diff options
author | Brian Behlendorf <[email protected]> | 2010-05-28 13:45:14 -0700 |
---|---|---|
committer | Brian Behlendorf <[email protected]> | 2010-05-28 13:45:14 -0700 |
commit | 428870ff734fdaccc342b33fc53cf94724409a46 (patch) | |
tree | 164e83c0ceda52a843795ed7cd9e95637d02c177 /module/zfs/spa.c | |
parent | 6119cb885a976e175a6e827894accf657ff1984f (diff) |
Update core ZFS code from build 121 to build 141.
Diffstat (limited to 'module/zfs/spa.c')
-rw-r--r-- | module/zfs/spa.c | 2134 |
1 files changed, 1630 insertions, 504 deletions
diff --git a/module/zfs/spa.c b/module/zfs/spa.c index d7ed23e6f..d7c5de0d3 100644 --- a/module/zfs/spa.c +++ b/module/zfs/spa.c @@ -20,8 +20,7 @@ */ /* - * Copyright 2009 Sun Microsystems, Inc. All rights reserved. - * Use is subject to license terms. + * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. */ /* @@ -35,13 +34,14 @@ #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/ddt.h> #include <sys/vdev_impl.h> #include <sys/metaslab.h> +#include <sys/metaslab_impl.h> #include <sys/uberblock_impl.h> #include <sys/txg.h> #include <sys/avl.h> @@ -57,57 +57,78 @@ #include <sys/arc.h> #include <sys/callb.h> #include <sys/systeminfo.h> -#include <sys/sunddi.h> #include <sys/spa_boot.h> #include <sys/zfs_ioctl.h> +#include <sys/dsl_scan.h> #ifdef _KERNEL +#include <sys/bootprops.h> +#include <sys/callb.h> +#include <sys/cpupart.h> +#include <sys/pool.h> +#include <sys/sysdc.h> #include <sys/zone.h> #endif /* _KERNEL */ #include "zfs_prop.h" #include "zfs_comutil.h" -enum zti_modes { +typedef enum zti_modes { zti_mode_fixed, /* value is # of threads (min 1) */ zti_mode_online_percent, /* value is % of online CPUs */ - zti_mode_tune, /* fill from zio_taskq_tune_* */ + zti_mode_batch, /* cpu-intensive; value is ignored */ + zti_mode_null, /* don't create a taskq */ zti_nmodes -}; +} zti_modes_t; -#define ZTI_THREAD_FIX(n) { zti_mode_fixed, (n) } -#define ZTI_THREAD_PCT(n) { zti_mode_online_percent, (n) } -#define ZTI_THREAD_TUNE { zti_mode_tune, 0 } +#define ZTI_FIX(n) { zti_mode_fixed, (n) } +#define ZTI_PCT(n) { zti_mode_online_percent, (n) } +#define ZTI_BATCH { zti_mode_batch, 0 } +#define ZTI_NULL { zti_mode_null, 0 } -#define ZTI_THREAD_ONE ZTI_THREAD_FIX(1) +#define ZTI_ONE ZTI_FIX(1) typedef struct zio_taskq_info { - const char *zti_name; - struct { - enum zti_modes zti_mode; - uint_t zti_value; - } zti_nthreads[ZIO_TASKQ_TYPES]; + enum zti_modes zti_mode; + uint_t zti_value; } zio_taskq_info_t; static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = { - "issue", "intr" + "issue", "issue_high", "intr", "intr_high" }; -const zio_taskq_info_t zio_taskqs[ZIO_TYPES] = { - /* ISSUE INTR */ - { "spa_zio_null", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } }, - { "spa_zio_read", { ZTI_THREAD_FIX(8), ZTI_THREAD_TUNE } }, - { "spa_zio_write", { ZTI_THREAD_TUNE, ZTI_THREAD_FIX(8) } }, - { "spa_zio_free", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } }, - { "spa_zio_claim", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } }, - { "spa_zio_ioctl", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } }, +/* + * Define the taskq threads for the following I/O types: + * NULL, READ, WRITE, FREE, CLAIM, and IOCTL + */ +const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = { + /* ISSUE ISSUE_HIGH INTR INTR_HIGH */ + { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, + { ZTI_FIX(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL }, + { ZTI_BATCH, ZTI_FIX(5), ZTI_FIX(8), ZTI_FIX(5) }, + { ZTI_FIX(100), ZTI_NULL, ZTI_ONE, ZTI_NULL }, + { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, + { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, }; -enum zti_modes zio_taskq_tune_mode = zti_mode_online_percent; -uint_t zio_taskq_tune_value = 80; /* #threads = 80% of # online CPUs */ - -static void spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx); +static dsl_syncfunc_t spa_sync_props; static boolean_t spa_has_active_shared_spare(spa_t *spa); +static int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config, + spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, + char **ereport); + +uint_t zio_taskq_batch_pct = 100; /* 1 thread per cpu in pset */ +id_t zio_taskq_psrset_bind = PS_NONE; +boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */ +uint_t zio_taskq_basedc = 80; /* base duty cycle */ + +boolean_t spa_create_process = B_TRUE; /* no process ==> no sysdc */ + +/* + * 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" /* * ========================================================================== @@ -144,7 +165,7 @@ static void spa_prop_get_config(spa_t *spa, nvlist_t **nvp) { uint64_t size; - uint64_t used; + uint64_t alloc; uint64_t cap, version; zprop_source_t src = ZPROP_SRC_NONE; spa_config_dirent_t *dp; @@ -152,17 +173,20 @@ spa_prop_get_config(spa_t *spa, nvlist_t **nvp) ASSERT(MUTEX_HELD(&spa->spa_props_lock)); if (spa->spa_root_vdev != NULL) { - size = spa_get_space(spa); - used = spa_get_alloc(spa); + alloc = metaslab_class_get_alloc(spa_normal_class(spa)); + size = metaslab_class_get_space(spa_normal_class(spa)); 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); + spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src); + spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL, + size - alloc, src); - cap = (size == 0) ? 0 : (used * 100 / size); + cap = (size == 0) ? 0 : (alloc * 100 / size); spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src); + spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL, + ddt_get_pool_dedup_ratio(spa), src); + spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL, spa->spa_root_vdev->vdev_state, src); @@ -197,9 +221,9 @@ spa_prop_get_config(spa_t *spa, nvlist_t **nvp) int spa_prop_get(spa_t *spa, nvlist_t **nvp) { + objset_t *mos = spa->spa_meta_objset; 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); @@ -212,7 +236,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) spa_prop_get_config(spa, nvp); /* If no pool property object, no more prop to get. */ - if (spa->spa_pool_props_object == 0) { + if (mos == NULL || spa->spa_pool_props_object == 0) { mutex_exit(&spa->spa_props_lock); return (0); } @@ -371,12 +395,14 @@ spa_prop_validate(spa_t *spa, nvlist_t *props) break; } - if (error = dmu_objset_open(strval, DMU_OST_ZFS, - DS_MODE_USER | DS_MODE_READONLY, &os)) + if (error = dmu_objset_hold(strval, FTAG, &os)) break; - /* We don't support gzip bootable datasets */ - if ((error = dsl_prop_get_integer(strval, + /* Must be ZPL and not gzip compressed. */ + + if (dmu_objset_type(os) != DMU_OST_ZFS) { + error = ENOTSUP; + } else if ((error = dsl_prop_get_integer(strval, zfs_prop_to_name(ZFS_PROP_COMPRESSION), &compress, NULL)) == 0 && !BOOTFS_COMPRESS_VALID(compress)) { @@ -384,7 +410,7 @@ spa_prop_validate(spa_t *spa, nvlist_t *props) } else { objnum = dmu_objset_id(os); } - dmu_objset_close(os); + dmu_objset_rele(os, FTAG); } break; @@ -432,6 +458,16 @@ spa_prop_validate(spa_t *spa, nvlist_t *props) strcmp(slash, "/..") == 0) error = EINVAL; break; + + case ZPOOL_PROP_DEDUPDITTO: + if (spa_version(spa) < SPA_VERSION_DEDUP) + error = ENOTSUP; + else + error = nvpair_value_uint64(elem, &intval); + if (error == 0 && + intval != 0 && intval < ZIO_DEDUPDITTO_MIN) + error = EINVAL; + break; } if (error) @@ -565,63 +601,185 @@ spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub) offsetof(spa_error_entry_t, se_avl)); } -/* - * Activate an uninitialized pool. - */ -static void -spa_activate(spa_t *spa, int mode) +static taskq_t * +spa_taskq_create(spa_t *spa, const char *name, enum zti_modes mode, + uint_t value) { - ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); + uint_t flags = TASKQ_PREPOPULATE; + boolean_t batch = B_FALSE; - spa->spa_state = POOL_STATE_ACTIVE; - spa->spa_mode = mode; + switch (mode) { + case zti_mode_null: + return (NULL); /* no taskq needed */ - spa->spa_normal_class = metaslab_class_create(zfs_metaslab_ops); - spa->spa_log_class = metaslab_class_create(zfs_metaslab_ops); + case zti_mode_fixed: + ASSERT3U(value, >=, 1); + value = MAX(value, 1); + break; + case zti_mode_batch: + batch = B_TRUE; + flags |= TASKQ_THREADS_CPU_PCT; + value = zio_taskq_batch_pct; + break; + + case zti_mode_online_percent: + flags |= TASKQ_THREADS_CPU_PCT; + break; + + default: + panic("unrecognized mode for %s taskq (%u:%u) in " + "spa_activate()", + name, mode, value); + break; + } + + if (zio_taskq_sysdc && spa->spa_proc != &p0) { + if (batch) + flags |= TASKQ_DC_BATCH; + + return (taskq_create_sysdc(name, value, 50, INT_MAX, + spa->spa_proc, zio_taskq_basedc, flags)); + } + return (taskq_create_proc(name, value, maxclsyspri, 50, INT_MAX, + spa->spa_proc, flags)); +} + +static void +spa_create_zio_taskqs(spa_t *spa) +{ for (int t = 0; t < ZIO_TYPES; t++) { - const zio_taskq_info_t *ztip = &zio_taskqs[t]; for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { - enum zti_modes mode = ztip->zti_nthreads[q].zti_mode; - uint_t value = ztip->zti_nthreads[q].zti_value; + const zio_taskq_info_t *ztip = &zio_taskqs[t][q]; + enum zti_modes mode = ztip->zti_mode; + uint_t value = ztip->zti_value; char name[32]; (void) snprintf(name, sizeof (name), - "%s_%s", ztip->zti_name, zio_taskq_types[q]); + "%s_%s", zio_type_name[t], zio_taskq_types[q]); - if (mode == zti_mode_tune) { - mode = zio_taskq_tune_mode; - value = zio_taskq_tune_value; - if (mode == zti_mode_tune) - mode = zti_mode_online_percent; - } + spa->spa_zio_taskq[t][q] = + spa_taskq_create(spa, name, mode, value); + } + } +} + +#ifdef _KERNEL +static void +spa_thread(void *arg) +{ + callb_cpr_t cprinfo; - switch (mode) { - case zti_mode_fixed: - ASSERT3U(value, >=, 1); - value = MAX(value, 1); + spa_t *spa = arg; + user_t *pu = PTOU(curproc); - spa->spa_zio_taskq[t][q] = taskq_create(name, - value, maxclsyspri, 50, INT_MAX, - TASKQ_PREPOPULATE); - break; + CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr, + spa->spa_name); - case zti_mode_online_percent: - spa->spa_zio_taskq[t][q] = taskq_create(name, - value, maxclsyspri, 50, INT_MAX, - TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT); - break; + ASSERT(curproc != &p0); + (void) snprintf(pu->u_psargs, sizeof (pu->u_psargs), + "zpool-%s", spa->spa_name); + (void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm)); - case zti_mode_tune: - default: - panic("unrecognized mode for " - "zio_taskqs[%u]->zti_nthreads[%u] (%u:%u) " - "in spa_activate()", - t, q, mode, value); - break; + /* bind this thread to the requested psrset */ + if (zio_taskq_psrset_bind != PS_NONE) { + pool_lock(); + mutex_enter(&cpu_lock); + mutex_enter(&pidlock); + mutex_enter(&curproc->p_lock); + + if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind, + 0, NULL, NULL) == 0) { + curthread->t_bind_pset = zio_taskq_psrset_bind; + } else { + cmn_err(CE_WARN, + "Couldn't bind process for zfs pool \"%s\" to " + "pset %d\n", spa->spa_name, zio_taskq_psrset_bind); + } + + mutex_exit(&curproc->p_lock); + mutex_exit(&pidlock); + mutex_exit(&cpu_lock); + pool_unlock(); + } + + if (zio_taskq_sysdc) { + sysdc_thread_enter(curthread, 100, 0); + } + + spa->spa_proc = curproc; + spa->spa_did = curthread->t_did; + + spa_create_zio_taskqs(spa); + + mutex_enter(&spa->spa_proc_lock); + ASSERT(spa->spa_proc_state == SPA_PROC_CREATED); + + spa->spa_proc_state = SPA_PROC_ACTIVE; + cv_broadcast(&spa->spa_proc_cv); + + CALLB_CPR_SAFE_BEGIN(&cprinfo); + while (spa->spa_proc_state == SPA_PROC_ACTIVE) + cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); + CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock); + + ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE); + spa->spa_proc_state = SPA_PROC_GONE; + spa->spa_proc = &p0; + cv_broadcast(&spa->spa_proc_cv); + CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */ + + mutex_enter(&curproc->p_lock); + lwp_exit(); +} +#endif + +/* + * Activate an uninitialized pool. + */ +static void +spa_activate(spa_t *spa, int mode) +{ + ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); + + spa->spa_state = POOL_STATE_ACTIVE; + spa->spa_mode = mode; + + spa->spa_normal_class = metaslab_class_create(spa, zfs_metaslab_ops); + spa->spa_log_class = metaslab_class_create(spa, zfs_metaslab_ops); + + /* Try to create a covering process */ + mutex_enter(&spa->spa_proc_lock); + ASSERT(spa->spa_proc_state == SPA_PROC_NONE); + ASSERT(spa->spa_proc == &p0); + spa->spa_did = 0; + + /* Only create a process if we're going to be around a while. */ + if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) { + if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri, + NULL, 0) == 0) { + spa->spa_proc_state = SPA_PROC_CREATED; + while (spa->spa_proc_state == SPA_PROC_CREATED) { + cv_wait(&spa->spa_proc_cv, + &spa->spa_proc_lock); } + ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); + ASSERT(spa->spa_proc != &p0); + ASSERT(spa->spa_did != 0); + } else { +#ifdef _KERNEL + cmn_err(CE_WARN, + "Couldn't create process for zfs pool \"%s\"\n", + spa->spa_name); +#endif } } + mutex_exit(&spa->spa_proc_lock); + + /* If we didn't create a process, we need to create our taskqs. */ + if (spa->spa_proc == &p0) { + spa_create_zio_taskqs(spa); + } list_create(&spa->spa_config_dirty_list, sizeof (vdev_t), offsetof(vdev_t, vdev_config_dirty_node)); @@ -658,7 +816,8 @@ spa_deactivate(spa_t *spa) 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]); + if (spa->spa_zio_taskq[t][q] != NULL) + taskq_destroy(spa->spa_zio_taskq[t][q]); spa->spa_zio_taskq[t][q] = NULL; } } @@ -679,6 +838,31 @@ spa_deactivate(spa_t *spa) avl_destroy(&spa->spa_errlist_last); spa->spa_state = POOL_STATE_UNINITIALIZED; + + mutex_enter(&spa->spa_proc_lock); + if (spa->spa_proc_state != SPA_PROC_NONE) { + ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE); + spa->spa_proc_state = SPA_PROC_DEACTIVATE; + cv_broadcast(&spa->spa_proc_cv); + while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) { + ASSERT(spa->spa_proc != &p0); + cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock); + } + ASSERT(spa->spa_proc_state == SPA_PROC_GONE); + spa->spa_proc_state = SPA_PROC_NONE; + } + ASSERT(spa->spa_proc == &p0); + mutex_exit(&spa->spa_proc_lock); + + /* + * We want to make sure spa_thread() has actually exited the ZFS + * module, so that the module can't be unloaded out from underneath + * it. + */ + if (spa->spa_did != 0) { + thread_join(spa->spa_did); + spa->spa_did = 0; + } } /* @@ -759,14 +943,19 @@ spa_unload(spa_t *spa) spa->spa_async_zio_root = NULL; } + bpobj_close(&spa->spa_deferred_bpobj); + /* * Close the dsl pool. */ if (spa->spa_dsl_pool) { dsl_pool_close(spa->spa_dsl_pool); spa->spa_dsl_pool = NULL; + spa->spa_meta_objset = NULL; } + ddt_unload(spa); + spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* @@ -919,7 +1108,7 @@ spa_load_spares(spa_t *spa) 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); + spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE); 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++) @@ -1045,7 +1234,7 @@ spa_load_l2cache(spa_t *spa) 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); + sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE); VERIFY(nvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0); out: @@ -1099,26 +1288,23 @@ spa_check_removed(vdev_t *vd) * that the label does not contain the most up-to-date information. */ void -spa_load_log_state(spa_t *spa) +spa_load_log_state(spa_t *spa, nvlist_t *nv) { - nvlist_t *nv, *nvroot, **child; - uint64_t is_log; - uint_t children; - vdev_t *rvd = spa->spa_root_vdev; - - VERIFY(load_nvlist(spa, spa->spa_config_object, &nv) == 0); - VERIFY(nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); - VERIFY(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, - &child, &children) == 0); + vdev_t *ovd, *rvd = spa->spa_root_vdev; - for (int c = 0; c < children; c++) { - vdev_t *tvd = rvd->vdev_child[c]; + /* + * Load the original root vdev tree from the passed config. + */ + spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); + VERIFY(spa_config_parse(spa, &ovd, nv, NULL, 0, VDEV_ALLOC_LOAD) == 0); - if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, - &is_log) == 0 && is_log) - vdev_load_log_state(tvd, child[c]); + for (int c = 0; c < rvd->vdev_children; c++) { + vdev_t *cvd = rvd->vdev_child[c]; + if (cvd->vdev_islog) + vdev_load_log_state(cvd, ovd->vdev_child[c]); } - nvlist_free(nv); + vdev_free(ovd); + spa_config_exit(spa, SCL_ALL, FTAG); } /* @@ -1133,7 +1319,7 @@ spa_check_logs(spa_t *spa) 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; + spa_set_log_state(spa, SPA_LOG_MISSING); return (1); } break; @@ -1141,23 +1327,344 @@ spa_check_logs(spa_t *spa) return (0); } +static boolean_t +spa_passivate_log(spa_t *spa) +{ + vdev_t *rvd = spa->spa_root_vdev; + boolean_t slog_found = B_FALSE; + + ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); + + if (!spa_has_slogs(spa)) + return (B_FALSE); + + for (int c = 0; c < rvd->vdev_children; c++) { + vdev_t *tvd = rvd->vdev_child[c]; + metaslab_group_t *mg = tvd->vdev_mg; + + if (tvd->vdev_islog) { + metaslab_group_passivate(mg); + slog_found = B_TRUE; + } + } + + return (slog_found); +} + +static void +spa_activate_log(spa_t *spa) +{ + vdev_t *rvd = spa->spa_root_vdev; + + ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER)); + + for (int c = 0; c < rvd->vdev_children; c++) { + vdev_t *tvd = rvd->vdev_child[c]; + metaslab_group_t *mg = tvd->vdev_mg; + + if (tvd->vdev_islog) + metaslab_group_activate(mg); + } +} + +int +spa_offline_log(spa_t *spa) +{ + int error = 0; + + if ((error = dmu_objset_find(spa_name(spa), zil_vdev_offline, + NULL, DS_FIND_CHILDREN)) == 0) { + + /* + * We successfully offlined the log device, sync out the + * current txg so that the "stubby" block can be removed + * by zil_sync(). + */ + txg_wait_synced(spa->spa_dsl_pool, 0); + } + return (error); +} + +static void +spa_aux_check_removed(spa_aux_vdev_t *sav) +{ + for (int i = 0; i < sav->sav_count; i++) + spa_check_removed(sav->sav_vdevs[i]); +} + +void +spa_claim_notify(zio_t *zio) +{ + spa_t *spa = zio->io_spa; + + if (zio->io_error) + return; + + mutex_enter(&spa->spa_props_lock); /* any mutex will do */ + if (spa->spa_claim_max_txg < zio->io_bp->blk_birth) + spa->spa_claim_max_txg = zio->io_bp->blk_birth; + mutex_exit(&spa->spa_props_lock); +} + +typedef struct spa_load_error { + uint64_t sle_meta_count; + uint64_t sle_data_count; +} spa_load_error_t; + +static void +spa_load_verify_done(zio_t *zio) +{ + blkptr_t *bp = zio->io_bp; + spa_load_error_t *sle = zio->io_private; + dmu_object_type_t type = BP_GET_TYPE(bp); + int error = zio->io_error; + + if (error) { + if ((BP_GET_LEVEL(bp) != 0 || dmu_ot[type].ot_metadata) && + type != DMU_OT_INTENT_LOG) + atomic_add_64(&sle->sle_meta_count, 1); + else + atomic_add_64(&sle->sle_data_count, 1); + } + zio_data_buf_free(zio->io_data, zio->io_size); +} + +/*ARGSUSED*/ +static int +spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, + arc_buf_t *pbuf, const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg) +{ + if (bp != NULL) { + zio_t *rio = arg; + size_t size = BP_GET_PSIZE(bp); + void *data = zio_data_buf_alloc(size); + + zio_nowait(zio_read(rio, spa, bp, data, size, + spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB, + ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL | + ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb)); + } + return (0); +} + +static int +spa_load_verify(spa_t *spa) +{ + zio_t *rio; + spa_load_error_t sle = { 0 }; + zpool_rewind_policy_t policy; + boolean_t verify_ok = B_FALSE; + int error; + + zpool_get_rewind_policy(spa->spa_config, &policy); + + if (policy.zrp_request & ZPOOL_NEVER_REWIND) + return (0); + + rio = zio_root(spa, NULL, &sle, + ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); + + error = traverse_pool(spa, spa->spa_verify_min_txg, + TRAVERSE_PRE | TRAVERSE_PREFETCH, spa_load_verify_cb, rio); + + (void) zio_wait(rio); + + spa->spa_load_meta_errors = sle.sle_meta_count; + spa->spa_load_data_errors = sle.sle_data_count; + + if (!error && sle.sle_meta_count <= policy.zrp_maxmeta && + sle.sle_data_count <= policy.zrp_maxdata) { + verify_ok = B_TRUE; + spa->spa_load_txg = spa->spa_uberblock.ub_txg; + spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp; + } else { + spa->spa_load_max_txg = spa->spa_uberblock.ub_txg; + } + + if (error) { + if (error != ENXIO && error != EIO) + error = EIO; + return (error); + } + + return (verify_ok ? 0 : EIO); +} + +/* + * Find a value in the pool props object. + */ +static void +spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val) +{ + (void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object, + zpool_prop_to_name(prop), sizeof (uint64_t), 1, val); +} + +/* + * Find a value in the pool directory object. + */ +static int +spa_dir_prop(spa_t *spa, const char *name, uint64_t *val) +{ + return (zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + name, sizeof (uint64_t), 1, val)); +} + +static int +spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err) +{ + vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux); + return (err); +} + +/* + * Fix up config after a partly-completed split. This is done with the + * ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off + * pool have that entry in their config, but only the splitting one contains + * a list of all the guids of the vdevs that are being split off. + * + * This function determines what to do with that list: either rejoin + * all the disks to the pool, or complete the splitting process. To attempt + * the rejoin, each disk that is offlined is marked online again, and + * we do a reopen() call. If the vdev label for every disk that was + * marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL) + * then we call vdev_split() on each disk, and complete the split. + * + * Otherwise we leave the config alone, with all the vdevs in place in + * the original pool. + */ +static void +spa_try_repair(spa_t *spa, nvlist_t *config) +{ + uint_t extracted; + uint64_t *glist; + uint_t i, gcount; + nvlist_t *nvl; + vdev_t **vd; + boolean_t attempt_reopen; + + if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0) + return; + + /* check that the config is complete */ + if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, + &glist, &gcount) != 0) + return; + + vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP); + + /* attempt to online all the vdevs & validate */ + attempt_reopen = B_TRUE; + for (i = 0; i < gcount; i++) { + if (glist[i] == 0) /* vdev is hole */ + continue; + + vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE); + if (vd[i] == NULL) { + /* + * Don't bother attempting to reopen the disks; + * just do the split. + */ + attempt_reopen = B_FALSE; + } else { + /* attempt to re-online it */ + vd[i]->vdev_offline = B_FALSE; + } + } + + if (attempt_reopen) { + vdev_reopen(spa->spa_root_vdev); + + /* check each device to see what state it's in */ + for (extracted = 0, i = 0; i < gcount; i++) { + if (vd[i] != NULL && + vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL) + break; + ++extracted; + } + } + + /* + * If every disk has been moved to the new pool, or if we never + * even attempted to look at them, then we split them off for + * good. + */ + if (!attempt_reopen || gcount == extracted) { + for (i = 0; i < gcount; i++) + if (vd[i] != NULL) + vdev_split(vd[i]); + vdev_reopen(spa->spa_root_vdev); + } + + kmem_free(vd, gcount * sizeof (vdev_t *)); +} + +static int +spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type, + boolean_t mosconfig) +{ + nvlist_t *config = spa->spa_config; + char *ereport = FM_EREPORT_ZFS_POOL; + int error; + uint64_t pool_guid; + nvlist_t *nvl; + + if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) + return (EINVAL); + + /* + * 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, + &spa->spa_ubsync.ub_version) != 0) + spa->spa_ubsync.ub_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; + } else { + spa->spa_load_guid = pool_guid; + + if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, + &nvl) == 0) { + VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting, + KM_SLEEP) == 0); + } + + error = spa_load_impl(spa, pool_guid, config, state, type, + mosconfig, &ereport); + } + + 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 = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE; + spa->spa_ena = 0; + + return (error); +} + /* * 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) +spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config, + spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig, + char **ereport) { 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; int orig_mode = spa->spa_mode; - char *ereport = FM_EREPORT_ZFS_POOL; + int parse; + uint64_t obj; /* * If this is an untrusted config, access the pool in read-only mode. @@ -1170,29 +1677,11 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) 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; - } + if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot)) + return (EINVAL); - spa->spa_load_guid = pool_guid; + parse = (type == SPA_IMPORT_EXISTING ? + VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT); /* * Create "The Godfather" zio to hold all async IOs @@ -1206,15 +1695,17 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) * 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); + error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, parse); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) - goto out; + return (error); ASSERT(spa->spa_root_vdev == rvd); - ASSERT(spa_guid(spa) == pool_guid); + + if (type != SPA_IMPORT_ASSEMBLE) { + ASSERT(spa_guid(spa) == pool_guid); + } /* * Try to open all vdevs, loading each label in the process. @@ -1223,26 +1714,31 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) error = vdev_open(rvd); spa_config_exit(spa, SCL_ALL, FTAG); if (error != 0) - goto out; + return (error); /* * We need to validate the vdev labels against the configuration that * we have in hand, which is dependent on the setting of mosconfig. If * mosconfig is true then we're validating the vdev labels based on - * that config. Otherwise, we're validating against the cached config + * that config. Otherwise, we're validating against the cached config * (zpool.cache) that was read when we loaded the zfs module, and then * later we will recursively call spa_load() and validate against * the vdev config. + * + * If we're assembling a new pool that's been split off from an + * existing pool, the labels haven't yet been updated so we skip + * validation for now. */ - 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 (type != SPA_IMPORT_ASSEMBLE) { + spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); + error = vdev_validate(rvd); + spa_config_exit(spa, SCL_ALL, FTAG); - if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) { - error = ENXIO; - goto out; + if (error != 0) + return (error); + + if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) + return (ENXIO); } /* @@ -1253,32 +1749,29 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) /* * 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 (ub->ub_txg == 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO)); /* * 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 (ub->ub_version > SPA_VERSION) + return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP)); /* * 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; + if (mosconfig && type != SPA_IMPORT_ASSEMBLE && + rvd->vdev_guid_sum != ub->ub_guid_sum) + return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO)); + + if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) { + spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); + spa_try_repair(spa, config); + spa_config_exit(spa, SCL_ALL, FTAG); + nvlist_free(spa->spa_config_splitting); + spa->spa_config_splitting = NULL; } /* @@ -1286,41 +1779,34 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) */ spa->spa_state = POOL_STATE_ACTIVE; spa->spa_ubsync = spa->spa_uberblock; - spa->spa_first_txg = spa_last_synced_txg(spa) + 1; + spa->spa_verify_min_txg = spa->spa_extreme_rewind ? + TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1; + spa->spa_first_txg = spa->spa_last_ubsync_txg ? + spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1; + spa->spa_claim_max_txg = spa->spa_first_txg; + spa->spa_prev_software_version = ub->ub_software_version; + 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; - } + if (error) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); 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 (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); if (!mosconfig) { - nvlist_t *newconfig; uint64_t hostid; + nvlist_t *policy = NULL, *nvconfig; - 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 (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); - if (!spa_is_root(spa) && nvlist_lookup_uint64(newconfig, + if (!spa_is_root(spa) && nvlist_lookup_uint64(nvconfig, ZPOOL_CONFIG_HOSTID, &hostid) == 0) { char *hostname; unsigned long myhostid = 0; - VERIFY(nvlist_lookup_string(newconfig, + VERIFY(nvlist_lookup_string(nvconfig, ZPOOL_CONFIG_HOSTNAME, &hostname) == 0); #ifdef _KERNEL @@ -1334,185 +1820,133 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) #endif /* _KERNEL */ if (hostid != 0 && myhostid != 0 && hostid != myhostid) { + nvlist_free(nvconfig); 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; + return (EBADF); } } + if (nvlist_lookup_nvlist(spa->spa_config, + ZPOOL_REWIND_POLICY, &policy) == 0) + VERIFY(nvlist_add_nvlist(nvconfig, + ZPOOL_REWIND_POLICY, policy) == 0); - spa_config_set(spa, newconfig); + spa_config_set(spa, nvconfig); spa_unload(spa); spa_deactivate(spa); spa_activate(spa, orig_mode); - return (spa_load(spa, newconfig, state, B_TRUE)); + return (spa_load(spa, state, SPA_IMPORT_EXISTING, 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; - } + if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj); + if (error != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * 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; - } + error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + + error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION, + &spa->spa_creation_version); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * 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 = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); - 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; - } + error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB, + &spa->spa_errlog_scrub); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); /* * 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; - } + error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + + /* + * If we're assembling the pool from the split-off vdevs of + * an existing pool, we don't want to attach the spares & cache + * devices. + */ /* * 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) { + error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object); + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { 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->spa_spares.sav_config) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_spares(spa); spa_config_exit(spa, SCL_ALL, FTAG); + } else if (error == 0) { + spa->spa_spares.sav_sync = B_TRUE; } /* * 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, + error = spa_dir_prop(spa, DMU_POOL_L2CACHE, &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) { + if (error != 0 && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + if (error == 0 && type != SPA_IMPORT_ASSEMBLE) { 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->spa_l2cache.sav_config) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa_load_l2cache(spa); spa_config_exit(spa, SCL_ALL, FTAG); + } else if (error == 0) { + spa->spa_l2cache.sav_sync = B_TRUE; } - spa_load_log_state(spa); - - 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; - } + error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object); + if (error && error != ENOENT) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); 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); - (void) zap_lookup(spa->spa_meta_objset, - spa->spa_pool_props_object, - zpool_prop_to_name(ZPOOL_PROP_AUTOEXPAND), - sizeof (uint64_t), 1, &spa->spa_autoexpand); + uint64_t autoreplace; + + spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs); + spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace); + spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation); + spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode); + spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand); + spa_prop_find(spa, ZPOOL_PROP_DEDUPDITTO, + &spa->spa_dedup_ditto); + + spa->spa_autoreplace = (autoreplace != 0); } /* @@ -1522,8 +1956,18 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) * unopenable vdevs so that the normal autoreplace handler can take * over. */ - if (autoreplace && state != SPA_LOAD_TRYIMPORT) + if (spa->spa_autoreplace && state != SPA_LOAD_TRYIMPORT) { spa_check_removed(spa->spa_root_vdev); + /* + * For the import case, this is done in spa_import(), because + * at this point we're using the spare definitions from + * the MOS config, not necessarily from the userland config. + */ + if (state != SPA_LOAD_IMPORT) { + spa_aux_check_removed(&spa->spa_spares); + spa_aux_check_removed(&spa->spa_l2cache); + } + } /* * Load the vdev state for all toplevel vdevs. @@ -1541,12 +1985,48 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) * 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 (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) + return (ENXIO); + + /* + * Load the DDTs (dedup tables). + */ + error = ddt_load(spa); + if (error != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + + spa_update_dspace(spa); + + if (state != SPA_LOAD_TRYIMPORT) { + error = spa_load_verify(spa); + if (error) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, + error)); } - if (spa_writeable(spa)) { + /* + * Load the intent log state and check log integrity. If we're + * assembling a pool from a split, the log is not transferred over. + */ + if (type != SPA_IMPORT_ASSEMBLE) { + nvlist_t *nvconfig; + + if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0) + return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + + VERIFY(nvlist_lookup_nvlist(nvconfig, ZPOOL_CONFIG_VDEV_TREE, + &nvroot) == 0); + spa_load_log_state(spa, nvroot); + nvlist_free(nvconfig); + + if (spa_check_logs(spa)) { + *ereport = FM_EREPORT_ZFS_LOG_REPLAY; + return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO)); + } + } + + if (spa_writeable(spa) && (state == SPA_LOAD_RECOVER || + spa->spa_load_max_txg == UINT64_MAX)) { dmu_tx_t *tx; int need_update = B_FALSE; @@ -1555,21 +2035,32 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) /* * Claim log blocks that haven't been committed yet. * This must all happen in a single txg. + * Note: spa_claim_max_txg is updated by spa_claim_notify(), + * invoked from zil_claim_log_block()'s i/o done callback. + * Price of rollback is that we abandon the log. */ + spa->spa_claiming = B_TRUE; + 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_log_state = SPA_LOG_GOOD; + spa->spa_claiming = B_FALSE; + + spa_set_log_state(spa, SPA_LOG_GOOD); spa->spa_sync_on = B_TRUE; txg_sync_start(spa->spa_dsl_pool); /* - * Wait for all claims to sync. + * Wait for all claims to sync. We sync up to the highest + * claimed log block birth time so that claimed log blocks + * don't appear to be from the future. spa_claim_max_txg + * will have been set for us by either zil_check_log_chain() + * (invoked from spa_check_logs()) or zil_claim() above. */ - txg_wait_synced(spa->spa_dsl_pool, 0); + txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg); /* * If the config cache is stale, or we have uninitialized @@ -1579,7 +2070,8 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) * in-core spa_config and update the disk labels. */ if (config_cache_txg != spa->spa_config_txg || - state == SPA_LOAD_IMPORT || spa->spa_load_verbatim) + state == SPA_LOAD_IMPORT || spa->spa_load_verbatim || + state == SPA_LOAD_RECOVER) need_update = B_TRUE; for (int c = 0; c < rvd->vdev_children; c++) @@ -1596,19 +2088,101 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) /* * Check all DTLs to see if anything needs resilvering. */ - if (vdev_resilver_needed(rvd, NULL, NULL)) + if (!dsl_scan_resilvering(spa->spa_dsl_pool) && + vdev_resilver_needed(rvd, NULL, NULL)) spa_async_request(spa, SPA_ASYNC_RESILVER); + + /* + * Delete any inconsistent datasets. + */ + (void) dmu_objset_find(spa_name(spa), + dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN); + + /* + * Clean up any stale temporary dataset userrefs. + */ + dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool); } - 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 (0); +} - return (error); +static int +spa_load_retry(spa_t *spa, spa_load_state_t state, int mosconfig) +{ + spa_unload(spa); + spa_deactivate(spa); + + spa->spa_load_max_txg--; + + spa_activate(spa, spa_mode_global); + spa_async_suspend(spa); + + return (spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig)); +} + +static int +spa_load_best(spa_t *spa, spa_load_state_t state, int mosconfig, + uint64_t max_request, int rewind_flags) +{ + nvlist_t *config = NULL; + int load_error, rewind_error; + uint64_t safe_rewind_txg; + uint64_t min_txg; + + if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) { + spa->spa_load_max_txg = spa->spa_load_txg; + spa_set_log_state(spa, SPA_LOG_CLEAR); + } else { + spa->spa_load_max_txg = max_request; + } + + load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING, + mosconfig); + if (load_error == 0) + return (0); + + if (spa->spa_root_vdev != NULL) + config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); + + spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg; + spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp; + + if (rewind_flags & ZPOOL_NEVER_REWIND) { + nvlist_free(config); + return (load_error); + } + + /* Price of rolling back is discarding txgs, including log */ + if (state == SPA_LOAD_RECOVER) + spa_set_log_state(spa, SPA_LOG_CLEAR); + + spa->spa_load_max_txg = spa->spa_last_ubsync_txg; + safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE; + min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ? + TXG_INITIAL : safe_rewind_txg; + + /* + * Continue as long as we're finding errors, we're still within + * the acceptable rewind range, and we're still finding uberblocks + */ + while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg && + spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) { + if (spa->spa_load_max_txg < safe_rewind_txg) + spa->spa_extreme_rewind = B_TRUE; + rewind_error = spa_load_retry(spa, state, mosconfig); + } + + if (config) + spa_rewind_data_to_nvlist(spa, config); + + spa->spa_extreme_rewind = B_FALSE; + spa->spa_load_max_txg = UINT64_MAX; + + if (config && (rewind_error || state != SPA_LOAD_RECOVER)) + spa_config_set(spa, config); + + return (state == SPA_LOAD_RECOVER ? rewind_error : load_error); } /* @@ -1624,7 +2198,8 @@ out: * ambiguous state. */ static int -spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config) +spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t *nvpolicy, + nvlist_t **config) { spa_t *spa; int error; @@ -1648,11 +2223,23 @@ spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config) mutex_exit(&spa_namespace_lock); return (ENOENT); } + if (spa->spa_state == POOL_STATE_UNINITIALIZED) { + spa_load_state_t state = SPA_LOAD_OPEN; + zpool_rewind_policy_t policy; + + zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config, + &policy); + if (policy.zrp_request & ZPOOL_DO_REWIND) + state = SPA_LOAD_RECOVER; spa_activate(spa, spa_mode_global); - error = spa_load(spa, spa->spa_config, SPA_LOAD_OPEN, B_FALSE); + if (state != SPA_LOAD_RECOVER) + spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; + + error = spa_load_best(spa, state, B_FALSE, policy.zrp_txg, + policy.zrp_request); if (error == EBADF) { /* @@ -1677,38 +2264,49 @@ spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config) * 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); + if (config != NULL && spa->spa_config) + VERIFY(nvlist_dup(spa->spa_config, config, + KM_SLEEP) == 0); spa_unload(spa); spa_deactivate(spa); - spa->spa_last_open_failed = B_TRUE; + spa->spa_last_open_failed = error; 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); + if (locked) { + spa->spa_last_open_failed = 0; + spa->spa_last_ubsync_txg = 0; + spa->spa_load_txg = 0; + mutex_exit(&spa_namespace_lock); + } + + *spapp = spa; + return (0); } int +spa_open_rewind(const char *name, spa_t **spapp, void *tag, nvlist_t *policy, + nvlist_t **config) +{ + return (spa_open_common(name, spapp, tag, policy, config)); +} + +int spa_open(const char *name, spa_t **spapp, void *tag) { - return (spa_open_common(name, spapp, tag, NULL)); + return (spa_open_common(name, spapp, tag, NULL, NULL)); } /* @@ -1779,7 +2377,7 @@ spa_add_spares(spa_t *spa, nvlist_t *config) if (spa_spare_exists(guid, &pool, NULL) && pool != 0ULL) { VERIFY(nvlist_lookup_uint64_array( - spares[i], ZPOOL_CONFIG_STATS, + spares[i], ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) == 0); vs->vs_state = VDEV_STATE_CANT_OPEN; vs->vs_aux = VDEV_AUX_SPARED; @@ -1836,7 +2434,8 @@ spa_add_l2cache(spa_t *spa, nvlist_t *config) ASSERT(vd != NULL); VERIFY(nvlist_lookup_uint64_array(l2cache[i], - ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0); + ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) + == 0); vdev_get_stats(vd, vs); } } @@ -1849,7 +2448,7 @@ spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen) spa_t *spa; *config = NULL; - error = spa_open_common(name, &spa, FTAG, config); + error = spa_open_common(name, &spa, FTAG, NULL, config); if (spa != NULL) { /* @@ -2093,7 +2692,7 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, uint64_t txg = TXG_INITIAL; nvlist_t **spares, **l2cache; uint_t nspares, nl2cache; - uint64_t version; + uint64_t version, obj; /* * If this pool already exists, return failure. @@ -2109,11 +2708,9 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, */ (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); - spa = spa_add(pool, altroot); + spa = spa_add(pool, NULL, altroot); spa_activate(spa, spa_mode_global); - spa->spa_uberblock.ub_txg = txg - 1; - if (props && (error = spa_prop_validate(spa, props))) { spa_deactivate(spa); spa_remove(spa); @@ -2125,6 +2722,9 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, &version) != 0) version = SPA_VERSION; ASSERT(version <= SPA_VERSION); + + spa->spa_first_txg = txg; + spa->spa_uberblock.ub_txg = txg - 1; spa->spa_uberblock.ub_version = version; spa->spa_ubsync = spa->spa_uberblock; @@ -2200,6 +2800,13 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg); spa->spa_meta_objset = dp->dp_meta_objset; + /* + * Create DDTs (dedup tables). + */ + ddt_create(spa); + + spa_update_dspace(spa); + tx = dmu_tx_create_assigned(dp, txg); /* @@ -2215,6 +2822,12 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, cmn_err(CE_PANIC, "failed to add pool config"); } + if (zap_add(spa->spa_meta_objset, + DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION, + sizeof (uint64_t), 1, &version, tx) != 0) { + cmn_err(CE_PANIC, "failed to add pool version"); + } + /* Newly created pools with the right version are always deflated. */ if (version >= SPA_VERSION_RAIDZ_DEFLATE) { spa->spa_deflate = TRUE; @@ -2226,20 +2839,20 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, } /* - * Create the deferred-free bplist object. Turn off compression + * Create the deferred-free bpobj. 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, + obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx); + dmu_object_set_compress(spa->spa_meta_objset, 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"); + DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ, + sizeof (uint64_t), 1, &obj, tx) != 0) { + cmn_err(CE_PANIC, "failed to add bpobj"); } + VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj, + spa->spa_meta_objset, obj)); /* * Create the pool's history object. @@ -2254,9 +2867,10 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE); spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND); + if (props != NULL) { spa_configfile_set(spa, props, B_FALSE); - spa_sync_props(spa, props, CRED(), tx); + spa_sync_props(spa, props, tx); } dmu_tx_commit(tx); @@ -2387,7 +3001,17 @@ spa_import_rootpool(char *devpath, char *devid) /* * Read the label from the boot device and generate a configuration. */ - if ((config = spa_generate_rootconf(devpath, devid, &guid)) == NULL) { + config = spa_generate_rootconf(devpath, devid, &guid); +#if defined(_OBP) && defined(_KERNEL) + if (config == NULL) { + if (strstr(devpath, "/iscsi/ssd") != NULL) { + /* iscsi boot */ + get_iscsi_bootpath_phy(devpath); + config = spa_generate_rootconf(devpath, devid, &guid); + } + } +#endif + if (config == NULL) { cmn_err(CE_NOTE, "Can not read the pool label from '%s'", devpath); return (EIO); @@ -2406,7 +3030,7 @@ spa_import_rootpool(char *devpath, char *devid) spa_remove(spa); } - spa = spa_add(pname, NULL); + spa = spa_add(pname, config, NULL); spa->spa_is_root = B_TRUE; spa->spa_load_verbatim = B_TRUE; @@ -2462,7 +3086,6 @@ spa_import_rootpool(char *devpath, char *devid) goto out; } - VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); error = 0; spa_history_log_version(spa, LOG_POOL_IMPORT); out: @@ -2495,12 +3118,10 @@ spa_import_verbatim(const char *pool, nvlist_t *config, nvlist_t *props) (void) nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); - spa = spa_add(pool, altroot); + spa = spa_add(pool, config, altroot); spa->spa_load_verbatim = B_TRUE; - VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); - if (props != NULL) spa_configfile_set(spa, props, B_FALSE); @@ -2520,6 +3141,8 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) { spa_t *spa; char *altroot = NULL; + spa_load_state_t state = SPA_LOAD_IMPORT; + zpool_rewind_policy_t policy; int error; nvlist_t *nvroot; nvlist_t **spares, **l2cache; @@ -2529,17 +3152,21 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) * If a pool with this name exists, return failure. */ mutex_enter(&spa_namespace_lock); - if ((spa = spa_lookup(pool)) != NULL) { + if (spa_lookup(pool) != NULL) { mutex_exit(&spa_namespace_lock); return (EEXIST); } + zpool_get_rewind_policy(config, &policy); + if (policy.zrp_request & ZPOOL_DO_REWIND) + state = SPA_LOAD_RECOVER; + /* * 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 = spa_add(pool, config, altroot); spa_activate(spa, spa_mode_global); /* @@ -2552,7 +3179,16 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) * because the user-supplied config is actually the one to trust when * doing an import. */ - error = spa_load(spa, config, SPA_LOAD_IMPORT, B_TRUE); + if (state != SPA_LOAD_RECOVER) + spa->spa_last_ubsync_txg = spa->spa_load_txg = 0; + error = spa_load_best(spa, state, B_TRUE, policy.zrp_txg, + policy.zrp_request); + + /* + * Propagate anything learned about failing or best txgs + * back to caller + */ + spa_rewind_data_to_nvlist(spa, config); spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* @@ -2592,8 +3228,6 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) return (error); } - spa_async_resume(spa); - /* * Override any spares and level 2 cache devices as specified by * the user, as these may have correct device names/devids, etc. @@ -2629,6 +3263,14 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) spa->spa_l2cache.sav_sync = B_TRUE; } + /* + * Check for any removed devices. + */ + if (spa->spa_autoreplace) { + spa_aux_check_removed(&spa->spa_spares); + spa_aux_check_removed(&spa->spa_l2cache); + } + if (spa_writeable(spa)) { /* * Update the config cache to include the newly-imported pool. @@ -2636,6 +3278,8 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); } + spa_async_resume(spa); + /* * It's possible that the pool was expanded while it was exported. * We kick off an async task to handle this for us. @@ -2648,13 +3292,6 @@ spa_import(const char *pool, nvlist_t *config, nvlist_t *props) return (0); } - -/* - * 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) { @@ -2674,7 +3311,7 @@ spa_tryimport(nvlist_t *tryconfig) * Create and initialize the spa structure. */ mutex_enter(&spa_namespace_lock); - spa = spa_add(TRYIMPORT_NAME, NULL); + spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL); spa_activate(spa, FREAD); /* @@ -2682,7 +3319,7 @@ spa_tryimport(nvlist_t *tryconfig) * Pass TRUE for mosconfig because the user-supplied config * is actually the one to trust when doing an import. */ - error = spa_load(spa, tryconfig, SPA_LOAD_TRYIMPORT, B_TRUE); + error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING, B_TRUE); /* * If 'tryconfig' was at least parsable, return the current config. @@ -2827,7 +3464,8 @@ spa_export_common(char *pool, int new_state, nvlist_t **oldconfig, if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); spa->spa_state = new_state; - spa->spa_final_txg = spa_last_synced_txg(spa) + 1; + spa->spa_final_txg = spa_last_synced_txg(spa) + + TXG_DEFER_SIZE + 1; vdev_config_dirty(spa->spa_root_vdev); spa_config_exit(spa, SCL_ALL, FTAG); } @@ -2897,7 +3535,7 @@ spa_reset(char *pool) int spa_vdev_add(spa_t *spa, nvlist_t *nvroot) { - uint64_t txg; + uint64_t txg, id; int error; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd, *tvd; @@ -2938,9 +3576,19 @@ spa_vdev_add(spa_t *spa, nvlist_t *nvroot) * Transfer each new top-level vdev from vd to rvd. */ for (int c = 0; c < vd->vdev_children; c++) { + + /* + * Set the vdev id to the first hole, if one exists. + */ + for (id = 0; id < rvd->vdev_children; id++) { + if (rvd->vdev_child[id]->vdev_ishole) { + vdev_free(rvd->vdev_child[id]); + break; + } + } tvd = vd->vdev_child[c]; vdev_remove_child(vd, tvd); - tvd->vdev_id = rvd->vdev_children; + tvd->vdev_id = id; vdev_add_child(rvd, tvd); vdev_config_dirty(tvd); } @@ -2997,7 +3645,7 @@ spa_vdev_add(spa_t *spa, nvlist_t *nvroot) int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) { - uint64_t txg, open_txg; + uint64_t txg, dtl_max_txg; vdev_t *rvd = spa->spa_root_vdev; vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd; vdev_ops_t *pvops; @@ -3123,6 +3771,7 @@ spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) */ vdev_remove_child(newrootvd, newvd); newvd->vdev_id = pvd->vdev_children; + newvd->vdev_crtxg = oldvd->vdev_crtxg; vdev_add_child(pvd, newvd); tvd = newvd->vdev_top; @@ -3132,13 +3781,14 @@ spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) 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(). + * Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account + * for any dmu_sync-ed blocks. It will propagate upward when + * spa_vdev_exit() calls vdev_dtl_reassess(). */ - open_txg = txg + TXG_CONCURRENT_STATES - 1; + dtl_max_txg = txg + TXG_CONCURRENT_STATES; - vdev_dtl_dirty(newvd, DTL_MISSING, - TXG_INITIAL, open_txg - TXG_INITIAL + 1); + vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL, + dtl_max_txg - TXG_INITIAL); if (newvd->vdev_isspare) { spa_spare_activate(newvd); @@ -3154,10 +3804,18 @@ spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) */ vdev_dirty(tvd, VDD_DTL, newvd, txg); - (void) spa_vdev_exit(spa, newrootvd, open_txg, 0); + /* + * Restart the resilver + */ + dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg); - spa_history_internal_log(LOG_POOL_VDEV_ATTACH, spa, NULL, - CRED(), "%s vdev=%s %s vdev=%s", + /* + * Commit the config + */ + (void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0); + + spa_history_log_internal(LOG_POOL_VDEV_ATTACH, spa, NULL, + "%s vdev=%s %s vdev=%s", replacing && newvd_isspare ? "spare in" : replacing ? "replace" : "attach", newvdpath, replacing ? "for" : "to", oldvdpath); @@ -3165,11 +3823,6 @@ spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) spa_strfree(oldvdpath); spa_strfree(newvdpath); - /* - * Kick off a resilver to update newvd. - */ - VERIFY3U(spa_scrub(spa, POOL_SCRUB_RESILVER), ==, 0); - return (0); } @@ -3188,6 +3841,7 @@ spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) boolean_t unspare = B_FALSE; uint64_t unspare_guid; size_t len; + char *vdpath; txg = spa_vdev_enter(spa); @@ -3354,6 +4008,7 @@ spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) * But first make sure we're not on any *other* txg's DTL list, to * prevent vd from being accessed after it's freed. */ + vdpath = spa_strdup(vd->vdev_path); for (int t = 0; t < TXG_SIZE; t++) (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t); vd->vdev_detached = B_TRUE; @@ -3363,6 +4018,10 @@ spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) error = spa_vdev_exit(spa, vd, txg, 0); + spa_history_log_internal(LOG_POOL_VDEV_DETACH, spa, NULL, + "vdev=%s", vdpath); + spa_strfree(vdpath); + /* * 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 @@ -3379,7 +4038,8 @@ spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) continue; spa_open_ref(spa, FTAG); mutex_exit(&spa_namespace_lock); - (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); + (void) spa_vdev_remove(spa, unspare_guid, + B_TRUE); mutex_enter(&spa_namespace_lock); spa_close(spa, FTAG); } @@ -3389,6 +4049,289 @@ spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) return (error); } +/* + * Split a set of devices from their mirrors, and create a new pool from them. + */ +int +spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config, + nvlist_t *props, boolean_t exp) +{ + int error = 0; + uint64_t txg, *glist; + spa_t *newspa; + uint_t c, children, lastlog; + nvlist_t **child, *nvl, *tmp; + dmu_tx_t *tx; + char *altroot = NULL; + vdev_t *rvd, **vml = NULL; /* vdev modify list */ + boolean_t activate_slog; + + if (!spa_writeable(spa)) + return (EROFS); + + txg = spa_vdev_enter(spa); + + /* clear the log and flush everything up to now */ + activate_slog = spa_passivate_log(spa); + (void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); + error = spa_offline_log(spa); + txg = spa_vdev_config_enter(spa); + + if (activate_slog) + spa_activate_log(spa); + + if (error != 0) + return (spa_vdev_exit(spa, NULL, txg, error)); + + /* check new spa name before going any further */ + if (spa_lookup(newname) != NULL) + return (spa_vdev_exit(spa, NULL, txg, EEXIST)); + + /* + * scan through all the children to ensure they're all mirrors + */ + if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 || + nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child, + &children) != 0) + return (spa_vdev_exit(spa, NULL, txg, EINVAL)); + + /* first, check to ensure we've got the right child count */ + rvd = spa->spa_root_vdev; + lastlog = 0; + for (c = 0; c < rvd->vdev_children; c++) { + vdev_t *vd = rvd->vdev_child[c]; + + /* don't count the holes & logs as children */ + if (vd->vdev_islog || vd->vdev_ishole) { + if (lastlog == 0) + lastlog = c; + continue; + } + + lastlog = 0; + } + if (children != (lastlog != 0 ? lastlog : rvd->vdev_children)) + return (spa_vdev_exit(spa, NULL, txg, EINVAL)); + + /* next, ensure no spare or cache devices are part of the split */ + if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 || + nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0) + return (spa_vdev_exit(spa, NULL, txg, EINVAL)); + + vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP); + glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP); + + /* then, loop over each vdev and validate it */ + for (c = 0; c < children; c++) { + uint64_t is_hole = 0; + + (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, + &is_hole); + + if (is_hole != 0) { + if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole || + spa->spa_root_vdev->vdev_child[c]->vdev_islog) { + continue; + } else { + error = EINVAL; + break; + } + } + + /* which disk is going to be split? */ + if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID, + &glist[c]) != 0) { + error = EINVAL; + break; + } + + /* look it up in the spa */ + vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE); + if (vml[c] == NULL) { + error = ENODEV; + break; + } + + /* make sure there's nothing stopping the split */ + if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops || + vml[c]->vdev_islog || + vml[c]->vdev_ishole || + vml[c]->vdev_isspare || + vml[c]->vdev_isl2cache || + !vdev_writeable(vml[c]) || + vml[c]->vdev_children != 0 || + vml[c]->vdev_state != VDEV_STATE_HEALTHY || + c != spa->spa_root_vdev->vdev_child[c]->vdev_id) { + error = EINVAL; + break; + } + + if (vdev_dtl_required(vml[c])) { + error = EBUSY; + break; + } + + /* we need certain info from the top level */ + VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY, + vml[c]->vdev_top->vdev_ms_array) == 0); + VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT, + vml[c]->vdev_top->vdev_ms_shift) == 0); + VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE, + vml[c]->vdev_top->vdev_asize) == 0); + VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT, + vml[c]->vdev_top->vdev_ashift) == 0); + } + + if (error != 0) { + kmem_free(vml, children * sizeof (vdev_t *)); + kmem_free(glist, children * sizeof (uint64_t)); + return (spa_vdev_exit(spa, NULL, txg, error)); + } + + /* stop writers from using the disks */ + for (c = 0; c < children; c++) { + if (vml[c] != NULL) + vml[c]->vdev_offline = B_TRUE; + } + vdev_reopen(spa->spa_root_vdev); + + /* + * Temporarily record the splitting vdevs in the spa config. This + * will disappear once the config is regenerated. + */ + VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY(nvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, + glist, children) == 0); + kmem_free(glist, children * sizeof (uint64_t)); + + mutex_enter(&spa->spa_props_lock); + VERIFY(nvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, + nvl) == 0); + mutex_exit(&spa->spa_props_lock); + spa->spa_config_splitting = nvl; + vdev_config_dirty(spa->spa_root_vdev); + + /* configure and create the new pool */ + VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname) == 0); + VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, + exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE) == 0); + VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, + spa_version(spa)) == 0); + VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, + spa->spa_config_txg) == 0); + VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, + spa_generate_guid(NULL)) == 0); + (void) nvlist_lookup_string(props, + zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); + + /* add the new pool to the namespace */ + newspa = spa_add(newname, config, altroot); + newspa->spa_config_txg = spa->spa_config_txg; + spa_set_log_state(newspa, SPA_LOG_CLEAR); + + /* release the spa config lock, retaining the namespace lock */ + spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); + + if (zio_injection_enabled) + zio_handle_panic_injection(spa, FTAG, 1); + + spa_activate(newspa, spa_mode_global); + spa_async_suspend(newspa); + + /* create the new pool from the disks of the original pool */ + error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE, B_TRUE); + if (error) + goto out; + + /* if that worked, generate a real config for the new pool */ + if (newspa->spa_root_vdev != NULL) { + VERIFY(nvlist_alloc(&newspa->spa_config_splitting, + NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY(nvlist_add_uint64(newspa->spa_config_splitting, + ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa)) == 0); + spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL, + B_TRUE)); + } + + /* set the props */ + if (props != NULL) { + spa_configfile_set(newspa, props, B_FALSE); + error = spa_prop_set(newspa, props); + if (error) + goto out; + } + + /* flush everything */ + txg = spa_vdev_config_enter(newspa); + vdev_config_dirty(newspa->spa_root_vdev); + (void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG); + + if (zio_injection_enabled) + zio_handle_panic_injection(spa, FTAG, 2); + + spa_async_resume(newspa); + + /* finally, update the original pool's config */ + txg = spa_vdev_config_enter(spa); + tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); + error = dmu_tx_assign(tx, TXG_WAIT); + if (error != 0) + dmu_tx_abort(tx); + for (c = 0; c < children; c++) { + if (vml[c] != NULL) { + vdev_split(vml[c]); + if (error == 0) + spa_history_log_internal(LOG_POOL_VDEV_DETACH, + spa, tx, "vdev=%s", + vml[c]->vdev_path); + vdev_free(vml[c]); + } + } + vdev_config_dirty(spa->spa_root_vdev); + spa->spa_config_splitting = NULL; + nvlist_free(nvl); + if (error == 0) + dmu_tx_commit(tx); + (void) spa_vdev_exit(spa, NULL, txg, 0); + + if (zio_injection_enabled) + zio_handle_panic_injection(spa, FTAG, 3); + + /* split is complete; log a history record */ + spa_history_log_internal(LOG_POOL_SPLIT, newspa, NULL, + "split new pool %s from pool %s", newname, spa_name(spa)); + + kmem_free(vml, children * sizeof (vdev_t *)); + + /* if we're not going to mount the filesystems in userland, export */ + if (exp) + error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL, + B_FALSE, B_FALSE); + + return (error); + +out: + spa_unload(newspa); + spa_deactivate(newspa); + spa_remove(newspa); + + txg = spa_vdev_config_enter(spa); + + /* re-online all offlined disks */ + for (c = 0; c < children; c++) { + if (vml[c] != NULL) + vml[c]->vdev_offline = B_FALSE; + } + vdev_reopen(spa->spa_root_vdev); + + nvlist_free(spa->spa_config_splitting); + spa->spa_config_splitting = NULL; + (void) spa_vdev_exit(spa, NULL, txg, error); + + kmem_free(vml, children * sizeof (vdev_t *)); + return (error); +} + static nvlist_t * spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid) { @@ -3431,16 +4374,113 @@ spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count, } /* + * Evacuate the device. + */ +static int +spa_vdev_remove_evacuate(spa_t *spa, vdev_t *vd) +{ + uint64_t txg; + int error = 0; + + ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); + ASSERT(vd == vd->vdev_top); + + /* + * Evacuate the device. We don't hold the config lock as writer + * since we need to do I/O but we do keep the + * spa_namespace_lock held. Once this completes the device + * should no longer have any blocks allocated on it. + */ + if (vd->vdev_islog) { + if (vd->vdev_stat.vs_alloc != 0) + error = spa_offline_log(spa); + } else { + error = ENOTSUP; + } + + if (error) + return (error); + + /* + * The evacuation succeeded. Remove any remaining MOS metadata + * associated with this vdev, and wait for these changes to sync. + */ + ASSERT3U(vd->vdev_stat.vs_alloc, ==, 0); + txg = spa_vdev_config_enter(spa); + vd->vdev_removing = B_TRUE; + vdev_dirty(vd, 0, NULL, txg); + vdev_config_dirty(vd); + spa_vdev_config_exit(spa, NULL, txg, 0, FTAG); + + return (0); +} + +/* + * Complete the removal by cleaning up the namespace. + */ +static void +spa_vdev_remove_from_namespace(spa_t *spa, vdev_t *vd) +{ + vdev_t *rvd = spa->spa_root_vdev; + uint64_t id = vd->vdev_id; + boolean_t last_vdev = (id == (rvd->vdev_children - 1)); + + ASSERT(MUTEX_HELD(&spa_namespace_lock)); + ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); + ASSERT(vd == vd->vdev_top); + + /* + * Only remove any devices which are empty. + */ + if (vd->vdev_stat.vs_alloc != 0) + return; + + (void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); + + if (list_link_active(&vd->vdev_state_dirty_node)) + vdev_state_clean(vd); + if (list_link_active(&vd->vdev_config_dirty_node)) + vdev_config_clean(vd); + + vdev_free(vd); + + if (last_vdev) { + vdev_compact_children(rvd); + } else { + vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops); + vdev_add_child(rvd, vd); + } + vdev_config_dirty(rvd); + + /* + * Reassess the health of our root vdev. + */ + vdev_reopen(rvd); +} + +/* + * Remove a device from the pool - + * + * Removing a device from the vdev namespace requires several steps + * and can take a significant amount of time. As a result we use + * the spa_vdev_config_[enter/exit] functions which allow us to + * grab and release the spa_config_lock while still holding the namespace + * lock. During each step the configuration is synced out. + */ + +/* * Remove a device from the pool. Currently, this supports removing only hot - * spares and level 2 ARC devices. + * spares, slogs, and level 2 ARC devices. */ int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare) { vdev_t *vd; + metaslab_group_t *mg; nvlist_t **spares, **l2cache, *nv; - uint_t nspares, nl2cache; uint64_t txg = 0; + uint_t nspares, nl2cache; int error = 0; boolean_t locked = MUTEX_HELD(&spa_namespace_lock); @@ -3476,6 +4516,49 @@ spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare) ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv); spa_load_l2cache(spa); spa->spa_l2cache.sav_sync = B_TRUE; + } else if (vd != NULL && vd->vdev_islog) { + ASSERT(!locked); + ASSERT(vd == vd->vdev_top); + + /* + * XXX - Once we have bp-rewrite this should + * become the common case. + */ + + mg = vd->vdev_mg; + + /* + * Stop allocating from this vdev. + */ + metaslab_group_passivate(mg); + + /* + * Wait for the youngest allocations and frees to sync, + * and then wait for the deferral of those frees to finish. + */ + spa_vdev_config_exit(spa, NULL, + txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG); + + /* + * Attempt to evacuate the vdev. + */ + error = spa_vdev_remove_evacuate(spa, vd); + + txg = spa_vdev_config_enter(spa); + + /* + * If we couldn't evacuate the vdev, unwind. + */ + if (error) { + metaslab_group_activate(mg); + return (spa_vdev_exit(spa, NULL, txg, error)); + } + + /* + * Clean up the vdev namespace. + */ + spa_vdev_remove_from_namespace(spa, vd); + } else if (vd != NULL) { /* * Normal vdevs cannot be removed (yet). @@ -3517,6 +4600,7 @@ spa_vdev_resilver_done_hunt(vdev_t *vd) newvd = vd->vdev_child[1]; if (vdev_dtl_empty(newvd, DTL_MISSING) && + vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) return (oldvd); } @@ -3530,6 +4614,7 @@ spa_vdev_resilver_done_hunt(vdev_t *vd) if (newvd->vdev_unspare && vdev_dtl_empty(newvd, DTL_MISSING) && + vdev_dtl_empty(newvd, DTL_OUTAGE) && !vdev_dtl_required(oldvd)) { newvd->vdev_unspare = 0; return (oldvd); @@ -3576,36 +4661,41 @@ spa_vdev_resilver_done(spa_t *spa) } /* - * Update the stored path or FRU for this vdev. Dirty the vdev configuration, - * relying on spa_vdev_enter/exit() to synchronize the labels and cache. + * Update the stored path or FRU for this vdev. */ int spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value, boolean_t ispath) { vdev_t *vd; - uint64_t txg; + boolean_t sync = B_FALSE; - txg = spa_vdev_enter(spa); + spa_vdev_state_enter(spa, SCL_ALL); if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) - return (spa_vdev_exit(spa, NULL, txg, ENOENT)); + return (spa_vdev_state_exit(spa, NULL, ENOENT)); if (!vd->vdev_ops->vdev_op_leaf) - return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); + return (spa_vdev_state_exit(spa, NULL, ENOTSUP)); if (ispath) { - spa_strfree(vd->vdev_path); - vd->vdev_path = spa_strdup(value); + if (strcmp(value, vd->vdev_path) != 0) { + spa_strfree(vd->vdev_path); + vd->vdev_path = spa_strdup(value); + sync = B_TRUE; + } } else { - if (vd->vdev_fru != NULL) + if (vd->vdev_fru == NULL) { + vd->vdev_fru = spa_strdup(value); + sync = B_TRUE; + } else if (strcmp(value, vd->vdev_fru) != 0) { spa_strfree(vd->vdev_fru); - vd->vdev_fru = spa_strdup(value); + vd->vdev_fru = spa_strdup(value); + sync = B_TRUE; + } } - vdev_config_dirty(vd->vdev_top); - - return (spa_vdev_exit(spa, NULL, txg, 0)); + return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0)); } int @@ -3622,40 +4712,38 @@ spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru) /* * ========================================================================== - * SPA Scrubbing + * SPA Scanning * ========================================================================== */ int -spa_scrub(spa_t *spa, pool_scrub_type_t type) +spa_scan_stop(spa_t *spa) +{ + ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); + if (dsl_scan_resilvering(spa->spa_dsl_pool)) + return (EBUSY); + return (dsl_scan_cancel(spa->spa_dsl_pool)); +} + +int +spa_scan(spa_t *spa, pool_scan_func_t func) { ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); - if ((uint_t)type >= POOL_SCRUB_TYPES) + if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE) 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 && + if (func == POOL_SCAN_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); - } + return (dsl_scan(spa->spa_dsl_pool, func)); } /* @@ -3668,9 +4756,20 @@ static void spa_async_remove(spa_t *spa, vdev_t *vd) { if (vd->vdev_remove_wanted) { - vd->vdev_remove_wanted = 0; + vd->vdev_remove_wanted = B_FALSE; + vd->vdev_delayed_close = B_FALSE; vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE); - vdev_clear(spa, vd); + + /* + * We want to clear the stats, but we don't want to do a full + * vdev_clear() as that will cause us to throw away + * degraded/faulted state as well as attempt to reopen the + * device, all of which is a waste. + */ + vd->vdev_stat.vs_read_errors = 0; + vd->vdev_stat.vs_write_errors = 0; + vd->vdev_stat.vs_checksum_errors = 0; + vdev_state_dirty(vd->vdev_top); } @@ -3682,7 +4781,7 @@ static void spa_async_probe(spa_t *spa, vdev_t *vd) { if (vd->vdev_probe_wanted) { - vd->vdev_probe_wanted = 0; + vd->vdev_probe_wanted = B_FALSE; vdev_reopen(vd); /* vdev_open() does the actual probe */ } @@ -3737,24 +4836,23 @@ spa_async_thread(spa_t *spa) * See if the config needs to be updated. */ if (tasks & SPA_ASYNC_CONFIG_UPDATE) { - uint64_t oldsz, space_update; + uint64_t old_space, new_space; mutex_enter(&spa_namespace_lock); - oldsz = spa_get_space(spa); + old_space = metaslab_class_get_space(spa_normal_class(spa)); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); - space_update = spa_get_space(spa) - oldsz; + new_space = metaslab_class_get_space(spa_normal_class(spa)); mutex_exit(&spa_namespace_lock); /* * If the pool grew as a result of the config update, * then log an internal history event. */ - if (space_update) { - spa_history_internal_log(LOG_POOL_VDEV_ONLINE, - spa, NULL, CRED(), + if (new_space != old_space) { + spa_history_log_internal(LOG_POOL_VDEV_ONLINE, + spa, NULL, "pool '%s' size: %llu(+%llu)", - spa_name(spa), spa_get_space(spa), - space_update); + spa_name(spa), new_space, new_space - old_space); } } @@ -3762,7 +4860,7 @@ spa_async_thread(spa_t *spa) * See if any devices need to be marked REMOVED. */ if (tasks & SPA_ASYNC_REMOVE) { - spa_vdev_state_enter(spa); + spa_vdev_state_enter(spa, SCL_NONE); 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]); @@ -3781,7 +4879,7 @@ spa_async_thread(spa_t *spa) * See if any devices need to be probed. */ if (tasks & SPA_ASYNC_PROBE) { - spa_vdev_state_enter(spa); + spa_vdev_state_enter(spa, SCL_NONE); spa_async_probe(spa, spa->spa_root_vdev); (void) spa_vdev_state_exit(spa, NULL, 0); } @@ -3796,7 +4894,7 @@ spa_async_thread(spa_t *spa) * Kick off a resilver. */ if (tasks & SPA_ASYNC_RESILVER) - VERIFY(spa_scrub(spa, POOL_SCRUB_RESILVER) == 0); + dsl_resilver_restart(spa->spa_dsl_pool, 0); /* * Let the world know that we're done. @@ -3842,6 +4940,7 @@ spa_async_dispatch(spa_t *spa) void spa_async_request(spa_t *spa, int task) { + zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task); mutex_enter(&spa->spa_async_lock); spa->spa_async_tasks |= task; mutex_exit(&spa->spa_async_lock); @@ -3853,37 +4952,22 @@ spa_async_request(spa_t *spa, int task) * ========================================================================== */ -static void -spa_sync_deferred_frees(spa_t *spa, uint64_t txg) +static int +bpobj_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) { - 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); + bpobj_t *bpo = arg; + bpobj_enqueue(bpo, bp, tx); + return (0); +} - tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg); - bplist_vacate(bpl, tx); +static int +spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) +{ + zio_t *zio = arg; - /* - * 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); + zio_nowait(zio_free_sync(zio, zio->io_spa, dmu_tx_get_txg(tx), bp, + zio->io_flags)); + return (0); } static void @@ -3950,7 +5034,7 @@ spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx, 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); + B_FALSE, VDEV_CONFIG_L2CACHE); VERIFY(nvlist_add_nvlist_array(nvroot, config, list, sav->sav_count) == 0); for (i = 0; i < sav->sav_count; i++) @@ -3990,7 +5074,7 @@ spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) * Set zpool properties. */ static void -spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) +spa_sync_props(void *arg1, void *arg2, dmu_tx_t *tx) { spa_t *spa = arg1; objset_t *mos = spa->spa_meta_objset; @@ -4041,8 +5125,6 @@ spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) * 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); @@ -4091,7 +5173,12 @@ spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) break; case ZPOOL_PROP_AUTOEXPAND: spa->spa_autoexpand = intval; - spa_async_request(spa, SPA_ASYNC_AUTOEXPAND); + if (tx->tx_txg != TXG_INITIAL) + spa_async_request(spa, + SPA_ASYNC_AUTOEXPAND); + break; + case ZPOOL_PROP_DEDUPDITTO: + spa->spa_dedup_ditto = intval; break; default: break; @@ -4101,8 +5188,8 @@ spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) /* 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", + spa_history_log_internal(LOG_POOL_PROPSET, + spa, tx, "%s %lld %s", nvpair_name(elem), intval, spa_name(spa)); } } @@ -4111,6 +5198,42 @@ spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) } /* + * Perform one-time upgrade on-disk changes. spa_version() does not + * reflect the new version this txg, so there must be no changes this + * txg to anything that the upgrade code depends on after it executes. + * Therefore this must be called after dsl_pool_sync() does the sync + * tasks. + */ +static void +spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx) +{ + dsl_pool_t *dp = spa->spa_dsl_pool; + + ASSERT(spa->spa_sync_pass == 1); + + 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 (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES && + spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) { + dsl_pool_upgrade_dir_clones(dp, tx); + + /* Keeping the freedir open increases spa_minref */ + spa->spa_minref += 3; + } +} + +/* * Sync the specified transaction group. New blocks may be dirtied as * part of the process, so we iterate until it converges. */ @@ -4119,11 +5242,11 @@ 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; + bpobj_t *defer_bpo = &spa->spa_deferred_bpobj; + bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK]; vdev_t *rvd = spa->spa_root_vdev; vdev_t *vd; dmu_tx_t *tx; - int dirty_vdevs; int error; /* @@ -4158,8 +5281,6 @@ spa_sync(spa_t *spa, uint64_t txg) } 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); /* @@ -4183,34 +5304,29 @@ spa_sync(spa_t *spa, uint64_t txg) } } - 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 anything has changed in this txg, or if someone is waiting + * for this txg to sync (eg, spa_vdev_remove()), 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); + !txg_list_empty(&dp->dp_sync_tasks, txg) || + ((dsl_scan_active(dp->dp_scan) || + txg_sync_waiting(dp)) && !spa_shutting_down(spa))) { + zio_t *zio = zio_root(spa, NULL, NULL, 0); + VERIFY3U(bpobj_iterate(defer_bpo, + spa_free_sync_cb, zio, tx), ==, 0); + VERIFY3U(zio_wait(zio), ==, 0); + } /* * Iterate to convergence. */ do { - spa->spa_sync_pass++; + int pass = ++spa->spa_sync_pass; spa_sync_config_object(spa, tx); spa_sync_aux_dev(spa, &spa->spa_spares, tx, @@ -4220,18 +5336,26 @@ spa_sync(spa_t *spa, uint64_t txg) 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++; + if (pass <= SYNC_PASS_DEFERRED_FREE) { + zio_t *zio = zio_root(spa, NULL, NULL, 0); + bplist_iterate(free_bpl, spa_free_sync_cb, + zio, tx); + VERIFY(zio_wait(zio) == 0); + } else { + bplist_iterate(free_bpl, bpobj_enqueue_cb, + defer_bpo, tx); } - bplist_sync(bpl, tx); - } while (dirty_vdevs); + ddt_sync(spa, txg); + dsl_scan_sync(dp, tx); + + while (vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) + vdev_sync(vd, txg); - bplist_close(bpl); + if (pass == 1) + spa_sync_upgrades(spa, tx); - dprintf("txg %llu passes %d\n", txg, spa->spa_sync_pass); + } while (dmu_objset_is_dirty(mos, txg)); /* * Rewrite the vdev configuration (which includes the uberblock) @@ -4302,10 +5426,7 @@ spa_sync(spa_t *spa, uint64_t txg) spa->spa_ubsync = spa->spa_uberblock; - /* - * Clean up the ZIL records for the synced txg. - */ - dsl_pool_zil_clean(dp); + dsl_pool_sync_done(dp, txg); /* * Update usable space statistics. @@ -4313,6 +5434,8 @@ spa_sync(spa_t *spa, uint64_t txg) while (vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg))) vdev_sync_done(vd, txg); + spa_update_dspace(spa); + /* * It had better be the case that we didn't dirty anything * since vdev_config_sync(). @@ -4320,10 +5443,13 @@ spa_sync(spa_t *spa, uint64_t txg) 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->spa_sync_pass = 0; spa_config_exit(spa, SCL_CONFIG, FTAG); + spa_handle_ignored_writes(spa); + /* * If any async tasks have been requested, kick them off. */ |