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_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
357 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
358 boolean_t result
= B_FALSE
;
360 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
361 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
363 if (ds
== dmu_buf_get_user(dbuf
))
366 dmu_buf_rele(dbuf
, tag
);
373 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
376 objset_t
*mos
= dp
->dp_meta_objset
;
380 dmu_object_info_t doi
;
382 ASSERT(dsl_pool_config_held(dp
));
384 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
388 /* Make sure dsobj has the correct object type. */
389 dmu_object_info_from_db(dbuf
, &doi
);
390 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
391 dmu_buf_rele(dbuf
, tag
);
392 return (SET_ERROR(EINVAL
));
395 ds
= dmu_buf_get_user(dbuf
);
397 dsl_dataset_t
*winner
= NULL
;
399 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
401 ds
->ds_object
= dsobj
;
402 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
403 list_link_init(&ds
->ds_synced_link
);
405 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
406 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
407 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
408 refcount_create(&ds
->ds_longholds
);
410 bplist_create(&ds
->ds_pending_deadlist
);
411 dsl_deadlist_open(&ds
->ds_deadlist
,
412 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
414 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
415 offsetof(dmu_sendarg_t
, dsa_link
));
418 err
= dsl_dir_hold_obj(dp
,
419 dsl_dataset_phys(ds
)->ds_dir_obj
, NULL
, ds
,
423 mutex_destroy(&ds
->ds_lock
);
424 mutex_destroy(&ds
->ds_opening_lock
);
425 mutex_destroy(&ds
->ds_sendstream_lock
);
426 refcount_destroy(&ds
->ds_longholds
);
427 bplist_destroy(&ds
->ds_pending_deadlist
);
428 dsl_deadlist_close(&ds
->ds_deadlist
);
429 kmem_free(ds
, sizeof (dsl_dataset_t
));
430 dmu_buf_rele(dbuf
, tag
);
434 if (!ds
->ds_is_snapshot
) {
435 ds
->ds_snapname
[0] = '\0';
436 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
437 err
= dsl_dataset_hold_obj(dp
,
438 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
441 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
442 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
443 DS_FIELD_BOOKMARK_NAMES
,
444 sizeof (ds
->ds_bookmarks
), 1,
446 if (zaperr
!= ENOENT
)
450 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
451 err
= dsl_dataset_get_snapname(ds
);
453 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
455 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
456 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
461 if (err
== 0 && !ds
->ds_is_snapshot
) {
462 err
= dsl_prop_get_int_ds(ds
,
463 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
466 err
= dsl_prop_get_int_ds(ds
,
467 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
471 ds
->ds_reserved
= ds
->ds_quota
= 0;
474 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict
, &ds
->ds_dbuf
);
476 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
478 if (err
!= 0 || winner
!= NULL
) {
479 bplist_destroy(&ds
->ds_pending_deadlist
);
480 dsl_deadlist_close(&ds
->ds_deadlist
);
482 dsl_dataset_rele(ds
->ds_prev
, ds
);
483 dsl_dir_rele(ds
->ds_dir
, ds
);
484 mutex_destroy(&ds
->ds_lock
);
485 mutex_destroy(&ds
->ds_opening_lock
);
486 mutex_destroy(&ds
->ds_sendstream_lock
);
487 refcount_destroy(&ds
->ds_longholds
);
488 kmem_free(ds
, sizeof (dsl_dataset_t
));
490 dmu_buf_rele(dbuf
, tag
);
496 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
499 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
500 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
501 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
502 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
503 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
509 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
,
510 void *tag
, dsl_dataset_t
**dsp
)
513 const char *snapname
;
517 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
521 ASSERT(dsl_pool_config_held(dp
));
522 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
524 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, dsp
);
526 err
= SET_ERROR(ENOENT
);
528 /* we may be looking for a snapshot */
529 if (err
== 0 && snapname
!= NULL
) {
532 if (*snapname
++ != '@') {
533 dsl_dataset_rele(*dsp
, tag
);
534 dsl_dir_rele(dd
, FTAG
);
535 return (SET_ERROR(ENOENT
));
538 dprintf("looking for snapshot '%s'\n", snapname
);
539 err
= dsl_dataset_snap_lookup(*dsp
, snapname
, &obj
);
541 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
542 dsl_dataset_rele(*dsp
, tag
);
545 mutex_enter(&ds
->ds_lock
);
546 if (ds
->ds_snapname
[0] == 0)
547 (void) strlcpy(ds
->ds_snapname
, snapname
,
548 sizeof (ds
->ds_snapname
));
549 mutex_exit(&ds
->ds_lock
);
554 dsl_dir_rele(dd
, FTAG
);
559 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
,
560 void *tag
, dsl_dataset_t
**dsp
)
562 int err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
565 if (!dsl_dataset_tryown(*dsp
, tag
)) {
566 dsl_dataset_rele(*dsp
, tag
);
568 return (SET_ERROR(EBUSY
));
574 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
,
575 void *tag
, dsl_dataset_t
**dsp
)
577 int err
= dsl_dataset_hold(dp
, name
, tag
, dsp
);
580 if (!dsl_dataset_tryown(*dsp
, tag
)) {
581 dsl_dataset_rele(*dsp
, tag
);
582 return (SET_ERROR(EBUSY
));
588 * See the comment above dsl_pool_hold() for details. In summary, a long
589 * hold is used to prevent destruction of a dataset while the pool hold
590 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
592 * The dataset and pool must be held when this function is called. After it
593 * is called, the pool hold may be released while the dataset is still held
597 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
599 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
600 (void) refcount_add(&ds
->ds_longholds
, tag
);
604 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
606 (void) refcount_remove(&ds
->ds_longholds
, tag
);
609 /* Return B_TRUE if there are any long holds on this dataset. */
611 dsl_dataset_long_held(dsl_dataset_t
*ds
)
613 return (!refcount_is_zero(&ds
->ds_longholds
));
617 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
620 (void) strcpy(name
, "mos");
622 dsl_dir_name(ds
->ds_dir
, name
);
623 VERIFY0(dsl_dataset_get_snapname(ds
));
624 if (ds
->ds_snapname
[0]) {
625 (void) strcat(name
, "@");
627 * We use a "recursive" mutex so that we
628 * can call dprintf_ds() with ds_lock held.
630 if (!MUTEX_HELD(&ds
->ds_lock
)) {
631 mutex_enter(&ds
->ds_lock
);
632 (void) strcat(name
, ds
->ds_snapname
);
633 mutex_exit(&ds
->ds_lock
);
635 (void) strcat(name
, ds
->ds_snapname
);
642 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
644 dmu_buf_rele(ds
->ds_dbuf
, tag
);
648 dsl_dataset_disown(dsl_dataset_t
*ds
, void *tag
)
650 ASSERT3P(ds
->ds_owner
, ==, tag
);
651 ASSERT(ds
->ds_dbuf
!= NULL
);
653 mutex_enter(&ds
->ds_lock
);
655 mutex_exit(&ds
->ds_lock
);
656 dsl_dataset_long_rele(ds
, tag
);
657 dsl_dataset_rele(ds
, tag
);
661 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
663 boolean_t gotit
= FALSE
;
665 mutex_enter(&ds
->ds_lock
);
666 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
668 dsl_dataset_long_hold(ds
, tag
);
671 mutex_exit(&ds
->ds_lock
);
676 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
677 uint64_t flags
, dmu_tx_t
*tx
)
679 dsl_pool_t
*dp
= dd
->dd_pool
;
681 dsl_dataset_phys_t
*dsphys
;
683 objset_t
*mos
= dp
->dp_meta_objset
;
686 origin
= dp
->dp_origin_snap
;
688 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
689 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
690 ASSERT(dmu_tx_is_syncing(tx
));
691 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
693 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
694 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
695 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
696 dmu_buf_will_dirty(dbuf
, tx
);
697 dsphys
= dbuf
->db_data
;
698 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
699 dsphys
->ds_dir_obj
= dd
->dd_object
;
700 dsphys
->ds_flags
= flags
;
701 dsphys
->ds_fsid_guid
= unique_create();
702 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
703 sizeof (dsphys
->ds_guid
));
704 dsphys
->ds_snapnames_zapobj
=
705 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
707 dsphys
->ds_creation_time
= gethrestime_sec();
708 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
710 if (origin
== NULL
) {
711 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
713 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
715 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
716 dsphys
->ds_prev_snap_txg
=
717 dsl_dataset_phys(origin
)->ds_creation_txg
;
718 dsphys
->ds_referenced_bytes
=
719 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
720 dsphys
->ds_compressed_bytes
=
721 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
722 dsphys
->ds_uncompressed_bytes
=
723 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
724 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
727 * Inherit flags that describe the dataset's contents
728 * (INCONSISTENT) or properties (Case Insensitive).
730 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
731 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
733 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
734 dsl_dataset_phys(origin
)->ds_num_children
++;
736 VERIFY0(dsl_dataset_hold_obj(dp
,
737 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
739 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
740 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
741 dsl_dataset_rele(ohds
, FTAG
);
743 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
744 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
745 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
747 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
749 VERIFY0(zap_add_int(mos
,
750 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
754 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
755 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
756 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
757 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
758 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
759 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
761 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
763 VERIFY0(zap_add_int(mos
,
764 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
769 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
770 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
772 dmu_buf_rele(dbuf
, FTAG
);
774 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
775 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
781 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
785 VERIFY0(dmu_objset_from_ds(ds
, &os
));
786 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
787 dsl_dataset_dirty(ds
, tx
);
791 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
792 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
, dmu_tx_t
*tx
)
794 dsl_pool_t
*dp
= pdd
->dd_pool
;
795 uint64_t dsobj
, ddobj
;
798 ASSERT(dmu_tx_is_syncing(tx
));
799 ASSERT(lastname
[0] != '@');
801 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
802 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
804 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
,
805 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
807 dsl_deleg_set_create_perms(dd
, tx
, cr
);
810 * Since we're creating a new node we know it's a leaf, so we can
811 * initialize the counts if the limit feature is active.
813 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
815 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
817 dsl_dir_zapify(dd
, tx
);
818 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
819 sizeof (cnt
), 1, &cnt
, tx
));
820 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
821 sizeof (cnt
), 1, &cnt
, tx
));
824 dsl_dir_rele(dd
, FTAG
);
827 * If we are creating a clone, make sure we zero out any stale
828 * data from the origin snapshots zil header.
830 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
833 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
834 dsl_dataset_zero_zil(ds
, tx
);
835 dsl_dataset_rele(ds
, FTAG
);
842 * The unique space in the head dataset can be calculated by subtracting
843 * the space used in the most recent snapshot, that is still being used
844 * in this file system, from the space currently in use. To figure out
845 * the space in the most recent snapshot still in use, we need to take
846 * the total space used in the snapshot and subtract out the space that
847 * has been freed up since the snapshot was taken.
850 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
853 uint64_t dlused
, dlcomp
, dluncomp
;
855 ASSERT(!ds
->ds_is_snapshot
);
857 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
858 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
862 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
864 ASSERT3U(dlused
, <=, mrs_used
);
865 dsl_dataset_phys(ds
)->ds_unique_bytes
=
866 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
868 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
869 SPA_VERSION_UNIQUE_ACCURATE
)
870 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
874 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
877 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
879 ASSERTV(uint64_t count
);
881 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
882 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
885 * The err should not be ENOENT, but a bug in a previous version
886 * of the code could cause upgrade_clones_cb() to not set
887 * ds_next_snap_obj when it should, leading to a missing entry.
888 * If we knew that the pool was created after
889 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
890 * ENOENT. However, at least we can check that we don't have
891 * too many entries in the next_clones_obj even after failing to
896 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
898 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
903 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
905 return (&dsl_dataset_phys(ds
)->ds_bp
);
909 dsl_dataset_set_blkptr(dsl_dataset_t
*ds
, blkptr_t
*bp
, dmu_tx_t
*tx
)
911 ASSERT(dmu_tx_is_syncing(tx
));
912 /* If it's the meta-objset, set dp_meta_rootbp */
914 tx
->tx_pool
->dp_meta_rootbp
= *bp
;
916 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
917 dsl_dataset_phys(ds
)->ds_bp
= *bp
;
922 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
924 return (ds
->ds_dir
->dd_pool
->dp_spa
);
928 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
932 if (ds
== NULL
) /* this is the meta-objset */
935 ASSERT(ds
->ds_objset
!= NULL
);
937 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
938 panic("dirtying snapshot!");
940 dp
= ds
->ds_dir
->dd_pool
;
942 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
943 /* up the hold count until we can be written out */
944 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
949 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
953 for (t
= 0; t
< TXG_SIZE
; t
++) {
954 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
962 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
966 if (!dmu_tx_is_syncing(tx
))
970 * If there's an fs-only reservation, any blocks that might become
971 * owned by the snapshot dataset must be accommodated by space
972 * outside of the reservation.
974 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
975 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
976 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
977 return (SET_ERROR(ENOSPC
));
980 * Propagate any reserved space for this snapshot to other
981 * snapshot checks in this sync group.
984 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
989 typedef struct dsl_dataset_snapshot_arg
{
990 nvlist_t
*ddsa_snaps
;
991 nvlist_t
*ddsa_props
;
992 nvlist_t
*ddsa_errors
;
994 } dsl_dataset_snapshot_arg_t
;
997 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
998 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1003 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1005 if (!dmu_tx_is_syncing(tx
))
1009 * We don't allow multiple snapshots of the same txg. If there
1010 * is already one, try again.
1012 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1013 return (SET_ERROR(EAGAIN
));
1016 * Check for conflicting snapshot name.
1018 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1020 return (SET_ERROR(EEXIST
));
1021 if (error
!= ENOENT
)
1025 * We don't allow taking snapshots of inconsistent datasets, such as
1026 * those into which we are currently receiving. However, if we are
1027 * creating this snapshot as part of a receive, this check will be
1028 * executed atomically with respect to the completion of the receive
1029 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1030 * case we ignore this, knowing it will be fixed up for us shortly in
1031 * dmu_recv_end_sync().
1033 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1034 return (SET_ERROR(EBUSY
));
1037 * Skip the check for temporary snapshots or if we have already checked
1038 * the counts in dsl_dataset_snapshot_check. This means we really only
1039 * check the count here when we're receiving a stream.
1041 if (cnt
!= 0 && cr
!= NULL
) {
1042 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1043 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1048 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1056 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1058 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1059 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1064 * Pre-compute how many total new snapshots will be created for each
1065 * level in the tree and below. This is needed for validating the
1066 * snapshot limit when either taking a recursive snapshot or when
1067 * taking multiple snapshots.
1069 * The problem is that the counts are not actually adjusted when
1070 * we are checking, only when we finally sync. For a single snapshot,
1071 * this is easy, the count will increase by 1 at each node up the tree,
1072 * but its more complicated for the recursive/multiple snapshot case.
1074 * The dsl_fs_ss_limit_check function does recursively check the count
1075 * at each level up the tree but since it is validating each snapshot
1076 * independently we need to be sure that we are validating the complete
1077 * count for the entire set of snapshots. We do this by rolling up the
1078 * counts for each component of the name into an nvlist and then
1079 * checking each of those cases with the aggregated count.
1081 * This approach properly handles not only the recursive snapshot
1082 * case (where we get all of those on the ddsa_snaps list) but also
1083 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1084 * validate the limit on 'a' using a count of 2).
1086 * We validate the snapshot names in the third loop and only report
1089 if (dmu_tx_is_syncing(tx
)) {
1091 nvlist_t
*cnt_track
= NULL
;
1092 cnt_track
= fnvlist_alloc();
1094 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1096 /* Rollup aggregated counts into the cnt_track list */
1097 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1099 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1103 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1104 pdelim
= strchr(nm
, '@');
1110 if (nvlist_lookup_uint64(cnt_track
, nm
,
1112 /* update existing entry */
1113 fnvlist_add_uint64(cnt_track
, nm
,
1117 fnvlist_add_uint64(cnt_track
, nm
, 1);
1120 pdelim
= strrchr(nm
, '/');
1123 } while (pdelim
!= NULL
);
1126 kmem_free(nm
, MAXPATHLEN
);
1128 /* Check aggregated counts at each level */
1129 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1130 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1136 name
= nvpair_name(pair
);
1137 cnt
= fnvpair_value_uint64(pair
);
1140 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1142 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1143 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1145 dsl_dataset_rele(ds
, FTAG
);
1149 if (ddsa
->ddsa_errors
!= NULL
)
1150 fnvlist_add_int32(ddsa
->ddsa_errors
,
1153 /* only report one error for this check */
1157 nvlist_free(cnt_track
);
1160 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1161 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1165 char dsname
[MAXNAMELEN
];
1167 name
= nvpair_name(pair
);
1168 if (strlen(name
) >= MAXNAMELEN
)
1169 error
= SET_ERROR(ENAMETOOLONG
);
1171 atp
= strchr(name
, '@');
1173 error
= SET_ERROR(EINVAL
);
1175 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1178 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1180 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1181 error
= dsl_dataset_snapshot_check_impl(ds
,
1182 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1183 dsl_dataset_rele(ds
, FTAG
);
1187 if (ddsa
->ddsa_errors
!= NULL
) {
1188 fnvlist_add_int32(ddsa
->ddsa_errors
,
1199 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1202 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1204 dsl_dataset_phys_t
*dsphys
;
1205 uint64_t dsobj
, crtxg
;
1206 objset_t
*mos
= dp
->dp_meta_objset
;
1207 ASSERTV(static zil_header_t zero_zil
);
1208 ASSERTV(objset_t
*os
);
1210 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1213 * If we are on an old pool, the zil must not be active, in which
1214 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1216 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1217 dmu_objset_from_ds(ds
, &os
) != 0 ||
1218 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1219 sizeof (zero_zil
)) == 0);
1221 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1224 * The origin's ds_creation_txg has to be < TXG_INITIAL
1226 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1231 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1232 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1233 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1234 dmu_buf_will_dirty(dbuf
, tx
);
1235 dsphys
= dbuf
->db_data
;
1236 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1237 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1238 dsphys
->ds_fsid_guid
= unique_create();
1239 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1240 sizeof (dsphys
->ds_guid
));
1241 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1242 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1243 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1244 dsphys
->ds_num_children
= 1;
1245 dsphys
->ds_creation_time
= gethrestime_sec();
1246 dsphys
->ds_creation_txg
= crtxg
;
1247 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1248 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1249 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1250 dsphys
->ds_uncompressed_bytes
=
1251 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1252 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1253 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1254 dmu_buf_rele(dbuf
, FTAG
);
1256 ASSERT3U(ds
->ds_prev
!= 0, ==,
1257 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1259 uint64_t next_clones_obj
=
1260 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1261 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1263 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1264 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1266 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1267 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1268 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1269 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1270 } else if (next_clones_obj
!= 0) {
1271 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1272 dsphys
->ds_next_snap_obj
, tx
);
1273 VERIFY0(zap_add_int(mos
,
1274 next_clones_obj
, dsobj
, tx
));
1279 * If we have a reference-reservation on this dataset, we will
1280 * need to increase the amount of refreservation being charged
1281 * since our unique space is going to zero.
1283 if (ds
->ds_reserved
) {
1285 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1286 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1288 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1292 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1293 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1294 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1295 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1296 dsl_deadlist_close(&ds
->ds_deadlist
);
1297 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1298 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1299 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1300 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1302 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1303 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1304 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1305 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1306 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1307 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1309 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1310 snapname
, 8, 1, &dsobj
, tx
));
1313 dsl_dataset_rele(ds
->ds_prev
, ds
);
1314 VERIFY0(dsl_dataset_hold_obj(dp
,
1315 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1317 dsl_scan_ds_snapshotted(ds
, tx
);
1319 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1321 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1325 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1327 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1328 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1331 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1332 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1335 char dsname
[MAXNAMELEN
];
1337 name
= nvpair_name(pair
);
1338 atp
= strchr(name
, '@');
1339 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1340 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1342 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1343 if (ddsa
->ddsa_props
!= NULL
) {
1344 dsl_props_set_sync_impl(ds
->ds_prev
,
1345 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1347 dsl_dataset_rele(ds
, FTAG
);
1352 * The snapshots must all be in the same pool.
1353 * All-or-nothing: if there are any failures, nothing will be modified.
1356 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1358 dsl_dataset_snapshot_arg_t ddsa
;
1360 boolean_t needsuspend
;
1364 nvlist_t
*suspended
= NULL
;
1366 pair
= nvlist_next_nvpair(snaps
, NULL
);
1369 firstname
= nvpair_name(pair
);
1371 error
= spa_open(firstname
, &spa
, FTAG
);
1374 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1375 spa_close(spa
, FTAG
);
1378 suspended
= fnvlist_alloc();
1379 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1380 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1381 char fsname
[MAXNAMELEN
];
1382 char *snapname
= nvpair_name(pair
);
1386 atp
= strchr(snapname
, '@');
1388 error
= SET_ERROR(EINVAL
);
1391 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1393 error
= zil_suspend(fsname
, &cookie
);
1396 fnvlist_add_uint64(suspended
, fsname
,
1401 ddsa
.ddsa_snaps
= snaps
;
1402 ddsa
.ddsa_props
= props
;
1403 ddsa
.ddsa_errors
= errors
;
1404 ddsa
.ddsa_cr
= CRED();
1407 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1408 dsl_dataset_snapshot_sync
, &ddsa
,
1409 fnvlist_num_pairs(snaps
) * 3);
1412 if (suspended
!= NULL
) {
1413 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1414 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1415 zil_resume((void *)(uintptr_t)
1416 fnvpair_value_uint64(pair
));
1418 fnvlist_free(suspended
);
1423 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1424 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1425 char *snapname
= nvpair_name(pair
);
1426 zvol_create_minors(snapname
);
1434 typedef struct dsl_dataset_snapshot_tmp_arg
{
1435 const char *ddsta_fsname
;
1436 const char *ddsta_snapname
;
1437 minor_t ddsta_cleanup_minor
;
1438 const char *ddsta_htag
;
1439 } dsl_dataset_snapshot_tmp_arg_t
;
1442 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1444 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1445 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1449 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1453 /* NULL cred means no limit check for tmp snapshot */
1454 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1455 tx
, B_FALSE
, 0, NULL
);
1457 dsl_dataset_rele(ds
, FTAG
);
1461 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1462 dsl_dataset_rele(ds
, FTAG
);
1463 return (SET_ERROR(ENOTSUP
));
1465 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1468 dsl_dataset_rele(ds
, FTAG
);
1472 dsl_dataset_rele(ds
, FTAG
);
1477 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1479 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1480 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1483 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1485 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1486 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1487 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1488 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1490 dsl_dataset_rele(ds
, FTAG
);
1494 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1495 minor_t cleanup_minor
, const char *htag
)
1497 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1500 boolean_t needsuspend
;
1503 ddsta
.ddsta_fsname
= fsname
;
1504 ddsta
.ddsta_snapname
= snapname
;
1505 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1506 ddsta
.ddsta_htag
= htag
;
1508 error
= spa_open(fsname
, &spa
, FTAG
);
1511 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1512 spa_close(spa
, FTAG
);
1515 error
= zil_suspend(fsname
, &cookie
);
1520 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1521 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3);
1530 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1532 ASSERT(dmu_tx_is_syncing(tx
));
1533 ASSERT(ds
->ds_objset
!= NULL
);
1534 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1537 * in case we had to change ds_fsid_guid when we opened it,
1540 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1541 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1543 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1547 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1550 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1553 nvlist_t
*propval
= fnvlist_alloc();
1554 nvlist_t
*val
= fnvlist_alloc();
1556 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1559 * There may be missing entries in ds_next_clones_obj
1560 * due to a bug in a previous version of the code.
1561 * Only trust it if it has the right number of entries.
1563 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1564 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1567 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1)
1569 for (zap_cursor_init(&zc
, mos
,
1570 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1571 zap_cursor_retrieve(&zc
, &za
) == 0;
1572 zap_cursor_advance(&zc
)) {
1573 dsl_dataset_t
*clone
;
1574 char buf
[ZFS_MAXNAMELEN
];
1575 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1576 za
.za_first_integer
, FTAG
, &clone
));
1577 dsl_dir_name(clone
->ds_dir
, buf
);
1578 fnvlist_add_boolean(val
, buf
);
1579 dsl_dataset_rele(clone
, FTAG
);
1581 zap_cursor_fini(&zc
);
1582 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1583 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
), propval
);
1586 nvlist_free(propval
);
1590 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1592 uint64_t refd
, avail
, uobjs
, aobjs
, ratio
;
1593 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
1595 ASSERT(dsl_pool_config_held(dp
));
1597 ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1598 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1599 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1601 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
, ratio
);
1602 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
1603 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1605 if (ds
->ds_is_snapshot
) {
1606 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
, ratio
);
1607 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
1608 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1609 get_clones_stat(ds
, nv
);
1611 dsl_dir_stats(ds
->ds_dir
, nv
);
1614 dsl_dataset_space(ds
, &refd
, &avail
, &uobjs
, &aobjs
);
1615 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
, avail
);
1616 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
, refd
);
1618 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
1619 dsl_dataset_phys(ds
)->ds_creation_time
);
1620 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
1621 dsl_dataset_phys(ds
)->ds_creation_txg
);
1622 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
1624 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
1626 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
1627 dsl_dataset_phys(ds
)->ds_guid
);
1628 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
1629 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1630 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
1632 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
1634 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
1635 DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
1637 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
1638 uint64_t written
, comp
, uncomp
;
1639 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1640 dsl_dataset_t
*prev
;
1643 err
= dsl_dataset_hold_obj(dp
,
1644 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
1646 err
= dsl_dataset_space_written(prev
, ds
, &written
,
1648 dsl_dataset_rele(prev
, FTAG
);
1650 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
1659 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
1661 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1662 ASSERT(dsl_pool_config_held(dp
));
1664 stat
->dds_creation_txg
= dsl_dataset_phys(ds
)->ds_creation_txg
;
1665 stat
->dds_inconsistent
=
1666 dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
;
1667 stat
->dds_guid
= dsl_dataset_phys(ds
)->ds_guid
;
1668 stat
->dds_origin
[0] = '\0';
1669 if (ds
->ds_is_snapshot
) {
1670 stat
->dds_is_snapshot
= B_TRUE
;
1671 stat
->dds_num_clones
=
1672 dsl_dataset_phys(ds
)->ds_num_children
- 1;
1674 stat
->dds_is_snapshot
= B_FALSE
;
1675 stat
->dds_num_clones
= 0;
1677 if (dsl_dir_is_clone(ds
->ds_dir
)) {
1680 VERIFY0(dsl_dataset_hold_obj(dp
,
1681 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
,
1683 dsl_dataset_name(ods
, stat
->dds_origin
);
1684 dsl_dataset_rele(ods
, FTAG
);
1690 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
1692 return (ds
->ds_fsid_guid
);
1696 dsl_dataset_space(dsl_dataset_t
*ds
,
1697 uint64_t *refdbytesp
, uint64_t *availbytesp
,
1698 uint64_t *usedobjsp
, uint64_t *availobjsp
)
1700 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1701 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
1702 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
1704 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
1705 if (ds
->ds_quota
!= 0) {
1707 * Adjust available bytes according to refquota
1709 if (*refdbytesp
< ds
->ds_quota
)
1710 *availbytesp
= MIN(*availbytesp
,
1711 ds
->ds_quota
- *refdbytesp
);
1715 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
1716 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
1720 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
1722 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1725 if (dsl_dataset_phys(ds
)->ds_bp
.blk_birth
>
1726 dsl_dataset_phys(snap
)->ds_creation_txg
) {
1727 objset_t
*os
, *os_snap
;
1729 * It may be that only the ZIL differs, because it was
1730 * reset in the head. Don't count that as being
1733 if (dmu_objset_from_ds(ds
, &os
) != 0)
1735 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
1737 return (bcmp(&os
->os_phys
->os_meta_dnode
,
1738 &os_snap
->os_phys
->os_meta_dnode
,
1739 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
1744 typedef struct dsl_dataset_rename_snapshot_arg
{
1745 const char *ddrsa_fsname
;
1746 const char *ddrsa_oldsnapname
;
1747 const char *ddrsa_newsnapname
;
1748 boolean_t ddrsa_recursive
;
1750 } dsl_dataset_rename_snapshot_arg_t
;
1754 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
1755 dsl_dataset_t
*hds
, void *arg
)
1757 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1761 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1763 /* ignore nonexistent snapshots */
1764 return (error
== ENOENT
? 0 : error
);
1767 /* new name should not exist */
1768 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
1770 error
= SET_ERROR(EEXIST
);
1771 else if (error
== ENOENT
)
1774 /* dataset name + 1 for the "@" + the new snapshot name must fit */
1775 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
1776 strlen(ddrsa
->ddrsa_newsnapname
) >= MAXNAMELEN
)
1777 error
= SET_ERROR(ENAMETOOLONG
);
1783 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1785 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1786 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1790 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
1794 if (ddrsa
->ddrsa_recursive
) {
1795 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1796 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
1799 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
1801 dsl_dataset_rele(hds
, FTAG
);
1806 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
1807 dsl_dataset_t
*hds
, void *arg
)
1809 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1812 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
1815 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1816 ASSERT(error
== 0 || error
== ENOENT
);
1817 if (error
== ENOENT
) {
1818 /* ignore nonexistent snapshots */
1822 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
1824 /* log before we change the name */
1825 spa_history_log_internal_ds(ds
, "rename", tx
,
1826 "-> @%s", ddrsa
->ddrsa_newsnapname
);
1828 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
1830 mutex_enter(&ds
->ds_lock
);
1831 (void) strcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
);
1832 mutex_exit(&ds
->ds_lock
);
1833 VERIFY0(zap_add(dp
->dp_meta_objset
,
1834 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
1835 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
1837 dsl_dataset_rele(ds
, FTAG
);
1842 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1844 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1845 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1848 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
1849 ddrsa
->ddrsa_tx
= tx
;
1850 if (ddrsa
->ddrsa_recursive
) {
1851 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1852 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
1855 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
1857 dsl_dataset_rele(hds
, FTAG
);
1861 dsl_dataset_rename_snapshot(const char *fsname
,
1862 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
1865 char *oldname
, *newname
;
1869 dsl_dataset_rename_snapshot_arg_t ddrsa
;
1871 ddrsa
.ddrsa_fsname
= fsname
;
1872 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
1873 ddrsa
.ddrsa_newsnapname
= newsnapname
;
1874 ddrsa
.ddrsa_recursive
= recursive
;
1876 error
= dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
1877 dsl_dataset_rename_snapshot_sync
, &ddrsa
, 1);
1880 return (SET_ERROR(error
));
1883 oldname
= kmem_asprintf("%s@%s", fsname
, oldsnapname
);
1884 newname
= kmem_asprintf("%s@%s", fsname
, newsnapname
);
1885 zvol_rename_minors(oldname
, newname
);
1894 * If we're doing an ownership handoff, we need to make sure that there is
1895 * only one long hold on the dataset. We're not allowed to change anything here
1896 * so we don't permanently release the long hold or regular hold here. We want
1897 * to do this only when syncing to avoid the dataset unexpectedly going away
1898 * when we release the long hold.
1901 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
1905 if (!dmu_tx_is_syncing(tx
))
1908 if (owner
!= NULL
) {
1909 VERIFY3P(ds
->ds_owner
, ==, owner
);
1910 dsl_dataset_long_rele(ds
, owner
);
1913 held
= dsl_dataset_long_held(ds
);
1916 dsl_dataset_long_hold(ds
, owner
);
1919 return (SET_ERROR(EBUSY
));
1924 typedef struct dsl_dataset_rollback_arg
{
1925 const char *ddra_fsname
;
1927 nvlist_t
*ddra_result
;
1928 } dsl_dataset_rollback_arg_t
;
1931 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
1933 dsl_dataset_rollback_arg_t
*ddra
= arg
;
1934 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1936 int64_t unused_refres_delta
;
1939 nvlist_t
*proprequest
, *bookmarks
;
1941 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
1945 /* must not be a snapshot */
1946 if (ds
->ds_is_snapshot
) {
1947 dsl_dataset_rele(ds
, FTAG
);
1948 return (SET_ERROR(EINVAL
));
1951 /* must have a most recent snapshot */
1952 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
1953 dsl_dataset_rele(ds
, FTAG
);
1954 return (SET_ERROR(EINVAL
));
1957 /* must not have any bookmarks after the most recent snapshot */
1958 proprequest
= fnvlist_alloc();
1959 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
1960 bookmarks
= fnvlist_alloc();
1961 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
1962 fnvlist_free(proprequest
);
1965 for (pair
= nvlist_next_nvpair(bookmarks
, NULL
);
1966 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
1968 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
1969 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
1970 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
1971 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
1972 fnvlist_free(bookmarks
);
1973 dsl_dataset_rele(ds
, FTAG
);
1974 return (SET_ERROR(EEXIST
));
1977 fnvlist_free(bookmarks
);
1979 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
1981 dsl_dataset_rele(ds
, FTAG
);
1986 * Check if the snap we are rolling back to uses more than
1989 if (ds
->ds_quota
!= 0 &&
1990 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
1991 dsl_dataset_rele(ds
, FTAG
);
1992 return (SET_ERROR(EDQUOT
));
1996 * When we do the clone swap, we will temporarily use more space
1997 * due to the refreservation (the head will no longer have any
1998 * unique space, so the entire amount of the refreservation will need
1999 * to be free). We will immediately destroy the clone, freeing
2000 * this space, but the freeing happens over many txg's.
2002 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2003 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2005 if (unused_refres_delta
> 0 &&
2006 unused_refres_delta
>
2007 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2008 dsl_dataset_rele(ds
, FTAG
);
2009 return (SET_ERROR(ENOSPC
));
2012 dsl_dataset_rele(ds
, FTAG
);
2017 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2019 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2020 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2021 dsl_dataset_t
*ds
, *clone
;
2023 char namebuf
[ZFS_MAXNAMELEN
];
2025 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2027 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2028 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2030 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2031 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, tx
);
2033 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2035 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2036 dsl_dataset_zero_zil(ds
, tx
);
2038 dsl_destroy_head_sync_impl(clone
, tx
);
2040 dsl_dataset_rele(clone
, FTAG
);
2041 dsl_dataset_rele(ds
, FTAG
);
2045 * Rolls back the given filesystem or volume to the most recent snapshot.
2046 * The name of the most recent snapshot will be returned under key "target"
2047 * in the result nvlist.
2050 * - The existing dataset MUST be owned by the specified owner at entry
2051 * - Upon return, dataset will still be held by the same owner, whether we
2054 * This mode is required any time the existing filesystem is mounted. See
2055 * notes above zfs_suspend_fs() for further details.
2058 dsl_dataset_rollback(const char *fsname
, void *owner
, nvlist_t
*result
)
2060 dsl_dataset_rollback_arg_t ddra
;
2062 ddra
.ddra_fsname
= fsname
;
2063 ddra
.ddra_owner
= owner
;
2064 ddra
.ddra_result
= result
;
2066 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2067 dsl_dataset_rollback_sync
, &ddra
, 1));
2070 struct promotenode
{
2075 typedef struct dsl_dataset_promote_arg
{
2076 const char *ddpa_clonename
;
2077 dsl_dataset_t
*ddpa_clone
;
2078 list_t shared_snaps
, origin_snaps
, clone_snaps
;
2079 dsl_dataset_t
*origin_origin
; /* origin of the origin */
2080 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
2083 } dsl_dataset_promote_arg_t
;
2085 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2086 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2088 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2091 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2093 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2094 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2096 struct promotenode
*snap
;
2097 dsl_dataset_t
*origin_ds
;
2102 err
= promote_hold(ddpa
, dp
, FTAG
);
2106 hds
= ddpa
->ddpa_clone
;
2108 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2109 promote_rele(ddpa
, FTAG
);
2110 return (SET_ERROR(EXDEV
));
2114 * Compute and check the amount of space to transfer. Since this is
2115 * so expensive, don't do the preliminary check.
2117 if (!dmu_tx_is_syncing(tx
)) {
2118 promote_rele(ddpa
, FTAG
);
2122 snap
= list_head(&ddpa
->shared_snaps
);
2123 origin_ds
= snap
->ds
;
2125 /* compute origin's new unique space */
2126 snap
= list_tail(&ddpa
->clone_snaps
);
2127 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2128 origin_ds
->ds_object
);
2129 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2130 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2131 &ddpa
->unique
, &unused
, &unused
);
2134 * Walk the snapshots that we are moving
2136 * Compute space to transfer. Consider the incremental changes
2137 * to used by each snapshot:
2138 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2139 * So each snapshot gave birth to:
2140 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2141 * So a sequence would look like:
2142 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2143 * Which simplifies to:
2144 * uN + kN + kN-1 + ... + k1 + k0
2145 * Note however, if we stop before we reach the ORIGIN we get:
2146 * uN + kN + kN-1 + ... + kM - uM-1
2149 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2150 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2151 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2152 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2153 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2154 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2155 dsl_dataset_t
*ds
= snap
->ds
;
2160 * If there are long holds, we won't be able to evict
2163 if (dsl_dataset_long_held(ds
)) {
2164 err
= SET_ERROR(EBUSY
);
2168 /* Check that the snapshot name does not conflict */
2169 VERIFY0(dsl_dataset_get_snapname(ds
));
2170 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2172 (void) strcpy(ddpa
->err_ds
, snap
->ds
->ds_snapname
);
2173 err
= SET_ERROR(EEXIST
);
2179 /* The very first snapshot does not have a deadlist */
2180 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2183 dsl_deadlist_space(&ds
->ds_deadlist
,
2184 &dlused
, &dlcomp
, &dluncomp
);
2185 ddpa
->used
+= dlused
;
2186 ddpa
->comp
+= dlcomp
;
2187 ddpa
->uncomp
+= dluncomp
;
2191 * If we are a clone of a clone then we never reached ORIGIN,
2192 * so we need to subtract out the clone origin's used space.
2194 if (ddpa
->origin_origin
) {
2196 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2198 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2200 dsl_dataset_phys(ddpa
->origin_origin
)->
2201 ds_uncompressed_bytes
;
2204 /* Check that there is enough space and limit headroom here */
2205 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2206 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2211 * Compute the amounts of space that will be used by snapshots
2212 * after the promotion (for both origin and clone). For each,
2213 * it is the amount of space that will be on all of their
2214 * deadlists (that was not born before their new origin).
2216 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2220 * Note, typically this will not be a clone of a clone,
2221 * so dd_origin_txg will be < TXG_INITIAL, so
2222 * these snaplist_space() -> dsl_deadlist_space_range()
2223 * calls will be fast because they do not have to
2224 * iterate over all bps.
2226 snap
= list_head(&ddpa
->origin_snaps
);
2227 err
= snaplist_space(&ddpa
->shared_snaps
,
2228 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2232 err
= snaplist_space(&ddpa
->clone_snaps
,
2233 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2236 ddpa
->cloneusedsnap
+= space
;
2238 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2239 DD_FLAG_USED_BREAKDOWN
) {
2240 err
= snaplist_space(&ddpa
->origin_snaps
,
2241 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2242 &ddpa
->originusedsnap
);
2248 promote_rele(ddpa
, FTAG
);
2253 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2255 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2256 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2258 struct promotenode
*snap
;
2259 dsl_dataset_t
*origin_ds
;
2260 dsl_dataset_t
*origin_head
;
2262 dsl_dir_t
*odd
= NULL
;
2263 uint64_t oldnext_obj
;
2266 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2267 hds
= ddpa
->ddpa_clone
;
2269 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2271 snap
= list_head(&ddpa
->shared_snaps
);
2272 origin_ds
= snap
->ds
;
2275 snap
= list_head(&ddpa
->origin_snaps
);
2276 origin_head
= snap
->ds
;
2279 * We need to explicitly open odd, since origin_ds's dd will be
2282 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2285 /* change origin's next snap */
2286 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2287 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2288 snap
= list_tail(&ddpa
->clone_snaps
);
2289 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2290 origin_ds
->ds_object
);
2291 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2293 /* change the origin's next clone */
2294 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2295 dsl_dataset_remove_from_next_clones(origin_ds
,
2296 snap
->ds
->ds_object
, tx
);
2297 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2298 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2303 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2304 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2305 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2306 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2307 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2308 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2309 origin_head
->ds_dir
->dd_origin_txg
=
2310 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2312 /* change dd_clone entries */
2313 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2314 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2315 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2316 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2317 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2318 hds
->ds_object
, tx
));
2320 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2321 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2322 origin_head
->ds_object
, tx
));
2323 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2324 dsl_dir_phys(dd
)->dd_clones
=
2325 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2326 DMU_OT_NONE
, 0, tx
);
2328 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2329 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2332 /* move snapshots to this dir */
2333 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2334 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2335 dsl_dataset_t
*ds
= snap
->ds
;
2338 * Property callbacks are registered to a particular
2339 * dsl_dir. Since ours is changing, evict the objset
2340 * so that they will be unregistered from the old dsl_dir.
2342 if (ds
->ds_objset
) {
2343 dmu_objset_evict(ds
->ds_objset
);
2344 ds
->ds_objset
= NULL
;
2347 /* move snap name entry */
2348 VERIFY0(dsl_dataset_get_snapname(ds
));
2349 VERIFY0(dsl_dataset_snap_remove(origin_head
,
2350 ds
->ds_snapname
, tx
, B_TRUE
));
2351 VERIFY0(zap_add(dp
->dp_meta_objset
,
2352 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
2353 8, 1, &ds
->ds_object
, tx
));
2354 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
2355 DD_FIELD_SNAPSHOT_COUNT
, tx
);
2357 /* change containing dsl_dir */
2358 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2359 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
2360 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
2361 ASSERT3P(ds
->ds_dir
, ==, odd
);
2362 dsl_dir_rele(ds
->ds_dir
, ds
);
2363 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
2364 NULL
, ds
, &ds
->ds_dir
));
2366 /* move any clone references */
2367 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
2368 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2372 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
2373 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2374 zap_cursor_retrieve(&zc
, &za
) == 0;
2375 zap_cursor_advance(&zc
)) {
2376 dsl_dataset_t
*cnds
;
2379 if (za
.za_first_integer
== oldnext_obj
) {
2381 * We've already moved the
2382 * origin's reference.
2387 VERIFY0(dsl_dataset_hold_obj(dp
,
2388 za
.za_first_integer
, FTAG
, &cnds
));
2389 o
= dsl_dir_phys(cnds
->ds_dir
)->
2390 dd_head_dataset_obj
;
2392 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2393 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
2394 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2395 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
2396 dsl_dataset_rele(cnds
, FTAG
);
2398 zap_cursor_fini(&zc
);
2401 ASSERT(!dsl_prop_hascb(ds
));
2405 * Change space accounting.
2406 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2407 * both be valid, or both be 0 (resulting in delta == 0). This
2408 * is true for each of {clone,origin} independently.
2411 delta
= ddpa
->cloneusedsnap
-
2412 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
2413 ASSERT3S(delta
, >=, 0);
2414 ASSERT3U(ddpa
->used
, >=, delta
);
2415 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2416 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
2417 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
2419 delta
= ddpa
->originusedsnap
-
2420 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
2421 ASSERT3S(delta
, <=, 0);
2422 ASSERT3U(ddpa
->used
, >=, -delta
);
2423 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2424 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
2425 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
2427 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
2429 /* log history record */
2430 spa_history_log_internal_ds(hds
, "promote", tx
, "");
2432 dsl_dir_rele(odd
, FTAG
);
2433 promote_rele(ddpa
, FTAG
);
2437 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2438 * (exclusive) and last_obj (inclusive). The list will be in reverse
2439 * order (last_obj will be the list_head()). If first_obj == 0, do all
2440 * snapshots back to this dataset's origin.
2443 snaplist_make(dsl_pool_t
*dp
,
2444 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
2446 uint64_t obj
= last_obj
;
2448 list_create(l
, sizeof (struct promotenode
),
2449 offsetof(struct promotenode
, link
));
2451 while (obj
!= first_obj
) {
2453 struct promotenode
*snap
;
2456 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
2457 ASSERT(err
!= ENOENT
);
2462 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
2464 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
2466 list_insert_tail(l
, snap
);
2467 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
2474 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
2476 struct promotenode
*snap
;
2479 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
2480 uint64_t used
, comp
, uncomp
;
2481 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2482 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
2489 snaplist_destroy(list_t
*l
, void *tag
)
2491 struct promotenode
*snap
;
2493 if (l
== NULL
|| !list_link_active(&l
->list_head
))
2496 while ((snap
= list_tail(l
)) != NULL
) {
2497 list_remove(l
, snap
);
2498 dsl_dataset_rele(snap
->ds
, tag
);
2499 kmem_free(snap
, sizeof (*snap
));
2505 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
2509 struct promotenode
*snap
;
2511 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
2515 dd
= ddpa
->ddpa_clone
->ds_dir
;
2517 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
2518 !dsl_dir_is_clone(dd
)) {
2519 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2520 return (SET_ERROR(EINVAL
));
2523 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
2524 &ddpa
->shared_snaps
, tag
);
2528 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
2529 &ddpa
->clone_snaps
, tag
);
2533 snap
= list_head(&ddpa
->shared_snaps
);
2534 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
2535 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
2536 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
2537 &ddpa
->origin_snaps
, tag
);
2541 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
2542 error
= dsl_dataset_hold_obj(dp
,
2543 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
2544 tag
, &ddpa
->origin_origin
);
2550 promote_rele(ddpa
, tag
);
2555 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
2557 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
2558 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
2559 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
2560 if (ddpa
->origin_origin
!= NULL
)
2561 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
2562 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2568 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2569 * in with the name. (It must be at least MAXNAMELEN bytes long.)
2572 dsl_dataset_promote(const char *name
, char *conflsnap
)
2574 dsl_dataset_promote_arg_t ddpa
= { 0 };
2580 * We will modify space proportional to the number of
2581 * snapshots. Compute numsnaps.
2583 error
= dmu_objset_hold(name
, FTAG
, &os
);
2586 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
2587 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
2589 dmu_objset_rele(os
, FTAG
);
2593 ddpa
.ddpa_clonename
= name
;
2594 ddpa
.err_ds
= conflsnap
;
2597 return (dsl_sync_task(name
, dsl_dataset_promote_check
,
2598 dsl_dataset_promote_sync
, &ddpa
, 2 + numsnaps
));
2602 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
2603 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
2605 int64_t unused_refres_delta
;
2607 /* they should both be heads */
2608 if (clone
->ds_is_snapshot
||
2609 origin_head
->ds_is_snapshot
)
2610 return (SET_ERROR(EINVAL
));
2612 /* if we are not forcing, the branch point should be just before them */
2613 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
2614 return (SET_ERROR(EINVAL
));
2616 /* clone should be the clone (unless they are unrelated) */
2617 if (clone
->ds_prev
!= NULL
&&
2618 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
2619 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
2620 return (SET_ERROR(EINVAL
));
2622 /* the clone should be a child of the origin */
2623 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
2624 return (SET_ERROR(EINVAL
));
2626 /* origin_head shouldn't be modified unless 'force' */
2628 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
2629 return (SET_ERROR(ETXTBSY
));
2631 /* origin_head should have no long holds (e.g. is not mounted) */
2632 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
2633 return (SET_ERROR(EBUSY
));
2635 /* check amount of any unconsumed refreservation */
2636 unused_refres_delta
=
2637 (int64_t)MIN(origin_head
->ds_reserved
,
2638 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2639 (int64_t)MIN(origin_head
->ds_reserved
,
2640 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2642 if (unused_refres_delta
> 0 &&
2643 unused_refres_delta
>
2644 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
2645 return (SET_ERROR(ENOSPC
));
2647 /* clone can't be over the head's refquota */
2648 if (origin_head
->ds_quota
!= 0 &&
2649 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
2650 origin_head
->ds_quota
)
2651 return (SET_ERROR(EDQUOT
));
2657 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
2658 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
2660 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2661 int64_t unused_refres_delta
;
2663 ASSERT(clone
->ds_reserved
== 0);
2664 ASSERT(origin_head
->ds_quota
== 0 ||
2665 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
);
2666 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
2668 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
2669 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
2671 if (clone
->ds_objset
!= NULL
) {
2672 dmu_objset_evict(clone
->ds_objset
);
2673 clone
->ds_objset
= NULL
;
2676 if (origin_head
->ds_objset
!= NULL
) {
2677 dmu_objset_evict(origin_head
->ds_objset
);
2678 origin_head
->ds_objset
= NULL
;
2681 unused_refres_delta
=
2682 (int64_t)MIN(origin_head
->ds_reserved
,
2683 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2684 (int64_t)MIN(origin_head
->ds_reserved
,
2685 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2688 * Reset origin's unique bytes, if it exists.
2690 if (clone
->ds_prev
) {
2691 dsl_dataset_t
*origin
= clone
->ds_prev
;
2692 uint64_t comp
, uncomp
;
2694 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
2695 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2696 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
2697 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
2703 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
2704 dsl_dataset_phys(origin_head
)->ds_bp
=
2705 dsl_dataset_phys(clone
)->ds_bp
;
2706 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
2709 /* set dd_*_bytes */
2711 int64_t dused
, dcomp
, duncomp
;
2712 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
2713 uint64_t odl_used
, odl_comp
, odl_uncomp
;
2715 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
2716 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
2718 dsl_deadlist_space(&clone
->ds_deadlist
,
2719 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2720 dsl_deadlist_space(&origin_head
->ds_deadlist
,
2721 &odl_used
, &odl_comp
, &odl_uncomp
);
2723 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
2725 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
2727 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
2729 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
2731 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
2733 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
2736 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
2737 dused
, dcomp
, duncomp
, tx
);
2738 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
2739 -dused
, -dcomp
, -duncomp
, tx
);
2742 * The difference in the space used by snapshots is the
2743 * difference in snapshot space due to the head's
2744 * deadlist (since that's the only thing that's
2745 * changing that affects the snapused).
2747 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2748 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2749 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2750 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
2751 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2752 &odl_used
, &odl_comp
, &odl_uncomp
);
2753 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
2754 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
2757 /* swap ds_*_bytes */
2758 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
2759 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
2760 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
2761 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
2762 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
2763 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
2764 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
2765 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2767 /* apply any parent delta for change in unconsumed refreservation */
2768 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
2769 unused_refres_delta
, 0, 0, tx
);
2774 dsl_deadlist_close(&clone
->ds_deadlist
);
2775 dsl_deadlist_close(&origin_head
->ds_deadlist
);
2776 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
2777 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2778 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
2779 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2780 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
2781 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
2783 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
2785 spa_history_log_internal_ds(clone
, "clone swap", tx
,
2786 "parent=%s", origin_head
->ds_dir
->dd_myname
);
2790 * Given a pool name and a dataset object number in that pool,
2791 * return the name of that dataset.
2794 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
2800 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
2804 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
2806 dsl_dataset_name(ds
, buf
);
2807 dsl_dataset_rele(ds
, FTAG
);
2809 dsl_pool_rele(dp
, FTAG
);
2815 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
2816 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
2820 ASSERT3S(asize
, >, 0);
2823 * *ref_rsrv is the portion of asize that will come from any
2824 * unconsumed refreservation space.
2828 mutex_enter(&ds
->ds_lock
);
2830 * Make a space adjustment for reserved bytes.
2832 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2834 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2836 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2838 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
2841 if (!check_quota
|| ds
->ds_quota
== 0) {
2842 mutex_exit(&ds
->ds_lock
);
2846 * If they are requesting more space, and our current estimate
2847 * is over quota, they get to try again unless the actual
2848 * on-disk is over quota and there are no pending changes (which
2849 * may free up space for us).
2851 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
2854 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
2855 error
= SET_ERROR(ERESTART
);
2857 error
= SET_ERROR(EDQUOT
);
2859 mutex_exit(&ds
->ds_lock
);
2864 typedef struct dsl_dataset_set_qr_arg
{
2865 const char *ddsqra_name
;
2866 zprop_source_t ddsqra_source
;
2867 uint64_t ddsqra_value
;
2868 } dsl_dataset_set_qr_arg_t
;
2873 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
2875 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2876 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2881 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
2882 return (SET_ERROR(ENOTSUP
));
2884 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
2888 if (ds
->ds_is_snapshot
) {
2889 dsl_dataset_rele(ds
, FTAG
);
2890 return (SET_ERROR(EINVAL
));
2893 error
= dsl_prop_predict(ds
->ds_dir
,
2894 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
2895 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
2897 dsl_dataset_rele(ds
, FTAG
);
2902 dsl_dataset_rele(ds
, FTAG
);
2906 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
2907 newval
< ds
->ds_reserved
) {
2908 dsl_dataset_rele(ds
, FTAG
);
2909 return (SET_ERROR(ENOSPC
));
2912 dsl_dataset_rele(ds
, FTAG
);
2917 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
2919 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2920 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2924 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
2926 dsl_prop_set_sync_impl(ds
,
2927 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
2928 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
2929 &ddsqra
->ddsqra_value
, tx
);
2931 VERIFY0(dsl_prop_get_int_ds(ds
,
2932 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
2934 if (ds
->ds_quota
!= newval
) {
2935 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2936 ds
->ds_quota
= newval
;
2938 dsl_dataset_rele(ds
, FTAG
);
2942 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
2945 dsl_dataset_set_qr_arg_t ddsqra
;
2947 ddsqra
.ddsqra_name
= dsname
;
2948 ddsqra
.ddsqra_source
= source
;
2949 ddsqra
.ddsqra_value
= refquota
;
2951 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
2952 dsl_dataset_set_refquota_sync
, &ddsqra
, 0));
2956 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
2958 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2959 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2962 uint64_t newval
, unique
;
2964 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
2965 return (SET_ERROR(ENOTSUP
));
2967 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
2971 if (ds
->ds_is_snapshot
) {
2972 dsl_dataset_rele(ds
, FTAG
);
2973 return (SET_ERROR(EINVAL
));
2976 error
= dsl_prop_predict(ds
->ds_dir
,
2977 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
2978 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
2980 dsl_dataset_rele(ds
, FTAG
);
2985 * If we are doing the preliminary check in open context, the
2986 * space estimates may be inaccurate.
2988 if (!dmu_tx_is_syncing(tx
)) {
2989 dsl_dataset_rele(ds
, FTAG
);
2993 mutex_enter(&ds
->ds_lock
);
2994 if (!DS_UNIQUE_IS_ACCURATE(ds
))
2995 dsl_dataset_recalc_head_uniq(ds
);
2996 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
2997 mutex_exit(&ds
->ds_lock
);
2999 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3000 uint64_t delta
= MAX(unique
, newval
) -
3001 MAX(unique
, ds
->ds_reserved
);
3004 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3005 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3006 dsl_dataset_rele(ds
, FTAG
);
3007 return (SET_ERROR(ENOSPC
));
3011 dsl_dataset_rele(ds
, FTAG
);
3016 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3017 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3023 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3024 source
, sizeof (value
), 1, &value
, tx
);
3026 VERIFY0(dsl_prop_get_int_ds(ds
,
3027 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3029 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3030 mutex_enter(&ds
->ds_dir
->dd_lock
);
3031 mutex_enter(&ds
->ds_lock
);
3032 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3033 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3034 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3035 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3036 ds
->ds_reserved
= newval
;
3037 mutex_exit(&ds
->ds_lock
);
3039 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3040 mutex_exit(&ds
->ds_dir
->dd_lock
);
3044 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3046 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3047 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3050 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3051 dsl_dataset_set_refreservation_sync_impl(ds
,
3052 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3053 dsl_dataset_rele(ds
, FTAG
);
3057 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3058 uint64_t refreservation
)
3060 dsl_dataset_set_qr_arg_t ddsqra
;
3062 ddsqra
.ddsqra_name
= dsname
;
3063 ddsqra
.ddsqra_source
= source
;
3064 ddsqra
.ddsqra_value
= refreservation
;
3066 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3067 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0));
3071 * Return (in *usedp) the amount of space written in new that is not
3072 * present in oldsnap. New may be a snapshot or the head. Old must be
3073 * a snapshot before new, in new's filesystem (or its origin). If not then
3074 * fail and return EINVAL.
3076 * The written space is calculated by considering two components: First, we
3077 * ignore any freed space, and calculate the written as new's used space
3078 * minus old's used space. Next, we add in the amount of space that was freed
3079 * between the two snapshots, thus reducing new's used space relative to old's.
3080 * Specifically, this is the space that was born before old->ds_creation_txg,
3081 * and freed before new (ie. on new's deadlist or a previous deadlist).
3083 * space freed [---------------------]
3084 * snapshots ---O-------O--------O-------O------
3088 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3089 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3093 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3095 ASSERT(dsl_pool_config_held(dp
));
3098 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3099 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3102 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3103 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3106 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3107 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3109 snapobj
= new->ds_object
;
3110 while (snapobj
!= oldsnap
->ds_object
) {
3111 dsl_dataset_t
*snap
;
3112 uint64_t used
, comp
, uncomp
;
3114 if (snapobj
== new->ds_object
) {
3117 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3122 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3123 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3125 * The blocks in the deadlist can not be born after
3126 * ds_prev_snap_txg, so get the whole deadlist space,
3127 * which is more efficient (especially for old-format
3128 * deadlists). Unfortunately the deadlist code
3129 * doesn't have enough information to make this
3130 * optimization itself.
3132 dsl_deadlist_space(&snap
->ds_deadlist
,
3133 &used
, &comp
, &uncomp
);
3135 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3136 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3137 &used
, &comp
, &uncomp
);
3144 * If we get to the beginning of the chain of snapshots
3145 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3146 * was not a snapshot of/before new.
3148 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3150 dsl_dataset_rele(snap
, FTAG
);
3152 err
= SET_ERROR(EINVAL
);
3161 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3162 * lastsnap, and all snapshots in between are deleted.
3164 * blocks that would be freed [---------------------------]
3165 * snapshots ---O-------O--------O-------O--------O
3166 * firstsnap lastsnap
3168 * This is the set of blocks that were born after the snap before firstsnap,
3169 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3170 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3171 * We calculate this by iterating over the relevant deadlists (from the snap
3172 * after lastsnap, backward to the snap after firstsnap), summing up the
3173 * space on the deadlist that was born after the snap before firstsnap.
3176 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3177 dsl_dataset_t
*lastsnap
,
3178 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3182 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3184 ASSERT(firstsnap
->ds_is_snapshot
);
3185 ASSERT(lastsnap
->ds_is_snapshot
);
3188 * Check that the snapshots are in the same dsl_dir, and firstsnap
3189 * is before lastsnap.
3191 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3192 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3193 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3194 return (SET_ERROR(EINVAL
));
3196 *usedp
= *compp
= *uncompp
= 0;
3198 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3199 while (snapobj
!= firstsnap
->ds_object
) {
3201 uint64_t used
, comp
, uncomp
;
3203 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3207 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3208 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3209 &used
, &comp
, &uncomp
);
3214 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3215 ASSERT3U(snapobj
, !=, 0);
3216 dsl_dataset_rele(ds
, FTAG
);
3222 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3223 * For example, they could both be snapshots of the same filesystem, and
3224 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3225 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3226 * filesystem. Or 'earlier' could be the origin's origin.
3228 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3231 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3232 uint64_t earlier_txg
)
3234 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3237 dsl_dataset_t
*origin
;
3239 ASSERT(dsl_pool_config_held(dp
));
3240 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3242 if (earlier_txg
== 0)
3243 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3245 if (later
->ds_is_snapshot
&&
3246 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3249 if (later
->ds_dir
== earlier
->ds_dir
)
3251 if (!dsl_dir_is_clone(later
->ds_dir
))
3254 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3256 error
= dsl_dataset_hold_obj(dp
,
3257 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3260 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3261 dsl_dataset_rele(origin
, FTAG
);
3267 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
3269 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
3270 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
3273 #if defined(_KERNEL) && defined(HAVE_SPL)
3274 EXPORT_SYMBOL(dsl_dataset_hold
);
3275 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
3276 EXPORT_SYMBOL(dsl_dataset_own
);
3277 EXPORT_SYMBOL(dsl_dataset_own_obj
);
3278 EXPORT_SYMBOL(dsl_dataset_name
);
3279 EXPORT_SYMBOL(dsl_dataset_rele
);
3280 EXPORT_SYMBOL(dsl_dataset_disown
);
3281 EXPORT_SYMBOL(dsl_dataset_tryown
);
3282 EXPORT_SYMBOL(dsl_dataset_create_sync
);
3283 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
3284 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
3285 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
3286 EXPORT_SYMBOL(dsl_dataset_promote
);
3287 EXPORT_SYMBOL(dsl_dataset_user_hold
);
3288 EXPORT_SYMBOL(dsl_dataset_user_release
);
3289 EXPORT_SYMBOL(dsl_dataset_get_holds
);
3290 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
3291 EXPORT_SYMBOL(dsl_dataset_set_blkptr
);
3292 EXPORT_SYMBOL(dsl_dataset_get_spa
);
3293 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
3294 EXPORT_SYMBOL(dsl_dataset_space_written
);
3295 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
3296 EXPORT_SYMBOL(dsl_dataset_sync
);
3297 EXPORT_SYMBOL(dsl_dataset_block_born
);
3298 EXPORT_SYMBOL(dsl_dataset_block_kill
);
3299 EXPORT_SYMBOL(dsl_dataset_block_freeable
);
3300 EXPORT_SYMBOL(dsl_dataset_prev_snap_txg
);
3301 EXPORT_SYMBOL(dsl_dataset_dirty
);
3302 EXPORT_SYMBOL(dsl_dataset_stats
);
3303 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
3304 EXPORT_SYMBOL(dsl_dataset_space
);
3305 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
3306 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
3307 EXPORT_SYMBOL(dsl_dataset_check_quota
);
3308 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
3309 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);