external/cddl/osnet/dist/uts/common/fs/zfs/spa_config.c

   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 2008 Sun Microsystems, Inc.  All rights reserved.
  24  * Use is subject to license terms.
  25  */
  26
  27 #include <sys/spa.h>
  28 #include <sys/spa_impl.h>
  29 #include <sys/nvpair.h>
  30 #include <sys/uio.h>
  31 #include <sys/fs/zfs.h>
  32 #include <sys/vdev_impl.h>
  33 #include <sys/zfs_ioctl.h>
  34 #include <sys/utsname.h>
  35 #include <sys/systeminfo.h>
  36 #include <sys/sunddi.h>
  37 #ifdef _KERNEL
  38 #include <sys/kobj.h>
  39 #endif
  40
  41 /*
  42  * Pool configuration repository.
  43  *
  44  * Pool configuration is stored as a packed nvlist on the filesystem.  By
  45  * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
  46  * (when the ZFS module is loaded).  Pools can also have the 'cachefile'
  47  * property set that allows them to be stored in an alternate location until
  48  * the control of external software.
  49  *
  50  * For each cache file, we have a single nvlist which holds all the
  51  * configuration information.  When the module loads, we read this information
  52  * from /etc/zfs/zpool.cache and populate the SPA namespace.  This namespace is
  53  * maintained independently in spa.c.  Whenever the namespace is modified, or
  54  * the configuration of a pool is changed, we call spa_config_sync(), which
  55  * walks through all the active pools and writes the configuration to disk.
  56  */
  57
  58 static uint64_t spa_config_generation = 1;
  59
  60 /*
  61  * This can be overridden in userland to preserve an alternate namespace for
  62  * userland pools when doing testing.
  63  */
  64 const char *spa_config_path = ZPOOL_CACHE;
  65
  66 /*
  67  * Called when the module is first loaded, this routine loads the configuration
  68  * file into the SPA namespace.  It does not actually open or load the pools; it
  69  * only populates the namespace.
  70  */
  71 void
  72 spa_config_load(void)
  73 {
  74         void *buf = NULL;
  75         nvlist_t *nvlist, *child;
  76         nvpair_t *nvpair;
  77         spa_t *spa;
  78         char *pathname;
  79         struct _buf *file;
  80         uint64_t fsize;
  81
  82         /*
  83          * Open the configuration file.
  84          */
  85         pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
  86         (void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path);
  87
  88         file = kobj_open_file(pathname);
  89
  90         kmem_free(pathname, MAXPATHLEN);
  91
  92         if (file == (struct _buf *)-1)
  93                 return;
  94
  95         if (kobj_get_filesize(file, &fsize) != 0)
  96                 goto out;
  97
  98         buf = kmem_alloc(fsize, KM_SLEEP);
  99
 100         /*
 101          * Read the nvlist from the file.
 102          */
 103         if (kobj_read_file(file, buf, fsize, 0) < 0)
 104                 goto out;
 105
 106         /*
 107          * Unpack the nvlist.
 108          */
 109         if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
 110                 goto out;
 111
 112         /*
 113          * Iterate over all elements in the nvlist, creating a new spa_t for
 114          * each one with the specified configuration.
 115          */
 116         mutex_enter(&spa_namespace_lock);
 117         nvpair = NULL;
 118         while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
 119
 120                 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
 121                         continue;
 122
 123                 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
 124
 125                 if (spa_lookup(nvpair_name(nvpair)) != NULL)
 126                         continue;
 127                 spa = spa_add(nvpair_name(nvpair), NULL);
 128
 129                 /*
 130                  * We blindly duplicate the configuration here.  If it's
 131                  * invalid, we will catch it when the pool is first opened.
 132                  */
 133                 VERIFY(nvlist_dup(child, &spa->spa_config, 0) == 0);
 134         }
 135         mutex_exit(&spa_namespace_lock);
 136
 137         nvlist_free(nvlist);
 138
 139 out:
 140         if (buf != NULL)
 141                 kmem_free(buf, fsize);
 142
 143         kobj_close_file(file);
 144 }
 145
 146 static void
 147 spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
 148 {
 149         size_t buflen;
 150         char *buf;
 151         vnode_t *vp;
 152         int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX;
 153         char *temp;
 154         /*
 155          * If the nvlist is empty (NULL), then remove the old cachefile.
 156          */
 157         if (nvl == NULL) {
 158                 (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE);
 159                 return;
 160         }
 161
 162         /*
 163          * Pack the configuration into a buffer.
 164          */
 165         VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
 166
 167         buf = kmem_alloc(buflen, KM_SLEEP);
 168         temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
 169
 170         VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
 171             KM_SLEEP) == 0);
 172
 173         /*
 174          * Write the configuration to disk.  We need to do the traditional
 175          * 'write to temporary file, sync, move over original' to make sure we
 176          * always have a consistent view of the data.
 177          */
 178         (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
 179
 180         if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
 181                 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
 182                     0, RLIM64_INFINITY, kcred, NULL) == 0 &&
 183                     VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {
 184                         (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE);
 185                 }
 186                 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL);
 187                 VN_RELE(vp);
 188         }
 189
 190         (void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
 191
 192         kmem_free(buf, buflen);
 193         kmem_free(temp, MAXPATHLEN);
 194 }
 195
 196 /*
 197  * Synchronize pool configuration to disk.  This must be called with the
 198  * namespace lock held.
 199  */
 200 void
 201 spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
 202 {
 203         spa_config_dirent_t *dp, *tdp;
 204         nvlist_t *nvl;
 205
 206         ASSERT(MUTEX_HELD(&spa_namespace_lock));
 207
 208         /*
 209          * Iterate over all cachefiles for the pool, past or present.  When the
 210          * cachefile is changed, the new one is pushed onto this list, allowing
 211          * us to update previous cachefiles that no longer contain this pool.
 212          */
 213         for (dp = list_head(&target->spa_config_list); dp != NULL;
 214             dp = list_next(&target->spa_config_list, dp)) {
 215                 spa_t *spa = NULL;
 216                 if (dp->scd_path == NULL)
 217                         continue;
 218
 219                 /*
 220                  * Iterate over all pools, adding any matching pools to 'nvl'.
 221                  */
 222                 nvl = NULL;
 223                 while ((spa = spa_next(spa)) != NULL) {
 224                         if (spa == target && removing)
 225                                 continue;
 226                         mutex_enter(&spa->spa_props_lock);
 227                         tdp = list_head(&spa->spa_config_list);
 228                         if (spa->spa_config == NULL ||
 229                             tdp->scd_path == NULL ||
 230                             strcmp(tdp->scd_path, dp->scd_path) != 0) {
 231                                 mutex_exit(&spa->spa_props_lock);
 232                                 continue;
 233                         }
 234
 235                         if (nvl == NULL)
 236                                 VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
 237                                     KM_SLEEP) == 0);
 238
 239                         VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
 240                             spa->spa_config) == 0);
 241                         mutex_exit(&spa->spa_props_lock);
 242                 }
 243
 244                 spa_config_write(dp, nvl);
 245                 nvlist_free(nvl);
 246         }
 247
 248         /*
 249          * Remove any config entries older than the current one.
 250          */
 251         dp = list_head(&target->spa_config_list);
 252         while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
 253                 list_remove(&target->spa_config_list, tdp);
 254                 if (tdp->scd_path != NULL)
 255                         spa_strfree(tdp->scd_path);
 256                 kmem_free(tdp, sizeof (spa_config_dirent_t));
 257         }
 258
 259         spa_config_generation++;
 260
 261         if (postsysevent)
 262                 spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC);
 263 }
 264
 265 /*
 266  * Sigh.  Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
 267  * and we don't want to allow the local zone to see all the pools anyway.
 268  * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
 269  * information for all pool visible within the zone.
 270  */
 271 nvlist_t *
 272 spa_all_configs(uint64_t *generation)
 273 {
 274         nvlist_t *pools;
 275         spa_t *spa = NULL;
 276
 277         if (*generation == spa_config_generation)
 278                 return (NULL);
 279
 280         VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 281
 282         mutex_enter(&spa_namespace_lock);
 283         while ((spa = spa_next(spa)) != NULL) {
 284                 if (INGLOBALZONE(curproc) ||
 285                     zone_dataset_visible(spa_name(spa), NULL)) {
 286                         mutex_enter(&spa->spa_props_lock);
 287                         VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
 288                             spa->spa_config) == 0);
 289                         mutex_exit(&spa->spa_props_lock);
 290                 }
 291         }
 292         *generation = spa_config_generation;
 293         mutex_exit(&spa_namespace_lock);
 294
 295         return (pools);
 296 }
 297
 298 void
 299 spa_config_set(spa_t *spa, nvlist_t *config)
 300 {
 301         mutex_enter(&spa->spa_props_lock);
 302         if (spa->spa_config != NULL)
 303                 nvlist_free(spa->spa_config);
 304         spa->spa_config = config;
 305         mutex_exit(&spa->spa_props_lock);
 306 }
 307
 308 /*
 309  * Generate the pool's configuration based on the current in-core state.
 310  * We infer whether to generate a complete config or just one top-level config
 311  * based on whether vd is the root vdev.
 312  */
 313 nvlist_t *
 314 spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
 315 {
 316         nvlist_t *config, *nvroot;
 317         vdev_t *rvd = spa->spa_root_vdev;
 318         unsigned long hostid = 0;
 319         boolean_t locked = B_FALSE;
 320
 321         if (vd == NULL) {
 322                 vd = rvd;
 323                 locked = B_TRUE;
 324                 spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
 325         }
 326
 327         ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
 328             (SCL_CONFIG | SCL_STATE));
 329
 330         /*
 331          * If txg is -1, report the current value of spa->spa_config_txg.
 332          */
 333         if (txg == -1ULL)
 334                 txg = spa->spa_config_txg;
 335
 336         VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 337
 338         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
 339             spa_version(spa)) == 0);
 340         VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
 341             spa_name(spa)) == 0);
 342         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
 343             spa_state(spa)) == 0);
 344         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
 345             txg) == 0);
 346         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
 347             spa_guid(spa)) == 0);
 348         (void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
 349         if (hostid != 0) {
 350                 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
 351                     hostid) == 0);
 352         }
 353         VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
 354             utsname.nodename) == 0);
 355
 356         if (vd != rvd) {
 357                 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
 358                     vd->vdev_top->vdev_guid) == 0);
 359                 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
 360                     vd->vdev_guid) == 0);
 361                 if (vd->vdev_isspare)
 362                         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
 363                             1ULL) == 0);
 364                 if (vd->vdev_islog)
 365                         VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
 366                             1ULL) == 0);
 367                 vd = vd->vdev_top;              /* label contains top config */
 368         }
 369
 370         nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE, B_FALSE);
 371         VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
 372         nvlist_free(nvroot);
 373
 374         if (locked)
 375                 spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
 376
 377         return (config);
 378 }
 379
 380 /*
 381  * For a pool that's not currently a booting rootpool, update all disk labels,
 382  * generate a fresh config based on the current in-core state, and sync the
 383  * global config cache.
 384  */
 385 void
 386 spa_config_update(spa_t *spa, int what)
 387 {
 388         spa_config_update_common(spa, what, FALSE);
 389 }
 390
 391 /*
 392  * Update all disk labels, generate a fresh config based on the current
 393  * in-core state, and sync the global config cache (do not sync the config
 394  * cache if this is a booting rootpool).
 395  */
 396 void
 397 spa_config_update_common(spa_t *spa, int what, boolean_t isroot)
 398 {
 399         vdev_t *rvd = spa->spa_root_vdev;
 400         uint64_t txg;
 401         int c;
 402
 403         ASSERT(MUTEX_HELD(&spa_namespace_lock));
 404
 405         spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
 406         txg = spa_last_synced_txg(spa) + 1;
 407         if (what == SPA_CONFIG_UPDATE_POOL) {
 408                 vdev_config_dirty(rvd);
 409         } else {
 410                 /*
 411                  * If we have top-level vdevs that were added but have
 412                  * not yet been prepared for allocation, do that now.
 413                  * (It's safe now because the config cache is up to date,
 414                  * so it will be able to translate the new DVAs.)
 415                  * See comments in spa_vdev_add() for full details.
 416                  */
 417                 for (c = 0; c < rvd->vdev_children; c++) {
 418                         vdev_t *tvd = rvd->vdev_child[c];
 419                         if (tvd->vdev_ms_array == 0) {
 420                                 vdev_init(tvd, txg);
 421                                 vdev_config_dirty(tvd);
 422                         }
 423                 }
 424         }
 425         spa_config_exit(spa, SCL_ALL, FTAG);
 426
 427         /*
 428          * Wait for the mosconfig to be regenerated and synced.
 429          */
 430         txg_wait_synced(spa->spa_dsl_pool, txg);
 431
 432         /*
 433          * Update the global config cache to reflect the new mosconfig.
 434          */
 435         if (!isroot)
 436                 spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL);
 437
 438         if (what == SPA_CONFIG_UPDATE_POOL)
 439                 spa_config_update_common(spa, SPA_CONFIG_UPDATE_VDEVS, isroot);
 440 }