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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
27 #include <sys/zfs_context.h>
28 #include <sys/spa_impl.h>
30 #include <sys/zio_checksum.h>
31 #include <sys/zio_compress.h>
33 #include <sys/dmu_tx.h>
36 #include <sys/vdev_impl.h>
37 #include <sys/metaslab.h>
38 #include <sys/uberblock_impl.h>
41 #include <sys/unique.h>
42 #include <sys/dsl_pool.h>
43 #include <sys/dsl_dir.h>
44 #include <sys/dsl_prop.h>
45 #include <sys/fm/util.h>
46 #include <sys/dsl_scan.h>
47 #include <sys/fs/zfs.h>
48 #include <sys/metaslab_impl.h>
51 #include <sys/kstat.h>
53 #include "zfeature_common.h"
58 * There are four basic locks for managing spa_t structures:
60 * spa_namespace_lock (global mutex)
62 * This lock must be acquired to do any of the following:
64 * - Lookup a spa_t by name
65 * - Add or remove a spa_t from the namespace
66 * - Increase spa_refcount from non-zero
67 * - Check if spa_refcount is zero
69 * - add/remove/attach/detach devices
70 * - Held for the duration of create/destroy/import/export
72 * It does not need to handle recursion. A create or destroy may
73 * reference objects (files or zvols) in other pools, but by
74 * definition they must have an existing reference, and will never need
75 * to lookup a spa_t by name.
77 * spa_refcount (per-spa refcount_t protected by mutex)
79 * This reference count keep track of any active users of the spa_t. The
80 * spa_t cannot be destroyed or freed while this is non-zero. Internally,
81 * the refcount is never really 'zero' - opening a pool implicitly keeps
82 * some references in the DMU. Internally we check against spa_minref, but
83 * present the image of a zero/non-zero value to consumers.
85 * spa_config_lock[] (per-spa array of rwlocks)
87 * This protects the spa_t from config changes, and must be held in
88 * the following circumstances:
90 * - RW_READER to perform I/O to the spa
91 * - RW_WRITER to change the vdev config
93 * The locking order is fairly straightforward:
95 * spa_namespace_lock -> spa_refcount
97 * The namespace lock must be acquired to increase the refcount from 0
98 * or to check if it is zero.
100 * spa_refcount -> spa_config_lock[]
102 * There must be at least one valid reference on the spa_t to acquire
105 * spa_namespace_lock -> spa_config_lock[]
107 * The namespace lock must always be taken before the config lock.
110 * The spa_namespace_lock can be acquired directly and is globally visible.
112 * The namespace is manipulated using the following functions, all of which
113 * require the spa_namespace_lock to be held.
115 * spa_lookup() Lookup a spa_t by name.
117 * spa_add() Create a new spa_t in the namespace.
119 * spa_remove() Remove a spa_t from the namespace. This also
120 * frees up any memory associated with the spa_t.
122 * spa_next() Returns the next spa_t in the system, or the
123 * first if NULL is passed.
125 * spa_evict_all() Shutdown and remove all spa_t structures in
128 * spa_guid_exists() Determine whether a pool/device guid exists.
130 * The spa_refcount is manipulated using the following functions:
132 * spa_open_ref() Adds a reference to the given spa_t. Must be
133 * called with spa_namespace_lock held if the
134 * refcount is currently zero.
136 * spa_close() Remove a reference from the spa_t. This will
137 * not free the spa_t or remove it from the
138 * namespace. No locking is required.
140 * spa_refcount_zero() Returns true if the refcount is currently
141 * zero. Must be called with spa_namespace_lock
144 * The spa_config_lock[] is an array of rwlocks, ordered as follows:
145 * SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
146 * spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
148 * To read the configuration, it suffices to hold one of these locks as reader.
149 * To modify the configuration, you must hold all locks as writer. To modify
150 * vdev state without altering the vdev tree's topology (e.g. online/offline),
151 * you must hold SCL_STATE and SCL_ZIO as writer.
153 * We use these distinct config locks to avoid recursive lock entry.
154 * For example, spa_sync() (which holds SCL_CONFIG as reader) induces
155 * block allocations (SCL_ALLOC), which may require reading space maps
156 * from disk (dmu_read() -> zio_read() -> SCL_ZIO).
158 * The spa config locks cannot be normal rwlocks because we need the
159 * ability to hand off ownership. For example, SCL_ZIO is acquired
160 * by the issuing thread and later released by an interrupt thread.
161 * They do, however, obey the usual write-wanted semantics to prevent
162 * writer (i.e. system administrator) starvation.
164 * The lock acquisition rules are as follows:
167 * Protects changes to the vdev tree topology, such as vdev
168 * add/remove/attach/detach. Protects the dirty config list
169 * (spa_config_dirty_list) and the set of spares and l2arc devices.
172 * Protects changes to pool state and vdev state, such as vdev
173 * online/offline/fault/degrade/clear. Protects the dirty state list
174 * (spa_state_dirty_list) and global pool state (spa_state).
177 * Protects changes to metaslab groups and classes.
178 * Held as reader by metaslab_alloc() and metaslab_claim().
181 * Held by bp-level zios (those which have no io_vd upon entry)
182 * to prevent changes to the vdev tree. The bp-level zio implicitly
183 * protects all of its vdev child zios, which do not hold SCL_ZIO.
186 * Protects changes to metaslab groups and classes.
187 * Held as reader by metaslab_free(). SCL_FREE is distinct from
188 * SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
189 * blocks in zio_done() while another i/o that holds either
190 * SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
193 * Held as reader to prevent changes to the vdev tree during trivial
194 * inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the
195 * other locks, and lower than all of them, to ensure that it's safe
196 * to acquire regardless of caller context.
198 * In addition, the following rules apply:
200 * (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
201 * The lock ordering is SCL_CONFIG > spa_props_lock.
203 * (b) I/O operations on leaf vdevs. For any zio operation that takes
204 * an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
205 * or zio_write_phys() -- the caller must ensure that the config cannot
206 * cannot change in the interim, and that the vdev cannot be reopened.
207 * SCL_STATE as reader suffices for both.
209 * The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
211 * spa_vdev_enter() Acquire the namespace lock and the config lock
214 * spa_vdev_exit() Release the config lock, wait for all I/O
215 * to complete, sync the updated configs to the
216 * cache, and release the namespace lock.
218 * vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
219 * Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
220 * locking is, always, based on spa_namespace_lock and spa_config_lock[].
222 * spa_rename() is also implemented within this file since it requires
223 * manipulation of the namespace.
226 static avl_tree_t spa_namespace_avl
;
227 kmutex_t spa_namespace_lock
;
228 static kcondvar_t spa_namespace_cv
;
229 static int spa_active_count
;
230 int spa_max_replication_override
= SPA_DVAS_PER_BP
;
232 static kmutex_t spa_spare_lock
;
233 static avl_tree_t spa_spare_avl
;
234 static kmutex_t spa_l2cache_lock
;
235 static avl_tree_t spa_l2cache_avl
;
237 kmem_cache_t
*spa_buffer_pool
;
241 * Expiration time in units of zfs_txg_synctime_ms. This value has two
242 * meanings. First it is used to determine when the spa_deadman logic
243 * should fire. By default the spa_deadman will fire if spa_sync has
244 * not completed in 1000 * zfs_txg_synctime_ms (i.e. 1000 seconds).
245 * Secondly, the value determines if an I/O is considered "hung".
246 * Any I/O that has not completed in zfs_deadman_synctime is considered
247 * "hung" resulting in a zevent being posted.
248 * 1000 zfs_txg_synctime_ms (i.e. 1000 seconds).
250 unsigned long zfs_deadman_synctime
= 1000ULL;
253 * By default the deadman is enabled.
255 int zfs_deadman_enabled
= 1;
258 * ==========================================================================
260 * ==========================================================================
263 spa_config_lock_init(spa_t
*spa
)
267 for (i
= 0; i
< SCL_LOCKS
; i
++) {
268 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
269 mutex_init(&scl
->scl_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
270 cv_init(&scl
->scl_cv
, NULL
, CV_DEFAULT
, NULL
);
271 refcount_create_untracked(&scl
->scl_count
);
272 scl
->scl_writer
= NULL
;
273 scl
->scl_write_wanted
= 0;
278 spa_config_lock_destroy(spa_t
*spa
)
282 for (i
= 0; i
< SCL_LOCKS
; i
++) {
283 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
284 mutex_destroy(&scl
->scl_lock
);
285 cv_destroy(&scl
->scl_cv
);
286 refcount_destroy(&scl
->scl_count
);
287 ASSERT(scl
->scl_writer
== NULL
);
288 ASSERT(scl
->scl_write_wanted
== 0);
293 spa_config_tryenter(spa_t
*spa
, int locks
, void *tag
, krw_t rw
)
297 for (i
= 0; i
< SCL_LOCKS
; i
++) {
298 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
299 if (!(locks
& (1 << i
)))
301 mutex_enter(&scl
->scl_lock
);
302 if (rw
== RW_READER
) {
303 if (scl
->scl_writer
|| scl
->scl_write_wanted
) {
304 mutex_exit(&scl
->scl_lock
);
305 spa_config_exit(spa
, locks
^ (1 << i
), tag
);
309 ASSERT(scl
->scl_writer
!= curthread
);
310 if (!refcount_is_zero(&scl
->scl_count
)) {
311 mutex_exit(&scl
->scl_lock
);
312 spa_config_exit(spa
, locks
^ (1 << i
), tag
);
315 scl
->scl_writer
= curthread
;
317 (void) refcount_add(&scl
->scl_count
, tag
);
318 mutex_exit(&scl
->scl_lock
);
324 spa_config_enter(spa_t
*spa
, int locks
, void *tag
, krw_t rw
)
329 ASSERT3U(SCL_LOCKS
, <, sizeof (wlocks_held
) * NBBY
);
331 for (i
= 0; i
< SCL_LOCKS
; i
++) {
332 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
333 if (scl
->scl_writer
== curthread
)
334 wlocks_held
|= (1 << i
);
335 if (!(locks
& (1 << i
)))
337 mutex_enter(&scl
->scl_lock
);
338 if (rw
== RW_READER
) {
339 while (scl
->scl_writer
|| scl
->scl_write_wanted
) {
340 cv_wait(&scl
->scl_cv
, &scl
->scl_lock
);
343 ASSERT(scl
->scl_writer
!= curthread
);
344 while (!refcount_is_zero(&scl
->scl_count
)) {
345 scl
->scl_write_wanted
++;
346 cv_wait(&scl
->scl_cv
, &scl
->scl_lock
);
347 scl
->scl_write_wanted
--;
349 scl
->scl_writer
= curthread
;
351 (void) refcount_add(&scl
->scl_count
, tag
);
352 mutex_exit(&scl
->scl_lock
);
354 ASSERT(wlocks_held
<= locks
);
358 spa_config_exit(spa_t
*spa
, int locks
, void *tag
)
362 for (i
= SCL_LOCKS
- 1; i
>= 0; i
--) {
363 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
364 if (!(locks
& (1 << i
)))
366 mutex_enter(&scl
->scl_lock
);
367 ASSERT(!refcount_is_zero(&scl
->scl_count
));
368 if (refcount_remove(&scl
->scl_count
, tag
) == 0) {
369 ASSERT(scl
->scl_writer
== NULL
||
370 scl
->scl_writer
== curthread
);
371 scl
->scl_writer
= NULL
; /* OK in either case */
372 cv_broadcast(&scl
->scl_cv
);
374 mutex_exit(&scl
->scl_lock
);
379 spa_config_held(spa_t
*spa
, int locks
, krw_t rw
)
381 int i
, locks_held
= 0;
383 for (i
= 0; i
< SCL_LOCKS
; i
++) {
384 spa_config_lock_t
*scl
= &spa
->spa_config_lock
[i
];
385 if (!(locks
& (1 << i
)))
387 if ((rw
== RW_READER
&& !refcount_is_zero(&scl
->scl_count
)) ||
388 (rw
== RW_WRITER
&& scl
->scl_writer
== curthread
))
389 locks_held
|= 1 << i
;
396 * ==========================================================================
397 * SPA namespace functions
398 * ==========================================================================
402 * Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
403 * Returns NULL if no matching spa_t is found.
406 spa_lookup(const char *name
)
408 static spa_t search
; /* spa_t is large; don't allocate on stack */
413 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
415 (void) strlcpy(search
.spa_name
, name
, sizeof (search
.spa_name
));
418 * If it's a full dataset name, figure out the pool name and
421 cp
= strpbrk(search
.spa_name
, "/@");
425 spa
= avl_find(&spa_namespace_avl
, &search
, &where
);
431 * Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
432 * If the zfs_deadman_enabled flag is set then it inspects all vdev queues
433 * looking for potentially hung I/Os.
436 spa_deadman(void *arg
)
440 zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
441 (gethrtime() - spa
->spa_sync_starttime
) / NANOSEC
,
442 ++spa
->spa_deadman_calls
);
443 if (zfs_deadman_enabled
)
444 vdev_deadman(spa
->spa_root_vdev
);
446 spa
->spa_deadman_tqid
= taskq_dispatch_delay(system_taskq
,
447 spa_deadman
, spa
, TQ_PUSHPAGE
, ddi_get_lbolt() +
448 NSEC_TO_TICK(spa
->spa_deadman_synctime
));
452 * Create an uninitialized spa_t with the given name. Requires
453 * spa_namespace_lock. The caller must ensure that the spa_t doesn't already
454 * exist by calling spa_lookup() first.
457 spa_add(const char *name
, nvlist_t
*config
, const char *altroot
)
460 spa_config_dirent_t
*dp
;
463 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
465 spa
= kmem_zalloc(sizeof (spa_t
), KM_PUSHPAGE
| KM_NODEBUG
);
467 mutex_init(&spa
->spa_async_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
468 mutex_init(&spa
->spa_errlist_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
469 mutex_init(&spa
->spa_errlog_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
470 mutex_init(&spa
->spa_history_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
471 mutex_init(&spa
->spa_proc_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
472 mutex_init(&spa
->spa_props_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
473 mutex_init(&spa
->spa_scrub_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
474 mutex_init(&spa
->spa_suspend_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
475 mutex_init(&spa
->spa_vdev_top_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
477 cv_init(&spa
->spa_async_cv
, NULL
, CV_DEFAULT
, NULL
);
478 cv_init(&spa
->spa_proc_cv
, NULL
, CV_DEFAULT
, NULL
);
479 cv_init(&spa
->spa_scrub_io_cv
, NULL
, CV_DEFAULT
, NULL
);
480 cv_init(&spa
->spa_suspend_cv
, NULL
, CV_DEFAULT
, NULL
);
482 for (t
= 0; t
< TXG_SIZE
; t
++)
483 bplist_create(&spa
->spa_free_bplist
[t
]);
485 (void) strlcpy(spa
->spa_name
, name
, sizeof (spa
->spa_name
));
486 spa
->spa_state
= POOL_STATE_UNINITIALIZED
;
487 spa
->spa_freeze_txg
= UINT64_MAX
;
488 spa
->spa_final_txg
= UINT64_MAX
;
489 spa
->spa_load_max_txg
= UINT64_MAX
;
491 spa
->spa_proc_state
= SPA_PROC_NONE
;
493 spa
->spa_deadman_synctime
= zfs_deadman_synctime
*
494 zfs_txg_synctime_ms
* MICROSEC
;
496 refcount_create(&spa
->spa_refcount
);
497 spa_config_lock_init(spa
);
500 avl_add(&spa_namespace_avl
, spa
);
503 * Set the alternate root, if there is one.
506 spa
->spa_root
= spa_strdup(altroot
);
511 * Every pool starts with the default cachefile
513 list_create(&spa
->spa_config_list
, sizeof (spa_config_dirent_t
),
514 offsetof(spa_config_dirent_t
, scd_link
));
516 dp
= kmem_zalloc(sizeof (spa_config_dirent_t
), KM_PUSHPAGE
);
517 dp
->scd_path
= altroot
? NULL
: spa_strdup(spa_config_path
);
518 list_insert_head(&spa
->spa_config_list
, dp
);
520 VERIFY(nvlist_alloc(&spa
->spa_load_info
, NV_UNIQUE_NAME
,
523 if (config
!= NULL
) {
526 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_FEATURES_FOR_READ
,
528 VERIFY(nvlist_dup(features
, &spa
->spa_label_features
,
532 VERIFY(nvlist_dup(config
, &spa
->spa_config
, 0) == 0);
535 if (spa
->spa_label_features
== NULL
) {
536 VERIFY(nvlist_alloc(&spa
->spa_label_features
, NV_UNIQUE_NAME
,
540 spa
->spa_debug
= ((zfs_flags
& ZFS_DEBUG_SPA
) != 0);
546 * Removes a spa_t from the namespace, freeing up any memory used. Requires
547 * spa_namespace_lock. This is called only after the spa_t has been closed and
551 spa_remove(spa_t
*spa
)
553 spa_config_dirent_t
*dp
;
556 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
557 ASSERT(spa
->spa_state
== POOL_STATE_UNINITIALIZED
);
559 nvlist_free(spa
->spa_config_splitting
);
561 avl_remove(&spa_namespace_avl
, spa
);
562 cv_broadcast(&spa_namespace_cv
);
565 spa_strfree(spa
->spa_root
);
569 while ((dp
= list_head(&spa
->spa_config_list
)) != NULL
) {
570 list_remove(&spa
->spa_config_list
, dp
);
571 if (dp
->scd_path
!= NULL
)
572 spa_strfree(dp
->scd_path
);
573 kmem_free(dp
, sizeof (spa_config_dirent_t
));
576 list_destroy(&spa
->spa_config_list
);
578 nvlist_free(spa
->spa_label_features
);
579 nvlist_free(spa
->spa_load_info
);
580 spa_config_set(spa
, NULL
);
582 refcount_destroy(&spa
->spa_refcount
);
584 spa_stats_destroy(spa
);
585 spa_config_lock_destroy(spa
);
587 for (t
= 0; t
< TXG_SIZE
; t
++)
588 bplist_destroy(&spa
->spa_free_bplist
[t
]);
590 cv_destroy(&spa
->spa_async_cv
);
591 cv_destroy(&spa
->spa_proc_cv
);
592 cv_destroy(&spa
->spa_scrub_io_cv
);
593 cv_destroy(&spa
->spa_suspend_cv
);
595 mutex_destroy(&spa
->spa_async_lock
);
596 mutex_destroy(&spa
->spa_errlist_lock
);
597 mutex_destroy(&spa
->spa_errlog_lock
);
598 mutex_destroy(&spa
->spa_history_lock
);
599 mutex_destroy(&spa
->spa_proc_lock
);
600 mutex_destroy(&spa
->spa_props_lock
);
601 mutex_destroy(&spa
->spa_scrub_lock
);
602 mutex_destroy(&spa
->spa_suspend_lock
);
603 mutex_destroy(&spa
->spa_vdev_top_lock
);
605 kmem_free(spa
, sizeof (spa_t
));
609 * Given a pool, return the next pool in the namespace, or NULL if there is
610 * none. If 'prev' is NULL, return the first pool.
613 spa_next(spa_t
*prev
)
615 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
618 return (AVL_NEXT(&spa_namespace_avl
, prev
));
620 return (avl_first(&spa_namespace_avl
));
624 * ==========================================================================
625 * SPA refcount functions
626 * ==========================================================================
630 * Add a reference to the given spa_t. Must have at least one reference, or
631 * have the namespace lock held.
634 spa_open_ref(spa_t
*spa
, void *tag
)
636 ASSERT(refcount_count(&spa
->spa_refcount
) >= spa
->spa_minref
||
637 MUTEX_HELD(&spa_namespace_lock
));
638 (void) refcount_add(&spa
->spa_refcount
, tag
);
642 * Remove a reference to the given spa_t. Must have at least one reference, or
643 * have the namespace lock held.
646 spa_close(spa_t
*spa
, void *tag
)
648 ASSERT(refcount_count(&spa
->spa_refcount
) > spa
->spa_minref
||
649 MUTEX_HELD(&spa_namespace_lock
));
650 (void) refcount_remove(&spa
->spa_refcount
, tag
);
654 * Check to see if the spa refcount is zero. Must be called with
655 * spa_namespace_lock held. We really compare against spa_minref, which is the
656 * number of references acquired when opening a pool
659 spa_refcount_zero(spa_t
*spa
)
661 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
663 return (refcount_count(&spa
->spa_refcount
) == spa
->spa_minref
);
667 * ==========================================================================
668 * SPA spare and l2cache tracking
669 * ==========================================================================
673 * Hot spares and cache devices are tracked using the same code below,
674 * for 'auxiliary' devices.
677 typedef struct spa_aux
{
685 spa_aux_compare(const void *a
, const void *b
)
687 const spa_aux_t
*sa
= a
;
688 const spa_aux_t
*sb
= b
;
690 if (sa
->aux_guid
< sb
->aux_guid
)
692 else if (sa
->aux_guid
> sb
->aux_guid
)
699 spa_aux_add(vdev_t
*vd
, avl_tree_t
*avl
)
705 search
.aux_guid
= vd
->vdev_guid
;
706 if ((aux
= avl_find(avl
, &search
, &where
)) != NULL
) {
709 aux
= kmem_zalloc(sizeof (spa_aux_t
), KM_PUSHPAGE
);
710 aux
->aux_guid
= vd
->vdev_guid
;
712 avl_insert(avl
, aux
, where
);
717 spa_aux_remove(vdev_t
*vd
, avl_tree_t
*avl
)
723 search
.aux_guid
= vd
->vdev_guid
;
724 aux
= avl_find(avl
, &search
, &where
);
728 if (--aux
->aux_count
== 0) {
729 avl_remove(avl
, aux
);
730 kmem_free(aux
, sizeof (spa_aux_t
));
731 } else if (aux
->aux_pool
== spa_guid(vd
->vdev_spa
)) {
732 aux
->aux_pool
= 0ULL;
737 spa_aux_exists(uint64_t guid
, uint64_t *pool
, int *refcnt
, avl_tree_t
*avl
)
739 spa_aux_t search
, *found
;
741 search
.aux_guid
= guid
;
742 found
= avl_find(avl
, &search
, NULL
);
746 *pool
= found
->aux_pool
;
753 *refcnt
= found
->aux_count
;
758 return (found
!= NULL
);
762 spa_aux_activate(vdev_t
*vd
, avl_tree_t
*avl
)
764 spa_aux_t search
, *found
;
767 search
.aux_guid
= vd
->vdev_guid
;
768 found
= avl_find(avl
, &search
, &where
);
769 ASSERT(found
!= NULL
);
770 ASSERT(found
->aux_pool
== 0ULL);
772 found
->aux_pool
= spa_guid(vd
->vdev_spa
);
776 * Spares are tracked globally due to the following constraints:
778 * - A spare may be part of multiple pools.
779 * - A spare may be added to a pool even if it's actively in use within
781 * - A spare in use in any pool can only be the source of a replacement if
782 * the target is a spare in the same pool.
784 * We keep track of all spares on the system through the use of a reference
785 * counted AVL tree. When a vdev is added as a spare, or used as a replacement
786 * spare, then we bump the reference count in the AVL tree. In addition, we set
787 * the 'vdev_isspare' member to indicate that the device is a spare (active or
788 * inactive). When a spare is made active (used to replace a device in the
789 * pool), we also keep track of which pool its been made a part of.
791 * The 'spa_spare_lock' protects the AVL tree. These functions are normally
792 * called under the spa_namespace lock as part of vdev reconfiguration. The
793 * separate spare lock exists for the status query path, which does not need to
794 * be completely consistent with respect to other vdev configuration changes.
798 spa_spare_compare(const void *a
, const void *b
)
800 return (spa_aux_compare(a
, b
));
804 spa_spare_add(vdev_t
*vd
)
806 mutex_enter(&spa_spare_lock
);
807 ASSERT(!vd
->vdev_isspare
);
808 spa_aux_add(vd
, &spa_spare_avl
);
809 vd
->vdev_isspare
= B_TRUE
;
810 mutex_exit(&spa_spare_lock
);
814 spa_spare_remove(vdev_t
*vd
)
816 mutex_enter(&spa_spare_lock
);
817 ASSERT(vd
->vdev_isspare
);
818 spa_aux_remove(vd
, &spa_spare_avl
);
819 vd
->vdev_isspare
= B_FALSE
;
820 mutex_exit(&spa_spare_lock
);
824 spa_spare_exists(uint64_t guid
, uint64_t *pool
, int *refcnt
)
828 mutex_enter(&spa_spare_lock
);
829 found
= spa_aux_exists(guid
, pool
, refcnt
, &spa_spare_avl
);
830 mutex_exit(&spa_spare_lock
);
836 spa_spare_activate(vdev_t
*vd
)
838 mutex_enter(&spa_spare_lock
);
839 ASSERT(vd
->vdev_isspare
);
840 spa_aux_activate(vd
, &spa_spare_avl
);
841 mutex_exit(&spa_spare_lock
);
845 * Level 2 ARC devices are tracked globally for the same reasons as spares.
846 * Cache devices currently only support one pool per cache device, and so
847 * for these devices the aux reference count is currently unused beyond 1.
851 spa_l2cache_compare(const void *a
, const void *b
)
853 return (spa_aux_compare(a
, b
));
857 spa_l2cache_add(vdev_t
*vd
)
859 mutex_enter(&spa_l2cache_lock
);
860 ASSERT(!vd
->vdev_isl2cache
);
861 spa_aux_add(vd
, &spa_l2cache_avl
);
862 vd
->vdev_isl2cache
= B_TRUE
;
863 mutex_exit(&spa_l2cache_lock
);
867 spa_l2cache_remove(vdev_t
*vd
)
869 mutex_enter(&spa_l2cache_lock
);
870 ASSERT(vd
->vdev_isl2cache
);
871 spa_aux_remove(vd
, &spa_l2cache_avl
);
872 vd
->vdev_isl2cache
= B_FALSE
;
873 mutex_exit(&spa_l2cache_lock
);
877 spa_l2cache_exists(uint64_t guid
, uint64_t *pool
)
881 mutex_enter(&spa_l2cache_lock
);
882 found
= spa_aux_exists(guid
, pool
, NULL
, &spa_l2cache_avl
);
883 mutex_exit(&spa_l2cache_lock
);
889 spa_l2cache_activate(vdev_t
*vd
)
891 mutex_enter(&spa_l2cache_lock
);
892 ASSERT(vd
->vdev_isl2cache
);
893 spa_aux_activate(vd
, &spa_l2cache_avl
);
894 mutex_exit(&spa_l2cache_lock
);
898 * ==========================================================================
900 * ==========================================================================
904 * Lock the given spa_t for the purpose of adding or removing a vdev.
905 * Grabs the global spa_namespace_lock plus the spa config lock for writing.
906 * It returns the next transaction group for the spa_t.
909 spa_vdev_enter(spa_t
*spa
)
911 mutex_enter(&spa
->spa_vdev_top_lock
);
912 mutex_enter(&spa_namespace_lock
);
913 return (spa_vdev_config_enter(spa
));
917 * Internal implementation for spa_vdev_enter(). Used when a vdev
918 * operation requires multiple syncs (i.e. removing a device) while
919 * keeping the spa_namespace_lock held.
922 spa_vdev_config_enter(spa_t
*spa
)
924 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
926 spa_config_enter(spa
, SCL_ALL
, spa
, RW_WRITER
);
928 return (spa_last_synced_txg(spa
) + 1);
932 * Used in combination with spa_vdev_config_enter() to allow the syncing
933 * of multiple transactions without releasing the spa_namespace_lock.
936 spa_vdev_config_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
, int error
, char *tag
)
938 int config_changed
= B_FALSE
;
940 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
941 ASSERT(txg
> spa_last_synced_txg(spa
));
943 spa
->spa_pending_vdev
= NULL
;
948 vdev_dtl_reassess(spa
->spa_root_vdev
, 0, 0, B_FALSE
);
950 if (error
== 0 && !list_is_empty(&spa
->spa_config_dirty_list
)) {
951 config_changed
= B_TRUE
;
952 spa
->spa_config_generation
++;
956 * Verify the metaslab classes.
958 ASSERT(metaslab_class_validate(spa_normal_class(spa
)) == 0);
959 ASSERT(metaslab_class_validate(spa_log_class(spa
)) == 0);
961 spa_config_exit(spa
, SCL_ALL
, spa
);
964 * Panic the system if the specified tag requires it. This
965 * is useful for ensuring that configurations are updated
968 if (zio_injection_enabled
)
969 zio_handle_panic_injection(spa
, tag
, 0);
972 * Note: this txg_wait_synced() is important because it ensures
973 * that there won't be more than one config change per txg.
974 * This allows us to use the txg as the generation number.
977 txg_wait_synced(spa
->spa_dsl_pool
, txg
);
980 ASSERT(!vd
->vdev_detached
|| vd
->vdev_dtl_smo
.smo_object
== 0);
981 spa_config_enter(spa
, SCL_ALL
, spa
, RW_WRITER
);
983 spa_config_exit(spa
, SCL_ALL
, spa
);
987 * If the config changed, update the config cache.
990 spa_config_sync(spa
, B_FALSE
, B_TRUE
);
994 * Unlock the spa_t after adding or removing a vdev. Besides undoing the
995 * locking of spa_vdev_enter(), we also want make sure the transactions have
996 * synced to disk, and then update the global configuration cache with the new
1000 spa_vdev_exit(spa_t
*spa
, vdev_t
*vd
, uint64_t txg
, int error
)
1002 spa_vdev_config_exit(spa
, vd
, txg
, error
, FTAG
);
1003 mutex_exit(&spa_namespace_lock
);
1004 mutex_exit(&spa
->spa_vdev_top_lock
);
1010 * Lock the given spa_t for the purpose of changing vdev state.
1013 spa_vdev_state_enter(spa_t
*spa
, int oplocks
)
1015 int locks
= SCL_STATE_ALL
| oplocks
;
1018 * Root pools may need to read of the underlying devfs filesystem
1019 * when opening up a vdev. Unfortunately if we're holding the
1020 * SCL_ZIO lock it will result in a deadlock when we try to issue
1021 * the read from the root filesystem. Instead we "prefetch"
1022 * the associated vnodes that we need prior to opening the
1023 * underlying devices and cache them so that we can prevent
1024 * any I/O when we are doing the actual open.
1026 if (spa_is_root(spa
)) {
1027 int low
= locks
& ~(SCL_ZIO
- 1);
1028 int high
= locks
& ~low
;
1030 spa_config_enter(spa
, high
, spa
, RW_WRITER
);
1031 vdev_hold(spa
->spa_root_vdev
);
1032 spa_config_enter(spa
, low
, spa
, RW_WRITER
);
1034 spa_config_enter(spa
, locks
, spa
, RW_WRITER
);
1036 spa
->spa_vdev_locks
= locks
;
1040 spa_vdev_state_exit(spa_t
*spa
, vdev_t
*vd
, int error
)
1042 boolean_t config_changed
= B_FALSE
;
1044 if (vd
!= NULL
|| error
== 0)
1045 vdev_dtl_reassess(vd
? vd
->vdev_top
: spa
->spa_root_vdev
,
1049 vdev_state_dirty(vd
->vdev_top
);
1050 config_changed
= B_TRUE
;
1051 spa
->spa_config_generation
++;
1054 if (spa_is_root(spa
))
1055 vdev_rele(spa
->spa_root_vdev
);
1057 ASSERT3U(spa
->spa_vdev_locks
, >=, SCL_STATE_ALL
);
1058 spa_config_exit(spa
, spa
->spa_vdev_locks
, spa
);
1061 * If anything changed, wait for it to sync. This ensures that,
1062 * from the system administrator's perspective, zpool(1M) commands
1063 * are synchronous. This is important for things like zpool offline:
1064 * when the command completes, you expect no further I/O from ZFS.
1067 txg_wait_synced(spa
->spa_dsl_pool
, 0);
1070 * If the config changed, update the config cache.
1072 if (config_changed
) {
1073 mutex_enter(&spa_namespace_lock
);
1074 spa_config_sync(spa
, B_FALSE
, B_TRUE
);
1075 mutex_exit(&spa_namespace_lock
);
1082 * ==========================================================================
1083 * Miscellaneous functions
1084 * ==========================================================================
1088 spa_activate_mos_feature(spa_t
*spa
, const char *feature
)
1090 (void) nvlist_add_boolean(spa
->spa_label_features
, feature
);
1091 vdev_config_dirty(spa
->spa_root_vdev
);
1095 spa_deactivate_mos_feature(spa_t
*spa
, const char *feature
)
1097 (void) nvlist_remove_all(spa
->spa_label_features
, feature
);
1098 vdev_config_dirty(spa
->spa_root_vdev
);
1105 spa_rename(const char *name
, const char *newname
)
1111 * Lookup the spa_t and grab the config lock for writing. We need to
1112 * actually open the pool so that we can sync out the necessary labels.
1113 * It's OK to call spa_open() with the namespace lock held because we
1114 * allow recursive calls for other reasons.
1116 mutex_enter(&spa_namespace_lock
);
1117 if ((err
= spa_open(name
, &spa
, FTAG
)) != 0) {
1118 mutex_exit(&spa_namespace_lock
);
1122 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1124 avl_remove(&spa_namespace_avl
, spa
);
1125 (void) strlcpy(spa
->spa_name
, newname
, sizeof (spa
->spa_name
));
1126 avl_add(&spa_namespace_avl
, spa
);
1129 * Sync all labels to disk with the new names by marking the root vdev
1130 * dirty and waiting for it to sync. It will pick up the new pool name
1133 vdev_config_dirty(spa
->spa_root_vdev
);
1135 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1137 txg_wait_synced(spa
->spa_dsl_pool
, 0);
1140 * Sync the updated config cache.
1142 spa_config_sync(spa
, B_FALSE
, B_TRUE
);
1144 spa_close(spa
, FTAG
);
1146 mutex_exit(&spa_namespace_lock
);
1152 * Return the spa_t associated with given pool_guid, if it exists. If
1153 * device_guid is non-zero, determine whether the pool exists *and* contains
1154 * a device with the specified device_guid.
1157 spa_by_guid(uint64_t pool_guid
, uint64_t device_guid
)
1160 avl_tree_t
*t
= &spa_namespace_avl
;
1162 ASSERT(MUTEX_HELD(&spa_namespace_lock
));
1164 for (spa
= avl_first(t
); spa
!= NULL
; spa
= AVL_NEXT(t
, spa
)) {
1165 if (spa
->spa_state
== POOL_STATE_UNINITIALIZED
)
1167 if (spa
->spa_root_vdev
== NULL
)
1169 if (spa_guid(spa
) == pool_guid
) {
1170 if (device_guid
== 0)
1173 if (vdev_lookup_by_guid(spa
->spa_root_vdev
,
1174 device_guid
) != NULL
)
1178 * Check any devices we may be in the process of adding.
1180 if (spa
->spa_pending_vdev
) {
1181 if (vdev_lookup_by_guid(spa
->spa_pending_vdev
,
1182 device_guid
) != NULL
)
1192 * Determine whether a pool with the given pool_guid exists.
1195 spa_guid_exists(uint64_t pool_guid
, uint64_t device_guid
)
1197 return (spa_by_guid(pool_guid
, device_guid
) != NULL
);
1201 spa_strdup(const char *s
)
1207 new = kmem_alloc(len
+ 1, KM_PUSHPAGE
);
1215 spa_strfree(char *s
)
1217 kmem_free(s
, strlen(s
) + 1);
1221 spa_get_random(uint64_t range
)
1227 (void) random_get_pseudo_bytes((void *)&r
, sizeof (uint64_t));
1233 spa_generate_guid(spa_t
*spa
)
1235 uint64_t guid
= spa_get_random(-1ULL);
1238 while (guid
== 0 || spa_guid_exists(spa_guid(spa
), guid
))
1239 guid
= spa_get_random(-1ULL);
1241 while (guid
== 0 || spa_guid_exists(guid
, 0))
1242 guid
= spa_get_random(-1ULL);
1249 sprintf_blkptr(char *buf
, const blkptr_t
*bp
)
1252 char *checksum
= NULL
;
1253 char *compress
= NULL
;
1256 if (BP_GET_TYPE(bp
) & DMU_OT_NEWTYPE
) {
1257 dmu_object_byteswap_t bswap
=
1258 DMU_OT_BYTESWAP(BP_GET_TYPE(bp
));
1259 (void) snprintf(type
, sizeof (type
), "bswap %s %s",
1260 DMU_OT_IS_METADATA(BP_GET_TYPE(bp
)) ?
1261 "metadata" : "data",
1262 dmu_ot_byteswap
[bswap
].ob_name
);
1264 (void) strlcpy(type
, dmu_ot
[BP_GET_TYPE(bp
)].ot_name
,
1267 checksum
= zio_checksum_table
[BP_GET_CHECKSUM(bp
)].ci_name
;
1268 compress
= zio_compress_table
[BP_GET_COMPRESS(bp
)].ci_name
;
1271 SPRINTF_BLKPTR(snprintf
, ' ', buf
, bp
, type
, checksum
, compress
);
1275 spa_freeze(spa_t
*spa
)
1277 uint64_t freeze_txg
= 0;
1279 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1280 if (spa
->spa_freeze_txg
== UINT64_MAX
) {
1281 freeze_txg
= spa_last_synced_txg(spa
) + TXG_SIZE
;
1282 spa
->spa_freeze_txg
= freeze_txg
;
1284 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1285 if (freeze_txg
!= 0)
1286 txg_wait_synced(spa_get_dsl(spa
), freeze_txg
);
1290 * This is a stripped-down version of strtoull, suitable only for converting
1291 * lowercase hexidecimal numbers that don't overflow.
1294 strtonum(const char *str
, char **nptr
)
1300 while ((c
= *str
) != '\0') {
1301 if (c
>= '0' && c
<= '9')
1303 else if (c
>= 'a' && c
<= 'f')
1304 digit
= 10 + c
- 'a';
1315 *nptr
= (char *)str
;
1321 * ==========================================================================
1322 * Accessor functions
1323 * ==========================================================================
1327 spa_shutting_down(spa_t
*spa
)
1329 return (spa
->spa_async_suspended
);
1333 spa_get_dsl(spa_t
*spa
)
1335 return (spa
->spa_dsl_pool
);
1339 spa_is_initializing(spa_t
*spa
)
1341 return (spa
->spa_is_initializing
);
1345 spa_get_rootblkptr(spa_t
*spa
)
1347 return (&spa
->spa_ubsync
.ub_rootbp
);
1351 spa_set_rootblkptr(spa_t
*spa
, const blkptr_t
*bp
)
1353 spa
->spa_uberblock
.ub_rootbp
= *bp
;
1357 spa_altroot(spa_t
*spa
, char *buf
, size_t buflen
)
1359 if (spa
->spa_root
== NULL
)
1362 (void) strncpy(buf
, spa
->spa_root
, buflen
);
1366 spa_sync_pass(spa_t
*spa
)
1368 return (spa
->spa_sync_pass
);
1372 spa_name(spa_t
*spa
)
1374 return (spa
->spa_name
);
1378 spa_guid(spa_t
*spa
)
1380 dsl_pool_t
*dp
= spa_get_dsl(spa
);
1384 * If we fail to parse the config during spa_load(), we can go through
1385 * the error path (which posts an ereport) and end up here with no root
1386 * vdev. We stash the original pool guid in 'spa_config_guid' to handle
1389 if (spa
->spa_root_vdev
== NULL
)
1390 return (spa
->spa_config_guid
);
1392 guid
= spa
->spa_last_synced_guid
!= 0 ?
1393 spa
->spa_last_synced_guid
: spa
->spa_root_vdev
->vdev_guid
;
1396 * Return the most recently synced out guid unless we're
1397 * in syncing context.
1399 if (dp
&& dsl_pool_sync_context(dp
))
1400 return (spa
->spa_root_vdev
->vdev_guid
);
1406 spa_load_guid(spa_t
*spa
)
1409 * This is a GUID that exists solely as a reference for the
1410 * purposes of the arc. It is generated at load time, and
1411 * is never written to persistent storage.
1413 return (spa
->spa_load_guid
);
1417 spa_last_synced_txg(spa_t
*spa
)
1419 return (spa
->spa_ubsync
.ub_txg
);
1423 spa_first_txg(spa_t
*spa
)
1425 return (spa
->spa_first_txg
);
1429 spa_syncing_txg(spa_t
*spa
)
1431 return (spa
->spa_syncing_txg
);
1435 spa_state(spa_t
*spa
)
1437 return (spa
->spa_state
);
1441 spa_load_state(spa_t
*spa
)
1443 return (spa
->spa_load_state
);
1447 spa_freeze_txg(spa_t
*spa
)
1449 return (spa
->spa_freeze_txg
);
1454 spa_get_asize(spa_t
*spa
, uint64_t lsize
)
1457 * The worst case is single-sector max-parity RAID-Z blocks, in which
1458 * case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
1459 * times the size; so just assume that. Add to this the fact that
1460 * we can have up to 3 DVAs per bp, and one more factor of 2 because
1461 * the block may be dittoed with up to 3 DVAs by ddt_sync().
1463 return (lsize
* (VDEV_RAIDZ_MAXPARITY
+ 1) * SPA_DVAS_PER_BP
* 2);
1467 spa_get_dspace(spa_t
*spa
)
1469 return (spa
->spa_dspace
);
1473 spa_update_dspace(spa_t
*spa
)
1475 spa
->spa_dspace
= metaslab_class_get_dspace(spa_normal_class(spa
)) +
1476 ddt_get_dedup_dspace(spa
);
1480 * Return the failure mode that has been set to this pool. The default
1481 * behavior will be to block all I/Os when a complete failure occurs.
1484 spa_get_failmode(spa_t
*spa
)
1486 return (spa
->spa_failmode
);
1490 spa_suspended(spa_t
*spa
)
1492 return (spa
->spa_suspended
);
1496 spa_version(spa_t
*spa
)
1498 return (spa
->spa_ubsync
.ub_version
);
1502 spa_deflate(spa_t
*spa
)
1504 return (spa
->spa_deflate
);
1508 spa_normal_class(spa_t
*spa
)
1510 return (spa
->spa_normal_class
);
1514 spa_log_class(spa_t
*spa
)
1516 return (spa
->spa_log_class
);
1520 spa_max_replication(spa_t
*spa
)
1523 * As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
1524 * handle BPs with more than one DVA allocated. Set our max
1525 * replication level accordingly.
1527 if (spa_version(spa
) < SPA_VERSION_DITTO_BLOCKS
)
1529 return (MIN(SPA_DVAS_PER_BP
, spa_max_replication_override
));
1533 spa_prev_software_version(spa_t
*spa
)
1535 return (spa
->spa_prev_software_version
);
1539 spa_deadman_synctime(spa_t
*spa
)
1541 return (spa
->spa_deadman_synctime
);
1545 dva_get_dsize_sync(spa_t
*spa
, const dva_t
*dva
)
1547 uint64_t asize
= DVA_GET_ASIZE(dva
);
1548 uint64_t dsize
= asize
;
1550 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_READER
) != 0);
1552 if (asize
!= 0 && spa
->spa_deflate
) {
1553 vdev_t
*vd
= vdev_lookup_top(spa
, DVA_GET_VDEV(dva
));
1554 dsize
= (asize
>> SPA_MINBLOCKSHIFT
) * vd
->vdev_deflate_ratio
;
1561 bp_get_dsize_sync(spa_t
*spa
, const blkptr_t
*bp
)
1566 for (d
= 0; d
< SPA_DVAS_PER_BP
; d
++)
1567 dsize
+= dva_get_dsize_sync(spa
, &bp
->blk_dva
[d
]);
1573 bp_get_dsize(spa_t
*spa
, const blkptr_t
*bp
)
1578 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
1580 for (d
= 0; d
< SPA_DVAS_PER_BP
; d
++)
1581 dsize
+= dva_get_dsize_sync(spa
, &bp
->blk_dva
[d
]);
1583 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
1589 * ==========================================================================
1590 * Initialization and Termination
1591 * ==========================================================================
1595 spa_name_compare(const void *a1
, const void *a2
)
1597 const spa_t
*s1
= a1
;
1598 const spa_t
*s2
= a2
;
1601 s
= strcmp(s1
->spa_name
, s2
->spa_name
);
1618 mutex_init(&spa_namespace_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1619 mutex_init(&spa_spare_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1620 mutex_init(&spa_l2cache_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1621 cv_init(&spa_namespace_cv
, NULL
, CV_DEFAULT
, NULL
);
1623 avl_create(&spa_namespace_avl
, spa_name_compare
, sizeof (spa_t
),
1624 offsetof(spa_t
, spa_avl
));
1626 avl_create(&spa_spare_avl
, spa_spare_compare
, sizeof (spa_aux_t
),
1627 offsetof(spa_aux_t
, aux_avl
));
1629 avl_create(&spa_l2cache_avl
, spa_l2cache_compare
, sizeof (spa_aux_t
),
1630 offsetof(spa_aux_t
, aux_avl
));
1632 spa_mode_global
= mode
;
1641 vdev_cache_stat_init();
1644 zpool_feature_init();
1656 vdev_cache_stat_fini();
1665 avl_destroy(&spa_namespace_avl
);
1666 avl_destroy(&spa_spare_avl
);
1667 avl_destroy(&spa_l2cache_avl
);
1669 cv_destroy(&spa_namespace_cv
);
1670 mutex_destroy(&spa_namespace_lock
);
1671 mutex_destroy(&spa_spare_lock
);
1672 mutex_destroy(&spa_l2cache_lock
);
1676 * Return whether this pool has slogs. No locking needed.
1677 * It's not a problem if the wrong answer is returned as it's only for
1678 * performance and not correctness
1681 spa_has_slogs(spa_t
*spa
)
1683 return (spa
->spa_log_class
->mc_rotor
!= NULL
);
1687 spa_get_log_state(spa_t
*spa
)
1689 return (spa
->spa_log_state
);
1693 spa_set_log_state(spa_t
*spa
, spa_log_state_t state
)
1695 spa
->spa_log_state
= state
;
1699 spa_is_root(spa_t
*spa
)
1701 return (spa
->spa_is_root
);
1705 spa_writeable(spa_t
*spa
)
1707 return (!!(spa
->spa_mode
& FWRITE
));
1711 spa_mode(spa_t
*spa
)
1713 return (spa
->spa_mode
);
1717 spa_bootfs(spa_t
*spa
)
1719 return (spa
->spa_bootfs
);
1723 spa_delegation(spa_t
*spa
)
1725 return (spa
->spa_delegation
);
1729 spa_meta_objset(spa_t
*spa
)
1731 return (spa
->spa_meta_objset
);
1735 spa_dedup_checksum(spa_t
*spa
)
1737 return (spa
->spa_dedup_checksum
);
1741 * Reset pool scan stat per scan pass (or reboot).
1744 spa_scan_stat_init(spa_t
*spa
)
1746 /* data not stored on disk */
1747 spa
->spa_scan_pass_start
= gethrestime_sec();
1748 spa
->spa_scan_pass_exam
= 0;
1749 vdev_scan_stat_init(spa
->spa_root_vdev
);
1753 * Get scan stats for zpool status reports
1756 spa_scan_get_stats(spa_t
*spa
, pool_scan_stat_t
*ps
)
1758 dsl_scan_t
*scn
= spa
->spa_dsl_pool
? spa
->spa_dsl_pool
->dp_scan
: NULL
;
1760 if (scn
== NULL
|| scn
->scn_phys
.scn_func
== POOL_SCAN_NONE
)
1761 return (SET_ERROR(ENOENT
));
1762 bzero(ps
, sizeof (pool_scan_stat_t
));
1764 /* data stored on disk */
1765 ps
->pss_func
= scn
->scn_phys
.scn_func
;
1766 ps
->pss_start_time
= scn
->scn_phys
.scn_start_time
;
1767 ps
->pss_end_time
= scn
->scn_phys
.scn_end_time
;
1768 ps
->pss_to_examine
= scn
->scn_phys
.scn_to_examine
;
1769 ps
->pss_examined
= scn
->scn_phys
.scn_examined
;
1770 ps
->pss_to_process
= scn
->scn_phys
.scn_to_process
;
1771 ps
->pss_processed
= scn
->scn_phys
.scn_processed
;
1772 ps
->pss_errors
= scn
->scn_phys
.scn_errors
;
1773 ps
->pss_state
= scn
->scn_phys
.scn_state
;
1775 /* data not stored on disk */
1776 ps
->pss_pass_start
= spa
->spa_scan_pass_start
;
1777 ps
->pss_pass_exam
= spa
->spa_scan_pass_exam
;
1783 spa_debug_enabled(spa_t
*spa
)
1785 return (spa
->spa_debug
);
1788 #if defined(_KERNEL) && defined(HAVE_SPL)
1789 /* Namespace manipulation */
1790 EXPORT_SYMBOL(spa_lookup
);
1791 EXPORT_SYMBOL(spa_add
);
1792 EXPORT_SYMBOL(spa_remove
);
1793 EXPORT_SYMBOL(spa_next
);
1795 /* Refcount functions */
1796 EXPORT_SYMBOL(spa_open_ref
);
1797 EXPORT_SYMBOL(spa_close
);
1798 EXPORT_SYMBOL(spa_refcount_zero
);
1800 /* Pool configuration lock */
1801 EXPORT_SYMBOL(spa_config_tryenter
);
1802 EXPORT_SYMBOL(spa_config_enter
);
1803 EXPORT_SYMBOL(spa_config_exit
);
1804 EXPORT_SYMBOL(spa_config_held
);
1806 /* Pool vdev add/remove lock */
1807 EXPORT_SYMBOL(spa_vdev_enter
);
1808 EXPORT_SYMBOL(spa_vdev_exit
);
1810 /* Pool vdev state change lock */
1811 EXPORT_SYMBOL(spa_vdev_state_enter
);
1812 EXPORT_SYMBOL(spa_vdev_state_exit
);
1814 /* Accessor functions */
1815 EXPORT_SYMBOL(spa_shutting_down
);
1816 EXPORT_SYMBOL(spa_get_dsl
);
1817 EXPORT_SYMBOL(spa_get_rootblkptr
);
1818 EXPORT_SYMBOL(spa_set_rootblkptr
);
1819 EXPORT_SYMBOL(spa_altroot
);
1820 EXPORT_SYMBOL(spa_sync_pass
);
1821 EXPORT_SYMBOL(spa_name
);
1822 EXPORT_SYMBOL(spa_guid
);
1823 EXPORT_SYMBOL(spa_last_synced_txg
);
1824 EXPORT_SYMBOL(spa_first_txg
);
1825 EXPORT_SYMBOL(spa_syncing_txg
);
1826 EXPORT_SYMBOL(spa_version
);
1827 EXPORT_SYMBOL(spa_state
);
1828 EXPORT_SYMBOL(spa_load_state
);
1829 EXPORT_SYMBOL(spa_freeze_txg
);
1830 EXPORT_SYMBOL(spa_get_asize
);
1831 EXPORT_SYMBOL(spa_get_dspace
);
1832 EXPORT_SYMBOL(spa_update_dspace
);
1833 EXPORT_SYMBOL(spa_deflate
);
1834 EXPORT_SYMBOL(spa_normal_class
);
1835 EXPORT_SYMBOL(spa_log_class
);
1836 EXPORT_SYMBOL(spa_max_replication
);
1837 EXPORT_SYMBOL(spa_prev_software_version
);
1838 EXPORT_SYMBOL(spa_get_failmode
);
1839 EXPORT_SYMBOL(spa_suspended
);
1840 EXPORT_SYMBOL(spa_bootfs
);
1841 EXPORT_SYMBOL(spa_delegation
);
1842 EXPORT_SYMBOL(spa_meta_objset
);
1844 /* Miscellaneous support routines */
1845 EXPORT_SYMBOL(spa_rename
);
1846 EXPORT_SYMBOL(spa_guid_exists
);
1847 EXPORT_SYMBOL(spa_strdup
);
1848 EXPORT_SYMBOL(spa_strfree
);
1849 EXPORT_SYMBOL(spa_get_random
);
1850 EXPORT_SYMBOL(spa_generate_guid
);
1851 EXPORT_SYMBOL(sprintf_blkptr
);
1852 EXPORT_SYMBOL(spa_freeze
);
1853 EXPORT_SYMBOL(spa_upgrade
);
1854 EXPORT_SYMBOL(spa_evict_all
);
1855 EXPORT_SYMBOL(spa_lookup_by_guid
);
1856 EXPORT_SYMBOL(spa_has_spare
);
1857 EXPORT_SYMBOL(dva_get_dsize_sync
);
1858 EXPORT_SYMBOL(bp_get_dsize_sync
);
1859 EXPORT_SYMBOL(bp_get_dsize
);
1860 EXPORT_SYMBOL(spa_has_slogs
);
1861 EXPORT_SYMBOL(spa_is_root
);
1862 EXPORT_SYMBOL(spa_writeable
);
1863 EXPORT_SYMBOL(spa_mode
);
1865 EXPORT_SYMBOL(spa_namespace_lock
);
1867 module_param(zfs_deadman_synctime
, ulong
, 0644);
1868 MODULE_PARM_DESC(zfs_deadman_synctime
,"Expire in units of zfs_txg_synctime_ms");
1870 module_param(zfs_deadman_enabled
, int, 0644);
1871 MODULE_PARM_DESC(zfs_deadman_enabled
, "Enable deadman timer");