spa_config.c revision 185171
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/zfs_context.h> 28#include <sys/spa.h> 29#include <sys/spa_impl.h> 30#include <sys/nvpair.h> 31#include <sys/uio.h> 32#include <sys/fs/zfs.h> 33#include <sys/vdev_impl.h> 34#include <sys/zfs_ioctl.h> 35#include <sys/utsname.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 58static 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 */ 64const 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 */ 71void 72spa_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 87 (void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path); 88 89 file = kobj_open_file(pathname); 90 91 if (file == (struct _buf *)-1) { 92 ZFS_LOG(1, "Cannot open %s.", pathname); 93 goto out; 94 } 95 96 if (kobj_get_filesize(file, &fsize) != 0) { 97 ZFS_LOG(1, "Cannot get size of %s.", pathname); 98 goto out; 99 } 100 101 buf = kmem_alloc(fsize, KM_SLEEP); 102 103 /* 104 * Read the nvlist from the file. 105 */ 106 if (kobj_read_file(file, buf, fsize, 0) < 0) { 107 ZFS_LOG(1, "Cannot read %s.", pathname); 108 goto out; 109 } 110 111 /* 112 * Unpack the nvlist. 113 */ 114 if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0) 115 goto out; 116 117 ZFS_LOG(1, "File %s loaded.", pathname); 118 119 /* 120 * Iterate over all elements in the nvlist, creating a new spa_t for 121 * each one with the specified configuration. 122 */ 123 mutex_enter(&spa_namespace_lock); 124 nvpair = NULL; 125 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) { 126 127 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST) 128 continue; 129 130 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0); 131 132 if (spa_lookup(nvpair_name(nvpair)) != NULL) 133 continue; 134 spa = spa_add(nvpair_name(nvpair), NULL); 135 136 /* 137 * We blindly duplicate the configuration here. If it's 138 * invalid, we will catch it when the pool is first opened. 139 */ 140 VERIFY(nvlist_dup(child, &spa->spa_config, 0) == 0); 141 } 142 mutex_exit(&spa_namespace_lock); 143 144 nvlist_free(nvlist); 145 146out: 147 kmem_free(pathname, MAXPATHLEN); 148 if (buf != NULL) 149 kmem_free(buf, fsize); 150 if (file != (struct _buf *)-1) 151 kobj_close_file(file); 152} 153 154static void 155spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl) 156{ 157 size_t buflen; 158 char *buf; 159 vnode_t *vp; 160 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX; 161 char *temp; 162 163 /* 164 * If the nvlist is empty (NULL), then remove the old cachefile. 165 */ 166 if (nvl == NULL) { 167 (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE); 168 return; 169 } 170 171 /* 172 * Pack the configuration into a buffer. 173 */ 174 VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0); 175 176 buf = kmem_alloc(buflen, KM_SLEEP); 177 temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 178 179 VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR, 180 KM_SLEEP) == 0); 181 182 /* 183 * Write the configuration to disk. We need to do the traditional 184 * 'write to temporary file, sync, move over original' to make sure we 185 * always have a consistent view of the data. 186 */ 187 (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path); 188 189 if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) { 190 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE, 191 0, RLIM64_INFINITY, kcred, NULL) == 0 && 192 VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) { 193 (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE); 194 } 195 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL); 196 VN_RELE(vp); 197 } 198 199 (void) vn_remove(temp, UIO_SYSSPACE, RMFILE); 200 201 kmem_free(buf, buflen); 202 kmem_free(temp, MAXPATHLEN); 203} 204 205/* 206 * Synchronize pool configuration to disk. This must be called with the 207 * namespace lock held. 208 */ 209void 210spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent) 211{ 212 spa_config_dirent_t *dp, *tdp; 213 nvlist_t *nvl; 214 215 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 216 217 /* 218 * Iterate over all cachefiles for the pool, past or present. When the 219 * cachefile is changed, the new one is pushed onto this list, allowing 220 * us to update previous cachefiles that no longer contain this pool. 221 */ 222 for (dp = list_head(&target->spa_config_list); dp != NULL; 223 dp = list_next(&target->spa_config_list, dp)) { 224 spa_t *spa = NULL; 225 if (dp->scd_path == NULL) 226 continue; 227 228 /* 229 * Iterate over all pools, adding any matching pools to 'nvl'. 230 */ 231 nvl = NULL; 232 while ((spa = spa_next(spa)) != NULL) { 233 if (spa == target && removing) 234 continue; 235 236 mutex_enter(&spa->spa_props_lock); 237 tdp = list_head(&spa->spa_config_list); 238 if (spa->spa_config == NULL || 239 tdp->scd_path == NULL || 240 strcmp(tdp->scd_path, dp->scd_path) != 0) { 241 mutex_exit(&spa->spa_props_lock); 242 continue; 243 } 244 245 if (nvl == NULL) 246 VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, 247 KM_SLEEP) == 0); 248 249 VERIFY(nvlist_add_nvlist(nvl, spa->spa_name, 250 spa->spa_config) == 0); 251 mutex_exit(&spa->spa_props_lock); 252 } 253 254 spa_config_write(dp, nvl); 255 nvlist_free(nvl); 256 } 257 258 /* 259 * Remove any config entries older than the current one. 260 */ 261 dp = list_head(&target->spa_config_list); 262 while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) { 263 list_remove(&target->spa_config_list, tdp); 264 if (tdp->scd_path != NULL) 265 spa_strfree(tdp->scd_path); 266 kmem_free(tdp, sizeof (spa_config_dirent_t)); 267 } 268 269 spa_config_generation++; 270 271 if (postsysevent) 272 spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC); 273} 274 275/* 276 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache, 277 * and we don't want to allow the local zone to see all the pools anyway. 278 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration 279 * information for all pool visible within the zone. 280 */ 281nvlist_t * 282spa_all_configs(uint64_t *generation) 283{ 284 nvlist_t *pools; 285 spa_t *spa = NULL; 286 287 if (*generation == spa_config_generation) 288 return (NULL); 289 290 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0); 291 292 mutex_enter(&spa_namespace_lock); 293 while ((spa = spa_next(spa)) != NULL) { 294 if (INGLOBALZONE(curthread) || 295 zone_dataset_visible(spa_name(spa), NULL)) { 296 mutex_enter(&spa->spa_props_lock); 297 VERIFY(nvlist_add_nvlist(pools, spa_name(spa), 298 spa->spa_config) == 0); 299 mutex_exit(&spa->spa_props_lock); 300 } 301 } 302 *generation = spa_config_generation; 303 mutex_exit(&spa_namespace_lock); 304 305 return (pools); 306} 307 308void 309spa_config_set(spa_t *spa, nvlist_t *config) 310{ 311 mutex_enter(&spa->spa_props_lock); 312 if (spa->spa_config != NULL) 313 nvlist_free(spa->spa_config); 314 spa->spa_config = config; 315 mutex_exit(&spa->spa_props_lock); 316} 317 318/* 319 * Generate the pool's configuration based on the current in-core state. 320 * We infer whether to generate a complete config or just one top-level config 321 * based on whether vd is the root vdev. 322 */ 323nvlist_t * 324spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats) 325{ 326 nvlist_t *config, *nvroot; 327 vdev_t *rvd = spa->spa_root_vdev; 328 unsigned long hostid = 0; 329 boolean_t locked = B_FALSE; 330 331 if (vd == NULL) { 332 vd = rvd; 333 locked = B_TRUE; 334 spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER); 335 } 336 337 ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) == 338 (SCL_CONFIG | SCL_STATE)); 339 340 /* 341 * If txg is -1, report the current value of spa->spa_config_txg. 342 */ 343 if (txg == -1ULL) 344 txg = spa->spa_config_txg; 345 346 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0); 347 348 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, 349 spa_version(spa)) == 0); 350 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, 351 spa_name(spa)) == 0); 352 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, 353 spa_state(spa)) == 0); 354 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, 355 txg) == 0); 356 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, 357 spa_guid(spa)) == 0); 358 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid); 359 if (hostid != 0) { 360 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 361 hostid) == 0); 362 } 363 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 364 utsname.nodename) == 0); 365 366 if (vd != rvd) { 367 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID, 368 vd->vdev_top->vdev_guid) == 0); 369 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID, 370 vd->vdev_guid) == 0); 371 if (vd->vdev_isspare) 372 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE, 373 1ULL) == 0); 374 if (vd->vdev_islog) 375 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG, 376 1ULL) == 0); 377 vd = vd->vdev_top; /* label contains top config */ 378 } 379 380 nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE, B_FALSE); 381 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); 382 nvlist_free(nvroot); 383 384 if (locked) 385 spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG); 386 387 return (config); 388} 389 390/* 391 * For a pool that's not currently a booting rootpool, update all disk labels, 392 * generate a fresh config based on the current in-core state, and sync the 393 * global config cache. 394 */ 395void 396spa_config_update(spa_t *spa, int what) 397{ 398 spa_config_update_common(spa, what, FALSE); 399} 400 401/* 402 * Update all disk labels, generate a fresh config based on the current 403 * in-core state, and sync the global config cache (do not sync the config 404 * cache if this is a booting rootpool). 405 */ 406void 407spa_config_update_common(spa_t *spa, int what, boolean_t isroot) 408{ 409 vdev_t *rvd = spa->spa_root_vdev; 410 uint64_t txg; 411 int c; 412 413 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 414 415 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); 416 txg = spa_last_synced_txg(spa) + 1; 417 if (what == SPA_CONFIG_UPDATE_POOL) { 418 vdev_config_dirty(rvd); 419 } else { 420 /* 421 * If we have top-level vdevs that were added but have 422 * not yet been prepared for allocation, do that now. 423 * (It's safe now because the config cache is up to date, 424 * so it will be able to translate the new DVAs.) 425 * See comments in spa_vdev_add() for full details. 426 */ 427 for (c = 0; c < rvd->vdev_children; c++) { 428 vdev_t *tvd = rvd->vdev_child[c]; 429 if (tvd->vdev_ms_array == 0) { 430 vdev_init(tvd, txg); 431 vdev_config_dirty(tvd); 432 } 433 } 434 } 435 spa_config_exit(spa, SCL_ALL, FTAG); 436 437 /* 438 * Wait for the mosconfig to be regenerated and synced. 439 */ 440 txg_wait_synced(spa->spa_dsl_pool, txg); 441 442 /* 443 * Update the global config cache to reflect the new mosconfig. 444 */ 445 if (!isroot) 446 spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL); 447 448 if (what == SPA_CONFIG_UPDATE_POOL) 449 spa_config_update_common(spa, SPA_CONFIG_UPDATE_VDEVS, isroot); 450} 451