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) 2013 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
;
68 int zfs_autoimport_disable
= 0;
71 * Called when the module is first loaded, this routine loads the configuration
72 * file into the SPA namespace. It does not actually open or load the pools; it
73 * only populates the namespace.
79 nvlist_t
*nvlist
, *child
;
85 if (zfs_autoimport_disable
)
89 * Open the configuration file.
91 pathname
= kmem_alloc(MAXPATHLEN
, KM_PUSHPAGE
);
93 (void) snprintf(pathname
, MAXPATHLEN
, "%s%s",
94 (rootdir
!= NULL
) ? "./" : "", spa_config_path
);
96 file
= kobj_open_file(pathname
);
98 kmem_free(pathname
, MAXPATHLEN
);
100 if (file
== (struct _buf
*)-1)
103 if (kobj_get_filesize(file
, &fsize
) != 0)
106 buf
= kmem_alloc(fsize
, KM_PUSHPAGE
| KM_NODEBUG
);
109 * Read the nvlist from the file.
111 if (kobj_read_file(file
, buf
, fsize
, 0) < 0)
117 if (nvlist_unpack(buf
, fsize
, &nvlist
, KM_PUSHPAGE
) != 0)
121 * Iterate over all elements in the nvlist, creating a new spa_t for
122 * each one with the specified configuration.
124 mutex_enter(&spa_namespace_lock
);
126 while ((nvpair
= nvlist_next_nvpair(nvlist
, nvpair
)) != NULL
) {
127 if (nvpair_type(nvpair
) != DATA_TYPE_NVLIST
)
130 VERIFY(nvpair_value_nvlist(nvpair
, &child
) == 0);
132 if (spa_lookup(nvpair_name(nvpair
)) != NULL
)
134 (void) spa_add(nvpair_name(nvpair
), child
, NULL
);
136 mutex_exit(&spa_namespace_lock
);
142 kmem_free(buf
, fsize
);
144 kobj_close_file(file
);
148 spa_config_write(spa_config_dirent_t
*dp
, nvlist_t
*nvl
)
153 int oflags
= FWRITE
| FTRUNC
| FCREAT
| FOFFMAX
;
157 * If the nvlist is empty (NULL), then remove the old cachefile.
160 (void) vn_remove(dp
->scd_path
, UIO_SYSSPACE
, RMFILE
);
165 * Pack the configuration into a buffer.
167 VERIFY(nvlist_size(nvl
, &buflen
, NV_ENCODE_XDR
) == 0);
169 buf
= kmem_alloc(buflen
, KM_PUSHPAGE
| KM_NODEBUG
);
170 temp
= kmem_zalloc(MAXPATHLEN
, KM_PUSHPAGE
);
172 VERIFY(nvlist_pack(nvl
, &buf
, &buflen
, NV_ENCODE_XDR
,
176 * Write the configuration to disk. We need to do the traditional
177 * 'write to temporary file, sync, move over original' to make sure we
178 * always have a consistent view of the data.
180 (void) snprintf(temp
, MAXPATHLEN
, "%s.tmp", dp
->scd_path
);
182 if (vn_open(temp
, UIO_SYSSPACE
, oflags
, 0644, &vp
, CRCREAT
, 0) == 0) {
183 if (vn_rdwr(UIO_WRITE
, vp
, buf
, buflen
, 0, UIO_SYSSPACE
,
184 0, RLIM64_INFINITY
, kcred
, NULL
) == 0 &&
185 VOP_FSYNC(vp
, FSYNC
, kcred
, NULL
) == 0) {
186 (void) vn_rename(temp
, dp
->scd_path
, UIO_SYSSPACE
);
188 (void) VOP_CLOSE(vp
, oflags
, 1, 0, kcred
, NULL
);
191 (void) vn_remove(temp
, UIO_SYSSPACE
, RMFILE
);
193 kmem_free(buf
, buflen
);
194 kmem_free(temp
, MAXPATHLEN
);
198 * Synchronize pool configuration to disk. This must be called with the
199 * namespace lock held. Synchronizing the pool cache is typically done after
200 * the configuration has been synced to the MOS. This exposes a window where
201 * the MOS config will have been updated but the cache file has not. If
202 * the system were to crash at that instant then the cached config may not
203 * contain the correct information to open the pool and an explicity import
207 spa_config_sync(spa_t
*target
, boolean_t removing
, boolean_t postsysevent
)
209 spa_config_dirent_t
*dp
, *tdp
;
212 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
214 if (rootdir
== NULL
|| !(spa_mode_global
& FWRITE
))
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.
222 for (dp
= list_head(&target
->spa_config_list
); dp
!= NULL
;
223 dp
= list_next(&target
->spa_config_list
, dp
)) {
225 if (dp
->scd_path
== NULL
)
229 * Iterate over all pools, adding any matching pools to 'nvl'.
232 while ((spa
= spa_next(spa
)) != NULL
) {
234 * Skip over our own pool if we're about to remove
235 * ourselves from the spa namespace or any pool that
236 * is readonly. Since we cannot guarantee that a
237 * readonly pool would successfully import upon reboot,
238 * we don't allow them to be written to the cache file.
240 if ((spa
== target
&& removing
) ||
244 mutex_enter(&spa
->spa_props_lock
);
245 tdp
= list_head(&spa
->spa_config_list
);
246 if (spa
->spa_config
== NULL
||
247 tdp
->scd_path
== NULL
||
248 strcmp(tdp
->scd_path
, dp
->scd_path
) != 0) {
249 mutex_exit(&spa
->spa_props_lock
);
254 VERIFY(nvlist_alloc(&nvl
, NV_UNIQUE_NAME
,
257 VERIFY(nvlist_add_nvlist(nvl
, spa
->spa_name
,
258 spa
->spa_config
) == 0);
259 mutex_exit(&spa
->spa_props_lock
);
262 spa_config_write(dp
, nvl
);
267 * Remove any config entries older than the current one.
269 dp
= list_head(&target
->spa_config_list
);
270 while ((tdp
= list_next(&target
->spa_config_list
, dp
)) != NULL
) {
271 list_remove(&target
->spa_config_list
, tdp
);
272 if (tdp
->scd_path
!= NULL
)
273 spa_strfree(tdp
->scd_path
);
274 kmem_free(tdp
, sizeof (spa_config_dirent_t
));
277 spa_config_generation
++;
280 spa_event_notify(target
, NULL
, FM_EREPORT_ZFS_CONFIG_SYNC
);
284 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
285 * and we don't want to allow the local zone to see all the pools anyway.
286 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
287 * information for all pool visible within the zone.
290 spa_all_configs(uint64_t *generation
)
295 if (*generation
== spa_config_generation
)
298 VERIFY(nvlist_alloc(&pools
, NV_UNIQUE_NAME
, KM_PUSHPAGE
) == 0);
300 mutex_enter(&spa_namespace_lock
);
301 while ((spa
= spa_next(spa
)) != NULL
) {
302 if (INGLOBALZONE(curproc
) ||
303 zone_dataset_visible(spa_name(spa
), NULL
)) {
304 mutex_enter(&spa
->spa_props_lock
);
305 VERIFY(nvlist_add_nvlist(pools
, spa_name(spa
),
306 spa
->spa_config
) == 0);
307 mutex_exit(&spa
->spa_props_lock
);
310 *generation
= spa_config_generation
;
311 mutex_exit(&spa_namespace_lock
);
317 spa_config_set(spa_t
*spa
, nvlist_t
*config
)
319 mutex_enter(&spa
->spa_props_lock
);
320 if (spa
->spa_config
!= NULL
)
321 nvlist_free(spa
->spa_config
);
322 spa
->spa_config
= config
;
323 mutex_exit(&spa
->spa_props_lock
);
327 * Generate the pool's configuration based on the current in-core state.
329 * We infer whether to generate a complete config or just one top-level config
330 * based on whether vd is the root vdev.
333 spa_config_generate(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
, int getstats
)
335 nvlist_t
*config
, *nvroot
;
336 vdev_t
*rvd
= spa
->spa_root_vdev
;
337 unsigned long hostid
= 0;
338 boolean_t locked
= B_FALSE
;
344 spa_config_enter(spa
, SCL_CONFIG
| SCL_STATE
, FTAG
, RW_READER
);
347 ASSERT(spa_config_held(spa
, SCL_CONFIG
| SCL_STATE
, RW_READER
) ==
348 (SCL_CONFIG
| SCL_STATE
));
351 * If txg is -1, report the current value of spa->spa_config_txg.
354 txg
= spa
->spa_config_txg
;
356 VERIFY(nvlist_alloc(&config
, NV_UNIQUE_NAME
, KM_PUSHPAGE
) == 0);
358 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_VERSION
,
359 spa_version(spa
)) == 0);
360 VERIFY(nvlist_add_string(config
, ZPOOL_CONFIG_POOL_NAME
,
361 spa_name(spa
)) == 0);
362 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
363 spa_state(spa
)) == 0);
364 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_POOL_TXG
,
366 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
367 spa_guid(spa
)) == 0);
368 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_ERRATA
,
369 spa
->spa_errata
) == 0);
370 VERIFY(spa
->spa_comment
== NULL
|| nvlist_add_string(config
,
371 ZPOOL_CONFIG_COMMENT
, spa
->spa_comment
) == 0);
375 hostid
= zone_get_hostid(NULL
);
378 * We're emulating the system's hostid in userland, so we can't use
381 (void) ddi_strtoul(hw_serial
, NULL
, 10, &hostid
);
384 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_HOSTID
,
387 VERIFY(nvlist_add_string(config
, ZPOOL_CONFIG_HOSTNAME
,
388 utsname
.nodename
) == 0);
391 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_TOP_GUID
,
392 vd
->vdev_top
->vdev_guid
) == 0);
393 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_GUID
,
394 vd
->vdev_guid
) == 0);
395 if (vd
->vdev_isspare
)
396 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_IS_SPARE
,
399 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_IS_LOG
,
401 vd
= vd
->vdev_top
; /* label contains top config */
404 * Only add the (potentially large) split information
405 * in the mos config, and not in the vdev labels
407 if (spa
->spa_config_splitting
!= NULL
)
408 VERIFY(nvlist_add_nvlist(config
, ZPOOL_CONFIG_SPLIT
,
409 spa
->spa_config_splitting
) == 0);
413 * Add the top-level config. We even add this on pools which
414 * don't support holes in the namespace.
416 vdev_top_config_generate(spa
, config
);
419 * If we're splitting, record the original pool's guid.
421 if (spa
->spa_config_splitting
!= NULL
&&
422 nvlist_lookup_uint64(spa
->spa_config_splitting
,
423 ZPOOL_CONFIG_SPLIT_GUID
, &split_guid
) == 0) {
424 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_SPLIT_GUID
,
428 nvroot
= vdev_config_generate(spa
, vd
, getstats
, 0);
429 VERIFY(nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, nvroot
) == 0);
433 * Store what's necessary for reading the MOS in the label.
435 VERIFY(nvlist_add_nvlist(config
, ZPOOL_CONFIG_FEATURES_FOR_READ
,
436 spa
->spa_label_features
) == 0);
438 if (getstats
&& spa_load_state(spa
) == SPA_LOAD_NONE
) {
439 ddt_histogram_t
*ddh
;
443 ddh
= kmem_zalloc(sizeof (ddt_histogram_t
), KM_PUSHPAGE
);
444 ddt_get_dedup_histogram(spa
, ddh
);
445 VERIFY(nvlist_add_uint64_array(config
,
446 ZPOOL_CONFIG_DDT_HISTOGRAM
,
447 (uint64_t *)ddh
, sizeof (*ddh
) / sizeof (uint64_t)) == 0);
448 kmem_free(ddh
, sizeof (ddt_histogram_t
));
450 ddo
= kmem_zalloc(sizeof (ddt_object_t
), KM_PUSHPAGE
);
451 ddt_get_dedup_object_stats(spa
, ddo
);
452 VERIFY(nvlist_add_uint64_array(config
,
453 ZPOOL_CONFIG_DDT_OBJ_STATS
,
454 (uint64_t *)ddo
, sizeof (*ddo
) / sizeof (uint64_t)) == 0);
455 kmem_free(ddo
, sizeof (ddt_object_t
));
457 dds
= kmem_zalloc(sizeof (ddt_stat_t
), KM_PUSHPAGE
);
458 ddt_get_dedup_stats(spa
, dds
);
459 VERIFY(nvlist_add_uint64_array(config
,
460 ZPOOL_CONFIG_DDT_STATS
,
461 (uint64_t *)dds
, sizeof (*dds
) / sizeof (uint64_t)) == 0);
462 kmem_free(dds
, sizeof (ddt_stat_t
));
466 spa_config_exit(spa
, SCL_CONFIG
| SCL_STATE
, FTAG
);
472 * Update all disk labels, generate a fresh config based on the current
473 * in-core state, and sync the global config cache (do not sync the config
474 * cache if this is a booting rootpool).
477 spa_config_update(spa_t
*spa
, int what
)
479 vdev_t
*rvd
= spa
->spa_root_vdev
;
483 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
485 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
486 txg
= spa_last_synced_txg(spa
) + 1;
487 if (what
== SPA_CONFIG_UPDATE_POOL
) {
488 vdev_config_dirty(rvd
);
491 * If we have top-level vdevs that were added but have
492 * not yet been prepared for allocation, do that now.
493 * (It's safe now because the config cache is up to date,
494 * so it will be able to translate the new DVAs.)
495 * See comments in spa_vdev_add() for full details.
497 for (c
= 0; c
< rvd
->vdev_children
; c
++) {
498 vdev_t
*tvd
= rvd
->vdev_child
[c
];
499 if (tvd
->vdev_ms_array
== 0)
500 vdev_metaslab_set_size(tvd
);
501 vdev_expand(tvd
, txg
);
504 spa_config_exit(spa
, SCL_ALL
, FTAG
);
507 * Wait for the mosconfig to be regenerated and synced.
509 txg_wait_synced(spa
->spa_dsl_pool
, txg
);
512 * Update the global config cache to reflect the new mosconfig.
514 if (!spa
->spa_is_root
)
515 spa_config_sync(spa
, B_FALSE
, what
!= SPA_CONFIG_UPDATE_POOL
);
517 if (what
== SPA_CONFIG_UPDATE_POOL
)
518 spa_config_update(spa
, SPA_CONFIG_UPDATE_VDEVS
);
521 #if defined(_KERNEL) && defined(HAVE_SPL)
522 EXPORT_SYMBOL(spa_config_sync
);
523 EXPORT_SYMBOL(spa_config_load
);
524 EXPORT_SYMBOL(spa_all_configs
);
525 EXPORT_SYMBOL(spa_config_set
);
526 EXPORT_SYMBOL(spa_config_generate
);
527 EXPORT_SYMBOL(spa_config_update
);
529 module_param(spa_config_path
, charp
, 0444);
530 MODULE_PARM_DESC(spa_config_path
, "SPA config file (/etc/zfs/zpool.cache)");
532 module_param(zfs_autoimport_disable
, int, 0644);
533 MODULE_PARM_DESC(zfs_autoimport_disable
, "Disable pool import at module load");