4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
24 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
25 * Copyright (c) 2014 RackTop Systems.
26 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
27 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
28 * Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
31 #include <sys/dmu_objset.h>
32 #include <sys/dsl_dataset.h>
33 #include <sys/dsl_dir.h>
34 #include <sys/dsl_prop.h>
35 #include <sys/dsl_synctask.h>
36 #include <sys/dmu_traverse.h>
37 #include <sys/dmu_impl.h>
38 #include <sys/dmu_tx.h>
42 #include <sys/zfeature.h>
43 #include <sys/unique.h>
44 #include <sys/zfs_context.h>
45 #include <sys/zfs_ioctl.h>
47 #include <sys/zfs_znode.h>
48 #include <sys/zfs_onexit.h>
50 #include <sys/dsl_scan.h>
51 #include <sys/dsl_deadlist.h>
52 #include <sys/dsl_destroy.h>
53 #include <sys/dsl_userhold.h>
54 #include <sys/dsl_bookmark.h>
55 #include <sys/policy.h>
56 #include <sys/dmu_send.h>
57 #include <sys/zio_compress.h>
58 #include <zfs_fletcher.h>
59 #include <sys/zio_checksum.h>
62 * The SPA supports block sizes up to 16MB. However, very large blocks
63 * can have an impact on i/o latency (e.g. tying up a spinning disk for
64 * ~300ms), and also potentially on the memory allocator. Therefore,
65 * we do not allow the recordsize to be set larger than zfs_max_recordsize
66 * (default 1MB). Larger blocks can be created by changing this tunable,
67 * and pools with larger blocks can always be imported and used, regardless
70 int zfs_max_recordsize
= 1 * 1024 * 1024;
72 #define SWITCH64(x, y) \
74 uint64_t __tmp = (x); \
79 #define DS_REF_MAX (1ULL << 62)
81 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
83 extern int spa_asize_inflation
;
86 * Figure out how much of this delta should be propagated to the dsl_dir
87 * layer. If there's a refreservation, that space has already been
88 * partially accounted for in our ancestors.
91 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
93 dsl_dataset_phys_t
*ds_phys
;
94 uint64_t old_bytes
, new_bytes
;
96 if (ds
->ds_reserved
== 0)
99 ds_phys
= dsl_dataset_phys(ds
);
100 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
101 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
103 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
104 return (new_bytes
- old_bytes
);
108 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
110 int used
, compressed
, uncompressed
;
114 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
115 compressed
= BP_GET_PSIZE(bp
);
116 uncompressed
= BP_GET_UCSIZE(bp
);
118 dprintf_bp(bp
, "ds=%p", ds
);
120 ASSERT(dmu_tx_is_syncing(tx
));
121 /* It could have been compressed away to nothing */
124 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
125 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
127 dsl_pool_mos_diduse_space(tx
->tx_pool
,
128 used
, compressed
, uncompressed
);
132 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
133 mutex_enter(&ds
->ds_lock
);
134 delta
= parent_delta(ds
, used
);
135 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
136 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
137 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
138 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
140 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
141 ds
->ds_feature_activation_needed
[SPA_FEATURE_LARGE_BLOCKS
] =
145 f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
146 if (f
!= SPA_FEATURE_NONE
)
147 ds
->ds_feature_activation_needed
[f
] = B_TRUE
;
149 mutex_exit(&ds
->ds_lock
);
150 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
151 compressed
, uncompressed
, tx
);
152 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
153 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
157 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
160 int used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
161 int compressed
= BP_GET_PSIZE(bp
);
162 int uncompressed
= BP_GET_UCSIZE(bp
);
167 ASSERT(dmu_tx_is_syncing(tx
));
168 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
171 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
172 dsl_pool_mos_diduse_space(tx
->tx_pool
,
173 -used
, -compressed
, -uncompressed
);
176 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
178 ASSERT(!ds
->ds_is_snapshot
);
179 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
181 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
184 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
185 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
187 mutex_enter(&ds
->ds_lock
);
188 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
189 !DS_UNIQUE_IS_ACCURATE(ds
));
190 delta
= parent_delta(ds
, -used
);
191 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
192 mutex_exit(&ds
->ds_lock
);
193 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
194 delta
, -compressed
, -uncompressed
, tx
);
195 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
196 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
198 dprintf_bp(bp
, "putting on dead list: %s", "");
201 * We are here as part of zio's write done callback,
202 * which means we're a zio interrupt thread. We can't
203 * call dsl_deadlist_insert() now because it may block
204 * waiting for I/O. Instead, put bp on the deferred
205 * queue and let dsl_pool_sync() finish the job.
207 bplist_append(&ds
->ds_pending_deadlist
, bp
);
209 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
211 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
212 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
213 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
214 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
215 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
216 ds
->ds_object
&& bp
->blk_birth
>
217 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
218 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
219 mutex_enter(&ds
->ds_prev
->ds_lock
);
220 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
221 mutex_exit(&ds
->ds_prev
->ds_lock
);
223 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
224 dsl_dir_transfer_space(ds
->ds_dir
, used
,
225 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
228 mutex_enter(&ds
->ds_lock
);
229 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
230 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
231 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
232 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
233 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
234 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
235 mutex_exit(&ds
->ds_lock
);
241 dsl_dataset_prev_snap_txg(dsl_dataset_t
*ds
)
243 uint64_t trysnap
= 0;
248 * The snapshot creation could fail, but that would cause an
249 * incorrect FALSE return, which would only result in an
250 * overestimation of the amount of space that an operation would
251 * consume, which is OK.
253 * There's also a small window where we could miss a pending
254 * snapshot, because we could set the sync task in the quiescing
255 * phase. So this should only be used as a guess.
257 if (ds
->ds_trysnap_txg
>
258 spa_last_synced_txg(ds
->ds_dir
->dd_pool
->dp_spa
))
259 trysnap
= ds
->ds_trysnap_txg
;
260 return (MAX(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, trysnap
));
264 dsl_dataset_block_freeable(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
267 if (blk_birth
<= dsl_dataset_prev_snap_txg(ds
) ||
268 (bp
!= NULL
&& BP_IS_HOLE(bp
)))
271 ddt_prefetch(dsl_dataset_get_spa(ds
), bp
);
277 * We have to release the fsid syncronously or we risk that a subsequent
278 * mount of the same dataset will fail to unique_insert the fsid. This
279 * failure would manifest itself as the fsid of this dataset changing
280 * between mounts which makes NFS clients quite unhappy.
283 dsl_dataset_evict_sync(void *dbu
)
285 dsl_dataset_t
*ds
= dbu
;
287 ASSERT(ds
->ds_owner
== NULL
);
289 unique_remove(ds
->ds_fsid_guid
);
293 dsl_dataset_evict_async(void *dbu
)
295 dsl_dataset_t
*ds
= dbu
;
297 ASSERT(ds
->ds_owner
== NULL
);
301 if (ds
->ds_objset
!= NULL
)
302 dmu_objset_evict(ds
->ds_objset
);
305 dsl_dataset_rele(ds
->ds_prev
, ds
);
309 bplist_destroy(&ds
->ds_pending_deadlist
);
310 if (ds
->ds_deadlist
.dl_os
!= NULL
)
311 dsl_deadlist_close(&ds
->ds_deadlist
);
313 dsl_dir_async_rele(ds
->ds_dir
, ds
);
315 ASSERT(!list_link_active(&ds
->ds_synced_link
));
317 list_destroy(&ds
->ds_prop_cbs
);
318 mutex_destroy(&ds
->ds_lock
);
319 mutex_destroy(&ds
->ds_opening_lock
);
320 mutex_destroy(&ds
->ds_sendstream_lock
);
321 refcount_destroy(&ds
->ds_longholds
);
323 kmem_free(ds
, sizeof (dsl_dataset_t
));
327 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
329 dsl_dataset_phys_t
*headphys
;
332 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
333 objset_t
*mos
= dp
->dp_meta_objset
;
335 if (ds
->ds_snapname
[0])
337 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
340 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
344 headphys
= headdbuf
->db_data
;
345 err
= zap_value_search(dp
->dp_meta_objset
,
346 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
347 if (err
!= 0 && zfs_recover
== B_TRUE
) {
349 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
350 "SNAPOBJ=%llu-ERR=%d",
351 (unsigned long long)ds
->ds_object
, err
);
353 dmu_buf_rele(headdbuf
, FTAG
);
358 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
360 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
361 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
365 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
370 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
371 value
, mt
, NULL
, 0, NULL
);
372 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
373 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
378 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
381 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
382 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
386 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
388 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
393 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
394 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
395 err
= zap_remove(mos
, snapobj
, name
, tx
);
397 if (err
== 0 && adj_cnt
)
398 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
399 DD_FIELD_SNAPSHOT_COUNT
, tx
);
405 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
407 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
408 boolean_t result
= B_FALSE
;
410 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
411 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
413 if (ds
== dmu_buf_get_user(dbuf
))
416 dmu_buf_rele(dbuf
, tag
);
423 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
426 objset_t
*mos
= dp
->dp_meta_objset
;
430 dmu_object_info_t doi
;
432 ASSERT(dsl_pool_config_held(dp
));
434 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
438 /* Make sure dsobj has the correct object type. */
439 dmu_object_info_from_db(dbuf
, &doi
);
440 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
441 dmu_buf_rele(dbuf
, tag
);
442 return (SET_ERROR(EINVAL
));
445 ds
= dmu_buf_get_user(dbuf
);
447 dsl_dataset_t
*winner
= NULL
;
449 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
451 ds
->ds_object
= dsobj
;
452 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
453 list_link_init(&ds
->ds_synced_link
);
455 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
456 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
457 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
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 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
868 * Inherit flags that describe the dataset's contents
869 * (INCONSISTENT) or properties (Case Insensitive).
871 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
872 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
874 for (f
= 0; f
< SPA_FEATURES
; f
++) {
875 if (origin
->ds_feature_inuse
[f
])
876 dsl_dataset_activate_feature(dsobj
, f
, tx
);
879 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
880 dsl_dataset_phys(origin
)->ds_num_children
++;
882 VERIFY0(dsl_dataset_hold_obj(dp
,
883 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
885 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
886 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
887 dsl_dataset_rele(ohds
, FTAG
);
889 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
890 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
891 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
893 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
895 VERIFY0(zap_add_int(mos
,
896 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
900 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
901 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
902 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
903 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
904 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
905 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
907 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
909 VERIFY0(zap_add_int(mos
,
910 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
915 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
916 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
918 dmu_buf_rele(dbuf
, FTAG
);
920 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
921 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
927 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
931 VERIFY0(dmu_objset_from_ds(ds
, &os
));
932 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
933 dsl_dataset_dirty(ds
, tx
);
937 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
938 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
, dmu_tx_t
*tx
)
940 dsl_pool_t
*dp
= pdd
->dd_pool
;
941 uint64_t dsobj
, ddobj
;
944 ASSERT(dmu_tx_is_syncing(tx
));
945 ASSERT(lastname
[0] != '@');
947 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
948 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
950 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
,
951 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
953 dsl_deleg_set_create_perms(dd
, tx
, cr
);
956 * Since we're creating a new node we know it's a leaf, so we can
957 * initialize the counts if the limit feature is active.
959 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
961 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
963 dsl_dir_zapify(dd
, tx
);
964 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
965 sizeof (cnt
), 1, &cnt
, tx
));
966 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
967 sizeof (cnt
), 1, &cnt
, tx
));
970 dsl_dir_rele(dd
, FTAG
);
973 * If we are creating a clone, make sure we zero out any stale
974 * data from the origin snapshots zil header.
976 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
979 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
980 dsl_dataset_zero_zil(ds
, tx
);
981 dsl_dataset_rele(ds
, FTAG
);
988 * The unique space in the head dataset can be calculated by subtracting
989 * the space used in the most recent snapshot, that is still being used
990 * in this file system, from the space currently in use. To figure out
991 * the space in the most recent snapshot still in use, we need to take
992 * the total space used in the snapshot and subtract out the space that
993 * has been freed up since the snapshot was taken.
996 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
999 uint64_t dlused
, dlcomp
, dluncomp
;
1001 ASSERT(!ds
->ds_is_snapshot
);
1003 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1004 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1008 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1010 ASSERT3U(dlused
, <=, mrs_used
);
1011 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1012 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1014 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1015 SPA_VERSION_UNIQUE_ACCURATE
)
1016 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1020 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1023 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1025 ASSERTV(uint64_t count
);
1027 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1028 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1031 * The err should not be ENOENT, but a bug in a previous version
1032 * of the code could cause upgrade_clones_cb() to not set
1033 * ds_next_snap_obj when it should, leading to a missing entry.
1034 * If we knew that the pool was created after
1035 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1036 * ENOENT. However, at least we can check that we don't have
1037 * too many entries in the next_clones_obj even after failing to
1042 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1044 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1049 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1051 return (&dsl_dataset_phys(ds
)->ds_bp
);
1055 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1057 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1061 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1065 if (ds
== NULL
) /* this is the meta-objset */
1068 ASSERT(ds
->ds_objset
!= NULL
);
1070 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1071 panic("dirtying snapshot!");
1073 dp
= ds
->ds_dir
->dd_pool
;
1075 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1076 /* up the hold count until we can be written out */
1077 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1082 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
1086 for (t
= 0; t
< TXG_SIZE
; t
++) {
1087 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1095 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1099 if (!dmu_tx_is_syncing(tx
))
1103 * If there's an fs-only reservation, any blocks that might become
1104 * owned by the snapshot dataset must be accommodated by space
1105 * outside of the reservation.
1107 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1108 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1109 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1110 return (SET_ERROR(ENOSPC
));
1113 * Propagate any reserved space for this snapshot to other
1114 * snapshot checks in this sync group.
1117 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1122 typedef struct dsl_dataset_snapshot_arg
{
1123 nvlist_t
*ddsa_snaps
;
1124 nvlist_t
*ddsa_props
;
1125 nvlist_t
*ddsa_errors
;
1127 } dsl_dataset_snapshot_arg_t
;
1130 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1131 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1136 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1138 if (!dmu_tx_is_syncing(tx
))
1142 * We don't allow multiple snapshots of the same txg. If there
1143 * is already one, try again.
1145 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1146 return (SET_ERROR(EAGAIN
));
1149 * Check for conflicting snapshot name.
1151 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1153 return (SET_ERROR(EEXIST
));
1154 if (error
!= ENOENT
)
1158 * We don't allow taking snapshots of inconsistent datasets, such as
1159 * those into which we are currently receiving. However, if we are
1160 * creating this snapshot as part of a receive, this check will be
1161 * executed atomically with respect to the completion of the receive
1162 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1163 * case we ignore this, knowing it will be fixed up for us shortly in
1164 * dmu_recv_end_sync().
1166 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1167 return (SET_ERROR(EBUSY
));
1170 * Skip the check for temporary snapshots or if we have already checked
1171 * the counts in dsl_dataset_snapshot_check. This means we really only
1172 * check the count here when we're receiving a stream.
1174 if (cnt
!= 0 && cr
!= NULL
) {
1175 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1176 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1181 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1189 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1191 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1192 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1197 * Pre-compute how many total new snapshots will be created for each
1198 * level in the tree and below. This is needed for validating the
1199 * snapshot limit when either taking a recursive snapshot or when
1200 * taking multiple snapshots.
1202 * The problem is that the counts are not actually adjusted when
1203 * we are checking, only when we finally sync. For a single snapshot,
1204 * this is easy, the count will increase by 1 at each node up the tree,
1205 * but its more complicated for the recursive/multiple snapshot case.
1207 * The dsl_fs_ss_limit_check function does recursively check the count
1208 * at each level up the tree but since it is validating each snapshot
1209 * independently we need to be sure that we are validating the complete
1210 * count for the entire set of snapshots. We do this by rolling up the
1211 * counts for each component of the name into an nvlist and then
1212 * checking each of those cases with the aggregated count.
1214 * This approach properly handles not only the recursive snapshot
1215 * case (where we get all of those on the ddsa_snaps list) but also
1216 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1217 * validate the limit on 'a' using a count of 2).
1219 * We validate the snapshot names in the third loop and only report
1222 if (dmu_tx_is_syncing(tx
)) {
1224 nvlist_t
*cnt_track
= NULL
;
1225 cnt_track
= fnvlist_alloc();
1227 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1229 /* Rollup aggregated counts into the cnt_track list */
1230 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1232 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1236 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1237 pdelim
= strchr(nm
, '@');
1243 if (nvlist_lookup_uint64(cnt_track
, nm
,
1245 /* update existing entry */
1246 fnvlist_add_uint64(cnt_track
, nm
,
1250 fnvlist_add_uint64(cnt_track
, nm
, 1);
1253 pdelim
= strrchr(nm
, '/');
1256 } while (pdelim
!= NULL
);
1259 kmem_free(nm
, MAXPATHLEN
);
1261 /* Check aggregated counts at each level */
1262 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1263 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1269 name
= nvpair_name(pair
);
1270 cnt
= fnvpair_value_uint64(pair
);
1273 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1275 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1276 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1278 dsl_dataset_rele(ds
, FTAG
);
1282 if (ddsa
->ddsa_errors
!= NULL
)
1283 fnvlist_add_int32(ddsa
->ddsa_errors
,
1286 /* only report one error for this check */
1290 nvlist_free(cnt_track
);
1293 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1294 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1297 char *name
, *atp
= NULL
;
1298 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1300 name
= nvpair_name(pair
);
1301 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1302 error
= SET_ERROR(ENAMETOOLONG
);
1304 atp
= strchr(name
, '@');
1306 error
= SET_ERROR(EINVAL
);
1308 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1311 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1313 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1314 error
= dsl_dataset_snapshot_check_impl(ds
,
1315 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1316 dsl_dataset_rele(ds
, FTAG
);
1320 if (ddsa
->ddsa_errors
!= NULL
) {
1321 fnvlist_add_int32(ddsa
->ddsa_errors
,
1332 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1335 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1337 dsl_dataset_phys_t
*dsphys
;
1338 uint64_t dsobj
, crtxg
;
1339 objset_t
*mos
= dp
->dp_meta_objset
;
1341 ASSERTV(static zil_header_t zero_zil
);
1342 ASSERTV(objset_t
*os
);
1344 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1347 * If we are on an old pool, the zil must not be active, in which
1348 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1350 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1351 dmu_objset_from_ds(ds
, &os
) != 0 ||
1352 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1353 sizeof (zero_zil
)) == 0);
1355 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1358 * The origin's ds_creation_txg has to be < TXG_INITIAL
1360 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1365 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1366 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1367 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1368 dmu_buf_will_dirty(dbuf
, tx
);
1369 dsphys
= dbuf
->db_data
;
1370 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1371 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1372 dsphys
->ds_fsid_guid
= unique_create();
1373 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1374 sizeof (dsphys
->ds_guid
));
1375 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1376 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1377 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1378 dsphys
->ds_num_children
= 1;
1379 dsphys
->ds_creation_time
= gethrestime_sec();
1380 dsphys
->ds_creation_txg
= crtxg
;
1381 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1382 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1383 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1384 dsphys
->ds_uncompressed_bytes
=
1385 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1386 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1387 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1388 dmu_buf_rele(dbuf
, FTAG
);
1390 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1391 if (ds
->ds_feature_inuse
[f
])
1392 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1395 ASSERT3U(ds
->ds_prev
!= 0, ==,
1396 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1398 uint64_t next_clones_obj
=
1399 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1400 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1402 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1403 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1405 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1406 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1407 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1408 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1409 } else if (next_clones_obj
!= 0) {
1410 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1411 dsphys
->ds_next_snap_obj
, tx
);
1412 VERIFY0(zap_add_int(mos
,
1413 next_clones_obj
, dsobj
, tx
));
1418 * If we have a reference-reservation on this dataset, we will
1419 * need to increase the amount of refreservation being charged
1420 * since our unique space is going to zero.
1422 if (ds
->ds_reserved
) {
1424 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1425 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1427 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1431 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1432 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1433 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1434 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1435 dsl_deadlist_close(&ds
->ds_deadlist
);
1436 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1437 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1438 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1439 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1441 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1442 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1443 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1444 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1445 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1446 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1448 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1449 snapname
, 8, 1, &dsobj
, tx
));
1452 dsl_dataset_rele(ds
->ds_prev
, ds
);
1453 VERIFY0(dsl_dataset_hold_obj(dp
,
1454 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1456 dsl_scan_ds_snapshotted(ds
, tx
);
1458 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1460 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1464 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1466 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1467 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1470 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1471 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1474 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1476 name
= nvpair_name(pair
);
1477 atp
= strchr(name
, '@');
1478 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1479 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1481 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1482 if (ddsa
->ddsa_props
!= NULL
) {
1483 dsl_props_set_sync_impl(ds
->ds_prev
,
1484 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1486 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1487 dsl_dataset_rele(ds
, FTAG
);
1492 * The snapshots must all be in the same pool.
1493 * All-or-nothing: if there are any failures, nothing will be modified.
1496 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1498 dsl_dataset_snapshot_arg_t ddsa
;
1500 boolean_t needsuspend
;
1504 nvlist_t
*suspended
= NULL
;
1506 pair
= nvlist_next_nvpair(snaps
, NULL
);
1509 firstname
= nvpair_name(pair
);
1511 error
= spa_open(firstname
, &spa
, FTAG
);
1514 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1515 spa_close(spa
, FTAG
);
1518 suspended
= fnvlist_alloc();
1519 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1520 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1521 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1522 char *snapname
= nvpair_name(pair
);
1526 atp
= strchr(snapname
, '@');
1528 error
= SET_ERROR(EINVAL
);
1531 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1533 error
= zil_suspend(fsname
, &cookie
);
1536 fnvlist_add_uint64(suspended
, fsname
,
1541 ddsa
.ddsa_snaps
= snaps
;
1542 ddsa
.ddsa_props
= props
;
1543 ddsa
.ddsa_errors
= errors
;
1544 ddsa
.ddsa_cr
= CRED();
1547 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1548 dsl_dataset_snapshot_sync
, &ddsa
,
1549 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1552 if (suspended
!= NULL
) {
1553 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1554 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1555 zil_resume((void *)(uintptr_t)
1556 fnvpair_value_uint64(pair
));
1558 fnvlist_free(suspended
);
1564 typedef struct dsl_dataset_snapshot_tmp_arg
{
1565 const char *ddsta_fsname
;
1566 const char *ddsta_snapname
;
1567 minor_t ddsta_cleanup_minor
;
1568 const char *ddsta_htag
;
1569 } dsl_dataset_snapshot_tmp_arg_t
;
1572 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1574 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1575 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1579 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1583 /* NULL cred means no limit check for tmp snapshot */
1584 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1585 tx
, B_FALSE
, 0, NULL
);
1587 dsl_dataset_rele(ds
, FTAG
);
1591 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1592 dsl_dataset_rele(ds
, FTAG
);
1593 return (SET_ERROR(ENOTSUP
));
1595 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1598 dsl_dataset_rele(ds
, FTAG
);
1602 dsl_dataset_rele(ds
, FTAG
);
1607 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1609 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1610 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1611 dsl_dataset_t
*ds
= NULL
;
1613 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1615 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1616 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1617 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1618 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1620 dsl_dataset_rele(ds
, FTAG
);
1624 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1625 minor_t cleanup_minor
, const char *htag
)
1627 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1630 boolean_t needsuspend
;
1633 ddsta
.ddsta_fsname
= fsname
;
1634 ddsta
.ddsta_snapname
= snapname
;
1635 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1636 ddsta
.ddsta_htag
= htag
;
1638 error
= spa_open(fsname
, &spa
, FTAG
);
1641 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1642 spa_close(spa
, FTAG
);
1645 error
= zil_suspend(fsname
, &cookie
);
1650 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1651 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1660 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1664 ASSERT(dmu_tx_is_syncing(tx
));
1665 ASSERT(ds
->ds_objset
!= NULL
);
1666 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1669 * in case we had to change ds_fsid_guid when we opened it,
1672 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1673 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1675 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1676 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1677 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1678 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1679 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1680 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1681 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1682 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1683 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1684 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1685 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1686 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1687 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1690 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1692 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1693 if (ds
->ds_feature_activation_needed
[f
]) {
1694 if (ds
->ds_feature_inuse
[f
])
1696 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1697 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1703 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1706 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1709 nvlist_t
*propval
= fnvlist_alloc();
1710 nvlist_t
*val
= fnvlist_alloc();
1712 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1715 * There may be missing entries in ds_next_clones_obj
1716 * due to a bug in a previous version of the code.
1717 * Only trust it if it has the right number of entries.
1719 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1720 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1723 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1)
1725 for (zap_cursor_init(&zc
, mos
,
1726 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1727 zap_cursor_retrieve(&zc
, &za
) == 0;
1728 zap_cursor_advance(&zc
)) {
1729 dsl_dataset_t
*clone
;
1730 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1731 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1732 za
.za_first_integer
, FTAG
, &clone
));
1733 dsl_dir_name(clone
->ds_dir
, buf
);
1734 fnvlist_add_boolean(val
, buf
);
1735 dsl_dataset_rele(clone
, FTAG
);
1737 zap_cursor_fini(&zc
);
1738 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1739 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
), propval
);
1742 nvlist_free(propval
);
1746 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1748 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1750 if (dsl_dataset_has_resume_receive_state(ds
)) {
1753 uint8_t *compressed
;
1755 nvlist_t
*token_nv
= fnvlist_alloc();
1756 size_t packed_size
, compressed_size
;
1759 char buf
[MAXNAMELEN
];
1762 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1763 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1764 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1766 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1767 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1768 fnvlist_add_uint64(token_nv
, "object", val
);
1770 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1771 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1772 fnvlist_add_uint64(token_nv
, "offset", val
);
1774 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1775 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1776 fnvlist_add_uint64(token_nv
, "bytes", val
);
1778 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1779 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1780 fnvlist_add_uint64(token_nv
, "toguid", val
);
1782 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1783 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1784 fnvlist_add_string(token_nv
, "toname", buf
);
1786 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1787 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1788 fnvlist_add_boolean(token_nv
, "largeblockok");
1790 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1791 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1792 fnvlist_add_boolean(token_nv
, "embedok");
1794 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1795 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1796 fnvlist_add_boolean(token_nv
, "compressok");
1798 packed
= fnvlist_pack(token_nv
, &packed_size
);
1799 fnvlist_free(token_nv
);
1800 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1802 compressed_size
= gzip_compress(packed
, compressed
,
1803 packed_size
, packed_size
, 6);
1805 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1807 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1808 for (i
= 0; i
< compressed_size
; i
++) {
1809 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1811 str
[compressed_size
* 2] = '\0';
1812 propval
= kmem_asprintf("%u-%llx-%llx-%s",
1813 ZFS_SEND_RESUME_TOKEN_VERSION
,
1814 (longlong_t
)cksum
.zc_word
[0],
1815 (longlong_t
)packed_size
, str
);
1816 dsl_prop_nvlist_add_string(nv
,
1817 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
1818 kmem_free(packed
, packed_size
);
1819 kmem_free(str
, compressed_size
* 2 + 1);
1820 kmem_free(compressed
, packed_size
);
1826 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1828 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1829 uint64_t refd
, avail
, uobjs
, aobjs
, ratio
;
1831 ASSERT(dsl_pool_config_held(dp
));
1833 ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1834 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1835 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1837 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
, ratio
);
1838 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
1839 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1841 if (ds
->ds_is_snapshot
) {
1842 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
, ratio
);
1843 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
1844 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1845 get_clones_stat(ds
, nv
);
1847 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
1848 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1849 dsl_dataset_name(ds
->ds_prev
, buf
);
1850 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
, buf
);
1853 dsl_dir_stats(ds
->ds_dir
, nv
);
1856 dsl_dataset_space(ds
, &refd
, &avail
, &uobjs
, &aobjs
);
1857 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
, avail
);
1858 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
, refd
);
1860 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
1861 dsl_dataset_phys(ds
)->ds_creation_time
);
1862 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
1863 dsl_dataset_phys(ds
)->ds_creation_txg
);
1864 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
1866 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
1868 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
1869 dsl_dataset_phys(ds
)->ds_guid
);
1870 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
1871 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1872 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
1874 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
1876 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
1877 DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
1879 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
1880 uint64_t written
, comp
, uncomp
;
1881 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1882 dsl_dataset_t
*prev
;
1885 err
= dsl_dataset_hold_obj(dp
,
1886 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
1888 err
= dsl_dataset_space_written(prev
, ds
, &written
,
1890 dsl_dataset_rele(prev
, FTAG
);
1892 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
1898 if (!dsl_dataset_is_snapshot(ds
)) {
1899 /* 6 extra bytes for /%recv */
1900 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
1901 dsl_dataset_t
*recv_ds
;
1904 * A failed "newfs" (e.g. full) resumable receive leaves
1905 * the stats set on this dataset. Check here for the prop.
1907 get_receive_resume_stats(ds
, nv
);
1910 * A failed incremental resumable receive leaves the
1911 * stats set on our child named "%recv". Check the child
1914 dsl_dataset_name(ds
, recvname
);
1915 if (strlcat(recvname
, "/", sizeof (recvname
)) <
1916 sizeof (recvname
) &&
1917 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
1918 sizeof (recvname
) &&
1919 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
1920 get_receive_resume_stats(recv_ds
, nv
);
1921 dsl_dataset_rele(recv_ds
, FTAG
);
1927 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
1929 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1930 ASSERT(dsl_pool_config_held(dp
));
1932 stat
->dds_creation_txg
= dsl_dataset_phys(ds
)->ds_creation_txg
;
1933 stat
->dds_inconsistent
=
1934 dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
;
1935 stat
->dds_guid
= dsl_dataset_phys(ds
)->ds_guid
;
1936 stat
->dds_origin
[0] = '\0';
1937 if (ds
->ds_is_snapshot
) {
1938 stat
->dds_is_snapshot
= B_TRUE
;
1939 stat
->dds_num_clones
=
1940 dsl_dataset_phys(ds
)->ds_num_children
- 1;
1942 stat
->dds_is_snapshot
= B_FALSE
;
1943 stat
->dds_num_clones
= 0;
1945 if (dsl_dir_is_clone(ds
->ds_dir
)) {
1948 VERIFY0(dsl_dataset_hold_obj(dp
,
1949 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
,
1951 dsl_dataset_name(ods
, stat
->dds_origin
);
1952 dsl_dataset_rele(ods
, FTAG
);
1958 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
1960 return (ds
->ds_fsid_guid
);
1964 dsl_dataset_space(dsl_dataset_t
*ds
,
1965 uint64_t *refdbytesp
, uint64_t *availbytesp
,
1966 uint64_t *usedobjsp
, uint64_t *availobjsp
)
1968 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1969 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
1970 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
1972 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
1973 if (ds
->ds_quota
!= 0) {
1975 * Adjust available bytes according to refquota
1977 if (*refdbytesp
< ds
->ds_quota
)
1978 *availbytesp
= MIN(*availbytesp
,
1979 ds
->ds_quota
- *refdbytesp
);
1983 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
1984 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
1988 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
1990 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1993 if (dsl_dataset_phys(ds
)->ds_bp
.blk_birth
>
1994 dsl_dataset_phys(snap
)->ds_creation_txg
) {
1995 objset_t
*os
, *os_snap
;
1997 * It may be that only the ZIL differs, because it was
1998 * reset in the head. Don't count that as being
2001 if (dmu_objset_from_ds(ds
, &os
) != 0)
2003 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2005 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2006 &os_snap
->os_phys
->os_meta_dnode
,
2007 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2012 typedef struct dsl_dataset_rename_snapshot_arg
{
2013 const char *ddrsa_fsname
;
2014 const char *ddrsa_oldsnapname
;
2015 const char *ddrsa_newsnapname
;
2016 boolean_t ddrsa_recursive
;
2018 } dsl_dataset_rename_snapshot_arg_t
;
2022 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2023 dsl_dataset_t
*hds
, void *arg
)
2025 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2029 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2031 /* ignore nonexistent snapshots */
2032 return (error
== ENOENT
? 0 : error
);
2035 /* new name should not exist */
2036 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2038 error
= SET_ERROR(EEXIST
);
2039 else if (error
== ENOENT
)
2042 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2043 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2044 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2045 error
= SET_ERROR(ENAMETOOLONG
);
2051 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2053 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2054 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2058 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2062 if (ddrsa
->ddrsa_recursive
) {
2063 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2064 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2067 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2069 dsl_dataset_rele(hds
, FTAG
);
2074 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2075 dsl_dataset_t
*hds
, void *arg
)
2077 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2080 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2083 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2084 ASSERT(error
== 0 || error
== ENOENT
);
2085 if (error
== ENOENT
) {
2086 /* ignore nonexistent snapshots */
2090 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2092 /* log before we change the name */
2093 spa_history_log_internal_ds(ds
, "rename", tx
,
2094 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2096 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2098 mutex_enter(&ds
->ds_lock
);
2099 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2100 sizeof (ds
->ds_snapname
));
2101 mutex_exit(&ds
->ds_lock
);
2102 VERIFY0(zap_add(dp
->dp_meta_objset
,
2103 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2104 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2105 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2106 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2108 dsl_dataset_rele(ds
, FTAG
);
2113 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2115 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2116 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2117 dsl_dataset_t
*hds
= NULL
;
2119 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2120 ddrsa
->ddrsa_tx
= tx
;
2121 if (ddrsa
->ddrsa_recursive
) {
2122 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2123 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2126 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2128 dsl_dataset_rele(hds
, FTAG
);
2132 dsl_dataset_rename_snapshot(const char *fsname
,
2133 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2135 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2137 ddrsa
.ddrsa_fsname
= fsname
;
2138 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2139 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2140 ddrsa
.ddrsa_recursive
= recursive
;
2142 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2143 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2144 1, ZFS_SPACE_CHECK_RESERVED
));
2148 * If we're doing an ownership handoff, we need to make sure that there is
2149 * only one long hold on the dataset. We're not allowed to change anything here
2150 * so we don't permanently release the long hold or regular hold here. We want
2151 * to do this only when syncing to avoid the dataset unexpectedly going away
2152 * when we release the long hold.
2155 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2159 if (!dmu_tx_is_syncing(tx
))
2162 if (owner
!= NULL
) {
2163 VERIFY3P(ds
->ds_owner
, ==, owner
);
2164 dsl_dataset_long_rele(ds
, owner
);
2167 held
= dsl_dataset_long_held(ds
);
2170 dsl_dataset_long_hold(ds
, owner
);
2173 return (SET_ERROR(EBUSY
));
2178 typedef struct dsl_dataset_rollback_arg
{
2179 const char *ddra_fsname
;
2181 nvlist_t
*ddra_result
;
2182 } dsl_dataset_rollback_arg_t
;
2185 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2187 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2188 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2190 int64_t unused_refres_delta
;
2193 nvlist_t
*proprequest
, *bookmarks
;
2195 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2199 /* must not be a snapshot */
2200 if (ds
->ds_is_snapshot
) {
2201 dsl_dataset_rele(ds
, FTAG
);
2202 return (SET_ERROR(EINVAL
));
2205 /* must have a most recent snapshot */
2206 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2207 dsl_dataset_rele(ds
, FTAG
);
2208 return (SET_ERROR(EINVAL
));
2211 /* must not have any bookmarks after the most recent snapshot */
2212 proprequest
= fnvlist_alloc();
2213 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2214 bookmarks
= fnvlist_alloc();
2215 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2216 fnvlist_free(proprequest
);
2219 for (pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2220 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2222 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2223 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2224 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2225 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2226 fnvlist_free(bookmarks
);
2227 dsl_dataset_rele(ds
, FTAG
);
2228 return (SET_ERROR(EEXIST
));
2231 fnvlist_free(bookmarks
);
2233 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2235 dsl_dataset_rele(ds
, FTAG
);
2240 * Check if the snap we are rolling back to uses more than
2243 if (ds
->ds_quota
!= 0 &&
2244 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2245 dsl_dataset_rele(ds
, FTAG
);
2246 return (SET_ERROR(EDQUOT
));
2250 * When we do the clone swap, we will temporarily use more space
2251 * due to the refreservation (the head will no longer have any
2252 * unique space, so the entire amount of the refreservation will need
2253 * to be free). We will immediately destroy the clone, freeing
2254 * this space, but the freeing happens over many txg's.
2256 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2257 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2259 if (unused_refres_delta
> 0 &&
2260 unused_refres_delta
>
2261 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2262 dsl_dataset_rele(ds
, FTAG
);
2263 return (SET_ERROR(ENOSPC
));
2266 dsl_dataset_rele(ds
, FTAG
);
2271 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2273 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2274 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2275 dsl_dataset_t
*ds
, *clone
;
2277 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2279 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2281 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2282 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2284 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2285 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, tx
);
2287 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2289 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2290 dsl_dataset_zero_zil(ds
, tx
);
2292 dsl_destroy_head_sync_impl(clone
, tx
);
2294 dsl_dataset_rele(clone
, FTAG
);
2295 dsl_dataset_rele(ds
, FTAG
);
2299 * Rolls back the given filesystem or volume to the most recent snapshot.
2300 * The name of the most recent snapshot will be returned under key "target"
2301 * in the result nvlist.
2304 * - The existing dataset MUST be owned by the specified owner at entry
2305 * - Upon return, dataset will still be held by the same owner, whether we
2308 * This mode is required any time the existing filesystem is mounted. See
2309 * notes above zfs_suspend_fs() for further details.
2312 dsl_dataset_rollback(const char *fsname
, void *owner
, nvlist_t
*result
)
2314 dsl_dataset_rollback_arg_t ddra
;
2316 ddra
.ddra_fsname
= fsname
;
2317 ddra
.ddra_owner
= owner
;
2318 ddra
.ddra_result
= result
;
2320 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2321 dsl_dataset_rollback_sync
, &ddra
,
2322 1, ZFS_SPACE_CHECK_RESERVED
));
2325 struct promotenode
{
2330 typedef struct dsl_dataset_promote_arg
{
2331 const char *ddpa_clonename
;
2332 dsl_dataset_t
*ddpa_clone
;
2333 list_t shared_snaps
, origin_snaps
, clone_snaps
;
2334 dsl_dataset_t
*origin_origin
; /* origin of the origin */
2335 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
2338 } dsl_dataset_promote_arg_t
;
2340 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2341 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2343 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2346 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2348 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2349 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2351 struct promotenode
*snap
;
2352 dsl_dataset_t
*origin_ds
;
2356 size_t max_snap_len
;
2358 err
= promote_hold(ddpa
, dp
, FTAG
);
2362 hds
= ddpa
->ddpa_clone
;
2363 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2365 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2366 promote_rele(ddpa
, FTAG
);
2367 return (SET_ERROR(EXDEV
));
2371 * Compute and check the amount of space to transfer. Since this is
2372 * so expensive, don't do the preliminary check.
2374 if (!dmu_tx_is_syncing(tx
)) {
2375 promote_rele(ddpa
, FTAG
);
2379 snap
= list_head(&ddpa
->shared_snaps
);
2381 err
= SET_ERROR(ENOENT
);
2384 origin_ds
= snap
->ds
;
2386 /* compute origin's new unique space */
2387 snap
= list_tail(&ddpa
->clone_snaps
);
2388 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2389 origin_ds
->ds_object
);
2390 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2391 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2392 &ddpa
->unique
, &unused
, &unused
);
2395 * Walk the snapshots that we are moving
2397 * Compute space to transfer. Consider the incremental changes
2398 * to used by each snapshot:
2399 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2400 * So each snapshot gave birth to:
2401 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2402 * So a sequence would look like:
2403 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2404 * Which simplifies to:
2405 * uN + kN + kN-1 + ... + k1 + k0
2406 * Note however, if we stop before we reach the ORIGIN we get:
2407 * uN + kN + kN-1 + ... + kM - uM-1
2410 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2411 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2412 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2413 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2414 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2415 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2416 dsl_dataset_t
*ds
= snap
->ds
;
2421 * If there are long holds, we won't be able to evict
2424 if (dsl_dataset_long_held(ds
)) {
2425 err
= SET_ERROR(EBUSY
);
2429 /* Check that the snapshot name does not conflict */
2430 VERIFY0(dsl_dataset_get_snapname(ds
));
2431 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2432 err
= SET_ERROR(ENAMETOOLONG
);
2435 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2437 (void) strcpy(ddpa
->err_ds
, snap
->ds
->ds_snapname
);
2438 err
= SET_ERROR(EEXIST
);
2444 /* The very first snapshot does not have a deadlist */
2445 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2448 dsl_deadlist_space(&ds
->ds_deadlist
,
2449 &dlused
, &dlcomp
, &dluncomp
);
2450 ddpa
->used
+= dlused
;
2451 ddpa
->comp
+= dlcomp
;
2452 ddpa
->uncomp
+= dluncomp
;
2456 * If we are a clone of a clone then we never reached ORIGIN,
2457 * so we need to subtract out the clone origin's used space.
2459 if (ddpa
->origin_origin
) {
2461 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2463 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2465 dsl_dataset_phys(ddpa
->origin_origin
)->
2466 ds_uncompressed_bytes
;
2469 /* Check that there is enough space and limit headroom here */
2470 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2471 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2476 * Compute the amounts of space that will be used by snapshots
2477 * after the promotion (for both origin and clone). For each,
2478 * it is the amount of space that will be on all of their
2479 * deadlists (that was not born before their new origin).
2481 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2485 * Note, typically this will not be a clone of a clone,
2486 * so dd_origin_txg will be < TXG_INITIAL, so
2487 * these snaplist_space() -> dsl_deadlist_space_range()
2488 * calls will be fast because they do not have to
2489 * iterate over all bps.
2491 snap
= list_head(&ddpa
->origin_snaps
);
2493 err
= SET_ERROR(ENOENT
);
2496 err
= snaplist_space(&ddpa
->shared_snaps
,
2497 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2501 err
= snaplist_space(&ddpa
->clone_snaps
,
2502 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2505 ddpa
->cloneusedsnap
+= space
;
2507 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2508 DD_FLAG_USED_BREAKDOWN
) {
2509 err
= snaplist_space(&ddpa
->origin_snaps
,
2510 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2511 &ddpa
->originusedsnap
);
2517 promote_rele(ddpa
, FTAG
);
2522 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2524 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2525 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2527 struct promotenode
*snap
;
2528 dsl_dataset_t
*origin_ds
;
2529 dsl_dataset_t
*origin_head
;
2531 dsl_dir_t
*odd
= NULL
;
2532 uint64_t oldnext_obj
;
2535 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2536 hds
= ddpa
->ddpa_clone
;
2538 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2540 snap
= list_head(&ddpa
->shared_snaps
);
2541 origin_ds
= snap
->ds
;
2544 snap
= list_head(&ddpa
->origin_snaps
);
2545 origin_head
= snap
->ds
;
2548 * We need to explicitly open odd, since origin_ds's dd will be
2551 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2554 /* change origin's next snap */
2555 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2556 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2557 snap
= list_tail(&ddpa
->clone_snaps
);
2558 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2559 origin_ds
->ds_object
);
2560 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2562 /* change the origin's next clone */
2563 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2564 dsl_dataset_remove_from_next_clones(origin_ds
,
2565 snap
->ds
->ds_object
, tx
);
2566 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2567 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2572 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2573 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2574 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2575 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2576 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2577 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2578 origin_head
->ds_dir
->dd_origin_txg
=
2579 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2581 /* change dd_clone entries */
2582 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2583 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2584 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2585 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2586 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2587 hds
->ds_object
, tx
));
2589 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2590 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2591 origin_head
->ds_object
, tx
));
2592 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2593 dsl_dir_phys(dd
)->dd_clones
=
2594 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2595 DMU_OT_NONE
, 0, tx
);
2597 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2598 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2601 /* move snapshots to this dir */
2602 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2603 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2604 dsl_dataset_t
*ds
= snap
->ds
;
2607 * Property callbacks are registered to a particular
2608 * dsl_dir. Since ours is changing, evict the objset
2609 * so that they will be unregistered from the old dsl_dir.
2611 if (ds
->ds_objset
) {
2612 dmu_objset_evict(ds
->ds_objset
);
2613 ds
->ds_objset
= NULL
;
2616 /* move snap name entry */
2617 VERIFY0(dsl_dataset_get_snapname(ds
));
2618 VERIFY0(dsl_dataset_snap_remove(origin_head
,
2619 ds
->ds_snapname
, tx
, B_TRUE
));
2620 VERIFY0(zap_add(dp
->dp_meta_objset
,
2621 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
2622 8, 1, &ds
->ds_object
, tx
));
2623 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
2624 DD_FIELD_SNAPSHOT_COUNT
, tx
);
2626 /* change containing dsl_dir */
2627 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2628 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
2629 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
2630 ASSERT3P(ds
->ds_dir
, ==, odd
);
2631 dsl_dir_rele(ds
->ds_dir
, ds
);
2632 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
2633 NULL
, ds
, &ds
->ds_dir
));
2635 /* move any clone references */
2636 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
2637 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2641 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
2642 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2643 zap_cursor_retrieve(&zc
, &za
) == 0;
2644 zap_cursor_advance(&zc
)) {
2645 dsl_dataset_t
*cnds
;
2648 if (za
.za_first_integer
== oldnext_obj
) {
2650 * We've already moved the
2651 * origin's reference.
2656 VERIFY0(dsl_dataset_hold_obj(dp
,
2657 za
.za_first_integer
, FTAG
, &cnds
));
2658 o
= dsl_dir_phys(cnds
->ds_dir
)->
2659 dd_head_dataset_obj
;
2661 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2662 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
2663 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2664 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
2665 dsl_dataset_rele(cnds
, FTAG
);
2667 zap_cursor_fini(&zc
);
2670 ASSERT(!dsl_prop_hascb(ds
));
2674 * Change space accounting.
2675 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2676 * both be valid, or both be 0 (resulting in delta == 0). This
2677 * is true for each of {clone,origin} independently.
2680 delta
= ddpa
->cloneusedsnap
-
2681 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
2682 ASSERT3S(delta
, >=, 0);
2683 ASSERT3U(ddpa
->used
, >=, delta
);
2684 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2685 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
2686 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
2688 delta
= ddpa
->originusedsnap
-
2689 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
2690 ASSERT3S(delta
, <=, 0);
2691 ASSERT3U(ddpa
->used
, >=, -delta
);
2692 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2693 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
2694 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
2696 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
2698 /* log history record */
2699 spa_history_log_internal_ds(hds
, "promote", tx
, "");
2701 dsl_dir_rele(odd
, FTAG
);
2702 promote_rele(ddpa
, FTAG
);
2706 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2707 * (exclusive) and last_obj (inclusive). The list will be in reverse
2708 * order (last_obj will be the list_head()). If first_obj == 0, do all
2709 * snapshots back to this dataset's origin.
2712 snaplist_make(dsl_pool_t
*dp
,
2713 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
2715 uint64_t obj
= last_obj
;
2717 list_create(l
, sizeof (struct promotenode
),
2718 offsetof(struct promotenode
, link
));
2720 while (obj
!= first_obj
) {
2722 struct promotenode
*snap
;
2725 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
2726 ASSERT(err
!= ENOENT
);
2731 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
2733 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
2735 list_insert_tail(l
, snap
);
2736 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
2743 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
2745 struct promotenode
*snap
;
2748 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
2749 uint64_t used
, comp
, uncomp
;
2750 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2751 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
2758 snaplist_destroy(list_t
*l
, void *tag
)
2760 struct promotenode
*snap
;
2762 if (l
== NULL
|| !list_link_active(&l
->list_head
))
2765 while ((snap
= list_tail(l
)) != NULL
) {
2766 list_remove(l
, snap
);
2767 dsl_dataset_rele(snap
->ds
, tag
);
2768 kmem_free(snap
, sizeof (*snap
));
2774 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
2778 struct promotenode
*snap
;
2780 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
2784 dd
= ddpa
->ddpa_clone
->ds_dir
;
2786 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
2787 !dsl_dir_is_clone(dd
)) {
2788 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2789 return (SET_ERROR(EINVAL
));
2792 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
2793 &ddpa
->shared_snaps
, tag
);
2797 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
2798 &ddpa
->clone_snaps
, tag
);
2802 snap
= list_head(&ddpa
->shared_snaps
);
2803 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
2804 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
2805 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
2806 &ddpa
->origin_snaps
, tag
);
2810 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
2811 error
= dsl_dataset_hold_obj(dp
,
2812 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
2813 tag
, &ddpa
->origin_origin
);
2819 promote_rele(ddpa
, tag
);
2824 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
2826 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
2827 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
2828 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
2829 if (ddpa
->origin_origin
!= NULL
)
2830 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
2831 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2837 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2838 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
2841 dsl_dataset_promote(const char *name
, char *conflsnap
)
2843 dsl_dataset_promote_arg_t ddpa
= { 0 };
2849 * We will modify space proportional to the number of
2850 * snapshots. Compute numsnaps.
2852 error
= dmu_objset_hold(name
, FTAG
, &os
);
2855 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
2856 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
2858 dmu_objset_rele(os
, FTAG
);
2862 ddpa
.ddpa_clonename
= name
;
2863 ddpa
.err_ds
= conflsnap
;
2866 return (dsl_sync_task(name
, dsl_dataset_promote_check
,
2867 dsl_dataset_promote_sync
, &ddpa
,
2868 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
));
2872 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
2873 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
2876 * "slack" factor for received datasets with refquota set on them.
2877 * See the bottom of this function for details on its use.
2879 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
2880 spa_asize_inflation
;
2881 int64_t unused_refres_delta
;
2883 /* they should both be heads */
2884 if (clone
->ds_is_snapshot
||
2885 origin_head
->ds_is_snapshot
)
2886 return (SET_ERROR(EINVAL
));
2888 /* if we are not forcing, the branch point should be just before them */
2889 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
2890 return (SET_ERROR(EINVAL
));
2892 /* clone should be the clone (unless they are unrelated) */
2893 if (clone
->ds_prev
!= NULL
&&
2894 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
2895 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
2896 return (SET_ERROR(EINVAL
));
2898 /* the clone should be a child of the origin */
2899 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
2900 return (SET_ERROR(EINVAL
));
2902 /* origin_head shouldn't be modified unless 'force' */
2904 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
2905 return (SET_ERROR(ETXTBSY
));
2907 /* origin_head should have no long holds (e.g. is not mounted) */
2908 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
2909 return (SET_ERROR(EBUSY
));
2911 /* check amount of any unconsumed refreservation */
2912 unused_refres_delta
=
2913 (int64_t)MIN(origin_head
->ds_reserved
,
2914 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2915 (int64_t)MIN(origin_head
->ds_reserved
,
2916 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2918 if (unused_refres_delta
> 0 &&
2919 unused_refres_delta
>
2920 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
2921 return (SET_ERROR(ENOSPC
));
2924 * The clone can't be too much over the head's refquota.
2926 * To ensure that the entire refquota can be used, we allow one
2927 * transaction to exceed the the refquota. Therefore, this check
2928 * needs to also allow for the space referenced to be more than the
2929 * refquota. The maximum amount of space that one transaction can use
2930 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
2931 * overage ensures that we are able to receive a filesystem that
2932 * exceeds the refquota on the source system.
2934 * So that overage is the refquota_slack we use below.
2936 if (origin_head
->ds_quota
!= 0 &&
2937 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
2938 origin_head
->ds_quota
+ refquota_slack
)
2939 return (SET_ERROR(EDQUOT
));
2945 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
2946 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
2949 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2950 int64_t unused_refres_delta
;
2952 ASSERT(clone
->ds_reserved
== 0);
2954 * NOTE: On DEBUG kernels there could be a race between this and
2955 * the check function if spa_asize_inflation is adjusted...
2957 ASSERT(origin_head
->ds_quota
== 0 ||
2958 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
2959 DMU_MAX_ACCESS
* spa_asize_inflation
);
2960 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
2963 * Swap per-dataset feature flags.
2965 for (f
= 0; f
< SPA_FEATURES
; f
++) {
2966 boolean_t clone_inuse
;
2967 boolean_t origin_head_inuse
;
2969 if (!(spa_feature_table
[f
].fi_flags
&
2970 ZFEATURE_FLAG_PER_DATASET
)) {
2971 ASSERT(!clone
->ds_feature_inuse
[f
]);
2972 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
2976 clone_inuse
= clone
->ds_feature_inuse
[f
];
2977 origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
2980 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
2981 clone
->ds_feature_inuse
[f
] = B_FALSE
;
2983 if (origin_head_inuse
) {
2984 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
2986 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
2989 dsl_dataset_activate_feature(origin_head
->ds_object
,
2991 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
2993 if (origin_head_inuse
) {
2994 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
2995 clone
->ds_feature_inuse
[f
] = B_TRUE
;
2999 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3000 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3002 if (clone
->ds_objset
!= NULL
) {
3003 dmu_objset_evict(clone
->ds_objset
);
3004 clone
->ds_objset
= NULL
;
3007 if (origin_head
->ds_objset
!= NULL
) {
3008 dmu_objset_evict(origin_head
->ds_objset
);
3009 origin_head
->ds_objset
= NULL
;
3012 unused_refres_delta
=
3013 (int64_t)MIN(origin_head
->ds_reserved
,
3014 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3015 (int64_t)MIN(origin_head
->ds_reserved
,
3016 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3019 * Reset origin's unique bytes, if it exists.
3021 if (clone
->ds_prev
) {
3022 dsl_dataset_t
*origin
= clone
->ds_prev
;
3023 uint64_t comp
, uncomp
;
3025 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3026 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3027 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3028 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3034 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3035 dsl_dataset_phys(origin_head
)->ds_bp
=
3036 dsl_dataset_phys(clone
)->ds_bp
;
3037 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3040 /* set dd_*_bytes */
3042 int64_t dused
, dcomp
, duncomp
;
3043 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3044 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3046 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3047 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3049 dsl_deadlist_space(&clone
->ds_deadlist
,
3050 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3051 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3052 &odl_used
, &odl_comp
, &odl_uncomp
);
3054 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3056 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3058 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3060 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3062 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3064 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3067 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3068 dused
, dcomp
, duncomp
, tx
);
3069 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3070 -dused
, -dcomp
, -duncomp
, tx
);
3073 * The difference in the space used by snapshots is the
3074 * difference in snapshot space due to the head's
3075 * deadlist (since that's the only thing that's
3076 * changing that affects the snapused).
3078 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3079 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3080 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3081 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3082 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3083 &odl_used
, &odl_comp
, &odl_uncomp
);
3084 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3085 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3088 /* swap ds_*_bytes */
3089 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3090 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3091 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3092 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3093 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3094 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3095 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3096 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3098 /* apply any parent delta for change in unconsumed refreservation */
3099 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3100 unused_refres_delta
, 0, 0, tx
);
3105 dsl_deadlist_close(&clone
->ds_deadlist
);
3106 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3107 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3108 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3109 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3110 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3111 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3112 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3114 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3116 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3117 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3121 * Given a pool name and a dataset object number in that pool,
3122 * return the name of that dataset.
3125 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3131 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3135 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3137 dsl_dataset_name(ds
, buf
);
3138 dsl_dataset_rele(ds
, FTAG
);
3140 dsl_pool_rele(dp
, FTAG
);
3146 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3147 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3151 ASSERT3S(asize
, >, 0);
3154 * *ref_rsrv is the portion of asize that will come from any
3155 * unconsumed refreservation space.
3159 mutex_enter(&ds
->ds_lock
);
3161 * Make a space adjustment for reserved bytes.
3163 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3165 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3167 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3169 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3172 if (!check_quota
|| ds
->ds_quota
== 0) {
3173 mutex_exit(&ds
->ds_lock
);
3177 * If they are requesting more space, and our current estimate
3178 * is over quota, they get to try again unless the actual
3179 * on-disk is over quota and there are no pending changes (which
3180 * may free up space for us).
3182 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3185 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3186 error
= SET_ERROR(ERESTART
);
3188 error
= SET_ERROR(EDQUOT
);
3190 mutex_exit(&ds
->ds_lock
);
3195 typedef struct dsl_dataset_set_qr_arg
{
3196 const char *ddsqra_name
;
3197 zprop_source_t ddsqra_source
;
3198 uint64_t ddsqra_value
;
3199 } dsl_dataset_set_qr_arg_t
;
3204 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3206 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3207 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3212 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3213 return (SET_ERROR(ENOTSUP
));
3215 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3219 if (ds
->ds_is_snapshot
) {
3220 dsl_dataset_rele(ds
, FTAG
);
3221 return (SET_ERROR(EINVAL
));
3224 error
= dsl_prop_predict(ds
->ds_dir
,
3225 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3226 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3228 dsl_dataset_rele(ds
, FTAG
);
3233 dsl_dataset_rele(ds
, FTAG
);
3237 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3238 newval
< ds
->ds_reserved
) {
3239 dsl_dataset_rele(ds
, FTAG
);
3240 return (SET_ERROR(ENOSPC
));
3243 dsl_dataset_rele(ds
, FTAG
);
3248 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3250 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3251 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3252 dsl_dataset_t
*ds
= NULL
;
3255 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3257 dsl_prop_set_sync_impl(ds
,
3258 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3259 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3260 &ddsqra
->ddsqra_value
, tx
);
3262 VERIFY0(dsl_prop_get_int_ds(ds
,
3263 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3265 if (ds
->ds_quota
!= newval
) {
3266 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3267 ds
->ds_quota
= newval
;
3269 dsl_dataset_rele(ds
, FTAG
);
3273 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3276 dsl_dataset_set_qr_arg_t ddsqra
;
3278 ddsqra
.ddsqra_name
= dsname
;
3279 ddsqra
.ddsqra_source
= source
;
3280 ddsqra
.ddsqra_value
= refquota
;
3282 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3283 dsl_dataset_set_refquota_sync
, &ddsqra
, 0, ZFS_SPACE_CHECK_NONE
));
3287 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3289 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3290 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3293 uint64_t newval
, unique
;
3295 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3296 return (SET_ERROR(ENOTSUP
));
3298 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3302 if (ds
->ds_is_snapshot
) {
3303 dsl_dataset_rele(ds
, FTAG
);
3304 return (SET_ERROR(EINVAL
));
3307 error
= dsl_prop_predict(ds
->ds_dir
,
3308 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3309 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3311 dsl_dataset_rele(ds
, FTAG
);
3316 * If we are doing the preliminary check in open context, the
3317 * space estimates may be inaccurate.
3319 if (!dmu_tx_is_syncing(tx
)) {
3320 dsl_dataset_rele(ds
, FTAG
);
3324 mutex_enter(&ds
->ds_lock
);
3325 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3326 dsl_dataset_recalc_head_uniq(ds
);
3327 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3328 mutex_exit(&ds
->ds_lock
);
3330 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3331 uint64_t delta
= MAX(unique
, newval
) -
3332 MAX(unique
, ds
->ds_reserved
);
3335 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3336 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3337 dsl_dataset_rele(ds
, FTAG
);
3338 return (SET_ERROR(ENOSPC
));
3342 dsl_dataset_rele(ds
, FTAG
);
3347 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3348 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3354 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3355 source
, sizeof (value
), 1, &value
, tx
);
3357 VERIFY0(dsl_prop_get_int_ds(ds
,
3358 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3360 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3361 mutex_enter(&ds
->ds_dir
->dd_lock
);
3362 mutex_enter(&ds
->ds_lock
);
3363 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3364 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3365 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3366 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3367 ds
->ds_reserved
= newval
;
3368 mutex_exit(&ds
->ds_lock
);
3370 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3371 mutex_exit(&ds
->ds_dir
->dd_lock
);
3375 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3377 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3378 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3379 dsl_dataset_t
*ds
= NULL
;
3381 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3382 dsl_dataset_set_refreservation_sync_impl(ds
,
3383 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3384 dsl_dataset_rele(ds
, FTAG
);
3388 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3389 uint64_t refreservation
)
3391 dsl_dataset_set_qr_arg_t ddsqra
;
3393 ddsqra
.ddsqra_name
= dsname
;
3394 ddsqra
.ddsqra_source
= source
;
3395 ddsqra
.ddsqra_value
= refreservation
;
3397 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3398 dsl_dataset_set_refreservation_sync
, &ddsqra
,
3399 0, ZFS_SPACE_CHECK_NONE
));
3403 * Return (in *usedp) the amount of space written in new that is not
3404 * present in oldsnap. New may be a snapshot or the head. Old must be
3405 * a snapshot before new, in new's filesystem (or its origin). If not then
3406 * fail and return EINVAL.
3408 * The written space is calculated by considering two components: First, we
3409 * ignore any freed space, and calculate the written as new's used space
3410 * minus old's used space. Next, we add in the amount of space that was freed
3411 * between the two snapshots, thus reducing new's used space relative to old's.
3412 * Specifically, this is the space that was born before old->ds_creation_txg,
3413 * and freed before new (ie. on new's deadlist or a previous deadlist).
3415 * space freed [---------------------]
3416 * snapshots ---O-------O--------O-------O------
3420 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3421 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3425 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3427 ASSERT(dsl_pool_config_held(dp
));
3430 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3431 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3434 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3435 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3438 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3439 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3441 snapobj
= new->ds_object
;
3442 while (snapobj
!= oldsnap
->ds_object
) {
3443 dsl_dataset_t
*snap
;
3444 uint64_t used
, comp
, uncomp
;
3446 if (snapobj
== new->ds_object
) {
3449 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3454 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3455 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3457 * The blocks in the deadlist can not be born after
3458 * ds_prev_snap_txg, so get the whole deadlist space,
3459 * which is more efficient (especially for old-format
3460 * deadlists). Unfortunately the deadlist code
3461 * doesn't have enough information to make this
3462 * optimization itself.
3464 dsl_deadlist_space(&snap
->ds_deadlist
,
3465 &used
, &comp
, &uncomp
);
3467 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3468 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3469 &used
, &comp
, &uncomp
);
3476 * If we get to the beginning of the chain of snapshots
3477 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3478 * was not a snapshot of/before new.
3480 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3482 dsl_dataset_rele(snap
, FTAG
);
3484 err
= SET_ERROR(EINVAL
);
3493 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3494 * lastsnap, and all snapshots in between are deleted.
3496 * blocks that would be freed [---------------------------]
3497 * snapshots ---O-------O--------O-------O--------O
3498 * firstsnap lastsnap
3500 * This is the set of blocks that were born after the snap before firstsnap,
3501 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3502 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3503 * We calculate this by iterating over the relevant deadlists (from the snap
3504 * after lastsnap, backward to the snap after firstsnap), summing up the
3505 * space on the deadlist that was born after the snap before firstsnap.
3508 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3509 dsl_dataset_t
*lastsnap
,
3510 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3514 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3516 ASSERT(firstsnap
->ds_is_snapshot
);
3517 ASSERT(lastsnap
->ds_is_snapshot
);
3520 * Check that the snapshots are in the same dsl_dir, and firstsnap
3521 * is before lastsnap.
3523 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3524 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3525 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3526 return (SET_ERROR(EINVAL
));
3528 *usedp
= *compp
= *uncompp
= 0;
3530 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3531 while (snapobj
!= firstsnap
->ds_object
) {
3533 uint64_t used
, comp
, uncomp
;
3535 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3539 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3540 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3541 &used
, &comp
, &uncomp
);
3546 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3547 ASSERT3U(snapobj
, !=, 0);
3548 dsl_dataset_rele(ds
, FTAG
);
3554 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3555 * For example, they could both be snapshots of the same filesystem, and
3556 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3557 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3558 * filesystem. Or 'earlier' could be the origin's origin.
3560 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3563 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3564 uint64_t earlier_txg
)
3566 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3569 dsl_dataset_t
*origin
;
3571 ASSERT(dsl_pool_config_held(dp
));
3572 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3574 if (earlier_txg
== 0)
3575 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3577 if (later
->ds_is_snapshot
&&
3578 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3581 if (later
->ds_dir
== earlier
->ds_dir
)
3583 if (!dsl_dir_is_clone(later
->ds_dir
))
3586 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3588 error
= dsl_dataset_hold_obj(dp
,
3589 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3592 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3593 dsl_dataset_rele(origin
, FTAG
);
3598 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
3600 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
3601 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
3605 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
3607 dmu_object_info_t doi
;
3609 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
3610 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
3614 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
3616 return (dsl_dataset_is_zapified(ds
) &&
3617 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
3618 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
3621 #if defined(_KERNEL) && defined(HAVE_SPL)
3623 module_param(zfs_max_recordsize
, int, 0644);
3624 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3626 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
3627 module_param(zfs_max_recordsize
, int, 0444);
3628 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3631 EXPORT_SYMBOL(dsl_dataset_hold
);
3632 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
3633 EXPORT_SYMBOL(dsl_dataset_own
);
3634 EXPORT_SYMBOL(dsl_dataset_own_obj
);
3635 EXPORT_SYMBOL(dsl_dataset_name
);
3636 EXPORT_SYMBOL(dsl_dataset_rele
);
3637 EXPORT_SYMBOL(dsl_dataset_disown
);
3638 EXPORT_SYMBOL(dsl_dataset_tryown
);
3639 EXPORT_SYMBOL(dsl_dataset_create_sync
);
3640 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
3641 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
3642 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
3643 EXPORT_SYMBOL(dsl_dataset_promote
);
3644 EXPORT_SYMBOL(dsl_dataset_user_hold
);
3645 EXPORT_SYMBOL(dsl_dataset_user_release
);
3646 EXPORT_SYMBOL(dsl_dataset_get_holds
);
3647 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
3648 EXPORT_SYMBOL(dsl_dataset_get_spa
);
3649 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
3650 EXPORT_SYMBOL(dsl_dataset_space_written
);
3651 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
3652 EXPORT_SYMBOL(dsl_dataset_sync
);
3653 EXPORT_SYMBOL(dsl_dataset_block_born
);
3654 EXPORT_SYMBOL(dsl_dataset_block_kill
);
3655 EXPORT_SYMBOL(dsl_dataset_block_freeable
);
3656 EXPORT_SYMBOL(dsl_dataset_prev_snap_txg
);
3657 EXPORT_SYMBOL(dsl_dataset_dirty
);
3658 EXPORT_SYMBOL(dsl_dataset_stats
);
3659 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
3660 EXPORT_SYMBOL(dsl_dataset_space
);
3661 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
3662 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
3663 EXPORT_SYMBOL(dsl_dataset_check_quota
);
3664 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
3665 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);