OpenZFS 2605, 6980, 6902

2605 want to resume interrupted zfs send
Reviewed by: George Wilson <[email protected]>
Reviewed by: Paul Dagnelie <[email protected]>
Reviewed by: Richard Elling <[email protected]>
Reviewed by: Xin Li <[email protected]>
Reviewed by: Arne Jansen <[email protected]>
Approved by: Dan McDonald <[email protected]>
Ported-by: kernelOfTruth <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/2605
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/9c3fd12

6980 6902 causes zfs send to break due to 32-bit/64-bit struct mismatch
Reviewed by: Paul Dagnelie <[email protected]>
Reviewed by: George Wilson <[email protected]>
Approved by: Robert Mustacchi <[email protected]>
Ported by: Brian Behlendorf <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/6980
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/ea4a67f

Porting notes:
- All rsend and snapshop tests enabled and updated for Linux.
- Fix misuse of input argument in traverse_visitbp().
- Fix ISO C90 warnings and errors.
- Fix gcc 'missing braces around initializer' in
  'struct send_thread_arg to_arg =' warning.
- Replace 4 argument fletcher_4_native() with 3 argument version,
  this change was made in OpenZFS 4185 which has not been ported.
- Part of the sections for 'zfs receive' and 'zfs send' was
  rewritten and reordered to approximate upstream.
- Fix mktree xattr creation, 'user.' prefix required.
- Minor fixes to newly enabled test cases
- Long holds for volumes allowed during receive for minor registration.

7 files changed, 65 insertions, 17 deletions

diff --git a/include/sys/dmu_impl.h b/include/sys/dmu_impl.h
index 75d094f08..d700d1d17 100644
--- a/include/sys/dmu_impl.h
+++ b/include/sys/dmu_impl.h
@@ -24,7 +24,7 @@
  */
 /*
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef _SYS_DMU_IMPL_H
@@ -272,6 +272,8 @@ typedef struct dmu_sendarg {
 	uint64_t dsa_featureflags;
 	uint64_t dsa_last_data_object;
 	uint64_t dsa_last_data_offset;
+	uint64_t dsa_resume_object;
+	uint64_t dsa_resume_offset;
 } dmu_sendarg_t;
 
 void dmu_object_zapify(objset_t *, uint64_t, dmu_object_type_t, dmu_tx_t *);
diff --git a/include/sys/dmu_send.h b/include/sys/dmu_send.h
index 2442a1f8a..871f56254 100644
--- a/include/sys/dmu_send.h
+++ b/include/sys/dmu_send.h
@@ -36,10 +36,13 @@ struct vnode;
 struct dsl_dataset;
 struct drr_begin;
 struct avl_tree;
+struct dmu_replay_record;
 
-int dmu_send(const char *tosnap, const char *fromsnap,
-    boolean_t embedok, boolean_t large_block_ok,
-    int outfd, struct vnode *vp, offset_t *off);
+extern const char *recv_clone_name;
+
+int dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
+    boolean_t large_block_ok, int outfd, uint64_t resumeobj, uint64_t resumeoff,
+    struct vnode *vp, offset_t *off);
 int dmu_send_estimate(struct dsl_dataset *ds, struct dsl_dataset *fromds,
     uint64_t *sizep);
 int dmu_send_estimate_from_txg(struct dsl_dataset *ds, uint64_t fromtxg,
@@ -50,12 +53,14 @@ int dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
 
 typedef struct dmu_recv_cookie {
 	struct dsl_dataset *drc_ds;
+	struct dmu_replay_record *drc_drr_begin;
 	struct drr_begin *drc_drrb;
 	const char *drc_tofs;
 	const char *drc_tosnap;
 	boolean_t drc_newfs;
 	boolean_t drc_byteswap;
 	boolean_t drc_force;
+	boolean_t drc_resumable;
 	struct avl_tree *drc_guid_to_ds_map;
 	zio_cksum_t drc_cksum;
 	uint64_t drc_newsnapobj;
@@ -63,8 +68,9 @@ typedef struct dmu_recv_cookie {
 	cred_t *drc_cred;
 } dmu_recv_cookie_t;
 
-int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *drrb,
-    boolean_t force, char *origin, dmu_recv_cookie_t *drc);
+int dmu_recv_begin(char *tofs, char *tosnap,
+    struct dmu_replay_record *drr_begin,
+    boolean_t force, boolean_t resumable, char *origin, dmu_recv_cookie_t *drc);
 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
     int cleanup_fd, uint64_t *action_handlep);
 int dmu_recv_end(dmu_recv_cookie_t *drc, void *owner);
diff --git a/include/sys/dmu_traverse.h b/include/sys/dmu_traverse.h
index 544b721e4..c010edd44 100644
--- a/include/sys/dmu_traverse.h
+++ b/include/sys/dmu_traverse.h
@@ -54,6 +54,8 @@ typedef int (blkptr_cb_t)(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 
 int traverse_dataset(struct dsl_dataset *ds,
     uint64_t txg_start, int flags, blkptr_cb_t func, void *arg);
+int traverse_dataset_resume(struct dsl_dataset *ds, uint64_t txg_start,
+    zbookmark_phys_t *resume, int flags, blkptr_cb_t func, void *arg);
 int traverse_dataset_destroyed(spa_t *spa, blkptr_t *blkptr,
     uint64_t txg_start, zbookmark_phys_t *resume, int flags,
     blkptr_cb_t func, void *arg);
diff --git a/include/sys/dsl_dataset.h b/include/sys/dsl_dataset.h
index f3e64a772..195a271c9 100644
--- a/include/sys/dsl_dataset.h
+++ b/include/sys/dsl_dataset.h
@@ -99,6 +99,18 @@ struct dsl_pool;
 #define	DS_FIELD_LARGE_DNODE "org.zfsonlinux:large_dnode"
 
 /*
+ * These fields are set on datasets that are in the middle of a resumable
+ * receive, and allow the sender to resume the send if it is interrupted.
+ */
+#define	DS_FIELD_RESUME_FROMGUID "com.delphix:resume_fromguid"
+#define	DS_FIELD_RESUME_TONAME "com.delphix:resume_toname"
+#define	DS_FIELD_RESUME_TOGUID "com.delphix:resume_toguid"
+#define	DS_FIELD_RESUME_OBJECT "com.delphix:resume_object"
+#define	DS_FIELD_RESUME_OFFSET "com.delphix:resume_offset"
+#define	DS_FIELD_RESUME_BYTES "com.delphix:resume_bytes"
+#define	DS_FIELD_RESUME_EMBEDOK "com.delphix:resume_embedok"
+
+/*
  * DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
  * name lookups should be performed case-insensitively.
  */
@@ -191,6 +203,14 @@ typedef struct dsl_dataset {
 	kmutex_t ds_sendstream_lock;
 	list_t ds_sendstreams;
 
+	/*
+	 * When in the middle of a resumable receive, tracks how much
+	 * progress we have made.
+	 */
+	uint64_t ds_resume_object[TXG_SIZE];
+	uint64_t ds_resume_offset[TXG_SIZE];
+	uint64_t ds_resume_bytes[TXG_SIZE];
+
 	/* Protected by our dsl_dir's dd_lock */
 	list_t ds_prop_cbs;
 
@@ -242,6 +262,7 @@ int dsl_dataset_own_obj(struct dsl_pool *dp, uint64_t dsobj,
 void dsl_dataset_disown(dsl_dataset_t *ds, void *tag);
 void dsl_dataset_name(dsl_dataset_t *ds, char *name);
 boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, void *tag);
+boolean_t dsl_dataset_has_owner(dsl_dataset_t *ds);
 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,
@@ -322,6 +343,8 @@ int dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx,
 void dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
     zprop_source_t source, uint64_t value, dmu_tx_t *tx);
 void dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx);
+boolean_t dsl_dataset_is_zapified(dsl_dataset_t *ds);
+boolean_t dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds);
 int dsl_dataset_rollback(const char *fsname, void *owner, nvlist_t *result);
 
 void dsl_dataset_deactivate_feature(uint64_t dsobj,
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
index 0935dca77..8a581eee5 100644
--- a/include/sys/fs/zfs.h
+++ b/include/sys/fs/zfs.h
@@ -158,6 +158,7 @@ typedef enum {
 	ZFS_PROP_REDUNDANT_METADATA,
 	ZFS_PROP_OVERLAY,
 	ZFS_PROP_PREV_SNAP,
+	ZFS_PROP_RECEIVE_RESUME_TOKEN,
 	ZFS_NUM_PROPS
 } zfs_prop_t;
 
diff --git a/include/sys/zfs_ioctl.h b/include/sys/zfs_ioctl.h
index a978524e2..ee6b9f650 100644
--- a/include/sys/zfs_ioctl.h
+++ b/include/sys/zfs_ioctl.h
@@ -90,16 +90,16 @@ typedef enum drr_headertype {
  * Feature flags for zfs send streams (flags in drr_versioninfo)
  */
 
-#define	DMU_BACKUP_FEATURE_DEDUP		(1<<0)
-#define	DMU_BACKUP_FEATURE_DEDUPPROPS		(1<<1)
-#define	DMU_BACKUP_FEATURE_SA_SPILL		(1<<2)
+#define	DMU_BACKUP_FEATURE_DEDUP		(1 << 0)
+#define	DMU_BACKUP_FEATURE_DEDUPPROPS		(1 << 1)
+#define	DMU_BACKUP_FEATURE_SA_SPILL		(1 << 2)
 /* flags #3 - #15 are reserved for incompatible closed-source implementations */
-#define	DMU_BACKUP_FEATURE_EMBED_DATA		(1<<16)
-#define	DMU_BACKUP_FEATURE_EMBED_DATA_LZ4	(1<<17)
+#define	DMU_BACKUP_FEATURE_EMBED_DATA		(1 << 16)
+#define	DMU_BACKUP_FEATURE_EMBED_DATA_LZ4	(1 << 17)
 /* flag #18 is reserved for a Delphix feature */
-#define	DMU_BACKUP_FEATURE_LARGE_BLOCKS		(1<<19)
-/* flag #20 is reserved for resumable streams */
-#define	DMU_BACKUP_FEATURE_LARGE_DNODE		(1<<21)
+#define	DMU_BACKUP_FEATURE_LARGE_BLOCKS		(1 << 19)
+#define	DMU_BACKUP_FEATURE_RESUMING		(1 << 20)
+#define	DMU_BACKUP_FEATURE_LARGE_DNODE		(1 << 21)
 
 /*
  * Mask of all supported backup features
@@ -107,11 +107,16 @@ typedef enum drr_headertype {
 #define	DMU_BACKUP_FEATURE_MASK	(DMU_BACKUP_FEATURE_DEDUP | \
     DMU_BACKUP_FEATURE_DEDUPPROPS | DMU_BACKUP_FEATURE_SA_SPILL | \
     DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_EMBED_DATA_LZ4 | \
-    DMU_BACKUP_FEATURE_LARGE_BLOCKS | DMU_BACKUP_FEATURE_LARGE_DNODE)
+    DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_LARGE_BLOCKS | \
+    DMU_BACKUP_FEATURE_LARGE_DNODE)
 
 /* Are all features in the given flag word currently supported? */
 #define	DMU_STREAM_SUPPORTED(x)	(!((x) & ~DMU_BACKUP_FEATURE_MASK))
 
+typedef enum dmu_send_resume_token_version {
+	ZFS_SEND_RESUME_TOKEN_VERSION = 1
+} dmu_send_resume_token_version_t;
+
 /*
  * The drr_versioninfo field of the dmu_replay_record has the
  * following layout:
@@ -335,6 +340,12 @@ typedef enum zfs_case {
 	ZFS_CASE_MIXED
 } zfs_case_t;
 
+/*
+ * Note: this struct must have the same layout in 32-bit and 64-bit, so
+ * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
+ * kernel.  Therefore, we add padding to it so that no "hidden" padding
+ * is automatically added on 64-bit (but not on 32-bit).
+ */
 typedef struct zfs_cmd {
 	char		zc_name[MAXPATHLEN];	/* name of pool or dataset */
 	uint64_t	zc_nvlist_src;		/* really (char *) */
@@ -363,14 +374,15 @@ typedef struct zfs_cmd {
 	uint64_t	zc_iflags;		/* internal to zfs(7fs) */
 	zfs_share_t	zc_share;
 	dmu_objset_stats_t zc_objset_stats;
-	struct drr_begin zc_begin_record;
+	dmu_replay_record_t zc_begin_record;
 	zinject_record_t zc_inject_record;
 	uint32_t	zc_defer_destroy;
 	uint32_t	zc_flags;
 	uint64_t	zc_action_handle;
 	int		zc_cleanup_fd;
 	uint8_t		zc_simple;
-	uint8_t		zc_pad[3];		/* alignment */
+	boolean_t	zc_resumable;
+	uint8_t		zc_pad[2];		/* alignment */
 	uint64_t	zc_sendobj;
 	uint64_t	zc_fromobj;
 	uint64_t	zc_createtxg;
diff --git a/include/sys/zvol.h b/include/sys/zvol.h
index c3e386f0b..00ed220d3 100644
--- a/include/sys/zvol.h
+++ b/include/sys/zvol.h
@@ -32,6 +32,8 @@
 #define	ZVOL_OBJ		1ULL
 #define	ZVOL_ZAP_OBJ		2ULL
 
+extern void *zvol_tag;
+
 extern void zvol_create_minors(spa_t *spa, const char *name, boolean_t async);
 extern void zvol_remove_minors(spa_t *spa, const char *name, boolean_t async);
 extern void zvol_rename_minors(spa_t *spa, const char *oldname,