4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
28 * Virtual Device Labels
29 * ---------------------
31 * The vdev label serves several distinct purposes:
33 * 1. Uniquely identify this device as part of a ZFS pool and confirm its
34 * identity within the pool.
36 * 2. Verify that all the devices given in a configuration are present
39 * 3. Determine the uberblock for the pool.
41 * 4. In case of an import operation, determine the configuration of the
42 * toplevel vdev of which it is a part.
44 * 5. If an import operation cannot find all the devices in the pool,
45 * provide enough information to the administrator to determine which
46 * devices are missing.
48 * It is important to note that while the kernel is responsible for writing the
49 * label, it only consumes the information in the first three cases. The
50 * latter information is only consumed in userland when determining the
51 * configuration to import a pool.
57 * Before describing the contents of the label, it's important to understand how
58 * the labels are written and updated with respect to the uberblock.
60 * When the pool configuration is altered, either because it was newly created
61 * or a device was added, we want to update all the labels such that we can deal
62 * with fatal failure at any point. To this end, each disk has two labels which
63 * are updated before and after the uberblock is synced. Assuming we have
64 * labels and an uberblock with the following transaction groups:
67 * +------+ +------+ +------+
69 * | t10 | | t10 | | t10 |
71 * +------+ +------+ +------+
73 * In this stable state, the labels and the uberblock were all updated within
74 * the same transaction group (10). Each label is mirrored and checksummed, so
75 * that we can detect when we fail partway through writing the label.
77 * In order to identify which labels are valid, the labels are written in the
80 * 1. For each vdev, update 'L1' to the new label
81 * 2. Update the uberblock
82 * 3. For each vdev, update 'L2' to the new label
84 * Given arbitrary failure, we can determine the correct label to use based on
85 * the transaction group. If we fail after updating L1 but before updating the
86 * UB, we will notice that L1's transaction group is greater than the uberblock,
87 * so L2 must be valid. If we fail after writing the uberblock but before
88 * writing L2, we will notice that L2's transaction group is less than L1, and
89 * therefore L1 is valid.
91 * Another added complexity is that not every label is updated when the config
92 * is synced. If we add a single device, we do not want to have to re-write
93 * every label for every device in the pool. This means that both L1 and L2 may
94 * be older than the pool uberblock, because the necessary information is stored
101 * The vdev label consists of two distinct parts, and is wrapped within the
102 * vdev_label_t structure. The label includes 8k of padding to permit legacy
103 * VTOC disk labels, but is otherwise ignored.
105 * The first half of the label is a packed nvlist which contains pool wide
106 * properties, per-vdev properties, and configuration information. It is
107 * described in more detail below.
109 * The latter half of the label consists of a redundant array of uberblocks.
110 * These uberblocks are updated whenever a transaction group is committed,
111 * or when the configuration is updated. When a pool is loaded, we scan each
112 * vdev for the 'best' uberblock.
115 * Configuration Information
116 * -------------------------
118 * The nvlist describing the pool and vdev contains the following elements:
120 * version ZFS on-disk version
123 * txg Transaction group in which this label was written
124 * pool_guid Unique identifier for this pool
125 * vdev_tree An nvlist describing vdev tree.
127 * An nvlist of the features necessary for reading the MOS.
129 * Each leaf device label also contains the following:
131 * top_guid Unique ID for top-level vdev in which this is contained
132 * guid Unique ID for the leaf vdev
134 * The 'vs' configuration follows the format described in 'spa_config.c'.
137 #include <sys/zfs_context.h>
139 #include <sys/spa_impl.h>
142 #include <sys/vdev.h>
143 #include <sys/vdev_impl.h>
144 #include <sys/uberblock_impl.h>
145 #include <sys/metaslab.h>
146 #include <sys/metaslab_impl.h>
148 #include <sys/dsl_scan.h>
150 #include <sys/fs/zfs.h>
151 #include <sys/trim_map.h>
153 static boolean_t vdev_trim_on_init = B_TRUE;
154 SYSCTL_DECL(_vfs_zfs_vdev);
155 SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, trim_on_init, CTLFLAG_RW,
156 &vdev_trim_on_init, 0, "Enable/disable full vdev trim on initialisation");
159 * Basic routines to read and write from a vdev label.
160 * Used throughout the rest of this file.
163 vdev_label_offset(uint64_t psize, int l, uint64_t offset)
165 ASSERT(offset < sizeof (vdev_label_t));
166 ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0);
168 return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
169 0 : psize - VDEV_LABELS * sizeof (vdev_label_t)));
173 * Returns back the vdev label associated with the passed in offset.
176 vdev_label_number(uint64_t psize, uint64_t offset)
180 if (offset >= psize - VDEV_LABEL_END_SIZE) {
181 offset -= psize - VDEV_LABEL_END_SIZE;
182 offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t);
184 l = offset / sizeof (vdev_label_t);
185 return (l < VDEV_LABELS ? l : -1);
189 vdev_label_read(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset,
190 uint64_t size, zio_done_func_t *done, void *private, int flags)
192 ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) ==
194 ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
196 zio_nowait(zio_read_phys(zio, vd,
197 vdev_label_offset(vd->vdev_psize, l, offset),
198 size, buf, ZIO_CHECKSUM_LABEL, done, private,
199 ZIO_PRIORITY_SYNC_READ, flags, B_TRUE));
203 vdev_label_write(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset,
204 uint64_t size, zio_done_func_t *done, void *private, int flags)
206 ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL ||
207 (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
208 (SCL_CONFIG | SCL_STATE) &&
209 dsl_pool_sync_context(spa_get_dsl(zio->io_spa))));
210 ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
212 zio_nowait(zio_write_phys(zio, vd,
213 vdev_label_offset(vd->vdev_psize, l, offset),
214 size, buf, ZIO_CHECKSUM_LABEL, done, private,
215 ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE));
219 root_vdev_actions_getprogress(vdev_t *vd, nvlist_t *nvl)
221 spa_t *spa = vd->vdev_spa;
223 if (vd != spa->spa_root_vdev)
226 /* provide either current or previous scan information */
228 if (spa_scan_get_stats(spa, &ps) == 0) {
229 fnvlist_add_uint64_array(nvl,
230 ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps,
231 sizeof (pool_scan_stat_t) / sizeof (uint64_t));
234 pool_removal_stat_t prs;
235 if (spa_removal_get_stats(spa, &prs) == 0) {
236 fnvlist_add_uint64_array(nvl,
237 ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t *)&prs,
238 sizeof (prs) / sizeof (uint64_t));
241 pool_checkpoint_stat_t pcs;
242 if (spa_checkpoint_get_stats(spa, &pcs) == 0) {
243 fnvlist_add_uint64_array(nvl,
244 ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t *)&pcs,
245 sizeof (pcs) / sizeof (uint64_t));
250 * Generate the nvlist representing this vdev's config.
253 vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
254 vdev_config_flag_t flags)
257 vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
259 nv = fnvlist_alloc();
261 fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type);
262 if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)))
263 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id);
264 fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid);
266 if (vd->vdev_path != NULL)
267 fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path);
269 if (vd->vdev_devid != NULL)
270 fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid);
272 if (vd->vdev_physpath != NULL)
273 fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH,
276 if (vd->vdev_fru != NULL)
277 fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru);
279 if (vd->vdev_nparity != 0) {
280 ASSERT(strcmp(vd->vdev_ops->vdev_op_type,
281 VDEV_TYPE_RAIDZ) == 0);
284 * Make sure someone hasn't managed to sneak a fancy new vdev
285 * into a crufty old storage pool.
287 ASSERT(vd->vdev_nparity == 1 ||
288 (vd->vdev_nparity <= 2 &&
289 spa_version(spa) >= SPA_VERSION_RAIDZ2) ||
290 (vd->vdev_nparity <= 3 &&
291 spa_version(spa) >= SPA_VERSION_RAIDZ3));
294 * Note that we'll add the nparity tag even on storage pools
295 * that only support a single parity device -- older software
296 * will just ignore it.
298 fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vd->vdev_nparity);
301 if (vd->vdev_wholedisk != -1ULL)
302 fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
305 if (vd->vdev_not_present && !(flags & VDEV_CONFIG_MISSING))
306 fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1);
308 if (vd->vdev_isspare)
309 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1);
311 if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) &&
312 vd == vd->vdev_top) {
313 fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
315 fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
317 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift);
318 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE,
320 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog);
321 if (vd->vdev_removing) {
322 fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING,
327 if (vd->vdev_dtl_sm != NULL) {
328 fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL,
329 space_map_object(vd->vdev_dtl_sm));
332 if (vic->vic_mapping_object != 0) {
333 fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT,
334 vic->vic_mapping_object);
337 if (vic->vic_births_object != 0) {
338 fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS,
339 vic->vic_births_object);
342 if (vic->vic_prev_indirect_vdev != UINT64_MAX) {
343 fnvlist_add_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV,
344 vic->vic_prev_indirect_vdev);
348 fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg);
350 if (flags & VDEV_CONFIG_MOS) {
351 if (vd->vdev_leaf_zap != 0) {
352 ASSERT(vd->vdev_ops->vdev_op_leaf);
353 fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_LEAF_ZAP,
357 if (vd->vdev_top_zap != 0) {
358 ASSERT(vd == vd->vdev_top);
359 fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
367 vdev_get_stats(vd, &vs);
368 fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
369 (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t));
371 root_vdev_actions_getprogress(vd, nv);
374 * Note: this can be called from open context
375 * (spa_get_stats()), so we need the rwlock to prevent
376 * the mapping from being changed by condensing.
378 rw_enter(&vd->vdev_indirect_rwlock, RW_READER);
379 if (vd->vdev_indirect_mapping != NULL) {
380 ASSERT(vd->vdev_indirect_births != NULL);
381 vdev_indirect_mapping_t *vim =
382 vd->vdev_indirect_mapping;
383 fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE,
384 vdev_indirect_mapping_size(vim));
386 rw_exit(&vd->vdev_indirect_rwlock);
387 if (vd->vdev_mg != NULL &&
388 vd->vdev_mg->mg_fragmentation != ZFS_FRAG_INVALID) {
390 * Compute approximately how much memory would be used
391 * for the indirect mapping if this device were to
394 * Note: If the frag metric is invalid, then not
395 * enough metaslabs have been converted to have
398 uint64_t seg_count = 0;
401 * There are the same number of allocated segments
402 * as free segments, so we will have at least one
403 * entry per free segment.
405 for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
406 seg_count += vd->vdev_mg->mg_histogram[i];
410 * The maximum length of a mapping is SPA_MAXBLOCKSIZE,
411 * so we need at least one entry per SPA_MAXBLOCKSIZE
414 seg_count += vd->vdev_stat.vs_alloc / SPA_MAXBLOCKSIZE;
416 fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE,
418 sizeof (vdev_indirect_mapping_entry_phys_t));
422 if (!vd->vdev_ops->vdev_op_leaf) {
426 ASSERT(!vd->vdev_ishole);
428 child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *),
431 for (c = 0, idx = 0; c < vd->vdev_children; c++) {
432 vdev_t *cvd = vd->vdev_child[c];
435 * If we're generating an nvlist of removing
436 * vdevs then skip over any device which is
439 if ((flags & VDEV_CONFIG_REMOVING) &&
443 child[idx++] = vdev_config_generate(spa, cvd,
448 fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
452 for (c = 0; c < idx; c++)
453 nvlist_free(child[c]);
455 kmem_free(child, vd->vdev_children * sizeof (nvlist_t *));
458 const char *aux = NULL;
460 if (vd->vdev_offline && !vd->vdev_tmpoffline)
461 fnvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, B_TRUE);
462 if (vd->vdev_resilver_txg != 0)
463 fnvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG,
464 vd->vdev_resilver_txg);
465 if (vd->vdev_faulted)
466 fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE);
467 if (vd->vdev_degraded)
468 fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE);
469 if (vd->vdev_removed)
470 fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE);
471 if (vd->vdev_unspare)
472 fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE);
474 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE);
476 switch (vd->vdev_stat.vs_aux) {
477 case VDEV_AUX_ERR_EXCEEDED:
478 aux = "err_exceeded";
481 case VDEV_AUX_EXTERNAL:
487 fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux);
489 if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) {
490 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID,
499 * Generate a view of the top-level vdevs. If we currently have holes
500 * in the namespace, then generate an array which contains a list of holey
501 * vdevs. Additionally, add the number of top-level children that currently
505 vdev_top_config_generate(spa_t *spa, nvlist_t *config)
507 vdev_t *rvd = spa->spa_root_vdev;
511 array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP);
513 for (c = 0, idx = 0; c < rvd->vdev_children; c++) {
514 vdev_t *tvd = rvd->vdev_child[c];
516 if (tvd->vdev_ishole) {
522 VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY,
526 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN,
527 rvd->vdev_children) == 0);
529 kmem_free(array, rvd->vdev_children * sizeof (uint64_t));
533 * Returns the configuration from the label of the given vdev. For vdevs
534 * which don't have a txg value stored on their label (i.e. spares/cache)
535 * or have not been completely initialized (txg = 0) just return
536 * the configuration from the first valid label we find. Otherwise,
537 * find the most up-to-date label that does not exceed the specified
541 vdev_label_read_config(vdev_t *vd, uint64_t txg)
543 spa_t *spa = vd->vdev_spa;
544 nvlist_t *config = NULL;
548 uint64_t best_txg = 0;
549 uint64_t label_txg = 0;
551 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
552 ZIO_FLAG_SPECULATIVE;
554 ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
556 if (!vdev_readable(vd))
559 vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE);
560 vp = abd_to_buf(vp_abd);
563 for (int l = 0; l < VDEV_LABELS; l++) {
564 nvlist_t *label = NULL;
566 zio = zio_root(spa, NULL, NULL, flags);
568 vdev_label_read(zio, vd, l, vp_abd,
569 offsetof(vdev_label_t, vl_vdev_phys),
570 sizeof (vdev_phys_t), NULL, NULL, flags);
572 if (zio_wait(zio) == 0 &&
573 nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist),
576 * Auxiliary vdevs won't have txg values in their
577 * labels and newly added vdevs may not have been
578 * completely initialized so just return the
579 * configuration from the first valid label we
582 error = nvlist_lookup_uint64(label,
583 ZPOOL_CONFIG_POOL_TXG, &label_txg);
584 if ((error || label_txg == 0) && !config) {
587 } else if (label_txg <= txg && label_txg > best_txg) {
588 best_txg = label_txg;
590 config = fnvlist_dup(label);
600 if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) {
601 flags |= ZIO_FLAG_TRYHARD;
606 * We found a valid label but it didn't pass txg restrictions.
608 if (config == NULL && label_txg != 0) {
609 vdev_dbgmsg(vd, "label discarded as txg is too large "
610 "(%llu > %llu)", (u_longlong_t)label_txg,
620 * Determine if a device is in use. The 'spare_guid' parameter will be filled
621 * in with the device guid if this spare is active elsewhere on the system.
624 vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason,
625 uint64_t *spare_guid, uint64_t *l2cache_guid)
627 spa_t *spa = vd->vdev_spa;
628 uint64_t state, pool_guid, device_guid, txg, spare_pool;
635 *l2cache_guid = 0ULL;
638 * Read the label, if any, and perform some basic sanity checks.
640 if ((label = vdev_label_read_config(vd, -1ULL)) == NULL)
643 (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
646 if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
648 nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID,
649 &device_guid) != 0) {
654 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
655 (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID,
657 nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
666 * Check to see if this device indeed belongs to the pool it claims to
667 * be a part of. The only way this is allowed is if the device is a hot
668 * spare (which we check for later on).
670 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
671 !spa_guid_exists(pool_guid, device_guid) &&
672 !spa_spare_exists(device_guid, NULL, NULL) &&
673 !spa_l2cache_exists(device_guid, NULL))
677 * If the transaction group is zero, then this an initialized (but
678 * unused) label. This is only an error if the create transaction
679 * on-disk is the same as the one we're using now, in which case the
680 * user has attempted to add the same vdev multiple times in the same
683 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
684 txg == 0 && vdtxg == crtxg)
688 * Check to see if this is a spare device. We do an explicit check for
689 * spa_has_spare() here because it may be on our pending list of spares
690 * to add. We also check if it is an l2cache device.
692 if (spa_spare_exists(device_guid, &spare_pool, NULL) ||
693 spa_has_spare(spa, device_guid)) {
695 *spare_guid = device_guid;
698 case VDEV_LABEL_CREATE:
699 case VDEV_LABEL_L2CACHE:
702 case VDEV_LABEL_REPLACE:
703 return (!spa_has_spare(spa, device_guid) ||
706 case VDEV_LABEL_SPARE:
707 return (spa_has_spare(spa, device_guid));
712 * Check to see if this is an l2cache device.
714 if (spa_l2cache_exists(device_guid, NULL))
718 * We can't rely on a pool's state if it's been imported
719 * read-only. Instead we look to see if the pools is marked
720 * read-only in the namespace and set the state to active.
722 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
723 (spa = spa_by_guid(pool_guid, device_guid)) != NULL &&
724 spa_mode(spa) == FREAD)
725 state = POOL_STATE_ACTIVE;
728 * If the device is marked ACTIVE, then this device is in use by another
729 * pool on the system.
731 return (state == POOL_STATE_ACTIVE);
735 * Initialize a vdev label. We check to make sure each leaf device is not in
736 * use, and writable. We put down an initial label which we will later
737 * overwrite with a complete label. Note that it's important to do this
738 * sequentially, not in parallel, so that we catch cases of multiple use of the
739 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
743 vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
745 spa_t *spa = vd->vdev_spa;
756 uint64_t spare_guid, l2cache_guid;
757 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
759 ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
761 for (int c = 0; c < vd->vdev_children; c++)
762 if ((error = vdev_label_init(vd->vdev_child[c],
763 crtxg, reason)) != 0)
766 /* Track the creation time for this vdev */
767 vd->vdev_crtxg = crtxg;
769 if (!vd->vdev_ops->vdev_op_leaf || !spa_writeable(spa))
773 * Dead vdevs cannot be initialized.
775 if (vdev_is_dead(vd))
776 return (SET_ERROR(EIO));
779 * Determine if the vdev is in use.
781 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT &&
782 vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid))
783 return (SET_ERROR(EBUSY));
786 * If this is a request to add or replace a spare or l2cache device
787 * that is in use elsewhere on the system, then we must update the
788 * guid (which was initialized to a random value) to reflect the
789 * actual GUID (which is shared between multiple pools).
791 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE &&
792 spare_guid != 0ULL) {
793 uint64_t guid_delta = spare_guid - vd->vdev_guid;
795 vd->vdev_guid += guid_delta;
797 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
798 pvd->vdev_guid_sum += guid_delta;
801 * If this is a replacement, then we want to fallthrough to the
802 * rest of the code. If we're adding a spare, then it's already
803 * labeled appropriately and we can just return.
805 if (reason == VDEV_LABEL_SPARE)
807 ASSERT(reason == VDEV_LABEL_REPLACE ||
808 reason == VDEV_LABEL_SPLIT);
811 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE &&
812 l2cache_guid != 0ULL) {
813 uint64_t guid_delta = l2cache_guid - vd->vdev_guid;
815 vd->vdev_guid += guid_delta;
817 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
818 pvd->vdev_guid_sum += guid_delta;
821 * If this is a replacement, then we want to fallthrough to the
822 * rest of the code. If we're adding an l2cache, then it's
823 * already labeled appropriately and we can just return.
825 if (reason == VDEV_LABEL_L2CACHE)
827 ASSERT(reason == VDEV_LABEL_REPLACE);
831 * TRIM the whole thing, excluding the blank space and boot header
832 * as specified by ZFS On-Disk Specification (section 1.3), so that
833 * we start with a clean slate.
834 * It's just an optimization, so we don't care if it fails.
835 * Don't TRIM if removing so that we don't interfere with zpool
838 if (zfs_trim_enabled && vdev_trim_on_init && !vd->vdev_notrim &&
839 (reason == VDEV_LABEL_CREATE || reason == VDEV_LABEL_SPARE ||
840 reason == VDEV_LABEL_L2CACHE))
841 zio_wait(zio_trim(NULL, spa, vd, VDEV_SKIP_SIZE,
842 vd->vdev_psize - VDEV_SKIP_SIZE));
845 * Initialize its label.
847 vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE);
848 abd_zero(vp_abd, sizeof (vdev_phys_t));
849 vp = abd_to_buf(vp_abd);
852 * Generate a label describing the pool and our top-level vdev.
853 * We mark it as being from txg 0 to indicate that it's not
854 * really part of an active pool just yet. The labels will
855 * be written again with a meaningful txg by spa_sync().
857 if (reason == VDEV_LABEL_SPARE ||
858 (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) {
860 * For inactive hot spares, we generate a special label that
861 * identifies as a mutually shared hot spare. We write the
862 * label if we are adding a hot spare, or if we are removing an
863 * active hot spare (in which case we want to revert the
866 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
868 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
869 spa_version(spa)) == 0);
870 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
871 POOL_STATE_SPARE) == 0);
872 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
873 vd->vdev_guid) == 0);
874 } else if (reason == VDEV_LABEL_L2CACHE ||
875 (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) {
877 * For level 2 ARC devices, add a special label.
879 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
881 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
882 spa_version(spa)) == 0);
883 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
884 POOL_STATE_L2CACHE) == 0);
885 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
886 vd->vdev_guid) == 0);
890 if (reason == VDEV_LABEL_SPLIT)
891 txg = spa->spa_uberblock.ub_txg;
892 label = spa_config_generate(spa, vd, txg, B_FALSE);
895 * Add our creation time. This allows us to detect multiple
896 * vdev uses as described above, and automatically expires if we
899 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
904 buflen = sizeof (vp->vp_nvlist);
906 error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP);
910 /* EFAULT means nvlist_pack ran out of room */
911 return (error == EFAULT ? ENAMETOOLONG : EINVAL);
915 * Initialize uberblock template.
917 ub_abd = abd_alloc_linear(VDEV_UBERBLOCK_RING, B_TRUE);
918 abd_zero(ub_abd, VDEV_UBERBLOCK_RING);
919 abd_copy_from_buf(ub_abd, &spa->spa_uberblock, sizeof (uberblock_t));
920 ub = abd_to_buf(ub_abd);
923 /* Initialize the 2nd padding area. */
924 pad2 = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE);
925 abd_zero(pad2, VDEV_PAD_SIZE);
928 * Write everything in parallel.
931 zio = zio_root(spa, NULL, NULL, flags);
933 for (int l = 0; l < VDEV_LABELS; l++) {
935 vdev_label_write(zio, vd, l, vp_abd,
936 offsetof(vdev_label_t, vl_vdev_phys),
937 sizeof (vdev_phys_t), NULL, NULL, flags);
940 * Skip the 1st padding area.
941 * Zero out the 2nd padding area where it might have
942 * left over data from previous filesystem format.
944 vdev_label_write(zio, vd, l, pad2,
945 offsetof(vdev_label_t, vl_pad2),
946 VDEV_PAD_SIZE, NULL, NULL, flags);
948 vdev_label_write(zio, vd, l, ub_abd,
949 offsetof(vdev_label_t, vl_uberblock),
950 VDEV_UBERBLOCK_RING, NULL, NULL, flags);
953 error = zio_wait(zio);
955 if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) {
956 flags |= ZIO_FLAG_TRYHARD;
966 * If this vdev hasn't been previously identified as a spare, then we
967 * mark it as such only if a) we are labeling it as a spare, or b) it
968 * exists as a spare elsewhere in the system. Do the same for
969 * level 2 ARC devices.
971 if (error == 0 && !vd->vdev_isspare &&
972 (reason == VDEV_LABEL_SPARE ||
973 spa_spare_exists(vd->vdev_guid, NULL, NULL)))
976 if (error == 0 && !vd->vdev_isl2cache &&
977 (reason == VDEV_LABEL_L2CACHE ||
978 spa_l2cache_exists(vd->vdev_guid, NULL)))
985 vdev_label_write_pad2(vdev_t *vd, const char *buf, size_t size)
987 spa_t *spa = vd->vdev_spa;
990 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
993 if (size > VDEV_PAD_SIZE)
996 if (!vd->vdev_ops->vdev_op_leaf)
998 if (vdev_is_dead(vd))
1001 ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
1003 pad2 = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE);
1004 abd_zero(pad2, VDEV_PAD_SIZE);
1005 abd_copy_from_buf(pad2, buf, size);
1008 zio = zio_root(spa, NULL, NULL, flags);
1009 vdev_label_write(zio, vd, 0, pad2,
1010 offsetof(vdev_label_t, vl_pad2),
1011 VDEV_PAD_SIZE, NULL, NULL, flags);
1012 error = zio_wait(zio);
1013 if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) {
1014 flags |= ZIO_FLAG_TRYHARD;
1023 * ==========================================================================
1024 * uberblock load/sync
1025 * ==========================================================================
1029 * Consider the following situation: txg is safely synced to disk. We've
1030 * written the first uberblock for txg + 1, and then we lose power. When we
1031 * come back up, we fail to see the uberblock for txg + 1 because, say,
1032 * it was on a mirrored device and the replica to which we wrote txg + 1
1033 * is now offline. If we then make some changes and sync txg + 1, and then
1034 * the missing replica comes back, then for a few seconds we'll have two
1035 * conflicting uberblocks on disk with the same txg. The solution is simple:
1036 * among uberblocks with equal txg, choose the one with the latest timestamp.
1039 vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2)
1041 if (ub1->ub_txg < ub2->ub_txg)
1043 if (ub1->ub_txg > ub2->ub_txg)
1046 if (ub1->ub_timestamp < ub2->ub_timestamp)
1048 if (ub1->ub_timestamp > ub2->ub_timestamp)
1055 uberblock_t *ubl_ubbest; /* Best uberblock */
1056 vdev_t *ubl_vd; /* vdev associated with the above */
1060 vdev_uberblock_load_done(zio_t *zio)
1062 vdev_t *vd = zio->io_vd;
1063 spa_t *spa = zio->io_spa;
1064 zio_t *rio = zio->io_private;
1065 uberblock_t *ub = abd_to_buf(zio->io_abd);
1066 struct ubl_cbdata *cbp = rio->io_private;
1068 ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd));
1070 if (zio->io_error == 0 && uberblock_verify(ub) == 0) {
1071 mutex_enter(&rio->io_lock);
1072 if (ub->ub_txg <= spa->spa_load_max_txg &&
1073 vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) {
1075 * Keep track of the vdev in which this uberblock
1076 * was found. We will use this information later
1077 * to obtain the config nvlist associated with
1080 *cbp->ubl_ubbest = *ub;
1083 mutex_exit(&rio->io_lock);
1086 abd_free(zio->io_abd);
1090 vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags,
1091 struct ubl_cbdata *cbp)
1093 for (int c = 0; c < vd->vdev_children; c++)
1094 vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp);
1096 if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) {
1097 for (int l = 0; l < VDEV_LABELS; l++) {
1098 for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
1099 vdev_label_read(zio, vd, l,
1100 abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd),
1101 B_TRUE), VDEV_UBERBLOCK_OFFSET(vd, n),
1102 VDEV_UBERBLOCK_SIZE(vd),
1103 vdev_uberblock_load_done, zio, flags);
1110 * Reads the 'best' uberblock from disk along with its associated
1111 * configuration. First, we read the uberblock array of each label of each
1112 * vdev, keeping track of the uberblock with the highest txg in each array.
1113 * Then, we read the configuration from the same vdev as the best uberblock.
1116 vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config)
1119 spa_t *spa = rvd->vdev_spa;
1120 struct ubl_cbdata cb;
1121 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
1122 ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD;
1127 bzero(ub, sizeof (uberblock_t));
1133 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
1134 zio = zio_root(spa, NULL, &cb, flags);
1135 vdev_uberblock_load_impl(zio, rvd, flags, &cb);
1136 (void) zio_wait(zio);
1139 * It's possible that the best uberblock was discovered on a label
1140 * that has a configuration which was written in a future txg.
1141 * Search all labels on this vdev to find the configuration that
1142 * matches the txg for our uberblock.
1144 if (cb.ubl_vd != NULL) {
1145 vdev_dbgmsg(cb.ubl_vd, "best uberblock found for spa %s. "
1146 "txg %llu", spa->spa_name, (u_longlong_t)ub->ub_txg);
1148 *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg);
1149 if (*config == NULL && spa->spa_extreme_rewind) {
1150 vdev_dbgmsg(cb.ubl_vd, "failed to read label config. "
1151 "Trying again without txg restrictions.");
1152 *config = vdev_label_read_config(cb.ubl_vd, UINT64_MAX);
1154 if (*config == NULL) {
1155 vdev_dbgmsg(cb.ubl_vd, "failed to read label config");
1158 spa_config_exit(spa, SCL_ALL, FTAG);
1162 * On success, increment root zio's count of good writes.
1163 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
1166 vdev_uberblock_sync_done(zio_t *zio)
1168 uint64_t *good_writes = zio->io_private;
1170 if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0)
1171 atomic_inc_64(good_writes);
1175 * Write the uberblock to all labels of all leaves of the specified vdev.
1178 vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags)
1180 for (uint64_t c = 0; c < vd->vdev_children; c++)
1181 vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags);
1183 if (!vd->vdev_ops->vdev_op_leaf)
1186 if (!vdev_writeable(vd))
1189 int n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1);
1191 /* Copy the uberblock_t into the ABD */
1192 abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE);
1193 abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd));
1194 abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t));
1196 for (int l = 0; l < VDEV_LABELS; l++)
1197 vdev_label_write(zio, vd, l, ub_abd,
1198 VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd),
1199 vdev_uberblock_sync_done, zio->io_private,
1200 flags | ZIO_FLAG_DONT_PROPAGATE);
1205 /* Sync the uberblocks to all vdevs in svd[] */
1207 vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags)
1209 spa_t *spa = svd[0]->vdev_spa;
1211 uint64_t good_writes = 0;
1213 zio = zio_root(spa, NULL, &good_writes, flags);
1215 for (int v = 0; v < svdcount; v++)
1216 vdev_uberblock_sync(zio, ub, svd[v], flags);
1218 (void) zio_wait(zio);
1221 * Flush the uberblocks to disk. This ensures that the odd labels
1222 * are no longer needed (because the new uberblocks and the even
1223 * labels are safely on disk), so it is safe to overwrite them.
1225 zio = zio_root(spa, NULL, NULL, flags);
1227 for (int v = 0; v < svdcount; v++) {
1228 if (vdev_writeable(svd[v])) {
1229 zio_flush(zio, svd[v]);
1233 (void) zio_wait(zio);
1235 return (good_writes >= 1 ? 0 : EIO);
1239 * On success, increment the count of good writes for our top-level vdev.
1242 vdev_label_sync_done(zio_t *zio)
1244 uint64_t *good_writes = zio->io_private;
1246 if (zio->io_error == 0)
1247 atomic_inc_64(good_writes);
1251 * If there weren't enough good writes, indicate failure to the parent.
1254 vdev_label_sync_top_done(zio_t *zio)
1256 uint64_t *good_writes = zio->io_private;
1258 if (*good_writes == 0)
1259 zio->io_error = SET_ERROR(EIO);
1261 kmem_free(good_writes, sizeof (uint64_t));
1265 * We ignore errors for log and cache devices, simply free the private data.
1268 vdev_label_sync_ignore_done(zio_t *zio)
1270 kmem_free(zio->io_private, sizeof (uint64_t));
1274 * Write all even or odd labels to all leaves of the specified vdev.
1277 vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags)
1285 for (int c = 0; c < vd->vdev_children; c++)
1286 vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags);
1288 if (!vd->vdev_ops->vdev_op_leaf)
1291 if (!vdev_writeable(vd))
1295 * Generate a label describing the top-level config to which we belong.
1297 label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE);
1299 vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE);
1300 abd_zero(vp_abd, sizeof (vdev_phys_t));
1301 vp = abd_to_buf(vp_abd);
1303 buf = vp->vp_nvlist;
1304 buflen = sizeof (vp->vp_nvlist);
1306 if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) {
1307 for (; l < VDEV_LABELS; l += 2) {
1308 vdev_label_write(zio, vd, l, vp_abd,
1309 offsetof(vdev_label_t, vl_vdev_phys),
1310 sizeof (vdev_phys_t),
1311 vdev_label_sync_done, zio->io_private,
1312 flags | ZIO_FLAG_DONT_PROPAGATE);
1321 vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags)
1323 list_t *dl = &spa->spa_config_dirty_list;
1329 * Write the new labels to disk.
1331 zio = zio_root(spa, NULL, NULL, flags);
1333 for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) {
1334 uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t),
1337 ASSERT(!vd->vdev_ishole);
1339 zio_t *vio = zio_null(zio, spa, NULL,
1340 (vd->vdev_islog || vd->vdev_aux != NULL) ?
1341 vdev_label_sync_ignore_done : vdev_label_sync_top_done,
1342 good_writes, flags);
1343 vdev_label_sync(vio, vd, l, txg, flags);
1347 error = zio_wait(zio);
1350 * Flush the new labels to disk.
1352 zio = zio_root(spa, NULL, NULL, flags);
1354 for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd))
1357 (void) zio_wait(zio);
1363 * Sync the uberblock and any changes to the vdev configuration.
1365 * The order of operations is carefully crafted to ensure that
1366 * if the system panics or loses power at any time, the state on disk
1367 * is still transactionally consistent. The in-line comments below
1368 * describe the failure semantics at each stage.
1370 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1371 * at any time, you can just call it again, and it will resume its work.
1374 vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg)
1376 spa_t *spa = svd[0]->vdev_spa;
1377 uberblock_t *ub = &spa->spa_uberblock;
1379 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
1381 ASSERT(svdcount != 0);
1384 * Normally, we don't want to try too hard to write every label and
1385 * uberblock. If there is a flaky disk, we don't want the rest of the
1386 * sync process to block while we retry. But if we can't write a
1387 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1388 * bailing out and declaring the pool faulted.
1391 if ((flags & ZIO_FLAG_TRYHARD) != 0)
1393 flags |= ZIO_FLAG_TRYHARD;
1396 ASSERT(ub->ub_txg <= txg);
1399 * If this isn't a resync due to I/O errors,
1400 * and nothing changed in this transaction group,
1401 * and the vdev configuration hasn't changed,
1402 * then there's nothing to do.
1404 if (ub->ub_txg < txg &&
1405 uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE &&
1406 list_is_empty(&spa->spa_config_dirty_list))
1409 if (txg > spa_freeze_txg(spa))
1412 ASSERT(txg <= spa->spa_final_txg);
1415 * Flush the write cache of every disk that's been written to
1416 * in this transaction group. This ensures that all blocks
1417 * written in this txg will be committed to stable storage
1418 * before any uberblock that references them.
1420 zio_t *zio = zio_root(spa, NULL, NULL, flags);
1423 txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd != NULL;
1424 vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg)))
1427 (void) zio_wait(zio);
1430 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1431 * system dies in the middle of this process, that's OK: all of the
1432 * even labels that made it to disk will be newer than any uberblock,
1433 * and will therefore be considered invalid. The odd labels (L1, L3),
1434 * which have not yet been touched, will still be valid. We flush
1435 * the new labels to disk to ensure that all even-label updates
1436 * are committed to stable storage before the uberblock update.
1438 if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) {
1439 if ((flags & ZIO_FLAG_TRYHARD) != 0) {
1440 zfs_dbgmsg("vdev_label_sync_list() returned error %d "
1441 "for pool '%s' when syncing out the even labels "
1442 "of dirty vdevs", error, spa_name(spa));
1448 * Sync the uberblocks to all vdevs in svd[].
1449 * If the system dies in the middle of this step, there are two cases
1450 * to consider, and the on-disk state is consistent either way:
1452 * (1) If none of the new uberblocks made it to disk, then the
1453 * previous uberblock will be the newest, and the odd labels
1454 * (which had not yet been touched) will be valid with respect
1455 * to that uberblock.
1457 * (2) If one or more new uberblocks made it to disk, then they
1458 * will be the newest, and the even labels (which had all
1459 * been successfully committed) will be valid with respect
1460 * to the new uberblocks.
1462 if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) {
1463 if ((flags & ZIO_FLAG_TRYHARD) != 0) {
1464 zfs_dbgmsg("vdev_uberblock_sync_list() returned error "
1465 "%d for pool '%s'", error, spa_name(spa));
1471 * Sync out odd labels for every dirty vdev. If the system dies
1472 * in the middle of this process, the even labels and the new
1473 * uberblocks will suffice to open the pool. The next time
1474 * the pool is opened, the first thing we'll do -- before any
1475 * user data is modified -- is mark every vdev dirty so that
1476 * all labels will be brought up to date. We flush the new labels
1477 * to disk to ensure that all odd-label updates are committed to
1478 * stable storage before the next transaction group begins.
1480 if ((error = vdev_label_sync_list(spa, 1, txg, flags)) != 0) {
1481 if ((flags & ZIO_FLAG_TRYHARD) != 0) {
1482 zfs_dbgmsg("vdev_label_sync_list() returned error %d "
1483 "for pool '%s' when syncing out the odd labels of "
1484 "dirty vdevs", error, spa_name(spa));
1489 trim_thread_wakeup(spa);