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 (c) 2011, 2016 by Delphix. All rights reserved.
25 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
26 * Copyright (c) 2014 RackTop Systems.
27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
28 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
29 * Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
30 * Copyright 2017 Nexenta Systems, Inc.
33 #include <sys/dmu_objset.h>
34 #include <sys/dsl_dataset.h>
35 #include <sys/dsl_dir.h>
36 #include <sys/dsl_prop.h>
37 #include <sys/dsl_synctask.h>
38 #include <sys/dmu_traverse.h>
39 #include <sys/dmu_impl.h>
40 #include <sys/dmu_tx.h>
44 #include <sys/zfeature.h>
45 #include <sys/unique.h>
46 #include <sys/zfs_context.h>
47 #include <sys/zfs_ioctl.h>
49 #include <sys/zfs_znode.h>
50 #include <sys/zfs_onexit.h>
52 #include <sys/dsl_scan.h>
53 #include <sys/dsl_deadlist.h>
54 #include <sys/dsl_destroy.h>
55 #include <sys/dsl_userhold.h>
56 #include <sys/dsl_bookmark.h>
57 #include <sys/policy.h>
58 #include <sys/dmu_send.h>
59 #include <sys/zio_compress.h>
60 #include <zfs_fletcher.h>
61 #include <sys/zio_checksum.h>
64 * The SPA supports block sizes up to 16MB. However, very large blocks
65 * can have an impact on i/o latency (e.g. tying up a spinning disk for
66 * ~300ms), and also potentially on the memory allocator. Therefore,
67 * we do not allow the recordsize to be set larger than zfs_max_recordsize
68 * (default 1MB). Larger blocks can be created by changing this tunable,
69 * and pools with larger blocks can always be imported and used, regardless
72 int zfs_max_recordsize
= 1 * 1024 * 1024;
74 #define SWITCH64(x, y) \
76 uint64_t __tmp = (x); \
81 #define DS_REF_MAX (1ULL << 62)
83 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
85 extern int spa_asize_inflation
;
88 * Figure out how much of this delta should be propagated to the dsl_dir
89 * layer. If there's a refreservation, that space has already been
90 * partially accounted for in our ancestors.
93 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
95 dsl_dataset_phys_t
*ds_phys
;
96 uint64_t old_bytes
, new_bytes
;
98 if (ds
->ds_reserved
== 0)
101 ds_phys
= dsl_dataset_phys(ds
);
102 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
103 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
105 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
106 return (new_bytes
- old_bytes
);
110 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
112 int used
, compressed
, uncompressed
;
116 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
117 compressed
= BP_GET_PSIZE(bp
);
118 uncompressed
= BP_GET_UCSIZE(bp
);
120 dprintf_bp(bp
, "ds=%p", ds
);
122 ASSERT(dmu_tx_is_syncing(tx
));
123 /* It could have been compressed away to nothing */
126 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
127 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
129 dsl_pool_mos_diduse_space(tx
->tx_pool
,
130 used
, compressed
, uncompressed
);
134 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
135 mutex_enter(&ds
->ds_lock
);
136 delta
= parent_delta(ds
, used
);
137 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
138 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
139 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
140 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
142 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
143 ds
->ds_feature_activation_needed
[SPA_FEATURE_LARGE_BLOCKS
] =
147 f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
148 if (f
!= SPA_FEATURE_NONE
)
149 ds
->ds_feature_activation_needed
[f
] = B_TRUE
;
151 mutex_exit(&ds
->ds_lock
);
152 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
153 compressed
, uncompressed
, tx
);
154 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
155 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
159 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
162 int used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
163 int compressed
= BP_GET_PSIZE(bp
);
164 int uncompressed
= BP_GET_UCSIZE(bp
);
169 ASSERT(dmu_tx_is_syncing(tx
));
170 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
173 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
174 dsl_pool_mos_diduse_space(tx
->tx_pool
,
175 -used
, -compressed
, -uncompressed
);
178 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
180 ASSERT(!ds
->ds_is_snapshot
);
181 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
183 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
186 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
187 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
189 mutex_enter(&ds
->ds_lock
);
190 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
191 !DS_UNIQUE_IS_ACCURATE(ds
));
192 delta
= parent_delta(ds
, -used
);
193 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
194 mutex_exit(&ds
->ds_lock
);
195 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
196 delta
, -compressed
, -uncompressed
, tx
);
197 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
198 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
200 dprintf_bp(bp
, "putting on dead list: %s", "");
203 * We are here as part of zio's write done callback,
204 * which means we're a zio interrupt thread. We can't
205 * call dsl_deadlist_insert() now because it may block
206 * waiting for I/O. Instead, put bp on the deferred
207 * queue and let dsl_pool_sync() finish the job.
209 bplist_append(&ds
->ds_pending_deadlist
, bp
);
211 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
213 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
214 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
215 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
216 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
217 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
218 ds
->ds_object
&& bp
->blk_birth
>
219 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
220 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
221 mutex_enter(&ds
->ds_prev
->ds_lock
);
222 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
223 mutex_exit(&ds
->ds_prev
->ds_lock
);
225 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
226 dsl_dir_transfer_space(ds
->ds_dir
, used
,
227 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
230 mutex_enter(&ds
->ds_lock
);
231 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
232 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
233 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
234 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
235 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
236 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
237 mutex_exit(&ds
->ds_lock
);
243 dsl_dataset_prev_snap_txg(dsl_dataset_t
*ds
)
245 uint64_t trysnap
= 0;
250 * The snapshot creation could fail, but that would cause an
251 * incorrect FALSE return, which would only result in an
252 * overestimation of the amount of space that an operation would
253 * consume, which is OK.
255 * There's also a small window where we could miss a pending
256 * snapshot, because we could set the sync task in the quiescing
257 * phase. So this should only be used as a guess.
259 if (ds
->ds_trysnap_txg
>
260 spa_last_synced_txg(ds
->ds_dir
->dd_pool
->dp_spa
))
261 trysnap
= ds
->ds_trysnap_txg
;
262 return (MAX(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, trysnap
));
266 dsl_dataset_block_freeable(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
269 if (blk_birth
<= dsl_dataset_prev_snap_txg(ds
) ||
270 (bp
!= NULL
&& BP_IS_HOLE(bp
)))
273 ddt_prefetch(dsl_dataset_get_spa(ds
), bp
);
279 * We have to release the fsid syncronously or we risk that a subsequent
280 * mount of the same dataset will fail to unique_insert the fsid. This
281 * failure would manifest itself as the fsid of this dataset changing
282 * between mounts which makes NFS clients quite unhappy.
285 dsl_dataset_evict_sync(void *dbu
)
287 dsl_dataset_t
*ds
= dbu
;
289 ASSERT(ds
->ds_owner
== NULL
);
291 unique_remove(ds
->ds_fsid_guid
);
295 dsl_dataset_evict_async(void *dbu
)
297 dsl_dataset_t
*ds
= dbu
;
299 ASSERT(ds
->ds_owner
== NULL
);
303 if (ds
->ds_objset
!= NULL
)
304 dmu_objset_evict(ds
->ds_objset
);
307 dsl_dataset_rele(ds
->ds_prev
, ds
);
311 bplist_destroy(&ds
->ds_pending_deadlist
);
312 if (ds
->ds_deadlist
.dl_os
!= NULL
)
313 dsl_deadlist_close(&ds
->ds_deadlist
);
315 dsl_dir_async_rele(ds
->ds_dir
, ds
);
317 ASSERT(!list_link_active(&ds
->ds_synced_link
));
319 list_destroy(&ds
->ds_prop_cbs
);
320 mutex_destroy(&ds
->ds_lock
);
321 mutex_destroy(&ds
->ds_opening_lock
);
322 mutex_destroy(&ds
->ds_sendstream_lock
);
323 refcount_destroy(&ds
->ds_longholds
);
324 rrw_destroy(&ds
->ds_bp_rwlock
);
326 kmem_free(ds
, sizeof (dsl_dataset_t
));
330 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
332 dsl_dataset_phys_t
*headphys
;
335 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
336 objset_t
*mos
= dp
->dp_meta_objset
;
338 if (ds
->ds_snapname
[0])
340 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
343 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
347 headphys
= headdbuf
->db_data
;
348 err
= zap_value_search(dp
->dp_meta_objset
,
349 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
350 if (err
!= 0 && zfs_recover
== B_TRUE
) {
352 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
353 "SNAPOBJ=%llu-ERR=%d",
354 (unsigned long long)ds
->ds_object
, err
);
356 dmu_buf_rele(headdbuf
, FTAG
);
361 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
363 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
364 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
368 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
371 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
372 value
, mt
, NULL
, 0, NULL
);
373 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
374 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
379 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
382 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
383 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
387 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
389 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
392 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
393 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
394 err
= zap_remove(mos
, snapobj
, name
, tx
);
396 if (err
== 0 && adj_cnt
)
397 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
398 DD_FIELD_SNAPSHOT_COUNT
, tx
);
404 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
406 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
407 boolean_t result
= B_FALSE
;
409 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
410 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
412 if (ds
== dmu_buf_get_user(dbuf
))
415 dmu_buf_rele(dbuf
, tag
);
422 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
425 objset_t
*mos
= dp
->dp_meta_objset
;
429 dmu_object_info_t doi
;
431 ASSERT(dsl_pool_config_held(dp
));
433 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
437 /* Make sure dsobj has the correct object type. */
438 dmu_object_info_from_db(dbuf
, &doi
);
439 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
440 dmu_buf_rele(dbuf
, tag
);
441 return (SET_ERROR(EINVAL
));
444 ds
= dmu_buf_get_user(dbuf
);
446 dsl_dataset_t
*winner
= NULL
;
448 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
450 ds
->ds_object
= dsobj
;
451 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
452 list_link_init(&ds
->ds_synced_link
);
454 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
455 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
456 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
457 rrw_init(&ds
->ds_bp_rwlock
, B_FALSE
);
458 refcount_create(&ds
->ds_longholds
);
460 bplist_create(&ds
->ds_pending_deadlist
);
461 dsl_deadlist_open(&ds
->ds_deadlist
,
462 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
464 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
465 offsetof(dmu_sendarg_t
, dsa_link
));
467 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
468 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
470 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
473 for (f
= 0; f
< SPA_FEATURES
; f
++) {
474 if (!(spa_feature_table
[f
].fi_flags
&
475 ZFEATURE_FLAG_PER_DATASET
))
477 err
= zap_contains(mos
, dsobj
,
478 spa_feature_table
[f
].fi_guid
);
480 ds
->ds_feature_inuse
[f
] = B_TRUE
;
482 ASSERT3U(err
, ==, ENOENT
);
488 err
= dsl_dir_hold_obj(dp
,
489 dsl_dataset_phys(ds
)->ds_dir_obj
, NULL
, ds
, &ds
->ds_dir
);
491 mutex_destroy(&ds
->ds_lock
);
492 mutex_destroy(&ds
->ds_opening_lock
);
493 mutex_destroy(&ds
->ds_sendstream_lock
);
494 refcount_destroy(&ds
->ds_longholds
);
495 bplist_destroy(&ds
->ds_pending_deadlist
);
496 dsl_deadlist_close(&ds
->ds_deadlist
);
497 kmem_free(ds
, sizeof (dsl_dataset_t
));
498 dmu_buf_rele(dbuf
, tag
);
502 if (!ds
->ds_is_snapshot
) {
503 ds
->ds_snapname
[0] = '\0';
504 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
505 err
= dsl_dataset_hold_obj(dp
,
506 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
509 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
510 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
511 DS_FIELD_BOOKMARK_NAMES
,
512 sizeof (ds
->ds_bookmarks
), 1,
514 if (zaperr
!= ENOENT
)
518 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
519 err
= dsl_dataset_get_snapname(ds
);
521 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
523 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
524 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
529 if (err
== 0 && !ds
->ds_is_snapshot
) {
530 err
= dsl_prop_get_int_ds(ds
,
531 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
534 err
= dsl_prop_get_int_ds(ds
,
535 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
539 ds
->ds_reserved
= ds
->ds_quota
= 0;
542 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict_sync
,
543 dsl_dataset_evict_async
, &ds
->ds_dbuf
);
545 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
547 if (err
!= 0 || winner
!= NULL
) {
548 bplist_destroy(&ds
->ds_pending_deadlist
);
549 dsl_deadlist_close(&ds
->ds_deadlist
);
551 dsl_dataset_rele(ds
->ds_prev
, ds
);
552 dsl_dir_rele(ds
->ds_dir
, ds
);
553 mutex_destroy(&ds
->ds_lock
);
554 mutex_destroy(&ds
->ds_opening_lock
);
555 mutex_destroy(&ds
->ds_sendstream_lock
);
556 refcount_destroy(&ds
->ds_longholds
);
557 kmem_free(ds
, sizeof (dsl_dataset_t
));
559 dmu_buf_rele(dbuf
, tag
);
565 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
566 if (ds
->ds_fsid_guid
!=
567 dsl_dataset_phys(ds
)->ds_fsid_guid
) {
568 zfs_dbgmsg("ds_fsid_guid changed from "
569 "%llx to %llx for pool %s dataset id %llu",
571 dsl_dataset_phys(ds
)->ds_fsid_guid
,
572 (long long)ds
->ds_fsid_guid
,
573 spa_name(dp
->dp_spa
),
578 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
579 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
580 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
581 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
582 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
588 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
,
589 void *tag
, dsl_dataset_t
**dsp
)
592 const char *snapname
;
597 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
601 ASSERT(dsl_pool_config_held(dp
));
602 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
604 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
606 err
= SET_ERROR(ENOENT
);
608 /* we may be looking for a snapshot */
609 if (err
== 0 && snapname
!= NULL
) {
610 dsl_dataset_t
*snap_ds
;
612 if (*snapname
++ != '@') {
613 dsl_dataset_rele(ds
, tag
);
614 dsl_dir_rele(dd
, FTAG
);
615 return (SET_ERROR(ENOENT
));
618 dprintf("looking for snapshot '%s'\n", snapname
);
619 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
621 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &snap_ds
);
622 dsl_dataset_rele(ds
, tag
);
625 mutex_enter(&snap_ds
->ds_lock
);
626 if (snap_ds
->ds_snapname
[0] == 0)
627 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
628 sizeof (snap_ds
->ds_snapname
));
629 mutex_exit(&snap_ds
->ds_lock
);
635 dsl_dir_rele(dd
, FTAG
);
640 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
,
641 void *tag
, dsl_dataset_t
**dsp
)
643 int err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
646 if (!dsl_dataset_tryown(*dsp
, tag
)) {
647 dsl_dataset_rele(*dsp
, tag
);
649 return (SET_ERROR(EBUSY
));
655 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
,
656 void *tag
, dsl_dataset_t
**dsp
)
658 int err
= dsl_dataset_hold(dp
, name
, tag
, dsp
);
661 if (!dsl_dataset_tryown(*dsp
, tag
)) {
662 dsl_dataset_rele(*dsp
, tag
);
663 return (SET_ERROR(EBUSY
));
669 * See the comment above dsl_pool_hold() for details. In summary, a long
670 * hold is used to prevent destruction of a dataset while the pool hold
671 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
673 * The dataset and pool must be held when this function is called. After it
674 * is called, the pool hold may be released while the dataset is still held
678 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
680 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
681 (void) refcount_add(&ds
->ds_longholds
, tag
);
685 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
687 (void) refcount_remove(&ds
->ds_longholds
, tag
);
690 /* Return B_TRUE if there are any long holds on this dataset. */
692 dsl_dataset_long_held(dsl_dataset_t
*ds
)
694 return (!refcount_is_zero(&ds
->ds_longholds
));
698 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
701 (void) strcpy(name
, "mos");
703 dsl_dir_name(ds
->ds_dir
, name
);
704 VERIFY0(dsl_dataset_get_snapname(ds
));
705 if (ds
->ds_snapname
[0]) {
706 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
707 <, ZFS_MAX_DATASET_NAME_LEN
);
709 * We use a "recursive" mutex so that we
710 * can call dprintf_ds() with ds_lock held.
712 if (!MUTEX_HELD(&ds
->ds_lock
)) {
713 mutex_enter(&ds
->ds_lock
);
714 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
715 ZFS_MAX_DATASET_NAME_LEN
), <,
716 ZFS_MAX_DATASET_NAME_LEN
);
717 mutex_exit(&ds
->ds_lock
);
719 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
720 ZFS_MAX_DATASET_NAME_LEN
), <,
721 ZFS_MAX_DATASET_NAME_LEN
);
728 dsl_dataset_namelen(dsl_dataset_t
*ds
)
731 VERIFY0(dsl_dataset_get_snapname(ds
));
732 mutex_enter(&ds
->ds_lock
);
733 len
= strlen(ds
->ds_snapname
);
734 /* add '@' if ds is a snap */
737 len
+= dsl_dir_namelen(ds
->ds_dir
);
738 mutex_exit(&ds
->ds_lock
);
743 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
745 dmu_buf_rele(ds
->ds_dbuf
, tag
);
749 dsl_dataset_disown(dsl_dataset_t
*ds
, void *tag
)
751 ASSERT3P(ds
->ds_owner
, ==, tag
);
752 ASSERT(ds
->ds_dbuf
!= NULL
);
754 mutex_enter(&ds
->ds_lock
);
756 mutex_exit(&ds
->ds_lock
);
757 dsl_dataset_long_rele(ds
, tag
);
758 dsl_dataset_rele(ds
, tag
);
762 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
764 boolean_t gotit
= FALSE
;
766 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
767 mutex_enter(&ds
->ds_lock
);
768 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
770 dsl_dataset_long_hold(ds
, tag
);
773 mutex_exit(&ds
->ds_lock
);
778 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
781 mutex_enter(&ds
->ds_lock
);
782 rv
= (ds
->ds_owner
!= NULL
);
783 mutex_exit(&ds
->ds_lock
);
788 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
790 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
791 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
794 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
796 spa_feature_incr(spa
, f
, tx
);
797 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
799 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
800 sizeof (zero
), 1, &zero
, tx
));
804 dsl_dataset_deactivate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
806 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
807 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
809 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
811 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
812 spa_feature_decr(spa
, f
, tx
);
816 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
817 uint64_t flags
, dmu_tx_t
*tx
)
819 dsl_pool_t
*dp
= dd
->dd_pool
;
821 dsl_dataset_phys_t
*dsphys
;
823 objset_t
*mos
= dp
->dp_meta_objset
;
826 origin
= dp
->dp_origin_snap
;
828 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
829 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
830 ASSERT(dmu_tx_is_syncing(tx
));
831 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
833 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
834 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
835 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
836 dmu_buf_will_dirty(dbuf
, tx
);
837 dsphys
= dbuf
->db_data
;
838 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
839 dsphys
->ds_dir_obj
= dd
->dd_object
;
840 dsphys
->ds_flags
= flags
;
841 dsphys
->ds_fsid_guid
= unique_create();
842 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
843 sizeof (dsphys
->ds_guid
));
844 dsphys
->ds_snapnames_zapobj
=
845 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
847 dsphys
->ds_creation_time
= gethrestime_sec();
848 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
850 if (origin
== NULL
) {
851 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
854 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
856 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
857 dsphys
->ds_prev_snap_txg
=
858 dsl_dataset_phys(origin
)->ds_creation_txg
;
859 dsphys
->ds_referenced_bytes
=
860 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
861 dsphys
->ds_compressed_bytes
=
862 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
863 dsphys
->ds_uncompressed_bytes
=
864 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
865 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
866 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
867 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
870 * Inherit flags that describe the dataset's contents
871 * (INCONSISTENT) or properties (Case Insensitive).
873 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
874 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
876 for (f
= 0; f
< SPA_FEATURES
; f
++) {
877 if (origin
->ds_feature_inuse
[f
])
878 dsl_dataset_activate_feature(dsobj
, f
, tx
);
881 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
882 dsl_dataset_phys(origin
)->ds_num_children
++;
884 VERIFY0(dsl_dataset_hold_obj(dp
,
885 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
887 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
888 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
889 dsl_dataset_rele(ohds
, FTAG
);
891 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
892 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
893 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
895 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
897 VERIFY0(zap_add_int(mos
,
898 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
902 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
903 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
904 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
905 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
906 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
907 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
909 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
911 VERIFY0(zap_add_int(mos
,
912 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
917 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
918 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
920 dmu_buf_rele(dbuf
, FTAG
);
922 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
923 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
929 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
933 VERIFY0(dmu_objset_from_ds(ds
, &os
));
934 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
935 dsl_dataset_dirty(ds
, tx
);
939 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
940 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
, dmu_tx_t
*tx
)
942 dsl_pool_t
*dp
= pdd
->dd_pool
;
943 uint64_t dsobj
, ddobj
;
946 ASSERT(dmu_tx_is_syncing(tx
));
947 ASSERT(lastname
[0] != '@');
949 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
950 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
952 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
,
953 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
955 dsl_deleg_set_create_perms(dd
, tx
, cr
);
958 * Since we're creating a new node we know it's a leaf, so we can
959 * initialize the counts if the limit feature is active.
961 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
963 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
965 dsl_dir_zapify(dd
, tx
);
966 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
967 sizeof (cnt
), 1, &cnt
, tx
));
968 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
969 sizeof (cnt
), 1, &cnt
, tx
));
972 dsl_dir_rele(dd
, FTAG
);
975 * If we are creating a clone, make sure we zero out any stale
976 * data from the origin snapshots zil header.
978 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
981 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
982 dsl_dataset_zero_zil(ds
, tx
);
983 dsl_dataset_rele(ds
, FTAG
);
990 * The unique space in the head dataset can be calculated by subtracting
991 * the space used in the most recent snapshot, that is still being used
992 * in this file system, from the space currently in use. To figure out
993 * the space in the most recent snapshot still in use, we need to take
994 * the total space used in the snapshot and subtract out the space that
995 * has been freed up since the snapshot was taken.
998 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1001 uint64_t dlused
, dlcomp
, dluncomp
;
1003 ASSERT(!ds
->ds_is_snapshot
);
1005 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1006 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1010 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1012 ASSERT3U(dlused
, <=, mrs_used
);
1013 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1014 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1016 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1017 SPA_VERSION_UNIQUE_ACCURATE
)
1018 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1022 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1025 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1027 ASSERTV(uint64_t count
);
1029 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1030 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1033 * The err should not be ENOENT, but a bug in a previous version
1034 * of the code could cause upgrade_clones_cb() to not set
1035 * ds_next_snap_obj when it should, leading to a missing entry.
1036 * If we knew that the pool was created after
1037 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1038 * ENOENT. However, at least we can check that we don't have
1039 * too many entries in the next_clones_obj even after failing to
1044 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1046 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1051 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1053 return (&dsl_dataset_phys(ds
)->ds_bp
);
1057 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1059 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1063 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1067 if (ds
== NULL
) /* this is the meta-objset */
1070 ASSERT(ds
->ds_objset
!= NULL
);
1072 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1073 panic("dirtying snapshot!");
1075 dp
= ds
->ds_dir
->dd_pool
;
1077 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1078 /* up the hold count until we can be written out */
1079 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1084 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
1088 for (t
= 0; t
< TXG_SIZE
; t
++) {
1089 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1097 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1101 if (!dmu_tx_is_syncing(tx
))
1105 * If there's an fs-only reservation, any blocks that might become
1106 * owned by the snapshot dataset must be accommodated by space
1107 * outside of the reservation.
1109 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1110 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1111 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1112 return (SET_ERROR(ENOSPC
));
1115 * Propagate any reserved space for this snapshot to other
1116 * snapshot checks in this sync group.
1119 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1124 typedef struct dsl_dataset_snapshot_arg
{
1125 nvlist_t
*ddsa_snaps
;
1126 nvlist_t
*ddsa_props
;
1127 nvlist_t
*ddsa_errors
;
1129 } dsl_dataset_snapshot_arg_t
;
1132 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1133 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1138 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1140 if (!dmu_tx_is_syncing(tx
))
1144 * We don't allow multiple snapshots of the same txg. If there
1145 * is already one, try again.
1147 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1148 return (SET_ERROR(EAGAIN
));
1151 * Check for conflicting snapshot name.
1153 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1155 return (SET_ERROR(EEXIST
));
1156 if (error
!= ENOENT
)
1160 * We don't allow taking snapshots of inconsistent datasets, such as
1161 * those into which we are currently receiving. However, if we are
1162 * creating this snapshot as part of a receive, this check will be
1163 * executed atomically with respect to the completion of the receive
1164 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1165 * case we ignore this, knowing it will be fixed up for us shortly in
1166 * dmu_recv_end_sync().
1168 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1169 return (SET_ERROR(EBUSY
));
1172 * Skip the check for temporary snapshots or if we have already checked
1173 * the counts in dsl_dataset_snapshot_check. This means we really only
1174 * check the count here when we're receiving a stream.
1176 if (cnt
!= 0 && cr
!= NULL
) {
1177 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1178 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1183 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1191 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1193 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1194 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1199 * Pre-compute how many total new snapshots will be created for each
1200 * level in the tree and below. This is needed for validating the
1201 * snapshot limit when either taking a recursive snapshot or when
1202 * taking multiple snapshots.
1204 * The problem is that the counts are not actually adjusted when
1205 * we are checking, only when we finally sync. For a single snapshot,
1206 * this is easy, the count will increase by 1 at each node up the tree,
1207 * but its more complicated for the recursive/multiple snapshot case.
1209 * The dsl_fs_ss_limit_check function does recursively check the count
1210 * at each level up the tree but since it is validating each snapshot
1211 * independently we need to be sure that we are validating the complete
1212 * count for the entire set of snapshots. We do this by rolling up the
1213 * counts for each component of the name into an nvlist and then
1214 * checking each of those cases with the aggregated count.
1216 * This approach properly handles not only the recursive snapshot
1217 * case (where we get all of those on the ddsa_snaps list) but also
1218 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1219 * validate the limit on 'a' using a count of 2).
1221 * We validate the snapshot names in the third loop and only report
1224 if (dmu_tx_is_syncing(tx
)) {
1226 nvlist_t
*cnt_track
= NULL
;
1227 cnt_track
= fnvlist_alloc();
1229 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1231 /* Rollup aggregated counts into the cnt_track list */
1232 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1234 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1238 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1239 pdelim
= strchr(nm
, '@');
1245 if (nvlist_lookup_uint64(cnt_track
, nm
,
1247 /* update existing entry */
1248 fnvlist_add_uint64(cnt_track
, nm
,
1252 fnvlist_add_uint64(cnt_track
, nm
, 1);
1255 pdelim
= strrchr(nm
, '/');
1258 } while (pdelim
!= NULL
);
1261 kmem_free(nm
, MAXPATHLEN
);
1263 /* Check aggregated counts at each level */
1264 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1265 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1271 name
= nvpair_name(pair
);
1272 cnt
= fnvpair_value_uint64(pair
);
1275 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1277 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1278 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1280 dsl_dataset_rele(ds
, FTAG
);
1284 if (ddsa
->ddsa_errors
!= NULL
)
1285 fnvlist_add_int32(ddsa
->ddsa_errors
,
1288 /* only report one error for this check */
1292 nvlist_free(cnt_track
);
1295 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1296 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1299 char *name
, *atp
= NULL
;
1300 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1302 name
= nvpair_name(pair
);
1303 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1304 error
= SET_ERROR(ENAMETOOLONG
);
1306 atp
= strchr(name
, '@');
1308 error
= SET_ERROR(EINVAL
);
1310 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1313 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1315 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1316 error
= dsl_dataset_snapshot_check_impl(ds
,
1317 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1318 dsl_dataset_rele(ds
, FTAG
);
1322 if (ddsa
->ddsa_errors
!= NULL
) {
1323 fnvlist_add_int32(ddsa
->ddsa_errors
,
1334 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1337 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1339 dsl_dataset_phys_t
*dsphys
;
1340 uint64_t dsobj
, crtxg
;
1341 objset_t
*mos
= dp
->dp_meta_objset
;
1343 ASSERTV(static zil_header_t zero_zil
);
1344 ASSERTV(objset_t
*os
);
1346 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1349 * If we are on an old pool, the zil must not be active, in which
1350 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1352 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1353 dmu_objset_from_ds(ds
, &os
) != 0 ||
1354 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1355 sizeof (zero_zil
)) == 0);
1357 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1360 * The origin's ds_creation_txg has to be < TXG_INITIAL
1362 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1367 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1368 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1369 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1370 dmu_buf_will_dirty(dbuf
, tx
);
1371 dsphys
= dbuf
->db_data
;
1372 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1373 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1374 dsphys
->ds_fsid_guid
= unique_create();
1375 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1376 sizeof (dsphys
->ds_guid
));
1377 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1378 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1379 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1380 dsphys
->ds_num_children
= 1;
1381 dsphys
->ds_creation_time
= gethrestime_sec();
1382 dsphys
->ds_creation_txg
= crtxg
;
1383 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1384 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1385 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1386 dsphys
->ds_uncompressed_bytes
=
1387 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1388 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1389 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1390 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1391 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1392 dmu_buf_rele(dbuf
, FTAG
);
1394 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1395 if (ds
->ds_feature_inuse
[f
])
1396 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1399 ASSERT3U(ds
->ds_prev
!= 0, ==,
1400 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1402 uint64_t next_clones_obj
=
1403 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1404 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1406 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1407 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1409 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1410 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1411 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1412 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1413 } else if (next_clones_obj
!= 0) {
1414 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1415 dsphys
->ds_next_snap_obj
, tx
);
1416 VERIFY0(zap_add_int(mos
,
1417 next_clones_obj
, dsobj
, tx
));
1422 * If we have a reference-reservation on this dataset, we will
1423 * need to increase the amount of refreservation being charged
1424 * since our unique space is going to zero.
1426 if (ds
->ds_reserved
) {
1428 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1429 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1431 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1435 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1436 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1437 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1438 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1439 dsl_deadlist_close(&ds
->ds_deadlist
);
1440 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1441 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1442 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1443 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1445 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1446 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1447 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1448 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1449 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1450 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1452 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1453 snapname
, 8, 1, &dsobj
, tx
));
1456 dsl_dataset_rele(ds
->ds_prev
, ds
);
1457 VERIFY0(dsl_dataset_hold_obj(dp
,
1458 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1460 dsl_scan_ds_snapshotted(ds
, tx
);
1462 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1464 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1468 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1470 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1471 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1474 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1475 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1478 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1480 name
= nvpair_name(pair
);
1481 atp
= strchr(name
, '@');
1482 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1483 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1485 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1486 if (ddsa
->ddsa_props
!= NULL
) {
1487 dsl_props_set_sync_impl(ds
->ds_prev
,
1488 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1490 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1491 dsl_dataset_rele(ds
, FTAG
);
1496 * The snapshots must all be in the same pool.
1497 * All-or-nothing: if there are any failures, nothing will be modified.
1500 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1502 dsl_dataset_snapshot_arg_t ddsa
;
1504 boolean_t needsuspend
;
1508 nvlist_t
*suspended
= NULL
;
1510 pair
= nvlist_next_nvpair(snaps
, NULL
);
1513 firstname
= nvpair_name(pair
);
1515 error
= spa_open(firstname
, &spa
, FTAG
);
1518 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1519 spa_close(spa
, FTAG
);
1522 suspended
= fnvlist_alloc();
1523 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1524 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1525 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1526 char *snapname
= nvpair_name(pair
);
1530 atp
= strchr(snapname
, '@');
1532 error
= SET_ERROR(EINVAL
);
1535 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1537 error
= zil_suspend(fsname
, &cookie
);
1540 fnvlist_add_uint64(suspended
, fsname
,
1545 ddsa
.ddsa_snaps
= snaps
;
1546 ddsa
.ddsa_props
= props
;
1547 ddsa
.ddsa_errors
= errors
;
1548 ddsa
.ddsa_cr
= CRED();
1551 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1552 dsl_dataset_snapshot_sync
, &ddsa
,
1553 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1556 if (suspended
!= NULL
) {
1557 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1558 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1559 zil_resume((void *)(uintptr_t)
1560 fnvpair_value_uint64(pair
));
1562 fnvlist_free(suspended
);
1568 typedef struct dsl_dataset_snapshot_tmp_arg
{
1569 const char *ddsta_fsname
;
1570 const char *ddsta_snapname
;
1571 minor_t ddsta_cleanup_minor
;
1572 const char *ddsta_htag
;
1573 } dsl_dataset_snapshot_tmp_arg_t
;
1576 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1578 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1579 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1583 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1587 /* NULL cred means no limit check for tmp snapshot */
1588 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1589 tx
, B_FALSE
, 0, NULL
);
1591 dsl_dataset_rele(ds
, FTAG
);
1595 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1596 dsl_dataset_rele(ds
, FTAG
);
1597 return (SET_ERROR(ENOTSUP
));
1599 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1602 dsl_dataset_rele(ds
, FTAG
);
1606 dsl_dataset_rele(ds
, FTAG
);
1611 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1613 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1614 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1615 dsl_dataset_t
*ds
= NULL
;
1617 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1619 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1620 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1621 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1622 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1624 dsl_dataset_rele(ds
, FTAG
);
1628 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1629 minor_t cleanup_minor
, const char *htag
)
1631 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1634 boolean_t needsuspend
;
1637 ddsta
.ddsta_fsname
= fsname
;
1638 ddsta
.ddsta_snapname
= snapname
;
1639 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1640 ddsta
.ddsta_htag
= htag
;
1642 error
= spa_open(fsname
, &spa
, FTAG
);
1645 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1646 spa_close(spa
, FTAG
);
1649 error
= zil_suspend(fsname
, &cookie
);
1654 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1655 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1664 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1668 ASSERT(dmu_tx_is_syncing(tx
));
1669 ASSERT(ds
->ds_objset
!= NULL
);
1670 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1673 * in case we had to change ds_fsid_guid when we opened it,
1676 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1677 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1679 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1680 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1681 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1682 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1683 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1684 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1685 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1686 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1687 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1688 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1689 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1690 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1691 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1694 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1696 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1697 if (ds
->ds_feature_activation_needed
[f
]) {
1698 if (ds
->ds_feature_inuse
[f
])
1700 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1701 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1707 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1710 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1713 nvlist_t
*propval
= fnvlist_alloc();
1714 nvlist_t
*val
= fnvlist_alloc();
1716 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1719 * There may be missing entries in ds_next_clones_obj
1720 * due to a bug in a previous version of the code.
1721 * Only trust it if it has the right number of entries.
1723 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1724 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1727 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1)
1729 for (zap_cursor_init(&zc
, mos
,
1730 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1731 zap_cursor_retrieve(&zc
, &za
) == 0;
1732 zap_cursor_advance(&zc
)) {
1733 dsl_dataset_t
*clone
;
1734 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1735 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1736 za
.za_first_integer
, FTAG
, &clone
));
1737 dsl_dir_name(clone
->ds_dir
, buf
);
1738 fnvlist_add_boolean(val
, buf
);
1739 dsl_dataset_rele(clone
, FTAG
);
1741 zap_cursor_fini(&zc
);
1742 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1743 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
), propval
);
1746 nvlist_free(propval
);
1750 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1752 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1754 if (dsl_dataset_has_resume_receive_state(ds
)) {
1757 uint8_t *compressed
;
1759 nvlist_t
*token_nv
= fnvlist_alloc();
1760 size_t packed_size
, compressed_size
;
1763 char buf
[MAXNAMELEN
];
1766 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1767 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1768 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1770 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1771 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1772 fnvlist_add_uint64(token_nv
, "object", val
);
1774 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1775 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1776 fnvlist_add_uint64(token_nv
, "offset", val
);
1778 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1779 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1780 fnvlist_add_uint64(token_nv
, "bytes", val
);
1782 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1783 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1784 fnvlist_add_uint64(token_nv
, "toguid", val
);
1786 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1787 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1788 fnvlist_add_string(token_nv
, "toname", buf
);
1790 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1791 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1792 fnvlist_add_boolean(token_nv
, "largeblockok");
1794 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1795 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1796 fnvlist_add_boolean(token_nv
, "embedok");
1798 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1799 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1800 fnvlist_add_boolean(token_nv
, "compressok");
1802 packed
= fnvlist_pack(token_nv
, &packed_size
);
1803 fnvlist_free(token_nv
);
1804 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1806 compressed_size
= gzip_compress(packed
, compressed
,
1807 packed_size
, packed_size
, 6);
1809 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1811 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1812 for (i
= 0; i
< compressed_size
; i
++) {
1813 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1815 str
[compressed_size
* 2] = '\0';
1816 propval
= kmem_asprintf("%u-%llx-%llx-%s",
1817 ZFS_SEND_RESUME_TOKEN_VERSION
,
1818 (longlong_t
)cksum
.zc_word
[0],
1819 (longlong_t
)packed_size
, str
);
1820 dsl_prop_nvlist_add_string(nv
,
1821 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
1822 kmem_free(packed
, packed_size
);
1823 kmem_free(str
, compressed_size
* 2 + 1);
1824 kmem_free(compressed
, packed_size
);
1830 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1832 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1833 uint64_t refd
, avail
, uobjs
, aobjs
, ratio
;
1835 ASSERT(dsl_pool_config_held(dp
));
1837 ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1838 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1839 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1841 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
, ratio
);
1842 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
1843 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1845 if (ds
->ds_is_snapshot
) {
1846 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
, ratio
);
1847 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
1848 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1849 get_clones_stat(ds
, nv
);
1851 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
1852 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1853 dsl_dataset_name(ds
->ds_prev
, buf
);
1854 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
, buf
);
1857 dsl_dir_stats(ds
->ds_dir
, nv
);
1860 dsl_dataset_space(ds
, &refd
, &avail
, &uobjs
, &aobjs
);
1861 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
, avail
);
1862 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
, refd
);
1864 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
1865 dsl_dataset_phys(ds
)->ds_creation_time
);
1866 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
1867 dsl_dataset_phys(ds
)->ds_creation_txg
);
1868 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
1870 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
1872 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
1873 dsl_dataset_phys(ds
)->ds_guid
);
1874 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
1875 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1876 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
1878 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
1880 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
1881 DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
1883 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
1884 uint64_t written
, comp
, uncomp
;
1885 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1886 dsl_dataset_t
*prev
;
1889 err
= dsl_dataset_hold_obj(dp
,
1890 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
1892 err
= dsl_dataset_space_written(prev
, ds
, &written
,
1894 dsl_dataset_rele(prev
, FTAG
);
1896 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
1902 if (!dsl_dataset_is_snapshot(ds
)) {
1903 /* 6 extra bytes for /%recv */
1904 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
1905 dsl_dataset_t
*recv_ds
;
1908 * A failed "newfs" (e.g. full) resumable receive leaves
1909 * the stats set on this dataset. Check here for the prop.
1911 get_receive_resume_stats(ds
, nv
);
1914 * A failed incremental resumable receive leaves the
1915 * stats set on our child named "%recv". Check the child
1918 dsl_dataset_name(ds
, recvname
);
1919 if (strlcat(recvname
, "/", sizeof (recvname
)) <
1920 sizeof (recvname
) &&
1921 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
1922 sizeof (recvname
) &&
1923 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
1924 get_receive_resume_stats(recv_ds
, nv
);
1925 dsl_dataset_rele(recv_ds
, FTAG
);
1931 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
1933 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1934 ASSERT(dsl_pool_config_held(dp
));
1936 stat
->dds_creation_txg
= dsl_dataset_phys(ds
)->ds_creation_txg
;
1937 stat
->dds_inconsistent
=
1938 dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
;
1939 stat
->dds_guid
= dsl_dataset_phys(ds
)->ds_guid
;
1940 stat
->dds_origin
[0] = '\0';
1941 if (ds
->ds_is_snapshot
) {
1942 stat
->dds_is_snapshot
= B_TRUE
;
1943 stat
->dds_num_clones
=
1944 dsl_dataset_phys(ds
)->ds_num_children
- 1;
1946 stat
->dds_is_snapshot
= B_FALSE
;
1947 stat
->dds_num_clones
= 0;
1949 if (dsl_dir_is_clone(ds
->ds_dir
)) {
1952 VERIFY0(dsl_dataset_hold_obj(dp
,
1953 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
,
1955 dsl_dataset_name(ods
, stat
->dds_origin
);
1956 dsl_dataset_rele(ods
, FTAG
);
1962 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
1964 return (ds
->ds_fsid_guid
);
1968 dsl_dataset_space(dsl_dataset_t
*ds
,
1969 uint64_t *refdbytesp
, uint64_t *availbytesp
,
1970 uint64_t *usedobjsp
, uint64_t *availobjsp
)
1972 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1973 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
1974 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
1976 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
1977 if (ds
->ds_quota
!= 0) {
1979 * Adjust available bytes according to refquota
1981 if (*refdbytesp
< ds
->ds_quota
)
1982 *availbytesp
= MIN(*availbytesp
,
1983 ds
->ds_quota
- *refdbytesp
);
1987 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1988 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
1989 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1990 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
1994 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
1996 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
1999 ASSERT(dsl_pool_config_held(dp
));
2002 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2003 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2004 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2005 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2006 objset_t
*os
, *os_snap
;
2008 * It may be that only the ZIL differs, because it was
2009 * reset in the head. Don't count that as being
2012 if (dmu_objset_from_ds(ds
, &os
) != 0)
2014 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2016 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2017 &os_snap
->os_phys
->os_meta_dnode
,
2018 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2023 typedef struct dsl_dataset_rename_snapshot_arg
{
2024 const char *ddrsa_fsname
;
2025 const char *ddrsa_oldsnapname
;
2026 const char *ddrsa_newsnapname
;
2027 boolean_t ddrsa_recursive
;
2029 } dsl_dataset_rename_snapshot_arg_t
;
2033 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2034 dsl_dataset_t
*hds
, void *arg
)
2036 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2040 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2042 /* ignore nonexistent snapshots */
2043 return (error
== ENOENT
? 0 : error
);
2046 /* new name should not exist */
2047 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2049 error
= SET_ERROR(EEXIST
);
2050 else if (error
== ENOENT
)
2053 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2054 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2055 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2056 error
= SET_ERROR(ENAMETOOLONG
);
2062 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2064 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2065 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2069 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2073 if (ddrsa
->ddrsa_recursive
) {
2074 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2075 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2078 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2080 dsl_dataset_rele(hds
, FTAG
);
2085 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2086 dsl_dataset_t
*hds
, void *arg
)
2088 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2091 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2094 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2095 ASSERT(error
== 0 || error
== ENOENT
);
2096 if (error
== ENOENT
) {
2097 /* ignore nonexistent snapshots */
2101 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2103 /* log before we change the name */
2104 spa_history_log_internal_ds(ds
, "rename", tx
,
2105 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2107 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2109 mutex_enter(&ds
->ds_lock
);
2110 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2111 sizeof (ds
->ds_snapname
));
2112 mutex_exit(&ds
->ds_lock
);
2113 VERIFY0(zap_add(dp
->dp_meta_objset
,
2114 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2115 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2116 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2117 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2119 dsl_dataset_rele(ds
, FTAG
);
2124 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2126 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2127 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2128 dsl_dataset_t
*hds
= NULL
;
2130 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2131 ddrsa
->ddrsa_tx
= tx
;
2132 if (ddrsa
->ddrsa_recursive
) {
2133 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2134 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2137 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2139 dsl_dataset_rele(hds
, FTAG
);
2143 dsl_dataset_rename_snapshot(const char *fsname
,
2144 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2146 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2148 ddrsa
.ddrsa_fsname
= fsname
;
2149 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2150 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2151 ddrsa
.ddrsa_recursive
= recursive
;
2153 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2154 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2155 1, ZFS_SPACE_CHECK_RESERVED
));
2159 * If we're doing an ownership handoff, we need to make sure that there is
2160 * only one long hold on the dataset. We're not allowed to change anything here
2161 * so we don't permanently release the long hold or regular hold here. We want
2162 * to do this only when syncing to avoid the dataset unexpectedly going away
2163 * when we release the long hold.
2166 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2170 if (!dmu_tx_is_syncing(tx
))
2173 if (owner
!= NULL
) {
2174 VERIFY3P(ds
->ds_owner
, ==, owner
);
2175 dsl_dataset_long_rele(ds
, owner
);
2178 held
= dsl_dataset_long_held(ds
);
2181 dsl_dataset_long_hold(ds
, owner
);
2184 return (SET_ERROR(EBUSY
));
2189 typedef struct dsl_dataset_rollback_arg
{
2190 const char *ddra_fsname
;
2192 nvlist_t
*ddra_result
;
2193 } dsl_dataset_rollback_arg_t
;
2196 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2198 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2199 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2201 int64_t unused_refres_delta
;
2204 nvlist_t
*proprequest
, *bookmarks
;
2206 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2210 /* must not be a snapshot */
2211 if (ds
->ds_is_snapshot
) {
2212 dsl_dataset_rele(ds
, FTAG
);
2213 return (SET_ERROR(EINVAL
));
2216 /* must have a most recent snapshot */
2217 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2218 dsl_dataset_rele(ds
, FTAG
);
2219 return (SET_ERROR(EINVAL
));
2222 /* must not have any bookmarks after the most recent snapshot */
2223 proprequest
= fnvlist_alloc();
2224 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2225 bookmarks
= fnvlist_alloc();
2226 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2227 fnvlist_free(proprequest
);
2230 for (pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2231 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2233 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2234 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2235 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2236 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2237 fnvlist_free(bookmarks
);
2238 dsl_dataset_rele(ds
, FTAG
);
2239 return (SET_ERROR(EEXIST
));
2242 fnvlist_free(bookmarks
);
2244 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2246 dsl_dataset_rele(ds
, FTAG
);
2251 * Check if the snap we are rolling back to uses more than
2254 if (ds
->ds_quota
!= 0 &&
2255 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2256 dsl_dataset_rele(ds
, FTAG
);
2257 return (SET_ERROR(EDQUOT
));
2261 * When we do the clone swap, we will temporarily use more space
2262 * due to the refreservation (the head will no longer have any
2263 * unique space, so the entire amount of the refreservation will need
2264 * to be free). We will immediately destroy the clone, freeing
2265 * this space, but the freeing happens over many txg's.
2267 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2268 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2270 if (unused_refres_delta
> 0 &&
2271 unused_refres_delta
>
2272 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2273 dsl_dataset_rele(ds
, FTAG
);
2274 return (SET_ERROR(ENOSPC
));
2277 dsl_dataset_rele(ds
, FTAG
);
2282 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2284 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2285 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2286 dsl_dataset_t
*ds
, *clone
;
2288 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2290 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2292 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2293 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2295 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2296 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, tx
);
2298 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2300 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2301 dsl_dataset_zero_zil(ds
, tx
);
2303 dsl_destroy_head_sync_impl(clone
, tx
);
2305 dsl_dataset_rele(clone
, FTAG
);
2306 dsl_dataset_rele(ds
, FTAG
);
2310 * Rolls back the given filesystem or volume to the most recent snapshot.
2311 * The name of the most recent snapshot will be returned under key "target"
2312 * in the result nvlist.
2315 * - The existing dataset MUST be owned by the specified owner at entry
2316 * - Upon return, dataset will still be held by the same owner, whether we
2319 * This mode is required any time the existing filesystem is mounted. See
2320 * notes above zfs_suspend_fs() for further details.
2323 dsl_dataset_rollback(const char *fsname
, void *owner
, nvlist_t
*result
)
2325 dsl_dataset_rollback_arg_t ddra
;
2327 ddra
.ddra_fsname
= fsname
;
2328 ddra
.ddra_owner
= owner
;
2329 ddra
.ddra_result
= result
;
2331 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2332 dsl_dataset_rollback_sync
, &ddra
,
2333 1, ZFS_SPACE_CHECK_RESERVED
));
2336 struct promotenode
{
2341 typedef struct dsl_dataset_promote_arg
{
2342 const char *ddpa_clonename
;
2343 dsl_dataset_t
*ddpa_clone
;
2344 list_t shared_snaps
, origin_snaps
, clone_snaps
;
2345 dsl_dataset_t
*origin_origin
; /* origin of the origin */
2346 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
2349 } dsl_dataset_promote_arg_t
;
2351 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2352 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2354 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2357 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2359 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2360 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2362 struct promotenode
*snap
;
2363 dsl_dataset_t
*origin_ds
;
2367 size_t max_snap_len
;
2369 err
= promote_hold(ddpa
, dp
, FTAG
);
2373 hds
= ddpa
->ddpa_clone
;
2374 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2376 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2377 promote_rele(ddpa
, FTAG
);
2378 return (SET_ERROR(EXDEV
));
2382 * Compute and check the amount of space to transfer. Since this is
2383 * so expensive, don't do the preliminary check.
2385 if (!dmu_tx_is_syncing(tx
)) {
2386 promote_rele(ddpa
, FTAG
);
2390 snap
= list_head(&ddpa
->shared_snaps
);
2392 err
= SET_ERROR(ENOENT
);
2395 origin_ds
= snap
->ds
;
2397 /* compute origin's new unique space */
2398 snap
= list_tail(&ddpa
->clone_snaps
);
2399 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2400 origin_ds
->ds_object
);
2401 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2402 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2403 &ddpa
->unique
, &unused
, &unused
);
2406 * Walk the snapshots that we are moving
2408 * Compute space to transfer. Consider the incremental changes
2409 * to used by each snapshot:
2410 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2411 * So each snapshot gave birth to:
2412 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2413 * So a sequence would look like:
2414 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2415 * Which simplifies to:
2416 * uN + kN + kN-1 + ... + k1 + k0
2417 * Note however, if we stop before we reach the ORIGIN we get:
2418 * uN + kN + kN-1 + ... + kM - uM-1
2421 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2422 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2423 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2424 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2425 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2426 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2427 dsl_dataset_t
*ds
= snap
->ds
;
2432 * If there are long holds, we won't be able to evict
2435 if (dsl_dataset_long_held(ds
)) {
2436 err
= SET_ERROR(EBUSY
);
2440 /* Check that the snapshot name does not conflict */
2441 VERIFY0(dsl_dataset_get_snapname(ds
));
2442 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2443 err
= SET_ERROR(ENAMETOOLONG
);
2446 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2448 (void) strcpy(ddpa
->err_ds
, snap
->ds
->ds_snapname
);
2449 err
= SET_ERROR(EEXIST
);
2455 /* The very first snapshot does not have a deadlist */
2456 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2459 dsl_deadlist_space(&ds
->ds_deadlist
,
2460 &dlused
, &dlcomp
, &dluncomp
);
2461 ddpa
->used
+= dlused
;
2462 ddpa
->comp
+= dlcomp
;
2463 ddpa
->uncomp
+= dluncomp
;
2467 * If we are a clone of a clone then we never reached ORIGIN,
2468 * so we need to subtract out the clone origin's used space.
2470 if (ddpa
->origin_origin
) {
2472 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2474 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2476 dsl_dataset_phys(ddpa
->origin_origin
)->
2477 ds_uncompressed_bytes
;
2480 /* Check that there is enough space and limit headroom here */
2481 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2482 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2487 * Compute the amounts of space that will be used by snapshots
2488 * after the promotion (for both origin and clone). For each,
2489 * it is the amount of space that will be on all of their
2490 * deadlists (that was not born before their new origin).
2492 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2496 * Note, typically this will not be a clone of a clone,
2497 * so dd_origin_txg will be < TXG_INITIAL, so
2498 * these snaplist_space() -> dsl_deadlist_space_range()
2499 * calls will be fast because they do not have to
2500 * iterate over all bps.
2502 snap
= list_head(&ddpa
->origin_snaps
);
2504 err
= SET_ERROR(ENOENT
);
2507 err
= snaplist_space(&ddpa
->shared_snaps
,
2508 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2512 err
= snaplist_space(&ddpa
->clone_snaps
,
2513 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2516 ddpa
->cloneusedsnap
+= space
;
2518 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2519 DD_FLAG_USED_BREAKDOWN
) {
2520 err
= snaplist_space(&ddpa
->origin_snaps
,
2521 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2522 &ddpa
->originusedsnap
);
2528 promote_rele(ddpa
, FTAG
);
2533 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2535 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2536 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2538 struct promotenode
*snap
;
2539 dsl_dataset_t
*origin_ds
;
2540 dsl_dataset_t
*origin_head
;
2542 dsl_dir_t
*odd
= NULL
;
2543 uint64_t oldnext_obj
;
2546 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2547 hds
= ddpa
->ddpa_clone
;
2549 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2551 snap
= list_head(&ddpa
->shared_snaps
);
2552 origin_ds
= snap
->ds
;
2555 snap
= list_head(&ddpa
->origin_snaps
);
2556 origin_head
= snap
->ds
;
2559 * We need to explicitly open odd, since origin_ds's dd will be
2562 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2565 /* change origin's next snap */
2566 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2567 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2568 snap
= list_tail(&ddpa
->clone_snaps
);
2569 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2570 origin_ds
->ds_object
);
2571 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2573 /* change the origin's next clone */
2574 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2575 dsl_dataset_remove_from_next_clones(origin_ds
,
2576 snap
->ds
->ds_object
, tx
);
2577 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2578 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2583 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2584 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2585 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2586 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2587 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2588 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2589 origin_head
->ds_dir
->dd_origin_txg
=
2590 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2592 /* change dd_clone entries */
2593 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2594 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2595 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2596 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2597 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2598 hds
->ds_object
, tx
));
2600 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2601 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2602 origin_head
->ds_object
, tx
));
2603 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2604 dsl_dir_phys(dd
)->dd_clones
=
2605 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2606 DMU_OT_NONE
, 0, tx
);
2608 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2609 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2612 /* move snapshots to this dir */
2613 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2614 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2615 dsl_dataset_t
*ds
= snap
->ds
;
2618 * Property callbacks are registered to a particular
2619 * dsl_dir. Since ours is changing, evict the objset
2620 * so that they will be unregistered from the old dsl_dir.
2622 if (ds
->ds_objset
) {
2623 dmu_objset_evict(ds
->ds_objset
);
2624 ds
->ds_objset
= NULL
;
2627 /* move snap name entry */
2628 VERIFY0(dsl_dataset_get_snapname(ds
));
2629 VERIFY0(dsl_dataset_snap_remove(origin_head
,
2630 ds
->ds_snapname
, tx
, B_TRUE
));
2631 VERIFY0(zap_add(dp
->dp_meta_objset
,
2632 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
2633 8, 1, &ds
->ds_object
, tx
));
2634 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
2635 DD_FIELD_SNAPSHOT_COUNT
, tx
);
2637 /* change containing dsl_dir */
2638 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2639 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
2640 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
2641 ASSERT3P(ds
->ds_dir
, ==, odd
);
2642 dsl_dir_rele(ds
->ds_dir
, ds
);
2643 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
2644 NULL
, ds
, &ds
->ds_dir
));
2646 /* move any clone references */
2647 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
2648 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2652 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
2653 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2654 zap_cursor_retrieve(&zc
, &za
) == 0;
2655 zap_cursor_advance(&zc
)) {
2656 dsl_dataset_t
*cnds
;
2659 if (za
.za_first_integer
== oldnext_obj
) {
2661 * We've already moved the
2662 * origin's reference.
2667 VERIFY0(dsl_dataset_hold_obj(dp
,
2668 za
.za_first_integer
, FTAG
, &cnds
));
2669 o
= dsl_dir_phys(cnds
->ds_dir
)->
2670 dd_head_dataset_obj
;
2672 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2673 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
2674 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2675 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
2676 dsl_dataset_rele(cnds
, FTAG
);
2678 zap_cursor_fini(&zc
);
2681 ASSERT(!dsl_prop_hascb(ds
));
2685 * Change space accounting.
2686 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2687 * both be valid, or both be 0 (resulting in delta == 0). This
2688 * is true for each of {clone,origin} independently.
2691 delta
= ddpa
->cloneusedsnap
-
2692 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
2693 ASSERT3S(delta
, >=, 0);
2694 ASSERT3U(ddpa
->used
, >=, delta
);
2695 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2696 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
2697 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
2699 delta
= ddpa
->originusedsnap
-
2700 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
2701 ASSERT3S(delta
, <=, 0);
2702 ASSERT3U(ddpa
->used
, >=, -delta
);
2703 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2704 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
2705 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
2707 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
2709 /* log history record */
2710 spa_history_log_internal_ds(hds
, "promote", tx
, "");
2712 dsl_dir_rele(odd
, FTAG
);
2713 promote_rele(ddpa
, FTAG
);
2717 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2718 * (exclusive) and last_obj (inclusive). The list will be in reverse
2719 * order (last_obj will be the list_head()). If first_obj == 0, do all
2720 * snapshots back to this dataset's origin.
2723 snaplist_make(dsl_pool_t
*dp
,
2724 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
2726 uint64_t obj
= last_obj
;
2728 list_create(l
, sizeof (struct promotenode
),
2729 offsetof(struct promotenode
, link
));
2731 while (obj
!= first_obj
) {
2733 struct promotenode
*snap
;
2736 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
2737 ASSERT(err
!= ENOENT
);
2742 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
2744 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
2746 list_insert_tail(l
, snap
);
2747 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
2754 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
2756 struct promotenode
*snap
;
2759 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
2760 uint64_t used
, comp
, uncomp
;
2761 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2762 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
2769 snaplist_destroy(list_t
*l
, void *tag
)
2771 struct promotenode
*snap
;
2773 if (l
== NULL
|| !list_link_active(&l
->list_head
))
2776 while ((snap
= list_tail(l
)) != NULL
) {
2777 list_remove(l
, snap
);
2778 dsl_dataset_rele(snap
->ds
, tag
);
2779 kmem_free(snap
, sizeof (*snap
));
2785 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
2789 struct promotenode
*snap
;
2791 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
2795 dd
= ddpa
->ddpa_clone
->ds_dir
;
2797 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
2798 !dsl_dir_is_clone(dd
)) {
2799 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2800 return (SET_ERROR(EINVAL
));
2803 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
2804 &ddpa
->shared_snaps
, tag
);
2808 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
2809 &ddpa
->clone_snaps
, tag
);
2813 snap
= list_head(&ddpa
->shared_snaps
);
2814 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
2815 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
2816 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
2817 &ddpa
->origin_snaps
, tag
);
2821 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
2822 error
= dsl_dataset_hold_obj(dp
,
2823 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
2824 tag
, &ddpa
->origin_origin
);
2830 promote_rele(ddpa
, tag
);
2835 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
2837 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
2838 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
2839 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
2840 if (ddpa
->origin_origin
!= NULL
)
2841 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
2842 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2848 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2849 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
2852 dsl_dataset_promote(const char *name
, char *conflsnap
)
2854 dsl_dataset_promote_arg_t ddpa
= { 0 };
2860 * We will modify space proportional to the number of
2861 * snapshots. Compute numsnaps.
2863 error
= dmu_objset_hold(name
, FTAG
, &os
);
2866 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
2867 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
2869 dmu_objset_rele(os
, FTAG
);
2873 ddpa
.ddpa_clonename
= name
;
2874 ddpa
.err_ds
= conflsnap
;
2877 return (dsl_sync_task(name
, dsl_dataset_promote_check
,
2878 dsl_dataset_promote_sync
, &ddpa
,
2879 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
));
2883 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
2884 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
2887 * "slack" factor for received datasets with refquota set on them.
2888 * See the bottom of this function for details on its use.
2890 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
2891 spa_asize_inflation
;
2892 int64_t unused_refres_delta
;
2894 /* they should both be heads */
2895 if (clone
->ds_is_snapshot
||
2896 origin_head
->ds_is_snapshot
)
2897 return (SET_ERROR(EINVAL
));
2899 /* if we are not forcing, the branch point should be just before them */
2900 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
2901 return (SET_ERROR(EINVAL
));
2903 /* clone should be the clone (unless they are unrelated) */
2904 if (clone
->ds_prev
!= NULL
&&
2905 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
2906 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
2907 return (SET_ERROR(EINVAL
));
2909 /* the clone should be a child of the origin */
2910 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
2911 return (SET_ERROR(EINVAL
));
2913 /* origin_head shouldn't be modified unless 'force' */
2915 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
2916 return (SET_ERROR(ETXTBSY
));
2918 /* origin_head should have no long holds (e.g. is not mounted) */
2919 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
2920 return (SET_ERROR(EBUSY
));
2922 /* check amount of any unconsumed refreservation */
2923 unused_refres_delta
=
2924 (int64_t)MIN(origin_head
->ds_reserved
,
2925 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2926 (int64_t)MIN(origin_head
->ds_reserved
,
2927 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2929 if (unused_refres_delta
> 0 &&
2930 unused_refres_delta
>
2931 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
2932 return (SET_ERROR(ENOSPC
));
2935 * The clone can't be too much over the head's refquota.
2937 * To ensure that the entire refquota can be used, we allow one
2938 * transaction to exceed the the refquota. Therefore, this check
2939 * needs to also allow for the space referenced to be more than the
2940 * refquota. The maximum amount of space that one transaction can use
2941 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
2942 * overage ensures that we are able to receive a filesystem that
2943 * exceeds the refquota on the source system.
2945 * So that overage is the refquota_slack we use below.
2947 if (origin_head
->ds_quota
!= 0 &&
2948 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
2949 origin_head
->ds_quota
+ refquota_slack
)
2950 return (SET_ERROR(EDQUOT
));
2956 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
2957 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
2960 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2961 int64_t unused_refres_delta
;
2964 ASSERT(clone
->ds_reserved
== 0);
2966 * NOTE: On DEBUG kernels there could be a race between this and
2967 * the check function if spa_asize_inflation is adjusted...
2969 ASSERT(origin_head
->ds_quota
== 0 ||
2970 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
2971 DMU_MAX_ACCESS
* spa_asize_inflation
);
2972 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
2975 * Swap per-dataset feature flags.
2977 for (f
= 0; f
< SPA_FEATURES
; f
++) {
2978 boolean_t clone_inuse
;
2979 boolean_t origin_head_inuse
;
2981 if (!(spa_feature_table
[f
].fi_flags
&
2982 ZFEATURE_FLAG_PER_DATASET
)) {
2983 ASSERT(!clone
->ds_feature_inuse
[f
]);
2984 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
2988 clone_inuse
= clone
->ds_feature_inuse
[f
];
2989 origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
2992 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
2993 clone
->ds_feature_inuse
[f
] = B_FALSE
;
2995 if (origin_head_inuse
) {
2996 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
2998 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
3001 dsl_dataset_activate_feature(origin_head
->ds_object
,
3003 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
3005 if (origin_head_inuse
) {
3006 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
3007 clone
->ds_feature_inuse
[f
] = B_TRUE
;
3011 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3012 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3014 if (clone
->ds_objset
!= NULL
) {
3015 dmu_objset_evict(clone
->ds_objset
);
3016 clone
->ds_objset
= NULL
;
3019 if (origin_head
->ds_objset
!= NULL
) {
3020 dmu_objset_evict(origin_head
->ds_objset
);
3021 origin_head
->ds_objset
= NULL
;
3024 unused_refres_delta
=
3025 (int64_t)MIN(origin_head
->ds_reserved
,
3026 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3027 (int64_t)MIN(origin_head
->ds_reserved
,
3028 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3031 * Reset origin's unique bytes, if it exists.
3033 if (clone
->ds_prev
) {
3034 dsl_dataset_t
*origin
= clone
->ds_prev
;
3035 uint64_t comp
, uncomp
;
3037 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3038 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3039 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3040 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3045 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3046 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3047 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3048 dsl_dataset_phys(origin_head
)->ds_bp
=
3049 dsl_dataset_phys(clone
)->ds_bp
;
3050 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3051 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3052 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3055 /* set dd_*_bytes */
3057 int64_t dused
, dcomp
, duncomp
;
3058 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3059 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3061 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3062 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3064 dsl_deadlist_space(&clone
->ds_deadlist
,
3065 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3066 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3067 &odl_used
, &odl_comp
, &odl_uncomp
);
3069 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3071 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3073 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3075 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3077 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3079 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3082 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3083 dused
, dcomp
, duncomp
, tx
);
3084 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3085 -dused
, -dcomp
, -duncomp
, tx
);
3088 * The difference in the space used by snapshots is the
3089 * difference in snapshot space due to the head's
3090 * deadlist (since that's the only thing that's
3091 * changing that affects the snapused).
3093 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3094 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3095 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3096 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3097 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3098 &odl_used
, &odl_comp
, &odl_uncomp
);
3099 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3100 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3103 /* swap ds_*_bytes */
3104 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3105 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3106 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3107 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3108 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3109 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3110 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3111 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3113 /* apply any parent delta for change in unconsumed refreservation */
3114 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3115 unused_refres_delta
, 0, 0, tx
);
3120 dsl_deadlist_close(&clone
->ds_deadlist
);
3121 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3122 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3123 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3124 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3125 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3126 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3127 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3129 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3131 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3132 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3136 * Given a pool name and a dataset object number in that pool,
3137 * return the name of that dataset.
3140 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3146 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3150 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3152 dsl_dataset_name(ds
, buf
);
3153 dsl_dataset_rele(ds
, FTAG
);
3155 dsl_pool_rele(dp
, FTAG
);
3161 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3162 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3166 ASSERT3S(asize
, >, 0);
3169 * *ref_rsrv is the portion of asize that will come from any
3170 * unconsumed refreservation space.
3174 mutex_enter(&ds
->ds_lock
);
3176 * Make a space adjustment for reserved bytes.
3178 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3180 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3182 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3184 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3187 if (!check_quota
|| ds
->ds_quota
== 0) {
3188 mutex_exit(&ds
->ds_lock
);
3192 * If they are requesting more space, and our current estimate
3193 * is over quota, they get to try again unless the actual
3194 * on-disk is over quota and there are no pending changes (which
3195 * may free up space for us).
3197 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3200 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3201 error
= SET_ERROR(ERESTART
);
3203 error
= SET_ERROR(EDQUOT
);
3205 mutex_exit(&ds
->ds_lock
);
3210 typedef struct dsl_dataset_set_qr_arg
{
3211 const char *ddsqra_name
;
3212 zprop_source_t ddsqra_source
;
3213 uint64_t ddsqra_value
;
3214 } dsl_dataset_set_qr_arg_t
;
3219 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3221 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3222 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3227 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3228 return (SET_ERROR(ENOTSUP
));
3230 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3234 if (ds
->ds_is_snapshot
) {
3235 dsl_dataset_rele(ds
, FTAG
);
3236 return (SET_ERROR(EINVAL
));
3239 error
= dsl_prop_predict(ds
->ds_dir
,
3240 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3241 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3243 dsl_dataset_rele(ds
, FTAG
);
3248 dsl_dataset_rele(ds
, FTAG
);
3252 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3253 newval
< ds
->ds_reserved
) {
3254 dsl_dataset_rele(ds
, FTAG
);
3255 return (SET_ERROR(ENOSPC
));
3258 dsl_dataset_rele(ds
, FTAG
);
3263 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3265 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3266 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3267 dsl_dataset_t
*ds
= NULL
;
3270 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3272 dsl_prop_set_sync_impl(ds
,
3273 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3274 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3275 &ddsqra
->ddsqra_value
, tx
);
3277 VERIFY0(dsl_prop_get_int_ds(ds
,
3278 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3280 if (ds
->ds_quota
!= newval
) {
3281 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3282 ds
->ds_quota
= newval
;
3284 dsl_dataset_rele(ds
, FTAG
);
3288 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3291 dsl_dataset_set_qr_arg_t ddsqra
;
3293 ddsqra
.ddsqra_name
= dsname
;
3294 ddsqra
.ddsqra_source
= source
;
3295 ddsqra
.ddsqra_value
= refquota
;
3297 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3298 dsl_dataset_set_refquota_sync
, &ddsqra
, 0, ZFS_SPACE_CHECK_NONE
));
3302 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3304 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3305 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3308 uint64_t newval
, unique
;
3310 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3311 return (SET_ERROR(ENOTSUP
));
3313 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3317 if (ds
->ds_is_snapshot
) {
3318 dsl_dataset_rele(ds
, FTAG
);
3319 return (SET_ERROR(EINVAL
));
3322 error
= dsl_prop_predict(ds
->ds_dir
,
3323 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3324 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3326 dsl_dataset_rele(ds
, FTAG
);
3331 * If we are doing the preliminary check in open context, the
3332 * space estimates may be inaccurate.
3334 if (!dmu_tx_is_syncing(tx
)) {
3335 dsl_dataset_rele(ds
, FTAG
);
3339 mutex_enter(&ds
->ds_lock
);
3340 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3341 dsl_dataset_recalc_head_uniq(ds
);
3342 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3343 mutex_exit(&ds
->ds_lock
);
3345 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3346 uint64_t delta
= MAX(unique
, newval
) -
3347 MAX(unique
, ds
->ds_reserved
);
3350 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3351 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3352 dsl_dataset_rele(ds
, FTAG
);
3353 return (SET_ERROR(ENOSPC
));
3357 dsl_dataset_rele(ds
, FTAG
);
3362 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3363 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3369 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3370 source
, sizeof (value
), 1, &value
, tx
);
3372 VERIFY0(dsl_prop_get_int_ds(ds
,
3373 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3375 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3376 mutex_enter(&ds
->ds_dir
->dd_lock
);
3377 mutex_enter(&ds
->ds_lock
);
3378 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3379 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3380 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3381 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3382 ds
->ds_reserved
= newval
;
3383 mutex_exit(&ds
->ds_lock
);
3385 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3386 mutex_exit(&ds
->ds_dir
->dd_lock
);
3390 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3392 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3393 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3394 dsl_dataset_t
*ds
= NULL
;
3396 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3397 dsl_dataset_set_refreservation_sync_impl(ds
,
3398 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3399 dsl_dataset_rele(ds
, FTAG
);
3403 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3404 uint64_t refreservation
)
3406 dsl_dataset_set_qr_arg_t ddsqra
;
3408 ddsqra
.ddsqra_name
= dsname
;
3409 ddsqra
.ddsqra_source
= source
;
3410 ddsqra
.ddsqra_value
= refreservation
;
3412 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3413 dsl_dataset_set_refreservation_sync
, &ddsqra
,
3414 0, ZFS_SPACE_CHECK_NONE
));
3418 * Return (in *usedp) the amount of space written in new that is not
3419 * present in oldsnap. New may be a snapshot or the head. Old must be
3420 * a snapshot before new, in new's filesystem (or its origin). If not then
3421 * fail and return EINVAL.
3423 * The written space is calculated by considering two components: First, we
3424 * ignore any freed space, and calculate the written as new's used space
3425 * minus old's used space. Next, we add in the amount of space that was freed
3426 * between the two snapshots, thus reducing new's used space relative to old's.
3427 * Specifically, this is the space that was born before old->ds_creation_txg,
3428 * and freed before new (ie. on new's deadlist or a previous deadlist).
3430 * space freed [---------------------]
3431 * snapshots ---O-------O--------O-------O------
3435 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3436 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3440 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3442 ASSERT(dsl_pool_config_held(dp
));
3445 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3446 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3449 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3450 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3453 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3454 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3456 snapobj
= new->ds_object
;
3457 while (snapobj
!= oldsnap
->ds_object
) {
3458 dsl_dataset_t
*snap
;
3459 uint64_t used
, comp
, uncomp
;
3461 if (snapobj
== new->ds_object
) {
3464 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3469 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3470 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3472 * The blocks in the deadlist can not be born after
3473 * ds_prev_snap_txg, so get the whole deadlist space,
3474 * which is more efficient (especially for old-format
3475 * deadlists). Unfortunately the deadlist code
3476 * doesn't have enough information to make this
3477 * optimization itself.
3479 dsl_deadlist_space(&snap
->ds_deadlist
,
3480 &used
, &comp
, &uncomp
);
3482 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3483 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3484 &used
, &comp
, &uncomp
);
3491 * If we get to the beginning of the chain of snapshots
3492 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3493 * was not a snapshot of/before new.
3495 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3497 dsl_dataset_rele(snap
, FTAG
);
3499 err
= SET_ERROR(EINVAL
);
3508 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3509 * lastsnap, and all snapshots in between are deleted.
3511 * blocks that would be freed [---------------------------]
3512 * snapshots ---O-------O--------O-------O--------O
3513 * firstsnap lastsnap
3515 * This is the set of blocks that were born after the snap before firstsnap,
3516 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3517 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3518 * We calculate this by iterating over the relevant deadlists (from the snap
3519 * after lastsnap, backward to the snap after firstsnap), summing up the
3520 * space on the deadlist that was born after the snap before firstsnap.
3523 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3524 dsl_dataset_t
*lastsnap
,
3525 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3529 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3531 ASSERT(firstsnap
->ds_is_snapshot
);
3532 ASSERT(lastsnap
->ds_is_snapshot
);
3535 * Check that the snapshots are in the same dsl_dir, and firstsnap
3536 * is before lastsnap.
3538 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3539 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3540 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3541 return (SET_ERROR(EINVAL
));
3543 *usedp
= *compp
= *uncompp
= 0;
3545 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3546 while (snapobj
!= firstsnap
->ds_object
) {
3548 uint64_t used
, comp
, uncomp
;
3550 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3554 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3555 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3556 &used
, &comp
, &uncomp
);
3561 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3562 ASSERT3U(snapobj
, !=, 0);
3563 dsl_dataset_rele(ds
, FTAG
);
3569 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3570 * For example, they could both be snapshots of the same filesystem, and
3571 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3572 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3573 * filesystem. Or 'earlier' could be the origin's origin.
3575 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3578 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3579 uint64_t earlier_txg
)
3581 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3584 dsl_dataset_t
*origin
;
3586 ASSERT(dsl_pool_config_held(dp
));
3587 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3589 if (earlier_txg
== 0)
3590 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3592 if (later
->ds_is_snapshot
&&
3593 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3596 if (later
->ds_dir
== earlier
->ds_dir
)
3598 if (!dsl_dir_is_clone(later
->ds_dir
))
3601 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3603 error
= dsl_dataset_hold_obj(dp
,
3604 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3607 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3608 dsl_dataset_rele(origin
, FTAG
);
3613 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
3615 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
3616 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
3620 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
3622 dmu_object_info_t doi
;
3624 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
3625 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
3629 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
3631 return (dsl_dataset_is_zapified(ds
) &&
3632 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
3633 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
3636 #if defined(_KERNEL) && defined(HAVE_SPL)
3638 module_param(zfs_max_recordsize
, int, 0644);
3639 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3641 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
3642 module_param(zfs_max_recordsize
, int, 0444);
3643 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3646 EXPORT_SYMBOL(dsl_dataset_hold
);
3647 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
3648 EXPORT_SYMBOL(dsl_dataset_own
);
3649 EXPORT_SYMBOL(dsl_dataset_own_obj
);
3650 EXPORT_SYMBOL(dsl_dataset_name
);
3651 EXPORT_SYMBOL(dsl_dataset_rele
);
3652 EXPORT_SYMBOL(dsl_dataset_disown
);
3653 EXPORT_SYMBOL(dsl_dataset_tryown
);
3654 EXPORT_SYMBOL(dsl_dataset_create_sync
);
3655 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
3656 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
3657 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
3658 EXPORT_SYMBOL(dsl_dataset_promote
);
3659 EXPORT_SYMBOL(dsl_dataset_user_hold
);
3660 EXPORT_SYMBOL(dsl_dataset_user_release
);
3661 EXPORT_SYMBOL(dsl_dataset_get_holds
);
3662 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
3663 EXPORT_SYMBOL(dsl_dataset_get_spa
);
3664 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
3665 EXPORT_SYMBOL(dsl_dataset_space_written
);
3666 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
3667 EXPORT_SYMBOL(dsl_dataset_sync
);
3668 EXPORT_SYMBOL(dsl_dataset_block_born
);
3669 EXPORT_SYMBOL(dsl_dataset_block_kill
);
3670 EXPORT_SYMBOL(dsl_dataset_block_freeable
);
3671 EXPORT_SYMBOL(dsl_dataset_prev_snap_txg
);
3672 EXPORT_SYMBOL(dsl_dataset_dirty
);
3673 EXPORT_SYMBOL(dsl_dataset_stats
);
3674 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
3675 EXPORT_SYMBOL(dsl_dataset_space
);
3676 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
3677 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
3678 EXPORT_SYMBOL(dsl_dataset_check_quota
);
3679 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
3680 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);