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 2007 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #pragma ident "%Z%%M% %I% %E% SMI" 28 29 #include <sys/spa.h> 30 #include <sys/spa_impl.h> 31 #include <sys/nvpair.h> 32 #include <sys/uio.h> 33 #include <sys/fs/zfs.h> 34 #include <sys/vdev_impl.h> 35 #include <sys/zfs_ioctl.h> 36 #include <sys/utsname.h> 37 #include <sys/systeminfo.h> 38 #include <sys/sunddi.h> 39 #ifdef _KERNEL 40 #include <sys/kobj.h> 41 #endif 42 43 /* 44 * Pool configuration repository. 45 * 46 * Pool configuration is stored as a packed nvlist on the filesystem. By 47 * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot 48 * (when the ZFS module is loaded). Pools can also have the 'cachefile' 49 * property set that allows them to be stored in an alternate location until 50 * the control of external software. 51 * 52 * For each cache file, we have a single nvlist which holds all the 53 * configuration information. When the module loads, we read this information 54 * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is 55 * maintained independently in spa.c. Whenever the namespace is modified, or 56 * the configuration of a pool is changed, we call spa_config_sync(), which 57 * walks through all the active pools and writes the configuration to disk. 58 */ 59 60 static uint64_t spa_config_generation = 1; 61 62 /* 63 * This can be overridden in userland to preserve an alternate namespace for 64 * userland pools when doing testing. 65 */ 66 const char *spa_config_dir = ZPOOL_CACHE_DIR; 67 68 /* 69 * Called when the module is first loaded, this routine loads the configuration 70 * file into the SPA namespace. It does not actually open or load the pools; it 71 * only populates the namespace. 72 */ 73 void 74 spa_config_load(void) 75 { 76 void *buf = NULL; 77 nvlist_t *nvlist, *child; 78 nvpair_t *nvpair; 79 spa_t *spa; 80 char pathname[128]; 81 struct _buf *file; 82 uint64_t fsize; 83 84 /* 85 * Open the configuration file. 86 */ 87 (void) snprintf(pathname, sizeof (pathname), "%s%s/%s", 88 (rootdir != NULL) ? "./" : "", spa_config_dir, ZPOOL_CACHE_FILE); 89 90 file = kobj_open_file(pathname); 91 if (file == (struct _buf *)-1) 92 return; 93 94 if (kobj_get_filesize(file, &fsize) != 0) 95 goto out; 96 97 buf = kmem_alloc(fsize, KM_SLEEP); 98 99 /* 100 * Read the nvlist from the file. 101 */ 102 if (kobj_read_file(file, buf, fsize, 0) < 0) 103 goto out; 104 105 /* 106 * Unpack the nvlist. 107 */ 108 if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0) 109 goto out; 110 111 /* 112 * Iterate over all elements in the nvlist, creating a new spa_t for 113 * each one with the specified configuration. 114 */ 115 mutex_enter(&spa_namespace_lock); 116 nvpair = NULL; 117 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) { 118 119 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST) 120 continue; 121 122 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0); 123 124 if (spa_lookup(nvpair_name(nvpair)) != NULL) 125 continue; 126 spa = spa_add(nvpair_name(nvpair), NULL); 127 128 /* 129 * We blindly duplicate the configuration here. If it's 130 * invalid, we will catch it when the pool is first opened. 131 */ 132 VERIFY(nvlist_dup(child, &spa->spa_config, 0) == 0); 133 } 134 mutex_exit(&spa_namespace_lock); 135 136 nvlist_free(nvlist); 137 138 out: 139 if (buf != NULL) 140 kmem_free(buf, fsize); 141 142 kobj_close_file(file); 143 } 144 145 /* 146 * This function is called when destroying or exporting a pool. It walks the 147 * list of active pools, and searches for any that match the given cache file. 148 * If there is only one cachefile, then the file is removed immediately, 149 * because we won't see the pool when iterating in spa_config_sync(). 150 */ 151 void 152 spa_config_check(const char *dir, const char *file) 153 { 154 size_t count = 0; 155 char pathname[128]; 156 spa_t *spa; 157 158 if (dir != NULL && strcmp(dir, "none") == 0) 159 return; 160 161 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 162 spa = NULL; 163 while ((spa = spa_next(spa)) != NULL) { 164 if (dir == NULL) { 165 if (spa->spa_config_dir == NULL) 166 count++; 167 } else { 168 if (spa->spa_config_dir && 169 strcmp(spa->spa_config_dir, dir) == 0 && 170 strcmp(spa->spa_config_file, file) == 0) 171 count++; 172 } 173 } 174 175 if (count == 1) { 176 if (dir == NULL) { 177 dir = spa_config_dir; 178 file = ZPOOL_CACHE_FILE; 179 } 180 181 (void) snprintf(pathname, sizeof (pathname), 182 "%s/%s", dir, file); 183 (void) vn_remove(pathname, UIO_SYSSPACE, RMFILE); 184 } 185 } 186 187 typedef struct spa_config_entry { 188 list_t sc_link; 189 const char *sc_dir; 190 const char *sc_file; 191 nvlist_t *sc_nvl; 192 } spa_config_entry_t; 193 194 static void 195 spa_config_entry_add(list_t *listp, spa_t *spa) 196 { 197 spa_config_entry_t *entry; 198 const char *dir, *file; 199 200 mutex_enter(&spa->spa_config_cache_lock); 201 if (!spa->spa_config || !spa->spa_name) { 202 mutex_exit(&spa->spa_config_cache_lock); 203 return; 204 } 205 206 if (spa->spa_config_dir) { 207 dir = spa->spa_config_dir; 208 file = spa->spa_config_file; 209 } else { 210 dir = spa_config_dir; 211 file = ZPOOL_CACHE_FILE; 212 } 213 214 if (strcmp(dir, "none") == 0) { 215 mutex_exit(&spa->spa_config_cache_lock); 216 return; 217 } 218 219 for (entry = list_head(listp); entry != NULL; 220 entry = list_next(listp, entry)) { 221 if (strcmp(entry->sc_dir, dir) == 0 && 222 strcmp(entry->sc_file, file) == 0) 223 break; 224 } 225 226 if (entry == NULL) { 227 entry = kmem_alloc(sizeof (spa_config_entry_t), KM_SLEEP); 228 entry->sc_dir = dir; 229 entry->sc_file = file; 230 VERIFY(nvlist_alloc(&entry->sc_nvl, NV_UNIQUE_NAME, 231 KM_SLEEP) == 0); 232 list_insert_tail(listp, entry); 233 } 234 235 VERIFY(nvlist_add_nvlist(entry->sc_nvl, spa->spa_name, 236 spa->spa_config) == 0); 237 mutex_exit(&spa->spa_config_cache_lock); 238 } 239 240 static void 241 spa_config_entry_write(spa_config_entry_t *entry) 242 { 243 nvlist_t *config = entry->sc_nvl; 244 size_t buflen; 245 char *buf; 246 vnode_t *vp; 247 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX; 248 char pathname[128]; 249 char pathname2[128]; 250 251 /* 252 * Pack the configuration into a buffer. 253 */ 254 VERIFY(nvlist_size(config, &buflen, NV_ENCODE_XDR) == 0); 255 256 buf = kmem_alloc(buflen, KM_SLEEP); 257 258 VERIFY(nvlist_pack(config, &buf, &buflen, NV_ENCODE_XDR, 259 KM_SLEEP) == 0); 260 261 /* 262 * Write the configuration to disk. We need to do the traditional 263 * 'write to temporary file, sync, move over original' to make sure we 264 * always have a consistent view of the data. 265 */ 266 (void) snprintf(pathname, sizeof (pathname), "%s/.%s", entry->sc_dir, 267 entry->sc_file); 268 269 if (vn_open(pathname, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) != 0) 270 goto out; 271 272 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE, 273 0, RLIM64_INFINITY, kcred, NULL) == 0 && 274 VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) { 275 (void) snprintf(pathname2, sizeof (pathname2), "%s/%s", 276 entry->sc_dir, entry->sc_file); 277 (void) vn_rename(pathname, pathname2, UIO_SYSSPACE); 278 } 279 280 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL); 281 VN_RELE(vp); 282 283 out: 284 (void) vn_remove(pathname, UIO_SYSSPACE, RMFILE); 285 kmem_free(buf, buflen); 286 } 287 288 /* 289 * Synchronize all pools to disk. This must be called with the namespace lock 290 * held. 291 */ 292 void 293 spa_config_sync(void) 294 { 295 spa_t *spa = NULL; 296 list_t files = { 0 }; 297 spa_config_entry_t *entry; 298 299 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 300 301 list_create(&files, sizeof (spa_config_entry_t), 302 offsetof(spa_config_entry_t, sc_link)); 303 304 /* 305 * Add all known pools to the configuration list, ignoring those with 306 * alternate root paths. 307 */ 308 spa = NULL; 309 while ((spa = spa_next(spa)) != NULL) 310 spa_config_entry_add(&files, spa); 311 312 while ((entry = list_head(&files)) != NULL) { 313 spa_config_entry_write(entry); 314 list_remove(&files, entry); 315 nvlist_free(entry->sc_nvl); 316 kmem_free(entry, sizeof (spa_config_entry_t)); 317 } 318 319 spa_config_generation++; 320 } 321 322 /* 323 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache, 324 * and we don't want to allow the local zone to see all the pools anyway. 325 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration 326 * information for all pool visible within the zone. 327 */ 328 nvlist_t * 329 spa_all_configs(uint64_t *generation) 330 { 331 nvlist_t *pools; 332 spa_t *spa; 333 334 if (*generation == spa_config_generation) 335 return (NULL); 336 337 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0); 338 339 spa = NULL; 340 mutex_enter(&spa_namespace_lock); 341 while ((spa = spa_next(spa)) != NULL) { 342 if (INGLOBALZONE(curproc) || 343 zone_dataset_visible(spa_name(spa), NULL)) { 344 mutex_enter(&spa->spa_config_cache_lock); 345 VERIFY(nvlist_add_nvlist(pools, spa_name(spa), 346 spa->spa_config) == 0); 347 mutex_exit(&spa->spa_config_cache_lock); 348 } 349 } 350 mutex_exit(&spa_namespace_lock); 351 352 *generation = spa_config_generation; 353 354 return (pools); 355 } 356 357 void 358 spa_config_set(spa_t *spa, nvlist_t *config) 359 { 360 mutex_enter(&spa->spa_config_cache_lock); 361 if (spa->spa_config != NULL) 362 nvlist_free(spa->spa_config); 363 spa->spa_config = config; 364 mutex_exit(&spa->spa_config_cache_lock); 365 } 366 367 /* 368 * Generate the pool's configuration based on the current in-core state. 369 * We infer whether to generate a complete config or just one top-level config 370 * based on whether vd is the root vdev. 371 */ 372 nvlist_t * 373 spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats) 374 { 375 nvlist_t *config, *nvroot; 376 vdev_t *rvd = spa->spa_root_vdev; 377 unsigned long hostid = 0; 378 379 ASSERT(spa_config_held(spa, RW_READER) || 380 spa_config_held(spa, RW_WRITER)); 381 382 if (vd == NULL) 383 vd = rvd; 384 385 /* 386 * If txg is -1, report the current value of spa->spa_config_txg. 387 */ 388 if (txg == -1ULL) 389 txg = spa->spa_config_txg; 390 391 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0); 392 393 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, 394 spa_version(spa)) == 0); 395 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, 396 spa_name(spa)) == 0); 397 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, 398 spa_state(spa)) == 0); 399 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, 400 txg) == 0); 401 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, 402 spa_guid(spa)) == 0); 403 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid); 404 if (hostid != 0) { 405 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 406 hostid) == 0); 407 } 408 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 409 utsname.nodename) == 0); 410 411 if (vd != rvd) { 412 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID, 413 vd->vdev_top->vdev_guid) == 0); 414 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID, 415 vd->vdev_guid) == 0); 416 if (vd->vdev_isspare) 417 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE, 418 1ULL) == 0); 419 if (vd->vdev_islog) 420 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG, 421 1ULL) == 0); 422 vd = vd->vdev_top; /* label contains top config */ 423 } 424 425 nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE); 426 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); 427 nvlist_free(nvroot); 428 429 return (config); 430 } 431 432 /* 433 * Update all disk labels, generate a fresh config based on the current 434 * in-core state, and sync the global config cache. 435 */ 436 void 437 spa_config_update(spa_t *spa, int what) 438 { 439 vdev_t *rvd = spa->spa_root_vdev; 440 uint64_t txg; 441 int c; 442 443 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 444 445 spa_config_enter(spa, RW_WRITER, FTAG); 446 txg = spa_last_synced_txg(spa) + 1; 447 if (what == SPA_CONFIG_UPDATE_POOL) { 448 vdev_config_dirty(rvd); 449 } else { 450 /* 451 * If we have top-level vdevs that were added but have 452 * not yet been prepared for allocation, do that now. 453 * (It's safe now because the config cache is up to date, 454 * so it will be able to translate the new DVAs.) 455 * See comments in spa_vdev_add() for full details. 456 */ 457 for (c = 0; c < rvd->vdev_children; c++) { 458 vdev_t *tvd = rvd->vdev_child[c]; 459 if (tvd->vdev_ms_array == 0) { 460 vdev_init(tvd, txg); 461 vdev_config_dirty(tvd); 462 } 463 } 464 } 465 spa_config_exit(spa, FTAG); 466 467 /* 468 * Wait for the mosconfig to be regenerated and synced. 469 */ 470 txg_wait_synced(spa->spa_dsl_pool, txg); 471 472 /* 473 * Update the global config cache to reflect the new mosconfig. 474 */ 475 spa_config_sync(); 476 477 if (what == SPA_CONFIG_UPDATE_POOL) 478 spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS); 479 } 480