+ boolean_t vdev_forcefault; /* force online fault */
+ uint8_t vdev_tmpoffline; /* device taken offline temporarily? */
+ uint8_t vdev_detached; /* device detached? */
+ uint8_t vdev_cant_read; /* vdev is failing all reads */
+ uint8_t vdev_cant_write; /* vdev is failing all writes */
+ uint64_t vdev_isspare; /* was a hot spare */
+ uint64_t vdev_isl2cache; /* was a l2cache device */
+ vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
+ vdev_cache_t vdev_cache; /* physical block cache */
+ spa_aux_vdev_t *vdev_aux; /* for l2cache vdevs */
+ zio_t *vdev_probe_zio; /* root of current probe */
+
+ /*
+ * For DTrace to work in userland (libzpool) context, these fields must
+ * remain at the end of the structure. DTrace will use the kernel's
+ * CTF definition for 'struct vdev', and since the size of a kmutex_t is
+ * larger in userland, the offsets for the rest fields would be
+ * incorrect.
+ */
+ kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
+ kmutex_t vdev_stat_lock; /* vdev_stat */
+ kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */
+};
+
+#define VDEV_SKIP_SIZE (8 << 10)
+#define VDEV_BOOT_HEADER_SIZE (8 << 10)
+#define VDEV_PHYS_SIZE (112 << 10)
+#define VDEV_UBERBLOCK_RING (128 << 10)
+
+#define VDEV_UBERBLOCK_SHIFT(vd) \
+ MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT)
+#define VDEV_UBERBLOCK_COUNT(vd) \
+ (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
+#define VDEV_UBERBLOCK_OFFSET(vd, n) \
+ offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
+#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
+
+/* ZFS boot block */
+#define VDEV_BOOT_MAGIC 0x2f5b007b10cULL
+#define VDEV_BOOT_VERSION 1 /* version number */
+
+typedef struct vdev_boot_header {
+ uint64_t vb_magic; /* VDEV_BOOT_MAGIC */
+ uint64_t vb_version; /* VDEV_BOOT_VERSION */
+ uint64_t vb_offset; /* start offset (bytes) */
+ uint64_t vb_size; /* size (bytes) */
+ char vb_pad[VDEV_BOOT_HEADER_SIZE - 4 * sizeof (uint64_t)];
+} vdev_boot_header_t;
+
+typedef struct vdev_phys {
+ char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_block_tail_t)];
+ zio_block_tail_t vp_zbt;
+} vdev_phys_t;
+
+typedef struct vdev_label {
+ char vl_pad[VDEV_SKIP_SIZE]; /* 8K */
+ vdev_boot_header_t vl_boot_header; /* 8K */
+ vdev_phys_t vl_vdev_phys; /* 112K */
+ char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
+} vdev_label_t; /* 256K total */
+
+/*
+ * vdev_dirty() flags
+ */
+#define VDD_METASLAB 0x01
+#define VDD_DTL 0x02
+
+/*
+ * Size and offset of embedded boot loader region on each label.
+ * The total size of the first two labels plus the boot area is 4MB.
+ */
+#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
+#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
+
+/*
+ * Size of label regions at the start and end of each leaf device.
+ */
+#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
+#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
+#define VDEV_LABELS 4
+
+#define VDEV_ALLOC_LOAD 0
+#define VDEV_ALLOC_ADD 1
+#define VDEV_ALLOC_SPARE 2
+#define VDEV_ALLOC_L2CACHE 3
+
+/*
+ * Allocate or free a vdev
+ */
+extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
+ vdev_t *parent, uint_t id, int alloctype);
+extern void vdev_free(vdev_t *vd);
+
+/*
+ * Add or remove children and parents
+ */
+extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
+extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
+extern void vdev_compact_children(vdev_t *pvd);
+extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
+extern void vdev_remove_parent(vdev_t *cvd);
+
+/*
+ * vdev sync load and sync
+ */
+extern void vdev_load(vdev_t *vd);
+extern void vdev_sync(vdev_t *vd, uint64_t txg);
+extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
+extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
+
+/*
+ * Available vdev types.
+ */
+extern vdev_ops_t vdev_root_ops;
+extern vdev_ops_t vdev_mirror_ops;
+extern vdev_ops_t vdev_replacing_ops;
+extern vdev_ops_t vdev_raidz_ops;
+extern vdev_ops_t vdev_disk_ops;
+extern vdev_ops_t vdev_file_ops;
+extern vdev_ops_t vdev_missing_ops;
+extern vdev_ops_t vdev_spare_ops;
+
+/*
+ * Common size functions
+ */
+extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
+extern uint64_t vdev_get_rsize(vdev_t *vd);
+
+/*
+ * zdb uses this tunable, so it must be declared here to make lint happy.
+ */
+extern int zfs_vdev_cache_size;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_VDEV_IMPL_H */
diff --git a/module/zfs/include/sys/zap.h b/module/zfs/include/sys/zap.h
new file mode 100644
index 000000000..f88cc068b
--- /dev/null
+++ b/module/zfs/include/sys/zap.h
@@ -0,0 +1,425 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZAP_H
+#define _SYS_ZAP_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * ZAP - ZFS Attribute Processor
+ *
+ * The ZAP is a module which sits on top of the DMU (Data Management
+ * Unit) and implements a higher-level storage primitive using DMU
+ * objects. Its primary consumer is the ZPL (ZFS Posix Layer).
+ *
+ * A "zapobj" is a DMU object which the ZAP uses to stores attributes.
+ * Users should use only zap routines to access a zapobj - they should
+ * not access the DMU object directly using DMU routines.
+ *
+ * The attributes stored in a zapobj are name-value pairs. The name is
+ * a zero-terminated string of up to ZAP_MAXNAMELEN bytes (including
+ * terminating NULL). The value is an array of integers, which may be
+ * 1, 2, 4, or 8 bytes long. The total space used by the array (number
+ * of integers * integer length) can be up to ZAP_MAXVALUELEN bytes.
+ * Note that an 8-byte integer value can be used to store the location
+ * (object number) of another dmu object (which may be itself a zapobj).
+ * Note that you can use a zero-length attribute to store a single bit
+ * of information - the attribute is present or not.
+ *
+ * The ZAP routines are thread-safe. However, you must observe the
+ * DMU's restriction that a transaction may not be operated on
+ * concurrently.
+ *
+ * Any of the routines that return an int may return an I/O error (EIO
+ * or ECHECKSUM).
+ *
+ *
+ * Implementation / Performance Notes:
+ *
+ * The ZAP is intended to operate most efficiently on attributes with
+ * short (49 bytes or less) names and single 8-byte values, for which
+ * the microzap will be used. The ZAP should be efficient enough so
+ * that the user does not need to cache these attributes.
+ *
+ * The ZAP's locking scheme makes its routines thread-safe. Operations
+ * on different zapobjs will be processed concurrently. Operations on
+ * the same zapobj which only read data will be processed concurrently.
+ * Operations on the same zapobj which modify data will be processed
+ * concurrently when there are many attributes in the zapobj (because
+ * the ZAP uses per-block locking - more than 128 * (number of cpus)
+ * small attributes will suffice).
+ */
+
+/*
+ * We're using zero-terminated byte strings (ie. ASCII or UTF-8 C
+ * strings) for the names of attributes, rather than a byte string
+ * bounded by an explicit length. If some day we want to support names
+ * in character sets which have embedded zeros (eg. UTF-16, UTF-32),
+ * we'll have to add routines for using length-bounded strings.
+ */
+
+#include <sys/dmu.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZAP_MAXNAMELEN 256
+#define ZAP_MAXVALUELEN 1024
+
+/*
+ * The matchtype specifies which entry will be accessed.
+ * MT_EXACT: only find an exact match (non-normalized)
+ * MT_FIRST: find the "first" normalized (case and Unicode
+ * form) match; the designated "first" match will not change as long
+ * as the set of entries with this normalization doesn't change
+ * MT_BEST: if there is an exact match, find that, otherwise find the
+ * first normalized match
+ */
+typedef enum matchtype
+{
+ MT_EXACT,
+ MT_BEST,
+ MT_FIRST
+} matchtype_t;
+
+/*
+ * Create a new zapobj with no attributes and return its object number.
+ * MT_EXACT will cause the zap object to only support MT_EXACT lookups,
+ * otherwise any matchtype can be used for lookups.
+ *
+ * normflags specifies what normalization will be done. values are:
+ * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
+ * only)
+ * U8_TEXTPREP_TOLOWER: case normalization will be performed.
+ * MT_FIRST/MT_BEST matching will find entries that match without
+ * regard to case (eg. looking for "foo" can find an entry "Foo").
+ * Eventually, other flags will permit unicode normalization as well.
+ */
+uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+
+/*
+ * Create a new zapobj with no attributes from the given (unallocated)
+ * object number.
+ */
+int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+int zap_create_claim_norm(objset_t *ds, uint64_t obj,
+ int normflags, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+
+/*
+ * The zapobj passed in must be a valid ZAP object for all of the
+ * following routines.
+ */
+
+/*
+ * Destroy this zapobj and all its attributes.
+ *
+ * Frees the object number using dmu_object_free.
+ */
+int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
+
+/*
+ * Manipulate attributes.
+ *
+ * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
+ */
+
+/*
+ * Retrieve the contents of the attribute with the given name.
+ *
+ * If the requested attribute does not exist, the call will fail and
+ * return ENOENT.
+ *
+ * If 'integer_size' is smaller than the attribute's integer size, the
+ * call will fail and return EINVAL.
+ *
+ * If 'integer_size' is equal to or larger than the attribute's integer
+ * size, the call will succeed and return 0. * When converting to a
+ * larger integer size, the integers will be treated as unsigned (ie. no
+ * sign-extension will be performed).
+ *
+ * 'num_integers' is the length (in integers) of 'buf'.
+ *
+ * If the attribute is longer than the buffer, as many integers as will
+ * fit will be transferred to 'buf'. If the entire attribute was not
+ * transferred, the call will return EOVERFLOW.
+ *
+ * If rn_len is nonzero, realname will be set to the name of the found
+ * entry (which may be different from the requested name if matchtype is
+ * not MT_EXACT).
+ *
+ * If normalization_conflictp is not NULL, it will be set if there is
+ * another name with the same case/unicode normalized form.
+ */
+int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
+ uint64_t integer_size, uint64_t num_integers, void *buf);
+int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
+ uint64_t integer_size, uint64_t num_integers, void *buf,
+ matchtype_t mt, char *realname, int rn_len,
+ boolean_t *normalization_conflictp);
+
+/*
+ * Create an attribute with the given name and value.
+ *
+ * If an attribute with the given name already exists, the call will
+ * fail and return EEXIST.
+ */
+int zap_add(objset_t *ds, uint64_t zapobj, const char *name,
+ int integer_size, uint64_t num_integers,
+ const void *val, dmu_tx_t *tx);
+
+/*
+ * Set the attribute with the given name to the given value. If an
+ * attribute with the given name does not exist, it will be created. If
+ * an attribute with the given name already exists, the previous value
+ * will be overwritten. The integer_size may be different from the
+ * existing attribute's integer size, in which case the attribute's
+ * integer size will be updated to the new value.
+ */
+int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
+ int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
+
+/*
+ * Get the length (in integers) and the integer size of the specified
+ * attribute.
+ *
+ * If the requested attribute does not exist, the call will fail and
+ * return ENOENT.
+ */
+int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
+ uint64_t *integer_size, uint64_t *num_integers);
+
+/*
+ * Remove the specified attribute.
+ *
+ * If the specified attribute does not exist, the call will fail and
+ * return ENOENT.
+ */
+int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
+int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
+ matchtype_t mt, dmu_tx_t *tx);
+
+/*
+ * Returns (in *count) the number of attributes in the specified zap
+ * object.
+ */
+int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
+
+
+/*
+ * Returns (in name) the name of the entry whose (value & mask)
+ * (za_first_integer) is value, or ENOENT if not found. The string
+ * pointed to by name must be at least 256 bytes long. If mask==0, the
+ * match must be exact (ie, same as mask=-1ULL).
+ */
+int zap_value_search(objset_t *os, uint64_t zapobj,
+ uint64_t value, uint64_t mask, char *name);
+
+/*
+ * Transfer all the entries from fromobj into intoobj. Only works on
+ * int_size=8 num_integers=1 values. Fails if there are any duplicated
+ * entries.
+ */
+int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
+
+/*
+ * Manipulate entries where the name + value are the "same" (the name is
+ * a stringified version of the value).
+ */
+int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
+int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
+int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
+
+struct zap;
+struct zap_leaf;
+typedef struct zap_cursor {
+ /* This structure is opaque! */
+ objset_t *zc_objset;
+ struct zap *zc_zap;
+ struct zap_leaf *zc_leaf;
+ uint64_t zc_zapobj;
+ uint64_t zc_hash;
+ uint32_t zc_cd;
+} zap_cursor_t;
+
+typedef struct {
+ int za_integer_length;
+ /*
+ * za_normalization_conflict will be set if there are additional
+ * entries with this normalized form (eg, "foo" and "Foo").
+ */
+ boolean_t za_normalization_conflict;
+ uint64_t za_num_integers;
+ uint64_t za_first_integer; /* no sign extension for <8byte ints */
+ char za_name[MAXNAMELEN];
+} zap_attribute_t;
+
+/*
+ * The interface for listing all the attributes of a zapobj can be
+ * thought of as cursor moving down a list of the attributes one by
+ * one. The cookie returned by the zap_cursor_serialize routine is
+ * persistent across system calls (and across reboot, even).
+ */
+
+/*
+ * Initialize a zap cursor, pointing to the "first" attribute of the
+ * zapobj. You must _fini the cursor when you are done with it.
+ */
+void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
+void zap_cursor_fini(zap_cursor_t *zc);
+
+/*
+ * Get the attribute currently pointed to by the cursor. Returns
+ * ENOENT if at the end of the attributes.
+ */
+int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
+
+/*
+ * Advance the cursor to the next attribute.
+ */
+void zap_cursor_advance(zap_cursor_t *zc);
+
+/*
+ * Get a persistent cookie pointing to the current position of the zap
+ * cursor. The low 4 bits in the cookie are always zero, and thus can
+ * be used as to differentiate a serialized cookie from a different type
+ * of value. The cookie will be less than 2^32 as long as there are
+ * fewer than 2^22 (4.2 million) entries in the zap object.
+ */
+uint64_t zap_cursor_serialize(zap_cursor_t *zc);
+
+/*
+ * Initialize a zap cursor pointing to the position recorded by
+ * zap_cursor_serialize (in the "serialized" argument). You can also
+ * use a "serialized" argument of 0 to start at the beginning of the
+ * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to
+ * zap_cursor_init(...).)
+ */
+void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
+ uint64_t zapobj, uint64_t serialized);
+
+
+#define ZAP_HISTOGRAM_SIZE 10
+
+typedef struct zap_stats {
+ /*
+ * Size of the pointer table (in number of entries).
+ * This is always a power of 2, or zero if it's a microzap.
+ * In general, it should be considerably greater than zs_num_leafs.
+ */
+ uint64_t zs_ptrtbl_len;
+
+ uint64_t zs_blocksize; /* size of zap blocks */
+
+ /*
+ * The number of blocks used. Note that some blocks may be
+ * wasted because old ptrtbl's and large name/value blocks are
+ * not reused. (Although their space is reclaimed, we don't
+ * reuse those offsets in the object.)
+ */
+ uint64_t zs_num_blocks;
+
+ /*
+ * Pointer table values from zap_ptrtbl in the zap_phys_t
+ */
+ uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */
+ uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */
+ uint64_t zs_ptrtbl_zt_blk; /* starting block number */
+ uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */
+ uint64_t zs_ptrtbl_zt_shift; /* bits to index it */
+
+ /*
+ * Values of the other members of the zap_phys_t
+ */
+ uint64_t zs_block_type; /* ZBT_HEADER */
+ uint64_t zs_magic; /* ZAP_MAGIC */
+ uint64_t zs_num_leafs; /* The number of leaf blocks */
+ uint64_t zs_num_entries; /* The number of zap entries */
+ uint64_t zs_salt; /* salt to stir into hash function */
+
+ /*
+ * Histograms. For all histograms, the last index
+ * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
+ * than what can be represented. For example
+ * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
+ * of leafs with more than 45 entries.
+ */
+
+ /*
+ * zs_leafs_with_n_pointers[n] is the number of leafs with
+ * 2^n pointers to it.
+ */
+ uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
+
+ /*
+ * zs_leafs_with_n_entries[n] is the number of leafs with
+ * [n*5, (n+1)*5) entries. In the current implementation, there
+ * can be at most 55 entries in any block, but there may be
+ * fewer if the name or value is large, or the block is not
+ * completely full.
+ */
+ uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
+
+ /*
+ * zs_leafs_n_tenths_full[n] is the number of leafs whose
+ * fullness is in the range [n/10, (n+1)/10).
+ */
+ uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
+
+ /*
+ * zs_entries_using_n_chunks[n] is the number of entries which
+ * consume n 24-byte chunks. (Note, large names/values only use
+ * one chunk, but contribute to zs_num_blocks_large.)
+ */
+ uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
+
+ /*
+ * zs_buckets_with_n_entries[n] is the number of buckets (each
+ * leaf has 64 buckets) with n entries.
+ * zs_buckets_with_n_entries[1] should be very close to
+ * zs_num_entries.
+ */
+ uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
+} zap_stats_t;
+
+/*
+ * Get statistics about a ZAP object. Note: you need to be aware of the
+ * internal implementation of the ZAP to correctly interpret some of the
+ * statistics. This interface shouldn't be relied on unless you really
+ * know what you're doing.
+ */
+int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZAP_H */
diff --git a/module/zfs/include/sys/zap_impl.h b/module/zfs/include/sys/zap_impl.h
new file mode 100644
index 000000000..0dc02ab6b
--- /dev/null
+++ b/module/zfs/include/sys/zap_impl.h
@@ -0,0 +1,218 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZAP_IMPL_H
+#define _SYS_ZAP_IMPL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zap.h>
+#include <sys/zfs_context.h>
+#include <sys/avl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern int fzap_default_block_shift;
+
+#define ZAP_MAGIC 0x2F52AB2ABULL
+
+#define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_f.zap_block_shift)
+
+#define ZAP_MAXCD (uint32_t)(-1)
+#define ZAP_HASHBITS 28
+#define MZAP_ENT_LEN 64
+#define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
+#define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT
+#define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT)
+
+typedef struct mzap_ent_phys {
+ uint64_t mze_value;
+ uint32_t mze_cd;
+ uint16_t mze_pad; /* in case we want to chain them someday */
+ char mze_name[MZAP_NAME_LEN];
+} mzap_ent_phys_t;
+
+typedef struct mzap_phys {
+ uint64_t mz_block_type; /* ZBT_MICRO */
+ uint64_t mz_salt;
+ uint64_t mz_normflags;
+ uint64_t mz_pad[5];
+ mzap_ent_phys_t mz_chunk[1];
+ /* actually variable size depending on block size */
+} mzap_phys_t;
+
+typedef struct mzap_ent {
+ avl_node_t mze_node;
+ int mze_chunkid;
+ uint64_t mze_hash;
+ mzap_ent_phys_t mze_phys;
+} mzap_ent_t;
+
+
+/*
+ * The (fat) zap is stored in one object. It is an array of
+ * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
+ *
+ * ptrtbl fits in first block:
+ * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
+ *
+ * ptrtbl too big for first block:
+ * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
+ *
+ */
+
+struct dmu_buf;
+struct zap_leaf;
+
+#define ZBT_LEAF ((1ULL << 63) + 0)
+#define ZBT_HEADER ((1ULL << 63) + 1)
+#define ZBT_MICRO ((1ULL << 63) + 3)
+/* any other values are ptrtbl blocks */
+
+/*
+ * the embedded pointer table takes up half a block:
+ * block size / entry size (2^3) / 2
+ */
+#define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
+
+/*
+ * The embedded pointer table starts half-way through the block. Since
+ * the pointer table itself is half the block, it starts at (64-bit)
+ * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
+ */
+#define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
+ ((uint64_t *)(zap)->zap_f.zap_phys) \
+ [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
+
+/*
+ * TAKE NOTE:
+ * If zap_phys_t is modified, zap_byteswap() must be modified.
+ */
+typedef struct zap_phys {
+ uint64_t zap_block_type; /* ZBT_HEADER */
+ uint64_t zap_magic; /* ZAP_MAGIC */
+
+ struct zap_table_phys {
+ uint64_t zt_blk; /* starting block number */
+ uint64_t zt_numblks; /* number of blocks */
+ uint64_t zt_shift; /* bits to index it */
+ uint64_t zt_nextblk; /* next (larger) copy start block */
+ uint64_t zt_blks_copied; /* number source blocks copied */
+ } zap_ptrtbl;
+
+ uint64_t zap_freeblk; /* the next free block */
+ uint64_t zap_num_leafs; /* number of leafs */
+ uint64_t zap_num_entries; /* number of entries */
+ uint64_t zap_salt; /* salt to stir into hash function */
+ uint64_t zap_normflags; /* flags for u8_textprep_str() */
+ /*
+ * This structure is followed by padding, and then the embedded
+ * pointer table. The embedded pointer table takes up second
+ * half of the block. It is accessed using the
+ * ZAP_EMBEDDED_PTRTBL_ENT() macro.
+ */
+} zap_phys_t;
+
+typedef struct zap_table_phys zap_table_phys_t;
+
+typedef struct zap {
+ objset_t *zap_objset;
+ uint64_t zap_object;
+ struct dmu_buf *zap_dbuf;
+ krwlock_t zap_rwlock;
+ boolean_t zap_ismicro;
+ int zap_normflags;
+ uint64_t zap_salt;
+ union {
+ struct {
+ zap_phys_t *zap_phys;
+
+ /*
+ * zap_num_entries_mtx protects
+ * zap_num_entries
+ */
+ kmutex_t zap_num_entries_mtx;
+ int zap_block_shift;
+ } zap_fat;
+ struct {
+ mzap_phys_t *zap_phys;
+ int16_t zap_num_entries;
+ int16_t zap_num_chunks;
+ int16_t zap_alloc_next;
+ avl_tree_t zap_avl;
+ } zap_micro;
+ } zap_u;
+} zap_t;
+
+typedef struct zap_name {
+ zap_t *zn_zap;
+ const char *zn_name_orij;
+ uint64_t zn_hash;
+ matchtype_t zn_matchtype;
+ const char *zn_name_norm;
+ char zn_normbuf[ZAP_MAXNAMELEN];
+} zap_name_t;
+
+#define zap_f zap_u.zap_fat
+#define zap_m zap_u.zap_micro
+
+boolean_t zap_match(zap_name_t *zn, const char *matchname);
+int zap_lockdir(objset_t *os, uint64_t obj, dmu_tx_t *tx,
+ krw_t lti, boolean_t fatreader, boolean_t adding, zap_t **zapp);
+void zap_unlockdir(zap_t *zap);
+void zap_evict(dmu_buf_t *db, void *vmzap);
+zap_name_t *zap_name_alloc(zap_t *zap, const char *name, matchtype_t mt);
+void zap_name_free(zap_name_t *zn);
+
+#define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n))))
+
+void fzap_byteswap(void *buf, size_t size);
+int fzap_count(zap_t *zap, uint64_t *count);
+int fzap_lookup(zap_name_t *zn,
+ uint64_t integer_size, uint64_t num_integers, void *buf,
+ char *realname, int rn_len, boolean_t *normalization_conflictp);
+int fzap_add(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers,
+ const void *val, dmu_tx_t *tx);
+int fzap_update(zap_name_t *zn,
+ int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
+int fzap_length(zap_name_t *zn,
+ uint64_t *integer_size, uint64_t *num_integers);
+int fzap_remove(zap_name_t *zn, dmu_tx_t *tx);
+int fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za);
+void fzap_get_stats(zap_t *zap, zap_stats_t *zs);
+void zap_put_leaf(struct zap_leaf *l);
+
+int fzap_add_cd(zap_name_t *zn,
+ uint64_t integer_size, uint64_t num_integers,
+ const void *val, uint32_t cd, dmu_tx_t *tx);
+void fzap_upgrade(zap_t *zap, dmu_tx_t *tx);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZAP_IMPL_H */
diff --git a/module/zfs/include/sys/zap_leaf.h b/module/zfs/include/sys/zap_leaf.h
new file mode 100644
index 000000000..14144e059
--- /dev/null
+++ b/module/zfs/include/sys/zap_leaf.h
@@ -0,0 +1,244 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZAP_LEAF_H
+#define _SYS_ZAP_LEAF_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct zap;
+
+#define ZAP_LEAF_MAGIC 0x2AB1EAF
+
+/* chunk size = 24 bytes */
+#define ZAP_LEAF_CHUNKSIZE 24
+
+/*
+ * The amount of space available for chunks is:
+ * block size (1<<l->l_bs) - hash entry size (2) * number of hash
+ * entries - header space (2*chunksize)
+ */
+#define ZAP_LEAF_NUMCHUNKS(l) \
+ (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
+ ZAP_LEAF_CHUNKSIZE - 2)
+
+/*
+ * The amount of space within the chunk available for the array is:
+ * chunk size - space for type (1) - space for next pointer (2)
+ */
+#define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
+
+#define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
+ (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
+
+/*
+ * Low water mark: when there are only this many chunks free, start
+ * growing the ptrtbl. Ideally, this should be larger than a
+ * "reasonably-sized" entry. 20 chunks is more than enough for the
+ * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
+ * while still being only around 3% for 16k blocks.
+ */
+#define ZAP_LEAF_LOW_WATER (20)
+
+/*
+ * The leaf hash table has block size / 2^5 (32) number of entries,
+ * which should be more than enough for the maximum number of entries,
+ * which is less than block size / CHUNKSIZE (24) / minimum number of
+ * chunks per entry (3).
+ */
+#define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
+#define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
+
+/*
+ * The chunks start immediately after the hash table. The end of the
+ * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
+ * chunk_t.
+ */
+#define ZAP_LEAF_CHUNK(l, idx) \
+ ((zap_leaf_chunk_t *) \
+ ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
+#define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
+
+typedef enum zap_chunk_type {
+ ZAP_CHUNK_FREE = 253,
+ ZAP_CHUNK_ENTRY = 252,
+ ZAP_CHUNK_ARRAY = 251,
+ ZAP_CHUNK_TYPE_MAX = 250
+} zap_chunk_type_t;
+
+#define ZLF_ENTRIES_CDSORTED (1<<0)
+
+/*
+ * TAKE NOTE:
+ * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
+ */
+typedef struct zap_leaf_phys {
+ struct zap_leaf_header {
+ uint64_t lh_block_type; /* ZBT_LEAF */
+ uint64_t lh_pad1;
+ uint64_t lh_prefix; /* hash prefix of this leaf */
+ uint32_t lh_magic; /* ZAP_LEAF_MAGIC */
+ uint16_t lh_nfree; /* number free chunks */
+ uint16_t lh_nentries; /* number of entries */
+ uint16_t lh_prefix_len; /* num bits used to id this */
+
+/* above is accessable to zap, below is zap_leaf private */
+
+ uint16_t lh_freelist; /* chunk head of free list */
+ uint8_t lh_flags; /* ZLF_* flags */
+ uint8_t lh_pad2[11];
+ } l_hdr; /* 2 24-byte chunks */
+
+ /*
+ * The header is followed by a hash table with
+ * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is
+ * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
+ * zap_leaf_chunk structures. These structures are accessed
+ * with the ZAP_LEAF_CHUNK() macro.
+ */
+
+ uint16_t l_hash[1];
+} zap_leaf_phys_t;
+
+typedef union zap_leaf_chunk {
+ struct zap_leaf_entry {
+ uint8_t le_type; /* always ZAP_CHUNK_ENTRY */
+ uint8_t le_int_size; /* size of ints */
+ uint16_t le_next; /* next entry in hash chain */
+ uint16_t le_name_chunk; /* first chunk of the name */
+ uint16_t le_name_length; /* bytes in name, incl null */
+ uint16_t le_value_chunk; /* first chunk of the value */
+ uint16_t le_value_length; /* value length in ints */
+ uint32_t le_cd; /* collision differentiator */
+ uint64_t le_hash; /* hash value of the name */
+ } l_entry;
+ struct zap_leaf_array {
+ uint8_t la_type; /* always ZAP_CHUNK_ARRAY */
+ uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
+ uint16_t la_next; /* next blk or CHAIN_END */
+ } l_array;
+ struct zap_leaf_free {
+ uint8_t lf_type; /* always ZAP_CHUNK_FREE */
+ uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
+ uint16_t lf_next; /* next in free list, or CHAIN_END */
+ } l_free;
+} zap_leaf_chunk_t;
+
+typedef struct zap_leaf {
+ krwlock_t l_rwlock;
+ uint64_t l_blkid; /* 1<<ZAP_BLOCK_SHIFT byte block off */
+ int l_bs; /* block size shift */
+ dmu_buf_t *l_dbuf;
+ zap_leaf_phys_t *l_phys;
+} zap_leaf_t;
+
+
+typedef struct zap_entry_handle {
+ /* below is set by zap_leaf.c and is public to zap.c */
+ uint64_t zeh_num_integers;
+ uint64_t zeh_hash;
+ uint32_t zeh_cd;
+ uint8_t zeh_integer_size;
+
+ /* below is private to zap_leaf.c */
+ uint16_t zeh_fakechunk;
+ uint16_t *zeh_chunkp;
+ zap_leaf_t *zeh_leaf;
+} zap_entry_handle_t;
+
+/*
+ * Return a handle to the named entry, or ENOENT if not found. The hash
+ * value must equal zap_hash(name).
+ */
+extern int zap_leaf_lookup(zap_leaf_t *l,
+ zap_name_t *zn, zap_entry_handle_t *zeh);
+
+/*
+ * Return a handle to the entry with this hash+cd, or the entry with the
+ * next closest hash+cd.
+ */
+extern int zap_leaf_lookup_closest(zap_leaf_t *l,
+ uint64_t hash, uint32_t cd, zap_entry_handle_t *zeh);
+
+/*
+ * Read the first num_integers in the attribute. Integer size
+ * conversion will be done without sign extension. Return EINVAL if
+ * integer_size is too small. Return EOVERFLOW if there are more than
+ * num_integers in the attribute.
+ */
+extern int zap_entry_read(const zap_entry_handle_t *zeh,
+ uint8_t integer_size, uint64_t num_integers, void *buf);
+
+extern int zap_entry_read_name(const zap_entry_handle_t *zeh,
+ uint16_t buflen, char *buf);
+
+/*
+ * Replace the value of an existing entry.
+ *
+ * zap_entry_update may fail if it runs out of space (ENOSPC).
+ */
+extern int zap_entry_update(zap_entry_handle_t *zeh,
+ uint8_t integer_size, uint64_t num_integers, const void *buf);
+
+/*
+ * Remove an entry.
+ */
+extern void zap_entry_remove(zap_entry_handle_t *zeh);
+
+/*
+ * Create an entry. An equal entry must not exist, and this entry must
+ * belong in this leaf (according to its hash value). Fills in the
+ * entry handle on success. Returns 0 on success or ENOSPC on failure.
+ */
+extern int zap_entry_create(zap_leaf_t *l,
+ const char *name, uint64_t h, uint32_t cd,
+ uint8_t integer_size, uint64_t num_integers, const void *buf,
+ zap_entry_handle_t *zeh);
+
+/*
+ * Return true if there are additional entries with the same normalized
+ * form.
+ */
+extern boolean_t zap_entry_normalization_conflict(zap_entry_handle_t *zeh,
+ zap_name_t *zn, const char *name, zap_t *zap);
+
+/*
+ * Other stuff.
+ */
+
+extern void zap_leaf_init(zap_leaf_t *l, boolean_t sort);
+extern void zap_leaf_byteswap(zap_leaf_phys_t *buf, int len);
+extern void zap_leaf_split(zap_leaf_t *l, zap_leaf_t *nl, boolean_t sort);
+extern void zap_leaf_stats(zap_t *zap, zap_leaf_t *l, zap_stats_t *zs);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZAP_LEAF_H */
diff --git a/module/zfs/include/sys/zfs_acl.h b/module/zfs/include/sys/zfs_acl.h
new file mode 100644
index 000000000..bd91b33d1
--- /dev/null
+++ b/module/zfs/include/sys/zfs_acl.h
@@ -0,0 +1,214 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_ACL_H
+#define _SYS_FS_ZFS_ACL_H
+
+#ifdef _KERNEL
+#include <sys/isa_defs.h>
+#include <sys/types32.h>
+#endif
+#include <sys/acl.h>
+#include <sys/dmu.h>
+#include <sys/zfs_fuid.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct znode_phys;
+
+#define ACE_SLOT_CNT 6
+#define ZFS_ACL_VERSION_INITIAL 0ULL
+#define ZFS_ACL_VERSION_FUID 1ULL
+#define ZFS_ACL_VERSION ZFS_ACL_VERSION_FUID
+
+/*
+ * ZFS ACLs are store in various forms.
+ * Files created with ACL version ZFS_ACL_VERSION_INITIAL
+ * will all be created with fixed length ACEs of type
+ * zfs_oldace_t.
+ *
+ * Files with ACL version ZFS_ACL_VERSION_FUID will be created
+ * with various sized ACEs. The abstraction entries will utilize
+ * zfs_ace_hdr_t, normal user/group entries will use zfs_ace_t
+ * and some specialized CIFS ACEs will use zfs_object_ace_t.
+ */
+
+/*
+ * All ACEs have a common hdr. For
+ * owner@, group@, and everyone@ this is all
+ * thats needed.
+ */
+typedef struct zfs_ace_hdr {
+ uint16_t z_type;
+ uint16_t z_flags;
+ uint32_t z_access_mask;
+} zfs_ace_hdr_t;
+
+typedef zfs_ace_hdr_t zfs_ace_abstract_t;
+
+/*
+ * Standard ACE
+ */
+typedef struct zfs_ace {
+ zfs_ace_hdr_t z_hdr;
+ uint64_t z_fuid;
+} zfs_ace_t;
+
+/*
+ * The following type only applies to ACE_ACCESS_ALLOWED|DENIED_OBJECT_ACE_TYPE
+ * and will only be set/retrieved in a CIFS context.
+ */
+
+typedef struct zfs_object_ace {
+ zfs_ace_t z_ace;
+ uint8_t z_object_type[16]; /* object type */
+ uint8_t z_inherit_type[16]; /* inherited object type */
+} zfs_object_ace_t;
+
+typedef struct zfs_oldace {
+ uint32_t z_fuid; /* "who" */
+ uint32_t z_access_mask; /* access mask */
+ uint16_t z_flags; /* flags, i.e inheritance */
+ uint16_t z_type; /* type of entry allow/deny */
+} zfs_oldace_t;
+
+typedef struct zfs_acl_phys_v0 {
+ uint64_t z_acl_extern_obj; /* ext acl pieces */
+ uint32_t z_acl_count; /* Number of ACEs */
+ uint16_t z_acl_version; /* acl version */
+ uint16_t z_acl_pad; /* pad */
+ zfs_oldace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
+} zfs_acl_phys_v0_t;
+
+#define ZFS_ACE_SPACE (sizeof (zfs_oldace_t) * ACE_SLOT_CNT)
+
+typedef struct zfs_acl_phys {
+ uint64_t z_acl_extern_obj; /* ext acl pieces */
+ uint32_t z_acl_size; /* Number of bytes in ACL */
+ uint16_t z_acl_version; /* acl version */
+ uint16_t z_acl_count; /* ace count */
+ uint8_t z_ace_data[ZFS_ACE_SPACE]; /* space for embedded ACEs */
+} zfs_acl_phys_t;
+
+
+
+typedef struct acl_ops {
+ uint32_t (*ace_mask_get) (void *acep); /* get access mask */
+ void (*ace_mask_set) (void *acep,
+ uint32_t mask); /* set access mask */
+ uint16_t (*ace_flags_get) (void *acep); /* get flags */
+ void (*ace_flags_set) (void *acep,
+ uint16_t flags); /* set flags */
+ uint16_t (*ace_type_get)(void *acep); /* get type */
+ void (*ace_type_set)(void *acep,
+ uint16_t type); /* set type */
+ uint64_t (*ace_who_get)(void *acep); /* get who/fuid */
+ void (*ace_who_set)(void *acep,
+ uint64_t who); /* set who/fuid */
+ size_t (*ace_size)(void *acep); /* how big is this ace */
+ size_t (*ace_abstract_size)(void); /* sizeof abstract entry */
+ int (*ace_mask_off)(void); /* off of access mask in ace */
+ int (*ace_data)(void *acep, void **datap);
+ /* ptr to data if any */
+} acl_ops_t;
+
+/*
+ * A zfs_acl_t structure is composed of a list of zfs_acl_node_t's.
+ * Each node will have one or more ACEs associated with it. You will
+ * only have multiple nodes during a chmod operation. Normally only
+ * one node is required.
+ */
+typedef struct zfs_acl_node {
+ list_node_t z_next; /* Next chunk of ACEs */
+ void *z_acldata; /* pointer into actual ACE(s) */
+ void *z_allocdata; /* pointer to kmem allocated memory */
+ size_t z_allocsize; /* Size of blob in bytes */
+ size_t z_size; /* length of ACL data */
+ int z_ace_count; /* number of ACEs in this acl node */
+ int z_ace_idx; /* ace iterator positioned on */
+} zfs_acl_node_t;
+
+typedef struct zfs_acl {
+ int z_acl_count; /* Number of ACEs */
+ size_t z_acl_bytes; /* Number of bytes in ACL */
+ uint_t z_version; /* version of ACL */
+ void *z_next_ace; /* pointer to next ACE */
+ int z_hints; /* ACL hints (ZFS_INHERIT_ACE ...) */
+ zfs_acl_node_t *z_curr_node; /* current node iterator is handling */
+ list_t z_acl; /* chunks of ACE data */
+ acl_ops_t z_ops; /* ACL operations */
+ boolean_t z_has_fuids; /* FUIDs present in ACL? */
+} zfs_acl_t;
+
+#define ACL_DATA_ALLOCED 0x1
+#define ZFS_ACL_SIZE(aclcnt) (sizeof (ace_t) * (aclcnt))
+
+/*
+ * Property values for acl_mode and acl_inherit.
+ *
+ * acl_mode can take discard, noallow, groupmask and passthrough.
+ * whereas acl_inherit has secure instead of groupmask.
+ */
+
+#define ZFS_ACL_DISCARD 0
+#define ZFS_ACL_NOALLOW 1
+#define ZFS_ACL_GROUPMASK 2
+#define ZFS_ACL_PASSTHROUGH 3
+#define ZFS_ACL_RESTRICTED 4
+#define ZFS_ACL_PASSTHROUGH_X 5
+
+struct znode;
+struct zfsvfs;
+struct zfs_fuid_info;
+
+#ifdef _KERNEL
+void zfs_perm_init(struct znode *, struct znode *, int, vattr_t *,
+ dmu_tx_t *, cred_t *, zfs_acl_t *, zfs_fuid_info_t **);
+int zfs_getacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
+int zfs_setacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
+void zfs_acl_rele(void *);
+void zfs_oldace_byteswap(ace_t *, int);
+void zfs_ace_byteswap(void *, size_t, boolean_t);
+extern int zfs_zaccess(struct znode *, int, int, boolean_t, cred_t *);
+extern int zfs_zaccess_rwx(struct znode *, mode_t, int, cred_t *);
+extern int zfs_zaccess_unix(struct znode *, mode_t, cred_t *);
+extern int zfs_acl_access(struct znode *, int, cred_t *);
+int zfs_acl_chmod_setattr(struct znode *, zfs_acl_t **, uint64_t);
+int zfs_zaccess_delete(struct znode *, struct znode *, cred_t *);
+int zfs_zaccess_rename(struct znode *, struct znode *,
+ struct znode *, struct znode *, cred_t *cr);
+void zfs_acl_free(zfs_acl_t *);
+int zfs_vsec_2_aclp(struct zfsvfs *, vtype_t, vsecattr_t *, zfs_acl_t **);
+int zfs_aclset_common(struct znode *, zfs_acl_t *, cred_t *,
+ struct zfs_fuid_info **, dmu_tx_t *);
+
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+#endif /* _SYS_FS_ZFS_ACL_H */
diff --git a/module/zfs/include/sys/zfs_context.h b/module/zfs/include/sys/zfs_context.h
new file mode 100644
index 000000000..a5be3e130
--- /dev/null
+++ b/module/zfs/include/sys/zfs_context.h
@@ -0,0 +1,73 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZFS_CONTEXT_H
+#define _SYS_ZFS_CONTEXT_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <sys/note.h>
+#include <sys/types.h>
+#include <sys/t_lock.h>
+#include <sys/atomic.h>
+#include <sys/sysmacros.h>
+#include <sys/bitmap.h>
+#include <sys/cmn_err.h>
+#include <sys/kmem.h>
+#include <sys/taskq.h>
+#include <sys/buf.h>
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/cpuvar.h>
+#include <sys/kobj.h>
+#include <sys/conf.h>
+#include <sys/disp.h>
+#include <sys/debug.h>
+#include <sys/random.h>
+#include <sys/byteorder.h>
+#include <sys/systm.h>
+#include <sys/list.h>
+#include <sys/uio.h>
+#include <sys/dirent.h>
+#include <sys/time.h>
+#include <vm/seg_kmem.h>
+#include <sys/zone.h>
+#include <sys/uio.h>
+#include <sys/zfs_debug.h>
+#include <sys/sysevent.h>
+#include <sys/sysevent/eventdefs.h>
+#include <sys/fm/util.h>
+
+#define CPU_SEQID (CPU->cpu_seqid)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZFS_CONTEXT_H */
diff --git a/module/zfs/include/sys/zfs_ctldir.h b/module/zfs/include/sys/zfs_ctldir.h
new file mode 100644
index 000000000..ce29625d1
--- /dev/null
+++ b/module/zfs/include/sys/zfs_ctldir.h
@@ -0,0 +1,74 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _ZFS_CTLDIR_H
+#define _ZFS_CTLDIR_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/pathname.h>
+#include <sys/vnode.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/zfs_znode.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZFS_CTLDIR_NAME ".zfs"
+
+#define zfs_has_ctldir(zdp) \
+ ((zdp)->z_id == (zdp)->z_zfsvfs->z_root && \
+ ((zdp)->z_zfsvfs->z_ctldir != NULL))
+#define zfs_show_ctldir(zdp) \
+ (zfs_has_ctldir(zdp) && \
+ ((zdp)->z_zfsvfs->z_show_ctldir))
+
+void zfsctl_create(zfsvfs_t *);
+void zfsctl_destroy(zfsvfs_t *);
+vnode_t *zfsctl_root(znode_t *);
+void zfsctl_init(void);
+void zfsctl_fini(void);
+
+int zfsctl_rename_snapshot(const char *from, const char *to);
+int zfsctl_destroy_snapshot(const char *snapname, int force);
+int zfsctl_umount_snapshots(vfs_t *, int, cred_t *);
+
+int zfsctl_root_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
+ int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
+ int *direntflags, pathname_t *realpnp);
+
+int zfsctl_make_fid(zfsvfs_t *zfsvfsp, uint64_t object, uint32_t gen,
+ fid_t *fidp);
+int zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp);
+
+#define ZFSCTL_INO_ROOT 0x1
+#define ZFSCTL_INO_SNAPDIR 0x2
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _ZFS_CTLDIR_H */
diff --git a/module/zfs/include/sys/zfs_debug.h b/module/zfs/include/sys/zfs_debug.h
new file mode 100644
index 000000000..450ac1c81
--- /dev/null
+++ b/module/zfs/include/sys/zfs_debug.h
@@ -0,0 +1,75 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZFS_DEBUG_H
+#define _SYS_ZFS_DEBUG_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef TRUE
+#define TRUE 1
+#endif
+
+#ifndef FALSE
+#define FALSE 0
+#endif
+
+/*
+ * ZFS debugging
+ */
+
+#if defined(DEBUG) || !defined(_KERNEL)
+#define ZFS_DEBUG
+#endif
+
+extern int zfs_flags;
+
+#define ZFS_DEBUG_DPRINTF 0x0001
+#define ZFS_DEBUG_DBUF_VERIFY 0x0002
+#define ZFS_DEBUG_DNODE_VERIFY 0x0004
+#define ZFS_DEBUG_SNAPNAMES 0x0008
+#define ZFS_DEBUG_MODIFY 0x0010
+
+#ifdef ZFS_DEBUG
+extern void __dprintf(const char *file, const char *func,
+ int line, const char *fmt, ...);
+#define dprintf(...) \
+ if (zfs_flags & ZFS_DEBUG_DPRINTF) \
+ __dprintf(__FILE__, __func__, __LINE__, __VA_ARGS__)
+#else
+#define dprintf(...) ((void)0)
+#endif /* ZFS_DEBUG */
+
+extern void zfs_panic_recover(const char *fmt, ...);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZFS_DEBUG_H */
diff --git a/module/zfs/include/sys/zfs_dir.h b/module/zfs/include/sys/zfs_dir.h
new file mode 100644
index 000000000..ebb66e8ae
--- /dev/null
+++ b/module/zfs/include/sys/zfs_dir.h
@@ -0,0 +1,76 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_DIR_H
+#define _SYS_FS_ZFS_DIR_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/pathname.h>
+#include <sys/dmu.h>
+#include <sys/zfs_znode.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* zfs_dirent_lock() flags */
+#define ZNEW 0x0001 /* entry should not exist */
+#define ZEXISTS 0x0002 /* entry should exist */
+#define ZSHARED 0x0004 /* shared access (zfs_dirlook()) */
+#define ZXATTR 0x0008 /* we want the xattr dir */
+#define ZRENAMING 0x0010 /* znode is being renamed */
+#define ZCILOOK 0x0020 /* case-insensitive lookup requested */
+#define ZCIEXACT 0x0040 /* c-i requires c-s match (rename) */
+
+/* mknode flags */
+#define IS_ROOT_NODE 0x01 /* create a root node */
+#define IS_XATTR 0x02 /* create an extended attribute node */
+#define IS_REPLAY 0x04 /* we are replaying intent log */
+
+extern int zfs_dirent_lock(zfs_dirlock_t **, znode_t *, char *, znode_t **,
+ int, int *, pathname_t *);
+extern void zfs_dirent_unlock(zfs_dirlock_t *);
+extern int zfs_link_create(zfs_dirlock_t *, znode_t *, dmu_tx_t *, int);
+extern int zfs_link_destroy(zfs_dirlock_t *, znode_t *, dmu_tx_t *, int,
+ boolean_t *);
+extern int zfs_dirlook(znode_t *, char *, vnode_t **, int, int *,
+ pathname_t *);
+extern void zfs_mknode(znode_t *, vattr_t *, dmu_tx_t *, cred_t *,
+ uint_t, znode_t **, int, zfs_acl_t *, zfs_fuid_info_t **);
+extern void zfs_rmnode(znode_t *);
+extern void zfs_dl_name_switch(zfs_dirlock_t *dl, char *new, char **old);
+extern boolean_t zfs_dirempty(znode_t *);
+extern void zfs_unlinked_add(znode_t *, dmu_tx_t *);
+extern void zfs_unlinked_drain(zfsvfs_t *zfsvfs);
+extern int zfs_sticky_remove_access(znode_t *, znode_t *, cred_t *cr);
+extern int zfs_get_xattrdir(znode_t *, vnode_t **, cred_t *, int);
+extern int zfs_make_xattrdir(znode_t *, vattr_t *, vnode_t **, cred_t *);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_FS_ZFS_DIR_H */
diff --git a/module/zfs/include/sys/zfs_fuid.h b/module/zfs/include/sys/zfs_fuid.h
new file mode 100644
index 000000000..810ffc81a
--- /dev/null
+++ b/module/zfs/include/sys/zfs_fuid.h
@@ -0,0 +1,125 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_FUID_H
+#define _SYS_FS_ZFS_FUID_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef _KERNEL
+#include <sys/kidmap.h>
+#include <sys/sid.h>
+#include <sys/dmu.h>
+#include <sys/zfs_vfsops.h>
+#endif
+#include <sys/avl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef enum {
+ ZFS_OWNER,
+ ZFS_GROUP,
+ ZFS_ACE_USER,
+ ZFS_ACE_GROUP
+} zfs_fuid_type_t;
+
+/*
+ * Estimate space needed for one more fuid table entry.
+ * for now assume its current size + 1K
+ */
+#define FUID_SIZE_ESTIMATE(z) (z->z_fuid_size + (SPA_MINBLOCKSIZE << 1))
+
+#define FUID_INDEX(x) (x >> 32)
+#define FUID_RID(x) (x & 0xffffffff)
+#define FUID_ENCODE(idx, rid) ((idx << 32) | rid)
+/*
+ * FUIDs cause problems for the intent log
+ * we need to replay the creation of the FUID,
+ * but we can't count on the idmapper to be around
+ * and during replay the FUID index may be different than
+ * before. Also, if an ACL has 100 ACEs and 12 different
+ * domains we don't want to log 100 domain strings, but rather
+ * just the unique 12.
+ */
+
+/*
+ * The FUIDs in the log will index into
+ * domain string table and the bottom half will be the rid.
+ * Used for mapping ephemeral uid/gid during ACL setting to FUIDs
+ */
+typedef struct zfs_fuid {
+ list_node_t z_next;
+ uint64_t z_id; /* uid/gid being converted to fuid */
+ uint64_t z_domidx; /* index in AVL domain table */
+ uint64_t z_logfuid; /* index for domain in log */
+} zfs_fuid_t;
+
+/* list of unique domains */
+typedef struct zfs_fuid_domain {
+ list_node_t z_next;
+ uint64_t z_domidx; /* AVL tree idx */
+ const char *z_domain; /* domain string */
+} zfs_fuid_domain_t;
+
+/*
+ * FUID information necessary for logging create, setattr, and setacl.
+ */
+typedef struct zfs_fuid_info {
+ list_t z_fuids;
+ list_t z_domains;
+ uint64_t z_fuid_owner;
+ uint64_t z_fuid_group;
+ char **z_domain_table; /* Used during replay */
+ uint32_t z_fuid_cnt; /* How many fuids in z_fuids */
+ uint32_t z_domain_cnt; /* How many domains */
+ size_t z_domain_str_sz; /* len of domain strings z_domain list */
+} zfs_fuid_info_t;
+
+#ifdef _KERNEL
+struct znode;
+extern uid_t zfs_fuid_map_id(zfsvfs_t *, uint64_t, cred_t *, zfs_fuid_type_t);
+extern void zfs_fuid_destroy(zfsvfs_t *);
+extern uint64_t zfs_fuid_create_cred(zfsvfs_t *, zfs_fuid_type_t,
+ dmu_tx_t *, cred_t *, zfs_fuid_info_t **);
+extern uint64_t zfs_fuid_create(zfsvfs_t *, uint64_t, cred_t *, zfs_fuid_type_t,
+ dmu_tx_t *, zfs_fuid_info_t **);
+extern void zfs_fuid_map_ids(struct znode *zp, cred_t *cr, uid_t *uid,
+ uid_t *gid);
+extern zfs_fuid_info_t *zfs_fuid_info_alloc(void);
+extern void zfs_fuid_info_free();
+extern boolean_t zfs_groupmember(zfsvfs_t *, uint64_t, cred_t *);
+#endif
+
+char *zfs_fuid_idx_domain(avl_tree_t *, uint32_t);
+uint64_t zfs_fuid_table_load(objset_t *, uint64_t, avl_tree_t *, avl_tree_t *);
+void zfs_fuid_table_destroy(avl_tree_t *, avl_tree_t *);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_FS_ZFS_FUID_H */
diff --git a/module/zfs/include/sys/zfs_ioctl.h b/module/zfs/include/sys/zfs_ioctl.h
new file mode 100644
index 000000000..1692608bb
--- /dev/null
+++ b/module/zfs/include/sys/zfs_ioctl.h
@@ -0,0 +1,196 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZFS_IOCTL_H
+#define _SYS_ZFS_IOCTL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/cred.h>
+#include <sys/dmu.h>
+#include <sys/zio.h>
+#include <sys/dsl_deleg.h>
+
+#ifdef _KERNEL
+#include <sys/nvpair.h>
+#endif /* _KERNEL */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Property values for snapdir
+ */
+#define ZFS_SNAPDIR_HIDDEN 0
+#define ZFS_SNAPDIR_VISIBLE 1
+
+#define DMU_BACKUP_STREAM_VERSION (1ULL)
+#define DMU_BACKUP_HEADER_VERSION (2ULL)
+#define DMU_BACKUP_MAGIC 0x2F5bacbacULL
+
+#define DRR_FLAG_CLONE (1<<0)
+#define DRR_FLAG_CI_DATA (1<<1)
+
+/*
+ * zfs ioctl command structure
+ */
+typedef struct dmu_replay_record {
+ enum {
+ DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
+ DRR_WRITE, DRR_FREE, DRR_END,
+ } drr_type;
+ uint32_t drr_payloadlen;
+ union {
+ struct drr_begin {
+ uint64_t drr_magic;
+ uint64_t drr_version;
+ uint64_t drr_creation_time;
+ dmu_objset_type_t drr_type;
+ uint32_t drr_flags;
+ uint64_t drr_toguid;
+ uint64_t drr_fromguid;
+ char drr_toname[MAXNAMELEN];
+ } drr_begin;
+ struct drr_end {
+ zio_cksum_t drr_checksum;
+ } drr_end;
+ struct drr_object {
+ uint64_t drr_object;
+ dmu_object_type_t drr_type;
+ dmu_object_type_t drr_bonustype;
+ uint32_t drr_blksz;
+ uint32_t drr_bonuslen;
+ uint8_t drr_checksum;
+ uint8_t drr_compress;
+ uint8_t drr_pad[6];
+ /* bonus content follows */
+ } drr_object;
+ struct drr_freeobjects {
+ uint64_t drr_firstobj;
+ uint64_t drr_numobjs;
+ } drr_freeobjects;
+ struct drr_write {
+ uint64_t drr_object;
+ dmu_object_type_t drr_type;
+ uint32_t drr_pad;
+ uint64_t drr_offset;
+ uint64_t drr_length;
+ /* content follows */
+ } drr_write;
+ struct drr_free {
+ uint64_t drr_object;
+ uint64_t drr_offset;
+ uint64_t drr_length;
+ } drr_free;
+ } drr_u;
+} dmu_replay_record_t;
+
+typedef struct zinject_record {
+ uint64_t zi_objset;
+ uint64_t zi_object;
+ uint64_t zi_start;
+ uint64_t zi_end;
+ uint64_t zi_guid;
+ uint32_t zi_level;
+ uint32_t zi_error;
+ uint64_t zi_type;
+ uint32_t zi_freq;
+ uint32_t zi_pad; /* pad out to 64 bit alignment */
+} zinject_record_t;
+
+#define ZINJECT_NULL 0x1
+#define ZINJECT_FLUSH_ARC 0x2
+#define ZINJECT_UNLOAD_SPA 0x4
+
+typedef struct zfs_share {
+ uint64_t z_exportdata;
+ uint64_t z_sharedata;
+ uint64_t z_sharetype; /* 0 = share, 1 = unshare */
+ uint64_t z_sharemax; /* max length of share string */
+} zfs_share_t;
+
+/*
+ * ZFS file systems may behave the usual, POSIX-compliant way, where
+ * name lookups are case-sensitive. They may also be set up so that
+ * all the name lookups are case-insensitive, or so that only some
+ * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
+ */
+typedef enum zfs_case {
+ ZFS_CASE_SENSITIVE,
+ ZFS_CASE_INSENSITIVE,
+ ZFS_CASE_MIXED
+} zfs_case_t;
+
+typedef struct zfs_cmd {
+ char zc_name[MAXPATHLEN];
+ char zc_value[MAXPATHLEN * 2];
+ char zc_string[MAXNAMELEN];
+ uint64_t zc_guid;
+ uint64_t zc_nvlist_conf; /* really (char *) */
+ uint64_t zc_nvlist_conf_size;
+ uint64_t zc_nvlist_src; /* really (char *) */
+ uint64_t zc_nvlist_src_size;
+ uint64_t zc_nvlist_dst; /* really (char *) */
+ uint64_t zc_nvlist_dst_size;
+ uint64_t zc_cookie;
+ uint64_t zc_objset_type;
+ uint64_t zc_perm_action;
+ uint64_t zc_history; /* really (char *) */
+ uint64_t zc_history_len;
+ uint64_t zc_history_offset;
+ uint64_t zc_obj;
+ zfs_share_t zc_share;
+ dmu_objset_stats_t zc_objset_stats;
+ struct drr_begin zc_begin_record;
+ zinject_record_t zc_inject_record;
+} zfs_cmd_t;
+
+#define ZVOL_MAX_MINOR (1 << 16)
+#define ZFS_MIN_MINOR (ZVOL_MAX_MINOR + 1)
+
+#ifdef _KERNEL
+
+typedef struct zfs_creat {
+ nvlist_t *zct_zplprops;
+ nvlist_t *zct_props;
+} zfs_creat_t;
+
+extern dev_info_t *zfs_dip;
+
+extern int zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr);
+extern int zfs_secpolicy_rename_perms(const char *from,
+ const char *to, cred_t *cr);
+extern int zfs_secpolicy_destroy_perms(const char *name, cred_t *cr);
+extern int zfs_busy(void);
+extern int zfs_unmount_snap(char *, void *);
+
+#endif /* _KERNEL */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZFS_IOCTL_H */
diff --git a/module/zfs/include/sys/zfs_rlock.h b/module/zfs/include/sys/zfs_rlock.h
new file mode 100644
index 000000000..f302b663e
--- /dev/null
+++ b/module/zfs/include/sys/zfs_rlock.h
@@ -0,0 +1,89 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_RLOCK_H
+#define _SYS_FS_ZFS_RLOCK_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef _KERNEL
+
+#include <sys/zfs_znode.h>
+
+typedef enum {
+ RL_READER,
+ RL_WRITER,
+ RL_APPEND
+} rl_type_t;
+
+typedef struct rl {
+ znode_t *r_zp; /* znode this lock applies to */
+ avl_node_t r_node; /* avl node link */
+ uint64_t r_off; /* file range offset */
+ uint64_t r_len; /* file range length */
+ uint_t r_cnt; /* range reference count in tree */
+ rl_type_t r_type; /* range type */
+ kcondvar_t r_wr_cv; /* cv for waiting writers */
+ kcondvar_t r_rd_cv; /* cv for waiting readers */
+ uint8_t r_proxy; /* acting for original range */
+ uint8_t r_write_wanted; /* writer wants to lock this range */
+ uint8_t r_read_wanted; /* reader wants to lock this range */
+} rl_t;
+
+/*
+ * Lock a range (offset, length) as either shared (READER)
+ * or exclusive (WRITER or APPEND). APPEND is a special type that
+ * is converted to WRITER that specified to lock from the start of the
+ * end of file. zfs_range_lock() returns the range lock structure.
+ */
+rl_t *zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type);
+
+/*
+ * Unlock range and destroy range lock structure.
+ */
+void zfs_range_unlock(rl_t *rl);
+
+/*
+ * Reduce range locked as RW_WRITER from whole file to specified range.
+ * Asserts the whole file was previously locked.
+ */
+void zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len);
+
+/*
+ * AVL comparison function used to compare range locks
+ */
+int zfs_range_compare(const void *arg1, const void *arg2);
+
+#endif /* _KERNEL */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_FS_ZFS_RLOCK_H */
diff --git a/module/zfs/include/sys/zfs_vfsops.h b/module/zfs/include/sys/zfs_vfsops.h
new file mode 100644
index 000000000..87b75e6e7
--- /dev/null
+++ b/module/zfs/include/sys/zfs_vfsops.h
@@ -0,0 +1,140 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_VFSOPS_H
+#define _SYS_FS_ZFS_VFSOPS_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/isa_defs.h>
+#include <sys/types32.h>
+#include <sys/list.h>
+#include <sys/vfs.h>
+#include <sys/zil.h>
+#include <sys/rrwlock.h>
+#include <sys/zfs_ioctl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct zfsvfs zfsvfs_t;
+
+struct zfsvfs {
+ vfs_t *z_vfs; /* generic fs struct */
+ zfsvfs_t *z_parent; /* parent fs */
+ objset_t *z_os; /* objset reference */
+ uint64_t z_root; /* id of root znode */
+ uint64_t z_unlinkedobj; /* id of unlinked zapobj */
+ uint64_t z_max_blksz; /* maximum block size for files */
+ uint64_t z_assign; /* TXG_NOWAIT or set by zil_replay() */
+ uint64_t z_fuid_obj; /* fuid table object number */
+ uint64_t z_fuid_size; /* fuid table size */
+ avl_tree_t z_fuid_idx; /* fuid tree keyed by index */
+ avl_tree_t z_fuid_domain; /* fuid tree keyed by domain */
+ krwlock_t z_fuid_lock; /* fuid lock */
+ boolean_t z_fuid_loaded; /* fuid tables are loaded */
+ struct zfs_fuid_info *z_fuid_replay; /* fuid info for replay */
+ zilog_t *z_log; /* intent log pointer */
+ uint_t z_acl_mode; /* acl chmod/mode behavior */
+ uint_t z_acl_inherit; /* acl inheritance behavior */
+ zfs_case_t z_case; /* case-sense */
+ boolean_t z_utf8; /* utf8-only */
+ int z_norm; /* normalization flags */
+ boolean_t z_atime; /* enable atimes mount option */
+ boolean_t z_unmounted; /* unmounted */
+ rrwlock_t z_teardown_lock;
+ krwlock_t z_teardown_inactive_lock;
+ list_t z_all_znodes; /* all vnodes in the fs */
+ kmutex_t z_znodes_lock; /* lock for z_all_znodes */
+ vnode_t *z_ctldir; /* .zfs directory pointer */
+ boolean_t z_show_ctldir; /* expose .zfs in the root dir */
+ boolean_t z_issnap; /* true if this is a snapshot */
+ boolean_t z_vscan; /* virus scan on/off */
+ boolean_t z_use_fuids; /* version allows fuids */
+ kmutex_t z_online_recv_lock; /* recv in prog grabs as WRITER */
+ uint64_t z_version; /* ZPL version */
+#define ZFS_OBJ_MTX_SZ 64
+ kmutex_t z_hold_mtx[ZFS_OBJ_MTX_SZ]; /* znode hold locks */
+};
+
+/*
+ * Normal filesystems (those not under .zfs/snapshot) have a total
+ * file ID size limited to 12 bytes (including the length field) due to
+ * NFSv2 protocol's limitation of 32 bytes for a filehandle. For historical
+ * reasons, this same limit is being imposed by the Solaris NFSv3 implementation
+ * (although the NFSv3 protocol actually permits a maximum of 64 bytes). It
+ * is not possible to expand beyond 12 bytes without abandoning support
+ * of NFSv2.
+ *
+ * For normal filesystems, we partition up the available space as follows:
+ * 2 bytes fid length (required)
+ * 6 bytes object number (48 bits)
+ * 4 bytes generation number (32 bits)
+ *
+ * We reserve only 48 bits for the object number, as this is the limit
+ * currently defined and imposed by the DMU.
+ */
+typedef struct zfid_short {
+ uint16_t zf_len;
+ uint8_t zf_object[6]; /* obj[i] = obj >> (8 * i) */
+ uint8_t zf_gen[4]; /* gen[i] = gen >> (8 * i) */
+} zfid_short_t;
+
+/*
+ * Filesystems under .zfs/snapshot have a total file ID size of 22 bytes
+ * (including the length field). This makes files under .zfs/snapshot
+ * accessible by NFSv3 and NFSv4, but not NFSv2.
+ *
+ * For files under .zfs/snapshot, we partition up the available space
+ * as follows:
+ * 2 bytes fid length (required)
+ * 6 bytes object number (48 bits)
+ * 4 bytes generation number (32 bits)
+ * 6 bytes objset id (48 bits)
+ * 4 bytes currently just zero (32 bits)
+ *
+ * We reserve only 48 bits for the object number and objset id, as these are
+ * the limits currently defined and imposed by the DMU.
+ */
+typedef struct zfid_long {
+ zfid_short_t z_fid;
+ uint8_t zf_setid[6]; /* obj[i] = obj >> (8 * i) */
+ uint8_t zf_setgen[4]; /* gen[i] = gen >> (8 * i) */
+} zfid_long_t;
+
+#define SHORT_FID_LEN (sizeof (zfid_short_t) - sizeof (uint16_t))
+#define LONG_FID_LEN (sizeof (zfid_long_t) - sizeof (uint16_t))
+
+extern uint_t zfs_fsyncer_key;
+
+extern int zfs_suspend_fs(zfsvfs_t *zfsvfs, char *osname, int *mode);
+extern int zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_FS_ZFS_VFSOPS_H */
diff --git a/module/zfs/include/sys/zfs_znode.h b/module/zfs/include/sys/zfs_znode.h
new file mode 100644
index 000000000..a5416525c
--- /dev/null
+++ b/module/zfs/include/sys/zfs_znode.h
@@ -0,0 +1,356 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_FS_ZFS_ZNODE_H
+#define _SYS_FS_ZFS_ZNODE_H
+
+#ifdef _KERNEL
+#include <sys/isa_defs.h>
+#include <sys/types32.h>
+#include <sys/attr.h>
+#include <sys/list.h>
+#include <sys/dmu.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/rrwlock.h>
+#endif
+#include <sys/zfs_acl.h>
+#include <sys/zil.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Additional file level attributes, that are stored
+ * in the upper half of zp_flags
+ */
+#define ZFS_READONLY 0x0000000100000000
+#define ZFS_HIDDEN 0x0000000200000000
+#define ZFS_SYSTEM 0x0000000400000000
+#define ZFS_ARCHIVE 0x0000000800000000
+#define ZFS_IMMUTABLE 0x0000001000000000
+#define ZFS_NOUNLINK 0x0000002000000000
+#define ZFS_APPENDONLY 0x0000004000000000
+#define ZFS_NODUMP 0x0000008000000000
+#define ZFS_OPAQUE 0x0000010000000000
+#define ZFS_AV_QUARANTINED 0x0000020000000000
+#define ZFS_AV_MODIFIED 0x0000040000000000
+
+#define ZFS_ATTR_SET(zp, attr, value) \
+{ \
+ if (value) \
+ zp->z_phys->zp_flags |= attr; \
+ else \
+ zp->z_phys->zp_flags &= ~attr; \
+}
+
+/*
+ * Define special zfs pflags
+ */
+#define ZFS_XATTR 0x1 /* is an extended attribute */
+#define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
+#define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
+#define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */
+#define ZFS_ACL_PROTECTED 0x10 /* ACL protected */
+#define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */
+#define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */
+#define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */
+
+/*
+ * Is ID ephemeral?
+ */
+#define IS_EPHEMERAL(x) (x > MAXUID)
+
+/*
+ * Should we use FUIDs?
+ */
+#define USE_FUIDS(version, os) (version >= ZPL_VERSION_FUID &&\
+ spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
+
+#define MASTER_NODE_OBJ 1
+
+/*
+ * Special attributes for master node.
+ */
+#define ZFS_FSID "FSID"
+#define ZFS_UNLINKED_SET "DELETE_QUEUE"
+#define ZFS_ROOT_OBJ "ROOT"
+#define ZPL_VERSION_STR "VERSION"
+#define ZFS_FUID_TABLES "FUID"
+
+#define ZFS_MAX_BLOCKSIZE (SPA_MAXBLOCKSIZE)
+
+/* Path component length */
+/*
+ * The generic fs code uses MAXNAMELEN to represent
+ * what the largest component length is. Unfortunately,
+ * this length includes the terminating NULL. ZFS needs
+ * to tell the users via pathconf() and statvfs() what the
+ * true maximum length of a component is, excluding the NULL.
+ */
+#define ZFS_MAXNAMELEN (MAXNAMELEN - 1)
+
+/*
+ * Convert mode bits (zp_mode) to BSD-style DT_* values for storing in
+ * the directory entries.
+ */
+#define IFTODT(mode) (((mode) & S_IFMT) >> 12)
+
+/*
+ * The directory entry has the type (currently unused on Solaris) in the
+ * top 4 bits, and the object number in the low 48 bits. The "middle"
+ * 12 bits are unused.
+ */
+#define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
+#define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
+
+/*
+ * This is the persistent portion of the znode. It is stored
+ * in the "bonus buffer" of the file. Short symbolic links
+ * are also stored in the bonus buffer.
+ */
+typedef struct znode_phys {
+ uint64_t zp_atime[2]; /* 0 - last file access time */
+ uint64_t zp_mtime[2]; /* 16 - last file modification time */
+ uint64_t zp_ctime[2]; /* 32 - last file change time */
+ uint64_t zp_crtime[2]; /* 48 - creation time */
+ uint64_t zp_gen; /* 64 - generation (txg of creation) */
+ uint64_t zp_mode; /* 72 - file mode bits */
+ uint64_t zp_size; /* 80 - size of file */
+ uint64_t zp_parent; /* 88 - directory parent (`..') */
+ uint64_t zp_links; /* 96 - number of links to file */
+ uint64_t zp_xattr; /* 104 - DMU object for xattrs */
+ uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
+ uint64_t zp_flags; /* 120 - persistent flags */
+ uint64_t zp_uid; /* 128 - file owner */
+ uint64_t zp_gid; /* 136 - owning group */
+ uint64_t zp_zap; /* 144 - extra attributes */
+ uint64_t zp_pad[3]; /* 152 - future */
+ zfs_acl_phys_t zp_acl; /* 176 - 263 ACL */
+ /*
+ * Data may pad out any remaining bytes in the znode buffer, eg:
+ *
+ * |<---------------------- dnode_phys (512) ------------------------>|
+ * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
+ * |<---- znode (264) ---->|<---- data (56) ---->|
+ *
+ * At present, we use this space for the following:
+ * - symbolic links
+ * - 32-byte anti-virus scanstamp (regular files only)
+ */
+} znode_phys_t;
+
+/*
+ * Directory entry locks control access to directory entries.
+ * They are used to protect creates, deletes, and renames.
+ * Each directory znode has a mutex and a list of locked names.
+ */
+#ifdef _KERNEL
+typedef struct zfs_dirlock {
+ char *dl_name; /* directory entry being locked */
+ uint32_t dl_sharecnt; /* 0 if exclusive, > 0 if shared */
+ uint16_t dl_namesize; /* set if dl_name was allocated */
+ kcondvar_t dl_cv; /* wait for entry to be unlocked */
+ struct znode *dl_dzp; /* directory znode */
+ struct zfs_dirlock *dl_next; /* next in z_dirlocks list */
+} zfs_dirlock_t;
+
+typedef struct znode {
+ struct zfsvfs *z_zfsvfs;
+ vnode_t *z_vnode;
+ uint64_t z_id; /* object ID for this znode */
+ kmutex_t z_lock; /* znode modification lock */
+ krwlock_t z_map_lock; /* page map lock */
+ krwlock_t z_parent_lock; /* parent lock for directories */
+ krwlock_t z_name_lock; /* "master" lock for dirent locks */
+ zfs_dirlock_t *z_dirlocks; /* directory entry lock list */
+ kmutex_t z_range_lock; /* protects changes to z_range_avl */
+ avl_tree_t z_range_avl; /* avl tree of file range locks */
+ uint8_t z_unlinked; /* file has been unlinked */
+ uint8_t z_atime_dirty; /* atime needs to be synced */
+ uint8_t z_zn_prefetch; /* Prefetch znodes? */
+ uint_t z_blksz; /* block size in bytes */
+ uint_t z_seq; /* modification sequence number */
+ uint64_t z_mapcnt; /* number of pages mapped to file */
+ uint64_t z_last_itx; /* last ZIL itx on this znode */
+ uint64_t z_gen; /* generation (same as zp_gen) */
+ uint32_t z_sync_cnt; /* synchronous open count */
+ kmutex_t z_acl_lock; /* acl data lock */
+ list_node_t z_link_node; /* all znodes in fs link */
+ /*
+ * These are dmu managed fields.
+ */
+ znode_phys_t *z_phys; /* pointer to persistent znode */
+ dmu_buf_t *z_dbuf; /* buffer containing the z_phys */
+} znode_t;
+
+
+/*
+ * Range locking rules
+ * --------------------
+ * 1. When truncating a file (zfs_create, zfs_setattr, zfs_space) the whole
+ * file range needs to be locked as RL_WRITER. Only then can the pages be
+ * freed etc and zp_size reset. zp_size must be set within range lock.
+ * 2. For writes and punching holes (zfs_write & zfs_space) just the range
+ * being written or freed needs to be locked as RL_WRITER.
+ * Multiple writes at the end of the file must coordinate zp_size updates
+ * to ensure data isn't lost. A compare and swap loop is currently used
+ * to ensure the file size is at least the offset last written.
+ * 3. For reads (zfs_read, zfs_get_data & zfs_putapage) just the range being
+ * read needs to be locked as RL_READER. A check against zp_size can then
+ * be made for reading beyond end of file.
+ */
+
+/*
+ * Convert between znode pointers and vnode pointers
+ */
+#define ZTOV(ZP) ((ZP)->z_vnode)
+#define VTOZ(VP) ((znode_t *)(VP)->v_data)
+
+/*
+ * ZFS_ENTER() is called on entry to each ZFS vnode and vfs operation.
+ * ZFS_EXIT() must be called before exitting the vop.
+ * ZFS_VERIFY_ZP() verifies the znode is valid.
+ */
+#define ZFS_ENTER(zfsvfs) \
+ { \
+ rrw_enter(&(zfsvfs)->z_teardown_lock, RW_READER, FTAG); \
+ if ((zfsvfs)->z_unmounted) { \
+ ZFS_EXIT(zfsvfs); \
+ return (EIO); \
+ } \
+ }
+
+#define ZFS_EXIT(zfsvfs) rrw_exit(&(zfsvfs)->z_teardown_lock, FTAG)
+
+#define ZFS_VERIFY_ZP(zp) \
+ if ((zp)->z_dbuf == NULL) { \
+ ZFS_EXIT((zp)->z_zfsvfs); \
+ return (EIO); \
+ } \
+
+/*
+ * Macros for dealing with dmu_buf_hold
+ */
+#define ZFS_OBJ_HASH(obj_num) ((obj_num) & (ZFS_OBJ_MTX_SZ - 1))
+#define ZFS_OBJ_MUTEX(zfsvfs, obj_num) \
+ (&(zfsvfs)->z_hold_mtx[ZFS_OBJ_HASH(obj_num)])
+#define ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num) \
+ mutex_enter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
+#define ZFS_OBJ_HOLD_TRYENTER(zfsvfs, obj_num) \
+ mutex_tryenter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
+#define ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num) \
+ mutex_exit(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
+
+/*
+ * Macros to encode/decode ZFS stored time values from/to struct timespec
+ */
+#define ZFS_TIME_ENCODE(tp, stmp) \
+{ \
+ (stmp)[0] = (uint64_t)(tp)->tv_sec; \
+ (stmp)[1] = (uint64_t)(tp)->tv_nsec; \
+}
+
+#define ZFS_TIME_DECODE(tp, stmp) \
+{ \
+ (tp)->tv_sec = (time_t)(stmp)[0]; \
+ (tp)->tv_nsec = (long)(stmp)[1]; \
+}
+
+/*
+ * Timestamp defines
+ */
+#define ACCESSED (AT_ATIME)
+#define STATE_CHANGED (AT_CTIME)
+#define CONTENT_MODIFIED (AT_MTIME | AT_CTIME)
+
+#define ZFS_ACCESSTIME_STAMP(zfsvfs, zp) \
+ if ((zfsvfs)->z_atime && !((zfsvfs)->z_vfs->vfs_flag & VFS_RDONLY)) \
+ zfs_time_stamper(zp, ACCESSED, NULL)
+
+extern int zfs_init_fs(zfsvfs_t *, znode_t **);
+extern void zfs_set_dataprop(objset_t *);
+extern void zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *,
+ dmu_tx_t *tx);
+extern void zfs_time_stamper(znode_t *, uint_t, dmu_tx_t *);
+extern void zfs_time_stamper_locked(znode_t *, uint_t, dmu_tx_t *);
+extern void zfs_grow_blocksize(znode_t *, uint64_t, dmu_tx_t *);
+extern int zfs_freesp(znode_t *, uint64_t, uint64_t, int, boolean_t);
+extern void zfs_znode_init(void);
+extern void zfs_znode_fini(void);
+extern int zfs_zget(zfsvfs_t *, uint64_t, znode_t **);
+extern int zfs_rezget(znode_t *);
+extern void zfs_zinactive(znode_t *);
+extern void zfs_znode_delete(znode_t *, dmu_tx_t *);
+extern void zfs_znode_free(znode_t *);
+extern void zfs_remove_op_tables();
+extern int zfs_create_op_tables();
+extern int zfs_sync(vfs_t *vfsp, short flag, cred_t *cr);
+extern dev_t zfs_cmpldev(uint64_t);
+extern int zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value);
+extern int zfs_set_version(const char *name, uint64_t newvers);
+extern int zfs_get_stats(objset_t *os, nvlist_t *nv);
+extern void zfs_znode_dmu_fini(znode_t *);
+
+extern void zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name, vsecattr_t *, zfs_fuid_info_t *,
+ vattr_t *vap);
+extern int zfs_log_create_txtype(zil_create_t, vsecattr_t *vsecp,
+ vattr_t *vap);
+extern void zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, char *name);
+extern void zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name);
+extern void zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name, char *link);
+extern void zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp);
+extern void zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, offset_t off, ssize_t len, int ioflag);
+extern void zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, uint64_t off, uint64_t len);
+extern void zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp);
+extern void zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
+ vsecattr_t *vsecp, zfs_fuid_info_t *fuidp);
+extern void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap);
+extern void zfs_upgrade(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
+
+extern caddr_t zfs_map_page(page_t *, enum seg_rw);
+extern void zfs_unmap_page(page_t *, caddr_t);
+
+extern zil_get_data_t zfs_get_data;
+extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
+extern int zfsfstype;
+
+#endif /* _KERNEL */
+
+extern int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_FS_ZFS_ZNODE_H */
diff --git a/module/zfs/include/sys/zil.h b/module/zfs/include/sys/zil.h
new file mode 100644
index 000000000..4d02d14f7
--- /dev/null
+++ b/module/zfs/include/sys/zil.h
@@ -0,0 +1,382 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZIL_H
+#define _SYS_ZIL_H
+
+#include <sys/types.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Intent log format:
+ *
+ * Each objset has its own intent log. The log header (zil_header_t)
+ * for objset N's intent log is kept in the Nth object of the SPA's
+ * intent_log objset. The log header points to a chain of log blocks,
+ * each of which contains log records (i.e., transactions) followed by
+ * a log block trailer (zil_trailer_t). The format of a log record
+ * depends on the record (or transaction) type, but all records begin
+ * with a common structure that defines the type, length, and txg.
+ */
+
+/*
+ * Intent log header - this on disk structure holds fields to manage
+ * the log. All fields are 64 bit to easily handle cross architectures.
+ */
+typedef struct zil_header {
+ uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
+ uint64_t zh_replay_seq; /* highest replayed sequence number */
+ blkptr_t zh_log; /* log chain */
+ uint64_t zh_claim_seq; /* highest claimed sequence number */
+ uint64_t zh_pad[5];
+} zil_header_t;
+
+/*
+ * Log block trailer - structure at the end of the header and each log block
+ *
+ * The zit_bt contains a zbt_cksum which for the intent log is
+ * the sequence number of this log block. A seq of 0 is invalid.
+ * The zbt_cksum is checked by the SPA against the sequence
+ * number passed in the blk_cksum field of the blkptr_t
+ */
+typedef struct zil_trailer {
+ uint64_t zit_pad;
+ blkptr_t zit_next_blk; /* next block in chain */
+ uint64_t zit_nused; /* bytes in log block used */
+ zio_block_tail_t zit_bt; /* block trailer */
+} zil_trailer_t;
+
+#define ZIL_MIN_BLKSZ 4096ULL
+#define ZIL_MAX_BLKSZ SPA_MAXBLOCKSIZE
+#define ZIL_BLK_DATA_SZ(lwb) ((lwb)->lwb_sz - sizeof (zil_trailer_t))
+
+/*
+ * The words of a log block checksum.
+ */
+#define ZIL_ZC_GUID_0 0
+#define ZIL_ZC_GUID_1 1
+#define ZIL_ZC_OBJSET 2
+#define ZIL_ZC_SEQ 3
+
+typedef enum zil_create {
+ Z_FILE,
+ Z_DIR,
+ Z_XATTRDIR,
+} zil_create_t;
+
+/*
+ * size of xvattr log section.
+ * its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps
+ * for create time and a single 64 bit integer for all of the attributes,
+ * and 4 64 bit integers (32 bytes) for the scanstamp.
+ *
+ */
+
+#define ZIL_XVAT_SIZE(mapsize) \
+ sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \
+ (sizeof (uint64_t) * 7)
+
+/*
+ * Size of ACL in log. The ACE data is padded out to properly align
+ * on 8 byte boundary.
+ */
+
+#define ZIL_ACE_LENGTH(x) (roundup(x, sizeof (uint64_t)))
+
+/*
+ * Intent log transaction types and record structures
+ */
+#define TX_CREATE 1 /* Create file */
+#define TX_MKDIR 2 /* Make directory */
+#define TX_MKXATTR 3 /* Make XATTR directory */
+#define TX_SYMLINK 4 /* Create symbolic link to a file */
+#define TX_REMOVE 5 /* Remove file */
+#define TX_RMDIR 6 /* Remove directory */
+#define TX_LINK 7 /* Create hard link to a file */
+#define TX_RENAME 8 /* Rename a file */
+#define TX_WRITE 9 /* File write */
+#define TX_TRUNCATE 10 /* Truncate a file */
+#define TX_SETATTR 11 /* Set file attributes */
+#define TX_ACL_V0 12 /* Set old formatted ACL */
+#define TX_ACL 13 /* Set ACL */
+#define TX_CREATE_ACL 14 /* create with ACL */
+#define TX_CREATE_ATTR 15 /* create + attrs */
+#define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */
+#define TX_MKDIR_ACL 17 /* mkdir with ACL */
+#define TX_MKDIR_ATTR 18 /* mkdir with attr */
+#define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */
+#define TX_MAX_TYPE 20 /* Max transaction type */
+
+/*
+ * The transactions for mkdir, symlink, remove, rmdir, link, and rename
+ * may have the following bit set, indicating the original request
+ * specified case-insensitive handling of names.
+ */
+#define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */
+
+/*
+ * Format of log records.
+ * The fields are carefully defined to allow them to be aligned
+ * and sized the same on sparc & intel architectures.
+ * Each log record has a common structure at the beginning.
+ *
+ * Note, lrc_seq holds two different sequence numbers. Whilst in memory
+ * it contains the transaction sequence number. The log record on
+ * disk holds the sequence number of all log records which is used to
+ * ensure we don't replay the same record. The two sequence numbers are
+ * different because the transactions can now be pushed out of order.
+ */
+typedef struct { /* common log record header */
+ uint64_t lrc_txtype; /* intent log transaction type */
+ uint64_t lrc_reclen; /* transaction record length */
+ uint64_t lrc_txg; /* dmu transaction group number */
+ uint64_t lrc_seq; /* see comment above */
+} lr_t;
+
+/*
+ * Handle option extended vattr attributes.
+ *
+ * Whenever new attributes are added the version number
+ * will need to be updated as will code in
+ * zfs_log.c and zfs_replay.c
+ */
+typedef struct {
+ uint32_t lr_attr_masksize; /* number of elements in array */
+ uint32_t lr_attr_bitmap; /* First entry of array */
+ /* remainder of array and any additional fields */
+} lr_attr_t;
+
+/*
+ * log record for creates without optional ACL.
+ * This log record does support optional xvattr_t attributes.
+ */
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_doid; /* object id of directory */
+ uint64_t lr_foid; /* object id of created file object */
+ uint64_t lr_mode; /* mode of object */
+ uint64_t lr_uid; /* uid of object */
+ uint64_t lr_gid; /* gid of object */
+ uint64_t lr_gen; /* generation (txg of creation) */
+ uint64_t lr_crtime[2]; /* creation time */
+ uint64_t lr_rdev; /* rdev of object to create */
+ /* name of object to create follows this */
+ /* for symlinks, link content follows name */
+ /* for creates with xvattr data, the name follows the xvattr info */
+} lr_create_t;
+
+/*
+ * FUID ACL record will be an array of ACEs from the original ACL.
+ * If this array includes ephemeral IDs, the record will also include
+ * an array of log-specific FUIDs to replace the ephemeral IDs.
+ * Only one copy of each unique domain will be present, so the log-specific
+ * FUIDs will use an index into a compressed domain table. On replay this
+ * information will be used to construct real FUIDs (and bypass idmap,
+ * since it may not be available).
+ */
+
+/*
+ * Log record for creates with optional ACL
+ * This log record is also used for recording any FUID
+ * information needed for replaying the create. If the
+ * file doesn't have any actual ACEs then the lr_aclcnt
+ * would be zero.
+ */
+typedef struct {
+ lr_create_t lr_create; /* common create portion */
+ uint64_t lr_aclcnt; /* number of ACEs in ACL */
+ uint64_t lr_domcnt; /* number of unique domains */
+ uint64_t lr_fuidcnt; /* number of real fuids */
+ uint64_t lr_acl_bytes; /* number of bytes in ACL */
+ uint64_t lr_acl_flags; /* ACL flags */
+ /* lr_acl_bytes number of variable sized ace's follows */
+ /* if create is also setting xvattr's, then acl data follows xvattr */
+ /* if ACE FUIDs are needed then they will follow the xvattr_t */
+ /* Following the FUIDs will be the domain table information. */
+ /* The FUIDs for the owner and group will be in the lr_create */
+ /* portion of the record. */
+ /* name follows ACL data */
+} lr_acl_create_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_doid; /* obj id of directory */
+ /* name of object to remove follows this */
+} lr_remove_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_doid; /* obj id of directory */
+ uint64_t lr_link_obj; /* obj id of link */
+ /* name of object to link follows this */
+} lr_link_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_sdoid; /* obj id of source directory */
+ uint64_t lr_tdoid; /* obj id of target directory */
+ /* 2 strings: names of source and destination follow this */
+} lr_rename_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_foid; /* file object to write */
+ uint64_t lr_offset; /* offset to write to */
+ uint64_t lr_length; /* user data length to write */
+ uint64_t lr_blkoff; /* offset represented by lr_blkptr */
+ blkptr_t lr_blkptr; /* spa block pointer for replay */
+ /* write data will follow for small writes */
+} lr_write_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_foid; /* object id of file to truncate */
+ uint64_t lr_offset; /* offset to truncate from */
+ uint64_t lr_length; /* length to truncate */
+} lr_truncate_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_foid; /* file object to change attributes */
+ uint64_t lr_mask; /* mask of attributes to set */
+ uint64_t lr_mode; /* mode to set */
+ uint64_t lr_uid; /* uid to set */
+ uint64_t lr_gid; /* gid to set */
+ uint64_t lr_size; /* size to set */
+ uint64_t lr_atime[2]; /* access time */
+ uint64_t lr_mtime[2]; /* modification time */
+ /* optional attribute lr_attr_t may be here */
+} lr_setattr_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_foid; /* obj id of file */
+ uint64_t lr_aclcnt; /* number of acl entries */
+ /* lr_aclcnt number of ace_t entries follow this */
+} lr_acl_v0_t;
+
+typedef struct {
+ lr_t lr_common; /* common portion of log record */
+ uint64_t lr_foid; /* obj id of file */
+ uint64_t lr_aclcnt; /* number of ACEs in ACL */
+ uint64_t lr_domcnt; /* number of unique domains */
+ uint64_t lr_fuidcnt; /* number of real fuids */
+ uint64_t lr_acl_bytes; /* number of bytes in ACL */
+ uint64_t lr_acl_flags; /* ACL flags */
+ /* lr_acl_bytes number of variable sized ace's follows */
+} lr_acl_t;
+
+/*
+ * ZIL structure definitions, interface function prototype and globals.
+ */
+
+/*
+ * ZFS intent log transaction structure
+ */
+typedef enum {
+ WR_INDIRECT, /* indirect - a large write (dmu_sync() data */
+ /* and put blkptr in log, rather than actual data) */
+ WR_COPIED, /* immediate - data is copied into lr_write_t */
+ WR_NEED_COPY, /* immediate - data needs to be copied if pushed */
+} itx_wr_state_t;
+
+typedef struct itx {
+ list_node_t itx_node; /* linkage on zl_itx_list */
+ void *itx_private; /* type-specific opaque data */
+ itx_wr_state_t itx_wr_state; /* write state */
+ uint8_t itx_sync; /* synchronous transaction */
+ uint64_t itx_sod; /* record size on disk */
+ lr_t itx_lr; /* common part of log record */
+ /* followed by type-specific part of lr_xx_t and its immediate data */
+} itx_t;
+
+
+/*
+ * zgd_t is passed through dmu_sync() to the callback routine zfs_get_done()
+ * to handle the cleanup of the dmu_sync() buffer write
+ */
+typedef struct {
+ zilog_t *zgd_zilog; /* zilog */
+ blkptr_t *zgd_bp; /* block pointer */
+ struct rl *zgd_rl; /* range lock */
+} zgd_t;
+
+
+typedef void zil_parse_blk_func_t(zilog_t *zilog, blkptr_t *bp, void *arg,
+ uint64_t txg);
+typedef void zil_parse_lr_func_t(zilog_t *zilog, lr_t *lr, void *arg,
+ uint64_t txg);
+typedef int zil_replay_func_t();
+typedef void zil_replay_cleaner_t();
+typedef int zil_get_data_t(void *arg, lr_write_t *lr, char *dbuf, zio_t *zio);
+
+extern uint64_t zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func,
+ zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg);
+
+extern void zil_init(void);
+extern void zil_fini(void);
+
+extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys);
+extern void zil_free(zilog_t *zilog);
+
+extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data);
+extern void zil_close(zilog_t *zilog);
+
+extern void zil_replay(objset_t *os, void *arg, uint64_t *txgp,
+ zil_replay_func_t *replay_func[TX_MAX_TYPE],
+ zil_replay_cleaner_t *replay_cleaner);
+extern void zil_destroy(zilog_t *zilog, boolean_t keep_first);
+extern void zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx);
+
+extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize);
+extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx);
+
+extern void zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid);
+
+extern int zil_claim(char *osname, void *txarg);
+extern int zil_check_log_chain(char *osname, void *txarg);
+extern int zil_clear_log_chain(char *osname, void *txarg);
+extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx);
+extern void zil_clean(zilog_t *zilog);
+extern int zil_is_committed(zilog_t *zilog);
+
+extern int zil_suspend(zilog_t *zilog);
+extern void zil_resume(zilog_t *zilog);
+
+extern void zil_add_block(zilog_t *zilog, blkptr_t *bp);
+
+extern int zil_disable;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZIL_H */
diff --git a/module/zfs/include/sys/zil_impl.h b/module/zfs/include/sys/zil_impl.h
new file mode 100644
index 000000000..0fc800b96
--- /dev/null
+++ b/module/zfs/include/sys/zil_impl.h
@@ -0,0 +1,109 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZIL_IMPL_H
+#define _SYS_ZIL_IMPL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zil.h>
+#include <sys/dmu_objset.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Log write buffer.
+ */
+typedef struct lwb {
+ zilog_t *lwb_zilog; /* back pointer to log struct */
+ blkptr_t lwb_blk; /* on disk address of this log blk */
+ int lwb_nused; /* # used bytes in buffer */
+ int lwb_sz; /* size of block and buffer */
+ char *lwb_buf; /* log write buffer */
+ zio_t *lwb_zio; /* zio for this buffer */
+ uint64_t lwb_max_txg; /* highest txg in this lwb */
+ txg_handle_t lwb_txgh; /* txg handle for txg_exit() */
+ list_node_t lwb_node; /* zilog->zl_lwb_list linkage */
+} lwb_t;
+
+/*
+ * Vdev flushing: during a zil_commit(), we build up an AVL tree of the vdevs
+ * we've touched so we know which ones need a write cache flush at the end.
+ */
+typedef struct zil_vdev_node {
+ uint64_t zv_vdev; /* vdev to be flushed */
+ avl_node_t zv_node; /* AVL tree linkage */
+} zil_vdev_node_t;
+
+/*
+ * Stable storage intent log management structure. One per dataset.
+ */
+struct zilog {
+ kmutex_t zl_lock; /* protects most zilog_t fields */
+ struct dsl_pool *zl_dmu_pool; /* DSL pool */
+ spa_t *zl_spa; /* handle for read/write log */
+ const zil_header_t *zl_header; /* log header buffer */
+ objset_t *zl_os; /* object set we're logging */
+ zil_get_data_t *zl_get_data; /* callback to get object content */
+ zio_t *zl_root_zio; /* log writer root zio */
+ uint64_t zl_itx_seq; /* next itx sequence number */
+ uint64_t zl_commit_seq; /* committed upto this number */
+ uint64_t zl_lr_seq; /* log record sequence number */
+ uint64_t zl_destroy_txg; /* txg of last zil_destroy() */
+ uint64_t zl_replay_seq[TXG_SIZE]; /* seq of last replayed rec */
+ uint32_t zl_suspend; /* log suspend count */
+ kcondvar_t zl_cv_writer; /* log writer thread completion */
+ kcondvar_t zl_cv_suspend; /* log suspend completion */
+ uint8_t zl_suspending; /* log is currently suspending */
+ uint8_t zl_keep_first; /* keep first log block in destroy */
+ uint8_t zl_stop_replay; /* don't replay any further */
+ uint8_t zl_stop_sync; /* for debugging */
+ uint8_t zl_writer; /* boolean: write setup in progress */
+ uint8_t zl_log_error; /* boolean: log write error */
+ list_t zl_itx_list; /* in-memory itx list */
+ uint64_t zl_itx_list_sz; /* total size of records on list */
+ uint64_t zl_cur_used; /* current commit log size used */
+ uint64_t zl_prev_used; /* previous commit log size used */
+ list_t zl_lwb_list; /* in-flight log write list */
+ kmutex_t zl_vdev_lock; /* protects zl_vdev_tree */
+ avl_tree_t zl_vdev_tree; /* vdevs to flush in zil_commit() */
+ taskq_t *zl_clean_taskq; /* runs lwb and itx clean tasks */
+ avl_tree_t zl_dva_tree; /* track DVAs during log parse */
+ clock_t zl_replay_time; /* lbolt of when replay started */
+ uint64_t zl_replay_blks; /* number of log blocks replayed */
+};
+
+typedef struct zil_dva_node {
+ dva_t zn_dva;
+ avl_node_t zn_node;
+} zil_dva_node_t;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZIL_IMPL_H */
diff --git a/module/zfs/include/sys/zio.h b/module/zfs/include/sys/zio.h
new file mode 100644
index 000000000..4de78dfee
--- /dev/null
+++ b/module/zfs/include/sys/zio.h
@@ -0,0 +1,424 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _ZIO_H
+#define _ZIO_H
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/avl.h>
+#include <sys/fs/zfs.h>
+#include <sys/zio_impl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZBT_MAGIC 0x210da7ab10c7a11ULL /* zio data bloc tail */
+
+typedef struct zio_block_tail {
+ uint64_t zbt_magic; /* for validation, endianness */
+ zio_cksum_t zbt_cksum; /* 256-bit checksum */
+} zio_block_tail_t;
+
+/*
+ * Gang block headers are self-checksumming and contain an array
+ * of block pointers.
+ */
+#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
+#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
+ sizeof (zio_block_tail_t)) / sizeof (blkptr_t))
+#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
+ sizeof (zio_block_tail_t) - \
+ (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
+ sizeof (uint64_t))
+
+typedef struct zio_gbh {
+ blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
+ uint64_t zg_filler[SPA_GBH_FILLER];
+ zio_block_tail_t zg_tail;
+} zio_gbh_phys_t;
+
+enum zio_checksum {
+ ZIO_CHECKSUM_INHERIT = 0,
+ ZIO_CHECKSUM_ON,
+ ZIO_CHECKSUM_OFF,
+ ZIO_CHECKSUM_LABEL,
+ ZIO_CHECKSUM_GANG_HEADER,
+ ZIO_CHECKSUM_ZILOG,
+ ZIO_CHECKSUM_FLETCHER_2,
+ ZIO_CHECKSUM_FLETCHER_4,
+ ZIO_CHECKSUM_SHA256,
+ ZIO_CHECKSUM_FUNCTIONS
+};
+
+#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_2
+#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
+
+enum zio_compress {
+ ZIO_COMPRESS_INHERIT = 0,
+ ZIO_COMPRESS_ON,
+ ZIO_COMPRESS_OFF,
+ ZIO_COMPRESS_LZJB,
+ ZIO_COMPRESS_EMPTY,
+ ZIO_COMPRESS_GZIP_1,
+ ZIO_COMPRESS_GZIP_2,
+ ZIO_COMPRESS_GZIP_3,
+ ZIO_COMPRESS_GZIP_4,
+ ZIO_COMPRESS_GZIP_5,
+ ZIO_COMPRESS_GZIP_6,
+ ZIO_COMPRESS_GZIP_7,
+ ZIO_COMPRESS_GZIP_8,
+ ZIO_COMPRESS_GZIP_9,
+ ZIO_COMPRESS_FUNCTIONS
+};
+
+#define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
+#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
+
+#define ZIO_FAILURE_MODE_WAIT 0
+#define ZIO_FAILURE_MODE_CONTINUE 1
+#define ZIO_FAILURE_MODE_PANIC 2
+
+#define ZIO_PRIORITY_NOW (zio_priority_table[0])
+#define ZIO_PRIORITY_SYNC_READ (zio_priority_table[1])
+#define ZIO_PRIORITY_SYNC_WRITE (zio_priority_table[2])
+#define ZIO_PRIORITY_ASYNC_READ (zio_priority_table[3])
+#define ZIO_PRIORITY_ASYNC_WRITE (zio_priority_table[4])
+#define ZIO_PRIORITY_FREE (zio_priority_table[5])
+#define ZIO_PRIORITY_CACHE_FILL (zio_priority_table[6])
+#define ZIO_PRIORITY_LOG_WRITE (zio_priority_table[7])
+#define ZIO_PRIORITY_RESILVER (zio_priority_table[8])
+#define ZIO_PRIORITY_SCRUB (zio_priority_table[9])
+#define ZIO_PRIORITY_TABLE_SIZE 10
+
+#define ZIO_FLAG_MUSTSUCCEED 0x00000
+#define ZIO_FLAG_CANFAIL 0x00001
+#define ZIO_FLAG_SPECULATIVE 0x00002
+#define ZIO_FLAG_CONFIG_WRITER 0x00004
+#define ZIO_FLAG_DONT_RETRY 0x00008
+
+#define ZIO_FLAG_DONT_CACHE 0x00010
+#define ZIO_FLAG_DONT_QUEUE 0x00020
+#define ZIO_FLAG_DONT_AGGREGATE 0x00040
+#define ZIO_FLAG_DONT_PROPAGATE 0x00080
+
+#define ZIO_FLAG_IO_BYPASS 0x00100
+#define ZIO_FLAG_IO_REPAIR 0x00200
+#define ZIO_FLAG_IO_RETRY 0x00400
+#define ZIO_FLAG_IO_REWRITE 0x00800
+
+#define ZIO_FLAG_PROBE 0x01000
+#define ZIO_FLAG_RESILVER 0x02000
+#define ZIO_FLAG_SCRUB 0x04000
+#define ZIO_FLAG_SCRUB_THREAD 0x08000
+
+#define ZIO_FLAG_GANG_CHILD 0x10000
+
+#define ZIO_FLAG_GANG_INHERIT \
+ (ZIO_FLAG_CANFAIL | \
+ ZIO_FLAG_SPECULATIVE | \
+ ZIO_FLAG_CONFIG_WRITER | \
+ ZIO_FLAG_DONT_RETRY | \
+ ZIO_FLAG_DONT_CACHE | \
+ ZIO_FLAG_DONT_AGGREGATE | \
+ ZIO_FLAG_RESILVER | \
+ ZIO_FLAG_SCRUB | \
+ ZIO_FLAG_SCRUB_THREAD)
+
+#define ZIO_FLAG_VDEV_INHERIT \
+ (ZIO_FLAG_GANG_INHERIT | \
+ ZIO_FLAG_IO_REPAIR | \
+ ZIO_FLAG_IO_RETRY | \
+ ZIO_FLAG_PROBE)
+
+#define ZIO_PIPELINE_CONTINUE 0x100
+#define ZIO_PIPELINE_STOP 0x101
+
+#define ZIO_GANG_CHILD_FLAGS(zio) \
+ (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
+ ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
+
+enum zio_child {
+ ZIO_CHILD_VDEV = 0,
+ ZIO_CHILD_GANG,
+ ZIO_CHILD_LOGICAL,
+ ZIO_CHILD_TYPES
+};
+
+enum zio_wait_type {
+ ZIO_WAIT_READY = 0,
+ ZIO_WAIT_DONE,
+ ZIO_WAIT_TYPES
+};
+
+/*
+ * We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
+ * graveyard) to indicate checksum errors and fragmentation.
+ */
+#define ECKSUM EBADE
+#define EFRAGS EBADR
+
+typedef struct zio zio_t;
+typedef void zio_done_func_t(zio_t *zio);
+
+extern uint8_t zio_priority_table[ZIO_PRIORITY_TABLE_SIZE];
+extern char *zio_type_name[ZIO_TYPES];
+
+/*
+ * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
+ * identifies any block in the pool. By convention, the meta-objset (MOS)
+ * is objset 0, the meta-dnode is object 0, the root block (osphys_t) is
+ * level -1 of the meta-dnode, and intent log blocks (which are chained
+ * off the root block) have blkid == sequence number. In summary:
+ *
+ * mos is objset 0
+ * meta-dnode is object 0
+ * root block is <objset, 0, -1, 0>
+ * intent log is <objset, 0, -1, ZIL sequence number>
+ *
+ * Note: this structure is called a bookmark because its first purpose was
+ * to remember where to resume a pool-wide traverse. The absolute ordering
+ * for block visitation during traversal is defined in compare_bookmark().
+ *
+ * Note: this structure is passed between userland and the kernel.
+ * Therefore it must not change size or alignment between 32/64 bit
+ * compilation options.
+ */
+typedef struct zbookmark {
+ uint64_t zb_objset;
+ uint64_t zb_object;
+ int64_t zb_level;
+ uint64_t zb_blkid;
+} zbookmark_t;
+
+typedef struct zio_prop {
+ enum zio_checksum zp_checksum;
+ enum zio_compress zp_compress;
+ dmu_object_type_t zp_type;
+ uint8_t zp_level;
+ uint8_t zp_ndvas;
+} zio_prop_t;
+
+typedef struct zio_gang_node {
+ zio_gbh_phys_t *gn_gbh;
+ struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
+} zio_gang_node_t;
+
+typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
+ zio_gang_node_t *gn, void *data);
+
+typedef void zio_transform_func_t(zio_t *zio, void *data, uint64_t size);
+
+typedef struct zio_transform {
+ void *zt_orig_data;
+ uint64_t zt_orig_size;
+ uint64_t zt_bufsize;
+ zio_transform_func_t *zt_transform;
+ struct zio_transform *zt_next;
+} zio_transform_t;
+
+typedef int zio_pipe_stage_t(zio_t *zio);
+
+/*
+ * The io_reexecute flags are distinct from io_flags because the child must
+ * be able to propagate them to the parent. The normal io_flags are local
+ * to the zio, not protected by any lock, and not modifiable by children;
+ * the reexecute flags are protected by io_lock, modifiable by children,
+ * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
+ */
+#define ZIO_REEXECUTE_NOW 0x01
+#define ZIO_REEXECUTE_SUSPEND 0x02
+
+struct zio {
+ /* Core information about this I/O */
+ zbookmark_t io_bookmark;
+ zio_prop_t io_prop;
+ zio_type_t io_type;
+ enum zio_child io_child_type;
+ int io_cmd;
+ uint8_t io_priority;
+ uint8_t io_reexecute;
+ uint8_t io_async_root;
+ uint64_t io_txg;
+ spa_t *io_spa;
+ blkptr_t *io_bp;
+ blkptr_t io_bp_copy;
+ zio_t *io_parent;
+ zio_t *io_child;
+ zio_t *io_sibling_prev;
+ zio_t *io_sibling_next;
+ zio_t *io_logical;
+ zio_transform_t *io_transform_stack;
+
+ /* Callback info */
+ zio_done_func_t *io_ready;
+ zio_done_func_t *io_done;
+ void *io_private;
+ blkptr_t io_bp_orig;
+
+ /* Data represented by this I/O */
+ void *io_data;
+ uint64_t io_size;
+
+ /* Stuff for the vdev stack */
+ vdev_t *io_vd;
+ void *io_vsd;
+ zio_done_func_t *io_vsd_free;
+ uint64_t io_offset;
+ uint64_t io_deadline;
+ avl_node_t io_offset_node;
+ avl_node_t io_deadline_node;
+ avl_tree_t *io_vdev_tree;
+ zio_t *io_delegate_list;
+ zio_t *io_delegate_next;
+
+ /* Internal pipeline state */
+ int io_flags;
+ zio_stage_t io_stage;
+ uint32_t io_pipeline;
+ int io_orig_flags;
+ zio_stage_t io_orig_stage;
+ uint32_t io_orig_pipeline;
+ int io_error;
+ int io_child_error[ZIO_CHILD_TYPES];
+ uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
+ uint64_t *io_stall;
+ zio_gang_node_t *io_gang_tree;
+ void *io_executor;
+ void *io_waiter;
+ kmutex_t io_lock;
+ kcondvar_t io_cv;
+
+ /* FMA state */
+ uint64_t io_ena;
+};
+
+extern zio_t *zio_null(zio_t *pio, spa_t *spa,
+ zio_done_func_t *done, void *private, int flags);
+
+extern zio_t *zio_root(spa_t *spa,
+ zio_done_func_t *done, void *private, int flags);
+
+extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, void *data,
+ uint64_t size, zio_done_func_t *done, void *private,
+ int priority, int flags, const zbookmark_t *zb);
+
+extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ void *data, uint64_t size, zio_prop_t *zp,
+ zio_done_func_t *ready, zio_done_func_t *done, void *private,
+ int priority, int flags, const zbookmark_t *zb);
+
+extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ void *data, uint64_t size, zio_done_func_t *done, void *private,
+ int priority, int flags, zbookmark_t *zb);
+
+extern void zio_skip_write(zio_t *zio);
+
+extern zio_t *zio_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, int flags);
+
+extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, int flags);
+
+extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
+ zio_done_func_t *done, void *private, int priority, int flags);
+
+extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
+ uint64_t size, void *data, int checksum,
+ zio_done_func_t *done, void *private, int priority, int flags,
+ boolean_t labels);
+
+extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
+ uint64_t size, void *data, int checksum,
+ zio_done_func_t *done, void *private, int priority, int flags,
+ boolean_t labels);
+
+extern int zio_alloc_blk(spa_t *spa, uint64_t size, blkptr_t *new_bp,
+ blkptr_t *old_bp, uint64_t txg);
+extern void zio_free_blk(spa_t *spa, blkptr_t *bp, uint64_t txg);
+extern void zio_flush(zio_t *zio, vdev_t *vd);
+
+extern int zio_wait(zio_t *zio);
+extern void zio_nowait(zio_t *zio);
+extern void zio_execute(zio_t *zio);
+extern void zio_interrupt(zio_t *zio);
+
+extern void *zio_buf_alloc(size_t size);
+extern void zio_buf_free(void *buf, size_t size);
+extern void *zio_data_buf_alloc(size_t size);
+extern void zio_data_buf_free(void *buf, size_t size);
+
+extern void zio_resubmit_stage_async(void *);
+
+extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
+ uint64_t offset, void *data, uint64_t size, int type, int priority,
+ int flags, zio_done_func_t *done, void *private);
+
+extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
+ void *data, uint64_t size, int type, int priority,
+ int flags, zio_done_func_t *done, void *private);
+
+extern void zio_vdev_io_bypass(zio_t *zio);
+extern void zio_vdev_io_reissue(zio_t *zio);
+extern void zio_vdev_io_redone(zio_t *zio);
+
+extern void zio_checksum_verified(zio_t *zio);
+extern int zio_worst_error(int e1, int e2);
+
+extern uint8_t zio_checksum_select(uint8_t child, uint8_t parent);
+extern uint8_t zio_compress_select(uint8_t child, uint8_t parent);
+
+extern void zio_suspend(spa_t *spa, zio_t *zio);
+extern void zio_resume(spa_t *spa);
+extern void zio_resume_wait(spa_t *spa);
+
+/*
+ * Initial setup and teardown.
+ */
+extern void zio_init(void);
+extern void zio_fini(void);
+
+/*
+ * Fault injection
+ */
+struct zinject_record;
+extern uint32_t zio_injection_enabled;
+extern int zio_inject_fault(char *name, int flags, int *id,
+ struct zinject_record *record);
+extern int zio_inject_list_next(int *id, char *name, size_t buflen,
+ struct zinject_record *record);
+extern int zio_clear_fault(int id);
+extern int zio_handle_fault_injection(zio_t *zio, int error);
+extern int zio_handle_device_injection(vdev_t *vd, int error);
+extern int zio_handle_label_injection(zio_t *zio, int error);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _ZIO_H */
diff --git a/module/zfs/include/sys/zio_checksum.h b/module/zfs/include/sys/zio_checksum.h
new file mode 100644
index 000000000..da407399d
--- /dev/null
+++ b/module/zfs/include/sys/zio_checksum.h
@@ -0,0 +1,73 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZIO_CHECKSUM_H
+#define _SYS_ZIO_CHECKSUM_H
+
+#include <sys/zio.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Signature for checksum functions.
+ */
+typedef void zio_checksum_t(const void *data, uint64_t size, zio_cksum_t *zcp);
+
+/*
+ * Information about each checksum function.
+ */
+typedef struct zio_checksum_info {
+ zio_checksum_t *ci_func[2]; /* checksum function for each byteorder */
+ int ci_correctable; /* number of correctable bits */
+ int ci_zbt; /* uses zio block tail? */
+ char *ci_name; /* descriptive name */
+} zio_checksum_info_t;
+
+extern zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS];
+
+/*
+ * Checksum routines.
+ */
+extern zio_checksum_t fletcher_2_native;
+extern zio_checksum_t fletcher_4_native;
+extern zio_checksum_t fletcher_4_incremental_native;
+
+extern zio_checksum_t fletcher_2_byteswap;
+extern zio_checksum_t fletcher_4_byteswap;
+extern zio_checksum_t fletcher_4_incremental_byteswap;
+
+extern zio_checksum_t zio_checksum_SHA256;
+
+extern void zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
+ void *data, uint64_t size);
+extern int zio_checksum_error(zio_t *zio);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZIO_CHECKSUM_H */
diff --git a/module/zfs/include/sys/zio_compress.h b/module/zfs/include/sys/zio_compress.h
new file mode 100644
index 000000000..66ee8d45b
--- /dev/null
+++ b/module/zfs/include/sys/zio_compress.h
@@ -0,0 +1,82 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZIO_COMPRESS_H
+#define _SYS_ZIO_COMPRESS_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zio.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Common signature for all zio compress/decompress functions.
+ */
+typedef size_t zio_compress_func_t(void *src, void *dst,
+ size_t s_len, size_t d_len, int);
+typedef int zio_decompress_func_t(void *src, void *dst,
+ size_t s_len, size_t d_len, int);
+
+/*
+ * Information about each compression function.
+ */
+typedef struct zio_compress_info {
+ zio_compress_func_t *ci_compress; /* compression function */
+ zio_decompress_func_t *ci_decompress; /* decompression function */
+ int ci_level; /* level parameter */
+ char *ci_name; /* algorithm name */
+} zio_compress_info_t;
+
+extern zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS];
+
+/*
+ * Compression routines.
+ */
+extern size_t lzjb_compress(void *src, void *dst, size_t s_len, size_t d_len,
+ int level);
+extern int lzjb_decompress(void *src, void *dst, size_t s_len, size_t d_len,
+ int level);
+extern size_t gzip_compress(void *src, void *dst, size_t s_len, size_t d_len,
+ int level);
+extern int gzip_decompress(void *src, void *dst, size_t s_len, size_t d_len,
+ int level);
+
+/*
+ * Compress and decompress data if necessary.
+ */
+extern int zio_compress_data(int cpfunc, void *src, uint64_t srcsize,
+ void **destp, uint64_t *destsizep, uint64_t *destbufsizep);
+extern int zio_decompress_data(int cpfunc, void *src, uint64_t srcsize,
+ void *dest, uint64_t destsize);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZIO_COMPRESS_H */
diff --git a/module/zfs/include/sys/zio_impl.h b/module/zfs/include/sys/zio_impl.h
new file mode 100644
index 000000000..e7503b733
--- /dev/null
+++ b/module/zfs/include/sys/zio_impl.h
@@ -0,0 +1,143 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _ZIO_IMPL_H
+#define _ZIO_IMPL_H
+
+#include <sys/zfs_context.h>
+#include <sys/zio.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * I/O Groups: pipeline stage definitions.
+ */
+typedef enum zio_stage {
+ ZIO_STAGE_OPEN = 0, /* RWFCI */
+
+ ZIO_STAGE_ISSUE_ASYNC, /* -W--- */
+
+ ZIO_STAGE_READ_BP_INIT, /* R---- */
+ ZIO_STAGE_WRITE_BP_INIT, /* -W--- */
+
+ ZIO_STAGE_CHECKSUM_GENERATE, /* -W--- */
+
+ ZIO_STAGE_GANG_ASSEMBLE, /* RWFC- */
+ ZIO_STAGE_GANG_ISSUE, /* RWFC- */
+
+ ZIO_STAGE_DVA_ALLOCATE, /* -W--- */
+ ZIO_STAGE_DVA_FREE, /* --F-- */
+ ZIO_STAGE_DVA_CLAIM, /* ---C- */
+
+ ZIO_STAGE_READY, /* RWFCI */
+
+ ZIO_STAGE_VDEV_IO_START, /* RW--I */
+ ZIO_STAGE_VDEV_IO_DONE, /* RW--I */
+ ZIO_STAGE_VDEV_IO_ASSESS, /* RW--I */
+
+ ZIO_STAGE_CHECKSUM_VERIFY, /* R---- */
+
+ ZIO_STAGE_DONE, /* RWFCI */
+ ZIO_STAGES
+} zio_stage_t;
+
+#define ZIO_INTERLOCK_STAGES \
+ ((1U << ZIO_STAGE_READY) | \
+ (1U << ZIO_STAGE_DONE))
+
+#define ZIO_INTERLOCK_PIPELINE \
+ ZIO_INTERLOCK_STAGES
+
+#define ZIO_VDEV_IO_STAGES \
+ ((1U << ZIO_STAGE_VDEV_IO_START) | \
+ (1U << ZIO_STAGE_VDEV_IO_DONE) | \
+ (1U << ZIO_STAGE_VDEV_IO_ASSESS))
+
+#define ZIO_VDEV_CHILD_PIPELINE \
+ (ZIO_VDEV_IO_STAGES | \
+ (1U << ZIO_STAGE_DONE))
+
+#define ZIO_READ_COMMON_STAGES \
+ (ZIO_INTERLOCK_STAGES | \
+ ZIO_VDEV_IO_STAGES | \
+ (1U << ZIO_STAGE_CHECKSUM_VERIFY))
+
+#define ZIO_READ_PHYS_PIPELINE \
+ ZIO_READ_COMMON_STAGES
+
+#define ZIO_READ_PIPELINE \
+ (ZIO_READ_COMMON_STAGES | \
+ (1U << ZIO_STAGE_READ_BP_INIT))
+
+#define ZIO_WRITE_COMMON_STAGES \
+ (ZIO_INTERLOCK_STAGES | \
+ ZIO_VDEV_IO_STAGES | \
+ (1U << ZIO_STAGE_ISSUE_ASYNC) | \
+ (1U << ZIO_STAGE_CHECKSUM_GENERATE))
+
+#define ZIO_WRITE_PHYS_PIPELINE \
+ ZIO_WRITE_COMMON_STAGES
+
+#define ZIO_REWRITE_PIPELINE \
+ (ZIO_WRITE_COMMON_STAGES | \
+ (1U << ZIO_STAGE_WRITE_BP_INIT))
+
+#define ZIO_WRITE_PIPELINE \
+ (ZIO_WRITE_COMMON_STAGES | \
+ (1U << ZIO_STAGE_WRITE_BP_INIT) | \
+ (1U << ZIO_STAGE_DVA_ALLOCATE))
+
+#define ZIO_GANG_STAGES \
+ ((1U << ZIO_STAGE_GANG_ASSEMBLE) | \
+ (1U << ZIO_STAGE_GANG_ISSUE))
+
+#define ZIO_FREE_PIPELINE \
+ (ZIO_INTERLOCK_STAGES | \
+ (1U << ZIO_STAGE_DVA_FREE))
+
+#define ZIO_CLAIM_PIPELINE \
+ (ZIO_INTERLOCK_STAGES | \
+ (1U << ZIO_STAGE_DVA_CLAIM))
+
+#define ZIO_IOCTL_PIPELINE \
+ (ZIO_INTERLOCK_STAGES | \
+ (1U << ZIO_STAGE_VDEV_IO_START) | \
+ (1U << ZIO_STAGE_VDEV_IO_ASSESS))
+
+#define ZIO_CONFIG_LOCK_BLOCKING_STAGES \
+ ((1U << ZIO_STAGE_VDEV_IO_START) | \
+ (1U << ZIO_STAGE_DVA_ALLOCATE) | \
+ (1U << ZIO_STAGE_DVA_CLAIM))
+
+extern void zio_inject_init(void);
+extern void zio_inject_fini(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _ZIO_IMPL_H */
diff --git a/module/zfs/include/sys/zvol.h b/module/zfs/include/sys/zvol.h
new file mode 100644
index 000000000..06adc667e
--- /dev/null
+++ b/module/zfs/include/sys/zvol.h
@@ -0,0 +1,70 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#ifndef _SYS_ZVOL_H
+#define _SYS_ZVOL_H
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZVOL_OBJ 1ULL
+#define ZVOL_ZAP_OBJ 2ULL
+
+#ifdef _KERNEL
+extern int zvol_check_volsize(uint64_t volsize, uint64_t blocksize);
+extern int zvol_check_volblocksize(uint64_t volblocksize);
+extern int zvol_get_stats(objset_t *os, nvlist_t *nv);
+extern void zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
+extern int zvol_create_minor(const char *, major_t);
+extern int zvol_remove_minor(const char *);
+extern int zvol_set_volsize(const char *, major_t, uint64_t);
+extern int zvol_set_volblocksize(const char *, uint64_t);
+
+extern int zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr);
+extern int zvol_dump(dev_t dev, caddr_t addr, daddr_t offset, int nblocks);
+extern int zvol_close(dev_t dev, int flag, int otyp, cred_t *cr);
+extern int zvol_strategy(buf_t *bp);
+extern int zvol_read(dev_t dev, uio_t *uiop, cred_t *cr);
+extern int zvol_write(dev_t dev, uio_t *uiop, cred_t *cr);
+extern int zvol_aread(dev_t dev, struct aio_req *aio, cred_t *cr);
+extern int zvol_awrite(dev_t dev, struct aio_req *aio, cred_t *cr);
+extern int zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr,
+ int *rvalp);
+extern int zvol_busy(void);
+extern void zvol_init(void);
+extern void zvol_fini(void);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _SYS_ZVOL_H */
diff --git a/module/zfs/lzjb.c b/module/zfs/lzjb.c
new file mode 100644
index 000000000..7fcde8475
--- /dev/null
+++ b/module/zfs/lzjb.c
@@ -0,0 +1,128 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * We keep our own copy of this algorithm for 2 main reasons:
+ * 1. If we didn't, anyone modifying common/os/compress.c would
+ * directly break our on disk format
+ * 2. Our version of lzjb does not have a number of checks that the
+ * common/os version needs and uses
+ * In particular, we are adding the "feature" that compress() can
+ * take a destination buffer size and return -1 if the data will not
+ * compress to d_len or less.
+ */
+
+#include <sys/types.h>
+
+#define MATCH_BITS 6
+#define MATCH_MIN 3
+#define MATCH_MAX ((1 << MATCH_BITS) + (MATCH_MIN - 1))
+#define OFFSET_MASK ((1 << (16 - MATCH_BITS)) - 1)
+#define LEMPEL_SIZE 256
+
+/*ARGSUSED*/
+size_t
+lzjb_compress(void *s_start, void *d_start, size_t s_len, size_t d_len, int n)
+{
+ uchar_t *src = s_start;
+ uchar_t *dst = d_start;
+ uchar_t *cpy, *copymap;
+ int copymask = 1 << (NBBY - 1);
+ int mlen, offset;
+ uint16_t *hp;
+ uint16_t lempel[LEMPEL_SIZE]; /* uninitialized; see above */
+
+ while (src < (uchar_t *)s_start + s_len) {
+ if ((copymask <<= 1) == (1 << NBBY)) {
+ if (dst >= (uchar_t *)d_start + d_len - 1 - 2 * NBBY) {
+ if (d_len != s_len)
+ return (s_len);
+ mlen = s_len;
+ for (src = s_start, dst = d_start; mlen; mlen--)
+ *dst++ = *src++;
+ return (s_len);
+ }
+ copymask = 1;
+ copymap = dst;
+ *dst++ = 0;
+ }
+ if (src > (uchar_t *)s_start + s_len - MATCH_MAX) {
+ *dst++ = *src++;
+ continue;
+ }
+ hp = &lempel[((src[0] + 13) ^ (src[1] - 13) ^ src[2]) &
+ (LEMPEL_SIZE - 1)];
+ offset = (intptr_t)(src - *hp) & OFFSET_MASK;
+ *hp = (uint16_t)(uintptr_t)src;
+ cpy = src - offset;
+ if (cpy >= (uchar_t *)s_start && cpy != src &&
+ src[0] == cpy[0] && src[1] == cpy[1] && src[2] == cpy[2]) {
+ *copymap |= copymask;
+ for (mlen = MATCH_MIN; mlen < MATCH_MAX; mlen++)
+ if (src[mlen] != cpy[mlen])
+ break;
+ *dst++ = ((mlen - MATCH_MIN) << (NBBY - MATCH_BITS)) |
+ (offset >> NBBY);
+ *dst++ = (uchar_t)offset;
+ src += mlen;
+ } else {
+ *dst++ = *src++;
+ }
+ }
+ return (dst - (uchar_t *)d_start);
+}
+
+/*ARGSUSED*/
+int
+lzjb_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len, int n)
+{
+ uchar_t *src = s_start;
+ uchar_t *dst = d_start;
+ uchar_t *d_end = (uchar_t *)d_start + d_len;
+ uchar_t *cpy, copymap;
+ int copymask = 1 << (NBBY - 1);
+
+ while (dst < d_end) {
+ if ((copymask <<= 1) == (1 << NBBY)) {
+ copymask = 1;
+ copymap = *src++;
+ }
+ if (copymap & copymask) {
+ int mlen = (src[0] >> (NBBY - MATCH_BITS)) + MATCH_MIN;
+ int offset = ((src[0] << NBBY) | src[1]) & OFFSET_MASK;
+ src += 2;
+ if ((cpy = dst - offset) < (uchar_t *)d_start)
+ return (-1);
+ while (--mlen >= 0 && dst < d_end)
+ *dst++ = *cpy++;
+ } else {
+ *dst++ = *src++;
+ }
+ }
+ return (0);
+}
diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
new file mode 100644
index 000000000..87727fac2
--- /dev/null
+++ b/module/zfs/metaslab.c
@@ -0,0 +1,1049 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa_impl.h>
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/space_map.h>
+#include <sys/metaslab_impl.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+
+uint64_t metaslab_aliquot = 512ULL << 10;
+uint64_t metaslab_gang_bang = SPA_MAXBLOCKSIZE + 1; /* force gang blocks */
+
+/*
+ * ==========================================================================
+ * Metaslab classes
+ * ==========================================================================
+ */
+metaslab_class_t *
+metaslab_class_create(void)
+{
+ metaslab_class_t *mc;
+
+ mc = kmem_zalloc(sizeof (metaslab_class_t), KM_SLEEP);
+
+ mc->mc_rotor = NULL;
+
+ return (mc);
+}
+
+void
+metaslab_class_destroy(metaslab_class_t *mc)
+{
+ metaslab_group_t *mg;
+
+ while ((mg = mc->mc_rotor) != NULL) {
+ metaslab_class_remove(mc, mg);
+ metaslab_group_destroy(mg);
+ }
+
+ kmem_free(mc, sizeof (metaslab_class_t));
+}
+
+void
+metaslab_class_add(metaslab_class_t *mc, metaslab_group_t *mg)
+{
+ metaslab_group_t *mgprev, *mgnext;
+
+ ASSERT(mg->mg_class == NULL);
+
+ if ((mgprev = mc->mc_rotor) == NULL) {
+ mg->mg_prev = mg;
+ mg->mg_next = mg;
+ } else {
+ mgnext = mgprev->mg_next;
+ mg->mg_prev = mgprev;
+ mg->mg_next = mgnext;
+ mgprev->mg_next = mg;
+ mgnext->mg_prev = mg;
+ }
+ mc->mc_rotor = mg;
+ mg->mg_class = mc;
+}
+
+void
+metaslab_class_remove(metaslab_class_t *mc, metaslab_group_t *mg)
+{
+ metaslab_group_t *mgprev, *mgnext;
+
+ ASSERT(mg->mg_class == mc);
+
+ mgprev = mg->mg_prev;
+ mgnext = mg->mg_next;
+
+ if (mg == mgnext) {
+ mc->mc_rotor = NULL;
+ } else {
+ mc->mc_rotor = mgnext;
+ mgprev->mg_next = mgnext;
+ mgnext->mg_prev = mgprev;
+ }
+
+ mg->mg_prev = NULL;
+ mg->mg_next = NULL;
+ mg->mg_class = NULL;
+}
+
+/*
+ * ==========================================================================
+ * Metaslab groups
+ * ==========================================================================
+ */
+static int
+metaslab_compare(const void *x1, const void *x2)
+{
+ const metaslab_t *m1 = x1;
+ const metaslab_t *m2 = x2;
+
+ if (m1->ms_weight < m2->ms_weight)
+ return (1);
+ if (m1->ms_weight > m2->ms_weight)
+ return (-1);
+
+ /*
+ * If the weights are identical, use the offset to force uniqueness.
+ */
+ if (m1->ms_map.sm_start < m2->ms_map.sm_start)
+ return (-1);
+ if (m1->ms_map.sm_start > m2->ms_map.sm_start)
+ return (1);
+
+ ASSERT3P(m1, ==, m2);
+
+ return (0);
+}
+
+metaslab_group_t *
+metaslab_group_create(metaslab_class_t *mc, vdev_t *vd)
+{
+ metaslab_group_t *mg;
+
+ mg = kmem_zalloc(sizeof (metaslab_group_t), KM_SLEEP);
+ mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
+ avl_create(&mg->mg_metaslab_tree, metaslab_compare,
+ sizeof (metaslab_t), offsetof(struct metaslab, ms_group_node));
+ mg->mg_aliquot = metaslab_aliquot * MAX(1, vd->vdev_children);
+ mg->mg_vd = vd;
+ metaslab_class_add(mc, mg);
+
+ return (mg);
+}
+
+void
+metaslab_group_destroy(metaslab_group_t *mg)
+{
+ avl_destroy(&mg->mg_metaslab_tree);
+ mutex_destroy(&mg->mg_lock);
+ kmem_free(mg, sizeof (metaslab_group_t));
+}
+
+static void
+metaslab_group_add(metaslab_group_t *mg, metaslab_t *msp)
+{
+ mutex_enter(&mg->mg_lock);
+ ASSERT(msp->ms_group == NULL);
+ msp->ms_group = mg;
+ msp->ms_weight = 0;
+ avl_add(&mg->mg_metaslab_tree, msp);
+ mutex_exit(&mg->mg_lock);
+}
+
+static void
+metaslab_group_remove(metaslab_group_t *mg, metaslab_t *msp)
+{
+ mutex_enter(&mg->mg_lock);
+ ASSERT(msp->ms_group == mg);
+ avl_remove(&mg->mg_metaslab_tree, msp);
+ msp->ms_group = NULL;
+ mutex_exit(&mg->mg_lock);
+}
+
+static void
+metaslab_group_sort(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
+{
+ /*
+ * Although in principle the weight can be any value, in
+ * practice we do not use values in the range [1, 510].
+ */
+ ASSERT(weight >= SPA_MINBLOCKSIZE-1 || weight == 0);
+ ASSERT(MUTEX_HELD(&msp->ms_lock));
+
+ mutex_enter(&mg->mg_lock);
+ ASSERT(msp->ms_group == mg);
+ avl_remove(&mg->mg_metaslab_tree, msp);
+ msp->ms_weight = weight;
+ avl_add(&mg->mg_metaslab_tree, msp);
+ mutex_exit(&mg->mg_lock);
+}
+
+/*
+ * ==========================================================================
+ * The first-fit block allocator
+ * ==========================================================================
+ */
+static void
+metaslab_ff_load(space_map_t *sm)
+{
+ ASSERT(sm->sm_ppd == NULL);
+ sm->sm_ppd = kmem_zalloc(64 * sizeof (uint64_t), KM_SLEEP);
+}
+
+static void
+metaslab_ff_unload(space_map_t *sm)
+{
+ kmem_free(sm->sm_ppd, 64 * sizeof (uint64_t));
+ sm->sm_ppd = NULL;
+}
+
+static uint64_t
+metaslab_ff_alloc(space_map_t *sm, uint64_t size)
+{
+ avl_tree_t *t = &sm->sm_root;
+ uint64_t align = size & -size;
+ uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;
+ space_seg_t *ss, ssearch;
+ avl_index_t where;
+
+ ssearch.ss_start = *cursor;
+ ssearch.ss_end = *cursor + size;
+
+ ss = avl_find(t, &ssearch, &where);
+ if (ss == NULL)
+ ss = avl_nearest(t, where, AVL_AFTER);
+
+ while (ss != NULL) {
+ uint64_t offset = P2ROUNDUP(ss->ss_start, align);
+
+ if (offset + size <= ss->ss_end) {
+ *cursor = offset + size;
+ return (offset);
+ }
+ ss = AVL_NEXT(t, ss);
+ }
+
+ /*
+ * If we know we've searched the whole map (*cursor == 0), give up.
+ * Otherwise, reset the cursor to the beginning and try again.
+ */
+ if (*cursor == 0)
+ return (-1ULL);
+
+ *cursor = 0;
+ return (metaslab_ff_alloc(sm, size));
+}
+
+/* ARGSUSED */
+static void
+metaslab_ff_claim(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ /* No need to update cursor */
+}
+
+/* ARGSUSED */
+static void
+metaslab_ff_free(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ /* No need to update cursor */
+}
+
+static space_map_ops_t metaslab_ff_ops = {
+ metaslab_ff_load,
+ metaslab_ff_unload,
+ metaslab_ff_alloc,
+ metaslab_ff_claim,
+ metaslab_ff_free
+};
+
+/*
+ * ==========================================================================
+ * Metaslabs
+ * ==========================================================================
+ */
+metaslab_t *
+metaslab_init(metaslab_group_t *mg, space_map_obj_t *smo,
+ uint64_t start, uint64_t size, uint64_t txg)
+{
+ vdev_t *vd = mg->mg_vd;
+ metaslab_t *msp;
+
+ msp = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
+ mutex_init(&msp->ms_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ msp->ms_smo_syncing = *smo;
+
+ /*
+ * We create the main space map here, but we don't create the
+ * allocmaps and freemaps until metaslab_sync_done(). This serves
+ * two purposes: it allows metaslab_sync_done() to detect the
+ * addition of new space; and for debugging, it ensures that we'd
+ * data fault on any attempt to use this metaslab before it's ready.
+ */
+ space_map_create(&msp->ms_map, start, size,
+ vd->vdev_ashift, &msp->ms_lock);
+
+ metaslab_group_add(mg, msp);
+
+ /*
+ * If we're opening an existing pool (txg == 0) or creating
+ * a new one (txg == TXG_INITIAL), all space is available now.
+ * If we're adding space to an existing pool, the new space
+ * does not become available until after this txg has synced.
+ */
+ if (txg <= TXG_INITIAL)
+ metaslab_sync_done(msp, 0);
+
+ if (txg != 0) {
+ /*
+ * The vdev is dirty, but the metaslab isn't -- it just needs
+ * to have metaslab_sync_done() invoked from vdev_sync_done().
+ * [We could just dirty the metaslab, but that would cause us
+ * to allocate a space map object for it, which is wasteful
+ * and would mess up the locality logic in metaslab_weight().]
+ */
+ ASSERT(TXG_CLEAN(txg) == spa_last_synced_txg(vd->vdev_spa));
+ vdev_dirty(vd, 0, NULL, txg);
+ vdev_dirty(vd, VDD_METASLAB, msp, TXG_CLEAN(txg));
+ }
+
+ return (msp);
+}
+
+void
+metaslab_fini(metaslab_t *msp)
+{
+ metaslab_group_t *mg = msp->ms_group;
+ int t;
+
+ vdev_space_update(mg->mg_vd, -msp->ms_map.sm_size,
+ -msp->ms_smo.smo_alloc, B_TRUE);
+
+ metaslab_group_remove(mg, msp);
+
+ mutex_enter(&msp->ms_lock);
+
+ space_map_unload(&msp->ms_map);
+ space_map_destroy(&msp->ms_map);
+
+ for (t = 0; t < TXG_SIZE; t++) {
+ space_map_destroy(&msp->ms_allocmap[t]);
+ space_map_destroy(&msp->ms_freemap[t]);
+ }
+
+ mutex_exit(&msp->ms_lock);
+ mutex_destroy(&msp->ms_lock);
+
+ kmem_free(msp, sizeof (metaslab_t));
+}
+
+#define METASLAB_WEIGHT_PRIMARY (1ULL << 63)
+#define METASLAB_WEIGHT_SECONDARY (1ULL << 62)
+#define METASLAB_ACTIVE_MASK \
+ (METASLAB_WEIGHT_PRIMARY | METASLAB_WEIGHT_SECONDARY)
+#define METASLAB_SMO_BONUS_MULTIPLIER 2
+
+static uint64_t
+metaslab_weight(metaslab_t *msp)
+{
+ metaslab_group_t *mg = msp->ms_group;
+ space_map_t *sm = &msp->ms_map;
+ space_map_obj_t *smo = &msp->ms_smo;
+ vdev_t *vd = mg->mg_vd;
+ uint64_t weight, space;
+
+ ASSERT(MUTEX_HELD(&msp->ms_lock));
+
+ /*
+ * The baseline weight is the metaslab's free space.
+ */
+ space = sm->sm_size - smo->smo_alloc;
+ weight = space;
+
+ /*
+ * Modern disks have uniform bit density and constant angular velocity.
+ * Therefore, the outer recording zones are faster (higher bandwidth)
+ * than the inner zones by the ratio of outer to inner track diameter,
+ * which is typically around 2:1. We account for this by assigning
+ * higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
+ * In effect, this means that we'll select the metaslab with the most
+ * free bandwidth rather than simply the one with the most free space.
+ */
+ weight = 2 * weight -
+ ((sm->sm_start >> vd->vdev_ms_shift) * weight) / vd->vdev_ms_count;
+ ASSERT(weight >= space && weight <= 2 * space);
+
+ /*
+ * For locality, assign higher weight to metaslabs we've used before.
+ */
+ if (smo->smo_object != 0)
+ weight *= METASLAB_SMO_BONUS_MULTIPLIER;
+ ASSERT(weight >= space &&
+ weight <= 2 * METASLAB_SMO_BONUS_MULTIPLIER * space);
+
+ /*
+ * If this metaslab is one we're actively using, adjust its weight to
+ * make it preferable to any inactive metaslab so we'll polish it off.
+ */
+ weight |= (msp->ms_weight & METASLAB_ACTIVE_MASK);
+
+ return (weight);
+}
+
+static int
+metaslab_activate(metaslab_t *msp, uint64_t activation_weight)
+{
+ space_map_t *sm = &msp->ms_map;
+
+ ASSERT(MUTEX_HELD(&msp->ms_lock));
+
+ if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
+ int error = space_map_load(sm, &metaslab_ff_ops,
+ SM_FREE, &msp->ms_smo,
+ msp->ms_group->mg_vd->vdev_spa->spa_meta_objset);
+ if (error) {
+ metaslab_group_sort(msp->ms_group, msp, 0);
+ return (error);
+ }
+ metaslab_group_sort(msp->ms_group, msp,
+ msp->ms_weight | activation_weight);
+ }
+ ASSERT(sm->sm_loaded);
+ ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
+
+ return (0);
+}
+
+static void
+metaslab_passivate(metaslab_t *msp, uint64_t size)
+{
+ /*
+ * If size < SPA_MINBLOCKSIZE, then we will not allocate from
+ * this metaslab again. In that case, it had better be empty,
+ * or we would be leaving space on the table.
+ */
+ ASSERT(size >= SPA_MINBLOCKSIZE || msp->ms_map.sm_space == 0);
+ metaslab_group_sort(msp->ms_group, msp, MIN(msp->ms_weight, size));
+ ASSERT((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0);
+}
+
+/*
+ * Write a metaslab to disk in the context of the specified transaction group.
+ */
+void
+metaslab_sync(metaslab_t *msp, uint64_t txg)
+{
+ vdev_t *vd = msp->ms_group->mg_vd;
+ spa_t *spa = vd->vdev_spa;
+ objset_t *mos = spa->spa_meta_objset;
+ space_map_t *allocmap = &msp->ms_allocmap[txg & TXG_MASK];
+ space_map_t *freemap = &msp->ms_freemap[txg & TXG_MASK];
+ space_map_t *freed_map = &msp->ms_freemap[TXG_CLEAN(txg) & TXG_MASK];
+ space_map_t *sm = &msp->ms_map;
+ space_map_obj_t *smo = &msp->ms_smo_syncing;
+ dmu_buf_t *db;
+ dmu_tx_t *tx;
+ int t;
+
+ tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
+
+ /*
+ * The only state that can actually be changing concurrently with
+ * metaslab_sync() is the metaslab's ms_map. No other thread can
+ * be modifying this txg's allocmap, freemap, freed_map, or smo.
+ * Therefore, we only hold ms_lock to satify space_map ASSERTs.
+ * We drop it whenever we call into the DMU, because the DMU
+ * can call down to us (e.g. via zio_free()) at any time.
+ */
+ mutex_enter(&msp->ms_lock);
+
+ if (smo->smo_object == 0) {
+ ASSERT(smo->smo_objsize == 0);
+ ASSERT(smo->smo_alloc == 0);
+ mutex_exit(&msp->ms_lock);
+ smo->smo_object = dmu_object_alloc(mos,
+ DMU_OT_SPACE_MAP, 1 << SPACE_MAP_BLOCKSHIFT,
+ DMU_OT_SPACE_MAP_HEADER, sizeof (*smo), tx);
+ ASSERT(smo->smo_object != 0);
+ dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
+ (sm->sm_start >> vd->vdev_ms_shift),
+ sizeof (uint64_t), &smo->smo_object, tx);
+ mutex_enter(&msp->ms_lock);
+ }
+
+ space_map_walk(freemap, space_map_add, freed_map);
+
+ if (sm->sm_loaded && spa_sync_pass(spa) == 1 && smo->smo_objsize >=
+ 2 * sizeof (uint64_t) * avl_numnodes(&sm->sm_root)) {
+ /*
+ * The in-core space map representation is twice as compact
+ * as the on-disk one, so it's time to condense the latter
+ * by generating a pure allocmap from first principles.
+ *
+ * This metaslab is 100% allocated,
+ * minus the content of the in-core map (sm),
+ * minus what's been freed this txg (freed_map),
+ * minus allocations from txgs in the future
+ * (because they haven't been committed yet).
+ */
+ space_map_vacate(allocmap, NULL, NULL);
+ space_map_vacate(freemap, NULL, NULL);
+
+ space_map_add(allocmap, allocmap->sm_start, allocmap->sm_size);
+
+ space_map_walk(sm, space_map_remove, allocmap);
+ space_map_walk(freed_map, space_map_remove, allocmap);
+
+ for (t = 1; t < TXG_CONCURRENT_STATES; t++)
+ space_map_walk(&msp->ms_allocmap[(txg + t) & TXG_MASK],
+ space_map_remove, allocmap);
+
+ mutex_exit(&msp->ms_lock);
+ space_map_truncate(smo, mos, tx);
+ mutex_enter(&msp->ms_lock);
+ }
+
+ space_map_sync(allocmap, SM_ALLOC, smo, mos, tx);
+ space_map_sync(freemap, SM_FREE, smo, mos, tx);
+
+ mutex_exit(&msp->ms_lock);
+
+ VERIFY(0 == dmu_bonus_hold(mos, smo->smo_object, FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+ ASSERT3U(db->db_size, >=, sizeof (*smo));
+ bcopy(smo, db->db_data, sizeof (*smo));
+ dmu_buf_rele(db, FTAG);
+
+ dmu_tx_commit(tx);
+}
+
+/*
+ * Called after a transaction group has completely synced to mark
+ * all of the metaslab's free space as usable.
+ */
+void
+metaslab_sync_done(metaslab_t *msp, uint64_t txg)
+{
+ space_map_obj_t *smo = &msp->ms_smo;
+ space_map_obj_t *smosync = &msp->ms_smo_syncing;
+ space_map_t *sm = &msp->ms_map;
+ space_map_t *freed_map = &msp->ms_freemap[TXG_CLEAN(txg) & TXG_MASK];
+ metaslab_group_t *mg = msp->ms_group;
+ vdev_t *vd = mg->mg_vd;
+ int t;
+
+ mutex_enter(&msp->ms_lock);
+
+ /*
+ * If this metaslab is just becoming available, initialize its
+ * allocmaps and freemaps and add its capacity to the vdev.
+ */
+ if (freed_map->sm_size == 0) {
+ for (t = 0; t < TXG_SIZE; t++) {
+ space_map_create(&msp->ms_allocmap[t], sm->sm_start,
+ sm->sm_size, sm->sm_shift, sm->sm_lock);
+ space_map_create(&msp->ms_freemap[t], sm->sm_start,
+ sm->sm_size, sm->sm_shift, sm->sm_lock);
+ }
+ vdev_space_update(vd, sm->sm_size, 0, B_TRUE);
+ }
+
+ vdev_space_update(vd, 0, smosync->smo_alloc - smo->smo_alloc, B_TRUE);
+
+ ASSERT(msp->ms_allocmap[txg & TXG_MASK].sm_space == 0);
+ ASSERT(msp->ms_freemap[txg & TXG_MASK].sm_space == 0);
+
+ /*
+ * If there's a space_map_load() in progress, wait for it to complete
+ * so that we have a consistent view of the in-core space map.
+ * Then, add everything we freed in this txg to the map.
+ */
+ space_map_load_wait(sm);
+ space_map_vacate(freed_map, sm->sm_loaded ? space_map_free : NULL, sm);
+
+ *smo = *smosync;
+
+ /*
+ * If the map is loaded but no longer active, evict it as soon as all
+ * future allocations have synced. (If we unloaded it now and then
+ * loaded a moment later, the map wouldn't reflect those allocations.)
+ */
+ if (sm->sm_loaded && (msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
+ int evictable = 1;
+
+ for (t = 1; t < TXG_CONCURRENT_STATES; t++)
+ if (msp->ms_allocmap[(txg + t) & TXG_MASK].sm_space)
+ evictable = 0;
+
+ if (evictable)
+ space_map_unload(sm);
+ }
+
+ metaslab_group_sort(mg, msp, metaslab_weight(msp));
+
+ mutex_exit(&msp->ms_lock);
+}
+
+static uint64_t
+metaslab_distance(metaslab_t *msp, dva_t *dva)
+{
+ uint64_t ms_shift = msp->ms_group->mg_vd->vdev_ms_shift;
+ uint64_t offset = DVA_GET_OFFSET(dva) >> ms_shift;
+ uint64_t start = msp->ms_map.sm_start >> ms_shift;
+
+ if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
+ return (1ULL << 63);
+
+ if (offset < start)
+ return ((start - offset) << ms_shift);
+ if (offset > start)
+ return ((offset - start) << ms_shift);
+ return (0);
+}
+
+static uint64_t
+metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
+ uint64_t min_distance, dva_t *dva, int d)
+{
+ metaslab_t *msp = NULL;
+ uint64_t offset = -1ULL;
+ avl_tree_t *t = &mg->mg_metaslab_tree;
+ uint64_t activation_weight;
+ uint64_t target_distance;
+ int i;
+
+ activation_weight = METASLAB_WEIGHT_PRIMARY;
+ for (i = 0; i < d; i++)
+ if (DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id)
+ activation_weight = METASLAB_WEIGHT_SECONDARY;
+
+ for (;;) {
+ mutex_enter(&mg->mg_lock);
+ for (msp = avl_first(t); msp; msp = AVL_NEXT(t, msp)) {
+ if (msp->ms_weight < size) {
+ mutex_exit(&mg->mg_lock);
+ return (-1ULL);
+ }
+
+ if (activation_weight == METASLAB_WEIGHT_PRIMARY)
+ break;
+
+ target_distance = min_distance +
+ (msp->ms_smo.smo_alloc ? 0 : min_distance >> 1);
+
+ for (i = 0; i < d; i++)
+ if (metaslab_distance(msp, &dva[i]) <
+ target_distance)
+ break;
+ if (i == d)
+ break;
+ }
+ mutex_exit(&mg->mg_lock);
+ if (msp == NULL)
+ return (-1ULL);
+
+ mutex_enter(&msp->ms_lock);
+
+ /*
+ * Ensure that the metaslab we have selected is still
+ * capable of handling our request. It's possible that
+ * another thread may have changed the weight while we
+ * were blocked on the metaslab lock.
+ */
+ if (msp->ms_weight < size) {
+ mutex_exit(&msp->ms_lock);
+ continue;
+ }
+
+ if ((msp->ms_weight & METASLAB_WEIGHT_SECONDARY) &&
+ activation_weight == METASLAB_WEIGHT_PRIMARY) {
+ metaslab_passivate(msp,
+ msp->ms_weight & ~METASLAB_ACTIVE_MASK);
+ mutex_exit(&msp->ms_lock);
+ continue;
+ }
+
+ if (metaslab_activate(msp, activation_weight) != 0) {
+ mutex_exit(&msp->ms_lock);
+ continue;
+ }
+
+ if ((offset = space_map_alloc(&msp->ms_map, size)) != -1ULL)
+ break;
+
+ metaslab_passivate(msp, size - 1);
+
+ mutex_exit(&msp->ms_lock);
+ }
+
+ if (msp->ms_allocmap[txg & TXG_MASK].sm_space == 0)
+ vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);
+
+ space_map_add(&msp->ms_allocmap[txg & TXG_MASK], offset, size);
+
+ mutex_exit(&msp->ms_lock);
+
+ return (offset);
+}
+
+/*
+ * Allocate a block for the specified i/o.
+ */
+static int
+metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
+ dva_t *dva, int d, dva_t *hintdva, uint64_t txg, int flags)
+{
+ metaslab_group_t *mg, *rotor;
+ vdev_t *vd;
+ int dshift = 3;
+ int all_zero;
+ uint64_t offset = -1ULL;
+ uint64_t asize;
+ uint64_t distance;
+
+ ASSERT(!DVA_IS_VALID(&dva[d]));
+
+ /*
+ * For testing, make some blocks above a certain size be gang blocks.
+ */
+ if (psize >= metaslab_gang_bang && (lbolt & 3) == 0)
+ return (ENOSPC);
+
+ /*
+ * Start at the rotor and loop through all mgs until we find something.
+ * Note that there's no locking on mc_rotor or mc_allocated because
+ * nothing actually breaks if we miss a few updates -- we just won't
+ * allocate quite as evenly. It all balances out over time.
+ *
+ * If we are doing ditto or log blocks, try to spread them across
+ * consecutive vdevs. If we're forced to reuse a vdev before we've
+ * allocated all of our ditto blocks, then try and spread them out on
+ * that vdev as much as possible. If it turns out to not be possible,
+ * gradually lower our standards until anything becomes acceptable.
+ * Also, allocating on consecutive vdevs (as opposed to random vdevs)
+ * gives us hope of containing our fault domains to something we're
+ * able to reason about. Otherwise, any two top-level vdev failures
+ * will guarantee the loss of data. With consecutive allocation,
+ * only two adjacent top-level vdev failures will result in data loss.
+ *
+ * If we are doing gang blocks (hintdva is non-NULL), try to keep
+ * ourselves on the same vdev as our gang block header. That
+ * way, we can hope for locality in vdev_cache, plus it makes our
+ * fault domains something tractable.
+ */
+ if (hintdva) {
+ vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));
+ if (flags & METASLAB_HINTBP_AVOID)
+ mg = vd->vdev_mg->mg_next;
+ else
+ mg = vd->vdev_mg;
+ } else if (d != 0) {
+ vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
+ mg = vd->vdev_mg->mg_next;
+ } else {
+ mg = mc->mc_rotor;
+ }
+
+ /*
+ * If the hint put us into the wrong class, just follow the rotor.
+ */
+ if (mg->mg_class != mc)
+ mg = mc->mc_rotor;
+
+ rotor = mg;
+top:
+ all_zero = B_TRUE;
+ do {
+ vd = mg->mg_vd;
+ /*
+ * Don't allocate from faulted devices.
+ */
+ if (!vdev_allocatable(vd))
+ goto next;
+ /*
+ * Avoid writing single-copy data to a failing vdev
+ */
+ if ((vd->vdev_stat.vs_write_errors > 0 ||
+ vd->vdev_state < VDEV_STATE_HEALTHY) &&
+ d == 0 && dshift == 3) {
+ all_zero = B_FALSE;
+ goto next;
+ }
+
+ ASSERT(mg->mg_class == mc);
+
+ distance = vd->vdev_asize >> dshift;
+ if (distance <= (1ULL << vd->vdev_ms_shift))
+ distance = 0;
+ else
+ all_zero = B_FALSE;
+
+ asize = vdev_psize_to_asize(vd, psize);
+ ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);
+
+ offset = metaslab_group_alloc(mg, asize, txg, distance, dva, d);
+ if (offset != -1ULL) {
+ /*
+ * If we've just selected this metaslab group,
+ * figure out whether the corresponding vdev is
+ * over- or under-used relative to the pool,
+ * and set an allocation bias to even it out.
+ */
+ if (mc->mc_allocated == 0) {
+ vdev_stat_t *vs = &vd->vdev_stat;
+ uint64_t alloc, space;
+ int64_t vu, su;
+
+ alloc = spa_get_alloc(spa);
+ space = spa_get_space(spa);
+
+ /*
+ * Determine percent used in units of 0..1024.
+ * (This is just to avoid floating point.)
+ */
+ vu = (vs->vs_alloc << 10) / (vs->vs_space + 1);
+ su = (alloc << 10) / (space + 1);
+
+ /*
+ * Bias by at most +/- 25% of the aliquot.
+ */
+ mg->mg_bias = ((su - vu) *
+ (int64_t)mg->mg_aliquot) / (1024 * 4);
+ }
+
+ if (atomic_add_64_nv(&mc->mc_allocated, asize) >=
+ mg->mg_aliquot + mg->mg_bias) {
+ mc->mc_rotor = mg->mg_next;
+ mc->mc_allocated = 0;
+ }
+
+ DVA_SET_VDEV(&dva[d], vd->vdev_id);
+ DVA_SET_OFFSET(&dva[d], offset);
+ DVA_SET_GANG(&dva[d], !!(flags & METASLAB_GANG_HEADER));
+ DVA_SET_ASIZE(&dva[d], asize);
+
+ return (0);
+ }
+next:
+ mc->mc_rotor = mg->mg_next;
+ mc->mc_allocated = 0;
+ } while ((mg = mg->mg_next) != rotor);
+
+ if (!all_zero) {
+ dshift++;
+ ASSERT(dshift < 64);
+ goto top;
+ }
+
+ bzero(&dva[d], sizeof (dva_t));
+
+ return (ENOSPC);
+}
+
+/*
+ * Free the block represented by DVA in the context of the specified
+ * transaction group.
+ */
+static void
+metaslab_free_dva(spa_t *spa, const dva_t *dva, uint64_t txg, boolean_t now)
+{
+ uint64_t vdev = DVA_GET_VDEV(dva);
+ uint64_t offset = DVA_GET_OFFSET(dva);
+ uint64_t size = DVA_GET_ASIZE(dva);
+ vdev_t *vd;
+ metaslab_t *msp;
+
+ ASSERT(DVA_IS_VALID(dva));
+
+ if (txg > spa_freeze_txg(spa))
+ return;
+
+ if ((vd = vdev_lookup_top(spa, vdev)) == NULL ||
+ (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
+ cmn_err(CE_WARN, "metaslab_free_dva(): bad DVA %llu:%llu",
+ (u_longlong_t)vdev, (u_longlong_t)offset);
+ ASSERT(0);
+ return;
+ }
+
+ msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
+
+ if (DVA_GET_GANG(dva))
+ size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);
+
+ mutex_enter(&msp->ms_lock);
+
+ if (now) {
+ space_map_remove(&msp->ms_allocmap[txg & TXG_MASK],
+ offset, size);
+ space_map_free(&msp->ms_map, offset, size);
+ } else {
+ if (msp->ms_freemap[txg & TXG_MASK].sm_space == 0)
+ vdev_dirty(vd, VDD_METASLAB, msp, txg);
+ space_map_add(&msp->ms_freemap[txg & TXG_MASK], offset, size);
+ }
+
+ mutex_exit(&msp->ms_lock);
+}
+
+/*
+ * Intent log support: upon opening the pool after a crash, notify the SPA
+ * of blocks that the intent log has allocated for immediate write, but
+ * which are still considered free by the SPA because the last transaction
+ * group didn't commit yet.
+ */
+static int
+metaslab_claim_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
+{
+ uint64_t vdev = DVA_GET_VDEV(dva);
+ uint64_t offset = DVA_GET_OFFSET(dva);
+ uint64_t size = DVA_GET_ASIZE(dva);
+ vdev_t *vd;
+ metaslab_t *msp;
+ int error;
+
+ ASSERT(DVA_IS_VALID(dva));
+
+ if ((vd = vdev_lookup_top(spa, vdev)) == NULL ||
+ (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count)
+ return (ENXIO);
+
+ msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
+
+ if (DVA_GET_GANG(dva))
+ size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);
+
+ mutex_enter(&msp->ms_lock);
+
+ error = metaslab_activate(msp, METASLAB_WEIGHT_SECONDARY);
+ if (error || txg == 0) { /* txg == 0 indicates dry run */
+ mutex_exit(&msp->ms_lock);
+ return (error);
+ }
+
+ space_map_claim(&msp->ms_map, offset, size);
+
+ if (spa_mode & FWRITE) { /* don't dirty if we're zdb(1M) */
+ if (msp->ms_allocmap[txg & TXG_MASK].sm_space == 0)
+ vdev_dirty(vd, VDD_METASLAB, msp, txg);
+ space_map_add(&msp->ms_allocmap[txg & TXG_MASK], offset, size);
+ }
+
+ mutex_exit(&msp->ms_lock);
+
+ return (0);
+}
+
+int
+metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize, blkptr_t *bp,
+ int ndvas, uint64_t txg, blkptr_t *hintbp, int flags)
+{
+ dva_t *dva = bp->blk_dva;
+ dva_t *hintdva = hintbp->blk_dva;
+ int error = 0;
+
+ ASSERT(bp->blk_birth == 0);
+
+ spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
+
+ if (mc->mc_rotor == NULL) { /* no vdevs in this class */
+ spa_config_exit(spa, SCL_ALLOC, FTAG);
+ return (ENOSPC);
+ }
+
+ ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
+ ASSERT(BP_GET_NDVAS(bp) == 0);
+ ASSERT(hintbp == NULL || ndvas <= BP_GET_NDVAS(hintbp));
+
+ for (int d = 0; d < ndvas; d++) {
+ error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
+ txg, flags);
+ if (error) {
+ for (d--; d >= 0; d--) {
+ metaslab_free_dva(spa, &dva[d], txg, B_TRUE);
+ bzero(&dva[d], sizeof (dva_t));
+ }
+ spa_config_exit(spa, SCL_ALLOC, FTAG);
+ return (error);
+ }
+ }
+ ASSERT(error == 0);
+ ASSERT(BP_GET_NDVAS(bp) == ndvas);
+
+ spa_config_exit(spa, SCL_ALLOC, FTAG);
+
+ bp->blk_birth = txg;
+
+ return (0);
+}
+
+void
+metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now)
+{
+ const dva_t *dva = bp->blk_dva;
+ int ndvas = BP_GET_NDVAS(bp);
+
+ ASSERT(!BP_IS_HOLE(bp));
+ ASSERT(!now || bp->blk_birth >= spa->spa_syncing_txg);
+
+ spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);
+
+ for (int d = 0; d < ndvas; d++)
+ metaslab_free_dva(spa, &dva[d], txg, now);
+
+ spa_config_exit(spa, SCL_FREE, FTAG);
+}
+
+int
+metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
+{
+ const dva_t *dva = bp->blk_dva;
+ int ndvas = BP_GET_NDVAS(bp);
+ int error = 0;
+
+ ASSERT(!BP_IS_HOLE(bp));
+
+ if (txg != 0) {
+ /*
+ * First do a dry run to make sure all DVAs are claimable,
+ * so we don't have to unwind from partial failures below.
+ */
+ if ((error = metaslab_claim(spa, bp, 0)) != 0)
+ return (error);
+ }
+
+ spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
+
+ for (int d = 0; d < ndvas; d++)
+ if ((error = metaslab_claim_dva(spa, &dva[d], txg)) != 0)
+ break;
+
+ spa_config_exit(spa, SCL_ALLOC, FTAG);
+
+ ASSERT(error == 0 || txg == 0);
+
+ return (error);
+}
diff --git a/module/zfs/refcount.c b/module/zfs/refcount.c
new file mode 100644
index 000000000..f1b3b23fe
--- /dev/null
+++ b/module/zfs/refcount.c
@@ -0,0 +1,195 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/refcount.h>
+
+#if defined(DEBUG) || !defined(_KERNEL)
+
+#ifdef _KERNEL
+int reference_tracking_enable = FALSE; /* runs out of memory too easily */
+#else
+int reference_tracking_enable = TRUE;
+#endif
+int reference_history = 4; /* tunable */
+
+static kmem_cache_t *reference_cache;
+static kmem_cache_t *reference_history_cache;
+
+void
+refcount_init(void)
+{
+ reference_cache = kmem_cache_create("reference_cache",
+ sizeof (reference_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
+
+ reference_history_cache = kmem_cache_create("reference_history_cache",
+ sizeof (uint64_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
+}
+
+void
+refcount_fini(void)
+{
+ kmem_cache_destroy(reference_cache);
+ kmem_cache_destroy(reference_history_cache);
+}
+
+void
+refcount_create(refcount_t *rc)
+{
+ mutex_init(&rc->rc_mtx, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&rc->rc_list, sizeof (reference_t),
+ offsetof(reference_t, ref_link));
+ list_create(&rc->rc_removed, sizeof (reference_t),
+ offsetof(reference_t, ref_link));
+ rc->rc_count = 0;
+ rc->rc_removed_count = 0;
+}
+
+void
+refcount_destroy_many(refcount_t *rc, uint64_t number)
+{
+ reference_t *ref;
+
+ ASSERT(rc->rc_count == number);
+ while (ref = list_head(&rc->rc_list)) {
+ list_remove(&rc->rc_list, ref);
+ kmem_cache_free(reference_cache, ref);
+ }
+ list_destroy(&rc->rc_list);
+
+ while (ref = list_head(&rc->rc_removed)) {
+ list_remove(&rc->rc_removed, ref);
+ kmem_cache_free(reference_history_cache, ref->ref_removed);
+ kmem_cache_free(reference_cache, ref);
+ }
+ list_destroy(&rc->rc_removed);
+ mutex_destroy(&rc->rc_mtx);
+}
+
+void
+refcount_destroy(refcount_t *rc)
+{
+ refcount_destroy_many(rc, 0);
+}
+
+int
+refcount_is_zero(refcount_t *rc)
+{
+ ASSERT(rc->rc_count >= 0);
+ return (rc->rc_count == 0);
+}
+
+int64_t
+refcount_count(refcount_t *rc)
+{
+ ASSERT(rc->rc_count >= 0);
+ return (rc->rc_count);
+}
+
+int64_t
+refcount_add_many(refcount_t *rc, uint64_t number, void *holder)
+{
+ reference_t *ref;
+ int64_t count;
+
+ if (reference_tracking_enable) {
+ ref = kmem_cache_alloc(reference_cache, KM_SLEEP);
+ ref->ref_holder = holder;
+ ref->ref_number = number;
+ }
+ mutex_enter(&rc->rc_mtx);
+ ASSERT(rc->rc_count >= 0);
+ if (reference_tracking_enable)
+ list_insert_head(&rc->rc_list, ref);
+ rc->rc_count += number;
+ count = rc->rc_count;
+ mutex_exit(&rc->rc_mtx);
+
+ return (count);
+}
+
+int64_t
+refcount_add(refcount_t *rc, void *holder)
+{
+ return (refcount_add_many(rc, 1, holder));
+}
+
+int64_t
+refcount_remove_many(refcount_t *rc, uint64_t number, void *holder)
+{
+ reference_t *ref;
+ int64_t count;
+
+ mutex_enter(&rc->rc_mtx);
+ ASSERT(rc->rc_count >= number);
+
+ if (!reference_tracking_enable) {
+ rc->rc_count -= number;
+ count = rc->rc_count;
+ mutex_exit(&rc->rc_mtx);
+ return (count);
+ }
+
+ for (ref = list_head(&rc->rc_list); ref;
+ ref = list_next(&rc->rc_list, ref)) {
+ if (ref->ref_holder == holder && ref->ref_number == number) {
+ list_remove(&rc->rc_list, ref);
+ if (reference_history > 0) {
+ ref->ref_removed =
+ kmem_cache_alloc(reference_history_cache,
+ KM_SLEEP);
+ list_insert_head(&rc->rc_removed, ref);
+ rc->rc_removed_count++;
+ if (rc->rc_removed_count >= reference_history) {
+ ref = list_tail(&rc->rc_removed);
+ list_remove(&rc->rc_removed, ref);
+ kmem_cache_free(reference_history_cache,
+ ref->ref_removed);
+ kmem_cache_free(reference_cache, ref);
+ rc->rc_removed_count--;
+ }
+ } else {
+ kmem_cache_free(reference_cache, ref);
+ }
+ rc->rc_count -= number;
+ count = rc->rc_count;
+ mutex_exit(&rc->rc_mtx);
+ return (count);
+ }
+ }
+ panic("No such hold %p on refcount %llx", holder,
+ (u_longlong_t)(uintptr_t)rc);
+ return (-1);
+}
+
+int64_t
+refcount_remove(refcount_t *rc, void *holder)
+{
+ return (refcount_remove_many(rc, 1, holder));
+}
+
+#endif
diff --git a/module/zfs/rrwlock.c b/module/zfs/rrwlock.c
new file mode 100644
index 000000000..710685dbc
--- /dev/null
+++ b/module/zfs/rrwlock.c
@@ -0,0 +1,249 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/refcount.h>
+#include <sys/rrwlock.h>
+
+/*
+ * This file contains the implementation of a re-entrant read
+ * reader/writer lock (aka "rrwlock").
+ *
+ * This is a normal reader/writer lock with the additional feature
+ * of allowing threads who have already obtained a read lock to
+ * re-enter another read lock (re-entrant read) - even if there are
+ * waiting writers.
+ *
+ * Callers who have not obtained a read lock give waiting writers priority.
+ *
+ * The rrwlock_t lock does not allow re-entrant writers, nor does it
+ * allow a re-entrant mix of reads and writes (that is, it does not
+ * allow a caller who has already obtained a read lock to be able to
+ * then grab a write lock without first dropping all read locks, and
+ * vice versa).
+ *
+ * The rrwlock_t uses tsd (thread specific data) to keep a list of
+ * nodes (rrw_node_t), where each node keeps track of which specific
+ * lock (rrw_node_t::rn_rrl) the thread has grabbed. Since re-entering
+ * should be rare, a thread that grabs multiple reads on the same rrwlock_t
+ * will store multiple rrw_node_ts of the same 'rrn_rrl'. Nodes on the
+ * tsd list can represent a different rrwlock_t. This allows a thread
+ * to enter multiple and unique rrwlock_ts for read locks at the same time.
+ *
+ * Since using tsd exposes some overhead, the rrwlock_t only needs to
+ * keep tsd data when writers are waiting. If no writers are waiting, then
+ * a reader just bumps the anonymous read count (rr_anon_rcount) - no tsd
+ * is needed. Once a writer attempts to grab the lock, readers then
+ * keep tsd data and bump the linked readers count (rr_linked_rcount).
+ *
+ * If there are waiting writers and there are anonymous readers, then a
+ * reader doesn't know if it is a re-entrant lock. But since it may be one,
+ * we allow the read to proceed (otherwise it could deadlock). Since once
+ * waiting writers are active, readers no longer bump the anonymous count,
+ * the anonymous readers will eventually flush themselves out. At this point,
+ * readers will be able to tell if they are a re-entrant lock (have a
+ * rrw_node_t entry for the lock) or not. If they are a re-entrant lock, then
+ * we must let the proceed. If they are not, then the reader blocks for the
+ * waiting writers. Hence, we do not starve writers.
+ */
+
+/* global key for TSD */
+uint_t rrw_tsd_key;
+
+typedef struct rrw_node {
+ struct rrw_node *rn_next;
+ rrwlock_t *rn_rrl;
+} rrw_node_t;
+
+static rrw_node_t *
+rrn_find(rrwlock_t *rrl)
+{
+ rrw_node_t *rn;
+
+ if (refcount_count(&rrl->rr_linked_rcount) == 0)
+ return (NULL);
+
+ for (rn = tsd_get(rrw_tsd_key); rn != NULL; rn = rn->rn_next) {
+ if (rn->rn_rrl == rrl)
+ return (rn);
+ }
+ return (NULL);
+}
+
+/*
+ * Add a node to the head of the singly linked list.
+ */
+static void
+rrn_add(rrwlock_t *rrl)
+{
+ rrw_node_t *rn;
+
+ rn = kmem_alloc(sizeof (*rn), KM_SLEEP);
+ rn->rn_rrl = rrl;
+ rn->rn_next = tsd_get(rrw_tsd_key);
+ VERIFY(tsd_set(rrw_tsd_key, rn) == 0);
+}
+
+/*
+ * If a node is found for 'rrl', then remove the node from this
+ * thread's list and return TRUE; otherwise return FALSE.
+ */
+static boolean_t
+rrn_find_and_remove(rrwlock_t *rrl)
+{
+ rrw_node_t *rn;
+ rrw_node_t *prev = NULL;
+
+ if (refcount_count(&rrl->rr_linked_rcount) == 0)
+ return (NULL);
+
+ for (rn = tsd_get(rrw_tsd_key); rn != NULL; rn = rn->rn_next) {
+ if (rn->rn_rrl == rrl) {
+ if (prev)
+ prev->rn_next = rn->rn_next;
+ else
+ VERIFY(tsd_set(rrw_tsd_key, rn->rn_next) == 0);
+ kmem_free(rn, sizeof (*rn));
+ return (B_TRUE);
+ }
+ prev = rn;
+ }
+ return (B_FALSE);
+}
+
+void
+rrw_init(rrwlock_t *rrl)
+{
+ mutex_init(&rrl->rr_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&rrl->rr_cv, NULL, CV_DEFAULT, NULL);
+ rrl->rr_writer = NULL;
+ refcount_create(&rrl->rr_anon_rcount);
+ refcount_create(&rrl->rr_linked_rcount);
+ rrl->rr_writer_wanted = B_FALSE;
+}
+
+void
+rrw_destroy(rrwlock_t *rrl)
+{
+ mutex_destroy(&rrl->rr_lock);
+ cv_destroy(&rrl->rr_cv);
+ ASSERT(rrl->rr_writer == NULL);
+ refcount_destroy(&rrl->rr_anon_rcount);
+ refcount_destroy(&rrl->rr_linked_rcount);
+}
+
+static void
+rrw_enter_read(rrwlock_t *rrl, void *tag)
+{
+ mutex_enter(&rrl->rr_lock);
+ ASSERT(rrl->rr_writer != curthread);
+ ASSERT(refcount_count(&rrl->rr_anon_rcount) >= 0);
+
+ while (rrl->rr_writer || (rrl->rr_writer_wanted &&
+ refcount_is_zero(&rrl->rr_anon_rcount) &&
+ rrn_find(rrl) == NULL))
+ cv_wait(&rrl->rr_cv, &rrl->rr_lock);
+
+ if (rrl->rr_writer_wanted) {
+ /* may or may not be a re-entrant enter */
+ rrn_add(rrl);
+ (void) refcount_add(&rrl->rr_linked_rcount, tag);
+ } else {
+ (void) refcount_add(&rrl->rr_anon_rcount, tag);
+ }
+ ASSERT(rrl->rr_writer == NULL);
+ mutex_exit(&rrl->rr_lock);
+}
+
+static void
+rrw_enter_write(rrwlock_t *rrl)
+{
+ mutex_enter(&rrl->rr_lock);
+ ASSERT(rrl->rr_writer != curthread);
+
+ while (refcount_count(&rrl->rr_anon_rcount) > 0 ||
+ refcount_count(&rrl->rr_linked_rcount) > 0 ||
+ rrl->rr_writer != NULL) {
+ rrl->rr_writer_wanted = B_TRUE;
+ cv_wait(&rrl->rr_cv, &rrl->rr_lock);
+ }
+ rrl->rr_writer_wanted = B_FALSE;
+ rrl->rr_writer = curthread;
+ mutex_exit(&rrl->rr_lock);
+}
+
+void
+rrw_enter(rrwlock_t *rrl, krw_t rw, void *tag)
+{
+ if (rw == RW_READER)
+ rrw_enter_read(rrl, tag);
+ else
+ rrw_enter_write(rrl);
+}
+
+void
+rrw_exit(rrwlock_t *rrl, void *tag)
+{
+ mutex_enter(&rrl->rr_lock);
+ ASSERT(!refcount_is_zero(&rrl->rr_anon_rcount) ||
+ !refcount_is_zero(&rrl->rr_linked_rcount) ||
+ rrl->rr_writer != NULL);
+
+ if (rrl->rr_writer == NULL) {
+ if (rrn_find_and_remove(rrl)) {
+ if (refcount_remove(&rrl->rr_linked_rcount, tag) == 0)
+ cv_broadcast(&rrl->rr_cv);
+
+ } else {
+ if (refcount_remove(&rrl->rr_anon_rcount, tag) == 0)
+ cv_broadcast(&rrl->rr_cv);
+ }
+ } else {
+ ASSERT(rrl->rr_writer == curthread);
+ ASSERT(refcount_is_zero(&rrl->rr_anon_rcount) &&
+ refcount_is_zero(&rrl->rr_linked_rcount));
+ rrl->rr_writer = NULL;
+ cv_broadcast(&rrl->rr_cv);
+ }
+ mutex_exit(&rrl->rr_lock);
+}
+
+boolean_t
+rrw_held(rrwlock_t *rrl, krw_t rw)
+{
+ boolean_t held;
+
+ mutex_enter(&rrl->rr_lock);
+ if (rw == RW_WRITER) {
+ held = (rrl->rr_writer == curthread);
+ } else {
+ held = (!refcount_is_zero(&rrl->rr_anon_rcount) ||
+ !refcount_is_zero(&rrl->rr_linked_rcount));
+ }
+ mutex_exit(&rrl->rr_lock);
+
+ return (held);
+}
diff --git a/module/zfs/sha256.c b/module/zfs/sha256.c
new file mode 100644
index 000000000..ca7076cb6
--- /dev/null
+++ b/module/zfs/sha256.c
@@ -0,0 +1,129 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+
+/*
+ * SHA-256 checksum, as specified in FIPS 180-3, available at:
+ * http://csrc.nist.gov/publications/PubsFIPS.html
+ *
+ * This is a very compact implementation of SHA-256.
+ * It is designed to be simple and portable, not to be fast.
+ */
+
+/*
+ * The literal definitions of Ch() and Maj() according to FIPS 180-3 are:
+ *
+ * Ch(x, y, z) (x & y) ^ (~x & z)
+ * Maj(x, y, z) (x & y) ^ (x & z) ^ (y & z)
+ *
+ * We use equivalent logical reductions here that require one less op.
+ */
+#define Ch(x, y, z) ((z) ^ ((x) & ((y) ^ (z))))
+#define Maj(x, y, z) (((x) & (y)) ^ ((z) & ((x) ^ (y))))
+#define Rot32(x, s) (((x) >> s) | ((x) << (32 - s)))
+#define SIGMA0(x) (Rot32(x, 2) ^ Rot32(x, 13) ^ Rot32(x, 22))
+#define SIGMA1(x) (Rot32(x, 6) ^ Rot32(x, 11) ^ Rot32(x, 25))
+#define sigma0(x) (Rot32(x, 7) ^ Rot32(x, 18) ^ ((x) >> 3))
+#define sigma1(x) (Rot32(x, 17) ^ Rot32(x, 19) ^ ((x) >> 10))
+
+static const uint32_t SHA256_K[64] = {
+ 0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
+ 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
+ 0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
+ 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
+ 0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
+ 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
+ 0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
+ 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
+ 0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
+ 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
+ 0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
+ 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
+ 0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
+ 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
+ 0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
+ 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
+};
+
+static void
+SHA256Transform(uint32_t *H, const uint8_t *cp)
+{
+ uint32_t a, b, c, d, e, f, g, h, t, T1, T2, W[64];
+
+ for (t = 0; t < 16; t++, cp += 4)
+ W[t] = (cp[0] << 24) | (cp[1] << 16) | (cp[2] << 8) | cp[3];
+
+ for (t = 16; t < 64; t++)
+ W[t] = sigma1(W[t - 2]) + W[t - 7] +
+ sigma0(W[t - 15]) + W[t - 16];
+
+ a = H[0]; b = H[1]; c = H[2]; d = H[3];
+ e = H[4]; f = H[5]; g = H[6]; h = H[7];
+
+ for (t = 0; t < 64; t++) {
+ T1 = h + SIGMA1(e) + Ch(e, f, g) + SHA256_K[t] + W[t];
+ T2 = SIGMA0(a) + Maj(a, b, c);
+ h = g; g = f; f = e; e = d + T1;
+ d = c; c = b; b = a; a = T1 + T2;
+ }
+
+ H[0] += a; H[1] += b; H[2] += c; H[3] += d;
+ H[4] += e; H[5] += f; H[6] += g; H[7] += h;
+}
+
+void
+zio_checksum_SHA256(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ uint32_t H[8] = { 0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
+ 0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19 };
+ uint8_t pad[128];
+ int i, padsize;
+
+ for (i = 0; i < (size & ~63ULL); i += 64)
+ SHA256Transform(H, (uint8_t *)buf + i);
+
+ for (padsize = 0; i < size; i++)
+ pad[padsize++] = *((uint8_t *)buf + i);
+
+ for (pad[padsize++] = 0x80; (padsize & 63) != 56; padsize++)
+ pad[padsize] = 0;
+
+ for (i = 56; i >= 0; i -= 8)
+ pad[padsize++] = (size << 3) >> i;
+
+ for (i = 0; i < padsize; i += 64)
+ SHA256Transform(H, pad + i);
+
+ ZIO_SET_CHECKSUM(zcp,
+ (uint64_t)H[0] << 32 | H[1],
+ (uint64_t)H[2] << 32 | H[3],
+ (uint64_t)H[4] << 32 | H[5],
+ (uint64_t)H[6] << 32 | H[7]);
+}
diff --git a/module/zfs/spa.c b/module/zfs/spa.c
new file mode 100644
index 000000000..fb1b96f8b
--- /dev/null
+++ b/module/zfs/spa.c
@@ -0,0 +1,4301 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * This file contains all the routines used when modifying on-disk SPA state.
+ * This includes opening, importing, destroying, exporting a pool, and syncing a
+ * pool.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/fm/fs/zfs.h>
+#include <sys/spa_impl.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+#include <sys/zio_compress.h>
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/zap.h>
+#include <sys/zil.h>
+#include <sys/vdev_impl.h>
+#include <sys/metaslab.h>
+#include <sys/uberblock_impl.h>
+#include <sys/txg.h>
+#include <sys/avl.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dmu_objset.h>
+#include <sys/unique.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_synctask.h>
+#include <sys/fs/zfs.h>
+#include <sys/arc.h>
+#include <sys/callb.h>
+#include <sys/systeminfo.h>
+#include <sys/sunddi.h>
+#include <sys/spa_boot.h>
+
+#include "zfs_prop.h"
+#include "zfs_comutil.h"
+
+int zio_taskq_threads[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
+ /* ISSUE INTR */
+ { 1, 1 }, /* ZIO_TYPE_NULL */
+ { 1, 8 }, /* ZIO_TYPE_READ */
+ { 8, 1 }, /* ZIO_TYPE_WRITE */
+ { 1, 1 }, /* ZIO_TYPE_FREE */
+ { 1, 1 }, /* ZIO_TYPE_CLAIM */
+ { 1, 1 }, /* ZIO_TYPE_IOCTL */
+};
+
+static void spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx);
+static boolean_t spa_has_active_shared_spare(spa_t *spa);
+
+/*
+ * ==========================================================================
+ * SPA properties routines
+ * ==========================================================================
+ */
+
+/*
+ * Add a (source=src, propname=propval) list to an nvlist.
+ */
+static void
+spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval,
+ uint64_t intval, zprop_source_t src)
+{
+ const char *propname = zpool_prop_to_name(prop);
+ nvlist_t *propval;
+
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0);
+
+ if (strval != NULL)
+ VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0);
+ else
+ VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, intval) == 0);
+
+ VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0);
+ nvlist_free(propval);
+}
+
+/*
+ * Get property values from the spa configuration.
+ */
+static void
+spa_prop_get_config(spa_t *spa, nvlist_t **nvp)
+{
+ uint64_t size = spa_get_space(spa);
+ uint64_t used = spa_get_alloc(spa);
+ uint64_t cap, version;
+ zprop_source_t src = ZPROP_SRC_NONE;
+ spa_config_dirent_t *dp;
+
+ ASSERT(MUTEX_HELD(&spa->spa_props_lock));
+
+ /*
+ * readonly properties
+ */
+ spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_USED, NULL, used, src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_AVAILABLE, NULL, size - used, src);
+
+ cap = (size == 0) ? 0 : (used * 100 / size);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);
+
+ spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
+ spa->spa_root_vdev->vdev_state, src);
+
+ /*
+ * settable properties that are not stored in the pool property object.
+ */
+ version = spa_version(spa);
+ if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION))
+ src = ZPROP_SRC_DEFAULT;
+ else
+ src = ZPROP_SRC_LOCAL;
+ spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src);
+
+ if (spa->spa_root != NULL)
+ spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
+ 0, ZPROP_SRC_LOCAL);
+
+ if ((dp = list_head(&spa->spa_config_list)) != NULL) {
+ if (dp->scd_path == NULL) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
+ "none", 0, ZPROP_SRC_LOCAL);
+ } else if (strcmp(dp->scd_path, spa_config_path) != 0) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
+ dp->scd_path, 0, ZPROP_SRC_LOCAL);
+ }
+ }
+}
+
+/*
+ * Get zpool property values.
+ */
+int
+spa_prop_get(spa_t *spa, nvlist_t **nvp)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ objset_t *mos = spa->spa_meta_objset;
+ int err;
+
+ VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ mutex_enter(&spa->spa_props_lock);
+
+ /*
+ * Get properties from the spa config.
+ */
+ spa_prop_get_config(spa, nvp);
+
+ /* If no pool property object, no more prop to get. */
+ if (spa->spa_pool_props_object == 0) {
+ mutex_exit(&spa->spa_props_lock);
+ return (0);
+ }
+
+ /*
+ * Get properties from the MOS pool property object.
+ */
+ for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object);
+ (err = zap_cursor_retrieve(&zc, &za)) == 0;
+ zap_cursor_advance(&zc)) {
+ uint64_t intval = 0;
+ char *strval = NULL;
+ zprop_source_t src = ZPROP_SRC_DEFAULT;
+ zpool_prop_t prop;
+
+ if ((prop = zpool_name_to_prop(za.za_name)) == ZPROP_INVAL)
+ continue;
+
+ switch (za.za_integer_length) {
+ case 8:
+ /* integer property */
+ if (za.za_first_integer !=
+ zpool_prop_default_numeric(prop))
+ src = ZPROP_SRC_LOCAL;
+
+ if (prop == ZPOOL_PROP_BOOTFS) {
+ dsl_pool_t *dp;
+ dsl_dataset_t *ds = NULL;
+
+ dp = spa_get_dsl(spa);
+ rw_enter(&dp->dp_config_rwlock, RW_READER);
+ if (err = dsl_dataset_hold_obj(dp,
+ za.za_first_integer, FTAG, &ds)) {
+ rw_exit(&dp->dp_config_rwlock);
+ break;
+ }
+
+ strval = kmem_alloc(
+ MAXNAMELEN + strlen(MOS_DIR_NAME) + 1,
+ KM_SLEEP);
+ dsl_dataset_name(ds, strval);
+ dsl_dataset_rele(ds, FTAG);
+ rw_exit(&dp->dp_config_rwlock);
+ } else {
+ strval = NULL;
+ intval = za.za_first_integer;
+ }
+
+ spa_prop_add_list(*nvp, prop, strval, intval, src);
+
+ if (strval != NULL)
+ kmem_free(strval,
+ MAXNAMELEN + strlen(MOS_DIR_NAME) + 1);
+
+ break;
+
+ case 1:
+ /* string property */
+ strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
+ err = zap_lookup(mos, spa->spa_pool_props_object,
+ za.za_name, 1, za.za_num_integers, strval);
+ if (err) {
+ kmem_free(strval, za.za_num_integers);
+ break;
+ }
+ spa_prop_add_list(*nvp, prop, strval, 0, src);
+ kmem_free(strval, za.za_num_integers);
+ break;
+
+ default:
+ break;
+ }
+ }
+ zap_cursor_fini(&zc);
+ mutex_exit(&spa->spa_props_lock);
+out:
+ if (err && err != ENOENT) {
+ nvlist_free(*nvp);
+ *nvp = NULL;
+ return (err);
+ }
+
+ return (0);
+}
+
+/*
+ * Validate the given pool properties nvlist and modify the list
+ * for the property values to be set.
+ */
+static int
+spa_prop_validate(spa_t *spa, nvlist_t *props)
+{
+ nvpair_t *elem;
+ int error = 0, reset_bootfs = 0;
+ uint64_t objnum;
+
+ elem = NULL;
+ while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
+ zpool_prop_t prop;
+ char *propname, *strval;
+ uint64_t intval;
+ objset_t *os;
+ char *slash;
+
+ propname = nvpair_name(elem);
+
+ if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL)
+ return (EINVAL);
+
+ switch (prop) {
+ case ZPOOL_PROP_VERSION:
+ error = nvpair_value_uint64(elem, &intval);
+ if (!error &&
+ (intval < spa_version(spa) || intval > SPA_VERSION))
+ error = EINVAL;
+ break;
+
+ case ZPOOL_PROP_DELEGATION:
+ case ZPOOL_PROP_AUTOREPLACE:
+ case ZPOOL_PROP_LISTSNAPS:
+ error = nvpair_value_uint64(elem, &intval);
+ if (!error && intval > 1)
+ error = EINVAL;
+ break;
+
+ case ZPOOL_PROP_BOOTFS:
+ if (spa_version(spa) < SPA_VERSION_BOOTFS) {
+ error = ENOTSUP;
+ break;
+ }
+
+ /*
+ * Make sure the vdev config is bootable
+ */
+ if (!vdev_is_bootable(spa->spa_root_vdev)) {
+ error = ENOTSUP;
+ break;
+ }
+
+ reset_bootfs = 1;
+
+ error = nvpair_value_string(elem, &strval);
+
+ if (!error) {
+ uint64_t compress;
+
+ if (strval == NULL || strval[0] == '\0') {
+ objnum = zpool_prop_default_numeric(
+ ZPOOL_PROP_BOOTFS);
+ break;
+ }
+
+ if (error = dmu_objset_open(strval, DMU_OST_ZFS,
+ DS_MODE_USER | DS_MODE_READONLY, &os))
+ break;
+
+ /* We don't support gzip bootable datasets */
+ if ((error = dsl_prop_get_integer(strval,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION),
+ &compress, NULL)) == 0 &&
+ !BOOTFS_COMPRESS_VALID(compress)) {
+ error = ENOTSUP;
+ } else {
+ objnum = dmu_objset_id(os);
+ }
+ dmu_objset_close(os);
+ }
+ break;
+
+ case ZPOOL_PROP_FAILUREMODE:
+ error = nvpair_value_uint64(elem, &intval);
+ if (!error && (intval < ZIO_FAILURE_MODE_WAIT ||
+ intval > ZIO_FAILURE_MODE_PANIC))
+ error = EINVAL;
+
+ /*
+ * This is a special case which only occurs when
+ * the pool has completely failed. This allows
+ * the user to change the in-core failmode property
+ * without syncing it out to disk (I/Os might
+ * currently be blocked). We do this by returning
+ * EIO to the caller (spa_prop_set) to trick it
+ * into thinking we encountered a property validation
+ * error.
+ */
+ if (!error && spa_suspended(spa)) {
+ spa->spa_failmode = intval;
+ error = EIO;
+ }
+ break;
+
+ case ZPOOL_PROP_CACHEFILE:
+ if ((error = nvpair_value_string(elem, &strval)) != 0)
+ break;
+
+ if (strval[0] == '\0')
+ break;
+
+ if (strcmp(strval, "none") == 0)
+ break;
+
+ if (strval[0] != '/') {
+ error = EINVAL;
+ break;
+ }
+
+ slash = strrchr(strval, '/');
+ ASSERT(slash != NULL);
+
+ if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
+ strcmp(slash, "/..") == 0)
+ error = EINVAL;
+ break;
+ }
+
+ if (error)
+ break;
+ }
+
+ if (!error && reset_bootfs) {
+ error = nvlist_remove(props,
+ zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING);
+
+ if (!error) {
+ error = nvlist_add_uint64(props,
+ zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum);
+ }
+ }
+
+ return (error);
+}
+
+int
+spa_prop_set(spa_t *spa, nvlist_t *nvp)
+{
+ int error;
+
+ if ((error = spa_prop_validate(spa, nvp)) != 0)
+ return (error);
+
+ return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props,
+ spa, nvp, 3));
+}
+
+/*
+ * If the bootfs property value is dsobj, clear it.
+ */
+void
+spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx)
+{
+ if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) {
+ VERIFY(zap_remove(spa->spa_meta_objset,
+ spa->spa_pool_props_object,
+ zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0);
+ spa->spa_bootfs = 0;
+ }
+}
+
+/*
+ * ==========================================================================
+ * SPA state manipulation (open/create/destroy/import/export)
+ * ==========================================================================
+ */
+
+static int
+spa_error_entry_compare(const void *a, const void *b)
+{
+ spa_error_entry_t *sa = (spa_error_entry_t *)a;
+ spa_error_entry_t *sb = (spa_error_entry_t *)b;
+ int ret;
+
+ ret = bcmp(&sa->se_bookmark, &sb->se_bookmark,
+ sizeof (zbookmark_t));
+
+ if (ret < 0)
+ return (-1);
+ else if (ret > 0)
+ return (1);
+ else
+ return (0);
+}
+
+/*
+ * Utility function which retrieves copies of the current logs and
+ * re-initializes them in the process.
+ */
+void
+spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub)
+{
+ ASSERT(MUTEX_HELD(&spa->spa_errlist_lock));
+
+ bcopy(&spa->spa_errlist_last, last, sizeof (avl_tree_t));
+ bcopy(&spa->spa_errlist_scrub, scrub, sizeof (avl_tree_t));
+
+ avl_create(&spa->spa_errlist_scrub,
+ spa_error_entry_compare, sizeof (spa_error_entry_t),
+ offsetof(spa_error_entry_t, se_avl));
+ avl_create(&spa->spa_errlist_last,
+ spa_error_entry_compare, sizeof (spa_error_entry_t),
+ offsetof(spa_error_entry_t, se_avl));
+}
+
+/*
+ * Activate an uninitialized pool.
+ */
+static void
+spa_activate(spa_t *spa)
+{
+ ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
+
+ spa->spa_state = POOL_STATE_ACTIVE;
+
+ spa->spa_normal_class = metaslab_class_create();
+ spa->spa_log_class = metaslab_class_create();
+
+ for (int t = 0; t < ZIO_TYPES; t++) {
+ for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
+ spa->spa_zio_taskq[t][q] = taskq_create("spa_zio",
+ zio_taskq_threads[t][q], maxclsyspri, 50,
+ INT_MAX, TASKQ_PREPOPULATE);
+ }
+ }
+
+ list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
+ offsetof(vdev_t, vdev_config_dirty_node));
+ list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
+ offsetof(vdev_t, vdev_state_dirty_node));
+
+ txg_list_create(&spa->spa_vdev_txg_list,
+ offsetof(struct vdev, vdev_txg_node));
+
+ avl_create(&spa->spa_errlist_scrub,
+ spa_error_entry_compare, sizeof (spa_error_entry_t),
+ offsetof(spa_error_entry_t, se_avl));
+ avl_create(&spa->spa_errlist_last,
+ spa_error_entry_compare, sizeof (spa_error_entry_t),
+ offsetof(spa_error_entry_t, se_avl));
+}
+
+/*
+ * Opposite of spa_activate().
+ */
+static void
+spa_deactivate(spa_t *spa)
+{
+ ASSERT(spa->spa_sync_on == B_FALSE);
+ ASSERT(spa->spa_dsl_pool == NULL);
+ ASSERT(spa->spa_root_vdev == NULL);
+
+ ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED);
+
+ txg_list_destroy(&spa->spa_vdev_txg_list);
+
+ list_destroy(&spa->spa_config_dirty_list);
+ list_destroy(&spa->spa_state_dirty_list);
+
+ for (int t = 0; t < ZIO_TYPES; t++) {
+ for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
+ taskq_destroy(spa->spa_zio_taskq[t][q]);
+ spa->spa_zio_taskq[t][q] = NULL;
+ }
+ }
+
+ metaslab_class_destroy(spa->spa_normal_class);
+ spa->spa_normal_class = NULL;
+
+ metaslab_class_destroy(spa->spa_log_class);
+ spa->spa_log_class = NULL;
+
+ /*
+ * If this was part of an import or the open otherwise failed, we may
+ * still have errors left in the queues. Empty them just in case.
+ */
+ spa_errlog_drain(spa);
+
+ avl_destroy(&spa->spa_errlist_scrub);
+ avl_destroy(&spa->spa_errlist_last);
+
+ spa->spa_state = POOL_STATE_UNINITIALIZED;
+}
+
+/*
+ * Verify a pool configuration, and construct the vdev tree appropriately. This
+ * will create all the necessary vdevs in the appropriate layout, with each vdev
+ * in the CLOSED state. This will prep the pool before open/creation/import.
+ * All vdev validation is done by the vdev_alloc() routine.
+ */
+static int
+spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent,
+ uint_t id, int atype)
+{
+ nvlist_t **child;
+ uint_t c, children;
+ int error;
+
+ if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0)
+ return (error);
+
+ if ((*vdp)->vdev_ops->vdev_op_leaf)
+ return (0);
+
+ error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+ &child, &children);
+
+ if (error == ENOENT)
+ return (0);
+
+ if (error) {
+ vdev_free(*vdp);
+ *vdp = NULL;
+ return (EINVAL);
+ }
+
+ for (c = 0; c < children; c++) {
+ vdev_t *vd;
+ if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c,
+ atype)) != 0) {
+ vdev_free(*vdp);
+ *vdp = NULL;
+ return (error);
+ }
+ }
+
+ ASSERT(*vdp != NULL);
+
+ return (0);
+}
+
+/*
+ * Opposite of spa_load().
+ */
+static void
+spa_unload(spa_t *spa)
+{
+ int i;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ /*
+ * Stop async tasks.
+ */
+ spa_async_suspend(spa);
+
+ /*
+ * Stop syncing.
+ */
+ if (spa->spa_sync_on) {
+ txg_sync_stop(spa->spa_dsl_pool);
+ spa->spa_sync_on = B_FALSE;
+ }
+
+ /*
+ * Wait for any outstanding async I/O to complete.
+ */
+ mutex_enter(&spa->spa_async_root_lock);
+ while (spa->spa_async_root_count != 0)
+ cv_wait(&spa->spa_async_root_cv, &spa->spa_async_root_lock);
+ mutex_exit(&spa->spa_async_root_lock);
+
+ /*
+ * Drop and purge level 2 cache
+ */
+ spa_l2cache_drop(spa);
+
+ /*
+ * Close the dsl pool.
+ */
+ if (spa->spa_dsl_pool) {
+ dsl_pool_close(spa->spa_dsl_pool);
+ spa->spa_dsl_pool = NULL;
+ }
+
+ /*
+ * Close all vdevs.
+ */
+ if (spa->spa_root_vdev)
+ vdev_free(spa->spa_root_vdev);
+ ASSERT(spa->spa_root_vdev == NULL);
+
+ for (i = 0; i < spa->spa_spares.sav_count; i++)
+ vdev_free(spa->spa_spares.sav_vdevs[i]);
+ if (spa->spa_spares.sav_vdevs) {
+ kmem_free(spa->spa_spares.sav_vdevs,
+ spa->spa_spares.sav_count * sizeof (void *));
+ spa->spa_spares.sav_vdevs = NULL;
+ }
+ if (spa->spa_spares.sav_config) {
+ nvlist_free(spa->spa_spares.sav_config);
+ spa->spa_spares.sav_config = NULL;
+ }
+ spa->spa_spares.sav_count = 0;
+
+ for (i = 0; i < spa->spa_l2cache.sav_count; i++)
+ vdev_free(spa->spa_l2cache.sav_vdevs[i]);
+ if (spa->spa_l2cache.sav_vdevs) {
+ kmem_free(spa->spa_l2cache.sav_vdevs,
+ spa->spa_l2cache.sav_count * sizeof (void *));
+ spa->spa_l2cache.sav_vdevs = NULL;
+ }
+ if (spa->spa_l2cache.sav_config) {
+ nvlist_free(spa->spa_l2cache.sav_config);
+ spa->spa_l2cache.sav_config = NULL;
+ }
+ spa->spa_l2cache.sav_count = 0;
+
+ spa->spa_async_suspended = 0;
+}
+
+/*
+ * Load (or re-load) the current list of vdevs describing the active spares for
+ * this pool. When this is called, we have some form of basic information in
+ * 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and
+ * then re-generate a more complete list including status information.
+ */
+static void
+spa_load_spares(spa_t *spa)
+{
+ nvlist_t **spares;
+ uint_t nspares;
+ int i;
+ vdev_t *vd, *tvd;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ /*
+ * First, close and free any existing spare vdevs.
+ */
+ for (i = 0; i < spa->spa_spares.sav_count; i++) {
+ vd = spa->spa_spares.sav_vdevs[i];
+
+ /* Undo the call to spa_activate() below */
+ if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
+ B_FALSE)) != NULL && tvd->vdev_isspare)
+ spa_spare_remove(tvd);
+ vdev_close(vd);
+ vdev_free(vd);
+ }
+
+ if (spa->spa_spares.sav_vdevs)
+ kmem_free(spa->spa_spares.sav_vdevs,
+ spa->spa_spares.sav_count * sizeof (void *));
+
+ if (spa->spa_spares.sav_config == NULL)
+ nspares = 0;
+ else
+ VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0);
+
+ spa->spa_spares.sav_count = (int)nspares;
+ spa->spa_spares.sav_vdevs = NULL;
+
+ if (nspares == 0)
+ return;
+
+ /*
+ * Construct the array of vdevs, opening them to get status in the
+ * process. For each spare, there is potentially two different vdev_t
+ * structures associated with it: one in the list of spares (used only
+ * for basic validation purposes) and one in the active vdev
+ * configuration (if it's spared in). During this phase we open and
+ * validate each vdev on the spare list. If the vdev also exists in the
+ * active configuration, then we also mark this vdev as an active spare.
+ */
+ spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *),
+ KM_SLEEP);
+ for (i = 0; i < spa->spa_spares.sav_count; i++) {
+ VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0,
+ VDEV_ALLOC_SPARE) == 0);
+ ASSERT(vd != NULL);
+
+ spa->spa_spares.sav_vdevs[i] = vd;
+
+ if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
+ B_FALSE)) != NULL) {
+ if (!tvd->vdev_isspare)
+ spa_spare_add(tvd);
+
+ /*
+ * We only mark the spare active if we were successfully
+ * able to load the vdev. Otherwise, importing a pool
+ * with a bad active spare would result in strange
+ * behavior, because multiple pool would think the spare
+ * is actively in use.
+ *
+ * There is a vulnerability here to an equally bizarre
+ * circumstance, where a dead active spare is later
+ * brought back to life (onlined or otherwise). Given
+ * the rarity of this scenario, and the extra complexity
+ * it adds, we ignore the possibility.
+ */
+ if (!vdev_is_dead(tvd))
+ spa_spare_activate(tvd);
+ }
+
+ vd->vdev_top = vd;
+
+ if (vdev_open(vd) != 0)
+ continue;
+
+ if (vdev_validate_aux(vd) == 0)
+ spa_spare_add(vd);
+ }
+
+ /*
+ * Recompute the stashed list of spares, with status information
+ * this time.
+ */
+ VERIFY(nvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES,
+ DATA_TYPE_NVLIST_ARRAY) == 0);
+
+ spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *),
+ KM_SLEEP);
+ for (i = 0; i < spa->spa_spares.sav_count; i++)
+ spares[i] = vdev_config_generate(spa,
+ spa->spa_spares.sav_vdevs[i], B_TRUE, B_TRUE, B_FALSE);
+ VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, spares, spa->spa_spares.sav_count) == 0);
+ for (i = 0; i < spa->spa_spares.sav_count; i++)
+ nvlist_free(spares[i]);
+ kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *));
+}
+
+/*
+ * Load (or re-load) the current list of vdevs describing the active l2cache for
+ * this pool. When this is called, we have some form of basic information in
+ * 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and
+ * then re-generate a more complete list including status information.
+ * Devices which are already active have their details maintained, and are
+ * not re-opened.
+ */
+static void
+spa_load_l2cache(spa_t *spa)
+{
+ nvlist_t **l2cache;
+ uint_t nl2cache;
+ int i, j, oldnvdevs;
+ uint64_t guid, size;
+ vdev_t *vd, **oldvdevs, **newvdevs;
+ spa_aux_vdev_t *sav = &spa->spa_l2cache;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ if (sav->sav_config != NULL) {
+ VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
+ ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
+ newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);
+ } else {
+ nl2cache = 0;
+ }
+
+ oldvdevs = sav->sav_vdevs;
+ oldnvdevs = sav->sav_count;
+ sav->sav_vdevs = NULL;
+ sav->sav_count = 0;
+
+ /*
+ * Process new nvlist of vdevs.
+ */
+ for (i = 0; i < nl2cache; i++) {
+ VERIFY(nvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID,
+ &guid) == 0);
+
+ newvdevs[i] = NULL;
+ for (j = 0; j < oldnvdevs; j++) {
+ vd = oldvdevs[j];
+ if (vd != NULL && guid == vd->vdev_guid) {
+ /*
+ * Retain previous vdev for add/remove ops.
+ */
+ newvdevs[i] = vd;
+ oldvdevs[j] = NULL;
+ break;
+ }
+ }
+
+ if (newvdevs[i] == NULL) {
+ /*
+ * Create new vdev
+ */
+ VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0,
+ VDEV_ALLOC_L2CACHE) == 0);
+ ASSERT(vd != NULL);
+ newvdevs[i] = vd;
+
+ /*
+ * Commit this vdev as an l2cache device,
+ * even if it fails to open.
+ */
+ spa_l2cache_add(vd);
+
+ vd->vdev_top = vd;
+ vd->vdev_aux = sav;
+
+ spa_l2cache_activate(vd);
+
+ if (vdev_open(vd) != 0)
+ continue;
+
+ (void) vdev_validate_aux(vd);
+
+ if (!vdev_is_dead(vd)) {
+ size = vdev_get_rsize(vd);
+ l2arc_add_vdev(spa, vd,
+ VDEV_LABEL_START_SIZE,
+ size - VDEV_LABEL_START_SIZE);
+ }
+ }
+ }
+
+ /*
+ * Purge vdevs that were dropped
+ */
+ for (i = 0; i < oldnvdevs; i++) {
+ uint64_t pool;
+
+ vd = oldvdevs[i];
+ if (vd != NULL) {
+ if ((spa_mode & FWRITE) &&
+ spa_l2cache_exists(vd->vdev_guid, &pool) &&
+ pool != 0ULL &&
+ l2arc_vdev_present(vd)) {
+ l2arc_remove_vdev(vd);
+ }
+ (void) vdev_close(vd);
+ spa_l2cache_remove(vd);
+ }
+ }
+
+ if (oldvdevs)
+ kmem_free(oldvdevs, oldnvdevs * sizeof (void *));
+
+ if (sav->sav_config == NULL)
+ goto out;
+
+ sav->sav_vdevs = newvdevs;
+ sav->sav_count = (int)nl2cache;
+
+ /*
+ * Recompute the stashed list of l2cache devices, with status
+ * information this time.
+ */
+ VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
+ DATA_TYPE_NVLIST_ARRAY) == 0);
+
+ l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP);
+ for (i = 0; i < sav->sav_count; i++)
+ l2cache[i] = vdev_config_generate(spa,
+ sav->sav_vdevs[i], B_TRUE, B_FALSE, B_TRUE);
+ VERIFY(nvlist_add_nvlist_array(sav->sav_config,
+ ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0);
+out:
+ for (i = 0; i < sav->sav_count; i++)
+ nvlist_free(l2cache[i]);
+ if (sav->sav_count)
+ kmem_free(l2cache, sav->sav_count * sizeof (void *));
+}
+
+static int
+load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value)
+{
+ dmu_buf_t *db;
+ char *packed = NULL;
+ size_t nvsize = 0;
+ int error;
+ *value = NULL;
+
+ VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
+ nvsize = *(uint64_t *)db->db_data;
+ dmu_buf_rele(db, FTAG);
+
+ packed = kmem_alloc(nvsize, KM_SLEEP);
+ error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed);
+ if (error == 0)
+ error = nvlist_unpack(packed, nvsize, value, 0);
+ kmem_free(packed, nvsize);
+
+ return (error);
+}
+
+/*
+ * Checks to see if the given vdev could not be opened, in which case we post a
+ * sysevent to notify the autoreplace code that the device has been removed.
+ */
+static void
+spa_check_removed(vdev_t *vd)
+{
+ int c;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ spa_check_removed(vd->vdev_child[c]);
+
+ if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd)) {
+ zfs_post_autoreplace(vd->vdev_spa, vd);
+ spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_CHECK);
+ }
+}
+
+/*
+ * Check for missing log devices
+ */
+int
+spa_check_logs(spa_t *spa)
+{
+ switch (spa->spa_log_state) {
+ case SPA_LOG_MISSING:
+ /* need to recheck in case slog has been restored */
+ case SPA_LOG_UNKNOWN:
+ if (dmu_objset_find(spa->spa_name, zil_check_log_chain, NULL,
+ DS_FIND_CHILDREN)) {
+ spa->spa_log_state = SPA_LOG_MISSING;
+ return (1);
+ }
+ break;
+
+ case SPA_LOG_CLEAR:
+ (void) dmu_objset_find(spa->spa_name, zil_clear_log_chain, NULL,
+ DS_FIND_CHILDREN);
+ break;
+ }
+ spa->spa_log_state = SPA_LOG_GOOD;
+ return (0);
+}
+
+/*
+ * Load an existing storage pool, using the pool's builtin spa_config as a
+ * source of configuration information.
+ */
+static int
+spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig)
+{
+ int error = 0;
+ nvlist_t *nvroot = NULL;
+ vdev_t *rvd;
+ uberblock_t *ub = &spa->spa_uberblock;
+ uint64_t config_cache_txg = spa->spa_config_txg;
+ uint64_t pool_guid;
+ uint64_t version;
+ uint64_t autoreplace = 0;
+ char *ereport = FM_EREPORT_ZFS_POOL;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ spa->spa_load_state = state;
+
+ if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) ||
+ nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) {
+ error = EINVAL;
+ goto out;
+ }
+
+ /*
+ * Versioning wasn't explicitly added to the label until later, so if
+ * it's not present treat it as the initial version.
+ */
+ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) != 0)
+ version = SPA_VERSION_INITIAL;
+
+ (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
+ &spa->spa_config_txg);
+
+ if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) &&
+ spa_guid_exists(pool_guid, 0)) {
+ error = EEXIST;
+ goto out;
+ }
+
+ spa->spa_load_guid = pool_guid;
+
+ /*
+ * Parse the configuration into a vdev tree. We explicitly set the
+ * value that will be returned by spa_version() since parsing the
+ * configuration requires knowing the version number.
+ */
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa->spa_ubsync.ub_version = version;
+ error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_LOAD);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ if (error != 0)
+ goto out;
+
+ ASSERT(spa->spa_root_vdev == rvd);
+ ASSERT(spa_guid(spa) == pool_guid);
+
+ /*
+ * Try to open all vdevs, loading each label in the process.
+ */
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ error = vdev_open(rvd);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ if (error != 0)
+ goto out;
+
+ /*
+ * Validate the labels for all leaf vdevs. We need to grab the config
+ * lock because all label I/O is done with ZIO_FLAG_CONFIG_WRITER.
+ */
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ error = vdev_validate(rvd);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ if (error != 0)
+ goto out;
+
+ if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
+ error = ENXIO;
+ goto out;
+ }
+
+ /*
+ * Find the best uberblock.
+ */
+ vdev_uberblock_load(NULL, rvd, ub);
+
+ /*
+ * If we weren't able to find a single valid uberblock, return failure.
+ */
+ if (ub->ub_txg == 0) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = ENXIO;
+ goto out;
+ }
+
+ /*
+ * If the pool is newer than the code, we can't open it.
+ */
+ if (ub->ub_version > SPA_VERSION) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_VERSION_NEWER);
+ error = ENOTSUP;
+ goto out;
+ }
+
+ /*
+ * If the vdev guid sum doesn't match the uberblock, we have an
+ * incomplete configuration.
+ */
+ if (rvd->vdev_guid_sum != ub->ub_guid_sum && mosconfig) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_BAD_GUID_SUM);
+ error = ENXIO;
+ goto out;
+ }
+
+ /*
+ * Initialize internal SPA structures.
+ */
+ spa->spa_state = POOL_STATE_ACTIVE;
+ spa->spa_ubsync = spa->spa_uberblock;
+ spa->spa_first_txg = spa_last_synced_txg(spa) + 1;
+ error = dsl_pool_open(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
+ if (error) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ goto out;
+ }
+ spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;
+
+ if (zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
+ sizeof (uint64_t), 1, &spa->spa_config_object) != 0) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ if (!mosconfig) {
+ nvlist_t *newconfig;
+ uint64_t hostid;
+
+ if (load_nvlist(spa, spa->spa_config_object, &newconfig) != 0) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ if (!spa_is_root(spa) && nvlist_lookup_uint64(newconfig,
+ ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
+ char *hostname;
+ unsigned long myhostid = 0;
+
+ VERIFY(nvlist_lookup_string(newconfig,
+ ZPOOL_CONFIG_HOSTNAME, &hostname) == 0);
+
+ (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid);
+ if (hostid != 0 && myhostid != 0 &&
+ (unsigned long)hostid != myhostid) {
+ cmn_err(CE_WARN, "pool '%s' could not be "
+ "loaded as it was last accessed by "
+ "another system (host: %s hostid: 0x%lx). "
+ "See: http://www.sun.com/msg/ZFS-8000-EY",
+ spa_name(spa), hostname,
+ (unsigned long)hostid);
+ error = EBADF;
+ goto out;
+ }
+ }
+
+ spa_config_set(spa, newconfig);
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_activate(spa);
+
+ return (spa_load(spa, newconfig, state, B_TRUE));
+ }
+
+ if (zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPLIST,
+ sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj) != 0) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ /*
+ * Load the bit that tells us to use the new accounting function
+ * (raid-z deflation). If we have an older pool, this will not
+ * be present.
+ */
+ error = zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
+ sizeof (uint64_t), 1, &spa->spa_deflate);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ /*
+ * Load the persistent error log. If we have an older pool, this will
+ * not be present.
+ */
+ error = zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_LAST,
+ sizeof (uint64_t), 1, &spa->spa_errlog_last);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ error = zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_SCRUB,
+ sizeof (uint64_t), 1, &spa->spa_errlog_scrub);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ /*
+ * Load the history object. If we have an older pool, this
+ * will not be present.
+ */
+ error = zap_lookup(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_HISTORY,
+ sizeof (uint64_t), 1, &spa->spa_history);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ /*
+ * Load any hot spares for this pool.
+ */
+ error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_SPARES, sizeof (uint64_t), 1, &spa->spa_spares.sav_object);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+ if (error == 0) {
+ ASSERT(spa_version(spa) >= SPA_VERSION_SPARES);
+ if (load_nvlist(spa, spa->spa_spares.sav_object,
+ &spa->spa_spares.sav_config) != 0) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_spares(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+
+ /*
+ * Load any level 2 ARC devices for this pool.
+ */
+ error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_L2CACHE, sizeof (uint64_t), 1,
+ &spa->spa_l2cache.sav_object);
+ if (error != 0 && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+ if (error == 0) {
+ ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE);
+ if (load_nvlist(spa, spa->spa_l2cache.sav_object,
+ &spa->spa_l2cache.sav_config) != 0) {
+ vdev_set_state(rvd, B_TRUE,
+ VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_l2cache(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+
+ if (spa_check_logs(spa)) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_BAD_LOG);
+ error = ENXIO;
+ ereport = FM_EREPORT_ZFS_LOG_REPLAY;
+ goto out;
+ }
+
+
+ spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
+
+ error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_PROPS, sizeof (uint64_t), 1, &spa->spa_pool_props_object);
+
+ if (error && error != ENOENT) {
+ vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ error = EIO;
+ goto out;
+ }
+
+ if (error == 0) {
+ (void) zap_lookup(spa->spa_meta_objset,
+ spa->spa_pool_props_object,
+ zpool_prop_to_name(ZPOOL_PROP_BOOTFS),
+ sizeof (uint64_t), 1, &spa->spa_bootfs);
+ (void) zap_lookup(spa->spa_meta_objset,
+ spa->spa_pool_props_object,
+ zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE),
+ sizeof (uint64_t), 1, &autoreplace);
+ (void) zap_lookup(spa->spa_meta_objset,
+ spa->spa_pool_props_object,
+ zpool_prop_to_name(ZPOOL_PROP_DELEGATION),
+ sizeof (uint64_t), 1, &spa->spa_delegation);
+ (void) zap_lookup(spa->spa_meta_objset,
+ spa->spa_pool_props_object,
+ zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE),
+ sizeof (uint64_t), 1, &spa->spa_failmode);
+ }
+
+ /*
+ * If the 'autoreplace' property is set, then post a resource notifying
+ * the ZFS DE that it should not issue any faults for unopenable
+ * devices. We also iterate over the vdevs, and post a sysevent for any
+ * unopenable vdevs so that the normal autoreplace handler can take
+ * over.
+ */
+ if (autoreplace && state != SPA_LOAD_TRYIMPORT)
+ spa_check_removed(spa->spa_root_vdev);
+
+ /*
+ * Load the vdev state for all toplevel vdevs.
+ */
+ vdev_load(rvd);
+
+ /*
+ * Propagate the leaf DTLs we just loaded all the way up the tree.
+ */
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ vdev_dtl_reassess(rvd, 0, 0, B_FALSE);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ /*
+ * Check the state of the root vdev. If it can't be opened, it
+ * indicates one or more toplevel vdevs are faulted.
+ */
+ if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
+ error = ENXIO;
+ goto out;
+ }
+
+ if ((spa_mode & FWRITE) && state != SPA_LOAD_TRYIMPORT) {
+ dmu_tx_t *tx;
+ int need_update = B_FALSE;
+ int c;
+
+ /*
+ * Claim log blocks that haven't been committed yet.
+ * This must all happen in a single txg.
+ */
+ tx = dmu_tx_create_assigned(spa_get_dsl(spa),
+ spa_first_txg(spa));
+ (void) dmu_objset_find(spa_name(spa),
+ zil_claim, tx, DS_FIND_CHILDREN);
+ dmu_tx_commit(tx);
+
+ spa->spa_sync_on = B_TRUE;
+ txg_sync_start(spa->spa_dsl_pool);
+
+ /*
+ * Wait for all claims to sync.
+ */
+ txg_wait_synced(spa->spa_dsl_pool, 0);
+
+ /*
+ * If the config cache is stale, or we have uninitialized
+ * metaslabs (see spa_vdev_add()), then update the config.
+ */
+ if (config_cache_txg != spa->spa_config_txg ||
+ state == SPA_LOAD_IMPORT)
+ need_update = B_TRUE;
+
+ for (c = 0; c < rvd->vdev_children; c++)
+ if (rvd->vdev_child[c]->vdev_ms_array == 0)
+ need_update = B_TRUE;
+
+ /*
+ * Update the config cache asychronously in case we're the
+ * root pool, in which case the config cache isn't writable yet.
+ */
+ if (need_update)
+ spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
+ }
+
+ error = 0;
+out:
+ spa->spa_minref = refcount_count(&spa->spa_refcount);
+ if (error && error != EBADF)
+ zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0);
+ spa->spa_load_state = SPA_LOAD_NONE;
+ spa->spa_ena = 0;
+
+ return (error);
+}
+
+/*
+ * Pool Open/Import
+ *
+ * The import case is identical to an open except that the configuration is sent
+ * down from userland, instead of grabbed from the configuration cache. For the
+ * case of an open, the pool configuration will exist in the
+ * POOL_STATE_UNINITIALIZED state.
+ *
+ * The stats information (gen/count/ustats) is used to gather vdev statistics at
+ * the same time open the pool, without having to keep around the spa_t in some
+ * ambiguous state.
+ */
+static int
+spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config)
+{
+ spa_t *spa;
+ int error;
+ int locked = B_FALSE;
+
+ *spapp = NULL;
+
+ /*
+ * As disgusting as this is, we need to support recursive calls to this
+ * function because dsl_dir_open() is called during spa_load(), and ends
+ * up calling spa_open() again. The real fix is to figure out how to
+ * avoid dsl_dir_open() calling this in the first place.
+ */
+ if (mutex_owner(&spa_namespace_lock) != curthread) {
+ mutex_enter(&spa_namespace_lock);
+ locked = B_TRUE;
+ }
+
+ if ((spa = spa_lookup(pool)) == NULL) {
+ if (locked)
+ mutex_exit(&spa_namespace_lock);
+ return (ENOENT);
+ }
+ if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
+
+ spa_activate(spa);
+
+ error = spa_load(spa, spa->spa_config, SPA_LOAD_OPEN, B_FALSE);
+
+ if (error == EBADF) {
+ /*
+ * If vdev_validate() returns failure (indicated by
+ * EBADF), it indicates that one of the vdevs indicates
+ * that the pool has been exported or destroyed. If
+ * this is the case, the config cache is out of sync and
+ * we should remove the pool from the namespace.
+ */
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_config_sync(spa, B_TRUE, B_TRUE);
+ spa_remove(spa);
+ if (locked)
+ mutex_exit(&spa_namespace_lock);
+ return (ENOENT);
+ }
+
+ if (error) {
+ /*
+ * We can't open the pool, but we still have useful
+ * information: the state of each vdev after the
+ * attempted vdev_open(). Return this to the user.
+ */
+ if (config != NULL && spa->spa_root_vdev != NULL)
+ *config = spa_config_generate(spa, NULL, -1ULL,
+ B_TRUE);
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa->spa_last_open_failed = B_TRUE;
+ if (locked)
+ mutex_exit(&spa_namespace_lock);
+ *spapp = NULL;
+ return (error);
+ } else {
+ spa->spa_last_open_failed = B_FALSE;
+ }
+ }
+
+ spa_open_ref(spa, tag);
+
+ if (locked)
+ mutex_exit(&spa_namespace_lock);
+
+ *spapp = spa;
+
+ if (config != NULL)
+ *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
+
+ return (0);
+}
+
+int
+spa_open(const char *name, spa_t **spapp, void *tag)
+{
+ return (spa_open_common(name, spapp, tag, NULL));
+}
+
+/*
+ * Lookup the given spa_t, incrementing the inject count in the process,
+ * preventing it from being exported or destroyed.
+ */
+spa_t *
+spa_inject_addref(char *name)
+{
+ spa_t *spa;
+
+ mutex_enter(&spa_namespace_lock);
+ if ((spa = spa_lookup(name)) == NULL) {
+ mutex_exit(&spa_namespace_lock);
+ return (NULL);
+ }
+ spa->spa_inject_ref++;
+ mutex_exit(&spa_namespace_lock);
+
+ return (spa);
+}
+
+void
+spa_inject_delref(spa_t *spa)
+{
+ mutex_enter(&spa_namespace_lock);
+ spa->spa_inject_ref--;
+ mutex_exit(&spa_namespace_lock);
+}
+
+/*
+ * Add spares device information to the nvlist.
+ */
+static void
+spa_add_spares(spa_t *spa, nvlist_t *config)
+{
+ nvlist_t **spares;
+ uint_t i, nspares;
+ nvlist_t *nvroot;
+ uint64_t guid;
+ vdev_stat_t *vs;
+ uint_t vsc;
+ uint64_t pool;
+
+ if (spa->spa_spares.sav_count == 0)
+ return;
+
+ VERIFY(nvlist_lookup_nvlist(config,
+ ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
+ VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0);
+ if (nspares != 0) {
+ VERIFY(nvlist_add_nvlist_array(nvroot,
+ ZPOOL_CONFIG_SPARES, spares, nspares) == 0);
+ VERIFY(nvlist_lookup_nvlist_array(nvroot,
+ ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0);
+
+ /*
+ * Go through and find any spares which have since been
+ * repurposed as an active spare. If this is the case, update
+ * their status appropriately.
+ */
+ for (i = 0; i < nspares; i++) {
+ VERIFY(nvlist_lookup_uint64(spares[i],
+ ZPOOL_CONFIG_GUID, &guid) == 0);
+ if (spa_spare_exists(guid, &pool, NULL) &&
+ pool != 0ULL) {
+ VERIFY(nvlist_lookup_uint64_array(
+ spares[i], ZPOOL_CONFIG_STATS,
+ (uint64_t **)&vs, &vsc) == 0);
+ vs->vs_state = VDEV_STATE_CANT_OPEN;
+ vs->vs_aux = VDEV_AUX_SPARED;
+ }
+ }
+ }
+}
+
+/*
+ * Add l2cache device information to the nvlist, including vdev stats.
+ */
+static void
+spa_add_l2cache(spa_t *spa, nvlist_t *config)
+{
+ nvlist_t **l2cache;
+ uint_t i, j, nl2cache;
+ nvlist_t *nvroot;
+ uint64_t guid;
+ vdev_t *vd;
+ vdev_stat_t *vs;
+ uint_t vsc;
+
+ if (spa->spa_l2cache.sav_count == 0)
+ return;
+
+ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
+
+ VERIFY(nvlist_lookup_nvlist(config,
+ ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
+ VERIFY(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
+ if (nl2cache != 0) {
+ VERIFY(nvlist_add_nvlist_array(nvroot,
+ ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0);
+ VERIFY(nvlist_lookup_nvlist_array(nvroot,
+ ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
+
+ /*
+ * Update level 2 cache device stats.
+ */
+
+ for (i = 0; i < nl2cache; i++) {
+ VERIFY(nvlist_lookup_uint64(l2cache[i],
+ ZPOOL_CONFIG_GUID, &guid) == 0);
+
+ vd = NULL;
+ for (j = 0; j < spa->spa_l2cache.sav_count; j++) {
+ if (guid ==
+ spa->spa_l2cache.sav_vdevs[j]->vdev_guid) {
+ vd = spa->spa_l2cache.sav_vdevs[j];
+ break;
+ }
+ }
+ ASSERT(vd != NULL);
+
+ VERIFY(nvlist_lookup_uint64_array(l2cache[i],
+ ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0);
+ vdev_get_stats(vd, vs);
+ }
+ }
+
+ spa_config_exit(spa, SCL_CONFIG, FTAG);
+}
+
+int
+spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen)
+{
+ int error;
+ spa_t *spa;
+
+ *config = NULL;
+ error = spa_open_common(name, &spa, FTAG, config);
+
+ if (spa && *config != NULL) {
+ VERIFY(nvlist_add_uint64(*config, ZPOOL_CONFIG_ERRCOUNT,
+ spa_get_errlog_size(spa)) == 0);
+
+ if (spa_suspended(spa))
+ VERIFY(nvlist_add_uint64(*config,
+ ZPOOL_CONFIG_SUSPENDED, spa->spa_failmode) == 0);
+
+ spa_add_spares(spa, *config);
+ spa_add_l2cache(spa, *config);
+ }
+
+ /*
+ * We want to get the alternate root even for faulted pools, so we cheat
+ * and call spa_lookup() directly.
+ */
+ if (altroot) {
+ if (spa == NULL) {
+ mutex_enter(&spa_namespace_lock);
+ spa = spa_lookup(name);
+ if (spa)
+ spa_altroot(spa, altroot, buflen);
+ else
+ altroot[0] = '\0';
+ spa = NULL;
+ mutex_exit(&spa_namespace_lock);
+ } else {
+ spa_altroot(spa, altroot, buflen);
+ }
+ }
+
+ if (spa != NULL)
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+/*
+ * Validate that the auxiliary device array is well formed. We must have an
+ * array of nvlists, each which describes a valid leaf vdev. If this is an
+ * import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be
+ * specified, as long as they are well-formed.
+ */
+static int
+spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode,
+ spa_aux_vdev_t *sav, const char *config, uint64_t version,
+ vdev_labeltype_t label)
+{
+ nvlist_t **dev;
+ uint_t i, ndev;
+ vdev_t *vd;
+ int error;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ /*
+ * It's acceptable to have no devs specified.
+ */
+ if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0)
+ return (0);
+
+ if (ndev == 0)
+ return (EINVAL);
+
+ /*
+ * Make sure the pool is formatted with a version that supports this
+ * device type.
+ */
+ if (spa_version(spa) < version)
+ return (ENOTSUP);
+
+ /*
+ * Set the pending device list so we correctly handle device in-use
+ * checking.
+ */
+ sav->sav_pending = dev;
+ sav->sav_npending = ndev;
+
+ for (i = 0; i < ndev; i++) {
+ if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0,
+ mode)) != 0)
+ goto out;
+
+ if (!vd->vdev_ops->vdev_op_leaf) {
+ vdev_free(vd);
+ error = EINVAL;
+ goto out;
+ }
+
+ /*
+ * The L2ARC currently only supports disk devices in
+ * kernel context. For user-level testing, we allow it.
+ */
+#ifdef _KERNEL
+ if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) &&
+ strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) {
+ error = ENOTBLK;
+ goto out;
+ }
+#endif
+ vd->vdev_top = vd;
+
+ if ((error = vdev_open(vd)) == 0 &&
+ (error = vdev_label_init(vd, crtxg, label)) == 0) {
+ VERIFY(nvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID,
+ vd->vdev_guid) == 0);
+ }
+
+ vdev_free(vd);
+
+ if (error &&
+ (mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE))
+ goto out;
+ else
+ error = 0;
+ }
+
+out:
+ sav->sav_pending = NULL;
+ sav->sav_npending = 0;
+ return (error);
+}
+
+static int
+spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode)
+{
+ int error;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode,
+ &spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES,
+ VDEV_LABEL_SPARE)) != 0) {
+ return (error);
+ }
+
+ return (spa_validate_aux_devs(spa, nvroot, crtxg, mode,
+ &spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE,
+ VDEV_LABEL_L2CACHE));
+}
+
+static void
+spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs,
+ const char *config)
+{
+ int i;
+
+ if (sav->sav_config != NULL) {
+ nvlist_t **olddevs;
+ uint_t oldndevs;
+ nvlist_t **newdevs;
+
+ /*
+ * Generate new dev list by concatentating with the
+ * current dev list.
+ */
+ VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config,
+ &olddevs, &oldndevs) == 0);
+
+ newdevs = kmem_alloc(sizeof (void *) *
+ (ndevs + oldndevs), KM_SLEEP);
+ for (i = 0; i < oldndevs; i++)
+ VERIFY(nvlist_dup(olddevs[i], &newdevs[i],
+ KM_SLEEP) == 0);
+ for (i = 0; i < ndevs; i++)
+ VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs],
+ KM_SLEEP) == 0);
+
+ VERIFY(nvlist_remove(sav->sav_config, config,
+ DATA_TYPE_NVLIST_ARRAY) == 0);
+
+ VERIFY(nvlist_add_nvlist_array(sav->sav_config,
+ config, newdevs, ndevs + oldndevs) == 0);
+ for (i = 0; i < oldndevs + ndevs; i++)
+ nvlist_free(newdevs[i]);
+ kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *));
+ } else {
+ /*
+ * Generate a new dev list.
+ */
+ VERIFY(nvlist_alloc(&sav->sav_config, NV_UNIQUE_NAME,
+ KM_SLEEP) == 0);
+ VERIFY(nvlist_add_nvlist_array(sav->sav_config, config,
+ devs, ndevs) == 0);
+ }
+}
+
+/*
+ * Stop and drop level 2 ARC devices
+ */
+void
+spa_l2cache_drop(spa_t *spa)
+{
+ vdev_t *vd;
+ int i;
+ spa_aux_vdev_t *sav = &spa->spa_l2cache;
+
+ for (i = 0; i < sav->sav_count; i++) {
+ uint64_t pool;
+
+ vd = sav->sav_vdevs[i];
+ ASSERT(vd != NULL);
+
+ if ((spa_mode & FWRITE) &&
+ spa_l2cache_exists(vd->vdev_guid, &pool) && pool != 0ULL &&
+ l2arc_vdev_present(vd)) {
+ l2arc_remove_vdev(vd);
+ }
+ if (vd->vdev_isl2cache)
+ spa_l2cache_remove(vd);
+ vdev_clear_stats(vd);
+ (void) vdev_close(vd);
+ }
+}
+
+/*
+ * Pool Creation
+ */
+int
+spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
+ const char *history_str, nvlist_t *zplprops)
+{
+ spa_t *spa;
+ char *altroot = NULL;
+ vdev_t *rvd;
+ dsl_pool_t *dp;
+ dmu_tx_t *tx;
+ int c, error = 0;
+ uint64_t txg = TXG_INITIAL;
+ nvlist_t **spares, **l2cache;
+ uint_t nspares, nl2cache;
+ uint64_t version;
+
+ /*
+ * If this pool already exists, return failure.
+ */
+ mutex_enter(&spa_namespace_lock);
+ if (spa_lookup(pool) != NULL) {
+ mutex_exit(&spa_namespace_lock);
+ return (EEXIST);
+ }
+
+ /*
+ * Allocate a new spa_t structure.
+ */
+ (void) nvlist_lookup_string(props,
+ zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
+ spa = spa_add(pool, altroot);
+ spa_activate(spa);
+
+ spa->spa_uberblock.ub_txg = txg - 1;
+
+ if (props && (error = spa_prop_validate(spa, props))) {
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_remove(spa);
+ mutex_exit(&spa_namespace_lock);
+ return (error);
+ }
+
+ if (nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION),
+ &version) != 0)
+ version = SPA_VERSION;
+ ASSERT(version <= SPA_VERSION);
+ spa->spa_uberblock.ub_version = version;
+ spa->spa_ubsync = spa->spa_uberblock;
+
+ /*
+ * Create the root vdev.
+ */
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+ error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD);
+
+ ASSERT(error != 0 || rvd != NULL);
+ ASSERT(error != 0 || spa->spa_root_vdev == rvd);
+
+ if (error == 0 && !zfs_allocatable_devs(nvroot))
+ error = EINVAL;
+
+ if (error == 0 &&
+ (error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
+ (error = spa_validate_aux(spa, nvroot, txg,
+ VDEV_ALLOC_ADD)) == 0) {
+ for (c = 0; c < rvd->vdev_children; c++)
+ vdev_init(rvd->vdev_child[c], txg);
+ vdev_config_dirty(rvd);
+ }
+
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ if (error != 0) {
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_remove(spa);
+ mutex_exit(&spa_namespace_lock);
+ return (error);
+ }
+
+ /*
+ * Get the list of spares, if specified.
+ */
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
+ &spares, &nspares) == 0) {
+ VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME,
+ KM_SLEEP) == 0);
+ VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, spares, nspares) == 0);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_spares(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ spa->spa_spares.sav_sync = B_TRUE;
+ }
+
+ /*
+ * Get the list of level 2 cache devices, if specified.
+ */
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
+ &l2cache, &nl2cache) == 0) {
+ VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_l2cache(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ spa->spa_l2cache.sav_sync = B_TRUE;
+ }
+
+ spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg);
+ spa->spa_meta_objset = dp->dp_meta_objset;
+
+ tx = dmu_tx_create_assigned(dp, txg);
+
+ /*
+ * Create the pool config object.
+ */
+ spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset,
+ DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE,
+ DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx);
+
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
+ sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add pool config");
+ }
+
+ /* Newly created pools with the right version are always deflated. */
+ if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
+ spa->spa_deflate = TRUE;
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
+ sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add deflate");
+ }
+ }
+
+ /*
+ * Create the deferred-free bplist object. Turn off compression
+ * because sync-to-convergence takes longer if the blocksize
+ * keeps changing.
+ */
+ spa->spa_sync_bplist_obj = bplist_create(spa->spa_meta_objset,
+ 1 << 14, tx);
+ dmu_object_set_compress(spa->spa_meta_objset, spa->spa_sync_bplist_obj,
+ ZIO_COMPRESS_OFF, tx);
+
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPLIST,
+ sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add bplist");
+ }
+
+ /*
+ * Create the pool's history object.
+ */
+ if (version >= SPA_VERSION_ZPOOL_HISTORY)
+ spa_history_create_obj(spa, tx);
+
+ /*
+ * Set pool properties.
+ */
+ spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
+ spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
+ spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
+ if (props)
+ spa_sync_props(spa, props, CRED(), tx);
+
+ dmu_tx_commit(tx);
+
+ spa->spa_sync_on = B_TRUE;
+ txg_sync_start(spa->spa_dsl_pool);
+
+ /*
+ * We explicitly wait for the first transaction to complete so that our
+ * bean counters are appropriately updated.
+ */
+ txg_wait_synced(spa->spa_dsl_pool, txg);
+
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+
+ if (version >= SPA_VERSION_ZPOOL_HISTORY && history_str != NULL)
+ (void) spa_history_log(spa, history_str, LOG_CMD_POOL_CREATE);
+
+ mutex_exit(&spa_namespace_lock);
+
+ spa->spa_minref = refcount_count(&spa->spa_refcount);
+
+ return (0);
+}
+
+/*
+ * Import the given pool into the system. We set up the necessary spa_t and
+ * then call spa_load() to do the dirty work.
+ */
+static int
+spa_import_common(const char *pool, nvlist_t *config, nvlist_t *props,
+ boolean_t isroot, boolean_t allowfaulted)
+{
+ spa_t *spa;
+ char *altroot = NULL;
+ int error, loaderr;
+ nvlist_t *nvroot;
+ nvlist_t **spares, **l2cache;
+ uint_t nspares, nl2cache;
+
+ /*
+ * If a pool with this name exists, return failure.
+ */
+ mutex_enter(&spa_namespace_lock);
+ if ((spa = spa_lookup(pool)) != NULL) {
+ if (isroot) {
+ /*
+ * Remove the existing root pool from the
+ * namespace so that we can replace it with
+ * the correct config we just read in.
+ */
+ ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
+ spa_remove(spa);
+ } else {
+ mutex_exit(&spa_namespace_lock);
+ return (EEXIST);
+ }
+ }
+
+ /*
+ * Create and initialize the spa structure.
+ */
+ (void) nvlist_lookup_string(props,
+ zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
+ spa = spa_add(pool, altroot);
+ spa_activate(spa);
+
+ if (allowfaulted)
+ spa->spa_import_faulted = B_TRUE;
+ spa->spa_is_root = isroot;
+
+ /*
+ * Pass off the heavy lifting to spa_load().
+ * Pass TRUE for mosconfig (unless this is a root pool) because
+ * the user-supplied config is actually the one to trust when
+ * doing an import.
+ */
+ loaderr = error = spa_load(spa, config, SPA_LOAD_IMPORT, !isroot);
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ /*
+ * Toss any existing sparelist, as it doesn't have any validity anymore,
+ * and conflicts with spa_has_spare().
+ */
+ if (!isroot && spa->spa_spares.sav_config) {
+ nvlist_free(spa->spa_spares.sav_config);
+ spa->spa_spares.sav_config = NULL;
+ spa_load_spares(spa);
+ }
+ if (!isroot && spa->spa_l2cache.sav_config) {
+ nvlist_free(spa->spa_l2cache.sav_config);
+ spa->spa_l2cache.sav_config = NULL;
+ spa_load_l2cache(spa);
+ }
+
+ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
+ &nvroot) == 0);
+ if (error == 0)
+ error = spa_validate_aux(spa, nvroot, -1ULL, VDEV_ALLOC_SPARE);
+ if (error == 0)
+ error = spa_validate_aux(spa, nvroot, -1ULL,
+ VDEV_ALLOC_L2CACHE);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ if (error != 0 || (props && (error = spa_prop_set(spa, props)))) {
+ if (loaderr != 0 && loaderr != EINVAL && allowfaulted) {
+ /*
+ * If we failed to load the pool, but 'allowfaulted' is
+ * set, then manually set the config as if the config
+ * passed in was specified in the cache file.
+ */
+ error = 0;
+ spa->spa_import_faulted = B_FALSE;
+ if (spa->spa_config == NULL)
+ spa->spa_config = spa_config_generate(spa,
+ NULL, -1ULL, B_TRUE);
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+ } else {
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_remove(spa);
+ }
+ mutex_exit(&spa_namespace_lock);
+ return (error);
+ }
+
+ /*
+ * Override any spares and level 2 cache devices as specified by
+ * the user, as these may have correct device names/devids, etc.
+ */
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
+ &spares, &nspares) == 0) {
+ if (spa->spa_spares.sav_config)
+ VERIFY(nvlist_remove(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0);
+ else
+ VERIFY(nvlist_alloc(&spa->spa_spares.sav_config,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, spares, nspares) == 0);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_spares(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ spa->spa_spares.sav_sync = B_TRUE;
+ }
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
+ &l2cache, &nl2cache) == 0) {
+ if (spa->spa_l2cache.sav_config)
+ VERIFY(nvlist_remove(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0);
+ else
+ VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_load_l2cache(spa);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ spa->spa_l2cache.sav_sync = B_TRUE;
+ }
+
+ if (spa_mode & FWRITE) {
+ /*
+ * Update the config cache to include the newly-imported pool.
+ */
+ spa_config_update_common(spa, SPA_CONFIG_UPDATE_POOL, isroot);
+ }
+
+ spa->spa_import_faulted = B_FALSE;
+ mutex_exit(&spa_namespace_lock);
+
+ return (0);
+}
+
+#ifdef _KERNEL
+/*
+ * Build a "root" vdev for a top level vdev read in from a rootpool
+ * device label.
+ */
+static void
+spa_build_rootpool_config(nvlist_t *config)
+{
+ nvlist_t *nvtop, *nvroot;
+ uint64_t pgid;
+
+ /*
+ * Add this top-level vdev to the child array.
+ */
+ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtop)
+ == 0);
+ VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pgid)
+ == 0);
+
+ /*
+ * Put this pool's top-level vdevs into a root vdev.
+ */
+ VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT)
+ == 0);
+ VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0);
+ VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0);
+ VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
+ &nvtop, 1) == 0);
+
+ /*
+ * Replace the existing vdev_tree with the new root vdev in
+ * this pool's configuration (remove the old, add the new).
+ */
+ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
+ nvlist_free(nvroot);
+}
+
+/*
+ * Get the root pool information from the root disk, then import the root pool
+ * during the system boot up time.
+ */
+extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **);
+
+int
+spa_check_rootconf(char *devpath, char *devid, nvlist_t **bestconf,
+ uint64_t *besttxg)
+{
+ nvlist_t *config;
+ uint64_t txg;
+ int error;
+
+ if (error = vdev_disk_read_rootlabel(devpath, devid, &config))
+ return (error);
+
+ VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0);
+
+ if (bestconf != NULL)
+ *bestconf = config;
+ else
+ nvlist_free(config);
+ *besttxg = txg;
+ return (0);
+}
+
+boolean_t
+spa_rootdev_validate(nvlist_t *nv)
+{
+ uint64_t ival;
+
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
+ nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
+ nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
+ return (B_FALSE);
+
+ return (B_TRUE);
+}
+
+
+/*
+ * Given the boot device's physical path or devid, check if the device
+ * is in a valid state. If so, return the configuration from the vdev
+ * label.
+ */
+int
+spa_get_rootconf(char *devpath, char *devid, nvlist_t **bestconf)
+{
+ nvlist_t *conf = NULL;
+ uint64_t txg = 0;
+ nvlist_t *nvtop, **child;
+ char *type;
+ char *bootpath = NULL;
+ uint_t children, c;
+ char *tmp;
+ int error;
+
+ if (devpath && ((tmp = strchr(devpath, ' ')) != NULL))
+ *tmp = '\0';
+ if (error = spa_check_rootconf(devpath, devid, &conf, &txg)) {
+ cmn_err(CE_NOTE, "error reading device label");
+ return (error);
+ }
+ if (txg == 0) {
+ cmn_err(CE_NOTE, "this device is detached");
+ nvlist_free(conf);
+ return (EINVAL);
+ }
+
+ VERIFY(nvlist_lookup_nvlist(conf, ZPOOL_CONFIG_VDEV_TREE,
+ &nvtop) == 0);
+ VERIFY(nvlist_lookup_string(nvtop, ZPOOL_CONFIG_TYPE, &type) == 0);
+
+ if (strcmp(type, VDEV_TYPE_DISK) == 0) {
+ if (spa_rootdev_validate(nvtop)) {
+ goto out;
+ } else {
+ nvlist_free(conf);
+ return (EINVAL);
+ }
+ }
+
+ ASSERT(strcmp(type, VDEV_TYPE_MIRROR) == 0);
+
+ VERIFY(nvlist_lookup_nvlist_array(nvtop, ZPOOL_CONFIG_CHILDREN,
+ &child, &children) == 0);
+
+ /*
+ * Go thru vdevs in the mirror to see if the given device
+ * has the most recent txg. Only the device with the most
+ * recent txg has valid information and should be booted.
+ */
+ for (c = 0; c < children; c++) {
+ char *cdevid, *cpath;
+ uint64_t tmptxg;
+
+ if (nvlist_lookup_string(child[c], ZPOOL_CONFIG_PHYS_PATH,
+ &cpath) != 0)
+ return (EINVAL);
+ if (nvlist_lookup_string(child[c], ZPOOL_CONFIG_DEVID,
+ &cdevid) != 0)
+ return (EINVAL);
+ if ((spa_check_rootconf(cpath, cdevid, NULL,
+ &tmptxg) == 0) && (tmptxg > txg)) {
+ txg = tmptxg;
+ VERIFY(nvlist_lookup_string(child[c],
+ ZPOOL_CONFIG_PATH, &bootpath) == 0);
+ }
+ }
+
+ /* Does the best device match the one we've booted from? */
+ if (bootpath) {
+ cmn_err(CE_NOTE, "try booting from '%s'", bootpath);
+ return (EINVAL);
+ }
+out:
+ *bestconf = conf;
+ return (0);
+}
+
+/*
+ * Import a root pool.
+ *
+ * For x86. devpath_list will consist of devid and/or physpath name of
+ * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a").
+ * The GRUB "findroot" command will return the vdev we should boot.
+ *
+ * For Sparc, devpath_list consists the physpath name of the booting device
+ * no matter the rootpool is a single device pool or a mirrored pool.
+ * e.g.
+ * "/pci@1f,0/ide@d/disk@0,0:a"
+ */
+int
+spa_import_rootpool(char *devpath, char *devid)
+{
+ nvlist_t *conf = NULL;
+ char *pname;
+ int error;
+
+ /*
+ * Get the vdev pathname and configuation from the most
+ * recently updated vdev (highest txg).
+ */
+ if (error = spa_get_rootconf(devpath, devid, &conf))
+ goto msg_out;
+
+ /*
+ * Add type "root" vdev to the config.
+ */
+ spa_build_rootpool_config(conf);
+
+ VERIFY(nvlist_lookup_string(conf, ZPOOL_CONFIG_POOL_NAME, &pname) == 0);
+
+ /*
+ * We specify 'allowfaulted' for this to be treated like spa_open()
+ * instead of spa_import(). This prevents us from marking vdevs as
+ * persistently unavailable, and generates FMA ereports as if it were a
+ * pool open, not import.
+ */
+ error = spa_import_common(pname, conf, NULL, B_TRUE, B_TRUE);
+ ASSERT(error != EEXIST);
+
+ nvlist_free(conf);
+ return (error);
+
+msg_out:
+ cmn_err(CE_NOTE, "\n"
+ " *************************************************** \n"
+ " * This device is not bootable! * \n"
+ " * It is either offlined or detached or faulted. * \n"
+ " * Please try to boot from a different device. * \n"
+ " *************************************************** ");
+
+ return (error);
+}
+#endif
+
+/*
+ * Import a non-root pool into the system.
+ */
+int
+spa_import(const char *pool, nvlist_t *config, nvlist_t *props)
+{
+ return (spa_import_common(pool, config, props, B_FALSE, B_FALSE));
+}
+
+int
+spa_import_faulted(const char *pool, nvlist_t *config, nvlist_t *props)
+{
+ return (spa_import_common(pool, config, props, B_FALSE, B_TRUE));
+}
+
+
+/*
+ * This (illegal) pool name is used when temporarily importing a spa_t in order
+ * to get the vdev stats associated with the imported devices.
+ */
+#define TRYIMPORT_NAME "$import"
+
+nvlist_t *
+spa_tryimport(nvlist_t *tryconfig)
+{
+ nvlist_t *config = NULL;
+ char *poolname;
+ spa_t *spa;
+ uint64_t state;
+
+ if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname))
+ return (NULL);
+
+ if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state))
+ return (NULL);
+
+ /*
+ * Create and initialize the spa structure.
+ */
+ mutex_enter(&spa_namespace_lock);
+ spa = spa_add(TRYIMPORT_NAME, NULL);
+ spa_activate(spa);
+
+ /*
+ * Pass off the heavy lifting to spa_load().
+ * Pass TRUE for mosconfig because the user-supplied config
+ * is actually the one to trust when doing an import.
+ */
+ (void) spa_load(spa, tryconfig, SPA_LOAD_TRYIMPORT, B_TRUE);
+
+ /*
+ * If 'tryconfig' was at least parsable, return the current config.
+ */
+ if (spa->spa_root_vdev != NULL) {
+ config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
+ VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
+ poolname) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
+ state) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
+ spa->spa_uberblock.ub_timestamp) == 0);
+
+ /*
+ * If the bootfs property exists on this pool then we
+ * copy it out so that external consumers can tell which
+ * pools are bootable.
+ */
+ if (spa->spa_bootfs) {
+ char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+
+ /*
+ * We have to play games with the name since the
+ * pool was opened as TRYIMPORT_NAME.
+ */
+ if (dsl_dsobj_to_dsname(spa_name(spa),
+ spa->spa_bootfs, tmpname) == 0) {
+ char *cp;
+ char *dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+
+ cp = strchr(tmpname, '/');
+ if (cp == NULL) {
+ (void) strlcpy(dsname, tmpname,
+ MAXPATHLEN);
+ } else {
+ (void) snprintf(dsname, MAXPATHLEN,
+ "%s/%s", poolname, ++cp);
+ }
+ VERIFY(nvlist_add_string(config,
+ ZPOOL_CONFIG_BOOTFS, dsname) == 0);
+ kmem_free(dsname, MAXPATHLEN);
+ }
+ kmem_free(tmpname, MAXPATHLEN);
+ }
+
+ /*
+ * Add the list of hot spares and level 2 cache devices.
+ */
+ spa_add_spares(spa, config);
+ spa_add_l2cache(spa, config);
+ }
+
+ spa_unload(spa);
+ spa_deactivate(spa);
+ spa_remove(spa);
+ mutex_exit(&spa_namespace_lock);
+
+ return (config);
+}
+
+/*
+ * Pool export/destroy
+ *
+ * The act of destroying or exporting a pool is very simple. We make sure there
+ * is no more pending I/O and any references to the pool are gone. Then, we
+ * update the pool state and sync all the labels to disk, removing the
+ * configuration from the cache afterwards.
+ */
+static int
+spa_export_common(char *pool, int new_state, nvlist_t **oldconfig,
+ boolean_t force)
+{
+ spa_t *spa;
+
+ if (oldconfig)
+ *oldconfig = NULL;
+
+ if (!(spa_mode & FWRITE))
+ return (EROFS);
+
+ mutex_enter(&spa_namespace_lock);
+ if ((spa = spa_lookup(pool)) == NULL) {
+ mutex_exit(&spa_namespace_lock);
+ return (ENOENT);
+ }
+
+ /*
+ * Put a hold on the pool, drop the namespace lock, stop async tasks,
+ * reacquire the namespace lock, and see if we can export.
+ */
+ spa_open_ref(spa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+ spa_async_suspend(spa);
+ mutex_enter(&spa_namespace_lock);
+ spa_close(spa, FTAG);
+
+ /*
+ * The pool will be in core if it's openable,
+ * in which case we can modify its state.
+ */
+ if (spa->spa_state != POOL_STATE_UNINITIALIZED && spa->spa_sync_on) {
+ /*
+ * Objsets may be open only because they're dirty, so we
+ * have to force it to sync before checking spa_refcnt.
+ */
+ txg_wait_synced(spa->spa_dsl_pool, 0);
+
+ /*
+ * A pool cannot be exported or destroyed if there are active
+ * references. If we are resetting a pool, allow references by
+ * fault injection handlers.
+ */
+ if (!spa_refcount_zero(spa) ||
+ (spa->spa_inject_ref != 0 &&
+ new_state != POOL_STATE_UNINITIALIZED)) {
+ spa_async_resume(spa);
+ mutex_exit(&spa_namespace_lock);
+ return (EBUSY);
+ }
+
+ /*
+ * A pool cannot be exported if it has an active shared spare.
+ * This is to prevent other pools stealing the active spare
+ * from an exported pool. At user's own will, such pool can
+ * be forcedly exported.
+ */
+ if (!force && new_state == POOL_STATE_EXPORTED &&
+ spa_has_active_shared_spare(spa)) {
+ spa_async_resume(spa);
+ mutex_exit(&spa_namespace_lock);
+ return (EXDEV);
+ }
+
+ /*
+ * We want this to be reflected on every label,
+ * so mark them all dirty. spa_unload() will do the
+ * final sync that pushes these changes out.
+ */
+ if (new_state != POOL_STATE_UNINITIALIZED) {
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa->spa_state = new_state;
+ spa->spa_final_txg = spa_last_synced_txg(spa) + 1;
+ vdev_config_dirty(spa->spa_root_vdev);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+ }
+
+ spa_event_notify(spa, NULL, ESC_ZFS_POOL_DESTROY);
+
+ if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
+ spa_unload(spa);
+ spa_deactivate(spa);
+ }
+
+ if (oldconfig && spa->spa_config)
+ VERIFY(nvlist_dup(spa->spa_config, oldconfig, 0) == 0);
+
+ if (new_state != POOL_STATE_UNINITIALIZED) {
+ spa_config_sync(spa, B_TRUE, B_TRUE);
+ spa_remove(spa);
+ }
+ mutex_exit(&spa_namespace_lock);
+
+ return (0);
+}
+
+/*
+ * Destroy a storage pool.
+ */
+int
+spa_destroy(char *pool)
+{
+ return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL, B_FALSE));
+}
+
+/*
+ * Export a storage pool.
+ */
+int
+spa_export(char *pool, nvlist_t **oldconfig, boolean_t force)
+{
+ return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig, force));
+}
+
+/*
+ * Similar to spa_export(), this unloads the spa_t without actually removing it
+ * from the namespace in any way.
+ */
+int
+spa_reset(char *pool)
+{
+ return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL,
+ B_FALSE));
+}
+
+/*
+ * ==========================================================================
+ * Device manipulation
+ * ==========================================================================
+ */
+
+/*
+ * Add a device to a storage pool.
+ */
+int
+spa_vdev_add(spa_t *spa, nvlist_t *nvroot)
+{
+ uint64_t txg;
+ int c, error;
+ vdev_t *rvd = spa->spa_root_vdev;
+ vdev_t *vd, *tvd;
+ nvlist_t **spares, **l2cache;
+ uint_t nspares, nl2cache;
+
+ txg = spa_vdev_enter(spa);
+
+ if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0,
+ VDEV_ALLOC_ADD)) != 0)
+ return (spa_vdev_exit(spa, NULL, txg, error));
+
+ spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */
+
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares,
+ &nspares) != 0)
+ nspares = 0;
+
+ if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache,
+ &nl2cache) != 0)
+ nl2cache = 0;
+
+ if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0)
+ return (spa_vdev_exit(spa, vd, txg, EINVAL));
+
+ if (vd->vdev_children != 0 &&
+ (error = vdev_create(vd, txg, B_FALSE)) != 0)
+ return (spa_vdev_exit(spa, vd, txg, error));
+
+ /*
+ * We must validate the spares and l2cache devices after checking the
+ * children. Otherwise, vdev_inuse() will blindly overwrite the spare.
+ */
+ if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0)
+ return (spa_vdev_exit(spa, vd, txg, error));
+
+ /*
+ * Transfer each new top-level vdev from vd to rvd.
+ */
+ for (c = 0; c < vd->vdev_children; c++) {
+ tvd = vd->vdev_child[c];
+ vdev_remove_child(vd, tvd);
+ tvd->vdev_id = rvd->vdev_children;
+ vdev_add_child(rvd, tvd);
+ vdev_config_dirty(tvd);
+ }
+
+ if (nspares != 0) {
+ spa_set_aux_vdevs(&spa->spa_spares, spares, nspares,
+ ZPOOL_CONFIG_SPARES);
+ spa_load_spares(spa);
+ spa->spa_spares.sav_sync = B_TRUE;
+ }
+
+ if (nl2cache != 0) {
+ spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache,
+ ZPOOL_CONFIG_L2CACHE);
+ spa_load_l2cache(spa);
+ spa->spa_l2cache.sav_sync = B_TRUE;
+ }
+
+ /*
+ * We have to be careful when adding new vdevs to an existing pool.
+ * If other threads start allocating from these vdevs before we
+ * sync the config cache, and we lose power, then upon reboot we may
+ * fail to open the pool because there are DVAs that the config cache
+ * can't translate. Therefore, we first add the vdevs without
+ * initializing metaslabs; sync the config cache (via spa_vdev_exit());
+ * and then let spa_config_update() initialize the new metaslabs.
+ *
+ * spa_load() checks for added-but-not-initialized vdevs, so that
+ * if we lose power at any point in this sequence, the remaining
+ * steps will be completed the next time we load the pool.
+ */
+ (void) spa_vdev_exit(spa, vd, txg, 0);
+
+ mutex_enter(&spa_namespace_lock);
+ spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
+ mutex_exit(&spa_namespace_lock);
+
+ return (0);
+}
+
+/*
+ * Attach a device to a mirror. The arguments are the path to any device
+ * in the mirror, and the nvroot for the new device. If the path specifies
+ * a device that is not mirrored, we automatically insert the mirror vdev.
+ *
+ * If 'replacing' is specified, the new device is intended to replace the
+ * existing device; in this case the two devices are made into their own
+ * mirror using the 'replacing' vdev, which is functionally identical to
+ * the mirror vdev (it actually reuses all the same ops) but has a few
+ * extra rules: you can't attach to it after it's been created, and upon
+ * completion of resilvering, the first disk (the one being replaced)
+ * is automatically detached.
+ */
+int
+spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing)
+{
+ uint64_t txg, open_txg;
+ vdev_t *rvd = spa->spa_root_vdev;
+ vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd;
+ vdev_ops_t *pvops;
+ dmu_tx_t *tx;
+ char *oldvdpath, *newvdpath;
+ int newvd_isspare;
+ int error;
+
+ txg = spa_vdev_enter(spa);
+
+ oldvd = spa_lookup_by_guid(spa, guid, B_FALSE);
+
+ if (oldvd == NULL)
+ return (spa_vdev_exit(spa, NULL, txg, ENODEV));
+
+ if (!oldvd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+
+ pvd = oldvd->vdev_parent;
+
+ if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0,
+ VDEV_ALLOC_ADD)) != 0)
+ return (spa_vdev_exit(spa, NULL, txg, EINVAL));
+
+ if (newrootvd->vdev_children != 1)
+ return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));
+
+ newvd = newrootvd->vdev_child[0];
+
+ if (!newvd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));
+
+ if ((error = vdev_create(newrootvd, txg, replacing)) != 0)
+ return (spa_vdev_exit(spa, newrootvd, txg, error));
+
+ /*
+ * Spares can't replace logs
+ */
+ if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare)
+ return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
+
+ if (!replacing) {
+ /*
+ * For attach, the only allowable parent is a mirror or the root
+ * vdev.
+ */
+ if (pvd->vdev_ops != &vdev_mirror_ops &&
+ pvd->vdev_ops != &vdev_root_ops)
+ return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
+
+ pvops = &vdev_mirror_ops;
+ } else {
+ /*
+ * Active hot spares can only be replaced by inactive hot
+ * spares.
+ */
+ if (pvd->vdev_ops == &vdev_spare_ops &&
+ pvd->vdev_child[1] == oldvd &&
+ !spa_has_spare(spa, newvd->vdev_guid))
+ return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
+
+ /*
+ * If the source is a hot spare, and the parent isn't already a
+ * spare, then we want to create a new hot spare. Otherwise, we
+ * want to create a replacing vdev. The user is not allowed to
+ * attach to a spared vdev child unless the 'isspare' state is
+ * the same (spare replaces spare, non-spare replaces
+ * non-spare).
+ */
+ if (pvd->vdev_ops == &vdev_replacing_ops)
+ return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
+ else if (pvd->vdev_ops == &vdev_spare_ops &&
+ newvd->vdev_isspare != oldvd->vdev_isspare)
+ return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
+ else if (pvd->vdev_ops != &vdev_spare_ops &&
+ newvd->vdev_isspare)
+ pvops = &vdev_spare_ops;
+ else
+ pvops = &vdev_replacing_ops;
+ }
+
+ /*
+ * Compare the new device size with the replaceable/attachable
+ * device size.
+ */
+ if (newvd->vdev_psize < vdev_get_rsize(oldvd))
+ return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW));
+
+ /*
+ * The new device cannot have a higher alignment requirement
+ * than the top-level vdev.
+ */
+ if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift)
+ return (spa_vdev_exit(spa, newrootvd, txg, EDOM));
+
+ /*
+ * If this is an in-place replacement, update oldvd's path and devid
+ * to make it distinguishable from newvd, and unopenable from now on.
+ */
+ if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) {
+ spa_strfree(oldvd->vdev_path);
+ oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5,
+ KM_SLEEP);
+ (void) sprintf(oldvd->vdev_path, "%s/%s",
+ newvd->vdev_path, "old");
+ if (oldvd->vdev_devid != NULL) {
+ spa_strfree(oldvd->vdev_devid);
+ oldvd->vdev_devid = NULL;
+ }
+ }
+
+ /*
+ * If the parent is not a mirror, or if we're replacing, insert the new
+ * mirror/replacing/spare vdev above oldvd.
+ */
+ if (pvd->vdev_ops != pvops)
+ pvd = vdev_add_parent(oldvd, pvops);
+
+ ASSERT(pvd->vdev_top->vdev_parent == rvd);
+ ASSERT(pvd->vdev_ops == pvops);
+ ASSERT(oldvd->vdev_parent == pvd);
+
+ /*
+ * Extract the new device from its root and add it to pvd.
+ */
+ vdev_remove_child(newrootvd, newvd);
+ newvd->vdev_id = pvd->vdev_children;
+ vdev_add_child(pvd, newvd);
+
+ /*
+ * If newvd is smaller than oldvd, but larger than its rsize,
+ * the addition of newvd may have decreased our parent's asize.
+ */
+ pvd->vdev_asize = MIN(pvd->vdev_asize, newvd->vdev_asize);
+
+ tvd = newvd->vdev_top;
+ ASSERT(pvd->vdev_top == tvd);
+ ASSERT(tvd->vdev_parent == rvd);
+
+ vdev_config_dirty(tvd);
+
+ /*
+ * Set newvd's DTL to [TXG_INITIAL, open_txg]. It will propagate
+ * upward when spa_vdev_exit() calls vdev_dtl_reassess().
+ */
+ open_txg = txg + TXG_CONCURRENT_STATES - 1;
+
+ mutex_enter(&newvd->vdev_dtl_lock);
+ space_map_add(&newvd->vdev_dtl_map, TXG_INITIAL,
+ open_txg - TXG_INITIAL + 1);
+ mutex_exit(&newvd->vdev_dtl_lock);
+
+ if (newvd->vdev_isspare)
+ spa_spare_activate(newvd);
+ oldvdpath = spa_strdup(oldvd->vdev_path);
+ newvdpath = spa_strdup(newvd->vdev_path);
+ newvd_isspare = newvd->vdev_isspare;
+
+ /*
+ * Mark newvd's DTL dirty in this txg.
+ */
+ vdev_dirty(tvd, VDD_DTL, newvd, txg);
+
+ (void) spa_vdev_exit(spa, newrootvd, open_txg, 0);
+
+ tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
+ if (dmu_tx_assign(tx, TXG_WAIT) == 0) {
+ spa_history_internal_log(LOG_POOL_VDEV_ATTACH, spa, tx,
+ CRED(), "%s vdev=%s %s vdev=%s",
+ replacing && newvd_isspare ? "spare in" :
+ replacing ? "replace" : "attach", newvdpath,
+ replacing ? "for" : "to", oldvdpath);
+ dmu_tx_commit(tx);
+ } else {
+ dmu_tx_abort(tx);
+ }
+
+ spa_strfree(oldvdpath);
+ spa_strfree(newvdpath);
+
+ /*
+ * Kick off a resilver to update newvd.
+ */
+ VERIFY3U(spa_scrub(spa, POOL_SCRUB_RESILVER), ==, 0);
+
+ return (0);
+}
+
+/*
+ * Detach a device from a mirror or replacing vdev.
+ * If 'replace_done' is specified, only detach if the parent
+ * is a replacing vdev.
+ */
+int
+spa_vdev_detach(spa_t *spa, uint64_t guid, int replace_done)
+{
+ uint64_t txg;
+ int c, t, error;
+ vdev_t *rvd = spa->spa_root_vdev;
+ vdev_t *vd, *pvd, *cvd, *tvd;
+ boolean_t unspare = B_FALSE;
+ uint64_t unspare_guid;
+ size_t len;
+
+ txg = spa_vdev_enter(spa);
+
+ vd = spa_lookup_by_guid(spa, guid, B_FALSE);
+
+ if (vd == NULL)
+ return (spa_vdev_exit(spa, NULL, txg, ENODEV));
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+
+ pvd = vd->vdev_parent;
+
+ /*
+ * If replace_done is specified, only remove this device if it's
+ * the first child of a replacing vdev. For the 'spare' vdev, either
+ * disk can be removed.
+ */
+ if (replace_done) {
+ if (pvd->vdev_ops == &vdev_replacing_ops) {
+ if (vd->vdev_id != 0)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+ } else if (pvd->vdev_ops != &vdev_spare_ops) {
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+ }
+ }
+
+ ASSERT(pvd->vdev_ops != &vdev_spare_ops ||
+ spa_version(spa) >= SPA_VERSION_SPARES);
+
+ /*
+ * Only mirror, replacing, and spare vdevs support detach.
+ */
+ if (pvd->vdev_ops != &vdev_replacing_ops &&
+ pvd->vdev_ops != &vdev_mirror_ops &&
+ pvd->vdev_ops != &vdev_spare_ops)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+
+ /*
+ * If there's only one replica, you can't detach it.
+ */
+ if (pvd->vdev_children <= 1)
+ return (spa_vdev_exit(spa, NULL, txg, EBUSY));
+
+ /*
+ * If all siblings have non-empty DTLs, this device may have the only
+ * valid copy of the data, which means we cannot safely detach it.
+ *
+ * XXX -- as in the vdev_offline() case, we really want a more
+ * precise DTL check.
+ */
+ for (c = 0; c < pvd->vdev_children; c++) {
+ uint64_t dirty;
+
+ cvd = pvd->vdev_child[c];
+ if (cvd == vd)
+ continue;
+ if (vdev_is_dead(cvd))
+ continue;
+ mutex_enter(&cvd->vdev_dtl_lock);
+ dirty = cvd->vdev_dtl_map.sm_space |
+ cvd->vdev_dtl_scrub.sm_space;
+ mutex_exit(&cvd->vdev_dtl_lock);
+ if (!dirty)
+ break;
+ }
+
+ if (c == pvd->vdev_children)
+ return (spa_vdev_exit(spa, NULL, txg, EBUSY));
+
+ /*
+ * If we are detaching the second disk from a replacing vdev, then
+ * check to see if we changed the original vdev's path to have "/old"
+ * at the end in spa_vdev_attach(). If so, undo that change now.
+ */
+ if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id == 1 &&
+ pvd->vdev_child[0]->vdev_path != NULL &&
+ pvd->vdev_child[1]->vdev_path != NULL) {
+ ASSERT(pvd->vdev_child[1] == vd);
+ cvd = pvd->vdev_child[0];
+ len = strlen(vd->vdev_path);
+ if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
+ strcmp(cvd->vdev_path + len, "/old") == 0) {
+ spa_strfree(cvd->vdev_path);
+ cvd->vdev_path = spa_strdup(vd->vdev_path);
+ }
+ }
+
+ /*
+ * If we are detaching the original disk from a spare, then it implies
+ * that the spare should become a real disk, and be removed from the
+ * active spare list for the pool.
+ */
+ if (pvd->vdev_ops == &vdev_spare_ops &&
+ vd->vdev_id == 0)
+ unspare = B_TRUE;
+
+ /*
+ * Erase the disk labels so the disk can be used for other things.
+ * This must be done after all other error cases are handled,
+ * but before we disembowel vd (so we can still do I/O to it).
+ * But if we can't do it, don't treat the error as fatal --
+ * it may be that the unwritability of the disk is the reason
+ * it's being detached!
+ */
+ error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
+
+ /*
+ * Remove vd from its parent and compact the parent's children.
+ */
+ vdev_remove_child(pvd, vd);
+ vdev_compact_children(pvd);
+
+ /*
+ * Remember one of the remaining children so we can get tvd below.
+ */
+ cvd = pvd->vdev_child[0];
+
+ /*
+ * If we need to remove the remaining child from the list of hot spares,
+ * do it now, marking the vdev as no longer a spare in the process. We
+ * must do this before vdev_remove_parent(), because that can change the
+ * GUID if it creates a new toplevel GUID.
+ */
+ if (unspare) {
+ ASSERT(cvd->vdev_isspare);
+ spa_spare_remove(cvd);
+ unspare_guid = cvd->vdev_guid;
+ }
+
+ /*
+ * If the parent mirror/replacing vdev only has one child,
+ * the parent is no longer needed. Remove it from the tree.
+ */
+ if (pvd->vdev_children == 1)
+ vdev_remove_parent(cvd);
+
+ /*
+ * We don't set tvd until now because the parent we just removed
+ * may have been the previous top-level vdev.
+ */
+ tvd = cvd->vdev_top;
+ ASSERT(tvd->vdev_parent == rvd);
+
+ /*
+ * Reevaluate the parent vdev state.
+ */
+ vdev_propagate_state(cvd);
+
+ /*
+ * If the device we just detached was smaller than the others, it may be
+ * possible to add metaslabs (i.e. grow the pool). vdev_metaslab_init()
+ * can't fail because the existing metaslabs are already in core, so
+ * there's nothing to read from disk.
+ */
+ VERIFY(vdev_metaslab_init(tvd, txg) == 0);
+
+ vdev_config_dirty(tvd);
+
+ /*
+ * Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that
+ * vd->vdev_detached is set and free vd's DTL object in syncing context.
+ * But first make sure we're not on any *other* txg's DTL list, to
+ * prevent vd from being accessed after it's freed.
+ */
+ for (t = 0; t < TXG_SIZE; t++)
+ (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
+ vd->vdev_detached = B_TRUE;
+ vdev_dirty(tvd, VDD_DTL, vd, txg);
+
+ spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE);
+
+ error = spa_vdev_exit(spa, vd, txg, 0);
+
+ /*
+ * If this was the removal of the original device in a hot spare vdev,
+ * then we want to go through and remove the device from the hot spare
+ * list of every other pool.
+ */
+ if (unspare) {
+ spa = NULL;
+ mutex_enter(&spa_namespace_lock);
+ while ((spa = spa_next(spa)) != NULL) {
+ if (spa->spa_state != POOL_STATE_ACTIVE)
+ continue;
+ spa_open_ref(spa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+ (void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
+ mutex_enter(&spa_namespace_lock);
+ spa_close(spa, FTAG);
+ }
+ mutex_exit(&spa_namespace_lock);
+ }
+
+ return (error);
+}
+
+static nvlist_t *
+spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid)
+{
+ for (int i = 0; i < count; i++) {
+ uint64_t guid;
+
+ VERIFY(nvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID,
+ &guid) == 0);
+
+ if (guid == target_guid)
+ return (nvpp[i]);
+ }
+
+ return (NULL);
+}
+
+static void
+spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count,
+ nvlist_t *dev_to_remove)
+{
+ nvlist_t **newdev = NULL;
+
+ if (count > 1)
+ newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP);
+
+ for (int i = 0, j = 0; i < count; i++) {
+ if (dev[i] == dev_to_remove)
+ continue;
+ VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0);
+ }
+
+ VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0);
+ VERIFY(nvlist_add_nvlist_array(config, name, newdev, count - 1) == 0);
+
+ for (int i = 0; i < count - 1; i++)
+ nvlist_free(newdev[i]);
+
+ if (count > 1)
+ kmem_free(newdev, (count - 1) * sizeof (void *));
+}
+
+/*
+ * Remove a device from the pool. Currently, this supports removing only hot
+ * spares and level 2 ARC devices.
+ */
+int
+spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
+{
+ vdev_t *vd;
+ nvlist_t **spares, **l2cache, *nv;
+ uint_t nspares, nl2cache;
+ uint64_t txg;
+ int error = 0;
+
+ txg = spa_vdev_enter(spa);
+
+ vd = spa_lookup_by_guid(spa, guid, B_FALSE);
+
+ if (spa->spa_spares.sav_vdevs != NULL &&
+ nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 &&
+ (nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) {
+ /*
+ * Only remove the hot spare if it's not currently in use
+ * in this pool.
+ */
+ if (vd == NULL || unspare) {
+ spa_vdev_remove_aux(spa->spa_spares.sav_config,
+ ZPOOL_CONFIG_SPARES, spares, nspares, nv);
+ spa_load_spares(spa);
+ spa->spa_spares.sav_sync = B_TRUE;
+ } else {
+ error = EBUSY;
+ }
+ } else if (spa->spa_l2cache.sav_vdevs != NULL &&
+ nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 &&
+ (nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) {
+ /*
+ * Cache devices can always be removed.
+ */
+ spa_vdev_remove_aux(spa->spa_l2cache.sav_config,
+ ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
+ spa_load_l2cache(spa);
+ spa->spa_l2cache.sav_sync = B_TRUE;
+ } else if (vd != NULL) {
+ /*
+ * Normal vdevs cannot be removed (yet).
+ */
+ error = ENOTSUP;
+ } else {
+ /*
+ * There is no vdev of any kind with the specified guid.
+ */
+ error = ENOENT;
+ }
+
+ return (spa_vdev_exit(spa, NULL, txg, error));
+}
+
+/*
+ * Find any device that's done replacing, or a vdev marked 'unspare' that's
+ * current spared, so we can detach it.
+ */
+static vdev_t *
+spa_vdev_resilver_done_hunt(vdev_t *vd)
+{
+ vdev_t *newvd, *oldvd;
+ int c;
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]);
+ if (oldvd != NULL)
+ return (oldvd);
+ }
+
+ /*
+ * Check for a completed replacement.
+ */
+ if (vd->vdev_ops == &vdev_replacing_ops && vd->vdev_children == 2) {
+ oldvd = vd->vdev_child[0];
+ newvd = vd->vdev_child[1];
+
+ mutex_enter(&newvd->vdev_dtl_lock);
+ if (newvd->vdev_dtl_map.sm_space == 0 &&
+ newvd->vdev_dtl_scrub.sm_space == 0) {
+ mutex_exit(&newvd->vdev_dtl_lock);
+ return (oldvd);
+ }
+ mutex_exit(&newvd->vdev_dtl_lock);
+ }
+
+ /*
+ * Check for a completed resilver with the 'unspare' flag set.
+ */
+ if (vd->vdev_ops == &vdev_spare_ops && vd->vdev_children == 2) {
+ newvd = vd->vdev_child[0];
+ oldvd = vd->vdev_child[1];
+
+ mutex_enter(&newvd->vdev_dtl_lock);
+ if (newvd->vdev_unspare &&
+ newvd->vdev_dtl_map.sm_space == 0 &&
+ newvd->vdev_dtl_scrub.sm_space == 0) {
+ newvd->vdev_unspare = 0;
+ mutex_exit(&newvd->vdev_dtl_lock);
+ return (oldvd);
+ }
+ mutex_exit(&newvd->vdev_dtl_lock);
+ }
+
+ return (NULL);
+}
+
+static void
+spa_vdev_resilver_done(spa_t *spa)
+{
+ vdev_t *vd;
+ vdev_t *pvd;
+ uint64_t guid;
+ uint64_t pguid = 0;
+
+ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
+
+ while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) {
+ guid = vd->vdev_guid;
+ /*
+ * If we have just finished replacing a hot spared device, then
+ * we need to detach the parent's first child (the original hot
+ * spare) as well.
+ */
+ pvd = vd->vdev_parent;
+ if (pvd->vdev_parent->vdev_ops == &vdev_spare_ops &&
+ pvd->vdev_id == 0) {
+ ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
+ ASSERT(pvd->vdev_parent->vdev_children == 2);
+ pguid = pvd->vdev_parent->vdev_child[1]->vdev_guid;
+ }
+ spa_config_exit(spa, SCL_CONFIG, FTAG);
+ if (spa_vdev_detach(spa, guid, B_TRUE) != 0)
+ return;
+ if (pguid != 0 && spa_vdev_detach(spa, pguid, B_TRUE) != 0)
+ return;
+ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
+ }
+
+ spa_config_exit(spa, SCL_CONFIG, FTAG);
+}
+
+/*
+ * Update the stored path for this vdev. Dirty the vdev configuration, relying
+ * on spa_vdev_enter/exit() to synchronize the labels and cache.
+ */
+int
+spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath)
+{
+ vdev_t *vd;
+ uint64_t txg;
+
+ txg = spa_vdev_enter(spa);
+
+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) {
+ /*
+ * Determine if this is a reference to a hot spare device. If
+ * it is, update the path manually as there is no associated
+ * vdev_t that can be synced to disk.
+ */
+ nvlist_t **spares;
+ uint_t i, nspares;
+
+ if (spa->spa_spares.sav_config != NULL) {
+ VERIFY(nvlist_lookup_nvlist_array(
+ spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES,
+ &spares, &nspares) == 0);
+ for (i = 0; i < nspares; i++) {
+ uint64_t theguid;
+ VERIFY(nvlist_lookup_uint64(spares[i],
+ ZPOOL_CONFIG_GUID, &theguid) == 0);
+ if (theguid == guid) {
+ VERIFY(nvlist_add_string(spares[i],
+ ZPOOL_CONFIG_PATH, newpath) == 0);
+ spa_load_spares(spa);
+ spa->spa_spares.sav_sync = B_TRUE;
+ return (spa_vdev_exit(spa, NULL, txg,
+ 0));
+ }
+ }
+ }
+
+ return (spa_vdev_exit(spa, NULL, txg, ENOENT));
+ }
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+
+ spa_strfree(vd->vdev_path);
+ vd->vdev_path = spa_strdup(newpath);
+
+ vdev_config_dirty(vd->vdev_top);
+
+ return (spa_vdev_exit(spa, NULL, txg, 0));
+}
+
+/*
+ * ==========================================================================
+ * SPA Scrubbing
+ * ==========================================================================
+ */
+
+int
+spa_scrub(spa_t *spa, pool_scrub_type_t type)
+{
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
+
+ if ((uint_t)type >= POOL_SCRUB_TYPES)
+ return (ENOTSUP);
+
+ /*
+ * If a resilver was requested, but there is no DTL on a
+ * writeable leaf device, we have nothing to do.
+ */
+ if (type == POOL_SCRUB_RESILVER &&
+ !vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
+ spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
+ return (0);
+ }
+
+ if (type == POOL_SCRUB_EVERYTHING &&
+ spa->spa_dsl_pool->dp_scrub_func != SCRUB_FUNC_NONE &&
+ spa->spa_dsl_pool->dp_scrub_isresilver)
+ return (EBUSY);
+
+ if (type == POOL_SCRUB_EVERYTHING || type == POOL_SCRUB_RESILVER) {
+ return (dsl_pool_scrub_clean(spa->spa_dsl_pool));
+ } else if (type == POOL_SCRUB_NONE) {
+ return (dsl_pool_scrub_cancel(spa->spa_dsl_pool));
+ } else {
+ return (EINVAL);
+ }
+}
+
+/*
+ * ==========================================================================
+ * SPA async task processing
+ * ==========================================================================
+ */
+
+static void
+spa_async_remove(spa_t *spa, vdev_t *vd)
+{
+ if (vd->vdev_remove_wanted) {
+ vd->vdev_remove_wanted = 0;
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE);
+ vdev_clear(spa, vd);
+ vdev_state_dirty(vd->vdev_top);
+ }
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ spa_async_remove(spa, vd->vdev_child[c]);
+}
+
+static void
+spa_async_probe(spa_t *spa, vdev_t *vd)
+{
+ if (vd->vdev_probe_wanted) {
+ vd->vdev_probe_wanted = 0;
+ vdev_reopen(vd); /* vdev_open() does the actual probe */
+ }
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ spa_async_probe(spa, vd->vdev_child[c]);
+}
+
+static void
+spa_async_thread(spa_t *spa)
+{
+ int tasks;
+
+ ASSERT(spa->spa_sync_on);
+
+ mutex_enter(&spa->spa_async_lock);
+ tasks = spa->spa_async_tasks;
+ spa->spa_async_tasks = 0;
+ mutex_exit(&spa->spa_async_lock);
+
+ /*
+ * See if the config needs to be updated.
+ */
+ if (tasks & SPA_ASYNC_CONFIG_UPDATE) {
+ mutex_enter(&spa_namespace_lock);
+ spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
+ mutex_exit(&spa_namespace_lock);
+ }
+
+ /*
+ * See if any devices need to be marked REMOVED.
+ */
+ if (tasks & SPA_ASYNC_REMOVE) {
+ spa_vdev_state_enter(spa);
+ spa_async_remove(spa, spa->spa_root_vdev);
+ for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
+ spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]);
+ for (int i = 0; i < spa->spa_spares.sav_count; i++)
+ spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]);
+ (void) spa_vdev_state_exit(spa, NULL, 0);
+ }
+
+ /*
+ * See if any devices need to be probed.
+ */
+ if (tasks & SPA_ASYNC_PROBE) {
+ spa_vdev_state_enter(spa);
+ spa_async_probe(spa, spa->spa_root_vdev);
+ (void) spa_vdev_state_exit(spa, NULL, 0);
+ }
+
+ /*
+ * If any devices are done replacing, detach them.
+ */
+ if (tasks & SPA_ASYNC_RESILVER_DONE)
+ spa_vdev_resilver_done(spa);
+
+ /*
+ * Kick off a resilver.
+ */
+ if (tasks & SPA_ASYNC_RESILVER)
+ VERIFY(spa_scrub(spa, POOL_SCRUB_RESILVER) == 0);
+
+ /*
+ * Let the world know that we're done.
+ */
+ mutex_enter(&spa->spa_async_lock);
+ spa->spa_async_thread = NULL;
+ cv_broadcast(&spa->spa_async_cv);
+ mutex_exit(&spa->spa_async_lock);
+ thread_exit();
+}
+
+void
+spa_async_suspend(spa_t *spa)
+{
+ mutex_enter(&spa->spa_async_lock);
+ spa->spa_async_suspended++;
+ while (spa->spa_async_thread != NULL)
+ cv_wait(&spa->spa_async_cv, &spa->spa_async_lock);
+ mutex_exit(&spa->spa_async_lock);
+}
+
+void
+spa_async_resume(spa_t *spa)
+{
+ mutex_enter(&spa->spa_async_lock);
+ ASSERT(spa->spa_async_suspended != 0);
+ spa->spa_async_suspended--;
+ mutex_exit(&spa->spa_async_lock);
+}
+
+static void
+spa_async_dispatch(spa_t *spa)
+{
+ mutex_enter(&spa->spa_async_lock);
+ if (spa->spa_async_tasks && !spa->spa_async_suspended &&
+ spa->spa_async_thread == NULL &&
+ rootdir != NULL && !vn_is_readonly(rootdir))
+ spa->spa_async_thread = thread_create(NULL, 0,
+ spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri);
+ mutex_exit(&spa->spa_async_lock);
+}
+
+void
+spa_async_request(spa_t *spa, int task)
+{
+ mutex_enter(&spa->spa_async_lock);
+ spa->spa_async_tasks |= task;
+ mutex_exit(&spa->spa_async_lock);
+}
+
+/*
+ * ==========================================================================
+ * SPA syncing routines
+ * ==========================================================================
+ */
+
+static void
+spa_sync_deferred_frees(spa_t *spa, uint64_t txg)
+{
+ bplist_t *bpl = &spa->spa_sync_bplist;
+ dmu_tx_t *tx;
+ blkptr_t blk;
+ uint64_t itor = 0;
+ zio_t *zio;
+ int error;
+ uint8_t c = 1;
+
+ zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
+
+ while (bplist_iterate(bpl, &itor, &blk) == 0) {
+ ASSERT(blk.blk_birth < txg);
+ zio_nowait(zio_free(zio, spa, txg, &blk, NULL, NULL,
+ ZIO_FLAG_MUSTSUCCEED));
+ }
+
+ error = zio_wait(zio);
+ ASSERT3U(error, ==, 0);
+
+ tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+ bplist_vacate(bpl, tx);
+
+ /*
+ * Pre-dirty the first block so we sync to convergence faster.
+ * (Usually only the first block is needed.)
+ */
+ dmu_write(spa->spa_meta_objset, spa->spa_sync_bplist_obj, 0, 1, &c, tx);
+ dmu_tx_commit(tx);
+}
+
+static void
+spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx)
+{
+ char *packed = NULL;
+ size_t bufsize;
+ size_t nvsize = 0;
+ dmu_buf_t *db;
+
+ VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0);
+
+ /*
+ * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
+ * information. This avoids the dbuf_will_dirty() path and
+ * saves us a pre-read to get data we don't actually care about.
+ */
+ bufsize = P2ROUNDUP(nvsize, SPA_CONFIG_BLOCKSIZE);
+ packed = kmem_alloc(bufsize, KM_SLEEP);
+
+ VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
+ KM_SLEEP) == 0);
+ bzero(packed + nvsize, bufsize - nvsize);
+
+ dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);
+
+ kmem_free(packed, bufsize);
+
+ VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+ *(uint64_t *)db->db_data = nvsize;
+ dmu_buf_rele(db, FTAG);
+}
+
+static void
+spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx,
+ const char *config, const char *entry)
+{
+ nvlist_t *nvroot;
+ nvlist_t **list;
+ int i;
+
+ if (!sav->sav_sync)
+ return;
+
+ /*
+ * Update the MOS nvlist describing the list of available devices.
+ * spa_validate_aux() will have already made sure this nvlist is
+ * valid and the vdevs are labeled appropriately.
+ */
+ if (sav->sav_object == 0) {
+ sav->sav_object = dmu_object_alloc(spa->spa_meta_objset,
+ DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE,
+ sizeof (uint64_t), tx);
+ VERIFY(zap_update(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1,
+ &sav->sav_object, tx) == 0);
+ }
+
+ VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ if (sav->sav_count == 0) {
+ VERIFY(nvlist_add_nvlist_array(nvroot, config, NULL, 0) == 0);
+ } else {
+ list = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP);
+ for (i = 0; i < sav->sav_count; i++)
+ list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
+ B_FALSE, B_FALSE, B_TRUE);
+ VERIFY(nvlist_add_nvlist_array(nvroot, config, list,
+ sav->sav_count) == 0);
+ for (i = 0; i < sav->sav_count; i++)
+ nvlist_free(list[i]);
+ kmem_free(list, sav->sav_count * sizeof (void *));
+ }
+
+ spa_sync_nvlist(spa, sav->sav_object, nvroot, tx);
+ nvlist_free(nvroot);
+
+ sav->sav_sync = B_FALSE;
+}
+
+static void
+spa_sync_config_object(spa_t *spa, dmu_tx_t *tx)
+{
+ nvlist_t *config;
+
+ if (list_is_empty(&spa->spa_config_dirty_list))
+ return;
+
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+
+ config = spa_config_generate(spa, spa->spa_root_vdev,
+ dmu_tx_get_txg(tx), B_FALSE);
+
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ if (spa->spa_config_syncing)
+ nvlist_free(spa->spa_config_syncing);
+ spa->spa_config_syncing = config;
+
+ spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
+}
+
+/*
+ * Set zpool properties.
+ */
+static void
+spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ spa_t *spa = arg1;
+ objset_t *mos = spa->spa_meta_objset;
+ nvlist_t *nvp = arg2;
+ nvpair_t *elem;
+ uint64_t intval;
+ char *strval;
+ zpool_prop_t prop;
+ const char *propname;
+ zprop_type_t proptype;
+ spa_config_dirent_t *dp;
+
+ mutex_enter(&spa->spa_props_lock);
+
+ elem = NULL;
+ while ((elem = nvlist_next_nvpair(nvp, elem))) {
+ switch (prop = zpool_name_to_prop(nvpair_name(elem))) {
+ case ZPOOL_PROP_VERSION:
+ /*
+ * Only set version for non-zpool-creation cases
+ * (set/import). spa_create() needs special care
+ * for version setting.
+ */
+ if (tx->tx_txg != TXG_INITIAL) {
+ VERIFY(nvpair_value_uint64(elem,
+ &intval) == 0);
+ ASSERT(intval <= SPA_VERSION);
+ ASSERT(intval >= spa_version(spa));
+ spa->spa_uberblock.ub_version = intval;
+ vdev_config_dirty(spa->spa_root_vdev);
+ }
+ break;
+
+ case ZPOOL_PROP_ALTROOT:
+ /*
+ * 'altroot' is a non-persistent property. It should
+ * have been set temporarily at creation or import time.
+ */
+ ASSERT(spa->spa_root != NULL);
+ break;
+
+ case ZPOOL_PROP_CACHEFILE:
+ /*
+ * 'cachefile' is a non-persistent property, but note
+ * an async request that the config cache needs to be
+ * udpated.
+ */
+ VERIFY(nvpair_value_string(elem, &strval) == 0);
+
+ dp = kmem_alloc(sizeof (spa_config_dirent_t), KM_SLEEP);
+
+ if (strval[0] == '\0')
+ dp->scd_path = spa_strdup(spa_config_path);
+ else if (strcmp(strval, "none") == 0)
+ dp->scd_path = NULL;
+ else
+ dp->scd_path = spa_strdup(strval);
+
+ list_insert_head(&spa->spa_config_list, dp);
+ spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
+ break;
+ default:
+ /*
+ * Set pool property values in the poolprops mos object.
+ */
+ if (spa->spa_pool_props_object == 0) {
+ objset_t *mos = spa->spa_meta_objset;
+
+ VERIFY((spa->spa_pool_props_object =
+ zap_create(mos, DMU_OT_POOL_PROPS,
+ DMU_OT_NONE, 0, tx)) > 0);
+
+ VERIFY(zap_update(mos,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
+ 8, 1, &spa->spa_pool_props_object, tx)
+ == 0);
+ }
+
+ /* normalize the property name */
+ propname = zpool_prop_to_name(prop);
+ proptype = zpool_prop_get_type(prop);
+
+ if (nvpair_type(elem) == DATA_TYPE_STRING) {
+ ASSERT(proptype == PROP_TYPE_STRING);
+ VERIFY(nvpair_value_string(elem, &strval) == 0);
+ VERIFY(zap_update(mos,
+ spa->spa_pool_props_object, propname,
+ 1, strlen(strval) + 1, strval, tx) == 0);
+
+ } else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
+ VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+
+ if (proptype == PROP_TYPE_INDEX) {
+ const char *unused;
+ VERIFY(zpool_prop_index_to_string(
+ prop, intval, &unused) == 0);
+ }
+ VERIFY(zap_update(mos,
+ spa->spa_pool_props_object, propname,
+ 8, 1, &intval, tx) == 0);
+ } else {
+ ASSERT(0); /* not allowed */
+ }
+
+ switch (prop) {
+ case ZPOOL_PROP_DELEGATION:
+ spa->spa_delegation = intval;
+ break;
+ case ZPOOL_PROP_BOOTFS:
+ spa->spa_bootfs = intval;
+ break;
+ case ZPOOL_PROP_FAILUREMODE:
+ spa->spa_failmode = intval;
+ break;
+ default:
+ break;
+ }
+ }
+
+ /* log internal history if this is not a zpool create */
+ if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY &&
+ tx->tx_txg != TXG_INITIAL) {
+ spa_history_internal_log(LOG_POOL_PROPSET,
+ spa, tx, cr, "%s %lld %s",
+ nvpair_name(elem), intval, spa_name(spa));
+ }
+ }
+
+ mutex_exit(&spa->spa_props_lock);
+}
+
+/*
+ * Sync the specified transaction group. New blocks may be dirtied as
+ * part of the process, so we iterate until it converges.
+ */
+void
+spa_sync(spa_t *spa, uint64_t txg)
+{
+ dsl_pool_t *dp = spa->spa_dsl_pool;
+ objset_t *mos = spa->spa_meta_objset;
+ bplist_t *bpl = &spa->spa_sync_bplist;
+ vdev_t *rvd = spa->spa_root_vdev;
+ vdev_t *vd;
+ dmu_tx_t *tx;
+ int dirty_vdevs;
+ int error;
+
+ /*
+ * Lock out configuration changes.
+ */
+ spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
+
+ spa->spa_syncing_txg = txg;
+ spa->spa_sync_pass = 0;
+
+ /*
+ * If there are any pending vdev state changes, convert them
+ * into config changes that go out with this transaction group.
+ */
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) {
+ vdev_state_clean(vd);
+ vdev_config_dirty(vd);
+ }
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ VERIFY(0 == bplist_open(bpl, mos, spa->spa_sync_bplist_obj));
+
+ tx = dmu_tx_create_assigned(dp, txg);
+
+ /*
+ * If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
+ * set spa_deflate if we have no raid-z vdevs.
+ */
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) {
+ int i;
+
+ for (i = 0; i < rvd->vdev_children; i++) {
+ vd = rvd->vdev_child[i];
+ if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE)
+ break;
+ }
+ if (i == rvd->vdev_children) {
+ spa->spa_deflate = TRUE;
+ VERIFY(0 == zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
+ sizeof (uint64_t), 1, &spa->spa_deflate, tx));
+ }
+ }
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
+ dsl_pool_create_origin(dp, tx);
+
+ /* Keeping the origin open increases spa_minref */
+ spa->spa_minref += 3;
+ }
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
+ dsl_pool_upgrade_clones(dp, tx);
+ }
+
+ /*
+ * If anything has changed in this txg, push the deferred frees
+ * from the previous txg. If not, leave them alone so that we
+ * don't generate work on an otherwise idle system.
+ */
+ if (!txg_list_empty(&dp->dp_dirty_datasets, txg) ||
+ !txg_list_empty(&dp->dp_dirty_dirs, txg) ||
+ !txg_list_empty(&dp->dp_sync_tasks, txg))
+ spa_sync_deferred_frees(spa, txg);
+
+ /*
+ * Iterate to convergence.
+ */
+ do {
+ spa->spa_sync_pass++;
+
+ spa_sync_config_object(spa, tx);
+ spa_sync_aux_dev(spa, &spa->spa_spares, tx,
+ ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES);
+ spa_sync_aux_dev(spa, &spa->spa_l2cache, tx,
+ ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE);
+ spa_errlog_sync(spa, txg);
+ dsl_pool_sync(dp, txg);
+
+ dirty_vdevs = 0;
+ while (vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) {
+ vdev_sync(vd, txg);
+ dirty_vdevs++;
+ }
+
+ bplist_sync(bpl, tx);
+ } while (dirty_vdevs);
+
+ bplist_close(bpl);
+
+ dprintf("txg %llu passes %d\n", txg, spa->spa_sync_pass);
+
+ /*
+ * Rewrite the vdev configuration (which includes the uberblock)
+ * to commit the transaction group.
+ *
+ * If there are no dirty vdevs, we sync the uberblock to a few
+ * random top-level vdevs that are known to be visible in the
+ * config cache (see spa_vdev_add() for a complete description).
+ * If there *are* dirty vdevs, sync the uberblock to all vdevs.
+ */
+ for (;;) {
+ /*
+ * We hold SCL_STATE to prevent vdev open/close/etc.
+ * while we're attempting to write the vdev labels.
+ */
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+
+ if (list_is_empty(&spa->spa_config_dirty_list)) {
+ vdev_t *svd[SPA_DVAS_PER_BP];
+ int svdcount = 0;
+ int children = rvd->vdev_children;
+ int c0 = spa_get_random(children);
+ int c;
+
+ for (c = 0; c < children; c++) {
+ vd = rvd->vdev_child[(c0 + c) % children];
+ if (vd->vdev_ms_array == 0 || vd->vdev_islog)
+ continue;
+ svd[svdcount++] = vd;
+ if (svdcount == SPA_DVAS_PER_BP)
+ break;
+ }
+ error = vdev_config_sync(svd, svdcount, txg);
+ } else {
+ error = vdev_config_sync(rvd->vdev_child,
+ rvd->vdev_children, txg);
+ }
+
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ if (error == 0)
+ break;
+ zio_suspend(spa, NULL);
+ zio_resume_wait(spa);
+ }
+ dmu_tx_commit(tx);
+
+ /*
+ * Clear the dirty config list.
+ */
+ while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL)
+ vdev_config_clean(vd);
+
+ /*
+ * Now that the new config has synced transactionally,
+ * let it become visible to the config cache.
+ */
+ if (spa->spa_config_syncing != NULL) {
+ spa_config_set(spa, spa->spa_config_syncing);
+ spa->spa_config_txg = txg;
+ spa->spa_config_syncing = NULL;
+ }
+
+ spa->spa_ubsync = spa->spa_uberblock;
+
+ /*
+ * Clean up the ZIL records for the synced txg.
+ */
+ dsl_pool_zil_clean(dp);
+
+ /*
+ * Update usable space statistics.
+ */
+ while (vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))
+ vdev_sync_done(vd, txg);
+
+ /*
+ * It had better be the case that we didn't dirty anything
+ * since vdev_config_sync().
+ */
+ ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
+ ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
+ ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg));
+ ASSERT(bpl->bpl_queue == NULL);
+
+ spa_config_exit(spa, SCL_CONFIG, FTAG);
+
+ /*
+ * If any async tasks have been requested, kick them off.
+ */
+ spa_async_dispatch(spa);
+}
+
+/*
+ * Sync all pools. We don't want to hold the namespace lock across these
+ * operations, so we take a reference on the spa_t and drop the lock during the
+ * sync.
+ */
+void
+spa_sync_allpools(void)
+{
+ spa_t *spa = NULL;
+ mutex_enter(&spa_namespace_lock);
+ while ((spa = spa_next(spa)) != NULL) {
+ if (spa_state(spa) != POOL_STATE_ACTIVE || spa_suspended(spa))
+ continue;
+ spa_open_ref(spa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+ txg_wait_synced(spa_get_dsl(spa), 0);
+ mutex_enter(&spa_namespace_lock);
+ spa_close(spa, FTAG);
+ }
+ mutex_exit(&spa_namespace_lock);
+}
+
+/*
+ * ==========================================================================
+ * Miscellaneous routines
+ * ==========================================================================
+ */
+
+/*
+ * Remove all pools in the system.
+ */
+void
+spa_evict_all(void)
+{
+ spa_t *spa;
+
+ /*
+ * Remove all cached state. All pools should be closed now,
+ * so every spa in the AVL tree should be unreferenced.
+ */
+ mutex_enter(&spa_namespace_lock);
+ while ((spa = spa_next(NULL)) != NULL) {
+ /*
+ * Stop async tasks. The async thread may need to detach
+ * a device that's been replaced, which requires grabbing
+ * spa_namespace_lock, so we must drop it here.
+ */
+ spa_open_ref(spa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+ spa_async_suspend(spa);
+ mutex_enter(&spa_namespace_lock);
+ spa_close(spa, FTAG);
+
+ if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
+ spa_unload(spa);
+ spa_deactivate(spa);
+ }
+ spa_remove(spa);
+ }
+ mutex_exit(&spa_namespace_lock);
+}
+
+vdev_t *
+spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t l2cache)
+{
+ vdev_t *vd;
+ int i;
+
+ if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL)
+ return (vd);
+
+ if (l2cache) {
+ for (i = 0; i < spa->spa_l2cache.sav_count; i++) {
+ vd = spa->spa_l2cache.sav_vdevs[i];
+ if (vd->vdev_guid == guid)
+ return (vd);
+ }
+ }
+
+ return (NULL);
+}
+
+void
+spa_upgrade(spa_t *spa, uint64_t version)
+{
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+ /*
+ * This should only be called for a non-faulted pool, and since a
+ * future version would result in an unopenable pool, this shouldn't be
+ * possible.
+ */
+ ASSERT(spa->spa_uberblock.ub_version <= SPA_VERSION);
+ ASSERT(version >= spa->spa_uberblock.ub_version);
+
+ spa->spa_uberblock.ub_version = version;
+ vdev_config_dirty(spa->spa_root_vdev);
+
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ txg_wait_synced(spa_get_dsl(spa), 0);
+}
+
+boolean_t
+spa_has_spare(spa_t *spa, uint64_t guid)
+{
+ int i;
+ uint64_t spareguid;
+ spa_aux_vdev_t *sav = &spa->spa_spares;
+
+ for (i = 0; i < sav->sav_count; i++)
+ if (sav->sav_vdevs[i]->vdev_guid == guid)
+ return (B_TRUE);
+
+ for (i = 0; i < sav->sav_npending; i++) {
+ if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID,
+ &spareguid) == 0 && spareguid == guid)
+ return (B_TRUE);
+ }
+
+ return (B_FALSE);
+}
+
+/*
+ * Check if a pool has an active shared spare device.
+ * Note: reference count of an active spare is 2, as a spare and as a replace
+ */
+static boolean_t
+spa_has_active_shared_spare(spa_t *spa)
+{
+ int i, refcnt;
+ uint64_t pool;
+ spa_aux_vdev_t *sav = &spa->spa_spares;
+
+ for (i = 0; i < sav->sav_count; i++) {
+ if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool,
+ &refcnt) && pool != 0ULL && pool == spa_guid(spa) &&
+ refcnt > 2)
+ return (B_TRUE);
+ }
+
+ return (B_FALSE);
+}
+
+/*
+ * Post a sysevent corresponding to the given event. The 'name' must be one of
+ * the event definitions in sys/sysevent/eventdefs.h. The payload will be
+ * filled in from the spa and (optionally) the vdev. This doesn't do anything
+ * in the userland libzpool, as we don't want consumers to misinterpret ztest
+ * or zdb as real changes.
+ */
+void
+spa_event_notify(spa_t *spa, vdev_t *vd, const char *name)
+{
+#ifdef _KERNEL
+ sysevent_t *ev;
+ sysevent_attr_list_t *attr = NULL;
+ sysevent_value_t value;
+ sysevent_id_t eid;
+
+ ev = sysevent_alloc(EC_ZFS, (char *)name, SUNW_KERN_PUB "zfs",
+ SE_SLEEP);
+
+ value.value_type = SE_DATA_TYPE_STRING;
+ value.value.sv_string = spa_name(spa);
+ if (sysevent_add_attr(&attr, ZFS_EV_POOL_NAME, &value, SE_SLEEP) != 0)
+ goto done;
+
+ value.value_type = SE_DATA_TYPE_UINT64;
+ value.value.sv_uint64 = spa_guid(spa);
+ if (sysevent_add_attr(&attr, ZFS_EV_POOL_GUID, &value, SE_SLEEP) != 0)
+ goto done;
+
+ if (vd) {
+ value.value_type = SE_DATA_TYPE_UINT64;
+ value.value.sv_uint64 = vd->vdev_guid;
+ if (sysevent_add_attr(&attr, ZFS_EV_VDEV_GUID, &value,
+ SE_SLEEP) != 0)
+ goto done;
+
+ if (vd->vdev_path) {
+ value.value_type = SE_DATA_TYPE_STRING;
+ value.value.sv_string = vd->vdev_path;
+ if (sysevent_add_attr(&attr, ZFS_EV_VDEV_PATH,
+ &value, SE_SLEEP) != 0)
+ goto done;
+ }
+ }
+
+ if (sysevent_attach_attributes(ev, attr) != 0)
+ goto done;
+ attr = NULL;
+
+ (void) log_sysevent(ev, SE_SLEEP, &eid);
+
+done:
+ if (attr)
+ sysevent_free_attr(attr);
+ sysevent_free(ev);
+#endif
+}
diff --git a/module/zfs/spa_boot.c b/module/zfs/spa_boot.c
new file mode 100644
index 000000000..49e9e5019
--- /dev/null
+++ b/module/zfs/spa_boot.c
@@ -0,0 +1,47 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/sunddi.h>
+
+char *
+spa_get_bootprop(char *propname)
+{
+ char *value;
+
+ if (ddi_prop_lookup_string(DDI_DEV_T_ANY, ddi_root_node(),
+ DDI_PROP_DONTPASS, propname, &value) != DDI_SUCCESS)
+ return (NULL);
+ return (value);
+}
+
+void
+spa_free_bootprop(char *value)
+{
+ ddi_prop_free(value);
+}
diff --git a/module/zfs/spa_config.c b/module/zfs/spa_config.c
new file mode 100644
index 000000000..ee425a916
--- /dev/null
+++ b/module/zfs/spa_config.c
@@ -0,0 +1,444 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/nvpair.h>
+#include <sys/uio.h>
+#include <sys/fs/zfs.h>
+#include <sys/vdev_impl.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/utsname.h>
+#include <sys/systeminfo.h>
+#include <sys/sunddi.h>
+#ifdef _KERNEL
+#include <sys/kobj.h>
+#endif
+
+/*
+ * Pool configuration repository.
+ *
+ * Pool configuration is stored as a packed nvlist on the filesystem. By
+ * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
+ * (when the ZFS module is loaded). Pools can also have the 'cachefile'
+ * property set that allows them to be stored in an alternate location until
+ * the control of external software.
+ *
+ * For each cache file, we have a single nvlist which holds all the
+ * configuration information. When the module loads, we read this information
+ * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
+ * maintained independently in spa.c. Whenever the namespace is modified, or
+ * the configuration of a pool is changed, we call spa_config_sync(), which
+ * walks through all the active pools and writes the configuration to disk.
+ */
+
+static uint64_t spa_config_generation = 1;
+
+/*
+ * This can be overridden in userland to preserve an alternate namespace for
+ * userland pools when doing testing.
+ */
+const char *spa_config_path = ZPOOL_CACHE;
+
+/*
+ * Called when the module is first loaded, this routine loads the configuration
+ * file into the SPA namespace. It does not actually open or load the pools; it
+ * only populates the namespace.
+ */
+void
+spa_config_load(void)
+{
+ void *buf = NULL;
+ nvlist_t *nvlist, *child;
+ nvpair_t *nvpair;
+ spa_t *spa;
+ char *pathname;
+ struct _buf *file;
+ uint64_t fsize;
+
+ /*
+ * Open the configuration file.
+ */
+ pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+
+ (void) snprintf(pathname, MAXPATHLEN, "%s%s",
+ (rootdir != NULL) ? "./" : "", spa_config_path);
+
+ file = kobj_open_file(pathname);
+
+ kmem_free(pathname, MAXPATHLEN);
+
+ if (file == (struct _buf *)-1)
+ return;
+
+ if (kobj_get_filesize(file, &fsize) != 0)
+ goto out;
+
+ buf = kmem_alloc(fsize, KM_SLEEP);
+
+ /*
+ * Read the nvlist from the file.
+ */
+ if (kobj_read_file(file, buf, fsize, 0) < 0)
+ goto out;
+
+ /*
+ * Unpack the nvlist.
+ */
+ if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
+ goto out;
+
+ /*
+ * Iterate over all elements in the nvlist, creating a new spa_t for
+ * each one with the specified configuration.
+ */
+ mutex_enter(&spa_namespace_lock);
+ nvpair = NULL;
+ while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
+
+ if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
+ continue;
+
+ VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
+
+ if (spa_lookup(nvpair_name(nvpair)) != NULL)
+ continue;
+ spa = spa_add(nvpair_name(nvpair), NULL);
+
+ /*
+ * We blindly duplicate the configuration here. If it's
+ * invalid, we will catch it when the pool is first opened.
+ */
+ VERIFY(nvlist_dup(child, &spa->spa_config, 0) == 0);
+ }
+ mutex_exit(&spa_namespace_lock);
+
+ nvlist_free(nvlist);
+
+out:
+ if (buf != NULL)
+ kmem_free(buf, fsize);
+
+ kobj_close_file(file);
+}
+
+static void
+spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
+{
+ size_t buflen;
+ char *buf;
+ vnode_t *vp;
+ int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX;
+ char *temp;
+
+ /*
+ * If the nvlist is empty (NULL), then remove the old cachefile.
+ */
+ if (nvl == NULL) {
+ (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE);
+ return;
+ }
+
+ /*
+ * Pack the configuration into a buffer.
+ */
+ VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
+
+ buf = kmem_alloc(buflen, KM_SLEEP);
+ temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
+
+ VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
+ KM_SLEEP) == 0);
+
+ /*
+ * Write the configuration to disk. We need to do the traditional
+ * 'write to temporary file, sync, move over original' to make sure we
+ * always have a consistent view of the data.
+ */
+ (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
+
+ if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
+ if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
+ 0, RLIM64_INFINITY, kcred, NULL) == 0 &&
+ VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {
+ (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE);
+ }
+ (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL);
+ VN_RELE(vp);
+ }
+
+ (void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
+
+ kmem_free(buf, buflen);
+ kmem_free(temp, MAXPATHLEN);
+}
+
+/*
+ * Synchronize pool configuration to disk. This must be called with the
+ * namespace lock held.
+ */
+void
+spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
+{
+ spa_config_dirent_t *dp, *tdp;
+ nvlist_t *nvl;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ /*
+ * Iterate over all cachefiles for the pool, past or present. When the
+ * cachefile is changed, the new one is pushed onto this list, allowing
+ * us to update previous cachefiles that no longer contain this pool.
+ */
+ for (dp = list_head(&target->spa_config_list); dp != NULL;
+ dp = list_next(&target->spa_config_list, dp)) {
+ spa_t *spa = NULL;
+ if (dp->scd_path == NULL)
+ continue;
+
+ /*
+ * Iterate over all pools, adding any matching pools to 'nvl'.
+ */
+ nvl = NULL;
+ while ((spa = spa_next(spa)) != NULL) {
+ if (spa == target && removing)
+ continue;
+
+ mutex_enter(&spa->spa_props_lock);
+ tdp = list_head(&spa->spa_config_list);
+ if (spa->spa_config == NULL ||
+ tdp->scd_path == NULL ||
+ strcmp(tdp->scd_path, dp->scd_path) != 0) {
+ mutex_exit(&spa->spa_props_lock);
+ continue;
+ }
+
+ if (nvl == NULL)
+ VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
+ KM_SLEEP) == 0);
+
+ VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
+ spa->spa_config) == 0);
+ mutex_exit(&spa->spa_props_lock);
+ }
+
+ spa_config_write(dp, nvl);
+ nvlist_free(nvl);
+ }
+
+ /*
+ * Remove any config entries older than the current one.
+ */
+ dp = list_head(&target->spa_config_list);
+ while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
+ list_remove(&target->spa_config_list, tdp);
+ if (tdp->scd_path != NULL)
+ spa_strfree(tdp->scd_path);
+ kmem_free(tdp, sizeof (spa_config_dirent_t));
+ }
+
+ spa_config_generation++;
+
+ if (postsysevent)
+ spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC);
+}
+
+/*
+ * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
+ * and we don't want to allow the local zone to see all the pools anyway.
+ * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
+ * information for all pool visible within the zone.
+ */
+nvlist_t *
+spa_all_configs(uint64_t *generation)
+{
+ nvlist_t *pools;
+ spa_t *spa = NULL;
+
+ if (*generation == spa_config_generation)
+ return (NULL);
+
+ VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ mutex_enter(&spa_namespace_lock);
+ while ((spa = spa_next(spa)) != NULL) {
+ if (INGLOBALZONE(curproc) ||
+ zone_dataset_visible(spa_name(spa), NULL)) {
+ mutex_enter(&spa->spa_props_lock);
+ VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
+ spa->spa_config) == 0);
+ mutex_exit(&spa->spa_props_lock);
+ }
+ }
+ *generation = spa_config_generation;
+ mutex_exit(&spa_namespace_lock);
+
+ return (pools);
+}
+
+void
+spa_config_set(spa_t *spa, nvlist_t *config)
+{
+ mutex_enter(&spa->spa_props_lock);
+ if (spa->spa_config != NULL)
+ nvlist_free(spa->spa_config);
+ spa->spa_config = config;
+ mutex_exit(&spa->spa_props_lock);
+}
+
+/*
+ * Generate the pool's configuration based on the current in-core state.
+ * We infer whether to generate a complete config or just one top-level config
+ * based on whether vd is the root vdev.
+ */
+nvlist_t *
+spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
+{
+ nvlist_t *config, *nvroot;
+ vdev_t *rvd = spa->spa_root_vdev;
+ unsigned long hostid = 0;
+ boolean_t locked = B_FALSE;
+
+ if (vd == NULL) {
+ vd = rvd;
+ locked = B_TRUE;
+ spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
+ }
+
+ ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
+ (SCL_CONFIG | SCL_STATE));
+
+ /*
+ * If txg is -1, report the current value of spa->spa_config_txg.
+ */
+ if (txg == -1ULL)
+ txg = spa->spa_config_txg;
+
+ VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
+ spa_version(spa)) == 0);
+ VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
+ spa_name(spa)) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
+ spa_state(spa)) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
+ txg) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
+ spa_guid(spa)) == 0);
+ (void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
+ if (hostid != 0) {
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
+ hostid) == 0);
+ }
+ VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
+ utsname.nodename) == 0);
+
+ if (vd != rvd) {
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
+ vd->vdev_top->vdev_guid) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
+ vd->vdev_guid) == 0);
+ if (vd->vdev_isspare)
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
+ 1ULL) == 0);
+ if (vd->vdev_islog)
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
+ 1ULL) == 0);
+ vd = vd->vdev_top; /* label contains top config */
+ }
+
+ nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE, B_FALSE);
+ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
+ nvlist_free(nvroot);
+
+ if (locked)
+ spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
+
+ return (config);
+}
+
+/*
+ * For a pool that's not currently a booting rootpool, update all disk labels,
+ * generate a fresh config based on the current in-core state, and sync the
+ * global config cache.
+ */
+void
+spa_config_update(spa_t *spa, int what)
+{
+ spa_config_update_common(spa, what, FALSE);
+}
+
+/*
+ * Update all disk labels, generate a fresh config based on the current
+ * in-core state, and sync the global config cache (do not sync the config
+ * cache if this is a booting rootpool).
+ */
+void
+spa_config_update_common(spa_t *spa, int what, boolean_t isroot)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+ uint64_t txg;
+ int c;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ txg = spa_last_synced_txg(spa) + 1;
+ if (what == SPA_CONFIG_UPDATE_POOL) {
+ vdev_config_dirty(rvd);
+ } else {
+ /*
+ * If we have top-level vdevs that were added but have
+ * not yet been prepared for allocation, do that now.
+ * (It's safe now because the config cache is up to date,
+ * so it will be able to translate the new DVAs.)
+ * See comments in spa_vdev_add() for full details.
+ */
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ if (tvd->vdev_ms_array == 0) {
+ vdev_init(tvd, txg);
+ vdev_config_dirty(tvd);
+ }
+ }
+ }
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ /*
+ * Wait for the mosconfig to be regenerated and synced.
+ */
+ txg_wait_synced(spa->spa_dsl_pool, txg);
+
+ /*
+ * Update the global config cache to reflect the new mosconfig.
+ */
+ if (!isroot)
+ spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL);
+
+ if (what == SPA_CONFIG_UPDATE_POOL)
+ spa_config_update_common(spa, SPA_CONFIG_UPDATE_VDEVS, isroot);
+}
diff --git a/module/zfs/spa_errlog.c b/module/zfs/spa_errlog.c
new file mode 100644
index 000000000..c642bd768
--- /dev/null
+++ b/module/zfs/spa_errlog.c
@@ -0,0 +1,437 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * Routines to manage the on-disk persistent error log.
+ *
+ * Each pool stores a log of all logical data errors seen during normal
+ * operation. This is actually the union of two distinct logs: the last log,
+ * and the current log. All errors seen are logged to the current log. When a
+ * scrub completes, the current log becomes the last log, the last log is thrown
+ * out, and the current log is reinitialized. This way, if an error is somehow
+ * corrected, a new scrub will show that that it no longer exists, and will be
+ * deleted from the log when the scrub completes.
+ *
+ * The log is stored using a ZAP object whose key is a string form of the
+ * zbookmark tuple (objset, object, level, blkid), and whose contents is an
+ * optional 'objset:object' human-readable string describing the data. When an
+ * error is first logged, this string will be empty, indicating that no name is
+ * known. This prevents us from having to issue a potentially large amount of
+ * I/O to discover the object name during an error path. Instead, we do the
+ * calculation when the data is requested, storing the result so future queries
+ * will be faster.
+ *
+ * This log is then shipped into an nvlist where the key is the dataset name and
+ * the value is the object name. Userland is then responsible for uniquifying
+ * this list and displaying it to the user.
+ */
+
+#include <sys/dmu_tx.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/zap.h>
+#include <sys/zio.h>
+
+/*
+ * This is a stripped-down version of strtoull, suitable only for converting
+ * lowercase hexidecimal numbers that don't overflow.
+ */
+#ifdef _KERNEL
+static uint64_t
+strtonum(char *str, char **nptr)
+{
+ uint64_t val = 0;
+ char c;
+ int digit;
+
+ while ((c = *str) != '\0') {
+ if (c >= '0' && c <= '9')
+ digit = c - '0';
+ else if (c >= 'a' && c <= 'f')
+ digit = 10 + c - 'a';
+ else
+ break;
+
+ val *= 16;
+ val += digit;
+
+ str++;
+ }
+
+ *nptr = str;
+
+ return (val);
+}
+#endif
+
+/*
+ * Convert a bookmark to a string.
+ */
+static void
+bookmark_to_name(zbookmark_t *zb, char *buf, size_t len)
+{
+ (void) snprintf(buf, len, "%llx:%llx:%llx:%llx",
+ (u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object,
+ (u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid);
+}
+
+/*
+ * Convert a string to a bookmark
+ */
+#ifdef _KERNEL
+static void
+name_to_bookmark(char *buf, zbookmark_t *zb)
+{
+ zb->zb_objset = strtonum(buf, &buf);
+ ASSERT(*buf == ':');
+ zb->zb_object = strtonum(buf + 1, &buf);
+ ASSERT(*buf == ':');
+ zb->zb_level = (int)strtonum(buf + 1, &buf);
+ ASSERT(*buf == ':');
+ zb->zb_blkid = strtonum(buf + 1, &buf);
+ ASSERT(*buf == '\0');
+}
+#endif
+
+/*
+ * Log an uncorrectable error to the persistent error log. We add it to the
+ * spa's list of pending errors. The changes are actually synced out to disk
+ * during spa_errlog_sync().
+ */
+void
+spa_log_error(spa_t *spa, zio_t *zio)
+{
+ zbookmark_t *zb = &zio->io_logical->io_bookmark;
+ spa_error_entry_t search;
+ spa_error_entry_t *new;
+ avl_tree_t *tree;
+ avl_index_t where;
+
+ /*
+ * If we are trying to import a pool, ignore any errors, as we won't be
+ * writing to the pool any time soon.
+ */
+ if (spa->spa_load_state == SPA_LOAD_TRYIMPORT)
+ return;
+
+ mutex_enter(&spa->spa_errlist_lock);
+
+ /*
+ * If we have had a request to rotate the log, log it to the next list
+ * instead of the current one.
+ */
+ if (spa->spa_scrub_active || spa->spa_scrub_finished)
+ tree = &spa->spa_errlist_scrub;
+ else
+ tree = &spa->spa_errlist_last;
+
+ search.se_bookmark = *zb;
+ if (avl_find(tree, &search, &where) != NULL) {
+ mutex_exit(&spa->spa_errlist_lock);
+ return;
+ }
+
+ new = kmem_zalloc(sizeof (spa_error_entry_t), KM_SLEEP);
+ new->se_bookmark = *zb;
+ avl_insert(tree, new, where);
+
+ mutex_exit(&spa->spa_errlist_lock);
+}
+
+/*
+ * Return the number of errors currently in the error log. This is actually the
+ * sum of both the last log and the current log, since we don't know the union
+ * of these logs until we reach userland.
+ */
+uint64_t
+spa_get_errlog_size(spa_t *spa)
+{
+ uint64_t total = 0, count;
+
+ mutex_enter(&spa->spa_errlog_lock);
+ if (spa->spa_errlog_scrub != 0 &&
+ zap_count(spa->spa_meta_objset, spa->spa_errlog_scrub,
+ &count) == 0)
+ total += count;
+
+ if (spa->spa_errlog_last != 0 && !spa->spa_scrub_finished &&
+ zap_count(spa->spa_meta_objset, spa->spa_errlog_last,
+ &count) == 0)
+ total += count;
+ mutex_exit(&spa->spa_errlog_lock);
+
+ mutex_enter(&spa->spa_errlist_lock);
+ total += avl_numnodes(&spa->spa_errlist_last);
+ total += avl_numnodes(&spa->spa_errlist_scrub);
+ mutex_exit(&spa->spa_errlist_lock);
+
+ return (total);
+}
+
+#ifdef _KERNEL
+static int
+process_error_log(spa_t *spa, uint64_t obj, void *addr, size_t *count)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ zbookmark_t zb;
+
+ if (obj == 0)
+ return (0);
+
+ for (zap_cursor_init(&zc, spa->spa_meta_objset, obj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+
+ if (*count == 0) {
+ zap_cursor_fini(&zc);
+ return (ENOMEM);
+ }
+
+ name_to_bookmark(za.za_name, &zb);
+
+ if (copyout(&zb, (char *)addr +
+ (*count - 1) * sizeof (zbookmark_t),
+ sizeof (zbookmark_t)) != 0)
+ return (EFAULT);
+
+ *count -= 1;
+ }
+
+ zap_cursor_fini(&zc);
+
+ return (0);
+}
+
+static int
+process_error_list(avl_tree_t *list, void *addr, size_t *count)
+{
+ spa_error_entry_t *se;
+
+ for (se = avl_first(list); se != NULL; se = AVL_NEXT(list, se)) {
+
+ if (*count == 0)
+ return (ENOMEM);
+
+ if (copyout(&se->se_bookmark, (char *)addr +
+ (*count - 1) * sizeof (zbookmark_t),
+ sizeof (zbookmark_t)) != 0)
+ return (EFAULT);
+
+ *count -= 1;
+ }
+
+ return (0);
+}
+#endif
+
+/*
+ * Copy all known errors to userland as an array of bookmarks. This is
+ * actually a union of the on-disk last log and current log, as well as any
+ * pending error requests.
+ *
+ * Because the act of reading the on-disk log could cause errors to be
+ * generated, we have two separate locks: one for the error log and one for the
+ * in-core error lists. We only need the error list lock to log and error, so
+ * we grab the error log lock while we read the on-disk logs, and only pick up
+ * the error list lock when we are finished.
+ */
+int
+spa_get_errlog(spa_t *spa, void *uaddr, size_t *count)
+{
+ int ret = 0;
+
+#ifdef _KERNEL
+ mutex_enter(&spa->spa_errlog_lock);
+
+ ret = process_error_log(spa, spa->spa_errlog_scrub, uaddr, count);
+
+ if (!ret && !spa->spa_scrub_finished)
+ ret = process_error_log(spa, spa->spa_errlog_last, uaddr,
+ count);
+
+ mutex_enter(&spa->spa_errlist_lock);
+ if (!ret)
+ ret = process_error_list(&spa->spa_errlist_scrub, uaddr,
+ count);
+ if (!ret)
+ ret = process_error_list(&spa->spa_errlist_last, uaddr,
+ count);
+ mutex_exit(&spa->spa_errlist_lock);
+
+ mutex_exit(&spa->spa_errlog_lock);
+#endif
+
+ return (ret);
+}
+
+/*
+ * Called when a scrub completes. This simply set a bit which tells which AVL
+ * tree to add new errors. spa_errlog_sync() is responsible for actually
+ * syncing the changes to the underlying objects.
+ */
+void
+spa_errlog_rotate(spa_t *spa)
+{
+ mutex_enter(&spa->spa_errlist_lock);
+ spa->spa_scrub_finished = B_TRUE;
+ mutex_exit(&spa->spa_errlist_lock);
+}
+
+/*
+ * Discard any pending errors from the spa_t. Called when unloading a faulted
+ * pool, as the errors encountered during the open cannot be synced to disk.
+ */
+void
+spa_errlog_drain(spa_t *spa)
+{
+ spa_error_entry_t *se;
+ void *cookie;
+
+ mutex_enter(&spa->spa_errlist_lock);
+
+ cookie = NULL;
+ while ((se = avl_destroy_nodes(&spa->spa_errlist_last,
+ &cookie)) != NULL)
+ kmem_free(se, sizeof (spa_error_entry_t));
+ cookie = NULL;
+ while ((se = avl_destroy_nodes(&spa->spa_errlist_scrub,
+ &cookie)) != NULL)
+ kmem_free(se, sizeof (spa_error_entry_t));
+
+ mutex_exit(&spa->spa_errlist_lock);
+}
+
+/*
+ * Process a list of errors into the current on-disk log.
+ */
+static void
+sync_error_list(spa_t *spa, avl_tree_t *t, uint64_t *obj, dmu_tx_t *tx)
+{
+ spa_error_entry_t *se;
+ char buf[64];
+ void *cookie;
+
+ if (avl_numnodes(t) != 0) {
+ /* create log if necessary */
+ if (*obj == 0)
+ *obj = zap_create(spa->spa_meta_objset,
+ DMU_OT_ERROR_LOG, DMU_OT_NONE,
+ 0, tx);
+
+ /* add errors to the current log */
+ for (se = avl_first(t); se != NULL; se = AVL_NEXT(t, se)) {
+ char *name = se->se_name ? se->se_name : "";
+
+ bookmark_to_name(&se->se_bookmark, buf, sizeof (buf));
+
+ (void) zap_update(spa->spa_meta_objset,
+ *obj, buf, 1, strlen(name) + 1, name, tx);
+ }
+
+ /* purge the error list */
+ cookie = NULL;
+ while ((se = avl_destroy_nodes(t, &cookie)) != NULL)
+ kmem_free(se, sizeof (spa_error_entry_t));
+ }
+}
+
+/*
+ * Sync the error log out to disk. This is a little tricky because the act of
+ * writing the error log requires the spa_errlist_lock. So, we need to lock the
+ * error lists, take a copy of the lists, and then reinitialize them. Then, we
+ * drop the error list lock and take the error log lock, at which point we
+ * do the errlog processing. Then, if we encounter an I/O error during this
+ * process, we can successfully add the error to the list. Note that this will
+ * result in the perpetual recycling of errors, but it is an unlikely situation
+ * and not a performance critical operation.
+ */
+void
+spa_errlog_sync(spa_t *spa, uint64_t txg)
+{
+ dmu_tx_t *tx;
+ avl_tree_t scrub, last;
+ int scrub_finished;
+
+ mutex_enter(&spa->spa_errlist_lock);
+
+ /*
+ * Bail out early under normal circumstances.
+ */
+ if (avl_numnodes(&spa->spa_errlist_scrub) == 0 &&
+ avl_numnodes(&spa->spa_errlist_last) == 0 &&
+ !spa->spa_scrub_finished) {
+ mutex_exit(&spa->spa_errlist_lock);
+ return;
+ }
+
+ spa_get_errlists(spa, &last, &scrub);
+ scrub_finished = spa->spa_scrub_finished;
+ spa->spa_scrub_finished = B_FALSE;
+
+ mutex_exit(&spa->spa_errlist_lock);
+ mutex_enter(&spa->spa_errlog_lock);
+
+ tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+
+ /*
+ * Sync out the current list of errors.
+ */
+ sync_error_list(spa, &last, &spa->spa_errlog_last, tx);
+
+ /*
+ * Rotate the log if necessary.
+ */
+ if (scrub_finished) {
+ if (spa->spa_errlog_last != 0)
+ VERIFY(dmu_object_free(spa->spa_meta_objset,
+ spa->spa_errlog_last, tx) == 0);
+ spa->spa_errlog_last = spa->spa_errlog_scrub;
+ spa->spa_errlog_scrub = 0;
+
+ sync_error_list(spa, &scrub, &spa->spa_errlog_last, tx);
+ }
+
+ /*
+ * Sync out any pending scrub errors.
+ */
+ sync_error_list(spa, &scrub, &spa->spa_errlog_scrub, tx);
+
+ /*
+ * Update the MOS to reflect the new values.
+ */
+ (void) zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_ERRLOG_LAST, sizeof (uint64_t), 1,
+ &spa->spa_errlog_last, tx);
+ (void) zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_ERRLOG_SCRUB, sizeof (uint64_t), 1,
+ &spa->spa_errlog_scrub, tx);
+
+ dmu_tx_commit(tx);
+
+ mutex_exit(&spa->spa_errlog_lock);
+}
diff --git a/module/zfs/spa_history.c b/module/zfs/spa_history.c
new file mode 100644
index 000000000..c997240c1
--- /dev/null
+++ b/module/zfs/spa_history.c
@@ -0,0 +1,428 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/zap.h>
+#include <sys/dsl_synctask.h>
+#include <sys/dmu_tx.h>
+#include <sys/dmu_objset.h>
+#include <sys/utsname.h>
+#include <sys/cmn_err.h>
+#include <sys/sunddi.h>
+#ifdef _KERNEL
+#include <sys/zone.h>
+#endif
+
+/*
+ * Routines to manage the on-disk history log.
+ *
+ * The history log is stored as a dmu object containing
+ * <packed record length, record nvlist> tuples.
+ *
+ * Where "record nvlist" is a nvlist containing uint64_ts and strings, and
+ * "packed record length" is the packed length of the "record nvlist" stored
+ * as a little endian uint64_t.
+ *
+ * The log is implemented as a ring buffer, though the original creation
+ * of the pool ('zpool create') is never overwritten.
+ *
+ * The history log is tracked as object 'spa_t::spa_history'. The bonus buffer
+ * of 'spa_history' stores the offsets for logging/retrieving history as
+ * 'spa_history_phys_t'. 'sh_pool_create_len' is the ending offset in bytes of
+ * where the 'zpool create' record is stored. This allows us to never
+ * overwrite the original creation of the pool. 'sh_phys_max_off' is the
+ * physical ending offset in bytes of the log. This tells you the length of
+ * the buffer. 'sh_eof' is the logical EOF (in bytes). Whenever a record
+ * is added, 'sh_eof' is incremented by the the size of the record.
+ * 'sh_eof' is never decremented. 'sh_bof' is the logical BOF (in bytes).
+ * This is where the consumer should start reading from after reading in
+ * the 'zpool create' portion of the log.
+ *
+ * 'sh_records_lost' keeps track of how many records have been overwritten
+ * and permanently lost.
+ */
+
+/* convert a logical offset to physical */
+static uint64_t
+spa_history_log_to_phys(uint64_t log_off, spa_history_phys_t *shpp)
+{
+ uint64_t phys_len;
+
+ phys_len = shpp->sh_phys_max_off - shpp->sh_pool_create_len;
+ return ((log_off - shpp->sh_pool_create_len) % phys_len
+ + shpp->sh_pool_create_len);
+}
+
+void
+spa_history_create_obj(spa_t *spa, dmu_tx_t *tx)
+{
+ dmu_buf_t *dbp;
+ spa_history_phys_t *shpp;
+ objset_t *mos = spa->spa_meta_objset;
+
+ ASSERT(spa->spa_history == 0);
+ spa->spa_history = dmu_object_alloc(mos, DMU_OT_SPA_HISTORY,
+ SPA_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS,
+ sizeof (spa_history_phys_t), tx);
+
+ VERIFY(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_HISTORY, sizeof (uint64_t), 1,
+ &spa->spa_history, tx) == 0);
+
+ VERIFY(0 == dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
+ ASSERT(dbp->db_size >= sizeof (spa_history_phys_t));
+
+ shpp = dbp->db_data;
+ dmu_buf_will_dirty(dbp, tx);
+
+ /*
+ * Figure out maximum size of history log. We set it at
+ * 1% of pool size, with a max of 32MB and min of 128KB.
+ */
+ shpp->sh_phys_max_off = spa_get_dspace(spa) / 100;
+ shpp->sh_phys_max_off = MIN(shpp->sh_phys_max_off, 32<<20);
+ shpp->sh_phys_max_off = MAX(shpp->sh_phys_max_off, 128<<10);
+
+ dmu_buf_rele(dbp, FTAG);
+}
+
+/*
+ * Change 'sh_bof' to the beginning of the next record.
+ */
+static int
+spa_history_advance_bof(spa_t *spa, spa_history_phys_t *shpp)
+{
+ objset_t *mos = spa->spa_meta_objset;
+ uint64_t firstread, reclen, phys_bof;
+ char buf[sizeof (reclen)];
+ int err;
+
+ phys_bof = spa_history_log_to_phys(shpp->sh_bof, shpp);
+ firstread = MIN(sizeof (reclen), shpp->sh_phys_max_off - phys_bof);
+
+ if ((err = dmu_read(mos, spa->spa_history, phys_bof, firstread,
+ buf)) != 0)
+ return (err);
+ if (firstread != sizeof (reclen)) {
+ if ((err = dmu_read(mos, spa->spa_history,
+ shpp->sh_pool_create_len, sizeof (reclen) - firstread,
+ buf + firstread)) != 0)
+ return (err);
+ }
+
+ reclen = LE_64(*((uint64_t *)buf));
+ shpp->sh_bof += reclen + sizeof (reclen);
+ shpp->sh_records_lost++;
+ return (0);
+}
+
+static int
+spa_history_write(spa_t *spa, void *buf, uint64_t len, spa_history_phys_t *shpp,
+ dmu_tx_t *tx)
+{
+ uint64_t firstwrite, phys_eof;
+ objset_t *mos = spa->spa_meta_objset;
+ int err;
+
+ ASSERT(MUTEX_HELD(&spa->spa_history_lock));
+
+ /* see if we need to reset logical BOF */
+ while (shpp->sh_phys_max_off - shpp->sh_pool_create_len -
+ (shpp->sh_eof - shpp->sh_bof) <= len) {
+ if ((err = spa_history_advance_bof(spa, shpp)) != 0) {
+ return (err);
+ }
+ }
+
+ phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
+ firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
+ shpp->sh_eof += len;
+ dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);
+
+ len -= firstwrite;
+ if (len > 0) {
+ /* write out the rest at the beginning of physical file */
+ dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
+ len, (char *)buf + firstwrite, tx);
+ }
+
+ return (0);
+}
+
+static char *
+spa_history_zone()
+{
+#ifdef _KERNEL
+ return (curproc->p_zone->zone_name);
+#else
+ return ("global");
+#endif
+}
+
+/*
+ * Write out a history event.
+ */
+static void
+spa_history_log_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+ spa_t *spa = arg1;
+ history_arg_t *hap = arg2;
+ const char *history_str = hap->ha_history_str;
+ objset_t *mos = spa->spa_meta_objset;
+ dmu_buf_t *dbp;
+ spa_history_phys_t *shpp;
+ size_t reclen;
+ uint64_t le_len;
+ nvlist_t *nvrecord;
+ char *record_packed = NULL;
+ int ret;
+
+ /*
+ * If we have an older pool that doesn't have a command
+ * history object, create it now.
+ */
+ mutex_enter(&spa->spa_history_lock);
+ if (!spa->spa_history)
+ spa_history_create_obj(spa, tx);
+ mutex_exit(&spa->spa_history_lock);
+
+ /*
+ * Get the offset of where we need to write via the bonus buffer.
+ * Update the offset when the write completes.
+ */
+ VERIFY(0 == dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
+ shpp = dbp->db_data;
+
+ dmu_buf_will_dirty(dbp, tx);
+
+#ifdef ZFS_DEBUG
+ {
+ dmu_object_info_t doi;
+ dmu_object_info_from_db(dbp, &doi);
+ ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
+ }
+#endif
+
+ VERIFY(nvlist_alloc(&nvrecord, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_TIME,
+ gethrestime_sec()) == 0);
+ VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_WHO,
+ (uint64_t)crgetuid(cr)) == 0);
+ if (hap->ha_zone[0] != '\0')
+ VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_ZONE,
+ hap->ha_zone) == 0);
+#ifdef _KERNEL
+ VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_HOST,
+ utsname.nodename) == 0);
+#endif
+ if (hap->ha_log_type == LOG_CMD_POOL_CREATE ||
+ hap->ha_log_type == LOG_CMD_NORMAL) {
+ VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_CMD,
+ history_str) == 0);
+ } else {
+ VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_INT_EVENT,
+ hap->ha_event) == 0);
+ VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_TXG,
+ tx->tx_txg) == 0);
+ VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_INT_STR,
+ history_str) == 0);
+ }
+
+ VERIFY(nvlist_size(nvrecord, &reclen, NV_ENCODE_XDR) == 0);
+ record_packed = kmem_alloc(reclen, KM_SLEEP);
+
+ VERIFY(nvlist_pack(nvrecord, &record_packed, &reclen,
+ NV_ENCODE_XDR, KM_SLEEP) == 0);
+
+ mutex_enter(&spa->spa_history_lock);
+ if (hap->ha_log_type == LOG_CMD_POOL_CREATE)
+ VERIFY(shpp->sh_eof == shpp->sh_pool_create_len);
+
+ /* write out the packed length as little endian */
+ le_len = LE_64((uint64_t)reclen);
+ ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
+ if (!ret)
+ ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
+
+ if (!ret && hap->ha_log_type == LOG_CMD_POOL_CREATE) {
+ shpp->sh_pool_create_len += sizeof (le_len) + reclen;
+ shpp->sh_bof = shpp->sh_pool_create_len;
+ }
+
+ mutex_exit(&spa->spa_history_lock);
+ nvlist_free(nvrecord);
+ kmem_free(record_packed, reclen);
+ dmu_buf_rele(dbp, FTAG);
+
+ if (hap->ha_log_type == LOG_INTERNAL) {
+ kmem_free((void*)hap->ha_history_str, HIS_MAX_RECORD_LEN);
+ kmem_free(hap, sizeof (history_arg_t));
+ }
+}
+
+/*
+ * Write out a history event.
+ */
+int
+spa_history_log(spa_t *spa, const char *history_str, history_log_type_t what)
+{
+ history_arg_t ha;
+
+ ASSERT(what != LOG_INTERNAL);
+
+ ha.ha_history_str = history_str;
+ ha.ha_log_type = what;
+ (void) strlcpy(ha.ha_zone, spa_history_zone(), sizeof (ha.ha_zone));
+ return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_history_log_sync,
+ spa, &ha, 0));
+}
+
+/*
+ * Read out the command history.
+ */
+int
+spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
+{
+ objset_t *mos = spa->spa_meta_objset;
+ dmu_buf_t *dbp;
+ uint64_t read_len, phys_read_off, phys_eof;
+ uint64_t leftover = 0;
+ spa_history_phys_t *shpp;
+ int err;
+
+ /*
+ * If the command history doesn't exist (older pool),
+ * that's ok, just return ENOENT.
+ */
+ if (!spa->spa_history)
+ return (ENOENT);
+
+ if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0)
+ return (err);
+ shpp = dbp->db_data;
+
+#ifdef ZFS_DEBUG
+ {
+ dmu_object_info_t doi;
+ dmu_object_info_from_db(dbp, &doi);
+ ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
+ }
+#endif
+
+ mutex_enter(&spa->spa_history_lock);
+ phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
+
+ if (*offp < shpp->sh_pool_create_len) {
+ /* read in just the zpool create history */
+ phys_read_off = *offp;
+ read_len = MIN(*len, shpp->sh_pool_create_len -
+ phys_read_off);
+ } else {
+ /*
+ * Need to reset passed in offset to BOF if the passed in
+ * offset has since been overwritten.
+ */
+ *offp = MAX(*offp, shpp->sh_bof);
+ phys_read_off = spa_history_log_to_phys(*offp, shpp);
+
+ /*
+ * Read up to the minimum of what the user passed down or
+ * the EOF (physical or logical). If we hit physical EOF,
+ * use 'leftover' to read from the physical BOF.
+ */
+ if (phys_read_off <= phys_eof) {
+ read_len = MIN(*len, phys_eof - phys_read_off);
+ } else {
+ read_len = MIN(*len,
+ shpp->sh_phys_max_off - phys_read_off);
+ if (phys_read_off + *len > shpp->sh_phys_max_off) {
+ leftover = MIN(*len - read_len,
+ phys_eof - shpp->sh_pool_create_len);
+ }
+ }
+ }
+
+ /* offset for consumer to use next */
+ *offp += read_len + leftover;
+
+ /* tell the consumer how much you actually read */
+ *len = read_len + leftover;
+
+ if (read_len == 0) {
+ mutex_exit(&spa->spa_history_lock);
+ dmu_buf_rele(dbp, FTAG);
+ return (0);
+ }
+
+ err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf);
+ if (leftover && err == 0) {
+ err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
+ leftover, buf + read_len);
+ }
+ mutex_exit(&spa->spa_history_lock);
+
+ dmu_buf_rele(dbp, FTAG);
+ return (err);
+}
+
+void
+spa_history_internal_log(history_internal_events_t event, spa_t *spa,
+ dmu_tx_t *tx, cred_t *cr, const char *fmt, ...)
+{
+ history_arg_t *hap;
+ char *str;
+ va_list adx;
+
+ /*
+ * If this is part of creating a pool, not everything is
+ * initialized yet, so don't bother logging the internal events.
+ */
+ if (tx->tx_txg == TXG_INITIAL)
+ return;
+
+ hap = kmem_alloc(sizeof (history_arg_t), KM_SLEEP);
+ str = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
+
+ va_start(adx, fmt);
+ (void) vsnprintf(str, HIS_MAX_RECORD_LEN, fmt, adx);
+ va_end(adx);
+
+ hap->ha_log_type = LOG_INTERNAL;
+ hap->ha_history_str = str;
+ hap->ha_event = event;
+ hap->ha_zone[0] = '\0';
+
+ if (dmu_tx_is_syncing(tx)) {
+ spa_history_log_sync(spa, hap, cr, tx);
+ } else {
+ dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
+ spa_history_log_sync, spa, hap, 0, tx);
+ }
+ /* spa_history_log_sync() will free hap and str */
+}
diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
new file mode 100644
index 000000000..36046e6df
--- /dev/null
+++ b/module/zfs/spa_misc.c
@@ -0,0 +1,1410 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa_impl.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+#include <sys/zio_compress.h>
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/zap.h>
+#include <sys/zil.h>
+#include <sys/vdev_impl.h>
+#include <sys/metaslab.h>
+#include <sys/uberblock_impl.h>
+#include <sys/txg.h>
+#include <sys/avl.h>
+#include <sys/unique.h>
+#include <sys/dsl_pool.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_prop.h>
+#include <sys/fs/zfs.h>
+#include <sys/metaslab_impl.h>
+#include <sys/sunddi.h>
+#include <sys/arc.h>
+#include "zfs_prop.h"
+
+/*
+ * SPA locking
+ *
+ * There are four basic locks for managing spa_t structures:
+ *
+ * spa_namespace_lock (global mutex)
+ *
+ * This lock must be acquired to do any of the following:
+ *
+ * - Lookup a spa_t by name
+ * - Add or remove a spa_t from the namespace
+ * - Increase spa_refcount from non-zero
+ * - Check if spa_refcount is zero
+ * - Rename a spa_t
+ * - add/remove/attach/detach devices
+ * - Held for the duration of create/destroy/import/export
+ *
+ * It does not need to handle recursion. A create or destroy may
+ * reference objects (files or zvols) in other pools, but by
+ * definition they must have an existing reference, and will never need
+ * to lookup a spa_t by name.
+ *
+ * spa_refcount (per-spa refcount_t protected by mutex)
+ *
+ * This reference count keep track of any active users of the spa_t. The
+ * spa_t cannot be destroyed or freed while this is non-zero. Internally,
+ * the refcount is never really 'zero' - opening a pool implicitly keeps
+ * some references in the DMU. Internally we check against spa_minref, but
+ * present the image of a zero/non-zero value to consumers.
+ *
+ * spa_config_lock[] (per-spa array of rwlocks)
+ *
+ * This protects the spa_t from config changes, and must be held in
+ * the following circumstances:
+ *
+ * - RW_READER to perform I/O to the spa
+ * - RW_WRITER to change the vdev config
+ *
+ * The locking order is fairly straightforward:
+ *
+ * spa_namespace_lock -> spa_refcount
+ *
+ * The namespace lock must be acquired to increase the refcount from 0
+ * or to check if it is zero.
+ *
+ * spa_refcount -> spa_config_lock[]
+ *
+ * There must be at least one valid reference on the spa_t to acquire
+ * the config lock.
+ *
+ * spa_namespace_lock -> spa_config_lock[]
+ *
+ * The namespace lock must always be taken before the config lock.
+ *
+ *
+ * The spa_namespace_lock can be acquired directly and is globally visible.
+ *
+ * The namespace is manipulated using the following functions, all of which
+ * require the spa_namespace_lock to be held.
+ *
+ * spa_lookup() Lookup a spa_t by name.
+ *
+ * spa_add() Create a new spa_t in the namespace.
+ *
+ * spa_remove() Remove a spa_t from the namespace. This also
+ * frees up any memory associated with the spa_t.
+ *
+ * spa_next() Returns the next spa_t in the system, or the
+ * first if NULL is passed.
+ *
+ * spa_evict_all() Shutdown and remove all spa_t structures in
+ * the system.
+ *
+ * spa_guid_exists() Determine whether a pool/device guid exists.
+ *
+ * The spa_refcount is manipulated using the following functions:
+ *
+ * spa_open_ref() Adds a reference to the given spa_t. Must be
+ * called with spa_namespace_lock held if the
+ * refcount is currently zero.
+ *
+ * spa_close() Remove a reference from the spa_t. This will
+ * not free the spa_t or remove it from the
+ * namespace. No locking is required.
+ *
+ * spa_refcount_zero() Returns true if the refcount is currently
+ * zero. Must be called with spa_namespace_lock
+ * held.
+ *
+ * The spa_config_lock[] is an array of rwlocks, ordered as follows:
+ * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
+ * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
+ *
+ * To read the configuration, it suffices to hold one of these locks as reader.
+ * To modify the configuration, you must hold all locks as writer. To modify
+ * vdev state without altering the vdev tree's topology (e.g. online/offline),
+ * you must hold SCL_STATE and SCL_ZIO as writer.
+ *
+ * We use these distinct config locks to avoid recursive lock entry.
+ * For example, spa_sync() (which holds SCL_CONFIG as reader) induces
+ * block allocations (SCL_ALLOC), which may require reading space maps
+ * from disk (dmu_read() -> zio_read() -> SCL_ZIO).
+ *
+ * The spa config locks cannot be normal rwlocks because we need the
+ * ability to hand off ownership. For example, SCL_ZIO is acquired
+ * by the issuing thread and later released by an interrupt thread.
+ * They do, however, obey the usual write-wanted semantics to prevent
+ * writer (i.e. system administrator) starvation.
+ *
+ * The lock acquisition rules are as follows:
+ *
+ * SCL_CONFIG
+ * Protects changes to the vdev tree topology, such as vdev
+ * add/remove/attach/detach. Protects the dirty config list
+ * (spa_config_dirty_list) and the set of spares and l2arc devices.
+ *
+ * SCL_STATE
+ * Protects changes to pool state and vdev state, such as vdev
+ * online/offline/fault/degrade/clear. Protects the dirty state list
+ * (spa_state_dirty_list) and global pool state (spa_state).
+ *
+ * SCL_ALLOC
+ * Protects changes to metaslab groups and classes.
+ * Held as reader by metaslab_alloc() and metaslab_claim().
+ *
+ * SCL_ZIO
+ * Held by bp-level zios (those which have no io_vd upon entry)
+ * to prevent changes to the vdev tree. The bp-level zio implicitly
+ * protects all of its vdev child zios, which do not hold SCL_ZIO.
+ *
+ * SCL_FREE
+ * Protects changes to metaslab groups and classes.
+ * Held as reader by metaslab_free(). SCL_FREE is distinct from
+ * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
+ * blocks in zio_done() while another i/o that holds either
+ * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
+ *
+ * SCL_VDEV
+ * Held as reader to prevent changes to the vdev tree during trivial
+ * inquiries such as bp_get_dasize(). SCL_VDEV is distinct from the
+ * other locks, and lower than all of them, to ensure that it's safe
+ * to acquire regardless of caller context.
+ *
+ * In addition, the following rules apply:
+ *
+ * (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
+ * The lock ordering is SCL_CONFIG > spa_props_lock.
+ *
+ * (b) I/O operations on leaf vdevs. For any zio operation that takes
+ * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
+ * or zio_write_phys() -- the caller must ensure that the config cannot
+ * cannot change in the interim, and that the vdev cannot be reopened.
+ * SCL_STATE as reader suffices for both.
+ *
+ * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
+ *
+ * spa_vdev_enter() Acquire the namespace lock and the config lock
+ * for writing.
+ *
+ * spa_vdev_exit() Release the config lock, wait for all I/O
+ * to complete, sync the updated configs to the
+ * cache, and release the namespace lock.
+ *
+ * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
+ * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
+ * locking is, always, based on spa_namespace_lock and spa_config_lock[].
+ *
+ * spa_rename() is also implemented within this file since is requires
+ * manipulation of the namespace.
+ */
+
+static avl_tree_t spa_namespace_avl;
+kmutex_t spa_namespace_lock;
+static kcondvar_t spa_namespace_cv;
+static int spa_active_count;
+int spa_max_replication_override = SPA_DVAS_PER_BP;
+
+static kmutex_t spa_spare_lock;
+static avl_tree_t spa_spare_avl;
+static kmutex_t spa_l2cache_lock;
+static avl_tree_t spa_l2cache_avl;
+
+kmem_cache_t *spa_buffer_pool;
+int spa_mode;
+
+#ifdef ZFS_DEBUG
+/* Everything except dprintf is on by default in debug builds */
+int zfs_flags = ~ZFS_DEBUG_DPRINTF;
+#else
+int zfs_flags = 0;
+#endif
+
+/*
+ * zfs_recover can be set to nonzero to attempt to recover from
+ * otherwise-fatal errors, typically caused by on-disk corruption. When
+ * set, calls to zfs_panic_recover() will turn into warning messages.
+ */
+int zfs_recover = 0;
+
+
+/*
+ * ==========================================================================
+ * SPA config locking
+ * ==========================================================================
+ */
+static void
+spa_config_lock_init(spa_t *spa)
+{
+ for (int i = 0; i < SCL_LOCKS; i++) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
+ refcount_create(&scl->scl_count);
+ scl->scl_writer = NULL;
+ scl->scl_write_wanted = 0;
+ }
+}
+
+static void
+spa_config_lock_destroy(spa_t *spa)
+{
+ for (int i = 0; i < SCL_LOCKS; i++) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ mutex_destroy(&scl->scl_lock);
+ cv_destroy(&scl->scl_cv);
+ refcount_destroy(&scl->scl_count);
+ ASSERT(scl->scl_writer == NULL);
+ ASSERT(scl->scl_write_wanted == 0);
+ }
+}
+
+int
+spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw)
+{
+ for (int i = 0; i < SCL_LOCKS; i++) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ if (!(locks & (1 << i)))
+ continue;
+ mutex_enter(&scl->scl_lock);
+ if (rw == RW_READER) {
+ if (scl->scl_writer || scl->scl_write_wanted) {
+ mutex_exit(&scl->scl_lock);
+ spa_config_exit(spa, locks ^ (1 << i), tag);
+ return (0);
+ }
+ } else {
+ ASSERT(scl->scl_writer != curthread);
+ if (!refcount_is_zero(&scl->scl_count)) {
+ mutex_exit(&scl->scl_lock);
+ spa_config_exit(spa, locks ^ (1 << i), tag);
+ return (0);
+ }
+ scl->scl_writer = curthread;
+ }
+ (void) refcount_add(&scl->scl_count, tag);
+ mutex_exit(&scl->scl_lock);
+ }
+ return (1);
+}
+
+void
+spa_config_enter(spa_t *spa, int locks, void *tag, krw_t rw)
+{
+ for (int i = 0; i < SCL_LOCKS; i++) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ if (!(locks & (1 << i)))
+ continue;
+ mutex_enter(&scl->scl_lock);
+ if (rw == RW_READER) {
+ while (scl->scl_writer || scl->scl_write_wanted) {
+ cv_wait(&scl->scl_cv, &scl->scl_lock);
+ }
+ } else {
+ ASSERT(scl->scl_writer != curthread);
+ while (!refcount_is_zero(&scl->scl_count)) {
+ scl->scl_write_wanted++;
+ cv_wait(&scl->scl_cv, &scl->scl_lock);
+ scl->scl_write_wanted--;
+ }
+ scl->scl_writer = curthread;
+ }
+ (void) refcount_add(&scl->scl_count, tag);
+ mutex_exit(&scl->scl_lock);
+ }
+}
+
+void
+spa_config_exit(spa_t *spa, int locks, void *tag)
+{
+ for (int i = SCL_LOCKS - 1; i >= 0; i--) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ if (!(locks & (1 << i)))
+ continue;
+ mutex_enter(&scl->scl_lock);
+ ASSERT(!refcount_is_zero(&scl->scl_count));
+ if (refcount_remove(&scl->scl_count, tag) == 0) {
+ ASSERT(scl->scl_writer == NULL ||
+ scl->scl_writer == curthread);
+ scl->scl_writer = NULL; /* OK in either case */
+ cv_broadcast(&scl->scl_cv);
+ }
+ mutex_exit(&scl->scl_lock);
+ }
+}
+
+int
+spa_config_held(spa_t *spa, int locks, krw_t rw)
+{
+ int locks_held = 0;
+
+ for (int i = 0; i < SCL_LOCKS; i++) {
+ spa_config_lock_t *scl = &spa->spa_config_lock[i];
+ if (!(locks & (1 << i)))
+ continue;
+ if ((rw == RW_READER && !refcount_is_zero(&scl->scl_count)) ||
+ (rw == RW_WRITER && scl->scl_writer == curthread))
+ locks_held |= 1 << i;
+ }
+
+ return (locks_held);
+}
+
+/*
+ * ==========================================================================
+ * SPA namespace functions
+ * ==========================================================================
+ */
+
+/*
+ * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
+ * Returns NULL if no matching spa_t is found.
+ */
+spa_t *
+spa_lookup(const char *name)
+{
+ static spa_t search; /* spa_t is large; don't allocate on stack */
+ spa_t *spa;
+ avl_index_t where;
+ char c;
+ char *cp;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ /*
+ * If it's a full dataset name, figure out the pool name and
+ * just use that.
+ */
+ cp = strpbrk(name, "/@");
+ if (cp) {
+ c = *cp;
+ *cp = '\0';
+ }
+
+ (void) strlcpy(search.spa_name, name, sizeof (search.spa_name));
+ spa = avl_find(&spa_namespace_avl, &search, &where);
+
+ if (cp)
+ *cp = c;
+
+ return (spa);
+}
+
+/*
+ * Create an uninitialized spa_t with the given name. Requires
+ * spa_namespace_lock. The caller must ensure that the spa_t doesn't already
+ * exist by calling spa_lookup() first.
+ */
+spa_t *
+spa_add(const char *name, const char *altroot)
+{
+ spa_t *spa;
+ spa_config_dirent_t *dp;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);
+
+ mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_async_root_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_sync_bplist.bpl_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&spa->spa_async_root_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
+
+ (void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
+ spa->spa_state = POOL_STATE_UNINITIALIZED;
+ spa->spa_freeze_txg = UINT64_MAX;
+ spa->spa_final_txg = UINT64_MAX;
+
+ refcount_create(&spa->spa_refcount);
+ spa_config_lock_init(spa);
+
+ avl_add(&spa_namespace_avl, spa);
+
+ mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ /*
+ * Set the alternate root, if there is one.
+ */
+ if (altroot) {
+ spa->spa_root = spa_strdup(altroot);
+ spa_active_count++;
+ }
+
+ /*
+ * Every pool starts with the default cachefile
+ */
+ list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
+ offsetof(spa_config_dirent_t, scd_link));
+
+ dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
+ dp->scd_path = spa_strdup(spa_config_path);
+ list_insert_head(&spa->spa_config_list, dp);
+
+ return (spa);
+}
+
+/*
+ * Removes a spa_t from the namespace, freeing up any memory used. Requires
+ * spa_namespace_lock. This is called only after the spa_t has been closed and
+ * deactivated.
+ */
+void
+spa_remove(spa_t *spa)
+{
+ spa_config_dirent_t *dp;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+ ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
+
+ avl_remove(&spa_namespace_avl, spa);
+ cv_broadcast(&spa_namespace_cv);
+
+ if (spa->spa_root) {
+ spa_strfree(spa->spa_root);
+ spa_active_count--;
+ }
+
+ while ((dp = list_head(&spa->spa_config_list)) != NULL) {
+ list_remove(&spa->spa_config_list, dp);
+ if (dp->scd_path != NULL)
+ spa_strfree(dp->scd_path);
+ kmem_free(dp, sizeof (spa_config_dirent_t));
+ }
+
+ list_destroy(&spa->spa_config_list);
+
+ spa_config_set(spa, NULL);
+
+ refcount_destroy(&spa->spa_refcount);
+
+ spa_config_lock_destroy(spa);
+
+ cv_destroy(&spa->spa_async_cv);
+ cv_destroy(&spa->spa_async_root_cv);
+ cv_destroy(&spa->spa_scrub_io_cv);
+ cv_destroy(&spa->spa_suspend_cv);
+
+ mutex_destroy(&spa->spa_async_lock);
+ mutex_destroy(&spa->spa_async_root_lock);
+ mutex_destroy(&spa->spa_scrub_lock);
+ mutex_destroy(&spa->spa_errlog_lock);
+ mutex_destroy(&spa->spa_errlist_lock);
+ mutex_destroy(&spa->spa_sync_bplist.bpl_lock);
+ mutex_destroy(&spa->spa_history_lock);
+ mutex_destroy(&spa->spa_props_lock);
+ mutex_destroy(&spa->spa_suspend_lock);
+
+ kmem_free(spa, sizeof (spa_t));
+}
+
+/*
+ * Given a pool, return the next pool in the namespace, or NULL if there is
+ * none. If 'prev' is NULL, return the first pool.
+ */
+spa_t *
+spa_next(spa_t *prev)
+{
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ if (prev)
+ return (AVL_NEXT(&spa_namespace_avl, prev));
+ else
+ return (avl_first(&spa_namespace_avl));
+}
+
+/*
+ * ==========================================================================
+ * SPA refcount functions
+ * ==========================================================================
+ */
+
+/*
+ * Add a reference to the given spa_t. Must have at least one reference, or
+ * have the namespace lock held.
+ */
+void
+spa_open_ref(spa_t *spa, void *tag)
+{
+ ASSERT(refcount_count(&spa->spa_refcount) >= spa->spa_minref ||
+ MUTEX_HELD(&spa_namespace_lock));
+ (void) refcount_add(&spa->spa_refcount, tag);
+}
+
+/*
+ * Remove a reference to the given spa_t. Must have at least one reference, or
+ * have the namespace lock held.
+ */
+void
+spa_close(spa_t *spa, void *tag)
+{
+ ASSERT(refcount_count(&spa->spa_refcount) > spa->spa_minref ||
+ MUTEX_HELD(&spa_namespace_lock));
+ (void) refcount_remove(&spa->spa_refcount, tag);
+}
+
+/*
+ * Check to see if the spa refcount is zero. Must be called with
+ * spa_namespace_lock held. We really compare against spa_minref, which is the
+ * number of references acquired when opening a pool
+ */
+boolean_t
+spa_refcount_zero(spa_t *spa)
+{
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ return (refcount_count(&spa->spa_refcount) == spa->spa_minref);
+}
+
+/*
+ * ==========================================================================
+ * SPA spare and l2cache tracking
+ * ==========================================================================
+ */
+
+/*
+ * Hot spares and cache devices are tracked using the same code below,
+ * for 'auxiliary' devices.
+ */
+
+typedef struct spa_aux {
+ uint64_t aux_guid;
+ uint64_t aux_pool;
+ avl_node_t aux_avl;
+ int aux_count;
+} spa_aux_t;
+
+static int
+spa_aux_compare(const void *a, const void *b)
+{
+ const spa_aux_t *sa = a;
+ const spa_aux_t *sb = b;
+
+ if (sa->aux_guid < sb->aux_guid)
+ return (-1);
+ else if (sa->aux_guid > sb->aux_guid)
+ return (1);
+ else
+ return (0);
+}
+
+void
+spa_aux_add(vdev_t *vd, avl_tree_t *avl)
+{
+ avl_index_t where;
+ spa_aux_t search;
+ spa_aux_t *aux;
+
+ search.aux_guid = vd->vdev_guid;
+ if ((aux = avl_find(avl, &search, &where)) != NULL) {
+ aux->aux_count++;
+ } else {
+ aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
+ aux->aux_guid = vd->vdev_guid;
+ aux->aux_count = 1;
+ avl_insert(avl, aux, where);
+ }
+}
+
+void
+spa_aux_remove(vdev_t *vd, avl_tree_t *avl)
+{
+ spa_aux_t search;
+ spa_aux_t *aux;
+ avl_index_t where;
+
+ search.aux_guid = vd->vdev_guid;
+ aux = avl_find(avl, &search, &where);
+
+ ASSERT(aux != NULL);
+
+ if (--aux->aux_count == 0) {
+ avl_remove(avl, aux);
+ kmem_free(aux, sizeof (spa_aux_t));
+ } else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
+ aux->aux_pool = 0ULL;
+ }
+}
+
+boolean_t
+spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl)
+{
+ spa_aux_t search, *found;
+
+ search.aux_guid = guid;
+ found = avl_find(avl, &search, NULL);
+
+ if (pool) {
+ if (found)
+ *pool = found->aux_pool;
+ else
+ *pool = 0ULL;
+ }
+
+ if (refcnt) {
+ if (found)
+ *refcnt = found->aux_count;
+ else
+ *refcnt = 0;
+ }
+
+ return (found != NULL);
+}
+
+void
+spa_aux_activate(vdev_t *vd, avl_tree_t *avl)
+{
+ spa_aux_t search, *found;
+ avl_index_t where;
+
+ search.aux_guid = vd->vdev_guid;
+ found = avl_find(avl, &search, &where);
+ ASSERT(found != NULL);
+ ASSERT(found->aux_pool == 0ULL);
+
+ found->aux_pool = spa_guid(vd->vdev_spa);
+}
+
+/*
+ * Spares are tracked globally due to the following constraints:
+ *
+ * - A spare may be part of multiple pools.
+ * - A spare may be added to a pool even if it's actively in use within
+ * another pool.
+ * - A spare in use in any pool can only be the source of a replacement if
+ * the target is a spare in the same pool.
+ *
+ * We keep track of all spares on the system through the use of a reference
+ * counted AVL tree. When a vdev is added as a spare, or used as a replacement
+ * spare, then we bump the reference count in the AVL tree. In addition, we set
+ * the 'vdev_isspare' member to indicate that the device is a spare (active or
+ * inactive). When a spare is made active (used to replace a device in the
+ * pool), we also keep track of which pool its been made a part of.
+ *
+ * The 'spa_spare_lock' protects the AVL tree. These functions are normally
+ * called under the spa_namespace lock as part of vdev reconfiguration. The
+ * separate spare lock exists for the status query path, which does not need to
+ * be completely consistent with respect to other vdev configuration changes.
+ */
+
+static int
+spa_spare_compare(const void *a, const void *b)
+{
+ return (spa_aux_compare(a, b));
+}
+
+void
+spa_spare_add(vdev_t *vd)
+{
+ mutex_enter(&spa_spare_lock);
+ ASSERT(!vd->vdev_isspare);
+ spa_aux_add(vd, &spa_spare_avl);
+ vd->vdev_isspare = B_TRUE;
+ mutex_exit(&spa_spare_lock);
+}
+
+void
+spa_spare_remove(vdev_t *vd)
+{
+ mutex_enter(&spa_spare_lock);
+ ASSERT(vd->vdev_isspare);
+ spa_aux_remove(vd, &spa_spare_avl);
+ vd->vdev_isspare = B_FALSE;
+ mutex_exit(&spa_spare_lock);
+}
+
+boolean_t
+spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt)
+{
+ boolean_t found;
+
+ mutex_enter(&spa_spare_lock);
+ found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
+ mutex_exit(&spa_spare_lock);
+
+ return (found);
+}
+
+void
+spa_spare_activate(vdev_t *vd)
+{
+ mutex_enter(&spa_spare_lock);
+ ASSERT(vd->vdev_isspare);
+ spa_aux_activate(vd, &spa_spare_avl);
+ mutex_exit(&spa_spare_lock);
+}
+
+/*
+ * Level 2 ARC devices are tracked globally for the same reasons as spares.
+ * Cache devices currently only support one pool per cache device, and so
+ * for these devices the aux reference count is currently unused beyond 1.
+ */
+
+static int
+spa_l2cache_compare(const void *a, const void *b)
+{
+ return (spa_aux_compare(a, b));
+}
+
+void
+spa_l2cache_add(vdev_t *vd)
+{
+ mutex_enter(&spa_l2cache_lock);
+ ASSERT(!vd->vdev_isl2cache);
+ spa_aux_add(vd, &spa_l2cache_avl);
+ vd->vdev_isl2cache = B_TRUE;
+ mutex_exit(&spa_l2cache_lock);
+}
+
+void
+spa_l2cache_remove(vdev_t *vd)
+{
+ mutex_enter(&spa_l2cache_lock);
+ ASSERT(vd->vdev_isl2cache);
+ spa_aux_remove(vd, &spa_l2cache_avl);
+ vd->vdev_isl2cache = B_FALSE;
+ mutex_exit(&spa_l2cache_lock);
+}
+
+boolean_t
+spa_l2cache_exists(uint64_t guid, uint64_t *pool)
+{
+ boolean_t found;
+
+ mutex_enter(&spa_l2cache_lock);
+ found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
+ mutex_exit(&spa_l2cache_lock);
+
+ return (found);
+}
+
+void
+spa_l2cache_activate(vdev_t *vd)
+{
+ mutex_enter(&spa_l2cache_lock);
+ ASSERT(vd->vdev_isl2cache);
+ spa_aux_activate(vd, &spa_l2cache_avl);
+ mutex_exit(&spa_l2cache_lock);
+}
+
+void
+spa_l2cache_space_update(vdev_t *vd, int64_t space, int64_t alloc)
+{
+ vdev_space_update(vd, space, alloc, B_FALSE);
+}
+
+/*
+ * ==========================================================================
+ * SPA vdev locking
+ * ==========================================================================
+ */
+
+/*
+ * Lock the given spa_t for the purpose of adding or removing a vdev.
+ * Grabs the global spa_namespace_lock plus the spa config lock for writing.
+ * It returns the next transaction group for the spa_t.
+ */
+uint64_t
+spa_vdev_enter(spa_t *spa)
+{
+ mutex_enter(&spa_namespace_lock);
+
+ spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
+
+ return (spa_last_synced_txg(spa) + 1);
+}
+
+/*
+ * Unlock the spa_t after adding or removing a vdev. Besides undoing the
+ * locking of spa_vdev_enter(), we also want make sure the transactions have
+ * synced to disk, and then update the global configuration cache with the new
+ * information.
+ */
+int
+spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error)
+{
+ int config_changed = B_FALSE;
+
+ ASSERT(txg > spa_last_synced_txg(spa));
+
+ spa->spa_pending_vdev = NULL;
+
+ /*
+ * Reassess the DTLs.
+ */
+ vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE);
+
+ /*
+ * If the config changed, notify the scrub thread that it must restart.
+ */
+ if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
+ dsl_pool_scrub_restart(spa->spa_dsl_pool);
+ config_changed = B_TRUE;
+ }
+
+ spa_config_exit(spa, SCL_ALL, spa);
+
+ /*
+ * Note: this txg_wait_synced() is important because it ensures
+ * that there won't be more than one config change per txg.
+ * This allows us to use the txg as the generation number.
+ */
+ if (error == 0)
+ txg_wait_synced(spa->spa_dsl_pool, txg);
+
+ if (vd != NULL) {
+ ASSERT(!vd->vdev_detached || vd->vdev_dtl.smo_object == 0);
+ vdev_free(vd);
+ }
+
+ /*
+ * If the config changed, update the config cache.
+ */
+ if (config_changed)
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+
+ mutex_exit(&spa_namespace_lock);
+
+ return (error);
+}
+
+/*
+ * Lock the given spa_t for the purpose of changing vdev state.
+ */
+void
+spa_vdev_state_enter(spa_t *spa)
+{
+ spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER);
+}
+
+int
+spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error)
+{
+ if (vd != NULL)
+ vdev_state_dirty(vd->vdev_top);
+
+ spa_config_exit(spa, SCL_STATE_ALL, spa);
+
+ return (error);
+}
+
+/*
+ * ==========================================================================
+ * Miscellaneous functions
+ * ==========================================================================
+ */
+
+/*
+ * Rename a spa_t.
+ */
+int
+spa_rename(const char *name, const char *newname)
+{
+ spa_t *spa;
+ int err;
+
+ /*
+ * Lookup the spa_t and grab the config lock for writing. We need to
+ * actually open the pool so that we can sync out the necessary labels.
+ * It's OK to call spa_open() with the namespace lock held because we
+ * allow recursive calls for other reasons.
+ */
+ mutex_enter(&spa_namespace_lock);
+ if ((err = spa_open(name, &spa, FTAG)) != 0) {
+ mutex_exit(&spa_namespace_lock);
+ return (err);
+ }
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+ avl_remove(&spa_namespace_avl, spa);
+ (void) strlcpy(spa->spa_name, newname, sizeof (spa->spa_name));
+ avl_add(&spa_namespace_avl, spa);
+
+ /*
+ * Sync all labels to disk with the new names by marking the root vdev
+ * dirty and waiting for it to sync. It will pick up the new pool name
+ * during the sync.
+ */
+ vdev_config_dirty(spa->spa_root_vdev);
+
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ txg_wait_synced(spa->spa_dsl_pool, 0);
+
+ /*
+ * Sync the updated config cache.
+ */
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+
+ spa_close(spa, FTAG);
+
+ mutex_exit(&spa_namespace_lock);
+
+ return (0);
+}
+
+
+/*
+ * Determine whether a pool with given pool_guid exists. If device_guid is
+ * non-zero, determine whether the pool exists *and* contains a device with the
+ * specified device_guid.
+ */
+boolean_t
+spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
+{
+ spa_t *spa;
+ avl_tree_t *t = &spa_namespace_avl;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+
+ for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
+ if (spa->spa_state == POOL_STATE_UNINITIALIZED)
+ continue;
+ if (spa->spa_root_vdev == NULL)
+ continue;
+ if (spa_guid(spa) == pool_guid) {
+ if (device_guid == 0)
+ break;
+
+ if (vdev_lookup_by_guid(spa->spa_root_vdev,
+ device_guid) != NULL)
+ break;
+
+ /*
+ * Check any devices we may be in the process of adding.
+ */
+ if (spa->spa_pending_vdev) {
+ if (vdev_lookup_by_guid(spa->spa_pending_vdev,
+ device_guid) != NULL)
+ break;
+ }
+ }
+ }
+
+ return (spa != NULL);
+}
+
+char *
+spa_strdup(const char *s)
+{
+ size_t len;
+ char *new;
+
+ len = strlen(s);
+ new = kmem_alloc(len + 1, KM_SLEEP);
+ bcopy(s, new, len);
+ new[len] = '\0';
+
+ return (new);
+}
+
+void
+spa_strfree(char *s)
+{
+ kmem_free(s, strlen(s) + 1);
+}
+
+uint64_t
+spa_get_random(uint64_t range)
+{
+ uint64_t r;
+
+ ASSERT(range != 0);
+
+ (void) random_get_pseudo_bytes((void *)&r, sizeof (uint64_t));
+
+ return (r % range);
+}
+
+void
+sprintf_blkptr(char *buf, int len, const blkptr_t *bp)
+{
+ int d;
+
+ if (bp == NULL) {
+ (void) snprintf(buf, len, "<NULL>");
+ return;
+ }
+
+ if (BP_IS_HOLE(bp)) {
+ (void) snprintf(buf, len, "<hole>");
+ return;
+ }
+
+ (void) snprintf(buf, len, "[L%llu %s] %llxL/%llxP ",
+ (u_longlong_t)BP_GET_LEVEL(bp),
+ dmu_ot[BP_GET_TYPE(bp)].ot_name,
+ (u_longlong_t)BP_GET_LSIZE(bp),
+ (u_longlong_t)BP_GET_PSIZE(bp));
+
+ for (d = 0; d < BP_GET_NDVAS(bp); d++) {
+ const dva_t *dva = &bp->blk_dva[d];
+ (void) snprintf(buf + strlen(buf), len - strlen(buf),
+ "DVA[%d]=<%llu:%llx:%llx> ", d,
+ (u_longlong_t)DVA_GET_VDEV(dva),
+ (u_longlong_t)DVA_GET_OFFSET(dva),
+ (u_longlong_t)DVA_GET_ASIZE(dva));
+ }
+
+ (void) snprintf(buf + strlen(buf), len - strlen(buf),
+ "%s %s %s %s birth=%llu fill=%llu cksum=%llx:%llx:%llx:%llx",
+ zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name,
+ zio_compress_table[BP_GET_COMPRESS(bp)].ci_name,
+ BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE",
+ BP_IS_GANG(bp) ? "gang" : "contiguous",
+ (u_longlong_t)bp->blk_birth,
+ (u_longlong_t)bp->blk_fill,
+ (u_longlong_t)bp->blk_cksum.zc_word[0],
+ (u_longlong_t)bp->blk_cksum.zc_word[1],
+ (u_longlong_t)bp->blk_cksum.zc_word[2],
+ (u_longlong_t)bp->blk_cksum.zc_word[3]);
+}
+
+void
+spa_freeze(spa_t *spa)
+{
+ uint64_t freeze_txg = 0;
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ if (spa->spa_freeze_txg == UINT64_MAX) {
+ freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
+ spa->spa_freeze_txg = freeze_txg;
+ }
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ if (freeze_txg != 0)
+ txg_wait_synced(spa_get_dsl(spa), freeze_txg);
+}
+
+void
+zfs_panic_recover(const char *fmt, ...)
+{
+ va_list adx;
+
+ va_start(adx, fmt);
+ vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
+ va_end(adx);
+}
+
+/*
+ * ==========================================================================
+ * Accessor functions
+ * ==========================================================================
+ */
+
+boolean_t
+spa_shutting_down(spa_t *spa)
+{
+ return (spa->spa_async_suspended);
+}
+
+dsl_pool_t *
+spa_get_dsl(spa_t *spa)
+{
+ return (spa->spa_dsl_pool);
+}
+
+blkptr_t *
+spa_get_rootblkptr(spa_t *spa)
+{
+ return (&spa->spa_ubsync.ub_rootbp);
+}
+
+void
+spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp)
+{
+ spa->spa_uberblock.ub_rootbp = *bp;
+}
+
+void
+spa_altroot(spa_t *spa, char *buf, size_t buflen)
+{
+ if (spa->spa_root == NULL)
+ buf[0] = '\0';
+ else
+ (void) strncpy(buf, spa->spa_root, buflen);
+}
+
+int
+spa_sync_pass(spa_t *spa)
+{
+ return (spa->spa_sync_pass);
+}
+
+char *
+spa_name(spa_t *spa)
+{
+ return (spa->spa_name);
+}
+
+uint64_t
+spa_guid(spa_t *spa)
+{
+ /*
+ * If we fail to parse the config during spa_load(), we can go through
+ * the error path (which posts an ereport) and end up here with no root
+ * vdev. We stash the original pool guid in 'spa_load_guid' to handle
+ * this case.
+ */
+ if (spa->spa_root_vdev != NULL)
+ return (spa->spa_root_vdev->vdev_guid);
+ else
+ return (spa->spa_load_guid);
+}
+
+uint64_t
+spa_last_synced_txg(spa_t *spa)
+{
+ return (spa->spa_ubsync.ub_txg);
+}
+
+uint64_t
+spa_first_txg(spa_t *spa)
+{
+ return (spa->spa_first_txg);
+}
+
+pool_state_t
+spa_state(spa_t *spa)
+{
+ return (spa->spa_state);
+}
+
+uint64_t
+spa_freeze_txg(spa_t *spa)
+{
+ return (spa->spa_freeze_txg);
+}
+
+/*
+ * Return how much space is allocated in the pool (ie. sum of all asize)
+ */
+uint64_t
+spa_get_alloc(spa_t *spa)
+{
+ return (spa->spa_root_vdev->vdev_stat.vs_alloc);
+}
+
+/*
+ * Return how much (raid-z inflated) space there is in the pool.
+ */
+uint64_t
+spa_get_space(spa_t *spa)
+{
+ return (spa->spa_root_vdev->vdev_stat.vs_space);
+}
+
+/*
+ * Return the amount of raid-z-deflated space in the pool.
+ */
+uint64_t
+spa_get_dspace(spa_t *spa)
+{
+ if (spa->spa_deflate)
+ return (spa->spa_root_vdev->vdev_stat.vs_dspace);
+ else
+ return (spa->spa_root_vdev->vdev_stat.vs_space);
+}
+
+/* ARGSUSED */
+uint64_t
+spa_get_asize(spa_t *spa, uint64_t lsize)
+{
+ /*
+ * For now, the worst case is 512-byte RAID-Z blocks, in which
+ * case the space requirement is exactly 2x; so just assume that.
+ * Add to this the fact that we can have up to 3 DVAs per bp, and
+ * we have to multiply by a total of 6x.
+ */
+ return (lsize * 6);
+}
+
+/*
+ * Return the failure mode that has been set to this pool. The default
+ * behavior will be to block all I/Os when a complete failure occurs.
+ */
+uint8_t
+spa_get_failmode(spa_t *spa)
+{
+ return (spa->spa_failmode);
+}
+
+boolean_t
+spa_suspended(spa_t *spa)
+{
+ return (spa->spa_suspended);
+}
+
+uint64_t
+spa_version(spa_t *spa)
+{
+ return (spa->spa_ubsync.ub_version);
+}
+
+int
+spa_max_replication(spa_t *spa)
+{
+ /*
+ * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
+ * handle BPs with more than one DVA allocated. Set our max
+ * replication level accordingly.
+ */
+ if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
+ return (1);
+ return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
+}
+
+uint64_t
+bp_get_dasize(spa_t *spa, const blkptr_t *bp)
+{
+ int sz = 0, i;
+
+ if (!spa->spa_deflate)
+ return (BP_GET_ASIZE(bp));
+
+ spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
+ for (i = 0; i < SPA_DVAS_PER_BP; i++) {
+ vdev_t *vd =
+ vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[i]));
+ if (vd)
+ sz += (DVA_GET_ASIZE(&bp->blk_dva[i]) >>
+ SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio;
+ }
+ spa_config_exit(spa, SCL_VDEV, FTAG);
+ return (sz);
+}
+
+/*
+ * ==========================================================================
+ * Initialization and Termination
+ * ==========================================================================
+ */
+
+static int
+spa_name_compare(const void *a1, const void *a2)
+{
+ const spa_t *s1 = a1;
+ const spa_t *s2 = a2;
+ int s;
+
+ s = strcmp(s1->spa_name, s2->spa_name);
+ if (s > 0)
+ return (1);
+ if (s < 0)
+ return (-1);
+ return (0);
+}
+
+int
+spa_busy(void)
+{
+ return (spa_active_count);
+}
+
+void
+spa_boot_init()
+{
+ spa_config_load();
+}
+
+void
+spa_init(int mode)
+{
+ mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);
+
+ avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
+ offsetof(spa_t, spa_avl));
+
+ avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
+ offsetof(spa_aux_t, aux_avl));
+
+ avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
+ offsetof(spa_aux_t, aux_avl));
+
+ spa_mode = mode;
+
+ refcount_init();
+ unique_init();
+ zio_init();
+ dmu_init();
+ zil_init();
+ vdev_cache_stat_init();
+ zfs_prop_init();
+ zpool_prop_init();
+ spa_config_load();
+ l2arc_start();
+}
+
+void
+spa_fini(void)
+{
+ l2arc_stop();
+
+ spa_evict_all();
+
+ vdev_cache_stat_fini();
+ zil_fini();
+ dmu_fini();
+ zio_fini();
+ unique_fini();
+ refcount_fini();
+
+ avl_destroy(&spa_namespace_avl);
+ avl_destroy(&spa_spare_avl);
+ avl_destroy(&spa_l2cache_avl);
+
+ cv_destroy(&spa_namespace_cv);
+ mutex_destroy(&spa_namespace_lock);
+ mutex_destroy(&spa_spare_lock);
+ mutex_destroy(&spa_l2cache_lock);
+}
+
+/*
+ * Return whether this pool has slogs. No locking needed.
+ * It's not a problem if the wrong answer is returned as it's only for
+ * performance and not correctness
+ */
+boolean_t
+spa_has_slogs(spa_t *spa)
+{
+ return (spa->spa_log_class->mc_rotor != NULL);
+}
+
+/*
+ * Return whether this pool is the root pool.
+ */
+boolean_t
+spa_is_root(spa_t *spa)
+{
+ return (spa->spa_is_root);
+}
diff --git a/module/zfs/space_map.c b/module/zfs/space_map.c
new file mode 100644
index 000000000..0a1fd59ea
--- /dev/null
+++ b/module/zfs/space_map.c
@@ -0,0 +1,506 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/zio.h>
+#include <sys/space_map.h>
+
+/*
+ * Space map routines.
+ * NOTE: caller is responsible for all locking.
+ */
+static int
+space_map_seg_compare(const void *x1, const void *x2)
+{
+ const space_seg_t *s1 = x1;
+ const space_seg_t *s2 = x2;
+
+ if (s1->ss_start < s2->ss_start) {
+ if (s1->ss_end > s2->ss_start)
+ return (0);
+ return (-1);
+ }
+ if (s1->ss_start > s2->ss_start) {
+ if (s1->ss_start < s2->ss_end)
+ return (0);
+ return (1);
+ }
+ return (0);
+}
+
+void
+space_map_create(space_map_t *sm, uint64_t start, uint64_t size, uint8_t shift,
+ kmutex_t *lp)
+{
+ bzero(sm, sizeof (*sm));
+
+ avl_create(&sm->sm_root, space_map_seg_compare,
+ sizeof (space_seg_t), offsetof(struct space_seg, ss_node));
+
+ sm->sm_start = start;
+ sm->sm_size = size;
+ sm->sm_shift = shift;
+ sm->sm_lock = lp;
+}
+
+void
+space_map_destroy(space_map_t *sm)
+{
+ ASSERT(!sm->sm_loaded && !sm->sm_loading);
+ VERIFY3U(sm->sm_space, ==, 0);
+ avl_destroy(&sm->sm_root);
+}
+
+void
+space_map_add(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ avl_index_t where;
+ space_seg_t ssearch, *ss_before, *ss_after, *ss;
+ uint64_t end = start + size;
+ int merge_before, merge_after;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+ VERIFY(size != 0);
+ VERIFY3U(start, >=, sm->sm_start);
+ VERIFY3U(end, <=, sm->sm_start + sm->sm_size);
+ VERIFY(sm->sm_space + size <= sm->sm_size);
+ VERIFY(P2PHASE(start, 1ULL << sm->sm_shift) == 0);
+ VERIFY(P2PHASE(size, 1ULL << sm->sm_shift) == 0);
+
+ ssearch.ss_start = start;
+ ssearch.ss_end = end;
+ ss = avl_find(&sm->sm_root, &ssearch, &where);
+
+ if (ss != NULL && ss->ss_start <= start && ss->ss_end >= end) {
+ zfs_panic_recover("zfs: allocating allocated segment"
+ "(offset=%llu size=%llu)\n",
+ (longlong_t)start, (longlong_t)size);
+ return;
+ }
+
+ /* Make sure we don't overlap with either of our neighbors */
+ VERIFY(ss == NULL);
+
+ ss_before = avl_nearest(&sm->sm_root, where, AVL_BEFORE);
+ ss_after = avl_nearest(&sm->sm_root, where, AVL_AFTER);
+
+ merge_before = (ss_before != NULL && ss_before->ss_end == start);
+ merge_after = (ss_after != NULL && ss_after->ss_start == end);
+
+ if (merge_before && merge_after) {
+ avl_remove(&sm->sm_root, ss_before);
+ ss_after->ss_start = ss_before->ss_start;
+ kmem_free(ss_before, sizeof (*ss_before));
+ } else if (merge_before) {
+ ss_before->ss_end = end;
+ } else if (merge_after) {
+ ss_after->ss_start = start;
+ } else {
+ ss = kmem_alloc(sizeof (*ss), KM_SLEEP);
+ ss->ss_start = start;
+ ss->ss_end = end;
+ avl_insert(&sm->sm_root, ss, where);
+ }
+
+ sm->sm_space += size;
+}
+
+void
+space_map_remove(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ avl_index_t where;
+ space_seg_t ssearch, *ss, *newseg;
+ uint64_t end = start + size;
+ int left_over, right_over;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+ VERIFY(size != 0);
+ VERIFY(P2PHASE(start, 1ULL << sm->sm_shift) == 0);
+ VERIFY(P2PHASE(size, 1ULL << sm->sm_shift) == 0);
+
+ ssearch.ss_start = start;
+ ssearch.ss_end = end;
+ ss = avl_find(&sm->sm_root, &ssearch, &where);
+
+ /* Make sure we completely overlap with someone */
+ if (ss == NULL) {
+ zfs_panic_recover("zfs: freeing free segment "
+ "(offset=%llu size=%llu)",
+ (longlong_t)start, (longlong_t)size);
+ return;
+ }
+ VERIFY3U(ss->ss_start, <=, start);
+ VERIFY3U(ss->ss_end, >=, end);
+ VERIFY(sm->sm_space - size <= sm->sm_size);
+
+ left_over = (ss->ss_start != start);
+ right_over = (ss->ss_end != end);
+
+ if (left_over && right_over) {
+ newseg = kmem_alloc(sizeof (*newseg), KM_SLEEP);
+ newseg->ss_start = end;
+ newseg->ss_end = ss->ss_end;
+ ss->ss_end = start;
+ avl_insert_here(&sm->sm_root, newseg, ss, AVL_AFTER);
+ } else if (left_over) {
+ ss->ss_end = start;
+ } else if (right_over) {
+ ss->ss_start = end;
+ } else {
+ avl_remove(&sm->sm_root, ss);
+ kmem_free(ss, sizeof (*ss));
+ }
+
+ sm->sm_space -= size;
+}
+
+int
+space_map_contains(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ avl_index_t where;
+ space_seg_t ssearch, *ss;
+ uint64_t end = start + size;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+ VERIFY(size != 0);
+ VERIFY(P2PHASE(start, 1ULL << sm->sm_shift) == 0);
+ VERIFY(P2PHASE(size, 1ULL << sm->sm_shift) == 0);
+
+ ssearch.ss_start = start;
+ ssearch.ss_end = end;
+ ss = avl_find(&sm->sm_root, &ssearch, &where);
+
+ return (ss != NULL && ss->ss_start <= start && ss->ss_end >= end);
+}
+
+void
+space_map_vacate(space_map_t *sm, space_map_func_t *func, space_map_t *mdest)
+{
+ space_seg_t *ss;
+ void *cookie = NULL;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ while ((ss = avl_destroy_nodes(&sm->sm_root, &cookie)) != NULL) {
+ if (func != NULL)
+ func(mdest, ss->ss_start, ss->ss_end - ss->ss_start);
+ kmem_free(ss, sizeof (*ss));
+ }
+ sm->sm_space = 0;
+}
+
+void
+space_map_walk(space_map_t *sm, space_map_func_t *func, space_map_t *mdest)
+{
+ space_seg_t *ss;
+
+ for (ss = avl_first(&sm->sm_root); ss; ss = AVL_NEXT(&sm->sm_root, ss))
+ func(mdest, ss->ss_start, ss->ss_end - ss->ss_start);
+}
+
+void
+space_map_excise(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ avl_tree_t *t = &sm->sm_root;
+ avl_index_t where;
+ space_seg_t *ss, search;
+ uint64_t end = start + size;
+ uint64_t rm_start, rm_end;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ search.ss_start = start;
+ search.ss_end = start;
+
+ for (;;) {
+ ss = avl_find(t, &search, &where);
+
+ if (ss == NULL)
+ ss = avl_nearest(t, where, AVL_AFTER);
+
+ if (ss == NULL || ss->ss_start >= end)
+ break;
+
+ rm_start = MAX(ss->ss_start, start);
+ rm_end = MIN(ss->ss_end, end);
+
+ space_map_remove(sm, rm_start, rm_end - rm_start);
+ }
+}
+
+/*
+ * Replace smd with the union of smd and sms.
+ */
+void
+space_map_union(space_map_t *smd, space_map_t *sms)
+{
+ avl_tree_t *t = &sms->sm_root;
+ space_seg_t *ss;
+
+ ASSERT(MUTEX_HELD(smd->sm_lock));
+
+ /*
+ * For each source segment, remove any intersections with the
+ * destination, then add the source segment to the destination.
+ */
+ for (ss = avl_first(t); ss != NULL; ss = AVL_NEXT(t, ss)) {
+ space_map_excise(smd, ss->ss_start, ss->ss_end - ss->ss_start);
+ space_map_add(smd, ss->ss_start, ss->ss_end - ss->ss_start);
+ }
+}
+
+/*
+ * Wait for any in-progress space_map_load() to complete.
+ */
+void
+space_map_load_wait(space_map_t *sm)
+{
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ while (sm->sm_loading)
+ cv_wait(&sm->sm_load_cv, sm->sm_lock);
+}
+
+/*
+ * Note: space_map_load() will drop sm_lock across dmu_read() calls.
+ * The caller must be OK with this.
+ */
+int
+space_map_load(space_map_t *sm, space_map_ops_t *ops, uint8_t maptype,
+ space_map_obj_t *smo, objset_t *os)
+{
+ uint64_t *entry, *entry_map, *entry_map_end;
+ uint64_t bufsize, size, offset, end, space;
+ uint64_t mapstart = sm->sm_start;
+ int error = 0;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ space_map_load_wait(sm);
+
+ if (sm->sm_loaded)
+ return (0);
+
+ sm->sm_loading = B_TRUE;
+ end = smo->smo_objsize;
+ space = smo->smo_alloc;
+
+ ASSERT(sm->sm_ops == NULL);
+ VERIFY3U(sm->sm_space, ==, 0);
+
+ if (maptype == SM_FREE) {
+ space_map_add(sm, sm->sm_start, sm->sm_size);
+ space = sm->sm_size - space;
+ }
+
+ bufsize = 1ULL << SPACE_MAP_BLOCKSHIFT;
+ entry_map = zio_buf_alloc(bufsize);
+
+ mutex_exit(sm->sm_lock);
+ if (end > bufsize)
+ dmu_prefetch(os, smo->smo_object, bufsize, end - bufsize);
+ mutex_enter(sm->sm_lock);
+
+ for (offset = 0; offset < end; offset += bufsize) {
+ size = MIN(end - offset, bufsize);
+ VERIFY(P2PHASE(size, sizeof (uint64_t)) == 0);
+ VERIFY(size != 0);
+
+ dprintf("object=%llu offset=%llx size=%llx\n",
+ smo->smo_object, offset, size);
+
+ mutex_exit(sm->sm_lock);
+ error = dmu_read(os, smo->smo_object, offset, size, entry_map);
+ mutex_enter(sm->sm_lock);
+ if (error != 0)
+ break;
+
+ entry_map_end = entry_map + (size / sizeof (uint64_t));
+ for (entry = entry_map; entry < entry_map_end; entry++) {
+ uint64_t e = *entry;
+
+ if (SM_DEBUG_DECODE(e)) /* Skip debug entries */
+ continue;
+
+ (SM_TYPE_DECODE(e) == maptype ?
+ space_map_add : space_map_remove)(sm,
+ (SM_OFFSET_DECODE(e) << sm->sm_shift) + mapstart,
+ SM_RUN_DECODE(e) << sm->sm_shift);
+ }
+ }
+
+ if (error == 0) {
+ VERIFY3U(sm->sm_space, ==, space);
+
+ sm->sm_loaded = B_TRUE;
+ sm->sm_ops = ops;
+ if (ops != NULL)
+ ops->smop_load(sm);
+ } else {
+ space_map_vacate(sm, NULL, NULL);
+ }
+
+ zio_buf_free(entry_map, bufsize);
+
+ sm->sm_loading = B_FALSE;
+
+ cv_broadcast(&sm->sm_load_cv);
+
+ return (error);
+}
+
+void
+space_map_unload(space_map_t *sm)
+{
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ if (sm->sm_loaded && sm->sm_ops != NULL)
+ sm->sm_ops->smop_unload(sm);
+
+ sm->sm_loaded = B_FALSE;
+ sm->sm_ops = NULL;
+
+ space_map_vacate(sm, NULL, NULL);
+}
+
+uint64_t
+space_map_alloc(space_map_t *sm, uint64_t size)
+{
+ uint64_t start;
+
+ start = sm->sm_ops->smop_alloc(sm, size);
+ if (start != -1ULL)
+ space_map_remove(sm, start, size);
+ return (start);
+}
+
+void
+space_map_claim(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ sm->sm_ops->smop_claim(sm, start, size);
+ space_map_remove(sm, start, size);
+}
+
+void
+space_map_free(space_map_t *sm, uint64_t start, uint64_t size)
+{
+ space_map_add(sm, start, size);
+ sm->sm_ops->smop_free(sm, start, size);
+}
+
+/*
+ * Note: space_map_sync() will drop sm_lock across dmu_write() calls.
+ */
+void
+space_map_sync(space_map_t *sm, uint8_t maptype,
+ space_map_obj_t *smo, objset_t *os, dmu_tx_t *tx)
+{
+ spa_t *spa = dmu_objset_spa(os);
+ void *cookie = NULL;
+ space_seg_t *ss;
+ uint64_t bufsize, start, size, run_len;
+ uint64_t *entry, *entry_map, *entry_map_end;
+
+ ASSERT(MUTEX_HELD(sm->sm_lock));
+
+ if (sm->sm_space == 0)
+ return;
+
+ dprintf("object %4llu, txg %llu, pass %d, %c, count %lu, space %llx\n",
+ smo->smo_object, dmu_tx_get_txg(tx), spa_sync_pass(spa),
+ maptype == SM_ALLOC ? 'A' : 'F', avl_numnodes(&sm->sm_root),
+ sm->sm_space);
+
+ if (maptype == SM_ALLOC)
+ smo->smo_alloc += sm->sm_space;
+ else
+ smo->smo_alloc -= sm->sm_space;
+
+ bufsize = (8 + avl_numnodes(&sm->sm_root)) * sizeof (uint64_t);
+ bufsize = MIN(bufsize, 1ULL << SPACE_MAP_BLOCKSHIFT);
+ entry_map = zio_buf_alloc(bufsize);
+ entry_map_end = entry_map + (bufsize / sizeof (uint64_t));
+ entry = entry_map;
+
+ *entry++ = SM_DEBUG_ENCODE(1) |
+ SM_DEBUG_ACTION_ENCODE(maptype) |
+ SM_DEBUG_SYNCPASS_ENCODE(spa_sync_pass(spa)) |
+ SM_DEBUG_TXG_ENCODE(dmu_tx_get_txg(tx));
+
+ while ((ss = avl_destroy_nodes(&sm->sm_root, &cookie)) != NULL) {
+ size = ss->ss_end - ss->ss_start;
+ start = (ss->ss_start - sm->sm_start) >> sm->sm_shift;
+
+ sm->sm_space -= size;
+ size >>= sm->sm_shift;
+
+ while (size) {
+ run_len = MIN(size, SM_RUN_MAX);
+
+ if (entry == entry_map_end) {
+ mutex_exit(sm->sm_lock);
+ dmu_write(os, smo->smo_object, smo->smo_objsize,
+ bufsize, entry_map, tx);
+ mutex_enter(sm->sm_lock);
+ smo->smo_objsize += bufsize;
+ entry = entry_map;
+ }
+
+ *entry++ = SM_OFFSET_ENCODE(start) |
+ SM_TYPE_ENCODE(maptype) |
+ SM_RUN_ENCODE(run_len);
+
+ start += run_len;
+ size -= run_len;
+ }
+ kmem_free(ss, sizeof (*ss));
+ }
+
+ if (entry != entry_map) {
+ size = (entry - entry_map) * sizeof (uint64_t);
+ mutex_exit(sm->sm_lock);
+ dmu_write(os, smo->smo_object, smo->smo_objsize,
+ size, entry_map, tx);
+ mutex_enter(sm->sm_lock);
+ smo->smo_objsize += size;
+ }
+
+ zio_buf_free(entry_map, bufsize);
+
+ VERIFY3U(sm->sm_space, ==, 0);
+}
+
+void
+space_map_truncate(space_map_obj_t *smo, objset_t *os, dmu_tx_t *tx)
+{
+ VERIFY(dmu_free_range(os, smo->smo_object, 0, -1ULL, tx) == 0);
+
+ smo->smo_objsize = 0;
+ smo->smo_alloc = 0;
+}
diff --git a/module/zfs/txg.c b/module/zfs/txg.c
new file mode 100644
index 000000000..2bbf2f086
--- /dev/null
+++ b/module/zfs/txg.c
@@ -0,0 +1,627 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/txg_impl.h>
+#include <sys/dmu_impl.h>
+#include <sys/dsl_pool.h>
+#include <sys/callb.h>
+
+/*
+ * Pool-wide transaction groups.
+ */
+
+static void txg_sync_thread(dsl_pool_t *dp);
+static void txg_quiesce_thread(dsl_pool_t *dp);
+
+int zfs_txg_timeout = 30; /* max seconds worth of delta per txg */
+
+/*
+ * Prepare the txg subsystem.
+ */
+void
+txg_init(dsl_pool_t *dp, uint64_t txg)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ int c;
+ bzero(tx, sizeof (tx_state_t));
+
+ tx->tx_cpu = kmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
+
+ for (c = 0; c < max_ncpus; c++) {
+ int i;
+
+ mutex_init(&tx->tx_cpu[c].tc_lock, NULL, MUTEX_DEFAULT, NULL);
+ for (i = 0; i < TXG_SIZE; i++) {
+ cv_init(&tx->tx_cpu[c].tc_cv[i], NULL, CV_DEFAULT,
+ NULL);
+ }
+ }
+
+ rw_init(&tx->tx_suspend, NULL, RW_DEFAULT, NULL);
+ mutex_init(&tx->tx_sync_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ tx->tx_open_txg = txg;
+}
+
+/*
+ * Close down the txg subsystem.
+ */
+void
+txg_fini(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ int c;
+
+ ASSERT(tx->tx_threads == 0);
+
+ rw_destroy(&tx->tx_suspend);
+ mutex_destroy(&tx->tx_sync_lock);
+
+ for (c = 0; c < max_ncpus; c++) {
+ int i;
+
+ mutex_destroy(&tx->tx_cpu[c].tc_lock);
+ for (i = 0; i < TXG_SIZE; i++)
+ cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
+ }
+
+ kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
+
+ bzero(tx, sizeof (tx_state_t));
+}
+
+/*
+ * Start syncing transaction groups.
+ */
+void
+txg_sync_start(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ mutex_enter(&tx->tx_sync_lock);
+
+ dprintf("pool %p\n", dp);
+
+ ASSERT(tx->tx_threads == 0);
+
+ tx->tx_threads = 2;
+
+ tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
+ dp, 0, &p0, TS_RUN, minclsyspri);
+
+ /*
+ * The sync thread can need a larger-than-default stack size on
+ * 32-bit x86. This is due in part to nested pools and
+ * scrub_visitbp() recursion.
+ */
+ tx->tx_sync_thread = thread_create(NULL, 12<<10, txg_sync_thread,
+ dp, 0, &p0, TS_RUN, minclsyspri);
+
+ mutex_exit(&tx->tx_sync_lock);
+}
+
+static void
+txg_thread_enter(tx_state_t *tx, callb_cpr_t *cpr)
+{
+ CALLB_CPR_INIT(cpr, &tx->tx_sync_lock, callb_generic_cpr, FTAG);
+ mutex_enter(&tx->tx_sync_lock);
+}
+
+static void
+txg_thread_exit(tx_state_t *tx, callb_cpr_t *cpr, kthread_t **tpp)
+{
+ ASSERT(*tpp != NULL);
+ *tpp = NULL;
+ tx->tx_threads--;
+ cv_broadcast(&tx->tx_exit_cv);
+ CALLB_CPR_EXIT(cpr); /* drops &tx->tx_sync_lock */
+ thread_exit();
+}
+
+static void
+txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
+{
+ CALLB_CPR_SAFE_BEGIN(cpr);
+
+ if (time)
+ (void) cv_timedwait(cv, &tx->tx_sync_lock, lbolt + time);
+ else
+ cv_wait(cv, &tx->tx_sync_lock);
+
+ CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
+}
+
+/*
+ * Stop syncing transaction groups.
+ */
+void
+txg_sync_stop(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ dprintf("pool %p\n", dp);
+ /*
+ * Finish off any work in progress.
+ */
+ ASSERT(tx->tx_threads == 2);
+ txg_wait_synced(dp, 0);
+
+ /*
+ * Wake all sync threads and wait for them to die.
+ */
+ mutex_enter(&tx->tx_sync_lock);
+
+ ASSERT(tx->tx_threads == 2);
+
+ tx->tx_exiting = 1;
+
+ cv_broadcast(&tx->tx_quiesce_more_cv);
+ cv_broadcast(&tx->tx_quiesce_done_cv);
+ cv_broadcast(&tx->tx_sync_more_cv);
+
+ while (tx->tx_threads != 0)
+ cv_wait(&tx->tx_exit_cv, &tx->tx_sync_lock);
+
+ tx->tx_exiting = 0;
+
+ mutex_exit(&tx->tx_sync_lock);
+}
+
+uint64_t
+txg_hold_open(dsl_pool_t *dp, txg_handle_t *th)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ tx_cpu_t *tc = &tx->tx_cpu[CPU_SEQID];
+ uint64_t txg;
+
+ mutex_enter(&tc->tc_lock);
+
+ txg = tx->tx_open_txg;
+ tc->tc_count[txg & TXG_MASK]++;
+
+ th->th_cpu = tc;
+ th->th_txg = txg;
+
+ return (txg);
+}
+
+void
+txg_rele_to_quiesce(txg_handle_t *th)
+{
+ tx_cpu_t *tc = th->th_cpu;
+
+ mutex_exit(&tc->tc_lock);
+}
+
+void
+txg_rele_to_sync(txg_handle_t *th)
+{
+ tx_cpu_t *tc = th->th_cpu;
+ int g = th->th_txg & TXG_MASK;
+
+ mutex_enter(&tc->tc_lock);
+ ASSERT(tc->tc_count[g] != 0);
+ if (--tc->tc_count[g] == 0)
+ cv_broadcast(&tc->tc_cv[g]);
+ mutex_exit(&tc->tc_lock);
+
+ th->th_cpu = NULL; /* defensive */
+}
+
+static void
+txg_quiesce(dsl_pool_t *dp, uint64_t txg)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ int g = txg & TXG_MASK;
+ int c;
+
+ /*
+ * Grab all tx_cpu locks so nobody else can get into this txg.
+ */
+ for (c = 0; c < max_ncpus; c++)
+ mutex_enter(&tx->tx_cpu[c].tc_lock);
+
+ ASSERT(txg == tx->tx_open_txg);
+ tx->tx_open_txg++;
+
+ /*
+ * Now that we've incremented tx_open_txg, we can let threads
+ * enter the next transaction group.
+ */
+ for (c = 0; c < max_ncpus; c++)
+ mutex_exit(&tx->tx_cpu[c].tc_lock);
+
+ /*
+ * Quiesce the transaction group by waiting for everyone to txg_exit().
+ */
+ for (c = 0; c < max_ncpus; c++) {
+ tx_cpu_t *tc = &tx->tx_cpu[c];
+ mutex_enter(&tc->tc_lock);
+ while (tc->tc_count[g] != 0)
+ cv_wait(&tc->tc_cv[g], &tc->tc_lock);
+ mutex_exit(&tc->tc_lock);
+ }
+}
+
+static void
+txg_sync_thread(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ callb_cpr_t cpr;
+ uint64_t start, delta;
+
+ txg_thread_enter(tx, &cpr);
+
+ start = delta = 0;
+ for (;;) {
+ uint64_t timer, timeout = zfs_txg_timeout * hz;
+ uint64_t txg;
+
+ /*
+ * We sync when we're scrubbing, there's someone waiting
+ * on us, or the quiesce thread has handed off a txg to
+ * us, or we have reached our timeout.
+ */
+ timer = (delta >= timeout ? 0 : timeout - delta);
+ while ((dp->dp_scrub_func == SCRUB_FUNC_NONE ||
+ spa_shutting_down(dp->dp_spa)) &&
+ !tx->tx_exiting && timer > 0 &&
+ tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
+ tx->tx_quiesced_txg == 0) {
+ dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
+ tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
+ txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
+ delta = lbolt - start;
+ timer = (delta > timeout ? 0 : timeout - delta);
+ }
+
+ /*
+ * Wait until the quiesce thread hands off a txg to us,
+ * prompting it to do so if necessary.
+ */
+ while (!tx->tx_exiting && tx->tx_quiesced_txg == 0) {
+ if (tx->tx_quiesce_txg_waiting < tx->tx_open_txg+1)
+ tx->tx_quiesce_txg_waiting = tx->tx_open_txg+1;
+ cv_broadcast(&tx->tx_quiesce_more_cv);
+ txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
+ }
+
+ if (tx->tx_exiting)
+ txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
+
+ rw_enter(&tx->tx_suspend, RW_WRITER);
+
+ /*
+ * Consume the quiesced txg which has been handed off to
+ * us. This may cause the quiescing thread to now be
+ * able to quiesce another txg, so we must signal it.
+ */
+ txg = tx->tx_quiesced_txg;
+ tx->tx_quiesced_txg = 0;
+ tx->tx_syncing_txg = txg;
+ cv_broadcast(&tx->tx_quiesce_more_cv);
+ rw_exit(&tx->tx_suspend);
+
+ dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
+ txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
+ mutex_exit(&tx->tx_sync_lock);
+
+ start = lbolt;
+ spa_sync(dp->dp_spa, txg);
+ delta = lbolt - start;
+
+ mutex_enter(&tx->tx_sync_lock);
+ rw_enter(&tx->tx_suspend, RW_WRITER);
+ tx->tx_synced_txg = txg;
+ tx->tx_syncing_txg = 0;
+ rw_exit(&tx->tx_suspend);
+ cv_broadcast(&tx->tx_sync_done_cv);
+ }
+}
+
+static void
+txg_quiesce_thread(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ callb_cpr_t cpr;
+
+ txg_thread_enter(tx, &cpr);
+
+ for (;;) {
+ uint64_t txg;
+
+ /*
+ * We quiesce when there's someone waiting on us.
+ * However, we can only have one txg in "quiescing" or
+ * "quiesced, waiting to sync" state. So we wait until
+ * the "quiesced, waiting to sync" txg has been consumed
+ * by the sync thread.
+ */
+ while (!tx->tx_exiting &&
+ (tx->tx_open_txg >= tx->tx_quiesce_txg_waiting ||
+ tx->tx_quiesced_txg != 0))
+ txg_thread_wait(tx, &cpr, &tx->tx_quiesce_more_cv, 0);
+
+ if (tx->tx_exiting)
+ txg_thread_exit(tx, &cpr, &tx->tx_quiesce_thread);
+
+ txg = tx->tx_open_txg;
+ dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
+ txg, tx->tx_quiesce_txg_waiting,
+ tx->tx_sync_txg_waiting);
+ mutex_exit(&tx->tx_sync_lock);
+ txg_quiesce(dp, txg);
+ mutex_enter(&tx->tx_sync_lock);
+
+ /*
+ * Hand this txg off to the sync thread.
+ */
+ dprintf("quiesce done, handing off txg %llu\n", txg);
+ tx->tx_quiesced_txg = txg;
+ cv_broadcast(&tx->tx_sync_more_cv);
+ cv_broadcast(&tx->tx_quiesce_done_cv);
+ }
+}
+
+/*
+ * Delay this thread by 'ticks' if we are still in the open transaction
+ * group and there is already a waiting txg quiesing or quiesced. Abort
+ * the delay if this txg stalls or enters the quiesing state.
+ */
+void
+txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ int timeout = lbolt + ticks;
+
+ /* don't delay if this txg could transition to quiesing immediately */
+ if (tx->tx_open_txg > txg ||
+ tx->tx_syncing_txg == txg-1 || tx->tx_synced_txg == txg-1)
+ return;
+
+ mutex_enter(&tx->tx_sync_lock);
+ if (tx->tx_open_txg > txg || tx->tx_synced_txg == txg-1) {
+ mutex_exit(&tx->tx_sync_lock);
+ return;
+ }
+
+ while (lbolt < timeout &&
+ tx->tx_syncing_txg < txg-1 && !txg_stalled(dp))
+ (void) cv_timedwait(&tx->tx_quiesce_more_cv, &tx->tx_sync_lock,
+ timeout);
+
+ mutex_exit(&tx->tx_sync_lock);
+}
+
+void
+txg_wait_synced(dsl_pool_t *dp, uint64_t txg)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ mutex_enter(&tx->tx_sync_lock);
+ ASSERT(tx->tx_threads == 2);
+ if (txg == 0)
+ txg = tx->tx_open_txg;
+ if (tx->tx_sync_txg_waiting < txg)
+ tx->tx_sync_txg_waiting = txg;
+ dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
+ txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
+ while (tx->tx_synced_txg < txg) {
+ dprintf("broadcasting sync more "
+ "tx_synced=%llu waiting=%llu dp=%p\n",
+ tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
+ cv_broadcast(&tx->tx_sync_more_cv);
+ cv_wait(&tx->tx_sync_done_cv, &tx->tx_sync_lock);
+ }
+ mutex_exit(&tx->tx_sync_lock);
+}
+
+void
+txg_wait_open(dsl_pool_t *dp, uint64_t txg)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ mutex_enter(&tx->tx_sync_lock);
+ ASSERT(tx->tx_threads == 2);
+ if (txg == 0)
+ txg = tx->tx_open_txg + 1;
+ if (tx->tx_quiesce_txg_waiting < txg)
+ tx->tx_quiesce_txg_waiting = txg;
+ dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
+ txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
+ while (tx->tx_open_txg < txg) {
+ cv_broadcast(&tx->tx_quiesce_more_cv);
+ cv_wait(&tx->tx_quiesce_done_cv, &tx->tx_sync_lock);
+ }
+ mutex_exit(&tx->tx_sync_lock);
+}
+
+boolean_t
+txg_stalled(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ return (tx->tx_quiesce_txg_waiting > tx->tx_open_txg);
+}
+
+boolean_t
+txg_sync_waiting(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ return (tx->tx_syncing_txg <= tx->tx_sync_txg_waiting ||
+ tx->tx_quiesced_txg != 0);
+}
+
+void
+txg_suspend(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ /* XXX some code paths suspend when they are already suspended! */
+ rw_enter(&tx->tx_suspend, RW_READER);
+}
+
+void
+txg_resume(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+ rw_exit(&tx->tx_suspend);
+}
+
+/*
+ * Per-txg object lists.
+ */
+void
+txg_list_create(txg_list_t *tl, size_t offset)
+{
+ int t;
+
+ mutex_init(&tl->tl_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ tl->tl_offset = offset;
+
+ for (t = 0; t < TXG_SIZE; t++)
+ tl->tl_head[t] = NULL;
+}
+
+void
+txg_list_destroy(txg_list_t *tl)
+{
+ int t;
+
+ for (t = 0; t < TXG_SIZE; t++)
+ ASSERT(txg_list_empty(tl, t));
+
+ mutex_destroy(&tl->tl_lock);
+}
+
+int
+txg_list_empty(txg_list_t *tl, uint64_t txg)
+{
+ return (tl->tl_head[txg & TXG_MASK] == NULL);
+}
+
+/*
+ * Add an entry to the list.
+ * Returns 0 if it's a new entry, 1 if it's already there.
+ */
+int
+txg_list_add(txg_list_t *tl, void *p, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+ int already_on_list;
+
+ mutex_enter(&tl->tl_lock);
+ already_on_list = tn->tn_member[t];
+ if (!already_on_list) {
+ tn->tn_member[t] = 1;
+ tn->tn_next[t] = tl->tl_head[t];
+ tl->tl_head[t] = tn;
+ }
+ mutex_exit(&tl->tl_lock);
+
+ return (already_on_list);
+}
+
+/*
+ * Remove the head of the list and return it.
+ */
+void *
+txg_list_remove(txg_list_t *tl, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn;
+ void *p = NULL;
+
+ mutex_enter(&tl->tl_lock);
+ if ((tn = tl->tl_head[t]) != NULL) {
+ p = (char *)tn - tl->tl_offset;
+ tl->tl_head[t] = tn->tn_next[t];
+ tn->tn_next[t] = NULL;
+ tn->tn_member[t] = 0;
+ }
+ mutex_exit(&tl->tl_lock);
+
+ return (p);
+}
+
+/*
+ * Remove a specific item from the list and return it.
+ */
+void *
+txg_list_remove_this(txg_list_t *tl, void *p, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn, **tp;
+
+ mutex_enter(&tl->tl_lock);
+
+ for (tp = &tl->tl_head[t]; (tn = *tp) != NULL; tp = &tn->tn_next[t]) {
+ if ((char *)tn - tl->tl_offset == p) {
+ *tp = tn->tn_next[t];
+ tn->tn_next[t] = NULL;
+ tn->tn_member[t] = 0;
+ mutex_exit(&tl->tl_lock);
+ return (p);
+ }
+ }
+
+ mutex_exit(&tl->tl_lock);
+
+ return (NULL);
+}
+
+int
+txg_list_member(txg_list_t *tl, void *p, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+
+ return (tn->tn_member[t]);
+}
+
+/*
+ * Walk a txg list -- only safe if you know it's not changing.
+ */
+void *
+txg_list_head(txg_list_t *tl, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn = tl->tl_head[t];
+
+ return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
+}
+
+void *
+txg_list_next(txg_list_t *tl, void *p, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+
+ tn = tn->tn_next[t];
+
+ return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
+}
diff --git a/module/zfs/uberblock.c b/module/zfs/uberblock.c
new file mode 100644
index 000000000..34d7e0c3a
--- /dev/null
+++ b/module/zfs/uberblock.c
@@ -0,0 +1,63 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/uberblock_impl.h>
+#include <sys/vdev_impl.h>
+
+int
+uberblock_verify(uberblock_t *ub)
+{
+ if (ub->ub_magic == BSWAP_64((uint64_t)UBERBLOCK_MAGIC))
+ byteswap_uint64_array(ub, sizeof (uberblock_t));
+
+ if (ub->ub_magic != UBERBLOCK_MAGIC)
+ return (EINVAL);
+
+ return (0);
+}
+
+/*
+ * Update the uberblock and return a boolean value indicating whether
+ * anything changed in this transaction group.
+ */
+int
+uberblock_update(uberblock_t *ub, vdev_t *rvd, uint64_t txg)
+{
+ ASSERT(ub->ub_txg < txg);
+
+ /*
+ * We explicitly do not set ub_version here, so that older versions
+ * continue to be written with the previous uberblock version.
+ */
+ ub->ub_magic = UBERBLOCK_MAGIC;
+ ub->ub_txg = txg;
+ ub->ub_guid_sum = rvd->vdev_guid_sum;
+ ub->ub_timestamp = gethrestime_sec();
+
+ return (ub->ub_rootbp.blk_birth == txg);
+}
diff --git a/module/zfs/unique.c b/module/zfs/unique.c
new file mode 100644
index 000000000..fbe7b619a
--- /dev/null
+++ b/module/zfs/unique.c
@@ -0,0 +1,116 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/avl.h>
+#include <sys/unique.h>
+
+static avl_tree_t unique_avl;
+static kmutex_t unique_mtx;
+
+typedef struct unique {
+ avl_node_t un_link;
+ uint64_t un_value;
+} unique_t;
+
+#define UNIQUE_MASK ((1ULL << UNIQUE_BITS) - 1)
+
+static int
+unique_compare(const void *a, const void *b)
+{
+ const unique_t *una = a;
+ const unique_t *unb = b;
+
+ if (una->un_value < unb->un_value)
+ return (-1);
+ if (una->un_value > unb->un_value)
+ return (+1);
+ return (0);
+}
+
+void
+unique_init(void)
+{
+ avl_create(&unique_avl, unique_compare,
+ sizeof (unique_t), offsetof(unique_t, un_link));
+ mutex_init(&unique_mtx, NULL, MUTEX_DEFAULT, NULL);
+}
+
+void
+unique_fini(void)
+{
+ avl_destroy(&unique_avl);
+ mutex_destroy(&unique_mtx);
+}
+
+uint64_t
+unique_create(void)
+{
+ uint64_t value = unique_insert(0);
+ unique_remove(value);
+ return (value);
+}
+
+uint64_t
+unique_insert(uint64_t value)
+{
+ avl_index_t idx;
+ unique_t *un = kmem_alloc(sizeof (unique_t), KM_SLEEP);
+
+ un->un_value = value;
+
+ mutex_enter(&unique_mtx);
+ while (un->un_value == 0 || un->un_value & ~UNIQUE_MASK ||
+ avl_find(&unique_avl, un, &idx)) {
+ mutex_exit(&unique_mtx);
+ (void) random_get_pseudo_bytes((void*)&un->un_value,
+ sizeof (un->un_value));
+ un->un_value &= UNIQUE_MASK;
+ mutex_enter(&unique_mtx);
+ }
+
+ avl_insert(&unique_avl, un, idx);
+ mutex_exit(&unique_mtx);
+
+ return (un->un_value);
+}
+
+void
+unique_remove(uint64_t value)
+{
+ unique_t un_tofind;
+ unique_t *un;
+
+ un_tofind.un_value = value;
+ mutex_enter(&unique_mtx);
+ un = avl_find(&unique_avl, &un_tofind, NULL);
+ if (un != NULL) {
+ avl_remove(&unique_avl, un);
+ kmem_free(un, sizeof (unique_t));
+ }
+ mutex_exit(&unique_mtx);
+}
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
new file mode 100644
index 000000000..16a27e514
--- /dev/null
+++ b/module/zfs/vdev.c
@@ -0,0 +1,2425 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/fm/fs/zfs.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/dmu.h>
+#include <sys/dmu_tx.h>
+#include <sys/vdev_impl.h>
+#include <sys/uberblock_impl.h>
+#include <sys/metaslab.h>
+#include <sys/metaslab_impl.h>
+#include <sys/space_map.h>
+#include <sys/zio.h>
+#include <sys/zap.h>
+#include <sys/fs/zfs.h>
+#include <sys/arc.h>
+
+/*
+ * Virtual device management.
+ */
+
+static vdev_ops_t *vdev_ops_table[] = {
+ &vdev_root_ops,
+ &vdev_raidz_ops,
+ &vdev_mirror_ops,
+ &vdev_replacing_ops,
+ &vdev_spare_ops,
+ &vdev_disk_ops,
+ &vdev_file_ops,
+ &vdev_missing_ops,
+ NULL
+};
+
+/* maximum scrub/resilver I/O queue per leaf vdev */
+int zfs_scrub_limit = 10;
+
+/*
+ * Given a vdev type, return the appropriate ops vector.
+ */
+static vdev_ops_t *
+vdev_getops(const char *type)
+{
+ vdev_ops_t *ops, **opspp;
+
+ for (opspp = vdev_ops_table; (ops = *opspp) != NULL; opspp++)
+ if (strcmp(ops->vdev_op_type, type) == 0)
+ break;
+
+ return (ops);
+}
+
+/*
+ * Default asize function: return the MAX of psize with the asize of
+ * all children. This is what's used by anything other than RAID-Z.
+ */
+uint64_t
+vdev_default_asize(vdev_t *vd, uint64_t psize)
+{
+ uint64_t asize = P2ROUNDUP(psize, 1ULL << vd->vdev_top->vdev_ashift);
+ uint64_t csize;
+ uint64_t c;
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ csize = vdev_psize_to_asize(vd->vdev_child[c], psize);
+ asize = MAX(asize, csize);
+ }
+
+ return (asize);
+}
+
+/*
+ * Get the replaceable or attachable device size.
+ * If the parent is a mirror or raidz, the replaceable size is the minimum
+ * psize of all its children. For the rest, just return our own psize.
+ *
+ * e.g.
+ * psize rsize
+ * root - -
+ * mirror/raidz - -
+ * disk1 20g 20g
+ * disk2 40g 20g
+ * disk3 80g 80g
+ */
+uint64_t
+vdev_get_rsize(vdev_t *vd)
+{
+ vdev_t *pvd, *cvd;
+ uint64_t c, rsize;
+
+ pvd = vd->vdev_parent;
+
+ /*
+ * If our parent is NULL or the root, just return our own psize.
+ */
+ if (pvd == NULL || pvd->vdev_parent == NULL)
+ return (vd->vdev_psize);
+
+ rsize = 0;
+
+ for (c = 0; c < pvd->vdev_children; c++) {
+ cvd = pvd->vdev_child[c];
+ rsize = MIN(rsize - 1, cvd->vdev_psize - 1) + 1;
+ }
+
+ return (rsize);
+}
+
+vdev_t *
+vdev_lookup_top(spa_t *spa, uint64_t vdev)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
+
+ if (vdev < rvd->vdev_children) {
+ ASSERT(rvd->vdev_child[vdev] != NULL);
+ return (rvd->vdev_child[vdev]);
+ }
+
+ return (NULL);
+}
+
+vdev_t *
+vdev_lookup_by_guid(vdev_t *vd, uint64_t guid)
+{
+ int c;
+ vdev_t *mvd;
+
+ if (vd->vdev_guid == guid)
+ return (vd);
+
+ for (c = 0; c < vd->vdev_children; c++)
+ if ((mvd = vdev_lookup_by_guid(vd->vdev_child[c], guid)) !=
+ NULL)
+ return (mvd);
+
+ return (NULL);
+}
+
+void
+vdev_add_child(vdev_t *pvd, vdev_t *cvd)
+{
+ size_t oldsize, newsize;
+ uint64_t id = cvd->vdev_id;
+ vdev_t **newchild;
+
+ ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+ ASSERT(cvd->vdev_parent == NULL);
+
+ cvd->vdev_parent = pvd;
+
+ if (pvd == NULL)
+ return;
+
+ ASSERT(id >= pvd->vdev_children || pvd->vdev_child[id] == NULL);
+
+ oldsize = pvd->vdev_children * sizeof (vdev_t *);
+ pvd->vdev_children = MAX(pvd->vdev_children, id + 1);
+ newsize = pvd->vdev_children * sizeof (vdev_t *);
+
+ newchild = kmem_zalloc(newsize, KM_SLEEP);
+ if (pvd->vdev_child != NULL) {
+ bcopy(pvd->vdev_child, newchild, oldsize);
+ kmem_free(pvd->vdev_child, oldsize);
+ }
+
+ pvd->vdev_child = newchild;
+ pvd->vdev_child[id] = cvd;
+
+ cvd->vdev_top = (pvd->vdev_top ? pvd->vdev_top: cvd);
+ ASSERT(cvd->vdev_top->vdev_parent->vdev_parent == NULL);
+
+ /*
+ * Walk up all ancestors to update guid sum.
+ */
+ for (; pvd != NULL; pvd = pvd->vdev_parent)
+ pvd->vdev_guid_sum += cvd->vdev_guid_sum;
+
+ if (cvd->vdev_ops->vdev_op_leaf)
+ cvd->vdev_spa->spa_scrub_maxinflight += zfs_scrub_limit;
+}
+
+void
+vdev_remove_child(vdev_t *pvd, vdev_t *cvd)
+{
+ int c;
+ uint_t id = cvd->vdev_id;
+
+ ASSERT(cvd->vdev_parent == pvd);
+
+ if (pvd == NULL)
+ return;
+
+ ASSERT(id < pvd->vdev_children);
+ ASSERT(pvd->vdev_child[id] == cvd);
+
+ pvd->vdev_child[id] = NULL;
+ cvd->vdev_parent = NULL;
+
+ for (c = 0; c < pvd->vdev_children; c++)
+ if (pvd->vdev_child[c])
+ break;
+
+ if (c == pvd->vdev_children) {
+ kmem_free(pvd->vdev_child, c * sizeof (vdev_t *));
+ pvd->vdev_child = NULL;
+ pvd->vdev_children = 0;
+ }
+
+ /*
+ * Walk up all ancestors to update guid sum.
+ */
+ for (; pvd != NULL; pvd = pvd->vdev_parent)
+ pvd->vdev_guid_sum -= cvd->vdev_guid_sum;
+
+ if (cvd->vdev_ops->vdev_op_leaf)
+ cvd->vdev_spa->spa_scrub_maxinflight -= zfs_scrub_limit;
+}
+
+/*
+ * Remove any holes in the child array.
+ */
+void
+vdev_compact_children(vdev_t *pvd)
+{
+ vdev_t **newchild, *cvd;
+ int oldc = pvd->vdev_children;
+ int newc, c;
+
+ ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ for (c = newc = 0; c < oldc; c++)
+ if (pvd->vdev_child[c])
+ newc++;
+
+ newchild = kmem_alloc(newc * sizeof (vdev_t *), KM_SLEEP);
+
+ for (c = newc = 0; c < oldc; c++) {
+ if ((cvd = pvd->vdev_child[c]) != NULL) {
+ newchild[newc] = cvd;
+ cvd->vdev_id = newc++;
+ }
+ }
+
+ kmem_free(pvd->vdev_child, oldc * sizeof (vdev_t *));
+ pvd->vdev_child = newchild;
+ pvd->vdev_children = newc;
+}
+
+/*
+ * Allocate and minimally initialize a vdev_t.
+ */
+static vdev_t *
+vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
+{
+ vdev_t *vd;
+
+ vd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
+
+ if (spa->spa_root_vdev == NULL) {
+ ASSERT(ops == &vdev_root_ops);
+ spa->spa_root_vdev = vd;
+ }
+
+ if (guid == 0) {
+ if (spa->spa_root_vdev == vd) {
+ /*
+ * The root vdev's guid will also be the pool guid,
+ * which must be unique among all pools.
+ */
+ while (guid == 0 || spa_guid_exists(guid, 0))
+ guid = spa_get_random(-1ULL);
+ } else {
+ /*
+ * Any other vdev's guid must be unique within the pool.
+ */
+ while (guid == 0 ||
+ spa_guid_exists(spa_guid(spa), guid))
+ guid = spa_get_random(-1ULL);
+ }
+ ASSERT(!spa_guid_exists(spa_guid(spa), guid));
+ }
+
+ vd->vdev_spa = spa;
+ vd->vdev_id = id;
+ vd->vdev_guid = guid;
+ vd->vdev_guid_sum = guid;
+ vd->vdev_ops = ops;
+ vd->vdev_state = VDEV_STATE_CLOSED;
+
+ mutex_init(&vd->vdev_dtl_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
+ space_map_create(&vd->vdev_dtl_map, 0, -1ULL, 0, &vd->vdev_dtl_lock);
+ space_map_create(&vd->vdev_dtl_scrub, 0, -1ULL, 0, &vd->vdev_dtl_lock);
+ txg_list_create(&vd->vdev_ms_list,
+ offsetof(struct metaslab, ms_txg_node));
+ txg_list_create(&vd->vdev_dtl_list,
+ offsetof(struct vdev, vdev_dtl_node));
+ vd->vdev_stat.vs_timestamp = gethrtime();
+ vdev_queue_init(vd);
+ vdev_cache_init(vd);
+
+ return (vd);
+}
+
+/*
+ * Allocate a new vdev. The 'alloctype' is used to control whether we are
+ * creating a new vdev or loading an existing one - the behavior is slightly
+ * different for each case.
+ */
+int
+vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
+ int alloctype)
+{
+ vdev_ops_t *ops;
+ char *type;
+ uint64_t guid = 0, islog, nparity;
+ vdev_t *vd;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
+ return (EINVAL);
+
+ if ((ops = vdev_getops(type)) == NULL)
+ return (EINVAL);
+
+ /*
+ * If this is a load, get the vdev guid from the nvlist.
+ * Otherwise, vdev_alloc_common() will generate one for us.
+ */
+ if (alloctype == VDEV_ALLOC_LOAD) {
+ uint64_t label_id;
+
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, &label_id) ||
+ label_id != id)
+ return (EINVAL);
+
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
+ return (EINVAL);
+ } else if (alloctype == VDEV_ALLOC_SPARE) {
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
+ return (EINVAL);
+ } else if (alloctype == VDEV_ALLOC_L2CACHE) {
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
+ return (EINVAL);
+ }
+
+ /*
+ * The first allocated vdev must be of type 'root'.
+ */
+ if (ops != &vdev_root_ops && spa->spa_root_vdev == NULL)
+ return (EINVAL);
+
+ /*
+ * Determine whether we're a log vdev.
+ */
+ islog = 0;
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &islog);
+ if (islog && spa_version(spa) < SPA_VERSION_SLOGS)
+ return (ENOTSUP);
+
+ /*
+ * Set the nparity property for RAID-Z vdevs.
+ */
+ nparity = -1ULL;
+ if (ops == &vdev_raidz_ops) {
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
+ &nparity) == 0) {
+ /*
+ * Currently, we can only support 2 parity devices.
+ */
+ if (nparity == 0 || nparity > 2)
+ return (EINVAL);
+ /*
+ * Older versions can only support 1 parity device.
+ */
+ if (nparity == 2 &&
+ spa_version(spa) < SPA_VERSION_RAID6)
+ return (ENOTSUP);
+ } else {
+ /*
+ * We require the parity to be specified for SPAs that
+ * support multiple parity levels.
+ */
+ if (spa_version(spa) >= SPA_VERSION_RAID6)
+ return (EINVAL);
+ /*
+ * Otherwise, we default to 1 parity device for RAID-Z.
+ */
+ nparity = 1;
+ }
+ } else {
+ nparity = 0;
+ }
+ ASSERT(nparity != -1ULL);
+
+ vd = vdev_alloc_common(spa, id, guid, ops);
+
+ vd->vdev_islog = islog;
+ vd->vdev_nparity = nparity;
+
+ if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &vd->vdev_path) == 0)
+ vd->vdev_path = spa_strdup(vd->vdev_path);
+ if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &vd->vdev_devid) == 0)
+ vd->vdev_devid = spa_strdup(vd->vdev_devid);
+ if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
+ &vd->vdev_physpath) == 0)
+ vd->vdev_physpath = spa_strdup(vd->vdev_physpath);
+
+ /*
+ * Set the whole_disk property. If it's not specified, leave the value
+ * as -1.
+ */
+ if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
+ &vd->vdev_wholedisk) != 0)
+ vd->vdev_wholedisk = -1ULL;
+
+ /*
+ * Look for the 'not present' flag. This will only be set if the device
+ * was not present at the time of import.
+ */
+ if (!spa->spa_import_faulted)
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
+ &vd->vdev_not_present);
+
+ /*
+ * Get the alignment requirement.
+ */
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &vd->vdev_ashift);
+
+ /*
+ * If we're a top-level vdev, try to load the allocation parameters.
+ */
+ if (parent && !parent->vdev_parent && alloctype == VDEV_ALLOC_LOAD) {
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
+ &vd->vdev_ms_array);
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
+ &vd->vdev_ms_shift);
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASIZE,
+ &vd->vdev_asize);
+ }
+
+ /*
+ * If we're a leaf vdev, try to load the DTL object and other state.
+ */
+ if (vd->vdev_ops->vdev_op_leaf &&
+ (alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_L2CACHE)) {
+ if (alloctype == VDEV_ALLOC_LOAD) {
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
+ &vd->vdev_dtl.smo_object);
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
+ &vd->vdev_unspare);
+ }
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE,
+ &vd->vdev_offline);
+
+ /*
+ * When importing a pool, we want to ignore the persistent fault
+ * state, as the diagnosis made on another system may not be
+ * valid in the current context.
+ */
+ if (spa->spa_load_state == SPA_LOAD_OPEN) {
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED,
+ &vd->vdev_faulted);
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DEGRADED,
+ &vd->vdev_degraded);
+ (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED,
+ &vd->vdev_removed);
+ }
+ }
+
+ /*
+ * Add ourselves to the parent's list of children.
+ */
+ vdev_add_child(parent, vd);
+
+ *vdp = vd;
+
+ return (0);
+}
+
+void
+vdev_free(vdev_t *vd)
+{
+ int c;
+ spa_t *spa = vd->vdev_spa;
+
+ /*
+ * vdev_free() implies closing the vdev first. This is simpler than
+ * trying to ensure complicated semantics for all callers.
+ */
+ vdev_close(vd);
+
+ ASSERT(!list_link_active(&vd->vdev_config_dirty_node));
+
+ /*
+ * Free all children.
+ */
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_free(vd->vdev_child[c]);
+
+ ASSERT(vd->vdev_child == NULL);
+ ASSERT(vd->vdev_guid_sum == vd->vdev_guid);
+
+ /*
+ * Discard allocation state.
+ */
+ if (vd == vd->vdev_top)
+ vdev_metaslab_fini(vd);
+
+ ASSERT3U(vd->vdev_stat.vs_space, ==, 0);
+ ASSERT3U(vd->vdev_stat.vs_dspace, ==, 0);
+ ASSERT3U(vd->vdev_stat.vs_alloc, ==, 0);
+
+ /*
+ * Remove this vdev from its parent's child list.
+ */
+ vdev_remove_child(vd->vdev_parent, vd);
+
+ ASSERT(vd->vdev_parent == NULL);
+
+ /*
+ * Clean up vdev structure.
+ */
+ vdev_queue_fini(vd);
+ vdev_cache_fini(vd);
+
+ if (vd->vdev_path)
+ spa_strfree(vd->vdev_path);
+ if (vd->vdev_devid)
+ spa_strfree(vd->vdev_devid);
+ if (vd->vdev_physpath)
+ spa_strfree(vd->vdev_physpath);
+
+ if (vd->vdev_isspare)
+ spa_spare_remove(vd);
+ if (vd->vdev_isl2cache)
+ spa_l2cache_remove(vd);
+
+ txg_list_destroy(&vd->vdev_ms_list);
+ txg_list_destroy(&vd->vdev_dtl_list);
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_unload(&vd->vdev_dtl_map);
+ space_map_destroy(&vd->vdev_dtl_map);
+ space_map_vacate(&vd->vdev_dtl_scrub, NULL, NULL);
+ space_map_destroy(&vd->vdev_dtl_scrub);
+ mutex_exit(&vd->vdev_dtl_lock);
+ mutex_destroy(&vd->vdev_dtl_lock);
+ mutex_destroy(&vd->vdev_stat_lock);
+ mutex_destroy(&vd->vdev_probe_lock);
+
+ if (vd == spa->spa_root_vdev)
+ spa->spa_root_vdev = NULL;
+
+ kmem_free(vd, sizeof (vdev_t));
+}
+
+/*
+ * Transfer top-level vdev state from svd to tvd.
+ */
+static void
+vdev_top_transfer(vdev_t *svd, vdev_t *tvd)
+{
+ spa_t *spa = svd->vdev_spa;
+ metaslab_t *msp;
+ vdev_t *vd;
+ int t;
+
+ ASSERT(tvd == tvd->vdev_top);
+
+ tvd->vdev_ms_array = svd->vdev_ms_array;
+ tvd->vdev_ms_shift = svd->vdev_ms_shift;
+ tvd->vdev_ms_count = svd->vdev_ms_count;
+
+ svd->vdev_ms_array = 0;
+ svd->vdev_ms_shift = 0;
+ svd->vdev_ms_count = 0;
+
+ tvd->vdev_mg = svd->vdev_mg;
+ tvd->vdev_ms = svd->vdev_ms;
+
+ svd->vdev_mg = NULL;
+ svd->vdev_ms = NULL;
+
+ if (tvd->vdev_mg != NULL)
+ tvd->vdev_mg->mg_vd = tvd;
+
+ tvd->vdev_stat.vs_alloc = svd->vdev_stat.vs_alloc;
+ tvd->vdev_stat.vs_space = svd->vdev_stat.vs_space;
+ tvd->vdev_stat.vs_dspace = svd->vdev_stat.vs_dspace;
+
+ svd->vdev_stat.vs_alloc = 0;
+ svd->vdev_stat.vs_space = 0;
+ svd->vdev_stat.vs_dspace = 0;
+
+ for (t = 0; t < TXG_SIZE; t++) {
+ while ((msp = txg_list_remove(&svd->vdev_ms_list, t)) != NULL)
+ (void) txg_list_add(&tvd->vdev_ms_list, msp, t);
+ while ((vd = txg_list_remove(&svd->vdev_dtl_list, t)) != NULL)
+ (void) txg_list_add(&tvd->vdev_dtl_list, vd, t);
+ if (txg_list_remove_this(&spa->spa_vdev_txg_list, svd, t))
+ (void) txg_list_add(&spa->spa_vdev_txg_list, tvd, t);
+ }
+
+ if (list_link_active(&svd->vdev_config_dirty_node)) {
+ vdev_config_clean(svd);
+ vdev_config_dirty(tvd);
+ }
+
+ if (list_link_active(&svd->vdev_state_dirty_node)) {
+ vdev_state_clean(svd);
+ vdev_state_dirty(tvd);
+ }
+
+ tvd->vdev_deflate_ratio = svd->vdev_deflate_ratio;
+ svd->vdev_deflate_ratio = 0;
+
+ tvd->vdev_islog = svd->vdev_islog;
+ svd->vdev_islog = 0;
+}
+
+static void
+vdev_top_update(vdev_t *tvd, vdev_t *vd)
+{
+ int c;
+
+ if (vd == NULL)
+ return;
+
+ vd->vdev_top = tvd;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_top_update(tvd, vd->vdev_child[c]);
+}
+
+/*
+ * Add a mirror/replacing vdev above an existing vdev.
+ */
+vdev_t *
+vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops)
+{
+ spa_t *spa = cvd->vdev_spa;
+ vdev_t *pvd = cvd->vdev_parent;
+ vdev_t *mvd;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ mvd = vdev_alloc_common(spa, cvd->vdev_id, 0, ops);
+
+ mvd->vdev_asize = cvd->vdev_asize;
+ mvd->vdev_ashift = cvd->vdev_ashift;
+ mvd->vdev_state = cvd->vdev_state;
+
+ vdev_remove_child(pvd, cvd);
+ vdev_add_child(pvd, mvd);
+ cvd->vdev_id = mvd->vdev_children;
+ vdev_add_child(mvd, cvd);
+ vdev_top_update(cvd->vdev_top, cvd->vdev_top);
+
+ if (mvd == mvd->vdev_top)
+ vdev_top_transfer(cvd, mvd);
+
+ return (mvd);
+}
+
+/*
+ * Remove a 1-way mirror/replacing vdev from the tree.
+ */
+void
+vdev_remove_parent(vdev_t *cvd)
+{
+ vdev_t *mvd = cvd->vdev_parent;
+ vdev_t *pvd = mvd->vdev_parent;
+
+ ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ ASSERT(mvd->vdev_children == 1);
+ ASSERT(mvd->vdev_ops == &vdev_mirror_ops ||
+ mvd->vdev_ops == &vdev_replacing_ops ||
+ mvd->vdev_ops == &vdev_spare_ops);
+ cvd->vdev_ashift = mvd->vdev_ashift;
+
+ vdev_remove_child(mvd, cvd);
+ vdev_remove_child(pvd, mvd);
+ /*
+ * If cvd will replace mvd as a top-level vdev, preserve mvd's guid.
+ * Otherwise, we could have detached an offline device, and when we
+ * go to import the pool we'll think we have two top-level vdevs,
+ * instead of a different version of the same top-level vdev.
+ */
+ if (mvd->vdev_top == mvd)
+ cvd->vdev_guid = cvd->vdev_guid_sum = mvd->vdev_guid;
+ cvd->vdev_id = mvd->vdev_id;
+ vdev_add_child(pvd, cvd);
+ vdev_top_update(cvd->vdev_top, cvd->vdev_top);
+
+ if (cvd == cvd->vdev_top)
+ vdev_top_transfer(mvd, cvd);
+
+ ASSERT(mvd->vdev_children == 0);
+ vdev_free(mvd);
+}
+
+int
+vdev_metaslab_init(vdev_t *vd, uint64_t txg)
+{
+ spa_t *spa = vd->vdev_spa;
+ objset_t *mos = spa->spa_meta_objset;
+ metaslab_class_t *mc;
+ uint64_t m;
+ uint64_t oldc = vd->vdev_ms_count;
+ uint64_t newc = vd->vdev_asize >> vd->vdev_ms_shift;
+ metaslab_t **mspp;
+ int error;
+
+ if (vd->vdev_ms_shift == 0) /* not being allocated from yet */
+ return (0);
+
+ ASSERT(oldc <= newc);
+
+ if (vd->vdev_islog)
+ mc = spa->spa_log_class;
+ else
+ mc = spa->spa_normal_class;
+
+ if (vd->vdev_mg == NULL)
+ vd->vdev_mg = metaslab_group_create(mc, vd);
+
+ mspp = kmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);
+
+ if (oldc != 0) {
+ bcopy(vd->vdev_ms, mspp, oldc * sizeof (*mspp));
+ kmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
+ }
+
+ vd->vdev_ms = mspp;
+ vd->vdev_ms_count = newc;
+
+ for (m = oldc; m < newc; m++) {
+ space_map_obj_t smo = { 0, 0, 0 };
+ if (txg == 0) {
+ uint64_t object = 0;
+ error = dmu_read(mos, vd->vdev_ms_array,
+ m * sizeof (uint64_t), sizeof (uint64_t), &object);
+ if (error)
+ return (error);
+ if (object != 0) {
+ dmu_buf_t *db;
+ error = dmu_bonus_hold(mos, object, FTAG, &db);
+ if (error)
+ return (error);
+ ASSERT3U(db->db_size, >=, sizeof (smo));
+ bcopy(db->db_data, &smo, sizeof (smo));
+ ASSERT3U(smo.smo_object, ==, object);
+ dmu_buf_rele(db, FTAG);
+ }
+ }
+ vd->vdev_ms[m] = metaslab_init(vd->vdev_mg, &smo,
+ m << vd->vdev_ms_shift, 1ULL << vd->vdev_ms_shift, txg);
+ }
+
+ return (0);
+}
+
+void
+vdev_metaslab_fini(vdev_t *vd)
+{
+ uint64_t m;
+ uint64_t count = vd->vdev_ms_count;
+
+ if (vd->vdev_ms != NULL) {
+ for (m = 0; m < count; m++)
+ if (vd->vdev_ms[m] != NULL)
+ metaslab_fini(vd->vdev_ms[m]);
+ kmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
+ vd->vdev_ms = NULL;
+ }
+}
+
+typedef struct vdev_probe_stats {
+ boolean_t vps_readable;
+ boolean_t vps_writeable;
+ int vps_flags;
+ zio_t *vps_root;
+ vdev_t *vps_vd;
+} vdev_probe_stats_t;
+
+static void
+vdev_probe_done(zio_t *zio)
+{
+ vdev_probe_stats_t *vps = zio->io_private;
+ vdev_t *vd = vps->vps_vd;
+
+ if (zio->io_type == ZIO_TYPE_READ) {
+ ASSERT(zio->io_vd == vd);
+ if (zio->io_error == 0)
+ vps->vps_readable = 1;
+ if (zio->io_error == 0 && (spa_mode & FWRITE)) {
+ zio_nowait(zio_write_phys(vps->vps_root, vd,
+ zio->io_offset, zio->io_size, zio->io_data,
+ ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
+ ZIO_PRIORITY_SYNC_WRITE, vps->vps_flags, B_TRUE));
+ } else {
+ zio_buf_free(zio->io_data, zio->io_size);
+ }
+ } else if (zio->io_type == ZIO_TYPE_WRITE) {
+ ASSERT(zio->io_vd == vd);
+ if (zio->io_error == 0)
+ vps->vps_writeable = 1;
+ zio_buf_free(zio->io_data, zio->io_size);
+ } else if (zio->io_type == ZIO_TYPE_NULL) {
+ ASSERT(zio->io_vd == NULL);
+ ASSERT(zio == vps->vps_root);
+
+ vd->vdev_cant_read |= !vps->vps_readable;
+ vd->vdev_cant_write |= !vps->vps_writeable;
+
+ if (vdev_readable(vd) &&
+ (vdev_writeable(vd) || !(spa_mode & FWRITE))) {
+ zio->io_error = 0;
+ } else {
+ ASSERT(zio->io_error != 0);
+ zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
+ zio->io_spa, vd, NULL, 0, 0);
+ zio->io_error = ENXIO;
+ }
+ kmem_free(vps, sizeof (*vps));
+ }
+}
+
+/*
+ * Determine whether this device is accessible by reading and writing
+ * to several known locations: the pad regions of each vdev label
+ * but the first (which we leave alone in case it contains a VTOC).
+ */
+zio_t *
+vdev_probe(vdev_t *vd, zio_t *pio)
+{
+ spa_t *spa = vd->vdev_spa;
+ vdev_probe_stats_t *vps;
+ zio_t *zio;
+
+ vps = kmem_zalloc(sizeof (*vps), KM_SLEEP);
+
+ vps->vps_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_PROBE |
+ ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE | ZIO_FLAG_DONT_RETRY;
+
+ if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
+ /*
+ * vdev_cant_read and vdev_cant_write can only transition
+ * from TRUE to FALSE when we have the SCL_ZIO lock as writer;
+ * otherwise they can only transition from FALSE to TRUE.
+ * This ensures that any zio looking at these values can
+ * assume that failures persist for the life of the I/O.
+ * That's important because when a device has intermittent
+ * connectivity problems, we want to ensure that they're
+ * ascribed to the device (ENXIO) and not the zio (EIO).
+ *
+ * Since we hold SCL_ZIO as writer here, clear both values
+ * so the probe can reevaluate from first principles.
+ */
+ vps->vps_flags |= ZIO_FLAG_CONFIG_WRITER;
+ vd->vdev_cant_read = B_FALSE;
+ vd->vdev_cant_write = B_FALSE;
+ }
+
+ ASSERT(vd->vdev_ops->vdev_op_leaf);
+
+ zio = zio_null(pio, spa, vdev_probe_done, vps, vps->vps_flags);
+
+ vps->vps_root = zio;
+ vps->vps_vd = vd;
+
+ for (int l = 1; l < VDEV_LABELS; l++) {
+ zio_nowait(zio_read_phys(zio, vd,
+ vdev_label_offset(vd->vdev_psize, l,
+ offsetof(vdev_label_t, vl_pad)),
+ VDEV_SKIP_SIZE, zio_buf_alloc(VDEV_SKIP_SIZE),
+ ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
+ ZIO_PRIORITY_SYNC_READ, vps->vps_flags, B_TRUE));
+ }
+
+ return (zio);
+}
+
+/*
+ * Prepare a virtual device for access.
+ */
+int
+vdev_open(vdev_t *vd)
+{
+ int error;
+ int c;
+ uint64_t osize = 0;
+ uint64_t asize, psize;
+ uint64_t ashift = 0;
+
+ ASSERT(vd->vdev_state == VDEV_STATE_CLOSED ||
+ vd->vdev_state == VDEV_STATE_CANT_OPEN ||
+ vd->vdev_state == VDEV_STATE_OFFLINE);
+
+ vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
+
+ if (!vd->vdev_removed && vd->vdev_faulted) {
+ ASSERT(vd->vdev_children == 0);
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
+ VDEV_AUX_ERR_EXCEEDED);
+ return (ENXIO);
+ } else if (vd->vdev_offline) {
+ ASSERT(vd->vdev_children == 0);
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE, VDEV_AUX_NONE);
+ return (ENXIO);
+ }
+
+ error = vd->vdev_ops->vdev_op_open(vd, &osize, &ashift);
+
+ if (zio_injection_enabled && error == 0)
+ error = zio_handle_device_injection(vd, ENXIO);
+
+ if (error) {
+ if (vd->vdev_removed &&
+ vd->vdev_stat.vs_aux != VDEV_AUX_OPEN_FAILED)
+ vd->vdev_removed = B_FALSE;
+
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ vd->vdev_stat.vs_aux);
+ return (error);
+ }
+
+ vd->vdev_removed = B_FALSE;
+
+ if (vd->vdev_degraded) {
+ ASSERT(vd->vdev_children == 0);
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
+ VDEV_AUX_ERR_EXCEEDED);
+ } else {
+ vd->vdev_state = VDEV_STATE_HEALTHY;
+ }
+
+ for (c = 0; c < vd->vdev_children; c++)
+ if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
+ VDEV_AUX_NONE);
+ break;
+ }
+
+ osize = P2ALIGN(osize, (uint64_t)sizeof (vdev_label_t));
+
+ if (vd->vdev_children == 0) {
+ if (osize < SPA_MINDEVSIZE) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_TOO_SMALL);
+ return (EOVERFLOW);
+ }
+ psize = osize;
+ asize = osize - (VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE);
+ } else {
+ if (vd->vdev_parent != NULL && osize < SPA_MINDEVSIZE -
+ (VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE)) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_TOO_SMALL);
+ return (EOVERFLOW);
+ }
+ psize = 0;
+ asize = osize;
+ }
+
+ vd->vdev_psize = psize;
+
+ if (vd->vdev_asize == 0) {
+ /*
+ * This is the first-ever open, so use the computed values.
+ * For testing purposes, a higher ashift can be requested.
+ */
+ vd->vdev_asize = asize;
+ vd->vdev_ashift = MAX(ashift, vd->vdev_ashift);
+ } else {
+ /*
+ * Make sure the alignment requirement hasn't increased.
+ */
+ if (ashift > vd->vdev_top->vdev_ashift) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_BAD_LABEL);
+ return (EINVAL);
+ }
+
+ /*
+ * Make sure the device hasn't shrunk.
+ */
+ if (asize < vd->vdev_asize) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_BAD_LABEL);
+ return (EINVAL);
+ }
+
+ /*
+ * If all children are healthy and the asize has increased,
+ * then we've experienced dynamic LUN growth.
+ */
+ if (vd->vdev_state == VDEV_STATE_HEALTHY &&
+ asize > vd->vdev_asize) {
+ vd->vdev_asize = asize;
+ }
+ }
+
+ /*
+ * Ensure we can issue some IO before declaring the
+ * vdev open for business.
+ */
+ if (vd->vdev_ops->vdev_op_leaf &&
+ (error = zio_wait(vdev_probe(vd, NULL))) != 0) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_IO_FAILURE);
+ return (error);
+ }
+
+ /*
+ * If this is a top-level vdev, compute the raidz-deflation
+ * ratio. Note, we hard-code in 128k (1<<17) because it is the
+ * current "typical" blocksize. Even if SPA_MAXBLOCKSIZE
+ * changes, this algorithm must never change, or we will
+ * inconsistently account for existing bp's.
+ */
+ if (vd->vdev_top == vd) {
+ vd->vdev_deflate_ratio = (1<<17) /
+ (vdev_psize_to_asize(vd, 1<<17) >> SPA_MINBLOCKSHIFT);
+ }
+
+ /*
+ * If a leaf vdev has a DTL, and seems healthy, then kick off a
+ * resilver. But don't do this if we are doing a reopen for a
+ * scrub, since this would just restart the scrub we are already
+ * doing.
+ */
+ if (vd->vdev_children == 0 && !vd->vdev_spa->spa_scrub_reopen) {
+ mutex_enter(&vd->vdev_dtl_lock);
+ if (vd->vdev_dtl_map.sm_space != 0 && vdev_writeable(vd))
+ spa_async_request(vd->vdev_spa, SPA_ASYNC_RESILVER);
+ mutex_exit(&vd->vdev_dtl_lock);
+ }
+
+ return (0);
+}
+
+/*
+ * Called once the vdevs are all opened, this routine validates the label
+ * contents. This needs to be done before vdev_load() so that we don't
+ * inadvertently do repair I/Os to the wrong device.
+ *
+ * This function will only return failure if one of the vdevs indicates that it
+ * has since been destroyed or exported. This is only possible if
+ * /etc/zfs/zpool.cache was readonly at the time. Otherwise, the vdev state
+ * will be updated but the function will return 0.
+ */
+int
+vdev_validate(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+ int c;
+ nvlist_t *label;
+ uint64_t guid, top_guid;
+ uint64_t state;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ if (vdev_validate(vd->vdev_child[c]) != 0)
+ return (EBADF);
+
+ /*
+ * If the device has already failed, or was marked offline, don't do
+ * any further validation. Otherwise, label I/O will fail and we will
+ * overwrite the previous state.
+ */
+ if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) {
+
+ if ((label = vdev_label_read_config(vd)) == NULL) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_BAD_LABEL);
+ return (0);
+ }
+
+ if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID,
+ &guid) != 0 || guid != spa_guid(spa)) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ nvlist_free(label);
+ return (0);
+ }
+
+ /*
+ * If this vdev just became a top-level vdev because its
+ * sibling was detached, it will have adopted the parent's
+ * vdev guid -- but the label may or may not be on disk yet.
+ * Fortunately, either version of the label will have the
+ * same top guid, so if we're a top-level vdev, we can
+ * safely compare to that instead.
+ */
+ if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID,
+ &guid) != 0 ||
+ nvlist_lookup_uint64(label, ZPOOL_CONFIG_TOP_GUID,
+ &top_guid) != 0 ||
+ (vd->vdev_guid != guid &&
+ (vd->vdev_guid != top_guid || vd != vd->vdev_top))) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ nvlist_free(label);
+ return (0);
+ }
+
+ if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
+ &state) != 0) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ nvlist_free(label);
+ return (0);
+ }
+
+ nvlist_free(label);
+
+ if (spa->spa_load_state == SPA_LOAD_OPEN &&
+ state != POOL_STATE_ACTIVE)
+ return (EBADF);
+
+ /*
+ * If we were able to open and validate a vdev that was
+ * previously marked permanently unavailable, clear that state
+ * now.
+ */
+ if (vd->vdev_not_present)
+ vd->vdev_not_present = 0;
+ }
+
+ return (0);
+}
+
+/*
+ * Close a virtual device.
+ */
+void
+vdev_close(vdev_t *vd)
+{
+ vd->vdev_ops->vdev_op_close(vd);
+
+ vdev_cache_purge(vd);
+
+ /*
+ * We record the previous state before we close it, so that if we are
+ * doing a reopen(), we don't generate FMA ereports if we notice that
+ * it's still faulted.
+ */
+ vd->vdev_prevstate = vd->vdev_state;
+
+ if (vd->vdev_offline)
+ vd->vdev_state = VDEV_STATE_OFFLINE;
+ else
+ vd->vdev_state = VDEV_STATE_CLOSED;
+ vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
+}
+
+void
+vdev_reopen(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+
+ ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
+
+ vdev_close(vd);
+ (void) vdev_open(vd);
+
+ /*
+ * Call vdev_validate() here to make sure we have the same device.
+ * Otherwise, a device with an invalid label could be successfully
+ * opened in response to vdev_reopen().
+ */
+ if (vd->vdev_aux) {
+ (void) vdev_validate_aux(vd);
+ if (vdev_readable(vd) && vdev_writeable(vd) &&
+ !l2arc_vdev_present(vd)) {
+ uint64_t size = vdev_get_rsize(vd);
+ l2arc_add_vdev(spa, vd,
+ VDEV_LABEL_START_SIZE,
+ size - VDEV_LABEL_START_SIZE);
+ }
+ } else {
+ (void) vdev_validate(vd);
+ }
+
+ /*
+ * Reassess parent vdev's health.
+ */
+ vdev_propagate_state(vd);
+}
+
+int
+vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
+{
+ int error;
+
+ /*
+ * Normally, partial opens (e.g. of a mirror) are allowed.
+ * For a create, however, we want to fail the request if
+ * there are any components we can't open.
+ */
+ error = vdev_open(vd);
+
+ if (error || vd->vdev_state != VDEV_STATE_HEALTHY) {
+ vdev_close(vd);
+ return (error ? error : ENXIO);
+ }
+
+ /*
+ * Recursively initialize all labels.
+ */
+ if ((error = vdev_label_init(vd, txg, isreplacing ?
+ VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
+ vdev_close(vd);
+ return (error);
+ }
+
+ return (0);
+}
+
+/*
+ * The is the latter half of vdev_create(). It is distinct because it
+ * involves initiating transactions in order to do metaslab creation.
+ * For creation, we want to try to create all vdevs at once and then undo it
+ * if anything fails; this is much harder if we have pending transactions.
+ */
+void
+vdev_init(vdev_t *vd, uint64_t txg)
+{
+ /*
+ * Aim for roughly 200 metaslabs per vdev.
+ */
+ vd->vdev_ms_shift = highbit(vd->vdev_asize / 200);
+ vd->vdev_ms_shift = MAX(vd->vdev_ms_shift, SPA_MAXBLOCKSHIFT);
+
+ /*
+ * Initialize the vdev's metaslabs. This can't fail because
+ * there's nothing to read when creating all new metaslabs.
+ */
+ VERIFY(vdev_metaslab_init(vd, txg) == 0);
+}
+
+void
+vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
+{
+ ASSERT(vd == vd->vdev_top);
+ ASSERT(ISP2(flags));
+
+ if (flags & VDD_METASLAB)
+ (void) txg_list_add(&vd->vdev_ms_list, arg, txg);
+
+ if (flags & VDD_DTL)
+ (void) txg_list_add(&vd->vdev_dtl_list, arg, txg);
+
+ (void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
+}
+
+void
+vdev_dtl_dirty(space_map_t *sm, uint64_t txg, uint64_t size)
+{
+ mutex_enter(sm->sm_lock);
+ if (!space_map_contains(sm, txg, size))
+ space_map_add(sm, txg, size);
+ mutex_exit(sm->sm_lock);
+}
+
+int
+vdev_dtl_contains(space_map_t *sm, uint64_t txg, uint64_t size)
+{
+ int dirty;
+
+ /*
+ * Quick test without the lock -- covers the common case that
+ * there are no dirty time segments.
+ */
+ if (sm->sm_space == 0)
+ return (0);
+
+ mutex_enter(sm->sm_lock);
+ dirty = space_map_contains(sm, txg, size);
+ mutex_exit(sm->sm_lock);
+
+ return (dirty);
+}
+
+/*
+ * Reassess DTLs after a config change or scrub completion.
+ */
+void
+vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done)
+{
+ spa_t *spa = vd->vdev_spa;
+ int c;
+
+ ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
+
+ if (vd->vdev_children == 0) {
+ mutex_enter(&vd->vdev_dtl_lock);
+ if (scrub_txg != 0 &&
+ (spa->spa_scrub_started || spa->spa_scrub_errors == 0)) {
+ /* XXX should check scrub_done? */
+ /*
+ * We completed a scrub up to scrub_txg. If we
+ * did it without rebooting, then the scrub dtl
+ * will be valid, so excise the old region and
+ * fold in the scrub dtl. Otherwise, leave the
+ * dtl as-is if there was an error.
+ */
+ space_map_excise(&vd->vdev_dtl_map, 0, scrub_txg);
+ space_map_union(&vd->vdev_dtl_map, &vd->vdev_dtl_scrub);
+ }
+ if (scrub_done)
+ space_map_vacate(&vd->vdev_dtl_scrub, NULL, NULL);
+ mutex_exit(&vd->vdev_dtl_lock);
+
+ if (txg != 0)
+ vdev_dirty(vd->vdev_top, VDD_DTL, vd, txg);
+ return;
+ }
+
+ /*
+ * Make sure the DTLs are always correct under the scrub lock.
+ */
+ if (vd == spa->spa_root_vdev)
+ mutex_enter(&spa->spa_scrub_lock);
+
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_vacate(&vd->vdev_dtl_map, NULL, NULL);
+ space_map_vacate(&vd->vdev_dtl_scrub, NULL, NULL);
+ mutex_exit(&vd->vdev_dtl_lock);
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ vdev_t *cvd = vd->vdev_child[c];
+ vdev_dtl_reassess(cvd, txg, scrub_txg, scrub_done);
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_union(&vd->vdev_dtl_map, &cvd->vdev_dtl_map);
+ space_map_union(&vd->vdev_dtl_scrub, &cvd->vdev_dtl_scrub);
+ mutex_exit(&vd->vdev_dtl_lock);
+ }
+
+ if (vd == spa->spa_root_vdev)
+ mutex_exit(&spa->spa_scrub_lock);
+}
+
+static int
+vdev_dtl_load(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+ space_map_obj_t *smo = &vd->vdev_dtl;
+ objset_t *mos = spa->spa_meta_objset;
+ dmu_buf_t *db;
+ int error;
+
+ ASSERT(vd->vdev_children == 0);
+
+ if (smo->smo_object == 0)
+ return (0);
+
+ if ((error = dmu_bonus_hold(mos, smo->smo_object, FTAG, &db)) != 0)
+ return (error);
+
+ ASSERT3U(db->db_size, >=, sizeof (*smo));
+ bcopy(db->db_data, smo, sizeof (*smo));
+ dmu_buf_rele(db, FTAG);
+
+ mutex_enter(&vd->vdev_dtl_lock);
+ error = space_map_load(&vd->vdev_dtl_map, NULL, SM_ALLOC, smo, mos);
+ mutex_exit(&vd->vdev_dtl_lock);
+
+ return (error);
+}
+
+void
+vdev_dtl_sync(vdev_t *vd, uint64_t txg)
+{
+ spa_t *spa = vd->vdev_spa;
+ space_map_obj_t *smo = &vd->vdev_dtl;
+ space_map_t *sm = &vd->vdev_dtl_map;
+ objset_t *mos = spa->spa_meta_objset;
+ space_map_t smsync;
+ kmutex_t smlock;
+ dmu_buf_t *db;
+ dmu_tx_t *tx;
+
+ tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+
+ if (vd->vdev_detached) {
+ if (smo->smo_object != 0) {
+ int err = dmu_object_free(mos, smo->smo_object, tx);
+ ASSERT3U(err, ==, 0);
+ smo->smo_object = 0;
+ }
+ dmu_tx_commit(tx);
+ return;
+ }
+
+ if (smo->smo_object == 0) {
+ ASSERT(smo->smo_objsize == 0);
+ ASSERT(smo->smo_alloc == 0);
+ smo->smo_object = dmu_object_alloc(mos,
+ DMU_OT_SPACE_MAP, 1 << SPACE_MAP_BLOCKSHIFT,
+ DMU_OT_SPACE_MAP_HEADER, sizeof (*smo), tx);
+ ASSERT(smo->smo_object != 0);
+ vdev_config_dirty(vd->vdev_top);
+ }
+
+ mutex_init(&smlock, NULL, MUTEX_DEFAULT, NULL);
+
+ space_map_create(&smsync, sm->sm_start, sm->sm_size, sm->sm_shift,
+ &smlock);
+
+ mutex_enter(&smlock);
+
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_walk(sm, space_map_add, &smsync);
+ mutex_exit(&vd->vdev_dtl_lock);
+
+ space_map_truncate(smo, mos, tx);
+ space_map_sync(&smsync, SM_ALLOC, smo, mos, tx);
+
+ space_map_destroy(&smsync);
+
+ mutex_exit(&smlock);
+ mutex_destroy(&smlock);
+
+ VERIFY(0 == dmu_bonus_hold(mos, smo->smo_object, FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+ ASSERT3U(db->db_size, >=, sizeof (*smo));
+ bcopy(smo, db->db_data, sizeof (*smo));
+ dmu_buf_rele(db, FTAG);
+
+ dmu_tx_commit(tx);
+}
+
+/*
+ * Determine if resilver is needed, and if so the txg range.
+ */
+boolean_t
+vdev_resilver_needed(vdev_t *vd, uint64_t *minp, uint64_t *maxp)
+{
+ boolean_t needed = B_FALSE;
+ uint64_t thismin = UINT64_MAX;
+ uint64_t thismax = 0;
+
+ if (vd->vdev_children == 0) {
+ mutex_enter(&vd->vdev_dtl_lock);
+ if (vd->vdev_dtl_map.sm_space != 0 && vdev_writeable(vd)) {
+ space_seg_t *ss;
+
+ ss = avl_first(&vd->vdev_dtl_map.sm_root);
+ thismin = ss->ss_start - 1;
+ ss = avl_last(&vd->vdev_dtl_map.sm_root);
+ thismax = ss->ss_end;
+ needed = B_TRUE;
+ }
+ mutex_exit(&vd->vdev_dtl_lock);
+ } else {
+ int c;
+ for (c = 0; c < vd->vdev_children; c++) {
+ vdev_t *cvd = vd->vdev_child[c];
+ uint64_t cmin, cmax;
+
+ if (vdev_resilver_needed(cvd, &cmin, &cmax)) {
+ thismin = MIN(thismin, cmin);
+ thismax = MAX(thismax, cmax);
+ needed = B_TRUE;
+ }
+ }
+ }
+
+ if (needed && minp) {
+ *minp = thismin;
+ *maxp = thismax;
+ }
+ return (needed);
+}
+
+void
+vdev_load(vdev_t *vd)
+{
+ int c;
+
+ /*
+ * Recursively load all children.
+ */
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_load(vd->vdev_child[c]);
+
+ /*
+ * If this is a top-level vdev, initialize its metaslabs.
+ */
+ if (vd == vd->vdev_top &&
+ (vd->vdev_ashift == 0 || vd->vdev_asize == 0 ||
+ vdev_metaslab_init(vd, 0) != 0))
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+
+ /*
+ * If this is a leaf vdev, load its DTL.
+ */
+ if (vd->vdev_ops->vdev_op_leaf && vdev_dtl_load(vd) != 0)
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+}
+
+/*
+ * The special vdev case is used for hot spares and l2cache devices. Its
+ * sole purpose it to set the vdev state for the associated vdev. To do this,
+ * we make sure that we can open the underlying device, then try to read the
+ * label, and make sure that the label is sane and that it hasn't been
+ * repurposed to another pool.
+ */
+int
+vdev_validate_aux(vdev_t *vd)
+{
+ nvlist_t *label;
+ uint64_t guid, version;
+ uint64_t state;
+
+ if (!vdev_readable(vd))
+ return (0);
+
+ if ((label = vdev_label_read_config(vd)) == NULL) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ return (-1);
+ }
+
+ if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_VERSION, &version) != 0 ||
+ version > SPA_VERSION ||
+ nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0 ||
+ guid != vd->vdev_guid ||
+ nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, &state) != 0) {
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ nvlist_free(label);
+ return (-1);
+ }
+
+ /*
+ * We don't actually check the pool state here. If it's in fact in
+ * use by another pool, we update this fact on the fly when requested.
+ */
+ nvlist_free(label);
+ return (0);
+}
+
+void
+vdev_sync_done(vdev_t *vd, uint64_t txg)
+{
+ metaslab_t *msp;
+
+ while (msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg)))
+ metaslab_sync_done(msp, txg);
+}
+
+void
+vdev_sync(vdev_t *vd, uint64_t txg)
+{
+ spa_t *spa = vd->vdev_spa;
+ vdev_t *lvd;
+ metaslab_t *msp;
+ dmu_tx_t *tx;
+
+ if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0) {
+ ASSERT(vd == vd->vdev_top);
+ tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+ vd->vdev_ms_array = dmu_object_alloc(spa->spa_meta_objset,
+ DMU_OT_OBJECT_ARRAY, 0, DMU_OT_NONE, 0, tx);
+ ASSERT(vd->vdev_ms_array != 0);
+ vdev_config_dirty(vd);
+ dmu_tx_commit(tx);
+ }
+
+ while ((msp = txg_list_remove(&vd->vdev_ms_list, txg)) != NULL) {
+ metaslab_sync(msp, txg);
+ (void) txg_list_add(&vd->vdev_ms_list, msp, TXG_CLEAN(txg));
+ }
+
+ while ((lvd = txg_list_remove(&vd->vdev_dtl_list, txg)) != NULL)
+ vdev_dtl_sync(lvd, txg);
+
+ (void) txg_list_add(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg));
+}
+
+uint64_t
+vdev_psize_to_asize(vdev_t *vd, uint64_t psize)
+{
+ return (vd->vdev_ops->vdev_op_asize(vd, psize));
+}
+
+/*
+ * Mark the given vdev faulted. A faulted vdev behaves as if the device could
+ * not be opened, and no I/O is attempted.
+ */
+int
+vdev_fault(spa_t *spa, uint64_t guid)
+{
+ vdev_t *vd;
+
+ spa_vdev_state_enter(spa);
+
+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
+ return (spa_vdev_state_exit(spa, NULL, ENODEV));
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
+
+ /*
+ * Faulted state takes precedence over degraded.
+ */
+ vd->vdev_faulted = 1ULL;
+ vd->vdev_degraded = 0ULL;
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED, VDEV_AUX_ERR_EXCEEDED);
+
+ /*
+ * If marking the vdev as faulted cause the top-level vdev to become
+ * unavailable, then back off and simply mark the vdev as degraded
+ * instead.
+ */
+ if (vdev_is_dead(vd->vdev_top) && vd->vdev_aux == NULL) {
+ vd->vdev_degraded = 1ULL;
+ vd->vdev_faulted = 0ULL;
+
+ /*
+ * If we reopen the device and it's not dead, only then do we
+ * mark it degraded.
+ */
+ vdev_reopen(vd);
+
+ if (vdev_readable(vd)) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED,
+ VDEV_AUX_ERR_EXCEEDED);
+ }
+ }
+
+ return (spa_vdev_state_exit(spa, vd, 0));
+}
+
+/*
+ * Mark the given vdev degraded. A degraded vdev is purely an indication to the
+ * user that something is wrong. The vdev continues to operate as normal as far
+ * as I/O is concerned.
+ */
+int
+vdev_degrade(spa_t *spa, uint64_t guid)
+{
+ vdev_t *vd;
+
+ spa_vdev_state_enter(spa);
+
+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
+ return (spa_vdev_state_exit(spa, NULL, ENODEV));
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
+
+ /*
+ * If the vdev is already faulted, then don't do anything.
+ */
+ if (vd->vdev_faulted || vd->vdev_degraded)
+ return (spa_vdev_state_exit(spa, NULL, 0));
+
+ vd->vdev_degraded = 1ULL;
+ if (!vdev_is_dead(vd))
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED,
+ VDEV_AUX_ERR_EXCEEDED);
+
+ return (spa_vdev_state_exit(spa, vd, 0));
+}
+
+/*
+ * Online the given vdev. If 'unspare' is set, it implies two things. First,
+ * any attached spare device should be detached when the device finishes
+ * resilvering. Second, the online should be treated like a 'test' online case,
+ * so no FMA events are generated if the device fails to open.
+ */
+int
+vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
+{
+ vdev_t *vd;
+
+ spa_vdev_state_enter(spa);
+
+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
+ return (spa_vdev_state_exit(spa, NULL, ENODEV));
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
+
+ vd->vdev_offline = B_FALSE;
+ vd->vdev_tmpoffline = B_FALSE;
+ vd->vdev_checkremove = !!(flags & ZFS_ONLINE_CHECKREMOVE);
+ vd->vdev_forcefault = !!(flags & ZFS_ONLINE_FORCEFAULT);
+ vdev_reopen(vd->vdev_top);
+ vd->vdev_checkremove = vd->vdev_forcefault = B_FALSE;
+
+ if (newstate)
+ *newstate = vd->vdev_state;
+ if ((flags & ZFS_ONLINE_UNSPARE) &&
+ !vdev_is_dead(vd) && vd->vdev_parent &&
+ vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
+ vd->vdev_parent->vdev_child[0] == vd)
+ vd->vdev_unspare = B_TRUE;
+
+ (void) spa_vdev_state_exit(spa, vd, 0);
+
+ VERIFY3U(spa_scrub(spa, POOL_SCRUB_RESILVER), ==, 0);
+
+ return (0);
+}
+
+int
+vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
+{
+ vdev_t *vd;
+
+ spa_vdev_state_enter(spa);
+
+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
+ return (spa_vdev_state_exit(spa, NULL, ENODEV));
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
+
+ /*
+ * If the device isn't already offline, try to offline it.
+ */
+ if (!vd->vdev_offline) {
+ /*
+ * If this device's top-level vdev has a non-empty DTL,
+ * don't allow the device to be offlined.
+ *
+ * XXX -- make this more precise by allowing the offline
+ * as long as the remaining devices don't have any DTL holes.
+ */
+ if (vd->vdev_top->vdev_dtl_map.sm_space != 0)
+ return (spa_vdev_state_exit(spa, NULL, EBUSY));
+
+ /*
+ * Offline this device and reopen its top-level vdev.
+ * If this action results in the top-level vdev becoming
+ * unusable, undo it and fail the request.
+ */
+ vd->vdev_offline = B_TRUE;
+ vdev_reopen(vd->vdev_top);
+ if (vdev_is_dead(vd->vdev_top) && vd->vdev_aux == NULL) {
+ vd->vdev_offline = B_FALSE;
+ vdev_reopen(vd->vdev_top);
+ return (spa_vdev_state_exit(spa, NULL, EBUSY));
+ }
+ }
+
+ vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);
+
+ return (spa_vdev_state_exit(spa, vd, 0));
+}
+
+/*
+ * Clear the error counts associated with this vdev. Unlike vdev_online() and
+ * vdev_offline(), we assume the spa config is locked. We also clear all
+ * children. If 'vd' is NULL, then the user wants to clear all vdevs.
+ */
+void
+vdev_clear(spa_t *spa, vdev_t *vd)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+
+ ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
+
+ if (vd == NULL)
+ vd = rvd;
+
+ vd->vdev_stat.vs_read_errors = 0;
+ vd->vdev_stat.vs_write_errors = 0;
+ vd->vdev_stat.vs_checksum_errors = 0;
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ vdev_clear(spa, vd->vdev_child[c]);
+
+ /*
+ * If we're in the FAULTED state or have experienced failed I/O, then
+ * clear the persistent state and attempt to reopen the device. We
+ * also mark the vdev config dirty, so that the new faulted state is
+ * written out to disk.
+ */
+ if (vd->vdev_faulted || vd->vdev_degraded ||
+ !vdev_readable(vd) || !vdev_writeable(vd)) {
+
+ vd->vdev_faulted = vd->vdev_degraded = 0;
+ vd->vdev_cant_read = B_FALSE;
+ vd->vdev_cant_write = B_FALSE;
+
+ vdev_reopen(vd);
+
+ if (vd != rvd)
+ vdev_state_dirty(vd->vdev_top);
+
+ if (vd->vdev_aux == NULL && !vdev_is_dead(vd))
+ spa_async_request(spa, SPA_ASYNC_RESILVER);
+
+ spa_event_notify(spa, vd, ESC_ZFS_VDEV_CLEAR);
+ }
+}
+
+boolean_t
+vdev_is_dead(vdev_t *vd)
+{
+ return (vd->vdev_state < VDEV_STATE_DEGRADED);
+}
+
+boolean_t
+vdev_readable(vdev_t *vd)
+{
+ return (!vdev_is_dead(vd) && !vd->vdev_cant_read);
+}
+
+boolean_t
+vdev_writeable(vdev_t *vd)
+{
+ return (!vdev_is_dead(vd) && !vd->vdev_cant_write);
+}
+
+boolean_t
+vdev_allocatable(vdev_t *vd)
+{
+ /*
+ * We currently allow allocations from vdevs which maybe in the
+ * process of reopening (i.e. VDEV_STATE_CLOSED). If the device
+ * fails to reopen then we'll catch it later when we're holding
+ * the proper locks.
+ */
+ return (!(vdev_is_dead(vd) && vd->vdev_state != VDEV_STATE_CLOSED) &&
+ !vd->vdev_cant_write);
+}
+
+boolean_t
+vdev_accessible(vdev_t *vd, zio_t *zio)
+{
+ ASSERT(zio->io_vd == vd);
+
+ if (vdev_is_dead(vd) || vd->vdev_remove_wanted)
+ return (B_FALSE);
+
+ if (zio->io_type == ZIO_TYPE_READ)
+ return (!vd->vdev_cant_read);
+
+ if (zio->io_type == ZIO_TYPE_WRITE)
+ return (!vd->vdev_cant_write);
+
+ return (B_TRUE);
+}
+
+/*
+ * Get statistics for the given vdev.
+ */
+void
+vdev_get_stats(vdev_t *vd, vdev_stat_t *vs)
+{
+ vdev_t *rvd = vd->vdev_spa->spa_root_vdev;
+
+ mutex_enter(&vd->vdev_stat_lock);
+ bcopy(&vd->vdev_stat, vs, sizeof (*vs));
+ vs->vs_scrub_errors = vd->vdev_spa->spa_scrub_errors;
+ vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
+ vs->vs_state = vd->vdev_state;
+ vs->vs_rsize = vdev_get_rsize(vd);
+ mutex_exit(&vd->vdev_stat_lock);
+
+ /*
+ * If we're getting stats on the root vdev, aggregate the I/O counts
+ * over all top-level vdevs (i.e. the direct children of the root).
+ */
+ if (vd == rvd) {
+ for (int c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *cvd = rvd->vdev_child[c];
+ vdev_stat_t *cvs = &cvd->vdev_stat;
+
+ mutex_enter(&vd->vdev_stat_lock);
+ for (int t = 0; t < ZIO_TYPES; t++) {
+ vs->vs_ops[t] += cvs->vs_ops[t];
+ vs->vs_bytes[t] += cvs->vs_bytes[t];
+ }
+ vs->vs_scrub_examined += cvs->vs_scrub_examined;
+ mutex_exit(&vd->vdev_stat_lock);
+ }
+ }
+}
+
+void
+vdev_clear_stats(vdev_t *vd)
+{
+ mutex_enter(&vd->vdev_stat_lock);
+ vd->vdev_stat.vs_space = 0;
+ vd->vdev_stat.vs_dspace = 0;
+ vd->vdev_stat.vs_alloc = 0;
+ mutex_exit(&vd->vdev_stat_lock);
+}
+
+void
+vdev_stat_update(zio_t *zio, uint64_t psize)
+{
+ vdev_t *rvd = zio->io_spa->spa_root_vdev;
+ vdev_t *vd = zio->io_vd ? zio->io_vd : rvd;
+ vdev_t *pvd;
+ uint64_t txg = zio->io_txg;
+ vdev_stat_t *vs = &vd->vdev_stat;
+ zio_type_t type = zio->io_type;
+ int flags = zio->io_flags;
+
+ /*
+ * If this i/o is a gang leader, it didn't do any actual work.
+ */
+ if (zio->io_gang_tree)
+ return;
+
+ if (zio->io_error == 0) {
+ /*
+ * If this is a root i/o, don't count it -- we've already
+ * counted the top-level vdevs, and vdev_get_stats() will
+ * aggregate them when asked. This reduces contention on
+ * the root vdev_stat_lock and implicitly handles blocks
+ * that compress away to holes, for which there is no i/o.
+ * (Holes never create vdev children, so all the counters
+ * remain zero, which is what we want.)
+ *
+ * Note: this only applies to successful i/o (io_error == 0)
+ * because unlike i/o counts, errors are not additive.
+ * When reading a ditto block, for example, failure of
+ * one top-level vdev does not imply a root-level error.
+ */
+ if (vd == rvd)
+ return;
+
+ ASSERT(vd == zio->io_vd);
+ if (!(flags & ZIO_FLAG_IO_BYPASS)) {
+ mutex_enter(&vd->vdev_stat_lock);
+ vs->vs_ops[type]++;
+ vs->vs_bytes[type] += psize;
+ mutex_exit(&vd->vdev_stat_lock);
+ }
+ if (flags & ZIO_FLAG_IO_REPAIR) {
+ ASSERT(zio->io_delegate_list == NULL);
+ mutex_enter(&vd->vdev_stat_lock);
+ if (flags & ZIO_FLAG_SCRUB_THREAD)
+ vs->vs_scrub_repaired += psize;
+ else
+ vs->vs_self_healed += psize;
+ mutex_exit(&vd->vdev_stat_lock);
+ }
+ return;
+ }
+
+ if (flags & ZIO_FLAG_SPECULATIVE)
+ return;
+
+ mutex_enter(&vd->vdev_stat_lock);
+ if (type == ZIO_TYPE_READ) {
+ if (zio->io_error == ECKSUM)
+ vs->vs_checksum_errors++;
+ else
+ vs->vs_read_errors++;
+ }
+ if (type == ZIO_TYPE_WRITE)
+ vs->vs_write_errors++;
+ mutex_exit(&vd->vdev_stat_lock);
+
+ if (type == ZIO_TYPE_WRITE && txg != 0 && vd->vdev_children == 0) {
+ if (flags & ZIO_FLAG_SCRUB_THREAD) {
+ ASSERT(flags & ZIO_FLAG_IO_REPAIR);
+ for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
+ vdev_dtl_dirty(&pvd->vdev_dtl_scrub, txg, 1);
+ }
+ if (!(flags & ZIO_FLAG_IO_REPAIR)) {
+ if (vdev_dtl_contains(&vd->vdev_dtl_map, txg, 1))
+ return;
+ vdev_dirty(vd->vdev_top, VDD_DTL, vd, txg);
+ for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
+ vdev_dtl_dirty(&pvd->vdev_dtl_map, txg, 1);
+ }
+ }
+}
+
+void
+vdev_scrub_stat_update(vdev_t *vd, pool_scrub_type_t type, boolean_t complete)
+{
+ int c;
+ vdev_stat_t *vs = &vd->vdev_stat;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_scrub_stat_update(vd->vdev_child[c], type, complete);
+
+ mutex_enter(&vd->vdev_stat_lock);
+
+ if (type == POOL_SCRUB_NONE) {
+ /*
+ * Update completion and end time. Leave everything else alone
+ * so we can report what happened during the previous scrub.
+ */
+ vs->vs_scrub_complete = complete;
+ vs->vs_scrub_end = gethrestime_sec();
+ } else {
+ vs->vs_scrub_type = type;
+ vs->vs_scrub_complete = 0;
+ vs->vs_scrub_examined = 0;
+ vs->vs_scrub_repaired = 0;
+ vs->vs_scrub_start = gethrestime_sec();
+ vs->vs_scrub_end = 0;
+ }
+
+ mutex_exit(&vd->vdev_stat_lock);
+}
+
+/*
+ * Update the in-core space usage stats for this vdev and the root vdev.
+ */
+void
+vdev_space_update(vdev_t *vd, int64_t space_delta, int64_t alloc_delta,
+ boolean_t update_root)
+{
+ int64_t dspace_delta = space_delta;
+ spa_t *spa = vd->vdev_spa;
+ vdev_t *rvd = spa->spa_root_vdev;
+
+ ASSERT(vd == vd->vdev_top);
+
+ /*
+ * Apply the inverse of the psize-to-asize (ie. RAID-Z) space-expansion
+ * factor. We must calculate this here and not at the root vdev
+ * because the root vdev's psize-to-asize is simply the max of its
+ * childrens', thus not accurate enough for us.
+ */
+ ASSERT((dspace_delta & (SPA_MINBLOCKSIZE-1)) == 0);
+ dspace_delta = (dspace_delta >> SPA_MINBLOCKSHIFT) *
+ vd->vdev_deflate_ratio;
+
+ mutex_enter(&vd->vdev_stat_lock);
+ vd->vdev_stat.vs_space += space_delta;
+ vd->vdev_stat.vs_alloc += alloc_delta;
+ vd->vdev_stat.vs_dspace += dspace_delta;
+ mutex_exit(&vd->vdev_stat_lock);
+
+ if (update_root) {
+ ASSERT(rvd == vd->vdev_parent);
+ ASSERT(vd->vdev_ms_count != 0);
+
+ /*
+ * Don't count non-normal (e.g. intent log) space as part of
+ * the pool's capacity.
+ */
+ if (vd->vdev_mg->mg_class != spa->spa_normal_class)
+ return;
+
+ mutex_enter(&rvd->vdev_stat_lock);
+ rvd->vdev_stat.vs_space += space_delta;
+ rvd->vdev_stat.vs_alloc += alloc_delta;
+ rvd->vdev_stat.vs_dspace += dspace_delta;
+ mutex_exit(&rvd->vdev_stat_lock);
+ }
+}
+
+/*
+ * Mark a top-level vdev's config as dirty, placing it on the dirty list
+ * so that it will be written out next time the vdev configuration is synced.
+ * If the root vdev is specified (vdev_top == NULL), dirty all top-level vdevs.
+ */
+void
+vdev_config_dirty(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+ vdev_t *rvd = spa->spa_root_vdev;
+ int c;
+
+ /*
+ * If this is an aux vdev (as with l2cache devices), then we update the
+ * vdev config manually and set the sync flag.
+ */
+ if (vd->vdev_aux != NULL) {
+ spa_aux_vdev_t *sav = vd->vdev_aux;
+ nvlist_t **aux;
+ uint_t naux;
+
+ for (c = 0; c < sav->sav_count; c++) {
+ if (sav->sav_vdevs[c] == vd)
+ break;
+ }
+
+ if (c == sav->sav_count) {
+ /*
+ * We're being removed. There's nothing more to do.
+ */
+ ASSERT(sav->sav_sync == B_TRUE);
+ return;
+ }
+
+ sav->sav_sync = B_TRUE;
+
+ VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
+ ZPOOL_CONFIG_L2CACHE, &aux, &naux) == 0);
+
+ ASSERT(c < naux);
+
+ /*
+ * Setting the nvlist in the middle if the array is a little
+ * sketchy, but it will work.
+ */
+ nvlist_free(aux[c]);
+ aux[c] = vdev_config_generate(spa, vd, B_TRUE, B_FALSE, B_TRUE);
+
+ return;
+ }
+
+ /*
+ * The dirty list is protected by the SCL_CONFIG lock. The caller
+ * must either hold SCL_CONFIG as writer, or must be the sync thread
+ * (which holds SCL_CONFIG as reader). There's only one sync thread,
+ * so this is sufficient to ensure mutual exclusion.
+ */
+ ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) ||
+ (dsl_pool_sync_context(spa_get_dsl(spa)) &&
+ spa_config_held(spa, SCL_CONFIG, RW_READER)));
+
+ if (vd == rvd) {
+ for (c = 0; c < rvd->vdev_children; c++)
+ vdev_config_dirty(rvd->vdev_child[c]);
+ } else {
+ ASSERT(vd == vd->vdev_top);
+
+ if (!list_link_active(&vd->vdev_config_dirty_node))
+ list_insert_head(&spa->spa_config_dirty_list, vd);
+ }
+}
+
+void
+vdev_config_clean(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+
+ ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) ||
+ (dsl_pool_sync_context(spa_get_dsl(spa)) &&
+ spa_config_held(spa, SCL_CONFIG, RW_READER)));
+
+ ASSERT(list_link_active(&vd->vdev_config_dirty_node));
+ list_remove(&spa->spa_config_dirty_list, vd);
+}
+
+/*
+ * Mark a top-level vdev's state as dirty, so that the next pass of
+ * spa_sync() can convert this into vdev_config_dirty(). We distinguish
+ * the state changes from larger config changes because they require
+ * much less locking, and are often needed for administrative actions.
+ */
+void
+vdev_state_dirty(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+
+ ASSERT(vd == vd->vdev_top);
+
+ /*
+ * The state list is protected by the SCL_STATE lock. The caller
+ * must either hold SCL_STATE as writer, or must be the sync thread
+ * (which holds SCL_STATE as reader). There's only one sync thread,
+ * so this is sufficient to ensure mutual exclusion.
+ */
+ ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) ||
+ (dsl_pool_sync_context(spa_get_dsl(spa)) &&
+ spa_config_held(spa, SCL_STATE, RW_READER)));
+
+ if (!list_link_active(&vd->vdev_state_dirty_node))
+ list_insert_head(&spa->spa_state_dirty_list, vd);
+}
+
+void
+vdev_state_clean(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+
+ ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) ||
+ (dsl_pool_sync_context(spa_get_dsl(spa)) &&
+ spa_config_held(spa, SCL_STATE, RW_READER)));
+
+ ASSERT(list_link_active(&vd->vdev_state_dirty_node));
+ list_remove(&spa->spa_state_dirty_list, vd);
+}
+
+/*
+ * Propagate vdev state up from children to parent.
+ */
+void
+vdev_propagate_state(vdev_t *vd)
+{
+ vdev_t *rvd = vd->vdev_spa->spa_root_vdev;
+ int degraded = 0, faulted = 0;
+ int corrupted = 0;
+ int c;
+ vdev_t *child;
+
+ if (vd->vdev_children > 0) {
+ for (c = 0; c < vd->vdev_children; c++) {
+ child = vd->vdev_child[c];
+
+ if (!vdev_readable(child) ||
+ (!vdev_writeable(child) && (spa_mode & FWRITE))) {
+ /*
+ * Root special: if there is a top-level log
+ * device, treat the root vdev as if it were
+ * degraded.
+ */
+ if (child->vdev_islog && vd == rvd)
+ degraded++;
+ else
+ faulted++;
+ } else if (child->vdev_state <= VDEV_STATE_DEGRADED) {
+ degraded++;
+ }
+
+ if (child->vdev_stat.vs_aux == VDEV_AUX_CORRUPT_DATA)
+ corrupted++;
+ }
+
+ vd->vdev_ops->vdev_op_state_change(vd, faulted, degraded);
+
+ /*
+ * Root special: if there is a top-level vdev that cannot be
+ * opened due to corrupted metadata, then propagate the root
+ * vdev's aux state as 'corrupt' rather than 'insufficient
+ * replicas'.
+ */
+ if (corrupted && vd == rvd &&
+ rvd->vdev_state == VDEV_STATE_CANT_OPEN)
+ vdev_set_state(rvd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_CORRUPT_DATA);
+ }
+
+ if (vd->vdev_parent)
+ vdev_propagate_state(vd->vdev_parent);
+}
+
+/*
+ * Set a vdev's state. If this is during an open, we don't update the parent
+ * state, because we're in the process of opening children depth-first.
+ * Otherwise, we propagate the change to the parent.
+ *
+ * If this routine places a device in a faulted state, an appropriate ereport is
+ * generated.
+ */
+void
+vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
+{
+ uint64_t save_state;
+ spa_t *spa = vd->vdev_spa;
+
+ if (state == vd->vdev_state) {
+ vd->vdev_stat.vs_aux = aux;
+ return;
+ }
+
+ save_state = vd->vdev_state;
+
+ vd->vdev_state = state;
+ vd->vdev_stat.vs_aux = aux;
+
+ /*
+ * If we are setting the vdev state to anything but an open state, then
+ * always close the underlying device. Otherwise, we keep accessible
+ * but invalid devices open forever. We don't call vdev_close() itself,
+ * because that implies some extra checks (offline, etc) that we don't
+ * want here. This is limited to leaf devices, because otherwise
+ * closing the device will affect other children.
+ */
+ if (vdev_is_dead(vd) && vd->vdev_ops->vdev_op_leaf)
+ vd->vdev_ops->vdev_op_close(vd);
+
+ if (vd->vdev_removed &&
+ state == VDEV_STATE_CANT_OPEN &&
+ (aux == VDEV_AUX_OPEN_FAILED || vd->vdev_checkremove)) {
+ /*
+ * If the previous state is set to VDEV_STATE_REMOVED, then this
+ * device was previously marked removed and someone attempted to
+ * reopen it. If this failed due to a nonexistent device, then
+ * keep the device in the REMOVED state. We also let this be if
+ * it is one of our special test online cases, which is only
+ * attempting to online the device and shouldn't generate an FMA
+ * fault.
+ */
+ vd->vdev_state = VDEV_STATE_REMOVED;
+ vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
+ } else if (state == VDEV_STATE_REMOVED) {
+ /*
+ * Indicate to the ZFS DE that this device has been removed, and
+ * any recent errors should be ignored.
+ */
+ zfs_post_remove(spa, vd);
+ vd->vdev_removed = B_TRUE;
+ } else if (state == VDEV_STATE_CANT_OPEN) {
+ /*
+ * If we fail to open a vdev during an import, we mark it as
+ * "not available", which signifies that it was never there to
+ * begin with. Failure to open such a device is not considered
+ * an error.
+ */
+ if (spa->spa_load_state == SPA_LOAD_IMPORT &&
+ !spa->spa_import_faulted &&
+ vd->vdev_ops->vdev_op_leaf)
+ vd->vdev_not_present = 1;
+
+ /*
+ * Post the appropriate ereport. If the 'prevstate' field is
+ * set to something other than VDEV_STATE_UNKNOWN, it indicates
+ * that this is part of a vdev_reopen(). In this case, we don't
+ * want to post the ereport if the device was already in the
+ * CANT_OPEN state beforehand.
+ *
+ * If the 'checkremove' flag is set, then this is an attempt to
+ * online the device in response to an insertion event. If we
+ * hit this case, then we have detected an insertion event for a
+ * faulted or offline device that wasn't in the removed state.
+ * In this scenario, we don't post an ereport because we are
+ * about to replace the device, or attempt an online with
+ * vdev_forcefault, which will generate the fault for us.
+ */
+ if ((vd->vdev_prevstate != state || vd->vdev_forcefault) &&
+ !vd->vdev_not_present && !vd->vdev_checkremove &&
+ vd != spa->spa_root_vdev) {
+ const char *class;
+
+ switch (aux) {
+ case VDEV_AUX_OPEN_FAILED:
+ class = FM_EREPORT_ZFS_DEVICE_OPEN_FAILED;
+ break;
+ case VDEV_AUX_CORRUPT_DATA:
+ class = FM_EREPORT_ZFS_DEVICE_CORRUPT_DATA;
+ break;
+ case VDEV_AUX_NO_REPLICAS:
+ class = FM_EREPORT_ZFS_DEVICE_NO_REPLICAS;
+ break;
+ case VDEV_AUX_BAD_GUID_SUM:
+ class = FM_EREPORT_ZFS_DEVICE_BAD_GUID_SUM;
+ break;
+ case VDEV_AUX_TOO_SMALL:
+ class = FM_EREPORT_ZFS_DEVICE_TOO_SMALL;
+ break;
+ case VDEV_AUX_BAD_LABEL:
+ class = FM_EREPORT_ZFS_DEVICE_BAD_LABEL;
+ break;
+ case VDEV_AUX_IO_FAILURE:
+ class = FM_EREPORT_ZFS_IO_FAILURE;
+ break;
+ default:
+ class = FM_EREPORT_ZFS_DEVICE_UNKNOWN;
+ }
+
+ zfs_ereport_post(class, spa, vd, NULL, save_state, 0);
+ }
+
+ /* Erase any notion of persistent removed state */
+ vd->vdev_removed = B_FALSE;
+ } else {
+ vd->vdev_removed = B_FALSE;
+ }
+
+ if (!isopen)
+ vdev_propagate_state(vd);
+}
+
+/*
+ * Check the vdev configuration to ensure that it's capable of supporting
+ * a root pool. Currently, we do not support RAID-Z or partial configuration.
+ * In addition, only a single top-level vdev is allowed and none of the leaves
+ * can be wholedisks.
+ */
+boolean_t
+vdev_is_bootable(vdev_t *vd)
+{
+ int c;
+
+ if (!vd->vdev_ops->vdev_op_leaf) {
+ char *vdev_type = vd->vdev_ops->vdev_op_type;
+
+ if (strcmp(vdev_type, VDEV_TYPE_ROOT) == 0 &&
+ vd->vdev_children > 1) {
+ return (B_FALSE);
+ } else if (strcmp(vdev_type, VDEV_TYPE_RAIDZ) == 0 ||
+ strcmp(vdev_type, VDEV_TYPE_MISSING) == 0) {
+ return (B_FALSE);
+ }
+ } else if (vd->vdev_wholedisk == 1) {
+ return (B_FALSE);
+ }
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ if (!vdev_is_bootable(vd->vdev_child[c]))
+ return (B_FALSE);
+ }
+ return (B_TRUE);
+}
diff --git a/module/zfs/vdev_cache.c b/module/zfs/vdev_cache.c
new file mode 100644
index 000000000..5a7b59f6e
--- /dev/null
+++ b/module/zfs/vdev_cache.c
@@ -0,0 +1,425 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/kstat.h>
+
+/*
+ * Virtual device read-ahead caching.
+ *
+ * This file implements a simple LRU read-ahead cache. When the DMU reads
+ * a given block, it will often want other, nearby blocks soon thereafter.
+ * We take advantage of this by reading a larger disk region and caching
+ * the result. In the best case, this can turn 128 back-to-back 512-byte
+ * reads into a single 64k read followed by 127 cache hits; this reduces
+ * latency dramatically. In the worst case, it can turn an isolated 512-byte
+ * read into a 64k read, which doesn't affect latency all that much but is
+ * terribly wasteful of bandwidth. A more intelligent version of the cache
+ * could keep track of access patterns and not do read-ahead unless it sees
+ * at least two temporally close I/Os to the same region. Currently, only
+ * metadata I/O is inflated. A futher enhancement could take advantage of
+ * more semantic information about the I/O. And it could use something
+ * faster than an AVL tree; that was chosen solely for convenience.
+ *
+ * There are five cache operations: allocate, fill, read, write, evict.
+ *
+ * (1) Allocate. This reserves a cache entry for the specified region.
+ * We separate the allocate and fill operations so that multiple threads
+ * don't generate I/O for the same cache miss.
+ *
+ * (2) Fill. When the I/O for a cache miss completes, the fill routine
+ * places the data in the previously allocated cache entry.
+ *
+ * (3) Read. Read data from the cache.
+ *
+ * (4) Write. Update cache contents after write completion.
+ *
+ * (5) Evict. When allocating a new entry, we evict the oldest (LRU) entry
+ * if the total cache size exceeds zfs_vdev_cache_size.
+ */
+
+/*
+ * These tunables are for performance analysis.
+ */
+/*
+ * All i/os smaller than zfs_vdev_cache_max will be turned into
+ * 1<<zfs_vdev_cache_bshift byte reads by the vdev_cache (aka software
+ * track buffer). At most zfs_vdev_cache_size bytes will be kept in each
+ * vdev's vdev_cache.
+ */
+int zfs_vdev_cache_max = 1<<14; /* 16KB */
+int zfs_vdev_cache_size = 10ULL << 20; /* 10MB */
+int zfs_vdev_cache_bshift = 16;
+
+#define VCBS (1 << zfs_vdev_cache_bshift) /* 64KB */
+
+kstat_t *vdc_ksp = NULL;
+
+typedef struct vdc_stats {
+ kstat_named_t vdc_stat_delegations;
+ kstat_named_t vdc_stat_hits;
+ kstat_named_t vdc_stat_misses;
+} vdc_stats_t;
+
+static vdc_stats_t vdc_stats = {
+ { "delegations", KSTAT_DATA_UINT64 },
+ { "hits", KSTAT_DATA_UINT64 },
+ { "misses", KSTAT_DATA_UINT64 }
+};
+
+#define VDCSTAT_BUMP(stat) atomic_add_64(&vdc_stats.stat.value.ui64, 1);
+
+static int
+vdev_cache_offset_compare(const void *a1, const void *a2)
+{
+ const vdev_cache_entry_t *ve1 = a1;
+ const vdev_cache_entry_t *ve2 = a2;
+
+ if (ve1->ve_offset < ve2->ve_offset)
+ return (-1);
+ if (ve1->ve_offset > ve2->ve_offset)
+ return (1);
+ return (0);
+}
+
+static int
+vdev_cache_lastused_compare(const void *a1, const void *a2)
+{
+ const vdev_cache_entry_t *ve1 = a1;
+ const vdev_cache_entry_t *ve2 = a2;
+
+ if (ve1->ve_lastused < ve2->ve_lastused)
+ return (-1);
+ if (ve1->ve_lastused > ve2->ve_lastused)
+ return (1);
+
+ /*
+ * Among equally old entries, sort by offset to ensure uniqueness.
+ */
+ return (vdev_cache_offset_compare(a1, a2));
+}
+
+/*
+ * Evict the specified entry from the cache.
+ */
+static void
+vdev_cache_evict(vdev_cache_t *vc, vdev_cache_entry_t *ve)
+{
+ ASSERT(MUTEX_HELD(&vc->vc_lock));
+ ASSERT(ve->ve_fill_io == NULL);
+ ASSERT(ve->ve_data != NULL);
+
+ avl_remove(&vc->vc_lastused_tree, ve);
+ avl_remove(&vc->vc_offset_tree, ve);
+ zio_buf_free(ve->ve_data, VCBS);
+ kmem_free(ve, sizeof (vdev_cache_entry_t));
+}
+
+/*
+ * Allocate an entry in the cache. At the point we don't have the data,
+ * we're just creating a placeholder so that multiple threads don't all
+ * go off and read the same blocks.
+ */
+static vdev_cache_entry_t *
+vdev_cache_allocate(zio_t *zio)
+{
+ vdev_cache_t *vc = &zio->io_vd->vdev_cache;
+ uint64_t offset = P2ALIGN(zio->io_offset, VCBS);
+ vdev_cache_entry_t *ve;
+
+ ASSERT(MUTEX_HELD(&vc->vc_lock));
+
+ if (zfs_vdev_cache_size == 0)
+ return (NULL);
+
+ /*
+ * If adding a new entry would exceed the cache size,
+ * evict the oldest entry (LRU).
+ */
+ if ((avl_numnodes(&vc->vc_lastused_tree) << zfs_vdev_cache_bshift) >
+ zfs_vdev_cache_size) {
+ ve = avl_first(&vc->vc_lastused_tree);
+ if (ve->ve_fill_io != NULL)
+ return (NULL);
+ ASSERT(ve->ve_hits != 0);
+ vdev_cache_evict(vc, ve);
+ }
+
+ ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_SLEEP);
+ ve->ve_offset = offset;
+ ve->ve_lastused = lbolt;
+ ve->ve_data = zio_buf_alloc(VCBS);
+
+ avl_add(&vc->vc_offset_tree, ve);
+ avl_add(&vc->vc_lastused_tree, ve);
+
+ return (ve);
+}
+
+static void
+vdev_cache_hit(vdev_cache_t *vc, vdev_cache_entry_t *ve, zio_t *zio)
+{
+ uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
+
+ ASSERT(MUTEX_HELD(&vc->vc_lock));
+ ASSERT(ve->ve_fill_io == NULL);
+
+ if (ve->ve_lastused != lbolt) {
+ avl_remove(&vc->vc_lastused_tree, ve);
+ ve->ve_lastused = lbolt;
+ avl_add(&vc->vc_lastused_tree, ve);
+ }
+
+ ve->ve_hits++;
+ bcopy(ve->ve_data + cache_phase, zio->io_data, zio->io_size);
+}
+
+/*
+ * Fill a previously allocated cache entry with data.
+ */
+static void
+vdev_cache_fill(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ vdev_cache_t *vc = &vd->vdev_cache;
+ vdev_cache_entry_t *ve = zio->io_private;
+ zio_t *dio;
+
+ ASSERT(zio->io_size == VCBS);
+
+ /*
+ * Add data to the cache.
+ */
+ mutex_enter(&vc->vc_lock);
+
+ ASSERT(ve->ve_fill_io == zio);
+ ASSERT(ve->ve_offset == zio->io_offset);
+ ASSERT(ve->ve_data == zio->io_data);
+
+ ve->ve_fill_io = NULL;
+
+ /*
+ * Even if this cache line was invalidated by a missed write update,
+ * any reads that were queued up before the missed update are still
+ * valid, so we can satisfy them from this line before we evict it.
+ */
+ for (dio = zio->io_delegate_list; dio; dio = dio->io_delegate_next)
+ vdev_cache_hit(vc, ve, dio);
+
+ if (zio->io_error || ve->ve_missed_update)
+ vdev_cache_evict(vc, ve);
+
+ mutex_exit(&vc->vc_lock);
+
+ while ((dio = zio->io_delegate_list) != NULL) {
+ zio->io_delegate_list = dio->io_delegate_next;
+ dio->io_delegate_next = NULL;
+ dio->io_error = zio->io_error;
+ zio_execute(dio);
+ }
+}
+
+/*
+ * Read data from the cache. Returns 0 on cache hit, errno on a miss.
+ */
+int
+vdev_cache_read(zio_t *zio)
+{
+ vdev_cache_t *vc = &zio->io_vd->vdev_cache;
+ vdev_cache_entry_t *ve, ve_search;
+ uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
+ uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
+ zio_t *fio;
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ);
+
+ if (zio->io_flags & ZIO_FLAG_DONT_CACHE)
+ return (EINVAL);
+
+ if (zio->io_size > zfs_vdev_cache_max)
+ return (EOVERFLOW);
+
+ /*
+ * If the I/O straddles two or more cache blocks, don't cache it.
+ */
+ if (P2BOUNDARY(zio->io_offset, zio->io_size, VCBS))
+ return (EXDEV);
+
+ ASSERT(cache_phase + zio->io_size <= VCBS);
+
+ mutex_enter(&vc->vc_lock);
+
+ ve_search.ve_offset = cache_offset;
+ ve = avl_find(&vc->vc_offset_tree, &ve_search, NULL);
+
+ if (ve != NULL) {
+ if (ve->ve_missed_update) {
+ mutex_exit(&vc->vc_lock);
+ return (ESTALE);
+ }
+
+ if ((fio = ve->ve_fill_io) != NULL) {
+ zio->io_delegate_next = fio->io_delegate_list;
+ fio->io_delegate_list = zio;
+ zio_vdev_io_bypass(zio);
+ mutex_exit(&vc->vc_lock);
+ VDCSTAT_BUMP(vdc_stat_delegations);
+ return (0);
+ }
+
+ vdev_cache_hit(vc, ve, zio);
+ zio_vdev_io_bypass(zio);
+
+ mutex_exit(&vc->vc_lock);
+ zio_execute(zio);
+ VDCSTAT_BUMP(vdc_stat_hits);
+ return (0);
+ }
+
+ ve = vdev_cache_allocate(zio);
+
+ if (ve == NULL) {
+ mutex_exit(&vc->vc_lock);
+ return (ENOMEM);
+ }
+
+ fio = zio_vdev_delegated_io(zio->io_vd, cache_offset,
+ ve->ve_data, VCBS, ZIO_TYPE_READ, ZIO_PRIORITY_CACHE_FILL,
+ ZIO_FLAG_DONT_CACHE, vdev_cache_fill, ve);
+
+ ve->ve_fill_io = fio;
+ fio->io_delegate_list = zio;
+ zio_vdev_io_bypass(zio);
+
+ mutex_exit(&vc->vc_lock);
+ zio_nowait(fio);
+ VDCSTAT_BUMP(vdc_stat_misses);
+
+ return (0);
+}
+
+/*
+ * Update cache contents upon write completion.
+ */
+void
+vdev_cache_write(zio_t *zio)
+{
+ vdev_cache_t *vc = &zio->io_vd->vdev_cache;
+ vdev_cache_entry_t *ve, ve_search;
+ uint64_t io_start = zio->io_offset;
+ uint64_t io_end = io_start + zio->io_size;
+ uint64_t min_offset = P2ALIGN(io_start, VCBS);
+ uint64_t max_offset = P2ROUNDUP(io_end, VCBS);
+ avl_index_t where;
+
+ ASSERT(zio->io_type == ZIO_TYPE_WRITE);
+
+ mutex_enter(&vc->vc_lock);
+
+ ve_search.ve_offset = min_offset;
+ ve = avl_find(&vc->vc_offset_tree, &ve_search, &where);
+
+ if (ve == NULL)
+ ve = avl_nearest(&vc->vc_offset_tree, where, AVL_AFTER);
+
+ while (ve != NULL && ve->ve_offset < max_offset) {
+ uint64_t start = MAX(ve->ve_offset, io_start);
+ uint64_t end = MIN(ve->ve_offset + VCBS, io_end);
+
+ if (ve->ve_fill_io != NULL) {
+ ve->ve_missed_update = 1;
+ } else {
+ bcopy((char *)zio->io_data + start - io_start,
+ ve->ve_data + start - ve->ve_offset, end - start);
+ }
+ ve = AVL_NEXT(&vc->vc_offset_tree, ve);
+ }
+ mutex_exit(&vc->vc_lock);
+}
+
+void
+vdev_cache_purge(vdev_t *vd)
+{
+ vdev_cache_t *vc = &vd->vdev_cache;
+ vdev_cache_entry_t *ve;
+
+ mutex_enter(&vc->vc_lock);
+ while ((ve = avl_first(&vc->vc_offset_tree)) != NULL)
+ vdev_cache_evict(vc, ve);
+ mutex_exit(&vc->vc_lock);
+}
+
+void
+vdev_cache_init(vdev_t *vd)
+{
+ vdev_cache_t *vc = &vd->vdev_cache;
+
+ mutex_init(&vc->vc_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ avl_create(&vc->vc_offset_tree, vdev_cache_offset_compare,
+ sizeof (vdev_cache_entry_t),
+ offsetof(struct vdev_cache_entry, ve_offset_node));
+
+ avl_create(&vc->vc_lastused_tree, vdev_cache_lastused_compare,
+ sizeof (vdev_cache_entry_t),
+ offsetof(struct vdev_cache_entry, ve_lastused_node));
+}
+
+void
+vdev_cache_fini(vdev_t *vd)
+{
+ vdev_cache_t *vc = &vd->vdev_cache;
+
+ vdev_cache_purge(vd);
+
+ avl_destroy(&vc->vc_offset_tree);
+ avl_destroy(&vc->vc_lastused_tree);
+
+ mutex_destroy(&vc->vc_lock);
+}
+
+void
+vdev_cache_stat_init(void)
+{
+ vdc_ksp = kstat_create("zfs", 0, "vdev_cache_stats", "misc",
+ KSTAT_TYPE_NAMED, sizeof (vdc_stats) / sizeof (kstat_named_t),
+ KSTAT_FLAG_VIRTUAL);
+ if (vdc_ksp != NULL) {
+ vdc_ksp->ks_data = &vdc_stats;
+ kstat_install(vdc_ksp);
+ }
+}
+
+void
+vdev_cache_stat_fini(void)
+{
+ if (vdc_ksp != NULL) {
+ kstat_delete(vdc_ksp);
+ vdc_ksp = NULL;
+ }
+}
diff --git a/module/zfs/vdev_file.c b/module/zfs/vdev_file.c
new file mode 100644
index 000000000..dc0e920bf
--- /dev/null
+++ b/module/zfs/vdev_file.c
@@ -0,0 +1,188 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_file.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/fs/zfs.h>
+#include <sys/fm/fs/zfs.h>
+
+/*
+ * Virtual device vector for files.
+ */
+
+static int
+vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift)
+{
+ vdev_file_t *vf;
+ vnode_t *vp;
+ vattr_t vattr;
+ int error;
+
+ /*
+ * We must have a pathname, and it must be absolute.
+ */
+ if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
+ vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
+ return (EINVAL);
+ }
+
+ vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
+
+ /*
+ * We always open the files from the root of the global zone, even if
+ * we're in a local zone. If the user has gotten to this point, the
+ * administrator has already decided that the pool should be available
+ * to local zone users, so the underlying devices should be as well.
+ */
+ ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
+ error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
+ spa_mode | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
+
+ if (error) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
+ return (error);
+ }
+
+ vf->vf_vnode = vp;
+
+#ifdef _KERNEL
+ /*
+ * Make sure it's a regular file.
+ */
+ if (vp->v_type != VREG) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
+ return (ENODEV);
+ }
+#endif
+ /*
+ * Determine the physical size of the file.
+ */
+ vattr.va_mask = AT_SIZE;
+ error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
+ if (error) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
+ return (error);
+ }
+
+ *psize = vattr.va_size;
+ *ashift = SPA_MINBLOCKSHIFT;
+
+ return (0);
+}
+
+static void
+vdev_file_close(vdev_t *vd)
+{
+ vdev_file_t *vf = vd->vdev_tsd;
+
+ if (vf == NULL)
+ return;
+
+ if (vf->vf_vnode != NULL) {
+ (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
+ (void) VOP_CLOSE(vf->vf_vnode, spa_mode, 1, 0, kcred, NULL);
+ VN_RELE(vf->vf_vnode);
+ }
+
+ kmem_free(vf, sizeof (vdev_file_t));
+ vd->vdev_tsd = NULL;
+}
+
+static int
+vdev_file_io_start(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ vdev_file_t *vf = vd->vdev_tsd;
+ ssize_t resid;
+
+ if (zio->io_type == ZIO_TYPE_IOCTL) {
+ /* XXPOLICY */
+ if (!vdev_readable(vd)) {
+ zio->io_error = ENXIO;
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+
+ switch (zio->io_cmd) {
+ case DKIOCFLUSHWRITECACHE:
+ zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
+ kcred, NULL);
+ break;
+ default:
+ zio->io_error = ENOTSUP;
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+
+ zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
+ UIO_READ : UIO_WRITE, vf->vf_vnode, zio->io_data,
+ zio->io_size, zio->io_offset, UIO_SYSSPACE,
+ 0, RLIM64_INFINITY, kcred, &resid);
+
+ if (resid != 0 && zio->io_error == 0)
+ zio->io_error = ENOSPC;
+
+ zio_interrupt(zio);
+
+ return (ZIO_PIPELINE_STOP);
+}
+
+/* ARGSUSED */
+static void
+vdev_file_io_done(zio_t *zio)
+{
+}
+
+vdev_ops_t vdev_file_ops = {
+ vdev_file_open,
+ vdev_file_close,
+ vdev_default_asize,
+ vdev_file_io_start,
+ vdev_file_io_done,
+ NULL,
+ VDEV_TYPE_FILE, /* name of this vdev type */
+ B_TRUE /* leaf vdev */
+};
+
+/*
+ * From userland we access disks just like files.
+ */
+#ifndef _KERNEL
+
+vdev_ops_t vdev_disk_ops = {
+ vdev_file_open,
+ vdev_file_close,
+ vdev_default_asize,
+ vdev_file_io_start,
+ vdev_file_io_done,
+ NULL,
+ VDEV_TYPE_DISK, /* name of this vdev type */
+ B_TRUE /* leaf vdev */
+};
+
+#endif
diff --git a/module/zfs/vdev_label.c b/module/zfs/vdev_label.c
new file mode 100644
index 000000000..bf930466f
--- /dev/null
+++ b/module/zfs/vdev_label.c
@@ -0,0 +1,1078 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * Virtual Device Labels
+ * ---------------------
+ *
+ * The vdev label serves several distinct purposes:
+ *
+ * 1. Uniquely identify this device as part of a ZFS pool and confirm its
+ * identity within the pool.
+ *
+ * 2. Verify that all the devices given in a configuration are present
+ * within the pool.
+ *
+ * 3. Determine the uberblock for the pool.
+ *
+ * 4. In case of an import operation, determine the configuration of the
+ * toplevel vdev of which it is a part.
+ *
+ * 5. If an import operation cannot find all the devices in the pool,
+ * provide enough information to the administrator to determine which
+ * devices are missing.
+ *
+ * It is important to note that while the kernel is responsible for writing the
+ * label, it only consumes the information in the first three cases. The
+ * latter information is only consumed in userland when determining the
+ * configuration to import a pool.
+ *
+ *
+ * Label Organization
+ * ------------------
+ *
+ * Before describing the contents of the label, it's important to understand how
+ * the labels are written and updated with respect to the uberblock.
+ *
+ * When the pool configuration is altered, either because it was newly created
+ * or a device was added, we want to update all the labels such that we can deal
+ * with fatal failure at any point. To this end, each disk has two labels which
+ * are updated before and after the uberblock is synced. Assuming we have
+ * labels and an uberblock with the following transaction groups:
+ *
+ * L1 UB L2
+ * +------+ +------+ +------+
+ * | | | | | |
+ * | t10 | | t10 | | t10 |
+ * | | | | | |
+ * +------+ +------+ +------+
+ *
+ * In this stable state, the labels and the uberblock were all updated within
+ * the same transaction group (10). Each label is mirrored and checksummed, so
+ * that we can detect when we fail partway through writing the label.
+ *
+ * In order to identify which labels are valid, the labels are written in the
+ * following manner:
+ *
+ * 1. For each vdev, update 'L1' to the new label
+ * 2. Update the uberblock
+ * 3. For each vdev, update 'L2' to the new label
+ *
+ * Given arbitrary failure, we can determine the correct label to use based on
+ * the transaction group. If we fail after updating L1 but before updating the
+ * UB, we will notice that L1's transaction group is greater than the uberblock,
+ * so L2 must be valid. If we fail after writing the uberblock but before
+ * writing L2, we will notice that L2's transaction group is less than L1, and
+ * therefore L1 is valid.
+ *
+ * Another added complexity is that not every label is updated when the config
+ * is synced. If we add a single device, we do not want to have to re-write
+ * every label for every device in the pool. This means that both L1 and L2 may
+ * be older than the pool uberblock, because the necessary information is stored
+ * on another vdev.
+ *
+ *
+ * On-disk Format
+ * --------------
+ *
+ * The vdev label consists of two distinct parts, and is wrapped within the
+ * vdev_label_t structure. The label includes 8k of padding to permit legacy
+ * VTOC disk labels, but is otherwise ignored.
+ *
+ * The first half of the label is a packed nvlist which contains pool wide
+ * properties, per-vdev properties, and configuration information. It is
+ * described in more detail below.
+ *
+ * The latter half of the label consists of a redundant array of uberblocks.
+ * These uberblocks are updated whenever a transaction group is committed,
+ * or when the configuration is updated. When a pool is loaded, we scan each
+ * vdev for the 'best' uberblock.
+ *
+ *
+ * Configuration Information
+ * -------------------------
+ *
+ * The nvlist describing the pool and vdev contains the following elements:
+ *
+ * version ZFS on-disk version
+ * name Pool name
+ * state Pool state
+ * txg Transaction group in which this label was written
+ * pool_guid Unique identifier for this pool
+ * vdev_tree An nvlist describing vdev tree.
+ *
+ * Each leaf device label also contains the following:
+ *
+ * top_guid Unique ID for top-level vdev in which this is contained
+ * guid Unique ID for the leaf vdev
+ *
+ * The 'vs' configuration follows the format described in 'spa_config.c'.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/dmu.h>
+#include <sys/zap.h>
+#include <sys/vdev.h>
+#include <sys/vdev_impl.h>
+#include <sys/uberblock_impl.h>
+#include <sys/metaslab.h>
+#include <sys/zio.h>
+#include <sys/fs/zfs.h>
+
+/*
+ * Basic routines to read and write from a vdev label.
+ * Used throughout the rest of this file.
+ */
+uint64_t
+vdev_label_offset(uint64_t psize, int l, uint64_t offset)
+{
+ ASSERT(offset < sizeof (vdev_label_t));
+ ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0);
+
+ return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
+ 0 : psize - VDEV_LABELS * sizeof (vdev_label_t)));
+}
+
+/*
+ * Returns back the vdev label associated with the passed in offset.
+ */
+int
+vdev_label_number(uint64_t psize, uint64_t offset)
+{
+ int l;
+
+ if (offset >= psize - VDEV_LABEL_END_SIZE) {
+ offset -= psize - VDEV_LABEL_END_SIZE;
+ offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t);
+ }
+ l = offset / sizeof (vdev_label_t);
+ return (l < VDEV_LABELS ? l : -1);
+}
+
+static void
+vdev_label_read(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset,
+ uint64_t size, zio_done_func_t *done, void *private, int flags)
+{
+ ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) ==
+ SCL_STATE_ALL);
+ ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
+
+ zio_nowait(zio_read_phys(zio, vd,
+ vdev_label_offset(vd->vdev_psize, l, offset),
+ size, buf, ZIO_CHECKSUM_LABEL, done, private,
+ ZIO_PRIORITY_SYNC_READ, flags, B_TRUE));
+}
+
+static void
+vdev_label_write(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset,
+ uint64_t size, zio_done_func_t *done, void *private, int flags)
+{
+ ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL ||
+ (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
+ (SCL_CONFIG | SCL_STATE) &&
+ dsl_pool_sync_context(spa_get_dsl(zio->io_spa))));
+ ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
+
+ zio_nowait(zio_write_phys(zio, vd,
+ vdev_label_offset(vd->vdev_psize, l, offset),
+ size, buf, ZIO_CHECKSUM_LABEL, done, private,
+ ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE));
+}
+
+/*
+ * Generate the nvlist representing this vdev's config.
+ */
+nvlist_t *
+vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
+ boolean_t isspare, boolean_t isl2cache)
+{
+ nvlist_t *nv = NULL;
+
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE,
+ vd->vdev_ops->vdev_op_type) == 0);
+ if (!isspare && !isl2cache)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id)
+ == 0);
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid) == 0);
+
+ if (vd->vdev_path != NULL)
+ VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PATH,
+ vd->vdev_path) == 0);
+
+ if (vd->vdev_devid != NULL)
+ VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_DEVID,
+ vd->vdev_devid) == 0);
+
+ if (vd->vdev_physpath != NULL)
+ VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH,
+ vd->vdev_physpath) == 0);
+
+ if (vd->vdev_nparity != 0) {
+ ASSERT(strcmp(vd->vdev_ops->vdev_op_type,
+ VDEV_TYPE_RAIDZ) == 0);
+
+ /*
+ * Make sure someone hasn't managed to sneak a fancy new vdev
+ * into a crufty old storage pool.
+ */
+ ASSERT(vd->vdev_nparity == 1 ||
+ (vd->vdev_nparity == 2 &&
+ spa_version(spa) >= SPA_VERSION_RAID6));
+
+ /*
+ * Note that we'll add the nparity tag even on storage pools
+ * that only support a single parity device -- older software
+ * will just ignore it.
+ */
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY,
+ vd->vdev_nparity) == 0);
+ }
+
+ if (vd->vdev_wholedisk != -1ULL)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
+ vd->vdev_wholedisk) == 0);
+
+ if (vd->vdev_not_present)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1) == 0);
+
+ if (vd->vdev_isspare)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1) == 0);
+
+ if (!isspare && !isl2cache && vd == vd->vdev_top) {
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
+ vd->vdev_ms_array) == 0);
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
+ vd->vdev_ms_shift) == 0);
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT,
+ vd->vdev_ashift) == 0);
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE,
+ vd->vdev_asize) == 0);
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG,
+ vd->vdev_islog) == 0);
+ }
+
+ if (vd->vdev_dtl.smo_object != 0)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DTL,
+ vd->vdev_dtl.smo_object) == 0);
+
+ if (getstats) {
+ vdev_stat_t vs;
+ vdev_get_stats(vd, &vs);
+ VERIFY(nvlist_add_uint64_array(nv, ZPOOL_CONFIG_STATS,
+ (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t)) == 0);
+ }
+
+ if (!vd->vdev_ops->vdev_op_leaf) {
+ nvlist_t **child;
+ int c;
+
+ child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *),
+ KM_SLEEP);
+
+ for (c = 0; c < vd->vdev_children; c++)
+ child[c] = vdev_config_generate(spa, vd->vdev_child[c],
+ getstats, isspare, isl2cache);
+
+ VERIFY(nvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+ child, vd->vdev_children) == 0);
+
+ for (c = 0; c < vd->vdev_children; c++)
+ nvlist_free(child[c]);
+
+ kmem_free(child, vd->vdev_children * sizeof (nvlist_t *));
+
+ } else {
+ if (vd->vdev_offline && !vd->vdev_tmpoffline)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE,
+ B_TRUE) == 0);
+ if (vd->vdev_faulted)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED,
+ B_TRUE) == 0);
+ if (vd->vdev_degraded)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED,
+ B_TRUE) == 0);
+ if (vd->vdev_removed)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED,
+ B_TRUE) == 0);
+ if (vd->vdev_unspare)
+ VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE,
+ B_TRUE) == 0);
+ }
+
+ return (nv);
+}
+
+nvlist_t *
+vdev_label_read_config(vdev_t *vd)
+{
+ spa_t *spa = vd->vdev_spa;
+ nvlist_t *config = NULL;
+ vdev_phys_t *vp;
+ zio_t *zio;
+ int flags =
+ ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
+
+ ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
+
+ if (!vdev_readable(vd))
+ return (NULL);
+
+ vp = zio_buf_alloc(sizeof (vdev_phys_t));
+
+ for (int l = 0; l < VDEV_LABELS; l++) {
+
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ vdev_label_read(zio, vd, l, vp,
+ offsetof(vdev_label_t, vl_vdev_phys),
+ sizeof (vdev_phys_t), NULL, NULL, flags);
+
+ if (zio_wait(zio) == 0 &&
+ nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist),
+ &config, 0) == 0)
+ break;
+
+ if (config != NULL) {
+ nvlist_free(config);
+ config = NULL;
+ }
+ }
+
+ zio_buf_free(vp, sizeof (vdev_phys_t));
+
+ return (config);
+}
+
+/*
+ * Determine if a device is in use. The 'spare_guid' parameter will be filled
+ * in with the device guid if this spare is active elsewhere on the system.
+ */
+static boolean_t
+vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason,
+ uint64_t *spare_guid, uint64_t *l2cache_guid)
+{
+ spa_t *spa = vd->vdev_spa;
+ uint64_t state, pool_guid, device_guid, txg, spare_pool;
+ uint64_t vdtxg = 0;
+ nvlist_t *label;
+
+ if (spare_guid)
+ *spare_guid = 0ULL;
+ if (l2cache_guid)
+ *l2cache_guid = 0ULL;
+
+ /*
+ * Read the label, if any, and perform some basic sanity checks.
+ */
+ if ((label = vdev_label_read_config(vd)) == NULL)
+ return (B_FALSE);
+
+ (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
+ &vdtxg);
+
+ if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
+ &state) != 0 ||
+ nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID,
+ &device_guid) != 0) {
+ nvlist_free(label);
+ return (B_FALSE);
+ }
+
+ if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
+ (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID,
+ &pool_guid) != 0 ||
+ nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
+ &txg) != 0)) {
+ nvlist_free(label);
+ return (B_FALSE);
+ }
+
+ nvlist_free(label);
+
+ /*
+ * Check to see if this device indeed belongs to the pool it claims to
+ * be a part of. The only way this is allowed is if the device is a hot
+ * spare (which we check for later on).
+ */
+ if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
+ !spa_guid_exists(pool_guid, device_guid) &&
+ !spa_spare_exists(device_guid, NULL, NULL) &&
+ !spa_l2cache_exists(device_guid, NULL))
+ return (B_FALSE);
+
+ /*
+ * If the transaction group is zero, then this an initialized (but
+ * unused) label. This is only an error if the create transaction
+ * on-disk is the same as the one we're using now, in which case the
+ * user has attempted to add the same vdev multiple times in the same
+ * transaction.
+ */
+ if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
+ txg == 0 && vdtxg == crtxg)
+ return (B_TRUE);
+
+ /*
+ * Check to see if this is a spare device. We do an explicit check for
+ * spa_has_spare() here because it may be on our pending list of spares
+ * to add. We also check if it is an l2cache device.
+ */
+ if (spa_spare_exists(device_guid, &spare_pool, NULL) ||
+ spa_has_spare(spa, device_guid)) {
+ if (spare_guid)
+ *spare_guid = device_guid;
+
+ switch (reason) {
+ case VDEV_LABEL_CREATE:
+ case VDEV_LABEL_L2CACHE:
+ return (B_TRUE);
+
+ case VDEV_LABEL_REPLACE:
+ return (!spa_has_spare(spa, device_guid) ||
+ spare_pool != 0ULL);
+
+ case VDEV_LABEL_SPARE:
+ return (spa_has_spare(spa, device_guid));
+ }
+ }
+
+ /*
+ * Check to see if this is an l2cache device.
+ */
+ if (spa_l2cache_exists(device_guid, NULL))
+ return (B_TRUE);
+
+ /*
+ * If the device is marked ACTIVE, then this device is in use by another
+ * pool on the system.
+ */
+ return (state == POOL_STATE_ACTIVE);
+}
+
+/*
+ * Initialize a vdev label. We check to make sure each leaf device is not in
+ * use, and writable. We put down an initial label which we will later
+ * overwrite with a complete label. Note that it's important to do this
+ * sequentially, not in parallel, so that we catch cases of multiple use of the
+ * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
+ * itself.
+ */
+int
+vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
+{
+ spa_t *spa = vd->vdev_spa;
+ nvlist_t *label;
+ vdev_phys_t *vp;
+ vdev_boot_header_t *vb;
+ uberblock_t *ub;
+ zio_t *zio;
+ char *buf;
+ size_t buflen;
+ int error;
+ uint64_t spare_guid, l2cache_guid;
+ int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
+
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ if ((error = vdev_label_init(vd->vdev_child[c],
+ crtxg, reason)) != 0)
+ return (error);
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (0);
+
+ /*
+ * Dead vdevs cannot be initialized.
+ */
+ if (vdev_is_dead(vd))
+ return (EIO);
+
+ /*
+ * Determine if the vdev is in use.
+ */
+ if (reason != VDEV_LABEL_REMOVE &&
+ vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid))
+ return (EBUSY);
+
+ ASSERT(reason != VDEV_LABEL_REMOVE ||
+ vdev_inuse(vd, crtxg, reason, NULL, NULL));
+
+ /*
+ * If this is a request to add or replace a spare or l2cache device
+ * that is in use elsewhere on the system, then we must update the
+ * guid (which was initialized to a random value) to reflect the
+ * actual GUID (which is shared between multiple pools).
+ */
+ if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE &&
+ spare_guid != 0ULL) {
+ uint64_t guid_delta = spare_guid - vd->vdev_guid;
+
+ vd->vdev_guid += guid_delta;
+
+ for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
+ pvd->vdev_guid_sum += guid_delta;
+
+ /*
+ * If this is a replacement, then we want to fallthrough to the
+ * rest of the code. If we're adding a spare, then it's already
+ * labeled appropriately and we can just return.
+ */
+ if (reason == VDEV_LABEL_SPARE)
+ return (0);
+ ASSERT(reason == VDEV_LABEL_REPLACE);
+ }
+
+ if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE &&
+ l2cache_guid != 0ULL) {
+ uint64_t guid_delta = l2cache_guid - vd->vdev_guid;
+
+ vd->vdev_guid += guid_delta;
+
+ for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
+ pvd->vdev_guid_sum += guid_delta;
+
+ /*
+ * If this is a replacement, then we want to fallthrough to the
+ * rest of the code. If we're adding an l2cache, then it's
+ * already labeled appropriately and we can just return.
+ */
+ if (reason == VDEV_LABEL_L2CACHE)
+ return (0);
+ ASSERT(reason == VDEV_LABEL_REPLACE);
+ }
+
+ /*
+ * Initialize its label.
+ */
+ vp = zio_buf_alloc(sizeof (vdev_phys_t));
+ bzero(vp, sizeof (vdev_phys_t));
+
+ /*
+ * Generate a label describing the pool and our top-level vdev.
+ * We mark it as being from txg 0 to indicate that it's not
+ * really part of an active pool just yet. The labels will
+ * be written again with a meaningful txg by spa_sync().
+ */
+ if (reason == VDEV_LABEL_SPARE ||
+ (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) {
+ /*
+ * For inactive hot spares, we generate a special label that
+ * identifies as a mutually shared hot spare. We write the
+ * label if we are adding a hot spare, or if we are removing an
+ * active hot spare (in which case we want to revert the
+ * labels).
+ */
+ VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
+ spa_version(spa)) == 0);
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
+ POOL_STATE_SPARE) == 0);
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
+ vd->vdev_guid) == 0);
+ } else if (reason == VDEV_LABEL_L2CACHE ||
+ (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) {
+ /*
+ * For level 2 ARC devices, add a special label.
+ */
+ VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
+ spa_version(spa)) == 0);
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
+ POOL_STATE_L2CACHE) == 0);
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
+ vd->vdev_guid) == 0);
+ } else {
+ label = spa_config_generate(spa, vd, 0ULL, B_FALSE);
+
+ /*
+ * Add our creation time. This allows us to detect multiple
+ * vdev uses as described above, and automatically expires if we
+ * fail.
+ */
+ VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
+ crtxg) == 0);
+ }
+
+ buf = vp->vp_nvlist;
+ buflen = sizeof (vp->vp_nvlist);
+
+ error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP);
+ if (error != 0) {
+ nvlist_free(label);
+ zio_buf_free(vp, sizeof (vdev_phys_t));
+ /* EFAULT means nvlist_pack ran out of room */
+ return (error == EFAULT ? ENAMETOOLONG : EINVAL);
+ }
+
+ /*
+ * Initialize boot block header.
+ */
+ vb = zio_buf_alloc(sizeof (vdev_boot_header_t));
+ bzero(vb, sizeof (vdev_boot_header_t));
+ vb->vb_magic = VDEV_BOOT_MAGIC;
+ vb->vb_version = VDEV_BOOT_VERSION;
+ vb->vb_offset = VDEV_BOOT_OFFSET;
+ vb->vb_size = VDEV_BOOT_SIZE;
+
+ /*
+ * Initialize uberblock template.
+ */
+ ub = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd));
+ bzero(ub, VDEV_UBERBLOCK_SIZE(vd));
+ *ub = spa->spa_uberblock;
+ ub->ub_txg = 0;
+
+ /*
+ * Write everything in parallel.
+ */
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ for (int l = 0; l < VDEV_LABELS; l++) {
+
+ vdev_label_write(zio, vd, l, vp,
+ offsetof(vdev_label_t, vl_vdev_phys),
+ sizeof (vdev_phys_t), NULL, NULL, flags);
+
+ vdev_label_write(zio, vd, l, vb,
+ offsetof(vdev_label_t, vl_boot_header),
+ sizeof (vdev_boot_header_t), NULL, NULL, flags);
+
+ for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
+ vdev_label_write(zio, vd, l, ub,
+ VDEV_UBERBLOCK_OFFSET(vd, n),
+ VDEV_UBERBLOCK_SIZE(vd), NULL, NULL, flags);
+ }
+ }
+
+ error = zio_wait(zio);
+
+ nvlist_free(label);
+ zio_buf_free(ub, VDEV_UBERBLOCK_SIZE(vd));
+ zio_buf_free(vb, sizeof (vdev_boot_header_t));
+ zio_buf_free(vp, sizeof (vdev_phys_t));
+
+ /*
+ * If this vdev hasn't been previously identified as a spare, then we
+ * mark it as such only if a) we are labeling it as a spare, or b) it
+ * exists as a spare elsewhere in the system. Do the same for
+ * level 2 ARC devices.
+ */
+ if (error == 0 && !vd->vdev_isspare &&
+ (reason == VDEV_LABEL_SPARE ||
+ spa_spare_exists(vd->vdev_guid, NULL, NULL)))
+ spa_spare_add(vd);
+
+ if (error == 0 && !vd->vdev_isl2cache &&
+ (reason == VDEV_LABEL_L2CACHE ||
+ spa_l2cache_exists(vd->vdev_guid, NULL)))
+ spa_l2cache_add(vd);
+
+ return (error);
+}
+
+/*
+ * ==========================================================================
+ * uberblock load/sync
+ * ==========================================================================
+ */
+
+/*
+ * Consider the following situation: txg is safely synced to disk. We've
+ * written the first uberblock for txg + 1, and then we lose power. When we
+ * come back up, we fail to see the uberblock for txg + 1 because, say,
+ * it was on a mirrored device and the replica to which we wrote txg + 1
+ * is now offline. If we then make some changes and sync txg + 1, and then
+ * the missing replica comes back, then for a new seconds we'll have two
+ * conflicting uberblocks on disk with the same txg. The solution is simple:
+ * among uberblocks with equal txg, choose the one with the latest timestamp.
+ */
+static int
+vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2)
+{
+ if (ub1->ub_txg < ub2->ub_txg)
+ return (-1);
+ if (ub1->ub_txg > ub2->ub_txg)
+ return (1);
+
+ if (ub1->ub_timestamp < ub2->ub_timestamp)
+ return (-1);
+ if (ub1->ub_timestamp > ub2->ub_timestamp)
+ return (1);
+
+ return (0);
+}
+
+static void
+vdev_uberblock_load_done(zio_t *zio)
+{
+ zio_t *rio = zio->io_private;
+ uberblock_t *ub = zio->io_data;
+ uberblock_t *ubbest = rio->io_private;
+
+ ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(zio->io_vd));
+
+ if (zio->io_error == 0 && uberblock_verify(ub) == 0) {
+ mutex_enter(&rio->io_lock);
+ if (vdev_uberblock_compare(ub, ubbest) > 0)
+ *ubbest = *ub;
+ mutex_exit(&rio->io_lock);
+ }
+
+ zio_buf_free(zio->io_data, zio->io_size);
+}
+
+void
+vdev_uberblock_load(zio_t *zio, vdev_t *vd, uberblock_t *ubbest)
+{
+ spa_t *spa = vd->vdev_spa;
+ vdev_t *rvd = spa->spa_root_vdev;
+ int flags =
+ ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
+
+ if (vd == rvd) {
+ ASSERT(zio == NULL);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ zio = zio_root(spa, NULL, ubbest, flags);
+ bzero(ubbest, sizeof (uberblock_t));
+ }
+
+ ASSERT(zio != NULL);
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ vdev_uberblock_load(zio, vd->vdev_child[c], ubbest);
+
+ if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) {
+ for (int l = 0; l < VDEV_LABELS; l++) {
+ for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
+ vdev_label_read(zio, vd, l,
+ zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)),
+ VDEV_UBERBLOCK_OFFSET(vd, n),
+ VDEV_UBERBLOCK_SIZE(vd),
+ vdev_uberblock_load_done, zio, flags);
+ }
+ }
+ }
+
+ if (vd == rvd) {
+ (void) zio_wait(zio);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+}
+
+/*
+ * On success, increment root zio's count of good writes.
+ * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
+ */
+static void
+vdev_uberblock_sync_done(zio_t *zio)
+{
+ uint64_t *good_writes = zio->io_private;
+
+ if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0)
+ atomic_add_64(good_writes, 1);
+}
+
+/*
+ * Write the uberblock to all labels of all leaves of the specified vdev.
+ */
+static void
+vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags)
+{
+ uberblock_t *ubbuf;
+ int n;
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags);
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return;
+
+ if (!vdev_writeable(vd))
+ return;
+
+ n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1);
+
+ ubbuf = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd));
+ bzero(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
+ *ubbuf = *ub;
+
+ for (int l = 0; l < VDEV_LABELS; l++)
+ vdev_label_write(zio, vd, l, ubbuf,
+ VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd),
+ vdev_uberblock_sync_done, zio->io_private,
+ flags | ZIO_FLAG_DONT_PROPAGATE);
+
+ zio_buf_free(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
+}
+
+int
+vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags)
+{
+ spa_t *spa = svd[0]->vdev_spa;
+ zio_t *zio;
+ uint64_t good_writes = 0;
+
+ zio = zio_root(spa, NULL, &good_writes, flags);
+
+ for (int v = 0; v < svdcount; v++)
+ vdev_uberblock_sync(zio, ub, svd[v], flags);
+
+ (void) zio_wait(zio);
+
+ /*
+ * Flush the uberblocks to disk. This ensures that the odd labels
+ * are no longer needed (because the new uberblocks and the even
+ * labels are safely on disk), so it is safe to overwrite them.
+ */
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ for (int v = 0; v < svdcount; v++)
+ zio_flush(zio, svd[v]);
+
+ (void) zio_wait(zio);
+
+ return (good_writes >= 1 ? 0 : EIO);
+}
+
+/*
+ * On success, increment the count of good writes for our top-level vdev.
+ */
+static void
+vdev_label_sync_done(zio_t *zio)
+{
+ uint64_t *good_writes = zio->io_private;
+
+ if (zio->io_error == 0)
+ atomic_add_64(good_writes, 1);
+}
+
+/*
+ * If there weren't enough good writes, indicate failure to the parent.
+ */
+static void
+vdev_label_sync_top_done(zio_t *zio)
+{
+ uint64_t *good_writes = zio->io_private;
+
+ if (*good_writes == 0)
+ zio->io_error = EIO;
+
+ kmem_free(good_writes, sizeof (uint64_t));
+}
+
+/*
+ * We ignore errors for log and cache devices, simply free the private data.
+ */
+static void
+vdev_label_sync_ignore_done(zio_t *zio)
+{
+ kmem_free(zio->io_private, sizeof (uint64_t));
+}
+
+/*
+ * Write all even or odd labels to all leaves of the specified vdev.
+ */
+static void
+vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags)
+{
+ nvlist_t *label;
+ vdev_phys_t *vp;
+ char *buf;
+ size_t buflen;
+
+ for (int c = 0; c < vd->vdev_children; c++)
+ vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags);
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return;
+
+ if (!vdev_writeable(vd))
+ return;
+
+ /*
+ * Generate a label describing the top-level config to which we belong.
+ */
+ label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE);
+
+ vp = zio_buf_alloc(sizeof (vdev_phys_t));
+ bzero(vp, sizeof (vdev_phys_t));
+
+ buf = vp->vp_nvlist;
+ buflen = sizeof (vp->vp_nvlist);
+
+ if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) {
+ for (; l < VDEV_LABELS; l += 2) {
+ vdev_label_write(zio, vd, l, vp,
+ offsetof(vdev_label_t, vl_vdev_phys),
+ sizeof (vdev_phys_t),
+ vdev_label_sync_done, zio->io_private,
+ flags | ZIO_FLAG_DONT_PROPAGATE);
+ }
+ }
+
+ zio_buf_free(vp, sizeof (vdev_phys_t));
+ nvlist_free(label);
+}
+
+int
+vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags)
+{
+ list_t *dl = &spa->spa_config_dirty_list;
+ vdev_t *vd;
+ zio_t *zio;
+ int error;
+
+ /*
+ * Write the new labels to disk.
+ */
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) {
+ uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t),
+ KM_SLEEP);
+ zio_t *vio = zio_null(zio, spa,
+ (vd->vdev_islog || vd->vdev_aux != NULL) ?
+ vdev_label_sync_ignore_done : vdev_label_sync_top_done,
+ good_writes, flags);
+ vdev_label_sync(vio, vd, l, txg, flags);
+ zio_nowait(vio);
+ }
+
+ error = zio_wait(zio);
+
+ /*
+ * Flush the new labels to disk.
+ */
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd))
+ zio_flush(zio, vd);
+
+ (void) zio_wait(zio);
+
+ return (error);
+}
+
+/*
+ * Sync the uberblock and any changes to the vdev configuration.
+ *
+ * The order of operations is carefully crafted to ensure that
+ * if the system panics or loses power at any time, the state on disk
+ * is still transactionally consistent. The in-line comments below
+ * describe the failure semantics at each stage.
+ *
+ * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
+ * at any time, you can just call it again, and it will resume its work.
+ */
+int
+vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg)
+{
+ spa_t *spa = svd[0]->vdev_spa;
+ uberblock_t *ub = &spa->spa_uberblock;
+ vdev_t *vd;
+ zio_t *zio;
+ int error;
+ int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
+
+ ASSERT(ub->ub_txg <= txg);
+
+ /*
+ * If this isn't a resync due to I/O errors,
+ * and nothing changed in this transaction group,
+ * and the vdev configuration hasn't changed,
+ * then there's nothing to do.
+ */
+ if (ub->ub_txg < txg &&
+ uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE &&
+ list_is_empty(&spa->spa_config_dirty_list))
+ return (0);
+
+ if (txg > spa_freeze_txg(spa))
+ return (0);
+
+ ASSERT(txg <= spa->spa_final_txg);
+
+ /*
+ * Flush the write cache of every disk that's been written to
+ * in this transaction group. This ensures that all blocks
+ * written in this txg will be committed to stable storage
+ * before any uberblock that references them.
+ */
+ zio = zio_root(spa, NULL, NULL, flags);
+
+ for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd;
+ vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg)))
+ zio_flush(zio, vd);
+
+ (void) zio_wait(zio);
+
+ /*
+ * Sync out the even labels (L0, L2) for every dirty vdev. If the
+ * system dies in the middle of this process, that's OK: all of the
+ * even labels that made it to disk will be newer than any uberblock,
+ * and will therefore be considered invalid. The odd labels (L1, L3),
+ * which have not yet been touched, will still be valid. We flush
+ * the new labels to disk to ensure that all even-label updates
+ * are committed to stable storage before the uberblock update.
+ */
+ if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0)
+ return (error);
+
+ /*
+ * Sync the uberblocks to all vdevs in svd[].
+ * If the system dies in the middle of this step, there are two cases
+ * to consider, and the on-disk state is consistent either way:
+ *
+ * (1) If none of the new uberblocks made it to disk, then the
+ * previous uberblock will be the newest, and the odd labels
+ * (which had not yet been touched) will be valid with respect
+ * to that uberblock.
+ *
+ * (2) If one or more new uberblocks made it to disk, then they
+ * will be the newest, and the even labels (which had all
+ * been successfully committed) will be valid with respect
+ * to the new uberblocks.
+ */
+ if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0)
+ return (error);
+
+ /*
+ * Sync out odd labels for every dirty vdev. If the system dies
+ * in the middle of this process, the even labels and the new
+ * uberblocks will suffice to open the pool. The next time
+ * the pool is opened, the first thing we'll do -- before any
+ * user data is modified -- is mark every vdev dirty so that
+ * all labels will be brought up to date. We flush the new labels
+ * to disk to ensure that all odd-label updates are committed to
+ * stable storage before the next transaction group begins.
+ */
+ return (vdev_label_sync_list(spa, 1, txg, flags));
+}
diff --git a/module/zfs/vdev_mirror.c b/module/zfs/vdev_mirror.c
new file mode 100644
index 000000000..c4629ff45
--- /dev/null
+++ b/module/zfs/vdev_mirror.c
@@ -0,0 +1,480 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/fs/zfs.h>
+
+/*
+ * Virtual device vector for mirroring.
+ */
+
+typedef struct mirror_child {
+ vdev_t *mc_vd;
+ uint64_t mc_offset;
+ int mc_error;
+ uint8_t mc_tried;
+ uint8_t mc_skipped;
+ uint8_t mc_speculative;
+} mirror_child_t;
+
+typedef struct mirror_map {
+ int mm_children;
+ int mm_replacing;
+ int mm_preferred;
+ int mm_root;
+ mirror_child_t mm_child[1];
+} mirror_map_t;
+
+int vdev_mirror_shift = 21;
+
+static void
+vdev_mirror_map_free(zio_t *zio)
+{
+ mirror_map_t *mm = zio->io_vsd;
+
+ kmem_free(mm, offsetof(mirror_map_t, mm_child[mm->mm_children]));
+}
+
+static mirror_map_t *
+vdev_mirror_map_alloc(zio_t *zio)
+{
+ mirror_map_t *mm = NULL;
+ mirror_child_t *mc;
+ vdev_t *vd = zio->io_vd;
+ int c, d;
+
+ if (vd == NULL) {
+ dva_t *dva = zio->io_bp->blk_dva;
+ spa_t *spa = zio->io_spa;
+
+ c = BP_GET_NDVAS(zio->io_bp);
+
+ mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP);
+ mm->mm_children = c;
+ mm->mm_replacing = B_FALSE;
+ mm->mm_preferred = spa_get_random(c);
+ mm->mm_root = B_TRUE;
+
+ /*
+ * Check the other, lower-index DVAs to see if they're on
+ * the same vdev as the child we picked. If they are, use
+ * them since they are likely to have been allocated from
+ * the primary metaslab in use at the time, and hence are
+ * more likely to have locality with single-copy data.
+ */
+ for (c = mm->mm_preferred, d = c - 1; d >= 0; d--) {
+ if (DVA_GET_VDEV(&dva[d]) == DVA_GET_VDEV(&dva[c]))
+ mm->mm_preferred = d;
+ }
+
+ for (c = 0; c < mm->mm_children; c++) {
+ mc = &mm->mm_child[c];
+
+ mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c]));
+ mc->mc_offset = DVA_GET_OFFSET(&dva[c]);
+ }
+ } else {
+ c = vd->vdev_children;
+
+ mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP);
+ mm->mm_children = c;
+ mm->mm_replacing = (vd->vdev_ops == &vdev_replacing_ops ||
+ vd->vdev_ops == &vdev_spare_ops);
+ mm->mm_preferred = mm->mm_replacing ? 0 :
+ (zio->io_offset >> vdev_mirror_shift) % c;
+ mm->mm_root = B_FALSE;
+
+ for (c = 0; c < mm->mm_children; c++) {
+ mc = &mm->mm_child[c];
+ mc->mc_vd = vd->vdev_child[c];
+ mc->mc_offset = zio->io_offset;
+ }
+ }
+
+ zio->io_vsd = mm;
+ zio->io_vsd_free = vdev_mirror_map_free;
+ return (mm);
+}
+
+static int
+vdev_mirror_open(vdev_t *vd, uint64_t *asize, uint64_t *ashift)
+{
+ vdev_t *cvd;
+ uint64_t c;
+ int numerrors = 0;
+ int ret, lasterror = 0;
+
+ if (vd->vdev_children == 0) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
+ return (EINVAL);
+ }
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ cvd = vd->vdev_child[c];
+
+ if ((ret = vdev_open(cvd)) != 0) {
+ lasterror = ret;
+ numerrors++;
+ continue;
+ }
+
+ *asize = MIN(*asize - 1, cvd->vdev_asize - 1) + 1;
+ *ashift = MAX(*ashift, cvd->vdev_ashift);
+ }
+
+ if (numerrors == vd->vdev_children) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
+ return (lasterror);
+ }
+
+ return (0);
+}
+
+static void
+vdev_mirror_close(vdev_t *vd)
+{
+ uint64_t c;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_close(vd->vdev_child[c]);
+}
+
+static void
+vdev_mirror_child_done(zio_t *zio)
+{
+ mirror_child_t *mc = zio->io_private;
+
+ mc->mc_error = zio->io_error;
+ mc->mc_tried = 1;
+ mc->mc_skipped = 0;
+}
+
+static void
+vdev_mirror_scrub_done(zio_t *zio)
+{
+ mirror_child_t *mc = zio->io_private;
+
+ if (zio->io_error == 0) {
+ zio_t *pio = zio->io_parent;
+ mutex_enter(&pio->io_lock);
+ ASSERT3U(zio->io_size, >=, pio->io_size);
+ bcopy(zio->io_data, pio->io_data, pio->io_size);
+ mutex_exit(&pio->io_lock);
+ }
+
+ zio_buf_free(zio->io_data, zio->io_size);
+
+ mc->mc_error = zio->io_error;
+ mc->mc_tried = 1;
+ mc->mc_skipped = 0;
+}
+
+/*
+ * Try to find a child whose DTL doesn't contain the block we want to read.
+ * If we can't, try the read on any vdev we haven't already tried.
+ */
+static int
+vdev_mirror_child_select(zio_t *zio)
+{
+ mirror_map_t *mm = zio->io_vsd;
+ mirror_child_t *mc;
+ uint64_t txg = zio->io_txg;
+ int i, c;
+
+ ASSERT(zio->io_bp == NULL || zio->io_bp->blk_birth == txg);
+
+ /*
+ * Try to find a child whose DTL doesn't contain the block to read.
+ * If a child is known to be completely inaccessible (indicated by
+ * vdev_readable() returning B_FALSE), don't even try.
+ */
+ for (i = 0, c = mm->mm_preferred; i < mm->mm_children; i++, c++) {
+ if (c >= mm->mm_children)
+ c = 0;
+ mc = &mm->mm_child[c];
+ if (mc->mc_tried || mc->mc_skipped)
+ continue;
+ if (!vdev_readable(mc->mc_vd)) {
+ mc->mc_error = ENXIO;
+ mc->mc_tried = 1; /* don't even try */
+ mc->mc_skipped = 1;
+ continue;
+ }
+ if (!vdev_dtl_contains(&mc->mc_vd->vdev_dtl_map, txg, 1))
+ return (c);
+ mc->mc_error = ESTALE;
+ mc->mc_skipped = 1;
+ mc->mc_speculative = 1;
+ }
+
+ /*
+ * Every device is either missing or has this txg in its DTL.
+ * Look for any child we haven't already tried before giving up.
+ */
+ for (c = 0; c < mm->mm_children; c++)
+ if (!mm->mm_child[c].mc_tried)
+ return (c);
+
+ /*
+ * Every child failed. There's no place left to look.
+ */
+ return (-1);
+}
+
+static int
+vdev_mirror_io_start(zio_t *zio)
+{
+ mirror_map_t *mm;
+ mirror_child_t *mc;
+ int c, children;
+
+ mm = vdev_mirror_map_alloc(zio);
+
+ if (zio->io_type == ZIO_TYPE_READ) {
+ if ((zio->io_flags & ZIO_FLAG_SCRUB) && !mm->mm_replacing) {
+ /*
+ * For scrubbing reads we need to allocate a read
+ * buffer for each child and issue reads to all
+ * children. If any child succeeds, it will copy its
+ * data into zio->io_data in vdev_mirror_scrub_done.
+ */
+ for (c = 0; c < mm->mm_children; c++) {
+ mc = &mm->mm_child[c];
+ zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
+ mc->mc_vd, mc->mc_offset,
+ zio_buf_alloc(zio->io_size), zio->io_size,
+ zio->io_type, zio->io_priority, 0,
+ vdev_mirror_scrub_done, mc));
+ }
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+ /*
+ * For normal reads just pick one child.
+ */
+ c = vdev_mirror_child_select(zio);
+ children = (c >= 0);
+ } else {
+ ASSERT(zio->io_type == ZIO_TYPE_WRITE);
+
+ /*
+ * If this is a resilvering I/O to a replacing vdev,
+ * only the last child should be written -- unless the
+ * first child happens to have a DTL entry here as well.
+ * All other writes go to all children.
+ */
+ if ((zio->io_flags & ZIO_FLAG_RESILVER) && mm->mm_replacing &&
+ !vdev_dtl_contains(&mm->mm_child[0].mc_vd->vdev_dtl_map,
+ zio->io_txg, 1)) {
+ c = mm->mm_children - 1;
+ children = 1;
+ } else {
+ c = 0;
+ children = mm->mm_children;
+ }
+ }
+
+ while (children--) {
+ mc = &mm->mm_child[c];
+ zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
+ mc->mc_vd, mc->mc_offset, zio->io_data, zio->io_size,
+ zio->io_type, zio->io_priority, 0,
+ vdev_mirror_child_done, mc));
+ c++;
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+vdev_mirror_worst_error(mirror_map_t *mm)
+{
+ int error[2] = { 0, 0 };
+
+ for (int c = 0; c < mm->mm_children; c++) {
+ mirror_child_t *mc = &mm->mm_child[c];
+ int s = mc->mc_speculative;
+ error[s] = zio_worst_error(error[s], mc->mc_error);
+ }
+
+ return (error[0] ? error[0] : error[1]);
+}
+
+static void
+vdev_mirror_io_done(zio_t *zio)
+{
+ mirror_map_t *mm = zio->io_vsd;
+ mirror_child_t *mc;
+ int c;
+ int good_copies = 0;
+ int unexpected_errors = 0;
+
+ for (c = 0; c < mm->mm_children; c++) {
+ mc = &mm->mm_child[c];
+
+ if (mc->mc_error) {
+ if (!mc->mc_skipped)
+ unexpected_errors++;
+ } else if (mc->mc_tried) {
+ good_copies++;
+ }
+ }
+
+ if (zio->io_type == ZIO_TYPE_WRITE) {
+ /*
+ * XXX -- for now, treat partial writes as success.
+ *
+ * Now that we support write reallocation, it would be better
+ * to treat partial failure as real failure unless there are
+ * no non-degraded top-level vdevs left, and not update DTLs
+ * if we intend to reallocate.
+ */
+ /* XXPOLICY */
+ if (good_copies != mm->mm_children) {
+ /*
+ * Always require at least one good copy.
+ *
+ * For ditto blocks (io_vd == NULL), require
+ * all copies to be good.
+ *
+ * XXX -- for replacing vdevs, there's no great answer.
+ * If the old device is really dead, we may not even
+ * be able to access it -- so we only want to
+ * require good writes to the new device. But if
+ * the new device turns out to be flaky, we want
+ * to be able to detach it -- which requires all
+ * writes to the old device to have succeeded.
+ */
+ if (good_copies == 0 || zio->io_vd == NULL)
+ zio->io_error = vdev_mirror_worst_error(mm);
+ }
+ return;
+ }
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ);
+
+ /*
+ * If we don't have a good copy yet, keep trying other children.
+ */
+ /* XXPOLICY */
+ if (good_copies == 0 && (c = vdev_mirror_child_select(zio)) != -1) {
+ ASSERT(c >= 0 && c < mm->mm_children);
+ mc = &mm->mm_child[c];
+ zio_vdev_io_redone(zio);
+ zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
+ mc->mc_vd, mc->mc_offset, zio->io_data, zio->io_size,
+ ZIO_TYPE_READ, zio->io_priority, 0,
+ vdev_mirror_child_done, mc));
+ return;
+ }
+
+ /* XXPOLICY */
+ if (good_copies == 0) {
+ zio->io_error = vdev_mirror_worst_error(mm);
+ ASSERT(zio->io_error != 0);
+ }
+
+ if (good_copies && (spa_mode & FWRITE) &&
+ (unexpected_errors ||
+ (zio->io_flags & ZIO_FLAG_RESILVER) ||
+ ((zio->io_flags & ZIO_FLAG_SCRUB) && mm->mm_replacing))) {
+ /*
+ * Use the good data we have in hand to repair damaged children.
+ */
+ for (c = 0; c < mm->mm_children; c++) {
+ /*
+ * Don't rewrite known good children.
+ * Not only is it unnecessary, it could
+ * actually be harmful: if the system lost
+ * power while rewriting the only good copy,
+ * there would be no good copies left!
+ */
+ mc = &mm->mm_child[c];
+
+ if (mc->mc_error == 0) {
+ if (mc->mc_tried)
+ continue;
+ if (!(zio->io_flags & ZIO_FLAG_SCRUB) &&
+ !vdev_dtl_contains(&mc->mc_vd->vdev_dtl_map,
+ zio->io_txg, 1))
+ continue;
+ mc->mc_error = ESTALE;
+ }
+
+ zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
+ mc->mc_vd, mc->mc_offset,
+ zio->io_data, zio->io_size,
+ ZIO_TYPE_WRITE, zio->io_priority,
+ ZIO_FLAG_IO_REPAIR, NULL, NULL));
+ }
+ }
+}
+
+static void
+vdev_mirror_state_change(vdev_t *vd, int faulted, int degraded)
+{
+ if (faulted == vd->vdev_children)
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_NO_REPLICAS);
+ else if (degraded + faulted != 0)
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
+ else
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
+}
+
+vdev_ops_t vdev_mirror_ops = {
+ vdev_mirror_open,
+ vdev_mirror_close,
+ vdev_default_asize,
+ vdev_mirror_io_start,
+ vdev_mirror_io_done,
+ vdev_mirror_state_change,
+ VDEV_TYPE_MIRROR, /* name of this vdev type */
+ B_FALSE /* not a leaf vdev */
+};
+
+vdev_ops_t vdev_replacing_ops = {
+ vdev_mirror_open,
+ vdev_mirror_close,
+ vdev_default_asize,
+ vdev_mirror_io_start,
+ vdev_mirror_io_done,
+ vdev_mirror_state_change,
+ VDEV_TYPE_REPLACING, /* name of this vdev type */
+ B_FALSE /* not a leaf vdev */
+};
+
+vdev_ops_t vdev_spare_ops = {
+ vdev_mirror_open,
+ vdev_mirror_close,
+ vdev_default_asize,
+ vdev_mirror_io_start,
+ vdev_mirror_io_done,
+ vdev_mirror_state_change,
+ VDEV_TYPE_SPARE, /* name of this vdev type */
+ B_FALSE /* not a leaf vdev */
+};
diff --git a/module/zfs/vdev_missing.c b/module/zfs/vdev_missing.c
new file mode 100644
index 000000000..731f7d3dc
--- /dev/null
+++ b/module/zfs/vdev_missing.c
@@ -0,0 +1,85 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * The 'missing' vdev is a special vdev type used only during import. It
+ * signifies a placeholder in the root vdev for some vdev that we know is
+ * missing. We pass it down to the kernel to allow the rest of the
+ * configuration to parsed and an attempt made to open all available devices.
+ * Because its GUID is always 0, we know that the guid sum will mismatch and we
+ * won't be able to open the pool anyway.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/fs/zfs.h>
+#include <sys/zio.h>
+
+/* ARGSUSED */
+static int
+vdev_missing_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift)
+{
+ /*
+ * Really this should just fail. But then the root vdev will be in the
+ * faulted state with VDEV_AUX_NO_REPLICAS, when what we really want is
+ * VDEV_AUX_BAD_GUID_SUM. So we pretend to succeed, knowing that we
+ * will fail the GUID sum check before ever trying to open the pool.
+ */
+ *psize = SPA_MINDEVSIZE;
+ *ashift = SPA_MINBLOCKSHIFT;
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+vdev_missing_close(vdev_t *vd)
+{
+}
+
+/* ARGSUSED */
+static int
+vdev_missing_io_start(zio_t *zio)
+{
+ zio->io_error = ENOTSUP;
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/* ARGSUSED */
+static void
+vdev_missing_io_done(zio_t *zio)
+{
+}
+
+vdev_ops_t vdev_missing_ops = {
+ vdev_missing_open,
+ vdev_missing_close,
+ vdev_default_asize,
+ vdev_missing_io_start,
+ vdev_missing_io_done,
+ NULL,
+ VDEV_TYPE_MISSING, /* name of this vdev type */
+ B_TRUE /* leaf vdev */
+};
diff --git a/module/zfs/vdev_queue.c b/module/zfs/vdev_queue.c
new file mode 100644
index 000000000..46fca0e3b
--- /dev/null
+++ b/module/zfs/vdev_queue.c
@@ -0,0 +1,308 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/avl.h>
+
+/*
+ * These tunables are for performance analysis.
+ */
+/*
+ * zfs_vdev_max_pending is the maximum number of i/os concurrently
+ * pending to each device. zfs_vdev_min_pending is the initial number
+ * of i/os pending to each device (before it starts ramping up to
+ * max_pending).
+ */
+int zfs_vdev_max_pending = 35;
+int zfs_vdev_min_pending = 4;
+
+/* deadline = pri + (lbolt >> time_shift) */
+int zfs_vdev_time_shift = 6;
+
+/* exponential I/O issue ramp-up rate */
+int zfs_vdev_ramp_rate = 2;
+
+/*
+ * i/os will be aggregated into a single large i/o up to
+ * zfs_vdev_aggregation_limit bytes long.
+ */
+int zfs_vdev_aggregation_limit = SPA_MAXBLOCKSIZE;
+
+/*
+ * Virtual device vector for disk I/O scheduling.
+ */
+int
+vdev_queue_deadline_compare(const void *x1, const void *x2)
+{
+ const zio_t *z1 = x1;
+ const zio_t *z2 = x2;
+
+ if (z1->io_deadline < z2->io_deadline)
+ return (-1);
+ if (z1->io_deadline > z2->io_deadline)
+ return (1);
+
+ if (z1->io_offset < z2->io_offset)
+ return (-1);
+ if (z1->io_offset > z2->io_offset)
+ return (1);
+
+ if (z1 < z2)
+ return (-1);
+ if (z1 > z2)
+ return (1);
+
+ return (0);
+}
+
+int
+vdev_queue_offset_compare(const void *x1, const void *x2)
+{
+ const zio_t *z1 = x1;
+ const zio_t *z2 = x2;
+
+ if (z1->io_offset < z2->io_offset)
+ return (-1);
+ if (z1->io_offset > z2->io_offset)
+ return (1);
+
+ if (z1 < z2)
+ return (-1);
+ if (z1 > z2)
+ return (1);
+
+ return (0);
+}
+
+void
+vdev_queue_init(vdev_t *vd)
+{
+ vdev_queue_t *vq = &vd->vdev_queue;
+
+ mutex_init(&vq->vq_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ avl_create(&vq->vq_deadline_tree, vdev_queue_deadline_compare,
+ sizeof (zio_t), offsetof(struct zio, io_deadline_node));
+
+ avl_create(&vq->vq_read_tree, vdev_queue_offset_compare,
+ sizeof (zio_t), offsetof(struct zio, io_offset_node));
+
+ avl_create(&vq->vq_write_tree, vdev_queue_offset_compare,
+ sizeof (zio_t), offsetof(struct zio, io_offset_node));
+
+ avl_create(&vq->vq_pending_tree, vdev_queue_offset_compare,
+ sizeof (zio_t), offsetof(struct zio, io_offset_node));
+}
+
+void
+vdev_queue_fini(vdev_t *vd)
+{
+ vdev_queue_t *vq = &vd->vdev_queue;
+
+ avl_destroy(&vq->vq_deadline_tree);
+ avl_destroy(&vq->vq_read_tree);
+ avl_destroy(&vq->vq_write_tree);
+ avl_destroy(&vq->vq_pending_tree);
+
+ mutex_destroy(&vq->vq_lock);
+}
+
+static void
+vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio)
+{
+ avl_add(&vq->vq_deadline_tree, zio);
+ avl_add(zio->io_vdev_tree, zio);
+}
+
+static void
+vdev_queue_io_remove(vdev_queue_t *vq, zio_t *zio)
+{
+ avl_remove(&vq->vq_deadline_tree, zio);
+ avl_remove(zio->io_vdev_tree, zio);
+}
+
+static void
+vdev_queue_agg_io_done(zio_t *aio)
+{
+ zio_t *dio;
+ uint64_t offset = 0;
+
+ while ((dio = aio->io_delegate_list) != NULL) {
+ if (aio->io_type == ZIO_TYPE_READ)
+ bcopy((char *)aio->io_data + offset, dio->io_data,
+ dio->io_size);
+ offset += dio->io_size;
+ aio->io_delegate_list = dio->io_delegate_next;
+ dio->io_delegate_next = NULL;
+ dio->io_error = aio->io_error;
+ zio_execute(dio);
+ }
+ ASSERT3U(offset, ==, aio->io_size);
+
+ zio_buf_free(aio->io_data, aio->io_size);
+}
+
+#define IS_ADJACENT(io, nio) \
+ ((io)->io_offset + (io)->io_size == (nio)->io_offset)
+
+static zio_t *
+vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
+{
+ zio_t *fio, *lio, *aio, *dio;
+ avl_tree_t *tree;
+ uint64_t size;
+
+ ASSERT(MUTEX_HELD(&vq->vq_lock));
+
+ if (avl_numnodes(&vq->vq_pending_tree) >= pending_limit ||
+ avl_numnodes(&vq->vq_deadline_tree) == 0)
+ return (NULL);
+
+ fio = lio = avl_first(&vq->vq_deadline_tree);
+
+ tree = fio->io_vdev_tree;
+ size = fio->io_size;
+
+ while ((dio = AVL_PREV(tree, fio)) != NULL && IS_ADJACENT(dio, fio) &&
+ !((dio->io_flags | fio->io_flags) & ZIO_FLAG_DONT_AGGREGATE) &&
+ size + dio->io_size <= zfs_vdev_aggregation_limit) {
+ dio->io_delegate_next = fio;
+ fio = dio;
+ size += dio->io_size;
+ }
+
+ while ((dio = AVL_NEXT(tree, lio)) != NULL && IS_ADJACENT(lio, dio) &&
+ !((lio->io_flags | dio->io_flags) & ZIO_FLAG_DONT_AGGREGATE) &&
+ size + dio->io_size <= zfs_vdev_aggregation_limit) {
+ lio->io_delegate_next = dio;
+ lio = dio;
+ size += dio->io_size;
+ }
+
+ if (fio != lio) {
+ char *buf = zio_buf_alloc(size);
+ uint64_t offset = 0;
+
+ ASSERT(size <= zfs_vdev_aggregation_limit);
+
+ aio = zio_vdev_delegated_io(fio->io_vd, fio->io_offset,
+ buf, size, fio->io_type, ZIO_PRIORITY_NOW,
+ ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE,
+ vdev_queue_agg_io_done, NULL);
+
+ aio->io_delegate_list = fio;
+
+ for (dio = fio; dio != NULL; dio = dio->io_delegate_next) {
+ ASSERT(dio->io_type == aio->io_type);
+ ASSERT(dio->io_vdev_tree == tree);
+ if (dio->io_type == ZIO_TYPE_WRITE)
+ bcopy(dio->io_data, buf + offset, dio->io_size);
+ offset += dio->io_size;
+ vdev_queue_io_remove(vq, dio);
+ zio_vdev_io_bypass(dio);
+ }
+
+ ASSERT(offset == size);
+
+ avl_add(&vq->vq_pending_tree, aio);
+
+ return (aio);
+ }
+
+ ASSERT(fio->io_vdev_tree == tree);
+ vdev_queue_io_remove(vq, fio);
+
+ avl_add(&vq->vq_pending_tree, fio);
+
+ return (fio);
+}
+
+zio_t *
+vdev_queue_io(zio_t *zio)
+{
+ vdev_queue_t *vq = &zio->io_vd->vdev_queue;
+ zio_t *nio;
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
+
+ if (zio->io_flags & ZIO_FLAG_DONT_QUEUE)
+ return (zio);
+
+ zio->io_flags |= ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE;
+
+ if (zio->io_type == ZIO_TYPE_READ)
+ zio->io_vdev_tree = &vq->vq_read_tree;
+ else
+ zio->io_vdev_tree = &vq->vq_write_tree;
+
+ mutex_enter(&vq->vq_lock);
+
+ zio->io_deadline = (lbolt64 >> zfs_vdev_time_shift) + zio->io_priority;
+
+ vdev_queue_io_add(vq, zio);
+
+ nio = vdev_queue_io_to_issue(vq, zfs_vdev_min_pending);
+
+ mutex_exit(&vq->vq_lock);
+
+ if (nio == NULL)
+ return (NULL);
+
+ if (nio->io_done == vdev_queue_agg_io_done) {
+ zio_nowait(nio);
+ return (NULL);
+ }
+
+ return (nio);
+}
+
+void
+vdev_queue_io_done(zio_t *zio)
+{
+ vdev_queue_t *vq = &zio->io_vd->vdev_queue;
+
+ mutex_enter(&vq->vq_lock);
+
+ avl_remove(&vq->vq_pending_tree, zio);
+
+ for (int i = 0; i < zfs_vdev_ramp_rate; i++) {
+ zio_t *nio = vdev_queue_io_to_issue(vq, zfs_vdev_max_pending);
+ if (nio == NULL)
+ break;
+ mutex_exit(&vq->vq_lock);
+ if (nio->io_done == vdev_queue_agg_io_done) {
+ zio_nowait(nio);
+ } else {
+ zio_vdev_io_reissue(nio);
+ zio_execute(nio);
+ }
+ mutex_enter(&vq->vq_lock);
+ }
+
+ mutex_exit(&vq->vq_lock);
+}
diff --git a/module/zfs/vdev_raidz.c b/module/zfs/vdev_raidz.c
new file mode 100644
index 000000000..69e314468
--- /dev/null
+++ b/module/zfs/vdev_raidz.c
@@ -0,0 +1,1209 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+#include <sys/fs/zfs.h>
+#include <sys/fm/fs/zfs.h>
+
+/*
+ * Virtual device vector for RAID-Z.
+ *
+ * This vdev supports both single and double parity. For single parity, we
+ * use a simple XOR of all the data columns. For double parity, we use both
+ * the simple XOR as well as a technique described in "The mathematics of
+ * RAID-6" by H. Peter Anvin. This technique defines a Galois field, GF(2^8),
+ * over the integers expressable in a single byte. Briefly, the operations on
+ * the field are defined as follows:
+ *
+ * o addition (+) is represented by a bitwise XOR
+ * o subtraction (-) is therefore identical to addition: A + B = A - B
+ * o multiplication of A by 2 is defined by the following bitwise expression:
+ * (A * 2)_7 = A_6
+ * (A * 2)_6 = A_5
+ * (A * 2)_5 = A_4
+ * (A * 2)_4 = A_3 + A_7
+ * (A * 2)_3 = A_2 + A_7
+ * (A * 2)_2 = A_1 + A_7
+ * (A * 2)_1 = A_0
+ * (A * 2)_0 = A_7
+ *
+ * In C, multiplying by 2 is therefore ((a << 1) ^ ((a & 0x80) ? 0x1d : 0)).
+ *
+ * Observe that any number in the field (except for 0) can be expressed as a
+ * power of 2 -- a generator for the field. We store a table of the powers of
+ * 2 and logs base 2 for quick look ups, and exploit the fact that A * B can
+ * be rewritten as 2^(log_2(A) + log_2(B)) (where '+' is normal addition rather
+ * than field addition). The inverse of a field element A (A^-1) is A^254.
+ *
+ * The two parity columns, P and Q, over several data columns, D_0, ... D_n-1,
+ * can be expressed by field operations:
+ *
+ * P = D_0 + D_1 + ... + D_n-2 + D_n-1
+ * Q = 2^n-1 * D_0 + 2^n-2 * D_1 + ... + 2^1 * D_n-2 + 2^0 * D_n-1
+ * = ((...((D_0) * 2 + D_1) * 2 + ...) * 2 + D_n-2) * 2 + D_n-1
+ *
+ * See the reconstruction code below for how P and Q can used individually or
+ * in concert to recover missing data columns.
+ */
+
+typedef struct raidz_col {
+ uint64_t rc_devidx; /* child device index for I/O */
+ uint64_t rc_offset; /* device offset */
+ uint64_t rc_size; /* I/O size */
+ void *rc_data; /* I/O data */
+ int rc_error; /* I/O error for this device */
+ uint8_t rc_tried; /* Did we attempt this I/O column? */
+ uint8_t rc_skipped; /* Did we skip this I/O column? */
+} raidz_col_t;
+
+typedef struct raidz_map {
+ uint64_t rm_cols; /* Column count */
+ uint64_t rm_bigcols; /* Number of oversized columns */
+ uint64_t rm_asize; /* Actual total I/O size */
+ uint64_t rm_missingdata; /* Count of missing data devices */
+ uint64_t rm_missingparity; /* Count of missing parity devices */
+ uint64_t rm_firstdatacol; /* First data column/parity count */
+ raidz_col_t rm_col[1]; /* Flexible array of I/O columns */
+} raidz_map_t;
+
+#define VDEV_RAIDZ_P 0
+#define VDEV_RAIDZ_Q 1
+
+#define VDEV_RAIDZ_MAXPARITY 2
+
+#define VDEV_RAIDZ_MUL_2(a) (((a) << 1) ^ (((a) & 0x80) ? 0x1d : 0))
+
+/*
+ * These two tables represent powers and logs of 2 in the Galois field defined
+ * above. These values were computed by repeatedly multiplying by 2 as above.
+ */
+static const uint8_t vdev_raidz_pow2[256] = {
+ 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
+ 0x1d, 0x3a, 0x74, 0xe8, 0xcd, 0x87, 0x13, 0x26,
+ 0x4c, 0x98, 0x2d, 0x5a, 0xb4, 0x75, 0xea, 0xc9,
+ 0x8f, 0x03, 0x06, 0x0c, 0x18, 0x30, 0x60, 0xc0,
+ 0x9d, 0x27, 0x4e, 0x9c, 0x25, 0x4a, 0x94, 0x35,
+ 0x6a, 0xd4, 0xb5, 0x77, 0xee, 0xc1, 0x9f, 0x23,
+ 0x46, 0x8c, 0x05, 0x0a, 0x14, 0x28, 0x50, 0xa0,
+ 0x5d, 0xba, 0x69, 0xd2, 0xb9, 0x6f, 0xde, 0xa1,
+ 0x5f, 0xbe, 0x61, 0xc2, 0x99, 0x2f, 0x5e, 0xbc,
+ 0x65, 0xca, 0x89, 0x0f, 0x1e, 0x3c, 0x78, 0xf0,
+ 0xfd, 0xe7, 0xd3, 0xbb, 0x6b, 0xd6, 0xb1, 0x7f,
+ 0xfe, 0xe1, 0xdf, 0xa3, 0x5b, 0xb6, 0x71, 0xe2,
+ 0xd9, 0xaf, 0x43, 0x86, 0x11, 0x22, 0x44, 0x88,
+ 0x0d, 0x1a, 0x34, 0x68, 0xd0, 0xbd, 0x67, 0xce,
+ 0x81, 0x1f, 0x3e, 0x7c, 0xf8, 0xed, 0xc7, 0x93,
+ 0x3b, 0x76, 0xec, 0xc5, 0x97, 0x33, 0x66, 0xcc,
+ 0x85, 0x17, 0x2e, 0x5c, 0xb8, 0x6d, 0xda, 0xa9,
+ 0x4f, 0x9e, 0x21, 0x42, 0x84, 0x15, 0x2a, 0x54,
+ 0xa8, 0x4d, 0x9a, 0x29, 0x52, 0xa4, 0x55, 0xaa,
+ 0x49, 0x92, 0x39, 0x72, 0xe4, 0xd5, 0xb7, 0x73,
+ 0xe6, 0xd1, 0xbf, 0x63, 0xc6, 0x91, 0x3f, 0x7e,
+ 0xfc, 0xe5, 0xd7, 0xb3, 0x7b, 0xf6, 0xf1, 0xff,
+ 0xe3, 0xdb, 0xab, 0x4b, 0x96, 0x31, 0x62, 0xc4,
+ 0x95, 0x37, 0x6e, 0xdc, 0xa5, 0x57, 0xae, 0x41,
+ 0x82, 0x19, 0x32, 0x64, 0xc8, 0x8d, 0x07, 0x0e,
+ 0x1c, 0x38, 0x70, 0xe0, 0xdd, 0xa7, 0x53, 0xa6,
+ 0x51, 0xa2, 0x59, 0xb2, 0x79, 0xf2, 0xf9, 0xef,
+ 0xc3, 0x9b, 0x2b, 0x56, 0xac, 0x45, 0x8a, 0x09,
+ 0x12, 0x24, 0x48, 0x90, 0x3d, 0x7a, 0xf4, 0xf5,
+ 0xf7, 0xf3, 0xfb, 0xeb, 0xcb, 0x8b, 0x0b, 0x16,
+ 0x2c, 0x58, 0xb0, 0x7d, 0xfa, 0xe9, 0xcf, 0x83,
+ 0x1b, 0x36, 0x6c, 0xd8, 0xad, 0x47, 0x8e, 0x01
+};
+static const uint8_t vdev_raidz_log2[256] = {
+ 0x00, 0x00, 0x01, 0x19, 0x02, 0x32, 0x1a, 0xc6,
+ 0x03, 0xdf, 0x33, 0xee, 0x1b, 0x68, 0xc7, 0x4b,
+ 0x04, 0x64, 0xe0, 0x0e, 0x34, 0x8d, 0xef, 0x81,
+ 0x1c, 0xc1, 0x69, 0xf8, 0xc8, 0x08, 0x4c, 0x71,
+ 0x05, 0x8a, 0x65, 0x2f, 0xe1, 0x24, 0x0f, 0x21,
+ 0x35, 0x93, 0x8e, 0xda, 0xf0, 0x12, 0x82, 0x45,
+ 0x1d, 0xb5, 0xc2, 0x7d, 0x6a, 0x27, 0xf9, 0xb9,
+ 0xc9, 0x9a, 0x09, 0x78, 0x4d, 0xe4, 0x72, 0xa6,
+ 0x06, 0xbf, 0x8b, 0x62, 0x66, 0xdd, 0x30, 0xfd,
+ 0xe2, 0x98, 0x25, 0xb3, 0x10, 0x91, 0x22, 0x88,
+ 0x36, 0xd0, 0x94, 0xce, 0x8f, 0x96, 0xdb, 0xbd,
+ 0xf1, 0xd2, 0x13, 0x5c, 0x83, 0x38, 0x46, 0x40,
+ 0x1e, 0x42, 0xb6, 0xa3, 0xc3, 0x48, 0x7e, 0x6e,
+ 0x6b, 0x3a, 0x28, 0x54, 0xfa, 0x85, 0xba, 0x3d,
+ 0xca, 0x5e, 0x9b, 0x9f, 0x0a, 0x15, 0x79, 0x2b,
+ 0x4e, 0xd4, 0xe5, 0xac, 0x73, 0xf3, 0xa7, 0x57,
+ 0x07, 0x70, 0xc0, 0xf7, 0x8c, 0x80, 0x63, 0x0d,
+ 0x67, 0x4a, 0xde, 0xed, 0x31, 0xc5, 0xfe, 0x18,
+ 0xe3, 0xa5, 0x99, 0x77, 0x26, 0xb8, 0xb4, 0x7c,
+ 0x11, 0x44, 0x92, 0xd9, 0x23, 0x20, 0x89, 0x2e,
+ 0x37, 0x3f, 0xd1, 0x5b, 0x95, 0xbc, 0xcf, 0xcd,
+ 0x90, 0x87, 0x97, 0xb2, 0xdc, 0xfc, 0xbe, 0x61,
+ 0xf2, 0x56, 0xd3, 0xab, 0x14, 0x2a, 0x5d, 0x9e,
+ 0x84, 0x3c, 0x39, 0x53, 0x47, 0x6d, 0x41, 0xa2,
+ 0x1f, 0x2d, 0x43, 0xd8, 0xb7, 0x7b, 0xa4, 0x76,
+ 0xc4, 0x17, 0x49, 0xec, 0x7f, 0x0c, 0x6f, 0xf6,
+ 0x6c, 0xa1, 0x3b, 0x52, 0x29, 0x9d, 0x55, 0xaa,
+ 0xfb, 0x60, 0x86, 0xb1, 0xbb, 0xcc, 0x3e, 0x5a,
+ 0xcb, 0x59, 0x5f, 0xb0, 0x9c, 0xa9, 0xa0, 0x51,
+ 0x0b, 0xf5, 0x16, 0xeb, 0x7a, 0x75, 0x2c, 0xd7,
+ 0x4f, 0xae, 0xd5, 0xe9, 0xe6, 0xe7, 0xad, 0xe8,
+ 0x74, 0xd6, 0xf4, 0xea, 0xa8, 0x50, 0x58, 0xaf,
+};
+
+/*
+ * Multiply a given number by 2 raised to the given power.
+ */
+static uint8_t
+vdev_raidz_exp2(uint_t a, int exp)
+{
+ if (a == 0)
+ return (0);
+
+ ASSERT(exp >= 0);
+ ASSERT(vdev_raidz_log2[a] > 0 || a == 1);
+
+ exp += vdev_raidz_log2[a];
+ if (exp > 255)
+ exp -= 255;
+
+ return (vdev_raidz_pow2[exp]);
+}
+
+static void
+vdev_raidz_map_free(zio_t *zio)
+{
+ raidz_map_t *rm = zio->io_vsd;
+ int c;
+
+ for (c = 0; c < rm->rm_firstdatacol; c++)
+ zio_buf_free(rm->rm_col[c].rc_data, rm->rm_col[c].rc_size);
+
+ kmem_free(rm, offsetof(raidz_map_t, rm_col[rm->rm_cols]));
+}
+
+static raidz_map_t *
+vdev_raidz_map_alloc(zio_t *zio, uint64_t unit_shift, uint64_t dcols,
+ uint64_t nparity)
+{
+ raidz_map_t *rm;
+ uint64_t b = zio->io_offset >> unit_shift;
+ uint64_t s = zio->io_size >> unit_shift;
+ uint64_t f = b % dcols;
+ uint64_t o = (b / dcols) << unit_shift;
+ uint64_t q, r, c, bc, col, acols, coff, devidx;
+
+ q = s / (dcols - nparity);
+ r = s - q * (dcols - nparity);
+ bc = (r == 0 ? 0 : r + nparity);
+
+ acols = (q == 0 ? bc : dcols);
+
+ rm = kmem_alloc(offsetof(raidz_map_t, rm_col[acols]), KM_SLEEP);
+
+ rm->rm_cols = acols;
+ rm->rm_bigcols = bc;
+ rm->rm_asize = 0;
+ rm->rm_missingdata = 0;
+ rm->rm_missingparity = 0;
+ rm->rm_firstdatacol = nparity;
+
+ for (c = 0; c < acols; c++) {
+ col = f + c;
+ coff = o;
+ if (col >= dcols) {
+ col -= dcols;
+ coff += 1ULL << unit_shift;
+ }
+ rm->rm_col[c].rc_devidx = col;
+ rm->rm_col[c].rc_offset = coff;
+ rm->rm_col[c].rc_size = (q + (c < bc)) << unit_shift;
+ rm->rm_col[c].rc_data = NULL;
+ rm->rm_col[c].rc_error = 0;
+ rm->rm_col[c].rc_tried = 0;
+ rm->rm_col[c].rc_skipped = 0;
+ rm->rm_asize += rm->rm_col[c].rc_size;
+ }
+
+ rm->rm_asize = roundup(rm->rm_asize, (nparity + 1) << unit_shift);
+
+ for (c = 0; c < rm->rm_firstdatacol; c++)
+ rm->rm_col[c].rc_data = zio_buf_alloc(rm->rm_col[c].rc_size);
+
+ rm->rm_col[c].rc_data = zio->io_data;
+
+ for (c = c + 1; c < acols; c++)
+ rm->rm_col[c].rc_data = (char *)rm->rm_col[c - 1].rc_data +
+ rm->rm_col[c - 1].rc_size;
+
+ /*
+ * If all data stored spans all columns, there's a danger that parity
+ * will always be on the same device and, since parity isn't read
+ * during normal operation, that that device's I/O bandwidth won't be
+ * used effectively. We therefore switch the parity every 1MB.
+ *
+ * ... at least that was, ostensibly, the theory. As a practical
+ * matter unless we juggle the parity between all devices evenly, we
+ * won't see any benefit. Further, occasional writes that aren't a
+ * multiple of the LCM of the number of children and the minimum
+ * stripe width are sufficient to avoid pessimal behavior.
+ * Unfortunately, this decision created an implicit on-disk format
+ * requirement that we need to support for all eternity, but only
+ * for single-parity RAID-Z.
+ */
+ ASSERT(rm->rm_cols >= 2);
+ ASSERT(rm->rm_col[0].rc_size == rm->rm_col[1].rc_size);
+
+ if (rm->rm_firstdatacol == 1 && (zio->io_offset & (1ULL << 20))) {
+ devidx = rm->rm_col[0].rc_devidx;
+ o = rm->rm_col[0].rc_offset;
+ rm->rm_col[0].rc_devidx = rm->rm_col[1].rc_devidx;
+ rm->rm_col[0].rc_offset = rm->rm_col[1].rc_offset;
+ rm->rm_col[1].rc_devidx = devidx;
+ rm->rm_col[1].rc_offset = o;
+ }
+
+ zio->io_vsd = rm;
+ zio->io_vsd_free = vdev_raidz_map_free;
+ return (rm);
+}
+
+static void
+vdev_raidz_generate_parity_p(raidz_map_t *rm)
+{
+ uint64_t *p, *src, pcount, ccount, i;
+ int c;
+
+ pcount = rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]);
+
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ src = rm->rm_col[c].rc_data;
+ p = rm->rm_col[VDEV_RAIDZ_P].rc_data;
+ ccount = rm->rm_col[c].rc_size / sizeof (src[0]);
+
+ if (c == rm->rm_firstdatacol) {
+ ASSERT(ccount == pcount);
+ for (i = 0; i < ccount; i++, p++, src++) {
+ *p = *src;
+ }
+ } else {
+ ASSERT(ccount <= pcount);
+ for (i = 0; i < ccount; i++, p++, src++) {
+ *p ^= *src;
+ }
+ }
+ }
+}
+
+static void
+vdev_raidz_generate_parity_pq(raidz_map_t *rm)
+{
+ uint64_t *q, *p, *src, pcount, ccount, mask, i;
+ int c;
+
+ pcount = rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]);
+ ASSERT(rm->rm_col[VDEV_RAIDZ_P].rc_size ==
+ rm->rm_col[VDEV_RAIDZ_Q].rc_size);
+
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ src = rm->rm_col[c].rc_data;
+ p = rm->rm_col[VDEV_RAIDZ_P].rc_data;
+ q = rm->rm_col[VDEV_RAIDZ_Q].rc_data;
+ ccount = rm->rm_col[c].rc_size / sizeof (src[0]);
+
+ if (c == rm->rm_firstdatacol) {
+ ASSERT(ccount == pcount || ccount == 0);
+ for (i = 0; i < ccount; i++, p++, q++, src++) {
+ *q = *src;
+ *p = *src;
+ }
+ for (; i < pcount; i++, p++, q++, src++) {
+ *q = 0;
+ *p = 0;
+ }
+ } else {
+ ASSERT(ccount <= pcount);
+
+ /*
+ * Rather than multiplying each byte individually (as
+ * described above), we are able to handle 8 at once
+ * by generating a mask based on the high bit in each
+ * byte and using that to conditionally XOR in 0x1d.
+ */
+ for (i = 0; i < ccount; i++, p++, q++, src++) {
+ mask = *q & 0x8080808080808080ULL;
+ mask = (mask << 1) - (mask >> 7);
+ *q = ((*q << 1) & 0xfefefefefefefefeULL) ^
+ (mask & 0x1d1d1d1d1d1d1d1dULL);
+ *q ^= *src;
+ *p ^= *src;
+ }
+
+ /*
+ * Treat short columns as though they are full of 0s.
+ */
+ for (; i < pcount; i++, q++) {
+ mask = *q & 0x8080808080808080ULL;
+ mask = (mask << 1) - (mask >> 7);
+ *q = ((*q << 1) & 0xfefefefefefefefeULL) ^
+ (mask & 0x1d1d1d1d1d1d1d1dULL);
+ }
+ }
+ }
+}
+
+static void
+vdev_raidz_reconstruct_p(raidz_map_t *rm, int x)
+{
+ uint64_t *dst, *src, xcount, ccount, count, i;
+ int c;
+
+ xcount = rm->rm_col[x].rc_size / sizeof (src[0]);
+ ASSERT(xcount <= rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]));
+ ASSERT(xcount > 0);
+
+ src = rm->rm_col[VDEV_RAIDZ_P].rc_data;
+ dst = rm->rm_col[x].rc_data;
+ for (i = 0; i < xcount; i++, dst++, src++) {
+ *dst = *src;
+ }
+
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ src = rm->rm_col[c].rc_data;
+ dst = rm->rm_col[x].rc_data;
+
+ if (c == x)
+ continue;
+
+ ccount = rm->rm_col[c].rc_size / sizeof (src[0]);
+ count = MIN(ccount, xcount);
+
+ for (i = 0; i < count; i++, dst++, src++) {
+ *dst ^= *src;
+ }
+ }
+}
+
+static void
+vdev_raidz_reconstruct_q(raidz_map_t *rm, int x)
+{
+ uint64_t *dst, *src, xcount, ccount, count, mask, i;
+ uint8_t *b;
+ int c, j, exp;
+
+ xcount = rm->rm_col[x].rc_size / sizeof (src[0]);
+ ASSERT(xcount <= rm->rm_col[VDEV_RAIDZ_Q].rc_size / sizeof (src[0]));
+
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ src = rm->rm_col[c].rc_data;
+ dst = rm->rm_col[x].rc_data;
+
+ if (c == x)
+ ccount = 0;
+ else
+ ccount = rm->rm_col[c].rc_size / sizeof (src[0]);
+
+ count = MIN(ccount, xcount);
+
+ if (c == rm->rm_firstdatacol) {
+ for (i = 0; i < count; i++, dst++, src++) {
+ *dst = *src;
+ }
+ for (; i < xcount; i++, dst++) {
+ *dst = 0;
+ }
+
+ } else {
+ /*
+ * For an explanation of this, see the comment in
+ * vdev_raidz_generate_parity_pq() above.
+ */
+ for (i = 0; i < count; i++, dst++, src++) {
+ mask = *dst & 0x8080808080808080ULL;
+ mask = (mask << 1) - (mask >> 7);
+ *dst = ((*dst << 1) & 0xfefefefefefefefeULL) ^
+ (mask & 0x1d1d1d1d1d1d1d1dULL);
+ *dst ^= *src;
+ }
+
+ for (; i < xcount; i++, dst++) {
+ mask = *dst & 0x8080808080808080ULL;
+ mask = (mask << 1) - (mask >> 7);
+ *dst = ((*dst << 1) & 0xfefefefefefefefeULL) ^
+ (mask & 0x1d1d1d1d1d1d1d1dULL);
+ }
+ }
+ }
+
+ src = rm->rm_col[VDEV_RAIDZ_Q].rc_data;
+ dst = rm->rm_col[x].rc_data;
+ exp = 255 - (rm->rm_cols - 1 - x);
+
+ for (i = 0; i < xcount; i++, dst++, src++) {
+ *dst ^= *src;
+ for (j = 0, b = (uint8_t *)dst; j < 8; j++, b++) {
+ *b = vdev_raidz_exp2(*b, exp);
+ }
+ }
+}
+
+static void
+vdev_raidz_reconstruct_pq(raidz_map_t *rm, int x, int y)
+{
+ uint8_t *p, *q, *pxy, *qxy, *xd, *yd, tmp, a, b, aexp, bexp;
+ void *pdata, *qdata;
+ uint64_t xsize, ysize, i;
+
+ ASSERT(x < y);
+ ASSERT(x >= rm->rm_firstdatacol);
+ ASSERT(y < rm->rm_cols);
+
+ ASSERT(rm->rm_col[x].rc_size >= rm->rm_col[y].rc_size);
+
+ /*
+ * Move the parity data aside -- we're going to compute parity as
+ * though columns x and y were full of zeros -- Pxy and Qxy. We want to
+ * reuse the parity generation mechanism without trashing the actual
+ * parity so we make those columns appear to be full of zeros by
+ * setting their lengths to zero.
+ */
+ pdata = rm->rm_col[VDEV_RAIDZ_P].rc_data;
+ qdata = rm->rm_col[VDEV_RAIDZ_Q].rc_data;
+ xsize = rm->rm_col[x].rc_size;
+ ysize = rm->rm_col[y].rc_size;
+
+ rm->rm_col[VDEV_RAIDZ_P].rc_data =
+ zio_buf_alloc(rm->rm_col[VDEV_RAIDZ_P].rc_size);
+ rm->rm_col[VDEV_RAIDZ_Q].rc_data =
+ zio_buf_alloc(rm->rm_col[VDEV_RAIDZ_Q].rc_size);
+ rm->rm_col[x].rc_size = 0;
+ rm->rm_col[y].rc_size = 0;
+
+ vdev_raidz_generate_parity_pq(rm);
+
+ rm->rm_col[x].rc_size = xsize;
+ rm->rm_col[y].rc_size = ysize;
+
+ p = pdata;
+ q = qdata;
+ pxy = rm->rm_col[VDEV_RAIDZ_P].rc_data;
+ qxy = rm->rm_col[VDEV_RAIDZ_Q].rc_data;
+ xd = rm->rm_col[x].rc_data;
+ yd = rm->rm_col[y].rc_data;
+
+ /*
+ * We now have:
+ * Pxy = P + D_x + D_y
+ * Qxy = Q + 2^(ndevs - 1 - x) * D_x + 2^(ndevs - 1 - y) * D_y
+ *
+ * We can then solve for D_x:
+ * D_x = A * (P + Pxy) + B * (Q + Qxy)
+ * where
+ * A = 2^(x - y) * (2^(x - y) + 1)^-1
+ * B = 2^(ndevs - 1 - x) * (2^(x - y) + 1)^-1
+ *
+ * With D_x in hand, we can easily solve for D_y:
+ * D_y = P + Pxy + D_x
+ */
+
+ a = vdev_raidz_pow2[255 + x - y];
+ b = vdev_raidz_pow2[255 - (rm->rm_cols - 1 - x)];
+ tmp = 255 - vdev_raidz_log2[a ^ 1];
+
+ aexp = vdev_raidz_log2[vdev_raidz_exp2(a, tmp)];
+ bexp = vdev_raidz_log2[vdev_raidz_exp2(b, tmp)];
+
+ for (i = 0; i < xsize; i++, p++, q++, pxy++, qxy++, xd++, yd++) {
+ *xd = vdev_raidz_exp2(*p ^ *pxy, aexp) ^
+ vdev_raidz_exp2(*q ^ *qxy, bexp);
+
+ if (i < ysize)
+ *yd = *p ^ *pxy ^ *xd;
+ }
+
+ zio_buf_free(rm->rm_col[VDEV_RAIDZ_P].rc_data,
+ rm->rm_col[VDEV_RAIDZ_P].rc_size);
+ zio_buf_free(rm->rm_col[VDEV_RAIDZ_Q].rc_data,
+ rm->rm_col[VDEV_RAIDZ_Q].rc_size);
+
+ /*
+ * Restore the saved parity data.
+ */
+ rm->rm_col[VDEV_RAIDZ_P].rc_data = pdata;
+ rm->rm_col[VDEV_RAIDZ_Q].rc_data = qdata;
+}
+
+
+static int
+vdev_raidz_open(vdev_t *vd, uint64_t *asize, uint64_t *ashift)
+{
+ vdev_t *cvd;
+ uint64_t nparity = vd->vdev_nparity;
+ int c, error;
+ int lasterror = 0;
+ int numerrors = 0;
+
+ ASSERT(nparity > 0);
+
+ if (nparity > VDEV_RAIDZ_MAXPARITY ||
+ vd->vdev_children < nparity + 1) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
+ return (EINVAL);
+ }
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ cvd = vd->vdev_child[c];
+
+ if ((error = vdev_open(cvd)) != 0) {
+ lasterror = error;
+ numerrors++;
+ continue;
+ }
+
+ *asize = MIN(*asize - 1, cvd->vdev_asize - 1) + 1;
+ *ashift = MAX(*ashift, cvd->vdev_ashift);
+ }
+
+ *asize *= vd->vdev_children;
+
+ if (numerrors > nparity) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
+ return (lasterror);
+ }
+
+ return (0);
+}
+
+static void
+vdev_raidz_close(vdev_t *vd)
+{
+ int c;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_close(vd->vdev_child[c]);
+}
+
+static uint64_t
+vdev_raidz_asize(vdev_t *vd, uint64_t psize)
+{
+ uint64_t asize;
+ uint64_t ashift = vd->vdev_top->vdev_ashift;
+ uint64_t cols = vd->vdev_children;
+ uint64_t nparity = vd->vdev_nparity;
+
+ asize = ((psize - 1) >> ashift) + 1;
+ asize += nparity * ((asize + cols - nparity - 1) / (cols - nparity));
+ asize = roundup(asize, nparity + 1) << ashift;
+
+ return (asize);
+}
+
+static void
+vdev_raidz_child_done(zio_t *zio)
+{
+ raidz_col_t *rc = zio->io_private;
+
+ rc->rc_error = zio->io_error;
+ rc->rc_tried = 1;
+ rc->rc_skipped = 0;
+}
+
+static int
+vdev_raidz_io_start(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ vdev_t *tvd = vd->vdev_top;
+ vdev_t *cvd;
+ blkptr_t *bp = zio->io_bp;
+ raidz_map_t *rm;
+ raidz_col_t *rc;
+ int c;
+
+ rm = vdev_raidz_map_alloc(zio, tvd->vdev_ashift, vd->vdev_children,
+ vd->vdev_nparity);
+
+ ASSERT3U(rm->rm_asize, ==, vdev_psize_to_asize(vd, zio->io_size));
+
+ if (zio->io_type == ZIO_TYPE_WRITE) {
+ /*
+ * Generate RAID parity in the first virtual columns.
+ */
+ if (rm->rm_firstdatacol == 1)
+ vdev_raidz_generate_parity_p(rm);
+ else
+ vdev_raidz_generate_parity_pq(rm);
+
+ for (c = 0; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ cvd = vd->vdev_child[rc->rc_devidx];
+ zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
+ rc->rc_offset, rc->rc_data, rc->rc_size,
+ zio->io_type, zio->io_priority, 0,
+ vdev_raidz_child_done, rc));
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ);
+
+ /*
+ * Iterate over the columns in reverse order so that we hit the parity
+ * last -- any errors along the way will force us to read the parity
+ * data.
+ */
+ for (c = rm->rm_cols - 1; c >= 0; c--) {
+ rc = &rm->rm_col[c];
+ cvd = vd->vdev_child[rc->rc_devidx];
+ if (!vdev_readable(cvd)) {
+ if (c >= rm->rm_firstdatacol)
+ rm->rm_missingdata++;
+ else
+ rm->rm_missingparity++;
+ rc->rc_error = ENXIO;
+ rc->rc_tried = 1; /* don't even try */
+ rc->rc_skipped = 1;
+ continue;
+ }
+ if (vdev_dtl_contains(&cvd->vdev_dtl_map, bp->blk_birth, 1)) {
+ if (c >= rm->rm_firstdatacol)
+ rm->rm_missingdata++;
+ else
+ rm->rm_missingparity++;
+ rc->rc_error = ESTALE;
+ rc->rc_skipped = 1;
+ continue;
+ }
+ if (c >= rm->rm_firstdatacol || rm->rm_missingdata > 0 ||
+ (zio->io_flags & ZIO_FLAG_SCRUB)) {
+ zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
+ rc->rc_offset, rc->rc_data, rc->rc_size,
+ zio->io_type, zio->io_priority, 0,
+ vdev_raidz_child_done, rc));
+ }
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/*
+ * Report a checksum error for a child of a RAID-Z device.
+ */
+static void
+raidz_checksum_error(zio_t *zio, raidz_col_t *rc)
+{
+ vdev_t *vd = zio->io_vd->vdev_child[rc->rc_devidx];
+
+ if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
+ mutex_enter(&vd->vdev_stat_lock);
+ vd->vdev_stat.vs_checksum_errors++;
+ mutex_exit(&vd->vdev_stat_lock);
+ }
+
+ if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE))
+ zfs_ereport_post(FM_EREPORT_ZFS_CHECKSUM,
+ zio->io_spa, vd, zio, rc->rc_offset, rc->rc_size);
+}
+
+/*
+ * Generate the parity from the data columns. If we tried and were able to
+ * read the parity without error, verify that the generated parity matches the
+ * data we read. If it doesn't, we fire off a checksum error. Return the
+ * number such failures.
+ */
+static int
+raidz_parity_verify(zio_t *zio, raidz_map_t *rm)
+{
+ void *orig[VDEV_RAIDZ_MAXPARITY];
+ int c, ret = 0;
+ raidz_col_t *rc;
+
+ for (c = 0; c < rm->rm_firstdatacol; c++) {
+ rc = &rm->rm_col[c];
+ if (!rc->rc_tried || rc->rc_error != 0)
+ continue;
+ orig[c] = zio_buf_alloc(rc->rc_size);
+ bcopy(rc->rc_data, orig[c], rc->rc_size);
+ }
+
+ if (rm->rm_firstdatacol == 1)
+ vdev_raidz_generate_parity_p(rm);
+ else
+ vdev_raidz_generate_parity_pq(rm);
+
+ for (c = 0; c < rm->rm_firstdatacol; c++) {
+ rc = &rm->rm_col[c];
+ if (!rc->rc_tried || rc->rc_error != 0)
+ continue;
+ if (bcmp(orig[c], rc->rc_data, rc->rc_size) != 0) {
+ raidz_checksum_error(zio, rc);
+ rc->rc_error = ECKSUM;
+ ret++;
+ }
+ zio_buf_free(orig[c], rc->rc_size);
+ }
+
+ return (ret);
+}
+
+static uint64_t raidz_corrected_p;
+static uint64_t raidz_corrected_q;
+static uint64_t raidz_corrected_pq;
+
+static int
+vdev_raidz_worst_error(raidz_map_t *rm)
+{
+ int error = 0;
+
+ for (int c = 0; c < rm->rm_cols; c++)
+ error = zio_worst_error(error, rm->rm_col[c].rc_error);
+
+ return (error);
+}
+
+static void
+vdev_raidz_io_done(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ vdev_t *cvd;
+ raidz_map_t *rm = zio->io_vsd;
+ raidz_col_t *rc, *rc1;
+ int unexpected_errors = 0;
+ int parity_errors = 0;
+ int parity_untried = 0;
+ int data_errors = 0;
+ int total_errors = 0;
+ int n, c, c1;
+
+ ASSERT(zio->io_bp != NULL); /* XXX need to add code to enforce this */
+
+ ASSERT(rm->rm_missingparity <= rm->rm_firstdatacol);
+ ASSERT(rm->rm_missingdata <= rm->rm_cols - rm->rm_firstdatacol);
+
+ for (c = 0; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+
+ if (rc->rc_error) {
+ ASSERT(rc->rc_error != ECKSUM); /* child has no bp */
+
+ if (c < rm->rm_firstdatacol)
+ parity_errors++;
+ else
+ data_errors++;
+
+ if (!rc->rc_skipped)
+ unexpected_errors++;
+
+ total_errors++;
+ } else if (c < rm->rm_firstdatacol && !rc->rc_tried) {
+ parity_untried++;
+ }
+ }
+
+ if (zio->io_type == ZIO_TYPE_WRITE) {
+ /*
+ * XXX -- for now, treat partial writes as a success.
+ * (If we couldn't write enough columns to reconstruct
+ * the data, the I/O failed. Otherwise, good enough.)
+ *
+ * Now that we support write reallocation, it would be better
+ * to treat partial failure as real failure unless there are
+ * no non-degraded top-level vdevs left, and not update DTLs
+ * if we intend to reallocate.
+ */
+ /* XXPOLICY */
+ if (total_errors > rm->rm_firstdatacol)
+ zio->io_error = vdev_raidz_worst_error(rm);
+
+ return;
+ }
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ);
+ /*
+ * There are three potential phases for a read:
+ * 1. produce valid data from the columns read
+ * 2. read all disks and try again
+ * 3. perform combinatorial reconstruction
+ *
+ * Each phase is progressively both more expensive and less likely to
+ * occur. If we encounter more errors than we can repair or all phases
+ * fail, we have no choice but to return an error.
+ */
+
+ /*
+ * If the number of errors we saw was correctable -- less than or equal
+ * to the number of parity disks read -- attempt to produce data that
+ * has a valid checksum. Naturally, this case applies in the absence of
+ * any errors.
+ */
+ if (total_errors <= rm->rm_firstdatacol - parity_untried) {
+ switch (data_errors) {
+ case 0:
+ if (zio_checksum_error(zio) == 0) {
+ /*
+ * If we read parity information (unnecessarily
+ * as it happens since no reconstruction was
+ * needed) regenerate and verify the parity.
+ * We also regenerate parity when resilvering
+ * so we can write it out to the failed device
+ * later.
+ */
+ if (parity_errors + parity_untried <
+ rm->rm_firstdatacol ||
+ (zio->io_flags & ZIO_FLAG_RESILVER)) {
+ n = raidz_parity_verify(zio, rm);
+ unexpected_errors += n;
+ ASSERT(parity_errors + n <=
+ rm->rm_firstdatacol);
+ }
+ goto done;
+ }
+ break;
+
+ case 1:
+ /*
+ * We either attempt to read all the parity columns or
+ * none of them. If we didn't try to read parity, we
+ * wouldn't be here in the correctable case. There must
+ * also have been fewer parity errors than parity
+ * columns or, again, we wouldn't be in this code path.
+ */
+ ASSERT(parity_untried == 0);
+ ASSERT(parity_errors < rm->rm_firstdatacol);
+
+ /*
+ * Find the column that reported the error.
+ */
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ if (rc->rc_error != 0)
+ break;
+ }
+ ASSERT(c != rm->rm_cols);
+ ASSERT(!rc->rc_skipped || rc->rc_error == ENXIO ||
+ rc->rc_error == ESTALE);
+
+ if (rm->rm_col[VDEV_RAIDZ_P].rc_error == 0) {
+ vdev_raidz_reconstruct_p(rm, c);
+ } else {
+ ASSERT(rm->rm_firstdatacol > 1);
+ vdev_raidz_reconstruct_q(rm, c);
+ }
+
+ if (zio_checksum_error(zio) == 0) {
+ if (rm->rm_col[VDEV_RAIDZ_P].rc_error == 0)
+ atomic_inc_64(&raidz_corrected_p);
+ else
+ atomic_inc_64(&raidz_corrected_q);
+
+ /*
+ * If there's more than one parity disk that
+ * was successfully read, confirm that the
+ * other parity disk produced the correct data.
+ * This routine is suboptimal in that it
+ * regenerates both the parity we wish to test
+ * as well as the parity we just used to
+ * perform the reconstruction, but this should
+ * be a relatively uncommon case, and can be
+ * optimized if it becomes a problem.
+ * We also regenerate parity when resilvering
+ * so we can write it out to the failed device
+ * later.
+ */
+ if (parity_errors < rm->rm_firstdatacol - 1 ||
+ (zio->io_flags & ZIO_FLAG_RESILVER)) {
+ n = raidz_parity_verify(zio, rm);
+ unexpected_errors += n;
+ ASSERT(parity_errors + n <=
+ rm->rm_firstdatacol);
+ }
+
+ goto done;
+ }
+ break;
+
+ case 2:
+ /*
+ * Two data column errors require double parity.
+ */
+ ASSERT(rm->rm_firstdatacol == 2);
+
+ /*
+ * Find the two columns that reported errors.
+ */
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ if (rc->rc_error != 0)
+ break;
+ }
+ ASSERT(c != rm->rm_cols);
+ ASSERT(!rc->rc_skipped || rc->rc_error == ENXIO ||
+ rc->rc_error == ESTALE);
+
+ for (c1 = c++; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ if (rc->rc_error != 0)
+ break;
+ }
+ ASSERT(c != rm->rm_cols);
+ ASSERT(!rc->rc_skipped || rc->rc_error == ENXIO ||
+ rc->rc_error == ESTALE);
+
+ vdev_raidz_reconstruct_pq(rm, c1, c);
+
+ if (zio_checksum_error(zio) == 0) {
+ atomic_inc_64(&raidz_corrected_pq);
+ goto done;
+ }
+ break;
+
+ default:
+ ASSERT(rm->rm_firstdatacol <= 2);
+ ASSERT(0);
+ }
+ }
+
+ /*
+ * This isn't a typical situation -- either we got a read error or
+ * a child silently returned bad data. Read every block so we can
+ * try again with as much data and parity as we can track down. If
+ * we've already been through once before, all children will be marked
+ * as tried so we'll proceed to combinatorial reconstruction.
+ */
+ unexpected_errors = 1;
+ rm->rm_missingdata = 0;
+ rm->rm_missingparity = 0;
+
+ for (c = 0; c < rm->rm_cols; c++) {
+ if (rm->rm_col[c].rc_tried)
+ continue;
+
+ zio_vdev_io_redone(zio);
+ do {
+ rc = &rm->rm_col[c];
+ if (rc->rc_tried)
+ continue;
+ zio_nowait(zio_vdev_child_io(zio, NULL,
+ vd->vdev_child[rc->rc_devidx],
+ rc->rc_offset, rc->rc_data, rc->rc_size,
+ zio->io_type, zio->io_priority, 0,
+ vdev_raidz_child_done, rc));
+ } while (++c < rm->rm_cols);
+
+ return;
+ }
+
+ /*
+ * At this point we've attempted to reconstruct the data given the
+ * errors we detected, and we've attempted to read all columns. There
+ * must, therefore, be one or more additional problems -- silent errors
+ * resulting in invalid data rather than explicit I/O errors resulting
+ * in absent data. Before we attempt combinatorial reconstruction make
+ * sure we have a chance of coming up with the right answer.
+ */
+ if (total_errors >= rm->rm_firstdatacol) {
+ zio->io_error = vdev_raidz_worst_error(rm);
+ /*
+ * If there were exactly as many device errors as parity
+ * columns, yet we couldn't reconstruct the data, then at
+ * least one device must have returned bad data silently.
+ */
+ if (total_errors == rm->rm_firstdatacol)
+ zio->io_error = zio_worst_error(zio->io_error, ECKSUM);
+ goto done;
+ }
+
+ if (rm->rm_col[VDEV_RAIDZ_P].rc_error == 0) {
+ /*
+ * Attempt to reconstruct the data from parity P.
+ */
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ void *orig;
+ rc = &rm->rm_col[c];
+
+ orig = zio_buf_alloc(rc->rc_size);
+ bcopy(rc->rc_data, orig, rc->rc_size);
+ vdev_raidz_reconstruct_p(rm, c);
+
+ if (zio_checksum_error(zio) == 0) {
+ zio_buf_free(orig, rc->rc_size);
+ atomic_inc_64(&raidz_corrected_p);
+
+ /*
+ * If this child didn't know that it returned
+ * bad data, inform it.
+ */
+ if (rc->rc_tried && rc->rc_error == 0)
+ raidz_checksum_error(zio, rc);
+ rc->rc_error = ECKSUM;
+ goto done;
+ }
+
+ bcopy(orig, rc->rc_data, rc->rc_size);
+ zio_buf_free(orig, rc->rc_size);
+ }
+ }
+
+ if (rm->rm_firstdatacol > 1 && rm->rm_col[VDEV_RAIDZ_Q].rc_error == 0) {
+ /*
+ * Attempt to reconstruct the data from parity Q.
+ */
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) {
+ void *orig;
+ rc = &rm->rm_col[c];
+
+ orig = zio_buf_alloc(rc->rc_size);
+ bcopy(rc->rc_data, orig, rc->rc_size);
+ vdev_raidz_reconstruct_q(rm, c);
+
+ if (zio_checksum_error(zio) == 0) {
+ zio_buf_free(orig, rc->rc_size);
+ atomic_inc_64(&raidz_corrected_q);
+
+ /*
+ * If this child didn't know that it returned
+ * bad data, inform it.
+ */
+ if (rc->rc_tried && rc->rc_error == 0)
+ raidz_checksum_error(zio, rc);
+ rc->rc_error = ECKSUM;
+ goto done;
+ }
+
+ bcopy(orig, rc->rc_data, rc->rc_size);
+ zio_buf_free(orig, rc->rc_size);
+ }
+ }
+
+ if (rm->rm_firstdatacol > 1 &&
+ rm->rm_col[VDEV_RAIDZ_P].rc_error == 0 &&
+ rm->rm_col[VDEV_RAIDZ_Q].rc_error == 0) {
+ /*
+ * Attempt to reconstruct the data from both P and Q.
+ */
+ for (c = rm->rm_firstdatacol; c < rm->rm_cols - 1; c++) {
+ void *orig, *orig1;
+ rc = &rm->rm_col[c];
+
+ orig = zio_buf_alloc(rc->rc_size);
+ bcopy(rc->rc_data, orig, rc->rc_size);
+
+ for (c1 = c + 1; c1 < rm->rm_cols; c1++) {
+ rc1 = &rm->rm_col[c1];
+
+ orig1 = zio_buf_alloc(rc1->rc_size);
+ bcopy(rc1->rc_data, orig1, rc1->rc_size);
+
+ vdev_raidz_reconstruct_pq(rm, c, c1);
+
+ if (zio_checksum_error(zio) == 0) {
+ zio_buf_free(orig, rc->rc_size);
+ zio_buf_free(orig1, rc1->rc_size);
+ atomic_inc_64(&raidz_corrected_pq);
+
+ /*
+ * If these children didn't know they
+ * returned bad data, inform them.
+ */
+ if (rc->rc_tried && rc->rc_error == 0)
+ raidz_checksum_error(zio, rc);
+ if (rc1->rc_tried && rc1->rc_error == 0)
+ raidz_checksum_error(zio, rc1);
+
+ rc->rc_error = ECKSUM;
+ rc1->rc_error = ECKSUM;
+
+ goto done;
+ }
+
+ bcopy(orig1, rc1->rc_data, rc1->rc_size);
+ zio_buf_free(orig1, rc1->rc_size);
+ }
+
+ bcopy(orig, rc->rc_data, rc->rc_size);
+ zio_buf_free(orig, rc->rc_size);
+ }
+ }
+
+ /*
+ * All combinations failed to checksum. Generate checksum ereports for
+ * all children.
+ */
+ zio->io_error = ECKSUM;
+
+ if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
+ for (c = 0; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ zfs_ereport_post(FM_EREPORT_ZFS_CHECKSUM,
+ zio->io_spa, vd->vdev_child[rc->rc_devidx], zio,
+ rc->rc_offset, rc->rc_size);
+ }
+ }
+
+done:
+ zio_checksum_verified(zio);
+
+ if (zio->io_error == 0 && (spa_mode & FWRITE) &&
+ (unexpected_errors || (zio->io_flags & ZIO_FLAG_RESILVER))) {
+ /*
+ * Use the good data we have in hand to repair damaged children.
+ */
+ for (c = 0; c < rm->rm_cols; c++) {
+ rc = &rm->rm_col[c];
+ cvd = vd->vdev_child[rc->rc_devidx];
+
+ if (rc->rc_error == 0)
+ continue;
+
+ zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
+ rc->rc_offset, rc->rc_data, rc->rc_size,
+ ZIO_TYPE_WRITE, zio->io_priority,
+ ZIO_FLAG_IO_REPAIR, NULL, NULL));
+ }
+ }
+}
+
+static void
+vdev_raidz_state_change(vdev_t *vd, int faulted, int degraded)
+{
+ if (faulted > vd->vdev_nparity)
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_NO_REPLICAS);
+ else if (degraded + faulted != 0)
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
+ else
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
+}
+
+vdev_ops_t vdev_raidz_ops = {
+ vdev_raidz_open,
+ vdev_raidz_close,
+ vdev_raidz_asize,
+ vdev_raidz_io_start,
+ vdev_raidz_io_done,
+ vdev_raidz_state_change,
+ VDEV_TYPE_RAIDZ, /* name of this vdev type */
+ B_FALSE /* not a leaf vdev */
+};
diff --git a/module/zfs/vdev_root.c b/module/zfs/vdev_root.c
new file mode 100644
index 000000000..88383f002
--- /dev/null
+++ b/module/zfs/vdev_root.c
@@ -0,0 +1,118 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+#include <sys/fs/zfs.h>
+
+/*
+ * Virtual device vector for the pool's root vdev.
+ */
+
+/*
+ * We should be able to tolerate one failure with absolutely no damage
+ * to our metadata. Two failures will take out space maps, a bunch of
+ * indirect block trees, meta dnodes, dnodes, etc. Probably not a happy
+ * place to live. When we get smarter, we can liberalize this policy.
+ * e.g. If we haven't lost two consecutive top-level vdevs, then we are
+ * probably fine. Adding bean counters during alloc/free can make this
+ * future guesswork more accurate.
+ */
+static int
+too_many_errors(vdev_t *vd, int numerrors)
+{
+ ASSERT3U(numerrors, <=, vd->vdev_children);
+ return (numerrors > 0);
+}
+
+static int
+vdev_root_open(vdev_t *vd, uint64_t *asize, uint64_t *ashift)
+{
+ int c;
+ int lasterror = 0;
+ int numerrors = 0;
+
+ if (vd->vdev_children == 0) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
+ return (EINVAL);
+ }
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ vdev_t *cvd = vd->vdev_child[c];
+ int error;
+
+ if ((error = vdev_open(cvd)) != 0 &&
+ !cvd->vdev_islog) {
+ lasterror = error;
+ numerrors++;
+ continue;
+ }
+ }
+
+ if (too_many_errors(vd, numerrors)) {
+ vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
+ return (lasterror);
+ }
+
+ *asize = 0;
+ *ashift = 0;
+
+ return (0);
+}
+
+static void
+vdev_root_close(vdev_t *vd)
+{
+ int c;
+
+ for (c = 0; c < vd->vdev_children; c++)
+ vdev_close(vd->vdev_child[c]);
+}
+
+static void
+vdev_root_state_change(vdev_t *vd, int faulted, int degraded)
+{
+ if (too_many_errors(vd, faulted)) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+ VDEV_AUX_NO_REPLICAS);
+ } else if (degraded) {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
+ } else {
+ vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
+ }
+}
+
+vdev_ops_t vdev_root_ops = {
+ vdev_root_open,
+ vdev_root_close,
+ vdev_default_asize,
+ NULL, /* io_start - not applicable to the root */
+ NULL, /* io_done - not applicable to the root */
+ vdev_root_state_change,
+ VDEV_TYPE_ROOT, /* name of this vdev type */
+ B_FALSE /* not a leaf vdev */
+};
diff --git a/module/zfs/zap.c b/module/zfs/zap.c
new file mode 100644
index 000000000..ca859ec35
--- /dev/null
+++ b/module/zfs/zap.c
@@ -0,0 +1,1136 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+
+/*
+ * This file contains the top half of the zfs directory structure
+ * implementation. The bottom half is in zap_leaf.c.
+ *
+ * The zdir is an extendable hash data structure. There is a table of
+ * pointers to buckets (zap_t->zd_data->zd_leafs). The buckets are
+ * each a constant size and hold a variable number of directory entries.
+ * The buckets (aka "leaf nodes") are implemented in zap_leaf.c.
+ *
+ * The pointer table holds a power of 2 number of pointers.
+ * (1<<zap_t->zd_data->zd_phys->zd_prefix_len). The bucket pointed to
+ * by the pointer at index i in the table holds entries whose hash value
+ * has a zd_prefix_len - bit prefix
+ */
+
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/zfs_context.h>
+#include <sys/zfs_znode.h>
+#include <sys/zap.h>
+#include <sys/refcount.h>
+#include <sys/zap_impl.h>
+#include <sys/zap_leaf.h>
+
+int fzap_default_block_shift = 14; /* 16k blocksize */
+
+static void zap_leaf_pageout(dmu_buf_t *db, void *vl);
+static uint64_t zap_allocate_blocks(zap_t *zap, int nblocks);
+
+
+void
+fzap_byteswap(void *vbuf, size_t size)
+{
+ uint64_t block_type;
+
+ block_type = *(uint64_t *)vbuf;
+
+ if (block_type == ZBT_LEAF || block_type == BSWAP_64(ZBT_LEAF))
+ zap_leaf_byteswap(vbuf, size);
+ else {
+ /* it's a ptrtbl block */
+ byteswap_uint64_array(vbuf, size);
+ }
+}
+
+void
+fzap_upgrade(zap_t *zap, dmu_tx_t *tx)
+{
+ dmu_buf_t *db;
+ zap_leaf_t *l;
+ int i;
+ zap_phys_t *zp;
+
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+ zap->zap_ismicro = FALSE;
+
+ (void) dmu_buf_update_user(zap->zap_dbuf, zap, zap,
+ &zap->zap_f.zap_phys, zap_evict);
+
+ mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
+ zap->zap_f.zap_block_shift = highbit(zap->zap_dbuf->db_size) - 1;
+
+ zp = zap->zap_f.zap_phys;
+ /*
+ * explicitly zero it since it might be coming from an
+ * initialized microzap
+ */
+ bzero(zap->zap_dbuf->db_data, zap->zap_dbuf->db_size);
+ zp->zap_block_type = ZBT_HEADER;
+ zp->zap_magic = ZAP_MAGIC;
+
+ zp->zap_ptrtbl.zt_shift = ZAP_EMBEDDED_PTRTBL_SHIFT(zap);
+
+ zp->zap_freeblk = 2; /* block 1 will be the first leaf */
+ zp->zap_num_leafs = 1;
+ zp->zap_num_entries = 0;
+ zp->zap_salt = zap->zap_salt;
+ zp->zap_normflags = zap->zap_normflags;
+
+ /* block 1 will be the first leaf */
+ for (i = 0; i < (1<<zp->zap_ptrtbl.zt_shift); i++)
+ ZAP_EMBEDDED_PTRTBL_ENT(zap, i) = 1;
+
+ /*
+ * set up block 1 - the first leaf
+ */
+ VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ 1<<FZAP_BLOCK_SHIFT(zap), FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+
+ l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
+ l->l_dbuf = db;
+ l->l_phys = db->db_data;
+
+ zap_leaf_init(l, zp->zap_normflags != 0);
+
+ kmem_free(l, sizeof (zap_leaf_t));
+ dmu_buf_rele(db, FTAG);
+}
+
+static int
+zap_tryupgradedir(zap_t *zap, dmu_tx_t *tx)
+{
+ if (RW_WRITE_HELD(&zap->zap_rwlock))
+ return (1);
+ if (rw_tryupgrade(&zap->zap_rwlock)) {
+ dmu_buf_will_dirty(zap->zap_dbuf, tx);
+ return (1);
+ }
+ return (0);
+}
+
+/*
+ * Generic routines for dealing with the pointer & cookie tables.
+ */
+
+static int
+zap_table_grow(zap_t *zap, zap_table_phys_t *tbl,
+ void (*transfer_func)(const uint64_t *src, uint64_t *dst, int n),
+ dmu_tx_t *tx)
+{
+ uint64_t b, newblk;
+ dmu_buf_t *db_old, *db_new;
+ int err;
+ int bs = FZAP_BLOCK_SHIFT(zap);
+ int hepb = 1<<(bs-4);
+ /* hepb = half the number of entries in a block */
+
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+ ASSERT(tbl->zt_blk != 0);
+ ASSERT(tbl->zt_numblks > 0);
+
+ if (tbl->zt_nextblk != 0) {
+ newblk = tbl->zt_nextblk;
+ } else {
+ newblk = zap_allocate_blocks(zap, tbl->zt_numblks * 2);
+ tbl->zt_nextblk = newblk;
+ ASSERT3U(tbl->zt_blks_copied, ==, 0);
+ dmu_prefetch(zap->zap_objset, zap->zap_object,
+ tbl->zt_blk << bs, tbl->zt_numblks << bs);
+ }
+
+ /*
+ * Copy the ptrtbl from the old to new location.
+ */
+
+ b = tbl->zt_blks_copied;
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (tbl->zt_blk + b) << bs, FTAG, &db_old);
+ if (err)
+ return (err);
+
+ /* first half of entries in old[b] go to new[2*b+0] */
+ VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (newblk + 2*b+0) << bs, FTAG, &db_new));
+ dmu_buf_will_dirty(db_new, tx);
+ transfer_func(db_old->db_data, db_new->db_data, hepb);
+ dmu_buf_rele(db_new, FTAG);
+
+ /* second half of entries in old[b] go to new[2*b+1] */
+ VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (newblk + 2*b+1) << bs, FTAG, &db_new));
+ dmu_buf_will_dirty(db_new, tx);
+ transfer_func((uint64_t *)db_old->db_data + hepb,
+ db_new->db_data, hepb);
+ dmu_buf_rele(db_new, FTAG);
+
+ dmu_buf_rele(db_old, FTAG);
+
+ tbl->zt_blks_copied++;
+
+ dprintf("copied block %llu of %llu\n",
+ tbl->zt_blks_copied, tbl->zt_numblks);
+
+ if (tbl->zt_blks_copied == tbl->zt_numblks) {
+ (void) dmu_free_range(zap->zap_objset, zap->zap_object,
+ tbl->zt_blk << bs, tbl->zt_numblks << bs, tx);
+
+ tbl->zt_blk = newblk;
+ tbl->zt_numblks *= 2;
+ tbl->zt_shift++;
+ tbl->zt_nextblk = 0;
+ tbl->zt_blks_copied = 0;
+
+ dprintf("finished; numblocks now %llu (%lluk entries)\n",
+ tbl->zt_numblks, 1<<(tbl->zt_shift-10));
+ }
+
+ return (0);
+}
+
+static int
+zap_table_store(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t val,
+ dmu_tx_t *tx)
+{
+ int err;
+ uint64_t blk, off;
+ int bs = FZAP_BLOCK_SHIFT(zap);
+ dmu_buf_t *db;
+
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+ ASSERT(tbl->zt_blk != 0);
+
+ dprintf("storing %llx at index %llx\n", val, idx);
+
+ blk = idx >> (bs-3);
+ off = idx & ((1<<(bs-3))-1);
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (tbl->zt_blk + blk) << bs, FTAG, &db);
+ if (err)
+ return (err);
+ dmu_buf_will_dirty(db, tx);
+
+ if (tbl->zt_nextblk != 0) {
+ uint64_t idx2 = idx * 2;
+ uint64_t blk2 = idx2 >> (bs-3);
+ uint64_t off2 = idx2 & ((1<<(bs-3))-1);
+ dmu_buf_t *db2;
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (tbl->zt_nextblk + blk2) << bs, FTAG, &db2);
+ if (err) {
+ dmu_buf_rele(db, FTAG);
+ return (err);
+ }
+ dmu_buf_will_dirty(db2, tx);
+ ((uint64_t *)db2->db_data)[off2] = val;
+ ((uint64_t *)db2->db_data)[off2+1] = val;
+ dmu_buf_rele(db2, FTAG);
+ }
+
+ ((uint64_t *)db->db_data)[off] = val;
+ dmu_buf_rele(db, FTAG);
+
+ return (0);
+}
+
+static int
+zap_table_load(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t *valp)
+{
+ uint64_t blk, off;
+ int err;
+ dmu_buf_t *db;
+ int bs = FZAP_BLOCK_SHIFT(zap);
+
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+
+ blk = idx >> (bs-3);
+ off = idx & ((1<<(bs-3))-1);
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (tbl->zt_blk + blk) << bs, FTAG, &db);
+ if (err)
+ return (err);
+ *valp = ((uint64_t *)db->db_data)[off];
+ dmu_buf_rele(db, FTAG);
+
+ if (tbl->zt_nextblk != 0) {
+ /*
+ * read the nextblk for the sake of i/o error checking,
+ * so that zap_table_load() will catch errors for
+ * zap_table_store.
+ */
+ blk = (idx*2) >> (bs-3);
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (tbl->zt_nextblk + blk) << bs, FTAG, &db);
+ dmu_buf_rele(db, FTAG);
+ }
+ return (err);
+}
+
+/*
+ * Routines for growing the ptrtbl.
+ */
+
+static void
+zap_ptrtbl_transfer(const uint64_t *src, uint64_t *dst, int n)
+{
+ int i;
+ for (i = 0; i < n; i++) {
+ uint64_t lb = src[i];
+ dst[2*i+0] = lb;
+ dst[2*i+1] = lb;
+ }
+}
+
+static int
+zap_grow_ptrtbl(zap_t *zap, dmu_tx_t *tx)
+{
+ /* In case things go horribly wrong. */
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_shift >= ZAP_HASHBITS-2)
+ return (ENOSPC);
+
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks == 0) {
+ /*
+ * We are outgrowing the "embedded" ptrtbl (the one
+ * stored in the header block). Give it its own entire
+ * block, which will double the size of the ptrtbl.
+ */
+ uint64_t newblk;
+ dmu_buf_t *db_new;
+ int err;
+
+ ASSERT3U(zap->zap_f.zap_phys->zap_ptrtbl.zt_shift, ==,
+ ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
+ ASSERT3U(zap->zap_f.zap_phys->zap_ptrtbl.zt_blk, ==, 0);
+
+ newblk = zap_allocate_blocks(zap, 1);
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ newblk << FZAP_BLOCK_SHIFT(zap), FTAG, &db_new);
+ if (err)
+ return (err);
+ dmu_buf_will_dirty(db_new, tx);
+ zap_ptrtbl_transfer(&ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
+ db_new->db_data, 1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
+ dmu_buf_rele(db_new, FTAG);
+
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_blk = newblk;
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks = 1;
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_shift++;
+
+ ASSERT3U(1ULL << zap->zap_f.zap_phys->zap_ptrtbl.zt_shift, ==,
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks <<
+ (FZAP_BLOCK_SHIFT(zap)-3));
+
+ return (0);
+ } else {
+ return (zap_table_grow(zap, &zap->zap_f.zap_phys->zap_ptrtbl,
+ zap_ptrtbl_transfer, tx));
+ }
+}
+
+static void
+zap_increment_num_entries(zap_t *zap, int delta, dmu_tx_t *tx)
+{
+ dmu_buf_will_dirty(zap->zap_dbuf, tx);
+ mutex_enter(&zap->zap_f.zap_num_entries_mtx);
+ ASSERT(delta > 0 || zap->zap_f.zap_phys->zap_num_entries >= -delta);
+ zap->zap_f.zap_phys->zap_num_entries += delta;
+ mutex_exit(&zap->zap_f.zap_num_entries_mtx);
+}
+
+static uint64_t
+zap_allocate_blocks(zap_t *zap, int nblocks)
+{
+ uint64_t newblk;
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+ newblk = zap->zap_f.zap_phys->zap_freeblk;
+ zap->zap_f.zap_phys->zap_freeblk += nblocks;
+ return (newblk);
+}
+
+static zap_leaf_t *
+zap_create_leaf(zap_t *zap, dmu_tx_t *tx)
+{
+ void *winner;
+ zap_leaf_t *l = kmem_alloc(sizeof (zap_leaf_t), KM_SLEEP);
+
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+
+ rw_init(&l->l_rwlock, 0, 0, 0);
+ rw_enter(&l->l_rwlock, RW_WRITER);
+ l->l_blkid = zap_allocate_blocks(zap, 1);
+ l->l_dbuf = NULL;
+ l->l_phys = NULL;
+
+ VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ l->l_blkid << FZAP_BLOCK_SHIFT(zap), NULL, &l->l_dbuf));
+ winner = dmu_buf_set_user(l->l_dbuf, l, &l->l_phys, zap_leaf_pageout);
+ ASSERT(winner == NULL);
+ dmu_buf_will_dirty(l->l_dbuf, tx);
+
+ zap_leaf_init(l, zap->zap_normflags != 0);
+
+ zap->zap_f.zap_phys->zap_num_leafs++;
+
+ return (l);
+}
+
+int
+fzap_count(zap_t *zap, uint64_t *count)
+{
+ ASSERT(!zap->zap_ismicro);
+ mutex_enter(&zap->zap_f.zap_num_entries_mtx); /* unnecessary */
+ *count = zap->zap_f.zap_phys->zap_num_entries;
+ mutex_exit(&zap->zap_f.zap_num_entries_mtx);
+ return (0);
+}
+
+/*
+ * Routines for obtaining zap_leaf_t's
+ */
+
+void
+zap_put_leaf(zap_leaf_t *l)
+{
+ rw_exit(&l->l_rwlock);
+ dmu_buf_rele(l->l_dbuf, NULL);
+}
+
+_NOTE(ARGSUSED(0))
+static void
+zap_leaf_pageout(dmu_buf_t *db, void *vl)
+{
+ zap_leaf_t *l = vl;
+
+ rw_destroy(&l->l_rwlock);
+ kmem_free(l, sizeof (zap_leaf_t));
+}
+
+static zap_leaf_t *
+zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
+{
+ zap_leaf_t *l, *winner;
+
+ ASSERT(blkid != 0);
+
+ l = kmem_alloc(sizeof (zap_leaf_t), KM_SLEEP);
+ rw_init(&l->l_rwlock, 0, 0, 0);
+ rw_enter(&l->l_rwlock, RW_WRITER);
+ l->l_blkid = blkid;
+ l->l_bs = highbit(db->db_size)-1;
+ l->l_dbuf = db;
+ l->l_phys = NULL;
+
+ winner = dmu_buf_set_user(db, l, &l->l_phys, zap_leaf_pageout);
+
+ rw_exit(&l->l_rwlock);
+ if (winner != NULL) {
+ /* someone else set it first */
+ zap_leaf_pageout(NULL, l);
+ l = winner;
+ }
+
+ /*
+ * lhr_pad was previously used for the next leaf in the leaf
+ * chain. There should be no chained leafs (as we have removed
+ * support for them).
+ */
+ ASSERT3U(l->l_phys->l_hdr.lh_pad1, ==, 0);
+
+ /*
+ * There should be more hash entries than there can be
+ * chunks to put in the hash table
+ */
+ ASSERT3U(ZAP_LEAF_HASH_NUMENTRIES(l), >, ZAP_LEAF_NUMCHUNKS(l) / 3);
+
+ /* The chunks should begin at the end of the hash table */
+ ASSERT3P(&ZAP_LEAF_CHUNK(l, 0), ==,
+ &l->l_phys->l_hash[ZAP_LEAF_HASH_NUMENTRIES(l)]);
+
+ /* The chunks should end at the end of the block */
+ ASSERT3U((uintptr_t)&ZAP_LEAF_CHUNK(l, ZAP_LEAF_NUMCHUNKS(l)) -
+ (uintptr_t)l->l_phys, ==, l->l_dbuf->db_size);
+
+ return (l);
+}
+
+static int
+zap_get_leaf_byblk(zap_t *zap, uint64_t blkid, dmu_tx_t *tx, krw_t lt,
+ zap_leaf_t **lp)
+{
+ dmu_buf_t *db;
+ zap_leaf_t *l;
+ int bs = FZAP_BLOCK_SHIFT(zap);
+ int err;
+
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ blkid << bs, NULL, &db);
+ if (err)
+ return (err);
+
+ ASSERT3U(db->db_object, ==, zap->zap_object);
+ ASSERT3U(db->db_offset, ==, blkid << bs);
+ ASSERT3U(db->db_size, ==, 1 << bs);
+ ASSERT(blkid != 0);
+
+ l = dmu_buf_get_user(db);
+
+ if (l == NULL)
+ l = zap_open_leaf(blkid, db);
+
+ rw_enter(&l->l_rwlock, lt);
+ /*
+ * Must lock before dirtying, otherwise l->l_phys could change,
+ * causing ASSERT below to fail.
+ */
+ if (lt == RW_WRITER)
+ dmu_buf_will_dirty(db, tx);
+ ASSERT3U(l->l_blkid, ==, blkid);
+ ASSERT3P(l->l_dbuf, ==, db);
+ ASSERT3P(l->l_phys, ==, l->l_dbuf->db_data);
+ ASSERT3U(l->l_phys->l_hdr.lh_block_type, ==, ZBT_LEAF);
+ ASSERT3U(l->l_phys->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);
+
+ *lp = l;
+ return (0);
+}
+
+static int
+zap_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t *valp)
+{
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks == 0) {
+ ASSERT3U(idx, <,
+ (1ULL << zap->zap_f.zap_phys->zap_ptrtbl.zt_shift));
+ *valp = ZAP_EMBEDDED_PTRTBL_ENT(zap, idx);
+ return (0);
+ } else {
+ return (zap_table_load(zap, &zap->zap_f.zap_phys->zap_ptrtbl,
+ idx, valp));
+ }
+}
+
+static int
+zap_set_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t blk, dmu_tx_t *tx)
+{
+ ASSERT(tx != NULL);
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_blk == 0) {
+ ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) = blk;
+ return (0);
+ } else {
+ return (zap_table_store(zap, &zap->zap_f.zap_phys->zap_ptrtbl,
+ idx, blk, tx));
+ }
+}
+
+static int
+zap_deref_leaf(zap_t *zap, uint64_t h, dmu_tx_t *tx, krw_t lt, zap_leaf_t **lp)
+{
+ uint64_t idx, blk;
+ int err;
+
+ ASSERT(zap->zap_dbuf == NULL ||
+ zap->zap_f.zap_phys == zap->zap_dbuf->db_data);
+ ASSERT3U(zap->zap_f.zap_phys->zap_magic, ==, ZAP_MAGIC);
+ idx = ZAP_HASH_IDX(h, zap->zap_f.zap_phys->zap_ptrtbl.zt_shift);
+ err = zap_idx_to_blk(zap, idx, &blk);
+ if (err != 0)
+ return (err);
+ err = zap_get_leaf_byblk(zap, blk, tx, lt, lp);
+
+ ASSERT(err || ZAP_HASH_IDX(h, (*lp)->l_phys->l_hdr.lh_prefix_len) ==
+ (*lp)->l_phys->l_hdr.lh_prefix);
+ return (err);
+}
+
+static int
+zap_expand_leaf(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx, zap_leaf_t **lp)
+{
+ zap_t *zap = zn->zn_zap;
+ uint64_t hash = zn->zn_hash;
+ zap_leaf_t *nl;
+ int prefix_diff, i, err;
+ uint64_t sibling;
+ int old_prefix_len = l->l_phys->l_hdr.lh_prefix_len;
+
+ ASSERT3U(old_prefix_len, <=, zap->zap_f.zap_phys->zap_ptrtbl.zt_shift);
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+
+ ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
+ l->l_phys->l_hdr.lh_prefix);
+
+ if (zap_tryupgradedir(zap, tx) == 0 ||
+ old_prefix_len == zap->zap_f.zap_phys->zap_ptrtbl.zt_shift) {
+ /* We failed to upgrade, or need to grow the pointer table */
+ objset_t *os = zap->zap_objset;
+ uint64_t object = zap->zap_object;
+
+ zap_put_leaf(l);
+ zap_unlockdir(zap);
+ err = zap_lockdir(os, object, tx, RW_WRITER,
+ FALSE, FALSE, &zn->zn_zap);
+ zap = zn->zn_zap;
+ if (err)
+ return (err);
+ ASSERT(!zap->zap_ismicro);
+
+ while (old_prefix_len ==
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_shift) {
+ err = zap_grow_ptrtbl(zap, tx);
+ if (err)
+ return (err);
+ }
+
+ err = zap_deref_leaf(zap, hash, tx, RW_WRITER, &l);
+ if (err)
+ return (err);
+
+ if (l->l_phys->l_hdr.lh_prefix_len != old_prefix_len) {
+ /* it split while our locks were down */
+ *lp = l;
+ return (0);
+ }
+ }
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+ ASSERT3U(old_prefix_len, <, zap->zap_f.zap_phys->zap_ptrtbl.zt_shift);
+ ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
+ l->l_phys->l_hdr.lh_prefix);
+
+ prefix_diff = zap->zap_f.zap_phys->zap_ptrtbl.zt_shift -
+ (old_prefix_len + 1);
+ sibling = (ZAP_HASH_IDX(hash, old_prefix_len + 1) | 1) << prefix_diff;
+
+ /* check for i/o errors before doing zap_leaf_split */
+ for (i = 0; i < (1ULL<<prefix_diff); i++) {
+ uint64_t blk;
+ err = zap_idx_to_blk(zap, sibling+i, &blk);
+ if (err)
+ return (err);
+ ASSERT3U(blk, ==, l->l_blkid);
+ }
+
+ nl = zap_create_leaf(zap, tx);
+ zap_leaf_split(l, nl, zap->zap_normflags != 0);
+
+ /* set sibling pointers */
+ for (i = 0; i < (1ULL<<prefix_diff); i++) {
+ err = zap_set_idx_to_blk(zap, sibling+i, nl->l_blkid, tx);
+ ASSERT3U(err, ==, 0); /* we checked for i/o errors above */
+ }
+
+ if (hash & (1ULL << (64 - l->l_phys->l_hdr.lh_prefix_len))) {
+ /* we want the sibling */
+ zap_put_leaf(l);
+ *lp = nl;
+ } else {
+ zap_put_leaf(nl);
+ *lp = l;
+ }
+
+ return (0);
+}
+
+static void
+zap_put_leaf_maybe_grow_ptrtbl(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx)
+{
+ zap_t *zap = zn->zn_zap;
+ int shift = zap->zap_f.zap_phys->zap_ptrtbl.zt_shift;
+ int leaffull = (l->l_phys->l_hdr.lh_prefix_len == shift &&
+ l->l_phys->l_hdr.lh_nfree < ZAP_LEAF_LOW_WATER);
+
+ zap_put_leaf(l);
+
+ if (leaffull || zap->zap_f.zap_phys->zap_ptrtbl.zt_nextblk) {
+ int err;
+
+ /*
+ * We are in the middle of growing the pointer table, or
+ * this leaf will soon make us grow it.
+ */
+ if (zap_tryupgradedir(zap, tx) == 0) {
+ objset_t *os = zap->zap_objset;
+ uint64_t zapobj = zap->zap_object;
+
+ zap_unlockdir(zap);
+ err = zap_lockdir(os, zapobj, tx,
+ RW_WRITER, FALSE, FALSE, &zn->zn_zap);
+ zap = zn->zn_zap;
+ if (err)
+ return;
+ }
+
+ /* could have finished growing while our locks were down */
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_shift == shift)
+ (void) zap_grow_ptrtbl(zap, tx);
+ }
+}
+
+
+static int
+fzap_checksize(const char *name, uint64_t integer_size, uint64_t num_integers)
+{
+ if (name && strlen(name) > ZAP_MAXNAMELEN)
+ return (E2BIG);
+
+ /* Only integer sizes supported by C */
+ switch (integer_size) {
+ case 1:
+ case 2:
+ case 4:
+ case 8:
+ break;
+ default:
+ return (EINVAL);
+ }
+
+ if (integer_size * num_integers > ZAP_MAXVALUELEN)
+ return (E2BIG);
+
+ return (0);
+}
+
+/*
+ * Routines for manipulating attributes.
+ */
+int
+fzap_lookup(zap_name_t *zn,
+ uint64_t integer_size, uint64_t num_integers, void *buf,
+ char *realname, int rn_len, boolean_t *ncp)
+{
+ zap_leaf_t *l;
+ int err;
+ zap_entry_handle_t zeh;
+
+ err = fzap_checksize(zn->zn_name_orij, integer_size, num_integers);
+ if (err != 0)
+ return (err);
+
+ err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
+ if (err != 0)
+ return (err);
+ err = zap_leaf_lookup(l, zn, &zeh);
+ if (err == 0) {
+ err = zap_entry_read(&zeh, integer_size, num_integers, buf);
+ (void) zap_entry_read_name(&zeh, rn_len, realname);
+ if (ncp) {
+ *ncp = zap_entry_normalization_conflict(&zeh,
+ zn, NULL, zn->zn_zap);
+ }
+ }
+
+ zap_put_leaf(l);
+ return (err);
+}
+
+int
+fzap_add_cd(zap_name_t *zn,
+ uint64_t integer_size, uint64_t num_integers,
+ const void *val, uint32_t cd, dmu_tx_t *tx)
+{
+ zap_leaf_t *l;
+ int err;
+ zap_entry_handle_t zeh;
+ zap_t *zap = zn->zn_zap;
+
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+ ASSERT(!zap->zap_ismicro);
+ ASSERT(fzap_checksize(zn->zn_name_orij,
+ integer_size, num_integers) == 0);
+
+ err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
+ if (err != 0)
+ return (err);
+retry:
+ err = zap_leaf_lookup(l, zn, &zeh);
+ if (err == 0) {
+ err = EEXIST;
+ goto out;
+ }
+ if (err != ENOENT)
+ goto out;
+
+ err = zap_entry_create(l, zn->zn_name_orij, zn->zn_hash, cd,
+ integer_size, num_integers, val, &zeh);
+
+ if (err == 0) {
+ zap_increment_num_entries(zap, 1, tx);
+ } else if (err == EAGAIN) {
+ err = zap_expand_leaf(zn, l, tx, &l);
+ zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
+ if (err == 0)
+ goto retry;
+ }
+
+out:
+ if (zap != NULL)
+ zap_put_leaf_maybe_grow_ptrtbl(zn, l, tx);
+ return (err);
+}
+
+int
+fzap_add(zap_name_t *zn,
+ uint64_t integer_size, uint64_t num_integers,
+ const void *val, dmu_tx_t *tx)
+{
+ int err = fzap_checksize(zn->zn_name_orij, integer_size, num_integers);
+ if (err != 0)
+ return (err);
+
+ return (fzap_add_cd(zn, integer_size, num_integers,
+ val, ZAP_MAXCD, tx));
+}
+
+int
+fzap_update(zap_name_t *zn,
+ int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx)
+{
+ zap_leaf_t *l;
+ int err, create;
+ zap_entry_handle_t zeh;
+ zap_t *zap = zn->zn_zap;
+
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+ err = fzap_checksize(zn->zn_name_orij, integer_size, num_integers);
+ if (err != 0)
+ return (err);
+
+ err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
+ if (err != 0)
+ return (err);
+retry:
+ err = zap_leaf_lookup(l, zn, &zeh);
+ create = (err == ENOENT);
+ ASSERT(err == 0 || err == ENOENT);
+
+ if (create) {
+ err = zap_entry_create(l, zn->zn_name_orij, zn->zn_hash,
+ ZAP_MAXCD, integer_size, num_integers, val, &zeh);
+ if (err == 0)
+ zap_increment_num_entries(zap, 1, tx);
+ } else {
+ err = zap_entry_update(&zeh, integer_size, num_integers, val);
+ }
+
+ if (err == EAGAIN) {
+ err = zap_expand_leaf(zn, l, tx, &l);
+ zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
+ if (err == 0)
+ goto retry;
+ }
+
+ if (zap != NULL)
+ zap_put_leaf_maybe_grow_ptrtbl(zn, l, tx);
+ return (err);
+}
+
+int
+fzap_length(zap_name_t *zn,
+ uint64_t *integer_size, uint64_t *num_integers)
+{
+ zap_leaf_t *l;
+ int err;
+ zap_entry_handle_t zeh;
+
+ err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
+ if (err != 0)
+ return (err);
+ err = zap_leaf_lookup(l, zn, &zeh);
+ if (err != 0)
+ goto out;
+
+ if (integer_size)
+ *integer_size = zeh.zeh_integer_size;
+ if (num_integers)
+ *num_integers = zeh.zeh_num_integers;
+out:
+ zap_put_leaf(l);
+ return (err);
+}
+
+int
+fzap_remove(zap_name_t *zn, dmu_tx_t *tx)
+{
+ zap_leaf_t *l;
+ int err;
+ zap_entry_handle_t zeh;
+
+ err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, tx, RW_WRITER, &l);
+ if (err != 0)
+ return (err);
+ err = zap_leaf_lookup(l, zn, &zeh);
+ if (err == 0) {
+ zap_entry_remove(&zeh);
+ zap_increment_num_entries(zn->zn_zap, -1, tx);
+ }
+ zap_put_leaf(l);
+ return (err);
+}
+
+/*
+ * Helper functions for consumers.
+ */
+
+int
+zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask,
+ char *name)
+{
+ zap_cursor_t zc;
+ zap_attribute_t *za;
+ int err;
+
+ if (mask == 0)
+ mask = -1ULL;
+
+ za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
+ for (zap_cursor_init(&zc, os, zapobj);
+ (err = zap_cursor_retrieve(&zc, za)) == 0;
+ zap_cursor_advance(&zc)) {
+ if ((za->za_first_integer & mask) == (value & mask)) {
+ (void) strcpy(name, za->za_name);
+ break;
+ }
+ }
+ zap_cursor_fini(&zc);
+ kmem_free(za, sizeof (zap_attribute_t));
+ return (err);
+}
+
+int
+zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx)
+{
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ int err;
+
+ for (zap_cursor_init(&zc, os, fromobj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ (void) zap_cursor_advance(&zc)) {
+ if (za.za_integer_length != 8 || za.za_num_integers != 1)
+ return (EINVAL);
+ err = zap_add(os, intoobj, za.za_name,
+ 8, 1, &za.za_first_integer, tx);
+ if (err)
+ return (err);
+ }
+ zap_cursor_fini(&zc);
+ return (0);
+}
+
+int
+zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
+{
+ char name[20];
+
+ (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
+ return (zap_add(os, obj, name, 8, 1, &value, tx));
+}
+
+int
+zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
+{
+ char name[20];
+
+ (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
+ return (zap_remove(os, obj, name, tx));
+}
+
+int
+zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value)
+{
+ char name[20];
+
+ (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
+ return (zap_lookup(os, obj, name, 8, 1, &value));
+}
+
+/*
+ * Routines for iterating over the attributes.
+ */
+
+int
+fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za)
+{
+ int err = ENOENT;
+ zap_entry_handle_t zeh;
+ zap_leaf_t *l;
+
+ /* retrieve the next entry at or after zc_hash/zc_cd */
+ /* if no entry, return ENOENT */
+
+ if (zc->zc_leaf &&
+ (ZAP_HASH_IDX(zc->zc_hash,
+ zc->zc_leaf->l_phys->l_hdr.lh_prefix_len) !=
+ zc->zc_leaf->l_phys->l_hdr.lh_prefix)) {
+ rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);
+ zap_put_leaf(zc->zc_leaf);
+ zc->zc_leaf = NULL;
+ }
+
+again:
+ if (zc->zc_leaf == NULL) {
+ err = zap_deref_leaf(zap, zc->zc_hash, NULL, RW_READER,
+ &zc->zc_leaf);
+ if (err != 0)
+ return (err);
+ } else {
+ rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);
+ }
+ l = zc->zc_leaf;
+
+ err = zap_leaf_lookup_closest(l, zc->zc_hash, zc->zc_cd, &zeh);
+
+ if (err == ENOENT) {
+ uint64_t nocare =
+ (1ULL << (64 - l->l_phys->l_hdr.lh_prefix_len)) - 1;
+ zc->zc_hash = (zc->zc_hash & ~nocare) + nocare + 1;
+ zc->zc_cd = 0;
+ if (l->l_phys->l_hdr.lh_prefix_len == 0 || zc->zc_hash == 0) {
+ zc->zc_hash = -1ULL;
+ } else {
+ zap_put_leaf(zc->zc_leaf);
+ zc->zc_leaf = NULL;
+ goto again;
+ }
+ }
+
+ if (err == 0) {
+ zc->zc_hash = zeh.zeh_hash;
+ zc->zc_cd = zeh.zeh_cd;
+ za->za_integer_length = zeh.zeh_integer_size;
+ za->za_num_integers = zeh.zeh_num_integers;
+ if (zeh.zeh_num_integers == 0) {
+ za->za_first_integer = 0;
+ } else {
+ err = zap_entry_read(&zeh, 8, 1, &za->za_first_integer);
+ ASSERT(err == 0 || err == EOVERFLOW);
+ }
+ err = zap_entry_read_name(&zeh,
+ sizeof (za->za_name), za->za_name);
+ ASSERT(err == 0);
+
+ za->za_normalization_conflict =
+ zap_entry_normalization_conflict(&zeh,
+ NULL, za->za_name, zap);
+ }
+ rw_exit(&zc->zc_leaf->l_rwlock);
+ return (err);
+}
+
+
+static void
+zap_stats_ptrtbl(zap_t *zap, uint64_t *tbl, int len, zap_stats_t *zs)
+{
+ int i, err;
+ uint64_t lastblk = 0;
+
+ /*
+ * NB: if a leaf has more pointers than an entire ptrtbl block
+ * can hold, then it'll be accounted for more than once, since
+ * we won't have lastblk.
+ */
+ for (i = 0; i < len; i++) {
+ zap_leaf_t *l;
+
+ if (tbl[i] == lastblk)
+ continue;
+ lastblk = tbl[i];
+
+ err = zap_get_leaf_byblk(zap, tbl[i], NULL, RW_READER, &l);
+ if (err == 0) {
+ zap_leaf_stats(zap, l, zs);
+ zap_put_leaf(l);
+ }
+ }
+}
+
+void
+fzap_get_stats(zap_t *zap, zap_stats_t *zs)
+{
+ int bs = FZAP_BLOCK_SHIFT(zap);
+ zs->zs_blocksize = 1ULL << bs;
+
+ /*
+ * Set zap_phys_t fields
+ */
+ zs->zs_num_leafs = zap->zap_f.zap_phys->zap_num_leafs;
+ zs->zs_num_entries = zap->zap_f.zap_phys->zap_num_entries;
+ zs->zs_num_blocks = zap->zap_f.zap_phys->zap_freeblk;
+ zs->zs_block_type = zap->zap_f.zap_phys->zap_block_type;
+ zs->zs_magic = zap->zap_f.zap_phys->zap_magic;
+ zs->zs_salt = zap->zap_f.zap_phys->zap_salt;
+
+ /*
+ * Set zap_ptrtbl fields
+ */
+ zs->zs_ptrtbl_len = 1ULL << zap->zap_f.zap_phys->zap_ptrtbl.zt_shift;
+ zs->zs_ptrtbl_nextblk = zap->zap_f.zap_phys->zap_ptrtbl.zt_nextblk;
+ zs->zs_ptrtbl_blks_copied =
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_blks_copied;
+ zs->zs_ptrtbl_zt_blk = zap->zap_f.zap_phys->zap_ptrtbl.zt_blk;
+ zs->zs_ptrtbl_zt_numblks = zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks;
+ zs->zs_ptrtbl_zt_shift = zap->zap_f.zap_phys->zap_ptrtbl.zt_shift;
+
+ if (zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks == 0) {
+ /* the ptrtbl is entirely in the header block. */
+ zap_stats_ptrtbl(zap, &ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
+ 1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap), zs);
+ } else {
+ int b;
+
+ dmu_prefetch(zap->zap_objset, zap->zap_object,
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_blk << bs,
+ zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks << bs);
+
+ for (b = 0; b < zap->zap_f.zap_phys->zap_ptrtbl.zt_numblks;
+ b++) {
+ dmu_buf_t *db;
+ int err;
+
+ err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
+ (zap->zap_f.zap_phys->zap_ptrtbl.zt_blk + b) << bs,
+ FTAG, &db);
+ if (err == 0) {
+ zap_stats_ptrtbl(zap, db->db_data,
+ 1<<(bs-3), zs);
+ dmu_buf_rele(db, FTAG);
+ }
+ }
+ }
+}
diff --git a/module/zfs/zap_leaf.c b/module/zfs/zap_leaf.c
new file mode 100644
index 000000000..da498b6bc
--- /dev/null
+++ b/module/zfs/zap_leaf.c
@@ -0,0 +1,853 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * The 512-byte leaf is broken into 32 16-byte chunks.
+ * chunk number n means l_chunk[n], even though the header precedes it.
+ * the names are stored null-terminated.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/zap.h>
+#include <sys/zap_impl.h>
+#include <sys/zap_leaf.h>
+#include <sys/spa.h>
+#include <sys/dmu.h>
+
+static uint16_t *zap_leaf_rehash_entry(zap_leaf_t *l, uint16_t entry);
+
+#define CHAIN_END 0xffff /* end of the chunk chain */
+
+/* half the (current) minimum block size */
+#define MAX_ARRAY_BYTES (8<<10)
+
+#define LEAF_HASH(l, h) \
+ ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \
+ ((h) >> (64 - ZAP_LEAF_HASH_SHIFT(l)-(l)->l_phys->l_hdr.lh_prefix_len)))
+
+#define LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[LEAF_HASH(l, h)])
+
+
+static void
+zap_memset(void *a, int c, size_t n)
+{
+ char *cp = a;
+ char *cpend = cp + n;
+
+ while (cp < cpend)
+ *cp++ = c;
+}
+
+static void
+stv(int len, void *addr, uint64_t value)
+{
+ switch (len) {
+ case 1:
+ *(uint8_t *)addr = value;
+ return;
+ case 2:
+ *(uint16_t *)addr = value;
+ return;
+ case 4:
+ *(uint32_t *)addr = value;
+ return;
+ case 8:
+ *(uint64_t *)addr = value;
+ return;
+ }
+ ASSERT(!"bad int len");
+}
+
+static uint64_t
+ldv(int len, const void *addr)
+{
+ switch (len) {
+ case 1:
+ return (*(uint8_t *)addr);
+ case 2:
+ return (*(uint16_t *)addr);
+ case 4:
+ return (*(uint32_t *)addr);
+ case 8:
+ return (*(uint64_t *)addr);
+ }
+ ASSERT(!"bad int len");
+ return (0xFEEDFACEDEADBEEFULL);
+}
+
+void
+zap_leaf_byteswap(zap_leaf_phys_t *buf, int size)
+{
+ int i;
+ zap_leaf_t l;
+ l.l_bs = highbit(size)-1;
+ l.l_phys = buf;
+
+ buf->l_hdr.lh_block_type = BSWAP_64(buf->l_hdr.lh_block_type);
+ buf->l_hdr.lh_prefix = BSWAP_64(buf->l_hdr.lh_prefix);
+ buf->l_hdr.lh_magic = BSWAP_32(buf->l_hdr.lh_magic);
+ buf->l_hdr.lh_nfree = BSWAP_16(buf->l_hdr.lh_nfree);
+ buf->l_hdr.lh_nentries = BSWAP_16(buf->l_hdr.lh_nentries);
+ buf->l_hdr.lh_prefix_len = BSWAP_16(buf->l_hdr.lh_prefix_len);
+ buf->l_hdr.lh_freelist = BSWAP_16(buf->l_hdr.lh_freelist);
+
+ for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++)
+ buf->l_hash[i] = BSWAP_16(buf->l_hash[i]);
+
+ for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
+ zap_leaf_chunk_t *lc = &ZAP_LEAF_CHUNK(&l, i);
+ struct zap_leaf_entry *le;
+
+ switch (lc->l_free.lf_type) {
+ case ZAP_CHUNK_ENTRY:
+ le = &lc->l_entry;
+
+ le->le_type = BSWAP_8(le->le_type);
+ le->le_int_size = BSWAP_8(le->le_int_size);
+ le->le_next = BSWAP_16(le->le_next);
+ le->le_name_chunk = BSWAP_16(le->le_name_chunk);
+ le->le_name_length = BSWAP_16(le->le_name_length);
+ le->le_value_chunk = BSWAP_16(le->le_value_chunk);
+ le->le_value_length = BSWAP_16(le->le_value_length);
+ le->le_cd = BSWAP_32(le->le_cd);
+ le->le_hash = BSWAP_64(le->le_hash);
+ break;
+ case ZAP_CHUNK_FREE:
+ lc->l_free.lf_type = BSWAP_8(lc->l_free.lf_type);
+ lc->l_free.lf_next = BSWAP_16(lc->l_free.lf_next);
+ break;
+ case ZAP_CHUNK_ARRAY:
+ lc->l_array.la_type = BSWAP_8(lc->l_array.la_type);
+ lc->l_array.la_next = BSWAP_16(lc->l_array.la_next);
+ /* la_array doesn't need swapping */
+ break;
+ default:
+ ASSERT(!"bad leaf type");
+ }
+ }
+}
+
+void
+zap_leaf_init(zap_leaf_t *l, boolean_t sort)
+{
+ int i;
+
+ l->l_bs = highbit(l->l_dbuf->db_size)-1;
+ zap_memset(&l->l_phys->l_hdr, 0, sizeof (struct zap_leaf_header));
+ zap_memset(l->l_phys->l_hash, CHAIN_END, 2*ZAP_LEAF_HASH_NUMENTRIES(l));
+ for (i = 0; i < ZAP_LEAF_NUMCHUNKS(l); i++) {
+ ZAP_LEAF_CHUNK(l, i).l_free.lf_type = ZAP_CHUNK_FREE;
+ ZAP_LEAF_CHUNK(l, i).l_free.lf_next = i+1;
+ }
+ ZAP_LEAF_CHUNK(l, ZAP_LEAF_NUMCHUNKS(l)-1).l_free.lf_next = CHAIN_END;
+ l->l_phys->l_hdr.lh_block_type = ZBT_LEAF;
+ l->l_phys->l_hdr.lh_magic = ZAP_LEAF_MAGIC;
+ l->l_phys->l_hdr.lh_nfree = ZAP_LEAF_NUMCHUNKS(l);
+ if (sort)
+ l->l_phys->l_hdr.lh_flags |= ZLF_ENTRIES_CDSORTED;
+}
+
+/*
+ * Routines which manipulate leaf chunks (l_chunk[]).
+ */
+
+static uint16_t
+zap_leaf_chunk_alloc(zap_leaf_t *l)
+{
+ int chunk;
+
+ ASSERT(l->l_phys->l_hdr.lh_nfree > 0);
+
+ chunk = l->l_phys->l_hdr.lh_freelist;
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT3U(ZAP_LEAF_CHUNK(l, chunk).l_free.lf_type, ==, ZAP_CHUNK_FREE);
+
+ l->l_phys->l_hdr.lh_freelist = ZAP_LEAF_CHUNK(l, chunk).l_free.lf_next;
+
+ l->l_phys->l_hdr.lh_nfree--;
+
+ return (chunk);
+}
+
+static void
+zap_leaf_chunk_free(zap_leaf_t *l, uint16_t chunk)
+{
+ struct zap_leaf_free *zlf = &ZAP_LEAF_CHUNK(l, chunk).l_free;
+ ASSERT3U(l->l_phys->l_hdr.lh_nfree, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT(zlf->lf_type != ZAP_CHUNK_FREE);
+
+ zlf->lf_type = ZAP_CHUNK_FREE;
+ zlf->lf_next = l->l_phys->l_hdr.lh_freelist;
+ bzero(zlf->lf_pad, sizeof (zlf->lf_pad)); /* help it to compress */
+ l->l_phys->l_hdr.lh_freelist = chunk;
+
+ l->l_phys->l_hdr.lh_nfree++;
+}
+
+/*
+ * Routines which manipulate leaf arrays (zap_leaf_array type chunks).
+ */
+
+static uint16_t
+zap_leaf_array_create(zap_leaf_t *l, const char *buf,
+ int integer_size, int num_integers)
+{
+ uint16_t chunk_head;
+ uint16_t *chunkp = &chunk_head;
+ int byten = 0;
+ uint64_t value;
+ int shift = (integer_size-1)*8;
+ int len = num_integers;
+
+ ASSERT3U(num_integers * integer_size, <, MAX_ARRAY_BYTES);
+
+ while (len > 0) {
+ uint16_t chunk = zap_leaf_chunk_alloc(l);
+ struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ int i;
+
+ la->la_type = ZAP_CHUNK_ARRAY;
+ for (i = 0; i < ZAP_LEAF_ARRAY_BYTES; i++) {
+ if (byten == 0)
+ value = ldv(integer_size, buf);
+ la->la_array[i] = value >> shift;
+ value <<= 8;
+ if (++byten == integer_size) {
+ byten = 0;
+ buf += integer_size;
+ if (--len == 0)
+ break;
+ }
+ }
+
+ *chunkp = chunk;
+ chunkp = &la->la_next;
+ }
+ *chunkp = CHAIN_END;
+
+ return (chunk_head);
+}
+
+static void
+zap_leaf_array_free(zap_leaf_t *l, uint16_t *chunkp)
+{
+ uint16_t chunk = *chunkp;
+
+ *chunkp = CHAIN_END;
+
+ while (chunk != CHAIN_END) {
+ int nextchunk = ZAP_LEAF_CHUNK(l, chunk).l_array.la_next;
+ ASSERT3U(ZAP_LEAF_CHUNK(l, chunk).l_array.la_type, ==,
+ ZAP_CHUNK_ARRAY);
+ zap_leaf_chunk_free(l, chunk);
+ chunk = nextchunk;
+ }
+}
+
+/* array_len and buf_len are in integers, not bytes */
+static void
+zap_leaf_array_read(zap_leaf_t *l, uint16_t chunk,
+ int array_int_len, int array_len, int buf_int_len, uint64_t buf_len,
+ char *buf)
+{
+ int len = MIN(array_len, buf_len);
+ int byten = 0;
+ uint64_t value = 0;
+
+ ASSERT3U(array_int_len, <=, buf_int_len);
+
+ /* Fast path for one 8-byte integer */
+ if (array_int_len == 8 && buf_int_len == 8 && len == 1) {
+ struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ uint8_t *ip = la->la_array;
+ uint64_t *buf64 = (uint64_t *)buf;
+
+ *buf64 = (uint64_t)ip[0] << 56 | (uint64_t)ip[1] << 48 |
+ (uint64_t)ip[2] << 40 | (uint64_t)ip[3] << 32 |
+ (uint64_t)ip[4] << 24 | (uint64_t)ip[5] << 16 |
+ (uint64_t)ip[6] << 8 | (uint64_t)ip[7];
+ return;
+ }
+
+ /* Fast path for an array of 1-byte integers (eg. the entry name) */
+ if (array_int_len == 1 && buf_int_len == 1 &&
+ buf_len > array_len + ZAP_LEAF_ARRAY_BYTES) {
+ while (chunk != CHAIN_END) {
+ struct zap_leaf_array *la =
+ &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ bcopy(la->la_array, buf, ZAP_LEAF_ARRAY_BYTES);
+ buf += ZAP_LEAF_ARRAY_BYTES;
+ chunk = la->la_next;
+ }
+ return;
+ }
+
+ while (len > 0) {
+ struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ int i;
+
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ for (i = 0; i < ZAP_LEAF_ARRAY_BYTES && len > 0; i++) {
+ value = (value << 8) | la->la_array[i];
+ byten++;
+ if (byten == array_int_len) {
+ stv(buf_int_len, buf, value);
+ byten = 0;
+ len--;
+ if (len == 0)
+ return;
+ buf += buf_int_len;
+ }
+ }
+ chunk = la->la_next;
+ }
+}
+
+/*
+ * Only to be used on 8-bit arrays.
+ * array_len is actual len in bytes (not encoded le_value_length).
+ * namenorm is null-terminated.
+ */
+static boolean_t
+zap_leaf_array_match(zap_leaf_t *l, zap_name_t *zn, int chunk, int array_len)
+{
+ int bseen = 0;
+
+ if (zn->zn_matchtype == MT_FIRST) {
+ char *thisname = kmem_alloc(array_len, KM_SLEEP);
+ boolean_t match;
+
+ zap_leaf_array_read(l, chunk, 1, array_len, 1,
+ array_len, thisname);
+ match = zap_match(zn, thisname);
+ kmem_free(thisname, array_len);
+ return (match);
+ }
+
+ /* Fast path for exact matching */
+ while (bseen < array_len) {
+ struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ int toread = MIN(array_len - bseen, ZAP_LEAF_ARRAY_BYTES);
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ if (bcmp(la->la_array, zn->zn_name_orij + bseen, toread))
+ break;
+ chunk = la->la_next;
+ bseen += toread;
+ }
+ return (bseen == array_len);
+}
+
+/*
+ * Routines which manipulate leaf entries.
+ */
+
+int
+zap_leaf_lookup(zap_leaf_t *l, zap_name_t *zn, zap_entry_handle_t *zeh)
+{
+ uint16_t *chunkp;
+ struct zap_leaf_entry *le;
+
+ ASSERT3U(l->l_phys->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);
+
+again:
+ for (chunkp = LEAF_HASH_ENTPTR(l, zn->zn_hash);
+ *chunkp != CHAIN_END; chunkp = &le->le_next) {
+ uint16_t chunk = *chunkp;
+ le = ZAP_LEAF_ENTRY(l, chunk);
+
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ if (le->le_hash != zn->zn_hash)
+ continue;
+
+ /*
+ * NB: the entry chain is always sorted by cd on
+ * normalized zap objects, so this will find the
+ * lowest-cd match for MT_FIRST.
+ */
+ ASSERT(zn->zn_matchtype == MT_EXACT ||
+ (l->l_phys->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED));
+ if (zap_leaf_array_match(l, zn, le->le_name_chunk,
+ le->le_name_length)) {
+ zeh->zeh_num_integers = le->le_value_length;
+ zeh->zeh_integer_size = le->le_int_size;
+ zeh->zeh_cd = le->le_cd;
+ zeh->zeh_hash = le->le_hash;
+ zeh->zeh_chunkp = chunkp;
+ zeh->zeh_leaf = l;
+ return (0);
+ }
+ }
+
+ /*
+ * NB: we could of course do this in one pass, but that would be
+ * a pain. We'll see if MT_BEST is even used much.
+ */
+ if (zn->zn_matchtype == MT_BEST) {
+ zn->zn_matchtype = MT_FIRST;
+ goto again;
+ }
+
+ return (ENOENT);
+}
+
+/* Return (h1,cd1 >= h2,cd2) */
+#define HCD_GTEQ(h1, cd1, h2, cd2) \
+ ((h1 > h2) ? TRUE : ((h1 == h2 && cd1 >= cd2) ? TRUE : FALSE))
+
+int
+zap_leaf_lookup_closest(zap_leaf_t *l,
+ uint64_t h, uint32_t cd, zap_entry_handle_t *zeh)
+{
+ uint16_t chunk;
+ uint64_t besth = -1ULL;
+ uint32_t bestcd = ZAP_MAXCD;
+ uint16_t bestlh = ZAP_LEAF_HASH_NUMENTRIES(l)-1;
+ uint16_t lh;
+ struct zap_leaf_entry *le;
+
+ ASSERT3U(l->l_phys->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);
+
+ for (lh = LEAF_HASH(l, h); lh <= bestlh; lh++) {
+ for (chunk = l->l_phys->l_hash[lh];
+ chunk != CHAIN_END; chunk = le->le_next) {
+ le = ZAP_LEAF_ENTRY(l, chunk);
+
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ if (HCD_GTEQ(le->le_hash, le->le_cd, h, cd) &&
+ HCD_GTEQ(besth, bestcd, le->le_hash, le->le_cd)) {
+ ASSERT3U(bestlh, >=, lh);
+ bestlh = lh;
+ besth = le->le_hash;
+ bestcd = le->le_cd;
+
+ zeh->zeh_num_integers = le->le_value_length;
+ zeh->zeh_integer_size = le->le_int_size;
+ zeh->zeh_cd = le->le_cd;
+ zeh->zeh_hash = le->le_hash;
+ zeh->zeh_fakechunk = chunk;
+ zeh->zeh_chunkp = &zeh->zeh_fakechunk;
+ zeh->zeh_leaf = l;
+ }
+ }
+ }
+
+ return (bestcd == ZAP_MAXCD ? ENOENT : 0);
+}
+
+int
+zap_entry_read(const zap_entry_handle_t *zeh,
+ uint8_t integer_size, uint64_t num_integers, void *buf)
+{
+ struct zap_leaf_entry *le =
+ ZAP_LEAF_ENTRY(zeh->zeh_leaf, *zeh->zeh_chunkp);
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ if (le->le_int_size > integer_size)
+ return (EINVAL);
+
+ zap_leaf_array_read(zeh->zeh_leaf, le->le_value_chunk, le->le_int_size,
+ le->le_value_length, integer_size, num_integers, buf);
+
+ if (zeh->zeh_num_integers > num_integers)
+ return (EOVERFLOW);
+ return (0);
+
+}
+
+int
+zap_entry_read_name(const zap_entry_handle_t *zeh, uint16_t buflen, char *buf)
+{
+ struct zap_leaf_entry *le =
+ ZAP_LEAF_ENTRY(zeh->zeh_leaf, *zeh->zeh_chunkp);
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ zap_leaf_array_read(zeh->zeh_leaf, le->le_name_chunk, 1,
+ le->le_name_length, 1, buflen, buf);
+ if (le->le_name_length > buflen)
+ return (EOVERFLOW);
+ return (0);
+}
+
+int
+zap_entry_update(zap_entry_handle_t *zeh,
+ uint8_t integer_size, uint64_t num_integers, const void *buf)
+{
+ int delta_chunks;
+ zap_leaf_t *l = zeh->zeh_leaf;
+ struct zap_leaf_entry *le = ZAP_LEAF_ENTRY(l, *zeh->zeh_chunkp);
+
+ delta_chunks = ZAP_LEAF_ARRAY_NCHUNKS(num_integers * integer_size) -
+ ZAP_LEAF_ARRAY_NCHUNKS(le->le_value_length * le->le_int_size);
+
+ if ((int)l->l_phys->l_hdr.lh_nfree < delta_chunks)
+ return (EAGAIN);
+
+ /*
+ * We should search other chained leaves (via
+ * zap_entry_remove,create?) otherwise returning EAGAIN will
+ * just send us into an infinite loop if we have to chain
+ * another leaf block, rather than being able to split this
+ * block.
+ */
+
+ zap_leaf_array_free(l, &le->le_value_chunk);
+ le->le_value_chunk =
+ zap_leaf_array_create(l, buf, integer_size, num_integers);
+ le->le_value_length = num_integers;
+ le->le_int_size = integer_size;
+ return (0);
+}
+
+void
+zap_entry_remove(zap_entry_handle_t *zeh)
+{
+ uint16_t entry_chunk;
+ struct zap_leaf_entry *le;
+ zap_leaf_t *l = zeh->zeh_leaf;
+
+ ASSERT3P(zeh->zeh_chunkp, !=, &zeh->zeh_fakechunk);
+
+ entry_chunk = *zeh->zeh_chunkp;
+ le = ZAP_LEAF_ENTRY(l, entry_chunk);
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ zap_leaf_array_free(l, &le->le_name_chunk);
+ zap_leaf_array_free(l, &le->le_value_chunk);
+
+ *zeh->zeh_chunkp = le->le_next;
+ zap_leaf_chunk_free(l, entry_chunk);
+
+ l->l_phys->l_hdr.lh_nentries--;
+}
+
+int
+zap_entry_create(zap_leaf_t *l, const char *name, uint64_t h, uint32_t cd,
+ uint8_t integer_size, uint64_t num_integers, const void *buf,
+ zap_entry_handle_t *zeh)
+{
+ uint16_t chunk;
+ uint16_t *chunkp;
+ struct zap_leaf_entry *le;
+ uint64_t namelen, valuelen;
+ int numchunks;
+
+ valuelen = integer_size * num_integers;
+ namelen = strlen(name) + 1;
+ ASSERT(namelen >= 2);
+
+ numchunks = 1 + ZAP_LEAF_ARRAY_NCHUNKS(namelen) +
+ ZAP_LEAF_ARRAY_NCHUNKS(valuelen);
+ if (numchunks > ZAP_LEAF_NUMCHUNKS(l))
+ return (E2BIG);
+
+ if (cd == ZAP_MAXCD) {
+ /* find the lowest unused cd */
+ if (l->l_phys->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED) {
+ cd = 0;
+
+ for (chunk = *LEAF_HASH_ENTPTR(l, h);
+ chunk != CHAIN_END; chunk = le->le_next) {
+ le = ZAP_LEAF_ENTRY(l, chunk);
+ if (le->le_cd > cd)
+ break;
+ if (le->le_hash == h) {
+ ASSERT3U(cd, ==, le->le_cd);
+ cd++;
+ }
+ }
+ } else {
+ /* old unsorted format; do it the O(n^2) way */
+ for (cd = 0; cd < ZAP_MAXCD; cd++) {
+ for (chunk = *LEAF_HASH_ENTPTR(l, h);
+ chunk != CHAIN_END; chunk = le->le_next) {
+ le = ZAP_LEAF_ENTRY(l, chunk);
+ if (le->le_hash == h &&
+ le->le_cd == cd) {
+ break;
+ }
+ }
+ /* If this cd is not in use, we are good. */
+ if (chunk == CHAIN_END)
+ break;
+ }
+ }
+ /*
+ * we would run out of space in a block before we could
+ * have ZAP_MAXCD entries
+ */
+ ASSERT3U(cd, <, ZAP_MAXCD);
+ }
+
+ if (l->l_phys->l_hdr.lh_nfree < numchunks)
+ return (EAGAIN);
+
+ /* make the entry */
+ chunk = zap_leaf_chunk_alloc(l);
+ le = ZAP_LEAF_ENTRY(l, chunk);
+ le->le_type = ZAP_CHUNK_ENTRY;
+ le->le_name_chunk = zap_leaf_array_create(l, name, 1, namelen);
+ le->le_name_length = namelen;
+ le->le_value_chunk =
+ zap_leaf_array_create(l, buf, integer_size, num_integers);
+ le->le_value_length = num_integers;
+ le->le_int_size = integer_size;
+ le->le_hash = h;
+ le->le_cd = cd;
+
+ /* link it into the hash chain */
+ /* XXX if we did the search above, we could just use that */
+ chunkp = zap_leaf_rehash_entry(l, chunk);
+
+ l->l_phys->l_hdr.lh_nentries++;
+
+ zeh->zeh_leaf = l;
+ zeh->zeh_num_integers = num_integers;
+ zeh->zeh_integer_size = le->le_int_size;
+ zeh->zeh_cd = le->le_cd;
+ zeh->zeh_hash = le->le_hash;
+ zeh->zeh_chunkp = chunkp;
+
+ return (0);
+}
+
+/*
+ * Determine if there is another entry with the same normalized form.
+ * For performance purposes, either zn or name must be provided (the
+ * other can be NULL). Note, there usually won't be any hash
+ * conflicts, in which case we don't need the concatenated/normalized
+ * form of the name. But all callers have one of these on hand anyway,
+ * so might as well take advantage. A cleaner but slower interface
+ * would accept neither argument, and compute the normalized name as
+ * needed (using zap_name_alloc(zap_entry_read_name(zeh))).
+ */
+boolean_t
+zap_entry_normalization_conflict(zap_entry_handle_t *zeh, zap_name_t *zn,
+ const char *name, zap_t *zap)
+{
+ uint64_t chunk;
+ struct zap_leaf_entry *le;
+ boolean_t allocdzn = B_FALSE;
+
+ if (zap->zap_normflags == 0)
+ return (B_FALSE);
+
+ for (chunk = *LEAF_HASH_ENTPTR(zeh->zeh_leaf, zeh->zeh_hash);
+ chunk != CHAIN_END; chunk = le->le_next) {
+ le = ZAP_LEAF_ENTRY(zeh->zeh_leaf, chunk);
+ if (le->le_hash != zeh->zeh_hash)
+ continue;
+ if (le->le_cd == zeh->zeh_cd)
+ continue;
+
+ if (zn == NULL) {
+ zn = zap_name_alloc(zap, name, MT_FIRST);
+ allocdzn = B_TRUE;
+ }
+ if (zap_leaf_array_match(zeh->zeh_leaf, zn,
+ le->le_name_chunk, le->le_name_length)) {
+ if (allocdzn)
+ zap_name_free(zn);
+ return (B_TRUE);
+ }
+ }
+ if (allocdzn)
+ zap_name_free(zn);
+ return (B_FALSE);
+}
+
+/*
+ * Routines for transferring entries between leafs.
+ */
+
+static uint16_t *
+zap_leaf_rehash_entry(zap_leaf_t *l, uint16_t entry)
+{
+ struct zap_leaf_entry *le = ZAP_LEAF_ENTRY(l, entry);
+ struct zap_leaf_entry *le2;
+ uint16_t *chunkp;
+
+ /*
+ * keep the entry chain sorted by cd
+ * NB: this will not cause problems for unsorted leafs, though
+ * it is unnecessary there.
+ */
+ for (chunkp = LEAF_HASH_ENTPTR(l, le->le_hash);
+ *chunkp != CHAIN_END; chunkp = &le2->le_next) {
+ le2 = ZAP_LEAF_ENTRY(l, *chunkp);
+ if (le2->le_cd > le->le_cd)
+ break;
+ }
+
+ le->le_next = *chunkp;
+ *chunkp = entry;
+ return (chunkp);
+}
+
+static uint16_t
+zap_leaf_transfer_array(zap_leaf_t *l, uint16_t chunk, zap_leaf_t *nl)
+{
+ uint16_t new_chunk;
+ uint16_t *nchunkp = &new_chunk;
+
+ while (chunk != CHAIN_END) {
+ uint16_t nchunk = zap_leaf_chunk_alloc(nl);
+ struct zap_leaf_array *nla =
+ &ZAP_LEAF_CHUNK(nl, nchunk).l_array;
+ struct zap_leaf_array *la =
+ &ZAP_LEAF_CHUNK(l, chunk).l_array;
+ int nextchunk = la->la_next;
+
+ ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(l));
+ ASSERT3U(nchunk, <, ZAP_LEAF_NUMCHUNKS(l));
+
+ *nla = *la; /* structure assignment */
+
+ zap_leaf_chunk_free(l, chunk);
+ chunk = nextchunk;
+ *nchunkp = nchunk;
+ nchunkp = &nla->la_next;
+ }
+ *nchunkp = CHAIN_END;
+ return (new_chunk);
+}
+
+static void
+zap_leaf_transfer_entry(zap_leaf_t *l, int entry, zap_leaf_t *nl)
+{
+ struct zap_leaf_entry *le, *nle;
+ uint16_t chunk;
+
+ le = ZAP_LEAF_ENTRY(l, entry);
+ ASSERT3U(le->le_type, ==, ZAP_CHUNK_ENTRY);
+
+ chunk = zap_leaf_chunk_alloc(nl);
+ nle = ZAP_LEAF_ENTRY(nl, chunk);
+ *nle = *le; /* structure assignment */
+
+ (void) zap_leaf_rehash_entry(nl, chunk);
+
+ nle->le_name_chunk = zap_leaf_transfer_array(l, le->le_name_chunk, nl);
+ nle->le_value_chunk =
+ zap_leaf_transfer_array(l, le->le_value_chunk, nl);
+
+ zap_leaf_chunk_free(l, entry);
+
+ l->l_phys->l_hdr.lh_nentries--;
+ nl->l_phys->l_hdr.lh_nentries++;
+}
+
+/*
+ * Transfer the entries whose hash prefix ends in 1 to the new leaf.
+ */
+void
+zap_leaf_split(zap_leaf_t *l, zap_leaf_t *nl, boolean_t sort)
+{
+ int i;
+ int bit = 64 - 1 - l->l_phys->l_hdr.lh_prefix_len;
+
+ /* set new prefix and prefix_len */
+ l->l_phys->l_hdr.lh_prefix <<= 1;
+ l->l_phys->l_hdr.lh_prefix_len++;
+ nl->l_phys->l_hdr.lh_prefix = l->l_phys->l_hdr.lh_prefix | 1;
+ nl->l_phys->l_hdr.lh_prefix_len = l->l_phys->l_hdr.lh_prefix_len;
+
+ /* break existing hash chains */
+ zap_memset(l->l_phys->l_hash, CHAIN_END, 2*ZAP_LEAF_HASH_NUMENTRIES(l));
+
+ if (sort)
+ l->l_phys->l_hdr.lh_flags |= ZLF_ENTRIES_CDSORTED;
+
+ /*
+ * Transfer entries whose hash bit 'bit' is set to nl; rehash
+ * the remaining entries
+ *
+ * NB: We could find entries via the hashtable instead. That
+ * would be O(hashents+numents) rather than O(numblks+numents),
+ * but this accesses memory more sequentially, and when we're
+ * called, the block is usually pretty full.
+ */
+ for (i = 0; i < ZAP_LEAF_NUMCHUNKS(l); i++) {
+ struct zap_leaf_entry *le = ZAP_LEAF_ENTRY(l, i);
+ if (le->le_type != ZAP_CHUNK_ENTRY)
+ continue;
+
+ if (le->le_hash & (1ULL << bit))
+ zap_leaf_transfer_entry(l, i, nl);
+ else
+ (void) zap_leaf_rehash_entry(l, i);
+ }
+}
+
+void
+zap_leaf_stats(zap_t *zap, zap_leaf_t *l, zap_stats_t *zs)
+{
+ int i, n;
+
+ n = zap->zap_f.zap_phys->zap_ptrtbl.zt_shift -
+ l->l_phys->l_hdr.lh_prefix_len;
+ n = MIN(n, ZAP_HISTOGRAM_SIZE-1);
+ zs->zs_leafs_with_2n_pointers[n]++;
+
+
+ n = l->l_phys->l_hdr.lh_nentries/5;
+ n = MIN(n, ZAP_HISTOGRAM_SIZE-1);
+ zs->zs_blocks_with_n5_entries[n]++;
+
+ n = ((1<<FZAP_BLOCK_SHIFT(zap)) -
+ l->l_phys->l_hdr.lh_nfree * (ZAP_LEAF_ARRAY_BYTES+1))*10 /
+ (1<<FZAP_BLOCK_SHIFT(zap));
+ n = MIN(n, ZAP_HISTOGRAM_SIZE-1);
+ zs->zs_blocks_n_tenths_full[n]++;
+
+ for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(l); i++) {
+ int nentries = 0;
+ int chunk = l->l_phys->l_hash[i];
+
+ while (chunk != CHAIN_END) {
+ struct zap_leaf_entry *le =
+ ZAP_LEAF_ENTRY(l, chunk);
+
+ n = 1 + ZAP_LEAF_ARRAY_NCHUNKS(le->le_name_length) +
+ ZAP_LEAF_ARRAY_NCHUNKS(le->le_value_length *
+ le->le_int_size);
+ n = MIN(n, ZAP_HISTOGRAM_SIZE-1);
+ zs->zs_entries_using_n_chunks[n]++;
+
+ chunk = le->le_next;
+ nentries++;
+ }
+
+ n = nentries;
+ n = MIN(n, ZAP_HISTOGRAM_SIZE-1);
+ zs->zs_buckets_with_n_entries[n]++;
+ }
+}
diff --git a/module/zfs/zap_micro.c b/module/zfs/zap_micro.c
new file mode 100644
index 000000000..abba42775
--- /dev/null
+++ b/module/zfs/zap_micro.c
@@ -0,0 +1,1069 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/zfs_context.h>
+#include <sys/zap.h>
+#include <sys/refcount.h>
+#include <sys/zap_impl.h>
+#include <sys/zap_leaf.h>
+#include <sys/avl.h>
+
+#ifdef _KERNEL
+#include <sys/sunddi.h>
+#endif
+
+static int mzap_upgrade(zap_t **zapp, dmu_tx_t *tx);
+
+
+static uint64_t
+zap_hash(zap_t *zap, const char *normname)
+{
+ const uint8_t *cp;
+ uint8_t c;
+ uint64_t crc = zap->zap_salt;
+
+ /* NB: name must already be normalized, if necessary */
+
+ ASSERT(crc != 0);
+ ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
+ for (cp = (const uint8_t *)normname; (c = *cp) != '\0'; cp++) {
+ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ c) & 0xFF];
+ }
+
+ /*
+ * Only use 28 bits, since we need 4 bits in the cookie for the
+ * collision differentiator. We MUST use the high bits, since
+ * those are the ones that we first pay attention to when
+ * chosing the bucket.
+ */
+ crc &= ~((1ULL << (64 - ZAP_HASHBITS)) - 1);
+
+ return (crc);
+}
+
+static int
+zap_normalize(zap_t *zap, const char *name, char *namenorm)
+{
+ size_t inlen, outlen;
+ int err;
+
+ inlen = strlen(name) + 1;
+ outlen = ZAP_MAXNAMELEN;
+
+ err = 0;
+ (void) u8_textprep_str((char *)name, &inlen, namenorm, &outlen,
+ zap->zap_normflags | U8_TEXTPREP_IGNORE_NULL, U8_UNICODE_LATEST,
+ &err);
+
+ return (err);
+}
+
+boolean_t
+zap_match(zap_name_t *zn, const char *matchname)
+{
+ if (zn->zn_matchtype == MT_FIRST) {
+ char norm[ZAP_MAXNAMELEN];
+
+ if (zap_normalize(zn->zn_zap, matchname, norm) != 0)
+ return (B_FALSE);
+
+ return (strcmp(zn->zn_name_norm, norm) == 0);
+ } else {
+ /* MT_BEST or MT_EXACT */
+ return (strcmp(zn->zn_name_orij, matchname) == 0);
+ }
+}
+
+void
+zap_name_free(zap_name_t *zn)
+{
+ kmem_free(zn, sizeof (zap_name_t));
+}
+
+/* XXX combine this with zap_lockdir()? */
+zap_name_t *
+zap_name_alloc(zap_t *zap, const char *name, matchtype_t mt)
+{
+ zap_name_t *zn = kmem_alloc(sizeof (zap_name_t), KM_SLEEP);
+
+ zn->zn_zap = zap;
+ zn->zn_name_orij = name;
+ zn->zn_matchtype = mt;
+ if (zap->zap_normflags) {
+ if (zap_normalize(zap, name, zn->zn_normbuf) != 0) {
+ zap_name_free(zn);
+ return (NULL);
+ }
+ zn->zn_name_norm = zn->zn_normbuf;
+ } else {
+ if (mt != MT_EXACT) {
+ zap_name_free(zn);
+ return (NULL);
+ }
+ zn->zn_name_norm = zn->zn_name_orij;
+ }
+
+ zn->zn_hash = zap_hash(zap, zn->zn_name_norm);
+ return (zn);
+}
+
+static void
+mzap_byteswap(mzap_phys_t *buf, size_t size)
+{
+ int i, max;
+ buf->mz_block_type = BSWAP_64(buf->mz_block_type);
+ buf->mz_salt = BSWAP_64(buf->mz_salt);
+ buf->mz_normflags = BSWAP_64(buf->mz_normflags);
+ max = (size / MZAP_ENT_LEN) - 1;
+ for (i = 0; i < max; i++) {
+ buf->mz_chunk[i].mze_value =
+ BSWAP_64(buf->mz_chunk[i].mze_value);
+ buf->mz_chunk[i].mze_cd =
+ BSWAP_32(buf->mz_chunk[i].mze_cd);
+ }
+}
+
+void
+zap_byteswap(void *buf, size_t size)
+{
+ uint64_t block_type;
+
+ block_type = *(uint64_t *)buf;
+
+ if (block_type == ZBT_MICRO || block_type == BSWAP_64(ZBT_MICRO)) {
+ /* ASSERT(magic == ZAP_LEAF_MAGIC); */
+ mzap_byteswap(buf, size);
+ } else {
+ fzap_byteswap(buf, size);
+ }
+}
+
+static int
+mze_compare(const void *arg1, const void *arg2)
+{
+ const mzap_ent_t *mze1 = arg1;
+ const mzap_ent_t *mze2 = arg2;
+
+ if (mze1->mze_hash > mze2->mze_hash)
+ return (+1);
+ if (mze1->mze_hash < mze2->mze_hash)
+ return (-1);
+ if (mze1->mze_phys.mze_cd > mze2->mze_phys.mze_cd)
+ return (+1);
+ if (mze1->mze_phys.mze_cd < mze2->mze_phys.mze_cd)
+ return (-1);
+ return (0);
+}
+
+static void
+mze_insert(zap_t *zap, int chunkid, uint64_t hash, mzap_ent_phys_t *mzep)
+{
+ mzap_ent_t *mze;
+
+ ASSERT(zap->zap_ismicro);
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+ ASSERT(mzep->mze_cd < ZAP_MAXCD);
+
+ mze = kmem_alloc(sizeof (mzap_ent_t), KM_SLEEP);
+ mze->mze_chunkid = chunkid;
+ mze->mze_hash = hash;
+ mze->mze_phys = *mzep;
+ avl_add(&zap->zap_m.zap_avl, mze);
+}
+
+static mzap_ent_t *
+mze_find(zap_name_t *zn)
+{
+ mzap_ent_t mze_tofind;
+ mzap_ent_t *mze;
+ avl_index_t idx;
+ avl_tree_t *avl = &zn->zn_zap->zap_m.zap_avl;
+
+ ASSERT(zn->zn_zap->zap_ismicro);
+ ASSERT(RW_LOCK_HELD(&zn->zn_zap->zap_rwlock));
+
+ if (strlen(zn->zn_name_norm) >= sizeof (mze_tofind.mze_phys.mze_name))
+ return (NULL);
+
+ mze_tofind.mze_hash = zn->zn_hash;
+ mze_tofind.mze_phys.mze_cd = 0;
+
+again:
+ mze = avl_find(avl, &mze_tofind, &idx);
+ if (mze == NULL)
+ mze = avl_nearest(avl, idx, AVL_AFTER);
+ for (; mze && mze->mze_hash == zn->zn_hash; mze = AVL_NEXT(avl, mze)) {
+ if (zap_match(zn, mze->mze_phys.mze_name))
+ return (mze);
+ }
+ if (zn->zn_matchtype == MT_BEST) {
+ zn->zn_matchtype = MT_FIRST;
+ goto again;
+ }
+ return (NULL);
+}
+
+static uint32_t
+mze_find_unused_cd(zap_t *zap, uint64_t hash)
+{
+ mzap_ent_t mze_tofind;
+ mzap_ent_t *mze;
+ avl_index_t idx;
+ avl_tree_t *avl = &zap->zap_m.zap_avl;
+ uint32_t cd;
+
+ ASSERT(zap->zap_ismicro);
+ ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
+
+ mze_tofind.mze_hash = hash;
+ mze_tofind.mze_phys.mze_cd = 0;
+
+ cd = 0;
+ for (mze = avl_find(avl, &mze_tofind, &idx);
+ mze && mze->mze_hash == hash; mze = AVL_NEXT(avl, mze)) {
+ if (mze->mze_phys.mze_cd != cd)
+ break;
+ cd++;
+ }
+
+ return (cd);
+}
+
+static void
+mze_remove(zap_t *zap, mzap_ent_t *mze)
+{
+ ASSERT(zap->zap_ismicro);
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+
+ avl_remove(&zap->zap_m.zap_avl, mze);
+ kmem_free(mze, sizeof (mzap_ent_t));
+}
+
+static void
+mze_destroy(zap_t *zap)
+{
+ mzap_ent_t *mze;
+ void *avlcookie = NULL;
+
+ while (mze = avl_destroy_nodes(&zap->zap_m.zap_avl, &avlcookie))
+ kmem_free(mze, sizeof (mzap_ent_t));
+ avl_destroy(&zap->zap_m.zap_avl);
+}
+
+static zap_t *
+mzap_open(objset_t *os, uint64_t obj, dmu_buf_t *db)
+{
+ zap_t *winner;
+ zap_t *zap;
+ int i;
+
+ ASSERT3U(MZAP_ENT_LEN, ==, sizeof (mzap_ent_phys_t));
+
+ zap = kmem_zalloc(sizeof (zap_t), KM_SLEEP);
+ rw_init(&zap->zap_rwlock, 0, 0, 0);
+ rw_enter(&zap->zap_rwlock, RW_WRITER);
+ zap->zap_objset = os;
+ zap->zap_object = obj;
+ zap->zap_dbuf = db;
+
+ if (*(uint64_t *)db->db_data != ZBT_MICRO) {
+ mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
+ zap->zap_f.zap_block_shift = highbit(db->db_size) - 1;
+ } else {
+ zap->zap_ismicro = TRUE;
+ }
+
+ /*
+ * Make sure that zap_ismicro is set before we let others see
+ * it, because zap_lockdir() checks zap_ismicro without the lock
+ * held.
+ */
+ winner = dmu_buf_set_user(db, zap, &zap->zap_m.zap_phys, zap_evict);
+
+ if (winner != NULL) {
+ rw_exit(&zap->zap_rwlock);
+ rw_destroy(&zap->zap_rwlock);
+ if (!zap->zap_ismicro)
+ mutex_destroy(&zap->zap_f.zap_num_entries_mtx);
+ kmem_free(zap, sizeof (zap_t));
+ return (winner);
+ }
+
+ if (zap->zap_ismicro) {
+ zap->zap_salt = zap->zap_m.zap_phys->mz_salt;
+ zap->zap_normflags = zap->zap_m.zap_phys->mz_normflags;
+ zap->zap_m.zap_num_chunks = db->db_size / MZAP_ENT_LEN - 1;
+ avl_create(&zap->zap_m.zap_avl, mze_compare,
+ sizeof (mzap_ent_t), offsetof(mzap_ent_t, mze_node));
+
+ for (i = 0; i < zap->zap_m.zap_num_chunks; i++) {
+ mzap_ent_phys_t *mze =
+ &zap->zap_m.zap_phys->mz_chunk[i];
+ if (mze->mze_name[0]) {
+ zap_name_t *zn;
+
+ zap->zap_m.zap_num_entries++;
+ zn = zap_name_alloc(zap, mze->mze_name,
+ MT_EXACT);
+ mze_insert(zap, i, zn->zn_hash, mze);
+ zap_name_free(zn);
+ }
+ }
+ } else {
+ zap->zap_salt = zap->zap_f.zap_phys->zap_salt;
+ zap->zap_normflags = zap->zap_f.zap_phys->zap_normflags;
+
+ ASSERT3U(sizeof (struct zap_leaf_header), ==,
+ 2*ZAP_LEAF_CHUNKSIZE);
+
+ /*
+ * The embedded pointer table should not overlap the
+ * other members.
+ */
+ ASSERT3P(&ZAP_EMBEDDED_PTRTBL_ENT(zap, 0), >,
+ &zap->zap_f.zap_phys->zap_salt);
+
+ /*
+ * The embedded pointer table should end at the end of
+ * the block
+ */
+ ASSERT3U((uintptr_t)&ZAP_EMBEDDED_PTRTBL_ENT(zap,
+ 1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)) -
+ (uintptr_t)zap->zap_f.zap_phys, ==,
+ zap->zap_dbuf->db_size);
+ }
+ rw_exit(&zap->zap_rwlock);
+ return (zap);
+}
+
+int
+zap_lockdir(objset_t *os, uint64_t obj, dmu_tx_t *tx,
+ krw_t lti, boolean_t fatreader, boolean_t adding, zap_t **zapp)
+{
+ zap_t *zap;
+ dmu_buf_t *db;
+ krw_t lt;
+ int err;
+
+ *zapp = NULL;
+
+ err = dmu_buf_hold(os, obj, 0, NULL, &db);
+ if (err)
+ return (err);
+
+#ifdef ZFS_DEBUG
+ {
+ dmu_object_info_t doi;
+ dmu_object_info_from_db(db, &doi);
+ ASSERT(dmu_ot[doi.doi_type].ot_byteswap == zap_byteswap);
+ }
+#endif
+
+ zap = dmu_buf_get_user(db);
+ if (zap == NULL)
+ zap = mzap_open(os, obj, db);
+
+ /*
+ * We're checking zap_ismicro without the lock held, in order to
+ * tell what type of lock we want. Once we have some sort of
+ * lock, see if it really is the right type. In practice this
+ * can only be different if it was upgraded from micro to fat,
+ * and micro wanted WRITER but fat only needs READER.
+ */
+ lt = (!zap->zap_ismicro && fatreader) ? RW_READER : lti;
+ rw_enter(&zap->zap_rwlock, lt);
+ if (lt != ((!zap->zap_ismicro && fatreader) ? RW_READER : lti)) {
+ /* it was upgraded, now we only need reader */
+ ASSERT(lt == RW_WRITER);
+ ASSERT(RW_READER ==
+ (!zap->zap_ismicro && fatreader) ? RW_READER : lti);
+ rw_downgrade(&zap->zap_rwlock);
+ lt = RW_READER;
+ }
+
+ zap->zap_objset = os;
+
+ if (lt == RW_WRITER)
+ dmu_buf_will_dirty(db, tx);
+
+ ASSERT3P(zap->zap_dbuf, ==, db);
+
+ ASSERT(!zap->zap_ismicro ||
+ zap->zap_m.zap_num_entries <= zap->zap_m.zap_num_chunks);
+ if (zap->zap_ismicro && tx && adding &&
+ zap->zap_m.zap_num_entries == zap->zap_m.zap_num_chunks) {
+ uint64_t newsz = db->db_size + SPA_MINBLOCKSIZE;
+ if (newsz > MZAP_MAX_BLKSZ) {
+ dprintf("upgrading obj %llu: num_entries=%u\n",
+ obj, zap->zap_m.zap_num_entries);
+ *zapp = zap;
+ return (mzap_upgrade(zapp, tx));
+ }
+ err = dmu_object_set_blocksize(os, obj, newsz, 0, tx);
+ ASSERT3U(err, ==, 0);
+ zap->zap_m.zap_num_chunks =
+ db->db_size / MZAP_ENT_LEN - 1;
+ }
+
+ *zapp = zap;
+ return (0);
+}
+
+void
+zap_unlockdir(zap_t *zap)
+{
+ rw_exit(&zap->zap_rwlock);
+ dmu_buf_rele(zap->zap_dbuf, NULL);
+}
+
+static int
+mzap_upgrade(zap_t **zapp, dmu_tx_t *tx)
+{
+ mzap_phys_t *mzp;
+ int i, sz, nchunks, err;
+ zap_t *zap = *zapp;
+
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+
+ sz = zap->zap_dbuf->db_size;
+ mzp = kmem_alloc(sz, KM_SLEEP);
+ bcopy(zap->zap_dbuf->db_data, mzp, sz);
+ nchunks = zap->zap_m.zap_num_chunks;
+
+ err = dmu_object_set_blocksize(zap->zap_objset, zap->zap_object,
+ 1ULL << fzap_default_block_shift, 0, tx);
+ if (err) {
+ kmem_free(mzp, sz);
+ return (err);
+ }
+
+ dprintf("upgrading obj=%llu with %u chunks\n",
+ zap->zap_object, nchunks);
+ /* XXX destroy the avl later, so we can use the stored hash value */
+ mze_destroy(zap);
+
+ fzap_upgrade(zap, tx);
+
+ for (i = 0; i < nchunks; i++) {
+ int err;
+ mzap_ent_phys_t *mze = &mzp->mz_chunk[i];
+ zap_name_t *zn;
+ if (mze->mze_name[0] == 0)
+ continue;
+ dprintf("adding %s=%llu\n",
+ mze->mze_name, mze->mze_value);
+ zn = zap_name_alloc(zap, mze->mze_name, MT_EXACT);
+ err = fzap_add_cd(zn, 8, 1, &mze->mze_value, mze->mze_cd, tx);
+ zap = zn->zn_zap; /* fzap_add_cd() may change zap */
+ zap_name_free(zn);
+ if (err)
+ break;
+ }
+ kmem_free(mzp, sz);
+ *zapp = zap;
+ return (err);
+}
+
+static void
+mzap_create_impl(objset_t *os, uint64_t obj, int normflags, dmu_tx_t *tx)
+{
+ dmu_buf_t *db;
+ mzap_phys_t *zp;
+
+ VERIFY(0 == dmu_buf_hold(os, obj, 0, FTAG, &db));
+
+#ifdef ZFS_DEBUG
+ {
+ dmu_object_info_t doi;
+ dmu_object_info_from_db(db, &doi);
+ ASSERT(dmu_ot[doi.doi_type].ot_byteswap == zap_byteswap);
+ }
+#endif
+
+ dmu_buf_will_dirty(db, tx);
+ zp = db->db_data;
+ zp->mz_block_type = ZBT_MICRO;
+ zp->mz_salt = ((uintptr_t)db ^ (uintptr_t)tx ^ (obj << 1)) | 1ULL;
+ zp->mz_normflags = normflags;
+ dmu_buf_rele(db, FTAG);
+}
+
+int
+zap_create_claim(objset_t *os, uint64_t obj, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ return (zap_create_claim_norm(os, obj,
+ 0, ot, bonustype, bonuslen, tx));
+}
+
+int
+zap_create_claim_norm(objset_t *os, uint64_t obj, int normflags,
+ dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ int err;
+
+ err = dmu_object_claim(os, obj, ot, 0, bonustype, bonuslen, tx);
+ if (err != 0)
+ return (err);
+ mzap_create_impl(os, obj, normflags, tx);
+ return (0);
+}
+
+uint64_t
+zap_create(objset_t *os, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ return (zap_create_norm(os, 0, ot, bonustype, bonuslen, tx));
+}
+
+uint64_t
+zap_create_norm(objset_t *os, int normflags, dmu_object_type_t ot,
+ dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+{
+ uint64_t obj = dmu_object_alloc(os, ot, 0, bonustype, bonuslen, tx);
+
+ mzap_create_impl(os, obj, normflags, tx);
+ return (obj);
+}
+
+int
+zap_destroy(objset_t *os, uint64_t zapobj, dmu_tx_t *tx)
+{
+ /*
+ * dmu_object_free will free the object number and free the
+ * data. Freeing the data will cause our pageout function to be
+ * called, which will destroy our data (zap_leaf_t's and zap_t).
+ */
+
+ return (dmu_object_free(os, zapobj, tx));
+}
+
+_NOTE(ARGSUSED(0))
+void
+zap_evict(dmu_buf_t *db, void *vzap)
+{
+ zap_t *zap = vzap;
+
+ rw_destroy(&zap->zap_rwlock);
+
+ if (zap->zap_ismicro)
+ mze_destroy(zap);
+ else
+ mutex_destroy(&zap->zap_f.zap_num_entries_mtx);
+
+ kmem_free(zap, sizeof (zap_t));
+}
+
+int
+zap_count(objset_t *os, uint64_t zapobj, uint64_t *count)
+{
+ zap_t *zap;
+ int err;
+
+ err = zap_lockdir(os, zapobj, NULL, RW_READER, TRUE, FALSE, &zap);
+ if (err)
+ return (err);
+ if (!zap->zap_ismicro) {
+ err = fzap_count(zap, count);
+ } else {
+ *count = zap->zap_m.zap_num_entries;
+ }
+ zap_unlockdir(zap);
+ return (err);
+}
+
+/*
+ * zn may be NULL; if not specified, it will be computed if needed.
+ * See also the comment above zap_entry_normalization_conflict().
+ */
+static boolean_t
+mzap_normalization_conflict(zap_t *zap, zap_name_t *zn, mzap_ent_t *mze)
+{
+ mzap_ent_t *other;
+ int direction = AVL_BEFORE;
+ boolean_t allocdzn = B_FALSE;
+
+ if (zap->zap_normflags == 0)
+ return (B_FALSE);
+
+again:
+ for (other = avl_walk(&zap->zap_m.zap_avl, mze, direction);
+ other && other->mze_hash == mze->mze_hash;
+ other = avl_walk(&zap->zap_m.zap_avl, other, direction)) {
+
+ if (zn == NULL) {
+ zn = zap_name_alloc(zap, mze->mze_phys.mze_name,
+ MT_FIRST);
+ allocdzn = B_TRUE;
+ }
+ if (zap_match(zn, other->mze_phys.mze_name)) {
+ if (allocdzn)
+ zap_name_free(zn);
+ return (B_TRUE);
+ }
+ }
+
+ if (direction == AVL_BEFORE) {
+ direction = AVL_AFTER;
+ goto again;
+ }
+
+ if (allocdzn)
+ zap_name_free(zn);
+ return (B_FALSE);
+}
+
+/*
+ * Routines for manipulating attributes.
+ */
+
+int
+zap_lookup(objset_t *os, uint64_t zapobj, const char *name,
+ uint64_t integer_size, uint64_t num_integers, void *buf)
+{
+ return (zap_lookup_norm(os, zapobj, name, integer_size,
+ num_integers, buf, MT_EXACT, NULL, 0, NULL));
+}
+
+int
+zap_lookup_norm(objset_t *os, uint64_t zapobj, const char *name,
+ uint64_t integer_size, uint64_t num_integers, void *buf,
+ matchtype_t mt, char *realname, int rn_len,
+ boolean_t *ncp)
+{
+ zap_t *zap;
+ int err;
+ mzap_ent_t *mze;
+ zap_name_t *zn;
+
+ err = zap_lockdir(os, zapobj, NULL, RW_READER, TRUE, FALSE, &zap);
+ if (err)
+ return (err);
+ zn = zap_name_alloc(zap, name, mt);
+ if (zn == NULL) {
+ zap_unlockdir(zap);
+ return (ENOTSUP);
+ }
+
+ if (!zap->zap_ismicro) {
+ err = fzap_lookup(zn, integer_size, num_integers, buf,
+ realname, rn_len, ncp);
+ } else {
+ mze = mze_find(zn);
+ if (mze == NULL) {
+ err = ENOENT;
+ } else {
+ if (num_integers < 1) {
+ err = EOVERFLOW;
+ } else if (integer_size != 8) {
+ err = EINVAL;
+ } else {
+ *(uint64_t *)buf = mze->mze_phys.mze_value;
+ (void) strlcpy(realname,
+ mze->mze_phys.mze_name, rn_len);
+ if (ncp) {
+ *ncp = mzap_normalization_conflict(zap,
+ zn, mze);
+ }
+ }
+ }
+ }
+ zap_name_free(zn);
+ zap_unlockdir(zap);
+ return (err);
+}
+
+int
+zap_length(objset_t *os, uint64_t zapobj, const char *name,
+ uint64_t *integer_size, uint64_t *num_integers)
+{
+ zap_t *zap;
+ int err;
+ mzap_ent_t *mze;
+ zap_name_t *zn;
+
+ err = zap_lockdir(os, zapobj, NULL, RW_READER, TRUE, FALSE, &zap);
+ if (err)
+ return (err);
+ zn = zap_name_alloc(zap, name, MT_EXACT);
+ if (zn == NULL) {
+ zap_unlockdir(zap);
+ return (ENOTSUP);
+ }
+ if (!zap->zap_ismicro) {
+ err = fzap_length(zn, integer_size, num_integers);
+ } else {
+ mze = mze_find(zn);
+ if (mze == NULL) {
+ err = ENOENT;
+ } else {
+ if (integer_size)
+ *integer_size = 8;
+ if (num_integers)
+ *num_integers = 1;
+ }
+ }
+ zap_name_free(zn);
+ zap_unlockdir(zap);
+ return (err);
+}
+
+static void
+mzap_addent(zap_name_t *zn, uint64_t value)
+{
+ int i;
+ zap_t *zap = zn->zn_zap;
+ int start = zap->zap_m.zap_alloc_next;
+ uint32_t cd;
+
+ dprintf("obj=%llu %s=%llu\n", zap->zap_object,
+ zn->zn_name_orij, value);
+ ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
+
+#ifdef ZFS_DEBUG
+ for (i = 0; i < zap->zap_m.zap_num_chunks; i++) {
+ mzap_ent_phys_t *mze = &zap->zap_m.zap_phys->mz_chunk[i];
+ ASSERT(strcmp(zn->zn_name_orij, mze->mze_name) != 0);
+ }
+#endif
+
+ cd = mze_find_unused_cd(zap, zn->zn_hash);
+ /* given the limited size of the microzap, this can't happen */
+ ASSERT(cd != ZAP_MAXCD);
+
+again:
+ for (i = start; i < zap->zap_m.zap_num_chunks; i++) {
+ mzap_ent_phys_t *mze = &zap->zap_m.zap_phys->mz_chunk[i];
+ if (mze->mze_name[0] == 0) {
+ mze->mze_value = value;
+ mze->mze_cd = cd;
+ (void) strcpy(mze->mze_name, zn->zn_name_orij);
+ zap->zap_m.zap_num_entries++;
+ zap->zap_m.zap_alloc_next = i+1;
+ if (zap->zap_m.zap_alloc_next ==
+ zap->zap_m.zap_num_chunks)
+ zap->zap_m.zap_alloc_next = 0;
+ mze_insert(zap, i, zn->zn_hash, mze);
+ return;
+ }
+ }
+ if (start != 0) {
+ start = 0;
+ goto again;
+ }
+ ASSERT(!"out of entries!");
+}
+
+int
+zap_add(objset_t *os, uint64_t zapobj, const char *name,
+ int integer_size, uint64_t num_integers,
+ const void *val, dmu_tx_t *tx)
+{
+ zap_t *zap;
+ int err;
+ mzap_ent_t *mze;
+ const uint64_t *intval = val;
+ zap_name_t *zn;
+
+ err = zap_lockdir(os, zapobj, tx, RW_WRITER, TRUE, TRUE, &zap);
+ if (err)
+ return (err);
+ zn = zap_name_alloc(zap, name, MT_EXACT);
+ if (zn == NULL) {
+ zap_unlockdir(zap);
+ return (ENOTSUP);
+ }
+ if (!zap->zap_ismicro) {
+ err = fzap_add(zn, integer_size, num_integers, val, tx);
+ zap = zn->zn_zap; /* fzap_add() may change zap */
+ } else if (integer_size != 8 || num_integers != 1 ||
+ strlen(name) >= MZAP_NAME_LEN) {
+ dprintf("upgrading obj %llu: intsz=%u numint=%llu name=%s\n",
+ zapobj, integer_size, num_integers, name);
+ err = mzap_upgrade(&zn->zn_zap, tx);
+ if (err == 0)
+ err = fzap_add(zn, integer_size, num_integers, val, tx);
+ zap = zn->zn_zap; /* fzap_add() may change zap */
+ } else {
+ mze = mze_find(zn);
+ if (mze != NULL) {
+ err = EEXIST;
+ } else {
+ mzap_addent(zn, *intval);
+ }
+ }
+ ASSERT(zap == zn->zn_zap);
+ zap_name_free(zn);
+ if (zap != NULL) /* may be NULL if fzap_add() failed */
+ zap_unlockdir(zap);
+ return (err);
+}
+
+int
+zap_update(objset_t *os, uint64_t zapobj, const char *name,
+ int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx)
+{
+ zap_t *zap;
+ mzap_ent_t *mze;
+ const uint64_t *intval = val;
+ zap_name_t *zn;
+ int err;
+
+ err = zap_lockdir(os, zapobj, tx, RW_WRITER, TRUE, TRUE, &zap);
+ if (err)
+ return (err);
+ zn = zap_name_alloc(zap, name, MT_EXACT);
+ if (zn == NULL) {
+ zap_unlockdir(zap);
+ return (ENOTSUP);
+ }
+ if (!zap->zap_ismicro) {
+ err = fzap_update(zn, integer_size, num_integers, val, tx);
+ zap = zn->zn_zap; /* fzap_update() may change zap */
+ } else if (integer_size != 8 || num_integers != 1 ||
+ strlen(name) >= MZAP_NAME_LEN) {
+ dprintf("upgrading obj %llu: intsz=%u numint=%llu name=%s\n",
+ zapobj, integer_size, num_integers, name);
+ err = mzap_upgrade(&zn->zn_zap, tx);
+ if (err == 0)
+ err = fzap_update(zn, integer_size, num_integers,
+ val, tx);
+ zap = zn->zn_zap; /* fzap_update() may change zap */
+ } else {
+ mze = mze_find(zn);
+ if (mze != NULL) {
+ mze->mze_phys.mze_value = *intval;
+ zap->zap_m.zap_phys->mz_chunk
+ [mze->mze_chunkid].mze_value = *intval;
+ } else {
+ mzap_addent(zn, *intval);
+ }
+ }
+ ASSERT(zap == zn->zn_zap);
+ zap_name_free(zn);
+ if (zap != NULL) /* may be NULL if fzap_upgrade() failed */
+ zap_unlockdir(zap);
+ return (err);
+}
+
+int
+zap_remove(objset_t *os, uint64_t zapobj, const char *name, dmu_tx_t *tx)
+{
+ return (zap_remove_norm(os, zapobj, name, MT_EXACT, tx));
+}
+
+int
+zap_remove_norm(objset_t *os, uint64_t zapobj, const char *name,
+ matchtype_t mt, dmu_tx_t *tx)
+{
+ zap_t *zap;
+ int err;
+ mzap_ent_t *mze;
+ zap_name_t *zn;
+
+ err = zap_lockdir(os, zapobj, tx, RW_WRITER, TRUE, FALSE, &zap);
+ if (err)
+ return (err);
+ zn = zap_name_alloc(zap, name, mt);
+ if (zn == NULL) {
+ zap_unlockdir(zap);
+ return (ENOTSUP);
+ }
+ if (!zap->zap_ismicro) {
+ err = fzap_remove(zn, tx);
+ } else {
+ mze = mze_find(zn);
+ if (mze == NULL) {
+ err = ENOENT;
+ } else {
+ zap->zap_m.zap_num_entries--;
+ bzero(&zap->zap_m.zap_phys->mz_chunk[mze->mze_chunkid],
+ sizeof (mzap_ent_phys_t));
+ mze_remove(zap, mze);
+ }
+ }
+ zap_name_free(zn);
+ zap_unlockdir(zap);
+ return (err);
+}
+
+/*
+ * Routines for iterating over the attributes.
+ */
+
+/*
+ * We want to keep the high 32 bits of the cursor zero if we can, so
+ * that 32-bit programs can access this. So use a small hash value so
+ * we can fit 4 bits of cd into the 32-bit cursor.
+ *
+ * [ 4 zero bits | 32-bit collision differentiator | 28-bit hash value ]
+ */
+void
+zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *os, uint64_t zapobj,
+ uint64_t serialized)
+{
+ zc->zc_objset = os;
+ zc->zc_zap = NULL;
+ zc->zc_leaf = NULL;
+ zc->zc_zapobj = zapobj;
+ if (serialized == -1ULL) {
+ zc->zc_hash = -1ULL;
+ zc->zc_cd = 0;
+ } else {
+ zc->zc_hash = serialized << (64-ZAP_HASHBITS);
+ zc->zc_cd = serialized >> ZAP_HASHBITS;
+ if (zc->zc_cd >= ZAP_MAXCD) /* corrupt serialized */
+ zc->zc_cd = 0;
+ }
+}
+
+void
+zap_cursor_init(zap_cursor_t *zc, objset_t *os, uint64_t zapobj)
+{
+ zap_cursor_init_serialized(zc, os, zapobj, 0);
+}
+
+void
+zap_cursor_fini(zap_cursor_t *zc)
+{
+ if (zc->zc_zap) {
+ rw_enter(&zc->zc_zap->zap_rwlock, RW_READER);
+ zap_unlockdir(zc->zc_zap);
+ zc->zc_zap = NULL;
+ }
+ if (zc->zc_leaf) {
+ rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);
+ zap_put_leaf(zc->zc_leaf);
+ zc->zc_leaf = NULL;
+ }
+ zc->zc_objset = NULL;
+}
+
+uint64_t
+zap_cursor_serialize(zap_cursor_t *zc)
+{
+ if (zc->zc_hash == -1ULL)
+ return (-1ULL);
+ ASSERT((zc->zc_hash & (ZAP_MAXCD-1)) == 0);
+ ASSERT(zc->zc_cd < ZAP_MAXCD);
+ return ((zc->zc_hash >> (64-ZAP_HASHBITS)) |
+ ((uint64_t)zc->zc_cd << ZAP_HASHBITS));
+}
+
+int
+zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za)
+{
+ int err;
+ avl_index_t idx;
+ mzap_ent_t mze_tofind;
+ mzap_ent_t *mze;
+
+ if (zc->zc_hash == -1ULL)
+ return (ENOENT);
+
+ if (zc->zc_zap == NULL) {
+ err = zap_lockdir(zc->zc_objset, zc->zc_zapobj, NULL,
+ RW_READER, TRUE, FALSE, &zc->zc_zap);
+ if (err)
+ return (err);
+ } else {
+ rw_enter(&zc->zc_zap->zap_rwlock, RW_READER);
+ }
+ if (!zc->zc_zap->zap_ismicro) {
+ err = fzap_cursor_retrieve(zc->zc_zap, zc, za);
+ } else {
+ err = ENOENT;
+
+ mze_tofind.mze_hash = zc->zc_hash;
+ mze_tofind.mze_phys.mze_cd = zc->zc_cd;
+
+ mze = avl_find(&zc->zc_zap->zap_m.zap_avl, &mze_tofind, &idx);
+ if (mze == NULL) {
+ mze = avl_nearest(&zc->zc_zap->zap_m.zap_avl,
+ idx, AVL_AFTER);
+ }
+ if (mze) {
+ ASSERT(0 == bcmp(&mze->mze_phys,
+ &zc->zc_zap->zap_m.zap_phys->mz_chunk
+ [mze->mze_chunkid], sizeof (mze->mze_phys)));
+
+ za->za_normalization_conflict =
+ mzap_normalization_conflict(zc->zc_zap, NULL, mze);
+ za->za_integer_length = 8;
+ za->za_num_integers = 1;
+ za->za_first_integer = mze->mze_phys.mze_value;
+ (void) strcpy(za->za_name, mze->mze_phys.mze_name);
+ zc->zc_hash = mze->mze_hash;
+ zc->zc_cd = mze->mze_phys.mze_cd;
+ err = 0;
+ } else {
+ zc->zc_hash = -1ULL;
+ }
+ }
+ rw_exit(&zc->zc_zap->zap_rwlock);
+ return (err);
+}
+
+void
+zap_cursor_advance(zap_cursor_t *zc)
+{
+ if (zc->zc_hash == -1ULL)
+ return;
+ zc->zc_cd++;
+ if (zc->zc_cd >= ZAP_MAXCD) {
+ zc->zc_cd = 0;
+ zc->zc_hash += 1ULL<<(64-ZAP_HASHBITS);
+ if (zc->zc_hash == 0) /* EOF */
+ zc->zc_hash = -1ULL;
+ }
+}
+
+int
+zap_get_stats(objset_t *os, uint64_t zapobj, zap_stats_t *zs)
+{
+ int err;
+ zap_t *zap;
+
+ err = zap_lockdir(os, zapobj, NULL, RW_READER, TRUE, FALSE, &zap);
+ if (err)
+ return (err);
+
+ bzero(zs, sizeof (zap_stats_t));
+
+ if (zap->zap_ismicro) {
+ zs->zs_blocksize = zap->zap_dbuf->db_size;
+ zs->zs_num_entries = zap->zap_m.zap_num_entries;
+ zs->zs_num_blocks = 1;
+ } else {
+ fzap_get_stats(zap, zs);
+ }
+ zap_unlockdir(zap);
+ return (0);
+}
diff --git a/module/zfs/zfs_acl.c b/module/zfs/zfs_acl.c
new file mode 100644
index 000000000..341dc4dfe
--- /dev/null
+++ b/module/zfs/zfs_acl.c
@@ -0,0 +1,2680 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/time.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/resource.h>
+#include <sys/vfs.h>
+#include <sys/vnode.h>
+#include <sys/sid.h>
+#include <sys/file.h>
+#include <sys/stat.h>
+#include <sys/kmem.h>
+#include <sys/cmn_err.h>
+#include <sys/errno.h>
+#include <sys/unistd.h>
+#include <sys/sdt.h>
+#include <sys/fs/zfs.h>
+#include <sys/mode.h>
+#include <sys/policy.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_fuid.h>
+#include <sys/zfs_acl.h>
+#include <sys/zfs_dir.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/dmu.h>
+#include <sys/dnode.h>
+#include <sys/zap.h>
+#include "fs/fs_subr.h"
+#include <acl/acl_common.h>
+
+#define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
+#define DENY ACE_ACCESS_DENIED_ACE_TYPE
+#define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
+#define MIN_ACE_TYPE ALLOW
+
+#define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
+#define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
+ ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
+#define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
+ ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
+#define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
+ ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
+#define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
+
+#define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
+ ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
+ ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
+ ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
+
+#define WRITE_MASK (WRITE_MASK_DATA|ACE_WRITE_ATTRIBUTES|ACE_WRITE_ACL|\
+ ACE_WRITE_OWNER|ACE_DELETE|ACE_DELETE_CHILD)
+
+#define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
+ ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
+
+#define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
+ ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
+
+#define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
+ ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
+
+#define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
+
+#define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
+ ZFS_ACL_PROTECTED)
+
+#define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
+ ZFS_ACL_OBJ_ACE)
+
+static uint16_t
+zfs_ace_v0_get_type(void *acep)
+{
+ return (((zfs_oldace_t *)acep)->z_type);
+}
+
+static uint16_t
+zfs_ace_v0_get_flags(void *acep)
+{
+ return (((zfs_oldace_t *)acep)->z_flags);
+}
+
+static uint32_t
+zfs_ace_v0_get_mask(void *acep)
+{
+ return (((zfs_oldace_t *)acep)->z_access_mask);
+}
+
+static uint64_t
+zfs_ace_v0_get_who(void *acep)
+{
+ return (((zfs_oldace_t *)acep)->z_fuid);
+}
+
+static void
+zfs_ace_v0_set_type(void *acep, uint16_t type)
+{
+ ((zfs_oldace_t *)acep)->z_type = type;
+}
+
+static void
+zfs_ace_v0_set_flags(void *acep, uint16_t flags)
+{
+ ((zfs_oldace_t *)acep)->z_flags = flags;
+}
+
+static void
+zfs_ace_v0_set_mask(void *acep, uint32_t mask)
+{
+ ((zfs_oldace_t *)acep)->z_access_mask = mask;
+}
+
+static void
+zfs_ace_v0_set_who(void *acep, uint64_t who)
+{
+ ((zfs_oldace_t *)acep)->z_fuid = who;
+}
+
+/*ARGSUSED*/
+static size_t
+zfs_ace_v0_size(void *acep)
+{
+ return (sizeof (zfs_oldace_t));
+}
+
+static size_t
+zfs_ace_v0_abstract_size(void)
+{
+ return (sizeof (zfs_oldace_t));
+}
+
+static int
+zfs_ace_v0_mask_off(void)
+{
+ return (offsetof(zfs_oldace_t, z_access_mask));
+}
+
+/*ARGSUSED*/
+static int
+zfs_ace_v0_data(void *acep, void **datap)
+{
+ *datap = NULL;
+ return (0);
+}
+
+static acl_ops_t zfs_acl_v0_ops = {
+ zfs_ace_v0_get_mask,
+ zfs_ace_v0_set_mask,
+ zfs_ace_v0_get_flags,
+ zfs_ace_v0_set_flags,
+ zfs_ace_v0_get_type,
+ zfs_ace_v0_set_type,
+ zfs_ace_v0_get_who,
+ zfs_ace_v0_set_who,
+ zfs_ace_v0_size,
+ zfs_ace_v0_abstract_size,
+ zfs_ace_v0_mask_off,
+ zfs_ace_v0_data
+};
+
+static uint16_t
+zfs_ace_fuid_get_type(void *acep)
+{
+ return (((zfs_ace_hdr_t *)acep)->z_type);
+}
+
+static uint16_t
+zfs_ace_fuid_get_flags(void *acep)
+{
+ return (((zfs_ace_hdr_t *)acep)->z_flags);
+}
+
+static uint32_t
+zfs_ace_fuid_get_mask(void *acep)
+{
+ return (((zfs_ace_hdr_t *)acep)->z_access_mask);
+}
+
+static uint64_t
+zfs_ace_fuid_get_who(void *args)
+{
+ uint16_t entry_type;
+ zfs_ace_t *acep = args;
+
+ entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
+
+ if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
+ entry_type == ACE_EVERYONE)
+ return (-1);
+ return (((zfs_ace_t *)acep)->z_fuid);
+}
+
+static void
+zfs_ace_fuid_set_type(void *acep, uint16_t type)
+{
+ ((zfs_ace_hdr_t *)acep)->z_type = type;
+}
+
+static void
+zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
+{
+ ((zfs_ace_hdr_t *)acep)->z_flags = flags;
+}
+
+static void
+zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
+{
+ ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
+}
+
+static void
+zfs_ace_fuid_set_who(void *arg, uint64_t who)
+{
+ zfs_ace_t *acep = arg;
+
+ uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
+
+ if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
+ entry_type == ACE_EVERYONE)
+ return;
+ acep->z_fuid = who;
+}
+
+static size_t
+zfs_ace_fuid_size(void *acep)
+{
+ zfs_ace_hdr_t *zacep = acep;
+ uint16_t entry_type;
+
+ switch (zacep->z_type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ return (sizeof (zfs_object_ace_t));
+ case ALLOW:
+ case DENY:
+ entry_type =
+ (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
+ if (entry_type == ACE_OWNER ||
+ entry_type == OWNING_GROUP ||
+ entry_type == ACE_EVERYONE)
+ return (sizeof (zfs_ace_hdr_t));
+ /*FALLTHROUGH*/
+ default:
+ return (sizeof (zfs_ace_t));
+ }
+}
+
+static size_t
+zfs_ace_fuid_abstract_size(void)
+{
+ return (sizeof (zfs_ace_hdr_t));
+}
+
+static int
+zfs_ace_fuid_mask_off(void)
+{
+ return (offsetof(zfs_ace_hdr_t, z_access_mask));
+}
+
+static int
+zfs_ace_fuid_data(void *acep, void **datap)
+{
+ zfs_ace_t *zacep = acep;
+ zfs_object_ace_t *zobjp;
+
+ switch (zacep->z_hdr.z_type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ zobjp = acep;
+ *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
+ return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
+ default:
+ *datap = NULL;
+ return (0);
+ }
+}
+
+static acl_ops_t zfs_acl_fuid_ops = {
+ zfs_ace_fuid_get_mask,
+ zfs_ace_fuid_set_mask,
+ zfs_ace_fuid_get_flags,
+ zfs_ace_fuid_set_flags,
+ zfs_ace_fuid_get_type,
+ zfs_ace_fuid_set_type,
+ zfs_ace_fuid_get_who,
+ zfs_ace_fuid_set_who,
+ zfs_ace_fuid_size,
+ zfs_ace_fuid_abstract_size,
+ zfs_ace_fuid_mask_off,
+ zfs_ace_fuid_data
+};
+
+static int
+zfs_acl_version(int version)
+{
+ if (version < ZPL_VERSION_FUID)
+ return (ZFS_ACL_VERSION_INITIAL);
+ else
+ return (ZFS_ACL_VERSION_FUID);
+}
+
+static int
+zfs_acl_version_zp(znode_t *zp)
+{
+ return (zfs_acl_version(zp->z_zfsvfs->z_version));
+}
+
+static zfs_acl_t *
+zfs_acl_alloc(int vers)
+{
+ zfs_acl_t *aclp;
+
+ aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
+ list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
+ offsetof(zfs_acl_node_t, z_next));
+ aclp->z_version = vers;
+ if (vers == ZFS_ACL_VERSION_FUID)
+ aclp->z_ops = zfs_acl_fuid_ops;
+ else
+ aclp->z_ops = zfs_acl_v0_ops;
+ return (aclp);
+}
+
+static zfs_acl_node_t *
+zfs_acl_node_alloc(size_t bytes)
+{
+ zfs_acl_node_t *aclnode;
+
+ aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
+ if (bytes) {
+ aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
+ aclnode->z_allocdata = aclnode->z_acldata;
+ aclnode->z_allocsize = bytes;
+ aclnode->z_size = bytes;
+ }
+
+ return (aclnode);
+}
+
+static void
+zfs_acl_node_free(zfs_acl_node_t *aclnode)
+{
+ if (aclnode->z_allocsize)
+ kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
+ kmem_free(aclnode, sizeof (zfs_acl_node_t));
+}
+
+static void
+zfs_acl_release_nodes(zfs_acl_t *aclp)
+{
+ zfs_acl_node_t *aclnode;
+
+ while (aclnode = list_head(&aclp->z_acl)) {
+ list_remove(&aclp->z_acl, aclnode);
+ zfs_acl_node_free(aclnode);
+ }
+ aclp->z_acl_count = 0;
+ aclp->z_acl_bytes = 0;
+}
+
+void
+zfs_acl_free(zfs_acl_t *aclp)
+{
+ zfs_acl_release_nodes(aclp);
+ list_destroy(&aclp->z_acl);
+ kmem_free(aclp, sizeof (zfs_acl_t));
+}
+
+static boolean_t
+zfs_acl_valid_ace_type(uint_t type, uint_t flags)
+{
+ uint16_t entry_type;
+
+ switch (type) {
+ case ALLOW:
+ case DENY:
+ case ACE_SYSTEM_AUDIT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_ACE_TYPE:
+ entry_type = flags & ACE_TYPE_FLAGS;
+ return (entry_type == ACE_OWNER ||
+ entry_type == OWNING_GROUP ||
+ entry_type == ACE_EVERYONE || entry_type == 0 ||
+ entry_type == ACE_IDENTIFIER_GROUP);
+ default:
+ if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
+ return (B_TRUE);
+ }
+ return (B_FALSE);
+}
+
+static boolean_t
+zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
+{
+ /*
+ * first check type of entry
+ */
+
+ if (!zfs_acl_valid_ace_type(type, iflags))
+ return (B_FALSE);
+
+ switch (type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ if (aclp->z_version < ZFS_ACL_VERSION_FUID)
+ return (B_FALSE);
+ aclp->z_hints |= ZFS_ACL_OBJ_ACE;
+ }
+
+ /*
+ * next check inheritance level flags
+ */
+
+ if (obj_type == VDIR &&
+ (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
+ aclp->z_hints |= ZFS_INHERIT_ACE;
+
+ if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
+ if ((iflags & (ACE_FILE_INHERIT_ACE|
+ ACE_DIRECTORY_INHERIT_ACE)) == 0) {
+ return (B_FALSE);
+ }
+ }
+
+ return (B_TRUE);
+}
+
+static void *
+zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
+ uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
+{
+ zfs_acl_node_t *aclnode;
+
+ if (start == NULL) {
+ aclnode = list_head(&aclp->z_acl);
+ if (aclnode == NULL)
+ return (NULL);
+
+ aclp->z_next_ace = aclnode->z_acldata;
+ aclp->z_curr_node = aclnode;
+ aclnode->z_ace_idx = 0;
+ }
+
+ aclnode = aclp->z_curr_node;
+
+ if (aclnode == NULL)
+ return (NULL);
+
+ if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
+ aclnode = list_next(&aclp->z_acl, aclnode);
+ if (aclnode == NULL)
+ return (NULL);
+ else {
+ aclp->z_curr_node = aclnode;
+ aclnode->z_ace_idx = 0;
+ aclp->z_next_ace = aclnode->z_acldata;
+ }
+ }
+
+ if (aclnode->z_ace_idx < aclnode->z_ace_count) {
+ void *acep = aclp->z_next_ace;
+ size_t ace_size;
+
+ /*
+ * Make sure we don't overstep our bounds
+ */
+ ace_size = aclp->z_ops.ace_size(acep);
+
+ if (((caddr_t)acep + ace_size) >
+ ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
+ return (NULL);
+ }
+
+ *iflags = aclp->z_ops.ace_flags_get(acep);
+ *type = aclp->z_ops.ace_type_get(acep);
+ *access_mask = aclp->z_ops.ace_mask_get(acep);
+ *who = aclp->z_ops.ace_who_get(acep);
+ aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
+ aclnode->z_ace_idx++;
+ return ((void *)acep);
+ }
+ return (NULL);
+}
+
+/*ARGSUSED*/
+static uint64_t
+zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
+ uint16_t *flags, uint16_t *type, uint32_t *mask)
+{
+ zfs_acl_t *aclp = datap;
+ zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
+ uint64_t who;
+
+ acep = zfs_acl_next_ace(aclp, acep, &who, mask,
+ flags, type);
+ return ((uint64_t)(uintptr_t)acep);
+}
+
+static zfs_acl_node_t *
+zfs_acl_curr_node(zfs_acl_t *aclp)
+{
+ ASSERT(aclp->z_curr_node);
+ return (aclp->z_curr_node);
+}
+
+/*
+ * Copy ACE to internal ZFS format.
+ * While processing the ACL each ACE will be validated for correctness.
+ * ACE FUIDs will be created later.
+ */
+int
+zfs_copy_ace_2_fuid(vtype_t obj_type, zfs_acl_t *aclp, void *datap,
+ zfs_ace_t *z_acl, int aclcnt, size_t *size)
+{
+ int i;
+ uint16_t entry_type;
+ zfs_ace_t *aceptr = z_acl;
+ ace_t *acep = datap;
+ zfs_object_ace_t *zobjacep;
+ ace_object_t *aceobjp;
+
+ for (i = 0; i != aclcnt; i++) {
+ aceptr->z_hdr.z_access_mask = acep->a_access_mask;
+ aceptr->z_hdr.z_flags = acep->a_flags;
+ aceptr->z_hdr.z_type = acep->a_type;
+ entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
+ if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
+ entry_type != ACE_EVERYONE) {
+ if (!aclp->z_has_fuids)
+ aclp->z_has_fuids = IS_EPHEMERAL(acep->a_who);
+ aceptr->z_fuid = (uint64_t)acep->a_who;
+ }
+
+ /*
+ * Make sure ACE is valid
+ */
+ if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
+ aceptr->z_hdr.z_flags) != B_TRUE)
+ return (EINVAL);
+
+ switch (acep->a_type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ zobjacep = (zfs_object_ace_t *)aceptr;
+ aceobjp = (ace_object_t *)acep;
+
+ bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
+ sizeof (aceobjp->a_obj_type));
+ bcopy(aceobjp->a_inherit_obj_type,
+ zobjacep->z_inherit_type,
+ sizeof (aceobjp->a_inherit_obj_type));
+ acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
+ break;
+ default:
+ acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
+ }
+
+ aceptr = (zfs_ace_t *)((caddr_t)aceptr +
+ aclp->z_ops.ace_size(aceptr));
+ }
+
+ *size = (caddr_t)aceptr - (caddr_t)z_acl;
+
+ return (0);
+}
+
+/*
+ * Copy ZFS ACEs to fixed size ace_t layout
+ */
+static void
+zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
+ void *datap, int filter)
+{
+ uint64_t who;
+ uint32_t access_mask;
+ uint16_t iflags, type;
+ zfs_ace_hdr_t *zacep = NULL;
+ ace_t *acep = datap;
+ ace_object_t *objacep;
+ zfs_object_ace_t *zobjacep;
+ size_t ace_size;
+ uint16_t entry_type;
+
+ while (zacep = zfs_acl_next_ace(aclp, zacep,
+ &who, &access_mask, &iflags, &type)) {
+
+ switch (type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ if (filter) {
+ continue;
+ }
+ zobjacep = (zfs_object_ace_t *)zacep;
+ objacep = (ace_object_t *)acep;
+ bcopy(zobjacep->z_object_type,
+ objacep->a_obj_type,
+ sizeof (zobjacep->z_object_type));
+ bcopy(zobjacep->z_inherit_type,
+ objacep->a_inherit_obj_type,
+ sizeof (zobjacep->z_inherit_type));
+ ace_size = sizeof (ace_object_t);
+ break;
+ default:
+ ace_size = sizeof (ace_t);
+ break;
+ }
+
+ entry_type = (iflags & ACE_TYPE_FLAGS);
+ if ((entry_type != ACE_OWNER &&
+ entry_type != OWNING_GROUP &&
+ entry_type != ACE_EVERYONE)) {
+ acep->a_who = zfs_fuid_map_id(zfsvfs, who,
+ cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
+ ZFS_ACE_GROUP : ZFS_ACE_USER);
+ } else {
+ acep->a_who = (uid_t)(int64_t)who;
+ }
+ acep->a_access_mask = access_mask;
+ acep->a_flags = iflags;
+ acep->a_type = type;
+ acep = (ace_t *)((caddr_t)acep + ace_size);
+ }
+}
+
+static int
+zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
+ zfs_oldace_t *z_acl, int aclcnt, size_t *size)
+{
+ int i;
+ zfs_oldace_t *aceptr = z_acl;
+
+ for (i = 0; i != aclcnt; i++, aceptr++) {
+ aceptr->z_access_mask = acep[i].a_access_mask;
+ aceptr->z_type = acep[i].a_type;
+ aceptr->z_flags = acep[i].a_flags;
+ aceptr->z_fuid = acep[i].a_who;
+ /*
+ * Make sure ACE is valid
+ */
+ if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
+ aceptr->z_flags) != B_TRUE)
+ return (EINVAL);
+ }
+ *size = (caddr_t)aceptr - (caddr_t)z_acl;
+ return (0);
+}
+
+/*
+ * convert old ACL format to new
+ */
+void
+zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp)
+{
+ zfs_oldace_t *oldaclp;
+ int i;
+ uint16_t type, iflags;
+ uint32_t access_mask;
+ uint64_t who;
+ void *cookie = NULL;
+ zfs_acl_node_t *newaclnode;
+
+ ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
+ /*
+ * First create the ACE in a contiguous piece of memory
+ * for zfs_copy_ace_2_fuid().
+ *
+ * We only convert an ACL once, so this won't happen
+ * everytime.
+ */
+ oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
+ KM_SLEEP);
+ i = 0;
+ while (cookie = zfs_acl_next_ace(aclp, cookie, &who,
+ &access_mask, &iflags, &type)) {
+ oldaclp[i].z_flags = iflags;
+ oldaclp[i].z_type = type;
+ oldaclp[i].z_fuid = who;
+ oldaclp[i++].z_access_mask = access_mask;
+ }
+
+ newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
+ sizeof (zfs_object_ace_t));
+ aclp->z_ops = zfs_acl_fuid_ops;
+ VERIFY(zfs_copy_ace_2_fuid(ZTOV(zp)->v_type, aclp, oldaclp,
+ newaclnode->z_acldata, aclp->z_acl_count,
+ &newaclnode->z_size) == 0);
+ newaclnode->z_ace_count = aclp->z_acl_count;
+ aclp->z_version = ZFS_ACL_VERSION;
+ kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
+
+ /*
+ * Release all previous ACL nodes
+ */
+
+ zfs_acl_release_nodes(aclp);
+
+ list_insert_head(&aclp->z_acl, newaclnode);
+
+ aclp->z_acl_bytes = newaclnode->z_size;
+ aclp->z_acl_count = newaclnode->z_ace_count;
+
+}
+
+/*
+ * Convert unix access mask to v4 access mask
+ */
+static uint32_t
+zfs_unix_to_v4(uint32_t access_mask)
+{
+ uint32_t new_mask = 0;
+
+ if (access_mask & S_IXOTH)
+ new_mask |= ACE_EXECUTE;
+ if (access_mask & S_IWOTH)
+ new_mask |= ACE_WRITE_DATA;
+ if (access_mask & S_IROTH)
+ new_mask |= ACE_READ_DATA;
+ return (new_mask);
+}
+
+static void
+zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
+ uint16_t access_type, uint64_t fuid, uint16_t entry_type)
+{
+ uint16_t type = entry_type & ACE_TYPE_FLAGS;
+
+ aclp->z_ops.ace_mask_set(acep, access_mask);
+ aclp->z_ops.ace_type_set(acep, access_type);
+ aclp->z_ops.ace_flags_set(acep, entry_type);
+ if ((type != ACE_OWNER && type != OWNING_GROUP &&
+ type != ACE_EVERYONE))
+ aclp->z_ops.ace_who_set(acep, fuid);
+}
+
+/*
+ * Determine mode of file based on ACL.
+ * Also, create FUIDs for any User/Group ACEs
+ */
+static uint64_t
+zfs_mode_fuid_compute(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
+ zfs_fuid_info_t **fuidp, dmu_tx_t *tx)
+{
+ int entry_type;
+ mode_t mode;
+ mode_t seen = 0;
+ zfs_ace_hdr_t *acep = NULL;
+ uint64_t who;
+ uint16_t iflags, type;
+ uint32_t access_mask;
+
+ mode = (zp->z_phys->zp_mode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
+
+ while (acep = zfs_acl_next_ace(aclp, acep, &who,
+ &access_mask, &iflags, &type)) {
+
+ if (!zfs_acl_valid_ace_type(type, iflags))
+ continue;
+
+ entry_type = (iflags & ACE_TYPE_FLAGS);
+
+ /*
+ * Skip over owner@, group@ or everyone@ inherit only ACEs
+ */
+ if ((iflags & ACE_INHERIT_ONLY_ACE) &&
+ (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
+ entry_type == OWNING_GROUP))
+ continue;
+
+ if (entry_type == ACE_OWNER) {
+ if ((access_mask & ACE_READ_DATA) &&
+ (!(seen & S_IRUSR))) {
+ seen |= S_IRUSR;
+ if (type == ALLOW) {
+ mode |= S_IRUSR;
+ }
+ }
+ if ((access_mask & ACE_WRITE_DATA) &&
+ (!(seen & S_IWUSR))) {
+ seen |= S_IWUSR;
+ if (type == ALLOW) {
+ mode |= S_IWUSR;
+ }
+ }
+ if ((access_mask & ACE_EXECUTE) &&
+ (!(seen & S_IXUSR))) {
+ seen |= S_IXUSR;
+ if (type == ALLOW) {
+ mode |= S_IXUSR;
+ }
+ }
+ } else if (entry_type == OWNING_GROUP) {
+ if ((access_mask & ACE_READ_DATA) &&
+ (!(seen & S_IRGRP))) {
+ seen |= S_IRGRP;
+ if (type == ALLOW) {
+ mode |= S_IRGRP;
+ }
+ }
+ if ((access_mask & ACE_WRITE_DATA) &&
+ (!(seen & S_IWGRP))) {
+ seen |= S_IWGRP;
+ if (type == ALLOW) {
+ mode |= S_IWGRP;
+ }
+ }
+ if ((access_mask & ACE_EXECUTE) &&
+ (!(seen & S_IXGRP))) {
+ seen |= S_IXGRP;
+ if (type == ALLOW) {
+ mode |= S_IXGRP;
+ }
+ }
+ } else if (entry_type == ACE_EVERYONE) {
+ if ((access_mask & ACE_READ_DATA)) {
+ if (!(seen & S_IRUSR)) {
+ seen |= S_IRUSR;
+ if (type == ALLOW) {
+ mode |= S_IRUSR;
+ }
+ }
+ if (!(seen & S_IRGRP)) {
+ seen |= S_IRGRP;
+ if (type == ALLOW) {
+ mode |= S_IRGRP;
+ }
+ }
+ if (!(seen & S_IROTH)) {
+ seen |= S_IROTH;
+ if (type == ALLOW) {
+ mode |= S_IROTH;
+ }
+ }
+ }
+ if ((access_mask & ACE_WRITE_DATA)) {
+ if (!(seen & S_IWUSR)) {
+ seen |= S_IWUSR;
+ if (type == ALLOW) {
+ mode |= S_IWUSR;
+ }
+ }
+ if (!(seen & S_IWGRP)) {
+ seen |= S_IWGRP;
+ if (type == ALLOW) {
+ mode |= S_IWGRP;
+ }
+ }
+ if (!(seen & S_IWOTH)) {
+ seen |= S_IWOTH;
+ if (type == ALLOW) {
+ mode |= S_IWOTH;
+ }
+ }
+ }
+ if ((access_mask & ACE_EXECUTE)) {
+ if (!(seen & S_IXUSR)) {
+ seen |= S_IXUSR;
+ if (type == ALLOW) {
+ mode |= S_IXUSR;
+ }
+ }
+ if (!(seen & S_IXGRP)) {
+ seen |= S_IXGRP;
+ if (type == ALLOW) {
+ mode |= S_IXGRP;
+ }
+ }
+ if (!(seen & S_IXOTH)) {
+ seen |= S_IXOTH;
+ if (type == ALLOW) {
+ mode |= S_IXOTH;
+ }
+ }
+ }
+ }
+ /*
+ * Now handle FUID create for user/group ACEs
+ */
+ if (entry_type == 0 || entry_type == ACE_IDENTIFIER_GROUP) {
+ aclp->z_ops.ace_who_set(acep,
+ zfs_fuid_create(zp->z_zfsvfs, who, cr,
+ (entry_type == 0) ? ZFS_ACE_USER : ZFS_ACE_GROUP,
+ tx, fuidp));
+ }
+ }
+ return (mode);
+}
+
+static zfs_acl_t *
+zfs_acl_node_read_internal(znode_t *zp, boolean_t will_modify)
+{
+ zfs_acl_t *aclp;
+ zfs_acl_node_t *aclnode;
+
+ aclp = zfs_acl_alloc(zp->z_phys->zp_acl.z_acl_version);
+
+ /*
+ * Version 0 to 1 znode_acl_phys has the size/count fields swapped.
+ * Version 0 didn't have a size field, only a count.
+ */
+ if (zp->z_phys->zp_acl.z_acl_version == ZFS_ACL_VERSION_INITIAL) {
+ aclp->z_acl_count = zp->z_phys->zp_acl.z_acl_size;
+ aclp->z_acl_bytes = ZFS_ACL_SIZE(aclp->z_acl_count);
+ } else {
+ aclp->z_acl_count = zp->z_phys->zp_acl.z_acl_count;
+ aclp->z_acl_bytes = zp->z_phys->zp_acl.z_acl_size;
+ }
+
+ aclnode = zfs_acl_node_alloc(will_modify ? aclp->z_acl_bytes : 0);
+ aclnode->z_ace_count = aclp->z_acl_count;
+ if (will_modify) {
+ bcopy(zp->z_phys->zp_acl.z_ace_data, aclnode->z_acldata,
+ aclp->z_acl_bytes);
+ } else {
+ aclnode->z_size = aclp->z_acl_bytes;
+ aclnode->z_acldata = &zp->z_phys->zp_acl.z_ace_data[0];
+ }
+
+ list_insert_head(&aclp->z_acl, aclnode);
+
+ return (aclp);
+}
+
+/*
+ * Read an external acl object.
+ */
+static int
+zfs_acl_node_read(znode_t *zp, zfs_acl_t **aclpp, boolean_t will_modify)
+{
+ uint64_t extacl = zp->z_phys->zp_acl.z_acl_extern_obj;
+ zfs_acl_t *aclp;
+ size_t aclsize;
+ size_t acl_count;
+ zfs_acl_node_t *aclnode;
+ int error;
+
+ ASSERT(MUTEX_HELD(&zp->z_acl_lock));
+
+ if (zp->z_phys->zp_acl.z_acl_extern_obj == 0) {
+ *aclpp = zfs_acl_node_read_internal(zp, will_modify);
+ return (0);
+ }
+
+ aclp = zfs_acl_alloc(zp->z_phys->zp_acl.z_acl_version);
+ if (zp->z_phys->zp_acl.z_acl_version == ZFS_ACL_VERSION_INITIAL) {
+ zfs_acl_phys_v0_t *zacl0 =
+ (zfs_acl_phys_v0_t *)&zp->z_phys->zp_acl;
+
+ aclsize = ZFS_ACL_SIZE(zacl0->z_acl_count);
+ acl_count = zacl0->z_acl_count;
+ } else {
+ aclsize = zp->z_phys->zp_acl.z_acl_size;
+ acl_count = zp->z_phys->zp_acl.z_acl_count;
+ if (aclsize == 0)
+ aclsize = acl_count * sizeof (zfs_ace_t);
+ }
+ aclnode = zfs_acl_node_alloc(aclsize);
+ list_insert_head(&aclp->z_acl, aclnode);
+ error = dmu_read(zp->z_zfsvfs->z_os, extacl, 0,
+ aclsize, aclnode->z_acldata);
+ aclnode->z_ace_count = acl_count;
+ aclp->z_acl_count = acl_count;
+ aclp->z_acl_bytes = aclsize;
+
+ if (error != 0) {
+ zfs_acl_free(aclp);
+ /* convert checksum errors into IO errors */
+ if (error == ECKSUM)
+ error = EIO;
+ return (error);
+ }
+
+ *aclpp = aclp;
+ return (0);
+}
+
+/*
+ * common code for setting ACLs.
+ *
+ * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
+ * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
+ * already checked the acl and knows whether to inherit.
+ */
+int
+zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
+ zfs_fuid_info_t **fuidp, dmu_tx_t *tx)
+{
+ int error;
+ znode_phys_t *zphys = zp->z_phys;
+ zfs_acl_phys_t *zacl = &zphys->zp_acl;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ uint64_t aoid = zphys->zp_acl.z_acl_extern_obj;
+ uint64_t off = 0;
+ dmu_object_type_t otype;
+ zfs_acl_node_t *aclnode;
+
+ ASSERT(MUTEX_HELD(&zp->z_lock));
+ ASSERT(MUTEX_HELD(&zp->z_acl_lock));
+
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+
+ zphys->zp_mode = zfs_mode_fuid_compute(zp, aclp, cr, fuidp, tx);
+
+ /*
+ * Decide which opbject type to use. If we are forced to
+ * use old ACL format than transform ACL into zfs_oldace_t
+ * layout.
+ */
+ if (!zfsvfs->z_use_fuids) {
+ otype = DMU_OT_OLDACL;
+ } else {
+ if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
+ (zfsvfs->z_version >= ZPL_VERSION_FUID))
+ zfs_acl_xform(zp, aclp);
+ ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
+ otype = DMU_OT_ACL;
+ }
+
+ if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
+ /*
+ * If ACL was previously external and we are now
+ * converting to new ACL format then release old
+ * ACL object and create a new one.
+ */
+ if (aoid && aclp->z_version != zacl->z_acl_version) {
+ error = dmu_object_free(zfsvfs->z_os,
+ zp->z_phys->zp_acl.z_acl_extern_obj, tx);
+ if (error)
+ return (error);
+ aoid = 0;
+ }
+ if (aoid == 0) {
+ aoid = dmu_object_alloc(zfsvfs->z_os,
+ otype, aclp->z_acl_bytes,
+ otype == DMU_OT_ACL ? DMU_OT_SYSACL : DMU_OT_NONE,
+ otype == DMU_OT_ACL ? DN_MAX_BONUSLEN : 0, tx);
+ } else {
+ (void) dmu_object_set_blocksize(zfsvfs->z_os, aoid,
+ aclp->z_acl_bytes, 0, tx);
+ }
+ zphys->zp_acl.z_acl_extern_obj = aoid;
+ for (aclnode = list_head(&aclp->z_acl); aclnode;
+ aclnode = list_next(&aclp->z_acl, aclnode)) {
+ if (aclnode->z_ace_count == 0)
+ continue;
+ dmu_write(zfsvfs->z_os, aoid, off,
+ aclnode->z_size, aclnode->z_acldata, tx);
+ off += aclnode->z_size;
+ }
+ } else {
+ void *start = zacl->z_ace_data;
+ /*
+ * Migrating back embedded?
+ */
+ if (zphys->zp_acl.z_acl_extern_obj) {
+ error = dmu_object_free(zfsvfs->z_os,
+ zp->z_phys->zp_acl.z_acl_extern_obj, tx);
+ if (error)
+ return (error);
+ zphys->zp_acl.z_acl_extern_obj = 0;
+ }
+
+ for (aclnode = list_head(&aclp->z_acl); aclnode;
+ aclnode = list_next(&aclp->z_acl, aclnode)) {
+ if (aclnode->z_ace_count == 0)
+ continue;
+ bcopy(aclnode->z_acldata, start, aclnode->z_size);
+ start = (caddr_t)start + aclnode->z_size;
+ }
+ }
+
+ /*
+ * If Old version then swap count/bytes to match old
+ * layout of znode_acl_phys_t.
+ */
+ if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
+ zphys->zp_acl.z_acl_size = aclp->z_acl_count;
+ zphys->zp_acl.z_acl_count = aclp->z_acl_bytes;
+ } else {
+ zphys->zp_acl.z_acl_size = aclp->z_acl_bytes;
+ zphys->zp_acl.z_acl_count = aclp->z_acl_count;
+ }
+
+ zphys->zp_acl.z_acl_version = aclp->z_version;
+
+ /*
+ * Replace ACL wide bits, but first clear them.
+ */
+ zp->z_phys->zp_flags &= ~ZFS_ACL_WIDE_FLAGS;
+
+ zp->z_phys->zp_flags |= aclp->z_hints;
+
+ if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
+ zp->z_phys->zp_flags |= ZFS_ACL_TRIVIAL;
+
+ zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
+ return (0);
+}
+
+/*
+ * Update access mask for prepended ACE
+ *
+ * This applies the "groupmask" value for aclmode property.
+ */
+static void
+zfs_acl_prepend_fixup(zfs_acl_t *aclp, void *acep, void *origacep,
+ mode_t mode, uint64_t owner)
+{
+ int rmask, wmask, xmask;
+ int user_ace;
+ uint16_t aceflags;
+ uint32_t origmask, acepmask;
+ uint64_t fuid;
+
+ aceflags = aclp->z_ops.ace_flags_get(acep);
+ fuid = aclp->z_ops.ace_who_get(acep);
+ origmask = aclp->z_ops.ace_mask_get(origacep);
+ acepmask = aclp->z_ops.ace_mask_get(acep);
+
+ user_ace = (!(aceflags &
+ (ACE_OWNER|ACE_GROUP|ACE_IDENTIFIER_GROUP)));
+
+ if (user_ace && (fuid == owner)) {
+ rmask = S_IRUSR;
+ wmask = S_IWUSR;
+ xmask = S_IXUSR;
+ } else {
+ rmask = S_IRGRP;
+ wmask = S_IWGRP;
+ xmask = S_IXGRP;
+ }
+
+ if (origmask & ACE_READ_DATA) {
+ if (mode & rmask) {
+ acepmask &= ~ACE_READ_DATA;
+ } else {
+ acepmask |= ACE_READ_DATA;
+ }
+ }
+
+ if (origmask & ACE_WRITE_DATA) {
+ if (mode & wmask) {
+ acepmask &= ~ACE_WRITE_DATA;
+ } else {
+ acepmask |= ACE_WRITE_DATA;
+ }
+ }
+
+ if (origmask & ACE_APPEND_DATA) {
+ if (mode & wmask) {
+ acepmask &= ~ACE_APPEND_DATA;
+ } else {
+ acepmask |= ACE_APPEND_DATA;
+ }
+ }
+
+ if (origmask & ACE_EXECUTE) {
+ if (mode & xmask) {
+ acepmask &= ~ACE_EXECUTE;
+ } else {
+ acepmask |= ACE_EXECUTE;
+ }
+ }
+ aclp->z_ops.ace_mask_set(acep, acepmask);
+}
+
+/*
+ * Apply mode to canonical six ACEs.
+ */
+static void
+zfs_acl_fixup_canonical_six(zfs_acl_t *aclp, mode_t mode)
+{
+ zfs_acl_node_t *aclnode = list_tail(&aclp->z_acl);
+ void *acep;
+ int maskoff = aclp->z_ops.ace_mask_off();
+ size_t abstract_size = aclp->z_ops.ace_abstract_size();
+
+ ASSERT(aclnode != NULL);
+
+ acep = (void *)((caddr_t)aclnode->z_acldata +
+ aclnode->z_size - (abstract_size * 6));
+
+ /*
+ * Fixup final ACEs to match the mode
+ */
+
+ adjust_ace_pair_common(acep, maskoff, abstract_size,
+ (mode & 0700) >> 6); /* owner@ */
+
+ acep = (caddr_t)acep + (abstract_size * 2);
+
+ adjust_ace_pair_common(acep, maskoff, abstract_size,
+ (mode & 0070) >> 3); /* group@ */
+
+ acep = (caddr_t)acep + (abstract_size * 2);
+ adjust_ace_pair_common(acep, maskoff,
+ abstract_size, mode); /* everyone@ */
+}
+
+
+static int
+zfs_acl_ace_match(zfs_acl_t *aclp, void *acep, int allow_deny,
+ int entry_type, int accessmask)
+{
+ uint32_t mask = aclp->z_ops.ace_mask_get(acep);
+ uint16_t type = aclp->z_ops.ace_type_get(acep);
+ uint16_t flags = aclp->z_ops.ace_flags_get(acep);
+
+ return (mask == accessmask && type == allow_deny &&
+ ((flags & ACE_TYPE_FLAGS) == entry_type));
+}
+
+/*
+ * Can prepended ACE be reused?
+ */
+static int
+zfs_reuse_deny(zfs_acl_t *aclp, void *acep, void *prevacep)
+{
+ int okay_masks;
+ uint16_t prevtype;
+ uint16_t prevflags;
+ uint16_t flags;
+ uint32_t mask, prevmask;
+
+ if (prevacep == NULL)
+ return (B_FALSE);
+
+ prevtype = aclp->z_ops.ace_type_get(prevacep);
+ prevflags = aclp->z_ops.ace_flags_get(prevacep);
+ flags = aclp->z_ops.ace_flags_get(acep);
+ mask = aclp->z_ops.ace_mask_get(acep);
+ prevmask = aclp->z_ops.ace_mask_get(prevacep);
+
+ if (prevtype != DENY)
+ return (B_FALSE);
+
+ if (prevflags != (flags & ACE_IDENTIFIER_GROUP))
+ return (B_FALSE);
+
+ okay_masks = (mask & OKAY_MASK_BITS);
+
+ if (prevmask & ~okay_masks)
+ return (B_FALSE);
+
+ return (B_TRUE);
+}
+
+
+/*
+ * Insert new ACL node into chain of zfs_acl_node_t's
+ *
+ * This will result in two possible results.
+ * 1. If the ACL is currently just a single zfs_acl_node and
+ * we are prepending the entry then current acl node will have
+ * a new node inserted above it.
+ *
+ * 2. If we are inserting in the middle of current acl node then
+ * the current node will be split in two and new node will be inserted
+ * in between the two split nodes.
+ */
+static zfs_acl_node_t *
+zfs_acl_ace_insert(zfs_acl_t *aclp, void *acep)
+{
+ zfs_acl_node_t *newnode;
+ zfs_acl_node_t *trailernode = NULL;
+ zfs_acl_node_t *currnode = zfs_acl_curr_node(aclp);
+ int curr_idx = aclp->z_curr_node->z_ace_idx;
+ int trailer_count;
+ size_t oldsize;
+
+ newnode = zfs_acl_node_alloc(aclp->z_ops.ace_size(acep));
+ newnode->z_ace_count = 1;
+
+ oldsize = currnode->z_size;
+
+ if (curr_idx != 1) {
+ trailernode = zfs_acl_node_alloc(0);
+ trailernode->z_acldata = acep;
+
+ trailer_count = currnode->z_ace_count - curr_idx + 1;
+ currnode->z_ace_count = curr_idx - 1;
+ currnode->z_size = (caddr_t)acep - (caddr_t)currnode->z_acldata;
+ trailernode->z_size = oldsize - currnode->z_size;
+ trailernode->z_ace_count = trailer_count;
+ }
+
+ aclp->z_acl_count += 1;
+ aclp->z_acl_bytes += aclp->z_ops.ace_size(acep);
+
+ if (curr_idx == 1)
+ list_insert_before(&aclp->z_acl, currnode, newnode);
+ else
+ list_insert_after(&aclp->z_acl, currnode, newnode);
+ if (trailernode) {
+ list_insert_after(&aclp->z_acl, newnode, trailernode);
+ aclp->z_curr_node = trailernode;
+ trailernode->z_ace_idx = 1;
+ }
+
+ return (newnode);
+}
+
+/*
+ * Prepend deny ACE
+ */
+static void *
+zfs_acl_prepend_deny(znode_t *zp, zfs_acl_t *aclp, void *acep,
+ mode_t mode)
+{
+ zfs_acl_node_t *aclnode;
+ void *newacep;
+ uint64_t fuid;
+ uint16_t flags;
+
+ aclnode = zfs_acl_ace_insert(aclp, acep);
+ newacep = aclnode->z_acldata;
+ fuid = aclp->z_ops.ace_who_get(acep);
+ flags = aclp->z_ops.ace_flags_get(acep);
+ zfs_set_ace(aclp, newacep, 0, DENY, fuid, (flags & ACE_TYPE_FLAGS));
+ zfs_acl_prepend_fixup(aclp, newacep, acep, mode, zp->z_phys->zp_uid);
+
+ return (newacep);
+}
+
+/*
+ * Split an inherited ACE into inherit_only ACE
+ * and original ACE with inheritance flags stripped off.
+ */
+static void
+zfs_acl_split_ace(zfs_acl_t *aclp, zfs_ace_hdr_t *acep)
+{
+ zfs_acl_node_t *aclnode;
+ zfs_acl_node_t *currnode;
+ void *newacep;
+ uint16_t type, flags;
+ uint32_t mask;
+ uint64_t fuid;
+
+ type = aclp->z_ops.ace_type_get(acep);
+ flags = aclp->z_ops.ace_flags_get(acep);
+ mask = aclp->z_ops.ace_mask_get(acep);
+ fuid = aclp->z_ops.ace_who_get(acep);
+
+ aclnode = zfs_acl_ace_insert(aclp, acep);
+ newacep = aclnode->z_acldata;
+
+ aclp->z_ops.ace_type_set(newacep, type);
+ aclp->z_ops.ace_flags_set(newacep, flags | ACE_INHERIT_ONLY_ACE);
+ aclp->z_ops.ace_mask_set(newacep, mask);
+ aclp->z_ops.ace_type_set(newacep, type);
+ aclp->z_ops.ace_who_set(newacep, fuid);
+ aclp->z_next_ace = acep;
+ flags &= ~ALL_INHERIT;
+ aclp->z_ops.ace_flags_set(acep, flags);
+ currnode = zfs_acl_curr_node(aclp);
+ ASSERT(currnode->z_ace_idx >= 1);
+ currnode->z_ace_idx -= 1;
+}
+
+/*
+ * Are ACES started at index i, the canonical six ACES?
+ */
+static int
+zfs_have_canonical_six(zfs_acl_t *aclp)
+{
+ void *acep;
+ zfs_acl_node_t *aclnode = list_tail(&aclp->z_acl);
+ int i = 0;
+ size_t abstract_size = aclp->z_ops.ace_abstract_size();
+
+ ASSERT(aclnode != NULL);
+
+ if (aclnode->z_ace_count < 6)
+ return (0);
+
+ acep = (void *)((caddr_t)aclnode->z_acldata +
+ aclnode->z_size - (aclp->z_ops.ace_abstract_size() * 6));
+
+ if ((zfs_acl_ace_match(aclp, (caddr_t)acep + (abstract_size * i++),
+ DENY, ACE_OWNER, 0) &&
+ zfs_acl_ace_match(aclp, (caddr_t)acep + (abstract_size * i++),
+ ALLOW, ACE_OWNER, OWNER_ALLOW_MASK) &&
+ zfs_acl_ace_match(aclp, (caddr_t)acep + (abstract_size * i++), DENY,
+ OWNING_GROUP, 0) && zfs_acl_ace_match(aclp, (caddr_t)acep +
+ (abstract_size * i++),
+ ALLOW, OWNING_GROUP, 0) &&
+ zfs_acl_ace_match(aclp, (caddr_t)acep + (abstract_size * i++),
+ DENY, ACE_EVERYONE, EVERYONE_DENY_MASK) &&
+ zfs_acl_ace_match(aclp, (caddr_t)acep + (abstract_size * i++),
+ ALLOW, ACE_EVERYONE, EVERYONE_ALLOW_MASK))) {
+ return (1);
+ } else {
+ return (0);
+ }
+}
+
+
+/*
+ * Apply step 1g, to group entries
+ *
+ * Need to deal with corner case where group may have
+ * greater permissions than owner. If so then limit
+ * group permissions, based on what extra permissions
+ * group has.
+ */
+static void
+zfs_fixup_group_entries(zfs_acl_t *aclp, void *acep, void *prevacep,
+ mode_t mode)
+{
+ uint32_t prevmask = aclp->z_ops.ace_mask_get(prevacep);
+ uint32_t mask = aclp->z_ops.ace_mask_get(acep);
+ uint16_t prevflags = aclp->z_ops.ace_flags_get(prevacep);
+ mode_t extramode = (mode >> 3) & 07;
+ mode_t ownermode = (mode >> 6);
+
+ if (prevflags & ACE_IDENTIFIER_GROUP) {
+
+ extramode &= ~ownermode;
+
+ if (extramode) {
+ if (extramode & S_IROTH) {
+ prevmask &= ~ACE_READ_DATA;
+ mask &= ~ACE_READ_DATA;
+ }
+ if (extramode & S_IWOTH) {
+ prevmask &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
+ mask &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
+ }
+ if (extramode & S_IXOTH) {
+ prevmask &= ~ACE_EXECUTE;
+ mask &= ~ACE_EXECUTE;
+ }
+ }
+ }
+ aclp->z_ops.ace_mask_set(acep, mask);
+ aclp->z_ops.ace_mask_set(prevacep, prevmask);
+}
+
+/*
+ * Apply the chmod algorithm as described
+ * in PSARC/2002/240
+ */
+static void
+zfs_acl_chmod(znode_t *zp, uint64_t mode, zfs_acl_t *aclp)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ void *acep = NULL, *prevacep = NULL;
+ uint64_t who;
+ int i;
+ int entry_type;
+ int reuse_deny;
+ int need_canonical_six = 1;
+ uint16_t iflags, type;
+ uint32_t access_mask;
+
+ ASSERT(MUTEX_HELD(&zp->z_acl_lock));
+ ASSERT(MUTEX_HELD(&zp->z_lock));
+
+ aclp->z_hints = (zp->z_phys->zp_flags & V4_ACL_WIDE_FLAGS);
+
+ /*
+ * If discard then just discard all ACL nodes which
+ * represent the ACEs.
+ *
+ * New owner@/group@/everone@ ACEs will be added
+ * later.
+ */
+ if (zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
+ zfs_acl_release_nodes(aclp);
+
+ while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
+ &iflags, &type)) {
+
+ entry_type = (iflags & ACE_TYPE_FLAGS);
+ iflags = (iflags & ALL_INHERIT);
+
+ if ((type != ALLOW && type != DENY) ||
+ (iflags & ACE_INHERIT_ONLY_ACE)) {
+ if (iflags)
+ aclp->z_hints |= ZFS_INHERIT_ACE;
+ switch (type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ aclp->z_hints |= ZFS_ACL_OBJ_ACE;
+ break;
+ }
+ goto nextace;
+ }
+
+ /*
+ * Need to split ace into two?
+ */
+ if ((iflags & (ACE_FILE_INHERIT_ACE|
+ ACE_DIRECTORY_INHERIT_ACE)) &&
+ (!(iflags & ACE_INHERIT_ONLY_ACE))) {
+ zfs_acl_split_ace(aclp, acep);
+ aclp->z_hints |= ZFS_INHERIT_ACE;
+ goto nextace;
+ }
+
+ if (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
+ (entry_type == OWNING_GROUP)) {
+ access_mask &= ~OGE_CLEAR;
+ aclp->z_ops.ace_mask_set(acep, access_mask);
+ goto nextace;
+ } else {
+ reuse_deny = B_TRUE;
+ if (type == ALLOW) {
+
+ /*
+ * Check preceding ACE if any, to see
+ * if we need to prepend a DENY ACE.
+ * This is only applicable when the acl_mode
+ * property == groupmask.
+ */
+ if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK) {
+
+ reuse_deny = zfs_reuse_deny(aclp, acep,
+ prevacep);
+
+ if (!reuse_deny) {
+ prevacep =
+ zfs_acl_prepend_deny(zp,
+ aclp, acep, mode);
+ } else {
+ zfs_acl_prepend_fixup(
+ aclp, prevacep,
+ acep, mode,
+ zp->z_phys->zp_uid);
+ }
+ zfs_fixup_group_entries(aclp, acep,
+ prevacep, mode);
+
+ }
+ }
+ }
+nextace:
+ prevacep = acep;
+ }
+
+ /*
+ * Check out last six aces, if we have six.
+ */
+
+ if (aclp->z_acl_count >= 6) {
+ if (zfs_have_canonical_six(aclp)) {
+ need_canonical_six = 0;
+ }
+ }
+
+ if (need_canonical_six) {
+ size_t abstract_size = aclp->z_ops.ace_abstract_size();
+ void *zacep;
+ zfs_acl_node_t *aclnode =
+ zfs_acl_node_alloc(abstract_size * 6);
+
+ aclnode->z_size = abstract_size * 6;
+ aclnode->z_ace_count = 6;
+ aclp->z_acl_bytes += aclnode->z_size;
+ list_insert_tail(&aclp->z_acl, aclnode);
+
+ zacep = aclnode->z_acldata;
+
+ i = 0;
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++),
+ 0, DENY, -1, ACE_OWNER);
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++),
+ OWNER_ALLOW_MASK, ALLOW, -1, ACE_OWNER);
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++), 0,
+ DENY, -1, OWNING_GROUP);
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++), 0,
+ ALLOW, -1, OWNING_GROUP);
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++),
+ EVERYONE_DENY_MASK, DENY, -1, ACE_EVERYONE);
+ zfs_set_ace(aclp, (caddr_t)zacep + (abstract_size * i++),
+ EVERYONE_ALLOW_MASK, ALLOW, -1, ACE_EVERYONE);
+ aclp->z_acl_count += 6;
+ }
+
+ zfs_acl_fixup_canonical_six(aclp, mode);
+}
+
+int
+zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
+{
+ int error;
+
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&zp->z_acl_lock);
+ *aclp = NULL;
+ error = zfs_acl_node_read(zp, aclp, B_TRUE);
+ if (error == 0)
+ zfs_acl_chmod(zp, mode, *aclp);
+ mutex_exit(&zp->z_acl_lock);
+ mutex_exit(&zp->z_lock);
+ return (error);
+}
+
+/*
+ * strip off write_owner and write_acl
+ */
+static void
+zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
+{
+ uint32_t mask = aclp->z_ops.ace_mask_get(acep);
+
+ if ((zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
+ (aclp->z_ops.ace_type_get(acep) == ALLOW)) {
+ mask &= ~RESTRICTED_CLEAR;
+ aclp->z_ops.ace_mask_set(acep, mask);
+ }
+}
+
+/*
+ * Should ACE be inherited?
+ */
+static int
+zfs_ace_can_use(znode_t *zp, uint16_t acep_flags)
+{
+ int vtype = ZTOV(zp)->v_type;
+ int iflags = (acep_flags & 0xf);
+
+ if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
+ return (1);
+ else if (iflags & ACE_FILE_INHERIT_ACE)
+ return (!((vtype == VDIR) &&
+ (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
+ return (0);
+}
+
+/*
+ * inherit inheritable ACEs from parent
+ */
+static zfs_acl_t *
+zfs_acl_inherit(znode_t *zp, zfs_acl_t *paclp, uint64_t mode,
+ boolean_t *need_chmod)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ void *pacep;
+ void *acep, *acep2;
+ zfs_acl_node_t *aclnode, *aclnode2;
+ zfs_acl_t *aclp = NULL;
+ uint64_t who;
+ uint32_t access_mask;
+ uint16_t iflags, newflags, type;
+ size_t ace_size;
+ void *data1, *data2;
+ size_t data1sz, data2sz;
+ boolean_t vdir = ZTOV(zp)->v_type == VDIR;
+ boolean_t vreg = ZTOV(zp)->v_type == VREG;
+ boolean_t passthrough, passthrough_x, noallow;
+
+ passthrough_x =
+ zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
+ passthrough = passthrough_x ||
+ zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
+ noallow =
+ zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
+
+ *need_chmod = B_TRUE;
+ pacep = NULL;
+ aclp = zfs_acl_alloc(paclp->z_version);
+ if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD)
+ return (aclp);
+ while (pacep = zfs_acl_next_ace(paclp, pacep, &who,
+ &access_mask, &iflags, &type)) {
+
+ /*
+ * don't inherit bogus ACEs
+ */
+ if (!zfs_acl_valid_ace_type(type, iflags))
+ continue;
+
+ if (noallow && type == ALLOW)
+ continue;
+
+ ace_size = aclp->z_ops.ace_size(pacep);
+
+ if (!zfs_ace_can_use(zp, iflags))
+ continue;
+
+ /*
+ * If owner@, group@, or everyone@ inheritable
+ * then zfs_acl_chmod() isn't needed.
+ */
+ if (passthrough &&
+ ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
+ ((iflags & OWNING_GROUP) ==
+ OWNING_GROUP)) && (vreg || (vdir && (iflags &
+ ACE_DIRECTORY_INHERIT_ACE)))) {
+ *need_chmod = B_FALSE;
+
+ if (!vdir && passthrough_x &&
+ ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
+ access_mask &= ~ACE_EXECUTE;
+ }
+ }
+
+ aclnode = zfs_acl_node_alloc(ace_size);
+ list_insert_tail(&aclp->z_acl, aclnode);
+ acep = aclnode->z_acldata;
+
+ zfs_set_ace(aclp, acep, access_mask, type,
+ who, iflags|ACE_INHERITED_ACE);
+
+ /*
+ * Copy special opaque data if any
+ */
+ if ((data1sz = paclp->z_ops.ace_data(pacep, &data1)) != 0) {
+ VERIFY((data2sz = aclp->z_ops.ace_data(acep,
+ &data2)) == data1sz);
+ bcopy(data1, data2, data2sz);
+ }
+ aclp->z_acl_count++;
+ aclnode->z_ace_count++;
+ aclp->z_acl_bytes += aclnode->z_size;
+ newflags = aclp->z_ops.ace_flags_get(acep);
+
+ if (vdir)
+ aclp->z_hints |= ZFS_INHERIT_ACE;
+
+ if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
+ newflags &= ~ALL_INHERIT;
+ aclp->z_ops.ace_flags_set(acep,
+ newflags|ACE_INHERITED_ACE);
+ zfs_restricted_update(zfsvfs, aclp, acep);
+ continue;
+ }
+
+ ASSERT(vdir);
+
+ newflags = aclp->z_ops.ace_flags_get(acep);
+ if ((iflags & (ACE_FILE_INHERIT_ACE |
+ ACE_DIRECTORY_INHERIT_ACE)) !=
+ ACE_FILE_INHERIT_ACE) {
+ aclnode2 = zfs_acl_node_alloc(ace_size);
+ list_insert_tail(&aclp->z_acl, aclnode2);
+ acep2 = aclnode2->z_acldata;
+ zfs_set_ace(aclp, acep2,
+ access_mask, type, who,
+ iflags|ACE_INHERITED_ACE);
+ newflags |= ACE_INHERIT_ONLY_ACE;
+ aclp->z_ops.ace_flags_set(acep, newflags);
+ newflags &= ~ALL_INHERIT;
+ aclp->z_ops.ace_flags_set(acep2,
+ newflags|ACE_INHERITED_ACE);
+
+ /*
+ * Copy special opaque data if any
+ */
+ if ((data1sz = aclp->z_ops.ace_data(acep,
+ &data1)) != 0) {
+ VERIFY((data2sz =
+ aclp->z_ops.ace_data(acep2,
+ &data2)) == data1sz);
+ bcopy(data1, data2, data1sz);
+ }
+ aclp->z_acl_count++;
+ aclnode2->z_ace_count++;
+ aclp->z_acl_bytes += aclnode->z_size;
+ zfs_restricted_update(zfsvfs, aclp, acep2);
+ } else {
+ newflags |= ACE_INHERIT_ONLY_ACE;
+ aclp->z_ops.ace_flags_set(acep,
+ newflags|ACE_INHERITED_ACE);
+ }
+ }
+ return (aclp);
+}
+
+/*
+ * Create file system object initial permissions
+ * including inheritable ACEs.
+ */
+void
+zfs_perm_init(znode_t *zp, znode_t *parent, int flag,
+ vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
+ zfs_acl_t *setaclp, zfs_fuid_info_t **fuidp)
+{
+ uint64_t mode, fuid, fgid;
+ int error;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zfs_acl_t *aclp = NULL;
+ zfs_acl_t *paclp;
+ xvattr_t *xvap = (xvattr_t *)vap;
+ gid_t gid;
+ boolean_t need_chmod = B_TRUE;
+
+ if (setaclp)
+ aclp = setaclp;
+
+ mode = MAKEIMODE(vap->va_type, vap->va_mode);
+
+ /*
+ * Determine uid and gid.
+ */
+ if ((flag & (IS_ROOT_NODE | IS_REPLAY)) ||
+ ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
+ fuid = zfs_fuid_create(zfsvfs, vap->va_uid, cr,
+ ZFS_OWNER, tx, fuidp);
+ fgid = zfs_fuid_create(zfsvfs, vap->va_gid, cr,
+ ZFS_GROUP, tx, fuidp);
+ gid = vap->va_gid;
+ } else {
+ fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER, tx, cr, fuidp);
+ fgid = 0;
+ if (vap->va_mask & AT_GID) {
+ fgid = zfs_fuid_create(zfsvfs, vap->va_gid, cr,
+ ZFS_GROUP, tx, fuidp);
+ gid = vap->va_gid;
+ if (fgid != parent->z_phys->zp_gid &&
+ !groupmember(vap->va_gid, cr) &&
+ secpolicy_vnode_create_gid(cr) != 0)
+ fgid = 0;
+ }
+ if (fgid == 0) {
+ if (parent->z_phys->zp_mode & S_ISGID) {
+ fgid = parent->z_phys->zp_gid;
+ gid = zfs_fuid_map_id(zfsvfs, fgid,
+ cr, ZFS_GROUP);
+ } else {
+ fgid = zfs_fuid_create_cred(zfsvfs,
+ ZFS_GROUP, tx, cr, fuidp);
+ gid = crgetgid(cr);
+ }
+ }
+ }
+
+ /*
+ * If we're creating a directory, and the parent directory has the
+ * set-GID bit set, set in on the new directory.
+ * Otherwise, if the user is neither privileged nor a member of the
+ * file's new group, clear the file's set-GID bit.
+ */
+
+ if ((parent->z_phys->zp_mode & S_ISGID) && (vap->va_type == VDIR)) {
+ mode |= S_ISGID;
+ } else {
+ if ((mode & S_ISGID) &&
+ secpolicy_vnode_setids_setgids(cr, gid) != 0)
+ mode &= ~S_ISGID;
+ }
+
+ zp->z_phys->zp_uid = fuid;
+ zp->z_phys->zp_gid = fgid;
+ zp->z_phys->zp_mode = mode;
+
+ if (aclp == NULL) {
+ mutex_enter(&parent->z_lock);
+ if ((ZTOV(parent)->v_type == VDIR &&
+ (parent->z_phys->zp_flags & ZFS_INHERIT_ACE)) &&
+ !(zp->z_phys->zp_flags & ZFS_XATTR)) {
+ mutex_enter(&parent->z_acl_lock);
+ VERIFY(0 == zfs_acl_node_read(parent, &paclp, B_FALSE));
+ mutex_exit(&parent->z_acl_lock);
+ aclp = zfs_acl_inherit(zp, paclp, mode, &need_chmod);
+ zfs_acl_free(paclp);
+ } else {
+ aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
+ }
+ mutex_exit(&parent->z_lock);
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&zp->z_acl_lock);
+ if (need_chmod)
+ zfs_acl_chmod(zp, mode, aclp);
+ } else {
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&zp->z_acl_lock);
+ }
+
+ /* Force auto_inherit on all new directory objects */
+ if (vap->va_type == VDIR)
+ aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
+
+ error = zfs_aclset_common(zp, aclp, cr, fuidp, tx);
+
+ /* Set optional attributes if any */
+ if (vap->va_mask & AT_XVATTR)
+ zfs_xvattr_set(zp, xvap);
+
+ mutex_exit(&zp->z_lock);
+ mutex_exit(&zp->z_acl_lock);
+ ASSERT3U(error, ==, 0);
+
+ if (aclp != setaclp)
+ zfs_acl_free(aclp);
+}
+
+/*
+ * Retrieve a files ACL
+ */
+int
+zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
+{
+ zfs_acl_t *aclp;
+ ulong_t mask;
+ int error;
+ int count = 0;
+ int largeace = 0;
+
+ mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
+ VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
+
+ if (error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr))
+ return (error);
+
+ if (mask == 0)
+ return (ENOSYS);
+
+ mutex_enter(&zp->z_acl_lock);
+
+ error = zfs_acl_node_read(zp, &aclp, B_FALSE);
+ if (error != 0) {
+ mutex_exit(&zp->z_acl_lock);
+ return (error);
+ }
+
+ /*
+ * Scan ACL to determine number of ACEs
+ */
+ if ((zp->z_phys->zp_flags & ZFS_ACL_OBJ_ACE) &&
+ !(mask & VSA_ACE_ALLTYPES)) {
+ void *zacep = NULL;
+ uint64_t who;
+ uint32_t access_mask;
+ uint16_t type, iflags;
+
+ while (zacep = zfs_acl_next_ace(aclp, zacep,
+ &who, &access_mask, &iflags, &type)) {
+ switch (type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ largeace++;
+ continue;
+ default:
+ count++;
+ }
+ }
+ vsecp->vsa_aclcnt = count;
+ } else
+ count = aclp->z_acl_count;
+
+ if (mask & VSA_ACECNT) {
+ vsecp->vsa_aclcnt = count;
+ }
+
+ if (mask & VSA_ACE) {
+ size_t aclsz;
+
+ zfs_acl_node_t *aclnode = list_head(&aclp->z_acl);
+
+ aclsz = count * sizeof (ace_t) +
+ sizeof (ace_object_t) * largeace;
+
+ vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
+ vsecp->vsa_aclentsz = aclsz;
+
+ if (aclp->z_version == ZFS_ACL_VERSION_FUID)
+ zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
+ vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
+ else {
+ bcopy(aclnode->z_acldata, vsecp->vsa_aclentp,
+ count * sizeof (ace_t));
+ }
+ }
+ if (mask & VSA_ACE_ACLFLAGS) {
+ vsecp->vsa_aclflags = 0;
+ if (zp->z_phys->zp_flags & ZFS_ACL_DEFAULTED)
+ vsecp->vsa_aclflags |= ACL_DEFAULTED;
+ if (zp->z_phys->zp_flags & ZFS_ACL_PROTECTED)
+ vsecp->vsa_aclflags |= ACL_PROTECTED;
+ if (zp->z_phys->zp_flags & ZFS_ACL_AUTO_INHERIT)
+ vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
+ }
+
+ mutex_exit(&zp->z_acl_lock);
+
+ zfs_acl_free(aclp);
+
+ return (0);
+}
+
+int
+zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
+ vsecattr_t *vsecp, zfs_acl_t **zaclp)
+{
+ zfs_acl_t *aclp;
+ zfs_acl_node_t *aclnode;
+ int aclcnt = vsecp->vsa_aclcnt;
+ int error;
+
+ if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
+ return (EINVAL);
+
+ aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
+
+ aclp->z_hints = 0;
+ aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
+ if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
+ if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
+ (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
+ aclcnt, &aclnode->z_size)) != 0) {
+ zfs_acl_free(aclp);
+ zfs_acl_node_free(aclnode);
+ return (error);
+ }
+ } else {
+ if ((error = zfs_copy_ace_2_fuid(obj_type, aclp,
+ vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
+ &aclnode->z_size)) != 0) {
+ zfs_acl_free(aclp);
+ zfs_acl_node_free(aclnode);
+ return (error);
+ }
+ }
+ aclp->z_acl_bytes = aclnode->z_size;
+ aclnode->z_ace_count = aclcnt;
+ aclp->z_acl_count = aclcnt;
+ list_insert_head(&aclp->z_acl, aclnode);
+
+ /*
+ * If flags are being set then add them to z_hints
+ */
+ if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
+ if (vsecp->vsa_aclflags & ACL_PROTECTED)
+ aclp->z_hints |= ZFS_ACL_PROTECTED;
+ if (vsecp->vsa_aclflags & ACL_DEFAULTED)
+ aclp->z_hints |= ZFS_ACL_DEFAULTED;
+ if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
+ aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
+ }
+
+ *zaclp = aclp;
+
+ return (0);
+}
+
+/*
+ * Set a files ACL
+ */
+int
+zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zilog_t *zilog = zfsvfs->z_log;
+ ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
+ dmu_tx_t *tx;
+ int error;
+ zfs_acl_t *aclp;
+ zfs_fuid_info_t *fuidp = NULL;
+
+ if (mask == 0)
+ return (ENOSYS);
+
+ if (zp->z_phys->zp_flags & ZFS_IMMUTABLE)
+ return (EPERM);
+
+ if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr))
+ return (error);
+
+ error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, &aclp);
+ if (error)
+ return (error);
+
+ /*
+ * If ACL wide flags aren't being set then preserve any
+ * existing flags.
+ */
+ if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
+ aclp->z_hints |= (zp->z_phys->zp_flags & V4_ACL_WIDE_FLAGS);
+ }
+top:
+ if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)) {
+ zfs_acl_free(aclp);
+ return (error);
+ }
+
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&zp->z_acl_lock);
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+
+ if (zp->z_phys->zp_acl.z_acl_extern_obj) {
+ /* Are we upgrading ACL? */
+ if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
+ zp->z_phys->zp_acl.z_acl_version ==
+ ZFS_ACL_VERSION_INITIAL) {
+ dmu_tx_hold_free(tx,
+ zp->z_phys->zp_acl.z_acl_extern_obj,
+ 0, DMU_OBJECT_END);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
+ 0, aclp->z_acl_bytes);
+ } else {
+ dmu_tx_hold_write(tx,
+ zp->z_phys->zp_acl.z_acl_extern_obj,
+ 0, aclp->z_acl_bytes);
+ }
+ } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
+ }
+ if (aclp->z_has_fuids) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ mutex_exit(&zp->z_acl_lock);
+ mutex_exit(&zp->z_lock);
+
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ zfs_acl_free(aclp);
+ return (error);
+ }
+
+ error = zfs_aclset_common(zp, aclp, cr, &fuidp, tx);
+ ASSERT(error == 0);
+
+ zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
+
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+ zfs_acl_free(aclp);
+ dmu_tx_commit(tx);
+done:
+ mutex_exit(&zp->z_acl_lock);
+ mutex_exit(&zp->z_lock);
+
+ return (error);
+}
+
+/*
+ * working_mode returns the permissions that were not granted
+ */
+static int
+zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
+ boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
+{
+ zfs_acl_t *aclp;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+ uid_t uid = crgetuid(cr);
+ uint64_t who;
+ uint16_t type, iflags;
+ uint16_t entry_type;
+ uint32_t access_mask;
+ uint32_t deny_mask = 0;
+ zfs_ace_hdr_t *acep = NULL;
+ boolean_t checkit;
+ uid_t fowner;
+ uid_t gowner;
+
+ /*
+ * Short circuit empty requests
+ */
+ if (v4_mode == 0)
+ return (0);
+
+ *check_privs = B_TRUE;
+
+ if (zfsvfs->z_assign >= TXG_INITIAL) { /* ZIL replay */
+ *working_mode = 0;
+ return (0);
+ }
+
+ *working_mode = v4_mode;
+
+ if ((v4_mode & WRITE_MASK) &&
+ (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
+ (!IS_DEVVP(ZTOV(zp)))) {
+ *check_privs = B_FALSE;
+ return (EROFS);
+ }
+
+ /*
+ * Only check for READONLY on non-directories.
+ */
+ if ((v4_mode & WRITE_MASK_DATA) &&
+ (((ZTOV(zp)->v_type != VDIR) &&
+ (zp->z_phys->zp_flags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
+ (ZTOV(zp)->v_type == VDIR &&
+ (zp->z_phys->zp_flags & ZFS_IMMUTABLE)))) {
+ *check_privs = B_FALSE;
+ return (EPERM);
+ }
+
+ if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
+ (zp->z_phys->zp_flags & ZFS_NOUNLINK)) {
+ *check_privs = B_FALSE;
+ return (EPERM);
+ }
+
+ if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
+ (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED))) {
+ *check_privs = B_FALSE;
+ return (EACCES);
+ }
+
+ /*
+ * The caller requested that the ACL check be skipped. This
+ * would only happen if the caller checked VOP_ACCESS() with a
+ * 32 bit ACE mask and already had the appropriate permissions.
+ */
+ if (skipaclchk) {
+ *working_mode = 0;
+ return (0);
+ }
+
+ zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
+
+ mutex_enter(&zp->z_acl_lock);
+
+ error = zfs_acl_node_read(zp, &aclp, B_FALSE);
+ if (error != 0) {
+ mutex_exit(&zp->z_acl_lock);
+ return (error);
+ }
+
+ while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
+ &iflags, &type)) {
+
+ if (!zfs_acl_valid_ace_type(type, iflags))
+ continue;
+
+ if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
+ continue;
+
+ entry_type = (iflags & ACE_TYPE_FLAGS);
+
+ checkit = B_FALSE;
+
+ switch (entry_type) {
+ case ACE_OWNER:
+ if (uid == fowner)
+ checkit = B_TRUE;
+ break;
+ case OWNING_GROUP:
+ who = gowner;
+ /*FALLTHROUGH*/
+ case ACE_IDENTIFIER_GROUP:
+ checkit = zfs_groupmember(zfsvfs, who, cr);
+ break;
+ case ACE_EVERYONE:
+ checkit = B_TRUE;
+ break;
+
+ /* USER Entry */
+ default:
+ if (entry_type == 0) {
+ uid_t newid;
+
+ newid = zfs_fuid_map_id(zfsvfs, who, cr,
+ ZFS_ACE_USER);
+ if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
+ uid == newid)
+ checkit = B_TRUE;
+ break;
+ } else {
+ zfs_acl_free(aclp);
+ mutex_exit(&zp->z_acl_lock);
+ return (EIO);
+ }
+ }
+
+ if (checkit) {
+ uint32_t mask_matched = (access_mask & *working_mode);
+
+ if (mask_matched) {
+ if (type == DENY)
+ deny_mask |= mask_matched;
+
+ *working_mode &= ~mask_matched;
+ }
+ }
+
+ /* Are we done? */
+ if (*working_mode == 0)
+ break;
+ }
+
+ mutex_exit(&zp->z_acl_lock);
+ zfs_acl_free(aclp);
+
+ /* Put the found 'denies' back on the working mode */
+ if (deny_mask) {
+ *working_mode |= deny_mask;
+ return (EACCES);
+ } else if (*working_mode) {
+ return (-1);
+ }
+
+ return (0);
+}
+
+static int
+zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
+ cred_t *cr)
+{
+ if (*working_mode != ACE_WRITE_DATA)
+ return (EACCES);
+
+ return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
+ check_privs, B_FALSE, cr));
+}
+
+/*
+ * Determine whether Access should be granted/denied, invoking least
+ * priv subsytem when a deny is determined.
+ */
+int
+zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
+{
+ uint32_t working_mode;
+ int error;
+ int is_attr;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ boolean_t check_privs;
+ znode_t *xzp;
+ znode_t *check_zp = zp;
+
+ is_attr = ((zp->z_phys->zp_flags & ZFS_XATTR) &&
+ (ZTOV(zp)->v_type == VDIR));
+
+ /*
+ * If attribute then validate against base file
+ */
+ if (is_attr) {
+ if ((error = zfs_zget(zp->z_zfsvfs,
+ zp->z_phys->zp_parent, &xzp)) != 0) {
+ return (error);
+ }
+
+ check_zp = xzp;
+
+ /*
+ * fixup mode to map to xattr perms
+ */
+
+ if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
+ mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
+ mode |= ACE_WRITE_NAMED_ATTRS;
+ }
+
+ if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
+ mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
+ mode |= ACE_READ_NAMED_ATTRS;
+ }
+ }
+
+ if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
+ &check_privs, skipaclchk, cr)) == 0) {
+ if (is_attr)
+ VN_RELE(ZTOV(xzp));
+ return (0);
+ }
+
+ if (error && !check_privs) {
+ if (is_attr)
+ VN_RELE(ZTOV(xzp));
+ return (error);
+ }
+
+ if (error && (flags & V_APPEND)) {
+ error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
+ }
+
+ if (error && check_privs) {
+ uid_t owner;
+ mode_t checkmode = 0;
+
+ owner = zfs_fuid_map_id(zfsvfs, check_zp->z_phys->zp_uid, cr,
+ ZFS_OWNER);
+
+ /*
+ * First check for implicit owner permission on
+ * read_acl/read_attributes
+ */
+
+ error = 0;
+ ASSERT(working_mode != 0);
+
+ if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
+ owner == crgetuid(cr)))
+ working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
+
+ if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
+ ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
+ checkmode |= VREAD;
+ if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
+ ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
+ checkmode |= VWRITE;
+ if (working_mode & ACE_EXECUTE)
+ checkmode |= VEXEC;
+
+ if (checkmode)
+ error = secpolicy_vnode_access(cr, ZTOV(check_zp),
+ owner, checkmode);
+
+ if (error == 0 && (working_mode & ACE_WRITE_OWNER))
+ error = secpolicy_vnode_chown(cr, B_TRUE);
+ if (error == 0 && (working_mode & ACE_WRITE_ACL))
+ error = secpolicy_vnode_setdac(cr, owner);
+
+ if (error == 0 && (working_mode &
+ (ACE_DELETE|ACE_DELETE_CHILD)))
+ error = secpolicy_vnode_remove(cr);
+
+ if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
+ error = secpolicy_vnode_chown(cr, B_FALSE);
+ }
+ if (error == 0) {
+ /*
+ * See if any bits other than those already checked
+ * for are still present. If so then return EACCES
+ */
+ if (working_mode & ~(ZFS_CHECKED_MASKS)) {
+ error = EACCES;
+ }
+ }
+ }
+
+ if (is_attr)
+ VN_RELE(ZTOV(xzp));
+
+ return (error);
+}
+
+/*
+ * Translate traditional unix VREAD/VWRITE/VEXEC mode into
+ * native ACL format and call zfs_zaccess()
+ */
+int
+zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
+{
+ return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
+}
+
+/*
+ * Access function for secpolicy_vnode_setattr
+ */
+int
+zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
+{
+ int v4_mode = zfs_unix_to_v4(mode >> 6);
+
+ return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
+}
+
+static int
+zfs_delete_final_check(znode_t *zp, znode_t *dzp,
+ mode_t missing_perms, cred_t *cr)
+{
+ int error;
+ uid_t downer;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ downer = zfs_fuid_map_id(zfsvfs, dzp->z_phys->zp_uid, cr, ZFS_OWNER);
+
+ error = secpolicy_vnode_access(cr, ZTOV(dzp), downer, missing_perms);
+
+ if (error == 0)
+ error = zfs_sticky_remove_access(dzp, zp, cr);
+
+ return (error);
+}
+
+/*
+ * Determine whether Access should be granted/deny, without
+ * consulting least priv subsystem.
+ *
+ *
+ * The following chart is the recommended NFSv4 enforcement for
+ * ability to delete an object.
+ *
+ * -------------------------------------------------------
+ * | Parent Dir | Target Object Permissions |
+ * | permissions | |
+ * -------------------------------------------------------
+ * | | ACL Allows | ACL Denies| Delete |
+ * | | Delete | Delete | unspecified|
+ * -------------------------------------------------------
+ * | ACL Allows | Permit | Permit | Permit |
+ * | DELETE_CHILD | |
+ * -------------------------------------------------------
+ * | ACL Denies | Permit | Deny | Deny |
+ * | DELETE_CHILD | | | |
+ * -------------------------------------------------------
+ * | ACL specifies | | | |
+ * | only allow | Permit | Permit | Permit |
+ * | write and | | | |
+ * | execute | | | |
+ * -------------------------------------------------------
+ * | ACL denies | | | |
+ * | write and | Permit | Deny | Deny |
+ * | execute | | | |
+ * -------------------------------------------------------
+ * ^
+ * |
+ * No search privilege, can't even look up file?
+ *
+ */
+int
+zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
+{
+ uint32_t dzp_working_mode = 0;
+ uint32_t zp_working_mode = 0;
+ int dzp_error, zp_error;
+ mode_t missing_perms;
+ boolean_t dzpcheck_privs = B_TRUE;
+ boolean_t zpcheck_privs = B_TRUE;
+
+ /*
+ * We want specific DELETE permissions to
+ * take precedence over WRITE/EXECUTE. We don't
+ * want an ACL such as this to mess us up.
+ * user:joe:write_data:deny,user:joe:delete:allow
+ *
+ * However, deny permissions may ultimately be overridden
+ * by secpolicy_vnode_access().
+ *
+ * We will ask for all of the necessary permissions and then
+ * look at the working modes from the directory and target object
+ * to determine what was found.
+ */
+
+ if (zp->z_phys->zp_flags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
+ return (EPERM);
+
+ /*
+ * First row
+ * If the directory permissions allow the delete, we are done.
+ */
+ if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
+ &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
+ return (0);
+
+ /*
+ * If target object has delete permission then we are done
+ */
+ if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
+ &zpcheck_privs, B_FALSE, cr)) == 0)
+ return (0);
+
+ ASSERT(dzp_error && zp_error);
+
+ if (!dzpcheck_privs)
+ return (dzp_error);
+ if (!zpcheck_privs)
+ return (zp_error);
+
+ /*
+ * Second row
+ *
+ * If directory returns EACCES then delete_child was denied
+ * due to deny delete_child. In this case send the request through
+ * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
+ * since that *could* allow the delete based on write/execute permission
+ * and we want delete permissions to override write/execute.
+ */
+
+ if (dzp_error == EACCES)
+ return (secpolicy_vnode_remove(cr));
+
+ /*
+ * Third Row
+ * only need to see if we have write/execute on directory.
+ */
+
+ if ((dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
+ &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
+ return (zfs_sticky_remove_access(dzp, zp, cr));
+
+ if (!dzpcheck_privs)
+ return (dzp_error);
+
+ /*
+ * Fourth row
+ */
+
+ missing_perms = (dzp_working_mode & ACE_WRITE_DATA) ? VWRITE : 0;
+ missing_perms |= (dzp_working_mode & ACE_EXECUTE) ? VEXEC : 0;
+
+ ASSERT(missing_perms);
+
+ return (zfs_delete_final_check(zp, dzp, missing_perms, cr));
+
+}
+
+int
+zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
+ znode_t *tzp, cred_t *cr)
+{
+ int add_perm;
+ int error;
+
+ if (szp->z_phys->zp_flags & ZFS_AV_QUARANTINED)
+ return (EACCES);
+
+ add_perm = (ZTOV(szp)->v_type == VDIR) ?
+ ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
+
+ /*
+ * Rename permissions are combination of delete permission +
+ * add file/subdir permission.
+ */
+
+ /*
+ * first make sure we do the delete portion.
+ *
+ * If that succeeds then check for add_file/add_subdir permissions
+ */
+
+ if (error = zfs_zaccess_delete(sdzp, szp, cr))
+ return (error);
+
+ /*
+ * If we have a tzp, see if we can delete it?
+ */
+ if (tzp) {
+ if (error = zfs_zaccess_delete(tdzp, tzp, cr))
+ return (error);
+ }
+
+ /*
+ * Now check for add permissions
+ */
+ error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
+
+ return (error);
+}
diff --git a/module/zfs/zfs_byteswap.c b/module/zfs/zfs_byteswap.c
new file mode 100644
index 000000000..ab97f83eb
--- /dev/null
+++ b/module/zfs/zfs_byteswap.c
@@ -0,0 +1,175 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/vfs.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_acl.h>
+
+void
+zfs_oldace_byteswap(ace_t *ace, int ace_cnt)
+{
+ int i;
+
+ for (i = 0; i != ace_cnt; i++, ace++) {
+ ace->a_who = BSWAP_32(ace->a_who);
+ ace->a_access_mask = BSWAP_32(ace->a_access_mask);
+ ace->a_flags = BSWAP_16(ace->a_flags);
+ ace->a_type = BSWAP_16(ace->a_type);
+ }
+}
+
+/*
+ * swap ace_t and ace_oject_t
+ */
+void
+zfs_ace_byteswap(void *buf, size_t size, boolean_t zfs_layout)
+{
+ caddr_t end;
+ caddr_t ptr;
+ zfs_ace_t *zacep;
+ ace_t *acep;
+ uint16_t entry_type;
+ size_t entry_size;
+ int ace_type;
+
+ end = (caddr_t)buf + size;
+ ptr = buf;
+
+ while (ptr < end) {
+ if (zfs_layout) {
+ zacep = (zfs_ace_t *)ptr;
+ zacep->z_hdr.z_access_mask =
+ BSWAP_32(zacep->z_hdr.z_access_mask);
+ zacep->z_hdr.z_flags = BSWAP_16(zacep->z_hdr.z_flags);
+ ace_type = zacep->z_hdr.z_type =
+ BSWAP_16(zacep->z_hdr.z_type);
+ entry_type = zacep->z_hdr.z_flags & ACE_TYPE_FLAGS;
+ } else {
+ acep = (ace_t *)ptr;
+ acep->a_access_mask = BSWAP_32(acep->a_access_mask);
+ acep->a_flags = BSWAP_16(acep->a_flags);
+ ace_type = acep->a_type = BSWAP_16(acep->a_type);
+ acep->a_who = BSWAP_32(acep->a_who);
+ entry_type = acep->a_flags & ACE_TYPE_FLAGS;
+ }
+ switch (entry_type) {
+ case ACE_OWNER:
+ case ACE_EVERYONE:
+ case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
+ entry_size = zfs_layout ?
+ sizeof (zfs_ace_hdr_t) : sizeof (ace_t);
+ break;
+ case ACE_IDENTIFIER_GROUP:
+ default:
+ if (zfs_layout) {
+ zacep->z_fuid = BSWAP_64(zacep->z_fuid);
+ }
+ switch (ace_type) {
+ case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
+ case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
+ case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
+ entry_size = zfs_layout ?
+ sizeof (zfs_object_ace_t) :
+ sizeof (ace_object_t);
+ break;
+ default:
+ entry_size = zfs_layout ? sizeof (zfs_ace_t) :
+ sizeof (ace_t);
+ break;
+ }
+ }
+ ptr = ptr + entry_size;
+ }
+}
+
+/* ARGSUSED */
+void
+zfs_oldacl_byteswap(void *buf, size_t size)
+{
+ int cnt;
+
+ /*
+ * Arggh, since we don't know how many ACEs are in
+ * the array, we have to swap the entire block
+ */
+
+ cnt = size / sizeof (ace_t);
+
+ zfs_oldace_byteswap((ace_t *)buf, cnt);
+}
+
+/* ARGSUSED */
+void
+zfs_acl_byteswap(void *buf, size_t size)
+{
+ zfs_ace_byteswap(buf, size, B_TRUE);
+}
+
+void
+zfs_znode_byteswap(void *buf, size_t size)
+{
+ znode_phys_t *zp = buf;
+
+ ASSERT(size >= sizeof (znode_phys_t));
+
+ zp->zp_crtime[0] = BSWAP_64(zp->zp_crtime[0]);
+ zp->zp_crtime[1] = BSWAP_64(zp->zp_crtime[1]);
+ zp->zp_atime[0] = BSWAP_64(zp->zp_atime[0]);
+ zp->zp_atime[1] = BSWAP_64(zp->zp_atime[1]);
+ zp->zp_mtime[0] = BSWAP_64(zp->zp_mtime[0]);
+ zp->zp_mtime[1] = BSWAP_64(zp->zp_mtime[1]);
+ zp->zp_ctime[0] = BSWAP_64(zp->zp_ctime[0]);
+ zp->zp_ctime[1] = BSWAP_64(zp->zp_ctime[1]);
+ zp->zp_gen = BSWAP_64(zp->zp_gen);
+ zp->zp_mode = BSWAP_64(zp->zp_mode);
+ zp->zp_size = BSWAP_64(zp->zp_size);
+ zp->zp_parent = BSWAP_64(zp->zp_parent);
+ zp->zp_links = BSWAP_64(zp->zp_links);
+ zp->zp_xattr = BSWAP_64(zp->zp_xattr);
+ zp->zp_rdev = BSWAP_64(zp->zp_rdev);
+ zp->zp_flags = BSWAP_64(zp->zp_flags);
+ zp->zp_uid = BSWAP_64(zp->zp_uid);
+ zp->zp_gid = BSWAP_64(zp->zp_gid);
+ zp->zp_zap = BSWAP_64(zp->zp_zap);
+ zp->zp_pad[0] = BSWAP_64(zp->zp_pad[0]);
+ zp->zp_pad[1] = BSWAP_64(zp->zp_pad[1]);
+ zp->zp_pad[2] = BSWAP_64(zp->zp_pad[2]);
+
+ zp->zp_acl.z_acl_extern_obj = BSWAP_64(zp->zp_acl.z_acl_extern_obj);
+ zp->zp_acl.z_acl_size = BSWAP_32(zp->zp_acl.z_acl_size);
+ zp->zp_acl.z_acl_version = BSWAP_16(zp->zp_acl.z_acl_version);
+ zp->zp_acl.z_acl_count = BSWAP_16(zp->zp_acl.z_acl_count);
+ if (zp->zp_acl.z_acl_version == ZFS_ACL_VERSION) {
+ zfs_acl_byteswap((void *)&zp->zp_acl.z_ace_data[0],
+ ZFS_ACE_SPACE);
+ } else
+ zfs_oldace_byteswap((ace_t *)&zp->zp_acl.z_ace_data[0],
+ ACE_SLOT_CNT);
+}
diff --git a/module/zfs/zfs_ctldir.c b/module/zfs/zfs_ctldir.c
new file mode 100644
index 000000000..208fc3629
--- /dev/null
+++ b/module/zfs/zfs_ctldir.c
@@ -0,0 +1,1159 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * ZFS control directory (a.k.a. ".zfs")
+ *
+ * This directory provides a common location for all ZFS meta-objects.
+ * Currently, this is only the 'snapshot' directory, but this may expand in the
+ * future. The elements are built using the GFS primitives, as the hierarchy
+ * does not actually exist on disk.
+ *
+ * For 'snapshot', we don't want to have all snapshots always mounted, because
+ * this would take up a huge amount of space in /etc/mnttab. We have three
+ * types of objects:
+ *
+ * ctldir ------> snapshotdir -------> snapshot
+ * |
+ * |
+ * V
+ * mounted fs
+ *
+ * The 'snapshot' node contains just enough information to lookup '..' and act
+ * as a mountpoint for the snapshot. Whenever we lookup a specific snapshot, we
+ * perform an automount of the underlying filesystem and return the
+ * corresponding vnode.
+ *
+ * All mounts are handled automatically by the kernel, but unmounts are
+ * (currently) handled from user land. The main reason is that there is no
+ * reliable way to auto-unmount the filesystem when it's "no longer in use".
+ * When the user unmounts a filesystem, we call zfsctl_unmount(), which
+ * unmounts any snapshots within the snapshot directory.
+ *
+ * The '.zfs', '.zfs/snapshot', and all directories created under
+ * '.zfs/snapshot' (ie: '.zfs/snapshot/<snapname>') are all GFS nodes and
+ * share the same vfs_t as the head filesystem (what '.zfs' lives under).
+ *
+ * File systems mounted ontop of the GFS nodes '.zfs/snapshot/<snapname>'
+ * (ie: snapshots) are ZFS nodes and have their own unique vfs_t.
+ * However, vnodes within these mounted on file systems have their v_vfsp
+ * fields set to the head filesystem to make NFS happy (see
+ * zfsctl_snapdir_lookup()). We VFS_HOLD the head filesystem's vfs_t
+ * so that it cannot be freed until all snapshots have been unmounted.
+ */
+
+#include <fs/fs_subr.h>
+#include <sys/zfs_ctldir.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/vfs_opreg.h>
+#include <sys/gfs.h>
+#include <sys/stat.h>
+#include <sys/dmu.h>
+#include <sys/dsl_deleg.h>
+#include <sys/mount.h>
+#include <sys/sunddi.h>
+
+#include "zfs_namecheck.h"
+
+typedef struct zfsctl_node {
+ gfs_dir_t zc_gfs_private;
+ uint64_t zc_id;
+ timestruc_t zc_cmtime; /* ctime and mtime, always the same */
+} zfsctl_node_t;
+
+typedef struct zfsctl_snapdir {
+ zfsctl_node_t sd_node;
+ kmutex_t sd_lock;
+ avl_tree_t sd_snaps;
+} zfsctl_snapdir_t;
+
+typedef struct {
+ char *se_name;
+ vnode_t *se_root;
+ avl_node_t se_node;
+} zfs_snapentry_t;
+
+static int
+snapentry_compare(const void *a, const void *b)
+{
+ const zfs_snapentry_t *sa = a;
+ const zfs_snapentry_t *sb = b;
+ int ret = strcmp(sa->se_name, sb->se_name);
+
+ if (ret < 0)
+ return (-1);
+ else if (ret > 0)
+ return (1);
+ else
+ return (0);
+}
+
+vnodeops_t *zfsctl_ops_root;
+vnodeops_t *zfsctl_ops_snapdir;
+vnodeops_t *zfsctl_ops_snapshot;
+
+static const fs_operation_def_t zfsctl_tops_root[];
+static const fs_operation_def_t zfsctl_tops_snapdir[];
+static const fs_operation_def_t zfsctl_tops_snapshot[];
+
+static vnode_t *zfsctl_mknode_snapdir(vnode_t *);
+static vnode_t *zfsctl_snapshot_mknode(vnode_t *, uint64_t objset);
+static int zfsctl_unmount_snap(zfs_snapentry_t *, int, cred_t *);
+
+static gfs_opsvec_t zfsctl_opsvec[] = {
+ { ".zfs", zfsctl_tops_root, &zfsctl_ops_root },
+ { ".zfs/snapshot", zfsctl_tops_snapdir, &zfsctl_ops_snapdir },
+ { ".zfs/snapshot/vnode", zfsctl_tops_snapshot, &zfsctl_ops_snapshot },
+ { NULL }
+};
+
+/*
+ * Root directory elements. We have only a single static entry, 'snapshot'.
+ */
+static gfs_dirent_t zfsctl_root_entries[] = {
+ { "snapshot", zfsctl_mknode_snapdir, GFS_CACHE_VNODE },
+ { NULL }
+};
+
+/* include . and .. in the calculation */
+#define NROOT_ENTRIES ((sizeof (zfsctl_root_entries) / \
+ sizeof (gfs_dirent_t)) + 1)
+
+
+/*
+ * Initialize the various GFS pieces we'll need to create and manipulate .zfs
+ * directories. This is called from the ZFS init routine, and initializes the
+ * vnode ops vectors that we'll be using.
+ */
+void
+zfsctl_init(void)
+{
+ VERIFY(gfs_make_opsvec(zfsctl_opsvec) == 0);
+}
+
+void
+zfsctl_fini(void)
+{
+ /*
+ * Remove vfsctl vnode ops
+ */
+ if (zfsctl_ops_root)
+ vn_freevnodeops(zfsctl_ops_root);
+ if (zfsctl_ops_snapdir)
+ vn_freevnodeops(zfsctl_ops_snapdir);
+ if (zfsctl_ops_snapshot)
+ vn_freevnodeops(zfsctl_ops_snapshot);
+
+ zfsctl_ops_root = NULL;
+ zfsctl_ops_snapdir = NULL;
+ zfsctl_ops_snapshot = NULL;
+}
+
+/*
+ * Return the inode number associated with the 'snapshot' directory.
+ */
+/* ARGSUSED */
+static ino64_t
+zfsctl_root_inode_cb(vnode_t *vp, int index)
+{
+ ASSERT(index == 0);
+ return (ZFSCTL_INO_SNAPDIR);
+}
+
+/*
+ * Create the '.zfs' directory. This directory is cached as part of the VFS
+ * structure. This results in a hold on the vfs_t. The code in zfs_umount()
+ * therefore checks against a vfs_count of 2 instead of 1. This reference
+ * is removed when the ctldir is destroyed in the unmount.
+ */
+void
+zfsctl_create(zfsvfs_t *zfsvfs)
+{
+ vnode_t *vp, *rvp;
+ zfsctl_node_t *zcp;
+
+ ASSERT(zfsvfs->z_ctldir == NULL);
+
+ vp = gfs_root_create(sizeof (zfsctl_node_t), zfsvfs->z_vfs,
+ zfsctl_ops_root, ZFSCTL_INO_ROOT, zfsctl_root_entries,
+ zfsctl_root_inode_cb, MAXNAMELEN, NULL, NULL);
+ zcp = vp->v_data;
+ zcp->zc_id = ZFSCTL_INO_ROOT;
+
+ VERIFY(VFS_ROOT(zfsvfs->z_vfs, &rvp) == 0);
+ ZFS_TIME_DECODE(&zcp->zc_cmtime, VTOZ(rvp)->z_phys->zp_crtime);
+ VN_RELE(rvp);
+
+ /*
+ * We're only faking the fact that we have a root of a filesystem for
+ * the sake of the GFS interfaces. Undo the flag manipulation it did
+ * for us.
+ */
+ vp->v_flag &= ~(VROOT | VNOCACHE | VNOMAP | VNOSWAP | VNOMOUNT);
+
+ zfsvfs->z_ctldir = vp;
+}
+
+/*
+ * Destroy the '.zfs' directory. Only called when the filesystem is unmounted.
+ * There might still be more references if we were force unmounted, but only
+ * new zfs_inactive() calls can occur and they don't reference .zfs
+ */
+void
+zfsctl_destroy(zfsvfs_t *zfsvfs)
+{
+ VN_RELE(zfsvfs->z_ctldir);
+ zfsvfs->z_ctldir = NULL;
+}
+
+/*
+ * Given a root znode, retrieve the associated .zfs directory.
+ * Add a hold to the vnode and return it.
+ */
+vnode_t *
+zfsctl_root(znode_t *zp)
+{
+ ASSERT(zfs_has_ctldir(zp));
+ VN_HOLD(zp->z_zfsvfs->z_ctldir);
+ return (zp->z_zfsvfs->z_ctldir);
+}
+
+/*
+ * Common open routine. Disallow any write access.
+ */
+/* ARGSUSED */
+static int
+zfsctl_common_open(vnode_t **vpp, int flags, cred_t *cr, caller_context_t *ct)
+{
+ if (flags & FWRITE)
+ return (EACCES);
+
+ return (0);
+}
+
+/*
+ * Common close routine. Nothing to do here.
+ */
+/* ARGSUSED */
+static int
+zfsctl_common_close(vnode_t *vpp, int flags, int count, offset_t off,
+ cred_t *cr, caller_context_t *ct)
+{
+ return (0);
+}
+
+/*
+ * Common access routine. Disallow writes.
+ */
+/* ARGSUSED */
+static int
+zfsctl_common_access(vnode_t *vp, int mode, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ if (mode & VWRITE)
+ return (EACCES);
+
+ return (0);
+}
+
+/*
+ * Common getattr function. Fill in basic information.
+ */
+static void
+zfsctl_common_getattr(vnode_t *vp, vattr_t *vap)
+{
+ zfsctl_node_t *zcp = vp->v_data;
+ timestruc_t now;
+
+ vap->va_uid = 0;
+ vap->va_gid = 0;
+ vap->va_rdev = 0;
+ /*
+ * We are a purly virtual object, so we have no
+ * blocksize or allocated blocks.
+ */
+ vap->va_blksize = 0;
+ vap->va_nblocks = 0;
+ vap->va_seq = 0;
+ vap->va_fsid = vp->v_vfsp->vfs_dev;
+ vap->va_mode = S_IRUSR | S_IXUSR | S_IRGRP | S_IXGRP |
+ S_IROTH | S_IXOTH;
+ vap->va_type = VDIR;
+ /*
+ * We live in the now (for atime).
+ */
+ gethrestime(&now);
+ vap->va_atime = now;
+ vap->va_mtime = vap->va_ctime = zcp->zc_cmtime;
+}
+
+/*ARGSUSED*/
+static int
+zfsctl_common_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
+{
+ zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+ zfsctl_node_t *zcp = vp->v_data;
+ uint64_t object = zcp->zc_id;
+ zfid_short_t *zfid;
+ int i;
+
+ ZFS_ENTER(zfsvfs);
+
+ if (fidp->fid_len < SHORT_FID_LEN) {
+ fidp->fid_len = SHORT_FID_LEN;
+ ZFS_EXIT(zfsvfs);
+ return (ENOSPC);
+ }
+
+ zfid = (zfid_short_t *)fidp;
+
+ zfid->zf_len = SHORT_FID_LEN;
+
+ for (i = 0; i < sizeof (zfid->zf_object); i++)
+ zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
+
+ /* .zfs znodes always have a generation number of 0 */
+ for (i = 0; i < sizeof (zfid->zf_gen); i++)
+ zfid->zf_gen[i] = 0;
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/*
+ * .zfs inode namespace
+ *
+ * We need to generate unique inode numbers for all files and directories
+ * within the .zfs pseudo-filesystem. We use the following scheme:
+ *
+ * ENTRY ZFSCTL_INODE
+ * .zfs 1
+ * .zfs/snapshot 2
+ * .zfs/snapshot/<snap> objectid(snap)
+ */
+
+#define ZFSCTL_INO_SNAP(id) (id)
+
+/*
+ * Get root directory attributes.
+ */
+/* ARGSUSED */
+static int
+zfsctl_root_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+
+ ZFS_ENTER(zfsvfs);
+ vap->va_nodeid = ZFSCTL_INO_ROOT;
+ vap->va_nlink = vap->va_size = NROOT_ENTRIES;
+
+ zfsctl_common_getattr(vp, vap);
+ ZFS_EXIT(zfsvfs);
+
+ return (0);
+}
+
+/*
+ * Special case the handling of "..".
+ */
+/* ARGSUSED */
+int
+zfsctl_root_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
+ int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
+ int *direntflags, pathname_t *realpnp)
+{
+ zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
+ int err;
+
+ /*
+ * No extended attributes allowed under .zfs
+ */
+ if (flags & LOOKUP_XATTR)
+ return (EINVAL);
+
+ ZFS_ENTER(zfsvfs);
+
+ if (strcmp(nm, "..") == 0) {
+ err = VFS_ROOT(dvp->v_vfsp, vpp);
+ } else {
+ err = gfs_vop_lookup(dvp, nm, vpp, pnp, flags, rdir,
+ cr, ct, direntflags, realpnp);
+ }
+
+ ZFS_EXIT(zfsvfs);
+
+ return (err);
+}
+
+static const fs_operation_def_t zfsctl_tops_root[] = {
+ { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
+ { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
+ { VOPNAME_IOCTL, { .error = fs_inval } },
+ { VOPNAME_GETATTR, { .vop_getattr = zfsctl_root_getattr } },
+ { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
+ { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
+ { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_root_lookup } },
+ { VOPNAME_SEEK, { .vop_seek = fs_seek } },
+ { VOPNAME_INACTIVE, { .vop_inactive = gfs_vop_inactive } },
+ { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
+ { NULL }
+};
+
+static int
+zfsctl_snapshot_zname(vnode_t *vp, const char *name, int len, char *zname)
+{
+ objset_t *os = ((zfsvfs_t *)((vp)->v_vfsp->vfs_data))->z_os;
+
+ if (snapshot_namecheck(name, NULL, NULL) != 0)
+ return (EILSEQ);
+ dmu_objset_name(os, zname);
+ if (strlen(zname) + 1 + strlen(name) >= len)
+ return (ENAMETOOLONG);
+ (void) strcat(zname, "@");
+ (void) strcat(zname, name);
+ return (0);
+}
+
+static int
+zfsctl_unmount_snap(zfs_snapentry_t *sep, int fflags, cred_t *cr)
+{
+ vnode_t *svp = sep->se_root;
+ int error;
+
+ ASSERT(vn_ismntpt(svp));
+
+ /* this will be dropped by dounmount() */
+ if ((error = vn_vfswlock(svp)) != 0)
+ return (error);
+
+ VN_HOLD(svp);
+ error = dounmount(vn_mountedvfs(svp), fflags, cr);
+ if (error) {
+ VN_RELE(svp);
+ return (error);
+ }
+ VFS_RELE(svp->v_vfsp);
+ /*
+ * We can't use VN_RELE(), as that will try to invoke
+ * zfsctl_snapdir_inactive(), which would cause us to destroy
+ * the sd_lock mutex held by our caller.
+ */
+ ASSERT(svp->v_count == 1);
+ gfs_vop_inactive(svp, cr, NULL);
+
+ kmem_free(sep->se_name, strlen(sep->se_name) + 1);
+ kmem_free(sep, sizeof (zfs_snapentry_t));
+
+ return (0);
+}
+
+static void
+zfsctl_rename_snap(zfsctl_snapdir_t *sdp, zfs_snapentry_t *sep, const char *nm)
+{
+ avl_index_t where;
+ vfs_t *vfsp;
+ refstr_t *pathref;
+ char newpath[MAXNAMELEN];
+ char *tail;
+
+ ASSERT(MUTEX_HELD(&sdp->sd_lock));
+ ASSERT(sep != NULL);
+
+ vfsp = vn_mountedvfs(sep->se_root);
+ ASSERT(vfsp != NULL);
+
+ vfs_lock_wait(vfsp);
+
+ /*
+ * Change the name in the AVL tree.
+ */
+ avl_remove(&sdp->sd_snaps, sep);
+ kmem_free(sep->se_name, strlen(sep->se_name) + 1);
+ sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
+ (void) strcpy(sep->se_name, nm);
+ VERIFY(avl_find(&sdp->sd_snaps, sep, &where) == NULL);
+ avl_insert(&sdp->sd_snaps, sep, where);
+
+ /*
+ * Change the current mountpoint info:
+ * - update the tail of the mntpoint path
+ * - update the tail of the resource path
+ */
+ pathref = vfs_getmntpoint(vfsp);
+ (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
+ VERIFY((tail = strrchr(newpath, '/')) != NULL);
+ *(tail+1) = '\0';
+ ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
+ (void) strcat(newpath, nm);
+ refstr_rele(pathref);
+ vfs_setmntpoint(vfsp, newpath);
+
+ pathref = vfs_getresource(vfsp);
+ (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
+ VERIFY((tail = strrchr(newpath, '@')) != NULL);
+ *(tail+1) = '\0';
+ ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
+ (void) strcat(newpath, nm);
+ refstr_rele(pathref);
+ vfs_setresource(vfsp, newpath);
+
+ vfs_unlock(vfsp);
+}
+
+/*ARGSUSED*/
+static int
+zfsctl_snapdir_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm,
+ cred_t *cr, caller_context_t *ct, int flags)
+{
+ zfsctl_snapdir_t *sdp = sdvp->v_data;
+ zfs_snapentry_t search, *sep;
+ zfsvfs_t *zfsvfs;
+ avl_index_t where;
+ char from[MAXNAMELEN], to[MAXNAMELEN];
+ char real[MAXNAMELEN];
+ int err;
+
+ zfsvfs = sdvp->v_vfsp->vfs_data;
+ ZFS_ENTER(zfsvfs);
+
+ if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
+ err = dmu_snapshot_realname(zfsvfs->z_os, snm, real,
+ MAXNAMELEN, NULL);
+ if (err == 0) {
+ snm = real;
+ } else if (err != ENOTSUP) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ }
+
+ ZFS_EXIT(zfsvfs);
+
+ err = zfsctl_snapshot_zname(sdvp, snm, MAXNAMELEN, from);
+ if (!err)
+ err = zfsctl_snapshot_zname(tdvp, tnm, MAXNAMELEN, to);
+ if (!err)
+ err = zfs_secpolicy_rename_perms(from, to, cr);
+ if (err)
+ return (err);
+
+ /*
+ * Cannot move snapshots out of the snapdir.
+ */
+ if (sdvp != tdvp)
+ return (EINVAL);
+
+ if (strcmp(snm, tnm) == 0)
+ return (0);
+
+ mutex_enter(&sdp->sd_lock);
+
+ search.se_name = (char *)snm;
+ if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) == NULL) {
+ mutex_exit(&sdp->sd_lock);
+ return (ENOENT);
+ }
+
+ err = dmu_objset_rename(from, to, B_FALSE);
+ if (err == 0)
+ zfsctl_rename_snap(sdp, sep, tnm);
+
+ mutex_exit(&sdp->sd_lock);
+
+ return (err);
+}
+
+/* ARGSUSED */
+static int
+zfsctl_snapdir_remove(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
+ caller_context_t *ct, int flags)
+{
+ zfsctl_snapdir_t *sdp = dvp->v_data;
+ zfs_snapentry_t *sep;
+ zfs_snapentry_t search;
+ zfsvfs_t *zfsvfs;
+ char snapname[MAXNAMELEN];
+ char real[MAXNAMELEN];
+ int err;
+
+ zfsvfs = dvp->v_vfsp->vfs_data;
+ ZFS_ENTER(zfsvfs);
+
+ if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
+
+ err = dmu_snapshot_realname(zfsvfs->z_os, name, real,
+ MAXNAMELEN, NULL);
+ if (err == 0) {
+ name = real;
+ } else if (err != ENOTSUP) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ }
+
+ ZFS_EXIT(zfsvfs);
+
+ err = zfsctl_snapshot_zname(dvp, name, MAXNAMELEN, snapname);
+ if (!err)
+ err = zfs_secpolicy_destroy_perms(snapname, cr);
+ if (err)
+ return (err);
+
+ mutex_enter(&sdp->sd_lock);
+
+ search.se_name = name;
+ sep = avl_find(&sdp->sd_snaps, &search, NULL);
+ if (sep) {
+ avl_remove(&sdp->sd_snaps, sep);
+ err = zfsctl_unmount_snap(sep, MS_FORCE, cr);
+ if (err)
+ avl_add(&sdp->sd_snaps, sep);
+ else
+ err = dmu_objset_destroy(snapname);
+ } else {
+ err = ENOENT;
+ }
+
+ mutex_exit(&sdp->sd_lock);
+
+ return (err);
+}
+
+/*
+ * This creates a snapshot under '.zfs/snapshot'.
+ */
+/* ARGSUSED */
+static int
+zfsctl_snapdir_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp,
+ cred_t *cr, caller_context_t *cc, int flags, vsecattr_t *vsecp)
+{
+ zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
+ char name[MAXNAMELEN];
+ int err;
+ static enum symfollow follow = NO_FOLLOW;
+ static enum uio_seg seg = UIO_SYSSPACE;
+
+ if (snapshot_namecheck(dirname, NULL, NULL) != 0)
+ return (EILSEQ);
+
+ dmu_objset_name(zfsvfs->z_os, name);
+
+ *vpp = NULL;
+
+ err = zfs_secpolicy_snapshot_perms(name, cr);
+ if (err)
+ return (err);
+
+ if (err == 0) {
+ err = dmu_objset_snapshot(name, dirname, B_FALSE);
+ if (err)
+ return (err);
+ err = lookupnameat(dirname, seg, follow, NULL, vpp, dvp);
+ }
+
+ return (err);
+}
+
+/*
+ * Lookup entry point for the 'snapshot' directory. Try to open the
+ * snapshot if it exist, creating the pseudo filesystem vnode as necessary.
+ * Perform a mount of the associated dataset on top of the vnode.
+ */
+/* ARGSUSED */
+static int
+zfsctl_snapdir_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
+ int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
+ int *direntflags, pathname_t *realpnp)
+{
+ zfsctl_snapdir_t *sdp = dvp->v_data;
+ objset_t *snap;
+ char snapname[MAXNAMELEN];
+ char real[MAXNAMELEN];
+ char *mountpoint;
+ zfs_snapentry_t *sep, search;
+ struct mounta margs;
+ vfs_t *vfsp;
+ size_t mountpoint_len;
+ avl_index_t where;
+ zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
+ int err;
+
+ /*
+ * No extended attributes allowed under .zfs
+ */
+ if (flags & LOOKUP_XATTR)
+ return (EINVAL);
+
+ ASSERT(dvp->v_type == VDIR);
+
+ if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0)
+ return (0);
+
+ /*
+ * If we get a recursive call, that means we got called
+ * from the domount() code while it was trying to look up the
+ * spec (which looks like a local path for zfs). We need to
+ * add some flag to domount() to tell it not to do this lookup.
+ */
+ if (MUTEX_HELD(&sdp->sd_lock))
+ return (ENOENT);
+
+ ZFS_ENTER(zfsvfs);
+
+ if (flags & FIGNORECASE) {
+ boolean_t conflict = B_FALSE;
+
+ err = dmu_snapshot_realname(zfsvfs->z_os, nm, real,
+ MAXNAMELEN, &conflict);
+ if (err == 0) {
+ nm = real;
+ } else if (err != ENOTSUP) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ if (realpnp)
+ (void) strlcpy(realpnp->pn_buf, nm,
+ realpnp->pn_bufsize);
+ if (conflict && direntflags)
+ *direntflags = ED_CASE_CONFLICT;
+ }
+
+ mutex_enter(&sdp->sd_lock);
+ search.se_name = (char *)nm;
+ if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) != NULL) {
+ *vpp = sep->se_root;
+ VN_HOLD(*vpp);
+ err = traverse(vpp);
+ if (err) {
+ VN_RELE(*vpp);
+ *vpp = NULL;
+ } else if (*vpp == sep->se_root) {
+ /*
+ * The snapshot was unmounted behind our backs,
+ * try to remount it.
+ */
+ goto domount;
+ } else {
+ /*
+ * VROOT was set during the traverse call. We need
+ * to clear it since we're pretending to be part
+ * of our parent's vfs.
+ */
+ (*vpp)->v_flag &= ~VROOT;
+ }
+ mutex_exit(&sdp->sd_lock);
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+
+ /*
+ * The requested snapshot is not currently mounted, look it up.
+ */
+ err = zfsctl_snapshot_zname(dvp, nm, MAXNAMELEN, snapname);
+ if (err) {
+ mutex_exit(&sdp->sd_lock);
+ ZFS_EXIT(zfsvfs);
+ /*
+ * handle "ls *" or "?" in a graceful manner,
+ * forcing EILSEQ to ENOENT.
+ * Since shell ultimately passes "*" or "?" as name to lookup
+ */
+ return (err == EILSEQ ? ENOENT : err);
+ }
+ if (dmu_objset_open(snapname, DMU_OST_ZFS,
+ DS_MODE_USER | DS_MODE_READONLY, &snap) != 0) {
+ mutex_exit(&sdp->sd_lock);
+ ZFS_EXIT(zfsvfs);
+ return (ENOENT);
+ }
+
+ sep = kmem_alloc(sizeof (zfs_snapentry_t), KM_SLEEP);
+ sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
+ (void) strcpy(sep->se_name, nm);
+ *vpp = sep->se_root = zfsctl_snapshot_mknode(dvp, dmu_objset_id(snap));
+ avl_insert(&sdp->sd_snaps, sep, where);
+
+ dmu_objset_close(snap);
+domount:
+ mountpoint_len = strlen(refstr_value(dvp->v_vfsp->vfs_mntpt)) +
+ strlen("/.zfs/snapshot/") + strlen(nm) + 1;
+ mountpoint = kmem_alloc(mountpoint_len, KM_SLEEP);
+ (void) snprintf(mountpoint, mountpoint_len, "%s/.zfs/snapshot/%s",
+ refstr_value(dvp->v_vfsp->vfs_mntpt), nm);
+
+ margs.spec = snapname;
+ margs.dir = mountpoint;
+ margs.flags = MS_SYSSPACE | MS_NOMNTTAB;
+ margs.fstype = "zfs";
+ margs.dataptr = NULL;
+ margs.datalen = 0;
+ margs.optptr = NULL;
+ margs.optlen = 0;
+
+ err = domount("zfs", &margs, *vpp, kcred, &vfsp);
+ kmem_free(mountpoint, mountpoint_len);
+
+ if (err == 0) {
+ /*
+ * Return the mounted root rather than the covered mount point.
+ * Takes the GFS vnode at .zfs/snapshot/<snapname> and returns
+ * the ZFS vnode mounted on top of the GFS node. This ZFS
+ * vnode is the root the newly created vfsp.
+ */
+ VFS_RELE(vfsp);
+ err = traverse(vpp);
+ }
+
+ if (err == 0) {
+ /*
+ * Fix up the root vnode mounted on .zfs/snapshot/<snapname>.
+ *
+ * This is where we lie about our v_vfsp in order to
+ * make .zfs/snapshot/<snapname> accessible over NFS
+ * without requiring manual mounts of <snapname>.
+ */
+ ASSERT(VTOZ(*vpp)->z_zfsvfs != zfsvfs);
+ VTOZ(*vpp)->z_zfsvfs->z_parent = zfsvfs;
+ (*vpp)->v_vfsp = zfsvfs->z_vfs;
+ (*vpp)->v_flag &= ~VROOT;
+ }
+ mutex_exit(&sdp->sd_lock);
+ ZFS_EXIT(zfsvfs);
+
+ /*
+ * If we had an error, drop our hold on the vnode and
+ * zfsctl_snapshot_inactive() will clean up.
+ */
+ if (err) {
+ VN_RELE(*vpp);
+ *vpp = NULL;
+ }
+ return (err);
+}
+
+/* ARGSUSED */
+static int
+zfsctl_snapdir_readdir_cb(vnode_t *vp, void *dp, int *eofp,
+ offset_t *offp, offset_t *nextp, void *data, int flags)
+{
+ zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+ char snapname[MAXNAMELEN];
+ uint64_t id, cookie;
+ boolean_t case_conflict;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+
+ cookie = *offp;
+ error = dmu_snapshot_list_next(zfsvfs->z_os, MAXNAMELEN, snapname, &id,
+ &cookie, &case_conflict);
+ if (error) {
+ ZFS_EXIT(zfsvfs);
+ if (error == ENOENT) {
+ *eofp = 1;
+ return (0);
+ }
+ return (error);
+ }
+
+ if (flags & V_RDDIR_ENTFLAGS) {
+ edirent_t *eodp = dp;
+
+ (void) strcpy(eodp->ed_name, snapname);
+ eodp->ed_ino = ZFSCTL_INO_SNAP(id);
+ eodp->ed_eflags = case_conflict ? ED_CASE_CONFLICT : 0;
+ } else {
+ struct dirent64 *odp = dp;
+
+ (void) strcpy(odp->d_name, snapname);
+ odp->d_ino = ZFSCTL_INO_SNAP(id);
+ }
+ *nextp = cookie;
+
+ ZFS_EXIT(zfsvfs);
+
+ return (0);
+}
+
+/*
+ * pvp is the '.zfs' directory (zfsctl_node_t).
+ * Creates vp, which is '.zfs/snapshot' (zfsctl_snapdir_t).
+ *
+ * This function is the callback to create a GFS vnode for '.zfs/snapshot'
+ * when a lookup is performed on .zfs for "snapshot".
+ */
+vnode_t *
+zfsctl_mknode_snapdir(vnode_t *pvp)
+{
+ vnode_t *vp;
+ zfsctl_snapdir_t *sdp;
+
+ vp = gfs_dir_create(sizeof (zfsctl_snapdir_t), pvp,
+ zfsctl_ops_snapdir, NULL, NULL, MAXNAMELEN,
+ zfsctl_snapdir_readdir_cb, NULL);
+ sdp = vp->v_data;
+ sdp->sd_node.zc_id = ZFSCTL_INO_SNAPDIR;
+ sdp->sd_node.zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
+ mutex_init(&sdp->sd_lock, NULL, MUTEX_DEFAULT, NULL);
+ avl_create(&sdp->sd_snaps, snapentry_compare,
+ sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node));
+ return (vp);
+}
+
+/* ARGSUSED */
+static int
+zfsctl_snapdir_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+ zfsctl_snapdir_t *sdp = vp->v_data;
+
+ ZFS_ENTER(zfsvfs);
+ zfsctl_common_getattr(vp, vap);
+ vap->va_nodeid = gfs_file_inode(vp);
+ vap->va_nlink = vap->va_size = avl_numnodes(&sdp->sd_snaps) + 2;
+ ZFS_EXIT(zfsvfs);
+
+ return (0);
+}
+
+/* ARGSUSED */
+static void
+zfsctl_snapdir_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
+{
+ zfsctl_snapdir_t *sdp = vp->v_data;
+ void *private;
+
+ private = gfs_dir_inactive(vp);
+ if (private != NULL) {
+ ASSERT(avl_numnodes(&sdp->sd_snaps) == 0);
+ mutex_destroy(&sdp->sd_lock);
+ avl_destroy(&sdp->sd_snaps);
+ kmem_free(private, sizeof (zfsctl_snapdir_t));
+ }
+}
+
+static const fs_operation_def_t zfsctl_tops_snapdir[] = {
+ { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
+ { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
+ { VOPNAME_IOCTL, { .error = fs_inval } },
+ { VOPNAME_GETATTR, { .vop_getattr = zfsctl_snapdir_getattr } },
+ { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
+ { VOPNAME_RENAME, { .vop_rename = zfsctl_snapdir_rename } },
+ { VOPNAME_RMDIR, { .vop_rmdir = zfsctl_snapdir_remove } },
+ { VOPNAME_MKDIR, { .vop_mkdir = zfsctl_snapdir_mkdir } },
+ { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
+ { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_snapdir_lookup } },
+ { VOPNAME_SEEK, { .vop_seek = fs_seek } },
+ { VOPNAME_INACTIVE, { .vop_inactive = zfsctl_snapdir_inactive } },
+ { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
+ { NULL }
+};
+
+/*
+ * pvp is the GFS vnode '.zfs/snapshot'.
+ *
+ * This creates a GFS node under '.zfs/snapshot' representing each
+ * snapshot. This newly created GFS node is what we mount snapshot
+ * vfs_t's ontop of.
+ */
+static vnode_t *
+zfsctl_snapshot_mknode(vnode_t *pvp, uint64_t objset)
+{
+ vnode_t *vp;
+ zfsctl_node_t *zcp;
+
+ vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp,
+ zfsctl_ops_snapshot, NULL, NULL, MAXNAMELEN, NULL, NULL);
+ zcp = vp->v_data;
+ zcp->zc_id = objset;
+ VFS_HOLD(vp->v_vfsp);
+
+ return (vp);
+}
+
+static void
+zfsctl_snapshot_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
+{
+ zfsctl_snapdir_t *sdp;
+ zfs_snapentry_t *sep, *next;
+ vnode_t *dvp;
+
+ VERIFY(gfs_dir_lookup(vp, "..", &dvp, cr, 0, NULL, NULL) == 0);
+ sdp = dvp->v_data;
+
+ mutex_enter(&sdp->sd_lock);
+
+ if (vp->v_count > 1) {
+ mutex_exit(&sdp->sd_lock);
+ return;
+ }
+ ASSERT(!vn_ismntpt(vp));
+
+ sep = avl_first(&sdp->sd_snaps);
+ while (sep != NULL) {
+ next = AVL_NEXT(&sdp->sd_snaps, sep);
+
+ if (sep->se_root == vp) {
+ avl_remove(&sdp->sd_snaps, sep);
+ kmem_free(sep->se_name, strlen(sep->se_name) + 1);
+ kmem_free(sep, sizeof (zfs_snapentry_t));
+ break;
+ }
+ sep = next;
+ }
+ ASSERT(sep != NULL);
+
+ mutex_exit(&sdp->sd_lock);
+ VN_RELE(dvp);
+ VFS_RELE(vp->v_vfsp);
+
+ /*
+ * Dispose of the vnode for the snapshot mount point.
+ * This is safe to do because once this entry has been removed
+ * from the AVL tree, it can't be found again, so cannot become
+ * "active". If we lookup the same name again we will end up
+ * creating a new vnode.
+ */
+ gfs_vop_inactive(vp, cr, ct);
+}
+
+
+/*
+ * These VP's should never see the light of day. They should always
+ * be covered.
+ */
+static const fs_operation_def_t zfsctl_tops_snapshot[] = {
+ VOPNAME_INACTIVE, { .vop_inactive = zfsctl_snapshot_inactive },
+ NULL, NULL
+};
+
+int
+zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ vnode_t *dvp, *vp;
+ zfsctl_snapdir_t *sdp;
+ zfsctl_node_t *zcp;
+ zfs_snapentry_t *sep;
+ int error;
+
+ ASSERT(zfsvfs->z_ctldir != NULL);
+ error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
+ NULL, 0, NULL, kcred, NULL, NULL, NULL);
+ if (error != 0)
+ return (error);
+ sdp = dvp->v_data;
+
+ mutex_enter(&sdp->sd_lock);
+ sep = avl_first(&sdp->sd_snaps);
+ while (sep != NULL) {
+ vp = sep->se_root;
+ zcp = vp->v_data;
+ if (zcp->zc_id == objsetid)
+ break;
+
+ sep = AVL_NEXT(&sdp->sd_snaps, sep);
+ }
+
+ if (sep != NULL) {
+ VN_HOLD(vp);
+ /*
+ * Return the mounted root rather than the covered mount point.
+ * Takes the GFS vnode at .zfs/snapshot/<snapshot objsetid>
+ * and returns the ZFS vnode mounted on top of the GFS node.
+ * This ZFS vnode is the root of the vfs for objset 'objsetid'.
+ */
+ error = traverse(&vp);
+ if (error == 0) {
+ if (vp == sep->se_root)
+ error = EINVAL;
+ else
+ *zfsvfsp = VTOZ(vp)->z_zfsvfs;
+ }
+ mutex_exit(&sdp->sd_lock);
+ VN_RELE(vp);
+ } else {
+ error = EINVAL;
+ mutex_exit(&sdp->sd_lock);
+ }
+
+ VN_RELE(dvp);
+
+ return (error);
+}
+
+/*
+ * Unmount any snapshots for the given filesystem. This is called from
+ * zfs_umount() - if we have a ctldir, then go through and unmount all the
+ * snapshots.
+ */
+int
+zfsctl_umount_snapshots(vfs_t *vfsp, int fflags, cred_t *cr)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ vnode_t *dvp;
+ zfsctl_snapdir_t *sdp;
+ zfs_snapentry_t *sep, *next;
+ int error;
+
+ ASSERT(zfsvfs->z_ctldir != NULL);
+ error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
+ NULL, 0, NULL, cr, NULL, NULL, NULL);
+ if (error != 0)
+ return (error);
+ sdp = dvp->v_data;
+
+ mutex_enter(&sdp->sd_lock);
+
+ sep = avl_first(&sdp->sd_snaps);
+ while (sep != NULL) {
+ next = AVL_NEXT(&sdp->sd_snaps, sep);
+
+ /*
+ * If this snapshot is not mounted, then it must
+ * have just been unmounted by somebody else, and
+ * will be cleaned up by zfsctl_snapdir_inactive().
+ */
+ if (vn_ismntpt(sep->se_root)) {
+ avl_remove(&sdp->sd_snaps, sep);
+ error = zfsctl_unmount_snap(sep, fflags, cr);
+ if (error) {
+ avl_add(&sdp->sd_snaps, sep);
+ break;
+ }
+ }
+ sep = next;
+ }
+
+ mutex_exit(&sdp->sd_lock);
+ VN_RELE(dvp);
+
+ return (error);
+}
diff --git a/module/zfs/zfs_dir.c b/module/zfs/zfs_dir.c
new file mode 100644
index 000000000..1ec493264
--- /dev/null
+++ b/module/zfs/zfs_dir.c
@@ -0,0 +1,977 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/time.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/resource.h>
+#include <sys/vfs.h>
+#include <sys/vnode.h>
+#include <sys/file.h>
+#include <sys/mode.h>
+#include <sys/kmem.h>
+#include <sys/uio.h>
+#include <sys/pathname.h>
+#include <sys/cmn_err.h>
+#include <sys/errno.h>
+#include <sys/stat.h>
+#include <sys/unistd.h>
+#include <sys/sunddi.h>
+#include <sys/random.h>
+#include <sys/policy.h>
+#include <sys/zfs_dir.h>
+#include <sys/zfs_acl.h>
+#include <sys/fs/zfs.h>
+#include "fs/fs_subr.h"
+#include <sys/zap.h>
+#include <sys/dmu.h>
+#include <sys/atomic.h>
+#include <sys/zfs_ctldir.h>
+#include <sys/zfs_fuid.h>
+#include <sys/dnlc.h>
+#include <sys/extdirent.h>
+
+/*
+ * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups
+ * of names after deciding which is the appropriate lookup interface.
+ */
+static int
+zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, char *name, boolean_t exact,
+ boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid)
+{
+ int error;
+
+ if (zfsvfs->z_norm) {
+ matchtype_t mt = MT_FIRST;
+ boolean_t conflict = B_FALSE;
+ size_t bufsz = 0;
+ char *buf = NULL;
+
+ if (rpnp) {
+ buf = rpnp->pn_buf;
+ bufsz = rpnp->pn_bufsize;
+ }
+ if (exact)
+ mt = MT_EXACT;
+ /*
+ * In the non-mixed case we only expect there would ever
+ * be one match, but we need to use the normalizing lookup.
+ */
+ error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1,
+ zoid, mt, buf, bufsz, &conflict);
+ if (!error && deflags)
+ *deflags = conflict ? ED_CASE_CONFLICT : 0;
+ } else {
+ error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid);
+ }
+ *zoid = ZFS_DIRENT_OBJ(*zoid);
+
+ if (error == ENOENT && update)
+ dnlc_update(ZTOV(dzp), name, DNLC_NO_VNODE);
+
+ return (error);
+}
+
+/*
+ * Lock a directory entry. A dirlock on <dzp, name> protects that name
+ * in dzp's directory zap object. As long as you hold a dirlock, you can
+ * assume two things: (1) dzp cannot be reaped, and (2) no other thread
+ * can change the zap entry for (i.e. link or unlink) this name.
+ *
+ * Input arguments:
+ * dzp - znode for directory
+ * name - name of entry to lock
+ * flag - ZNEW: if the entry already exists, fail with EEXIST.
+ * ZEXISTS: if the entry does not exist, fail with ENOENT.
+ * ZSHARED: allow concurrent access with other ZSHARED callers.
+ * ZXATTR: we want dzp's xattr directory
+ * ZCILOOK: On a mixed sensitivity file system,
+ * this lookup should be case-insensitive.
+ * ZCIEXACT: On a purely case-insensitive file system,
+ * this lookup should be case-sensitive.
+ * ZRENAMING: we are locking for renaming, force narrow locks
+ *
+ * Output arguments:
+ * zpp - pointer to the znode for the entry (NULL if there isn't one)
+ * dlpp - pointer to the dirlock for this entry (NULL on error)
+ * direntflags - (case-insensitive lookup only)
+ * flags if multiple case-sensitive matches exist in directory
+ * realpnp - (case-insensitive lookup only)
+ * actual name matched within the directory
+ *
+ * Return value: 0 on success or errno on failure.
+ *
+ * NOTE: Always checks for, and rejects, '.' and '..'.
+ * NOTE: For case-insensitive file systems we take wide locks (see below),
+ * but return znode pointers to a single match.
+ */
+int
+zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp,
+ int flag, int *direntflags, pathname_t *realpnp)
+{
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zfs_dirlock_t *dl;
+ boolean_t update;
+ boolean_t exact;
+ uint64_t zoid;
+ vnode_t *vp = NULL;
+ int error = 0;
+ int cmpflags;
+
+ *zpp = NULL;
+ *dlpp = NULL;
+
+ /*
+ * Verify that we are not trying to lock '.', '..', or '.zfs'
+ */
+ if (name[0] == '.' &&
+ (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) ||
+ zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)
+ return (EEXIST);
+
+ /*
+ * Case sensitivity and normalization preferences are set when
+ * the file system is created. These are stored in the
+ * zfsvfs->z_case and zfsvfs->z_norm fields. These choices
+ * affect what vnodes can be cached in the DNLC, how we
+ * perform zap lookups, and the "width" of our dirlocks.
+ *
+ * A normal dirlock locks a single name. Note that with
+ * normalization a name can be composed multiple ways, but
+ * when normalized, these names all compare equal. A wide
+ * dirlock locks multiple names. We need these when the file
+ * system is supporting mixed-mode access. It is sometimes
+ * necessary to lock all case permutations of file name at
+ * once so that simultaneous case-insensitive/case-sensitive
+ * behaves as rationally as possible.
+ */
+
+ /*
+ * Decide if exact matches should be requested when performing
+ * a zap lookup on file systems supporting case-insensitive
+ * access.
+ */
+ exact =
+ ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) ||
+ ((zfsvfs->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK));
+
+ /*
+ * Only look in or update the DNLC if we are looking for the
+ * name on a file system that does not require normalization
+ * or case folding. We can also look there if we happen to be
+ * on a non-normalizing, mixed sensitivity file system IF we
+ * are looking for the exact name.
+ *
+ * Maybe can add TO-UPPERed version of name to dnlc in ci-only
+ * case for performance improvement?
+ */
+ update = !zfsvfs->z_norm ||
+ ((zfsvfs->z_case == ZFS_CASE_MIXED) &&
+ !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK));
+
+ /*
+ * ZRENAMING indicates we are in a situation where we should
+ * take narrow locks regardless of the file system's
+ * preferences for normalizing and case folding. This will
+ * prevent us deadlocking trying to grab the same wide lock
+ * twice if the two names happen to be case-insensitive
+ * matches.
+ */
+ if (flag & ZRENAMING)
+ cmpflags = 0;
+ else
+ cmpflags = zfsvfs->z_norm;
+
+ /*
+ * Wait until there are no locks on this name.
+ */
+ rw_enter(&dzp->z_name_lock, RW_READER);
+ mutex_enter(&dzp->z_lock);
+ for (;;) {
+ if (dzp->z_unlinked) {
+ mutex_exit(&dzp->z_lock);
+ rw_exit(&dzp->z_name_lock);
+ return (ENOENT);
+ }
+ for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) {
+ if ((u8_strcmp(name, dl->dl_name, 0, cmpflags,
+ U8_UNICODE_LATEST, &error) == 0) || error != 0)
+ break;
+ }
+ if (error != 0) {
+ mutex_exit(&dzp->z_lock);
+ rw_exit(&dzp->z_name_lock);
+ return (ENOENT);
+ }
+ if (dl == NULL) {
+ /*
+ * Allocate a new dirlock and add it to the list.
+ */
+ dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP);
+ cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL);
+ dl->dl_name = name;
+ dl->dl_sharecnt = 0;
+ dl->dl_namesize = 0;
+ dl->dl_dzp = dzp;
+ dl->dl_next = dzp->z_dirlocks;
+ dzp->z_dirlocks = dl;
+ break;
+ }
+ if ((flag & ZSHARED) && dl->dl_sharecnt != 0)
+ break;
+ cv_wait(&dl->dl_cv, &dzp->z_lock);
+ }
+
+ if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) {
+ /*
+ * We're the second shared reference to dl. Make a copy of
+ * dl_name in case the first thread goes away before we do.
+ * Note that we initialize the new name before storing its
+ * pointer into dl_name, because the first thread may load
+ * dl->dl_name at any time. He'll either see the old value,
+ * which is his, or the new shared copy; either is OK.
+ */
+ dl->dl_namesize = strlen(dl->dl_name) + 1;
+ name = kmem_alloc(dl->dl_namesize, KM_SLEEP);
+ bcopy(dl->dl_name, name, dl->dl_namesize);
+ dl->dl_name = name;
+ }
+
+ mutex_exit(&dzp->z_lock);
+
+ /*
+ * We have a dirlock on the name. (Note that it is the dirlock,
+ * not the dzp's z_lock, that protects the name in the zap object.)
+ * See if there's an object by this name; if so, put a hold on it.
+ */
+ if (flag & ZXATTR) {
+ zoid = dzp->z_phys->zp_xattr;
+ error = (zoid == 0 ? ENOENT : 0);
+ } else {
+ if (update)
+ vp = dnlc_lookup(ZTOV(dzp), name);
+ if (vp == DNLC_NO_VNODE) {
+ VN_RELE(vp);
+ error = ENOENT;
+ } else if (vp) {
+ if (flag & ZNEW) {
+ zfs_dirent_unlock(dl);
+ VN_RELE(vp);
+ return (EEXIST);
+ }
+ *dlpp = dl;
+ *zpp = VTOZ(vp);
+ return (0);
+ } else {
+ error = zfs_match_find(zfsvfs, dzp, name, exact,
+ update, direntflags, realpnp, &zoid);
+ }
+ }
+ if (error) {
+ if (error != ENOENT || (flag & ZEXISTS)) {
+ zfs_dirent_unlock(dl);
+ return (error);
+ }
+ } else {
+ if (flag & ZNEW) {
+ zfs_dirent_unlock(dl);
+ return (EEXIST);
+ }
+ error = zfs_zget(zfsvfs, zoid, zpp);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ return (error);
+ }
+ if (!(flag & ZXATTR) && update)
+ dnlc_update(ZTOV(dzp), name, ZTOV(*zpp));
+ }
+
+ *dlpp = dl;
+
+ return (0);
+}
+
+/*
+ * Unlock this directory entry and wake anyone who was waiting for it.
+ */
+void
+zfs_dirent_unlock(zfs_dirlock_t *dl)
+{
+ znode_t *dzp = dl->dl_dzp;
+ zfs_dirlock_t **prev_dl, *cur_dl;
+
+ mutex_enter(&dzp->z_lock);
+ rw_exit(&dzp->z_name_lock);
+ if (dl->dl_sharecnt > 1) {
+ dl->dl_sharecnt--;
+ mutex_exit(&dzp->z_lock);
+ return;
+ }
+ prev_dl = &dzp->z_dirlocks;
+ while ((cur_dl = *prev_dl) != dl)
+ prev_dl = &cur_dl->dl_next;
+ *prev_dl = dl->dl_next;
+ cv_broadcast(&dl->dl_cv);
+ mutex_exit(&dzp->z_lock);
+
+ if (dl->dl_namesize != 0)
+ kmem_free(dl->dl_name, dl->dl_namesize);
+ cv_destroy(&dl->dl_cv);
+ kmem_free(dl, sizeof (*dl));
+}
+
+/*
+ * Look up an entry in a directory.
+ *
+ * NOTE: '.' and '..' are handled as special cases because
+ * no directory entries are actually stored for them. If this is
+ * the root of a filesystem, then '.zfs' is also treated as a
+ * special pseudo-directory.
+ */
+int
+zfs_dirlook(znode_t *dzp, char *name, vnode_t **vpp, int flags,
+ int *deflg, pathname_t *rpnp)
+{
+ zfs_dirlock_t *dl;
+ znode_t *zp;
+ int error = 0;
+
+ if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
+ *vpp = ZTOV(dzp);
+ VN_HOLD(*vpp);
+ } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ /*
+ * If we are a snapshot mounted under .zfs, return
+ * the vp for the snapshot directory.
+ */
+ if (dzp->z_phys->zp_parent == dzp->z_id &&
+ zfsvfs->z_parent != zfsvfs) {
+ error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
+ "snapshot", vpp, NULL, 0, NULL, kcred,
+ NULL, NULL, NULL);
+ return (error);
+ }
+ rw_enter(&dzp->z_parent_lock, RW_READER);
+ error = zfs_zget(zfsvfs, dzp->z_phys->zp_parent, &zp);
+ if (error == 0)
+ *vpp = ZTOV(zp);
+ rw_exit(&dzp->z_parent_lock);
+ } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) {
+ *vpp = zfsctl_root(dzp);
+ } else {
+ int zf;
+
+ zf = ZEXISTS | ZSHARED;
+ if (flags & FIGNORECASE)
+ zf |= ZCILOOK;
+
+ error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp);
+ if (error == 0) {
+ *vpp = ZTOV(zp);
+ zfs_dirent_unlock(dl);
+ dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
+ }
+ rpnp = NULL;
+ }
+
+ if ((flags & FIGNORECASE) && rpnp && !error)
+ (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize);
+
+ return (error);
+}
+
+/*
+ * unlinked Set (formerly known as the "delete queue") Error Handling
+ *
+ * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
+ * don't specify the name of the entry that we will be manipulating. We
+ * also fib and say that we won't be adding any new entries to the
+ * unlinked set, even though we might (this is to lower the minimum file
+ * size that can be deleted in a full filesystem). So on the small
+ * chance that the nlink list is using a fat zap (ie. has more than
+ * 2000 entries), we *may* not pre-read a block that's needed.
+ * Therefore it is remotely possible for some of the assertions
+ * regarding the unlinked set below to fail due to i/o error. On a
+ * nondebug system, this will result in the space being leaked.
+ */
+void
+zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ ASSERT(zp->z_unlinked);
+ ASSERT3U(zp->z_phys->zp_links, ==, 0);
+
+ VERIFY3U(0, ==,
+ zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
+}
+
+/*
+ * Clean up any znodes that had no links when we either crashed or
+ * (force) umounted the file system.
+ */
+void
+zfs_unlinked_drain(zfsvfs_t *zfsvfs)
+{
+ zap_cursor_t zc;
+ zap_attribute_t zap;
+ dmu_object_info_t doi;
+ znode_t *zp;
+ int error;
+
+ /*
+ * Interate over the contents of the unlinked set.
+ */
+ for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
+ zap_cursor_retrieve(&zc, &zap) == 0;
+ zap_cursor_advance(&zc)) {
+
+ /*
+ * See what kind of object we have in list
+ */
+
+ error = dmu_object_info(zfsvfs->z_os,
+ zap.za_first_integer, &doi);
+ if (error != 0)
+ continue;
+
+ ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
+ (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
+ /*
+ * We need to re-mark these list entries for deletion,
+ * so we pull them back into core and set zp->z_unlinked.
+ */
+ error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
+
+ /*
+ * We may pick up znodes that are already marked for deletion.
+ * This could happen during the purge of an extended attribute
+ * directory. All we need to do is skip over them, since they
+ * are already in the system marked z_unlinked.
+ */
+ if (error != 0)
+ continue;
+
+ zp->z_unlinked = B_TRUE;
+ VN_RELE(ZTOV(zp));
+ }
+ zap_cursor_fini(&zc);
+}
+
+/*
+ * Delete the entire contents of a directory. Return a count
+ * of the number of entries that could not be deleted. If we encounter
+ * an error, return a count of at least one so that the directory stays
+ * in the unlinked set.
+ *
+ * NOTE: this function assumes that the directory is inactive,
+ * so there is no need to lock its entries before deletion.
+ * Also, it assumes the directory contents is *only* regular
+ * files.
+ */
+static int
+zfs_purgedir(znode_t *dzp)
+{
+ zap_cursor_t zc;
+ zap_attribute_t zap;
+ znode_t *xzp;
+ dmu_tx_t *tx;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zfs_dirlock_t dl;
+ int skipped = 0;
+ int error;
+
+ for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
+ (error = zap_cursor_retrieve(&zc, &zap)) == 0;
+ zap_cursor_advance(&zc)) {
+ error = zfs_zget(zfsvfs,
+ ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
+ if (error) {
+ skipped += 1;
+ continue;
+ }
+
+ ASSERT((ZTOV(xzp)->v_type == VREG) ||
+ (ZTOV(xzp)->v_type == VLNK));
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, dzp->z_id);
+ dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
+ dmu_tx_hold_bonus(tx, xzp->z_id);
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ VN_RELE(ZTOV(xzp));
+ skipped += 1;
+ continue;
+ }
+ bzero(&dl, sizeof (dl));
+ dl.dl_dzp = dzp;
+ dl.dl_name = zap.za_name;
+
+ error = zfs_link_destroy(&dl, xzp, tx, 0, NULL);
+ if (error)
+ skipped += 1;
+ dmu_tx_commit(tx);
+
+ VN_RELE(ZTOV(xzp));
+ }
+ zap_cursor_fini(&zc);
+ if (error != ENOENT)
+ skipped += 1;
+ return (skipped);
+}
+
+void
+zfs_rmnode(znode_t *zp)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ objset_t *os = zfsvfs->z_os;
+ znode_t *xzp = NULL;
+ dmu_tx_t *tx;
+ uint64_t acl_obj;
+ int error;
+
+ ASSERT(ZTOV(zp)->v_count == 0);
+ ASSERT(zp->z_phys->zp_links == 0);
+
+ /*
+ * If this is a ZIL replay then leave the object in the unlinked set.
+ * Otherwise we can get a deadlock, because the delete can be
+ * quite large and span multiple tx's and txgs, but each replay
+ * creates a tx to atomically run the replay function and mark the
+ * replay record as complete. We deadlock trying to start a tx in
+ * a new txg to further the deletion but can't because the replay
+ * tx hasn't finished.
+ *
+ * We actually delete the object if we get a failure to create an
+ * object in zil_replay_log_record(), or after calling zil_replay().
+ */
+ if (zfsvfs->z_assign >= TXG_INITIAL) {
+ zfs_znode_dmu_fini(zp);
+ zfs_znode_free(zp);
+ return;
+ }
+
+ /*
+ * If this is an attribute directory, purge its contents.
+ */
+ if (ZTOV(zp)->v_type == VDIR && (zp->z_phys->zp_flags & ZFS_XATTR)) {
+ if (zfs_purgedir(zp) != 0) {
+ /*
+ * Not enough space to delete some xattrs.
+ * Leave it in the unlinked set.
+ */
+ zfs_znode_dmu_fini(zp);
+ zfs_znode_free(zp);
+ return;
+ }
+ }
+
+ /*
+ * Free up all the data in the file.
+ */
+ error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
+ if (error) {
+ /*
+ * Not enough space. Leave the file in the unlinked set.
+ */
+ zfs_znode_dmu_fini(zp);
+ zfs_znode_free(zp);
+ return;
+ }
+
+ /*
+ * If the file has extended attributes, we're going to unlink
+ * the xattr dir.
+ */
+ if (zp->z_phys->zp_xattr) {
+ error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
+ ASSERT(error == 0);
+ }
+
+ acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
+
+ /*
+ * Set up the final transaction.
+ */
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
+ if (xzp) {
+ dmu_tx_hold_bonus(tx, xzp->z_id);
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
+ }
+ if (acl_obj)
+ dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ /*
+ * Not enough space to delete the file. Leave it in the
+ * unlinked set, leaking it until the fs is remounted (at
+ * which point we'll call zfs_unlinked_drain() to process it).
+ */
+ dmu_tx_abort(tx);
+ zfs_znode_dmu_fini(zp);
+ zfs_znode_free(zp);
+ goto out;
+ }
+
+ if (xzp) {
+ dmu_buf_will_dirty(xzp->z_dbuf, tx);
+ mutex_enter(&xzp->z_lock);
+ xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
+ xzp->z_phys->zp_links = 0; /* no more links to it */
+ mutex_exit(&xzp->z_lock);
+ zfs_unlinked_add(xzp, tx);
+ }
+
+ /* Remove this znode from the unlinked set */
+ VERIFY3U(0, ==,
+ zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
+
+ zfs_znode_delete(zp, tx);
+
+ dmu_tx_commit(tx);
+out:
+ if (xzp)
+ VN_RELE(ZTOV(xzp));
+}
+
+static uint64_t
+zfs_dirent(znode_t *zp)
+{
+ uint64_t de = zp->z_id;
+ if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
+ de |= IFTODT((zp)->z_phys->zp_mode) << 60;
+ return (de);
+}
+
+/*
+ * Link zp into dl. Can only fail if zp has been unlinked.
+ */
+int
+zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
+{
+ znode_t *dzp = dl->dl_dzp;
+ vnode_t *vp = ZTOV(zp);
+ uint64_t value;
+ int zp_is_dir = (vp->v_type == VDIR);
+ int error;
+
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+ mutex_enter(&zp->z_lock);
+
+ if (!(flag & ZRENAMING)) {
+ if (zp->z_unlinked) { /* no new links to unlinked zp */
+ ASSERT(!(flag & (ZNEW | ZEXISTS)));
+ mutex_exit(&zp->z_lock);
+ return (ENOENT);
+ }
+ zp->z_phys->zp_links++;
+ }
+ zp->z_phys->zp_parent = dzp->z_id; /* dzp is now zp's parent */
+
+ if (!(flag & ZNEW))
+ zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
+ mutex_exit(&zp->z_lock);
+
+ dmu_buf_will_dirty(dzp->z_dbuf, tx);
+ mutex_enter(&dzp->z_lock);
+ dzp->z_phys->zp_size++; /* one dirent added */
+ dzp->z_phys->zp_links += zp_is_dir; /* ".." link from zp */
+ zfs_time_stamper_locked(dzp, CONTENT_MODIFIED, tx);
+ mutex_exit(&dzp->z_lock);
+
+ value = zfs_dirent(zp);
+ error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name,
+ 8, 1, &value, tx);
+ ASSERT(error == 0);
+
+ dnlc_update(ZTOV(dzp), dl->dl_name, vp);
+
+ return (0);
+}
+
+/*
+ * Unlink zp from dl, and mark zp for deletion if this was the last link.
+ * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
+ * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
+ * If it's non-NULL, we use it to indicate whether the znode needs deletion,
+ * and it's the caller's job to do it.
+ */
+int
+zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
+ boolean_t *unlinkedp)
+{
+ znode_t *dzp = dl->dl_dzp;
+ vnode_t *vp = ZTOV(zp);
+ int zp_is_dir = (vp->v_type == VDIR);
+ boolean_t unlinked = B_FALSE;
+ int error;
+
+ dnlc_remove(ZTOV(dzp), dl->dl_name);
+
+ if (!(flag & ZRENAMING)) {
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+
+ if (vn_vfswlock(vp)) /* prevent new mounts on zp */
+ return (EBUSY);
+
+ if (vn_ismntpt(vp)) { /* don't remove mount point */
+ vn_vfsunlock(vp);
+ return (EBUSY);
+ }
+
+ mutex_enter(&zp->z_lock);
+ if (zp_is_dir && !zfs_dirempty(zp)) { /* dir not empty */
+ mutex_exit(&zp->z_lock);
+ vn_vfsunlock(vp);
+ return (EEXIST);
+ }
+ if (zp->z_phys->zp_links <= zp_is_dir) {
+ zfs_panic_recover("zfs: link count on %s is %u, "
+ "should be at least %u",
+ zp->z_vnode->v_path ? zp->z_vnode->v_path :
+ "<unknown>", (int)zp->z_phys->zp_links,
+ zp_is_dir + 1);
+ zp->z_phys->zp_links = zp_is_dir + 1;
+ }
+ if (--zp->z_phys->zp_links == zp_is_dir) {
+ zp->z_unlinked = B_TRUE;
+ zp->z_phys->zp_links = 0;
+ unlinked = B_TRUE;
+ } else {
+ zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
+ }
+ mutex_exit(&zp->z_lock);
+ vn_vfsunlock(vp);
+ }
+
+ dmu_buf_will_dirty(dzp->z_dbuf, tx);
+ mutex_enter(&dzp->z_lock);
+ dzp->z_phys->zp_size--; /* one dirent removed */
+ dzp->z_phys->zp_links -= zp_is_dir; /* ".." link from zp */
+ zfs_time_stamper_locked(dzp, CONTENT_MODIFIED, tx);
+ mutex_exit(&dzp->z_lock);
+
+ if (zp->z_zfsvfs->z_norm) {
+ if (((zp->z_zfsvfs->z_case == ZFS_CASE_INSENSITIVE) &&
+ (flag & ZCIEXACT)) ||
+ ((zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) &&
+ !(flag & ZCILOOK)))
+ error = zap_remove_norm(zp->z_zfsvfs->z_os,
+ dzp->z_id, dl->dl_name, MT_EXACT, tx);
+ else
+ error = zap_remove_norm(zp->z_zfsvfs->z_os,
+ dzp->z_id, dl->dl_name, MT_FIRST, tx);
+ } else {
+ error = zap_remove(zp->z_zfsvfs->z_os,
+ dzp->z_id, dl->dl_name, tx);
+ }
+ ASSERT(error == 0);
+
+ if (unlinkedp != NULL)
+ *unlinkedp = unlinked;
+ else if (unlinked)
+ zfs_unlinked_add(zp, tx);
+
+ return (0);
+}
+
+/*
+ * Indicate whether the directory is empty. Works with or without z_lock
+ * held, but can only be consider a hint in the latter case. Returns true
+ * if only "." and ".." remain and there's no work in progress.
+ */
+boolean_t
+zfs_dirempty(znode_t *dzp)
+{
+ return (dzp->z_phys->zp_size == 2 && dzp->z_dirlocks == 0);
+}
+
+int
+zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ znode_t *xzp;
+ dmu_tx_t *tx;
+ int error;
+ zfs_fuid_info_t *fuidp = NULL;
+
+ *xvpp = NULL;
+
+ if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
+ return (error);
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
+ if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT)
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ return (error);
+ }
+ zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, NULL, &fuidp);
+ ASSERT(xzp->z_phys->zp_parent == zp->z_id);
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+ zp->z_phys->zp_xattr = xzp->z_id;
+
+ (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
+ xzp, "", NULL, fuidp, vap);
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+ dmu_tx_commit(tx);
+
+ *xvpp = ZTOV(xzp);
+
+ return (0);
+}
+
+/*
+ * Return a znode for the extended attribute directory for zp.
+ * ** If the directory does not already exist, it is created **
+ *
+ * IN: zp - znode to obtain attribute directory from
+ * cr - credentials of caller
+ * flags - flags from the VOP_LOOKUP call
+ *
+ * OUT: xzpp - pointer to extended attribute znode
+ *
+ * RETURN: 0 on success
+ * error number on failure
+ */
+int
+zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ znode_t *xzp;
+ zfs_dirlock_t *dl;
+ vattr_t va;
+ int error;
+top:
+ error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL);
+ if (error)
+ return (error);
+
+ if (xzp != NULL) {
+ *xvpp = ZTOV(xzp);
+ zfs_dirent_unlock(dl);
+ return (0);
+ }
+
+ ASSERT(zp->z_phys->zp_xattr == 0);
+
+ if (!(flags & CREATE_XATTR_DIR)) {
+ zfs_dirent_unlock(dl);
+ return (ENOENT);
+ }
+
+ if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
+ zfs_dirent_unlock(dl);
+ return (EROFS);
+ }
+
+ /*
+ * The ability to 'create' files in an attribute
+ * directory comes from the write_xattr permission on the base file.
+ *
+ * The ability to 'search' an attribute directory requires
+ * read_xattr permission on the base file.
+ *
+ * Once in a directory the ability to read/write attributes
+ * is controlled by the permissions on the attribute file.
+ */
+ va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
+ va.va_type = VDIR;
+ va.va_mode = S_IFDIR | S_ISVTX | 0777;
+ zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
+
+ error = zfs_make_xattrdir(zp, &va, xvpp, cr);
+ zfs_dirent_unlock(dl);
+
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ /* NB: we already did dmu_tx_wait() if necessary */
+ goto top;
+ }
+
+ return (error);
+}
+
+/*
+ * Decide whether it is okay to remove within a sticky directory.
+ *
+ * In sticky directories, write access is not sufficient;
+ * you can remove entries from a directory only if:
+ *
+ * you own the directory,
+ * you own the entry,
+ * the entry is a plain file and you have write access,
+ * or you are privileged (checked in secpolicy...).
+ *
+ * The function returns 0 if remove access is granted.
+ */
+int
+zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
+{
+ uid_t uid;
+ uid_t downer;
+ uid_t fowner;
+ zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
+
+ if (zdp->z_zfsvfs->z_assign >= TXG_INITIAL) /* ZIL replay */
+ return (0);
+
+ if ((zdp->z_phys->zp_mode & S_ISVTX) == 0)
+ return (0);
+
+ downer = zfs_fuid_map_id(zfsvfs, zdp->z_phys->zp_uid, cr, ZFS_OWNER);
+ fowner = zfs_fuid_map_id(zfsvfs, zp->z_phys->zp_uid, cr, ZFS_OWNER);
+
+ if ((uid = crgetuid(cr)) == downer || uid == fowner ||
+ (ZTOV(zp)->v_type == VREG &&
+ zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
+ return (0);
+ else
+ return (secpolicy_vnode_remove(cr));
+}
diff --git a/module/zfs/zfs_fm.c b/module/zfs/zfs_fm.c
new file mode 100644
index 000000000..236d69e7e
--- /dev/null
+++ b/module/zfs/zfs_fm.c
@@ -0,0 +1,362 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/vdev.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio.h>
+
+#include <sys/fm/fs/zfs.h>
+#include <sys/fm/protocol.h>
+#include <sys/fm/util.h>
+#include <sys/sysevent.h>
+
+/*
+ * This general routine is responsible for generating all the different ZFS
+ * ereports. The payload is dependent on the class, and which arguments are
+ * supplied to the function:
+ *
+ * EREPORT POOL VDEV IO
+ * block X X X
+ * data X X
+ * device X X
+ * pool X
+ *
+ * If we are in a loading state, all errors are chained together by the same
+ * SPA-wide ENA (Error Numeric Association).
+ *
+ * For isolated I/O requests, we get the ENA from the zio_t. The propagation
+ * gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want
+ * to chain together all ereports associated with a logical piece of data. For
+ * read I/Os, there are basically three 'types' of I/O, which form a roughly
+ * layered diagram:
+ *
+ * +---------------+
+ * | Aggregate I/O | No associated logical data or device
+ * +---------------+
+ * |
+ * V
+ * +---------------+ Reads associated with a piece of logical data.
+ * | Read I/O | This includes reads on behalf of RAID-Z,
+ * +---------------+ mirrors, gang blocks, retries, etc.
+ * |
+ * V
+ * +---------------+ Reads associated with a particular device, but
+ * | Physical I/O | no logical data. Issued as part of vdev caching
+ * +---------------+ and I/O aggregation.
+ *
+ * Note that 'physical I/O' here is not the same terminology as used in the rest
+ * of ZIO. Typically, 'physical I/O' simply means that there is no attached
+ * blockpointer. But I/O with no associated block pointer can still be related
+ * to a logical piece of data (i.e. RAID-Z requests).
+ *
+ * Purely physical I/O always have unique ENAs. They are not related to a
+ * particular piece of logical data, and therefore cannot be chained together.
+ * We still generate an ereport, but the DE doesn't correlate it with any
+ * logical piece of data. When such an I/O fails, the delegated I/O requests
+ * will issue a retry, which will trigger the 'real' ereport with the correct
+ * ENA.
+ *
+ * We keep track of the ENA for a ZIO chain through the 'io_logical' member.
+ * When a new logical I/O is issued, we set this to point to itself. Child I/Os
+ * then inherit this pointer, so that when it is first set subsequent failures
+ * will use the same ENA. For vdev cache fill and queue aggregation I/O,
+ * this pointer is set to NULL, and no ereport will be generated (since it
+ * doesn't actually correspond to any particular device or piece of data,
+ * and the caller will always retry without caching or queueing anyway).
+ */
+void
+zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
+ uint64_t stateoroffset, uint64_t size)
+{
+#ifdef _KERNEL
+ nvlist_t *ereport, *detector;
+ uint64_t ena;
+ char class[64];
+ int state;
+
+ /*
+ * If we are doing a spa_tryimport(), ignore errors.
+ */
+ if (spa->spa_load_state == SPA_LOAD_TRYIMPORT)
+ return;
+
+ /*
+ * If we are in the middle of opening a pool, and the previous attempt
+ * failed, don't bother logging any new ereports - we're just going to
+ * get the same diagnosis anyway.
+ */
+ if (spa->spa_load_state != SPA_LOAD_NONE &&
+ spa->spa_last_open_failed)
+ return;
+
+ if (zio != NULL) {
+ /*
+ * If this is not a read or write zio, ignore the error. This
+ * can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
+ */
+ if (zio->io_type != ZIO_TYPE_READ &&
+ zio->io_type != ZIO_TYPE_WRITE)
+ return;
+
+ /*
+ * Ignore any errors from speculative I/Os, as failure is an
+ * expected result.
+ */
+ if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
+ return;
+
+ /*
+ * If the vdev has already been marked as failing due to a
+ * failed probe, then ignore any subsequent I/O errors, as the
+ * DE will automatically fault the vdev on the first such
+ * failure.
+ */
+ if (vd != NULL &&
+ (!vdev_readable(vd) || !vdev_writeable(vd)) &&
+ strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) != 0)
+ return;
+ }
+
+ if ((ereport = fm_nvlist_create(NULL)) == NULL)
+ return;
+
+ if ((detector = fm_nvlist_create(NULL)) == NULL) {
+ fm_nvlist_destroy(ereport, FM_NVA_FREE);
+ return;
+ }
+
+ /*
+ * Serialize ereport generation
+ */
+ mutex_enter(&spa->spa_errlist_lock);
+
+ /*
+ * Determine the ENA to use for this event. If we are in a loading
+ * state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use
+ * a root zio-wide ENA. Otherwise, simply use a unique ENA.
+ */
+ if (spa->spa_load_state != SPA_LOAD_NONE) {
+ if (spa->spa_ena == 0)
+ spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
+ ena = spa->spa_ena;
+ } else if (zio != NULL && zio->io_logical != NULL) {
+ if (zio->io_logical->io_ena == 0)
+ zio->io_logical->io_ena =
+ fm_ena_generate(0, FM_ENA_FMT1);
+ ena = zio->io_logical->io_ena;
+ } else {
+ ena = fm_ena_generate(0, FM_ENA_FMT1);
+ }
+
+ /*
+ * Construct the full class, detector, and other standard FMA fields.
+ */
+ (void) snprintf(class, sizeof (class), "%s.%s",
+ ZFS_ERROR_CLASS, subclass);
+
+ fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa),
+ vd != NULL ? vd->vdev_guid : 0);
+
+ fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL);
+
+ /*
+ * Construct the per-ereport payload, depending on which parameters are
+ * passed in.
+ */
+
+ /*
+ * If we are importing a faulted pool, then we treat it like an open,
+ * not an import. Otherwise, the DE will ignore all faults during
+ * import, since the default behavior is to mark the devices as
+ * persistently unavailable, not leave them in the faulted state.
+ */
+ state = spa->spa_import_faulted ? SPA_LOAD_OPEN : spa->spa_load_state;
+
+ /*
+ * Generic payload members common to all ereports.
+ */
+ fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL,
+ DATA_TYPE_STRING, spa_name(spa), FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
+ DATA_TYPE_UINT64, spa_guid(spa),
+ FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
+ state, NULL);
+
+ if (spa != NULL) {
+ fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
+ DATA_TYPE_STRING,
+ spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
+ FM_EREPORT_FAILMODE_WAIT :
+ spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
+ FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC,
+ NULL);
+ }
+
+ if (vd != NULL) {
+ vdev_t *pvd = vd->vdev_parent;
+
+ fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
+ DATA_TYPE_UINT64, vd->vdev_guid,
+ FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
+ DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
+ if (vd->vdev_path)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
+ DATA_TYPE_STRING, vd->vdev_path, NULL);
+ if (vd->vdev_devid)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
+ DATA_TYPE_STRING, vd->vdev_devid, NULL);
+
+ if (pvd != NULL) {
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID,
+ DATA_TYPE_UINT64, pvd->vdev_guid,
+ FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
+ DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type,
+ NULL);
+ if (pvd->vdev_path)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
+ DATA_TYPE_STRING, pvd->vdev_path, NULL);
+ if (pvd->vdev_devid)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
+ DATA_TYPE_STRING, pvd->vdev_devid, NULL);
+ }
+ }
+
+ if (zio != NULL) {
+ /*
+ * Payload common to all I/Os.
+ */
+ fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
+ DATA_TYPE_INT32, zio->io_error, NULL);
+
+ /*
+ * If the 'size' parameter is non-zero, it indicates this is a
+ * RAID-Z or other I/O where the physical offset and length are
+ * provided for us, instead of within the zio_t.
+ */
+ if (vd != NULL) {
+ if (size)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
+ DATA_TYPE_UINT64, stateoroffset,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
+ DATA_TYPE_UINT64, size, NULL);
+ else
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
+ DATA_TYPE_UINT64, zio->io_offset,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
+ DATA_TYPE_UINT64, zio->io_size, NULL);
+ }
+
+ /*
+ * Payload for I/Os with corresponding logical information.
+ */
+ if (zio->io_logical != NULL)
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET,
+ DATA_TYPE_UINT64,
+ zio->io_logical->io_bookmark.zb_objset,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
+ DATA_TYPE_UINT64,
+ zio->io_logical->io_bookmark.zb_object,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
+ DATA_TYPE_INT64,
+ zio->io_logical->io_bookmark.zb_level,
+ FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
+ DATA_TYPE_UINT64,
+ zio->io_logical->io_bookmark.zb_blkid, NULL);
+ } else if (vd != NULL) {
+ /*
+ * If we have a vdev but no zio, this is a device fault, and the
+ * 'stateoroffset' parameter indicates the previous state of the
+ * vdev.
+ */
+ fm_payload_set(ereport,
+ FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
+ DATA_TYPE_UINT64, stateoroffset, NULL);
+ }
+ mutex_exit(&spa->spa_errlist_lock);
+
+ fm_ereport_post(ereport, EVCH_SLEEP);
+
+ fm_nvlist_destroy(ereport, FM_NVA_FREE);
+ fm_nvlist_destroy(detector, FM_NVA_FREE);
+#endif
+}
+
+static void
+zfs_post_common(spa_t *spa, vdev_t *vd, const char *name)
+{
+#ifdef _KERNEL
+ nvlist_t *resource;
+ char class[64];
+
+ if ((resource = fm_nvlist_create(NULL)) == NULL)
+ return;
+
+ (void) snprintf(class, sizeof (class), "%s.%s.%s", FM_RSRC_RESOURCE,
+ ZFS_ERROR_CLASS, name);
+ VERIFY(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION) == 0);
+ VERIFY(nvlist_add_string(resource, FM_CLASS, class) == 0);
+ VERIFY(nvlist_add_uint64(resource,
+ FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa)) == 0);
+ if (vd)
+ VERIFY(nvlist_add_uint64(resource,
+ FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid) == 0);
+
+ fm_ereport_post(resource, EVCH_SLEEP);
+
+ fm_nvlist_destroy(resource, FM_NVA_FREE);
+#endif
+}
+
+/*
+ * The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
+ * has been removed from the system. This will cause the DE to ignore any
+ * recent I/O errors, inferring that they are due to the asynchronous device
+ * removal.
+ */
+void
+zfs_post_remove(spa_t *spa, vdev_t *vd)
+{
+ zfs_post_common(spa, vd, FM_RESOURCE_REMOVED);
+}
+
+/*
+ * The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
+ * has the 'autoreplace' property set, and therefore any broken vdevs will be
+ * handled by higher level logic, and no vdev fault should be generated.
+ */
+void
+zfs_post_autoreplace(spa_t *spa, vdev_t *vd)
+{
+ zfs_post_common(spa, vd, FM_RESOURCE_AUTOREPLACE);
+}
diff --git a/module/zfs/zfs_fuid.c b/module/zfs/zfs_fuid.c
new file mode 100644
index 000000000..7cb505258
--- /dev/null
+++ b/module/zfs/zfs_fuid.c
@@ -0,0 +1,704 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/sunddi.h>
+#include <sys/dmu.h>
+#include <sys/avl.h>
+#include <sys/zap.h>
+#include <sys/refcount.h>
+#include <sys/nvpair.h>
+#ifdef _KERNEL
+#include <sys/kidmap.h>
+#include <sys/sid.h>
+#include <sys/zfs_vfsops.h>
+#include <sys/zfs_znode.h>
+#endif
+#include <sys/zfs_fuid.h>
+
+/*
+ * FUID Domain table(s).
+ *
+ * The FUID table is stored as a packed nvlist of an array
+ * of nvlists which contain an index, domain string and offset
+ *
+ * During file system initialization the nvlist(s) are read and
+ * two AVL trees are created. One tree is keyed by the index number
+ * and the other by the domain string. Nodes are never removed from
+ * trees, but new entries may be added. If a new entry is added then the
+ * on-disk packed nvlist will also be updated.
+ */
+
+#define FUID_IDX "fuid_idx"
+#define FUID_DOMAIN "fuid_domain"
+#define FUID_OFFSET "fuid_offset"
+#define FUID_NVP_ARRAY "fuid_nvlist"
+
+typedef struct fuid_domain {
+ avl_node_t f_domnode;
+ avl_node_t f_idxnode;
+ ksiddomain_t *f_ksid;
+ uint64_t f_idx;
+} fuid_domain_t;
+
+static char *nulldomain = "";
+
+/*
+ * Compare two indexes.
+ */
+static int
+idx_compare(const void *arg1, const void *arg2)
+{
+ const fuid_domain_t *node1 = arg1;
+ const fuid_domain_t *node2 = arg2;
+
+ if (node1->f_idx < node2->f_idx)
+ return (-1);
+ else if (node1->f_idx > node2->f_idx)
+ return (1);
+ return (0);
+}
+
+/*
+ * Compare two domain strings.
+ */
+static int
+domain_compare(const void *arg1, const void *arg2)
+{
+ const fuid_domain_t *node1 = arg1;
+ const fuid_domain_t *node2 = arg2;
+ int val;
+
+ val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
+ if (val == 0)
+ return (0);
+ return (val > 0 ? 1 : -1);
+}
+
+/*
+ * load initial fuid domain and idx trees. This function is used by
+ * both the kernel and zdb.
+ */
+uint64_t
+zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
+ avl_tree_t *domain_tree)
+{
+ dmu_buf_t *db;
+ uint64_t fuid_size;
+
+ avl_create(idx_tree, idx_compare,
+ sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
+ avl_create(domain_tree, domain_compare,
+ sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
+
+ VERIFY(0 == dmu_bonus_hold(os, fuid_obj, FTAG, &db));
+ fuid_size = *(uint64_t *)db->db_data;
+ dmu_buf_rele(db, FTAG);
+
+ if (fuid_size) {
+ nvlist_t **fuidnvp;
+ nvlist_t *nvp = NULL;
+ uint_t count;
+ char *packed;
+ int i;
+
+ packed = kmem_alloc(fuid_size, KM_SLEEP);
+ VERIFY(dmu_read(os, fuid_obj, 0, fuid_size, packed) == 0);
+ VERIFY(nvlist_unpack(packed, fuid_size,
+ &nvp, 0) == 0);
+ VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
+ &fuidnvp, &count) == 0);
+
+ for (i = 0; i != count; i++) {
+ fuid_domain_t *domnode;
+ char *domain;
+ uint64_t idx;
+
+ VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
+ &domain) == 0);
+ VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
+ &idx) == 0);
+
+ domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
+
+ domnode->f_idx = idx;
+ domnode->f_ksid = ksid_lookupdomain(domain);
+ avl_add(idx_tree, domnode);
+ avl_add(domain_tree, domnode);
+ }
+ nvlist_free(nvp);
+ kmem_free(packed, fuid_size);
+ }
+ return (fuid_size);
+}
+
+void
+zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
+{
+ fuid_domain_t *domnode;
+ void *cookie;
+
+ cookie = NULL;
+ while (domnode = avl_destroy_nodes(domain_tree, &cookie))
+ ksiddomain_rele(domnode->f_ksid);
+
+ avl_destroy(domain_tree);
+ cookie = NULL;
+ while (domnode = avl_destroy_nodes(idx_tree, &cookie))
+ kmem_free(domnode, sizeof (fuid_domain_t));
+ avl_destroy(idx_tree);
+}
+
+char *
+zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
+{
+ fuid_domain_t searchnode, *findnode;
+ avl_index_t loc;
+
+ searchnode.f_idx = idx;
+
+ findnode = avl_find(idx_tree, &searchnode, &loc);
+
+ return (findnode ? findnode->f_ksid->kd_name : nulldomain);
+}
+
+#ifdef _KERNEL
+/*
+ * Load the fuid table(s) into memory.
+ */
+static void
+zfs_fuid_init(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
+{
+ int error = 0;
+
+ rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
+
+ if (zfsvfs->z_fuid_loaded) {
+ rw_exit(&zfsvfs->z_fuid_lock);
+ return;
+ }
+
+ if (zfsvfs->z_fuid_obj == 0) {
+
+ /* first make sure we need to allocate object */
+
+ error = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
+ ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
+ if (error == ENOENT && tx != NULL) {
+ zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
+ DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
+ sizeof (uint64_t), tx);
+ VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
+ ZFS_FUID_TABLES, sizeof (uint64_t), 1,
+ &zfsvfs->z_fuid_obj, tx) == 0);
+ }
+ }
+
+ if (zfsvfs->z_fuid_obj != 0) {
+ zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
+ zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
+ &zfsvfs->z_fuid_domain);
+ zfsvfs->z_fuid_loaded = B_TRUE;
+ }
+
+ rw_exit(&zfsvfs->z_fuid_lock);
+}
+
+/*
+ * Query domain table for a given domain.
+ *
+ * If domain isn't found it is added to AVL trees and
+ * the results are pushed out to disk.
+ */
+int
+zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
+ dmu_tx_t *tx)
+{
+ fuid_domain_t searchnode, *findnode;
+ avl_index_t loc;
+ krw_t rw = RW_READER;
+
+ /*
+ * If the dummy "nobody" domain then return an index of 0
+ * to cause the created FUID to be a standard POSIX id
+ * for the user nobody.
+ */
+ if (domain[0] == '\0') {
+ *retdomain = nulldomain;
+ return (0);
+ }
+
+ searchnode.f_ksid = ksid_lookupdomain(domain);
+ if (retdomain) {
+ *retdomain = searchnode.f_ksid->kd_name;
+ }
+ if (!zfsvfs->z_fuid_loaded)
+ zfs_fuid_init(zfsvfs, tx);
+
+retry:
+ rw_enter(&zfsvfs->z_fuid_lock, rw);
+ findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
+
+ if (findnode) {
+ rw_exit(&zfsvfs->z_fuid_lock);
+ ksiddomain_rele(searchnode.f_ksid);
+ return (findnode->f_idx);
+ } else {
+ fuid_domain_t *domnode;
+ nvlist_t *nvp;
+ nvlist_t **fuids;
+ uint64_t retidx;
+ size_t nvsize = 0;
+ char *packed;
+ dmu_buf_t *db;
+ int i = 0;
+
+ if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
+ rw_exit(&zfsvfs->z_fuid_lock);
+ rw = RW_WRITER;
+ goto retry;
+ }
+
+ domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
+ domnode->f_ksid = searchnode.f_ksid;
+
+ retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
+
+ avl_add(&zfsvfs->z_fuid_domain, domnode);
+ avl_add(&zfsvfs->z_fuid_idx, domnode);
+ /*
+ * Now resync the on-disk nvlist.
+ */
+ VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+ domnode = avl_first(&zfsvfs->z_fuid_domain);
+ fuids = kmem_alloc(retidx * sizeof (void *), KM_SLEEP);
+ while (domnode) {
+ VERIFY(nvlist_alloc(&fuids[i],
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
+ domnode->f_idx) == 0);
+ VERIFY(nvlist_add_uint64(fuids[i],
+ FUID_OFFSET, 0) == 0);
+ VERIFY(nvlist_add_string(fuids[i++], FUID_DOMAIN,
+ domnode->f_ksid->kd_name) == 0);
+ domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode);
+ }
+ VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
+ fuids, retidx) == 0);
+ for (i = 0; i != retidx; i++)
+ nvlist_free(fuids[i]);
+ kmem_free(fuids, retidx * sizeof (void *));
+ VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
+ packed = kmem_alloc(nvsize, KM_SLEEP);
+ VERIFY(nvlist_pack(nvp, &packed, &nvsize,
+ NV_ENCODE_XDR, KM_SLEEP) == 0);
+ nvlist_free(nvp);
+ zfsvfs->z_fuid_size = nvsize;
+ dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
+ zfsvfs->z_fuid_size, packed, tx);
+ kmem_free(packed, zfsvfs->z_fuid_size);
+ VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
+ FTAG, &db));
+ dmu_buf_will_dirty(db, tx);
+ *(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
+ dmu_buf_rele(db, FTAG);
+
+ rw_exit(&zfsvfs->z_fuid_lock);
+ return (retidx);
+ }
+}
+
+/*
+ * Query domain table by index, returning domain string
+ *
+ * Returns a pointer from an avl node of the domain string.
+ *
+ */
+static char *
+zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
+{
+ char *domain;
+
+ if (idx == 0 || !zfsvfs->z_use_fuids)
+ return (NULL);
+
+ if (!zfsvfs->z_fuid_loaded)
+ zfs_fuid_init(zfsvfs, NULL);
+
+ rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
+
+ if (zfsvfs->z_fuid_obj)
+ domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
+ else
+ domain = nulldomain;
+ rw_exit(&zfsvfs->z_fuid_lock);
+
+ ASSERT(domain);
+ return (domain);
+}
+
+void
+zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
+{
+ *uidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_phys->zp_uid,
+ cr, ZFS_OWNER);
+ *gidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_phys->zp_gid,
+ cr, ZFS_GROUP);
+}
+
+uid_t
+zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
+ cred_t *cr, zfs_fuid_type_t type)
+{
+ uint32_t index = FUID_INDEX(fuid);
+ char *domain;
+ uid_t id;
+
+ if (index == 0)
+ return (fuid);
+
+ domain = zfs_fuid_find_by_idx(zfsvfs, index);
+ ASSERT(domain != NULL);
+
+ if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
+ (void) kidmap_getuidbysid(crgetzone(cr), domain,
+ FUID_RID(fuid), &id);
+ } else {
+ (void) kidmap_getgidbysid(crgetzone(cr), domain,
+ FUID_RID(fuid), &id);
+ }
+ return (id);
+}
+
+/*
+ * Add a FUID node to the list of fuid's being created for this
+ * ACL
+ *
+ * If ACL has multiple domains, then keep only one copy of each unique
+ * domain.
+ */
+static void
+zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
+ uint64_t idx, uint64_t id, zfs_fuid_type_t type)
+{
+ zfs_fuid_t *fuid;
+ zfs_fuid_domain_t *fuid_domain;
+ zfs_fuid_info_t *fuidp;
+ uint64_t fuididx;
+ boolean_t found = B_FALSE;
+
+ if (*fuidpp == NULL)
+ *fuidpp = zfs_fuid_info_alloc();
+
+ fuidp = *fuidpp;
+ /*
+ * First find fuid domain index in linked list
+ *
+ * If one isn't found then create an entry.
+ */
+
+ for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
+ fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
+ fuid_domain), fuididx++) {
+ if (idx == fuid_domain->z_domidx) {
+ found = B_TRUE;
+ break;
+ }
+ }
+
+ if (!found) {
+ fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
+ fuid_domain->z_domain = domain;
+ fuid_domain->z_domidx = idx;
+ list_insert_tail(&fuidp->z_domains, fuid_domain);
+ fuidp->z_domain_str_sz += strlen(domain) + 1;
+ fuidp->z_domain_cnt++;
+ }
+
+ if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
+ /*
+ * Now allocate fuid entry and add it on the end of the list
+ */
+
+ fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
+ fuid->z_id = id;
+ fuid->z_domidx = idx;
+ fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
+
+ list_insert_tail(&fuidp->z_fuids, fuid);
+ fuidp->z_fuid_cnt++;
+ } else {
+ if (type == ZFS_OWNER)
+ fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
+ else
+ fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
+ }
+}
+
+/*
+ * Create a file system FUID, based on information in the users cred
+ */
+uint64_t
+zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
+ dmu_tx_t *tx, cred_t *cr, zfs_fuid_info_t **fuidp)
+{
+ uint64_t idx;
+ ksid_t *ksid;
+ uint32_t rid;
+ char *kdomain;
+ const char *domain;
+ uid_t id;
+
+ VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
+
+ ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
+ if (ksid) {
+ id = ksid_getid(ksid);
+ } else {
+ if (type == ZFS_OWNER)
+ id = crgetuid(cr);
+ else
+ id = crgetgid(cr);
+ }
+
+ if (!zfsvfs->z_use_fuids || (!IS_EPHEMERAL(id)))
+ return ((uint64_t)id);
+
+ rid = ksid_getrid(ksid);
+ domain = ksid_getdomain(ksid);
+
+ idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+
+ zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
+
+ return (FUID_ENCODE(idx, rid));
+}
+
+/*
+ * Create a file system FUID for an ACL ace
+ * or a chown/chgrp of the file.
+ * This is similar to zfs_fuid_create_cred, except that
+ * we can't find the domain + rid information in the
+ * cred. Instead we have to query Winchester for the
+ * domain and rid.
+ *
+ * During replay operations the domain+rid information is
+ * found in the zfs_fuid_info_t that the replay code has
+ * attached to the zfsvfs of the file system.
+ */
+uint64_t
+zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
+ zfs_fuid_type_t type, dmu_tx_t *tx, zfs_fuid_info_t **fuidpp)
+{
+ const char *domain;
+ char *kdomain;
+ uint32_t fuid_idx = FUID_INDEX(id);
+ uint32_t rid;
+ idmap_stat status;
+ uint64_t idx;
+ boolean_t is_replay = (zfsvfs->z_assign >= TXG_INITIAL);
+ zfs_fuid_t *zfuid = NULL;
+ zfs_fuid_info_t *fuidp;
+
+ /*
+ * If POSIX ID, or entry is already a FUID then
+ * just return the id
+ *
+ * We may also be handed an already FUID'ized id via
+ * chmod.
+ */
+
+ if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
+ return (id);
+
+ if (is_replay) {
+ fuidp = zfsvfs->z_fuid_replay;
+
+ /*
+ * If we are passed an ephemeral id, but no
+ * fuid_info was logged then return NOBODY.
+ * This is most likely a result of idmap service
+ * not being available.
+ */
+ if (fuidp == NULL)
+ return (UID_NOBODY);
+
+ switch (type) {
+ case ZFS_ACE_USER:
+ case ZFS_ACE_GROUP:
+ zfuid = list_head(&fuidp->z_fuids);
+ rid = FUID_RID(zfuid->z_logfuid);
+ idx = FUID_INDEX(zfuid->z_logfuid);
+ break;
+ case ZFS_OWNER:
+ rid = FUID_RID(fuidp->z_fuid_owner);
+ idx = FUID_INDEX(fuidp->z_fuid_owner);
+ break;
+ case ZFS_GROUP:
+ rid = FUID_RID(fuidp->z_fuid_group);
+ idx = FUID_INDEX(fuidp->z_fuid_group);
+ break;
+ };
+ domain = fuidp->z_domain_table[idx -1];
+ } else {
+ if (type == ZFS_OWNER || type == ZFS_ACE_USER)
+ status = kidmap_getsidbyuid(crgetzone(cr), id,
+ &domain, &rid);
+ else
+ status = kidmap_getsidbygid(crgetzone(cr), id,
+ &domain, &rid);
+
+ if (status != 0) {
+ /*
+ * When returning nobody we will need to
+ * make a dummy fuid table entry for logging
+ * purposes.
+ */
+ rid = UID_NOBODY;
+ domain = nulldomain;
+ }
+ }
+
+ idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+
+ if (!is_replay)
+ zfs_fuid_node_add(fuidpp, kdomain, rid, idx, id, type);
+ else if (zfuid != NULL) {
+ list_remove(&fuidp->z_fuids, zfuid);
+ kmem_free(zfuid, sizeof (zfs_fuid_t));
+ }
+ return (FUID_ENCODE(idx, rid));
+}
+
+void
+zfs_fuid_destroy(zfsvfs_t *zfsvfs)
+{
+ rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
+ if (!zfsvfs->z_fuid_loaded) {
+ rw_exit(&zfsvfs->z_fuid_lock);
+ return;
+ }
+ zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
+ rw_exit(&zfsvfs->z_fuid_lock);
+}
+
+/*
+ * Allocate zfs_fuid_info for tracking FUIDs created during
+ * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
+ */
+zfs_fuid_info_t *
+zfs_fuid_info_alloc(void)
+{
+ zfs_fuid_info_t *fuidp;
+
+ fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
+ list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
+ offsetof(zfs_fuid_domain_t, z_next));
+ list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
+ offsetof(zfs_fuid_t, z_next));
+ return (fuidp);
+}
+
+/*
+ * Release all memory associated with zfs_fuid_info_t
+ */
+void
+zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
+{
+ zfs_fuid_t *zfuid;
+ zfs_fuid_domain_t *zdomain;
+
+ while ((zfuid = list_head(&fuidp->z_fuids)) != NULL) {
+ list_remove(&fuidp->z_fuids, zfuid);
+ kmem_free(zfuid, sizeof (zfs_fuid_t));
+ }
+
+ if (fuidp->z_domain_table != NULL)
+ kmem_free(fuidp->z_domain_table,
+ (sizeof (char **)) * fuidp->z_domain_cnt);
+
+ while ((zdomain = list_head(&fuidp->z_domains)) != NULL) {
+ list_remove(&fuidp->z_domains, zdomain);
+ kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
+ }
+
+ kmem_free(fuidp, sizeof (zfs_fuid_info_t));
+}
+
+/*
+ * Check to see if id is a groupmember. If cred
+ * has ksid info then sidlist is checked first
+ * and if still not found then POSIX groups are checked
+ *
+ * Will use a straight FUID compare when possible.
+ */
+boolean_t
+zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
+{
+ ksid_t *ksid = crgetsid(cr, KSID_GROUP);
+ uid_t gid;
+
+ if (ksid) {
+ int i;
+ ksid_t *ksid_groups;
+ ksidlist_t *ksidlist = crgetsidlist(cr);
+ uint32_t idx = FUID_INDEX(id);
+ uint32_t rid = FUID_RID(id);
+
+ ASSERT(ksidlist);
+ ksid_groups = ksidlist->ksl_sids;
+
+ for (i = 0; i != ksidlist->ksl_nsid; i++) {
+ if (idx == 0) {
+ if (id != IDMAP_WK_CREATOR_GROUP_GID &&
+ id == ksid_groups[i].ks_id) {
+ return (B_TRUE);
+ }
+ } else {
+ char *domain;
+
+ domain = zfs_fuid_find_by_idx(zfsvfs, idx);
+ ASSERT(domain != NULL);
+
+ if (strcmp(domain,
+ IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
+ return (B_FALSE);
+
+ if ((strcmp(domain,
+ ksid_groups[i].ks_domain->kd_name) == 0) &&
+ rid == ksid_groups[i].ks_rid)
+ return (B_TRUE);
+ }
+ }
+ }
+
+ /*
+ * Not found in ksidlist, check posix groups
+ */
+ gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
+ return (groupmember(gid, cr));
+}
+#endif
diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c
new file mode 100644
index 000000000..b6ad57451
--- /dev/null
+++ b/module/zfs/zfs_ioctl.c
@@ -0,0 +1,3175 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/errno.h>
+#include <sys/uio.h>
+#include <sys/buf.h>
+#include <sys/modctl.h>
+#include <sys/open.h>
+#include <sys/file.h>
+#include <sys/kmem.h>
+#include <sys/conf.h>
+#include <sys/cmn_err.h>
+#include <sys/stat.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zfs_znode.h>
+#include <sys/zap.h>
+#include <sys/spa.h>
+#include <sys/spa_impl.h>
+#include <sys/vdev.h>
+#include <sys/vdev_impl.h>
+#include <sys/dmu.h>
+#include <sys/dsl_dir.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_deleg.h>
+#include <sys/dmu_objset.h>
+#include <sys/ddi.h>
+#include <sys/sunddi.h>
+#include <sys/sunldi.h>
+#include <sys/policy.h>
+#include <sys/zone.h>
+#include <sys/nvpair.h>
+#include <sys/pathname.h>
+#include <sys/mount.h>
+#include <sys/sdt.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_ctldir.h>
+#include <sys/zfs_dir.h>
+#include <sys/zvol.h>
+#include <sharefs/share.h>
+#include <sys/dmu_objset.h>
+
+#include "zfs_namecheck.h"
+#include "zfs_prop.h"
+#include "zfs_deleg.h"
+
+extern struct modlfs zfs_modlfs;
+
+extern void zfs_init(void);
+extern void zfs_fini(void);
+
+ldi_ident_t zfs_li = NULL;
+dev_info_t *zfs_dip;
+
+typedef int zfs_ioc_func_t(zfs_cmd_t *);
+typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
+
+typedef struct zfs_ioc_vec {
+ zfs_ioc_func_t *zvec_func;
+ zfs_secpolicy_func_t *zvec_secpolicy;
+ enum {
+ NO_NAME,
+ POOL_NAME,
+ DATASET_NAME
+ } zvec_namecheck;
+ boolean_t zvec_his_log;
+} zfs_ioc_vec_t;
+
+static void clear_props(char *dataset, nvlist_t *props);
+static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
+ boolean_t *);
+int zfs_set_prop_nvlist(const char *, nvlist_t *);
+
+/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
+void
+__dprintf(const char *file, const char *func, int line, const char *fmt, ...)
+{
+ const char *newfile;
+ char buf[256];
+ va_list adx;
+
+ /*
+ * Get rid of annoying "../common/" prefix to filename.
+ */
+ newfile = strrchr(file, '/');
+ if (newfile != NULL) {
+ newfile = newfile + 1; /* Get rid of leading / */
+ } else {
+ newfile = file;
+ }
+
+ va_start(adx, fmt);
+ (void) vsnprintf(buf, sizeof (buf), fmt, adx);
+ va_end(adx);
+
+ /*
+ * To get this data, use the zfs-dprintf probe as so:
+ * dtrace -q -n 'zfs-dprintf \
+ * /stringof(arg0) == "dbuf.c"/ \
+ * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
+ * arg0 = file name
+ * arg1 = function name
+ * arg2 = line number
+ * arg3 = message
+ */
+ DTRACE_PROBE4(zfs__dprintf,
+ char *, newfile, char *, func, int, line, char *, buf);
+}
+
+static void
+history_str_free(char *buf)
+{
+ kmem_free(buf, HIS_MAX_RECORD_LEN);
+}
+
+static char *
+history_str_get(zfs_cmd_t *zc)
+{
+ char *buf;
+
+ if (zc->zc_history == NULL)
+ return (NULL);
+
+ buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
+ if (copyinstr((void *)(uintptr_t)zc->zc_history,
+ buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
+ history_str_free(buf);
+ return (NULL);
+ }
+
+ buf[HIS_MAX_RECORD_LEN -1] = '\0';
+
+ return (buf);
+}
+
+/*
+ * Check to see if the named dataset is currently defined as bootable
+ */
+static boolean_t
+zfs_is_bootfs(const char *name)
+{
+ spa_t *spa;
+ boolean_t ret = B_FALSE;
+
+ if (spa_open(name, &spa, FTAG) == 0) {
+ if (spa->spa_bootfs) {
+ objset_t *os;
+
+ if (dmu_objset_open(name, DMU_OST_ZFS,
+ DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
+ ret = (dmu_objset_id(os) == spa->spa_bootfs);
+ dmu_objset_close(os);
+ }
+ }
+ spa_close(spa, FTAG);
+ }
+ return (ret);
+}
+
+/*
+ * zfs_earlier_version
+ *
+ * Return non-zero if the spa version is less than requested version.
+ */
+static int
+zfs_earlier_version(const char *name, int version)
+{
+ spa_t *spa;
+
+ if (spa_open(name, &spa, FTAG) == 0) {
+ if (spa_version(spa) < version) {
+ spa_close(spa, FTAG);
+ return (1);
+ }
+ spa_close(spa, FTAG);
+ }
+ return (0);
+}
+
+/*
+ * zpl_earlier_version
+ *
+ * Return TRUE if the ZPL version is less than requested version.
+ */
+static boolean_t
+zpl_earlier_version(const char *name, int version)
+{
+ objset_t *os;
+ boolean_t rc = B_TRUE;
+
+ if (dmu_objset_open(name, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
+ uint64_t zplversion;
+
+ if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
+ rc = zplversion < version;
+ dmu_objset_close(os);
+ }
+ return (rc);
+}
+
+static void
+zfs_log_history(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ char *buf;
+
+ if ((buf = history_str_get(zc)) == NULL)
+ return;
+
+ if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
+ if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
+ (void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
+ spa_close(spa, FTAG);
+ }
+ history_str_free(buf);
+}
+
+/*
+ * Policy for top-level read operations (list pools). Requires no privileges,
+ * and can be used in the local zone, as there is no associated dataset.
+ */
+/* ARGSUSED */
+static int
+zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (0);
+}
+
+/*
+ * Policy for dataset read operations (list children, get statistics). Requires
+ * no privileges, but must be visible in the local zone.
+ */
+/* ARGSUSED */
+static int
+zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
+{
+ if (INGLOBALZONE(curproc) ||
+ zone_dataset_visible(zc->zc_name, NULL))
+ return (0);
+
+ return (ENOENT);
+}
+
+static int
+zfs_dozonecheck(const char *dataset, cred_t *cr)
+{
+ uint64_t zoned;
+ int writable = 1;
+
+ /*
+ * The dataset must be visible by this zone -- check this first
+ * so they don't see EPERM on something they shouldn't know about.
+ */
+ if (!INGLOBALZONE(curproc) &&
+ !zone_dataset_visible(dataset, &writable))
+ return (ENOENT);
+
+ if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
+ return (ENOENT);
+
+ if (INGLOBALZONE(curproc)) {
+ /*
+ * If the fs is zoned, only root can access it from the
+ * global zone.
+ */
+ if (secpolicy_zfs(cr) && zoned)
+ return (EPERM);
+ } else {
+ /*
+ * If we are in a local zone, the 'zoned' property must be set.
+ */
+ if (!zoned)
+ return (EPERM);
+
+ /* must be writable by this zone */
+ if (!writable)
+ return (EPERM);
+ }
+ return (0);
+}
+
+int
+zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
+{
+ int error;
+
+ error = zfs_dozonecheck(name, cr);
+ if (error == 0) {
+ error = secpolicy_zfs(cr);
+ if (error)
+ error = dsl_deleg_access(name, perm, cr);
+ }
+ return (error);
+}
+
+static int
+zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
+{
+ /*
+ * Check permissions for special properties.
+ */
+ switch (prop) {
+ case ZFS_PROP_ZONED:
+ /*
+ * Disallow setting of 'zoned' from within a local zone.
+ */
+ if (!INGLOBALZONE(curproc))
+ return (EPERM);
+ break;
+
+ case ZFS_PROP_QUOTA:
+ if (!INGLOBALZONE(curproc)) {
+ uint64_t zoned;
+ char setpoint[MAXNAMELEN];
+ /*
+ * Unprivileged users are allowed to modify the
+ * quota on things *under* (ie. contained by)
+ * the thing they own.
+ */
+ if (dsl_prop_get_integer(name, "zoned", &zoned,
+ setpoint))
+ return (EPERM);
+ if (!zoned || strlen(name) <= strlen(setpoint))
+ return (EPERM);
+ }
+ break;
+ }
+
+ return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
+}
+
+int
+zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
+{
+ int error;
+
+ error = zfs_dozonecheck(zc->zc_name, cr);
+ if (error)
+ return (error);
+
+ /*
+ * permission to set permissions will be evaluated later in
+ * dsl_deleg_can_allow()
+ */
+ return (0);
+}
+
+int
+zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
+{
+ int error;
+ error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_ROLLBACK, cr);
+ if (error == 0)
+ error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_MOUNT, cr);
+ return (error);
+}
+
+int
+zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_SEND, cr));
+}
+
+int
+zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
+{
+ if (!INGLOBALZONE(curproc))
+ return (EPERM);
+
+ if (secpolicy_nfs(cr) == 0) {
+ return (0);
+ } else {
+ vnode_t *vp;
+ int error;
+
+ if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
+ NO_FOLLOW, NULL, &vp)) != 0)
+ return (error);
+
+ /* Now make sure mntpnt and dataset are ZFS */
+
+ if (vp->v_vfsp->vfs_fstype != zfsfstype ||
+ (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
+ zc->zc_name) != 0)) {
+ VN_RELE(vp);
+ return (EPERM);
+ }
+
+ VN_RELE(vp);
+ return (dsl_deleg_access(zc->zc_name,
+ ZFS_DELEG_PERM_SHARE, cr));
+ }
+}
+
+static int
+zfs_get_parent(const char *datasetname, char *parent, int parentsize)
+{
+ char *cp;
+
+ /*
+ * Remove the @bla or /bla from the end of the name to get the parent.
+ */
+ (void) strncpy(parent, datasetname, parentsize);
+ cp = strrchr(parent, '@');
+ if (cp != NULL) {
+ cp[0] = '\0';
+ } else {
+ cp = strrchr(parent, '/');
+ if (cp == NULL)
+ return (ENOENT);
+ cp[0] = '\0';
+ }
+
+ return (0);
+}
+
+int
+zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
+{
+ int error;
+
+ if ((error = zfs_secpolicy_write_perms(name,
+ ZFS_DELEG_PERM_MOUNT, cr)) != 0)
+ return (error);
+
+ return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
+}
+
+static int
+zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
+}
+
+/*
+ * Must have sys_config privilege to check the iscsi permission
+ */
+/* ARGSUSED */
+static int
+zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (secpolicy_zfs(cr));
+}
+
+int
+zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
+{
+ char parentname[MAXNAMELEN];
+ int error;
+
+ if ((error = zfs_secpolicy_write_perms(from,
+ ZFS_DELEG_PERM_RENAME, cr)) != 0)
+ return (error);
+
+ if ((error = zfs_secpolicy_write_perms(from,
+ ZFS_DELEG_PERM_MOUNT, cr)) != 0)
+ return (error);
+
+ if ((error = zfs_get_parent(to, parentname,
+ sizeof (parentname))) != 0)
+ return (error);
+
+ if ((error = zfs_secpolicy_write_perms(parentname,
+ ZFS_DELEG_PERM_CREATE, cr)) != 0)
+ return (error);
+
+ if ((error = zfs_secpolicy_write_perms(parentname,
+ ZFS_DELEG_PERM_MOUNT, cr)) != 0)
+ return (error);
+
+ return (error);
+}
+
+static int
+zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
+}
+
+static int
+zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
+{
+ char parentname[MAXNAMELEN];
+ objset_t *clone;
+ int error;
+
+ error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_PROMOTE, cr);
+ if (error)
+ return (error);
+
+ error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &clone);
+
+ if (error == 0) {
+ dsl_dataset_t *pclone = NULL;
+ dsl_dir_t *dd;
+ dd = clone->os->os_dsl_dataset->ds_dir;
+
+ rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
+ error = dsl_dataset_hold_obj(dd->dd_pool,
+ dd->dd_phys->dd_origin_obj, FTAG, &pclone);
+ rw_exit(&dd->dd_pool->dp_config_rwlock);
+ if (error) {
+ dmu_objset_close(clone);
+ return (error);
+ }
+
+ error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_MOUNT, cr);
+
+ dsl_dataset_name(pclone, parentname);
+ dmu_objset_close(clone);
+ dsl_dataset_rele(pclone, FTAG);
+ if (error == 0)
+ error = zfs_secpolicy_write_perms(parentname,
+ ZFS_DELEG_PERM_PROMOTE, cr);
+ }
+ return (error);
+}
+
+static int
+zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
+{
+ int error;
+
+ if ((error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
+ return (error);
+
+ if ((error = zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_MOUNT, cr)) != 0)
+ return (error);
+
+ return (zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_CREATE, cr));
+}
+
+int
+zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
+{
+ int error;
+
+ if ((error = zfs_secpolicy_write_perms(name,
+ ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
+ return (error);
+
+ error = zfs_secpolicy_write_perms(name,
+ ZFS_DELEG_PERM_MOUNT, cr);
+
+ return (error);
+}
+
+static int
+zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
+{
+
+ return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
+}
+
+static int
+zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
+{
+ char parentname[MAXNAMELEN];
+ int error;
+
+ if ((error = zfs_get_parent(zc->zc_name, parentname,
+ sizeof (parentname))) != 0)
+ return (error);
+
+ if (zc->zc_value[0] != '\0') {
+ if ((error = zfs_secpolicy_write_perms(zc->zc_value,
+ ZFS_DELEG_PERM_CLONE, cr)) != 0)
+ return (error);
+ }
+
+ if ((error = zfs_secpolicy_write_perms(parentname,
+ ZFS_DELEG_PERM_CREATE, cr)) != 0)
+ return (error);
+
+ error = zfs_secpolicy_write_perms(parentname,
+ ZFS_DELEG_PERM_MOUNT, cr);
+
+ return (error);
+}
+
+static int
+zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
+{
+ int error;
+
+ error = secpolicy_fs_unmount(cr, NULL);
+ if (error) {
+ error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
+ }
+ return (error);
+}
+
+/*
+ * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
+ * SYS_CONFIG privilege, which is not available in a local zone.
+ */
+/* ARGSUSED */
+static int
+zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
+{
+ if (secpolicy_sys_config(cr, B_FALSE) != 0)
+ return (EPERM);
+
+ return (0);
+}
+
+/*
+ * Just like zfs_secpolicy_config, except that we will check for
+ * mount permission on the dataset for permission to create/remove
+ * the minor nodes.
+ */
+static int
+zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
+{
+ if (secpolicy_sys_config(cr, B_FALSE) != 0) {
+ return (dsl_deleg_access(zc->zc_name,
+ ZFS_DELEG_PERM_MOUNT, cr));
+ }
+
+ return (0);
+}
+
+/*
+ * Policy for fault injection. Requires all privileges.
+ */
+/* ARGSUSED */
+static int
+zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
+{
+ return (secpolicy_zinject(cr));
+}
+
+static int
+zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
+{
+ zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
+
+ if (prop == ZPROP_INVAL) {
+ if (!zfs_prop_user(zc->zc_value))
+ return (EINVAL);
+ return (zfs_secpolicy_write_perms(zc->zc_name,
+ ZFS_DELEG_PERM_USERPROP, cr));
+ } else {
+ if (!zfs_prop_inheritable(prop))
+ return (EINVAL);
+ return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
+ }
+}
+
+/*
+ * Returns the nvlist as specified by the user in the zfs_cmd_t.
+ */
+static int
+get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
+{
+ char *packed;
+ int error;
+ nvlist_t *list = NULL;
+
+ /*
+ * Read in and unpack the user-supplied nvlist.
+ */
+ if (size == 0)
+ return (EINVAL);
+
+ packed = kmem_alloc(size, KM_SLEEP);
+
+ if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
+ kmem_free(packed, size);
+ return (error);
+ }
+
+ if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
+ kmem_free(packed, size);
+ return (error);
+ }
+
+ kmem_free(packed, size);
+
+ *nvp = list;
+ return (0);
+}
+
+static int
+put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
+{
+ char *packed = NULL;
+ size_t size;
+ int error;
+
+ VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
+
+ if (size > zc->zc_nvlist_dst_size) {
+ error = ENOMEM;
+ } else {
+ packed = kmem_alloc(size, KM_SLEEP);
+ VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
+ KM_SLEEP) == 0);
+ error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
+ size);
+ kmem_free(packed, size);
+ }
+
+ zc->zc_nvlist_dst_size = size;
+ return (error);
+}
+
+static int
+zfs_ioc_pool_create(zfs_cmd_t *zc)
+{
+ int error;
+ nvlist_t *config, *props = NULL;
+ nvlist_t *rootprops = NULL;
+ nvlist_t *zplprops = NULL;
+ char *buf;
+
+ if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
+ &config))
+ return (error);
+
+ if (zc->zc_nvlist_src_size != 0 && (error =
+ get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
+ nvlist_free(config);
+ return (error);
+ }
+
+ if (props) {
+ nvlist_t *nvl = NULL;
+ uint64_t version = SPA_VERSION;
+
+ (void) nvlist_lookup_uint64(props,
+ zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
+ if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
+ error = EINVAL;
+ goto pool_props_bad;
+ }
+ (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
+ if (nvl) {
+ error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
+ if (error != 0) {
+ nvlist_free(config);
+ nvlist_free(props);
+ return (error);
+ }
+ (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
+ }
+ VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ error = zfs_fill_zplprops_root(version, rootprops,
+ zplprops, NULL);
+ if (error)
+ goto pool_props_bad;
+ }
+
+ buf = history_str_get(zc);
+
+ error = spa_create(zc->zc_name, config, props, buf, zplprops);
+
+ /*
+ * Set the remaining root properties
+ */
+ if (!error &&
+ (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
+ (void) spa_destroy(zc->zc_name);
+
+ if (buf != NULL)
+ history_str_free(buf);
+
+pool_props_bad:
+ nvlist_free(rootprops);
+ nvlist_free(zplprops);
+ nvlist_free(config);
+ nvlist_free(props);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_destroy(zfs_cmd_t *zc)
+{
+ int error;
+ zfs_log_history(zc);
+ error = spa_destroy(zc->zc_name);
+ return (error);
+}
+
+static int
+zfs_ioc_pool_import(zfs_cmd_t *zc)
+{
+ int error;
+ nvlist_t *config, *props = NULL;
+ uint64_t guid;
+
+ if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
+ &config)) != 0)
+ return (error);
+
+ if (zc->zc_nvlist_src_size != 0 && (error =
+ get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
+ nvlist_free(config);
+ return (error);
+ }
+
+ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
+ guid != zc->zc_guid)
+ error = EINVAL;
+ else if (zc->zc_cookie)
+ error = spa_import_faulted(zc->zc_name, config,
+ props);
+ else
+ error = spa_import(zc->zc_name, config, props);
+
+ nvlist_free(config);
+
+ if (props)
+ nvlist_free(props);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_export(zfs_cmd_t *zc)
+{
+ int error;
+ boolean_t force = (boolean_t)zc->zc_cookie;
+
+ zfs_log_history(zc);
+ error = spa_export(zc->zc_name, NULL, force);
+ return (error);
+}
+
+static int
+zfs_ioc_pool_configs(zfs_cmd_t *zc)
+{
+ nvlist_t *configs;
+ int error;
+
+ if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
+ return (EEXIST);
+
+ error = put_nvlist(zc, configs);
+
+ nvlist_free(configs);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_stats(zfs_cmd_t *zc)
+{
+ nvlist_t *config;
+ int error;
+ int ret = 0;
+
+ error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
+ sizeof (zc->zc_value));
+
+ if (config != NULL) {
+ ret = put_nvlist(zc, config);
+ nvlist_free(config);
+
+ /*
+ * The config may be present even if 'error' is non-zero.
+ * In this case we return success, and preserve the real errno
+ * in 'zc_cookie'.
+ */
+ zc->zc_cookie = error;
+ } else {
+ ret = error;
+ }
+
+ return (ret);
+}
+
+/*
+ * Try to import the given pool, returning pool stats as appropriate so that
+ * user land knows which devices are available and overall pool health.
+ */
+static int
+zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
+{
+ nvlist_t *tryconfig, *config;
+ int error;
+
+ if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
+ &tryconfig)) != 0)
+ return (error);
+
+ config = spa_tryimport(tryconfig);
+
+ nvlist_free(tryconfig);
+
+ if (config == NULL)
+ return (EINVAL);
+
+ error = put_nvlist(zc, config);
+ nvlist_free(config);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_scrub(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ error = spa_scrub(spa, zc->zc_cookie);
+
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_freeze(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+
+ error = spa_open(zc->zc_name, &spa, FTAG);
+ if (error == 0) {
+ spa_freeze(spa);
+ spa_close(spa, FTAG);
+ }
+ return (error);
+}
+
+static int
+zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
+ spa_close(spa, FTAG);
+ return (EINVAL);
+ }
+
+ spa_upgrade(spa, zc->zc_cookie);
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_get_history(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ char *hist_buf;
+ uint64_t size;
+ int error;
+
+ if ((size = zc->zc_history_len) == 0)
+ return (EINVAL);
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
+ spa_close(spa, FTAG);
+ return (ENOTSUP);
+ }
+
+ hist_buf = kmem_alloc(size, KM_SLEEP);
+ if ((error = spa_history_get(spa, &zc->zc_history_offset,
+ &zc->zc_history_len, hist_buf)) == 0) {
+ error = xcopyout(hist_buf,
+ (char *)(uintptr_t)zc->zc_history,
+ zc->zc_history_len);
+ }
+
+ spa_close(spa, FTAG);
+ kmem_free(hist_buf, size);
+ return (error);
+}
+
+static int
+zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
+{
+ int error;
+
+ if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
+ return (error);
+
+ return (0);
+}
+
+static int
+zfs_ioc_obj_to_path(zfs_cmd_t *zc)
+{
+ objset_t *osp;
+ int error;
+
+ if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
+ DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
+ return (error);
+ error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
+ sizeof (zc->zc_value));
+ dmu_objset_close(osp);
+
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_add(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+ nvlist_t *config, **l2cache, **spares;
+ uint_t nl2cache = 0, nspares = 0;
+
+ error = spa_open(zc->zc_name, &spa, FTAG);
+ if (error != 0)
+ return (error);
+
+ error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
+ &config);
+ (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
+ &l2cache, &nl2cache);
+
+ (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
+ &spares, &nspares);
+
+ /*
+ * A root pool with concatenated devices is not supported.
+ * Thus, can not add a device to a root pool.
+ *
+ * Intent log device can not be added to a rootpool because
+ * during mountroot, zil is replayed, a seperated log device
+ * can not be accessed during the mountroot time.
+ *
+ * l2cache and spare devices are ok to be added to a rootpool.
+ */
+ if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
+ spa_close(spa, FTAG);
+ return (EDOM);
+ }
+
+ if (error == 0) {
+ error = spa_vdev_add(spa, config);
+ nvlist_free(config);
+ }
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_remove(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+
+ error = spa_open(zc->zc_name, &spa, FTAG);
+ if (error != 0)
+ return (error);
+ error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+ vdev_state_t newstate = VDEV_STATE_UNKNOWN;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+ switch (zc->zc_cookie) {
+ case VDEV_STATE_ONLINE:
+ error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
+ break;
+
+ case VDEV_STATE_OFFLINE:
+ error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
+ break;
+
+ case VDEV_STATE_FAULTED:
+ error = vdev_fault(spa, zc->zc_guid);
+ break;
+
+ case VDEV_STATE_DEGRADED:
+ error = vdev_degrade(spa, zc->zc_guid);
+ break;
+
+ default:
+ error = EINVAL;
+ }
+ zc->zc_cookie = newstate;
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_attach(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int replacing = zc->zc_cookie;
+ nvlist_t *config;
+ int error;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
+ &config)) == 0) {
+ error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
+ nvlist_free(config);
+ }
+
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_detach(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
+
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+static int
+zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ char *path = zc->zc_value;
+ uint64_t guid = zc->zc_guid;
+ int error;
+
+ error = spa_open(zc->zc_name, &spa, FTAG);
+ if (error != 0)
+ return (error);
+
+ error = spa_vdev_setpath(spa, guid, path);
+ spa_close(spa, FTAG);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_nvlist_dst_size size of buffer for property nvlist
+ *
+ * outputs:
+ * zc_objset_stats stats
+ * zc_nvlist_dst property nvlist
+ * zc_nvlist_dst_size size of property nvlist
+ */
+static int
+zfs_ioc_objset_stats(zfs_cmd_t *zc)
+{
+ objset_t *os = NULL;
+ int error;
+ nvlist_t *nv;
+
+ if (error = dmu_objset_open(zc->zc_name,
+ DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
+ return (error);
+
+ dmu_objset_fast_stat(os, &zc->zc_objset_stats);
+
+ if (zc->zc_nvlist_dst != 0 &&
+ (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
+ dmu_objset_stats(os, nv);
+ /*
+ * NB: zvol_get_stats() will read the objset contents,
+ * which we aren't supposed to do with a
+ * DS_MODE_USER hold, because it could be
+ * inconsistent. So this is a bit of a workaround...
+ */
+ if (!zc->zc_objset_stats.dds_inconsistent) {
+ if (dmu_objset_type(os) == DMU_OST_ZVOL)
+ VERIFY(zvol_get_stats(os, nv) == 0);
+ }
+ error = put_nvlist(zc, nv);
+ nvlist_free(nv);
+ }
+
+ dmu_objset_close(os);
+ return (error);
+}
+
+static int
+nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
+{
+ uint64_t value;
+ int error;
+
+ /*
+ * zfs_get_zplprop() will either find a value or give us
+ * the default value (if there is one).
+ */
+ if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
+ return (error);
+ VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
+ return (0);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_nvlist_dst_size size of buffer for zpl property nvlist
+ *
+ * outputs:
+ * zc_nvlist_dst zpl property nvlist
+ * zc_nvlist_dst_size size of zpl property nvlist
+ */
+static int
+zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
+{
+ objset_t *os;
+ int err;
+
+ if (err = dmu_objset_open(zc->zc_name,
+ DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
+ return (err);
+
+ dmu_objset_fast_stat(os, &zc->zc_objset_stats);
+
+ /*
+ * NB: nvl_add_zplprop() will read the objset contents,
+ * which we aren't supposed to do with a DS_MODE_USER
+ * hold, because it could be inconsistent.
+ */
+ if (zc->zc_nvlist_dst != NULL &&
+ !zc->zc_objset_stats.dds_inconsistent &&
+ dmu_objset_type(os) == DMU_OST_ZFS) {
+ nvlist_t *nv;
+
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
+ (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
+ (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
+ (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
+ err = put_nvlist(zc, nv);
+ nvlist_free(nv);
+ } else {
+ err = ENOENT;
+ }
+ dmu_objset_close(os);
+ return (err);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_cookie zap cursor
+ * zc_nvlist_dst_size size of buffer for property nvlist
+ *
+ * outputs:
+ * zc_name name of next filesystem
+ * zc_objset_stats stats
+ * zc_nvlist_dst property nvlist
+ * zc_nvlist_dst_size size of property nvlist
+ */
+static int
+zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
+{
+ objset_t *os;
+ int error;
+ char *p;
+
+ if (error = dmu_objset_open(zc->zc_name,
+ DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
+ if (error == ENOENT)
+ error = ESRCH;
+ return (error);
+ }
+
+ p = strrchr(zc->zc_name, '/');
+ if (p == NULL || p[1] != '\0')
+ (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
+ p = zc->zc_name + strlen(zc->zc_name);
+
+ do {
+ error = dmu_dir_list_next(os,
+ sizeof (zc->zc_name) - (p - zc->zc_name), p,
+ NULL, &zc->zc_cookie);
+ if (error == ENOENT)
+ error = ESRCH;
+ } while (error == 0 && !INGLOBALZONE(curproc) &&
+ !zone_dataset_visible(zc->zc_name, NULL));
+ dmu_objset_close(os);
+
+ /*
+ * If it's a hidden dataset (ie. with a '$' in its name), don't
+ * try to get stats for it. Userland will skip over it.
+ */
+ if (error == 0 && strchr(zc->zc_name, '$') == NULL)
+ error = zfs_ioc_objset_stats(zc); /* fill in the stats */
+
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_cookie zap cursor
+ * zc_nvlist_dst_size size of buffer for property nvlist
+ *
+ * outputs:
+ * zc_name name of next snapshot
+ * zc_objset_stats stats
+ * zc_nvlist_dst property nvlist
+ * zc_nvlist_dst_size size of property nvlist
+ */
+static int
+zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
+{
+ objset_t *os;
+ int error;
+
+ error = dmu_objset_open(zc->zc_name,
+ DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
+ if (error)
+ return (error == ENOENT ? ESRCH : error);
+
+ /*
+ * A dataset name of maximum length cannot have any snapshots,
+ * so exit immediately.
+ */
+ if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
+ dmu_objset_close(os);
+ return (ESRCH);
+ }
+
+ error = dmu_snapshot_list_next(os,
+ sizeof (zc->zc_name) - strlen(zc->zc_name),
+ zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
+ dmu_objset_close(os);
+ if (error == 0)
+ error = zfs_ioc_objset_stats(zc); /* fill in the stats */
+ else if (error == ENOENT)
+ error = ESRCH;
+
+ /* if we failed, undo the @ that we tacked on to zc_name */
+ if (error)
+ *strchr(zc->zc_name, '@') = '\0';
+ return (error);
+}
+
+int
+zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
+{
+ nvpair_t *elem;
+ int error;
+ uint64_t intval;
+ char *strval;
+
+ /*
+ * First validate permission to set all of the properties
+ */
+ elem = NULL;
+ while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
+ const char *propname = nvpair_name(elem);
+ zfs_prop_t prop = zfs_name_to_prop(propname);
+
+ if (prop == ZPROP_INVAL) {
+ /*
+ * If this is a user-defined property, it must be a
+ * string, and there is no further validation to do.
+ */
+ if (!zfs_prop_user(propname) ||
+ nvpair_type(elem) != DATA_TYPE_STRING)
+ return (EINVAL);
+
+ if (error = zfs_secpolicy_write_perms(name,
+ ZFS_DELEG_PERM_USERPROP, CRED()))
+ return (error);
+ continue;
+ }
+
+ if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
+ return (error);
+
+ /*
+ * Check that this value is valid for this pool version
+ */
+ switch (prop) {
+ case ZFS_PROP_COMPRESSION:
+ /*
+ * If the user specified gzip compression, make sure
+ * the SPA supports it. We ignore any errors here since
+ * we'll catch them later.
+ */
+ if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
+ nvpair_value_uint64(elem, &intval) == 0) {
+ if (intval >= ZIO_COMPRESS_GZIP_1 &&
+ intval <= ZIO_COMPRESS_GZIP_9 &&
+ zfs_earlier_version(name,
+ SPA_VERSION_GZIP_COMPRESSION))
+ return (ENOTSUP);
+
+ /*
+ * If this is a bootable dataset then
+ * verify that the compression algorithm
+ * is supported for booting. We must return
+ * something other than ENOTSUP since it
+ * implies a downrev pool version.
+ */
+ if (zfs_is_bootfs(name) &&
+ !BOOTFS_COMPRESS_VALID(intval))
+ return (ERANGE);
+ }
+ break;
+
+ case ZFS_PROP_COPIES:
+ if (zfs_earlier_version(name,
+ SPA_VERSION_DITTO_BLOCKS))
+ return (ENOTSUP);
+ break;
+
+ case ZFS_PROP_SHARESMB:
+ if (zpl_earlier_version(name, ZPL_VERSION_FUID))
+ return (ENOTSUP);
+ break;
+
+ case ZFS_PROP_ACLINHERIT:
+ if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
+ nvpair_value_uint64(elem, &intval) == 0)
+ if (intval == ZFS_ACL_PASSTHROUGH_X &&
+ zfs_earlier_version(name,
+ SPA_VERSION_PASSTHROUGH_X))
+ return (ENOTSUP);
+ }
+ }
+
+ elem = NULL;
+ while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
+ const char *propname = nvpair_name(elem);
+ zfs_prop_t prop = zfs_name_to_prop(propname);
+
+ if (prop == ZPROP_INVAL) {
+ VERIFY(nvpair_value_string(elem, &strval) == 0);
+ error = dsl_prop_set(name, propname, 1,
+ strlen(strval) + 1, strval);
+ if (error == 0)
+ continue;
+ else
+ return (error);
+ }
+
+ switch (prop) {
+ case ZFS_PROP_QUOTA:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = dsl_dir_set_quota(name, intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_REFQUOTA:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = dsl_dataset_set_quota(name, intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_RESERVATION:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = dsl_dir_set_reservation(name,
+ intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_REFRESERVATION:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = dsl_dataset_set_reservation(name,
+ intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_VOLSIZE:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = zvol_set_volsize(name,
+ ddi_driver_major(zfs_dip), intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_VOLBLOCKSIZE:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = zvol_set_volblocksize(name, intval)) != 0)
+ return (error);
+ break;
+
+ case ZFS_PROP_VERSION:
+ if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
+ (error = zfs_set_version(name, intval)) != 0)
+ return (error);
+ break;
+
+ default:
+ if (nvpair_type(elem) == DATA_TYPE_STRING) {
+ if (zfs_prop_get_type(prop) !=
+ PROP_TYPE_STRING)
+ return (EINVAL);
+ VERIFY(nvpair_value_string(elem, &strval) == 0);
+ if ((error = dsl_prop_set(name,
+ nvpair_name(elem), 1, strlen(strval) + 1,
+ strval)) != 0)
+ return (error);
+ } else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
+ const char *unused;
+
+ VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+
+ switch (zfs_prop_get_type(prop)) {
+ case PROP_TYPE_NUMBER:
+ break;
+ case PROP_TYPE_STRING:
+ return (EINVAL);
+ case PROP_TYPE_INDEX:
+ if (zfs_prop_index_to_string(prop,
+ intval, &unused) != 0)
+ return (EINVAL);
+ break;
+ default:
+ cmn_err(CE_PANIC,
+ "unknown property type");
+ break;
+ }
+
+ if ((error = dsl_prop_set(name, propname,
+ 8, 1, &intval)) != 0)
+ return (error);
+ } else {
+ return (EINVAL);
+ }
+ break;
+ }
+ }
+
+ return (0);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_value name of property to inherit
+ * zc_nvlist_src{_size} nvlist of properties to apply
+ * zc_cookie clear existing local props?
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_set_prop(zfs_cmd_t *zc)
+{
+ nvlist_t *nvl;
+ int error;
+
+ if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &nvl)) != 0)
+ return (error);
+
+ if (zc->zc_cookie) {
+ nvlist_t *origprops;
+ objset_t *os;
+
+ if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
+ if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
+ clear_props(zc->zc_name, origprops);
+ nvlist_free(origprops);
+ }
+ dmu_objset_close(os);
+ }
+
+ }
+
+ error = zfs_set_prop_nvlist(zc->zc_name, nvl);
+
+ nvlist_free(nvl);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_value name of property to inherit
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_inherit_prop(zfs_cmd_t *zc)
+{
+ /* the property name has been validated by zfs_secpolicy_inherit() */
+ return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
+}
+
+static int
+zfs_ioc_pool_set_props(zfs_cmd_t *zc)
+{
+ nvlist_t *props;
+ spa_t *spa;
+ int error;
+
+ if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &props)))
+ return (error);
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
+ nvlist_free(props);
+ return (error);
+ }
+
+ error = spa_prop_set(spa, props);
+
+ nvlist_free(props);
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+static int
+zfs_ioc_pool_get_props(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+ nvlist_t *nvp = NULL;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ error = spa_prop_get(spa, &nvp);
+
+ if (error == 0 && zc->zc_nvlist_dst != NULL)
+ error = put_nvlist(zc, nvp);
+ else
+ error = EFAULT;
+
+ spa_close(spa, FTAG);
+
+ if (nvp)
+ nvlist_free(nvp);
+ return (error);
+}
+
+static int
+zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
+{
+ nvlist_t *nvp;
+ int error;
+ uint32_t uid;
+ uint32_t gid;
+ uint32_t *groups;
+ uint_t group_cnt;
+ cred_t *usercred;
+
+ if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &nvp)) != 0) {
+ return (error);
+ }
+
+ if ((error = nvlist_lookup_uint32(nvp,
+ ZFS_DELEG_PERM_UID, &uid)) != 0) {
+ nvlist_free(nvp);
+ return (EPERM);
+ }
+
+ if ((error = nvlist_lookup_uint32(nvp,
+ ZFS_DELEG_PERM_GID, &gid)) != 0) {
+ nvlist_free(nvp);
+ return (EPERM);
+ }
+
+ if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
+ &groups, &group_cnt)) != 0) {
+ nvlist_free(nvp);
+ return (EPERM);
+ }
+ usercred = cralloc();
+ if ((crsetugid(usercred, uid, gid) != 0) ||
+ (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
+ nvlist_free(nvp);
+ crfree(usercred);
+ return (EPERM);
+ }
+ nvlist_free(nvp);
+ error = dsl_deleg_access(zc->zc_name,
+ zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
+ crfree(usercred);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_nvlist_src{_size} nvlist of delegated permissions
+ * zc_perm_action allow/unallow flag
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_set_fsacl(zfs_cmd_t *zc)
+{
+ int error;
+ nvlist_t *fsaclnv = NULL;
+
+ if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &fsaclnv)) != 0)
+ return (error);
+
+ /*
+ * Verify nvlist is constructed correctly
+ */
+ if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
+ nvlist_free(fsaclnv);
+ return (EINVAL);
+ }
+
+ /*
+ * If we don't have PRIV_SYS_MOUNT, then validate
+ * that user is allowed to hand out each permission in
+ * the nvlist(s)
+ */
+
+ error = secpolicy_zfs(CRED());
+ if (error) {
+ if (zc->zc_perm_action == B_FALSE) {
+ error = dsl_deleg_can_allow(zc->zc_name,
+ fsaclnv, CRED());
+ } else {
+ error = dsl_deleg_can_unallow(zc->zc_name,
+ fsaclnv, CRED());
+ }
+ }
+
+ if (error == 0)
+ error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
+
+ nvlist_free(fsaclnv);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ *
+ * outputs:
+ * zc_nvlist_src{_size} nvlist of delegated permissions
+ */
+static int
+zfs_ioc_get_fsacl(zfs_cmd_t *zc)
+{
+ nvlist_t *nvp;
+ int error;
+
+ if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
+ error = put_nvlist(zc, nvp);
+ nvlist_free(nvp);
+ }
+
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of volume
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_create_minor(zfs_cmd_t *zc)
+{
+ return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
+}
+
+/*
+ * inputs:
+ * zc_name name of volume
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_remove_minor(zfs_cmd_t *zc)
+{
+ return (zvol_remove_minor(zc->zc_name));
+}
+
+/*
+ * Search the vfs list for a specified resource. Returns a pointer to it
+ * or NULL if no suitable entry is found. The caller of this routine
+ * is responsible for releasing the returned vfs pointer.
+ */
+static vfs_t *
+zfs_get_vfs(const char *resource)
+{
+ struct vfs *vfsp;
+ struct vfs *vfs_found = NULL;
+
+ vfs_list_read_lock();
+ vfsp = rootvfs;
+ do {
+ if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
+ VFS_HOLD(vfsp);
+ vfs_found = vfsp;
+ break;
+ }
+ vfsp = vfsp->vfs_next;
+ } while (vfsp != rootvfs);
+ vfs_list_unlock();
+ return (vfs_found);
+}
+
+/* ARGSUSED */
+static void
+zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
+{
+ zfs_creat_t *zct = arg;
+
+ zfs_create_fs(os, cr, zct->zct_zplprops, tx);
+}
+
+#define ZFS_PROP_UNDEFINED ((uint64_t)-1)
+
+/*
+ * inputs:
+ * createprops list of properties requested by creator
+ * default_zplver zpl version to use if unspecified in createprops
+ * fuids_ok fuids allowed in this version of the spa?
+ * os parent objset pointer (NULL if root fs)
+ *
+ * outputs:
+ * zplprops values for the zplprops we attach to the master node object
+ * is_ci true if requested file system will be purely case-insensitive
+ *
+ * Determine the settings for utf8only, normalization and
+ * casesensitivity. Specific values may have been requested by the
+ * creator and/or we can inherit values from the parent dataset. If
+ * the file system is of too early a vintage, a creator can not
+ * request settings for these properties, even if the requested
+ * setting is the default value. We don't actually want to create dsl
+ * properties for these, so remove them from the source nvlist after
+ * processing.
+ */
+static int
+zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
+ boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
+ boolean_t *is_ci)
+{
+ uint64_t zplver = default_zplver;
+ uint64_t sense = ZFS_PROP_UNDEFINED;
+ uint64_t norm = ZFS_PROP_UNDEFINED;
+ uint64_t u8 = ZFS_PROP_UNDEFINED;
+
+ ASSERT(zplprops != NULL);
+
+ /*
+ * Pull out creator prop choices, if any.
+ */
+ if (createprops) {
+ (void) nvlist_lookup_uint64(createprops,
+ zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
+ (void) nvlist_lookup_uint64(createprops,
+ zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
+ (void) nvlist_remove_all(createprops,
+ zfs_prop_to_name(ZFS_PROP_NORMALIZE));
+ (void) nvlist_lookup_uint64(createprops,
+ zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
+ (void) nvlist_remove_all(createprops,
+ zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
+ (void) nvlist_lookup_uint64(createprops,
+ zfs_prop_to_name(ZFS_PROP_CASE), &sense);
+ (void) nvlist_remove_all(createprops,
+ zfs_prop_to_name(ZFS_PROP_CASE));
+ }
+
+ /*
+ * If the zpl version requested is whacky or the file system
+ * or pool is version is too "young" to support normalization
+ * and the creator tried to set a value for one of the props,
+ * error out.
+ */
+ if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
+ (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
+ (zplver < ZPL_VERSION_NORMALIZATION &&
+ (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
+ sense != ZFS_PROP_UNDEFINED)))
+ return (ENOTSUP);
+
+ /*
+ * Put the version in the zplprops
+ */
+ VERIFY(nvlist_add_uint64(zplprops,
+ zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
+
+ if (norm == ZFS_PROP_UNDEFINED)
+ VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
+ VERIFY(nvlist_add_uint64(zplprops,
+ zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
+
+ /*
+ * If we're normalizing, names must always be valid UTF-8 strings.
+ */
+ if (norm)
+ u8 = 1;
+ if (u8 == ZFS_PROP_UNDEFINED)
+ VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
+ VERIFY(nvlist_add_uint64(zplprops,
+ zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
+
+ if (sense == ZFS_PROP_UNDEFINED)
+ VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
+ VERIFY(nvlist_add_uint64(zplprops,
+ zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
+
+ if (is_ci)
+ *is_ci = (sense == ZFS_CASE_INSENSITIVE);
+
+ return (0);
+}
+
+static int
+zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
+ nvlist_t *zplprops, boolean_t *is_ci)
+{
+ boolean_t fuids_ok = B_TRUE;
+ uint64_t zplver = ZPL_VERSION;
+ objset_t *os = NULL;
+ char parentname[MAXNAMELEN];
+ char *cp;
+ int error;
+
+ (void) strlcpy(parentname, dataset, sizeof (parentname));
+ cp = strrchr(parentname, '/');
+ ASSERT(cp != NULL);
+ cp[0] = '\0';
+
+ if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
+ zplver = ZPL_VERSION_FUID - 1;
+ fuids_ok = B_FALSE;
+ }
+
+ /*
+ * Open parent object set so we can inherit zplprop values.
+ */
+ if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
+ return (error);
+
+ error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
+ zplprops, is_ci);
+ dmu_objset_close(os);
+ return (error);
+}
+
+static int
+zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
+ nvlist_t *zplprops, boolean_t *is_ci)
+{
+ boolean_t fuids_ok = B_TRUE;
+ uint64_t zplver = ZPL_VERSION;
+ int error;
+
+ if (spa_vers < SPA_VERSION_FUID) {
+ zplver = ZPL_VERSION_FUID - 1;
+ fuids_ok = B_FALSE;
+ }
+
+ error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
+ zplprops, is_ci);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_objset_type type of objset to create (fs vs zvol)
+ * zc_name name of new objset
+ * zc_value name of snapshot to clone from (may be empty)
+ * zc_nvlist_src{_size} nvlist of properties to apply
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_create(zfs_cmd_t *zc)
+{
+ objset_t *clone;
+ int error = 0;
+ zfs_creat_t zct;
+ nvlist_t *nvprops = NULL;
+ void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
+ dmu_objset_type_t type = zc->zc_objset_type;
+
+ switch (type) {
+
+ case DMU_OST_ZFS:
+ cbfunc = zfs_create_cb;
+ break;
+
+ case DMU_OST_ZVOL:
+ cbfunc = zvol_create_cb;
+ break;
+
+ default:
+ cbfunc = NULL;
+ break;
+ }
+ if (strchr(zc->zc_name, '@') ||
+ strchr(zc->zc_name, '%'))
+ return (EINVAL);
+
+ if (zc->zc_nvlist_src != NULL &&
+ (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &nvprops)) != 0)
+ return (error);
+
+ zct.zct_zplprops = NULL;
+ zct.zct_props = nvprops;
+
+ if (zc->zc_value[0] != '\0') {
+ /*
+ * We're creating a clone of an existing snapshot.
+ */
+ zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
+ if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
+ nvlist_free(nvprops);
+ return (EINVAL);
+ }
+
+ error = dmu_objset_open(zc->zc_value, type,
+ DS_MODE_USER | DS_MODE_READONLY, &clone);
+ if (error) {
+ nvlist_free(nvprops);
+ return (error);
+ }
+
+ error = dmu_objset_create(zc->zc_name, type, clone, 0,
+ NULL, NULL);
+ if (error) {
+ dmu_objset_close(clone);
+ nvlist_free(nvprops);
+ return (error);
+ }
+ dmu_objset_close(clone);
+ } else {
+ boolean_t is_insensitive = B_FALSE;
+
+ if (cbfunc == NULL) {
+ nvlist_free(nvprops);
+ return (EINVAL);
+ }
+
+ if (type == DMU_OST_ZVOL) {
+ uint64_t volsize, volblocksize;
+
+ if (nvprops == NULL ||
+ nvlist_lookup_uint64(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLSIZE),
+ &volsize) != 0) {
+ nvlist_free(nvprops);
+ return (EINVAL);
+ }
+
+ if ((error = nvlist_lookup_uint64(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
+ &volblocksize)) != 0 && error != ENOENT) {
+ nvlist_free(nvprops);
+ return (EINVAL);
+ }
+
+ if (error != 0)
+ volblocksize = zfs_prop_default_numeric(
+ ZFS_PROP_VOLBLOCKSIZE);
+
+ if ((error = zvol_check_volblocksize(
+ volblocksize)) != 0 ||
+ (error = zvol_check_volsize(volsize,
+ volblocksize)) != 0) {
+ nvlist_free(nvprops);
+ return (error);
+ }
+ } else if (type == DMU_OST_ZFS) {
+ int error;
+
+ /*
+ * We have to have normalization and
+ * case-folding flags correct when we do the
+ * file system creation, so go figure them out
+ * now.
+ */
+ VERIFY(nvlist_alloc(&zct.zct_zplprops,
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ error = zfs_fill_zplprops(zc->zc_name, nvprops,
+ zct.zct_zplprops, &is_insensitive);
+ if (error != 0) {
+ nvlist_free(nvprops);
+ nvlist_free(zct.zct_zplprops);
+ return (error);
+ }
+ }
+ error = dmu_objset_create(zc->zc_name, type, NULL,
+ is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
+ nvlist_free(zct.zct_zplprops);
+ }
+
+ /*
+ * It would be nice to do this atomically.
+ */
+ if (error == 0) {
+ if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
+ (void) dmu_objset_destroy(zc->zc_name);
+ }
+ nvlist_free(nvprops);
+ return (error);
+}
+
+struct snap_prop_arg {
+ nvlist_t *nvprops;
+ const char *snapname;
+};
+
+static int
+set_snap_props(char *name, void *arg)
+{
+ struct snap_prop_arg *snpa = arg;
+ int len = strlen(name) + strlen(snpa->snapname) + 2;
+ char *buf = kmem_alloc(len, KM_SLEEP);
+ int err;
+
+ (void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
+ err = zfs_set_prop_nvlist(buf, snpa->nvprops);
+ if (err)
+ (void) dmu_objset_destroy(buf);
+ kmem_free(buf, len);
+ return (err);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_value short name of snapshot
+ * zc_cookie recursive flag
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_snapshot(zfs_cmd_t *zc)
+{
+ nvlist_t *nvprops = NULL;
+ int error;
+ boolean_t recursive = zc->zc_cookie;
+
+ if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
+ return (EINVAL);
+
+ if (zc->zc_nvlist_src != NULL &&
+ (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &nvprops)) != 0)
+ return (error);
+
+ error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
+
+ /*
+ * It would be nice to do this atomically.
+ */
+ if (error == 0) {
+ struct snap_prop_arg snpa;
+ snpa.nvprops = nvprops;
+ snpa.snapname = zc->zc_value;
+ if (recursive) {
+ error = dmu_objset_find(zc->zc_name,
+ set_snap_props, &snpa, DS_FIND_CHILDREN);
+ if (error) {
+ (void) dmu_snapshots_destroy(zc->zc_name,
+ zc->zc_value);
+ }
+ } else {
+ error = set_snap_props(zc->zc_name, &snpa);
+ }
+ }
+ nvlist_free(nvprops);
+ return (error);
+}
+
+int
+zfs_unmount_snap(char *name, void *arg)
+{
+ vfs_t *vfsp = NULL;
+
+ if (arg) {
+ char *snapname = arg;
+ int len = strlen(name) + strlen(snapname) + 2;
+ char *buf = kmem_alloc(len, KM_SLEEP);
+
+ (void) strcpy(buf, name);
+ (void) strcat(buf, "@");
+ (void) strcat(buf, snapname);
+ vfsp = zfs_get_vfs(buf);
+ kmem_free(buf, len);
+ } else if (strchr(name, '@')) {
+ vfsp = zfs_get_vfs(name);
+ }
+
+ if (vfsp) {
+ /*
+ * Always force the unmount for snapshots.
+ */
+ int flag = MS_FORCE;
+ int err;
+
+ if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
+ VFS_RELE(vfsp);
+ return (err);
+ }
+ VFS_RELE(vfsp);
+ if ((err = dounmount(vfsp, flag, kcred)) != 0)
+ return (err);
+ }
+ return (0);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_value short name of snapshot
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
+{
+ int err;
+
+ if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
+ return (EINVAL);
+ err = dmu_objset_find(zc->zc_name,
+ zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
+ if (err)
+ return (err);
+ return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
+}
+
+/*
+ * inputs:
+ * zc_name name of dataset to destroy
+ * zc_objset_type type of objset
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_destroy(zfs_cmd_t *zc)
+{
+ if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
+ int err = zfs_unmount_snap(zc->zc_name, NULL);
+ if (err)
+ return (err);
+ }
+
+ return (dmu_objset_destroy(zc->zc_name));
+}
+
+/*
+ * inputs:
+ * zc_name name of dataset to rollback (to most recent snapshot)
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_rollback(zfs_cmd_t *zc)
+{
+ objset_t *os;
+ int error;
+ zfsvfs_t *zfsvfs = NULL;
+
+ /*
+ * Get the zfsvfs for the receiving objset. There
+ * won't be one if we're operating on a zvol, if the
+ * objset doesn't exist yet, or is not mounted.
+ */
+ error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
+ if (error)
+ return (error);
+
+ if (dmu_objset_type(os) == DMU_OST_ZFS) {
+ mutex_enter(&os->os->os_user_ptr_lock);
+ zfsvfs = dmu_objset_get_user(os);
+ if (zfsvfs != NULL)
+ VFS_HOLD(zfsvfs->z_vfs);
+ mutex_exit(&os->os->os_user_ptr_lock);
+ }
+
+ if (zfsvfs != NULL) {
+ char *osname;
+ int mode;
+
+ osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
+ error = zfs_suspend_fs(zfsvfs, osname, &mode);
+ if (error == 0) {
+ int resume_err;
+
+ ASSERT(strcmp(osname, zc->zc_name) == 0);
+ error = dmu_objset_rollback(os);
+ resume_err = zfs_resume_fs(zfsvfs, osname, mode);
+ error = error ? error : resume_err;
+ } else {
+ dmu_objset_close(os);
+ }
+ kmem_free(osname, MAXNAMELEN);
+ VFS_RELE(zfsvfs->z_vfs);
+ } else {
+ error = dmu_objset_rollback(os);
+ }
+ /* Note, the dmu_objset_rollback() releases the objset for us. */
+
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name old name of dataset
+ * zc_value new name of dataset
+ * zc_cookie recursive flag (only valid for snapshots)
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_rename(zfs_cmd_t *zc)
+{
+ boolean_t recursive = zc->zc_cookie & 1;
+
+ zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
+ if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
+ strchr(zc->zc_value, '%'))
+ return (EINVAL);
+
+ /*
+ * Unmount snapshot unless we're doing a recursive rename,
+ * in which case the dataset code figures out which snapshots
+ * to unmount.
+ */
+ if (!recursive && strchr(zc->zc_name, '@') != NULL &&
+ zc->zc_objset_type == DMU_OST_ZFS) {
+ int err = zfs_unmount_snap(zc->zc_name, NULL);
+ if (err)
+ return (err);
+ }
+ return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
+}
+
+static void
+clear_props(char *dataset, nvlist_t *props)
+{
+ zfs_cmd_t *zc;
+ nvpair_t *prop;
+
+ if (props == NULL)
+ return;
+ zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
+ (void) strcpy(zc->zc_name, dataset);
+ for (prop = nvlist_next_nvpair(props, NULL); prop;
+ prop = nvlist_next_nvpair(props, prop)) {
+ (void) strcpy(zc->zc_value, nvpair_name(prop));
+ if (zfs_secpolicy_inherit(zc, CRED()) == 0)
+ (void) zfs_ioc_inherit_prop(zc);
+ }
+ kmem_free(zc, sizeof (zfs_cmd_t));
+}
+
+/*
+ * inputs:
+ * zc_name name of containing filesystem
+ * zc_nvlist_src{_size} nvlist of properties to apply
+ * zc_value name of snapshot to create
+ * zc_string name of clone origin (if DRR_FLAG_CLONE)
+ * zc_cookie file descriptor to recv from
+ * zc_begin_record the BEGIN record of the stream (not byteswapped)
+ * zc_guid force flag
+ *
+ * outputs:
+ * zc_cookie number of bytes read
+ */
+static int
+zfs_ioc_recv(zfs_cmd_t *zc)
+{
+ file_t *fp;
+ objset_t *os;
+ dmu_recv_cookie_t drc;
+ zfsvfs_t *zfsvfs = NULL;
+ boolean_t force = (boolean_t)zc->zc_guid;
+ int error, fd;
+ offset_t off;
+ nvlist_t *props = NULL;
+ nvlist_t *origprops = NULL;
+ objset_t *origin = NULL;
+ char *tosnap;
+ char tofs[ZFS_MAXNAMELEN];
+
+ if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
+ strchr(zc->zc_value, '@') == NULL ||
+ strchr(zc->zc_value, '%'))
+ return (EINVAL);
+
+ (void) strcpy(tofs, zc->zc_value);
+ tosnap = strchr(tofs, '@');
+ *tosnap = '\0';
+ tosnap++;
+
+ if (zc->zc_nvlist_src != NULL &&
+ (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
+ &props)) != 0)
+ return (error);
+
+ fd = zc->zc_cookie;
+ fp = getf(fd);
+ if (fp == NULL) {
+ nvlist_free(props);
+ return (EBADF);
+ }
+
+ if (dmu_objset_open(tofs, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
+ /*
+ * Try to get the zfsvfs for the receiving objset.
+ * There won't be one if we're operating on a zvol,
+ * if the objset doesn't exist yet, or is not mounted.
+ */
+ mutex_enter(&os->os->os_user_ptr_lock);
+ if (zfsvfs = dmu_objset_get_user(os)) {
+ if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
+ mutex_exit(&os->os->os_user_ptr_lock);
+ dmu_objset_close(os);
+ zfsvfs = NULL;
+ error = EBUSY;
+ goto out;
+ }
+ VFS_HOLD(zfsvfs->z_vfs);
+ }
+ mutex_exit(&os->os->os_user_ptr_lock);
+
+ /*
+ * If new properties are supplied, they are to completely
+ * replace the existing ones, so stash away the existing ones.
+ */
+ if (props)
+ (void) dsl_prop_get_all(os, &origprops, TRUE);
+
+ dmu_objset_close(os);
+ }
+
+ if (zc->zc_string[0]) {
+ error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &origin);
+ if (error)
+ goto out;
+ }
+
+ error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
+ force, origin, zfsvfs != NULL, &drc);
+ if (origin)
+ dmu_objset_close(origin);
+ if (error)
+ goto out;
+
+ /*
+ * Reset properties. We do this before we receive the stream
+ * so that the properties are applied to the new data.
+ */
+ if (props) {
+ clear_props(tofs, origprops);
+ /*
+ * XXX - Note, this is all-or-nothing; should be best-effort.
+ */
+ (void) zfs_set_prop_nvlist(tofs, props);
+ }
+
+ off = fp->f_offset;
+ error = dmu_recv_stream(&drc, fp->f_vnode, &off);
+
+ if (error == 0 && zfsvfs) {
+ char *osname;
+ int mode;
+
+ /* online recv */
+ osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
+ error = zfs_suspend_fs(zfsvfs, osname, &mode);
+ if (error == 0) {
+ int resume_err;
+
+ error = dmu_recv_end(&drc);
+ resume_err = zfs_resume_fs(zfsvfs, osname, mode);
+ error = error ? error : resume_err;
+ } else {
+ dmu_recv_abort_cleanup(&drc);
+ }
+ kmem_free(osname, MAXNAMELEN);
+ } else if (error == 0) {
+ error = dmu_recv_end(&drc);
+ }
+
+ zc->zc_cookie = off - fp->f_offset;
+ if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
+ fp->f_offset = off;
+
+ /*
+ * On error, restore the original props.
+ */
+ if (error && props) {
+ clear_props(tofs, props);
+ (void) zfs_set_prop_nvlist(tofs, origprops);
+ }
+out:
+ if (zfsvfs) {
+ mutex_exit(&zfsvfs->z_online_recv_lock);
+ VFS_RELE(zfsvfs->z_vfs);
+ }
+ nvlist_free(props);
+ nvlist_free(origprops);
+ releasef(fd);
+ return (error);
+}
+
+/*
+ * inputs:
+ * zc_name name of snapshot to send
+ * zc_value short name of incremental fromsnap (may be empty)
+ * zc_cookie file descriptor to send stream to
+ * zc_obj fromorigin flag (mutually exclusive with zc_value)
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_send(zfs_cmd_t *zc)
+{
+ objset_t *fromsnap = NULL;
+ objset_t *tosnap;
+ file_t *fp;
+ int error;
+ offset_t off;
+
+ error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &tosnap);
+ if (error)
+ return (error);
+
+ if (zc->zc_value[0] != '\0') {
+ char *buf;
+ char *cp;
+
+ buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+ (void) strncpy(buf, zc->zc_name, MAXPATHLEN);
+ cp = strchr(buf, '@');
+ if (cp)
+ *(cp+1) = 0;
+ (void) strncat(buf, zc->zc_value, MAXPATHLEN);
+ error = dmu_objset_open(buf, DMU_OST_ANY,
+ DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
+ kmem_free(buf, MAXPATHLEN);
+ if (error) {
+ dmu_objset_close(tosnap);
+ return (error);
+ }
+ }
+
+ fp = getf(zc->zc_cookie);
+ if (fp == NULL) {
+ dmu_objset_close(tosnap);
+ if (fromsnap)
+ dmu_objset_close(fromsnap);
+ return (EBADF);
+ }
+
+ off = fp->f_offset;
+ error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
+
+ if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
+ fp->f_offset = off;
+ releasef(zc->zc_cookie);
+ if (fromsnap)
+ dmu_objset_close(fromsnap);
+ dmu_objset_close(tosnap);
+ return (error);
+}
+
+static int
+zfs_ioc_inject_fault(zfs_cmd_t *zc)
+{
+ int id, error;
+
+ error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
+ &zc->zc_inject_record);
+
+ if (error == 0)
+ zc->zc_guid = (uint64_t)id;
+
+ return (error);
+}
+
+static int
+zfs_ioc_clear_fault(zfs_cmd_t *zc)
+{
+ return (zio_clear_fault((int)zc->zc_guid));
+}
+
+static int
+zfs_ioc_inject_list_next(zfs_cmd_t *zc)
+{
+ int id = (int)zc->zc_guid;
+ int error;
+
+ error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
+ &zc->zc_inject_record);
+
+ zc->zc_guid = id;
+
+ return (error);
+}
+
+static int
+zfs_ioc_error_log(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ int error;
+ size_t count = (size_t)zc->zc_nvlist_dst_size;
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
+ &count);
+ if (error == 0)
+ zc->zc_nvlist_dst_size = count;
+ else
+ zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
+
+ spa_close(spa, FTAG);
+
+ return (error);
+}
+
+static int
+zfs_ioc_clear(zfs_cmd_t *zc)
+{
+ spa_t *spa;
+ vdev_t *vd;
+ int error;
+
+ /*
+ * On zpool clear we also fix up missing slogs
+ */
+ mutex_enter(&spa_namespace_lock);
+ spa = spa_lookup(zc->zc_name);
+ if (spa == NULL) {
+ mutex_exit(&spa_namespace_lock);
+ return (EIO);
+ }
+ if (spa->spa_log_state == SPA_LOG_MISSING) {
+ /* we need to let spa_open/spa_load clear the chains */
+ spa->spa_log_state = SPA_LOG_CLEAR;
+ }
+ mutex_exit(&spa_namespace_lock);
+
+ if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
+ return (error);
+
+ spa_vdev_state_enter(spa);
+
+ if (zc->zc_guid == 0) {
+ vd = NULL;
+ } else {
+ vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
+ if (vd == NULL) {
+ (void) spa_vdev_state_exit(spa, NULL, ENODEV);
+ spa_close(spa, FTAG);
+ return (ENODEV);
+ }
+ }
+
+ vdev_clear(spa, vd);
+
+ (void) spa_vdev_state_exit(spa, NULL, 0);
+
+ /*
+ * Resume any suspended I/Os.
+ */
+ zio_resume(spa);
+
+ spa_close(spa, FTAG);
+
+ return (0);
+}
+
+/*
+ * inputs:
+ * zc_name name of filesystem
+ * zc_value name of origin snapshot
+ *
+ * outputs: none
+ */
+static int
+zfs_ioc_promote(zfs_cmd_t *zc)
+{
+ char *cp;
+
+ /*
+ * We don't need to unmount *all* the origin fs's snapshots, but
+ * it's easier.
+ */
+ cp = strchr(zc->zc_value, '@');
+ if (cp)
+ *cp = '\0';
+ (void) dmu_objset_find(zc->zc_value,
+ zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
+ return (dsl_dataset_promote(zc->zc_name));
+}
+
+/*
+ * We don't want to have a hard dependency
+ * against some special symbols in sharefs
+ * nfs, and smbsrv. Determine them if needed when
+ * the first file system is shared.
+ * Neither sharefs, nfs or smbsrv are unloadable modules.
+ */
+int (*znfsexport_fs)(void *arg);
+int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
+int (*zsmbexport_fs)(void *arg, boolean_t add_share);
+
+int zfs_nfsshare_inited;
+int zfs_smbshare_inited;
+
+ddi_modhandle_t nfs_mod;
+ddi_modhandle_t sharefs_mod;
+ddi_modhandle_t smbsrv_mod;
+kmutex_t zfs_share_lock;
+
+static int
+zfs_init_sharefs()
+{
+ int error;
+
+ ASSERT(MUTEX_HELD(&zfs_share_lock));
+ /* Both NFS and SMB shares also require sharetab support. */
+ if (sharefs_mod == NULL && ((sharefs_mod =
+ ddi_modopen("fs/sharefs",
+ KRTLD_MODE_FIRST, &error)) == NULL)) {
+ return (ENOSYS);
+ }
+ if (zshare_fs == NULL && ((zshare_fs =
+ (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
+ ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
+ return (ENOSYS);
+ }
+ return (0);
+}
+
+static int
+zfs_ioc_share(zfs_cmd_t *zc)
+{
+ int error;
+ int opcode;
+
+ switch (zc->zc_share.z_sharetype) {
+ case ZFS_SHARE_NFS:
+ case ZFS_UNSHARE_NFS:
+ if (zfs_nfsshare_inited == 0) {
+ mutex_enter(&zfs_share_lock);
+ if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
+ KRTLD_MODE_FIRST, &error)) == NULL)) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ if (znfsexport_fs == NULL &&
+ ((znfsexport_fs = (int (*)(void *))
+ ddi_modsym(nfs_mod,
+ "nfs_export", &error)) == NULL)) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ error = zfs_init_sharefs();
+ if (error) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ zfs_nfsshare_inited = 1;
+ mutex_exit(&zfs_share_lock);
+ }
+ break;
+ case ZFS_SHARE_SMB:
+ case ZFS_UNSHARE_SMB:
+ if (zfs_smbshare_inited == 0) {
+ mutex_enter(&zfs_share_lock);
+ if (smbsrv_mod == NULL && ((smbsrv_mod =
+ ddi_modopen("drv/smbsrv",
+ KRTLD_MODE_FIRST, &error)) == NULL)) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ if (zsmbexport_fs == NULL && ((zsmbexport_fs =
+ (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
+ "smb_server_share", &error)) == NULL)) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ error = zfs_init_sharefs();
+ if (error) {
+ mutex_exit(&zfs_share_lock);
+ return (ENOSYS);
+ }
+ zfs_smbshare_inited = 1;
+ mutex_exit(&zfs_share_lock);
+ }
+ break;
+ default:
+ return (EINVAL);
+ }
+
+ switch (zc->zc_share.z_sharetype) {
+ case ZFS_SHARE_NFS:
+ case ZFS_UNSHARE_NFS:
+ if (error =
+ znfsexport_fs((void *)
+ (uintptr_t)zc->zc_share.z_exportdata))
+ return (error);
+ break;
+ case ZFS_SHARE_SMB:
+ case ZFS_UNSHARE_SMB:
+ if (error = zsmbexport_fs((void *)
+ (uintptr_t)zc->zc_share.z_exportdata,
+ zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
+ B_TRUE : B_FALSE)) {
+ return (error);
+ }
+ break;
+ }
+
+ opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
+ zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
+ SHAREFS_ADD : SHAREFS_REMOVE;
+
+ /*
+ * Add or remove share from sharetab
+ */
+ error = zshare_fs(opcode,
+ (void *)(uintptr_t)zc->zc_share.z_sharedata,
+ zc->zc_share.z_sharemax);
+
+ return (error);
+
+}
+
+/*
+ * pool create, destroy, and export don't log the history as part of
+ * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
+ * do the logging of those commands.
+ */
+static zfs_ioc_vec_t zfs_ioc_vec[] = {
+ { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE },
+ { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
+ { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
+ { zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
+ { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
+ { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
+ { zfs_ioc_dataset_list_next, zfs_secpolicy_read,
+ DATASET_NAME, B_FALSE },
+ { zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
+ DATASET_NAME, B_FALSE },
+ { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
+ { zfs_ioc_create_minor, zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
+ { zfs_ioc_remove_minor, zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
+ { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
+ { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
+ { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
+ { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE },
+ { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
+ { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
+ { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
+ { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
+ { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
+ { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
+ { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
+ { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
+ { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
+ { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
+ { zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
+ { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE },
+ { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
+ { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
+ { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
+ { zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
+ DATASET_NAME, B_FALSE },
+ { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
+ { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
+};
+
+static int
+zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
+{
+ zfs_cmd_t *zc;
+ uint_t vec;
+ int error, rc;
+
+ if (getminor(dev) != 0)
+ return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
+
+ vec = cmd - ZFS_IOC;
+ ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
+
+ if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
+ return (EINVAL);
+
+ zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
+
+ error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
+
+ if (error == 0)
+ error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
+
+ /*
+ * Ensure that all pool/dataset names are valid before we pass down to
+ * the lower layers.
+ */
+ if (error == 0) {
+ zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
+ switch (zfs_ioc_vec[vec].zvec_namecheck) {
+ case POOL_NAME:
+ if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
+ error = EINVAL;
+ break;
+
+ case DATASET_NAME:
+ if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
+ error = EINVAL;
+ break;
+
+ case NO_NAME:
+ break;
+ }
+ }
+
+ if (error == 0)
+ error = zfs_ioc_vec[vec].zvec_func(zc);
+
+ rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
+ if (error == 0) {
+ error = rc;
+ if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
+ zfs_log_history(zc);
+ }
+
+ kmem_free(zc, sizeof (zfs_cmd_t));
+ return (error);
+}
+
+static int
+zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
+{
+ if (cmd != DDI_ATTACH)
+ return (DDI_FAILURE);
+
+ if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
+ DDI_PSEUDO, 0) == DDI_FAILURE)
+ return (DDI_FAILURE);
+
+ zfs_dip = dip;
+
+ ddi_report_dev(dip);
+
+ return (DDI_SUCCESS);
+}
+
+static int
+zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
+{
+ if (spa_busy() || zfs_busy() || zvol_busy())
+ return (DDI_FAILURE);
+
+ if (cmd != DDI_DETACH)
+ return (DDI_FAILURE);
+
+ zfs_dip = NULL;
+
+ ddi_prop_remove_all(dip);
+ ddi_remove_minor_node(dip, NULL);
+
+ return (DDI_SUCCESS);
+}
+
+/*ARGSUSED*/
+static int
+zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
+{
+ switch (infocmd) {
+ case DDI_INFO_DEVT2DEVINFO:
+ *result = zfs_dip;
+ return (DDI_SUCCESS);
+
+ case DDI_INFO_DEVT2INSTANCE:
+ *result = (void *)0;
+ return (DDI_SUCCESS);
+ }
+
+ return (DDI_FAILURE);
+}
+
+/*
+ * OK, so this is a little weird.
+ *
+ * /dev/zfs is the control node, i.e. minor 0.
+ * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
+ *
+ * /dev/zfs has basically nothing to do except serve up ioctls,
+ * so most of the standard driver entry points are in zvol.c.
+ */
+static struct cb_ops zfs_cb_ops = {
+ zvol_open, /* open */
+ zvol_close, /* close */
+ zvol_strategy, /* strategy */
+ nodev, /* print */
+ zvol_dump, /* dump */
+ zvol_read, /* read */
+ zvol_write, /* write */
+ zfsdev_ioctl, /* ioctl */
+ nodev, /* devmap */
+ nodev, /* mmap */
+ nodev, /* segmap */
+ nochpoll, /* poll */
+ ddi_prop_op, /* prop_op */
+ NULL, /* streamtab */
+ D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
+ CB_REV, /* version */
+ nodev, /* async read */
+ nodev, /* async write */
+};
+
+static struct dev_ops zfs_dev_ops = {
+ DEVO_REV, /* version */
+ 0, /* refcnt */
+ zfs_info, /* info */
+ nulldev, /* identify */
+ nulldev, /* probe */
+ zfs_attach, /* attach */
+ zfs_detach, /* detach */
+ nodev, /* reset */
+ &zfs_cb_ops, /* driver operations */
+ NULL, /* no bus operations */
+ NULL, /* power */
+ ddi_quiesce_not_needed, /* quiesce */
+};
+
+static struct modldrv zfs_modldrv = {
+ &mod_driverops,
+ "ZFS storage pool",
+ &zfs_dev_ops
+};
+
+static struct modlinkage modlinkage = {
+ MODREV_1,
+ (void *)&zfs_modlfs,
+ (void *)&zfs_modldrv,
+ NULL
+};
+
+
+uint_t zfs_fsyncer_key;
+extern uint_t rrw_tsd_key;
+
+int
+_init(void)
+{
+ int error;
+
+ spa_init(FREAD | FWRITE);
+ zfs_init();
+ zvol_init();
+
+ if ((error = mod_install(&modlinkage)) != 0) {
+ zvol_fini();
+ zfs_fini();
+ spa_fini();
+ return (error);
+ }
+
+ tsd_create(&zfs_fsyncer_key, NULL);
+ tsd_create(&rrw_tsd_key, NULL);
+
+ error = ldi_ident_from_mod(&modlinkage, &zfs_li);
+ ASSERT(error == 0);
+ mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ return (0);
+}
+
+int
+_fini(void)
+{
+ int error;
+
+ if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
+ return (EBUSY);
+
+ if ((error = mod_remove(&modlinkage)) != 0)
+ return (error);
+
+ zvol_fini();
+ zfs_fini();
+ spa_fini();
+ if (zfs_nfsshare_inited)
+ (void) ddi_modclose(nfs_mod);
+ if (zfs_smbshare_inited)
+ (void) ddi_modclose(smbsrv_mod);
+ if (zfs_nfsshare_inited || zfs_smbshare_inited)
+ (void) ddi_modclose(sharefs_mod);
+
+ tsd_destroy(&zfs_fsyncer_key);
+ ldi_ident_release(zfs_li);
+ zfs_li = NULL;
+ mutex_destroy(&zfs_share_lock);
+
+ return (error);
+}
+
+int
+_info(struct modinfo *modinfop)
+{
+ return (mod_info(&modlinkage, modinfop));
+}
diff --git a/module/zfs/zfs_log.c b/module/zfs/zfs_log.c
new file mode 100644
index 000000000..11cd4c264
--- /dev/null
+++ b/module/zfs/zfs_log.c
@@ -0,0 +1,694 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/cmn_err.h>
+#include <sys/kmem.h>
+#include <sys/thread.h>
+#include <sys/file.h>
+#include <sys/vfs.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_dir.h>
+#include <sys/zil.h>
+#include <sys/zil_impl.h>
+#include <sys/byteorder.h>
+#include <sys/policy.h>
+#include <sys/stat.h>
+#include <sys/mode.h>
+#include <sys/acl.h>
+#include <sys/dmu.h>
+#include <sys/spa.h>
+#include <sys/zfs_fuid.h>
+#include <sys/ddi.h>
+
+/*
+ * All the functions in this file are used to construct the log entries
+ * to record transactions. They allocate * an intent log transaction
+ * structure (itx_t) and save within it all the information necessary to
+ * possibly replay the transaction. The itx is then assigned a sequence
+ * number and inserted in the in-memory list anchored in the zilog.
+ */
+
+int
+zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
+{
+ int isxvattr = (vap->va_mask & AT_XVATTR);
+ switch (type) {
+ case Z_FILE:
+ if (vsecp == NULL && !isxvattr)
+ return (TX_CREATE);
+ if (vsecp && isxvattr)
+ return (TX_CREATE_ACL_ATTR);
+ if (vsecp)
+ return (TX_CREATE_ACL);
+ else
+ return (TX_CREATE_ATTR);
+ /*NOTREACHED*/
+ case Z_DIR:
+ if (vsecp == NULL && !isxvattr)
+ return (TX_MKDIR);
+ if (vsecp && isxvattr)
+ return (TX_MKDIR_ACL_ATTR);
+ if (vsecp)
+ return (TX_MKDIR_ACL);
+ else
+ return (TX_MKDIR_ATTR);
+ case Z_XATTRDIR:
+ return (TX_MKXATTR);
+ }
+ ASSERT(0);
+ return (TX_MAX_TYPE);
+}
+
+/*
+ * build up the log data necessary for logging xvattr_t
+ * First lr_attr_t is initialized. following the lr_attr_t
+ * is the mapsize and attribute bitmap copied from the xvattr_t.
+ * Following the bitmap and bitmapsize two 64 bit words are reserved
+ * for the create time which may be set. Following the create time
+ * records a single 64 bit integer which has the bits to set on
+ * replay for the xvattr.
+ */
+static void
+zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
+{
+ uint32_t *bitmap;
+ uint64_t *attrs;
+ uint64_t *crtime;
+ xoptattr_t *xoap;
+ void *scanstamp;
+ int i;
+
+ xoap = xva_getxoptattr(xvap);
+ ASSERT(xoap);
+
+ lrattr->lr_attr_masksize = xvap->xva_mapsize;
+ bitmap = &lrattr->lr_attr_bitmap;
+ for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
+ *bitmap = xvap->xva_reqattrmap[i];
+ }
+
+ /* Now pack the attributes up in a single uint64_t */
+ attrs = (uint64_t *)bitmap;
+ crtime = attrs + 1;
+ scanstamp = (caddr_t)(crtime + 2);
+ *attrs = 0;
+ if (XVA_ISSET_REQ(xvap, XAT_READONLY))
+ *attrs |= (xoap->xoa_readonly == 0) ? 0 :
+ XAT0_READONLY;
+ if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
+ *attrs |= (xoap->xoa_hidden == 0) ? 0 :
+ XAT0_HIDDEN;
+ if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
+ *attrs |= (xoap->xoa_system == 0) ? 0 :
+ XAT0_SYSTEM;
+ if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
+ *attrs |= (xoap->xoa_archive == 0) ? 0 :
+ XAT0_ARCHIVE;
+ if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
+ *attrs |= (xoap->xoa_immutable == 0) ? 0 :
+ XAT0_IMMUTABLE;
+ if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
+ *attrs |= (xoap->xoa_nounlink == 0) ? 0 :
+ XAT0_NOUNLINK;
+ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
+ *attrs |= (xoap->xoa_appendonly == 0) ? 0 :
+ XAT0_APPENDONLY;
+ if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
+ *attrs |= (xoap->xoa_opaque == 0) ? 0 :
+ XAT0_APPENDONLY;
+ if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
+ *attrs |= (xoap->xoa_nodump == 0) ? 0 :
+ XAT0_NODUMP;
+ if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
+ *attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
+ XAT0_AV_QUARANTINED;
+ if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
+ *attrs |= (xoap->xoa_av_modified == 0) ? 0 :
+ XAT0_AV_MODIFIED;
+ if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
+ ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
+ if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
+ bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
+}
+
+static void *
+zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
+{
+ zfs_fuid_t *zfuid;
+ uint64_t *fuidloc = start;
+
+ /* First copy in the ACE FUIDs */
+ for (zfuid = list_head(&fuidp->z_fuids); zfuid;
+ zfuid = list_next(&fuidp->z_fuids, zfuid)) {
+ *fuidloc++ = zfuid->z_logfuid;
+ }
+ return (fuidloc);
+}
+
+
+static void *
+zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
+{
+ zfs_fuid_domain_t *zdomain;
+
+ /* now copy in the domain info, if any */
+ if (fuidp->z_domain_str_sz != 0) {
+ for (zdomain = list_head(&fuidp->z_domains); zdomain;
+ zdomain = list_next(&fuidp->z_domains, zdomain)) {
+ bcopy((void *)zdomain->z_domain, start,
+ strlen(zdomain->z_domain) + 1);
+ start = (caddr_t)start +
+ strlen(zdomain->z_domain) + 1;
+ }
+ }
+ return (start);
+}
+
+/*
+ * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR,
+ * TX_MKDIR_ATTR and TX_MKXATTR
+ * transactions.
+ *
+ * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
+ * domain information appended prior to the name. In this case the
+ * uid/gid in the log record will be a log centric FUID.
+ *
+ * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
+ * may contain attributes, ACL and optional fuid information.
+ *
+ * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
+ * and ACL and normal users/groups in the ACEs.
+ *
+ * There may be an optional xvattr attribute information similar
+ * to zfs_log_setattr.
+ *
+ * Also, after the file name "domain" strings may be appended.
+ */
+void
+zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
+ zfs_fuid_info_t *fuidp, vattr_t *vap)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_create_t *lr;
+ lr_acl_create_t *lracl;
+ size_t aclsize;
+ size_t xvatsize = 0;
+ size_t txsize;
+ xvattr_t *xvap = (xvattr_t *)vap;
+ void *end;
+ size_t lrsize;
+ size_t namesize = strlen(name) + 1;
+ size_t fuidsz = 0;
+
+ if (zilog == NULL)
+ return;
+
+ /*
+ * If we have FUIDs present then add in space for
+ * domains and ACE fuid's if any.
+ */
+ if (fuidp) {
+ fuidsz += fuidp->z_domain_str_sz;
+ fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
+ }
+
+ if (vap->va_mask & AT_XVATTR)
+ xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
+
+ if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
+ (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
+ (int)txtype == TX_MKXATTR) {
+ txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
+ lrsize = sizeof (*lr);
+ } else {
+ aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0;
+ txsize =
+ sizeof (lr_acl_create_t) + namesize + fuidsz +
+ ZIL_ACE_LENGTH(aclsize) + xvatsize;
+ lrsize = sizeof (lr_acl_create_t);
+ }
+
+ itx = zil_itx_create(txtype, txsize);
+
+ lr = (lr_create_t *)&itx->itx_lr;
+ lr->lr_doid = dzp->z_id;
+ lr->lr_foid = zp->z_id;
+ lr->lr_mode = zp->z_phys->zp_mode;
+ if (!IS_EPHEMERAL(zp->z_phys->zp_uid)) {
+ lr->lr_uid = (uint64_t)zp->z_phys->zp_uid;
+ } else {
+ lr->lr_uid = fuidp->z_fuid_owner;
+ }
+ if (!IS_EPHEMERAL(zp->z_phys->zp_gid)) {
+ lr->lr_gid = (uint64_t)zp->z_phys->zp_gid;
+ } else {
+ lr->lr_gid = fuidp->z_fuid_group;
+ }
+ lr->lr_gen = zp->z_phys->zp_gen;
+ lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
+ lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
+ lr->lr_rdev = zp->z_phys->zp_rdev;
+
+ /*
+ * Fill in xvattr info if any
+ */
+ if (vap->va_mask & AT_XVATTR) {
+ zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
+ end = (caddr_t)lr + lrsize + xvatsize;
+ } else {
+ end = (caddr_t)lr + lrsize;
+ }
+
+ /* Now fill in any ACL info */
+
+ if (vsecp) {
+ lracl = (lr_acl_create_t *)&itx->itx_lr;
+ lracl->lr_aclcnt = vsecp->vsa_aclcnt;
+ lracl->lr_acl_bytes = aclsize;
+ lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
+ lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
+ if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
+ lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
+ else
+ lracl->lr_acl_flags = 0;
+
+ bcopy(vsecp->vsa_aclentp, end, aclsize);
+ end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
+ }
+
+ /* drop in FUID info */
+ if (fuidp) {
+ end = zfs_log_fuid_ids(fuidp, end);
+ end = zfs_log_fuid_domains(fuidp, end);
+ }
+ /*
+ * Now place file name in log record
+ */
+ bcopy(name, end, namesize);
+
+ seq = zil_itx_assign(zilog, itx, tx);
+ dzp->z_last_itx = seq;
+ zp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
+ */
+void
+zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, char *name)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_remove_t *lr;
+ size_t namesize = strlen(name) + 1;
+
+ if (zilog == NULL)
+ return;
+
+ itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
+ lr = (lr_remove_t *)&itx->itx_lr;
+ lr->lr_doid = dzp->z_id;
+ bcopy(name, (char *)(lr + 1), namesize);
+
+ seq = zil_itx_assign(zilog, itx, tx);
+ dzp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_link() handles TX_LINK transactions.
+ */
+void
+zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_link_t *lr;
+ size_t namesize = strlen(name) + 1;
+
+ if (zilog == NULL)
+ return;
+
+ itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
+ lr = (lr_link_t *)&itx->itx_lr;
+ lr->lr_doid = dzp->z_id;
+ lr->lr_link_obj = zp->z_id;
+ bcopy(name, (char *)(lr + 1), namesize);
+
+ seq = zil_itx_assign(zilog, itx, tx);
+ dzp->z_last_itx = seq;
+ zp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_symlink() handles TX_SYMLINK transactions.
+ */
+void
+zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *dzp, znode_t *zp, char *name, char *link)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_create_t *lr;
+ size_t namesize = strlen(name) + 1;
+ size_t linksize = strlen(link) + 1;
+
+ if (zilog == NULL)
+ return;
+
+ itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
+ lr = (lr_create_t *)&itx->itx_lr;
+ lr->lr_doid = dzp->z_id;
+ lr->lr_foid = zp->z_id;
+ lr->lr_mode = zp->z_phys->zp_mode;
+ lr->lr_uid = zp->z_phys->zp_uid;
+ lr->lr_gid = zp->z_phys->zp_gid;
+ lr->lr_gen = zp->z_phys->zp_gen;
+ lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
+ lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
+ bcopy(name, (char *)(lr + 1), namesize);
+ bcopy(link, (char *)(lr + 1) + namesize, linksize);
+
+ seq = zil_itx_assign(zilog, itx, tx);
+ dzp->z_last_itx = seq;
+ zp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_rename() handles TX_RENAME transactions.
+ */
+void
+zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
+ znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_rename_t *lr;
+ size_t snamesize = strlen(sname) + 1;
+ size_t dnamesize = strlen(dname) + 1;
+
+ if (zilog == NULL)
+ return;
+
+ itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
+ lr = (lr_rename_t *)&itx->itx_lr;
+ lr->lr_sdoid = sdzp->z_id;
+ lr->lr_tdoid = tdzp->z_id;
+ bcopy(sname, (char *)(lr + 1), snamesize);
+ bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
+
+ seq = zil_itx_assign(zilog, itx, tx);
+ sdzp->z_last_itx = seq;
+ tdzp->z_last_itx = seq;
+ szp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_write() handles TX_WRITE transactions.
+ */
+ssize_t zfs_immediate_write_sz = 32768;
+
+#define ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \
+ sizeof (lr_write_t))
+
+void
+zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, offset_t off, ssize_t resid, int ioflag)
+{
+ itx_wr_state_t write_state;
+ boolean_t slogging;
+ uintptr_t fsync_cnt;
+
+ if (zilog == NULL || zp->z_unlinked)
+ return;
+
+ /*
+ * Writes are handled in three different ways:
+ *
+ * WR_INDIRECT:
+ * In this mode, if we need to commit the write later, then the block
+ * is immediately written into the file system (using dmu_sync),
+ * and a pointer to the block is put into the log record.
+ * When the txg commits the block is linked in.
+ * This saves additionally writing the data into the log record.
+ * There are a few requirements for this to occur:
+ * - write is greater than zfs_immediate_write_sz
+ * - not using slogs (as slogs are assumed to always be faster
+ * than writing into the main pool)
+ * - the write occupies only one block
+ * WR_COPIED:
+ * If we know we'll immediately be committing the
+ * transaction (FSYNC or FDSYNC), the we allocate a larger
+ * log record here for the data and copy the data in.
+ * WR_NEED_COPY:
+ * Otherwise we don't allocate a buffer, and *if* we need to
+ * flush the write later then a buffer is allocated and
+ * we retrieve the data using the dmu.
+ */
+ slogging = spa_has_slogs(zilog->zl_spa);
+ if (resid > zfs_immediate_write_sz && !slogging && resid <= zp->z_blksz)
+ write_state = WR_INDIRECT;
+ else if (ioflag & (FSYNC | FDSYNC))
+ write_state = WR_COPIED;
+ else
+ write_state = WR_NEED_COPY;
+
+ if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
+ (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
+ }
+
+ while (resid) {
+ itx_t *itx;
+ lr_write_t *lr;
+ ssize_t len;
+
+ /*
+ * If the write would overflow the largest block then split it.
+ */
+ if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
+ len = SPA_MAXBLOCKSIZE >> 1;
+ else
+ len = resid;
+
+ itx = zil_itx_create(txtype, sizeof (*lr) +
+ (write_state == WR_COPIED ? len : 0));
+ lr = (lr_write_t *)&itx->itx_lr;
+ if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
+ zp->z_id, off, len, lr + 1) != 0) {
+ kmem_free(itx, offsetof(itx_t, itx_lr) +
+ itx->itx_lr.lrc_reclen);
+ itx = zil_itx_create(txtype, sizeof (*lr));
+ lr = (lr_write_t *)&itx->itx_lr;
+ write_state = WR_NEED_COPY;
+ }
+
+ itx->itx_wr_state = write_state;
+ if (write_state == WR_NEED_COPY)
+ itx->itx_sod += len;
+ lr->lr_foid = zp->z_id;
+ lr->lr_offset = off;
+ lr->lr_length = len;
+ lr->lr_blkoff = 0;
+ BP_ZERO(&lr->lr_blkptr);
+
+ itx->itx_private = zp->z_zfsvfs;
+
+ if ((zp->z_sync_cnt != 0) || (fsync_cnt != 0) ||
+ (ioflag & (FSYNC | FDSYNC)))
+ itx->itx_sync = B_TRUE;
+ else
+ itx->itx_sync = B_FALSE;
+
+ zp->z_last_itx = zil_itx_assign(zilog, itx, tx);
+
+ off += len;
+ resid -= len;
+ }
+}
+
+/*
+ * zfs_log_truncate() handles TX_TRUNCATE transactions.
+ */
+void
+zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, uint64_t off, uint64_t len)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_truncate_t *lr;
+
+ if (zilog == NULL || zp->z_unlinked)
+ return;
+
+ itx = zil_itx_create(txtype, sizeof (*lr));
+ lr = (lr_truncate_t *)&itx->itx_lr;
+ lr->lr_foid = zp->z_id;
+ lr->lr_offset = off;
+ lr->lr_length = len;
+
+ itx->itx_sync = (zp->z_sync_cnt != 0);
+ seq = zil_itx_assign(zilog, itx, tx);
+ zp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_setattr() handles TX_SETATTR transactions.
+ */
+void
+zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
+ znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_setattr_t *lr;
+ xvattr_t *xvap = (xvattr_t *)vap;
+ size_t recsize = sizeof (lr_setattr_t);
+ void *start;
+
+
+ if (zilog == NULL || zp->z_unlinked)
+ return;
+
+ /*
+ * If XVATTR set, then log record size needs to allow
+ * for lr_attr_t + xvattr mask, mapsize and create time
+ * plus actual attribute values
+ */
+ if (vap->va_mask & AT_XVATTR)
+ recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
+
+ if (fuidp)
+ recsize += fuidp->z_domain_str_sz;
+
+ itx = zil_itx_create(txtype, recsize);
+ lr = (lr_setattr_t *)&itx->itx_lr;
+ lr->lr_foid = zp->z_id;
+ lr->lr_mask = (uint64_t)mask_applied;
+ lr->lr_mode = (uint64_t)vap->va_mode;
+ if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
+ lr->lr_uid = fuidp->z_fuid_owner;
+ else
+ lr->lr_uid = (uint64_t)vap->va_uid;
+
+ if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
+ lr->lr_gid = fuidp->z_fuid_group;
+ else
+ lr->lr_gid = (uint64_t)vap->va_gid;
+
+ lr->lr_size = (uint64_t)vap->va_size;
+ ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
+ ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
+ start = (lr_setattr_t *)(lr + 1);
+ if (vap->va_mask & AT_XVATTR) {
+ zfs_log_xvattr((lr_attr_t *)start, xvap);
+ start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
+ }
+
+ /*
+ * Now stick on domain information if any on end
+ */
+
+ if (fuidp)
+ (void) zfs_log_fuid_domains(fuidp, start);
+
+ itx->itx_sync = (zp->z_sync_cnt != 0);
+ seq = zil_itx_assign(zilog, itx, tx);
+ zp->z_last_itx = seq;
+}
+
+/*
+ * zfs_log_acl() handles TX_ACL transactions.
+ */
+void
+zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
+ vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
+{
+ itx_t *itx;
+ uint64_t seq;
+ lr_acl_v0_t *lrv0;
+ lr_acl_t *lr;
+ int txtype;
+ int lrsize;
+ size_t txsize;
+ size_t aclbytes = vsecp->vsa_aclentsz;
+
+ if (zilog == NULL || zp->z_unlinked)
+ return;
+
+ txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
+ TX_ACL_V0 : TX_ACL;
+
+ if (txtype == TX_ACL)
+ lrsize = sizeof (*lr);
+ else
+ lrsize = sizeof (*lrv0);
+
+ txsize = lrsize +
+ ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
+ (fuidp ? fuidp->z_domain_str_sz : 0) +
+ sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
+
+ itx = zil_itx_create(txtype, txsize);
+
+ lr = (lr_acl_t *)&itx->itx_lr;
+ lr->lr_foid = zp->z_id;
+ if (txtype == TX_ACL) {
+ lr->lr_acl_bytes = aclbytes;
+ lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
+ lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
+ if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
+ lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
+ else
+ lr->lr_acl_flags = 0;
+ }
+ lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
+
+ if (txtype == TX_ACL_V0) {
+ lrv0 = (lr_acl_v0_t *)lr;
+ bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
+ } else {
+ void *start = (ace_t *)(lr + 1);
+
+ bcopy(vsecp->vsa_aclentp, start, aclbytes);
+
+ start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
+
+ if (fuidp) {
+ start = zfs_log_fuid_ids(fuidp, start);
+ (void) zfs_log_fuid_domains(fuidp, start);
+ }
+ }
+
+ itx->itx_sync = (zp->z_sync_cnt != 0);
+ seq = zil_itx_assign(zilog, itx, tx);
+ zp->z_last_itx = seq;
+}
diff --git a/module/zfs/zfs_replay.c b/module/zfs/zfs_replay.c
new file mode 100644
index 000000000..85b79703a
--- /dev/null
+++ b/module/zfs/zfs_replay.c
@@ -0,0 +1,878 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/cmn_err.h>
+#include <sys/kmem.h>
+#include <sys/thread.h>
+#include <sys/file.h>
+#include <sys/fcntl.h>
+#include <sys/vfs.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_dir.h>
+#include <sys/zfs_acl.h>
+#include <sys/zfs_fuid.h>
+#include <sys/spa.h>
+#include <sys/zil.h>
+#include <sys/byteorder.h>
+#include <sys/stat.h>
+#include <sys/mode.h>
+#include <sys/acl.h>
+#include <sys/atomic.h>
+#include <sys/cred.h>
+
+/*
+ * Functions to replay ZFS intent log (ZIL) records
+ * The functions are called through a function vector (zfs_replay_vector)
+ * which is indexed by the transaction type.
+ */
+
+static void
+zfs_init_vattr(vattr_t *vap, uint64_t mask, uint64_t mode,
+ uint64_t uid, uint64_t gid, uint64_t rdev, uint64_t nodeid)
+{
+ bzero(vap, sizeof (*vap));
+ vap->va_mask = (uint_t)mask;
+ vap->va_type = IFTOVT(mode);
+ vap->va_mode = mode & MODEMASK;
+ vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
+ vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
+ vap->va_rdev = zfs_cmpldev(rdev);
+ vap->va_nodeid = nodeid;
+}
+
+/* ARGSUSED */
+static int
+zfs_replay_error(zfsvfs_t *zfsvfs, lr_t *lr, boolean_t byteswap)
+{
+ return (ENOTSUP);
+}
+
+static void
+zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
+{
+ xoptattr_t *xoap = NULL;
+ uint64_t *attrs;
+ uint64_t *crtime;
+ uint32_t *bitmap;
+ void *scanstamp;
+ int i;
+
+ xvap->xva_vattr.va_mask |= AT_XVATTR;
+ if ((xoap = xva_getxoptattr(xvap)) == NULL) {
+ xvap->xva_vattr.va_mask &= ~AT_XVATTR; /* shouldn't happen */
+ return;
+ }
+
+ ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
+
+ bitmap = &lrattr->lr_attr_bitmap;
+ for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
+ xvap->xva_reqattrmap[i] = *bitmap;
+
+ attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
+ crtime = attrs + 1;
+ scanstamp = (caddr_t)(crtime + 2);
+
+ if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
+ xoap->xoa_hidden = ((*attrs & XAT0_HIDDEN) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
+ xoap->xoa_system = ((*attrs & XAT0_SYSTEM) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
+ xoap->xoa_archive = ((*attrs & XAT0_ARCHIVE) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_READONLY))
+ xoap->xoa_readonly = ((*attrs & XAT0_READONLY) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
+ xoap->xoa_immutable = ((*attrs & XAT0_IMMUTABLE) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
+ xoap->xoa_nounlink = ((*attrs & XAT0_NOUNLINK) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
+ xoap->xoa_appendonly = ((*attrs & XAT0_APPENDONLY) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
+ xoap->xoa_nodump = ((*attrs & XAT0_NODUMP) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
+ xoap->xoa_opaque = ((*attrs & XAT0_OPAQUE) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
+ xoap->xoa_av_modified = ((*attrs & XAT0_AV_MODIFIED) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
+ xoap->xoa_av_quarantined =
+ ((*attrs & XAT0_AV_QUARANTINED) != 0);
+ if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
+ ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime);
+ if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
+ bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ);
+}
+
+static int
+zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
+{
+ uint64_t uid_idx;
+ uint64_t gid_idx;
+ int domcnt = 0;
+
+ uid_idx = FUID_INDEX(uid);
+ gid_idx = FUID_INDEX(gid);
+ if (uid_idx)
+ domcnt++;
+ if (gid_idx > 0 && gid_idx != uid_idx)
+ domcnt++;
+
+ return (domcnt);
+}
+
+static void *
+zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
+ int domcnt)
+{
+ int i;
+
+ for (i = 0; i != domcnt; i++) {
+ fuid_infop->z_domain_table[i] = start;
+ start = (caddr_t)start + strlen(start) + 1;
+ }
+
+ return (start);
+}
+
+/*
+ * Set the uid/gid in the fuid_info structure.
+ */
+static void
+zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
+{
+ /*
+ * If owner or group are log specific FUIDs then slurp up
+ * domain information and build zfs_fuid_info_t
+ */
+ if (IS_EPHEMERAL(uid))
+ fuid_infop->z_fuid_owner = uid;
+
+ if (IS_EPHEMERAL(gid))
+ fuid_infop->z_fuid_group = gid;
+}
+
+/*
+ * Load fuid domains into fuid_info_t
+ */
+static zfs_fuid_info_t *
+zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
+{
+ int domcnt;
+
+ zfs_fuid_info_t *fuid_infop;
+
+ fuid_infop = zfs_fuid_info_alloc();
+
+ domcnt = zfs_replay_domain_cnt(uid, gid);
+
+ if (domcnt == 0)
+ return (fuid_infop);
+
+ fuid_infop->z_domain_table =
+ kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
+
+ zfs_replay_fuid_ugid(fuid_infop, uid, gid);
+
+ fuid_infop->z_domain_cnt = domcnt;
+ *end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
+ return (fuid_infop);
+}
+
+/*
+ * load zfs_fuid_t's and fuid_domains into fuid_info_t
+ */
+static zfs_fuid_info_t *
+zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
+ uint64_t gid)
+{
+ uint64_t *log_fuid = (uint64_t *)start;
+ zfs_fuid_info_t *fuid_infop;
+ int i;
+
+ fuid_infop = zfs_fuid_info_alloc();
+ fuid_infop->z_domain_cnt = domcnt;
+
+ fuid_infop->z_domain_table =
+ kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
+
+ for (i = 0; i != idcnt; i++) {
+ zfs_fuid_t *zfuid;
+
+ zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
+ zfuid->z_logfuid = *log_fuid;
+ zfuid->z_id = -1;
+ zfuid->z_domidx = 0;
+ list_insert_tail(&fuid_infop->z_fuids, zfuid);
+ log_fuid++;
+ }
+
+ zfs_replay_fuid_ugid(fuid_infop, uid, gid);
+
+ *end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
+ return (fuid_infop);
+}
+
+static void
+zfs_replay_swap_attrs(lr_attr_t *lrattr)
+{
+ /* swap the lr_attr structure */
+ byteswap_uint32_array(lrattr, sizeof (*lrattr));
+ /* swap the bitmap */
+ byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) *
+ sizeof (uint32_t));
+ /* swap the attributes, create time + 64 bit word for attributes */
+ byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) *
+ (lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t));
+}
+
+/*
+ * Replay file create with optional ACL, xvattr information as well
+ * as option FUID information.
+ */
+static int
+zfs_replay_create_acl(zfsvfs_t *zfsvfs,
+ lr_acl_create_t *lracl, boolean_t byteswap)
+{
+ char *name = NULL; /* location determined later */
+ lr_create_t *lr = (lr_create_t *)lracl;
+ znode_t *dzp;
+ vnode_t *vp = NULL;
+ xvattr_t xva;
+ int vflg = 0;
+ vsecattr_t vsec = { 0 };
+ lr_attr_t *lrattr;
+ void *aclstart;
+ void *fuidstart;
+ size_t xvatlen = 0;
+ uint64_t txtype;
+ int error;
+
+ if (byteswap) {
+ byteswap_uint64_array(lracl, sizeof (*lracl));
+ txtype = (int)lr->lr_common.lrc_txtype;
+ if (txtype == TX_CREATE_ACL_ATTR ||
+ txtype == TX_MKDIR_ACL_ATTR) {
+ lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
+ zfs_replay_swap_attrs(lrattr);
+ xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
+ }
+
+ aclstart = (caddr_t)(lracl + 1) + xvatlen;
+ zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
+ /* swap fuids */
+ if (lracl->lr_fuidcnt) {
+ byteswap_uint64_array((caddr_t)aclstart +
+ ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
+ lracl->lr_fuidcnt * sizeof (uint64_t));
+ }
+ }
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
+ return (error);
+
+ xva_init(&xva);
+ zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
+ lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
+
+ /*
+ * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
+ * eventually end up in zfs_mknode(), which assigns the object's
+ * creation time and generation number. The generic VOP_CREATE()
+ * doesn't have either concept, so we smuggle the values inside
+ * the vattr's otherwise unused va_ctime and va_nblocks fields.
+ */
+ ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
+ xva.xva_vattr.va_nblocks = lr->lr_gen;
+
+ error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
+ if (error != ENOENT)
+ goto bail;
+
+ if (lr->lr_common.lrc_txtype & TX_CI)
+ vflg |= FIGNORECASE;
+ switch ((int)lr->lr_common.lrc_txtype) {
+ case TX_CREATE_ACL:
+ aclstart = (caddr_t)(lracl + 1);
+ fuidstart = (caddr_t)aclstart +
+ ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
+ zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
+ (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
+ lr->lr_uid, lr->lr_gid);
+ /*FALLTHROUGH*/
+ case TX_CREATE_ACL_ATTR:
+ if (name == NULL) {
+ lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
+ xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
+ xva.xva_vattr.va_mask |= AT_XVATTR;
+ zfs_replay_xvattr(lrattr, &xva);
+ }
+ vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
+ vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
+ vsec.vsa_aclcnt = lracl->lr_aclcnt;
+ vsec.vsa_aclentsz = lracl->lr_acl_bytes;
+ vsec.vsa_aclflags = lracl->lr_acl_flags;
+ if (zfsvfs->z_fuid_replay == NULL) {
+ fuidstart = (caddr_t)(lracl + 1) + xvatlen +
+ ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuids(fuidstart,
+ (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
+ lr->lr_uid, lr->lr_gid);
+ }
+
+ error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
+ 0, 0, &vp, kcred, vflg, NULL, &vsec);
+ break;
+ case TX_MKDIR_ACL:
+ aclstart = (caddr_t)(lracl + 1);
+ fuidstart = (caddr_t)aclstart +
+ ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
+ zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
+ (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
+ lr->lr_uid, lr->lr_gid);
+ /*FALLTHROUGH*/
+ case TX_MKDIR_ACL_ATTR:
+ if (name == NULL) {
+ lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
+ xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
+ zfs_replay_xvattr(lrattr, &xva);
+ }
+ vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
+ vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
+ vsec.vsa_aclcnt = lracl->lr_aclcnt;
+ vsec.vsa_aclentsz = lracl->lr_acl_bytes;
+ vsec.vsa_aclflags = lracl->lr_acl_flags;
+ if (zfsvfs->z_fuid_replay == NULL) {
+ fuidstart = (caddr_t)(lracl + 1) + xvatlen +
+ ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuids(fuidstart,
+ (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
+ lr->lr_uid, lr->lr_gid);
+ }
+ error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
+ &vp, kcred, NULL, vflg, &vsec);
+ break;
+ default:
+ error = ENOTSUP;
+ }
+
+bail:
+ if (error == 0 && vp != NULL)
+ VN_RELE(vp);
+
+ VN_RELE(ZTOV(dzp));
+
+ zfs_fuid_info_free(zfsvfs->z_fuid_replay);
+ zfsvfs->z_fuid_replay = NULL;
+
+ return (error);
+}
+
+static int
+zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap)
+{
+ char *name = NULL; /* location determined later */
+ char *link; /* symlink content follows name */
+ znode_t *dzp;
+ vnode_t *vp = NULL;
+ xvattr_t xva;
+ int vflg = 0;
+ size_t lrsize = sizeof (lr_create_t);
+ lr_attr_t *lrattr;
+ void *start;
+ size_t xvatlen;
+ uint64_t txtype;
+ int error;
+
+ if (byteswap) {
+ byteswap_uint64_array(lr, sizeof (*lr));
+ txtype = (int)lr->lr_common.lrc_txtype;
+ if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
+ zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
+ }
+
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
+ return (error);
+
+ xva_init(&xva);
+ zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
+ lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
+
+ /*
+ * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
+ * eventually end up in zfs_mknode(), which assigns the object's
+ * creation time and generation number. The generic VOP_CREATE()
+ * doesn't have either concept, so we smuggle the values inside
+ * the vattr's otherwise unused va_ctime and va_nblocks fields.
+ */
+ ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
+ xva.xva_vattr.va_nblocks = lr->lr_gen;
+
+ error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
+ if (error != ENOENT)
+ goto out;
+
+ if (lr->lr_common.lrc_txtype & TX_CI)
+ vflg |= FIGNORECASE;
+
+ /*
+ * Symlinks don't have fuid info, and CIFS never creates
+ * symlinks.
+ *
+ * The _ATTR versions will grab the fuid info in their subcases.
+ */
+ if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
+ (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
+ (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
+ start = (lr + 1);
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuid_domain(start, &start,
+ lr->lr_uid, lr->lr_gid);
+ }
+
+ switch ((int)lr->lr_common.lrc_txtype) {
+ case TX_CREATE_ATTR:
+ lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
+ xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
+ zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
+ start = (caddr_t)(lr + 1) + xvatlen;
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuid_domain(start, &start,
+ lr->lr_uid, lr->lr_gid);
+ name = (char *)start;
+
+ /*FALLTHROUGH*/
+ case TX_CREATE:
+ if (name == NULL)
+ name = (char *)start;
+
+ error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
+ 0, 0, &vp, kcred, vflg, NULL, NULL);
+ break;
+ case TX_MKDIR_ATTR:
+ lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
+ xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
+ zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
+ start = (caddr_t)(lr + 1) + xvatlen;
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuid_domain(start, &start,
+ lr->lr_uid, lr->lr_gid);
+ name = (char *)start;
+
+ /*FALLTHROUGH*/
+ case TX_MKDIR:
+ if (name == NULL)
+ name = (char *)(lr + 1);
+
+ error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
+ &vp, kcred, NULL, vflg, NULL);
+ break;
+ case TX_MKXATTR:
+ name = (char *)(lr + 1);
+ error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
+ break;
+ case TX_SYMLINK:
+ name = (char *)(lr + 1);
+ link = name + strlen(name) + 1;
+ error = VOP_SYMLINK(ZTOV(dzp), name, &xva.xva_vattr,
+ link, kcred, NULL, vflg);
+ break;
+ default:
+ error = ENOTSUP;
+ }
+
+out:
+ if (error == 0 && vp != NULL)
+ VN_RELE(vp);
+
+ VN_RELE(ZTOV(dzp));
+
+ if (zfsvfs->z_fuid_replay)
+ zfs_fuid_info_free(zfsvfs->z_fuid_replay);
+ zfsvfs->z_fuid_replay = NULL;
+ return (error);
+}
+
+static int
+zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap)
+{
+ char *name = (char *)(lr + 1); /* name follows lr_remove_t */
+ znode_t *dzp;
+ int error;
+ int vflg = 0;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
+ return (error);
+
+ if (lr->lr_common.lrc_txtype & TX_CI)
+ vflg |= FIGNORECASE;
+
+ switch ((int)lr->lr_common.lrc_txtype) {
+ case TX_REMOVE:
+ error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg);
+ break;
+ case TX_RMDIR:
+ error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg);
+ break;
+ default:
+ error = ENOTSUP;
+ }
+
+ VN_RELE(ZTOV(dzp));
+
+ return (error);
+}
+
+static int
+zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap)
+{
+ char *name = (char *)(lr + 1); /* name follows lr_link_t */
+ znode_t *dzp, *zp;
+ int error;
+ int vflg = 0;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
+ return (error);
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
+ VN_RELE(ZTOV(dzp));
+ return (error);
+ }
+
+ if (lr->lr_common.lrc_txtype & TX_CI)
+ vflg |= FIGNORECASE;
+
+ error = VOP_LINK(ZTOV(dzp), ZTOV(zp), name, kcred, NULL, vflg);
+
+ VN_RELE(ZTOV(zp));
+ VN_RELE(ZTOV(dzp));
+
+ return (error);
+}
+
+static int
+zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
+{
+ char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
+ char *tname = sname + strlen(sname) + 1;
+ znode_t *sdzp, *tdzp;
+ int error;
+ int vflg = 0;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
+ return (error);
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
+ VN_RELE(ZTOV(sdzp));
+ return (error);
+ }
+
+ if (lr->lr_common.lrc_txtype & TX_CI)
+ vflg |= FIGNORECASE;
+
+ error = VOP_RENAME(ZTOV(sdzp), sname, ZTOV(tdzp), tname, kcred,
+ NULL, vflg);
+
+ VN_RELE(ZTOV(tdzp));
+ VN_RELE(ZTOV(sdzp));
+
+ return (error);
+}
+
+static int
+zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
+{
+ char *data = (char *)(lr + 1); /* data follows lr_write_t */
+ znode_t *zp;
+ int error;
+ ssize_t resid;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
+ /*
+ * As we can log writes out of order, it's possible the
+ * file has been removed. In this case just drop the write
+ * and return success.
+ */
+ if (error == ENOENT)
+ error = 0;
+ return (error);
+ }
+
+ error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, lr->lr_length,
+ lr->lr_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
+
+ VN_RELE(ZTOV(zp));
+
+ return (error);
+}
+
+static int
+zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap)
+{
+ znode_t *zp;
+ flock64_t fl;
+ int error;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
+ /*
+ * As we can log truncates out of order, it's possible the
+ * file has been removed. In this case just drop the truncate
+ * and return success.
+ */
+ if (error == ENOENT)
+ error = 0;
+ return (error);
+ }
+
+ bzero(&fl, sizeof (fl));
+ fl.l_type = F_WRLCK;
+ fl.l_whence = 0;
+ fl.l_start = lr->lr_offset;
+ fl.l_len = lr->lr_length;
+
+ error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
+ lr->lr_offset, kcred, NULL);
+
+ VN_RELE(ZTOV(zp));
+
+ return (error);
+}
+
+static int
+zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap)
+{
+ znode_t *zp;
+ xvattr_t xva;
+ vattr_t *vap = &xva.xva_vattr;
+ int error;
+ void *start;
+
+ xva_init(&xva);
+ if (byteswap) {
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ if ((lr->lr_mask & AT_XVATTR) &&
+ zfsvfs->z_version >= ZPL_VERSION_INITIAL)
+ zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
+ }
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
+ /*
+ * As we can log setattrs out of order, it's possible the
+ * file has been removed. In this case just drop the setattr
+ * and return success.
+ */
+ if (error == ENOENT)
+ error = 0;
+ return (error);
+ }
+
+ zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
+ lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
+
+ vap->va_size = lr->lr_size;
+ ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
+ ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
+
+ /*
+ * Fill in xvattr_t portions if necessary.
+ */
+
+ start = (lr_setattr_t *)(lr + 1);
+ if (vap->va_mask & AT_XVATTR) {
+ zfs_replay_xvattr((lr_attr_t *)start, &xva);
+ start = (caddr_t)start +
+ ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
+ } else
+ xva.xva_vattr.va_mask &= ~AT_XVATTR;
+
+ zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
+ lr->lr_uid, lr->lr_gid);
+
+ error = VOP_SETATTR(ZTOV(zp), vap, 0, kcred, NULL);
+
+ zfs_fuid_info_free(zfsvfs->z_fuid_replay);
+ zfsvfs->z_fuid_replay = NULL;
+ VN_RELE(ZTOV(zp));
+
+ return (error);
+}
+
+static int
+zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap)
+{
+ ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */
+ vsecattr_t vsa;
+ znode_t *zp;
+ int error;
+
+ if (byteswap) {
+ byteswap_uint64_array(lr, sizeof (*lr));
+ zfs_oldace_byteswap(ace, lr->lr_aclcnt);
+ }
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
+ /*
+ * As we can log acls out of order, it's possible the
+ * file has been removed. In this case just drop the acl
+ * and return success.
+ */
+ if (error == ENOENT)
+ error = 0;
+ return (error);
+ }
+
+ bzero(&vsa, sizeof (vsa));
+ vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
+ vsa.vsa_aclcnt = lr->lr_aclcnt;
+ vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
+ vsa.vsa_aclflags = 0;
+ vsa.vsa_aclentp = ace;
+
+ error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
+
+ VN_RELE(ZTOV(zp));
+
+ return (error);
+}
+
+/*
+ * Replaying ACLs is complicated by FUID support.
+ * The log record may contain some optional data
+ * to be used for replaying FUID's. These pieces
+ * are the actual FUIDs that were created initially.
+ * The FUID table index may no longer be valid and
+ * during zfs_create() a new index may be assigned.
+ * Because of this the log will contain the original
+ * doman+rid in order to create a new FUID.
+ *
+ * The individual ACEs may contain an ephemeral uid/gid which is no
+ * longer valid and will need to be replaced with an actual FUID.
+ *
+ */
+static int
+zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap)
+{
+ ace_t *ace = (ace_t *)(lr + 1);
+ vsecattr_t vsa;
+ znode_t *zp;
+ int error;
+
+ if (byteswap) {
+ byteswap_uint64_array(lr, sizeof (*lr));
+ zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
+ if (lr->lr_fuidcnt) {
+ byteswap_uint64_array((caddr_t)ace +
+ ZIL_ACE_LENGTH(lr->lr_acl_bytes),
+ lr->lr_fuidcnt * sizeof (uint64_t));
+ }
+ }
+
+ if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
+ /*
+ * As we can log acls out of order, it's possible the
+ * file has been removed. In this case just drop the acl
+ * and return success.
+ */
+ if (error == ENOENT)
+ error = 0;
+ return (error);
+ }
+
+ bzero(&vsa, sizeof (vsa));
+ vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
+ vsa.vsa_aclcnt = lr->lr_aclcnt;
+ vsa.vsa_aclentp = ace;
+ vsa.vsa_aclentsz = lr->lr_acl_bytes;
+ vsa.vsa_aclflags = lr->lr_acl_flags;
+
+ if (lr->lr_fuidcnt) {
+ void *fuidstart = (caddr_t)ace +
+ ZIL_ACE_LENGTH(lr->lr_acl_bytes);
+
+ zfsvfs->z_fuid_replay =
+ zfs_replay_fuids(fuidstart, &fuidstart,
+ lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
+ }
+
+ error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
+
+ if (zfsvfs->z_fuid_replay)
+ zfs_fuid_info_free(zfsvfs->z_fuid_replay);
+
+ zfsvfs->z_fuid_replay = NULL;
+ VN_RELE(ZTOV(zp));
+
+ return (error);
+}
+
+/*
+ * Callback vectors for replaying records
+ */
+zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
+ zfs_replay_error, /* 0 no such transaction type */
+ zfs_replay_create, /* TX_CREATE */
+ zfs_replay_create, /* TX_MKDIR */
+ zfs_replay_create, /* TX_MKXATTR */
+ zfs_replay_create, /* TX_SYMLINK */
+ zfs_replay_remove, /* TX_REMOVE */
+ zfs_replay_remove, /* TX_RMDIR */
+ zfs_replay_link, /* TX_LINK */
+ zfs_replay_rename, /* TX_RENAME */
+ zfs_replay_write, /* TX_WRITE */
+ zfs_replay_truncate, /* TX_TRUNCATE */
+ zfs_replay_setattr, /* TX_SETATTR */
+ zfs_replay_acl_v0, /* TX_ACL_V0 */
+ zfs_replay_acl, /* TX_ACL */
+ zfs_replay_create_acl, /* TX_CREATE_ACL */
+ zfs_replay_create, /* TX_CREATE_ATTR */
+ zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */
+ zfs_replay_create_acl, /* TX_MKDIR_ACL */
+ zfs_replay_create, /* TX_MKDIR_ATTR */
+ zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */
+};
diff --git a/module/zfs/zfs_rlock.c b/module/zfs/zfs_rlock.c
new file mode 100644
index 000000000..f0a75b5fa
--- /dev/null
+++ b/module/zfs/zfs_rlock.c
@@ -0,0 +1,602 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+/*
+ * This file contains the code to implement file range locking in
+ * ZFS, although there isn't much specific to ZFS (all that comes to mind
+ * support for growing the blocksize).
+ *
+ * Interface
+ * ---------
+ * Defined in zfs_rlock.h but essentially:
+ * rl = zfs_range_lock(zp, off, len, lock_type);
+ * zfs_range_unlock(rl);
+ * zfs_range_reduce(rl, off, len);
+ *
+ * AVL tree
+ * --------
+ * An AVL tree is used to maintain the state of the existing ranges
+ * that are locked for exclusive (writer) or shared (reader) use.
+ * The starting range offset is used for searching and sorting the tree.
+ *
+ * Common case
+ * -----------
+ * The (hopefully) usual case is of no overlaps or contention for
+ * locks. On entry to zfs_lock_range() a rl_t is allocated; the tree
+ * searched that finds no overlap, and *this* rl_t is placed in the tree.
+ *
+ * Overlaps/Reference counting/Proxy locks
+ * ---------------------------------------
+ * The avl code only allows one node at a particular offset. Also it's very
+ * inefficient to search through all previous entries looking for overlaps
+ * (because the very 1st in the ordered list might be at offset 0 but
+ * cover the whole file).
+ * So this implementation uses reference counts and proxy range locks.
+ * Firstly, only reader locks use reference counts and proxy locks,
+ * because writer locks are exclusive.
+ * When a reader lock overlaps with another then a proxy lock is created
+ * for that range and replaces the original lock. If the overlap
+ * is exact then the reference count of the proxy is simply incremented.
+ * Otherwise, the proxy lock is split into smaller lock ranges and
+ * new proxy locks created for non overlapping ranges.
+ * The reference counts are adjusted accordingly.
+ * Meanwhile, the orginal lock is kept around (this is the callers handle)
+ * and its offset and length are used when releasing the lock.
+ *
+ * Thread coordination
+ * -------------------
+ * In order to make wakeups efficient and to ensure multiple continuous
+ * readers on a range don't starve a writer for the same range lock,
+ * two condition variables are allocated in each rl_t.
+ * If a writer (or reader) can't get a range it initialises the writer
+ * (or reader) cv; sets a flag saying there's a writer (or reader) waiting;
+ * and waits on that cv. When a thread unlocks that range it wakes up all
+ * writers then all readers before destroying the lock.
+ *
+ * Append mode writes
+ * ------------------
+ * Append mode writes need to lock a range at the end of a file.
+ * The offset of the end of the file is determined under the
+ * range locking mutex, and the lock type converted from RL_APPEND to
+ * RL_WRITER and the range locked.
+ *
+ * Grow block handling
+ * -------------------
+ * ZFS supports multiple block sizes currently upto 128K. The smallest
+ * block size is used for the file which is grown as needed. During this
+ * growth all other writers and readers must be excluded.
+ * So if the block size needs to be grown then the whole file is
+ * exclusively locked, then later the caller will reduce the lock
+ * range to just the range to be written using zfs_reduce_range.
+ */
+
+#include <sys/zfs_rlock.h>
+
+/*
+ * Check if a write lock can be grabbed, or wait and recheck until available.
+ */
+static void
+zfs_range_lock_writer(znode_t *zp, rl_t *new)
+{
+ avl_tree_t *tree = &zp->z_range_avl;
+ rl_t *rl;
+ avl_index_t where;
+ uint64_t end_size;
+ uint64_t off = new->r_off;
+ uint64_t len = new->r_len;
+
+ for (;;) {
+ /*
+ * Range locking is also used by zvol and uses a
+ * dummied up znode. However, for zvol, we don't need to
+ * append or grow blocksize, and besides we don't have
+ * a z_phys or z_zfsvfs - so skip that processing.
+ *
+ * Yes, this is ugly, and would be solved by not handling
+ * grow or append in range lock code. If that was done then
+ * we could make the range locking code generically available
+ * to other non-zfs consumers.
+ */
+ if (zp->z_vnode) { /* caller is ZPL */
+ /*
+ * If in append mode pick up the current end of file.
+ * This is done under z_range_lock to avoid races.
+ */
+ if (new->r_type == RL_APPEND)
+ new->r_off = zp->z_phys->zp_size;
+
+ /*
+ * If we need to grow the block size then grab the whole
+ * file range. This is also done under z_range_lock to
+ * avoid races.
+ */
+ end_size = MAX(zp->z_phys->zp_size, new->r_off + len);
+ if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
+ zp->z_blksz < zp->z_zfsvfs->z_max_blksz)) {
+ new->r_off = 0;
+ new->r_len = UINT64_MAX;
+ }
+ }
+
+ /*
+ * First check for the usual case of no locks
+ */
+ if (avl_numnodes(tree) == 0) {
+ new->r_type = RL_WRITER; /* convert to writer */
+ avl_add(tree, new);
+ return;
+ }
+
+ /*
+ * Look for any locks in the range.
+ */
+ rl = avl_find(tree, new, &where);
+ if (rl)
+ goto wait; /* already locked at same offset */
+
+ rl = (rl_t *)avl_nearest(tree, where, AVL_AFTER);
+ if (rl && (rl->r_off < new->r_off + new->r_len))
+ goto wait;
+
+ rl = (rl_t *)avl_nearest(tree, where, AVL_BEFORE);
+ if (rl && rl->r_off + rl->r_len > new->r_off)
+ goto wait;
+
+ new->r_type = RL_WRITER; /* convert possible RL_APPEND */
+ avl_insert(tree, new, where);
+ return;
+wait:
+ if (!rl->r_write_wanted) {
+ cv_init(&rl->r_wr_cv, NULL, CV_DEFAULT, NULL);
+ rl->r_write_wanted = B_TRUE;
+ }
+ cv_wait(&rl->r_wr_cv, &zp->z_range_lock);
+
+ /* reset to original */
+ new->r_off = off;
+ new->r_len = len;
+ }
+}
+
+/*
+ * If this is an original (non-proxy) lock then replace it by
+ * a proxy and return the proxy.
+ */
+static rl_t *
+zfs_range_proxify(avl_tree_t *tree, rl_t *rl)
+{
+ rl_t *proxy;
+
+ if (rl->r_proxy)
+ return (rl); /* already a proxy */
+
+ ASSERT3U(rl->r_cnt, ==, 1);
+ ASSERT(rl->r_write_wanted == B_FALSE);
+ ASSERT(rl->r_read_wanted == B_FALSE);
+ avl_remove(tree, rl);
+ rl->r_cnt = 0;
+
+ /* create a proxy range lock */
+ proxy = kmem_alloc(sizeof (rl_t), KM_SLEEP);
+ proxy->r_off = rl->r_off;
+ proxy->r_len = rl->r_len;
+ proxy->r_cnt = 1;
+ proxy->r_type = RL_READER;
+ proxy->r_proxy = B_TRUE;
+ proxy->r_write_wanted = B_FALSE;
+ proxy->r_read_wanted = B_FALSE;
+ avl_add(tree, proxy);
+
+ return (proxy);
+}
+
+/*
+ * Split the range lock at the supplied offset
+ * returning the *front* proxy.
+ */
+static rl_t *
+zfs_range_split(avl_tree_t *tree, rl_t *rl, uint64_t off)
+{
+ rl_t *front, *rear;
+
+ ASSERT3U(rl->r_len, >, 1);
+ ASSERT3U(off, >, rl->r_off);
+ ASSERT3U(off, <, rl->r_off + rl->r_len);
+ ASSERT(rl->r_write_wanted == B_FALSE);
+ ASSERT(rl->r_read_wanted == B_FALSE);
+
+ /* create the rear proxy range lock */
+ rear = kmem_alloc(sizeof (rl_t), KM_SLEEP);
+ rear->r_off = off;
+ rear->r_len = rl->r_off + rl->r_len - off;
+ rear->r_cnt = rl->r_cnt;
+ rear->r_type = RL_READER;
+ rear->r_proxy = B_TRUE;
+ rear->r_write_wanted = B_FALSE;
+ rear->r_read_wanted = B_FALSE;
+
+ front = zfs_range_proxify(tree, rl);
+ front->r_len = off - rl->r_off;
+
+ avl_insert_here(tree, rear, front, AVL_AFTER);
+ return (front);
+}
+
+/*
+ * Create and add a new proxy range lock for the supplied range.
+ */
+static void
+zfs_range_new_proxy(avl_tree_t *tree, uint64_t off, uint64_t len)
+{
+ rl_t *rl;
+
+ ASSERT(len);
+ rl = kmem_alloc(sizeof (rl_t), KM_SLEEP);
+ rl->r_off = off;
+ rl->r_len = len;
+ rl->r_cnt = 1;
+ rl->r_type = RL_READER;
+ rl->r_proxy = B_TRUE;
+ rl->r_write_wanted = B_FALSE;
+ rl->r_read_wanted = B_FALSE;
+ avl_add(tree, rl);
+}
+
+static void
+zfs_range_add_reader(avl_tree_t *tree, rl_t *new, rl_t *prev, avl_index_t where)
+{
+ rl_t *next;
+ uint64_t off = new->r_off;
+ uint64_t len = new->r_len;
+
+ /*
+ * prev arrives either:
+ * - pointing to an entry at the same offset
+ * - pointing to the entry with the closest previous offset whose
+ * range may overlap with the new range
+ * - null, if there were no ranges starting before the new one
+ */
+ if (prev) {
+ if (prev->r_off + prev->r_len <= off) {
+ prev = NULL;
+ } else if (prev->r_off != off) {
+ /*
+ * convert to proxy if needed then
+ * split this entry and bump ref count
+ */
+ prev = zfs_range_split(tree, prev, off);
+ prev = AVL_NEXT(tree, prev); /* move to rear range */
+ }
+ }
+ ASSERT((prev == NULL) || (prev->r_off == off));
+
+ if (prev)
+ next = prev;
+ else
+ next = (rl_t *)avl_nearest(tree, where, AVL_AFTER);
+
+ if (next == NULL || off + len <= next->r_off) {
+ /* no overlaps, use the original new rl_t in the tree */
+ avl_insert(tree, new, where);
+ return;
+ }
+
+ if (off < next->r_off) {
+ /* Add a proxy for initial range before the overlap */
+ zfs_range_new_proxy(tree, off, next->r_off - off);
+ }
+
+ new->r_cnt = 0; /* will use proxies in tree */
+ /*
+ * We now search forward through the ranges, until we go past the end
+ * of the new range. For each entry we make it a proxy if it
+ * isn't already, then bump its reference count. If there's any
+ * gaps between the ranges then we create a new proxy range.
+ */
+ for (prev = NULL; next; prev = next, next = AVL_NEXT(tree, next)) {
+ if (off + len <= next->r_off)
+ break;
+ if (prev && prev->r_off + prev->r_len < next->r_off) {
+ /* there's a gap */
+ ASSERT3U(next->r_off, >, prev->r_off + prev->r_len);
+ zfs_range_new_proxy(tree, prev->r_off + prev->r_len,
+ next->r_off - (prev->r_off + prev->r_len));
+ }
+ if (off + len == next->r_off + next->r_len) {
+ /* exact overlap with end */
+ next = zfs_range_proxify(tree, next);
+ next->r_cnt++;
+ return;
+ }
+ if (off + len < next->r_off + next->r_len) {
+ /* new range ends in the middle of this block */
+ next = zfs_range_split(tree, next, off + len);
+ next->r_cnt++;
+ return;
+ }
+ ASSERT3U(off + len, >, next->r_off + next->r_len);
+ next = zfs_range_proxify(tree, next);
+ next->r_cnt++;
+ }
+
+ /* Add the remaining end range. */
+ zfs_range_new_proxy(tree, prev->r_off + prev->r_len,
+ (off + len) - (prev->r_off + prev->r_len));
+}
+
+/*
+ * Check if a reader lock can be grabbed, or wait and recheck until available.
+ */
+static void
+zfs_range_lock_reader(znode_t *zp, rl_t *new)
+{
+ avl_tree_t *tree = &zp->z_range_avl;
+ rl_t *prev, *next;
+ avl_index_t where;
+ uint64_t off = new->r_off;
+ uint64_t len = new->r_len;
+
+ /*
+ * Look for any writer locks in the range.
+ */
+retry:
+ prev = avl_find(tree, new, &where);
+ if (prev == NULL)
+ prev = (rl_t *)avl_nearest(tree, where, AVL_BEFORE);
+
+ /*
+ * Check the previous range for a writer lock overlap.
+ */
+ if (prev && (off < prev->r_off + prev->r_len)) {
+ if ((prev->r_type == RL_WRITER) || (prev->r_write_wanted)) {
+ if (!prev->r_read_wanted) {
+ cv_init(&prev->r_rd_cv, NULL, CV_DEFAULT, NULL);
+ prev->r_read_wanted = B_TRUE;
+ }
+ cv_wait(&prev->r_rd_cv, &zp->z_range_lock);
+ goto retry;
+ }
+ if (off + len < prev->r_off + prev->r_len)
+ goto got_lock;
+ }
+
+ /*
+ * Search through the following ranges to see if there's
+ * write lock any overlap.
+ */
+ if (prev)
+ next = AVL_NEXT(tree, prev);
+ else
+ next = (rl_t *)avl_nearest(tree, where, AVL_AFTER);
+ for (; next; next = AVL_NEXT(tree, next)) {
+ if (off + len <= next->r_off)
+ goto got_lock;
+ if ((next->r_type == RL_WRITER) || (next->r_write_wanted)) {
+ if (!next->r_read_wanted) {
+ cv_init(&next->r_rd_cv, NULL, CV_DEFAULT, NULL);
+ next->r_read_wanted = B_TRUE;
+ }
+ cv_wait(&next->r_rd_cv, &zp->z_range_lock);
+ goto retry;
+ }
+ if (off + len <= next->r_off + next->r_len)
+ goto got_lock;
+ }
+
+got_lock:
+ /*
+ * Add the read lock, which may involve splitting existing
+ * locks and bumping ref counts (r_cnt).
+ */
+ zfs_range_add_reader(tree, new, prev, where);
+}
+
+/*
+ * Lock a range (offset, length) as either shared (RL_READER)
+ * or exclusive (RL_WRITER). Returns the range lock structure
+ * for later unlocking or reduce range (if entire file
+ * previously locked as RL_WRITER).
+ */
+rl_t *
+zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type)
+{
+ rl_t *new;
+
+ ASSERT(type == RL_READER || type == RL_WRITER || type == RL_APPEND);
+
+ new = kmem_alloc(sizeof (rl_t), KM_SLEEP);
+ new->r_zp = zp;
+ new->r_off = off;
+ new->r_len = len;
+ new->r_cnt = 1; /* assume it's going to be in the tree */
+ new->r_type = type;
+ new->r_proxy = B_FALSE;
+ new->r_write_wanted = B_FALSE;
+ new->r_read_wanted = B_FALSE;
+
+ mutex_enter(&zp->z_range_lock);
+ if (type == RL_READER) {
+ /*
+ * First check for the usual case of no locks
+ */
+ if (avl_numnodes(&zp->z_range_avl) == 0)
+ avl_add(&zp->z_range_avl, new);
+ else
+ zfs_range_lock_reader(zp, new);
+ } else
+ zfs_range_lock_writer(zp, new); /* RL_WRITER or RL_APPEND */
+ mutex_exit(&zp->z_range_lock);
+ return (new);
+}
+
+/*
+ * Unlock a reader lock
+ */
+static void
+zfs_range_unlock_reader(znode_t *zp, rl_t *remove)
+{
+ avl_tree_t *tree = &zp->z_range_avl;
+ rl_t *rl, *next;
+ uint64_t len;
+
+ /*
+ * The common case is when the remove entry is in the tree
+ * (cnt == 1) meaning there's been no other reader locks overlapping
+ * with this one. Otherwise the remove entry will have been
+ * removed from the tree and replaced by proxies (one or
+ * more ranges mapping to the entire range).
+ */
+ if (remove->r_cnt == 1) {
+ avl_remove(tree, remove);
+ if (remove->r_write_wanted) {
+ cv_broadcast(&remove->r_wr_cv);
+ cv_destroy(&remove->r_wr_cv);
+ }
+ if (remove->r_read_wanted) {
+ cv_broadcast(&remove->r_rd_cv);
+ cv_destroy(&remove->r_rd_cv);
+ }
+ } else {
+ ASSERT3U(remove->r_cnt, ==, 0);
+ ASSERT3U(remove->r_write_wanted, ==, 0);
+ ASSERT3U(remove->r_read_wanted, ==, 0);
+ /*
+ * Find start proxy representing this reader lock,
+ * then decrement ref count on all proxies
+ * that make up this range, freeing them as needed.
+ */
+ rl = avl_find(tree, remove, NULL);
+ ASSERT(rl);
+ ASSERT(rl->r_cnt);
+ ASSERT(rl->r_type == RL_READER);
+ for (len = remove->r_len; len != 0; rl = next) {
+ len -= rl->r_len;
+ if (len) {
+ next = AVL_NEXT(tree, rl);
+ ASSERT(next);
+ ASSERT(rl->r_off + rl->r_len == next->r_off);
+ ASSERT(next->r_cnt);
+ ASSERT(next->r_type == RL_READER);
+ }
+ rl->r_cnt--;
+ if (rl->r_cnt == 0) {
+ avl_remove(tree, rl);
+ if (rl->r_write_wanted) {
+ cv_broadcast(&rl->r_wr_cv);
+ cv_destroy(&rl->r_wr_cv);
+ }
+ if (rl->r_read_wanted) {
+ cv_broadcast(&rl->r_rd_cv);
+ cv_destroy(&rl->r_rd_cv);
+ }
+ kmem_free(rl, sizeof (rl_t));
+ }
+ }
+ }
+ kmem_free(remove, sizeof (rl_t));
+}
+
+/*
+ * Unlock range and destroy range lock structure.
+ */
+void
+zfs_range_unlock(rl_t *rl)
+{
+ znode_t *zp = rl->r_zp;
+
+ ASSERT(rl->r_type == RL_WRITER || rl->r_type == RL_READER);
+ ASSERT(rl->r_cnt == 1 || rl->r_cnt == 0);
+ ASSERT(!rl->r_proxy);
+
+ mutex_enter(&zp->z_range_lock);
+ if (rl->r_type == RL_WRITER) {
+ /* writer locks can't be shared or split */
+ avl_remove(&zp->z_range_avl, rl);
+ mutex_exit(&zp->z_range_lock);
+ if (rl->r_write_wanted) {
+ cv_broadcast(&rl->r_wr_cv);
+ cv_destroy(&rl->r_wr_cv);
+ }
+ if (rl->r_read_wanted) {
+ cv_broadcast(&rl->r_rd_cv);
+ cv_destroy(&rl->r_rd_cv);
+ }
+ kmem_free(rl, sizeof (rl_t));
+ } else {
+ /*
+ * lock may be shared, let zfs_range_unlock_reader()
+ * release the lock and free the rl_t
+ */
+ zfs_range_unlock_reader(zp, rl);
+ mutex_exit(&zp->z_range_lock);
+ }
+}
+
+/*
+ * Reduce range locked as RL_WRITER from whole file to specified range.
+ * Asserts the whole file is exclusivly locked and so there's only one
+ * entry in the tree.
+ */
+void
+zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len)
+{
+ znode_t *zp = rl->r_zp;
+
+ /* Ensure there are no other locks */
+ ASSERT(avl_numnodes(&zp->z_range_avl) == 1);
+ ASSERT(rl->r_off == 0);
+ ASSERT(rl->r_type == RL_WRITER);
+ ASSERT(!rl->r_proxy);
+ ASSERT3U(rl->r_len, ==, UINT64_MAX);
+ ASSERT3U(rl->r_cnt, ==, 1);
+
+ mutex_enter(&zp->z_range_lock);
+ rl->r_off = off;
+ rl->r_len = len;
+ mutex_exit(&zp->z_range_lock);
+ if (rl->r_write_wanted)
+ cv_broadcast(&rl->r_wr_cv);
+ if (rl->r_read_wanted)
+ cv_broadcast(&rl->r_rd_cv);
+}
+
+/*
+ * AVL comparison function used to order range locks
+ * Locks are ordered on the start offset of the range.
+ */
+int
+zfs_range_compare(const void *arg1, const void *arg2)
+{
+ const rl_t *rl1 = arg1;
+ const rl_t *rl2 = arg2;
+
+ if (rl1->r_off > rl2->r_off)
+ return (1);
+ if (rl1->r_off < rl2->r_off)
+ return (-1);
+ return (0);
+}
diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c
new file mode 100644
index 000000000..06b4dee46
--- /dev/null
+++ b/module/zfs/zfs_vfsops.c
@@ -0,0 +1,1652 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/kmem.h>
+#include <sys/pathname.h>
+#include <sys/vnode.h>
+#include <sys/vfs.h>
+#include <sys/vfs_opreg.h>
+#include <sys/mntent.h>
+#include <sys/mount.h>
+#include <sys/cmn_err.h>
+#include "fs/fs_subr.h"
+#include <sys/zfs_znode.h>
+#include <sys/zfs_dir.h>
+#include <sys/zil.h>
+#include <sys/fs/zfs.h>
+#include <sys/dmu.h>
+#include <sys/dsl_prop.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_deleg.h>
+#include <sys/spa.h>
+#include <sys/zap.h>
+#include <sys/varargs.h>
+#include <sys/policy.h>
+#include <sys/atomic.h>
+#include <sys/mkdev.h>
+#include <sys/modctl.h>
+#include <sys/refstr.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zfs_ctldir.h>
+#include <sys/zfs_fuid.h>
+#include <sys/bootconf.h>
+#include <sys/sunddi.h>
+#include <sys/dnlc.h>
+#include <sys/dmu_objset.h>
+#include <sys/spa_boot.h>
+
+int zfsfstype;
+vfsops_t *zfs_vfsops = NULL;
+static major_t zfs_major;
+static minor_t zfs_minor;
+static kmutex_t zfs_dev_mtx;
+
+static int zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr);
+static int zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr);
+static int zfs_mountroot(vfs_t *vfsp, enum whymountroot);
+static int zfs_root(vfs_t *vfsp, vnode_t **vpp);
+static int zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp);
+static int zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp);
+static void zfs_freevfs(vfs_t *vfsp);
+
+static const fs_operation_def_t zfs_vfsops_template[] = {
+ VFSNAME_MOUNT, { .vfs_mount = zfs_mount },
+ VFSNAME_MOUNTROOT, { .vfs_mountroot = zfs_mountroot },
+ VFSNAME_UNMOUNT, { .vfs_unmount = zfs_umount },
+ VFSNAME_ROOT, { .vfs_root = zfs_root },
+ VFSNAME_STATVFS, { .vfs_statvfs = zfs_statvfs },
+ VFSNAME_SYNC, { .vfs_sync = zfs_sync },
+ VFSNAME_VGET, { .vfs_vget = zfs_vget },
+ VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs },
+ NULL, NULL
+};
+
+static const fs_operation_def_t zfs_vfsops_eio_template[] = {
+ VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs },
+ NULL, NULL
+};
+
+/*
+ * We need to keep a count of active fs's.
+ * This is necessary to prevent our module
+ * from being unloaded after a umount -f
+ */
+static uint32_t zfs_active_fs_count = 0;
+
+static char *noatime_cancel[] = { MNTOPT_ATIME, NULL };
+static char *atime_cancel[] = { MNTOPT_NOATIME, NULL };
+static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
+static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
+
+/*
+ * MO_DEFAULT is not used since the default value is determined
+ * by the equivalent property.
+ */
+static mntopt_t mntopts[] = {
+ { MNTOPT_NOXATTR, noxattr_cancel, NULL, 0, NULL },
+ { MNTOPT_XATTR, xattr_cancel, NULL, 0, NULL },
+ { MNTOPT_NOATIME, noatime_cancel, NULL, 0, NULL },
+ { MNTOPT_ATIME, atime_cancel, NULL, 0, NULL }
+};
+
+static mntopts_t zfs_mntopts = {
+ sizeof (mntopts) / sizeof (mntopt_t),
+ mntopts
+};
+
+/*ARGSUSED*/
+int
+zfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
+{
+ /*
+ * Data integrity is job one. We don't want a compromised kernel
+ * writing to the storage pool, so we never sync during panic.
+ */
+ if (panicstr)
+ return (0);
+
+ /*
+ * SYNC_ATTR is used by fsflush() to force old filesystems like UFS
+ * to sync metadata, which they would otherwise cache indefinitely.
+ * Semantically, the only requirement is that the sync be initiated.
+ * The DMU syncs out txgs frequently, so there's nothing to do.
+ */
+ if (flag & SYNC_ATTR)
+ return (0);
+
+ if (vfsp != NULL) {
+ /*
+ * Sync a specific filesystem.
+ */
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+
+ ZFS_ENTER(zfsvfs);
+ if (zfsvfs->z_log != NULL)
+ zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
+ else
+ txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
+ ZFS_EXIT(zfsvfs);
+ } else {
+ /*
+ * Sync all ZFS filesystems. This is what happens when you
+ * run sync(1M). Unlike other filesystems, ZFS honors the
+ * request by waiting for all pools to commit all dirty data.
+ */
+ spa_sync_allpools();
+ }
+
+ return (0);
+}
+
+static int
+zfs_create_unique_device(dev_t *dev)
+{
+ major_t new_major;
+
+ do {
+ ASSERT3U(zfs_minor, <=, MAXMIN32);
+ minor_t start = zfs_minor;
+ do {
+ mutex_enter(&zfs_dev_mtx);
+ if (zfs_minor >= MAXMIN32) {
+ /*
+ * If we're still using the real major
+ * keep out of /dev/zfs and /dev/zvol minor
+ * number space. If we're using a getudev()'ed
+ * major number, we can use all of its minors.
+ */
+ if (zfs_major == ddi_name_to_major(ZFS_DRIVER))
+ zfs_minor = ZFS_MIN_MINOR;
+ else
+ zfs_minor = 0;
+ } else {
+ zfs_minor++;
+ }
+ *dev = makedevice(zfs_major, zfs_minor);
+ mutex_exit(&zfs_dev_mtx);
+ } while (vfs_devismounted(*dev) && zfs_minor != start);
+ if (zfs_minor == start) {
+ /*
+ * We are using all ~262,000 minor numbers for the
+ * current major number. Create a new major number.
+ */
+ if ((new_major = getudev()) == (major_t)-1) {
+ cmn_err(CE_WARN,
+ "zfs_mount: Can't get unique major "
+ "device number.");
+ return (-1);
+ }
+ mutex_enter(&zfs_dev_mtx);
+ zfs_major = new_major;
+ zfs_minor = 0;
+
+ mutex_exit(&zfs_dev_mtx);
+ } else {
+ break;
+ }
+ /* CONSTANTCONDITION */
+ } while (1);
+
+ return (0);
+}
+
+static void
+atime_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval == TRUE) {
+ zfsvfs->z_atime = TRUE;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0);
+ } else {
+ zfsvfs->z_atime = FALSE;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0);
+ }
+}
+
+static void
+xattr_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval == TRUE) {
+ /* XXX locking on vfs_flag? */
+ zfsvfs->z_vfs->vfs_flag |= VFS_XATTR;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0);
+ } else {
+ /* XXX locking on vfs_flag? */
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0);
+ }
+}
+
+static void
+blksz_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval < SPA_MINBLOCKSIZE ||
+ newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
+ newval = SPA_MAXBLOCKSIZE;
+
+ zfsvfs->z_max_blksz = newval;
+ zfsvfs->z_vfs->vfs_bsize = newval;
+}
+
+static void
+readonly_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval) {
+ /* XXX locking on vfs_flag? */
+ zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0);
+ } else {
+ /* XXX locking on vfs_flag? */
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0);
+ }
+}
+
+static void
+devices_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval == FALSE) {
+ zfsvfs->z_vfs->vfs_flag |= VFS_NODEVICES;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES, NULL, 0);
+ } else {
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_NODEVICES;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES, NULL, 0);
+ }
+}
+
+static void
+setuid_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval == FALSE) {
+ zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0);
+ } else {
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0);
+ }
+}
+
+static void
+exec_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ if (newval == FALSE) {
+ zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0);
+ } else {
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC;
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0);
+ }
+}
+
+/*
+ * The nbmand mount option can be changed at mount time.
+ * We can't allow it to be toggled on live file systems or incorrect
+ * behavior may be seen from cifs clients
+ *
+ * This property isn't registered via dsl_prop_register(), but this callback
+ * will be called when a file system is first mounted
+ */
+static void
+nbmand_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+ if (newval == FALSE) {
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0);
+ } else {
+ vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND);
+ vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0);
+ }
+}
+
+static void
+snapdir_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ zfsvfs->z_show_ctldir = newval;
+}
+
+static void
+vscan_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ zfsvfs->z_vscan = newval;
+}
+
+static void
+acl_mode_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ zfsvfs->z_acl_mode = newval;
+}
+
+static void
+acl_inherit_changed_cb(void *arg, uint64_t newval)
+{
+ zfsvfs_t *zfsvfs = arg;
+
+ zfsvfs->z_acl_inherit = newval;
+}
+
+static int
+zfs_register_callbacks(vfs_t *vfsp)
+{
+ struct dsl_dataset *ds = NULL;
+ objset_t *os = NULL;
+ zfsvfs_t *zfsvfs = NULL;
+ uint64_t nbmand;
+ int readonly, do_readonly = B_FALSE;
+ int setuid, do_setuid = B_FALSE;
+ int exec, do_exec = B_FALSE;
+ int devices, do_devices = B_FALSE;
+ int xattr, do_xattr = B_FALSE;
+ int atime, do_atime = B_FALSE;
+ int error = 0;
+
+ ASSERT(vfsp);
+ zfsvfs = vfsp->vfs_data;
+ ASSERT(zfsvfs);
+ os = zfsvfs->z_os;
+
+ /*
+ * The act of registering our callbacks will destroy any mount
+ * options we may have. In order to enable temporary overrides
+ * of mount options, we stash away the current values and
+ * restore them after we register the callbacks.
+ */
+ if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) {
+ readonly = B_TRUE;
+ do_readonly = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
+ readonly = B_FALSE;
+ do_readonly = B_TRUE;
+ }
+ if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
+ devices = B_FALSE;
+ setuid = B_FALSE;
+ do_devices = B_TRUE;
+ do_setuid = B_TRUE;
+ } else {
+ if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) {
+ devices = B_FALSE;
+ do_devices = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) {
+ devices = B_TRUE;
+ do_devices = B_TRUE;
+ }
+
+ if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
+ setuid = B_FALSE;
+ do_setuid = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
+ setuid = B_TRUE;
+ do_setuid = B_TRUE;
+ }
+ }
+ if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
+ exec = B_FALSE;
+ do_exec = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
+ exec = B_TRUE;
+ do_exec = B_TRUE;
+ }
+ if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
+ xattr = B_FALSE;
+ do_xattr = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
+ xattr = B_TRUE;
+ do_xattr = B_TRUE;
+ }
+ if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) {
+ atime = B_FALSE;
+ do_atime = B_TRUE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) {
+ atime = B_TRUE;
+ do_atime = B_TRUE;
+ }
+
+ /*
+ * nbmand is a special property. It can only be changed at
+ * mount time.
+ *
+ * This is weird, but it is documented to only be changeable
+ * at mount time.
+ */
+ if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
+ nbmand = B_FALSE;
+ } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) {
+ nbmand = B_TRUE;
+ } else {
+ char osname[MAXNAMELEN];
+
+ dmu_objset_name(os, osname);
+ if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand,
+ NULL)) {
+ return (error);
+ }
+ }
+
+ /*
+ * Register property callbacks.
+ *
+ * It would probably be fine to just check for i/o error from
+ * the first prop_register(), but I guess I like to go
+ * overboard...
+ */
+ ds = dmu_objset_ds(os);
+ error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "xattr", xattr_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "recordsize", blksz_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "readonly", readonly_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "devices", devices_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "setuid", setuid_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "exec", exec_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "snapdir", snapdir_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "aclmode", acl_mode_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "aclinherit", acl_inherit_changed_cb, zfsvfs);
+ error = error ? error : dsl_prop_register(ds,
+ "vscan", vscan_changed_cb, zfsvfs);
+ if (error)
+ goto unregister;
+
+ /*
+ * Invoke our callbacks to restore temporary mount options.
+ */
+ if (do_readonly)
+ readonly_changed_cb(zfsvfs, readonly);
+ if (do_setuid)
+ setuid_changed_cb(zfsvfs, setuid);
+ if (do_exec)
+ exec_changed_cb(zfsvfs, exec);
+ if (do_devices)
+ devices_changed_cb(zfsvfs, devices);
+ if (do_xattr)
+ xattr_changed_cb(zfsvfs, xattr);
+ if (do_atime)
+ atime_changed_cb(zfsvfs, atime);
+
+ nbmand_changed_cb(zfsvfs, nbmand);
+
+ return (0);
+
+unregister:
+ /*
+ * We may attempt to unregister some callbacks that are not
+ * registered, but this is OK; it will simply return ENOMSG,
+ * which we will ignore.
+ */
+ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
+ zfsvfs);
+ (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs);
+ return (error);
+
+}
+
+static int
+zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
+{
+ int error;
+
+ error = zfs_register_callbacks(zfsvfs->z_vfs);
+ if (error)
+ return (error);
+
+ /*
+ * Set the objset user_ptr to track its zfsvfs.
+ */
+ mutex_enter(&zfsvfs->z_os->os->os_user_ptr_lock);
+ dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
+ mutex_exit(&zfsvfs->z_os->os->os_user_ptr_lock);
+
+ /*
+ * If we are not mounting (ie: online recv), then we don't
+ * have to worry about replaying the log as we blocked all
+ * operations out since we closed the ZIL.
+ */
+ if (mounting) {
+ boolean_t readonly;
+
+ /*
+ * During replay we remove the read only flag to
+ * allow replays to succeed.
+ */
+ readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY;
+ zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
+
+ /*
+ * Parse and replay the intent log.
+ */
+ zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign,
+ zfs_replay_vector, zfs_unlinked_drain);
+
+ zfs_unlinked_drain(zfsvfs);
+ zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */
+ }
+
+ if (!zil_disable)
+ zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
+
+ return (0);
+}
+
+static void
+zfs_freezfsvfs(zfsvfs_t *zfsvfs)
+{
+ mutex_destroy(&zfsvfs->z_znodes_lock);
+ mutex_destroy(&zfsvfs->z_online_recv_lock);
+ list_destroy(&zfsvfs->z_all_znodes);
+ rrw_destroy(&zfsvfs->z_teardown_lock);
+ rw_destroy(&zfsvfs->z_teardown_inactive_lock);
+ rw_destroy(&zfsvfs->z_fuid_lock);
+ kmem_free(zfsvfs, sizeof (zfsvfs_t));
+}
+
+static int
+zfs_domount(vfs_t *vfsp, char *osname)
+{
+ dev_t mount_dev;
+ uint64_t recordsize, readonly;
+ int error = 0;
+ int mode;
+ zfsvfs_t *zfsvfs;
+ znode_t *zp = NULL;
+
+ ASSERT(vfsp);
+ ASSERT(osname);
+
+ /*
+ * Initialize the zfs-specific filesystem structure.
+ * Should probably make this a kmem cache, shuffle fields,
+ * and just bzero up to z_hold_mtx[].
+ */
+ zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
+ zfsvfs->z_vfs = vfsp;
+ zfsvfs->z_parent = zfsvfs;
+ zfsvfs->z_assign = TXG_NOWAIT;
+ zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
+ zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
+
+ mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
+ offsetof(znode_t, z_link_node));
+ rrw_init(&zfsvfs->z_teardown_lock);
+ rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
+ rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+
+ /* Initialize the generic filesystem structure. */
+ vfsp->vfs_bcount = 0;
+ vfsp->vfs_data = NULL;
+
+ if (zfs_create_unique_device(&mount_dev) == -1) {
+ error = ENODEV;
+ goto out;
+ }
+ ASSERT(vfs_devismounted(mount_dev) == 0);
+
+ if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize,
+ NULL))
+ goto out;
+
+ vfsp->vfs_dev = mount_dev;
+ vfsp->vfs_fstype = zfsfstype;
+ vfsp->vfs_bsize = recordsize;
+ vfsp->vfs_flag |= VFS_NOTRUNC;
+ vfsp->vfs_data = zfsvfs;
+
+ if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
+ goto out;
+
+ mode = DS_MODE_OWNER;
+ if (readonly)
+ mode |= DS_MODE_READONLY;
+
+ error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
+ if (error == EROFS) {
+ mode = DS_MODE_OWNER | DS_MODE_READONLY;
+ error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
+ &zfsvfs->z_os);
+ }
+
+ if (error)
+ goto out;
+
+ if (error = zfs_init_fs(zfsvfs, &zp))
+ goto out;
+
+ /* The call to zfs_init_fs leaves the vnode held, release it here. */
+ VN_RELE(ZTOV(zp));
+
+ /*
+ * Set features for file system.
+ */
+ zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
+ if (zfsvfs->z_use_fuids) {
+ vfs_set_feature(vfsp, VFSFT_XVATTR);
+ vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS);
+ vfs_set_feature(vfsp, VFSFT_ACEMASKONACCESS);
+ vfs_set_feature(vfsp, VFSFT_ACLONCREATE);
+ }
+ if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
+ vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
+ vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
+ vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE);
+ } else if (zfsvfs->z_case == ZFS_CASE_MIXED) {
+ vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
+ vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
+ }
+
+ if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
+ uint64_t pval;
+
+ ASSERT(mode & DS_MODE_READONLY);
+ atime_changed_cb(zfsvfs, B_FALSE);
+ readonly_changed_cb(zfsvfs, B_TRUE);
+ if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))
+ goto out;
+ xattr_changed_cb(zfsvfs, pval);
+ zfsvfs->z_issnap = B_TRUE;
+ } else {
+ error = zfsvfs_setup(zfsvfs, B_TRUE);
+ }
+
+ if (!zfsvfs->z_issnap)
+ zfsctl_create(zfsvfs);
+out:
+ if (error) {
+ if (zfsvfs->z_os)
+ dmu_objset_close(zfsvfs->z_os);
+ zfs_freezfsvfs(zfsvfs);
+ } else {
+ atomic_add_32(&zfs_active_fs_count, 1);
+ }
+
+ return (error);
+}
+
+void
+zfs_unregister_callbacks(zfsvfs_t *zfsvfs)
+{
+ objset_t *os = zfsvfs->z_os;
+ struct dsl_dataset *ds;
+
+ /*
+ * Unregister properties.
+ */
+ if (!dmu_objset_is_snapshot(os)) {
+ ds = dmu_objset_ds(os);
+ VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb,
+ zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "aclinherit",
+ acl_inherit_changed_cb, zfsvfs) == 0);
+
+ VERIFY(dsl_prop_unregister(ds, "vscan",
+ vscan_changed_cb, zfsvfs) == 0);
+ }
+}
+
+/*
+ * Convert a decimal digit string to a uint64_t integer.
+ */
+static int
+str_to_uint64(char *str, uint64_t *objnum)
+{
+ uint64_t num = 0;
+
+ while (*str) {
+ if (*str < '0' || *str > '9')
+ return (EINVAL);
+
+ num = num*10 + *str++ - '0';
+ }
+
+ *objnum = num;
+ return (0);
+}
+
+/*
+ * The boot path passed from the boot loader is in the form of
+ * "rootpool-name/root-filesystem-object-number'. Convert this
+ * string to a dataset name: "rootpool-name/root-filesystem-name".
+ */
+static int
+zfs_parse_bootfs(char *bpath, char *outpath)
+{
+ char *slashp;
+ uint64_t objnum;
+ int error;
+
+ if (*bpath == 0 || *bpath == '/')
+ return (EINVAL);
+
+ (void) strcpy(outpath, bpath);
+
+ slashp = strchr(bpath, '/');
+
+ /* if no '/', just return the pool name */
+ if (slashp == NULL) {
+ return (0);
+ }
+
+ /* if not a number, just return the root dataset name */
+ if (str_to_uint64(slashp+1, &objnum)) {
+ return (0);
+ }
+
+ *slashp = '\0';
+ error = dsl_dsobj_to_dsname(bpath, objnum, outpath);
+ *slashp = '/';
+
+ return (error);
+}
+
+static int
+zfs_mountroot(vfs_t *vfsp, enum whymountroot why)
+{
+ int error = 0;
+ static int zfsrootdone = 0;
+ zfsvfs_t *zfsvfs = NULL;
+ znode_t *zp = NULL;
+ vnode_t *vp = NULL;
+ char *zfs_bootfs;
+ char *zfs_devid;
+
+ ASSERT(vfsp);
+
+ /*
+ * The filesystem that we mount as root is defined in the
+ * boot property "zfs-bootfs" with a format of
+ * "poolname/root-dataset-objnum".
+ */
+ if (why == ROOT_INIT) {
+ if (zfsrootdone++)
+ return (EBUSY);
+ /*
+ * the process of doing a spa_load will require the
+ * clock to be set before we could (for example) do
+ * something better by looking at the timestamp on
+ * an uberblock, so just set it to -1.
+ */
+ clkset(-1);
+
+ if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) {
+ cmn_err(CE_NOTE, "spa_get_bootfs: can not get "
+ "bootfs name");
+ return (EINVAL);
+ }
+ zfs_devid = spa_get_bootprop("diskdevid");
+ error = spa_import_rootpool(rootfs.bo_name, zfs_devid);
+ if (zfs_devid)
+ spa_free_bootprop(zfs_devid);
+ if (error) {
+ spa_free_bootprop(zfs_bootfs);
+ cmn_err(CE_NOTE, "spa_import_rootpool: error %d",
+ error);
+ return (error);
+ }
+ if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) {
+ spa_free_bootprop(zfs_bootfs);
+ cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d",
+ error);
+ return (error);
+ }
+
+ spa_free_bootprop(zfs_bootfs);
+
+ if (error = vfs_lock(vfsp))
+ return (error);
+
+ if (error = zfs_domount(vfsp, rootfs.bo_name)) {
+ cmn_err(CE_NOTE, "zfs_domount: error %d", error);
+ goto out;
+ }
+
+ zfsvfs = (zfsvfs_t *)vfsp->vfs_data;
+ ASSERT(zfsvfs);
+ if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) {
+ cmn_err(CE_NOTE, "zfs_zget: error %d", error);
+ goto out;
+ }
+
+ vp = ZTOV(zp);
+ mutex_enter(&vp->v_lock);
+ vp->v_flag |= VROOT;
+ mutex_exit(&vp->v_lock);
+ rootvp = vp;
+
+ /*
+ * Leave rootvp held. The root file system is never unmounted.
+ */
+
+ vfs_add((struct vnode *)0, vfsp,
+ (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0);
+out:
+ vfs_unlock(vfsp);
+ return (error);
+ } else if (why == ROOT_REMOUNT) {
+ readonly_changed_cb(vfsp->vfs_data, B_FALSE);
+ vfsp->vfs_flag |= VFS_REMOUNT;
+
+ /* refresh mount options */
+ zfs_unregister_callbacks(vfsp->vfs_data);
+ return (zfs_register_callbacks(vfsp));
+
+ } else if (why == ROOT_UNMOUNT) {
+ zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data);
+ (void) zfs_sync(vfsp, 0, 0);
+ return (0);
+ }
+
+ /*
+ * if "why" is equal to anything else other than ROOT_INIT,
+ * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it.
+ */
+ return (ENOTSUP);
+}
+
+/*ARGSUSED*/
+static int
+zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
+{
+ char *osname;
+ pathname_t spn;
+ int error = 0;
+ uio_seg_t fromspace = (uap->flags & MS_SYSSPACE) ?
+ UIO_SYSSPACE : UIO_USERSPACE;
+ int canwrite;
+
+ if (mvp->v_type != VDIR)
+ return (ENOTDIR);
+
+ mutex_enter(&mvp->v_lock);
+ if ((uap->flags & MS_REMOUNT) == 0 &&
+ (uap->flags & MS_OVERLAY) == 0 &&
+ (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
+ mutex_exit(&mvp->v_lock);
+ return (EBUSY);
+ }
+ mutex_exit(&mvp->v_lock);
+
+ /*
+ * ZFS does not support passing unparsed data in via MS_DATA.
+ * Users should use the MS_OPTIONSTR interface; this means
+ * that all option parsing is already done and the options struct
+ * can be interrogated.
+ */
+ if ((uap->flags & MS_DATA) && uap->datalen > 0)
+ return (EINVAL);
+
+ /*
+ * Get the objset name (the "special" mount argument).
+ */
+ if (error = pn_get(uap->spec, fromspace, &spn))
+ return (error);
+
+ osname = spn.pn_path;
+
+ /*
+ * Check for mount privilege?
+ *
+ * If we don't have privilege then see if
+ * we have local permission to allow it
+ */
+ error = secpolicy_fs_mount(cr, mvp, vfsp);
+ if (error) {
+ error = dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr);
+ if (error == 0) {
+ vattr_t vattr;
+
+ /*
+ * Make sure user is the owner of the mount point
+ * or has sufficient privileges.
+ */
+
+ vattr.va_mask = AT_UID;
+
+ if (error = VOP_GETATTR(mvp, &vattr, 0, cr, NULL)) {
+ goto out;
+ }
+
+ if (secpolicy_vnode_owner(cr, vattr.va_uid) != 0 &&
+ VOP_ACCESS(mvp, VWRITE, 0, cr, NULL) != 0) {
+ error = EPERM;
+ goto out;
+ }
+
+ secpolicy_fs_mount_clearopts(cr, vfsp);
+ } else {
+ goto out;
+ }
+ }
+
+ /*
+ * Refuse to mount a filesystem if we are in a local zone and the
+ * dataset is not visible.
+ */
+ if (!INGLOBALZONE(curproc) &&
+ (!zone_dataset_visible(osname, &canwrite) || !canwrite)) {
+ error = EPERM;
+ goto out;
+ }
+
+ /*
+ * When doing a remount, we simply refresh our temporary properties
+ * according to those options set in the current VFS options.
+ */
+ if (uap->flags & MS_REMOUNT) {
+ /* refresh mount options */
+ zfs_unregister_callbacks(vfsp->vfs_data);
+ error = zfs_register_callbacks(vfsp);
+ goto out;
+ }
+
+ error = zfs_domount(vfsp, osname);
+
+out:
+ pn_free(&spn);
+ return (error);
+}
+
+static int
+zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ dev32_t d32;
+ uint64_t refdbytes, availbytes, usedobjs, availobjs;
+
+ ZFS_ENTER(zfsvfs);
+
+ dmu_objset_space(zfsvfs->z_os,
+ &refdbytes, &availbytes, &usedobjs, &availobjs);
+
+ /*
+ * The underlying storage pool actually uses multiple block sizes.
+ * We report the fragsize as the smallest block size we support,
+ * and we report our blocksize as the filesystem's maximum blocksize.
+ */
+ statp->f_frsize = 1UL << SPA_MINBLOCKSHIFT;
+ statp->f_bsize = zfsvfs->z_max_blksz;
+
+ /*
+ * The following report "total" blocks of various kinds in the
+ * file system, but reported in terms of f_frsize - the
+ * "fragment" size.
+ */
+
+ statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT;
+ statp->f_bfree = availbytes >> SPA_MINBLOCKSHIFT;
+ statp->f_bavail = statp->f_bfree; /* no root reservation */
+
+ /*
+ * statvfs() should really be called statufs(), because it assumes
+ * static metadata. ZFS doesn't preallocate files, so the best
+ * we can do is report the max that could possibly fit in f_files,
+ * and that minus the number actually used in f_ffree.
+ * For f_ffree, report the smaller of the number of object available
+ * and the number of blocks (each object will take at least a block).
+ */
+ statp->f_ffree = MIN(availobjs, statp->f_bfree);
+ statp->f_favail = statp->f_ffree; /* no "root reservation" */
+ statp->f_files = statp->f_ffree + usedobjs;
+
+ (void) cmpldev(&d32, vfsp->vfs_dev);
+ statp->f_fsid = d32;
+
+ /*
+ * We're a zfs filesystem.
+ */
+ (void) strcpy(statp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name);
+
+ statp->f_flag = vf_to_stf(vfsp->vfs_flag);
+
+ statp->f_namemax = ZFS_MAXNAMELEN;
+
+ /*
+ * We have all of 32 characters to stuff a string here.
+ * Is there anything useful we could/should provide?
+ */
+ bzero(statp->f_fstr, sizeof (statp->f_fstr));
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+static int
+zfs_root(vfs_t *vfsp, vnode_t **vpp)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ znode_t *rootzp;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+
+ error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
+ if (error == 0)
+ *vpp = ZTOV(rootzp);
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Teardown the zfsvfs::z_os.
+ *
+ * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
+ * and 'z_teardown_inactive_lock' held.
+ */
+static int
+zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
+{
+ znode_t *zp;
+
+ rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG);
+
+ if (!unmounting) {
+ /*
+ * We purge the parent filesystem's vfsp as the parent
+ * filesystem and all of its snapshots have their vnode's
+ * v_vfsp set to the parent's filesystem's vfsp. Note,
+ * 'z_parent' is self referential for non-snapshots.
+ */
+ (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
+ }
+
+ /*
+ * Close the zil. NB: Can't close the zil while zfs_inactive
+ * threads are blocked as zil_close can call zfs_inactive.
+ */
+ if (zfsvfs->z_log) {
+ zil_close(zfsvfs->z_log);
+ zfsvfs->z_log = NULL;
+ }
+
+ rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER);
+
+ /*
+ * If we are not unmounting (ie: online recv) and someone already
+ * unmounted this file system while we were doing the switcheroo,
+ * or a reopen of z_os failed then just bail out now.
+ */
+ if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) {
+ rw_exit(&zfsvfs->z_teardown_inactive_lock);
+ rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
+ return (EIO);
+ }
+
+ /*
+ * At this point there are no vops active, and any new vops will
+ * fail with EIO since we have z_teardown_lock for writer (only
+ * relavent for forced unmount).
+ *
+ * Release all holds on dbufs.
+ */
+ mutex_enter(&zfsvfs->z_znodes_lock);
+ for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL;
+ zp = list_next(&zfsvfs->z_all_znodes, zp))
+ if (zp->z_dbuf) {
+ ASSERT(ZTOV(zp)->v_count > 0);
+ zfs_znode_dmu_fini(zp);
+ }
+ mutex_exit(&zfsvfs->z_znodes_lock);
+
+ /*
+ * If we are unmounting, set the unmounted flag and let new vops
+ * unblock. zfs_inactive will have the unmounted behavior, and all
+ * other vops will fail with EIO.
+ */
+ if (unmounting) {
+ zfsvfs->z_unmounted = B_TRUE;
+ rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
+ rw_exit(&zfsvfs->z_teardown_inactive_lock);
+ }
+
+ /*
+ * z_os will be NULL if there was an error in attempting to reopen
+ * zfsvfs, so just return as the properties had already been
+ * unregistered and cached data had been evicted before.
+ */
+ if (zfsvfs->z_os == NULL)
+ return (0);
+
+ /*
+ * Unregister properties.
+ */
+ zfs_unregister_callbacks(zfsvfs);
+
+ /*
+ * Evict cached data
+ */
+ if (dmu_objset_evict_dbufs(zfsvfs->z_os)) {
+ txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
+ (void) dmu_objset_evict_dbufs(zfsvfs->z_os);
+ }
+
+ return (0);
+}
+
+/*ARGSUSED*/
+static int
+zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ objset_t *os;
+ int ret;
+
+ ret = secpolicy_fs_unmount(cr, vfsp);
+ if (ret) {
+ ret = dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource),
+ ZFS_DELEG_PERM_MOUNT, cr);
+ if (ret)
+ return (ret);
+ }
+
+ /*
+ * We purge the parent filesystem's vfsp as the parent filesystem
+ * and all of its snapshots have their vnode's v_vfsp set to the
+ * parent's filesystem's vfsp. Note, 'z_parent' is self
+ * referential for non-snapshots.
+ */
+ (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
+
+ /*
+ * Unmount any snapshots mounted under .zfs before unmounting the
+ * dataset itself.
+ */
+ if (zfsvfs->z_ctldir != NULL &&
+ (ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) {
+ return (ret);
+ }
+
+ if (!(fflag & MS_FORCE)) {
+ /*
+ * Check the number of active vnodes in the file system.
+ * Our count is maintained in the vfs structure, but the
+ * number is off by 1 to indicate a hold on the vfs
+ * structure itself.
+ *
+ * The '.zfs' directory maintains a reference of its
+ * own, and any active references underneath are
+ * reflected in the vnode count.
+ */
+ if (zfsvfs->z_ctldir == NULL) {
+ if (vfsp->vfs_count > 1)
+ return (EBUSY);
+ } else {
+ if (vfsp->vfs_count > 2 ||
+ zfsvfs->z_ctldir->v_count > 1)
+ return (EBUSY);
+ }
+ }
+
+ vfsp->vfs_flag |= VFS_UNMOUNTED;
+
+ VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0);
+ os = zfsvfs->z_os;
+
+ /*
+ * z_os will be NULL if there was an error in
+ * attempting to reopen zfsvfs.
+ */
+ if (os != NULL) {
+ /*
+ * Unset the objset user_ptr.
+ */
+ mutex_enter(&os->os->os_user_ptr_lock);
+ dmu_objset_set_user(os, NULL);
+ mutex_exit(&os->os->os_user_ptr_lock);
+
+ /*
+ * Finally release the objset
+ */
+ dmu_objset_close(os);
+ }
+
+ /*
+ * We can now safely destroy the '.zfs' directory node.
+ */
+ if (zfsvfs->z_ctldir != NULL)
+ zfsctl_destroy(zfsvfs);
+
+ return (0);
+}
+
+static int
+zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ znode_t *zp;
+ uint64_t object = 0;
+ uint64_t fid_gen = 0;
+ uint64_t gen_mask;
+ uint64_t zp_gen;
+ int i, err;
+
+ *vpp = NULL;
+
+ ZFS_ENTER(zfsvfs);
+
+ if (fidp->fid_len == LONG_FID_LEN) {
+ zfid_long_t *zlfid = (zfid_long_t *)fidp;
+ uint64_t objsetid = 0;
+ uint64_t setgen = 0;
+
+ for (i = 0; i < sizeof (zlfid->zf_setid); i++)
+ objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i);
+
+ for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
+ setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);
+
+ ZFS_EXIT(zfsvfs);
+
+ err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs);
+ if (err)
+ return (EINVAL);
+ ZFS_ENTER(zfsvfs);
+ }
+
+ if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) {
+ zfid_short_t *zfid = (zfid_short_t *)fidp;
+
+ for (i = 0; i < sizeof (zfid->zf_object); i++)
+ object |= ((uint64_t)zfid->zf_object[i]) << (8 * i);
+
+ for (i = 0; i < sizeof (zfid->zf_gen); i++)
+ fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
+ } else {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /* A zero fid_gen means we are in the .zfs control directories */
+ if (fid_gen == 0 &&
+ (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) {
+ *vpp = zfsvfs->z_ctldir;
+ ASSERT(*vpp != NULL);
+ if (object == ZFSCTL_INO_SNAPDIR) {
+ VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL,
+ 0, NULL, NULL, NULL, NULL, NULL) == 0);
+ } else {
+ VN_HOLD(*vpp);
+ }
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+
+ gen_mask = -1ULL >> (64 - 8 * i);
+
+ dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask);
+ if (err = zfs_zget(zfsvfs, object, &zp)) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ zp_gen = zp->z_phys->zp_gen & gen_mask;
+ if (zp_gen == 0)
+ zp_gen = 1;
+ if (zp->z_unlinked || zp_gen != fid_gen) {
+ dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen);
+ VN_RELE(ZTOV(zp));
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ *vpp = ZTOV(zp);
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/*
+ * Block out VOPs and close zfsvfs_t::z_os
+ *
+ * Note, if successful, then we return with the 'z_teardown_lock' and
+ * 'z_teardown_inactive_lock' write held.
+ */
+int
+zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *mode)
+{
+ int error;
+
+ if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0)
+ return (error);
+
+ *mode = zfsvfs->z_os->os_mode;
+ dmu_objset_name(zfsvfs->z_os, name);
+ dmu_objset_close(zfsvfs->z_os);
+
+ return (0);
+}
+
+/*
+ * Reopen zfsvfs_t::z_os and release VOPs.
+ */
+int
+zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode)
+{
+ int err;
+
+ ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock));
+ ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock));
+
+ err = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
+ if (err) {
+ zfsvfs->z_os = NULL;
+ } else {
+ znode_t *zp;
+
+ VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0);
+
+ /*
+ * Attempt to re-establish all the active znodes with
+ * their dbufs. If a zfs_rezget() fails, then we'll let
+ * any potential callers discover that via ZFS_ENTER_VERIFY_VP
+ * when they try to use their znode.
+ */
+ mutex_enter(&zfsvfs->z_znodes_lock);
+ for (zp = list_head(&zfsvfs->z_all_znodes); zp;
+ zp = list_next(&zfsvfs->z_all_znodes, zp)) {
+ (void) zfs_rezget(zp);
+ }
+ mutex_exit(&zfsvfs->z_znodes_lock);
+
+ }
+
+ /* release the VOPs */
+ rw_exit(&zfsvfs->z_teardown_inactive_lock);
+ rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
+
+ if (err) {
+ /*
+ * Since we couldn't reopen zfsvfs::z_os, force
+ * unmount this file system.
+ */
+ if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0)
+ (void) dounmount(zfsvfs->z_vfs, MS_FORCE, CRED());
+ }
+ return (err);
+}
+
+static void
+zfs_freevfs(vfs_t *vfsp)
+{
+ zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ int i;
+
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_destroy(&zfsvfs->z_hold_mtx[i]);
+
+ zfs_fuid_destroy(zfsvfs);
+ zfs_freezfsvfs(zfsvfs);
+
+ atomic_add_32(&zfs_active_fs_count, -1);
+}
+
+/*
+ * VFS_INIT() initialization. Note that there is no VFS_FINI(),
+ * so we can't safely do any non-idempotent initialization here.
+ * Leave that to zfs_init() and zfs_fini(), which are called
+ * from the module's _init() and _fini() entry points.
+ */
+/*ARGSUSED*/
+static int
+zfs_vfsinit(int fstype, char *name)
+{
+ int error;
+
+ zfsfstype = fstype;
+
+ /*
+ * Setup vfsops and vnodeops tables.
+ */
+ error = vfs_setfsops(fstype, zfs_vfsops_template, &zfs_vfsops);
+ if (error != 0) {
+ cmn_err(CE_WARN, "zfs: bad vfs ops template");
+ }
+
+ error = zfs_create_op_tables();
+ if (error) {
+ zfs_remove_op_tables();
+ cmn_err(CE_WARN, "zfs: bad vnode ops template");
+ (void) vfs_freevfsops_by_type(zfsfstype);
+ return (error);
+ }
+
+ mutex_init(&zfs_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
+
+ /*
+ * Unique major number for all zfs mounts.
+ * If we run out of 32-bit minors, we'll getudev() another major.
+ */
+ zfs_major = ddi_name_to_major(ZFS_DRIVER);
+ zfs_minor = ZFS_MIN_MINOR;
+
+ return (0);
+}
+
+void
+zfs_init(void)
+{
+ /*
+ * Initialize .zfs directory structures
+ */
+ zfsctl_init();
+
+ /*
+ * Initialize znode cache, vnode ops, etc...
+ */
+ zfs_znode_init();
+}
+
+void
+zfs_fini(void)
+{
+ zfsctl_fini();
+ zfs_znode_fini();
+}
+
+int
+zfs_busy(void)
+{
+ return (zfs_active_fs_count != 0);
+}
+
+int
+zfs_set_version(const char *name, uint64_t newvers)
+{
+ int error;
+ objset_t *os;
+ dmu_tx_t *tx;
+ uint64_t curvers;
+
+ /*
+ * XXX for now, require that the filesystem be unmounted. Would
+ * be nice to find the zfsvfs_t and just update that if
+ * possible.
+ */
+
+ if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
+ return (EINVAL);
+
+ error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_OWNER, &os);
+ if (error)
+ return (error);
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
+ 8, 1, &curvers);
+ if (error)
+ goto out;
+ if (newvers < curvers) {
+ error = EINVAL;
+ goto out;
+ }
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ goto out;
+ }
+ error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1,
+ &newvers, tx);
+
+ spa_history_internal_log(LOG_DS_UPGRADE,
+ dmu_objset_spa(os), tx, CRED(),
+ "oldver=%llu newver=%llu dataset = %llu", curvers, newvers,
+ dmu_objset_id(os));
+ dmu_tx_commit(tx);
+
+out:
+ dmu_objset_close(os);
+ return (error);
+}
+
+/*
+ * Read a property stored within the master node.
+ */
+int
+zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value)
+{
+ const char *pname;
+ int error = ENOENT;
+
+ /*
+ * Look up the file system's value for the property. For the
+ * version property, we look up a slightly different string.
+ */
+ if (prop == ZFS_PROP_VERSION)
+ pname = ZPL_VERSION_STR;
+ else
+ pname = zfs_prop_to_name(prop);
+
+ if (os != NULL)
+ error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value);
+
+ if (error == ENOENT) {
+ /* No value set, use the default value */
+ switch (prop) {
+ case ZFS_PROP_VERSION:
+ *value = ZPL_VERSION;
+ break;
+ case ZFS_PROP_NORMALIZE:
+ case ZFS_PROP_UTF8ONLY:
+ *value = 0;
+ break;
+ case ZFS_PROP_CASE:
+ *value = ZFS_CASE_SENSITIVE;
+ break;
+ default:
+ return (error);
+ }
+ error = 0;
+ }
+ return (error);
+}
+
+static vfsdef_t vfw = {
+ VFSDEF_VERSION,
+ MNTTYPE_ZFS,
+ zfs_vfsinit,
+ VSW_HASPROTO|VSW_CANRWRO|VSW_CANREMOUNT|VSW_VOLATILEDEV|VSW_STATS|
+ VSW_XID,
+ &zfs_mntopts
+};
+
+struct modlfs zfs_modlfs = {
+ &mod_fsops, "ZFS filesystem version " SPA_VERSION_STRING, &vfw
+};
diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
new file mode 100644
index 000000000..8e0037e37
--- /dev/null
+++ b/module/zfs/zfs_vnops.c
@@ -0,0 +1,4561 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/* Portions Copyright 2007 Jeremy Teo */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/time.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/resource.h>
+#include <sys/vfs.h>
+#include <sys/vfs_opreg.h>
+#include <sys/vnode.h>
+#include <sys/file.h>
+#include <sys/stat.h>
+#include <sys/kmem.h>
+#include <sys/taskq.h>
+#include <sys/uio.h>
+#include <sys/vmsystm.h>
+#include <sys/atomic.h>
+#include <sys/vm.h>
+#include <vm/seg_vn.h>
+#include <vm/pvn.h>
+#include <vm/as.h>
+#include <vm/kpm.h>
+#include <vm/seg_kpm.h>
+#include <sys/mman.h>
+#include <sys/pathname.h>
+#include <sys/cmn_err.h>
+#include <sys/errno.h>
+#include <sys/unistd.h>
+#include <sys/zfs_dir.h>
+#include <sys/zfs_acl.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/fs/zfs.h>
+#include <sys/dmu.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/dbuf.h>
+#include <sys/zap.h>
+#include <sys/dirent.h>
+#include <sys/policy.h>
+#include <sys/sunddi.h>
+#include <sys/filio.h>
+#include <sys/sid.h>
+#include "fs/fs_subr.h"
+#include <sys/zfs_ctldir.h>
+#include <sys/zfs_fuid.h>
+#include <sys/dnlc.h>
+#include <sys/zfs_rlock.h>
+#include <sys/extdirent.h>
+#include <sys/kidmap.h>
+#include <sys/cred_impl.h>
+#include <sys/attr.h>
+
+/*
+ * Programming rules.
+ *
+ * Each vnode op performs some logical unit of work. To do this, the ZPL must
+ * properly lock its in-core state, create a DMU transaction, do the work,
+ * record this work in the intent log (ZIL), commit the DMU transaction,
+ * and wait for the intent log to commit if it is a synchronous operation.
+ * Moreover, the vnode ops must work in both normal and log replay context.
+ * The ordering of events is important to avoid deadlocks and references
+ * to freed memory. The example below illustrates the following Big Rules:
+ *
+ * (1) A check must be made in each zfs thread for a mounted file system.
+ * This is done avoiding races using ZFS_ENTER(zfsvfs).
+ * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
+ * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
+ * can return EIO from the calling function.
+ *
+ * (2) VN_RELE() should always be the last thing except for zil_commit()
+ * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
+ * First, if it's the last reference, the vnode/znode
+ * can be freed, so the zp may point to freed memory. Second, the last
+ * reference will call zfs_zinactive(), which may induce a lot of work --
+ * pushing cached pages (which acquires range locks) and syncing out
+ * cached atime changes. Third, zfs_zinactive() may require a new tx,
+ * which could deadlock the system if you were already holding one.
+ *
+ * (3) All range locks must be grabbed before calling dmu_tx_assign(),
+ * as they can span dmu_tx_assign() calls.
+ *
+ * (4) Always pass zfsvfs->z_assign as the second argument to dmu_tx_assign().
+ * In normal operation, this will be TXG_NOWAIT. During ZIL replay,
+ * it will be a specific txg. Either way, dmu_tx_assign() never blocks.
+ * This is critical because we don't want to block while holding locks.
+ * Note, in particular, that if a lock is sometimes acquired before
+ * the tx assigns, and sometimes after (e.g. z_lock), then failing to
+ * use a non-blocking assign can deadlock the system. The scenario:
+ *
+ * Thread A has grabbed a lock before calling dmu_tx_assign().
+ * Thread B is in an already-assigned tx, and blocks for this lock.
+ * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
+ * forever, because the previous txg can't quiesce until B's tx commits.
+ *
+ * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
+ * then drop all locks, call dmu_tx_wait(), and try again.
+ *
+ * (5) If the operation succeeded, generate the intent log entry for it
+ * before dropping locks. This ensures that the ordering of events
+ * in the intent log matches the order in which they actually occurred.
+ *
+ * (6) At the end of each vnode op, the DMU tx must always commit,
+ * regardless of whether there were any errors.
+ *
+ * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
+ * to ensure that synchronous semantics are provided when necessary.
+ *
+ * In general, this is how things should be ordered in each vnode op:
+ *
+ * ZFS_ENTER(zfsvfs); // exit if unmounted
+ * top:
+ * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
+ * rw_enter(...); // grab any other locks you need
+ * tx = dmu_tx_create(...); // get DMU tx
+ * dmu_tx_hold_*(); // hold each object you might modify
+ * error = dmu_tx_assign(tx, zfsvfs->z_assign); // try to assign
+ * if (error) {
+ * rw_exit(...); // drop locks
+ * zfs_dirent_unlock(dl); // unlock directory entry
+ * VN_RELE(...); // release held vnodes
+ * if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ * dmu_tx_wait(tx);
+ * dmu_tx_abort(tx);
+ * goto top;
+ * }
+ * dmu_tx_abort(tx); // abort DMU tx
+ * ZFS_EXIT(zfsvfs); // finished in zfs
+ * return (error); // really out of space
+ * }
+ * error = do_real_work(); // do whatever this VOP does
+ * if (error == 0)
+ * zfs_log_*(...); // on success, make ZIL entry
+ * dmu_tx_commit(tx); // commit DMU tx -- error or not
+ * rw_exit(...); // drop locks
+ * zfs_dirent_unlock(dl); // unlock directory entry
+ * VN_RELE(...); // release held vnodes
+ * zil_commit(zilog, seq, foid); // synchronous when necessary
+ * ZFS_EXIT(zfsvfs); // finished in zfs
+ * return (error); // done, report error
+ */
+
+/* ARGSUSED */
+static int
+zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(*vpp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
+ ((flag & FAPPEND) == 0)) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
+ ZTOV(zp)->v_type == VREG &&
+ !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
+ zp->z_phys->zp_size > 0) {
+ if (fs_vscan(*vpp, cr, 0) != 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EACCES);
+ }
+ }
+
+ /* Keep a count of the synchronous opens in the znode */
+ if (flag & (FSYNC | FDSYNC))
+ atomic_inc_32(&zp->z_sync_cnt);
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/* ARGSUSED */
+static int
+zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ /* Decrement the synchronous opens in the znode */
+ if ((flag & (FSYNC | FDSYNC)) && (count == 1))
+ atomic_dec_32(&zp->z_sync_cnt);
+
+ /*
+ * Clean up any locks held by this process on the vp.
+ */
+ cleanlocks(vp, ddi_get_pid(), 0);
+ cleanshares(vp, ddi_get_pid());
+
+ if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
+ ZTOV(zp)->v_type == VREG &&
+ !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
+ zp->z_phys->zp_size > 0)
+ VERIFY(fs_vscan(vp, cr, 1) == 0);
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/*
+ * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
+ * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
+ */
+static int
+zfs_holey(vnode_t *vp, int cmd, offset_t *off)
+{
+ znode_t *zp = VTOZ(vp);
+ uint64_t noff = (uint64_t)*off; /* new offset */
+ uint64_t file_sz;
+ int error;
+ boolean_t hole;
+
+ file_sz = zp->z_phys->zp_size;
+ if (noff >= file_sz) {
+ return (ENXIO);
+ }
+
+ if (cmd == _FIO_SEEK_HOLE)
+ hole = B_TRUE;
+ else
+ hole = B_FALSE;
+
+ error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
+
+ /* end of file? */
+ if ((error == ESRCH) || (noff > file_sz)) {
+ /*
+ * Handle the virtual hole at the end of file.
+ */
+ if (hole) {
+ *off = file_sz;
+ return (0);
+ }
+ return (ENXIO);
+ }
+
+ if (noff < *off)
+ return (error);
+ *off = noff;
+ return (error);
+}
+
+/* ARGSUSED */
+static int
+zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
+ int *rvalp, caller_context_t *ct)
+{
+ offset_t off;
+ int error;
+ zfsvfs_t *zfsvfs;
+ znode_t *zp;
+
+ switch (com) {
+ case _FIOFFS:
+ return (zfs_sync(vp->v_vfsp, 0, cred));
+
+ /*
+ * The following two ioctls are used by bfu. Faking out,
+ * necessary to avoid bfu errors.
+ */
+ case _FIOGDIO:
+ case _FIOSDIO:
+ return (0);
+
+ case _FIO_SEEK_DATA:
+ case _FIO_SEEK_HOLE:
+ if (ddi_copyin((void *)data, &off, sizeof (off), flag))
+ return (EFAULT);
+
+ zp = VTOZ(vp);
+ zfsvfs = zp->z_zfsvfs;
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ /* offset parameter is in/out */
+ error = zfs_holey(vp, com, &off);
+ ZFS_EXIT(zfsvfs);
+ if (error)
+ return (error);
+ if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
+ return (EFAULT);
+ return (0);
+ }
+ return (ENOTTY);
+}
+
+/*
+ * Utility functions to map and unmap a single physical page. These
+ * are used to manage the mappable copies of ZFS file data, and therefore
+ * do not update ref/mod bits.
+ */
+caddr_t
+zfs_map_page(page_t *pp, enum seg_rw rw)
+{
+ if (kpm_enable)
+ return (hat_kpm_mapin(pp, 0));
+ ASSERT(rw == S_READ || rw == S_WRITE);
+ return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0),
+ (caddr_t)-1));
+}
+
+void
+zfs_unmap_page(page_t *pp, caddr_t addr)
+{
+ if (kpm_enable) {
+ hat_kpm_mapout(pp, 0, addr);
+ } else {
+ ppmapout(addr);
+ }
+}
+
+/*
+ * When a file is memory mapped, we must keep the IO data synchronized
+ * between the DMU cache and the memory mapped pages. What this means:
+ *
+ * On Write: If we find a memory mapped page, we write to *both*
+ * the page and the dmu buffer.
+ *
+ * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
+ * the file is memory mapped.
+ */
+static int
+mappedwrite(vnode_t *vp, int nbytes, uio_t *uio, dmu_tx_t *tx)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int64_t start, off;
+ int len = nbytes;
+ int error = 0;
+
+ start = uio->uio_loffset;
+ off = start & PAGEOFFSET;
+ for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
+ page_t *pp;
+ uint64_t bytes = MIN(PAGESIZE - off, len);
+ uint64_t woff = uio->uio_loffset;
+
+ /*
+ * We don't want a new page to "appear" in the middle of
+ * the file update (because it may not get the write
+ * update data), so we grab a lock to block
+ * zfs_getpage().
+ */
+ rw_enter(&zp->z_map_lock, RW_WRITER);
+ if (pp = page_lookup(vp, start, SE_SHARED)) {
+ caddr_t va;
+
+ rw_exit(&zp->z_map_lock);
+ va = zfs_map_page(pp, S_WRITE);
+ error = uiomove(va+off, bytes, UIO_WRITE, uio);
+ if (error == 0) {
+ dmu_write(zfsvfs->z_os, zp->z_id,
+ woff, bytes, va+off, tx);
+ }
+ zfs_unmap_page(pp, va);
+ page_unlock(pp);
+ } else {
+ error = dmu_write_uio(zfsvfs->z_os, zp->z_id,
+ uio, bytes, tx);
+ rw_exit(&zp->z_map_lock);
+ }
+ len -= bytes;
+ off = 0;
+ if (error)
+ break;
+ }
+ return (error);
+}
+
+/*
+ * When a file is memory mapped, we must keep the IO data synchronized
+ * between the DMU cache and the memory mapped pages. What this means:
+ *
+ * On Read: We "read" preferentially from memory mapped pages,
+ * else we default from the dmu buffer.
+ *
+ * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
+ * the file is memory mapped.
+ */
+static int
+mappedread(vnode_t *vp, int nbytes, uio_t *uio)
+{
+ znode_t *zp = VTOZ(vp);
+ objset_t *os = zp->z_zfsvfs->z_os;
+ int64_t start, off;
+ int len = nbytes;
+ int error = 0;
+
+ start = uio->uio_loffset;
+ off = start & PAGEOFFSET;
+ for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
+ page_t *pp;
+ uint64_t bytes = MIN(PAGESIZE - off, len);
+
+ if (pp = page_lookup(vp, start, SE_SHARED)) {
+ caddr_t va;
+
+ va = zfs_map_page(pp, S_READ);
+ error = uiomove(va + off, bytes, UIO_READ, uio);
+ zfs_unmap_page(pp, va);
+ page_unlock(pp);
+ } else {
+ error = dmu_read_uio(os, zp->z_id, uio, bytes);
+ }
+ len -= bytes;
+ off = 0;
+ if (error)
+ break;
+ }
+ return (error);
+}
+
+offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
+
+/*
+ * Read bytes from specified file into supplied buffer.
+ *
+ * IN: vp - vnode of file to be read from.
+ * uio - structure supplying read location, range info,
+ * and return buffer.
+ * ioflag - SYNC flags; used to provide FRSYNC semantics.
+ * cr - credentials of caller.
+ * ct - caller context
+ *
+ * OUT: uio - updated offset and range, buffer filled.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Side Effects:
+ * vp - atime updated if byte count > 0
+ */
+/* ARGSUSED */
+static int
+zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ objset_t *os;
+ ssize_t n, nbytes;
+ int error;
+ rl_t *rl;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ os = zfsvfs->z_os;
+
+ if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
+ ZFS_EXIT(zfsvfs);
+ return (EACCES);
+ }
+
+ /*
+ * Validate file offset
+ */
+ if (uio->uio_loffset < (offset_t)0) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * Fasttrack empty reads
+ */
+ if (uio->uio_resid == 0) {
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+
+ /*
+ * Check for mandatory locks
+ */
+ if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
+ if (error = chklock(vp, FREAD,
+ uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ }
+
+ /*
+ * If we're in FRSYNC mode, sync out this znode before reading it.
+ */
+ if (ioflag & FRSYNC)
+ zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
+
+ /*
+ * Lock the range against changes.
+ */
+ rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
+
+ /*
+ * If we are reading past end-of-file we can skip
+ * to the end; but we might still need to set atime.
+ */
+ if (uio->uio_loffset >= zp->z_phys->zp_size) {
+ error = 0;
+ goto out;
+ }
+
+ ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
+ n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
+
+ while (n > 0) {
+ nbytes = MIN(n, zfs_read_chunk_size -
+ P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
+
+ if (vn_has_cached_data(vp))
+ error = mappedread(vp, nbytes, uio);
+ else
+ error = dmu_read_uio(os, zp->z_id, uio, nbytes);
+ if (error) {
+ /* convert checksum errors into IO errors */
+ if (error == ECKSUM)
+ error = EIO;
+ break;
+ }
+
+ n -= nbytes;
+ }
+
+out:
+ zfs_range_unlock(rl);
+
+ ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Write the bytes to a file.
+ *
+ * IN: vp - vnode of file to be written to.
+ * uio - structure supplying write location, range info,
+ * and data buffer.
+ * ioflag - FAPPEND flag set if in append mode.
+ * cr - credentials of caller.
+ * ct - caller context (NFS/CIFS fem monitor only)
+ *
+ * OUT: uio - updated offset and range.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - ctime|mtime updated if byte count > 0
+ */
+/* ARGSUSED */
+static int
+zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ rlim64_t limit = uio->uio_llimit;
+ ssize_t start_resid = uio->uio_resid;
+ ssize_t tx_bytes;
+ uint64_t end_size;
+ dmu_tx_t *tx;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zilog_t *zilog;
+ offset_t woff;
+ ssize_t n, nbytes;
+ rl_t *rl;
+ int max_blksz = zfsvfs->z_max_blksz;
+ uint64_t pflags;
+ int error;
+
+ /*
+ * Fasttrack empty write
+ */
+ n = start_resid;
+ if (n == 0)
+ return (0);
+
+ if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
+ limit = MAXOFFSET_T;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ /*
+ * If immutable or not appending then return EPERM
+ */
+ pflags = zp->z_phys->zp_flags;
+ if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
+ ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
+ (uio->uio_loffset < zp->z_phys->zp_size))) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ zilog = zfsvfs->z_log;
+
+ /*
+ * Pre-fault the pages to ensure slow (eg NFS) pages
+ * don't hold up txg.
+ */
+ uio_prefaultpages(n, uio);
+
+ /*
+ * If in append mode, set the io offset pointer to eof.
+ */
+ if (ioflag & FAPPEND) {
+ /*
+ * Range lock for a file append:
+ * The value for the start of range will be determined by
+ * zfs_range_lock() (to guarantee append semantics).
+ * If this write will cause the block size to increase,
+ * zfs_range_lock() will lock the entire file, so we must
+ * later reduce the range after we grow the block size.
+ */
+ rl = zfs_range_lock(zp, 0, n, RL_APPEND);
+ if (rl->r_len == UINT64_MAX) {
+ /* overlocked, zp_size can't change */
+ woff = uio->uio_loffset = zp->z_phys->zp_size;
+ } else {
+ woff = uio->uio_loffset = rl->r_off;
+ }
+ } else {
+ woff = uio->uio_loffset;
+ /*
+ * Validate file offset
+ */
+ if (woff < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * If we need to grow the block size then zfs_range_lock()
+ * will lock a wider range than we request here.
+ * Later after growing the block size we reduce the range.
+ */
+ rl = zfs_range_lock(zp, woff, n, RL_WRITER);
+ }
+
+ if (woff >= limit) {
+ zfs_range_unlock(rl);
+ ZFS_EXIT(zfsvfs);
+ return (EFBIG);
+ }
+
+ if ((woff + n) > limit || woff > (limit - n))
+ n = limit - woff;
+
+ /*
+ * Check for mandatory locks
+ */
+ if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
+ (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
+ zfs_range_unlock(rl);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ end_size = MAX(zp->z_phys->zp_size, woff + n);
+
+ /*
+ * Write the file in reasonable size chunks. Each chunk is written
+ * in a separate transaction; this keeps the intent log records small
+ * and allows us to do more fine-grained space accounting.
+ */
+ while (n > 0) {
+ /*
+ * Start a transaction.
+ */
+ woff = uio->uio_loffset;
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (error == ERESTART &&
+ zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ continue;
+ }
+ dmu_tx_abort(tx);
+ break;
+ }
+
+ /*
+ * If zfs_range_lock() over-locked we grow the blocksize
+ * and then reduce the lock range. This will only happen
+ * on the first iteration since zfs_range_reduce() will
+ * shrink down r_len to the appropriate size.
+ */
+ if (rl->r_len == UINT64_MAX) {
+ uint64_t new_blksz;
+
+ if (zp->z_blksz > max_blksz) {
+ ASSERT(!ISP2(zp->z_blksz));
+ new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
+ } else {
+ new_blksz = MIN(end_size, max_blksz);
+ }
+ zfs_grow_blocksize(zp, new_blksz, tx);
+ zfs_range_reduce(rl, woff, n);
+ }
+
+ /*
+ * XXX - should we really limit each write to z_max_blksz?
+ * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
+ */
+ nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
+ rw_enter(&zp->z_map_lock, RW_READER);
+
+ tx_bytes = uio->uio_resid;
+ if (vn_has_cached_data(vp)) {
+ rw_exit(&zp->z_map_lock);
+ error = mappedwrite(vp, nbytes, uio, tx);
+ } else {
+ error = dmu_write_uio(zfsvfs->z_os, zp->z_id,
+ uio, nbytes, tx);
+ rw_exit(&zp->z_map_lock);
+ }
+ tx_bytes -= uio->uio_resid;
+
+ /*
+ * If we made no progress, we're done. If we made even
+ * partial progress, update the znode and ZIL accordingly.
+ */
+ if (tx_bytes == 0) {
+ dmu_tx_commit(tx);
+ ASSERT(error != 0);
+ break;
+ }
+
+ /*
+ * Clear Set-UID/Set-GID bits on successful write if not
+ * privileged and at least one of the excute bits is set.
+ *
+ * It would be nice to to this after all writes have
+ * been done, but that would still expose the ISUID/ISGID
+ * to another app after the partial write is committed.
+ *
+ * Note: we don't call zfs_fuid_map_id() here because
+ * user 0 is not an ephemeral uid.
+ */
+ mutex_enter(&zp->z_acl_lock);
+ if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
+ (S_IXUSR >> 6))) != 0 &&
+ (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
+ secpolicy_vnode_setid_retain(cr,
+ (zp->z_phys->zp_mode & S_ISUID) != 0 &&
+ zp->z_phys->zp_uid == 0) != 0) {
+ zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
+ }
+ mutex_exit(&zp->z_acl_lock);
+
+ /*
+ * Update time stamp. NOTE: This marks the bonus buffer as
+ * dirty, so we don't have to do it again for zp_size.
+ */
+ zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
+
+ /*
+ * Update the file size (zp_size) if it has changed;
+ * account for possible concurrent updates.
+ */
+ while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
+ (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
+ uio->uio_loffset);
+ zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
+ dmu_tx_commit(tx);
+
+ if (error != 0)
+ break;
+ ASSERT(tx_bytes == nbytes);
+ n -= nbytes;
+ }
+
+ zfs_range_unlock(rl);
+
+ /*
+ * If we're in replay mode, or we made no progress, return error.
+ * Otherwise, it's at least a partial write, so it's successful.
+ */
+ if (zfsvfs->z_assign >= TXG_INITIAL || uio->uio_resid == start_resid) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (ioflag & (FSYNC | FDSYNC))
+ zil_commit(zilog, zp->z_last_itx, zp->z_id);
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+void
+zfs_get_done(dmu_buf_t *db, void *vzgd)
+{
+ zgd_t *zgd = (zgd_t *)vzgd;
+ rl_t *rl = zgd->zgd_rl;
+ vnode_t *vp = ZTOV(rl->r_zp);
+
+ dmu_buf_rele(db, vzgd);
+ zfs_range_unlock(rl);
+ VN_RELE(vp);
+ zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
+ kmem_free(zgd, sizeof (zgd_t));
+}
+
+/*
+ * Get data to generate a TX_WRITE intent log record.
+ */
+int
+zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
+{
+ zfsvfs_t *zfsvfs = arg;
+ objset_t *os = zfsvfs->z_os;
+ znode_t *zp;
+ uint64_t off = lr->lr_offset;
+ dmu_buf_t *db;
+ rl_t *rl;
+ zgd_t *zgd;
+ int dlen = lr->lr_length; /* length of user data */
+ int error = 0;
+
+ ASSERT(zio);
+ ASSERT(dlen != 0);
+
+ /*
+ * Nothing to do if the file has been removed
+ */
+ if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
+ return (ENOENT);
+ if (zp->z_unlinked) {
+ VN_RELE(ZTOV(zp));
+ return (ENOENT);
+ }
+
+ /*
+ * Write records come in two flavors: immediate and indirect.
+ * For small writes it's cheaper to store the data with the
+ * log record (immediate); for large writes it's cheaper to
+ * sync the data and get a pointer to it (indirect) so that
+ * we don't have to write the data twice.
+ */
+ if (buf != NULL) { /* immediate write */
+ rl = zfs_range_lock(zp, off, dlen, RL_READER);
+ /* test for truncation needs to be done while range locked */
+ if (off >= zp->z_phys->zp_size) {
+ error = ENOENT;
+ goto out;
+ }
+ VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf));
+ } else { /* indirect write */
+ uint64_t boff; /* block starting offset */
+
+ /*
+ * Have to lock the whole block to ensure when it's
+ * written out and it's checksum is being calculated
+ * that no one can change the data. We need to re-check
+ * blocksize after we get the lock in case it's changed!
+ */
+ for (;;) {
+ if (ISP2(zp->z_blksz)) {
+ boff = P2ALIGN_TYPED(off, zp->z_blksz,
+ uint64_t);
+ } else {
+ boff = 0;
+ }
+ dlen = zp->z_blksz;
+ rl = zfs_range_lock(zp, boff, dlen, RL_READER);
+ if (zp->z_blksz == dlen)
+ break;
+ zfs_range_unlock(rl);
+ }
+ /* test for truncation needs to be done while range locked */
+ if (off >= zp->z_phys->zp_size) {
+ error = ENOENT;
+ goto out;
+ }
+ zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
+ zgd->zgd_rl = rl;
+ zgd->zgd_zilog = zfsvfs->z_log;
+ zgd->zgd_bp = &lr->lr_blkptr;
+ VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
+ ASSERT(boff == db->db_offset);
+ lr->lr_blkoff = off - boff;
+ error = dmu_sync(zio, db, &lr->lr_blkptr,
+ lr->lr_common.lrc_txg, zfs_get_done, zgd);
+ ASSERT((error && error != EINPROGRESS) ||
+ lr->lr_length <= zp->z_blksz);
+ if (error == 0)
+ zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
+ /*
+ * If we get EINPROGRESS, then we need to wait for a
+ * write IO initiated by dmu_sync() to complete before
+ * we can release this dbuf. We will finish everything
+ * up in the zfs_get_done() callback.
+ */
+ if (error == EINPROGRESS)
+ return (0);
+ dmu_buf_rele(db, zgd);
+ kmem_free(zgd, sizeof (zgd_t));
+ }
+out:
+ zfs_range_unlock(rl);
+ VN_RELE(ZTOV(zp));
+ return (error);
+}
+
+/*ARGSUSED*/
+static int
+zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if (flag & V_ACE_MASK)
+ error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
+ else
+ error = zfs_zaccess_rwx(zp, mode, flag, cr);
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Lookup an entry in a directory, or an extended attribute directory.
+ * If it exists, return a held vnode reference for it.
+ *
+ * IN: dvp - vnode of directory to search.
+ * nm - name of entry to lookup.
+ * pnp - full pathname to lookup [UNUSED].
+ * flags - LOOKUP_XATTR set if looking for an attribute.
+ * rdir - root directory vnode [UNUSED].
+ * cr - credentials of caller.
+ * ct - caller context
+ * direntflags - directory lookup flags
+ * realpnp - returned pathname.
+ *
+ * OUT: vpp - vnode of located entry, NULL if not found.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * NA
+ */
+/* ARGSUSED */
+static int
+zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp,
+ int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
+ int *direntflags, pathname_t *realpnp)
+{
+ znode_t *zdp = VTOZ(dvp);
+ zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zdp);
+
+ *vpp = NULL;
+
+ if (flags & LOOKUP_XATTR) {
+ /*
+ * If the xattr property is off, refuse the lookup request.
+ */
+ if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * We don't allow recursive attributes..
+ * Maybe someday we will.
+ */
+ if (zdp->z_phys->zp_flags & ZFS_XATTR) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ /*
+ * Do we have permission to get into attribute directory?
+ */
+
+ if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
+ B_FALSE, cr)) {
+ VN_RELE(*vpp);
+ *vpp = NULL;
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (dvp->v_type != VDIR) {
+ ZFS_EXIT(zfsvfs);
+ return (ENOTDIR);
+ }
+
+ /*
+ * Check accessibility of directory.
+ */
+
+ if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
+ NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+
+ error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
+ if (error == 0) {
+ /*
+ * Convert device special files
+ */
+ if (IS_DEVVP(*vpp)) {
+ vnode_t *svp;
+
+ svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
+ VN_RELE(*vpp);
+ if (svp == NULL)
+ error = ENOSYS;
+ else
+ *vpp = svp;
+ }
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Attempt to create a new entry in a directory. If the entry
+ * already exists, truncate the file if permissible, else return
+ * an error. Return the vp of the created or trunc'd file.
+ *
+ * IN: dvp - vnode of directory to put new file entry in.
+ * name - name of new file entry.
+ * vap - attributes of new file.
+ * excl - flag indicating exclusive or non-exclusive mode.
+ * mode - mode to open file with.
+ * cr - credentials of caller.
+ * flag - large file flag [UNUSED].
+ * ct - caller context
+ * vsecp - ACL to be set
+ *
+ * OUT: vpp - vnode of created or trunc'd entry.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * dvp - ctime|mtime updated if new entry created
+ * vp - ctime|mtime always, atime if new
+ */
+
+/* ARGSUSED */
+static int
+zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl,
+ int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct,
+ vsecattr_t *vsecp)
+{
+ znode_t *zp, *dzp = VTOZ(dvp);
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ objset_t *os;
+ zfs_dirlock_t *dl;
+ dmu_tx_t *tx;
+ int error;
+ zfs_acl_t *aclp = NULL;
+ zfs_fuid_info_t *fuidp = NULL;
+ ksid_t *ksid;
+ uid_t uid;
+ gid_t gid = crgetgid(cr);
+
+ /*
+ * If we have an ephemeral id, ACL, or XVATTR then
+ * make sure file system is at proper version
+ */
+
+ ksid = crgetsid(cr, KSID_OWNER);
+ if (ksid)
+ uid = ksid_getid(ksid);
+ else
+ uid = crgetuid(cr);
+
+ if (zfsvfs->z_use_fuids == B_FALSE &&
+ (vsecp || (vap->va_mask & AT_XVATTR) ||
+ IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
+ return (EINVAL);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ os = zfsvfs->z_os;
+ zilog = zfsvfs->z_log;
+
+ if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
+ NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+
+ if (vap->va_mask & AT_XVATTR) {
+ if ((error = secpolicy_xvattr((xvattr_t *)vap,
+ crgetuid(cr), cr, vap->va_type)) != 0) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ }
+top:
+ *vpp = NULL;
+
+ if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr))
+ vap->va_mode &= ~VSVTX;
+
+ if (*name == '\0') {
+ /*
+ * Null component name refers to the directory itself.
+ */
+ VN_HOLD(dvp);
+ zp = dzp;
+ dl = NULL;
+ error = 0;
+ } else {
+ /* possible VN_HOLD(zp) */
+ int zflg = 0;
+
+ if (flag & FIGNORECASE)
+ zflg |= ZCILOOK;
+
+ error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
+ NULL, NULL);
+ if (error) {
+ if (strcmp(name, "..") == 0)
+ error = EISDIR;
+ ZFS_EXIT(zfsvfs);
+ if (aclp)
+ zfs_acl_free(aclp);
+ return (error);
+ }
+ }
+ if (vsecp && aclp == NULL) {
+ error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
+ if (error) {
+ ZFS_EXIT(zfsvfs);
+ if (dl)
+ zfs_dirent_unlock(dl);
+ return (error);
+ }
+ }
+
+ if (zp == NULL) {
+ uint64_t txtype;
+
+ /*
+ * Create a new file object and update the directory
+ * to reference it.
+ */
+ if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
+ goto out;
+ }
+
+ /*
+ * We only support the creation of regular files in
+ * extended attribute directories.
+ */
+ if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
+ (vap->va_type != VREG)) {
+ error = EINVAL;
+ goto out;
+ }
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(uid) ||
+ IS_EPHEMERAL(gid)) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
+ FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+ dmu_tx_hold_bonus(tx, dzp->z_id);
+ dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
+ if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp) {
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
+ 0, SPA_MAXBLOCKSIZE);
+ }
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ if (error == ERESTART &&
+ zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ if (aclp)
+ zfs_acl_free(aclp);
+ return (error);
+ }
+ zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
+ (void) zfs_link_create(dl, zp, tx, ZNEW);
+ txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
+ if (flag & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_create(zilog, tx, txtype, dzp, zp, name,
+ vsecp, fuidp, vap);
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+ dmu_tx_commit(tx);
+ } else {
+ int aflags = (flag & FAPPEND) ? V_APPEND : 0;
+
+ /*
+ * A directory entry already exists for this name.
+ */
+ /*
+ * Can't truncate an existing file if in exclusive mode.
+ */
+ if (excl == EXCL) {
+ error = EEXIST;
+ goto out;
+ }
+ /*
+ * Can't open a directory for writing.
+ */
+ if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
+ error = EISDIR;
+ goto out;
+ }
+ /*
+ * Verify requested access to file.
+ */
+ if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
+ goto out;
+ }
+
+ mutex_enter(&dzp->z_lock);
+ dzp->z_seq++;
+ mutex_exit(&dzp->z_lock);
+
+ /*
+ * Truncate regular files if requested.
+ */
+ if ((ZTOV(zp)->v_type == VREG) &&
+ (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
+ /* we can't hold any locks when calling zfs_freesp() */
+ zfs_dirent_unlock(dl);
+ dl = NULL;
+ error = zfs_freesp(zp, 0, 0, mode, TRUE);
+ if (error == 0) {
+ vnevent_create(ZTOV(zp), ct);
+ }
+ }
+ }
+out:
+
+ if (dl)
+ zfs_dirent_unlock(dl);
+
+ if (error) {
+ if (zp)
+ VN_RELE(ZTOV(zp));
+ } else {
+ *vpp = ZTOV(zp);
+ /*
+ * If vnode is for a device return a specfs vnode instead.
+ */
+ if (IS_DEVVP(*vpp)) {
+ struct vnode *svp;
+
+ svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
+ VN_RELE(*vpp);
+ if (svp == NULL) {
+ error = ENOSYS;
+ }
+ *vpp = svp;
+ }
+ }
+ if (aclp)
+ zfs_acl_free(aclp);
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Remove an entry from a directory.
+ *
+ * IN: dvp - vnode of directory to remove entry from.
+ * name - name of entry to remove.
+ * cr - credentials of caller.
+ * ct - caller context
+ * flags - case flags
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * dvp - ctime|mtime
+ * vp - ctime (if nlink > 0)
+ */
+/*ARGSUSED*/
+static int
+zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
+ int flags)
+{
+ znode_t *zp, *dzp = VTOZ(dvp);
+ znode_t *xzp = NULL;
+ vnode_t *vp;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ uint64_t acl_obj, xattr_obj;
+ zfs_dirlock_t *dl;
+ dmu_tx_t *tx;
+ boolean_t may_delete_now, delete_now = FALSE;
+ boolean_t unlinked, toobig = FALSE;
+ uint64_t txtype;
+ pathname_t *realnmp = NULL;
+ pathname_t realnm;
+ int error;
+ int zflg = ZEXISTS;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ zilog = zfsvfs->z_log;
+
+ if (flags & FIGNORECASE) {
+ zflg |= ZCILOOK;
+ pn_alloc(&realnm);
+ realnmp = &realnm;
+ }
+
+top:
+ /*
+ * Attempt to lock directory; fail if entry doesn't exist.
+ */
+ if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
+ NULL, realnmp)) {
+ if (realnmp)
+ pn_free(realnmp);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ vp = ZTOV(zp);
+
+ if (error = zfs_zaccess_delete(dzp, zp, cr)) {
+ goto out;
+ }
+
+ /*
+ * Need to use rmdir for removing directories.
+ */
+ if (vp->v_type == VDIR) {
+ error = EPERM;
+ goto out;
+ }
+
+ vnevent_remove(vp, dvp, name, ct);
+
+ if (realnmp)
+ dnlc_remove(dvp, realnmp->pn_buf);
+ else
+ dnlc_remove(dvp, name);
+
+ mutex_enter(&vp->v_lock);
+ may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp);
+ mutex_exit(&vp->v_lock);
+
+ /*
+ * We may delete the znode now, or we may put it in the unlinked set;
+ * it depends on whether we're the last link, and on whether there are
+ * other holds on the vnode. So we dmu_tx_hold() the right things to
+ * allow for either case.
+ */
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ if (may_delete_now) {
+ toobig =
+ zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
+ /* if the file is too big, only hold_free a token amount */
+ dmu_tx_hold_free(tx, zp->z_id, 0,
+ (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
+ }
+
+ /* are there any extended attributes? */
+ if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
+ /* XXX - do we need this if we are deleting? */
+ dmu_tx_hold_bonus(tx, xattr_obj);
+ }
+
+ /* are there any additional acls */
+ if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
+ may_delete_now)
+ dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
+
+ /* charge as an update -- would be nice not to charge at all */
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
+
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ VN_RELE(vp);
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ if (realnmp)
+ pn_free(realnmp);
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ /*
+ * Remove the directory entry.
+ */
+ error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
+
+ if (error) {
+ dmu_tx_commit(tx);
+ goto out;
+ }
+
+ if (unlinked) {
+ mutex_enter(&vp->v_lock);
+ delete_now = may_delete_now && !toobig &&
+ vp->v_count == 1 && !vn_has_cached_data(vp) &&
+ zp->z_phys->zp_xattr == xattr_obj &&
+ zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
+ mutex_exit(&vp->v_lock);
+ }
+
+ if (delete_now) {
+ if (zp->z_phys->zp_xattr) {
+ error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
+ ASSERT3U(error, ==, 0);
+ ASSERT3U(xzp->z_phys->zp_links, ==, 2);
+ dmu_buf_will_dirty(xzp->z_dbuf, tx);
+ mutex_enter(&xzp->z_lock);
+ xzp->z_unlinked = 1;
+ xzp->z_phys->zp_links = 0;
+ mutex_exit(&xzp->z_lock);
+ zfs_unlinked_add(xzp, tx);
+ zp->z_phys->zp_xattr = 0; /* probably unnecessary */
+ }
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&vp->v_lock);
+ vp->v_count--;
+ ASSERT3U(vp->v_count, ==, 0);
+ mutex_exit(&vp->v_lock);
+ mutex_exit(&zp->z_lock);
+ zfs_znode_delete(zp, tx);
+ } else if (unlinked) {
+ zfs_unlinked_add(zp, tx);
+ }
+
+ txtype = TX_REMOVE;
+ if (flags & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_remove(zilog, tx, txtype, dzp, name);
+
+ dmu_tx_commit(tx);
+out:
+ if (realnmp)
+ pn_free(realnmp);
+
+ zfs_dirent_unlock(dl);
+
+ if (!delete_now) {
+ VN_RELE(vp);
+ } else if (xzp) {
+ /* this rele is delayed to prevent nesting transactions */
+ VN_RELE(ZTOV(xzp));
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Create a new directory and insert it into dvp using the name
+ * provided. Return a pointer to the inserted directory.
+ *
+ * IN: dvp - vnode of directory to add subdir to.
+ * dirname - name of new directory.
+ * vap - attributes of new directory.
+ * cr - credentials of caller.
+ * ct - caller context
+ * vsecp - ACL to be set
+ *
+ * OUT: vpp - vnode of created directory.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * dvp - ctime|mtime updated
+ * vp - ctime|mtime|atime updated
+ */
+/*ARGSUSED*/
+static int
+zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
+ caller_context_t *ct, int flags, vsecattr_t *vsecp)
+{
+ znode_t *zp, *dzp = VTOZ(dvp);
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ zfs_dirlock_t *dl;
+ uint64_t txtype;
+ dmu_tx_t *tx;
+ int error;
+ zfs_acl_t *aclp = NULL;
+ zfs_fuid_info_t *fuidp = NULL;
+ int zf = ZNEW;
+ ksid_t *ksid;
+ uid_t uid;
+ gid_t gid = crgetgid(cr);
+
+ ASSERT(vap->va_type == VDIR);
+
+ /*
+ * If we have an ephemeral id, ACL, or XVATTR then
+ * make sure file system is at proper version
+ */
+
+ ksid = crgetsid(cr, KSID_OWNER);
+ if (ksid)
+ uid = ksid_getid(ksid);
+ else
+ uid = crgetuid(cr);
+ if (zfsvfs->z_use_fuids == B_FALSE &&
+ (vsecp || (vap->va_mask & AT_XVATTR) ||
+ IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
+ return (EINVAL);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ zilog = zfsvfs->z_log;
+
+ if (dzp->z_phys->zp_flags & ZFS_XATTR) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ if (zfsvfs->z_utf8 && u8_validate(dirname,
+ strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+ if (flags & FIGNORECASE)
+ zf |= ZCILOOK;
+
+ if (vap->va_mask & AT_XVATTR)
+ if ((error = secpolicy_xvattr((xvattr_t *)vap,
+ crgetuid(cr), cr, vap->va_type)) != 0) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ /*
+ * First make sure the new directory doesn't exist.
+ */
+top:
+ *vpp = NULL;
+
+ if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
+ NULL, NULL)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
+ zfs_dirent_unlock(dl);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (vsecp && aclp == NULL) {
+ error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ }
+ /*
+ * Add a new entry to the directory.
+ */
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
+ dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
+ if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(uid) ||
+ IS_EPHEMERAL(gid)) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+ if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp)
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
+ 0, SPA_MAXBLOCKSIZE);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ if (aclp)
+ zfs_acl_free(aclp);
+ return (error);
+ }
+
+ /*
+ * Create new node.
+ */
+ zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
+
+ if (aclp)
+ zfs_acl_free(aclp);
+
+ /*
+ * Now put new name in parent dir.
+ */
+ (void) zfs_link_create(dl, zp, tx, ZNEW);
+
+ *vpp = ZTOV(zp);
+
+ txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
+ if (flags & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, fuidp, vap);
+
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+ dmu_tx_commit(tx);
+
+ zfs_dirent_unlock(dl);
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/*
+ * Remove a directory subdir entry. If the current working
+ * directory is the same as the subdir to be removed, the
+ * remove will fail.
+ *
+ * IN: dvp - vnode of directory to remove from.
+ * name - name of directory to be removed.
+ * cwd - vnode of current working directory.
+ * cr - credentials of caller.
+ * ct - caller context
+ * flags - case flags
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * dvp - ctime|mtime updated
+ */
+/*ARGSUSED*/
+static int
+zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
+ caller_context_t *ct, int flags)
+{
+ znode_t *dzp = VTOZ(dvp);
+ znode_t *zp;
+ vnode_t *vp;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ zfs_dirlock_t *dl;
+ dmu_tx_t *tx;
+ int error;
+ int zflg = ZEXISTS;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ zilog = zfsvfs->z_log;
+
+ if (flags & FIGNORECASE)
+ zflg |= ZCILOOK;
+top:
+ zp = NULL;
+
+ /*
+ * Attempt to lock directory; fail if entry doesn't exist.
+ */
+ if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
+ NULL, NULL)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ vp = ZTOV(zp);
+
+ if (error = zfs_zaccess_delete(dzp, zp, cr)) {
+ goto out;
+ }
+
+ if (vp->v_type != VDIR) {
+ error = ENOTDIR;
+ goto out;
+ }
+
+ if (vp == cwd) {
+ error = EINVAL;
+ goto out;
+ }
+
+ vnevent_rmdir(vp, dvp, name, ct);
+
+ /*
+ * Grab a lock on the directory to make sure that noone is
+ * trying to add (or lookup) entries while we are removing it.
+ */
+ rw_enter(&zp->z_name_lock, RW_WRITER);
+
+ /*
+ * Grab a lock on the parent pointer to make sure we play well
+ * with the treewalk and directory rename code.
+ */
+ rw_enter(&zp->z_parent_lock, RW_WRITER);
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ rw_exit(&zp->z_parent_lock);
+ rw_exit(&zp->z_name_lock);
+ zfs_dirent_unlock(dl);
+ VN_RELE(vp);
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
+
+ if (error == 0) {
+ uint64_t txtype = TX_RMDIR;
+ if (flags & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_remove(zilog, tx, txtype, dzp, name);
+ }
+
+ dmu_tx_commit(tx);
+
+ rw_exit(&zp->z_parent_lock);
+ rw_exit(&zp->z_name_lock);
+out:
+ zfs_dirent_unlock(dl);
+
+ VN_RELE(vp);
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Read as many directory entries as will fit into the provided
+ * buffer from the given directory cursor position (specified in
+ * the uio structure.
+ *
+ * IN: vp - vnode of directory to read.
+ * uio - structure supplying read location, range info,
+ * and return buffer.
+ * cr - credentials of caller.
+ * ct - caller context
+ * flags - case flags
+ *
+ * OUT: uio - updated offset and range, buffer filled.
+ * eofp - set to true if end-of-file detected.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - atime updated
+ *
+ * Note that the low 4 bits of the cookie returned by zap is always zero.
+ * This allows us to use the low range for "special" directory entries:
+ * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
+ * we use the offset 2 for the '.zfs' directory.
+ */
+/* ARGSUSED */
+static int
+zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
+ caller_context_t *ct, int flags)
+{
+ znode_t *zp = VTOZ(vp);
+ iovec_t *iovp;
+ edirent_t *eodp;
+ dirent64_t *odp;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ objset_t *os;
+ caddr_t outbuf;
+ size_t bufsize;
+ zap_cursor_t zc;
+ zap_attribute_t zap;
+ uint_t bytes_wanted;
+ uint64_t offset; /* must be unsigned; checks for < 1 */
+ int local_eof;
+ int outcount;
+ int error;
+ uint8_t prefetch;
+ boolean_t check_sysattrs;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ /*
+ * If we are not given an eof variable,
+ * use a local one.
+ */
+ if (eofp == NULL)
+ eofp = &local_eof;
+
+ /*
+ * Check for valid iov_len.
+ */
+ if (uio->uio_iov->iov_len <= 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * Quit if directory has been removed (posix)
+ */
+ if ((*eofp = zp->z_unlinked) != 0) {
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+
+ error = 0;
+ os = zfsvfs->z_os;
+ offset = uio->uio_loffset;
+ prefetch = zp->z_zn_prefetch;
+
+ /*
+ * Initialize the iterator cursor.
+ */
+ if (offset <= 3) {
+ /*
+ * Start iteration from the beginning of the directory.
+ */
+ zap_cursor_init(&zc, os, zp->z_id);
+ } else {
+ /*
+ * The offset is a serialized cursor.
+ */
+ zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
+ }
+
+ /*
+ * Get space to change directory entries into fs independent format.
+ */
+ iovp = uio->uio_iov;
+ bytes_wanted = iovp->iov_len;
+ if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
+ bufsize = bytes_wanted;
+ outbuf = kmem_alloc(bufsize, KM_SLEEP);
+ odp = (struct dirent64 *)outbuf;
+ } else {
+ bufsize = bytes_wanted;
+ odp = (struct dirent64 *)iovp->iov_base;
+ }
+ eodp = (struct edirent *)odp;
+
+ /*
+ * If this VFS supports the system attribute view interface; and
+ * we're looking at an extended attribute directory; and we care
+ * about normalization conflicts on this vfs; then we must check
+ * for normalization conflicts with the sysattr name space.
+ */
+ check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
+ (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
+ (flags & V_RDDIR_ENTFLAGS);
+
+ /*
+ * Transform to file-system independent format
+ */
+ outcount = 0;
+ while (outcount < bytes_wanted) {
+ ino64_t objnum;
+ ushort_t reclen;
+ off64_t *next;
+
+ /*
+ * Special case `.', `..', and `.zfs'.
+ */
+ if (offset == 0) {
+ (void) strcpy(zap.za_name, ".");
+ zap.za_normalization_conflict = 0;
+ objnum = zp->z_id;
+ } else if (offset == 1) {
+ (void) strcpy(zap.za_name, "..");
+ zap.za_normalization_conflict = 0;
+ objnum = zp->z_phys->zp_parent;
+ } else if (offset == 2 && zfs_show_ctldir(zp)) {
+ (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
+ zap.za_normalization_conflict = 0;
+ objnum = ZFSCTL_INO_ROOT;
+ } else {
+ /*
+ * Grab next entry.
+ */
+ if (error = zap_cursor_retrieve(&zc, &zap)) {
+ if ((*eofp = (error == ENOENT)) != 0)
+ break;
+ else
+ goto update;
+ }
+
+ if (zap.za_integer_length != 8 ||
+ zap.za_num_integers != 1) {
+ cmn_err(CE_WARN, "zap_readdir: bad directory "
+ "entry, obj = %lld, offset = %lld\n",
+ (u_longlong_t)zp->z_id,
+ (u_longlong_t)offset);
+ error = ENXIO;
+ goto update;
+ }
+
+ objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
+ /*
+ * MacOS X can extract the object type here such as:
+ * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
+ */
+
+ if (check_sysattrs && !zap.za_normalization_conflict) {
+ zap.za_normalization_conflict =
+ xattr_sysattr_casechk(zap.za_name);
+ }
+ }
+
+ if (flags & V_RDDIR_ENTFLAGS)
+ reclen = EDIRENT_RECLEN(strlen(zap.za_name));
+ else
+ reclen = DIRENT64_RECLEN(strlen(zap.za_name));
+
+ /*
+ * Will this entry fit in the buffer?
+ */
+ if (outcount + reclen > bufsize) {
+ /*
+ * Did we manage to fit anything in the buffer?
+ */
+ if (!outcount) {
+ error = EINVAL;
+ goto update;
+ }
+ break;
+ }
+ if (flags & V_RDDIR_ENTFLAGS) {
+ /*
+ * Add extended flag entry:
+ */
+ eodp->ed_ino = objnum;
+ eodp->ed_reclen = reclen;
+ /* NOTE: ed_off is the offset for the *next* entry */
+ next = &(eodp->ed_off);
+ eodp->ed_eflags = zap.za_normalization_conflict ?
+ ED_CASE_CONFLICT : 0;
+ (void) strncpy(eodp->ed_name, zap.za_name,
+ EDIRENT_NAMELEN(reclen));
+ eodp = (edirent_t *)((intptr_t)eodp + reclen);
+ } else {
+ /*
+ * Add normal entry:
+ */
+ odp->d_ino = objnum;
+ odp->d_reclen = reclen;
+ /* NOTE: d_off is the offset for the *next* entry */
+ next = &(odp->d_off);
+ (void) strncpy(odp->d_name, zap.za_name,
+ DIRENT64_NAMELEN(reclen));
+ odp = (dirent64_t *)((intptr_t)odp + reclen);
+ }
+ outcount += reclen;
+
+ ASSERT(outcount <= bufsize);
+
+ /* Prefetch znode */
+ if (prefetch)
+ dmu_prefetch(os, objnum, 0, 0);
+
+ /*
+ * Move to the next entry, fill in the previous offset.
+ */
+ if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
+ zap_cursor_advance(&zc);
+ offset = zap_cursor_serialize(&zc);
+ } else {
+ offset += 1;
+ }
+ *next = offset;
+ }
+ zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
+
+ if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
+ iovp->iov_base += outcount;
+ iovp->iov_len -= outcount;
+ uio->uio_resid -= outcount;
+ } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
+ /*
+ * Reset the pointer.
+ */
+ offset = uio->uio_loffset;
+ }
+
+update:
+ zap_cursor_fini(&zc);
+ if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
+ kmem_free(outbuf, bufsize);
+
+ if (error == ENOENT)
+ error = 0;
+
+ ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
+
+ uio->uio_loffset = offset;
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+ulong_t zfs_fsync_sync_cnt = 4;
+
+static int
+zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ /*
+ * Regardless of whether this is required for standards conformance,
+ * this is the logical behavior when fsync() is called on a file with
+ * dirty pages. We use B_ASYNC since the ZIL transactions are already
+ * going to be pushed out as part of the zil_commit().
+ */
+ if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) &&
+ (vp->v_type == VREG) && !(IS_SWAPVP(vp)))
+ (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct);
+
+ (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+
+/*
+ * Get the requested file attributes and place them in the provided
+ * vattr structure.
+ *
+ * IN: vp - vnode of file.
+ * vap - va_mask identifies requested attributes.
+ * If AT_XVATTR set, then optional attrs are requested
+ * flags - ATTR_NOACLCHECK (CIFS server context)
+ * cr - credentials of caller.
+ * ct - caller context
+ *
+ * OUT: vap - attribute values.
+ *
+ * RETURN: 0 (always succeeds)
+ */
+/* ARGSUSED */
+static int
+zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ znode_phys_t *pzp;
+ int error = 0;
+ uint64_t links;
+ xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
+ xoptattr_t *xoap = NULL;
+ boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ pzp = zp->z_phys;
+
+ mutex_enter(&zp->z_lock);
+
+ /*
+ * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
+ * Also, if we are the owner don't bother, since owner should
+ * always be allowed to read basic attributes of file.
+ */
+ if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
+ (pzp->zp_uid != crgetuid(cr))) {
+ if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
+ skipaclchk, cr)) {
+ mutex_exit(&zp->z_lock);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ }
+
+ /*
+ * Return all attributes. It's cheaper to provide the answer
+ * than to determine whether we were asked the question.
+ */
+
+ vap->va_type = vp->v_type;
+ vap->va_mode = pzp->zp_mode & MODEMASK;
+ zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
+ vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
+ vap->va_nodeid = zp->z_id;
+ if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
+ links = pzp->zp_links + 1;
+ else
+ links = pzp->zp_links;
+ vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
+ vap->va_size = pzp->zp_size;
+ vap->va_rdev = vp->v_rdev;
+ vap->va_seq = zp->z_seq;
+
+ /*
+ * Add in any requested optional attributes and the create time.
+ * Also set the corresponding bits in the returned attribute bitmap.
+ */
+ if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
+ if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
+ xoap->xoa_archive =
+ ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
+ XVA_SET_RTN(xvap, XAT_ARCHIVE);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
+ xoap->xoa_readonly =
+ ((pzp->zp_flags & ZFS_READONLY) != 0);
+ XVA_SET_RTN(xvap, XAT_READONLY);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
+ xoap->xoa_system =
+ ((pzp->zp_flags & ZFS_SYSTEM) != 0);
+ XVA_SET_RTN(xvap, XAT_SYSTEM);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
+ xoap->xoa_hidden =
+ ((pzp->zp_flags & ZFS_HIDDEN) != 0);
+ XVA_SET_RTN(xvap, XAT_HIDDEN);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
+ xoap->xoa_nounlink =
+ ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
+ XVA_SET_RTN(xvap, XAT_NOUNLINK);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
+ xoap->xoa_immutable =
+ ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
+ XVA_SET_RTN(xvap, XAT_IMMUTABLE);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
+ xoap->xoa_appendonly =
+ ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
+ XVA_SET_RTN(xvap, XAT_APPENDONLY);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
+ xoap->xoa_nodump =
+ ((pzp->zp_flags & ZFS_NODUMP) != 0);
+ XVA_SET_RTN(xvap, XAT_NODUMP);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
+ xoap->xoa_opaque =
+ ((pzp->zp_flags & ZFS_OPAQUE) != 0);
+ XVA_SET_RTN(xvap, XAT_OPAQUE);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
+ xoap->xoa_av_quarantined =
+ ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
+ XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
+ xoap->xoa_av_modified =
+ ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
+ XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
+ vp->v_type == VREG &&
+ (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
+ size_t len;
+ dmu_object_info_t doi;
+
+ /*
+ * Only VREG files have anti-virus scanstamps, so we
+ * won't conflict with symlinks in the bonus buffer.
+ */
+ dmu_object_info_from_db(zp->z_dbuf, &doi);
+ len = sizeof (xoap->xoa_av_scanstamp) +
+ sizeof (znode_phys_t);
+ if (len <= doi.doi_bonus_size) {
+ /*
+ * pzp points to the start of the
+ * znode_phys_t. pzp + 1 points to the
+ * first byte after the znode_phys_t.
+ */
+ (void) memcpy(xoap->xoa_av_scanstamp,
+ pzp + 1,
+ sizeof (xoap->xoa_av_scanstamp));
+ XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
+ }
+ }
+
+ if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
+ ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
+ XVA_SET_RTN(xvap, XAT_CREATETIME);
+ }
+ }
+
+ ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
+ ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
+ ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
+
+ mutex_exit(&zp->z_lock);
+
+ dmu_object_size_from_db(zp->z_dbuf, &vap->va_blksize, &vap->va_nblocks);
+
+ if (zp->z_blksz == 0) {
+ /*
+ * Block size hasn't been set; suggest maximal I/O transfers.
+ */
+ vap->va_blksize = zfsvfs->z_max_blksz;
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+/*
+ * Set the file attributes to the values contained in the
+ * vattr structure.
+ *
+ * IN: vp - vnode of file to be modified.
+ * vap - new attribute values.
+ * If AT_XVATTR set, then optional attrs are being set
+ * flags - ATTR_UTIME set if non-default time values provided.
+ * - ATTR_NOACLCHECK (CIFS context only).
+ * cr - credentials of caller.
+ * ct - caller context
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - ctime updated, mtime updated if size changed.
+ */
+/* ARGSUSED */
+static int
+zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ znode_phys_t *pzp;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zilog_t *zilog;
+ dmu_tx_t *tx;
+ vattr_t oldva;
+ uint_t mask = vap->va_mask;
+ uint_t saved_mask;
+ int trim_mask = 0;
+ uint64_t new_mode;
+ znode_t *attrzp;
+ int need_policy = FALSE;
+ int err;
+ zfs_fuid_info_t *fuidp = NULL;
+ xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
+ xoptattr_t *xoap;
+ zfs_acl_t *aclp = NULL;
+ boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
+
+ if (mask == 0)
+ return (0);
+
+ if (mask & AT_NOSET)
+ return (EINVAL);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ pzp = zp->z_phys;
+ zilog = zfsvfs->z_log;
+
+ /*
+ * Make sure that if we have ephemeral uid/gid or xvattr specified
+ * that file system is at proper version level
+ */
+
+ if (zfsvfs->z_use_fuids == B_FALSE &&
+ (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
+ ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
+ (mask & AT_XVATTR))) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ if (mask & AT_SIZE && vp->v_type == VDIR) {
+ ZFS_EXIT(zfsvfs);
+ return (EISDIR);
+ }
+
+ if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * If this is an xvattr_t, then get a pointer to the structure of
+ * optional attributes. If this is NULL, then we have a vattr_t.
+ */
+ xoap = xva_getxoptattr(xvap);
+
+ /*
+ * Immutable files can only alter immutable bit and atime
+ */
+ if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
+ ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
+ ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ /*
+ * Verify timestamps doesn't overflow 32 bits.
+ * ZFS can handle large timestamps, but 32bit syscalls can't
+ * handle times greater than 2039. This check should be removed
+ * once large timestamps are fully supported.
+ */
+ if (mask & (AT_ATIME | AT_MTIME)) {
+ if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
+ ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
+ ZFS_EXIT(zfsvfs);
+ return (EOVERFLOW);
+ }
+ }
+
+top:
+ attrzp = NULL;
+
+ if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
+ ZFS_EXIT(zfsvfs);
+ return (EROFS);
+ }
+
+ /*
+ * First validate permissions
+ */
+
+ if (mask & AT_SIZE) {
+ err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
+ if (err) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ /*
+ * XXX - Note, we are not providing any open
+ * mode flags here (like FNDELAY), so we may
+ * block if there are locks present... this
+ * should be addressed in openat().
+ */
+ /* XXX - would it be OK to generate a log record here? */
+ err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
+ if (err) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ }
+
+ if (mask & (AT_ATIME|AT_MTIME) ||
+ ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
+ XVA_ISSET_REQ(xvap, XAT_READONLY) ||
+ XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
+ XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
+ XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
+ need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
+ skipaclchk, cr);
+
+ if (mask & (AT_UID|AT_GID)) {
+ int idmask = (mask & (AT_UID|AT_GID));
+ int take_owner;
+ int take_group;
+
+ /*
+ * NOTE: even if a new mode is being set,
+ * we may clear S_ISUID/S_ISGID bits.
+ */
+
+ if (!(mask & AT_MODE))
+ vap->va_mode = pzp->zp_mode;
+
+ /*
+ * Take ownership or chgrp to group we are a member of
+ */
+
+ take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
+ take_group = (mask & AT_GID) &&
+ zfs_groupmember(zfsvfs, vap->va_gid, cr);
+
+ /*
+ * If both AT_UID and AT_GID are set then take_owner and
+ * take_group must both be set in order to allow taking
+ * ownership.
+ *
+ * Otherwise, send the check through secpolicy_vnode_setattr()
+ *
+ */
+
+ if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
+ ((idmask == AT_UID) && take_owner) ||
+ ((idmask == AT_GID) && take_group)) {
+ if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
+ skipaclchk, cr) == 0) {
+ /*
+ * Remove setuid/setgid for non-privileged users
+ */
+ secpolicy_setid_clear(vap, cr);
+ trim_mask = (mask & (AT_UID|AT_GID));
+ } else {
+ need_policy = TRUE;
+ }
+ } else {
+ need_policy = TRUE;
+ }
+ }
+
+ mutex_enter(&zp->z_lock);
+ oldva.va_mode = pzp->zp_mode;
+ zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
+ if (mask & AT_XVATTR) {
+ if ((need_policy == FALSE) &&
+ (XVA_ISSET_REQ(xvap, XAT_APPENDONLY) &&
+ xoap->xoa_appendonly !=
+ ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) ||
+ (XVA_ISSET_REQ(xvap, XAT_NOUNLINK) &&
+ xoap->xoa_nounlink !=
+ ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) ||
+ (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE) &&
+ xoap->xoa_immutable !=
+ ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) ||
+ (XVA_ISSET_REQ(xvap, XAT_NODUMP) &&
+ xoap->xoa_nodump !=
+ ((pzp->zp_flags & ZFS_NODUMP) != 0)) ||
+ (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED) &&
+ xoap->xoa_av_modified !=
+ ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) ||
+ ((XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED) &&
+ ((vp->v_type != VREG && xoap->xoa_av_quarantined) ||
+ xoap->xoa_av_quarantined !=
+ ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)))) ||
+ (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ||
+ (XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
+ need_policy = TRUE;
+ }
+ }
+
+ mutex_exit(&zp->z_lock);
+
+ if (mask & AT_MODE) {
+ if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
+ err = secpolicy_setid_setsticky_clear(vp, vap,
+ &oldva, cr);
+ if (err) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ trim_mask |= AT_MODE;
+ } else {
+ need_policy = TRUE;
+ }
+ }
+
+ if (need_policy) {
+ /*
+ * If trim_mask is set then take ownership
+ * has been granted or write_acl is present and user
+ * has the ability to modify mode. In that case remove
+ * UID|GID and or MODE from mask so that
+ * secpolicy_vnode_setattr() doesn't revoke it.
+ */
+
+ if (trim_mask) {
+ saved_mask = vap->va_mask;
+ vap->va_mask &= ~trim_mask;
+ }
+ err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
+ (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
+ if (err) {
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+
+ if (trim_mask)
+ vap->va_mask |= saved_mask;
+ }
+
+ /*
+ * secpolicy_vnode_setattr, or take ownership may have
+ * changed va_mask
+ */
+ mask = vap->va_mask;
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ if (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
+ ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid))) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+
+ if (mask & AT_MODE) {
+ uint64_t pmode = pzp->zp_mode;
+
+ new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
+
+ if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) {
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+ if (pzp->zp_acl.z_acl_extern_obj) {
+ /* Are we upgrading ACL from old V0 format to new V1 */
+ if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
+ pzp->zp_acl.z_acl_version ==
+ ZFS_ACL_VERSION_INITIAL) {
+ dmu_tx_hold_free(tx,
+ pzp->zp_acl.z_acl_extern_obj, 0,
+ DMU_OBJECT_END);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
+ 0, aclp->z_acl_bytes);
+ } else {
+ dmu_tx_hold_write(tx,
+ pzp->zp_acl.z_acl_extern_obj, 0,
+ aclp->z_acl_bytes);
+ }
+ } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
+ 0, aclp->z_acl_bytes);
+ }
+ }
+
+ if ((mask & (AT_UID | AT_GID)) && pzp->zp_xattr != 0) {
+ err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
+ if (err) {
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ if (aclp)
+ zfs_acl_free(aclp);
+ return (err);
+ }
+ dmu_tx_hold_bonus(tx, attrzp->z_id);
+ }
+
+ err = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (err) {
+ if (attrzp)
+ VN_RELE(ZTOV(attrzp));
+
+ if (aclp) {
+ zfs_acl_free(aclp);
+ aclp = NULL;
+ }
+
+ if (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (err);
+ }
+
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+
+ /*
+ * Set each attribute requested.
+ * We group settings according to the locks they need to acquire.
+ *
+ * Note: you cannot set ctime directly, although it will be
+ * updated as a side-effect of calling this function.
+ */
+
+ mutex_enter(&zp->z_lock);
+
+ if (mask & AT_MODE) {
+ mutex_enter(&zp->z_acl_lock);
+ zp->z_phys->zp_mode = new_mode;
+ err = zfs_aclset_common(zp, aclp, cr, &fuidp, tx);
+ ASSERT3U(err, ==, 0);
+ mutex_exit(&zp->z_acl_lock);
+ }
+
+ if (attrzp)
+ mutex_enter(&attrzp->z_lock);
+
+ if (mask & AT_UID) {
+ pzp->zp_uid = zfs_fuid_create(zfsvfs,
+ vap->va_uid, cr, ZFS_OWNER, tx, &fuidp);
+ if (attrzp) {
+ attrzp->z_phys->zp_uid = zfs_fuid_create(zfsvfs,
+ vap->va_uid, cr, ZFS_OWNER, tx, &fuidp);
+ }
+ }
+
+ if (mask & AT_GID) {
+ pzp->zp_gid = zfs_fuid_create(zfsvfs, vap->va_gid,
+ cr, ZFS_GROUP, tx, &fuidp);
+ if (attrzp)
+ attrzp->z_phys->zp_gid = zfs_fuid_create(zfsvfs,
+ vap->va_gid, cr, ZFS_GROUP, tx, &fuidp);
+ }
+
+ if (aclp)
+ zfs_acl_free(aclp);
+
+ if (attrzp)
+ mutex_exit(&attrzp->z_lock);
+
+ if (mask & AT_ATIME)
+ ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
+
+ if (mask & AT_MTIME)
+ ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
+
+ /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
+ if (mask & AT_SIZE)
+ zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
+ else if (mask != 0)
+ zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
+ /*
+ * Do this after setting timestamps to prevent timestamp
+ * update from toggling bit
+ */
+
+ if (xoap && (mask & AT_XVATTR)) {
+ if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
+ size_t len;
+ dmu_object_info_t doi;
+
+ ASSERT(vp->v_type == VREG);
+
+ /* Grow the bonus buffer if necessary. */
+ dmu_object_info_from_db(zp->z_dbuf, &doi);
+ len = sizeof (xoap->xoa_av_scanstamp) +
+ sizeof (znode_phys_t);
+ if (len > doi.doi_bonus_size)
+ VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
+ }
+ zfs_xvattr_set(zp, xvap);
+ }
+
+ if (mask != 0)
+ zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
+
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+ mutex_exit(&zp->z_lock);
+
+ if (attrzp)
+ VN_RELE(ZTOV(attrzp));
+
+ dmu_tx_commit(tx);
+
+ ZFS_EXIT(zfsvfs);
+ return (err);
+}
+
+typedef struct zfs_zlock {
+ krwlock_t *zl_rwlock; /* lock we acquired */
+ znode_t *zl_znode; /* znode we held */
+ struct zfs_zlock *zl_next; /* next in list */
+} zfs_zlock_t;
+
+/*
+ * Drop locks and release vnodes that were held by zfs_rename_lock().
+ */
+static void
+zfs_rename_unlock(zfs_zlock_t **zlpp)
+{
+ zfs_zlock_t *zl;
+
+ while ((zl = *zlpp) != NULL) {
+ if (zl->zl_znode != NULL)
+ VN_RELE(ZTOV(zl->zl_znode));
+ rw_exit(zl->zl_rwlock);
+ *zlpp = zl->zl_next;
+ kmem_free(zl, sizeof (*zl));
+ }
+}
+
+/*
+ * Search back through the directory tree, using the ".." entries.
+ * Lock each directory in the chain to prevent concurrent renames.
+ * Fail any attempt to move a directory into one of its own descendants.
+ * XXX - z_parent_lock can overlap with map or grow locks
+ */
+static int
+zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
+{
+ zfs_zlock_t *zl;
+ znode_t *zp = tdzp;
+ uint64_t rootid = zp->z_zfsvfs->z_root;
+ uint64_t *oidp = &zp->z_id;
+ krwlock_t *rwlp = &szp->z_parent_lock;
+ krw_t rw = RW_WRITER;
+
+ /*
+ * First pass write-locks szp and compares to zp->z_id.
+ * Later passes read-lock zp and compare to zp->z_parent.
+ */
+ do {
+ if (!rw_tryenter(rwlp, rw)) {
+ /*
+ * Another thread is renaming in this path.
+ * Note that if we are a WRITER, we don't have any
+ * parent_locks held yet.
+ */
+ if (rw == RW_READER && zp->z_id > szp->z_id) {
+ /*
+ * Drop our locks and restart
+ */
+ zfs_rename_unlock(&zl);
+ *zlpp = NULL;
+ zp = tdzp;
+ oidp = &zp->z_id;
+ rwlp = &szp->z_parent_lock;
+ rw = RW_WRITER;
+ continue;
+ } else {
+ /*
+ * Wait for other thread to drop its locks
+ */
+ rw_enter(rwlp, rw);
+ }
+ }
+
+ zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
+ zl->zl_rwlock = rwlp;
+ zl->zl_znode = NULL;
+ zl->zl_next = *zlpp;
+ *zlpp = zl;
+
+ if (*oidp == szp->z_id) /* We're a descendant of szp */
+ return (EINVAL);
+
+ if (*oidp == rootid) /* We've hit the top */
+ return (0);
+
+ if (rw == RW_READER) { /* i.e. not the first pass */
+ int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
+ if (error)
+ return (error);
+ zl->zl_znode = zp;
+ }
+ oidp = &zp->z_phys->zp_parent;
+ rwlp = &zp->z_parent_lock;
+ rw = RW_READER;
+
+ } while (zp->z_id != sdzp->z_id);
+
+ return (0);
+}
+
+/*
+ * Move an entry from the provided source directory to the target
+ * directory. Change the entry name as indicated.
+ *
+ * IN: sdvp - Source directory containing the "old entry".
+ * snm - Old entry name.
+ * tdvp - Target directory to contain the "new entry".
+ * tnm - New entry name.
+ * cr - credentials of caller.
+ * ct - caller context
+ * flags - case flags
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * sdvp,tdvp - ctime|mtime updated
+ */
+/*ARGSUSED*/
+static int
+zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
+ caller_context_t *ct, int flags)
+{
+ znode_t *tdzp, *szp, *tzp;
+ znode_t *sdzp = VTOZ(sdvp);
+ zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
+ zilog_t *zilog;
+ vnode_t *realvp;
+ zfs_dirlock_t *sdl, *tdl;
+ dmu_tx_t *tx;
+ zfs_zlock_t *zl;
+ int cmp, serr, terr;
+ int error = 0;
+ int zflg = 0;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(sdzp);
+ zilog = zfsvfs->z_log;
+
+ /*
+ * Make sure we have the real vp for the target directory.
+ */
+ if (VOP_REALVP(tdvp, &realvp, ct) == 0)
+ tdvp = realvp;
+
+ if (tdvp->v_vfsp != sdvp->v_vfsp) {
+ ZFS_EXIT(zfsvfs);
+ return (EXDEV);
+ }
+
+ tdzp = VTOZ(tdvp);
+ ZFS_VERIFY_ZP(tdzp);
+ if (zfsvfs->z_utf8 && u8_validate(tnm,
+ strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+
+ if (flags & FIGNORECASE)
+ zflg |= ZCILOOK;
+
+top:
+ szp = NULL;
+ tzp = NULL;
+ zl = NULL;
+
+ /*
+ * This is to prevent the creation of links into attribute space
+ * by renaming a linked file into/outof an attribute directory.
+ * See the comment in zfs_link() for why this is considered bad.
+ */
+ if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
+ (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * Lock source and target directory entries. To prevent deadlock,
+ * a lock ordering must be defined. We lock the directory with
+ * the smallest object id first, or if it's a tie, the one with
+ * the lexically first name.
+ */
+ if (sdzp->z_id < tdzp->z_id) {
+ cmp = -1;
+ } else if (sdzp->z_id > tdzp->z_id) {
+ cmp = 1;
+ } else {
+ /*
+ * First compare the two name arguments without
+ * considering any case folding.
+ */
+ int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
+
+ cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
+ ASSERT(error == 0 || !zfsvfs->z_utf8);
+ if (cmp == 0) {
+ /*
+ * POSIX: "If the old argument and the new argument
+ * both refer to links to the same existing file,
+ * the rename() function shall return successfully
+ * and perform no other action."
+ */
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+ /*
+ * If the file system is case-folding, then we may
+ * have some more checking to do. A case-folding file
+ * system is either supporting mixed case sensitivity
+ * access or is completely case-insensitive. Note
+ * that the file system is always case preserving.
+ *
+ * In mixed sensitivity mode case sensitive behavior
+ * is the default. FIGNORECASE must be used to
+ * explicitly request case insensitive behavior.
+ *
+ * If the source and target names provided differ only
+ * by case (e.g., a request to rename 'tim' to 'Tim'),
+ * we will treat this as a special case in the
+ * case-insensitive mode: as long as the source name
+ * is an exact match, we will allow this to proceed as
+ * a name-change request.
+ */
+ if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
+ (zfsvfs->z_case == ZFS_CASE_MIXED &&
+ flags & FIGNORECASE)) &&
+ u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
+ &error) == 0) {
+ /*
+ * case preserving rename request, require exact
+ * name matches
+ */
+ zflg |= ZCIEXACT;
+ zflg &= ~ZCILOOK;
+ }
+ }
+
+ if (cmp < 0) {
+ serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
+ ZEXISTS | zflg, NULL, NULL);
+ terr = zfs_dirent_lock(&tdl,
+ tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
+ } else {
+ terr = zfs_dirent_lock(&tdl,
+ tdzp, tnm, &tzp, zflg, NULL, NULL);
+ serr = zfs_dirent_lock(&sdl,
+ sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
+ NULL, NULL);
+ }
+
+ if (serr) {
+ /*
+ * Source entry invalid or not there.
+ */
+ if (!terr) {
+ zfs_dirent_unlock(tdl);
+ if (tzp)
+ VN_RELE(ZTOV(tzp));
+ }
+ if (strcmp(snm, "..") == 0)
+ serr = EINVAL;
+ ZFS_EXIT(zfsvfs);
+ return (serr);
+ }
+ if (terr) {
+ zfs_dirent_unlock(sdl);
+ VN_RELE(ZTOV(szp));
+ if (strcmp(tnm, "..") == 0)
+ terr = EINVAL;
+ ZFS_EXIT(zfsvfs);
+ return (terr);
+ }
+
+ /*
+ * Must have write access at the source to remove the old entry
+ * and write access at the target to create the new entry.
+ * Note that if target and source are the same, this can be
+ * done in a single check.
+ */
+
+ if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
+ goto out;
+
+ if (ZTOV(szp)->v_type == VDIR) {
+ /*
+ * Check to make sure rename is valid.
+ * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
+ */
+ if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
+ goto out;
+ }
+
+ /*
+ * Does target exist?
+ */
+ if (tzp) {
+ /*
+ * Source and target must be the same type.
+ */
+ if (ZTOV(szp)->v_type == VDIR) {
+ if (ZTOV(tzp)->v_type != VDIR) {
+ error = ENOTDIR;
+ goto out;
+ }
+ } else {
+ if (ZTOV(tzp)->v_type == VDIR) {
+ error = EISDIR;
+ goto out;
+ }
+ }
+ /*
+ * POSIX dictates that when the source and target
+ * entries refer to the same file object, rename
+ * must do nothing and exit without error.
+ */
+ if (szp->z_id == tzp->z_id) {
+ error = 0;
+ goto out;
+ }
+ }
+
+ vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
+ if (tzp)
+ vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
+
+ /*
+ * notify the target directory if it is not the same
+ * as source directory.
+ */
+ if (tdvp != sdvp) {
+ vnevent_rename_dest_dir(tdvp, ct);
+ }
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
+ dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
+ dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
+ dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
+ if (sdzp != tdzp)
+ dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
+ if (tzp)
+ dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
+ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (zl != NULL)
+ zfs_rename_unlock(&zl);
+ zfs_dirent_unlock(sdl);
+ zfs_dirent_unlock(tdl);
+ VN_RELE(ZTOV(szp));
+ if (tzp)
+ VN_RELE(ZTOV(tzp));
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (tzp) /* Attempt to remove the existing target */
+ error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
+
+ if (error == 0) {
+ error = zfs_link_create(tdl, szp, tx, ZRENAMING);
+ if (error == 0) {
+ szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
+
+ error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
+ ASSERT(error == 0);
+
+ zfs_log_rename(zilog, tx,
+ TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
+ sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
+
+ /* Update path information for the target vnode */
+ vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
+ }
+ }
+
+ dmu_tx_commit(tx);
+out:
+ if (zl != NULL)
+ zfs_rename_unlock(&zl);
+
+ zfs_dirent_unlock(sdl);
+ zfs_dirent_unlock(tdl);
+
+ VN_RELE(ZTOV(szp));
+ if (tzp)
+ VN_RELE(ZTOV(tzp));
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Insert the indicated symbolic reference entry into the directory.
+ *
+ * IN: dvp - Directory to contain new symbolic link.
+ * link - Name for new symlink entry.
+ * vap - Attributes of new entry.
+ * target - Target path of new symlink.
+ * cr - credentials of caller.
+ * ct - caller context
+ * flags - case flags
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * dvp - ctime|mtime updated
+ */
+/*ARGSUSED*/
+static int
+zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr,
+ caller_context_t *ct, int flags)
+{
+ znode_t *zp, *dzp = VTOZ(dvp);
+ zfs_dirlock_t *dl;
+ dmu_tx_t *tx;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ int len = strlen(link);
+ int error;
+ int zflg = ZNEW;
+ zfs_fuid_info_t *fuidp = NULL;
+
+ ASSERT(vap->va_type == VLNK);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ zilog = zfsvfs->z_log;
+
+ if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
+ NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+ if (flags & FIGNORECASE)
+ zflg |= ZCILOOK;
+top:
+ if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (len > MAXPATHLEN) {
+ ZFS_EXIT(zfsvfs);
+ return (ENAMETOOLONG);
+ }
+
+ /*
+ * Attempt to lock directory; fail if entry already exists.
+ */
+ error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
+ if (error) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
+ dmu_tx_hold_bonus(tx, dzp->z_id);
+ dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
+ if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE)
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
+ if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
+ if (zfsvfs->z_fuid_obj == 0) {
+ dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+ dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+ } else {
+ dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+ dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+ FUID_SIZE_ESTIMATE(zfsvfs));
+ }
+ }
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ dmu_buf_will_dirty(dzp->z_dbuf, tx);
+
+ /*
+ * Create a new object for the symlink.
+ * Put the link content into bonus buffer if it will fit;
+ * otherwise, store it just like any other file data.
+ */
+ if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
+ zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, NULL, &fuidp);
+ if (len != 0)
+ bcopy(link, zp->z_phys + 1, len);
+ } else {
+ dmu_buf_t *dbp;
+
+ zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, NULL, &fuidp);
+ /*
+ * Nothing can access the znode yet so no locking needed
+ * for growing the znode's blocksize.
+ */
+ zfs_grow_blocksize(zp, len, tx);
+
+ VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
+ zp->z_id, 0, FTAG, &dbp));
+ dmu_buf_will_dirty(dbp, tx);
+
+ ASSERT3U(len, <=, dbp->db_size);
+ bcopy(link, dbp->db_data, len);
+ dmu_buf_rele(dbp, FTAG);
+ }
+ zp->z_phys->zp_size = len;
+
+ /*
+ * Insert the new object into the directory.
+ */
+ (void) zfs_link_create(dl, zp, tx, ZNEW);
+out:
+ if (error == 0) {
+ uint64_t txtype = TX_SYMLINK;
+ if (flags & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
+ }
+ if (fuidp)
+ zfs_fuid_info_free(fuidp);
+
+ dmu_tx_commit(tx);
+
+ zfs_dirent_unlock(dl);
+
+ VN_RELE(ZTOV(zp));
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Return, in the buffer contained in the provided uio structure,
+ * the symbolic path referred to by vp.
+ *
+ * IN: vp - vnode of symbolic link.
+ * uoip - structure to contain the link path.
+ * cr - credentials of caller.
+ * ct - caller context
+ *
+ * OUT: uio - structure to contain the link path.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - atime updated
+ */
+/* ARGSUSED */
+static int
+zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ size_t bufsz;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ bufsz = (size_t)zp->z_phys->zp_size;
+ if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
+ error = uiomove(zp->z_phys + 1,
+ MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
+ } else {
+ dmu_buf_t *dbp;
+ error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
+ if (error) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+ error = uiomove(dbp->db_data,
+ MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
+ dmu_buf_rele(dbp, FTAG);
+ }
+
+ ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Insert a new entry into directory tdvp referencing svp.
+ *
+ * IN: tdvp - Directory to contain new entry.
+ * svp - vnode of new entry.
+ * name - name of new entry.
+ * cr - credentials of caller.
+ * ct - caller context
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * tdvp - ctime|mtime updated
+ * svp - ctime updated
+ */
+/* ARGSUSED */
+static int
+zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
+ caller_context_t *ct, int flags)
+{
+ znode_t *dzp = VTOZ(tdvp);
+ znode_t *tzp, *szp;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ zilog_t *zilog;
+ zfs_dirlock_t *dl;
+ dmu_tx_t *tx;
+ vnode_t *realvp;
+ int error;
+ int zf = ZNEW;
+ uid_t owner;
+
+ ASSERT(tdvp->v_type == VDIR);
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(dzp);
+ zilog = zfsvfs->z_log;
+
+ if (VOP_REALVP(svp, &realvp, ct) == 0)
+ svp = realvp;
+
+ if (svp->v_vfsp != tdvp->v_vfsp) {
+ ZFS_EXIT(zfsvfs);
+ return (EXDEV);
+ }
+ szp = VTOZ(svp);
+ ZFS_VERIFY_ZP(szp);
+
+ if (zfsvfs->z_utf8 && u8_validate(name,
+ strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EILSEQ);
+ }
+ if (flags & FIGNORECASE)
+ zf |= ZCILOOK;
+
+top:
+ /*
+ * We do not support links between attributes and non-attributes
+ * because of the potential security risk of creating links
+ * into "normal" file space in order to circumvent restrictions
+ * imposed in attribute space.
+ */
+ if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
+ (dzp->z_phys->zp_flags & ZFS_XATTR)) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ /*
+ * POSIX dictates that we return EPERM here.
+ * Better choices include ENOTSUP or EISDIR.
+ */
+ if (svp->v_type == VDIR) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
+ if (owner != crgetuid(cr) &&
+ secpolicy_basic_link(cr) != 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ /*
+ * Attempt to lock directory; fail if entry already exists.
+ */
+ error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
+ if (error) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, szp->z_id);
+ dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ zfs_dirent_unlock(dl);
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ error = zfs_link_create(dl, szp, tx, 0);
+
+ if (error == 0) {
+ uint64_t txtype = TX_LINK;
+ if (flags & FIGNORECASE)
+ txtype |= TX_CI;
+ zfs_log_link(zilog, tx, txtype, dzp, szp, name);
+ }
+
+ dmu_tx_commit(tx);
+
+ zfs_dirent_unlock(dl);
+
+ if (error == 0) {
+ vnevent_link(svp, ct);
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * zfs_null_putapage() is used when the file system has been force
+ * unmounted. It just drops the pages.
+ */
+/* ARGSUSED */
+static int
+zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp,
+ size_t *lenp, int flags, cred_t *cr)
+{
+ pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR);
+ return (0);
+}
+
+/*
+ * Push a page out to disk, klustering if possible.
+ *
+ * IN: vp - file to push page to.
+ * pp - page to push.
+ * flags - additional flags.
+ * cr - credentials of caller.
+ *
+ * OUT: offp - start of range pushed.
+ * lenp - len of range pushed.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * NOTE: callers must have locked the page to be pushed. On
+ * exit, the page (and all other pages in the kluster) must be
+ * unlocked.
+ */
+/* ARGSUSED */
+static int
+zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp,
+ size_t *lenp, int flags, cred_t *cr)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zilog_t *zilog = zfsvfs->z_log;
+ dmu_tx_t *tx;
+ rl_t *rl;
+ u_offset_t off, koff;
+ size_t len, klen;
+ uint64_t filesz;
+ int err;
+
+ filesz = zp->z_phys->zp_size;
+ off = pp->p_offset;
+ len = PAGESIZE;
+ /*
+ * If our blocksize is bigger than the page size, try to kluster
+ * muiltiple pages so that we write a full block (thus avoiding
+ * a read-modify-write).
+ */
+ if (off < filesz && zp->z_blksz > PAGESIZE) {
+ if (!ISP2(zp->z_blksz)) {
+ /* Only one block in the file. */
+ klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE);
+ koff = 0;
+ } else {
+ klen = zp->z_blksz;
+ koff = P2ALIGN(off, (u_offset_t)klen);
+ }
+ ASSERT(koff <= filesz);
+ if (koff + klen > filesz)
+ klen = P2ROUNDUP(filesz - koff, (uint64_t)PAGESIZE);
+ pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags);
+ }
+ ASSERT3U(btop(len), ==, btopr(len));
+top:
+ rl = zfs_range_lock(zp, off, len, RL_WRITER);
+ /*
+ * Can't push pages past end-of-file.
+ */
+ filesz = zp->z_phys->zp_size;
+ if (off >= filesz) {
+ /* ignore all pages */
+ err = 0;
+ goto out;
+ } else if (off + len > filesz) {
+ int npages = btopr(filesz - off);
+ page_t *trunc;
+
+ page_list_break(&pp, &trunc, npages);
+ /* ignore pages past end of file */
+ if (trunc)
+ pvn_write_done(trunc, flags);
+ len = filesz - off;
+ }
+
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_write(tx, zp->z_id, off, len);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ err = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (err != 0) {
+ if (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ zfs_range_unlock(rl);
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ err = 0;
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ goto out;
+ }
+
+ if (zp->z_blksz <= PAGESIZE) {
+ caddr_t va = zfs_map_page(pp, S_READ);
+ ASSERT3U(len, <=, PAGESIZE);
+ dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx);
+ zfs_unmap_page(pp, va);
+ } else {
+ err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx);
+ }
+
+ if (err == 0) {
+ zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
+ zfs_log_write(zilog, tx, TX_WRITE, zp, off, len, 0);
+ dmu_tx_commit(tx);
+ }
+
+out:
+ zfs_range_unlock(rl);
+ pvn_write_done(pp, (err ? B_ERROR : 0) | flags);
+ if (offp)
+ *offp = off;
+ if (lenp)
+ *lenp = len;
+
+ return (err);
+}
+
+/*
+ * Copy the portion of the file indicated from pages into the file.
+ * The pages are stored in a page list attached to the files vnode.
+ *
+ * IN: vp - vnode of file to push page data to.
+ * off - position in file to put data.
+ * len - amount of data to write.
+ * flags - flags to control the operation.
+ * cr - credentials of caller.
+ * ct - caller context.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - ctime|mtime updated
+ */
+/*ARGSUSED*/
+static int
+zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ page_t *pp;
+ size_t io_len;
+ u_offset_t io_off;
+ uint64_t filesz;
+ int error = 0;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if (len == 0) {
+ /*
+ * Search the entire vp list for pages >= off.
+ */
+ error = pvn_vplist_dirty(vp, (u_offset_t)off, zfs_putapage,
+ flags, cr);
+ goto out;
+ }
+
+ filesz = zp->z_phys->zp_size; /* get consistent copy of zp_size */
+ if (off > filesz) {
+ /* past end of file */
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+
+ len = MIN(len, filesz - off);
+
+ for (io_off = off; io_off < off + len; io_off += io_len) {
+ if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) {
+ pp = page_lookup(vp, io_off,
+ (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED);
+ } else {
+ pp = page_lookup_nowait(vp, io_off,
+ (flags & B_FREE) ? SE_EXCL : SE_SHARED);
+ }
+
+ if (pp != NULL && pvn_getdirty(pp, flags)) {
+ int err;
+
+ /*
+ * Found a dirty page to push
+ */
+ err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr);
+ if (err)
+ error = err;
+ } else {
+ io_len = PAGESIZE;
+ }
+ }
+out:
+ if ((flags & B_ASYNC) == 0)
+ zil_commit(zfsvfs->z_log, UINT64_MAX, zp->z_id);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*ARGSUSED*/
+void
+zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+
+ rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
+ if (zp->z_dbuf == NULL) {
+ /*
+ * The fs has been unmounted, or we did a
+ * suspend/resume and this file no longer exists.
+ */
+ if (vn_has_cached_data(vp)) {
+ (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage,
+ B_INVAL, cr);
+ }
+
+ mutex_enter(&zp->z_lock);
+ vp->v_count = 0; /* count arrives as 1 */
+ mutex_exit(&zp->z_lock);
+ rw_exit(&zfsvfs->z_teardown_inactive_lock);
+ zfs_znode_free(zp);
+ return;
+ }
+
+ /*
+ * Attempt to push any data in the page cache. If this fails
+ * we will get kicked out later in zfs_zinactive().
+ */
+ if (vn_has_cached_data(vp)) {
+ (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC,
+ cr);
+ }
+
+ if (zp->z_atime_dirty && zp->z_unlinked == 0) {
+ dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
+
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ } else {
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+ mutex_enter(&zp->z_lock);
+ zp->z_atime_dirty = 0;
+ mutex_exit(&zp->z_lock);
+ dmu_tx_commit(tx);
+ }
+ }
+
+ zfs_zinactive(zp);
+ rw_exit(&zfsvfs->z_teardown_inactive_lock);
+}
+
+/*
+ * Bounds-check the seek operation.
+ *
+ * IN: vp - vnode seeking within
+ * ooff - old file offset
+ * noffp - pointer to new file offset
+ * ct - caller context
+ *
+ * RETURN: 0 if success
+ * EINVAL if new offset invalid
+ */
+/* ARGSUSED */
+static int
+zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp,
+ caller_context_t *ct)
+{
+ if (vp->v_type == VDIR)
+ return (0);
+ return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0);
+}
+
+/*
+ * Pre-filter the generic locking function to trap attempts to place
+ * a mandatory lock on a memory mapped file.
+ */
+static int
+zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset,
+ flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ /*
+ * We are following the UFS semantics with respect to mapcnt
+ * here: If we see that the file is mapped already, then we will
+ * return an error, but we don't worry about races between this
+ * function and zfs_map().
+ */
+ if (zp->z_mapcnt > 0 && MANDMODE((mode_t)zp->z_phys->zp_mode)) {
+ ZFS_EXIT(zfsvfs);
+ return (EAGAIN);
+ }
+ error = fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * If we can't find a page in the cache, we will create a new page
+ * and fill it with file data. For efficiency, we may try to fill
+ * multiple pages at once (klustering).
+ */
+static int
+zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg,
+ caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw)
+{
+ znode_t *zp = VTOZ(vp);
+ page_t *pp, *cur_pp;
+ objset_t *os = zp->z_zfsvfs->z_os;
+ caddr_t va;
+ u_offset_t io_off, total;
+ uint64_t oid = zp->z_id;
+ size_t io_len;
+ uint64_t filesz;
+ int err;
+
+ /*
+ * If we are only asking for a single page don't bother klustering.
+ */
+ filesz = zp->z_phys->zp_size; /* get consistent copy of zp_size */
+ if (off >= filesz)
+ return (EFAULT);
+ if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) {
+ io_off = off;
+ io_len = PAGESIZE;
+ pp = page_create_va(vp, io_off, io_len, PG_WAIT, seg, addr);
+ } else {
+ /*
+ * Try to fill a kluster of pages (a blocks worth).
+ */
+ size_t klen;
+ u_offset_t koff;
+
+ if (!ISP2(zp->z_blksz)) {
+ /* Only one block in the file. */
+ klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE);
+ koff = 0;
+ } else {
+ /*
+ * It would be ideal to align our offset to the
+ * blocksize but doing so has resulted in some
+ * strange application crashes. For now, we
+ * leave the offset as is and only adjust the
+ * length if we are off the end of the file.
+ */
+ koff = off;
+ klen = plsz;
+ }
+ ASSERT(koff <= filesz);
+ if (koff + klen > filesz)
+ klen = P2ROUNDUP(filesz, (uint64_t)PAGESIZE) - koff;
+ ASSERT3U(off, >=, koff);
+ ASSERT3U(off, <, koff + klen);
+ pp = pvn_read_kluster(vp, off, seg, addr, &io_off,
+ &io_len, koff, klen, 0);
+ }
+ if (pp == NULL) {
+ /*
+ * Some other thread entered the page before us.
+ * Return to zfs_getpage to retry the lookup.
+ */
+ *pl = NULL;
+ return (0);
+ }
+
+ /*
+ * Fill the pages in the kluster.
+ */
+ cur_pp = pp;
+ for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) {
+ ASSERT3U(io_off, ==, cur_pp->p_offset);
+ va = zfs_map_page(cur_pp, S_WRITE);
+ err = dmu_read(os, oid, io_off, PAGESIZE, va);
+ zfs_unmap_page(cur_pp, va);
+ if (err) {
+ /* On error, toss the entire kluster */
+ pvn_read_done(pp, B_ERROR);
+ /* convert checksum errors into IO errors */
+ if (err == ECKSUM)
+ err = EIO;
+ return (err);
+ }
+ cur_pp = cur_pp->p_next;
+ }
+out:
+ /*
+ * Fill in the page list array from the kluster. If
+ * there are too many pages in the kluster, return
+ * as many pages as possible starting from the desired
+ * offset `off'.
+ * NOTE: the page list will always be null terminated.
+ */
+ pvn_plist_init(pp, pl, plsz, off, io_len, rw);
+
+ return (0);
+}
+
+/*
+ * Return pointers to the pages for the file region [off, off + len]
+ * in the pl array. If plsz is greater than len, this function may
+ * also return page pointers from before or after the specified
+ * region (i.e. some region [off', off' + plsz]). These additional
+ * pages are only returned if they are already in the cache, or were
+ * created as part of a klustered read.
+ *
+ * IN: vp - vnode of file to get data from.
+ * off - position in file to get data from.
+ * len - amount of data to retrieve.
+ * plsz - length of provided page list.
+ * seg - segment to obtain pages for.
+ * addr - virtual address of fault.
+ * rw - mode of created pages.
+ * cr - credentials of caller.
+ * ct - caller context.
+ *
+ * OUT: protp - protection mode of created pages.
+ * pl - list of pages created.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - atime updated
+ */
+/* ARGSUSED */
+static int
+zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp,
+ page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr,
+ enum seg_rw rw, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ page_t *pp, **pl0 = pl;
+ int need_unlock = 0, err = 0;
+ offset_t orig_off;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if (protp)
+ *protp = PROT_ALL;
+
+ /* no faultahead (for now) */
+ if (pl == NULL) {
+ ZFS_EXIT(zfsvfs);
+ return (0);
+ }
+
+ /* can't fault past EOF */
+ if (off >= zp->z_phys->zp_size) {
+ ZFS_EXIT(zfsvfs);
+ return (EFAULT);
+ }
+ orig_off = off;
+
+ /*
+ * If we already own the lock, then we must be page faulting
+ * in the middle of a write to this file (i.e., we are writing
+ * to this file using data from a mapped region of the file).
+ */
+ if (rw_owner(&zp->z_map_lock) != curthread) {
+ rw_enter(&zp->z_map_lock, RW_WRITER);
+ need_unlock = TRUE;
+ }
+
+ /*
+ * Loop through the requested range [off, off + len] looking
+ * for pages. If we don't find a page, we will need to create
+ * a new page and fill it with data from the file.
+ */
+ while (len > 0) {
+ if (plsz < PAGESIZE)
+ break;
+ if (pp = page_lookup(vp, off, SE_SHARED)) {
+ *pl++ = pp;
+ off += PAGESIZE;
+ addr += PAGESIZE;
+ len -= PAGESIZE;
+ plsz -= PAGESIZE;
+ } else {
+ err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw);
+ if (err)
+ goto out;
+ /*
+ * klustering may have changed our region
+ * to be block aligned.
+ */
+ if (((pp = *pl) != 0) && (off != pp->p_offset)) {
+ int delta = off - pp->p_offset;
+ len += delta;
+ off -= delta;
+ addr -= delta;
+ }
+ while (*pl) {
+ pl++;
+ off += PAGESIZE;
+ addr += PAGESIZE;
+ plsz -= PAGESIZE;
+ if (len > PAGESIZE)
+ len -= PAGESIZE;
+ else
+ len = 0;
+ }
+ }
+ }
+
+ /*
+ * Fill out the page array with any pages already in the cache.
+ */
+ while (plsz > 0) {
+ pp = page_lookup_nowait(vp, off, SE_SHARED);
+ if (pp == NULL)
+ break;
+ *pl++ = pp;
+ off += PAGESIZE;
+ plsz -= PAGESIZE;
+ }
+
+ ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
+out:
+ /*
+ * We can't grab the range lock for the page as reader which would
+ * stop truncation as this leads to deadlock. So we need to recheck
+ * the file size.
+ */
+ if (orig_off >= zp->z_phys->zp_size)
+ err = EFAULT;
+ if (err) {
+ /*
+ * Release any pages we have previously locked.
+ */
+ while (pl > pl0)
+ page_unlock(*--pl);
+ }
+
+ *pl = NULL;
+
+ if (need_unlock)
+ rw_exit(&zp->z_map_lock);
+
+ ZFS_EXIT(zfsvfs);
+ return (err);
+}
+
+/*
+ * Request a memory map for a section of a file. This code interacts
+ * with common code and the VM system as follows:
+ *
+ * common code calls mmap(), which ends up in smmap_common()
+ *
+ * this calls VOP_MAP(), which takes you into (say) zfs
+ *
+ * zfs_map() calls as_map(), passing segvn_create() as the callback
+ *
+ * segvn_create() creates the new segment and calls VOP_ADDMAP()
+ *
+ * zfs_addmap() updates z_mapcnt
+ */
+/*ARGSUSED*/
+static int
+zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp,
+ size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ segvn_crargs_t vn_a;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if ((prot & PROT_WRITE) &&
+ (zp->z_phys->zp_flags & (ZFS_IMMUTABLE | ZFS_READONLY |
+ ZFS_APPENDONLY))) {
+ ZFS_EXIT(zfsvfs);
+ return (EPERM);
+ }
+
+ if ((prot & (PROT_READ | PROT_EXEC)) &&
+ (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED)) {
+ ZFS_EXIT(zfsvfs);
+ return (EACCES);
+ }
+
+ if (vp->v_flag & VNOMAP) {
+ ZFS_EXIT(zfsvfs);
+ return (ENOSYS);
+ }
+
+ if (off < 0 || len > MAXOFFSET_T - off) {
+ ZFS_EXIT(zfsvfs);
+ return (ENXIO);
+ }
+
+ if (vp->v_type != VREG) {
+ ZFS_EXIT(zfsvfs);
+ return (ENODEV);
+ }
+
+ /*
+ * If file is locked, disallow mapping.
+ */
+ if (MANDMODE((mode_t)zp->z_phys->zp_mode) && vn_has_flocks(vp)) {
+ ZFS_EXIT(zfsvfs);
+ return (EAGAIN);
+ }
+
+ as_rangelock(as);
+ error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags);
+ if (error != 0) {
+ as_rangeunlock(as);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ vn_a.vp = vp;
+ vn_a.offset = (u_offset_t)off;
+ vn_a.type = flags & MAP_TYPE;
+ vn_a.prot = prot;
+ vn_a.maxprot = maxprot;
+ vn_a.cred = cr;
+ vn_a.amp = NULL;
+ vn_a.flags = flags & ~MAP_TYPE;
+ vn_a.szc = 0;
+ vn_a.lgrp_mem_policy_flags = 0;
+
+ error = as_map(as, *addrp, len, segvn_create, &vn_a);
+
+ as_rangeunlock(as);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/* ARGSUSED */
+static int
+zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
+ size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ uint64_t pages = btopr(len);
+
+ atomic_add_64(&VTOZ(vp)->z_mapcnt, pages);
+ return (0);
+}
+
+/*
+ * The reason we push dirty pages as part of zfs_delmap() is so that we get a
+ * more accurate mtime for the associated file. Since we don't have a way of
+ * detecting when the data was actually modified, we have to resort to
+ * heuristics. If an explicit msync() is done, then we mark the mtime when the
+ * last page is pushed. The problem occurs when the msync() call is omitted,
+ * which by far the most common case:
+ *
+ * open()
+ * mmap()
+ * <modify memory>
+ * munmap()
+ * close()
+ * <time lapse>
+ * putpage() via fsflush
+ *
+ * If we wait until fsflush to come along, we can have a modification time that
+ * is some arbitrary point in the future. In order to prevent this in the
+ * common case, we flush pages whenever a (MAP_SHARED, PROT_WRITE) mapping is
+ * torn down.
+ */
+/* ARGSUSED */
+static int
+zfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
+ size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr,
+ caller_context_t *ct)
+{
+ uint64_t pages = btopr(len);
+
+ ASSERT3U(VTOZ(vp)->z_mapcnt, >=, pages);
+ atomic_add_64(&VTOZ(vp)->z_mapcnt, -pages);
+
+ if ((flags & MAP_SHARED) && (prot & PROT_WRITE) &&
+ vn_has_cached_data(vp))
+ (void) VOP_PUTPAGE(vp, off, len, B_ASYNC, cr, ct);
+
+ return (0);
+}
+
+/*
+ * Free or allocate space in a file. Currently, this function only
+ * supports the `F_FREESP' command. However, this command is somewhat
+ * misnamed, as its functionality includes the ability to allocate as
+ * well as free space.
+ *
+ * IN: vp - vnode of file to free data in.
+ * cmd - action to take (only F_FREESP supported).
+ * bfp - section of file to free/alloc.
+ * flag - current file open mode flags.
+ * offset - current file offset.
+ * cr - credentials of caller [UNUSED].
+ * ct - caller context.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ *
+ * Timestamps:
+ * vp - ctime|mtime updated
+ */
+/* ARGSUSED */
+static int
+zfs_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag,
+ offset_t offset, cred_t *cr, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ uint64_t off, len;
+ int error;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+
+ if (cmd != F_FREESP) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ if (error = convoff(vp, bfp, 0, offset)) {
+ ZFS_EXIT(zfsvfs);
+ return (error);
+ }
+
+ if (bfp->l_len < 0) {
+ ZFS_EXIT(zfsvfs);
+ return (EINVAL);
+ }
+
+ off = bfp->l_start;
+ len = bfp->l_len; /* 0 means from off to end of file */
+
+ error = zfs_freesp(zp, off, len, flag, TRUE);
+
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*ARGSUSED*/
+static int
+zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ uint32_t gen;
+ uint64_t object = zp->z_id;
+ zfid_short_t *zfid;
+ int size, i;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ gen = (uint32_t)zp->z_gen;
+
+ size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
+ if (fidp->fid_len < size) {
+ fidp->fid_len = size;
+ ZFS_EXIT(zfsvfs);
+ return (ENOSPC);
+ }
+
+ zfid = (zfid_short_t *)fidp;
+
+ zfid->zf_len = size;
+
+ for (i = 0; i < sizeof (zfid->zf_object); i++)
+ zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
+
+ /* Must have a non-zero generation number to distinguish from .zfs */
+ if (gen == 0)
+ gen = 1;
+ for (i = 0; i < sizeof (zfid->zf_gen); i++)
+ zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
+
+ if (size == LONG_FID_LEN) {
+ uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
+ zfid_long_t *zlfid;
+
+ zlfid = (zfid_long_t *)fidp;
+
+ for (i = 0; i < sizeof (zlfid->zf_setid); i++)
+ zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
+
+ /* XXX - this should be the generation number for the objset */
+ for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
+ zlfid->zf_setgen[i] = 0;
+ }
+
+ ZFS_EXIT(zfsvfs);
+ return (0);
+}
+
+static int
+zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp, *xzp;
+ zfsvfs_t *zfsvfs;
+ zfs_dirlock_t *dl;
+ int error;
+
+ switch (cmd) {
+ case _PC_LINK_MAX:
+ *valp = ULONG_MAX;
+ return (0);
+
+ case _PC_FILESIZEBITS:
+ *valp = 64;
+ return (0);
+
+ case _PC_XATTR_EXISTS:
+ zp = VTOZ(vp);
+ zfsvfs = zp->z_zfsvfs;
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ *valp = 0;
+ error = zfs_dirent_lock(&dl, zp, "", &xzp,
+ ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
+ if (error == 0) {
+ zfs_dirent_unlock(dl);
+ if (!zfs_dirempty(xzp))
+ *valp = 1;
+ VN_RELE(ZTOV(xzp));
+ } else if (error == ENOENT) {
+ /*
+ * If there aren't extended attributes, it's the
+ * same as having zero of them.
+ */
+ error = 0;
+ }
+ ZFS_EXIT(zfsvfs);
+ return (error);
+
+ case _PC_SATTR_ENABLED:
+ case _PC_SATTR_EXISTS:
+ *valp = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
+ (vp->v_type == VREG || vp->v_type == VDIR);
+ return (0);
+
+ case _PC_ACL_ENABLED:
+ *valp = _ACL_ACE_ENABLED;
+ return (0);
+
+ case _PC_MIN_HOLE_SIZE:
+ *valp = (ulong_t)SPA_MINBLOCKSIZE;
+ return (0);
+
+ default:
+ return (fs_pathconf(vp, cmd, valp, cr, ct));
+ }
+}
+
+/*ARGSUSED*/
+static int
+zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+ boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ error = zfs_getacl(zp, vsecp, skipaclchk, cr);
+ ZFS_EXIT(zfsvfs);
+
+ return (error);
+}
+
+/*ARGSUSED*/
+static int
+zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
+ caller_context_t *ct)
+{
+ znode_t *zp = VTOZ(vp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ int error;
+ boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
+
+ ZFS_ENTER(zfsvfs);
+ ZFS_VERIFY_ZP(zp);
+ error = zfs_setacl(zp, vsecp, skipaclchk, cr);
+ ZFS_EXIT(zfsvfs);
+ return (error);
+}
+
+/*
+ * Predeclare these here so that the compiler assumes that
+ * this is an "old style" function declaration that does
+ * not include arguments => we won't get type mismatch errors
+ * in the initializations that follow.
+ */
+static int zfs_inval();
+static int zfs_isdir();
+
+static int
+zfs_inval()
+{
+ return (EINVAL);
+}
+
+static int
+zfs_isdir()
+{
+ return (EISDIR);
+}
+/*
+ * Directory vnode operations template
+ */
+vnodeops_t *zfs_dvnodeops;
+const fs_operation_def_t zfs_dvnodeops_template[] = {
+ VOPNAME_OPEN, { .vop_open = zfs_open },
+ VOPNAME_CLOSE, { .vop_close = zfs_close },
+ VOPNAME_READ, { .error = zfs_isdir },
+ VOPNAME_WRITE, { .error = zfs_isdir },
+ VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
+ VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
+ VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
+ VOPNAME_ACCESS, { .vop_access = zfs_access },
+ VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
+ VOPNAME_CREATE, { .vop_create = zfs_create },
+ VOPNAME_REMOVE, { .vop_remove = zfs_remove },
+ VOPNAME_LINK, { .vop_link = zfs_link },
+ VOPNAME_RENAME, { .vop_rename = zfs_rename },
+ VOPNAME_MKDIR, { .vop_mkdir = zfs_mkdir },
+ VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir },
+ VOPNAME_READDIR, { .vop_readdir = zfs_readdir },
+ VOPNAME_SYMLINK, { .vop_symlink = zfs_symlink },
+ VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
+ VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
+ VOPNAME_FID, { .vop_fid = zfs_fid },
+ VOPNAME_SEEK, { .vop_seek = zfs_seek },
+ VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
+ VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
+ VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
+ VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
+ NULL, NULL
+};
+
+/*
+ * Regular file vnode operations template
+ */
+vnodeops_t *zfs_fvnodeops;
+const fs_operation_def_t zfs_fvnodeops_template[] = {
+ VOPNAME_OPEN, { .vop_open = zfs_open },
+ VOPNAME_CLOSE, { .vop_close = zfs_close },
+ VOPNAME_READ, { .vop_read = zfs_read },
+ VOPNAME_WRITE, { .vop_write = zfs_write },
+ VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
+ VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
+ VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
+ VOPNAME_ACCESS, { .vop_access = zfs_access },
+ VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
+ VOPNAME_RENAME, { .vop_rename = zfs_rename },
+ VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
+ VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
+ VOPNAME_FID, { .vop_fid = zfs_fid },
+ VOPNAME_SEEK, { .vop_seek = zfs_seek },
+ VOPNAME_FRLOCK, { .vop_frlock = zfs_frlock },
+ VOPNAME_SPACE, { .vop_space = zfs_space },
+ VOPNAME_GETPAGE, { .vop_getpage = zfs_getpage },
+ VOPNAME_PUTPAGE, { .vop_putpage = zfs_putpage },
+ VOPNAME_MAP, { .vop_map = zfs_map },
+ VOPNAME_ADDMAP, { .vop_addmap = zfs_addmap },
+ VOPNAME_DELMAP, { .vop_delmap = zfs_delmap },
+ VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
+ VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
+ VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
+ VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
+ NULL, NULL
+};
+
+/*
+ * Symbolic link vnode operations template
+ */
+vnodeops_t *zfs_symvnodeops;
+const fs_operation_def_t zfs_symvnodeops_template[] = {
+ VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
+ VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
+ VOPNAME_ACCESS, { .vop_access = zfs_access },
+ VOPNAME_RENAME, { .vop_rename = zfs_rename },
+ VOPNAME_READLINK, { .vop_readlink = zfs_readlink },
+ VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
+ VOPNAME_FID, { .vop_fid = zfs_fid },
+ VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
+ VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
+ NULL, NULL
+};
+
+/*
+ * Extended attribute directory vnode operations template
+ * This template is identical to the directory vnodes
+ * operation template except for restricted operations:
+ * VOP_MKDIR()
+ * VOP_SYMLINK()
+ * Note that there are other restrictions embedded in:
+ * zfs_create() - restrict type to VREG
+ * zfs_link() - no links into/out of attribute space
+ * zfs_rename() - no moves into/out of attribute space
+ */
+vnodeops_t *zfs_xdvnodeops;
+const fs_operation_def_t zfs_xdvnodeops_template[] = {
+ VOPNAME_OPEN, { .vop_open = zfs_open },
+ VOPNAME_CLOSE, { .vop_close = zfs_close },
+ VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
+ VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
+ VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
+ VOPNAME_ACCESS, { .vop_access = zfs_access },
+ VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
+ VOPNAME_CREATE, { .vop_create = zfs_create },
+ VOPNAME_REMOVE, { .vop_remove = zfs_remove },
+ VOPNAME_LINK, { .vop_link = zfs_link },
+ VOPNAME_RENAME, { .vop_rename = zfs_rename },
+ VOPNAME_MKDIR, { .error = zfs_inval },
+ VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir },
+ VOPNAME_READDIR, { .vop_readdir = zfs_readdir },
+ VOPNAME_SYMLINK, { .error = zfs_inval },
+ VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
+ VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
+ VOPNAME_FID, { .vop_fid = zfs_fid },
+ VOPNAME_SEEK, { .vop_seek = zfs_seek },
+ VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
+ VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
+ VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
+ VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
+ NULL, NULL
+};
+
+/*
+ * Error vnode operations template
+ */
+vnodeops_t *zfs_evnodeops;
+const fs_operation_def_t zfs_evnodeops_template[] = {
+ VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
+ VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
+ NULL, NULL
+};
diff --git a/module/zfs/zfs_znode.c b/module/zfs/zfs_znode.c
new file mode 100644
index 000000000..25751ae5f
--- /dev/null
+++ b/module/zfs/zfs_znode.c
@@ -0,0 +1,1672 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/* Portions Copyright 2007 Jeremy Teo */
+
+#ifdef _KERNEL
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/time.h>
+#include <sys/systm.h>
+#include <sys/sysmacros.h>
+#include <sys/resource.h>
+#include <sys/mntent.h>
+#include <sys/mkdev.h>
+#include <sys/u8_textprep.h>
+#include <sys/dsl_dataset.h>
+#include <sys/vfs.h>
+#include <sys/vfs_opreg.h>
+#include <sys/vnode.h>
+#include <sys/file.h>
+#include <sys/kmem.h>
+#include <sys/errno.h>
+#include <sys/unistd.h>
+#include <sys/mode.h>
+#include <sys/atomic.h>
+#include <vm/pvn.h>
+#include "fs/fs_subr.h"
+#include <sys/zfs_dir.h>
+#include <sys/zfs_acl.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/zfs_rlock.h>
+#include <sys/zfs_fuid.h>
+#include <sys/fs/zfs.h>
+#include <sys/kidmap.h>
+#endif /* _KERNEL */
+
+#include <sys/dmu.h>
+#include <sys/refcount.h>
+#include <sys/stat.h>
+#include <sys/zap.h>
+#include <sys/zfs_znode.h>
+
+#include "zfs_prop.h"
+
+/*
+ * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
+ * turned on when DEBUG is also defined.
+ */
+#ifdef DEBUG
+#define ZNODE_STATS
+#endif /* DEBUG */
+
+#ifdef ZNODE_STATS
+#define ZNODE_STAT_ADD(stat) ((stat)++)
+#else
+#define ZNODE_STAT_ADD(stat) /* nothing */
+#endif /* ZNODE_STATS */
+
+#define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3))
+#define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1))
+
+/*
+ * Functions needed for userland (ie: libzpool) are not put under
+ * #ifdef_KERNEL; the rest of the functions have dependencies
+ * (such as VFS logic) that will not compile easily in userland.
+ */
+#ifdef _KERNEL
+static kmem_cache_t *znode_cache = NULL;
+
+/*ARGSUSED*/
+static void
+znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
+{
+ /*
+ * We should never drop all dbuf refs without first clearing
+ * the eviction callback.
+ */
+ panic("evicting znode %p\n", user_ptr);
+}
+
+/*ARGSUSED*/
+static int
+zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
+{
+ znode_t *zp = buf;
+
+ ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
+
+ zp->z_vnode = vn_alloc(kmflags);
+ if (zp->z_vnode == NULL) {
+ return (-1);
+ }
+ ZTOV(zp)->v_data = zp;
+
+ list_link_init(&zp->z_link_node);
+
+ mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
+ rw_init(&zp->z_map_lock, NULL, RW_DEFAULT, NULL);
+ rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
+ rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
+ mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
+ avl_create(&zp->z_range_avl, zfs_range_compare,
+ sizeof (rl_t), offsetof(rl_t, r_node));
+
+ zp->z_dbuf = NULL;
+ zp->z_dirlocks = NULL;
+ return (0);
+}
+
+/*ARGSUSED*/
+static void
+zfs_znode_cache_destructor(void *buf, void *arg)
+{
+ znode_t *zp = buf;
+
+ ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
+ ASSERT(ZTOV(zp)->v_data == zp);
+ vn_free(ZTOV(zp));
+ ASSERT(!list_link_active(&zp->z_link_node));
+ mutex_destroy(&zp->z_lock);
+ rw_destroy(&zp->z_map_lock);
+ rw_destroy(&zp->z_parent_lock);
+ rw_destroy(&zp->z_name_lock);
+ mutex_destroy(&zp->z_acl_lock);
+ avl_destroy(&zp->z_range_avl);
+ mutex_destroy(&zp->z_range_lock);
+
+ ASSERT(zp->z_dbuf == NULL);
+ ASSERT(zp->z_dirlocks == NULL);
+}
+
+#ifdef ZNODE_STATS
+static struct {
+ uint64_t zms_zfsvfs_invalid;
+ uint64_t zms_zfsvfs_unmounted;
+ uint64_t zms_zfsvfs_recheck_invalid;
+ uint64_t zms_obj_held;
+ uint64_t zms_vnode_locked;
+ uint64_t zms_not_only_dnlc;
+} znode_move_stats;
+#endif /* ZNODE_STATS */
+
+static void
+zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
+{
+ vnode_t *vp;
+
+ /* Copy fields. */
+ nzp->z_zfsvfs = ozp->z_zfsvfs;
+
+ /* Swap vnodes. */
+ vp = nzp->z_vnode;
+ nzp->z_vnode = ozp->z_vnode;
+ ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
+ ZTOV(ozp)->v_data = ozp;
+ ZTOV(nzp)->v_data = nzp;
+
+ nzp->z_id = ozp->z_id;
+ ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
+ ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
+ nzp->z_unlinked = ozp->z_unlinked;
+ nzp->z_atime_dirty = ozp->z_atime_dirty;
+ nzp->z_zn_prefetch = ozp->z_zn_prefetch;
+ nzp->z_blksz = ozp->z_blksz;
+ nzp->z_seq = ozp->z_seq;
+ nzp->z_mapcnt = ozp->z_mapcnt;
+ nzp->z_last_itx = ozp->z_last_itx;
+ nzp->z_gen = ozp->z_gen;
+ nzp->z_sync_cnt = ozp->z_sync_cnt;
+ nzp->z_phys = ozp->z_phys;
+ nzp->z_dbuf = ozp->z_dbuf;
+
+ /* Update back pointers. */
+ (void) dmu_buf_update_user(nzp->z_dbuf, ozp, nzp, &nzp->z_phys,
+ znode_evict_error);
+
+ /*
+ * Invalidate the original znode by clearing fields that provide a
+ * pointer back to the znode. Set the low bit of the vfs pointer to
+ * ensure that zfs_znode_move() recognizes the znode as invalid in any
+ * subsequent callback.
+ */
+ ozp->z_dbuf = NULL;
+ POINTER_INVALIDATE(&ozp->z_zfsvfs);
+}
+
+/*
+ * Wrapper function for ZFS_ENTER that returns 0 if successful and otherwise
+ * returns a non-zero error code.
+ */
+static int
+zfs_enter(zfsvfs_t *zfsvfs)
+{
+ ZFS_ENTER(zfsvfs);
+ return (0);
+}
+
+/*ARGSUSED*/
+static kmem_cbrc_t
+zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
+{
+ znode_t *ozp = buf, *nzp = newbuf;
+ zfsvfs_t *zfsvfs;
+ vnode_t *vp;
+
+ /*
+ * The znode is on the file system's list of known znodes if the vfs
+ * pointer is valid. We set the low bit of the vfs pointer when freeing
+ * the znode to invalidate it, and the memory patterns written by kmem
+ * (baddcafe and deadbeef) set at least one of the two low bits. A newly
+ * created znode sets the vfs pointer last of all to indicate that the
+ * znode is known and in a valid state to be moved by this function.
+ */
+ zfsvfs = ozp->z_zfsvfs;
+ if (!POINTER_IS_VALID(zfsvfs)) {
+ ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ /*
+ * Ensure that the filesystem is not unmounted during the move.
+ */
+ if (zfs_enter(zfsvfs) != 0) { /* ZFS_ENTER */
+ ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ mutex_enter(&zfsvfs->z_znodes_lock);
+ /*
+ * Recheck the vfs pointer in case the znode was removed just before
+ * acquiring the lock.
+ */
+ if (zfsvfs != ozp->z_zfsvfs) {
+ mutex_exit(&zfsvfs->z_znodes_lock);
+ ZFS_EXIT(zfsvfs);
+ ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck_invalid);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ /*
+ * At this point we know that as long as we hold z_znodes_lock, the
+ * znode cannot be freed and fields within the znode can be safely
+ * accessed. Now, prevent a race with zfs_zget().
+ */
+ if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
+ mutex_exit(&zfsvfs->z_znodes_lock);
+ ZFS_EXIT(zfsvfs);
+ ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
+ return (KMEM_CBRC_LATER);
+ }
+
+ vp = ZTOV(ozp);
+ if (mutex_tryenter(&vp->v_lock) == 0) {
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
+ mutex_exit(&zfsvfs->z_znodes_lock);
+ ZFS_EXIT(zfsvfs);
+ ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
+ return (KMEM_CBRC_LATER);
+ }
+
+ /* Only move znodes that are referenced _only_ by the DNLC. */
+ if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
+ mutex_exit(&vp->v_lock);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
+ mutex_exit(&zfsvfs->z_znodes_lock);
+ ZFS_EXIT(zfsvfs);
+ ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
+ return (KMEM_CBRC_LATER);
+ }
+
+ /*
+ * The znode is known and in a valid state to move. We're holding the
+ * locks needed to execute the critical section.
+ */
+ zfs_znode_move_impl(ozp, nzp);
+ mutex_exit(&vp->v_lock);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
+
+ list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
+ mutex_exit(&zfsvfs->z_znodes_lock);
+ ZFS_EXIT(zfsvfs);
+
+ return (KMEM_CBRC_YES);
+}
+
+void
+zfs_znode_init(void)
+{
+ /*
+ * Initialize zcache
+ */
+ ASSERT(znode_cache == NULL);
+ znode_cache = kmem_cache_create("zfs_znode_cache",
+ sizeof (znode_t), 0, zfs_znode_cache_constructor,
+ zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
+ kmem_cache_set_move(znode_cache, zfs_znode_move);
+}
+
+void
+zfs_znode_fini(void)
+{
+ /*
+ * Cleanup vfs & vnode ops
+ */
+ zfs_remove_op_tables();
+
+ /*
+ * Cleanup zcache
+ */
+ if (znode_cache)
+ kmem_cache_destroy(znode_cache);
+ znode_cache = NULL;
+}
+
+struct vnodeops *zfs_dvnodeops;
+struct vnodeops *zfs_fvnodeops;
+struct vnodeops *zfs_symvnodeops;
+struct vnodeops *zfs_xdvnodeops;
+struct vnodeops *zfs_evnodeops;
+
+void
+zfs_remove_op_tables()
+{
+ /*
+ * Remove vfs ops
+ */
+ ASSERT(zfsfstype);
+ (void) vfs_freevfsops_by_type(zfsfstype);
+ zfsfstype = 0;
+
+ /*
+ * Remove vnode ops
+ */
+ if (zfs_dvnodeops)
+ vn_freevnodeops(zfs_dvnodeops);
+ if (zfs_fvnodeops)
+ vn_freevnodeops(zfs_fvnodeops);
+ if (zfs_symvnodeops)
+ vn_freevnodeops(zfs_symvnodeops);
+ if (zfs_xdvnodeops)
+ vn_freevnodeops(zfs_xdvnodeops);
+ if (zfs_evnodeops)
+ vn_freevnodeops(zfs_evnodeops);
+
+ zfs_dvnodeops = NULL;
+ zfs_fvnodeops = NULL;
+ zfs_symvnodeops = NULL;
+ zfs_xdvnodeops = NULL;
+ zfs_evnodeops = NULL;
+}
+
+extern const fs_operation_def_t zfs_dvnodeops_template[];
+extern const fs_operation_def_t zfs_fvnodeops_template[];
+extern const fs_operation_def_t zfs_xdvnodeops_template[];
+extern const fs_operation_def_t zfs_symvnodeops_template[];
+extern const fs_operation_def_t zfs_evnodeops_template[];
+
+int
+zfs_create_op_tables()
+{
+ int error;
+
+ /*
+ * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
+ * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
+ * In this case we just return as the ops vectors are already set up.
+ */
+ if (zfs_dvnodeops)
+ return (0);
+
+ error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
+ &zfs_dvnodeops);
+ if (error)
+ return (error);
+
+ error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
+ &zfs_fvnodeops);
+ if (error)
+ return (error);
+
+ error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
+ &zfs_symvnodeops);
+ if (error)
+ return (error);
+
+ error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
+ &zfs_xdvnodeops);
+ if (error)
+ return (error);
+
+ error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
+ &zfs_evnodeops);
+
+ return (error);
+}
+
+/*
+ * zfs_init_fs - Initialize the zfsvfs struct and the file system
+ * incore "master" object. Verify version compatibility.
+ */
+int
+zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp)
+{
+ extern int zfsfstype;
+
+ objset_t *os = zfsvfs->z_os;
+ int i, error;
+ uint64_t fsid_guid;
+ uint64_t zval;
+
+ *zpp = NULL;
+
+ error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
+ if (error) {
+ return (error);
+ } else if (zfsvfs->z_version > ZPL_VERSION) {
+ (void) printf("Mismatched versions: File system "
+ "is version %llu on-disk format, which is "
+ "incompatible with this software version %lld!",
+ (u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
+ return (ENOTSUP);
+ }
+
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
+ return (error);
+ zfsvfs->z_norm = (int)zval;
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
+ return (error);
+ zfsvfs->z_utf8 = (zval != 0);
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
+ return (error);
+ zfsvfs->z_case = (uint_t)zval;
+ /*
+ * Fold case on file systems that are always or sometimes case
+ * insensitive.
+ */
+ if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
+ zfsvfs->z_case == ZFS_CASE_MIXED)
+ zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
+
+ /*
+ * The fsid is 64 bits, composed of an 8-bit fs type, which
+ * separates our fsid from any other filesystem types, and a
+ * 56-bit objset unique ID. The objset unique ID is unique to
+ * all objsets open on this system, provided by unique_create().
+ * The 8-bit fs type must be put in the low bits of fsid[1]
+ * because that's where other Solaris filesystems put it.
+ */
+ fsid_guid = dmu_objset_fsid_guid(os);
+ ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
+ zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid;
+ zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
+ zfsfstype & 0xFF;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
+ &zfsvfs->z_root);
+ if (error)
+ return (error);
+ ASSERT(zfsvfs->z_root != 0);
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
+ &zfsvfs->z_unlinkedobj);
+ if (error)
+ return (error);
+
+ /*
+ * Initialize zget mutex's
+ */
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
+
+ error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp);
+ if (error) {
+ /*
+ * On error, we destroy the mutexes here since it's not
+ * possible for the caller to determine if the mutexes were
+ * initialized properly.
+ */
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_destroy(&zfsvfs->z_hold_mtx[i]);
+ return (error);
+ }
+ ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root);
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
+ &zfsvfs->z_fuid_obj);
+ if (error == ENOENT)
+ error = 0;
+
+ return (0);
+}
+
+/*
+ * define a couple of values we need available
+ * for both 64 and 32 bit environments.
+ */
+#ifndef NBITSMINOR64
+#define NBITSMINOR64 32
+#endif
+#ifndef MAXMAJ64
+#define MAXMAJ64 0xffffffffUL
+#endif
+#ifndef MAXMIN64
+#define MAXMIN64 0xffffffffUL
+#endif
+
+/*
+ * Create special expldev for ZFS private use.
+ * Can't use standard expldev since it doesn't do
+ * what we want. The standard expldev() takes a
+ * dev32_t in LP64 and expands it to a long dev_t.
+ * We need an interface that takes a dev32_t in ILP32
+ * and expands it to a long dev_t.
+ */
+static uint64_t
+zfs_expldev(dev_t dev)
+{
+#ifndef _LP64
+ major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
+ return (((uint64_t)major << NBITSMINOR64) |
+ ((minor_t)dev & MAXMIN32));
+#else
+ return (dev);
+#endif
+}
+
+/*
+ * Special cmpldev for ZFS private use.
+ * Can't use standard cmpldev since it takes
+ * a long dev_t and compresses it to dev32_t in
+ * LP64. We need to do a compaction of a long dev_t
+ * to a dev32_t in ILP32.
+ */
+dev_t
+zfs_cmpldev(uint64_t dev)
+{
+#ifndef _LP64
+ minor_t minor = (minor_t)dev & MAXMIN64;
+ major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
+
+ if (major > MAXMAJ32 || minor > MAXMIN32)
+ return (NODEV32);
+
+ return (((dev32_t)major << NBITSMINOR32) | minor);
+#else
+ return (dev);
+#endif
+}
+
+static void
+zfs_znode_dmu_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db)
+{
+ znode_t *nzp;
+
+ ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
+ ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
+
+ mutex_enter(&zp->z_lock);
+
+ ASSERT(zp->z_dbuf == NULL);
+ zp->z_dbuf = db;
+ nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error);
+
+ /*
+ * there should be no
+ * concurrent zgets on this object.
+ */
+ if (nzp != NULL)
+ panic("existing znode %p for dbuf %p", (void *)nzp, (void *)db);
+
+ /*
+ * Slap on VROOT if we are the root znode
+ */
+ if (zp->z_id == zfsvfs->z_root)
+ ZTOV(zp)->v_flag |= VROOT;
+
+ mutex_exit(&zp->z_lock);
+ vn_exists(ZTOV(zp));
+}
+
+void
+zfs_znode_dmu_fini(znode_t *zp)
+{
+ dmu_buf_t *db = zp->z_dbuf;
+ ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
+ zp->z_unlinked ||
+ RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
+ ASSERT(zp->z_dbuf != NULL);
+ zp->z_dbuf = NULL;
+ VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL));
+ dmu_buf_rele(db, NULL);
+}
+
+/*
+ * Construct a new znode/vnode and intialize.
+ *
+ * This does not do a call to dmu_set_user() that is
+ * up to the caller to do, in case you don't want to
+ * return the znode
+ */
+static znode_t *
+zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz)
+{
+ znode_t *zp;
+ vnode_t *vp;
+
+ zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
+
+ ASSERT(zp->z_dirlocks == NULL);
+ ASSERT(zp->z_dbuf == NULL);
+ ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
+
+ /*
+ * Defer setting z_zfsvfs until the znode is ready to be a candidate for
+ * the zfs_znode_move() callback.
+ */
+ zp->z_phys = NULL;
+ zp->z_unlinked = 0;
+ zp->z_atime_dirty = 0;
+ zp->z_mapcnt = 0;
+ zp->z_last_itx = 0;
+ zp->z_id = db->db_object;
+ zp->z_blksz = blksz;
+ zp->z_seq = 0x7A4653;
+ zp->z_sync_cnt = 0;
+
+ vp = ZTOV(zp);
+ vn_reinit(vp);
+
+ zfs_znode_dmu_init(zfsvfs, zp, db);
+
+ zp->z_gen = zp->z_phys->zp_gen;
+
+ vp->v_vfsp = zfsvfs->z_parent->z_vfs;
+ vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode);
+
+ switch (vp->v_type) {
+ case VDIR:
+ if (zp->z_phys->zp_flags & ZFS_XATTR) {
+ vn_setops(vp, zfs_xdvnodeops);
+ vp->v_flag |= V_XATTRDIR;
+ } else {
+ vn_setops(vp, zfs_dvnodeops);
+ }
+ zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
+ break;
+ case VBLK:
+ case VCHR:
+ vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev);
+ /*FALLTHROUGH*/
+ case VFIFO:
+ case VSOCK:
+ case VDOOR:
+ vn_setops(vp, zfs_fvnodeops);
+ break;
+ case VREG:
+ vp->v_flag |= VMODSORT;
+ vn_setops(vp, zfs_fvnodeops);
+ break;
+ case VLNK:
+ vn_setops(vp, zfs_symvnodeops);
+ break;
+ default:
+ vn_setops(vp, zfs_evnodeops);
+ break;
+ }
+
+ mutex_enter(&zfsvfs->z_znodes_lock);
+ list_insert_tail(&zfsvfs->z_all_znodes, zp);
+ membar_producer();
+ /*
+ * Everything else must be valid before assigning z_zfsvfs makes the
+ * znode eligible for zfs_znode_move().
+ */
+ zp->z_zfsvfs = zfsvfs;
+ mutex_exit(&zfsvfs->z_znodes_lock);
+
+ VFS_HOLD(zfsvfs->z_vfs);
+ return (zp);
+}
+
+/*
+ * Create a new DMU object to hold a zfs znode.
+ *
+ * IN: dzp - parent directory for new znode
+ * vap - file attributes for new znode
+ * tx - dmu transaction id for zap operations
+ * cr - credentials of caller
+ * flag - flags:
+ * IS_ROOT_NODE - new object will be root
+ * IS_XATTR - new object is an attribute
+ * IS_REPLAY - intent log replay
+ * bonuslen - length of bonus buffer
+ * setaclp - File/Dir initial ACL
+ * fuidp - Tracks fuid allocation.
+ *
+ * OUT: zpp - allocated znode
+ *
+ */
+void
+zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
+ uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_t *setaclp,
+ zfs_fuid_info_t **fuidp)
+{
+ dmu_buf_t *db;
+ znode_phys_t *pzp;
+ zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
+ timestruc_t now;
+ uint64_t gen, obj;
+ int err;
+
+ ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
+
+ if (zfsvfs->z_assign >= TXG_INITIAL) { /* ZIL replay */
+ obj = vap->va_nodeid;
+ flag |= IS_REPLAY;
+ now = vap->va_ctime; /* see zfs_replay_create() */
+ gen = vap->va_nblocks; /* ditto */
+ } else {
+ obj = 0;
+ gethrestime(&now);
+ gen = dmu_tx_get_txg(tx);
+ }
+
+ /*
+ * Create a new DMU object.
+ */
+ /*
+ * There's currently no mechanism for pre-reading the blocks that will
+ * be to needed allocate a new object, so we accept the small chance
+ * that there will be an i/o error and we will fail one of the
+ * assertions below.
+ */
+ if (vap->va_type == VDIR) {
+ if (flag & IS_REPLAY) {
+ err = zap_create_claim_norm(zfsvfs->z_os, obj,
+ zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
+ DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
+ ASSERT3U(err, ==, 0);
+ } else {
+ obj = zap_create_norm(zfsvfs->z_os,
+ zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
+ DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
+ }
+ } else {
+ if (flag & IS_REPLAY) {
+ err = dmu_object_claim(zfsvfs->z_os, obj,
+ DMU_OT_PLAIN_FILE_CONTENTS, 0,
+ DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
+ ASSERT3U(err, ==, 0);
+ } else {
+ obj = dmu_object_alloc(zfsvfs->z_os,
+ DMU_OT_PLAIN_FILE_CONTENTS, 0,
+ DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
+ }
+ }
+ VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db));
+ dmu_buf_will_dirty(db, tx);
+
+ /*
+ * Initialize the znode physical data to zero.
+ */
+ ASSERT(db->db_size >= sizeof (znode_phys_t));
+ bzero(db->db_data, db->db_size);
+ pzp = db->db_data;
+
+ /*
+ * If this is the root, fix up the half-initialized parent pointer
+ * to reference the just-allocated physical data area.
+ */
+ if (flag & IS_ROOT_NODE) {
+ dzp->z_dbuf = db;
+ dzp->z_phys = pzp;
+ dzp->z_id = obj;
+ }
+
+ /*
+ * If parent is an xattr, so am I.
+ */
+ if (dzp->z_phys->zp_flags & ZFS_XATTR)
+ flag |= IS_XATTR;
+
+ if (vap->va_type == VBLK || vap->va_type == VCHR) {
+ pzp->zp_rdev = zfs_expldev(vap->va_rdev);
+ }
+
+ if (zfsvfs->z_use_fuids)
+ pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
+
+ if (vap->va_type == VDIR) {
+ pzp->zp_size = 2; /* contents ("." and "..") */
+ pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
+ }
+
+ pzp->zp_parent = dzp->z_id;
+ if (flag & IS_XATTR)
+ pzp->zp_flags |= ZFS_XATTR;
+
+ pzp->zp_gen = gen;
+
+ ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
+ ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
+
+ if (vap->va_mask & AT_ATIME) {
+ ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
+ } else {
+ ZFS_TIME_ENCODE(&now, pzp->zp_atime);
+ }
+
+ if (vap->va_mask & AT_MTIME) {
+ ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
+ } else {
+ ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
+ }
+
+ pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode);
+ if (!(flag & IS_ROOT_NODE)) {
+ ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
+ *zpp = zfs_znode_alloc(zfsvfs, db, 0);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
+ } else {
+ /*
+ * If we are creating the root node, the "parent" we
+ * passed in is the znode for the root.
+ */
+ *zpp = dzp;
+ }
+ zfs_perm_init(*zpp, dzp, flag, vap, tx, cr, setaclp, fuidp);
+}
+
+void
+zfs_xvattr_set(znode_t *zp, xvattr_t *xvap)
+{
+ xoptattr_t *xoap;
+
+ xoap = xva_getxoptattr(xvap);
+ ASSERT(xoap);
+
+ if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
+ ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime);
+ XVA_SET_RTN(xvap, XAT_CREATETIME);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
+ ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly);
+ XVA_SET_RTN(xvap, XAT_READONLY);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
+ ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden);
+ XVA_SET_RTN(xvap, XAT_HIDDEN);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
+ ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system);
+ XVA_SET_RTN(xvap, XAT_SYSTEM);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
+ ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive);
+ XVA_SET_RTN(xvap, XAT_ARCHIVE);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
+ ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable);
+ XVA_SET_RTN(xvap, XAT_IMMUTABLE);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
+ ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink);
+ XVA_SET_RTN(xvap, XAT_NOUNLINK);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
+ ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly);
+ XVA_SET_RTN(xvap, XAT_APPENDONLY);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
+ ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump);
+ XVA_SET_RTN(xvap, XAT_NODUMP);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
+ ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque);
+ XVA_SET_RTN(xvap, XAT_OPAQUE);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
+ ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
+ xoap->xoa_av_quarantined);
+ XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
+ ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified);
+ XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
+ }
+ if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
+ (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp,
+ sizeof (xoap->xoa_av_scanstamp));
+ zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP;
+ XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
+ }
+}
+
+int
+zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
+{
+ dmu_object_info_t doi;
+ dmu_buf_t *db;
+ znode_t *zp;
+ int err;
+
+ *zpp = NULL;
+
+ ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
+
+ err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
+ if (err) {
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (err);
+ }
+
+ dmu_object_info_from_db(db, &doi);
+ if (doi.doi_bonus_type != DMU_OT_ZNODE ||
+ doi.doi_bonus_size < sizeof (znode_phys_t)) {
+ dmu_buf_rele(db, NULL);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (EINVAL);
+ }
+
+ zp = dmu_buf_get_user(db);
+ if (zp != NULL) {
+ mutex_enter(&zp->z_lock);
+
+ /*
+ * Since we do immediate eviction of the z_dbuf, we
+ * should never find a dbuf with a znode that doesn't
+ * know about the dbuf.
+ */
+ ASSERT3P(zp->z_dbuf, ==, db);
+ ASSERT3U(zp->z_id, ==, obj_num);
+ if (zp->z_unlinked) {
+ err = ENOENT;
+ } else {
+ VN_HOLD(ZTOV(zp));
+ *zpp = zp;
+ err = 0;
+ }
+ dmu_buf_rele(db, NULL);
+ mutex_exit(&zp->z_lock);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (err);
+ }
+
+ /*
+ * Not found create new znode/vnode
+ */
+ zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ *zpp = zp;
+ return (0);
+}
+
+int
+zfs_rezget(znode_t *zp)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ dmu_object_info_t doi;
+ dmu_buf_t *db;
+ uint64_t obj_num = zp->z_id;
+ int err;
+
+ ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
+
+ err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
+ if (err) {
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (err);
+ }
+
+ dmu_object_info_from_db(db, &doi);
+ if (doi.doi_bonus_type != DMU_OT_ZNODE ||
+ doi.doi_bonus_size < sizeof (znode_phys_t)) {
+ dmu_buf_rele(db, NULL);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (EINVAL);
+ }
+
+ if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) {
+ dmu_buf_rele(db, NULL);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+ return (EIO);
+ }
+
+ zfs_znode_dmu_init(zfsvfs, zp, db);
+ zp->z_unlinked = (zp->z_phys->zp_links == 0);
+ zp->z_blksz = doi.doi_data_block_size;
+
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
+
+ return (0);
+}
+
+void
+zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ objset_t *os = zfsvfs->z_os;
+ uint64_t obj = zp->z_id;
+ uint64_t acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
+
+ ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
+ if (acl_obj)
+ VERIFY(0 == dmu_object_free(os, acl_obj, tx));
+ VERIFY(0 == dmu_object_free(os, obj, tx));
+ zfs_znode_dmu_fini(zp);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
+ zfs_znode_free(zp);
+}
+
+void
+zfs_zinactive(znode_t *zp)
+{
+ vnode_t *vp = ZTOV(zp);
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ uint64_t z_id = zp->z_id;
+
+ ASSERT(zp->z_dbuf && zp->z_phys);
+
+ /*
+ * Don't allow a zfs_zget() while were trying to release this znode
+ */
+ ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
+
+ mutex_enter(&zp->z_lock);
+ mutex_enter(&vp->v_lock);
+ vp->v_count--;
+ if (vp->v_count > 0 || vn_has_cached_data(vp)) {
+ /*
+ * If the hold count is greater than zero, somebody has
+ * obtained a new reference on this znode while we were
+ * processing it here, so we are done. If we still have
+ * mapped pages then we are also done, since we don't
+ * want to inactivate the znode until the pages get pushed.
+ *
+ * XXX - if vn_has_cached_data(vp) is true, but count == 0,
+ * this seems like it would leave the znode hanging with
+ * no chance to go inactive...
+ */
+ mutex_exit(&vp->v_lock);
+ mutex_exit(&zp->z_lock);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
+ return;
+ }
+ mutex_exit(&vp->v_lock);
+
+ /*
+ * If this was the last reference to a file with no links,
+ * remove the file from the file system.
+ */
+ if (zp->z_unlinked) {
+ mutex_exit(&zp->z_lock);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
+ zfs_rmnode(zp);
+ return;
+ }
+ mutex_exit(&zp->z_lock);
+ zfs_znode_dmu_fini(zp);
+ ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
+ zfs_znode_free(zp);
+}
+
+void
+zfs_znode_free(znode_t *zp)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+
+ vn_invalid(ZTOV(zp));
+
+ ASSERT(ZTOV(zp)->v_count == 0);
+
+ mutex_enter(&zfsvfs->z_znodes_lock);
+ POINTER_INVALIDATE(&zp->z_zfsvfs);
+ list_remove(&zfsvfs->z_all_znodes, zp);
+ mutex_exit(&zfsvfs->z_znodes_lock);
+
+ kmem_cache_free(znode_cache, zp);
+
+ VFS_RELE(zfsvfs->z_vfs);
+}
+
+void
+zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx)
+{
+ timestruc_t now;
+
+ ASSERT(MUTEX_HELD(&zp->z_lock));
+
+ gethrestime(&now);
+
+ if (tx) {
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+ zp->z_atime_dirty = 0;
+ zp->z_seq++;
+ } else {
+ zp->z_atime_dirty = 1;
+ }
+
+ if (flag & AT_ATIME)
+ ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime);
+
+ if (flag & AT_MTIME) {
+ ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime);
+ if (zp->z_zfsvfs->z_use_fuids)
+ zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED);
+ }
+
+ if (flag & AT_CTIME) {
+ ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime);
+ if (zp->z_zfsvfs->z_use_fuids)
+ zp->z_phys->zp_flags |= ZFS_ARCHIVE;
+ }
+}
+
+/*
+ * Update the requested znode timestamps with the current time.
+ * If we are in a transaction, then go ahead and mark the znode
+ * dirty in the transaction so the timestamps will go to disk.
+ * Otherwise, we will get pushed next time the znode is updated
+ * in a transaction, or when this znode eventually goes inactive.
+ *
+ * Why is this OK?
+ * 1 - Only the ACCESS time is ever updated outside of a transaction.
+ * 2 - Multiple consecutive updates will be collapsed into a single
+ * znode update by the transaction grouping semantics of the DMU.
+ */
+void
+zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx)
+{
+ mutex_enter(&zp->z_lock);
+ zfs_time_stamper_locked(zp, flag, tx);
+ mutex_exit(&zp->z_lock);
+}
+
+/*
+ * Grow the block size for a file.
+ *
+ * IN: zp - znode of file to free data in.
+ * size - requested block size
+ * tx - open transaction.
+ *
+ * NOTE: this function assumes that the znode is write locked.
+ */
+void
+zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
+{
+ int error;
+ u_longlong_t dummy;
+
+ if (size <= zp->z_blksz)
+ return;
+ /*
+ * If the file size is already greater than the current blocksize,
+ * we will not grow. If there is more than one block in a file,
+ * the blocksize cannot change.
+ */
+ if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz)
+ return;
+
+ error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
+ size, 0, tx);
+ if (error == ENOTSUP)
+ return;
+ ASSERT3U(error, ==, 0);
+
+ /* What blocksize did we actually get? */
+ dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy);
+}
+
+/*
+ * This is a dummy interface used when pvn_vplist_dirty() should *not*
+ * be calling back into the fs for a putpage(). E.g.: when truncating
+ * a file, the pages being "thrown away* don't need to be written out.
+ */
+/* ARGSUSED */
+static int
+zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
+ int flags, cred_t *cr)
+{
+ ASSERT(0);
+ return (0);
+}
+
+/*
+ * Increase the file length
+ *
+ * IN: zp - znode of file to free data in.
+ * end - new end-of-file
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ */
+static int
+zfs_extend(znode_t *zp, uint64_t end)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ dmu_tx_t *tx;
+ rl_t *rl;
+ uint64_t newblksz;
+ int error;
+
+ /*
+ * We will change zp_size, lock the whole file.
+ */
+ rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+
+ /*
+ * Nothing to do if file already at desired length.
+ */
+ if (end <= zp->z_phys->zp_size) {
+ zfs_range_unlock(rl);
+ return (0);
+ }
+top:
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ if (end > zp->z_blksz &&
+ (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
+ /*
+ * We are growing the file past the current block size.
+ */
+ if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
+ ASSERT(!ISP2(zp->z_blksz));
+ newblksz = MIN(end, SPA_MAXBLOCKSIZE);
+ } else {
+ newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
+ }
+ dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
+ } else {
+ newblksz = 0;
+ }
+
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ zfs_range_unlock(rl);
+ return (error);
+ }
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+
+ if (newblksz)
+ zfs_grow_blocksize(zp, newblksz, tx);
+
+ zp->z_phys->zp_size = end;
+
+ zfs_range_unlock(rl);
+
+ dmu_tx_commit(tx);
+
+ return (0);
+}
+
+/*
+ * Free space in a file.
+ *
+ * IN: zp - znode of file to free data in.
+ * off - start of section to free.
+ * len - length of section to free.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ */
+static int
+zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ rl_t *rl;
+ int error;
+
+ /*
+ * Lock the range being freed.
+ */
+ rl = zfs_range_lock(zp, off, len, RL_WRITER);
+
+ /*
+ * Nothing to do if file already at desired length.
+ */
+ if (off >= zp->z_phys->zp_size) {
+ zfs_range_unlock(rl);
+ return (0);
+ }
+
+ if (off + len > zp->z_phys->zp_size)
+ len = zp->z_phys->zp_size - off;
+
+ error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
+
+ zfs_range_unlock(rl);
+
+ return (error);
+}
+
+/*
+ * Truncate a file
+ *
+ * IN: zp - znode of file to free data in.
+ * end - new end-of-file.
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ */
+static int
+zfs_trunc(znode_t *zp, uint64_t end)
+{
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ vnode_t *vp = ZTOV(zp);
+ dmu_tx_t *tx;
+ rl_t *rl;
+ int error;
+
+ /*
+ * We will change zp_size, lock the whole file.
+ */
+ rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+
+ /*
+ * Nothing to do if file already at desired length.
+ */
+ if (end >= zp->z_phys->zp_size) {
+ zfs_range_unlock(rl);
+ return (0);
+ }
+
+ error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
+ if (error) {
+ zfs_range_unlock(rl);
+ return (error);
+ }
+top:
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto top;
+ }
+ dmu_tx_abort(tx);
+ zfs_range_unlock(rl);
+ return (error);
+ }
+ dmu_buf_will_dirty(zp->z_dbuf, tx);
+
+ zp->z_phys->zp_size = end;
+
+ dmu_tx_commit(tx);
+
+ zfs_range_unlock(rl);
+
+ /*
+ * Clear any mapped pages in the truncated region. This has to
+ * happen outside of the transaction to avoid the possibility of
+ * a deadlock with someone trying to push a page that we are
+ * about to invalidate.
+ */
+ rw_enter(&zp->z_map_lock, RW_WRITER);
+ if (vn_has_cached_data(vp)) {
+ page_t *pp;
+ uint64_t start = end & PAGEMASK;
+ int poff = end & PAGEOFFSET;
+
+ if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
+ /*
+ * We need to zero a partial page.
+ */
+ pagezero(pp, poff, PAGESIZE - poff);
+ start += PAGESIZE;
+ page_unlock(pp);
+ }
+ error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
+ B_INVAL | B_TRUNC, NULL);
+ ASSERT(error == 0);
+ }
+ rw_exit(&zp->z_map_lock);
+
+ return (0);
+}
+
+/*
+ * Free space in a file
+ *
+ * IN: zp - znode of file to free data in.
+ * off - start of range
+ * len - end of range (0 => EOF)
+ * flag - current file open mode flags.
+ * log - TRUE if this action should be logged
+ *
+ * RETURN: 0 if success
+ * error code if failure
+ */
+int
+zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
+{
+ vnode_t *vp = ZTOV(zp);
+ dmu_tx_t *tx;
+ zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+ zilog_t *zilog = zfsvfs->z_log;
+ int error;
+
+ if (off > zp->z_phys->zp_size) {
+ error = zfs_extend(zp, off+len);
+ if (error == 0 && log)
+ goto log;
+ else
+ return (error);
+ }
+
+ /*
+ * Check for any locks in the region to be freed.
+ */
+ if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) {
+ uint64_t length = (len ? len : zp->z_phys->zp_size - off);
+ if (error = chklock(vp, FWRITE, off, length, flag, NULL))
+ return (error);
+ }
+
+ if (len == 0) {
+ error = zfs_trunc(zp, off);
+ } else {
+ if ((error = zfs_free_range(zp, off, len)) == 0 &&
+ off + len > zp->z_phys->zp_size)
+ error = zfs_extend(zp, off+len);
+ }
+ if (error || !log)
+ return (error);
+log:
+ tx = dmu_tx_create(zfsvfs->z_os);
+ dmu_tx_hold_bonus(tx, zp->z_id);
+ error = dmu_tx_assign(tx, zfsvfs->z_assign);
+ if (error) {
+ if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
+ dmu_tx_wait(tx);
+ dmu_tx_abort(tx);
+ goto log;
+ }
+ dmu_tx_abort(tx);
+ return (error);
+ }
+
+ zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
+ zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
+
+ dmu_tx_commit(tx);
+ return (0);
+}
+
+void
+zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
+{
+ zfsvfs_t zfsvfs;
+ uint64_t moid, doid, version;
+ uint64_t sense = ZFS_CASE_SENSITIVE;
+ uint64_t norm = 0;
+ nvpair_t *elem;
+ int error;
+ znode_t *rootzp = NULL;
+ vnode_t *vp;
+ vattr_t vattr;
+ znode_t *zp;
+
+ /*
+ * First attempt to create master node.
+ */
+ /*
+ * In an empty objset, there are no blocks to read and thus
+ * there can be no i/o errors (which we assert below).
+ */
+ moid = MASTER_NODE_OBJ;
+ error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
+ DMU_OT_NONE, 0, tx);
+ ASSERT(error == 0);
+
+ /*
+ * Set starting attributes.
+ */
+ if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
+ version = ZPL_VERSION;
+ else
+ version = ZPL_VERSION_FUID - 1;
+ error = zap_update(os, moid, ZPL_VERSION_STR,
+ 8, 1, &version, tx);
+ elem = NULL;
+ while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
+ /* For the moment we expect all zpl props to be uint64_ts */
+ uint64_t val;
+ char *name;
+
+ ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
+ VERIFY(nvpair_value_uint64(elem, &val) == 0);
+ name = nvpair_name(elem);
+ if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
+ version = val;
+ error = zap_update(os, moid, ZPL_VERSION_STR,
+ 8, 1, &version, tx);
+ } else {
+ error = zap_update(os, moid, name, 8, 1, &val, tx);
+ }
+ ASSERT(error == 0);
+ if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
+ norm = val;
+ else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
+ sense = val;
+ }
+ ASSERT(version != 0);
+
+ /*
+ * Create a delete queue.
+ */
+ doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
+
+ error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx);
+ ASSERT(error == 0);
+
+ /*
+ * Create root znode. Create minimal znode/vnode/zfsvfs
+ * to allow zfs_mknode to work.
+ */
+ vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
+ vattr.va_type = VDIR;
+ vattr.va_mode = S_IFDIR|0755;
+ vattr.va_uid = crgetuid(cr);
+ vattr.va_gid = crgetgid(cr);
+
+ rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
+ rootzp->z_unlinked = 0;
+ rootzp->z_atime_dirty = 0;
+
+ vp = ZTOV(rootzp);
+ vn_reinit(vp);
+ vp->v_type = VDIR;
+
+ bzero(&zfsvfs, sizeof (zfsvfs_t));
+
+ zfsvfs.z_os = os;
+ zfsvfs.z_assign = TXG_NOWAIT;
+ zfsvfs.z_parent = &zfsvfs;
+ zfsvfs.z_version = version;
+ zfsvfs.z_use_fuids = USE_FUIDS(version, os);
+ zfsvfs.z_norm = norm;
+ /*
+ * Fold case on file systems that are always or sometimes case
+ * insensitive.
+ */
+ if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
+ zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
+
+ mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
+ offsetof(znode_t, z_link_node));
+
+ ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
+ rootzp->z_zfsvfs = &zfsvfs;
+ zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, NULL, NULL);
+ ASSERT3P(zp, ==, rootzp);
+ ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
+ error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
+ ASSERT(error == 0);
+ POINTER_INVALIDATE(&rootzp->z_zfsvfs);
+
+ ZTOV(rootzp)->v_count = 0;
+ dmu_buf_rele(rootzp->z_dbuf, NULL);
+ rootzp->z_dbuf = NULL;
+ kmem_cache_free(znode_cache, rootzp);
+}
+
+#endif /* _KERNEL */
+/*
+ * Given an object number, return its parent object number and whether
+ * or not the object is an extended attribute directory.
+ */
+static int
+zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir)
+{
+ dmu_buf_t *db;
+ dmu_object_info_t doi;
+ znode_phys_t *zp;
+ int error;
+
+ if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0)
+ return (error);
+
+ dmu_object_info_from_db(db, &doi);
+ if (doi.doi_bonus_type != DMU_OT_ZNODE ||
+ doi.doi_bonus_size < sizeof (znode_phys_t)) {
+ dmu_buf_rele(db, FTAG);
+ return (EINVAL);
+ }
+
+ zp = db->db_data;
+ *pobjp = zp->zp_parent;
+ *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) &&
+ S_ISDIR(zp->zp_mode);
+ dmu_buf_rele(db, FTAG);
+
+ return (0);
+}
+
+int
+zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
+{
+ char *path = buf + len - 1;
+ int error;
+
+ *path = '\0';
+
+ for (;;) {
+ uint64_t pobj;
+ char component[MAXNAMELEN + 2];
+ size_t complen;
+ int is_xattrdir;
+
+ if ((error = zfs_obj_to_pobj(osp, obj, &pobj,
+ &is_xattrdir)) != 0)
+ break;
+
+ if (pobj == obj) {
+ if (path[0] != '/')
+ *--path = '/';
+ break;
+ }
+
+ component[0] = '/';
+ if (is_xattrdir) {
+ (void) sprintf(component + 1, "<xattrdir>");
+ } else {
+ error = zap_value_search(osp, pobj, obj,
+ ZFS_DIRENT_OBJ(-1ULL), component + 1);
+ if (error != 0)
+ break;
+ }
+
+ complen = strlen(component);
+ path -= complen;
+ ASSERT(path >= buf);
+ bcopy(component, path, complen);
+ obj = pobj;
+ }
+
+ if (error == 0)
+ (void) memmove(buf, path, buf + len - path);
+ return (error);
+}
diff --git a/module/zfs/zil.c b/module/zfs/zil.c
new file mode 100644
index 000000000..95101882b
--- /dev/null
+++ b/module/zfs/zil.c
@@ -0,0 +1,1735 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/zap.h>
+#include <sys/arc.h>
+#include <sys/stat.h>
+#include <sys/resource.h>
+#include <sys/zil.h>
+#include <sys/zil_impl.h>
+#include <sys/dsl_dataset.h>
+#include <sys/vdev.h>
+#include <sys/dmu_tx.h>
+
+/*
+ * The zfs intent log (ZIL) saves transaction records of system calls
+ * that change the file system in memory with enough information
+ * to be able to replay them. These are stored in memory until
+ * either the DMU transaction group (txg) commits them to the stable pool
+ * and they can be discarded, or they are flushed to the stable log
+ * (also in the pool) due to a fsync, O_DSYNC or other synchronous
+ * requirement. In the event of a panic or power fail then those log
+ * records (transactions) are replayed.
+ *
+ * There is one ZIL per file system. Its on-disk (pool) format consists
+ * of 3 parts:
+ *
+ * - ZIL header
+ * - ZIL blocks
+ * - ZIL records
+ *
+ * A log record holds a system call transaction. Log blocks can
+ * hold many log records and the blocks are chained together.
+ * Each ZIL block contains a block pointer (blkptr_t) to the next
+ * ZIL block in the chain. The ZIL header points to the first
+ * block in the chain. Note there is not a fixed place in the pool
+ * to hold blocks. They are dynamically allocated and freed as
+ * needed from the blocks available. Figure X shows the ZIL structure:
+ */
+
+/*
+ * This global ZIL switch affects all pools
+ */
+int zil_disable = 0; /* disable intent logging */
+
+/*
+ * Tunable parameter for debugging or performance analysis. Setting
+ * zfs_nocacheflush will cause corruption on power loss if a volatile
+ * out-of-order write cache is enabled.
+ */
+boolean_t zfs_nocacheflush = B_FALSE;
+
+static kmem_cache_t *zil_lwb_cache;
+
+static int
+zil_dva_compare(const void *x1, const void *x2)
+{
+ const dva_t *dva1 = x1;
+ const dva_t *dva2 = x2;
+
+ if (DVA_GET_VDEV(dva1) < DVA_GET_VDEV(dva2))
+ return (-1);
+ if (DVA_GET_VDEV(dva1) > DVA_GET_VDEV(dva2))
+ return (1);
+
+ if (DVA_GET_OFFSET(dva1) < DVA_GET_OFFSET(dva2))
+ return (-1);
+ if (DVA_GET_OFFSET(dva1) > DVA_GET_OFFSET(dva2))
+ return (1);
+
+ return (0);
+}
+
+static void
+zil_dva_tree_init(avl_tree_t *t)
+{
+ avl_create(t, zil_dva_compare, sizeof (zil_dva_node_t),
+ offsetof(zil_dva_node_t, zn_node));
+}
+
+static void
+zil_dva_tree_fini(avl_tree_t *t)
+{
+ zil_dva_node_t *zn;
+ void *cookie = NULL;
+
+ while ((zn = avl_destroy_nodes(t, &cookie)) != NULL)
+ kmem_free(zn, sizeof (zil_dva_node_t));
+
+ avl_destroy(t);
+}
+
+static int
+zil_dva_tree_add(avl_tree_t *t, dva_t *dva)
+{
+ zil_dva_node_t *zn;
+ avl_index_t where;
+
+ if (avl_find(t, dva, &where) != NULL)
+ return (EEXIST);
+
+ zn = kmem_alloc(sizeof (zil_dva_node_t), KM_SLEEP);
+ zn->zn_dva = *dva;
+ avl_insert(t, zn, where);
+
+ return (0);
+}
+
+static zil_header_t *
+zil_header_in_syncing_context(zilog_t *zilog)
+{
+ return ((zil_header_t *)zilog->zl_header);
+}
+
+static void
+zil_init_log_chain(zilog_t *zilog, blkptr_t *bp)
+{
+ zio_cksum_t *zc = &bp->blk_cksum;
+
+ zc->zc_word[ZIL_ZC_GUID_0] = spa_get_random(-1ULL);
+ zc->zc_word[ZIL_ZC_GUID_1] = spa_get_random(-1ULL);
+ zc->zc_word[ZIL_ZC_OBJSET] = dmu_objset_id(zilog->zl_os);
+ zc->zc_word[ZIL_ZC_SEQ] = 1ULL;
+}
+
+/*
+ * Read a log block, make sure it's valid, and byteswap it if necessary.
+ */
+static int
+zil_read_log_block(zilog_t *zilog, const blkptr_t *bp, arc_buf_t **abufpp)
+{
+ blkptr_t blk = *bp;
+ zbookmark_t zb;
+ uint32_t aflags = ARC_WAIT;
+ int error;
+
+ zb.zb_objset = bp->blk_cksum.zc_word[ZIL_ZC_OBJSET];
+ zb.zb_object = 0;
+ zb.zb_level = -1;
+ zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ];
+
+ *abufpp = NULL;
+
+ /*
+ * We shouldn't be doing any scrubbing while we're doing log
+ * replay, it's OK to not lock.
+ */
+ error = arc_read_nolock(NULL, zilog->zl_spa, &blk,
+ arc_getbuf_func, abufpp, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL |
+ ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB, &aflags, &zb);
+
+ if (error == 0) {
+ char *data = (*abufpp)->b_data;
+ uint64_t blksz = BP_GET_LSIZE(bp);
+ zil_trailer_t *ztp = (zil_trailer_t *)(data + blksz) - 1;
+ zio_cksum_t cksum = bp->blk_cksum;
+
+ /*
+ * Validate the checksummed log block.
+ *
+ * Sequence numbers should be... sequential. The checksum
+ * verifier for the next block should be bp's checksum plus 1.
+ *
+ * Also check the log chain linkage and size used.
+ */
+ cksum.zc_word[ZIL_ZC_SEQ]++;
+
+ if (bcmp(&cksum, &ztp->zit_next_blk.blk_cksum,
+ sizeof (cksum)) || BP_IS_HOLE(&ztp->zit_next_blk) ||
+ (ztp->zit_nused > (blksz - sizeof (zil_trailer_t)))) {
+ error = ECKSUM;
+ }
+
+ if (error) {
+ VERIFY(arc_buf_remove_ref(*abufpp, abufpp) == 1);
+ *abufpp = NULL;
+ }
+ }
+
+ dprintf("error %d on %llu:%llu\n", error, zb.zb_objset, zb.zb_blkid);
+
+ return (error);
+}
+
+/*
+ * Parse the intent log, and call parse_func for each valid record within.
+ * Return the highest sequence number.
+ */
+uint64_t
+zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func,
+ zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg)
+{
+ const zil_header_t *zh = zilog->zl_header;
+ uint64_t claim_seq = zh->zh_claim_seq;
+ uint64_t seq = 0;
+ uint64_t max_seq = 0;
+ blkptr_t blk = zh->zh_log;
+ arc_buf_t *abuf;
+ char *lrbuf, *lrp;
+ zil_trailer_t *ztp;
+ int reclen, error;
+
+ if (BP_IS_HOLE(&blk))
+ return (max_seq);
+
+ /*
+ * Starting at the block pointed to by zh_log we read the log chain.
+ * For each block in the chain we strongly check that block to
+ * ensure its validity. We stop when an invalid block is found.
+ * For each block pointer in the chain we call parse_blk_func().
+ * For each record in each valid block we call parse_lr_func().
+ * If the log has been claimed, stop if we encounter a sequence
+ * number greater than the highest claimed sequence number.
+ */
+ zil_dva_tree_init(&zilog->zl_dva_tree);
+ for (;;) {
+ seq = blk.blk_cksum.zc_word[ZIL_ZC_SEQ];
+
+ if (claim_seq != 0 && seq > claim_seq)
+ break;
+
+ ASSERT(max_seq < seq);
+ max_seq = seq;
+
+ error = zil_read_log_block(zilog, &blk, &abuf);
+
+ if (parse_blk_func != NULL)
+ parse_blk_func(zilog, &blk, arg, txg);
+
+ if (error)
+ break;
+
+ lrbuf = abuf->b_data;
+ ztp = (zil_trailer_t *)(lrbuf + BP_GET_LSIZE(&blk)) - 1;
+ blk = ztp->zit_next_blk;
+
+ if (parse_lr_func == NULL) {
+ VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
+ continue;
+ }
+
+ for (lrp = lrbuf; lrp < lrbuf + ztp->zit_nused; lrp += reclen) {
+ lr_t *lr = (lr_t *)lrp;
+ reclen = lr->lrc_reclen;
+ ASSERT3U(reclen, >=, sizeof (lr_t));
+ parse_lr_func(zilog, lr, arg, txg);
+ }
+ VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
+ }
+ zil_dva_tree_fini(&zilog->zl_dva_tree);
+
+ return (max_seq);
+}
+
+/* ARGSUSED */
+static void
+zil_claim_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t first_txg)
+{
+ spa_t *spa = zilog->zl_spa;
+ int err;
+
+ /*
+ * Claim log block if not already committed and not already claimed.
+ */
+ if (bp->blk_birth >= first_txg &&
+ zil_dva_tree_add(&zilog->zl_dva_tree, BP_IDENTITY(bp)) == 0) {
+ err = zio_wait(zio_claim(NULL, spa, first_txg, bp, NULL, NULL,
+ ZIO_FLAG_MUSTSUCCEED));
+ ASSERT(err == 0);
+ }
+}
+
+static void
+zil_claim_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t first_txg)
+{
+ if (lrc->lrc_txtype == TX_WRITE) {
+ lr_write_t *lr = (lr_write_t *)lrc;
+ zil_claim_log_block(zilog, &lr->lr_blkptr, tx, first_txg);
+ }
+}
+
+/* ARGSUSED */
+static void
+zil_free_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t claim_txg)
+{
+ zio_free_blk(zilog->zl_spa, bp, dmu_tx_get_txg(tx));
+}
+
+static void
+zil_free_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t claim_txg)
+{
+ /*
+ * If we previously claimed it, we need to free it.
+ */
+ if (claim_txg != 0 && lrc->lrc_txtype == TX_WRITE) {
+ lr_write_t *lr = (lr_write_t *)lrc;
+ blkptr_t *bp = &lr->lr_blkptr;
+ if (bp->blk_birth >= claim_txg &&
+ !zil_dva_tree_add(&zilog->zl_dva_tree, BP_IDENTITY(bp))) {
+ (void) arc_free(NULL, zilog->zl_spa,
+ dmu_tx_get_txg(tx), bp, NULL, NULL, ARC_WAIT);
+ }
+ }
+}
+
+/*
+ * Create an on-disk intent log.
+ */
+static void
+zil_create(zilog_t *zilog)
+{
+ const zil_header_t *zh = zilog->zl_header;
+ lwb_t *lwb;
+ uint64_t txg = 0;
+ dmu_tx_t *tx = NULL;
+ blkptr_t blk;
+ int error = 0;
+
+ /*
+ * Wait for any previous destroy to complete.
+ */
+ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
+
+ ASSERT(zh->zh_claim_txg == 0);
+ ASSERT(zh->zh_replay_seq == 0);
+
+ blk = zh->zh_log;
+
+ /*
+ * If we don't already have an initial log block, allocate one now.
+ */
+ if (BP_IS_HOLE(&blk)) {
+ tx = dmu_tx_create(zilog->zl_os);
+ (void) dmu_tx_assign(tx, TXG_WAIT);
+ dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
+ txg = dmu_tx_get_txg(tx);
+
+ error = zio_alloc_blk(zilog->zl_spa, ZIL_MIN_BLKSZ, &blk,
+ NULL, txg);
+
+ if (error == 0)
+ zil_init_log_chain(zilog, &blk);
+ }
+
+ /*
+ * Allocate a log write buffer (lwb) for the first log block.
+ */
+ if (error == 0) {
+ lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
+ lwb->lwb_zilog = zilog;
+ lwb->lwb_blk = blk;
+ lwb->lwb_nused = 0;
+ lwb->lwb_sz = BP_GET_LSIZE(&lwb->lwb_blk);
+ lwb->lwb_buf = zio_buf_alloc(lwb->lwb_sz);
+ lwb->lwb_max_txg = txg;
+ lwb->lwb_zio = NULL;
+
+ mutex_enter(&zilog->zl_lock);
+ list_insert_tail(&zilog->zl_lwb_list, lwb);
+ mutex_exit(&zilog->zl_lock);
+ }
+
+ /*
+ * If we just allocated the first log block, commit our transaction
+ * and wait for zil_sync() to stuff the block poiner into zh_log.
+ * (zh is part of the MOS, so we cannot modify it in open context.)
+ */
+ if (tx != NULL) {
+ dmu_tx_commit(tx);
+ txg_wait_synced(zilog->zl_dmu_pool, txg);
+ }
+
+ ASSERT(bcmp(&blk, &zh->zh_log, sizeof (blk)) == 0);
+}
+
+/*
+ * In one tx, free all log blocks and clear the log header.
+ * If keep_first is set, then we're replaying a log with no content.
+ * We want to keep the first block, however, so that the first
+ * synchronous transaction doesn't require a txg_wait_synced()
+ * in zil_create(). We don't need to txg_wait_synced() here either
+ * when keep_first is set, because both zil_create() and zil_destroy()
+ * will wait for any in-progress destroys to complete.
+ */
+void
+zil_destroy(zilog_t *zilog, boolean_t keep_first)
+{
+ const zil_header_t *zh = zilog->zl_header;
+ lwb_t *lwb;
+ dmu_tx_t *tx;
+ uint64_t txg;
+
+ /*
+ * Wait for any previous destroy to complete.
+ */
+ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
+
+ if (BP_IS_HOLE(&zh->zh_log))
+ return;
+
+ tx = dmu_tx_create(zilog->zl_os);
+ (void) dmu_tx_assign(tx, TXG_WAIT);
+ dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
+ txg = dmu_tx_get_txg(tx);
+
+ mutex_enter(&zilog->zl_lock);
+
+ /*
+ * It is possible for the ZIL to get the previously mounted zilog
+ * structure of the same dataset if quickly remounted and the dbuf
+ * eviction has not completed. In this case we can see a non
+ * empty lwb list and keep_first will be set. We fix this by
+ * clearing the keep_first. This will be slower but it's very rare.
+ */
+ if (!list_is_empty(&zilog->zl_lwb_list) && keep_first)
+ keep_first = B_FALSE;
+
+ ASSERT3U(zilog->zl_destroy_txg, <, txg);
+ zilog->zl_destroy_txg = txg;
+ zilog->zl_keep_first = keep_first;
+
+ if (!list_is_empty(&zilog->zl_lwb_list)) {
+ ASSERT(zh->zh_claim_txg == 0);
+ ASSERT(!keep_first);
+ while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
+ list_remove(&zilog->zl_lwb_list, lwb);
+ if (lwb->lwb_buf != NULL)
+ zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
+ zio_free_blk(zilog->zl_spa, &lwb->lwb_blk, txg);
+ kmem_cache_free(zil_lwb_cache, lwb);
+ }
+ } else {
+ if (!keep_first) {
+ (void) zil_parse(zilog, zil_free_log_block,
+ zil_free_log_record, tx, zh->zh_claim_txg);
+ }
+ }
+ mutex_exit(&zilog->zl_lock);
+
+ dmu_tx_commit(tx);
+}
+
+/*
+ * zil_rollback_destroy() is only called by the rollback code.
+ * We already have a syncing tx. Rollback has exclusive access to the
+ * dataset, so we don't have to worry about concurrent zil access.
+ * The actual freeing of any log blocks occurs in zil_sync() later in
+ * this txg syncing phase.
+ */
+void
+zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx)
+{
+ const zil_header_t *zh = zilog->zl_header;
+ uint64_t txg;
+
+ if (BP_IS_HOLE(&zh->zh_log))
+ return;
+
+ txg = dmu_tx_get_txg(tx);
+ ASSERT3U(zilog->zl_destroy_txg, <, txg);
+ zilog->zl_destroy_txg = txg;
+ zilog->zl_keep_first = B_FALSE;
+
+ /*
+ * Ensure there's no outstanding ZIL IO. No lwbs or just the
+ * unused one that allocated in advance is ok.
+ */
+ ASSERT(zilog->zl_lwb_list.list_head.list_next ==
+ zilog->zl_lwb_list.list_head.list_prev);
+ (void) zil_parse(zilog, zil_free_log_block, zil_free_log_record,
+ tx, zh->zh_claim_txg);
+}
+
+int
+zil_claim(char *osname, void *txarg)
+{
+ dmu_tx_t *tx = txarg;
+ uint64_t first_txg = dmu_tx_get_txg(tx);
+ zilog_t *zilog;
+ zil_header_t *zh;
+ objset_t *os;
+ int error;
+
+ error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
+ if (error) {
+ cmn_err(CE_WARN, "can't open objset for %s", osname);
+ return (0);
+ }
+
+ zilog = dmu_objset_zil(os);
+ zh = zil_header_in_syncing_context(zilog);
+
+ /*
+ * Claim all log blocks if we haven't already done so, and remember
+ * the highest claimed sequence number. This ensures that if we can
+ * read only part of the log now (e.g. due to a missing device),
+ * but we can read the entire log later, we will not try to replay
+ * or destroy beyond the last block we successfully claimed.
+ */
+ ASSERT3U(zh->zh_claim_txg, <=, first_txg);
+ if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) {
+ zh->zh_claim_txg = first_txg;
+ zh->zh_claim_seq = zil_parse(zilog, zil_claim_log_block,
+ zil_claim_log_record, tx, first_txg);
+ dsl_dataset_dirty(dmu_objset_ds(os), tx);
+ }
+
+ ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1));
+ dmu_objset_close(os);
+ return (0);
+}
+
+/*
+ * Check the log by walking the log chain.
+ * Checksum errors are ok as they indicate the end of the chain.
+ * Any other error (no device or read failure) returns an error.
+ */
+/* ARGSUSED */
+int
+zil_check_log_chain(char *osname, void *txarg)
+{
+ zilog_t *zilog;
+ zil_header_t *zh;
+ blkptr_t blk;
+ arc_buf_t *abuf;
+ objset_t *os;
+ char *lrbuf;
+ zil_trailer_t *ztp;
+ int error;
+
+ error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
+ if (error) {
+ cmn_err(CE_WARN, "can't open objset for %s", osname);
+ return (0);
+ }
+
+ zilog = dmu_objset_zil(os);
+ zh = zil_header_in_syncing_context(zilog);
+ blk = zh->zh_log;
+ if (BP_IS_HOLE(&blk)) {
+ dmu_objset_close(os);
+ return (0); /* no chain */
+ }
+
+ for (;;) {
+ error = zil_read_log_block(zilog, &blk, &abuf);
+ if (error)
+ break;
+ lrbuf = abuf->b_data;
+ ztp = (zil_trailer_t *)(lrbuf + BP_GET_LSIZE(&blk)) - 1;
+ blk = ztp->zit_next_blk;
+ VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
+ }
+ dmu_objset_close(os);
+ if (error == ECKSUM)
+ return (0); /* normal end of chain */
+ return (error);
+}
+
+/*
+ * Clear a log chain
+ */
+/* ARGSUSED */
+int
+zil_clear_log_chain(char *osname, void *txarg)
+{
+ zilog_t *zilog;
+ zil_header_t *zh;
+ objset_t *os;
+ dmu_tx_t *tx;
+ int error;
+
+ error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
+ if (error) {
+ cmn_err(CE_WARN, "can't open objset for %s", osname);
+ return (0);
+ }
+
+ zilog = dmu_objset_zil(os);
+ tx = dmu_tx_create(zilog->zl_os);
+ (void) dmu_tx_assign(tx, TXG_WAIT);
+ zh = zil_header_in_syncing_context(zilog);
+ BP_ZERO(&zh->zh_log);
+ dsl_dataset_dirty(dmu_objset_ds(os), tx);
+ dmu_tx_commit(tx);
+ dmu_objset_close(os);
+ return (0);
+}
+
+static int
+zil_vdev_compare(const void *x1, const void *x2)
+{
+ uint64_t v1 = ((zil_vdev_node_t *)x1)->zv_vdev;
+ uint64_t v2 = ((zil_vdev_node_t *)x2)->zv_vdev;
+
+ if (v1 < v2)
+ return (-1);
+ if (v1 > v2)
+ return (1);
+
+ return (0);
+}
+
+void
+zil_add_block(zilog_t *zilog, blkptr_t *bp)
+{
+ avl_tree_t *t = &zilog->zl_vdev_tree;
+ avl_index_t where;
+ zil_vdev_node_t *zv, zvsearch;
+ int ndvas = BP_GET_NDVAS(bp);
+ int i;
+
+ if (zfs_nocacheflush)
+ return;
+
+ ASSERT(zilog->zl_writer);
+
+ /*
+ * Even though we're zl_writer, we still need a lock because the
+ * zl_get_data() callbacks may have dmu_sync() done callbacks
+ * that will run concurrently.
+ */
+ mutex_enter(&zilog->zl_vdev_lock);
+ for (i = 0; i < ndvas; i++) {
+ zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
+ if (avl_find(t, &zvsearch, &where) == NULL) {
+ zv = kmem_alloc(sizeof (*zv), KM_SLEEP);
+ zv->zv_vdev = zvsearch.zv_vdev;
+ avl_insert(t, zv, where);
+ }
+ }
+ mutex_exit(&zilog->zl_vdev_lock);
+}
+
+void
+zil_flush_vdevs(zilog_t *zilog)
+{
+ spa_t *spa = zilog->zl_spa;
+ avl_tree_t *t = &zilog->zl_vdev_tree;
+ void *cookie = NULL;
+ zil_vdev_node_t *zv;
+ zio_t *zio;
+
+ ASSERT(zilog->zl_writer);
+
+ /*
+ * We don't need zl_vdev_lock here because we're the zl_writer,
+ * and all zl_get_data() callbacks are done.
+ */
+ if (avl_numnodes(t) == 0)
+ return;
+
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+
+ zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
+
+ while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) {
+ vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev);
+ if (vd != NULL)
+ zio_flush(zio, vd);
+ kmem_free(zv, sizeof (*zv));
+ }
+
+ /*
+ * Wait for all the flushes to complete. Not all devices actually
+ * support the DKIOCFLUSHWRITECACHE ioctl, so it's OK if it fails.
+ */
+ (void) zio_wait(zio);
+
+ spa_config_exit(spa, SCL_STATE, FTAG);
+}
+
+/*
+ * Function called when a log block write completes
+ */
+static void
+zil_lwb_write_done(zio_t *zio)
+{
+ lwb_t *lwb = zio->io_private;
+ zilog_t *zilog = lwb->lwb_zilog;
+
+ ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
+ ASSERT(BP_GET_CHECKSUM(zio->io_bp) == ZIO_CHECKSUM_ZILOG);
+ ASSERT(BP_GET_TYPE(zio->io_bp) == DMU_OT_INTENT_LOG);
+ ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
+ ASSERT(BP_GET_BYTEORDER(zio->io_bp) == ZFS_HOST_BYTEORDER);
+ ASSERT(!BP_IS_GANG(zio->io_bp));
+ ASSERT(!BP_IS_HOLE(zio->io_bp));
+ ASSERT(zio->io_bp->blk_fill == 0);
+
+ /*
+ * Now that we've written this log block, we have a stable pointer
+ * to the next block in the chain, so it's OK to let the txg in
+ * which we allocated the next block sync.
+ */
+ txg_rele_to_sync(&lwb->lwb_txgh);
+
+ zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
+ mutex_enter(&zilog->zl_lock);
+ lwb->lwb_buf = NULL;
+ if (zio->io_error)
+ zilog->zl_log_error = B_TRUE;
+ mutex_exit(&zilog->zl_lock);
+}
+
+/*
+ * Initialize the io for a log block.
+ */
+static void
+zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb)
+{
+ zbookmark_t zb;
+
+ zb.zb_objset = lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET];
+ zb.zb_object = 0;
+ zb.zb_level = -1;
+ zb.zb_blkid = lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ];
+
+ if (zilog->zl_root_zio == NULL) {
+ zilog->zl_root_zio = zio_root(zilog->zl_spa, NULL, NULL,
+ ZIO_FLAG_CANFAIL);
+ }
+ if (lwb->lwb_zio == NULL) {
+ lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa,
+ 0, &lwb->lwb_blk, lwb->lwb_buf,
+ lwb->lwb_sz, zil_lwb_write_done, lwb,
+ ZIO_PRIORITY_LOG_WRITE, ZIO_FLAG_CANFAIL, &zb);
+ }
+}
+
+/*
+ * Start a log block write and advance to the next log block.
+ * Calls are serialized.
+ */
+static lwb_t *
+zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb)
+{
+ lwb_t *nlwb;
+ zil_trailer_t *ztp = (zil_trailer_t *)(lwb->lwb_buf + lwb->lwb_sz) - 1;
+ spa_t *spa = zilog->zl_spa;
+ blkptr_t *bp = &ztp->zit_next_blk;
+ uint64_t txg;
+ uint64_t zil_blksz;
+ int error;
+
+ ASSERT(lwb->lwb_nused <= ZIL_BLK_DATA_SZ(lwb));
+
+ /*
+ * Allocate the next block and save its address in this block
+ * before writing it in order to establish the log chain.
+ * Note that if the allocation of nlwb synced before we wrote
+ * the block that points at it (lwb), we'd leak it if we crashed.
+ * Therefore, we don't do txg_rele_to_sync() until zil_lwb_write_done().
+ */
+ txg = txg_hold_open(zilog->zl_dmu_pool, &lwb->lwb_txgh);
+ txg_rele_to_quiesce(&lwb->lwb_txgh);
+
+ /*
+ * Pick a ZIL blocksize. We request a size that is the
+ * maximum of the previous used size, the current used size and
+ * the amount waiting in the queue.
+ */
+ zil_blksz = MAX(zilog->zl_prev_used,
+ zilog->zl_cur_used + sizeof (*ztp));
+ zil_blksz = MAX(zil_blksz, zilog->zl_itx_list_sz + sizeof (*ztp));
+ zil_blksz = P2ROUNDUP_TYPED(zil_blksz, ZIL_MIN_BLKSZ, uint64_t);
+ if (zil_blksz > ZIL_MAX_BLKSZ)
+ zil_blksz = ZIL_MAX_BLKSZ;
+
+ BP_ZERO(bp);
+ /* pass the old blkptr in order to spread log blocks across devs */
+ error = zio_alloc_blk(spa, zil_blksz, bp, &lwb->lwb_blk, txg);
+ if (error) {
+ dmu_tx_t *tx = dmu_tx_create_assigned(zilog->zl_dmu_pool, txg);
+
+ /*
+ * We dirty the dataset to ensure that zil_sync() will
+ * be called to remove this lwb from our zl_lwb_list.
+ * Failing to do so, may leave an lwb with a NULL lwb_buf
+ * hanging around on the zl_lwb_list.
+ */
+ dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
+ dmu_tx_commit(tx);
+
+ /*
+ * Since we've just experienced an allocation failure so we
+ * terminate the current lwb and send it on its way.
+ */
+ ztp->zit_pad = 0;
+ ztp->zit_nused = lwb->lwb_nused;
+ ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum;
+ zio_nowait(lwb->lwb_zio);
+
+ /*
+ * By returning NULL the caller will call tx_wait_synced()
+ */
+ return (NULL);
+ }
+
+ ASSERT3U(bp->blk_birth, ==, txg);
+ ztp->zit_pad = 0;
+ ztp->zit_nused = lwb->lwb_nused;
+ ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum;
+ bp->blk_cksum = lwb->lwb_blk.blk_cksum;
+ bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++;
+
+ /*
+ * Allocate a new log write buffer (lwb).
+ */
+ nlwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
+
+ nlwb->lwb_zilog = zilog;
+ nlwb->lwb_blk = *bp;
+ nlwb->lwb_nused = 0;
+ nlwb->lwb_sz = BP_GET_LSIZE(&nlwb->lwb_blk);
+ nlwb->lwb_buf = zio_buf_alloc(nlwb->lwb_sz);
+ nlwb->lwb_max_txg = txg;
+ nlwb->lwb_zio = NULL;
+
+ /*
+ * Put new lwb at the end of the log chain
+ */
+ mutex_enter(&zilog->zl_lock);
+ list_insert_tail(&zilog->zl_lwb_list, nlwb);
+ mutex_exit(&zilog->zl_lock);
+
+ /* Record the block for later vdev flushing */
+ zil_add_block(zilog, &lwb->lwb_blk);
+
+ /*
+ * kick off the write for the old log block
+ */
+ dprintf_bp(&lwb->lwb_blk, "lwb %p txg %llu: ", lwb, txg);
+ ASSERT(lwb->lwb_zio);
+ zio_nowait(lwb->lwb_zio);
+
+ return (nlwb);
+}
+
+static lwb_t *
+zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
+{
+ lr_t *lrc = &itx->itx_lr; /* common log record */
+ lr_write_t *lr = (lr_write_t *)lrc;
+ uint64_t txg = lrc->lrc_txg;
+ uint64_t reclen = lrc->lrc_reclen;
+ uint64_t dlen;
+
+ if (lwb == NULL)
+ return (NULL);
+ ASSERT(lwb->lwb_buf != NULL);
+
+ if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY)
+ dlen = P2ROUNDUP_TYPED(
+ lr->lr_length, sizeof (uint64_t), uint64_t);
+ else
+ dlen = 0;
+
+ zilog->zl_cur_used += (reclen + dlen);
+
+ zil_lwb_write_init(zilog, lwb);
+
+ /*
+ * If this record won't fit in the current log block, start a new one.
+ */
+ if (lwb->lwb_nused + reclen + dlen > ZIL_BLK_DATA_SZ(lwb)) {
+ lwb = zil_lwb_write_start(zilog, lwb);
+ if (lwb == NULL)
+ return (NULL);
+ zil_lwb_write_init(zilog, lwb);
+ ASSERT(lwb->lwb_nused == 0);
+ if (reclen + dlen > ZIL_BLK_DATA_SZ(lwb)) {
+ txg_wait_synced(zilog->zl_dmu_pool, txg);
+ return (lwb);
+ }
+ }
+
+ /*
+ * Update the lrc_seq, to be log record sequence number. See zil.h
+ * Then copy the record to the log buffer.
+ */
+ lrc->lrc_seq = ++zilog->zl_lr_seq; /* we are single threaded */
+ bcopy(lrc, lwb->lwb_buf + lwb->lwb_nused, reclen);
+
+ /*
+ * If it's a write, fetch the data or get its blkptr as appropriate.
+ */
+ if (lrc->lrc_txtype == TX_WRITE) {
+ if (txg > spa_freeze_txg(zilog->zl_spa))
+ txg_wait_synced(zilog->zl_dmu_pool, txg);
+ if (itx->itx_wr_state != WR_COPIED) {
+ char *dbuf;
+ int error;
+
+ /* alignment is guaranteed */
+ lr = (lr_write_t *)(lwb->lwb_buf + lwb->lwb_nused);
+ if (dlen) {
+ ASSERT(itx->itx_wr_state == WR_NEED_COPY);
+ dbuf = lwb->lwb_buf + lwb->lwb_nused + reclen;
+ lr->lr_common.lrc_reclen += dlen;
+ } else {
+ ASSERT(itx->itx_wr_state == WR_INDIRECT);
+ dbuf = NULL;
+ }
+ error = zilog->zl_get_data(
+ itx->itx_private, lr, dbuf, lwb->lwb_zio);
+ if (error) {
+ ASSERT(error == ENOENT || error == EEXIST ||
+ error == EALREADY);
+ return (lwb);
+ }
+ }
+ }
+
+ lwb->lwb_nused += reclen + dlen;
+ lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg);
+ ASSERT3U(lwb->lwb_nused, <=, ZIL_BLK_DATA_SZ(lwb));
+ ASSERT3U(P2PHASE(lwb->lwb_nused, sizeof (uint64_t)), ==, 0);
+
+ return (lwb);
+}
+
+itx_t *
+zil_itx_create(uint64_t txtype, size_t lrsize)
+{
+ itx_t *itx;
+
+ lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t);
+
+ itx = kmem_alloc(offsetof(itx_t, itx_lr) + lrsize, KM_SLEEP);
+ itx->itx_lr.lrc_txtype = txtype;
+ itx->itx_lr.lrc_reclen = lrsize;
+ itx->itx_sod = lrsize; /* if write & WR_NEED_COPY will be increased */
+ itx->itx_lr.lrc_seq = 0; /* defensive */
+
+ return (itx);
+}
+
+uint64_t
+zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx)
+{
+ uint64_t seq;
+
+ ASSERT(itx->itx_lr.lrc_seq == 0);
+
+ mutex_enter(&zilog->zl_lock);
+ list_insert_tail(&zilog->zl_itx_list, itx);
+ zilog->zl_itx_list_sz += itx->itx_sod;
+ itx->itx_lr.lrc_txg = dmu_tx_get_txg(tx);
+ itx->itx_lr.lrc_seq = seq = ++zilog->zl_itx_seq;
+ mutex_exit(&zilog->zl_lock);
+
+ return (seq);
+}
+
+/*
+ * Free up all in-memory intent log transactions that have now been synced.
+ */
+static void
+zil_itx_clean(zilog_t *zilog)
+{
+ uint64_t synced_txg = spa_last_synced_txg(zilog->zl_spa);
+ uint64_t freeze_txg = spa_freeze_txg(zilog->zl_spa);
+ list_t clean_list;
+ itx_t *itx;
+
+ list_create(&clean_list, sizeof (itx_t), offsetof(itx_t, itx_node));
+
+ mutex_enter(&zilog->zl_lock);
+ /* wait for a log writer to finish walking list */
+ while (zilog->zl_writer) {
+ cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock);
+ }
+
+ /*
+ * Move the sync'd log transactions to a separate list so we can call
+ * kmem_free without holding the zl_lock.
+ *
+ * There is no need to set zl_writer as we don't drop zl_lock here
+ */
+ while ((itx = list_head(&zilog->zl_itx_list)) != NULL &&
+ itx->itx_lr.lrc_txg <= MIN(synced_txg, freeze_txg)) {
+ list_remove(&zilog->zl_itx_list, itx);
+ zilog->zl_itx_list_sz -= itx->itx_sod;
+ list_insert_tail(&clean_list, itx);
+ }
+ cv_broadcast(&zilog->zl_cv_writer);
+ mutex_exit(&zilog->zl_lock);
+
+ /* destroy sync'd log transactions */
+ while ((itx = list_head(&clean_list)) != NULL) {
+ list_remove(&clean_list, itx);
+ kmem_free(itx, offsetof(itx_t, itx_lr)
+ + itx->itx_lr.lrc_reclen);
+ }
+ list_destroy(&clean_list);
+}
+
+/*
+ * If there are any in-memory intent log transactions which have now been
+ * synced then start up a taskq to free them.
+ */
+void
+zil_clean(zilog_t *zilog)
+{
+ itx_t *itx;
+
+ mutex_enter(&zilog->zl_lock);
+ itx = list_head(&zilog->zl_itx_list);
+ if ((itx != NULL) &&
+ (itx->itx_lr.lrc_txg <= spa_last_synced_txg(zilog->zl_spa))) {
+ (void) taskq_dispatch(zilog->zl_clean_taskq,
+ (void (*)(void *))zil_itx_clean, zilog, TQ_NOSLEEP);
+ }
+ mutex_exit(&zilog->zl_lock);
+}
+
+static void
+zil_commit_writer(zilog_t *zilog, uint64_t seq, uint64_t foid)
+{
+ uint64_t txg;
+ uint64_t commit_seq = 0;
+ itx_t *itx, *itx_next = (itx_t *)-1;
+ lwb_t *lwb;
+ spa_t *spa;
+
+ zilog->zl_writer = B_TRUE;
+ ASSERT(zilog->zl_root_zio == NULL);
+ spa = zilog->zl_spa;
+
+ if (zilog->zl_suspend) {
+ lwb = NULL;
+ } else {
+ lwb = list_tail(&zilog->zl_lwb_list);
+ if (lwb == NULL) {
+ /*
+ * Return if there's nothing to flush before we
+ * dirty the fs by calling zil_create()
+ */
+ if (list_is_empty(&zilog->zl_itx_list)) {
+ zilog->zl_writer = B_FALSE;
+ return;
+ }
+ mutex_exit(&zilog->zl_lock);
+ zil_create(zilog);
+ mutex_enter(&zilog->zl_lock);
+ lwb = list_tail(&zilog->zl_lwb_list);
+ }
+ }
+
+ /* Loop through in-memory log transactions filling log blocks. */
+ DTRACE_PROBE1(zil__cw1, zilog_t *, zilog);
+ for (;;) {
+ /*
+ * Find the next itx to push:
+ * Push all transactions related to specified foid and all
+ * other transactions except TX_WRITE, TX_TRUNCATE,
+ * TX_SETATTR and TX_ACL for all other files.
+ */
+ if (itx_next != (itx_t *)-1)
+ itx = itx_next;
+ else
+ itx = list_head(&zilog->zl_itx_list);
+ for (; itx != NULL; itx = list_next(&zilog->zl_itx_list, itx)) {
+ if (foid == 0) /* push all foids? */
+ break;
+ if (itx->itx_sync) /* push all O_[D]SYNC */
+ break;
+ switch (itx->itx_lr.lrc_txtype) {
+ case TX_SETATTR:
+ case TX_WRITE:
+ case TX_TRUNCATE:
+ case TX_ACL:
+ /* lr_foid is same offset for these records */
+ if (((lr_write_t *)&itx->itx_lr)->lr_foid
+ != foid) {
+ continue; /* skip this record */
+ }
+ }
+ break;
+ }
+ if (itx == NULL)
+ break;
+
+ if ((itx->itx_lr.lrc_seq > seq) &&
+ ((lwb == NULL) || (lwb->lwb_nused == 0) ||
+ (lwb->lwb_nused + itx->itx_sod > ZIL_BLK_DATA_SZ(lwb)))) {
+ break;
+ }
+
+ /*
+ * Save the next pointer. Even though we soon drop
+ * zl_lock all threads that may change the list
+ * (another writer or zil_itx_clean) can't do so until
+ * they have zl_writer.
+ */
+ itx_next = list_next(&zilog->zl_itx_list, itx);
+ list_remove(&zilog->zl_itx_list, itx);
+ zilog->zl_itx_list_sz -= itx->itx_sod;
+ mutex_exit(&zilog->zl_lock);
+ txg = itx->itx_lr.lrc_txg;
+ ASSERT(txg);
+
+ if (txg > spa_last_synced_txg(spa) ||
+ txg > spa_freeze_txg(spa))
+ lwb = zil_lwb_commit(zilog, itx, lwb);
+ kmem_free(itx, offsetof(itx_t, itx_lr)
+ + itx->itx_lr.lrc_reclen);
+ mutex_enter(&zilog->zl_lock);
+ }
+ DTRACE_PROBE1(zil__cw2, zilog_t *, zilog);
+ /* determine commit sequence number */
+ itx = list_head(&zilog->zl_itx_list);
+ if (itx)
+ commit_seq = itx->itx_lr.lrc_seq;
+ else
+ commit_seq = zilog->zl_itx_seq;
+ mutex_exit(&zilog->zl_lock);
+
+ /* write the last block out */
+ if (lwb != NULL && lwb->lwb_zio != NULL)
+ lwb = zil_lwb_write_start(zilog, lwb);
+
+ zilog->zl_prev_used = zilog->zl_cur_used;
+ zilog->zl_cur_used = 0;
+
+ /*
+ * Wait if necessary for the log blocks to be on stable storage.
+ */
+ if (zilog->zl_root_zio) {
+ DTRACE_PROBE1(zil__cw3, zilog_t *, zilog);
+ (void) zio_wait(zilog->zl_root_zio);
+ zilog->zl_root_zio = NULL;
+ DTRACE_PROBE1(zil__cw4, zilog_t *, zilog);
+ zil_flush_vdevs(zilog);
+ }
+
+ if (zilog->zl_log_error || lwb == NULL) {
+ zilog->zl_log_error = 0;
+ txg_wait_synced(zilog->zl_dmu_pool, 0);
+ }
+
+ mutex_enter(&zilog->zl_lock);
+ zilog->zl_writer = B_FALSE;
+
+ ASSERT3U(commit_seq, >=, zilog->zl_commit_seq);
+ zilog->zl_commit_seq = commit_seq;
+}
+
+/*
+ * Push zfs transactions to stable storage up to the supplied sequence number.
+ * If foid is 0 push out all transactions, otherwise push only those
+ * for that file or might have been used to create that file.
+ */
+void
+zil_commit(zilog_t *zilog, uint64_t seq, uint64_t foid)
+{
+ if (zilog == NULL || seq == 0)
+ return;
+
+ mutex_enter(&zilog->zl_lock);
+
+ seq = MIN(seq, zilog->zl_itx_seq); /* cap seq at largest itx seq */
+
+ while (zilog->zl_writer) {
+ cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock);
+ if (seq < zilog->zl_commit_seq) {
+ mutex_exit(&zilog->zl_lock);
+ return;
+ }
+ }
+ zil_commit_writer(zilog, seq, foid); /* drops zl_lock */
+ /* wake up others waiting on the commit */
+ cv_broadcast(&zilog->zl_cv_writer);
+ mutex_exit(&zilog->zl_lock);
+}
+
+/*
+ * Called in syncing context to free committed log blocks and update log header.
+ */
+void
+zil_sync(zilog_t *zilog, dmu_tx_t *tx)
+{
+ zil_header_t *zh = zil_header_in_syncing_context(zilog);
+ uint64_t txg = dmu_tx_get_txg(tx);
+ spa_t *spa = zilog->zl_spa;
+ lwb_t *lwb;
+
+ mutex_enter(&zilog->zl_lock);
+
+ ASSERT(zilog->zl_stop_sync == 0);
+
+ zh->zh_replay_seq = zilog->zl_replay_seq[txg & TXG_MASK];
+
+ if (zilog->zl_destroy_txg == txg) {
+ blkptr_t blk = zh->zh_log;
+
+ ASSERT(list_head(&zilog->zl_lwb_list) == NULL);
+ ASSERT(spa_sync_pass(spa) == 1);
+
+ bzero(zh, sizeof (zil_header_t));
+ bzero(zilog->zl_replay_seq, sizeof (zilog->zl_replay_seq));
+
+ if (zilog->zl_keep_first) {
+ /*
+ * If this block was part of log chain that couldn't
+ * be claimed because a device was missing during
+ * zil_claim(), but that device later returns,
+ * then this block could erroneously appear valid.
+ * To guard against this, assign a new GUID to the new
+ * log chain so it doesn't matter what blk points to.
+ */
+ zil_init_log_chain(zilog, &blk);
+ zh->zh_log = blk;
+ }
+ }
+
+ for (;;) {
+ lwb = list_head(&zilog->zl_lwb_list);
+ if (lwb == NULL) {
+ mutex_exit(&zilog->zl_lock);
+ return;
+ }
+ zh->zh_log = lwb->lwb_blk;
+ if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg)
+ break;
+ list_remove(&zilog->zl_lwb_list, lwb);
+ zio_free_blk(spa, &lwb->lwb_blk, txg);
+ kmem_cache_free(zil_lwb_cache, lwb);
+
+ /*
+ * If we don't have anything left in the lwb list then
+ * we've had an allocation failure and we need to zero
+ * out the zil_header blkptr so that we don't end
+ * up freeing the same block twice.
+ */
+ if (list_head(&zilog->zl_lwb_list) == NULL)
+ BP_ZERO(&zh->zh_log);
+ }
+ mutex_exit(&zilog->zl_lock);
+}
+
+void
+zil_init(void)
+{
+ zil_lwb_cache = kmem_cache_create("zil_lwb_cache",
+ sizeof (struct lwb), 0, NULL, NULL, NULL, NULL, NULL, 0);
+}
+
+void
+zil_fini(void)
+{
+ kmem_cache_destroy(zil_lwb_cache);
+}
+
+zilog_t *
+zil_alloc(objset_t *os, zil_header_t *zh_phys)
+{
+ zilog_t *zilog;
+
+ zilog = kmem_zalloc(sizeof (zilog_t), KM_SLEEP);
+
+ zilog->zl_header = zh_phys;
+ zilog->zl_os = os;
+ zilog->zl_spa = dmu_objset_spa(os);
+ zilog->zl_dmu_pool = dmu_objset_pool(os);
+ zilog->zl_destroy_txg = TXG_INITIAL - 1;
+
+ mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ list_create(&zilog->zl_itx_list, sizeof (itx_t),
+ offsetof(itx_t, itx_node));
+
+ list_create(&zilog->zl_lwb_list, sizeof (lwb_t),
+ offsetof(lwb_t, lwb_node));
+
+ mutex_init(&zilog->zl_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
+
+ avl_create(&zilog->zl_vdev_tree, zil_vdev_compare,
+ sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node));
+
+ cv_init(&zilog->zl_cv_writer, NULL, CV_DEFAULT, NULL);
+ cv_init(&zilog->zl_cv_suspend, NULL, CV_DEFAULT, NULL);
+
+ return (zilog);
+}
+
+void
+zil_free(zilog_t *zilog)
+{
+ lwb_t *lwb;
+
+ zilog->zl_stop_sync = 1;
+
+ while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
+ list_remove(&zilog->zl_lwb_list, lwb);
+ if (lwb->lwb_buf != NULL)
+ zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
+ kmem_cache_free(zil_lwb_cache, lwb);
+ }
+ list_destroy(&zilog->zl_lwb_list);
+
+ avl_destroy(&zilog->zl_vdev_tree);
+ mutex_destroy(&zilog->zl_vdev_lock);
+
+ ASSERT(list_head(&zilog->zl_itx_list) == NULL);
+ list_destroy(&zilog->zl_itx_list);
+ mutex_destroy(&zilog->zl_lock);
+
+ cv_destroy(&zilog->zl_cv_writer);
+ cv_destroy(&zilog->zl_cv_suspend);
+
+ kmem_free(zilog, sizeof (zilog_t));
+}
+
+/*
+ * return true if the initial log block is not valid
+ */
+static boolean_t
+zil_empty(zilog_t *zilog)
+{
+ const zil_header_t *zh = zilog->zl_header;
+ arc_buf_t *abuf = NULL;
+
+ if (BP_IS_HOLE(&zh->zh_log))
+ return (B_TRUE);
+
+ if (zil_read_log_block(zilog, &zh->zh_log, &abuf) != 0)
+ return (B_TRUE);
+
+ VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
+ return (B_FALSE);
+}
+
+/*
+ * Open an intent log.
+ */
+zilog_t *
+zil_open(objset_t *os, zil_get_data_t *get_data)
+{
+ zilog_t *zilog = dmu_objset_zil(os);
+
+ zilog->zl_get_data = get_data;
+ zilog->zl_clean_taskq = taskq_create("zil_clean", 1, minclsyspri,
+ 2, 2, TASKQ_PREPOPULATE);
+
+ return (zilog);
+}
+
+/*
+ * Close an intent log.
+ */
+void
+zil_close(zilog_t *zilog)
+{
+ /*
+ * If the log isn't already committed, mark the objset dirty
+ * (so zil_sync() will be called) and wait for that txg to sync.
+ */
+ if (!zil_is_committed(zilog)) {
+ uint64_t txg;
+ dmu_tx_t *tx = dmu_tx_create(zilog->zl_os);
+ (void) dmu_tx_assign(tx, TXG_WAIT);
+ dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
+ txg = dmu_tx_get_txg(tx);
+ dmu_tx_commit(tx);
+ txg_wait_synced(zilog->zl_dmu_pool, txg);
+ }
+
+ taskq_destroy(zilog->zl_clean_taskq);
+ zilog->zl_clean_taskq = NULL;
+ zilog->zl_get_data = NULL;
+
+ zil_itx_clean(zilog);
+ ASSERT(list_head(&zilog->zl_itx_list) == NULL);
+}
+
+/*
+ * Suspend an intent log. While in suspended mode, we still honor
+ * synchronous semantics, but we rely on txg_wait_synced() to do it.
+ * We suspend the log briefly when taking a snapshot so that the snapshot
+ * contains all the data it's supposed to, and has an empty intent log.
+ */
+int
+zil_suspend(zilog_t *zilog)
+{
+ const zil_header_t *zh = zilog->zl_header;
+
+ mutex_enter(&zilog->zl_lock);
+ if (zh->zh_claim_txg != 0) { /* unplayed log */
+ mutex_exit(&zilog->zl_lock);
+ return (EBUSY);
+ }
+ if (zilog->zl_suspend++ != 0) {
+ /*
+ * Someone else already began a suspend.
+ * Just wait for them to finish.
+ */
+ while (zilog->zl_suspending)
+ cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock);
+ mutex_exit(&zilog->zl_lock);
+ return (0);
+ }
+ zilog->zl_suspending = B_TRUE;
+ mutex_exit(&zilog->zl_lock);
+
+ zil_commit(zilog, UINT64_MAX, 0);
+
+ /*
+ * Wait for any in-flight log writes to complete.
+ */
+ mutex_enter(&zilog->zl_lock);
+ while (zilog->zl_writer)
+ cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock);
+ mutex_exit(&zilog->zl_lock);
+
+ zil_destroy(zilog, B_FALSE);
+
+ mutex_enter(&zilog->zl_lock);
+ zilog->zl_suspending = B_FALSE;
+ cv_broadcast(&zilog->zl_cv_suspend);
+ mutex_exit(&zilog->zl_lock);
+
+ return (0);
+}
+
+void
+zil_resume(zilog_t *zilog)
+{
+ mutex_enter(&zilog->zl_lock);
+ ASSERT(zilog->zl_suspend != 0);
+ zilog->zl_suspend--;
+ mutex_exit(&zilog->zl_lock);
+}
+
+typedef struct zil_replay_arg {
+ objset_t *zr_os;
+ zil_replay_func_t **zr_replay;
+ zil_replay_cleaner_t *zr_replay_cleaner;
+ void *zr_arg;
+ uint64_t *zr_txgp;
+ boolean_t zr_byteswap;
+ char *zr_lrbuf;
+} zil_replay_arg_t;
+
+static void
+zil_replay_log_record(zilog_t *zilog, lr_t *lr, void *zra, uint64_t claim_txg)
+{
+ zil_replay_arg_t *zr = zra;
+ const zil_header_t *zh = zilog->zl_header;
+ uint64_t reclen = lr->lrc_reclen;
+ uint64_t txtype = lr->lrc_txtype;
+ char *name;
+ int pass, error, sunk;
+
+ if (zilog->zl_stop_replay)
+ return;
+
+ if (lr->lrc_txg < claim_txg) /* already committed */
+ return;
+
+ if (lr->lrc_seq <= zh->zh_replay_seq) /* already replayed */
+ return;
+
+ /* Strip case-insensitive bit, still present in log record */
+ txtype &= ~TX_CI;
+
+ /*
+ * Make a copy of the data so we can revise and extend it.
+ */
+ bcopy(lr, zr->zr_lrbuf, reclen);
+
+ /*
+ * The log block containing this lr may have been byteswapped
+ * so that we can easily examine common fields like lrc_txtype.
+ * However, the log is a mix of different data types, and only the
+ * replay vectors know how to byteswap their records. Therefore, if
+ * the lr was byteswapped, undo it before invoking the replay vector.
+ */
+ if (zr->zr_byteswap)
+ byteswap_uint64_array(zr->zr_lrbuf, reclen);
+
+ /*
+ * If this is a TX_WRITE with a blkptr, suck in the data.
+ */
+ if (txtype == TX_WRITE && reclen == sizeof (lr_write_t)) {
+ lr_write_t *lrw = (lr_write_t *)lr;
+ blkptr_t *wbp = &lrw->lr_blkptr;
+ uint64_t wlen = lrw->lr_length;
+ char *wbuf = zr->zr_lrbuf + reclen;
+
+ if (BP_IS_HOLE(wbp)) { /* compressed to a hole */
+ bzero(wbuf, wlen);
+ } else {
+ /*
+ * A subsequent write may have overwritten this block,
+ * in which case wbp may have been been freed and
+ * reallocated, and our read of wbp may fail with a
+ * checksum error. We can safely ignore this because
+ * the later write will provide the correct data.
+ */
+ zbookmark_t zb;
+
+ zb.zb_objset = dmu_objset_id(zilog->zl_os);
+ zb.zb_object = lrw->lr_foid;
+ zb.zb_level = -1;
+ zb.zb_blkid = lrw->lr_offset / BP_GET_LSIZE(wbp);
+
+ (void) zio_wait(zio_read(NULL, zilog->zl_spa,
+ wbp, wbuf, BP_GET_LSIZE(wbp), NULL, NULL,
+ ZIO_PRIORITY_SYNC_READ,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &zb));
+ (void) memmove(wbuf, wbuf + lrw->lr_blkoff, wlen);
+ }
+ }
+
+ /*
+ * Replay of large truncates can end up needing additional txs
+ * and a different txg. If they are nested within the replay tx
+ * as below then a hang is possible. So we do the truncate here
+ * and redo the truncate later (a no-op) and update the sequence
+ * number whilst in the replay tx. Fortunately, it's safe to repeat
+ * a truncate if we crash and the truncate commits. A create over
+ * an existing file will also come in as a TX_TRUNCATE record.
+ *
+ * Note, remove of large files and renames over large files is
+ * handled by putting the deleted object on a stable list
+ * and if necessary force deleting the object outside of the replay
+ * transaction using the zr_replay_cleaner.
+ */
+ if (txtype == TX_TRUNCATE) {
+ *zr->zr_txgp = TXG_NOWAIT;
+ error = zr->zr_replay[TX_TRUNCATE](zr->zr_arg, zr->zr_lrbuf,
+ zr->zr_byteswap);
+ if (error)
+ goto bad;
+ zr->zr_byteswap = 0; /* only byteswap once */
+ }
+
+ /*
+ * We must now do two things atomically: replay this log record,
+ * and update the log header to reflect the fact that we did so.
+ * We use the DMU's ability to assign into a specific txg to do this.
+ */
+ for (pass = 1, sunk = B_FALSE; /* CONSTANTCONDITION */; pass++) {
+ uint64_t replay_txg;
+ dmu_tx_t *replay_tx;
+
+ replay_tx = dmu_tx_create(zr->zr_os);
+ error = dmu_tx_assign(replay_tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(replay_tx);
+ break;
+ }
+
+ replay_txg = dmu_tx_get_txg(replay_tx);
+
+ if (txtype == 0 || txtype >= TX_MAX_TYPE) {
+ error = EINVAL;
+ } else {
+ /*
+ * On the first pass, arrange for the replay vector
+ * to fail its dmu_tx_assign(). That's the only way
+ * to ensure that those code paths remain well tested.
+ *
+ * Only byteswap (if needed) on the 1st pass.
+ */
+ *zr->zr_txgp = replay_txg - (pass == 1);
+ error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lrbuf,
+ zr->zr_byteswap && pass == 1);
+ *zr->zr_txgp = TXG_NOWAIT;
+ }
+
+ if (error == 0) {
+ dsl_dataset_dirty(dmu_objset_ds(zr->zr_os), replay_tx);
+ zilog->zl_replay_seq[replay_txg & TXG_MASK] =
+ lr->lrc_seq;
+ }
+
+ dmu_tx_commit(replay_tx);
+
+ if (!error)
+ return;
+
+ /*
+ * The DMU's dnode layer doesn't see removes until the txg
+ * commits, so a subsequent claim can spuriously fail with
+ * EEXIST. So if we receive any error other than ERESTART
+ * we try syncing out any removes then retrying the
+ * transaction.
+ */
+ if (error != ERESTART && !sunk) {
+ if (zr->zr_replay_cleaner)
+ zr->zr_replay_cleaner(zr->zr_arg);
+ txg_wait_synced(spa_get_dsl(zilog->zl_spa), 0);
+ sunk = B_TRUE;
+ continue; /* retry */
+ }
+
+ if (error != ERESTART)
+ break;
+
+ if (pass != 1)
+ txg_wait_open(spa_get_dsl(zilog->zl_spa),
+ replay_txg + 1);
+
+ dprintf("pass %d, retrying\n", pass);
+ }
+
+bad:
+ ASSERT(error && error != ERESTART);
+ name = kmem_alloc(MAXNAMELEN, KM_SLEEP);
+ dmu_objset_name(zr->zr_os, name);
+ cmn_err(CE_WARN, "ZFS replay transaction error %d, "
+ "dataset %s, seq 0x%llx, txtype %llu %s\n",
+ error, name, (u_longlong_t)lr->lrc_seq, (u_longlong_t)txtype,
+ (lr->lrc_txtype & TX_CI) ? "CI" : "");
+ zilog->zl_stop_replay = 1;
+ kmem_free(name, MAXNAMELEN);
+}
+
+/* ARGSUSED */
+static void
+zil_incr_blks(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
+{
+ zilog->zl_replay_blks++;
+}
+
+/*
+ * If this dataset has a non-empty intent log, replay it and destroy it.
+ */
+void
+zil_replay(objset_t *os, void *arg, uint64_t *txgp,
+ zil_replay_func_t *replay_func[TX_MAX_TYPE],
+ zil_replay_cleaner_t *replay_cleaner)
+{
+ zilog_t *zilog = dmu_objset_zil(os);
+ const zil_header_t *zh = zilog->zl_header;
+ zil_replay_arg_t zr;
+
+ if (zil_empty(zilog)) {
+ zil_destroy(zilog, B_TRUE);
+ return;
+ }
+
+ zr.zr_os = os;
+ zr.zr_replay = replay_func;
+ zr.zr_replay_cleaner = replay_cleaner;
+ zr.zr_arg = arg;
+ zr.zr_txgp = txgp;
+ zr.zr_byteswap = BP_SHOULD_BYTESWAP(&zh->zh_log);
+ zr.zr_lrbuf = kmem_alloc(2 * SPA_MAXBLOCKSIZE, KM_SLEEP);
+
+ /*
+ * Wait for in-progress removes to sync before starting replay.
+ */
+ txg_wait_synced(zilog->zl_dmu_pool, 0);
+
+ zilog->zl_stop_replay = 0;
+ zilog->zl_replay_time = lbolt;
+ ASSERT(zilog->zl_replay_blks == 0);
+ (void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr,
+ zh->zh_claim_txg);
+ kmem_free(zr.zr_lrbuf, 2 * SPA_MAXBLOCKSIZE);
+
+ zil_destroy(zilog, B_FALSE);
+ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
+}
+
+/*
+ * Report whether all transactions are committed
+ */
+int
+zil_is_committed(zilog_t *zilog)
+{
+ lwb_t *lwb;
+ int ret;
+
+ mutex_enter(&zilog->zl_lock);
+ while (zilog->zl_writer)
+ cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock);
+
+ /* recent unpushed intent log transactions? */
+ if (!list_is_empty(&zilog->zl_itx_list)) {
+ ret = B_FALSE;
+ goto out;
+ }
+
+ /* intent log never used? */
+ lwb = list_head(&zilog->zl_lwb_list);
+ if (lwb == NULL) {
+ ret = B_TRUE;
+ goto out;
+ }
+
+ /*
+ * more than 1 log buffer means zil_sync() hasn't yet freed
+ * entries after a txg has committed
+ */
+ if (list_next(&zilog->zl_lwb_list, lwb)) {
+ ret = B_FALSE;
+ goto out;
+ }
+
+ ASSERT(zil_empty(zilog));
+ ret = B_TRUE;
+out:
+ cv_broadcast(&zilog->zl_cv_writer);
+ mutex_exit(&zilog->zl_lock);
+ return (ret);
+}
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
new file mode 100644
index 000000000..d347920ea
--- /dev/null
+++ b/module/zfs/zio.c
@@ -0,0 +1,2273 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/fm/fs/zfs.h>
+#include <sys/spa.h>
+#include <sys/txg.h>
+#include <sys/spa_impl.h>
+#include <sys/vdev_impl.h>
+#include <sys/zio_impl.h>
+#include <sys/zio_compress.h>
+#include <sys/zio_checksum.h>
+
+/*
+ * ==========================================================================
+ * I/O priority table
+ * ==========================================================================
+ */
+uint8_t zio_priority_table[ZIO_PRIORITY_TABLE_SIZE] = {
+ 0, /* ZIO_PRIORITY_NOW */
+ 0, /* ZIO_PRIORITY_SYNC_READ */
+ 0, /* ZIO_PRIORITY_SYNC_WRITE */
+ 6, /* ZIO_PRIORITY_ASYNC_READ */
+ 4, /* ZIO_PRIORITY_ASYNC_WRITE */
+ 4, /* ZIO_PRIORITY_FREE */
+ 0, /* ZIO_PRIORITY_CACHE_FILL */
+ 0, /* ZIO_PRIORITY_LOG_WRITE */
+ 10, /* ZIO_PRIORITY_RESILVER */
+ 20, /* ZIO_PRIORITY_SCRUB */
+};
+
+/*
+ * ==========================================================================
+ * I/O type descriptions
+ * ==========================================================================
+ */
+char *zio_type_name[ZIO_TYPES] = {
+ "null", "read", "write", "free", "claim", "ioctl" };
+
+#define SYNC_PASS_DEFERRED_FREE 1 /* defer frees after this pass */
+#define SYNC_PASS_DONT_COMPRESS 4 /* don't compress after this pass */
+#define SYNC_PASS_REWRITE 1 /* rewrite new bps after this pass */
+
+/*
+ * ==========================================================================
+ * I/O kmem caches
+ * ==========================================================================
+ */
+kmem_cache_t *zio_cache;
+kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
+kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
+
+#ifdef _KERNEL
+extern vmem_t *zio_alloc_arena;
+#endif
+
+/*
+ * An allocating zio is one that either currently has the DVA allocate
+ * stage set or will have it later in its lifetime.
+ */
+#define IO_IS_ALLOCATING(zio) \
+ ((zio)->io_orig_pipeline & (1U << ZIO_STAGE_DVA_ALLOCATE))
+
+void
+zio_init(void)
+{
+ size_t c;
+ vmem_t *data_alloc_arena = NULL;
+
+#ifdef _KERNEL
+ data_alloc_arena = zio_alloc_arena;
+#endif
+ zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0,
+ NULL, NULL, NULL, NULL, NULL, 0);
+
+ /*
+ * For small buffers, we want a cache for each multiple of
+ * SPA_MINBLOCKSIZE. For medium-size buffers, we want a cache
+ * for each quarter-power of 2. For large buffers, we want
+ * a cache for each multiple of PAGESIZE.
+ */
+ for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
+ size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
+ size_t p2 = size;
+ size_t align = 0;
+
+ while (p2 & (p2 - 1))
+ p2 &= p2 - 1;
+
+ if (size <= 4 * SPA_MINBLOCKSIZE) {
+ align = SPA_MINBLOCKSIZE;
+ } else if (P2PHASE(size, PAGESIZE) == 0) {
+ align = PAGESIZE;
+ } else if (P2PHASE(size, p2 >> 2) == 0) {
+ align = p2 >> 2;
+ }
+
+ if (align != 0) {
+ char name[36];
+ (void) sprintf(name, "zio_buf_%lu", (ulong_t)size);
+ zio_buf_cache[c] = kmem_cache_create(name, size,
+ align, NULL, NULL, NULL, NULL, NULL, KMC_NODEBUG);
+
+ (void) sprintf(name, "zio_data_buf_%lu", (ulong_t)size);
+ zio_data_buf_cache[c] = kmem_cache_create(name, size,
+ align, NULL, NULL, NULL, NULL, data_alloc_arena,
+ KMC_NODEBUG);
+ }
+ }
+
+ while (--c != 0) {
+ ASSERT(zio_buf_cache[c] != NULL);
+ if (zio_buf_cache[c - 1] == NULL)
+ zio_buf_cache[c - 1] = zio_buf_cache[c];
+
+ ASSERT(zio_data_buf_cache[c] != NULL);
+ if (zio_data_buf_cache[c - 1] == NULL)
+ zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
+ }
+
+ zio_inject_init();
+}
+
+void
+zio_fini(void)
+{
+ size_t c;
+ kmem_cache_t *last_cache = NULL;
+ kmem_cache_t *last_data_cache = NULL;
+
+ for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
+ if (zio_buf_cache[c] != last_cache) {
+ last_cache = zio_buf_cache[c];
+ kmem_cache_destroy(zio_buf_cache[c]);
+ }
+ zio_buf_cache[c] = NULL;
+
+ if (zio_data_buf_cache[c] != last_data_cache) {
+ last_data_cache = zio_data_buf_cache[c];
+ kmem_cache_destroy(zio_data_buf_cache[c]);
+ }
+ zio_data_buf_cache[c] = NULL;
+ }
+
+ kmem_cache_destroy(zio_cache);
+
+ zio_inject_fini();
+}
+
+/*
+ * ==========================================================================
+ * Allocate and free I/O buffers
+ * ==========================================================================
+ */
+
+/*
+ * Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a
+ * crashdump if the kernel panics, so use it judiciously. Obviously, it's
+ * useful to inspect ZFS metadata, but if possible, we should avoid keeping
+ * excess / transient data in-core during a crashdump.
+ */
+void *
+zio_buf_alloc(size_t size)
+{
+ size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+ ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+ return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
+}
+
+/*
+ * Use zio_data_buf_alloc to allocate data. The data will not appear in a
+ * crashdump if the kernel panics. This exists so that we will limit the amount
+ * of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount
+ * of kernel heap dumped to disk when the kernel panics)
+ */
+void *
+zio_data_buf_alloc(size_t size)
+{
+ size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+ ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+ return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
+}
+
+void
+zio_buf_free(void *buf, size_t size)
+{
+ size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+ ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+ kmem_cache_free(zio_buf_cache[c], buf);
+}
+
+void
+zio_data_buf_free(void *buf, size_t size)
+{
+ size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+ ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+ kmem_cache_free(zio_data_buf_cache[c], buf);
+}
+
+/*
+ * ==========================================================================
+ * Push and pop I/O transform buffers
+ * ==========================================================================
+ */
+static void
+zio_push_transform(zio_t *zio, void *data, uint64_t size, uint64_t bufsize,
+ zio_transform_func_t *transform)
+{
+ zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP);
+
+ zt->zt_orig_data = zio->io_data;
+ zt->zt_orig_size = zio->io_size;
+ zt->zt_bufsize = bufsize;
+ zt->zt_transform = transform;
+
+ zt->zt_next = zio->io_transform_stack;
+ zio->io_transform_stack = zt;
+
+ zio->io_data = data;
+ zio->io_size = size;
+}
+
+static void
+zio_pop_transforms(zio_t *zio)
+{
+ zio_transform_t *zt;
+
+ while ((zt = zio->io_transform_stack) != NULL) {
+ if (zt->zt_transform != NULL)
+ zt->zt_transform(zio,
+ zt->zt_orig_data, zt->zt_orig_size);
+
+ zio_buf_free(zio->io_data, zt->zt_bufsize);
+
+ zio->io_data = zt->zt_orig_data;
+ zio->io_size = zt->zt_orig_size;
+ zio->io_transform_stack = zt->zt_next;
+
+ kmem_free(zt, sizeof (zio_transform_t));
+ }
+}
+
+/*
+ * ==========================================================================
+ * I/O transform callbacks for subblocks and decompression
+ * ==========================================================================
+ */
+static void
+zio_subblock(zio_t *zio, void *data, uint64_t size)
+{
+ ASSERT(zio->io_size > size);
+
+ if (zio->io_type == ZIO_TYPE_READ)
+ bcopy(zio->io_data, data, size);
+}
+
+static void
+zio_decompress(zio_t *zio, void *data, uint64_t size)
+{
+ if (zio->io_error == 0 &&
+ zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
+ zio->io_data, zio->io_size, data, size) != 0)
+ zio->io_error = EIO;
+}
+
+/*
+ * ==========================================================================
+ * I/O parent/child relationships and pipeline interlocks
+ * ==========================================================================
+ */
+
+static void
+zio_add_child(zio_t *pio, zio_t *zio)
+{
+ mutex_enter(&pio->io_lock);
+ if (zio->io_stage < ZIO_STAGE_READY)
+ pio->io_children[zio->io_child_type][ZIO_WAIT_READY]++;
+ if (zio->io_stage < ZIO_STAGE_DONE)
+ pio->io_children[zio->io_child_type][ZIO_WAIT_DONE]++;
+ zio->io_sibling_prev = NULL;
+ zio->io_sibling_next = pio->io_child;
+ if (pio->io_child != NULL)
+ pio->io_child->io_sibling_prev = zio;
+ pio->io_child = zio;
+ zio->io_parent = pio;
+ mutex_exit(&pio->io_lock);
+}
+
+static void
+zio_remove_child(zio_t *pio, zio_t *zio)
+{
+ zio_t *next, *prev;
+
+ ASSERT(zio->io_parent == pio);
+
+ mutex_enter(&pio->io_lock);
+ next = zio->io_sibling_next;
+ prev = zio->io_sibling_prev;
+ if (next != NULL)
+ next->io_sibling_prev = prev;
+ if (prev != NULL)
+ prev->io_sibling_next = next;
+ if (pio->io_child == zio)
+ pio->io_child = next;
+ mutex_exit(&pio->io_lock);
+}
+
+static boolean_t
+zio_wait_for_children(zio_t *zio, enum zio_child child, enum zio_wait_type wait)
+{
+ uint64_t *countp = &zio->io_children[child][wait];
+ boolean_t waiting = B_FALSE;
+
+ mutex_enter(&zio->io_lock);
+ ASSERT(zio->io_stall == NULL);
+ if (*countp != 0) {
+ zio->io_stage--;
+ zio->io_stall = countp;
+ waiting = B_TRUE;
+ }
+ mutex_exit(&zio->io_lock);
+
+ return (waiting);
+}
+
+static void
+zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait)
+{
+ uint64_t *countp = &pio->io_children[zio->io_child_type][wait];
+ int *errorp = &pio->io_child_error[zio->io_child_type];
+
+ mutex_enter(&pio->io_lock);
+ if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
+ *errorp = zio_worst_error(*errorp, zio->io_error);
+ pio->io_reexecute |= zio->io_reexecute;
+ ASSERT3U(*countp, >, 0);
+ if (--*countp == 0 && pio->io_stall == countp) {
+ pio->io_stall = NULL;
+ mutex_exit(&pio->io_lock);
+ zio_execute(pio);
+ } else {
+ mutex_exit(&pio->io_lock);
+ }
+}
+
+static void
+zio_inherit_child_errors(zio_t *zio, enum zio_child c)
+{
+ if (zio->io_child_error[c] != 0 && zio->io_error == 0)
+ zio->io_error = zio->io_child_error[c];
+}
+
+/*
+ * ==========================================================================
+ * Create the various types of I/O (read, write, free, etc)
+ * ==========================================================================
+ */
+static zio_t *
+zio_create(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ void *data, uint64_t size, zio_done_func_t *done, void *private,
+ zio_type_t type, int priority, int flags, vdev_t *vd, uint64_t offset,
+ const zbookmark_t *zb, uint8_t stage, uint32_t pipeline)
+{
+ zio_t *zio;
+
+ ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
+ ASSERT(P2PHASE(size, SPA_MINBLOCKSIZE) == 0);
+ ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0);
+
+ ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER));
+ ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER));
+ ASSERT(vd || stage == ZIO_STAGE_OPEN);
+
+ zio = kmem_cache_alloc(zio_cache, KM_SLEEP);
+ bzero(zio, sizeof (zio_t));
+
+ mutex_init(&zio->io_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL);
+
+ if (vd != NULL)
+ zio->io_child_type = ZIO_CHILD_VDEV;
+ else if (flags & ZIO_FLAG_GANG_CHILD)
+ zio->io_child_type = ZIO_CHILD_GANG;
+ else
+ zio->io_child_type = ZIO_CHILD_LOGICAL;
+
+ if (bp != NULL) {
+ zio->io_bp = bp;
+ zio->io_bp_copy = *bp;
+ zio->io_bp_orig = *bp;
+ if (type != ZIO_TYPE_WRITE)
+ zio->io_bp = &zio->io_bp_copy; /* so caller can free */
+ if (zio->io_child_type == ZIO_CHILD_LOGICAL) {
+ if (BP_IS_GANG(bp))
+ pipeline |= ZIO_GANG_STAGES;
+ zio->io_logical = zio;
+ }
+ }
+
+ zio->io_spa = spa;
+ zio->io_txg = txg;
+ zio->io_data = data;
+ zio->io_size = size;
+ zio->io_done = done;
+ zio->io_private = private;
+ zio->io_type = type;
+ zio->io_priority = priority;
+ zio->io_vd = vd;
+ zio->io_offset = offset;
+ zio->io_orig_flags = zio->io_flags = flags;
+ zio->io_orig_stage = zio->io_stage = stage;
+ zio->io_orig_pipeline = zio->io_pipeline = pipeline;
+
+ if (zb != NULL)
+ zio->io_bookmark = *zb;
+
+ if (pio != NULL) {
+ /*
+ * Logical I/Os can have logical, gang, or vdev children.
+ * Gang I/Os can have gang or vdev children.
+ * Vdev I/Os can only have vdev children.
+ * The following ASSERT captures all of these constraints.
+ */
+ ASSERT(zio->io_child_type <= pio->io_child_type);
+ if (zio->io_logical == NULL)
+ zio->io_logical = pio->io_logical;
+ zio_add_child(pio, zio);
+ }
+
+ return (zio);
+}
+
+static void
+zio_destroy(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+ uint8_t async_root = zio->io_async_root;
+
+ mutex_destroy(&zio->io_lock);
+ cv_destroy(&zio->io_cv);
+ kmem_cache_free(zio_cache, zio);
+
+ if (async_root) {
+ mutex_enter(&spa->spa_async_root_lock);
+ if (--spa->spa_async_root_count == 0)
+ cv_broadcast(&spa->spa_async_root_cv);
+ mutex_exit(&spa->spa_async_root_lock);
+ }
+}
+
+zio_t *
+zio_null(zio_t *pio, spa_t *spa, zio_done_func_t *done, void *private,
+ int flags)
+{
+ zio_t *zio;
+
+ zio = zio_create(pio, spa, 0, NULL, NULL, 0, done, private,
+ ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL,
+ ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE);
+
+ return (zio);
+}
+
+zio_t *
+zio_root(spa_t *spa, zio_done_func_t *done, void *private, int flags)
+{
+ return (zio_null(NULL, spa, done, private, flags));
+}
+
+zio_t *
+zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
+ void *data, uint64_t size, zio_done_func_t *done, void *private,
+ int priority, int flags, const zbookmark_t *zb)
+{
+ zio_t *zio;
+
+ zio = zio_create(pio, spa, bp->blk_birth, (blkptr_t *)bp,
+ data, size, done, private,
+ ZIO_TYPE_READ, priority, flags, NULL, 0, zb,
+ ZIO_STAGE_OPEN, ZIO_READ_PIPELINE);
+
+ return (zio);
+}
+
+void
+zio_skip_write(zio_t *zio)
+{
+ ASSERT(zio->io_type == ZIO_TYPE_WRITE);
+ ASSERT(zio->io_stage == ZIO_STAGE_READY);
+ ASSERT(!BP_IS_GANG(zio->io_bp));
+
+ zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
+}
+
+zio_t *
+zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ void *data, uint64_t size, zio_prop_t *zp,
+ zio_done_func_t *ready, zio_done_func_t *done, void *private,
+ int priority, int flags, const zbookmark_t *zb)
+{
+ zio_t *zio;
+
+ ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF &&
+ zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
+ zp->zp_compress >= ZIO_COMPRESS_OFF &&
+ zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
+ zp->zp_type < DMU_OT_NUMTYPES &&
+ zp->zp_level < 32 &&
+ zp->zp_ndvas > 0 &&
+ zp->zp_ndvas <= spa_max_replication(spa));
+ ASSERT(ready != NULL);
+
+ zio = zio_create(pio, spa, txg, bp, data, size, done, private,
+ ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
+ ZIO_STAGE_OPEN, ZIO_WRITE_PIPELINE);
+
+ zio->io_ready = ready;
+ zio->io_prop = *zp;
+
+ return (zio);
+}
+
+zio_t *
+zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data,
+ uint64_t size, zio_done_func_t *done, void *private, int priority,
+ int flags, zbookmark_t *zb)
+{
+ zio_t *zio;
+
+ zio = zio_create(pio, spa, txg, bp, data, size, done, private,
+ ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
+ ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE);
+
+ return (zio);
+}
+
+zio_t *
+zio_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, int flags)
+{
+ zio_t *zio;
+
+ ASSERT(!BP_IS_HOLE(bp));
+
+ if (bp->blk_fill == BLK_FILL_ALREADY_FREED)
+ return (zio_null(pio, spa, NULL, NULL, flags));
+
+ if (txg == spa->spa_syncing_txg &&
+ spa_sync_pass(spa) > SYNC_PASS_DEFERRED_FREE) {
+ bplist_enqueue_deferred(&spa->spa_sync_bplist, bp);
+ return (zio_null(pio, spa, NULL, NULL, flags));
+ }
+
+ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
+ done, private, ZIO_TYPE_FREE, ZIO_PRIORITY_FREE, flags,
+ NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_FREE_PIPELINE);
+
+ return (zio);
+}
+
+zio_t *
+zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
+ zio_done_func_t *done, void *private, int flags)
+{
+ zio_t *zio;
+
+ /*
+ * A claim is an allocation of a specific block. Claims are needed
+ * to support immediate writes in the intent log. The issue is that
+ * immediate writes contain committed data, but in a txg that was
+ * *not* committed. Upon opening the pool after an unclean shutdown,
+ * the intent log claims all blocks that contain immediate write data
+ * so that the SPA knows they're in use.
+ *
+ * All claims *must* be resolved in the first txg -- before the SPA
+ * starts allocating blocks -- so that nothing is allocated twice.
+ */
+ ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <, spa_first_txg(spa));
+ ASSERT3U(spa_first_txg(spa), <=, txg);
+
+ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
+ done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW, flags,
+ NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE);
+
+ return (zio);
+}
+
+zio_t *
+zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
+ zio_done_func_t *done, void *private, int priority, int flags)
+{
+ zio_t *zio;
+ int c;
+
+ if (vd->vdev_children == 0) {
+ zio = zio_create(pio, spa, 0, NULL, NULL, 0, done, private,
+ ZIO_TYPE_IOCTL, priority, flags, vd, 0, NULL,
+ ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE);
+
+ zio->io_cmd = cmd;
+ } else {
+ zio = zio_null(pio, spa, NULL, NULL, flags);
+
+ for (c = 0; c < vd->vdev_children; c++)
+ zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd,
+ done, private, priority, flags));
+ }
+
+ return (zio);
+}
+
+zio_t *
+zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
+ void *data, int checksum, zio_done_func_t *done, void *private,
+ int priority, int flags, boolean_t labels)
+{
+ zio_t *zio;
+
+ ASSERT(vd->vdev_children == 0);
+ ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
+ offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
+ ASSERT3U(offset + size, <=, vd->vdev_psize);
+
+ zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, done, private,
+ ZIO_TYPE_READ, priority, flags, vd, offset, NULL,
+ ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE);
+
+ zio->io_prop.zp_checksum = checksum;
+
+ return (zio);
+}
+
+zio_t *
+zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
+ void *data, int checksum, zio_done_func_t *done, void *private,
+ int priority, int flags, boolean_t labels)
+{
+ zio_t *zio;
+
+ ASSERT(vd->vdev_children == 0);
+ ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
+ offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
+ ASSERT3U(offset + size, <=, vd->vdev_psize);
+
+ zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, done, private,
+ ZIO_TYPE_WRITE, priority, flags, vd, offset, NULL,
+ ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE);
+
+ zio->io_prop.zp_checksum = checksum;
+
+ if (zio_checksum_table[checksum].ci_zbt) {
+ /*
+ * zbt checksums are necessarily destructive -- they modify
+ * the end of the write buffer to hold the verifier/checksum.
+ * Therefore, we must make a local copy in case the data is
+ * being written to multiple places in parallel.
+ */
+ void *wbuf = zio_buf_alloc(size);
+ bcopy(data, wbuf, size);
+ zio_push_transform(zio, wbuf, size, size, NULL);
+ }
+
+ return (zio);
+}
+
+/*
+ * Create a child I/O to do some work for us.
+ */
+zio_t *
+zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset,
+ void *data, uint64_t size, int type, int priority, int flags,
+ zio_done_func_t *done, void *private)
+{
+ uint32_t pipeline = ZIO_VDEV_CHILD_PIPELINE;
+ zio_t *zio;
+
+ ASSERT(vd->vdev_parent ==
+ (pio->io_vd ? pio->io_vd : pio->io_spa->spa_root_vdev));
+
+ if (type == ZIO_TYPE_READ && bp != NULL) {
+ /*
+ * If we have the bp, then the child should perform the
+ * checksum and the parent need not. This pushes error
+ * detection as close to the leaves as possible and
+ * eliminates redundant checksums in the interior nodes.
+ */
+ pipeline |= 1U << ZIO_STAGE_CHECKSUM_VERIFY;
+ pio->io_pipeline &= ~(1U << ZIO_STAGE_CHECKSUM_VERIFY);
+ }
+
+ if (vd->vdev_children == 0)
+ offset += VDEV_LABEL_START_SIZE;
+
+ zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size,
+ done, private, type, priority,
+ (pio->io_flags & ZIO_FLAG_VDEV_INHERIT) |
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | flags,
+ vd, offset, &pio->io_bookmark,
+ ZIO_STAGE_VDEV_IO_START - 1, pipeline);
+
+ return (zio);
+}
+
+zio_t *
+zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, void *data, uint64_t size,
+ int type, int priority, int flags, zio_done_func_t *done, void *private)
+{
+ zio_t *zio;
+
+ ASSERT(vd->vdev_ops->vdev_op_leaf);
+
+ zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
+ data, size, done, private, type, priority,
+ flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY,
+ vd, offset, NULL,
+ ZIO_STAGE_VDEV_IO_START - 1, ZIO_VDEV_CHILD_PIPELINE);
+
+ return (zio);
+}
+
+void
+zio_flush(zio_t *zio, vdev_t *vd)
+{
+ zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE,
+ NULL, NULL, ZIO_PRIORITY_NOW,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY));
+}
+
+/*
+ * ==========================================================================
+ * Prepare to read and write logical blocks
+ * ==========================================================================
+ */
+
+static int
+zio_read_bp_init(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+
+ if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF && zio->io_logical == zio) {
+ uint64_t csize = BP_GET_PSIZE(bp);
+ void *cbuf = zio_buf_alloc(csize);
+
+ zio_push_transform(zio, cbuf, csize, csize, zio_decompress);
+ }
+
+ if (!dmu_ot[BP_GET_TYPE(bp)].ot_metadata && BP_GET_LEVEL(bp) == 0)
+ zio->io_flags |= ZIO_FLAG_DONT_CACHE;
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_write_bp_init(zio_t *zio)
+{
+ zio_prop_t *zp = &zio->io_prop;
+ int compress = zp->zp_compress;
+ blkptr_t *bp = zio->io_bp;
+ void *cbuf;
+ uint64_t lsize = zio->io_size;
+ uint64_t csize = lsize;
+ uint64_t cbufsize = 0;
+ int pass = 1;
+
+ /*
+ * If our children haven't all reached the ready stage,
+ * wait for them and then repeat this pipeline stage.
+ */
+ if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY) ||
+ zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_READY))
+ return (ZIO_PIPELINE_STOP);
+
+ if (!IO_IS_ALLOCATING(zio))
+ return (ZIO_PIPELINE_CONTINUE);
+
+ ASSERT(compress != ZIO_COMPRESS_INHERIT);
+
+ if (bp->blk_birth == zio->io_txg) {
+ /*
+ * We're rewriting an existing block, which means we're
+ * working on behalf of spa_sync(). For spa_sync() to
+ * converge, it must eventually be the case that we don't
+ * have to allocate new blocks. But compression changes
+ * the blocksize, which forces a reallocate, and makes
+ * convergence take longer. Therefore, after the first
+ * few passes, stop compressing to ensure convergence.
+ */
+ pass = spa_sync_pass(zio->io_spa);
+ ASSERT(pass > 1);
+
+ if (pass > SYNC_PASS_DONT_COMPRESS)
+ compress = ZIO_COMPRESS_OFF;
+
+ /*
+ * Only MOS (objset 0) data should need to be rewritten.
+ */
+ ASSERT(zio->io_logical->io_bookmark.zb_objset == 0);
+
+ /* Make sure someone doesn't change their mind on overwrites */
+ ASSERT(MIN(zp->zp_ndvas + BP_IS_GANG(bp),
+ spa_max_replication(zio->io_spa)) == BP_GET_NDVAS(bp));
+ }
+
+ if (compress != ZIO_COMPRESS_OFF) {
+ if (!zio_compress_data(compress, zio->io_data, zio->io_size,
+ &cbuf, &csize, &cbufsize)) {
+ compress = ZIO_COMPRESS_OFF;
+ } else if (csize != 0) {
+ zio_push_transform(zio, cbuf, csize, cbufsize, NULL);
+ }
+ }
+
+ /*
+ * The final pass of spa_sync() must be all rewrites, but the first
+ * few passes offer a trade-off: allocating blocks defers convergence,
+ * but newly allocated blocks are sequential, so they can be written
+ * to disk faster. Therefore, we allow the first few passes of
+ * spa_sync() to allocate new blocks, but force rewrites after that.
+ * There should only be a handful of blocks after pass 1 in any case.
+ */
+ if (bp->blk_birth == zio->io_txg && BP_GET_PSIZE(bp) == csize &&
+ pass > SYNC_PASS_REWRITE) {
+ ASSERT(csize != 0);
+ uint32_t gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;
+ zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages;
+ zio->io_flags |= ZIO_FLAG_IO_REWRITE;
+ } else {
+ BP_ZERO(bp);
+ zio->io_pipeline = ZIO_WRITE_PIPELINE;
+ }
+
+ if (csize == 0) {
+ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+ } else {
+ ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER);
+ BP_SET_LSIZE(bp, lsize);
+ BP_SET_PSIZE(bp, csize);
+ BP_SET_COMPRESS(bp, compress);
+ BP_SET_CHECKSUM(bp, zp->zp_checksum);
+ BP_SET_TYPE(bp, zp->zp_type);
+ BP_SET_LEVEL(bp, zp->zp_level);
+ BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/*
+ * ==========================================================================
+ * Execute the I/O pipeline
+ * ==========================================================================
+ */
+
+static void
+zio_taskq_dispatch(zio_t *zio, enum zio_taskq_type q)
+{
+ zio_type_t t = zio->io_type;
+
+ /*
+ * If we're a config writer, the normal issue and interrupt threads
+ * may all be blocked waiting for the config lock. In this case,
+ * select the otherwise-unused taskq for ZIO_TYPE_NULL.
+ */
+ if (zio->io_flags & ZIO_FLAG_CONFIG_WRITER)
+ t = ZIO_TYPE_NULL;
+
+ /*
+ * A similar issue exists for the L2ARC write thread until L2ARC 2.0.
+ */
+ if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux)
+ t = ZIO_TYPE_NULL;
+
+ (void) taskq_dispatch(zio->io_spa->spa_zio_taskq[t][q],
+ (task_func_t *)zio_execute, zio, TQ_SLEEP);
+}
+
+static boolean_t
+zio_taskq_member(zio_t *zio, enum zio_taskq_type q)
+{
+ kthread_t *executor = zio->io_executor;
+ spa_t *spa = zio->io_spa;
+
+ for (zio_type_t t = 0; t < ZIO_TYPES; t++)
+ if (taskq_member(spa->spa_zio_taskq[t][q], executor))
+ return (B_TRUE);
+
+ return (B_FALSE);
+}
+
+static int
+zio_issue_async(zio_t *zio)
+{
+ zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE);
+
+ return (ZIO_PIPELINE_STOP);
+}
+
+void
+zio_interrupt(zio_t *zio)
+{
+ zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT);
+}
+
+/*
+ * Execute the I/O pipeline until one of the following occurs:
+ * (1) the I/O completes; (2) the pipeline stalls waiting for
+ * dependent child I/Os; (3) the I/O issues, so we're waiting
+ * for an I/O completion interrupt; (4) the I/O is delegated by
+ * vdev-level caching or aggregation; (5) the I/O is deferred
+ * due to vdev-level queueing; (6) the I/O is handed off to
+ * another thread. In all cases, the pipeline stops whenever
+ * there's no CPU work; it never burns a thread in cv_wait().
+ *
+ * There's no locking on io_stage because there's no legitimate way
+ * for multiple threads to be attempting to process the same I/O.
+ */
+static zio_pipe_stage_t *zio_pipeline[ZIO_STAGES];
+
+void
+zio_execute(zio_t *zio)
+{
+ zio->io_executor = curthread;
+
+ while (zio->io_stage < ZIO_STAGE_DONE) {
+ uint32_t pipeline = zio->io_pipeline;
+ zio_stage_t stage = zio->io_stage;
+ int rv;
+
+ ASSERT(!MUTEX_HELD(&zio->io_lock));
+
+ while (((1U << ++stage) & pipeline) == 0)
+ continue;
+
+ ASSERT(stage <= ZIO_STAGE_DONE);
+ ASSERT(zio->io_stall == NULL);
+
+ /*
+ * If we are in interrupt context and this pipeline stage
+ * will grab a config lock that is held across I/O,
+ * issue async to avoid deadlock.
+ */
+ if (((1U << stage) & ZIO_CONFIG_LOCK_BLOCKING_STAGES) &&
+ zio->io_vd == NULL &&
+ zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
+ zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE);
+ return;
+ }
+
+ zio->io_stage = stage;
+ rv = zio_pipeline[stage](zio);
+
+ if (rv == ZIO_PIPELINE_STOP)
+ return;
+
+ ASSERT(rv == ZIO_PIPELINE_CONTINUE);
+ }
+}
+
+/*
+ * ==========================================================================
+ * Initiate I/O, either sync or async
+ * ==========================================================================
+ */
+int
+zio_wait(zio_t *zio)
+{
+ int error;
+
+ ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
+ ASSERT(zio->io_executor == NULL);
+
+ zio->io_waiter = curthread;
+
+ zio_execute(zio);
+
+ mutex_enter(&zio->io_lock);
+ while (zio->io_executor != NULL)
+ cv_wait(&zio->io_cv, &zio->io_lock);
+ mutex_exit(&zio->io_lock);
+
+ error = zio->io_error;
+ zio_destroy(zio);
+
+ return (error);
+}
+
+void
+zio_nowait(zio_t *zio)
+{
+ ASSERT(zio->io_executor == NULL);
+
+ if (zio->io_parent == NULL && zio->io_child_type == ZIO_CHILD_LOGICAL) {
+ /*
+ * This is a logical async I/O with no parent to wait for it.
+ * Attach it to the pool's global async root zio so that
+ * spa_unload() has a way of waiting for async I/O to finish.
+ */
+ spa_t *spa = zio->io_spa;
+ zio->io_async_root = B_TRUE;
+ mutex_enter(&spa->spa_async_root_lock);
+ spa->spa_async_root_count++;
+ mutex_exit(&spa->spa_async_root_lock);
+ }
+
+ zio_execute(zio);
+}
+
+/*
+ * ==========================================================================
+ * Reexecute or suspend/resume failed I/O
+ * ==========================================================================
+ */
+
+static void
+zio_reexecute(zio_t *pio)
+{
+ zio_t *zio, *zio_next;
+
+ pio->io_flags = pio->io_orig_flags;
+ pio->io_stage = pio->io_orig_stage;
+ pio->io_pipeline = pio->io_orig_pipeline;
+ pio->io_reexecute = 0;
+ pio->io_error = 0;
+ for (int c = 0; c < ZIO_CHILD_TYPES; c++)
+ pio->io_child_error[c] = 0;
+
+ if (IO_IS_ALLOCATING(pio)) {
+ /*
+ * Remember the failed bp so that the io_ready() callback
+ * can update its accounting upon reexecution. The block
+ * was already freed in zio_done(); we indicate this with
+ * a fill count of -1 so that zio_free() knows to skip it.
+ */
+ blkptr_t *bp = pio->io_bp;
+ ASSERT(bp->blk_birth == 0 || bp->blk_birth == pio->io_txg);
+ bp->blk_fill = BLK_FILL_ALREADY_FREED;
+ pio->io_bp_orig = *bp;
+ BP_ZERO(bp);
+ }
+
+ /*
+ * As we reexecute pio's children, new children could be created.
+ * New children go to the head of the io_child list, however,
+ * so we will (correctly) not reexecute them. The key is that
+ * the remainder of the io_child list, from 'zio_next' onward,
+ * cannot be affected by any side effects of reexecuting 'zio'.
+ */
+ for (zio = pio->io_child; zio != NULL; zio = zio_next) {
+ zio_next = zio->io_sibling_next;
+ mutex_enter(&pio->io_lock);
+ pio->io_children[zio->io_child_type][ZIO_WAIT_READY]++;
+ pio->io_children[zio->io_child_type][ZIO_WAIT_DONE]++;
+ mutex_exit(&pio->io_lock);
+ zio_reexecute(zio);
+ }
+
+ /*
+ * Now that all children have been reexecuted, execute the parent.
+ */
+ zio_execute(pio);
+}
+
+void
+zio_suspend(spa_t *spa, zio_t *zio)
+{
+ if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC)
+ fm_panic("Pool '%s' has encountered an uncorrectable I/O "
+ "failure and the failure mode property for this pool "
+ "is set to panic.", spa_name(spa));
+
+ zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, 0, 0);
+
+ mutex_enter(&spa->spa_suspend_lock);
+
+ if (spa->spa_suspend_zio_root == NULL)
+ spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL, 0);
+
+ spa->spa_suspended = B_TRUE;
+
+ if (zio != NULL) {
+ ASSERT(zio != spa->spa_suspend_zio_root);
+ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
+ ASSERT(zio->io_parent == NULL);
+ ASSERT(zio->io_stage == ZIO_STAGE_DONE);
+ zio_add_child(spa->spa_suspend_zio_root, zio);
+ }
+
+ mutex_exit(&spa->spa_suspend_lock);
+}
+
+void
+zio_resume(spa_t *spa)
+{
+ zio_t *pio, *zio;
+
+ /*
+ * Reexecute all previously suspended i/o.
+ */
+ mutex_enter(&spa->spa_suspend_lock);
+ spa->spa_suspended = B_FALSE;
+ cv_broadcast(&spa->spa_suspend_cv);
+ pio = spa->spa_suspend_zio_root;
+ spa->spa_suspend_zio_root = NULL;
+ mutex_exit(&spa->spa_suspend_lock);
+
+ if (pio == NULL)
+ return;
+
+ while ((zio = pio->io_child) != NULL) {
+ zio_remove_child(pio, zio);
+ zio->io_parent = NULL;
+ zio_reexecute(zio);
+ }
+
+ ASSERT(pio->io_children[ZIO_CHILD_LOGICAL][ZIO_WAIT_DONE] == 0);
+
+ (void) zio_wait(pio);
+}
+
+void
+zio_resume_wait(spa_t *spa)
+{
+ mutex_enter(&spa->spa_suspend_lock);
+ while (spa_suspended(spa))
+ cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock);
+ mutex_exit(&spa->spa_suspend_lock);
+}
+
+/*
+ * ==========================================================================
+ * Gang blocks.
+ *
+ * A gang block is a collection of small blocks that looks to the DMU
+ * like one large block. When zio_dva_allocate() cannot find a block
+ * of the requested size, due to either severe fragmentation or the pool
+ * being nearly full, it calls zio_write_gang_block() to construct the
+ * block from smaller fragments.
+ *
+ * A gang block consists of a gang header (zio_gbh_phys_t) and up to
+ * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like
+ * an indirect block: it's an array of block pointers. It consumes
+ * only one sector and hence is allocatable regardless of fragmentation.
+ * The gang header's bps point to its gang members, which hold the data.
+ *
+ * Gang blocks are self-checksumming, using the bp's <vdev, offset, txg>
+ * as the verifier to ensure uniqueness of the SHA256 checksum.
+ * Critically, the gang block bp's blk_cksum is the checksum of the data,
+ * not the gang header. This ensures that data block signatures (needed for
+ * deduplication) are independent of how the block is physically stored.
+ *
+ * Gang blocks can be nested: a gang member may itself be a gang block.
+ * Thus every gang block is a tree in which root and all interior nodes are
+ * gang headers, and the leaves are normal blocks that contain user data.
+ * The root of the gang tree is called the gang leader.
+ *
+ * To perform any operation (read, rewrite, free, claim) on a gang block,
+ * zio_gang_assemble() first assembles the gang tree (minus data leaves)
+ * in the io_gang_tree field of the original logical i/o by recursively
+ * reading the gang leader and all gang headers below it. This yields
+ * an in-core tree containing the contents of every gang header and the
+ * bps for every constituent of the gang block.
+ *
+ * With the gang tree now assembled, zio_gang_issue() just walks the gang tree
+ * and invokes a callback on each bp. To free a gang block, zio_gang_issue()
+ * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp.
+ * zio_claim_gang() provides a similarly trivial wrapper for zio_claim().
+ * zio_read_gang() is a wrapper around zio_read() that omits reading gang
+ * headers, since we already have those in io_gang_tree. zio_rewrite_gang()
+ * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite()
+ * of the gang header plus zio_checksum_compute() of the data to update the
+ * gang header's blk_cksum as described above.
+ *
+ * The two-phase assemble/issue model solves the problem of partial failure --
+ * what if you'd freed part of a gang block but then couldn't read the
+ * gang header for another part? Assembling the entire gang tree first
+ * ensures that all the necessary gang header I/O has succeeded before
+ * starting the actual work of free, claim, or write. Once the gang tree
+ * is assembled, free and claim are in-memory operations that cannot fail.
+ *
+ * In the event that a gang write fails, zio_dva_unallocate() walks the
+ * gang tree to immediately free (i.e. insert back into the space map)
+ * everything we've allocated. This ensures that we don't get ENOSPC
+ * errors during repeated suspend/resume cycles due to a flaky device.
+ *
+ * Gang rewrites only happen during sync-to-convergence. If we can't assemble
+ * the gang tree, we won't modify the block, so we can safely defer the free
+ * (knowing that the block is still intact). If we *can* assemble the gang
+ * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free
+ * each constituent bp and we can allocate a new block on the next sync pass.
+ *
+ * In all cases, the gang tree allows complete recovery from partial failure.
+ * ==========================================================================
+ */
+
+static zio_t *
+zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data)
+{
+ if (gn != NULL)
+ return (pio);
+
+ return (zio_read(pio, pio->io_spa, bp, data, BP_GET_PSIZE(bp),
+ NULL, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
+ &pio->io_bookmark));
+}
+
+zio_t *
+zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data)
+{
+ zio_t *zio;
+
+ if (gn != NULL) {
+ zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
+ gn->gn_gbh, SPA_GANGBLOCKSIZE, NULL, NULL, pio->io_priority,
+ ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
+ /*
+ * As we rewrite each gang header, the pipeline will compute
+ * a new gang block header checksum for it; but no one will
+ * compute a new data checksum, so we do that here. The one
+ * exception is the gang leader: the pipeline already computed
+ * its data checksum because that stage precedes gang assembly.
+ * (Presently, nothing actually uses interior data checksums;
+ * this is just good hygiene.)
+ */
+ if (gn != pio->io_logical->io_gang_tree) {
+ zio_checksum_compute(zio, BP_GET_CHECKSUM(bp),
+ data, BP_GET_PSIZE(bp));
+ }
+ } else {
+ zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
+ data, BP_GET_PSIZE(bp), NULL, NULL, pio->io_priority,
+ ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
+ }
+
+ return (zio);
+}
+
+/* ARGSUSED */
+zio_t *
+zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data)
+{
+ return (zio_free(pio, pio->io_spa, pio->io_txg, bp,
+ NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
+}
+
+/* ARGSUSED */
+zio_t *
+zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, void *data)
+{
+ return (zio_claim(pio, pio->io_spa, pio->io_txg, bp,
+ NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
+}
+
+static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = {
+ NULL,
+ zio_read_gang,
+ zio_rewrite_gang,
+ zio_free_gang,
+ zio_claim_gang,
+ NULL
+};
+
+static void zio_gang_tree_assemble_done(zio_t *zio);
+
+static zio_gang_node_t *
+zio_gang_node_alloc(zio_gang_node_t **gnpp)
+{
+ zio_gang_node_t *gn;
+
+ ASSERT(*gnpp == NULL);
+
+ gn = kmem_zalloc(sizeof (*gn), KM_SLEEP);
+ gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE);
+ *gnpp = gn;
+
+ return (gn);
+}
+
+static void
+zio_gang_node_free(zio_gang_node_t **gnpp)
+{
+ zio_gang_node_t *gn = *gnpp;
+
+ for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
+ ASSERT(gn->gn_child[g] == NULL);
+
+ zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE);
+ kmem_free(gn, sizeof (*gn));
+ *gnpp = NULL;
+}
+
+static void
+zio_gang_tree_free(zio_gang_node_t **gnpp)
+{
+ zio_gang_node_t *gn = *gnpp;
+
+ if (gn == NULL)
+ return;
+
+ for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
+ zio_gang_tree_free(&gn->gn_child[g]);
+
+ zio_gang_node_free(gnpp);
+}
+
+static void
+zio_gang_tree_assemble(zio_t *lio, blkptr_t *bp, zio_gang_node_t **gnpp)
+{
+ zio_gang_node_t *gn = zio_gang_node_alloc(gnpp);
+
+ ASSERT(lio->io_logical == lio);
+ ASSERT(BP_IS_GANG(bp));
+
+ zio_nowait(zio_read(lio, lio->io_spa, bp, gn->gn_gbh,
+ SPA_GANGBLOCKSIZE, zio_gang_tree_assemble_done, gn,
+ lio->io_priority, ZIO_GANG_CHILD_FLAGS(lio), &lio->io_bookmark));
+}
+
+static void
+zio_gang_tree_assemble_done(zio_t *zio)
+{
+ zio_t *lio = zio->io_logical;
+ zio_gang_node_t *gn = zio->io_private;
+ blkptr_t *bp = zio->io_bp;
+
+ ASSERT(zio->io_parent == lio);
+ ASSERT(zio->io_child == NULL);
+
+ if (zio->io_error)
+ return;
+
+ if (BP_SHOULD_BYTESWAP(bp))
+ byteswap_uint64_array(zio->io_data, zio->io_size);
+
+ ASSERT(zio->io_data == gn->gn_gbh);
+ ASSERT(zio->io_size == SPA_GANGBLOCKSIZE);
+ ASSERT(gn->gn_gbh->zg_tail.zbt_magic == ZBT_MAGIC);
+
+ for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
+ blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
+ if (!BP_IS_GANG(gbp))
+ continue;
+ zio_gang_tree_assemble(lio, gbp, &gn->gn_child[g]);
+ }
+}
+
+static void
+zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, void *data)
+{
+ zio_t *lio = pio->io_logical;
+ zio_t *zio;
+
+ ASSERT(BP_IS_GANG(bp) == !!gn);
+ ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(lio->io_bp));
+ ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == lio->io_gang_tree);
+
+ /*
+ * If you're a gang header, your data is in gn->gn_gbh.
+ * If you're a gang member, your data is in 'data' and gn == NULL.
+ */
+ zio = zio_gang_issue_func[lio->io_type](pio, bp, gn, data);
+
+ if (gn != NULL) {
+ ASSERT(gn->gn_gbh->zg_tail.zbt_magic == ZBT_MAGIC);
+
+ for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
+ blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
+ if (BP_IS_HOLE(gbp))
+ continue;
+ zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data);
+ data = (char *)data + BP_GET_PSIZE(gbp);
+ }
+ }
+
+ if (gn == lio->io_gang_tree)
+ ASSERT3P((char *)lio->io_data + lio->io_size, ==, data);
+
+ if (zio != pio)
+ zio_nowait(zio);
+}
+
+static int
+zio_gang_assemble(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+
+ ASSERT(BP_IS_GANG(bp) && zio == zio->io_logical);
+
+ zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree);
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_gang_issue(zio_t *zio)
+{
+ zio_t *lio = zio->io_logical;
+ blkptr_t *bp = zio->io_bp;
+
+ if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE))
+ return (ZIO_PIPELINE_STOP);
+
+ ASSERT(BP_IS_GANG(bp) && zio == lio);
+
+ if (zio->io_child_error[ZIO_CHILD_GANG] == 0)
+ zio_gang_tree_issue(lio, lio->io_gang_tree, bp, lio->io_data);
+ else
+ zio_gang_tree_free(&lio->io_gang_tree);
+
+ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static void
+zio_write_gang_member_ready(zio_t *zio)
+{
+ zio_t *pio = zio->io_parent;
+ zio_t *lio = zio->io_logical;
+ dva_t *cdva = zio->io_bp->blk_dva;
+ dva_t *pdva = pio->io_bp->blk_dva;
+ uint64_t asize;
+
+ if (BP_IS_HOLE(zio->io_bp))
+ return;
+
+ ASSERT(BP_IS_HOLE(&zio->io_bp_orig));
+
+ ASSERT(zio->io_child_type == ZIO_CHILD_GANG);
+ ASSERT3U(zio->io_prop.zp_ndvas, ==, lio->io_prop.zp_ndvas);
+ ASSERT3U(zio->io_prop.zp_ndvas, <=, BP_GET_NDVAS(zio->io_bp));
+ ASSERT3U(pio->io_prop.zp_ndvas, <=, BP_GET_NDVAS(pio->io_bp));
+ ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp));
+
+ mutex_enter(&pio->io_lock);
+ for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) {
+ ASSERT(DVA_GET_GANG(&pdva[d]));
+ asize = DVA_GET_ASIZE(&pdva[d]);
+ asize += DVA_GET_ASIZE(&cdva[d]);
+ DVA_SET_ASIZE(&pdva[d], asize);
+ }
+ mutex_exit(&pio->io_lock);
+}
+
+static int
+zio_write_gang_block(zio_t *pio)
+{
+ spa_t *spa = pio->io_spa;
+ blkptr_t *bp = pio->io_bp;
+ zio_t *lio = pio->io_logical;
+ zio_t *zio;
+ zio_gang_node_t *gn, **gnpp;
+ zio_gbh_phys_t *gbh;
+ uint64_t txg = pio->io_txg;
+ uint64_t resid = pio->io_size;
+ uint64_t lsize;
+ int ndvas = lio->io_prop.zp_ndvas;
+ int gbh_ndvas = MIN(ndvas + 1, spa_max_replication(spa));
+ zio_prop_t zp;
+ int error;
+
+ error = metaslab_alloc(spa, spa->spa_normal_class, SPA_GANGBLOCKSIZE,
+ bp, gbh_ndvas, txg, pio == lio ? NULL : lio->io_bp,
+ METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER);
+ if (error) {
+ pio->io_error = error;
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+
+ if (pio == lio) {
+ gnpp = &lio->io_gang_tree;
+ } else {
+ gnpp = pio->io_private;
+ ASSERT(pio->io_ready == zio_write_gang_member_ready);
+ }
+
+ gn = zio_gang_node_alloc(gnpp);
+ gbh = gn->gn_gbh;
+ bzero(gbh, SPA_GANGBLOCKSIZE);
+
+ /*
+ * Create the gang header.
+ */
+ zio = zio_rewrite(pio, spa, txg, bp, gbh, SPA_GANGBLOCKSIZE, NULL, NULL,
+ pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
+
+ /*
+ * Create and nowait the gang children.
+ */
+ for (int g = 0; resid != 0; resid -= lsize, g++) {
+ lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g),
+ SPA_MINBLOCKSIZE);
+ ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid);
+
+ zp.zp_checksum = lio->io_prop.zp_checksum;
+ zp.zp_compress = ZIO_COMPRESS_OFF;
+ zp.zp_type = DMU_OT_NONE;
+ zp.zp_level = 0;
+ zp.zp_ndvas = lio->io_prop.zp_ndvas;
+
+ zio_nowait(zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
+ (char *)pio->io_data + (pio->io_size - resid), lsize, &zp,
+ zio_write_gang_member_ready, NULL, &gn->gn_child[g],
+ pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
+ &pio->io_bookmark));
+ }
+
+ /*
+ * Set pio's pipeline to just wait for zio to finish.
+ */
+ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+
+ zio_nowait(zio);
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/*
+ * ==========================================================================
+ * Allocate and free blocks
+ * ==========================================================================
+ */
+
+static int
+zio_dva_allocate(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+ metaslab_class_t *mc = spa->spa_normal_class;
+ blkptr_t *bp = zio->io_bp;
+ int error;
+
+ ASSERT(BP_IS_HOLE(bp));
+ ASSERT3U(BP_GET_NDVAS(bp), ==, 0);
+ ASSERT3U(zio->io_prop.zp_ndvas, >, 0);
+ ASSERT3U(zio->io_prop.zp_ndvas, <=, spa_max_replication(spa));
+ ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));
+
+ error = metaslab_alloc(spa, mc, zio->io_size, bp,
+ zio->io_prop.zp_ndvas, zio->io_txg, NULL, 0);
+
+ if (error) {
+ if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE)
+ return (zio_write_gang_block(zio));
+ zio->io_error = error;
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_dva_free(zio_t *zio)
+{
+ metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE);
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_dva_claim(zio_t *zio)
+{
+ int error;
+
+ error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg);
+ if (error)
+ zio->io_error = error;
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/*
+ * Undo an allocation. This is used by zio_done() when an I/O fails
+ * and we want to give back the block we just allocated.
+ * This handles both normal blocks and gang blocks.
+ */
+static void
+zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
+{
+ spa_t *spa = zio->io_spa;
+ boolean_t now = !(zio->io_flags & ZIO_FLAG_IO_REWRITE);
+
+ ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp));
+
+ if (zio->io_bp == bp && !now) {
+ /*
+ * This is a rewrite for sync-to-convergence.
+ * We can't do a metaslab_free(NOW) because bp wasn't allocated
+ * during this sync pass, which means that metaslab_sync()
+ * already committed the allocation.
+ */
+ ASSERT(DVA_EQUAL(BP_IDENTITY(bp),
+ BP_IDENTITY(&zio->io_bp_orig)));
+ ASSERT(spa_sync_pass(spa) > 1);
+
+ if (BP_IS_GANG(bp) && gn == NULL) {
+ /*
+ * This is a gang leader whose gang header(s) we
+ * couldn't read now, so defer the free until later.
+ * The block should still be intact because without
+ * the headers, we'd never even start the rewrite.
+ */
+ bplist_enqueue_deferred(&spa->spa_sync_bplist, bp);
+ return;
+ }
+ }
+
+ if (!BP_IS_HOLE(bp))
+ metaslab_free(spa, bp, bp->blk_birth, now);
+
+ if (gn != NULL) {
+ for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
+ zio_dva_unallocate(zio, gn->gn_child[g],
+ &gn->gn_gbh->zg_blkptr[g]);
+ }
+ }
+}
+
+/*
+ * Try to allocate an intent log block. Return 0 on success, errno on failure.
+ */
+int
+zio_alloc_blk(spa_t *spa, uint64_t size, blkptr_t *new_bp, blkptr_t *old_bp,
+ uint64_t txg)
+{
+ int error;
+
+ error = metaslab_alloc(spa, spa->spa_log_class, size,
+ new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID);
+
+ if (error)
+ error = metaslab_alloc(spa, spa->spa_normal_class, size,
+ new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID);
+
+ if (error == 0) {
+ BP_SET_LSIZE(new_bp, size);
+ BP_SET_PSIZE(new_bp, size);
+ BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
+ BP_SET_CHECKSUM(new_bp, ZIO_CHECKSUM_ZILOG);
+ BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
+ BP_SET_LEVEL(new_bp, 0);
+ BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER);
+ }
+
+ return (error);
+}
+
+/*
+ * Free an intent log block. We know it can't be a gang block, so there's
+ * nothing to do except metaslab_free() it.
+ */
+void
+zio_free_blk(spa_t *spa, blkptr_t *bp, uint64_t txg)
+{
+ ASSERT(!BP_IS_GANG(bp));
+
+ metaslab_free(spa, bp, txg, B_FALSE);
+}
+
+/*
+ * ==========================================================================
+ * Read and write to physical devices
+ * ==========================================================================
+ */
+
+static void
+zio_vdev_io_probe_done(zio_t *zio)
+{
+ zio_t *dio;
+ vdev_t *vd = zio->io_private;
+
+ mutex_enter(&vd->vdev_probe_lock);
+ ASSERT(vd->vdev_probe_zio == zio);
+ vd->vdev_probe_zio = NULL;
+ mutex_exit(&vd->vdev_probe_lock);
+
+ while ((dio = zio->io_delegate_list) != NULL) {
+ zio->io_delegate_list = dio->io_delegate_next;
+ dio->io_delegate_next = NULL;
+ if (!vdev_accessible(vd, dio))
+ dio->io_error = ENXIO;
+ zio_execute(dio);
+ }
+}
+
+/*
+ * Probe the device to determine whether I/O failure is specific to this
+ * zio (e.g. a bad sector) or affects the entire vdev (e.g. unplugged).
+ */
+static int
+zio_vdev_io_probe(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ zio_t *pio = NULL;
+ boolean_t created_pio = B_FALSE;
+
+ /*
+ * Don't probe the probe.
+ */
+ if (zio->io_flags & ZIO_FLAG_PROBE)
+ return (ZIO_PIPELINE_CONTINUE);
+
+ /*
+ * To prevent 'probe storms' when a device fails, we create
+ * just one probe i/o at a time. All zios that want to probe
+ * this vdev will join the probe zio's io_delegate_list.
+ */
+ mutex_enter(&vd->vdev_probe_lock);
+
+ if ((pio = vd->vdev_probe_zio) == NULL) {
+ vd->vdev_probe_zio = pio = zio_root(zio->io_spa,
+ zio_vdev_io_probe_done, vd, ZIO_FLAG_CANFAIL);
+ created_pio = B_TRUE;
+ vd->vdev_probe_wanted = B_TRUE;
+ spa_async_request(zio->io_spa, SPA_ASYNC_PROBE);
+ }
+
+ zio->io_delegate_next = pio->io_delegate_list;
+ pio->io_delegate_list = zio;
+
+ mutex_exit(&vd->vdev_probe_lock);
+
+ if (created_pio) {
+ zio_nowait(vdev_probe(vd, pio));
+ zio_nowait(pio);
+ }
+
+ return (ZIO_PIPELINE_STOP);
+}
+
+static int
+zio_vdev_io_start(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ uint64_t align;
+ spa_t *spa = zio->io_spa;
+
+ ASSERT(zio->io_error == 0);
+ ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0);
+
+ if (vd == NULL) {
+ if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
+ spa_config_enter(spa, SCL_ZIO, zio, RW_READER);
+
+ /*
+ * The mirror_ops handle multiple DVAs in a single BP.
+ */
+ return (vdev_mirror_ops.vdev_op_io_start(zio));
+ }
+
+ align = 1ULL << vd->vdev_top->vdev_ashift;
+
+ if (P2PHASE(zio->io_size, align) != 0) {
+ uint64_t asize = P2ROUNDUP(zio->io_size, align);
+ char *abuf = zio_buf_alloc(asize);
+ ASSERT(vd == vd->vdev_top);
+ if (zio->io_type == ZIO_TYPE_WRITE) {
+ bcopy(zio->io_data, abuf, zio->io_size);
+ bzero(abuf + zio->io_size, asize - zio->io_size);
+ }
+ zio_push_transform(zio, abuf, asize, asize, zio_subblock);
+ }
+
+ ASSERT(P2PHASE(zio->io_offset, align) == 0);
+ ASSERT(P2PHASE(zio->io_size, align) == 0);
+ ASSERT(zio->io_type != ZIO_TYPE_WRITE || (spa_mode & FWRITE));
+
+ if (vd->vdev_ops->vdev_op_leaf &&
+ (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE)) {
+
+ if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio) == 0)
+ return (ZIO_PIPELINE_STOP);
+
+ if ((zio = vdev_queue_io(zio)) == NULL)
+ return (ZIO_PIPELINE_STOP);
+
+ if (!vdev_accessible(vd, zio)) {
+ zio->io_error = ENXIO;
+ zio_interrupt(zio);
+ return (ZIO_PIPELINE_STOP);
+ }
+
+ }
+
+ return (vd->vdev_ops->vdev_op_io_start(zio));
+}
+
+static int
+zio_vdev_io_done(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+ vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops;
+ boolean_t unexpected_error = B_FALSE;
+
+ if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE))
+ return (ZIO_PIPELINE_STOP);
+
+ ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
+
+ if (vd != NULL && vd->vdev_ops->vdev_op_leaf) {
+
+ vdev_queue_io_done(zio);
+
+ if (zio->io_type == ZIO_TYPE_WRITE)
+ vdev_cache_write(zio);
+
+ if (zio_injection_enabled && zio->io_error == 0)
+ zio->io_error = zio_handle_device_injection(vd, EIO);
+
+ if (zio_injection_enabled && zio->io_error == 0)
+ zio->io_error = zio_handle_label_injection(zio, EIO);
+
+ if (zio->io_error) {
+ if (!vdev_accessible(vd, zio)) {
+ zio->io_error = ENXIO;
+ } else {
+ unexpected_error = B_TRUE;
+ }
+ }
+ }
+
+ ops->vdev_op_io_done(zio);
+
+ if (unexpected_error)
+ return (zio_vdev_io_probe(zio));
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_vdev_io_assess(zio_t *zio)
+{
+ vdev_t *vd = zio->io_vd;
+
+ if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE))
+ return (ZIO_PIPELINE_STOP);
+
+ if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
+ spa_config_exit(zio->io_spa, SCL_ZIO, zio);
+
+ if (zio->io_vsd != NULL) {
+ zio->io_vsd_free(zio);
+ zio->io_vsd = NULL;
+ }
+
+ if (zio_injection_enabled && zio->io_error == 0)
+ zio->io_error = zio_handle_fault_injection(zio, EIO);
+
+ /*
+ * If the I/O failed, determine whether we should attempt to retry it.
+ */
+ if (zio->io_error && vd == NULL &&
+ !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) {
+ ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */
+ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */
+ zio->io_error = 0;
+ zio->io_flags |= ZIO_FLAG_IO_RETRY |
+ ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE;
+ zio->io_stage = ZIO_STAGE_VDEV_IO_START - 1;
+ zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE);
+ return (ZIO_PIPELINE_STOP);
+ }
+
+ /*
+ * If we got an error on a leaf device, convert it to ENXIO
+ * if the device is not accessible at all.
+ */
+ if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
+ !vdev_accessible(vd, zio))
+ zio->io_error = ENXIO;
+
+ /*
+ * If we can't write to an interior vdev (mirror or RAID-Z),
+ * set vdev_cant_write so that we stop trying to allocate from it.
+ */
+ if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
+ vd != NULL && !vd->vdev_ops->vdev_op_leaf)
+ vd->vdev_cant_write = B_TRUE;
+
+ if (zio->io_error)
+ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+void
+zio_vdev_io_reissue(zio_t *zio)
+{
+ ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
+ ASSERT(zio->io_error == 0);
+
+ zio->io_stage--;
+}
+
+void
+zio_vdev_io_redone(zio_t *zio)
+{
+ ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE);
+
+ zio->io_stage--;
+}
+
+void
+zio_vdev_io_bypass(zio_t *zio)
+{
+ ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
+ ASSERT(zio->io_error == 0);
+
+ zio->io_flags |= ZIO_FLAG_IO_BYPASS;
+ zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS - 1;
+}
+
+/*
+ * ==========================================================================
+ * Generate and verify checksums
+ * ==========================================================================
+ */
+static int
+zio_checksum_generate(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ enum zio_checksum checksum;
+
+ if (bp == NULL) {
+ /*
+ * This is zio_write_phys().
+ * We're either generating a label checksum, or none at all.
+ */
+ checksum = zio->io_prop.zp_checksum;
+
+ if (checksum == ZIO_CHECKSUM_OFF)
+ return (ZIO_PIPELINE_CONTINUE);
+
+ ASSERT(checksum == ZIO_CHECKSUM_LABEL);
+ } else {
+ if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) {
+ ASSERT(!IO_IS_ALLOCATING(zio));
+ checksum = ZIO_CHECKSUM_GANG_HEADER;
+ } else {
+ checksum = BP_GET_CHECKSUM(bp);
+ }
+ }
+
+ zio_checksum_compute(zio, checksum, zio->io_data, zio->io_size);
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_checksum_verify(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ int error;
+
+ if (bp == NULL) {
+ /*
+ * This is zio_read_phys().
+ * We're either verifying a label checksum, or nothing at all.
+ */
+ if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF)
+ return (ZIO_PIPELINE_CONTINUE);
+
+ ASSERT(zio->io_prop.zp_checksum == ZIO_CHECKSUM_LABEL);
+ }
+
+ if ((error = zio_checksum_error(zio)) != 0) {
+ zio->io_error = error;
+ if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
+ zfs_ereport_post(FM_EREPORT_ZFS_CHECKSUM,
+ zio->io_spa, zio->io_vd, zio, 0, 0);
+ }
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+/*
+ * Called by RAID-Z to ensure we don't compute the checksum twice.
+ */
+void
+zio_checksum_verified(zio_t *zio)
+{
+ zio->io_pipeline &= ~(1U << ZIO_STAGE_CHECKSUM_VERIFY);
+}
+
+/*
+ * ==========================================================================
+ * Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
+ * An error of 0 indictes success. ENXIO indicates whole-device failure,
+ * which may be transient (e.g. unplugged) or permament. ECKSUM and EIO
+ * indicate errors that are specific to one I/O, and most likely permanent.
+ * Any other error is presumed to be worse because we weren't expecting it.
+ * ==========================================================================
+ */
+int
+zio_worst_error(int e1, int e2)
+{
+ static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO };
+ int r1, r2;
+
+ for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++)
+ if (e1 == zio_error_rank[r1])
+ break;
+
+ for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++)
+ if (e2 == zio_error_rank[r2])
+ break;
+
+ return (r1 > r2 ? e1 : e2);
+}
+
+/*
+ * ==========================================================================
+ * I/O completion
+ * ==========================================================================
+ */
+static int
+zio_ready(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ zio_t *pio = zio->io_parent;
+
+ if (zio->io_ready) {
+ if (BP_IS_GANG(bp) &&
+ zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY))
+ return (ZIO_PIPELINE_STOP);
+
+ ASSERT(IO_IS_ALLOCATING(zio));
+ ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp));
+ ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);
+
+ zio->io_ready(zio);
+ }
+
+ if (bp != NULL && bp != &zio->io_bp_copy)
+ zio->io_bp_copy = *bp;
+
+ if (zio->io_error)
+ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+
+ if (pio != NULL)
+ zio_notify_parent(pio, zio, ZIO_WAIT_READY);
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
+static int
+zio_done(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+ zio_t *pio = zio->io_parent;
+ zio_t *lio = zio->io_logical;
+ blkptr_t *bp = zio->io_bp;
+ vdev_t *vd = zio->io_vd;
+ uint64_t psize = zio->io_size;
+
+ /*
+ * If our of children haven't all completed,
+ * wait for them and then repeat this pipeline stage.
+ */
+ if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE) ||
+ zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE) ||
+ zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE))
+ return (ZIO_PIPELINE_STOP);
+
+ for (int c = 0; c < ZIO_CHILD_TYPES; c++)
+ for (int w = 0; w < ZIO_WAIT_TYPES; w++)
+ ASSERT(zio->io_children[c][w] == 0);
+
+ if (bp != NULL) {
+ ASSERT(bp->blk_pad[0] == 0);
+ ASSERT(bp->blk_pad[1] == 0);
+ ASSERT(bp->blk_pad[2] == 0);
+ ASSERT(bcmp(bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 ||
+ (pio != NULL && bp == pio->io_bp));
+ if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(bp) &&
+ !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
+ ASSERT(!BP_SHOULD_BYTESWAP(bp));
+ ASSERT3U(zio->io_prop.zp_ndvas, <=, BP_GET_NDVAS(bp));
+ ASSERT(BP_COUNT_GANG(bp) == 0 ||
+ (BP_COUNT_GANG(bp) == BP_GET_NDVAS(bp)));
+ }
+ }
+
+ /*
+ * If there were child vdev or gang errors, they apply to us now.
+ */
+ zio_inherit_child_errors(zio, ZIO_CHILD_VDEV);
+ zio_inherit_child_errors(zio, ZIO_CHILD_GANG);
+
+ zio_pop_transforms(zio); /* note: may set zio->io_error */
+
+ vdev_stat_update(zio, psize);
+
+ if (zio->io_error) {
+ /*
+ * If this I/O is attached to a particular vdev,
+ * generate an error message describing the I/O failure
+ * at the block level. We ignore these errors if the
+ * device is currently unavailable.
+ */
+ if (zio->io_error != ECKSUM && vd != NULL && !vdev_is_dead(vd))
+ zfs_ereport_post(FM_EREPORT_ZFS_IO, spa, vd, zio, 0, 0);
+
+ if ((zio->io_error == EIO ||
+ !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) && zio == lio) {
+ /*
+ * For logical I/O requests, tell the SPA to log the
+ * error and generate a logical data ereport.
+ */
+ spa_log_error(spa, zio);
+ zfs_ereport_post(FM_EREPORT_ZFS_DATA, spa, NULL, zio,
+ 0, 0);
+ }
+ }
+
+ if (zio->io_error && zio == lio) {
+ /*
+ * Determine whether zio should be reexecuted. This will
+ * propagate all the way to the root via zio_notify_parent().
+ */
+ ASSERT(vd == NULL && bp != NULL);
+
+ if (IO_IS_ALLOCATING(zio))
+ if (zio->io_error != ENOSPC)
+ zio->io_reexecute |= ZIO_REEXECUTE_NOW;
+ else
+ zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
+
+ if ((zio->io_type == ZIO_TYPE_READ ||
+ zio->io_type == ZIO_TYPE_FREE) &&
+ zio->io_error == ENXIO &&
+ spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE)
+ zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
+
+ if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute)
+ zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
+ }
+
+ /*
+ * If there were logical child errors, they apply to us now.
+ * We defer this until now to avoid conflating logical child
+ * errors with errors that happened to the zio itself when
+ * updating vdev stats and reporting FMA events above.
+ */
+ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL);
+
+ if (zio->io_reexecute) {
+ /*
+ * This is a logical I/O that wants to reexecute.
+ *
+ * Reexecute is top-down. When an i/o fails, if it's not
+ * the root, it simply notifies its parent and sticks around.
+ * The parent, seeing that it still has children in zio_done(),
+ * does the same. This percolates all the way up to the root.
+ * The root i/o will reexecute or suspend the entire tree.
+ *
+ * This approach ensures that zio_reexecute() honors
+ * all the original i/o dependency relationships, e.g.
+ * parents not executing until children are ready.
+ */
+ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
+
+ if (IO_IS_ALLOCATING(zio))
+ zio_dva_unallocate(zio, zio->io_gang_tree, bp);
+
+ zio_gang_tree_free(&zio->io_gang_tree);
+
+ if (pio != NULL) {
+ /*
+ * We're not a root i/o, so there's nothing to do
+ * but notify our parent. Don't propagate errors
+ * upward since we haven't permanently failed yet.
+ */
+ zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE;
+ zio_notify_parent(pio, zio, ZIO_WAIT_DONE);
+ } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) {
+ /*
+ * We'd fail again if we reexecuted now, so suspend
+ * until conditions improve (e.g. device comes online).
+ */
+ zio_suspend(spa, zio);
+ } else {
+ /*
+ * Reexecution is potentially a huge amount of work.
+ * Hand it off to the otherwise-unused claim taskq.
+ */
+ (void) taskq_dispatch(
+ spa->spa_zio_taskq[ZIO_TYPE_CLAIM][ZIO_TASKQ_ISSUE],
+ (task_func_t *)zio_reexecute, zio, TQ_SLEEP);
+ }
+ return (ZIO_PIPELINE_STOP);
+ }
+
+ ASSERT(zio->io_child == NULL);
+ ASSERT(zio->io_reexecute == 0);
+ ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL));
+
+ if (zio->io_done)
+ zio->io_done(zio);
+
+ zio_gang_tree_free(&zio->io_gang_tree);
+
+ ASSERT(zio->io_delegate_list == NULL);
+ ASSERT(zio->io_delegate_next == NULL);
+
+ if (pio != NULL) {
+ zio_remove_child(pio, zio);
+ zio_notify_parent(pio, zio, ZIO_WAIT_DONE);
+ }
+
+ if (zio->io_waiter != NULL) {
+ mutex_enter(&zio->io_lock);
+ zio->io_executor = NULL;
+ cv_broadcast(&zio->io_cv);
+ mutex_exit(&zio->io_lock);
+ } else {
+ zio_destroy(zio);
+ }
+
+ return (ZIO_PIPELINE_STOP);
+}
+
+/*
+ * ==========================================================================
+ * I/O pipeline definition
+ * ==========================================================================
+ */
+static zio_pipe_stage_t *zio_pipeline[ZIO_STAGES] = {
+ NULL,
+ zio_issue_async,
+ zio_read_bp_init,
+ zio_write_bp_init,
+ zio_checksum_generate,
+ zio_gang_assemble,
+ zio_gang_issue,
+ zio_dva_allocate,
+ zio_dva_free,
+ zio_dva_claim,
+ zio_ready,
+ zio_vdev_io_start,
+ zio_vdev_io_done,
+ zio_vdev_io_assess,
+ zio_checksum_verify,
+ zio_done
+};
diff --git a/module/zfs/zio_checksum.c b/module/zfs/zio_checksum.c
new file mode 100644
index 000000000..bf7fe733f
--- /dev/null
+++ b/module/zfs/zio_checksum.c
@@ -0,0 +1,206 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/zio_checksum.h>
+
+/*
+ * Checksum vectors.
+ *
+ * In the SPA, everything is checksummed. We support checksum vectors
+ * for three distinct reasons:
+ *
+ * 1. Different kinds of data need different levels of protection.
+ * For SPA metadata, we always want a very strong checksum.
+ * For user data, we let users make the trade-off between speed
+ * and checksum strength.
+ *
+ * 2. Cryptographic hash and MAC algorithms are an area of active research.
+ * It is likely that in future hash functions will be at least as strong
+ * as current best-of-breed, and may be substantially faster as well.
+ * We want the ability to take advantage of these new hashes as soon as
+ * they become available.
+ *
+ * 3. If someone develops hardware that can compute a strong hash quickly,
+ * we want the ability to take advantage of that hardware.
+ *
+ * Of course, we don't want a checksum upgrade to invalidate existing
+ * data, so we store the checksum *function* in five bits of the DVA.
+ * This gives us room for up to 32 different checksum functions.
+ *
+ * When writing a block, we always checksum it with the latest-and-greatest
+ * checksum function of the appropriate strength. When reading a block,
+ * we compare the expected checksum against the actual checksum, which we
+ * compute via the checksum function specified in the DVA encoding.
+ */
+
+/*ARGSUSED*/
+static void
+zio_checksum_off(const void *buf, uint64_t size, zio_cksum_t *zcp)
+{
+ ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
+}
+
+zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = {
+ {{NULL, NULL}, 0, 0, "inherit"},
+ {{NULL, NULL}, 0, 0, "on"},
+ {{zio_checksum_off, zio_checksum_off}, 0, 0, "off"},
+ {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, "label"},
+ {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, "gang_header"},
+ {{fletcher_2_native, fletcher_2_byteswap}, 0, 1, "zilog"},
+ {{fletcher_2_native, fletcher_2_byteswap}, 0, 0, "fletcher2"},
+ {{fletcher_4_native, fletcher_4_byteswap}, 1, 0, "fletcher4"},
+ {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 0, "SHA256"},
+};
+
+uint8_t
+zio_checksum_select(uint8_t child, uint8_t parent)
+{
+ ASSERT(child < ZIO_CHECKSUM_FUNCTIONS);
+ ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS);
+ ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);
+
+ if (child == ZIO_CHECKSUM_INHERIT)
+ return (parent);
+
+ if (child == ZIO_CHECKSUM_ON)
+ return (ZIO_CHECKSUM_ON_VALUE);
+
+ return (child);
+}
+
+/*
+ * Set the external verifier for a gang block based on <vdev, offset, txg>,
+ * a tuple which is guaranteed to be unique for the life of the pool.
+ */
+static void
+zio_checksum_gang_verifier(zio_cksum_t *zcp, blkptr_t *bp)
+{
+ dva_t *dva = BP_IDENTITY(bp);
+ uint64_t txg = bp->blk_birth;
+
+ ASSERT(BP_IS_GANG(bp));
+
+ ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0);
+}
+
+/*
+ * Set the external verifier for a label block based on its offset.
+ * The vdev is implicit, and the txg is unknowable at pool open time --
+ * hence the logic in vdev_uberblock_load() to find the most recent copy.
+ */
+static void
+zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset)
+{
+ ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0);
+}
+
+/*
+ * Generate the checksum.
+ */
+void
+zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
+ void *data, uint64_t size)
+{
+ blkptr_t *bp = zio->io_bp;
+ uint64_t offset = zio->io_offset;
+ zio_block_tail_t *zbt = (zio_block_tail_t *)((char *)data + size) - 1;
+ zio_checksum_info_t *ci = &zio_checksum_table[checksum];
+ zio_cksum_t zbt_cksum;
+
+ ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS);
+ ASSERT(ci->ci_func[0] != NULL);
+
+ if (ci->ci_zbt) {
+ if (checksum == ZIO_CHECKSUM_GANG_HEADER)
+ zio_checksum_gang_verifier(&zbt->zbt_cksum, bp);
+ else if (checksum == ZIO_CHECKSUM_LABEL)
+ zio_checksum_label_verifier(&zbt->zbt_cksum, offset);
+ else
+ bp->blk_cksum = zbt->zbt_cksum;
+ zbt->zbt_magic = ZBT_MAGIC;
+ ci->ci_func[0](data, size, &zbt_cksum);
+ zbt->zbt_cksum = zbt_cksum;
+ } else {
+ ci->ci_func[0](data, size, &bp->blk_cksum);
+ }
+}
+
+int
+zio_checksum_error(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum :
+ (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp)));
+ int byteswap;
+ void *data = zio->io_data;
+ uint64_t size = (bp == NULL ? zio->io_size :
+ (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp)));
+ uint64_t offset = zio->io_offset;
+ zio_block_tail_t *zbt = (zio_block_tail_t *)((char *)data + size) - 1;
+ zio_checksum_info_t *ci = &zio_checksum_table[checksum];
+ zio_cksum_t actual_cksum, expected_cksum, verifier;
+
+ if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL)
+ return (EINVAL);
+
+ if (ci->ci_zbt) {
+ if (checksum == ZIO_CHECKSUM_GANG_HEADER)
+ zio_checksum_gang_verifier(&verifier, bp);
+ else if (checksum == ZIO_CHECKSUM_LABEL)
+ zio_checksum_label_verifier(&verifier, offset);
+ else
+ verifier = bp->blk_cksum;
+
+ byteswap = (zbt->zbt_magic == BSWAP_64(ZBT_MAGIC));
+
+ if (byteswap)
+ byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));
+
+ expected_cksum = zbt->zbt_cksum;
+ zbt->zbt_cksum = verifier;
+ ci->ci_func[byteswap](data, size, &actual_cksum);
+ zbt->zbt_cksum = expected_cksum;
+
+ if (byteswap)
+ byteswap_uint64_array(&expected_cksum,
+ sizeof (zio_cksum_t));
+ } else {
+ ASSERT(!BP_IS_GANG(bp));
+ byteswap = BP_SHOULD_BYTESWAP(bp);
+ expected_cksum = bp->blk_cksum;
+ ci->ci_func[byteswap](data, size, &actual_cksum);
+ }
+
+ if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
+ return (ECKSUM);
+
+ if (zio_injection_enabled && !zio->io_error)
+ return (zio_handle_fault_injection(zio, ECKSUM));
+
+ return (0);
+}
diff --git a/module/zfs/zio_compress.c b/module/zfs/zio_compress.c
new file mode 100644
index 000000000..c563be4eb
--- /dev/null
+++ b/module/zfs/zio_compress.c
@@ -0,0 +1,148 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#pragma ident "%Z%%M% %I% %E% SMI"
+
+#include <sys/zfs_context.h>
+#include <sys/compress.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/zio_compress.h>
+
+/*
+ * Compression vectors.
+ */
+
+zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS] = {
+ {NULL, NULL, 0, "inherit"},
+ {NULL, NULL, 0, "on"},
+ {NULL, NULL, 0, "uncompressed"},
+ {lzjb_compress, lzjb_decompress, 0, "lzjb"},
+ {NULL, NULL, 0, "empty"},
+ {gzip_compress, gzip_decompress, 1, "gzip-1"},
+ {gzip_compress, gzip_decompress, 2, "gzip-2"},
+ {gzip_compress, gzip_decompress, 3, "gzip-3"},
+ {gzip_compress, gzip_decompress, 4, "gzip-4"},
+ {gzip_compress, gzip_decompress, 5, "gzip-5"},
+ {gzip_compress, gzip_decompress, 6, "gzip-6"},
+ {gzip_compress, gzip_decompress, 7, "gzip-7"},
+ {gzip_compress, gzip_decompress, 8, "gzip-8"},
+ {gzip_compress, gzip_decompress, 9, "gzip-9"},
+};
+
+uint8_t
+zio_compress_select(uint8_t child, uint8_t parent)
+{
+ ASSERT(child < ZIO_COMPRESS_FUNCTIONS);
+ ASSERT(parent < ZIO_COMPRESS_FUNCTIONS);
+ ASSERT(parent != ZIO_COMPRESS_INHERIT && parent != ZIO_COMPRESS_ON);
+
+ if (child == ZIO_COMPRESS_INHERIT)
+ return (parent);
+
+ if (child == ZIO_COMPRESS_ON)
+ return (ZIO_COMPRESS_ON_VALUE);
+
+ return (child);
+}
+
+int
+zio_compress_data(int cpfunc, void *src, uint64_t srcsize, void **destp,
+ uint64_t *destsizep, uint64_t *destbufsizep)
+{
+ uint64_t *word, *word_end;
+ uint64_t ciosize, gapsize, destbufsize;
+ zio_compress_info_t *ci = &zio_compress_table[cpfunc];
+ char *dest;
+ uint_t allzero;
+
+ ASSERT((uint_t)cpfunc < ZIO_COMPRESS_FUNCTIONS);
+ ASSERT((uint_t)cpfunc == ZIO_COMPRESS_EMPTY || ci->ci_compress != NULL);
+
+ /*
+ * If the data is all zeroes, we don't even need to allocate
+ * a block for it. We indicate this by setting *destsizep = 0.
+ */
+ allzero = 1;
+ word = src;
+ word_end = (uint64_t *)(uintptr_t)((uintptr_t)word + srcsize);
+ while (word < word_end) {
+ if (*word++ != 0) {
+ allzero = 0;
+ break;
+ }
+ }
+ if (allzero) {
+ *destp = NULL;
+ *destsizep = 0;
+ *destbufsizep = 0;
+ return (1);
+ }
+
+ if (cpfunc == ZIO_COMPRESS_EMPTY)
+ return (0);
+
+ /* Compress at least 12.5% */
+ destbufsize = P2ALIGN(srcsize - (srcsize >> 3), SPA_MINBLOCKSIZE);
+ if (destbufsize == 0)
+ return (0);
+ dest = zio_buf_alloc(destbufsize);
+
+ ciosize = ci->ci_compress(src, dest, (size_t)srcsize,
+ (size_t)destbufsize, ci->ci_level);
+ if (ciosize > destbufsize) {
+ zio_buf_free(dest, destbufsize);
+ return (0);
+ }
+
+ /* Cool. We compressed at least as much as we were hoping to. */
+
+ /* For security, make sure we don't write random heap crap to disk */
+ gapsize = P2ROUNDUP(ciosize, SPA_MINBLOCKSIZE) - ciosize;
+ if (gapsize != 0) {
+ bzero(dest + ciosize, gapsize);
+ ciosize += gapsize;
+ }
+
+ ASSERT3U(ciosize, <=, destbufsize);
+ ASSERT(P2PHASE(ciosize, SPA_MINBLOCKSIZE) == 0);
+ *destp = dest;
+ *destsizep = ciosize;
+ *destbufsizep = destbufsize;
+
+ return (1);
+}
+
+int
+zio_decompress_data(int cpfunc, void *src, uint64_t srcsize,
+ void *dest, uint64_t destsize)
+{
+ zio_compress_info_t *ci = &zio_compress_table[cpfunc];
+
+ ASSERT((uint_t)cpfunc < ZIO_COMPRESS_FUNCTIONS);
+
+ return (ci->ci_decompress(src, dest, srcsize, destsize, ci->ci_level));
+}
diff --git a/module/zfs/zio_inject.c b/module/zfs/zio_inject.c
new file mode 100644
index 000000000..b3469fdd5
--- /dev/null
+++ b/module/zfs/zio_inject.c
@@ -0,0 +1,370 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * ZFS fault injection
+ *
+ * To handle fault injection, we keep track of a series of zinject_record_t
+ * structures which describe which logical block(s) should be injected with a
+ * fault. These are kept in a global list. Each record corresponds to a given
+ * spa_t and maintains a special hold on the spa_t so that it cannot be deleted
+ * or exported while the injection record exists.
+ *
+ * Device level injection is done using the 'zi_guid' field. If this is set, it
+ * means that the error is destined for a particular device, not a piece of
+ * data.
+ *
+ * This is a rather poor data structure and algorithm, but we don't expect more
+ * than a few faults at any one time, so it should be sufficient for our needs.
+ */
+
+#include <sys/arc.h>
+#include <sys/zio_impl.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/spa_impl.h>
+#include <sys/vdev_impl.h>
+#include <sys/fs/zfs.h>
+
+uint32_t zio_injection_enabled;
+
+typedef struct inject_handler {
+ int zi_id;
+ spa_t *zi_spa;
+ zinject_record_t zi_record;
+ list_node_t zi_link;
+} inject_handler_t;
+
+static list_t inject_handlers;
+static krwlock_t inject_lock;
+static int inject_next_id = 1;
+
+/*
+ * Returns true if the given record matches the I/O in progress.
+ */
+static boolean_t
+zio_match_handler(zbookmark_t *zb, uint64_t type,
+ zinject_record_t *record, int error)
+{
+ /*
+ * Check for a match against the MOS, which is based on type
+ */
+ if (zb->zb_objset == 0 && record->zi_objset == 0 &&
+ record->zi_object == 0) {
+ if (record->zi_type == DMU_OT_NONE ||
+ type == record->zi_type)
+ return (record->zi_freq == 0 ||
+ spa_get_random(100) < record->zi_freq);
+ else
+ return (B_FALSE);
+ }
+
+ /*
+ * Check for an exact match.
+ */
+ if (zb->zb_objset == record->zi_objset &&
+ zb->zb_object == record->zi_object &&
+ zb->zb_level == record->zi_level &&
+ zb->zb_blkid >= record->zi_start &&
+ zb->zb_blkid <= record->zi_end &&
+ error == record->zi_error)
+ return (record->zi_freq == 0 ||
+ spa_get_random(100) < record->zi_freq);
+
+ return (B_FALSE);
+}
+
+/*
+ * Determine if the I/O in question should return failure. Returns the errno
+ * to be returned to the caller.
+ */
+int
+zio_handle_fault_injection(zio_t *zio, int error)
+{
+ int ret = 0;
+ inject_handler_t *handler;
+
+ /*
+ * Ignore I/O not associated with any logical data.
+ */
+ if (zio->io_logical == NULL)
+ return (0);
+
+ /*
+ * Currently, we only support fault injection on reads.
+ */
+ if (zio->io_type != ZIO_TYPE_READ)
+ return (0);
+
+ rw_enter(&inject_lock, RW_READER);
+
+ for (handler = list_head(&inject_handlers); handler != NULL;
+ handler = list_next(&inject_handlers, handler)) {
+
+ /* Ignore errors not destined for this pool */
+ if (zio->io_spa != handler->zi_spa)
+ continue;
+
+ /* Ignore device errors */
+ if (handler->zi_record.zi_guid != 0)
+ continue;
+
+ /* If this handler matches, return EIO */
+ if (zio_match_handler(&zio->io_logical->io_bookmark,
+ zio->io_bp ? BP_GET_TYPE(zio->io_bp) : DMU_OT_NONE,
+ &handler->zi_record, error)) {
+ ret = error;
+ break;
+ }
+ }
+
+ rw_exit(&inject_lock);
+
+ return (ret);
+}
+
+/*
+ * Determine if the zio is part of a label update and has an injection
+ * handler associated with that portion of the label. Currently, we
+ * allow error injection in either the nvlist or the uberblock region of
+ * of the vdev label.
+ */
+int
+zio_handle_label_injection(zio_t *zio, int error)
+{
+ inject_handler_t *handler;
+ vdev_t *vd = zio->io_vd;
+ uint64_t offset = zio->io_offset;
+ int label;
+ int ret = 0;
+
+ if (offset + zio->io_size > VDEV_LABEL_START_SIZE &&
+ offset < vd->vdev_psize - VDEV_LABEL_END_SIZE)
+ return (0);
+
+ rw_enter(&inject_lock, RW_READER);
+
+ for (handler = list_head(&inject_handlers); handler != NULL;
+ handler = list_next(&inject_handlers, handler)) {
+ uint64_t start = handler->zi_record.zi_start;
+ uint64_t end = handler->zi_record.zi_end;
+
+ /* Ignore device only faults */
+ if (handler->zi_record.zi_start == 0)
+ continue;
+
+ /*
+ * The injection region is the relative offsets within a
+ * vdev label. We must determine the label which is being
+ * updated and adjust our region accordingly.
+ */
+ label = vdev_label_number(vd->vdev_psize, offset);
+ start = vdev_label_offset(vd->vdev_psize, label, start);
+ end = vdev_label_offset(vd->vdev_psize, label, end);
+
+ if (zio->io_vd->vdev_guid == handler->zi_record.zi_guid &&
+ (offset >= start && offset <= end)) {
+ ret = error;
+ break;
+ }
+ }
+ rw_exit(&inject_lock);
+ return (ret);
+}
+
+
+int
+zio_handle_device_injection(vdev_t *vd, int error)
+{
+ inject_handler_t *handler;
+ int ret = 0;
+
+ rw_enter(&inject_lock, RW_READER);
+
+ for (handler = list_head(&inject_handlers); handler != NULL;
+ handler = list_next(&inject_handlers, handler)) {
+
+ /* Ignore label specific faults */
+ if (handler->zi_record.zi_start != 0)
+ continue;
+
+ if (vd->vdev_guid == handler->zi_record.zi_guid) {
+ if (handler->zi_record.zi_error == error) {
+ /*
+ * For a failed open, pretend like the device
+ * has gone away.
+ */
+ if (error == ENXIO)
+ vd->vdev_stat.vs_aux =
+ VDEV_AUX_OPEN_FAILED;
+ ret = error;
+ break;
+ }
+ if (handler->zi_record.zi_error == ENXIO) {
+ ret = EIO;
+ break;
+ }
+ }
+ }
+
+ rw_exit(&inject_lock);
+
+ return (ret);
+}
+
+/*
+ * Create a new handler for the given record. We add it to the list, adding
+ * a reference to the spa_t in the process. We increment zio_injection_enabled,
+ * which is the switch to trigger all fault injection.
+ */
+int
+zio_inject_fault(char *name, int flags, int *id, zinject_record_t *record)
+{
+ inject_handler_t *handler;
+ int error;
+ spa_t *spa;
+
+ /*
+ * If this is pool-wide metadata, make sure we unload the corresponding
+ * spa_t, so that the next attempt to load it will trigger the fault.
+ * We call spa_reset() to unload the pool appropriately.
+ */
+ if (flags & ZINJECT_UNLOAD_SPA)
+ if ((error = spa_reset(name)) != 0)
+ return (error);
+
+ if (!(flags & ZINJECT_NULL)) {
+ /*
+ * spa_inject_ref() will add an injection reference, which will
+ * prevent the pool from being removed from the namespace while
+ * still allowing it to be unloaded.
+ */
+ if ((spa = spa_inject_addref(name)) == NULL)
+ return (ENOENT);
+
+ handler = kmem_alloc(sizeof (inject_handler_t), KM_SLEEP);
+
+ rw_enter(&inject_lock, RW_WRITER);
+
+ *id = handler->zi_id = inject_next_id++;
+ handler->zi_spa = spa;
+ handler->zi_record = *record;
+ list_insert_tail(&inject_handlers, handler);
+ atomic_add_32(&zio_injection_enabled, 1);
+
+ rw_exit(&inject_lock);
+ }
+
+ /*
+ * Flush the ARC, so that any attempts to read this data will end up
+ * going to the ZIO layer. Note that this is a little overkill, but
+ * we don't have the necessary ARC interfaces to do anything else, and
+ * fault injection isn't a performance critical path.
+ */
+ if (flags & ZINJECT_FLUSH_ARC)
+ arc_flush(NULL);
+
+ return (0);
+}
+
+/*
+ * Returns the next record with an ID greater than that supplied to the
+ * function. Used to iterate over all handlers in the system.
+ */
+int
+zio_inject_list_next(int *id, char *name, size_t buflen,
+ zinject_record_t *record)
+{
+ inject_handler_t *handler;
+ int ret;
+
+ mutex_enter(&spa_namespace_lock);
+ rw_enter(&inject_lock, RW_READER);
+
+ for (handler = list_head(&inject_handlers); handler != NULL;
+ handler = list_next(&inject_handlers, handler))
+ if (handler->zi_id > *id)
+ break;
+
+ if (handler) {
+ *record = handler->zi_record;
+ *id = handler->zi_id;
+ (void) strncpy(name, spa_name(handler->zi_spa), buflen);
+ ret = 0;
+ } else {
+ ret = ENOENT;
+ }
+
+ rw_exit(&inject_lock);
+ mutex_exit(&spa_namespace_lock);
+
+ return (ret);
+}
+
+/*
+ * Clear the fault handler with the given identifier, or return ENOENT if none
+ * exists.
+ */
+int
+zio_clear_fault(int id)
+{
+ inject_handler_t *handler;
+ int ret;
+
+ rw_enter(&inject_lock, RW_WRITER);
+
+ for (handler = list_head(&inject_handlers); handler != NULL;
+ handler = list_next(&inject_handlers, handler))
+ if (handler->zi_id == id)
+ break;
+
+ if (handler == NULL) {
+ ret = ENOENT;
+ } else {
+ list_remove(&inject_handlers, handler);
+ spa_inject_delref(handler->zi_spa);
+ kmem_free(handler, sizeof (inject_handler_t));
+ atomic_add_32(&zio_injection_enabled, -1);
+ ret = 0;
+ }
+
+ rw_exit(&inject_lock);
+
+ return (ret);
+}
+
+void
+zio_inject_init(void)
+{
+ rw_init(&inject_lock, NULL, RW_DEFAULT, NULL);
+ list_create(&inject_handlers, sizeof (inject_handler_t),
+ offsetof(inject_handler_t, zi_link));
+}
+
+void
+zio_inject_fini(void)
+{
+ list_destroy(&inject_handlers);
+ rw_destroy(&inject_lock);
+}
diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
new file mode 100644
index 000000000..4e993060c
--- /dev/null
+++ b/module/zfs/zvol.c
@@ -0,0 +1,1722 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * ZFS volume emulation driver.
+ *
+ * Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes.
+ * Volumes are accessed through the symbolic links named:
+ *
+ * /dev/zvol/dsk/<pool_name>/<dataset_name>
+ * /dev/zvol/rdsk/<pool_name>/<dataset_name>
+ *
+ * These links are created by the ZFS-specific devfsadm link generator.
+ * Volumes are persistent through reboot. No user command needs to be
+ * run before opening and using a device.
+ */
+
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/errno.h>
+#include <sys/uio.h>
+#include <sys/buf.h>
+#include <sys/modctl.h>
+#include <sys/open.h>
+#include <sys/kmem.h>
+#include <sys/conf.h>
+#include <sys/cmn_err.h>
+#include <sys/stat.h>
+#include <sys/zap.h>
+#include <sys/spa.h>
+#include <sys/zio.h>
+#include <sys/dmu_traverse.h>
+#include <sys/dnode.h>
+#include <sys/dsl_dataset.h>
+#include <sys/dsl_prop.h>
+#include <sys/dkio.h>
+#include <sys/efi_partition.h>
+#include <sys/byteorder.h>
+#include <sys/pathname.h>
+#include <sys/ddi.h>
+#include <sys/sunddi.h>
+#include <sys/crc32.h>
+#include <sys/dirent.h>
+#include <sys/policy.h>
+#include <sys/fs/zfs.h>
+#include <sys/zfs_ioctl.h>
+#include <sys/mkdev.h>
+#include <sys/zil.h>
+#include <sys/refcount.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_rlock.h>
+#include <sys/vdev_disk.h>
+#include <sys/vdev_impl.h>
+#include <sys/zvol.h>
+#include <sys/dumphdr.h>
+
+#include "zfs_namecheck.h"
+
+static void *zvol_state;
+
+#define ZVOL_DUMPSIZE "dumpsize"
+
+/*
+ * This lock protects the zvol_state structure from being modified
+ * while it's being used, e.g. an open that comes in before a create
+ * finishes. It also protects temporary opens of the dataset so that,
+ * e.g., an open doesn't get a spurious EBUSY.
+ */
+static kmutex_t zvol_state_lock;
+static uint32_t zvol_minors;
+
+typedef struct zvol_extent {
+ list_node_t ze_node;
+ dva_t ze_dva; /* dva associated with this extent */
+ uint64_t ze_nblks; /* number of blocks in extent */
+} zvol_extent_t;
+
+/*
+ * The in-core state of each volume.
+ */
+typedef struct zvol_state {
+ char zv_name[MAXPATHLEN]; /* pool/dd name */
+ uint64_t zv_volsize; /* amount of space we advertise */
+ uint64_t zv_volblocksize; /* volume block size */
+ minor_t zv_minor; /* minor number */
+ uint8_t zv_min_bs; /* minimum addressable block shift */
+ uint8_t zv_flags; /* readonly; dumpified */
+ objset_t *zv_objset; /* objset handle */
+ uint32_t zv_mode; /* DS_MODE_* flags at open time */
+ uint32_t zv_open_count[OTYPCNT]; /* open counts */
+ uint32_t zv_total_opens; /* total open count */
+ zilog_t *zv_zilog; /* ZIL handle */
+ list_t zv_extents; /* List of extents for dump */
+ uint64_t zv_txg_assign; /* txg to assign during ZIL replay */
+ znode_t zv_znode; /* for range locking */
+} zvol_state_t;
+
+/*
+ * zvol specific flags
+ */
+#define ZVOL_RDONLY 0x1
+#define ZVOL_DUMPIFIED 0x2
+#define ZVOL_EXCL 0x4
+
+/*
+ * zvol maximum transfer in one DMU tx.
+ */
+int zvol_maxphys = DMU_MAX_ACCESS/2;
+
+extern int zfs_set_prop_nvlist(const char *, nvlist_t *);
+static int zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio);
+static int zvol_dumpify(zvol_state_t *zv);
+static int zvol_dump_fini(zvol_state_t *zv);
+static int zvol_dump_init(zvol_state_t *zv, boolean_t resize);
+
+static void
+zvol_size_changed(zvol_state_t *zv, major_t maj)
+{
+ dev_t dev = makedevice(maj, zv->zv_minor);
+
+ VERIFY(ddi_prop_update_int64(dev, zfs_dip,
+ "Size", zv->zv_volsize) == DDI_SUCCESS);
+ VERIFY(ddi_prop_update_int64(dev, zfs_dip,
+ "Nblocks", lbtodb(zv->zv_volsize)) == DDI_SUCCESS);
+
+ /* Notify specfs to invalidate the cached size */
+ spec_size_invalidate(dev, VBLK);
+ spec_size_invalidate(dev, VCHR);
+}
+
+int
+zvol_check_volsize(uint64_t volsize, uint64_t blocksize)
+{
+ if (volsize == 0)
+ return (EINVAL);
+
+ if (volsize % blocksize != 0)
+ return (EINVAL);
+
+#ifdef _ILP32
+ if (volsize - 1 > SPEC_MAXOFFSET_T)
+ return (EOVERFLOW);
+#endif
+ return (0);
+}
+
+int
+zvol_check_volblocksize(uint64_t volblocksize)
+{
+ if (volblocksize < SPA_MINBLOCKSIZE ||
+ volblocksize > SPA_MAXBLOCKSIZE ||
+ !ISP2(volblocksize))
+ return (EDOM);
+
+ return (0);
+}
+
+static void
+zvol_readonly_changed_cb(void *arg, uint64_t newval)
+{
+ zvol_state_t *zv = arg;
+
+ if (newval)
+ zv->zv_flags |= ZVOL_RDONLY;
+ else
+ zv->zv_flags &= ~ZVOL_RDONLY;
+}
+
+int
+zvol_get_stats(objset_t *os, nvlist_t *nv)
+{
+ int error;
+ dmu_object_info_t doi;
+ uint64_t val;
+
+
+ error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val);
+ if (error)
+ return (error);
+
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val);
+
+ error = dmu_object_info(os, ZVOL_OBJ, &doi);
+
+ if (error == 0) {
+ dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE,
+ doi.doi_data_block_size);
+ }
+
+ return (error);
+}
+
+/*
+ * Find a free minor number.
+ */
+static minor_t
+zvol_minor_alloc(void)
+{
+ minor_t minor;
+
+ ASSERT(MUTEX_HELD(&zvol_state_lock));
+
+ for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++)
+ if (ddi_get_soft_state(zvol_state, minor) == NULL)
+ return (minor);
+
+ return (0);
+}
+
+static zvol_state_t *
+zvol_minor_lookup(const char *name)
+{
+ minor_t minor;
+ zvol_state_t *zv;
+
+ ASSERT(MUTEX_HELD(&zvol_state_lock));
+
+ for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) {
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL)
+ continue;
+ if (strcmp(zv->zv_name, name) == 0)
+ break;
+ }
+
+ return (zv);
+}
+
+/* extent mapping arg */
+struct maparg {
+ zvol_state_t *ma_zv;
+ uint64_t ma_blks;
+};
+
+/*ARGSUSED*/
+static int
+zvol_map_block(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
+ const dnode_phys_t *dnp, void *arg)
+{
+ struct maparg *ma = arg;
+ zvol_extent_t *ze;
+ int bs = ma->ma_zv->zv_volblocksize;
+
+ if (bp == NULL || zb->zb_object != ZVOL_OBJ || zb->zb_level != 0)
+ return (0);
+
+ VERIFY3U(ma->ma_blks, ==, zb->zb_blkid);
+ ma->ma_blks++;
+
+ /* Abort immediately if we have encountered gang blocks */
+ if (BP_IS_GANG(bp))
+ return (EFRAGS);
+
+ /*
+ * See if the block is at the end of the previous extent.
+ */
+ ze = list_tail(&ma->ma_zv->zv_extents);
+ if (ze &&
+ DVA_GET_VDEV(BP_IDENTITY(bp)) == DVA_GET_VDEV(&ze->ze_dva) &&
+ DVA_GET_OFFSET(BP_IDENTITY(bp)) ==
+ DVA_GET_OFFSET(&ze->ze_dva) + ze->ze_nblks * bs) {
+ ze->ze_nblks++;
+ return (0);
+ }
+
+ dprintf_bp(bp, "%s", "next blkptr:");
+
+ /* start a new extent */
+ ze = kmem_zalloc(sizeof (zvol_extent_t), KM_SLEEP);
+ ze->ze_dva = bp->blk_dva[0]; /* structure assignment */
+ ze->ze_nblks = 1;
+ list_insert_tail(&ma->ma_zv->zv_extents, ze);
+ return (0);
+}
+
+static void
+zvol_free_extents(zvol_state_t *zv)
+{
+ zvol_extent_t *ze;
+
+ while (ze = list_head(&zv->zv_extents)) {
+ list_remove(&zv->zv_extents, ze);
+ kmem_free(ze, sizeof (zvol_extent_t));
+ }
+}
+
+static int
+zvol_get_lbas(zvol_state_t *zv)
+{
+ struct maparg ma;
+ int err;
+
+ ma.ma_zv = zv;
+ ma.ma_blks = 0;
+ zvol_free_extents(zv);
+
+ err = traverse_dataset(dmu_objset_ds(zv->zv_objset), 0,
+ TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, zvol_map_block, &ma);
+ if (err || ma.ma_blks != (zv->zv_volsize / zv->zv_volblocksize)) {
+ zvol_free_extents(zv);
+ return (err ? err : EIO);
+ }
+
+ return (0);
+}
+
+/* ARGSUSED */
+void
+zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
+{
+ zfs_creat_t *zct = arg;
+ nvlist_t *nvprops = zct->zct_props;
+ int error;
+ uint64_t volblocksize, volsize;
+
+ VERIFY(nvlist_lookup_uint64(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0);
+ if (nvlist_lookup_uint64(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0)
+ volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
+
+ /*
+ * These properties must be removed from the list so the generic
+ * property setting step won't apply to them.
+ */
+ VERIFY(nvlist_remove_all(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0);
+ (void) nvlist_remove_all(nvprops,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE));
+
+ error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize,
+ DMU_OT_NONE, 0, tx);
+ ASSERT(error == 0);
+
+ error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
+ DMU_OT_NONE, 0, tx);
+ ASSERT(error == 0);
+
+ error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx);
+ ASSERT(error == 0);
+}
+
+/*
+ * Replay a TX_WRITE ZIL transaction that didn't get committed
+ * after a system failure
+ */
+static int
+zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap)
+{
+ objset_t *os = zv->zv_objset;
+ char *data = (char *)(lr + 1); /* data follows lr_write_t */
+ uint64_t off = lr->lr_offset;
+ uint64_t len = lr->lr_length;
+ dmu_tx_t *tx;
+ int error;
+
+ if (byteswap)
+ byteswap_uint64_array(lr, sizeof (*lr));
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_write(tx, ZVOL_OBJ, off, len);
+ error = dmu_tx_assign(tx, zv->zv_txg_assign);
+ if (error) {
+ dmu_tx_abort(tx);
+ } else {
+ dmu_write(os, ZVOL_OBJ, off, len, data, tx);
+ dmu_tx_commit(tx);
+ }
+
+ return (error);
+}
+
+/* ARGSUSED */
+static int
+zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap)
+{
+ return (ENOTSUP);
+}
+
+/*
+ * Callback vectors for replaying records.
+ * Only TX_WRITE is needed for zvol.
+ */
+zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = {
+ zvol_replay_err, /* 0 no such transaction type */
+ zvol_replay_err, /* TX_CREATE */
+ zvol_replay_err, /* TX_MKDIR */
+ zvol_replay_err, /* TX_MKXATTR */
+ zvol_replay_err, /* TX_SYMLINK */
+ zvol_replay_err, /* TX_REMOVE */
+ zvol_replay_err, /* TX_RMDIR */
+ zvol_replay_err, /* TX_LINK */
+ zvol_replay_err, /* TX_RENAME */
+ zvol_replay_write, /* TX_WRITE */
+ zvol_replay_err, /* TX_TRUNCATE */
+ zvol_replay_err, /* TX_SETATTR */
+ zvol_replay_err, /* TX_ACL */
+};
+
+/*
+ * Create a minor node (plus a whole lot more) for the specified volume.
+ */
+int
+zvol_create_minor(const char *name, major_t maj)
+{
+ zvol_state_t *zv;
+ objset_t *os;
+ dmu_object_info_t doi;
+ uint64_t volsize;
+ minor_t minor = 0;
+ struct pathname linkpath;
+ int ds_mode = DS_MODE_OWNER;
+ vnode_t *vp = NULL;
+ char *devpath;
+ size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + strlen(name) + 1;
+ char chrbuf[30], blkbuf[30];
+ int error;
+
+ mutex_enter(&zvol_state_lock);
+
+ if ((zv = zvol_minor_lookup(name)) != NULL) {
+ mutex_exit(&zvol_state_lock);
+ return (EEXIST);
+ }
+
+ if (strchr(name, '@') != 0)
+ ds_mode |= DS_MODE_READONLY;
+
+ error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os);
+
+ if (error) {
+ mutex_exit(&zvol_state_lock);
+ return (error);
+ }
+
+ error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
+
+ if (error) {
+ dmu_objset_close(os);
+ mutex_exit(&zvol_state_lock);
+ return (error);
+ }
+
+ /*
+ * If there's an existing /dev/zvol symlink, try to use the
+ * same minor number we used last time.
+ */
+ devpath = kmem_alloc(devpathlen, KM_SLEEP);
+
+ (void) sprintf(devpath, "%s%s", ZVOL_FULL_DEV_DIR, name);
+
+ error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp);
+
+ kmem_free(devpath, devpathlen);
+
+ if (error == 0 && vp->v_type != VLNK)
+ error = EINVAL;
+
+ if (error == 0) {
+ pn_alloc(&linkpath);
+ error = pn_getsymlink(vp, &linkpath, kcred);
+ if (error == 0) {
+ char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV);
+ if (ms != NULL) {
+ ms += strlen(ZVOL_PSEUDO_DEV);
+ minor = stoi(&ms);
+ }
+ }
+ pn_free(&linkpath);
+ }
+
+ if (vp != NULL)
+ VN_RELE(vp);
+
+ /*
+ * If we found a minor but it's already in use, we must pick a new one.
+ */
+ if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL)
+ minor = 0;
+
+ if (minor == 0)
+ minor = zvol_minor_alloc();
+
+ if (minor == 0) {
+ dmu_objset_close(os);
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+
+ if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) {
+ dmu_objset_close(os);
+ mutex_exit(&zvol_state_lock);
+ return (EAGAIN);
+ }
+
+ (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME,
+ (char *)name);
+
+ (void) sprintf(chrbuf, "%uc,raw", minor);
+
+ if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR,
+ minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
+ ddi_soft_state_free(zvol_state, minor);
+ dmu_objset_close(os);
+ mutex_exit(&zvol_state_lock);
+ return (EAGAIN);
+ }
+
+ (void) sprintf(blkbuf, "%uc", minor);
+
+ if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK,
+ minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
+ ddi_remove_minor_node(zfs_dip, chrbuf);
+ ddi_soft_state_free(zvol_state, minor);
+ dmu_objset_close(os);
+ mutex_exit(&zvol_state_lock);
+ return (EAGAIN);
+ }
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+
+ (void) strcpy(zv->zv_name, name);
+ zv->zv_min_bs = DEV_BSHIFT;
+ zv->zv_minor = minor;
+ zv->zv_volsize = volsize;
+ zv->zv_objset = os;
+ zv->zv_mode = ds_mode;
+ zv->zv_zilog = zil_open(os, zvol_get_data);
+ mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL);
+ avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare,
+ sizeof (rl_t), offsetof(rl_t, r_node));
+ list_create(&zv->zv_extents, sizeof (zvol_extent_t),
+ offsetof(zvol_extent_t, ze_node));
+ /* get and cache the blocksize */
+ error = dmu_object_info(os, ZVOL_OBJ, &doi);
+ ASSERT(error == 0);
+ zv->zv_volblocksize = doi.doi_data_block_size;
+
+ zil_replay(os, zv, &zv->zv_txg_assign, zvol_replay_vector, NULL);
+ zvol_size_changed(zv, maj);
+
+ /* XXX this should handle the possible i/o error */
+ VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset),
+ "readonly", zvol_readonly_changed_cb, zv) == 0);
+
+ zvol_minors++;
+
+ mutex_exit(&zvol_state_lock);
+
+ return (0);
+}
+
+/*
+ * Remove minor node for the specified volume.
+ */
+int
+zvol_remove_minor(const char *name)
+{
+ zvol_state_t *zv;
+ char namebuf[30];
+
+ mutex_enter(&zvol_state_lock);
+
+ if ((zv = zvol_minor_lookup(name)) == NULL) {
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+
+ if (zv->zv_total_opens != 0) {
+ mutex_exit(&zvol_state_lock);
+ return (EBUSY);
+ }
+
+ (void) sprintf(namebuf, "%uc,raw", zv->zv_minor);
+ ddi_remove_minor_node(zfs_dip, namebuf);
+
+ (void) sprintf(namebuf, "%uc", zv->zv_minor);
+ ddi_remove_minor_node(zfs_dip, namebuf);
+
+ VERIFY(dsl_prop_unregister(dmu_objset_ds(zv->zv_objset),
+ "readonly", zvol_readonly_changed_cb, zv) == 0);
+
+ zil_close(zv->zv_zilog);
+ zv->zv_zilog = NULL;
+ dmu_objset_close(zv->zv_objset);
+ zv->zv_objset = NULL;
+ avl_destroy(&zv->zv_znode.z_range_avl);
+ mutex_destroy(&zv->zv_znode.z_range_lock);
+
+ ddi_soft_state_free(zvol_state, zv->zv_minor);
+
+ zvol_minors--;
+
+ mutex_exit(&zvol_state_lock);
+
+ return (0);
+}
+
+int
+zvol_prealloc(zvol_state_t *zv)
+{
+ objset_t *os = zv->zv_objset;
+ dmu_tx_t *tx;
+ uint64_t refd, avail, usedobjs, availobjs;
+ uint64_t resid = zv->zv_volsize;
+ uint64_t off = 0;
+
+ /* Check the space usage before attempting to allocate the space */
+ dmu_objset_space(os, &refd, &avail, &usedobjs, &availobjs);
+ if (avail < zv->zv_volsize)
+ return (ENOSPC);
+
+ /* Free old extents if they exist */
+ zvol_free_extents(zv);
+
+ while (resid != 0) {
+ int error;
+ uint64_t bytes = MIN(resid, SPA_MAXBLOCKSIZE);
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ (void) dmu_free_long_range(os, ZVOL_OBJ, 0, off);
+ return (error);
+ }
+ dmu_prealloc(os, ZVOL_OBJ, off, bytes, tx);
+ dmu_tx_commit(tx);
+ off += bytes;
+ resid -= bytes;
+ }
+ txg_wait_synced(dmu_objset_pool(os), 0);
+
+ return (0);
+}
+
+int
+zvol_update_volsize(zvol_state_t *zv, major_t maj, uint64_t volsize)
+{
+ dmu_tx_t *tx;
+ int error;
+
+ ASSERT(MUTEX_HELD(&zvol_state_lock));
+
+ tx = dmu_tx_create(zv->zv_objset);
+ dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ return (error);
+ }
+
+ error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1,
+ &volsize, tx);
+ dmu_tx_commit(tx);
+
+ if (error == 0)
+ error = dmu_free_long_range(zv->zv_objset,
+ ZVOL_OBJ, volsize, DMU_OBJECT_END);
+
+ /*
+ * If we are using a faked-up state (zv_minor == 0) then don't
+ * try to update the in-core zvol state.
+ */
+ if (error == 0 && zv->zv_minor) {
+ zv->zv_volsize = volsize;
+ zvol_size_changed(zv, maj);
+ }
+ return (error);
+}
+
+int
+zvol_set_volsize(const char *name, major_t maj, uint64_t volsize)
+{
+ zvol_state_t *zv;
+ int error;
+ dmu_object_info_t doi;
+ uint64_t old_volsize = 0ULL;
+ zvol_state_t state = { 0 };
+
+ mutex_enter(&zvol_state_lock);
+
+ if ((zv = zvol_minor_lookup(name)) == NULL) {
+ /*
+ * If we are doing a "zfs clone -o volsize=", then the
+ * minor node won't exist yet.
+ */
+ error = dmu_objset_open(name, DMU_OST_ZVOL, DS_MODE_OWNER,
+ &state.zv_objset);
+ if (error != 0)
+ goto out;
+ zv = &state;
+ }
+ old_volsize = zv->zv_volsize;
+
+ if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 ||
+ (error = zvol_check_volsize(volsize,
+ doi.doi_data_block_size)) != 0)
+ goto out;
+
+ if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY)) {
+ error = EROFS;
+ goto out;
+ }
+
+ error = zvol_update_volsize(zv, maj, volsize);
+
+ /*
+ * Reinitialize the dump area to the new size. If we
+ * failed to resize the dump area then restore the it back to
+ * it's original size.
+ */
+ if (error == 0 && zv->zv_flags & ZVOL_DUMPIFIED) {
+ if ((error = zvol_dumpify(zv)) != 0 ||
+ (error = dumpvp_resize()) != 0) {
+ (void) zvol_update_volsize(zv, maj, old_volsize);
+ error = zvol_dumpify(zv);
+ }
+ }
+
+out:
+ if (state.zv_objset)
+ dmu_objset_close(state.zv_objset);
+
+ mutex_exit(&zvol_state_lock);
+
+ return (error);
+}
+
+int
+zvol_set_volblocksize(const char *name, uint64_t volblocksize)
+{
+ zvol_state_t *zv;
+ dmu_tx_t *tx;
+ int error;
+ boolean_t needlock;
+
+ /*
+ * The lock may already be held if we are being called from
+ * zvol_dump_init().
+ */
+ needlock = !MUTEX_HELD(&zvol_state_lock);
+ if (needlock)
+ mutex_enter(&zvol_state_lock);
+
+ if ((zv = zvol_minor_lookup(name)) == NULL) {
+ if (needlock)
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+ if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY)) {
+ if (needlock)
+ mutex_exit(&zvol_state_lock);
+ return (EROFS);
+ }
+
+ tx = dmu_tx_create(zv->zv_objset);
+ dmu_tx_hold_bonus(tx, ZVOL_OBJ);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ } else {
+ error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ,
+ volblocksize, 0, tx);
+ if (error == ENOTSUP)
+ error = EBUSY;
+ dmu_tx_commit(tx);
+ if (error == 0)
+ zv->zv_volblocksize = volblocksize;
+ }
+
+ if (needlock)
+ mutex_exit(&zvol_state_lock);
+
+ return (error);
+}
+
+/*ARGSUSED*/
+int
+zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr)
+{
+ minor_t minor = getminor(*devp);
+ zvol_state_t *zv;
+
+ if (minor == 0) /* This is the control device */
+ return (0);
+
+ mutex_enter(&zvol_state_lock);
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL) {
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+
+ ASSERT(zv->zv_objset != NULL);
+
+ if ((flag & FWRITE) &&
+ (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY))) {
+ mutex_exit(&zvol_state_lock);
+ return (EROFS);
+ }
+ if (zv->zv_flags & ZVOL_EXCL) {
+ mutex_exit(&zvol_state_lock);
+ return (EBUSY);
+ }
+ if (flag & FEXCL) {
+ if (zv->zv_total_opens != 0) {
+ mutex_exit(&zvol_state_lock);
+ return (EBUSY);
+ }
+ zv->zv_flags |= ZVOL_EXCL;
+ }
+
+ if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) {
+ zv->zv_open_count[otyp]++;
+ zv->zv_total_opens++;
+ }
+
+ mutex_exit(&zvol_state_lock);
+
+ return (0);
+}
+
+/*ARGSUSED*/
+int
+zvol_close(dev_t dev, int flag, int otyp, cred_t *cr)
+{
+ minor_t minor = getminor(dev);
+ zvol_state_t *zv;
+
+ if (minor == 0) /* This is the control device */
+ return (0);
+
+ mutex_enter(&zvol_state_lock);
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL) {
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+
+ if (zv->zv_flags & ZVOL_EXCL) {
+ ASSERT(zv->zv_total_opens == 1);
+ zv->zv_flags &= ~ZVOL_EXCL;
+ }
+
+ /*
+ * If the open count is zero, this is a spurious close.
+ * That indicates a bug in the kernel / DDI framework.
+ */
+ ASSERT(zv->zv_open_count[otyp] != 0);
+ ASSERT(zv->zv_total_opens != 0);
+
+ /*
+ * You may get multiple opens, but only one close.
+ */
+ zv->zv_open_count[otyp]--;
+ zv->zv_total_opens--;
+
+ mutex_exit(&zvol_state_lock);
+
+ return (0);
+}
+
+static void
+zvol_get_done(dmu_buf_t *db, void *vzgd)
+{
+ zgd_t *zgd = (zgd_t *)vzgd;
+ rl_t *rl = zgd->zgd_rl;
+
+ dmu_buf_rele(db, vzgd);
+ zfs_range_unlock(rl);
+ zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
+ kmem_free(zgd, sizeof (zgd_t));
+}
+
+/*
+ * Get data to generate a TX_WRITE intent log record.
+ */
+static int
+zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
+{
+ zvol_state_t *zv = arg;
+ objset_t *os = zv->zv_objset;
+ dmu_buf_t *db;
+ rl_t *rl;
+ zgd_t *zgd;
+ uint64_t boff; /* block starting offset */
+ int dlen = lr->lr_length; /* length of user data */
+ int error;
+
+ ASSERT(zio);
+ ASSERT(dlen != 0);
+
+ /*
+ * Write records come in two flavors: immediate and indirect.
+ * For small writes it's cheaper to store the data with the
+ * log record (immediate); for large writes it's cheaper to
+ * sync the data and get a pointer to it (indirect) so that
+ * we don't have to write the data twice.
+ */
+ if (buf != NULL) /* immediate write */
+ return (dmu_read(os, ZVOL_OBJ, lr->lr_offset, dlen, buf));
+
+ zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
+ zgd->zgd_zilog = zv->zv_zilog;
+ zgd->zgd_bp = &lr->lr_blkptr;
+
+ /*
+ * Lock the range of the block to ensure that when the data is
+ * written out and its checksum is being calculated that no other
+ * thread can change the block.
+ */
+ boff = P2ALIGN_TYPED(lr->lr_offset, zv->zv_volblocksize, uint64_t);
+ rl = zfs_range_lock(&zv->zv_znode, boff, zv->zv_volblocksize,
+ RL_READER);
+ zgd->zgd_rl = rl;
+
+ VERIFY(0 == dmu_buf_hold(os, ZVOL_OBJ, lr->lr_offset, zgd, &db));
+ error = dmu_sync(zio, db, &lr->lr_blkptr,
+ lr->lr_common.lrc_txg, zvol_get_done, zgd);
+ if (error == 0)
+ zil_add_block(zv->zv_zilog, &lr->lr_blkptr);
+ /*
+ * If we get EINPROGRESS, then we need to wait for a
+ * write IO initiated by dmu_sync() to complete before
+ * we can release this dbuf. We will finish everything
+ * up in the zvol_get_done() callback.
+ */
+ if (error == EINPROGRESS)
+ return (0);
+ dmu_buf_rele(db, zgd);
+ zfs_range_unlock(rl);
+ kmem_free(zgd, sizeof (zgd_t));
+ return (error);
+}
+
+/*
+ * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions.
+ *
+ * We store data in the log buffers if it's small enough.
+ * Otherwise we will later flush the data out via dmu_sync().
+ */
+ssize_t zvol_immediate_write_sz = 32768;
+
+static void
+zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t len)
+{
+ uint32_t blocksize = zv->zv_volblocksize;
+ lr_write_t *lr;
+
+ while (len) {
+ ssize_t nbytes = MIN(len, blocksize - P2PHASE(off, blocksize));
+ itx_t *itx = zil_itx_create(TX_WRITE, sizeof (*lr));
+
+ itx->itx_wr_state =
+ len > zvol_immediate_write_sz ? WR_INDIRECT : WR_NEED_COPY;
+ itx->itx_private = zv;
+ lr = (lr_write_t *)&itx->itx_lr;
+ lr->lr_foid = ZVOL_OBJ;
+ lr->lr_offset = off;
+ lr->lr_length = nbytes;
+ lr->lr_blkoff = off - P2ALIGN_TYPED(off, blocksize, uint64_t);
+ BP_ZERO(&lr->lr_blkptr);
+
+ (void) zil_itx_assign(zv->zv_zilog, itx, tx);
+ len -= nbytes;
+ off += nbytes;
+ }
+}
+
+static int
+zvol_dumpio_vdev(vdev_t *vd, void *addr, uint64_t offset, uint64_t size,
+ boolean_t doread, boolean_t isdump)
+{
+ vdev_disk_t *dvd;
+ int c;
+ int numerrors = 0;
+
+ for (c = 0; c < vd->vdev_children; c++) {
+ ASSERT(vd->vdev_ops == &vdev_mirror_ops);
+ int err = zvol_dumpio_vdev(vd->vdev_child[c],
+ addr, offset, size, doread, isdump);
+ if (err != 0) {
+ numerrors++;
+ } else if (doread) {
+ break;
+ }
+ }
+
+ if (!vd->vdev_ops->vdev_op_leaf)
+ return (numerrors < vd->vdev_children ? 0 : EIO);
+
+ if (doread && !vdev_readable(vd))
+ return (EIO);
+ else if (!doread && !vdev_writeable(vd))
+ return (EIO);
+
+ dvd = vd->vdev_tsd;
+ ASSERT3P(dvd, !=, NULL);
+ offset += VDEV_LABEL_START_SIZE;
+
+ if (ddi_in_panic() || isdump) {
+ ASSERT(!doread);
+ if (doread)
+ return (EIO);
+ return (ldi_dump(dvd->vd_lh, addr, lbtodb(offset),
+ lbtodb(size)));
+ } else {
+ return (vdev_disk_physio(dvd->vd_lh, addr, size, offset,
+ doread ? B_READ : B_WRITE));
+ }
+}
+
+static int
+zvol_dumpio(zvol_state_t *zv, void *addr, uint64_t offset, uint64_t size,
+ boolean_t doread, boolean_t isdump)
+{
+ vdev_t *vd;
+ int error;
+ zvol_extent_t *ze;
+ spa_t *spa = dmu_objset_spa(zv->zv_objset);
+
+ /* Must be sector aligned, and not stradle a block boundary. */
+ if (P2PHASE(offset, DEV_BSIZE) || P2PHASE(size, DEV_BSIZE) ||
+ P2BOUNDARY(offset, size, zv->zv_volblocksize)) {
+ return (EINVAL);
+ }
+ ASSERT(size <= zv->zv_volblocksize);
+
+ /* Locate the extent this belongs to */
+ ze = list_head(&zv->zv_extents);
+ while (offset >= ze->ze_nblks * zv->zv_volblocksize) {
+ offset -= ze->ze_nblks * zv->zv_volblocksize;
+ ze = list_next(&zv->zv_extents, ze);
+ }
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ vd = vdev_lookup_top(spa, DVA_GET_VDEV(&ze->ze_dva));
+ offset += DVA_GET_OFFSET(&ze->ze_dva);
+ error = zvol_dumpio_vdev(vd, addr, offset, size, doread, isdump);
+ spa_config_exit(spa, SCL_STATE, FTAG);
+ return (error);
+}
+
+int
+zvol_strategy(buf_t *bp)
+{
+ zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev));
+ uint64_t off, volsize;
+ size_t resid;
+ char *addr;
+ objset_t *os;
+ rl_t *rl;
+ int error = 0;
+ boolean_t doread = bp->b_flags & B_READ;
+ boolean_t is_dump = zv->zv_flags & ZVOL_DUMPIFIED;
+
+ if (zv == NULL) {
+ bioerror(bp, ENXIO);
+ biodone(bp);
+ return (0);
+ }
+
+ if (getminor(bp->b_edev) == 0) {
+ bioerror(bp, EINVAL);
+ biodone(bp);
+ return (0);
+ }
+
+ if (!(bp->b_flags & B_READ) &&
+ (zv->zv_flags & ZVOL_RDONLY ||
+ zv->zv_mode & DS_MODE_READONLY)) {
+ bioerror(bp, EROFS);
+ biodone(bp);
+ return (0);
+ }
+
+ off = ldbtob(bp->b_blkno);
+ volsize = zv->zv_volsize;
+
+ os = zv->zv_objset;
+ ASSERT(os != NULL);
+
+ bp_mapin(bp);
+ addr = bp->b_un.b_addr;
+ resid = bp->b_bcount;
+
+ if (resid > 0 && (off < 0 || off >= volsize)) {
+ bioerror(bp, EIO);
+ biodone(bp);
+ return (0);
+ }
+
+ /*
+ * There must be no buffer changes when doing a dmu_sync() because
+ * we can't change the data whilst calculating the checksum.
+ */
+ rl = zfs_range_lock(&zv->zv_znode, off, resid,
+ doread ? RL_READER : RL_WRITER);
+
+ while (resid != 0 && off < volsize) {
+ size_t size = MIN(resid, zvol_maxphys);
+ if (is_dump) {
+ size = MIN(size, P2END(off, zv->zv_volblocksize) - off);
+ error = zvol_dumpio(zv, addr, off, size,
+ doread, B_FALSE);
+ } else if (doread) {
+ error = dmu_read(os, ZVOL_OBJ, off, size, addr);
+ } else {
+ dmu_tx_t *tx = dmu_tx_create(os);
+ dmu_tx_hold_write(tx, ZVOL_OBJ, off, size);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ } else {
+ dmu_write(os, ZVOL_OBJ, off, size, addr, tx);
+ zvol_log_write(zv, tx, off, size);
+ dmu_tx_commit(tx);
+ }
+ }
+ if (error) {
+ /* convert checksum errors into IO errors */
+ if (error == ECKSUM)
+ error = EIO;
+ break;
+ }
+ off += size;
+ addr += size;
+ resid -= size;
+ }
+ zfs_range_unlock(rl);
+
+ if ((bp->b_resid = resid) == bp->b_bcount)
+ bioerror(bp, off > volsize ? EINVAL : error);
+
+ if (!(bp->b_flags & B_ASYNC) && !doread && !zil_disable && !is_dump)
+ zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ);
+ biodone(bp);
+
+ return (0);
+}
+
+/*
+ * Set the buffer count to the zvol maximum transfer.
+ * Using our own routine instead of the default minphys()
+ * means that for larger writes we write bigger buffers on X86
+ * (128K instead of 56K) and flush the disk write cache less often
+ * (every zvol_maxphys - currently 1MB) instead of minphys (currently
+ * 56K on X86 and 128K on sparc).
+ */
+void
+zvol_minphys(struct buf *bp)
+{
+ if (bp->b_bcount > zvol_maxphys)
+ bp->b_bcount = zvol_maxphys;
+}
+
+int
+zvol_dump(dev_t dev, caddr_t addr, daddr_t blkno, int nblocks)
+{
+ minor_t minor = getminor(dev);
+ zvol_state_t *zv;
+ int error = 0;
+ uint64_t size;
+ uint64_t boff;
+ uint64_t resid;
+
+ if (minor == 0) /* This is the control device */
+ return (ENXIO);
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL)
+ return (ENXIO);
+
+ boff = ldbtob(blkno);
+ resid = ldbtob(nblocks);
+
+ VERIFY3U(boff + resid, <=, zv->zv_volsize);
+
+ while (resid) {
+ size = MIN(resid, P2END(boff, zv->zv_volblocksize) - boff);
+ error = zvol_dumpio(zv, addr, boff, size, B_FALSE, B_TRUE);
+ if (error)
+ break;
+ boff += size;
+ addr += size;
+ resid -= size;
+ }
+
+ return (error);
+}
+
+/*ARGSUSED*/
+int
+zvol_read(dev_t dev, uio_t *uio, cred_t *cr)
+{
+ minor_t minor = getminor(dev);
+ zvol_state_t *zv;
+ uint64_t volsize;
+ rl_t *rl;
+ int error = 0;
+
+ if (minor == 0) /* This is the control device */
+ return (ENXIO);
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL)
+ return (ENXIO);
+
+ volsize = zv->zv_volsize;
+ if (uio->uio_resid > 0 &&
+ (uio->uio_loffset < 0 || uio->uio_loffset >= volsize))
+ return (EIO);
+
+ if (zv->zv_flags & ZVOL_DUMPIFIED) {
+ error = physio(zvol_strategy, NULL, dev, B_READ,
+ zvol_minphys, uio);
+ return (error);
+ }
+
+ rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid,
+ RL_READER);
+ while (uio->uio_resid > 0 && uio->uio_loffset < volsize) {
+ uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1);
+
+ /* don't read past the end */
+ if (bytes > volsize - uio->uio_loffset)
+ bytes = volsize - uio->uio_loffset;
+
+ error = dmu_read_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes);
+ if (error) {
+ /* convert checksum errors into IO errors */
+ if (error == ECKSUM)
+ error = EIO;
+ break;
+ }
+ }
+ zfs_range_unlock(rl);
+ return (error);
+}
+
+/*ARGSUSED*/
+int
+zvol_write(dev_t dev, uio_t *uio, cred_t *cr)
+{
+ minor_t minor = getminor(dev);
+ zvol_state_t *zv;
+ uint64_t volsize;
+ rl_t *rl;
+ int error = 0;
+
+ if (minor == 0) /* This is the control device */
+ return (ENXIO);
+
+ zv = ddi_get_soft_state(zvol_state, minor);
+ if (zv == NULL)
+ return (ENXIO);
+
+ volsize = zv->zv_volsize;
+ if (uio->uio_resid > 0 &&
+ (uio->uio_loffset < 0 || uio->uio_loffset >= volsize))
+ return (EIO);
+
+ if (zv->zv_flags & ZVOL_DUMPIFIED) {
+ error = physio(zvol_strategy, NULL, dev, B_WRITE,
+ zvol_minphys, uio);
+ return (error);
+ }
+
+ rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid,
+ RL_WRITER);
+ while (uio->uio_resid > 0 && uio->uio_loffset < volsize) {
+ uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1);
+ uint64_t off = uio->uio_loffset;
+ dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);
+
+ if (bytes > volsize - off) /* don't write past the end */
+ bytes = volsize - off;
+
+ dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ break;
+ }
+ error = dmu_write_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes, tx);
+ if (error == 0)
+ zvol_log_write(zv, tx, off, bytes);
+ dmu_tx_commit(tx);
+
+ if (error)
+ break;
+ }
+ zfs_range_unlock(rl);
+ return (error);
+}
+
+int
+zvol_getefi(void *arg, int flag, uint64_t vs, uint8_t bs)
+{
+ struct uuid uuid = EFI_RESERVED;
+ efi_gpe_t gpe = { 0 };
+ uint32_t crc;
+ dk_efi_t efi;
+ int length;
+ char *ptr;
+
+ if (ddi_copyin(arg, &efi, sizeof (dk_efi_t), flag))
+ return (EFAULT);
+ ptr = (char *)(uintptr_t)efi.dki_data_64;
+ length = efi.dki_length;
+ /*
+ * Some clients may attempt to request a PMBR for the
+ * zvol. Currently this interface will return EINVAL to
+ * such requests. These requests could be supported by
+ * adding a check for lba == 0 and consing up an appropriate
+ * PMBR.
+ */
+ if (efi.dki_lba < 1 || efi.dki_lba > 2 || length <= 0)
+ return (EINVAL);
+
+ gpe.efi_gpe_StartingLBA = LE_64(34ULL);
+ gpe.efi_gpe_EndingLBA = LE_64((vs >> bs) - 1);
+ UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid);
+
+ if (efi.dki_lba == 1) {
+ efi_gpt_t gpt = { 0 };
+
+ gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE);
+ gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT);
+ gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt));
+ gpt.efi_gpt_MyLBA = LE_64(1ULL);
+ gpt.efi_gpt_FirstUsableLBA = LE_64(34ULL);
+ gpt.efi_gpt_LastUsableLBA = LE_64((vs >> bs) - 1);
+ gpt.efi_gpt_PartitionEntryLBA = LE_64(2ULL);
+ gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1);
+ gpt.efi_gpt_SizeOfPartitionEntry =
+ LE_32(sizeof (efi_gpe_t));
+ CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table);
+ gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc);
+ CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table);
+ gpt.efi_gpt_HeaderCRC32 = LE_32(~crc);
+ if (ddi_copyout(&gpt, ptr, MIN(sizeof (gpt), length),
+ flag))
+ return (EFAULT);
+ ptr += sizeof (gpt);
+ length -= sizeof (gpt);
+ }
+ if (length > 0 && ddi_copyout(&gpe, ptr, MIN(sizeof (gpe),
+ length), flag))
+ return (EFAULT);
+ return (0);
+}
+
+/*
+ * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I).
+ */
+/*ARGSUSED*/
+int
+zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
+{
+ zvol_state_t *zv;
+ struct dk_cinfo dki;
+ struct dk_minfo dkm;
+ struct dk_callback *dkc;
+ int error = 0;
+ rl_t *rl;
+
+ mutex_enter(&zvol_state_lock);
+
+ zv = ddi_get_soft_state(zvol_state, getminor(dev));
+
+ if (zv == NULL) {
+ mutex_exit(&zvol_state_lock);
+ return (ENXIO);
+ }
+
+ switch (cmd) {
+
+ case DKIOCINFO:
+ bzero(&dki, sizeof (dki));
+ (void) strcpy(dki.dki_cname, "zvol");
+ (void) strcpy(dki.dki_dname, "zvol");
+ dki.dki_ctype = DKC_UNKNOWN;
+ dki.dki_maxtransfer = 1 << (SPA_MAXBLOCKSHIFT - zv->zv_min_bs);
+ mutex_exit(&zvol_state_lock);
+ if (ddi_copyout(&dki, (void *)arg, sizeof (dki), flag))
+ error = EFAULT;
+ return (error);
+
+ case DKIOCGMEDIAINFO:
+ bzero(&dkm, sizeof (dkm));
+ dkm.dki_lbsize = 1U << zv->zv_min_bs;
+ dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs;
+ dkm.dki_media_type = DK_UNKNOWN;
+ mutex_exit(&zvol_state_lock);
+ if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag))
+ error = EFAULT;
+ return (error);
+
+ case DKIOCGETEFI:
+ {
+ uint64_t vs = zv->zv_volsize;
+ uint8_t bs = zv->zv_min_bs;
+
+ mutex_exit(&zvol_state_lock);
+ error = zvol_getefi((void *)arg, flag, vs, bs);
+ return (error);
+ }
+
+ case DKIOCFLUSHWRITECACHE:
+ dkc = (struct dk_callback *)arg;
+ zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ);
+ if ((flag & FKIOCTL) && dkc != NULL && dkc->dkc_callback) {
+ (*dkc->dkc_callback)(dkc->dkc_cookie, error);
+ error = 0;
+ }
+ break;
+
+ case DKIOCGGEOM:
+ case DKIOCGVTOC:
+ /*
+ * commands using these (like prtvtoc) expect ENOTSUP
+ * since we're emulating an EFI label
+ */
+ error = ENOTSUP;
+ break;
+
+ case DKIOCDUMPINIT:
+ rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize,
+ RL_WRITER);
+ error = zvol_dumpify(zv);
+ zfs_range_unlock(rl);
+ break;
+
+ case DKIOCDUMPFINI:
+ rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize,
+ RL_WRITER);
+ error = zvol_dump_fini(zv);
+ zfs_range_unlock(rl);
+ break;
+
+ default:
+ error = ENOTTY;
+ break;
+
+ }
+ mutex_exit(&zvol_state_lock);
+ return (error);
+}
+
+int
+zvol_busy(void)
+{
+ return (zvol_minors != 0);
+}
+
+void
+zvol_init(void)
+{
+ VERIFY(ddi_soft_state_init(&zvol_state, sizeof (zvol_state_t), 1) == 0);
+ mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL);
+}
+
+void
+zvol_fini(void)
+{
+ mutex_destroy(&zvol_state_lock);
+ ddi_soft_state_fini(&zvol_state);
+}
+
+static boolean_t
+zvol_is_swap(zvol_state_t *zv)
+{
+ vnode_t *vp;
+ boolean_t ret = B_FALSE;
+ char *devpath;
+ size_t devpathlen;
+ int error;
+
+ devpathlen = strlen(ZVOL_FULL_DEV_DIR) + strlen(zv->zv_name) + 1;
+ devpath = kmem_alloc(devpathlen, KM_SLEEP);
+ (void) sprintf(devpath, "%s%s", ZVOL_FULL_DEV_DIR, zv->zv_name);
+ error = lookupname(devpath, UIO_SYSSPACE, FOLLOW, NULLVPP, &vp);
+ kmem_free(devpath, devpathlen);
+
+ ret = !error && IS_SWAPVP(common_specvp(vp));
+
+ if (vp != NULL)
+ VN_RELE(vp);
+
+ return (ret);
+}
+
+static int
+zvol_dump_init(zvol_state_t *zv, boolean_t resize)
+{
+ dmu_tx_t *tx;
+ int error = 0;
+ objset_t *os = zv->zv_objset;
+ nvlist_t *nv = NULL;
+
+ ASSERT(MUTEX_HELD(&zvol_state_lock));
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ return (error);
+ }
+
+ /*
+ * If we are resizing the dump device then we only need to
+ * update the refreservation to match the newly updated
+ * zvolsize. Otherwise, we save off the original state of the
+ * zvol so that we can restore them if the zvol is ever undumpified.
+ */
+ if (resize) {
+ error = zap_update(os, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1,
+ &zv->zv_volsize, tx);
+ } else {
+ uint64_t checksum, compress, refresrv, vbs;
+
+ error = dsl_prop_get_integer(zv->zv_name,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION), &compress, NULL);
+ error = error ? error : dsl_prop_get_integer(zv->zv_name,
+ zfs_prop_to_name(ZFS_PROP_CHECKSUM), &checksum, NULL);
+ error = error ? error : dsl_prop_get_integer(zv->zv_name,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &refresrv, NULL);
+ error = error ? error : dsl_prop_get_integer(zv->zv_name,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &vbs, NULL);
+
+ error = error ? error : zap_update(os, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1,
+ &compress, tx);
+ error = error ? error : zap_update(os, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum, tx);
+ error = error ? error : zap_update(os, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1,
+ &refresrv, tx);
+ error = error ? error : zap_update(os, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1,
+ &vbs, tx);
+ }
+ dmu_tx_commit(tx);
+
+ /* Truncate the file */
+ if (!error)
+ error = dmu_free_long_range(zv->zv_objset,
+ ZVOL_OBJ, 0, DMU_OBJECT_END);
+
+ if (error)
+ return (error);
+
+ /*
+ * We only need update the zvol's property if we are initializing
+ * the dump area for the first time.
+ */
+ if (!resize) {
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 0) == 0);
+ VERIFY(nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION),
+ ZIO_COMPRESS_OFF) == 0);
+ VERIFY(nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_CHECKSUM),
+ ZIO_CHECKSUM_OFF) == 0);
+ VERIFY(nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
+ SPA_MAXBLOCKSIZE) == 0);
+
+ error = zfs_set_prop_nvlist(zv->zv_name, nv);
+ nvlist_free(nv);
+
+ if (error)
+ return (error);
+ }
+
+ /* Allocate the space for the dump */
+ error = zvol_prealloc(zv);
+ return (error);
+}
+
+static int
+zvol_dumpify(zvol_state_t *zv)
+{
+ int error = 0;
+ uint64_t dumpsize = 0;
+ dmu_tx_t *tx;
+ objset_t *os = zv->zv_objset;
+
+ if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY))
+ return (EROFS);
+
+ /*
+ * We do not support swap devices acting as dump devices.
+ */
+ if (zvol_is_swap(zv))
+ return (ENOTSUP);
+
+ if (zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE,
+ 8, 1, &dumpsize) != 0 || dumpsize != zv->zv_volsize) {
+ boolean_t resize = (dumpsize > 0) ? B_TRUE : B_FALSE;
+
+ if ((error = zvol_dump_init(zv, resize)) != 0) {
+ (void) zvol_dump_fini(zv);
+ return (error);
+ }
+ }
+
+ /*
+ * Build up our lba mapping.
+ */
+ error = zvol_get_lbas(zv);
+ if (error) {
+ (void) zvol_dump_fini(zv);
+ return (error);
+ }
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ (void) zvol_dump_fini(zv);
+ return (error);
+ }
+
+ zv->zv_flags |= ZVOL_DUMPIFIED;
+ error = zap_update(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 8, 1,
+ &zv->zv_volsize, tx);
+ dmu_tx_commit(tx);
+
+ if (error) {
+ (void) zvol_dump_fini(zv);
+ return (error);
+ }
+
+ txg_wait_synced(dmu_objset_pool(os), 0);
+ return (0);
+}
+
+static int
+zvol_dump_fini(zvol_state_t *zv)
+{
+ dmu_tx_t *tx;
+ objset_t *os = zv->zv_objset;
+ nvlist_t *nv;
+ int error = 0;
+ uint64_t checksum, compress, refresrv, vbs;
+
+ /*
+ * Attempt to restore the zvol back to its pre-dumpified state.
+ * This is a best-effort attempt as it's possible that not all
+ * of these properties were initialized during the dumpify process
+ * (i.e. error during zvol_dump_init).
+ */
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ return (error);
+ }
+ (void) zap_remove(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, tx);
+ dmu_tx_commit(tx);
+
+ (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum);
+ (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, &compress);
+ (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &refresrv);
+ (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, &vbs);
+
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ (void) nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum);
+ (void) nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_COMPRESSION), compress);
+ (void) nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_REFRESERVATION), refresrv);
+ (void) nvlist_add_uint64(nv,
+ zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), vbs);
+ (void) zfs_set_prop_nvlist(zv->zv_name, nv);
+ nvlist_free(nv);
+
+ zvol_free_extents(zv);
+ zv->zv_flags &= ~ZVOL_DUMPIFIED;
+ (void) dmu_free_long_range(os, ZVOL_OBJ, 0, DMU_OBJECT_END);
+
+ return (0);
+}