Merge commit 'refs/top-bases/feature-branch' into feature-branch

ZFS.RELEASE

@@ -1 +1 @@
-http://dlc.sun.com/osol/on/downloads/b108/on-src.tar.bz2
+http://dlc.sun.com/osol/on/downloads/b117/on-src.tar.bz2

cmd/zdb/zdb.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -101,7 +101,9 @@ usage(void)
 	(void) fprintf(stderr, "        -C cached pool configuration\n");
 	(void) fprintf(stderr, "	-i intent logs\n");
 	(void) fprintf(stderr, "	-b block statistics\n");
-	(void) fprintf(stderr, "	-c checksum all data blocks\n");
+	(void) fprintf(stderr, "	-m metaslabs\n");
+	(void) fprintf(stderr, "	-c checksum all metadata (twice for "
+	    "all data) blocks\n");
 	(void) fprintf(stderr, "	-s report stats on zdb's I/O\n");
 	(void) fprintf(stderr, "	-S <user|all>:<cksum_alg|all> -- "
 	    "dump blkptr signatures\n");
@@ -124,6 +126,11 @@ usage(void)
 	exit(1);
 }
 
+/*
+ * Called for usage errors that are discovered after a call to spa_open(),
+ * dmu_bonus_hold(), or pool_match().  abort() is called for other errors.
+ */
+
 static void
 fatal(const char *fmt, ...)
 {
@@ -135,7 +142,7 @@ fatal(const char *fmt, ...)
 	va_end(ap);
 	(void) fprintf(stderr, "\n");
 
-	abort();
+	exit(1);
 }
 
 static void
@@ -208,7 +215,7 @@ dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size)
 	size_t nvsize = *(uint64_t *)data;
 	char *packed = umem_alloc(nvsize, UMEM_NOFAIL);
 
-	VERIFY(0 == dmu_read(os, object, 0, nvsize, packed));
+	VERIFY(0 == dmu_read(os, object, 0, nvsize, packed, DMU_READ_PREFETCH));
 
 	VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0);
 
@@ -434,7 +441,7 @@ dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
 	alloc = 0;
 	for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) {
 		VERIFY(0 == dmu_read(os, smo->smo_object, offset,
-		    sizeof (entry), &entry));
+		    sizeof (entry), &entry, DMU_READ_PREFETCH));
 		if (SM_DEBUG_DECODE(entry)) {
 			(void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n",
 			    (u_longlong_t)(offset / sizeof (entry)),
@@ -465,6 +472,21 @@ dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
 	}
 }
 
+static void
+dump_metaslab_stats(metaslab_t *msp)
+{
+	char maxbuf[5];
+	space_map_t *sm = &msp->ms_map;
+	avl_tree_t *t = sm->sm_pp_root;
+	int free_pct = sm->sm_space * 100 / sm->sm_size;
+
+	nicenum(space_map_maxsize(sm), maxbuf);
+
+	(void) printf("\t %20s %10lu   %7s  %6s   %4s %4d%%\n",
+	    "segments", avl_numnodes(t), "maxsize", maxbuf,
+	    "freepct", free_pct);
+}
+
 static void
 dump_metaslab(metaslab_t *msp)
 {
@@ -475,22 +497,28 @@ dump_metaslab(metaslab_t *msp)
 
 	nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf);
 
-	if (dump_opt['d'] <= 5) {
-		(void) printf("\t%10llx   %10llu   %5s\n",
-		    (u_longlong_t)msp->ms_map.sm_start,
-		    (u_longlong_t)smo->smo_object,
-		    freebuf);
-		return;
-	}
-
 	(void) printf(
-	    "\tvdev %llu   offset %08llx   spacemap %4llu   free %5s\n",
+	    "\tvdev %5llu   offset %12llx   spacemap %6llu   free    %5s\n",
 	    (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start,
 	    (u_longlong_t)smo->smo_object, freebuf);
 
-	ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift));
+	if (dump_opt['m'] > 1) {
+		mutex_enter(&msp->ms_lock);
+		VERIFY(space_map_load(&msp->ms_map, zfs_metaslab_ops,
+		    SM_FREE, &msp->ms_smo, spa->spa_meta_objset) == 0);
+		dump_metaslab_stats(msp);
+		space_map_unload(&msp->ms_map);
+		mutex_exit(&msp->ms_lock);
+	}
+
+	if (dump_opt['d'] > 5 || dump_opt['m'] > 2) {
+		ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift));
+
+		mutex_enter(&msp->ms_lock);
+		dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map);
+		mutex_exit(&msp->ms_lock);
+	}
 
-	dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map);
 }
 
 static void
@@ -505,14 +533,12 @@ dump_metaslabs(spa_t *spa)
 	for (c = 0; c < rvd->vdev_children; c++) {
 		vd = rvd->vdev_child[c];
 
-		(void) printf("\n    vdev %llu\n\n", (u_longlong_t)vd->vdev_id);
+		(void) printf("\t%-10s   %-19s   %-15s   %-10s\n",
+		    "vdev", "offset", "spacemap", "free");
+		(void) printf("\t%10s   %19s   %15s   %10s\n",
+		    "----------", "-------------------",
+		    "---------------", "-------------");
 
-		if (dump_opt['d'] <= 5) {
-			(void) printf("\t%10s   %10s   %5s\n",
-			    "offset", "spacemap", "free");
-			(void) printf("\t%10s   %10s   %5s\n",
-			    "------", "--------", "----");
-		}
 		for (m = 0; m < vd->vdev_ms_count; m++)
 			dump_metaslab(vd->vdev_ms[m]);
 		(void) printf("\n");
@@ -916,6 +942,7 @@ dump_uidgid(objset_t *os, znode_phys_t *zp)
 		/* first find the fuid object.  It lives in the master node */
 		VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES,
 		    8, 1, &fuid_obj) == 0);
+		zfs_fuid_avl_tree_create(&idx_tree, &domain_tree);
 		(void) zfs_fuid_table_load(os, fuid_obj,
 		    &idx_tree, &domain_tree);
 		fuid_table_loaded = B_TRUE;
@@ -1019,6 +1046,8 @@ static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = {
 	dump_packed_nvlist,	/* FUID nvlist size		*/
 	dump_zap,		/* DSL dataset next clones	*/
 	dump_zap,		/* DSL scrub queue		*/
+	dump_zap,		/* ZFS user/group used		*/
+	dump_zap,		/* ZFS user/group quota		*/
 };
 
 static void
@@ -1082,6 +1111,14 @@ dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header)
 	}
 
 	if (verbosity >= 4) {
+		(void) printf("\tdnode flags: %s%s\n",
+		    (dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) ?
+		    "USED_BYTES " : "",
+		    (dn->dn_phys->dn_flags & DNODE_FLAG_USERUSED_ACCOUNTED) ?
+		    "USERUSED_ACCOUNTED " : "");
+		(void) printf("\tdnode maxblkid: %llu\n",
+		    (longlong_t)dn->dn_phys->dn_maxblkid);
+
 		object_viewer[doi.doi_bonus_type](os, object, bonus, bsize);
 		object_viewer[doi.doi_type](os, object, NULL, 0);
 		*print_header = 1;
@@ -1136,7 +1173,7 @@ dump_dir(objset_t *os)
 	uint64_t object, object_count;
 	uint64_t refdbytes, usedobjs, scratch;
 	char numbuf[8];
-	char blkbuf[BP_SPRINTF_LEN];
+	char blkbuf[BP_SPRINTF_LEN + 20];
 	char osname[MAXNAMELEN];
 	char *type = "UNKNOWN";
 	int verbosity = dump_opt['d'];
@@ -1162,8 +1199,8 @@ dump_dir(objset_t *os)
 	nicenum(refdbytes, numbuf);
 
 	if (verbosity >= 4) {
-		(void) strcpy(blkbuf, ", rootbp ");
+		(void) sprintf(blkbuf + strlen(blkbuf), ", rootbp ");
-		sprintf_blkptr(blkbuf + strlen(blkbuf),
+		(void) sprintf_blkptr(blkbuf + strlen(blkbuf),
 		    BP_SPRINTF_LEN - strlen(blkbuf), os->os->os_rootbp);
 	} else {
 		blkbuf[0] = '\0';
@@ -1198,7 +1235,12 @@ dump_dir(objset_t *os)
 	}
 
 	dump_object(os, 0, verbosity, &print_header);
-	object_count = 1;
+	object_count = 0;
+	if (os->os->os_userused_dnode &&
+	    os->os->os_userused_dnode->dn_type != 0) {
+		dump_object(os, DMU_USERUSED_OBJECT, verbosity, &print_header);
+		dump_object(os, DMU_GROUPUSED_OBJECT, verbosity, &print_header);
+	}
 
 	object = 0;
 	while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) {
@@ -1210,8 +1252,10 @@ dump_dir(objset_t *os)
 
 	(void) printf("\n");
 
-	if (error != ESRCH)
+	if (error != ESRCH) {
-		fatal("dmu_object_next() = %d", error);
+		(void) fprintf(stderr, "dmu_object_next() = %d\n", error);
+		abort();
+	}
 }
 
 static void
@@ -1394,7 +1438,8 @@ static space_map_ops_t zdb_space_map_ops = {
 	zdb_space_map_unload,
 	NULL,	/* alloc */
 	zdb_space_map_claim,
-	NULL	/* free */
+	NULL,	/* free */
+	NULL	/* maxsize */
 };
 
 static void
@@ -1504,13 +1549,25 @@ zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
 {
 	zdb_cb_t *zcb = arg;
 	char blkbuf[BP_SPRINTF_LEN];
+	dmu_object_type_t type;
+	boolean_t is_l0_metadata;
 
 	if (bp == NULL)
 		return (0);
 
-	zdb_count_block(spa, zcb, bp, BP_GET_TYPE(bp));
+	type = BP_GET_TYPE(bp);
 
-	if (dump_opt['c'] || dump_opt['S']) {
+	zdb_count_block(spa, zcb, bp, type);
+
+	/*
+	 * if we do metadata-only checksumming there's no need to checksum
+	 * indirect blocks here because it is done during traverse
+	 */
+	is_l0_metadata = (BP_GET_LEVEL(bp) == 0 && type < DMU_OT_NUMTYPES &&
+	    dmu_ot[type].ot_metadata);
+
+	if (dump_opt['c'] > 1 || dump_opt['S'] ||
+	    (dump_opt['c'] && is_l0_metadata)) {
 		int ioerr, size;
 		void *data;
 
@@ -1522,7 +1579,7 @@ zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
 		free(data);
 
 		/* We expect io errors on intent log */
-		if (ioerr && BP_GET_TYPE(bp) != DMU_OT_INTENT_LOG) {
+		if (ioerr && type != DMU_OT_INTENT_LOG) {
 			zcb->zcb_haderrors = 1;
 			zcb->zcb_errors[ioerr]++;
 
@@ -1570,8 +1627,9 @@ dump_block_stats(spa_t *spa)
 	int c, e;
 
 	if (!dump_opt['S']) {
-		(void) printf("\nTraversing all blocks %s%s%s%s...\n",
+		(void) printf("\nTraversing all blocks %s%s%s%s%s...\n",
 		    (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "",
+		    (dump_opt['c'] == 1) ? "metadata " : "",
 		    dump_opt['c'] ? "checksums " : "",
 		    (dump_opt['c'] && !dump_opt['L']) ? "and verify " : "",
 		    !dump_opt['L'] ? "nothing leaked " : "");
@@ -1771,14 +1829,17 @@ dump_zpool(spa_t *spa)
 	if (dump_opt['u'])
 		dump_uberblock(&spa->spa_uberblock);
 
-	if (dump_opt['d'] || dump_opt['i']) {
+	if (dump_opt['d'] || dump_opt['i'] || dump_opt['m']) {
 		dump_dir(dp->dp_meta_objset);
 		if (dump_opt['d'] >= 3) {
 			dump_bplist(dp->dp_meta_objset,
 			    spa->spa_sync_bplist_obj, "Deferred frees");
 			dump_dtl(spa->spa_root_vdev, 0);
-			dump_metaslabs(spa);
 		}
+
+		if (dump_opt['d'] >= 3 || dump_opt['m'])
+			dump_metaslabs(spa);
+
 		(void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL,
 		    DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
 	}
@@ -2254,13 +2315,14 @@ main(int argc, char **argv)
 
 	dprintf_setup(&argc, argv);
 
-	while ((c = getopt(argc, argv, "udibcsvCLS:U:lRep:t:")) != -1) {
+	while ((c = getopt(argc, argv, "udibcmsvCLS:U:lRep:t:")) != -1) {
 		switch (c) {
 		case 'u':
 		case 'd':
 		case 'i':
 		case 'b':
 		case 'c':
+		case 'm':
 		case 's':
 		case 'C':
 		case 'l':
@@ -2396,7 +2458,7 @@ main(int argc, char **argv)
 			}
 
 			if (error == 0)
-				error = spa_import_faulted(argv[0],
+				error = spa_import_verbatim(argv[0],
 				    exported_conf, nvl);
 
 			nvlist_free(nvl);

cmd/zdb/zdb_il.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * Print intent log header and statistics.
  */
@@ -345,8 +343,10 @@ dump_intent_log(zilog_t *zilog)
 	if (zh->zh_log.blk_birth == 0 || verbose < 2)
 		return;
 
-	(void) printf("\n    ZIL header: claim_txg %llu, seq %llu\n",
+	(void) printf("\n    ZIL header: claim_txg %llu, claim_seq %llu",
-	    (u_longlong_t)zh->zh_claim_txg, (u_longlong_t)zh->zh_replay_seq);
+	    (u_longlong_t)zh->zh_claim_txg, (u_longlong_t)zh->zh_claim_seq);
+	(void) printf(" replay_seq %llu, flags 0x%llx\n",
+	    (u_longlong_t)zh->zh_replay_seq, (u_longlong_t)zh->zh_flags);
 
 	if (verbose >= 4)
 		print_log_bp(&zh->zh_log, "\n\tfirst block: ");

cmd/zfs/zfs_iter.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -53,11 +53,14 @@ typedef struct zfs_node {
 } zfs_node_t;
 
 typedef struct callback_data {
-	uu_avl_t	*cb_avl;
+	uu_avl_t		*cb_avl;
-	int		cb_flags;
+	int			cb_flags;
-	zfs_type_t	cb_types;
+	zfs_type_t		cb_types;
-	zfs_sort_column_t *cb_sortcol;
+	zfs_sort_column_t	*cb_sortcol;
-	zprop_list_t	**cb_proplist;
+	zprop_list_t		**cb_proplist;
+	int			cb_depth_limit;
+	int			cb_depth;
+	uint8_t			cb_props_table[ZFS_NUM_PROPS];
 } callback_data_t;
 
 uu_avl_pool_t *avl_pool;
@@ -98,10 +101,17 @@ zfs_callback(zfs_handle_t *zhp, void *data)
 		uu_avl_node_init(node, &node->zn_avlnode, avl_pool);
 		if (uu_avl_find(cb->cb_avl, node, cb->cb_sortcol,
 		    &idx) == NULL) {
-			if (cb->cb_proplist &&
+			if (cb->cb_proplist) {
-			    zfs_expand_proplist(zhp, cb->cb_proplist) != 0) {
+				if ((*cb->cb_proplist) &&
-				free(node);
+				    !(*cb->cb_proplist)->pl_all)
-				return (-1);
+					zfs_prune_proplist(zhp,
+					    cb->cb_props_table);
+
+				if (zfs_expand_proplist(zhp, cb->cb_proplist)
+				    != 0) {
+					free(node);
+					return (-1);
+				}
 			}
 			uu_avl_insert(cb->cb_avl, node, idx);
 			dontclose = 1;
@@ -113,11 +123,15 @@ zfs_callback(zfs_handle_t *zhp, void *data)
 	/*
 	 * Recurse if necessary.
 	 */
-	if (cb->cb_flags & ZFS_ITER_RECURSE) {
+	if (cb->cb_flags & ZFS_ITER_RECURSE &&
+	    ((cb->cb_flags & ZFS_ITER_DEPTH_LIMIT) == 0 ||
+	    cb->cb_depth < cb->cb_depth_limit)) {
+		cb->cb_depth++;
 		if (zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM)
 			(void) zfs_iter_filesystems(zhp, zfs_callback, data);
 		if ((zfs_get_type(zhp) != ZFS_TYPE_SNAPSHOT) && include_snaps)
 			(void) zfs_iter_snapshots(zhp, zfs_callback, data);
+		cb->cb_depth--;
 	}
 
 	if (!dontclose)
@@ -325,10 +339,10 @@ zfs_sort(const void *larg, const void *rarg, void *data)
 
 int
 zfs_for_each(int argc, char **argv, int flags, zfs_type_t types,
-    zfs_sort_column_t *sortcol, zprop_list_t **proplist,
+    zfs_sort_column_t *sortcol, zprop_list_t **proplist, int limit,
     zfs_iter_f callback, void *data)
 {
-	callback_data_t cb;
+	callback_data_t cb = {0};
 	int ret = 0;
 	zfs_node_t *node;
 	uu_avl_walk_t *walk;
@@ -346,6 +360,45 @@ zfs_for_each(int argc, char **argv, int flags, zfs_type_t types,
 	cb.cb_flags = flags;
 	cb.cb_proplist = proplist;
 	cb.cb_types = types;
+	cb.cb_depth_limit = limit;
+	/*
+	 * If cb_proplist is provided then in the zfs_handles created  we
+	 * retain only those properties listed in cb_proplist and sortcol.
+	 * The rest are pruned. So, the caller should make sure that no other
+	 * properties other than those listed in cb_proplist/sortcol are
+	 * accessed.
+	 *
+	 * If cb_proplist is NULL then we retain all the properties.  We
+	 * always retain the zoned property, which some other properties
+	 * need (userquota & friends), and the createtxg property, which
+	 * we need to sort snapshots.
+	 */
+	if (cb.cb_proplist && *cb.cb_proplist) {
+		zprop_list_t *p = *cb.cb_proplist;
+
+		while (p) {
+			if (p->pl_prop >= ZFS_PROP_TYPE &&
+			    p->pl_prop < ZFS_NUM_PROPS) {
+				cb.cb_props_table[p->pl_prop] = B_TRUE;
+			}
+			p = p->pl_next;
+		}
+
+		while (sortcol) {
+			if (sortcol->sc_prop >= ZFS_PROP_TYPE &&
+			    sortcol->sc_prop < ZFS_NUM_PROPS) {
+				cb.cb_props_table[sortcol->sc_prop] = B_TRUE;
+			}
+			sortcol = sortcol->sc_next;
+		}
+
+		cb.cb_props_table[ZFS_PROP_ZONED] = B_TRUE;
+		cb.cb_props_table[ZFS_PROP_CREATETXG] = B_TRUE;
+	} else {
+		(void) memset(cb.cb_props_table, B_TRUE,
+		    sizeof (cb.cb_props_table));
+	}
+
 	if ((cb.cb_avl = uu_avl_create(avl_pool, NULL, UU_DEFAULT)) == NULL) {
 		(void) fprintf(stderr,
 		    gettext("internal error: out of memory\n"));

cmd/zfs/zfs_iter.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -41,9 +41,10 @@ typedef struct zfs_sort_column {
 #define	ZFS_ITER_RECURSE	   (1 << 0)
 #define	ZFS_ITER_ARGS_CAN_BE_PATHS (1 << 1)
 #define	ZFS_ITER_PROP_LISTSNAPS    (1 << 2)
+#define	ZFS_ITER_DEPTH_LIMIT	   (1 << 3)
 
 int zfs_for_each(int, char **, int options, zfs_type_t,
-    zfs_sort_column_t *, zprop_list_t **, zfs_iter_f, void *);
+    zfs_sort_column_t *, zprop_list_t **, int, zfs_iter_f, void *);
 int zfs_add_sort_column(zfs_sort_column_t **, const char *, boolean_t);
 void zfs_free_sort_columns(zfs_sort_column_t *);
 

cmd/zfs/zfs_main.c

@@ -39,12 +39,14 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <zone.h>
+#include <grp.h>
+#include <pwd.h>
 #include <sys/mkdev.h>
 #include <sys/mntent.h>
 #include <sys/mnttab.h>
 #include <sys/mount.h>
 #include <sys/stat.h>
-#include <sys/avl.h>
+#include <sys/fs/zfs.h>
 
 #include <libzfs.h>
 #include <libuutil.h>
@@ -56,6 +58,7 @@ libzfs_handle_t *g_zfs;
 
 static FILE *mnttab_file;
 static char history_str[HIS_MAX_RECORD_LEN];
+const char *pypath = "/usr/lib/zfs/pyzfs.py";
 
 static int zfs_do_clone(int argc, char **argv);
 static int zfs_do_create(int argc, char **argv);
@@ -75,8 +78,8 @@ static int zfs_do_unshare(int argc, char **argv);
 static int zfs_do_send(int argc, char **argv);
 static int zfs_do_receive(int argc, char **argv);
 static int zfs_do_promote(int argc, char **argv);
-static int zfs_do_allow(int argc, char **argv);
+static int zfs_do_userspace(int argc, char **argv);
-static int zfs_do_unallow(int argc, char **argv);
+static int zfs_do_python(int argc, char **argv);
 
 /*
  * Enable a reasonable set of defaults for libumem debugging on DEBUG builds.
@@ -116,7 +119,9 @@ typedef enum {
 	HELP_UNMOUNT,
 	HELP_UNSHARE,
 	HELP_ALLOW,
-	HELP_UNALLOW
+	HELP_UNALLOW,
+	HELP_USERSPACE,
+	HELP_GROUPSPACE
 } zfs_help_t;
 
 typedef struct zfs_command {
@@ -150,6 +155,8 @@ static zfs_command_t command_table[] = {
 	{ "get", 	zfs_do_get,		HELP_GET		},
 	{ "inherit",	zfs_do_inherit,		HELP_INHERIT		},
 	{ "upgrade",	zfs_do_upgrade,		HELP_UPGRADE		},
+	{ "userspace",	zfs_do_userspace,	HELP_USERSPACE		},
+	{ "groupspace",	zfs_do_userspace,	HELP_GROUPSPACE		},
 	{ NULL },
 	{ "mount",	zfs_do_mount,		HELP_MOUNT		},
 	{ "unmount",	zfs_do_unmount,		HELP_UNMOUNT		},
@@ -159,9 +166,9 @@ static zfs_command_t command_table[] = {
 	{ "send",	zfs_do_send,		HELP_SEND		},
 	{ "receive",	zfs_do_receive,		HELP_RECEIVE		},
 	{ NULL },
-	{ "allow",	zfs_do_allow,		HELP_ALLOW		},
+	{ "allow",	zfs_do_python,		HELP_ALLOW		},
 	{ NULL },
-	{ "unallow",	zfs_do_unallow,		HELP_UNALLOW		},
+	{ "unallow",	zfs_do_python,		HELP_UNALLOW		},
 };
 
 #define	NCOMMAND	(sizeof (command_table) / sizeof (command_table[0]))
@@ -184,8 +191,8 @@ get_usage(zfs_help_t idx)
 		return (gettext("\tdestroy [-rRf] "
 		    "<filesystem|volume|snapshot>\n"));
 	case HELP_GET:
-		return (gettext("\tget [-rHp] [-o field[,...]] "
+		return (gettext("\tget [-rHp] [-d max] "
-		    "[-s source[,...]]\n"
+		    "[-o field[,...]] [-s source[,...]]\n"
 		    "\t    <\"all\" | property[,...]> "
 		    "[filesystem|volume|snapshot] ...\n"));
 	case HELP_INHERIT:
@@ -195,8 +202,8 @@ get_usage(zfs_help_t idx)
 		return (gettext("\tupgrade [-v]\n"
 		    "\tupgrade [-r] [-V version] <-a | filesystem ...>\n"));
 	case HELP_LIST:
-		return (gettext("\tlist [-rH] [-o property[,...]] "
+		return (gettext("\tlist [-rH][-d max] "
-		    "[-t type[,...]] [-s property] ...\n"
+		    "[-o property[,...]] [-t type[,...]] [-s property] ...\n"
 		    "\t    [-S property] ... "
 		    "[filesystem|volume|snapshot] ...\n"));
 	case HELP_MOUNT:
@@ -232,7 +239,8 @@ get_usage(zfs_help_t idx)
 		return (gettext("\tunshare [-f] "
 		    "<-a | filesystem|mountpoint>\n"));
 	case HELP_ALLOW:
-		return (gettext("\tallow [-ldug] "
+		return (gettext("\tallow <filesystem|volume>\n"
+		    "\tallow [-ldug] "
 		    "<\"everyone\"|user|group>[,...] <perm|@setname>[,...]\n"
 		    "\t    <filesystem|volume>\n"
 		    "\tallow [-ld] -e <perm|@setname>[,...] "
@@ -250,6 +258,14 @@ get_usage(zfs_help_t idx)
 		    "<filesystem|volume>\n"
 		    "\tunallow [-r] -s @setname [<perm|@setname>[,...]] "
 		    "<filesystem|volume>\n"));
+	case HELP_USERSPACE:
+		return (gettext("\tuserspace [-hniHp] [-o field[,...]] "
+		    "[-sS field] ... [-t type[,...]]\n"
+		    "\t    <filesystem|snapshot>\n"));
+	case HELP_GROUPSPACE:
+		return (gettext("\tgroupspace [-hniHpU] [-o field[,...]] "
+		    "[-sS field] ... [-t type[,...]]\n"
+		    "\t    <filesystem|snapshot>\n"));
 	}
 
 	abort();
@@ -311,7 +327,6 @@ usage(boolean_t requested)
 {
 	int i;
 	boolean_t show_properties = B_FALSE;
-	boolean_t show_permissions = B_FALSE;
 	FILE *fp = requested ? stdout : stderr;
 
 	if (current_command == NULL) {
@@ -342,13 +357,7 @@ usage(boolean_t requested)
 	    strcmp(current_command->name, "list") == 0))
 		show_properties = B_TRUE;
 
-	if (current_command != NULL &&
-	    (strcmp(current_command->name, "allow") == 0 ||
-	    strcmp(current_command->name, "unallow") == 0))
-		show_permissions = B_TRUE;
-
 	if (show_properties) {
-
 		(void) fprintf(fp,
 		    gettext("\nThe following properties are supported:\n"));
 
@@ -359,16 +368,26 @@ usage(boolean_t requested)
 		(void) zprop_iter(usage_prop_cb, fp, B_FALSE, B_TRUE,
 		    ZFS_TYPE_DATASET);
 
+		(void) fprintf(fp, "\t%-15s ", "userused@...");
+		(void) fprintf(fp, " NO       NO   <size>\n");
+		(void) fprintf(fp, "\t%-15s ", "groupused@...");
+		(void) fprintf(fp, " NO       NO   <size>\n");
+		(void) fprintf(fp, "\t%-15s ", "userquota@...");
+		(void) fprintf(fp, "YES       NO   <size> | none\n");
+		(void) fprintf(fp, "\t%-15s ", "groupquota@...");
+		(void) fprintf(fp, "YES       NO   <size> | none\n");
+
 		(void) fprintf(fp, gettext("\nSizes are specified in bytes "
 		    "with standard units such as K, M, G, etc.\n"));
 		(void) fprintf(fp, gettext("\nUser-defined properties can "
 		    "be specified by using a name containing a colon (:).\n"));
-
+		(void) fprintf(fp, gettext("\nThe {user|group}{used|quota}@ "
-	} else if (show_permissions) {
+		    "properties must be appended with\n"
-		(void) fprintf(fp,
+		    "a user or group specifier of one of these forms:\n"
-		    gettext("\nThe following permissions are supported:\n"));
+		    "    POSIX name      (eg: \"matt\")\n"
-
+		    "    POSIX id        (eg: \"126829\")\n"
-		zfs_deleg_permissions();
+		    "    SMB name@domain (eg: \"matt@sun\")\n"
+		    "    SMB SID         (eg: \"S-1-234-567-89\")\n"));
 	} else {
 		(void) fprintf(fp,
 		    gettext("\nFor the property list, run: %s\n"),
@@ -415,6 +434,27 @@ parseprop(nvlist_t *props)
 	return (0);
 }
 
+static int
+parse_depth(char *opt, int *flags)
+{
+	char *tmp;
+	int depth;
+
+	depth = (int)strtol(opt, &tmp, 0);
+	if (*tmp) {
+		(void) fprintf(stderr,
+		    gettext("%s is not an integer\n"), optarg);
+		usage(B_FALSE);
+	}
+	if (depth < 0) {
+		(void) fprintf(stderr,
+		    gettext("Depth can not be negative.\n"));
+		usage(B_FALSE);
+	}
+	*flags |= (ZFS_ITER_DEPTH_LIMIT|ZFS_ITER_RECURSE);
+	return (depth);
+}
+
 /*
  * zfs clone [-p] [-o prop=value] ... <snap> <fs | vol>
  *
@@ -1063,6 +1103,17 @@ get_callback(zfs_handle_t *zhp, void *data)
 			zprop_print_one_property(zfs_get_name(zhp), cbp,
 			    zfs_prop_to_name(pl->pl_prop),
 			    buf, sourcetype, source);
+		} else if (zfs_prop_userquota(pl->pl_user_prop)) {
+			sourcetype = ZPROP_SRC_LOCAL;
+
+			if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
+			    buf, sizeof (buf), cbp->cb_literal) != 0) {
+				sourcetype = ZPROP_SRC_NONE;
+				(void) strlcpy(buf, "-", sizeof (buf));
+			}
+
+			zprop_print_one_property(zfs_get_name(zhp), cbp,
+			    pl->pl_user_prop, buf, sourcetype, source);
 		} else {
 			if (nvlist_lookup_nvlist(userprop,
 			    pl->pl_user_prop, &propval) != 0) {
@@ -1102,6 +1153,7 @@ zfs_do_get(int argc, char **argv)
 	int i, c, flags = 0;
 	char *value, *fields;
 	int ret;
+	int limit = 0;
 	zprop_list_t fake_name = { 0 };
 
 	/*
@@ -1115,11 +1167,14 @@ zfs_do_get(int argc, char **argv)
 	cb.cb_type = ZFS_TYPE_DATASET;
 
 	/* check options */
-	while ((c = getopt(argc, argv, ":o:s:rHp")) != -1) {
+	while ((c = getopt(argc, argv, ":d:o:s:rHp")) != -1) {
 		switch (c) {
 		case 'p':
 			cb.cb_literal = B_TRUE;
 			break;
+		case 'd':
+			limit = parse_depth(optarg, &flags);
+			break;
 		case 'r':
 			flags |= ZFS_ITER_RECURSE;
 			break;
@@ -1250,7 +1305,7 @@ zfs_do_get(int argc, char **argv)
 
 	/* run for each object */
 	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET, NULL,
-	    &cb.cb_proplist, get_callback, &cb);
+	    &cb.cb_proplist, limit, get_callback, &cb);
 
 	if (cb.cb_proplist == &fake_name)
 		zprop_free_list(fake_name.pl_next);
@@ -1363,10 +1418,10 @@ zfs_do_inherit(int argc, char **argv)
 
 	if (flags & ZFS_ITER_RECURSE) {
 		ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
-		    NULL, NULL, inherit_recurse_cb, propname);
+		    NULL, NULL, 0, inherit_recurse_cb, propname);
 	} else {
 		ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
-		    NULL, NULL, inherit_cb, propname);
+		    NULL, NULL, 0, inherit_cb, propname);
 	}
 
 	return (ret);
@@ -1435,21 +1490,30 @@ upgrade_set_callback(zfs_handle_t *zhp, void *data)
 {
 	upgrade_cbdata_t *cb = data;
 	int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
+	int i;
+	static struct { int zplver; int spaver; } table[] = {
+		{ZPL_VERSION_FUID, SPA_VERSION_FUID},
+		{ZPL_VERSION_USERSPACE, SPA_VERSION_USERSPACE},
+		{0, 0}
+	};
 
-	if (cb->cb_version >= ZPL_VERSION_FUID) {
-		int spa_version;
 
-		if (zfs_spa_version(zhp, &spa_version) < 0)
+	for (i = 0; table[i].zplver; i++) {
-			return (-1);
+		if (cb->cb_version >= table[i].zplver) {
+			int spa_version;
 
-		if (spa_version < SPA_VERSION_FUID) {
+			if (zfs_spa_version(zhp, &spa_version) < 0)
-			/* can't upgrade */
+				return (-1);
-			(void) printf(gettext("%s: can not be upgraded; "
+
-			    "the pool version needs to first be upgraded\nto "
+			if (spa_version < table[i].spaver) {
-			    "version %d\n\n"),
+				/* can't upgrade */
-			    zfs_get_name(zhp), SPA_VERSION_FUID);
+				(void) printf(gettext("%s: can not be "
-			cb->cb_numfailed++;
+				    "upgraded; the pool version needs to first "
-			return (0);
+				    "be upgraded\nto version %d\n\n"),
+				    zfs_get_name(zhp), table[i].spaver);
+				cb->cb_numfailed++;
+				return (0);
+			}
 		}
 	}
 
@@ -1550,6 +1614,8 @@ zfs_do_upgrade(int argc, char **argv)
 		(void) printf(gettext(" 2   Enhanced directory entries\n"));
 		(void) printf(gettext(" 3   Case insensitive and File system "
 		    "unique identifer (FUID)\n"));
+		(void) printf(gettext(" 4   userquota, groupquota "
+		    "properties\n"));
 		(void) printf(gettext("\nFor more information on a particular "
 		    "version, including supported releases, see:\n\n"));
 		(void) printf("http://www.opensolaris.org/os/community/zfs/"
@@ -1561,7 +1627,7 @@ zfs_do_upgrade(int argc, char **argv)
 		if (cb.cb_version == 0)
 			cb.cb_version = ZPL_VERSION;
 		ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_FILESYSTEM,
-		    NULL, NULL, upgrade_set_callback, &cb);
+		    NULL, NULL, 0, upgrade_set_callback, &cb);
 		(void) printf(gettext("%llu filesystems upgraded\n"),
 		    cb.cb_numupgraded);
 		if (cb.cb_numsamegraded) {
@@ -1579,14 +1645,14 @@ zfs_do_upgrade(int argc, char **argv)
 
 		flags |= ZFS_ITER_RECURSE;
 		ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
-		    NULL, NULL, upgrade_list_callback, &cb);
+		    NULL, NULL, 0, upgrade_list_callback, &cb);
 
 		found = cb.cb_foundone;
 		cb.cb_foundone = B_FALSE;
 		cb.cb_newer = B_TRUE;
 
 		ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
-		    NULL, NULL, upgrade_list_callback, &cb);
+		    NULL, NULL, 0, upgrade_list_callback, &cb);
 
 		if (!cb.cb_foundone && !found) {
 			(void) printf(gettext("All filesystems are "
@@ -1598,11 +1664,90 @@ zfs_do_upgrade(int argc, char **argv)
 }
 
 /*
- * list [-rH] [-o property[,property]...] [-t type[,type]...]
+ * zfs userspace
+ */
+static int
+userspace_cb(void *arg, const char *domain, uid_t rid, uint64_t space)
+{
+	zfs_userquota_prop_t *typep = arg;
+	zfs_userquota_prop_t p = *typep;
+	char *name = NULL;
+	char *ug, *propname;
+	char namebuf[32];
+	char sizebuf[32];
+
+	if (domain == NULL || domain[0] == '\0') {
+		if (p == ZFS_PROP_GROUPUSED || p == ZFS_PROP_GROUPQUOTA) {
+			struct group *g = getgrgid(rid);
+			if (g)
+				name = g->gr_name;
+		} else {
+			struct passwd *p = getpwuid(rid);
+			if (p)
+				name = p->pw_name;
+		}
+	}
+
+	if (p == ZFS_PROP_GROUPUSED || p == ZFS_PROP_GROUPQUOTA)
+		ug = "group";
+	else
+		ug = "user";
+
+	if (p == ZFS_PROP_USERUSED || p == ZFS_PROP_GROUPUSED)
+		propname = "used";
+	else
+		propname = "quota";
+
+	if (name == NULL) {
+		(void) snprintf(namebuf, sizeof (namebuf),
+		    "%llu", (longlong_t)rid);
+		name = namebuf;
+	}
+	zfs_nicenum(space, sizebuf, sizeof (sizebuf));
+
+	(void) printf("%s %s %s%c%s %s\n", propname, ug, domain,
+	    domain[0] ? '-' : ' ', name, sizebuf);
+
+	return (0);
+}
+
+static int
+zfs_do_userspace(int argc, char **argv)
+{
+	zfs_handle_t *zhp;
+	zfs_userquota_prop_t p;
+	int error;
+
+	/*
+	 * Try the python version.  If the execv fails, we'll continue
+	 * and do a simplistic implementation.
+	 */
+	(void) execv(pypath, argv-1);
+
+	(void) printf("internal error: %s not found\n"
+	    "falling back on built-in implementation, "
+	    "some features will not work\n", pypath);
+
+	if ((zhp = zfs_open(g_zfs, argv[argc-1], ZFS_TYPE_DATASET)) == NULL)
+		return (1);
+
+	(void) printf("PROP TYPE NAME VALUE\n");
+
+	for (p = 0; p < ZFS_NUM_USERQUOTA_PROPS; p++) {
+		error = zfs_userspace(zhp, p, userspace_cb, &p);
+		if (error)
+			break;
+	}
+	return (error);
+}
+
+/*
+ * list [-r][-d max] [-H] [-o property[,property]...] [-t type[,type]...]
  *      [-s property [-s property]...] [-S property [-S property]...]
  *      <dataset> ...
  *
  * 	-r	Recurse over all children
+ * 	-d	Limit recursion by depth.
  * 	-H	Scripted mode; elide headers and separate columns by tabs
  * 	-o	Control which fields to display.
  * 	-t	Control which object types to display.
@@ -1685,7 +1830,6 @@ print_dataset(zfs_handle_t *zhp, zprop_list_t *pl, boolean_t scripted)
 			first = B_FALSE;
 		}
 
-		right_justify = B_FALSE;
 		if (pl->pl_prop != ZPROP_INVAL) {
 			if (zfs_prop_get(zhp, pl->pl_prop, property,
 			    sizeof (property), NULL, NULL, 0, B_FALSE) != 0)
@@ -1694,6 +1838,13 @@ print_dataset(zfs_handle_t *zhp, zprop_list_t *pl, boolean_t scripted)
 				propstr = property;
 
 			right_justify = zfs_prop_align_right(pl->pl_prop);
+		} else if (zfs_prop_userquota(pl->pl_user_prop)) {
+			if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
+			    property, sizeof (property), B_FALSE) != 0)
+				propstr = "-";
+			else
+				propstr = property;
+			right_justify = B_TRUE;
 		} else {
 			if (nvlist_lookup_nvlist(userprops,
 			    pl->pl_user_prop, &propval) != 0)
@@ -1701,6 +1852,7 @@ print_dataset(zfs_handle_t *zhp, zprop_list_t *pl, boolean_t scripted)
 			else
 				verify(nvlist_lookup_string(propval,
 				    ZPROP_VALUE, &propstr) == 0);
+			right_justify = B_FALSE;
 		}
 
 		width = pl->pl_width;
@@ -1752,16 +1904,20 @@ zfs_do_list(int argc, char **argv)
 	char *fields = NULL;
 	list_cbdata_t cb = { 0 };
 	char *value;
+	int limit = 0;
 	int ret;
 	zfs_sort_column_t *sortcol = NULL;
 	int flags = ZFS_ITER_PROP_LISTSNAPS | ZFS_ITER_ARGS_CAN_BE_PATHS;
 
 	/* check options */
-	while ((c = getopt(argc, argv, ":o:rt:Hs:S:")) != -1) {
+	while ((c = getopt(argc, argv, ":d:o:rt:Hs:S:")) != -1) {
 		switch (c) {
 		case 'o':
 			fields = optarg;
 			break;
+		case 'd':
+			limit = parse_depth(optarg, &flags);
+			break;
 		case 'r':
 			flags |= ZFS_ITER_RECURSE;
 			break;
@@ -1852,7 +2008,7 @@ zfs_do_list(int argc, char **argv)
 	cb.cb_first = B_TRUE;
 
 	ret = zfs_for_each(argc, argv, flags, types, sortcol, &cb.cb_proplist,
-	    list_callback, &cb);
+	    limit, list_callback, &cb);
 
 	zprop_free_list(cb.cb_proplist);
 	zfs_free_sort_columns(sortcol);
@@ -2235,7 +2391,7 @@ zfs_do_set(int argc, char **argv)
 	}
 
 	ret = zfs_for_each(argc - 2, argv + 2, NULL,
-	    ZFS_TYPE_DATASET, NULL, NULL, set_callback, &cb);
+	    ZFS_TYPE_DATASET, NULL, NULL, 0, set_callback, &cb);
 
 	return (ret);
 }
@@ -2495,390 +2651,6 @@ zfs_do_receive(int argc, char **argv)
 	return (err != 0);
 }
 
-typedef struct allow_cb {
-	int  a_permcnt;
-	size_t a_treeoffset;
-} allow_cb_t;
-
-static void
-zfs_print_perms(avl_tree_t *tree)
-{
-	zfs_perm_node_t *permnode;
-
-	permnode = avl_first(tree);
-	while (permnode != NULL) {
-		(void) printf("%s", permnode->z_pname);
-		permnode = AVL_NEXT(tree, permnode);
-		if (permnode)
-			(void) printf(",");
-		else
-			(void) printf("\n");
-	}
-}
-
-/*
- * Iterate over user/groups/everyone/... and the call perm_iter
- * function to print actual permission when tree has >0 nodes.
- */
-static void
-zfs_iter_perms(avl_tree_t *tree, const char *banner, allow_cb_t *cb)
-{
-	zfs_allow_node_t *item;
-	avl_tree_t *ptree;
-
-	item = avl_first(tree);
-	while (item) {
-		ptree = (void *)((char *)item + cb->a_treeoffset);
-		if (avl_numnodes(ptree)) {
-			if (cb->a_permcnt++ == 0)
-				(void) printf("%s\n", banner);
-			(void) printf("\t%s", item->z_key);
-			/*
-			 * Avoid an extra space being printed
-			 * for "everyone" which is keyed with a null
-			 * string
-			 */
-			if (item->z_key[0] != '\0')
-				(void) printf(" ");
-			zfs_print_perms(ptree);
-		}
-		item = AVL_NEXT(tree, item);
-	}
-}
-
-#define	LINES "-------------------------------------------------------------\n"
-static int
-zfs_print_allows(char *ds)
-{
-	zfs_allow_t *curperms, *perms;
-	zfs_handle_t *zhp;
-	allow_cb_t allowcb = { 0 };
-	char banner[MAXPATHLEN];
-
-	if (ds[0] == '-')
-		usage(B_FALSE);
-
-	if (strrchr(ds, '@')) {
-		(void) fprintf(stderr, gettext("Snapshots don't have 'allow'"
-		    " permissions\n"));
-		return (1);
-	}
-	if ((zhp = zfs_open(g_zfs, ds, ZFS_TYPE_DATASET)) == NULL)
-		return (1);
-
-	if (zfs_perm_get(zhp, &perms)) {
-		(void) fprintf(stderr,
-		    gettext("Failed to retrieve 'allows' on %s\n"), ds);
-		zfs_close(zhp);
-		return (1);
-	}
-
-	zfs_close(zhp);
-
-	if (perms != NULL)
-		(void) printf("%s", LINES);
-	for (curperms = perms; curperms; curperms = curperms->z_next) {
-
-		(void) snprintf(banner, sizeof (banner),
-		    gettext("Permission sets on (%s)"), curperms->z_setpoint);
-		allowcb.a_treeoffset =
-		    offsetof(zfs_allow_node_t, z_localdescend);
-		allowcb.a_permcnt = 0;
-		zfs_iter_perms(&curperms->z_sets, banner, &allowcb);
-
-		(void) snprintf(banner, sizeof (banner),
-		    gettext("Create time permissions on (%s)"),
-		    curperms->z_setpoint);
-		allowcb.a_treeoffset =
-		    offsetof(zfs_allow_node_t, z_localdescend);
-		allowcb.a_permcnt = 0;
-		zfs_iter_perms(&curperms->z_crperms, banner, &allowcb);
-
-
-		(void) snprintf(banner, sizeof (banner),
-		    gettext("Local permissions on (%s)"), curperms->z_setpoint);
-		allowcb.a_treeoffset = offsetof(zfs_allow_node_t, z_local);
-		allowcb.a_permcnt = 0;
-		zfs_iter_perms(&curperms->z_user, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_group, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_everyone, banner, &allowcb);
-
-		(void) snprintf(banner, sizeof (banner),
-		    gettext("Descendent permissions on (%s)"),
-		    curperms->z_setpoint);
-		allowcb.a_treeoffset = offsetof(zfs_allow_node_t, z_descend);
-		allowcb.a_permcnt = 0;
-		zfs_iter_perms(&curperms->z_user, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_group, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_everyone, banner, &allowcb);
-
-		(void) snprintf(banner, sizeof (banner),
-		    gettext("Local+Descendent permissions on (%s)"),
-		    curperms->z_setpoint);
-		allowcb.a_treeoffset =
-		    offsetof(zfs_allow_node_t, z_localdescend);
-		allowcb.a_permcnt = 0;
-		zfs_iter_perms(&curperms->z_user, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_group, banner, &allowcb);
-		zfs_iter_perms(&curperms->z_everyone, banner, &allowcb);
-
-		(void) printf("%s", LINES);
-	}
-	zfs_free_allows(perms);
-	return (0);
-}
-
-#define	ALLOWOPTIONS "ldcsu:g:e"
-#define	UNALLOWOPTIONS "ldcsu:g:er"
-
-/*
- * Validate options, and build necessary datastructure to display/remove/add
- * permissions.
- * Returns 0 - If permissions should be added/removed
- * Returns 1 - If permissions should be displayed.
- * Returns -1 - on failure
- */
-int
-parse_allow_args(int *argc, char **argv[], boolean_t unallow,
-    char **ds, int *recurse, nvlist_t **zperms)
-{
-	int c;
-	char *options = unallow ? UNALLOWOPTIONS : ALLOWOPTIONS;
-	zfs_deleg_inherit_t deleg_type = ZFS_DELEG_NONE;
-	zfs_deleg_who_type_t who_type = ZFS_DELEG_WHO_UNKNOWN;
-	char *who = NULL;
-	char *perms = NULL;
-	zfs_handle_t *zhp;
-
-	while ((c = getopt(*argc, *argv, options)) != -1) {
-		switch (c) {
-		case 'l':
-			if (who_type == ZFS_DELEG_CREATE ||
-			    who_type == ZFS_DELEG_NAMED_SET)
-				usage(B_FALSE);
-
-			deleg_type |= ZFS_DELEG_PERM_LOCAL;
-			break;
-		case 'd':
-			if (who_type == ZFS_DELEG_CREATE ||
-			    who_type == ZFS_DELEG_NAMED_SET)
-				usage(B_FALSE);
-
-			deleg_type |= ZFS_DELEG_PERM_DESCENDENT;
-			break;
-		case 'r':
-			*recurse = B_TRUE;
-			break;
-		case 'c':
-			if (who_type != ZFS_DELEG_WHO_UNKNOWN)
-				usage(B_FALSE);
-			if (deleg_type)
-				usage(B_FALSE);
-			who_type = ZFS_DELEG_CREATE;
-			break;
-		case 's':
-			if (who_type != ZFS_DELEG_WHO_UNKNOWN)
-				usage(B_FALSE);
-			if (deleg_type)
-				usage(B_FALSE);
-			who_type = ZFS_DELEG_NAMED_SET;
-			break;
-		case 'u':
-			if (who_type != ZFS_DELEG_WHO_UNKNOWN)
-				usage(B_FALSE);
-			who_type = ZFS_DELEG_USER;
-			who = optarg;
-			break;
-		case 'g':
-			if (who_type != ZFS_DELEG_WHO_UNKNOWN)
-				usage(B_FALSE);
-			who_type = ZFS_DELEG_GROUP;
-			who = optarg;
-			break;
-		case 'e':
-			if (who_type != ZFS_DELEG_WHO_UNKNOWN)
-				usage(B_FALSE);
-			who_type = ZFS_DELEG_EVERYONE;
-			break;
-		default:
-			usage(B_FALSE);
-			break;
-		}
-	}
-
-	if (deleg_type == 0)
-		deleg_type = ZFS_DELEG_PERM_LOCALDESCENDENT;
-
-	*argc -= optind;
-	*argv += optind;
-
-	if (unallow == B_FALSE && *argc == 1) {
-		/*
-		 * Only print permissions if no options were processed
-		 */
-		if (optind == 1)
-			return (1);
-		else
-			usage(B_FALSE);
-	}
-
-	/*
-	 * initialize variables for zfs_build_perms based on number
-	 * of arguments.
-	 * 3 arguments ==>	zfs [un]allow joe perm,perm,perm <dataset> or
-	 *			zfs [un]allow -s @set1 perm,perm <dataset>
-	 * 2 arguments ==>	zfs [un]allow -c perm,perm <dataset> or
-	 *			zfs [un]allow -u|-g <name> perm <dataset> or
-	 *			zfs [un]allow -e perm,perm <dataset>
-	 *			zfs unallow joe <dataset>
-	 *			zfs unallow -s @set1 <dataset>
-	 * 1 argument  ==>	zfs [un]allow -e <dataset> or
-	 *			zfs [un]allow -c <dataset>
-	 */
-
-	switch (*argc) {
-	case 3:
-		perms = (*argv)[1];
-		who = (*argv)[0];
-		*ds = (*argv)[2];
-
-		/*
-		 * advance argc/argv for do_allow cases.
-		 * for do_allow case make sure who have a know who type
-		 * and its not a permission set.
-		 */
-		if (unallow == B_TRUE) {
-			*argc -= 2;
-			*argv += 2;
-		} else if (who_type != ZFS_DELEG_WHO_UNKNOWN &&
-		    who_type != ZFS_DELEG_NAMED_SET)
-			usage(B_FALSE);
-		break;
-
-	case 2:
-		if (unallow == B_TRUE && (who_type == ZFS_DELEG_EVERYONE ||
-		    who_type == ZFS_DELEG_CREATE || who != NULL)) {
-			perms = (*argv)[0];
-			*ds = (*argv)[1];
-		} else {
-			if (unallow == B_FALSE &&
-			    (who_type == ZFS_DELEG_WHO_UNKNOWN ||
-			    who_type == ZFS_DELEG_NAMED_SET))
-				usage(B_FALSE);
-			else if (who_type == ZFS_DELEG_WHO_UNKNOWN ||
-			    who_type == ZFS_DELEG_NAMED_SET)
-				who = (*argv)[0];
-			else if (who_type != ZFS_DELEG_NAMED_SET)
-				perms = (*argv)[0];
-			*ds = (*argv)[1];
-		}
-		if (unallow == B_TRUE) {
-			(*argc)--;
-			(*argv)++;
-		}
-		break;
-
-	case 1:
-		if (unallow == B_FALSE)
-			usage(B_FALSE);
-		if (who == NULL && who_type != ZFS_DELEG_CREATE &&
-		    who_type != ZFS_DELEG_EVERYONE)
-			usage(B_FALSE);
-		*ds = (*argv)[0];
-		break;
-
-	default:
-		usage(B_FALSE);
-	}
-
-	if (strrchr(*ds, '@')) {
-		(void) fprintf(stderr,
-		    gettext("Can't set or remove 'allow' permissions "
-		    "on snapshots.\n"));
-			return (-1);
-	}
-
-	if ((zhp = zfs_open(g_zfs, *ds, ZFS_TYPE_DATASET)) == NULL)
-		return (-1);
-
-	if ((zfs_build_perms(zhp, who, perms,
-	    who_type, deleg_type, zperms)) != 0) {
-		zfs_close(zhp);
-		return (-1);
-	}
-	zfs_close(zhp);
-	return (0);
-}
-
-static int
-zfs_do_allow(int argc, char **argv)
-{
-	char *ds;
-	nvlist_t *zperms = NULL;
-	zfs_handle_t *zhp;
-	int unused;
-	int ret;
-
-	if ((ret = parse_allow_args(&argc, &argv, B_FALSE, &ds,
-	    &unused, &zperms)) == -1)
-		return (1);
-
-	if (ret == 1)
-		return (zfs_print_allows(argv[0]));
-
-	if ((zhp = zfs_open(g_zfs, ds, ZFS_TYPE_DATASET)) == NULL)
-		return (1);
-
-	if (zfs_perm_set(zhp, zperms)) {
-		zfs_close(zhp);
-		nvlist_free(zperms);
-		return (1);
-	}
-	nvlist_free(zperms);
-	zfs_close(zhp);
-
-	return (0);
-}
-
-static int
-unallow_callback(zfs_handle_t *zhp, void *data)
-{
-	nvlist_t *nvp = (nvlist_t *)data;
-	int error;
-
-	error = zfs_perm_remove(zhp, nvp);
-	if (error) {
-		(void) fprintf(stderr, gettext("Failed to remove permissions "
-		    "on %s\n"), zfs_get_name(zhp));
-	}
-	return (error);
-}
-
-static int
-zfs_do_unallow(int argc, char **argv)
-{
-	int recurse = B_FALSE;
-	char *ds;
-	int error;
-	nvlist_t *zperms = NULL;
-	int flags = 0;
-
-	if (parse_allow_args(&argc, &argv, B_TRUE,
-	    &ds, &recurse, &zperms) == -1)
-		return (1);
-
-	if (recurse)
-		flags |= ZFS_ITER_RECURSE;
-	error = zfs_for_each(argc, argv, flags,
-	    ZFS_TYPE_FILESYSTEM|ZFS_TYPE_VOLUME, NULL,
-	    NULL, unallow_callback, (void *)zperms);
-
-	if (zperms)
-		nvlist_free(zperms);
-
-	return (error);
-}
-
 typedef struct get_all_cbdata {
 	zfs_handle_t	**cb_handles;
 	size_t		cb_alloc;
@@ -3944,6 +3716,15 @@ zfs_do_unshare(int argc, char **argv)
 	return (unshare_unmount(OP_SHARE, argc, argv));
 }
 
+/* ARGSUSED */
+static int
+zfs_do_python(int argc, char **argv)
+{
+	(void) execv(pypath, argv-1);
+	(void) printf("internal error: %s not found\n", pypath);
+	return (-1);
+}
+
 /*
  * Called when invoked as /etc/fs/zfs/mount.  Do the mount if the mountpoint is
  * 'legacy'.  Otherwise, complain that use should be using 'zfs mount'.
@@ -4197,6 +3978,7 @@ main(int argc, char **argv)
 		/*
 		 * Run the appropriate command.
 		 */
+		libzfs_mnttab_cache(g_zfs, B_TRUE);
 		if (find_command_idx(cmdname, &i) == 0) {
 			current_command = &command_table[i];
 			ret = command_table[i].func(argc - 1, argv + 1);
@@ -4209,6 +3991,7 @@ main(int argc, char **argv)
 			    "command '%s'\n"), cmdname);
 			usage(B_FALSE);
 		}
+		libzfs_mnttab_cache(g_zfs, B_FALSE);
 	}
 
 	(void) fclose(mnttab_file);

cmd/zinject/zinject.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * ZFS Fault Injector
  *
@@ -224,7 +222,7 @@ usage(void)
 	    "\t\tClear the particular record (if given a numeric ID), or\n"
 	    "\t\tall records if 'all' is specificed.\n"
 	    "\n"
-	    "\tzinject -d device [-e errno] [-L <nvlist|uber>] pool\n"
+	    "\tzinject -d device [-e errno] [-L <nvlist|uber>] [-F] pool\n"
 	    "\t\tInject a fault into a particular device or the device's\n"
 	    "\t\tlabel.  Label injection can either be 'nvlist' or 'uber'.\n"
 	    "\t\t'errno' can either be 'nxio' (the default) or 'io'.\n"
@@ -516,7 +514,7 @@ main(int argc, char **argv)
 		return (0);
 	}
 
-	while ((c = getopt(argc, argv, ":ab:d:f:qhc:t:l:mr:e:uL:")) != -1) {
+	while ((c = getopt(argc, argv, ":ab:d:f:Fqhc:t:l:mr:e:uL:")) != -1) {
 		switch (c) {
 		case 'a':
 			flags |= ZINJECT_FLUSH_ARC;
@@ -553,6 +551,9 @@ main(int argc, char **argv)
 				return (1);
 			}
 			break;
+		case 'F':
+			record.zi_failfast = B_TRUE;
+			break;
 		case 'h':
 			usage();
 			return (0);

cmd/zpool/zpool_main.c

@@ -376,12 +376,11 @@ add_prop_list(const char *propname, char *propval, nvlist_t **props,
 		}
 		normnm = zpool_prop_to_name(prop);
 	} else {
-		if ((fprop = zfs_name_to_prop(propname)) == ZPROP_INVAL) {
+		if ((fprop = zfs_name_to_prop(propname)) != ZPROP_INVAL) {
-			(void) fprintf(stderr, gettext("property '%s' is "
+			normnm = zfs_prop_to_name(fprop);
-			    "not a valid file system property\n"), propname);
+		} else {
-			return (2);
+			normnm = propname;
 		}
-		normnm = zfs_prop_to_name(fprop);
 	}
 
 	if (nvlist_lookup_string(proplist, normnm, &strval) == 0 &&
@@ -979,14 +978,189 @@ max_width(zpool_handle_t *zhp, nvlist_t *nv, int depth, int max)
 	return (max);
 }
 
+typedef struct spare_cbdata {
+	uint64_t	cb_guid;
+	zpool_handle_t	*cb_zhp;
+} spare_cbdata_t;
+
+static boolean_t
+find_vdev(nvlist_t *nv, uint64_t search)
+{
+	uint64_t guid;
+	nvlist_t **child;
+	uint_t c, children;
+
+	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
+	    search == guid)
+		return (B_TRUE);
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) == 0) {
+		for (c = 0; c < children; c++)
+			if (find_vdev(child[c], search))
+				return (B_TRUE);
+	}
+
+	return (B_FALSE);
+}
+
+static int
+find_spare(zpool_handle_t *zhp, void *data)
+{
+	spare_cbdata_t *cbp = data;
+	nvlist_t *config, *nvroot;
+
+	config = zpool_get_config(zhp, NULL);
+	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
+	    &nvroot) == 0);
+
+	if (find_vdev(nvroot, cbp->cb_guid)) {
+		cbp->cb_zhp = zhp;
+		return (1);
+	}
+
+	zpool_close(zhp);
+	return (0);
+}
+
+/*
+ * Print out configuration state as requested by status_callback.
+ */
+void
+print_status_config(zpool_handle_t *zhp, const char *name, nvlist_t *nv,
+    int namewidth, int depth, boolean_t isspare)
+{
+	nvlist_t **child;
+	uint_t c, children;
+	vdev_stat_t *vs;
+	char rbuf[6], wbuf[6], cbuf[6], repaired[7];
+	char *vname;
+	uint64_t notpresent;
+	spare_cbdata_t cb;
+	char *state;
+
+	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
+	    (uint64_t **)&vs, &c) == 0);
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) != 0)
+		children = 0;
+
+	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
+	if (isspare) {
+		/*
+		 * For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
+		 * online drives.
+		 */
+		if (vs->vs_aux == VDEV_AUX_SPARED)
+			state = "INUSE";
+		else if (vs->vs_state == VDEV_STATE_HEALTHY)
+			state = "AVAIL";
+	}
+
+	(void) printf("\t%*s%-*s  %-8s", depth, "", namewidth - depth,
+	    name, state);
+
+	if (!isspare) {
+		zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
+		zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
+		zfs_nicenum(vs->vs_checksum_errors, cbuf, sizeof (cbuf));
+		(void) printf(" %5s %5s %5s", rbuf, wbuf, cbuf);
+	}
+
+	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
+	    &notpresent) == 0) {
+		char *path;
+		verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
+		(void) printf("  was %s", path);
+	} else if (vs->vs_aux != 0) {
+		(void) printf("  ");
+
+		switch (vs->vs_aux) {
+		case VDEV_AUX_OPEN_FAILED:
+			(void) printf(gettext("cannot open"));
+			break;
+
+		case VDEV_AUX_BAD_GUID_SUM:
+			(void) printf(gettext("missing device"));
+			break;
+
+		case VDEV_AUX_NO_REPLICAS:
+			(void) printf(gettext("insufficient replicas"));
+			break;
+
+		case VDEV_AUX_VERSION_NEWER:
+			(void) printf(gettext("newer version"));
+			break;
+
+		case VDEV_AUX_SPARED:
+			verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
+			    &cb.cb_guid) == 0);
+			if (zpool_iter(g_zfs, find_spare, &cb) == 1) {
+				if (strcmp(zpool_get_name(cb.cb_zhp),
+				    zpool_get_name(zhp)) == 0)
+					(void) printf(gettext("currently in "
+					    "use"));
+				else
+					(void) printf(gettext("in use by "
+					    "pool '%s'"),
+					    zpool_get_name(cb.cb_zhp));
+				zpool_close(cb.cb_zhp);
+			} else {
+				(void) printf(gettext("currently in use"));
+			}
+			break;
+
+		case VDEV_AUX_ERR_EXCEEDED:
+			(void) printf(gettext("too many errors"));
+			break;
+
+		case VDEV_AUX_IO_FAILURE:
+			(void) printf(gettext("experienced I/O failures"));
+			break;
+
+		case VDEV_AUX_BAD_LOG:
+			(void) printf(gettext("bad intent log"));
+			break;
+
+		default:
+			(void) printf(gettext("corrupted data"));
+			break;
+		}
+	} else if (vs->vs_scrub_repaired != 0 && children == 0) {
+		/*
+		 * Report bytes resilvered/repaired on leaf devices.
+		 */
+		zfs_nicenum(vs->vs_scrub_repaired, repaired, sizeof (repaired));
+		(void) printf(gettext("  %s %s"), repaired,
+		    (vs->vs_scrub_type == POOL_SCRUB_RESILVER) ?
+		    "resilvered" : "repaired");
+	}
+
+	(void) printf("\n");
+
+	for (c = 0; c < children; c++) {
+		uint64_t is_log = B_FALSE;
+
+		/* Don't print logs here */
+		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
+		    &is_log);
+		if (is_log)
+			continue;
+		vname = zpool_vdev_name(g_zfs, zhp, child[c]);
+		print_status_config(zhp, vname, child[c],
+		    namewidth, depth + 2, isspare);
+		free(vname);
+	}
+}
+
 
 /*
  * Print the configuration of an exported pool.  Iterate over all vdevs in the
  * pool, printing out the name and status for each one.
  */
 void
-print_import_config(const char *name, nvlist_t *nv, int namewidth, int depth,
+print_import_config(const char *name, nvlist_t *nv, int namewidth, int depth)
-    boolean_t print_logs)
 {
 	nvlist_t **child;
 	uint_t c, children;
@@ -1043,12 +1217,11 @@ print_import_config(const char *name, nvlist_t *nv, int namewidth, int depth,
 
 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
 		    &is_log);
-		if ((is_log && !print_logs) || (!is_log && print_logs))
+		if (is_log)
 			continue;
 
 		vname = zpool_vdev_name(g_zfs, NULL, child[c]);
-		print_import_config(vname, child[c],
+		print_import_config(vname, child[c], namewidth, depth + 2);
-		    namewidth, depth + 2, B_FALSE);
 		free(vname);
 	}
 
@@ -1073,6 +1246,43 @@ print_import_config(const char *name, nvlist_t *nv, int namewidth, int depth,
 	}
 }
 
+/*
+ * Print log vdevs.
+ * Logs are recorded as top level vdevs in the main pool child array
+ * but with "is_log" set to 1. We use either print_status_config() or
+ * print_import_config() to print the top level logs then any log
+ * children (eg mirrored slogs) are printed recursively - which
+ * works because only the top level vdev is marked "is_log"
+ */
+static void
+print_logs(zpool_handle_t *zhp, nvlist_t *nv, int namewidth, boolean_t verbose)
+{
+	uint_t c, children;
+	nvlist_t **child;
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child,
+	    &children) != 0)
+		return;
+
+	(void) printf(gettext("\tlogs\n"));
+
+	for (c = 0; c < children; c++) {
+		uint64_t is_log = B_FALSE;
+		char *name;
+
+		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
+		    &is_log);
+		if (!is_log)
+			continue;
+		name = zpool_vdev_name(g_zfs, zhp, child[c]);
+		if (verbose)
+			print_status_config(zhp, name, child[c], namewidth,
+			    2, B_FALSE);
+		else
+			print_import_config(name, child[c], namewidth, 2);
+		free(name);
+	}
+}
 /*
  * Display the status for the given pool.
  */
@@ -1241,11 +1451,9 @@ show_import(nvlist_t *config)
 	if (namewidth < 10)
 		namewidth = 10;
 
-	print_import_config(name, nvroot, namewidth, 0, B_FALSE);
+	print_import_config(name, nvroot, namewidth, 0);
-	if (num_logs(nvroot) > 0) {
+	if (num_logs(nvroot) > 0)
-		(void) printf(gettext("\tlogs\n"));
+		print_logs(NULL, nvroot, namewidth, B_FALSE);
-		print_import_config(name, nvroot, namewidth, 0, B_TRUE);
-	}
 
 	if (reason == ZPOOL_STATUS_BAD_GUID_SUM) {
 		(void) printf(gettext("\n\tAdditional devices are known to "
@@ -2427,10 +2635,14 @@ zpool_do_online(int argc, char **argv)
 	zpool_handle_t *zhp;
 	int ret = 0;
 	vdev_state_t newstate;
+	int flags = 0;
 
 	/* check options */
-	while ((c = getopt(argc, argv, "t")) != -1) {
+	while ((c = getopt(argc, argv, "et")) != -1) {
 		switch (c) {
+		case 'e':
+			flags |= ZFS_ONLINE_EXPAND;
+			break;
 		case 't':
 		case '?':
 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
@@ -2458,7 +2670,7 @@ zpool_do_online(int argc, char **argv)
 		return (1);
 
 	for (i = 1; i < argc; i++) {
-		if (zpool_vdev_online(zhp, argv[i], 0, &newstate) == 0) {
+		if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
 			if (newstate != VDEV_STATE_HEALTHY) {
 				(void) printf(gettext("warning: device '%s' "
 				    "onlined, but remains in faulted state\n"),
@@ -2715,181 +2927,6 @@ print_scrub_status(nvlist_t *nvroot)
 	    (u_longlong_t)(minutes_left / 60), (uint_t)(minutes_left % 60));
 }
 
-typedef struct spare_cbdata {
-	uint64_t	cb_guid;
-	zpool_handle_t	*cb_zhp;
-} spare_cbdata_t;
-
-static boolean_t
-find_vdev(nvlist_t *nv, uint64_t search)
-{
-	uint64_t guid;
-	nvlist_t **child;
-	uint_t c, children;
-
-	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
-	    search == guid)
-		return (B_TRUE);
-
-	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
-	    &child, &children) == 0) {
-		for (c = 0; c < children; c++)
-			if (find_vdev(child[c], search))
-				return (B_TRUE);
-	}
-
-	return (B_FALSE);
-}
-
-static int
-find_spare(zpool_handle_t *zhp, void *data)
-{
-	spare_cbdata_t *cbp = data;
-	nvlist_t *config, *nvroot;
-
-	config = zpool_get_config(zhp, NULL);
-	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
-	    &nvroot) == 0);
-
-	if (find_vdev(nvroot, cbp->cb_guid)) {
-		cbp->cb_zhp = zhp;
-		return (1);
-	}
-
-	zpool_close(zhp);
-	return (0);
-}
-
-/*
- * Print out configuration state as requested by status_callback.
- */
-void
-print_status_config(zpool_handle_t *zhp, const char *name, nvlist_t *nv,
-    int namewidth, int depth, boolean_t isspare, boolean_t print_logs)
-{
-	nvlist_t **child;
-	uint_t c, children;
-	vdev_stat_t *vs;
-	char rbuf[6], wbuf[6], cbuf[6], repaired[7];
-	char *vname;
-	uint64_t notpresent;
-	spare_cbdata_t cb;
-	char *state;
-
-	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
-	    (uint64_t **)&vs, &c) == 0);
-
-	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
-	    &child, &children) != 0)
-		children = 0;
-
-	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
-	if (isspare) {
-		/*
-		 * For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
-		 * online drives.
-		 */
-		if (vs->vs_aux == VDEV_AUX_SPARED)
-			state = "INUSE";
-		else if (vs->vs_state == VDEV_STATE_HEALTHY)
-			state = "AVAIL";
-	}
-
-	(void) printf("\t%*s%-*s  %-8s", depth, "", namewidth - depth,
-	    name, state);
-
-	if (!isspare) {
-		zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
-		zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
-		zfs_nicenum(vs->vs_checksum_errors, cbuf, sizeof (cbuf));
-		(void) printf(" %5s %5s %5s", rbuf, wbuf, cbuf);
-	}
-
-	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
-	    &notpresent) == 0) {
-		char *path;
-		verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
-		(void) printf("  was %s", path);
-	} else if (vs->vs_aux != 0) {
-		(void) printf("  ");
-
-		switch (vs->vs_aux) {
-		case VDEV_AUX_OPEN_FAILED:
-			(void) printf(gettext("cannot open"));
-			break;
-
-		case VDEV_AUX_BAD_GUID_SUM:
-			(void) printf(gettext("missing device"));
-			break;
-
-		case VDEV_AUX_NO_REPLICAS:
-			(void) printf(gettext("insufficient replicas"));
-			break;
-
-		case VDEV_AUX_VERSION_NEWER:
-			(void) printf(gettext("newer version"));
-			break;
-
-		case VDEV_AUX_SPARED:
-			verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
-			    &cb.cb_guid) == 0);
-			if (zpool_iter(g_zfs, find_spare, &cb) == 1) {
-				if (strcmp(zpool_get_name(cb.cb_zhp),
-				    zpool_get_name(zhp)) == 0)
-					(void) printf(gettext("currently in "
-					    "use"));
-				else
-					(void) printf(gettext("in use by "
-					    "pool '%s'"),
-					    zpool_get_name(cb.cb_zhp));
-				zpool_close(cb.cb_zhp);
-			} else {
-				(void) printf(gettext("currently in use"));
-			}
-			break;
-
-		case VDEV_AUX_ERR_EXCEEDED:
-			(void) printf(gettext("too many errors"));
-			break;
-
-		case VDEV_AUX_IO_FAILURE:
-			(void) printf(gettext("experienced I/O failures"));
-			break;
-
-		case VDEV_AUX_BAD_LOG:
-			(void) printf(gettext("bad intent log"));
-			break;
-
-		default:
-			(void) printf(gettext("corrupted data"));
-			break;
-		}
-	} else if (vs->vs_scrub_repaired != 0 && children == 0) {
-		/*
-		 * Report bytes resilvered/repaired on leaf devices.
-		 */
-		zfs_nicenum(vs->vs_scrub_repaired, repaired, sizeof (repaired));
-		(void) printf(gettext("  %s %s"), repaired,
-		    (vs->vs_scrub_type == POOL_SCRUB_RESILVER) ?
-		    "resilvered" : "repaired");
-	}
-
-	(void) printf("\n");
-
-	for (c = 0; c < children; c++) {
-		uint64_t is_log = B_FALSE;
-
-		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
-		    &is_log);
-		if ((is_log && !print_logs) || (!is_log && print_logs))
-			continue;
-		vname = zpool_vdev_name(g_zfs, zhp, child[c]);
-		print_status_config(zhp, vname, child[c],
-		    namewidth, depth + 2, isspare, B_FALSE);
-		free(vname);
-	}
-}
-
 static void
 print_error_log(zpool_handle_t *zhp)
 {
@@ -2940,7 +2977,7 @@ print_spares(zpool_handle_t *zhp, nvlist_t **spares, uint_t nspares,
 	for (i = 0; i < nspares; i++) {
 		name = zpool_vdev_name(g_zfs, zhp, spares[i]);
 		print_status_config(zhp, name, spares[i],
-		    namewidth, 2, B_TRUE, B_FALSE);
+		    namewidth, 2, B_TRUE);
 		free(name);
 	}
 }
@@ -2960,7 +2997,7 @@ print_l2cache(zpool_handle_t *zhp, nvlist_t **l2cache, uint_t nl2cache,
 	for (i = 0; i < nl2cache; i++) {
 		name = zpool_vdev_name(g_zfs, zhp, l2cache[i]);
 		print_status_config(zhp, name, l2cache[i],
-		    namewidth, 2, B_FALSE, B_FALSE);
+		    namewidth, 2, B_FALSE);
 		free(name);
 	}
 }
@@ -3190,11 +3227,10 @@ status_callback(zpool_handle_t *zhp, void *data)
 		(void) printf(gettext("\t%-*s  %-8s %5s %5s %5s\n"), namewidth,
 		    "NAME", "STATE", "READ", "WRITE", "CKSUM");
 		print_status_config(zhp, zpool_get_name(zhp), nvroot,
-		    namewidth, 0, B_FALSE, B_FALSE);
+		    namewidth, 0, B_FALSE);
-		if (num_logs(nvroot) > 0)
-			print_status_config(zhp, "logs", nvroot, namewidth, 0,
-			    B_FALSE, B_TRUE);
 
+		if (num_logs(nvroot) > 0)
+			print_logs(zhp, nvroot, namewidth, B_TRUE);
 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
 		    &l2cache, &nl2cache) == 0)
 			print_l2cache(zhp, l2cache, nl2cache, namewidth);
@@ -3418,7 +3454,7 @@ zpool_do_upgrade(int argc, char **argv)
 
 
 	/* check options */
-	while ((c = getopt(argc, argv, "avV:")) != -1) {
+	while ((c = getopt(argc, argv, ":avV:")) != -1) {
 		switch (c) {
 		case 'a':
 			cb.cb_all = B_TRUE;
@@ -3435,6 +3471,11 @@ zpool_do_upgrade(int argc, char **argv)
 				usage(B_FALSE);
 			}
 			break;
+		case ':':
+			(void) fprintf(stderr, gettext("missing argument for "
+			    "'%c' option\n"), optopt);
+			usage(B_FALSE);
+			break;
 		case '?':
 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
 			    optopt);
@@ -3495,8 +3536,9 @@ zpool_do_upgrade(int argc, char **argv)
 		(void) printf(gettext(" 11  Improved scrub performance\n"));
 		(void) printf(gettext(" 12  Snapshot properties\n"));
 		(void) printf(gettext(" 13  snapused property\n"));
-		(void) printf(gettext(" 14  passthrough-x aclinherit "
+		(void) printf(gettext(" 14  passthrough-x aclinherit\n"));
-		    "support\n"));
+		(void) printf(gettext(" 15  user/group space accounting\n"));
+		(void) printf(gettext(" 16  stmf property support\n"));
 		(void) printf(gettext("For more information on a particular "
 		    "version, including supported releases, see:\n\n"));
 		(void) printf("http://www.opensolaris.org/os/community/zfs/"

cmd/ztest/ztest.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -76,6 +76,7 @@
 #include <sys/spa.h>
 #include <sys/dmu.h>
 #include <sys/txg.h>
+#include <sys/dbuf.h>
 #include <sys/zap.h>
 #include <sys/dmu_objset.h>
 #include <sys/poll.h>
@@ -92,6 +93,7 @@
 #include <sys/vdev_file.h>
 #include <sys/spa_impl.h>
 #include <sys/dsl_prop.h>
+#include <sys/dsl_dataset.h>
 #include <sys/refcount.h>
 #include <stdio.h>
 #include <stdio_ext.h>
@@ -162,6 +164,7 @@ typedef void ztest_func_t(ztest_args_t *);
  * Note: these aren't static because we want dladdr() to work.
  */
 ztest_func_t ztest_dmu_read_write;
+ztest_func_t ztest_dmu_read_write_zcopy;
 ztest_func_t ztest_dmu_write_parallel;
 ztest_func_t ztest_dmu_object_alloc_free;
 ztest_func_t ztest_dmu_commit_callbacks;
@@ -171,6 +174,7 @@ ztest_func_t ztest_traverse;
 ztest_func_t ztest_dsl_prop_get_set;
 ztest_func_t ztest_dmu_objset_create_destroy;
 ztest_func_t ztest_dmu_snapshot_create_destroy;
+ztest_func_t ztest_dsl_dataset_promote_busy;
 ztest_func_t ztest_spa_create_destroy;
 ztest_func_t ztest_fault_inject;
 ztest_func_t ztest_spa_rename;
@@ -197,6 +201,7 @@ uint64_t zopt_rarely = 60;		/* every 60 seconds */
 
 ztest_info_t ztest_info[] = {
 	{ ztest_dmu_read_write,			1,	&zopt_always	},
+	{ ztest_dmu_read_write_zcopy,		1,	&zopt_always	},
 	{ ztest_dmu_write_parallel,		30,	&zopt_always	},
 	{ ztest_dmu_object_alloc_free,		1,	&zopt_always	},
 	{ ztest_dmu_commit_callbacks,		10,	&zopt_always	},
@@ -210,6 +215,7 @@ ztest_info_t ztest_info[] = {
 	{ ztest_spa_rename,			1,	&zopt_rarely	},
 	{ ztest_vdev_attach_detach,		1,	&zopt_rarely	},
 	{ ztest_vdev_LUN_growth,		1,	&zopt_rarely	},
+	{ ztest_dsl_dataset_promote_busy,	1,	&zopt_rarely	},
 	{ ztest_vdev_add_remove,		1,	&zopt_vdevtime	},
 	{ ztest_vdev_aux_add_remove,		1,	&zopt_vdevtime	},
 	{ ztest_scrub,				1,	&zopt_vdevtime	},
@@ -255,9 +261,11 @@ static ztest_shared_t *ztest_shared;
 static int ztest_random_fd;
 static int ztest_dump_core = 1;
 
+static uint64_t metaslab_sz;
 static boolean_t ztest_exiting;
 
 extern uint64_t metaslab_gang_bang;
+extern uint64_t metaslab_df_alloc_threshold;
 
 #define	ZTEST_DIROBJ		1
 #define	ZTEST_MICROZAP_OBJ	2
@@ -959,7 +967,7 @@ ztest_vdev_aux_add_remove(ztest_args_t *za)
 		 * of devices that have pending state changes.
 		 */
 		if (ztest_random(2) == 0)
-			(void) vdev_online(spa, guid, B_FALSE, NULL);
+			(void) vdev_online(spa, guid, 0, NULL);
 
 		error = spa_vdev_remove(spa, guid, B_FALSE);
 		if (error != 0 && error != EBUSY)
@@ -1037,7 +1045,7 @@ ztest_vdev_attach_detach(ztest_args_t *za)
 	}
 
 	oldguid = oldvd->vdev_guid;
-	oldsize = vdev_get_rsize(oldvd);
+	oldsize = vdev_get_min_asize(oldvd);
 	oldvd_is_log = oldvd->vdev_top->vdev_islog;
 	(void) strcpy(oldpath, oldvd->vdev_path);
 	pvd = oldvd->vdev_parent;
@@ -1073,7 +1081,7 @@ ztest_vdev_attach_detach(ztest_args_t *za)
 	}
 
 	if (newvd) {
-		newsize = vdev_get_rsize(newvd);
+		newsize = vdev_get_min_asize(newvd);
 	} else {
 		/*
 		 * Make newsize a little bigger or smaller than oldsize.
@@ -1148,6 +1156,95 @@ ztest_vdev_attach_detach(ztest_args_t *za)
 	(void) pthread_mutex_unlock(&ztest_shared->zs_vdev_lock);
 }
 
+/*
+ * Callback function which expands the physical size of the vdev.
+ */
+vdev_t *
+grow_vdev(vdev_t *vd, void *arg)
+{
+	spa_t *spa = vd->vdev_spa;
+	size_t *newsize = arg;
+	size_t fsize;
+	int fd;
+
+	ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE);
+	ASSERT(vd->vdev_ops->vdev_op_leaf);
+
+	if ((fd = open(vd->vdev_path, O_RDWR)) == -1)
+		return (vd);
+
+	fsize = lseek(fd, 0, SEEK_END);
+	(void) ftruncate(fd, *newsize);
+
+	if (zopt_verbose >= 6) {
+		(void) printf("%s grew from %lu to %lu bytes\n",
+		    vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize);
+	}
+	(void) close(fd);
+	return (NULL);
+}
+
+/*
+ * Callback function which expands a given vdev by calling vdev_online().
+ */
+/* ARGSUSED */
+vdev_t *
+online_vdev(vdev_t *vd, void *arg)
+{
+	spa_t *spa = vd->vdev_spa;
+	vdev_t *tvd = vd->vdev_top;
+	vdev_t *pvd = vd->vdev_parent;
+	uint64_t guid = vd->vdev_guid;
+
+	ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE);
+	ASSERT(vd->vdev_ops->vdev_op_leaf);
+
+	/* Calling vdev_online will initialize the new metaslabs */
+	spa_config_exit(spa, SCL_STATE, spa);
+	(void) vdev_online(spa, guid, ZFS_ONLINE_EXPAND, NULL);
+	spa_config_enter(spa, SCL_STATE, spa, RW_READER);
+
+	/*
+	 * Since we dropped the lock we need to ensure that we're
+	 * still talking to the original vdev. It's possible this
+	 * vdev may have been detached/replaced while we were
+	 * trying to online it.
+	 */
+	if (vd != vdev_lookup_by_guid(tvd, guid) || vd->vdev_parent != pvd) {
+		if (zopt_verbose >= 6) {
+			(void) printf("vdev %p has disappeared, was "
+			    "guid %llu\n", (void *)vd, (u_longlong_t)guid);
+		}
+		return (vd);
+	}
+	return (NULL);
+}
+
+/*
+ * Traverse the vdev tree calling the supplied function.
+ * We continue to walk the tree until we either have walked all
+ * children or we receive a non-NULL return from the callback.
+ * If a NULL callback is passed, then we just return back the first
+ * leaf vdev we encounter.
+ */
+vdev_t *
+vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg)
+{
+	if (vd->vdev_ops->vdev_op_leaf) {
+		if (func == NULL)
+			return (vd);
+		else
+			return (func(vd, arg));
+	}
+
+	for (uint_t c = 0; c < vd->vdev_children; c++) {
+		vdev_t *cvd = vd->vdev_child[c];
+		if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL)
+			return (cvd);
+	}
+	return (NULL);
+}
+
 /*
  * Verify that dynamic LUN growth works as expected.
  */
@@ -1155,44 +1252,108 @@ void
 ztest_vdev_LUN_growth(ztest_args_t *za)
 {
 	spa_t *spa = za->za_spa;
-	char dev_name[MAXPATHLEN];
+	vdev_t *vd, *tvd = NULL;
-	uint64_t leaves = MAX(zopt_mirrors, 1) * zopt_raidz;
+	size_t psize, newsize;
-	uint64_t vdev;
+	uint64_t spa_newsize, spa_cursize, ms_count;
-	size_t fsize;
-	int fd;
 
-	(void) pthread_mutex_lock(&ztest_shared->zs_vdev_lock);
+	(void) mutex_lock(&ztest_shared->zs_vdev_lock);
+	mutex_enter(&spa_namespace_lock);
+	spa_config_enter(spa, SCL_STATE, spa, RW_READER);
 
-	/*
+	while (tvd == NULL || tvd->vdev_islog) {
-	 * Pick a random leaf vdev.
+		uint64_t vdev;
-	 */
-	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
-	vdev = ztest_random(spa->spa_root_vdev->vdev_children * leaves);
-	spa_config_exit(spa, SCL_VDEV, FTAG);
 
-	(void) sprintf(dev_name, ztest_dev_template, zopt_dir, zopt_pool, vdev);
+		vdev = ztest_random(spa->spa_root_vdev->vdev_children);
-
+		tvd = spa->spa_root_vdev->vdev_child[vdev];
-	if ((fd = open(dev_name, O_RDWR)) != -1) {
-		/*
-		 * Determine the size.
-		 */
-		fsize = lseek(fd, 0, SEEK_END);
-
-		/*
-		 * If it's less than 2x the original size, grow by around 3%.
-		 */
-		if (fsize < 2 * zopt_vdev_size) {
-			size_t newsize = fsize + ztest_random(fsize / 32);
-			(void) ftruncate(fd, newsize);
-			if (zopt_verbose >= 6) {
-				(void) printf("%s grew from %lu to %lu bytes\n",
-				    dev_name, (ulong_t)fsize, (ulong_t)newsize);
-			}
-		}
-		(void) close(fd);
 	}
 
-	(void) pthread_mutex_unlock(&ztest_shared->zs_vdev_lock);
+	/*
+	 * Determine the size of the first leaf vdev associated with
+	 * our top-level device.
+	 */
+	vd = vdev_walk_tree(tvd, NULL, NULL);
+	ASSERT3P(vd, !=, NULL);
+	ASSERT(vd->vdev_ops->vdev_op_leaf);
+
+	psize = vd->vdev_psize;
+
+	/*
+	 * We only try to expand the vdev if it's less than 4x its
+	 * original size and it has a valid psize.
+	 */
+	if (psize == 0 || psize >= 4 * zopt_vdev_size) {
+		spa_config_exit(spa, SCL_STATE, spa);
+		mutex_exit(&spa_namespace_lock);
+		(void) mutex_unlock(&ztest_shared->zs_vdev_lock);
+		return;
+	}
+	ASSERT(psize > 0);
+	newsize = psize + psize / 8;
+	ASSERT3U(newsize, >, psize);
+
+	if (zopt_verbose >= 6) {
+		(void) printf("Expanding vdev %s from %lu to %lu\n",
+		    vd->vdev_path, (ulong_t)psize, (ulong_t)newsize);
+	}
+
+	spa_cursize = spa_get_space(spa);
+	ms_count = tvd->vdev_ms_count;
+
+	/*
+	 * Growing the vdev is a two step process:
+	 *	1). expand the physical size (i.e. relabel)
+	 *	2). online the vdev to create the new metaslabs
+	 */
+	if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL ||
+	    vdev_walk_tree(tvd, online_vdev, NULL) != NULL ||
+	    tvd->vdev_state != VDEV_STATE_HEALTHY) {
+		if (zopt_verbose >= 5) {
+			(void) printf("Could not expand LUN because "
+			    "some vdevs were not healthy\n");
+		}
+		(void) spa_config_exit(spa, SCL_STATE, spa);
+		mutex_exit(&spa_namespace_lock);
+		(void) mutex_unlock(&ztest_shared->zs_vdev_lock);
+		return;
+	}
+
+	(void) spa_config_exit(spa, SCL_STATE, spa);
+	mutex_exit(&spa_namespace_lock);
+
+	/*
+	 * Expanding the LUN will update the config asynchronously,
+	 * thus we must wait for the async thread to complete any
+	 * pending tasks before proceeding.
+	 */
+	mutex_enter(&spa->spa_async_lock);
+	while (spa->spa_async_thread != NULL || spa->spa_async_tasks)
+		cv_wait(&spa->spa_async_cv, &spa->spa_async_lock);
+	mutex_exit(&spa->spa_async_lock);
+
+	spa_config_enter(spa, SCL_STATE, spa, RW_READER);
+	spa_newsize = spa_get_space(spa);
+
+	/*
+	 * Make sure we were able to grow the pool.
+	 */
+	if (ms_count >= tvd->vdev_ms_count ||
+	    spa_cursize >= spa_newsize) {
+		(void) printf("Top-level vdev metaslab count: "
+		    "before %llu, after %llu\n",
+		    (u_longlong_t)ms_count,
+		    (u_longlong_t)tvd->vdev_ms_count);
+		fatal(0, "LUN expansion failed: before %llu, "
+		    "after %llu\n", spa_cursize, spa_newsize);
+	} else if (zopt_verbose >= 5) {
+		char oldnumbuf[6], newnumbuf[6];
+
+		nicenum(spa_cursize, oldnumbuf);
+		nicenum(spa_newsize, newnumbuf);
+		(void) printf("%s grew from %s to %s\n",
+		    spa->spa_name, oldnumbuf, newnumbuf);
+	}
+	spa_config_exit(spa, SCL_STATE, spa);
+	(void) mutex_unlock(&ztest_shared->zs_vdev_lock);
 }
 
 /* ARGSUSED */
@@ -1438,7 +1599,8 @@ ztest_dmu_snapshot_create_destroy(ztest_args_t *za)
 	error = dmu_objset_destroy(snapname);
 	if (error != 0 && error != ENOENT)
 		fatal(0, "dmu_objset_destroy() = %d", error);
-	error = dmu_objset_snapshot(osname, strchr(snapname, '@')+1, FALSE);
+	error = dmu_objset_snapshot(osname, strchr(snapname, '@')+1,
+	    NULL, FALSE);
 	if (error == ENOSPC)
 		ztest_record_enospc("dmu_take_snapshot");
 	else if (error != 0 && error != EEXIST)
@@ -1446,6 +1608,148 @@ ztest_dmu_snapshot_create_destroy(ztest_args_t *za)
 	(void) pthread_rwlock_unlock(&ztest_shared->zs_name_lock);
 }
 
+/*
+ * Cleanup non-standard snapshots and clones.
+ */
+void
+ztest_dsl_dataset_cleanup(char *osname, uint64_t curval)
+{
+	char snap1name[100];
+	char clone1name[100];
+	char snap2name[100];
+	char clone2name[100];
+	char snap3name[100];
+	int error;
+
+	(void) snprintf(snap1name, 100, "%s@s1_%llu", osname, curval);
+	(void) snprintf(clone1name, 100, "%s/c1_%llu", osname, curval);
+	(void) snprintf(snap2name, 100, "%s@s2_%llu", clone1name, curval);
+	(void) snprintf(clone2name, 100, "%s/c2_%llu", osname, curval);
+	(void) snprintf(snap3name, 100, "%s@s3_%llu", clone1name, curval);
+
+	error = dmu_objset_destroy(clone2name);
+	if (error && error != ENOENT)
+		fatal(0, "dmu_objset_destroy(%s) = %d", clone2name, error);
+	error = dmu_objset_destroy(snap3name);
+	if (error && error != ENOENT)
+		fatal(0, "dmu_objset_destroy(%s) = %d", snap3name, error);
+	error = dmu_objset_destroy(snap2name);
+	if (error && error != ENOENT)
+		fatal(0, "dmu_objset_destroy(%s) = %d", snap2name, error);
+	error = dmu_objset_destroy(clone1name);
+	if (error && error != ENOENT)
+		fatal(0, "dmu_objset_destroy(%s) = %d", clone1name, error);
+	error = dmu_objset_destroy(snap1name);
+	if (error && error != ENOENT)
+		fatal(0, "dmu_objset_destroy(%s) = %d", snap1name, error);
+}
+
+/*
+ * Verify dsl_dataset_promote handles EBUSY
+ */
+void
+ztest_dsl_dataset_promote_busy(ztest_args_t *za)
+{
+	int error;
+	objset_t *os = za->za_os;
+	objset_t *clone;
+	dsl_dataset_t *ds;
+	char snap1name[100];
+	char clone1name[100];
+	char snap2name[100];
+	char clone2name[100];
+	char snap3name[100];
+	char osname[MAXNAMELEN];
+	uint64_t curval = za->za_instance;
+
+	(void) rw_rdlock(&ztest_shared->zs_name_lock);
+
+	dmu_objset_name(os, osname);
+	ztest_dsl_dataset_cleanup(osname, curval);
+
+	(void) snprintf(snap1name, 100, "%s@s1_%llu", osname, curval);
+	(void) snprintf(clone1name, 100, "%s/c1_%llu", osname, curval);
+	(void) snprintf(snap2name, 100, "%s@s2_%llu", clone1name, curval);
+	(void) snprintf(clone2name, 100, "%s/c2_%llu", osname, curval);
+	(void) snprintf(snap3name, 100, "%s@s3_%llu", clone1name, curval);
+
+	error = dmu_objset_snapshot(osname, strchr(snap1name, '@')+1,
+	    NULL, FALSE);
+	if (error && error != EEXIST) {
+		if (error == ENOSPC) {
+			ztest_record_enospc("dmu_take_snapshot");
+			goto out;
+		}
+		fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error);
+	}
+
+	error = dmu_objset_open(snap1name, DMU_OST_OTHER,
+	    DS_MODE_USER | DS_MODE_READONLY, &clone);
+	if (error)
+		fatal(0, "dmu_open_snapshot(%s) = %d", snap1name, error);
+
+	error = dmu_objset_create(clone1name, DMU_OST_OTHER, clone, 0,
+	    NULL, NULL);
+	dmu_objset_close(clone);
+	if (error) {
+		if (error == ENOSPC) {
+			ztest_record_enospc("dmu_objset_create");
+			goto out;
+		}
+		fatal(0, "dmu_objset_create(%s) = %d", clone1name, error);
+	}
+
+	error = dmu_objset_snapshot(clone1name, strchr(snap2name, '@')+1,
+	    NULL, FALSE);
+	if (error && error != EEXIST) {
+		if (error == ENOSPC) {
+			ztest_record_enospc("dmu_take_snapshot");
+			goto out;
+		}
+		fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error);
+	}
+
+	error = dmu_objset_snapshot(clone1name, strchr(snap3name, '@')+1,
+	    NULL, FALSE);
+	if (error && error != EEXIST) {
+		if (error == ENOSPC) {
+			ztest_record_enospc("dmu_take_snapshot");
+			goto out;
+		}
+		fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error);
+	}
+
+	error = dmu_objset_open(snap3name, DMU_OST_OTHER,
+	    DS_MODE_USER | DS_MODE_READONLY, &clone);
+	if (error)
+		fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error);
+
+	error = dmu_objset_create(clone2name, DMU_OST_OTHER, clone, 0,
+	    NULL, NULL);
+	dmu_objset_close(clone);
+	if (error) {
+		if (error == ENOSPC) {
+			ztest_record_enospc("dmu_objset_create");
+			goto out;
+		}
+		fatal(0, "dmu_objset_create(%s) = %d", clone2name, error);
+	}
+
+	error = dsl_dataset_own(snap1name, DS_MODE_READONLY, FTAG, &ds);
+	if (error)
+		fatal(0, "dsl_dataset_own(%s) = %d", snap1name, error);
+	error = dsl_dataset_promote(clone2name);
+	if (error != EBUSY)
+		fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name,
+		    error);
+	dsl_dataset_disown(ds, FTAG);
+
+out:
+	ztest_dsl_dataset_cleanup(osname, curval);
+
+	(void) rw_unlock(&ztest_shared->zs_name_lock);
+}
+
 /*
  * Verify that dmu_object_{alloc,free} work as expected.
  */
@@ -1469,7 +1773,7 @@ ztest_dmu_object_alloc_free(ztest_args_t *za)
 	 * Create a batch object if necessary, and record it in the directory.
 	 */
 	VERIFY3U(0, ==, dmu_read(os, ZTEST_DIROBJ, za->za_diroff,
-	    sizeof (uint64_t), &batchobj));
+	    sizeof (uint64_t), &batchobj, DMU_READ_PREFETCH));
 	if (batchobj == 0) {
 		tx = dmu_tx_create(os);
 		dmu_tx_hold_write(tx, ZTEST_DIROBJ, za->za_diroff,
@@ -1494,7 +1798,7 @@ ztest_dmu_object_alloc_free(ztest_args_t *za)
 	 */
 	for (b = 0; b < batchsize; b++) {
 		VERIFY3U(0, ==, dmu_read(os, batchobj, b * sizeof (uint64_t),
-		    sizeof (uint64_t), &object));
+		    sizeof (uint64_t), &object, DMU_READ_PREFETCH));
 		if (object == 0)
 			continue;
 		/*
@@ -1529,7 +1833,7 @@ ztest_dmu_object_alloc_free(ztest_args_t *za)
 		 * We expect the word at endoff to be our object number.
 		 */
 		VERIFY(0 == dmu_read(os, object, endoff,
-		    sizeof (uint64_t), &temp));
+		    sizeof (uint64_t), &temp, DMU_READ_PREFETCH));
 
 		if (temp != object) {
 			fatal(0, "bad data in %s, got %llu, expected %llu",
@@ -1714,7 +2018,7 @@ ztest_dmu_read_write(ztest_args_t *za)
 	 * Read the directory info.  If it's the first time, set things up.
 	 */
 	VERIFY(0 == dmu_read(os, ZTEST_DIROBJ, za->za_diroff,
-	    sizeof (dd), &dd));
+	    sizeof (dd), &dd, DMU_READ_PREFETCH));
 	if (dd.dd_chunk == 0) {
 		ASSERT(dd.dd_packobj == 0);
 		ASSERT(dd.dd_bigobj == 0);
@@ -1776,9 +2080,11 @@ ztest_dmu_read_write(ztest_args_t *za)
 	/*
 	 * Read the current contents of our objects.
 	 */
-	error = dmu_read(os, dd.dd_packobj, packoff, packsize, packbuf);
+	error = dmu_read(os, dd.dd_packobj, packoff, packsize, packbuf,
+	    DMU_READ_PREFETCH);
 	ASSERT3U(error, ==, 0);
-	error = dmu_read(os, dd.dd_bigobj, bigoff, bigsize, bigbuf);
+	error = dmu_read(os, dd.dd_bigobj, bigoff, bigsize, bigbuf,
+	    DMU_READ_PREFETCH);
 	ASSERT3U(error, ==, 0);
 
 	/*
@@ -1884,9 +2190,9 @@ ztest_dmu_read_write(ztest_args_t *za)
 		void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);
 
 		VERIFY(0 == dmu_read(os, dd.dd_packobj, packoff,
-		    packsize, packcheck));
+		    packsize, packcheck, DMU_READ_PREFETCH));
 		VERIFY(0 == dmu_read(os, dd.dd_bigobj, bigoff,
-		    bigsize, bigcheck));
+		    bigsize, bigcheck, DMU_READ_PREFETCH));
 
 		ASSERT(bcmp(packbuf, packcheck, packsize) == 0);
 		ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0);
@@ -1899,6 +2205,314 @@ ztest_dmu_read_write(ztest_args_t *za)
 	umem_free(bigbuf, bigsize);
 }
 
+void
+compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf,
+    uint64_t bigsize, uint64_t n, dmu_read_write_dir_t dd, uint64_t txg)
+{
+	uint64_t i;
+	bufwad_t *pack;
+	bufwad_t *bigH;
+	bufwad_t *bigT;
+
+	/*
+	 * For each index from n to n + s, verify that the existing bufwad
+	 * in packobj matches the bufwads at the head and tail of the
+	 * corresponding chunk in bigobj.  Then update all three bufwads
+	 * with the new values we want to write out.
+	 */
+	for (i = 0; i < s; i++) {
+		/* LINTED */
+		pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t));
+		/* LINTED */
+		bigH = (bufwad_t *)((char *)bigbuf + i * dd.dd_chunk);
+		/* LINTED */
+		bigT = (bufwad_t *)((char *)bigH + dd.dd_chunk) - 1;
+
+		ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize);
+		ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize);
+
+		if (pack->bw_txg > txg)
+			fatal(0, "future leak: got %llx, open txg is %llx",
+			    pack->bw_txg, txg);
+
+		if (pack->bw_data != 0 && pack->bw_index != n + i)
+			fatal(0, "wrong index: got %llx, wanted %llx+%llx",
+			    pack->bw_index, n, i);
+
+		if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0)
+			fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH);
+
+		if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0)
+			fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT);
+
+		pack->bw_index = n + i;
+		pack->bw_txg = txg;
+		pack->bw_data = 1 + ztest_random(-2ULL);
+
+		*bigH = *pack;
+		*bigT = *pack;
+	}
+}
+
+void
+ztest_dmu_read_write_zcopy(ztest_args_t *za)
+{
+	objset_t *os = za->za_os;
+	dmu_read_write_dir_t dd;
+	dmu_tx_t *tx;
+	uint64_t i;
+	int error;
+	uint64_t n, s, txg;
+	bufwad_t *packbuf, *bigbuf;
+	uint64_t packoff, packsize, bigoff, bigsize;
+	uint64_t regions = 997;
+	uint64_t stride = 123456789ULL;
+	uint64_t width = 9;
+	dmu_buf_t *bonus_db;
+	arc_buf_t **bigbuf_arcbufs;
+	dmu_object_info_t *doi = &za->za_doi;
+
+	/*
+	 * This test uses two objects, packobj and bigobj, that are always
+	 * updated together (i.e. in the same tx) so that their contents are
+	 * in sync and can be compared.  Their contents relate to each other
+	 * in a simple way: packobj is a dense array of 'bufwad' structures,
+	 * while bigobj is a sparse array of the same bufwads.  Specifically,
+	 * for any index n, there are three bufwads that should be identical:
+	 *
+	 *	packobj, at offset n * sizeof (bufwad_t)
+	 *	bigobj, at the head of the nth chunk
+	 *	bigobj, at the tail of the nth chunk
+	 *
+	 * The chunk size is set equal to bigobj block size so that
+	 * dmu_assign_arcbuf() can be tested for object updates.
+	 */
+
+	/*
+	 * Read the directory info.  If it's the first time, set things up.
+	 */
+	VERIFY(0 == dmu_read(os, ZTEST_DIROBJ, za->za_diroff,
+	    sizeof (dd), &dd, DMU_READ_PREFETCH));
+	if (dd.dd_chunk == 0) {
+		ASSERT(dd.dd_packobj == 0);
+		ASSERT(dd.dd_bigobj == 0);
+		tx = dmu_tx_create(os);
+		dmu_tx_hold_write(tx, ZTEST_DIROBJ, za->za_diroff, sizeof (dd));
+		dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+		error = dmu_tx_assign(tx, TXG_WAIT);
+		if (error) {
+			ztest_record_enospc("create r/w directory");
+			dmu_tx_abort(tx);
+			return;
+		}
+
+		dd.dd_packobj = dmu_object_alloc(os, DMU_OT_UINT64_OTHER, 0,
+		    DMU_OT_NONE, 0, tx);
+		dd.dd_bigobj = dmu_object_alloc(os, DMU_OT_UINT64_OTHER, 0,
+		    DMU_OT_NONE, 0, tx);
+		ztest_set_random_blocksize(os, dd.dd_packobj, tx);
+		ztest_set_random_blocksize(os, dd.dd_bigobj, tx);
+
+		VERIFY(dmu_object_info(os, dd.dd_bigobj, doi) == 0);
+		ASSERT(doi->doi_data_block_size >= 2 * sizeof (bufwad_t));
+		ASSERT(ISP2(doi->doi_data_block_size));
+		dd.dd_chunk = doi->doi_data_block_size;
+
+		dmu_write(os, ZTEST_DIROBJ, za->za_diroff, sizeof (dd), &dd,
+		    tx);
+		dmu_tx_commit(tx);
+	} else {
+		VERIFY(dmu_object_info(os, dd.dd_bigobj, doi) == 0);
+		VERIFY(ISP2(doi->doi_data_block_size));
+		VERIFY(dd.dd_chunk == doi->doi_data_block_size);
+		VERIFY(dd.dd_chunk >= 2 * sizeof (bufwad_t));
+	}
+
+	/*
+	 * Pick a random index and compute the offsets into packobj and bigobj.
+	 */
+	n = ztest_random(regions) * stride + ztest_random(width);
+	s = 1 + ztest_random(width - 1);
+
+	packoff = n * sizeof (bufwad_t);
+	packsize = s * sizeof (bufwad_t);
+
+	bigoff = n * dd.dd_chunk;
+	bigsize = s * dd.dd_chunk;
+
+	packbuf = umem_zalloc(packsize, UMEM_NOFAIL);
+	bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL);
+
+	VERIFY(dmu_bonus_hold(os, dd.dd_bigobj, FTAG, &bonus_db) == 0);
+
+	bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL);
+
+	/*
+	 * Iteration 0 test zcopy for DB_UNCACHED dbufs.
+	 * Iteration 1 test zcopy to already referenced dbufs.
+	 * Iteration 2 test zcopy to dirty dbuf in the same txg.
+	 * Iteration 3 test zcopy to dbuf dirty in previous txg.
+	 * Iteration 4 test zcopy when dbuf is no longer dirty.
+	 * Iteration 5 test zcopy when it can't be done.
+	 * Iteration 6 one more zcopy write.
+	 */
+	for (i = 0; i < 7; i++) {
+		uint64_t j;
+		uint64_t off;
+
+		/*
+		 * In iteration 5 (i == 5) use arcbufs
+		 * that don't match bigobj blksz to test
+		 * dmu_assign_arcbuf() when it can't directly
+		 * assign an arcbuf to a dbuf.
+		 */
+		for (j = 0; j < s; j++) {
+			if (i != 5) {
+				bigbuf_arcbufs[j] =
+				    dmu_request_arcbuf(bonus_db,
+				    dd.dd_chunk);
+			} else {
+				bigbuf_arcbufs[2 * j] =
+				    dmu_request_arcbuf(bonus_db,
+				    dd.dd_chunk / 2);
+				bigbuf_arcbufs[2 * j + 1] =
+				    dmu_request_arcbuf(bonus_db,
+				    dd.dd_chunk / 2);
+			}
+		}
+
+		/*
+		 * Get a tx for the mods to both packobj and bigobj.
+		 */
+		tx = dmu_tx_create(os);
+
+		dmu_tx_hold_write(tx, dd.dd_packobj, packoff, packsize);
+		dmu_tx_hold_write(tx, dd.dd_bigobj, bigoff, bigsize);
+
+		if (ztest_random(100) == 0) {
+			error = -1;
+		} else {
+			error = dmu_tx_assign(tx, TXG_WAIT);
+		}
+
+		if (error) {
+			if (error != -1) {
+				ztest_record_enospc("dmu r/w range");
+			}
+			dmu_tx_abort(tx);
+			umem_free(packbuf, packsize);
+			umem_free(bigbuf, bigsize);
+			for (j = 0; j < s; j++) {
+				if (i != 5) {
+					dmu_return_arcbuf(bigbuf_arcbufs[j]);
+				} else {
+					dmu_return_arcbuf(
+					    bigbuf_arcbufs[2 * j]);
+					dmu_return_arcbuf(
+					    bigbuf_arcbufs[2 * j + 1]);
+				}
+			}
+			umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
+			dmu_buf_rele(bonus_db, FTAG);
+			return;
+		}
+
+		txg = dmu_tx_get_txg(tx);
+
+		/*
+		 * 50% of the time don't read objects in the 1st iteration to
+		 * test dmu_assign_arcbuf() for the case when there're no
+		 * existing dbufs for the specified offsets.
+		 */
+		if (i != 0 || ztest_random(2) != 0) {
+			error = dmu_read(os, dd.dd_packobj, packoff,
+			    packsize, packbuf, DMU_READ_PREFETCH);
+			ASSERT3U(error, ==, 0);
+			error = dmu_read(os, dd.dd_bigobj, bigoff, bigsize,
+			    bigbuf, DMU_READ_PREFETCH);
+			ASSERT3U(error, ==, 0);
+		}
+		compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize,
+		    n, dd, txg);
+
+		/*
+		 * We've verified all the old bufwads, and made new ones.
+		 * Now write them out.
+		 */
+		dmu_write(os, dd.dd_packobj, packoff, packsize, packbuf, tx);
+		if (zopt_verbose >= 6) {
+			(void) printf("writing offset %llx size %llx"
+			    " txg %llx\n",
+			    (u_longlong_t)bigoff,
+			    (u_longlong_t)bigsize,
+			    (u_longlong_t)txg);
+		}
+		for (off = bigoff, j = 0; j < s; j++, off += dd.dd_chunk) {
+			dmu_buf_t *dbt;
+			if (i != 5) {
+				bcopy((caddr_t)bigbuf + (off - bigoff),
+				    bigbuf_arcbufs[j]->b_data, dd.dd_chunk);
+			} else {
+				bcopy((caddr_t)bigbuf + (off - bigoff),
+				    bigbuf_arcbufs[2 * j]->b_data,
+				    dd.dd_chunk / 2);
+				bcopy((caddr_t)bigbuf + (off - bigoff) +
+				    dd.dd_chunk / 2,
+				    bigbuf_arcbufs[2 * j + 1]->b_data,
+				    dd.dd_chunk / 2);
+			}
+
+			if (i == 1) {
+				VERIFY(dmu_buf_hold(os, dd.dd_bigobj, off,
+				    FTAG, &dbt) == 0);
+			}
+			if (i != 5) {
+				dmu_assign_arcbuf(bonus_db, off,
+				    bigbuf_arcbufs[j], tx);
+			} else {
+				dmu_assign_arcbuf(bonus_db, off,
+				    bigbuf_arcbufs[2 * j], tx);
+				dmu_assign_arcbuf(bonus_db,
+				    off + dd.dd_chunk / 2,
+				    bigbuf_arcbufs[2 * j + 1], tx);
+			}
+			if (i == 1) {
+				dmu_buf_rele(dbt, FTAG);
+			}
+		}
+		dmu_tx_commit(tx);
+
+		/*
+		 * Sanity check the stuff we just wrote.
+		 */
+		{
+			void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
+			void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);
+
+			VERIFY(0 == dmu_read(os, dd.dd_packobj, packoff,
+			    packsize, packcheck, DMU_READ_PREFETCH));
+			VERIFY(0 == dmu_read(os, dd.dd_bigobj, bigoff,
+			    bigsize, bigcheck, DMU_READ_PREFETCH));
+
+			ASSERT(bcmp(packbuf, packcheck, packsize) == 0);
+			ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0);
+
+			umem_free(packcheck, packsize);
+			umem_free(bigcheck, bigsize);
+		}
+		if (i == 2) {
+			txg_wait_open(dmu_objset_pool(os), 0);
+		} else if (i == 3) {
+			txg_wait_synced(dmu_objset_pool(os), 0);
+		}
+	}
+
+	dmu_buf_rele(bonus_db, FTAG);
+	umem_free(packbuf, packsize);
+	umem_free(bigbuf, bigsize);
+	umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
+}
+
 void
 ztest_dmu_check_future_leak(ztest_args_t *za)
 {
@@ -1948,6 +2562,8 @@ ztest_dmu_write_parallel(ztest_args_t *za)
 	uint64_t blkoff;
 	zbookmark_t zb;
 	dmu_tx_t *tx = dmu_tx_create(os);
+	dmu_buf_t *bonus_db;
+	arc_buf_t *abuf = NULL;
 
 	dmu_objset_name(os, osname);
 
@@ -1976,6 +2592,12 @@ ztest_dmu_write_parallel(ztest_args_t *za)
 		}
 	}
 
+	if (off != -1ULL && P2PHASE(off, bs) == 0 && !do_free &&
+	    ztest_random(8) == 0) {
+		VERIFY(dmu_bonus_hold(os, ZTEST_DIROBJ, FTAG, &bonus_db) == 0);
+		abuf = dmu_request_arcbuf(bonus_db, bs);
+	}
+
 	txg_how = ztest_random(2) == 0 ? TXG_WAIT : TXG_NOWAIT;
 	error = dmu_tx_assign(tx, txg_how);
 	if (error) {
@@ -1986,6 +2608,10 @@ ztest_dmu_write_parallel(ztest_args_t *za)
 			ztest_record_enospc("dmu write parallel");
 		}
 		dmu_tx_abort(tx);
+		if (abuf != NULL) {
+			dmu_return_arcbuf(abuf);
+			dmu_buf_rele(bonus_db, FTAG);
+		}
 		return;
 	}
 	txg = dmu_tx_get_txg(tx);
@@ -2040,8 +2666,12 @@ ztest_dmu_write_parallel(ztest_args_t *za)
 		za->za_dbuf = NULL;
 	} else if (do_free) {
 		VERIFY(dmu_free_range(os, ZTEST_DIROBJ, off, bs, tx) == 0);
-	} else {
+	} else if (abuf == NULL) {
 		dmu_write(os, ZTEST_DIROBJ, off, btsize, wbt, tx);
+	} else {
+		bcopy(wbt, abuf->b_data, btsize);
+		dmu_assign_arcbuf(bonus_db, off, abuf, tx);
+		dmu_buf_rele(bonus_db, FTAG);
 	}
 
 	(void) pthread_mutex_unlock(lp);
@@ -2077,16 +2707,20 @@ ztest_dmu_write_parallel(ztest_args_t *za)
 	dmu_buf_rele(db, FTAG);
 	za->za_dbuf = NULL;
 
-	(void) pthread_mutex_unlock(lp);
+	if (error) {
-
+		(void) mutex_unlock(lp);
-	if (error)
 		return;
+	}
 
-	if (blk.blk_birth == 0)		/* concurrent free */
+	if (blk.blk_birth == 0)	{	/* concurrent free */
+		(void) mutex_unlock(lp);
 		return;
+	}
 
 	txg_suspend(dmu_objset_pool(os));
 
+	(void) mutex_unlock(lp);
+
 	ASSERT(blk.blk_fill == 1);
 	ASSERT3U(BP_GET_TYPE(&blk), ==, DMU_OT_UINT64_OTHER);
 	ASSERT3U(BP_GET_LEVEL(&blk), ==, 0);
@@ -2159,7 +2793,7 @@ ztest_zap(ztest_args_t *za)
 	 * Create a new object if necessary, and record it in the directory.
 	 */
 	VERIFY(0 == dmu_read(os, ZTEST_DIROBJ, za->za_diroff,
-	    sizeof (uint64_t), &object));
+	    sizeof (uint64_t), &object, DMU_READ_PREFETCH));
 
 	if (object == 0) {
 		tx = dmu_tx_create(os);
@@ -3011,7 +3645,7 @@ ztest_verify_blocks(char *pool)
 	isa = strdup(isa);
 	/* LINTED */
 	(void) sprintf(bin,
-	    "/usr/sbin%.*s/zdb -bc%s%s -U /tmp/zpool.cache %s",
+	    "/usr/sbin%.*s/zdb -bcc%s%s -U /tmp/zpool.cache %s",
 	    isalen,
 	    isa,
 	    zopt_verbose >= 3 ? "s" : "",
@@ -3158,7 +3792,7 @@ ztest_resume(void *arg)
 		spa_vdev_state_enter(spa);
 		vdev_clear(spa, NULL);
 		(void) spa_vdev_state_exit(spa, NULL, 0);
-		zio_resume(spa);
+		(void) zio_resume(spa);
 	}
 	return (NULL);
 }
@@ -3434,6 +4068,10 @@ ztest_run(char *pool)
 		(void) snprintf(name, 100, "%s/%s_%d", pool, pool, d);
 		if (zopt_verbose >= 3)
 			(void) printf("Destroying %s to free up space\n", name);
+
+		/* Cleanup any non-standard clones and snapshots */
+		ztest_dsl_dataset_cleanup(name, za[d].za_instance);
+
 		(void) dmu_objset_find(name, ztest_destroy_cb, &za[d],
 		    DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
 		(void) pthread_rwlock_unlock(&ztest_shared->zs_name_lock);
@@ -3520,6 +4158,8 @@ ztest_init(char *pool)
 	if (error)
 		fatal(0, "spa_open() = %d", error);
 
+	metaslab_sz = 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
+
 	if (zopt_verbose >= 3)
 		show_pool_stats(spa);
 
@@ -3611,6 +4251,9 @@ main(int argc, char **argv)
 			zi->zi_call_time = 0;
 		}
 
+		/* Set the allocation switch size */
+		metaslab_df_alloc_threshold = ztest_random(metaslab_sz / 4) + 1;
+
 		pid = fork();
 
 		if (pid == -1)

lib/libzfs/include/libzfs.h

@@ -116,6 +116,7 @@ enum {
 	EZFS_VDEVNOTSUP,	/* unsupported vdev type */
 	EZFS_NOTSUP,		/* ops not supported on this dataset */
 	EZFS_ACTIVE_SPARE,	/* pool has active shared spare devices */
+	EZFS_UNPLAYED_LOGS,	/* log device has unplayed logs */
 	EZFS_UNKNOWN
 };
 
@@ -178,6 +179,7 @@ extern const char *libzfs_error_action(libzfs_handle_t *);
 extern const char *libzfs_error_description(libzfs_handle_t *);
 extern void libzfs_mnttab_init(libzfs_handle_t *);
 extern void libzfs_mnttab_fini(libzfs_handle_t *);
+extern void libzfs_mnttab_cache(libzfs_handle_t *, boolean_t);
 extern int libzfs_mnttab_find(libzfs_handle_t *, const char *,
     struct mnttab *);
 extern void libzfs_mnttab_add(libzfs_handle_t *, const char *,
@@ -229,6 +231,8 @@ extern int zpool_vdev_clear(zpool_handle_t *, uint64_t);
 
 extern nvlist_t *zpool_find_vdev(zpool_handle_t *, const char *, boolean_t *,
     boolean_t *, boolean_t *);
+extern nvlist_t *zpool_find_vdev_by_physpath(zpool_handle_t *, const char *,
+    boolean_t *, boolean_t *, boolean_t *);
 extern int zpool_label_disk(libzfs_handle_t *, zpool_handle_t *, char *);
 
 /*
@@ -335,7 +339,8 @@ extern int zpool_stage_history(libzfs_handle_t *, const char *);
 extern void zpool_obj_to_path(zpool_handle_t *, uint64_t, uint64_t, char *,
     size_t len);
 extern int zfs_ioctl(libzfs_handle_t *, int, struct zfs_cmd *);
-extern int zpool_get_physpath(zpool_handle_t *, char *);
+extern int zpool_get_physpath(zpool_handle_t *, char *, size_t);
+
 /*
  * Basic handle manipulations.  These functions do not create or destroy the
  * underlying datasets, only the references to them.
@@ -368,6 +373,10 @@ extern int zfs_prop_get(zfs_handle_t *, zfs_prop_t, char *, size_t,
     zprop_source_t *, char *, size_t, boolean_t);
 extern int zfs_prop_get_numeric(zfs_handle_t *, zfs_prop_t, uint64_t *,
     zprop_source_t *, char *, size_t);
+extern int zfs_prop_get_userquota_int(zfs_handle_t *zhp, const char *propname,
+    uint64_t *propvalue);
+extern int zfs_prop_get_userquota(zfs_handle_t *zhp, const char *propname,
+    char *propbuf, int proplen, boolean_t literal);
 extern uint64_t zfs_prop_get_int(zfs_handle_t *, zfs_prop_t);
 extern int zfs_prop_inherit(zfs_handle_t *, const char *);
 extern const char *zfs_prop_values(zfs_prop_t);
@@ -384,6 +393,7 @@ typedef struct zprop_list {
 } zprop_list_t;
 
 extern int zfs_expand_proplist(zfs_handle_t *, zprop_list_t **);
+extern void zfs_prune_proplist(zfs_handle_t *, uint8_t *);
 
 #define	ZFS_MOUNTPOINT_NONE	"none"
 #define	ZFS_MOUNTPOINT_LEGACY	"legacy"
@@ -454,6 +464,12 @@ extern int zfs_send(zfs_handle_t *, const char *, const char *,
     boolean_t, boolean_t, boolean_t, boolean_t, int);
 extern int zfs_promote(zfs_handle_t *);
 
+typedef int (*zfs_userspace_cb_t)(void *arg, const char *domain,
+    uid_t rid, uint64_t space);
+
+extern int zfs_userspace(zfs_handle_t *zhp, zfs_userquota_prop_t type,
+    zfs_userspace_cb_t func, void *arg);
+
 typedef struct recvflags {
 	/* print informational messages (ie, -v was specified) */
 	int verbose : 1;
@@ -491,17 +507,6 @@ extern boolean_t zfs_dataset_exists(libzfs_handle_t *, const char *,
     zfs_type_t);
 extern int zfs_spa_version(zfs_handle_t *, int *);
 
-/*
- * dataset permission functions.
- */
-extern int zfs_perm_set(zfs_handle_t *, nvlist_t *);
-extern int zfs_perm_remove(zfs_handle_t *, nvlist_t *);
-extern int zfs_build_perms(zfs_handle_t *, char *, char *,
-    zfs_deleg_who_type_t, zfs_deleg_inherit_t, nvlist_t **nvlist_t);
-extern int zfs_perm_get(zfs_handle_t *, zfs_allow_t **);
-extern void zfs_free_allows(zfs_allow_t *);
-extern void zfs_deleg_permissions(void);
-
 /*
  * Mount support functions.
  */
@@ -536,7 +541,7 @@ extern boolean_t zfs_is_shared_iscsi(zfs_handle_t *);
 extern int zfs_share_iscsi(zfs_handle_t *);
 extern int zfs_unshare_iscsi(zfs_handle_t *);
 extern int zfs_iscsi_perm_check(libzfs_handle_t *, char *, ucred_t *);
-extern int zfs_deleg_share_nfs(libzfs_handle_t *, char *, char *,
+extern int zfs_deleg_share_nfs(libzfs_handle_t *, char *, char *, char *,
     void *, void *, int, zfs_share_op_t);
 
 /*
@@ -574,6 +579,15 @@ extern int zpool_remove_zvol_links(zpool_handle_t *);
 /* is this zvol valid for use as a dump device? */
 extern int zvol_check_dump_config(char *);
 
+/*
+ * Management interfaces for SMB ACL files
+ */
+
+int zfs_smb_acl_add(libzfs_handle_t *, char *, char *, char *);
+int zfs_smb_acl_remove(libzfs_handle_t *, char *, char *, char *);
+int zfs_smb_acl_purge(libzfs_handle_t *, char *, char *);
+int zfs_smb_acl_rename(libzfs_handle_t *, char *, char *, char *, char *);
+
 /*
  * Enable and disable datasets within a pool by mounting/unmounting and
  * sharing/unsharing them.

lib/libzfs/include/libzfs_impl.h

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -63,6 +63,7 @@ struct libzfs_handle {
 	int libzfs_printerr;
 	void *libzfs_sharehdl; /* libshare handle */
 	uint_t libzfs_shareflags;
+	boolean_t libzfs_mnttab_enable;
 	avl_tree_t libzfs_mnttab_cache;
 };
 #define	ZFSSHARE_MISS	0x01	/* Didn't find entry in cache */
@@ -78,6 +79,7 @@ struct zfs_handle {
 	nvlist_t *zfs_user_props;
 	boolean_t zfs_mntcheck;
 	char *zfs_mntopts;
+	uint8_t *zfs_props_table;
 };
 
 /*
@@ -185,7 +187,7 @@ extern int zfs_init_libshare(libzfs_handle_t *, int);
 extern void zfs_uninit_libshare(libzfs_handle_t *);
 extern int zfs_parse_options(char *, zfs_share_proto_t);
 
-extern int zfs_unshare_proto(zfs_handle_t *zhp,
+extern int zfs_unshare_proto(zfs_handle_t *,
     const char *, zfs_share_proto_t *);
 #ifdef	__cplusplus
 }

lib/libzfs/libzfs_changelist.c

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  *
  * Portions Copyright 2007 Ramprakash Jelari
@@ -218,6 +218,7 @@ changelist_postfix(prop_changelist_t *clp)
 
 		boolean_t sharenfs;
 		boolean_t sharesmb;
+		boolean_t mounted;
 
 		/*
 		 * If we are in the global zone, but this dataset is exported
@@ -272,20 +273,29 @@ changelist_postfix(prop_changelist_t *clp)
 		    shareopts, sizeof (shareopts), NULL, NULL, 0,
 		    B_FALSE) == 0) && (strcmp(shareopts, "off") != 0));
 
-		if ((cn->cn_mounted || clp->cl_waslegacy || sharenfs ||
+		mounted = zfs_is_mounted(cn->cn_handle, NULL);
-		    sharesmb) && !zfs_is_mounted(cn->cn_handle, NULL) &&
+
-		    zfs_mount(cn->cn_handle, NULL, 0) != 0)
+		if (!mounted && (cn->cn_mounted ||
-			errors++;
+		    ((sharenfs || sharesmb || clp->cl_waslegacy) &&
+		    (zfs_prop_get_int(cn->cn_handle,
+		    ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_ON)))) {
+
+			if (zfs_mount(cn->cn_handle, NULL, 0) != 0)
+				errors++;
+			else
+				mounted = TRUE;
+		}
 
 		/*
-		 * We always re-share even if the filesystem is currently
+		 * If the file system is mounted we always re-share even
-		 * shared, so that we can adopt any new options.
+		 * if the filesystem is currently shared, so that we can
+		 * adopt any new options.
 		 */
-		if (sharenfs)
+		if (sharenfs && mounted)
 			errors += zfs_share_nfs(cn->cn_handle);
 		else if (cn->cn_shared || clp->cl_waslegacy)
 			errors += zfs_unshare_nfs(cn->cn_handle, NULL);
-		if (sharesmb)
+		if (sharesmb && mounted)
 			errors += zfs_share_smb(cn->cn_handle);
 		else if (cn->cn_shared || clp->cl_waslegacy)
 			errors += zfs_unshare_smb(cn->cn_handle, NULL);
@@ -621,8 +631,6 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
 	} else if (prop == ZFS_PROP_VOLSIZE) {
 		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
-	} else if (prop == ZFS_PROP_VERSION) {
-		clp->cl_prop = ZFS_PROP_MOUNTPOINT;
 	} else {
 		clp->cl_prop = prop;
 	}

lib/libzfs/libzfs_graph.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * Iterate over all children of the current object.  This includes the normal
  * dataset hierarchy, but also arbitrary hierarchies due to clones.  We want to
@@ -399,13 +397,6 @@ iterate_children(libzfs_handle_t *hdl, zfs_graph_t *zgp, const char *dataset)
 	for ((void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
 	    ioctl(hdl->libzfs_fd, ZFS_IOC_DATASET_LIST_NEXT, &zc) == 0;
 	    (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name))) {
-
-		/*
-		 * Ignore private dataset names.
-		 */
-		if (dataset_name_hidden(zc.zc_name))
-			continue;
-
 		/*
 		 * Get statistics for this dataset, to determine the type of the
 		 * dataset and clone statistics.  If this fails, the dataset has

lib/libzfs/libzfs_pool.c

@@ -42,6 +42,7 @@
 #include <sys/zfs_ioctl.h>
 #include <sys/zio.h>
 #include <strings.h>
+#include <dlfcn.h>
 
 #include "zfs_namecheck.h"
 #include "zfs_prop.h"
@@ -55,6 +56,10 @@ static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
 #define	BOOTCMD	"installboot(1M)"
 #endif
 
+#define	DISK_ROOT	"/dev/dsk"
+#define	RDISK_ROOT	"/dev/rdsk"
+#define	BACKUP_SLICE	"s2"
+
 /*
  * ====================================================================
  *   zpool property functions
@@ -627,6 +632,12 @@ zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
 }
 
 
+/*
+ * Don't start the slice at the default block of 34; many storage
+ * devices will use a stripe width of 128k, so start there instead.
+ */
+#define	NEW_START_BLOCK	256
+
 /*
  * Validate the given pool name, optionally putting an extended error message in
  * 'buf'.
@@ -1369,46 +1380,90 @@ zpool_scrub(zpool_handle_t *zhp, pool_scrub_type_t type)
 }
 
 /*
+ * Find a vdev that matches the search criteria specified. We use the
+ * the nvpair name to determine how we should look for the device.
  * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
  * spare; but FALSE if its an INUSE spare.
  */
 static nvlist_t *
-vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
+vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
-    boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
+    boolean_t *l2cache, boolean_t *log)
 {
 	uint_t c, children;
 	nvlist_t **child;
-	uint64_t theguid, present;
-	char *path;
-	uint64_t wholedisk = 0;
 	nvlist_t *ret;
 	uint64_t is_log;
+	char *srchkey;
+	nvpair_t *pair = nvlist_next_nvpair(search, NULL);
 
-	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &theguid) == 0);
+	/* Nothing to look for */
+	if (search == NULL || pair == NULL)
+		return (NULL);
 
-	if (search == NULL &&
+	/* Obtain the key we will use to search */
-	    nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &present) == 0) {
+	srchkey = nvpair_name(pair);
-		/*
+
-		 * If the device has never been present since import, the only
+	switch (nvpair_type(pair)) {
-		 * reliable way to match the vdev is by GUID.
+	case DATA_TYPE_UINT64: {
-		 */
+		uint64_t srchval, theguid, present;
-		if (theguid == guid)
+
-			return (nv);
+		verify(nvpair_value_uint64(pair, &srchval) == 0);
-	} else if (search != NULL &&
+		if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
-	    nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
+			if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
-		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
+			    &present) == 0) {
-		    &wholedisk);
+				/*
-		if (wholedisk) {
+				 * If the device has never been present since
-			/*
+				 * import, the only reliable way to match the
-			 * For whole disks, the internal path has 's0', but the
+				 * vdev is by GUID.
-			 * path passed in by the user doesn't.
+				 */
-			 */
+				verify(nvlist_lookup_uint64(nv,
-			if (strlen(search) == strlen(path) - 2 &&
+				    ZPOOL_CONFIG_GUID, &theguid) == 0);
-			    strncmp(search, path, strlen(search)) == 0)
+				if (theguid == srchval)
-				return (nv);
+					return (nv);
-		} else if (strcmp(search, path) == 0) {
+			}
-			return (nv);
 		}
+		break;
+	}
+
+	case DATA_TYPE_STRING: {
+		char *srchval, *val;
+
+		verify(nvpair_value_string(pair, &srchval) == 0);
+		if (nvlist_lookup_string(nv, srchkey, &val) != 0)
+			break;
+
+		/*
+		 * Search for the requested value. We special case the search
+		 * for ZPOOL_CONFIG_PATH when it's a wholedisk. Otherwise,
+		 * all other searches are simple string compares.
+		 */
+		if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && val) {
+			uint64_t wholedisk = 0;
+
+			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
+			    &wholedisk);
+			if (wholedisk) {
+				/*
+				 * For whole disks, the internal path has 's0',
+				 * but the path passed in by the user doesn't.
+				 */
+				if (strlen(srchval) == strlen(val) - 2 &&
+				    strncmp(srchval, val, strlen(srchval)) == 0)
+					return (nv);
+				break;
+			}
+		}
+
+		/*
+		 * Common case
+		 */
+		if (strcmp(srchval, val) == 0)
+			return (nv);
+		break;
+	}
+
+	default:
+		break;
 	}
 
 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
@@ -1416,7 +1471,7 @@ vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
 		return (NULL);
 
 	for (c = 0; c < children; c++) {
-		if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
+		if ((ret = vdev_to_nvlist_iter(child[c], search,
 		    avail_spare, l2cache, NULL)) != NULL) {
 			/*
 			 * The 'is_log' value is only set for the toplevel
@@ -1437,7 +1492,7 @@ vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
 	    &child, &children) == 0) {
 		for (c = 0; c < children; c++) {
-			if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
+			if ((ret = vdev_to_nvlist_iter(child[c], search,
 			    avail_spare, l2cache, NULL)) != NULL) {
 				*avail_spare = B_TRUE;
 				return (ret);
@@ -1448,7 +1503,7 @@ vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
 	    &child, &children) == 0) {
 		for (c = 0; c < children; c++) {
-			if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
+			if ((ret = vdev_to_nvlist_iter(child[c], search,
 			    avail_spare, l2cache, NULL)) != NULL) {
 				*l2cache = B_TRUE;
 				return (ret);
@@ -1459,24 +1514,48 @@ vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
 	return (NULL);
 }
 
+/*
+ * Given a physical path (minus the "/devices" prefix), find the
+ * associated vdev.
+ */
+nvlist_t *
+zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
+    boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
+{
+	nvlist_t *search, *nvroot, *ret;
+
+	verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+	verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
+
+	verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
+	    &nvroot) == 0);
+
+	*avail_spare = B_FALSE;
+	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
+	nvlist_free(search);
+
+	return (ret);
+}
+
 nvlist_t *
 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
     boolean_t *l2cache, boolean_t *log)
 {
 	char buf[MAXPATHLEN];
-	const char *search;
 	char *end;
-	nvlist_t *nvroot;
+	nvlist_t *nvroot, *search, *ret;
 	uint64_t guid;
 
+	verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
 	guid = strtoull(path, &end, 10);
 	if (guid != 0 && *end == '\0') {
-		search = NULL;
+		verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
 	} else if (path[0] != '/') {
 		(void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
-		search = buf;
+		verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
 	} else {
-		search = path;
+		verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
 	}
 
 	verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
@@ -1486,8 +1565,10 @@ zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
 	*l2cache = B_FALSE;
 	if (log != NULL)
 		*log = B_FALSE;
-	return (vdev_to_nvlist_iter(nvroot, search, guid, avail_spare,
+	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
-	    l2cache, log));
+	nvlist_free(search);
+
+	return (ret);
 }
 
 static int
@@ -1504,80 +1585,141 @@ vdev_online(nvlist_t *nv)
 }
 
 /*
- * Get phys_path for a root pool
+ * Helper function for zpool_get_physpaths().
- * Return 0 on success; non-zeron on failure.
  */
-int
+static int
-zpool_get_physpath(zpool_handle_t *zhp, char *physpath)
+vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
+    size_t *bytes_written)
 {
-	nvlist_t *vdev_root;
+	size_t bytes_left, pos, rsz;
-	nvlist_t **child;
+	char *tmppath;
-	uint_t count;
+	const char *format;
-	int i;
 
-	/*
+	if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
-	 * Make sure this is a root pool, as phys_path doesn't mean
+	    &tmppath) != 0)
-	 * anything to a non-root pool.
+		return (EZFS_NODEVICE);
-	 */
-	if (!pool_is_bootable(zhp))
-		return (-1);
 
-	verify(nvlist_lookup_nvlist(zhp->zpool_config,
+	pos = *bytes_written;
-	    ZPOOL_CONFIG_VDEV_TREE, &vdev_root) == 0);
+	bytes_left = physpath_size - pos;
+	format = (pos == 0) ? "%s" : " %s";
 
-	if (nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
+	rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
-	    &child, &count) != 0)
+	*bytes_written += rsz;
-		return (-2);
 
-	for (i = 0; i < count; i++) {
+	if (rsz >= bytes_left) {
-		nvlist_t **child2;
+		/* if physpath was not copied properly, clear it */
-		uint_t count2;
+		if (bytes_left != 0) {
-		char *type;
+			physpath[pos] = 0;
-		char *tmppath;
+		}
-		int j;
+		return (EZFS_NOSPC);
+	}
+	return (0);
+}
 
-		if (nvlist_lookup_string(child[i], ZPOOL_CONFIG_TYPE, &type)
+static int
-		    != 0)
+vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
-			return (-3);
+    size_t *rsz, boolean_t is_spare)
+{
+	char *type;
+	int ret;
 
-		if (strcmp(type, VDEV_TYPE_DISK) == 0) {
+	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
-			if (!vdev_online(child[i]))
+		return (EZFS_INVALCONFIG);
-				return (-8);
-			verify(nvlist_lookup_string(child[i],
-			    ZPOOL_CONFIG_PHYS_PATH, &tmppath) == 0);
-			(void) strncpy(physpath, tmppath, strlen(tmppath));
-		} else if (strcmp(type, VDEV_TYPE_MIRROR) == 0) {
-			if (nvlist_lookup_nvlist_array(child[i],
-			    ZPOOL_CONFIG_CHILDREN, &child2, &count2) != 0)
-				return (-4);
 
-			for (j = 0; j < count2; j++) {
+	if (strcmp(type, VDEV_TYPE_DISK) == 0) {
-				if (!vdev_online(child2[j]))
+		/*
-					return (-8);
+		 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
-				if (nvlist_lookup_string(child2[j],
+		 * For a spare vdev, we only want to boot from the active
-				    ZPOOL_CONFIG_PHYS_PATH, &tmppath) != 0)
+		 * spare device.
-					return (-5);
+		 */
+		if (is_spare) {
+			uint64_t spare = 0;
+			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
+			    &spare);
+			if (!spare)
+				return (EZFS_INVALCONFIG);
+		}
 
-				if ((strlen(physpath) + strlen(tmppath)) >
+		if (vdev_online(nv)) {
-				    MAXNAMELEN)
+			if ((ret = vdev_get_one_physpath(nv, physpath,
-					return (-6);
+			    phypath_size, rsz)) != 0)
+				return (ret);
+		}
+	} else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
+	    strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
+	    (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
+		nvlist_t **child;
+		uint_t count;
+		int i, ret;
 
-				if (strlen(physpath) == 0) {
+		if (nvlist_lookup_nvlist_array(nv,
-					(void) strncpy(physpath, tmppath,
+		    ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
-					    strlen(tmppath));
+			return (EZFS_INVALCONFIG);
-				} else {
+
-					(void) strcat(physpath, " ");
+		for (i = 0; i < count; i++) {
-					(void) strcat(physpath, tmppath);
+			ret = vdev_get_physpaths(child[i], physpath,
-				}
+			    phypath_size, rsz, is_spare);
-			}
+			if (ret == EZFS_NOSPC)
-		} else {
+				return (ret);
-			return (-7);
 		}
 	}
 
+	return (EZFS_POOL_INVALARG);
+}
+
+/*
+ * Get phys_path for a root pool config.
+ * Return 0 on success; non-zero on failure.
+ */
+static int
+zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
+{
+	size_t rsz;
+	nvlist_t *vdev_root;
+	nvlist_t **child;
+	uint_t count;
+	char *type;
+
+	rsz = 0;
+
+	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
+	    &vdev_root) != 0)
+		return (EZFS_INVALCONFIG);
+
+	if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
+	    nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
+	    &child, &count) != 0)
+		return (EZFS_INVALCONFIG);
+
+	/*
+	 * root pool can not have EFI labeled disks and can only have
+	 * a single top-level vdev.
+	 */
+	if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
+	    pool_uses_efi(vdev_root))
+		return (EZFS_POOL_INVALARG);
+
+	(void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
+	    B_FALSE);
+
+	/* No online devices */
+	if (rsz == 0)
+		return (EZFS_NODEVICE);
+
 	return (0);
 }
 
+/*
+ * Get phys_path for a root pool
+ * Return 0 on success; non-zero on failure.
+ */
+int
+zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
+{
+	return (zpool_get_config_physpath(zhp->zpool_config, physpath,
+	    phypath_size));
+}
+
 /*
  * Returns TRUE if the given guid corresponds to the given type.
  * This is used to check for hot spares (INUSE or not), and level 2 cache
@@ -1606,6 +1748,45 @@ is_guid_type(zpool_handle_t *zhp, uint64_t guid, const char *type)
 	return (B_FALSE);
 }
 
+/*
+ * If the device has being dynamically expanded then we need to relabel
+ * the disk to use the new unallocated space.
+ */
+static int
+zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
+{
+	char path[MAXPATHLEN];
+	char errbuf[1024];
+	int fd, error;
+	int (*_efi_use_whole_disk)(int);
+
+	if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
+	    "efi_use_whole_disk")) == NULL)
+		return (-1);
+
+	(void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
+
+	if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
+		    "relabel '%s': unable to open device"), name);
+		return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
+	}
+
+	/*
+	 * It's possible that we might encounter an error if the device
+	 * does not have any unallocated space left. If so, we simply
+	 * ignore that error and continue on.
+	 */
+	error = _efi_use_whole_disk(fd);
+	(void) close(fd);
+	if (error && error != VT_ENOSPC) {
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
+		    "relabel '%s': unable to read disk capacity"), name);
+		return (zfs_error(hdl, EZFS_NOCAP, errbuf));
+	}
+	return (0);
+}
+
 /*
  * Bring the specified vdev online.   The 'flags' parameter is a set of the
  * ZFS_ONLINE_* flags.
@@ -1617,15 +1798,20 @@ zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
 	zfs_cmd_t zc = { 0 };
 	char msg[1024];
 	nvlist_t *tgt;
-	boolean_t avail_spare, l2cache;
+	boolean_t avail_spare, l2cache, islog;
 	libzfs_handle_t *hdl = zhp->zpool_hdl;
 
-	(void) snprintf(msg, sizeof (msg),
+	if (flags & ZFS_ONLINE_EXPAND) {
-	    dgettext(TEXT_DOMAIN, "cannot online %s"), path);
+		(void) snprintf(msg, sizeof (msg),
+		    dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
+	} else {
+		(void) snprintf(msg, sizeof (msg),
+		    dgettext(TEXT_DOMAIN, "cannot online %s"), path);
+	}
 
 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
-	    NULL)) == NULL)
+	    &islog)) == NULL)
 		return (zfs_error(hdl, EZFS_NODEVICE, msg));
 
 	verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
@@ -1634,6 +1820,31 @@ zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
 	    is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE)
 		return (zfs_error(hdl, EZFS_ISSPARE, msg));
 
+	if (flags & ZFS_ONLINE_EXPAND ||
+	    zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
+		char *pathname = NULL;
+		uint64_t wholedisk = 0;
+
+		(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
+		    &wholedisk);
+		verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
+		    &pathname) == 0);
+
+		/*
+		 * XXX - L2ARC 1.0 devices can't support expansion.
+		 */
+		if (l2cache) {
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "cannot expand cache devices"));
+			return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
+		}
+
+		if (wholedisk) {
+			pathname += strlen(DISK_ROOT) + 1;
+			(void) zpool_relabel_disk(zhp->zpool_hdl, pathname);
+		}
+	}
+
 	zc.zc_cookie = VDEV_STATE_ONLINE;
 	zc.zc_obj = flags;
 
@@ -1684,6 +1895,12 @@ zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
 		 */
 		return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
 
+	case EEXIST:
+		/*
+		 * The log device has unplayed logs
+		 */
+		return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
+
 	default:
 		return (zpool_standard_error(hdl, errno, msg));
 	}
@@ -1888,6 +2105,14 @@ zpool_vdev_attach(zpool_handle_t *zhp,
 			(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please "
 			    "be sure to invoke %s to make '%s' bootable.\n"),
 			    BOOTCMD, new_disk);
+
+			/*
+			 * XXX need a better way to prevent user from
+			 * booting up a half-baked vdev.
+			 */
+			(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
+			    "sure to wait until resilver is done "
+			    "before rebooting.\n"));
 		}
 		return (0);
 	}
@@ -2803,14 +3028,6 @@ zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
 	free(mntpnt);
 }
 
-#define	RDISK_ROOT	"/dev/rdsk"
-#define	BACKUP_SLICE	"s2"
-/*
- * Don't start the slice at the default block of 34; many storage
- * devices will use a stripe width of 128k, so start there instead.
- */
-#define	NEW_START_BLOCK	256
-
 /*
  * Read the EFI label from the config, if a label does not exist then
  * pass back the error to the caller. If the caller has passed a non-NULL

lib/libzfs/libzfs_sendrecv.c

@@ -237,6 +237,8 @@ send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv)
 		zfs_prop_t prop = zfs_name_to_prop(propname);
 		nvlist_t *propnv;
 
+		assert(zfs_prop_user(propname) || prop != ZPROP_INVAL);
+
 		if (!zfs_prop_user(propname) && zfs_prop_readonly(prop))
 			continue;
 
@@ -594,12 +596,18 @@ dump_filesystem(zfs_handle_t *zhp, void *arg)
 			    zhp->zfs_name, sdd->fromsnap);
 			sdd->err = B_TRUE;
 		} else if (!sdd->seento) {
-			(void) fprintf(stderr,
+			if (sdd->fromsnap) {
-			    "WARNING: could not send %s@%s:\n"
+				(void) fprintf(stderr,
-			    "incremental source (%s@%s) "
+				    "WARNING: could not send %s@%s:\n"
-			    "is not earlier than it\n",
+				    "incremental source (%s@%s) "
-			    zhp->zfs_name, sdd->tosnap,
+				    "is not earlier than it\n",
-			    zhp->zfs_name, sdd->fromsnap);
+				    zhp->zfs_name, sdd->tosnap,
+				    zhp->zfs_name, sdd->fromsnap);
+			} else {
+				(void) fprintf(stderr, "WARNING: "
+				    "could not send %s@%s: does not exist\n",
+				    zhp->zfs_name, sdd->tosnap);
+			}
 			sdd->err = B_TRUE;
 		}
 	} else {

lib/libzfs/libzfs_util.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -210,6 +210,9 @@ libzfs_error_description(libzfs_handle_t *hdl)
 	case EZFS_ACTIVE_SPARE:
 		return (dgettext(TEXT_DOMAIN, "pool has active shared spare "
 		    "device"));
+	case EZFS_UNPLAYED_LOGS:
+		return (dgettext(TEXT_DOMAIN, "log device has unplayed intent "
+		    "logs"));
 	case EZFS_UNKNOWN:
 		return (dgettext(TEXT_DOMAIN, "unknown error"));
 	default:
@@ -364,6 +367,11 @@ zfs_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
 	case ENOTSUP:
 		zfs_verror(hdl, EZFS_BADVERSION, fmt, ap);
 		break;
+	case EAGAIN:
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+		    "pool I/O is currently suspended"));
+		zfs_verror(hdl, EZFS_POOLUNAVAIL, fmt, ap);
+		break;
 	default:
 		zfs_error_aux(hdl, strerror(errno));
 		zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);
@@ -437,6 +445,11 @@ zpool_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
 	case EDQUOT:
 		zfs_verror(hdl, EZFS_NOSPC, fmt, ap);
 		return (-1);
+	case EAGAIN:
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+		    "pool I/O is currently suspended"));
+		zfs_verror(hdl, EZFS_POOLUNAVAIL, fmt, ap);
+		break;
 
 	default:
 		zfs_error_aux(hdl, strerror(error));
@@ -575,6 +588,7 @@ libzfs_init(void)
 
 	zfs_prop_init();
 	zpool_prop_init();
+	libzfs_mnttab_init(hdl);
 
 	return (hdl);
 }
@@ -592,6 +606,7 @@ libzfs_fini(libzfs_handle_t *hdl)
 		(void) free(hdl->libzfs_log_str);
 	zpool_free_handles(hdl);
 	namespace_clear(hdl);
+	libzfs_mnttab_fini(hdl);
 	free(hdl);
 }
 
@@ -1209,7 +1224,7 @@ addlist(libzfs_handle_t *hdl, char *propname, zprop_list_t **listp,
 	 * dataset property,
 	 */
 	if (prop == ZPROP_INVAL && (type == ZFS_TYPE_POOL ||
-	    !zfs_prop_user(propname))) {
+	    (!zfs_prop_user(propname) && !zfs_prop_userquota(propname)))) {
 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 		    "invalid property '%s'"), propname);
 		return (zfs_error(hdl, EZFS_BADPROP,

lib/libzpool/include/sys/zfs_context.h

@@ -58,6 +58,7 @@ extern "C" {
 #include <atomic.h>
 #include <dirent.h>
 #include <time.h>
+#include <libsysevent.h>
 #include <sys/note.h>
 #include <sys/types.h>
 #include <sys/cred.h>
@@ -72,6 +73,7 @@ extern "C" {
 #include <sys/kstat.h>
 #include <sys/u8_textprep.h>
 #include <sys/sysevent/eventdefs.h>
+#include <sys/sysevent/dev.h>
 
 /*
  * Debugging
@@ -317,6 +319,7 @@ typedef void (task_func_t)(void *);
 #define	TASKQ_PREPOPULATE	0x0001
 #define	TASKQ_CPR_SAFE		0x0002	/* Use CPR safe protocol */
 #define	TASKQ_DYNAMIC		0x0004	/* Use dynamic thread scheduling */
+#define	TASKQ_THREADS_CPU_PCT	0x0008	/* Use dynamic thread scheduling */
 
 #define	TQ_SLEEP	KM_SLEEP	/* Can block for memory */
 #define	TQ_NOSLEEP	KM_NOSLEEP	/* cannot block for memory; may fail */
@@ -542,6 +545,10 @@ typedef struct ksiddomain {
 ksiddomain_t *ksid_lookupdomain(const char *);
 void ksiddomain_rele(ksiddomain_t *);
 
+#define	DDI_SLEEP	KM_SLEEP
+#define	ddi_log_sysevent(_a, _b, _c, _d, _e, _f, _g) \
+	sysevent_post_event(_c, _d, _b, "libzpool", _e, _f)
+
 #ifdef	__cplusplus
 }
 #endif

lib/libzpool/taskq.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -174,6 +174,19 @@ taskq_create(const char *name, int nthreads, pri_t pri,
 	taskq_t *tq = kmem_zalloc(sizeof (taskq_t), KM_SLEEP);
 	int t;
 
+	if (flags & TASKQ_THREADS_CPU_PCT) {
+		int pct;
+		ASSERT3S(nthreads, >=, 0);
+		ASSERT3S(nthreads, <=, 100);
+		pct = MIN(nthreads, 100);
+		pct = MAX(pct, 0);
+
+		nthreads = (sysconf(_SC_NPROCESSORS_ONLN) * pct) / 100;
+		nthreads = MAX(nthreads, 1);	/* need at least 1 thread */
+	} else {
+		ASSERT3S(nthreads, >=, 1);
+	}
+
 	rw_init(&tq->tq_threadlock, NULL, RW_DEFAULT, NULL);
 	mutex_init(&tq->tq_lock, NULL, MUTEX_DEFAULT, NULL);
 	cv_init(&tq->tq_dispatch_cv, NULL, CV_DEFAULT, NULL);

module/zcommon/include/sys/fm/fs/zfs.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_FM_FS_ZFS_H
 #define	_SYS_FM_FS_ZFS_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #ifdef	__cplusplus
 extern "C" {
 #endif
@@ -57,6 +55,7 @@ extern "C" {
 #define	FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE	"vdev_type"
 #define	FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH	"vdev_path"
 #define	FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID	"vdev_devid"
+#define	FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU		"vdev_fru"
 #define	FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID	"parent_guid"
 #define	FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE	"parent_type"
 #define	FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH	"parent_path"

module/zcommon/include/sys/fs/zfs.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -48,6 +48,10 @@ typedef enum {
 #define	ZFS_TYPE_DATASET	\
 	(ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME | ZFS_TYPE_SNAPSHOT)
 
+#define	ZAP_MAXNAMELEN 256
+#define	ZAP_MAXVALUELEN (1024 * 8)
+#define	ZAP_OLDMAXVALUELEN 1024
+
 /*
  * Dataset properties are identified by these constants and must be added to
  * the end of this list to ensure that external consumers are not affected
@@ -105,9 +109,21 @@ typedef enum {
 	ZFS_PROP_USEDDS,
 	ZFS_PROP_USEDCHILD,
 	ZFS_PROP_USEDREFRESERV,
+	ZFS_PROP_USERACCOUNTING,	/* not exposed to the user */
+	ZFS_PROP_STMF_SHAREINFO,	/* not exposed to the user */
 	ZFS_NUM_PROPS
 } zfs_prop_t;
 
+typedef enum {
+	ZFS_PROP_USERUSED,
+	ZFS_PROP_USERQUOTA,
+	ZFS_PROP_GROUPUSED,
+	ZFS_PROP_GROUPQUOTA,
+	ZFS_NUM_USERQUOTA_PROPS
+} zfs_userquota_prop_t;
+
+extern const char *zfs_userquota_prop_prefixes[ZFS_NUM_USERQUOTA_PROPS];
+
 /*
  * Pool properties are identified by these constants and must be added to the
  * end of this list to ensure that external consumers are not affected
@@ -130,6 +146,7 @@ typedef enum {
 	ZPOOL_PROP_CACHEFILE,
 	ZPOOL_PROP_FAILUREMODE,
 	ZPOOL_PROP_LISTSNAPS,
+	ZPOOL_PROP_AUTOEXPAND,
 	ZPOOL_NUM_PROPS
 } zpool_prop_t;
 
@@ -169,6 +186,7 @@ boolean_t zfs_prop_setonce(zfs_prop_t);
 const char *zfs_prop_to_name(zfs_prop_t);
 zfs_prop_t zfs_name_to_prop(const char *);
 boolean_t zfs_prop_user(const char *);
+boolean_t zfs_prop_userquota(const char *name);
 int zfs_prop_index_to_string(zfs_prop_t, uint64_t, const char **);
 int zfs_prop_string_to_index(zfs_prop_t, const char *, uint64_t *);
 boolean_t zfs_prop_valid_for_type(int, zfs_type_t);
@@ -213,6 +231,9 @@ typedef enum {
 #define	ZFS_DELEG_PERM_GID	"gid"
 #define	ZFS_DELEG_PERM_GROUPS	"groups"
 
+#define	ZFS_SMB_ACL_SRC		"src"
+#define	ZFS_SMB_ACL_TARGET	"target"
+
 typedef enum {
 	ZFS_CANMOUNT_OFF = 0,
 	ZFS_CANMOUNT_ON = 1,
@@ -226,6 +247,13 @@ typedef enum zfs_share_op {
 	ZFS_UNSHARE_SMB = 3
 } zfs_share_op_t;
 
+typedef enum zfs_smb_acl_op {
+	ZFS_SMB_ACL_ADD,
+	ZFS_SMB_ACL_REMOVE,
+	ZFS_SMB_ACL_RENAME,
+	ZFS_SMB_ACL_PURGE
+} zfs_smb_acl_op_t;
+
 typedef enum zfs_cache_type {
 	ZFS_CACHE_NONE = 0,
 	ZFS_CACHE_METADATA = 1,
@@ -250,13 +278,16 @@ typedef enum zfs_cache_type {
 #define	SPA_VERSION_12			12ULL
 #define	SPA_VERSION_13			13ULL
 #define	SPA_VERSION_14			14ULL
+#define	SPA_VERSION_15			15ULL
+#define	SPA_VERSION_16			16ULL
 /*
  * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk
- * format change. Go to usr/src/grub/grub-0.95/stage2/{zfs-include/, fsys_zfs*},
+ * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*},
- * and do the appropriate changes.
+ * and do the appropriate changes.  Also bump the version number in
+ * usr/src/grub/capability.
  */
-#define	SPA_VERSION			SPA_VERSION_14
+#define	SPA_VERSION			SPA_VERSION_16
-#define	SPA_VERSION_STRING		"14"
+#define	SPA_VERSION_STRING		"16"
 
 /*
  * Symbolic names for the changes that caused a SPA_VERSION switch.
@@ -292,6 +323,8 @@ typedef enum zfs_cache_type {
 #define	SPA_VERSION_SNAP_PROPS		SPA_VERSION_12
 #define	SPA_VERSION_USED_BREAKDOWN	SPA_VERSION_13
 #define	SPA_VERSION_PASSTHROUGH_X	SPA_VERSION_14
+#define	SPA_VERSION_USERSPACE		SPA_VERSION_15
+#define	SPA_VERSION_STMF_PROP		SPA_VERSION_16
 
 /*
  * ZPL version - rev'd whenever an incompatible on-disk format change
@@ -299,19 +332,21 @@ typedef enum zfs_cache_type {
  * also update the version_table[] and help message in zfs_prop.c.
  *
  * When changing, be sure to teach GRUB how to read the new format!
- * See usr/src/grub/grub-0.95/stage2/{zfs-include/,fsys_zfs*}
+ * See usr/src/grub/grub-0.97/stage2/{zfs-include/,fsys_zfs*}
  */
 #define	ZPL_VERSION_1			1ULL
 #define	ZPL_VERSION_2			2ULL
 #define	ZPL_VERSION_3			3ULL
-#define	ZPL_VERSION			ZPL_VERSION_3
+#define	ZPL_VERSION_4			4ULL
-#define	ZPL_VERSION_STRING		"3"
+#define	ZPL_VERSION			ZPL_VERSION_4
+#define	ZPL_VERSION_STRING		"4"
 
 #define	ZPL_VERSION_INITIAL		ZPL_VERSION_1
 #define	ZPL_VERSION_DIRENT_TYPE		ZPL_VERSION_2
 #define	ZPL_VERSION_FUID		ZPL_VERSION_3
 #define	ZPL_VERSION_NORMALIZATION	ZPL_VERSION_3
 #define	ZPL_VERSION_SYSATTR		ZPL_VERSION_3
+#define	ZPL_VERSION_USERSPACE		ZPL_VERSION_4
 
 /*
  * The following are configuration names used in the nvlist describing a pool's
@@ -361,6 +396,7 @@ typedef enum zfs_cache_type {
 #define	ZPOOL_CONFIG_FAULTED		"faulted"
 #define	ZPOOL_CONFIG_DEGRADED		"degraded"
 #define	ZPOOL_CONFIG_REMOVED		"removed"
+#define	ZPOOL_CONFIG_FRU		"fru"
 
 #define	VDEV_TYPE_ROOT			"root"
 #define	VDEV_TYPE_MIRROR		"mirror"
@@ -503,7 +539,7 @@ typedef struct vdev_stat {
 /*
  * And here are the things we need with /dev, etc. in front of them.
  */
-#define	ZVOL_PSEUDO_DEV		"/devices/pseudo/zvol@0:"
+#define	ZVOL_PSEUDO_DEV		"/devices/pseudo/zfs@0:"
 #define	ZVOL_FULL_DEV_DIR	"/dev/" ZVOL_DEV_DIR "/"
 
 #define	ZVOL_PROP_NAME		"name"
@@ -531,6 +567,7 @@ typedef enum zfs_ioc {
 	ZFS_IOC_VDEV_ATTACH,
 	ZFS_IOC_VDEV_DETACH,
 	ZFS_IOC_VDEV_SETPATH,
+	ZFS_IOC_VDEV_SETFRU,
 	ZFS_IOC_OBJSET_STATS,
 	ZFS_IOC_OBJSET_ZPLPROPS,
 	ZFS_IOC_DATASET_LIST_NEXT,
@@ -560,7 +597,11 @@ typedef enum zfs_ioc {
 	ZFS_IOC_GET_FSACL,
 	ZFS_IOC_ISCSI_PERM_CHECK,
 	ZFS_IOC_SHARE,
-	ZFS_IOC_INHERIT_PROP
+	ZFS_IOC_INHERIT_PROP,
+	ZFS_IOC_SMB_ACL,
+	ZFS_IOC_USERSPACE_ONE,
+	ZFS_IOC_USERSPACE_MANY,
+	ZFS_IOC_USERSPACE_UPGRADE
 } zfs_ioc_t;
 
 /*
@@ -602,6 +643,7 @@ typedef enum {
 #define	ZFS_ONLINE_CHECKREMOVE	0x1
 #define	ZFS_ONLINE_UNSPARE	0x2
 #define	ZFS_ONLINE_FORCEFAULT	0x4
+#define	ZFS_ONLINE_EXPAND	0x8
 #define	ZFS_OFFLINE_TEMPORARY	0x1
 
 /*

module/zcommon/include/zfs_deleg.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_ZFS_DELEG_H
 #define	_ZFS_DELEG_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/fs/zfs.h>
 
 #ifdef	__cplusplus
@@ -59,6 +57,10 @@ typedef enum {
 	ZFS_DELEG_NOTE_USERPROP,
 	ZFS_DELEG_NOTE_MOUNT,
 	ZFS_DELEG_NOTE_SHARE,
+	ZFS_DELEG_NOTE_USERQUOTA,
+	ZFS_DELEG_NOTE_GROUPQUOTA,
+	ZFS_DELEG_NOTE_USERUSED,
+	ZFS_DELEG_NOTE_GROUPUSED,
 	ZFS_DELEG_NOTE_NONE
 } zfs_deleg_note_t;
 

module/zcommon/include/zfs_namecheck.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_ZFS_NAMECHECK_H
 #define	_ZFS_NAMECHECK_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #ifdef	__cplusplus
 extern "C" {
 #endif
@@ -50,7 +48,6 @@ typedef enum {
 int pool_namecheck(const char *, namecheck_err_t *, char *);
 int dataset_namecheck(const char *, namecheck_err_t *, char *);
 int mountpoint_namecheck(const char *, namecheck_err_t *);
-int dataset_name_hidden(const char *);
 int snapshot_namecheck(const char *, namecheck_err_t *, char *);
 int permset_namecheck(const char *, namecheck_err_t *, char *);
 

module/zcommon/zfs_deleg.c

@@ -19,13 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #if defined(_KERNEL)
 #include <sys/systm.h>
 #include <sys/sunddi.h>
@@ -66,6 +63,10 @@ zfs_deleg_perm_tab_t zfs_deleg_perm_tab[] = {
 	{ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE },
 	{ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_NONE },
 	{ZFS_DELEG_PERM_USERPROP, ZFS_DELEG_NOTE_USERPROP },
+	{ZFS_DELEG_PERM_USERQUOTA, ZFS_DELEG_NOTE_USERQUOTA },
+	{ZFS_DELEG_PERM_GROUPQUOTA, ZFS_DELEG_NOTE_GROUPQUOTA },
+	{ZFS_DELEG_PERM_USERUSED, ZFS_DELEG_NOTE_USERUSED },
+	{ZFS_DELEG_PERM_GROUPUSED, ZFS_DELEG_NOTE_GROUPUSED },
 	{NULL, ZFS_DELEG_NOTE_NONE }
 };
 

module/zcommon/zfs_namecheck.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * Common name validation routines for ZFS.  These routines are shared by the
  * userland code as well as the ioctl() layer to ensure that we don't
@@ -345,19 +343,3 @@ pool_namecheck(const char *pool, namecheck_err_t *why, char *what)
 
 	return (0);
 }
-
-/*
- * Check if the dataset name is private for internal usage.
- * '$' is reserved for internal dataset names. e.g. "$MOS"
- *
- * Return 1 if the given name is used internally.
- * Return 0 if it is not.
- */
-int
-dataset_name_hidden(const char *name)
-{
-	if (strchr(name, '$') != NULL)
-		return (1);
-
-	return (0);
-}

module/zcommon/zfs_prop.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -43,6 +43,14 @@
 
 static zprop_desc_t zfs_prop_table[ZFS_NUM_PROPS];
 
+/* Note this is indexed by zfs_userquota_prop_t, keep the order the same */
+const char *zfs_userquota_prop_prefixes[] = {
+	"userused@",
+	"userquota@",
+	"groupused@",
+	"groupquota@"
+};
+
 zprop_desc_t *
 zfs_prop_get_table(void)
 {
@@ -133,6 +141,7 @@ zfs_prop_init(void)
 		{ "1",		1 },
 		{ "2",		2 },
 		{ "3",		3 },
+		{ "4",		4 },
 		{ "current",	ZPL_VERSION },
 		{ NULL }
 	};
@@ -218,7 +227,7 @@ zfs_prop_init(void)
 	/* default index properties */
 	register_index(ZFS_PROP_VERSION, "version", 0, PROP_DEFAULT,
 	    ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
-	    "1 | 2 | 3 | current", "VERSION", version_table);
+	    "1 | 2 | 3 | 4 | current", "VERSION", version_table);
 	register_index(ZFS_PROP_CANMOUNT, "canmount", ZFS_CANMOUNT_ON,
 	    PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, "on | off | noauto",
 	    "CANMOUNT", canmount_table);
@@ -305,8 +314,13 @@ zfs_prop_init(void)
 	    PROP_READONLY, ZFS_TYPE_DATASET, "NAME");
 	register_hidden(ZFS_PROP_ISCSIOPTIONS, "iscsioptions", PROP_TYPE_STRING,
 	    PROP_INHERIT, ZFS_TYPE_VOLUME, "ISCSIOPTIONS");
+	register_hidden(ZFS_PROP_STMF_SHAREINFO, "stmf_sbd_lu",
+	    PROP_TYPE_STRING, PROP_INHERIT, ZFS_TYPE_VOLUME,
+	    "STMF_SBD_LU");
 	register_hidden(ZFS_PROP_GUID, "guid", PROP_TYPE_NUMBER, PROP_READONLY,
 	    ZFS_TYPE_DATASET, "GUID");
+	register_hidden(ZFS_PROP_USERACCOUNTING, "useraccounting",
+	    PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, NULL);
 
 	/* oddball properties */
 	register_impl(ZFS_PROP_CREATION, "creation", PROP_TYPE_NUMBER, 0, NULL,
@@ -330,7 +344,6 @@ zfs_name_to_prop(const char *propname)
 	return (zprop_name_to_prop(propname, ZFS_TYPE_DATASET));
 }
 
-
 /*
  * For user property names, we allow all lowercase alphanumeric characters, plus
  * a few useful punctuation characters.
@@ -367,6 +380,26 @@ zfs_prop_user(const char *name)
 	return (B_TRUE);
 }
 
+/*
+ * Returns true if this is a valid userspace-type property (one with a '@').
+ * Note that after the @, any character is valid (eg, another @, for SID
+ * user@domain).
+ */
+boolean_t
+zfs_prop_userquota(const char *name)
+{
+	zfs_userquota_prop_t prop;
+
+	for (prop = 0; prop < ZFS_NUM_USERQUOTA_PROPS; prop++) {
+		if (strncmp(name, zfs_userquota_prop_prefixes[prop],
+		    strlen(zfs_userquota_prop_prefixes[prop])) == 0) {
+			return (B_TRUE);
+		}
+	}
+
+	return (B_FALSE);
+}
+
 /*
  * Tables of index types, plus functions to convert between the user view
  * (strings) and internal representation (uint64_t).

module/zcommon/zpool_prop.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -96,6 +96,8 @@ zpool_prop_init(void)
 	    ZFS_TYPE_POOL, "on | off", "REPLACE", boolean_table);
 	register_index(ZPOOL_PROP_LISTSNAPS, "listsnapshots", 0, PROP_DEFAULT,
 	    ZFS_TYPE_POOL, "on | off", "LISTSNAPS", boolean_table);
+	register_index(ZPOOL_PROP_AUTOEXPAND, "autoexpand", 0, PROP_DEFAULT,
+	    ZFS_TYPE_POOL, "on | off", "EXPAND", boolean_table);
 
 	/* default index properties */
 	register_index(ZPOOL_PROP_FAILUREMODE, "failmode",

module/zcommon/zprop_common.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * Common routines used by zfs and zpool property management.
  */
@@ -205,9 +203,6 @@ propname_match(const char *p, size_t len, zprop_desc_t *prop_entry)
 #ifndef _KERNEL
 	const char *colname = prop_entry->pd_colname;
 	int c;
-
-	if (colname == NULL)
-		return (B_FALSE);
 #endif
 
 	if (len == strlen(propname) &&
@@ -215,7 +210,7 @@ propname_match(const char *p, size_t len, zprop_desc_t *prop_entry)
 		return (B_TRUE);
 
 #ifndef _KERNEL
-	if (len != strlen(colname))
+	if (colname == NULL || len != strlen(colname))
 		return (B_FALSE);
 
 	for (c = 0; c < len; c++)

module/zfs/arc.c

@@ -124,6 +124,7 @@
 #include <sys/arc.h>
 #include <sys/refcount.h>
 #include <sys/vdev.h>
+#include <sys/vdev_impl.h>
 #ifdef _KERNEL
 #include <sys/vmsystm.h>
 #include <vm/anon.h>
@@ -397,6 +398,7 @@ static arc_state_t	*arc_l2c_only;
 
 static int		arc_no_grow;	/* Don't try to grow cache size */
 static uint64_t		arc_tempreserve;
+static uint64_t		arc_loaned_bytes;
 static uint64_t		arc_meta_used;
 static uint64_t		arc_meta_limit;
 static uint64_t		arc_meta_max = 0;
@@ -610,7 +612,7 @@ typedef struct l2arc_write_callback {
 struct l2arc_buf_hdr {
 	/* protected by arc_buf_hdr  mutex */
 	l2arc_dev_t	*b_dev;			/* L2ARC device */
-	daddr_t		b_daddr;		/* disk address, offset byte */
+	uint64_t	b_daddr;		/* disk address, offset byte */
 };
 
 typedef struct l2arc_data_free {
@@ -1203,6 +1205,41 @@ arc_buf_alloc(spa_t *spa, int size, void *tag, arc_buf_contents_t type)
 	return (buf);
 }
 
+static char *arc_onloan_tag = "onloan";
+
+/*
+ * Loan out an anonymous arc buffer. Loaned buffers are not counted as in
+ * flight data by arc_tempreserve_space() until they are "returned". Loaned
+ * buffers must be returned to the arc before they can be used by the DMU or
+ * freed.
+ */
+arc_buf_t *
+arc_loan_buf(spa_t *spa, int size)
+{
+	arc_buf_t *buf;
+
+	buf = arc_buf_alloc(spa, size, arc_onloan_tag, ARC_BUFC_DATA);
+
+	atomic_add_64(&arc_loaned_bytes, size);
+	return (buf);
+}
+
+/*
+ * Return a loaned arc buffer to the arc.
+ */
+void
+arc_return_buf(arc_buf_t *buf, void *tag)
+{
+	arc_buf_hdr_t *hdr = buf->b_hdr;
+
+	ASSERT(hdr->b_state == arc_anon);
+	ASSERT(buf->b_data != NULL);
+	VERIFY(refcount_remove(&hdr->b_refcnt, arc_onloan_tag) == 0);
+	VERIFY(refcount_add(&hdr->b_refcnt, tag) == 1);
+
+	atomic_add_64(&arc_loaned_bytes, -hdr->b_size);
+}
+
 static arc_buf_t *
 arc_buf_clone(arc_buf_t *from)
 {
@@ -2504,7 +2541,6 @@ arc_read(zio_t *pio, spa_t *spa, blkptr_t *bp, arc_buf_t *pbuf,
     uint32_t *arc_flags, const zbookmark_t *zb)
 {
 	int err;
-	arc_buf_hdr_t *hdr = pbuf->b_hdr;
 
 	ASSERT(!refcount_is_zero(&pbuf->b_hdr->b_refcnt));
 	ASSERT3U((char *)bp - (char *)pbuf->b_data, <, pbuf->b_hdr->b_size);
@@ -2512,9 +2548,8 @@ arc_read(zio_t *pio, spa_t *spa, blkptr_t *bp, arc_buf_t *pbuf,
 
 	err = arc_read_nolock(pio, spa, bp, done, private, priority,
 	    zio_flags, arc_flags, zb);
-
-	ASSERT3P(hdr, ==, pbuf->b_hdr);
 	rw_exit(&pbuf->b_lock);
+
 	return (err);
 }
 
@@ -2604,7 +2639,7 @@ top:
 		uint64_t size = BP_GET_LSIZE(bp);
 		arc_callback_t	*acb;
 		vdev_t *vd = NULL;
-		daddr_t addr;
+		uint64_t addr;
 		boolean_t devw = B_FALSE;
 
 		if (hdr == NULL) {
@@ -2923,6 +2958,7 @@ arc_release(arc_buf_t *buf, void *tag)
 	kmutex_t *hash_lock;
 	l2arc_buf_hdr_t *l2hdr;
 	uint64_t buf_size;
+	boolean_t released = B_FALSE;
 
 	rw_enter(&buf->b_lock, RW_WRITER);
 	hdr = buf->b_hdr;
@@ -2938,12 +2974,12 @@ arc_release(arc_buf_t *buf, void *tag)
 		ASSERT(buf->b_efunc == NULL);
 		arc_buf_thaw(buf);
 		rw_exit(&buf->b_lock);
-		return;
+		released = B_TRUE;
+	} else {
+		hash_lock = HDR_LOCK(hdr);
+		mutex_enter(hash_lock);
 	}
 
-	hash_lock = HDR_LOCK(hdr);
-	mutex_enter(hash_lock);
-
 	l2hdr = hdr->b_l2hdr;
 	if (l2hdr) {
 		mutex_enter(&l2arc_buflist_mtx);
@@ -2951,6 +2987,9 @@ arc_release(arc_buf_t *buf, void *tag)
 		buf_size = hdr->b_size;
 	}
 
+	if (released)
+		goto out;
+
 	/*
 	 * Do we have more than one buf?
 	 */
@@ -3018,6 +3057,7 @@ arc_release(arc_buf_t *buf, void *tag)
 	buf->b_efunc = NULL;
 	buf->b_private = NULL;
 
+out:
 	if (l2hdr) {
 		list_remove(l2hdr->b_dev->l2ad_buflist, hdr);
 		kmem_free(l2hdr, sizeof (l2arc_buf_hdr_t));
@@ -3311,10 +3351,9 @@ arc_free(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
 }
 
 static int
-arc_memory_throttle(uint64_t reserve, uint64_t txg)
+arc_memory_throttle(uint64_t reserve, uint64_t inflight_data, uint64_t txg)
 {
 #ifdef _KERNEL
-	uint64_t inflight_data = arc_anon->arcs_size;
 	uint64_t available_memory = ptob(freemem);
 	static uint64_t page_load = 0;
 	static uint64_t last_txg = 0;
@@ -3376,6 +3415,7 @@ int
 arc_tempreserve_space(uint64_t reserve, uint64_t txg)
 {
 	int error;
+	uint64_t anon_size;
 
 #ifdef ZFS_DEBUG
 	/*
@@ -3391,12 +3431,19 @@ arc_tempreserve_space(uint64_t reserve, uint64_t txg)
 	if (reserve > arc_c)
 		return (ENOMEM);
 
+	/*
+	 * Don't count loaned bufs as in flight dirty data to prevent long
+	 * network delays from blocking transactions that are ready to be
+	 * assigned to a txg.
+	 */
+	anon_size = MAX((int64_t)(arc_anon->arcs_size - arc_loaned_bytes), 0);
+
 	/*
 	 * Writes will, almost always, require additional memory allocations
 	 * in order to compress/encrypt/etc the data.  We therefor need to
 	 * make sure that there is sufficient available memory for this.
 	 */
-	if (error = arc_memory_throttle(reserve, txg))
+	if (error = arc_memory_throttle(reserve, anon_size, txg))
 		return (error);
 
 	/*
@@ -3406,8 +3453,9 @@ arc_tempreserve_space(uint64_t reserve, uint64_t txg)
 	 * Note: if two requests come in concurrently, we might let them
 	 * both succeed, when one of them should fail.  Not a huge deal.
 	 */
-	if (reserve + arc_tempreserve + arc_anon->arcs_size > arc_c / 2 &&
+
-	    arc_anon->arcs_size > arc_c / 4) {
+	if (reserve + arc_tempreserve + anon_size > arc_c / 2 &&
+	    anon_size > arc_c / 4) {
 		dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
 		    "anon_data=%lluK tempreserve=%lluK arc_c=%lluK\n",
 		    arc_tempreserve>>10,
@@ -3592,6 +3640,8 @@ arc_fini(void)
 	mutex_destroy(&zfs_write_limit_lock);
 
 	buf_fini();
+
+	ASSERT(arc_loaned_bytes == 0);
 }
 
 /*
@@ -4486,7 +4536,7 @@ l2arc_vdev_present(vdev_t *vd)
  * validated the vdev and opened it.
  */
 void
-l2arc_add_vdev(spa_t *spa, vdev_t *vd, uint64_t start, uint64_t end)
+l2arc_add_vdev(spa_t *spa, vdev_t *vd)
 {
 	l2arc_dev_t *adddev;
 
@@ -4500,8 +4550,8 @@ l2arc_add_vdev(spa_t *spa, vdev_t *vd, uint64_t start, uint64_t end)
 	adddev->l2ad_vdev = vd;
 	adddev->l2ad_write = l2arc_write_max;
 	adddev->l2ad_boost = l2arc_write_boost;
-	adddev->l2ad_start = start;
+	adddev->l2ad_start = VDEV_LABEL_START_SIZE;
-	adddev->l2ad_end = end;
+	adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd);
 	adddev->l2ad_hand = adddev->l2ad_start;
 	adddev->l2ad_evict = adddev->l2ad_start;
 	adddev->l2ad_first = B_TRUE;

module/zfs/dbuf.c

@@ -329,7 +329,7 @@ dbuf_verify(dmu_buf_impl_t *db)
 		if (db->db_parent == dn->dn_dbuf) {
 			/* db is pointed to by the dnode */
 			/* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */
-			if (db->db.db_object == DMU_META_DNODE_OBJECT)
+			if (DMU_OBJECT_IS_SPECIAL(db->db.db_object))
 				ASSERT(db->db_parent == NULL);
 			else
 				ASSERT(db->db_parent != NULL);
@@ -465,15 +465,15 @@ dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t *flags)
 	ASSERT(db->db_buf == NULL);
 
 	if (db->db_blkid == DB_BONUS_BLKID) {
-		int bonuslen = dn->dn_bonuslen;
+		int bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen);
 
 		ASSERT3U(bonuslen, <=, db->db.db_size);
 		db->db.db_data = zio_buf_alloc(DN_MAX_BONUSLEN);
 		arc_space_consume(DN_MAX_BONUSLEN, ARC_SPACE_OTHER);
 		if (bonuslen < DN_MAX_BONUSLEN)
 			bzero(db->db.db_data, DN_MAX_BONUSLEN);
-		bcopy(DN_BONUS(dn->dn_phys), db->db.db_data,
+		if (bonuslen)
-		    bonuslen);
+			bcopy(DN_BONUS(dn->dn_phys), db->db.db_data, bonuslen);
 		dbuf_update_data(db);
 		db->db_state = DB_CACHED;
 		mutex_exit(&db->db_mtx);
@@ -908,15 +908,11 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 	 * Shouldn't dirty a regular buffer in syncing context.  Private
 	 * objects may be dirtied in syncing context, but only if they
 	 * were already pre-dirtied in open context.
-	 * XXX We may want to prohibit dirtying in syncing context even
-	 * if they did pre-dirty.
 	 */
 	ASSERT(!dmu_tx_is_syncing(tx) ||
 	    BP_IS_HOLE(dn->dn_objset->os_rootbp) ||
-	    dn->dn_object == DMU_META_DNODE_OBJECT ||
+	    DMU_OBJECT_IS_SPECIAL(dn->dn_object) ||
-	    dn->dn_objset->os_dsl_dataset == NULL ||
+	    dn->dn_objset->os_dsl_dataset == NULL);
-	    dsl_dir_is_private(dn->dn_objset->os_dsl_dataset->ds_dir));
-
 	/*
 	 * We make this assert for private objects as well, but after we
 	 * check if we're already dirty.  They are allowed to re-dirty
@@ -975,7 +971,8 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 	/*
 	 * Only valid if not already dirty.
 	 */
-	ASSERT(dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
+	ASSERT(dn->dn_object == 0 ||
+	    dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
 	    (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));
 
 	ASSERT3U(dn->dn_nlevels, >, db->db_level);
@@ -987,15 +984,13 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 
 	/*
 	 * We should only be dirtying in syncing context if it's the
-	 * mos, a spa os, or we're initializing the os.  However, we are
+	 * mos or we're initializing the os or it's a special object.
-	 * allowed to dirty in syncing context provided we already
+	 * However, we are allowed to dirty in syncing context provided
-	 * dirtied it in open context.  Hence we must make this
+	 * we already dirtied it in open context.  Hence we must make
-	 * assertion only if we're not already dirty.
+	 * this assertion only if we're not already dirty.
 	 */
-	ASSERT(!dmu_tx_is_syncing(tx) ||
+	ASSERT(!dmu_tx_is_syncing(tx) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) ||
-	    os->os_dsl_dataset == NULL ||
+	    os->os_dsl_dataset == NULL || BP_IS_HOLE(os->os_rootbp));
-	    !dsl_dir_is_private(os->os_dsl_dataset->ds_dir) ||
-	    !BP_IS_HOLE(os->os_rootbp));
 	ASSERT(db->db.db_size != 0);
 
 	dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);
@@ -1311,6 +1306,68 @@ dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx)
 	mutex_exit(&db->db_mtx);
 }
 
+/*
+ * Directly assign a provided arc buf to a given dbuf if it's not referenced
+ * by anybody except our caller. Otherwise copy arcbuf's contents to dbuf.
+ */
+void
+dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx)
+{
+	ASSERT(!refcount_is_zero(&db->db_holds));
+	ASSERT(db->db_dnode->dn_object != DMU_META_DNODE_OBJECT);
+	ASSERT(db->db_blkid != DB_BONUS_BLKID);
+	ASSERT(db->db_level == 0);
+	ASSERT(DBUF_GET_BUFC_TYPE(db) == ARC_BUFC_DATA);
+	ASSERT(buf != NULL);
+	ASSERT(arc_buf_size(buf) == db->db.db_size);
+	ASSERT(tx->tx_txg != 0);
+
+	arc_return_buf(buf, db);
+	ASSERT(arc_released(buf));
+
+	mutex_enter(&db->db_mtx);
+
+	while (db->db_state == DB_READ || db->db_state == DB_FILL)
+		cv_wait(&db->db_changed, &db->db_mtx);
+
+	ASSERT(db->db_state == DB_CACHED || db->db_state == DB_UNCACHED);
+
+	if (db->db_state == DB_CACHED &&
+	    refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) {
+		mutex_exit(&db->db_mtx);
+		(void) dbuf_dirty(db, tx);
+		bcopy(buf->b_data, db->db.db_data, db->db.db_size);
+		VERIFY(arc_buf_remove_ref(buf, db) == 1);
+		return;
+	}
+
+	if (db->db_state == DB_CACHED) {
+		dbuf_dirty_record_t *dr = db->db_last_dirty;
+
+		ASSERT(db->db_buf != NULL);
+		if (dr != NULL && dr->dr_txg == tx->tx_txg) {
+			ASSERT(dr->dt.dl.dr_data == db->db_buf);
+			if (!arc_released(db->db_buf)) {
+				ASSERT(dr->dt.dl.dr_override_state ==
+				    DR_OVERRIDDEN);
+				arc_release(db->db_buf, db);
+			}
+			dr->dt.dl.dr_data = buf;
+			VERIFY(arc_buf_remove_ref(db->db_buf, db) == 1);
+		} else if (dr == NULL || dr->dt.dl.dr_data != db->db_buf) {
+			arc_release(db->db_buf, db);
+			VERIFY(arc_buf_remove_ref(db->db_buf, db) == 1);
+		}
+		db->db_buf = NULL;
+	}
+	ASSERT(db->db_buf == NULL);
+	dbuf_set_data(db, buf);
+	db->db_state = DB_FILL;
+	mutex_exit(&db->db_mtx);
+	(void) dbuf_dirty(db, tx);
+	dbuf_fill_done(db, tx);
+}
+
 /*
  * "Clear" the contents of this dbuf.  This will mark the dbuf
  * EVICTING and clear *most* of its references.  Unfortunetely,
@@ -1855,6 +1912,19 @@ dmu_buf_get_user(dmu_buf_t *db_fake)
 	return (db->db_user_ptr);
 }
 
+boolean_t
+dmu_buf_freeable(dmu_buf_t *dbuf)
+{
+	boolean_t res = B_FALSE;
+	dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf;
+
+	if (db->db_blkptr)
+		res = dsl_dataset_block_freeable(db->db_objset->os_dsl_dataset,
+		    db->db_blkptr->blk_birth);
+
+	return (res);
+}
+
 static void
 dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db)
 {
@@ -1943,7 +2013,6 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
 	dnode_t *dn = db->db_dnode;
 	objset_impl_t *os = dn->dn_objset;
 	uint64_t txg = tx->tx_txg;
-	int blksz;
 
 	ASSERT(dmu_tx_is_syncing(tx));
 
@@ -2049,32 +2118,25 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
 		return;
 	}
 
-	if (db->db_state != DB_NOFILL) {
+	if (db->db_state != DB_NOFILL &&
-		blksz = arc_buf_size(*datap);
+	    dn->dn_object != DMU_META_DNODE_OBJECT &&
-
+	    refcount_count(&db->db_holds) > 1 &&
-		if (dn->dn_object != DMU_META_DNODE_OBJECT) {
+	    *datap == db->db_buf) {
-			/*
+		/*
-			 * If this buffer is currently "in use" (i.e., there
+		 * If this buffer is currently "in use" (i.e., there
-			 * are active holds and db_data still references it),
+		 * are active holds and db_data still references it),
-			 * then make a copy before we start the write so that
+		 * then make a copy before we start the write so that
-			 * any modifications from the open txg will not leak
+		 * any modifications from the open txg will not leak
-			 * into this write.
+		 * into this write.
-			 *
+		 *
-			 * NOTE: this copy does not need to be made for
+		 * NOTE: this copy does not need to be made for
-			 * objects only modified in the syncing context (e.g.
+		 * objects only modified in the syncing context (e.g.
-			 * DNONE_DNODE blocks).
+		 * DNONE_DNODE blocks).
-			 */
+		 */
-			if (refcount_count(&db->db_holds) > 1 &&
+		int blksz = arc_buf_size(*datap);
-			    *datap == db->db_buf) {
+		arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
-				arc_buf_contents_t type =
+		*datap = arc_buf_alloc(os->os_spa, blksz, db, type);
-				    DBUF_GET_BUFC_TYPE(db);
+		bcopy(db->db.db_data, (*datap)->b_data, blksz);
-				*datap =
-				    arc_buf_alloc(os->os_spa, blksz, db, type);
-				bcopy(db->db.db_data, (*datap)->b_data, blksz);
-			}
-		}
-
-		ASSERT(*datap != NULL);
 	}
 	db->db_data_pending = dr;
 

module/zfs/dmu.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -85,6 +85,8 @@ const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
 	{	byteswap_uint64_array,	TRUE,	"FUID table size"	},
 	{	zap_byteswap,		TRUE,	"DSL dataset next clones"},
 	{	zap_byteswap,		TRUE,	"scrub work queue"	},
+	{	zap_byteswap,		TRUE,	"ZFS user/group used"	},
+	{	zap_byteswap,		TRUE,	"ZFS user/group quota"	},
 };
 
 int
@@ -180,22 +182,22 @@ dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
  * whose dnodes are in the same block.
  */
 static int
-dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
+dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
-    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
+    int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags)
 {
 	dsl_pool_t *dp = NULL;
 	dmu_buf_t **dbp;
 	uint64_t blkid, nblks, i;
-	uint32_t flags;
+	uint32_t dbuf_flags;
 	int err;
 	zio_t *zio;
 	hrtime_t start;
 
 	ASSERT(length <= DMU_MAX_ACCESS);
 
-	flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT;
+	dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT;
-	if (length > zfetch_array_rd_sz)
+	if (flags & DMU_READ_NO_PREFETCH || length > zfetch_array_rd_sz)
-		flags |= DB_RF_NOPREFETCH;
+		dbuf_flags |= DB_RF_NOPREFETCH;
 
 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
 	if (dn->dn_datablkshift) {
@@ -233,7 +235,7 @@ dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
 		/* initiate async i/o */
 		if (read) {
 			rw_exit(&dn->dn_struct_rwlock);
-			(void) dbuf_read(db, zio, flags);
+			(void) dbuf_read(db, zio, dbuf_flags);
 			rw_enter(&dn->dn_struct_rwlock, RW_READER);
 		}
 		dbp[i] = &db->db;
@@ -285,7 +287,7 @@ dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
 		return (err);
 
 	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
-	    numbufsp, dbpp);
+	    numbufsp, dbpp, DMU_READ_PREFETCH);
 
 	dnode_rele(dn, FTAG);
 
@@ -300,7 +302,7 @@ dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
 	int err;
 
 	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
-	    numbufsp, dbpp);
+	    numbufsp, dbpp, DMU_READ_PREFETCH);
 
 	return (err);
 }
@@ -442,7 +444,8 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
 	object_size = align == 1 ? dn->dn_datablksz :
 	    (dn->dn_maxblkid + 1) << dn->dn_datablkshift;
 
-	if (trunc || (end = offset + length) > object_size)
+	end = offset + length;
+	if (trunc || end > object_size)
 		end = object_size;
 	if (end <= offset)
 		return (0);
@@ -450,6 +453,7 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
 
 	while (length) {
 		start = end;
+		/* assert(offset <= start) */
 		err = get_next_chunk(dn, &start, offset);
 		if (err)
 			return (err);
@@ -540,7 +544,7 @@ dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
 
 int
 dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
-    void *buf)
+    void *buf, uint32_t flags)
 {
 	dnode_t *dn;
 	dmu_buf_t **dbp;
@@ -570,7 +574,7 @@ dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
 		 * to be reading in parallel.
 		 */
 		err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
-		    TRUE, FTAG, &numbufs, &dbp);
+		    TRUE, FTAG, &numbufs, &dbp, flags);
 		if (err)
 			break;
 
@@ -810,6 +814,58 @@ dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
 }
 #endif
 
+/*
+ * Allocate a loaned anonymous arc buffer.
+ */
+arc_buf_t *
+dmu_request_arcbuf(dmu_buf_t *handle, int size)
+{
+	dnode_t *dn = ((dmu_buf_impl_t *)handle)->db_dnode;
+
+	return (arc_loan_buf(dn->dn_objset->os_spa, size));
+}
+
+/*
+ * Free a loaned arc buffer.
+ */
+void
+dmu_return_arcbuf(arc_buf_t *buf)
+{
+	arc_return_buf(buf, FTAG);
+	VERIFY(arc_buf_remove_ref(buf, FTAG) == 1);
+}
+
+/*
+ * When possible directly assign passed loaned arc buffer to a dbuf.
+ * If this is not possible copy the contents of passed arc buf via
+ * dmu_write().
+ */
+void
+dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf,
+    dmu_tx_t *tx)
+{
+	dnode_t *dn = ((dmu_buf_impl_t *)handle)->db_dnode;
+	dmu_buf_impl_t *db;
+	uint32_t blksz = (uint32_t)arc_buf_size(buf);
+	uint64_t blkid;
+
+	rw_enter(&dn->dn_struct_rwlock, RW_READER);
+	blkid = dbuf_whichblock(dn, offset);
+	VERIFY((db = dbuf_hold(dn, blkid, FTAG)) != NULL);
+	rw_exit(&dn->dn_struct_rwlock);
+
+	if (offset == db->db.db_offset && blksz == db->db.db_size) {
+		dbuf_assign_arcbuf(db, buf, tx);
+		dbuf_rele(db, FTAG);
+	} else {
+		dbuf_rele(db, FTAG);
+		ASSERT(dn->dn_objset->os.os == dn->dn_objset);
+		dmu_write(&dn->dn_objset->os, dn->dn_object, offset, blksz,
+		    buf->b_data, tx);
+		dmu_return_arcbuf(buf);
+	}
+}
+
 typedef struct {
 	dbuf_dirty_record_t	*dr;
 	dmu_sync_cb_t		*done;

module/zfs/dmu_object.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/dmu.h>
 #include <sys/dmu_objset.h>
 #include <sys/dmu_tx.h>
@@ -108,22 +106,56 @@ dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
 
 int
 dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
-    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
+    int blocksize, dmu_object_type_t bonustype, int bonuslen)
 {
 	dnode_t *dn;
+	dmu_tx_t *tx;
+	int nblkptr;
 	int err;
 
-	if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
+	if (object == DMU_META_DNODE_OBJECT)
 		return (EBADF);
 
 	err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED,
 	    FTAG, &dn);
 	if (err)
 		return (err);
+
+	if (dn->dn_type == ot && dn->dn_datablksz == blocksize &&
+	    dn->dn_bonustype == bonustype && dn->dn_bonuslen == bonuslen) {
+		/* nothing is changing, this is a noop */
+		dnode_rele(dn, FTAG);
+		return (0);
+	}
+
+	nblkptr = 1 + ((DN_MAX_BONUSLEN - bonuslen) >> SPA_BLKPTRSHIFT);
+
+	/*
+	 * If we are losing blkptrs or changing the block size this must
+	 * be a new file instance.   We must clear out the previous file
+	 * contents before we can change this type of metadata in the dnode.
+	 */
+	if (dn->dn_nblkptr > nblkptr || dn->dn_datablksz != blocksize) {
+		err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END);
+		if (err)
+			goto out;
+	}
+
+	tx = dmu_tx_create(os);
+	dmu_tx_hold_bonus(tx, object);
+	err = dmu_tx_assign(tx, TXG_WAIT);
+	if (err) {
+		dmu_tx_abort(tx);
+		goto out;
+	}
+
 	dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, tx);
+
+	dmu_tx_commit(tx);
+out:
 	dnode_rele(dn, FTAG);
 
-	return (0);
+	return (err);
 }
 
 int

module/zfs/dmu_objset.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -164,10 +164,15 @@ dmu_objset_byteswap(void *buf, size_t size)
 {
 	objset_phys_t *osp = buf;
 
-	ASSERT(size == sizeof (objset_phys_t));
+	ASSERT(size == OBJSET_OLD_PHYS_SIZE || size == sizeof (objset_phys_t));
 	dnode_byteswap(&osp->os_meta_dnode);
 	byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t));
 	osp->os_type = BSWAP_64(osp->os_type);
+	osp->os_flags = BSWAP_64(osp->os_flags);
+	if (size == sizeof (objset_phys_t)) {
+		dnode_byteswap(&osp->os_userused_dnode);
+		dnode_byteswap(&osp->os_groupused_dnode);
+	}
 }
 
 int
@@ -210,12 +215,30 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
 				err = EIO;
 			return (err);
 		}
+
+		/* Increase the blocksize if we are permitted. */
+		if (spa_version(spa) >= SPA_VERSION_USERSPACE &&
+		    arc_buf_size(osi->os_phys_buf) < sizeof (objset_phys_t)) {
+			arc_buf_t *buf = arc_buf_alloc(spa,
+			    sizeof (objset_phys_t), &osi->os_phys_buf,
+			    ARC_BUFC_METADATA);
+			bzero(buf->b_data, sizeof (objset_phys_t));
+			bcopy(osi->os_phys_buf->b_data, buf->b_data,
+			    arc_buf_size(osi->os_phys_buf));
+			(void) arc_buf_remove_ref(osi->os_phys_buf,
+			    &osi->os_phys_buf);
+			osi->os_phys_buf = buf;
+		}
+
 		osi->os_phys = osi->os_phys_buf->b_data;
+		osi->os_flags = osi->os_phys->os_flags;
 	} else {
-		osi->os_phys_buf = arc_buf_alloc(spa, sizeof (objset_phys_t),
+		int size = spa_version(spa) >= SPA_VERSION_USERSPACE ?
+		    sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE;
+		osi->os_phys_buf = arc_buf_alloc(spa, size,
 		    &osi->os_phys_buf, ARC_BUFC_METADATA);
 		osi->os_phys = osi->os_phys_buf->b_data;
-		bzero(osi->os_phys, sizeof (objset_phys_t));
+		bzero(osi->os_phys, size);
 	}
 
 	/*
@@ -276,6 +299,12 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
 
 	osi->os_meta_dnode = dnode_special_open(osi,
 	    &osi->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT);
+	if (arc_buf_size(osi->os_phys_buf) >= sizeof (objset_phys_t)) {
+		osi->os_userused_dnode = dnode_special_open(osi,
+		    &osi->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT);
+		osi->os_groupused_dnode = dnode_special_open(osi,
+		    &osi->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT);
+	}
 
 	/*
 	 * We should be the only thread trying to do this because we
@@ -456,13 +485,15 @@ dmu_objset_evict(dsl_dataset_t *ds, void *arg)
 	os.os = osi;
 	(void) dmu_objset_evict_dbufs(&os);
 
-	ASSERT3P(list_head(&osi->os_dnodes), ==, osi->os_meta_dnode);
-	ASSERT3P(list_tail(&osi->os_dnodes), ==, osi->os_meta_dnode);
-	ASSERT3P(list_head(&osi->os_meta_dnode->dn_dbufs), ==, NULL);
-
 	dnode_special_close(osi->os_meta_dnode);
+	if (osi->os_userused_dnode) {
+		dnode_special_close(osi->os_userused_dnode);
+		dnode_special_close(osi->os_groupused_dnode);
+	}
 	zil_free(osi->os_zil);
 
+	ASSERT3P(list_head(&osi->os_dnodes), ==, NULL);
+
 	VERIFY(arc_buf_remove_ref(osi->os_phys_buf, &osi->os_phys_buf) == 1);
 	mutex_destroy(&osi->os_lock);
 	mutex_destroy(&osi->os_obj_lock);
@@ -520,6 +551,10 @@ dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
 	ASSERT(type != DMU_OST_ANY);
 	ASSERT(type < DMU_OST_NUMTYPES);
 	osi->os_phys->os_type = type;
+	if (dmu_objset_userused_enabled(osi)) {
+		osi->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
+		osi->os_flags = osi->os_phys->os_flags;
+	}
 
 	dsl_dataset_dirty(ds, tx);
 
@@ -704,13 +739,33 @@ struct snaparg {
 	char *snapname;
 	char failed[MAXPATHLEN];
 	boolean_t checkperms;
-	list_t objsets;
+	nvlist_t *props;
 };
 
-struct osnode {
+static int
-	list_node_t node;
+snapshot_check(void *arg1, void *arg2, dmu_tx_t *tx)
-	objset_t *os;
+{
-};
+	objset_t *os = arg1;
+	struct snaparg *sn = arg2;
+
+	/* The props have already been checked by zfs_check_userprops(). */
+
+	return (dsl_dataset_snapshot_check(os->os->os_dsl_dataset,
+	    sn->snapname, tx));
+}
+
+static void
+snapshot_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+	objset_t *os = arg1;
+	dsl_dataset_t *ds = os->os->os_dsl_dataset;
+	struct snaparg *sn = arg2;
+
+	dsl_dataset_snapshot_sync(ds, sn->snapname, cr, tx);
+
+	if (sn->props)
+		dsl_props_set_sync(ds->ds_prev, sn->props, cr, tx);
+}
 
 static int
 dmu_objset_snapshot_one(char *name, void *arg)
@@ -747,13 +802,8 @@ dmu_objset_snapshot_one(char *name, void *arg)
 	 */
 	err = zil_suspend(dmu_objset_zil(os));
 	if (err == 0) {
-		struct osnode *osn;
+		dsl_sync_task_create(sn->dstg, snapshot_check,
-		dsl_sync_task_create(sn->dstg, dsl_dataset_snapshot_check,
+		    snapshot_sync, os, sn, 3);
-		    dsl_dataset_snapshot_sync, os->os->os_dsl_dataset,
-		    sn->snapname, 3);
-		osn = kmem_alloc(sizeof (struct osnode), KM_SLEEP);
-		osn->os = os;
-		list_insert_tail(&sn->objsets, osn);
 	} else {
 		dmu_objset_close(os);
 	}
@@ -762,11 +812,11 @@ dmu_objset_snapshot_one(char *name, void *arg)
 }
 
 int
-dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive)
+dmu_objset_snapshot(char *fsname, char *snapname,
+    nvlist_t *props, boolean_t recursive)
 {
 	dsl_sync_task_t *dst;
-	struct osnode *osn;
+	struct snaparg sn;
-	struct snaparg sn = { 0 };
 	spa_t *spa;
 	int err;
 
@@ -778,8 +828,7 @@ dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive)
 
 	sn.dstg = dsl_sync_task_group_create(spa_get_dsl(spa));
 	sn.snapname = snapname;
-	list_create(&sn.objsets, sizeof (struct osnode),
+	sn.props = props;
-	    offsetof(struct osnode, node));
 
 	if (recursive) {
 		sn.checkperms = B_TRUE;
@@ -790,27 +839,19 @@ dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive)
 		err = dmu_objset_snapshot_one(fsname, &sn);
 	}
 
-	if (err)
+	if (err == 0)
-		goto out;
+		err = dsl_sync_task_group_wait(sn.dstg);
-
-	err = dsl_sync_task_group_wait(sn.dstg);
 
 	for (dst = list_head(&sn.dstg->dstg_tasks); dst;
 	    dst = list_next(&sn.dstg->dstg_tasks, dst)) {
-		dsl_dataset_t *ds = dst->dst_arg1;
+		objset_t *os = dst->dst_arg1;
+		dsl_dataset_t *ds = os->os->os_dsl_dataset;
 		if (dst->dst_err)
 			dsl_dataset_name(ds, sn.failed);
+		zil_resume(dmu_objset_zil(os));
+		dmu_objset_close(os);
 	}
 
-out:
-	while (osn = list_head(&sn.objsets)) {
-		list_remove(&sn.objsets, osn);
-		zil_resume(dmu_objset_zil(osn->os));
-		dmu_objset_close(osn->os);
-		kmem_free(osn, sizeof (struct osnode));
-	}
-	list_destroy(&sn.objsets);
-
 	if (err)
 		(void) strcpy(fsname, sn.failed);
 	dsl_sync_task_group_destroy(sn.dstg);
@@ -819,7 +860,7 @@ out:
 }
 
 static void
-dmu_objset_sync_dnodes(list_t *list, dmu_tx_t *tx)
+dmu_objset_sync_dnodes(list_t *list, list_t *newlist, dmu_tx_t *tx)
 {
 	dnode_t *dn;
 
@@ -827,14 +868,20 @@ dmu_objset_sync_dnodes(list_t *list, dmu_tx_t *tx)
 		ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
 		ASSERT(dn->dn_dbuf->db_data_pending);
 		/*
-		 * Initialize dn_zio outside dnode_sync()
+		 * Initialize dn_zio outside dnode_sync() because the
-		 * to accomodate meta-dnode
+		 * meta-dnode needs to set it ouside dnode_sync().
 		 */
 		dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio;
 		ASSERT(dn->dn_zio);
 
 		ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS);
 		list_remove(list, dn);
+
+		if (newlist) {
+			(void) dnode_add_ref(dn, newlist);
+			list_insert_tail(newlist, dn);
+		}
+
 		dnode_sync(dn, tx);
 	}
 }
@@ -853,9 +900,12 @@ ready(zio_t *zio, arc_buf_t *abuf, void *arg)
 	ASSERT(BP_GET_LEVEL(bp) == 0);
 
 	/*
-	 * Update rootbp fill count.
+	 * Update rootbp fill count: it should be the number of objects
+	 * allocated in the object set (not counting the "special"
+	 * objects that are stored in the objset_phys_t -- the meta
+	 * dnode and user/group accounting objects).
 	 */
-	bp->blk_fill = 1;	/* count the meta-dnode */
+	bp->blk_fill = 0;
 	for (int i = 0; i < dnp->dn_nblkptr; i++)
 		bp->blk_fill += dnp->dn_blkptr[i].blk_fill;
 
@@ -878,6 +928,7 @@ dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
 	writeprops_t wp = { 0 };
 	zio_t *zio;
 	list_t *list;
+	list_t *newlist = NULL;
 	dbuf_dirty_record_t *dr;
 
 	dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
@@ -915,20 +966,41 @@ dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
 	}
 
 	arc_release(os->os_phys_buf, &os->os_phys_buf);
+
 	zio = arc_write(pio, os->os_spa, &wp, DMU_OS_IS_L2CACHEABLE(os),
 	    tx->tx_txg, os->os_rootbp, os->os_phys_buf, ready, NULL, os,
 	    ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
 
 	/*
-	 * Sync meta-dnode - the parent IO for the sync is the root block
+	 * Sync special dnodes - the parent IO for the sync is the root block
 	 */
 	os->os_meta_dnode->dn_zio = zio;
 	dnode_sync(os->os_meta_dnode, tx);
 
+	os->os_phys->os_flags = os->os_flags;
+
+	if (os->os_userused_dnode &&
+	    os->os_userused_dnode->dn_type != DMU_OT_NONE) {
+		os->os_userused_dnode->dn_zio = zio;
+		dnode_sync(os->os_userused_dnode, tx);
+		os->os_groupused_dnode->dn_zio = zio;
+		dnode_sync(os->os_groupused_dnode, tx);
+	}
+
 	txgoff = tx->tx_txg & TXG_MASK;
 
-	dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], tx);
+	if (dmu_objset_userused_enabled(os)) {
-	dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], tx);
+		newlist = &os->os_synced_dnodes;
+		/*
+		 * We must create the list here because it uses the
+		 * dn_dirty_link[] of this txg.
+		 */
+		list_create(newlist, sizeof (dnode_t),
+		    offsetof(dnode_t, dn_dirty_link[txgoff]));
+	}
+
+	dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx);
+	dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx);
 
 	list = &os->os_meta_dnode->dn_dirty_records[txgoff];
 	while (dr = list_head(list)) {
@@ -945,6 +1017,145 @@ dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
 	zio_nowait(zio);
 }
 
+static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES];
+
+void
+dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb)
+{
+	used_cbs[ost] = cb;
+}
+
+boolean_t
+dmu_objset_userused_enabled(objset_impl_t *os)
+{
+	return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE &&
+	    used_cbs[os->os_phys->os_type] &&
+	    os->os_userused_dnode);
+}
+
+void
+dmu_objset_do_userquota_callbacks(objset_impl_t *os, dmu_tx_t *tx)
+{
+	dnode_t *dn;
+	list_t *list = &os->os_synced_dnodes;
+	static const char zerobuf[DN_MAX_BONUSLEN] = {0};
+
+	ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os));
+
+	while (dn = list_head(list)) {
+		dmu_object_type_t bonustype;
+
+		ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object));
+		ASSERT(dn->dn_oldphys);
+		ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE ||
+		    dn->dn_phys->dn_flags &
+		    DNODE_FLAG_USERUSED_ACCOUNTED);
+
+		/* Allocate the user/groupused objects if necessary. */
+		if (os->os_userused_dnode->dn_type == DMU_OT_NONE) {
+			VERIFY(0 == zap_create_claim(&os->os,
+			    DMU_USERUSED_OBJECT,
+			    DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
+			VERIFY(0 == zap_create_claim(&os->os,
+			    DMU_GROUPUSED_OBJECT,
+			    DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
+		}
+
+		/*
+		 * If the object was not previously
+		 * accounted, pretend that it was free.
+		 */
+		if (!(dn->dn_oldphys->dn_flags &
+		    DNODE_FLAG_USERUSED_ACCOUNTED)) {
+			bzero(dn->dn_oldphys, sizeof (dnode_phys_t));
+		}
+
+		/*
+		 * If the object was freed, use the previous bonustype.
+		 */
+		bonustype = dn->dn_phys->dn_bonustype ?
+		    dn->dn_phys->dn_bonustype : dn->dn_oldphys->dn_bonustype;
+		ASSERT(dn->dn_phys->dn_type != 0 ||
+		    (bcmp(DN_BONUS(dn->dn_phys), zerobuf,
+		    DN_MAX_BONUSLEN) == 0 &&
+		    DN_USED_BYTES(dn->dn_phys) == 0));
+		ASSERT(dn->dn_oldphys->dn_type != 0 ||
+		    (bcmp(DN_BONUS(dn->dn_oldphys), zerobuf,
+		    DN_MAX_BONUSLEN) == 0 &&
+		    DN_USED_BYTES(dn->dn_oldphys) == 0));
+		used_cbs[os->os_phys->os_type](&os->os, bonustype,
+		    DN_BONUS(dn->dn_oldphys), DN_BONUS(dn->dn_phys),
+		    DN_USED_BYTES(dn->dn_oldphys),
+		    DN_USED_BYTES(dn->dn_phys), tx);
+
+		/*
+		 * The mutex is needed here for interlock with dnode_allocate.
+		 */
+		mutex_enter(&dn->dn_mtx);
+		zio_buf_free(dn->dn_oldphys, sizeof (dnode_phys_t));
+		dn->dn_oldphys = NULL;
+		mutex_exit(&dn->dn_mtx);
+
+		list_remove(list, dn);
+		dnode_rele(dn, list);
+	}
+}
+
+boolean_t
+dmu_objset_userspace_present(objset_t *os)
+{
+	return (os->os->os_phys->os_flags &
+	    OBJSET_FLAG_USERACCOUNTING_COMPLETE);
+}
+
+int
+dmu_objset_userspace_upgrade(objset_t *os)
+{
+	uint64_t obj;
+	int err = 0;
+
+	if (dmu_objset_userspace_present(os))
+		return (0);
+	if (!dmu_objset_userused_enabled(os->os))
+		return (ENOTSUP);
+	if (dmu_objset_is_snapshot(os))
+		return (EINVAL);
+
+	/*
+	 * We simply need to mark every object dirty, so that it will be
+	 * synced out and now accounted.  If this is called
+	 * concurrently, or if we already did some work before crashing,
+	 * that's fine, since we track each object's accounted state
+	 * independently.
+	 */
+
+	for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
+		dmu_tx_t *tx = dmu_tx_create(os);
+		dmu_buf_t *db;
+		int objerr;
+
+		if (issig(JUSTLOOKING) && issig(FORREAL))
+			return (EINTR);
+
+		objerr = dmu_bonus_hold(os, obj, FTAG, &db);
+		if (objerr)
+			continue;
+		dmu_tx_hold_bonus(tx, obj);
+		objerr = dmu_tx_assign(tx, TXG_WAIT);
+		if (objerr) {
+			dmu_tx_abort(tx);
+			continue;
+		}
+		dmu_buf_will_dirty(db, tx);
+		dmu_buf_rele(db, FTAG);
+		dmu_tx_commit(tx);
+	}
+
+	os->os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
+	txg_wait_synced(dmu_objset_pool(os), 0);
+	return (0);
+}
+
 void
 dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
     uint64_t *usedobjsp, uint64_t *availobjsp)
@@ -978,6 +1189,8 @@ dmu_objset_stats(objset_t *os, nvlist_t *nv)
 
 	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE,
 	    os->os->os_phys->os_type);
+	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING,
+	    dmu_objset_userspace_present(os));
 }
 
 int

module/zfs/dmu_send.c

@@ -161,7 +161,9 @@ backup_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
 	if (issig(JUSTLOOKING) && issig(FORREAL))
 		return (EINTR);
 
-	if (bp == NULL && zb->zb_object == 0) {
+	if (zb->zb_object != 0 && DMU_OBJECT_IS_SPECIAL(zb->zb_object)) {
+		return (0);
+	} else if (bp == NULL && zb->zb_object == 0) {
 		uint64_t span = BP_SPAN(dnp, zb->zb_level);
 		uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT;
 		err = dump_freeobjects(ba, dnobj, span >> DNODE_SHIFT);
@@ -775,11 +777,6 @@ restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
 	dmu_tx_t *tx;
 	void *data = NULL;
 
-	err = dmu_object_info(os, drro->drr_object, NULL);
-
-	if (err != 0 && err != ENOENT)
-		return (EINVAL);
-
 	if (drro->drr_type == DMU_OT_NONE ||
 	    drro->drr_type >= DMU_OT_NUMTYPES ||
 	    drro->drr_bonustype >= DMU_OT_NUMTYPES ||
@@ -792,18 +789,21 @@ restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
 		return (EINVAL);
 	}
 
+	err = dmu_object_info(os, drro->drr_object, NULL);
+
+	if (err != 0 && err != ENOENT)
+		return (EINVAL);
+
 	if (drro->drr_bonuslen) {
 		data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8));
 		if (ra->err)
 			return (ra->err);
 	}
 
-	tx = dmu_tx_create(os);
-
 	if (err == ENOENT) {
 		/* currently free, want to be allocated */
+		tx = dmu_tx_create(os);
 		dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
-		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, 1);
 		err = dmu_tx_assign(tx, TXG_WAIT);
 		if (err) {
 			dmu_tx_abort(tx);
@@ -812,28 +812,22 @@ restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
 		err = dmu_object_claim(os, drro->drr_object,
 		    drro->drr_type, drro->drr_blksz,
 		    drro->drr_bonustype, drro->drr_bonuslen, tx);
+		dmu_tx_commit(tx);
 	} else {
 		/* currently allocated, want to be allocated */
-		dmu_tx_hold_bonus(tx, drro->drr_object);
-		/*
-		 * We may change blocksize and delete old content,
-		 * so need to hold_write and hold_free.
-		 */
-		dmu_tx_hold_write(tx, drro->drr_object, 0, 1);
-		dmu_tx_hold_free(tx, drro->drr_object, 0, DMU_OBJECT_END);
-		err = dmu_tx_assign(tx, TXG_WAIT);
-		if (err) {
-			dmu_tx_abort(tx);
-			return (err);
-		}
-
 		err = dmu_object_reclaim(os, drro->drr_object,
 		    drro->drr_type, drro->drr_blksz,
-		    drro->drr_bonustype, drro->drr_bonuslen, tx);
+		    drro->drr_bonustype, drro->drr_bonuslen);
 	}
-	if (err) {
+	if (err)
-		dmu_tx_commit(tx);
 		return (EINVAL);
+
+	tx = dmu_tx_create(os);
+	dmu_tx_hold_bonus(tx, drro->drr_object);
+	err = dmu_tx_assign(tx, TXG_WAIT);
+	if (err) {
+		dmu_tx_abort(tx);
+		return (err);
 	}
 
 	dmu_object_set_checksum(os, drro->drr_object, drro->drr_checksum, tx);

module/zfs/dmu_traverse.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -64,6 +64,9 @@ struct traverse_data {
 	void *td_arg;
 };
 
+static int traverse_dnode(struct traverse_data *td, const dnode_phys_t *dnp,
+    arc_buf_t *buf, uint64_t objset, uint64_t object);
+
 /* ARGSUSED */
 static void
 traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
@@ -189,7 +192,7 @@ traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
 		}
 	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
 		uint32_t flags = ARC_WAIT;
-		int i, j;
+		int i;
 		int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
 
 		err = arc_read(NULL, td->td_spa, bp, pbuf,
@@ -201,20 +204,15 @@ traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
 		/* recursively visitbp() blocks below this */
 		dnp = buf->b_data;
 		for (i = 0; i < epb && err == 0; i++, dnp++) {
-			for (j = 0; j < dnp->dn_nblkptr; j++) {
+			err = traverse_dnode(td, dnp, buf, zb->zb_objset,
-				SET_BOOKMARK(&czb, zb->zb_objset,
+			    zb->zb_blkid * epb + i);
-				    zb->zb_blkid * epb + i,
+			if (err)
-				    dnp->dn_nlevels - 1, j);
+				break;
-				err = traverse_visitbp(td, dnp, buf,
-				    (blkptr_t *)&dnp->dn_blkptr[j], &czb);
-				if (err)
-					break;
-			}
 		}
 	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
 		uint32_t flags = ARC_WAIT;
 		objset_phys_t *osp;
-		int j;
+		dnode_phys_t *dnp;
 
 		err = arc_read_nolock(NULL, td->td_spa, bp,
 		    arc_getbuf_func, &buf,
@@ -223,20 +221,19 @@ traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
 			return (err);
 
 		osp = buf->b_data;
-		/*
-		 * traverse_zil is just here for zdb's leak checking.
-		 * For other consumers, there will be no ZIL blocks.
-		 */
 		traverse_zil(td, &osp->os_zil_header);
 
-		for (j = 0; j < osp->os_meta_dnode.dn_nblkptr; j++) {
+		dnp = &osp->os_meta_dnode;
-			SET_BOOKMARK(&czb, zb->zb_objset, 0,
+		err = traverse_dnode(td, dnp, buf, zb->zb_objset, 0);
-			    osp->os_meta_dnode.dn_nlevels - 1, j);
+		if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) {
-			err = traverse_visitbp(td, &osp->os_meta_dnode, buf,
+			dnp = &osp->os_userused_dnode;
-			    (blkptr_t *)&osp->os_meta_dnode.dn_blkptr[j],
+			err = traverse_dnode(td, dnp, buf, zb->zb_objset,
-			    &czb);
+			    DMU_USERUSED_OBJECT);
-			if (err)
+		}
-				break;
+		if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) {
+			dnp = &osp->os_groupused_dnode;
+			err = traverse_dnode(td, dnp, buf, zb->zb_objset,
+			    DMU_GROUPUSED_OBJECT);
 		}
 	}
 
@@ -249,6 +246,23 @@ traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
 	return (err);
 }
 
+static int
+traverse_dnode(struct traverse_data *td, const dnode_phys_t *dnp,
+    arc_buf_t *buf, uint64_t objset, uint64_t object)
+{
+	int j, err = 0;
+	zbookmark_t czb;
+
+	for (j = 0; j < dnp->dn_nblkptr; j++) {
+		SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
+		err = traverse_visitbp(td, dnp, buf,
+		    (blkptr_t *)&dnp->dn_blkptr[j], &czb);
+		if (err)
+			break;
+	}
+	return (err);
+}
+
 /* ARGSUSED */
 static int
 traverse_prefetcher(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,

module/zfs/dmu_tx.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -162,6 +162,41 @@ dmu_tx_check_ioerr(zio_t *zio, dnode_t *dn, int level, uint64_t blkid)
 	return (err);
 }
 
+static void
+dmu_tx_count_indirects(dmu_tx_hold_t *txh, dmu_buf_impl_t *db,
+    boolean_t freeable, dmu_buf_impl_t **history)
+{
+	int i = db->db_level + 1;
+	dnode_t *dn = db->db_dnode;
+
+	if (i >= dn->dn_nlevels)
+		return;
+
+	db = db->db_parent;
+	if (db == NULL) {
+		uint64_t lvls = dn->dn_nlevels - i;
+
+		txh->txh_space_towrite += lvls << dn->dn_indblkshift;
+		return;
+	}
+
+	if (db != history[i]) {
+		dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
+		uint64_t space = 1ULL << dn->dn_indblkshift;
+
+		freeable = (db->db_blkptr && (freeable ||
+		    dsl_dataset_block_freeable(ds, db->db_blkptr->blk_birth)));
+		if (freeable)
+			txh->txh_space_tooverwrite += space;
+		else
+			txh->txh_space_towrite += space;
+		if (db->db_blkptr)
+			txh->txh_space_tounref += space;
+		history[i] = db;
+		dmu_tx_count_indirects(txh, db, freeable, history);
+	}
+}
+
 /* ARGSUSED */
 static void
 dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
@@ -179,18 +214,24 @@ dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
 	min_ibs = DN_MIN_INDBLKSHIFT;
 	max_ibs = DN_MAX_INDBLKSHIFT;
 
-	/*
-	 * For i/o error checking, read the first and last level-0
-	 * blocks (if they are not aligned), and all the level-1 blocks.
-	 */
-
 	if (dn) {
+		dmu_buf_impl_t *last[DN_MAX_LEVELS];
+		int nlvls = dn->dn_nlevels;
+		int delta;
+
+		/*
+		 * For i/o error checking, read the first and last level-0
+		 * blocks (if they are not aligned), and all the level-1 blocks.
+		 */
 		if (dn->dn_maxblkid == 0) {
-			if ((off > 0 || len < dn->dn_datablksz) &&
+			delta = dn->dn_datablksz;
-			    off < dn->dn_datablksz) {
+			start = (off < dn->dn_datablksz) ? 0 : 1;
+			end = (off+len <= dn->dn_datablksz) ? 0 : 1;
+			if (start == 0 && (off > 0 || len < dn->dn_datablksz)) {
 				err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
 				if (err)
 					goto out;
+				delta -= off;
 			}
 		} else {
 			zio_t *zio = zio_root(dn->dn_objset->os_spa,
@@ -215,10 +256,9 @@ dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
 			}
 
 			/* level-1 blocks */
-			if (dn->dn_nlevels > 1) {
+			if (nlvls > 1) {
-				start >>= dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+				int shft = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
-				end >>= dn->dn_indblkshift - SPA_BLKPTRSHIFT;
+				for (i = (start>>shft)+1; i < end>>shft; i++) {
-				for (i = start+1; i < end; i++) {
 					err = dmu_tx_check_ioerr(zio, dn, 1, i);
 					if (err)
 						goto out;
@@ -228,20 +268,70 @@ dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
 			err = zio_wait(zio);
 			if (err)
 				goto out;
+			delta = P2NPHASE(off, dn->dn_datablksz);
 		}
-	}
 
-	/*
+		if (dn->dn_maxblkid > 0) {
-	 * If there's more than one block, the blocksize can't change,
+			/*
-	 * so we can make a more precise estimate.  Alternatively,
+			 * The blocksize can't change,
-	 * if the dnode's ibs is larger than max_ibs, always use that.
+			 * so we can make a more precise estimate.
-	 * This ensures that if we reduce DN_MAX_INDBLKSHIFT,
+			 */
-	 * the code will still work correctly on existing pools.
+			ASSERT(dn->dn_datablkshift != 0);
-	 */
-	if (dn && (dn->dn_maxblkid != 0 || dn->dn_indblkshift > max_ibs)) {
-		min_ibs = max_ibs = dn->dn_indblkshift;
-		if (dn->dn_datablkshift != 0)
 			min_bs = max_bs = dn->dn_datablkshift;
+			min_ibs = max_ibs = dn->dn_indblkshift;
+		} else if (dn->dn_indblkshift > max_ibs) {
+			/*
+			 * This ensures that if we reduce DN_MAX_INDBLKSHIFT,
+			 * the code will still work correctly on older pools.
+			 */
+			min_ibs = max_ibs = dn->dn_indblkshift;
+		}
+
+		/*
+		 * If this write is not off the end of the file
+		 * we need to account for overwrites/unref.
+		 */
+		if (start <= dn->dn_maxblkid)
+			bzero(last, sizeof (dmu_buf_impl_t *) * DN_MAX_LEVELS);
+		while (start <= dn->dn_maxblkid) {
+			spa_t *spa = txh->txh_tx->tx_pool->dp_spa;
+			dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
+			dmu_buf_impl_t *db;
+
+			rw_enter(&dn->dn_struct_rwlock, RW_READER);
+			db = dbuf_hold_level(dn, 0, start, FTAG);
+			rw_exit(&dn->dn_struct_rwlock);
+			if (db->db_blkptr && dsl_dataset_block_freeable(ds,
+			    db->db_blkptr->blk_birth)) {
+				dprintf_bp(db->db_blkptr, "can free old%s", "");
+				txh->txh_space_tooverwrite += dn->dn_datablksz;
+				txh->txh_space_tounref += dn->dn_datablksz;
+				dmu_tx_count_indirects(txh, db, TRUE, last);
+			} else {
+				txh->txh_space_towrite += dn->dn_datablksz;
+				if (db->db_blkptr)
+					txh->txh_space_tounref +=
+					    bp_get_dasize(spa, db->db_blkptr);
+				dmu_tx_count_indirects(txh, db, FALSE, last);
+			}
+			dbuf_rele(db, FTAG);
+			if (++start > end) {
+				/*
+				 * Account for new indirects appearing
+				 * before this IO gets assigned into a txg.
+				 */
+				bits = 64 - min_bs;
+				epbs = min_ibs - SPA_BLKPTRSHIFT;
+				for (bits -= epbs * (nlvls - 1);
+				    bits >= 0; bits -= epbs)
+					txh->txh_fudge += 1ULL << max_ibs;
+				goto out;
+			}
+			off += delta;
+			if (len >= delta)
+				len -= delta;
+			delta = dn->dn_datablksz;
+		}
 	}
 
 	/*
@@ -264,20 +354,22 @@ dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
 	for (bits = 64 - min_bs; bits >= 0; bits -= epbs) {
 		start >>= epbs;
 		end >>= epbs;
-		/*
+		ASSERT3U(end, >=, start);
-		 * If we increase the number of levels of indirection,
-		 * we'll need new blkid=0 indirect blocks.  If start == 0,
-		 * we're already accounting for that blocks; and if end == 0,
-		 * we can't increase the number of levels beyond that.
-		 */
-		if (start != 0 && end != 0)
-			txh->txh_space_towrite += 1ULL << max_ibs;
 		txh->txh_space_towrite += (end - start + 1) << max_ibs;
+		if (start != 0) {
+			/*
+			 * We also need a new blkid=0 indirect block
+			 * to reference any existing file data.
+			 */
+			txh->txh_space_towrite += 1ULL << max_ibs;
+		}
 	}
 
-	ASSERT(txh->txh_space_towrite < 2 * DMU_MAX_ACCESS);
-
 out:
+	if (txh->txh_space_towrite + txh->txh_space_tooverwrite >
+	    2 * DMU_MAX_ACCESS)
+		err = EFBIG;
+
 	if (err)
 		txh->txh_tx->tx_err = err;
 }
@@ -294,6 +386,7 @@ dmu_tx_count_dnode(dmu_tx_hold_t *txh)
 	    dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,
 	    dn->dn_dbuf->db_blkptr->blk_birth)) {
 		txh->txh_space_tooverwrite += space;
+		txh->txh_space_tounref += space;
 	} else {
 		txh->txh_space_towrite += space;
 		if (dn && dn->dn_dbuf->db_blkptr)
@@ -537,7 +630,7 @@ dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len)
 }
 
 void
-dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name)
+dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name)
 {
 	dmu_tx_hold_t *txh;
 	dnode_t *dn;
@@ -586,9 +679,9 @@ dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name)
 			txh->txh_space_tooverwrite += SPA_MAXBLOCKSIZE;
 		} else {
 			txh->txh_space_towrite += SPA_MAXBLOCKSIZE;
-			txh->txh_space_tounref +=
-			    BP_GET_ASIZE(dn->dn_phys->dn_blkptr);
 		}
+		if (dn->dn_phys->dn_blkptr[0].blk_birth)
+			txh->txh_space_tounref += SPA_MAXBLOCKSIZE;
 		return;
 	}
 
@@ -605,12 +698,9 @@ dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name)
 		}
 	}
 
-	/*
+	err = zap_count_write(&dn->dn_objset->os, dn->dn_object, name, add,
-	 * 3 blocks overwritten: target leaf, ptrtbl block, header block
+	    &txh->txh_space_towrite, &txh->txh_space_tooverwrite,
-	 * 3 new blocks written if adding: new split leaf, 2 grown ptrtbl blocks
+	    txh->txh_dnode->dn_datablkshift);
-	 */
-	dmu_tx_count_write(txh, dn->dn_maxblkid * dn->dn_datablksz,
-	    (3 + (add ? 3 : 0)) << dn->dn_datablkshift);
 
 	/*
 	 * If the modified blocks are scattered to the four winds,
@@ -618,7 +708,10 @@ dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name)
 	 */
 	epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
 	for (nblocks = dn->dn_maxblkid >> epbs; nblocks != 0; nblocks >>= epbs)
-		txh->txh_space_towrite += 3 << dn->dn_indblkshift;
+		if (dn->dn_objset->os_dsl_dataset->ds_phys->ds_prev_snap_obj)
+			txh->txh_space_towrite += 3 << dn->dn_indblkshift;
+		else
+			txh->txh_space_tooverwrite += 3 << dn->dn_indblkshift;
 }
 
 void

module/zfs/dnode.c

@@ -156,7 +156,7 @@ dnode_verify(dnode_t *dn)
 	}
 	if (dn->dn_phys->dn_type != DMU_OT_NONE)
 		ASSERT3U(dn->dn_phys->dn_nlevels, <=, dn->dn_nlevels);
-	ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || dn->dn_dbuf != NULL);
+	ASSERT(DMU_OBJECT_IS_SPECIAL(dn->dn_object) || dn->dn_dbuf != NULL);
 	if (dn->dn_dbuf != NULL) {
 		ASSERT3P(dn->dn_phys, ==,
 		    (dnode_phys_t *)dn->dn_dbuf->db.db_data +
@@ -321,6 +321,7 @@ dnode_destroy(dnode_t *dn)
 	}
 	ASSERT(NULL == list_head(&dn->dn_dbufs));
 #endif
+	ASSERT(dn->dn_oldphys == NULL);
 
 	mutex_enter(&os->os_lock);
 	list_remove(&os->os_dnodes, dn);
@@ -417,8 +418,7 @@ void
 dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
 {
-	int i, nblkptr;
+	int nblkptr;
-	dmu_buf_impl_t *db = NULL;
 
 	ASSERT3U(blocksize, >=, SPA_MINBLOCKSIZE);
 	ASSERT3U(blocksize, <=, SPA_MAXBLOCKSIZE);
@@ -430,42 +430,25 @@ dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
 	ASSERT3U(bonustype, <, DMU_OT_NUMTYPES);
 	ASSERT3U(bonuslen, <=, DN_MAX_BONUSLEN);
 
-	for (i = 0; i < TXG_SIZE; i++)
-		ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
-
 	/* clean up any unreferenced dbufs */
 	dnode_evict_dbufs(dn);
-	ASSERT3P(list_head(&dn->dn_dbufs), ==, NULL);
 
-	/*
-	 * XXX I should really have a generation number to tell if we
-	 * need to do this...
-	 */
-	if (blocksize != dn->dn_datablksz ||
-	    dn->dn_bonustype != bonustype || dn->dn_bonuslen != bonuslen) {
-		/* free all old data */
-		dnode_free_range(dn, 0, -1ULL, tx);
-	}
-
-	nblkptr = 1 + ((DN_MAX_BONUSLEN - bonuslen) >> SPA_BLKPTRSHIFT);
-
-	/* change blocksize */
 	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
-	if (blocksize != dn->dn_datablksz &&
-	    (!BP_IS_HOLE(&dn->dn_phys->dn_blkptr[0]) ||
-	    list_head(&dn->dn_dbufs) != NULL)) {
-		db = dbuf_hold(dn, 0, FTAG);
-		dbuf_new_size(db, blocksize, tx);
-	}
-	dnode_setdblksz(dn, blocksize);
 	dnode_setdirty(dn, tx);
-	dn->dn_next_bonuslen[tx->tx_txg&TXG_MASK] = bonuslen;
+	if (dn->dn_datablksz != blocksize) {
-	dn->dn_next_blksz[tx->tx_txg&TXG_MASK] = blocksize;
+		/* change blocksize */
+		ASSERT(dn->dn_maxblkid == 0 &&
+		    (BP_IS_HOLE(&dn->dn_phys->dn_blkptr[0]) ||
+		    dnode_block_freed(dn, 0)));
+		dnode_setdblksz(dn, blocksize);
+		dn->dn_next_blksz[tx->tx_txg&TXG_MASK] = blocksize;
+	}
+	if (dn->dn_bonuslen != bonuslen)
+		dn->dn_next_bonuslen[tx->tx_txg&TXG_MASK] = bonuslen;
+	nblkptr = 1 + ((DN_MAX_BONUSLEN - bonuslen) >> SPA_BLKPTRSHIFT);
 	if (dn->dn_nblkptr != nblkptr)
 		dn->dn_next_nblkptr[tx->tx_txg&TXG_MASK] = nblkptr;
 	rw_exit(&dn->dn_struct_rwlock);
-	if (db)
-		dbuf_rele(db, FTAG);
 
 	/* change type */
 	dn->dn_type = ot;
@@ -569,6 +552,22 @@ dnode_hold_impl(objset_impl_t *os, uint64_t object, int flag,
 	 */
 	ASSERT(spa_config_held(os->os_spa, SCL_ALL, RW_WRITER) == 0);
 
+	if (object == DMU_USERUSED_OBJECT || object == DMU_GROUPUSED_OBJECT) {
+		dn = (object == DMU_USERUSED_OBJECT) ?
+		    os->os_userused_dnode : os->os_groupused_dnode;
+		if (dn == NULL)
+			return (ENOENT);
+		type = dn->dn_type;
+		if ((flag & DNODE_MUST_BE_ALLOCATED) && type == DMU_OT_NONE)
+			return (ENOENT);
+		if ((flag & DNODE_MUST_BE_FREE) && type != DMU_OT_NONE)
+			return (EEXIST);
+		DNODE_VERIFY(dn);
+		(void) refcount_add(&dn->dn_holds, tag);
+		*dnp = dn;
+		return (0);
+	}
+
 	if (object == 0 || object >= DN_MAX_OBJECT)
 		return (EINVAL);
 
@@ -627,7 +626,8 @@ dnode_hold_impl(objset_impl_t *os, uint64_t object, int flag,
 	type = dn->dn_type;
 	if (dn->dn_free_txg ||
 	    ((flag & DNODE_MUST_BE_ALLOCATED) && type == DMU_OT_NONE) ||
-	    ((flag & DNODE_MUST_BE_FREE) && type != DMU_OT_NONE)) {
+	    ((flag & DNODE_MUST_BE_FREE) &&
+	    (type != DMU_OT_NONE || dn->dn_oldphys))) {
 		mutex_exit(&dn->dn_mtx);
 		dbuf_rele(db, FTAG);
 		return (type == DMU_OT_NONE ? ENOENT : EEXIST);
@@ -692,8 +692,10 @@ dnode_setdirty(dnode_t *dn, dmu_tx_t *tx)
 	objset_impl_t *os = dn->dn_objset;
 	uint64_t txg = tx->tx_txg;
 
-	if (dn->dn_object == DMU_META_DNODE_OBJECT)
+	if (DMU_OBJECT_IS_SPECIAL(dn->dn_object)) {
+		dsl_dataset_dirty(os->os_dsl_dataset, tx);
 		return;
+	}
 
 	DNODE_VERIFY(dn);
 
@@ -1189,11 +1191,6 @@ dnode_block_freed(dnode_t *dn, uint64_t blkid)
 	if (dn->dn_free_txg)
 		return (TRUE);
 
-	/*
-	 * If dn_datablkshift is not set, then there's only a single
-	 * block, in which case there will never be a free range so it
-	 * won't matter.
-	 */
 	range_tofind.fr_blkid = blkid;
 	mutex_enter(&dn->dn_mtx);
 	for (i = 0; i < TXG_SIZE; i++) {
@@ -1278,7 +1275,7 @@ dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
 	dprintf("probing object %llu offset %llx level %d of %u\n",
 	    dn->dn_object, *offset, lvl, dn->dn_phys->dn_nlevels);
 
-	hole = flags & DNODE_FIND_HOLE;
+	hole = ((flags & DNODE_FIND_HOLE) != 0);
 	inc = (flags & DNODE_FIND_BACKWARDS) ? -1 : 1;
 	ASSERT(txg == 0 || !hole);
 
@@ -1325,16 +1322,7 @@ dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
 
 		for (i = (*offset >> span) & (blkfill - 1);
 		    i >= 0 && i < blkfill; i += inc) {
-			boolean_t newcontents = B_TRUE;
+			if ((dnp[i].dn_type == DMU_OT_NONE) == hole)
-			if (txg) {
-				int j;
-				newcontents = B_FALSE;
-				for (j = 0; j < dnp[i].dn_nblkptr; j++) {
-					if (dnp[i].dn_blkptr[j].blk_birth > txg)
-						newcontents = B_TRUE;
-				}
-			}
-			if (!dnp[i].dn_type == hole && newcontents)
 				break;
 			*offset += (1ULL << span) * inc;
 		}

module/zfs/dnode_sync.c

@@ -504,9 +504,6 @@ dnode_sync_free(dnode_t *dn, dmu_tx_t *tx)
 
 /*
  * Write out the dnode's dirty buffers.
- *
- * NOTE: The dnode is kept in memory by being dirty.  Once the
- * dirty bit is cleared, it may be evicted.  Beware of this!
  */
 void
 dnode_sync(dnode_t *dn, dmu_tx_t *tx)
@@ -515,20 +512,33 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 	dnode_phys_t *dnp = dn->dn_phys;
 	int txgoff = tx->tx_txg & TXG_MASK;
 	list_t *list = &dn->dn_dirty_records[txgoff];
+	static const dnode_phys_t zerodn = { 0 };
 
 	ASSERT(dmu_tx_is_syncing(tx));
 	ASSERT(dnp->dn_type != DMU_OT_NONE || dn->dn_allocated_txg);
+	ASSERT(dnp->dn_type != DMU_OT_NONE ||
+	    bcmp(dnp, &zerodn, DNODE_SIZE) == 0);
 	DNODE_VERIFY(dn);
 
 	ASSERT(dn->dn_dbuf == NULL || arc_released(dn->dn_dbuf->db_buf));
 
+	if (dmu_objset_userused_enabled(dn->dn_objset) &&
+	    !DMU_OBJECT_IS_SPECIAL(dn->dn_object)) {
+		ASSERT(dn->dn_oldphys == NULL);
+		dn->dn_oldphys = zio_buf_alloc(sizeof (dnode_phys_t));
+		*dn->dn_oldphys = *dn->dn_phys; /* struct assignment */
+		dn->dn_phys->dn_flags |= DNODE_FLAG_USERUSED_ACCOUNTED;
+	} else {
+		/* Once we account for it, we should always account for it. */
+		ASSERT(!(dn->dn_phys->dn_flags &
+		    DNODE_FLAG_USERUSED_ACCOUNTED));
+	}
+
 	mutex_enter(&dn->dn_mtx);
 	if (dn->dn_allocated_txg == tx->tx_txg) {
 		/* The dnode is newly allocated or reallocated */
 		if (dnp->dn_type == DMU_OT_NONE) {
 			/* this is a first alloc, not a realloc */
-			/* XXX shouldn't the phys already be zeroed? */
-			bzero(dnp, DNODE_CORE_SIZE);
 			dnp->dn_nlevels = 1;
 			dnp->dn_nblkptr = dn->dn_nblkptr;
 		}
@@ -626,7 +636,7 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 
 	dbuf_sync_list(list, tx);
 
-	if (dn->dn_object != DMU_META_DNODE_OBJECT) {
+	if (!DMU_OBJECT_IS_SPECIAL(dn->dn_object)) {
 		ASSERT3P(list_head(list), ==, NULL);
 		dnode_rele(dn, (void *)(uintptr_t)tx->tx_txg);
 	}

module/zfs/dsl_dataset.c

@@ -229,7 +229,7 @@ dsl_dataset_prev_snap_txg(dsl_dataset_t *ds)
 	return (MAX(ds->ds_phys->ds_prev_snap_txg, trysnap));
 }
 
-int
+boolean_t
 dsl_dataset_block_freeable(dsl_dataset_t *ds, uint64_t blk_birth)
 {
 	return (blk_birth > dsl_dataset_prev_snap_txg(ds));
@@ -548,6 +548,7 @@ dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, int flags, void *owner,
 		return (err);
 	if (!dsl_dataset_tryown(*dsp, DS_MODE_IS_INCONSISTENT(flags), owner)) {
 		dsl_dataset_rele(*dsp, owner);
+		*dsp = NULL;
 		return (EBUSY);
 	}
 	return (0);
@@ -974,6 +975,27 @@ dsl_dataset_destroy(dsl_dataset_t *ds, void *tag)
 		(void) dmu_free_object(os, obj);
 	}
 
+	/*
+	 * We need to sync out all in-flight IO before we try to evict
+	 * (the dataset evict func is trying to clear the cached entries
+	 * for this dataset in the ARC).
+	 */
+	txg_wait_synced(dd->dd_pool, 0);
+
+	/*
+	 * If we managed to free all the objects in open
+	 * context, the user space accounting should be zero.
+	 */
+	if (ds->ds_phys->ds_bp.blk_fill == 0 &&
+	    dmu_objset_userused_enabled(os->os)) {
+		uint64_t count;
+
+		ASSERT(zap_count(os, DMU_USERUSED_OBJECT, &count) != 0 ||
+		    count == 0);
+		ASSERT(zap_count(os, DMU_GROUPUSED_OBJECT, &count) != 0 ||
+		    count == 0);
+	}
+
 	dmu_objset_close(os);
 	if (err != ESRCH)
 		goto out;
@@ -1058,7 +1080,6 @@ dsl_dataset_get_user_ptr(dsl_dataset_t *ds)
 	return (ds->ds_user_ptr);
 }
 
-
 blkptr_t *
 dsl_dataset_get_blkptr(dsl_dataset_t *ds)
 {
@@ -1164,8 +1185,18 @@ kill_blkptr(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
 	if (bp == NULL)
 		return (0);
 
-	ASSERT3U(bp->blk_birth, >, ka->ds->ds_phys->ds_prev_snap_txg);
+	if ((zb->zb_level == -1ULL && zb->zb_blkid != 0) ||
-	(void) dsl_dataset_block_kill(ka->ds, bp, ka->zio, ka->tx);
+	    (zb->zb_object != 0 && dnp == NULL)) {
+		/*
+		 * It's a block in the intent log.  It has no
+		 * accounting, so just free it.
+		 */
+		VERIFY3U(0, ==, dsl_free(ka->zio, ka->tx->tx_pool,
+		    ka->tx->tx_txg, bp, NULL, NULL, ARC_NOWAIT));
+	} else {
+		ASSERT3U(bp->blk_birth, >, ka->ds->ds_phys->ds_prev_snap_txg);
+		(void) dsl_dataset_block_kill(ka->ds, bp, ka->zio, ka->tx);
+	}
 
 	return (0);
 }
@@ -1209,13 +1240,7 @@ dsl_dataset_rollback_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
 
 	dmu_buf_will_dirty(ds->ds_dbuf, tx);
 
-	/*
-	 * Before the roll back destroy the zil.
-	 */
 	if (ds->ds_user_ptr != NULL) {
-		zil_rollback_destroy(
-		    ((objset_impl_t *)ds->ds_user_ptr)->os_zil, tx);
-
 		/*
 		 * We need to make sure that the objset_impl_t is reopened after
 		 * we do the rollback, otherwise it will have the wrong
@@ -1248,7 +1273,10 @@ dsl_dataset_rollback_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
 	    ds->ds_phys->ds_deadlist_obj));
 
 	{
-		/* Free blkptrs that we gave birth to */
+		/*
+		 * Free blkptrs that we gave birth to - this covers
+		 * claimed but not played log blocks too.
+		 */
 		zio_t *zio;
 		struct killarg ka;
 
@@ -1262,8 +1290,7 @@ dsl_dataset_rollback_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
 		(void) zio_wait(zio);
 	}
 
-	ASSERT(!(ds->ds_phys->ds_flags & DS_FLAG_UNIQUE_ACCURATE) ||
+	ASSERT(!DS_UNIQUE_IS_ACCURATE(ds) || ds->ds_phys->ds_unique_bytes == 0);
-	    ds->ds_phys->ds_unique_bytes == 0);
 
 	if (ds->ds_prev && ds->ds_prev != ds->ds_dir->dd_pool->dp_origin_snap) {
 		/* Change our contents to that of the prev snapshot */
@@ -1481,7 +1508,7 @@ dsl_dataset_destroy_sync(void *arg1, void *tag, cred_t *cr, dmu_tx_t *tx)
 		dmu_buf_will_dirty(ds_prev->ds_dbuf, tx);
 		if (after_branch_point &&
 		    ds_prev->ds_phys->ds_next_clones_obj != 0) {
-			VERIFY(0 == zap_remove_int(mos,
+			VERIFY3U(0, ==, zap_remove_int(mos,
 			    ds_prev->ds_phys->ds_next_clones_obj, obj, tx));
 			if (ds->ds_phys->ds_next_snap_obj != 0) {
 				VERIFY(0 == zap_add_int(mos,
@@ -1654,7 +1681,7 @@ dsl_dataset_destroy_sync(void *arg1, void *tag, cred_t *cr, dmu_tx_t *tx)
 		err = traverse_dataset(ds, ds->ds_phys->ds_prev_snap_txg,
 		    TRAVERSE_POST, kill_blkptr, &ka);
 		ASSERT3U(err, ==, 0);
-		ASSERT(spa_version(dp->dp_spa) < SPA_VERSION_UNIQUE_ACCURATE ||
+		ASSERT(!DS_UNIQUE_IS_ACCURATE(ds) ||
 		    ds->ds_phys->ds_unique_bytes == 0);
 	}
 
@@ -2583,7 +2610,7 @@ snaplist_destroy(list_t *l, boolean_t own)
 {
 	struct promotenode *snap;
 
-	if (!list_link_active(&l->list_head))
+	if (!l || !list_link_active(&l->list_head))
 		return;
 
 	while ((snap = list_tail(l)) != NULL) {

module/zfs/dsl_dir.c

@@ -227,24 +227,11 @@ dsl_dir_namelen(dsl_dir_t *dd)
 	return (result);
 }
 
-int
-dsl_dir_is_private(dsl_dir_t *dd)
-{
-	int rv = FALSE;
-
-	if (dd->dd_parent && dsl_dir_is_private(dd->dd_parent))
-		rv = TRUE;
-	if (dataset_name_hidden(dd->dd_myname))
-		rv = TRUE;
-	return (rv);
-}
-
-
 static int
 getcomponent(const char *path, char *component, const char **nextp)
 {
 	char *p;
-	if (path == NULL)
+	if ((path == NULL) || (path[0] == '\0'))
 		return (ENOENT);
 	/* This would be a good place to reserve some namespace... */
 	p = strpbrk(path, "/@");

module/zfs/dsl_pool.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -90,6 +90,9 @@ dsl_pool_open_impl(spa_t *spa, uint64_t txg)
 	mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&dp->dp_scrub_cancel_lock, NULL, MUTEX_DEFAULT, NULL);
 
+	dp->dp_vnrele_taskq = taskq_create("zfs_vn_rele_taskq", 1, minclsyspri,
+	    1, 4, 0);
+
 	return (dp);
 }
 
@@ -129,14 +132,15 @@ dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
 			goto out;
 		err = dsl_dataset_hold_obj(dp, dd->dd_phys->dd_head_dataset_obj,
 		    FTAG, &ds);
-		if (err)
+		if (err == 0) {
-			goto out;
+			err = dsl_dataset_hold_obj(dp,
-		err = dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
+			    ds->ds_phys->ds_prev_snap_obj, dp,
-		    dp, &dp->dp_origin_snap);
+			    &dp->dp_origin_snap);
-		if (err)
+			dsl_dataset_rele(ds, FTAG);
-			goto out;
+		}
-		dsl_dataset_rele(ds, FTAG);
 		dsl_dir_close(dd, dp);
+		if (err)
+			goto out;
 	}
 
 	/* get scrub status */
@@ -226,6 +230,7 @@ dsl_pool_close(dsl_pool_t *dp)
 	rw_destroy(&dp->dp_config_rwlock);
 	mutex_destroy(&dp->dp_lock);
 	mutex_destroy(&dp->dp_scrub_cancel_lock);
+	taskq_destroy(dp->dp_vnrele_taskq);
 	if (dp->dp_blkstats)
 		kmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
 	kmem_free(dp, sizeof (dsl_pool_t));
@@ -296,24 +301,52 @@ dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
 	tx = dmu_tx_create_assigned(dp, txg);
 
 	dp->dp_read_overhead = 0;
+	start = gethrtime();
+
 	zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
 	while (ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) {
-		if (!list_link_active(&ds->ds_synced_link))
+		/*
-			list_insert_tail(&dp->dp_synced_datasets, ds);
+		 * We must not sync any non-MOS datasets twice, because
-		else
+		 * we may have taken a snapshot of them.  However, we
-			dmu_buf_rele(ds->ds_dbuf, ds);
+		 * may sync newly-created datasets on pass 2.
+		 */
+		ASSERT(!list_link_active(&ds->ds_synced_link));
+		list_insert_tail(&dp->dp_synced_datasets, ds);
 		dsl_dataset_sync(ds, zio, tx);
 	}
 	DTRACE_PROBE(pool_sync__1setup);
-
-	start = gethrtime();
 	err = zio_wait(zio);
+
 	write_time = gethrtime() - start;
 	ASSERT(err == 0);
 	DTRACE_PROBE(pool_sync__2rootzio);
 
-	while (dstg = txg_list_remove(&dp->dp_sync_tasks, txg))
+	for (ds = list_head(&dp->dp_synced_datasets); ds;
+	    ds = list_next(&dp->dp_synced_datasets, ds))
+		dmu_objset_do_userquota_callbacks(ds->ds_user_ptr, tx);
+
+	/*
+	 * Sync the datasets again to push out the changes due to
+	 * userquota updates.  This must be done before we process the
+	 * sync tasks, because that could cause a snapshot of a dataset
+	 * whose ds_bp will be rewritten when we do this 2nd sync.
+	 */
+	zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
+	while (ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) {
+		ASSERT(list_link_active(&ds->ds_synced_link));
+		dmu_buf_rele(ds->ds_dbuf, ds);
+		dsl_dataset_sync(ds, zio, tx);
+	}
+	err = zio_wait(zio);
+
+	while (dstg = txg_list_remove(&dp->dp_sync_tasks, txg)) {
+		/*
+		 * No more sync tasks should have been added while we
+		 * were syncing.
+		 */
+		ASSERT(spa_sync_pass(dp->dp_spa) == 1);
 		dsl_sync_task_group_sync(dstg, tx);
+	}
 	DTRACE_PROBE(pool_sync__3task);
 
 	start = gethrtime();
@@ -611,3 +644,9 @@ dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx)
 	dsl_dataset_rele(ds, FTAG);
 	rw_exit(&dp->dp_config_rwlock);
 }
+
+taskq_t *
+dsl_pool_vnrele_taskq(dsl_pool_t *dp)
+{
+	return (dp->dp_vnrele_taskq);
+}

module/zfs/dsl_prop.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/dmu.h>
 #include <sys/dmu_objset.h>
 #include <sys/dmu_tx.h>
@@ -415,6 +413,34 @@ dsl_prop_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
 	    "%s=%s dataset = %llu", psa->name, valstr, ds->ds_object);
 }
 
+void
+dsl_props_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+{
+	dsl_dataset_t *ds = arg1;
+	nvlist_t *nvl = arg2;
+	nvpair_t *elem = NULL;
+
+	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
+		struct prop_set_arg psa;
+
+		psa.name = nvpair_name(elem);
+
+		if (nvpair_type(elem) == DATA_TYPE_STRING) {
+			VERIFY(nvpair_value_string(elem,
+			    (char **)&psa.buf) == 0);
+			psa.intsz = 1;
+			psa.numints = strlen(psa.buf) + 1;
+		} else {
+			uint64_t intval;
+			VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+			psa.intsz = sizeof (intval);
+			psa.numints = 1;
+			psa.buf = &intval;
+		}
+		dsl_prop_set_sync(ds, &psa, cr, tx);
+	}
+}
+
 void
 dsl_prop_set_uint64_sync(dsl_dir_t *dd, const char *name, uint64_t val,
     cred_t *cr, dmu_tx_t *tx)
@@ -438,6 +464,7 @@ dsl_prop_set(const char *dsname, const char *propname,
     int intsz, int numints, const void *buf)
 {
 	dsl_dataset_t *ds;
+	uint64_t version;
 	int err;
 	struct prop_set_arg psa;
 
@@ -447,15 +474,19 @@ dsl_prop_set(const char *dsname, const char *propname,
 	 */
 	if (strlen(propname) >= ZAP_MAXNAMELEN)
 		return (ENAMETOOLONG);
-	if (intsz * numints >= ZAP_MAXVALUELEN)
-		return (E2BIG);
 
 	err = dsl_dataset_hold(dsname, FTAG, &ds);
 	if (err)
 		return (err);
 
+	version = spa_version(ds->ds_dir->dd_pool->dp_spa);
+	if (intsz * numints >= (version < SPA_VERSION_STMF_PROP ?
+	    ZAP_OLDMAXVALUELEN : ZAP_MAXVALUELEN)) {
+		dsl_dataset_rele(ds, FTAG);
+		return (E2BIG);
+	}
 	if (dsl_dataset_is_snapshot(ds) &&
-	    spa_version(ds->ds_dir->dd_pool->dp_spa) < SPA_VERSION_SNAP_PROPS) {
+	    version < SPA_VERSION_SNAP_PROPS) {
 		dsl_dataset_rele(ds, FTAG);
 		return (ENOTSUP);
 	}
@@ -471,6 +502,50 @@ dsl_prop_set(const char *dsname, const char *propname,
 	return (err);
 }
 
+int
+dsl_props_set(const char *dsname, nvlist_t *nvl)
+{
+	dsl_dataset_t *ds;
+	uint64_t version;
+	nvpair_t *elem = NULL;
+	int err;
+
+	if (err = dsl_dataset_hold(dsname, FTAG, &ds))
+		return (err);
+	/*
+	 * Do these checks before the syncfunc, since it can't fail.
+	 */
+	version = spa_version(ds->ds_dir->dd_pool->dp_spa);
+	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
+		if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
+			dsl_dataset_rele(ds, FTAG);
+			return (ENAMETOOLONG);
+		}
+		if (nvpair_type(elem) == DATA_TYPE_STRING) {
+			char *valstr;
+			VERIFY(nvpair_value_string(elem, &valstr) == 0);
+			if (strlen(valstr) >= (version <
+			    SPA_VERSION_STMF_PROP ?
+			    ZAP_OLDMAXVALUELEN : ZAP_MAXVALUELEN)) {
+				dsl_dataset_rele(ds, FTAG);
+				return (E2BIG);
+			}
+		}
+	}
+
+	if (dsl_dataset_is_snapshot(ds) &&
+	    version < SPA_VERSION_SNAP_PROPS) {
+		dsl_dataset_rele(ds, FTAG);
+		return (ENOTSUP);
+	}
+
+	err = dsl_sync_task_do(ds->ds_dir->dd_pool,
+	    NULL, dsl_props_set_sync, ds, nvl, 2);
+
+	dsl_dataset_rele(ds, FTAG);
+	return (err);
+}
+
 /*
  * Iterate over all properties for this dataset and return them in an nvlist.
  */

module/zfs/dsl_scrub.c

@@ -45,6 +45,8 @@ typedef int (scrub_cb_t)(dsl_pool_t *, const blkptr_t *, const zbookmark_t *);
 
 static scrub_cb_t dsl_pool_scrub_clean_cb;
 static dsl_syncfunc_t dsl_pool_scrub_cancel_sync;
+static void scrub_visitdnode(dsl_pool_t *dp, dnode_phys_t *dnp, arc_buf_t *buf,
+    uint64_t objset, uint64_t object);
 
 int zfs_scrub_min_time = 1; /* scrub for at least 1 sec each txg */
 int zfs_resilver_min_time = 3; /* resilver for at least 3 sec each txg */
@@ -348,6 +350,12 @@ traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
 	if (bp->blk_birth <= dp->dp_scrub_min_txg)
 		return;
 
+	/*
+	 * One block ("stubby") can be allocated a long time ago; we
+	 * want to visit that one because it has been allocated
+	 * (on-disk) even if it hasn't been claimed (even though for
+	 * plain scrub there's nothing to do to it).
+	 */
 	if (claim_txg == 0 && bp->blk_birth >= spa_first_txg(dp->dp_spa))
 		return;
 
@@ -373,6 +381,11 @@ traverse_zil_record(zilog_t *zilog, lr_t *lrc, void *arg, uint64_t claim_txg)
 		if (bp->blk_birth <= dp->dp_scrub_min_txg)
 			return;
 
+		/*
+		 * birth can be < claim_txg if this record's txg is
+		 * already txg sync'ed (but this log block contains
+		 * other records that are not synced)
+		 */
 		if (claim_txg == 0 || bp->blk_birth < claim_txg)
 			return;
 
@@ -472,7 +485,7 @@ scrub_visitbp(dsl_pool_t *dp, dnode_phys_t *dnp,
 	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
 		uint32_t flags = ARC_WAIT;
 		dnode_phys_t *child_dnp;
-		int i, j;
+		int i;
 		int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
 
 		err = arc_read(NULL, dp->dp_spa, bp, pbuf,
@@ -487,20 +500,12 @@ scrub_visitbp(dsl_pool_t *dp, dnode_phys_t *dnp,
 		child_dnp = buf->b_data;
 
 		for (i = 0; i < epb; i++, child_dnp++) {
-			for (j = 0; j < child_dnp->dn_nblkptr; j++) {
+			scrub_visitdnode(dp, child_dnp, buf, zb->zb_objset,
-				zbookmark_t czb;
+			    zb->zb_blkid * epb + i);
-
-				SET_BOOKMARK(&czb, zb->zb_objset,
-				    zb->zb_blkid * epb + i,
-				    child_dnp->dn_nlevels - 1, j);
-				scrub_visitbp(dp, child_dnp, buf,
-				    &child_dnp->dn_blkptr[j], &czb);
-			}
 		}
 	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
 		uint32_t flags = ARC_WAIT;
 		objset_phys_t *osp;
-		int j;
 
 		err = arc_read_nolock(NULL, dp->dp_spa, bp,
 		    arc_getbuf_func, &buf,
@@ -516,13 +521,13 @@ scrub_visitbp(dsl_pool_t *dp, dnode_phys_t *dnp,
 
 		traverse_zil(dp, &osp->os_zil_header);
 
-		for (j = 0; j < osp->os_meta_dnode.dn_nblkptr; j++) {
+		scrub_visitdnode(dp, &osp->os_meta_dnode,
-			zbookmark_t czb;
+		    buf, zb->zb_objset, 0);
-
+		if (arc_buf_size(buf) >= sizeof (objset_phys_t)) {
-			SET_BOOKMARK(&czb, zb->zb_objset, 0,
+			scrub_visitdnode(dp, &osp->os_userused_dnode,
-			    osp->os_meta_dnode.dn_nlevels - 1, j);
+			    buf, zb->zb_objset, 0);
-			scrub_visitbp(dp, &osp->os_meta_dnode, buf,
+			scrub_visitdnode(dp, &osp->os_groupused_dnode,
-			    &osp->os_meta_dnode.dn_blkptr[j], &czb);
+			    buf, zb->zb_objset, 0);
 		}
 	}
 
@@ -531,6 +536,21 @@ scrub_visitbp(dsl_pool_t *dp, dnode_phys_t *dnp,
 		(void) arc_buf_remove_ref(buf, &buf);
 }
 
+static void
+scrub_visitdnode(dsl_pool_t *dp, dnode_phys_t *dnp, arc_buf_t *buf,
+    uint64_t objset, uint64_t object)
+{
+	int j;
+
+	for (j = 0; j < dnp->dn_nblkptr; j++) {
+		zbookmark_t czb;
+
+		SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
+		scrub_visitbp(dp, dnp, buf, &dnp->dn_blkptr[j], &czb);
+	}
+
+}
+
 static void
 scrub_visit_rootbp(dsl_pool_t *dp, dsl_dataset_t *ds, blkptr_t *bp)
 {

module/zfs/fletcher.c

@@ -19,11 +19,111 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
+/*
+ * Fletcher Checksums
+ * ------------------
+ *
+ * ZFS's 2nd and 4th order Fletcher checksums are defined by the following
+ * recurrence relations:
+ *
+ *	a  = a    + f
+ *	 i    i-1    i-1
+ *
+ *	b  = b    + a
+ *	 i    i-1    i
+ *
+ *	c  = c    + b		(fletcher-4 only)
+ *	 i    i-1    i
+ *
+ *	d  = d    + c		(fletcher-4 only)
+ *	 i    i-1    i
+ *
+ * Where
+ *	a_0 = b_0 = c_0 = d_0 = 0
+ * and
+ *	f_0 .. f_(n-1) are the input data.
+ *
+ * Using standard techniques, these translate into the following series:
+ *
+ *	     __n_			     __n_
+ *	     \   |			     \   |
+ *	a  =  >     f			b  =  >     i * f
+ *	 n   /___|   n - i		 n   /___|	 n - i
+ *	     i = 1			     i = 1
+ *
+ *
+ *	     __n_			     __n_
+ *	     \   |  i*(i+1)		     \   |  i*(i+1)*(i+2)
+ *	c  =  >     ------- f		d  =  >     ------------- f
+ *	 n   /___|     2     n - i	 n   /___|	  6	   n - i
+ *	     i = 1			     i = 1
+ *
+ * For fletcher-2, the f_is are 64-bit, and [ab]_i are 64-bit accumulators.
+ * Since the additions are done mod (2^64), errors in the high bits may not
+ * be noticed.  For this reason, fletcher-2 is deprecated.
+ *
+ * For fletcher-4, the f_is are 32-bit, and [abcd]_i are 64-bit accumulators.
+ * A conservative estimate of how big the buffer can get before we overflow
+ * can be estimated using f_i = 0xffffffff for all i:
+ *
+ * % bc
+ *  f=2^32-1;d=0; for (i = 1; d<2^64; i++) { d += f*i*(i+1)*(i+2)/6 }; (i-1)*4
+ * 2264
+ *  quit
+ * %
+ *
+ * So blocks of up to 2k will not overflow.  Our largest block size is
+ * 128k, which has 32k 4-byte words, so we can compute the largest possible
+ * accumulators, then divide by 2^64 to figure the max amount of overflow:
+ *
+ * % bc
+ *  a=b=c=d=0; f=2^32-1; for (i=1; i<=32*1024; i++) { a+=f; b+=a; c+=b; d+=c }
+ *  a/2^64;b/2^64;c/2^64;d/2^64
+ * 0
+ * 0
+ * 1365
+ * 11186858
+ *  quit
+ * %
+ *
+ * So a and b cannot overflow.  To make sure each bit of input has some
+ * effect on the contents of c and d, we can look at what the factors of
+ * the coefficients in the equations for c_n and d_n are.  The number of 2s
+ * in the factors determines the lowest set bit in the multiplier.  Running
+ * through the cases for n*(n+1)/2 reveals that the highest power of 2 is
+ * 2^14, and for n*(n+1)*(n+2)/6 it is 2^15.  So while some data may overflow
+ * the 64-bit accumulators, every bit of every f_i effects every accumulator,
+ * even for 128k blocks.
+ *
+ * If we wanted to make a stronger version of fletcher4 (fletcher4c?),
+ * we could do our calculations mod (2^32 - 1) by adding in the carries
+ * periodically, and store the number of carries in the top 32-bits.
+ *
+ * --------------------
+ * Checksum Performance
+ * --------------------
+ *
+ * There are two interesting components to checksum performance: cached and
+ * uncached performance.  With cached data, fletcher-2 is about four times
+ * faster than fletcher-4.  With uncached data, the performance difference is
+ * negligible, since the cost of a cache fill dominates the processing time.
+ * Even though fletcher-4 is slower than fletcher-2, it is still a pretty
+ * efficient pass over the data.
+ *
+ * In normal operation, the data which is being checksummed is in a buffer
+ * which has been filled either by:
+ *
+ *	1. a compression step, which will be mostly cached, or
+ *	2. a bcopy() or copyin(), which will be uncached (because the
+ *	   copy is cache-bypassing).
+ *
+ * For both cached and uncached data, both fletcher checksums are much faster
+ * than sha-256, and slower than 'off', which doesn't touch the data at all.
+ */
 
 #include <sys/types.h>
 #include <sys/sysmacros.h>

module/zfs/include/sys/arc.h

@@ -85,6 +85,8 @@ void *arc_data_buf_alloc(uint64_t space);
 void arc_data_buf_free(void *buf, uint64_t space);
 arc_buf_t *arc_buf_alloc(spa_t *spa, int size, void *tag,
     arc_buf_contents_t type);
+arc_buf_t *arc_loan_buf(spa_t *spa, int size);
+void arc_return_buf(arc_buf_t *buf, void *tag);
 void arc_buf_add_ref(arc_buf_t *buf, void *tag);
 int arc_buf_remove_ref(arc_buf_t *buf, void *tag);
 int arc_buf_size(arc_buf_t *buf);
@@ -134,7 +136,7 @@ void arc_fini(void);
  * Level 2 ARC
  */
 
-void l2arc_add_vdev(spa_t *spa, vdev_t *vd, uint64_t start, uint64_t end);
+void l2arc_add_vdev(spa_t *spa, vdev_t *vd);
 void l2arc_remove_vdev(vdev_t *vd);
 boolean_t l2arc_vdev_present(vdev_t *vd);
 void l2arc_init(void);

module/zfs/include/sys/dbuf.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -264,6 +264,7 @@ void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx);
 void dmu_buf_will_not_fill(dmu_buf_t *db, dmu_tx_t *tx);
 void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx);
 void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx);
+void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx);
 dbuf_dirty_record_t *dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
 
 void dbuf_clear(dmu_buf_impl_t *db);

module/zfs/include/sys/dmu.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -60,6 +60,7 @@ struct zbookmark;
 struct spa;
 struct nvlist;
 struct objset_impl;
+struct arc_buf;
 
 typedef struct objset objset_t;
 typedef struct dmu_tx dmu_tx_t;
@@ -114,6 +115,8 @@ typedef enum dmu_object_type {
 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
 	DMU_OT_NEXT_CLONES,		/* ZAP */
 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
+	DMU_OT_USERGROUP_USED,		/* ZAP */
+	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
 	DMU_OT_NUMTYPES
 } dmu_object_type_t;
 
@@ -156,6 +159,9 @@ void zfs_znode_byteswap(void *buf, size_t size);
 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
 
+#define	DMU_USERUSED_OBJECT	(-1ULL)
+#define	DMU_GROUPUSED_OBJECT	(-2ULL)
+
 /*
  * Public routines to create, destroy, open, and close objsets.
  */
@@ -171,7 +177,8 @@ int dmu_objset_create(const char *name, dmu_objset_type_t type,
 int dmu_objset_destroy(const char *name);
 int dmu_snapshots_destroy(char *fsname, char *snapname);
 int dmu_objset_rollback(objset_t *os);
-int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
+int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
+    boolean_t recursive);
 int dmu_objset_rename(const char *name, const char *newname,
     boolean_t recursive);
 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
@@ -235,7 +242,7 @@ uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
-    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+    int blocksize, dmu_object_type_t bonustype, int bonuslen);
 
 /*
  * Free an object from this objset.
@@ -397,6 +404,11 @@ void *dmu_buf_get_user(dmu_buf_t *db);
  */
 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
 
+/*
+ * Tells if the given dbuf is freeable.
+ */
+boolean_t dmu_buf_freeable(dmu_buf_t *);
+
 /*
  * You must create a transaction, then hold the objects which you will
  * (or might) modify as part of this transaction.  Then you must assign
@@ -422,7 +434,7 @@ dmu_tx_t *dmu_tx_create(objset_t *os);
 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
     uint64_t len);
-void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
+void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
 void dmu_tx_abort(dmu_tx_t *tx);
 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
@@ -465,8 +477,10 @@ int dmu_free_object(objset_t *os, uint64_t object);
  * Canfail routines will return 0 on success, or an errno if there is a
  * nonrecoverable I/O error.
  */
+#define	DMU_READ_PREFETCH	0 /* prefetch */
+#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
-	void *buf);
+	void *buf, uint32_t flags);
 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
 	const void *buf, dmu_tx_t *tx);
 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
@@ -476,6 +490,10 @@ int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
     dmu_tx_t *tx);
 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
     uint64_t size, struct page *pp, dmu_tx_t *tx);
+struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
+void dmu_return_arcbuf(struct arc_buf *buf);
+void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
+    dmu_tx_t *tx);
 
 extern int zfs_prefetch_disable;
 
@@ -582,6 +600,12 @@ extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
     int maxlen, boolean_t *conflict);
 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
     uint64_t *idp, uint64_t *offp);
+
+typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
+    void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
+    dmu_tx_t *tx);
+extern void dmu_objset_register_type(dmu_objset_type_t ost,
+    objset_used_cb_t *cb);
 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
 extern void *dmu_objset_get_user(objset_t *os);
 

module/zfs/include/sys/dmu_objset.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -42,12 +42,20 @@ struct dsl_dataset;
 struct dmu_tx;
 struct objset_impl;
 
+#define	OBJSET_PHYS_SIZE 2048
+#define	OBJSET_OLD_PHYS_SIZE 1024
+
+#define	OBJSET_FLAG_USERACCOUNTING_COMPLETE	(1ULL<<0)
+
 typedef struct objset_phys {
 	dnode_phys_t os_meta_dnode;
 	zil_header_t os_zil_header;
 	uint64_t os_type;
-	char os_pad[1024 - sizeof (dnode_phys_t) - sizeof (zil_header_t) -
+	uint64_t os_flags;
-	    sizeof (uint64_t)];
+	char os_pad[OBJSET_PHYS_SIZE - sizeof (dnode_phys_t)*3 -
+	    sizeof (zil_header_t) - sizeof (uint64_t)*2];
+	dnode_phys_t os_userused_dnode;
+	dnode_phys_t os_groupused_dnode;
 } objset_phys_t;
 
 struct objset {
@@ -62,6 +70,8 @@ typedef struct objset_impl {
 	arc_buf_t *os_phys_buf;
 	objset_phys_t *os_phys;
 	dnode_t *os_meta_dnode;
+	dnode_t *os_userused_dnode;
+	dnode_t *os_groupused_dnode;
 	zilog_t *os_zil;
 	objset_t os;
 	uint8_t os_checksum;	/* can change, under dsl_dir's locks */
@@ -74,6 +84,8 @@ typedef struct objset_impl {
 	struct dmu_tx *os_synctx; /* XXX sketchy */
 	blkptr_t *os_rootbp;
 	zil_header_t os_zil_header;
+	list_t os_synced_dnodes;
+	uint64_t os_flags;
 
 	/* Protected by os_obj_lock */
 	kmutex_t os_obj_lock;
@@ -92,6 +104,7 @@ typedef struct objset_impl {
 } objset_impl_t;
 
 #define	DMU_META_DNODE_OBJECT	0
+#define	DMU_OBJECT_IS_SPECIAL(obj) ((int64_t)(obj) <= 0)
 
 #define	DMU_OS_IS_L2CACHEABLE(os)				\
 	((os)->os_secondary_cache == ZFS_CACHE_ALL ||		\
@@ -106,7 +119,8 @@ int dmu_objset_create(const char *name, dmu_objset_type_t type,
     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
 int dmu_objset_destroy(const char *name);
 int dmu_objset_rollback(objset_t *os);
-int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
+int dmu_objset_snapshot(char *fsname, char *snapname, nvlist_t *props,
+    boolean_t recursive);
 void dmu_objset_stats(objset_t *os, nvlist_t *nv);
 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
@@ -127,6 +141,10 @@ objset_impl_t *dmu_objset_create_impl(spa_t *spa, struct dsl_dataset *ds,
 int dmu_objset_open_impl(spa_t *spa, struct dsl_dataset *ds, blkptr_t *bp,
     objset_impl_t **osip);
 void dmu_objset_evict(struct dsl_dataset *ds, void *arg);
+void dmu_objset_do_userquota_callbacks(objset_impl_t *os, dmu_tx_t *tx);
+boolean_t dmu_objset_userused_enabled(objset_impl_t *os);
+int dmu_objset_userspace_upgrade(objset_t *os);
+boolean_t dmu_objset_userspace_present(objset_t *os);
 
 #ifdef	__cplusplus
 }

module/zfs/include/sys/dnode.h

@@ -98,7 +98,8 @@ enum dnode_dirtycontext {
 };
 
 /* Is dn_used in bytes?  if not, it's in multiples of SPA_MINBLOCKSIZE */
-#define	DNODE_FLAG_USED_BYTES	(1<<0)
+#define	DNODE_FLAG_USED_BYTES		(1<<0)
+#define	DNODE_FLAG_USERUSED_ACCOUNTED	(1<<1)
 
 typedef struct dnode_phys {
 	uint8_t dn_type;		/* dmu_object_type_t */
@@ -131,10 +132,7 @@ typedef struct dnode {
 	 */
 	krwlock_t dn_struct_rwlock;
 
-	/*
+	/* Our link on dn_objset->os_dnodes list; protected by os_lock.  */
-	 * Our link on dataset's dd_dnodes list.
-	 * Protected by dd_accounting_mtx.
-	 */
 	list_node_t dn_link;
 
 	/* immutable: */
@@ -191,6 +189,9 @@ typedef struct dnode {
 	/* parent IO for current sync write */
 	zio_t *dn_zio;
 
+	/* used in syncing context */
+	dnode_phys_t *dn_oldphys;
+
 	/* holds prefetch structure */
 	struct zfetch	dn_zfetch;
 } dnode_t;

module/zfs/include/sys/dsl_dataset.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -195,7 +195,7 @@ void dsl_dataset_sync(dsl_dataset_t *os, zio_t *zio, dmu_tx_t *tx);
 void dsl_dataset_block_born(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx);
 int dsl_dataset_block_kill(dsl_dataset_t *ds, blkptr_t *bp, zio_t *pio,
     dmu_tx_t *tx);
-int dsl_dataset_block_freeable(dsl_dataset_t *ds, uint64_t blk_birth);
+boolean_t dsl_dataset_block_freeable(dsl_dataset_t *ds, uint64_t blk_birth);
 uint64_t dsl_dataset_prev_snap_txg(dsl_dataset_t *ds);
 
 void dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx);

module/zfs/include/sys/dsl_deleg.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_DSL_DELEG_H
 #define	_SYS_DSL_DELEG_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/dmu.h>
 #include <sys/dsl_pool.h>
 #include <sys/zfs_context.h>
@@ -51,6 +49,10 @@ extern "C" {
 #define	ZFS_DELEG_PERM_ALLOW		"allow"
 #define	ZFS_DELEG_PERM_USERPROP		"userprop"
 #define	ZFS_DELEG_PERM_VSCAN		"vscan"
+#define	ZFS_DELEG_PERM_USERQUOTA	"userquota"
+#define	ZFS_DELEG_PERM_GROUPQUOTA	"groupquota"
+#define	ZFS_DELEG_PERM_USERUSED		"userused"
+#define	ZFS_DELEG_PERM_GROUPUSED	"groupused"
 
 /*
  * Note: the names of properties that are marked delegatable are also

module/zfs/include/sys/dsl_dir.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -107,7 +107,6 @@ int dsl_dir_open_obj(dsl_pool_t *dp, uint64_t ddobj,
     const char *tail, void *tag, dsl_dir_t **);
 void dsl_dir_name(dsl_dir_t *dd, char *buf);
 int dsl_dir_namelen(dsl_dir_t *dd);
-int dsl_dir_is_private(dsl_dir_t *dd);
 uint64_t dsl_dir_create_sync(dsl_pool_t *dp, dsl_dir_t *pds,
     const char *name, dmu_tx_t *tx);
 dsl_checkfunc_t dsl_dir_destroy_check;

module/zfs/include/sys/dsl_pool.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -77,6 +77,7 @@ typedef struct dsl_pool {
 	struct dsl_dir *dp_mos_dir;
 	struct dsl_dataset *dp_origin_snap;
 	uint64_t dp_root_dir_obj;
+	struct taskq *dp_vnrele_taskq;
 
 	/* No lock needed - sync context only */
 	blkptr_t dp_meta_rootbp;
@@ -143,6 +144,8 @@ int dsl_pool_scrub_clean(dsl_pool_t *dp);
 void dsl_pool_scrub_sync(dsl_pool_t *dp, dmu_tx_t *tx);
 void dsl_pool_scrub_restart(dsl_pool_t *dp);
 
+taskq_t *dsl_pool_vnrele_taskq(dsl_pool_t *dp);
+
 #ifdef	__cplusplus
 }
 #endif

module/zfs/include/sys/dsl_prop.h

@@ -19,18 +19,17 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_DSL_PROP_H
 #define	_SYS_DSL_PROP_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/dmu.h>
 #include <sys/dsl_pool.h>
 #include <sys/zfs_context.h>
+#include <sys/dsl_synctask.h>
 
 #ifdef	__cplusplus
 extern "C" {
@@ -66,8 +65,10 @@ int dsl_prop_get_ds(struct dsl_dataset *ds, const char *propname,
 int dsl_prop_get_dd(struct dsl_dir *dd, const char *propname,
     int intsz, int numints, void *buf, char *setpoint);
 
+dsl_syncfunc_t dsl_props_set_sync;
 int dsl_prop_set(const char *ddname, const char *propname,
     int intsz, int numints, const void *buf);
+int dsl_props_set(const char *dsname, nvlist_t *nvl);
 void dsl_prop_set_uint64_sync(dsl_dir_t *dd, const char *name, uint64_t val,
     cred_t *cr, dmu_tx_t *tx);
 

module/zfs/include/sys/metaslab.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -39,6 +39,8 @@ extern "C" {
 typedef struct metaslab_class metaslab_class_t;
 typedef struct metaslab_group metaslab_group_t;
 
+extern space_map_ops_t *zfs_metaslab_ops;
+
 extern metaslab_t *metaslab_init(metaslab_group_t *mg, space_map_obj_t *smo,
     uint64_t start, uint64_t size, uint64_t txg);
 extern void metaslab_fini(metaslab_t *msp);
@@ -55,7 +57,7 @@ extern void metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg,
     boolean_t now);
 extern int metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg);
 
-extern metaslab_class_t *metaslab_class_create(void);
+extern metaslab_class_t *metaslab_class_create(space_map_ops_t *ops);
 extern void metaslab_class_destroy(metaslab_class_t *mc);
 extern void metaslab_class_add(metaslab_class_t *mc, metaslab_group_t *mg);
 extern void metaslab_class_remove(metaslab_class_t *mc, metaslab_group_t *mg);

module/zfs/include/sys/metaslab_impl.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef _SYS_METASLAB_IMPL_H
 #define	_SYS_METASLAB_IMPL_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/metaslab.h>
 #include <sys/space_map.h>
 #include <sys/vdev.h>
@@ -41,6 +39,7 @@ extern "C" {
 struct metaslab_class {
 	metaslab_group_t	*mc_rotor;
 	uint64_t		mc_allocated;
+	space_map_ops_t		*mc_ops;
 };
 
 struct metaslab_group {

module/zfs/include/sys/spa.h

@@ -324,12 +324,9 @@ extern int spa_get_stats(const char *pool, nvlist_t **config,
     char *altroot, size_t buflen);
 extern int spa_create(const char *pool, nvlist_t *config, nvlist_t *props,
     const char *history_str, nvlist_t *zplprops);
-extern int spa_check_rootconf(char *devpath, char *devid,
-    nvlist_t **bestconf, uint64_t *besttxg);
-extern boolean_t spa_rootdev_validate(nvlist_t *nv);
 extern int spa_import_rootpool(char *devpath, char *devid);
 extern int spa_import(const char *pool, nvlist_t *config, nvlist_t *props);
-extern int spa_import_faulted(const char *, nvlist_t *, nvlist_t *);
+extern int spa_import_verbatim(const char *, nvlist_t *, nvlist_t *);
 extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
 extern int spa_destroy(char *pool);
 extern int spa_export(char *pool, nvlist_t **oldconfig, boolean_t force,
@@ -347,6 +344,7 @@ extern void spa_inject_delref(spa_t *spa);
 #define	SPA_ASYNC_PROBE		0x04
 #define	SPA_ASYNC_RESILVER_DONE	0x08
 #define	SPA_ASYNC_RESILVER	0x10
+#define	SPA_ASYNC_AUTOEXPAND	0x20
 
 /* device manipulation */
 extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
@@ -356,6 +354,7 @@ extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
     int replace_done);
 extern int spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare);
 extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
+extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
 
 /* spare state (which is global across all pools) */
 extern void spa_spare_add(vdev_t *vd);

module/zfs/include/sys/spa_boot.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef _SYS_SPA_BOOT_H
 #define	_SYS_SPA_BOOT_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/nvpair.h>
 
 #ifdef	__cplusplus
@@ -36,7 +34,6 @@ extern "C" {
 
 extern char *spa_get_bootprop(char *prop);
 extern void spa_free_bootprop(char *prop);
-extern int spa_get_rootconf(char *devpath, char *devid, nvlist_t **bestconf_p);
 
 #ifdef	__cplusplus
 }

module/zfs/include/sys/spa_impl.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -141,9 +141,6 @@ struct spa {
 	int		spa_async_suspended;	/* async tasks suspended */
 	kcondvar_t	spa_async_cv;		/* wait for thread_exit() */
 	uint16_t	spa_async_tasks;	/* async task mask */
-	kmutex_t	spa_async_root_lock;	/* protects async root count */
-	uint64_t	spa_async_root_count;	/* number of async root zios */
-	kcondvar_t	spa_async_root_cv;	/* notify when count == 0 */
 	char		*spa_root;		/* alternate root directory */
 	uint64_t	spa_ena;		/* spa-wide ereport ENA */
 	boolean_t	spa_last_open_failed;	/* true if last open faled */
@@ -163,15 +160,16 @@ struct spa {
 	uint64_t	spa_failmode;		/* failure mode for the pool */
 	uint64_t	spa_delegation;		/* delegation on/off */
 	list_t		spa_config_list;	/* previous cache file(s) */
+	zio_t		*spa_async_zio_root;	/* root of all async I/O */
 	zio_t		*spa_suspend_zio_root;	/* root of all suspended I/O */
 	kmutex_t	spa_suspend_lock;	/* protects suspend_zio_root */
 	kcondvar_t	spa_suspend_cv;		/* notification of resume */
 	uint8_t		spa_suspended;		/* pool is suspended */
-	boolean_t	spa_import_faulted;	/* allow faulted vdevs */
 	boolean_t	spa_is_root;		/* pool is root */
 	int		spa_minref;		/* num refs when first opened */
 	int		spa_mode;		/* FREAD | FWRITE */
 	spa_log_state_t spa_log_state;		/* log state */
+	uint64_t	spa_autoexpand;		/* lun expansion on/off */
 	/*
 	 * spa_refcnt & spa_config_lock must be the last elements
 	 * because refcount_t changes size based on compilation options.

module/zfs/include/sys/space_map.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -46,12 +46,14 @@ typedef struct space_map {
 	uint8_t		sm_loading;	/* map loading? */
 	kcondvar_t	sm_load_cv;	/* map load completion */
 	space_map_ops_t	*sm_ops;	/* space map block picker ops vector */
+	avl_tree_t	*sm_pp_root;	/* picker-private AVL tree */
 	void		*sm_ppd;	/* picker-private data */
 	kmutex_t	*sm_lock;	/* pointer to lock that protects map */
 } space_map_t;
 
 typedef struct space_seg {
 	avl_node_t	ss_node;	/* AVL node */
+	avl_node_t	ss_pp_node;	/* AVL picker-private node */
 	uint64_t	ss_start;	/* starting offset of this segment */
 	uint64_t	ss_end;		/* ending offset (non-inclusive) */
 } space_seg_t;
@@ -74,6 +76,7 @@ struct space_map_ops {
 	uint64_t (*smop_alloc)(space_map_t *sm, uint64_t size);
 	void	(*smop_claim)(space_map_t *sm, uint64_t start, uint64_t size);
 	void	(*smop_free)(space_map_t *sm, uint64_t start, uint64_t size);
+	uint64_t (*smop_max)(space_map_t *sm);
 };
 
 /*
@@ -152,6 +155,7 @@ extern void space_map_unload(space_map_t *sm);
 extern uint64_t space_map_alloc(space_map_t *sm, uint64_t size);
 extern void space_map_claim(space_map_t *sm, uint64_t start, uint64_t size);
 extern void space_map_free(space_map_t *sm, uint64_t start, uint64_t size);
+extern uint64_t space_map_maxsize(space_map_t *sm);
 
 extern void space_map_sync(space_map_t *sm, uint8_t maptype,
     space_map_obj_t *smo, objset_t *os, dmu_tx_t *tx);

module/zfs/include/sys/vdev.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -50,7 +50,6 @@ extern int vdev_open(vdev_t *);
 extern int vdev_validate(vdev_t *);
 extern void vdev_close(vdev_t *);
 extern int vdev_create(vdev_t *, uint64_t txg, boolean_t isreplace);
-extern void vdev_init(vdev_t *, uint64_t txg);
 extern void vdev_reopen(vdev_t *);
 extern int vdev_validate_aux(vdev_t *vd);
 extern zio_t *vdev_probe(vdev_t *vd, zio_t *pio);
@@ -71,6 +70,8 @@ extern boolean_t vdev_resilver_needed(vdev_t *vd,
 
 extern int vdev_metaslab_init(vdev_t *vd, uint64_t txg);
 extern void vdev_metaslab_fini(vdev_t *vd);
+extern void vdev_metaslab_set_size(vdev_t *);
+extern void vdev_expand(vdev_t *vd, uint64_t txg);
 
 extern void vdev_get_stats(vdev_t *vd, vdev_stat_t *vs);
 extern void vdev_clear_stats(vdev_t *vd);
@@ -113,7 +114,8 @@ extern void vdev_queue_io_done(zio_t *zio);
 
 extern void vdev_config_dirty(vdev_t *vd);
 extern void vdev_config_clean(vdev_t *vd);
-extern int vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg);
+extern int vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg,
+    boolean_t);
 
 extern void vdev_state_dirty(vdev_t *vd);
 extern void vdev_state_clean(vdev_t *vd);

module/zfs/include/sys/vdev_impl.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -113,6 +113,7 @@ struct vdev {
 	uint64_t	vdev_guid;	/* unique ID for this vdev	*/
 	uint64_t	vdev_guid_sum;	/* self guid + all child guids	*/
 	uint64_t	vdev_asize;	/* allocatable device capacity	*/
+	uint64_t	vdev_min_asize;	/* min acceptable asize		*/
 	uint64_t	vdev_ashift;	/* block alignment shift	*/
 	uint64_t	vdev_state;	/* see VDEV_STATE_* #defines	*/
 	uint64_t	vdev_prevstate;	/* used when reopening a vdev	*/
@@ -125,6 +126,7 @@ struct vdev {
 	uint64_t	vdev_children;	/* number of children		*/
 	space_map_t	vdev_dtl[DTL_TYPES]; /* in-core dirty time logs	*/
 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
+	boolean_t	vdev_expanding;	/* expand the vdev?		*/
 
 	/*
 	 * Top-level vdev state.
@@ -159,6 +161,7 @@ struct vdev {
 	char		*vdev_path;	/* vdev path (if any)		*/
 	char		*vdev_devid;	/* vdev devid (if any)		*/
 	char		*vdev_physpath;	/* vdev device path (if any)	*/
+	char		*vdev_fru;	/* physical FRU location	*/
 	uint64_t	vdev_not_present; /* not present during import	*/
 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
 	hrtime_t	vdev_last_try;	/* last reopen time		*/
@@ -188,8 +191,9 @@ struct vdev {
 	kmutex_t	vdev_probe_lock; /* protects vdev_probe_zio	*/
 };
 
-#define	VDEV_SKIP_SIZE		(8 << 10)
+#define	VDEV_PAD_SIZE		(8 << 10)
-#define	VDEV_BOOT_HEADER_SIZE	(8 << 10)
+/* 2 padding areas (vl_pad1 and vl_pad2) to skip */
+#define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
 #define	VDEV_PHYS_SIZE		(112 << 10)
 #define	VDEV_UBERBLOCK_RING	(128 << 10)
 
@@ -201,26 +205,14 @@ struct vdev {
 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
 
-/* ZFS boot block */
-#define	VDEV_BOOT_MAGIC		0x2f5b007b10cULL
-#define	VDEV_BOOT_VERSION	1		/* version number	*/
-
-typedef struct vdev_boot_header {
-	uint64_t	vb_magic;		/* VDEV_BOOT_MAGIC	*/
-	uint64_t	vb_version;		/* VDEV_BOOT_VERSION	*/
-	uint64_t	vb_offset;		/* start offset	(bytes) */
-	uint64_t	vb_size;		/* size (bytes)		*/
-	char		vb_pad[VDEV_BOOT_HEADER_SIZE - 4 * sizeof (uint64_t)];
-} vdev_boot_header_t;
-
 typedef struct vdev_phys {
 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_block_tail_t)];
 	zio_block_tail_t vp_zbt;
 } vdev_phys_t;
 
 typedef struct vdev_label {
-	char		vl_pad[VDEV_SKIP_SIZE];			/*   8K	*/
+	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K */
-	vdev_boot_header_t vl_boot_header;			/*   8K	*/
+	char		vl_pad2[VDEV_PAD_SIZE];			/*  8K */
 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
 } vdev_label_t;							/* 256K total */
@@ -249,6 +241,7 @@ typedef struct vdev_label {
 #define	VDEV_ALLOC_ADD		1
 #define	VDEV_ALLOC_SPARE	2
 #define	VDEV_ALLOC_L2CACHE	3
+#define	VDEV_ALLOC_ROOTPOOL	4
 
 /*
  * Allocate or free a vdev
@@ -269,6 +262,7 @@ extern void vdev_remove_parent(vdev_t *cvd);
 /*
  * vdev sync load and sync
  */
+extern void vdev_load_log_state(vdev_t *vd, nvlist_t *nv);
 extern void vdev_load(vdev_t *vd);
 extern void vdev_sync(vdev_t *vd, uint64_t txg);
 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
@@ -290,7 +284,8 @@ extern vdev_ops_t vdev_spare_ops;
  * Common size functions
  */
 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
-extern uint64_t vdev_get_rsize(vdev_t *vd);
+extern uint64_t vdev_get_min_asize(vdev_t *vd);
+extern void vdev_set_min_asize(vdev_t *vd);
 
 /*
  * zdb uses this tunable, so it must be declared here to make lint happy.

module/zfs/include/sys/zap.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_ZAP_H
 #define	_SYS_ZAP_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * ZAP - ZFS Attribute Processor
  *
@@ -87,9 +85,6 @@
 extern "C" {
 #endif
 
-#define	ZAP_MAXNAMELEN 256
-#define	ZAP_MAXVALUELEN 1024
-
 /*
  * The matchtype specifies which entry will be accessed.
  * MT_EXACT: only find an exact match (non-normalized)
@@ -186,6 +181,10 @@ int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
     matchtype_t mt, char *realname, int rn_len,
     boolean_t *normalization_conflictp);
 
+int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
+    int add, uint64_t *towrite, uint64_t *tooverwrite,
+    uint64_t dn_datablkshift);
+
 /*
  * Create an attribute with the given name and value.
  *

module/zfs/include/sys/zap_impl.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_ZAP_IMPL_H
 #define	_SYS_ZAP_IMPL_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/zap.h>
 #include <sys/zfs_context.h>
 #include <sys/avl.h>
@@ -195,6 +193,8 @@ int fzap_count(zap_t *zap, uint64_t *count);
 int fzap_lookup(zap_name_t *zn,
     uint64_t integer_size, uint64_t num_integers, void *buf,
     char *realname, int rn_len, boolean_t *normalization_conflictp);
+int fzap_count_write(zap_name_t *zn, int add, uint64_t *towrite,
+    uint64_t *tooverwrite);
 int fzap_add(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers,
     const void *val, dmu_tx_t *tx);
 int fzap_update(zap_name_t *zn,

module/zfs/include/sys/zfs_acl.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -114,8 +114,6 @@ typedef struct zfs_acl_phys {
 	uint8_t		z_ace_data[ZFS_ACE_SPACE]; /* space for embedded ACEs */
 } zfs_acl_phys_t;
 
-
-
 typedef struct acl_ops {
 	uint32_t	(*ace_mask_get) (void *acep); /* get  access mask */
 	void 		(*ace_mask_set) (void *acep,
@@ -161,12 +159,21 @@ typedef struct zfs_acl {
 	zfs_acl_node_t	*z_curr_node;	/* current node iterator is handling */
 	list_t		z_acl;		/* chunks of ACE data */
 	acl_ops_t	z_ops;		/* ACL operations */
-	boolean_t	z_has_fuids;	/* FUIDs present in ACL? */
 } zfs_acl_t;
 
 #define	ACL_DATA_ALLOCED	0x1
 #define	ZFS_ACL_SIZE(aclcnt)	(sizeof (ace_t) * (aclcnt))
 
+struct zfs_fuid_info;
+
+typedef struct zfs_acl_ids {
+	uint64_t		z_fuid;		/* file owner fuid */
+	uint64_t		z_fgid;		/* file group owner fuid */
+	uint64_t		z_mode;		/* mode to set on create */
+	zfs_acl_t		*z_aclp;	/* ACL to create with file */
+	struct zfs_fuid_info 	*z_fuidp;	/* for tracking fuids for log */
+} zfs_acl_ids_t;
+
 /*
  * Property values for acl_mode and acl_inherit.
  *
@@ -183,16 +190,18 @@ typedef struct zfs_acl {
 
 struct znode;
 struct zfsvfs;
-struct zfs_fuid_info;
 
 #ifdef _KERNEL
-void zfs_perm_init(struct znode *, struct znode *, int, vattr_t *,
+int zfs_acl_ids_create(struct znode *, int, vattr_t *,
-    dmu_tx_t *, cred_t *, zfs_acl_t *, zfs_fuid_info_t **);
+    cred_t *, vsecattr_t *, zfs_acl_ids_t *);
+void zfs_acl_ids_free(zfs_acl_ids_t *);
+boolean_t zfs_acl_ids_overquota(struct zfsvfs *, zfs_acl_ids_t *);
 int zfs_getacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
 int zfs_setacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
 void zfs_acl_rele(void *);
 void zfs_oldace_byteswap(ace_t *, int);
 void zfs_ace_byteswap(void *, size_t, boolean_t);
+extern boolean_t zfs_has_access(struct znode *zp, cred_t *cr);
 extern int zfs_zaccess(struct znode *, int, int, boolean_t, cred_t *);
 extern int zfs_zaccess_rwx(struct znode *, mode_t, int, cred_t *);
 extern int zfs_zaccess_unix(struct znode *, mode_t, cred_t *);
@@ -202,9 +211,9 @@ int zfs_zaccess_delete(struct znode *, struct znode *, cred_t *);
 int zfs_zaccess_rename(struct znode *, struct znode *,
     struct znode *, struct znode *, cred_t *cr);
 void zfs_acl_free(zfs_acl_t *);
-int zfs_vsec_2_aclp(struct zfsvfs *, vtype_t, vsecattr_t *, zfs_acl_t **);
+int zfs_vsec_2_aclp(struct zfsvfs *, vtype_t, vsecattr_t *, cred_t *,
-int zfs_aclset_common(struct znode *, zfs_acl_t *, cred_t *,
+    struct zfs_fuid_info **, zfs_acl_t **);
-    struct zfs_fuid_info **, dmu_tx_t *);
+int zfs_aclset_common(struct znode *, zfs_acl_t *, cred_t *, dmu_tx_t *);
 
 #endif
 

module/zfs/include/sys/zfs_context.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef _SYS_ZFS_CONTEXT_H
 #define	_SYS_ZFS_CONTEXT_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #ifdef	__cplusplus
 extern "C" {
 #endif
@@ -62,6 +60,7 @@ extern "C" {
 #include <sys/zfs_debug.h>
 #include <sys/sysevent.h>
 #include <sys/sysevent/eventdefs.h>
+#include <sys/sysevent/dev.h>
 #include <sys/fm/util.h>
 
 #define	CPU_SEQID	(CPU->cpu_seqid)

module/zfs/include/sys/zfs_ctldir.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_ZFS_CTLDIR_H
 #define	_ZFS_CTLDIR_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/pathname.h>
 #include <sys/vnode.h>
 #include <sys/zfs_vfsops.h>
@@ -66,6 +64,7 @@ int zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp);
 
 #define	ZFSCTL_INO_ROOT		0x1
 #define	ZFSCTL_INO_SNAPDIR	0x2
+#define	ZFSCTL_INO_SHARES	0x3
 
 #ifdef	__cplusplus
 }

module/zfs/include/sys/zfs_dir.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_FS_ZFS_DIR_H
 #define	_SYS_FS_ZFS_DIR_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/pathname.h>
 #include <sys/dmu.h>
 #include <sys/zfs_znode.h>
@@ -59,7 +57,7 @@ extern int zfs_link_destroy(zfs_dirlock_t *, znode_t *, dmu_tx_t *, int,
 extern int zfs_dirlook(znode_t *, char *, vnode_t **, int, int *,
     pathname_t *);
 extern void zfs_mknode(znode_t *, vattr_t *, dmu_tx_t *, cred_t *,
-    uint_t, znode_t **, int, zfs_acl_t *, zfs_fuid_info_t **);
+    uint_t, znode_t **, int, zfs_acl_ids_t *);
 extern void zfs_rmnode(znode_t *);
 extern void zfs_dl_name_switch(zfs_dirlock_t *dl, char *new, char **old);
 extern boolean_t zfs_dirempty(znode_t *);

module/zfs/include/sys/zfs_fuid.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_FS_ZFS_FUID_H
 #define	_SYS_FS_ZFS_FUID_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #ifdef _KERNEL
 #include <sys/kidmap.h>
 #include <sys/sid.h>
@@ -51,11 +49,11 @@ typedef enum {
  * Estimate space needed for one more fuid table entry.
  * for now assume its current size + 1K
  */
-#define	FUID_SIZE_ESTIMATE(z) (z->z_fuid_size + (SPA_MINBLOCKSIZE << 1))
+#define	FUID_SIZE_ESTIMATE(z) ((z)->z_fuid_size + (SPA_MINBLOCKSIZE << 1))
 
-#define	FUID_INDEX(x)	(x >> 32)
+#define	FUID_INDEX(x)	((x) >> 32)
-#define	FUID_RID(x)	(x & 0xffffffff)
+#define	FUID_RID(x)	((x) & 0xffffffff)
-#define	FUID_ENCODE(idx, rid) ((idx << 32) | rid)
+#define	FUID_ENCODE(idx, rid) (((uint64_t)(idx) << 32) | (rid))
 /*
  * FUIDs cause problems for the intent log
  * we need to replay the creation of the FUID,
@@ -104,17 +102,23 @@ struct znode;
 extern uid_t zfs_fuid_map_id(zfsvfs_t *, uint64_t, cred_t *, zfs_fuid_type_t);
 extern void zfs_fuid_destroy(zfsvfs_t *);
 extern uint64_t zfs_fuid_create_cred(zfsvfs_t *, zfs_fuid_type_t,
-    dmu_tx_t *, cred_t *, zfs_fuid_info_t **);
+    cred_t *, zfs_fuid_info_t **);
 extern uint64_t zfs_fuid_create(zfsvfs_t *, uint64_t, cred_t *, zfs_fuid_type_t,
-    dmu_tx_t *, zfs_fuid_info_t **);
+    zfs_fuid_info_t **);
-extern void zfs_fuid_map_ids(struct znode *zp, cred_t *cr, uid_t *uid,
+extern void zfs_fuid_map_ids(struct znode *zp, cred_t *cr,
-    uid_t *gid);
+    uid_t *uid, uid_t *gid);
 extern zfs_fuid_info_t *zfs_fuid_info_alloc(void);
-extern void zfs_fuid_info_free();
+extern void zfs_fuid_info_free(zfs_fuid_info_t *);
 extern boolean_t zfs_groupmember(zfsvfs_t *, uint64_t, cred_t *);
+void zfs_fuid_sync(zfsvfs_t *, dmu_tx_t *);
+extern int zfs_fuid_find_by_domain(zfsvfs_t *, const char *domain,
+    char **retdomain, boolean_t addok);
+extern const char *zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx);
+extern void zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
 #endif
 
 char *zfs_fuid_idx_domain(avl_tree_t *, uint32_t);
+void zfs_fuid_avl_tree_create(avl_tree_t *, avl_tree_t *);
 uint64_t zfs_fuid_table_load(objset_t *, uint64_t, avl_tree_t *, avl_tree_t *);
 void zfs_fuid_table_destroy(avl_tree_t *, avl_tree_t *);
 

module/zfs/include/sys/zfs_ioctl.h

@@ -19,15 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #ifndef	_SYS_ZFS_IOCTL_H
 #define	_SYS_ZFS_IOCTL_H
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/cred.h>
 #include <sys/dmu.h>
 #include <sys/zio.h>
@@ -118,7 +116,7 @@ typedef struct zinject_record {
 	uint32_t	zi_error;
 	uint64_t	zi_type;
 	uint32_t	zi_freq;
-	uint32_t	zi_pad;	/* pad out to 64 bit alignment */
+	uint32_t	zi_failfast;
 } zinject_record_t;
 
 #define	ZINJECT_NULL		0x1
@@ -162,12 +160,20 @@ typedef struct zfs_cmd {
 	uint64_t 	zc_history_len;
 	uint64_t	zc_history_offset;
 	uint64_t	zc_obj;
+	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;
 	zinject_record_t zc_inject_record;
 } zfs_cmd_t;
 
+typedef struct zfs_useracct {
+	char zu_domain[256];
+	uid_t zu_rid;
+	uint32_t zu_pad;
+	uint64_t zu_space;
+} zfs_useracct_t;
+
 #define	ZVOL_MAX_MINOR	(1 << 16)
 #define	ZFS_MIN_MINOR	(ZVOL_MAX_MINOR + 1)
 

module/zfs/include/sys/zfs_vfsops.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -53,6 +53,7 @@ struct zfsvfs {
 	avl_tree_t	z_fuid_domain;	/* fuid tree keyed by domain */
 	krwlock_t	z_fuid_lock;	/* fuid lock */
 	boolean_t	z_fuid_loaded;	/* fuid tables are loaded */
+	boolean_t	z_fuid_dirty;   /* need to sync fuid table ? */
 	struct zfs_fuid_info	*z_fuid_replay; /* fuid info for replay */
 	zilog_t		*z_log;		/* intent log pointer */
 	uint_t		z_acl_mode;	/* acl chmod/mode behavior */
@@ -72,8 +73,12 @@ struct zfsvfs {
 	boolean_t	z_vscan;	/* virus scan on/off */
 	boolean_t	z_use_fuids;	/* version allows fuids */
 	boolean_t	z_replay;	/* set during ZIL replay */
-	kmutex_t	z_online_recv_lock; /* recv in prog grabs as WRITER */
+	kmutex_t	z_online_recv_lock; /* held while recv in progress */
 	uint64_t	z_version;	/* ZPL version */
+	uint64_t	z_shares_dir;	/* hidden shares dir */
+	kmutex_t	z_lock;
+	uint64_t	z_userquota_obj;
+	uint64_t	z_groupquota_obj;
 #define	ZFS_OBJ_MTX_SZ	64
 	kmutex_t	z_hold_mtx[ZFS_OBJ_MTX_SZ];	/* znode hold locks */
 };
@@ -130,6 +135,17 @@ extern uint_t zfs_fsyncer_key;
 
 extern int zfs_suspend_fs(zfsvfs_t *zfsvfs, char *osname, int *mode);
 extern int zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode);
+extern int zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    const char *domain, uint64_t rid, uint64_t *valuep);
+extern int zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    uint64_t *cookiep, void *vbuf, uint64_t *bufsizep);
+extern int zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    const char *domain, uint64_t rid, uint64_t quota);
+extern boolean_t zfs_usergroup_overquota(zfsvfs_t *zfsvfs,
+    boolean_t isgroup, uint64_t fuid);
+extern int zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers);
+extern int zfsvfs_create(const char *name, int mode, zfsvfs_t **zvp);
+extern void zfsvfs_free(zfsvfs_t *zfsvfs);
 
 #ifdef	__cplusplus
 }

module/zfs/include/sys/zfs_znode.h

@@ -93,12 +93,15 @@ extern "C" {
 
 /*
  * Special attributes for master node.
+ * "userquota@" and "groupquota@" are also valid (from
+ * zfs_userquota_prop_prefixes[]).
  */
 #define	ZFS_FSID		"FSID"
 #define	ZFS_UNLINKED_SET	"DELETE_QUEUE"
 #define	ZFS_ROOT_OBJ		"ROOT"
 #define	ZPL_VERSION_STR		"VERSION"
 #define	ZFS_FUID_TABLES		"FUID"
+#define	ZFS_SHARES_DIR		"SHARES"
 
 #define	ZFS_MAX_BLOCKSIZE	(SPA_MAXBLOCKSIZE)
 
@@ -309,7 +312,6 @@ extern int	zfs_create_op_tables();
 extern int	zfs_sync(vfs_t *vfsp, short flag, cred_t *cr);
 extern dev_t	zfs_cmpldev(uint64_t);
 extern int	zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value);
-extern int	zfs_set_version(const char *name, uint64_t newvers);
 extern int	zfs_get_stats(objset_t *os, nvlist_t *nv);
 extern void	zfs_znode_dmu_fini(znode_t *);
 
@@ -336,6 +338,7 @@ extern void zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp);
 extern void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap);
 extern void zfs_upgrade(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
+extern int zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
 
 extern caddr_t zfs_map_page(page_t *, enum seg_rw);
 extern void zfs_unmap_page(page_t *, caddr_t);

module/zfs/include/sys/zil.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -56,9 +56,15 @@ typedef struct zil_header {
 	uint64_t zh_replay_seq;	/* highest replayed sequence number */
 	blkptr_t zh_log;	/* log chain */
 	uint64_t zh_claim_seq;	/* highest claimed sequence number */
-	uint64_t zh_pad[5];
+	uint64_t zh_flags;	/* header flags */
+	uint64_t zh_pad[4];
 } zil_header_t;
 
+/*
+ * zh_flags bit settings
+ */
+#define	ZIL_REPLAY_NEEDED 0x1	/* replay needed - internal only */
+
 /*
  * Log block trailer - structure at the end of the header and each log block
  *
@@ -299,7 +305,27 @@ typedef struct {
  */
 
 /*
- * ZFS intent log transaction structure
+ * Writes are handled in three different ways:
+ *
+ * WR_INDIRECT:
+ *    In this mode, if we need to commit the write later, then the block
+ *    is immediately written into the file system (using dmu_sync),
+ *    and a pointer to the block is put into the log record.
+ *    When the txg commits the block is linked in.
+ *    This saves additionally writing the data into the log record.
+ *    There are a few requirements for this to occur:
+ *	- write is greater than zfs/zvol_immediate_write_sz
+ *	- not using slogs (as slogs are assumed to always be faster
+ *	  than writing into the main pool)
+ *	- the write occupies only one block
+ * WR_COPIED:
+ *    If we know we'll immediately be committing the
+ *    transaction (FSYNC or FDSYNC), the we allocate a larger
+ *    log record here for the data and copy the data in.
+ * WR_NEED_COPY:
+ *    Otherwise we don't allocate a buffer, and *if* we need to
+ *    flush the write later then a buffer is allocated and
+ *    we retrieve the data using the dmu.
  */
 typedef enum {
 	WR_INDIRECT,	/* indirect - a large write (dmu_sync() data */
@@ -359,9 +385,9 @@ extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx);
 
 extern void	zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid);
 
+extern int	zil_vdev_offline(char *osname, void *txarg);
 extern int	zil_claim(char *osname, void *txarg);
 extern int	zil_check_log_chain(char *osname, void *txarg);
-extern int	zil_clear_log_chain(char *osname, void *txarg);
 extern void	zil_sync(zilog_t *zilog, dmu_tx_t *tx);
 extern void	zil_clean(zilog_t *zilog);
 extern int	zil_is_committed(zilog_t *zilog);

module/zfs/include/sys/zil_impl.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -101,6 +101,9 @@ typedef struct zil_dva_node {
 	avl_node_t	zn_node;
 } zil_dva_node_t;
 
+#define	ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \
+    sizeof (lr_write_t))
+
 #ifdef	__cplusplus
 }
 #endif

module/zfs/include/sys/zio.h

@@ -76,7 +76,7 @@ enum zio_checksum {
 	ZIO_CHECKSUM_FUNCTIONS
 };
 
-#define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_2
+#define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_4
 #define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
 
 enum zio_compress {
@@ -116,30 +116,33 @@ enum zio_compress {
 #define	ZIO_PRIORITY_SCRUB		(zio_priority_table[9])
 #define	ZIO_PRIORITY_TABLE_SIZE		10
 
-#define	ZIO_FLAG_MUSTSUCCEED		0x00000
+#define	ZIO_FLAG_MUSTSUCCEED		0x000000
-#define	ZIO_FLAG_CANFAIL		0x00001
+#define	ZIO_FLAG_CANFAIL		0x000001
-#define	ZIO_FLAG_SPECULATIVE		0x00002
+#define	ZIO_FLAG_SPECULATIVE		0x000002
-#define	ZIO_FLAG_CONFIG_WRITER		0x00004
+#define	ZIO_FLAG_CONFIG_WRITER		0x000004
-#define	ZIO_FLAG_DONT_RETRY		0x00008
+#define	ZIO_FLAG_DONT_RETRY		0x000008
 
-#define	ZIO_FLAG_DONT_CACHE		0x00010
+#define	ZIO_FLAG_DONT_CACHE		0x000010
-#define	ZIO_FLAG_DONT_QUEUE		0x00020
+#define	ZIO_FLAG_DONT_QUEUE		0x000020
-#define	ZIO_FLAG_DONT_AGGREGATE		0x00040
+#define	ZIO_FLAG_DONT_AGGREGATE		0x000040
-#define	ZIO_FLAG_DONT_PROPAGATE		0x00080
+#define	ZIO_FLAG_DONT_PROPAGATE		0x000080
 
-#define	ZIO_FLAG_IO_BYPASS		0x00100
+#define	ZIO_FLAG_IO_BYPASS		0x000100
-#define	ZIO_FLAG_IO_REPAIR		0x00200
+#define	ZIO_FLAG_IO_REPAIR		0x000200
-#define	ZIO_FLAG_IO_RETRY		0x00400
+#define	ZIO_FLAG_IO_RETRY		0x000400
-#define	ZIO_FLAG_IO_REWRITE		0x00800
+#define	ZIO_FLAG_IO_REWRITE		0x000800
 
-#define	ZIO_FLAG_SELF_HEAL		0x01000
+#define	ZIO_FLAG_SELF_HEAL		0x001000
-#define	ZIO_FLAG_RESILVER		0x02000
+#define	ZIO_FLAG_RESILVER		0x002000
-#define	ZIO_FLAG_SCRUB			0x04000
+#define	ZIO_FLAG_SCRUB			0x004000
-#define	ZIO_FLAG_SCRUB_THREAD		0x08000
+#define	ZIO_FLAG_SCRUB_THREAD		0x008000
 
-#define	ZIO_FLAG_PROBE			0x10000
+#define	ZIO_FLAG_PROBE			0x010000
-#define	ZIO_FLAG_GANG_CHILD		0x20000
+#define	ZIO_FLAG_GANG_CHILD		0x020000
-#define	ZIO_FLAG_RAW			0x40000
+#define	ZIO_FLAG_RAW			0x040000
+#define	ZIO_FLAG_GODFATHER		0x080000
+
+#define	ZIO_FLAG_TRYHARD		0x100000
 
 #define	ZIO_FLAG_GANG_INHERIT		\
 	(ZIO_FLAG_CANFAIL |		\
@@ -157,7 +160,8 @@ enum zio_compress {
 	(ZIO_FLAG_GANG_INHERIT |	\
 	ZIO_FLAG_IO_REPAIR |		\
 	ZIO_FLAG_IO_RETRY |		\
-	ZIO_FLAG_PROBE)
+	ZIO_FLAG_PROBE |		\
+	ZIO_FLAG_TRYHARD)
 
 #define	ZIO_FLAG_AGG_INHERIT		\
 	(ZIO_FLAG_DONT_AGGREGATE |	\
@@ -281,7 +285,6 @@ struct zio {
 	int		io_cmd;
 	uint8_t		io_priority;
 	uint8_t		io_reexecute;
-	uint8_t		io_async_root;
 	uint8_t		io_state[ZIO_WAIT_TYPES];
 	uint64_t	io_txg;
 	spa_t		*io_spa;
@@ -324,6 +327,7 @@ struct zio {
 	int		io_child_error[ZIO_CHILD_TYPES];
 	uint64_t	io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
 	uint64_t	*io_stall;
+	zio_t		*io_gang_leader;
 	zio_gang_node_t	*io_gang_tree;
 	void		*io_executor;
 	void		*io_waiter;
@@ -415,7 +419,7 @@ extern uint8_t zio_checksum_select(uint8_t child, uint8_t parent);
 extern uint8_t zio_compress_select(uint8_t child, uint8_t parent);
 
 extern void zio_suspend(spa_t *spa, zio_t *zio);
-extern void zio_resume(spa_t *spa);
+extern int zio_resume(spa_t *spa);
 extern void zio_resume_wait(spa_t *spa);
 
 /*
@@ -435,7 +439,7 @@ extern int zio_inject_list_next(int *id, char *name, size_t buflen,
     struct zinject_record *record);
 extern int zio_clear_fault(int id);
 extern int zio_handle_fault_injection(zio_t *zio, int error);
-extern int zio_handle_device_injection(vdev_t *vd, int error);
+extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
 extern int zio_handle_label_injection(zio_t *zio, int error);
 
 #ifdef	__cplusplus

module/zfs/metaslab.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -35,19 +35,36 @@
 uint64_t metaslab_aliquot = 512ULL << 10;
 uint64_t metaslab_gang_bang = SPA_MAXBLOCKSIZE + 1;	/* force gang blocks */
 
+/*
+ * Minimum size which forces the dynamic allocator to change
+ * it's allocation strategy. Once the space map cannot satisfy
+ * an allocation of this size then it switches to using more
+ * aggressive strategy (i.e search by size rather than offset).
+ */
+uint64_t metaslab_df_alloc_threshold = SPA_MAXBLOCKSIZE;
+
+/*
+ * The minimum free space, in percent, which must be available
+ * in a space map to continue allocations in a first-fit fashion.
+ * Once the space_map's free space drops below this level we dynamically
+ * switch to using best-fit allocations.
+ */
+int metaslab_df_free_pct = 30;
+
 /*
  * ==========================================================================
  * Metaslab classes
  * ==========================================================================
  */
 metaslab_class_t *
-metaslab_class_create(void)
+metaslab_class_create(space_map_ops_t *ops)
 {
 	metaslab_class_t *mc;
 
 	mc = kmem_zalloc(sizeof (metaslab_class_t), KM_SLEEP);
 
 	mc->mc_rotor = NULL;
+	mc->mc_ops = ops;
 
 	return (mc);
 }
@@ -202,30 +219,14 @@ metaslab_group_sort(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
 }
 
 /*
- * ==========================================================================
+ * This is a helper function that can be used by the allocator to find
- * The first-fit block allocator
+ * a suitable block to allocate. This will search the specified AVL
- * ==========================================================================
+ * tree looking for a block that matches the specified criteria.
  */
-static void
-metaslab_ff_load(space_map_t *sm)
-{
-	ASSERT(sm->sm_ppd == NULL);
-	sm->sm_ppd = kmem_zalloc(64 * sizeof (uint64_t), KM_SLEEP);
-}
-
-static void
-metaslab_ff_unload(space_map_t *sm)
-{
-	kmem_free(sm->sm_ppd, 64 * sizeof (uint64_t));
-	sm->sm_ppd = NULL;
-}
-
 static uint64_t
-metaslab_ff_alloc(space_map_t *sm, uint64_t size)
+metaslab_block_picker(avl_tree_t *t, uint64_t *cursor, uint64_t size,
+    uint64_t align)
 {
-	avl_tree_t *t = &sm->sm_root;
-	uint64_t align = size & -size;
-	uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;
 	space_seg_t *ss, ssearch;
 	avl_index_t where;
 
@@ -254,7 +255,37 @@ metaslab_ff_alloc(space_map_t *sm, uint64_t size)
 		return (-1ULL);
 
 	*cursor = 0;
-	return (metaslab_ff_alloc(sm, size));
+	return (metaslab_block_picker(t, cursor, size, align));
+}
+
+/*
+ * ==========================================================================
+ * The first-fit block allocator
+ * ==========================================================================
+ */
+static void
+metaslab_ff_load(space_map_t *sm)
+{
+	ASSERT(sm->sm_ppd == NULL);
+	sm->sm_ppd = kmem_zalloc(64 * sizeof (uint64_t), KM_SLEEP);
+	sm->sm_pp_root = NULL;
+}
+
+static void
+metaslab_ff_unload(space_map_t *sm)
+{
+	kmem_free(sm->sm_ppd, 64 * sizeof (uint64_t));
+	sm->sm_ppd = NULL;
+}
+
+static uint64_t
+metaslab_ff_alloc(space_map_t *sm, uint64_t size)
+{
+	avl_tree_t *t = &sm->sm_root;
+	uint64_t align = size & -size;
+	uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;
+
+	return (metaslab_block_picker(t, cursor, size, align));
 }
 
 /* ARGSUSED */
@@ -276,9 +307,136 @@ static space_map_ops_t metaslab_ff_ops = {
 	metaslab_ff_unload,
 	metaslab_ff_alloc,
 	metaslab_ff_claim,
-	metaslab_ff_free
+	metaslab_ff_free,
+	NULL	/* maxsize */
 };
 
+/*
+ * Dynamic block allocator -
+ * Uses the first fit allocation scheme until space get low and then
+ * adjusts to a best fit allocation method. Uses metaslab_df_alloc_threshold
+ * and metaslab_df_free_pct to determine when to switch the allocation scheme.
+ */
+
+uint64_t
+metaslab_df_maxsize(space_map_t *sm)
+{
+	avl_tree_t *t = sm->sm_pp_root;
+	space_seg_t *ss;
+
+	if (t == NULL || (ss = avl_last(t)) == NULL)
+		return (0ULL);
+
+	return (ss->ss_end - ss->ss_start);
+}
+
+static int
+metaslab_df_seg_compare(const void *x1, const void *x2)
+{
+	const space_seg_t *s1 = x1;
+	const space_seg_t *s2 = x2;
+	uint64_t ss_size1 = s1->ss_end - s1->ss_start;
+	uint64_t ss_size2 = s2->ss_end - s2->ss_start;
+
+	if (ss_size1 < ss_size2)
+		return (-1);
+	if (ss_size1 > ss_size2)
+		return (1);
+
+	if (s1->ss_start < s2->ss_start)
+		return (-1);
+	if (s1->ss_start > s2->ss_start)
+		return (1);
+
+	return (0);
+}
+
+static void
+metaslab_df_load(space_map_t *sm)
+{
+	space_seg_t *ss;
+
+	ASSERT(sm->sm_ppd == NULL);
+	sm->sm_ppd = kmem_zalloc(64 * sizeof (uint64_t), KM_SLEEP);
+
+	sm->sm_pp_root = kmem_alloc(sizeof (avl_tree_t), KM_SLEEP);
+	avl_create(sm->sm_pp_root, metaslab_df_seg_compare,
+	    sizeof (space_seg_t), offsetof(struct space_seg, ss_pp_node));
+
+	for (ss = avl_first(&sm->sm_root); ss; ss = AVL_NEXT(&sm->sm_root, ss))
+		avl_add(sm->sm_pp_root, ss);
+}
+
+static void
+metaslab_df_unload(space_map_t *sm)
+{
+	void *cookie = NULL;
+
+	kmem_free(sm->sm_ppd, 64 * sizeof (uint64_t));
+	sm->sm_ppd = NULL;
+
+	while (avl_destroy_nodes(sm->sm_pp_root, &cookie) != NULL) {
+		/* tear down the tree */
+	}
+
+	avl_destroy(sm->sm_pp_root);
+	kmem_free(sm->sm_pp_root, sizeof (avl_tree_t));
+	sm->sm_pp_root = NULL;
+}
+
+static uint64_t
+metaslab_df_alloc(space_map_t *sm, uint64_t size)
+{
+	avl_tree_t *t = &sm->sm_root;
+	uint64_t align = size & -size;
+	uint64_t *cursor = (uint64_t *)sm->sm_ppd + highbit(align) - 1;
+	uint64_t max_size = metaslab_df_maxsize(sm);
+	int free_pct = sm->sm_space * 100 / sm->sm_size;
+
+	ASSERT(MUTEX_HELD(sm->sm_lock));
+	ASSERT3U(avl_numnodes(&sm->sm_root), ==, avl_numnodes(sm->sm_pp_root));
+
+	if (max_size < size)
+		return (-1ULL);
+
+	/*
+	 * If we're running low on space switch to using the size
+	 * sorted AVL tree (best-fit).
+	 */
+	if (max_size < metaslab_df_alloc_threshold ||
+	    free_pct < metaslab_df_free_pct) {
+		t = sm->sm_pp_root;
+		*cursor = 0;
+	}
+
+	return (metaslab_block_picker(t, cursor, size, 1ULL));
+}
+
+/* ARGSUSED */
+static void
+metaslab_df_claim(space_map_t *sm, uint64_t start, uint64_t size)
+{
+	/* No need to update cursor */
+}
+
+/* ARGSUSED */
+static void
+metaslab_df_free(space_map_t *sm, uint64_t start, uint64_t size)
+{
+	/* No need to update cursor */
+}
+
+static space_map_ops_t metaslab_df_ops = {
+	metaslab_df_load,
+	metaslab_df_unload,
+	metaslab_df_alloc,
+	metaslab_df_claim,
+	metaslab_df_free,
+	metaslab_df_maxsize
+};
+
+space_map_ops_t *zfs_metaslab_ops = &metaslab_df_ops;
+
 /*
  * ==========================================================================
  * Metaslabs
@@ -414,20 +572,28 @@ metaslab_weight(metaslab_t *msp)
 }
 
 static int
-metaslab_activate(metaslab_t *msp, uint64_t activation_weight)
+metaslab_activate(metaslab_t *msp, uint64_t activation_weight, uint64_t size)
 {
 	space_map_t *sm = &msp->ms_map;
+	space_map_ops_t *sm_ops = msp->ms_group->mg_class->mc_ops;
 
 	ASSERT(MUTEX_HELD(&msp->ms_lock));
 
 	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
-		int error = space_map_load(sm, &metaslab_ff_ops,
+		int error = space_map_load(sm, sm_ops, SM_FREE, &msp->ms_smo,
-		    SM_FREE, &msp->ms_smo,
 		    msp->ms_group->mg_vd->vdev_spa->spa_meta_objset);
 		if (error) {
 			metaslab_group_sort(msp->ms_group, msp, 0);
 			return (error);
 		}
+
+		/*
+		 * If we were able to load the map then make sure
+		 * that this map is still able to satisfy our request.
+		 */
+		if (msp->ms_weight < size)
+			return (ENOSPC);
+
 		metaslab_group_sort(msp->ms_group, msp,
 		    msp->ms_weight | activation_weight);
 	}
@@ -636,11 +802,16 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
 	int i;
 
 	activation_weight = METASLAB_WEIGHT_PRIMARY;
-	for (i = 0; i < d; i++)
+	for (i = 0; i < d; i++) {
-		if (DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id)
+		if (DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
 			activation_weight = METASLAB_WEIGHT_SECONDARY;
+			break;
+		}
+	}
 
 	for (;;) {
+		boolean_t was_active;
+
 		mutex_enter(&mg->mg_lock);
 		for (msp = avl_first(t); msp; msp = AVL_NEXT(t, msp)) {
 			if (msp->ms_weight < size) {
@@ -648,6 +819,7 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
 				return (-1ULL);
 			}
 
+			was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
 			if (activation_weight == METASLAB_WEIGHT_PRIMARY)
 				break;
 
@@ -673,7 +845,9 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
 		 * another thread may have changed the weight while we
 		 * were blocked on the metaslab lock.
 		 */
-		if (msp->ms_weight < size) {
+		if (msp->ms_weight < size || (was_active &&
+		    !(msp->ms_weight & METASLAB_ACTIVE_MASK) &&
+		    activation_weight == METASLAB_WEIGHT_PRIMARY)) {
 			mutex_exit(&msp->ms_lock);
 			continue;
 		}
@@ -686,7 +860,7 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t size, uint64_t txg,
 			continue;
 		}
 
-		if (metaslab_activate(msp, activation_weight) != 0) {
+		if (metaslab_activate(msp, activation_weight, size) != 0) {
 			mutex_exit(&msp->ms_lock);
 			continue;
 		}
@@ -869,7 +1043,7 @@ next:
 		goto top;
 	}
 
-	if (!zio_lock) {
+	if (!allocatable && !zio_lock) {
 		dshift = 3;
 		zio_lock = B_TRUE;
 		goto top;
@@ -955,7 +1129,7 @@ metaslab_claim_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
 
 	mutex_enter(&msp->ms_lock);
 
-	error = metaslab_activate(msp, METASLAB_WEIGHT_SECONDARY);
+	error = metaslab_activate(msp, METASLAB_WEIGHT_SECONDARY, 0);
 	if (error || txg == 0) {	/* txg == 0 indicates dry run */
 		mutex_exit(&msp->ms_lock);
 		return (error);

module/zfs/spa_config.c

@@ -432,10 +432,9 @@ spa_config_update_common(spa_t *spa, int what, boolean_t isroot)
 		 */
 		for (c = 0; c < rvd->vdev_children; c++) {
 			vdev_t *tvd = rvd->vdev_child[c];
-			if (tvd->vdev_ms_array == 0) {
+			if (tvd->vdev_ms_array == 0)
-				vdev_init(tvd, txg);
+				vdev_metaslab_set_size(tvd);
-				vdev_config_dirty(tvd);
+			vdev_expand(tvd, txg);
-			}
 		}
 	}
 	spa_config_exit(spa, SCL_ALL, FTAG);

module/zfs/spa_errlog.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * Routines to manage the on-disk persistent error log.
  *
@@ -61,8 +59,8 @@
  * lowercase hexidecimal numbers that don't overflow.
  */
 #ifdef _KERNEL
-static uint64_t
+uint64_t
-strtonum(char *str, char **nptr)
+strtonum(const char *str, char **nptr)
 {
 	uint64_t val = 0;
 	char c;
@@ -82,7 +80,8 @@ strtonum(char *str, char **nptr)
 		str++;
 	}
 
-	*nptr = str;
+	if (nptr)
+		*nptr = (char *)str;
 
 	return (val);
 }

module/zfs/spa_history.c

@@ -20,12 +20,10 @@
  */
 
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/spa.h>
 #include <sys/spa_impl.h>
 #include <sys/zap.h>
@@ -127,12 +125,12 @@ spa_history_advance_bof(spa_t *spa, spa_history_phys_t *shpp)
 	firstread = MIN(sizeof (reclen), shpp->sh_phys_max_off - phys_bof);
 
 	if ((err = dmu_read(mos, spa->spa_history, phys_bof, firstread,
-	    buf)) != 0)
+	    buf, DMU_READ_PREFETCH)) != 0)
 		return (err);
 	if (firstread != sizeof (reclen)) {
 		if ((err = dmu_read(mos, spa->spa_history,
 		    shpp->sh_pool_create_len, sizeof (reclen) - firstread,
-		    buf + firstread)) != 0)
+		    buf + firstread, DMU_READ_PREFETCH)) != 0)
 			return (err);
 	}
 
@@ -380,10 +378,11 @@ spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
 		return (0);
 	}
 
-	err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf);
+	err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
+	    DMU_READ_PREFETCH);
 	if (leftover && err == 0) {
 		err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
-		    leftover, buf + read_len);
+		    leftover, buf + read_len, DMU_READ_PREFETCH);
 	}
 	mutex_exit(&spa->spa_history_lock);
 

module/zfs/spa_misc.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -425,7 +425,6 @@ spa_add(const char *name, const char *altroot)
 	spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);
 
 	mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
-	mutex_init(&spa->spa_async_root_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
@@ -434,7 +433,6 @@ spa_add(const char *name, const char *altroot)
 	mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
 
 	cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
-	cv_init(&spa->spa_async_root_cv, NULL, CV_DEFAULT, NULL);
 	cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
 	cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
 
@@ -508,12 +506,10 @@ spa_remove(spa_t *spa)
 	spa_config_lock_destroy(spa);
 
 	cv_destroy(&spa->spa_async_cv);
-	cv_destroy(&spa->spa_async_root_cv);
 	cv_destroy(&spa->spa_scrub_io_cv);
 	cv_destroy(&spa->spa_suspend_cv);
 
 	mutex_destroy(&spa->spa_async_lock);
-	mutex_destroy(&spa->spa_async_root_lock);
 	mutex_destroy(&spa->spa_scrub_lock);
 	mutex_destroy(&spa->spa_errlog_lock);
 	mutex_destroy(&spa->spa_errlist_lock);

module/zfs/space_map.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -116,12 +116,23 @@ space_map_add(space_map_t *sm, uint64_t start, uint64_t size)
 
 	if (merge_before && merge_after) {
 		avl_remove(&sm->sm_root, ss_before);
+		if (sm->sm_pp_root) {
+			avl_remove(sm->sm_pp_root, ss_before);
+			avl_remove(sm->sm_pp_root, ss_after);
+		}
 		ss_after->ss_start = ss_before->ss_start;
 		kmem_free(ss_before, sizeof (*ss_before));
+		ss = ss_after;
 	} else if (merge_before) {
 		ss_before->ss_end = end;
+		if (sm->sm_pp_root)
+			avl_remove(sm->sm_pp_root, ss_before);
+		ss = ss_before;
 	} else if (merge_after) {
 		ss_after->ss_start = start;
+		if (sm->sm_pp_root)
+			avl_remove(sm->sm_pp_root, ss_after);
+		ss = ss_after;
 	} else {
 		ss = kmem_alloc(sizeof (*ss), KM_SLEEP);
 		ss->ss_start = start;
@@ -129,6 +140,9 @@ space_map_add(space_map_t *sm, uint64_t start, uint64_t size)
 		avl_insert(&sm->sm_root, ss, where);
 	}
 
+	if (sm->sm_pp_root)
+		avl_add(sm->sm_pp_root, ss);
+
 	sm->sm_space += size;
 }
 
@@ -163,12 +177,17 @@ space_map_remove(space_map_t *sm, uint64_t start, uint64_t size)
 	left_over = (ss->ss_start != start);
 	right_over = (ss->ss_end != end);
 
+	if (sm->sm_pp_root)
+		avl_remove(sm->sm_pp_root, ss);
+
 	if (left_over && right_over) {
 		newseg = kmem_alloc(sizeof (*newseg), KM_SLEEP);
 		newseg->ss_start = end;
 		newseg->ss_end = ss->ss_end;
 		ss->ss_end = start;
 		avl_insert_here(&sm->sm_root, newseg, ss, AVL_AFTER);
+		if (sm->sm_pp_root)
+			avl_add(sm->sm_pp_root, newseg);
 	} else if (left_over) {
 		ss->ss_end = start;
 	} else if (right_over) {
@@ -176,8 +195,12 @@ space_map_remove(space_map_t *sm, uint64_t start, uint64_t size)
 	} else {
 		avl_remove(&sm->sm_root, ss);
 		kmem_free(ss, sizeof (*ss));
+		ss = NULL;
 	}
 
+	if (sm->sm_pp_root && ss != NULL)
+		avl_add(sm->sm_pp_root, ss);
+
 	sm->sm_space -= size;
 }
 
@@ -288,7 +311,8 @@ space_map_load(space_map_t *sm, space_map_ops_t *ops, uint8_t maptype,
 		    smo->smo_object, offset, size);
 
 		mutex_exit(sm->sm_lock);
-		error = dmu_read(os, smo->smo_object, offset, size, entry_map);
+		error = dmu_read(os, smo->smo_object, offset, size, entry_map,
+		    DMU_READ_PREFETCH);
 		mutex_enter(sm->sm_lock);
 		if (error != 0)
 			break;
@@ -341,6 +365,15 @@ space_map_unload(space_map_t *sm)
 	space_map_vacate(sm, NULL, NULL);
 }
 
+uint64_t
+space_map_maxsize(space_map_t *sm)
+{
+	if (sm->sm_loaded && sm->sm_ops != NULL)
+		return (sm->sm_ops->smop_max(sm));
+	else
+		return (-1ULL);
+}
+
 uint64_t
 space_map_alloc(space_map_t *sm, uint64_t size)
 {

module/zfs/vdev.c

@@ -39,6 +39,7 @@
 #include <sys/zap.h>
 #include <sys/fs/zfs.h>
 #include <sys/arc.h>
+#include <sys/zil.h>
 
 /*
  * Virtual device management.
@@ -83,9 +84,8 @@ vdev_default_asize(vdev_t *vd, uint64_t psize)
 {
 	uint64_t asize = P2ROUNDUP(psize, 1ULL << vd->vdev_top->vdev_ashift);
 	uint64_t csize;
-	uint64_t c;
 
-	for (c = 0; c < vd->vdev_children; c++) {
+	for (int c = 0; c < vd->vdev_children; c++) {
 		csize = vdev_psize_to_asize(vd->vdev_child[c], psize);
 		asize = MAX(asize, csize);
 	}
@@ -94,40 +94,47 @@ vdev_default_asize(vdev_t *vd, uint64_t psize)
 }
 
 /*
- * Get the replaceable or attachable device size.
+ * Get the minimum allocatable size. We define the allocatable size as
- * If the parent is a mirror or raidz, the replaceable size is the minimum
+ * the vdev's asize rounded to the nearest metaslab. This allows us to
- * psize of all its children. For the rest, just return our own psize.
+ * replace or attach devices which don't have the same physical size but
- *
+ * can still satisfy the same number of allocations.
- * e.g.
- *			psize	rsize
- * root			-	-
- *	mirror/raidz	-	-
- *	    disk1	20g	20g
- *	    disk2 	40g	20g
- *	disk3 		80g	80g
  */
 uint64_t
-vdev_get_rsize(vdev_t *vd)
+vdev_get_min_asize(vdev_t *vd)
 {
-	vdev_t *pvd, *cvd;
+	vdev_t *pvd = vd->vdev_parent;
-	uint64_t c, rsize;
-
-	pvd = vd->vdev_parent;
 
 	/*
-	 * If our parent is NULL or the root, just return our own psize.
+	 * The our parent is NULL (inactive spare or cache) or is the root,
+	 * just return our own asize.
 	 */
-	if (pvd == NULL || pvd->vdev_parent == NULL)
+	if (pvd == NULL)
-		return (vd->vdev_psize);
+		return (vd->vdev_asize);
 
-	rsize = 0;
+	/*
+	 * The top-level vdev just returns the allocatable size rounded
+	 * to the nearest metaslab.
+	 */
+	if (vd == vd->vdev_top)
+		return (P2ALIGN(vd->vdev_asize, 1ULL << vd->vdev_ms_shift));
 
-	for (c = 0; c < pvd->vdev_children; c++) {
+	/*
-		cvd = pvd->vdev_child[c];
+	 * The allocatable space for a raidz vdev is N * sizeof(smallest child),
-		rsize = MIN(rsize - 1, cvd->vdev_psize - 1) + 1;
+	 * so each child must provide at least 1/Nth of its asize.
-	}
+	 */
+	if (pvd->vdev_ops == &vdev_raidz_ops)
+		return (pvd->vdev_min_asize / pvd->vdev_children);
 
-	return (rsize);
+	return (pvd->vdev_min_asize);
+}
+
+void
+vdev_set_min_asize(vdev_t *vd)
+{
+	vd->vdev_min_asize = vdev_get_min_asize(vd);
+
+	for (int c = 0; c < vd->vdev_children; c++)
+		vdev_set_min_asize(vd->vdev_child[c]);
 }
 
 vdev_t *
@@ -148,13 +155,12 @@ vdev_lookup_top(spa_t *spa, uint64_t vdev)
 vdev_t *
 vdev_lookup_by_guid(vdev_t *vd, uint64_t guid)
 {
-	int c;
 	vdev_t *mvd;
 
 	if (vd->vdev_guid == guid)
 		return (vd);
 
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++)
 		if ((mvd = vdev_lookup_by_guid(vd->vdev_child[c], guid)) !=
 		    NULL)
 			return (mvd);
@@ -250,17 +256,17 @@ vdev_compact_children(vdev_t *pvd)
 {
 	vdev_t **newchild, *cvd;
 	int oldc = pvd->vdev_children;
-	int newc, c;
+	int newc;
 
 	ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
 
-	for (c = newc = 0; c < oldc; c++)
+	for (int c = newc = 0; c < oldc; c++)
 		if (pvd->vdev_child[c])
 			newc++;
 
 	newchild = kmem_alloc(newc * sizeof (vdev_t *), KM_SLEEP);
 
-	for (c = newc = 0; c < oldc; c++) {
+	for (int c = newc = 0; c < oldc; c++) {
 		if ((cvd = pvd->vdev_child[c]) != NULL) {
 			newchild[newc] = cvd;
 			cvd->vdev_id = newc++;
@@ -372,6 +378,9 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	} else if (alloctype == VDEV_ALLOC_L2CACHE) {
 		if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
 			return (EINVAL);
+	} else if (alloctype == VDEV_ALLOC_ROOTPOOL) {
+		if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
+			return (EINVAL);
 	}
 
 	/*
@@ -435,6 +444,8 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
 	    &vd->vdev_physpath) == 0)
 		vd->vdev_physpath = spa_strdup(vd->vdev_physpath);
+	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_FRU, &vd->vdev_fru) == 0)
+		vd->vdev_fru = spa_strdup(vd->vdev_fru);
 
 	/*
 	 * Set the whole_disk property.  If it's not specified, leave the value
@@ -448,9 +459,8 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	 * Look for the 'not present' flag.  This will only be set if the device
 	 * was not present at the time of import.
 	 */
-	if (!spa->spa_import_faulted)
+	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
-		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
+	    &vd->vdev_not_present);
-		    &vd->vdev_not_present);
 
 	/*
 	 * Get the alignment requirement.
@@ -473,13 +483,23 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	 * If we're a leaf vdev, try to load the DTL object and other state.
 	 */
 	if (vd->vdev_ops->vdev_op_leaf &&
-	    (alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_L2CACHE)) {
+	    (alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_L2CACHE ||
+	    alloctype == VDEV_ALLOC_ROOTPOOL)) {
 		if (alloctype == VDEV_ALLOC_LOAD) {
 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
 			    &vd->vdev_dtl_smo.smo_object);
 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
 			    &vd->vdev_unspare);
 		}
+
+		if (alloctype == VDEV_ALLOC_ROOTPOOL) {
+			uint64_t spare = 0;
+
+			if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
+			    &spare) == 0 && spare)
+				spa_spare_add(vd);
+		}
+
 		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE,
 		    &vd->vdev_offline);
 
@@ -511,7 +531,6 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 void
 vdev_free(vdev_t *vd)
 {
-	int c;
 	spa_t *spa = vd->vdev_spa;
 
 	/*
@@ -525,7 +544,7 @@ vdev_free(vdev_t *vd)
 	/*
 	 * Free all children.
 	 */
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++)
 		vdev_free(vd->vdev_child[c]);
 
 	ASSERT(vd->vdev_child == NULL);
@@ -560,6 +579,8 @@ vdev_free(vdev_t *vd)
 		spa_strfree(vd->vdev_devid);
 	if (vd->vdev_physpath)
 		spa_strfree(vd->vdev_physpath);
+	if (vd->vdev_fru)
+		spa_strfree(vd->vdev_fru);
 
 	if (vd->vdev_isspare)
 		spa_spare_remove(vd);
@@ -653,14 +674,12 @@ vdev_top_transfer(vdev_t *svd, vdev_t *tvd)
 static void
 vdev_top_update(vdev_t *tvd, vdev_t *vd)
 {
-	int c;
-
 	if (vd == NULL)
 		return;
 
 	vd->vdev_top = tvd;
 
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++)
 		vdev_top_update(tvd, vd->vdev_child[c]);
 }
 
@@ -679,6 +698,7 @@ vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops)
 	mvd = vdev_alloc_common(spa, cvd->vdev_id, 0, ops);
 
 	mvd->vdev_asize = cvd->vdev_asize;
+	mvd->vdev_min_asize = cvd->vdev_min_asize;
 	mvd->vdev_ashift = cvd->vdev_ashift;
 	mvd->vdev_state = cvd->vdev_state;
 
@@ -751,6 +771,15 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 	if (vd->vdev_ms_shift == 0)	/* not being allocated from yet */
 		return (0);
 
+	/*
+	 * Compute the raidz-deflation ratio.  Note, we hard-code
+	 * in 128k (1 << 17) because it is the current "typical" blocksize.
+	 * Even if SPA_MAXBLOCKSIZE changes, this algorithm must never change,
+	 * or we will inconsistently account for existing bp's.
+	 */
+	vd->vdev_deflate_ratio = (1 << 17) /
+	    (vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
+
 	ASSERT(oldc <= newc);
 
 	if (vd->vdev_islog)
@@ -776,7 +805,8 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 		if (txg == 0) {
 			uint64_t object = 0;
 			error = dmu_read(mos, vd->vdev_ms_array,
-			    m * sizeof (uint64_t), sizeof (uint64_t), &object);
+			    m * sizeof (uint64_t), sizeof (uint64_t), &object,
+			    DMU_READ_PREFETCH);
 			if (error)
 				return (error);
 			if (object != 0) {
@@ -903,7 +933,7 @@ vdev_probe(vdev_t *vd, zio_t *zio)
 
 		vps->vps_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_PROBE |
 		    ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE |
-		    ZIO_FLAG_DONT_RETRY;
+		    ZIO_FLAG_TRYHARD;
 
 		if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
 			/*
@@ -950,8 +980,8 @@ vdev_probe(vdev_t *vd, zio_t *zio)
 	for (int l = 1; l < VDEV_LABELS; l++) {
 		zio_nowait(zio_read_phys(pio, vd,
 		    vdev_label_offset(vd->vdev_psize, l,
-		    offsetof(vdev_label_t, vl_pad)),
+		    offsetof(vdev_label_t, vl_pad2)),
-		    VDEV_SKIP_SIZE, zio_buf_alloc(VDEV_SKIP_SIZE),
+		    VDEV_PAD_SIZE, zio_buf_alloc(VDEV_PAD_SIZE),
 		    ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
 		    ZIO_PRIORITY_SYNC_READ, vps->vps_flags, B_TRUE));
 	}
@@ -971,7 +1001,6 @@ vdev_open(vdev_t *vd)
 {
 	spa_t *spa = vd->vdev_spa;
 	int error;
-	int c;
 	uint64_t osize = 0;
 	uint64_t asize, psize;
 	uint64_t ashift = 0;
@@ -983,6 +1012,9 @@ vdev_open(vdev_t *vd)
 	    vd->vdev_state == VDEV_STATE_OFFLINE);
 
 	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
+	vd->vdev_cant_read = B_FALSE;
+	vd->vdev_cant_write = B_FALSE;
+	vd->vdev_min_asize = vdev_get_min_asize(vd);
 
 	if (!vd->vdev_removed && vd->vdev_faulted) {
 		ASSERT(vd->vdev_children == 0);
@@ -998,7 +1030,7 @@ vdev_open(vdev_t *vd)
 	error = vd->vdev_ops->vdev_op_open(vd, &osize, &ashift);
 
 	if (zio_injection_enabled && error == 0)
-		error = zio_handle_device_injection(vd, ENXIO);
+		error = zio_handle_device_injection(vd, NULL, ENXIO);
 
 	if (error) {
 		if (vd->vdev_removed &&
@@ -1020,12 +1052,13 @@ vdev_open(vdev_t *vd)
 		vd->vdev_state = VDEV_STATE_HEALTHY;
 	}
 
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++) {
 		if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
 			vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
 			    VDEV_AUX_NONE);
 			break;
 		}
+	}
 
 	osize = P2ALIGN(osize, (uint64_t)sizeof (vdev_label_t));
 
@@ -1050,6 +1083,15 @@ vdev_open(vdev_t *vd)
 
 	vd->vdev_psize = psize;
 
+	/*
+	 * Make sure the allocatable size hasn't shrunk.
+	 */
+	if (asize < vd->vdev_min_asize) {
+		vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+		    VDEV_AUX_BAD_LABEL);
+		return (EINVAL);
+	}
+
 	if (vd->vdev_asize == 0) {
 		/*
 		 * This is the first-ever open, so use the computed values.
@@ -1066,26 +1108,19 @@ vdev_open(vdev_t *vd)
 			    VDEV_AUX_BAD_LABEL);
 			return (EINVAL);
 		}
-
-		/*
-		 * Make sure the device hasn't shrunk.
-		 */
-		if (asize < vd->vdev_asize) {
-			vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
-			    VDEV_AUX_BAD_LABEL);
-			return (EINVAL);
-		}
-
-		/*
-		 * If all children are healthy and the asize has increased,
-		 * then we've experienced dynamic LUN growth.
-		 */
-		if (vd->vdev_state == VDEV_STATE_HEALTHY &&
-		    asize > vd->vdev_asize) {
-			vd->vdev_asize = asize;
-		}
 	}
 
+	/*
+	 * If all children are healthy and the asize has increased,
+	 * then we've experienced dynamic LUN growth.  If automatic
+	 * expansion is enabled then use the additional space.
+	 */
+	if (vd->vdev_state == VDEV_STATE_HEALTHY && asize > vd->vdev_asize &&
+	    (vd->vdev_expanding || spa->spa_autoexpand))
+		vd->vdev_asize = asize;
+
+	vdev_set_min_asize(vd);
+
 	/*
 	 * Ensure we can issue some IO before declaring the
 	 * vdev open for business.
@@ -1097,18 +1132,6 @@ vdev_open(vdev_t *vd)
 		return (error);
 	}
 
-	/*
-	 * If this is a top-level vdev, compute the raidz-deflation
-	 * ratio.  Note, we hard-code in 128k (1<<17) because it is the
-	 * current "typical" blocksize.  Even if SPA_MAXBLOCKSIZE
-	 * changes, this algorithm must never change, or we will
-	 * inconsistently account for existing bp's.
-	 */
-	if (vd->vdev_top == vd) {
-		vd->vdev_deflate_ratio = (1<<17) /
-		    (vdev_psize_to_asize(vd, 1<<17) >> SPA_MINBLOCKSHIFT);
-	}
-
 	/*
 	 * If a leaf vdev has a DTL, and seems healthy, then kick off a
 	 * resilver.  But don't do this if we are doing a reopen for a scrub,
@@ -1135,12 +1158,11 @@ int
 vdev_validate(vdev_t *vd)
 {
 	spa_t *spa = vd->vdev_spa;
-	int c;
 	nvlist_t *label;
 	uint64_t guid, top_guid;
 	uint64_t state;
 
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++)
 		if (vdev_validate(vd->vdev_child[c]) != 0)
 			return (EBADF);
 
@@ -1226,7 +1248,7 @@ vdev_close(vdev_t *vd)
 	vdev_cache_purge(vd);
 
 	/*
-	 * We record the previous state before we close it, so  that if we are
+	 * We record the previous state before we close it, so that if we are
 	 * doing a reopen(), we don't generate FMA ereports if we notice that
 	 * it's still faulted.
 	 */
@@ -1257,12 +1279,9 @@ vdev_reopen(vdev_t *vd)
 	if (vd->vdev_aux) {
 		(void) vdev_validate_aux(vd);
 		if (vdev_readable(vd) && vdev_writeable(vd) &&
-		    !l2arc_vdev_present(vd)) {
+		    vd->vdev_aux == &spa->spa_l2cache &&
-			uint64_t size = vdev_get_rsize(vd);
+		    !l2arc_vdev_present(vd))
-			l2arc_add_vdev(spa, vd,
+			l2arc_add_vdev(spa, vd);
-			    VDEV_LABEL_START_SIZE,
-			    size - VDEV_LABEL_START_SIZE);
-		}
 	} else {
 		(void) vdev_validate(vd);
 	}
@@ -1302,26 +1321,14 @@ vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
 	return (0);
 }
 
-/*
- * The is the latter half of vdev_create().  It is distinct because it
- * involves initiating transactions in order to do metaslab creation.
- * For creation, we want to try to create all vdevs at once and then undo it
- * if anything fails; this is much harder if we have pending transactions.
- */
 void
-vdev_init(vdev_t *vd, uint64_t txg)
+vdev_metaslab_set_size(vdev_t *vd)
 {
 	/*
 	 * Aim for roughly 200 metaslabs per vdev.
 	 */
 	vd->vdev_ms_shift = highbit(vd->vdev_asize / 200);
 	vd->vdev_ms_shift = MAX(vd->vdev_ms_shift, SPA_MAXBLOCKSHIFT);
-
-	/*
-	 * Initialize the vdev's metaslabs.  This can't fail because
-	 * there's nothing to read when creating all new metaslabs.
-	 */
-	VERIFY(vdev_metaslab_init(vd, txg) == 0);
 }
 
 void
@@ -1879,7 +1886,7 @@ vdev_degrade(spa_t *spa, uint64_t guid)
 int
 vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 {
-	vdev_t *vd;
+	vdev_t *vd, *tvd, *pvd, *rvd = spa->spa_root_vdev;
 
 	spa_vdev_state_enter(spa);
 
@@ -1889,13 +1896,26 @@ vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 	if (!vd->vdev_ops->vdev_op_leaf)
 		return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
 
+	tvd = vd->vdev_top;
 	vd->vdev_offline = B_FALSE;
 	vd->vdev_tmpoffline = B_FALSE;
 	vd->vdev_checkremove = !!(flags & ZFS_ONLINE_CHECKREMOVE);
 	vd->vdev_forcefault = !!(flags & ZFS_ONLINE_FORCEFAULT);
-	vdev_reopen(vd->vdev_top);
+
+	/* XXX - L2ARC 1.0 does not support expansion */
+	if (!vd->vdev_aux) {
+		for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
+			pvd->vdev_expanding = !!(flags & ZFS_ONLINE_EXPAND);
+	}
+
+	vdev_reopen(tvd);
 	vd->vdev_checkremove = vd->vdev_forcefault = B_FALSE;
 
+	if (!vd->vdev_aux) {
+		for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
+			pvd->vdev_expanding = B_FALSE;
+	}
+
 	if (newstate)
 		*newstate = vd->vdev_state;
 	if ((flags & ZFS_ONLINE_UNSPARE) &&
@@ -1904,13 +1924,21 @@ vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 	    vd->vdev_parent->vdev_child[0] == vd)
 		vd->vdev_unspare = B_TRUE;
 
+	if ((flags & ZFS_ONLINE_EXPAND) || spa->spa_autoexpand) {
+
+		/* XXX - L2ARC 1.0 does not support expansion */
+		if (vd->vdev_aux)
+			return (spa_vdev_state_exit(spa, vd, ENOTSUP));
+		spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
+	}
 	return (spa_vdev_state_exit(spa, vd, 0));
 }
 
 int
 vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
 {
-	vdev_t *vd;
+	vdev_t *vd, *tvd;
+	int error;
 
 	spa_vdev_state_enter(spa);
 
@@ -1920,34 +1948,58 @@ vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
 	if (!vd->vdev_ops->vdev_op_leaf)
 		return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
 
+	tvd = vd->vdev_top;
+
 	/*
 	 * If the device isn't already offline, try to offline it.
 	 */
 	if (!vd->vdev_offline) {
 		/*
 		 * If this device has the only valid copy of some data,
-		 * don't allow it to be offlined.
+		 * don't allow it to be offlined. Log devices are always
+		 * expendable.
 		 */
-		if (vd->vdev_aux == NULL && vdev_dtl_required(vd))
+		if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
+		    vdev_dtl_required(vd))
 			return (spa_vdev_state_exit(spa, NULL, EBUSY));
 
 		/*
 		 * Offline this device and reopen its top-level vdev.
-		 * If this action results in the top-level vdev becoming
+		 * If the top-level vdev is a log device then just offline
-		 * unusable, undo it and fail the request.
+		 * it. Otherwise, if this action results in the top-level
+		 * vdev becoming unusable, undo it and fail the request.
 		 */
 		vd->vdev_offline = B_TRUE;
-		vdev_reopen(vd->vdev_top);
+		vdev_reopen(tvd);
-		if (vd->vdev_aux == NULL && vdev_is_dead(vd->vdev_top)) {
+
+		if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
+		    vdev_is_dead(tvd)) {
 			vd->vdev_offline = B_FALSE;
-			vdev_reopen(vd->vdev_top);
+			vdev_reopen(tvd);
 			return (spa_vdev_state_exit(spa, NULL, EBUSY));
 		}
 	}
 
 	vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);
 
-	return (spa_vdev_state_exit(spa, vd, 0));
+	if (!tvd->vdev_islog || !vdev_is_dead(tvd))
+		return (spa_vdev_state_exit(spa, vd, 0));
+
+	(void) spa_vdev_state_exit(spa, vd, 0);
+
+	error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
+	    NULL, DS_FIND_CHILDREN);
+	if (error) {
+		(void) vdev_online(spa, guid, 0, NULL);
+		return (error);
+	}
+	/*
+	 * If we successfully offlined the log device then we need to
+	 * 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 (0);
 }
 
 /*
@@ -2062,7 +2114,9 @@ vdev_get_stats(vdev_t *vd, vdev_stat_t *vs)
 	vs->vs_scrub_errors = vd->vdev_spa->spa_scrub_errors;
 	vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
 	vs->vs_state = vd->vdev_state;
-	vs->vs_rsize = vdev_get_rsize(vd);
+	vs->vs_rsize = vdev_get_min_asize(vd);
+	if (vd->vdev_ops->vdev_op_leaf)
+		vs->vs_rsize += VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE;
 	mutex_exit(&vd->vdev_stat_lock);
 
 	/*
@@ -2155,14 +2209,24 @@ vdev_stat_update(zio_t *zio, uint64_t psize)
 	if (flags & ZIO_FLAG_SPECULATIVE)
 		return;
 
+	/*
+	 * If this is an I/O error that is going to be retried, then ignore the
+	 * error.  Otherwise, the user may interpret B_FAILFAST I/O errors as
+	 * hard errors, when in reality they can happen for any number of
+	 * innocuous reasons (bus resets, MPxIO link failure, etc).
+	 */
+	if (zio->io_error == EIO &&
+	    !(zio->io_flags & ZIO_FLAG_IO_RETRY))
+		return;
+
 	mutex_enter(&vd->vdev_stat_lock);
-	if (type == ZIO_TYPE_READ) {
+	if (type == ZIO_TYPE_READ && !vdev_is_dead(vd)) {
 		if (zio->io_error == ECKSUM)
 			vs->vs_checksum_errors++;
 		else
 			vs->vs_read_errors++;
 	}
-	if (type == ZIO_TYPE_WRITE)
+	if (type == ZIO_TYPE_WRITE && !vdev_is_dead(vd))
 		vs->vs_write_errors++;
 	mutex_exit(&vd->vdev_stat_lock);
 
@@ -2205,10 +2269,9 @@ vdev_stat_update(zio_t *zio, uint64_t psize)
 void
 vdev_scrub_stat_update(vdev_t *vd, pool_scrub_type_t type, boolean_t complete)
 {
-	int c;
 	vdev_stat_t *vs = &vd->vdev_stat;
 
-	for (c = 0; c < vd->vdev_children; c++)
+	for (int c = 0; c < vd->vdev_children; c++)
 		vdev_scrub_stat_update(vd->vdev_child[c], type, complete);
 
 	mutex_enter(&vd->vdev_stat_lock);
@@ -2252,6 +2315,7 @@ vdev_space_update(vdev_t *vd, int64_t space_delta, int64_t alloc_delta,
 	 * childrens', thus not accurate enough for us.
 	 */
 	ASSERT((dspace_delta & (SPA_MINBLOCKSIZE-1)) == 0);
+	ASSERT(vd->vdev_deflate_ratio != 0 || vd->vdev_isl2cache);
 	dspace_delta = (dspace_delta >> SPA_MINBLOCKSHIFT) *
 	    vd->vdev_deflate_ratio;
 
@@ -2293,8 +2357,8 @@ vdev_config_dirty(vdev_t *vd)
 	int c;
 
 	/*
-	 * If this is an aux vdev (as with l2cache devices), then we update the
+	 * If this is an aux vdev (as with l2cache and spare devices), then we
-	 * vdev config manually and set the sync flag.
+	 * update the vdev config manually and set the sync flag.
 	 */
 	if (vd->vdev_aux != NULL) {
 		spa_aux_vdev_t *sav = vd->vdev_aux;
@@ -2316,8 +2380,11 @@ vdev_config_dirty(vdev_t *vd)
 
 		sav->sav_sync = B_TRUE;
 
-		VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
+		if (nvlist_lookup_nvlist_array(sav->sav_config,
-		    ZPOOL_CONFIG_L2CACHE, &aux, &naux) == 0);
+		    ZPOOL_CONFIG_L2CACHE, &aux, &naux) != 0) {
+			VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
+			    ZPOOL_CONFIG_SPARES, &aux, &naux) == 0);
+		}
 
 		ASSERT(c < naux);
 
@@ -2415,11 +2482,10 @@ vdev_propagate_state(vdev_t *vd)
 	vdev_t *rvd = spa->spa_root_vdev;
 	int degraded = 0, faulted = 0;
 	int corrupted = 0;
-	int c;
 	vdev_t *child;
 
 	if (vd->vdev_children > 0) {
-		for (c = 0; c < vd->vdev_children; c++) {
+		for (int c = 0; c < vd->vdev_children; c++) {
 			child = vd->vdev_child[c];
 
 			if (!vdev_readable(child) ||
@@ -2523,7 +2589,6 @@ vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
 		 * an error.
 		 */
 		if (spa->spa_load_state == SPA_LOAD_IMPORT &&
-		    !spa->spa_import_faulted &&
 		    vd->vdev_ops->vdev_op_leaf)
 			vd->vdev_not_present = 1;
 
@@ -2582,8 +2647,8 @@ vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
 		vd->vdev_removed = B_FALSE;
 	}
 
-	if (!isopen)
+	if (!isopen && vd->vdev_parent)
-		vdev_propagate_state(vd);
+		vdev_propagate_state(vd->vdev_parent);
 }
 
 /*
@@ -2595,8 +2660,6 @@ vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
 boolean_t
 vdev_is_bootable(vdev_t *vd)
 {
-	int c;
-
 	if (!vd->vdev_ops->vdev_op_leaf) {
 		char *vdev_type = vd->vdev_ops->vdev_op_type;
 
@@ -2611,9 +2674,53 @@ vdev_is_bootable(vdev_t *vd)
 		return (B_FALSE);
 	}
 
-	for (c = 0; c < vd->vdev_children; c++) {
+	for (int c = 0; c < vd->vdev_children; c++) {
 		if (!vdev_is_bootable(vd->vdev_child[c]))
 			return (B_FALSE);
 	}
 	return (B_TRUE);
 }
+
+void
+vdev_load_log_state(vdev_t *vd, nvlist_t *nv)
+{
+	uint_t children;
+	nvlist_t **child;
+	uint64_t val;
+	spa_t *spa = vd->vdev_spa;
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) == 0) {
+		for (int c = 0; c < children; c++)
+			vdev_load_log_state(vd->vdev_child[c], child[c]);
+	}
+
+	if (vd->vdev_ops->vdev_op_leaf && nvlist_lookup_uint64(nv,
+	    ZPOOL_CONFIG_OFFLINE, &val) == 0 && val) {
+
+		/*
+		 * It would be nice to call vdev_offline()
+		 * directly but the pool isn't fully loaded and
+		 * the txg threads have not been started yet.
+		 */
+		spa_config_enter(spa, SCL_STATE_ALL, FTAG, RW_WRITER);
+		vd->vdev_offline = val;
+		vdev_reopen(vd->vdev_top);
+		spa_config_exit(spa, SCL_STATE_ALL, FTAG);
+	}
+}
+
+/*
+ * Expand a vdev if possible.
+ */
+void
+vdev_expand(vdev_t *vd, uint64_t txg)
+{
+	ASSERT(vd->vdev_top == vd);
+	ASSERT(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+
+	if ((vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count) {
+		VERIFY(vdev_metaslab_init(vd, txg) == 0);
+		vdev_config_dirty(vd);
+	}
+}

module/zfs/vdev_label.c

@@ -233,6 +233,10 @@ vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
 		VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH,
 		    vd->vdev_physpath) == 0);
 
+	if (vd->vdev_fru != NULL)
+		VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_FRU,
+		    vd->vdev_fru) == 0);
+
 	if (vd->vdev_nparity != 0) {
 		ASSERT(strcmp(vd->vdev_ops->vdev_op_type,
 		    VDEV_TYPE_RAIDZ) == 0);
@@ -335,8 +339,8 @@ vdev_label_read_config(vdev_t *vd)
 	nvlist_t *config = NULL;
 	vdev_phys_t *vp;
 	zio_t *zio;
-	int flags =
+	int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
-	    ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
+	    ZIO_FLAG_SPECULATIVE;
 
 	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
 
@@ -345,6 +349,7 @@ vdev_label_read_config(vdev_t *vd)
 
 	vp = zio_buf_alloc(sizeof (vdev_phys_t));
 
+retry:
 	for (int l = 0; l < VDEV_LABELS; l++) {
 
 		zio = zio_root(spa, NULL, NULL, flags);
@@ -364,6 +369,11 @@ vdev_label_read_config(vdev_t *vd)
 		}
 	}
 
+	if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) {
+		flags |= ZIO_FLAG_TRYHARD;
+		goto retry;
+	}
+
 	zio_buf_free(vp, sizeof (vdev_phys_t));
 
 	return (config);
@@ -488,7 +498,7 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 	spa_t *spa = vd->vdev_spa;
 	nvlist_t *label;
 	vdev_phys_t *vp;
-	vdev_boot_header_t *vb;
+	char *pad2;
 	uberblock_t *ub;
 	zio_t *zio;
 	char *buf;
@@ -629,16 +639,6 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 		return (error == EFAULT ? ENAMETOOLONG : EINVAL);
 	}
 
-	/*
-	 * Initialize boot block header.
-	 */
-	vb = zio_buf_alloc(sizeof (vdev_boot_header_t));
-	bzero(vb, sizeof (vdev_boot_header_t));
-	vb->vb_magic = VDEV_BOOT_MAGIC;
-	vb->vb_version = VDEV_BOOT_VERSION;
-	vb->vb_offset = VDEV_BOOT_OFFSET;
-	vb->vb_size = VDEV_BOOT_SIZE;
-
 	/*
 	 * Initialize uberblock template.
 	 */
@@ -647,9 +647,14 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 	*ub = spa->spa_uberblock;
 	ub->ub_txg = 0;
 
+	/* Initialize the 2nd padding area. */
+	pad2 = zio_buf_alloc(VDEV_PAD_SIZE);
+	bzero(pad2, VDEV_PAD_SIZE);
+
 	/*
 	 * Write everything in parallel.
 	 */
+retry:
 	zio = zio_root(spa, NULL, NULL, flags);
 
 	for (int l = 0; l < VDEV_LABELS; l++) {
@@ -658,9 +663,14 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 		    offsetof(vdev_label_t, vl_vdev_phys),
 		    sizeof (vdev_phys_t), NULL, NULL, flags);
 
-		vdev_label_write(zio, vd, l, vb,
+		/*
-		    offsetof(vdev_label_t, vl_boot_header),
+		 * Skip the 1st padding area.
-		    sizeof (vdev_boot_header_t), NULL, NULL, flags);
+		 * Zero out the 2nd padding area where it might have
+		 * left over data from previous filesystem format.
+		 */
+		vdev_label_write(zio, vd, l, pad2,
+		    offsetof(vdev_label_t, vl_pad2),
+		    VDEV_PAD_SIZE, NULL, NULL, flags);
 
 		for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
 			vdev_label_write(zio, vd, l, ub,
@@ -671,9 +681,14 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 
 	error = zio_wait(zio);
 
+	if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) {
+		flags |= ZIO_FLAG_TRYHARD;
+		goto retry;
+	}
+
 	nvlist_free(label);
+	zio_buf_free(pad2, VDEV_PAD_SIZE);
 	zio_buf_free(ub, VDEV_UBERBLOCK_SIZE(vd));
-	zio_buf_free(vb, sizeof (vdev_boot_header_t));
 	zio_buf_free(vp, sizeof (vdev_phys_t));
 
 	/*
@@ -757,8 +772,8 @@ vdev_uberblock_load(zio_t *zio, vdev_t *vd, uberblock_t *ubbest)
 {
 	spa_t *spa = vd->vdev_spa;
 	vdev_t *rvd = spa->spa_root_vdev;
-	int flags =
+	int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
-	    ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
+	    ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD;
 
 	if (vd == rvd) {
 		ASSERT(zio == NULL);
@@ -996,7 +1011,7 @@ vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags)
  * at any time, you can just call it again, and it will resume its work.
  */
 int
-vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg)
+vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg, boolean_t tryhard)
 {
 	spa_t *spa = svd[0]->vdev_spa;
 	uberblock_t *ub = &spa->spa_uberblock;
@@ -1005,6 +1020,16 @@ vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg)
 	int error;
 	int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
 
+	/*
+	 * Normally, we don't want to try too hard to write every label and
+	 * uberblock.  If there is a flaky disk, we don't want the rest of the
+	 * sync process to block while we retry.  But if we can't write a
+	 * single label out, we should retry with ZIO_FLAG_TRYHARD before
+	 * bailing out and declaring the pool faulted.
+	 */
+	if (tryhard)
+		flags |= ZIO_FLAG_TRYHARD;
+
 	ASSERT(ub->ub_txg <= txg);
 
 	/*

module/zfs/vdev_queue.c

@@ -48,10 +48,11 @@ int zfs_vdev_time_shift = 6;
 int zfs_vdev_ramp_rate = 2;
 
 /*
- * i/os will be aggregated into a single large i/o up to
+ * To reduce IOPs, we aggregate small adjacent i/os into one large i/o.
- * zfs_vdev_aggregation_limit bytes long.
+ * For read i/os, we also aggregate across small adjacency gaps.
  */
 int zfs_vdev_aggregation_limit = SPA_MAXBLOCKSIZE;
+int zfs_vdev_read_gap_limit = 32 << 10;
 
 /*
  * Virtual device vector for disk I/O scheduling.
@@ -159,16 +160,23 @@ vdev_queue_agg_io_done(zio_t *aio)
 	zio_buf_free(aio->io_data, aio->io_size);
 }
 
-#define	IS_ADJACENT(io, nio) \
+/*
-	((io)->io_offset + (io)->io_size == (nio)->io_offset)
+ * Compute the range spanned by two i/os, which is the endpoint of the last
+ * (lio->io_offset + lio->io_size) minus start of the first (fio->io_offset).
+ * Conveniently, the gap between fio and lio is given by -IO_SPAN(lio, fio);
+ * thus fio and lio are adjacent if and only if IO_SPAN(lio, fio) == 0.
+ */
+#define	IO_SPAN(fio, lio) ((lio)->io_offset + (lio)->io_size - (fio)->io_offset)
+#define	IO_GAP(fio, lio) (-IO_SPAN(lio, fio))
 
 static zio_t *
 vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 {
 	zio_t *fio, *lio, *aio, *dio, *nio;
 	avl_tree_t *t;
-	uint64_t size;
 	int flags;
+	uint64_t maxspan = zfs_vdev_aggregation_limit;
+	uint64_t maxgap;
 
 	ASSERT(MUTEX_HELD(&vq->vq_lock));
 
@@ -179,8 +187,8 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 	fio = lio = avl_first(&vq->vq_deadline_tree);
 
 	t = fio->io_vdev_tree;
-	size = fio->io_size;
 	flags = fio->io_flags & ZIO_FLAG_AGG_INHERIT;
+	maxgap = (t == &vq->vq_read_tree) ? zfs_vdev_read_gap_limit : 0;
 
 	if (!(flags & ZIO_FLAG_DONT_AGGREGATE)) {
 		/*
@@ -191,22 +199,18 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 		 * scrub/resilver, can be preserved in the aggregate.
 		 */
 		while ((dio = AVL_PREV(t, fio)) != NULL &&
-		    IS_ADJACENT(dio, fio) &&
 		    (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
-		    size + dio->io_size <= zfs_vdev_aggregation_limit) {
+		    IO_SPAN(dio, lio) <= maxspan && IO_GAP(dio, fio) <= maxgap)
 			fio = dio;
-			size += dio->io_size;
+
-		}
 		while ((dio = AVL_NEXT(t, lio)) != NULL &&
-		    IS_ADJACENT(lio, dio) &&
 		    (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
-		    size + dio->io_size <= zfs_vdev_aggregation_limit) {
+		    IO_SPAN(fio, dio) <= maxspan && IO_GAP(lio, dio) <= maxgap)
 			lio = dio;
-			size += dio->io_size;
-		}
 	}
 
 	if (fio != lio) {
+		uint64_t size = IO_SPAN(fio, lio);
 		ASSERT(size <= zfs_vdev_aggregation_limit);
 
 		aio = zio_vdev_delegated_io(fio->io_vd, fio->io_offset,
@@ -214,9 +218,10 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 		    flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE,
 		    vdev_queue_agg_io_done, NULL);
 
-		/* We want to process lio, then stop */
+		nio = fio;
-		lio = AVL_NEXT(t, lio);
+		do {
-		for (dio = fio; dio != lio; dio = nio) {
+			dio = nio;
+			nio = AVL_NEXT(t, dio);
 			ASSERT(dio->io_type == aio->io_type);
 			ASSERT(dio->io_vdev_tree == t);
 
@@ -224,13 +229,12 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 				bcopy(dio->io_data, (char *)aio->io_data +
 				    (dio->io_offset - aio->io_offset),
 				    dio->io_size);
-			nio = AVL_NEXT(t, dio);
 
 			zio_add_child(dio, aio);
 			vdev_queue_io_remove(vq, dio);
 			zio_vdev_io_bypass(dio);
 			zio_execute(dio);
-		}
+		} while (dio != lio);
 
 		avl_add(&vq->vq_pending_tree, aio);
 

module/zfs/vdev_raidz.c

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -697,7 +697,7 @@ vdev_raidz_io_start(zio_t *zio)
 			continue;
 		}
 		if (c >= rm->rm_firstdatacol || rm->rm_missingdata > 0 ||
-		    (zio->io_flags & ZIO_FLAG_SCRUB)) {
+		    (zio->io_flags & (ZIO_FLAG_SCRUB | ZIO_FLAG_RESILVER))) {
 			zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
 			    rc->rc_offset, rc->rc_data, rc->rc_size,
 			    zio->io_type, zio->io_priority, 0,

module/zfs/zap.c

@@ -19,13 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
-
 /*
  * This file contains the top half of the zfs directory structure
  * implementation. The bottom half is in zap_leaf.c.
@@ -45,6 +42,7 @@
 #include <sys/dmu.h>
 #include <sys/zfs_context.h>
 #include <sys/zfs_znode.h>
+#include <sys/fs/zfs.h>
 #include <sys/zap.h>
 #include <sys/refcount.h>
 #include <sys/zap_impl.h>
@@ -1157,3 +1155,58 @@ fzap_get_stats(zap_t *zap, zap_stats_t *zs)
 		}
 	}
 }
+
+int
+fzap_count_write(zap_name_t *zn, int add, uint64_t *towrite,
+    uint64_t *tooverwrite)
+{
+	zap_t *zap = zn->zn_zap;
+	zap_leaf_t *l;
+	int err;
+
+	/*
+	 * Account for the header block of the fatzap.
+	 */
+	if (!add && dmu_buf_freeable(zap->zap_dbuf)) {
+		tooverwrite += zap->zap_dbuf->db_size;
+	} else {
+		towrite += zap->zap_dbuf->db_size;
+	}
+
+	/*
+	 * Account for the pointer table blocks.
+	 * If we are adding we need to account for the following cases :
+	 * - If the pointer table is embedded, this operation could force an
+	 *   external pointer table.
+	 * - If this already has an external pointer table this operation
+	 *   could extend the table.
+	 */
+	if (add) {
+		if (zap->zap_f.zap_phys->zap_ptrtbl.zt_blk == 0)
+			towrite += zap->zap_dbuf->db_size;
+		else
+			towrite += (zap->zap_dbuf->db_size * 3);
+	}
+
+	/*
+	 * Now, check if the block containing leaf is freeable
+	 * and account accordingly.
+	 */
+	err = zap_deref_leaf(zap, zn->zn_hash, NULL, RW_READER, &l);
+	if (err != 0) {
+		return (err);
+	}
+
+	if (!add && dmu_buf_freeable(l->l_dbuf)) {
+		tooverwrite += l->l_dbuf->db_size;
+	} else {
+		/*
+		 * If this an add operation, the leaf block could split.
+		 * Hence, we need to account for an additional leaf block.
+		 */
+		towrite += (add ? 2 : 1) * l->l_dbuf->db_size;
+	}
+
+	zap_put_leaf(l);
+	return (0);
+}

module/zfs/zap_leaf.c

@@ -19,24 +19,23 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 /*
  * The 512-byte leaf is broken into 32 16-byte chunks.
  * chunk number n means l_chunk[n], even though the header precedes it.
  * the names are stored null-terminated.
  */
 
+#include <sys/spa.h>
+#include <sys/dmu.h>
 #include <sys/zfs_context.h>
+#include <sys/fs/zfs.h>
 #include <sys/zap.h>
 #include <sys/zap_impl.h>
 #include <sys/zap_leaf.h>
-#include <sys/spa.h>
-#include <sys/dmu.h>
 
 static uint16_t *zap_leaf_rehash_entry(zap_leaf_t *l, uint16_t entry);
 

module/zfs/zap_micro.c

@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/spa.h>
 #include <sys/dmu.h>
 #include <sys/zfs_context.h>
@@ -78,8 +76,8 @@ zap_normalize(zap_t *zap, const char *name, char *namenorm)
 
 	err = 0;
 	(void) u8_textprep_str((char *)name, &inlen, namenorm, &outlen,
-	    zap->zap_normflags | U8_TEXTPREP_IGNORE_NULL, U8_UNICODE_LATEST,
+	    zap->zap_normflags | U8_TEXTPREP_IGNORE_NULL |
-	    &err);
+	    U8_TEXTPREP_IGNORE_INVALID, U8_UNICODE_LATEST, &err);
 
 	return (err);
 }
@@ -1106,3 +1104,79 @@ zap_get_stats(objset_t *os, uint64_t zapobj, zap_stats_t *zs)
 	zap_unlockdir(zap);
 	return (0);
 }
+
+int
+zap_count_write(objset_t *os, uint64_t zapobj, const char *name, int add,
+    uint64_t *towrite, uint64_t *tooverwrite, uint64_t dn_datablkshift)
+{
+	zap_t *zap;
+	int err = 0;
+
+
+	/*
+	 * Since, we don't have a name, we cannot figure out which blocks will
+	 * be affected in this operation. So, account for the worst case :
+	 * - 3 blocks overwritten: target leaf, ptrtbl block, header block
+	 * - 4 new blocks written if adding:
+	 * 	- 2 blocks for possibly split leaves,
+	 * 	- 2 grown ptrtbl blocks
+	 *
+	 * This also accomodates the case where an add operation to a fairly
+	 * large microzap results in a promotion to fatzap.
+	 */
+	if (name == NULL) {
+		*towrite += (3 + (add ? 4 : 0)) * SPA_MAXBLOCKSIZE;
+		return (err);
+	}
+
+	/*
+	 * We lock the zap with adding ==  FALSE. Because, if we pass
+	 * the actual value of add, it could trigger a mzap_upgrade().
+	 * At present we are just evaluating the possibility of this operation
+	 * and hence we donot want to trigger an upgrade.
+	 */
+	err = zap_lockdir(os, zapobj, NULL, RW_READER, TRUE, FALSE, &zap);
+	if (err)
+		return (err);
+
+	if (!zap->zap_ismicro) {
+		zap_name_t *zn = zap_name_alloc(zap, name, MT_EXACT);
+		if (zn) {
+			err = fzap_count_write(zn, add, towrite,
+			    tooverwrite);
+			zap_name_free(zn);
+		} else {
+			/*
+			 * We treat this case as similar to (name == NULL)
+			 */
+			*towrite += (3 + (add ? 4 : 0)) * SPA_MAXBLOCKSIZE;
+		}
+	} else {
+		if (!add) {
+			if (dmu_buf_freeable(zap->zap_dbuf))
+				*tooverwrite += SPA_MAXBLOCKSIZE;
+			else
+				*towrite += SPA_MAXBLOCKSIZE;
+		} else {
+			/*
+			 * We are here if we are adding and (name != NULL).
+			 * It is hard to find out if this add will promote this
+			 * microzap to fatzap. Hence, we assume the worst case
+			 * and account for the blocks assuming this microzap
+			 * would be promoted to a fatzap.
+			 *
+			 * 1 block overwritten  : header block
+			 * 4 new blocks written : 2 new split leaf, 2 grown
+			 *			ptrtbl blocks
+			 */
+			if (dmu_buf_freeable(zap->zap_dbuf))
+				*tooverwrite += 1 << dn_datablkshift;
+			else
+				*towrite += 1 << dn_datablkshift;
+			*towrite += 4 << dn_datablkshift;
+		}
+	}
+
+	zap_unlockdir(zap);
+	return (err);
+}

module/zfs/zfs_acl.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -65,15 +65,16 @@
     ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
 #define	OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
     ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
-#define	WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
 
 #define	ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
     ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
     ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
     ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
 
-#define	WRITE_MASK (WRITE_MASK_DATA|ACE_WRITE_ATTRIBUTES|ACE_WRITE_ACL|\
+#define	WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
-    ACE_WRITE_OWNER|ACE_DELETE|ACE_DELETE_CHILD)
+#define	WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
+    ACE_DELETE|ACE_DELETE_CHILD)
+#define	WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
 
 #define	OGE_CLEAR	(ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
     ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
@@ -538,8 +539,9 @@ zfs_acl_curr_node(zfs_acl_t *aclp)
  * ACE FUIDs will be created later.
  */
 int
-zfs_copy_ace_2_fuid(vtype_t obj_type, zfs_acl_t *aclp, void *datap,
+zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
-    zfs_ace_t *z_acl, int aclcnt, size_t *size)
+    void *datap, zfs_ace_t *z_acl, int aclcnt, size_t *size,
+    zfs_fuid_info_t **fuidp, cred_t *cr)
 {
 	int i;
 	uint16_t entry_type;
@@ -555,9 +557,9 @@ zfs_copy_ace_2_fuid(vtype_t obj_type, zfs_acl_t *aclp, void *datap,
 		entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
 		if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
 		    entry_type != ACE_EVERYONE) {
-			if (!aclp->z_has_fuids)
+			aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
-				aclp->z_has_fuids = IS_EPHEMERAL(acep->a_who);
+			    cr, (entry_type == 0) ?
-			aceptr->z_fuid = (uint64_t)acep->a_who;
+			    ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
 		}
 
 		/*
@@ -682,7 +684,7 @@ zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
  * convert old ACL format to new
  */
 void
-zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp)
+zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
 {
 	zfs_oldace_t *oldaclp;
 	int i;
@@ -714,9 +716,9 @@ zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp)
 	newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
 	    sizeof (zfs_object_ace_t));
 	aclp->z_ops = zfs_acl_fuid_ops;
-	VERIFY(zfs_copy_ace_2_fuid(ZTOV(zp)->v_type, aclp, oldaclp,
+	VERIFY(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
-	    newaclnode->z_acldata, aclp->z_acl_count,
+	    oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
-	    &newaclnode->z_size) == 0);
+	    &newaclnode->z_size, NULL, cr) == 0);
 	newaclnode->z_ace_count = aclp->z_acl_count;
 	aclp->z_version = ZFS_ACL_VERSION;
 	kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
@@ -770,8 +772,7 @@ zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
  * Also, create FUIDs for any User/Group ACEs
  */
 static uint64_t
-zfs_mode_fuid_compute(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
+zfs_mode_compute(znode_t *zp, zfs_acl_t *aclp)
-    zfs_fuid_info_t **fuidp, dmu_tx_t *tx)
 {
 	int		entry_type;
 	mode_t		mode;
@@ -905,15 +906,6 @@ zfs_mode_fuid_compute(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
 				}
 			}
 		}
-		/*
-		 * Now handle FUID create for user/group ACEs
-		 */
-		if (entry_type == 0 || entry_type == ACE_IDENTIFIER_GROUP) {
-			aclp->z_ops.ace_who_set(acep,
-			    zfs_fuid_create(zp->z_zfsvfs, who, cr,
-			    (entry_type == 0) ? ZFS_ACE_USER : ZFS_ACE_GROUP,
-			    tx, fuidp));
-		}
 	}
 	return (mode);
 }
@@ -989,7 +981,7 @@ zfs_acl_node_read(znode_t *zp, zfs_acl_t **aclpp, boolean_t will_modify)
 	aclnode = zfs_acl_node_alloc(aclsize);
 	list_insert_head(&aclp->z_acl, aclnode);
 	error = dmu_read(zp->z_zfsvfs->z_os, extacl, 0,
-	    aclsize, aclnode->z_acldata);
+	    aclsize, aclnode->z_acldata, DMU_READ_PREFETCH);
 	aclnode->z_ace_count = acl_count;
 	aclp->z_acl_count = acl_count;
 	aclp->z_acl_bytes = aclsize;
@@ -1014,8 +1006,7 @@ zfs_acl_node_read(znode_t *zp, zfs_acl_t **aclpp, boolean_t will_modify)
  * already checked the acl and knows whether to inherit.
  */
 int
-zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
+zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
-    zfs_fuid_info_t **fuidp, dmu_tx_t *tx)
 {
 	int		error;
 	znode_phys_t	*zphys = zp->z_phys;
@@ -1026,12 +1017,9 @@ zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
 	dmu_object_type_t otype;
 	zfs_acl_node_t	*aclnode;
 
-	ASSERT(MUTEX_HELD(&zp->z_lock));
-	ASSERT(MUTEX_HELD(&zp->z_acl_lock));
-
 	dmu_buf_will_dirty(zp->z_dbuf, tx);
 
-	zphys->zp_mode = zfs_mode_fuid_compute(zp, aclp, cr, fuidp, tx);
+	zphys->zp_mode = zfs_mode_compute(zp, aclp);
 
 	/*
 	 * Decide which opbject type to use.  If we are forced to
@@ -1043,7 +1031,7 @@ zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
 	} else {
 		if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
 		    (zfsvfs->z_version >= ZPL_VERSION_FUID))
-			zfs_acl_xform(zp, aclp);
+			zfs_acl_xform(zp, aclp, cr);
 		ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
 		otype = DMU_OT_ACL;
 	}
@@ -1125,7 +1113,6 @@ zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr,
 	if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
 		zp->z_phys->zp_flags |= ZFS_ACL_TRIVIAL;
 
-	zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
 	return (0);
 }
 
@@ -1336,7 +1323,7 @@ zfs_acl_ace_insert(zfs_acl_t *aclp, void  *acep)
  * Prepend deny ACE
  */
 static void *
-zfs_acl_prepend_deny(znode_t *zp, zfs_acl_t *aclp, void *acep,
+zfs_acl_prepend_deny(uint64_t uid, zfs_acl_t *aclp, void *acep,
     mode_t mode)
 {
 	zfs_acl_node_t *aclnode;
@@ -1349,7 +1336,7 @@ zfs_acl_prepend_deny(znode_t *zp, zfs_acl_t *aclp, void *acep,
 	fuid = aclp->z_ops.ace_who_get(acep);
 	flags = aclp->z_ops.ace_flags_get(acep);
 	zfs_set_ace(aclp, newacep, 0, DENY, fuid, (flags & ACE_TYPE_FLAGS));
-	zfs_acl_prepend_fixup(aclp, newacep, acep, mode, zp->z_phys->zp_uid);
+	zfs_acl_prepend_fixup(aclp, newacep, acep, mode, uid);
 
 	return (newacep);
 }
@@ -1473,9 +1460,9 @@ zfs_fixup_group_entries(zfs_acl_t *aclp, void *acep, void *prevacep,
  * in PSARC/2002/240
  */
 static void
-zfs_acl_chmod(znode_t *zp, uint64_t mode, zfs_acl_t *aclp)
+zfs_acl_chmod(zfsvfs_t *zfsvfs, uint64_t uid,
+    uint64_t mode, zfs_acl_t *aclp)
 {
-	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
 	void		*acep = NULL, *prevacep = NULL;
 	uint64_t	who;
 	int 		i;
@@ -1485,11 +1472,6 @@ zfs_acl_chmod(znode_t *zp, uint64_t mode, zfs_acl_t *aclp)
 	uint16_t	iflags, type;
 	uint32_t	access_mask;
 
-	ASSERT(MUTEX_HELD(&zp->z_acl_lock));
-	ASSERT(MUTEX_HELD(&zp->z_lock));
-
-	aclp->z_hints = (zp->z_phys->zp_flags & V4_ACL_WIDE_FLAGS);
-
 	/*
 	 * If discard then just discard all ACL nodes which
 	 * represent the ACEs.
@@ -1554,17 +1536,15 @@ zfs_acl_chmod(znode_t *zp, uint64_t mode, zfs_acl_t *aclp)
 
 					if (!reuse_deny) {
 						prevacep =
-						    zfs_acl_prepend_deny(zp,
+						    zfs_acl_prepend_deny(uid,
 						    aclp, acep, mode);
 					} else {
 						zfs_acl_prepend_fixup(
 						    aclp, prevacep,
-						    acep, mode,
+						    acep, mode, uid);
-						    zp->z_phys->zp_uid);
 					}
 					zfs_fixup_group_entries(aclp, acep,
 					    prevacep, mode);
-
 				}
 			}
 		}
@@ -1623,8 +1603,10 @@ zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
 	mutex_enter(&zp->z_acl_lock);
 	*aclp = NULL;
 	error = zfs_acl_node_read(zp, aclp, B_TRUE);
-	if (error == 0)
+	if (error == 0) {
-		zfs_acl_chmod(zp, mode, *aclp);
+		(*aclp)->z_hints = zp->z_phys->zp_flags & V4_ACL_WIDE_FLAGS;
+		zfs_acl_chmod(zp->z_zfsvfs, zp->z_phys->zp_uid, mode, *aclp);
+	}
 	mutex_exit(&zp->z_acl_lock);
 	mutex_exit(&zp->z_lock);
 	return (error);
@@ -1649,9 +1631,8 @@ zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
  * Should ACE be inherited?
  */
 static int
-zfs_ace_can_use(znode_t *zp, uint16_t acep_flags)
+zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
 {
-	int vtype = ZTOV(zp)->v_type;
 	int	iflags = (acep_flags & 0xf);
 
 	if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
@@ -1666,10 +1647,9 @@ zfs_ace_can_use(znode_t *zp, uint16_t acep_flags)
  * inherit inheritable ACEs from parent
  */
 static zfs_acl_t *
-zfs_acl_inherit(znode_t *zp, zfs_acl_t *paclp, uint64_t mode,
+zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
-    boolean_t *need_chmod)
+    uint64_t mode, boolean_t *need_chmod)
 {
-	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
 	void		*pacep;
 	void		*acep, *acep2;
 	zfs_acl_node_t  *aclnode, *aclnode2;
@@ -1680,8 +1660,8 @@ zfs_acl_inherit(znode_t *zp, zfs_acl_t *paclp, uint64_t mode,
 	size_t		ace_size;
 	void		*data1, *data2;
 	size_t		data1sz, data2sz;
-	boolean_t	vdir = ZTOV(zp)->v_type == VDIR;
+	boolean_t	vdir = vtype == VDIR;
-	boolean_t	vreg = ZTOV(zp)->v_type == VREG;
+	boolean_t	vreg = vtype == VREG;
 	boolean_t	passthrough, passthrough_x, noallow;
 
 	passthrough_x =
@@ -1710,7 +1690,7 @@ zfs_acl_inherit(znode_t *zp, zfs_acl_t *paclp, uint64_t mode,
 
 		ace_size = aclp->z_ops.ace_size(pacep);
 
-		if (!zfs_ace_can_use(zp, iflags))
+		if (!zfs_ace_can_use(vtype, iflags))
 			continue;
 
 		/*
@@ -1806,55 +1786,58 @@ zfs_acl_inherit(znode_t *zp, zfs_acl_t *paclp, uint64_t mode,
  * Create file system object initial permissions
  * including inheritable ACEs.
  */
-void
+int
-zfs_perm_init(znode_t *zp, znode_t *parent, int flag,
+zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
-    vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
+    vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
-    zfs_acl_t *setaclp, zfs_fuid_info_t **fuidp)
 {
-	uint64_t	mode, fuid, fgid;
 	int		error;
-	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
+	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
-	zfs_acl_t	*aclp = NULL;
 	zfs_acl_t	*paclp;
-	xvattr_t	*xvap = (xvattr_t *)vap;
 	gid_t		gid;
 	boolean_t	need_chmod = B_TRUE;
 
-	if (setaclp)
+	bzero(acl_ids, sizeof (zfs_acl_ids_t));
-		aclp = setaclp;
+	acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
 
-	mode = MAKEIMODE(vap->va_type, vap->va_mode);
+	if (vsecp)
+		if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
+		    &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
+			return (error);
 
 	/*
 	 * Determine uid and gid.
 	 */
 	if ((flag & (IS_ROOT_NODE | IS_REPLAY)) ||
 	    ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
-		fuid = zfs_fuid_create(zfsvfs, vap->va_uid, cr,
+		acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
-		    ZFS_OWNER, tx, fuidp);
+		    (uint64_t)vap->va_uid, cr,
-		fgid = zfs_fuid_create(zfsvfs, vap->va_gid, cr,
+		    ZFS_OWNER, &acl_ids->z_fuidp);
-		    ZFS_GROUP, tx, fuidp);
+		acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
+		    (uint64_t)vap->va_gid, cr,
+		    ZFS_GROUP, &acl_ids->z_fuidp);
 		gid = vap->va_gid;
 	} else {
-		fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER, tx, cr, fuidp);
+		acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
-		fgid = 0;
+		    cr, &acl_ids->z_fuidp);
+		acl_ids->z_fgid = 0;
 		if (vap->va_mask & AT_GID)  {
-			fgid = zfs_fuid_create(zfsvfs, vap->va_gid, cr,
+			acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
-			    ZFS_GROUP, tx, fuidp);
+			    (uint64_t)vap->va_gid,
+			    cr, ZFS_GROUP, &acl_ids->z_fuidp);
 			gid = vap->va_gid;
-			if (fgid != parent->z_phys->zp_gid &&
+			if (acl_ids->z_fgid != dzp->z_phys->zp_gid &&
 			    !groupmember(vap->va_gid, cr) &&
 			    secpolicy_vnode_create_gid(cr) != 0)
-				fgid = 0;
+				acl_ids->z_fgid = 0;
 		}
-		if (fgid == 0) {
+		if (acl_ids->z_fgid == 0) {
-			if (parent->z_phys->zp_mode & S_ISGID) {
+			if (dzp->z_phys->zp_mode & S_ISGID) {
-				fgid = parent->z_phys->zp_gid;
+				acl_ids->z_fgid = dzp->z_phys->zp_gid;
-				gid = zfs_fuid_map_id(zfsvfs, fgid,
+				gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
 				    cr, ZFS_GROUP);
 			} else {
-				fgid = zfs_fuid_create_cred(zfsvfs,
+				acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
-				    ZFS_GROUP, tx, cr, fuidp);
+				    ZFS_GROUP, cr, &acl_ids->z_fuidp);
 				gid = crgetgid(cr);
 			}
 		}
@@ -1867,57 +1850,61 @@ zfs_perm_init(znode_t *zp, znode_t *parent, int flag,
 	 * file's new group, clear the file's set-GID bit.
 	 */
 
-	if ((parent->z_phys->zp_mode & S_ISGID) && (vap->va_type == VDIR)) {
+	if (!(flag & IS_ROOT_NODE) && (dzp->z_phys->zp_mode & S_ISGID) &&
-		mode |= S_ISGID;
+	    (vap->va_type == VDIR)) {
+		acl_ids->z_mode |= S_ISGID;
 	} else {
-		if ((mode & S_ISGID) &&
+		if ((acl_ids->z_mode & S_ISGID) &&
 		    secpolicy_vnode_setids_setgids(cr, gid) != 0)
-			mode &= ~S_ISGID;
+			acl_ids->z_mode &= ~S_ISGID;
 	}
 
-	zp->z_phys->zp_uid = fuid;
+	if (acl_ids->z_aclp == NULL) {
-	zp->z_phys->zp_gid = fgid;
+		mutex_enter(&dzp->z_lock);
-	zp->z_phys->zp_mode = mode;
+		if (!(flag & IS_ROOT_NODE) && (ZTOV(dzp)->v_type == VDIR &&
-
+		    (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE)) &&
-	if (aclp == NULL) {
+		    !(dzp->z_phys->zp_flags & ZFS_XATTR)) {
-		mutex_enter(&parent->z_lock);
+			mutex_enter(&dzp->z_acl_lock);
-		if ((ZTOV(parent)->v_type == VDIR &&
+			VERIFY(0 == zfs_acl_node_read(dzp, &paclp, B_FALSE));
-		    (parent->z_phys->zp_flags & ZFS_INHERIT_ACE)) &&
+			mutex_exit(&dzp->z_acl_lock);
-		    !(zp->z_phys->zp_flags & ZFS_XATTR)) {
+			acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
-			mutex_enter(&parent->z_acl_lock);
+			    vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
-			VERIFY(0 == zfs_acl_node_read(parent, &paclp, B_FALSE));
-			mutex_exit(&parent->z_acl_lock);
-			aclp = zfs_acl_inherit(zp, paclp, mode, &need_chmod);
 			zfs_acl_free(paclp);
 		} else {
-			aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
+			acl_ids->z_aclp =
+			    zfs_acl_alloc(zfs_acl_version_zp(dzp));
+		}
+		mutex_exit(&dzp->z_lock);
+		if (need_chmod) {
+			acl_ids->z_aclp->z_hints = (vap->va_type == VDIR) ?
+			    ZFS_ACL_AUTO_INHERIT : 0;
+			zfs_acl_chmod(zfsvfs, acl_ids->z_fuid,
+			    acl_ids->z_mode, acl_ids->z_aclp);
 		}
-		mutex_exit(&parent->z_lock);
-		mutex_enter(&zp->z_lock);
-		mutex_enter(&zp->z_acl_lock);
-		if (need_chmod)
-			zfs_acl_chmod(zp, mode, aclp);
-	} else {
-		mutex_enter(&zp->z_lock);
-		mutex_enter(&zp->z_acl_lock);
 	}
 
-	/* Force auto_inherit on all new directory objects */
+	return (0);
-	if (vap->va_type == VDIR)
+}
-		aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
 
-	error = zfs_aclset_common(zp, aclp, cr, fuidp, tx);
+/*
+ * Free ACL and fuid_infop, but not the acl_ids structure
+ */
+void
+zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
+{
+	if (acl_ids->z_aclp)
+		zfs_acl_free(acl_ids->z_aclp);
+	if (acl_ids->z_fuidp)
+		zfs_fuid_info_free(acl_ids->z_fuidp);
+	acl_ids->z_aclp = NULL;
+	acl_ids->z_fuidp = NULL;
+}
 
-	/* Set optional attributes if any */
+boolean_t
-	if (vap->va_mask & AT_XVATTR)
+zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
-		zfs_xvattr_set(zp, xvap);
+{
-
+	return (zfs_usergroup_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
-	mutex_exit(&zp->z_lock);
+	    zfs_usergroup_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
-	mutex_exit(&zp->z_acl_lock);
-	ASSERT3U(error, ==, 0);
-
-	if (aclp != setaclp)
-		zfs_acl_free(aclp);
 }
 
 /*
@@ -2018,7 +2005,7 @@ zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
 
 int
 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
-    vsecattr_t *vsecp, zfs_acl_t **zaclp)
+    vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
 {
 	zfs_acl_t *aclp;
 	zfs_acl_node_t *aclnode;
@@ -2041,9 +2028,9 @@ zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
 			return (error);
 		}
 	} else {
-		if ((error = zfs_copy_ace_2_fuid(obj_type, aclp,
+		if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
 		    vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
-		    &aclnode->z_size)) != 0) {
+		    &aclnode->z_size, fuidp, cr)) != 0) {
 			zfs_acl_free(aclp);
 			zfs_acl_node_free(aclnode);
 			return (error);
@@ -2084,6 +2071,7 @@ zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
 	int		error;
 	zfs_acl_t	*aclp;
 	zfs_fuid_info_t	*fuidp = NULL;
+	boolean_t	fuid_dirtied;
 
 	if (mask == 0)
 		return (ENOSYS);
@@ -2094,7 +2082,8 @@ zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
 	if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr))
 		return (error);
 
-	error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, &aclp);
+	error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
+	    &aclp);
 	if (error)
 		return (error);
 
@@ -2135,18 +2124,9 @@ top:
 	} else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
 		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
 	}
-	if (aclp->z_has_fuids) {
+	fuid_dirtied = zfsvfs->z_fuid_dirty;
-		if (zfsvfs->z_fuid_obj == 0) {
+	if (fuid_dirtied)
-			dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+		zfs_fuid_txhold(zfsvfs, tx);
-			dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-			dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
-		} else {
-			dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-			dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-		}
-	}
 
 	error = dmu_tx_assign(tx, TXG_NOWAIT);
 	if (error) {
@@ -2163,9 +2143,13 @@ top:
 		return (error);
 	}
 
-	error = zfs_aclset_common(zp, aclp, cr, &fuidp, tx);
+	error = zfs_aclset_common(zp, aclp, cr, tx);
 	ASSERT(error == 0);
 
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
+
+	zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
 	zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
 
 	if (fuidp)
@@ -2180,45 +2164,17 @@ done:
 }
 
 /*
- * working_mode returns the permissions that were not granted
+ * Check accesses of interest (AoI) against attributes of the dataset
+ * such as read-only.  Returns zero if no AoI conflict with dataset
+ * attributes, otherwise an appropriate errno is returned.
  */
 static int
-zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
+zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
-    boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
 {
-	zfs_acl_t	*aclp;
-	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
-	int		error;
-	uid_t		uid = crgetuid(cr);
-	uint64_t 	who;
-	uint16_t	type, iflags;
-	uint16_t	entry_type;
-	uint32_t	access_mask;
-	uint32_t	deny_mask = 0;
-	zfs_ace_hdr_t	*acep = NULL;
-	boolean_t	checkit;
-	uid_t		fowner;
-	uid_t		gowner;
-
-	/*
-	 * Short circuit empty requests
-	 */
-	if (v4_mode == 0)
-		return (0);
-
-	*check_privs = B_TRUE;
-
-	if (zfsvfs->z_replay) {
-		*working_mode = 0;
-		return (0);
-	}
-
-	*working_mode = v4_mode;
-
 	if ((v4_mode & WRITE_MASK) &&
 	    (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
-	    (!IS_DEVVP(ZTOV(zp)))) {
+	    (!IS_DEVVP(ZTOV(zp)) ||
-		*check_privs = B_FALSE;
+	    (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
 		return (EROFS);
 	}
 
@@ -2230,31 +2186,64 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
 	    (zp->z_phys->zp_flags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
 	    (ZTOV(zp)->v_type == VDIR &&
 	    (zp->z_phys->zp_flags & ZFS_IMMUTABLE)))) {
-		*check_privs = B_FALSE;
 		return (EPERM);
 	}
 
 	if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
 	    (zp->z_phys->zp_flags & ZFS_NOUNLINK)) {
-		*check_privs = B_FALSE;
 		return (EPERM);
 	}
 
 	if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
 	    (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED))) {
-		*check_privs = B_FALSE;
 		return (EACCES);
 	}
 
-	/*
+	return (0);
-	 * The caller requested that the ACL check be skipped.  This
+}
-	 * would only happen if the caller checked VOP_ACCESS() with a
+
-	 * 32 bit ACE mask and already had the appropriate permissions.
+/*
-	 */
+ * The primary usage of this function is to loop through all of the
-	if (skipaclchk) {
+ * ACEs in the znode, determining what accesses of interest (AoI) to
-		*working_mode = 0;
+ * the caller are allowed or denied.  The AoI are expressed as bits in
-		return (0);
+ * the working_mode parameter.  As each ACE is processed, bits covered
-	}
+ * by that ACE are removed from the working_mode.  This removal
+ * facilitates two things.  The first is that when the working mode is
+ * empty (= 0), we know we've looked at all the AoI. The second is
+ * that the ACE interpretation rules don't allow a later ACE to undo
+ * something granted or denied by an earlier ACE.  Removing the
+ * discovered access or denial enforces this rule.  At the end of
+ * processing the ACEs, all AoI that were found to be denied are
+ * placed into the working_mode, giving the caller a mask of denied
+ * accesses.  Returns:
+ *	0		if all AoI granted
+ *	EACCESS 	if the denied mask is non-zero
+ *	other error	if abnormal failure (e.g., IO error)
+ *
+ * A secondary usage of the function is to determine if any of the
+ * AoI are granted.  If an ACE grants any access in
+ * the working_mode, we immediately short circuit out of the function.
+ * This mode is chosen by setting anyaccess to B_TRUE.  The
+ * working_mode is not a denied access mask upon exit if the function
+ * is used in this manner.
+ */
+static int
+zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
+    boolean_t anyaccess, cred_t *cr)
+{
+	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
+	zfs_acl_t	*aclp;
+	int		error;
+	uid_t		uid = crgetuid(cr);
+	uint64_t 	who;
+	uint16_t	type, iflags;
+	uint16_t	entry_type;
+	uint32_t	access_mask;
+	uint32_t	deny_mask = 0;
+	zfs_ace_hdr_t	*acep = NULL;
+	boolean_t	checkit;
+	uid_t		fowner;
+	uid_t		gowner;
 
 	zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
 
@@ -2268,6 +2257,7 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
 
 	while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
 	    &iflags, &type)) {
+		uint32_t mask_matched;
 
 		if (!zfs_acl_valid_ace_type(type, iflags))
 			continue;
@@ -2275,6 +2265,11 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
 		if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
 			continue;
 
+		/* Skip ACE if it does not affect any AoI */
+		mask_matched = (access_mask & *working_mode);
+		if (!mask_matched)
+			continue;
+
 		entry_type = (iflags & ACE_TYPE_FLAGS);
 
 		checkit = B_FALSE;
@@ -2313,14 +2308,24 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
 		}
 
 		if (checkit) {
-			uint32_t mask_matched = (access_mask & *working_mode);
+			if (type == DENY) {
-
+				DTRACE_PROBE3(zfs__ace__denies,
-			if (mask_matched) {
+				    znode_t *, zp,
-				if (type == DENY)
+				    zfs_ace_hdr_t *, acep,
-					deny_mask |= mask_matched;
+				    uint32_t, mask_matched);
-
+				deny_mask |= mask_matched;
-				*working_mode &= ~mask_matched;
+			} else {
+				DTRACE_PROBE3(zfs__ace__allows,
+				    znode_t *, zp,
+				    zfs_ace_hdr_t *, acep,
+				    uint32_t, mask_matched);
+				if (anyaccess) {
+					mutex_exit(&zp->z_acl_lock);
+					zfs_acl_free(aclp);
+					return (0);
+				}
 			}
+			*working_mode &= ~mask_matched;
 		}
 
 		/* Are we done? */
@@ -2342,6 +2347,69 @@ zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
 	return (0);
 }
 
+/*
+ * Return true if any access whatsoever granted, we don't actually
+ * care what access is granted.
+ */
+boolean_t
+zfs_has_access(znode_t *zp, cred_t *cr)
+{
+	uint32_t have = ACE_ALL_PERMS;
+
+	if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
+		uid_t		owner;
+
+		owner = zfs_fuid_map_id(zp->z_zfsvfs,
+		    zp->z_phys->zp_uid, cr, ZFS_OWNER);
+
+		return (
+		    secpolicy_vnode_access(cr, ZTOV(zp), owner, VREAD) == 0 ||
+		    secpolicy_vnode_access(cr, ZTOV(zp), owner, VWRITE) == 0 ||
+		    secpolicy_vnode_access(cr, ZTOV(zp), owner, VEXEC) == 0 ||
+		    secpolicy_vnode_chown(cr, B_TRUE) == 0 ||
+		    secpolicy_vnode_chown(cr, B_FALSE) == 0 ||
+		    secpolicy_vnode_setdac(cr, owner) == 0 ||
+		    secpolicy_vnode_remove(cr) == 0);
+	}
+	return (B_TRUE);
+}
+
+static int
+zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
+    boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
+{
+	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
+	int err;
+
+	*working_mode = v4_mode;
+	*check_privs = B_TRUE;
+
+	/*
+	 * Short circuit empty requests
+	 */
+	if (v4_mode == 0 || zfsvfs->z_replay) {
+		*working_mode = 0;
+		return (0);
+	}
+
+	if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
+		*check_privs = B_FALSE;
+		return (err);
+	}
+
+	/*
+	 * The caller requested that the ACL check be skipped.  This
+	 * would only happen if the caller checked VOP_ACCESS() with a
+	 * 32 bit ACE mask and already had the appropriate permissions.
+	 */
+	if (skipaclchk) {
+		*working_mode = 0;
+		return (0);
+	}
+
+	return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
+}
+
 static int
 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
     cred_t *cr)

module/zfs/zfs_ctldir.c

@@ -114,12 +114,16 @@ snapentry_compare(const void *a, const void *b)
 vnodeops_t *zfsctl_ops_root;
 vnodeops_t *zfsctl_ops_snapdir;
 vnodeops_t *zfsctl_ops_snapshot;
+vnodeops_t *zfsctl_ops_shares;
+vnodeops_t *zfsctl_ops_shares_dir;
 
 static const fs_operation_def_t zfsctl_tops_root[];
 static const fs_operation_def_t zfsctl_tops_snapdir[];
 static const fs_operation_def_t zfsctl_tops_snapshot[];
+static const fs_operation_def_t zfsctl_tops_shares[];
 
 static vnode_t *zfsctl_mknode_snapdir(vnode_t *);
+static vnode_t *zfsctl_mknode_shares(vnode_t *);
 static vnode_t *zfsctl_snapshot_mknode(vnode_t *, uint64_t objset);
 static int zfsctl_unmount_snap(zfs_snapentry_t *, int, cred_t *);
 
@@ -127,14 +131,18 @@ static gfs_opsvec_t zfsctl_opsvec[] = {
 	{ ".zfs", zfsctl_tops_root, &zfsctl_ops_root },
 	{ ".zfs/snapshot", zfsctl_tops_snapdir, &zfsctl_ops_snapdir },
 	{ ".zfs/snapshot/vnode", zfsctl_tops_snapshot, &zfsctl_ops_snapshot },
+	{ ".zfs/shares", zfsctl_tops_shares, &zfsctl_ops_shares_dir },
+	{ ".zfs/shares/vnode", zfsctl_tops_shares, &zfsctl_ops_shares },
 	{ NULL }
 };
 
 /*
- * Root directory elements.  We have only a single static entry, 'snapshot'.
+ * Root directory elements.  We only have two entries
+ * snapshot and shares.
  */
 static gfs_dirent_t zfsctl_root_entries[] = {
 	{ "snapshot", zfsctl_mknode_snapdir, GFS_CACHE_VNODE },
+	{ "shares", zfsctl_mknode_shares, GFS_CACHE_VNODE },
 	{ NULL }
 };
 
@@ -166,21 +174,34 @@ zfsctl_fini(void)
 		vn_freevnodeops(zfsctl_ops_snapdir);
 	if (zfsctl_ops_snapshot)
 		vn_freevnodeops(zfsctl_ops_snapshot);
+	if (zfsctl_ops_shares)
+		vn_freevnodeops(zfsctl_ops_shares);
+	if (zfsctl_ops_shares_dir)
+		vn_freevnodeops(zfsctl_ops_shares_dir);
 
 	zfsctl_ops_root = NULL;
 	zfsctl_ops_snapdir = NULL;
 	zfsctl_ops_snapshot = NULL;
+	zfsctl_ops_shares = NULL;
+	zfsctl_ops_shares_dir = NULL;
 }
 
 /*
- * Return the inode number associated with the 'snapshot' directory.
+ * Return the inode number associated with the 'snapshot' or
+ * 'shares' directory.
  */
 /* ARGSUSED */
 static ino64_t
 zfsctl_root_inode_cb(vnode_t *vp, int index)
 {
-	ASSERT(index == 0);
+	zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
-	return (ZFSCTL_INO_SNAPDIR);
+
+	ASSERT(index <= 2);
+
+	if (index == 0)
+		return (ZFSCTL_INO_SNAPDIR);
+
+	return (zfsvfs->z_shares_dir);
 }
 
 /*
@@ -348,6 +369,30 @@ zfsctl_common_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
 	return (0);
 }
 
+
+/*ARGSUSED*/
+static int
+zfsctl_shares_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
+{
+	zfsvfs_t	*zfsvfs = vp->v_vfsp->vfs_data;
+	znode_t		*dzp;
+	int		error;
+
+	ZFS_ENTER(zfsvfs);
+
+	if (zfsvfs->z_shares_dir == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (ENOTSUP);
+	}
+
+	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
+		error = VOP_FID(ZTOV(dzp), fidp, ct);
+		VN_RELE(ZTOV(dzp));
+	}
+
+	ZFS_EXIT(zfsvfs);
+	return (error);
+}
 /*
  * .zfs inode namespace
  *
@@ -478,7 +523,7 @@ zfsctl_unmount_snap(zfs_snapentry_t *sep, int fflags, cred_t *cr)
 		VN_RELE(svp);
 		return (error);
 	}
-	VFS_RELE(svp->v_vfsp);
+
 	/*
 	 * We can't use VN_RELE(), as that will try to invoke
 	 * zfsctl_snapdir_inactive(), which would cause us to destroy
@@ -691,7 +736,7 @@ zfsctl_snapdir_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t  **vpp,
 		return (err);
 
 	if (err == 0) {
-		err = dmu_objset_snapshot(name, dirname, B_FALSE);
+		err = dmu_objset_snapshot(name, dirname, NULL, B_FALSE);
 		if (err)
 			return (err);
 		err = lookupnameat(dirname, seg, follow, NULL, vpp, dvp);
@@ -732,9 +777,6 @@ zfsctl_snapdir_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
 
 	ASSERT(dvp->v_type == VDIR);
 
-	if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0)
-		return (0);
-
 	/*
 	 * If we get a recursive call, that means we got called
 	 * from the domount() code while it was trying to look up the
@@ -746,6 +788,11 @@ zfsctl_snapdir_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
 
 	ZFS_ENTER(zfsvfs);
 
+	if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (0);
+	}
+
 	if (flags & FIGNORECASE) {
 		boolean_t conflict = B_FALSE;
 
@@ -844,7 +891,7 @@ domount:
 		 * Return the mounted root rather than the covered mount point.
 		 * Takes the GFS vnode at .zfs/snapshot/<snapname> and returns
 		 * the ZFS vnode mounted on top of the GFS node.  This ZFS
-		 * vnode is the root the newly created vfsp.
+		 * vnode is the root of the newly created vfsp.
 		 */
 		VFS_RELE(vfsp);
 		err = traverse(vpp);
@@ -877,6 +924,37 @@ domount:
 	return (err);
 }
 
+/* ARGSUSED */
+static int
+zfsctl_shares_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
+    int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
+    int *direntflags, pathname_t *realpnp)
+{
+	zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
+	znode_t *dzp;
+	int error;
+
+	ZFS_ENTER(zfsvfs);
+
+	if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (0);
+	}
+
+	if (zfsvfs->z_shares_dir == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (ENOTSUP);
+	}
+	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0)
+		error = VOP_LOOKUP(ZTOV(dzp), nm, vpp, pnp,
+		    flags, rdir, cr, ct, direntflags, realpnp);
+
+	VN_RELE(ZTOV(dzp));
+	ZFS_EXIT(zfsvfs);
+
+	return (error);
+}
+
 /* ARGSUSED */
 static int
 zfsctl_snapdir_readdir_cb(vnode_t *vp, void *dp, int *eofp,
@@ -921,6 +999,33 @@ zfsctl_snapdir_readdir_cb(vnode_t *vp, void *dp, int *eofp,
 	return (0);
 }
 
+/* ARGSUSED */
+static int
+zfsctl_shares_readdir(vnode_t *vp, uio_t *uiop, cred_t *cr, int *eofp,
+    caller_context_t *ct, int flags)
+{
+	zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+	znode_t *dzp;
+	int error;
+
+	ZFS_ENTER(zfsvfs);
+
+	if (zfsvfs->z_shares_dir == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (ENOTSUP);
+	}
+	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
+		error = VOP_READDIR(ZTOV(dzp), uiop, cr, eofp, ct, flags);
+		VN_RELE(ZTOV(dzp));
+	} else {
+		*eofp = 1;
+		error = ENOENT;
+	}
+
+	ZFS_EXIT(zfsvfs);
+	return (error);
+}
+
 /*
  * pvp is the '.zfs' directory (zfsctl_node_t).
  * Creates vp, which is '.zfs/snapshot' (zfsctl_snapdir_t).
@@ -946,6 +1051,45 @@ zfsctl_mknode_snapdir(vnode_t *pvp)
 	return (vp);
 }
 
+vnode_t *
+zfsctl_mknode_shares(vnode_t *pvp)
+{
+	vnode_t *vp;
+	zfsctl_node_t *sdp;
+
+	vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp,
+	    zfsctl_ops_shares, NULL, NULL, MAXNAMELEN,
+	    NULL, NULL);
+	sdp = vp->v_data;
+	sdp->zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
+	return (vp);
+
+}
+
+/* ARGSUSED */
+static int
+zfsctl_shares_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
+    caller_context_t *ct)
+{
+	zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+	znode_t *dzp;
+	int error;
+
+	ZFS_ENTER(zfsvfs);
+	if (zfsvfs->z_shares_dir == 0) {
+		ZFS_EXIT(zfsvfs);
+		return (ENOTSUP);
+	}
+	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
+		error = VOP_GETATTR(ZTOV(dzp), vap, flags, cr, ct);
+		VN_RELE(ZTOV(dzp));
+	}
+	ZFS_EXIT(zfsvfs);
+	return (error);
+
+
+}
+
 /* ARGSUSED */
 static int
 zfsctl_snapdir_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
@@ -996,6 +1140,20 @@ static const fs_operation_def_t zfsctl_tops_snapdir[] = {
 	{ NULL }
 };
 
+static const fs_operation_def_t zfsctl_tops_shares[] = {
+	{ VOPNAME_OPEN,		{ .vop_open = zfsctl_common_open }	},
+	{ VOPNAME_CLOSE,	{ .vop_close = zfsctl_common_close }	},
+	{ VOPNAME_IOCTL,	{ .error = fs_inval }			},
+	{ VOPNAME_GETATTR,	{ .vop_getattr = zfsctl_shares_getattr } },
+	{ VOPNAME_ACCESS,	{ .vop_access = zfsctl_common_access }	},
+	{ VOPNAME_READDIR,	{ .vop_readdir = zfsctl_shares_readdir } },
+	{ VOPNAME_LOOKUP,	{ .vop_lookup = zfsctl_shares_lookup }	},
+	{ VOPNAME_SEEK,		{ .vop_seek = fs_seek }			},
+	{ VOPNAME_INACTIVE,	{ .vop_inactive = gfs_vop_inactive } },
+	{ VOPNAME_FID,		{ .vop_fid = zfsctl_shares_fid } },
+	{ NULL }
+};
+
 /*
  * pvp is the GFS vnode '.zfs/snapshot'.
  *
@@ -1013,7 +1171,6 @@ zfsctl_snapshot_mknode(vnode_t *pvp, uint64_t objset)
 	    zfsctl_ops_snapshot, NULL, NULL, MAXNAMELEN, NULL, NULL);
 	zcp = vp->v_data;
 	zcp->zc_id = objset;
-	VFS_HOLD(vp->v_vfsp);
 
 	return (vp);
 }
@@ -1052,7 +1209,6 @@ zfsctl_snapshot_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
 
 	mutex_exit(&sdp->sd_lock);
 	VN_RELE(dvp);
-	VFS_RELE(vp->v_vfsp);
 
 	/*
 	 * Dispose of the vnode for the snapshot mount point.

module/zfs/zfs_dir.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -805,44 +805,49 @@ zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
 	znode_t *xzp;
 	dmu_tx_t *tx;
 	int error;
-	zfs_fuid_info_t *fuidp = NULL;
+	zfs_acl_ids_t acl_ids;
+	boolean_t fuid_dirtied;
 
 	*xvpp = NULL;
 
 	if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
 		return (error);
 
+	if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
+	    &acl_ids)) != 0)
+		return (error);
+	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
+		zfs_acl_ids_free(&acl_ids);
+		return (EDQUOT);
+	}
+
 	tx = dmu_tx_create(zfsvfs->z_os);
 	dmu_tx_hold_bonus(tx, zp->z_id);
 	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
-	if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
+	fuid_dirtied = zfsvfs->z_fuid_dirty;
-		if (zfsvfs->z_fuid_obj == 0) {
+	if (fuid_dirtied)
-			dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+		zfs_fuid_txhold(zfsvfs, tx);
-			dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-			dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
-		} else {
-			dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-			dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-		}
-	}
 	error = dmu_tx_assign(tx, TXG_NOWAIT);
 	if (error) {
+		zfs_acl_ids_free(&acl_ids);
 		if (error == ERESTART)
 			dmu_tx_wait(tx);
 		dmu_tx_abort(tx);
 		return (error);
 	}
-	zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, NULL, &fuidp);
+	zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, 0, &acl_ids);
+
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
+
 	ASSERT(xzp->z_phys->zp_parent == zp->z_id);
 	dmu_buf_will_dirty(zp->z_dbuf, tx);
 	zp->z_phys->zp_xattr = xzp->z_id;
 
 	(void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
-	    xzp, "", NULL, fuidp, vap);
+	    xzp, "", NULL, acl_ids.z_fuidp, vap);
-	if (fuidp)
+
-		zfs_fuid_info_free(fuidp);
+	zfs_acl_ids_free(&acl_ids);
 	dmu_tx_commit(tx);
 
 	*xvpp = ZTOV(xzp);

module/zfs/zfs_fm.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -96,7 +96,6 @@ zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
 	nvlist_t *ereport, *detector;
 	uint64_t ena;
 	char class[64];
-	int state;
 
 	/*
 	 * If we are doing a spa_tryimport(), ignore errors.
@@ -130,15 +129,39 @@ zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
 			return;
 
 		/*
-		 * If the vdev has already been marked as failing due to a
+		 * If this I/O is not a retry I/O, don't post an ereport.
-		 * failed probe, then ignore any subsequent I/O errors, as the
+		 * Otherwise, we risk making bad diagnoses based on B_FAILFAST
-		 * DE will automatically fault the vdev on the first such
+		 * I/Os.
-		 * failure.
 		 */
-		if (vd != NULL &&
+		if (zio->io_error == EIO &&
-		    (!vdev_readable(vd) || !vdev_writeable(vd)) &&
+		    !(zio->io_flags & ZIO_FLAG_IO_RETRY))
-		    strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) != 0)
 			return;
+
+		if (vd != NULL) {
+			/*
+			 * If the vdev has already been marked as failing due
+			 * to a failed probe, then ignore any subsequent I/O
+			 * errors, as the DE will automatically fault the vdev
+			 * on the first such failure.  This also catches cases
+			 * where vdev_remove_wanted is set and the device has
+			 * not yet been asynchronously placed into the REMOVED
+			 * state.
+			 */
+			if (zio->io_vd == vd &&
+			    !vdev_accessible(vd, zio) &&
+			    strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) != 0)
+				return;
+
+			/*
+			 * Ignore checksum errors for reads from DTL regions of
+			 * leaf vdevs.
+			 */
+			if (zio->io_type == ZIO_TYPE_READ &&
+			    zio->io_error == ECKSUM &&
+			    vd->vdev_ops->vdev_op_leaf &&
+			    vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1))
+				return;
+		}
 	}
 
 	if ((ereport = fm_nvlist_create(NULL)) == NULL)
@@ -188,14 +211,6 @@ zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
 	 * passed in.
 	 */
 
-	/*
-	 * If we are importing a faulted pool, then we treat it like an open,
-	 * not an import.  Otherwise, the DE will ignore all faults during
-	 * import, since the default behavior is to mark the devices as
-	 * persistently unavailable, not leave them in the faulted state.
-	 */
-	state = spa->spa_import_faulted ? SPA_LOAD_OPEN : spa->spa_load_state;
-
 	/*
 	 * Generic payload members common to all ereports.
 	 */
@@ -203,7 +218,7 @@ zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
 	    DATA_TYPE_STRING, spa_name(spa), FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
 	    DATA_TYPE_UINT64, spa_guid(spa),
 	    FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
-	    state, NULL);
+	    spa->spa_load_state, NULL);
 
 	if (spa != NULL) {
 		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
@@ -222,14 +237,18 @@ zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
 		    DATA_TYPE_UINT64, vd->vdev_guid,
 		    FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
 		    DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
-		if (vd->vdev_path)
+		if (vd->vdev_path != NULL)
 			fm_payload_set(ereport,
 			    FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
 			    DATA_TYPE_STRING, vd->vdev_path, NULL);
-		if (vd->vdev_devid)
+		if (vd->vdev_devid != NULL)
 			fm_payload_set(ereport,
 			    FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
 			    DATA_TYPE_STRING, vd->vdev_devid, NULL);
+		if (vd->vdev_fru != NULL)
+			fm_payload_set(ereport,
+			    FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU,
+			    DATA_TYPE_STRING, vd->vdev_fru, NULL);
 
 		if (pvd != NULL) {
 			fm_payload_set(ereport,

module/zfs/zfs_fuid.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -47,8 +47,10 @@
  * During file system initialization the nvlist(s) are read and
  * two AVL trees are created.  One tree is keyed by the index number
  * and the other by the domain string.  Nodes are never removed from
- * trees, but new entries may be added.  If a new entry is added then the
+ * trees, but new entries may be added.  If a new entry is added then
- * on-disk packed nvlist will also be updated.
+ * the zfsvfs->z_fuid_dirty flag is set to true and the caller will then
+ * be responsible for calling zfs_fuid_sync() to sync the changes to disk.
+ *
  */
 
 #define	FUID_IDX	"fuid_idx"
@@ -97,6 +99,15 @@ domain_compare(const void *arg1, const void *arg2)
 	return (val > 0 ? 1 : -1);
 }
 
+void
+zfs_fuid_avl_tree_create(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
+{
+	avl_create(idx_tree, idx_compare,
+	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
+	avl_create(domain_tree, domain_compare,
+	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
+}
+
 /*
  * load initial fuid domain and idx trees.  This function is used by
  * both the kernel and zdb.
@@ -108,12 +119,9 @@ zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
 	dmu_buf_t *db;
 	uint64_t fuid_size;
 
-	avl_create(idx_tree, idx_compare,
+	ASSERT(fuid_obj != 0);
-	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
+	VERIFY(0 == dmu_bonus_hold(os, fuid_obj,
-	avl_create(domain_tree, domain_compare,
+	    FTAG, &db));
-	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
-
-	VERIFY(0 == dmu_bonus_hold(os, fuid_obj, FTAG, &db));
 	fuid_size = *(uint64_t *)db->db_data;
 	dmu_buf_rele(db, FTAG);
 
@@ -125,7 +133,8 @@ zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
 		int i;
 
 		packed = kmem_alloc(fuid_size, KM_SLEEP);
-		VERIFY(dmu_read(os, fuid_obj, 0, fuid_size, packed) == 0);
+		VERIFY(dmu_read(os, fuid_obj, 0,
+		    fuid_size, packed, DMU_READ_PREFETCH) == 0);
 		VERIFY(nvlist_unpack(packed, fuid_size,
 		    &nvp, 0) == 0);
 		VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
@@ -189,10 +198,8 @@ zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
  * Load the fuid table(s) into memory.
  */
 static void
-zfs_fuid_init(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
+zfs_fuid_init(zfsvfs_t *zfsvfs)
 {
-	int error = 0;
-
 	rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
 
 	if (zfsvfs->z_fuid_loaded) {
@@ -200,41 +207,101 @@ zfs_fuid_init(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
 		return;
 	}
 
-	if (zfsvfs->z_fuid_obj == 0) {
+	zfs_fuid_avl_tree_create(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
-
-		/* first make sure we need to allocate object */
-
-		error = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
-		    ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
-		if (error == ENOENT && tx != NULL) {
-			zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
-			    DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
-			    sizeof (uint64_t), tx);
-			VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
-			    ZFS_FUID_TABLES, sizeof (uint64_t), 1,
-			    &zfsvfs->z_fuid_obj, tx) == 0);
-		}
-	}
 
+	(void) zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
+	    ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
 	if (zfsvfs->z_fuid_obj != 0) {
 		zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
 		    zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
 		    &zfsvfs->z_fuid_domain);
-		zfsvfs->z_fuid_loaded = B_TRUE;
 	}
 
+	zfsvfs->z_fuid_loaded = B_TRUE;
+	rw_exit(&zfsvfs->z_fuid_lock);
+}
+
+/*
+ * sync out AVL trees to persistent storage.
+ */
+void
+zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
+{
+	nvlist_t *nvp;
+	nvlist_t **fuids;
+	size_t nvsize = 0;
+	char *packed;
+	dmu_buf_t *db;
+	fuid_domain_t *domnode;
+	int numnodes;
+	int i;
+
+	if (!zfsvfs->z_fuid_dirty) {
+		return;
+	}
+
+	rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
+
+	/*
+	 * First see if table needs to be created?
+	 */
+	if (zfsvfs->z_fuid_obj == 0) {
+		zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
+		    DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
+		    sizeof (uint64_t), tx);
+		VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
+		    ZFS_FUID_TABLES, sizeof (uint64_t), 1,
+		    &zfsvfs->z_fuid_obj, tx) == 0);
+	}
+
+	VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+
+	numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
+	fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
+	for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
+	    domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
+		VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
+		VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
+		    domnode->f_idx) == 0);
+		VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
+		VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
+		    domnode->f_ksid->kd_name) == 0);
+	}
+	VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
+	    fuids, numnodes) == 0);
+	for (i = 0; i != numnodes; i++)
+		nvlist_free(fuids[i]);
+	kmem_free(fuids, numnodes * sizeof (void *));
+	VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
+	packed = kmem_alloc(nvsize, KM_SLEEP);
+	VERIFY(nvlist_pack(nvp, &packed, &nvsize,
+	    NV_ENCODE_XDR, KM_SLEEP) == 0);
+	nvlist_free(nvp);
+	zfsvfs->z_fuid_size = nvsize;
+	dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
+	    zfsvfs->z_fuid_size, packed, tx);
+	kmem_free(packed, zfsvfs->z_fuid_size);
+	VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
+	    FTAG, &db));
+	dmu_buf_will_dirty(db, tx);
+	*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
+	dmu_buf_rele(db, FTAG);
+
+	zfsvfs->z_fuid_dirty = B_FALSE;
 	rw_exit(&zfsvfs->z_fuid_lock);
 }
 
 /*
  * Query domain table for a given domain.
  *
- * If domain isn't found it is added to AVL trees and
+ * If domain isn't found and addok is set, it is added to AVL trees and
- * the results are pushed out to disk.
+ * the zfsvfs->z_fuid_dirty flag will be set to TRUE.  It will then be
+ * necessary for the caller or another thread to detect the dirty table
+ * and sync out the changes.
  */
 int
-zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
+zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain,
-    dmu_tx_t *tx)
+    char **retdomain, boolean_t addok)
 {
 	fuid_domain_t searchnode, *findnode;
 	avl_index_t loc;
@@ -246,16 +313,16 @@ zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
 	 * for the user nobody.
 	 */
 	if (domain[0] == '\0') {
-		*retdomain = nulldomain;
+		if (retdomain)
+			*retdomain = nulldomain;
 		return (0);
 	}
 
 	searchnode.f_ksid = ksid_lookupdomain(domain);
-	if (retdomain) {
+	if (retdomain)
 		*retdomain = searchnode.f_ksid->kd_name;
-	}
 	if (!zfsvfs->z_fuid_loaded)
-		zfs_fuid_init(zfsvfs, tx);
+		zfs_fuid_init(zfsvfs);
 
 retry:
 	rw_enter(&zfsvfs->z_fuid_lock, rw);
@@ -265,15 +332,9 @@ retry:
 		rw_exit(&zfsvfs->z_fuid_lock);
 		ksiddomain_rele(searchnode.f_ksid);
 		return (findnode->f_idx);
-	} else {
+	} else if (addok) {
 		fuid_domain_t *domnode;
-		nvlist_t *nvp;
-		nvlist_t **fuids;
 		uint64_t retidx;
-		size_t nvsize = 0;
-		char *packed;
-		dmu_buf_t *db;
-		int i = 0;
 
 		if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
 			rw_exit(&zfsvfs->z_fuid_lock);
@@ -288,46 +349,11 @@ retry:
 
 		avl_add(&zfsvfs->z_fuid_domain, domnode);
 		avl_add(&zfsvfs->z_fuid_idx, domnode);
-		/*
+		zfsvfs->z_fuid_dirty = B_TRUE;
-		 * Now resync the on-disk nvlist.
-		 */
-		VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-
-		domnode = avl_first(&zfsvfs->z_fuid_domain);
-		fuids = kmem_alloc(retidx * sizeof (void *), KM_SLEEP);
-		while (domnode) {
-			VERIFY(nvlist_alloc(&fuids[i],
-			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
-			VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
-			    domnode->f_idx) == 0);
-			VERIFY(nvlist_add_uint64(fuids[i],
-			    FUID_OFFSET, 0) == 0);
-			VERIFY(nvlist_add_string(fuids[i++], FUID_DOMAIN,
-			    domnode->f_ksid->kd_name) == 0);
-			domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode);
-		}
-		VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
-		    fuids, retidx) == 0);
-		for (i = 0; i != retidx; i++)
-			nvlist_free(fuids[i]);
-		kmem_free(fuids, retidx * sizeof (void *));
-		VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
-		packed = kmem_alloc(nvsize, KM_SLEEP);
-		VERIFY(nvlist_pack(nvp, &packed, &nvsize,
-		    NV_ENCODE_XDR, KM_SLEEP) == 0);
-		nvlist_free(nvp);
-		zfsvfs->z_fuid_size = nvsize;
-		dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
-		    zfsvfs->z_fuid_size, packed, tx);
-		kmem_free(packed, zfsvfs->z_fuid_size);
-		VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
-		    FTAG, &db));
-		dmu_buf_will_dirty(db, tx);
-		*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
-		dmu_buf_rele(db, FTAG);
-
 		rw_exit(&zfsvfs->z_fuid_lock);
 		return (retidx);
+	} else {
+		return (-1);
 	}
 }
 
@@ -337,7 +363,7 @@ retry:
  * Returns a pointer from an avl node of the domain string.
  *
  */
-static char *
+const char *
 zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
 {
 	char *domain;
@@ -346,7 +372,7 @@ zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
 		return (NULL);
 
 	if (!zfsvfs->z_fuid_loaded)
-		zfs_fuid_init(zfsvfs, NULL);
+		zfs_fuid_init(zfsvfs);
 
 	rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
 
@@ -374,7 +400,7 @@ zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
     cred_t *cr, zfs_fuid_type_t type)
 {
 	uint32_t index = FUID_INDEX(fuid);
-	char *domain;
+	const char *domain;
 	uid_t id;
 
 	if (index == 0)
@@ -439,6 +465,7 @@ zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
 	}
 
 	if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
+
 		/*
 		 * Now allocate fuid entry and add it on the end of the list
 		 */
@@ -463,7 +490,7 @@ zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
  */
 uint64_t
 zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
-    dmu_tx_t *tx, cred_t *cr, zfs_fuid_info_t **fuidp)
+    cred_t *cr, zfs_fuid_info_t **fuidp)
 {
 	uint64_t	idx;
 	ksid_t		*ksid;
@@ -490,7 +517,7 @@ zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
 	rid = ksid_getrid(ksid);
 	domain = ksid_getdomain(ksid);
 
-	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
 
 	zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
 
@@ -511,7 +538,7 @@ zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
  */
 uint64_t
 zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
-    zfs_fuid_type_t type, dmu_tx_t *tx, zfs_fuid_info_t **fuidpp)
+    zfs_fuid_type_t type, zfs_fuid_info_t **fuidpp)
 {
 	const char *domain;
 	char *kdomain;
@@ -581,10 +608,11 @@ zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
 		}
 	}
 
-	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, tx);
+	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
 
 	if (!zfsvfs->z_replay)
-		zfs_fuid_node_add(fuidpp, kdomain, rid, idx, id, type);
+		zfs_fuid_node_add(fuidpp, kdomain,
+		    rid, idx, id, type);
 	else if (zfuid != NULL) {
 		list_remove(&fuidp->z_fuids, zfuid);
 		kmem_free(zfuid, sizeof (zfs_fuid_t));
@@ -658,16 +686,15 @@ boolean_t
 zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
 {
 	ksid_t		*ksid = crgetsid(cr, KSID_GROUP);
+	ksidlist_t	*ksidlist = crgetsidlist(cr);
 	uid_t		gid;
 
-	if (ksid) {
+	if (ksid && ksidlist) {
 		int 		i;
 		ksid_t		*ksid_groups;
-		ksidlist_t	*ksidlist = crgetsidlist(cr);
 		uint32_t	idx = FUID_INDEX(id);
 		uint32_t	rid = FUID_RID(id);
 
-		ASSERT(ksidlist);
 		ksid_groups = ksidlist->ksl_sids;
 
 		for (i = 0; i != ksidlist->ksl_nsid; i++) {
@@ -677,7 +704,7 @@ zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
 					return (B_TRUE);
 				}
 			} else {
-				char *domain;
+				const char *domain;
 
 				domain = zfs_fuid_find_by_idx(zfsvfs, idx);
 				ASSERT(domain != NULL);
@@ -700,4 +727,19 @@ zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
 	gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
 	return (groupmember(gid, cr));
 }
+
+void
+zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
+{
+	if (zfsvfs->z_fuid_obj == 0) {
+		dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+		    FUID_SIZE_ESTIMATE(zfsvfs));
+		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
+	} else {
+		dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+		dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+		    FUID_SIZE_ESTIMATE(zfsvfs));
+	}
+}
 #endif

module/zfs/zfs_log.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -467,9 +467,6 @@ zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
  */
 ssize_t zfs_immediate_write_sz = 32768;
 
-#define	ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \
-    sizeof (lr_write_t))
-
 void
 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 	znode_t *zp, offset_t off, ssize_t resid, int ioflag)
@@ -483,29 +480,6 @@ zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 
 	ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */
 
-	/*
-	 * Writes are handled in three different ways:
-	 *
-	 * WR_INDIRECT:
-	 *    In this mode, if we need to commit the write later, then the block
-	 *    is immediately written into the file system (using dmu_sync),
-	 *    and a pointer to the block is put into the log record.
-	 *    When the txg commits the block is linked in.
-	 *    This saves additionally writing the data into the log record.
-	 *    There are a few requirements for this to occur:
-	 *	- write is greater than zfs_immediate_write_sz
-	 *	- not using slogs (as slogs are assumed to always be faster
-	 *	  than writing into the main pool)
-	 *	- the write occupies only one block
-	 * WR_COPIED:
-	 *    If we know we'll immediately be committing the
-	 *    transaction (FSYNC or FDSYNC), the we allocate a larger
-	 *    log record here for the data and copy the data in.
-	 * WR_NEED_COPY:
-	 *    Otherwise we don't allocate a buffer, and *if* we need to
-	 *    flush the write later then a buffer is allocated and
-	 *    we retrieve the data using the dmu.
-	 */
 	slogging = spa_has_slogs(zilog->zl_spa);
 	if (resid > zfs_immediate_write_sz && !slogging && resid <= zp->z_blksz)
 		write_state = WR_INDIRECT;
@@ -535,7 +509,7 @@ zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 		    (write_state == WR_COPIED ? len : 0));
 		lr = (lr_write_t *)&itx->itx_lr;
 		if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
-		    zp->z_id, off, len, lr + 1) != 0) {
+		    zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
 			kmem_free(itx, offsetof(itx_t, itx_lr) +
 			    itx->itx_lr.lrc_reclen);
 			itx = zil_itx_create(txtype, sizeof (*lr));

module/zfs/zfs_vfsops.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -67,6 +67,8 @@ static major_t zfs_major;
 static minor_t zfs_minor;
 static kmutex_t	zfs_dev_mtx;
 
+extern int sys_shutdown;
+
 static int zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr);
 static int zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr);
 static int zfs_mountroot(vfs_t *vfsp, enum whymountroot);
@@ -145,12 +147,24 @@ zfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
 		 * Sync a specific filesystem.
 		 */
 		zfsvfs_t *zfsvfs = vfsp->vfs_data;
+		dsl_pool_t *dp;
 
 		ZFS_ENTER(zfsvfs);
+		dp = dmu_objset_pool(zfsvfs->z_os);
+
+		/*
+		 * If the system is shutting down, then skip any
+		 * filesystems which may exist on a suspended pool.
+		 */
+		if (sys_shutdown && spa_suspended(dp->dp_spa)) {
+			ZFS_EXIT(zfsvfs);
+			return (0);
+		}
+
 		if (zfsvfs->z_log != NULL)
 			zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
 		else
-			txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
+			txg_wait_synced(dp, 0);
 		ZFS_EXIT(zfsvfs);
 	} else {
 		/*
@@ -554,6 +568,393 @@ unregister:
 
 }
 
+static void
+uidacct(objset_t *os, boolean_t isgroup, uint64_t fuid,
+    int64_t delta, dmu_tx_t *tx)
+{
+	uint64_t used = 0;
+	char buf[32];
+	int err;
+	uint64_t obj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
+
+	if (delta == 0)
+		return;
+
+	(void) snprintf(buf, sizeof (buf), "%llx", (longlong_t)fuid);
+	err = zap_lookup(os, obj, buf, 8, 1, &used);
+	ASSERT(err == 0 || err == ENOENT);
+	/* no underflow/overflow */
+	ASSERT(delta > 0 || used >= -delta);
+	ASSERT(delta < 0 || used + delta > used);
+	used += delta;
+	if (used == 0)
+		err = zap_remove(os, obj, buf, tx);
+	else
+		err = zap_update(os, obj, buf, 8, 1, &used, tx);
+	ASSERT(err == 0);
+}
+
+static void
+zfs_space_delta_cb(objset_t *os, dmu_object_type_t bonustype,
+    void *oldbonus, void *newbonus,
+    uint64_t oldused, uint64_t newused, dmu_tx_t *tx)
+{
+	znode_phys_t *oldznp = oldbonus;
+	znode_phys_t *newznp = newbonus;
+
+	if (bonustype != DMU_OT_ZNODE)
+		return;
+
+	/* We charge 512 for the dnode (if it's allocated). */
+	if (oldznp->zp_gen != 0)
+		oldused += DNODE_SIZE;
+	if (newznp->zp_gen != 0)
+		newused += DNODE_SIZE;
+
+	if (oldznp->zp_uid == newznp->zp_uid) {
+		uidacct(os, B_FALSE, oldznp->zp_uid, newused-oldused, tx);
+	} else {
+		uidacct(os, B_FALSE, oldznp->zp_uid, -oldused, tx);
+		uidacct(os, B_FALSE, newznp->zp_uid, newused, tx);
+	}
+
+	if (oldznp->zp_gid == newznp->zp_gid) {
+		uidacct(os, B_TRUE, oldznp->zp_gid, newused-oldused, tx);
+	} else {
+		uidacct(os, B_TRUE, oldznp->zp_gid, -oldused, tx);
+		uidacct(os, B_TRUE, newznp->zp_gid, newused, tx);
+	}
+}
+
+static void
+fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr,
+    char *domainbuf, int buflen, uid_t *ridp)
+{
+	extern uint64_t strtonum(const char *str, char **nptr);
+	uint64_t fuid;
+	const char *domain;
+
+	fuid = strtonum(fuidstr, NULL);
+
+	domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid));
+	if (domain)
+		(void) strlcpy(domainbuf, domain, buflen);
+	else
+		domainbuf[0] = '\0';
+	*ridp = FUID_RID(fuid);
+}
+
+static uint64_t
+zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type)
+{
+	switch (type) {
+	case ZFS_PROP_USERUSED:
+		return (DMU_USERUSED_OBJECT);
+	case ZFS_PROP_GROUPUSED:
+		return (DMU_GROUPUSED_OBJECT);
+	case ZFS_PROP_USERQUOTA:
+		return (zfsvfs->z_userquota_obj);
+	case ZFS_PROP_GROUPQUOTA:
+		return (zfsvfs->z_groupquota_obj);
+	}
+	return (0);
+}
+
+int
+zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    uint64_t *cookiep, void *vbuf, uint64_t *bufsizep)
+{
+	int error;
+	zap_cursor_t zc;
+	zap_attribute_t za;
+	zfs_useracct_t *buf = vbuf;
+	uint64_t obj;
+
+	if (!dmu_objset_userspace_present(zfsvfs->z_os))
+		return (ENOTSUP);
+
+	obj = zfs_userquota_prop_to_obj(zfsvfs, type);
+	if (obj == 0) {
+		*bufsizep = 0;
+		return (0);
+	}
+
+	for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep);
+	    (error = zap_cursor_retrieve(&zc, &za)) == 0;
+	    zap_cursor_advance(&zc)) {
+		if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) >
+		    *bufsizep)
+			break;
+
+		fuidstr_to_sid(zfsvfs, za.za_name,
+		    buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid);
+
+		buf->zu_space = za.za_first_integer;
+		buf++;
+	}
+	if (error == ENOENT)
+		error = 0;
+
+	ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep);
+	*bufsizep = (uintptr_t)buf - (uintptr_t)vbuf;
+	*cookiep = zap_cursor_serialize(&zc);
+	zap_cursor_fini(&zc);
+	return (error);
+}
+
+/*
+ * buf must be big enough (eg, 32 bytes)
+ */
+static int
+id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid,
+    char *buf, boolean_t addok)
+{
+	uint64_t fuid;
+	int domainid = 0;
+
+	if (domain && domain[0]) {
+		domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok);
+		if (domainid == -1)
+			return (ENOENT);
+	}
+	fuid = FUID_ENCODE(domainid, rid);
+	(void) sprintf(buf, "%llx", (longlong_t)fuid);
+	return (0);
+}
+
+int
+zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    const char *domain, uint64_t rid, uint64_t *valp)
+{
+	char buf[32];
+	int err;
+	uint64_t obj;
+
+	*valp = 0;
+
+	if (!dmu_objset_userspace_present(zfsvfs->z_os))
+		return (ENOTSUP);
+
+	obj = zfs_userquota_prop_to_obj(zfsvfs, type);
+	if (obj == 0)
+		return (0);
+
+	err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE);
+	if (err)
+		return (err);
+
+	err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp);
+	if (err == ENOENT)
+		err = 0;
+	return (err);
+}
+
+int
+zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type,
+    const char *domain, uint64_t rid, uint64_t quota)
+{
+	char buf[32];
+	int err;
+	dmu_tx_t *tx;
+	uint64_t *objp;
+	boolean_t fuid_dirtied;
+
+	if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
+		return (EINVAL);
+
+	if (zfsvfs->z_version < ZPL_VERSION_USERSPACE)
+		return (ENOTSUP);
+
+	objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj :
+	    &zfsvfs->z_groupquota_obj;
+
+	err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE);
+	if (err)
+		return (err);
+	fuid_dirtied = zfsvfs->z_fuid_dirty;
+
+	tx = dmu_tx_create(zfsvfs->z_os);
+	dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
+	if (*objp == 0) {
+		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
+		    zfs_userquota_prop_prefixes[type]);
+	}
+	if (fuid_dirtied)
+		zfs_fuid_txhold(zfsvfs, tx);
+	err = dmu_tx_assign(tx, TXG_WAIT);
+	if (err) {
+		dmu_tx_abort(tx);
+		return (err);
+	}
+
+	mutex_enter(&zfsvfs->z_lock);
+	if (*objp == 0) {
+		*objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA,
+		    DMU_OT_NONE, 0, tx);
+		VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
+		    zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
+	}
+	mutex_exit(&zfsvfs->z_lock);
+
+	if (quota == 0) {
+		err = zap_remove(zfsvfs->z_os, *objp, buf, tx);
+		if (err == ENOENT)
+			err = 0;
+	} else {
+		err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, &quota, tx);
+	}
+	ASSERT(err == 0);
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
+	dmu_tx_commit(tx);
+	return (err);
+}
+
+boolean_t
+zfs_usergroup_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid)
+{
+	char buf[32];
+	uint64_t used, quota, usedobj, quotaobj;
+	int err;
+
+	usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
+	quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;
+
+	if (quotaobj == 0 || zfsvfs->z_replay)
+		return (B_FALSE);
+
+	(void) sprintf(buf, "%llx", (longlong_t)fuid);
+	err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, &quota);
+	if (err != 0)
+		return (B_FALSE);
+
+	err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used);
+	if (err != 0)
+		return (B_FALSE);
+	return (used >= quota);
+}
+
+int
+zfsvfs_create(const char *osname, int mode, zfsvfs_t **zvp)
+{
+	objset_t *os;
+	zfsvfs_t *zfsvfs;
+	uint64_t zval;
+	int i, error;
+
+	if (error = dsl_prop_get_integer(osname, "readonly", &zval, NULL))
+		return (error);
+	if (zval)
+		mode |= DS_MODE_READONLY;
+
+	error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &os);
+	if (error == EROFS) {
+		mode |= DS_MODE_READONLY;
+		error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &os);
+	}
+	if (error)
+		return (error);
+
+	/*
+	 * Initialize the zfs-specific filesystem structure.
+	 * Should probably make this a kmem cache, shuffle fields,
+	 * and just bzero up to z_hold_mtx[].
+	 */
+	zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
+	zfsvfs->z_vfs = NULL;
+	zfsvfs->z_parent = zfsvfs;
+	zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
+	zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
+	zfsvfs->z_os = os;
+
+	error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
+	if (error) {
+		goto out;
+	} else if (zfsvfs->z_version > ZPL_VERSION) {
+		(void) printf("Mismatched versions:  File system "
+		    "is version %llu on-disk format, which is "
+		    "incompatible with this software version %lld!",
+		    (u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
+		error = ENOTSUP;
+		goto out;
+	}
+
+	if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
+		goto out;
+	zfsvfs->z_norm = (int)zval;
+
+	if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
+		goto out;
+	zfsvfs->z_utf8 = (zval != 0);
+
+	if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
+		goto out;
+	zfsvfs->z_case = (uint_t)zval;
+
+	/*
+	 * Fold case on file systems that are always or sometimes case
+	 * insensitive.
+	 */
+	if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
+	    zfsvfs->z_case == ZFS_CASE_MIXED)
+		zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
+
+	zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
+
+	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
+	    &zfsvfs->z_root);
+	if (error)
+		goto out;
+	ASSERT(zfsvfs->z_root != 0);
+
+	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
+	    &zfsvfs->z_unlinkedobj);
+	if (error)
+		goto out;
+
+	error = zap_lookup(os, MASTER_NODE_OBJ,
+	    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
+	    8, 1, &zfsvfs->z_userquota_obj);
+	if (error && error != ENOENT)
+		goto out;
+
+	error = zap_lookup(os, MASTER_NODE_OBJ,
+	    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
+	    8, 1, &zfsvfs->z_groupquota_obj);
+	if (error && error != ENOENT)
+		goto out;
+
+	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
+	    &zfsvfs->z_fuid_obj);
+	if (error && error != ENOENT)
+		goto out;
+
+	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
+	    &zfsvfs->z_shares_dir);
+	if (error && error != ENOENT)
+		goto out;
+
+	mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
+	mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL);
+	mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL);
+	list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
+	    offsetof(znode_t, z_link_node));
+	rrw_init(&zfsvfs->z_teardown_lock);
+	rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
+	rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+		mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
+
+	*zvp = zfsvfs;
+	return (0);
+
+out:
+	dmu_objset_close(os);
+	*zvp = NULL;
+	kmem_free(zfsvfs, sizeof (zfsvfs_t));
+	return (error);
+}
+
 static int
 zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 {
@@ -570,6 +971,12 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 	dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
 	mutex_exit(&zfsvfs->z_os->os->os_user_ptr_lock);
 
+	zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
+	if (zil_disable) {
+		zil_destroy(zfsvfs->z_log, 0);
+		zfsvfs->z_log = NULL;
+	}
+
 	/*
 	 * If we are not mounting (ie: online recv), then we don't
 	 * have to worry about replaying the log as we blocked all
@@ -588,11 +995,7 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 		else
 			zfs_unlinked_drain(zfsvfs);
 
-		zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
+		if (zfsvfs->z_log) {
-		if (zil_disable) {
-			zil_destroy(zfsvfs->z_log, 0);
-			zfsvfs->z_log = NULL;
-		} else {
 			/*
 			 * Parse and replay the intent log.
 			 *
@@ -630,49 +1033,63 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
 	return (0);
 }
 
-static void
+void
-zfs_freezfsvfs(zfsvfs_t *zfsvfs)
+zfsvfs_free(zfsvfs_t *zfsvfs)
 {
+	int i;
+	extern krwlock_t zfsvfs_lock; /* in zfs_znode.c */
+
+	/*
+	 * This is a barrier to prevent the filesystem from going away in
+	 * zfs_znode_move() until we can safely ensure that the filesystem is
+	 * not unmounted. We consider the filesystem valid before the barrier
+	 * and invalid after the barrier.
+	 */
+	rw_enter(&zfsvfs_lock, RW_READER);
+	rw_exit(&zfsvfs_lock);
+
+	zfs_fuid_destroy(zfsvfs);
+
 	mutex_destroy(&zfsvfs->z_znodes_lock);
 	mutex_destroy(&zfsvfs->z_online_recv_lock);
+	mutex_destroy(&zfsvfs->z_lock);
 	list_destroy(&zfsvfs->z_all_znodes);
 	rrw_destroy(&zfsvfs->z_teardown_lock);
 	rw_destroy(&zfsvfs->z_teardown_inactive_lock);
 	rw_destroy(&zfsvfs->z_fuid_lock);
+	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+		mutex_destroy(&zfsvfs->z_hold_mtx[i]);
 	kmem_free(zfsvfs, sizeof (zfsvfs_t));
 }
 
+static void
+zfs_set_fuid_feature(zfsvfs_t *zfsvfs)
+{
+	zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
+	if (zfsvfs->z_use_fuids && zfsvfs->z_vfs) {
+		vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR);
+		vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS);
+		vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS);
+		vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE);
+		vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER);
+	}
+}
+
 static int
 zfs_domount(vfs_t *vfsp, char *osname)
 {
 	dev_t mount_dev;
-	uint64_t recordsize, readonly;
+	uint64_t recordsize, fsid_guid;
 	int error = 0;
-	int mode;
 	zfsvfs_t *zfsvfs;
-	znode_t *zp = NULL;
 
 	ASSERT(vfsp);
 	ASSERT(osname);
 
-	/*
+	error = zfsvfs_create(osname, DS_MODE_OWNER, &zfsvfs);
-	 * Initialize the zfs-specific filesystem structure.
+	if (error)
-	 * Should probably make this a kmem cache, shuffle fields,
+		return (error);
-	 * and just bzero up to z_hold_mtx[].
-	 */
-	zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
 	zfsvfs->z_vfs = vfsp;
-	zfsvfs->z_parent = zfsvfs;
-	zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
-	zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
-
-	mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
-	mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL);
-	list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
-	    offsetof(znode_t, z_link_node));
-	rrw_init(&zfsvfs->z_teardown_lock);
-	rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
-	rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
 
 	/* Initialize the generic filesystem structure. */
 	vfsp->vfs_bcount = 0;
@@ -694,39 +1111,24 @@ zfs_domount(vfs_t *vfsp, char *osname)
 	vfsp->vfs_flag |= VFS_NOTRUNC;
 	vfsp->vfs_data = zfsvfs;
 
-	if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
+	/*
-		goto out;
+	 * The fsid is 64 bits, composed of an 8-bit fs type, which
-
+	 * separates our fsid from any other filesystem types, and a
-	mode = DS_MODE_OWNER;
+	 * 56-bit objset unique ID.  The objset unique ID is unique to
-	if (readonly)
+	 * all objsets open on this system, provided by unique_create().
-		mode |= DS_MODE_READONLY;
+	 * The 8-bit fs type must be put in the low bits of fsid[1]
-
+	 * because that's where other Solaris filesystems put it.
-	error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
+	 */
-	if (error == EROFS) {
+	fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os);
-		mode = DS_MODE_OWNER | DS_MODE_READONLY;
+	ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
-		error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
+	vfsp->vfs_fsid.val[0] = fsid_guid;
-		    &zfsvfs->z_os);
+	vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
-	}
+	    zfsfstype & 0xFF;
-
-	if (error)
-		goto out;
-
-	if (error = zfs_init_fs(zfsvfs, &zp))
-		goto out;
-
-	/* The call to zfs_init_fs leaves the vnode held, release it here. */
-	VN_RELE(ZTOV(zp));
 
 	/*
 	 * Set features for file system.
 	 */
-	zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
+	zfs_set_fuid_feature(zfsvfs);
-	if (zfsvfs->z_use_fuids) {
-		vfs_set_feature(vfsp, VFSFT_XVATTR);
-		vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS);
-		vfs_set_feature(vfsp, VFSFT_ACEMASKONACCESS);
-		vfs_set_feature(vfsp, VFSFT_ACLONCREATE);
-	}
 	if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
 		vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
 		vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
@@ -739,13 +1141,16 @@ zfs_domount(vfs_t *vfsp, char *osname)
 	if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
 		uint64_t pval;
 
-		ASSERT(mode & DS_MODE_READONLY);
 		atime_changed_cb(zfsvfs, B_FALSE);
 		readonly_changed_cb(zfsvfs, B_TRUE);
 		if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))
 			goto out;
 		xattr_changed_cb(zfsvfs, pval);
 		zfsvfs->z_issnap = B_TRUE;
+
+		mutex_enter(&zfsvfs->z_os->os->os_user_ptr_lock);
+		dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
+		mutex_exit(&zfsvfs->z_os->os->os_user_ptr_lock);
 	} else {
 		error = zfsvfs_setup(zfsvfs, B_TRUE);
 	}
@@ -754,9 +1159,8 @@ zfs_domount(vfs_t *vfsp, char *osname)
 		zfsctl_create(zfsvfs);
 out:
 	if (error) {
-		if (zfsvfs->z_os)
+		dmu_objset_close(zfsvfs->z_os);
-			dmu_objset_close(zfsvfs->z_os);
+		zfsvfs_free(zfsvfs);
-		zfs_freezfsvfs(zfsvfs);
 	} else {
 		atomic_add_32(&zfs_active_fs_count, 1);
 	}
@@ -1067,6 +1471,13 @@ zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
 
 	error = zfs_domount(vfsp, osname);
 
+	/*
+	 * Add an extra VFS_HOLD on our parent vfs so that it can't
+	 * disappear due to a forced unmount.
+	 */
+	if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap)
+		VFS_HOLD(mvp->v_vfsp);
+
 out:
 	pn_free(&spn);
 	return (error);
@@ -1426,15 +1837,16 @@ zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp)
  * 'z_teardown_inactive_lock' write held.
  */
 int
-zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *mode)
+zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *modep)
 {
 	int error;
 
 	if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0)
 		return (error);
 
-	*mode = zfsvfs->z_os->os_mode;
+	*modep = zfsvfs->z_os->os_mode;
-	dmu_objset_name(zfsvfs->z_os, name);
+	if (name)
+		dmu_objset_name(zfsvfs->z_os, name);
 	dmu_objset_close(zfsvfs->z_os);
 
 	return (0);
@@ -1493,13 +1905,15 @@ static void
 zfs_freevfs(vfs_t *vfsp)
 {
 	zfsvfs_t *zfsvfs = vfsp->vfs_data;
-	int i;
 
-	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+	/*
-		mutex_destroy(&zfsvfs->z_hold_mtx[i]);
+	 * If this is a snapshot, we have an extra VFS_HOLD on our parent
+	 * from zfs_mount().  Release it here.
+	 */
+	if (zfsvfs->z_issnap)
+		VFS_RELE(zfsvfs->z_parent->z_vfs);
 
-	zfs_fuid_destroy(zfsvfs);
+	zfsvfs_free(zfsvfs);
-	zfs_freezfsvfs(zfsvfs);
 
 	atomic_add_32(&zfs_active_fs_count, -1);
 }
@@ -1558,6 +1972,8 @@ zfs_init(void)
 	 * Initialize znode cache, vnode ops, etc...
 	 */
 	zfs_znode_init();
+
+	dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb);
 }
 
 void
@@ -1574,54 +1990,46 @@ zfs_busy(void)
 }
 
 int
-zfs_set_version(const char *name, uint64_t newvers)
+zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers)
 {
 	int error;
-	objset_t *os;
+	objset_t *os = zfsvfs->z_os;
 	dmu_tx_t *tx;
-	uint64_t curvers;
-
-	/*
-	 * XXX for now, require that the filesystem be unmounted.  Would
-	 * be nice to find the zfsvfs_t and just update that if
-	 * possible.
-	 */
 
 	if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
 		return (EINVAL);
 
-	error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_OWNER, &os);
+	if (newvers < zfsvfs->z_version)
-	if (error)
+		return (EINVAL);
-		return (error);
-
-	error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
-	    8, 1, &curvers);
-	if (error)
-		goto out;
-	if (newvers < curvers) {
-		error = EINVAL;
-		goto out;
-	}
 
 	tx = dmu_tx_create(os);
-	dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR);
+	dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
 	error = dmu_tx_assign(tx, TXG_WAIT);
 	if (error) {
 		dmu_tx_abort(tx);
-		goto out;
+		return (error);
+	}
+	error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
+	    8, 1, &newvers, tx);
+
+	if (error) {
+		dmu_tx_commit(tx);
+		return (error);
 	}
-	error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1,
-	    &newvers, tx);
 
 	spa_history_internal_log(LOG_DS_UPGRADE,
 	    dmu_objset_spa(os), tx, CRED(),
-	    "oldver=%llu newver=%llu dataset = %llu", curvers, newvers,
+	    "oldver=%llu newver=%llu dataset = %llu",
-	    dmu_objset_id(os));
+	    zfsvfs->z_version, newvers, dmu_objset_id(os));
+
 	dmu_tx_commit(tx);
 
-out:
+	zfsvfs->z_version = newvers;
-	dmu_objset_close(os);
+
-	return (error);
+	if (zfsvfs->z_version >= ZPL_VERSION_FUID)
+		zfs_set_fuid_feature(zfsvfs);
+
+	return (0);
 }
 
 /*

module/zfs/zfs_vnops.c

@@ -101,6 +101,7 @@
  *	pushing cached pages (which acquires range locks) and syncing out
  *	cached atime changes.  Third, zfs_zinactive() may require a new tx,
  *	which could deadlock the system if you were already holding one.
+ *	If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
  *
  *  (3)	All range locks must be grabbed before calling dmu_tx_assign(),
  *	as they can span dmu_tx_assign() calls.
@@ -363,7 +364,8 @@ update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid)
 			caddr_t va;
 
 			va = zfs_map_page(pp, S_WRITE);
-			(void) dmu_read(os, oid, start+off, nbytes, va+off);
+			(void) dmu_read(os, oid, start+off, nbytes, va+off,
+			    DMU_READ_PREFETCH);
 			zfs_unmap_page(pp, va);
 			page_unlock(pp);
 		}
@@ -567,6 +569,7 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
 	int		max_blksz = zfsvfs->z_max_blksz;
 	uint64_t	pflags;
 	int		error;
+	arc_buf_t	*abuf;
 
 	/*
 	 * Fasttrack empty write
@@ -663,10 +666,46 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
 	 * and allows us to do more fine-grained space accounting.
 	 */
 	while (n > 0) {
+		abuf = NULL;
+		woff = uio->uio_loffset;
+
+again:
+		if (zfs_usergroup_overquota(zfsvfs,
+		    B_FALSE, zp->z_phys->zp_uid) ||
+		    zfs_usergroup_overquota(zfsvfs,
+		    B_TRUE, zp->z_phys->zp_gid)) {
+			if (abuf != NULL)
+				dmu_return_arcbuf(abuf);
+			error = EDQUOT;
+			break;
+		}
+
+		/*
+		 * If dmu_assign_arcbuf() is expected to execute with minimum
+		 * overhead loan an arc buffer and copy user data to it before
+		 * we enter a txg.  This avoids holding a txg forever while we
+		 * pagefault on a hanging NFS server mapping.
+		 */
+		if (abuf == NULL && n >= max_blksz &&
+		    woff >= zp->z_phys->zp_size &&
+		    P2PHASE(woff, max_blksz) == 0 &&
+		    zp->z_blksz == max_blksz) {
+			size_t cbytes;
+
+			abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
+			ASSERT(abuf != NULL);
+			ASSERT(arc_buf_size(abuf) == max_blksz);
+			if (error = uiocopy(abuf->b_data, max_blksz,
+			    UIO_WRITE, uio, &cbytes)) {
+				dmu_return_arcbuf(abuf);
+				break;
+			}
+			ASSERT(cbytes == max_blksz);
+		}
+
 		/*
 		 * Start a transaction.
 		 */
-		woff = uio->uio_loffset;
 		tx = dmu_tx_create(zfsvfs->z_os);
 		dmu_tx_hold_bonus(tx, zp->z_id);
 		dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
@@ -675,9 +714,11 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
 			if (error == ERESTART) {
 				dmu_tx_wait(tx);
 				dmu_tx_abort(tx);
-				continue;
+				goto again;
 			}
 			dmu_tx_abort(tx);
+			if (abuf != NULL)
+				dmu_return_arcbuf(abuf);
 			break;
 		}
 
@@ -706,12 +747,22 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
 		 */
 		nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
 
-		tx_bytes = uio->uio_resid;
+		if (abuf == NULL) {
-		error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio, nbytes, tx);
+			tx_bytes = uio->uio_resid;
-		tx_bytes -= uio->uio_resid;
+			error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
-		if (tx_bytes && vn_has_cached_data(vp))
+			    nbytes, tx);
+			tx_bytes -= uio->uio_resid;
+		} else {
+			tx_bytes = nbytes;
+			ASSERT(tx_bytes == max_blksz);
+			dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
+			ASSERT(tx_bytes <= uio->uio_resid);
+			uioskip(uio, tx_bytes);
+		}
+		if (tx_bytes && vn_has_cached_data(vp)) {
 			update_pages(vp, woff,
 			    tx_bytes, zfsvfs->z_os, zp->z_id);
+		}
 
 		/*
 		 * If we made no progress, we're done.  If we made even
@@ -791,10 +842,15 @@ zfs_get_done(dmu_buf_t *db, void *vzgd)
 	zgd_t *zgd = (zgd_t *)vzgd;
 	rl_t *rl = zgd->zgd_rl;
 	vnode_t *vp = ZTOV(rl->r_zp);
+	objset_t *os = rl->r_zp->z_zfsvfs->z_os;
 
 	dmu_buf_rele(db, vzgd);
 	zfs_range_unlock(rl);
-	VN_RELE(vp);
+	/*
+	 * Release the vnode asynchronously as we currently have the
+	 * txg stopped from syncing.
+	 */
+	VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
 	zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
 	kmem_free(zgd, sizeof (zgd_t));
 }
@@ -824,7 +880,12 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 	if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
 		return (ENOENT);
 	if (zp->z_unlinked) {
-		VN_RELE(ZTOV(zp));
+		/*
+		 * Release the vnode asynchronously as we currently have the
+		 * txg stopped from syncing.
+		 */
+		VN_RELE_ASYNC(ZTOV(zp),
+		    dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
 		return (ENOENT);
 	}
 
@@ -842,7 +903,8 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 			error = ENOENT;
 			goto out;
 		}
-		VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf));
+		VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
+		    DMU_READ_NO_PREFETCH));
 	} else { /* indirect write */
 		uint64_t boff; /* block starting offset */
 
@@ -896,7 +958,11 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 	}
 out:
 	zfs_range_unlock(rl);
-	VN_RELE(ZTOV(zp));
+	/*
+	 * Release the vnode asynchronously as we currently have the
+	 * txg stopped from syncing.
+	 */
+	VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
 	return (error);
 }
 
@@ -1074,11 +1140,11 @@ zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl,
 	zfs_dirlock_t	*dl;
 	dmu_tx_t	*tx;
 	int		error;
-	zfs_acl_t	*aclp = NULL;
-	zfs_fuid_info_t *fuidp = NULL;
 	ksid_t		*ksid;
 	uid_t		uid;
 	gid_t		gid = crgetgid(cr);
+	zfs_acl_ids_t	acl_ids;
+	boolean_t	fuid_dirtied;
 
 	/*
 	 * If we have an ephemeral id, ACL, or XVATTR then
@@ -1141,21 +1207,9 @@ top:
 			if (strcmp(name, "..") == 0)
 				error = EISDIR;
 			ZFS_EXIT(zfsvfs);
-			if (aclp)
-				zfs_acl_free(aclp);
 			return (error);
 		}
 	}
-	if (vsecp && aclp == NULL) {
-		error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
-		if (error) {
-			ZFS_EXIT(zfsvfs);
-			if (dl)
-				zfs_dirent_unlock(dl);
-			return (error);
-		}
-	}
-
 	if (zp == NULL) {
 		uint64_t txtype;
 
@@ -1177,30 +1231,28 @@ top:
 			goto out;
 		}
 
+		if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
+		    &acl_ids)) != 0)
+			goto out;
+		if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
+			error = EDQUOT;
+			goto out;
+		}
+
 		tx = dmu_tx_create(os);
 		dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
-		if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(uid) ||
+		fuid_dirtied = zfsvfs->z_fuid_dirty;
-		    IS_EPHEMERAL(gid)) {
+		if (fuid_dirtied)
-			if (zfsvfs->z_fuid_obj == 0) {
+			zfs_fuid_txhold(zfsvfs, tx);
-				dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
-				dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-				    FUID_SIZE_ESTIMATE(zfsvfs));
-				dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
-				    FALSE, NULL);
-			} else {
-				dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-				dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-				    FUID_SIZE_ESTIMATE(zfsvfs));
-			}
-		}
 		dmu_tx_hold_bonus(tx, dzp->z_id);
 		dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
-		if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp) {
+		if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
 			dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
 			    0, SPA_MAXBLOCKSIZE);
 		}
 		error = dmu_tx_assign(tx, TXG_NOWAIT);
 		if (error) {
+			zfs_acl_ids_free(&acl_ids);
 			zfs_dirent_unlock(dl);
 			if (error == ERESTART) {
 				dmu_tx_wait(tx);
@@ -1209,19 +1261,21 @@ top:
 			}
 			dmu_tx_abort(tx);
 			ZFS_EXIT(zfsvfs);
-			if (aclp)
-				zfs_acl_free(aclp);
 			return (error);
 		}
-		zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
+		zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
+
+		if (fuid_dirtied)
+			zfs_fuid_sync(zfsvfs, tx);
+
 		(void) zfs_link_create(dl, zp, tx, ZNEW);
+
 		txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
 		if (flag & FIGNORECASE)
 			txtype |= TX_CI;
 		zfs_log_create(zilog, tx, txtype, dzp, zp, name,
-		    vsecp, fuidp, vap);
+		    vsecp, acl_ids.z_fuidp, vap);
-		if (fuidp)
+		zfs_acl_ids_free(&acl_ids);
-			zfs_fuid_info_free(fuidp);
 		dmu_tx_commit(tx);
 	} else {
 		int aflags = (flag & FAPPEND) ? V_APPEND : 0;
@@ -1292,8 +1346,6 @@ out:
 			*vpp = svp;
 		}
 	}
-	if (aclp)
-		zfs_acl_free(aclp);
 
 	ZFS_EXIT(zfsvfs);
 	return (error);
@@ -1528,12 +1580,12 @@ zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
 	uint64_t	txtype;
 	dmu_tx_t	*tx;
 	int		error;
-	zfs_acl_t	*aclp = NULL;
-	zfs_fuid_info_t	*fuidp = NULL;
 	int		zf = ZNEW;
 	ksid_t		*ksid;
 	uid_t		uid;
 	gid_t		gid = crgetgid(cr);
+	zfs_acl_ids_t	acl_ids;
+	boolean_t	fuid_dirtied;
 
 	ASSERT(vap->va_type == VDIR);
 
@@ -1594,38 +1646,33 @@ top:
 		return (error);
 	}
 
-	if (vsecp && aclp == NULL) {
+	if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
-		error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
+	    &acl_ids)) != 0) {
-		if (error) {
+		zfs_dirent_unlock(dl);
-			zfs_dirent_unlock(dl);
+		ZFS_EXIT(zfsvfs);
-			ZFS_EXIT(zfsvfs);
+		return (error);
-			return (error);
-		}
 	}
+	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
+		zfs_dirent_unlock(dl);
+		ZFS_EXIT(zfsvfs);
+		return (EDQUOT);
+	}
+
 	/*
 	 * Add a new entry to the directory.
 	 */
 	tx = dmu_tx_create(zfsvfs->z_os);
 	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
 	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
-	if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(uid) ||
+	fuid_dirtied = zfsvfs->z_fuid_dirty;
-	    IS_EPHEMERAL(gid)) {
+	if (fuid_dirtied)
-		if (zfsvfs->z_fuid_obj == 0) {
+		zfs_fuid_txhold(zfsvfs, tx);
-			dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+	if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
-			dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-			dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
-		} else {
-			dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-			dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-		}
-	}
-	if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp)
 		dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
 		    0, SPA_MAXBLOCKSIZE);
 	error = dmu_tx_assign(tx, TXG_NOWAIT);
 	if (error) {
+		zfs_acl_ids_free(&acl_ids);
 		zfs_dirent_unlock(dl);
 		if (error == ERESTART) {
 			dmu_tx_wait(tx);
@@ -1634,19 +1681,16 @@ top:
 		}
 		dmu_tx_abort(tx);
 		ZFS_EXIT(zfsvfs);
-		if (aclp)
-			zfs_acl_free(aclp);
 		return (error);
 	}
 
 	/*
 	 * Create new node.
 	 */
-	zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
+	zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
-
-	if (aclp)
-		zfs_acl_free(aclp);
 
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
 	/*
 	 * Now put new name in parent dir.
 	 */
@@ -1657,10 +1701,10 @@ top:
 	txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
 	if (flags & FIGNORECASE)
 		txtype |= TX_CI;
-	zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, fuidp, vap);
+	zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
+	    acl_ids.z_fuidp, vap);
 
-	if (fuidp)
+	zfs_acl_ids_free(&acl_ids);
-		zfs_fuid_info_free(fuidp);
 	dmu_tx_commit(tx);
 
 	zfs_dirent_unlock(dl);
@@ -1969,6 +2013,21 @@ zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
 			}
 		}
 
+		if (flags & V_RDDIR_ACCFILTER) {
+			/*
+			 * If we have no access at all, don't include
+			 * this entry in the returned information
+			 */
+			znode_t	*ezp;
+			if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
+				goto skip_entry;
+			if (!zfs_has_access(ezp, cr)) {
+				VN_RELE(ZTOV(ezp));
+				goto skip_entry;
+			}
+			VN_RELE(ZTOV(ezp));
+		}
+
 		if (flags & V_RDDIR_ENTFLAGS)
 			reclen = EDIRENT_RECLEN(strlen(zap.za_name));
 		else
@@ -2020,6 +2079,7 @@ zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
 		if (prefetch)
 			dmu_prefetch(os, objnum, 0, 0);
 
+	skip_entry:
 		/*
 		 * Move to the next entry, fill in the previous offset.
 		 */
@@ -2120,8 +2180,6 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 	ZFS_VERIFY_ZP(zp);
 	pzp = zp->z_phys;
 
-	mutex_enter(&zp->z_lock);
-
 	/*
 	 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
 	 * Also, if we are the owner don't bother, since owner should
@@ -2131,7 +2189,6 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 	    (pzp->zp_uid != crgetuid(cr))) {
 		if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
 		    skipaclchk, cr)) {
-			mutex_exit(&zp->z_lock);
 			ZFS_EXIT(zfsvfs);
 			return (error);
 		}
@@ -2142,6 +2199,7 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 	 * than to determine whether we were asked the question.
 	 */
 
+	mutex_enter(&zp->z_lock);
 	vap->va_type = vp->v_type;
 	vap->va_mode = pzp->zp_mode & MODEMASK;
 	zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
@@ -2312,6 +2370,7 @@ zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 	uint_t		saved_mask;
 	int		trim_mask = 0;
 	uint64_t	new_mode;
+	uint64_t	new_uid, new_gid;
 	znode_t		*attrzp;
 	int		need_policy = FALSE;
 	int		err;
@@ -2320,6 +2379,7 @@ zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 	xoptattr_t	*xoap;
 	zfs_acl_t	*aclp = NULL;
 	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
+	boolean_t fuid_dirtied = B_FALSE;
 
 	if (mask == 0)
 		return (0);
@@ -2610,30 +2670,14 @@ top:
 
 	tx = dmu_tx_create(zfsvfs->z_os);
 	dmu_tx_hold_bonus(tx, zp->z_id);
-	if (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
-	    ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid))) {
-		if (zfsvfs->z_fuid_obj == 0) {
-			dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
-			dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-			dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
-		} else {
-			dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-			dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-		}
-	}
 
 	if (mask & AT_MODE) {
 		uint64_t pmode = pzp->zp_mode;
 
 		new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
 
-		if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) {
+		if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
-			dmu_tx_abort(tx);
+			goto out;
-			ZFS_EXIT(zfsvfs);
-			return (err);
-		}
 		if (pzp->zp_acl.z_acl_extern_obj) {
 			/* Are we upgrading ACL from old V0 format to new V1 */
 			if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
@@ -2655,36 +2699,53 @@ top:
 		}
 	}
 
-	if ((mask & (AT_UID | AT_GID)) && pzp->zp_xattr != 0) {
+	if (mask & (AT_UID | AT_GID)) {
-		err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
+		if (pzp->zp_xattr) {
-		if (err) {
+			err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
-			dmu_tx_abort(tx);
+			if (err)
-			ZFS_EXIT(zfsvfs);
+				goto out;
-			if (aclp)
+			dmu_tx_hold_bonus(tx, attrzp->z_id);
-				zfs_acl_free(aclp);
+		}
-			return (err);
+		if (mask & AT_UID) {
+			new_uid = zfs_fuid_create(zfsvfs,
+			    (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
+			if (new_uid != pzp->zp_uid &&
+			    zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
+				err = EDQUOT;
+				goto out;
+			}
+		}
+
+		if (mask & AT_GID) {
+			new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
+			    cr, ZFS_GROUP, &fuidp);
+			if (new_gid != pzp->zp_gid &&
+			    zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
+				err = EDQUOT;
+				goto out;
+			}
+		}
+		fuid_dirtied = zfsvfs->z_fuid_dirty;
+		if (fuid_dirtied) {
+			if (zfsvfs->z_fuid_obj == 0) {
+				dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
+				dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
+				    FUID_SIZE_ESTIMATE(zfsvfs));
+				dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
+				    FALSE, NULL);
+			} else {
+				dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
+				dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
+				    FUID_SIZE_ESTIMATE(zfsvfs));
+			}
 		}
-		dmu_tx_hold_bonus(tx, attrzp->z_id);
 	}
 
 	err = dmu_tx_assign(tx, TXG_NOWAIT);
 	if (err) {
-		if (attrzp)
+		if (err == ERESTART)
-			VN_RELE(ZTOV(attrzp));
-
-		if (aclp) {
-			zfs_acl_free(aclp);
-			aclp = NULL;
-		}
-
-		if (err == ERESTART) {
 			dmu_tx_wait(tx);
-			dmu_tx_abort(tx);
+		goto out;
-			goto top;
-		}
-		dmu_tx_abort(tx);
-		ZFS_EXIT(zfsvfs);
-		return (err);
 	}
 
 	dmu_buf_will_dirty(zp->z_dbuf, tx);
@@ -2702,7 +2763,7 @@ top:
 	if (mask & AT_MODE) {
 		mutex_enter(&zp->z_acl_lock);
 		zp->z_phys->zp_mode = new_mode;
-		err = zfs_aclset_common(zp, aclp, cr, &fuidp, tx);
+		err = zfs_aclset_common(zp, aclp, cr, tx);
 		ASSERT3U(err, ==, 0);
 		mutex_exit(&zp->z_acl_lock);
 	}
@@ -2711,25 +2772,17 @@ top:
 		mutex_enter(&attrzp->z_lock);
 
 	if (mask & AT_UID) {
-		pzp->zp_uid = zfs_fuid_create(zfsvfs,
+		pzp->zp_uid = new_uid;
-		    vap->va_uid, cr, ZFS_OWNER, tx, &fuidp);
+		if (attrzp)
-		if (attrzp) {
+			attrzp->z_phys->zp_uid = new_uid;
-			attrzp->z_phys->zp_uid = zfs_fuid_create(zfsvfs,
-			    vap->va_uid,  cr, ZFS_OWNER, tx, &fuidp);
-		}
 	}
 
 	if (mask & AT_GID) {
-		pzp->zp_gid = zfs_fuid_create(zfsvfs, vap->va_gid,
+		pzp->zp_gid = new_gid;
-		    cr, ZFS_GROUP, tx, &fuidp);
 		if (attrzp)
-			attrzp->z_phys->zp_gid = zfs_fuid_create(zfsvfs,
+			attrzp->z_phys->zp_gid = new_gid;
-			    vap->va_gid, cr, ZFS_GROUP, tx, &fuidp);
 	}
 
-	if (aclp)
-		zfs_acl_free(aclp);
-
 	if (attrzp)
 		mutex_exit(&attrzp->z_lock);
 
@@ -2791,17 +2844,35 @@ top:
 		zfs_xvattr_set(zp, xvap);
 	}
 
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
+
 	if (mask != 0)
 		zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
 
-	if (fuidp)
-		zfs_fuid_info_free(fuidp);
 	mutex_exit(&zp->z_lock);
 
+out:
 	if (attrzp)
 		VN_RELE(ZTOV(attrzp));
 
-	dmu_tx_commit(tx);
+	if (aclp) {
+		zfs_acl_free(aclp);
+		aclp = NULL;
+	}
+
+	if (fuidp) {
+		zfs_fuid_info_free(fuidp);
+		fuidp = NULL;
+	}
+
+	if (err)
+		dmu_tx_abort(tx);
+	else
+		dmu_tx_commit(tx);
+
+	if (err == ERESTART)
+		goto top;
 
 	ZFS_EXIT(zfsvfs);
 	return (err);
@@ -3232,7 +3303,8 @@ zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr,
 	int		len = strlen(link);
 	int		error;
 	int		zflg = ZNEW;
-	zfs_fuid_info_t *fuidp = NULL;
+	zfs_acl_ids_t	acl_ids;
+	boolean_t	fuid_dirtied;
 
 	ASSERT(vap->va_type == VLNK);
 
@@ -3267,26 +3339,25 @@ top:
 		return (error);
 	}
 
+	VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
+	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
+		zfs_acl_ids_free(&acl_ids);
+		zfs_dirent_unlock(dl);
+		ZFS_EXIT(zfsvfs);
+		return (EDQUOT);
+	}
 	tx = dmu_tx_create(zfsvfs->z_os);
+	fuid_dirtied = zfsvfs->z_fuid_dirty;
 	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
 	dmu_tx_hold_bonus(tx, dzp->z_id);
 	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
-	if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE)
+	if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
 		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
-	if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
+	if (fuid_dirtied)
-		if (zfsvfs->z_fuid_obj == 0) {
+		zfs_fuid_txhold(zfsvfs, tx);
-			dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
-			dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-			dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
-		} else {
-			dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
-			dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
-			    FUID_SIZE_ESTIMATE(zfsvfs));
-		}
-	}
 	error = dmu_tx_assign(tx, TXG_NOWAIT);
 	if (error) {
+		zfs_acl_ids_free(&acl_ids);
 		zfs_dirent_unlock(dl);
 		if (error == ERESTART) {
 			dmu_tx_wait(tx);
@@ -3306,13 +3377,16 @@ top:
 	 * otherwise, store it just like any other file data.
 	 */
 	if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
-		zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, NULL, &fuidp);
+		zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
 		if (len != 0)
 			bcopy(link, zp->z_phys + 1, len);
 	} else {
 		dmu_buf_t *dbp;
 
-		zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, NULL, &fuidp);
+		zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
+
+		if (fuid_dirtied)
+			zfs_fuid_sync(zfsvfs, tx);
 		/*
 		 * Nothing can access the znode yet so no locking needed
 		 * for growing the znode's blocksize.
@@ -3333,15 +3407,14 @@ top:
 	 * Insert the new object into the directory.
 	 */
 	(void) zfs_link_create(dl, zp, tx, ZNEW);
-out:
 	if (error == 0) {
 		uint64_t txtype = TX_SYMLINK;
 		if (flags & FIGNORECASE)
 			txtype |= TX_CI;
 		zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
 	}
-	if (fuidp)
+
-		zfs_fuid_info_free(fuidp);
+	zfs_acl_ids_free(&acl_ids);
 
 	dmu_tx_commit(tx);
 
@@ -3618,6 +3691,12 @@ zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp,
 			pvn_write_done(trunc, flags);
 		len = filesz - off;
 	}
+
+	if (zfs_usergroup_overquota(zfsvfs, B_FALSE, zp->z_phys->zp_uid) ||
+	    zfs_usergroup_overquota(zfsvfs, B_TRUE, zp->z_phys->zp_gid)) {
+		err = EDQUOT;
+		goto out;
+	}
 top:
 	tx = dmu_tx_create(zfsvfs->z_os);
 	dmu_tx_hold_write(tx, zp->z_id, off, len);
@@ -3705,7 +3784,7 @@ zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr,
 	else
 		io_off = 0;
 	if (len > 0 && ISP2(blksz))
-		io_len = P2ROUNDUP_TYPED(len + (io_off - off), blksz, size_t);
+		io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t);
 	else
 		io_len = 0;
 
@@ -3869,7 +3948,8 @@ zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset,
  * If we can't find a page in the cache, we will create a new page
  * and fill it with file data.  For efficiency, we may try to fill
  * multiple pages at once (klustering) to fill up the supplied page
- * list.
+ * list.  Note that the pages to be filled are held with an exclusive
+ * lock to prevent access by other threads while they are being filled.
  */
 static int
 zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg,
@@ -3888,7 +3968,8 @@ zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg,
 		 */
 		io_off = off;
 		io_len = PAGESIZE;
-		pp = page_create_va(vp, io_off, io_len, PG_WAIT, seg, addr);
+		pp = page_create_va(vp, io_off, io_len,
+		    PG_EXCL | PG_WAIT, seg, addr);
 	} else {
 		/*
 		 * Try to find enough pages to fill the page list
@@ -3913,7 +3994,8 @@ zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg,
 
 		ASSERT3U(io_off, ==, cur_pp->p_offset);
 		va = zfs_map_page(cur_pp, S_WRITE);
-		err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va);
+		err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va,
+		    DMU_READ_PREFETCH);
 		zfs_unmap_page(cur_pp, va);
 		if (err) {
 			/* On error, toss the entire kluster */
@@ -3991,7 +4073,7 @@ zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp,
 		*protp = PROT_ALL;
 
 	/*
-	 * Loop through the requested range [off, off + len] looking
+	 * Loop through the requested range [off, off + len) looking
 	 * for pages.  If we don't find a page, we will need to create
 	 * a new page and fill it with data from the file.
 	 */
@@ -4337,6 +4419,11 @@ zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 		    (vp->v_type == VREG || vp->v_type == VDIR);
 		return (0);
 
+	case _PC_ACCESS_FILTERING:
+		*valp = vfs_has_feature(vp->v_vfsp, VFSFT_ACCESS_FILTER) &&
+		    vp->v_type == VDIR;
+		return (0);
+
 	case _PC_ACL_ENABLED:
 		*valp = _ACL_ACE_ENABLED;
 		return (0);
@@ -4488,6 +4575,22 @@ const fs_operation_def_t zfs_symvnodeops_template[] = {
 	NULL,			NULL
 };
 
+/*
+ * special share hidden files vnode operations template
+ */
+vnodeops_t *zfs_sharevnodeops;
+const fs_operation_def_t zfs_sharevnodeops_template[] = {
+	VOPNAME_GETATTR,	{ .vop_getattr = zfs_getattr },
+	VOPNAME_ACCESS,		{ .vop_access = zfs_access },
+	VOPNAME_INACTIVE,	{ .vop_inactive = zfs_inactive },
+	VOPNAME_FID,		{ .vop_fid = zfs_fid },
+	VOPNAME_PATHCONF,	{ .vop_pathconf = zfs_pathconf },
+	VOPNAME_GETSECATTR,	{ .vop_getsecattr = zfs_getsecattr },
+	VOPNAME_SETSECATTR,	{ .vop_setsecattr = zfs_setsecattr },
+	VOPNAME_VNEVENT,	{ .vop_vnevent = fs_vnevent_support },
+	NULL,			NULL
+};
+
 /*
  * Extended attribute directory vnode operations template
  *	This template is identical to the directory vnodes

module/zfs/zfs_znode.c

@@ -87,6 +87,12 @@
  * (such as VFS logic) that will not compile easily in userland.
  */
 #ifdef _KERNEL
+/*
+ * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
+ * be freed before it can be safely accessed.
+ */
+krwlock_t zfsvfs_lock;
+
 static kmem_cache_t *znode_cache = NULL;
 
 /*ARGSUSED*/
@@ -154,8 +160,9 @@ zfs_znode_cache_destructor(void *buf, void *arg)
 #ifdef	ZNODE_STATS
 static struct {
 	uint64_t zms_zfsvfs_invalid;
+	uint64_t zms_zfsvfs_recheck1;
 	uint64_t zms_zfsvfs_unmounted;
-	uint64_t zms_zfsvfs_recheck_invalid;
+	uint64_t zms_zfsvfs_recheck2;
 	uint64_t zms_obj_held;
 	uint64_t zms_vnode_locked;
 	uint64_t zms_not_only_dnlc;
@@ -206,17 +213,6 @@ zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
 	POINTER_INVALIDATE(&ozp->z_zfsvfs);
 }
 
-/*
- * Wrapper function for ZFS_ENTER that returns 0 if successful and otherwise
- * returns a non-zero error code.
- */
-static int
-zfs_enter(zfsvfs_t *zfsvfs)
-{
-	ZFS_ENTER(zfsvfs);
-	return (0);
-}
-
 /*ARGSUSED*/
 static kmem_cbrc_t
 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
@@ -240,12 +236,32 @@ zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
 	}
 
 	/*
-	 * Ensure that the filesystem is not unmounted during the move.
+	 * Close a small window in which it's possible that the filesystem could
+	 * be unmounted and freed, and zfsvfs, though valid in the previous
+	 * statement, could point to unrelated memory by the time we try to
+	 * prevent the filesystem from being unmounted.
 	 */
-	if (zfs_enter(zfsvfs) != 0) {		/* ZFS_ENTER */
+	rw_enter(&zfsvfs_lock, RW_WRITER);
+	if (zfsvfs != ozp->z_zfsvfs) {
+		rw_exit(&zfsvfs_lock);
+		ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
+		return (KMEM_CBRC_DONT_KNOW);
+	}
+
+	/*
+	 * If the znode is still valid, then so is the file system. We know that
+	 * no valid file system can be freed while we hold zfsvfs_lock, so we
+	 * can safely ensure that the filesystem is not and will not be
+	 * unmounted. The next statement is equivalent to ZFS_ENTER().
+	 */
+	rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
+	if (zfsvfs->z_unmounted) {
+		ZFS_EXIT(zfsvfs);
+		rw_exit(&zfsvfs_lock);
 		ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
 		return (KMEM_CBRC_DONT_KNOW);
 	}
+	rw_exit(&zfsvfs_lock);
 
 	mutex_enter(&zfsvfs->z_znodes_lock);
 	/*
@@ -255,7 +271,7 @@ zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
 	if (zfsvfs != ozp->z_zfsvfs) {
 		mutex_exit(&zfsvfs->z_znodes_lock);
 		ZFS_EXIT(zfsvfs);
-		ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck_invalid);
+		ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
 		return (KMEM_CBRC_DONT_KNOW);
 	}
 
@@ -311,6 +327,7 @@ zfs_znode_init(void)
 	/*
 	 * Initialize zcache
 	 */
+	rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
 	ASSERT(znode_cache == NULL);
 	znode_cache = kmem_cache_create("zfs_znode_cache",
 	    sizeof (znode_t), 0, zfs_znode_cache_constructor,
@@ -332,6 +349,7 @@ zfs_znode_fini(void)
 	if (znode_cache)
 		kmem_cache_destroy(znode_cache);
 	znode_cache = NULL;
+	rw_destroy(&zfsvfs_lock);
 }
 
 struct vnodeops *zfs_dvnodeops;
@@ -339,6 +357,7 @@ struct vnodeops *zfs_fvnodeops;
 struct vnodeops *zfs_symvnodeops;
 struct vnodeops *zfs_xdvnodeops;
 struct vnodeops *zfs_evnodeops;
+struct vnodeops *zfs_sharevnodeops;
 
 void
 zfs_remove_op_tables()
@@ -363,12 +382,15 @@ zfs_remove_op_tables()
 		vn_freevnodeops(zfs_xdvnodeops);
 	if (zfs_evnodeops)
 		vn_freevnodeops(zfs_evnodeops);
+	if (zfs_sharevnodeops)
+		vn_freevnodeops(zfs_sharevnodeops);
 
 	zfs_dvnodeops = NULL;
 	zfs_fvnodeops = NULL;
 	zfs_symvnodeops = NULL;
 	zfs_xdvnodeops = NULL;
 	zfs_evnodeops = NULL;
+	zfs_sharevnodeops = NULL;
 }
 
 extern const fs_operation_def_t zfs_dvnodeops_template[];
@@ -376,6 +398,7 @@ extern const fs_operation_def_t zfs_fvnodeops_template[];
 extern const fs_operation_def_t zfs_xdvnodeops_template[];
 extern const fs_operation_def_t zfs_symvnodeops_template[];
 extern const fs_operation_def_t zfs_evnodeops_template[];
+extern const fs_operation_def_t zfs_sharevnodeops_template[];
 
 int
 zfs_create_op_tables()
@@ -412,103 +435,58 @@ zfs_create_op_tables()
 
 	error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
 	    &zfs_evnodeops);
+	if (error)
+		return (error);
+
+	error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
+	    &zfs_sharevnodeops);
 
 	return (error);
 }
 
-/*
- * zfs_init_fs - Initialize the zfsvfs struct and the file system
- *	incore "master" object.  Verify version compatibility.
- */
 int
-zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp)
+zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
 {
-	extern int zfsfstype;
+	zfs_acl_ids_t acl_ids;
+	vattr_t vattr;
+	znode_t *sharezp;
+	vnode_t *vp;
+	znode_t *zp;
+	int error;
 
-	objset_t	*os = zfsvfs->z_os;
+	vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
-	int		i, error;
+	vattr.va_type = VDIR;
-	uint64_t fsid_guid;
+	vattr.va_mode = S_IFDIR|0555;
-	uint64_t zval;
+	vattr.va_uid = crgetuid(kcred);
+	vattr.va_gid = crgetgid(kcred);
 
-	*zpp = NULL;
+	sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
+	sharezp->z_unlinked = 0;
+	sharezp->z_atime_dirty = 0;
+	sharezp->z_zfsvfs = zfsvfs;
 
-	error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
+	vp = ZTOV(sharezp);
-	if (error) {
+	vn_reinit(vp);
-		return (error);
+	vp->v_type = VDIR;
-	} else if (zfsvfs->z_version > ZPL_VERSION) {
-		(void) printf("Mismatched versions:  File system "
-		    "is version %llu on-disk format, which is "
-		    "incompatible with this software version %lld!",
-		    (u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
-		return (ENOTSUP);
-	}
 
-	if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
+	VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
-		return (error);
+	    kcred, NULL, &acl_ids));
-	zfsvfs->z_norm = (int)zval;
+	zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE,
-	if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
+	    &zp, 0, &acl_ids);
-		return (error);
+	ASSERT3P(zp, ==, sharezp);
-	zfsvfs->z_utf8 = (zval != 0);
+	ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
-	if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
+	POINTER_INVALIDATE(&sharezp->z_zfsvfs);
-		return (error);
+	error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
-	zfsvfs->z_case = (uint_t)zval;
+	    ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
-	/*
+	zfsvfs->z_shares_dir = sharezp->z_id;
-	 * Fold case on file systems that are always or sometimes case
-	 * insensitive.
-	 */
-	if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
-	    zfsvfs->z_case == ZFS_CASE_MIXED)
-		zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
 
-	/*
+	zfs_acl_ids_free(&acl_ids);
-	 * The fsid is 64 bits, composed of an 8-bit fs type, which
+	ZTOV(sharezp)->v_count = 0;
-	 * separates our fsid from any other filesystem types, and a
+	dmu_buf_rele(sharezp->z_dbuf, NULL);
-	 * 56-bit objset unique ID.  The objset unique ID is unique to
+	sharezp->z_dbuf = NULL;
-	 * all objsets open on this system, provided by unique_create().
+	kmem_cache_free(znode_cache, sharezp);
-	 * The 8-bit fs type must be put in the low bits of fsid[1]
-	 * because that's where other Solaris filesystems put it.
-	 */
-	fsid_guid = dmu_objset_fsid_guid(os);
-	ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
-	zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid;
-	zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
-	    zfsfstype & 0xFF;
 
-	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
+	return (error);
-	    &zfsvfs->z_root);
-	if (error)
-		return (error);
-	ASSERT(zfsvfs->z_root != 0);
-
-	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
-	    &zfsvfs->z_unlinkedobj);
-	if (error)
-		return (error);
-
-	/*
-	 * Initialize zget mutex's
-	 */
-	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
-		mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
-
-	error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp);
-	if (error) {
-		/*
-		 * On error, we destroy the mutexes here since it's not
-		 * possible for the caller to determine if the mutexes were
-		 * initialized properly.
-		 */
-		for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
-			mutex_destroy(&zfsvfs->z_hold_mtx[i]);
-		return (error);
-	}
-	ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root);
-	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
-	    &zfsvfs->z_fuid_obj);
-	if (error == ENOENT)
-		error = 0;
-
-	return (0);
 }
 
 /*
@@ -676,7 +654,10 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz)
 		break;
 	case VREG:
 		vp->v_flag |= VMODSORT;
-		vn_setops(vp, zfs_fvnodeops);
+		if (zp->z_phys->zp_parent == zfsvfs->z_shares_dir)
+			vn_setops(vp, zfs_sharevnodeops);
+		else
+			vn_setops(vp, zfs_fvnodeops);
 		break;
 	case VLNK:
 		vn_setops(vp, zfs_symvnodeops);
@@ -720,8 +701,7 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz)
  */
 void
 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
-    uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_t *setaclp,
+    uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_ids_t *acl_ids)
-    zfs_fuid_info_t **fuidp)
 {
 	dmu_buf_t	*db;
 	znode_phys_t	*pzp;
@@ -846,7 +826,12 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
 		 */
 		*zpp = dzp;
 	}
-	zfs_perm_init(*zpp, dzp, flag, vap, tx, cr, setaclp, fuidp);
+	pzp->zp_uid = acl_ids->z_fuid;
+	pzp->zp_gid = acl_ids->z_fgid;
+	pzp->zp_mode = acl_ids->z_mode;
+	VERIFY(0 == zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
+	if (vap->va_mask & AT_XVATTR)
+		zfs_xvattr_set(*zpp, (xvattr_t *)vap);
 }
 
 void
@@ -1474,7 +1459,7 @@ void
 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 {
 	zfsvfs_t	zfsvfs;
-	uint64_t	moid, doid, version;
+	uint64_t	moid, obj, version;
 	uint64_t	sense = ZFS_CASE_SENSITIVE;
 	uint64_t	norm = 0;
 	nvpair_t	*elem;
@@ -1483,6 +1468,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 	vnode_t		*vp;
 	vattr_t		vattr;
 	znode_t		*zp;
+	zfs_acl_ids_t	acl_ids;
 
 	/*
 	 * First attempt to create master node.
@@ -1499,12 +1485,12 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 	/*
 	 * Set starting attributes.
 	 */
-	if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
+	if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_USERSPACE)
 		version = ZPL_VERSION;
+	else if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
+		version = ZPL_VERSION_USERSPACE - 1;
 	else
 		version = ZPL_VERSION_FUID - 1;
-	error = zap_update(os, moid, ZPL_VERSION_STR,
-	    8, 1, &version, tx);
 	elem = NULL;
 	while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
 		/* For the moment we expect all zpl props to be uint64_ts */
@@ -1515,9 +1501,8 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 		VERIFY(nvpair_value_uint64(elem, &val) == 0);
 		name = nvpair_name(elem);
 		if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
-			version = val;
+			if (val < version)
-			error = zap_update(os, moid, ZPL_VERSION_STR,
+				version = val;
-			    8, 1, &version, tx);
 		} else {
 			error = zap_update(os, moid, name, 8, 1, &val, tx);
 		}
@@ -1528,13 +1513,14 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 			sense = val;
 	}
 	ASSERT(version != 0);
+	error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
 
 	/*
 	 * Create a delete queue.
 	 */
-	doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
+	obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
 
-	error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx);
+	error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
 	ASSERT(error == 0);
 
 	/*
@@ -1575,17 +1561,28 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
 
 	ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
 	rootzp->z_zfsvfs = &zfsvfs;
-	zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, NULL, NULL);
+	VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
+	    cr, NULL, &acl_ids));
+	zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, &acl_ids);
 	ASSERT3P(zp, ==, rootzp);
 	ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
 	error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
 	ASSERT(error == 0);
+	zfs_acl_ids_free(&acl_ids);
 	POINTER_INVALIDATE(&rootzp->z_zfsvfs);
 
 	ZTOV(rootzp)->v_count = 0;
 	dmu_buf_rele(rootzp->z_dbuf, NULL);
 	rootzp->z_dbuf = NULL;
 	kmem_cache_free(znode_cache, rootzp);
+
+	/*
+	 * Create shares directory
+	 */
+
+	error = zfs_create_share_dir(&zfsvfs, tx);
+
+	ASSERT(error == 0);
 }
 
 #endif /* _KERNEL */

module/zfs/zil.c

@@ -19,12 +19,13 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #include <sys/zfs_context.h>
 #include <sys/spa.h>
+#include <sys/spa_impl.h>
 #include <sys/dmu.h>
 #include <sys/zap.h>
 #include <sys/arc.h>
@@ -471,34 +472,22 @@ zil_destroy(zilog_t *zilog, boolean_t keep_first)
 }
 
 /*
- * zil_rollback_destroy() is only called by the rollback code.
+ * return true if the initial log block is not valid
- * We already have a syncing tx. Rollback has exclusive access to the
- * dataset, so we don't have to worry about concurrent zil access.
- * The actual freeing of any log blocks occurs in zil_sync() later in
- * this txg syncing phase.
  */
-void
+static boolean_t
-zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx)
+zil_empty(zilog_t *zilog)
 {
 	const zil_header_t *zh = zilog->zl_header;
-	uint64_t txg;
+	arc_buf_t *abuf = NULL;
 
 	if (BP_IS_HOLE(&zh->zh_log))
-		return;
+		return (B_TRUE);
 
-	txg = dmu_tx_get_txg(tx);
+	if (zil_read_log_block(zilog, &zh->zh_log, &abuf) != 0)
-	ASSERT3U(zilog->zl_destroy_txg, <, txg);
+		return (B_TRUE);
-	zilog->zl_destroy_txg = txg;
-	zilog->zl_keep_first = B_FALSE;
 
-	/*
+	VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
-	 * Ensure there's no outstanding ZIL IO.  No lwbs or just the
+	return (B_FALSE);
-	 * unused one that allocated in advance is ok.
-	 */
-	ASSERT(zilog->zl_lwb_list.list_head.list_next ==
-	    zilog->zl_lwb_list.list_head.list_prev);
-	(void) zil_parse(zilog, zil_free_log_block, zil_free_log_record,
-	    tx, zh->zh_claim_txg);
 }
 
 int
@@ -520,6 +509,30 @@ zil_claim(char *osname, void *txarg)
 	zilog = dmu_objset_zil(os);
 	zh = zil_header_in_syncing_context(zilog);
 
+	if (zilog->zl_spa->spa_log_state == SPA_LOG_CLEAR) {
+		if (!BP_IS_HOLE(&zh->zh_log))
+			zio_free_blk(zilog->zl_spa, &zh->zh_log, first_txg);
+		BP_ZERO(&zh->zh_log);
+		dsl_dataset_dirty(dmu_objset_ds(os), tx);
+	}
+
+	/*
+	 * Record here whether the zil has any records to replay.
+	 * If the header block pointer is null or the block points
+	 * to the stubby then we know there are no valid log records.
+	 * We use the header to store this state as the the zilog gets
+	 * freed later in dmu_objset_close().
+	 * The flags (and the rest of the header fields) are cleared in
+	 * zil_sync() as a result of a zil_destroy(), after replaying the log.
+	 *
+	 * Note, the intent log can be empty but still need the
+	 * stubby to be claimed.
+	 */
+	if (!zil_empty(zilog)) {
+		zh->zh_flags |= ZIL_REPLAY_NEEDED;
+		dsl_dataset_dirty(dmu_objset_ds(os), tx);
+	}
+
 	/*
 	 * Claim all log blocks if we haven't already done so, and remember
 	 * the highest claimed sequence number.  This ensures that if we can
@@ -587,36 +600,6 @@ zil_check_log_chain(char *osname, void *txarg)
 	return (error);
 }
 
-/*
- * Clear a log chain
- */
-/* ARGSUSED */
-int
-zil_clear_log_chain(char *osname, void *txarg)
-{
-	zilog_t *zilog;
-	zil_header_t *zh;
-	objset_t *os;
-	dmu_tx_t *tx;
-	int error;
-
-	error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
-	if (error) {
-		cmn_err(CE_WARN, "can't open objset for %s", osname);
-		return (0);
-	}
-
-	zilog = dmu_objset_zil(os);
-	tx = dmu_tx_create(zilog->zl_os);
-	(void) dmu_tx_assign(tx, TXG_WAIT);
-	zh = zil_header_in_syncing_context(zilog);
-	BP_ZERO(&zh->zh_log);
-	dsl_dataset_dirty(dmu_objset_ds(os), tx);
-	dmu_tx_commit(tx);
-	dmu_objset_close(os);
-	return (0);
-}
-
 static int
 zil_vdev_compare(const void *x1, const void *x2)
 {
@@ -719,18 +702,26 @@ zil_lwb_write_done(zio_t *zio)
 	ASSERT(zio->io_bp->blk_fill == 0);
 
 	/*
-	 * Now that we've written this log block, we have a stable pointer
+	 * Ensure the lwb buffer pointer is cleared before releasing
-	 * to the next block in the chain, so it's OK to let the txg in
+	 * the txg. If we have had an allocation failure and
-	 * which we allocated the next block sync.
+	 * the txg is waiting to sync then we want want zil_sync()
+	 * to remove the lwb so that it's not picked up as the next new
+	 * one in zil_commit_writer(). zil_sync() will only remove
+	 * the lwb if lwb_buf is null.
 	 */
-	txg_rele_to_sync(&lwb->lwb_txgh);
-
 	zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
 	mutex_enter(&zilog->zl_lock);
 	lwb->lwb_buf = NULL;
 	if (zio->io_error)
 		zilog->zl_log_error = B_TRUE;
 	mutex_exit(&zilog->zl_lock);
+
+	/*
+	 * Now that we've written this log block, we have a stable pointer
+	 * to the next block in the chain, so it's OK to let the txg in
+	 * which we allocated the next block sync.
+	 */
+	txg_rele_to_sync(&lwb->lwb_txgh);
 }
 
 /*
@@ -752,9 +743,9 @@ zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb)
 	}
 	if (lwb->lwb_zio == NULL) {
 		lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa,
-		    0, &lwb->lwb_blk, lwb->lwb_buf,
+		    0, &lwb->lwb_blk, lwb->lwb_buf, lwb->lwb_sz,
-		    lwb->lwb_sz, zil_lwb_write_done, lwb,
+		    zil_lwb_write_done, lwb, ZIO_PRIORITY_LOG_WRITE,
-		    ZIO_PRIORITY_LOG_WRITE, ZIO_FLAG_CANFAIL, &zb);
+		    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &zb);
 	}
 }
 
@@ -1040,7 +1031,7 @@ zil_clean(zilog_t *zilog)
 	if ((itx != NULL) &&
 	    (itx->itx_lr.lrc_txg <= spa_last_synced_txg(zilog->zl_spa))) {
 		(void) taskq_dispatch(zilog->zl_clean_taskq,
-		    (void (*)(void *))zil_itx_clean, zilog, TQ_NOSLEEP);
+		    (task_func_t *)zil_itx_clean, zilog, TQ_SLEEP);
 	}
 	mutex_exit(&zilog->zl_lock);
 }
@@ -1216,6 +1207,13 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
 	spa_t *spa = zilog->zl_spa;
 	lwb_t *lwb;
 
+	/*
+	 * We don't zero out zl_destroy_txg, so make sure we don't try
+	 * to destroy it twice.
+	 */
+	if (spa_sync_pass(spa) != 1)
+		return;
+
 	mutex_enter(&zilog->zl_lock);
 
 	ASSERT(zilog->zl_stop_sync == 0);
@@ -1226,7 +1224,6 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
 		blkptr_t blk = zh->zh_log;
 
 		ASSERT(list_head(&zilog->zl_lwb_list) == NULL);
-		ASSERT(spa_sync_pass(spa) == 1);
 
 		bzero(zh, sizeof (zil_header_t));
 		bzero(zilog->zl_replayed_seq, sizeof (zilog->zl_replayed_seq));
@@ -1245,12 +1242,7 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
 		}
 	}
 
-	for (;;) {
+	while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
-		lwb = list_head(&zilog->zl_lwb_list);
-		if (lwb == NULL) {
-			mutex_exit(&zilog->zl_lock);
-			return;
-		}
 		zh->zh_log = lwb->lwb_blk;
 		if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg)
 			break;
@@ -1343,25 +1335,6 @@ zil_free(zilog_t *zilog)
 	kmem_free(zilog, sizeof (zilog_t));
 }
 
-/*
- * return true if the initial log block is not valid
- */
-static boolean_t
-zil_empty(zilog_t *zilog)
-{
-	const zil_header_t *zh = zilog->zl_header;
-	arc_buf_t *abuf = NULL;
-
-	if (BP_IS_HOLE(&zh->zh_log))
-		return (B_TRUE);
-
-	if (zil_read_log_block(zilog, &zh->zh_log, &abuf) != 0)
-		return (B_TRUE);
-
-	VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
-	return (B_FALSE);
-}
-
 /*
  * Open an intent log.
  */
@@ -1417,7 +1390,7 @@ zil_suspend(zilog_t *zilog)
 	const zil_header_t *zh = zilog->zl_header;
 
 	mutex_enter(&zilog->zl_lock);
-	if (zh->zh_claim_txg != 0) {		/* unplayed log */
+	if (zh->zh_flags & ZIL_REPLAY_NEEDED) {		/* unplayed log */
 		mutex_exit(&zilog->zl_lock);
 		return (EBUSY);
 	}
@@ -1601,7 +1574,7 @@ zil_replay(objset_t *os, void *arg, zil_replay_func_t *replay_func[TX_MAX_TYPE])
 	const zil_header_t *zh = zilog->zl_header;
 	zil_replay_arg_t zr;
 
-	if (zil_empty(zilog)) {
+	if ((zh->zh_flags & ZIL_REPLAY_NEEDED) == 0) {
 		zil_destroy(zilog, B_TRUE);
 		return;
 	}
@@ -1671,3 +1644,24 @@ out:
 	mutex_exit(&zilog->zl_lock);
 	return (ret);
 }
+
+/* ARGSUSED */
+int
+zil_vdev_offline(char *osname, void *arg)
+{
+	objset_t *os;
+	zilog_t *zilog;
+	int error;
+
+	error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
+	if (error)
+		return (error);
+
+	zilog = dmu_objset_zil(os);
+	if (zil_suspend(zilog) != 0)
+		error = EEXIST;
+	else
+		zil_resume(zilog);
+	dmu_objset_close(os);
+	return (error);
+}