spa_config.c revision 168962
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/zfs_context.h> 30#include <sys/spa.h> 31#include <sys/spa_impl.h> 32#include <sys/nvpair.h> 33#include <sys/uio.h> 34#include <sys/fs/zfs.h> 35#include <sys/vdev_impl.h> 36#include <sys/zfs_ioctl.h> 37#include <sys/utsname.h> 38#include <sys/sunddi.h> 39#ifdef _KERNEL 40#include <sys/kobj.h> 41#endif 42 43/* 44 * Pool configuration repository. 45 * 46 * The configuration for all pools, in addition to being stored on disk, is 47 * stored in /etc/zfs/zpool.cache as a packed nvlist. The kernel maintains 48 * this list as pools are created, destroyed, or modified. 49 * 50 * We have a single nvlist which holds all the configuration information. When 51 * the module loads, we read this information from the cache and populate the 52 * SPA namespace. This namespace is maintained independently in spa.c. 53 * Whenever the namespace is modified, or the configuration of a pool is 54 * changed, we call spa_config_sync(), which walks through all the active pools 55 * 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_dir = ZPOOL_CACHE_DIR; 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[128]; 79 struct _buf *file; 80 uint64_t fsize; 81 82 /* 83 * Open the configuration file. 84 */ 85 (void) snprintf(pathname, sizeof (pathname), "%s/%s", 86 spa_config_dir, ZPOOL_CACHE_FILE); 87 88 file = kobj_open_file(pathname); 89 if (file == (struct _buf *)-1) { 90 ZFS_LOG(1, "Cannot open %s.", pathname); 91 return; 92 } 93 94 if (kobj_get_filesize(file, &fsize) != 0) { 95 ZFS_LOG(1, "Cannot get size of %s.", pathname); 96 goto out; 97 } 98 99 buf = kmem_alloc(fsize, KM_SLEEP); 100 101 /* 102 * Read the nvlist from the file. 103 */ 104 if (kobj_read_file(file, buf, fsize, 0) < 0) { 105 ZFS_LOG(1, "Cannot read %s.", pathname); 106 goto out; 107 } 108 109 /* 110 * Unpack the nvlist. 111 */ 112 if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0) 113 goto out; 114 115 ZFS_LOG(1, "File %s loaded.", pathname); 116 117 /* 118 * Iterate over all elements in the nvlist, creating a new spa_t for 119 * each one with the specified configuration. 120 */ 121 mutex_enter(&spa_namespace_lock); 122 nvpair = NULL; 123 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) { 124 125 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST) 126 continue; 127 128 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0); 129 130 if (spa_lookup(nvpair_name(nvpair)) != NULL) 131 continue; 132 spa = spa_add(nvpair_name(nvpair), NULL); 133 134 /* 135 * We blindly duplicate the configuration here. If it's 136 * invalid, we will catch it when the pool is first opened. 137 */ 138 VERIFY(nvlist_dup(child, &spa->spa_config, 0) == 0); 139 } 140 mutex_exit(&spa_namespace_lock); 141 142 nvlist_free(nvlist); 143 144out: 145 if (buf != NULL) 146 kmem_free(buf, fsize); 147 148 kobj_close_file(file); 149} 150 151/* 152 * Synchronize all pools to disk. This must be called with the namespace lock 153 * held. 154 */ 155void 156spa_config_sync(void) 157{ 158 spa_t *spa = NULL; 159 nvlist_t *config; 160 size_t buflen; 161 char *buf; 162 vnode_t *vp; 163 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX; 164 char pathname[128]; 165 char pathname2[128]; 166 167 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 168 169 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0); 170 171 /* 172 * Add all known pools to the configuration list, ignoring those with 173 * alternate root paths. 174 */ 175 spa = NULL; 176 while ((spa = spa_next(spa)) != NULL) { 177 mutex_enter(&spa->spa_config_cache_lock); 178 if (spa->spa_config && spa->spa_name && spa->spa_root == NULL) 179 VERIFY(nvlist_add_nvlist(config, spa->spa_name, 180 spa->spa_config) == 0); 181 mutex_exit(&spa->spa_config_cache_lock); 182 } 183 184 /* 185 * Pack the configuration into a buffer. 186 */ 187 VERIFY(nvlist_size(config, &buflen, NV_ENCODE_XDR) == 0); 188 189 buf = kmem_alloc(buflen, KM_SLEEP); 190 191 VERIFY(nvlist_pack(config, &buf, &buflen, NV_ENCODE_XDR, 192 KM_SLEEP) == 0); 193 194 /* 195 * Write the configuration to disk. We need to do the traditional 196 * 'write to temporary file, sync, move over original' to make sure we 197 * always have a consistent view of the data. 198 */ 199 (void) snprintf(pathname, sizeof (pathname), "%s/%s", spa_config_dir, 200 ZPOOL_CACHE_TMP); 201 202 if (vn_open(pathname, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) != 0) 203 goto out; 204 205 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE, 206 0, RLIM64_INFINITY, kcred, NULL) == 0 && 207 VOP_FSYNC(vp, FSYNC, kcred) == 0) { 208 (void) snprintf(pathname2, sizeof (pathname2), "%s/%s", 209 spa_config_dir, ZPOOL_CACHE_FILE); 210 (void) vn_rename(pathname, pathname2, UIO_SYSSPACE); 211 } 212 213 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred); 214 VN_RELE(vp); 215 216out: 217 (void) vn_remove(pathname, UIO_SYSSPACE, RMFILE); 218 spa_config_generation++; 219 220 kmem_free(buf, buflen); 221 nvlist_free(config); 222} 223 224/* 225 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache, 226 * and we don't want to allow the local zone to see all the pools anyway. 227 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration 228 * information for all pool visible within the zone. 229 */ 230nvlist_t * 231spa_all_configs(uint64_t *generation) 232{ 233 nvlist_t *pools; 234 spa_t *spa; 235 236 if (*generation == spa_config_generation) 237 return (NULL); 238 239 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0); 240 241 spa = NULL; 242 mutex_enter(&spa_namespace_lock); 243 while ((spa = spa_next(spa)) != NULL) { 244 if (INGLOBALZONE(curproc) || 245 zone_dataset_visible(spa_name(spa), NULL)) { 246 mutex_enter(&spa->spa_config_cache_lock); 247 VERIFY(nvlist_add_nvlist(pools, spa_name(spa), 248 spa->spa_config) == 0); 249 mutex_exit(&spa->spa_config_cache_lock); 250 } 251 } 252 mutex_exit(&spa_namespace_lock); 253 254 *generation = spa_config_generation; 255 256 return (pools); 257} 258 259void 260spa_config_set(spa_t *spa, nvlist_t *config) 261{ 262 mutex_enter(&spa->spa_config_cache_lock); 263 if (spa->spa_config != NULL) 264 nvlist_free(spa->spa_config); 265 spa->spa_config = config; 266 mutex_exit(&spa->spa_config_cache_lock); 267} 268 269/* 270 * Generate the pool's configuration based on the current in-core state. 271 * We infer whether to generate a complete config or just one top-level config 272 * based on whether vd is the root vdev. 273 */ 274nvlist_t * 275spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats) 276{ 277 nvlist_t *config, *nvroot; 278 vdev_t *rvd = spa->spa_root_vdev; 279 unsigned long hostid = 0; 280 281 ASSERT(spa_config_held(spa, RW_READER)); 282 283 if (vd == NULL) 284 vd = rvd; 285 286 /* 287 * If txg is -1, report the current value of spa->spa_config_txg. 288 */ 289 if (txg == -1ULL) 290 txg = spa->spa_config_txg; 291 292 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0); 293 294 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, 295 spa_version(spa)) == 0); 296 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, 297 spa_name(spa)) == 0); 298 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, 299 spa_state(spa)) == 0); 300 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, 301 txg) == 0); 302 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, 303 spa_guid(spa)) == 0); 304 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid); 305 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 306 hostid) == 0); 307 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 308 utsname.nodename) == 0); 309 310 if (vd != rvd) { 311 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID, 312 vd->vdev_top->vdev_guid) == 0); 313 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID, 314 vd->vdev_guid) == 0); 315 if (vd->vdev_isspare) 316 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE, 317 1ULL) == 0); 318 vd = vd->vdev_top; /* label contains top config */ 319 } 320 321 nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE); 322 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0); 323 nvlist_free(nvroot); 324 325 return (config); 326} 327 328/* 329 * Update all disk labels, generate a fresh config based on the current 330 * in-core state, and sync the global config cache. 331 */ 332void 333spa_config_update(spa_t *spa, int what) 334{ 335 vdev_t *rvd = spa->spa_root_vdev; 336 uint64_t txg; 337 int c; 338 339 ASSERT(MUTEX_HELD(&spa_namespace_lock)); 340 341 spa_config_enter(spa, RW_WRITER, FTAG); 342 txg = spa_last_synced_txg(spa) + 1; 343 if (what == SPA_CONFIG_UPDATE_POOL) { 344 vdev_config_dirty(rvd); 345 } else { 346 /* 347 * If we have top-level vdevs that were added but have 348 * not yet been prepared for allocation, do that now. 349 * (It's safe now because the config cache is up to date, 350 * so it will be able to translate the new DVAs.) 351 * See comments in spa_vdev_add() for full details. 352 */ 353 for (c = 0; c < rvd->vdev_children; c++) { 354 vdev_t *tvd = rvd->vdev_child[c]; 355 if (tvd->vdev_ms_array == 0) { 356 vdev_init(tvd, txg); 357 vdev_config_dirty(tvd); 358 } 359 } 360 } 361 spa_config_exit(spa, FTAG); 362 363 /* 364 * Wait for the mosconfig to be regenerated and synced. 365 */ 366 txg_wait_synced(spa->spa_dsl_pool, txg); 367 368 /* 369 * Update the global config cache to reflect the new mosconfig. 370 */ 371 spa_config_sync(); 372 373 if (what == SPA_CONFIG_UPDATE_POOL) 374 spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS); 375} 376