MFV r258371,r258372: 4101 metaslab_debug should allow for fine-grained control

4101 metaslab_debug should allow for fine-grained control 4102 space_maps should store more information about themselves 4103 space map object blocksize should be increased 4104 ::spa_space no longer works 4105 removing a mirrored log device results in a leaked object 4106 asynchronously load metaslab illumos/illumos-gate@0713e232b7 Note that some tunables have been removed and some new tunables have been added. Of particular note, FreeBSD-only knob vfs.zfs.space_map_last_hope is removed as it was a nop for some time now (after one of the previous merges from upstream). MFC after: 11 days Sponsored by: HybridCluster [merge]
2013-11-28 19:37:22 +00:00 · 2013-11-28 19:37:22 +00:00 · 9932b97e88
commit 9932b97e88
parent 78de33b79a
24 changed files with 2508 additions and 1447 deletions
--- a/cddl/contrib/opensolaris/cmd/zdb/zdb.c
+++ b/cddl/contrib/opensolaris/cmd/zdb/zdb.c
@ -242,7 +242,7 @@ const char histo_stars[] = "****************************************";
 const int histo_width = sizeof (histo_stars) - 1;

 static void
-dump_histogram(const uint64_t *histo, int size)
+dump_histogram(const uint64_t *histo, int size, int offset)
 {
 	int i;
 	int minidx = size - 1;
@ -263,7 +263,7 @@ dump_histogram(const uint64_t *histo, int size)

 	for (i = minidx; i <= maxidx; i++) {
 		(void) printf("\t\t\t%3u: %6llu %s\n",
-		    i, (u_longlong_t)histo[i],
+		    i + offset, (u_longlong_t)histo[i],
 		    &histo_stars[(max - histo[i]) * histo_width / max]);
 	}
 }
@ -316,19 +316,19 @@ dump_zap_stats(objset_t *os, uint64_t object)
 	    (u_longlong_t)zs.zs_salt);

 	(void) printf("\t\tLeafs with 2^n pointers:\n");
-	dump_histogram(zs.zs_leafs_with_2n_pointers, ZAP_HISTOGRAM_SIZE);
+	dump_histogram(zs.zs_leafs_with_2n_pointers, ZAP_HISTOGRAM_SIZE, 0);

 	(void) printf("\t\tBlocks with n*5 entries:\n");
-	dump_histogram(zs.zs_blocks_with_n5_entries, ZAP_HISTOGRAM_SIZE);
+	dump_histogram(zs.zs_blocks_with_n5_entries, ZAP_HISTOGRAM_SIZE, 0);

 	(void) printf("\t\tBlocks n/10 full:\n");
-	dump_histogram(zs.zs_blocks_n_tenths_full, ZAP_HISTOGRAM_SIZE);
+	dump_histogram(zs.zs_blocks_n_tenths_full, ZAP_HISTOGRAM_SIZE, 0);

 	(void) printf("\t\tEntries with n chunks:\n");
-	dump_histogram(zs.zs_entries_using_n_chunks, ZAP_HISTOGRAM_SIZE);
+	dump_histogram(zs.zs_entries_using_n_chunks, ZAP_HISTOGRAM_SIZE, 0);

 	(void) printf("\t\tBuckets with n entries:\n");
-	dump_histogram(zs.zs_buckets_with_n_entries, ZAP_HISTOGRAM_SIZE);
+	dump_histogram(zs.zs_buckets_with_n_entries, ZAP_HISTOGRAM_SIZE, 0);
 }

 /*ARGSUSED*/
@ -517,26 +517,85 @@ dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size)
 	zap_cursor_fini(&zc);
 }

+int
+get_dtl_refcount(vdev_t *vd)
+{
+	int refcount = 0;
+
+	if (vd->vdev_ops->vdev_op_leaf) {
+		space_map_t *sm = vd->vdev_dtl_sm;
+
+		if (sm != NULL &&
+		    sm->sm_dbuf->db_size == sizeof (space_map_phys_t))
+			return (1);
+		return (0);
+	}
+
+	for (int c = 0; c < vd->vdev_children; c++)
+		refcount += get_dtl_refcount(vd->vdev_child[c]);
+	return (refcount);
+}
+
+int
+get_metaslab_refcount(vdev_t *vd)
+{
+	int refcount = 0;
+
+	if (vd->vdev_top == vd) {
+		for (int m = 0; m < vd->vdev_ms_count; m++) {
+			space_map_t *sm = vd->vdev_ms[m]->ms_sm;
+
+			if (sm != NULL &&
+			    sm->sm_dbuf->db_size == sizeof (space_map_phys_t))
+				refcount++;
+		}
+	}
+	for (int c = 0; c < vd->vdev_children; c++)
+		refcount += get_metaslab_refcount(vd->vdev_child[c]);
+
+	return (refcount);
+}
+
+static int
+verify_spacemap_refcounts(spa_t *spa)
+{
+	int expected_refcount, actual_refcount;
+
+	expected_refcount = spa_feature_get_refcount(spa,
+	    &spa_feature_table[SPA_FEATURE_SPACEMAP_HISTOGRAM]);
+	actual_refcount = get_dtl_refcount(spa->spa_root_vdev);
+	actual_refcount += get_metaslab_refcount(spa->spa_root_vdev);
+
+	if (expected_refcount != actual_refcount) {
+		(void) printf("space map refcount mismatch: expected %d != "
+		    "actual %d\n", expected_refcount, actual_refcount);
+		return (2);
+	}
+	return (0);
+}
+
 static void
-dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
+dump_spacemap(objset_t *os, space_map_t *sm)
 {
 	uint64_t alloc, offset, entry;
-	uint8_t mapshift = sm->sm_shift;
-	uint64_t mapstart = sm->sm_start;
 	char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
 			    "INVALID", "INVALID", "INVALID", "INVALID" };

-	if (smo->smo_object == 0)
+	if (sm == NULL)
 		return;

 	/*
 	 * Print out the freelist entries in both encoded and decoded form.
 	 */
 	alloc = 0;
-	for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) {
-		VERIFY3U(0, ==, dmu_read(os, smo->smo_object, offset,
+	for (offset = 0; offset < space_map_length(sm);
+	    offset += sizeof (entry)) {
+		uint8_t mapshift = sm->sm_shift;
+
+		VERIFY0(dmu_read(os, space_map_object(sm), offset,
 		    sizeof (entry), &entry, DMU_READ_PREFETCH));
 		if (SM_DEBUG_DECODE(entry)) {
+
 			(void) printf("\t    [%6llu] %s: txg %llu, pass %llu\n",
 			    (u_longlong_t)(offset / sizeof (entry)),
 			    ddata[SM_DEBUG_ACTION_DECODE(entry)],
@ -548,10 +607,10 @@ dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
 			    (u_longlong_t)(offset / sizeof (entry)),
 			    SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
 			    (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
-			    mapshift) + mapstart),
+			    mapshift) + sm->sm_start),
 			    (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
-			    mapshift) + mapstart + (SM_RUN_DECODE(entry) <<
-			    mapshift)),
+			    mapshift) + sm->sm_start +
+			    (SM_RUN_DECODE(entry) << mapshift)),
 			    (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift));
 			if (SM_TYPE_DECODE(entry) == SM_ALLOC)
 				alloc += SM_RUN_DECODE(entry) << mapshift;
@ -559,10 +618,10 @@ dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
 				alloc -= SM_RUN_DECODE(entry) << mapshift;
 		}
 	}
-	if (alloc != smo->smo_alloc) {
+	if (alloc != space_map_allocated(sm)) {
 		(void) printf("space_map_object alloc (%llu) INCONSISTENT "
 		    "with space map summary (%llu)\n",
-		    (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc);
+		    (u_longlong_t)space_map_allocated(sm), (u_longlong_t)alloc);
 	}
 }

@ -570,15 +629,17 @@ static void
 dump_metaslab_stats(metaslab_t *msp)
 {
 	char maxbuf[32];
-	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;
+	range_tree_t *rt = msp->ms_tree;
+	avl_tree_t *t = &msp->ms_size_tree;
+	int free_pct = range_tree_space(rt) * 100 / msp->ms_size;

-	zdb_nicenum(space_map_maxsize(sm), maxbuf);
+	zdb_nicenum(metaslab_block_maxsize(msp), maxbuf);

 	(void) printf("\t %25s %10lu   %7s  %6s   %4s %4d%%\n",
 	    "segments", avl_numnodes(t), "maxsize", maxbuf,
 	    "freepct", free_pct);
+	(void) printf("\tIn-memory histogram:\n");
+	dump_histogram(rt->rt_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
 }

 static void
@ -586,33 +647,45 @@ dump_metaslab(metaslab_t *msp)
 {
 	vdev_t *vd = msp->ms_group->mg_vd;
 	spa_t *spa = vd->vdev_spa;
-	space_map_t *sm = msp->ms_map;
-	space_map_obj_t *smo = &msp->ms_smo;
+	space_map_t *sm = msp->ms_sm;
 	char freebuf[32];

-	zdb_nicenum(sm->sm_size - smo->smo_alloc, freebuf);
+	zdb_nicenum(msp->ms_size - space_map_allocated(sm), freebuf);

 	(void) printf(
 	    "\tmetaslab %6llu   offset %12llx   spacemap %6llu   free    %5s\n",
-	    (u_longlong_t)(sm->sm_start / sm->sm_size),
-	    (u_longlong_t)sm->sm_start, (u_longlong_t)smo->smo_object, freebuf);
+	    (u_longlong_t)msp->ms_id, (u_longlong_t)msp->ms_start,
+	    (u_longlong_t)space_map_object(sm), freebuf);

-	if (dump_opt['m'] > 1 && !dump_opt['L']) {
+	if (dump_opt['m'] > 2 && !dump_opt['L']) {
 		mutex_enter(&msp->ms_lock);
-		space_map_load_wait(sm);
-		if (!sm->sm_loaded)
-			VERIFY(space_map_load(sm, zfs_metaslab_ops,
-			    SM_FREE, smo, spa->spa_meta_objset) == 0);
+		metaslab_load_wait(msp);
+		if (!msp->ms_loaded) {
+			VERIFY0(metaslab_load(msp));
+			range_tree_stat_verify(msp->ms_tree);
+		}
 		dump_metaslab_stats(msp);
-		space_map_unload(sm);
+		metaslab_unload(msp);
 		mutex_exit(&msp->ms_lock);
 	}

-	if (dump_opt['d'] > 5 || dump_opt['m'] > 2) {
-		ASSERT(sm->sm_size == (1ULL << vd->vdev_ms_shift));
+	if (dump_opt['m'] > 1 && sm != NULL &&
+	    spa_feature_is_active(spa,
+	    &spa_feature_table[SPA_FEATURE_SPACEMAP_HISTOGRAM])) {
+		/*
+		 * The space map histogram represents free space in chunks
+		 * of sm_shift (i.e. bucket 0 refers to 2^sm_shift).
+		 */
+		(void) printf("\tOn-disk histogram:\n");
+		dump_histogram(sm->sm_phys->smp_histogram,
+		    SPACE_MAP_HISTOGRAM_SIZE(sm), sm->sm_shift);
+	}
+
+	if (dump_opt['d'] > 5 || dump_opt['m'] > 3) {
+		ASSERT(msp->ms_size == (1ULL << vd->vdev_ms_shift));

 		mutex_enter(&msp->ms_lock);
-		dump_spacemap(spa->spa_meta_objset, smo, sm);
+		dump_spacemap(spa->spa_meta_objset, msp->ms_sm);
 		mutex_exit(&msp->ms_lock);
 	}
 }
@ -801,9 +874,9 @@ dump_all_ddts(spa_t *spa)
 }

 static void
-dump_dtl_seg(space_map_t *sm, uint64_t start, uint64_t size)
+dump_dtl_seg(void *arg, uint64_t start, uint64_t size)
 {
-	char *prefix = (void *)sm;
+	char *prefix = arg;

 	(void) printf("%s [%llu,%llu) length %llu\n",
 	    prefix,
@ -833,17 +906,16 @@ dump_dtl(vdev_t *vd, int indent)
 	    required ? "DTL-required" : "DTL-expendable");

 	for (int t = 0; t < DTL_TYPES; t++) {
-		space_map_t *sm = &vd->vdev_dtl[t];
-		if (sm->sm_space == 0)
+		range_tree_t *rt = vd->vdev_dtl[t];
+		if (range_tree_space(rt) == 0)
 			continue;
 		(void) snprintf(prefix, sizeof (prefix), "\t%*s%s",
 		    indent + 2, "", name[t]);
-		mutex_enter(sm->sm_lock);
-		space_map_walk(sm, dump_dtl_seg, (void *)prefix);
-		mutex_exit(sm->sm_lock);
+		mutex_enter(rt->rt_lock);
+		range_tree_walk(rt, dump_dtl_seg, prefix);
+		mutex_exit(rt->rt_lock);
 		if (dump_opt['d'] > 5 && vd->vdev_children == 0)
-			dump_spacemap(spa->spa_meta_objset,
-			    &vd->vdev_dtl_smo, sm);
+			dump_spacemap(spa->spa_meta_objset, vd->vdev_dtl_sm);
 	}

 	for (int c = 0; c < vd->vdev_children; c++)
@ -2172,39 +2244,17 @@ zdb_blkptr_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 }

 static void
-zdb_leak(space_map_t *sm, uint64_t start, uint64_t size)
+zdb_leak(void *arg, uint64_t start, uint64_t size)
 {
-	vdev_t *vd = sm->sm_ppd;
+	vdev_t *vd = arg;

 	(void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
 	    (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size);
 }

-/* ARGSUSED */
-static void
-zdb_space_map_load(space_map_t *sm)
-{
-}
-
-static void
-zdb_space_map_unload(space_map_t *sm)
-{
-	space_map_vacate(sm, zdb_leak, sm);
-}
-
-/* ARGSUSED */
-static void
-zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size)
-{
-}
-
-static space_map_ops_t zdb_space_map_ops = {
-	zdb_space_map_load,
-	zdb_space_map_unload,
+static metaslab_ops_t zdb_metaslab_ops = {
 	NULL,	/* alloc */
-	zdb_space_map_claim,
-	NULL,	/* free */
-	NULL	/* maxsize */
+	NULL	/* fragmented */
 };

 static void
@ -2259,11 +2309,21 @@ zdb_leak_init(spa_t *spa, zdb_cb_t *zcb)
 			for (int m = 0; m < vd->vdev_ms_count; m++) {
 				metaslab_t *msp = vd->vdev_ms[m];
 				mutex_enter(&msp->ms_lock);
-				space_map_unload(msp->ms_map);
-				VERIFY(space_map_load(msp->ms_map,
-				    &zdb_space_map_ops, SM_ALLOC, &msp->ms_smo,
-				    spa->spa_meta_objset) == 0);
-				msp->ms_map->sm_ppd = vd;
+				metaslab_unload(msp);
+
+				/*
+				 * For leak detection, we overload the metaslab
+				 * ms_tree to contain allocated segments
+				 * instead of free segments. As a result,
+				 * we can't use the normal metaslab_load/unload
+				 * interfaces.
+				 */
+				if (msp->ms_sm != NULL) {
+					msp->ms_ops = &zdb_metaslab_ops;
+					VERIFY0(space_map_load(msp->ms_sm,
+					    msp->ms_tree, SM_ALLOC));
+					msp->ms_loaded = B_TRUE;
+				}
 				mutex_exit(&msp->ms_lock);
 			}
 		}
@ -2286,7 +2346,20 @@ zdb_leak_fini(spa_t *spa)
 			for (int m = 0; m < vd->vdev_ms_count; m++) {
 				metaslab_t *msp = vd->vdev_ms[m];
 				mutex_enter(&msp->ms_lock);
-				space_map_unload(msp->ms_map);
+
+				/*
+				 * The ms_tree has been overloaded to
+				 * contain allocated segments. Now that we
+				 * finished traversing all blocks, any
+				 * block that remains in the ms_tree
+				 * represents an allocated block that we
+				 * did not claim during the traversal.
+				 * Claimed blocks would have been removed
+				 * from the ms_tree.
+				 */
+				range_tree_vacate(msp->ms_tree, zdb_leak, vd);
+				msp->ms_loaded = B_FALSE;
+
 				mutex_exit(&msp->ms_lock);
 			}
 		}
@ -2489,7 +2562,7 @@ dump_block_stats(spa_t *spa)
 					    "(in 512-byte sectors): "
 					    "number of blocks\n");
 					dump_histogram(zb->zb_psize_histogram,
-					    PSIZE_HISTO_SIZE);
+					    PSIZE_HISTO_SIZE, 0);
 				}
 			}
 		}
@ -2659,6 +2732,9 @@ dump_zpool(spa_t *spa)
 	if (dump_opt['b'] || dump_opt['c'])
 		rc = dump_block_stats(spa);

+	if (rc == 0)
+		rc = verify_spacemap_refcounts(spa);
+
 	if (dump_opt['s'])
 		show_pool_stats(spa);

--- a/cddl/contrib/opensolaris/cmd/zpool/zpool-features.7
+++ b/cddl/contrib/opensolaris/cmd/zpool/zpool-features.7
@ -23,7 +23,7 @@
 .\"
 .\" $FreeBSD$
 .\"
-.Dd September 20, 2013
+.Dd October 08, 2013
 .Dt ZPOOL-FEATURES 7
 .Os
 .Sh NAME
@ -251,6 +251,24 @@ configuration.
 .\" .Xr dumpon 8
 .\" command to configure a
 .\" dump device on a pool comprised of multiple vdevs.
+.It Sy spacemap_histogram
+.Bl -column "READ\-ONLY COMPATIBLE" "com.delphix:spacemap_histogram"
+.It GUID Ta com.delphix:spacemap_histogram
+.It READ\-ONLY COMPATIBLE Ta yes
+.It DEPENDENCIES Ta none
+.El
+.Pp
+This features allows ZFS to maintain more information about how free space
+is organized within the pool. If this feature is 
+.Sy enabled ,
+ZFS will
+set this feature to
+.Sy active
+when a new space map object is created or
+an existing space map is upgraded to the new format.
+Once the feature is
+.Sy active ,
+it will remain in that state until the pool is destroyed.
 .El
 .Sh SEE ALSO
 .Xr zpool 8
--- a/cddl/contrib/opensolaris/cmd/ztest/ztest.c
+++ b/cddl/contrib/opensolaris/cmd/ztest/ztest.c
@ -5339,7 +5339,7 @@ ztest_deadman_thread(void *arg)
 		 * vdev_deadman() discovers that there hasn't been any recent
 		 * I/Os then it will end up aborting the tests.
 		 */
-		if (spa_suspended(spa)) {
+		if (spa_suspended(spa) || spa->spa_root_vdev == NULL) {
 			fatal(0, "aborting test after %llu seconds because "
 			    "pool has transitioned to a suspended state.",
 			    zfs_deadman_synctime_ms / 1000);
--- a/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.c
+++ b/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.c
@ -20,7 +20,7 @@
 */

 /*
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
 */
@ -163,4 +163,7 @@ zpool_feature_init(void)
 	zfeature_register(SPA_FEATURE_MULTI_VDEV_CRASH_DUMP,
 	    "com.joyent:multi_vdev_crash_dump", "multi_vdev_crash_dump",
 	    "Crash dumps to multiple vdev pools.", B_FALSE, B_FALSE, NULL);
+	zfeature_register(SPA_FEATURE_SPACEMAP_HISTOGRAM,
+	    "com.delphix:spacemap_histogram", "spacemap_histogram",
+	    "Spacemaps maintain space histograms.", B_TRUE, B_FALSE, NULL);
 }
--- a/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.h
+++ b/sys/cddl/contrib/opensolaris/common/zfs/zfeature_common.h
@ -20,7 +20,7 @@
 */

 /*
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
 */
@ -55,6 +55,7 @@ static enum spa_feature {
 	SPA_FEATURE_EMPTY_BPOBJ,
 	SPA_FEATURE_LZ4_COMPRESS,
 	SPA_FEATURE_MULTI_VDEV_CRASH_DUMP,
+	SPA_FEATURE_SPACEMAP_HISTOGRAM,
 	SPA_FEATURES
 } spa_feature_t;

--- a/sys/cddl/contrib/opensolaris/uts/common/Makefile.files
+++ b/sys/cddl/contrib/opensolaris/uts/common/Makefile.files
@ -21,7 +21,8 @@

 #
 # Copyright (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
-# Copyright (c) 2012 by Delphix. All rights reserved.
+# Copyright (c) 2012 Nexenta Systems, Inc. All rights reserved.
+# Copyright (c) 2013 by Delphix. All rights reserved.
 # Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
 #
 #
@ -62,6 +63,7 @@ ZFS_COMMON_OBJS +=		\
 	lz4.o			\
 	lzjb.o			\
 	metaslab.o		\
+	range_tree.o		\
 	refcount.o		\
 	rrwlock.o		\
 	sa.o			\
@ -72,6 +74,7 @@ ZFS_COMMON_OBJS +=		\
 	spa_history.o		\
 	spa_misc.o		\
 	space_map.o		\
+	space_reftree.o		\
 	txg.o			\
 	uberblock.o		\
 	unique.o		\
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dnode.c
@ -1339,7 +1339,7 @@ dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx)
 	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);

 	/* Check for any allocated blocks beyond the first */
-	if (dn->dn_phys->dn_maxblkid != 0)
+	if (dn->dn_maxblkid != 0)
 		goto fail;

 	mutex_enter(&dn->dn_dbufs_mtx);
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/metaslab.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/metaslab.c
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/range_tree.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/range_tree.c
@ -0,0 +1,391 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+/*
+ * Copyright (c) 2013 by Delphix. All rights reserved.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/spa.h>
+#include <sys/dmu.h>
+#include <sys/dnode.h>
+#include <sys/zio.h>
+#include <sys/range_tree.h>
+
+static kmem_cache_t *range_seg_cache;
+
+void
+range_tree_init(void)
+{
+	ASSERT(range_seg_cache == NULL);
+	range_seg_cache = kmem_cache_create("range_seg_cache",
+	    sizeof (range_seg_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
+}
+
+void
+range_tree_fini(void)
+{
+	kmem_cache_destroy(range_seg_cache);
+	range_seg_cache = NULL;
+}
+
+void
+range_tree_stat_verify(range_tree_t *rt)
+{
+	range_seg_t *rs;
+	uint64_t hist[RANGE_TREE_HISTOGRAM_SIZE] = { 0 };
+	int i;
+
+	for (rs = avl_first(&rt->rt_root); rs != NULL;
+	    rs = AVL_NEXT(&rt->rt_root, rs)) {
+		uint64_t size = rs->rs_end - rs->rs_start;
+		int idx	= highbit(size) - 1;
+
+		hist[idx]++;
+		ASSERT3U(hist[idx], !=, 0);
+	}
+
+	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
+		if (hist[i] != rt->rt_histogram[i]) {
+			zfs_dbgmsg("i=%d, hist=%p, hist=%llu, rt_hist=%llu",
+			    i, hist, hist[i], rt->rt_histogram[i]);
+		}
+		VERIFY3U(hist[i], ==, rt->rt_histogram[i]);
+	}
+}
+
+static void
+range_tree_stat_incr(range_tree_t *rt, range_seg_t *rs)
+{
+	uint64_t size = rs->rs_end - rs->rs_start;
+	int idx = highbit(size) - 1;
+
+	ASSERT3U(idx, <,
+	    sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+	rt->rt_histogram[idx]++;
+	ASSERT3U(rt->rt_histogram[idx], !=, 0);
+}
+
+static void
+range_tree_stat_decr(range_tree_t *rt, range_seg_t *rs)
+{
+	uint64_t size = rs->rs_end - rs->rs_start;
+	int idx = highbit(size) - 1;
+
+	ASSERT3U(idx, <,
+	    sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+	ASSERT3U(rt->rt_histogram[idx], !=, 0);
+	rt->rt_histogram[idx]--;
+}
+
+/*
+ * NOTE: caller is responsible for all locking.
+ */
+static int
+range_tree_seg_compare(const void *x1, const void *x2)
+{
+	const range_seg_t *r1 = x1;
+	const range_seg_t *r2 = x2;
+
+	if (r1->rs_start < r2->rs_start) {
+		if (r1->rs_end > r2->rs_start)
+			return (0);
+		return (-1);
+	}
+	if (r1->rs_start > r2->rs_start) {
+		if (r1->rs_start < r2->rs_end)
+			return (0);
+		return (1);
+	}
+	return (0);
+}
+
+range_tree_t *
+range_tree_create(range_tree_ops_t *ops, void *arg, kmutex_t *lp)
+{
+	range_tree_t *rt;
+
+	rt = kmem_zalloc(sizeof (range_tree_t), KM_SLEEP);
+
+	avl_create(&rt->rt_root, range_tree_seg_compare,
+	    sizeof (range_seg_t), offsetof(range_seg_t, rs_node));
+
+	rt->rt_lock = lp;
+	rt->rt_ops = ops;
+	rt->rt_arg = arg;
+
+	if (rt->rt_ops != NULL)
+		rt->rt_ops->rtop_create(rt, rt->rt_arg);
+
+	return (rt);
+}
+
+void
+range_tree_destroy(range_tree_t *rt)
+{
+	VERIFY0(rt->rt_space);
+
+	if (rt->rt_ops != NULL)
+		rt->rt_ops->rtop_destroy(rt, rt->rt_arg);
+
+	avl_destroy(&rt->rt_root);
+	kmem_free(rt, sizeof (*rt));
+}
+
+void
+range_tree_add(void *arg, uint64_t start, uint64_t size)
+{
+	range_tree_t *rt = arg;
+	avl_index_t where;
+	range_seg_t rsearch, *rs_before, *rs_after, *rs;
+	uint64_t end = start + size;
+	boolean_t merge_before, merge_after;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+	VERIFY(size != 0);
+
+	rsearch.rs_start = start;
+	rsearch.rs_end = end;
+	rs = avl_find(&rt->rt_root, &rsearch, &where);
+
+	if (rs != NULL && rs->rs_start <= start && rs->rs_end >= end) {
+		zfs_panic_recover("zfs: allocating allocated segment"
+		    "(offset=%llu size=%llu)\n",
+		    (longlong_t)start, (longlong_t)size);
+		return;
+	}
+
+	/* Make sure we don't overlap with either of our neighbors */
+	VERIFY(rs == NULL);
+
+	rs_before = avl_nearest(&rt->rt_root, where, AVL_BEFORE);
+	rs_after = avl_nearest(&rt->rt_root, where, AVL_AFTER);
+
+	merge_before = (rs_before != NULL && rs_before->rs_end == start);
+	merge_after = (rs_after != NULL && rs_after->rs_start == end);
+
+	if (merge_before && merge_after) {
+		avl_remove(&rt->rt_root, rs_before);
+		if (rt->rt_ops != NULL) {
+			rt->rt_ops->rtop_remove(rt, rs_before, rt->rt_arg);
+			rt->rt_ops->rtop_remove(rt, rs_after, rt->rt_arg);
+		}
+
+		range_tree_stat_decr(rt, rs_before);
+		range_tree_stat_decr(rt, rs_after);
+
+		rs_after->rs_start = rs_before->rs_start;
+		kmem_cache_free(range_seg_cache, rs_before);
+		rs = rs_after;
+	} else if (merge_before) {
+		if (rt->rt_ops != NULL)
+			rt->rt_ops->rtop_remove(rt, rs_before, rt->rt_arg);
+
+		range_tree_stat_decr(rt, rs_before);
+
+		rs_before->rs_end = end;
+		rs = rs_before;
+	} else if (merge_after) {
+		if (rt->rt_ops != NULL)
+			rt->rt_ops->rtop_remove(rt, rs_after, rt->rt_arg);
+
+		range_tree_stat_decr(rt, rs_after);
+
+		rs_after->rs_start = start;
+		rs = rs_after;
+	} else {
+		rs = kmem_cache_alloc(range_seg_cache, KM_SLEEP);
+		rs->rs_start = start;
+		rs->rs_end = end;
+		avl_insert(&rt->rt_root, rs, where);
+	}
+
+	if (rt->rt_ops != NULL)
+		rt->rt_ops->rtop_add(rt, rs, rt->rt_arg);
+
+	range_tree_stat_incr(rt, rs);
+	rt->rt_space += size;
+}
+
+void
+range_tree_remove(void *arg, uint64_t start, uint64_t size)
+{
+	range_tree_t *rt = arg;
+	avl_index_t where;
+	range_seg_t rsearch, *rs, *newseg;
+	uint64_t end = start + size;
+	boolean_t left_over, right_over;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+	VERIFY3U(size, !=, 0);
+	VERIFY3U(size, <=, rt->rt_space);
+
+	rsearch.rs_start = start;
+	rsearch.rs_end = end;
+	rs = avl_find(&rt->rt_root, &rsearch, &where);
+
+	/* Make sure we completely overlap with someone */
+	if (rs == NULL) {
+		zfs_panic_recover("zfs: freeing free segment "
+		    "(offset=%llu size=%llu)",
+		    (longlong_t)start, (longlong_t)size);
+		return;
+	}
+	VERIFY3U(rs->rs_start, <=, start);
+	VERIFY3U(rs->rs_end, >=, end);
+
+	left_over = (rs->rs_start != start);
+	right_over = (rs->rs_end != end);
+
+	range_tree_stat_decr(rt, rs);
+
+	if (rt->rt_ops != NULL)
+		rt->rt_ops->rtop_remove(rt, rs, rt->rt_arg);
+
+	if (left_over && right_over) {
+		newseg = kmem_cache_alloc(range_seg_cache, KM_SLEEP);
+		newseg->rs_start = end;
+		newseg->rs_end = rs->rs_end;
+		range_tree_stat_incr(rt, newseg);
+
+		rs->rs_end = start;
+
+		avl_insert_here(&rt->rt_root, newseg, rs, AVL_AFTER);
+		if (rt->rt_ops != NULL)
+			rt->rt_ops->rtop_add(rt, newseg, rt->rt_arg);
+	} else if (left_over) {
+		rs->rs_end = start;
+	} else if (right_over) {
+		rs->rs_start = end;
+	} else {
+		avl_remove(&rt->rt_root, rs);
+		kmem_cache_free(range_seg_cache, rs);
+		rs = NULL;
+	}
+
+	if (rs != NULL) {
+		range_tree_stat_incr(rt, rs);
+
+		if (rt->rt_ops != NULL)
+			rt->rt_ops->rtop_add(rt, rs, rt->rt_arg);
+	}
+
+	rt->rt_space -= size;
+}
+
+static range_seg_t *
+range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size,
+    avl_index_t *wherep)
+{
+	range_seg_t rsearch, *rs;
+	uint64_t end = start + size;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+	VERIFY(size != 0);
+
+	rsearch.rs_start = start;
+	rsearch.rs_end = end;
+	rs = avl_find(&rt->rt_root, &rsearch, wherep);
+
+	if (rs != NULL && rs->rs_start <= start && rs->rs_end >= end)
+		return (rs);
+	return (NULL);
+}
+
+void
+range_tree_verify(range_tree_t *rt, uint64_t off, uint64_t size)
+{
+	range_seg_t *rs;
+	avl_index_t where;
+
+	mutex_enter(rt->rt_lock);
+	rs = range_tree_find(rt, off, size, &where);
+	if (rs != NULL)
+		panic("freeing free block; rs=%p", (void *)rs);
+	mutex_exit(rt->rt_lock);
+}
+
+boolean_t
+range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size)
+{
+	avl_index_t where;
+
+	return (range_tree_find(rt, start, size, &where) != NULL);
+}
+
+void
+range_tree_swap(range_tree_t **rtsrc, range_tree_t **rtdst)
+{
+	range_tree_t *rt;
+
+	ASSERT(MUTEX_HELD((*rtsrc)->rt_lock));
+	ASSERT0(range_tree_space(*rtdst));
+	ASSERT0(avl_numnodes(&(*rtdst)->rt_root));
+
+	rt = *rtsrc;
+	*rtsrc = *rtdst;
+	*rtdst = rt;
+}
+
+void
+range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg)
+{
+	range_seg_t *rs;
+	void *cookie = NULL;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+
+	if (rt->rt_ops != NULL)
+		rt->rt_ops->rtop_vacate(rt, rt->rt_arg);
+
+	while ((rs = avl_destroy_nodes(&rt->rt_root, &cookie)) != NULL) {
+		if (func != NULL)
+			func(arg, rs->rs_start, rs->rs_end - rs->rs_start);
+		kmem_cache_free(range_seg_cache, rs);
+	}
+
+	bzero(rt->rt_histogram, sizeof (rt->rt_histogram));
+	rt->rt_space = 0;
+}
+
+void
+range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg)
+{
+	range_seg_t *rs;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+
+	for (rs = avl_first(&rt->rt_root); rs; rs = AVL_NEXT(&rt->rt_root, rs))
+		func(arg, rs->rs_start, rs->rs_end - rs->rs_start);
+}
+
+uint64_t
+range_tree_space(range_tree_t *rt)
+{
+	return (rt->rt_space);
+}
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
@ -1275,6 +1275,15 @@ spa_unload(spa_t *spa)

 	bpobj_close(&spa->spa_deferred_bpobj);

+	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+	/*
+	 * Close all vdevs.
+	 */
+	if (spa->spa_root_vdev)
+		vdev_free(spa->spa_root_vdev);
+	ASSERT(spa->spa_root_vdev == NULL);
+
 	/*
 	 * Close the dsl pool.
 	 */
@ -1286,20 +1295,12 @@ spa_unload(spa_t *spa)

 	ddt_unload(spa);

-	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

 	/*
 	 * Drop and purge level 2 cache
 	 */
 	spa_l2cache_drop(spa);

-	/*
-	 * Close all vdevs.
-	 */
-	if (spa->spa_root_vdev)
-		vdev_free(spa->spa_root_vdev);
-	ASSERT(spa->spa_root_vdev == NULL);
-
 	for (i = 0; i < spa->spa_spares.sav_count; i++)
 		vdev_free(spa->spa_spares.sav_vdevs[i]);
 	if (spa->spa_spares.sav_vdevs) {
@ -4767,7 +4768,9 @@ spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing)
 	vdev_dirty(tvd, VDD_DTL, newvd, txg);

 	/*
-	 * Restart the resilver
+	 * Schedule the resilver to restart in the future. We do this to
+	 * ensure that dmu_sync-ed blocks have been stitched into the
+	 * respective datasets.
 	 */
 	dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg);

@ -5396,7 +5399,7 @@ spa_vdev_remove_evacuate(spa_t *spa, vdev_t *vd)
 	ASSERT0(vd->vdev_stat.vs_alloc);
 	txg = spa_vdev_config_enter(spa);
 	vd->vdev_removing = B_TRUE;
-	vdev_dirty(vd, 0, NULL, txg);
+	vdev_dirty_leaves(vd, VDD_DTL, txg);
 	vdev_config_dirty(vd);
 	spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);

@ -6240,7 +6243,7 @@ spa_sync_props(void *arg, dmu_tx_t *tx)
 			ASSERT(zpool_prop_feature(nvpair_name(elem)));

 			fname = strchr(nvpair_name(elem), '@') + 1;
-			VERIFY3U(0, ==, zfeature_lookup_name(fname, &feature));
+			VERIFY0(zfeature_lookup_name(fname, &feature));

 			spa_feature_enable(spa, feature, tx);
 			spa_history_log_internal(spa, "set", tx,
@ -6248,7 +6251,7 @@ spa_sync_props(void *arg, dmu_tx_t *tx)
 			break;

 		case ZPOOL_PROP_VERSION:
-			VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+			intval = fnvpair_value_uint64(elem);
 			/*
 			 * The version is synced seperatly before other
 			 * properties and should be correct by now.
@ -6272,7 +6275,7 @@ spa_sync_props(void *arg, dmu_tx_t *tx)
 			 */
 			break;
 		case ZPOOL_PROP_COMMENT:
-			VERIFY(nvpair_value_string(elem, &strval) == 0);
+			strval = fnvpair_value_string(elem);
 			if (spa->spa_comment != NULL)
 				spa_strfree(spa->spa_comment);
 			spa->spa_comment = spa_strdup(strval);
@ -6304,23 +6307,23 @@ spa_sync_props(void *arg, dmu_tx_t *tx)

 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
 				ASSERT(proptype == PROP_TYPE_STRING);
-				VERIFY(nvpair_value_string(elem, &strval) == 0);
-				VERIFY(zap_update(mos,
+				strval = fnvpair_value_string(elem);
+				VERIFY0(zap_update(mos,
 				    spa->spa_pool_props_object, propname,
-				    1, strlen(strval) + 1, strval, tx) == 0);
+				    1, strlen(strval) + 1, strval, tx));
 				spa_history_log_internal(spa, "set", tx,
 				    "%s=%s", nvpair_name(elem), strval);
 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
-				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+				intval = fnvpair_value_uint64(elem);

 				if (proptype == PROP_TYPE_INDEX) {
 					const char *unused;
-					VERIFY(zpool_prop_index_to_string(
-					    prop, intval, &unused) == 0);
+					VERIFY0(zpool_prop_index_to_string(
+					    prop, intval, &unused));
 				}
-				VERIFY(zap_update(mos,
+				VERIFY0(zap_update(mos,
 				    spa->spa_pool_props_object, propname,
-				    8, 1, &intval, tx) == 0);
+				    8, 1, &intval, tx));
 				spa_history_log_internal(spa, "set", tx,
 				    "%s=%lld", nvpair_name(elem), intval);
 			} else {
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c
@ -497,6 +497,18 @@ spa_deadman(void *arg)
 {
 	spa_t *spa = arg;

+	/*
+	 * Disable the deadman timer if the pool is suspended.
+	 */
+	if (spa_suspended(spa)) {
+#ifdef illumos
+		VERIFY(cyclic_reprogram(spa->spa_deadman_cycid, CY_INFINITY));
+#else
+		/* Nothing.  just don't schedule any future callouts. */
+#endif
+		return;
+	}
+
 	zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
 	    (gethrtime() - spa->spa_sync_starttime) / NANOSEC,
 	    ++spa->spa_deadman_calls);
@ -1068,7 +1080,7 @@ spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag)
 		txg_wait_synced(spa->spa_dsl_pool, txg);

 	if (vd != NULL) {
-		ASSERT(!vd->vdev_detached || vd->vdev_dtl_smo.smo_object == 0);
+		ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL);
 		spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
 		vdev_free(vd);
 		spa_config_exit(spa, SCL_ALL, spa);
@ -1750,7 +1762,7 @@ spa_init(int mode)
 #endif /* illumos */
 	refcount_sysinit();
 	unique_init();
-	space_map_init();
+	range_tree_init();
 	zio_init();
 	lz4_init();
 	dmu_init();
@ -1780,7 +1792,7 @@ spa_fini(void)
 	dmu_fini();
 	lz4_fini();
 	zio_fini();
-	space_map_fini();
+	range_tree_fini();
 	unique_fini();
 	refcount_fini();

--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/space_map.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/space_map.c
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/space_reftree.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/space_reftree.c
@ -0,0 +1,159 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+/*
+ * Copyright (c) 2013 by Delphix. All rights reserved.
+ */
+
+#include <sys/zfs_context.h>
+#include <sys/range_tree.h>
+#include <sys/space_reftree.h>
+
+/*
+ * Space reference trees.
+ *
+ * A range tree is a collection of integers.  Every integer is either
+ * in the tree, or it's not.  A space reference tree generalizes
+ * the idea: it allows its members to have arbitrary reference counts,
+ * as opposed to the implicit reference count of 0 or 1 in a range tree.
+ * This representation comes in handy when computing the union or
+ * intersection of multiple space maps.  For example, the union of
+ * N range trees is the subset of the reference tree with refcnt >= 1.
+ * The intersection of N range trees is the subset with refcnt >= N.
+ *
+ * [It's very much like a Fourier transform.  Unions and intersections
+ * are hard to perform in the 'range tree domain', so we convert the trees
+ * into the 'reference count domain', where it's trivial, then invert.]
+ *
+ * vdev_dtl_reassess() uses computations of this form to determine
+ * DTL_MISSING and DTL_OUTAGE for interior vdevs -- e.g. a RAID-Z vdev
+ * has an outage wherever refcnt >= vdev_nparity + 1, and a mirror vdev
+ * has an outage wherever refcnt >= vdev_children.
+ */
+static int
+space_reftree_compare(const void *x1, const void *x2)
+{
+	const space_ref_t *sr1 = x1;
+	const space_ref_t *sr2 = x2;
+
+	if (sr1->sr_offset < sr2->sr_offset)
+		return (-1);
+	if (sr1->sr_offset > sr2->sr_offset)
+		return (1);
+
+	if (sr1 < sr2)
+		return (-1);
+	if (sr1 > sr2)
+		return (1);
+
+	return (0);
+}
+
+void
+space_reftree_create(avl_tree_t *t)
+{
+	avl_create(t, space_reftree_compare,
+	    sizeof (space_ref_t), offsetof(space_ref_t, sr_node));
+}
+
+void
+space_reftree_destroy(avl_tree_t *t)
+{
+	space_ref_t *sr;
+	void *cookie = NULL;
+
+	while ((sr = avl_destroy_nodes(t, &cookie)) != NULL)
+		kmem_free(sr, sizeof (*sr));
+
+	avl_destroy(t);
+}
+
+static void
+space_reftree_add_node(avl_tree_t *t, uint64_t offset, int64_t refcnt)
+{
+	space_ref_t *sr;
+
+	sr = kmem_alloc(sizeof (*sr), KM_SLEEP);
+	sr->sr_offset = offset;
+	sr->sr_refcnt = refcnt;
+
+	avl_add(t, sr);
+}
+
+void
+space_reftree_add_seg(avl_tree_t *t, uint64_t start, uint64_t end,
+	int64_t refcnt)
+{
+	space_reftree_add_node(t, start, refcnt);
+	space_reftree_add_node(t, end, -refcnt);
+}
+
+/*
+ * Convert (or add) a range tree into a reference tree.
+ */
+void
+space_reftree_add_map(avl_tree_t *t, range_tree_t *rt, int64_t refcnt)
+{
+	range_seg_t *rs;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+
+	for (rs = avl_first(&rt->rt_root); rs; rs = AVL_NEXT(&rt->rt_root, rs))
+		space_reftree_add_seg(t, rs->rs_start, rs->rs_end, refcnt);
+}
+
+/*
+ * Convert a reference tree into a range tree.  The range tree will contain
+ * all members of the reference tree for which refcnt >= minref.
+ */
+void
+space_reftree_generate_map(avl_tree_t *t, range_tree_t *rt, int64_t minref)
+{
+	uint64_t start = -1ULL;
+	int64_t refcnt = 0;
+	space_ref_t *sr;
+
+	ASSERT(MUTEX_HELD(rt->rt_lock));
+
+	range_tree_vacate(rt, NULL, NULL);
+
+	for (sr = avl_first(t); sr != NULL; sr = AVL_NEXT(t, sr)) {
+		refcnt += sr->sr_refcnt;
+		if (refcnt >= minref) {
+			if (start == -1ULL) {
+				start = sr->sr_offset;
+			}
+		} else {
+			if (start != -1ULL) {
+				uint64_t end = sr->sr_offset;
+				ASSERT(start <= end);
+				if (end > start)
+					range_tree_add(rt, start, end - start);
+				start = -1ULL;
+			}
+		}
+	}
+	ASSERT(refcnt == 0);
+	ASSERT(start == -1ULL);
+}
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/metaslab.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/metaslab.h
@ -20,7 +20,7 @@
 */
 /*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
 */

 #ifndef _SYS_METASLAB_H
@ -36,14 +36,25 @@
 extern "C" {
 #endif

-extern space_map_ops_t *zfs_metaslab_ops;
+typedef struct metaslab_ops {
+	uint64_t (*msop_alloc)(metaslab_t *msp, uint64_t size);
+	boolean_t (*msop_fragmented)(metaslab_t *msp);
+} metaslab_ops_t;

-extern metaslab_t *metaslab_init(metaslab_group_t *mg, space_map_obj_t *smo,
-    uint64_t start, uint64_t size, uint64_t txg);
-extern void metaslab_fini(metaslab_t *msp);
-extern void metaslab_sync(metaslab_t *msp, uint64_t txg);
-extern void metaslab_sync_done(metaslab_t *msp, uint64_t txg);
-extern void metaslab_sync_reassess(metaslab_group_t *mg);
+extern metaslab_ops_t *zfs_metaslab_ops;
+
+metaslab_t *metaslab_init(metaslab_group_t *mg, uint64_t id,
+    uint64_t object, uint64_t txg);
+void metaslab_fini(metaslab_t *msp);
+
+void metaslab_load_wait(metaslab_t *msp);
+int metaslab_load(metaslab_t *msp);
+void metaslab_unload(metaslab_t *msp);
+
+void metaslab_sync(metaslab_t *msp, uint64_t txg);
+void metaslab_sync_done(metaslab_t *msp, uint64_t txg);
+void metaslab_sync_reassess(metaslab_group_t *mg);
+uint64_t metaslab_block_maxsize(metaslab_t *msp);

 #define	METASLAB_HINTBP_FAVOR	0x0
 #define	METASLAB_HINTBP_AVOID	0x1
@ -51,32 +62,29 @@ extern void metaslab_sync_reassess(metaslab_group_t *mg);
 #define	METASLAB_GANG_CHILD	0x4
 #define	METASLAB_GANG_AVOID	0x8

-extern int metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
+int metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
    blkptr_t *bp, int ncopies, uint64_t txg, blkptr_t *hintbp, int flags);
-extern void metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg,
-    boolean_t now);
-extern int metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg);
-extern void metaslab_check_free(spa_t *spa, const blkptr_t *bp);
+void metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now);
+int metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg);
+void metaslab_check_free(spa_t *spa, const blkptr_t *bp);

-extern metaslab_class_t *metaslab_class_create(spa_t *spa,
-    space_map_ops_t *ops);
-extern void metaslab_class_destroy(metaslab_class_t *mc);
-extern int metaslab_class_validate(metaslab_class_t *mc);
+metaslab_class_t *metaslab_class_create(spa_t *spa, metaslab_ops_t *ops);
+void metaslab_class_destroy(metaslab_class_t *mc);
+int metaslab_class_validate(metaslab_class_t *mc);

-extern void metaslab_class_space_update(metaslab_class_t *mc,
+void metaslab_class_space_update(metaslab_class_t *mc,
    int64_t alloc_delta, int64_t defer_delta,
    int64_t space_delta, int64_t dspace_delta);
-extern uint64_t metaslab_class_get_alloc(metaslab_class_t *mc);
-extern uint64_t metaslab_class_get_space(metaslab_class_t *mc);
-extern uint64_t metaslab_class_get_dspace(metaslab_class_t *mc);
-extern uint64_t metaslab_class_get_deferred(metaslab_class_t *mc);
-extern uint64_t metaslab_class_get_minblocksize(metaslab_class_t *mc);
+uint64_t metaslab_class_get_alloc(metaslab_class_t *mc);
+uint64_t metaslab_class_get_space(metaslab_class_t *mc);
+uint64_t metaslab_class_get_dspace(metaslab_class_t *mc);
+uint64_t metaslab_class_get_deferred(metaslab_class_t *mc);
+uint64_t metaslab_class_get_minblocksize(metaslab_class_t *mc);

-extern metaslab_group_t *metaslab_group_create(metaslab_class_t *mc,
-    vdev_t *vd);
-extern void metaslab_group_destroy(metaslab_group_t *mg);
-extern void metaslab_group_activate(metaslab_group_t *mg);
-extern void metaslab_group_passivate(metaslab_group_t *mg);
+metaslab_group_t *metaslab_group_create(metaslab_class_t *mc, vdev_t *vd);
+void metaslab_group_destroy(metaslab_group_t *mg);
+void metaslab_group_activate(metaslab_group_t *mg);
+void metaslab_group_passivate(metaslab_group_t *mg);

 #ifdef	__cplusplus
 }
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/metaslab_impl.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/metaslab_impl.h
@ -32,6 +32,7 @@

 #include <sys/metaslab.h>
 #include <sys/space_map.h>
+#include <sys/range_tree.h>
 #include <sys/vdev.h>
 #include <sys/txg.h>
 #include <sys/avl.h>
@ -43,7 +44,7 @@ extern "C" {
 struct metaslab_class {
 	spa_t			*mc_spa;
 	metaslab_group_t	*mc_rotor;
-	space_map_ops_t		*mc_ops;
+	metaslab_ops_t		*mc_ops;
 	uint64_t		mc_aliquot;
 	uint64_t		mc_alloc_groups; /* # of allocatable groups */
 	uint64_t		mc_alloc;	/* total allocated space */
@ -57,7 +58,6 @@ struct metaslab_group {
 	kmutex_t		mg_lock;
 	avl_tree_t		mg_metaslab_tree;
 	uint64_t		mg_aliquot;
-	uint64_t		mg_bonus_area;
 	uint64_t		mg_alloc_failures;
 	boolean_t		mg_allocatable;		/* can we allocate? */
 	uint64_t		mg_free_capacity;	/* percentage free */
@ -65,45 +65,102 @@ struct metaslab_group {
 	int64_t			mg_activation_count;
 	metaslab_class_t	*mg_class;
 	vdev_t			*mg_vd;
+	taskq_t			*mg_taskq;
 	metaslab_group_t	*mg_prev;
 	metaslab_group_t	*mg_next;
 };

 /*
- * Each metaslab maintains an in-core free map (ms_map) that contains the
- * current list of free segments. As blocks are allocated, the allocated
- * segment is removed from the ms_map and added to a per txg allocation map.
- * As blocks are freed, they are added to the per txg free map. These per
- * txg maps allow us to process all allocations and frees in syncing context
- * where it is safe to update the on-disk space maps.
+ * This value defines the number of elements in the ms_lbas array. The value
+ * of 64 was chosen as it covers to cover all power of 2 buckets up to
+ * UINT64_MAX. This is the equivalent of highbit(UINT64_MAX).
+ */
+#define	MAX_LBAS	64
+
+/*
+ * Each metaslab maintains a set of in-core trees to track metaslab operations.
+ * The in-core free tree (ms_tree) contains the current list of free segments.
+ * As blocks are allocated, the allocated segment are removed from the ms_tree
+ * and added to a per txg allocation tree (ms_alloctree). As blocks are freed,
+ * they are added to the per txg free tree (ms_freetree). These per txg
+ * trees allow us to process all allocations and frees in syncing context
+ * where it is safe to update the on-disk space maps. One additional in-core
+ * tree is maintained to track deferred frees (ms_defertree). Once a block
+ * is freed it will move from the ms_freetree to the ms_defertree. A deferred
+ * free means that a block has been freed but cannot be used by the pool
+ * until TXG_DEFER_SIZE transactions groups later. For example, a block
+ * that is freed in txg 50 will not be available for reallocation until
+ * txg 52 (50 + TXG_DEFER_SIZE).  This provides a safety net for uberblock
+ * rollback. A pool could be safely rolled back TXG_DEFERS_SIZE
+ * transactions groups and ensure that no block has been reallocated.
 *
- * Each metaslab's free space is tracked in a space map object in the MOS,
+ * The simplified transition diagram looks like this:
+ *
+ *
+ *      ALLOCATE
+ *         |
+ *         V
+ *    free segment (ms_tree) --------> ms_alloctree ----> (write to space map)
+ *         ^
+ *         |
+ *         |                           ms_freetree <--- FREE
+ *         |                                 |
+ *         |                                 |
+ *         |                                 |
+ *         +----------- ms_defertree <-------+---------> (write to space map)
+ *
+ *
+ * Each metaslab's space is tracked in a single space map in the MOS,
 * which is only updated in syncing context. Each time we sync a txg,
- * we append the allocs and frees from that txg to the space map object.
- * When the txg is done syncing, metaslab_sync_done() updates ms_smo
- * to ms_smo_syncing. Everything in ms_smo is always safe to allocate.
+ * we append the allocs and frees from that txg to the space map.
+ * The pool space is only updated once all metaslabs have finished syncing.
 *
- * To load the in-core free map we read the space map object from disk.
+ * To load the in-core free tree we read the space map from disk.
 * This object contains a series of alloc and free records that are
 * combined to make up the list of all free segments in this metaslab. These
- * segments are represented in-core by the ms_map and are stored in an
+ * segments are represented in-core by the ms_tree and are stored in an
 * AVL tree.
 *
- * As the space map objects grows (as a result of the appends) it will
- * eventually become space-inefficient. When the space map object is
- * zfs_condense_pct/100 times the size of the minimal on-disk representation,
- * we rewrite it in its minimized form.
+ * As the space map grows (as a result of the appends) it will
+ * eventually become space-inefficient. When the metaslab's in-core free tree
+ * is zfs_condense_pct/100 times the size of the minimal on-disk
+ * representation, we rewrite it in its minimized form. If a metaslab
+ * needs to condense then we must set the ms_condensing flag to ensure
+ * that allocations are not performed on the metaslab that is being written.
 */
 struct metaslab {
-	kmutex_t	ms_lock;	/* metaslab lock		*/
-	space_map_obj_t	ms_smo;		/* synced space map object	*/
-	space_map_obj_t	ms_smo_syncing;	/* syncing space map object	*/
-	space_map_t	*ms_allocmap[TXG_SIZE];	/* allocated this txg	*/
-	space_map_t	*ms_freemap[TXG_SIZE];	/* freed this txg	*/
-	space_map_t	*ms_defermap[TXG_DEFER_SIZE];	/* deferred frees */
-	space_map_t	*ms_map;	/* in-core free space map	*/
+	kmutex_t	ms_lock;
+	kcondvar_t	ms_load_cv;
+	space_map_t	*ms_sm;
+	metaslab_ops_t	*ms_ops;
+	uint64_t	ms_id;
+	uint64_t	ms_start;
+	uint64_t	ms_size;
+
+	range_tree_t	*ms_alloctree[TXG_SIZE];
+	range_tree_t	*ms_freetree[TXG_SIZE];
+	range_tree_t	*ms_defertree[TXG_DEFER_SIZE];
+	range_tree_t	*ms_tree;
+
+	boolean_t	ms_condensing;	/* condensing? */
+	boolean_t	ms_loaded;
+	boolean_t	ms_loading;
+
 	int64_t		ms_deferspace;	/* sum of ms_defermap[] space	*/
 	uint64_t	ms_weight;	/* weight vs. others in group	*/
+	uint64_t	ms_factor;
+	uint64_t	ms_access_txg;
+
+	/*
+	 * The metaslab block allocators can optionally use a size-ordered
+	 * range tree and/or an array of LBAs. Not all allocators use
+	 * this functionality. The ms_size_tree should always contain the
+	 * same number of segments as the ms_tree. The only difference
+	 * is that the ms_size_tree is ordered by segment sizes.
+	 */
+	avl_tree_t	ms_size_tree;
+	uint64_t	ms_lbas[MAX_LBAS];
+
 	metaslab_group_t *ms_group;	/* metaslab group		*/
 	avl_node_t	ms_group_node;	/* node in metaslab group tree	*/
 	txg_node_t	ms_txg_node;	/* per-txg dirty metaslab links	*/
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/range_tree.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/range_tree.h
@ -0,0 +1,96 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * Copyright (c) 2013 by Delphix. All rights reserved.
+ */
+
+#ifndef _SYS_RANGE_TREE_H
+#define	_SYS_RANGE_TREE_H
+
+#include <sys/avl.h>
+#include <sys/dmu.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+#define	RANGE_TREE_HISTOGRAM_SIZE	64
+
+typedef struct range_tree_ops range_tree_ops_t;
+
+typedef struct range_tree {
+	avl_tree_t	rt_root;	/* offset-ordered segment AVL tree */
+	uint64_t	rt_space;	/* sum of all segments in the map */
+	range_tree_ops_t *rt_ops;
+	void		*rt_arg;
+
+	/*
+	 * The rt_histogram maintains a histogram of ranges. Each bucket,
+	 * rt_histogram[i], contains the number of ranges whose size is:
+	 * 2^i <= size of range in bytes < 2^(i+1)
+	 */
+	uint64_t	rt_histogram[RANGE_TREE_HISTOGRAM_SIZE];
+	kmutex_t	*rt_lock;	/* pointer to lock that protects map */
+} range_tree_t;
+
+typedef struct range_seg {
+	avl_node_t	rs_node;	/* AVL node */
+	avl_node_t	rs_pp_node;	/* AVL picker-private node */
+	uint64_t	rs_start;	/* starting offset of this segment */
+	uint64_t	rs_end;		/* ending offset (non-inclusive) */
+} range_seg_t;
+
+struct range_tree_ops {
+	void    (*rtop_create)(range_tree_t *rt, void *arg);
+	void    (*rtop_destroy)(range_tree_t *rt, void *arg);
+	void	(*rtop_add)(range_tree_t *rt, range_seg_t *rs, void *arg);
+	void    (*rtop_remove)(range_tree_t *rt, range_seg_t *rs, void *arg);
+	void	(*rtop_vacate)(range_tree_t *rt, void *arg);
+};
+
+typedef void range_tree_func_t(void *arg, uint64_t start, uint64_t size);
+
+void range_tree_init(void);
+void range_tree_fini(void);
+range_tree_t *range_tree_create(range_tree_ops_t *ops, void *arg, kmutex_t *lp);
+void range_tree_destroy(range_tree_t *rt);
+boolean_t range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size);
+uint64_t range_tree_space(range_tree_t *rt);
+void range_tree_verify(range_tree_t *rt, uint64_t start, uint64_t size);
+void range_tree_swap(range_tree_t **rtsrc, range_tree_t **rtdst);
+void range_tree_stat_verify(range_tree_t *rt);
+
+void range_tree_add(void *arg, uint64_t start, uint64_t size);
+void range_tree_remove(void *arg, uint64_t start, uint64_t size);
+
+void range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg);
+void range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _SYS_RANGE_TREE_H */
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/space_map.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/space_map.h
@ -24,66 +24,72 @@
 */

 /*
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
 */

 #ifndef _SYS_SPACE_MAP_H
 #define	_SYS_SPACE_MAP_H

 #include <sys/avl.h>
+#include <sys/range_tree.h>
 #include <sys/dmu.h>

 #ifdef	__cplusplus
 extern "C" {
 #endif

-typedef struct space_map_ops space_map_ops_t;
+/*
+ * The size of the space map object has increased to include a histogram.
+ * The SPACE_MAP_SIZE_V0 designates the original size and is used to
+ * maintain backward compatibility.
+ */
+#define	SPACE_MAP_SIZE_V0	(3 * sizeof (uint64_t))
+#define	SPACE_MAP_HISTOGRAM_SIZE(sm)			\
+	(sizeof ((sm)->sm_phys->smp_histogram) /	\
+	sizeof ((sm)->sm_phys->smp_histogram[0]))

+/*
+ * The space_map_phys is the on-disk representation of the space map.
+ * Consumers of space maps should never reference any of the members of this
+ * structure directly. These members may only be updated in syncing context.
+ *
+ * Note the smp_object is no longer used but remains in the structure
+ * for backward compatibility.
+ */
+typedef struct space_map_phys {
+	uint64_t	smp_object;	/* on-disk space map object */
+	uint64_t	smp_objsize;	/* size of the object */
+	uint64_t	smp_alloc;	/* space allocated from the map */
+	uint64_t	smp_pad[5];	/* reserved */
+
+	/*
+	 * The smp_histogram maintains a histogram of free regions. Each
+	 * bucket, smp_histogram[i], contains the number of free regions
+	 * whose size is:
+	 * 2^(i+sm_shift) <= size of free region in bytes < 2^(i+sm_shift+1)
+	 */
+	uint64_t	smp_histogram[32]; /* histogram of free space */
+} space_map_phys_t;
+
+/*
+ * The space map object defines a region of space, its size, how much is
+ * allocated, and the on-disk object that stores this information.
+ * Consumers of space maps may only access the members of this structure.
+ */
 typedef struct space_map {
-	avl_tree_t	sm_root;	/* offset-ordered segment AVL tree */
-	uint64_t	sm_space;	/* sum of all segments in the map */
 	uint64_t	sm_start;	/* start of map */
 	uint64_t	sm_size;	/* size of map */
 	uint8_t		sm_shift;	/* unit shift */
-	uint8_t		sm_loaded;	/* map loaded? */
-	uint8_t		sm_loading;	/* map loading? */
-	uint8_t		sm_condensing;	/* map condensing? */
-	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;	/* size-ordered, picker-private tree */
-	void		*sm_ppd;	/* picker-private data */
+	uint64_t	sm_length;	/* synced length */
+	uint64_t	sm_alloc;	/* synced space allocated */
+	objset_t	*sm_os;		/* objset for this map */
+	uint64_t	sm_object;	/* object id for this map */
+	uint32_t	sm_blksz;	/* block size for space map */
+	dmu_buf_t	*sm_dbuf;	/* space_map_phys_t dbuf */
+	space_map_phys_t *sm_phys;	/* on-disk space map */
 	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;
-
-typedef struct space_ref {
-	avl_node_t	sr_node;	/* AVL node */
-	uint64_t	sr_offset;	/* offset (start or end) */
-	int64_t		sr_refcnt;	/* associated reference count */
-} space_ref_t;
-
-typedef struct space_map_obj {
-	uint64_t	smo_object;	/* on-disk space map object */
-	uint64_t	smo_objsize;	/* size of the object */
-	uint64_t	smo_alloc;	/* space allocated from the map */
-} space_map_obj_t;
-
-struct space_map_ops {
-	void	(*smop_load)(space_map_t *sm);
-	void	(*smop_unload)(space_map_t *sm);
-	uint64_t (*smop_alloc)(space_map_t *sm, uint64_t size);
-	void	(*smop_claim)(space_map_t *sm, uint64_t start, uint64_t size);
-	void	(*smop_free)(space_map_t *sm, uint64_t start, uint64_t size);
-	uint64_t (*smop_max)(space_map_t *sm);
-	boolean_t (*smop_fragmented)(space_map_t *sm);
-};
-
 /*
 * debug entry
 *
@ -124,61 +130,45 @@ struct space_map_ops {

 #define	SM_RUN_MAX			SM_RUN_DECODE(~0ULL)

-#define	SM_ALLOC	0x0
-#define	SM_FREE		0x1
+typedef enum {
+	SM_ALLOC,
+	SM_FREE
+} maptype_t;

 /*
 * The data for a given space map can be kept on blocks of any size.
 * Larger blocks entail fewer i/o operations, but they also cause the
 * DMU to keep more data in-core, and also to waste more i/o bandwidth
 * when only a few blocks have changed since the last transaction group.
- * This could use a lot more research, but for now, set the freelist
- * block size to 4k (2^12).
+ * Rather than having a fixed block size for all space maps the block size
+ * can adjust as needed (see space_map_max_blksz). Set the initial block
+ * size for the space map to 4k.
 */
-#define	SPACE_MAP_BLOCKSHIFT	12
+#define	SPACE_MAP_INITIAL_BLOCKSIZE	(1ULL << 12)

-typedef void space_map_func_t(space_map_t *sm, uint64_t start, uint64_t size);
+int space_map_load(space_map_t *sm, range_tree_t *rt, maptype_t maptype);

-extern void space_map_init(void);
-extern void space_map_fini(void);
-extern void space_map_create(space_map_t *sm, uint64_t start, uint64_t size,
-    uint8_t shift, kmutex_t *lp);
-extern void space_map_destroy(space_map_t *sm);
-extern void space_map_add(space_map_t *sm, uint64_t start, uint64_t size);
-extern void space_map_remove(space_map_t *sm, uint64_t start, uint64_t size);
-extern boolean_t space_map_contains(space_map_t *sm,
-    uint64_t start, uint64_t size);
-extern space_seg_t *space_map_find(space_map_t *sm, uint64_t start,
-    uint64_t size, avl_index_t *wherep);
-extern void space_map_swap(space_map_t **msrc, space_map_t **mdest);
-extern void space_map_vacate(space_map_t *sm,
-    space_map_func_t *func, space_map_t *mdest);
-extern void space_map_walk(space_map_t *sm,
-    space_map_func_t *func, space_map_t *mdest);
+void space_map_histogram_clear(space_map_t *sm);
+void space_map_histogram_add(space_map_t *sm, range_tree_t *rt,
+    dmu_tx_t *tx);

-extern void space_map_load_wait(space_map_t *sm);
-extern int space_map_load(space_map_t *sm, space_map_ops_t *ops,
-    uint8_t maptype, space_map_obj_t *smo, objset_t *os);
-extern void space_map_unload(space_map_t *sm);
+void space_map_update(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);
+uint64_t space_map_object(space_map_t *sm);
+uint64_t space_map_allocated(space_map_t *sm);
+uint64_t space_map_length(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);
-extern void space_map_truncate(space_map_obj_t *smo,
-    objset_t *os, dmu_tx_t *tx);
+void space_map_write(space_map_t *sm, range_tree_t *rt, maptype_t maptype,
+    dmu_tx_t *tx);
+void space_map_truncate(space_map_t *sm, dmu_tx_t *tx);
+uint64_t space_map_alloc(objset_t *os, dmu_tx_t *tx);
+void space_map_free(space_map_t *sm, dmu_tx_t *tx);

-extern void space_map_ref_create(avl_tree_t *t);
-extern void space_map_ref_destroy(avl_tree_t *t);
-extern void space_map_ref_add_seg(avl_tree_t *t,
-    uint64_t start, uint64_t end, int64_t refcnt);
-extern void space_map_ref_add_map(avl_tree_t *t,
-    space_map_t *sm, int64_t refcnt);
-extern void space_map_ref_generate_map(avl_tree_t *t,
-    space_map_t *sm, int64_t minref);
+int space_map_open(space_map_t **smp, objset_t *os, uint64_t object,
+    uint64_t start, uint64_t size, uint8_t shift, kmutex_t *lp);
+void space_map_close(space_map_t *sm);
+
+int64_t space_map_alloc_delta(space_map_t *sm);

 #ifdef	__cplusplus
 }
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/space_reftree.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/space_reftree.h
@ -0,0 +1,57 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * Copyright (c) 2013 by Delphix. All rights reserved.
+ */
+
+#ifndef _SYS_SPACE_REFTREE_H
+#define	_SYS_SPACE_REFTREE_H
+
+#include <sys/range_tree.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+typedef struct space_ref {
+	avl_node_t	sr_node;	/* AVL node */
+	uint64_t	sr_offset;	/* range offset (start or end) */
+	int64_t		sr_refcnt;	/* associated reference count */
+} space_ref_t;
+
+void space_reftree_create(avl_tree_t *t);
+void space_reftree_destroy(avl_tree_t *t);
+void space_reftree_add_seg(avl_tree_t *t, uint64_t start, uint64_t end,
+    int64_t refcnt);
+void space_reftree_add_map(avl_tree_t *t, range_tree_t *rt, int64_t refcnt);
+void space_reftree_generate_map(avl_tree_t *t, range_tree_t *rt,
+    int64_t minref);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _SYS_SPACE_REFTREE_H */
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h
@ -163,7 +163,6 @@ struct vdev {
 	vdev_t		*vdev_parent;	/* parent vdev			*/
 	vdev_t		**vdev_child;	/* array of children		*/
 	uint64_t	vdev_children;	/* number of children		*/
-	space_map_t	vdev_dtl[DTL_TYPES]; /* in-core dirty time logs	*/
 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
 	boolean_t	vdev_expanding;	/* expand the vdev?		*/
 	boolean_t	vdev_reopening;	/* reopen in progress?		*/
@ -184,19 +183,21 @@ struct vdev {
 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
 	boolean_t	vdev_remove_wanted; /* async remove wanted?	*/
 	boolean_t	vdev_probe_wanted; /* async probe wanted?	*/
-	uint64_t	vdev_removing;	/* device is being removed?	*/
 	list_node_t	vdev_config_dirty_node; /* config dirty list	*/
 	list_node_t	vdev_state_dirty_node; /* state dirty list	*/
 	uint64_t	vdev_deflate_ratio; /* deflation ratio (x512)	*/
 	uint64_t	vdev_islog;	/* is an intent log device	*/
-	uint64_t	vdev_ishole;	/* is a hole in the namespace 	*/
+	uint64_t	vdev_removing;	/* device is being removed?	*/
+	boolean_t	vdev_ishole;	/* is a hole in the namespace 	*/

 	/*
 	 * Leaf vdev state.
 	 */
-	uint64_t	vdev_psize;	/* physical device capacity	*/
-	space_map_obj_t	vdev_dtl_smo;	/* dirty time log space map obj	*/
+	range_tree_t	*vdev_dtl[DTL_TYPES]; /* dirty time logs	*/
+	space_map_t	*vdev_dtl_sm;	/* dirty time log space map	*/
 	txg_node_t	vdev_dtl_node;	/* per-txg dirty DTL linkage	*/
+	uint64_t	vdev_dtl_object; /* DTL object			*/
+	uint64_t	vdev_psize;	/* physical device capacity	*/
 	uint64_t	vdev_wholedisk;	/* true if this is a whole disk */
 	uint64_t	vdev_offline;	/* persistent offline state	*/
 	uint64_t	vdev_faulted;	/* persistent faulted state	*/
@ -210,19 +211,18 @@ struct vdev {
 	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		*/
 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
 	boolean_t	vdev_notrim;	/* true if trim failed */
 	boolean_t	vdev_checkremove; /* temporary online test	*/
 	boolean_t	vdev_forcefault; /* force online fault		*/
 	boolean_t	vdev_splitting;	/* split or repair in progress  */
 	boolean_t	vdev_delayed_close; /* delayed device close?	*/
-	uint8_t		vdev_tmpoffline; /* device taken offline temporarily? */
-	uint8_t		vdev_detached;	/* device detached?		*/
-	uint8_t		vdev_cant_read;	/* vdev is failing all reads	*/
-	uint8_t		vdev_cant_write; /* vdev is failing all writes	*/
-	uint64_t	vdev_isspare;	/* was a hot spare		*/
-	uint64_t	vdev_isl2cache;	/* was a l2cache device		*/
+	boolean_t	vdev_tmpoffline; /* device taken offline temporarily? */
+	boolean_t	vdev_detached;	/* device detached?		*/
+	boolean_t	vdev_cant_read;	/* vdev is failing all reads	*/
+	boolean_t	vdev_cant_write; /* vdev is failing all writes	*/
+	boolean_t	vdev_isspare;	/* was a hot spare		*/
+	boolean_t	vdev_isl2cache;	/* was a l2cache device		*/
 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
 	vdev_cache_t	vdev_cache;	/* physical block cache		*/
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache vdevs		*/
@ -327,9 +327,11 @@ extern void vdev_remove_parent(vdev_t *cvd);
 extern void vdev_load_log_state(vdev_t *nvd, vdev_t *ovd);
 extern boolean_t vdev_log_state_valid(vdev_t *vd);
 extern void vdev_load(vdev_t *vd);
+extern int vdev_dtl_load(vdev_t *vd);
 extern void vdev_sync(vdev_t *vd, uint64_t txg);
 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
+extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);

 /*
 * Available vdev types.
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfeature.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfeature.h
@ -20,7 +20,7 @@
 */

 /*
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
 */

 #ifndef _SYS_ZFEATURE_H
@ -47,6 +47,7 @@ extern void spa_feature_incr(struct spa *, zfeature_info_t *, struct dmu_tx *);
 extern void spa_feature_decr(struct spa *, zfeature_info_t *, struct dmu_tx *);
 extern boolean_t spa_feature_is_enabled(struct spa *, zfeature_info_t *);
 extern boolean_t spa_feature_is_active(struct spa *, zfeature_info_t *);
+extern int spa_feature_get_refcount(struct spa *, zfeature_info_t *);

 #ifdef	__cplusplus
 }
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
@ -37,6 +37,7 @@
 #include <sys/metaslab.h>
 #include <sys/metaslab_impl.h>
 #include <sys/space_map.h>
+#include <sys/space_reftree.h>
 #include <sys/zio.h>
 #include <sys/zap.h>
 #include <sys/fs/zfs.h>
@ -371,7 +372,7 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
 	mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
 	for (int t = 0; t < DTL_TYPES; t++) {
-		space_map_create(&vd->vdev_dtl[t], 0, -1ULL, 0,
+		vd->vdev_dtl[t] = range_tree_create(NULL, NULL,
 		    &vd->vdev_dtl_lock);
 	}
 	txg_list_create(&vd->vdev_ms_list,
@ -557,7 +558,7 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	    alloctype == VDEV_ALLOC_ROOTPOOL)) {
 		if (alloctype == VDEV_ALLOC_LOAD) {
 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
-			    &vd->vdev_dtl_smo.smo_object);
+			    &vd->vdev_dtl_object);
 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
 			    &vd->vdev_unspare);
 		}
@ -679,9 +680,10 @@ vdev_free(vdev_t *vd)
 	txg_list_destroy(&vd->vdev_dtl_list);

 	mutex_enter(&vd->vdev_dtl_lock);
+	space_map_close(vd->vdev_dtl_sm);
 	for (int t = 0; t < DTL_TYPES; t++) {
-		space_map_unload(&vd->vdev_dtl[t]);
-		space_map_destroy(&vd->vdev_dtl[t]);
+		range_tree_vacate(vd->vdev_dtl[t], NULL, NULL);
+		range_tree_destroy(vd->vdev_dtl[t]);
 	}
 	mutex_exit(&vd->vdev_dtl_lock);

@ -896,27 +898,16 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 	vd->vdev_ms_count = newc;

 	for (m = oldc; m < newc; m++) {
-		space_map_obj_t smo = { 0, 0, 0 };
+		uint64_t object = 0;
+
 		if (txg == 0) {
-			uint64_t object = 0;
 			error = dmu_read(mos, vd->vdev_ms_array,
 			    m * sizeof (uint64_t), sizeof (uint64_t), &object,
 			    DMU_READ_PREFETCH);
 			if (error)
 				return (error);
-			if (object != 0) {
-				dmu_buf_t *db;
-				error = dmu_bonus_hold(mos, object, FTAG, &db);
-				if (error)
-					return (error);
-				ASSERT3U(db->db_size, >=, sizeof (smo));
-				bcopy(db->db_data, &smo, sizeof (smo));
-				ASSERT3U(smo.smo_object, ==, object);
-				dmu_buf_rele(db, FTAG);
-			}
 		}
-		vd->vdev_ms[m] = metaslab_init(vd->vdev_mg, &smo,
-		    m << vd->vdev_ms_shift, 1ULL << vd->vdev_ms_shift, txg);
+		vd->vdev_ms[m] = metaslab_init(vd->vdev_mg, m, object, txg);
 	}

 	if (txg == 0)
@ -944,9 +935,12 @@ vdev_metaslab_fini(vdev_t *vd)

 	if (vd->vdev_ms != NULL) {
 		metaslab_group_passivate(vd->vdev_mg);
-		for (m = 0; m < count; m++)
-			if (vd->vdev_ms[m] != NULL)
-				metaslab_fini(vd->vdev_ms[m]);
+		for (m = 0; m < count; m++) {
+			metaslab_t *msp = vd->vdev_ms[m];
+
+			if (msp != NULL)
+				metaslab_fini(msp);
+		}
 		kmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
 		vd->vdev_ms = NULL;
 	}
@ -1614,9 +1608,10 @@ vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
 	}

 	/*
-	 * Recursively initialize all labels.
+	 * Recursively load DTLs and initialize all labels.
 	 */
-	if ((error = vdev_label_init(vd, txg, isreplacing ?
+	if ((error = vdev_dtl_load(vd)) != 0 ||
+	    (error = vdev_label_init(vd, txg, isreplacing ?
 	    VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
 		vdev_close(vd);
 		return (error);
@ -1669,6 +1664,16 @@ vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
 	(void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
 }

+void
+vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg)
+{
+	for (int c = 0; c < vd->vdev_children; c++)
+		vdev_dirty_leaves(vd->vdev_child[c], flags, txg);
+
+	if (vd->vdev_ops->vdev_op_leaf)
+		vdev_dirty(vd->vdev_top, flags, vd, txg);
+}
+
 /*
 * DTLs.
 *
@ -1710,31 +1715,31 @@ vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
 void
 vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
 {
-	space_map_t *sm = &vd->vdev_dtl[t];
+	range_tree_t *rt = vd->vdev_dtl[t];

 	ASSERT(t < DTL_TYPES);
 	ASSERT(vd != vd->vdev_spa->spa_root_vdev);
 	ASSERT(spa_writeable(vd->vdev_spa));

-	mutex_enter(sm->sm_lock);
-	if (!space_map_contains(sm, txg, size))
-		space_map_add(sm, txg, size);
-	mutex_exit(sm->sm_lock);
+	mutex_enter(rt->rt_lock);
+	if (!range_tree_contains(rt, txg, size))
+		range_tree_add(rt, txg, size);
+	mutex_exit(rt->rt_lock);
 }

 boolean_t
 vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
 {
-	space_map_t *sm = &vd->vdev_dtl[t];
+	range_tree_t *rt = vd->vdev_dtl[t];
 	boolean_t dirty = B_FALSE;

 	ASSERT(t < DTL_TYPES);
 	ASSERT(vd != vd->vdev_spa->spa_root_vdev);

-	mutex_enter(sm->sm_lock);
-	if (sm->sm_space != 0)
-		dirty = space_map_contains(sm, txg, size);
-	mutex_exit(sm->sm_lock);
+	mutex_enter(rt->rt_lock);
+	if (range_tree_space(rt) != 0)
+		dirty = range_tree_contains(rt, txg, size);
+	mutex_exit(rt->rt_lock);

 	return (dirty);
 }
@ -1742,12 +1747,12 @@ vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
 boolean_t
 vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
 {
-	space_map_t *sm = &vd->vdev_dtl[t];
+	range_tree_t *rt = vd->vdev_dtl[t];
 	boolean_t empty;

-	mutex_enter(sm->sm_lock);
-	empty = (sm->sm_space == 0);
-	mutex_exit(sm->sm_lock);
+	mutex_enter(rt->rt_lock);
+	empty = (range_tree_space(rt) == 0);
+	mutex_exit(rt->rt_lock);

 	return (empty);
 }
@ -1758,14 +1763,14 @@ vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
 static uint64_t
 vdev_dtl_min(vdev_t *vd)
 {
-	space_seg_t *ss;
+	range_seg_t *rs;

 	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
-	ASSERT3U(vd->vdev_dtl[DTL_MISSING].sm_space, !=, 0);
+	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
 	ASSERT0(vd->vdev_children);

-	ss = avl_first(&vd->vdev_dtl[DTL_MISSING].sm_root);
-	return (ss->ss_start - 1);
+	rs = avl_first(&vd->vdev_dtl[DTL_MISSING]->rt_root);
+	return (rs->rs_start - 1);
 }

 /*
@ -1774,14 +1779,14 @@ vdev_dtl_min(vdev_t *vd)
 static uint64_t
 vdev_dtl_max(vdev_t *vd)
 {
-	space_seg_t *ss;
+	range_seg_t *rs;

 	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
-	ASSERT3U(vd->vdev_dtl[DTL_MISSING].sm_space, !=, 0);
+	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
 	ASSERT0(vd->vdev_children);

-	ss = avl_last(&vd->vdev_dtl[DTL_MISSING].sm_root);
-	return (ss->ss_end);
+	rs = avl_last(&vd->vdev_dtl[DTL_MISSING]->rt_root);
+	return (rs->rs_end);
 }

 /*
@ -1802,7 +1807,7 @@ vdev_dtl_should_excise(vdev_t *vd)
 	ASSERT0(vd->vdev_children);

 	if (vd->vdev_resilver_txg == 0 ||
-	    vd->vdev_dtl[DTL_MISSING].sm_space == 0)
+	    range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0)
 		return (B_TRUE);

 	/*
@ -1872,35 +1877,35 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done)
 			 * positive refcnt -- either 1 or 2.  We then convert
 			 * the reference tree into the new DTL_MISSING map.
 			 */
-			space_map_ref_create(&reftree);
-			space_map_ref_add_map(&reftree,
-			    &vd->vdev_dtl[DTL_MISSING], 1);
-			space_map_ref_add_seg(&reftree, 0, scrub_txg, -1);
-			space_map_ref_add_map(&reftree,
-			    &vd->vdev_dtl[DTL_SCRUB], 2);
-			space_map_ref_generate_map(&reftree,
-			    &vd->vdev_dtl[DTL_MISSING], 1);
-			space_map_ref_destroy(&reftree);
+			space_reftree_create(&reftree);
+			space_reftree_add_map(&reftree,
+			    vd->vdev_dtl[DTL_MISSING], 1);
+			space_reftree_add_seg(&reftree, 0, scrub_txg, -1);
+			space_reftree_add_map(&reftree,
+			    vd->vdev_dtl[DTL_SCRUB], 2);
+			space_reftree_generate_map(&reftree,
+			    vd->vdev_dtl[DTL_MISSING], 1);
+			space_reftree_destroy(&reftree);
 		}
-		space_map_vacate(&vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
-		space_map_walk(&vd->vdev_dtl[DTL_MISSING],
-		    space_map_add, &vd->vdev_dtl[DTL_PARTIAL]);
+		range_tree_vacate(vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
+		range_tree_walk(vd->vdev_dtl[DTL_MISSING],
+		    range_tree_add, vd->vdev_dtl[DTL_PARTIAL]);
 		if (scrub_done)
-			space_map_vacate(&vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
-		space_map_vacate(&vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
+			range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
+		range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
 		if (!vdev_readable(vd))
-			space_map_add(&vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
+			range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
 		else
-			space_map_walk(&vd->vdev_dtl[DTL_MISSING],
-			    space_map_add, &vd->vdev_dtl[DTL_OUTAGE]);
+			range_tree_walk(vd->vdev_dtl[DTL_MISSING],
+			    range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);

 		/*
 		 * If the vdev was resilvering and no longer has any
 		 * DTLs then reset its resilvering flag.
 		 */
 		if (vd->vdev_resilver_txg != 0 &&
-		    vd->vdev_dtl[DTL_MISSING].sm_space == 0 &&
-		    vd->vdev_dtl[DTL_OUTAGE].sm_space == 0)
+		    range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0 &&
+		    range_tree_space(vd->vdev_dtl[DTL_OUTAGE]) == 0)
 			vd->vdev_resilver_txg = 0;

 		mutex_exit(&vd->vdev_dtl_lock);
@ -1922,46 +1927,55 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done)
 			minref = vd->vdev_nparity + 1;	/* RAID-Z */
 		else
 			minref = vd->vdev_children;	/* any kind of mirror */
-		space_map_ref_create(&reftree);
+		space_reftree_create(&reftree);
 		for (int c = 0; c < vd->vdev_children; c++) {
 			vdev_t *cvd = vd->vdev_child[c];
 			mutex_enter(&cvd->vdev_dtl_lock);
-			space_map_ref_add_map(&reftree, &cvd->vdev_dtl[s], 1);
+			space_reftree_add_map(&reftree, cvd->vdev_dtl[s], 1);
 			mutex_exit(&cvd->vdev_dtl_lock);
 		}
-		space_map_ref_generate_map(&reftree, &vd->vdev_dtl[t], minref);
-		space_map_ref_destroy(&reftree);
+		space_reftree_generate_map(&reftree, vd->vdev_dtl[t], minref);
+		space_reftree_destroy(&reftree);
 	}
 	mutex_exit(&vd->vdev_dtl_lock);
 }

-static int
+int
 vdev_dtl_load(vdev_t *vd)
 {
 	spa_t *spa = vd->vdev_spa;
-	space_map_obj_t *smo = &vd->vdev_dtl_smo;
 	objset_t *mos = spa->spa_meta_objset;
-	dmu_buf_t *db;
-	int error;
+	int error = 0;

-	ASSERT(vd->vdev_children == 0);
+	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
+		ASSERT(!vd->vdev_ishole);

-	if (smo->smo_object == 0)
-		return (0);
+		error = space_map_open(&vd->vdev_dtl_sm, mos,
+		    vd->vdev_dtl_object, 0, -1ULL, 0, &vd->vdev_dtl_lock);
+		if (error)
+			return (error);
+		ASSERT(vd->vdev_dtl_sm != NULL);

-	ASSERT(!vd->vdev_ishole);
+		mutex_enter(&vd->vdev_dtl_lock);
+
+		/*
+		 * Now that we've opened the space_map we need to update
+		 * the in-core DTL.
+		 */
+		space_map_update(vd->vdev_dtl_sm);
+
+		error = space_map_load(vd->vdev_dtl_sm,
+		    vd->vdev_dtl[DTL_MISSING], SM_ALLOC);
+		mutex_exit(&vd->vdev_dtl_lock);

-	if ((error = dmu_bonus_hold(mos, smo->smo_object, FTAG, &db)) != 0)
 		return (error);
+	}

-	ASSERT3U(db->db_size, >=, sizeof (*smo));
-	bcopy(db->db_data, smo, sizeof (*smo));
-	dmu_buf_rele(db, FTAG);
-
-	mutex_enter(&vd->vdev_dtl_lock);
-	error = space_map_load(&vd->vdev_dtl[DTL_MISSING],
-	    NULL, SM_ALLOC, smo, mos);
-	mutex_exit(&vd->vdev_dtl_lock);
+	for (int c = 0; c < vd->vdev_children; c++) {
+		error = vdev_dtl_load(vd->vdev_child[c]);
+		if (error != 0)
+			break;
+	}

 	return (error);
 }
@ -1970,66 +1984,75 @@ void
 vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 {
 	spa_t *spa = vd->vdev_spa;
-	space_map_obj_t *smo = &vd->vdev_dtl_smo;
-	space_map_t *sm = &vd->vdev_dtl[DTL_MISSING];
+	range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
 	objset_t *mos = spa->spa_meta_objset;
-	space_map_t smsync;
-	kmutex_t smlock;
-	dmu_buf_t *db;
+	range_tree_t *rtsync;
+	kmutex_t rtlock;
 	dmu_tx_t *tx;
+	uint64_t object = space_map_object(vd->vdev_dtl_sm);

 	ASSERT(!vd->vdev_ishole);
+	ASSERT(vd->vdev_ops->vdev_op_leaf);

 	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);

-	if (vd->vdev_detached) {
-		if (smo->smo_object != 0) {
-			int err = dmu_object_free(mos, smo->smo_object, tx);
-			ASSERT0(err);
-			smo->smo_object = 0;
-		}
+	if (vd->vdev_detached || vd->vdev_top->vdev_removing) {
+		mutex_enter(&vd->vdev_dtl_lock);
+		space_map_free(vd->vdev_dtl_sm, tx);
+		space_map_close(vd->vdev_dtl_sm);
+		vd->vdev_dtl_sm = NULL;
+		mutex_exit(&vd->vdev_dtl_lock);
 		dmu_tx_commit(tx);
 		return;
 	}

-	if (smo->smo_object == 0) {
-		ASSERT(smo->smo_objsize == 0);
-		ASSERT(smo->smo_alloc == 0);
-		smo->smo_object = dmu_object_alloc(mos,
-		    DMU_OT_SPACE_MAP, 1 << SPACE_MAP_BLOCKSHIFT,
-		    DMU_OT_SPACE_MAP_HEADER, sizeof (*smo), tx);
-		ASSERT(smo->smo_object != 0);
+	if (vd->vdev_dtl_sm == NULL) {
+		uint64_t new_object;
+
+		new_object = space_map_alloc(mos, tx);
+		VERIFY3U(new_object, !=, 0);
+
+		VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
+		    0, -1ULL, 0, &vd->vdev_dtl_lock));
+		ASSERT(vd->vdev_dtl_sm != NULL);
+	}
+
+	bzero(&rtlock, sizeof(rtlock));
+	mutex_init(&rtlock, NULL, MUTEX_DEFAULT, NULL);
+
+	rtsync = range_tree_create(NULL, NULL, &rtlock);
+
+	mutex_enter(&rtlock);
+
+	mutex_enter(&vd->vdev_dtl_lock);
+	range_tree_walk(rt, range_tree_add, rtsync);
+	mutex_exit(&vd->vdev_dtl_lock);
+
+	space_map_truncate(vd->vdev_dtl_sm, tx);
+	space_map_write(vd->vdev_dtl_sm, rtsync, SM_ALLOC, tx);
+	range_tree_vacate(rtsync, NULL, NULL);
+
+	range_tree_destroy(rtsync);
+
+	mutex_exit(&rtlock);
+	mutex_destroy(&rtlock);
+
+	/*
+	 * If the object for the space map has changed then dirty
+	 * the top level so that we update the config.
+	 */
+	if (object != space_map_object(vd->vdev_dtl_sm)) {
+		zfs_dbgmsg("txg %llu, spa %s, DTL old object %llu, "
+		    "new object %llu", txg, spa_name(spa), object,
+		    space_map_object(vd->vdev_dtl_sm));
 		vdev_config_dirty(vd->vdev_top);
 	}

-	bzero(&smlock, sizeof (smlock));
-	mutex_init(&smlock, NULL, MUTEX_DEFAULT, NULL);
-
-	space_map_create(&smsync, sm->sm_start, sm->sm_size, sm->sm_shift,
-	    &smlock);
-
-	mutex_enter(&smlock);
+	dmu_tx_commit(tx);

 	mutex_enter(&vd->vdev_dtl_lock);
-	space_map_walk(sm, space_map_add, &smsync);
+	space_map_update(vd->vdev_dtl_sm);
 	mutex_exit(&vd->vdev_dtl_lock);
-
-	space_map_truncate(smo, mos, tx);
-	space_map_sync(&smsync, SM_ALLOC, smo, mos, tx);
-	space_map_vacate(&smsync, NULL, NULL);
-
-	space_map_destroy(&smsync);
-
-	mutex_exit(&smlock);
-	mutex_destroy(&smlock);
-
-	VERIFY(0 == dmu_bonus_hold(mos, smo->smo_object, FTAG, &db));
-	dmu_buf_will_dirty(db, tx);
-	ASSERT3U(db->db_size, >=, sizeof (*smo));
-	bcopy(smo, db->db_data, sizeof (*smo));
-	dmu_buf_rele(db, FTAG);
-
-	dmu_tx_commit(tx);
 }

 /*
@ -2078,7 +2101,7 @@ vdev_resilver_needed(vdev_t *vd, uint64_t *minp, uint64_t *maxp)

 	if (vd->vdev_children == 0) {
 		mutex_enter(&vd->vdev_dtl_lock);
-		if (vd->vdev_dtl[DTL_MISSING].sm_space != 0 &&
+		if (range_tree_space(vd->vdev_dtl[DTL_MISSING]) != 0 &&
 		    vdev_writeable(vd)) {

 			thismin = vdev_dtl_min(vd);
@ -2183,29 +2206,25 @@ vdev_remove(vdev_t *vd, uint64_t txg)

 	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

-	if (vd->vdev_dtl_smo.smo_object) {
-		ASSERT0(vd->vdev_dtl_smo.smo_alloc);
-		(void) dmu_object_free(mos, vd->vdev_dtl_smo.smo_object, tx);
-		vd->vdev_dtl_smo.smo_object = 0;
-	}
-
 	if (vd->vdev_ms != NULL) {
 		for (int m = 0; m < vd->vdev_ms_count; m++) {
 			metaslab_t *msp = vd->vdev_ms[m];

-			if (msp == NULL || msp->ms_smo.smo_object == 0)
+			if (msp == NULL || msp->ms_sm == NULL)
 				continue;

-			ASSERT0(msp->ms_smo.smo_alloc);
-			(void) dmu_object_free(mos, msp->ms_smo.smo_object, tx);
-			msp->ms_smo.smo_object = 0;
+			mutex_enter(&msp->ms_lock);
+			VERIFY0(space_map_allocated(msp->ms_sm));
+			space_map_free(msp->ms_sm, tx);
+			space_map_close(msp->ms_sm);
+			msp->ms_sm = NULL;
+			mutex_exit(&msp->ms_lock);
 		}
 	}

 	if (vd->vdev_ms_array) {
 		(void) dmu_object_free(mos, vd->vdev_ms_array, tx);
 		vd->vdev_ms_array = 0;
-		vd->vdev_ms_shift = 0;
 	}
 	dmu_tx_commit(tx);
 }
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
@ -289,9 +289,10 @@ vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
 			    vd->vdev_removing);
 	}

-	if (vd->vdev_dtl_smo.smo_object != 0)
+	if (vd->vdev_dtl_sm != NULL) {
 		fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL,
-		    vd->vdev_dtl_smo.smo_object);
+		    space_map_object(vd->vdev_dtl_sm));
+	}

 	if (vd->vdev_crtxg)
 		fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg);
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfeature.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfeature.c
@ -361,36 +361,46 @@ spa_feature_enable(spa_t *spa, zfeature_info_t *feature, dmu_tx_t *tx)
 	    spa->spa_feat_desc_obj, feature, FEATURE_ACTION_ENABLE, tx));
 }

-/*
- * If the specified feature has not yet been enabled, this function returns
- * ENOTSUP; otherwise, this function increments the feature's refcount (or
- * returns EOVERFLOW if the refcount cannot be incremented). This function must
- * be called from syncing context.
- */
 void
 spa_feature_incr(spa_t *spa, zfeature_info_t *feature, dmu_tx_t *tx)
 {
+	ASSERT(dmu_tx_is_syncing(tx));
 	ASSERT3U(spa_version(spa), >=, SPA_VERSION_FEATURES);
 	VERIFY3U(0, ==, feature_do_action(spa->spa_meta_objset,
 	    spa->spa_feat_for_read_obj, spa->spa_feat_for_write_obj,
 	    spa->spa_feat_desc_obj, feature, FEATURE_ACTION_INCR, tx));
 }

-/*
- * If the specified feature has not yet been enabled, this function returns
- * ENOTSUP; otherwise, this function decrements the feature's refcount (or
- * returns EOVERFLOW if the refcount is already 0). This function must
- * be called from syncing context.
- */
 void
 spa_feature_decr(spa_t *spa, zfeature_info_t *feature, dmu_tx_t *tx)
 {
+	ASSERT(dmu_tx_is_syncing(tx));
 	ASSERT3U(spa_version(spa), >=, SPA_VERSION_FEATURES);
 	VERIFY3U(0, ==, feature_do_action(spa->spa_meta_objset,
 	    spa->spa_feat_for_read_obj, spa->spa_feat_for_write_obj,
 	    spa->spa_feat_desc_obj, feature, FEATURE_ACTION_DECR, tx));
 }

+/*
+ * This interface is for debugging only. Normal consumers should use
+ * spa_feature_is_enabled/spa_feature_is_active.
+ */
+int
+spa_feature_get_refcount(spa_t *spa, zfeature_info_t *feature)
+{
+	int err;
+	uint64_t refcount;
+
+	if (spa_version(spa) < SPA_VERSION_FEATURES)
+		return (B_FALSE);
+
+	err = feature_get_refcount(spa->spa_meta_objset,
+	    spa->spa_feat_for_read_obj, spa->spa_feat_for_write_obj,
+	    feature, &refcount);
+	ASSERT(err == 0 || err == ENOTSUP);
+	return (err == 0 ? refcount : 0);
+}
+
 boolean_t
 spa_feature_is_enabled(spa_t *spa, zfeature_info_t *feature)
 {
--- a/sys/conf/files
+++ b/sys/conf/files
@ -205,6 +205,7 @@ cddl/contrib/opensolaris/uts/common/fs/zfs/gzip.c			optional zfs compile-with "$
 cddl/contrib/opensolaris/uts/common/fs/zfs/lz4.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/lzjb.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/metaslab.c			optional zfs compile-with "${ZFS_C}"
+cddl/contrib/opensolaris/uts/common/fs/zfs/range_tree.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/refcount.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/rrwlock.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/sa.c				optional zfs compile-with "${ZFS_C}"
@ -215,6 +216,7 @@ cddl/contrib/opensolaris/uts/common/fs/zfs/spa_errlog.c			optional zfs compile-w
 cddl/contrib/opensolaris/uts/common/fs/zfs/spa_history.c		optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/spa_misc.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/space_map.c			optional zfs compile-with "${ZFS_C}"
+cddl/contrib/opensolaris/uts/common/fs/zfs/space_reftree.c		optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/trim_map.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/txg.c			optional zfs compile-with "${ZFS_C}"
 cddl/contrib/opensolaris/uts/common/fs/zfs/uberblock.c			optional zfs compile-with "${ZFS_C}"