MFV r299440: 6736 ZFS per-vdev ZAPs

Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Don Brady <don.brady@intel.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>
Approved by: Richard Lowe <richlowe@richlowe.net>
Author: Joe Stein <joe.stein@delphix.com>

openzfs/openzfs@215198a6ad15cf4832370e2f19247abeb36b951a

10 files changed, 389 insertions, 69 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
index 2e039ed..173c949 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
@@ -1665,6 +1665,19 @@ spa_check_removed(vdev_t *vd)
 	}
 }
 
+static void
+spa_config_valid_zaps(vdev_t *vd, vdev_t *mvd)
+{
+	ASSERT3U(vd->vdev_children, ==, mvd->vdev_children);
+
+	vd->vdev_top_zap = mvd->vdev_top_zap;
+	vd->vdev_leaf_zap = mvd->vdev_leaf_zap;
+
+	for (uint64_t i = 0; i < vd->vdev_children; i++) {
+		spa_config_valid_zaps(vd->vdev_child[i], mvd->vdev_child[i]);
+	}
+}
+
 /*
  * Validate the current config against the MOS config
  */
@@ -1768,16 +1781,25 @@ spa_config_valid(spa_t *spa, nvlist_t *config)
 			spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
 
 			vdev_reopen(rvd);
-		} else if (mtvd->vdev_islog) {
+		} else {
+			if (mtvd->vdev_islog) {
+				/*
+				 * Load the slog device's state from the MOS
+				 * config since it's possible that the label
+				 * does not contain the most up-to-date
+				 * information.
+				 */
+				vdev_load_log_state(tvd, mtvd);
+				vdev_reopen(tvd);
+			}
+
 			/*
-			 * Load the slog device's state from the MOS config
-			 * since it's possible that the label does not
-			 * contain the most up-to-date information.
+			 * Per-vdev ZAP info is stored exclusively in the MOS.
 			 */
-			vdev_load_log_state(tvd, mtvd);
-			vdev_reopen(tvd);
+			spa_config_valid_zaps(tvd, mtvd);
 		}
 	}
+
 	vdev_free(mrvd);
 	spa_config_exit(spa, SCL_ALL, FTAG);
 
@@ -2210,6 +2232,34 @@ spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type,
 }
 
 /*
+ * Count the number of per-vdev ZAPs associated with all of the vdevs in the
+ * vdev tree rooted in the given vd, and ensure that each ZAP is present in the
+ * spa's per-vdev ZAP list.
+ */
+static uint64_t
+vdev_count_verify_zaps(vdev_t *vd)
+{
+	spa_t *spa = vd->vdev_spa;
+	uint64_t total = 0;
+	if (vd->vdev_top_zap != 0) {
+		total++;
+		ASSERT0(zap_lookup_int(spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps, vd->vdev_top_zap));
+	}
+	if (vd->vdev_leaf_zap != 0) {
+		total++;
+		ASSERT0(zap_lookup_int(spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps, vd->vdev_leaf_zap));
+	}
+
+	for (uint64_t i = 0; i < vd->vdev_children; i++) {
+		total += vdev_count_verify_zaps(vd->vdev_child[i]);
+	}
+
+	return (total);
+}
+
+/*
  * Load an existing storage pool, using the pool's builtin spa_config as a
  * source of configuration information.
  */
@@ -2638,6 +2688,39 @@ spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config,
 		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
 
 	/*
+	 * Load the per-vdev ZAP map. If we have an older pool, this will not
+	 * be present; in this case, defer its creation to a later time to
+	 * avoid dirtying the MOS this early / out of sync context. See
+	 * spa_sync_config_object.
+	 */
+
+	/* The sentinel is only available in the MOS config. */
+	nvlist_t *mos_config;
+	if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0)
+		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+	error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
+	    &spa->spa_all_vdev_zaps);
+
+	if (error != ENOENT && error != 0) {
+		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+	} else if (error == 0 && !nvlist_exists(mos_config,
+	    ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
+		/*
+		 * An older version of ZFS overwrote the sentinel value, so
+		 * we have orphaned per-vdev ZAPs in the MOS. Defer their
+		 * destruction to later; see spa_sync_config_object.
+		 */
+		spa->spa_avz_action = AVZ_ACTION_DESTROY;
+		/*
+		 * We're assuming that no vdevs have had their ZAPs created
+		 * before this. Better be sure of it.
+		 */
+		ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
+	}
+	nvlist_free(mos_config);
+
+	/*
 	 * If we're assembling the pool from the split-off vdevs of
 	 * an existing pool, we don't want to attach the spares & cache
 	 * devices.
@@ -5324,6 +5407,16 @@ spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
 		    vml[c]->vdev_top->vdev_asize) == 0);
 		VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
 		    vml[c]->vdev_top->vdev_ashift) == 0);
+
+		/* transfer per-vdev ZAPs */
+		ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0);
+		VERIFY0(nvlist_add_uint64(child[c],
+		    ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap));
+
+		ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0);
+		VERIFY0(nvlist_add_uint64(child[c],
+		    ZPOOL_CONFIG_VDEV_TOP_ZAP,
+		    vml[c]->vdev_parent->vdev_top_zap));
 	}
 
 	if (error != 0) {
@@ -5365,11 +5458,13 @@ spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
 	    spa->spa_config_txg) == 0);
 	VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
 	    spa_generate_guid(NULL)) == 0);
+	VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
 	(void) nvlist_lookup_string(props,
 	    zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
 
 	/* add the new pool to the namespace */
 	newspa = spa_add(newname, config, altroot);
+	newspa->spa_avz_action = AVZ_ACTION_REBUILD;
 	newspa->spa_config_txg = spa->spa_config_txg;
 	spa_set_log_state(newspa, SPA_LOG_CLEAR);
 
@@ -5434,9 +5529,11 @@ spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
 			if (error == 0)
 				spa_history_log_internal(spa, "detach", tx,
 				    "vdev=%s", vml[c]->vdev_path);
+
 			vdev_free(vml[c]);
 		}
 	}
+	spa->spa_avz_action = AVZ_ACTION_REBUILD;
 	vdev_config_dirty(spa->spa_root_vdev);
 	spa->spa_config_splitting = NULL;
 	nvlist_free(nvl);
@@ -6321,16 +6418,118 @@ spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx,
 	sav->sav_sync = B_FALSE;
 }
 
+/*
+ * Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t.
+ * The all-vdev ZAP must be empty.
+ */
+static void
+spa_avz_build(vdev_t *vd, uint64_t avz, dmu_tx_t *tx)
+{
+	spa_t *spa = vd->vdev_spa;
+	if (vd->vdev_top_zap != 0) {
+		VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
+		    vd->vdev_top_zap, tx));
+	}
+	if (vd->vdev_leaf_zap != 0) {
+		VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
+		    vd->vdev_leaf_zap, tx));
+	}
+	for (uint64_t i = 0; i < vd->vdev_children; i++) {
+		spa_avz_build(vd->vdev_child[i], avz, tx);
+	}
+}
+
 static void
 spa_sync_config_object(spa_t *spa, dmu_tx_t *tx)
 {
 	nvlist_t *config;
 
-	if (list_is_empty(&spa->spa_config_dirty_list))
+	/*
+	 * If the pool is being imported from a pre-per-vdev-ZAP version of ZFS,
+	 * its config may not be dirty but we still need to build per-vdev ZAPs.
+	 * Similarly, if the pool is being assembled (e.g. after a split), we
+	 * need to rebuild the AVZ although the config may not be dirty.
+	 */
+	if (list_is_empty(&spa->spa_config_dirty_list) &&
+	    spa->spa_avz_action == AVZ_ACTION_NONE)
 		return;
 
 	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
 
+	ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE ||
+	    spa->spa_all_vdev_zaps != 0);
+
+	if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
+		/* Make and build the new AVZ */
+		uint64_t new_avz = zap_create(spa->spa_meta_objset,
+		    DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
+		spa_avz_build(spa->spa_root_vdev, new_avz, tx);
+
+		/* Diff old AVZ with new one */
+		zap_cursor_t zc;
+		zap_attribute_t za;
+
+		for (zap_cursor_init(&zc, spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps);
+		    zap_cursor_retrieve(&zc, &za) == 0;
+		    zap_cursor_advance(&zc)) {
+			uint64_t vdzap = za.za_first_integer;
+			if (zap_lookup_int(spa->spa_meta_objset, new_avz,
+			    vdzap) == ENOENT) {
+				/*
+				 * ZAP is listed in old AVZ but not in new one;
+				 * destroy it
+				 */
+				VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap,
+				    tx));
+			}
+		}
+
+		zap_cursor_fini(&zc);
+
+		/* Destroy the old AVZ */
+		VERIFY0(zap_destroy(spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps, tx));
+
+		/* Replace the old AVZ in the dir obj with the new one */
+		VERIFY0(zap_update(spa->spa_meta_objset,
+		    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP,
+		    sizeof (new_avz), 1, &new_avz, tx));
+
+		spa->spa_all_vdev_zaps = new_avz;
+	} else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) {
+		zap_cursor_t zc;
+		zap_attribute_t za;
+
+		/* Walk through the AVZ and destroy all listed ZAPs */
+		for (zap_cursor_init(&zc, spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps);
+		    zap_cursor_retrieve(&zc, &za) == 0;
+		    zap_cursor_advance(&zc)) {
+			uint64_t zap = za.za_first_integer;
+			VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx));
+		}
+
+		zap_cursor_fini(&zc);
+
+		/* Destroy and unlink the AVZ itself */
+		VERIFY0(zap_destroy(spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps, tx));
+		VERIFY0(zap_remove(spa->spa_meta_objset,
+		    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx));
+		spa->spa_all_vdev_zaps = 0;
+	}
+
+	if (spa->spa_all_vdev_zaps == 0) {
+		spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset,
+		    DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT,
+		    DMU_POOL_VDEV_ZAP_MAP, tx);
+	}
+	spa->spa_avz_action = AVZ_ACTION_NONE;
+
+	/* Create ZAPs for vdevs that don't have them. */
+	vdev_construct_zaps(spa->spa_root_vdev, tx);
+
 	config = spa_config_generate(spa, spa->spa_root_vdev,
 	    dmu_tx_get_txg(tx), B_FALSE);
 
@@ -6738,6 +6937,21 @@ spa_sync(spa_t *spa, uint64_t txg)
 
 	} while (dmu_objset_is_dirty(mos, txg));
 
+	if (!list_is_empty(&spa->spa_config_dirty_list)) {
+		/*
+		 * Make sure that the number of ZAPs for all the vdevs matches
+		 * the number of ZAPs in the per-vdev ZAP list. This only gets
+		 * called if the config is dirty; otherwise there may be
+		 * outstanding AVZ operations that weren't completed in
+		 * spa_sync_config_object.
+		 */
+		uint64_t all_vdev_zap_entry_count;
+		ASSERT0(zap_count(spa->spa_meta_objset,
+		    spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count));
+		ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==,
+		    all_vdev_zap_entry_count);
+	}
+
 	/*
 	 * Rewrite the vdev configuration (which includes the uberblock)
 	 * to commit the transaction group.
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_config.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_config.c
index a5c17a5..882c552 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_config.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa_config.c
@@ -22,7 +22,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -123,7 +123,7 @@ spa_config_load(void)
 		if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
 			continue;
 
-		VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
+		child = fnvpair_value_nvlist(nvpair);
 
 		if (spa_lookup(nvpair_name(nvpair)) != NULL)
 			continue;
@@ -181,14 +181,9 @@ spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
 	/*
 	 * Pack the configuration into a buffer.
 	 */
-	VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
-
-	buf = kmem_alloc(buflen, KM_SLEEP);
+	buf = fnvlist_pack(nvl, &buflen);
 	temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
 
-	VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
-	    KM_SLEEP) == 0);
-
 	/*
 	 * Write the configuration to disk.  We need to do the traditional
 	 * 'write to temporary file, sync, move over original' to make sure we
@@ -209,7 +204,7 @@ spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
 
 	(void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
 
-	kmem_free(buf, buflen);
+	fnvlist_pack_free(buf, buflen);
 	kmem_free(temp, MAXPATHLEN);
 	return (err);
 }
@@ -276,11 +271,10 @@ spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
 			}
 
 			if (nvl == NULL)
-				VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
-				    KM_SLEEP) == 0);
+				nvl = fnvlist_alloc();
 
-			VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
-			    spa->spa_config) == 0);
+			fnvlist_add_nvlist(nvl, spa->spa_name,
+			    spa->spa_config);
 			mutex_exit(&spa->spa_props_lock);
 
 			if (nvlist_lookup_nvlist(nvl, spa->spa_name, &nvroot) == 0)
@@ -345,15 +339,15 @@ spa_all_configs(uint64_t *generation)
 	if (*generation == spa_config_generation)
 		return (NULL);
 
-	VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+	pools = fnvlist_alloc();
 
 	mutex_enter(&spa_namespace_lock);
 	while ((spa = spa_next(spa)) != NULL) {
 		if (INGLOBALZONE(curthread) ||
 		    zone_dataset_visible(spa_name(spa), NULL)) {
 			mutex_enter(&spa->spa_props_lock);
-			VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
-			    spa->spa_config) == 0);
+			fnvlist_add_nvlist(pools, spa_name(spa),
+			    spa->spa_config);
 			mutex_exit(&spa->spa_props_lock);
 		}
 	}
@@ -402,21 +396,17 @@ spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 	if (txg == -1ULL)
 		txg = spa->spa_config_txg;
 
-	VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-
-	VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
-	    spa_version(spa)) == 0);
-	VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
-	    spa_name(spa)) == 0);
-	VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
-	    spa_state(spa)) == 0);
-	VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
-	    txg) == 0);
-	VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
-	    spa_guid(spa)) == 0);
-	VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
-	    ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
+	config = fnvlist_alloc();
 
+	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
+	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, spa_name(spa));
+	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, spa_state(spa));
+	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, txg);
+	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_guid(spa));
+	if (spa->spa_comment != NULL) {
+		fnvlist_add_string(config, ZPOOL_CONFIG_COMMENT,
+		    spa->spa_comment);
+	}
 
 #ifdef	_KERNEL
 	hostid = zone_get_hostid(NULL);
@@ -428,23 +418,24 @@ spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 	(void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
 #endif	/* _KERNEL */
 	if (hostid != 0) {
-		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
-		    hostid) == 0);
+		fnvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, hostid);
 	}
-	VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
-	    utsname.nodename) == 0);
+	fnvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, utsname.nodename);
 
+	int config_gen_flags = 0;
 	if (vd != rvd) {
-		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
-		    vd->vdev_top->vdev_guid) == 0);
-		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
-		    vd->vdev_guid) == 0);
-		if (vd->vdev_isspare)
-			VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
-			    1ULL) == 0);
-		if (vd->vdev_islog)
-			VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
-			    1ULL) == 0);
+		fnvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
+		    vd->vdev_top->vdev_guid);
+		fnvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
+		    vd->vdev_guid);
+		if (vd->vdev_isspare) {
+			fnvlist_add_uint64(config,
+			    ZPOOL_CONFIG_IS_SPARE, 1ULL);
+		}
+		if (vd->vdev_islog) {
+			fnvlist_add_uint64(config,
+			    ZPOOL_CONFIG_IS_LOG, 1ULL);
+		}
 		vd = vd->vdev_top;		/* label contains top config */
 	} else {
 		/*
@@ -452,8 +443,12 @@ spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 		 * in the mos config, and not in the vdev labels
 		 */
 		if (spa->spa_config_splitting != NULL)
-			VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
-			    spa->spa_config_splitting) == 0);
+			fnvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
+			    spa->spa_config_splitting);
+		fnvlist_add_boolean(config,
+		    ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS);
+
+		config_gen_flags |= VDEV_CONFIG_MOS;
 	}
 
 	/*
@@ -468,19 +463,19 @@ spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 	if (spa->spa_config_splitting != NULL &&
 	    nvlist_lookup_uint64(spa->spa_config_splitting,
 	    ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
-		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
-		    split_guid) == 0);
+		fnvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
+		    split_guid);
 	}
 
-	nvroot = vdev_config_generate(spa, vd, getstats, 0);
-	VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
+	nvroot = vdev_config_generate(spa, vd, getstats, config_gen_flags);
+	fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot);
 	nvlist_free(nvroot);
 
 	/*
 	 * Store what's necessary for reading the MOS in the label.
 	 */
-	VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
-	    spa->spa_label_features) == 0);
+	fnvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
+	    spa->spa_label_features);
 
 	if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
 		ddt_histogram_t *ddh;
@@ -489,23 +484,23 @@ spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 
 		ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
 		ddt_get_dedup_histogram(spa, ddh);
-		VERIFY(nvlist_add_uint64_array(config,
+		fnvlist_add_uint64_array(config,
 		    ZPOOL_CONFIG_DDT_HISTOGRAM,
-		    (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
+		    (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t));
 		kmem_free(ddh, sizeof (ddt_histogram_t));
 
 		ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
 		ddt_get_dedup_object_stats(spa, ddo);
-		VERIFY(nvlist_add_uint64_array(config,
+		fnvlist_add_uint64_array(config,
 		    ZPOOL_CONFIG_DDT_OBJ_STATS,
-		    (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
+		    (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t));
 		kmem_free(ddo, sizeof (ddt_object_t));
 
 		dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
 		ddt_get_dedup_stats(spa, dds);
-		VERIFY(nvlist_add_uint64_array(config,
+		fnvlist_add_uint64_array(config,
 		    ZPOOL_CONFIG_DDT_STATS,
-		    (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
+		    (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t));
 		kmem_free(dds, sizeof (ddt_stat_t));
 	}
 
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
index 3bbe18f..8a8f29c 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
@@ -323,6 +323,7 @@ typedef struct dmu_buf {
 #define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
 #define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"
 #define	DMU_POOL_CHECKSUM_SALT		"org.illumos:checksum_salt"
+#define	DMU_POOL_VDEV_ZAP_MAP		"com.delphix:vdev_zap_map"
 
 /*
  * Allocate an object from this objset.  The range of object numbers
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h
index a93ba83..0a25b6c 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/spa_impl.h
@@ -117,6 +117,12 @@ typedef struct spa_taskqs {
 	taskq_t **stqs_taskq;
 } spa_taskqs_t;
 
+typedef enum spa_all_vdev_zap_action {
+	AVZ_ACTION_NONE = 0,
+	AVZ_ACTION_DESTROY,	/* Destroy all per-vdev ZAPs and the AVZ. */
+	AVZ_ACTION_REBUILD	/* Populate the new AVZ, see spa_avz_rebuild */
+} spa_avz_action_t;
+
 struct spa {
 	/*
 	 * Fields protected by spa_namespace_lock.
@@ -264,6 +270,9 @@ struct spa {
 	uint64_t	spa_deadman_calls;	/* number of deadman calls */
 	hrtime_t	spa_sync_starttime;	/* starting time fo spa_sync */
 	uint64_t	spa_deadman_synctime;	/* deadman expiration timer */
+	uint64_t	spa_all_vdev_zaps;	/* ZAP of per-vd ZAP obj #s */
+	spa_avz_action_t	spa_avz_action;	/* destroy/rebuild AVZ? */
+
 #ifdef illumos
 	/*
 	 * spa_iokstat_lock protects spa_iokstat and
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h
index 6312f03..1344260 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h
@@ -72,6 +72,10 @@ extern void vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
 extern boolean_t vdev_dtl_required(vdev_t *vd);
 extern boolean_t vdev_resilver_needed(vdev_t *vd,
     uint64_t *minp, uint64_t *maxp);
+extern void vdev_destroy_unlink_zap(vdev_t *vd, uint64_t zapobj,
+    dmu_tx_t *tx);
+extern uint64_t vdev_create_link_zap(vdev_t *vd, dmu_tx_t *tx);
+extern void vdev_construct_zaps(vdev_t *vd, dmu_tx_t *tx);
 
 extern void vdev_hold(vdev_t *);
 extern void vdev_rele(vdev_t *);
@@ -135,7 +139,8 @@ extern void vdev_state_clean(vdev_t *vd);
 typedef enum vdev_config_flag {
 	VDEV_CONFIG_SPARE = 1 << 0,
 	VDEV_CONFIG_L2CACHE = 1 << 1,
-	VDEV_CONFIG_REMOVING = 1 << 2
+	VDEV_CONFIG_REMOVING = 1 << 2,
+	VDEV_CONFIG_MOS = 1 << 3
 } vdev_config_flag_t;
 
 extern void vdev_top_config_generate(spa_t *spa, nvlist_t *config);
diff --git 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
index b0d242a..0c1f5ac 100644
--- 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
@@ -191,6 +191,7 @@ struct vdev {
 	uint64_t	vdev_islog;	/* is an intent log device	*/
 	uint64_t	vdev_removing;	/* device is being removed?	*/
 	boolean_t	vdev_ishole;	/* is a hole in the namespace 	*/
+	uint64_t	vdev_top_zap;
 
 	/*
 	 * Leaf vdev state.
@@ -234,6 +235,7 @@ struct vdev {
 	uint16_t	vdev_rotation_rate; /* rotational rate of the media */
 #define	VDEV_RATE_UNKNOWN	0
 #define	VDEV_RATE_NON_ROTATING	1
+	uint64_t	vdev_leaf_zap;
 
 	/*
 	 * For DTrace to work in userland (libzpool) context, these fields must
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
index 28ff6c9..9b08e35 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
@@ -609,6 +609,10 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 		    &vd->vdev_asize);
 		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVING,
 		    &vd->vdev_removing);
+		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
+		    &vd->vdev_top_zap);
+	} else {
+		ASSERT0(vd->vdev_top_zap);
 	}
 
 	if (parent && !parent->vdev_parent && alloctype != VDEV_ALLOC_ATTACH) {
@@ -620,9 +624,18 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 		    spa_log_class(spa) : spa_normal_class(spa), vd);
 	}
 
+	if (vd->vdev_ops->vdev_op_leaf &&
+	    (alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_SPLIT)) {
+		(void) nvlist_lookup_uint64(nv,
+		    ZPOOL_CONFIG_VDEV_LEAF_ZAP, &vd->vdev_leaf_zap);
+	} else {
+		ASSERT0(vd->vdev_leaf_zap);
+	}
+
 	/*
 	 * 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_ROOTPOOL)) {
@@ -783,10 +796,12 @@ vdev_top_transfer(vdev_t *svd, vdev_t *tvd)
 	tvd->vdev_ms_array = svd->vdev_ms_array;
 	tvd->vdev_ms_shift = svd->vdev_ms_shift;
 	tvd->vdev_ms_count = svd->vdev_ms_count;
+	tvd->vdev_top_zap = svd->vdev_top_zap;
 
 	svd->vdev_ms_array = 0;
 	svd->vdev_ms_shift = 0;
 	svd->vdev_ms_count = 0;
+	svd->vdev_top_zap = 0;
 
 	if (tvd->vdev_mg)
 		ASSERT3P(tvd->vdev_mg, ==, svd->vdev_mg);
@@ -2079,6 +2094,49 @@ vdev_dtl_load(vdev_t *vd)
 }
 
 void
+vdev_destroy_unlink_zap(vdev_t *vd, uint64_t zapobj, dmu_tx_t *tx)
+{
+	spa_t *spa = vd->vdev_spa;
+
+	VERIFY0(zap_destroy(spa->spa_meta_objset, zapobj, tx));
+	VERIFY0(zap_remove_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
+	    zapobj, tx));
+}
+
+uint64_t
+vdev_create_link_zap(vdev_t *vd, dmu_tx_t *tx)
+{
+	spa_t *spa = vd->vdev_spa;
+	uint64_t zap = zap_create(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA,
+	    DMU_OT_NONE, 0, tx);
+
+	ASSERT(zap != 0);
+	VERIFY0(zap_add_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
+	    zap, tx));
+
+	return (zap);
+}
+
+void
+vdev_construct_zaps(vdev_t *vd, dmu_tx_t *tx)
+{
+	if (vd->vdev_ops != &vdev_hole_ops &&
+	    vd->vdev_ops != &vdev_missing_ops &&
+	    vd->vdev_ops != &vdev_root_ops &&
+	    !vd->vdev_top->vdev_removing) {
+		if (vd->vdev_ops->vdev_op_leaf && vd->vdev_leaf_zap == 0) {
+			vd->vdev_leaf_zap = vdev_create_link_zap(vd, tx);
+		}
+		if (vd == vd->vdev_top && vd->vdev_top_zap == 0) {
+			vd->vdev_top_zap = vdev_create_link_zap(vd, tx);
+		}
+	}
+	for (uint64_t i = 0; i < vd->vdev_children; i++) {
+		vdev_construct_zaps(vd->vdev_child[i], tx);
+	}
+}
+
+void
 vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 {
 	spa_t *spa = vd->vdev_spa;
@@ -2100,6 +2158,18 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 		space_map_close(vd->vdev_dtl_sm);
 		vd->vdev_dtl_sm = NULL;
 		mutex_exit(&vd->vdev_dtl_lock);
+
+		/*
+		 * We only destroy the leaf ZAP for detached leaves or for
+		 * removed log devices. Removed data devices handle leaf ZAP
+		 * cleanup later, once cancellation is no longer possible.
+		 */
+		if (vd->vdev_leaf_zap != 0 && (vd->vdev_detached ||
+		    vd->vdev_top->vdev_islog)) {
+			vdev_destroy_unlink_zap(vd, vd->vdev_leaf_zap, tx);
+			vd->vdev_leaf_zap = 0;
+		}
+
 		dmu_tx_commit(tx);
 		return;
 	}
@@ -2303,6 +2373,8 @@ vdev_remove(vdev_t *vd, uint64_t txg)
 	dmu_tx_t *tx;
 
 	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
+	ASSERT(vd == vd->vdev_top);
+	ASSERT3U(txg, ==, spa_syncing_txg(spa));
 
 	if (vd->vdev_ms != NULL) {
 		metaslab_group_t *mg = vd->vdev_mg;
@@ -2344,6 +2416,11 @@ vdev_remove(vdev_t *vd, uint64_t txg)
 		(void) dmu_object_free(mos, vd->vdev_ms_array, tx);
 		vd->vdev_ms_array = 0;
 	}
+
+	if (vd->vdev_islog && vd->vdev_top_zap != 0) {
+		vdev_destroy_unlink_zap(vd, vd->vdev_top_zap, tx);
+		vd->vdev_top_zap = 0;
+	}
 	dmu_tx_commit(tx);
 }
 
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
index 77191e7..7901115 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_label.c
@@ -297,6 +297,20 @@ vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
 	if (vd->vdev_crtxg)
 		fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg);
 
+	if (flags & VDEV_CONFIG_MOS) {
+		if (vd->vdev_leaf_zap != 0) {
+			ASSERT(vd->vdev_ops->vdev_op_leaf);
+			fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_LEAF_ZAP,
+			    vd->vdev_leaf_zap);
+		}
+
+		if (vd->vdev_top_zap != 0) {
+			ASSERT(vd == vd->vdev_top);
+			fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
+			    vd->vdev_top_zap);
+		}
+	}
+
 	if (getstats) {
 		vdev_stat_t vs;
 		pool_scan_stat_t ps;
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zap.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zap.c
index 5fc2621..717368c 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zap.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zap.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  */
 
@@ -962,8 +962,8 @@ zap_create_link(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
 	uint64_t new_obj;
 
 	VERIFY((new_obj = zap_create(os, ot, DMU_OT_NONE, 0, tx)) > 0);
-	VERIFY(zap_add(os, parent_obj, name, sizeof (uint64_t), 1, &new_obj,
-	    tx) == 0);
+	VERIFY0(zap_add(os, parent_obj, name, sizeof (uint64_t), 1, &new_obj,
+	    tx));
 
 	return (new_obj);
 }
diff --git a/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h b/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
index de94fe7..ef24fc3 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
@@ -565,6 +565,9 @@ typedef struct zpool_rewind_policy {
 #define	ZPOOL_CONFIG_CAN_RDONLY		"can_rdonly"	/* not stored on disk */
 #define	ZPOOL_CONFIG_FEATURES_FOR_READ	"features_for_read"
 #define	ZPOOL_CONFIG_FEATURE_STATS	"feature_stats"	/* not stored on disk */
+#define	ZPOOL_CONFIG_VDEV_TOP_ZAP	"com.delphix:vdev_zap_top"
+#define	ZPOOL_CONFIG_VDEV_LEAF_ZAP	"com.delphix:vdev_zap_leaf"
+#define	ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS	"com.delphix:has_per_vdev_zaps"
 /*
  * The persistent vdev state is stored as separate values rather than a single
  * 'vdev_state' entry.  This is because a device can be in multiple states, such