1 /*
2 * CDDL HEADER START
3 *
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.
7 *
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.
12 *
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]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2013 by Delphix. All rights reserved.
25 */
26
27 /*
28 * Virtual Device Labels
29 * ---------------------
30 *
31 * The vdev label serves several distinct purposes:
32 *
33 * 1. Uniquely identify this device as part of a ZFS pool and confirm its
34 * identity within the pool.
35 *
36 * 2. Verify that all the devices given in a configuration are present
37 * within the pool.
38 *
39 * 3. Determine the uberblock for the pool.
40 *
41 * 4. In case of an import operation, determine the configuration of the
42 * toplevel vdev of which it is a part.
43 *
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.
47 *
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.
52 *
53 *
54 * Label Organization
55 * ------------------
56 *
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.
59 *
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:
65 *
66 * L1 UB L2
67 * +------+ +------+ +------+
68 * | | | | | |
69 * | t10 | | t10 | | t10 |
70 * | | | | | |
71 * +------+ +------+ +------+
72 *
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.
76 *
77 * In order to identify which labels are valid, the labels are written in the
78 * following manner:
79 *
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
83 *
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.
90 *
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
95 * on another vdev.
96 *
97 *
98 * On-disk Format
99 * --------------
100 *
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.
104 *
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.
108 *
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.
113 *
114 *
115 * Configuration Information
116 * -------------------------
117 *
118 * The nvlist describing the pool and vdev contains the following elements:
119 *
120 * version ZFS on-disk version
121 * name Pool name
122 * state Pool state
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.
126 * features_for_read
127 * An nvlist of the features necessary for reading the MOS.
128 *
129 * Each leaf device label also contains the following:
130 *
131 * top_guid Unique ID for top-level vdev in which this is contained
132 * guid Unique ID for the leaf vdev
133 *
134 * The 'vs' configuration follows the format described in 'spa_config.c'.
135 */
136
137 #include <sys/zfs_context.h>
138 #include <sys/spa.h>
139 #include <sys/spa_impl.h>
140 #include <sys/dmu.h>
141 #include <sys/zap.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/zio.h>
147 #include <sys/dsl_scan.h>
148 #include <sys/fs/zfs.h>
149
150 /*
151 * Basic routines to read and write from a vdev label.
152 * Used throughout the rest of this file.
153 */
154 uint64_t
155 vdev_label_offset(uint64_t psize, int l, uint64_t offset)
156 {
157 ASSERT(offset < sizeof (vdev_label_t));
158 ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0);
159
160 return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
161 0 : psize - VDEV_LABELS * sizeof (vdev_label_t)));
162 }
163
164 /*
165 * Returns back the vdev label associated with the passed in offset.
166 */
167 int
168 vdev_label_number(uint64_t psize, uint64_t offset)
169 {
170 int l;
171
172 if (offset >= psize - VDEV_LABEL_END_SIZE) {
173 offset -= psize - VDEV_LABEL_END_SIZE;
174 offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t);
175 }
176 l = offset / sizeof (vdev_label_t);
177 return (l < VDEV_LABELS ? l : -1);
178 }
179
180 static void
181 vdev_label_read(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset,
182 uint64_t size, zio_done_func_t *done, void *private, int flags)
183 {
184 ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) ==
185 SCL_STATE_ALL);
186 ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
187
188 zio_nowait(zio_read_phys(zio, vd,
189 vdev_label_offset(vd->vdev_psize, l, offset),
190 size, buf, ZIO_CHECKSUM_LABEL, done, private,
191 ZIO_PRIORITY_SYNC_READ, flags, B_TRUE));
192 }
193
194 static void
195 vdev_label_write(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset,
196 uint64_t size, zio_done_func_t *done, void *private, int flags)
197 {
198 ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL ||
199 (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
200 (SCL_CONFIG | SCL_STATE) &&
201 dsl_pool_sync_context(spa_get_dsl(zio->io_spa))));
202 ASSERT(flags & ZIO_FLAG_CONFIG_WRITER);
203
204 zio_nowait(zio_write_phys(zio, vd,
205 vdev_label_offset(vd->vdev_psize, l, offset),
206 size, buf, ZIO_CHECKSUM_LABEL, done, private,
207 ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE));
208 }
209
210 /*
211 * Generate the nvlist representing this vdev's config.
212 */
213 nvlist_t *
214 vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats,
215 vdev_config_flag_t flags)
216 {
217 nvlist_t *nv = NULL;
218
219 nv = fnvlist_alloc();
220
221 fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type);
222 if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)))
223 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id);
224 fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid);
225
226 if (vd->vdev_path != NULL)
227 fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path);
228
229 if (vd->vdev_devid != NULL)
230 fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid);
231
232 if (vd->vdev_physpath != NULL)
233 fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH,
234 vd->vdev_physpath);
235
236 if (vd->vdev_fru != NULL)
237 fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru);
238
239 if (vd->vdev_nparity != 0) {
240 ASSERT(strcmp(vd->vdev_ops->vdev_op_type,
241 VDEV_TYPE_RAIDZ) == 0);
242
243 /*
244 * Make sure someone hasn't managed to sneak a fancy new vdev
245 * into a crufty old storage pool.
246 */
247 ASSERT(vd->vdev_nparity == 1 ||
248 (vd->vdev_nparity <= 2 &&
249 spa_version(spa) >= SPA_VERSION_RAIDZ2) ||
250 (vd->vdev_nparity <= 3 &&
251 spa_version(spa) >= SPA_VERSION_RAIDZ3));
252
253 /*
254 * Note that we'll add the nparity tag even on storage pools
255 * that only support a single parity device -- older software
256 * will just ignore it.
257 */
258 fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vd->vdev_nparity);
259 }
260
261 if (vd->vdev_wholedisk != -1ULL)
262 fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
263 vd->vdev_wholedisk);
264
265 if (vd->vdev_not_present)
266 fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1);
267
268 if (vd->vdev_isspare)
269 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1);
270
271 if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) &&
272 vd == vd->vdev_top) {
273 fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
274 vd->vdev_ms_array);
275 fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
276 vd->vdev_ms_shift);
277 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift);
278 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE,
279 vd->vdev_asize);
280 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog);
281 if (vd->vdev_removing)
282 fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING,
283 vd->vdev_removing);
284 }
285
286 if (vd->vdev_dtl_smo.smo_object != 0)
287 fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL,
288 vd->vdev_dtl_smo.smo_object);
289
290 if (vd->vdev_crtxg)
291 fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg);
292
293 if (getstats) {
294 vdev_stat_t vs;
295 pool_scan_stat_t ps;
296
297 vdev_get_stats(vd, &vs);
298 fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
299 (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t));
300
301 /* provide either current or previous scan information */
302 if (spa_scan_get_stats(spa, &ps) == 0) {
303 fnvlist_add_uint64_array(nv,
304 ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps,
305 sizeof (pool_scan_stat_t) / sizeof (uint64_t));
306 }
307 }
308
309 if (!vd->vdev_ops->vdev_op_leaf) {
310 nvlist_t **child;
311 int c, idx;
312
313 ASSERT(!vd->vdev_ishole);
314
315 child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *),
316 KM_SLEEP);
317
318 for (c = 0, idx = 0; c < vd->vdev_children; c++) {
319 vdev_t *cvd = vd->vdev_child[c];
320
321 /*
322 * If we're generating an nvlist of removing
323 * vdevs then skip over any device which is
324 * not being removed.
325 */
326 if ((flags & VDEV_CONFIG_REMOVING) &&
327 !cvd->vdev_removing)
328 continue;
329
330 child[idx++] = vdev_config_generate(spa, cvd,
331 getstats, flags);
332 }
333
334 if (idx) {
335 fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
336 child, idx);
337 }
338
339 for (c = 0; c < idx; c++)
340 nvlist_free(child[c]);
341
342 kmem_free(child, vd->vdev_children * sizeof (nvlist_t *));
343
344 } else {
345 const char *aux = NULL;
346
347 if (vd->vdev_offline && !vd->vdev_tmpoffline)
348 fnvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, B_TRUE);
349 if (vd->vdev_resilver_txg != 0)
350 fnvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG,
351 vd->vdev_resilver_txg);
352 if (vd->vdev_faulted)
353 fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE);
354 if (vd->vdev_degraded)
355 fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE);
356 if (vd->vdev_removed)
357 fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE);
358 if (vd->vdev_unspare)
359 fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE);
360 if (vd->vdev_ishole)
361 fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE);
362
363 switch (vd->vdev_stat.vs_aux) {
364 case VDEV_AUX_ERR_EXCEEDED:
365 aux = "err_exceeded";
366 break;
367
368 case VDEV_AUX_EXTERNAL:
369 aux = "external";
370 break;
371 }
372
373 if (aux != NULL)
374 fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux);
375
376 if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) {
377 fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID,
378 vd->vdev_orig_guid);
379 }
380 }
381
382 return (nv);
383 }
384
385 /*
386 * Generate a view of the top-level vdevs. If we currently have holes
387 * in the namespace, then generate an array which contains a list of holey
388 * vdevs. Additionally, add the number of top-level children that currently
389 * exist.
390 */
391 void
392 vdev_top_config_generate(spa_t *spa, nvlist_t *config)
393 {
394 vdev_t *rvd = spa->spa_root_vdev;
395 uint64_t *array;
396 uint_t c, idx;
397
398 array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP);
399
400 for (c = 0, idx = 0; c < rvd->vdev_children; c++) {
401 vdev_t *tvd = rvd->vdev_child[c];
402
403 if (tvd->vdev_ishole)
404 array[idx++] = c;
405 }
406
407 if (idx) {
408 VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY,
409 array, idx) == 0);
410 }
411
412 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN,
413 rvd->vdev_children) == 0);
414
415 kmem_free(array, rvd->vdev_children * sizeof (uint64_t));
416 }
417
418 /*
419 * Returns the configuration from the label of the given vdev. For vdevs
420 * which don't have a txg value stored on their label (i.e. spares/cache)
421 * or have not been completely initialized (txg = 0) just return
422 * the configuration from the first valid label we find. Otherwise,
423 * find the most up-to-date label that does not exceed the specified
424 * 'txg' value.
425 */
426 nvlist_t *
427 vdev_label_read_config(vdev_t *vd, uint64_t txg)
428 {
429 spa_t *spa = vd->vdev_spa;
430 nvlist_t *config = NULL;
431 vdev_phys_t *vp;
432 zio_t *zio;
433 uint64_t best_txg = 0;
434 int error = 0;
435 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
436 ZIO_FLAG_SPECULATIVE;
437
438 ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
439
440 if (!vdev_readable(vd))
441 return (NULL);
442
443 vp = zio_buf_alloc(sizeof (vdev_phys_t));
444
445 retry:
446 for (int l = 0; l < VDEV_LABELS; l++) {
447 nvlist_t *label = NULL;
448
449 zio = zio_root(spa, NULL, NULL, flags);
450
451 vdev_label_read(zio, vd, l, vp,
452 offsetof(vdev_label_t, vl_vdev_phys),
453 sizeof (vdev_phys_t), NULL, NULL, flags);
454
455 if (zio_wait(zio) == 0 &&
456 nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist),
457 &label, 0) == 0) {
458 uint64_t label_txg = 0;
459
460 /*
461 * Auxiliary vdevs won't have txg values in their
462 * labels and newly added vdevs may not have been
463 * completely initialized so just return the
464 * configuration from the first valid label we
465 * encounter.
466 */
467 error = nvlist_lookup_uint64(label,
468 ZPOOL_CONFIG_POOL_TXG, &label_txg);
469 if ((error || label_txg == 0) && !config) {
470 config = label;
471 break;
472 } else if (label_txg <= txg && label_txg > best_txg) {
473 best_txg = label_txg;
474 nvlist_free(config);
475 config = fnvlist_dup(label);
476 }
477 }
478
479 if (label != NULL) {
480 nvlist_free(label);
481 label = NULL;
482 }
483 }
484
485 if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) {
486 flags |= ZIO_FLAG_TRYHARD;
487 goto retry;
488 }
489
490 zio_buf_free(vp, sizeof (vdev_phys_t));
491
492 return (config);
493 }
494
495 /*
496 * Determine if a device is in use. The 'spare_guid' parameter will be filled
497 * in with the device guid if this spare is active elsewhere on the system.
498 */
499 static boolean_t
500 vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason,
501 uint64_t *spare_guid, uint64_t *l2cache_guid)
502 {
503 spa_t *spa = vd->vdev_spa;
504 uint64_t state, pool_guid, device_guid, txg, spare_pool;
505 uint64_t vdtxg = 0;
506 nvlist_t *label;
507
508 if (spare_guid)
509 *spare_guid = 0ULL;
510 if (l2cache_guid)
511 *l2cache_guid = 0ULL;
512
513 /*
514 * Read the label, if any, and perform some basic sanity checks.
515 */
516 if ((label = vdev_label_read_config(vd, -1ULL)) == NULL)
517 return (B_FALSE);
518
519 (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
520 &vdtxg);
521
522 if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
523 &state) != 0 ||
524 nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID,
525 &device_guid) != 0) {
526 nvlist_free(label);
527 return (B_FALSE);
528 }
529
530 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
531 (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID,
532 &pool_guid) != 0 ||
533 nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
534 &txg) != 0)) {
535 nvlist_free(label);
536 return (B_FALSE);
537 }
538
539 nvlist_free(label);
540
541 /*
542 * Check to see if this device indeed belongs to the pool it claims to
543 * be a part of. The only way this is allowed is if the device is a hot
544 * spare (which we check for later on).
545 */
546 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
547 !spa_guid_exists(pool_guid, device_guid) &&
548 !spa_spare_exists(device_guid, NULL, NULL) &&
549 !spa_l2cache_exists(device_guid, NULL))
550 return (B_FALSE);
551
552 /*
553 * If the transaction group is zero, then this an initialized (but
554 * unused) label. This is only an error if the create transaction
555 * on-disk is the same as the one we're using now, in which case the
556 * user has attempted to add the same vdev multiple times in the same
557 * transaction.
558 */
559 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
560 txg == 0 && vdtxg == crtxg)
561 return (B_TRUE);
562
563 /*
564 * Check to see if this is a spare device. We do an explicit check for
565 * spa_has_spare() here because it may be on our pending list of spares
566 * to add. We also check if it is an l2cache device.
567 */
568 if (spa_spare_exists(device_guid, &spare_pool, NULL) ||
569 spa_has_spare(spa, device_guid)) {
570 if (spare_guid)
571 *spare_guid = device_guid;
572
573 switch (reason) {
574 case VDEV_LABEL_CREATE:
575 case VDEV_LABEL_L2CACHE:
576 return (B_TRUE);
577
578 case VDEV_LABEL_REPLACE:
579 return (!spa_has_spare(spa, device_guid) ||
580 spare_pool != 0ULL);
581
582 case VDEV_LABEL_SPARE:
583 return (spa_has_spare(spa, device_guid));
584 }
585 }
586
587 /*
588 * Check to see if this is an l2cache device.
589 */
590 if (spa_l2cache_exists(device_guid, NULL))
591 return (B_TRUE);
592
593 /*
594 * We can't rely on a pool's state if it's been imported
595 * read-only. Instead we look to see if the pools is marked
596 * read-only in the namespace and set the state to active.
597 */
598 if ((spa = spa_by_guid(pool_guid, device_guid)) != NULL &&
599 spa_mode(spa) == FREAD)
600 state = POOL_STATE_ACTIVE;
601
602 /*
603 * If the device is marked ACTIVE, then this device is in use by another
604 * pool on the system.
605 */
606 return (state == POOL_STATE_ACTIVE);
607 }
608
609 /*
610 * Initialize a vdev label. We check to make sure each leaf device is not in
611 * use, and writable. We put down an initial label which we will later
612 * overwrite with a complete label. Note that it's important to do this
613 * sequentially, not in parallel, so that we catch cases of multiple use of the
614 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
615 * itself.
616 */
617 int
618 vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
619 {
620 spa_t *spa = vd->vdev_spa;
621 nvlist_t *label;
622 vdev_phys_t *vp;
623 char *pad2;
624 uberblock_t *ub;
625 zio_t *zio;
626 char *buf;
627 size_t buflen;
628 int error;
629 uint64_t spare_guid, l2cache_guid;
630 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
631
632 ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
633
634 for (int c = 0; c < vd->vdev_children; c++)
635 if ((error = vdev_label_init(vd->vdev_child[c],
636 crtxg, reason)) != 0)
637 return (error);
638
639 /* Track the creation time for this vdev */
640 vd->vdev_crtxg = crtxg;
641
642 if (!vd->vdev_ops->vdev_op_leaf)
643 return (0);
644
645 /*
646 * Dead vdevs cannot be initialized.
647 */
648 if (vdev_is_dead(vd))
649 return (SET_ERROR(EIO));
650
651 /*
652 * Determine if the vdev is in use.
653 */
654 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT &&
655 vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid))
656 return (SET_ERROR(EBUSY));
657
658 /*
659 * If this is a request to add or replace a spare or l2cache device
660 * that is in use elsewhere on the system, then we must update the
661 * guid (which was initialized to a random value) to reflect the
662 * actual GUID (which is shared between multiple pools).
663 */
664 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE &&
665 spare_guid != 0ULL) {
666 uint64_t guid_delta = spare_guid - vd->vdev_guid;
667
668 vd->vdev_guid += guid_delta;
669
670 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
671 pvd->vdev_guid_sum += guid_delta;
672
673 /*
674 * If this is a replacement, then we want to fallthrough to the
675 * rest of the code. If we're adding a spare, then it's already
676 * labeled appropriately and we can just return.
677 */
678 if (reason == VDEV_LABEL_SPARE)
679 return (0);
680 ASSERT(reason == VDEV_LABEL_REPLACE ||
681 reason == VDEV_LABEL_SPLIT);
682 }
683
684 if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE &&
685 l2cache_guid != 0ULL) {
686 uint64_t guid_delta = l2cache_guid - vd->vdev_guid;
687
688 vd->vdev_guid += guid_delta;
689
690 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
691 pvd->vdev_guid_sum += guid_delta;
692
693 /*
694 * If this is a replacement, then we want to fallthrough to the
695 * rest of the code. If we're adding an l2cache, then it's
696 * already labeled appropriately and we can just return.
697 */
698 if (reason == VDEV_LABEL_L2CACHE)
699 return (0);
700 ASSERT(reason == VDEV_LABEL_REPLACE);
701 }
702
703 /*
704 * Initialize its label.
705 */
706 vp = zio_buf_alloc(sizeof (vdev_phys_t));
707 bzero(vp, sizeof (vdev_phys_t));
708
709 /*
710 * Generate a label describing the pool and our top-level vdev.
711 * We mark it as being from txg 0 to indicate that it's not
712 * really part of an active pool just yet. The labels will
713 * be written again with a meaningful txg by spa_sync().
714 */
715 if (reason == VDEV_LABEL_SPARE ||
716 (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) {
717 /*
718 * For inactive hot spares, we generate a special label that
719 * identifies as a mutually shared hot spare. We write the
720 * label if we are adding a hot spare, or if we are removing an
721 * active hot spare (in which case we want to revert the
722 * labels).
723 */
724 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
725
726 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
727 spa_version(spa)) == 0);
728 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
729 POOL_STATE_SPARE) == 0);
730 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
731 vd->vdev_guid) == 0);
732 } else if (reason == VDEV_LABEL_L2CACHE ||
733 (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) {
734 /*
735 * For level 2 ARC devices, add a special label.
736 */
737 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
738
739 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
740 spa_version(spa)) == 0);
741 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
742 POOL_STATE_L2CACHE) == 0);
743 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
744 vd->vdev_guid) == 0);
745 } else {
746 uint64_t txg = 0ULL;
747
748 if (reason == VDEV_LABEL_SPLIT)
749 txg = spa->spa_uberblock.ub_txg;
750 label = spa_config_generate(spa, vd, txg, B_FALSE);
751
752 /*
753 * Add our creation time. This allows us to detect multiple
754 * vdev uses as described above, and automatically expires if we
755 * fail.
756 */
757 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
758 crtxg) == 0);
759 }
760
761 buf = vp->vp_nvlist;
762 buflen = sizeof (vp->vp_nvlist);
763
764 error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP);
765 if (error != 0) {
766 nvlist_free(label);
767 zio_buf_free(vp, sizeof (vdev_phys_t));
768 /* EFAULT means nvlist_pack ran out of room */
769 return (error == EFAULT ? ENAMETOOLONG : EINVAL);
770 }
771
772 /*
773 * Initialize uberblock template.
774 */
775 ub = zio_buf_alloc(VDEV_UBERBLOCK_RING);
776 bzero(ub, VDEV_UBERBLOCK_RING);
777 *ub = spa->spa_uberblock;
778 ub->ub_txg = 0;
779
780 /* Initialize the 2nd padding area. */
781 pad2 = zio_buf_alloc(VDEV_PAD_SIZE);
782 bzero(pad2, VDEV_PAD_SIZE);
783
784 /*
785 * Write everything in parallel.
786 */
787 retry:
788 zio = zio_root(spa, NULL, NULL, flags);
789
790 for (int l = 0; l < VDEV_LABELS; l++) {
791
792 vdev_label_write(zio, vd, l, vp,
793 offsetof(vdev_label_t, vl_vdev_phys),
794 sizeof (vdev_phys_t), NULL, NULL, flags);
795
796 /*
797 * Skip the 1st padding area.
798 * Zero out the 2nd padding area where it might have
799 * left over data from previous filesystem format.
800 */
801 vdev_label_write(zio, vd, l, pad2,
802 offsetof(vdev_label_t, vl_pad2),
803 VDEV_PAD_SIZE, NULL, NULL, flags);
804
805 vdev_label_write(zio, vd, l, ub,
806 offsetof(vdev_label_t, vl_uberblock),
807 VDEV_UBERBLOCK_RING, NULL, NULL, flags);
808 }
809
810 error = zio_wait(zio);
811
812 if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) {
813 flags |= ZIO_FLAG_TRYHARD;
814 goto retry;
815 }
816
817 nvlist_free(label);
818 zio_buf_free(pad2, VDEV_PAD_SIZE);
819 zio_buf_free(ub, VDEV_UBERBLOCK_RING);
820 zio_buf_free(vp, sizeof (vdev_phys_t));
821
822 /*
823 * If this vdev hasn't been previously identified as a spare, then we
824 * mark it as such only if a) we are labeling it as a spare, or b) it
825 * exists as a spare elsewhere in the system. Do the same for
826 * level 2 ARC devices.
827 */
828 if (error == 0 && !vd->vdev_isspare &&
829 (reason == VDEV_LABEL_SPARE ||
830 spa_spare_exists(vd->vdev_guid, NULL, NULL)))
831 spa_spare_add(vd);
832
833 if (error == 0 && !vd->vdev_isl2cache &&
834 (reason == VDEV_LABEL_L2CACHE ||
835 spa_l2cache_exists(vd->vdev_guid, NULL)))
836 spa_l2cache_add(vd);
837
838 return (error);
839 }
840
841 /*
842 * ==========================================================================
843 * uberblock load/sync
844 * ==========================================================================
845 */
846
847 /*
848 * Consider the following situation: txg is safely synced to disk. We've
849 * written the first uberblock for txg + 1, and then we lose power. When we
850 * come back up, we fail to see the uberblock for txg + 1 because, say,
851 * it was on a mirrored device and the replica to which we wrote txg + 1
852 * is now offline. If we then make some changes and sync txg + 1, and then
853 * the missing replica comes back, then for a few seconds we'll have two
854 * conflicting uberblocks on disk with the same txg. The solution is simple:
855 * among uberblocks with equal txg, choose the one with the latest timestamp.
856 */
857 static int
858 vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2)
859 {
860 if (ub1->ub_txg < ub2->ub_txg)
861 return (-1);
862 if (ub1->ub_txg > ub2->ub_txg)
863 return (1);
864
865 if (ub1->ub_timestamp < ub2->ub_timestamp)
866 return (-1);
867 if (ub1->ub_timestamp > ub2->ub_timestamp)
868 return (1);
869
870 return (0);
871 }
872
873 struct ubl_cbdata {
874 uberblock_t *ubl_ubbest; /* Best uberblock */
875 vdev_t *ubl_vd; /* vdev associated with the above */
876 };
877
878 static void
879 vdev_uberblock_load_done(zio_t *zio)
880 {
881 vdev_t *vd = zio->io_vd;
882 spa_t *spa = zio->io_spa;
883 zio_t *rio = zio->io_private;
884 uberblock_t *ub = zio->io_data;
885 struct ubl_cbdata *cbp = rio->io_private;
886
887 ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd));
888
889 if (zio->io_error == 0 && uberblock_verify(ub) == 0) {
890 mutex_enter(&rio->io_lock);
891 if (ub->ub_txg <= spa->spa_load_max_txg &&
892 vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) {
893 /*
894 * Keep track of the vdev in which this uberblock
895 * was found. We will use this information later
896 * to obtain the config nvlist associated with
897 * this uberblock.
898 */
899 *cbp->ubl_ubbest = *ub;
900 cbp->ubl_vd = vd;
901 }
902 mutex_exit(&rio->io_lock);
903 }
904
905 zio_buf_free(zio->io_data, zio->io_size);
906 }
907
908 static void
909 vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags,
910 struct ubl_cbdata *cbp)
911 {
912 for (int c = 0; c < vd->vdev_children; c++)
913 vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp);
914
915 if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) {
916 for (int l = 0; l < VDEV_LABELS; l++) {
917 for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
918 vdev_label_read(zio, vd, l,
919 zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)),
920 VDEV_UBERBLOCK_OFFSET(vd, n),
921 VDEV_UBERBLOCK_SIZE(vd),
922 vdev_uberblock_load_done, zio, flags);
923 }
924 }
925 }
926 }
927
928 /*
929 * Reads the 'best' uberblock from disk along with its associated
930 * configuration. First, we read the uberblock array of each label of each
931 * vdev, keeping track of the uberblock with the highest txg in each array.
932 * Then, we read the configuration from the same vdev as the best uberblock.
933 */
934 void
935 vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config)
936 {
937 zio_t *zio;
938 spa_t *spa = rvd->vdev_spa;
939 struct ubl_cbdata cb;
940 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
941 ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD;
942
943 ASSERT(ub);
944 ASSERT(config);
945
946 bzero(ub, sizeof (uberblock_t));
947 *config = NULL;
948
949 cb.ubl_ubbest = ub;
950 cb.ubl_vd = NULL;
951
952 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
953 zio = zio_root(spa, NULL, &cb, flags);
954 vdev_uberblock_load_impl(zio, rvd, flags, &cb);
955 (void) zio_wait(zio);
956
957 /*
958 * It's possible that the best uberblock was discovered on a label
959 * that has a configuration which was written in a future txg.
960 * Search all labels on this vdev to find the configuration that
961 * matches the txg for our uberblock.
962 */
963 if (cb.ubl_vd != NULL)
964 *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg);
965 spa_config_exit(spa, SCL_ALL, FTAG);
966 }
967
968 /*
969 * On success, increment root zio's count of good writes.
970 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
971 */
972 static void
973 vdev_uberblock_sync_done(zio_t *zio)
974 {
975 uint64_t *good_writes = zio->io_private;
976
977 if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0)
978 atomic_add_64(good_writes, 1);
979 }
980
981 /*
982 * Write the uberblock to all labels of all leaves of the specified vdev.
983 */
984 static void
985 vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags)
986 {
987 uberblock_t *ubbuf;
988 int n;
989
990 for (int c = 0; c < vd->vdev_children; c++)
991 vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags);
992
993 if (!vd->vdev_ops->vdev_op_leaf)
994 return;
995
996 if (!vdev_writeable(vd))
997 return;
998
999 n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1);
1000
1001 ubbuf = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd));
1002 bzero(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
1003 *ubbuf = *ub;
1004
1005 for (int l = 0; l < VDEV_LABELS; l++)
1006 vdev_label_write(zio, vd, l, ubbuf,
1007 VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd),
1008 vdev_uberblock_sync_done, zio->io_private,
1009 flags | ZIO_FLAG_DONT_PROPAGATE);
1010
1011 zio_buf_free(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
1012 }
1013
1014 /* Sync the uberblocks to all vdevs in svd[] */
1015 int
1016 vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags)
1017 {
1018 spa_t *spa = svd[0]->vdev_spa;
1019 zio_t *zio;
1020 uint64_t good_writes = 0;
1021
1022 zio = zio_root(spa, NULL, &good_writes, flags);
1023
1024 for (int v = 0; v < svdcount; v++)
1025 vdev_uberblock_sync(zio, ub, svd[v], flags);
1026
1027 (void) zio_wait(zio);
1028
1029 /*
1030 * Flush the uberblocks to disk. This ensures that the odd labels
1031 * are no longer needed (because the new uberblocks and the even
1032 * labels are safely on disk), so it is safe to overwrite them.
1033 */
1034 zio = zio_root(spa, NULL, NULL, flags);
1035
1036 for (int v = 0; v < svdcount; v++)
1037 zio_flush(zio, svd[v]);
1038
1039 (void) zio_wait(zio);
1040
1041 return (good_writes >= 1 ? 0 : EIO);
1042 }
1043
1044 /*
1045 * On success, increment the count of good writes for our top-level vdev.
1046 */
1047 static void
1048 vdev_label_sync_done(zio_t *zio)
1049 {
1050 uint64_t *good_writes = zio->io_private;
1051
1052 if (zio->io_error == 0)
1053 atomic_add_64(good_writes, 1);
1054 }
1055
1056 /*
1057 * If there weren't enough good writes, indicate failure to the parent.
1058 */
1059 static void
1060 vdev_label_sync_top_done(zio_t *zio)
1061 {
1062 uint64_t *good_writes = zio->io_private;
1063
1064 if (*good_writes == 0)
1065 zio->io_error = SET_ERROR(EIO);
1066
1067 kmem_free(good_writes, sizeof (uint64_t));
1068 }
1069
1070 /*
1071 * We ignore errors for log and cache devices, simply free the private data.
1072 */
1073 static void
1074 vdev_label_sync_ignore_done(zio_t *zio)
1075 {
1076 kmem_free(zio->io_private, sizeof (uint64_t));
1077 }
1078
1079 /*
1080 * Write all even or odd labels to all leaves of the specified vdev.
1081 */
1082 static void
1083 vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags)
1084 {
1085 nvlist_t *label;
1086 vdev_phys_t *vp;
1087 char *buf;
1088 size_t buflen;
1089
1090 for (int c = 0; c < vd->vdev_children; c++)
1091 vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags);
1092
1093 if (!vd->vdev_ops->vdev_op_leaf)
1094 return;
1095
1096 if (!vdev_writeable(vd))
1097 return;
1098
1099 /*
1100 * Generate a label describing the top-level config to which we belong.
1101 */
1102 label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE);
1103
1104 vp = zio_buf_alloc(sizeof (vdev_phys_t));
1105 bzero(vp, sizeof (vdev_phys_t));
1106
1107 buf = vp->vp_nvlist;
1108 buflen = sizeof (vp->vp_nvlist);
1109
1110 if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) {
1111 for (; l < VDEV_LABELS; l += 2) {
1112 vdev_label_write(zio, vd, l, vp,
1113 offsetof(vdev_label_t, vl_vdev_phys),
1114 sizeof (vdev_phys_t),
1115 vdev_label_sync_done, zio->io_private,
1116 flags | ZIO_FLAG_DONT_PROPAGATE);
1117 }
1118 }
1119
1120 zio_buf_free(vp, sizeof (vdev_phys_t));
1121 nvlist_free(label);
1122 }
1123
1124 int
1125 vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags)
1126 {
1127 list_t *dl = &spa->spa_config_dirty_list;
1128 vdev_t *vd;
1129 zio_t *zio;
1130 int error;
1131
1132 /*
1133 * Write the new labels to disk.
1134 */
1135 zio = zio_root(spa, NULL, NULL, flags);
1136
1137 for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) {
1138 uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t),
1139 KM_SLEEP);
1140
1141 ASSERT(!vd->vdev_ishole);
1142
1143 zio_t *vio = zio_null(zio, spa, NULL,
1144 (vd->vdev_islog || vd->vdev_aux != NULL) ?
1145 vdev_label_sync_ignore_done : vdev_label_sync_top_done,
1146 good_writes, flags);
1147 vdev_label_sync(vio, vd, l, txg, flags);
1148 zio_nowait(vio);
1149 }
1150
1151 error = zio_wait(zio);
1152
1153 /*
1154 * Flush the new labels to disk.
1155 */
1156 zio = zio_root(spa, NULL, NULL, flags);
1157
1158 for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd))
1159 zio_flush(zio, vd);
1160
1161 (void) zio_wait(zio);
1162
1163 return (error);
1164 }
1165
1166 /*
1167 * Sync the uberblock and any changes to the vdev configuration.
1168 *
1169 * The order of operations is carefully crafted to ensure that
1170 * if the system panics or loses power at any time, the state on disk
1171 * is still transactionally consistent. The in-line comments below
1172 * describe the failure semantics at each stage.
1173 *
1174 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1175 * at any time, you can just call it again, and it will resume its work.
1176 */
1177 int
1178 vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg, boolean_t tryhard)
1179 {
1180 spa_t *spa = svd[0]->vdev_spa;
1181 uberblock_t *ub = &spa->spa_uberblock;
1182 vdev_t *vd;
1183 zio_t *zio;
1184 int error;
1185 int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
1186
1187 /*
1188 * Normally, we don't want to try too hard to write every label and
1189 * uberblock. If there is a flaky disk, we don't want the rest of the
1190 * sync process to block while we retry. But if we can't write a
1191 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1192 * bailing out and declaring the pool faulted.
1193 */
1194 if (tryhard)
1195 flags |= ZIO_FLAG_TRYHARD;
1196
1197 ASSERT(ub->ub_txg <= txg);
1198
1199 /*
1200 * If this isn't a resync due to I/O errors,
1201 * and nothing changed in this transaction group,
1202 * and the vdev configuration hasn't changed,
1203 * then there's nothing to do.
1204 */
1205 if (ub->ub_txg < txg &&
1206 uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE &&
1207 list_is_empty(&spa->spa_config_dirty_list))
1208 return (0);
1209
1210 if (txg > spa_freeze_txg(spa))
1211 return (0);
1212
1213 ASSERT(txg <= spa->spa_final_txg);
1214
1215 /*
1216 * Flush the write cache of every disk that's been written to
1217 * in this transaction group. This ensures that all blocks
1218 * written in this txg will be committed to stable storage
1219 * before any uberblock that references them.
1220 */
1221 zio = zio_root(spa, NULL, NULL, flags);
1222
1223 for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd;
1224 vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg)))
1225 zio_flush(zio, vd);
1226
1227 (void) zio_wait(zio);
1228
1229 /*
1230 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1231 * system dies in the middle of this process, that's OK: all of the
1232 * even labels that made it to disk will be newer than any uberblock,
1233 * and will therefore be considered invalid. The odd labels (L1, L3),
1234 * which have not yet been touched, will still be valid. We flush
1235 * the new labels to disk to ensure that all even-label updates
1236 * are committed to stable storage before the uberblock update.
1237 */
1238 if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0)
1239 return (error);
1240
1241 /*
1242 * Sync the uberblocks to all vdevs in svd[].
1243 * If the system dies in the middle of this step, there are two cases
1244 * to consider, and the on-disk state is consistent either way:
1245 *
1246 * (1) If none of the new uberblocks made it to disk, then the
1247 * previous uberblock will be the newest, and the odd labels
1248 * (which had not yet been touched) will be valid with respect
1249 * to that uberblock.
1250 *
1251 * (2) If one or more new uberblocks made it to disk, then they
1252 * will be the newest, and the even labels (which had all
1253 * been successfully committed) will be valid with respect
1254 * to the new uberblocks.
1255 */
1256 if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0)
1257 return (error);
1258
1259 /*
1260 * Sync out odd labels for every dirty vdev. If the system dies
1261 * in the middle of this process, the even labels and the new
1262 * uberblocks will suffice to open the pool. The next time
1263 * the pool is opened, the first thing we'll do -- before any
1264 * user data is modified -- is mark every vdev dirty so that
1265 * all labels will be brought up to date. We flush the new labels
1266 * to disk to ensure that all odd-label updates are committed to
1267 * stable storage before the next transaction group begins.
1268 */
1269 return (vdev_label_sync_list(spa, 1, txg, flags));
1270 }