1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5ea8dc4b6Seschrock * Common Development and Distribution License (the "License"). 6ea8dc4b6Seschrock * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 2199653d4eSeschrock 22fa9e4066Sahrens /* 23*379c004dSEric Schrock * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 24fa9e4066Sahrens * Use is subject to license terms. 25fa9e4066Sahrens */ 26fa9e4066Sahrens 27fa9e4066Sahrens /* 28fa9e4066Sahrens * This file contains all the routines used when modifying on-disk SPA state. 29fa9e4066Sahrens * This includes opening, importing, destroying, exporting a pool, and syncing a 30fa9e4066Sahrens * pool. 31fa9e4066Sahrens */ 32fa9e4066Sahrens 33fa9e4066Sahrens #include <sys/zfs_context.h> 34ea8dc4b6Seschrock #include <sys/fm/fs/zfs.h> 35fa9e4066Sahrens #include <sys/spa_impl.h> 36fa9e4066Sahrens #include <sys/zio.h> 37fa9e4066Sahrens #include <sys/zio_checksum.h> 38fa9e4066Sahrens #include <sys/zio_compress.h> 39fa9e4066Sahrens #include <sys/dmu.h> 40fa9e4066Sahrens #include <sys/dmu_tx.h> 41fa9e4066Sahrens #include <sys/zap.h> 42fa9e4066Sahrens #include <sys/zil.h> 43fa9e4066Sahrens #include <sys/vdev_impl.h> 44fa9e4066Sahrens #include <sys/metaslab.h> 45fa9e4066Sahrens #include <sys/uberblock_impl.h> 46fa9e4066Sahrens #include <sys/txg.h> 47fa9e4066Sahrens #include <sys/avl.h> 48fa9e4066Sahrens #include <sys/dmu_traverse.h> 49b1b8ab34Slling #include <sys/dmu_objset.h> 50fa9e4066Sahrens #include <sys/unique.h> 51fa9e4066Sahrens #include <sys/dsl_pool.h> 52b1b8ab34Slling #include <sys/dsl_dataset.h> 53fa9e4066Sahrens #include <sys/dsl_dir.h> 54fa9e4066Sahrens #include <sys/dsl_prop.h> 55b1b8ab34Slling #include <sys/dsl_synctask.h> 56fa9e4066Sahrens #include <sys/fs/zfs.h> 57fa94a07fSbrendan #include <sys/arc.h> 58fa9e4066Sahrens #include <sys/callb.h> 5995173954Sek #include <sys/systeminfo.h> 6095173954Sek #include <sys/sunddi.h> 61e7cbe64fSgw #include <sys/spa_boot.h> 62fa9e4066Sahrens 63990b4856Slling #include "zfs_prop.h" 64b7b97454Sperrin #include "zfs_comutil.h" 65990b4856Slling 66e14bb325SJeff Bonwick int zio_taskq_threads[ZIO_TYPES][ZIO_TASKQ_TYPES] = { 67e14bb325SJeff Bonwick /* ISSUE INTR */ 68e14bb325SJeff Bonwick { 1, 1 }, /* ZIO_TYPE_NULL */ 69e14bb325SJeff Bonwick { 1, 8 }, /* ZIO_TYPE_READ */ 70e14bb325SJeff Bonwick { 8, 1 }, /* ZIO_TYPE_WRITE */ 71e14bb325SJeff Bonwick { 1, 1 }, /* ZIO_TYPE_FREE */ 72e14bb325SJeff Bonwick { 1, 1 }, /* ZIO_TYPE_CLAIM */ 73e14bb325SJeff Bonwick { 1, 1 }, /* ZIO_TYPE_IOCTL */ 74e14bb325SJeff Bonwick }; 75416e0cd8Sek 76990b4856Slling static void spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx); 7789a89ebfSlling static boolean_t spa_has_active_shared_spare(spa_t *spa); 78990b4856Slling 79990b4856Slling /* 80990b4856Slling * ========================================================================== 81990b4856Slling * SPA properties routines 82990b4856Slling * ========================================================================== 83990b4856Slling */ 84990b4856Slling 85990b4856Slling /* 86990b4856Slling * Add a (source=src, propname=propval) list to an nvlist. 87990b4856Slling */ 889d82f4f6Slling static void 89990b4856Slling spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval, 90990b4856Slling uint64_t intval, zprop_source_t src) 91990b4856Slling { 92990b4856Slling const char *propname = zpool_prop_to_name(prop); 93990b4856Slling nvlist_t *propval; 94990b4856Slling 959d82f4f6Slling VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); 969d82f4f6Slling VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0); 97990b4856Slling 989d82f4f6Slling if (strval != NULL) 999d82f4f6Slling VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0); 1009d82f4f6Slling else 1019d82f4f6Slling VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, intval) == 0); 102990b4856Slling 1039d82f4f6Slling VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0); 104990b4856Slling nvlist_free(propval); 105990b4856Slling } 106990b4856Slling 107990b4856Slling /* 108990b4856Slling * Get property values from the spa configuration. 109990b4856Slling */ 1109d82f4f6Slling static void 111990b4856Slling spa_prop_get_config(spa_t *spa, nvlist_t **nvp) 112990b4856Slling { 113*379c004dSEric Schrock uint64_t size; 114*379c004dSEric Schrock uint64_t used; 115990b4856Slling uint64_t cap, version; 116990b4856Slling zprop_source_t src = ZPROP_SRC_NONE; 117c5904d13Seschrock spa_config_dirent_t *dp; 118990b4856Slling 119e14bb325SJeff Bonwick ASSERT(MUTEX_HELD(&spa->spa_props_lock)); 120e14bb325SJeff Bonwick 121*379c004dSEric Schrock if (spa->spa_root_vdev != NULL) { 122*379c004dSEric Schrock size = spa_get_space(spa); 123*379c004dSEric Schrock used = spa_get_alloc(spa); 124*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src); 125*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src); 126*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_USED, NULL, used, src); 127*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_AVAILABLE, NULL, 128*379c004dSEric Schrock size - used, src); 129*379c004dSEric Schrock 130*379c004dSEric Schrock cap = (size == 0) ? 0 : (used * 100 / size); 131*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src); 132*379c004dSEric Schrock 133*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL, 134*379c004dSEric Schrock spa->spa_root_vdev->vdev_state, src); 135*379c004dSEric Schrock 136*379c004dSEric Schrock version = spa_version(spa); 137*379c004dSEric Schrock if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) 138*379c004dSEric Schrock src = ZPROP_SRC_DEFAULT; 139*379c004dSEric Schrock else 140*379c004dSEric Schrock src = ZPROP_SRC_LOCAL; 141*379c004dSEric Schrock spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src); 142*379c004dSEric Schrock } 143990b4856Slling 1449d82f4f6Slling spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src); 145990b4856Slling 1469d82f4f6Slling if (spa->spa_root != NULL) 1479d82f4f6Slling spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root, 1489d82f4f6Slling 0, ZPROP_SRC_LOCAL); 149990b4856Slling 150c5904d13Seschrock if ((dp = list_head(&spa->spa_config_list)) != NULL) { 151c5904d13Seschrock if (dp->scd_path == NULL) { 1529d82f4f6Slling spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, 153c5904d13Seschrock "none", 0, ZPROP_SRC_LOCAL); 154c5904d13Seschrock } else if (strcmp(dp->scd_path, spa_config_path) != 0) { 1559d82f4f6Slling spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE, 156c5904d13Seschrock dp->scd_path, 0, ZPROP_SRC_LOCAL); 1572f8aaab3Seschrock } 1582f8aaab3Seschrock } 159990b4856Slling } 160990b4856Slling 161990b4856Slling /* 162990b4856Slling * Get zpool property values. 163990b4856Slling */ 164990b4856Slling int 165990b4856Slling spa_prop_get(spa_t *spa, nvlist_t **nvp) 166990b4856Slling { 167990b4856Slling zap_cursor_t zc; 168990b4856Slling zap_attribute_t za; 169990b4856Slling objset_t *mos = spa->spa_meta_objset; 170990b4856Slling int err; 171990b4856Slling 1729d82f4f6Slling VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); 173990b4856Slling 174e14bb325SJeff Bonwick mutex_enter(&spa->spa_props_lock); 175e14bb325SJeff Bonwick 176990b4856Slling /* 177990b4856Slling * Get properties from the spa config. 178990b4856Slling */ 1799d82f4f6Slling spa_prop_get_config(spa, nvp); 180990b4856Slling 181990b4856Slling /* If no pool property object, no more prop to get. */ 182990b4856Slling if (spa->spa_pool_props_object == 0) { 183990b4856Slling mutex_exit(&spa->spa_props_lock); 184990b4856Slling return (0); 185990b4856Slling } 186990b4856Slling 187990b4856Slling /* 188990b4856Slling * Get properties from the MOS pool property object. 189990b4856Slling */ 190990b4856Slling for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object); 191990b4856Slling (err = zap_cursor_retrieve(&zc, &za)) == 0; 192990b4856Slling zap_cursor_advance(&zc)) { 193990b4856Slling uint64_t intval = 0; 194990b4856Slling char *strval = NULL; 195990b4856Slling zprop_source_t src = ZPROP_SRC_DEFAULT; 196990b4856Slling zpool_prop_t prop; 197990b4856Slling 198990b4856Slling if ((prop = zpool_name_to_prop(za.za_name)) == ZPROP_INVAL) 199990b4856Slling continue; 200990b4856Slling 201990b4856Slling switch (za.za_integer_length) { 202990b4856Slling case 8: 203990b4856Slling /* integer property */ 204990b4856Slling if (za.za_first_integer != 205990b4856Slling zpool_prop_default_numeric(prop)) 206990b4856Slling src = ZPROP_SRC_LOCAL; 207990b4856Slling 208990b4856Slling if (prop == ZPOOL_PROP_BOOTFS) { 209990b4856Slling dsl_pool_t *dp; 210990b4856Slling dsl_dataset_t *ds = NULL; 211990b4856Slling 212990b4856Slling dp = spa_get_dsl(spa); 213990b4856Slling rw_enter(&dp->dp_config_rwlock, RW_READER); 214745cd3c5Smaybee if (err = dsl_dataset_hold_obj(dp, 215745cd3c5Smaybee za.za_first_integer, FTAG, &ds)) { 216990b4856Slling rw_exit(&dp->dp_config_rwlock); 217990b4856Slling break; 218990b4856Slling } 219990b4856Slling 220990b4856Slling strval = kmem_alloc( 221990b4856Slling MAXNAMELEN + strlen(MOS_DIR_NAME) + 1, 222990b4856Slling KM_SLEEP); 223990b4856Slling dsl_dataset_name(ds, strval); 224745cd3c5Smaybee dsl_dataset_rele(ds, FTAG); 225990b4856Slling rw_exit(&dp->dp_config_rwlock); 226990b4856Slling } else { 227990b4856Slling strval = NULL; 228990b4856Slling intval = za.za_first_integer; 229990b4856Slling } 230990b4856Slling 2319d82f4f6Slling spa_prop_add_list(*nvp, prop, strval, intval, src); 232990b4856Slling 233990b4856Slling if (strval != NULL) 234990b4856Slling kmem_free(strval, 235990b4856Slling MAXNAMELEN + strlen(MOS_DIR_NAME) + 1); 236990b4856Slling 237990b4856Slling break; 238990b4856Slling 239990b4856Slling case 1: 240990b4856Slling /* string property */ 241990b4856Slling strval = kmem_alloc(za.za_num_integers, KM_SLEEP); 242990b4856Slling err = zap_lookup(mos, spa->spa_pool_props_object, 243990b4856Slling za.za_name, 1, za.za_num_integers, strval); 244990b4856Slling if (err) { 245990b4856Slling kmem_free(strval, za.za_num_integers); 246990b4856Slling break; 247990b4856Slling } 2489d82f4f6Slling spa_prop_add_list(*nvp, prop, strval, 0, src); 249990b4856Slling kmem_free(strval, za.za_num_integers); 250990b4856Slling break; 251990b4856Slling 252990b4856Slling default: 253990b4856Slling break; 254990b4856Slling } 255990b4856Slling } 256990b4856Slling zap_cursor_fini(&zc); 257990b4856Slling mutex_exit(&spa->spa_props_lock); 258990b4856Slling out: 259990b4856Slling if (err && err != ENOENT) { 260990b4856Slling nvlist_free(*nvp); 2619d82f4f6Slling *nvp = NULL; 262990b4856Slling return (err); 263990b4856Slling } 264990b4856Slling 265990b4856Slling return (0); 266990b4856Slling } 267990b4856Slling 268990b4856Slling /* 269990b4856Slling * Validate the given pool properties nvlist and modify the list 270990b4856Slling * for the property values to be set. 271990b4856Slling */ 272990b4856Slling static int 273990b4856Slling spa_prop_validate(spa_t *spa, nvlist_t *props) 274990b4856Slling { 275990b4856Slling nvpair_t *elem; 276990b4856Slling int error = 0, reset_bootfs = 0; 277990b4856Slling uint64_t objnum; 278990b4856Slling 279990b4856Slling elem = NULL; 280990b4856Slling while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 281990b4856Slling zpool_prop_t prop; 282990b4856Slling char *propname, *strval; 283990b4856Slling uint64_t intval; 284990b4856Slling objset_t *os; 2852f8aaab3Seschrock char *slash; 286990b4856Slling 287990b4856Slling propname = nvpair_name(elem); 288990b4856Slling 289990b4856Slling if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) 290990b4856Slling return (EINVAL); 291990b4856Slling 292990b4856Slling switch (prop) { 293990b4856Slling case ZPOOL_PROP_VERSION: 294990b4856Slling error = nvpair_value_uint64(elem, &intval); 295990b4856Slling if (!error && 296990b4856Slling (intval < spa_version(spa) || intval > SPA_VERSION)) 297990b4856Slling error = EINVAL; 298990b4856Slling break; 299990b4856Slling 300990b4856Slling case ZPOOL_PROP_DELEGATION: 301990b4856Slling case ZPOOL_PROP_AUTOREPLACE: 302d5b5bb25SRich Morris case ZPOOL_PROP_LISTSNAPS: 303990b4856Slling error = nvpair_value_uint64(elem, &intval); 304990b4856Slling if (!error && intval > 1) 305990b4856Slling error = EINVAL; 306990b4856Slling break; 307990b4856Slling 308990b4856Slling case ZPOOL_PROP_BOOTFS: 309990b4856Slling if (spa_version(spa) < SPA_VERSION_BOOTFS) { 310990b4856Slling error = ENOTSUP; 311990b4856Slling break; 312990b4856Slling } 313990b4856Slling 314990b4856Slling /* 31515e6edf1Sgw * Make sure the vdev config is bootable 316990b4856Slling */ 31715e6edf1Sgw if (!vdev_is_bootable(spa->spa_root_vdev)) { 318990b4856Slling error = ENOTSUP; 319990b4856Slling break; 320990b4856Slling } 321990b4856Slling 322990b4856Slling reset_bootfs = 1; 323990b4856Slling 324990b4856Slling error = nvpair_value_string(elem, &strval); 325990b4856Slling 326990b4856Slling if (!error) { 32715e6edf1Sgw uint64_t compress; 32815e6edf1Sgw 329990b4856Slling if (strval == NULL || strval[0] == '\0') { 330990b4856Slling objnum = zpool_prop_default_numeric( 331990b4856Slling ZPOOL_PROP_BOOTFS); 332990b4856Slling break; 333990b4856Slling } 334990b4856Slling 335990b4856Slling if (error = dmu_objset_open(strval, DMU_OST_ZFS, 336745cd3c5Smaybee DS_MODE_USER | DS_MODE_READONLY, &os)) 337990b4856Slling break; 33815e6edf1Sgw 33915e6edf1Sgw /* We don't support gzip bootable datasets */ 34015e6edf1Sgw if ((error = dsl_prop_get_integer(strval, 34115e6edf1Sgw zfs_prop_to_name(ZFS_PROP_COMPRESSION), 34215e6edf1Sgw &compress, NULL)) == 0 && 34315e6edf1Sgw !BOOTFS_COMPRESS_VALID(compress)) { 34415e6edf1Sgw error = ENOTSUP; 34515e6edf1Sgw } else { 34615e6edf1Sgw objnum = dmu_objset_id(os); 34715e6edf1Sgw } 348990b4856Slling dmu_objset_close(os); 349990b4856Slling } 350990b4856Slling break; 351e14bb325SJeff Bonwick 3520a4e9518Sgw case ZPOOL_PROP_FAILUREMODE: 3530a4e9518Sgw error = nvpair_value_uint64(elem, &intval); 3540a4e9518Sgw if (!error && (intval < ZIO_FAILURE_MODE_WAIT || 3550a4e9518Sgw intval > ZIO_FAILURE_MODE_PANIC)) 3560a4e9518Sgw error = EINVAL; 3570a4e9518Sgw 3580a4e9518Sgw /* 3590a4e9518Sgw * This is a special case which only occurs when 3600a4e9518Sgw * the pool has completely failed. This allows 3610a4e9518Sgw * the user to change the in-core failmode property 3620a4e9518Sgw * without syncing it out to disk (I/Os might 3630a4e9518Sgw * currently be blocked). We do this by returning 3640a4e9518Sgw * EIO to the caller (spa_prop_set) to trick it 3650a4e9518Sgw * into thinking we encountered a property validation 3660a4e9518Sgw * error. 3670a4e9518Sgw */ 368e14bb325SJeff Bonwick if (!error && spa_suspended(spa)) { 3690a4e9518Sgw spa->spa_failmode = intval; 3700a4e9518Sgw error = EIO; 3710a4e9518Sgw } 3720a4e9518Sgw break; 3732f8aaab3Seschrock 3742f8aaab3Seschrock case ZPOOL_PROP_CACHEFILE: 3752f8aaab3Seschrock if ((error = nvpair_value_string(elem, &strval)) != 0) 3762f8aaab3Seschrock break; 3772f8aaab3Seschrock 3782f8aaab3Seschrock if (strval[0] == '\0') 3792f8aaab3Seschrock break; 3802f8aaab3Seschrock 3812f8aaab3Seschrock if (strcmp(strval, "none") == 0) 3822f8aaab3Seschrock break; 3832f8aaab3Seschrock 3842f8aaab3Seschrock if (strval[0] != '/') { 3852f8aaab3Seschrock error = EINVAL; 3862f8aaab3Seschrock break; 3872f8aaab3Seschrock } 3882f8aaab3Seschrock 3892f8aaab3Seschrock slash = strrchr(strval, '/'); 3902f8aaab3Seschrock ASSERT(slash != NULL); 3912f8aaab3Seschrock 3922f8aaab3Seschrock if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 3932f8aaab3Seschrock strcmp(slash, "/..") == 0) 3942f8aaab3Seschrock error = EINVAL; 3952f8aaab3Seschrock break; 396990b4856Slling } 397990b4856Slling 398990b4856Slling if (error) 399990b4856Slling break; 400990b4856Slling } 401990b4856Slling 402990b4856Slling if (!error && reset_bootfs) { 403990b4856Slling error = nvlist_remove(props, 404990b4856Slling zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING); 405990b4856Slling 406990b4856Slling if (!error) { 407990b4856Slling error = nvlist_add_uint64(props, 408990b4856Slling zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum); 409990b4856Slling } 410990b4856Slling } 411990b4856Slling 412990b4856Slling return (error); 413990b4856Slling } 414990b4856Slling 415*379c004dSEric Schrock void 416*379c004dSEric Schrock spa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync) 417*379c004dSEric Schrock { 418*379c004dSEric Schrock char *cachefile; 419*379c004dSEric Schrock spa_config_dirent_t *dp; 420*379c004dSEric Schrock 421*379c004dSEric Schrock if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), 422*379c004dSEric Schrock &cachefile) != 0) 423*379c004dSEric Schrock return; 424*379c004dSEric Schrock 425*379c004dSEric Schrock dp = kmem_alloc(sizeof (spa_config_dirent_t), 426*379c004dSEric Schrock KM_SLEEP); 427*379c004dSEric Schrock 428*379c004dSEric Schrock if (cachefile[0] == '\0') 429*379c004dSEric Schrock dp->scd_path = spa_strdup(spa_config_path); 430*379c004dSEric Schrock else if (strcmp(cachefile, "none") == 0) 431*379c004dSEric Schrock dp->scd_path = NULL; 432*379c004dSEric Schrock else 433*379c004dSEric Schrock dp->scd_path = spa_strdup(cachefile); 434*379c004dSEric Schrock 435*379c004dSEric Schrock list_insert_head(&spa->spa_config_list, dp); 436*379c004dSEric Schrock if (need_sync) 437*379c004dSEric Schrock spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); 438*379c004dSEric Schrock } 439*379c004dSEric Schrock 440990b4856Slling int 441990b4856Slling spa_prop_set(spa_t *spa, nvlist_t *nvp) 442990b4856Slling { 443990b4856Slling int error; 444*379c004dSEric Schrock nvpair_t *elem; 445*379c004dSEric Schrock boolean_t need_sync = B_FALSE; 446*379c004dSEric Schrock zpool_prop_t prop; 447990b4856Slling 448990b4856Slling if ((error = spa_prop_validate(spa, nvp)) != 0) 449990b4856Slling return (error); 450990b4856Slling 451*379c004dSEric Schrock elem = NULL; 452*379c004dSEric Schrock while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) { 453*379c004dSEric Schrock if ((prop = zpool_name_to_prop( 454*379c004dSEric Schrock nvpair_name(elem))) == ZPROP_INVAL) 455*379c004dSEric Schrock return (EINVAL); 456*379c004dSEric Schrock 457*379c004dSEric Schrock if (prop == ZPOOL_PROP_CACHEFILE || prop == ZPOOL_PROP_ALTROOT) 458*379c004dSEric Schrock continue; 459*379c004dSEric Schrock 460*379c004dSEric Schrock need_sync = B_TRUE; 461*379c004dSEric Schrock break; 462*379c004dSEric Schrock } 463*379c004dSEric Schrock 464*379c004dSEric Schrock if (need_sync) 465*379c004dSEric Schrock return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props, 466*379c004dSEric Schrock spa, nvp, 3)); 467*379c004dSEric Schrock else 468*379c004dSEric Schrock return (0); 469990b4856Slling } 470990b4856Slling 471990b4856Slling /* 472990b4856Slling * If the bootfs property value is dsobj, clear it. 473990b4856Slling */ 474990b4856Slling void 475990b4856Slling spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx) 476990b4856Slling { 477990b4856Slling if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) { 478990b4856Slling VERIFY(zap_remove(spa->spa_meta_objset, 479990b4856Slling spa->spa_pool_props_object, 480990b4856Slling zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0); 481990b4856Slling spa->spa_bootfs = 0; 482990b4856Slling } 483990b4856Slling } 484990b4856Slling 485fa9e4066Sahrens /* 486fa9e4066Sahrens * ========================================================================== 487fa9e4066Sahrens * SPA state manipulation (open/create/destroy/import/export) 488fa9e4066Sahrens * ========================================================================== 489fa9e4066Sahrens */ 490fa9e4066Sahrens 491ea8dc4b6Seschrock static int 492ea8dc4b6Seschrock spa_error_entry_compare(const void *a, const void *b) 493ea8dc4b6Seschrock { 494ea8dc4b6Seschrock spa_error_entry_t *sa = (spa_error_entry_t *)a; 495ea8dc4b6Seschrock spa_error_entry_t *sb = (spa_error_entry_t *)b; 496ea8dc4b6Seschrock int ret; 497ea8dc4b6Seschrock 498ea8dc4b6Seschrock ret = bcmp(&sa->se_bookmark, &sb->se_bookmark, 499ea8dc4b6Seschrock sizeof (zbookmark_t)); 500ea8dc4b6Seschrock 501ea8dc4b6Seschrock if (ret < 0) 502ea8dc4b6Seschrock return (-1); 503ea8dc4b6Seschrock else if (ret > 0) 504ea8dc4b6Seschrock return (1); 505ea8dc4b6Seschrock else 506ea8dc4b6Seschrock return (0); 507ea8dc4b6Seschrock } 508ea8dc4b6Seschrock 509ea8dc4b6Seschrock /* 510ea8dc4b6Seschrock * Utility function which retrieves copies of the current logs and 511ea8dc4b6Seschrock * re-initializes them in the process. 512ea8dc4b6Seschrock */ 513ea8dc4b6Seschrock void 514ea8dc4b6Seschrock spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub) 515ea8dc4b6Seschrock { 516ea8dc4b6Seschrock ASSERT(MUTEX_HELD(&spa->spa_errlist_lock)); 517ea8dc4b6Seschrock 518ea8dc4b6Seschrock bcopy(&spa->spa_errlist_last, last, sizeof (avl_tree_t)); 519ea8dc4b6Seschrock bcopy(&spa->spa_errlist_scrub, scrub, sizeof (avl_tree_t)); 520ea8dc4b6Seschrock 521ea8dc4b6Seschrock avl_create(&spa->spa_errlist_scrub, 522ea8dc4b6Seschrock spa_error_entry_compare, sizeof (spa_error_entry_t), 523ea8dc4b6Seschrock offsetof(spa_error_entry_t, se_avl)); 524ea8dc4b6Seschrock avl_create(&spa->spa_errlist_last, 525ea8dc4b6Seschrock spa_error_entry_compare, sizeof (spa_error_entry_t), 526ea8dc4b6Seschrock offsetof(spa_error_entry_t, se_avl)); 527ea8dc4b6Seschrock } 528ea8dc4b6Seschrock 529fa9e4066Sahrens /* 530fa9e4066Sahrens * Activate an uninitialized pool. 531fa9e4066Sahrens */ 532fa9e4066Sahrens static void 5338ad4d6ddSJeff Bonwick spa_activate(spa_t *spa, int mode) 534fa9e4066Sahrens { 535fa9e4066Sahrens ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 536fa9e4066Sahrens 537fa9e4066Sahrens spa->spa_state = POOL_STATE_ACTIVE; 5388ad4d6ddSJeff Bonwick spa->spa_mode = mode; 539fa9e4066Sahrens 540fa9e4066Sahrens spa->spa_normal_class = metaslab_class_create(); 5418654d025Sperrin spa->spa_log_class = metaslab_class_create(); 542fa9e4066Sahrens 543e14bb325SJeff Bonwick for (int t = 0; t < ZIO_TYPES; t++) { 544e14bb325SJeff Bonwick for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { 545e14bb325SJeff Bonwick spa->spa_zio_taskq[t][q] = taskq_create("spa_zio", 546e14bb325SJeff Bonwick zio_taskq_threads[t][q], maxclsyspri, 50, 547e14bb325SJeff Bonwick INT_MAX, TASKQ_PREPOPULATE); 548e14bb325SJeff Bonwick } 549fa9e4066Sahrens } 550fa9e4066Sahrens 551e14bb325SJeff Bonwick list_create(&spa->spa_config_dirty_list, sizeof (vdev_t), 552e14bb325SJeff Bonwick offsetof(vdev_t, vdev_config_dirty_node)); 553e14bb325SJeff Bonwick list_create(&spa->spa_state_dirty_list, sizeof (vdev_t), 554e14bb325SJeff Bonwick offsetof(vdev_t, vdev_state_dirty_node)); 555fa9e4066Sahrens 556fa9e4066Sahrens txg_list_create(&spa->spa_vdev_txg_list, 557fa9e4066Sahrens offsetof(struct vdev, vdev_txg_node)); 558ea8dc4b6Seschrock 559ea8dc4b6Seschrock avl_create(&spa->spa_errlist_scrub, 560ea8dc4b6Seschrock spa_error_entry_compare, sizeof (spa_error_entry_t), 561ea8dc4b6Seschrock offsetof(spa_error_entry_t, se_avl)); 562ea8dc4b6Seschrock avl_create(&spa->spa_errlist_last, 563ea8dc4b6Seschrock spa_error_entry_compare, sizeof (spa_error_entry_t), 564ea8dc4b6Seschrock offsetof(spa_error_entry_t, se_avl)); 565fa9e4066Sahrens } 566fa9e4066Sahrens 567fa9e4066Sahrens /* 568fa9e4066Sahrens * Opposite of spa_activate(). 569fa9e4066Sahrens */ 570fa9e4066Sahrens static void 571fa9e4066Sahrens spa_deactivate(spa_t *spa) 572fa9e4066Sahrens { 573fa9e4066Sahrens ASSERT(spa->spa_sync_on == B_FALSE); 574fa9e4066Sahrens ASSERT(spa->spa_dsl_pool == NULL); 575fa9e4066Sahrens ASSERT(spa->spa_root_vdev == NULL); 576fa9e4066Sahrens 577fa9e4066Sahrens ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED); 578fa9e4066Sahrens 579fa9e4066Sahrens txg_list_destroy(&spa->spa_vdev_txg_list); 580fa9e4066Sahrens 581e14bb325SJeff Bonwick list_destroy(&spa->spa_config_dirty_list); 582e14bb325SJeff Bonwick list_destroy(&spa->spa_state_dirty_list); 583fa9e4066Sahrens 584e14bb325SJeff Bonwick for (int t = 0; t < ZIO_TYPES; t++) { 585e14bb325SJeff Bonwick for (int q = 0; q < ZIO_TASKQ_TYPES; q++) { 586e14bb325SJeff Bonwick taskq_destroy(spa->spa_zio_taskq[t][q]); 587e14bb325SJeff Bonwick spa->spa_zio_taskq[t][q] = NULL; 588e14bb325SJeff Bonwick } 589fa9e4066Sahrens } 590fa9e4066Sahrens 591fa9e4066Sahrens metaslab_class_destroy(spa->spa_normal_class); 592fa9e4066Sahrens spa->spa_normal_class = NULL; 593fa9e4066Sahrens 5948654d025Sperrin metaslab_class_destroy(spa->spa_log_class); 5958654d025Sperrin spa->spa_log_class = NULL; 5968654d025Sperrin 597ea8dc4b6Seschrock /* 598ea8dc4b6Seschrock * If this was part of an import or the open otherwise failed, we may 599ea8dc4b6Seschrock * still have errors left in the queues. Empty them just in case. 600ea8dc4b6Seschrock */ 601ea8dc4b6Seschrock spa_errlog_drain(spa); 602ea8dc4b6Seschrock 603ea8dc4b6Seschrock avl_destroy(&spa->spa_errlist_scrub); 604ea8dc4b6Seschrock avl_destroy(&spa->spa_errlist_last); 605ea8dc4b6Seschrock 606fa9e4066Sahrens spa->spa_state = POOL_STATE_UNINITIALIZED; 607fa9e4066Sahrens } 608fa9e4066Sahrens 609fa9e4066Sahrens /* 610fa9e4066Sahrens * Verify a pool configuration, and construct the vdev tree appropriately. This 611fa9e4066Sahrens * will create all the necessary vdevs in the appropriate layout, with each vdev 612fa9e4066Sahrens * in the CLOSED state. This will prep the pool before open/creation/import. 613fa9e4066Sahrens * All vdev validation is done by the vdev_alloc() routine. 614fa9e4066Sahrens */ 61599653d4eSeschrock static int 61699653d4eSeschrock spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, 61799653d4eSeschrock uint_t id, int atype) 618fa9e4066Sahrens { 619fa9e4066Sahrens nvlist_t **child; 620fa9e4066Sahrens uint_t c, children; 62199653d4eSeschrock int error; 622fa9e4066Sahrens 62399653d4eSeschrock if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0) 62499653d4eSeschrock return (error); 625fa9e4066Sahrens 62699653d4eSeschrock if ((*vdp)->vdev_ops->vdev_op_leaf) 62799653d4eSeschrock return (0); 628fa9e4066Sahrens 629e14bb325SJeff Bonwick error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 630e14bb325SJeff Bonwick &child, &children); 631e14bb325SJeff Bonwick 632e14bb325SJeff Bonwick if (error == ENOENT) 633e14bb325SJeff Bonwick return (0); 634e14bb325SJeff Bonwick 635e14bb325SJeff Bonwick if (error) { 63699653d4eSeschrock vdev_free(*vdp); 63799653d4eSeschrock *vdp = NULL; 63899653d4eSeschrock return (EINVAL); 639fa9e4066Sahrens } 640fa9e4066Sahrens 641fa9e4066Sahrens for (c = 0; c < children; c++) { 64299653d4eSeschrock vdev_t *vd; 64399653d4eSeschrock if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c, 64499653d4eSeschrock atype)) != 0) { 64599653d4eSeschrock vdev_free(*vdp); 64699653d4eSeschrock *vdp = NULL; 64799653d4eSeschrock return (error); 648fa9e4066Sahrens } 649fa9e4066Sahrens } 650fa9e4066Sahrens 65199653d4eSeschrock ASSERT(*vdp != NULL); 65299653d4eSeschrock 65399653d4eSeschrock return (0); 654fa9e4066Sahrens } 655fa9e4066Sahrens 656fa9e4066Sahrens /* 657fa9e4066Sahrens * Opposite of spa_load(). 658fa9e4066Sahrens */ 659fa9e4066Sahrens static void 660fa9e4066Sahrens spa_unload(spa_t *spa) 661fa9e4066Sahrens { 66299653d4eSeschrock int i; 66399653d4eSeschrock 664e14bb325SJeff Bonwick ASSERT(MUTEX_HELD(&spa_namespace_lock)); 665e14bb325SJeff Bonwick 666ea8dc4b6Seschrock /* 667ea8dc4b6Seschrock * Stop async tasks. 668ea8dc4b6Seschrock */ 669ea8dc4b6Seschrock spa_async_suspend(spa); 670ea8dc4b6Seschrock 671fa9e4066Sahrens /* 672fa9e4066Sahrens * Stop syncing. 673fa9e4066Sahrens */ 674fa9e4066Sahrens if (spa->spa_sync_on) { 675fa9e4066Sahrens txg_sync_stop(spa->spa_dsl_pool); 676fa9e4066Sahrens spa->spa_sync_on = B_FALSE; 677fa9e4066Sahrens } 678fa9e4066Sahrens 679fa9e4066Sahrens /* 680e14bb325SJeff Bonwick * Wait for any outstanding async I/O to complete. 681fa9e4066Sahrens */ 682e14bb325SJeff Bonwick mutex_enter(&spa->spa_async_root_lock); 683e14bb325SJeff Bonwick while (spa->spa_async_root_count != 0) 684e14bb325SJeff Bonwick cv_wait(&spa->spa_async_root_cv, &spa->spa_async_root_lock); 685e14bb325SJeff Bonwick mutex_exit(&spa->spa_async_root_lock); 686fa9e4066Sahrens 687fa9e4066Sahrens /* 688fa9e4066Sahrens * Close the dsl pool. 689fa9e4066Sahrens */ 690fa9e4066Sahrens if (spa->spa_dsl_pool) { 691fa9e4066Sahrens dsl_pool_close(spa->spa_dsl_pool); 692fa9e4066Sahrens spa->spa_dsl_pool = NULL; 693fa9e4066Sahrens } 694fa9e4066Sahrens 6958ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 6968ad4d6ddSJeff Bonwick 6978ad4d6ddSJeff Bonwick /* 6988ad4d6ddSJeff Bonwick * Drop and purge level 2 cache 6998ad4d6ddSJeff Bonwick */ 7008ad4d6ddSJeff Bonwick spa_l2cache_drop(spa); 7018ad4d6ddSJeff Bonwick 702fa9e4066Sahrens /* 703fa9e4066Sahrens * Close all vdevs. 704fa9e4066Sahrens */ 7050e34b6a7Sbonwick if (spa->spa_root_vdev) 706fa9e4066Sahrens vdev_free(spa->spa_root_vdev); 7070e34b6a7Sbonwick ASSERT(spa->spa_root_vdev == NULL); 708ea8dc4b6Seschrock 709fa94a07fSbrendan for (i = 0; i < spa->spa_spares.sav_count; i++) 710fa94a07fSbrendan vdev_free(spa->spa_spares.sav_vdevs[i]); 711fa94a07fSbrendan if (spa->spa_spares.sav_vdevs) { 712fa94a07fSbrendan kmem_free(spa->spa_spares.sav_vdevs, 713fa94a07fSbrendan spa->spa_spares.sav_count * sizeof (void *)); 714fa94a07fSbrendan spa->spa_spares.sav_vdevs = NULL; 71599653d4eSeschrock } 716fa94a07fSbrendan if (spa->spa_spares.sav_config) { 717fa94a07fSbrendan nvlist_free(spa->spa_spares.sav_config); 718fa94a07fSbrendan spa->spa_spares.sav_config = NULL; 719fa94a07fSbrendan } 7202ce8af81SEric Schrock spa->spa_spares.sav_count = 0; 721fa94a07fSbrendan 722fa94a07fSbrendan for (i = 0; i < spa->spa_l2cache.sav_count; i++) 723fa94a07fSbrendan vdev_free(spa->spa_l2cache.sav_vdevs[i]); 724fa94a07fSbrendan if (spa->spa_l2cache.sav_vdevs) { 725fa94a07fSbrendan kmem_free(spa->spa_l2cache.sav_vdevs, 726fa94a07fSbrendan spa->spa_l2cache.sav_count * sizeof (void *)); 727fa94a07fSbrendan spa->spa_l2cache.sav_vdevs = NULL; 728fa94a07fSbrendan } 729fa94a07fSbrendan if (spa->spa_l2cache.sav_config) { 730fa94a07fSbrendan nvlist_free(spa->spa_l2cache.sav_config); 731fa94a07fSbrendan spa->spa_l2cache.sav_config = NULL; 73299653d4eSeschrock } 7332ce8af81SEric Schrock spa->spa_l2cache.sav_count = 0; 73499653d4eSeschrock 735ea8dc4b6Seschrock spa->spa_async_suspended = 0; 7368ad4d6ddSJeff Bonwick 7378ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 738fa9e4066Sahrens } 739fa9e4066Sahrens 74099653d4eSeschrock /* 74199653d4eSeschrock * Load (or re-load) the current list of vdevs describing the active spares for 74299653d4eSeschrock * this pool. When this is called, we have some form of basic information in 743fa94a07fSbrendan * 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and 744fa94a07fSbrendan * then re-generate a more complete list including status information. 74599653d4eSeschrock */ 74699653d4eSeschrock static void 74799653d4eSeschrock spa_load_spares(spa_t *spa) 74899653d4eSeschrock { 74999653d4eSeschrock nvlist_t **spares; 75099653d4eSeschrock uint_t nspares; 75199653d4eSeschrock int i; 75239c23413Seschrock vdev_t *vd, *tvd; 75399653d4eSeschrock 754e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 755e14bb325SJeff Bonwick 75699653d4eSeschrock /* 75799653d4eSeschrock * First, close and free any existing spare vdevs. 75899653d4eSeschrock */ 759fa94a07fSbrendan for (i = 0; i < spa->spa_spares.sav_count; i++) { 760fa94a07fSbrendan vd = spa->spa_spares.sav_vdevs[i]; 76139c23413Seschrock 76239c23413Seschrock /* Undo the call to spa_activate() below */ 763c5904d13Seschrock if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, 764c5904d13Seschrock B_FALSE)) != NULL && tvd->vdev_isspare) 76539c23413Seschrock spa_spare_remove(tvd); 76639c23413Seschrock vdev_close(vd); 76739c23413Seschrock vdev_free(vd); 76899653d4eSeschrock } 76939c23413Seschrock 770fa94a07fSbrendan if (spa->spa_spares.sav_vdevs) 771fa94a07fSbrendan kmem_free(spa->spa_spares.sav_vdevs, 772fa94a07fSbrendan spa->spa_spares.sav_count * sizeof (void *)); 77399653d4eSeschrock 774fa94a07fSbrendan if (spa->spa_spares.sav_config == NULL) 77599653d4eSeschrock nspares = 0; 77699653d4eSeschrock else 777fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, 77899653d4eSeschrock ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); 77999653d4eSeschrock 780fa94a07fSbrendan spa->spa_spares.sav_count = (int)nspares; 781fa94a07fSbrendan spa->spa_spares.sav_vdevs = NULL; 78299653d4eSeschrock 78399653d4eSeschrock if (nspares == 0) 78499653d4eSeschrock return; 78599653d4eSeschrock 78699653d4eSeschrock /* 78799653d4eSeschrock * Construct the array of vdevs, opening them to get status in the 78839c23413Seschrock * process. For each spare, there is potentially two different vdev_t 78939c23413Seschrock * structures associated with it: one in the list of spares (used only 79039c23413Seschrock * for basic validation purposes) and one in the active vdev 79139c23413Seschrock * configuration (if it's spared in). During this phase we open and 79239c23413Seschrock * validate each vdev on the spare list. If the vdev also exists in the 79339c23413Seschrock * active configuration, then we also mark this vdev as an active spare. 79499653d4eSeschrock */ 795fa94a07fSbrendan spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *), 796fa94a07fSbrendan KM_SLEEP); 797fa94a07fSbrendan for (i = 0; i < spa->spa_spares.sav_count; i++) { 79899653d4eSeschrock VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0, 79999653d4eSeschrock VDEV_ALLOC_SPARE) == 0); 80099653d4eSeschrock ASSERT(vd != NULL); 80199653d4eSeschrock 802fa94a07fSbrendan spa->spa_spares.sav_vdevs[i] = vd; 80399653d4eSeschrock 804c5904d13Seschrock if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid, 805c5904d13Seschrock B_FALSE)) != NULL) { 80639c23413Seschrock if (!tvd->vdev_isspare) 80739c23413Seschrock spa_spare_add(tvd); 80839c23413Seschrock 80939c23413Seschrock /* 81039c23413Seschrock * We only mark the spare active if we were successfully 81139c23413Seschrock * able to load the vdev. Otherwise, importing a pool 81239c23413Seschrock * with a bad active spare would result in strange 81339c23413Seschrock * behavior, because multiple pool would think the spare 81439c23413Seschrock * is actively in use. 81539c23413Seschrock * 81639c23413Seschrock * There is a vulnerability here to an equally bizarre 81739c23413Seschrock * circumstance, where a dead active spare is later 81839c23413Seschrock * brought back to life (onlined or otherwise). Given 81939c23413Seschrock * the rarity of this scenario, and the extra complexity 82039c23413Seschrock * it adds, we ignore the possibility. 82139c23413Seschrock */ 82239c23413Seschrock if (!vdev_is_dead(tvd)) 82339c23413Seschrock spa_spare_activate(tvd); 82439c23413Seschrock } 82539c23413Seschrock 826e14bb325SJeff Bonwick vd->vdev_top = vd; 827e14bb325SJeff Bonwick 82899653d4eSeschrock if (vdev_open(vd) != 0) 82999653d4eSeschrock continue; 83099653d4eSeschrock 831fa94a07fSbrendan if (vdev_validate_aux(vd) == 0) 832fa94a07fSbrendan spa_spare_add(vd); 83399653d4eSeschrock } 83499653d4eSeschrock 83599653d4eSeschrock /* 83699653d4eSeschrock * Recompute the stashed list of spares, with status information 83799653d4eSeschrock * this time. 83899653d4eSeschrock */ 839fa94a07fSbrendan VERIFY(nvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, 84099653d4eSeschrock DATA_TYPE_NVLIST_ARRAY) == 0); 84199653d4eSeschrock 842fa94a07fSbrendan spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *), 843fa94a07fSbrendan KM_SLEEP); 844fa94a07fSbrendan for (i = 0; i < spa->spa_spares.sav_count; i++) 845fa94a07fSbrendan spares[i] = vdev_config_generate(spa, 846fa94a07fSbrendan spa->spa_spares.sav_vdevs[i], B_TRUE, B_TRUE, B_FALSE); 847fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, 848fa94a07fSbrendan ZPOOL_CONFIG_SPARES, spares, spa->spa_spares.sav_count) == 0); 849fa94a07fSbrendan for (i = 0; i < spa->spa_spares.sav_count; i++) 85099653d4eSeschrock nvlist_free(spares[i]); 851fa94a07fSbrendan kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *)); 852fa94a07fSbrendan } 853fa94a07fSbrendan 854fa94a07fSbrendan /* 855fa94a07fSbrendan * Load (or re-load) the current list of vdevs describing the active l2cache for 856fa94a07fSbrendan * this pool. When this is called, we have some form of basic information in 857fa94a07fSbrendan * 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and 858fa94a07fSbrendan * then re-generate a more complete list including status information. 859fa94a07fSbrendan * Devices which are already active have their details maintained, and are 860fa94a07fSbrendan * not re-opened. 861fa94a07fSbrendan */ 862fa94a07fSbrendan static void 863fa94a07fSbrendan spa_load_l2cache(spa_t *spa) 864fa94a07fSbrendan { 865fa94a07fSbrendan nvlist_t **l2cache; 866fa94a07fSbrendan uint_t nl2cache; 867fa94a07fSbrendan int i, j, oldnvdevs; 868c5904d13Seschrock uint64_t guid, size; 869fa94a07fSbrendan vdev_t *vd, **oldvdevs, **newvdevs; 870fa94a07fSbrendan spa_aux_vdev_t *sav = &spa->spa_l2cache; 871fa94a07fSbrendan 872e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 873e14bb325SJeff Bonwick 874fa94a07fSbrendan if (sav->sav_config != NULL) { 875fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, 876fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); 877fa94a07fSbrendan newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP); 878fa94a07fSbrendan } else { 879fa94a07fSbrendan nl2cache = 0; 880fa94a07fSbrendan } 881fa94a07fSbrendan 882fa94a07fSbrendan oldvdevs = sav->sav_vdevs; 883fa94a07fSbrendan oldnvdevs = sav->sav_count; 884fa94a07fSbrendan sav->sav_vdevs = NULL; 885fa94a07fSbrendan sav->sav_count = 0; 886fa94a07fSbrendan 887fa94a07fSbrendan /* 888fa94a07fSbrendan * Process new nvlist of vdevs. 889fa94a07fSbrendan */ 890fa94a07fSbrendan for (i = 0; i < nl2cache; i++) { 891fa94a07fSbrendan VERIFY(nvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID, 892fa94a07fSbrendan &guid) == 0); 893fa94a07fSbrendan 894fa94a07fSbrendan newvdevs[i] = NULL; 895fa94a07fSbrendan for (j = 0; j < oldnvdevs; j++) { 896fa94a07fSbrendan vd = oldvdevs[j]; 897fa94a07fSbrendan if (vd != NULL && guid == vd->vdev_guid) { 898fa94a07fSbrendan /* 899fa94a07fSbrendan * Retain previous vdev for add/remove ops. 900fa94a07fSbrendan */ 901fa94a07fSbrendan newvdevs[i] = vd; 902fa94a07fSbrendan oldvdevs[j] = NULL; 903fa94a07fSbrendan break; 904fa94a07fSbrendan } 905fa94a07fSbrendan } 906fa94a07fSbrendan 907fa94a07fSbrendan if (newvdevs[i] == NULL) { 908fa94a07fSbrendan /* 909fa94a07fSbrendan * Create new vdev 910fa94a07fSbrendan */ 911fa94a07fSbrendan VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, 912fa94a07fSbrendan VDEV_ALLOC_L2CACHE) == 0); 913fa94a07fSbrendan ASSERT(vd != NULL); 914fa94a07fSbrendan newvdevs[i] = vd; 915fa94a07fSbrendan 916fa94a07fSbrendan /* 917fa94a07fSbrendan * Commit this vdev as an l2cache device, 918fa94a07fSbrendan * even if it fails to open. 919fa94a07fSbrendan */ 920fa94a07fSbrendan spa_l2cache_add(vd); 921fa94a07fSbrendan 922c5904d13Seschrock vd->vdev_top = vd; 923c5904d13Seschrock vd->vdev_aux = sav; 924c5904d13Seschrock 925c5904d13Seschrock spa_l2cache_activate(vd); 926c5904d13Seschrock 927fa94a07fSbrendan if (vdev_open(vd) != 0) 928fa94a07fSbrendan continue; 929fa94a07fSbrendan 930fa94a07fSbrendan (void) vdev_validate_aux(vd); 931fa94a07fSbrendan 932fa94a07fSbrendan if (!vdev_is_dead(vd)) { 933fa94a07fSbrendan size = vdev_get_rsize(vd); 934c5904d13Seschrock l2arc_add_vdev(spa, vd, 935c5904d13Seschrock VDEV_LABEL_START_SIZE, 936c5904d13Seschrock size - VDEV_LABEL_START_SIZE); 937fa94a07fSbrendan } 938fa94a07fSbrendan } 939fa94a07fSbrendan } 940fa94a07fSbrendan 941fa94a07fSbrendan /* 942fa94a07fSbrendan * Purge vdevs that were dropped 943fa94a07fSbrendan */ 944fa94a07fSbrendan for (i = 0; i < oldnvdevs; i++) { 945fa94a07fSbrendan uint64_t pool; 946fa94a07fSbrendan 947fa94a07fSbrendan vd = oldvdevs[i]; 948fa94a07fSbrendan if (vd != NULL) { 9498ad4d6ddSJeff Bonwick if (spa_l2cache_exists(vd->vdev_guid, &pool) && 9508ad4d6ddSJeff Bonwick pool != 0ULL && l2arc_vdev_present(vd)) 951fa94a07fSbrendan l2arc_remove_vdev(vd); 952fa94a07fSbrendan (void) vdev_close(vd); 953fa94a07fSbrendan spa_l2cache_remove(vd); 954fa94a07fSbrendan } 955fa94a07fSbrendan } 956fa94a07fSbrendan 957fa94a07fSbrendan if (oldvdevs) 958fa94a07fSbrendan kmem_free(oldvdevs, oldnvdevs * sizeof (void *)); 959fa94a07fSbrendan 960fa94a07fSbrendan if (sav->sav_config == NULL) 961fa94a07fSbrendan goto out; 962fa94a07fSbrendan 963fa94a07fSbrendan sav->sav_vdevs = newvdevs; 964fa94a07fSbrendan sav->sav_count = (int)nl2cache; 965fa94a07fSbrendan 966fa94a07fSbrendan /* 967fa94a07fSbrendan * Recompute the stashed list of l2cache devices, with status 968fa94a07fSbrendan * information this time. 969fa94a07fSbrendan */ 970fa94a07fSbrendan VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE, 971fa94a07fSbrendan DATA_TYPE_NVLIST_ARRAY) == 0); 972fa94a07fSbrendan 973fa94a07fSbrendan l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP); 974fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) 975fa94a07fSbrendan l2cache[i] = vdev_config_generate(spa, 976fa94a07fSbrendan sav->sav_vdevs[i], B_TRUE, B_FALSE, B_TRUE); 977fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(sav->sav_config, 978fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0); 979fa94a07fSbrendan out: 980fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) 981fa94a07fSbrendan nvlist_free(l2cache[i]); 982fa94a07fSbrendan if (sav->sav_count) 983fa94a07fSbrendan kmem_free(l2cache, sav->sav_count * sizeof (void *)); 98499653d4eSeschrock } 98599653d4eSeschrock 98699653d4eSeschrock static int 98799653d4eSeschrock load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value) 98899653d4eSeschrock { 98999653d4eSeschrock dmu_buf_t *db; 99099653d4eSeschrock char *packed = NULL; 99199653d4eSeschrock size_t nvsize = 0; 99299653d4eSeschrock int error; 99399653d4eSeschrock *value = NULL; 99499653d4eSeschrock 99599653d4eSeschrock VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); 99699653d4eSeschrock nvsize = *(uint64_t *)db->db_data; 99799653d4eSeschrock dmu_buf_rele(db, FTAG); 99899653d4eSeschrock 99999653d4eSeschrock packed = kmem_alloc(nvsize, KM_SLEEP); 100099653d4eSeschrock error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed); 100199653d4eSeschrock if (error == 0) 100299653d4eSeschrock error = nvlist_unpack(packed, nvsize, value, 0); 100399653d4eSeschrock kmem_free(packed, nvsize); 100499653d4eSeschrock 100599653d4eSeschrock return (error); 100699653d4eSeschrock } 100799653d4eSeschrock 10083d7072f8Seschrock /* 10093d7072f8Seschrock * Checks to see if the given vdev could not be opened, in which case we post a 10103d7072f8Seschrock * sysevent to notify the autoreplace code that the device has been removed. 10113d7072f8Seschrock */ 10123d7072f8Seschrock static void 10133d7072f8Seschrock spa_check_removed(vdev_t *vd) 10143d7072f8Seschrock { 10153d7072f8Seschrock int c; 10163d7072f8Seschrock 10173d7072f8Seschrock for (c = 0; c < vd->vdev_children; c++) 10183d7072f8Seschrock spa_check_removed(vd->vdev_child[c]); 10193d7072f8Seschrock 10203d7072f8Seschrock if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd)) { 10213d7072f8Seschrock zfs_post_autoreplace(vd->vdev_spa, vd); 10223d7072f8Seschrock spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_CHECK); 10233d7072f8Seschrock } 10243d7072f8Seschrock } 10253d7072f8Seschrock 1026b87f3af3Sperrin /* 1027b87f3af3Sperrin * Check for missing log devices 1028b87f3af3Sperrin */ 1029b87f3af3Sperrin int 1030b87f3af3Sperrin spa_check_logs(spa_t *spa) 1031b87f3af3Sperrin { 1032b87f3af3Sperrin switch (spa->spa_log_state) { 1033b87f3af3Sperrin case SPA_LOG_MISSING: 1034b87f3af3Sperrin /* need to recheck in case slog has been restored */ 1035b87f3af3Sperrin case SPA_LOG_UNKNOWN: 1036b87f3af3Sperrin if (dmu_objset_find(spa->spa_name, zil_check_log_chain, NULL, 1037b87f3af3Sperrin DS_FIND_CHILDREN)) { 1038b87f3af3Sperrin spa->spa_log_state = SPA_LOG_MISSING; 1039b87f3af3Sperrin return (1); 1040b87f3af3Sperrin } 1041b87f3af3Sperrin break; 1042b87f3af3Sperrin 1043b87f3af3Sperrin case SPA_LOG_CLEAR: 1044b87f3af3Sperrin (void) dmu_objset_find(spa->spa_name, zil_clear_log_chain, NULL, 1045b87f3af3Sperrin DS_FIND_CHILDREN); 1046b87f3af3Sperrin break; 1047b87f3af3Sperrin } 1048b87f3af3Sperrin spa->spa_log_state = SPA_LOG_GOOD; 1049b87f3af3Sperrin return (0); 1050b87f3af3Sperrin } 1051b87f3af3Sperrin 1052fa9e4066Sahrens /* 1053fa9e4066Sahrens * Load an existing storage pool, using the pool's builtin spa_config as a 1054ea8dc4b6Seschrock * source of configuration information. 1055fa9e4066Sahrens */ 1056fa9e4066Sahrens static int 1057ea8dc4b6Seschrock spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig) 1058fa9e4066Sahrens { 1059fa9e4066Sahrens int error = 0; 1060fa9e4066Sahrens nvlist_t *nvroot = NULL; 1061fa9e4066Sahrens vdev_t *rvd; 1062fa9e4066Sahrens uberblock_t *ub = &spa->spa_uberblock; 10630373e76bSbonwick uint64_t config_cache_txg = spa->spa_config_txg; 1064fa9e4066Sahrens uint64_t pool_guid; 106599653d4eSeschrock uint64_t version; 10663d7072f8Seschrock uint64_t autoreplace = 0; 10678ad4d6ddSJeff Bonwick int orig_mode = spa->spa_mode; 1068b87f3af3Sperrin char *ereport = FM_EREPORT_ZFS_POOL; 1069fa9e4066Sahrens 10708ad4d6ddSJeff Bonwick /* 10718ad4d6ddSJeff Bonwick * If this is an untrusted config, access the pool in read-only mode. 10728ad4d6ddSJeff Bonwick * This prevents things like resilvering recently removed devices. 10738ad4d6ddSJeff Bonwick */ 10748ad4d6ddSJeff Bonwick if (!mosconfig) 10758ad4d6ddSJeff Bonwick spa->spa_mode = FREAD; 10768ad4d6ddSJeff Bonwick 1077e14bb325SJeff Bonwick ASSERT(MUTEX_HELD(&spa_namespace_lock)); 1078e14bb325SJeff Bonwick 1079ea8dc4b6Seschrock spa->spa_load_state = state; 10800373e76bSbonwick 1081fa9e4066Sahrens if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) || 1082a9926bf0Sbonwick nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) { 1083ea8dc4b6Seschrock error = EINVAL; 1084ea8dc4b6Seschrock goto out; 1085ea8dc4b6Seschrock } 1086fa9e4066Sahrens 108799653d4eSeschrock /* 108899653d4eSeschrock * Versioning wasn't explicitly added to the label until later, so if 108999653d4eSeschrock * it's not present treat it as the initial version. 109099653d4eSeschrock */ 109199653d4eSeschrock if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) != 0) 1092e7437265Sahrens version = SPA_VERSION_INITIAL; 109399653d4eSeschrock 1094a9926bf0Sbonwick (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 1095a9926bf0Sbonwick &spa->spa_config_txg); 1096a9926bf0Sbonwick 10970373e76bSbonwick if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) && 1098ea8dc4b6Seschrock spa_guid_exists(pool_guid, 0)) { 1099ea8dc4b6Seschrock error = EEXIST; 1100ea8dc4b6Seschrock goto out; 1101ea8dc4b6Seschrock } 1102fa9e4066Sahrens 1103b5989ec7Seschrock spa->spa_load_guid = pool_guid; 1104b5989ec7Seschrock 1105fa9e4066Sahrens /* 110699653d4eSeschrock * Parse the configuration into a vdev tree. We explicitly set the 110799653d4eSeschrock * value that will be returned by spa_version() since parsing the 110899653d4eSeschrock * configuration requires knowing the version number. 1109fa9e4066Sahrens */ 1110e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 111199653d4eSeschrock spa->spa_ubsync.ub_version = version; 111299653d4eSeschrock error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_LOAD); 1113e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1114fa9e4066Sahrens 111599653d4eSeschrock if (error != 0) 1116ea8dc4b6Seschrock goto out; 1117fa9e4066Sahrens 11180e34b6a7Sbonwick ASSERT(spa->spa_root_vdev == rvd); 1119fa9e4066Sahrens ASSERT(spa_guid(spa) == pool_guid); 1120fa9e4066Sahrens 1121fa9e4066Sahrens /* 1122fa9e4066Sahrens * Try to open all vdevs, loading each label in the process. 1123fa9e4066Sahrens */ 1124e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 11250bf246f5Smc error = vdev_open(rvd); 1126e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 11270bf246f5Smc if (error != 0) 1128ea8dc4b6Seschrock goto out; 1129fa9e4066Sahrens 1130560e6e96Seschrock /* 1131560e6e96Seschrock * Validate the labels for all leaf vdevs. We need to grab the config 1132e14bb325SJeff Bonwick * lock because all label I/O is done with ZIO_FLAG_CONFIG_WRITER. 1133560e6e96Seschrock */ 11348ad4d6ddSJeff Bonwick if (mosconfig) { 11358ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 11368ad4d6ddSJeff Bonwick error = vdev_validate(rvd); 11378ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 11388ad4d6ddSJeff Bonwick if (error != 0) 11398ad4d6ddSJeff Bonwick goto out; 11408ad4d6ddSJeff Bonwick } 1141560e6e96Seschrock 1142560e6e96Seschrock if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) { 1143560e6e96Seschrock error = ENXIO; 1144560e6e96Seschrock goto out; 1145560e6e96Seschrock } 1146560e6e96Seschrock 1147fa9e4066Sahrens /* 1148fa9e4066Sahrens * Find the best uberblock. 1149fa9e4066Sahrens */ 1150e14bb325SJeff Bonwick vdev_uberblock_load(NULL, rvd, ub); 1151fa9e4066Sahrens 1152fa9e4066Sahrens /* 1153fa9e4066Sahrens * If we weren't able to find a single valid uberblock, return failure. 1154fa9e4066Sahrens */ 1155fa9e4066Sahrens if (ub->ub_txg == 0) { 1156eaca9bbdSeschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1157eaca9bbdSeschrock VDEV_AUX_CORRUPT_DATA); 1158ea8dc4b6Seschrock error = ENXIO; 1159ea8dc4b6Seschrock goto out; 1160ea8dc4b6Seschrock } 1161ea8dc4b6Seschrock 1162ea8dc4b6Seschrock /* 1163ea8dc4b6Seschrock * If the pool is newer than the code, we can't open it. 1164ea8dc4b6Seschrock */ 1165e7437265Sahrens if (ub->ub_version > SPA_VERSION) { 1166eaca9bbdSeschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1167eaca9bbdSeschrock VDEV_AUX_VERSION_NEWER); 1168ea8dc4b6Seschrock error = ENOTSUP; 1169ea8dc4b6Seschrock goto out; 1170fa9e4066Sahrens } 1171fa9e4066Sahrens 1172fa9e4066Sahrens /* 1173fa9e4066Sahrens * If the vdev guid sum doesn't match the uberblock, we have an 1174fa9e4066Sahrens * incomplete configuration. 1175fa9e4066Sahrens */ 1176ecc2d604Sbonwick if (rvd->vdev_guid_sum != ub->ub_guid_sum && mosconfig) { 1177ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1178ea8dc4b6Seschrock VDEV_AUX_BAD_GUID_SUM); 1179ea8dc4b6Seschrock error = ENXIO; 1180ea8dc4b6Seschrock goto out; 1181fa9e4066Sahrens } 1182fa9e4066Sahrens 1183fa9e4066Sahrens /* 1184fa9e4066Sahrens * Initialize internal SPA structures. 1185fa9e4066Sahrens */ 1186fa9e4066Sahrens spa->spa_state = POOL_STATE_ACTIVE; 1187fa9e4066Sahrens spa->spa_ubsync = spa->spa_uberblock; 1188fa9e4066Sahrens spa->spa_first_txg = spa_last_synced_txg(spa) + 1; 1189ea8dc4b6Seschrock error = dsl_pool_open(spa, spa->spa_first_txg, &spa->spa_dsl_pool); 1190ea8dc4b6Seschrock if (error) { 1191ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1192ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1193ea8dc4b6Seschrock goto out; 1194ea8dc4b6Seschrock } 1195fa9e4066Sahrens spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset; 1196fa9e4066Sahrens 1197ea8dc4b6Seschrock if (zap_lookup(spa->spa_meta_objset, 1198fa9e4066Sahrens DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG, 1199ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_config_object) != 0) { 1200ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1201ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1202ea8dc4b6Seschrock error = EIO; 1203ea8dc4b6Seschrock goto out; 1204ea8dc4b6Seschrock } 1205fa9e4066Sahrens 1206fa9e4066Sahrens if (!mosconfig) { 120799653d4eSeschrock nvlist_t *newconfig; 120895173954Sek uint64_t hostid; 1209fa9e4066Sahrens 121099653d4eSeschrock if (load_nvlist(spa, spa->spa_config_object, &newconfig) != 0) { 1211ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1212ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1213ea8dc4b6Seschrock error = EIO; 1214ea8dc4b6Seschrock goto out; 1215ea8dc4b6Seschrock } 1216fa9e4066Sahrens 121777650510SLin Ling if (!spa_is_root(spa) && nvlist_lookup_uint64(newconfig, 121877650510SLin Ling ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 121995173954Sek char *hostname; 122095173954Sek unsigned long myhostid = 0; 122195173954Sek 122295173954Sek VERIFY(nvlist_lookup_string(newconfig, 122395173954Sek ZPOOL_CONFIG_HOSTNAME, &hostname) == 0); 122495173954Sek 122595173954Sek (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid); 122617194a52Slling if (hostid != 0 && myhostid != 0 && 122717194a52Slling (unsigned long)hostid != myhostid) { 122895173954Sek cmn_err(CE_WARN, "pool '%s' could not be " 122995173954Sek "loaded as it was last accessed by " 123077650510SLin Ling "another system (host: %s hostid: 0x%lx). " 123195173954Sek "See: http://www.sun.com/msg/ZFS-8000-EY", 1232e14bb325SJeff Bonwick spa_name(spa), hostname, 123395173954Sek (unsigned long)hostid); 123495173954Sek error = EBADF; 123595173954Sek goto out; 123695173954Sek } 123795173954Sek } 123895173954Sek 1239fa9e4066Sahrens spa_config_set(spa, newconfig); 1240fa9e4066Sahrens spa_unload(spa); 1241fa9e4066Sahrens spa_deactivate(spa); 12428ad4d6ddSJeff Bonwick spa_activate(spa, orig_mode); 1243fa9e4066Sahrens 1244ea8dc4b6Seschrock return (spa_load(spa, newconfig, state, B_TRUE)); 1245fa9e4066Sahrens } 1246fa9e4066Sahrens 1247ea8dc4b6Seschrock if (zap_lookup(spa->spa_meta_objset, 1248fa9e4066Sahrens DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPLIST, 1249ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj) != 0) { 1250ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1251ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1252ea8dc4b6Seschrock error = EIO; 1253ea8dc4b6Seschrock goto out; 1254ea8dc4b6Seschrock } 1255fa9e4066Sahrens 125699653d4eSeschrock /* 125799653d4eSeschrock * Load the bit that tells us to use the new accounting function 125899653d4eSeschrock * (raid-z deflation). If we have an older pool, this will not 125999653d4eSeschrock * be present. 126099653d4eSeschrock */ 126199653d4eSeschrock error = zap_lookup(spa->spa_meta_objset, 126299653d4eSeschrock DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, 126399653d4eSeschrock sizeof (uint64_t), 1, &spa->spa_deflate); 126499653d4eSeschrock if (error != 0 && error != ENOENT) { 126599653d4eSeschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 126699653d4eSeschrock VDEV_AUX_CORRUPT_DATA); 126799653d4eSeschrock error = EIO; 126899653d4eSeschrock goto out; 126999653d4eSeschrock } 127099653d4eSeschrock 1271fa9e4066Sahrens /* 1272ea8dc4b6Seschrock * Load the persistent error log. If we have an older pool, this will 1273ea8dc4b6Seschrock * not be present. 1274fa9e4066Sahrens */ 1275ea8dc4b6Seschrock error = zap_lookup(spa->spa_meta_objset, 1276ea8dc4b6Seschrock DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_LAST, 1277ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_errlog_last); 1278d80c45e0Sbonwick if (error != 0 && error != ENOENT) { 1279ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1280ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1281ea8dc4b6Seschrock error = EIO; 1282ea8dc4b6Seschrock goto out; 1283ea8dc4b6Seschrock } 1284ea8dc4b6Seschrock 1285ea8dc4b6Seschrock error = zap_lookup(spa->spa_meta_objset, 1286ea8dc4b6Seschrock DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_SCRUB, 1287ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_errlog_scrub); 1288ea8dc4b6Seschrock if (error != 0 && error != ENOENT) { 1289ea8dc4b6Seschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1290ea8dc4b6Seschrock VDEV_AUX_CORRUPT_DATA); 1291ea8dc4b6Seschrock error = EIO; 1292ea8dc4b6Seschrock goto out; 1293ea8dc4b6Seschrock } 1294ea8dc4b6Seschrock 129506eeb2adSek /* 129606eeb2adSek * Load the history object. If we have an older pool, this 129706eeb2adSek * will not be present. 129806eeb2adSek */ 129906eeb2adSek error = zap_lookup(spa->spa_meta_objset, 130006eeb2adSek DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_HISTORY, 130106eeb2adSek sizeof (uint64_t), 1, &spa->spa_history); 130206eeb2adSek if (error != 0 && error != ENOENT) { 130306eeb2adSek vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 130406eeb2adSek VDEV_AUX_CORRUPT_DATA); 130506eeb2adSek error = EIO; 130606eeb2adSek goto out; 130706eeb2adSek } 130806eeb2adSek 130999653d4eSeschrock /* 131099653d4eSeschrock * Load any hot spares for this pool. 131199653d4eSeschrock */ 131299653d4eSeschrock error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, 1313fa94a07fSbrendan DMU_POOL_SPARES, sizeof (uint64_t), 1, &spa->spa_spares.sav_object); 131499653d4eSeschrock if (error != 0 && error != ENOENT) { 131599653d4eSeschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 131699653d4eSeschrock VDEV_AUX_CORRUPT_DATA); 131799653d4eSeschrock error = EIO; 131899653d4eSeschrock goto out; 131999653d4eSeschrock } 132099653d4eSeschrock if (error == 0) { 1321e7437265Sahrens ASSERT(spa_version(spa) >= SPA_VERSION_SPARES); 1322fa94a07fSbrendan if (load_nvlist(spa, spa->spa_spares.sav_object, 1323fa94a07fSbrendan &spa->spa_spares.sav_config) != 0) { 132499653d4eSeschrock vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 132599653d4eSeschrock VDEV_AUX_CORRUPT_DATA); 132699653d4eSeschrock error = EIO; 132799653d4eSeschrock goto out; 132899653d4eSeschrock } 132999653d4eSeschrock 1330e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 133199653d4eSeschrock spa_load_spares(spa); 1332e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 133399653d4eSeschrock } 133499653d4eSeschrock 1335fa94a07fSbrendan /* 1336fa94a07fSbrendan * Load any level 2 ARC devices for this pool. 1337fa94a07fSbrendan */ 1338fa94a07fSbrendan error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, 1339fa94a07fSbrendan DMU_POOL_L2CACHE, sizeof (uint64_t), 1, 1340fa94a07fSbrendan &spa->spa_l2cache.sav_object); 1341fa94a07fSbrendan if (error != 0 && error != ENOENT) { 1342fa94a07fSbrendan vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1343fa94a07fSbrendan VDEV_AUX_CORRUPT_DATA); 1344fa94a07fSbrendan error = EIO; 1345fa94a07fSbrendan goto out; 1346fa94a07fSbrendan } 1347fa94a07fSbrendan if (error == 0) { 1348fa94a07fSbrendan ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE); 1349fa94a07fSbrendan if (load_nvlist(spa, spa->spa_l2cache.sav_object, 1350fa94a07fSbrendan &spa->spa_l2cache.sav_config) != 0) { 1351fa94a07fSbrendan vdev_set_state(rvd, B_TRUE, 1352fa94a07fSbrendan VDEV_STATE_CANT_OPEN, 1353fa94a07fSbrendan VDEV_AUX_CORRUPT_DATA); 1354fa94a07fSbrendan error = EIO; 1355fa94a07fSbrendan goto out; 1356fa94a07fSbrendan } 1357fa94a07fSbrendan 1358e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1359fa94a07fSbrendan spa_load_l2cache(spa); 1360e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1361fa94a07fSbrendan } 1362fa94a07fSbrendan 1363b87f3af3Sperrin if (spa_check_logs(spa)) { 1364b87f3af3Sperrin vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1365b87f3af3Sperrin VDEV_AUX_BAD_LOG); 1366b87f3af3Sperrin error = ENXIO; 1367b87f3af3Sperrin ereport = FM_EREPORT_ZFS_LOG_REPLAY; 1368b87f3af3Sperrin goto out; 1369b87f3af3Sperrin } 1370b87f3af3Sperrin 1371b87f3af3Sperrin 1372990b4856Slling spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); 1373ecd6cf80Smarks 1374b1b8ab34Slling error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, 1375b1b8ab34Slling DMU_POOL_PROPS, sizeof (uint64_t), 1, &spa->spa_pool_props_object); 1376b1b8ab34Slling 1377b1b8ab34Slling if (error && error != ENOENT) { 1378b1b8ab34Slling vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN, 1379b1b8ab34Slling VDEV_AUX_CORRUPT_DATA); 1380b1b8ab34Slling error = EIO; 1381b1b8ab34Slling goto out; 1382b1b8ab34Slling } 1383b1b8ab34Slling 1384b1b8ab34Slling if (error == 0) { 1385b1b8ab34Slling (void) zap_lookup(spa->spa_meta_objset, 1386b1b8ab34Slling spa->spa_pool_props_object, 13873d7072f8Seschrock zpool_prop_to_name(ZPOOL_PROP_BOOTFS), 1388b1b8ab34Slling sizeof (uint64_t), 1, &spa->spa_bootfs); 13893d7072f8Seschrock (void) zap_lookup(spa->spa_meta_objset, 13903d7072f8Seschrock spa->spa_pool_props_object, 13913d7072f8Seschrock zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 13923d7072f8Seschrock sizeof (uint64_t), 1, &autoreplace); 1393ecd6cf80Smarks (void) zap_lookup(spa->spa_meta_objset, 1394ecd6cf80Smarks spa->spa_pool_props_object, 1395ecd6cf80Smarks zpool_prop_to_name(ZPOOL_PROP_DELEGATION), 1396ecd6cf80Smarks sizeof (uint64_t), 1, &spa->spa_delegation); 13970a4e9518Sgw (void) zap_lookup(spa->spa_meta_objset, 13980a4e9518Sgw spa->spa_pool_props_object, 13990a4e9518Sgw zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), 14000a4e9518Sgw sizeof (uint64_t), 1, &spa->spa_failmode); 1401b1b8ab34Slling } 1402b1b8ab34Slling 14033d7072f8Seschrock /* 14043d7072f8Seschrock * If the 'autoreplace' property is set, then post a resource notifying 14053d7072f8Seschrock * the ZFS DE that it should not issue any faults for unopenable 14063d7072f8Seschrock * devices. We also iterate over the vdevs, and post a sysevent for any 14073d7072f8Seschrock * unopenable vdevs so that the normal autoreplace handler can take 14083d7072f8Seschrock * over. 14093d7072f8Seschrock */ 1410b01c3b58Seschrock if (autoreplace && state != SPA_LOAD_TRYIMPORT) 14113d7072f8Seschrock spa_check_removed(spa->spa_root_vdev); 14123d7072f8Seschrock 1413ea8dc4b6Seschrock /* 1414560e6e96Seschrock * Load the vdev state for all toplevel vdevs. 1415ea8dc4b6Seschrock */ 1416560e6e96Seschrock vdev_load(rvd); 14170373e76bSbonwick 1418fa9e4066Sahrens /* 1419fa9e4066Sahrens * Propagate the leaf DTLs we just loaded all the way up the tree. 1420fa9e4066Sahrens */ 1421e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1422fa9e4066Sahrens vdev_dtl_reassess(rvd, 0, 0, B_FALSE); 1423e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 1424fa9e4066Sahrens 1425fa9e4066Sahrens /* 1426fa9e4066Sahrens * Check the state of the root vdev. If it can't be opened, it 1427fa9e4066Sahrens * indicates one or more toplevel vdevs are faulted. 1428fa9e4066Sahrens */ 1429ea8dc4b6Seschrock if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) { 1430ea8dc4b6Seschrock error = ENXIO; 1431ea8dc4b6Seschrock goto out; 1432ea8dc4b6Seschrock } 1433fa9e4066Sahrens 14348ad4d6ddSJeff Bonwick if (spa_writeable(spa)) { 14355dabedeeSbonwick dmu_tx_t *tx; 14360373e76bSbonwick int need_update = B_FALSE; 14378ad4d6ddSJeff Bonwick 14388ad4d6ddSJeff Bonwick ASSERT(state != SPA_LOAD_TRYIMPORT); 14395dabedeeSbonwick 14400373e76bSbonwick /* 14410373e76bSbonwick * Claim log blocks that haven't been committed yet. 14420373e76bSbonwick * This must all happen in a single txg. 14430373e76bSbonwick */ 14445dabedeeSbonwick tx = dmu_tx_create_assigned(spa_get_dsl(spa), 1445fa9e4066Sahrens spa_first_txg(spa)); 1446e14bb325SJeff Bonwick (void) dmu_objset_find(spa_name(spa), 14470b69c2f0Sahrens zil_claim, tx, DS_FIND_CHILDREN); 1448fa9e4066Sahrens dmu_tx_commit(tx); 1449fa9e4066Sahrens 1450fa9e4066Sahrens spa->spa_sync_on = B_TRUE; 1451fa9e4066Sahrens txg_sync_start(spa->spa_dsl_pool); 1452fa9e4066Sahrens 1453fa9e4066Sahrens /* 1454fa9e4066Sahrens * Wait for all claims to sync. 1455fa9e4066Sahrens */ 1456fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, 0); 14570e34b6a7Sbonwick 14580e34b6a7Sbonwick /* 14590373e76bSbonwick * If the config cache is stale, or we have uninitialized 14600373e76bSbonwick * metaslabs (see spa_vdev_add()), then update the config. 14610e34b6a7Sbonwick */ 14620373e76bSbonwick if (config_cache_txg != spa->spa_config_txg || 14630373e76bSbonwick state == SPA_LOAD_IMPORT) 14640373e76bSbonwick need_update = B_TRUE; 14650373e76bSbonwick 14668ad4d6ddSJeff Bonwick for (int c = 0; c < rvd->vdev_children; c++) 14670373e76bSbonwick if (rvd->vdev_child[c]->vdev_ms_array == 0) 14680373e76bSbonwick need_update = B_TRUE; 14690e34b6a7Sbonwick 14700e34b6a7Sbonwick /* 14710373e76bSbonwick * Update the config cache asychronously in case we're the 14720373e76bSbonwick * root pool, in which case the config cache isn't writable yet. 14730e34b6a7Sbonwick */ 14740373e76bSbonwick if (need_update) 14750373e76bSbonwick spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE); 14768ad4d6ddSJeff Bonwick 14778ad4d6ddSJeff Bonwick /* 14788ad4d6ddSJeff Bonwick * Check all DTLs to see if anything needs resilvering. 14798ad4d6ddSJeff Bonwick */ 14808ad4d6ddSJeff Bonwick if (vdev_resilver_needed(rvd, NULL, NULL)) 14818ad4d6ddSJeff Bonwick spa_async_request(spa, SPA_ASYNC_RESILVER); 1482fa9e4066Sahrens } 1483fa9e4066Sahrens 1484ea8dc4b6Seschrock error = 0; 1485ea8dc4b6Seschrock out: 1486088f3894Sahrens spa->spa_minref = refcount_count(&spa->spa_refcount); 148799653d4eSeschrock if (error && error != EBADF) 1488b87f3af3Sperrin zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0); 1489ea8dc4b6Seschrock spa->spa_load_state = SPA_LOAD_NONE; 1490ea8dc4b6Seschrock spa->spa_ena = 0; 1491ea8dc4b6Seschrock 1492ea8dc4b6Seschrock return (error); 1493fa9e4066Sahrens } 1494fa9e4066Sahrens 1495fa9e4066Sahrens /* 1496fa9e4066Sahrens * Pool Open/Import 1497fa9e4066Sahrens * 1498fa9e4066Sahrens * The import case is identical to an open except that the configuration is sent 1499fa9e4066Sahrens * down from userland, instead of grabbed from the configuration cache. For the 1500fa9e4066Sahrens * case of an open, the pool configuration will exist in the 15013d7072f8Seschrock * POOL_STATE_UNINITIALIZED state. 1502fa9e4066Sahrens * 1503fa9e4066Sahrens * The stats information (gen/count/ustats) is used to gather vdev statistics at 1504fa9e4066Sahrens * the same time open the pool, without having to keep around the spa_t in some 1505fa9e4066Sahrens * ambiguous state. 1506fa9e4066Sahrens */ 1507fa9e4066Sahrens static int 1508fa9e4066Sahrens spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config) 1509fa9e4066Sahrens { 1510fa9e4066Sahrens spa_t *spa; 1511fa9e4066Sahrens int error; 1512fa9e4066Sahrens int locked = B_FALSE; 1513fa9e4066Sahrens 1514fa9e4066Sahrens *spapp = NULL; 1515fa9e4066Sahrens 1516fa9e4066Sahrens /* 1517fa9e4066Sahrens * As disgusting as this is, we need to support recursive calls to this 1518fa9e4066Sahrens * function because dsl_dir_open() is called during spa_load(), and ends 1519fa9e4066Sahrens * up calling spa_open() again. The real fix is to figure out how to 1520fa9e4066Sahrens * avoid dsl_dir_open() calling this in the first place. 1521fa9e4066Sahrens */ 1522fa9e4066Sahrens if (mutex_owner(&spa_namespace_lock) != curthread) { 1523fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1524fa9e4066Sahrens locked = B_TRUE; 1525fa9e4066Sahrens } 1526fa9e4066Sahrens 1527fa9e4066Sahrens if ((spa = spa_lookup(pool)) == NULL) { 1528fa9e4066Sahrens if (locked) 1529fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1530fa9e4066Sahrens return (ENOENT); 1531fa9e4066Sahrens } 1532fa9e4066Sahrens if (spa->spa_state == POOL_STATE_UNINITIALIZED) { 1533fa9e4066Sahrens 15348ad4d6ddSJeff Bonwick spa_activate(spa, spa_mode_global); 1535fa9e4066Sahrens 15360373e76bSbonwick error = spa_load(spa, spa->spa_config, SPA_LOAD_OPEN, B_FALSE); 1537fa9e4066Sahrens 1538fa9e4066Sahrens if (error == EBADF) { 1539fa9e4066Sahrens /* 1540560e6e96Seschrock * If vdev_validate() returns failure (indicated by 1541560e6e96Seschrock * EBADF), it indicates that one of the vdevs indicates 1542560e6e96Seschrock * that the pool has been exported or destroyed. If 1543560e6e96Seschrock * this is the case, the config cache is out of sync and 1544560e6e96Seschrock * we should remove the pool from the namespace. 1545fa9e4066Sahrens */ 1546fa9e4066Sahrens spa_unload(spa); 1547fa9e4066Sahrens spa_deactivate(spa); 1548c5904d13Seschrock spa_config_sync(spa, B_TRUE, B_TRUE); 1549fa9e4066Sahrens spa_remove(spa); 1550fa9e4066Sahrens if (locked) 1551fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1552fa9e4066Sahrens return (ENOENT); 1553ea8dc4b6Seschrock } 1554ea8dc4b6Seschrock 1555ea8dc4b6Seschrock if (error) { 1556fa9e4066Sahrens /* 1557fa9e4066Sahrens * We can't open the pool, but we still have useful 1558fa9e4066Sahrens * information: the state of each vdev after the 1559fa9e4066Sahrens * attempted vdev_open(). Return this to the user. 1560fa9e4066Sahrens */ 1561e14bb325SJeff Bonwick if (config != NULL && spa->spa_root_vdev != NULL) 1562fa9e4066Sahrens *config = spa_config_generate(spa, NULL, -1ULL, 1563fa9e4066Sahrens B_TRUE); 1564fa9e4066Sahrens spa_unload(spa); 1565fa9e4066Sahrens spa_deactivate(spa); 1566ea8dc4b6Seschrock spa->spa_last_open_failed = B_TRUE; 1567fa9e4066Sahrens if (locked) 1568fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1569fa9e4066Sahrens *spapp = NULL; 1570fa9e4066Sahrens return (error); 1571ea8dc4b6Seschrock } else { 1572ea8dc4b6Seschrock spa->spa_last_open_failed = B_FALSE; 1573fa9e4066Sahrens } 1574fa9e4066Sahrens } 1575fa9e4066Sahrens 1576fa9e4066Sahrens spa_open_ref(spa, tag); 15773d7072f8Seschrock 1578fa9e4066Sahrens if (locked) 1579fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1580fa9e4066Sahrens 1581fa9e4066Sahrens *spapp = spa; 1582fa9e4066Sahrens 1583e14bb325SJeff Bonwick if (config != NULL) 1584fa9e4066Sahrens *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); 1585fa9e4066Sahrens 1586fa9e4066Sahrens return (0); 1587fa9e4066Sahrens } 1588fa9e4066Sahrens 1589fa9e4066Sahrens int 1590fa9e4066Sahrens spa_open(const char *name, spa_t **spapp, void *tag) 1591fa9e4066Sahrens { 1592fa9e4066Sahrens return (spa_open_common(name, spapp, tag, NULL)); 1593fa9e4066Sahrens } 1594fa9e4066Sahrens 1595ea8dc4b6Seschrock /* 1596ea8dc4b6Seschrock * Lookup the given spa_t, incrementing the inject count in the process, 1597ea8dc4b6Seschrock * preventing it from being exported or destroyed. 1598ea8dc4b6Seschrock */ 1599ea8dc4b6Seschrock spa_t * 1600ea8dc4b6Seschrock spa_inject_addref(char *name) 1601ea8dc4b6Seschrock { 1602ea8dc4b6Seschrock spa_t *spa; 1603ea8dc4b6Seschrock 1604ea8dc4b6Seschrock mutex_enter(&spa_namespace_lock); 1605ea8dc4b6Seschrock if ((spa = spa_lookup(name)) == NULL) { 1606ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1607ea8dc4b6Seschrock return (NULL); 1608ea8dc4b6Seschrock } 1609ea8dc4b6Seschrock spa->spa_inject_ref++; 1610ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1611ea8dc4b6Seschrock 1612ea8dc4b6Seschrock return (spa); 1613ea8dc4b6Seschrock } 1614ea8dc4b6Seschrock 1615ea8dc4b6Seschrock void 1616ea8dc4b6Seschrock spa_inject_delref(spa_t *spa) 1617ea8dc4b6Seschrock { 1618ea8dc4b6Seschrock mutex_enter(&spa_namespace_lock); 1619ea8dc4b6Seschrock spa->spa_inject_ref--; 1620ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1621ea8dc4b6Seschrock } 1622ea8dc4b6Seschrock 1623fa94a07fSbrendan /* 1624fa94a07fSbrendan * Add spares device information to the nvlist. 1625fa94a07fSbrendan */ 162699653d4eSeschrock static void 162799653d4eSeschrock spa_add_spares(spa_t *spa, nvlist_t *config) 162899653d4eSeschrock { 162999653d4eSeschrock nvlist_t **spares; 163099653d4eSeschrock uint_t i, nspares; 163199653d4eSeschrock nvlist_t *nvroot; 163299653d4eSeschrock uint64_t guid; 163399653d4eSeschrock vdev_stat_t *vs; 163499653d4eSeschrock uint_t vsc; 163539c23413Seschrock uint64_t pool; 163699653d4eSeschrock 1637fa94a07fSbrendan if (spa->spa_spares.sav_count == 0) 163899653d4eSeschrock return; 163999653d4eSeschrock 164099653d4eSeschrock VERIFY(nvlist_lookup_nvlist(config, 164199653d4eSeschrock ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1642fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, 164399653d4eSeschrock ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); 164499653d4eSeschrock if (nspares != 0) { 164599653d4eSeschrock VERIFY(nvlist_add_nvlist_array(nvroot, 164699653d4eSeschrock ZPOOL_CONFIG_SPARES, spares, nspares) == 0); 164799653d4eSeschrock VERIFY(nvlist_lookup_nvlist_array(nvroot, 164899653d4eSeschrock ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0); 164999653d4eSeschrock 165099653d4eSeschrock /* 165199653d4eSeschrock * Go through and find any spares which have since been 165299653d4eSeschrock * repurposed as an active spare. If this is the case, update 165399653d4eSeschrock * their status appropriately. 165499653d4eSeschrock */ 165599653d4eSeschrock for (i = 0; i < nspares; i++) { 165699653d4eSeschrock VERIFY(nvlist_lookup_uint64(spares[i], 165799653d4eSeschrock ZPOOL_CONFIG_GUID, &guid) == 0); 165889a89ebfSlling if (spa_spare_exists(guid, &pool, NULL) && 165989a89ebfSlling pool != 0ULL) { 166099653d4eSeschrock VERIFY(nvlist_lookup_uint64_array( 166199653d4eSeschrock spares[i], ZPOOL_CONFIG_STATS, 166299653d4eSeschrock (uint64_t **)&vs, &vsc) == 0); 166399653d4eSeschrock vs->vs_state = VDEV_STATE_CANT_OPEN; 166499653d4eSeschrock vs->vs_aux = VDEV_AUX_SPARED; 166599653d4eSeschrock } 166699653d4eSeschrock } 166799653d4eSeschrock } 166899653d4eSeschrock } 166999653d4eSeschrock 1670fa94a07fSbrendan /* 1671fa94a07fSbrendan * Add l2cache device information to the nvlist, including vdev stats. 1672fa94a07fSbrendan */ 1673fa94a07fSbrendan static void 1674fa94a07fSbrendan spa_add_l2cache(spa_t *spa, nvlist_t *config) 1675fa94a07fSbrendan { 1676fa94a07fSbrendan nvlist_t **l2cache; 1677fa94a07fSbrendan uint_t i, j, nl2cache; 1678fa94a07fSbrendan nvlist_t *nvroot; 1679fa94a07fSbrendan uint64_t guid; 1680fa94a07fSbrendan vdev_t *vd; 1681fa94a07fSbrendan vdev_stat_t *vs; 1682fa94a07fSbrendan uint_t vsc; 1683fa94a07fSbrendan 1684fa94a07fSbrendan if (spa->spa_l2cache.sav_count == 0) 1685fa94a07fSbrendan return; 1686fa94a07fSbrendan 1687e14bb325SJeff Bonwick spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); 1688fa94a07fSbrendan 1689fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist(config, 1690fa94a07fSbrendan ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1691fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, 1692fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); 1693fa94a07fSbrendan if (nl2cache != 0) { 1694fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(nvroot, 1695fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); 1696fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(nvroot, 1697fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0); 1698fa94a07fSbrendan 1699fa94a07fSbrendan /* 1700fa94a07fSbrendan * Update level 2 cache device stats. 1701fa94a07fSbrendan */ 1702fa94a07fSbrendan 1703fa94a07fSbrendan for (i = 0; i < nl2cache; i++) { 1704fa94a07fSbrendan VERIFY(nvlist_lookup_uint64(l2cache[i], 1705fa94a07fSbrendan ZPOOL_CONFIG_GUID, &guid) == 0); 1706fa94a07fSbrendan 1707fa94a07fSbrendan vd = NULL; 1708fa94a07fSbrendan for (j = 0; j < spa->spa_l2cache.sav_count; j++) { 1709fa94a07fSbrendan if (guid == 1710fa94a07fSbrendan spa->spa_l2cache.sav_vdevs[j]->vdev_guid) { 1711fa94a07fSbrendan vd = spa->spa_l2cache.sav_vdevs[j]; 1712fa94a07fSbrendan break; 1713fa94a07fSbrendan } 1714fa94a07fSbrendan } 1715fa94a07fSbrendan ASSERT(vd != NULL); 1716fa94a07fSbrendan 1717fa94a07fSbrendan VERIFY(nvlist_lookup_uint64_array(l2cache[i], 1718fa94a07fSbrendan ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0); 1719fa94a07fSbrendan vdev_get_stats(vd, vs); 1720fa94a07fSbrendan } 1721fa94a07fSbrendan } 1722fa94a07fSbrendan 1723e14bb325SJeff Bonwick spa_config_exit(spa, SCL_CONFIG, FTAG); 1724fa94a07fSbrendan } 1725fa94a07fSbrendan 1726fa9e4066Sahrens int 1727ea8dc4b6Seschrock spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen) 1728fa9e4066Sahrens { 1729fa9e4066Sahrens int error; 1730fa9e4066Sahrens spa_t *spa; 1731fa9e4066Sahrens 1732fa9e4066Sahrens *config = NULL; 1733fa9e4066Sahrens error = spa_open_common(name, &spa, FTAG, config); 1734fa9e4066Sahrens 173599653d4eSeschrock if (spa && *config != NULL) { 1736ea8dc4b6Seschrock VERIFY(nvlist_add_uint64(*config, ZPOOL_CONFIG_ERRCOUNT, 1737ea8dc4b6Seschrock spa_get_errlog_size(spa)) == 0); 1738ea8dc4b6Seschrock 1739e14bb325SJeff Bonwick if (spa_suspended(spa)) 1740e14bb325SJeff Bonwick VERIFY(nvlist_add_uint64(*config, 1741e14bb325SJeff Bonwick ZPOOL_CONFIG_SUSPENDED, spa->spa_failmode) == 0); 1742e14bb325SJeff Bonwick 174399653d4eSeschrock spa_add_spares(spa, *config); 1744fa94a07fSbrendan spa_add_l2cache(spa, *config); 174599653d4eSeschrock } 174699653d4eSeschrock 1747ea8dc4b6Seschrock /* 1748ea8dc4b6Seschrock * We want to get the alternate root even for faulted pools, so we cheat 1749ea8dc4b6Seschrock * and call spa_lookup() directly. 1750ea8dc4b6Seschrock */ 1751ea8dc4b6Seschrock if (altroot) { 1752ea8dc4b6Seschrock if (spa == NULL) { 1753ea8dc4b6Seschrock mutex_enter(&spa_namespace_lock); 1754ea8dc4b6Seschrock spa = spa_lookup(name); 1755ea8dc4b6Seschrock if (spa) 1756ea8dc4b6Seschrock spa_altroot(spa, altroot, buflen); 1757ea8dc4b6Seschrock else 1758ea8dc4b6Seschrock altroot[0] = '\0'; 1759ea8dc4b6Seschrock spa = NULL; 1760ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 1761ea8dc4b6Seschrock } else { 1762ea8dc4b6Seschrock spa_altroot(spa, altroot, buflen); 1763ea8dc4b6Seschrock } 1764ea8dc4b6Seschrock } 1765ea8dc4b6Seschrock 1766fa9e4066Sahrens if (spa != NULL) 1767fa9e4066Sahrens spa_close(spa, FTAG); 1768fa9e4066Sahrens 1769fa9e4066Sahrens return (error); 1770fa9e4066Sahrens } 1771fa9e4066Sahrens 177299653d4eSeschrock /* 1773fa94a07fSbrendan * Validate that the auxiliary device array is well formed. We must have an 1774fa94a07fSbrendan * array of nvlists, each which describes a valid leaf vdev. If this is an 1775fa94a07fSbrendan * import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be 1776fa94a07fSbrendan * specified, as long as they are well-formed. 177799653d4eSeschrock */ 177899653d4eSeschrock static int 1779fa94a07fSbrendan spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode, 1780fa94a07fSbrendan spa_aux_vdev_t *sav, const char *config, uint64_t version, 1781fa94a07fSbrendan vdev_labeltype_t label) 178299653d4eSeschrock { 1783fa94a07fSbrendan nvlist_t **dev; 1784fa94a07fSbrendan uint_t i, ndev; 178599653d4eSeschrock vdev_t *vd; 178699653d4eSeschrock int error; 178799653d4eSeschrock 1788e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 1789e14bb325SJeff Bonwick 179099653d4eSeschrock /* 1791fa94a07fSbrendan * It's acceptable to have no devs specified. 179299653d4eSeschrock */ 1793fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0) 179499653d4eSeschrock return (0); 179599653d4eSeschrock 1796fa94a07fSbrendan if (ndev == 0) 179799653d4eSeschrock return (EINVAL); 179899653d4eSeschrock 179999653d4eSeschrock /* 1800fa94a07fSbrendan * Make sure the pool is formatted with a version that supports this 1801fa94a07fSbrendan * device type. 180299653d4eSeschrock */ 1803fa94a07fSbrendan if (spa_version(spa) < version) 180499653d4eSeschrock return (ENOTSUP); 180599653d4eSeschrock 180639c23413Seschrock /* 1807fa94a07fSbrendan * Set the pending device list so we correctly handle device in-use 180839c23413Seschrock * checking. 180939c23413Seschrock */ 1810fa94a07fSbrendan sav->sav_pending = dev; 1811fa94a07fSbrendan sav->sav_npending = ndev; 181239c23413Seschrock 1813fa94a07fSbrendan for (i = 0; i < ndev; i++) { 1814fa94a07fSbrendan if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0, 181599653d4eSeschrock mode)) != 0) 181639c23413Seschrock goto out; 181799653d4eSeschrock 181899653d4eSeschrock if (!vd->vdev_ops->vdev_op_leaf) { 181999653d4eSeschrock vdev_free(vd); 182039c23413Seschrock error = EINVAL; 182139c23413Seschrock goto out; 182299653d4eSeschrock } 182399653d4eSeschrock 1824fa94a07fSbrendan /* 1825e14bb325SJeff Bonwick * The L2ARC currently only supports disk devices in 1826e14bb325SJeff Bonwick * kernel context. For user-level testing, we allow it. 1827fa94a07fSbrendan */ 1828e14bb325SJeff Bonwick #ifdef _KERNEL 1829fa94a07fSbrendan if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) && 1830fa94a07fSbrendan strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) { 1831fa94a07fSbrendan error = ENOTBLK; 1832fa94a07fSbrendan goto out; 1833fa94a07fSbrendan } 1834e14bb325SJeff Bonwick #endif 183599653d4eSeschrock vd->vdev_top = vd; 183699653d4eSeschrock 183739c23413Seschrock if ((error = vdev_open(vd)) == 0 && 1838fa94a07fSbrendan (error = vdev_label_init(vd, crtxg, label)) == 0) { 1839fa94a07fSbrendan VERIFY(nvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID, 184039c23413Seschrock vd->vdev_guid) == 0); 184139c23413Seschrock } 184299653d4eSeschrock 184399653d4eSeschrock vdev_free(vd); 184439c23413Seschrock 1845fa94a07fSbrendan if (error && 1846fa94a07fSbrendan (mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE)) 184739c23413Seschrock goto out; 184839c23413Seschrock else 184939c23413Seschrock error = 0; 185099653d4eSeschrock } 185199653d4eSeschrock 185239c23413Seschrock out: 1853fa94a07fSbrendan sav->sav_pending = NULL; 1854fa94a07fSbrendan sav->sav_npending = 0; 185539c23413Seschrock return (error); 185699653d4eSeschrock } 185799653d4eSeschrock 1858fa94a07fSbrendan static int 1859fa94a07fSbrendan spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode) 1860fa94a07fSbrendan { 1861fa94a07fSbrendan int error; 1862fa94a07fSbrendan 1863e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 1864e14bb325SJeff Bonwick 1865fa94a07fSbrendan if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode, 1866fa94a07fSbrendan &spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES, 1867fa94a07fSbrendan VDEV_LABEL_SPARE)) != 0) { 1868fa94a07fSbrendan return (error); 1869fa94a07fSbrendan } 1870fa94a07fSbrendan 1871fa94a07fSbrendan return (spa_validate_aux_devs(spa, nvroot, crtxg, mode, 1872fa94a07fSbrendan &spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE, 1873fa94a07fSbrendan VDEV_LABEL_L2CACHE)); 1874fa94a07fSbrendan } 1875fa94a07fSbrendan 1876fa94a07fSbrendan static void 1877fa94a07fSbrendan spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs, 1878fa94a07fSbrendan const char *config) 1879fa94a07fSbrendan { 1880fa94a07fSbrendan int i; 1881fa94a07fSbrendan 1882fa94a07fSbrendan if (sav->sav_config != NULL) { 1883fa94a07fSbrendan nvlist_t **olddevs; 1884fa94a07fSbrendan uint_t oldndevs; 1885fa94a07fSbrendan nvlist_t **newdevs; 1886fa94a07fSbrendan 1887fa94a07fSbrendan /* 1888fa94a07fSbrendan * Generate new dev list by concatentating with the 1889fa94a07fSbrendan * current dev list. 1890fa94a07fSbrendan */ 1891fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config, 1892fa94a07fSbrendan &olddevs, &oldndevs) == 0); 1893fa94a07fSbrendan 1894fa94a07fSbrendan newdevs = kmem_alloc(sizeof (void *) * 1895fa94a07fSbrendan (ndevs + oldndevs), KM_SLEEP); 1896fa94a07fSbrendan for (i = 0; i < oldndevs; i++) 1897fa94a07fSbrendan VERIFY(nvlist_dup(olddevs[i], &newdevs[i], 1898fa94a07fSbrendan KM_SLEEP) == 0); 1899fa94a07fSbrendan for (i = 0; i < ndevs; i++) 1900fa94a07fSbrendan VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs], 1901fa94a07fSbrendan KM_SLEEP) == 0); 1902fa94a07fSbrendan 1903fa94a07fSbrendan VERIFY(nvlist_remove(sav->sav_config, config, 1904fa94a07fSbrendan DATA_TYPE_NVLIST_ARRAY) == 0); 1905fa94a07fSbrendan 1906fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(sav->sav_config, 1907fa94a07fSbrendan config, newdevs, ndevs + oldndevs) == 0); 1908fa94a07fSbrendan for (i = 0; i < oldndevs + ndevs; i++) 1909fa94a07fSbrendan nvlist_free(newdevs[i]); 1910fa94a07fSbrendan kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *)); 1911fa94a07fSbrendan } else { 1912fa94a07fSbrendan /* 1913fa94a07fSbrendan * Generate a new dev list. 1914fa94a07fSbrendan */ 1915fa94a07fSbrendan VERIFY(nvlist_alloc(&sav->sav_config, NV_UNIQUE_NAME, 1916fa94a07fSbrendan KM_SLEEP) == 0); 1917fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(sav->sav_config, config, 1918fa94a07fSbrendan devs, ndevs) == 0); 1919fa94a07fSbrendan } 1920fa94a07fSbrendan } 1921fa94a07fSbrendan 1922fa94a07fSbrendan /* 1923fa94a07fSbrendan * Stop and drop level 2 ARC devices 1924fa94a07fSbrendan */ 1925fa94a07fSbrendan void 1926fa94a07fSbrendan spa_l2cache_drop(spa_t *spa) 1927fa94a07fSbrendan { 1928fa94a07fSbrendan vdev_t *vd; 1929fa94a07fSbrendan int i; 1930fa94a07fSbrendan spa_aux_vdev_t *sav = &spa->spa_l2cache; 1931fa94a07fSbrendan 1932fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) { 1933fa94a07fSbrendan uint64_t pool; 1934fa94a07fSbrendan 1935fa94a07fSbrendan vd = sav->sav_vdevs[i]; 1936fa94a07fSbrendan ASSERT(vd != NULL); 1937fa94a07fSbrendan 19388ad4d6ddSJeff Bonwick if (spa_l2cache_exists(vd->vdev_guid, &pool) && 19398ad4d6ddSJeff Bonwick pool != 0ULL && l2arc_vdev_present(vd)) 1940fa94a07fSbrendan l2arc_remove_vdev(vd); 1941fa94a07fSbrendan if (vd->vdev_isl2cache) 1942fa94a07fSbrendan spa_l2cache_remove(vd); 1943fa94a07fSbrendan vdev_clear_stats(vd); 1944fa94a07fSbrendan (void) vdev_close(vd); 1945fa94a07fSbrendan } 1946fa94a07fSbrendan } 1947fa94a07fSbrendan 1948fa9e4066Sahrens /* 1949fa9e4066Sahrens * Pool Creation 1950fa9e4066Sahrens */ 1951fa9e4066Sahrens int 1952990b4856Slling spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props, 19530a48a24eStimh const char *history_str, nvlist_t *zplprops) 1954fa9e4066Sahrens { 1955fa9e4066Sahrens spa_t *spa; 1956990b4856Slling char *altroot = NULL; 19570373e76bSbonwick vdev_t *rvd; 1958fa9e4066Sahrens dsl_pool_t *dp; 1959fa9e4066Sahrens dmu_tx_t *tx; 196099653d4eSeschrock int c, error = 0; 1961fa9e4066Sahrens uint64_t txg = TXG_INITIAL; 1962fa94a07fSbrendan nvlist_t **spares, **l2cache; 1963fa94a07fSbrendan uint_t nspares, nl2cache; 1964990b4856Slling uint64_t version; 1965fa9e4066Sahrens 1966fa9e4066Sahrens /* 1967fa9e4066Sahrens * If this pool already exists, return failure. 1968fa9e4066Sahrens */ 1969fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 1970fa9e4066Sahrens if (spa_lookup(pool) != NULL) { 1971fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 1972fa9e4066Sahrens return (EEXIST); 1973fa9e4066Sahrens } 1974fa9e4066Sahrens 1975fa9e4066Sahrens /* 1976fa9e4066Sahrens * Allocate a new spa_t structure. 1977fa9e4066Sahrens */ 1978990b4856Slling (void) nvlist_lookup_string(props, 1979990b4856Slling zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); 19800373e76bSbonwick spa = spa_add(pool, altroot); 19818ad4d6ddSJeff Bonwick spa_activate(spa, spa_mode_global); 1982fa9e4066Sahrens 1983fa9e4066Sahrens spa->spa_uberblock.ub_txg = txg - 1; 1984990b4856Slling 1985990b4856Slling if (props && (error = spa_prop_validate(spa, props))) { 1986990b4856Slling spa_unload(spa); 1987990b4856Slling spa_deactivate(spa); 1988990b4856Slling spa_remove(spa); 1989c5904d13Seschrock mutex_exit(&spa_namespace_lock); 1990990b4856Slling return (error); 1991990b4856Slling } 1992990b4856Slling 1993990b4856Slling if (nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION), 1994990b4856Slling &version) != 0) 1995990b4856Slling version = SPA_VERSION; 1996990b4856Slling ASSERT(version <= SPA_VERSION); 1997990b4856Slling spa->spa_uberblock.ub_version = version; 1998fa9e4066Sahrens spa->spa_ubsync = spa->spa_uberblock; 1999fa9e4066Sahrens 20000373e76bSbonwick /* 20010373e76bSbonwick * Create the root vdev. 20020373e76bSbonwick */ 2003e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 20040373e76bSbonwick 200599653d4eSeschrock error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD); 20060373e76bSbonwick 200799653d4eSeschrock ASSERT(error != 0 || rvd != NULL); 200899653d4eSeschrock ASSERT(error != 0 || spa->spa_root_vdev == rvd); 20090373e76bSbonwick 2010b7b97454Sperrin if (error == 0 && !zfs_allocatable_devs(nvroot)) 20110373e76bSbonwick error = EINVAL; 201299653d4eSeschrock 201399653d4eSeschrock if (error == 0 && 201499653d4eSeschrock (error = vdev_create(rvd, txg, B_FALSE)) == 0 && 2015fa94a07fSbrendan (error = spa_validate_aux(spa, nvroot, txg, 201699653d4eSeschrock VDEV_ALLOC_ADD)) == 0) { 201799653d4eSeschrock for (c = 0; c < rvd->vdev_children; c++) 201899653d4eSeschrock vdev_init(rvd->vdev_child[c], txg); 201999653d4eSeschrock vdev_config_dirty(rvd); 20200373e76bSbonwick } 20210373e76bSbonwick 2022e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2023fa9e4066Sahrens 202499653d4eSeschrock if (error != 0) { 2025fa9e4066Sahrens spa_unload(spa); 2026fa9e4066Sahrens spa_deactivate(spa); 2027fa9e4066Sahrens spa_remove(spa); 2028fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2029fa9e4066Sahrens return (error); 2030fa9e4066Sahrens } 2031fa9e4066Sahrens 203299653d4eSeschrock /* 203399653d4eSeschrock * Get the list of spares, if specified. 203499653d4eSeschrock */ 203599653d4eSeschrock if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 203699653d4eSeschrock &spares, &nspares) == 0) { 2037fa94a07fSbrendan VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME, 203899653d4eSeschrock KM_SLEEP) == 0); 2039fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, 204099653d4eSeschrock ZPOOL_CONFIG_SPARES, spares, nspares) == 0); 2041e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 204299653d4eSeschrock spa_load_spares(spa); 2043e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2044fa94a07fSbrendan spa->spa_spares.sav_sync = B_TRUE; 2045fa94a07fSbrendan } 2046fa94a07fSbrendan 2047fa94a07fSbrendan /* 2048fa94a07fSbrendan * Get the list of level 2 cache devices, if specified. 2049fa94a07fSbrendan */ 2050fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 2051fa94a07fSbrendan &l2cache, &nl2cache) == 0) { 2052fa94a07fSbrendan VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, 2053fa94a07fSbrendan NV_UNIQUE_NAME, KM_SLEEP) == 0); 2054fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, 2055fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); 2056e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 2057fa94a07fSbrendan spa_load_l2cache(spa); 2058e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2059fa94a07fSbrendan spa->spa_l2cache.sav_sync = B_TRUE; 206099653d4eSeschrock } 206199653d4eSeschrock 20620a48a24eStimh spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg); 2063fa9e4066Sahrens spa->spa_meta_objset = dp->dp_meta_objset; 2064fa9e4066Sahrens 2065fa9e4066Sahrens tx = dmu_tx_create_assigned(dp, txg); 2066fa9e4066Sahrens 2067fa9e4066Sahrens /* 2068fa9e4066Sahrens * Create the pool config object. 2069fa9e4066Sahrens */ 2070fa9e4066Sahrens spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset, 2071f7991ba4STim Haley DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE, 2072fa9e4066Sahrens DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx); 2073fa9e4066Sahrens 2074ea8dc4b6Seschrock if (zap_add(spa->spa_meta_objset, 2075fa9e4066Sahrens DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG, 2076ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) { 2077ea8dc4b6Seschrock cmn_err(CE_PANIC, "failed to add pool config"); 2078ea8dc4b6Seschrock } 2079fa9e4066Sahrens 2080990b4856Slling /* Newly created pools with the right version are always deflated. */ 2081990b4856Slling if (version >= SPA_VERSION_RAIDZ_DEFLATE) { 2082990b4856Slling spa->spa_deflate = TRUE; 2083990b4856Slling if (zap_add(spa->spa_meta_objset, 2084990b4856Slling DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, 2085990b4856Slling sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) { 2086990b4856Slling cmn_err(CE_PANIC, "failed to add deflate"); 2087990b4856Slling } 208899653d4eSeschrock } 208999653d4eSeschrock 2090fa9e4066Sahrens /* 2091fa9e4066Sahrens * Create the deferred-free bplist object. Turn off compression 2092fa9e4066Sahrens * because sync-to-convergence takes longer if the blocksize 2093fa9e4066Sahrens * keeps changing. 2094fa9e4066Sahrens */ 2095fa9e4066Sahrens spa->spa_sync_bplist_obj = bplist_create(spa->spa_meta_objset, 2096fa9e4066Sahrens 1 << 14, tx); 2097fa9e4066Sahrens dmu_object_set_compress(spa->spa_meta_objset, spa->spa_sync_bplist_obj, 2098fa9e4066Sahrens ZIO_COMPRESS_OFF, tx); 2099fa9e4066Sahrens 2100ea8dc4b6Seschrock if (zap_add(spa->spa_meta_objset, 2101fa9e4066Sahrens DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPLIST, 2102ea8dc4b6Seschrock sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj, tx) != 0) { 2103ea8dc4b6Seschrock cmn_err(CE_PANIC, "failed to add bplist"); 2104ea8dc4b6Seschrock } 2105fa9e4066Sahrens 210606eeb2adSek /* 210706eeb2adSek * Create the pool's history object. 210806eeb2adSek */ 2109990b4856Slling if (version >= SPA_VERSION_ZPOOL_HISTORY) 2110990b4856Slling spa_history_create_obj(spa, tx); 2111990b4856Slling 2112990b4856Slling /* 2113990b4856Slling * Set pool properties. 2114990b4856Slling */ 2115990b4856Slling spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS); 2116990b4856Slling spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION); 21170a4e9518Sgw spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE); 2118*379c004dSEric Schrock if (props != NULL) { 2119*379c004dSEric Schrock spa_configfile_set(spa, props, B_FALSE); 2120990b4856Slling spa_sync_props(spa, props, CRED(), tx); 2121*379c004dSEric Schrock } 212206eeb2adSek 2123fa9e4066Sahrens dmu_tx_commit(tx); 2124fa9e4066Sahrens 2125fa9e4066Sahrens spa->spa_sync_on = B_TRUE; 2126fa9e4066Sahrens txg_sync_start(spa->spa_dsl_pool); 2127fa9e4066Sahrens 2128fa9e4066Sahrens /* 2129fa9e4066Sahrens * We explicitly wait for the first transaction to complete so that our 2130fa9e4066Sahrens * bean counters are appropriately updated. 2131fa9e4066Sahrens */ 2132fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, txg); 2133fa9e4066Sahrens 2134c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 2135fa9e4066Sahrens 2136990b4856Slling if (version >= SPA_VERSION_ZPOOL_HISTORY && history_str != NULL) 2137228975ccSek (void) spa_history_log(spa, history_str, LOG_CMD_POOL_CREATE); 2138228975ccSek 2139fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2140fa9e4066Sahrens 2141088f3894Sahrens spa->spa_minref = refcount_count(&spa->spa_refcount); 2142088f3894Sahrens 2143fa9e4066Sahrens return (0); 2144fa9e4066Sahrens } 2145fa9e4066Sahrens 2146fa9e4066Sahrens /* 2147fa9e4066Sahrens * Import the given pool into the system. We set up the necessary spa_t and 2148fa9e4066Sahrens * then call spa_load() to do the dirty work. 2149fa9e4066Sahrens */ 2150e7cbe64fSgw static int 2151e7cbe64fSgw spa_import_common(const char *pool, nvlist_t *config, nvlist_t *props, 2152c5904d13Seschrock boolean_t isroot, boolean_t allowfaulted) 2153fa9e4066Sahrens { 2154fa9e4066Sahrens spa_t *spa; 2155990b4856Slling char *altroot = NULL; 2156c5904d13Seschrock int error, loaderr; 215799653d4eSeschrock nvlist_t *nvroot; 2158fa94a07fSbrendan nvlist_t **spares, **l2cache; 2159fa94a07fSbrendan uint_t nspares, nl2cache; 2160fa9e4066Sahrens 2161fa9e4066Sahrens /* 2162fa9e4066Sahrens * If a pool with this name exists, return failure. 2163fa9e4066Sahrens */ 2164fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 216500504c01SLin Ling if ((spa = spa_lookup(pool)) != NULL) { 216600504c01SLin Ling if (isroot) { 216700504c01SLin Ling /* 216800504c01SLin Ling * Remove the existing root pool from the 216900504c01SLin Ling * namespace so that we can replace it with 217000504c01SLin Ling * the correct config we just read in. 217100504c01SLin Ling */ 217200504c01SLin Ling ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED); 217300504c01SLin Ling spa_remove(spa); 217400504c01SLin Ling } else { 217500504c01SLin Ling mutex_exit(&spa_namespace_lock); 217600504c01SLin Ling return (EEXIST); 217700504c01SLin Ling } 2178fa9e4066Sahrens } 2179fa9e4066Sahrens 2180fa9e4066Sahrens /* 21810373e76bSbonwick * Create and initialize the spa structure. 2182fa9e4066Sahrens */ 2183990b4856Slling (void) nvlist_lookup_string(props, 2184990b4856Slling zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot); 21850373e76bSbonwick spa = spa_add(pool, altroot); 21868ad4d6ddSJeff Bonwick spa_activate(spa, spa_mode_global); 2187fa9e4066Sahrens 2188c5904d13Seschrock if (allowfaulted) 2189c5904d13Seschrock spa->spa_import_faulted = B_TRUE; 2190bf82a41bSeschrock spa->spa_is_root = isroot; 2191c5904d13Seschrock 21925dabedeeSbonwick /* 21930373e76bSbonwick * Pass off the heavy lifting to spa_load(). 2194088f3894Sahrens * Pass TRUE for mosconfig (unless this is a root pool) because 2195088f3894Sahrens * the user-supplied config is actually the one to trust when 2196088f3894Sahrens * doing an import. 21975dabedeeSbonwick */ 2198088f3894Sahrens loaderr = error = spa_load(spa, config, SPA_LOAD_IMPORT, !isroot); 2199fa9e4066Sahrens 2200e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 220199653d4eSeschrock /* 220299653d4eSeschrock * Toss any existing sparelist, as it doesn't have any validity anymore, 220399653d4eSeschrock * and conflicts with spa_has_spare(). 220499653d4eSeschrock */ 2205e7cbe64fSgw if (!isroot && spa->spa_spares.sav_config) { 2206fa94a07fSbrendan nvlist_free(spa->spa_spares.sav_config); 2207fa94a07fSbrendan spa->spa_spares.sav_config = NULL; 220899653d4eSeschrock spa_load_spares(spa); 220999653d4eSeschrock } 2210e7cbe64fSgw if (!isroot && spa->spa_l2cache.sav_config) { 2211fa94a07fSbrendan nvlist_free(spa->spa_l2cache.sav_config); 2212fa94a07fSbrendan spa->spa_l2cache.sav_config = NULL; 2213fa94a07fSbrendan spa_load_l2cache(spa); 2214fa94a07fSbrendan } 221599653d4eSeschrock 221699653d4eSeschrock VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 221799653d4eSeschrock &nvroot) == 0); 2218fa94a07fSbrendan if (error == 0) 2219fa94a07fSbrendan error = spa_validate_aux(spa, nvroot, -1ULL, VDEV_ALLOC_SPARE); 2220fa94a07fSbrendan if (error == 0) 2221fa94a07fSbrendan error = spa_validate_aux(spa, nvroot, -1ULL, 2222fa94a07fSbrendan VDEV_ALLOC_L2CACHE); 2223e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 222499653d4eSeschrock 2225*379c004dSEric Schrock if (props != NULL) 2226*379c004dSEric Schrock spa_configfile_set(spa, props, B_FALSE); 2227*379c004dSEric Schrock 22288ad4d6ddSJeff Bonwick if (error != 0 || (props && spa_writeable(spa) && 22298ad4d6ddSJeff Bonwick (error = spa_prop_set(spa, props)))) { 2230c5904d13Seschrock if (loaderr != 0 && loaderr != EINVAL && allowfaulted) { 2231c5904d13Seschrock /* 2232c5904d13Seschrock * If we failed to load the pool, but 'allowfaulted' is 2233c5904d13Seschrock * set, then manually set the config as if the config 2234c5904d13Seschrock * passed in was specified in the cache file. 2235c5904d13Seschrock */ 2236c5904d13Seschrock error = 0; 2237c5904d13Seschrock spa->spa_import_faulted = B_FALSE; 2238e14bb325SJeff Bonwick if (spa->spa_config == NULL) 2239c5904d13Seschrock spa->spa_config = spa_config_generate(spa, 2240c5904d13Seschrock NULL, -1ULL, B_TRUE); 2241c5904d13Seschrock spa_unload(spa); 2242c5904d13Seschrock spa_deactivate(spa); 2243c5904d13Seschrock spa_config_sync(spa, B_FALSE, B_TRUE); 2244c5904d13Seschrock } else { 2245c5904d13Seschrock spa_unload(spa); 2246c5904d13Seschrock spa_deactivate(spa); 2247c5904d13Seschrock spa_remove(spa); 2248c5904d13Seschrock } 2249fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2250fa9e4066Sahrens return (error); 2251fa9e4066Sahrens } 2252fa9e4066Sahrens 225399653d4eSeschrock /* 2254fa94a07fSbrendan * Override any spares and level 2 cache devices as specified by 2255fa94a07fSbrendan * the user, as these may have correct device names/devids, etc. 225699653d4eSeschrock */ 225799653d4eSeschrock if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 225899653d4eSeschrock &spares, &nspares) == 0) { 2259fa94a07fSbrendan if (spa->spa_spares.sav_config) 2260fa94a07fSbrendan VERIFY(nvlist_remove(spa->spa_spares.sav_config, 226199653d4eSeschrock ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0); 226299653d4eSeschrock else 2263fa94a07fSbrendan VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, 226499653d4eSeschrock NV_UNIQUE_NAME, KM_SLEEP) == 0); 2265fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config, 226699653d4eSeschrock ZPOOL_CONFIG_SPARES, spares, nspares) == 0); 2267e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 226899653d4eSeschrock spa_load_spares(spa); 2269e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2270fa94a07fSbrendan spa->spa_spares.sav_sync = B_TRUE; 2271fa94a07fSbrendan } 2272fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 2273fa94a07fSbrendan &l2cache, &nl2cache) == 0) { 2274fa94a07fSbrendan if (spa->spa_l2cache.sav_config) 2275fa94a07fSbrendan VERIFY(nvlist_remove(spa->spa_l2cache.sav_config, 2276fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0); 2277fa94a07fSbrendan else 2278fa94a07fSbrendan VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config, 2279fa94a07fSbrendan NV_UNIQUE_NAME, KM_SLEEP) == 0); 2280fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config, 2281fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0); 2282e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 2283fa94a07fSbrendan spa_load_l2cache(spa); 2284e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2285fa94a07fSbrendan spa->spa_l2cache.sav_sync = B_TRUE; 228699653d4eSeschrock } 228799653d4eSeschrock 22888ad4d6ddSJeff Bonwick if (spa_writeable(spa)) { 2289c5904d13Seschrock /* 2290c5904d13Seschrock * Update the config cache to include the newly-imported pool. 2291c5904d13Seschrock */ 2292e7cbe64fSgw spa_config_update_common(spa, SPA_CONFIG_UPDATE_POOL, isroot); 2293c5904d13Seschrock } 2294fa9e4066Sahrens 2295c5904d13Seschrock spa->spa_import_faulted = B_FALSE; 22963d7072f8Seschrock mutex_exit(&spa_namespace_lock); 22973d7072f8Seschrock 2298fa9e4066Sahrens return (0); 2299fa9e4066Sahrens } 2300fa9e4066Sahrens 2301e7cbe64fSgw #ifdef _KERNEL 2302e7cbe64fSgw /* 2303e7cbe64fSgw * Build a "root" vdev for a top level vdev read in from a rootpool 2304e7cbe64fSgw * device label. 2305e7cbe64fSgw */ 2306e7cbe64fSgw static void 2307e7cbe64fSgw spa_build_rootpool_config(nvlist_t *config) 2308e7cbe64fSgw { 2309e7cbe64fSgw nvlist_t *nvtop, *nvroot; 2310e7cbe64fSgw uint64_t pgid; 2311e7cbe64fSgw 2312e7cbe64fSgw /* 2313e7cbe64fSgw * Add this top-level vdev to the child array. 2314e7cbe64fSgw */ 2315e7cbe64fSgw VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtop) 2316e7cbe64fSgw == 0); 2317e7cbe64fSgw VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pgid) 2318e7cbe64fSgw == 0); 2319e7cbe64fSgw 2320e7cbe64fSgw /* 2321e7cbe64fSgw * Put this pool's top-level vdevs into a root vdev. 2322e7cbe64fSgw */ 2323e7cbe64fSgw VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2324e7cbe64fSgw VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) 2325e7cbe64fSgw == 0); 2326e7cbe64fSgw VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0); 2327e7cbe64fSgw VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0); 2328e7cbe64fSgw VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2329e7cbe64fSgw &nvtop, 1) == 0); 2330e7cbe64fSgw 2331e7cbe64fSgw /* 2332e7cbe64fSgw * Replace the existing vdev_tree with the new root vdev in 2333e7cbe64fSgw * this pool's configuration (remove the old, add the new). 2334e7cbe64fSgw */ 2335e7cbe64fSgw VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); 2336e7cbe64fSgw nvlist_free(nvroot); 2337e7cbe64fSgw } 2338e7cbe64fSgw 2339e7cbe64fSgw /* 2340e7cbe64fSgw * Get the root pool information from the root disk, then import the root pool 2341e7cbe64fSgw * during the system boot up time. 2342e7cbe64fSgw */ 2343f940fbb1SLin Ling extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **); 2344e7cbe64fSgw 2345051aabe6Staylor int 2346051aabe6Staylor spa_check_rootconf(char *devpath, char *devid, nvlist_t **bestconf, 2347e7cbe64fSgw uint64_t *besttxg) 2348e7cbe64fSgw { 2349e7cbe64fSgw nvlist_t *config; 2350e7cbe64fSgw uint64_t txg; 2351f940fbb1SLin Ling int error; 2352e7cbe64fSgw 2353f940fbb1SLin Ling if (error = vdev_disk_read_rootlabel(devpath, devid, &config)) 2354f940fbb1SLin Ling return (error); 2355e7cbe64fSgw 2356e7cbe64fSgw VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0); 2357e7cbe64fSgw 2358051aabe6Staylor if (bestconf != NULL) 2359e7cbe64fSgw *bestconf = config; 2360f940fbb1SLin Ling else 2361f940fbb1SLin Ling nvlist_free(config); 2362051aabe6Staylor *besttxg = txg; 2363051aabe6Staylor return (0); 2364e7cbe64fSgw } 2365e7cbe64fSgw 2366e7cbe64fSgw boolean_t 2367e7cbe64fSgw spa_rootdev_validate(nvlist_t *nv) 2368e7cbe64fSgw { 2369e7cbe64fSgw uint64_t ival; 2370e7cbe64fSgw 2371e7cbe64fSgw if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2372e7cbe64fSgw nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2373e7cbe64fSgw nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2374e7cbe64fSgw return (B_FALSE); 2375e7cbe64fSgw 2376e7cbe64fSgw return (B_TRUE); 2377e7cbe64fSgw } 2378e7cbe64fSgw 2379051aabe6Staylor 2380051aabe6Staylor /* 2381051aabe6Staylor * Given the boot device's physical path or devid, check if the device 2382051aabe6Staylor * is in a valid state. If so, return the configuration from the vdev 2383051aabe6Staylor * label. 2384051aabe6Staylor */ 2385051aabe6Staylor int 2386051aabe6Staylor spa_get_rootconf(char *devpath, char *devid, nvlist_t **bestconf) 2387051aabe6Staylor { 2388051aabe6Staylor nvlist_t *conf = NULL; 2389051aabe6Staylor uint64_t txg = 0; 2390051aabe6Staylor nvlist_t *nvtop, **child; 2391051aabe6Staylor char *type; 2392051aabe6Staylor char *bootpath = NULL; 2393051aabe6Staylor uint_t children, c; 2394051aabe6Staylor char *tmp; 2395f940fbb1SLin Ling int error; 2396051aabe6Staylor 2397051aabe6Staylor if (devpath && ((tmp = strchr(devpath, ' ')) != NULL)) 2398051aabe6Staylor *tmp = '\0'; 2399f940fbb1SLin Ling if (error = spa_check_rootconf(devpath, devid, &conf, &txg)) { 2400051aabe6Staylor cmn_err(CE_NOTE, "error reading device label"); 2401f940fbb1SLin Ling return (error); 2402051aabe6Staylor } 2403051aabe6Staylor if (txg == 0) { 2404051aabe6Staylor cmn_err(CE_NOTE, "this device is detached"); 2405051aabe6Staylor nvlist_free(conf); 2406051aabe6Staylor return (EINVAL); 2407051aabe6Staylor } 2408051aabe6Staylor 2409051aabe6Staylor VERIFY(nvlist_lookup_nvlist(conf, ZPOOL_CONFIG_VDEV_TREE, 2410051aabe6Staylor &nvtop) == 0); 2411051aabe6Staylor VERIFY(nvlist_lookup_string(nvtop, ZPOOL_CONFIG_TYPE, &type) == 0); 2412051aabe6Staylor 2413051aabe6Staylor if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2414051aabe6Staylor if (spa_rootdev_validate(nvtop)) { 2415051aabe6Staylor goto out; 2416051aabe6Staylor } else { 2417051aabe6Staylor nvlist_free(conf); 2418051aabe6Staylor return (EINVAL); 2419051aabe6Staylor } 2420051aabe6Staylor } 2421051aabe6Staylor 2422051aabe6Staylor ASSERT(strcmp(type, VDEV_TYPE_MIRROR) == 0); 2423051aabe6Staylor 2424051aabe6Staylor VERIFY(nvlist_lookup_nvlist_array(nvtop, ZPOOL_CONFIG_CHILDREN, 2425051aabe6Staylor &child, &children) == 0); 2426051aabe6Staylor 2427051aabe6Staylor /* 2428051aabe6Staylor * Go thru vdevs in the mirror to see if the given device 2429051aabe6Staylor * has the most recent txg. Only the device with the most 2430051aabe6Staylor * recent txg has valid information and should be booted. 2431051aabe6Staylor */ 2432051aabe6Staylor for (c = 0; c < children; c++) { 2433051aabe6Staylor char *cdevid, *cpath; 2434051aabe6Staylor uint64_t tmptxg; 2435051aabe6Staylor 2436ffb5616eSLin Ling cpath = NULL; 2437ffb5616eSLin Ling cdevid = NULL; 2438051aabe6Staylor if (nvlist_lookup_string(child[c], ZPOOL_CONFIG_PHYS_PATH, 2439ffb5616eSLin Ling &cpath) != 0 && nvlist_lookup_string(child[c], 2440ffb5616eSLin Ling ZPOOL_CONFIG_DEVID, &cdevid) != 0) 2441051aabe6Staylor return (EINVAL); 2442f4565e39SLin Ling if ((spa_check_rootconf(cpath, cdevid, NULL, 2443f4565e39SLin Ling &tmptxg) == 0) && (tmptxg > txg)) { 2444051aabe6Staylor txg = tmptxg; 2445051aabe6Staylor VERIFY(nvlist_lookup_string(child[c], 2446051aabe6Staylor ZPOOL_CONFIG_PATH, &bootpath) == 0); 2447051aabe6Staylor } 2448051aabe6Staylor } 2449051aabe6Staylor 2450051aabe6Staylor /* Does the best device match the one we've booted from? */ 2451051aabe6Staylor if (bootpath) { 2452051aabe6Staylor cmn_err(CE_NOTE, "try booting from '%s'", bootpath); 2453051aabe6Staylor return (EINVAL); 2454051aabe6Staylor } 2455051aabe6Staylor out: 2456051aabe6Staylor *bestconf = conf; 2457051aabe6Staylor return (0); 2458051aabe6Staylor } 2459051aabe6Staylor 2460e7cbe64fSgw /* 2461e7cbe64fSgw * Import a root pool. 2462e7cbe64fSgw * 2463051aabe6Staylor * For x86. devpath_list will consist of devid and/or physpath name of 2464051aabe6Staylor * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a"). 2465051aabe6Staylor * The GRUB "findroot" command will return the vdev we should boot. 2466e7cbe64fSgw * 2467e7cbe64fSgw * For Sparc, devpath_list consists the physpath name of the booting device 2468e7cbe64fSgw * no matter the rootpool is a single device pool or a mirrored pool. 2469e7cbe64fSgw * e.g. 2470e7cbe64fSgw * "/pci@1f,0/ide@d/disk@0,0:a" 2471e7cbe64fSgw */ 2472e7cbe64fSgw int 2473051aabe6Staylor spa_import_rootpool(char *devpath, char *devid) 2474e7cbe64fSgw { 2475e7cbe64fSgw nvlist_t *conf = NULL; 2476e7cbe64fSgw char *pname; 2477e7cbe64fSgw int error; 2478e7cbe64fSgw 2479e7cbe64fSgw /* 2480e7cbe64fSgw * Get the vdev pathname and configuation from the most 2481e7cbe64fSgw * recently updated vdev (highest txg). 2482e7cbe64fSgw */ 2483051aabe6Staylor if (error = spa_get_rootconf(devpath, devid, &conf)) 2484e7cbe64fSgw goto msg_out; 2485e7cbe64fSgw 2486e7cbe64fSgw /* 2487e7cbe64fSgw * Add type "root" vdev to the config. 2488e7cbe64fSgw */ 2489e7cbe64fSgw spa_build_rootpool_config(conf); 2490e7cbe64fSgw 2491e7cbe64fSgw VERIFY(nvlist_lookup_string(conf, ZPOOL_CONFIG_POOL_NAME, &pname) == 0); 2492e7cbe64fSgw 2493bf82a41bSeschrock /* 2494bf82a41bSeschrock * We specify 'allowfaulted' for this to be treated like spa_open() 2495bf82a41bSeschrock * instead of spa_import(). This prevents us from marking vdevs as 2496bf82a41bSeschrock * persistently unavailable, and generates FMA ereports as if it were a 2497bf82a41bSeschrock * pool open, not import. 2498bf82a41bSeschrock */ 2499bf82a41bSeschrock error = spa_import_common(pname, conf, NULL, B_TRUE, B_TRUE); 250000504c01SLin Ling ASSERT(error != EEXIST); 2501e7cbe64fSgw 2502e7cbe64fSgw nvlist_free(conf); 2503e7cbe64fSgw return (error); 2504e7cbe64fSgw 2505e7cbe64fSgw msg_out: 2506051aabe6Staylor cmn_err(CE_NOTE, "\n" 2507e7cbe64fSgw " *************************************************** \n" 2508e7cbe64fSgw " * This device is not bootable! * \n" 2509e7cbe64fSgw " * It is either offlined or detached or faulted. * \n" 2510e7cbe64fSgw " * Please try to boot from a different device. * \n" 2511051aabe6Staylor " *************************************************** "); 2512e7cbe64fSgw 2513e7cbe64fSgw return (error); 2514e7cbe64fSgw } 2515e7cbe64fSgw #endif 2516e7cbe64fSgw 2517e7cbe64fSgw /* 2518e7cbe64fSgw * Import a non-root pool into the system. 2519e7cbe64fSgw */ 2520e7cbe64fSgw int 2521e7cbe64fSgw spa_import(const char *pool, nvlist_t *config, nvlist_t *props) 2522e7cbe64fSgw { 2523c5904d13Seschrock return (spa_import_common(pool, config, props, B_FALSE, B_FALSE)); 2524e7cbe64fSgw } 2525e7cbe64fSgw 2526c5904d13Seschrock int 2527c5904d13Seschrock spa_import_faulted(const char *pool, nvlist_t *config, nvlist_t *props) 2528c5904d13Seschrock { 2529c5904d13Seschrock return (spa_import_common(pool, config, props, B_FALSE, B_TRUE)); 2530c5904d13Seschrock } 2531c5904d13Seschrock 2532c5904d13Seschrock 2533fa9e4066Sahrens /* 2534fa9e4066Sahrens * This (illegal) pool name is used when temporarily importing a spa_t in order 2535fa9e4066Sahrens * to get the vdev stats associated with the imported devices. 2536fa9e4066Sahrens */ 2537fa9e4066Sahrens #define TRYIMPORT_NAME "$import" 2538fa9e4066Sahrens 2539fa9e4066Sahrens nvlist_t * 2540fa9e4066Sahrens spa_tryimport(nvlist_t *tryconfig) 2541fa9e4066Sahrens { 2542fa9e4066Sahrens nvlist_t *config = NULL; 2543fa9e4066Sahrens char *poolname; 2544fa9e4066Sahrens spa_t *spa; 2545fa9e4066Sahrens uint64_t state; 2546fa9e4066Sahrens 2547fa9e4066Sahrens if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname)) 2548fa9e4066Sahrens return (NULL); 2549fa9e4066Sahrens 2550fa9e4066Sahrens if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state)) 2551fa9e4066Sahrens return (NULL); 2552fa9e4066Sahrens 2553fa9e4066Sahrens /* 25540373e76bSbonwick * Create and initialize the spa structure. 2555fa9e4066Sahrens */ 25560373e76bSbonwick mutex_enter(&spa_namespace_lock); 25570373e76bSbonwick spa = spa_add(TRYIMPORT_NAME, NULL); 25588ad4d6ddSJeff Bonwick spa_activate(spa, FREAD); 2559fa9e4066Sahrens 2560fa9e4066Sahrens /* 25610373e76bSbonwick * Pass off the heavy lifting to spa_load(). 2562ecc2d604Sbonwick * Pass TRUE for mosconfig because the user-supplied config 2563ecc2d604Sbonwick * is actually the one to trust when doing an import. 2564fa9e4066Sahrens */ 2565ecc2d604Sbonwick (void) spa_load(spa, tryconfig, SPA_LOAD_TRYIMPORT, B_TRUE); 2566fa9e4066Sahrens 2567fa9e4066Sahrens /* 2568fa9e4066Sahrens * If 'tryconfig' was at least parsable, return the current config. 2569fa9e4066Sahrens */ 2570fa9e4066Sahrens if (spa->spa_root_vdev != NULL) { 2571fa9e4066Sahrens config = spa_config_generate(spa, NULL, -1ULL, B_TRUE); 2572fa9e4066Sahrens VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, 2573fa9e4066Sahrens poolname) == 0); 2574fa9e4066Sahrens VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, 2575fa9e4066Sahrens state) == 0); 257695173954Sek VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP, 257795173954Sek spa->spa_uberblock.ub_timestamp) == 0); 257899653d4eSeschrock 2579e7cbe64fSgw /* 2580e7cbe64fSgw * If the bootfs property exists on this pool then we 2581e7cbe64fSgw * copy it out so that external consumers can tell which 2582e7cbe64fSgw * pools are bootable. 2583e7cbe64fSgw */ 2584e7cbe64fSgw if (spa->spa_bootfs) { 2585e7cbe64fSgw char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP); 2586e7cbe64fSgw 2587e7cbe64fSgw /* 2588e7cbe64fSgw * We have to play games with the name since the 2589e7cbe64fSgw * pool was opened as TRYIMPORT_NAME. 2590e7cbe64fSgw */ 2591e14bb325SJeff Bonwick if (dsl_dsobj_to_dsname(spa_name(spa), 2592e7cbe64fSgw spa->spa_bootfs, tmpname) == 0) { 2593e7cbe64fSgw char *cp; 2594e7cbe64fSgw char *dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP); 2595e7cbe64fSgw 2596e7cbe64fSgw cp = strchr(tmpname, '/'); 2597e7cbe64fSgw if (cp == NULL) { 2598e7cbe64fSgw (void) strlcpy(dsname, tmpname, 2599e7cbe64fSgw MAXPATHLEN); 2600e7cbe64fSgw } else { 2601e7cbe64fSgw (void) snprintf(dsname, MAXPATHLEN, 2602e7cbe64fSgw "%s/%s", poolname, ++cp); 2603e7cbe64fSgw } 2604e7cbe64fSgw VERIFY(nvlist_add_string(config, 2605e7cbe64fSgw ZPOOL_CONFIG_BOOTFS, dsname) == 0); 2606e7cbe64fSgw kmem_free(dsname, MAXPATHLEN); 2607e7cbe64fSgw } 2608e7cbe64fSgw kmem_free(tmpname, MAXPATHLEN); 2609e7cbe64fSgw } 2610e7cbe64fSgw 261199653d4eSeschrock /* 2612fa94a07fSbrendan * Add the list of hot spares and level 2 cache devices. 261399653d4eSeschrock */ 261499653d4eSeschrock spa_add_spares(spa, config); 2615fa94a07fSbrendan spa_add_l2cache(spa, config); 2616fa9e4066Sahrens } 2617fa9e4066Sahrens 2618fa9e4066Sahrens spa_unload(spa); 2619fa9e4066Sahrens spa_deactivate(spa); 2620fa9e4066Sahrens spa_remove(spa); 2621fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2622fa9e4066Sahrens 2623fa9e4066Sahrens return (config); 2624fa9e4066Sahrens } 2625fa9e4066Sahrens 2626fa9e4066Sahrens /* 2627fa9e4066Sahrens * Pool export/destroy 2628fa9e4066Sahrens * 2629fa9e4066Sahrens * The act of destroying or exporting a pool is very simple. We make sure there 2630fa9e4066Sahrens * is no more pending I/O and any references to the pool are gone. Then, we 2631fa9e4066Sahrens * update the pool state and sync all the labels to disk, removing the 2632394ab0cbSGeorge Wilson * configuration from the cache afterwards. If the 'hardforce' flag is set, then 2633394ab0cbSGeorge Wilson * we don't sync the labels or remove the configuration cache. 2634fa9e4066Sahrens */ 2635fa9e4066Sahrens static int 263689a89ebfSlling spa_export_common(char *pool, int new_state, nvlist_t **oldconfig, 2637394ab0cbSGeorge Wilson boolean_t force, boolean_t hardforce) 2638fa9e4066Sahrens { 2639fa9e4066Sahrens spa_t *spa; 2640fa9e4066Sahrens 264144cd46caSbillm if (oldconfig) 264244cd46caSbillm *oldconfig = NULL; 264344cd46caSbillm 26448ad4d6ddSJeff Bonwick if (!(spa_mode_global & FWRITE)) 2645fa9e4066Sahrens return (EROFS); 2646fa9e4066Sahrens 2647fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 2648fa9e4066Sahrens if ((spa = spa_lookup(pool)) == NULL) { 2649fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2650fa9e4066Sahrens return (ENOENT); 2651fa9e4066Sahrens } 2652fa9e4066Sahrens 2653ea8dc4b6Seschrock /* 2654ea8dc4b6Seschrock * Put a hold on the pool, drop the namespace lock, stop async tasks, 2655ea8dc4b6Seschrock * reacquire the namespace lock, and see if we can export. 2656ea8dc4b6Seschrock */ 2657ea8dc4b6Seschrock spa_open_ref(spa, FTAG); 2658ea8dc4b6Seschrock mutex_exit(&spa_namespace_lock); 2659ea8dc4b6Seschrock spa_async_suspend(spa); 2660ea8dc4b6Seschrock mutex_enter(&spa_namespace_lock); 2661ea8dc4b6Seschrock spa_close(spa, FTAG); 2662ea8dc4b6Seschrock 2663fa9e4066Sahrens /* 2664fa9e4066Sahrens * The pool will be in core if it's openable, 2665fa9e4066Sahrens * in which case we can modify its state. 2666fa9e4066Sahrens */ 2667fa9e4066Sahrens if (spa->spa_state != POOL_STATE_UNINITIALIZED && spa->spa_sync_on) { 2668fa9e4066Sahrens /* 2669fa9e4066Sahrens * Objsets may be open only because they're dirty, so we 2670fa9e4066Sahrens * have to force it to sync before checking spa_refcnt. 2671fa9e4066Sahrens */ 2672fa9e4066Sahrens txg_wait_synced(spa->spa_dsl_pool, 0); 2673fa9e4066Sahrens 2674ea8dc4b6Seschrock /* 2675ea8dc4b6Seschrock * A pool cannot be exported or destroyed if there are active 2676ea8dc4b6Seschrock * references. If we are resetting a pool, allow references by 2677ea8dc4b6Seschrock * fault injection handlers. 2678ea8dc4b6Seschrock */ 2679ea8dc4b6Seschrock if (!spa_refcount_zero(spa) || 2680ea8dc4b6Seschrock (spa->spa_inject_ref != 0 && 2681ea8dc4b6Seschrock new_state != POOL_STATE_UNINITIALIZED)) { 2682ea8dc4b6Seschrock spa_async_resume(spa); 2683fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2684fa9e4066Sahrens return (EBUSY); 2685fa9e4066Sahrens } 2686fa9e4066Sahrens 268789a89ebfSlling /* 268889a89ebfSlling * A pool cannot be exported if it has an active shared spare. 268989a89ebfSlling * This is to prevent other pools stealing the active spare 269089a89ebfSlling * from an exported pool. At user's own will, such pool can 269189a89ebfSlling * be forcedly exported. 269289a89ebfSlling */ 269389a89ebfSlling if (!force && new_state == POOL_STATE_EXPORTED && 269489a89ebfSlling spa_has_active_shared_spare(spa)) { 269589a89ebfSlling spa_async_resume(spa); 269689a89ebfSlling mutex_exit(&spa_namespace_lock); 269789a89ebfSlling return (EXDEV); 269889a89ebfSlling } 269989a89ebfSlling 2700fa9e4066Sahrens /* 2701fa9e4066Sahrens * We want this to be reflected on every label, 2702fa9e4066Sahrens * so mark them all dirty. spa_unload() will do the 2703fa9e4066Sahrens * final sync that pushes these changes out. 2704fa9e4066Sahrens */ 2705394ab0cbSGeorge Wilson if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) { 2706e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 2707ea8dc4b6Seschrock spa->spa_state = new_state; 27080373e76bSbonwick spa->spa_final_txg = spa_last_synced_txg(spa) + 1; 2709ea8dc4b6Seschrock vdev_config_dirty(spa->spa_root_vdev); 2710e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 2711ea8dc4b6Seschrock } 2712fa9e4066Sahrens } 2713fa9e4066Sahrens 27143d7072f8Seschrock spa_event_notify(spa, NULL, ESC_ZFS_POOL_DESTROY); 27153d7072f8Seschrock 2716fa9e4066Sahrens if (spa->spa_state != POOL_STATE_UNINITIALIZED) { 2717fa9e4066Sahrens spa_unload(spa); 2718fa9e4066Sahrens spa_deactivate(spa); 2719fa9e4066Sahrens } 2720fa9e4066Sahrens 272144cd46caSbillm if (oldconfig && spa->spa_config) 272244cd46caSbillm VERIFY(nvlist_dup(spa->spa_config, oldconfig, 0) == 0); 272344cd46caSbillm 2724ea8dc4b6Seschrock if (new_state != POOL_STATE_UNINITIALIZED) { 2725394ab0cbSGeorge Wilson if (!hardforce) 2726394ab0cbSGeorge Wilson spa_config_sync(spa, B_TRUE, B_TRUE); 2727ea8dc4b6Seschrock spa_remove(spa); 2728ea8dc4b6Seschrock } 2729fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 2730fa9e4066Sahrens 2731fa9e4066Sahrens return (0); 2732fa9e4066Sahrens } 2733fa9e4066Sahrens 2734fa9e4066Sahrens /* 2735fa9e4066Sahrens * Destroy a storage pool. 2736fa9e4066Sahrens */ 2737fa9e4066Sahrens int 2738fa9e4066Sahrens spa_destroy(char *pool) 2739fa9e4066Sahrens { 2740394ab0cbSGeorge Wilson return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL, 2741394ab0cbSGeorge Wilson B_FALSE, B_FALSE)); 2742fa9e4066Sahrens } 2743fa9e4066Sahrens 2744fa9e4066Sahrens /* 2745fa9e4066Sahrens * Export a storage pool. 2746fa9e4066Sahrens */ 2747fa9e4066Sahrens int 2748394ab0cbSGeorge Wilson spa_export(char *pool, nvlist_t **oldconfig, boolean_t force, 2749394ab0cbSGeorge Wilson boolean_t hardforce) 2750fa9e4066Sahrens { 2751394ab0cbSGeorge Wilson return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig, 2752394ab0cbSGeorge Wilson force, hardforce)); 2753fa9e4066Sahrens } 2754fa9e4066Sahrens 2755ea8dc4b6Seschrock /* 2756ea8dc4b6Seschrock * Similar to spa_export(), this unloads the spa_t without actually removing it 2757ea8dc4b6Seschrock * from the namespace in any way. 2758ea8dc4b6Seschrock */ 2759ea8dc4b6Seschrock int 2760ea8dc4b6Seschrock spa_reset(char *pool) 2761ea8dc4b6Seschrock { 276289a89ebfSlling return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL, 2763394ab0cbSGeorge Wilson B_FALSE, B_FALSE)); 2764ea8dc4b6Seschrock } 2765ea8dc4b6Seschrock 2766fa9e4066Sahrens /* 2767fa9e4066Sahrens * ========================================================================== 2768fa9e4066Sahrens * Device manipulation 2769fa9e4066Sahrens * ========================================================================== 2770fa9e4066Sahrens */ 2771fa9e4066Sahrens 2772fa9e4066Sahrens /* 27738654d025Sperrin * Add a device to a storage pool. 2774fa9e4066Sahrens */ 2775fa9e4066Sahrens int 2776fa9e4066Sahrens spa_vdev_add(spa_t *spa, nvlist_t *nvroot) 2777fa9e4066Sahrens { 2778fa9e4066Sahrens uint64_t txg; 27798ad4d6ddSJeff Bonwick int error; 2780fa9e4066Sahrens vdev_t *rvd = spa->spa_root_vdev; 27810e34b6a7Sbonwick vdev_t *vd, *tvd; 2782fa94a07fSbrendan nvlist_t **spares, **l2cache; 2783fa94a07fSbrendan uint_t nspares, nl2cache; 2784fa9e4066Sahrens 2785fa9e4066Sahrens txg = spa_vdev_enter(spa); 2786fa9e4066Sahrens 278799653d4eSeschrock if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0, 278899653d4eSeschrock VDEV_ALLOC_ADD)) != 0) 278999653d4eSeschrock return (spa_vdev_exit(spa, NULL, txg, error)); 2790fa9e4066Sahrens 2791e14bb325SJeff Bonwick spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */ 279299653d4eSeschrock 2793fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares, 2794fa94a07fSbrendan &nspares) != 0) 279599653d4eSeschrock nspares = 0; 279699653d4eSeschrock 2797fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache, 2798fa94a07fSbrendan &nl2cache) != 0) 2799fa94a07fSbrendan nl2cache = 0; 2800fa94a07fSbrendan 2801e14bb325SJeff Bonwick if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0) 2802fa9e4066Sahrens return (spa_vdev_exit(spa, vd, txg, EINVAL)); 2803fa9e4066Sahrens 2804e14bb325SJeff Bonwick if (vd->vdev_children != 0 && 2805e14bb325SJeff Bonwick (error = vdev_create(vd, txg, B_FALSE)) != 0) 2806e14bb325SJeff Bonwick return (spa_vdev_exit(spa, vd, txg, error)); 280799653d4eSeschrock 280839c23413Seschrock /* 2809fa94a07fSbrendan * We must validate the spares and l2cache devices after checking the 2810fa94a07fSbrendan * children. Otherwise, vdev_inuse() will blindly overwrite the spare. 281139c23413Seschrock */ 2812e14bb325SJeff Bonwick if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0) 281339c23413Seschrock return (spa_vdev_exit(spa, vd, txg, error)); 281439c23413Seschrock 281539c23413Seschrock /* 281639c23413Seschrock * Transfer each new top-level vdev from vd to rvd. 281739c23413Seschrock */ 28188ad4d6ddSJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) { 281939c23413Seschrock tvd = vd->vdev_child[c]; 282039c23413Seschrock vdev_remove_child(vd, tvd); 282139c23413Seschrock tvd->vdev_id = rvd->vdev_children; 282239c23413Seschrock vdev_add_child(rvd, tvd); 282339c23413Seschrock vdev_config_dirty(tvd); 282439c23413Seschrock } 282539c23413Seschrock 282699653d4eSeschrock if (nspares != 0) { 2827fa94a07fSbrendan spa_set_aux_vdevs(&spa->spa_spares, spares, nspares, 2828fa94a07fSbrendan ZPOOL_CONFIG_SPARES); 282999653d4eSeschrock spa_load_spares(spa); 2830fa94a07fSbrendan spa->spa_spares.sav_sync = B_TRUE; 2831fa94a07fSbrendan } 2832fa94a07fSbrendan 2833fa94a07fSbrendan if (nl2cache != 0) { 2834fa94a07fSbrendan spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache, 2835fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE); 2836fa94a07fSbrendan spa_load_l2cache(spa); 2837fa94a07fSbrendan spa->spa_l2cache.sav_sync = B_TRUE; 2838fa9e4066Sahrens } 2839fa9e4066Sahrens 2840fa9e4066Sahrens /* 28410e34b6a7Sbonwick * We have to be careful when adding new vdevs to an existing pool. 28420e34b6a7Sbonwick * If other threads start allocating from these vdevs before we 28430e34b6a7Sbonwick * sync the config cache, and we lose power, then upon reboot we may 28440e34b6a7Sbonwick * fail to open the pool because there are DVAs that the config cache 28450e34b6a7Sbonwick * can't translate. Therefore, we first add the vdevs without 28460e34b6a7Sbonwick * initializing metaslabs; sync the config cache (via spa_vdev_exit()); 28470373e76bSbonwick * and then let spa_config_update() initialize the new metaslabs. 28480e34b6a7Sbonwick * 28490e34b6a7Sbonwick * spa_load() checks for added-but-not-initialized vdevs, so that 28500e34b6a7Sbonwick * if we lose power at any point in this sequence, the remaining 28510e34b6a7Sbonwick * steps will be completed the next time we load the pool. 28520e34b6a7Sbonwick */ 28530373e76bSbonwick (void) spa_vdev_exit(spa, vd, txg, 0); 28540e34b6a7Sbonwick 28550373e76bSbonwick mutex_enter(&spa_namespace_lock); 28560373e76bSbonwick spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); 28570373e76bSbonwick mutex_exit(&spa_namespace_lock); 2858fa9e4066Sahrens 28590373e76bSbonwick return (0); 2860fa9e4066Sahrens } 2861fa9e4066Sahrens 2862fa9e4066Sahrens /* 2863fa9e4066Sahrens * Attach a device to a mirror. The arguments are the path to any device 2864fa9e4066Sahrens * in the mirror, and the nvroot for the new device. If the path specifies 2865fa9e4066Sahrens * a device that is not mirrored, we automatically insert the mirror vdev. 2866fa9e4066Sahrens * 2867fa9e4066Sahrens * If 'replacing' is specified, the new device is intended to replace the 2868fa9e4066Sahrens * existing device; in this case the two devices are made into their own 28693d7072f8Seschrock * mirror using the 'replacing' vdev, which is functionally identical to 2870fa9e4066Sahrens * the mirror vdev (it actually reuses all the same ops) but has a few 2871fa9e4066Sahrens * extra rules: you can't attach to it after it's been created, and upon 2872fa9e4066Sahrens * completion of resilvering, the first disk (the one being replaced) 2873fa9e4066Sahrens * is automatically detached. 2874fa9e4066Sahrens */ 2875fa9e4066Sahrens int 2876ea8dc4b6Seschrock spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing) 2877fa9e4066Sahrens { 2878fa9e4066Sahrens uint64_t txg, open_txg; 2879fa9e4066Sahrens vdev_t *rvd = spa->spa_root_vdev; 2880fa9e4066Sahrens vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd; 288199653d4eSeschrock vdev_ops_t *pvops; 28829b3f6b42SEric Kustarz dmu_tx_t *tx; 28839b3f6b42SEric Kustarz char *oldvdpath, *newvdpath; 28849b3f6b42SEric Kustarz int newvd_isspare; 28859b3f6b42SEric Kustarz int error; 2886fa9e4066Sahrens 2887fa9e4066Sahrens txg = spa_vdev_enter(spa); 2888fa9e4066Sahrens 2889c5904d13Seschrock oldvd = spa_lookup_by_guid(spa, guid, B_FALSE); 2890fa9e4066Sahrens 2891fa9e4066Sahrens if (oldvd == NULL) 2892fa9e4066Sahrens return (spa_vdev_exit(spa, NULL, txg, ENODEV)); 2893fa9e4066Sahrens 28940e34b6a7Sbonwick if (!oldvd->vdev_ops->vdev_op_leaf) 28950e34b6a7Sbonwick return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 28960e34b6a7Sbonwick 2897fa9e4066Sahrens pvd = oldvd->vdev_parent; 2898fa9e4066Sahrens 289999653d4eSeschrock if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0, 29003d7072f8Seschrock VDEV_ALLOC_ADD)) != 0) 29013d7072f8Seschrock return (spa_vdev_exit(spa, NULL, txg, EINVAL)); 29023d7072f8Seschrock 29033d7072f8Seschrock if (newrootvd->vdev_children != 1) 2904fa9e4066Sahrens return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); 2905fa9e4066Sahrens 2906fa9e4066Sahrens newvd = newrootvd->vdev_child[0]; 2907fa9e4066Sahrens 2908fa9e4066Sahrens if (!newvd->vdev_ops->vdev_op_leaf) 2909fa9e4066Sahrens return (spa_vdev_exit(spa, newrootvd, txg, EINVAL)); 2910fa9e4066Sahrens 291199653d4eSeschrock if ((error = vdev_create(newrootvd, txg, replacing)) != 0) 2912fa9e4066Sahrens return (spa_vdev_exit(spa, newrootvd, txg, error)); 2913fa9e4066Sahrens 29148654d025Sperrin /* 29158654d025Sperrin * Spares can't replace logs 29168654d025Sperrin */ 2917ee0eb9f2SEric Schrock if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare) 29188654d025Sperrin return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); 29198654d025Sperrin 292099653d4eSeschrock if (!replacing) { 292199653d4eSeschrock /* 292299653d4eSeschrock * For attach, the only allowable parent is a mirror or the root 292399653d4eSeschrock * vdev. 292499653d4eSeschrock */ 292599653d4eSeschrock if (pvd->vdev_ops != &vdev_mirror_ops && 292699653d4eSeschrock pvd->vdev_ops != &vdev_root_ops) 292799653d4eSeschrock return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); 292899653d4eSeschrock 292999653d4eSeschrock pvops = &vdev_mirror_ops; 293099653d4eSeschrock } else { 293199653d4eSeschrock /* 293299653d4eSeschrock * Active hot spares can only be replaced by inactive hot 293399653d4eSeschrock * spares. 293499653d4eSeschrock */ 293599653d4eSeschrock if (pvd->vdev_ops == &vdev_spare_ops && 293699653d4eSeschrock pvd->vdev_child[1] == oldvd && 293799653d4eSeschrock !spa_has_spare(spa, newvd->vdev_guid)) 293899653d4eSeschrock return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); 293999653d4eSeschrock 294099653d4eSeschrock /* 294199653d4eSeschrock * If the source is a hot spare, and the parent isn't already a 294299653d4eSeschrock * spare, then we want to create a new hot spare. Otherwise, we 294339c23413Seschrock * want to create a replacing vdev. The user is not allowed to 294439c23413Seschrock * attach to a spared vdev child unless the 'isspare' state is 294539c23413Seschrock * the same (spare replaces spare, non-spare replaces 294639c23413Seschrock * non-spare). 294799653d4eSeschrock */ 294899653d4eSeschrock if (pvd->vdev_ops == &vdev_replacing_ops) 294999653d4eSeschrock return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); 295039c23413Seschrock else if (pvd->vdev_ops == &vdev_spare_ops && 295139c23413Seschrock newvd->vdev_isspare != oldvd->vdev_isspare) 295239c23413Seschrock return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP)); 295399653d4eSeschrock else if (pvd->vdev_ops != &vdev_spare_ops && 295499653d4eSeschrock newvd->vdev_isspare) 295599653d4eSeschrock pvops = &vdev_spare_ops; 295699653d4eSeschrock else 295799653d4eSeschrock pvops = &vdev_replacing_ops; 295899653d4eSeschrock } 295999653d4eSeschrock 29602a79c5feSlling /* 29612a79c5feSlling * Compare the new device size with the replaceable/attachable 29622a79c5feSlling * device size. 29632a79c5feSlling */ 29642a79c5feSlling if (newvd->vdev_psize < vdev_get_rsize(oldvd)) 2965fa9e4066Sahrens return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW)); 2966fa9e4066Sahrens 2967ecc2d604Sbonwick /* 2968ecc2d604Sbonwick * The new device cannot have a higher alignment requirement 2969ecc2d604Sbonwick * than the top-level vdev. 2970ecc2d604Sbonwick */ 2971ecc2d604Sbonwick if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift) 2972fa9e4066Sahrens return (spa_vdev_exit(spa, newrootvd, txg, EDOM)); 2973fa9e4066Sahrens 2974fa9e4066Sahrens /* 2975fa9e4066Sahrens * If this is an in-place replacement, update oldvd's path and devid 2976fa9e4066Sahrens * to make it distinguishable from newvd, and unopenable from now on. 2977fa9e4066Sahrens */ 2978fa9e4066Sahrens if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) { 2979fa9e4066Sahrens spa_strfree(oldvd->vdev_path); 2980fa9e4066Sahrens oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5, 2981fa9e4066Sahrens KM_SLEEP); 2982fa9e4066Sahrens (void) sprintf(oldvd->vdev_path, "%s/%s", 2983fa9e4066Sahrens newvd->vdev_path, "old"); 2984fa9e4066Sahrens if (oldvd->vdev_devid != NULL) { 2985fa9e4066Sahrens spa_strfree(oldvd->vdev_devid); 2986fa9e4066Sahrens oldvd->vdev_devid = NULL; 2987fa9e4066Sahrens } 2988fa9e4066Sahrens } 2989fa9e4066Sahrens 2990fa9e4066Sahrens /* 299199653d4eSeschrock * If the parent is not a mirror, or if we're replacing, insert the new 299299653d4eSeschrock * mirror/replacing/spare vdev above oldvd. 2993fa9e4066Sahrens */ 2994fa9e4066Sahrens if (pvd->vdev_ops != pvops) 2995fa9e4066Sahrens pvd = vdev_add_parent(oldvd, pvops); 2996fa9e4066Sahrens 2997fa9e4066Sahrens ASSERT(pvd->vdev_top->vdev_parent == rvd); 2998fa9e4066Sahrens ASSERT(pvd->vdev_ops == pvops); 2999fa9e4066Sahrens ASSERT(oldvd->vdev_parent == pvd); 3000fa9e4066Sahrens 3001fa9e4066Sahrens /* 3002fa9e4066Sahrens * Extract the new device from its root and add it to pvd. 3003fa9e4066Sahrens */ 3004fa9e4066Sahrens vdev_remove_child(newrootvd, newvd); 3005fa9e4066Sahrens newvd->vdev_id = pvd->vdev_children; 3006fa9e4066Sahrens vdev_add_child(pvd, newvd); 3007fa9e4066Sahrens 3008ea8dc4b6Seschrock /* 3009ea8dc4b6Seschrock * If newvd is smaller than oldvd, but larger than its rsize, 3010ea8dc4b6Seschrock * the addition of newvd may have decreased our parent's asize. 3011ea8dc4b6Seschrock */ 3012ea8dc4b6Seschrock pvd->vdev_asize = MIN(pvd->vdev_asize, newvd->vdev_asize); 3013ea8dc4b6Seschrock 3014fa9e4066Sahrens tvd = newvd->vdev_top; 3015fa9e4066Sahrens ASSERT(pvd->vdev_top == tvd); 3016fa9e4066Sahrens ASSERT(tvd->vdev_parent == rvd); 3017fa9e4066Sahrens 3018fa9e4066Sahrens vdev_config_dirty(tvd); 3019fa9e4066Sahrens 3020fa9e4066Sahrens /* 3021fa9e4066Sahrens * Set newvd's DTL to [TXG_INITIAL, open_txg]. It will propagate 3022fa9e4066Sahrens * upward when spa_vdev_exit() calls vdev_dtl_reassess(). 3023fa9e4066Sahrens */ 3024fa9e4066Sahrens open_txg = txg + TXG_CONCURRENT_STATES - 1; 3025fa9e4066Sahrens 30268ad4d6ddSJeff Bonwick vdev_dtl_dirty(newvd, DTL_MISSING, 30278ad4d6ddSJeff Bonwick TXG_INITIAL, open_txg - TXG_INITIAL + 1); 3028fa9e4066Sahrens 302939c23413Seschrock if (newvd->vdev_isspare) 303039c23413Seschrock spa_spare_activate(newvd); 3031e14bb325SJeff Bonwick oldvdpath = spa_strdup(oldvd->vdev_path); 3032e14bb325SJeff Bonwick newvdpath = spa_strdup(newvd->vdev_path); 30339b3f6b42SEric Kustarz newvd_isspare = newvd->vdev_isspare; 3034ea8dc4b6Seschrock 3035fa9e4066Sahrens /* 3036fa9e4066Sahrens * Mark newvd's DTL dirty in this txg. 3037fa9e4066Sahrens */ 3038ecc2d604Sbonwick vdev_dirty(tvd, VDD_DTL, newvd, txg); 3039fa9e4066Sahrens 3040fa9e4066Sahrens (void) spa_vdev_exit(spa, newrootvd, open_txg, 0); 3041fa9e4066Sahrens 30429b3f6b42SEric Kustarz tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); 30439b3f6b42SEric Kustarz if (dmu_tx_assign(tx, TXG_WAIT) == 0) { 30449b3f6b42SEric Kustarz spa_history_internal_log(LOG_POOL_VDEV_ATTACH, spa, tx, 30459b3f6b42SEric Kustarz CRED(), "%s vdev=%s %s vdev=%s", 30469b3f6b42SEric Kustarz replacing && newvd_isspare ? "spare in" : 30479b3f6b42SEric Kustarz replacing ? "replace" : "attach", newvdpath, 30489b3f6b42SEric Kustarz replacing ? "for" : "to", oldvdpath); 30499b3f6b42SEric Kustarz dmu_tx_commit(tx); 30509b3f6b42SEric Kustarz } else { 30519b3f6b42SEric Kustarz dmu_tx_abort(tx); 30529b3f6b42SEric Kustarz } 30539b3f6b42SEric Kustarz 30549b3f6b42SEric Kustarz spa_strfree(oldvdpath); 30559b3f6b42SEric Kustarz spa_strfree(newvdpath); 30569b3f6b42SEric Kustarz 3057fa9e4066Sahrens /* 3058088f3894Sahrens * Kick off a resilver to update newvd. 3059fa9e4066Sahrens */ 3060088f3894Sahrens VERIFY3U(spa_scrub(spa, POOL_SCRUB_RESILVER), ==, 0); 3061fa9e4066Sahrens 3062fa9e4066Sahrens return (0); 3063fa9e4066Sahrens } 3064fa9e4066Sahrens 3065fa9e4066Sahrens /* 3066fa9e4066Sahrens * Detach a device from a mirror or replacing vdev. 3067fa9e4066Sahrens * If 'replace_done' is specified, only detach if the parent 3068fa9e4066Sahrens * is a replacing vdev. 3069fa9e4066Sahrens */ 3070fa9e4066Sahrens int 30718ad4d6ddSJeff Bonwick spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done) 3072fa9e4066Sahrens { 3073fa9e4066Sahrens uint64_t txg; 30748ad4d6ddSJeff Bonwick int error; 3075fa9e4066Sahrens vdev_t *rvd = spa->spa_root_vdev; 3076fa9e4066Sahrens vdev_t *vd, *pvd, *cvd, *tvd; 307799653d4eSeschrock boolean_t unspare = B_FALSE; 307899653d4eSeschrock uint64_t unspare_guid; 3079bf82a41bSeschrock size_t len; 3080fa9e4066Sahrens 3081fa9e4066Sahrens txg = spa_vdev_enter(spa); 3082fa9e4066Sahrens 3083c5904d13Seschrock vd = spa_lookup_by_guid(spa, guid, B_FALSE); 3084fa9e4066Sahrens 3085fa9e4066Sahrens if (vd == NULL) 3086fa9e4066Sahrens return (spa_vdev_exit(spa, NULL, txg, ENODEV)); 3087fa9e4066Sahrens 30880e34b6a7Sbonwick if (!vd->vdev_ops->vdev_op_leaf) 30890e34b6a7Sbonwick return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 30900e34b6a7Sbonwick 3091fa9e4066Sahrens pvd = vd->vdev_parent; 3092fa9e4066Sahrens 30938ad4d6ddSJeff Bonwick /* 30948ad4d6ddSJeff Bonwick * If the parent/child relationship is not as expected, don't do it. 30958ad4d6ddSJeff Bonwick * Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing 30968ad4d6ddSJeff Bonwick * vdev that's replacing B with C. The user's intent in replacing 30978ad4d6ddSJeff Bonwick * is to go from M(A,B) to M(A,C). If the user decides to cancel 30988ad4d6ddSJeff Bonwick * the replace by detaching C, the expected behavior is to end up 30998ad4d6ddSJeff Bonwick * M(A,B). But suppose that right after deciding to detach C, 31008ad4d6ddSJeff Bonwick * the replacement of B completes. We would have M(A,C), and then 31018ad4d6ddSJeff Bonwick * ask to detach C, which would leave us with just A -- not what 31028ad4d6ddSJeff Bonwick * the user wanted. To prevent this, we make sure that the 31038ad4d6ddSJeff Bonwick * parent/child relationship hasn't changed -- in this example, 31048ad4d6ddSJeff Bonwick * that C's parent is still the replacing vdev R. 31058ad4d6ddSJeff Bonwick */ 31068ad4d6ddSJeff Bonwick if (pvd->vdev_guid != pguid && pguid != 0) 31078ad4d6ddSJeff Bonwick return (spa_vdev_exit(spa, NULL, txg, EBUSY)); 31088ad4d6ddSJeff Bonwick 3109fa9e4066Sahrens /* 3110fa9e4066Sahrens * If replace_done is specified, only remove this device if it's 311199653d4eSeschrock * the first child of a replacing vdev. For the 'spare' vdev, either 311299653d4eSeschrock * disk can be removed. 311399653d4eSeschrock */ 311499653d4eSeschrock if (replace_done) { 311599653d4eSeschrock if (pvd->vdev_ops == &vdev_replacing_ops) { 311699653d4eSeschrock if (vd->vdev_id != 0) 311799653d4eSeschrock return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 311899653d4eSeschrock } else if (pvd->vdev_ops != &vdev_spare_ops) { 311999653d4eSeschrock return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 312099653d4eSeschrock } 312199653d4eSeschrock } 312299653d4eSeschrock 312399653d4eSeschrock ASSERT(pvd->vdev_ops != &vdev_spare_ops || 3124e7437265Sahrens spa_version(spa) >= SPA_VERSION_SPARES); 3125fa9e4066Sahrens 3126fa9e4066Sahrens /* 312799653d4eSeschrock * Only mirror, replacing, and spare vdevs support detach. 3128fa9e4066Sahrens */ 3129fa9e4066Sahrens if (pvd->vdev_ops != &vdev_replacing_ops && 313099653d4eSeschrock pvd->vdev_ops != &vdev_mirror_ops && 313199653d4eSeschrock pvd->vdev_ops != &vdev_spare_ops) 3132fa9e4066Sahrens return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 3133fa9e4066Sahrens 3134fa9e4066Sahrens /* 31358ad4d6ddSJeff Bonwick * If this device has the only valid copy of some data, 31368ad4d6ddSJeff Bonwick * we cannot safely detach it. 3137fa9e4066Sahrens */ 31388ad4d6ddSJeff Bonwick if (vdev_dtl_required(vd)) 3139fa9e4066Sahrens return (spa_vdev_exit(spa, NULL, txg, EBUSY)); 3140fa9e4066Sahrens 31418ad4d6ddSJeff Bonwick ASSERT(pvd->vdev_children >= 2); 3142fa9e4066Sahrens 3143bf82a41bSeschrock /* 3144bf82a41bSeschrock * If we are detaching the second disk from a replacing vdev, then 3145bf82a41bSeschrock * check to see if we changed the original vdev's path to have "/old" 3146bf82a41bSeschrock * at the end in spa_vdev_attach(). If so, undo that change now. 3147bf82a41bSeschrock */ 3148bf82a41bSeschrock if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id == 1 && 3149bf82a41bSeschrock pvd->vdev_child[0]->vdev_path != NULL && 3150bf82a41bSeschrock pvd->vdev_child[1]->vdev_path != NULL) { 3151bf82a41bSeschrock ASSERT(pvd->vdev_child[1] == vd); 3152bf82a41bSeschrock cvd = pvd->vdev_child[0]; 3153bf82a41bSeschrock len = strlen(vd->vdev_path); 3154bf82a41bSeschrock if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 && 3155bf82a41bSeschrock strcmp(cvd->vdev_path + len, "/old") == 0) { 3156bf82a41bSeschrock spa_strfree(cvd->vdev_path); 3157bf82a41bSeschrock cvd->vdev_path = spa_strdup(vd->vdev_path); 3158bf82a41bSeschrock } 3159bf82a41bSeschrock } 3160bf82a41bSeschrock 316199653d4eSeschrock /* 316299653d4eSeschrock * If we are detaching the original disk from a spare, then it implies 316399653d4eSeschrock * that the spare should become a real disk, and be removed from the 316499653d4eSeschrock * active spare list for the pool. 316599653d4eSeschrock */ 316699653d4eSeschrock if (pvd->vdev_ops == &vdev_spare_ops && 31678ad4d6ddSJeff Bonwick vd->vdev_id == 0 && pvd->vdev_child[1]->vdev_isspare) 316899653d4eSeschrock unspare = B_TRUE; 316999653d4eSeschrock 3170fa9e4066Sahrens /* 3171fa9e4066Sahrens * Erase the disk labels so the disk can be used for other things. 3172fa9e4066Sahrens * This must be done after all other error cases are handled, 3173fa9e4066Sahrens * but before we disembowel vd (so we can still do I/O to it). 3174fa9e4066Sahrens * But if we can't do it, don't treat the error as fatal -- 3175fa9e4066Sahrens * it may be that the unwritability of the disk is the reason 3176fa9e4066Sahrens * it's being detached! 3177fa9e4066Sahrens */ 317839c23413Seschrock error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE); 3179fa9e4066Sahrens 3180fa9e4066Sahrens /* 3181fa9e4066Sahrens * Remove vd from its parent and compact the parent's children. 3182fa9e4066Sahrens */ 3183fa9e4066Sahrens vdev_remove_child(pvd, vd); 3184fa9e4066Sahrens vdev_compact_children(pvd); 3185fa9e4066Sahrens 3186fa9e4066Sahrens /* 3187fa9e4066Sahrens * Remember one of the remaining children so we can get tvd below. 3188fa9e4066Sahrens */ 3189fa9e4066Sahrens cvd = pvd->vdev_child[0]; 3190fa9e4066Sahrens 319199653d4eSeschrock /* 319299653d4eSeschrock * If we need to remove the remaining child from the list of hot spares, 31938ad4d6ddSJeff Bonwick * do it now, marking the vdev as no longer a spare in the process. 31948ad4d6ddSJeff Bonwick * We must do this before vdev_remove_parent(), because that can 31958ad4d6ddSJeff Bonwick * change the GUID if it creates a new toplevel GUID. For a similar 31968ad4d6ddSJeff Bonwick * reason, we must remove the spare now, in the same txg as the detach; 31978ad4d6ddSJeff Bonwick * otherwise someone could attach a new sibling, change the GUID, and 31988ad4d6ddSJeff Bonwick * the subsequent attempt to spa_vdev_remove(unspare_guid) would fail. 319999653d4eSeschrock */ 320099653d4eSeschrock if (unspare) { 320199653d4eSeschrock ASSERT(cvd->vdev_isspare); 320239c23413Seschrock spa_spare_remove(cvd); 320399653d4eSeschrock unspare_guid = cvd->vdev_guid; 32048ad4d6ddSJeff Bonwick (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); 320599653d4eSeschrock } 320699653d4eSeschrock 3207fa9e4066Sahrens /* 3208fa9e4066Sahrens * If the parent mirror/replacing vdev only has one child, 3209fa9e4066Sahrens * the parent is no longer needed. Remove it from the tree. 3210fa9e4066Sahrens */ 3211fa9e4066Sahrens if (pvd->vdev_children == 1) 3212fa9e4066Sahrens vdev_remove_parent(cvd); 3213fa9e4066Sahrens 3214fa9e4066Sahrens /* 3215fa9e4066Sahrens * We don't set tvd until now because the parent we just removed 3216fa9e4066Sahrens * may have been the previous top-level vdev. 3217fa9e4066Sahrens */ 3218fa9e4066Sahrens tvd = cvd->vdev_top; 3219fa9e4066Sahrens ASSERT(tvd->vdev_parent == rvd); 3220fa9e4066Sahrens 3221fa9e4066Sahrens /* 322239c23413Seschrock * Reevaluate the parent vdev state. 3223fa9e4066Sahrens */ 32243d7072f8Seschrock vdev_propagate_state(cvd); 3225fa9e4066Sahrens 3226fa9e4066Sahrens /* 322739c23413Seschrock * If the device we just detached was smaller than the others, it may be 322839c23413Seschrock * possible to add metaslabs (i.e. grow the pool). vdev_metaslab_init() 322939c23413Seschrock * can't fail because the existing metaslabs are already in core, so 323039c23413Seschrock * there's nothing to read from disk. 3231fa9e4066Sahrens */ 3232ecc2d604Sbonwick VERIFY(vdev_metaslab_init(tvd, txg) == 0); 3233fa9e4066Sahrens 3234fa9e4066Sahrens vdev_config_dirty(tvd); 3235fa9e4066Sahrens 3236fa9e4066Sahrens /* 323739c23413Seschrock * Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that 323839c23413Seschrock * vd->vdev_detached is set and free vd's DTL object in syncing context. 323939c23413Seschrock * But first make sure we're not on any *other* txg's DTL list, to 324039c23413Seschrock * prevent vd from being accessed after it's freed. 3241fa9e4066Sahrens */ 32428ad4d6ddSJeff Bonwick for (int t = 0; t < TXG_SIZE; t++) 3243fa9e4066Sahrens (void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t); 3244ecc2d604Sbonwick vd->vdev_detached = B_TRUE; 3245ecc2d604Sbonwick vdev_dirty(tvd, VDD_DTL, vd, txg); 3246fa9e4066Sahrens 32473d7072f8Seschrock spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE); 32483d7072f8Seschrock 324999653d4eSeschrock error = spa_vdev_exit(spa, vd, txg, 0); 325099653d4eSeschrock 325199653d4eSeschrock /* 325239c23413Seschrock * If this was the removal of the original device in a hot spare vdev, 325339c23413Seschrock * then we want to go through and remove the device from the hot spare 325439c23413Seschrock * list of every other pool. 325599653d4eSeschrock */ 325699653d4eSeschrock if (unspare) { 32578ad4d6ddSJeff Bonwick spa_t *myspa = spa; 325899653d4eSeschrock spa = NULL; 325999653d4eSeschrock mutex_enter(&spa_namespace_lock); 326099653d4eSeschrock while ((spa = spa_next(spa)) != NULL) { 326199653d4eSeschrock if (spa->spa_state != POOL_STATE_ACTIVE) 326299653d4eSeschrock continue; 32638ad4d6ddSJeff Bonwick if (spa == myspa) 32648ad4d6ddSJeff Bonwick continue; 32659af0a4dfSJeff Bonwick spa_open_ref(spa, FTAG); 32669af0a4dfSJeff Bonwick mutex_exit(&spa_namespace_lock); 326799653d4eSeschrock (void) spa_vdev_remove(spa, unspare_guid, B_TRUE); 32689af0a4dfSJeff Bonwick mutex_enter(&spa_namespace_lock); 32699af0a4dfSJeff Bonwick spa_close(spa, FTAG); 327099653d4eSeschrock } 327199653d4eSeschrock mutex_exit(&spa_namespace_lock); 327299653d4eSeschrock } 327399653d4eSeschrock 327499653d4eSeschrock return (error); 327599653d4eSeschrock } 327699653d4eSeschrock 3277e14bb325SJeff Bonwick static nvlist_t * 3278e14bb325SJeff Bonwick spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid) 327999653d4eSeschrock { 3280e14bb325SJeff Bonwick for (int i = 0; i < count; i++) { 3281e14bb325SJeff Bonwick uint64_t guid; 328299653d4eSeschrock 3283e14bb325SJeff Bonwick VERIFY(nvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID, 3284e14bb325SJeff Bonwick &guid) == 0); 328599653d4eSeschrock 3286e14bb325SJeff Bonwick if (guid == target_guid) 3287e14bb325SJeff Bonwick return (nvpp[i]); 328899653d4eSeschrock } 328999653d4eSeschrock 3290e14bb325SJeff Bonwick return (NULL); 3291fa94a07fSbrendan } 3292fa94a07fSbrendan 3293e14bb325SJeff Bonwick static void 3294e14bb325SJeff Bonwick spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count, 3295e14bb325SJeff Bonwick nvlist_t *dev_to_remove) 3296fa94a07fSbrendan { 3297e14bb325SJeff Bonwick nvlist_t **newdev = NULL; 3298fa94a07fSbrendan 3299e14bb325SJeff Bonwick if (count > 1) 3300e14bb325SJeff Bonwick newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP); 3301fa94a07fSbrendan 3302e14bb325SJeff Bonwick for (int i = 0, j = 0; i < count; i++) { 3303e14bb325SJeff Bonwick if (dev[i] == dev_to_remove) 3304e14bb325SJeff Bonwick continue; 3305e14bb325SJeff Bonwick VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0); 3306fa94a07fSbrendan } 3307fa94a07fSbrendan 3308e14bb325SJeff Bonwick VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0); 3309e14bb325SJeff Bonwick VERIFY(nvlist_add_nvlist_array(config, name, newdev, count - 1) == 0); 3310fa94a07fSbrendan 3311e14bb325SJeff Bonwick for (int i = 0; i < count - 1; i++) 3312e14bb325SJeff Bonwick nvlist_free(newdev[i]); 3313fa94a07fSbrendan 3314e14bb325SJeff Bonwick if (count > 1) 3315e14bb325SJeff Bonwick kmem_free(newdev, (count - 1) * sizeof (void *)); 3316fa94a07fSbrendan } 3317fa94a07fSbrendan 3318fa94a07fSbrendan /* 3319fa94a07fSbrendan * Remove a device from the pool. Currently, this supports removing only hot 3320fa94a07fSbrendan * spares and level 2 ARC devices. 3321fa94a07fSbrendan */ 3322fa94a07fSbrendan int 3323fa94a07fSbrendan spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare) 3324fa94a07fSbrendan { 3325fa94a07fSbrendan vdev_t *vd; 3326e14bb325SJeff Bonwick nvlist_t **spares, **l2cache, *nv; 3327fa94a07fSbrendan uint_t nspares, nl2cache; 33288ad4d6ddSJeff Bonwick uint64_t txg = 0; 3329fa94a07fSbrendan int error = 0; 33308ad4d6ddSJeff Bonwick boolean_t locked = MUTEX_HELD(&spa_namespace_lock); 3331fa94a07fSbrendan 33328ad4d6ddSJeff Bonwick if (!locked) 33338ad4d6ddSJeff Bonwick txg = spa_vdev_enter(spa); 3334fa94a07fSbrendan 3335c5904d13Seschrock vd = spa_lookup_by_guid(spa, guid, B_FALSE); 3336fa94a07fSbrendan 3337fa94a07fSbrendan if (spa->spa_spares.sav_vdevs != NULL && 3338fa94a07fSbrendan nvlist_lookup_nvlist_array(spa->spa_spares.sav_config, 3339e14bb325SJeff Bonwick ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 && 3340e14bb325SJeff Bonwick (nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) { 3341e14bb325SJeff Bonwick /* 3342e14bb325SJeff Bonwick * Only remove the hot spare if it's not currently in use 3343e14bb325SJeff Bonwick * in this pool. 3344e14bb325SJeff Bonwick */ 3345e14bb325SJeff Bonwick if (vd == NULL || unspare) { 3346e14bb325SJeff Bonwick spa_vdev_remove_aux(spa->spa_spares.sav_config, 3347e14bb325SJeff Bonwick ZPOOL_CONFIG_SPARES, spares, nspares, nv); 3348e14bb325SJeff Bonwick spa_load_spares(spa); 3349e14bb325SJeff Bonwick spa->spa_spares.sav_sync = B_TRUE; 3350e14bb325SJeff Bonwick } else { 3351e14bb325SJeff Bonwick error = EBUSY; 3352e14bb325SJeff Bonwick } 3353e14bb325SJeff Bonwick } else if (spa->spa_l2cache.sav_vdevs != NULL && 3354fa94a07fSbrendan nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config, 3355e14bb325SJeff Bonwick ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 && 3356e14bb325SJeff Bonwick (nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) { 3357e14bb325SJeff Bonwick /* 3358e14bb325SJeff Bonwick * Cache devices can always be removed. 3359e14bb325SJeff Bonwick */ 3360e14bb325SJeff Bonwick spa_vdev_remove_aux(spa->spa_l2cache.sav_config, 3361e14bb325SJeff Bonwick ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv); 3362fa94a07fSbrendan spa_load_l2cache(spa); 3363fa94a07fSbrendan spa->spa_l2cache.sav_sync = B_TRUE; 3364e14bb325SJeff Bonwick } else if (vd != NULL) { 3365e14bb325SJeff Bonwick /* 3366e14bb325SJeff Bonwick * Normal vdevs cannot be removed (yet). 3367e14bb325SJeff Bonwick */ 3368e14bb325SJeff Bonwick error = ENOTSUP; 3369e14bb325SJeff Bonwick } else { 3370e14bb325SJeff Bonwick /* 3371e14bb325SJeff Bonwick * There is no vdev of any kind with the specified guid. 3372e14bb325SJeff Bonwick */ 3373e14bb325SJeff Bonwick error = ENOENT; 3374fa94a07fSbrendan } 337599653d4eSeschrock 33768ad4d6ddSJeff Bonwick if (!locked) 33778ad4d6ddSJeff Bonwick return (spa_vdev_exit(spa, NULL, txg, error)); 33788ad4d6ddSJeff Bonwick 33798ad4d6ddSJeff Bonwick return (error); 3380fa9e4066Sahrens } 3381fa9e4066Sahrens 3382fa9e4066Sahrens /* 33833d7072f8Seschrock * Find any device that's done replacing, or a vdev marked 'unspare' that's 33843d7072f8Seschrock * current spared, so we can detach it. 3385fa9e4066Sahrens */ 3386ea8dc4b6Seschrock static vdev_t * 33873d7072f8Seschrock spa_vdev_resilver_done_hunt(vdev_t *vd) 3388fa9e4066Sahrens { 3389ea8dc4b6Seschrock vdev_t *newvd, *oldvd; 3390fa9e4066Sahrens int c; 3391fa9e4066Sahrens 3392ea8dc4b6Seschrock for (c = 0; c < vd->vdev_children; c++) { 33933d7072f8Seschrock oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]); 3394ea8dc4b6Seschrock if (oldvd != NULL) 3395ea8dc4b6Seschrock return (oldvd); 3396ea8dc4b6Seschrock } 3397fa9e4066Sahrens 33983d7072f8Seschrock /* 33993d7072f8Seschrock * Check for a completed replacement. 34003d7072f8Seschrock */ 3401fa9e4066Sahrens if (vd->vdev_ops == &vdev_replacing_ops && vd->vdev_children == 2) { 3402ea8dc4b6Seschrock oldvd = vd->vdev_child[0]; 3403ea8dc4b6Seschrock newvd = vd->vdev_child[1]; 3404ea8dc4b6Seschrock 34058ad4d6ddSJeff Bonwick if (vdev_dtl_empty(newvd, DTL_MISSING) && 34068ad4d6ddSJeff Bonwick !vdev_dtl_required(oldvd)) 3407ea8dc4b6Seschrock return (oldvd); 3408fa9e4066Sahrens } 3409ea8dc4b6Seschrock 34103d7072f8Seschrock /* 34113d7072f8Seschrock * Check for a completed resilver with the 'unspare' flag set. 34123d7072f8Seschrock */ 34133d7072f8Seschrock if (vd->vdev_ops == &vdev_spare_ops && vd->vdev_children == 2) { 34143d7072f8Seschrock newvd = vd->vdev_child[0]; 34153d7072f8Seschrock oldvd = vd->vdev_child[1]; 34163d7072f8Seschrock 34173d7072f8Seschrock if (newvd->vdev_unspare && 34188ad4d6ddSJeff Bonwick vdev_dtl_empty(newvd, DTL_MISSING) && 34198ad4d6ddSJeff Bonwick !vdev_dtl_required(oldvd)) { 34203d7072f8Seschrock newvd->vdev_unspare = 0; 34213d7072f8Seschrock return (oldvd); 34223d7072f8Seschrock } 34233d7072f8Seschrock } 34243d7072f8Seschrock 3425ea8dc4b6Seschrock return (NULL); 3426fa9e4066Sahrens } 3427fa9e4066Sahrens 3428ea8dc4b6Seschrock static void 34293d7072f8Seschrock spa_vdev_resilver_done(spa_t *spa) 3430fa9e4066Sahrens { 34318ad4d6ddSJeff Bonwick vdev_t *vd, *pvd, *ppvd; 34328ad4d6ddSJeff Bonwick uint64_t guid, sguid, pguid, ppguid; 3433ea8dc4b6Seschrock 34348ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 3435ea8dc4b6Seschrock 34363d7072f8Seschrock while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) { 34378ad4d6ddSJeff Bonwick pvd = vd->vdev_parent; 34388ad4d6ddSJeff Bonwick ppvd = pvd->vdev_parent; 3439ea8dc4b6Seschrock guid = vd->vdev_guid; 34408ad4d6ddSJeff Bonwick pguid = pvd->vdev_guid; 34418ad4d6ddSJeff Bonwick ppguid = ppvd->vdev_guid; 34428ad4d6ddSJeff Bonwick sguid = 0; 344399653d4eSeschrock /* 344499653d4eSeschrock * If we have just finished replacing a hot spared device, then 344599653d4eSeschrock * we need to detach the parent's first child (the original hot 344699653d4eSeschrock * spare) as well. 344799653d4eSeschrock */ 34488ad4d6ddSJeff Bonwick if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0) { 344999653d4eSeschrock ASSERT(pvd->vdev_ops == &vdev_replacing_ops); 34508ad4d6ddSJeff Bonwick ASSERT(ppvd->vdev_children == 2); 34518ad4d6ddSJeff Bonwick sguid = ppvd->vdev_child[1]->vdev_guid; 345299653d4eSeschrock } 34538ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 34548ad4d6ddSJeff Bonwick if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0) 3455ea8dc4b6Seschrock return; 34568ad4d6ddSJeff Bonwick if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0) 345799653d4eSeschrock return; 34588ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 3459fa9e4066Sahrens } 3460fa9e4066Sahrens 34618ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 3462fa9e4066Sahrens } 3463fa9e4066Sahrens 3464c67d9675Seschrock /* 3465c67d9675Seschrock * Update the stored path for this vdev. Dirty the vdev configuration, relying 3466c67d9675Seschrock * on spa_vdev_enter/exit() to synchronize the labels and cache. 3467c67d9675Seschrock */ 3468c67d9675Seschrock int 3469c67d9675Seschrock spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath) 3470c67d9675Seschrock { 3471c5904d13Seschrock vdev_t *vd; 3472c67d9675Seschrock uint64_t txg; 3473c67d9675Seschrock 3474c67d9675Seschrock txg = spa_vdev_enter(spa); 3475c67d9675Seschrock 3476c5904d13Seschrock if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL) { 347799653d4eSeschrock /* 3478c5904d13Seschrock * Determine if this is a reference to a hot spare device. If 3479c5904d13Seschrock * it is, update the path manually as there is no associated 3480c5904d13Seschrock * vdev_t that can be synced to disk. 348199653d4eSeschrock */ 3482c5904d13Seschrock nvlist_t **spares; 3483c5904d13Seschrock uint_t i, nspares; 3484fa94a07fSbrendan 3485fa94a07fSbrendan if (spa->spa_spares.sav_config != NULL) { 3486fa94a07fSbrendan VERIFY(nvlist_lookup_nvlist_array( 3487fa94a07fSbrendan spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES, 3488fa94a07fSbrendan &spares, &nspares) == 0); 348999653d4eSeschrock for (i = 0; i < nspares; i++) { 349099653d4eSeschrock uint64_t theguid; 349199653d4eSeschrock VERIFY(nvlist_lookup_uint64(spares[i], 349299653d4eSeschrock ZPOOL_CONFIG_GUID, &theguid) == 0); 3493fa94a07fSbrendan if (theguid == guid) { 3494fa94a07fSbrendan VERIFY(nvlist_add_string(spares[i], 3495fa94a07fSbrendan ZPOOL_CONFIG_PATH, newpath) == 0); 3496fa94a07fSbrendan spa_load_spares(spa); 3497fa94a07fSbrendan spa->spa_spares.sav_sync = B_TRUE; 3498fa94a07fSbrendan return (spa_vdev_exit(spa, NULL, txg, 3499fa94a07fSbrendan 0)); 3500fa94a07fSbrendan } 350199653d4eSeschrock } 3502fa94a07fSbrendan } 350399653d4eSeschrock 3504fa94a07fSbrendan return (spa_vdev_exit(spa, NULL, txg, ENOENT)); 350599653d4eSeschrock } 3506c67d9675Seschrock 35070e34b6a7Sbonwick if (!vd->vdev_ops->vdev_op_leaf) 35080e34b6a7Sbonwick return (spa_vdev_exit(spa, NULL, txg, ENOTSUP)); 35090e34b6a7Sbonwick 3510c67d9675Seschrock spa_strfree(vd->vdev_path); 3511c67d9675Seschrock vd->vdev_path = spa_strdup(newpath); 3512c67d9675Seschrock 3513c67d9675Seschrock vdev_config_dirty(vd->vdev_top); 3514c67d9675Seschrock 3515c67d9675Seschrock return (spa_vdev_exit(spa, NULL, txg, 0)); 3516c67d9675Seschrock } 3517c67d9675Seschrock 3518fa9e4066Sahrens /* 3519fa9e4066Sahrens * ========================================================================== 3520fa9e4066Sahrens * SPA Scrubbing 3521fa9e4066Sahrens * ========================================================================== 3522fa9e4066Sahrens */ 3523fa9e4066Sahrens 3524ea8dc4b6Seschrock int 3525088f3894Sahrens spa_scrub(spa_t *spa, pool_scrub_type_t type) 3526fa9e4066Sahrens { 3527e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); 3528bb8b5132Sek 3529fa9e4066Sahrens if ((uint_t)type >= POOL_SCRUB_TYPES) 3530fa9e4066Sahrens return (ENOTSUP); 3531fa9e4066Sahrens 3532fa9e4066Sahrens /* 3533088f3894Sahrens * If a resilver was requested, but there is no DTL on a 3534088f3894Sahrens * writeable leaf device, we have nothing to do. 3535fa9e4066Sahrens */ 3536088f3894Sahrens if (type == POOL_SCRUB_RESILVER && 3537088f3894Sahrens !vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) { 3538088f3894Sahrens spa_async_request(spa, SPA_ASYNC_RESILVER_DONE); 3539ea8dc4b6Seschrock return (0); 3540ea8dc4b6Seschrock } 3541fa9e4066Sahrens 3542088f3894Sahrens if (type == POOL_SCRUB_EVERYTHING && 3543088f3894Sahrens spa->spa_dsl_pool->dp_scrub_func != SCRUB_FUNC_NONE && 3544088f3894Sahrens spa->spa_dsl_pool->dp_scrub_isresilver) 3545088f3894Sahrens return (EBUSY); 3546fa9e4066Sahrens 3547088f3894Sahrens if (type == POOL_SCRUB_EVERYTHING || type == POOL_SCRUB_RESILVER) { 3548088f3894Sahrens return (dsl_pool_scrub_clean(spa->spa_dsl_pool)); 3549088f3894Sahrens } else if (type == POOL_SCRUB_NONE) { 3550088f3894Sahrens return (dsl_pool_scrub_cancel(spa->spa_dsl_pool)); 3551ea8dc4b6Seschrock } else { 3552088f3894Sahrens return (EINVAL); 3553fa9e4066Sahrens } 3554fa9e4066Sahrens } 3555fa9e4066Sahrens 3556ea8dc4b6Seschrock /* 3557ea8dc4b6Seschrock * ========================================================================== 3558ea8dc4b6Seschrock * SPA async task processing 3559ea8dc4b6Seschrock * ========================================================================== 3560ea8dc4b6Seschrock */ 3561ea8dc4b6Seschrock 3562ea8dc4b6Seschrock static void 35633d7072f8Seschrock spa_async_remove(spa_t *spa, vdev_t *vd) 3564fa9e4066Sahrens { 356549cf58c0SBrendan Gregg - Sun Microsystems if (vd->vdev_remove_wanted) { 356649cf58c0SBrendan Gregg - Sun Microsystems vd->vdev_remove_wanted = 0; 356749cf58c0SBrendan Gregg - Sun Microsystems vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE); 3568e14bb325SJeff Bonwick vdev_clear(spa, vd); 3569e14bb325SJeff Bonwick vdev_state_dirty(vd->vdev_top); 3570ea8dc4b6Seschrock } 357149cf58c0SBrendan Gregg - Sun Microsystems 3572e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 357349cf58c0SBrendan Gregg - Sun Microsystems spa_async_remove(spa, vd->vdev_child[c]); 3574ea8dc4b6Seschrock } 3575fa9e4066Sahrens 3576e14bb325SJeff Bonwick static void 3577e14bb325SJeff Bonwick spa_async_probe(spa_t *spa, vdev_t *vd) 3578e14bb325SJeff Bonwick { 3579e14bb325SJeff Bonwick if (vd->vdev_probe_wanted) { 3580e14bb325SJeff Bonwick vd->vdev_probe_wanted = 0; 3581e14bb325SJeff Bonwick vdev_reopen(vd); /* vdev_open() does the actual probe */ 3582e14bb325SJeff Bonwick } 3583e14bb325SJeff Bonwick 3584e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 3585e14bb325SJeff Bonwick spa_async_probe(spa, vd->vdev_child[c]); 3586e14bb325SJeff Bonwick } 3587e14bb325SJeff Bonwick 3588ea8dc4b6Seschrock static void 3589ea8dc4b6Seschrock spa_async_thread(spa_t *spa) 3590ea8dc4b6Seschrock { 3591e14bb325SJeff Bonwick int tasks; 3592ea8dc4b6Seschrock 3593ea8dc4b6Seschrock ASSERT(spa->spa_sync_on); 3594ea8dc4b6Seschrock 3595ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3596ea8dc4b6Seschrock tasks = spa->spa_async_tasks; 3597ea8dc4b6Seschrock spa->spa_async_tasks = 0; 3598ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3599ea8dc4b6Seschrock 36000373e76bSbonwick /* 36010373e76bSbonwick * See if the config needs to be updated. 36020373e76bSbonwick */ 36030373e76bSbonwick if (tasks & SPA_ASYNC_CONFIG_UPDATE) { 36040373e76bSbonwick mutex_enter(&spa_namespace_lock); 36050373e76bSbonwick spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); 36060373e76bSbonwick mutex_exit(&spa_namespace_lock); 36070373e76bSbonwick } 36080373e76bSbonwick 3609ea8dc4b6Seschrock /* 36103d7072f8Seschrock * See if any devices need to be marked REMOVED. 3611ea8dc4b6Seschrock */ 3612e14bb325SJeff Bonwick if (tasks & SPA_ASYNC_REMOVE) { 3613e14bb325SJeff Bonwick spa_vdev_state_enter(spa); 36143d7072f8Seschrock spa_async_remove(spa, spa->spa_root_vdev); 3615e14bb325SJeff Bonwick for (int i = 0; i < spa->spa_l2cache.sav_count; i++) 361649cf58c0SBrendan Gregg - Sun Microsystems spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]); 3617e14bb325SJeff Bonwick for (int i = 0; i < spa->spa_spares.sav_count; i++) 361849cf58c0SBrendan Gregg - Sun Microsystems spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]); 3619e14bb325SJeff Bonwick (void) spa_vdev_state_exit(spa, NULL, 0); 3620e14bb325SJeff Bonwick } 3621e14bb325SJeff Bonwick 3622e14bb325SJeff Bonwick /* 3623e14bb325SJeff Bonwick * See if any devices need to be probed. 3624e14bb325SJeff Bonwick */ 3625e14bb325SJeff Bonwick if (tasks & SPA_ASYNC_PROBE) { 3626e14bb325SJeff Bonwick spa_vdev_state_enter(spa); 3627e14bb325SJeff Bonwick spa_async_probe(spa, spa->spa_root_vdev); 3628e14bb325SJeff Bonwick (void) spa_vdev_state_exit(spa, NULL, 0); 36293d7072f8Seschrock } 3630ea8dc4b6Seschrock 3631ea8dc4b6Seschrock /* 3632ea8dc4b6Seschrock * If any devices are done replacing, detach them. 3633ea8dc4b6Seschrock */ 36343d7072f8Seschrock if (tasks & SPA_ASYNC_RESILVER_DONE) 36353d7072f8Seschrock spa_vdev_resilver_done(spa); 3636fa9e4066Sahrens 3637ea8dc4b6Seschrock /* 3638ea8dc4b6Seschrock * Kick off a resilver. 3639ea8dc4b6Seschrock */ 3640088f3894Sahrens if (tasks & SPA_ASYNC_RESILVER) 3641088f3894Sahrens VERIFY(spa_scrub(spa, POOL_SCRUB_RESILVER) == 0); 3642ea8dc4b6Seschrock 3643ea8dc4b6Seschrock /* 3644ea8dc4b6Seschrock * Let the world know that we're done. 3645ea8dc4b6Seschrock */ 3646ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3647ea8dc4b6Seschrock spa->spa_async_thread = NULL; 3648ea8dc4b6Seschrock cv_broadcast(&spa->spa_async_cv); 3649ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3650ea8dc4b6Seschrock thread_exit(); 3651ea8dc4b6Seschrock } 3652ea8dc4b6Seschrock 3653ea8dc4b6Seschrock void 3654ea8dc4b6Seschrock spa_async_suspend(spa_t *spa) 3655ea8dc4b6Seschrock { 3656ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3657ea8dc4b6Seschrock spa->spa_async_suspended++; 3658ea8dc4b6Seschrock while (spa->spa_async_thread != NULL) 3659ea8dc4b6Seschrock cv_wait(&spa->spa_async_cv, &spa->spa_async_lock); 3660ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3661ea8dc4b6Seschrock } 3662ea8dc4b6Seschrock 3663ea8dc4b6Seschrock void 3664ea8dc4b6Seschrock spa_async_resume(spa_t *spa) 3665ea8dc4b6Seschrock { 3666ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3667ea8dc4b6Seschrock ASSERT(spa->spa_async_suspended != 0); 3668ea8dc4b6Seschrock spa->spa_async_suspended--; 3669ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3670ea8dc4b6Seschrock } 3671ea8dc4b6Seschrock 3672ea8dc4b6Seschrock static void 3673ea8dc4b6Seschrock spa_async_dispatch(spa_t *spa) 3674ea8dc4b6Seschrock { 3675ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3676ea8dc4b6Seschrock if (spa->spa_async_tasks && !spa->spa_async_suspended && 36770373e76bSbonwick spa->spa_async_thread == NULL && 36780373e76bSbonwick rootdir != NULL && !vn_is_readonly(rootdir)) 3679ea8dc4b6Seschrock spa->spa_async_thread = thread_create(NULL, 0, 3680ea8dc4b6Seschrock spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri); 3681ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3682ea8dc4b6Seschrock } 3683ea8dc4b6Seschrock 3684ea8dc4b6Seschrock void 3685ea8dc4b6Seschrock spa_async_request(spa_t *spa, int task) 3686ea8dc4b6Seschrock { 3687ea8dc4b6Seschrock mutex_enter(&spa->spa_async_lock); 3688ea8dc4b6Seschrock spa->spa_async_tasks |= task; 3689ea8dc4b6Seschrock mutex_exit(&spa->spa_async_lock); 3690fa9e4066Sahrens } 3691fa9e4066Sahrens 3692fa9e4066Sahrens /* 3693fa9e4066Sahrens * ========================================================================== 3694fa9e4066Sahrens * SPA syncing routines 3695fa9e4066Sahrens * ========================================================================== 3696fa9e4066Sahrens */ 3697fa9e4066Sahrens 3698fa9e4066Sahrens static void 3699fa9e4066Sahrens spa_sync_deferred_frees(spa_t *spa, uint64_t txg) 3700fa9e4066Sahrens { 3701fa9e4066Sahrens bplist_t *bpl = &spa->spa_sync_bplist; 3702fa9e4066Sahrens dmu_tx_t *tx; 3703fa9e4066Sahrens blkptr_t blk; 3704fa9e4066Sahrens uint64_t itor = 0; 3705fa9e4066Sahrens zio_t *zio; 3706fa9e4066Sahrens int error; 3707fa9e4066Sahrens uint8_t c = 1; 3708fa9e4066Sahrens 3709e14bb325SJeff Bonwick zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); 3710fa9e4066Sahrens 3711e14bb325SJeff Bonwick while (bplist_iterate(bpl, &itor, &blk) == 0) { 3712e14bb325SJeff Bonwick ASSERT(blk.blk_birth < txg); 3713e14bb325SJeff Bonwick zio_nowait(zio_free(zio, spa, txg, &blk, NULL, NULL, 3714e14bb325SJeff Bonwick ZIO_FLAG_MUSTSUCCEED)); 3715e14bb325SJeff Bonwick } 3716fa9e4066Sahrens 3717fa9e4066Sahrens error = zio_wait(zio); 3718fa9e4066Sahrens ASSERT3U(error, ==, 0); 3719fa9e4066Sahrens 3720fa9e4066Sahrens tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg); 3721fa9e4066Sahrens bplist_vacate(bpl, tx); 3722fa9e4066Sahrens 3723fa9e4066Sahrens /* 3724fa9e4066Sahrens * Pre-dirty the first block so we sync to convergence faster. 3725fa9e4066Sahrens * (Usually only the first block is needed.) 3726fa9e4066Sahrens */ 3727fa9e4066Sahrens dmu_write(spa->spa_meta_objset, spa->spa_sync_bplist_obj, 0, 1, &c, tx); 3728fa9e4066Sahrens dmu_tx_commit(tx); 3729fa9e4066Sahrens } 3730fa9e4066Sahrens 3731fa9e4066Sahrens static void 373299653d4eSeschrock spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) 3733fa9e4066Sahrens { 3734fa9e4066Sahrens char *packed = NULL; 3735f7991ba4STim Haley size_t bufsize; 3736fa9e4066Sahrens size_t nvsize = 0; 3737fa9e4066Sahrens dmu_buf_t *db; 3738fa9e4066Sahrens 373999653d4eSeschrock VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0); 3740fa9e4066Sahrens 3741f7991ba4STim Haley /* 3742f7991ba4STim Haley * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration 3743f7991ba4STim Haley * information. This avoids the dbuf_will_dirty() path and 3744f7991ba4STim Haley * saves us a pre-read to get data we don't actually care about. 3745f7991ba4STim Haley */ 3746f7991ba4STim Haley bufsize = P2ROUNDUP(nvsize, SPA_CONFIG_BLOCKSIZE); 3747f7991ba4STim Haley packed = kmem_alloc(bufsize, KM_SLEEP); 3748fa9e4066Sahrens 374999653d4eSeschrock VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, 3750ea8dc4b6Seschrock KM_SLEEP) == 0); 3751f7991ba4STim Haley bzero(packed + nvsize, bufsize - nvsize); 3752fa9e4066Sahrens 3753f7991ba4STim Haley dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx); 3754fa9e4066Sahrens 3755f7991ba4STim Haley kmem_free(packed, bufsize); 3756fa9e4066Sahrens 375799653d4eSeschrock VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); 3758fa9e4066Sahrens dmu_buf_will_dirty(db, tx); 3759fa9e4066Sahrens *(uint64_t *)db->db_data = nvsize; 3760ea8dc4b6Seschrock dmu_buf_rele(db, FTAG); 3761fa9e4066Sahrens } 3762fa9e4066Sahrens 376399653d4eSeschrock static void 3764fa94a07fSbrendan spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx, 3765fa94a07fSbrendan const char *config, const char *entry) 376699653d4eSeschrock { 376799653d4eSeschrock nvlist_t *nvroot; 3768fa94a07fSbrendan nvlist_t **list; 376999653d4eSeschrock int i; 377099653d4eSeschrock 3771fa94a07fSbrendan if (!sav->sav_sync) 377299653d4eSeschrock return; 377399653d4eSeschrock 377499653d4eSeschrock /* 3775fa94a07fSbrendan * Update the MOS nvlist describing the list of available devices. 3776fa94a07fSbrendan * spa_validate_aux() will have already made sure this nvlist is 37773d7072f8Seschrock * valid and the vdevs are labeled appropriately. 377899653d4eSeschrock */ 3779fa94a07fSbrendan if (sav->sav_object == 0) { 3780fa94a07fSbrendan sav->sav_object = dmu_object_alloc(spa->spa_meta_objset, 3781fa94a07fSbrendan DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE, 3782fa94a07fSbrendan sizeof (uint64_t), tx); 378399653d4eSeschrock VERIFY(zap_update(spa->spa_meta_objset, 3784fa94a07fSbrendan DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1, 3785fa94a07fSbrendan &sav->sav_object, tx) == 0); 378699653d4eSeschrock } 378799653d4eSeschrock 378899653d4eSeschrock VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0); 3789fa94a07fSbrendan if (sav->sav_count == 0) { 3790fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(nvroot, config, NULL, 0) == 0); 379199653d4eSeschrock } else { 3792fa94a07fSbrendan list = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP); 3793fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) 3794fa94a07fSbrendan list[i] = vdev_config_generate(spa, sav->sav_vdevs[i], 3795fa94a07fSbrendan B_FALSE, B_FALSE, B_TRUE); 3796fa94a07fSbrendan VERIFY(nvlist_add_nvlist_array(nvroot, config, list, 3797fa94a07fSbrendan sav->sav_count) == 0); 3798fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) 3799fa94a07fSbrendan nvlist_free(list[i]); 3800fa94a07fSbrendan kmem_free(list, sav->sav_count * sizeof (void *)); 380199653d4eSeschrock } 380299653d4eSeschrock 3803fa94a07fSbrendan spa_sync_nvlist(spa, sav->sav_object, nvroot, tx); 380406eeb2adSek nvlist_free(nvroot); 380599653d4eSeschrock 3806fa94a07fSbrendan sav->sav_sync = B_FALSE; 380799653d4eSeschrock } 380899653d4eSeschrock 380999653d4eSeschrock static void 381099653d4eSeschrock spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) 381199653d4eSeschrock { 381299653d4eSeschrock nvlist_t *config; 381399653d4eSeschrock 3814e14bb325SJeff Bonwick if (list_is_empty(&spa->spa_config_dirty_list)) 381599653d4eSeschrock return; 381699653d4eSeschrock 3817e14bb325SJeff Bonwick spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 3818e14bb325SJeff Bonwick 3819e14bb325SJeff Bonwick config = spa_config_generate(spa, spa->spa_root_vdev, 3820e14bb325SJeff Bonwick dmu_tx_get_txg(tx), B_FALSE); 3821e14bb325SJeff Bonwick 3822e14bb325SJeff Bonwick spa_config_exit(spa, SCL_STATE, FTAG); 382399653d4eSeschrock 382499653d4eSeschrock if (spa->spa_config_syncing) 382599653d4eSeschrock nvlist_free(spa->spa_config_syncing); 382699653d4eSeschrock spa->spa_config_syncing = config; 382799653d4eSeschrock 382899653d4eSeschrock spa_sync_nvlist(spa, spa->spa_config_object, config, tx); 382999653d4eSeschrock } 383099653d4eSeschrock 3831990b4856Slling /* 3832990b4856Slling * Set zpool properties. 3833990b4856Slling */ 3834b1b8ab34Slling static void 3835ecd6cf80Smarks spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx) 3836b1b8ab34Slling { 3837b1b8ab34Slling spa_t *spa = arg1; 3838b1b8ab34Slling objset_t *mos = spa->spa_meta_objset; 3839990b4856Slling nvlist_t *nvp = arg2; 3840990b4856Slling nvpair_t *elem; 38413d7072f8Seschrock uint64_t intval; 3842c5904d13Seschrock char *strval; 3843990b4856Slling zpool_prop_t prop; 3844990b4856Slling const char *propname; 3845990b4856Slling zprop_type_t proptype; 3846b1b8ab34Slling 3847e14bb325SJeff Bonwick mutex_enter(&spa->spa_props_lock); 3848e14bb325SJeff Bonwick 3849990b4856Slling elem = NULL; 3850990b4856Slling while ((elem = nvlist_next_nvpair(nvp, elem))) { 3851990b4856Slling switch (prop = zpool_name_to_prop(nvpair_name(elem))) { 3852990b4856Slling case ZPOOL_PROP_VERSION: 3853990b4856Slling /* 3854990b4856Slling * Only set version for non-zpool-creation cases 3855990b4856Slling * (set/import). spa_create() needs special care 3856990b4856Slling * for version setting. 3857990b4856Slling */ 3858990b4856Slling if (tx->tx_txg != TXG_INITIAL) { 3859990b4856Slling VERIFY(nvpair_value_uint64(elem, 3860990b4856Slling &intval) == 0); 3861990b4856Slling ASSERT(intval <= SPA_VERSION); 3862990b4856Slling ASSERT(intval >= spa_version(spa)); 3863990b4856Slling spa->spa_uberblock.ub_version = intval; 3864990b4856Slling vdev_config_dirty(spa->spa_root_vdev); 3865990b4856Slling } 3866ecd6cf80Smarks break; 3867990b4856Slling 3868990b4856Slling case ZPOOL_PROP_ALTROOT: 3869990b4856Slling /* 3870990b4856Slling * 'altroot' is a non-persistent property. It should 3871990b4856Slling * have been set temporarily at creation or import time. 3872990b4856Slling */ 3873990b4856Slling ASSERT(spa->spa_root != NULL); 3874b1b8ab34Slling break; 38753d7072f8Seschrock 38762f8aaab3Seschrock case ZPOOL_PROP_CACHEFILE: 3877990b4856Slling /* 3878*379c004dSEric Schrock * 'cachefile' is also a non-persisitent property. 3879990b4856Slling */ 38803d7072f8Seschrock break; 3881990b4856Slling default: 3882990b4856Slling /* 3883990b4856Slling * Set pool property values in the poolprops mos object. 3884990b4856Slling */ 3885990b4856Slling if (spa->spa_pool_props_object == 0) { 3886990b4856Slling objset_t *mos = spa->spa_meta_objset; 3887990b4856Slling 3888990b4856Slling VERIFY((spa->spa_pool_props_object = 3889990b4856Slling zap_create(mos, DMU_OT_POOL_PROPS, 3890990b4856Slling DMU_OT_NONE, 0, tx)) > 0); 3891990b4856Slling 3892990b4856Slling VERIFY(zap_update(mos, 3893990b4856Slling DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS, 3894990b4856Slling 8, 1, &spa->spa_pool_props_object, tx) 3895990b4856Slling == 0); 3896990b4856Slling } 3897990b4856Slling 3898990b4856Slling /* normalize the property name */ 3899990b4856Slling propname = zpool_prop_to_name(prop); 3900990b4856Slling proptype = zpool_prop_get_type(prop); 3901990b4856Slling 3902990b4856Slling if (nvpair_type(elem) == DATA_TYPE_STRING) { 3903990b4856Slling ASSERT(proptype == PROP_TYPE_STRING); 3904990b4856Slling VERIFY(nvpair_value_string(elem, &strval) == 0); 3905990b4856Slling VERIFY(zap_update(mos, 3906990b4856Slling spa->spa_pool_props_object, propname, 3907990b4856Slling 1, strlen(strval) + 1, strval, tx) == 0); 3908990b4856Slling 3909990b4856Slling } else if (nvpair_type(elem) == DATA_TYPE_UINT64) { 3910990b4856Slling VERIFY(nvpair_value_uint64(elem, &intval) == 0); 3911990b4856Slling 3912990b4856Slling if (proptype == PROP_TYPE_INDEX) { 3913990b4856Slling const char *unused; 3914990b4856Slling VERIFY(zpool_prop_index_to_string( 3915990b4856Slling prop, intval, &unused) == 0); 3916990b4856Slling } 3917990b4856Slling VERIFY(zap_update(mos, 3918990b4856Slling spa->spa_pool_props_object, propname, 3919990b4856Slling 8, 1, &intval, tx) == 0); 3920990b4856Slling } else { 3921990b4856Slling ASSERT(0); /* not allowed */ 3922990b4856Slling } 3923990b4856Slling 39240a4e9518Sgw switch (prop) { 39250a4e9518Sgw case ZPOOL_PROP_DELEGATION: 3926990b4856Slling spa->spa_delegation = intval; 39270a4e9518Sgw break; 39280a4e9518Sgw case ZPOOL_PROP_BOOTFS: 3929990b4856Slling spa->spa_bootfs = intval; 39300a4e9518Sgw break; 39310a4e9518Sgw case ZPOOL_PROP_FAILUREMODE: 39320a4e9518Sgw spa->spa_failmode = intval; 39330a4e9518Sgw break; 39340a4e9518Sgw default: 39350a4e9518Sgw break; 39360a4e9518Sgw } 3937990b4856Slling } 3938990b4856Slling 3939990b4856Slling /* log internal history if this is not a zpool create */ 3940990b4856Slling if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY && 3941990b4856Slling tx->tx_txg != TXG_INITIAL) { 3942990b4856Slling spa_history_internal_log(LOG_POOL_PROPSET, 3943990b4856Slling spa, tx, cr, "%s %lld %s", 3944e14bb325SJeff Bonwick nvpair_name(elem), intval, spa_name(spa)); 3945b1b8ab34Slling } 3946b1b8ab34Slling } 3947e14bb325SJeff Bonwick 3948e14bb325SJeff Bonwick mutex_exit(&spa->spa_props_lock); 3949b1b8ab34Slling } 3950b1b8ab34Slling 3951fa9e4066Sahrens /* 3952fa9e4066Sahrens * Sync the specified transaction group. New blocks may be dirtied as 3953fa9e4066Sahrens * part of the process, so we iterate until it converges. 3954fa9e4066Sahrens */ 3955fa9e4066Sahrens void 3956fa9e4066Sahrens spa_sync(spa_t *spa, uint64_t txg) 3957fa9e4066Sahrens { 3958fa9e4066Sahrens dsl_pool_t *dp = spa->spa_dsl_pool; 3959fa9e4066Sahrens objset_t *mos = spa->spa_meta_objset; 3960fa9e4066Sahrens bplist_t *bpl = &spa->spa_sync_bplist; 39610373e76bSbonwick vdev_t *rvd = spa->spa_root_vdev; 3962fa9e4066Sahrens vdev_t *vd; 3963fa9e4066Sahrens dmu_tx_t *tx; 3964fa9e4066Sahrens int dirty_vdevs; 3965e14bb325SJeff Bonwick int error; 3966fa9e4066Sahrens 3967fa9e4066Sahrens /* 3968fa9e4066Sahrens * Lock out configuration changes. 3969fa9e4066Sahrens */ 3970e14bb325SJeff Bonwick spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); 3971fa9e4066Sahrens 3972fa9e4066Sahrens spa->spa_syncing_txg = txg; 3973fa9e4066Sahrens spa->spa_sync_pass = 0; 3974fa9e4066Sahrens 3975e14bb325SJeff Bonwick /* 3976e14bb325SJeff Bonwick * If there are any pending vdev state changes, convert them 3977e14bb325SJeff Bonwick * into config changes that go out with this transaction group. 3978e14bb325SJeff Bonwick */ 3979e14bb325SJeff Bonwick spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 39808ad4d6ddSJeff Bonwick while (list_head(&spa->spa_state_dirty_list) != NULL) { 39818ad4d6ddSJeff Bonwick /* 39828ad4d6ddSJeff Bonwick * We need the write lock here because, for aux vdevs, 39838ad4d6ddSJeff Bonwick * calling vdev_config_dirty() modifies sav_config. 39848ad4d6ddSJeff Bonwick * This is ugly and will become unnecessary when we 39858ad4d6ddSJeff Bonwick * eliminate the aux vdev wart by integrating all vdevs 39868ad4d6ddSJeff Bonwick * into the root vdev tree. 39878ad4d6ddSJeff Bonwick */ 39888ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); 39898ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_WRITER); 39908ad4d6ddSJeff Bonwick while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) { 39918ad4d6ddSJeff Bonwick vdev_state_clean(vd); 39928ad4d6ddSJeff Bonwick vdev_config_dirty(vd); 39938ad4d6ddSJeff Bonwick } 39948ad4d6ddSJeff Bonwick spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); 39958ad4d6ddSJeff Bonwick spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); 3996e14bb325SJeff Bonwick } 3997e14bb325SJeff Bonwick spa_config_exit(spa, SCL_STATE, FTAG); 3998e14bb325SJeff Bonwick 3999ea8dc4b6Seschrock VERIFY(0 == bplist_open(bpl, mos, spa->spa_sync_bplist_obj)); 4000fa9e4066Sahrens 400199653d4eSeschrock tx = dmu_tx_create_assigned(dp, txg); 400299653d4eSeschrock 400399653d4eSeschrock /* 4004e7437265Sahrens * If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg, 400599653d4eSeschrock * set spa_deflate if we have no raid-z vdevs. 400699653d4eSeschrock */ 4007e7437265Sahrens if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE && 4008e7437265Sahrens spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) { 400999653d4eSeschrock int i; 401099653d4eSeschrock 401199653d4eSeschrock for (i = 0; i < rvd->vdev_children; i++) { 401299653d4eSeschrock vd = rvd->vdev_child[i]; 401399653d4eSeschrock if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE) 401499653d4eSeschrock break; 401599653d4eSeschrock } 401699653d4eSeschrock if (i == rvd->vdev_children) { 401799653d4eSeschrock spa->spa_deflate = TRUE; 401899653d4eSeschrock VERIFY(0 == zap_add(spa->spa_meta_objset, 401999653d4eSeschrock DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE, 402099653d4eSeschrock sizeof (uint64_t), 1, &spa->spa_deflate, tx)); 402199653d4eSeschrock } 402299653d4eSeschrock } 402399653d4eSeschrock 4024088f3894Sahrens if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN && 4025088f3894Sahrens spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) { 4026088f3894Sahrens dsl_pool_create_origin(dp, tx); 4027088f3894Sahrens 4028088f3894Sahrens /* Keeping the origin open increases spa_minref */ 4029088f3894Sahrens spa->spa_minref += 3; 4030088f3894Sahrens } 4031088f3894Sahrens 4032088f3894Sahrens if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES && 4033088f3894Sahrens spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) { 4034088f3894Sahrens dsl_pool_upgrade_clones(dp, tx); 4035088f3894Sahrens } 4036088f3894Sahrens 4037fa9e4066Sahrens /* 4038fa9e4066Sahrens * If anything has changed in this txg, push the deferred frees 4039fa9e4066Sahrens * from the previous txg. If not, leave them alone so that we 4040fa9e4066Sahrens * don't generate work on an otherwise idle system. 4041fa9e4066Sahrens */ 4042fa9e4066Sahrens if (!txg_list_empty(&dp->dp_dirty_datasets, txg) || 40431615a317Sek !txg_list_empty(&dp->dp_dirty_dirs, txg) || 40441615a317Sek !txg_list_empty(&dp->dp_sync_tasks, txg)) 4045fa9e4066Sahrens spa_sync_deferred_frees(spa, txg); 4046fa9e4066Sahrens 4047fa9e4066Sahrens /* 4048fa9e4066Sahrens * Iterate to convergence. 4049fa9e4066Sahrens */ 4050fa9e4066Sahrens do { 4051fa9e4066Sahrens spa->spa_sync_pass++; 4052fa9e4066Sahrens 4053fa9e4066Sahrens spa_sync_config_object(spa, tx); 4054fa94a07fSbrendan spa_sync_aux_dev(spa, &spa->spa_spares, tx, 4055fa94a07fSbrendan ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES); 4056fa94a07fSbrendan spa_sync_aux_dev(spa, &spa->spa_l2cache, tx, 4057fa94a07fSbrendan ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE); 4058ea8dc4b6Seschrock spa_errlog_sync(spa, txg); 4059fa9e4066Sahrens dsl_pool_sync(dp, txg); 4060fa9e4066Sahrens 4061fa9e4066Sahrens dirty_vdevs = 0; 4062fa9e4066Sahrens while (vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) { 4063fa9e4066Sahrens vdev_sync(vd, txg); 4064fa9e4066Sahrens dirty_vdevs++; 4065fa9e4066Sahrens } 4066fa9e4066Sahrens 4067fa9e4066Sahrens bplist_sync(bpl, tx); 4068fa9e4066Sahrens } while (dirty_vdevs); 4069fa9e4066Sahrens 4070fa9e4066Sahrens bplist_close(bpl); 4071fa9e4066Sahrens 4072fa9e4066Sahrens dprintf("txg %llu passes %d\n", txg, spa->spa_sync_pass); 4073fa9e4066Sahrens 4074fa9e4066Sahrens /* 4075fa9e4066Sahrens * Rewrite the vdev configuration (which includes the uberblock) 4076fa9e4066Sahrens * to commit the transaction group. 40770373e76bSbonwick * 407817f17c2dSbonwick * If there are no dirty vdevs, we sync the uberblock to a few 407917f17c2dSbonwick * random top-level vdevs that are known to be visible in the 4080e14bb325SJeff Bonwick * config cache (see spa_vdev_add() for a complete description). 4081e14bb325SJeff Bonwick * If there *are* dirty vdevs, sync the uberblock to all vdevs. 40820373e76bSbonwick */ 4083e14bb325SJeff Bonwick for (;;) { 4084e14bb325SJeff Bonwick /* 4085e14bb325SJeff Bonwick * We hold SCL_STATE to prevent vdev open/close/etc. 4086e14bb325SJeff Bonwick * while we're attempting to write the vdev labels. 4087e14bb325SJeff Bonwick */ 4088e14bb325SJeff Bonwick spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 4089e14bb325SJeff Bonwick 4090e14bb325SJeff Bonwick if (list_is_empty(&spa->spa_config_dirty_list)) { 4091e14bb325SJeff Bonwick vdev_t *svd[SPA_DVAS_PER_BP]; 4092e14bb325SJeff Bonwick int svdcount = 0; 4093e14bb325SJeff Bonwick int children = rvd->vdev_children; 4094e14bb325SJeff Bonwick int c0 = spa_get_random(children); 4095e14bb325SJeff Bonwick int c; 4096e14bb325SJeff Bonwick 4097e14bb325SJeff Bonwick for (c = 0; c < children; c++) { 4098e14bb325SJeff Bonwick vd = rvd->vdev_child[(c0 + c) % children]; 4099e14bb325SJeff Bonwick if (vd->vdev_ms_array == 0 || vd->vdev_islog) 4100e14bb325SJeff Bonwick continue; 4101e14bb325SJeff Bonwick svd[svdcount++] = vd; 4102e14bb325SJeff Bonwick if (svdcount == SPA_DVAS_PER_BP) 4103e14bb325SJeff Bonwick break; 4104e14bb325SJeff Bonwick } 4105e14bb325SJeff Bonwick error = vdev_config_sync(svd, svdcount, txg); 4106e14bb325SJeff Bonwick } else { 4107e14bb325SJeff Bonwick error = vdev_config_sync(rvd->vdev_child, 4108e14bb325SJeff Bonwick rvd->vdev_children, txg); 41090373e76bSbonwick } 4110e14bb325SJeff Bonwick 4111e14bb325SJeff Bonwick spa_config_exit(spa, SCL_STATE, FTAG); 4112e14bb325SJeff Bonwick 4113e14bb325SJeff Bonwick if (error == 0) 4114e14bb325SJeff Bonwick break; 4115e14bb325SJeff Bonwick zio_suspend(spa, NULL); 4116e14bb325SJeff Bonwick zio_resume_wait(spa); 41170373e76bSbonwick } 411899653d4eSeschrock dmu_tx_commit(tx); 411999653d4eSeschrock 41200373e76bSbonwick /* 41210373e76bSbonwick * Clear the dirty config list. 4122fa9e4066Sahrens */ 4123e14bb325SJeff Bonwick while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL) 41240373e76bSbonwick vdev_config_clean(vd); 41250373e76bSbonwick 41260373e76bSbonwick /* 41270373e76bSbonwick * Now that the new config has synced transactionally, 41280373e76bSbonwick * let it become visible to the config cache. 41290373e76bSbonwick */ 41300373e76bSbonwick if (spa->spa_config_syncing != NULL) { 41310373e76bSbonwick spa_config_set(spa, spa->spa_config_syncing); 41320373e76bSbonwick spa->spa_config_txg = txg; 41330373e76bSbonwick spa->spa_config_syncing = NULL; 41340373e76bSbonwick } 4135fa9e4066Sahrens 4136fa9e4066Sahrens spa->spa_ubsync = spa->spa_uberblock; 4137fa9e4066Sahrens 4138fa9e4066Sahrens /* 4139fa9e4066Sahrens * Clean up the ZIL records for the synced txg. 4140fa9e4066Sahrens */ 4141fa9e4066Sahrens dsl_pool_zil_clean(dp); 4142fa9e4066Sahrens 4143fa9e4066Sahrens /* 4144fa9e4066Sahrens * Update usable space statistics. 4145fa9e4066Sahrens */ 4146fa9e4066Sahrens while (vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg))) 4147fa9e4066Sahrens vdev_sync_done(vd, txg); 4148fa9e4066Sahrens 4149fa9e4066Sahrens /* 4150fa9e4066Sahrens * It had better be the case that we didn't dirty anything 415199653d4eSeschrock * since vdev_config_sync(). 4152fa9e4066Sahrens */ 4153fa9e4066Sahrens ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg)); 4154fa9e4066Sahrens ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg)); 4155fa9e4066Sahrens ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg)); 4156fa9e4066Sahrens ASSERT(bpl->bpl_queue == NULL); 4157fa9e4066Sahrens 4158e14bb325SJeff Bonwick spa_config_exit(spa, SCL_CONFIG, FTAG); 4159ea8dc4b6Seschrock 4160ea8dc4b6Seschrock /* 4161ea8dc4b6Seschrock * If any async tasks have been requested, kick them off. 4162ea8dc4b6Seschrock */ 4163ea8dc4b6Seschrock spa_async_dispatch(spa); 4164fa9e4066Sahrens } 4165fa9e4066Sahrens 4166fa9e4066Sahrens /* 4167fa9e4066Sahrens * Sync all pools. We don't want to hold the namespace lock across these 4168fa9e4066Sahrens * operations, so we take a reference on the spa_t and drop the lock during the 4169fa9e4066Sahrens * sync. 4170fa9e4066Sahrens */ 4171fa9e4066Sahrens void 4172fa9e4066Sahrens spa_sync_allpools(void) 4173fa9e4066Sahrens { 4174fa9e4066Sahrens spa_t *spa = NULL; 4175fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 4176fa9e4066Sahrens while ((spa = spa_next(spa)) != NULL) { 4177e14bb325SJeff Bonwick if (spa_state(spa) != POOL_STATE_ACTIVE || spa_suspended(spa)) 4178fa9e4066Sahrens continue; 4179fa9e4066Sahrens spa_open_ref(spa, FTAG); 4180fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 4181fa9e4066Sahrens txg_wait_synced(spa_get_dsl(spa), 0); 4182fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 4183fa9e4066Sahrens spa_close(spa, FTAG); 4184fa9e4066Sahrens } 4185fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 4186fa9e4066Sahrens } 4187fa9e4066Sahrens 4188fa9e4066Sahrens /* 4189fa9e4066Sahrens * ========================================================================== 4190fa9e4066Sahrens * Miscellaneous routines 4191fa9e4066Sahrens * ========================================================================== 4192fa9e4066Sahrens */ 4193fa9e4066Sahrens 4194fa9e4066Sahrens /* 4195fa9e4066Sahrens * Remove all pools in the system. 4196fa9e4066Sahrens */ 4197fa9e4066Sahrens void 4198fa9e4066Sahrens spa_evict_all(void) 4199fa9e4066Sahrens { 4200fa9e4066Sahrens spa_t *spa; 4201fa9e4066Sahrens 4202fa9e4066Sahrens /* 4203fa9e4066Sahrens * Remove all cached state. All pools should be closed now, 4204fa9e4066Sahrens * so every spa in the AVL tree should be unreferenced. 4205fa9e4066Sahrens */ 4206fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 4207fa9e4066Sahrens while ((spa = spa_next(NULL)) != NULL) { 4208fa9e4066Sahrens /* 4209ea8dc4b6Seschrock * Stop async tasks. The async thread may need to detach 4210ea8dc4b6Seschrock * a device that's been replaced, which requires grabbing 4211ea8dc4b6Seschrock * spa_namespace_lock, so we must drop it here. 4212fa9e4066Sahrens */ 4213fa9e4066Sahrens spa_open_ref(spa, FTAG); 4214fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 4215ea8dc4b6Seschrock spa_async_suspend(spa); 4216fa9e4066Sahrens mutex_enter(&spa_namespace_lock); 4217fa9e4066Sahrens spa_close(spa, FTAG); 4218fa9e4066Sahrens 4219fa9e4066Sahrens if (spa->spa_state != POOL_STATE_UNINITIALIZED) { 4220fa9e4066Sahrens spa_unload(spa); 4221fa9e4066Sahrens spa_deactivate(spa); 4222fa9e4066Sahrens } 4223fa9e4066Sahrens spa_remove(spa); 4224fa9e4066Sahrens } 4225fa9e4066Sahrens mutex_exit(&spa_namespace_lock); 4226fa9e4066Sahrens } 4227ea8dc4b6Seschrock 4228ea8dc4b6Seschrock vdev_t * 4229c5904d13Seschrock spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t l2cache) 4230ea8dc4b6Seschrock { 4231c5904d13Seschrock vdev_t *vd; 4232c5904d13Seschrock int i; 4233c5904d13Seschrock 4234c5904d13Seschrock if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL) 4235c5904d13Seschrock return (vd); 4236c5904d13Seschrock 4237c5904d13Seschrock if (l2cache) { 4238c5904d13Seschrock for (i = 0; i < spa->spa_l2cache.sav_count; i++) { 4239c5904d13Seschrock vd = spa->spa_l2cache.sav_vdevs[i]; 4240c5904d13Seschrock if (vd->vdev_guid == guid) 4241c5904d13Seschrock return (vd); 4242c5904d13Seschrock } 4243c5904d13Seschrock } 4244c5904d13Seschrock 4245c5904d13Seschrock return (NULL); 4246ea8dc4b6Seschrock } 4247eaca9bbdSeschrock 4248eaca9bbdSeschrock void 4249990b4856Slling spa_upgrade(spa_t *spa, uint64_t version) 4250eaca9bbdSeschrock { 4251e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 4252eaca9bbdSeschrock 4253eaca9bbdSeschrock /* 4254eaca9bbdSeschrock * This should only be called for a non-faulted pool, and since a 4255eaca9bbdSeschrock * future version would result in an unopenable pool, this shouldn't be 4256eaca9bbdSeschrock * possible. 4257eaca9bbdSeschrock */ 4258e7437265Sahrens ASSERT(spa->spa_uberblock.ub_version <= SPA_VERSION); 4259990b4856Slling ASSERT(version >= spa->spa_uberblock.ub_version); 4260eaca9bbdSeschrock 4261990b4856Slling spa->spa_uberblock.ub_version = version; 4262eaca9bbdSeschrock vdev_config_dirty(spa->spa_root_vdev); 4263eaca9bbdSeschrock 4264e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 426599653d4eSeschrock 426699653d4eSeschrock txg_wait_synced(spa_get_dsl(spa), 0); 426799653d4eSeschrock } 426899653d4eSeschrock 426999653d4eSeschrock boolean_t 427099653d4eSeschrock spa_has_spare(spa_t *spa, uint64_t guid) 427199653d4eSeschrock { 427299653d4eSeschrock int i; 427339c23413Seschrock uint64_t spareguid; 4274fa94a07fSbrendan spa_aux_vdev_t *sav = &spa->spa_spares; 427599653d4eSeschrock 4276fa94a07fSbrendan for (i = 0; i < sav->sav_count; i++) 4277fa94a07fSbrendan if (sav->sav_vdevs[i]->vdev_guid == guid) 427899653d4eSeschrock return (B_TRUE); 427999653d4eSeschrock 4280fa94a07fSbrendan for (i = 0; i < sav->sav_npending; i++) { 4281fa94a07fSbrendan if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID, 4282fa94a07fSbrendan &spareguid) == 0 && spareguid == guid) 428339c23413Seschrock return (B_TRUE); 428439c23413Seschrock } 428539c23413Seschrock 428699653d4eSeschrock return (B_FALSE); 4287eaca9bbdSeschrock } 4288b1b8ab34Slling 428989a89ebfSlling /* 429089a89ebfSlling * Check if a pool has an active shared spare device. 429189a89ebfSlling * Note: reference count of an active spare is 2, as a spare and as a replace 429289a89ebfSlling */ 429389a89ebfSlling static boolean_t 429489a89ebfSlling spa_has_active_shared_spare(spa_t *spa) 429589a89ebfSlling { 429689a89ebfSlling int i, refcnt; 429789a89ebfSlling uint64_t pool; 429889a89ebfSlling spa_aux_vdev_t *sav = &spa->spa_spares; 429989a89ebfSlling 430089a89ebfSlling for (i = 0; i < sav->sav_count; i++) { 430189a89ebfSlling if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool, 430289a89ebfSlling &refcnt) && pool != 0ULL && pool == spa_guid(spa) && 430389a89ebfSlling refcnt > 2) 430489a89ebfSlling return (B_TRUE); 430589a89ebfSlling } 430689a89ebfSlling 430789a89ebfSlling return (B_FALSE); 430889a89ebfSlling } 430989a89ebfSlling 43103d7072f8Seschrock /* 43113d7072f8Seschrock * Post a sysevent corresponding to the given event. The 'name' must be one of 43123d7072f8Seschrock * the event definitions in sys/sysevent/eventdefs.h. The payload will be 43133d7072f8Seschrock * filled in from the spa and (optionally) the vdev. This doesn't do anything 43143d7072f8Seschrock * in the userland libzpool, as we don't want consumers to misinterpret ztest 43153d7072f8Seschrock * or zdb as real changes. 43163d7072f8Seschrock */ 43173d7072f8Seschrock void 43183d7072f8Seschrock spa_event_notify(spa_t *spa, vdev_t *vd, const char *name) 43193d7072f8Seschrock { 43203d7072f8Seschrock #ifdef _KERNEL 43213d7072f8Seschrock sysevent_t *ev; 43223d7072f8Seschrock sysevent_attr_list_t *attr = NULL; 43233d7072f8Seschrock sysevent_value_t value; 43243d7072f8Seschrock sysevent_id_t eid; 43253d7072f8Seschrock 43263d7072f8Seschrock ev = sysevent_alloc(EC_ZFS, (char *)name, SUNW_KERN_PUB "zfs", 43273d7072f8Seschrock SE_SLEEP); 43283d7072f8Seschrock 43293d7072f8Seschrock value.value_type = SE_DATA_TYPE_STRING; 43303d7072f8Seschrock value.value.sv_string = spa_name(spa); 43313d7072f8Seschrock if (sysevent_add_attr(&attr, ZFS_EV_POOL_NAME, &value, SE_SLEEP) != 0) 43323d7072f8Seschrock goto done; 43333d7072f8Seschrock 43343d7072f8Seschrock value.value_type = SE_DATA_TYPE_UINT64; 43353d7072f8Seschrock value.value.sv_uint64 = spa_guid(spa); 43363d7072f8Seschrock if (sysevent_add_attr(&attr, ZFS_EV_POOL_GUID, &value, SE_SLEEP) != 0) 43373d7072f8Seschrock goto done; 43383d7072f8Seschrock 43393d7072f8Seschrock if (vd) { 43403d7072f8Seschrock value.value_type = SE_DATA_TYPE_UINT64; 43413d7072f8Seschrock value.value.sv_uint64 = vd->vdev_guid; 43423d7072f8Seschrock if (sysevent_add_attr(&attr, ZFS_EV_VDEV_GUID, &value, 43433d7072f8Seschrock SE_SLEEP) != 0) 43443d7072f8Seschrock goto done; 43453d7072f8Seschrock 43463d7072f8Seschrock if (vd->vdev_path) { 43473d7072f8Seschrock value.value_type = SE_DATA_TYPE_STRING; 43483d7072f8Seschrock value.value.sv_string = vd->vdev_path; 43493d7072f8Seschrock if (sysevent_add_attr(&attr, ZFS_EV_VDEV_PATH, 43503d7072f8Seschrock &value, SE_SLEEP) != 0) 43513d7072f8Seschrock goto done; 43523d7072f8Seschrock } 43533d7072f8Seschrock } 43543d7072f8Seschrock 4355b01c3b58Seschrock if (sysevent_attach_attributes(ev, attr) != 0) 4356b01c3b58Seschrock goto done; 4357b01c3b58Seschrock attr = NULL; 4358b01c3b58Seschrock 43593d7072f8Seschrock (void) log_sysevent(ev, SE_SLEEP, &eid); 43603d7072f8Seschrock 43613d7072f8Seschrock done: 43623d7072f8Seschrock if (attr) 43633d7072f8Seschrock sysevent_free_attr(attr); 43643d7072f8Seschrock sysevent_free(ev); 43653d7072f8Seschrock #endif 43663d7072f8Seschrock } 4367