upstream/illumos/illumos-gate: comparison usr/src/uts/common/fs/zfs/spa.c

equal deleted inserted replaced

-:733714f4dc24
+:2889e2596bd6
 #include <sys/callb.h>
 #include <sys/systeminfo.h>
 #include <sys/spa_boot.h>
 #include <sys/zfs_ioctl.h>
 #include <sys/dsl_scan.h>
+#include <sys/zfeature.h>
 #ifdef	_KERNEL
 #include <sys/bootprops.h>
 #include <sys/callb.h>
 #include <sys/cpupart.h>
 	{ ZTI_FIX(100),	ZTI_NULL,	ZTI_ONE,	ZTI_NULL },
 	{ ZTI_ONE,	ZTI_NULL,	ZTI_ONE,	ZTI_NULL },
 	{ ZTI_ONE,	ZTI_NULL,	ZTI_ONE,	ZTI_NULL },
 };
+static dsl_syncfunc_t spa_sync_version;
 static dsl_syncfunc_t spa_sync_props;
 static boolean_t spa_has_active_shared_spare(spa_t *spa);
 static int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config,
 spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
 char **ereport);
 */
 static void
 spa_prop_get_config(spa_t *spa, nvlist_t **nvp)
 {
 	vdev_t *rvd = spa->spa_root_vdev;
+	dsl_pool_t *pool = spa->spa_dsl_pool;
 	uint64_t size;
 	uint64_t alloc;
 	uint64_t space;
 	uint64_t cap, version;
 	zprop_source_t src = ZPROP_SRC_NONE;
 		if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION))
 			src = ZPROP_SRC_DEFAULT;
 		else
 			src = ZPROP_SRC_LOCAL;
 		spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src);
+	}
+	if (pool != NULL) {
+		dsl_dir_t *freedir = pool->dp_free_dir;
+		/*
+		 * The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
+		 * when opening pools before this version freedir will be NULL.
+		 */
+		if (freedir != NULL) {
+			spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
+			    freedir->dd_phys->dd_used_bytes, src);
+		} else {
+			spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
+			    NULL, 0, src);
+		}
 	}
 	spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
 	if (spa->spa_comment != NULL) {
 spa_prop_validate(spa_t *spa, nvlist_t *props)
 {
 	nvpair_t *elem;
 	int error = 0, reset_bootfs = 0;
 	uint64_t objnum;
+	boolean_t has_feature = B_FALSE;
 	elem = NULL;
 	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
-		zpool_prop_t prop;
-		char *propname, *strval;
 		uint64_t intval;
-		objset_t *os;
+		char *strval, *slash, *check, *fname;
-		char *slash, *check;
+		const char *propname = nvpair_name(elem);
+		zpool_prop_t prop = zpool_name_to_prop(propname);
-		propname = nvpair_name(elem);
-		if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL)
-			return (EINVAL);
 		switch (prop) {
+		case ZPROP_INVAL:
+			if (!zpool_prop_feature(propname)) {
+				error = EINVAL;
+				break;
+			}
+			/*
+			 * Sanitize the input.
+			 */
+			if (nvpair_type(elem) != DATA_TYPE_UINT64) {
+				error = EINVAL;
+				break;
+			}
+			if (nvpair_value_uint64(elem, &intval) != 0) {
+				error = EINVAL;
+				break;
+			}
+			if (intval != 0) {
+				error = EINVAL;
+				break;
+			}
+			fname = strchr(propname, '@') + 1;
+			if (zfeature_lookup_name(fname, NULL) != 0) {
+				error = EINVAL;
+				break;
+			}
+			has_feature = B_TRUE;
+			break;
 		case ZPOOL_PROP_VERSION:
 			error = nvpair_value_uint64(elem, &intval);
 			if (!error &&
-			    (intval < spa_version(spa) || intval > SPA_VERSION))
+			    (intval < spa_version(spa) ||
+			    intval > SPA_VERSION_BEFORE_FEATURES ||
+			    has_feature))
 				error = EINVAL;
 			break;
 		case ZPOOL_PROP_DELEGATION:
 		case ZPOOL_PROP_AUTOREPLACE:
 			reset_bootfs = 1;
 			error = nvpair_value_string(elem, &strval);
 			if (!error) {
+				objset_t *os;
 				uint64_t compress;
 				if (strval == NULL || strval[0] == '\0') {
 					objnum = zpool_prop_default_numeric(
 					    ZPOOL_PROP_BOOTFS);
 int
 spa_prop_set(spa_t *spa, nvlist_t *nvp)
 {
 	int error;
-	nvpair_t *elem;
+	nvpair_t *elem = NULL;
 	boolean_t need_sync = B_FALSE;
-	zpool_prop_t prop;
 	if ((error = spa_prop_validate(spa, nvp)) != 0)
 		return (error);
-	elem = NULL;
 	while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
-		if ((prop = zpool_name_to_prop(
+		zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));
-		    nvpair_name(elem))) == ZPROP_INVAL)
-			return (EINVAL);
 		if (prop == ZPOOL_PROP_CACHEFILE ||
 		    prop == ZPOOL_PROP_ALTROOT ||
 		    prop == ZPOOL_PROP_READONLY)
 			continue;
+		if (prop == ZPOOL_PROP_VERSION || prop == ZPROP_INVAL) {
+			uint64_t ver;
+			if (prop == ZPOOL_PROP_VERSION) {
+				VERIFY(nvpair_value_uint64(elem, &ver) == 0);
+			} else {
+				ASSERT(zpool_prop_feature(nvpair_name(elem)));
+				ver = SPA_VERSION_FEATURES;
+				need_sync = B_TRUE;
+			}
+			/* Save time if the version is already set. */
+			if (ver == spa_version(spa))
+				continue;
+			/*
+			 * In addition to the pool directory object, we might
+			 * create the pool properties object, the features for
+			 * read object, the features for write object, or the
+			 * feature descriptions object.
+			 */
+			error = dsl_sync_task_do(spa_get_dsl(spa), NULL,
+			    spa_sync_version, spa, &ver, 6);
+			if (error)
+				return (error);
+			continue;
+		}
 		need_sync = B_TRUE;
 		break;
 	}
-	if (need_sync)
+	if (need_sync) {
 		return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props,
-		    spa, nvp, 3));
+		    spa, nvp, 6));
-	else
+	}
-		return (0);
+	return (0);
 }
 /*
 * If the bootfs property value is dsobj, clear it.
 */
 	spa_load_error_t *sle = zio->io_private;
 	dmu_object_type_t type = BP_GET_TYPE(bp);
 	int error = zio->io_error;
 	if (error) {
-		if ((BP_GET_LEVEL(bp) != 0 || dmu_ot[type].ot_metadata) &&
+		if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) &&
 		    type != DMU_OT_INTENT_LOG)
 			atomic_add_64(&sle->sle_meta_count, 1);
 		else
 			atomic_add_64(&sle->sle_data_count, 1);
 	}
 		    &nvl) == 0) {
 			VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting,
 			    KM_SLEEP) == 0);
 		}
+		nvlist_free(spa->spa_load_info);
+		spa->spa_load_info = fnvlist_alloc();
 		gethrestime(&spa->spa_loaded_ts);
 		error = spa_load_impl(spa, pool_guid, config, state, type,
 		    mosconfig, &ereport);
 	}
 spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
 char **ereport)
 {
 	int error = 0;
 	nvlist_t *nvroot = NULL;
+	nvlist_t *label;
 	vdev_t *rvd;
 	uberblock_t *ub = &spa->spa_uberblock;
 	uint64_t children, config_cache_txg = spa->spa_config_txg;
 	int orig_mode = spa->spa_mode;
 	int parse;
 	uint64_t obj;
+	boolean_t missing_feat_write = B_FALSE;
 	/*
 	 * If this is an untrusted config, access the pool in read-only mode.
 	 * This prevents things like resilvering recently removed devices.
 	 */
 	}
 	/*
 	 * Find the best uberblock.
 	 */
-	vdev_uberblock_load(NULL, rvd, ub);
+	vdev_uberblock_load(rvd, ub, &label);
 	/*
 	 * If we weren't able to find a single valid uberblock, return failure.
 	 */
-	if (ub->ub_txg == 0)
+	if (ub->ub_txg == 0) {
+		nvlist_free(label);
 		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
+	}
-	/*
-	 * If the pool is newer than the code, we can't open it.
+	/*
-	 */
+	 * If the pool has an unsupported version we can't open it.
-	if (ub->ub_version > SPA_VERSION)
+	 */
+	if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) {
+		nvlist_free(label);
 		return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP));
+	}
+	if (ub->ub_version >= SPA_VERSION_FEATURES) {
+		nvlist_t *features;
+		/*
+		 * If we weren't able to find what's necessary for reading the
+		 * MOS in the label, return failure.
+		 */
+		if (label == NULL || nvlist_lookup_nvlist(label,
+		    ZPOOL_CONFIG_FEATURES_FOR_READ, &features) != 0) {
+			nvlist_free(label);
+			return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
+			    ENXIO));
+		}
+		/*
+		 * Update our in-core representation with the definitive values
+		 * from the label.
+		 */
+		nvlist_free(spa->spa_label_features);
+		VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0);
+	}
+	nvlist_free(label);
+	/*
+	 * Look through entries in the label nvlist's features_for_read. If
+	 * there is a feature listed there which we don't understand then we
+	 * cannot open a pool.
+	 */
+	if (ub->ub_version >= SPA_VERSION_FEATURES) {
+		nvlist_t *unsup_feat;
+		VERIFY(nvlist_alloc(&unsup_feat, NV_UNIQUE_NAME, KM_SLEEP) ==
+		    0);
+		for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features,
+		    NULL); nvp != NULL;
+		    nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) {
+			if (!zfeature_is_supported(nvpair_name(nvp))) {
+				VERIFY(nvlist_add_string(unsup_feat,
+				    nvpair_name(nvp), "") == 0);
+			}
+		}
+		if (!nvlist_empty(unsup_feat)) {
+			VERIFY(nvlist_add_nvlist(spa->spa_load_info,
+			    ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat) == 0);
+			nvlist_free(unsup_feat);
+			return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
+			    ENOTSUP));
+		}
+		nvlist_free(unsup_feat);
+	}
 	/*
 	 * If the vdev guid sum doesn't match the uberblock, we have an
 	 * incomplete configuration.  We first check to see if the pool
 	 * is aware of the complete config (i.e ZPOOL_CONFIG_VDEV_CHILDREN).
 	spa->spa_first_txg = spa->spa_last_ubsync_txg ?
 	    spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1;
 	spa->spa_claim_max_txg = spa->spa_first_txg;
 	spa->spa_prev_software_version = ub->ub_software_version;
-	error = dsl_pool_open(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
+	error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
 	if (error)
 		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
 	spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;
 	if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object) != 0)
+		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+	if (spa_version(spa) >= SPA_VERSION_FEATURES) {
+		boolean_t missing_feat_read = B_FALSE;
+		nvlist_t *unsup_feat;
+		if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
+		    &spa->spa_feat_for_read_obj) != 0) {
+			return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+		}
+		if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE,
+		    &spa->spa_feat_for_write_obj) != 0) {
+			return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+		}
+		if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS,
+		    &spa->spa_feat_desc_obj) != 0) {
+			return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+		}
+		VERIFY(nvlist_alloc(&unsup_feat, NV_UNIQUE_NAME, KM_SLEEP) ==
+		    0);
+		if (!feature_is_supported(spa->spa_meta_objset,
+		    spa->spa_feat_for_read_obj, spa->spa_feat_desc_obj,
+		    unsup_feat))
+			missing_feat_read = B_TRUE;
+		if (spa_writeable(spa) || state == SPA_LOAD_TRYIMPORT) {
+			if (!feature_is_supported(spa->spa_meta_objset,
+			    spa->spa_feat_for_write_obj, spa->spa_feat_desc_obj,
+			    unsup_feat))
+				missing_feat_write = B_TRUE;
+		}
+		if (!nvlist_empty(unsup_feat)) {
+			VERIFY(nvlist_add_nvlist(spa->spa_load_info,
+			    ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat) == 0);
+		}
+		nvlist_free(unsup_feat);
+		if (!missing_feat_read) {
+			fnvlist_add_boolean(spa->spa_load_info,
+			    ZPOOL_CONFIG_CAN_RDONLY);
+		}
+		/*
+		 * If the state is SPA_LOAD_TRYIMPORT, our objective is
+		 * twofold: to determine whether the pool is available for
+		 * import in read-write mode and (if it is not) whether the
+		 * pool is available for import in read-only mode. If the pool
+		 * is available for import in read-write mode, it is displayed
+		 * as available in userland; if it is not available for import
+		 * in read-only mode, it is displayed as unavailable in
+		 * userland. If the pool is available for import in read-only
+		 * mode but not read-write mode, it is displayed as unavailable
+		 * in userland with a special note that the pool is actually
+		 * available for open in read-only mode.
+		 *
+		 * As a result, if the state is SPA_LOAD_TRYIMPORT and we are
+		 * missing a feature for write, we must first determine whether
+		 * the pool can be opened read-only before returning to
+		 * userland in order to know whether to display the
+		 * abovementioned note.
+		 */
+		if (missing_feat_read || (missing_feat_write &&
+		    spa_writeable(spa))) {
+			return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
+			    ENOTSUP));
+		}
+	}
+	spa->spa_is_initializing = B_TRUE;
+	error = dsl_pool_open(spa->spa_dsl_pool);
+	spa->spa_is_initializing = B_FALSE;
+	if (error != 0)
 		return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
 	if (!mosconfig) {
 		uint64_t hostid;
 		nvlist_t *policy = NULL, *nvconfig;
 			    ENXIO));
 		}
 		nvlist_free(nvconfig);
 		/*
-		 * Now that we've validate the config, check the state of the
+		 * Now that we've validated the config, check the state of the
 		 * root vdev.  If it can't be opened, it indicates one or
 		 * more toplevel vdevs are faulted.
 		 */
 		if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
 			return (ENXIO);
 		if (spa_check_logs(spa)) {
 			*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
 			return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO));
 		}
+	}
+	if (missing_feat_write) {
+		ASSERT(state == SPA_LOAD_TRYIMPORT);
+		/*
+		 * At this point, we know that we can open the pool in
+		 * read-only mode but not read-write mode. We now have enough
+		 * information and can return to userland.
+		 */
+		return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP));
 	}
 	/*
 	 * We've successfully opened the pool, verify that we're ready
 	 * to start pushing transactions.
 	spa_async_suspend(spa);
 	return (spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig));
 }
+/*
+* If spa_load() fails this function will try loading prior txg's. If
+* 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool
+* will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this
+* function will not rewind the pool and will return the same error as
+* spa_load().
+*/
 static int
 spa_load_best(spa_t *spa, spa_load_state_t state, int mosconfig,
 uint64_t max_request, int rewind_flags)
 {
+	nvlist_t *loadinfo = NULL;
 	nvlist_t *config = NULL;
 	int load_error, rewind_error;
 	uint64_t safe_rewind_txg;
 	uint64_t min_txg;
 	if (rewind_flags & ZPOOL_NEVER_REWIND) {
 		nvlist_free(config);
 		return (load_error);
 	}
-	/* Price of rolling back is discarding txgs, including log */
+	if (state == SPA_LOAD_RECOVER) {
-	if (state == SPA_LOAD_RECOVER)
+		/* Price of rolling back is discarding txgs, including log */
 		spa_set_log_state(spa, SPA_LOG_CLEAR);
+	} else {
+		/*
+		 * If we aren't rolling back save the load info from our first
+		 * import attempt so that we can restore it after attempting
+		 * to rewind.
+		 */
+		loadinfo = spa->spa_load_info;
+		spa->spa_load_info = fnvlist_alloc();
+	}
 	spa->spa_load_max_txg = spa->spa_last_ubsync_txg;
 	safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE;
 	min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ?
 	    TXG_INITIAL : safe_rewind_txg;
 	spa->spa_load_max_txg = UINT64_MAX;
 	if (config && (rewind_error || state != SPA_LOAD_RECOVER))
 		spa_config_set(spa, config);
-	return (state == SPA_LOAD_RECOVER ? rewind_error : load_error);
+	if (state == SPA_LOAD_RECOVER) {
+		ASSERT3P(loadinfo, ==, NULL);
+		return (rewind_error);
+	} else {
+		/* Store the rewind info as part of the initial load info */
+		fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO,
+		    spa->spa_load_info);
+		/* Restore the initial load info */
+		fnvlist_free(spa->spa_load_info);
+		spa->spa_load_info = loadinfo;
+		return (load_error);
+	}
 }
 /*
 * Pool Open/Import
 *
 			vdev_get_stats(vd, vs);
 		}
 	}
 }
+static void
+spa_add_feature_stats(spa_t *spa, nvlist_t *config)
+{
+	nvlist_t *features;
+	zap_cursor_t zc;
+	zap_attribute_t za;
+	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
+	VERIFY(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+	if (spa->spa_feat_for_read_obj != 0) {
+		for (zap_cursor_init(&zc, spa->spa_meta_objset,
+		    spa->spa_feat_for_read_obj);
+		    zap_cursor_retrieve(&zc, &za) == 0;
+		    zap_cursor_advance(&zc)) {
+			ASSERT(za.za_integer_length == sizeof (uint64_t) &&
+			    za.za_num_integers == 1);
+			VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name,
+			    za.za_first_integer));
+		}
+		zap_cursor_fini(&zc);
+	}
+	if (spa->spa_feat_for_write_obj != 0) {
+		for (zap_cursor_init(&zc, spa->spa_meta_objset,
+		    spa->spa_feat_for_write_obj);
+		    zap_cursor_retrieve(&zc, &za) == 0;
+		    zap_cursor_advance(&zc)) {
+			ASSERT(za.za_integer_length == sizeof (uint64_t) &&
+			    za.za_num_integers == 1);
+			VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name,
+			    za.za_first_integer));
+		}
+		zap_cursor_fini(&zc);
+	}
+	VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
+	    features) == 0);
+	nvlist_free(features);
+}
 int
-spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen)
+spa_get_stats(const char *name, nvlist_t **config,
+char *altroot, size_t buflen)
 {
 	int error;
 	spa_t *spa;
 	*config = NULL;
 				    ZPOOL_CONFIG_SUSPENDED,
 				    spa->spa_failmode) == 0);
 			spa_add_spares(spa, *config);
 			spa_add_l2cache(spa, *config);
+			spa_add_feature_stats(spa, *config);
 		}
 	}
 	/*
 	 * We want to get the alternate root even for faulted pools, so we cheat
 	int error = 0;
 	uint64_t txg = TXG_INITIAL;
 	nvlist_t **spares, **l2cache;
 	uint_t nspares, nl2cache;
 	uint64_t version, obj;
+	boolean_t has_features;
 	/*
 	 * If this pool already exists, return failure.
 	 */
 	mutex_enter(&spa_namespace_lock);
 		spa_remove(spa);
 		mutex_exit(&spa_namespace_lock);
 		return (error);
 	}
-	if (nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION),
+	has_features = B_FALSE;
-	    &version) != 0)
+	for (nvpair_t *elem = nvlist_next_nvpair(props, NULL);
+	    elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
+		if (zpool_prop_feature(nvpair_name(elem)))
+			has_features = B_TRUE;
+	}
+	if (has_features || nvlist_lookup_uint64(props,
+	    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) {
 		version = SPA_VERSION;
-	ASSERT(version <= SPA_VERSION);
+	}
+	ASSERT(SPA_VERSION_IS_SUPPORTED(version));
 	spa->spa_first_txg = txg;
 	spa->spa_uberblock.ub_txg = txg - 1;
 	spa->spa_uberblock.ub_version = version;
 	spa->spa_ubsync = spa->spa_uberblock;
 		spa_load_l2cache(spa);
 		spa_config_exit(spa, SCL_ALL, FTAG);
 		spa->spa_l2cache.sav_sync = B_TRUE;
 	}
+	spa->spa_is_initializing = B_TRUE;
 	spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg);
 	spa->spa_meta_objset = dp->dp_meta_objset;
+	spa->spa_is_initializing = B_FALSE;
 	/*
 	 * Create DDTs (dedup tables).
 	 */
 	ddt_create(spa);
 	if (zap_add(spa->spa_meta_objset,
 	    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
 	    sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) {
 		cmn_err(CE_PANIC, "failed to add pool config");
 	}
+	if (spa_version(spa) >= SPA_VERSION_FEATURES)
+		spa_feature_create_zap_objects(spa, tx);
 	if (zap_add(spa->spa_meta_objset,
 	    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
 	    sizeof (uint64_t), 1, &version, tx) != 0) {
 		cmn_err(CE_PANIC, "failed to add pool version");
 			config = spa_generate_rootconf(devpath, devid, &guid);
 		}
 	}
 #endif
 	if (config == NULL) {
-		cmn_err(CE_NOTE, "Can not read the pool label from '%s'",
+		cmn_err(CE_NOTE, "Cannot read the pool label from '%s'",
 		    devpath);
 		return (EIO);
 	}
 	VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
 		    poolname) == 0);
 		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
 		    state) == 0);
 		VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
 		    spa->spa_uberblock.ub_timestamp) == 0);
+		VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
+		    spa->spa_load_info) == 0);
 		/*
 		 * If the bootfs property exists on this pool then we
 		 * copy it out so that external consumers can tell which
 		 * pools are bootable.
 	/*
 	 * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
 	 * information.  This avoids the dbuf_will_dirty() path and
 	 * saves us a pre-read to get data we don't actually care about.
 	 */
-	bufsize = P2ROUNDUP(nvsize, SPA_CONFIG_BLOCKSIZE);
+	bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
 	packed = kmem_alloc(bufsize, KM_SLEEP);
 	VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
 	    KM_SLEEP) == 0);
 	bzero(packed + nvsize, bufsize - nvsize);
 	spa->spa_config_syncing = config;
 	spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
 }
+static void
+spa_sync_version(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+	spa_t *spa = arg1;
+	uint64_t version = *(uint64_t *)arg2;
+	/*
+	 * Setting the version is special cased when first creating the pool.
+	 */
+	ASSERT(tx->tx_txg != TXG_INITIAL);
+	ASSERT(version <= SPA_VERSION);
+	ASSERT(version >= spa_version(spa));
+	spa->spa_uberblock.ub_version = version;
+	vdev_config_dirty(spa->spa_root_vdev);
+}
 /*
 * Set zpool properties.
 */
 static void
 spa_sync_props(void *arg1, void *arg2, dmu_tx_t *tx)
 {
 	spa_t *spa = arg1;
 	objset_t *mos = spa->spa_meta_objset;
 	nvlist_t *nvp = arg2;
-	nvpair_t *elem;
+	nvpair_t *elem = NULL;
-	uint64_t intval;
-	char *strval;
-	zpool_prop_t prop;
-	const char *propname;
-	zprop_type_t proptype;
 	mutex_enter(&spa->spa_props_lock);
-	elem = NULL;
 	while ((elem = nvlist_next_nvpair(nvp, elem))) {
+		uint64_t intval;
+		char *strval, *fname;
+		zpool_prop_t prop;
+		const char *propname;
+		zprop_type_t proptype;
+		zfeature_info_t *feature;
 		switch (prop = zpool_name_to_prop(nvpair_name(elem))) {
+		case ZPROP_INVAL:
+			/*
+			 * We checked this earlier in spa_prop_validate().
+			 */
+			ASSERT(zpool_prop_feature(nvpair_name(elem)));
+			fname = strchr(nvpair_name(elem), '@') + 1;
+			VERIFY3U(0, ==, zfeature_lookup_name(fname, &feature));
+			spa_feature_enable(spa, feature, tx);
+			break;
 		case ZPOOL_PROP_VERSION:
+			VERIFY(nvpair_value_uint64(elem, &intval) == 0);
 			/*
-			 * Only set version for non-zpool-creation cases
+			 * The version is synced seperatly before other
-			 * (set/import). spa_create() needs special care
+			 * properties and should be correct by now.
-			 * for version setting.
 			 */
-			if (tx->tx_txg != TXG_INITIAL) {
+			ASSERT3U(spa_version(spa), >=, intval);
-				VERIFY(nvpair_value_uint64(elem,
-				    &intval) == 0);
-				ASSERT(intval <= SPA_VERSION);
-				ASSERT(intval >= spa_version(spa));
-				spa->spa_uberblock.ub_version = intval;
-				vdev_config_dirty(spa->spa_root_vdev);
-			}
 			break;
 		case ZPOOL_PROP_ALTROOT:
 			/*
 			 * 'altroot' is a non-persistent property. It should
 		default:
 			/*
 			 * Set pool property values in the poolprops mos object.
 			 */
 			if (spa->spa_pool_props_object == 0) {
-				VERIFY((spa->spa_pool_props_object =
+				spa->spa_pool_props_object =
-				    zap_create(mos, DMU_OT_POOL_PROPS,
+				    zap_create_link(mos, DMU_OT_POOL_PROPS,
-				    DMU_OT_NONE, 0, tx)) > 0);
-				VERIFY(zap_update(mos,
 				    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
-				    8, 1, &spa->spa_pool_props_object, tx)
+				    tx);
-				    == 0);
 			}
 			/* normalize the property name */
 			propname = zpool_prop_to_name(prop);
 			proptype = zpool_prop_get_type(prop);
 		dsl_pool_upgrade_dir_clones(dp, tx);
 		/* Keeping the freedir open increases spa_minref */
 		spa->spa_minref += 3;
 	}
+	if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES &&
+	    spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
+		spa_feature_create_zap_objects(spa, tx);
+	}
 }
 /*
 * Sync the specified transaction group.  New blocks may be dirtied as
 * part of the process, so we iterate until it converges.

changeset 13700	2889e2596bd6
parent 13605	b5c2b5db80d6
child 13743	95aba6e49b9f