1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5441d80aaSlling * Common Development and Distribution License (the "License"). 6441d80aaSlling * 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 */ 21fa9e4066Sahrens /* 22*1195e687SMark J Musante * Copyright 2010 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens /* 27fa9e4066Sahrens * Virtual Device Labels 28fa9e4066Sahrens * --------------------- 29fa9e4066Sahrens * 30fa9e4066Sahrens * The vdev label serves several distinct purposes: 31fa9e4066Sahrens * 32fa9e4066Sahrens * 1. Uniquely identify this device as part of a ZFS pool and confirm its 33fa9e4066Sahrens * identity within the pool. 34fa9e4066Sahrens * 35fa9e4066Sahrens * 2. Verify that all the devices given in a configuration are present 36fa9e4066Sahrens * within the pool. 37fa9e4066Sahrens * 38fa9e4066Sahrens * 3. Determine the uberblock for the pool. 39fa9e4066Sahrens * 40fa9e4066Sahrens * 4. In case of an import operation, determine the configuration of the 41fa9e4066Sahrens * toplevel vdev of which it is a part. 42fa9e4066Sahrens * 43fa9e4066Sahrens * 5. If an import operation cannot find all the devices in the pool, 44fa9e4066Sahrens * provide enough information to the administrator to determine which 45fa9e4066Sahrens * devices are missing. 46fa9e4066Sahrens * 47fa9e4066Sahrens * It is important to note that while the kernel is responsible for writing the 48fa9e4066Sahrens * label, it only consumes the information in the first three cases. The 49fa9e4066Sahrens * latter information is only consumed in userland when determining the 50fa9e4066Sahrens * configuration to import a pool. 51fa9e4066Sahrens * 52fa9e4066Sahrens * 53fa9e4066Sahrens * Label Organization 54fa9e4066Sahrens * ------------------ 55fa9e4066Sahrens * 56fa9e4066Sahrens * Before describing the contents of the label, it's important to understand how 57fa9e4066Sahrens * the labels are written and updated with respect to the uberblock. 58fa9e4066Sahrens * 59fa9e4066Sahrens * When the pool configuration is altered, either because it was newly created 60fa9e4066Sahrens * or a device was added, we want to update all the labels such that we can deal 61fa9e4066Sahrens * with fatal failure at any point. To this end, each disk has two labels which 62fa9e4066Sahrens * are updated before and after the uberblock is synced. Assuming we have 633d7072f8Seschrock * labels and an uberblock with the following transaction groups: 64fa9e4066Sahrens * 65fa9e4066Sahrens * L1 UB L2 66fa9e4066Sahrens * +------+ +------+ +------+ 67fa9e4066Sahrens * | | | | | | 68fa9e4066Sahrens * | t10 | | t10 | | t10 | 69fa9e4066Sahrens * | | | | | | 70fa9e4066Sahrens * +------+ +------+ +------+ 71fa9e4066Sahrens * 72fa9e4066Sahrens * In this stable state, the labels and the uberblock were all updated within 73fa9e4066Sahrens * the same transaction group (10). Each label is mirrored and checksummed, so 74fa9e4066Sahrens * that we can detect when we fail partway through writing the label. 75fa9e4066Sahrens * 76fa9e4066Sahrens * In order to identify which labels are valid, the labels are written in the 77fa9e4066Sahrens * following manner: 78fa9e4066Sahrens * 79fa9e4066Sahrens * 1. For each vdev, update 'L1' to the new label 80fa9e4066Sahrens * 2. Update the uberblock 81fa9e4066Sahrens * 3. For each vdev, update 'L2' to the new label 82fa9e4066Sahrens * 83fa9e4066Sahrens * Given arbitrary failure, we can determine the correct label to use based on 84fa9e4066Sahrens * the transaction group. If we fail after updating L1 but before updating the 85fa9e4066Sahrens * UB, we will notice that L1's transaction group is greater than the uberblock, 86fa9e4066Sahrens * so L2 must be valid. If we fail after writing the uberblock but before 87fa9e4066Sahrens * writing L2, we will notice that L2's transaction group is less than L1, and 88fa9e4066Sahrens * therefore L1 is valid. 89fa9e4066Sahrens * 90fa9e4066Sahrens * Another added complexity is that not every label is updated when the config 91fa9e4066Sahrens * is synced. If we add a single device, we do not want to have to re-write 92fa9e4066Sahrens * every label for every device in the pool. This means that both L1 and L2 may 93fa9e4066Sahrens * be older than the pool uberblock, because the necessary information is stored 94fa9e4066Sahrens * on another vdev. 95fa9e4066Sahrens * 96fa9e4066Sahrens * 97fa9e4066Sahrens * On-disk Format 98fa9e4066Sahrens * -------------- 99fa9e4066Sahrens * 100fa9e4066Sahrens * The vdev label consists of two distinct parts, and is wrapped within the 101fa9e4066Sahrens * vdev_label_t structure. The label includes 8k of padding to permit legacy 102fa9e4066Sahrens * VTOC disk labels, but is otherwise ignored. 103fa9e4066Sahrens * 104fa9e4066Sahrens * The first half of the label is a packed nvlist which contains pool wide 105fa9e4066Sahrens * properties, per-vdev properties, and configuration information. It is 106fa9e4066Sahrens * described in more detail below. 107fa9e4066Sahrens * 108fa9e4066Sahrens * The latter half of the label consists of a redundant array of uberblocks. 109fa9e4066Sahrens * These uberblocks are updated whenever a transaction group is committed, 110fa9e4066Sahrens * or when the configuration is updated. When a pool is loaded, we scan each 111fa9e4066Sahrens * vdev for the 'best' uberblock. 112fa9e4066Sahrens * 113fa9e4066Sahrens * 114fa9e4066Sahrens * Configuration Information 115fa9e4066Sahrens * ------------------------- 116fa9e4066Sahrens * 117fa9e4066Sahrens * The nvlist describing the pool and vdev contains the following elements: 118fa9e4066Sahrens * 119fa9e4066Sahrens * version ZFS on-disk version 120fa9e4066Sahrens * name Pool name 121fa9e4066Sahrens * state Pool state 122fa9e4066Sahrens * txg Transaction group in which this label was written 123fa9e4066Sahrens * pool_guid Unique identifier for this pool 124fa9e4066Sahrens * vdev_tree An nvlist describing vdev tree. 125fa9e4066Sahrens * 126fa9e4066Sahrens * Each leaf device label also contains the following: 127fa9e4066Sahrens * 128fa9e4066Sahrens * top_guid Unique ID for top-level vdev in which this is contained 129fa9e4066Sahrens * guid Unique ID for the leaf vdev 130fa9e4066Sahrens * 131fa9e4066Sahrens * The 'vs' configuration follows the format described in 'spa_config.c'. 132fa9e4066Sahrens */ 133fa9e4066Sahrens 134fa9e4066Sahrens #include <sys/zfs_context.h> 135fa9e4066Sahrens #include <sys/spa.h> 136fa9e4066Sahrens #include <sys/spa_impl.h> 137fa9e4066Sahrens #include <sys/dmu.h> 138fa9e4066Sahrens #include <sys/zap.h> 139fa9e4066Sahrens #include <sys/vdev.h> 140fa9e4066Sahrens #include <sys/vdev_impl.h> 141fa9e4066Sahrens #include <sys/uberblock_impl.h> 142fa9e4066Sahrens #include <sys/metaslab.h> 143fa9e4066Sahrens #include <sys/zio.h> 144fa9e4066Sahrens #include <sys/fs/zfs.h> 145fa9e4066Sahrens 146fa9e4066Sahrens /* 147fa9e4066Sahrens * Basic routines to read and write from a vdev label. 148fa9e4066Sahrens * Used throughout the rest of this file. 149fa9e4066Sahrens */ 150fa9e4066Sahrens uint64_t 151fa9e4066Sahrens vdev_label_offset(uint64_t psize, int l, uint64_t offset) 152fa9e4066Sahrens { 153ecc2d604Sbonwick ASSERT(offset < sizeof (vdev_label_t)); 154e7437265Sahrens ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); 155ecc2d604Sbonwick 156fa9e4066Sahrens return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 157fa9e4066Sahrens 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); 158fa9e4066Sahrens } 159fa9e4066Sahrens 16021bf64a7Sgw /* 16121bf64a7Sgw * Returns back the vdev label associated with the passed in offset. 16221bf64a7Sgw */ 16321bf64a7Sgw int 16421bf64a7Sgw vdev_label_number(uint64_t psize, uint64_t offset) 16521bf64a7Sgw { 16621bf64a7Sgw int l; 16721bf64a7Sgw 16821bf64a7Sgw if (offset >= psize - VDEV_LABEL_END_SIZE) { 16921bf64a7Sgw offset -= psize - VDEV_LABEL_END_SIZE; 17021bf64a7Sgw offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t); 17121bf64a7Sgw } 17221bf64a7Sgw l = offset / sizeof (vdev_label_t); 17321bf64a7Sgw return (l < VDEV_LABELS ? l : -1); 17421bf64a7Sgw } 17521bf64a7Sgw 176fa9e4066Sahrens static void 177fa9e4066Sahrens vdev_label_read(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset, 178e14bb325SJeff Bonwick uint64_t size, zio_done_func_t *done, void *private, int flags) 179fa9e4066Sahrens { 180e14bb325SJeff Bonwick ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) == 181e14bb325SJeff Bonwick SCL_STATE_ALL); 182e14bb325SJeff Bonwick ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); 183fa9e4066Sahrens 184fa9e4066Sahrens zio_nowait(zio_read_phys(zio, vd, 185fa9e4066Sahrens vdev_label_offset(vd->vdev_psize, l, offset), 186fa9e4066Sahrens size, buf, ZIO_CHECKSUM_LABEL, done, private, 187e14bb325SJeff Bonwick ZIO_PRIORITY_SYNC_READ, flags, B_TRUE)); 188fa9e4066Sahrens } 189fa9e4066Sahrens 190fa9e4066Sahrens static void 191fa9e4066Sahrens vdev_label_write(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset, 19217f17c2dSbonwick uint64_t size, zio_done_func_t *done, void *private, int flags) 193fa9e4066Sahrens { 194e14bb325SJeff Bonwick ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL || 195e14bb325SJeff Bonwick (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) == 196e14bb325SJeff Bonwick (SCL_CONFIG | SCL_STATE) && 197e14bb325SJeff Bonwick dsl_pool_sync_context(spa_get_dsl(zio->io_spa)))); 198e14bb325SJeff Bonwick ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); 199fa9e4066Sahrens 200fa9e4066Sahrens zio_nowait(zio_write_phys(zio, vd, 201fa9e4066Sahrens vdev_label_offset(vd->vdev_psize, l, offset), 202fa9e4066Sahrens size, buf, ZIO_CHECKSUM_LABEL, done, private, 20317f17c2dSbonwick ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE)); 204fa9e4066Sahrens } 205fa9e4066Sahrens 206fa9e4066Sahrens /* 207fa9e4066Sahrens * Generate the nvlist representing this vdev's config. 208fa9e4066Sahrens */ 209fa9e4066Sahrens nvlist_t * 21099653d4eSeschrock vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, 211fa94a07fSbrendan boolean_t isspare, boolean_t isl2cache) 212fa9e4066Sahrens { 213fa9e4066Sahrens nvlist_t *nv = NULL; 214fa9e4066Sahrens 215ea8dc4b6Seschrock VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 216fa9e4066Sahrens 217fa9e4066Sahrens VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE, 218fa9e4066Sahrens vd->vdev_ops->vdev_op_type) == 0); 219fa94a07fSbrendan if (!isspare && !isl2cache) 22099653d4eSeschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id) 22199653d4eSeschrock == 0); 222fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid) == 0); 223fa9e4066Sahrens 224fa9e4066Sahrens if (vd->vdev_path != NULL) 225fa9e4066Sahrens VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, 226fa9e4066Sahrens vd->vdev_path) == 0); 227fa9e4066Sahrens 228fa9e4066Sahrens if (vd->vdev_devid != NULL) 229fa9e4066Sahrens VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, 230fa9e4066Sahrens vd->vdev_devid) == 0); 231fa9e4066Sahrens 2323d7072f8Seschrock if (vd->vdev_physpath != NULL) 2333d7072f8Seschrock VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, 2343d7072f8Seschrock vd->vdev_physpath) == 0); 2353d7072f8Seschrock 2366809eb4eSEric Schrock if (vd->vdev_fru != NULL) 2376809eb4eSEric Schrock VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_FRU, 2386809eb4eSEric Schrock vd->vdev_fru) == 0); 2396809eb4eSEric Schrock 24099653d4eSeschrock if (vd->vdev_nparity != 0) { 24199653d4eSeschrock ASSERT(strcmp(vd->vdev_ops->vdev_op_type, 24299653d4eSeschrock VDEV_TYPE_RAIDZ) == 0); 24399653d4eSeschrock 24499653d4eSeschrock /* 24599653d4eSeschrock * Make sure someone hasn't managed to sneak a fancy new vdev 24699653d4eSeschrock * into a crufty old storage pool. 24799653d4eSeschrock */ 24899653d4eSeschrock ASSERT(vd->vdev_nparity == 1 || 249f94275ceSAdam Leventhal (vd->vdev_nparity <= 2 && 250f94275ceSAdam Leventhal spa_version(spa) >= SPA_VERSION_RAIDZ2) || 251f94275ceSAdam Leventhal (vd->vdev_nparity <= 3 && 252f94275ceSAdam Leventhal spa_version(spa) >= SPA_VERSION_RAIDZ3)); 25399653d4eSeschrock 25499653d4eSeschrock /* 25599653d4eSeschrock * Note that we'll add the nparity tag even on storage pools 25699653d4eSeschrock * that only support a single parity device -- older software 25799653d4eSeschrock * will just ignore it. 25899653d4eSeschrock */ 25999653d4eSeschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, 26099653d4eSeschrock vd->vdev_nparity) == 0); 26199653d4eSeschrock } 26299653d4eSeschrock 263afefbcddSeschrock if (vd->vdev_wholedisk != -1ULL) 264afefbcddSeschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 265afefbcddSeschrock vd->vdev_wholedisk) == 0); 266afefbcddSeschrock 267ea8dc4b6Seschrock if (vd->vdev_not_present) 268ea8dc4b6Seschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1) == 0); 269ea8dc4b6Seschrock 27099653d4eSeschrock if (vd->vdev_isspare) 27199653d4eSeschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1) == 0); 27299653d4eSeschrock 273fa94a07fSbrendan if (!isspare && !isl2cache && vd == vd->vdev_top) { 274fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY, 275fa9e4066Sahrens vd->vdev_ms_array) == 0); 276fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT, 277fa9e4066Sahrens vd->vdev_ms_shift) == 0); 278fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, 279fa9e4066Sahrens vd->vdev_ashift) == 0); 280fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE, 281fa9e4066Sahrens vd->vdev_asize) == 0); 2828654d025Sperrin VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, 2838654d025Sperrin vd->vdev_islog) == 0); 284fa9e4066Sahrens } 285fa9e4066Sahrens 2868ad4d6ddSJeff Bonwick if (vd->vdev_dtl_smo.smo_object != 0) 287fa9e4066Sahrens VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DTL, 2888ad4d6ddSJeff Bonwick vd->vdev_dtl_smo.smo_object) == 0); 289fa9e4066Sahrens 29088ecc943SGeorge Wilson if (vd->vdev_crtxg) 29188ecc943SGeorge Wilson VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, 29288ecc943SGeorge Wilson vd->vdev_crtxg) == 0); 29388ecc943SGeorge Wilson 294fa9e4066Sahrens if (getstats) { 295fa9e4066Sahrens vdev_stat_t vs; 296fa9e4066Sahrens vdev_get_stats(vd, &vs); 297fa9e4066Sahrens VERIFY(nvlist_add_uint64_array(nv, ZPOOL_CONFIG_STATS, 298fa9e4066Sahrens (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t)) == 0); 299fa9e4066Sahrens } 300fa9e4066Sahrens 301fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) { 302fa9e4066Sahrens nvlist_t **child; 303fa9e4066Sahrens int c; 304fa9e4066Sahrens 30588ecc943SGeorge Wilson ASSERT(!vd->vdev_ishole); 30688ecc943SGeorge Wilson 307fa9e4066Sahrens child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *), 308fa9e4066Sahrens KM_SLEEP); 309fa9e4066Sahrens 310fa9e4066Sahrens for (c = 0; c < vd->vdev_children; c++) 31199653d4eSeschrock child[c] = vdev_config_generate(spa, vd->vdev_child[c], 312fa94a07fSbrendan getstats, isspare, isl2cache); 313fa9e4066Sahrens 314fa9e4066Sahrens VERIFY(nvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 315fa9e4066Sahrens child, vd->vdev_children) == 0); 316fa9e4066Sahrens 317fa9e4066Sahrens for (c = 0; c < vd->vdev_children; c++) 318fa9e4066Sahrens nvlist_free(child[c]); 319fa9e4066Sahrens 320fa9e4066Sahrens kmem_free(child, vd->vdev_children * sizeof (nvlist_t *)); 321441d80aaSlling 322441d80aaSlling } else { 323069f55e2SEric Schrock const char *aux = NULL; 324069f55e2SEric Schrock 325ecc2d604Sbonwick if (vd->vdev_offline && !vd->vdev_tmpoffline) 326441d80aaSlling VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, 327ecc2d604Sbonwick B_TRUE) == 0); 3283d7072f8Seschrock if (vd->vdev_faulted) 3293d7072f8Seschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, 3303d7072f8Seschrock B_TRUE) == 0); 3313d7072f8Seschrock if (vd->vdev_degraded) 3323d7072f8Seschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, 3333d7072f8Seschrock B_TRUE) == 0); 3343d7072f8Seschrock if (vd->vdev_removed) 3353d7072f8Seschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, 3363d7072f8Seschrock B_TRUE) == 0); 3373d7072f8Seschrock if (vd->vdev_unspare) 3383d7072f8Seschrock VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, 3393d7072f8Seschrock B_TRUE) == 0); 34088ecc943SGeorge Wilson if (vd->vdev_ishole) 34188ecc943SGeorge Wilson VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, 34288ecc943SGeorge Wilson B_TRUE) == 0); 343069f55e2SEric Schrock 344069f55e2SEric Schrock switch (vd->vdev_stat.vs_aux) { 345069f55e2SEric Schrock case VDEV_AUX_ERR_EXCEEDED: 346069f55e2SEric Schrock aux = "err_exceeded"; 347069f55e2SEric Schrock break; 348069f55e2SEric Schrock 349069f55e2SEric Schrock case VDEV_AUX_EXTERNAL: 350069f55e2SEric Schrock aux = "external"; 351069f55e2SEric Schrock break; 352069f55e2SEric Schrock } 353069f55e2SEric Schrock 354069f55e2SEric Schrock if (aux != NULL) 355069f55e2SEric Schrock VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, 356069f55e2SEric Schrock aux) == 0); 357*1195e687SMark J Musante 358*1195e687SMark J Musante if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) { 359*1195e687SMark J Musante VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID, 360*1195e687SMark J Musante vd->vdev_orig_guid) == 0); 361*1195e687SMark J Musante } 362fa9e4066Sahrens } 363fa9e4066Sahrens 364fa9e4066Sahrens return (nv); 365fa9e4066Sahrens } 366fa9e4066Sahrens 36788ecc943SGeorge Wilson /* 36888ecc943SGeorge Wilson * Generate a view of the top-level vdevs. If we currently have holes 36988ecc943SGeorge Wilson * in the namespace, then generate an array which contains a list of holey 37088ecc943SGeorge Wilson * vdevs. Additionally, add the number of top-level children that currently 37188ecc943SGeorge Wilson * exist. 37288ecc943SGeorge Wilson */ 37388ecc943SGeorge Wilson void 37488ecc943SGeorge Wilson vdev_top_config_generate(spa_t *spa, nvlist_t *config) 37588ecc943SGeorge Wilson { 37688ecc943SGeorge Wilson vdev_t *rvd = spa->spa_root_vdev; 37788ecc943SGeorge Wilson uint64_t *array; 37888ecc943SGeorge Wilson uint_t idx; 37988ecc943SGeorge Wilson 38088ecc943SGeorge Wilson array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP); 38188ecc943SGeorge Wilson 38288ecc943SGeorge Wilson idx = 0; 38388ecc943SGeorge Wilson for (int c = 0; c < rvd->vdev_children; c++) { 38488ecc943SGeorge Wilson vdev_t *tvd = rvd->vdev_child[c]; 38588ecc943SGeorge Wilson 38688ecc943SGeorge Wilson if (tvd->vdev_ishole) 38788ecc943SGeorge Wilson array[idx++] = c; 38888ecc943SGeorge Wilson } 38988ecc943SGeorge Wilson 390312c6e15SGeorge Wilson if (idx) { 391312c6e15SGeorge Wilson VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, 392312c6e15SGeorge Wilson array, idx) == 0); 393312c6e15SGeorge Wilson } 394312c6e15SGeorge Wilson 39588ecc943SGeorge Wilson VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, 39688ecc943SGeorge Wilson rvd->vdev_children) == 0); 39788ecc943SGeorge Wilson 39888ecc943SGeorge Wilson kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); 39988ecc943SGeorge Wilson } 40088ecc943SGeorge Wilson 401fa9e4066Sahrens nvlist_t * 402fa9e4066Sahrens vdev_label_read_config(vdev_t *vd) 403fa9e4066Sahrens { 4040373e76bSbonwick spa_t *spa = vd->vdev_spa; 405fa9e4066Sahrens nvlist_t *config = NULL; 406fa9e4066Sahrens vdev_phys_t *vp; 407fa9e4066Sahrens zio_t *zio; 4088956713aSEric Schrock int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | 4098956713aSEric Schrock ZIO_FLAG_SPECULATIVE; 410fa9e4066Sahrens 411e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL); 4120373e76bSbonwick 4130a4e9518Sgw if (!vdev_readable(vd)) 414fa9e4066Sahrens return (NULL); 415fa9e4066Sahrens 416fa9e4066Sahrens vp = zio_buf_alloc(sizeof (vdev_phys_t)); 417fa9e4066Sahrens 4188956713aSEric Schrock retry: 419e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 420fa9e4066Sahrens 421e14bb325SJeff Bonwick zio = zio_root(spa, NULL, NULL, flags); 422fa9e4066Sahrens 423fa9e4066Sahrens vdev_label_read(zio, vd, l, vp, 424fa9e4066Sahrens offsetof(vdev_label_t, vl_vdev_phys), 425e14bb325SJeff Bonwick sizeof (vdev_phys_t), NULL, NULL, flags); 426fa9e4066Sahrens 427fa9e4066Sahrens if (zio_wait(zio) == 0 && 428fa9e4066Sahrens nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist), 429ea8dc4b6Seschrock &config, 0) == 0) 430fa9e4066Sahrens break; 431fa9e4066Sahrens 432fa9e4066Sahrens if (config != NULL) { 433fa9e4066Sahrens nvlist_free(config); 434fa9e4066Sahrens config = NULL; 435fa9e4066Sahrens } 436fa9e4066Sahrens } 437fa9e4066Sahrens 4388956713aSEric Schrock if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) { 4398956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 4408956713aSEric Schrock goto retry; 4418956713aSEric Schrock } 4428956713aSEric Schrock 443fa9e4066Sahrens zio_buf_free(vp, sizeof (vdev_phys_t)); 444fa9e4066Sahrens 445fa9e4066Sahrens return (config); 446fa9e4066Sahrens } 447fa9e4066Sahrens 44839c23413Seschrock /* 44939c23413Seschrock * Determine if a device is in use. The 'spare_guid' parameter will be filled 45039c23413Seschrock * in with the device guid if this spare is active elsewhere on the system. 45139c23413Seschrock */ 45239c23413Seschrock static boolean_t 45339c23413Seschrock vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason, 454fa94a07fSbrendan uint64_t *spare_guid, uint64_t *l2cache_guid) 45539c23413Seschrock { 45639c23413Seschrock spa_t *spa = vd->vdev_spa; 45739c23413Seschrock uint64_t state, pool_guid, device_guid, txg, spare_pool; 45839c23413Seschrock uint64_t vdtxg = 0; 45939c23413Seschrock nvlist_t *label; 46039c23413Seschrock 46139c23413Seschrock if (spare_guid) 46239c23413Seschrock *spare_guid = 0ULL; 463fa94a07fSbrendan if (l2cache_guid) 464fa94a07fSbrendan *l2cache_guid = 0ULL; 46539c23413Seschrock 46639c23413Seschrock /* 46739c23413Seschrock * Read the label, if any, and perform some basic sanity checks. 46839c23413Seschrock */ 46939c23413Seschrock if ((label = vdev_label_read_config(vd)) == NULL) 47039c23413Seschrock return (B_FALSE); 47139c23413Seschrock 47239c23413Seschrock (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG, 47339c23413Seschrock &vdtxg); 47439c23413Seschrock 47539c23413Seschrock if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, 47639c23413Seschrock &state) != 0 || 47739c23413Seschrock nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, 47839c23413Seschrock &device_guid) != 0) { 47939c23413Seschrock nvlist_free(label); 48039c23413Seschrock return (B_FALSE); 48139c23413Seschrock } 48239c23413Seschrock 483fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 48439c23413Seschrock (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, 48539c23413Seschrock &pool_guid) != 0 || 48639c23413Seschrock nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, 48739c23413Seschrock &txg) != 0)) { 48839c23413Seschrock nvlist_free(label); 48939c23413Seschrock return (B_FALSE); 49039c23413Seschrock } 49139c23413Seschrock 49239c23413Seschrock nvlist_free(label); 49339c23413Seschrock 49439c23413Seschrock /* 49539c23413Seschrock * Check to see if this device indeed belongs to the pool it claims to 49639c23413Seschrock * be a part of. The only way this is allowed is if the device is a hot 49739c23413Seschrock * spare (which we check for later on). 49839c23413Seschrock */ 499fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 50039c23413Seschrock !spa_guid_exists(pool_guid, device_guid) && 50189a89ebfSlling !spa_spare_exists(device_guid, NULL, NULL) && 502fa94a07fSbrendan !spa_l2cache_exists(device_guid, NULL)) 50339c23413Seschrock return (B_FALSE); 50439c23413Seschrock 50539c23413Seschrock /* 50639c23413Seschrock * If the transaction group is zero, then this an initialized (but 50739c23413Seschrock * unused) label. This is only an error if the create transaction 50839c23413Seschrock * on-disk is the same as the one we're using now, in which case the 50939c23413Seschrock * user has attempted to add the same vdev multiple times in the same 51039c23413Seschrock * transaction. 51139c23413Seschrock */ 512fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 513fa94a07fSbrendan txg == 0 && vdtxg == crtxg) 51439c23413Seschrock return (B_TRUE); 51539c23413Seschrock 51639c23413Seschrock /* 51739c23413Seschrock * Check to see if this is a spare device. We do an explicit check for 51839c23413Seschrock * spa_has_spare() here because it may be on our pending list of spares 519fa94a07fSbrendan * to add. We also check if it is an l2cache device. 52039c23413Seschrock */ 52189a89ebfSlling if (spa_spare_exists(device_guid, &spare_pool, NULL) || 52239c23413Seschrock spa_has_spare(spa, device_guid)) { 52339c23413Seschrock if (spare_guid) 52439c23413Seschrock *spare_guid = device_guid; 52539c23413Seschrock 52639c23413Seschrock switch (reason) { 52739c23413Seschrock case VDEV_LABEL_CREATE: 528fa94a07fSbrendan case VDEV_LABEL_L2CACHE: 52939c23413Seschrock return (B_TRUE); 53039c23413Seschrock 53139c23413Seschrock case VDEV_LABEL_REPLACE: 53239c23413Seschrock return (!spa_has_spare(spa, device_guid) || 53339c23413Seschrock spare_pool != 0ULL); 53439c23413Seschrock 53539c23413Seschrock case VDEV_LABEL_SPARE: 53639c23413Seschrock return (spa_has_spare(spa, device_guid)); 53739c23413Seschrock } 53839c23413Seschrock } 53939c23413Seschrock 540fa94a07fSbrendan /* 541fa94a07fSbrendan * Check to see if this is an l2cache device. 542fa94a07fSbrendan */ 543fa94a07fSbrendan if (spa_l2cache_exists(device_guid, NULL)) 544fa94a07fSbrendan return (B_TRUE); 545fa94a07fSbrendan 54639c23413Seschrock /* 54739c23413Seschrock * If the device is marked ACTIVE, then this device is in use by another 54839c23413Seschrock * pool on the system. 54939c23413Seschrock */ 55039c23413Seschrock return (state == POOL_STATE_ACTIVE); 55139c23413Seschrock } 55239c23413Seschrock 55339c23413Seschrock /* 55439c23413Seschrock * Initialize a vdev label. We check to make sure each leaf device is not in 55539c23413Seschrock * use, and writable. We put down an initial label which we will later 55639c23413Seschrock * overwrite with a complete label. Note that it's important to do this 55739c23413Seschrock * sequentially, not in parallel, so that we catch cases of multiple use of the 55839c23413Seschrock * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with 55939c23413Seschrock * itself. 56039c23413Seschrock */ 56139c23413Seschrock int 56239c23413Seschrock vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) 563fa9e4066Sahrens { 564fa9e4066Sahrens spa_t *spa = vd->vdev_spa; 565fa9e4066Sahrens nvlist_t *label; 566fa9e4066Sahrens vdev_phys_t *vp; 567f83ffe1aSLin Ling char *pad2; 568ecc2d604Sbonwick uberblock_t *ub; 569fa9e4066Sahrens zio_t *zio; 570fa9e4066Sahrens char *buf; 571fa9e4066Sahrens size_t buflen; 572fa9e4066Sahrens int error; 573fa94a07fSbrendan uint64_t spare_guid, l2cache_guid; 574e14bb325SJeff Bonwick int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; 575fa9e4066Sahrens 576e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 5770373e76bSbonwick 578e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 57939c23413Seschrock if ((error = vdev_label_init(vd->vdev_child[c], 58039c23413Seschrock crtxg, reason)) != 0) 581fa9e4066Sahrens return (error); 582fa9e4066Sahrens 58388ecc943SGeorge Wilson /* Track the creation time for this vdev */ 58488ecc943SGeorge Wilson vd->vdev_crtxg = crtxg; 58588ecc943SGeorge Wilson 586fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) 587fa9e4066Sahrens return (0); 588fa9e4066Sahrens 589fa9e4066Sahrens /* 59039c23413Seschrock * Dead vdevs cannot be initialized. 591fa9e4066Sahrens */ 592fa9e4066Sahrens if (vdev_is_dead(vd)) 593fa9e4066Sahrens return (EIO); 594fa9e4066Sahrens 595fa9e4066Sahrens /* 59639c23413Seschrock * Determine if the vdev is in use. 597fa9e4066Sahrens */ 598*1195e687SMark J Musante if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT && 599fa94a07fSbrendan vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid)) 60039c23413Seschrock return (EBUSY); 60139c23413Seschrock 60239c23413Seschrock /* 603fa94a07fSbrendan * If this is a request to add or replace a spare or l2cache device 604fa94a07fSbrendan * that is in use elsewhere on the system, then we must update the 605fa94a07fSbrendan * guid (which was initialized to a random value) to reflect the 606fa94a07fSbrendan * actual GUID (which is shared between multiple pools). 60739c23413Seschrock */ 608fa94a07fSbrendan if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE && 609fa94a07fSbrendan spare_guid != 0ULL) { 61031157203SJeff Bonwick uint64_t guid_delta = spare_guid - vd->vdev_guid; 61199653d4eSeschrock 61231157203SJeff Bonwick vd->vdev_guid += guid_delta; 61331157203SJeff Bonwick 61431157203SJeff Bonwick for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) 61531157203SJeff Bonwick pvd->vdev_guid_sum += guid_delta; 61639c23413Seschrock 61799653d4eSeschrock /* 61839c23413Seschrock * If this is a replacement, then we want to fallthrough to the 61939c23413Seschrock * rest of the code. If we're adding a spare, then it's already 6203d7072f8Seschrock * labeled appropriately and we can just return. 62199653d4eSeschrock */ 62239c23413Seschrock if (reason == VDEV_LABEL_SPARE) 62339c23413Seschrock return (0); 624*1195e687SMark J Musante ASSERT(reason == VDEV_LABEL_REPLACE || 625*1195e687SMark J Musante reason == VDEV_LABEL_SPLIT); 626fa9e4066Sahrens } 627fa9e4066Sahrens 628fa94a07fSbrendan if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE && 629fa94a07fSbrendan l2cache_guid != 0ULL) { 63031157203SJeff Bonwick uint64_t guid_delta = l2cache_guid - vd->vdev_guid; 63131157203SJeff Bonwick 63231157203SJeff Bonwick vd->vdev_guid += guid_delta; 633fa94a07fSbrendan 63431157203SJeff Bonwick for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) 63531157203SJeff Bonwick pvd->vdev_guid_sum += guid_delta; 636fa94a07fSbrendan 637fa94a07fSbrendan /* 638fa94a07fSbrendan * If this is a replacement, then we want to fallthrough to the 639fa94a07fSbrendan * rest of the code. If we're adding an l2cache, then it's 640fa94a07fSbrendan * already labeled appropriately and we can just return. 641fa94a07fSbrendan */ 642fa94a07fSbrendan if (reason == VDEV_LABEL_L2CACHE) 643fa94a07fSbrendan return (0); 644fa94a07fSbrendan ASSERT(reason == VDEV_LABEL_REPLACE); 645fa94a07fSbrendan } 646fa94a07fSbrendan 647fa9e4066Sahrens /* 64839c23413Seschrock * Initialize its label. 649fa9e4066Sahrens */ 650fa9e4066Sahrens vp = zio_buf_alloc(sizeof (vdev_phys_t)); 651fa9e4066Sahrens bzero(vp, sizeof (vdev_phys_t)); 652fa9e4066Sahrens 653fa9e4066Sahrens /* 654fa9e4066Sahrens * Generate a label describing the pool and our top-level vdev. 655fa9e4066Sahrens * We mark it as being from txg 0 to indicate that it's not 656fa9e4066Sahrens * really part of an active pool just yet. The labels will 657fa9e4066Sahrens * be written again with a meaningful txg by spa_sync(). 658fa9e4066Sahrens */ 65939c23413Seschrock if (reason == VDEV_LABEL_SPARE || 66039c23413Seschrock (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) { 66139c23413Seschrock /* 66239c23413Seschrock * For inactive hot spares, we generate a special label that 66339c23413Seschrock * identifies as a mutually shared hot spare. We write the 66439c23413Seschrock * label if we are adding a hot spare, or if we are removing an 66539c23413Seschrock * active hot spare (in which case we want to revert the 66639c23413Seschrock * labels). 66739c23413Seschrock */ 66899653d4eSeschrock VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); 66999653d4eSeschrock 67099653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, 67199653d4eSeschrock spa_version(spa)) == 0); 67299653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, 67399653d4eSeschrock POOL_STATE_SPARE) == 0); 67499653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, 67599653d4eSeschrock vd->vdev_guid) == 0); 676fa94a07fSbrendan } else if (reason == VDEV_LABEL_L2CACHE || 677fa94a07fSbrendan (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) { 678fa94a07fSbrendan /* 679fa94a07fSbrendan * For level 2 ARC devices, add a special label. 680fa94a07fSbrendan */ 681fa94a07fSbrendan VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); 682fa94a07fSbrendan 683fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, 684fa94a07fSbrendan spa_version(spa)) == 0); 685fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, 686fa94a07fSbrendan POOL_STATE_L2CACHE) == 0); 687fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, 688fa94a07fSbrendan vd->vdev_guid) == 0); 68999653d4eSeschrock } else { 690*1195e687SMark J Musante uint64_t txg = 0ULL; 691*1195e687SMark J Musante 692*1195e687SMark J Musante if (reason == VDEV_LABEL_SPLIT) 693*1195e687SMark J Musante txg = spa->spa_uberblock.ub_txg; 694*1195e687SMark J Musante label = spa_config_generate(spa, vd, txg, B_FALSE); 69599653d4eSeschrock 69699653d4eSeschrock /* 69799653d4eSeschrock * Add our creation time. This allows us to detect multiple 69899653d4eSeschrock * vdev uses as described above, and automatically expires if we 69999653d4eSeschrock * fail. 70099653d4eSeschrock */ 70199653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, 70299653d4eSeschrock crtxg) == 0); 70399653d4eSeschrock } 704fa9e4066Sahrens 705fa9e4066Sahrens buf = vp->vp_nvlist; 706fa9e4066Sahrens buflen = sizeof (vp->vp_nvlist); 707fa9e4066Sahrens 708a75573b6Smmusante error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP); 709a75573b6Smmusante if (error != 0) { 710fa9e4066Sahrens nvlist_free(label); 711fa9e4066Sahrens zio_buf_free(vp, sizeof (vdev_phys_t)); 712a75573b6Smmusante /* EFAULT means nvlist_pack ran out of room */ 713a75573b6Smmusante return (error == EFAULT ? ENAMETOOLONG : EINVAL); 714fa9e4066Sahrens } 715fa9e4066Sahrens 716fa9e4066Sahrens /* 717fa9e4066Sahrens * Initialize uberblock template. 718fa9e4066Sahrens */ 719f64c0e34SEric Taylor ub = zio_buf_alloc(VDEV_UBERBLOCK_RING); 720f64c0e34SEric Taylor bzero(ub, VDEV_UBERBLOCK_RING); 721ecc2d604Sbonwick *ub = spa->spa_uberblock; 722ecc2d604Sbonwick ub->ub_txg = 0; 723fa9e4066Sahrens 724f83ffe1aSLin Ling /* Initialize the 2nd padding area. */ 725f83ffe1aSLin Ling pad2 = zio_buf_alloc(VDEV_PAD_SIZE); 726f83ffe1aSLin Ling bzero(pad2, VDEV_PAD_SIZE); 727f83ffe1aSLin Ling 728fa9e4066Sahrens /* 729fa9e4066Sahrens * Write everything in parallel. 730fa9e4066Sahrens */ 7318956713aSEric Schrock retry: 73217f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 733fa9e4066Sahrens 734e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 735fa9e4066Sahrens 736fa9e4066Sahrens vdev_label_write(zio, vd, l, vp, 737fa9e4066Sahrens offsetof(vdev_label_t, vl_vdev_phys), 73817f17c2dSbonwick sizeof (vdev_phys_t), NULL, NULL, flags); 739fa9e4066Sahrens 740f83ffe1aSLin Ling /* 741f83ffe1aSLin Ling * Skip the 1st padding area. 742f83ffe1aSLin Ling * Zero out the 2nd padding area where it might have 743f83ffe1aSLin Ling * left over data from previous filesystem format. 744f83ffe1aSLin Ling */ 745f83ffe1aSLin Ling vdev_label_write(zio, vd, l, pad2, 746f83ffe1aSLin Ling offsetof(vdev_label_t, vl_pad2), 747f83ffe1aSLin Ling VDEV_PAD_SIZE, NULL, NULL, flags); 748f83ffe1aSLin Ling 749f64c0e34SEric Taylor vdev_label_write(zio, vd, l, ub, 750f64c0e34SEric Taylor offsetof(vdev_label_t, vl_uberblock), 751f64c0e34SEric Taylor VDEV_UBERBLOCK_RING, NULL, NULL, flags); 752fa9e4066Sahrens } 753fa9e4066Sahrens 754fa9e4066Sahrens error = zio_wait(zio); 755fa9e4066Sahrens 7568956713aSEric Schrock if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { 7578956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 7588956713aSEric Schrock goto retry; 7598956713aSEric Schrock } 7608956713aSEric Schrock 761fa9e4066Sahrens nvlist_free(label); 762f83ffe1aSLin Ling zio_buf_free(pad2, VDEV_PAD_SIZE); 763f64c0e34SEric Taylor zio_buf_free(ub, VDEV_UBERBLOCK_RING); 764fa9e4066Sahrens zio_buf_free(vp, sizeof (vdev_phys_t)); 765fa9e4066Sahrens 76639c23413Seschrock /* 76739c23413Seschrock * If this vdev hasn't been previously identified as a spare, then we 7683d7072f8Seschrock * mark it as such only if a) we are labeling it as a spare, or b) it 769fa94a07fSbrendan * exists as a spare elsewhere in the system. Do the same for 770fa94a07fSbrendan * level 2 ARC devices. 77139c23413Seschrock */ 77239c23413Seschrock if (error == 0 && !vd->vdev_isspare && 77339c23413Seschrock (reason == VDEV_LABEL_SPARE || 77489a89ebfSlling spa_spare_exists(vd->vdev_guid, NULL, NULL))) 77539c23413Seschrock spa_spare_add(vd); 77699653d4eSeschrock 777fa94a07fSbrendan if (error == 0 && !vd->vdev_isl2cache && 778fa94a07fSbrendan (reason == VDEV_LABEL_L2CACHE || 779fa94a07fSbrendan spa_l2cache_exists(vd->vdev_guid, NULL))) 780fa94a07fSbrendan spa_l2cache_add(vd); 781fa94a07fSbrendan 78239c23413Seschrock return (error); 78399653d4eSeschrock } 78499653d4eSeschrock 785fa9e4066Sahrens /* 786fa9e4066Sahrens * ========================================================================== 787fa9e4066Sahrens * uberblock load/sync 788fa9e4066Sahrens * ========================================================================== 789fa9e4066Sahrens */ 790fa9e4066Sahrens 791fa9e4066Sahrens /* 792fa9e4066Sahrens * Consider the following situation: txg is safely synced to disk. We've 793fa9e4066Sahrens * written the first uberblock for txg + 1, and then we lose power. When we 794fa9e4066Sahrens * come back up, we fail to see the uberblock for txg + 1 because, say, 795fa9e4066Sahrens * it was on a mirrored device and the replica to which we wrote txg + 1 796fa9e4066Sahrens * is now offline. If we then make some changes and sync txg + 1, and then 797fa9e4066Sahrens * the missing replica comes back, then for a new seconds we'll have two 798fa9e4066Sahrens * conflicting uberblocks on disk with the same txg. The solution is simple: 799fa9e4066Sahrens * among uberblocks with equal txg, choose the one with the latest timestamp. 800fa9e4066Sahrens */ 801fa9e4066Sahrens static int 802fa9e4066Sahrens vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2) 803fa9e4066Sahrens { 804fa9e4066Sahrens if (ub1->ub_txg < ub2->ub_txg) 805fa9e4066Sahrens return (-1); 806fa9e4066Sahrens if (ub1->ub_txg > ub2->ub_txg) 807fa9e4066Sahrens return (1); 808fa9e4066Sahrens 809fa9e4066Sahrens if (ub1->ub_timestamp < ub2->ub_timestamp) 810fa9e4066Sahrens return (-1); 811fa9e4066Sahrens if (ub1->ub_timestamp > ub2->ub_timestamp) 812fa9e4066Sahrens return (1); 813fa9e4066Sahrens 814fa9e4066Sahrens return (0); 815fa9e4066Sahrens } 816fa9e4066Sahrens 817fa9e4066Sahrens static void 818fa9e4066Sahrens vdev_uberblock_load_done(zio_t *zio) 819fa9e4066Sahrens { 820468c413aSTim Haley spa_t *spa = zio->io_spa; 821e14bb325SJeff Bonwick zio_t *rio = zio->io_private; 822ecc2d604Sbonwick uberblock_t *ub = zio->io_data; 823e14bb325SJeff Bonwick uberblock_t *ubbest = rio->io_private; 824fa9e4066Sahrens 825ecc2d604Sbonwick ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(zio->io_vd)); 826fa9e4066Sahrens 827ea8dc4b6Seschrock if (zio->io_error == 0 && uberblock_verify(ub) == 0) { 828e14bb325SJeff Bonwick mutex_enter(&rio->io_lock); 829468c413aSTim Haley if (ub->ub_txg <= spa->spa_load_max_txg && 8302e551927SVictor Latushkin vdev_uberblock_compare(ub, ubbest) > 0) 831fa9e4066Sahrens *ubbest = *ub; 832e14bb325SJeff Bonwick mutex_exit(&rio->io_lock); 833fa9e4066Sahrens } 834fa9e4066Sahrens 835fa9e4066Sahrens zio_buf_free(zio->io_data, zio->io_size); 836fa9e4066Sahrens } 837fa9e4066Sahrens 838fa9e4066Sahrens void 839fa9e4066Sahrens vdev_uberblock_load(zio_t *zio, vdev_t *vd, uberblock_t *ubbest) 840fa9e4066Sahrens { 841e14bb325SJeff Bonwick spa_t *spa = vd->vdev_spa; 842e14bb325SJeff Bonwick vdev_t *rvd = spa->spa_root_vdev; 8438956713aSEric Schrock int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | 8448956713aSEric Schrock ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; 845e14bb325SJeff Bonwick 846e14bb325SJeff Bonwick if (vd == rvd) { 847e14bb325SJeff Bonwick ASSERT(zio == NULL); 848e14bb325SJeff Bonwick spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 849e14bb325SJeff Bonwick zio = zio_root(spa, NULL, ubbest, flags); 850e14bb325SJeff Bonwick bzero(ubbest, sizeof (uberblock_t)); 851e14bb325SJeff Bonwick } 852fa9e4066Sahrens 853e14bb325SJeff Bonwick ASSERT(zio != NULL); 854fa9e4066Sahrens 855e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 856e14bb325SJeff Bonwick vdev_uberblock_load(zio, vd->vdev_child[c], ubbest); 857fa9e4066Sahrens 858e14bb325SJeff Bonwick if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) { 859e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 860e14bb325SJeff Bonwick for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { 861e14bb325SJeff Bonwick vdev_label_read(zio, vd, l, 862e14bb325SJeff Bonwick zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)), 863e14bb325SJeff Bonwick VDEV_UBERBLOCK_OFFSET(vd, n), 864e14bb325SJeff Bonwick VDEV_UBERBLOCK_SIZE(vd), 865e14bb325SJeff Bonwick vdev_uberblock_load_done, zio, flags); 866e14bb325SJeff Bonwick } 867fa9e4066Sahrens } 868fa9e4066Sahrens } 869e14bb325SJeff Bonwick 870e14bb325SJeff Bonwick if (vd == rvd) { 871e14bb325SJeff Bonwick (void) zio_wait(zio); 872e14bb325SJeff Bonwick spa_config_exit(spa, SCL_ALL, FTAG); 873e14bb325SJeff Bonwick } 874fa9e4066Sahrens } 875fa9e4066Sahrens 876fa9e4066Sahrens /* 87717f17c2dSbonwick * On success, increment root zio's count of good writes. 8780373e76bSbonwick * We only get credit for writes to known-visible vdevs; see spa_vdev_add(). 879fa9e4066Sahrens */ 880fa9e4066Sahrens static void 881fa9e4066Sahrens vdev_uberblock_sync_done(zio_t *zio) 882fa9e4066Sahrens { 88317f17c2dSbonwick uint64_t *good_writes = zio->io_private; 884fa9e4066Sahrens 8850373e76bSbonwick if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0) 886fa9e4066Sahrens atomic_add_64(good_writes, 1); 887fa9e4066Sahrens } 888fa9e4066Sahrens 88917f17c2dSbonwick /* 89017f17c2dSbonwick * Write the uberblock to all labels of all leaves of the specified vdev. 89117f17c2dSbonwick */ 892fa9e4066Sahrens static void 893e14bb325SJeff Bonwick vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags) 894fa9e4066Sahrens { 89517f17c2dSbonwick uberblock_t *ubbuf; 896e14bb325SJeff Bonwick int n; 897fa9e4066Sahrens 898e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 899e14bb325SJeff Bonwick vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags); 900fa9e4066Sahrens 901fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) 902fa9e4066Sahrens return; 903fa9e4066Sahrens 904e14bb325SJeff Bonwick if (!vdev_writeable(vd)) 905fa9e4066Sahrens return; 906fa9e4066Sahrens 90717f17c2dSbonwick n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1); 908fa9e4066Sahrens 90917f17c2dSbonwick ubbuf = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)); 91017f17c2dSbonwick bzero(ubbuf, VDEV_UBERBLOCK_SIZE(vd)); 91117f17c2dSbonwick *ubbuf = *ub; 912fa9e4066Sahrens 913e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) 91417f17c2dSbonwick vdev_label_write(zio, vd, l, ubbuf, 915e14bb325SJeff Bonwick VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), 91617f17c2dSbonwick vdev_uberblock_sync_done, zio->io_private, 917e14bb325SJeff Bonwick flags | ZIO_FLAG_DONT_PROPAGATE); 918fa9e4066Sahrens 91917f17c2dSbonwick zio_buf_free(ubbuf, VDEV_UBERBLOCK_SIZE(vd)); 920fa9e4066Sahrens } 921fa9e4066Sahrens 92217f17c2dSbonwick int 92317f17c2dSbonwick vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags) 924fa9e4066Sahrens { 92517f17c2dSbonwick spa_t *spa = svd[0]->vdev_spa; 926e14bb325SJeff Bonwick zio_t *zio; 92717f17c2dSbonwick uint64_t good_writes = 0; 928fa9e4066Sahrens 929e14bb325SJeff Bonwick zio = zio_root(spa, NULL, &good_writes, flags); 930e14bb325SJeff Bonwick 931e14bb325SJeff Bonwick for (int v = 0; v < svdcount; v++) 932e14bb325SJeff Bonwick vdev_uberblock_sync(zio, ub, svd[v], flags); 933fa9e4066Sahrens 93417f17c2dSbonwick (void) zio_wait(zio); 935fa9e4066Sahrens 936fa9e4066Sahrens /* 93717f17c2dSbonwick * Flush the uberblocks to disk. This ensures that the odd labels 93817f17c2dSbonwick * are no longer needed (because the new uberblocks and the even 93917f17c2dSbonwick * labels are safely on disk), so it is safe to overwrite them. 940fa9e4066Sahrens */ 94117f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 942fa9e4066Sahrens 943e14bb325SJeff Bonwick for (int v = 0; v < svdcount; v++) 94417f17c2dSbonwick zio_flush(zio, svd[v]); 945fa9e4066Sahrens 94617f17c2dSbonwick (void) zio_wait(zio); 94717f17c2dSbonwick 94817f17c2dSbonwick return (good_writes >= 1 ? 0 : EIO); 949fa9e4066Sahrens } 950fa9e4066Sahrens 951fa9e4066Sahrens /* 95217f17c2dSbonwick * On success, increment the count of good writes for our top-level vdev. 953fa9e4066Sahrens */ 954fa9e4066Sahrens static void 95517f17c2dSbonwick vdev_label_sync_done(zio_t *zio) 956fa9e4066Sahrens { 95717f17c2dSbonwick uint64_t *good_writes = zio->io_private; 958fa9e4066Sahrens 959fa9e4066Sahrens if (zio->io_error == 0) 960fa9e4066Sahrens atomic_add_64(good_writes, 1); 961fa9e4066Sahrens } 962fa9e4066Sahrens 96317f17c2dSbonwick /* 96417f17c2dSbonwick * If there weren't enough good writes, indicate failure to the parent. 96517f17c2dSbonwick */ 966fa9e4066Sahrens static void 96717f17c2dSbonwick vdev_label_sync_top_done(zio_t *zio) 96817f17c2dSbonwick { 96917f17c2dSbonwick uint64_t *good_writes = zio->io_private; 97017f17c2dSbonwick 97117f17c2dSbonwick if (*good_writes == 0) 97217f17c2dSbonwick zio->io_error = EIO; 97317f17c2dSbonwick 97417f17c2dSbonwick kmem_free(good_writes, sizeof (uint64_t)); 97517f17c2dSbonwick } 97617f17c2dSbonwick 97751ece835Seschrock /* 9780430f8daSeschrock * We ignore errors for log and cache devices, simply free the private data. 97951ece835Seschrock */ 98051ece835Seschrock static void 9810430f8daSeschrock vdev_label_sync_ignore_done(zio_t *zio) 98251ece835Seschrock { 98351ece835Seschrock kmem_free(zio->io_private, sizeof (uint64_t)); 98451ece835Seschrock } 98551ece835Seschrock 98617f17c2dSbonwick /* 98717f17c2dSbonwick * Write all even or odd labels to all leaves of the specified vdev. 98817f17c2dSbonwick */ 98917f17c2dSbonwick static void 990e14bb325SJeff Bonwick vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags) 991fa9e4066Sahrens { 992fa9e4066Sahrens nvlist_t *label; 993fa9e4066Sahrens vdev_phys_t *vp; 994fa9e4066Sahrens char *buf; 995fa9e4066Sahrens size_t buflen; 996fa9e4066Sahrens 997e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 998e14bb325SJeff Bonwick vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags); 999fa9e4066Sahrens 1000fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) 1001fa9e4066Sahrens return; 1002fa9e4066Sahrens 1003e14bb325SJeff Bonwick if (!vdev_writeable(vd)) 1004fa9e4066Sahrens return; 1005fa9e4066Sahrens 1006fa9e4066Sahrens /* 1007fa9e4066Sahrens * Generate a label describing the top-level config to which we belong. 1008fa9e4066Sahrens */ 10090373e76bSbonwick label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE); 1010fa9e4066Sahrens 1011fa9e4066Sahrens vp = zio_buf_alloc(sizeof (vdev_phys_t)); 1012fa9e4066Sahrens bzero(vp, sizeof (vdev_phys_t)); 1013fa9e4066Sahrens 1014fa9e4066Sahrens buf = vp->vp_nvlist; 1015fa9e4066Sahrens buflen = sizeof (vp->vp_nvlist); 1016fa9e4066Sahrens 101717f17c2dSbonwick if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) { 101817f17c2dSbonwick for (; l < VDEV_LABELS; l += 2) { 101917f17c2dSbonwick vdev_label_write(zio, vd, l, vp, 102017f17c2dSbonwick offsetof(vdev_label_t, vl_vdev_phys), 102117f17c2dSbonwick sizeof (vdev_phys_t), 102217f17c2dSbonwick vdev_label_sync_done, zio->io_private, 1023e14bb325SJeff Bonwick flags | ZIO_FLAG_DONT_PROPAGATE); 102417f17c2dSbonwick } 102517f17c2dSbonwick } 1026fa9e4066Sahrens 1027fa9e4066Sahrens zio_buf_free(vp, sizeof (vdev_phys_t)); 1028fa9e4066Sahrens nvlist_free(label); 1029fa9e4066Sahrens } 1030fa9e4066Sahrens 103117f17c2dSbonwick int 1032e14bb325SJeff Bonwick vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags) 1033fa9e4066Sahrens { 1034e14bb325SJeff Bonwick list_t *dl = &spa->spa_config_dirty_list; 103517f17c2dSbonwick vdev_t *vd; 1036e14bb325SJeff Bonwick zio_t *zio; 1037fa9e4066Sahrens int error; 1038fa9e4066Sahrens 1039fa9e4066Sahrens /* 1040e14bb325SJeff Bonwick * Write the new labels to disk. 1041fa9e4066Sahrens */ 1042e14bb325SJeff Bonwick zio = zio_root(spa, NULL, NULL, flags); 1043fa9e4066Sahrens 104417f17c2dSbonwick for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) { 104517f17c2dSbonwick uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t), 104617f17c2dSbonwick KM_SLEEP); 104788ecc943SGeorge Wilson 104888ecc943SGeorge Wilson ASSERT(!vd->vdev_ishole); 104988ecc943SGeorge Wilson 1050a3f829aeSBill Moore zio_t *vio = zio_null(zio, spa, NULL, 10510430f8daSeschrock (vd->vdev_islog || vd->vdev_aux != NULL) ? 10520430f8daSeschrock vdev_label_sync_ignore_done : vdev_label_sync_top_done, 105317f17c2dSbonwick good_writes, flags); 1054e14bb325SJeff Bonwick vdev_label_sync(vio, vd, l, txg, flags); 105517f17c2dSbonwick zio_nowait(vio); 1056fa9e4066Sahrens } 1057e14bb325SJeff Bonwick 1058e14bb325SJeff Bonwick error = zio_wait(zio); 1059fa9e4066Sahrens 10608654d025Sperrin /* 106117f17c2dSbonwick * Flush the new labels to disk. 10628654d025Sperrin */ 106317f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 10648654d025Sperrin 106517f17c2dSbonwick for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) 106617f17c2dSbonwick zio_flush(zio, vd); 106717f17c2dSbonwick 106817f17c2dSbonwick (void) zio_wait(zio); 1069fa9e4066Sahrens 1070fa9e4066Sahrens return (error); 1071fa9e4066Sahrens } 1072fa9e4066Sahrens 1073fa9e4066Sahrens /* 107417f17c2dSbonwick * Sync the uberblock and any changes to the vdev configuration. 1075fa9e4066Sahrens * 1076fa9e4066Sahrens * The order of operations is carefully crafted to ensure that 1077fa9e4066Sahrens * if the system panics or loses power at any time, the state on disk 1078fa9e4066Sahrens * is still transactionally consistent. The in-line comments below 1079fa9e4066Sahrens * describe the failure semantics at each stage. 1080fa9e4066Sahrens * 108117f17c2dSbonwick * Moreover, vdev_config_sync() is designed to be idempotent: if it fails 1082fa9e4066Sahrens * at any time, you can just call it again, and it will resume its work. 1083fa9e4066Sahrens */ 1084e14bb325SJeff Bonwick int 10858956713aSEric Schrock vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg, boolean_t tryhard) 1086fa9e4066Sahrens { 108717f17c2dSbonwick spa_t *spa = svd[0]->vdev_spa; 1088fa9e4066Sahrens uberblock_t *ub = &spa->spa_uberblock; 10890373e76bSbonwick vdev_t *vd; 1090fa9e4066Sahrens zio_t *zio; 1091e14bb325SJeff Bonwick int error; 1092e14bb325SJeff Bonwick int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; 1093fa9e4066Sahrens 10948956713aSEric Schrock /* 10958956713aSEric Schrock * Normally, we don't want to try too hard to write every label and 10968956713aSEric Schrock * uberblock. If there is a flaky disk, we don't want the rest of the 10978956713aSEric Schrock * sync process to block while we retry. But if we can't write a 10988956713aSEric Schrock * single label out, we should retry with ZIO_FLAG_TRYHARD before 10998956713aSEric Schrock * bailing out and declaring the pool faulted. 11008956713aSEric Schrock */ 11018956713aSEric Schrock if (tryhard) 11028956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 11038956713aSEric Schrock 1104fa9e4066Sahrens ASSERT(ub->ub_txg <= txg); 1105fa9e4066Sahrens 1106fa9e4066Sahrens /* 110717f17c2dSbonwick * If this isn't a resync due to I/O errors, 110817f17c2dSbonwick * and nothing changed in this transaction group, 110917f17c2dSbonwick * and the vdev configuration hasn't changed, 11100373e76bSbonwick * then there's nothing to do. 1111fa9e4066Sahrens */ 111217f17c2dSbonwick if (ub->ub_txg < txg && 111317f17c2dSbonwick uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE && 1114e14bb325SJeff Bonwick list_is_empty(&spa->spa_config_dirty_list)) 1115e14bb325SJeff Bonwick return (0); 1116fa9e4066Sahrens 1117fa9e4066Sahrens if (txg > spa_freeze_txg(spa)) 1118e14bb325SJeff Bonwick return (0); 1119fa9e4066Sahrens 11200373e76bSbonwick ASSERT(txg <= spa->spa_final_txg); 11210373e76bSbonwick 1122fa9e4066Sahrens /* 1123fa9e4066Sahrens * Flush the write cache of every disk that's been written to 1124fa9e4066Sahrens * in this transaction group. This ensures that all blocks 1125fa9e4066Sahrens * written in this txg will be committed to stable storage 1126fa9e4066Sahrens * before any uberblock that references them. 1127fa9e4066Sahrens */ 112817f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 112917f17c2dSbonwick 1130fa9e4066Sahrens for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd; 113117f17c2dSbonwick vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg))) 113217f17c2dSbonwick zio_flush(zio, vd); 113317f17c2dSbonwick 1134fa9e4066Sahrens (void) zio_wait(zio); 1135fa9e4066Sahrens 1136fa9e4066Sahrens /* 1137fa9e4066Sahrens * Sync out the even labels (L0, L2) for every dirty vdev. If the 1138fa9e4066Sahrens * system dies in the middle of this process, that's OK: all of the 1139fa9e4066Sahrens * even labels that made it to disk will be newer than any uberblock, 1140fa9e4066Sahrens * and will therefore be considered invalid. The odd labels (L1, L3), 114117f17c2dSbonwick * which have not yet been touched, will still be valid. We flush 114217f17c2dSbonwick * the new labels to disk to ensure that all even-label updates 114317f17c2dSbonwick * are committed to stable storage before the uberblock update. 1144fa9e4066Sahrens */ 1145e14bb325SJeff Bonwick if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) 1146e14bb325SJeff Bonwick return (error); 1147fa9e4066Sahrens 1148fa9e4066Sahrens /* 1149e14bb325SJeff Bonwick * Sync the uberblocks to all vdevs in svd[]. 11500373e76bSbonwick * If the system dies in the middle of this step, there are two cases 11510373e76bSbonwick * to consider, and the on-disk state is consistent either way: 1152fa9e4066Sahrens * 1153fa9e4066Sahrens * (1) If none of the new uberblocks made it to disk, then the 1154fa9e4066Sahrens * previous uberblock will be the newest, and the odd labels 1155fa9e4066Sahrens * (which had not yet been touched) will be valid with respect 1156fa9e4066Sahrens * to that uberblock. 1157fa9e4066Sahrens * 1158fa9e4066Sahrens * (2) If one or more new uberblocks made it to disk, then they 1159fa9e4066Sahrens * will be the newest, and the even labels (which had all 1160fa9e4066Sahrens * been successfully committed) will be valid with respect 1161fa9e4066Sahrens * to the new uberblocks. 1162fa9e4066Sahrens */ 1163e14bb325SJeff Bonwick if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) 1164e14bb325SJeff Bonwick return (error); 1165fa9e4066Sahrens 1166fa9e4066Sahrens /* 1167fa9e4066Sahrens * Sync out odd labels for every dirty vdev. If the system dies 1168fa9e4066Sahrens * in the middle of this process, the even labels and the new 1169fa9e4066Sahrens * uberblocks will suffice to open the pool. The next time 1170fa9e4066Sahrens * the pool is opened, the first thing we'll do -- before any 1171fa9e4066Sahrens * user data is modified -- is mark every vdev dirty so that 117217f17c2dSbonwick * all labels will be brought up to date. We flush the new labels 117317f17c2dSbonwick * to disk to ensure that all odd-label updates are committed to 117417f17c2dSbonwick * stable storage before the next transaction group begins. 1175fa9e4066Sahrens */ 1176e14bb325SJeff Bonwick return (vdev_label_sync_list(spa, 1, txg, flags)); 1177fa9e4066Sahrens } 1178