Merge libzfs_core branch:

  includes MFV 238590, 238592, 247580

MFV 238590, 238592:
  In the first zfs ioctl restructuring phase, the libzfs_core library was
  introduced. It is a new thin library that wraps around kernel ioctl's.
  The idea is to provide a forward-compatible way of dealing with new
  features. Arguments are passed in nvlists and not random zfs_cmd fields,
  new-style ioctls are logged to pool history using a new method of
  history logging.

  http://blog.delphix.com/matt/2012/01/17/the-future-of-libzfs/

MFV 247580 [1]:
  To address issues of several deadlocks and race conditions the locking
  code around dsl_dataset was rewritten and the interface to synctasks
  was changed.

User-Visible Changes:
  "zfs snapshot" can create more arbitrary snapshots at once (atomically)
  "zfs destroy" destroys multiple snapshots at once
  "zfs recv" has improved performance

Backward Compatibility:
  I have extended the compatibility layer to support full backward
  compatibility by remapping or rewriting the responsible ioctl arguments.
  Old utilities are fully supported by the new kernel module.

Forward Compatibility:
  New utilities work with old kernels with the following restrictions:
    - creating, destroying, holding and releasing of multiple snapshots
      at once is not supported, this includes recursive (-r) commands

Illumos ZFS issues:
  2882 implement libzfs_core
  2900 "zfs snapshot" should be able to create multiple,
       arbitrary snapshots at once
  3464 zfs synctask code needs restructuring

References:
  https://www.illumos.org/issues/2882
  https://www.illumos.org/issues/2900
  https://www.illumos.org/issues/3464 [1]

MFC after:	1 month
Sponsored by:	Hybrid Logic Inc. [1]

2 files changed, 157 insertions, 80 deletions

diff --git a/cddl/contrib/opensolaris/cmd/zfs/zfs.8 b/cddl/contrib/opensolaris/cmd/zfs/zfs.8
index 9c882da..a7a51c9 100644
--- a/cddl/contrib/opensolaris/cmd/zfs/zfs.8
+++ b/cddl/contrib/opensolaris/cmd/zfs/zfs.8
@@ -28,7 +28,7 @@
 .\"
 .\" $FreeBSD$
 .\"
-.Dd March 1, 2013
+.Dd March 21, 2013
 .Dt ZFS 8
 .Os
 .Sh NAME
@@ -65,6 +65,7 @@
 .Op Fl r
 .Oo Fl o Ar property Ns = Ns Ar value Oc Ns ...
 .Ar filesystem@snapname Ns | Ns Ar volume@snapname
+.Ar filesystem@snapname Ns | Ns Ar volume@snapname Ns ...
 .Nm
 .Cm rollback
 .Op Fl rRf
@@ -1617,7 +1618,11 @@ multiple snapshots.
 Destroy (or mark for deferred deletion) all snapshots with this name in
 descendent file systems.
 .It Fl R
-Recursively destroy all dependents.
+Recursively destroy all clones of these snapshots, including the clones,
+snapshots, and children.
+If this flag is specified, the
+.Op fl d
+flag will have no effect.
 .It Fl n
 Do a dry-run ("No-op") deletion. No data will be deleted. This is useful in
 conjunction with the
@@ -1645,17 +1650,18 @@ behavior for mounted file systems in use.
 .Op Fl r
 .Oo Fl o Ar property Ns = Ns Ar value Oc Ns ...
 .Ar filesystem@snapname Ns | Ns volume@snapname
+.Ar filesystem@snapname Ns | Ns volume@snapname Ns ...
 .Xc
 .Pp
-Creates a snapshot with the given name. All previous modifications by
-successful system calls to the file system are part of the snapshot. See the
+Creates snapshots with the given names. All previous modifications by
+successful system calls to the file system are part of the snapshots.
+Snapshots are taken atomically, so that all snapshots correspond to the same
+moment in time. See the
 .Qq Sx Snapshots
 section for details.
 .Bl -tag -width indent
 .It Fl r
-Recursively create snapshots of all descendent datasets. Snapshots are taken
-atomically, so that all recursive snapshots correspond to the same moment in
-time.
+Recursively create snapshots of all descendent datasets
 .It Fl o Ar property Ns = Ns Ar value
 Sets the specified property; see
 .Qq Nm Cm create
diff --git a/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c b/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
index 1dfc82d..611740c 100644
--- a/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
+++ b/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
@@ -58,6 +58,7 @@
 #include <time.h>
 
 #include <libzfs.h>
+#include <libzfs_core.h>
 #include <zfs_prop.h>
 #include <zfs_deleg.h>
 #include <libuutil.h>
@@ -74,6 +75,7 @@ libzfs_handle_t *g_zfs;
 
 static FILE *mnttab_file;
 static char history_str[HIS_MAX_RECORD_LEN];
+static boolean_t log_history = B_TRUE;
 
 static int zfs_do_clone(int argc, char **argv);
 static int zfs_do_create(int argc, char **argv);
@@ -276,7 +278,7 @@ get_usage(zfs_help_t idx)
 		return (gettext("\tshare <-a | filesystem>\n"));
 	case HELP_SNAPSHOT:
 		return (gettext("\tsnapshot [-r] [-o property=value] ... "
-		    "<filesystem@snapname|volume@snapname>\n"));
+		    "<filesystem@snapname|volume@snapname> ...\n"));
 	case HELP_UNMOUNT:
 		return (gettext("\tunmount [-f] "
 		    "<-a | filesystem|mountpoint>\n"));
@@ -914,11 +916,12 @@ typedef struct destroy_cbdata {
 	boolean_t	cb_parsable;
 	boolean_t	cb_dryrun;
 	nvlist_t	*cb_nvl;
+	nvlist_t	*cb_batchedsnaps;
 
 	/* first snap in contiguous run */
-	zfs_handle_t	*cb_firstsnap;
+	char		*cb_firstsnap;
 	/* previous snap in contiguous run */
-	zfs_handle_t	*cb_prevsnap;
+	char		*cb_prevsnap;
 	int64_t		cb_snapused;
 	char		*cb_snapspec;
 } destroy_cbdata_t;
@@ -1010,9 +1013,27 @@ destroy_callback(zfs_handle_t *zhp, void *data)
 		zfs_close(zhp);
 		return (0);
 	}
+	if (cb->cb_dryrun) {
+		zfs_close(zhp);
+		return (0);
+	}
 
-	if (!cb->cb_dryrun) {
-		if (zfs_unmount(zhp, NULL, cb->cb_force ? MS_FORCE : 0) != 0 ||
+	/*
+	 * We batch up all contiguous snapshots (even of different
+	 * filesystems) and destroy them with one ioctl.  We can't
+	 * simply do all snap deletions and then all fs deletions,
+	 * because we must delete a clone before its origin.
+	 */
+	if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT) {
+		fnvlist_add_boolean(cb->cb_batchedsnaps, name);
+	} else {
+		int error = zfs_destroy_snaps_nvl(g_zfs,
+		    cb->cb_batchedsnaps, B_FALSE);
+		fnvlist_free(cb->cb_batchedsnaps);
+		cb->cb_batchedsnaps = fnvlist_alloc();
+
+		if (error != 0 ||
+		    zfs_unmount(zhp, NULL, cb->cb_force ? MS_FORCE : 0) != 0 ||
 		    zfs_destroy(zhp, cb->cb_defer_destroy) != 0) {
 			zfs_close(zhp);
 			return (-1);
@@ -1032,11 +1053,13 @@ destroy_print_cb(zfs_handle_t *zhp, void *arg)
 
 	if (nvlist_exists(cb->cb_nvl, name)) {
 		if (cb->cb_firstsnap == NULL)
-			cb->cb_firstsnap = zfs_handle_dup(zhp);
+			cb->cb_firstsnap = strdup(name);
 		if (cb->cb_prevsnap != NULL)
-			zfs_close(cb->cb_prevsnap);
+			free(cb->cb_prevsnap);
 		/* this snap continues the current range */
-		cb->cb_prevsnap = zfs_handle_dup(zhp);
+		cb->cb_prevsnap = strdup(name);
+		if (cb->cb_firstsnap == NULL || cb->cb_prevsnap == NULL)
+			nomem();
 		if (cb->cb_verbose) {
 			if (cb->cb_parsable) {
 				(void) printf("destroy\t%s\n", name);
@@ -1051,12 +1074,12 @@ destroy_print_cb(zfs_handle_t *zhp, void *arg)
 	} else if (cb->cb_firstsnap != NULL) {
 		/* end of this range */
 		uint64_t used = 0;
-		err = zfs_get_snapused_int(cb->cb_firstsnap,
+		err = lzc_snaprange_space(cb->cb_firstsnap,
 		    cb->cb_prevsnap, &used);
 		cb->cb_snapused += used;
-		zfs_close(cb->cb_firstsnap);
+		free(cb->cb_firstsnap);
 		cb->cb_firstsnap = NULL;
-		zfs_close(cb->cb_prevsnap);
+		free(cb->cb_prevsnap);
 		cb->cb_prevsnap = NULL;
 	}
 	zfs_close(zhp);
@@ -1073,13 +1096,13 @@ destroy_print_snapshots(zfs_handle_t *fs_zhp, destroy_cbdata_t *cb)
 	if (cb->cb_firstsnap != NULL) {
 		uint64_t used = 0;
 		if (err == 0) {
-			err = zfs_get_snapused_int(cb->cb_firstsnap,
+			err = lzc_snaprange_space(cb->cb_firstsnap,
 			    cb->cb_prevsnap, &used);
 		}
 		cb->cb_snapused += used;
-		zfs_close(cb->cb_firstsnap);
+		free(cb->cb_firstsnap);
 		cb->cb_firstsnap = NULL;
-		zfs_close(cb->cb_prevsnap);
+		free(cb->cb_prevsnap);
 		cb->cb_prevsnap = NULL;
 	}
 	return (err);
@@ -1166,8 +1189,10 @@ static int
 zfs_do_destroy(int argc, char **argv)
 {
 	destroy_cbdata_t cb = { 0 };
+	int rv = 0;
+	int err = 0;
 	int c;
-	zfs_handle_t *zhp;
+	zfs_handle_t *zhp = NULL;
 	char *at;
 	zfs_type_t type = ZFS_TYPE_DATASET;
 
@@ -1221,11 +1246,9 @@ zfs_do_destroy(int argc, char **argv)
 
 	at = strchr(argv[0], '@');
 	if (at != NULL) {
-		int err = 0;
 
 		/* Build the list of snaps to destroy in cb_nvl. */
-		if (nvlist_alloc(&cb.cb_nvl, NV_UNIQUE_NAME, 0) != 0)
-			nomem();
+		cb.cb_nvl = fnvlist_alloc();
 
 		*at = '\0';
 		zhp = zfs_open(g_zfs, argv[0],
@@ -1236,17 +1259,15 @@ zfs_do_destroy(int argc, char **argv)
 		cb.cb_snapspec = at + 1;
 		if (gather_snapshots(zfs_handle_dup(zhp), &cb) != 0 ||
 		    cb.cb_error) {
-			zfs_close(zhp);
-			nvlist_free(cb.cb_nvl);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
 		if (nvlist_empty(cb.cb_nvl)) {
 			(void) fprintf(stderr, gettext("could not find any "
 			    "snapshots to destroy; check snapshot names.\n"));
-			zfs_close(zhp);
-			nvlist_free(cb.cb_nvl);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
 		if (cb.cb_verbose) {
@@ -1265,18 +1286,26 @@ zfs_do_destroy(int argc, char **argv)
 		}
 
 		if (!cb.cb_dryrun) {
-			if (cb.cb_doclones)
+			if (cb.cb_doclones) {
+				cb.cb_batchedsnaps = fnvlist_alloc();
 				err = destroy_clones(&cb);
+				if (err == 0) {
+					err = zfs_destroy_snaps_nvl(g_zfs,
+					    cb.cb_batchedsnaps, B_FALSE);
+				}
+				if (err != 0) {
+					rv = 1;
+					goto out;
+				}
+			}
 			if (err == 0) {
-				err = zfs_destroy_snaps_nvl(zhp, cb.cb_nvl,
+				err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_nvl,
 				    cb.cb_defer_destroy);
 			}
 		}
 
-		zfs_close(zhp);
-		nvlist_free(cb.cb_nvl);
 		if (err != 0)
-			return (1);
+			rv = 1;
 	} else {
 		/* Open the given dataset */
 		if ((zhp = zfs_open(g_zfs, argv[0], type)) == NULL)
@@ -1297,8 +1326,8 @@ zfs_do_destroy(int argc, char **argv)
 			    zfs_get_name(zhp));
 			(void) fprintf(stderr, gettext("use 'zpool destroy %s' "
 			    "to destroy the pool itself\n"), zfs_get_name(zhp));
-			zfs_close(zhp);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
 		/*
@@ -1308,30 +1337,42 @@ zfs_do_destroy(int argc, char **argv)
 		if (!cb.cb_doclones &&
 		    zfs_iter_dependents(zhp, B_TRUE, destroy_check_dependent,
 		    &cb) != 0) {
-			zfs_close(zhp);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
 		if (cb.cb_error) {
-			zfs_close(zhp);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
+		cb.cb_batchedsnaps = fnvlist_alloc();
 		if (zfs_iter_dependents(zhp, B_FALSE, destroy_callback,
 		    &cb) != 0) {
-			zfs_close(zhp);
-			return (1);
+			rv = 1;
+			goto out;
 		}
 
 		/*
 		 * Do the real thing.  The callback will close the
 		 * handle regardless of whether it succeeds or not.
 		 */
-		if (destroy_callback(zhp, &cb) != 0)
-			return (1);
+		err = destroy_callback(zhp, &cb);
+		zhp = NULL;
+		if (err == 0) {
+			err = zfs_destroy_snaps_nvl(g_zfs,
+			    cb.cb_batchedsnaps, cb.cb_defer_destroy);
+		}
+		if (err != 0)
+			rv = 1;
 	}
 
-	return (0);
+out:
+	fnvlist_free(cb.cb_batchedsnaps);
+	fnvlist_free(cb.cb_nvl);
+	if (zhp != NULL)
+		zfs_close(zhp);
+	return (rv);
 }
 
 static boolean_t
@@ -1932,9 +1973,11 @@ upgrade_set_callback(zfs_handle_t *zhp, void *data)
 			/*
 			 * If they did "zfs upgrade -a", then we could
 			 * be doing ioctls to different pools.  We need
-			 * to log this history once to each pool.
+			 * to log this history once to each pool, and bypass
+			 * the normal history logging that happens in main().
 			 */
-			verify(zpool_stage_history(g_zfs, history_str) == 0);
+			(void) zpool_log_history(g_zfs, history_str);
+			log_history = B_FALSE;
 		}
 		if (zfs_prop_set(zhp, "version", verstr) == 0)
 			cb->cb_numupgraded++;
@@ -3472,6 +3515,32 @@ zfs_do_set(int argc, char **argv)
 	return (ret);
 }
 
+typedef struct snap_cbdata {
+	nvlist_t *sd_nvl;
+	boolean_t sd_recursive;
+	const char *sd_snapname;
+} snap_cbdata_t;
+
+static int
+zfs_snapshot_cb(zfs_handle_t *zhp, void *arg)
+{
+	snap_cbdata_t *sd = arg;
+	char *name;
+	int rv = 0;
+	int error;
+
+	error = asprintf(&name, "%s@%s", zfs_get_name(zhp), sd->sd_snapname);
+	if (error == -1)
+		nomem();
+	fnvlist_add_boolean(sd->sd_nvl, name);
+	free(name);
+
+	if (sd->sd_recursive)
+		rv = zfs_iter_filesystems(zhp, zfs_snapshot_cb, sd);
+	zfs_close(zhp);
+	return (rv);
+}
+
 /*
  * zfs snapshot [-r] [-o prop=value] ... <fs@snap>
  *
@@ -3481,13 +3550,16 @@ zfs_do_set(int argc, char **argv)
 static int
 zfs_do_snapshot(int argc, char **argv)
 {
-	boolean_t recursive = B_FALSE;
 	int ret = 0;
 	char c;
 	nvlist_t *props;
+	snap_cbdata_t sd = { 0 };
+	boolean_t multiple_snaps = B_FALSE;
 
 	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
 		nomem();
+	if (nvlist_alloc(&sd.sd_nvl, NV_UNIQUE_NAME, 0) != 0)
+		nomem();
 
 	/* check options */
 	while ((c = getopt(argc, argv, "ro:")) != -1) {
@@ -3497,7 +3569,8 @@ zfs_do_snapshot(int argc, char **argv)
 				return (1);
 			break;
 		case 'r':
-			recursive = B_TRUE;
+			sd.sd_recursive = B_TRUE;
+			multiple_snaps = B_TRUE;
 			break;
 		case '?':
 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
@@ -3514,18 +3587,35 @@ zfs_do_snapshot(int argc, char **argv)
 		(void) fprintf(stderr, gettext("missing snapshot argument\n"));
 		goto usage;
 	}
-	if (argc > 1) {
-		(void) fprintf(stderr, gettext("too many arguments\n"));
-		goto usage;
+
+	if (argc > 1)
+		multiple_snaps = B_TRUE;
+	for (; argc > 0; argc--, argv++) {
+		char *atp;
+		zfs_handle_t *zhp;
+
+		atp = strchr(argv[0], '@');
+		if (atp == NULL)
+			goto usage;
+		*atp = '\0';
+		sd.sd_snapname = atp + 1;
+		zhp = zfs_open(g_zfs, argv[0],
+		    ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
+		if (zhp == NULL)
+			goto usage;
+		if (zfs_snapshot_cb(zhp, &sd) != 0)
+			goto usage;
 	}
 
-	ret = zfs_snapshot(g_zfs, argv[0], recursive, props);
+	ret = zfs_snapshot_nvl(g_zfs, sd.sd_nvl, props);
+	nvlist_free(sd.sd_nvl);
 	nvlist_free(props);
-	if (ret && recursive)
+	if (ret != 0 && multiple_snaps)
 		(void) fprintf(stderr, gettext("no snapshots were created\n"));
 	return (ret != 0);
 
 usage:
+	nvlist_free(sd.sd_nvl);
 	nvlist_free(props);
 	usage(B_FALSE);
 	return (-1);
@@ -5068,28 +5158,12 @@ cleanup2:
 	return (error);
 }
 
-/*
- * zfs allow [-r] [-t] <tag> <snap> ...
- *
- *	-r	Recursively hold
- *	-t	Temporary hold (hidden option)
- *
- * Apply a user-hold with the given tag to the list of snapshots.
- */
 static int
 zfs_do_allow(int argc, char **argv)
 {
 	return (zfs_do_allow_unallow_impl(argc, argv, B_FALSE));
 }
 
-/*
- * zfs unallow [-r] [-t] <tag> <snap> ...
- *
- *	-r	Recursively hold
- *	-t	Temporary hold (hidden option)
- *
- * Apply a user-hold with the given tag to the list of snapshots.
- */
 static int
 zfs_do_unallow(int argc, char **argv)
 {
@@ -5103,7 +5177,6 @@ zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
 	int i;
 	const char *tag;
 	boolean_t recursive = B_FALSE;
-	boolean_t temphold = B_FALSE;
 	const char *opts = holding ? "rt" : "r";
 	int c;
 
@@ -5113,9 +5186,6 @@ zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
 		case 'r':
 			recursive = B_TRUE;
 			break;
-		case 't':
-			temphold = B_TRUE;
-			break;
 		case '?':
 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
 			    optopt);
@@ -5164,7 +5234,7 @@ zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
 		}
 		if (holding) {
 			if (zfs_hold(zhp, delim+1, tag, recursive,
-			    temphold, B_FALSE, -1, 0, 0) != 0)
+			    B_FALSE, -1) != 0)
 				++errors;
 		} else {
 			if (zfs_release(zhp, delim+1, tag, recursive) != 0)
@@ -5180,7 +5250,6 @@ zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
  * zfs hold [-r] [-t] <tag> <snap> ...
  *
  *	-r	Recursively hold
- *	-t	Temporary hold (hidden option)
  *
  * Apply a user-hold with the given tag to the list of snapshots.
  */
@@ -6602,8 +6671,7 @@ main(int argc, char **argv)
 		return (1);
 	}
 
-	zpool_set_history_str("zfs", argc, argv, history_str);
-	verify(zpool_stage_history(g_zfs, history_str) == 0);
+	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));
 
 	libzfs_print_on_error(g_zfs, B_TRUE);
 
@@ -6672,6 +6740,9 @@ main(int argc, char **argv)
 
 	(void) fclose(mnttab_file);
 
+	if (ret == 0 && log_history)
+		(void) zpool_log_history(g_zfs, history_str);
+
 	libzfs_fini(g_zfs);
 
 	/*