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) 2012 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         VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 220 
 221         VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE,
 222             vd->vdev_ops->vdev_op_type) == 0);
 223         if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)))
 224                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id)
 225                     == 0);
 226         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid) == 0);
 227 
 228         if (vd->vdev_path != NULL)
 229                 VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PATH,
 230                     vd->vdev_path) == 0);
 231 
 232         if (vd->vdev_devid != NULL)
 233                 VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_DEVID,
 234                     vd->vdev_devid) == 0);
 235 
 236         if (vd->vdev_physpath != NULL)
 237                 VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH,
 238                     vd->vdev_physpath) == 0);
 239 
 240         if (vd->vdev_fru != NULL)
 241                 VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_FRU,
 242                     vd->vdev_fru) == 0);
 243 
 244         if (vd->vdev_nparity != 0) {
 245                 ASSERT(strcmp(vd->vdev_ops->vdev_op_type,
 246                     VDEV_TYPE_RAIDZ) == 0);
 247 
 248                 /*
 249                  * Make sure someone hasn't managed to sneak a fancy new vdev
 250                  * into a crufty old storage pool.
 251                  */
 252                 ASSERT(vd->vdev_nparity == 1 ||
 253                     (vd->vdev_nparity <= 2 &&
 254                     spa_version(spa) >= SPA_VERSION_RAIDZ2) ||
 255                     (vd->vdev_nparity <= 3 &&
 256                     spa_version(spa) >= SPA_VERSION_RAIDZ3));
 257 
 258                 /*
 259                  * Note that we'll add the nparity tag even on storage pools
 260                  * that only support a single parity device -- older software
 261                  * will just ignore it.
 262                  */
 263                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY,
 264                     vd->vdev_nparity) == 0);
 265         }
 266 
 267         if (vd->vdev_wholedisk != -1ULL)
 268                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
 269                     vd->vdev_wholedisk) == 0);
 270 
 271         if (vd->vdev_not_present)
 272                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1) == 0);
 273 
 274         if (vd->vdev_isspare)
 275                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1) == 0);
 276 
 277         if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) &&
 278             vd == vd->vdev_top) {
 279                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
 280                     vd->vdev_ms_array) == 0);
 281                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
 282                     vd->vdev_ms_shift) == 0);
 283                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT,
 284                     vd->vdev_ashift) == 0);
 285                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE,
 286                     vd->vdev_asize) == 0);
 287                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG,
 288                     vd->vdev_islog) == 0);
 289                 if (vd->vdev_removing)
 290                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING,
 291                             vd->vdev_removing) == 0);
 292         }
 293 
 294         if (vd->vdev_dtl_smo.smo_object != 0)
 295                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DTL,
 296                     vd->vdev_dtl_smo.smo_object) == 0);
 297 
 298         if (vd->vdev_crtxg)
 299                 VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG,
 300                     vd->vdev_crtxg) == 0);
 301 
 302         if (getstats) {
 303                 vdev_stat_t vs;
 304                 pool_scan_stat_t ps;
 305 
 306                 vdev_get_stats(vd, &vs);
 307                 VERIFY(nvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
 308                     (uint64_t *)&vs, sizeof (vs) / sizeof (uint64_t)) == 0);
 309 
 310                 /* provide either current or previous scan information */
 311                 if (spa_scan_get_stats(spa, &ps) == 0) {
 312                         VERIFY(nvlist_add_uint64_array(nv,
 313                             ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps,
 314                             sizeof (pool_scan_stat_t) / sizeof (uint64_t))
 315                             == 0);
 316                 }
 317         }
 318 
 319         if (!vd->vdev_ops->vdev_op_leaf) {
 320                 nvlist_t **child;
 321                 int c, idx;
 322 
 323                 ASSERT(!vd->vdev_ishole);
 324 
 325                 child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *),
 326                     KM_SLEEP);
 327 
 328                 for (c = 0, idx = 0; c < vd->vdev_children; c++) {
 329                         vdev_t *cvd = vd->vdev_child[c];
 330 
 331                         /*
 332                          * If we're generating an nvlist of removing
 333                          * vdevs then skip over any device which is
 334                          * not being removed.
 335                          */
 336                         if ((flags & VDEV_CONFIG_REMOVING) &&
 337                             !cvd->vdev_removing)
 338                                 continue;
 339 
 340                         child[idx++] = vdev_config_generate(spa, cvd,
 341                             getstats, flags);
 342                 }
 343 
 344                 if (idx) {
 345                         VERIFY(nvlist_add_nvlist_array(nv,
 346                             ZPOOL_CONFIG_CHILDREN, child, idx) == 0);
 347                 }
 348 
 349                 for (c = 0; c < idx; c++)
 350                         nvlist_free(child[c]);
 351 
 352                 kmem_free(child, vd->vdev_children * sizeof (nvlist_t *));
 353 
 354         } else {
 355                 const char *aux = NULL;
 356 
 357                 if (vd->vdev_offline && !vd->vdev_tmpoffline)
 358                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE,
 359                             B_TRUE) == 0);
 360                 if (vd->vdev_resilvering)
 361                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVERING,
 362                             B_TRUE) == 0);
 363                 if (vd->vdev_faulted)
 364                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED,
 365                             B_TRUE) == 0);
 366                 if (vd->vdev_degraded)
 367                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED,
 368                             B_TRUE) == 0);
 369                 if (vd->vdev_removed)
 370                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED,
 371                             B_TRUE) == 0);
 372                 if (vd->vdev_unspare)
 373                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE,
 374                             B_TRUE) == 0);
 375                 if (vd->vdev_ishole)
 376                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE,
 377                             B_TRUE) == 0);
 378 
 379                 switch (vd->vdev_stat.vs_aux) {
 380                 case VDEV_AUX_ERR_EXCEEDED:
 381                         aux = "err_exceeded";
 382                         break;
 383 
 384                 case VDEV_AUX_EXTERNAL:
 385                         aux = "external";
 386                         break;
 387                 }
 388 
 389                 if (aux != NULL)
 390                         VERIFY(nvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE,
 391                             aux) == 0);
 392 
 393                 if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) {
 394                         VERIFY(nvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID,
 395                             vd->vdev_orig_guid) == 0);
 396                 }
 397         }
 398 
 399         return (nv);
 400 }
 401 
 402 /*
 403  * Generate a view of the top-level vdevs.  If we currently have holes
 404  * in the namespace, then generate an array which contains a list of holey
 405  * vdevs.  Additionally, add the number of top-level children that currently
 406  * exist.
 407  */
 408 void
 409 vdev_top_config_generate(spa_t *spa, nvlist_t *config)
 410 {
 411         vdev_t *rvd = spa->spa_root_vdev;
 412         uint64_t *array;
 413         uint_t c, idx;
 414 
 415         array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP);
 416 
 417         for (c = 0, idx = 0; c < rvd->vdev_children; c++) {
 418                 vdev_t *tvd = rvd->vdev_child[c];
 419 
 420                 if (tvd->vdev_ishole)
 421                         array[idx++] = c;
 422         }
 423 
 424         if (idx) {
 425                 VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY,
 426                     array, idx) == 0);
 427         }
 428 
 429         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN,
 430             rvd->vdev_children) == 0);
 431 
 432         kmem_free(array, rvd->vdev_children * sizeof (uint64_t));
 433 }
 434 
 435 /*
 436  * Returns the configuration from the label of the given vdev. If 'label' is
 437  * VDEV_BEST_LABEL, each label of the vdev will be read until a valid
 438  * configuration is found; otherwise, only the specified label will be read.
 439  */
 440 nvlist_t *
 441 vdev_label_read_config(vdev_t *vd, int label)
 442 {
 443         spa_t *spa = vd->vdev_spa;
 444         nvlist_t *config = NULL;
 445         vdev_phys_t *vp;
 446         zio_t *zio;
 447         int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
 448             ZIO_FLAG_SPECULATIVE;
 449 
 450         ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
 451 
 452         if (!vdev_readable(vd))
 453                 return (NULL);
 454 
 455         vp = zio_buf_alloc(sizeof (vdev_phys_t));
 456 
 457 retry:
 458         for (int l = 0; l < VDEV_LABELS; l++) {
 459                 if (label >= 0 && label < VDEV_LABELS && label != l)
 460                         continue;
 461 
 462                 zio = zio_root(spa, NULL, NULL, flags);
 463 
 464                 vdev_label_read(zio, vd, l, vp,
 465                     offsetof(vdev_label_t, vl_vdev_phys),
 466                     sizeof (vdev_phys_t), NULL, NULL, flags);
 467 
 468                 if (zio_wait(zio) == 0 &&
 469                     nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist),
 470                     &config, 0) == 0)
 471                         break;
 472 
 473                 if (config != NULL) {
 474                         nvlist_free(config);
 475                         config = NULL;
 476                 }
 477         }
 478 
 479         if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) {
 480                 flags |= ZIO_FLAG_TRYHARD;
 481                 goto retry;
 482         }
 483 
 484         zio_buf_free(vp, sizeof (vdev_phys_t));
 485 
 486         return (config);
 487 }
 488 
 489 /*
 490  * Determine if a device is in use.  The 'spare_guid' parameter will be filled
 491  * in with the device guid if this spare is active elsewhere on the system.
 492  */
 493 static boolean_t
 494 vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason,
 495     uint64_t *spare_guid, uint64_t *l2cache_guid)
 496 {
 497         spa_t *spa = vd->vdev_spa;
 498         uint64_t state, pool_guid, device_guid, txg, spare_pool;
 499         uint64_t vdtxg = 0;
 500         nvlist_t *label;
 501 
 502         if (spare_guid)
 503                 *spare_guid = 0ULL;
 504         if (l2cache_guid)
 505                 *l2cache_guid = 0ULL;
 506 
 507         /*
 508          * Read the label, if any, and perform some basic sanity checks.
 509          */
 510         if ((label = vdev_label_read_config(vd, VDEV_BEST_LABEL)) == NULL)
 511                 return (B_FALSE);
 512 
 513         (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
 514             &vdtxg);
 515 
 516         if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
 517             &state) != 0 ||
 518             nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID,
 519             &device_guid) != 0) {
 520                 nvlist_free(label);
 521                 return (B_FALSE);
 522         }
 523 
 524         if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
 525             (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID,
 526             &pool_guid) != 0 ||
 527             nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
 528             &txg) != 0)) {
 529                 nvlist_free(label);
 530                 return (B_FALSE);
 531         }
 532 
 533         nvlist_free(label);
 534 
 535         /*
 536          * Check to see if this device indeed belongs to the pool it claims to
 537          * be a part of.  The only way this is allowed is if the device is a hot
 538          * spare (which we check for later on).
 539          */
 540         if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
 541             !spa_guid_exists(pool_guid, device_guid) &&
 542             !spa_spare_exists(device_guid, NULL, NULL) &&
 543             !spa_l2cache_exists(device_guid, NULL))
 544                 return (B_FALSE);
 545 
 546         /*
 547          * If the transaction group is zero, then this an initialized (but
 548          * unused) label.  This is only an error if the create transaction
 549          * on-disk is the same as the one we're using now, in which case the
 550          * user has attempted to add the same vdev multiple times in the same
 551          * transaction.
 552          */
 553         if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
 554             txg == 0 && vdtxg == crtxg)
 555                 return (B_TRUE);
 556 
 557         /*
 558          * Check to see if this is a spare device.  We do an explicit check for
 559          * spa_has_spare() here because it may be on our pending list of spares
 560          * to add.  We also check if it is an l2cache device.
 561          */
 562         if (spa_spare_exists(device_guid, &spare_pool, NULL) ||
 563             spa_has_spare(spa, device_guid)) {
 564                 if (spare_guid)
 565                         *spare_guid = device_guid;
 566 
 567                 switch (reason) {
 568                 case VDEV_LABEL_CREATE:
 569                 case VDEV_LABEL_L2CACHE:
 570                         return (B_TRUE);
 571 
 572                 case VDEV_LABEL_REPLACE:
 573                         return (!spa_has_spare(spa, device_guid) ||
 574                             spare_pool != 0ULL);
 575 
 576                 case VDEV_LABEL_SPARE:
 577                         return (spa_has_spare(spa, device_guid));
 578                 }
 579         }
 580 
 581         /*
 582          * Check to see if this is an l2cache device.
 583          */
 584         if (spa_l2cache_exists(device_guid, NULL))
 585                 return (B_TRUE);
 586 
 587         /*
 588          * We can't rely on a pool's state if it's been imported
 589          * read-only.  Instead we look to see if the pools is marked
 590          * read-only in the namespace and set the state to active.
 591          */
 592         if ((spa = spa_by_guid(pool_guid, device_guid)) != NULL &&
 593             spa_mode(spa) == FREAD)
 594                 state = POOL_STATE_ACTIVE;
 595 
 596         /*
 597          * If the device is marked ACTIVE, then this device is in use by another
 598          * pool on the system.
 599          */
 600         return (state == POOL_STATE_ACTIVE);
 601 }
 602 
 603 /*
 604  * Initialize a vdev label.  We check to make sure each leaf device is not in
 605  * use, and writable.  We put down an initial label which we will later
 606  * overwrite with a complete label.  Note that it's important to do this
 607  * sequentially, not in parallel, so that we catch cases of multiple use of the
 608  * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
 609  * itself.
 610  */
 611 int
 612 vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason)
 613 {
 614         spa_t *spa = vd->vdev_spa;
 615         nvlist_t *label;
 616         vdev_phys_t *vp;
 617         char *pad2;
 618         uberblock_t *ub;
 619         zio_t *zio;
 620         char *buf;
 621         size_t buflen;
 622         int error;
 623         uint64_t spare_guid, l2cache_guid;
 624         int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
 625 
 626         ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
 627 
 628         for (int c = 0; c < vd->vdev_children; c++)
 629                 if ((error = vdev_label_init(vd->vdev_child[c],
 630                     crtxg, reason)) != 0)
 631                         return (error);
 632 
 633         /* Track the creation time for this vdev */
 634         vd->vdev_crtxg = crtxg;
 635 
 636         if (!vd->vdev_ops->vdev_op_leaf)
 637                 return (0);
 638 
 639         /*
 640          * Dead vdevs cannot be initialized.
 641          */
 642         if (vdev_is_dead(vd))
 643                 return (EIO);
 644 
 645         /*
 646          * Determine if the vdev is in use.
 647          */
 648         if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT &&
 649             vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid))
 650                 return (EBUSY);
 651 
 652         /*
 653          * If this is a request to add or replace a spare or l2cache device
 654          * that is in use elsewhere on the system, then we must update the
 655          * guid (which was initialized to a random value) to reflect the
 656          * actual GUID (which is shared between multiple pools).
 657          */
 658         if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE &&
 659             spare_guid != 0ULL) {
 660                 uint64_t guid_delta = spare_guid - vd->vdev_guid;
 661 
 662                 vd->vdev_guid += guid_delta;
 663 
 664                 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
 665                         pvd->vdev_guid_sum += guid_delta;
 666 
 667                 /*
 668                  * If this is a replacement, then we want to fallthrough to the
 669                  * rest of the code.  If we're adding a spare, then it's already
 670                  * labeled appropriately and we can just return.
 671                  */
 672                 if (reason == VDEV_LABEL_SPARE)
 673                         return (0);
 674                 ASSERT(reason == VDEV_LABEL_REPLACE ||
 675                     reason == VDEV_LABEL_SPLIT);
 676         }
 677 
 678         if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE &&
 679             l2cache_guid != 0ULL) {
 680                 uint64_t guid_delta = l2cache_guid - vd->vdev_guid;
 681 
 682                 vd->vdev_guid += guid_delta;
 683 
 684                 for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent)
 685                         pvd->vdev_guid_sum += guid_delta;
 686 
 687                 /*
 688                  * If this is a replacement, then we want to fallthrough to the
 689                  * rest of the code.  If we're adding an l2cache, then it's
 690                  * already labeled appropriately and we can just return.
 691                  */
 692                 if (reason == VDEV_LABEL_L2CACHE)
 693                         return (0);
 694                 ASSERT(reason == VDEV_LABEL_REPLACE);
 695         }
 696 
 697         /*
 698          * Initialize its label.
 699          */
 700         vp = zio_buf_alloc(sizeof (vdev_phys_t));
 701         bzero(vp, sizeof (vdev_phys_t));
 702 
 703         /*
 704          * Generate a label describing the pool and our top-level vdev.
 705          * We mark it as being from txg 0 to indicate that it's not
 706          * really part of an active pool just yet.  The labels will
 707          * be written again with a meaningful txg by spa_sync().
 708          */
 709         if (reason == VDEV_LABEL_SPARE ||
 710             (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) {
 711                 /*
 712                  * For inactive hot spares, we generate a special label that
 713                  * identifies as a mutually shared hot spare.  We write the
 714                  * label if we are adding a hot spare, or if we are removing an
 715                  * active hot spare (in which case we want to revert the
 716                  * labels).
 717                  */
 718                 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 719 
 720                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
 721                     spa_version(spa)) == 0);
 722                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
 723                     POOL_STATE_SPARE) == 0);
 724                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
 725                     vd->vdev_guid) == 0);
 726         } else if (reason == VDEV_LABEL_L2CACHE ||
 727             (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) {
 728                 /*
 729                  * For level 2 ARC devices, add a special label.
 730                  */
 731                 VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 732 
 733                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION,
 734                     spa_version(spa)) == 0);
 735                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE,
 736                     POOL_STATE_L2CACHE) == 0);
 737                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID,
 738                     vd->vdev_guid) == 0);
 739         } else {
 740                 uint64_t txg = 0ULL;
 741 
 742                 if (reason == VDEV_LABEL_SPLIT)
 743                         txg = spa->spa_uberblock.ub_txg;
 744                 label = spa_config_generate(spa, vd, txg, B_FALSE);
 745 
 746                 /*
 747                  * Add our creation time.  This allows us to detect multiple
 748                  * vdev uses as described above, and automatically expires if we
 749                  * fail.
 750                  */
 751                 VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG,
 752                     crtxg) == 0);
 753         }
 754 
 755         buf = vp->vp_nvlist;
 756         buflen = sizeof (vp->vp_nvlist);
 757 
 758         error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP);
 759         if (error != 0) {
 760                 nvlist_free(label);
 761                 zio_buf_free(vp, sizeof (vdev_phys_t));
 762                 /* EFAULT means nvlist_pack ran out of room */
 763                 return (error == EFAULT ? ENAMETOOLONG : EINVAL);
 764         }
 765 
 766         /*
 767          * Initialize uberblock template.
 768          */
 769         ub = zio_buf_alloc(VDEV_UBERBLOCK_RING);
 770         bzero(ub, VDEV_UBERBLOCK_RING);
 771         *ub = spa->spa_uberblock;
 772         ub->ub_txg = 0;
 773 
 774         /* Initialize the 2nd padding area. */
 775         pad2 = zio_buf_alloc(VDEV_PAD_SIZE);
 776         bzero(pad2, VDEV_PAD_SIZE);
 777 
 778         /*
 779          * Write everything in parallel.
 780          */
 781 retry:
 782         zio = zio_root(spa, NULL, NULL, flags);
 783 
 784         for (int l = 0; l < VDEV_LABELS; l++) {
 785 
 786                 vdev_label_write(zio, vd, l, vp,
 787                     offsetof(vdev_label_t, vl_vdev_phys),
 788                     sizeof (vdev_phys_t), NULL, NULL, flags);
 789 
 790                 /*
 791                  * Skip the 1st padding area.
 792                  * Zero out the 2nd padding area where it might have
 793                  * left over data from previous filesystem format.
 794                  */
 795                 vdev_label_write(zio, vd, l, pad2,
 796                     offsetof(vdev_label_t, vl_pad2),
 797                     VDEV_PAD_SIZE, NULL, NULL, flags);
 798 
 799                 vdev_label_write(zio, vd, l, ub,
 800                     offsetof(vdev_label_t, vl_uberblock),
 801                     VDEV_UBERBLOCK_RING, NULL, NULL, flags);
 802         }
 803 
 804         error = zio_wait(zio);
 805 
 806         if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) {
 807                 flags |= ZIO_FLAG_TRYHARD;
 808                 goto retry;
 809         }
 810 
 811         nvlist_free(label);
 812         zio_buf_free(pad2, VDEV_PAD_SIZE);
 813         zio_buf_free(ub, VDEV_UBERBLOCK_RING);
 814         zio_buf_free(vp, sizeof (vdev_phys_t));
 815 
 816         /*
 817          * If this vdev hasn't been previously identified as a spare, then we
 818          * mark it as such only if a) we are labeling it as a spare, or b) it
 819          * exists as a spare elsewhere in the system.  Do the same for
 820          * level 2 ARC devices.
 821          */
 822         if (error == 0 && !vd->vdev_isspare &&
 823             (reason == VDEV_LABEL_SPARE ||
 824             spa_spare_exists(vd->vdev_guid, NULL, NULL)))
 825                 spa_spare_add(vd);
 826 
 827         if (error == 0 && !vd->vdev_isl2cache &&
 828             (reason == VDEV_LABEL_L2CACHE ||
 829             spa_l2cache_exists(vd->vdev_guid, NULL)))
 830                 spa_l2cache_add(vd);
 831 
 832         return (error);
 833 }
 834 
 835 /*
 836  * ==========================================================================
 837  * uberblock load/sync
 838  * ==========================================================================
 839  */
 840 
 841 /*
 842  * Consider the following situation: txg is safely synced to disk.  We've
 843  * written the first uberblock for txg + 1, and then we lose power.  When we
 844  * come back up, we fail to see the uberblock for txg + 1 because, say,
 845  * it was on a mirrored device and the replica to which we wrote txg + 1
 846  * is now offline.  If we then make some changes and sync txg + 1, and then
 847  * the missing replica comes back, then for a few seconds we'll have two
 848  * conflicting uberblocks on disk with the same txg.  The solution is simple:
 849  * among uberblocks with equal txg, choose the one with the latest timestamp.
 850  */
 851 static int
 852 vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2)
 853 {
 854         if (ub1->ub_txg < ub2->ub_txg)
 855                 return (-1);
 856         if (ub1->ub_txg > ub2->ub_txg)
 857                 return (1);
 858 
 859         if (ub1->ub_timestamp < ub2->ub_timestamp)
 860                 return (-1);
 861         if (ub1->ub_timestamp > ub2->ub_timestamp)
 862                 return (1);
 863 
 864         return (0);
 865 }
 866 
 867 struct ubl_cbdata {
 868         uberblock_t     *ubl_ubbest;    /* Best uberblock */
 869         vdev_t          *ubl_vd;        /* vdev associated with the above */
 870         int             ubl_label;      /* Label associated with the above */
 871 };
 872 
 873 static void
 874 vdev_uberblock_load_done(zio_t *zio)
 875 {
 876         vdev_t *vd = zio->io_vd;
 877         spa_t *spa = zio->io_spa;
 878         zio_t *rio = zio->io_private;
 879         uberblock_t *ub = zio->io_data;
 880         struct ubl_cbdata *cbp = rio->io_private;
 881 
 882         ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd));
 883 
 884         if (zio->io_error == 0 && uberblock_verify(ub) == 0) {
 885                 mutex_enter(&rio->io_lock);
 886                 if (ub->ub_txg <= spa->spa_load_max_txg &&
 887                     vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) {
 888                         /*
 889                          * Keep track of the vdev and label in which this
 890                          * uberblock was found. We will use this information
 891                          * later to obtain the config nvlist associated with
 892                          * this uberblock.
 893                          */
 894                         *cbp->ubl_ubbest = *ub;
 895                         cbp->ubl_vd = vd;
 896                         cbp->ubl_label = vdev_label_number(vd->vdev_psize,
 897                             zio->io_offset);
 898                 }
 899                 mutex_exit(&rio->io_lock);
 900         }
 901 
 902         zio_buf_free(zio->io_data, zio->io_size);
 903 }
 904 
 905 static void
 906 vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags,
 907     struct ubl_cbdata *cbp)
 908 {
 909         for (int c = 0; c < vd->vdev_children; c++)
 910                 vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp);
 911 
 912         if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) {
 913                 for (int l = 0; l < VDEV_LABELS; l++) {
 914                         for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) {
 915                                 vdev_label_read(zio, vd, l,
 916                                     zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)),
 917                                     VDEV_UBERBLOCK_OFFSET(vd, n),
 918                                     VDEV_UBERBLOCK_SIZE(vd),
 919                                     vdev_uberblock_load_done, zio, flags);
 920                         }
 921                 }
 922         }
 923 }
 924 
 925 /*
 926  * Reads the 'best' uberblock from disk along with its associated
 927  * configuration. First, we read the uberblock array of each label of each
 928  * vdev, keeping track of the uberblock with the highest txg in each array.
 929  * Then, we read the configuration from the same label as the best uberblock.
 930  */
 931 void
 932 vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config)
 933 {
 934         int i;
 935         zio_t *zio;
 936         spa_t *spa = rvd->vdev_spa;
 937         struct ubl_cbdata cb;
 938         int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
 939             ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD;
 940 
 941         ASSERT(ub);
 942         ASSERT(config);
 943 
 944         bzero(ub, sizeof (uberblock_t));
 945         *config = NULL;
 946 
 947         cb.ubl_ubbest = ub;
 948         cb.ubl_vd = NULL;
 949 
 950         spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
 951         zio = zio_root(spa, NULL, &cb, flags);
 952         vdev_uberblock_load_impl(zio, rvd, flags, &cb);
 953         (void) zio_wait(zio);
 954         if (cb.ubl_vd != NULL) {
 955                 for (i = cb.ubl_label % 2; i < VDEV_LABELS; i += 2) {
 956                         *config = vdev_label_read_config(cb.ubl_vd, i);
 957                         if (*config != NULL)
 958                                 break;
 959                 }
 960         }
 961         spa_config_exit(spa, SCL_ALL, FTAG);
 962 }
 963 
 964 /*
 965  * On success, increment root zio's count of good writes.
 966  * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
 967  */
 968 static void
 969 vdev_uberblock_sync_done(zio_t *zio)
 970 {
 971         uint64_t *good_writes = zio->io_private;
 972 
 973         if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0)
 974                 atomic_add_64(good_writes, 1);
 975 }
 976 
 977 /*
 978  * Write the uberblock to all labels of all leaves of the specified vdev.
 979  */
 980 static void
 981 vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags)
 982 {
 983         uberblock_t *ubbuf;
 984         int n;
 985 
 986         for (int c = 0; c < vd->vdev_children; c++)
 987                 vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags);
 988 
 989         if (!vd->vdev_ops->vdev_op_leaf)
 990                 return;
 991 
 992         if (!vdev_writeable(vd))
 993                 return;
 994 
 995         n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1);
 996 
 997         ubbuf = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd));
 998         bzero(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
 999         *ubbuf = *ub;
1000 
1001         for (int l = 0; l < VDEV_LABELS; l++)
1002                 vdev_label_write(zio, vd, l, ubbuf,
1003                     VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd),
1004                     vdev_uberblock_sync_done, zio->io_private,
1005                     flags | ZIO_FLAG_DONT_PROPAGATE);
1006 
1007         zio_buf_free(ubbuf, VDEV_UBERBLOCK_SIZE(vd));
1008 }
1009 
1010 int
1011 vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags)
1012 {
1013         spa_t *spa = svd[0]->vdev_spa;
1014         zio_t *zio;
1015         uint64_t good_writes = 0;
1016 
1017         zio = zio_root(spa, NULL, &good_writes, flags);
1018 
1019         for (int v = 0; v < svdcount; v++)
1020                 vdev_uberblock_sync(zio, ub, svd[v], flags);
1021 
1022         (void) zio_wait(zio);
1023 
1024         /*
1025          * Flush the uberblocks to disk.  This ensures that the odd labels
1026          * are no longer needed (because the new uberblocks and the even
1027          * labels are safely on disk), so it is safe to overwrite them.
1028          */
1029         zio = zio_root(spa, NULL, NULL, flags);
1030 
1031         for (int v = 0; v < svdcount; v++)
1032                 zio_flush(zio, svd[v]);
1033 
1034         (void) zio_wait(zio);
1035 
1036         return (good_writes >= 1 ? 0 : EIO);
1037 }
1038 
1039 /*
1040  * On success, increment the count of good writes for our top-level vdev.
1041  */
1042 static void
1043 vdev_label_sync_done(zio_t *zio)
1044 {
1045         uint64_t *good_writes = zio->io_private;
1046 
1047         if (zio->io_error == 0)
1048                 atomic_add_64(good_writes, 1);
1049 }
1050 
1051 /*
1052  * If there weren't enough good writes, indicate failure to the parent.
1053  */
1054 static void
1055 vdev_label_sync_top_done(zio_t *zio)
1056 {
1057         uint64_t *good_writes = zio->io_private;
1058 
1059         if (*good_writes == 0)
1060                 zio->io_error = EIO;
1061 
1062         kmem_free(good_writes, sizeof (uint64_t));
1063 }
1064 
1065 /*
1066  * We ignore errors for log and cache devices, simply free the private data.
1067  */
1068 static void
1069 vdev_label_sync_ignore_done(zio_t *zio)
1070 {
1071         kmem_free(zio->io_private, sizeof (uint64_t));
1072 }
1073 
1074 /*
1075  * Write all even or odd labels to all leaves of the specified vdev.
1076  */
1077 static void
1078 vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags)
1079 {
1080         nvlist_t *label;
1081         vdev_phys_t *vp;
1082         char *buf;
1083         size_t buflen;
1084 
1085         for (int c = 0; c < vd->vdev_children; c++)
1086                 vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags);
1087 
1088         if (!vd->vdev_ops->vdev_op_leaf)
1089                 return;
1090 
1091         if (!vdev_writeable(vd))
1092                 return;
1093 
1094         /*
1095          * Generate a label describing the top-level config to which we belong.
1096          */
1097         label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE);
1098 
1099         vp = zio_buf_alloc(sizeof (vdev_phys_t));
1100         bzero(vp, sizeof (vdev_phys_t));
1101 
1102         buf = vp->vp_nvlist;
1103         buflen = sizeof (vp->vp_nvlist);
1104 
1105         if (nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP) == 0) {
1106                 for (; l < VDEV_LABELS; l += 2) {
1107                         vdev_label_write(zio, vd, l, vp,
1108                             offsetof(vdev_label_t, vl_vdev_phys),
1109                             sizeof (vdev_phys_t),
1110                             vdev_label_sync_done, zio->io_private,
1111                             flags | ZIO_FLAG_DONT_PROPAGATE);
1112                 }
1113         }
1114 
1115         zio_buf_free(vp, sizeof (vdev_phys_t));
1116         nvlist_free(label);
1117 }
1118 
1119 int
1120 vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags)
1121 {
1122         list_t *dl = &spa->spa_config_dirty_list;
1123         vdev_t *vd;
1124         zio_t *zio;
1125         int error;
1126 
1127         /*
1128          * Write the new labels to disk.
1129          */
1130         zio = zio_root(spa, NULL, NULL, flags);
1131 
1132         for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) {
1133                 uint64_t *good_writes = kmem_zalloc(sizeof (uint64_t),
1134                     KM_SLEEP);
1135 
1136                 ASSERT(!vd->vdev_ishole);
1137 
1138                 zio_t *vio = zio_null(zio, spa, NULL,
1139                     (vd->vdev_islog || vd->vdev_aux != NULL) ?
1140                     vdev_label_sync_ignore_done : vdev_label_sync_top_done,
1141                     good_writes, flags);
1142                 vdev_label_sync(vio, vd, l, txg, flags);
1143                 zio_nowait(vio);
1144         }
1145 
1146         error = zio_wait(zio);
1147 
1148         /*
1149          * Flush the new labels to disk.
1150          */
1151         zio = zio_root(spa, NULL, NULL, flags);
1152 
1153         for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd))
1154                 zio_flush(zio, vd);
1155 
1156         (void) zio_wait(zio);
1157 
1158         return (error);
1159 }
1160 
1161 /*
1162  * Sync the uberblock and any changes to the vdev configuration.
1163  *
1164  * The order of operations is carefully crafted to ensure that
1165  * if the system panics or loses power at any time, the state on disk
1166  * is still transactionally consistent.  The in-line comments below
1167  * describe the failure semantics at each stage.
1168  *
1169  * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1170  * at any time, you can just call it again, and it will resume its work.
1171  */
1172 int
1173 vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg, boolean_t tryhard)
1174 {
1175         spa_t *spa = svd[0]->vdev_spa;
1176         uberblock_t *ub = &spa->spa_uberblock;
1177         vdev_t *vd;
1178         zio_t *zio;
1179         int error;
1180         int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
1181 
1182         /*
1183          * Normally, we don't want to try too hard to write every label and
1184          * uberblock.  If there is a flaky disk, we don't want the rest of the
1185          * sync process to block while we retry.  But if we can't write a
1186          * single label out, we should retry with ZIO_FLAG_TRYHARD before
1187          * bailing out and declaring the pool faulted.
1188          */
1189         if (tryhard)
1190                 flags |= ZIO_FLAG_TRYHARD;
1191 
1192         ASSERT(ub->ub_txg <= txg);
1193 
1194         /*
1195          * If this isn't a resync due to I/O errors,
1196          * and nothing changed in this transaction group,
1197          * and the vdev configuration hasn't changed,
1198          * then there's nothing to do.
1199          */
1200         if (ub->ub_txg < txg &&
1201             uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE &&
1202             list_is_empty(&spa->spa_config_dirty_list))
1203                 return (0);
1204 
1205         if (txg > spa_freeze_txg(spa))
1206                 return (0);
1207 
1208         ASSERT(txg <= spa->spa_final_txg);
1209 
1210         /*
1211          * Flush the write cache of every disk that's been written to
1212          * in this transaction group.  This ensures that all blocks
1213          * written in this txg will be committed to stable storage
1214          * before any uberblock that references them.
1215          */
1216         zio = zio_root(spa, NULL, NULL, flags);
1217 
1218         for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd;
1219             vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg)))
1220                 zio_flush(zio, vd);
1221 
1222         (void) zio_wait(zio);
1223 
1224         /*
1225          * Sync out the even labels (L0, L2) for every dirty vdev.  If the
1226          * system dies in the middle of this process, that's OK: all of the
1227          * even labels that made it to disk will be newer than any uberblock,
1228          * and will therefore be considered invalid.  The odd labels (L1, L3),
1229          * which have not yet been touched, will still be valid.  We flush
1230          * the new labels to disk to ensure that all even-label updates
1231          * are committed to stable storage before the uberblock update.
1232          */
1233         if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0)
1234                 return (error);
1235 
1236         /*
1237          * Sync the uberblocks to all vdevs in svd[].
1238          * If the system dies in the middle of this step, there are two cases
1239          * to consider, and the on-disk state is consistent either way:
1240          *
1241          * (1)  If none of the new uberblocks made it to disk, then the
1242          *      previous uberblock will be the newest, and the odd labels
1243          *      (which had not yet been touched) will be valid with respect
1244          *      to that uberblock.
1245          *
1246          * (2)  If one or more new uberblocks made it to disk, then they
1247          *      will be the newest, and the even labels (which had all
1248          *      been successfully committed) will be valid with respect
1249          *      to the new uberblocks.
1250          */
1251         if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0)
1252                 return (error);
1253 
1254         /*
1255          * Sync out odd labels for every dirty vdev.  If the system dies
1256          * in the middle of this process, the even labels and the new
1257          * uberblocks will suffice to open the pool.  The next time
1258          * the pool is opened, the first thing we'll do -- before any
1259          * user data is modified -- is mark every vdev dirty so that
1260          * all labels will be brought up to date.  We flush the new labels
1261          * to disk to ensure that all odd-label updates are committed to
1262          * stable storage before the next transaction group begins.
1263          */
1264         return (vdev_label_sync_list(spa, 1, txg, flags));
1265 }