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.
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.
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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2012 by Delphix. All rights reserved.
29 #include <sys/spa_impl.h>
30 #include <sys/nvpair.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/systeminfo.h>
37 #include <sys/sunddi.h>
38 #include <sys/zfeature.h>
45 * Pool configuration repository.
47 * Pool configuration is stored as a packed nvlist on the filesystem. By
48 * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
49 * (when the ZFS module is loaded). Pools can also have the 'cachefile'
50 * property set that allows them to be stored in an alternate location until
51 * the control of external software.
53 * For each cache file, we have a single nvlist which holds all the
54 * configuration information. When the module loads, we read this information
55 * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
56 * maintained independently in spa.c. Whenever the namespace is modified, or
57 * the configuration of a pool is changed, we call spa_config_sync(), which
58 * walks through all the active pools and writes the configuration to disk.
61 static uint64_t spa_config_generation = 1;
64 * This can be overridden in userland to preserve an alternate namespace for
65 * userland pools when doing testing.
67 char *spa_config_path = ZPOOL_CACHE;
70 * Called when the module is first loaded, this routine loads the configuration
71 * file into the SPA namespace. It does not actually open or load the pools; it
72 * only populates the namespace.
78 nvlist_t *nvlist, *child;
85 * Open the configuration file.
87 pathname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
89 (void) snprintf(pathname, MAXPATHLEN, "%s%s",
90 (rootdir != NULL) ? "./" : "", spa_config_path);
92 file = kobj_open_file(pathname);
94 kmem_free(pathname, MAXPATHLEN);
96 if (file == (struct _buf *)-1)
99 if (kobj_get_filesize(file, &fsize) != 0)
102 buf = kmem_alloc(fsize, KM_PUSHPAGE | KM_NODEBUG);
105 * Read the nvlist from the file.
107 if (kobj_read_file(file, buf, fsize, 0) < 0)
113 if (nvlist_unpack(buf, fsize, &nvlist, KM_PUSHPAGE) != 0)
117 * Iterate over all elements in the nvlist, creating a new spa_t for
118 * each one with the specified configuration.
120 mutex_enter(&spa_namespace_lock);
122 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
123 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
126 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
128 if (spa_lookup(nvpair_name(nvpair)) != NULL)
130 (void) spa_add(nvpair_name(nvpair), child, NULL);
132 mutex_exit(&spa_namespace_lock);
138 kmem_free(buf, fsize);
140 kobj_close_file(file);
144 spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
149 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX;
153 * If the nvlist is empty (NULL), then remove the old cachefile.
156 (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE);
161 * Pack the configuration into a buffer.
163 VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
165 buf = kmem_alloc(buflen, KM_PUSHPAGE | KM_NODEBUG);
166 temp = kmem_zalloc(MAXPATHLEN, KM_PUSHPAGE);
168 VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
172 * Write the configuration to disk. We need to do the traditional
173 * 'write to temporary file, sync, move over original' to make sure we
174 * always have a consistent view of the data.
176 (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
178 if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
179 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
180 0, RLIM64_INFINITY, kcred, NULL) == 0 &&
181 VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {
182 (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE);
184 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL);
187 (void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
189 kmem_free(buf, buflen);
190 kmem_free(temp, MAXPATHLEN);
194 * Synchronize pool configuration to disk. This must be called with the
195 * namespace lock held.
198 spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
200 spa_config_dirent_t *dp, *tdp;
203 ASSERT(MUTEX_HELD(&spa_namespace_lock));
205 if (rootdir == NULL || !(spa_mode_global & FWRITE))
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.
213 for (dp = list_head(&target->spa_config_list); dp != NULL;
214 dp = list_next(&target->spa_config_list, dp)) {
216 if (dp->scd_path == NULL)
220 * Iterate over all pools, adding any matching pools to 'nvl'.
223 while ((spa = spa_next(spa)) != NULL) {
224 if (spa == target && removing)
227 mutex_enter(&spa->spa_props_lock);
228 tdp = list_head(&spa->spa_config_list);
229 if (spa->spa_config == NULL ||
230 tdp->scd_path == NULL ||
231 strcmp(tdp->scd_path, dp->scd_path) != 0) {
232 mutex_exit(&spa->spa_props_lock);
237 VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
240 VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
241 spa->spa_config) == 0);
242 mutex_exit(&spa->spa_props_lock);
245 spa_config_write(dp, nvl);
250 * Remove any config entries older than the current one.
252 dp = list_head(&target->spa_config_list);
253 while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
254 list_remove(&target->spa_config_list, tdp);
255 if (tdp->scd_path != NULL)
256 spa_strfree(tdp->scd_path);
257 kmem_free(tdp, sizeof (spa_config_dirent_t));
260 spa_config_generation++;
263 spa_event_notify(target, NULL, FM_EREPORT_ZFS_CONFIG_SYNC);
267 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
268 * and we don't want to allow the local zone to see all the pools anyway.
269 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
270 * information for all pool visible within the zone.
273 spa_all_configs(uint64_t *generation)
278 if (*generation == spa_config_generation)
281 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
283 mutex_enter(&spa_namespace_lock);
284 while ((spa = spa_next(spa)) != NULL) {
285 if (INGLOBALZONE(curproc) ||
286 zone_dataset_visible(spa_name(spa), NULL)) {
287 mutex_enter(&spa->spa_props_lock);
288 VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
289 spa->spa_config) == 0);
290 mutex_exit(&spa->spa_props_lock);
293 *generation = spa_config_generation;
294 mutex_exit(&spa_namespace_lock);
300 spa_config_set(spa_t *spa, nvlist_t *config)
302 mutex_enter(&spa->spa_props_lock);
303 if (spa->spa_config != NULL)
304 nvlist_free(spa->spa_config);
305 spa->spa_config = config;
306 mutex_exit(&spa->spa_props_lock);
310 * Generate the pool's configuration based on the current in-core state.
311 * We infer whether to generate a complete config or just one top-level config
312 * based on whether vd is the root vdev.
315 spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
317 nvlist_t *config, *nvroot;
318 vdev_t *rvd = spa->spa_root_vdev;
319 unsigned long hostid = 0;
320 boolean_t locked = B_FALSE;
326 spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
329 ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
330 (SCL_CONFIG | SCL_STATE));
333 * If txg is -1, report the current value of spa->spa_config_txg.
336 txg = spa->spa_config_txg;
338 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
340 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
341 spa_version(spa)) == 0);
342 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
343 spa_name(spa)) == 0);
344 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
345 spa_state(spa)) == 0);
346 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
348 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
349 spa_guid(spa)) == 0);
350 VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
351 ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
355 hostid = zone_get_hostid(NULL);
358 * We're emulating the system's hostid in userland, so we can't use
361 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
364 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
367 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
368 utsname.nodename) == 0);
371 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
372 vd->vdev_top->vdev_guid) == 0);
373 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
374 vd->vdev_guid) == 0);
375 if (vd->vdev_isspare)
376 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
379 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
381 vd = vd->vdev_top; /* label contains top config */
384 * Only add the (potentially large) split information
385 * in the mos config, and not in the vdev labels
387 if (spa->spa_config_splitting != NULL)
388 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
389 spa->spa_config_splitting) == 0);
393 * Add the top-level config. We even add this on pools which
394 * don't support holes in the namespace.
396 vdev_top_config_generate(spa, config);
399 * If we're splitting, record the original pool's guid.
401 if (spa->spa_config_splitting != NULL &&
402 nvlist_lookup_uint64(spa->spa_config_splitting,
403 ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
404 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
408 nvroot = vdev_config_generate(spa, vd, getstats, 0);
409 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
413 * Store what's necessary for reading the MOS in the label.
415 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
416 spa->spa_label_features) == 0);
418 if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
419 ddt_histogram_t *ddh;
423 ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_PUSHPAGE);
424 ddt_get_dedup_histogram(spa, ddh);
425 VERIFY(nvlist_add_uint64_array(config,
426 ZPOOL_CONFIG_DDT_HISTOGRAM,
427 (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
428 kmem_free(ddh, sizeof (ddt_histogram_t));
430 ddo = kmem_zalloc(sizeof (ddt_object_t), KM_PUSHPAGE);
431 ddt_get_dedup_object_stats(spa, ddo);
432 VERIFY(nvlist_add_uint64_array(config,
433 ZPOOL_CONFIG_DDT_OBJ_STATS,
434 (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
435 kmem_free(ddo, sizeof (ddt_object_t));
437 dds = kmem_zalloc(sizeof (ddt_stat_t), KM_PUSHPAGE);
438 ddt_get_dedup_stats(spa, dds);
439 VERIFY(nvlist_add_uint64_array(config,
440 ZPOOL_CONFIG_DDT_STATS,
441 (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
442 kmem_free(dds, sizeof (ddt_stat_t));
446 spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
452 * Update all disk labels, generate a fresh config based on the current
453 * in-core state, and sync the global config cache (do not sync the config
454 * cache if this is a booting rootpool).
457 spa_config_update(spa_t *spa, int what)
459 vdev_t *rvd = spa->spa_root_vdev;
463 ASSERT(MUTEX_HELD(&spa_namespace_lock));
465 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
466 txg = spa_last_synced_txg(spa) + 1;
467 if (what == SPA_CONFIG_UPDATE_POOL) {
468 vdev_config_dirty(rvd);
471 * If we have top-level vdevs that were added but have
472 * not yet been prepared for allocation, do that now.
473 * (It's safe now because the config cache is up to date,
474 * so it will be able to translate the new DVAs.)
475 * See comments in spa_vdev_add() for full details.
477 for (c = 0; c < rvd->vdev_children; c++) {
478 vdev_t *tvd = rvd->vdev_child[c];
479 if (tvd->vdev_ms_array == 0)
480 vdev_metaslab_set_size(tvd);
481 vdev_expand(tvd, txg);
484 spa_config_exit(spa, SCL_ALL, FTAG);
487 * Wait for the mosconfig to be regenerated and synced.
489 txg_wait_synced(spa->spa_dsl_pool, txg);
492 * Update the global config cache to reflect the new mosconfig.
494 if (!spa->spa_is_root)
495 spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL);
497 if (what == SPA_CONFIG_UPDATE_POOL)
498 spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
501 #if defined(_KERNEL) && defined(HAVE_SPL)
502 EXPORT_SYMBOL(spa_config_sync);
503 EXPORT_SYMBOL(spa_config_load);
504 EXPORT_SYMBOL(spa_all_configs);
505 EXPORT_SYMBOL(spa_config_set);
506 EXPORT_SYMBOL(spa_config_generate);
507 EXPORT_SYMBOL(spa_config_update);
509 module_param(spa_config_path, charp, 0444);
510 MODULE_PARM_DESC(spa_config_path, "SPA config file (/etc/zfs/zpool.cache)");