4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
24 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
25 * Copyright (c) 2014 RackTop Systems.
26 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
27 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
30 #include <sys/dmu_objset.h>
31 #include <sys/dsl_dataset.h>
32 #include <sys/dsl_dir.h>
33 #include <sys/dsl_prop.h>
34 #include <sys/dsl_synctask.h>
35 #include <sys/dmu_traverse.h>
36 #include <sys/dmu_impl.h>
37 #include <sys/dmu_tx.h>
41 #include <sys/zfeature.h>
42 #include <sys/unique.h>
43 #include <sys/zfs_context.h>
44 #include <sys/zfs_ioctl.h>
46 #include <sys/zfs_znode.h>
47 #include <sys/zfs_onexit.h>
49 #include <sys/dsl_scan.h>
50 #include <sys/dsl_deadlist.h>
51 #include <sys/dsl_destroy.h>
52 #include <sys/dsl_userhold.h>
53 #include <sys/dsl_bookmark.h>
54 #include <sys/policy.h>
57 * The SPA supports block sizes up to 16MB. However, very large blocks
58 * can have an impact on i/o latency (e.g. tying up a spinning disk for
59 * ~300ms), and also potentially on the memory allocator. Therefore,
60 * we do not allow the recordsize to be set larger than zfs_max_recordsize
61 * (default 1MB). Larger blocks can be created by changing this tunable,
62 * and pools with larger blocks can always be imported and used, regardless
65 int zfs_max_recordsize
= 1 * 1024 * 1024;
67 #define SWITCH64(x, y) \
69 uint64_t __tmp = (x); \
74 #define DS_REF_MAX (1ULL << 62)
76 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
79 * Figure out how much of this delta should be propogated to the dsl_dir
80 * layer. If there's a refreservation, that space has already been
81 * partially accounted for in our ancestors.
84 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
86 dsl_dataset_phys_t
*ds_phys
;
87 uint64_t old_bytes
, new_bytes
;
89 if (ds
->ds_reserved
== 0)
92 ds_phys
= dsl_dataset_phys(ds
);
93 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
94 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
96 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
97 return (new_bytes
- old_bytes
);
101 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
103 int used
, compressed
, uncompressed
;
106 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
107 compressed
= BP_GET_PSIZE(bp
);
108 uncompressed
= BP_GET_UCSIZE(bp
);
110 dprintf_bp(bp
, "ds=%p", ds
);
112 ASSERT(dmu_tx_is_syncing(tx
));
113 /* It could have been compressed away to nothing */
116 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
117 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
119 dsl_pool_mos_diduse_space(tx
->tx_pool
,
120 used
, compressed
, uncompressed
);
124 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
125 mutex_enter(&ds
->ds_lock
);
126 delta
= parent_delta(ds
, used
);
127 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
128 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
129 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
130 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
131 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
132 ds
->ds_feature_activation_needed
[SPA_FEATURE_LARGE_BLOCKS
] =
135 mutex_exit(&ds
->ds_lock
);
136 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
137 compressed
, uncompressed
, tx
);
138 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
139 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
143 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
146 int used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
147 int compressed
= BP_GET_PSIZE(bp
);
148 int uncompressed
= BP_GET_UCSIZE(bp
);
153 ASSERT(dmu_tx_is_syncing(tx
));
154 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
157 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
158 dsl_pool_mos_diduse_space(tx
->tx_pool
,
159 -used
, -compressed
, -uncompressed
);
162 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
164 ASSERT(!ds
->ds_is_snapshot
);
165 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
167 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
170 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
171 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
173 mutex_enter(&ds
->ds_lock
);
174 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
175 !DS_UNIQUE_IS_ACCURATE(ds
));
176 delta
= parent_delta(ds
, -used
);
177 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
178 mutex_exit(&ds
->ds_lock
);
179 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
180 delta
, -compressed
, -uncompressed
, tx
);
181 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
182 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
184 dprintf_bp(bp
, "putting on dead list: %s", "");
187 * We are here as part of zio's write done callback,
188 * which means we're a zio interrupt thread. We can't
189 * call dsl_deadlist_insert() now because it may block
190 * waiting for I/O. Instead, put bp on the deferred
191 * queue and let dsl_pool_sync() finish the job.
193 bplist_append(&ds
->ds_pending_deadlist
, bp
);
195 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
197 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
198 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
199 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
200 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
201 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
202 ds
->ds_object
&& bp
->blk_birth
>
203 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
204 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
205 mutex_enter(&ds
->ds_prev
->ds_lock
);
206 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
207 mutex_exit(&ds
->ds_prev
->ds_lock
);
209 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
210 dsl_dir_transfer_space(ds
->ds_dir
, used
,
211 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
214 mutex_enter(&ds
->ds_lock
);
215 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
216 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
217 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
218 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
219 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
220 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
221 mutex_exit(&ds
->ds_lock
);
227 dsl_dataset_prev_snap_txg(dsl_dataset_t
*ds
)
229 uint64_t trysnap
= 0;
234 * The snapshot creation could fail, but that would cause an
235 * incorrect FALSE return, which would only result in an
236 * overestimation of the amount of space that an operation would
237 * consume, which is OK.
239 * There's also a small window where we could miss a pending
240 * snapshot, because we could set the sync task in the quiescing
241 * phase. So this should only be used as a guess.
243 if (ds
->ds_trysnap_txg
>
244 spa_last_synced_txg(ds
->ds_dir
->dd_pool
->dp_spa
))
245 trysnap
= ds
->ds_trysnap_txg
;
246 return (MAX(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, trysnap
));
250 dsl_dataset_block_freeable(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
253 if (blk_birth
<= dsl_dataset_prev_snap_txg(ds
) ||
254 (bp
!= NULL
&& BP_IS_HOLE(bp
)))
257 ddt_prefetch(dsl_dataset_get_spa(ds
), bp
);
263 dsl_dataset_evict(void *dbu
)
265 dsl_dataset_t
*ds
= dbu
;
267 ASSERT(ds
->ds_owner
== NULL
);
271 unique_remove(ds
->ds_fsid_guid
);
273 if (ds
->ds_objset
!= NULL
)
274 dmu_objset_evict(ds
->ds_objset
);
277 dsl_dataset_rele(ds
->ds_prev
, ds
);
281 bplist_destroy(&ds
->ds_pending_deadlist
);
282 if (ds
->ds_deadlist
.dl_os
!= NULL
)
283 dsl_deadlist_close(&ds
->ds_deadlist
);
285 dsl_dir_async_rele(ds
->ds_dir
, ds
);
287 ASSERT(!list_link_active(&ds
->ds_synced_link
));
289 list_destroy(&ds
->ds_prop_cbs
);
290 mutex_destroy(&ds
->ds_lock
);
291 mutex_destroy(&ds
->ds_opening_lock
);
292 mutex_destroy(&ds
->ds_sendstream_lock
);
293 refcount_destroy(&ds
->ds_longholds
);
295 kmem_free(ds
, sizeof (dsl_dataset_t
));
299 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
301 dsl_dataset_phys_t
*headphys
;
304 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
305 objset_t
*mos
= dp
->dp_meta_objset
;
307 if (ds
->ds_snapname
[0])
309 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
312 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
316 headphys
= headdbuf
->db_data
;
317 err
= zap_value_search(dp
->dp_meta_objset
,
318 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
319 if (err
!= 0 && zfs_recover
== B_TRUE
) {
321 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
322 "SNAPOBJ=%llu-ERR=%d",
323 (unsigned long long)ds
->ds_object
, err
);
325 dmu_buf_rele(headdbuf
, FTAG
);
330 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
332 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
333 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
337 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
342 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
343 value
, mt
, NULL
, 0, NULL
);
344 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
345 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
350 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
353 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
354 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
358 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
360 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
365 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
366 if (err
== ENOTSUP
&& mt
== MT_FIRST
)
367 err
= zap_remove(mos
, snapobj
, name
, tx
);
369 if (err
== 0 && adj_cnt
)
370 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
371 DD_FIELD_SNAPSHOT_COUNT
, tx
);
377 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
379 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
380 boolean_t result
= B_FALSE
;
382 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
383 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
385 if (ds
== dmu_buf_get_user(dbuf
))
388 dmu_buf_rele(dbuf
, tag
);
395 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
398 objset_t
*mos
= dp
->dp_meta_objset
;
402 dmu_object_info_t doi
;
404 ASSERT(dsl_pool_config_held(dp
));
406 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
410 /* Make sure dsobj has the correct object type. */
411 dmu_object_info_from_db(dbuf
, &doi
);
412 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
413 dmu_buf_rele(dbuf
, tag
);
414 return (SET_ERROR(EINVAL
));
417 ds
= dmu_buf_get_user(dbuf
);
419 dsl_dataset_t
*winner
= NULL
;
421 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
423 ds
->ds_object
= dsobj
;
424 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
425 list_link_init(&ds
->ds_synced_link
);
427 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
428 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
429 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
430 refcount_create(&ds
->ds_longholds
);
432 bplist_create(&ds
->ds_pending_deadlist
);
433 dsl_deadlist_open(&ds
->ds_deadlist
,
434 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
436 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
437 offsetof(dmu_sendarg_t
, dsa_link
));
439 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
440 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
442 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
445 for (f
= 0; f
< SPA_FEATURES
; f
++) {
446 if (!(spa_feature_table
[f
].fi_flags
&
447 ZFEATURE_FLAG_PER_DATASET
))
449 err
= zap_contains(mos
, dsobj
,
450 spa_feature_table
[f
].fi_guid
);
452 ds
->ds_feature_inuse
[f
] = B_TRUE
;
454 ASSERT3U(err
, ==, ENOENT
);
460 err
= dsl_dir_hold_obj(dp
,
461 dsl_dataset_phys(ds
)->ds_dir_obj
, NULL
, ds
, &ds
->ds_dir
);
463 mutex_destroy(&ds
->ds_lock
);
464 mutex_destroy(&ds
->ds_opening_lock
);
465 mutex_destroy(&ds
->ds_sendstream_lock
);
466 refcount_destroy(&ds
->ds_longholds
);
467 bplist_destroy(&ds
->ds_pending_deadlist
);
468 dsl_deadlist_close(&ds
->ds_deadlist
);
469 kmem_free(ds
, sizeof (dsl_dataset_t
));
470 dmu_buf_rele(dbuf
, tag
);
474 if (!ds
->ds_is_snapshot
) {
475 ds
->ds_snapname
[0] = '\0';
476 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
477 err
= dsl_dataset_hold_obj(dp
,
478 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
481 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
482 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
483 DS_FIELD_BOOKMARK_NAMES
,
484 sizeof (ds
->ds_bookmarks
), 1,
486 if (zaperr
!= ENOENT
)
490 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
491 err
= dsl_dataset_get_snapname(ds
);
493 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
495 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
496 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
501 if (err
== 0 && !ds
->ds_is_snapshot
) {
502 err
= dsl_prop_get_int_ds(ds
,
503 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
506 err
= dsl_prop_get_int_ds(ds
,
507 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
511 ds
->ds_reserved
= ds
->ds_quota
= 0;
514 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict
, &ds
->ds_dbuf
);
516 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
518 if (err
!= 0 || winner
!= NULL
) {
519 bplist_destroy(&ds
->ds_pending_deadlist
);
520 dsl_deadlist_close(&ds
->ds_deadlist
);
522 dsl_dataset_rele(ds
->ds_prev
, ds
);
523 dsl_dir_rele(ds
->ds_dir
, ds
);
524 mutex_destroy(&ds
->ds_lock
);
525 mutex_destroy(&ds
->ds_opening_lock
);
526 mutex_destroy(&ds
->ds_sendstream_lock
);
527 refcount_destroy(&ds
->ds_longholds
);
528 kmem_free(ds
, sizeof (dsl_dataset_t
));
530 dmu_buf_rele(dbuf
, tag
);
536 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
539 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
540 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
541 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
542 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
543 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
549 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
,
550 void *tag
, dsl_dataset_t
**dsp
)
553 const char *snapname
;
558 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
562 ASSERT(dsl_pool_config_held(dp
));
563 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
565 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
567 err
= SET_ERROR(ENOENT
);
569 /* we may be looking for a snapshot */
570 if (err
== 0 && snapname
!= NULL
) {
571 dsl_dataset_t
*snap_ds
;
573 if (*snapname
++ != '@') {
574 dsl_dataset_rele(ds
, tag
);
575 dsl_dir_rele(dd
, FTAG
);
576 return (SET_ERROR(ENOENT
));
579 dprintf("looking for snapshot '%s'\n", snapname
);
580 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
582 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &snap_ds
);
583 dsl_dataset_rele(ds
, tag
);
586 mutex_enter(&snap_ds
->ds_lock
);
587 if (snap_ds
->ds_snapname
[0] == 0)
588 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
589 sizeof (snap_ds
->ds_snapname
));
590 mutex_exit(&snap_ds
->ds_lock
);
596 dsl_dir_rele(dd
, FTAG
);
601 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
,
602 void *tag
, dsl_dataset_t
**dsp
)
604 int err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
607 if (!dsl_dataset_tryown(*dsp
, tag
)) {
608 dsl_dataset_rele(*dsp
, tag
);
610 return (SET_ERROR(EBUSY
));
616 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
,
617 void *tag
, dsl_dataset_t
**dsp
)
619 int err
= dsl_dataset_hold(dp
, name
, tag
, dsp
);
622 if (!dsl_dataset_tryown(*dsp
, tag
)) {
623 dsl_dataset_rele(*dsp
, tag
);
624 return (SET_ERROR(EBUSY
));
630 * See the comment above dsl_pool_hold() for details. In summary, a long
631 * hold is used to prevent destruction of a dataset while the pool hold
632 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
634 * The dataset and pool must be held when this function is called. After it
635 * is called, the pool hold may be released while the dataset is still held
639 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
641 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
642 (void) refcount_add(&ds
->ds_longholds
, tag
);
646 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
648 (void) refcount_remove(&ds
->ds_longholds
, tag
);
651 /* Return B_TRUE if there are any long holds on this dataset. */
653 dsl_dataset_long_held(dsl_dataset_t
*ds
)
655 return (!refcount_is_zero(&ds
->ds_longholds
));
659 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
662 (void) strcpy(name
, "mos");
664 dsl_dir_name(ds
->ds_dir
, name
);
665 VERIFY0(dsl_dataset_get_snapname(ds
));
666 if (ds
->ds_snapname
[0]) {
667 (void) strcat(name
, "@");
669 * We use a "recursive" mutex so that we
670 * can call dprintf_ds() with ds_lock held.
672 if (!MUTEX_HELD(&ds
->ds_lock
)) {
673 mutex_enter(&ds
->ds_lock
);
674 (void) strcat(name
, ds
->ds_snapname
);
675 mutex_exit(&ds
->ds_lock
);
677 (void) strcat(name
, ds
->ds_snapname
);
684 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
686 dmu_buf_rele(ds
->ds_dbuf
, tag
);
690 dsl_dataset_disown(dsl_dataset_t
*ds
, void *tag
)
692 ASSERT3P(ds
->ds_owner
, ==, tag
);
693 ASSERT(ds
->ds_dbuf
!= NULL
);
695 mutex_enter(&ds
->ds_lock
);
697 mutex_exit(&ds
->ds_lock
);
698 dsl_dataset_long_rele(ds
, tag
);
699 dsl_dataset_rele(ds
, tag
);
703 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
705 boolean_t gotit
= FALSE
;
707 mutex_enter(&ds
->ds_lock
);
708 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
710 dsl_dataset_long_hold(ds
, tag
);
713 mutex_exit(&ds
->ds_lock
);
718 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
720 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
721 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
724 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
726 spa_feature_incr(spa
, f
, tx
);
727 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
729 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
730 sizeof (zero
), 1, &zero
, tx
));
734 dsl_dataset_deactivate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
736 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
737 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
739 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
741 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
742 spa_feature_decr(spa
, f
, tx
);
746 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
747 uint64_t flags
, dmu_tx_t
*tx
)
749 dsl_pool_t
*dp
= dd
->dd_pool
;
751 dsl_dataset_phys_t
*dsphys
;
753 objset_t
*mos
= dp
->dp_meta_objset
;
756 origin
= dp
->dp_origin_snap
;
758 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
759 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
760 ASSERT(dmu_tx_is_syncing(tx
));
761 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
763 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
764 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
765 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
766 dmu_buf_will_dirty(dbuf
, tx
);
767 dsphys
= dbuf
->db_data
;
768 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
769 dsphys
->ds_dir_obj
= dd
->dd_object
;
770 dsphys
->ds_flags
= flags
;
771 dsphys
->ds_fsid_guid
= unique_create();
772 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
773 sizeof (dsphys
->ds_guid
));
774 dsphys
->ds_snapnames_zapobj
=
775 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
777 dsphys
->ds_creation_time
= gethrestime_sec();
778 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
780 if (origin
== NULL
) {
781 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
784 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
786 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
787 dsphys
->ds_prev_snap_txg
=
788 dsl_dataset_phys(origin
)->ds_creation_txg
;
789 dsphys
->ds_referenced_bytes
=
790 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
791 dsphys
->ds_compressed_bytes
=
792 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
793 dsphys
->ds_uncompressed_bytes
=
794 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
795 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
798 * Inherit flags that describe the dataset's contents
799 * (INCONSISTENT) or properties (Case Insensitive).
801 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
802 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
804 for (f
= 0; f
< SPA_FEATURES
; f
++) {
805 if (origin
->ds_feature_inuse
[f
])
806 dsl_dataset_activate_feature(dsobj
, f
, tx
);
809 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
810 dsl_dataset_phys(origin
)->ds_num_children
++;
812 VERIFY0(dsl_dataset_hold_obj(dp
,
813 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
815 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
816 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
817 dsl_dataset_rele(ohds
, FTAG
);
819 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
820 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
821 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
823 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
825 VERIFY0(zap_add_int(mos
,
826 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
830 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
831 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
832 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
833 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
834 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
835 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
837 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
839 VERIFY0(zap_add_int(mos
,
840 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
845 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
846 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
848 dmu_buf_rele(dbuf
, FTAG
);
850 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
851 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
857 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
861 VERIFY0(dmu_objset_from_ds(ds
, &os
));
862 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
863 dsl_dataset_dirty(ds
, tx
);
867 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
868 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
, dmu_tx_t
*tx
)
870 dsl_pool_t
*dp
= pdd
->dd_pool
;
871 uint64_t dsobj
, ddobj
;
874 ASSERT(dmu_tx_is_syncing(tx
));
875 ASSERT(lastname
[0] != '@');
877 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
878 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
880 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
,
881 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
883 dsl_deleg_set_create_perms(dd
, tx
, cr
);
886 * Since we're creating a new node we know it's a leaf, so we can
887 * initialize the counts if the limit feature is active.
889 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
891 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
893 dsl_dir_zapify(dd
, tx
);
894 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
895 sizeof (cnt
), 1, &cnt
, tx
));
896 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
897 sizeof (cnt
), 1, &cnt
, tx
));
900 dsl_dir_rele(dd
, FTAG
);
903 * If we are creating a clone, make sure we zero out any stale
904 * data from the origin snapshots zil header.
906 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
909 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
910 dsl_dataset_zero_zil(ds
, tx
);
911 dsl_dataset_rele(ds
, FTAG
);
918 * The unique space in the head dataset can be calculated by subtracting
919 * the space used in the most recent snapshot, that is still being used
920 * in this file system, from the space currently in use. To figure out
921 * the space in the most recent snapshot still in use, we need to take
922 * the total space used in the snapshot and subtract out the space that
923 * has been freed up since the snapshot was taken.
926 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
929 uint64_t dlused
, dlcomp
, dluncomp
;
931 ASSERT(!ds
->ds_is_snapshot
);
933 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
934 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
938 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
940 ASSERT3U(dlused
, <=, mrs_used
);
941 dsl_dataset_phys(ds
)->ds_unique_bytes
=
942 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
944 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
945 SPA_VERSION_UNIQUE_ACCURATE
)
946 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
950 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
953 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
955 ASSERTV(uint64_t count
);
957 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
958 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
961 * The err should not be ENOENT, but a bug in a previous version
962 * of the code could cause upgrade_clones_cb() to not set
963 * ds_next_snap_obj when it should, leading to a missing entry.
964 * If we knew that the pool was created after
965 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
966 * ENOENT. However, at least we can check that we don't have
967 * too many entries in the next_clones_obj even after failing to
972 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
974 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
979 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
981 return (&dsl_dataset_phys(ds
)->ds_bp
);
985 dsl_dataset_set_blkptr(dsl_dataset_t
*ds
, blkptr_t
*bp
, dmu_tx_t
*tx
)
987 ASSERT(dmu_tx_is_syncing(tx
));
988 /* If it's the meta-objset, set dp_meta_rootbp */
990 tx
->tx_pool
->dp_meta_rootbp
= *bp
;
992 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
993 dsl_dataset_phys(ds
)->ds_bp
= *bp
;
998 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1000 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1004 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1008 if (ds
== NULL
) /* this is the meta-objset */
1011 ASSERT(ds
->ds_objset
!= NULL
);
1013 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1014 panic("dirtying snapshot!");
1016 dp
= ds
->ds_dir
->dd_pool
;
1018 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1019 /* up the hold count until we can be written out */
1020 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1025 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
1029 for (t
= 0; t
< TXG_SIZE
; t
++) {
1030 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1038 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1042 if (!dmu_tx_is_syncing(tx
))
1046 * If there's an fs-only reservation, any blocks that might become
1047 * owned by the snapshot dataset must be accommodated by space
1048 * outside of the reservation.
1050 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1051 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1052 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1053 return (SET_ERROR(ENOSPC
));
1056 * Propagate any reserved space for this snapshot to other
1057 * snapshot checks in this sync group.
1060 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1065 typedef struct dsl_dataset_snapshot_arg
{
1066 nvlist_t
*ddsa_snaps
;
1067 nvlist_t
*ddsa_props
;
1068 nvlist_t
*ddsa_errors
;
1070 } dsl_dataset_snapshot_arg_t
;
1073 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1074 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1079 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1081 if (!dmu_tx_is_syncing(tx
))
1085 * We don't allow multiple snapshots of the same txg. If there
1086 * is already one, try again.
1088 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1089 return (SET_ERROR(EAGAIN
));
1092 * Check for conflicting snapshot name.
1094 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1096 return (SET_ERROR(EEXIST
));
1097 if (error
!= ENOENT
)
1101 * We don't allow taking snapshots of inconsistent datasets, such as
1102 * those into which we are currently receiving. However, if we are
1103 * creating this snapshot as part of a receive, this check will be
1104 * executed atomically with respect to the completion of the receive
1105 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1106 * case we ignore this, knowing it will be fixed up for us shortly in
1107 * dmu_recv_end_sync().
1109 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1110 return (SET_ERROR(EBUSY
));
1113 * Skip the check for temporary snapshots or if we have already checked
1114 * the counts in dsl_dataset_snapshot_check. This means we really only
1115 * check the count here when we're receiving a stream.
1117 if (cnt
!= 0 && cr
!= NULL
) {
1118 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1119 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1124 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1132 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1134 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1135 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1140 * Pre-compute how many total new snapshots will be created for each
1141 * level in the tree and below. This is needed for validating the
1142 * snapshot limit when either taking a recursive snapshot or when
1143 * taking multiple snapshots.
1145 * The problem is that the counts are not actually adjusted when
1146 * we are checking, only when we finally sync. For a single snapshot,
1147 * this is easy, the count will increase by 1 at each node up the tree,
1148 * but its more complicated for the recursive/multiple snapshot case.
1150 * The dsl_fs_ss_limit_check function does recursively check the count
1151 * at each level up the tree but since it is validating each snapshot
1152 * independently we need to be sure that we are validating the complete
1153 * count for the entire set of snapshots. We do this by rolling up the
1154 * counts for each component of the name into an nvlist and then
1155 * checking each of those cases with the aggregated count.
1157 * This approach properly handles not only the recursive snapshot
1158 * case (where we get all of those on the ddsa_snaps list) but also
1159 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1160 * validate the limit on 'a' using a count of 2).
1162 * We validate the snapshot names in the third loop and only report
1165 if (dmu_tx_is_syncing(tx
)) {
1167 nvlist_t
*cnt_track
= NULL
;
1168 cnt_track
= fnvlist_alloc();
1170 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1172 /* Rollup aggregated counts into the cnt_track list */
1173 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1175 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1179 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1180 pdelim
= strchr(nm
, '@');
1186 if (nvlist_lookup_uint64(cnt_track
, nm
,
1188 /* update existing entry */
1189 fnvlist_add_uint64(cnt_track
, nm
,
1193 fnvlist_add_uint64(cnt_track
, nm
, 1);
1196 pdelim
= strrchr(nm
, '/');
1199 } while (pdelim
!= NULL
);
1202 kmem_free(nm
, MAXPATHLEN
);
1204 /* Check aggregated counts at each level */
1205 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1206 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1212 name
= nvpair_name(pair
);
1213 cnt
= fnvpair_value_uint64(pair
);
1216 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1218 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1219 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1221 dsl_dataset_rele(ds
, FTAG
);
1225 if (ddsa
->ddsa_errors
!= NULL
)
1226 fnvlist_add_int32(ddsa
->ddsa_errors
,
1229 /* only report one error for this check */
1233 nvlist_free(cnt_track
);
1236 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1237 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1241 char dsname
[MAXNAMELEN
];
1243 name
= nvpair_name(pair
);
1244 if (strlen(name
) >= MAXNAMELEN
)
1245 error
= SET_ERROR(ENAMETOOLONG
);
1247 atp
= strchr(name
, '@');
1249 error
= SET_ERROR(EINVAL
);
1251 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1254 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1256 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1257 error
= dsl_dataset_snapshot_check_impl(ds
,
1258 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1259 dsl_dataset_rele(ds
, FTAG
);
1263 if (ddsa
->ddsa_errors
!= NULL
) {
1264 fnvlist_add_int32(ddsa
->ddsa_errors
,
1275 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1278 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1280 dsl_dataset_phys_t
*dsphys
;
1281 uint64_t dsobj
, crtxg
;
1282 objset_t
*mos
= dp
->dp_meta_objset
;
1284 ASSERTV(static zil_header_t zero_zil
);
1285 ASSERTV(objset_t
*os
);
1287 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1290 * If we are on an old pool, the zil must not be active, in which
1291 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1293 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1294 dmu_objset_from_ds(ds
, &os
) != 0 ||
1295 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1296 sizeof (zero_zil
)) == 0);
1298 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1301 * The origin's ds_creation_txg has to be < TXG_INITIAL
1303 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1308 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1309 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1310 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1311 dmu_buf_will_dirty(dbuf
, tx
);
1312 dsphys
= dbuf
->db_data
;
1313 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1314 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1315 dsphys
->ds_fsid_guid
= unique_create();
1316 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1317 sizeof (dsphys
->ds_guid
));
1318 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1319 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1320 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1321 dsphys
->ds_num_children
= 1;
1322 dsphys
->ds_creation_time
= gethrestime_sec();
1323 dsphys
->ds_creation_txg
= crtxg
;
1324 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1325 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1326 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1327 dsphys
->ds_uncompressed_bytes
=
1328 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1329 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1330 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1331 dmu_buf_rele(dbuf
, FTAG
);
1333 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1334 if (ds
->ds_feature_inuse
[f
])
1335 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1338 ASSERT3U(ds
->ds_prev
!= 0, ==,
1339 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1341 uint64_t next_clones_obj
=
1342 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1343 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1345 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1346 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1348 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1349 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1350 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1351 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1352 } else if (next_clones_obj
!= 0) {
1353 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1354 dsphys
->ds_next_snap_obj
, tx
);
1355 VERIFY0(zap_add_int(mos
,
1356 next_clones_obj
, dsobj
, tx
));
1361 * If we have a reference-reservation on this dataset, we will
1362 * need to increase the amount of refreservation being charged
1363 * since our unique space is going to zero.
1365 if (ds
->ds_reserved
) {
1367 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1368 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1370 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1374 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1375 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1376 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1377 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1378 dsl_deadlist_close(&ds
->ds_deadlist
);
1379 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1380 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1381 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1382 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1384 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1385 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1386 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1387 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1388 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1389 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1391 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1392 snapname
, 8, 1, &dsobj
, tx
));
1395 dsl_dataset_rele(ds
->ds_prev
, ds
);
1396 VERIFY0(dsl_dataset_hold_obj(dp
,
1397 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1399 dsl_scan_ds_snapshotted(ds
, tx
);
1401 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1403 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1407 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1409 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1410 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1413 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1414 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1417 char dsname
[MAXNAMELEN
];
1419 name
= nvpair_name(pair
);
1420 atp
= strchr(name
, '@');
1421 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1422 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1424 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1425 if (ddsa
->ddsa_props
!= NULL
) {
1426 dsl_props_set_sync_impl(ds
->ds_prev
,
1427 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1429 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1430 dsl_dataset_rele(ds
, FTAG
);
1435 * The snapshots must all be in the same pool.
1436 * All-or-nothing: if there are any failures, nothing will be modified.
1439 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1441 dsl_dataset_snapshot_arg_t ddsa
;
1443 boolean_t needsuspend
;
1447 nvlist_t
*suspended
= NULL
;
1449 pair
= nvlist_next_nvpair(snaps
, NULL
);
1452 firstname
= nvpair_name(pair
);
1454 error
= spa_open(firstname
, &spa
, FTAG
);
1457 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1458 spa_close(spa
, FTAG
);
1461 suspended
= fnvlist_alloc();
1462 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1463 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1464 char fsname
[MAXNAMELEN
];
1465 char *snapname
= nvpair_name(pair
);
1469 atp
= strchr(snapname
, '@');
1471 error
= SET_ERROR(EINVAL
);
1474 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1476 error
= zil_suspend(fsname
, &cookie
);
1479 fnvlist_add_uint64(suspended
, fsname
,
1484 ddsa
.ddsa_snaps
= snaps
;
1485 ddsa
.ddsa_props
= props
;
1486 ddsa
.ddsa_errors
= errors
;
1487 ddsa
.ddsa_cr
= CRED();
1490 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1491 dsl_dataset_snapshot_sync
, &ddsa
,
1492 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1495 if (suspended
!= NULL
) {
1496 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1497 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1498 zil_resume((void *)(uintptr_t)
1499 fnvpair_value_uint64(pair
));
1501 fnvlist_free(suspended
);
1507 typedef struct dsl_dataset_snapshot_tmp_arg
{
1508 const char *ddsta_fsname
;
1509 const char *ddsta_snapname
;
1510 minor_t ddsta_cleanup_minor
;
1511 const char *ddsta_htag
;
1512 } dsl_dataset_snapshot_tmp_arg_t
;
1515 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1517 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1518 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1522 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1526 /* NULL cred means no limit check for tmp snapshot */
1527 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1528 tx
, B_FALSE
, 0, NULL
);
1530 dsl_dataset_rele(ds
, FTAG
);
1534 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1535 dsl_dataset_rele(ds
, FTAG
);
1536 return (SET_ERROR(ENOTSUP
));
1538 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1541 dsl_dataset_rele(ds
, FTAG
);
1545 dsl_dataset_rele(ds
, FTAG
);
1550 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1552 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1553 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1556 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1558 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1559 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1560 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1561 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1563 dsl_dataset_rele(ds
, FTAG
);
1567 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1568 minor_t cleanup_minor
, const char *htag
)
1570 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1573 boolean_t needsuspend
;
1576 ddsta
.ddsta_fsname
= fsname
;
1577 ddsta
.ddsta_snapname
= snapname
;
1578 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1579 ddsta
.ddsta_htag
= htag
;
1581 error
= spa_open(fsname
, &spa
, FTAG
);
1584 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1585 spa_close(spa
, FTAG
);
1588 error
= zil_suspend(fsname
, &cookie
);
1593 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1594 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1603 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1607 ASSERT(dmu_tx_is_syncing(tx
));
1608 ASSERT(ds
->ds_objset
!= NULL
);
1609 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1612 * in case we had to change ds_fsid_guid when we opened it,
1615 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1616 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1618 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1620 for (f
= 0; f
< SPA_FEATURES
; f
++) {
1621 if (ds
->ds_feature_activation_needed
[f
]) {
1622 if (ds
->ds_feature_inuse
[f
])
1624 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1625 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1631 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1634 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1637 nvlist_t
*propval
= fnvlist_alloc();
1638 nvlist_t
*val
= fnvlist_alloc();
1640 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1643 * There may be missing entries in ds_next_clones_obj
1644 * due to a bug in a previous version of the code.
1645 * Only trust it if it has the right number of entries.
1647 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1648 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1651 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1)
1653 for (zap_cursor_init(&zc
, mos
,
1654 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1655 zap_cursor_retrieve(&zc
, &za
) == 0;
1656 zap_cursor_advance(&zc
)) {
1657 dsl_dataset_t
*clone
;
1658 char buf
[ZFS_MAXNAMELEN
];
1659 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1660 za
.za_first_integer
, FTAG
, &clone
));
1661 dsl_dir_name(clone
->ds_dir
, buf
);
1662 fnvlist_add_boolean(val
, buf
);
1663 dsl_dataset_rele(clone
, FTAG
);
1665 zap_cursor_fini(&zc
);
1666 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1667 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
), propval
);
1670 nvlist_free(propval
);
1674 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1676 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1677 uint64_t refd
, avail
, uobjs
, aobjs
, ratio
;
1679 ASSERT(dsl_pool_config_held(dp
));
1681 ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1682 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1683 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1685 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
, ratio
);
1686 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
1687 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1689 if (ds
->ds_is_snapshot
) {
1690 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
, ratio
);
1691 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
1692 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1693 get_clones_stat(ds
, nv
);
1695 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
1696 char buf
[MAXNAMELEN
];
1697 dsl_dataset_name(ds
->ds_prev
, buf
);
1698 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
, buf
);
1701 dsl_dir_stats(ds
->ds_dir
, nv
);
1704 dsl_dataset_space(ds
, &refd
, &avail
, &uobjs
, &aobjs
);
1705 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
, avail
);
1706 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
, refd
);
1708 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
1709 dsl_dataset_phys(ds
)->ds_creation_time
);
1710 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
1711 dsl_dataset_phys(ds
)->ds_creation_txg
);
1712 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
1714 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
1716 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
1717 dsl_dataset_phys(ds
)->ds_guid
);
1718 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
1719 dsl_dataset_phys(ds
)->ds_unique_bytes
);
1720 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
1722 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
1724 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
1725 DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
1727 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
1728 uint64_t written
, comp
, uncomp
;
1729 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1730 dsl_dataset_t
*prev
;
1733 err
= dsl_dataset_hold_obj(dp
,
1734 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
1736 err
= dsl_dataset_space_written(prev
, ds
, &written
,
1738 dsl_dataset_rele(prev
, FTAG
);
1740 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
1749 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
1751 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1752 ASSERT(dsl_pool_config_held(dp
));
1754 stat
->dds_creation_txg
= dsl_dataset_phys(ds
)->ds_creation_txg
;
1755 stat
->dds_inconsistent
=
1756 dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
;
1757 stat
->dds_guid
= dsl_dataset_phys(ds
)->ds_guid
;
1758 stat
->dds_origin
[0] = '\0';
1759 if (ds
->ds_is_snapshot
) {
1760 stat
->dds_is_snapshot
= B_TRUE
;
1761 stat
->dds_num_clones
=
1762 dsl_dataset_phys(ds
)->ds_num_children
- 1;
1764 stat
->dds_is_snapshot
= B_FALSE
;
1765 stat
->dds_num_clones
= 0;
1767 if (dsl_dir_is_clone(ds
->ds_dir
)) {
1770 VERIFY0(dsl_dataset_hold_obj(dp
,
1771 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
,
1773 dsl_dataset_name(ods
, stat
->dds_origin
);
1774 dsl_dataset_rele(ods
, FTAG
);
1780 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
1782 return (ds
->ds_fsid_guid
);
1786 dsl_dataset_space(dsl_dataset_t
*ds
,
1787 uint64_t *refdbytesp
, uint64_t *availbytesp
,
1788 uint64_t *usedobjsp
, uint64_t *availobjsp
)
1790 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1791 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
1792 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
1794 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
1795 if (ds
->ds_quota
!= 0) {
1797 * Adjust available bytes according to refquota
1799 if (*refdbytesp
< ds
->ds_quota
)
1800 *availbytesp
= MIN(*availbytesp
,
1801 ds
->ds_quota
- *refdbytesp
);
1805 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
1806 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
1810 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
1812 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1815 if (dsl_dataset_phys(ds
)->ds_bp
.blk_birth
>
1816 dsl_dataset_phys(snap
)->ds_creation_txg
) {
1817 objset_t
*os
, *os_snap
;
1819 * It may be that only the ZIL differs, because it was
1820 * reset in the head. Don't count that as being
1823 if (dmu_objset_from_ds(ds
, &os
) != 0)
1825 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
1827 return (bcmp(&os
->os_phys
->os_meta_dnode
,
1828 &os_snap
->os_phys
->os_meta_dnode
,
1829 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
1834 typedef struct dsl_dataset_rename_snapshot_arg
{
1835 const char *ddrsa_fsname
;
1836 const char *ddrsa_oldsnapname
;
1837 const char *ddrsa_newsnapname
;
1838 boolean_t ddrsa_recursive
;
1840 } dsl_dataset_rename_snapshot_arg_t
;
1844 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
1845 dsl_dataset_t
*hds
, void *arg
)
1847 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1851 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1853 /* ignore nonexistent snapshots */
1854 return (error
== ENOENT
? 0 : error
);
1857 /* new name should not exist */
1858 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
1860 error
= SET_ERROR(EEXIST
);
1861 else if (error
== ENOENT
)
1864 /* dataset name + 1 for the "@" + the new snapshot name must fit */
1865 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
1866 strlen(ddrsa
->ddrsa_newsnapname
) >= MAXNAMELEN
)
1867 error
= SET_ERROR(ENAMETOOLONG
);
1873 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1875 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1876 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1880 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
1884 if (ddrsa
->ddrsa_recursive
) {
1885 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1886 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
1889 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
1891 dsl_dataset_rele(hds
, FTAG
);
1896 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
1897 dsl_dataset_t
*hds
, void *arg
)
1899 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1902 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
1905 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
1906 ASSERT(error
== 0 || error
== ENOENT
);
1907 if (error
== ENOENT
) {
1908 /* ignore nonexistent snapshots */
1912 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
1914 /* log before we change the name */
1915 spa_history_log_internal_ds(ds
, "rename", tx
,
1916 "-> @%s", ddrsa
->ddrsa_newsnapname
);
1918 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
1920 mutex_enter(&ds
->ds_lock
);
1921 (void) strcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
);
1922 mutex_exit(&ds
->ds_lock
);
1923 VERIFY0(zap_add(dp
->dp_meta_objset
,
1924 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
1925 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
1926 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
1927 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
1929 dsl_dataset_rele(ds
, FTAG
);
1934 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1936 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
1937 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1940 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
1941 ddrsa
->ddrsa_tx
= tx
;
1942 if (ddrsa
->ddrsa_recursive
) {
1943 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
1944 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
1947 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
1949 dsl_dataset_rele(hds
, FTAG
);
1953 dsl_dataset_rename_snapshot(const char *fsname
,
1954 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
1956 dsl_dataset_rename_snapshot_arg_t ddrsa
;
1958 ddrsa
.ddrsa_fsname
= fsname
;
1959 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
1960 ddrsa
.ddrsa_newsnapname
= newsnapname
;
1961 ddrsa
.ddrsa_recursive
= recursive
;
1963 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
1964 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
1965 1, ZFS_SPACE_CHECK_RESERVED
));
1969 * If we're doing an ownership handoff, we need to make sure that there is
1970 * only one long hold on the dataset. We're not allowed to change anything here
1971 * so we don't permanently release the long hold or regular hold here. We want
1972 * to do this only when syncing to avoid the dataset unexpectedly going away
1973 * when we release the long hold.
1976 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
1980 if (!dmu_tx_is_syncing(tx
))
1983 if (owner
!= NULL
) {
1984 VERIFY3P(ds
->ds_owner
, ==, owner
);
1985 dsl_dataset_long_rele(ds
, owner
);
1988 held
= dsl_dataset_long_held(ds
);
1991 dsl_dataset_long_hold(ds
, owner
);
1994 return (SET_ERROR(EBUSY
));
1999 typedef struct dsl_dataset_rollback_arg
{
2000 const char *ddra_fsname
;
2002 nvlist_t
*ddra_result
;
2003 } dsl_dataset_rollback_arg_t
;
2006 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2008 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2009 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2011 int64_t unused_refres_delta
;
2014 nvlist_t
*proprequest
, *bookmarks
;
2016 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2020 /* must not be a snapshot */
2021 if (ds
->ds_is_snapshot
) {
2022 dsl_dataset_rele(ds
, FTAG
);
2023 return (SET_ERROR(EINVAL
));
2026 /* must have a most recent snapshot */
2027 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2028 dsl_dataset_rele(ds
, FTAG
);
2029 return (SET_ERROR(EINVAL
));
2032 /* must not have any bookmarks after the most recent snapshot */
2033 proprequest
= fnvlist_alloc();
2034 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2035 bookmarks
= fnvlist_alloc();
2036 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2037 fnvlist_free(proprequest
);
2040 for (pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2041 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2043 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2044 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2045 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2046 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2047 fnvlist_free(bookmarks
);
2048 dsl_dataset_rele(ds
, FTAG
);
2049 return (SET_ERROR(EEXIST
));
2052 fnvlist_free(bookmarks
);
2054 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2056 dsl_dataset_rele(ds
, FTAG
);
2061 * Check if the snap we are rolling back to uses more than
2064 if (ds
->ds_quota
!= 0 &&
2065 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2066 dsl_dataset_rele(ds
, FTAG
);
2067 return (SET_ERROR(EDQUOT
));
2071 * When we do the clone swap, we will temporarily use more space
2072 * due to the refreservation (the head will no longer have any
2073 * unique space, so the entire amount of the refreservation will need
2074 * to be free). We will immediately destroy the clone, freeing
2075 * this space, but the freeing happens over many txg's.
2077 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2078 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2080 if (unused_refres_delta
> 0 &&
2081 unused_refres_delta
>
2082 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2083 dsl_dataset_rele(ds
, FTAG
);
2084 return (SET_ERROR(ENOSPC
));
2087 dsl_dataset_rele(ds
, FTAG
);
2092 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2094 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2095 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2096 dsl_dataset_t
*ds
, *clone
;
2098 char namebuf
[ZFS_MAXNAMELEN
];
2100 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2102 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2103 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2105 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2106 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, tx
);
2108 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2110 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2111 dsl_dataset_zero_zil(ds
, tx
);
2113 dsl_destroy_head_sync_impl(clone
, tx
);
2115 dsl_dataset_rele(clone
, FTAG
);
2116 dsl_dataset_rele(ds
, FTAG
);
2120 * Rolls back the given filesystem or volume to the most recent snapshot.
2121 * The name of the most recent snapshot will be returned under key "target"
2122 * in the result nvlist.
2125 * - The existing dataset MUST be owned by the specified owner at entry
2126 * - Upon return, dataset will still be held by the same owner, whether we
2129 * This mode is required any time the existing filesystem is mounted. See
2130 * notes above zfs_suspend_fs() for further details.
2133 dsl_dataset_rollback(const char *fsname
, void *owner
, nvlist_t
*result
)
2135 dsl_dataset_rollback_arg_t ddra
;
2137 ddra
.ddra_fsname
= fsname
;
2138 ddra
.ddra_owner
= owner
;
2139 ddra
.ddra_result
= result
;
2141 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2142 dsl_dataset_rollback_sync
, &ddra
,
2143 1, ZFS_SPACE_CHECK_RESERVED
));
2146 struct promotenode
{
2151 typedef struct dsl_dataset_promote_arg
{
2152 const char *ddpa_clonename
;
2153 dsl_dataset_t
*ddpa_clone
;
2154 list_t shared_snaps
, origin_snaps
, clone_snaps
;
2155 dsl_dataset_t
*origin_origin
; /* origin of the origin */
2156 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
2159 } dsl_dataset_promote_arg_t
;
2161 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2162 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2164 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2167 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2169 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2170 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2172 struct promotenode
*snap
;
2173 dsl_dataset_t
*origin_ds
;
2177 size_t max_snap_len
;
2179 err
= promote_hold(ddpa
, dp
, FTAG
);
2183 hds
= ddpa
->ddpa_clone
;
2184 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2186 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2187 promote_rele(ddpa
, FTAG
);
2188 return (SET_ERROR(EXDEV
));
2192 * Compute and check the amount of space to transfer. Since this is
2193 * so expensive, don't do the preliminary check.
2195 if (!dmu_tx_is_syncing(tx
)) {
2196 promote_rele(ddpa
, FTAG
);
2200 snap
= list_head(&ddpa
->shared_snaps
);
2201 origin_ds
= snap
->ds
;
2203 /* compute origin's new unique space */
2204 snap
= list_tail(&ddpa
->clone_snaps
);
2205 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2206 origin_ds
->ds_object
);
2207 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2208 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2209 &ddpa
->unique
, &unused
, &unused
);
2212 * Walk the snapshots that we are moving
2214 * Compute space to transfer. Consider the incremental changes
2215 * to used by each snapshot:
2216 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2217 * So each snapshot gave birth to:
2218 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2219 * So a sequence would look like:
2220 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2221 * Which simplifies to:
2222 * uN + kN + kN-1 + ... + k1 + k0
2223 * Note however, if we stop before we reach the ORIGIN we get:
2224 * uN + kN + kN-1 + ... + kM - uM-1
2227 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2228 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2229 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2230 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2231 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2232 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2233 dsl_dataset_t
*ds
= snap
->ds
;
2238 * If there are long holds, we won't be able to evict
2241 if (dsl_dataset_long_held(ds
)) {
2242 err
= SET_ERROR(EBUSY
);
2246 /* Check that the snapshot name does not conflict */
2247 VERIFY0(dsl_dataset_get_snapname(ds
));
2248 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2249 err
= SET_ERROR(ENAMETOOLONG
);
2252 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2254 (void) strcpy(ddpa
->err_ds
, snap
->ds
->ds_snapname
);
2255 err
= SET_ERROR(EEXIST
);
2261 /* The very first snapshot does not have a deadlist */
2262 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2265 dsl_deadlist_space(&ds
->ds_deadlist
,
2266 &dlused
, &dlcomp
, &dluncomp
);
2267 ddpa
->used
+= dlused
;
2268 ddpa
->comp
+= dlcomp
;
2269 ddpa
->uncomp
+= dluncomp
;
2273 * If we are a clone of a clone then we never reached ORIGIN,
2274 * so we need to subtract out the clone origin's used space.
2276 if (ddpa
->origin_origin
) {
2278 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2280 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2282 dsl_dataset_phys(ddpa
->origin_origin
)->
2283 ds_uncompressed_bytes
;
2286 /* Check that there is enough space and limit headroom here */
2287 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2288 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2293 * Compute the amounts of space that will be used by snapshots
2294 * after the promotion (for both origin and clone). For each,
2295 * it is the amount of space that will be on all of their
2296 * deadlists (that was not born before their new origin).
2298 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2302 * Note, typically this will not be a clone of a clone,
2303 * so dd_origin_txg will be < TXG_INITIAL, so
2304 * these snaplist_space() -> dsl_deadlist_space_range()
2305 * calls will be fast because they do not have to
2306 * iterate over all bps.
2308 snap
= list_head(&ddpa
->origin_snaps
);
2309 err
= snaplist_space(&ddpa
->shared_snaps
,
2310 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2314 err
= snaplist_space(&ddpa
->clone_snaps
,
2315 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2318 ddpa
->cloneusedsnap
+= space
;
2320 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2321 DD_FLAG_USED_BREAKDOWN
) {
2322 err
= snaplist_space(&ddpa
->origin_snaps
,
2323 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2324 &ddpa
->originusedsnap
);
2330 promote_rele(ddpa
, FTAG
);
2335 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2337 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2338 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2340 struct promotenode
*snap
;
2341 dsl_dataset_t
*origin_ds
;
2342 dsl_dataset_t
*origin_head
;
2344 dsl_dir_t
*odd
= NULL
;
2345 uint64_t oldnext_obj
;
2348 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2349 hds
= ddpa
->ddpa_clone
;
2351 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2353 snap
= list_head(&ddpa
->shared_snaps
);
2354 origin_ds
= snap
->ds
;
2357 snap
= list_head(&ddpa
->origin_snaps
);
2358 origin_head
= snap
->ds
;
2361 * We need to explicitly open odd, since origin_ds's dd will be
2364 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2367 /* change origin's next snap */
2368 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2369 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2370 snap
= list_tail(&ddpa
->clone_snaps
);
2371 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2372 origin_ds
->ds_object
);
2373 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2375 /* change the origin's next clone */
2376 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2377 dsl_dataset_remove_from_next_clones(origin_ds
,
2378 snap
->ds
->ds_object
, tx
);
2379 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2380 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2385 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2386 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2387 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2388 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2389 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2390 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2391 origin_head
->ds_dir
->dd_origin_txg
=
2392 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2394 /* change dd_clone entries */
2395 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2396 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2397 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2398 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2399 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2400 hds
->ds_object
, tx
));
2402 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2403 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2404 origin_head
->ds_object
, tx
));
2405 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2406 dsl_dir_phys(dd
)->dd_clones
=
2407 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2408 DMU_OT_NONE
, 0, tx
);
2410 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2411 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2414 /* move snapshots to this dir */
2415 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2416 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2417 dsl_dataset_t
*ds
= snap
->ds
;
2420 * Property callbacks are registered to a particular
2421 * dsl_dir. Since ours is changing, evict the objset
2422 * so that they will be unregistered from the old dsl_dir.
2424 if (ds
->ds_objset
) {
2425 dmu_objset_evict(ds
->ds_objset
);
2426 ds
->ds_objset
= NULL
;
2429 /* move snap name entry */
2430 VERIFY0(dsl_dataset_get_snapname(ds
));
2431 VERIFY0(dsl_dataset_snap_remove(origin_head
,
2432 ds
->ds_snapname
, tx
, B_TRUE
));
2433 VERIFY0(zap_add(dp
->dp_meta_objset
,
2434 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
2435 8, 1, &ds
->ds_object
, tx
));
2436 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
2437 DD_FIELD_SNAPSHOT_COUNT
, tx
);
2439 /* change containing dsl_dir */
2440 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2441 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
2442 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
2443 ASSERT3P(ds
->ds_dir
, ==, odd
);
2444 dsl_dir_rele(ds
->ds_dir
, ds
);
2445 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
2446 NULL
, ds
, &ds
->ds_dir
));
2448 /* move any clone references */
2449 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
2450 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2454 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
2455 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2456 zap_cursor_retrieve(&zc
, &za
) == 0;
2457 zap_cursor_advance(&zc
)) {
2458 dsl_dataset_t
*cnds
;
2461 if (za
.za_first_integer
== oldnext_obj
) {
2463 * We've already moved the
2464 * origin's reference.
2469 VERIFY0(dsl_dataset_hold_obj(dp
,
2470 za
.za_first_integer
, FTAG
, &cnds
));
2471 o
= dsl_dir_phys(cnds
->ds_dir
)->
2472 dd_head_dataset_obj
;
2474 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2475 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
2476 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2477 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
2478 dsl_dataset_rele(cnds
, FTAG
);
2480 zap_cursor_fini(&zc
);
2483 ASSERT(!dsl_prop_hascb(ds
));
2487 * Change space accounting.
2488 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2489 * both be valid, or both be 0 (resulting in delta == 0). This
2490 * is true for each of {clone,origin} independently.
2493 delta
= ddpa
->cloneusedsnap
-
2494 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
2495 ASSERT3S(delta
, >=, 0);
2496 ASSERT3U(ddpa
->used
, >=, delta
);
2497 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2498 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
2499 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
2501 delta
= ddpa
->originusedsnap
-
2502 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
2503 ASSERT3S(delta
, <=, 0);
2504 ASSERT3U(ddpa
->used
, >=, -delta
);
2505 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
2506 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
2507 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
2509 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
2511 /* log history record */
2512 spa_history_log_internal_ds(hds
, "promote", tx
, "");
2514 dsl_dir_rele(odd
, FTAG
);
2515 promote_rele(ddpa
, FTAG
);
2519 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2520 * (exclusive) and last_obj (inclusive). The list will be in reverse
2521 * order (last_obj will be the list_head()). If first_obj == 0, do all
2522 * snapshots back to this dataset's origin.
2525 snaplist_make(dsl_pool_t
*dp
,
2526 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
2528 uint64_t obj
= last_obj
;
2530 list_create(l
, sizeof (struct promotenode
),
2531 offsetof(struct promotenode
, link
));
2533 while (obj
!= first_obj
) {
2535 struct promotenode
*snap
;
2538 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
2539 ASSERT(err
!= ENOENT
);
2544 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
2546 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
2548 list_insert_tail(l
, snap
);
2549 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
2556 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
2558 struct promotenode
*snap
;
2561 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
2562 uint64_t used
, comp
, uncomp
;
2563 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2564 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
2571 snaplist_destroy(list_t
*l
, void *tag
)
2573 struct promotenode
*snap
;
2575 if (l
== NULL
|| !list_link_active(&l
->list_head
))
2578 while ((snap
= list_tail(l
)) != NULL
) {
2579 list_remove(l
, snap
);
2580 dsl_dataset_rele(snap
->ds
, tag
);
2581 kmem_free(snap
, sizeof (*snap
));
2587 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
2591 struct promotenode
*snap
;
2593 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
2597 dd
= ddpa
->ddpa_clone
->ds_dir
;
2599 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
2600 !dsl_dir_is_clone(dd
)) {
2601 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2602 return (SET_ERROR(EINVAL
));
2605 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
2606 &ddpa
->shared_snaps
, tag
);
2610 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
2611 &ddpa
->clone_snaps
, tag
);
2615 snap
= list_head(&ddpa
->shared_snaps
);
2616 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
2617 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
2618 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
2619 &ddpa
->origin_snaps
, tag
);
2623 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
2624 error
= dsl_dataset_hold_obj(dp
,
2625 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
2626 tag
, &ddpa
->origin_origin
);
2632 promote_rele(ddpa
, tag
);
2637 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
2639 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
2640 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
2641 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
2642 if (ddpa
->origin_origin
!= NULL
)
2643 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
2644 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
2650 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2651 * in with the name. (It must be at least MAXNAMELEN bytes long.)
2654 dsl_dataset_promote(const char *name
, char *conflsnap
)
2656 dsl_dataset_promote_arg_t ddpa
= { 0 };
2662 * We will modify space proportional to the number of
2663 * snapshots. Compute numsnaps.
2665 error
= dmu_objset_hold(name
, FTAG
, &os
);
2668 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
2669 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
2671 dmu_objset_rele(os
, FTAG
);
2675 ddpa
.ddpa_clonename
= name
;
2676 ddpa
.err_ds
= conflsnap
;
2679 return (dsl_sync_task(name
, dsl_dataset_promote_check
,
2680 dsl_dataset_promote_sync
, &ddpa
,
2681 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
));
2685 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
2686 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
2688 int64_t unused_refres_delta
;
2690 /* they should both be heads */
2691 if (clone
->ds_is_snapshot
||
2692 origin_head
->ds_is_snapshot
)
2693 return (SET_ERROR(EINVAL
));
2695 /* if we are not forcing, the branch point should be just before them */
2696 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
2697 return (SET_ERROR(EINVAL
));
2699 /* clone should be the clone (unless they are unrelated) */
2700 if (clone
->ds_prev
!= NULL
&&
2701 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
2702 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
2703 return (SET_ERROR(EINVAL
));
2705 /* the clone should be a child of the origin */
2706 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
2707 return (SET_ERROR(EINVAL
));
2709 /* origin_head shouldn't be modified unless 'force' */
2711 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
2712 return (SET_ERROR(ETXTBSY
));
2714 /* origin_head should have no long holds (e.g. is not mounted) */
2715 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
2716 return (SET_ERROR(EBUSY
));
2718 /* check amount of any unconsumed refreservation */
2719 unused_refres_delta
=
2720 (int64_t)MIN(origin_head
->ds_reserved
,
2721 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2722 (int64_t)MIN(origin_head
->ds_reserved
,
2723 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2725 if (unused_refres_delta
> 0 &&
2726 unused_refres_delta
>
2727 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
2728 return (SET_ERROR(ENOSPC
));
2730 /* clone can't be over the head's refquota */
2731 if (origin_head
->ds_quota
!= 0 &&
2732 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
2733 origin_head
->ds_quota
)
2734 return (SET_ERROR(EDQUOT
));
2740 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
2741 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
2744 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2745 int64_t unused_refres_delta
;
2747 ASSERT(clone
->ds_reserved
== 0);
2748 ASSERT(origin_head
->ds_quota
== 0 ||
2749 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
);
2750 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
2753 * Swap per-dataset feature flags.
2755 for (f
= 0; f
< SPA_FEATURES
; f
++) {
2756 boolean_t clone_inuse
;
2757 boolean_t origin_head_inuse
;
2759 if (!(spa_feature_table
[f
].fi_flags
&
2760 ZFEATURE_FLAG_PER_DATASET
)) {
2761 ASSERT(!clone
->ds_feature_inuse
[f
]);
2762 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
2766 clone_inuse
= clone
->ds_feature_inuse
[f
];
2767 origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
2770 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
2771 clone
->ds_feature_inuse
[f
] = B_FALSE
;
2773 if (origin_head_inuse
) {
2774 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
2776 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
2779 dsl_dataset_activate_feature(origin_head
->ds_object
,
2781 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
2783 if (origin_head_inuse
) {
2784 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
2785 clone
->ds_feature_inuse
[f
] = B_TRUE
;
2789 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
2790 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
2792 if (clone
->ds_objset
!= NULL
) {
2793 dmu_objset_evict(clone
->ds_objset
);
2794 clone
->ds_objset
= NULL
;
2797 if (origin_head
->ds_objset
!= NULL
) {
2798 dmu_objset_evict(origin_head
->ds_objset
);
2799 origin_head
->ds_objset
= NULL
;
2802 unused_refres_delta
=
2803 (int64_t)MIN(origin_head
->ds_reserved
,
2804 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
2805 (int64_t)MIN(origin_head
->ds_reserved
,
2806 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2809 * Reset origin's unique bytes, if it exists.
2811 if (clone
->ds_prev
) {
2812 dsl_dataset_t
*origin
= clone
->ds_prev
;
2813 uint64_t comp
, uncomp
;
2815 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
2816 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2817 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
2818 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
2824 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
2825 dsl_dataset_phys(origin_head
)->ds_bp
=
2826 dsl_dataset_phys(clone
)->ds_bp
;
2827 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
2830 /* set dd_*_bytes */
2832 int64_t dused
, dcomp
, duncomp
;
2833 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
2834 uint64_t odl_used
, odl_comp
, odl_uncomp
;
2836 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
2837 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
2839 dsl_deadlist_space(&clone
->ds_deadlist
,
2840 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2841 dsl_deadlist_space(&origin_head
->ds_deadlist
,
2842 &odl_used
, &odl_comp
, &odl_uncomp
);
2844 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
2846 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
2848 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
2850 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
2852 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
2854 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
2857 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
2858 dused
, dcomp
, duncomp
, tx
);
2859 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
2860 -dused
, -dcomp
, -duncomp
, tx
);
2863 * The difference in the space used by snapshots is the
2864 * difference in snapshot space due to the head's
2865 * deadlist (since that's the only thing that's
2866 * changing that affects the snapused).
2868 dsl_deadlist_space_range(&clone
->ds_deadlist
,
2869 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2870 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
2871 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
2872 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
2873 &odl_used
, &odl_comp
, &odl_uncomp
);
2874 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
2875 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
2878 /* swap ds_*_bytes */
2879 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
2880 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
2881 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
2882 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
2883 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
2884 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
2885 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
2886 dsl_dataset_phys(clone
)->ds_unique_bytes
);
2888 /* apply any parent delta for change in unconsumed refreservation */
2889 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
2890 unused_refres_delta
, 0, 0, tx
);
2895 dsl_deadlist_close(&clone
->ds_deadlist
);
2896 dsl_deadlist_close(&origin_head
->ds_deadlist
);
2897 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
2898 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2899 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
2900 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
2901 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
2902 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
2904 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
2906 spa_history_log_internal_ds(clone
, "clone swap", tx
,
2907 "parent=%s", origin_head
->ds_dir
->dd_myname
);
2911 * Given a pool name and a dataset object number in that pool,
2912 * return the name of that dataset.
2915 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
2921 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
2925 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
2927 dsl_dataset_name(ds
, buf
);
2928 dsl_dataset_rele(ds
, FTAG
);
2930 dsl_pool_rele(dp
, FTAG
);
2936 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
2937 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
2941 ASSERT3S(asize
, >, 0);
2944 * *ref_rsrv is the portion of asize that will come from any
2945 * unconsumed refreservation space.
2949 mutex_enter(&ds
->ds_lock
);
2951 * Make a space adjustment for reserved bytes.
2953 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2955 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2957 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
2959 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
2962 if (!check_quota
|| ds
->ds_quota
== 0) {
2963 mutex_exit(&ds
->ds_lock
);
2967 * If they are requesting more space, and our current estimate
2968 * is over quota, they get to try again unless the actual
2969 * on-disk is over quota and there are no pending changes (which
2970 * may free up space for us).
2972 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
2975 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
2976 error
= SET_ERROR(ERESTART
);
2978 error
= SET_ERROR(EDQUOT
);
2980 mutex_exit(&ds
->ds_lock
);
2985 typedef struct dsl_dataset_set_qr_arg
{
2986 const char *ddsqra_name
;
2987 zprop_source_t ddsqra_source
;
2988 uint64_t ddsqra_value
;
2989 } dsl_dataset_set_qr_arg_t
;
2994 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
2996 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
2997 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3002 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3003 return (SET_ERROR(ENOTSUP
));
3005 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3009 if (ds
->ds_is_snapshot
) {
3010 dsl_dataset_rele(ds
, FTAG
);
3011 return (SET_ERROR(EINVAL
));
3014 error
= dsl_prop_predict(ds
->ds_dir
,
3015 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3016 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3018 dsl_dataset_rele(ds
, FTAG
);
3023 dsl_dataset_rele(ds
, FTAG
);
3027 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3028 newval
< ds
->ds_reserved
) {
3029 dsl_dataset_rele(ds
, FTAG
);
3030 return (SET_ERROR(ENOSPC
));
3033 dsl_dataset_rele(ds
, FTAG
);
3038 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3040 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3041 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3045 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3047 dsl_prop_set_sync_impl(ds
,
3048 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3049 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3050 &ddsqra
->ddsqra_value
, tx
);
3052 VERIFY0(dsl_prop_get_int_ds(ds
,
3053 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3055 if (ds
->ds_quota
!= newval
) {
3056 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3057 ds
->ds_quota
= newval
;
3059 dsl_dataset_rele(ds
, FTAG
);
3063 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3066 dsl_dataset_set_qr_arg_t ddsqra
;
3068 ddsqra
.ddsqra_name
= dsname
;
3069 ddsqra
.ddsqra_source
= source
;
3070 ddsqra
.ddsqra_value
= refquota
;
3072 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3073 dsl_dataset_set_refquota_sync
, &ddsqra
, 0, ZFS_SPACE_CHECK_NONE
));
3077 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3079 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3080 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3083 uint64_t newval
, unique
;
3085 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3086 return (SET_ERROR(ENOTSUP
));
3088 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3092 if (ds
->ds_is_snapshot
) {
3093 dsl_dataset_rele(ds
, FTAG
);
3094 return (SET_ERROR(EINVAL
));
3097 error
= dsl_prop_predict(ds
->ds_dir
,
3098 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3099 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3101 dsl_dataset_rele(ds
, FTAG
);
3106 * If we are doing the preliminary check in open context, the
3107 * space estimates may be inaccurate.
3109 if (!dmu_tx_is_syncing(tx
)) {
3110 dsl_dataset_rele(ds
, FTAG
);
3114 mutex_enter(&ds
->ds_lock
);
3115 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3116 dsl_dataset_recalc_head_uniq(ds
);
3117 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3118 mutex_exit(&ds
->ds_lock
);
3120 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3121 uint64_t delta
= MAX(unique
, newval
) -
3122 MAX(unique
, ds
->ds_reserved
);
3125 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3126 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3127 dsl_dataset_rele(ds
, FTAG
);
3128 return (SET_ERROR(ENOSPC
));
3132 dsl_dataset_rele(ds
, FTAG
);
3137 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3138 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3144 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3145 source
, sizeof (value
), 1, &value
, tx
);
3147 VERIFY0(dsl_prop_get_int_ds(ds
,
3148 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3150 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3151 mutex_enter(&ds
->ds_dir
->dd_lock
);
3152 mutex_enter(&ds
->ds_lock
);
3153 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3154 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3155 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3156 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3157 ds
->ds_reserved
= newval
;
3158 mutex_exit(&ds
->ds_lock
);
3160 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3161 mutex_exit(&ds
->ds_dir
->dd_lock
);
3165 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3167 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3168 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3171 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3172 dsl_dataset_set_refreservation_sync_impl(ds
,
3173 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3174 dsl_dataset_rele(ds
, FTAG
);
3178 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3179 uint64_t refreservation
)
3181 dsl_dataset_set_qr_arg_t ddsqra
;
3183 ddsqra
.ddsqra_name
= dsname
;
3184 ddsqra
.ddsqra_source
= source
;
3185 ddsqra
.ddsqra_value
= refreservation
;
3187 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3188 dsl_dataset_set_refreservation_sync
, &ddsqra
,
3189 0, ZFS_SPACE_CHECK_NONE
));
3193 * Return (in *usedp) the amount of space written in new that is not
3194 * present in oldsnap. New may be a snapshot or the head. Old must be
3195 * a snapshot before new, in new's filesystem (or its origin). If not then
3196 * fail and return EINVAL.
3198 * The written space is calculated by considering two components: First, we
3199 * ignore any freed space, and calculate the written as new's used space
3200 * minus old's used space. Next, we add in the amount of space that was freed
3201 * between the two snapshots, thus reducing new's used space relative to old's.
3202 * Specifically, this is the space that was born before old->ds_creation_txg,
3203 * and freed before new (ie. on new's deadlist or a previous deadlist).
3205 * space freed [---------------------]
3206 * snapshots ---O-------O--------O-------O------
3210 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3211 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3215 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3217 ASSERT(dsl_pool_config_held(dp
));
3220 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3221 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3224 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3225 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3228 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3229 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3231 snapobj
= new->ds_object
;
3232 while (snapobj
!= oldsnap
->ds_object
) {
3233 dsl_dataset_t
*snap
;
3234 uint64_t used
, comp
, uncomp
;
3236 if (snapobj
== new->ds_object
) {
3239 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3244 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3245 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3247 * The blocks in the deadlist can not be born after
3248 * ds_prev_snap_txg, so get the whole deadlist space,
3249 * which is more efficient (especially for old-format
3250 * deadlists). Unfortunately the deadlist code
3251 * doesn't have enough information to make this
3252 * optimization itself.
3254 dsl_deadlist_space(&snap
->ds_deadlist
,
3255 &used
, &comp
, &uncomp
);
3257 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3258 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3259 &used
, &comp
, &uncomp
);
3266 * If we get to the beginning of the chain of snapshots
3267 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3268 * was not a snapshot of/before new.
3270 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3272 dsl_dataset_rele(snap
, FTAG
);
3274 err
= SET_ERROR(EINVAL
);
3283 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3284 * lastsnap, and all snapshots in between are deleted.
3286 * blocks that would be freed [---------------------------]
3287 * snapshots ---O-------O--------O-------O--------O
3288 * firstsnap lastsnap
3290 * This is the set of blocks that were born after the snap before firstsnap,
3291 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3292 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3293 * We calculate this by iterating over the relevant deadlists (from the snap
3294 * after lastsnap, backward to the snap after firstsnap), summing up the
3295 * space on the deadlist that was born after the snap before firstsnap.
3298 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3299 dsl_dataset_t
*lastsnap
,
3300 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3304 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3306 ASSERT(firstsnap
->ds_is_snapshot
);
3307 ASSERT(lastsnap
->ds_is_snapshot
);
3310 * Check that the snapshots are in the same dsl_dir, and firstsnap
3311 * is before lastsnap.
3313 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3314 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3315 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3316 return (SET_ERROR(EINVAL
));
3318 *usedp
= *compp
= *uncompp
= 0;
3320 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3321 while (snapobj
!= firstsnap
->ds_object
) {
3323 uint64_t used
, comp
, uncomp
;
3325 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3329 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3330 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3331 &used
, &comp
, &uncomp
);
3336 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3337 ASSERT3U(snapobj
, !=, 0);
3338 dsl_dataset_rele(ds
, FTAG
);
3344 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3345 * For example, they could both be snapshots of the same filesystem, and
3346 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3347 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3348 * filesystem. Or 'earlier' could be the origin's origin.
3350 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3353 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3354 uint64_t earlier_txg
)
3356 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3359 dsl_dataset_t
*origin
;
3361 ASSERT(dsl_pool_config_held(dp
));
3362 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3364 if (earlier_txg
== 0)
3365 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3367 if (later
->ds_is_snapshot
&&
3368 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3371 if (later
->ds_dir
== earlier
->ds_dir
)
3373 if (!dsl_dir_is_clone(later
->ds_dir
))
3376 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3378 error
= dsl_dataset_hold_obj(dp
,
3379 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3382 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3383 dsl_dataset_rele(origin
, FTAG
);
3388 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
3390 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
3391 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
3394 #if defined(_KERNEL) && defined(HAVE_SPL)
3396 module_param(zfs_max_recordsize
, int, 0644);
3397 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3399 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
3400 module_param(zfs_max_recordsize
, int, 0444);
3401 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
3404 EXPORT_SYMBOL(dsl_dataset_hold
);
3405 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
3406 EXPORT_SYMBOL(dsl_dataset_own
);
3407 EXPORT_SYMBOL(dsl_dataset_own_obj
);
3408 EXPORT_SYMBOL(dsl_dataset_name
);
3409 EXPORT_SYMBOL(dsl_dataset_rele
);
3410 EXPORT_SYMBOL(dsl_dataset_disown
);
3411 EXPORT_SYMBOL(dsl_dataset_tryown
);
3412 EXPORT_SYMBOL(dsl_dataset_create_sync
);
3413 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
3414 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
3415 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
3416 EXPORT_SYMBOL(dsl_dataset_promote
);
3417 EXPORT_SYMBOL(dsl_dataset_user_hold
);
3418 EXPORT_SYMBOL(dsl_dataset_user_release
);
3419 EXPORT_SYMBOL(dsl_dataset_get_holds
);
3420 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
3421 EXPORT_SYMBOL(dsl_dataset_set_blkptr
);
3422 EXPORT_SYMBOL(dsl_dataset_get_spa
);
3423 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
3424 EXPORT_SYMBOL(dsl_dataset_space_written
);
3425 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
3426 EXPORT_SYMBOL(dsl_dataset_sync
);
3427 EXPORT_SYMBOL(dsl_dataset_block_born
);
3428 EXPORT_SYMBOL(dsl_dataset_block_kill
);
3429 EXPORT_SYMBOL(dsl_dataset_block_freeable
);
3430 EXPORT_SYMBOL(dsl_dataset_prev_snap_txg
);
3431 EXPORT_SYMBOL(dsl_dataset_dirty
);
3432 EXPORT_SYMBOL(dsl_dataset_stats
);
3433 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
3434 EXPORT_SYMBOL(dsl_dataset_space
);
3435 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
3436 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
3437 EXPORT_SYMBOL(dsl_dataset_check_quota
);
3438 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
3439 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);