4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
25 * Copyright (c) 2014 RackTop Systems.
26 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
29 #include <sys/dmu_objset.h>
30 #include <sys/dsl_dataset.h>
31 #include <sys/dsl_dir.h>
32 #include <sys/dsl_prop.h>
33 #include <sys/dsl_synctask.h>
34 #include <sys/dmu_traverse.h>
35 #include <sys/dmu_impl.h>
36 #include <sys/dmu_tx.h>
40 #include <sys/zfeature.h>
41 #include <sys/unique.h>
42 #include <sys/zfs_context.h>
43 #include <sys/zfs_ioctl.h>
45 #include <sys/zfs_znode.h>
46 #include <sys/zfs_onexit.h>
48 #include <sys/dsl_scan.h>
49 #include <sys/dsl_deadlist.h>
50 #include <sys/dsl_destroy.h>
51 #include <sys/dsl_userhold.h>
52 #include <sys/dsl_bookmark.h>
54 #define SWITCH64(x, y) \
56 uint64_t __tmp = (x); \
61 #define DS_REF_MAX (1ULL << 62)
63 #define DSL_DEADLIST_BLOCKSIZE SPA_MAXBLOCKSIZE
65 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
68 * Figure out how much of this delta should be propogated to the dsl_dir
69 * layer. If there's a refreservation, that space has already been
70 * partially accounted for in our ancestors.
73 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
75 dsl_dataset_phys_t
*ds_phys
;
76 uint64_t old_bytes
, new_bytes
;
78 if (ds
->ds_reserved
== 0)
81 ds_phys
= dsl_dataset_phys(ds
);
82 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
83 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
85 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
86 return (new_bytes
- old_bytes
);
90 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
92 int used
, compressed
, uncompressed
;
95 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
96 compressed
= BP_GET_PSIZE(bp
);
97 uncompressed
= BP_GET_UCSIZE(bp
);
99 dprintf_bp(bp
, "ds=%p", ds
);
101 ASSERT(dmu_tx_is_syncing(tx
));
102 /* It could have been compressed away to nothing */
105 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
106 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
108 dsl_pool_mos_diduse_space(tx
->tx_pool
,
109 used
, compressed
, uncompressed
);
113 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
114 mutex_enter(&ds
->ds_lock
);
115 delta
= parent_delta(ds
, used
);
116 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
117 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
118 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
119 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
120 mutex_exit(&ds
->ds_lock
);
121 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
122 compressed
, uncompressed
, tx
);
123 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
124 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
128 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
131 int used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
132 int compressed
= BP_GET_PSIZE(bp
);
133 int uncompressed
= BP_GET_UCSIZE(bp
);
138 ASSERT(dmu_tx_is_syncing(tx
));
139 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
142 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
143 dsl_pool_mos_diduse_space(tx
->tx_pool
,
144 -used
, -compressed
, -uncompressed
);
147 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
149 ASSERT(!ds
->ds_is_snapshot
);
150 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
152 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
155 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
156 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
158 mutex_enter(&ds
->ds_lock
);
159 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
160 !DS_UNIQUE_IS_ACCURATE(ds
));
161 delta
= parent_delta(ds
, -used
);
162 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
163 mutex_exit(&ds
->ds_lock
);
164 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
165 delta
, -compressed
, -uncompressed
, tx
);
166 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
167 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
169 dprintf_bp(bp
, "putting on dead list: %s", "");
172 * We are here as part of zio's write done callback,
173 * which means we're a zio interrupt thread. We can't
174 * call dsl_deadlist_insert() now because it may block
175 * waiting for I/O. Instead, put bp on the deferred
176 * queue and let dsl_pool_sync() finish the job.
178 bplist_append(&ds
->ds_pending_deadlist
, bp
);
180 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
182 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
183 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
184 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
185 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
186 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
187 ds
->ds_object
&& bp
->blk_birth
>
188 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
189 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
190 mutex_enter(&ds
->ds_prev
->ds_lock
);
191 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
192 mutex_exit(&ds
->ds_prev
->ds_lock
);
194 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
195 dsl_dir_transfer_space(ds
->ds_dir
, used
,
196 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
199 mutex_enter(&ds
->ds_lock
);
200 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
201 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
202 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
203 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
204 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
205 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
206 mutex_exit(&ds
->ds_lock
);
212 dsl_dataset_prev_snap_txg(dsl_dataset_t
*ds
)
214 uint64_t trysnap
= 0;
219 * The snapshot creation could fail, but that would cause an
220 * incorrect FALSE return, which would only result in an
221 * overestimation of the amount of space that an operation would
222 * consume, which is OK.
224 * There's also a small window where we could miss a pending
225 * snapshot, because we could set the sync task in the quiescing
226 * phase. So this should only be used as a guess.
228 if (ds
->ds_trysnap_txg
>
229 spa_last_synced_txg(ds
->ds_dir
->dd_pool
->dp_spa
))
230 trysnap
= ds
->ds_trysnap_txg
;
231 return (MAX(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, trysnap
));
235 dsl_dataset_block_freeable(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
238 if (blk_birth
<= dsl_dataset_prev_snap_txg(ds
) ||
239 (bp
!= NULL
&& BP_IS_HOLE(bp
)))
242 ddt_prefetch(dsl_dataset_get_spa(ds
), bp
);
248 dsl_dataset_evict(void *dbu
)
250 dsl_dataset_t
*ds
= dbu
;
252 ASSERT(ds
->ds_owner
== NULL
);
256 unique_remove(ds
->ds_fsid_guid
);
258 if (ds
->ds_objset
!= NULL
)
259 dmu_objset_evict(ds
->ds_objset
);
262 dsl_dataset_rele(ds
->ds_prev
, ds
);
266 bplist_destroy(&ds
->ds_pending_deadlist
);
267 if (ds
->ds_deadlist
.dl_os
!= NULL
)
268 dsl_deadlist_close(&ds
->ds_deadlist
);
270 dsl_dir_async_rele(ds
->ds_dir
, ds
);
272 ASSERT(!list_link_active(&ds
->ds_synced_link
));
274 mutex_destroy(&ds
->ds_lock
);
275 mutex_destroy(&ds
->ds_opening_lock
);
276 mutex_destroy(&ds
->ds_sendstream_lock
);
277 refcount_destroy(&ds
->ds_longholds
);
279 kmem_free(ds
, sizeof (dsl_dataset_t
));
283 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
285 dsl_dataset_phys_t
*headphys
;
288 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
289 objset_t
*mos
= dp
->dp_meta_objset
;
291 if (ds
->ds_snapname
[0])
293 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
296 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
300 headphys
= headdbuf
->db_data
;
301 err
= zap_value_search(dp
->dp_meta_objset
,
302 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
303 dmu_buf_rele(headdbuf
, FTAG
);
308 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
310 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
311 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
315 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
320 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
321 value
, mt
, NULL
, 0, NULL
);
322 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
323 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
328 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
331 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
332 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
336 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
338 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
343 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
344 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
345 err
= zap_remove(mos
, snapobj
, name
, tx
);
347 if (err
== 0 && adj_cnt
)
348 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
349 DD_FIELD_SNAPSHOT_COUNT
, tx
);
355 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
358 objset_t
*mos
= dp
->dp_meta_objset
;
362 dmu_object_info_t doi
;
364 ASSERT(dsl_pool_config_held(dp
));
366 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
370 /* Make sure dsobj has the correct object type. */
371 dmu_object_info_from_db(dbuf
, &doi
);
372 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
373 dmu_buf_rele(dbuf
, tag
);
374 return (SET_ERROR(EINVAL
));
377 ds
= dmu_buf_get_user(dbuf
);
379 dsl_dataset_t
*winner
= NULL
;
381 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
383 ds
->ds_object
= dsobj
;
384 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
385 list_link_init(&ds
->ds_synced_link
);
387 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
388 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
389 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
390 refcount_create(&ds
->ds_longholds
);
392 bplist_create(&ds
->ds_pending_deadlist
);
393 dsl_deadlist_open(&ds
->ds_deadlist
,
394 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
396 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
397 offsetof(dmu_sendarg_t
, dsa_link
));
400 err
= dsl_dir_hold_obj(dp
,
401 dsl_dataset_phys(ds
)->ds_dir_obj
, NULL
, ds
,
405 mutex_destroy(&ds
->ds_lock
);
406 mutex_destroy(&ds
->ds_opening_lock
);
407 mutex_destroy(&ds
->ds_sendstream_lock
);
408 refcount_destroy(&ds
->ds_longholds
);
409 bplist_destroy(&ds
->ds_pending_deadlist
);
410 dsl_deadlist_close(&ds
->ds_deadlist
);
411 kmem_free(ds
, sizeof (dsl_dataset_t
));
412 dmu_buf_rele(dbuf
, tag
);
416 if (!ds
->ds_is_snapshot
) {
417 ds
->ds_snapname
[0] = '\0';
418 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
419 err
= dsl_dataset_hold_obj(dp
,
420 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
423 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
424 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
425 DS_FIELD_BOOKMARK_NAMES
,
426 sizeof (ds
->ds_bookmarks
), 1,
428 if (zaperr
!= ENOENT
)
432 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
433 err
= dsl_dataset_get_snapname(ds
);
435 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
437 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
438 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
443 if (err
== 0 && !ds
->ds_is_snapshot
) {
444 err
= dsl_prop_get_int_ds(ds
,
445 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
448 err
= dsl_prop_get_int_ds(ds
,
449 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
453 ds
->ds_reserved
= ds
->ds_quota
= 0;
456 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict
, &ds
->ds_dbuf
);
458 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
460 if (err
!= 0 || winner
!= NULL
) {
461 bplist_destroy(&ds
->ds_pending_deadlist
);
462 dsl_deadlist_close(&ds
->ds_deadlist
);
464 dsl_dataset_rele(ds
->ds_prev
, ds
);
465 dsl_dir_rele(ds
->ds_dir
, ds
);
466 mutex_destroy(&ds
->ds_lock
);
467 mutex_destroy(&ds
->ds_opening_lock
);
468 mutex_destroy(&ds
->ds_sendstream_lock
);
469 refcount_destroy(&ds
->ds_longholds
);
470 kmem_free(ds
, sizeof (dsl_dataset_t
));
472 dmu_buf_rele(dbuf
, tag
);
478 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
481 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
482 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
483 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
484 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
485 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
491 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
,
492 void *tag
, dsl_dataset_t
**dsp
)
495 const char *snapname
;
499 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
503 ASSERT(dsl_pool_config_held(dp
));
504 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
506 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, dsp
);
508 err
= SET_ERROR(ENOENT
);
510 /* we may be looking for a snapshot */
511 if (err
== 0 && snapname
!= NULL
) {
514 if (*snapname
++ != '@') {
515 dsl_dataset_rele(*dsp
, tag
);
516 dsl_dir_rele(dd
, FTAG
);
517 return (SET_ERROR(ENOENT
));
520 dprintf("looking for snapshot '%s'\n", snapname
);
521 err
= dsl_dataset_snap_lookup(*dsp
, snapname
, &obj
);
523 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
524 dsl_dataset_rele(*dsp
, tag
);
527 mutex_enter(&ds
->ds_lock
);
528 if (ds
->ds_snapname
[0] == 0)
529 (void) strlcpy(ds
->ds_snapname
, snapname
,
530 sizeof (ds
->ds_snapname
));
531 mutex_exit(&ds
->ds_lock
);
536 dsl_dir_rele(dd
, FTAG
);
541 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
,
542 void *tag
, dsl_dataset_t
**dsp
)
544 int err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
547 if (!dsl_dataset_tryown(*dsp
, tag
)) {
548 dsl_dataset_rele(*dsp
, tag
);
550 return (SET_ERROR(EBUSY
));
556 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
,
557 void *tag
, dsl_dataset_t
**dsp
)
559 int err
= dsl_dataset_hold(dp
, name
, tag
, dsp
);
562 if (!dsl_dataset_tryown(*dsp
, tag
)) {
563 dsl_dataset_rele(*dsp
, tag
);
564 return (SET_ERROR(EBUSY
));
570 * See the comment above dsl_pool_hold() for details. In summary, a long
571 * hold is used to prevent destruction of a dataset while the pool hold
572 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
574 * The dataset and pool must be held when this function is called. After it
575 * is called, the pool hold may be released while the dataset is still held
579 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
581 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
582 (void) refcount_add(&ds
->ds_longholds
, tag
);
586 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
588 (void) refcount_remove(&ds
->ds_longholds
, tag
);
591 /* Return B_TRUE if there are any long holds on this dataset. */
593 dsl_dataset_long_held(dsl_dataset_t
*ds
)
595 return (!refcount_is_zero(&ds
->ds_longholds
));
599 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
602 (void) strcpy(name
, "mos");
604 dsl_dir_name(ds
->ds_dir
, name
);
605 VERIFY0(dsl_dataset_get_snapname(ds
));
606 if (ds
->ds_snapname
[0]) {
607 (void) strcat(name
, "@");
609 * We use a "recursive" mutex so that we
610 * can call dprintf_ds() with ds_lock held.
612 if (!MUTEX_HELD(&ds
->ds_lock
)) {
613 mutex_enter(&ds
->ds_lock
);
614 (void) strcat(name
, ds
->ds_snapname
);
615 mutex_exit(&ds
->ds_lock
);
617 (void) strcat(name
, ds
->ds_snapname
);
624 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
626 dmu_buf_rele(ds
->ds_dbuf
, tag
);
630 dsl_dataset_disown(dsl_dataset_t
*ds
, void *tag
)
632 ASSERT3P(ds
->ds_owner
, ==, tag
);
633 ASSERT(ds
->ds_dbuf
!= NULL
);
635 mutex_enter(&ds
->ds_lock
);
637 mutex_exit(&ds
->ds_lock
);
638 dsl_dataset_long_rele(ds
, tag
);
639 dsl_dataset_rele(ds
, tag
);
643 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
645 boolean_t gotit
= FALSE
;
647 mutex_enter(&ds
->ds_lock
);
648 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
650 dsl_dataset_long_hold(ds
, tag
);
653 mutex_exit(&ds
->ds_lock
);
658 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
659 uint64_t flags
, dmu_tx_t
*tx
)
661 dsl_pool_t
*dp
= dd
->dd_pool
;
663 dsl_dataset_phys_t
*dsphys
;
665 objset_t
*mos
= dp
->dp_meta_objset
;
668 origin
= dp
->dp_origin_snap
;
670 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
671 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
672 ASSERT(dmu_tx_is_syncing(tx
));
673 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
675 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
676 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
677 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
678 dmu_buf_will_dirty(dbuf
, tx
);
679 dsphys
= dbuf
->db_data
;
680 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
681 dsphys
->ds_dir_obj
= dd
->dd_object
;
682 dsphys
->ds_flags
= flags
;
683 dsphys
->ds_fsid_guid
= unique_create();
684 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
685 sizeof (dsphys
->ds_guid
));
686 dsphys
->ds_snapnames_zapobj
=
687 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
689 dsphys
->ds_creation_time
= gethrestime_sec();
690 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
692 if (origin
== NULL
) {
693 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
695 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
697 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
698 dsphys
->ds_prev_snap_txg
=
699 dsl_dataset_phys(origin
)->ds_creation_txg
;
700 dsphys
->ds_referenced_bytes
=
701 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
702 dsphys
->ds_compressed_bytes
=
703 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
704 dsphys
->ds_uncompressed_bytes
=
705 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
706 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
709 * Inherit flags that describe the dataset's contents
710 * (INCONSISTENT) or properties (Case Insensitive).
712 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
713 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
715 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
716 dsl_dataset_phys(origin
)->ds_num_children
++;
718 VERIFY0(dsl_dataset_hold_obj(dp
,
719 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
721 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
722 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
723 dsl_dataset_rele(ohds
, FTAG
);
725 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
726 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
727 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
729 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
731 VERIFY0(zap_add_int(mos
,
732 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
736 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
737 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
738 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
739 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
740 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
741 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
743 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
745 VERIFY0(zap_add_int(mos
,
746 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
751 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
752 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
754 dmu_buf_rele(dbuf
, FTAG
);
756 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
757 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
763 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
767 VERIFY0(dmu_objset_from_ds(ds
, &os
));
768 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
769 dsl_dataset_dirty(ds
, tx
);
773 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
774 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
, dmu_tx_t
*tx
)
776 dsl_pool_t
*dp
= pdd
->dd_pool
;
777 uint64_t dsobj
, ddobj
;
780 ASSERT(dmu_tx_is_syncing(tx
));
781 ASSERT(lastname
[0] != '@');
783 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
784 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
786 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
,
787 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
789 dsl_deleg_set_create_perms(dd
, tx
, cr
);
792 * Since we're creating a new node we know it's a leaf, so we can
793 * initialize the counts if the limit feature is active.
795 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
797 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
799 dsl_dir_zapify(dd
, tx
);
800 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
801 sizeof (cnt
), 1, &cnt
, tx
));
802 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
803 sizeof (cnt
), 1, &cnt
, tx
));
806 dsl_dir_rele(dd
, FTAG
);
809 * If we are creating a clone, make sure we zero out any stale
810 * data from the origin snapshots zil header.
812 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
815 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
816 dsl_dataset_zero_zil(ds
, tx
);
817 dsl_dataset_rele(ds
, FTAG
);
824 * The unique space in the head dataset can be calculated by subtracting
825 * the space used in the most recent snapshot, that is still being used
826 * in this file system, from the space currently in use. To figure out
827 * the space in the most recent snapshot still in use, we need to take
828 * the total space used in the snapshot and subtract out the space that
829 * has been freed up since the snapshot was taken.
832 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
835 uint64_t dlused
, dlcomp
, dluncomp
;
837 ASSERT(!ds
->ds_is_snapshot
);
839 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
840 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
844 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
846 ASSERT3U(dlused
, <=, mrs_used
);
847 dsl_dataset_phys(ds
)->ds_unique_bytes
=
848 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
850 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
851 SPA_VERSION_UNIQUE_ACCURATE
)
852 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
856 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
859 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
861 ASSERTV(uint64_t count
);
863 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
864 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
867 * The err should not be ENOENT, but a bug in a previous version
868 * of the code could cause upgrade_clones_cb() to not set
869 * ds_next_snap_obj when it should, leading to a missing entry.
870 * If we knew that the pool was created after
871 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
872 * ENOENT. However, at least we can check that we don't have
873 * too many entries in the next_clones_obj even after failing to
878 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
880 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
885 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
887 return (&dsl_dataset_phys(ds
)->ds_bp
);
891 dsl_dataset_set_blkptr(dsl_dataset_t
*ds
, blkptr_t
*bp
, dmu_tx_t
*tx
)
893 ASSERT(dmu_tx_is_syncing(tx
));
894 /* If it's the meta-objset, set dp_meta_rootbp */
896 tx
->tx_pool
->dp_meta_rootbp
= *bp
;
898 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
899 dsl_dataset_phys(ds
)->ds_bp
= *bp
;
904 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
906 return (ds
->ds_dir
->dd_pool
->dp_spa
);
910 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
914 if (ds
== NULL
) /* this is the meta-objset */
917 ASSERT(ds
->ds_objset
!= NULL
);
919 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
920 panic("dirtying snapshot!");
922 dp
= ds
->ds_dir
->dd_pool
;
924 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
925 /* up the hold count until we can be written out */
926 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
931 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
935 for (t
= 0; t
< TXG_SIZE
; t
++) {
936 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
944 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
948 if (!dmu_tx_is_syncing(tx
))
952 * If there's an fs-only reservation, any blocks that might become
953 * owned by the snapshot dataset must be accommodated by space
954 * outside of the reservation.
956 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
957 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
958 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
959 return (SET_ERROR(ENOSPC
));
962 * Propagate any reserved space for this snapshot to other
963 * snapshot checks in this sync group.
966 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
971 typedef struct dsl_dataset_snapshot_arg
{
972 nvlist_t
*ddsa_snaps
;
973 nvlist_t
*ddsa_props
;
974 nvlist_t
*ddsa_errors
;
976 } dsl_dataset_snapshot_arg_t
;
979 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
980 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
985 ds
->ds_trysnap_txg
= tx
->tx_txg
;
987 if (!dmu_tx_is_syncing(tx
))
991 * We don't allow multiple snapshots of the same txg. If there
992 * is already one, try again.
994 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
995 return (SET_ERROR(EAGAIN
));
998 * Check for conflicting snapshot name.
1000 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1002 return (SET_ERROR(EEXIST
));
1003 if (error
!= ENOENT
)
1007 * We don't allow taking snapshots of inconsistent datasets, such as
1008 * those into which we are currently receiving. However, if we are
1009 * creating this snapshot as part of a receive, this check will be
1010 * executed atomically with respect to the completion of the receive
1011 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1012 * case we ignore this, knowing it will be fixed up for us shortly in
1013 * dmu_recv_end_sync().
1015 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1016 return (SET_ERROR(EBUSY
));
1019 * Skip the check for temporary snapshots or if we have already checked
1020 * the counts in dsl_dataset_snapshot_check. This means we really only
1021 * check the count here when we're receiving a stream.
1023 if (cnt
!= 0 && cr
!= NULL
) {
1024 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1025 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1030 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1038 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1040 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1041 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1046 * Pre-compute how many total new snapshots will be created for each
1047 * level in the tree and below. This is needed for validating the
1048 * snapshot limit when either taking a recursive snapshot or when
1049 * taking multiple snapshots.
1051 * The problem is that the counts are not actually adjusted when
1052 * we are checking, only when we finally sync. For a single snapshot,
1053 * this is easy, the count will increase by 1 at each node up the tree,
1054 * but its more complicated for the recursive/multiple snapshot case.
1056 * The dsl_fs_ss_limit_check function does recursively check the count
1057 * at each level up the tree but since it is validating each snapshot
1058 * independently we need to be sure that we are validating the complete
1059 * count for the entire set of snapshots. We do this by rolling up the
1060 * counts for each component of the name into an nvlist and then
1061 * checking each of those cases with the aggregated count.
1063 * This approach properly handles not only the recursive snapshot
1064 * case (where we get all of those on the ddsa_snaps list) but also
1065 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1066 * validate the limit on 'a' using a count of 2).
1068 * We validate the snapshot names in the third loop and only report
1071 if (dmu_tx_is_syncing(tx
)) {
1073 nvlist_t
*cnt_track
= NULL
;
1074 cnt_track
= fnvlist_alloc();
1076 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1078 /* Rollup aggregated counts into the cnt_track list */
1079 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1081 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1085 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1086 pdelim
= strchr(nm
, '@');
1092 if (nvlist_lookup_uint64(cnt_track
, nm
,
1094 /* update existing entry */
1095 fnvlist_add_uint64(cnt_track
, nm
,
1099 fnvlist_add_uint64(cnt_track
, nm
, 1);
1102 pdelim
= strrchr(nm
, '/');
1105 } while (pdelim
!= NULL
);
1108 kmem_free(nm
, MAXPATHLEN
);
1110 /* Check aggregated counts at each level */
1111 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1112 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1118 name
= nvpair_name(pair
);
1119 cnt
= fnvpair_value_uint64(pair
);
1122 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1124 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1125 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1127 dsl_dataset_rele(ds
, FTAG
);
1131 if (ddsa
->ddsa_errors
!= NULL
)
1132 fnvlist_add_int32(ddsa
->ddsa_errors
,
1135 /* only report one error for this check */
1139 nvlist_free(cnt_track
);
1142 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1143 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1147 char dsname
[MAXNAMELEN
];
1149 name
= nvpair_name(pair
);
1150 if (strlen(name
) >= MAXNAMELEN
)
1151 error
= SET_ERROR(ENAMETOOLONG
);
1153 atp
= strchr(name
, '@');
1155 error
= SET_ERROR(EINVAL
);
1157 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1160 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1162 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1163 error
= dsl_dataset_snapshot_check_impl(ds
,
1164 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1165 dsl_dataset_rele(ds
, FTAG
);
1169 if (ddsa
->ddsa_errors
!= NULL
) {
1170 fnvlist_add_int32(ddsa
->ddsa_errors
,
1181 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1184 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1186 dsl_dataset_phys_t
*dsphys
;
1187 uint64_t dsobj
, crtxg
;
1188 objset_t
*mos
= dp
->dp_meta_objset
;
1189 ASSERTV(static zil_header_t zero_zil
);
1190 ASSERTV(objset_t
*os
);
1192 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1195 * If we are on an old pool, the zil must not be active, in which
1196 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1198 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1199 dmu_objset_from_ds(ds
, &os
) != 0 ||
1200 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1201 sizeof (zero_zil
)) == 0);
1203 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1206 * The origin's ds_creation_txg has to be < TXG_INITIAL
1208 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1213 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1214 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1215 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1216 dmu_buf_will_dirty(dbuf
, tx
);
1217 dsphys
= dbuf
->db_data
;
1218 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1219 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1220 dsphys
->ds_fsid_guid
= unique_create();
1221 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1222 sizeof (dsphys
->ds_guid
));
1223 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1224 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1225 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1226 dsphys
->ds_num_children
= 1;
1227 dsphys
->ds_creation_time
= gethrestime_sec();
1228 dsphys
->ds_creation_txg
= crtxg
;
1229 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1230 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1231 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1232 dsphys
->ds_uncompressed_bytes
=
1233 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1234 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1235 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1236 dmu_buf_rele(dbuf
, FTAG
);
1238 ASSERT3U(ds
->ds_prev
!= 0, ==,
1239 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1241 uint64_t next_clones_obj
=
1242 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1243 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1245 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1246 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1248 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1249 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1250 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1251 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1252 } else if (next_clones_obj
!= 0) {
1253 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1254 dsphys
->ds_next_snap_obj
, tx
);
1255 VERIFY0(zap_add_int(mos
,
1256 next_clones_obj
, dsobj
, tx
));
1261 * If we have a reference-reservation on this dataset, we will
1262 * need to increase the amount of refreservation being charged
1263 * since our unique space is going to zero.
1265 if (ds
->ds_reserved
) {
1267 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1268 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1270 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1274 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1275 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1276 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1277 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1278 dsl_deadlist_close(&ds
->ds_deadlist
);
1279 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1280 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1281 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1282 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1284 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1285 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1286 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1287 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1288 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1289 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1291 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1292 snapname
, 8, 1, &dsobj
, tx
));
1295 dsl_dataset_rele(ds
->ds_prev
, ds
);
1296 VERIFY0(dsl_dataset_hold_obj(dp
,
1297 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1299 dsl_scan_ds_snapshotted(ds
, tx
);
1301 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1303 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1307 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1309 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1310 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1313 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1314 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1317 char dsname
[MAXNAMELEN
];
1319 name
= nvpair_name(pair
);
1320 atp
= strchr(name
, '@');
1321 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1322 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1324 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1325 if (ddsa
->ddsa_props
!= NULL
) {
1326 dsl_props_set_sync_impl(ds
->ds_prev
,
1327 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1329 dsl_dataset_rele(ds
, FTAG
);
1334 * The snapshots must all be in the same pool.
1335 * All-or-nothing: if there are any failures, nothing will be modified.
1338 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1340 dsl_dataset_snapshot_arg_t ddsa
;
1342 boolean_t needsuspend
;
1346 nvlist_t
*suspended
= NULL
;
1348 pair
= nvlist_next_nvpair(snaps
, NULL
);
1351 firstname
= nvpair_name(pair
);
1353 error
= spa_open(firstname
, &spa
, FTAG
);
1356 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1357 spa_close(spa
, FTAG
);
1360 suspended
= fnvlist_alloc();
1361 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1362 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1363 char fsname
[MAXNAMELEN
];
1364 char *snapname
= nvpair_name(pair
);
1368 atp
= strchr(snapname
, '@');
1370 error
= SET_ERROR(EINVAL
);
1373 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1375 error
= zil_suspend(fsname
, &cookie
);
1378 fnvlist_add_uint64(suspended
, fsname
,
1383 ddsa
.ddsa_snaps
= snaps
;
1384 ddsa
.ddsa_props
= props
;
1385 ddsa
.ddsa_errors
= errors
;
1386 ddsa
.ddsa_cr
= CRED();
1389 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1390 dsl_dataset_snapshot_sync
, &ddsa
,
1391 fnvlist_num_pairs(snaps
) * 3);
1394 if (suspended
!= NULL
) {
1395 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1396 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1397 zil_resume((void *)(uintptr_t)
1398 fnvpair_value_uint64(pair
));
1400 fnvlist_free(suspended
);
1405 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1406 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1407 char *snapname
= nvpair_name(pair
);
1408 zvol_create_minors(snapname
);
1416 typedef struct dsl_dataset_snapshot_tmp_arg
{
1417 const char *ddsta_fsname
;
1418 const char *ddsta_snapname
;
1419 minor_t ddsta_cleanup_minor
;
1420 const char *ddsta_htag
;
1421 } dsl_dataset_snapshot_tmp_arg_t
;
1424 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1426 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1427 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1431 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1435 /* NULL cred means no limit check for tmp snapshot */
1436 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1437 tx
, B_FALSE
, 0, NULL
);
1439 dsl_dataset_rele(ds
, FTAG
);
1443 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1444 dsl_dataset_rele(ds
, FTAG
);
1445 return (SET_ERROR(ENOTSUP
));
1447 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1450 dsl_dataset_rele(ds
, FTAG
);
1454 dsl_dataset_rele(ds
, FTAG
);
1459 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1461 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1462 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1465 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1467 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1468 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1469 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1470 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1472 dsl_dataset_rele(ds
, FTAG
);
1476 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1477 minor_t cleanup_minor
, const char *htag
)
1479 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1482 boolean_t needsuspend
;
1485 ddsta
.ddsta_fsname
= fsname
;
1486 ddsta
.ddsta_snapname
= snapname
;
1487 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1488 ddsta
.ddsta_htag
= htag
;
1490 error
= spa_open(fsname
, &spa
, FTAG
);
1493 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1494 spa_close(spa
, FTAG
);
1497 error
= zil_suspend(fsname
, &cookie
);
1502 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1503 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3);
1512 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1514 ASSERT(dmu_tx_is_syncing(tx
));
1515 ASSERT(ds
->ds_objset
!= NULL
);
1516 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1519 * in case we had to change ds_fsid_guid when we opened it,
1522 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1523 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1525 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1529 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1532 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1535 nvlist_t
*propval
= fnvlist_alloc();
1536 nvlist_t
*val
= fnvlist_alloc();
1538 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1541 * There may be missing entries in ds_next_clones_obj
1542 * due to a bug in a previous version of the code.
1543 * Only trust it if it has the right number of entries.
1545 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1546 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1549 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1)
1551 for (zap_cursor_init(&zc
, mos
,
1552 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1553 zap_cursor_retrieve(&zc
, &za
) == 0;
1554 zap_cursor_advance(&zc
)) {
1555 dsl_dataset_t
*clone
;
1556 char buf
[ZFS_MAXNAMELEN
];
1557 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1558 za
.za_first_integer
, FTAG
, &clone
));
1559 dsl_dir_name(clone
->ds_dir
, buf
);
1560 fnvlist_add_boolean(val
, buf
);
1561 dsl_dataset_rele(clone
, FTAG
);
1563 zap_cursor_fini(&zc
);
1564 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1565 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
), propval
);
1568 nvlist_free(propval
);
1572 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1574 uint64_t refd
, avail
, uobjs
, aobjs
, ratio
;
1575 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
1577 ASSERT(dsl_pool_config_held(dp
));
1579 ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1580 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1581 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1583 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
, ratio
);
1584 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
1585 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1587 if (ds
->ds_is_snapshot
) {
1588 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
, ratio
);
1589 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
1590 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1591 get_clones_stat(ds
, nv
);
1593 dsl_dir_stats(ds
->ds_dir
, nv
);
1596 dsl_dataset_space(ds
, &refd
, &avail
, &uobjs
, &aobjs
);
1597 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
, avail
);
1598 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
, refd
);
1600 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
1601 dsl_dataset_phys(ds
)->ds_creation_time
);
1602 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
1603 dsl_dataset_phys(ds
)->ds_creation_txg
);
1604 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
1606 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
1608 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
1609 dsl_dataset_phys(ds
)->ds_guid
);
1610 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
1611 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1612 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
1614 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
1616 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
1617 DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
1619 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
1620 uint64_t written
, comp
, uncomp
;
1621 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1622 dsl_dataset_t
*prev
;
1625 err
= dsl_dataset_hold_obj(dp
,
1626 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
1628 err
= dsl_dataset_space_written(prev
, ds
, &written
,
1630 dsl_dataset_rele(prev
, FTAG
);
1632 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
1641 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
1643 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1644 ASSERT(dsl_pool_config_held(dp
));
1646 stat
->dds_creation_txg
= dsl_dataset_phys(ds
)->ds_creation_txg
;
1647 stat
->dds_inconsistent
=
1648 dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
;
1649 stat
->dds_guid
= dsl_dataset_phys(ds
)->ds_guid
;
1650 stat
->dds_origin
[0] = '\0';
1651 if (ds
->ds_is_snapshot
) {
1652 stat
->dds_is_snapshot
= B_TRUE
;
1653 stat
->dds_num_clones
=
1654 dsl_dataset_phys(ds
)->ds_num_children
- 1;
1656 stat
->dds_is_snapshot
= B_FALSE
;
1657 stat
->dds_num_clones
= 0;
1659 if (dsl_dir_is_clone(ds
->ds_dir
)) {
1662 VERIFY0(dsl_dataset_hold_obj(dp
,
1663 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
,
1665 dsl_dataset_name(ods
, stat
->dds_origin
);
1666 dsl_dataset_rele(ods
, FTAG
);
1672 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
1674 return (ds
->ds_fsid_guid
);
1678 dsl_dataset_space(dsl_dataset_t
*ds
,
1679 uint64_t *refdbytesp
, uint64_t *availbytesp
,
1680 uint64_t *usedobjsp
, uint64_t *availobjsp
)
1682 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1683 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
1684 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
1686 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
1687 if (ds
->ds_quota
!= 0) {
1689 * Adjust available bytes according to refquota
1691 if (*refdbytesp
< ds
->ds_quota
)
1692 *availbytesp
= MIN(*availbytesp
,
1693 ds
->ds_quota
- *refdbytesp
);
1697 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
1698 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
1702 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
1704 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1707 if (dsl_dataset_phys(ds
)->ds_bp
.blk_birth
>
1708 dsl_dataset_phys(snap
)->ds_creation_txg
) {
1709 objset_t
*os
, *os_snap
;
1711 * It may be that only the ZIL differs, because it was
1712 * reset in the head. Don't count that as being
1715 if (dmu_objset_from_ds(ds
, &os
) != 0)
1717 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
1719 return (bcmp(&os
->os_phys
->os_meta_dnode
,
1720 &os_snap
->os_phys
->os_meta_dnode
,
1721 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
1726 typedef struct dsl_dataset_rename_snapshot_arg
{
1727 const char *ddrsa_fsname
;
1728 const char *ddrsa_oldsnapname
;
1729 const char *ddrsa_newsnapname
;
1730 boolean_t ddrsa_recursive
;
1732 } dsl_dataset_rename_snapshot_arg_t
;
1736 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
1737 dsl_dataset_t
*hds
, void *arg
)
1739 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1743 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1745 /* ignore nonexistent snapshots */
1746 return (error
== ENOENT
? 0 : error
);
1749 /* new name should not exist */
1750 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
1752 error
= SET_ERROR(EEXIST
);
1753 else if (error
== ENOENT
)
1756 /* dataset name + 1 for the "@" + the new snapshot name must fit */
1757 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
1758 strlen(ddrsa
->ddrsa_newsnapname
) >= MAXNAMELEN
)
1759 error
= SET_ERROR(ENAMETOOLONG
);
1765 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1767 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1768 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1772 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
1776 if (ddrsa
->ddrsa_recursive
) {
1777 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1778 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
1781 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
1783 dsl_dataset_rele(hds
, FTAG
);
1788 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
1789 dsl_dataset_t
*hds
, void *arg
)
1791 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1794 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
1797 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1798 ASSERT(error
== 0 || error
== ENOENT
);
1799 if (error
== ENOENT
) {
1800 /* ignore nonexistent snapshots */
1804 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
1806 /* log before we change the name */
1807 spa_history_log_internal_ds(ds
, "rename", tx
,
1808 "-> @%s", ddrsa
->ddrsa_newsnapname
);
1810 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
1812 mutex_enter(&ds
->ds_lock
);
1813 (void) strcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
);
1814 mutex_exit(&ds
->ds_lock
);
1815 VERIFY0(zap_add(dp
->dp_meta_objset
,
1816 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
1817 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
1819 dsl_dataset_rele(ds
, FTAG
);
1824 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1826 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1827 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1830 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
1831 ddrsa
->ddrsa_tx
= tx
;
1832 if (ddrsa
->ddrsa_recursive
) {
1833 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1834 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
1837 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
1839 dsl_dataset_rele(hds
, FTAG
);
1843 dsl_dataset_rename_snapshot(const char *fsname
,
1844 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
1847 char *oldname
, *newname
;
1851 dsl_dataset_rename_snapshot_arg_t ddrsa
;
1853 ddrsa
.ddrsa_fsname
= fsname
;
1854 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
1855 ddrsa
.ddrsa_newsnapname
= newsnapname
;
1856 ddrsa
.ddrsa_recursive
= recursive
;
1858 error
= dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
1859 dsl_dataset_rename_snapshot_sync
, &ddrsa
, 1);
1862 return (SET_ERROR(error
));
1865 oldname
= kmem_asprintf("%s@%s", fsname
, oldsnapname
);
1866 newname
= kmem_asprintf("%s@%s", fsname
, newsnapname
);
1867 zvol_rename_minors(oldname
, newname
);
1876 * If we're doing an ownership handoff, we need to make sure that there is
1877 * only one long hold on the dataset. We're not allowed to change anything here
1878 * so we don't permanently release the long hold or regular hold here. We want
1879 * to do this only when syncing to avoid the dataset unexpectedly going away
1880 * when we release the long hold.
1883 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
1887 if (!dmu_tx_is_syncing(tx
))
1890 if (owner
!= NULL
) {
1891 VERIFY3P(ds
->ds_owner
, ==, owner
);
1892 dsl_dataset_long_rele(ds
, owner
);
1895 held
= dsl_dataset_long_held(ds
);
1898 dsl_dataset_long_hold(ds
, owner
);
1901 return (SET_ERROR(EBUSY
));
1906 typedef struct dsl_dataset_rollback_arg
{
1907 const char *ddra_fsname
;
1909 nvlist_t
*ddra_result
;
1910 } dsl_dataset_rollback_arg_t
;
1913 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
1915 dsl_dataset_rollback_arg_t
*ddra
= arg
;
1916 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1918 int64_t unused_refres_delta
;
1921 nvlist_t
*proprequest
, *bookmarks
;
1923 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
1927 /* must not be a snapshot */
1928 if (ds
->ds_is_snapshot
) {
1929 dsl_dataset_rele(ds
, FTAG
);
1930 return (SET_ERROR(EINVAL
));
1933 /* must have a most recent snapshot */
1934 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
1935 dsl_dataset_rele(ds
, FTAG
);
1936 return (SET_ERROR(EINVAL
));
1939 /* must not have any bookmarks after the most recent snapshot */
1940 proprequest
= fnvlist_alloc();
1941 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
1942 bookmarks
= fnvlist_alloc();
1943 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
1944 fnvlist_free(proprequest
);
1947 for (pair
= nvlist_next_nvpair(bookmarks
, NULL
);
1948 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
1950 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
1951 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
1952 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
1953 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
1954 fnvlist_free(bookmarks
);
1955 dsl_dataset_rele(ds
, FTAG
);
1956 return (SET_ERROR(EEXIST
));
1959 fnvlist_free(bookmarks
);
1961 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
1963 dsl_dataset_rele(ds
, FTAG
);
1968 * Check if the snap we are rolling back to uses more than
1971 if (ds
->ds_quota
!= 0 &&
1972 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
1973 dsl_dataset_rele(ds
, FTAG
);
1974 return (SET_ERROR(EDQUOT
));
1978 * When we do the clone swap, we will temporarily use more space
1979 * due to the refreservation (the head will no longer have any
1980 * unique space, so the entire amount of the refreservation will need
1981 * to be free). We will immediately destroy the clone, freeing
1982 * this space, but the freeing happens over many txg's.
1984 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
1985 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1987 if (unused_refres_delta
> 0 &&
1988 unused_refres_delta
>
1989 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
1990 dsl_dataset_rele(ds
, FTAG
);
1991 return (SET_ERROR(ENOSPC
));
1994 dsl_dataset_rele(ds
, FTAG
);
1999 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2001 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2002 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2003 dsl_dataset_t
*ds
, *clone
;
2005 char namebuf
[ZFS_MAXNAMELEN
];
2007 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2009 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2010 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2012 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2013 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, tx
);
2015 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2017 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2018 dsl_dataset_zero_zil(ds
, tx
);
2020 dsl_destroy_head_sync_impl(clone
, tx
);
2022 dsl_dataset_rele(clone
, FTAG
);
2023 dsl_dataset_rele(ds
, FTAG
);
2027 * Rolls back the given filesystem or volume to the most recent snapshot.
2028 * The name of the most recent snapshot will be returned under key "target"
2029 * in the result nvlist.
2032 * - The existing dataset MUST be owned by the specified owner at entry
2033 * - Upon return, dataset will still be held by the same owner, whether we
2036 * This mode is required any time the existing filesystem is mounted. See
2037 * notes above zfs_suspend_fs() for further details.
2040 dsl_dataset_rollback(const char *fsname
, void *owner
, nvlist_t
*result
)
2042 dsl_dataset_rollback_arg_t ddra
;
2044 ddra
.ddra_fsname
= fsname
;
2045 ddra
.ddra_owner
= owner
;
2046 ddra
.ddra_result
= result
;
2048 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2049 dsl_dataset_rollback_sync
, &ddra
, 1));
2052 struct promotenode
{
2057 typedef struct dsl_dataset_promote_arg
{
2058 const char *ddpa_clonename
;
2059 dsl_dataset_t
*ddpa_clone
;
2060 list_t shared_snaps
, origin_snaps
, clone_snaps
;
2061 dsl_dataset_t
*origin_origin
; /* origin of the origin */
2062 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
2065 } dsl_dataset_promote_arg_t
;
2067 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2068 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2070 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2073 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2075 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2076 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2078 struct promotenode
*snap
;
2079 dsl_dataset_t
*origin_ds
;
2084 err
= promote_hold(ddpa
, dp
, FTAG
);
2088 hds
= ddpa
->ddpa_clone
;
2090 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2091 promote_rele(ddpa
, FTAG
);
2092 return (SET_ERROR(EXDEV
));
2096 * Compute and check the amount of space to transfer. Since this is
2097 * so expensive, don't do the preliminary check.
2099 if (!dmu_tx_is_syncing(tx
)) {
2100 promote_rele(ddpa
, FTAG
);
2104 snap
= list_head(&ddpa
->shared_snaps
);
2105 origin_ds
= snap
->ds
;
2107 /* compute origin's new unique space */
2108 snap
= list_tail(&ddpa
->clone_snaps
);
2109 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2110 origin_ds
->ds_object
);
2111 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2112 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2113 &ddpa
->unique
, &unused
, &unused
);
2116 * Walk the snapshots that we are moving
2118 * Compute space to transfer. Consider the incremental changes
2119 * to used by each snapshot:
2120 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2121 * So each snapshot gave birth to:
2122 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2123 * So a sequence would look like:
2124 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2125 * Which simplifies to:
2126 * uN + kN + kN-1 + ... + k1 + k0
2127 * Note however, if we stop before we reach the ORIGIN we get:
2128 * uN + kN + kN-1 + ... + kM - uM-1
2131 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2132 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2133 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2134 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2135 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2136 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2137 dsl_dataset_t
*ds
= snap
->ds
;
2142 * If there are long holds, we won't be able to evict
2145 if (dsl_dataset_long_held(ds
)) {
2146 err
= SET_ERROR(EBUSY
);
2150 /* Check that the snapshot name does not conflict */
2151 VERIFY0(dsl_dataset_get_snapname(ds
));
2152 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2154 (void) strcpy(ddpa
->err_ds
, snap
->ds
->ds_snapname
);
2155 err
= SET_ERROR(EEXIST
);
2161 /* The very first snapshot does not have a deadlist */
2162 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2165 dsl_deadlist_space(&ds
->ds_deadlist
,
2166 &dlused
, &dlcomp
, &dluncomp
);
2167 ddpa
->used
+= dlused
;
2168 ddpa
->comp
+= dlcomp
;
2169 ddpa
->uncomp
+= dluncomp
;
2173 * If we are a clone of a clone then we never reached ORIGIN,
2174 * so we need to subtract out the clone origin's used space.
2176 if (ddpa
->origin_origin
) {
2178 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2180 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2182 dsl_dataset_phys(ddpa
->origin_origin
)->
2183 ds_uncompressed_bytes
;
2186 /* Check that there is enough space and limit headroom here */
2187 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2188 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2193 * Compute the amounts of space that will be used by snapshots
2194 * after the promotion (for both origin and clone). For each,
2195 * it is the amount of space that will be on all of their
2196 * deadlists (that was not born before their new origin).
2198 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2202 * Note, typically this will not be a clone of a clone,
2203 * so dd_origin_txg will be < TXG_INITIAL, so
2204 * these snaplist_space() -> dsl_deadlist_space_range()
2205 * calls will be fast because they do not have to
2206 * iterate over all bps.
2208 snap
= list_head(&ddpa
->origin_snaps
);
2209 err
= snaplist_space(&ddpa
->shared_snaps
,
2210 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2214 err
= snaplist_space(&ddpa
->clone_snaps
,
2215 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2218 ddpa
->cloneusedsnap
+= space
;
2220 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2221 DD_FLAG_USED_BREAKDOWN
) {
2222 err
= snaplist_space(&ddpa
->origin_snaps
,
2223 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2224 &ddpa
->originusedsnap
);
2230 promote_rele(ddpa
, FTAG
);
2235 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2237 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2238 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2240 struct promotenode
*snap
;
2241 dsl_dataset_t
*origin_ds
;
2242 dsl_dataset_t
*origin_head
;
2244 dsl_dir_t
*odd
= NULL
;
2245 uint64_t oldnext_obj
;
2248 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2249 hds
= ddpa
->ddpa_clone
;
2251 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2253 snap
= list_head(&ddpa
->shared_snaps
);
2254 origin_ds
= snap
->ds
;
2257 snap
= list_head(&ddpa
->origin_snaps
);
2258 origin_head
= snap
->ds
;
2261 * We need to explicitly open odd, since origin_ds's dd will be
2264 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2267 /* change origin's next snap */
2268 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2269 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2270 snap
= list_tail(&ddpa
->clone_snaps
);
2271 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2272 origin_ds
->ds_object
);
2273 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2275 /* change the origin's next clone */
2276 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2277 dsl_dataset_remove_from_next_clones(origin_ds
,
2278 snap
->ds
->ds_object
, tx
);
2279 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2280 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2285 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2286 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2287 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2288 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2289 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2290 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2291 origin_head
->ds_dir
->dd_origin_txg
=
2292 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2294 /* change dd_clone entries */
2295 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2296 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2297 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2298 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2299 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2300 hds
->ds_object
, tx
));
2302 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2303 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2304 origin_head
->ds_object
, tx
));
2305 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2306 dsl_dir_phys(dd
)->dd_clones
=
2307 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2308 DMU_OT_NONE
, 0, tx
);
2310 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2311 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2314 /* move snapshots to this dir */
2315 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2316 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2317 dsl_dataset_t
*ds
= snap
->ds
;
2320 * Property callbacks are registered to a particular
2321 * dsl_dir. Since ours is changing, evict the objset
2322 * so that they will be unregistered from the old dsl_dir.
2324 if (ds
->ds_objset
) {
2325 dmu_objset_evict(ds
->ds_objset
);
2326 ds
->ds_objset
= NULL
;
2329 /* move snap name entry */
2330 VERIFY0(dsl_dataset_get_snapname(ds
));
2331 VERIFY0(dsl_dataset_snap_remove(origin_head
,
2332 ds
->ds_snapname
, tx
, B_TRUE
));
2333 VERIFY0(zap_add(dp
->dp_meta_objset
,
2334 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
2335 8, 1, &ds
->ds_object
, tx
));
2336 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
2337 DD_FIELD_SNAPSHOT_COUNT
, tx
);
2339 /* change containing dsl_dir */
2340 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2341 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
2342 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
2343 ASSERT3P(ds
->ds_dir
, ==, odd
);
2344 dsl_dir_rele(ds
->ds_dir
, ds
);
2345 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
2346 NULL
, ds
, &ds
->ds_dir
));
2348 /* move any clone references */
2349 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
2350 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2354 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
2355 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2356 zap_cursor_retrieve(&zc
, &za
) == 0;
2357 zap_cursor_advance(&zc
)) {
2358 dsl_dataset_t
*cnds
;
2361 if (za
.za_first_integer
== oldnext_obj
) {
2363 * We've already moved the
2364 * origin's reference.
2369 VERIFY0(dsl_dataset_hold_obj(dp
,
2370 za
.za_first_integer
, FTAG
, &cnds
));
2371 o
= dsl_dir_phys(cnds
->ds_dir
)->
2372 dd_head_dataset_obj
;
2374 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2375 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
2376 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2377 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
2378 dsl_dataset_rele(cnds
, FTAG
);
2380 zap_cursor_fini(&zc
);
2383 ASSERT(!dsl_prop_hascb(ds
));
2387 * Change space accounting.
2388 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2389 * both be valid, or both be 0 (resulting in delta == 0). This
2390 * is true for each of {clone,origin} independently.
2393 delta
= ddpa
->cloneusedsnap
-
2394 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
2395 ASSERT3S(delta
, >=, 0);
2396 ASSERT3U(ddpa
->used
, >=, delta
);
2397 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2398 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
2399 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
2401 delta
= ddpa
->originusedsnap
-
2402 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
2403 ASSERT3S(delta
, <=, 0);
2404 ASSERT3U(ddpa
->used
, >=, -delta
);
2405 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2406 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
2407 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
2409 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
2411 /* log history record */
2412 spa_history_log_internal_ds(hds
, "promote", tx
, "");
2414 dsl_dir_rele(odd
, FTAG
);
2415 promote_rele(ddpa
, FTAG
);
2419 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2420 * (exclusive) and last_obj (inclusive). The list will be in reverse
2421 * order (last_obj will be the list_head()). If first_obj == 0, do all
2422 * snapshots back to this dataset's origin.
2425 snaplist_make(dsl_pool_t
*dp
,
2426 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
2428 uint64_t obj
= last_obj
;
2430 list_create(l
, sizeof (struct promotenode
),
2431 offsetof(struct promotenode
, link
));
2433 while (obj
!= first_obj
) {
2435 struct promotenode
*snap
;
2438 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
2439 ASSERT(err
!= ENOENT
);
2444 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
2446 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
2448 list_insert_tail(l
, snap
);
2449 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
2456 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
2458 struct promotenode
*snap
;
2461 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
2462 uint64_t used
, comp
, uncomp
;
2463 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2464 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
2471 snaplist_destroy(list_t
*l
, void *tag
)
2473 struct promotenode
*snap
;
2475 if (l
== NULL
|| !list_link_active(&l
->list_head
))
2478 while ((snap
= list_tail(l
)) != NULL
) {
2479 list_remove(l
, snap
);
2480 dsl_dataset_rele(snap
->ds
, tag
);
2481 kmem_free(snap
, sizeof (*snap
));
2487 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
2491 struct promotenode
*snap
;
2493 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
2497 dd
= ddpa
->ddpa_clone
->ds_dir
;
2499 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
2500 !dsl_dir_is_clone(dd
)) {
2501 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2502 return (SET_ERROR(EINVAL
));
2505 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
2506 &ddpa
->shared_snaps
, tag
);
2510 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
2511 &ddpa
->clone_snaps
, tag
);
2515 snap
= list_head(&ddpa
->shared_snaps
);
2516 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
2517 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
2518 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
2519 &ddpa
->origin_snaps
, tag
);
2523 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
2524 error
= dsl_dataset_hold_obj(dp
,
2525 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
2526 tag
, &ddpa
->origin_origin
);
2532 promote_rele(ddpa
, tag
);
2537 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
2539 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
2540 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
2541 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
2542 if (ddpa
->origin_origin
!= NULL
)
2543 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
2544 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2550 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2551 * in with the name. (It must be at least MAXNAMELEN bytes long.)
2554 dsl_dataset_promote(const char *name
, char *conflsnap
)
2556 dsl_dataset_promote_arg_t ddpa
= { 0 };
2562 * We will modify space proportional to the number of
2563 * snapshots. Compute numsnaps.
2565 error
= dmu_objset_hold(name
, FTAG
, &os
);
2568 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
2569 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
2571 dmu_objset_rele(os
, FTAG
);
2575 ddpa
.ddpa_clonename
= name
;
2576 ddpa
.err_ds
= conflsnap
;
2579 return (dsl_sync_task(name
, dsl_dataset_promote_check
,
2580 dsl_dataset_promote_sync
, &ddpa
, 2 + numsnaps
));
2584 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
2585 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
2587 int64_t unused_refres_delta
;
2589 /* they should both be heads */
2590 if (clone
->ds_is_snapshot
||
2591 origin_head
->ds_is_snapshot
)
2592 return (SET_ERROR(EINVAL
));
2594 /* if we are not forcing, the branch point should be just before them */
2595 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
2596 return (SET_ERROR(EINVAL
));
2598 /* clone should be the clone (unless they are unrelated) */
2599 if (clone
->ds_prev
!= NULL
&&
2600 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
2601 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
2602 return (SET_ERROR(EINVAL
));
2604 /* the clone should be a child of the origin */
2605 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
2606 return (SET_ERROR(EINVAL
));
2608 /* origin_head shouldn't be modified unless 'force' */
2610 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
2611 return (SET_ERROR(ETXTBSY
));
2613 /* origin_head should have no long holds (e.g. is not mounted) */
2614 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
2615 return (SET_ERROR(EBUSY
));
2617 /* check amount of any unconsumed refreservation */
2618 unused_refres_delta
=
2619 (int64_t)MIN(origin_head
->ds_reserved
,
2620 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2621 (int64_t)MIN(origin_head
->ds_reserved
,
2622 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2624 if (unused_refres_delta
> 0 &&
2625 unused_refres_delta
>
2626 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
2627 return (SET_ERROR(ENOSPC
));
2629 /* clone can't be over the head's refquota */
2630 if (origin_head
->ds_quota
!= 0 &&
2631 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
2632 origin_head
->ds_quota
)
2633 return (SET_ERROR(EDQUOT
));
2639 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
2640 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
2642 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2643 int64_t unused_refres_delta
;
2645 ASSERT(clone
->ds_reserved
== 0);
2646 ASSERT(origin_head
->ds_quota
== 0 ||
2647 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
);
2648 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
2650 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
2651 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
2653 if (clone
->ds_objset
!= NULL
) {
2654 dmu_objset_evict(clone
->ds_objset
);
2655 clone
->ds_objset
= NULL
;
2658 if (origin_head
->ds_objset
!= NULL
) {
2659 dmu_objset_evict(origin_head
->ds_objset
);
2660 origin_head
->ds_objset
= NULL
;
2663 unused_refres_delta
=
2664 (int64_t)MIN(origin_head
->ds_reserved
,
2665 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2666 (int64_t)MIN(origin_head
->ds_reserved
,
2667 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2670 * Reset origin's unique bytes, if it exists.
2672 if (clone
->ds_prev
) {
2673 dsl_dataset_t
*origin
= clone
->ds_prev
;
2674 uint64_t comp
, uncomp
;
2676 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
2677 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2678 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
2679 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
2685 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
2686 dsl_dataset_phys(origin_head
)->ds_bp
=
2687 dsl_dataset_phys(clone
)->ds_bp
;
2688 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
2691 /* set dd_*_bytes */
2693 int64_t dused
, dcomp
, duncomp
;
2694 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
2695 uint64_t odl_used
, odl_comp
, odl_uncomp
;
2697 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
2698 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
2700 dsl_deadlist_space(&clone
->ds_deadlist
,
2701 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2702 dsl_deadlist_space(&origin_head
->ds_deadlist
,
2703 &odl_used
, &odl_comp
, &odl_uncomp
);
2705 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
2707 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
2709 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
2711 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
2713 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
2715 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
2718 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
2719 dused
, dcomp
, duncomp
, tx
);
2720 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
2721 -dused
, -dcomp
, -duncomp
, tx
);
2724 * The difference in the space used by snapshots is the
2725 * difference in snapshot space due to the head's
2726 * deadlist (since that's the only thing that's
2727 * changing that affects the snapused).
2729 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2730 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2731 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2732 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
2733 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2734 &odl_used
, &odl_comp
, &odl_uncomp
);
2735 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
2736 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
2739 /* swap ds_*_bytes */
2740 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
2741 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
2742 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
2743 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
2744 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
2745 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
2746 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
2747 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2749 /* apply any parent delta for change in unconsumed refreservation */
2750 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
2751 unused_refres_delta
, 0, 0, tx
);
2756 dsl_deadlist_close(&clone
->ds_deadlist
);
2757 dsl_deadlist_close(&origin_head
->ds_deadlist
);
2758 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
2759 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2760 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
2761 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2762 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
2763 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
2765 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
2767 spa_history_log_internal_ds(clone
, "clone swap", tx
,
2768 "parent=%s", origin_head
->ds_dir
->dd_myname
);
2772 * Given a pool name and a dataset object number in that pool,
2773 * return the name of that dataset.
2776 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
2782 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
2786 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
2788 dsl_dataset_name(ds
, buf
);
2789 dsl_dataset_rele(ds
, FTAG
);
2791 dsl_pool_rele(dp
, FTAG
);
2797 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
2798 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
2802 ASSERT3S(asize
, >, 0);
2805 * *ref_rsrv is the portion of asize that will come from any
2806 * unconsumed refreservation space.
2810 mutex_enter(&ds
->ds_lock
);
2812 * Make a space adjustment for reserved bytes.
2814 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2816 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2818 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2820 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
2823 if (!check_quota
|| ds
->ds_quota
== 0) {
2824 mutex_exit(&ds
->ds_lock
);
2828 * If they are requesting more space, and our current estimate
2829 * is over quota, they get to try again unless the actual
2830 * on-disk is over quota and there are no pending changes (which
2831 * may free up space for us).
2833 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
2836 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
2837 error
= SET_ERROR(ERESTART
);
2839 error
= SET_ERROR(EDQUOT
);
2841 mutex_exit(&ds
->ds_lock
);
2846 typedef struct dsl_dataset_set_qr_arg
{
2847 const char *ddsqra_name
;
2848 zprop_source_t ddsqra_source
;
2849 uint64_t ddsqra_value
;
2850 } dsl_dataset_set_qr_arg_t
;
2855 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
2857 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2858 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2863 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
2864 return (SET_ERROR(ENOTSUP
));
2866 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
2870 if (ds
->ds_is_snapshot
) {
2871 dsl_dataset_rele(ds
, FTAG
);
2872 return (SET_ERROR(EINVAL
));
2875 error
= dsl_prop_predict(ds
->ds_dir
,
2876 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
2877 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
2879 dsl_dataset_rele(ds
, FTAG
);
2884 dsl_dataset_rele(ds
, FTAG
);
2888 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
2889 newval
< ds
->ds_reserved
) {
2890 dsl_dataset_rele(ds
, FTAG
);
2891 return (SET_ERROR(ENOSPC
));
2894 dsl_dataset_rele(ds
, FTAG
);
2899 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
2901 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2902 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2906 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
2908 dsl_prop_set_sync_impl(ds
,
2909 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
2910 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
2911 &ddsqra
->ddsqra_value
, tx
);
2913 VERIFY0(dsl_prop_get_int_ds(ds
,
2914 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
2916 if (ds
->ds_quota
!= newval
) {
2917 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2918 ds
->ds_quota
= newval
;
2920 dsl_dataset_rele(ds
, FTAG
);
2924 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
2927 dsl_dataset_set_qr_arg_t ddsqra
;
2929 ddsqra
.ddsqra_name
= dsname
;
2930 ddsqra
.ddsqra_source
= source
;
2931 ddsqra
.ddsqra_value
= refquota
;
2933 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
2934 dsl_dataset_set_refquota_sync
, &ddsqra
, 0));
2938 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
2940 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2941 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2944 uint64_t newval
, unique
;
2946 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
2947 return (SET_ERROR(ENOTSUP
));
2949 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
2953 if (ds
->ds_is_snapshot
) {
2954 dsl_dataset_rele(ds
, FTAG
);
2955 return (SET_ERROR(EINVAL
));
2958 error
= dsl_prop_predict(ds
->ds_dir
,
2959 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
2960 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
2962 dsl_dataset_rele(ds
, FTAG
);
2967 * If we are doing the preliminary check in open context, the
2968 * space estimates may be inaccurate.
2970 if (!dmu_tx_is_syncing(tx
)) {
2971 dsl_dataset_rele(ds
, FTAG
);
2975 mutex_enter(&ds
->ds_lock
);
2976 if (!DS_UNIQUE_IS_ACCURATE(ds
))
2977 dsl_dataset_recalc_head_uniq(ds
);
2978 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
2979 mutex_exit(&ds
->ds_lock
);
2981 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
2982 uint64_t delta
= MAX(unique
, newval
) -
2983 MAX(unique
, ds
->ds_reserved
);
2986 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
2987 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
2988 dsl_dataset_rele(ds
, FTAG
);
2989 return (SET_ERROR(ENOSPC
));
2993 dsl_dataset_rele(ds
, FTAG
);
2998 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
2999 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3005 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3006 source
, sizeof (value
), 1, &value
, tx
);
3008 VERIFY0(dsl_prop_get_int_ds(ds
,
3009 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3011 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3012 mutex_enter(&ds
->ds_dir
->dd_lock
);
3013 mutex_enter(&ds
->ds_lock
);
3014 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3015 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3016 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3017 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3018 ds
->ds_reserved
= newval
;
3019 mutex_exit(&ds
->ds_lock
);
3021 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3022 mutex_exit(&ds
->ds_dir
->dd_lock
);
3026 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3028 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3029 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3032 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3033 dsl_dataset_set_refreservation_sync_impl(ds
,
3034 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3035 dsl_dataset_rele(ds
, FTAG
);
3039 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3040 uint64_t refreservation
)
3042 dsl_dataset_set_qr_arg_t ddsqra
;
3044 ddsqra
.ddsqra_name
= dsname
;
3045 ddsqra
.ddsqra_source
= source
;
3046 ddsqra
.ddsqra_value
= refreservation
;
3048 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3049 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0));
3053 * Return (in *usedp) the amount of space written in new that is not
3054 * present in oldsnap. New may be a snapshot or the head. Old must be
3055 * a snapshot before new, in new's filesystem (or its origin). If not then
3056 * fail and return EINVAL.
3058 * The written space is calculated by considering two components: First, we
3059 * ignore any freed space, and calculate the written as new's used space
3060 * minus old's used space. Next, we add in the amount of space that was freed
3061 * between the two snapshots, thus reducing new's used space relative to old's.
3062 * Specifically, this is the space that was born before old->ds_creation_txg,
3063 * and freed before new (ie. on new's deadlist or a previous deadlist).
3065 * space freed [---------------------]
3066 * snapshots ---O-------O--------O-------O------
3070 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3071 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3075 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3077 ASSERT(dsl_pool_config_held(dp
));
3080 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3081 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3084 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3085 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3088 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3089 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3091 snapobj
= new->ds_object
;
3092 while (snapobj
!= oldsnap
->ds_object
) {
3093 dsl_dataset_t
*snap
;
3094 uint64_t used
, comp
, uncomp
;
3096 if (snapobj
== new->ds_object
) {
3099 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3104 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3105 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3107 * The blocks in the deadlist can not be born after
3108 * ds_prev_snap_txg, so get the whole deadlist space,
3109 * which is more efficient (especially for old-format
3110 * deadlists). Unfortunately the deadlist code
3111 * doesn't have enough information to make this
3112 * optimization itself.
3114 dsl_deadlist_space(&snap
->ds_deadlist
,
3115 &used
, &comp
, &uncomp
);
3117 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3118 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3119 &used
, &comp
, &uncomp
);
3126 * If we get to the beginning of the chain of snapshots
3127 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3128 * was not a snapshot of/before new.
3130 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3132 dsl_dataset_rele(snap
, FTAG
);
3134 err
= SET_ERROR(EINVAL
);
3143 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3144 * lastsnap, and all snapshots in between are deleted.
3146 * blocks that would be freed [---------------------------]
3147 * snapshots ---O-------O--------O-------O--------O
3148 * firstsnap lastsnap
3150 * This is the set of blocks that were born after the snap before firstsnap,
3151 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3152 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3153 * We calculate this by iterating over the relevant deadlists (from the snap
3154 * after lastsnap, backward to the snap after firstsnap), summing up the
3155 * space on the deadlist that was born after the snap before firstsnap.
3158 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3159 dsl_dataset_t
*lastsnap
,
3160 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3164 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3166 ASSERT(firstsnap
->ds_is_snapshot
);
3167 ASSERT(lastsnap
->ds_is_snapshot
);
3170 * Check that the snapshots are in the same dsl_dir, and firstsnap
3171 * is before lastsnap.
3173 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3174 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3175 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3176 return (SET_ERROR(EINVAL
));
3178 *usedp
= *compp
= *uncompp
= 0;
3180 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3181 while (snapobj
!= firstsnap
->ds_object
) {
3183 uint64_t used
, comp
, uncomp
;
3185 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3189 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3190 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3191 &used
, &comp
, &uncomp
);
3196 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3197 ASSERT3U(snapobj
, !=, 0);
3198 dsl_dataset_rele(ds
, FTAG
);
3204 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3205 * For example, they could both be snapshots of the same filesystem, and
3206 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3207 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3208 * filesystem. Or 'earlier' could be the origin's origin.
3210 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3213 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3214 uint64_t earlier_txg
)
3216 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3219 dsl_dataset_t
*origin
;
3221 ASSERT(dsl_pool_config_held(dp
));
3222 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3224 if (earlier_txg
== 0)
3225 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3227 if (later
->ds_is_snapshot
&&
3228 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3231 if (later
->ds_dir
== earlier
->ds_dir
)
3233 if (!dsl_dir_is_clone(later
->ds_dir
))
3236 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3238 error
= dsl_dataset_hold_obj(dp
,
3239 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3242 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3243 dsl_dataset_rele(origin
, FTAG
);
3249 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
3251 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
3252 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
3255 #if defined(_KERNEL) && defined(HAVE_SPL)
3256 EXPORT_SYMBOL(dsl_dataset_hold
);
3257 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
3258 EXPORT_SYMBOL(dsl_dataset_own
);
3259 EXPORT_SYMBOL(dsl_dataset_own_obj
);
3260 EXPORT_SYMBOL(dsl_dataset_name
);
3261 EXPORT_SYMBOL(dsl_dataset_rele
);
3262 EXPORT_SYMBOL(dsl_dataset_disown
);
3263 EXPORT_SYMBOL(dsl_dataset_tryown
);
3264 EXPORT_SYMBOL(dsl_dataset_create_sync
);
3265 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
3266 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
3267 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
3268 EXPORT_SYMBOL(dsl_dataset_promote
);
3269 EXPORT_SYMBOL(dsl_dataset_user_hold
);
3270 EXPORT_SYMBOL(dsl_dataset_user_release
);
3271 EXPORT_SYMBOL(dsl_dataset_get_holds
);
3272 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
3273 EXPORT_SYMBOL(dsl_dataset_set_blkptr
);
3274 EXPORT_SYMBOL(dsl_dataset_get_spa
);
3275 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
3276 EXPORT_SYMBOL(dsl_dataset_space_written
);
3277 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
3278 EXPORT_SYMBOL(dsl_dataset_sync
);
3279 EXPORT_SYMBOL(dsl_dataset_block_born
);
3280 EXPORT_SYMBOL(dsl_dataset_block_kill
);
3281 EXPORT_SYMBOL(dsl_dataset_block_freeable
);
3282 EXPORT_SYMBOL(dsl_dataset_prev_snap_txg
);
3283 EXPORT_SYMBOL(dsl_dataset_dirty
);
3284 EXPORT_SYMBOL(dsl_dataset_stats
);
3285 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
3286 EXPORT_SYMBOL(dsl_dataset_space
);
3287 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
3288 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
3289 EXPORT_SYMBOL(dsl_dataset_check_quota
);
3290 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
3291 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);