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 */ 21ad135b5dSChristopher Siden 22fa9e4066Sahrens /* 233f9d6ad7SLin Ling * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 2486714001SSerapheim Dimitropoulos * Copyright (c) 2012, 2018 by Delphix. All rights reserved. 25*e0f1c0afSOlaf Faaland * Copyright 2019 Joyent, Inc. 26fa9e4066Sahrens */ 27fa9e4066Sahrens 28fa9e4066Sahrens /* 29fa9e4066Sahrens * Virtual Device Labels 30fa9e4066Sahrens * --------------------- 31fa9e4066Sahrens * 32fa9e4066Sahrens * The vdev label serves several distinct purposes: 33fa9e4066Sahrens * 34fa9e4066Sahrens * 1. Uniquely identify this device as part of a ZFS pool and confirm its 35fa9e4066Sahrens * identity within the pool. 36fa9e4066Sahrens * 37cfd63e1bSMatthew Ahrens * 2. Verify that all the devices given in a configuration are present 38fa9e4066Sahrens * within the pool. 39fa9e4066Sahrens * 40cfd63e1bSMatthew Ahrens * 3. Determine the uberblock for the pool. 41fa9e4066Sahrens * 42cfd63e1bSMatthew Ahrens * 4. In case of an import operation, determine the configuration of the 43fa9e4066Sahrens * toplevel vdev of which it is a part. 44fa9e4066Sahrens * 45cfd63e1bSMatthew Ahrens * 5. If an import operation cannot find all the devices in the pool, 46fa9e4066Sahrens * provide enough information to the administrator to determine which 47fa9e4066Sahrens * devices are missing. 48fa9e4066Sahrens * 49fa9e4066Sahrens * It is important to note that while the kernel is responsible for writing the 50fa9e4066Sahrens * label, it only consumes the information in the first three cases. The 51fa9e4066Sahrens * latter information is only consumed in userland when determining the 52fa9e4066Sahrens * configuration to import a pool. 53fa9e4066Sahrens * 54fa9e4066Sahrens * 55fa9e4066Sahrens * Label Organization 56fa9e4066Sahrens * ------------------ 57fa9e4066Sahrens * 58fa9e4066Sahrens * Before describing the contents of the label, it's important to understand how 59fa9e4066Sahrens * the labels are written and updated with respect to the uberblock. 60fa9e4066Sahrens * 61fa9e4066Sahrens * When the pool configuration is altered, either because it was newly created 62fa9e4066Sahrens * or a device was added, we want to update all the labels such that we can deal 63fa9e4066Sahrens * with fatal failure at any point. To this end, each disk has two labels which 64fa9e4066Sahrens * are updated before and after the uberblock is synced. Assuming we have 653d7072f8Seschrock * labels and an uberblock with the following transaction groups: 66fa9e4066Sahrens * 67fa9e4066Sahrens * L1 UB L2 68fa9e4066Sahrens * +------+ +------+ +------+ 69fa9e4066Sahrens * | | | | | | 70fa9e4066Sahrens * | t10 | | t10 | | t10 | 71fa9e4066Sahrens * | | | | | | 72fa9e4066Sahrens * +------+ +------+ +------+ 73fa9e4066Sahrens * 74fa9e4066Sahrens * In this stable state, the labels and the uberblock were all updated within 75fa9e4066Sahrens * the same transaction group (10). Each label is mirrored and checksummed, so 76fa9e4066Sahrens * that we can detect when we fail partway through writing the label. 77fa9e4066Sahrens * 78fa9e4066Sahrens * In order to identify which labels are valid, the labels are written in the 79fa9e4066Sahrens * following manner: 80fa9e4066Sahrens * 81cfd63e1bSMatthew Ahrens * 1. For each vdev, update 'L1' to the new label 82cfd63e1bSMatthew Ahrens * 2. Update the uberblock 83cfd63e1bSMatthew Ahrens * 3. For each vdev, update 'L2' to the new label 84fa9e4066Sahrens * 85fa9e4066Sahrens * Given arbitrary failure, we can determine the correct label to use based on 86fa9e4066Sahrens * the transaction group. If we fail after updating L1 but before updating the 87fa9e4066Sahrens * UB, we will notice that L1's transaction group is greater than the uberblock, 88fa9e4066Sahrens * so L2 must be valid. If we fail after writing the uberblock but before 89fa9e4066Sahrens * writing L2, we will notice that L2's transaction group is less than L1, and 90fa9e4066Sahrens * therefore L1 is valid. 91fa9e4066Sahrens * 92fa9e4066Sahrens * Another added complexity is that not every label is updated when the config 93fa9e4066Sahrens * is synced. If we add a single device, we do not want to have to re-write 94fa9e4066Sahrens * every label for every device in the pool. This means that both L1 and L2 may 95fa9e4066Sahrens * be older than the pool uberblock, because the necessary information is stored 96fa9e4066Sahrens * on another vdev. 97fa9e4066Sahrens * 98fa9e4066Sahrens * 99fa9e4066Sahrens * On-disk Format 100fa9e4066Sahrens * -------------- 101fa9e4066Sahrens * 102fa9e4066Sahrens * The vdev label consists of two distinct parts, and is wrapped within the 103fa9e4066Sahrens * vdev_label_t structure. The label includes 8k of padding to permit legacy 104fa9e4066Sahrens * VTOC disk labels, but is otherwise ignored. 105fa9e4066Sahrens * 106fa9e4066Sahrens * The first half of the label is a packed nvlist which contains pool wide 107fa9e4066Sahrens * properties, per-vdev properties, and configuration information. It is 108fa9e4066Sahrens * described in more detail below. 109fa9e4066Sahrens * 110fa9e4066Sahrens * The latter half of the label consists of a redundant array of uberblocks. 111fa9e4066Sahrens * These uberblocks are updated whenever a transaction group is committed, 112fa9e4066Sahrens * or when the configuration is updated. When a pool is loaded, we scan each 113fa9e4066Sahrens * vdev for the 'best' uberblock. 114fa9e4066Sahrens * 115fa9e4066Sahrens * 116fa9e4066Sahrens * Configuration Information 117fa9e4066Sahrens * ------------------------- 118fa9e4066Sahrens * 119fa9e4066Sahrens * The nvlist describing the pool and vdev contains the following elements: 120fa9e4066Sahrens * 121cfd63e1bSMatthew Ahrens * version ZFS on-disk version 122cfd63e1bSMatthew Ahrens * name Pool name 123cfd63e1bSMatthew Ahrens * state Pool state 124cfd63e1bSMatthew Ahrens * txg Transaction group in which this label was written 125cfd63e1bSMatthew Ahrens * pool_guid Unique identifier for this pool 126cfd63e1bSMatthew Ahrens * vdev_tree An nvlist describing vdev tree. 127ad135b5dSChristopher Siden * features_for_read 128ad135b5dSChristopher Siden * An nvlist of the features necessary for reading the MOS. 129fa9e4066Sahrens * 130fa9e4066Sahrens * Each leaf device label also contains the following: 131fa9e4066Sahrens * 132cfd63e1bSMatthew Ahrens * top_guid Unique ID for top-level vdev in which this is contained 133cfd63e1bSMatthew Ahrens * guid Unique ID for the leaf vdev 134fa9e4066Sahrens * 135fa9e4066Sahrens * The 'vs' configuration follows the format described in 'spa_config.c'. 136fa9e4066Sahrens */ 137fa9e4066Sahrens 138fa9e4066Sahrens #include <sys/zfs_context.h> 139fa9e4066Sahrens #include <sys/spa.h> 140fa9e4066Sahrens #include <sys/spa_impl.h> 141fa9e4066Sahrens #include <sys/dmu.h> 142fa9e4066Sahrens #include <sys/zap.h> 143fa9e4066Sahrens #include <sys/vdev.h> 144fa9e4066Sahrens #include <sys/vdev_impl.h> 145fa9e4066Sahrens #include <sys/uberblock_impl.h> 146fa9e4066Sahrens #include <sys/metaslab.h> 1475cabbc6bSPrashanth Sreenivasa #include <sys/metaslab_impl.h> 148fa9e4066Sahrens #include <sys/zio.h> 1493f9d6ad7SLin Ling #include <sys/dsl_scan.h> 150770499e1SDan Kimmel #include <sys/abd.h> 151fa9e4066Sahrens #include <sys/fs/zfs.h> 152fa9e4066Sahrens 153fa9e4066Sahrens /* 154fa9e4066Sahrens * Basic routines to read and write from a vdev label. 155fa9e4066Sahrens * Used throughout the rest of this file. 156fa9e4066Sahrens */ 157fa9e4066Sahrens uint64_t 158fa9e4066Sahrens vdev_label_offset(uint64_t psize, int l, uint64_t offset) 159fa9e4066Sahrens { 160ecc2d604Sbonwick ASSERT(offset < sizeof (vdev_label_t)); 161e7437265Sahrens ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); 162ecc2d604Sbonwick 163fa9e4066Sahrens return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 164fa9e4066Sahrens 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); 165fa9e4066Sahrens } 166fa9e4066Sahrens 16721bf64a7Sgw /* 16821bf64a7Sgw * Returns back the vdev label associated with the passed in offset. 16921bf64a7Sgw */ 17021bf64a7Sgw int 17121bf64a7Sgw vdev_label_number(uint64_t psize, uint64_t offset) 17221bf64a7Sgw { 17321bf64a7Sgw int l; 17421bf64a7Sgw 17521bf64a7Sgw if (offset >= psize - VDEV_LABEL_END_SIZE) { 17621bf64a7Sgw offset -= psize - VDEV_LABEL_END_SIZE; 17721bf64a7Sgw offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t); 17821bf64a7Sgw } 17921bf64a7Sgw l = offset / sizeof (vdev_label_t); 18021bf64a7Sgw return (l < VDEV_LABELS ? l : -1); 18121bf64a7Sgw } 18221bf64a7Sgw 183fa9e4066Sahrens static void 184770499e1SDan Kimmel vdev_label_read(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, 1859a686fbcSPaul Dagnelie uint64_t size, zio_done_func_t *done, void *private, int flags) 186fa9e4066Sahrens { 187e14bb325SJeff Bonwick ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) == 188e14bb325SJeff Bonwick SCL_STATE_ALL); 189e14bb325SJeff Bonwick ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); 190fa9e4066Sahrens 191fa9e4066Sahrens zio_nowait(zio_read_phys(zio, vd, 192fa9e4066Sahrens vdev_label_offset(vd->vdev_psize, l, offset), 193fa9e4066Sahrens size, buf, ZIO_CHECKSUM_LABEL, done, private, 194e14bb325SJeff Bonwick ZIO_PRIORITY_SYNC_READ, flags, B_TRUE)); 195fa9e4066Sahrens } 196fa9e4066Sahrens 197*e0f1c0afSOlaf Faaland void 198770499e1SDan Kimmel vdev_label_write(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, 1999a686fbcSPaul Dagnelie uint64_t size, zio_done_func_t *done, void *private, int flags) 200fa9e4066Sahrens { 201*e0f1c0afSOlaf Faaland #ifdef _KERNEL 202*e0f1c0afSOlaf Faaland /* 203*e0f1c0afSOlaf Faaland * This assert is invalid in the user-level ztest MMP code because 204*e0f1c0afSOlaf Faaland * the ztest thread is not in dsl_pool_sync_context. ZoL does not 205*e0f1c0afSOlaf Faaland * build the user-level code with DEBUG so this is not an issue there. 206*e0f1c0afSOlaf Faaland */ 207e14bb325SJeff Bonwick ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL || 208e14bb325SJeff Bonwick (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) == 209e14bb325SJeff Bonwick (SCL_CONFIG | SCL_STATE) && 210e14bb325SJeff Bonwick dsl_pool_sync_context(spa_get_dsl(zio->io_spa)))); 211*e0f1c0afSOlaf Faaland #endif 212e14bb325SJeff Bonwick ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); 213fa9e4066Sahrens 214fa9e4066Sahrens zio_nowait(zio_write_phys(zio, vd, 215fa9e4066Sahrens vdev_label_offset(vd->vdev_psize, l, offset), 216fa9e4066Sahrens size, buf, ZIO_CHECKSUM_LABEL, done, private, 21717f17c2dSbonwick ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE)); 218fa9e4066Sahrens } 219fa9e4066Sahrens 22086714001SSerapheim Dimitropoulos static void 22186714001SSerapheim Dimitropoulos root_vdev_actions_getprogress(vdev_t *vd, nvlist_t *nvl) 22286714001SSerapheim Dimitropoulos { 22386714001SSerapheim Dimitropoulos spa_t *spa = vd->vdev_spa; 22486714001SSerapheim Dimitropoulos 22586714001SSerapheim Dimitropoulos if (vd != spa->spa_root_vdev) 22686714001SSerapheim Dimitropoulos return; 22786714001SSerapheim Dimitropoulos 22886714001SSerapheim Dimitropoulos /* provide either current or previous scan information */ 22986714001SSerapheim Dimitropoulos pool_scan_stat_t ps; 23086714001SSerapheim Dimitropoulos if (spa_scan_get_stats(spa, &ps) == 0) { 23186714001SSerapheim Dimitropoulos fnvlist_add_uint64_array(nvl, 23286714001SSerapheim Dimitropoulos ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps, 23386714001SSerapheim Dimitropoulos sizeof (pool_scan_stat_t) / sizeof (uint64_t)); 23486714001SSerapheim Dimitropoulos } 23586714001SSerapheim Dimitropoulos 23686714001SSerapheim Dimitropoulos pool_removal_stat_t prs; 23786714001SSerapheim Dimitropoulos if (spa_removal_get_stats(spa, &prs) == 0) { 23886714001SSerapheim Dimitropoulos fnvlist_add_uint64_array(nvl, 23986714001SSerapheim Dimitropoulos ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t *)&prs, 24086714001SSerapheim Dimitropoulos sizeof (prs) / sizeof (uint64_t)); 24186714001SSerapheim Dimitropoulos } 24286714001SSerapheim Dimitropoulos 24386714001SSerapheim Dimitropoulos pool_checkpoint_stat_t pcs; 24486714001SSerapheim Dimitropoulos if (spa_checkpoint_get_stats(spa, &pcs) == 0) { 24586714001SSerapheim Dimitropoulos fnvlist_add_uint64_array(nvl, 24686714001SSerapheim Dimitropoulos ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t *)&pcs, 24786714001SSerapheim Dimitropoulos sizeof (pcs) / sizeof (uint64_t)); 24886714001SSerapheim Dimitropoulos } 24986714001SSerapheim Dimitropoulos } 25086714001SSerapheim Dimitropoulos 251fa9e4066Sahrens /* 252fa9e4066Sahrens * Generate the nvlist representing this vdev's config. 253fa9e4066Sahrens */ 254fa9e4066Sahrens nvlist_t * 25599653d4eSeschrock vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, 2563f9d6ad7SLin Ling vdev_config_flag_t flags) 257fa9e4066Sahrens { 258fa9e4066Sahrens nvlist_t *nv = NULL; 2595cabbc6bSPrashanth Sreenivasa vdev_indirect_config_t *vic = &vd->vdev_indirect_config; 260fa9e4066Sahrens 261b4952e17SGeorge Wilson nv = fnvlist_alloc(); 262fa9e4066Sahrens 263b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type); 2643f9d6ad7SLin Ling if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE))) 265b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id); 266b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid); 267fa9e4066Sahrens 268fa9e4066Sahrens if (vd->vdev_path != NULL) 269b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path); 270fa9e4066Sahrens 271fa9e4066Sahrens if (vd->vdev_devid != NULL) 272b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid); 273fa9e4066Sahrens 2743d7072f8Seschrock if (vd->vdev_physpath != NULL) 275b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, 276b4952e17SGeorge Wilson vd->vdev_physpath); 2773d7072f8Seschrock 2786809eb4eSEric Schrock if (vd->vdev_fru != NULL) 279b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru); 2806809eb4eSEric Schrock 28199653d4eSeschrock if (vd->vdev_nparity != 0) { 28299653d4eSeschrock ASSERT(strcmp(vd->vdev_ops->vdev_op_type, 28399653d4eSeschrock VDEV_TYPE_RAIDZ) == 0); 28499653d4eSeschrock 28599653d4eSeschrock /* 28699653d4eSeschrock * Make sure someone hasn't managed to sneak a fancy new vdev 28799653d4eSeschrock * into a crufty old storage pool. 28899653d4eSeschrock */ 28999653d4eSeschrock ASSERT(vd->vdev_nparity == 1 || 290f94275ceSAdam Leventhal (vd->vdev_nparity <= 2 && 291f94275ceSAdam Leventhal spa_version(spa) >= SPA_VERSION_RAIDZ2) || 292f94275ceSAdam Leventhal (vd->vdev_nparity <= 3 && 293f94275ceSAdam Leventhal spa_version(spa) >= SPA_VERSION_RAIDZ3)); 29499653d4eSeschrock 29599653d4eSeschrock /* 29699653d4eSeschrock * Note that we'll add the nparity tag even on storage pools 29799653d4eSeschrock * that only support a single parity device -- older software 29899653d4eSeschrock * will just ignore it. 29999653d4eSeschrock */ 300b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vd->vdev_nparity); 30199653d4eSeschrock } 30299653d4eSeschrock 303afefbcddSeschrock if (vd->vdev_wholedisk != -1ULL) 304b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 305b4952e17SGeorge Wilson vd->vdev_wholedisk); 306afefbcddSeschrock 3076f793812SPavel Zakharov if (vd->vdev_not_present && !(flags & VDEV_CONFIG_MISSING)) 308b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1); 309ea8dc4b6Seschrock 31099653d4eSeschrock if (vd->vdev_isspare) 311b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1); 31299653d4eSeschrock 3133f9d6ad7SLin Ling if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) && 3143f9d6ad7SLin Ling vd == vd->vdev_top) { 315b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY, 316b4952e17SGeorge Wilson vd->vdev_ms_array); 317b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT, 318b4952e17SGeorge Wilson vd->vdev_ms_shift); 319b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift); 320b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE, 321b4952e17SGeorge Wilson vd->vdev_asize); 322b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog); 3235cabbc6bSPrashanth Sreenivasa if (vd->vdev_removing) { 324b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING, 325b4952e17SGeorge Wilson vd->vdev_removing); 3265cabbc6bSPrashanth Sreenivasa } 327fa9e4066Sahrens } 328fa9e4066Sahrens 3290713e232SGeorge Wilson if (vd->vdev_dtl_sm != NULL) { 330b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL, 3310713e232SGeorge Wilson space_map_object(vd->vdev_dtl_sm)); 3320713e232SGeorge Wilson } 333fa9e4066Sahrens 3345cabbc6bSPrashanth Sreenivasa if (vic->vic_mapping_object != 0) { 3355cabbc6bSPrashanth Sreenivasa fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT, 3365cabbc6bSPrashanth Sreenivasa vic->vic_mapping_object); 3375cabbc6bSPrashanth Sreenivasa } 3385cabbc6bSPrashanth Sreenivasa 3395cabbc6bSPrashanth Sreenivasa if (vic->vic_births_object != 0) { 3405cabbc6bSPrashanth Sreenivasa fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS, 3415cabbc6bSPrashanth Sreenivasa vic->vic_births_object); 3425cabbc6bSPrashanth Sreenivasa } 3435cabbc6bSPrashanth Sreenivasa 3445cabbc6bSPrashanth Sreenivasa if (vic->vic_prev_indirect_vdev != UINT64_MAX) { 3455cabbc6bSPrashanth Sreenivasa fnvlist_add_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV, 3465cabbc6bSPrashanth Sreenivasa vic->vic_prev_indirect_vdev); 3475cabbc6bSPrashanth Sreenivasa } 3485cabbc6bSPrashanth Sreenivasa 34988ecc943SGeorge Wilson if (vd->vdev_crtxg) 350b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg); 35188ecc943SGeorge Wilson 352215198a6SJoe Stein if (flags & VDEV_CONFIG_MOS) { 353215198a6SJoe Stein if (vd->vdev_leaf_zap != 0) { 354215198a6SJoe Stein ASSERT(vd->vdev_ops->vdev_op_leaf); 355215198a6SJoe Stein fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_LEAF_ZAP, 356215198a6SJoe Stein vd->vdev_leaf_zap); 357215198a6SJoe Stein } 358215198a6SJoe Stein 359215198a6SJoe Stein if (vd->vdev_top_zap != 0) { 360215198a6SJoe Stein ASSERT(vd == vd->vdev_top); 361215198a6SJoe Stein fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP, 362215198a6SJoe Stein vd->vdev_top_zap); 363215198a6SJoe Stein } 364215198a6SJoe Stein } 365215198a6SJoe Stein 366fa9e4066Sahrens if (getstats) { 367fa9e4066Sahrens vdev_stat_t vs; 3683f9d6ad7SLin Ling 369fa9e4066Sahrens vdev_get_stats(vd, &vs); 370b4952e17SGeorge Wilson fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 371b4952e17SGeorge Wilson (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t)); 3723f9d6ad7SLin Ling 37386714001SSerapheim Dimitropoulos root_vdev_actions_getprogress(vd, nv); 3745cabbc6bSPrashanth Sreenivasa 3755cabbc6bSPrashanth Sreenivasa /* 3765cabbc6bSPrashanth Sreenivasa * Note: this can be called from open context 3775cabbc6bSPrashanth Sreenivasa * (spa_get_stats()), so we need the rwlock to prevent 3785cabbc6bSPrashanth Sreenivasa * the mapping from being changed by condensing. 3795cabbc6bSPrashanth Sreenivasa */ 3805cabbc6bSPrashanth Sreenivasa rw_enter(&vd->vdev_indirect_rwlock, RW_READER); 3815cabbc6bSPrashanth Sreenivasa if (vd->vdev_indirect_mapping != NULL) { 3825cabbc6bSPrashanth Sreenivasa ASSERT(vd->vdev_indirect_births != NULL); 3835cabbc6bSPrashanth Sreenivasa vdev_indirect_mapping_t *vim = 3845cabbc6bSPrashanth Sreenivasa vd->vdev_indirect_mapping; 3855cabbc6bSPrashanth Sreenivasa fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE, 3865cabbc6bSPrashanth Sreenivasa vdev_indirect_mapping_size(vim)); 3875cabbc6bSPrashanth Sreenivasa } 3885cabbc6bSPrashanth Sreenivasa rw_exit(&vd->vdev_indirect_rwlock); 3895cabbc6bSPrashanth Sreenivasa if (vd->vdev_mg != NULL && 3905cabbc6bSPrashanth Sreenivasa vd->vdev_mg->mg_fragmentation != ZFS_FRAG_INVALID) { 3915cabbc6bSPrashanth Sreenivasa /* 3925cabbc6bSPrashanth Sreenivasa * Compute approximately how much memory would be used 3935cabbc6bSPrashanth Sreenivasa * for the indirect mapping if this device were to 3945cabbc6bSPrashanth Sreenivasa * be removed. 3955cabbc6bSPrashanth Sreenivasa * 3965cabbc6bSPrashanth Sreenivasa * Note: If the frag metric is invalid, then not 3975cabbc6bSPrashanth Sreenivasa * enough metaslabs have been converted to have 3985cabbc6bSPrashanth Sreenivasa * histograms. 3995cabbc6bSPrashanth Sreenivasa */ 4005cabbc6bSPrashanth Sreenivasa uint64_t seg_count = 0; 401cfd63e1bSMatthew Ahrens uint64_t to_alloc = vd->vdev_stat.vs_alloc; 4025cabbc6bSPrashanth Sreenivasa 4035cabbc6bSPrashanth Sreenivasa /* 4045cabbc6bSPrashanth Sreenivasa * There are the same number of allocated segments 4055cabbc6bSPrashanth Sreenivasa * as free segments, so we will have at least one 406cfd63e1bSMatthew Ahrens * entry per free segment. However, small free 407cfd63e1bSMatthew Ahrens * segments (smaller than vdev_removal_max_span) 408cfd63e1bSMatthew Ahrens * will be combined with adjacent allocated segments 409cfd63e1bSMatthew Ahrens * as a single mapping. 4105cabbc6bSPrashanth Sreenivasa */ 4115cabbc6bSPrashanth Sreenivasa for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) { 412cfd63e1bSMatthew Ahrens if (1ULL << (i + 1) < vdev_removal_max_span) { 413cfd63e1bSMatthew Ahrens to_alloc += 414cfd63e1bSMatthew Ahrens vd->vdev_mg->mg_histogram[i] << 415cfd63e1bSMatthew Ahrens i + 1; 416cfd63e1bSMatthew Ahrens } else { 417cfd63e1bSMatthew Ahrens seg_count += 418cfd63e1bSMatthew Ahrens vd->vdev_mg->mg_histogram[i]; 419cfd63e1bSMatthew Ahrens } 4205cabbc6bSPrashanth Sreenivasa } 4215cabbc6bSPrashanth Sreenivasa 4225cabbc6bSPrashanth Sreenivasa /* 423cfd63e1bSMatthew Ahrens * The maximum length of a mapping is 424cfd63e1bSMatthew Ahrens * zfs_remove_max_segment, so we need at least one entry 425cfd63e1bSMatthew Ahrens * per zfs_remove_max_segment of allocated data. 4265cabbc6bSPrashanth Sreenivasa */ 427cfd63e1bSMatthew Ahrens seg_count += to_alloc / zfs_remove_max_segment; 4285cabbc6bSPrashanth Sreenivasa 4295cabbc6bSPrashanth Sreenivasa fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE, 4305cabbc6bSPrashanth Sreenivasa seg_count * 4315cabbc6bSPrashanth Sreenivasa sizeof (vdev_indirect_mapping_entry_phys_t)); 4325cabbc6bSPrashanth Sreenivasa } 433fa9e4066Sahrens } 434fa9e4066Sahrens 435fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) { 436fa9e4066Sahrens nvlist_t **child; 4373f9d6ad7SLin Ling int c, idx; 438fa9e4066Sahrens 43988ecc943SGeorge Wilson ASSERT(!vd->vdev_ishole); 44088ecc943SGeorge Wilson 441fa9e4066Sahrens child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *), 442fa9e4066Sahrens KM_SLEEP); 443fa9e4066Sahrens 4443f9d6ad7SLin Ling for (c = 0, idx = 0; c < vd->vdev_children; c++) { 4453f9d6ad7SLin Ling vdev_t *cvd = vd->vdev_child[c]; 4463f9d6ad7SLin Ling 4473f9d6ad7SLin Ling /* 4483f9d6ad7SLin Ling * If we're generating an nvlist of removing 4493f9d6ad7SLin Ling * vdevs then skip over any device which is 4503f9d6ad7SLin Ling * not being removed. 4513f9d6ad7SLin Ling */ 4523f9d6ad7SLin Ling if ((flags & VDEV_CONFIG_REMOVING) && 4533f9d6ad7SLin Ling !cvd->vdev_removing) 4543f9d6ad7SLin Ling continue; 455fa9e4066Sahrens 4563f9d6ad7SLin Ling child[idx++] = vdev_config_generate(spa, cvd, 4573f9d6ad7SLin Ling getstats, flags); 4583f9d6ad7SLin Ling } 4593f9d6ad7SLin Ling 4603f9d6ad7SLin Ling if (idx) { 461b4952e17SGeorge Wilson fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 462b4952e17SGeorge Wilson child, idx); 4633f9d6ad7SLin Ling } 464fa9e4066Sahrens 4653f9d6ad7SLin Ling for (c = 0; c < idx; c++) 466fa9e4066Sahrens nvlist_free(child[c]); 467fa9e4066Sahrens 468fa9e4066Sahrens kmem_free(child, vd->vdev_children * sizeof (nvlist_t *)); 469441d80aaSlling 470441d80aaSlling } else { 471069f55e2SEric Schrock const char *aux = NULL; 472069f55e2SEric Schrock 473ecc2d604Sbonwick if (vd->vdev_offline && !vd->vdev_tmpoffline) 474b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, B_TRUE); 475b4952e17SGeorge Wilson if (vd->vdev_resilver_txg != 0) 476b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG, 477b4952e17SGeorge Wilson vd->vdev_resilver_txg); 4783d7072f8Seschrock if (vd->vdev_faulted) 479b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE); 4803d7072f8Seschrock if (vd->vdev_degraded) 481b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE); 4823d7072f8Seschrock if (vd->vdev_removed) 483b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE); 4843d7072f8Seschrock if (vd->vdev_unspare) 485b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE); 48688ecc943SGeorge Wilson if (vd->vdev_ishole) 487b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE); 488069f55e2SEric Schrock 489069f55e2SEric Schrock switch (vd->vdev_stat.vs_aux) { 490069f55e2SEric Schrock case VDEV_AUX_ERR_EXCEEDED: 491069f55e2SEric Schrock aux = "err_exceeded"; 492069f55e2SEric Schrock break; 493069f55e2SEric Schrock 494069f55e2SEric Schrock case VDEV_AUX_EXTERNAL: 495069f55e2SEric Schrock aux = "external"; 496069f55e2SEric Schrock break; 497069f55e2SEric Schrock } 498069f55e2SEric Schrock 499069f55e2SEric Schrock if (aux != NULL) 500b4952e17SGeorge Wilson fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux); 5011195e687SMark J Musante 5021195e687SMark J Musante if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) { 503b4952e17SGeorge Wilson fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID, 504b4952e17SGeorge Wilson vd->vdev_orig_guid); 5051195e687SMark J Musante } 506fa9e4066Sahrens } 507fa9e4066Sahrens 508fa9e4066Sahrens return (nv); 509fa9e4066Sahrens } 510fa9e4066Sahrens 51188ecc943SGeorge Wilson /* 51288ecc943SGeorge Wilson * Generate a view of the top-level vdevs. If we currently have holes 51388ecc943SGeorge Wilson * in the namespace, then generate an array which contains a list of holey 51488ecc943SGeorge Wilson * vdevs. Additionally, add the number of top-level children that currently 51588ecc943SGeorge Wilson * exist. 51688ecc943SGeorge Wilson */ 51788ecc943SGeorge Wilson void 51888ecc943SGeorge Wilson vdev_top_config_generate(spa_t *spa, nvlist_t *config) 51988ecc943SGeorge Wilson { 52088ecc943SGeorge Wilson vdev_t *rvd = spa->spa_root_vdev; 52188ecc943SGeorge Wilson uint64_t *array; 5223f9d6ad7SLin Ling uint_t c, idx; 52388ecc943SGeorge Wilson 52488ecc943SGeorge Wilson array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP); 52588ecc943SGeorge Wilson 5263f9d6ad7SLin Ling for (c = 0, idx = 0; c < rvd->vdev_children; c++) { 52788ecc943SGeorge Wilson vdev_t *tvd = rvd->vdev_child[c]; 52888ecc943SGeorge Wilson 5295cabbc6bSPrashanth Sreenivasa if (tvd->vdev_ishole) { 53088ecc943SGeorge Wilson array[idx++] = c; 5315cabbc6bSPrashanth Sreenivasa } 53288ecc943SGeorge Wilson } 53388ecc943SGeorge Wilson 534312c6e15SGeorge Wilson if (idx) { 535312c6e15SGeorge Wilson VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, 536312c6e15SGeorge Wilson array, idx) == 0); 537312c6e15SGeorge Wilson } 538312c6e15SGeorge Wilson 53988ecc943SGeorge Wilson VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, 54088ecc943SGeorge Wilson rvd->vdev_children) == 0); 54188ecc943SGeorge Wilson 54288ecc943SGeorge Wilson kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); 54388ecc943SGeorge Wilson } 54488ecc943SGeorge Wilson 545ad135b5dSChristopher Siden /* 546dfbb9432SGeorge Wilson * Returns the configuration from the label of the given vdev. For vdevs 547dfbb9432SGeorge Wilson * which don't have a txg value stored on their label (i.e. spares/cache) 548dfbb9432SGeorge Wilson * or have not been completely initialized (txg = 0) just return 549dfbb9432SGeorge Wilson * the configuration from the first valid label we find. Otherwise, 550dfbb9432SGeorge Wilson * find the most up-to-date label that does not exceed the specified 551dfbb9432SGeorge Wilson * 'txg' value. 552ad135b5dSChristopher Siden */ 553fa9e4066Sahrens nvlist_t * 554dfbb9432SGeorge Wilson vdev_label_read_config(vdev_t *vd, uint64_t txg) 555fa9e4066Sahrens { 5560373e76bSbonwick spa_t *spa = vd->vdev_spa; 557fa9e4066Sahrens nvlist_t *config = NULL; 558fa9e4066Sahrens vdev_phys_t *vp; 559770499e1SDan Kimmel abd_t *vp_abd; 560fa9e4066Sahrens zio_t *zio; 561dfbb9432SGeorge Wilson uint64_t best_txg = 0; 562b6bf6e15SPavel Zakharov uint64_t label_txg = 0; 563dfbb9432SGeorge Wilson int error = 0; 5648956713aSEric Schrock int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | 5658956713aSEric Schrock ZIO_FLAG_SPECULATIVE; 566fa9e4066Sahrens 567e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL); 5680373e76bSbonwick 5690a4e9518Sgw if (!vdev_readable(vd)) 570fa9e4066Sahrens return (NULL); 571fa9e4066Sahrens 572770499e1SDan Kimmel vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); 573770499e1SDan Kimmel vp = abd_to_buf(vp_abd); 574fa9e4066Sahrens 5758956713aSEric Schrock retry: 576e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 577dfbb9432SGeorge Wilson nvlist_t *label = NULL; 578fa9e4066Sahrens 579e14bb325SJeff Bonwick zio = zio_root(spa, NULL, NULL, flags); 580fa9e4066Sahrens 581770499e1SDan Kimmel vdev_label_read(zio, vd, l, vp_abd, 582fa9e4066Sahrens offsetof(vdev_label_t, vl_vdev_phys), 583e14bb325SJeff Bonwick sizeof (vdev_phys_t), NULL, NULL, flags); 584fa9e4066Sahrens 585fa9e4066Sahrens if (zio_wait(zio) == 0 && 586fa9e4066Sahrens nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist), 587dfbb9432SGeorge Wilson &label, 0) == 0) { 588dfbb9432SGeorge Wilson /* 589dfbb9432SGeorge Wilson * Auxiliary vdevs won't have txg values in their 590dfbb9432SGeorge Wilson * labels and newly added vdevs may not have been 591dfbb9432SGeorge Wilson * completely initialized so just return the 592dfbb9432SGeorge Wilson * configuration from the first valid label we 593dfbb9432SGeorge Wilson * encounter. 594dfbb9432SGeorge Wilson */ 595dfbb9432SGeorge Wilson error = nvlist_lookup_uint64(label, 596dfbb9432SGeorge Wilson ZPOOL_CONFIG_POOL_TXG, &label_txg); 597dfbb9432SGeorge Wilson if ((error || label_txg == 0) && !config) { 598dfbb9432SGeorge Wilson config = label; 599dfbb9432SGeorge Wilson break; 600dfbb9432SGeorge Wilson } else if (label_txg <= txg && label_txg > best_txg) { 601dfbb9432SGeorge Wilson best_txg = label_txg; 602dfbb9432SGeorge Wilson nvlist_free(config); 603dfbb9432SGeorge Wilson config = fnvlist_dup(label); 604dfbb9432SGeorge Wilson } 605dfbb9432SGeorge Wilson } 606fa9e4066Sahrens 607dfbb9432SGeorge Wilson if (label != NULL) { 608dfbb9432SGeorge Wilson nvlist_free(label); 609dfbb9432SGeorge Wilson label = NULL; 610fa9e4066Sahrens } 611fa9e4066Sahrens } 612fa9e4066Sahrens 6138956713aSEric Schrock if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) { 6148956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 6158956713aSEric Schrock goto retry; 6168956713aSEric Schrock } 6178956713aSEric Schrock 618b6bf6e15SPavel Zakharov /* 619b6bf6e15SPavel Zakharov * We found a valid label but it didn't pass txg restrictions. 620b6bf6e15SPavel Zakharov */ 621b6bf6e15SPavel Zakharov if (config == NULL && label_txg != 0) { 622b6bf6e15SPavel Zakharov vdev_dbgmsg(vd, "label discarded as txg is too large " 623b6bf6e15SPavel Zakharov "(%llu > %llu)", (u_longlong_t)label_txg, 624b6bf6e15SPavel Zakharov (u_longlong_t)txg); 625b6bf6e15SPavel Zakharov } 626b6bf6e15SPavel Zakharov 627770499e1SDan Kimmel abd_free(vp_abd); 628fa9e4066Sahrens 629fa9e4066Sahrens return (config); 630fa9e4066Sahrens } 631fa9e4066Sahrens 63239c23413Seschrock /* 63339c23413Seschrock * Determine if a device is in use. The 'spare_guid' parameter will be filled 63439c23413Seschrock * in with the device guid if this spare is active elsewhere on the system. 63539c23413Seschrock */ 63639c23413Seschrock static boolean_t 63739c23413Seschrock vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason, 638fa94a07fSbrendan uint64_t *spare_guid, uint64_t *l2cache_guid) 63939c23413Seschrock { 64039c23413Seschrock spa_t *spa = vd->vdev_spa; 64139c23413Seschrock uint64_t state, pool_guid, device_guid, txg, spare_pool; 64239c23413Seschrock uint64_t vdtxg = 0; 64339c23413Seschrock nvlist_t *label; 64439c23413Seschrock 64539c23413Seschrock if (spare_guid) 64639c23413Seschrock *spare_guid = 0ULL; 647fa94a07fSbrendan if (l2cache_guid) 648fa94a07fSbrendan *l2cache_guid = 0ULL; 64939c23413Seschrock 65039c23413Seschrock /* 65139c23413Seschrock * Read the label, if any, and perform some basic sanity checks. 65239c23413Seschrock */ 653dfbb9432SGeorge Wilson if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) 65439c23413Seschrock return (B_FALSE); 65539c23413Seschrock 65639c23413Seschrock (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG, 65739c23413Seschrock &vdtxg); 65839c23413Seschrock 65939c23413Seschrock if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, 66039c23413Seschrock &state) != 0 || 66139c23413Seschrock nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, 66239c23413Seschrock &device_guid) != 0) { 66339c23413Seschrock nvlist_free(label); 66439c23413Seschrock return (B_FALSE); 66539c23413Seschrock } 66639c23413Seschrock 667fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 66839c23413Seschrock (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, 66939c23413Seschrock &pool_guid) != 0 || 67039c23413Seschrock nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, 67139c23413Seschrock &txg) != 0)) { 67239c23413Seschrock nvlist_free(label); 67339c23413Seschrock return (B_FALSE); 67439c23413Seschrock } 67539c23413Seschrock 67639c23413Seschrock nvlist_free(label); 67739c23413Seschrock 67839c23413Seschrock /* 67939c23413Seschrock * Check to see if this device indeed belongs to the pool it claims to 68039c23413Seschrock * be a part of. The only way this is allowed is if the device is a hot 68139c23413Seschrock * spare (which we check for later on). 68239c23413Seschrock */ 683fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 68439c23413Seschrock !spa_guid_exists(pool_guid, device_guid) && 68589a89ebfSlling !spa_spare_exists(device_guid, NULL, NULL) && 686fa94a07fSbrendan !spa_l2cache_exists(device_guid, NULL)) 68739c23413Seschrock return (B_FALSE); 68839c23413Seschrock 68939c23413Seschrock /* 69039c23413Seschrock * If the transaction group is zero, then this an initialized (but 69139c23413Seschrock * unused) label. This is only an error if the create transaction 69239c23413Seschrock * on-disk is the same as the one we're using now, in which case the 69339c23413Seschrock * user has attempted to add the same vdev multiple times in the same 69439c23413Seschrock * transaction. 69539c23413Seschrock */ 696fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 697fa94a07fSbrendan txg == 0 && vdtxg == crtxg) 69839c23413Seschrock return (B_TRUE); 69939c23413Seschrock 70039c23413Seschrock /* 70139c23413Seschrock * Check to see if this is a spare device. We do an explicit check for 70239c23413Seschrock * spa_has_spare() here because it may be on our pending list of spares 703fa94a07fSbrendan * to add. We also check if it is an l2cache device. 70439c23413Seschrock */ 70589a89ebfSlling if (spa_spare_exists(device_guid, &spare_pool, NULL) || 70639c23413Seschrock spa_has_spare(spa, device_guid)) { 70739c23413Seschrock if (spare_guid) 70839c23413Seschrock *spare_guid = device_guid; 70939c23413Seschrock 71039c23413Seschrock switch (reason) { 71139c23413Seschrock case VDEV_LABEL_CREATE: 712fa94a07fSbrendan case VDEV_LABEL_L2CACHE: 71339c23413Seschrock return (B_TRUE); 71439c23413Seschrock 71539c23413Seschrock case VDEV_LABEL_REPLACE: 71639c23413Seschrock return (!spa_has_spare(spa, device_guid) || 71739c23413Seschrock spare_pool != 0ULL); 71839c23413Seschrock 71939c23413Seschrock case VDEV_LABEL_SPARE: 72039c23413Seschrock return (spa_has_spare(spa, device_guid)); 72139c23413Seschrock } 72239c23413Seschrock } 72339c23413Seschrock 724fa94a07fSbrendan /* 725fa94a07fSbrendan * Check to see if this is an l2cache device. 726fa94a07fSbrendan */ 727fa94a07fSbrendan if (spa_l2cache_exists(device_guid, NULL)) 728fa94a07fSbrendan return (B_TRUE); 729fa94a07fSbrendan 730f9af39baSGeorge Wilson /* 731f9af39baSGeorge Wilson * We can't rely on a pool's state if it's been imported 732f9af39baSGeorge Wilson * read-only. Instead we look to see if the pools is marked 733f9af39baSGeorge Wilson * read-only in the namespace and set the state to active. 734f9af39baSGeorge Wilson */ 7355bdd995dSRichard Yao if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 7365bdd995dSRichard Yao (spa = spa_by_guid(pool_guid, device_guid)) != NULL && 737f9af39baSGeorge Wilson spa_mode(spa) == FREAD) 738f9af39baSGeorge Wilson state = POOL_STATE_ACTIVE; 739f9af39baSGeorge Wilson 74039c23413Seschrock /* 74139c23413Seschrock * If the device is marked ACTIVE, then this device is in use by another 74239c23413Seschrock * pool on the system. 74339c23413Seschrock */ 74439c23413Seschrock return (state == POOL_STATE_ACTIVE); 74539c23413Seschrock } 74639c23413Seschrock 74739c23413Seschrock /* 74839c23413Seschrock * Initialize a vdev label. We check to make sure each leaf device is not in 74939c23413Seschrock * use, and writable. We put down an initial label which we will later 75039c23413Seschrock * overwrite with a complete label. Note that it's important to do this 75139c23413Seschrock * sequentially, not in parallel, so that we catch cases of multiple use of the 75239c23413Seschrock * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with 75339c23413Seschrock * itself. 75439c23413Seschrock */ 75539c23413Seschrock int 75639c23413Seschrock vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) 757fa9e4066Sahrens { 758fa9e4066Sahrens spa_t *spa = vd->vdev_spa; 759fa9e4066Sahrens nvlist_t *label; 760fa9e4066Sahrens vdev_phys_t *vp; 761770499e1SDan Kimmel abd_t *vp_abd; 762770499e1SDan Kimmel abd_t *pad2; 763ecc2d604Sbonwick uberblock_t *ub; 764770499e1SDan Kimmel abd_t *ub_abd; 765fa9e4066Sahrens zio_t *zio; 766fa9e4066Sahrens char *buf; 767fa9e4066Sahrens size_t buflen; 768fa9e4066Sahrens int error; 769fa94a07fSbrendan uint64_t spare_guid, l2cache_guid; 770e14bb325SJeff Bonwick int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; 771fa9e4066Sahrens 772e14bb325SJeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); 7730373e76bSbonwick 774e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 77539c23413Seschrock if ((error = vdev_label_init(vd->vdev_child[c], 77639c23413Seschrock crtxg, reason)) != 0) 777fa9e4066Sahrens return (error); 778fa9e4066Sahrens 77988ecc943SGeorge Wilson /* Track the creation time for this vdev */ 78088ecc943SGeorge Wilson vd->vdev_crtxg = crtxg; 78188ecc943SGeorge Wilson 782973c78e9SGeorge Wilson if (!vd->vdev_ops->vdev_op_leaf || !spa_writeable(spa)) 783fa9e4066Sahrens return (0); 784fa9e4066Sahrens 785fa9e4066Sahrens /* 78639c23413Seschrock * Dead vdevs cannot be initialized. 787fa9e4066Sahrens */ 788fa9e4066Sahrens if (vdev_is_dead(vd)) 789be6fd75aSMatthew Ahrens return (SET_ERROR(EIO)); 790fa9e4066Sahrens 791fa9e4066Sahrens /* 79239c23413Seschrock * Determine if the vdev is in use. 793fa9e4066Sahrens */ 7941195e687SMark J Musante if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT && 795fa94a07fSbrendan vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid)) 796be6fd75aSMatthew Ahrens return (SET_ERROR(EBUSY)); 79739c23413Seschrock 79839c23413Seschrock /* 799fa94a07fSbrendan * If this is a request to add or replace a spare or l2cache device 800fa94a07fSbrendan * that is in use elsewhere on the system, then we must update the 801fa94a07fSbrendan * guid (which was initialized to a random value) to reflect the 802fa94a07fSbrendan * actual GUID (which is shared between multiple pools). 80339c23413Seschrock */ 804fa94a07fSbrendan if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE && 805fa94a07fSbrendan spare_guid != 0ULL) { 80631157203SJeff Bonwick uint64_t guid_delta = spare_guid - vd->vdev_guid; 80799653d4eSeschrock 80831157203SJeff Bonwick vd->vdev_guid += guid_delta; 80931157203SJeff Bonwick 81031157203SJeff Bonwick for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) 81131157203SJeff Bonwick pvd->vdev_guid_sum += guid_delta; 81239c23413Seschrock 81399653d4eSeschrock /* 81439c23413Seschrock * If this is a replacement, then we want to fallthrough to the 81539c23413Seschrock * rest of the code. If we're adding a spare, then it's already 8163d7072f8Seschrock * labeled appropriately and we can just return. 81799653d4eSeschrock */ 81839c23413Seschrock if (reason == VDEV_LABEL_SPARE) 81939c23413Seschrock return (0); 8201195e687SMark J Musante ASSERT(reason == VDEV_LABEL_REPLACE || 8211195e687SMark J Musante reason == VDEV_LABEL_SPLIT); 822fa9e4066Sahrens } 823fa9e4066Sahrens 824fa94a07fSbrendan if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE && 825fa94a07fSbrendan l2cache_guid != 0ULL) { 82631157203SJeff Bonwick uint64_t guid_delta = l2cache_guid - vd->vdev_guid; 82731157203SJeff Bonwick 82831157203SJeff Bonwick vd->vdev_guid += guid_delta; 829fa94a07fSbrendan 83031157203SJeff Bonwick for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) 83131157203SJeff Bonwick pvd->vdev_guid_sum += guid_delta; 832fa94a07fSbrendan 833fa94a07fSbrendan /* 834fa94a07fSbrendan * If this is a replacement, then we want to fallthrough to the 835fa94a07fSbrendan * rest of the code. If we're adding an l2cache, then it's 836fa94a07fSbrendan * already labeled appropriately and we can just return. 837fa94a07fSbrendan */ 838fa94a07fSbrendan if (reason == VDEV_LABEL_L2CACHE) 839fa94a07fSbrendan return (0); 840fa94a07fSbrendan ASSERT(reason == VDEV_LABEL_REPLACE); 841fa94a07fSbrendan } 842fa94a07fSbrendan 843fa9e4066Sahrens /* 84439c23413Seschrock * Initialize its label. 845fa9e4066Sahrens */ 846770499e1SDan Kimmel vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); 847770499e1SDan Kimmel abd_zero(vp_abd, sizeof (vdev_phys_t)); 848770499e1SDan Kimmel vp = abd_to_buf(vp_abd); 849fa9e4066Sahrens 850fa9e4066Sahrens /* 851fa9e4066Sahrens * Generate a label describing the pool and our top-level vdev. 852fa9e4066Sahrens * We mark it as being from txg 0 to indicate that it's not 853fa9e4066Sahrens * really part of an active pool just yet. The labels will 854fa9e4066Sahrens * be written again with a meaningful txg by spa_sync(). 855fa9e4066Sahrens */ 85639c23413Seschrock if (reason == VDEV_LABEL_SPARE || 85739c23413Seschrock (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) { 85839c23413Seschrock /* 85939c23413Seschrock * For inactive hot spares, we generate a special label that 86039c23413Seschrock * identifies as a mutually shared hot spare. We write the 86139c23413Seschrock * label if we are adding a hot spare, or if we are removing an 86239c23413Seschrock * active hot spare (in which case we want to revert the 86339c23413Seschrock * labels). 86439c23413Seschrock */ 86599653d4eSeschrock VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); 86699653d4eSeschrock 86799653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, 86899653d4eSeschrock spa_version(spa)) == 0); 86999653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, 87099653d4eSeschrock POOL_STATE_SPARE) == 0); 87199653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, 87299653d4eSeschrock vd->vdev_guid) == 0); 873fa94a07fSbrendan } else if (reason == VDEV_LABEL_L2CACHE || 874fa94a07fSbrendan (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) { 875fa94a07fSbrendan /* 876fa94a07fSbrendan * For level 2 ARC devices, add a special label. 877fa94a07fSbrendan */ 878fa94a07fSbrendan VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); 879fa94a07fSbrendan 880fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, 881fa94a07fSbrendan spa_version(spa)) == 0); 882fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, 883fa94a07fSbrendan POOL_STATE_L2CACHE) == 0); 884fa94a07fSbrendan VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, 885fa94a07fSbrendan vd->vdev_guid) == 0); 88699653d4eSeschrock } else { 8871195e687SMark J Musante uint64_t txg = 0ULL; 8881195e687SMark J Musante 8891195e687SMark J Musante if (reason == VDEV_LABEL_SPLIT) 8901195e687SMark J Musante txg = spa->spa_uberblock.ub_txg; 8911195e687SMark J Musante label = spa_config_generate(spa, vd, txg, B_FALSE); 89299653d4eSeschrock 89399653d4eSeschrock /* 89499653d4eSeschrock * Add our creation time. This allows us to detect multiple 89599653d4eSeschrock * vdev uses as described above, and automatically expires if we 89699653d4eSeschrock * fail. 89799653d4eSeschrock */ 89899653d4eSeschrock VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, 89999653d4eSeschrock crtxg) == 0); 90099653d4eSeschrock } 901fa9e4066Sahrens 902fa9e4066Sahrens buf = vp->vp_nvlist; 903fa9e4066Sahrens buflen = sizeof (vp->vp_nvlist); 904fa9e4066Sahrens 905a75573b6Smmusante error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP); 906a75573b6Smmusante if (error != 0) { 907fa9e4066Sahrens nvlist_free(label); 908770499e1SDan Kimmel abd_free(vp_abd); 909a75573b6Smmusante /* EFAULT means nvlist_pack ran out of room */ 910a75573b6Smmusante return (error == EFAULT ? ENAMETOOLONG : EINVAL); 911fa9e4066Sahrens } 912fa9e4066Sahrens 913fa9e4066Sahrens /* 914fa9e4066Sahrens * Initialize uberblock template. 915fa9e4066Sahrens */ 916770499e1SDan Kimmel ub_abd = abd_alloc_linear(VDEV_UBERBLOCK_RING, B_TRUE); 917770499e1SDan Kimmel abd_zero(ub_abd, VDEV_UBERBLOCK_RING); 918770499e1SDan Kimmel abd_copy_from_buf(ub_abd, &spa->spa_uberblock, sizeof (uberblock_t)); 919770499e1SDan Kimmel ub = abd_to_buf(ub_abd); 920ecc2d604Sbonwick ub->ub_txg = 0; 921fa9e4066Sahrens 922f83ffe1aSLin Ling /* Initialize the 2nd padding area. */ 923770499e1SDan Kimmel pad2 = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE); 924770499e1SDan Kimmel abd_zero(pad2, VDEV_PAD_SIZE); 925f83ffe1aSLin Ling 926fa9e4066Sahrens /* 927fa9e4066Sahrens * Write everything in parallel. 928fa9e4066Sahrens */ 9298956713aSEric Schrock retry: 93017f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 931fa9e4066Sahrens 932e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 933fa9e4066Sahrens 934770499e1SDan Kimmel vdev_label_write(zio, vd, l, vp_abd, 935fa9e4066Sahrens offsetof(vdev_label_t, vl_vdev_phys), 93617f17c2dSbonwick sizeof (vdev_phys_t), NULL, NULL, flags); 937fa9e4066Sahrens 938f83ffe1aSLin Ling /* 939f83ffe1aSLin Ling * Skip the 1st padding area. 940f83ffe1aSLin Ling * Zero out the 2nd padding area where it might have 941f83ffe1aSLin Ling * left over data from previous filesystem format. 942f83ffe1aSLin Ling */ 943f83ffe1aSLin Ling vdev_label_write(zio, vd, l, pad2, 944f83ffe1aSLin Ling offsetof(vdev_label_t, vl_pad2), 945f83ffe1aSLin Ling VDEV_PAD_SIZE, NULL, NULL, flags); 946f83ffe1aSLin Ling 947770499e1SDan Kimmel vdev_label_write(zio, vd, l, ub_abd, 948f64c0e34SEric Taylor offsetof(vdev_label_t, vl_uberblock), 949f64c0e34SEric Taylor VDEV_UBERBLOCK_RING, NULL, NULL, flags); 950fa9e4066Sahrens } 951fa9e4066Sahrens 952fa9e4066Sahrens error = zio_wait(zio); 953fa9e4066Sahrens 9548956713aSEric Schrock if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { 9558956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 9568956713aSEric Schrock goto retry; 9578956713aSEric Schrock } 9588956713aSEric Schrock 959fa9e4066Sahrens nvlist_free(label); 960770499e1SDan Kimmel abd_free(pad2); 961770499e1SDan Kimmel abd_free(ub_abd); 962770499e1SDan Kimmel abd_free(vp_abd); 963fa9e4066Sahrens 96439c23413Seschrock /* 96539c23413Seschrock * If this vdev hasn't been previously identified as a spare, then we 9663d7072f8Seschrock * mark it as such only if a) we are labeling it as a spare, or b) it 967fa94a07fSbrendan * exists as a spare elsewhere in the system. Do the same for 968fa94a07fSbrendan * level 2 ARC devices. 96939c23413Seschrock */ 97039c23413Seschrock if (error == 0 && !vd->vdev_isspare && 97139c23413Seschrock (reason == VDEV_LABEL_SPARE || 97289a89ebfSlling spa_spare_exists(vd->vdev_guid, NULL, NULL))) 97339c23413Seschrock spa_spare_add(vd); 97499653d4eSeschrock 975fa94a07fSbrendan if (error == 0 && !vd->vdev_isl2cache && 976fa94a07fSbrendan (reason == VDEV_LABEL_L2CACHE || 977fa94a07fSbrendan spa_l2cache_exists(vd->vdev_guid, NULL))) 978fa94a07fSbrendan spa_l2cache_add(vd); 979fa94a07fSbrendan 98039c23413Seschrock return (error); 98199653d4eSeschrock } 98299653d4eSeschrock 983fa9e4066Sahrens /* 984fa9e4066Sahrens * ========================================================================== 985fa9e4066Sahrens * uberblock load/sync 986fa9e4066Sahrens * ========================================================================== 987fa9e4066Sahrens */ 988fa9e4066Sahrens 989fa9e4066Sahrens /* 990fa9e4066Sahrens * Consider the following situation: txg is safely synced to disk. We've 991fa9e4066Sahrens * written the first uberblock for txg + 1, and then we lose power. When we 992fa9e4066Sahrens * come back up, we fail to see the uberblock for txg + 1 because, say, 993fa9e4066Sahrens * it was on a mirrored device and the replica to which we wrote txg + 1 994fa9e4066Sahrens * is now offline. If we then make some changes and sync txg + 1, and then 995ad135b5dSChristopher Siden * the missing replica comes back, then for a few seconds we'll have two 996fa9e4066Sahrens * conflicting uberblocks on disk with the same txg. The solution is simple: 997fa9e4066Sahrens * among uberblocks with equal txg, choose the one with the latest timestamp. 998fa9e4066Sahrens */ 999fa9e4066Sahrens static int 1000fa9e4066Sahrens vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2) 1001fa9e4066Sahrens { 1002fa9e4066Sahrens if (ub1->ub_txg < ub2->ub_txg) 1003fa9e4066Sahrens return (-1); 1004fa9e4066Sahrens if (ub1->ub_txg > ub2->ub_txg) 1005fa9e4066Sahrens return (1); 1006fa9e4066Sahrens 1007fa9e4066Sahrens if (ub1->ub_timestamp < ub2->ub_timestamp) 1008fa9e4066Sahrens return (-1); 1009fa9e4066Sahrens if (ub1->ub_timestamp > ub2->ub_timestamp) 1010fa9e4066Sahrens return (1); 1011fa9e4066Sahrens 1012fa9e4066Sahrens return (0); 1013fa9e4066Sahrens } 1014fa9e4066Sahrens 1015ad135b5dSChristopher Siden struct ubl_cbdata { 1016ad135b5dSChristopher Siden uberblock_t *ubl_ubbest; /* Best uberblock */ 1017ad135b5dSChristopher Siden vdev_t *ubl_vd; /* vdev associated with the above */ 1018ad135b5dSChristopher Siden }; 1019ad135b5dSChristopher Siden 1020fa9e4066Sahrens static void 1021fa9e4066Sahrens vdev_uberblock_load_done(zio_t *zio) 1022fa9e4066Sahrens { 1023ad135b5dSChristopher Siden vdev_t *vd = zio->io_vd; 1024468c413aSTim Haley spa_t *spa = zio->io_spa; 1025e14bb325SJeff Bonwick zio_t *rio = zio->io_private; 1026770499e1SDan Kimmel uberblock_t *ub = abd_to_buf(zio->io_abd); 1027ad135b5dSChristopher Siden struct ubl_cbdata *cbp = rio->io_private; 1028fa9e4066Sahrens 1029ad135b5dSChristopher Siden ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd)); 1030fa9e4066Sahrens 1031ea8dc4b6Seschrock if (zio->io_error == 0 && uberblock_verify(ub) == 0) { 1032e14bb325SJeff Bonwick mutex_enter(&rio->io_lock); 1033468c413aSTim Haley if (ub->ub_txg <= spa->spa_load_max_txg && 1034ad135b5dSChristopher Siden vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) { 1035ad135b5dSChristopher Siden /* 1036dfbb9432SGeorge Wilson * Keep track of the vdev in which this uberblock 1037dfbb9432SGeorge Wilson * was found. We will use this information later 1038dfbb9432SGeorge Wilson * to obtain the config nvlist associated with 1039ad135b5dSChristopher Siden * this uberblock. 1040ad135b5dSChristopher Siden */ 1041ad135b5dSChristopher Siden *cbp->ubl_ubbest = *ub; 1042ad135b5dSChristopher Siden cbp->ubl_vd = vd; 1043ad135b5dSChristopher Siden } 1044e14bb325SJeff Bonwick mutex_exit(&rio->io_lock); 1045fa9e4066Sahrens } 1046fa9e4066Sahrens 1047770499e1SDan Kimmel abd_free(zio->io_abd); 1048fa9e4066Sahrens } 1049fa9e4066Sahrens 1050ad135b5dSChristopher Siden static void 1051ad135b5dSChristopher Siden vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags, 1052ad135b5dSChristopher Siden struct ubl_cbdata *cbp) 1053fa9e4066Sahrens { 1054e14bb325SJeff Bonwick for (int c = 0; c < vd->vdev_children; c++) 1055ad135b5dSChristopher Siden vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp); 1056fa9e4066Sahrens 1057e14bb325SJeff Bonwick if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) { 1058e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) { 1059e14bb325SJeff Bonwick for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { 1060e14bb325SJeff Bonwick vdev_label_read(zio, vd, l, 1061770499e1SDan Kimmel abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd), 1062770499e1SDan Kimmel B_TRUE), VDEV_UBERBLOCK_OFFSET(vd, n), 1063e14bb325SJeff Bonwick VDEV_UBERBLOCK_SIZE(vd), 1064e14bb325SJeff Bonwick vdev_uberblock_load_done, zio, flags); 1065e14bb325SJeff Bonwick } 1066fa9e4066Sahrens } 1067fa9e4066Sahrens } 1068ad135b5dSChristopher Siden } 1069e14bb325SJeff Bonwick 1070ad135b5dSChristopher Siden /* 1071ad135b5dSChristopher Siden * Reads the 'best' uberblock from disk along with its associated 1072ad135b5dSChristopher Siden * configuration. First, we read the uberblock array of each label of each 1073ad135b5dSChristopher Siden * vdev, keeping track of the uberblock with the highest txg in each array. 1074dfbb9432SGeorge Wilson * Then, we read the configuration from the same vdev as the best uberblock. 1075ad135b5dSChristopher Siden */ 1076ad135b5dSChristopher Siden void 1077ad135b5dSChristopher Siden vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config) 1078ad135b5dSChristopher Siden { 1079ad135b5dSChristopher Siden zio_t *zio; 1080ad135b5dSChristopher Siden spa_t *spa = rvd->vdev_spa; 1081ad135b5dSChristopher Siden struct ubl_cbdata cb; 1082ad135b5dSChristopher Siden int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | 1083ad135b5dSChristopher Siden ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; 1084ad135b5dSChristopher Siden 1085ad135b5dSChristopher Siden ASSERT(ub); 1086ad135b5dSChristopher Siden ASSERT(config); 1087ad135b5dSChristopher Siden 1088ad135b5dSChristopher Siden bzero(ub, sizeof (uberblock_t)); 1089ad135b5dSChristopher Siden *config = NULL; 1090ad135b5dSChristopher Siden 1091ad135b5dSChristopher Siden cb.ubl_ubbest = ub; 1092ad135b5dSChristopher Siden cb.ubl_vd = NULL; 1093ad135b5dSChristopher Siden 1094ad135b5dSChristopher Siden spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 1095ad135b5dSChristopher Siden zio = zio_root(spa, NULL, &cb, flags); 1096ad135b5dSChristopher Siden vdev_uberblock_load_impl(zio, rvd, flags, &cb); 1097ad135b5dSChristopher Siden (void) zio_wait(zio); 1098dfbb9432SGeorge Wilson 1099dfbb9432SGeorge Wilson /* 1100dfbb9432SGeorge Wilson * It's possible that the best uberblock was discovered on a label 1101dfbb9432SGeorge Wilson * that has a configuration which was written in a future txg. 1102dfbb9432SGeorge Wilson * Search all labels on this vdev to find the configuration that 1103dfbb9432SGeorge Wilson * matches the txg for our uberblock. 1104dfbb9432SGeorge Wilson */ 11053ee8c80cSPavel Zakharov if (cb.ubl_vd != NULL) { 11063ee8c80cSPavel Zakharov vdev_dbgmsg(cb.ubl_vd, "best uberblock found for spa %s. " 11073ee8c80cSPavel Zakharov "txg %llu", spa->spa_name, (u_longlong_t)ub->ub_txg); 11083ee8c80cSPavel Zakharov 1109dfbb9432SGeorge Wilson *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg); 11106f793812SPavel Zakharov if (*config == NULL && spa->spa_extreme_rewind) { 11116f793812SPavel Zakharov vdev_dbgmsg(cb.ubl_vd, "failed to read label config. " 11126f793812SPavel Zakharov "Trying again without txg restrictions."); 11136f793812SPavel Zakharov *config = vdev_label_read_config(cb.ubl_vd, UINT64_MAX); 11146f793812SPavel Zakharov } 11153ee8c80cSPavel Zakharov if (*config == NULL) { 11163ee8c80cSPavel Zakharov vdev_dbgmsg(cb.ubl_vd, "failed to read label config"); 11173ee8c80cSPavel Zakharov } 11183ee8c80cSPavel Zakharov } 1119ad135b5dSChristopher Siden spa_config_exit(spa, SCL_ALL, FTAG); 1120fa9e4066Sahrens } 1121fa9e4066Sahrens 1122fa9e4066Sahrens /* 112317f17c2dSbonwick * On success, increment root zio's count of good writes. 11240373e76bSbonwick * We only get credit for writes to known-visible vdevs; see spa_vdev_add(). 1125fa9e4066Sahrens */ 1126fa9e4066Sahrens static void 1127fa9e4066Sahrens vdev_uberblock_sync_done(zio_t *zio) 1128fa9e4066Sahrens { 112917f17c2dSbonwick uint64_t *good_writes = zio->io_private; 1130fa9e4066Sahrens 11310373e76bSbonwick if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0) 11321a5e258fSJosef 'Jeff' Sipek atomic_inc_64(good_writes); 1133fa9e4066Sahrens } 1134fa9e4066Sahrens 113517f17c2dSbonwick /* 113617f17c2dSbonwick * Write the uberblock to all labels of all leaves of the specified vdev. 113717f17c2dSbonwick */ 1138fa9e4066Sahrens static void 1139a3b55830SMatthew Ahrens vdev_uberblock_sync(zio_t *zio, uint64_t *good_writes, 1140a3b55830SMatthew Ahrens uberblock_t *ub, vdev_t *vd, int flags) 1141fa9e4066Sahrens { 1142a3b55830SMatthew Ahrens for (uint64_t c = 0; c < vd->vdev_children; c++) { 1143a3b55830SMatthew Ahrens vdev_uberblock_sync(zio, good_writes, 1144a3b55830SMatthew Ahrens ub, vd->vdev_child[c], flags); 1145a3b55830SMatthew Ahrens } 1146fa9e4066Sahrens 1147fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) 1148fa9e4066Sahrens return; 1149fa9e4066Sahrens 1150e14bb325SJeff Bonwick if (!vdev_writeable(vd)) 1151fa9e4066Sahrens return; 1152fa9e4066Sahrens 1153*e0f1c0afSOlaf Faaland int m = spa_multihost(vd->vdev_spa) ? MMP_BLOCKS_PER_LABEL : 0; 1154*e0f1c0afSOlaf Faaland int n = ub->ub_txg % (VDEV_UBERBLOCK_COUNT(vd) - m); 1155fa9e4066Sahrens 1156770499e1SDan Kimmel /* Copy the uberblock_t into the ABD */ 1157770499e1SDan Kimmel abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE); 1158770499e1SDan Kimmel abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd)); 1159770499e1SDan Kimmel abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t)); 1160fa9e4066Sahrens 1161e14bb325SJeff Bonwick for (int l = 0; l < VDEV_LABELS; l++) 1162770499e1SDan Kimmel vdev_label_write(zio, vd, l, ub_abd, 1163e14bb325SJeff Bonwick VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), 1164a3b55830SMatthew Ahrens vdev_uberblock_sync_done, good_writes, 1165e14bb325SJeff Bonwick flags | ZIO_FLAG_DONT_PROPAGATE); 1166fa9e4066Sahrens 1167770499e1SDan Kimmel abd_free(ub_abd); 1168fa9e4066Sahrens } 1169fa9e4066Sahrens 11703e30c24aSWill Andrews /* Sync the uberblocks to all vdevs in svd[] */ 117117f17c2dSbonwick int 117217f17c2dSbonwick vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags) 1173fa9e4066Sahrens { 117417f17c2dSbonwick spa_t *spa = svd[0]->vdev_spa; 1175e14bb325SJeff Bonwick zio_t *zio; 117617f17c2dSbonwick uint64_t good_writes = 0; 1177fa9e4066Sahrens 1178a3b55830SMatthew Ahrens zio = zio_root(spa, NULL, NULL, flags); 1179e14bb325SJeff Bonwick 1180e14bb325SJeff Bonwick for (int v = 0; v < svdcount; v++) 1181a3b55830SMatthew Ahrens vdev_uberblock_sync(zio, &good_writes, ub, svd[v], flags); 1182fa9e4066Sahrens 118317f17c2dSbonwick (void) zio_wait(zio); 1184fa9e4066Sahrens 1185fa9e4066Sahrens /* 118617f17c2dSbonwick * Flush the uberblocks to disk. This ensures that the odd labels 118717f17c2dSbonwick * are no longer needed (because the new uberblocks and the even 118817f17c2dSbonwick * labels are safely on disk), so it is safe to overwrite them. 1189fa9e4066Sahrens */ 119017f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 1191fa9e4066Sahrens 11925cabbc6bSPrashanth Sreenivasa for (int v = 0; v < svdcount; v++) { 11935cabbc6bSPrashanth Sreenivasa if (vdev_writeable(svd[v])) { 11945cabbc6bSPrashanth Sreenivasa zio_flush(zio, svd[v]); 11955cabbc6bSPrashanth Sreenivasa } 11965cabbc6bSPrashanth Sreenivasa } 1197fa9e4066Sahrens 119817f17c2dSbonwick (void) zio_wait(zio); 119917f17c2dSbonwick 120017f17c2dSbonwick return (good_writes >= 1 ? 0 : EIO); 1201fa9e4066Sahrens } 1202fa9e4066Sahrens 1203fa9e4066Sahrens /* 120417f17c2dSbonwick * On success, increment the count of good writes for our top-level vdev. 1205fa9e4066Sahrens */ 1206fa9e4066Sahrens static void 120717f17c2dSbonwick vdev_label_sync_done(zio_t *zio) 1208fa9e4066Sahrens { 120917f17c2dSbonwick uint64_t *good_writes = zio->io_private; 1210fa9e4066Sahrens 1211fa9e4066Sahrens if (zio->io_error == 0) 12121a5e258fSJosef 'Jeff' Sipek atomic_inc_64(good_writes); 1213fa9e4066Sahrens } 1214fa9e4066Sahrens 121517f17c2dSbonwick /* 121617f17c2dSbonwick * If there weren't enough good writes, indicate failure to the parent. 121717f17c2dSbonwick */ 1218fa9e4066Sahrens static void 121917f17c2dSbonwick vdev_label_sync_top_done(zio_t *zio) 122017f17c2dSbonwick { 122117f17c2dSbonwick uint64_t *good_writes = zio->io_private; 122217f17c2dSbonwick 122317f17c2dSbonwick if (*good_writes == 0) 1224be6fd75aSMatthew Ahrens zio->io_error = SET_ERROR(EIO); 122517f17c2dSbonwick 122617f17c2dSbonwick kmem_free(good_writes, sizeof (uint64_t)); 122717f17c2dSbonwick } 122817f17c2dSbonwick 122951ece835Seschrock /* 12300430f8daSeschrock * We ignore errors for log and cache devices, simply free the private data. 123151ece835Seschrock */ 123251ece835Seschrock static void 12330430f8daSeschrock vdev_label_sync_ignore_done(zio_t *zio) 123451ece835Seschrock { 123551ece835Seschrock kmem_free(zio->io_private, sizeof (uint64_t)); 123651ece835Seschrock } 123751ece835Seschrock 123817f17c2dSbonwick /* 123917f17c2dSbonwick * Write all even or odd labels to all leaves of the specified vdev. 124017f17c2dSbonwick */ 124117f17c2dSbonwick static void 1242a3b55830SMatthew Ahrens vdev_label_sync(zio_t *zio, uint64_t *good_writes, 1243a3b55830SMatthew Ahrens vdev_t *vd, int l, uint64_t txg, int flags) 1244fa9e4066Sahrens { 1245fa9e4066Sahrens nvlist_t *label; 1246fa9e4066Sahrens vdev_phys_t *vp; 1247770499e1SDan Kimmel abd_t *vp_abd; 1248fa9e4066Sahrens char *buf; 1249fa9e4066Sahrens size_t buflen; 1250fa9e4066Sahrens 1251a3b55830SMatthew Ahrens for (int c = 0; c < vd->vdev_children; c++) { 1252a3b55830SMatthew Ahrens vdev_label_sync(zio, good_writes, 1253a3b55830SMatthew Ahrens vd->vdev_child[c], l, txg, flags); 1254a3b55830SMatthew Ahrens } 1255fa9e4066Sahrens 1256fa9e4066Sahrens if (!vd->vdev_ops->vdev_op_leaf) 1257fa9e4066Sahrens return; 1258fa9e4066Sahrens 1259e14bb325SJeff Bonwick if (!vdev_writeable(vd)) 1260fa9e4066Sahrens return; 1261fa9e4066Sahrens 1262fa9e4066Sahrens /* 1263fa9e4066Sahrens * Generate a label describing the top-level config to which we belong. 1264fa9e4066Sahrens */ 12650373e76bSbonwick label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE); 1266fa9e4066Sahrens 1267770499e1SDan Kimmel vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); 1268770499e1SDan Kimmel abd_zero(vp_abd, sizeof (vdev_phys_t)); 1269770499e1SDan Kimmel vp = abd_to_buf(vp_abd); 1270fa9e4066Sahrens 1271fa9e4066Sahrens buf = vp->vp_nvlist; 1272fa9e4066Sahrens buflen = sizeof (vp->vp_nvlist); 1273fa9e4066Sahrens 127417f17c2dSbonwick if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) { 127517f17c2dSbonwick for (; l < VDEV_LABELS; l += 2) { 1276770499e1SDan Kimmel vdev_label_write(zio, vd, l, vp_abd, 127717f17c2dSbonwick offsetof(vdev_label_t, vl_vdev_phys), 127817f17c2dSbonwick sizeof (vdev_phys_t), 1279a3b55830SMatthew Ahrens vdev_label_sync_done, good_writes, 1280e14bb325SJeff Bonwick flags | ZIO_FLAG_DONT_PROPAGATE); 128117f17c2dSbonwick } 128217f17c2dSbonwick } 1283fa9e4066Sahrens 1284770499e1SDan Kimmel abd_free(vp_abd); 1285fa9e4066Sahrens nvlist_free(label); 1286fa9e4066Sahrens } 1287fa9e4066Sahrens 128817f17c2dSbonwick int 1289e14bb325SJeff Bonwick vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags) 1290fa9e4066Sahrens { 1291e14bb325SJeff Bonwick list_t *dl = &spa->spa_config_dirty_list; 129217f17c2dSbonwick vdev_t *vd; 1293e14bb325SJeff Bonwick zio_t *zio; 1294fa9e4066Sahrens int error; 1295fa9e4066Sahrens 1296fa9e4066Sahrens /* 1297e14bb325SJeff Bonwick * Write the new labels to disk. 1298fa9e4066Sahrens */ 1299e14bb325SJeff Bonwick zio = zio_root(spa, NULL, NULL, flags); 1300fa9e4066Sahrens 130117f17c2dSbonwick for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) { 130217f17c2dSbonwick uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t), 130317f17c2dSbonwick KM_SLEEP); 130488ecc943SGeorge Wilson 130588ecc943SGeorge Wilson ASSERT(!vd->vdev_ishole); 130688ecc943SGeorge Wilson 1307a3f829aeSBill Moore zio_t *vio = zio_null(zio, spa, NULL, 13080430f8daSeschrock (vd->vdev_islog || vd->vdev_aux != NULL) ? 13090430f8daSeschrock vdev_label_sync_ignore_done : vdev_label_sync_top_done, 131017f17c2dSbonwick good_writes, flags); 1311a3b55830SMatthew Ahrens vdev_label_sync(vio, good_writes, vd, l, txg, flags); 131217f17c2dSbonwick zio_nowait(vio); 1313fa9e4066Sahrens } 1314e14bb325SJeff Bonwick 1315e14bb325SJeff Bonwick error = zio_wait(zio); 1316fa9e4066Sahrens 13178654d025Sperrin /* 131817f17c2dSbonwick * Flush the new labels to disk. 13198654d025Sperrin */ 132017f17c2dSbonwick zio = zio_root(spa, NULL, NULL, flags); 13218654d025Sperrin 132217f17c2dSbonwick for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) 132317f17c2dSbonwick zio_flush(zio, vd); 132417f17c2dSbonwick 132517f17c2dSbonwick (void) zio_wait(zio); 1326fa9e4066Sahrens 1327fa9e4066Sahrens return (error); 1328fa9e4066Sahrens } 1329fa9e4066Sahrens 1330fa9e4066Sahrens /* 133117f17c2dSbonwick * Sync the uberblock and any changes to the vdev configuration. 1332fa9e4066Sahrens * 1333fa9e4066Sahrens * The order of operations is carefully crafted to ensure that 1334fa9e4066Sahrens * if the system panics or loses power at any time, the state on disk 1335fa9e4066Sahrens * is still transactionally consistent. The in-line comments below 1336fa9e4066Sahrens * describe the failure semantics at each stage. 1337fa9e4066Sahrens * 133817f17c2dSbonwick * Moreover, vdev_config_sync() is designed to be idempotent: if it fails 1339fa9e4066Sahrens * at any time, you can just call it again, and it will resume its work. 1340fa9e4066Sahrens */ 1341e14bb325SJeff Bonwick int 1342eb5bb584SWill Andrews vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg) 1343fa9e4066Sahrens { 134417f17c2dSbonwick spa_t *spa = svd[0]->vdev_spa; 1345fa9e4066Sahrens uberblock_t *ub = &spa->spa_uberblock; 1346eb5bb584SWill Andrews int error = 0; 1347e14bb325SJeff Bonwick int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; 1348fa9e4066Sahrens 134986714001SSerapheim Dimitropoulos ASSERT(svdcount != 0); 1350eb5bb584SWill Andrews retry: 13518956713aSEric Schrock /* 13528956713aSEric Schrock * Normally, we don't want to try too hard to write every label and 13538956713aSEric Schrock * uberblock. If there is a flaky disk, we don't want the rest of the 13548956713aSEric Schrock * sync process to block while we retry. But if we can't write a 13558956713aSEric Schrock * single label out, we should retry with ZIO_FLAG_TRYHARD before 13568956713aSEric Schrock * bailing out and declaring the pool faulted. 13578956713aSEric Schrock */ 1358eb5bb584SWill Andrews if (error != 0) { 1359eb5bb584SWill Andrews if ((flags & ZIO_FLAG_TRYHARD) != 0) 1360eb5bb584SWill Andrews return (error); 13618956713aSEric Schrock flags |= ZIO_FLAG_TRYHARD; 1362eb5bb584SWill Andrews } 13638956713aSEric Schrock 1364fa9e4066Sahrens ASSERT(ub->ub_txg <= txg); 1365fa9e4066Sahrens 1366fa9e4066Sahrens /* 136717f17c2dSbonwick * If this isn't a resync due to I/O errors, 136817f17c2dSbonwick * and nothing changed in this transaction group, 136917f17c2dSbonwick * and the vdev configuration hasn't changed, 13700373e76bSbonwick * then there's nothing to do. 1371fa9e4066Sahrens */ 1372*e0f1c0afSOlaf Faaland if (ub->ub_txg < txg) { 1373*e0f1c0afSOlaf Faaland boolean_t changed = uberblock_update(ub, spa->spa_root_vdev, 1374*e0f1c0afSOlaf Faaland txg, spa->spa_mmp.mmp_delay); 1375*e0f1c0afSOlaf Faaland 1376*e0f1c0afSOlaf Faaland if (!changed && list_is_empty(&spa->spa_config_dirty_list)) 1377*e0f1c0afSOlaf Faaland return (0); 1378*e0f1c0afSOlaf Faaland } 1379fa9e4066Sahrens 1380fa9e4066Sahrens if (txg > spa_freeze_txg(spa)) 1381e14bb325SJeff Bonwick return (0); 1382fa9e4066Sahrens 13830373e76bSbonwick ASSERT(txg <= spa->spa_final_txg); 13840373e76bSbonwick 1385fa9e4066Sahrens /* 1386fa9e4066Sahrens * Flush the write cache of every disk that's been written to 1387fa9e4066Sahrens * in this transaction group. This ensures that all blocks 1388fa9e4066Sahrens * written in this txg will be committed to stable storage 1389fa9e4066Sahrens * before any uberblock that references them. 1390fa9e4066Sahrens */ 139186714001SSerapheim Dimitropoulos zio_t *zio = zio_root(spa, NULL, NULL, flags); 139217f17c2dSbonwick 139386714001SSerapheim Dimitropoulos for (vdev_t *vd = 139486714001SSerapheim Dimitropoulos txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd != NULL; 139517f17c2dSbonwick vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg))) 139617f17c2dSbonwick zio_flush(zio, vd); 139717f17c2dSbonwick 1398fa9e4066Sahrens (void) zio_wait(zio); 1399fa9e4066Sahrens 1400fa9e4066Sahrens /* 1401fa9e4066Sahrens * Sync out the even labels (L0, L2) for every dirty vdev. If the 1402fa9e4066Sahrens * system dies in the middle of this process, that's OK: all of the 1403fa9e4066Sahrens * even labels that made it to disk will be newer than any uberblock, 1404fa9e4066Sahrens * and will therefore be considered invalid. The odd labels (L1, L3), 140517f17c2dSbonwick * which have not yet been touched, will still be valid. We flush 140617f17c2dSbonwick * the new labels to disk to ensure that all even-label updates 140717f17c2dSbonwick * are committed to stable storage before the uberblock update. 1408fa9e4066Sahrens */ 140986714001SSerapheim Dimitropoulos if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) { 141086714001SSerapheim Dimitropoulos if ((flags & ZIO_FLAG_TRYHARD) != 0) { 141186714001SSerapheim Dimitropoulos zfs_dbgmsg("vdev_label_sync_list() returned error %d " 141286714001SSerapheim Dimitropoulos "for pool '%s' when syncing out the even labels " 141386714001SSerapheim Dimitropoulos "of dirty vdevs", error, spa_name(spa)); 141486714001SSerapheim Dimitropoulos } 1415eb5bb584SWill Andrews goto retry; 141686714001SSerapheim Dimitropoulos } 1417fa9e4066Sahrens 1418fa9e4066Sahrens /* 1419e14bb325SJeff Bonwick * Sync the uberblocks to all vdevs in svd[]. 14200373e76bSbonwick * If the system dies in the middle of this step, there are two cases 14210373e76bSbonwick * to consider, and the on-disk state is consistent either way: 1422fa9e4066Sahrens * 1423fa9e4066Sahrens * (1) If none of the new uberblocks made it to disk, then the 1424fa9e4066Sahrens * previous uberblock will be the newest, and the odd labels 1425fa9e4066Sahrens * (which had not yet been touched) will be valid with respect 1426fa9e4066Sahrens * to that uberblock. 1427fa9e4066Sahrens * 1428fa9e4066Sahrens * (2) If one or more new uberblocks made it to disk, then they 1429fa9e4066Sahrens * will be the newest, and the even labels (which had all 1430fa9e4066Sahrens * been successfully committed) will be valid with respect 1431fa9e4066Sahrens * to the new uberblocks. 1432fa9e4066Sahrens */ 143386714001SSerapheim Dimitropoulos if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) { 143486714001SSerapheim Dimitropoulos if ((flags & ZIO_FLAG_TRYHARD) != 0) { 143586714001SSerapheim Dimitropoulos zfs_dbgmsg("vdev_uberblock_sync_list() returned error " 143686714001SSerapheim Dimitropoulos "%d for pool '%s'", error, spa_name(spa)); 143786714001SSerapheim Dimitropoulos } 1438eb5bb584SWill Andrews goto retry; 143986714001SSerapheim Dimitropoulos } 1440fa9e4066Sahrens 1441*e0f1c0afSOlaf Faaland if (spa_multihost(spa)) 1442*e0f1c0afSOlaf Faaland mmp_update_uberblock(spa, ub); 1443*e0f1c0afSOlaf Faaland 1444fa9e4066Sahrens /* 1445fa9e4066Sahrens * Sync out odd labels for every dirty vdev. If the system dies 1446fa9e4066Sahrens * in the middle of this process, the even labels and the new 1447fa9e4066Sahrens * uberblocks will suffice to open the pool. The next time 1448fa9e4066Sahrens * the pool is opened, the first thing we'll do -- before any 1449fa9e4066Sahrens * user data is modified -- is mark every vdev dirty so that 145017f17c2dSbonwick * all labels will be brought up to date. We flush the new labels 145117f17c2dSbonwick * to disk to ensure that all odd-label updates are committed to 145217f17c2dSbonwick * stable storage before the next transaction group begins. 1453fa9e4066Sahrens */ 145486714001SSerapheim Dimitropoulos if ((error = vdev_label_sync_list(spa, 1, txg, flags)) != 0) { 145586714001SSerapheim Dimitropoulos if ((flags & ZIO_FLAG_TRYHARD) != 0) { 145686714001SSerapheim Dimitropoulos zfs_dbgmsg("vdev_label_sync_list() returned error %d " 145786714001SSerapheim Dimitropoulos "for pool '%s' when syncing out the odd labels of " 145886714001SSerapheim Dimitropoulos "dirty vdevs", error, spa_name(spa)); 145986714001SSerapheim Dimitropoulos } 1460eb5bb584SWill Andrews goto retry; 146186714001SSerapheim Dimitropoulos } 1462eb5bb584SWill Andrews 1463eb5bb584SWill Andrews return (0); 1464fa9e4066Sahrens } 1465