4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
25 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
26 * Copyright (c) 2014 RackTop Systems.
27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
28 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
29 * Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
30 * Copyright 2017 Nexenta Systems, Inc.
33 #include <sys/dmu_objset.h>
34 #include <sys/dsl_dataset.h>
35 #include <sys/dsl_dir.h>
36 #include <sys/dsl_prop.h>
37 #include <sys/dsl_synctask.h>
38 #include <sys/dmu_traverse.h>
39 #include <sys/dmu_impl.h>
40 #include <sys/dmu_tx.h>
44 #include <sys/zfeature.h>
45 #include <sys/unique.h>
46 #include <sys/zfs_context.h>
47 #include <sys/zfs_ioctl.h>
49 #include <sys/spa_impl.h>
51 #include <sys/zfs_znode.h>
52 #include <sys/zfs_onexit.h>
54 #include <sys/dsl_scan.h>
55 #include <sys/dsl_deadlist.h>
56 #include <sys/dsl_destroy.h>
57 #include <sys/dsl_userhold.h>
58 #include <sys/dsl_bookmark.h>
59 #include <sys/policy.h>
60 #include <sys/dmu_recv.h>
61 #include <sys/zio_compress.h>
62 #include <zfs_fletcher.h>
63 #include <sys/zio_checksum.h>
66 * The SPA supports block sizes up to 16MB. However, very large blocks
67 * can have an impact on i/o latency (e.g. tying up a spinning disk for
68 * ~300ms), and also potentially on the memory allocator. Therefore,
69 * we do not allow the recordsize to be set larger than zfs_max_recordsize
70 * (default 1MB). Larger blocks can be created by changing this tunable,
71 * and pools with larger blocks can always be imported and used, regardless
74 int zfs_max_recordsize
= 1 * 1024 * 1024;
76 #define SWITCH64(x, y) \
78 uint64_t __tmp = (x); \
83 #define DS_REF_MAX (1ULL << 62)
85 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
87 static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
,
88 uint64_t obj
, dmu_tx_t
*tx
);
89 static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
,
92 extern int spa_asize_inflation
;
94 static zil_header_t zero_zil
;
97 * Figure out how much of this delta should be propagated to the dsl_dir
98 * layer. If there's a refreservation, that space has already been
99 * partially accounted for in our ancestors.
102 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
104 dsl_dataset_phys_t
*ds_phys
;
105 uint64_t old_bytes
, new_bytes
;
107 if (ds
->ds_reserved
== 0)
110 ds_phys
= dsl_dataset_phys(ds
);
111 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
112 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
114 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
115 return (new_bytes
- old_bytes
);
119 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
121 int used
, compressed
, uncompressed
;
124 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
125 compressed
= BP_GET_PSIZE(bp
);
126 uncompressed
= BP_GET_UCSIZE(bp
);
128 dprintf_bp(bp
, "ds=%p", ds
);
130 ASSERT(dmu_tx_is_syncing(tx
));
131 /* It could have been compressed away to nothing */
134 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
135 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
137 dsl_pool_mos_diduse_space(tx
->tx_pool
,
138 used
, compressed
, uncompressed
);
142 ASSERT3U(bp
->blk_birth
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
143 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
144 mutex_enter(&ds
->ds_lock
);
145 delta
= parent_delta(ds
, used
);
146 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
147 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
148 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
149 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
151 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
152 ds
->ds_feature_activation_needed
[SPA_FEATURE_LARGE_BLOCKS
] =
156 spa_feature_t f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
157 if (f
!= SPA_FEATURE_NONE
)
158 ds
->ds_feature_activation_needed
[f
] = B_TRUE
;
160 mutex_exit(&ds
->ds_lock
);
161 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
162 compressed
, uncompressed
, tx
);
163 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
164 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
168 * Called when the specified segment has been remapped, and is thus no
169 * longer referenced in the head dataset. The vdev must be indirect.
171 * If the segment is referenced by a snapshot, put it on the remap deadlist.
172 * Otherwise, add this segment to the obsolete spacemap.
175 dsl_dataset_block_remapped(dsl_dataset_t
*ds
, uint64_t vdev
, uint64_t offset
,
176 uint64_t size
, uint64_t birth
, dmu_tx_t
*tx
)
178 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
180 ASSERT(dmu_tx_is_syncing(tx
));
181 ASSERT(birth
<= tx
->tx_txg
);
182 ASSERT(!ds
->ds_is_snapshot
);
184 if (birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
185 spa_vdev_indirect_mark_obsolete(spa
, vdev
, offset
, size
, tx
);
188 dva_t
*dva
= &fakebp
.blk_dva
[0];
192 mutex_enter(&ds
->ds_remap_deadlist_lock
);
193 if (!dsl_dataset_remap_deadlist_exists(ds
)) {
194 dsl_dataset_create_remap_deadlist(ds
, tx
);
196 mutex_exit(&ds
->ds_remap_deadlist_lock
);
199 fakebp
.blk_birth
= birth
;
200 DVA_SET_VDEV(dva
, vdev
);
201 DVA_SET_OFFSET(dva
, offset
);
202 DVA_SET_ASIZE(dva
, size
);
204 dsl_deadlist_insert(&ds
->ds_remap_deadlist
, &fakebp
, tx
);
209 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
212 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
214 int used
= bp_get_dsize_sync(spa
, bp
);
215 int compressed
= BP_GET_PSIZE(bp
);
216 int uncompressed
= BP_GET_UCSIZE(bp
);
221 ASSERT(dmu_tx_is_syncing(tx
));
222 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
225 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
226 dsl_pool_mos_diduse_space(tx
->tx_pool
,
227 -used
, -compressed
, -uncompressed
);
230 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
232 ASSERT(!ds
->ds_is_snapshot
);
233 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
235 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
238 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
239 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
241 mutex_enter(&ds
->ds_lock
);
242 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
243 !DS_UNIQUE_IS_ACCURATE(ds
));
244 delta
= parent_delta(ds
, -used
);
245 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
246 mutex_exit(&ds
->ds_lock
);
247 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
248 delta
, -compressed
, -uncompressed
, tx
);
249 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
250 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
252 dprintf_bp(bp
, "putting on dead list: %s", "");
255 * We are here as part of zio's write done callback,
256 * which means we're a zio interrupt thread. We can't
257 * call dsl_deadlist_insert() now because it may block
258 * waiting for I/O. Instead, put bp on the deferred
259 * queue and let dsl_pool_sync() finish the job.
261 bplist_append(&ds
->ds_pending_deadlist
, bp
);
263 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
265 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
266 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
267 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
268 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
269 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
270 ds
->ds_object
&& bp
->blk_birth
>
271 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
272 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
273 mutex_enter(&ds
->ds_prev
->ds_lock
);
274 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
275 mutex_exit(&ds
->ds_prev
->ds_lock
);
277 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
278 dsl_dir_transfer_space(ds
->ds_dir
, used
,
279 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
282 mutex_enter(&ds
->ds_lock
);
283 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
284 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
285 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
286 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
287 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
288 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
289 mutex_exit(&ds
->ds_lock
);
295 * We have to release the fsid syncronously or we risk that a subsequent
296 * mount of the same dataset will fail to unique_insert the fsid. This
297 * failure would manifest itself as the fsid of this dataset changing
298 * between mounts which makes NFS clients quite unhappy.
301 dsl_dataset_evict_sync(void *dbu
)
303 dsl_dataset_t
*ds
= dbu
;
305 ASSERT(ds
->ds_owner
== NULL
);
307 unique_remove(ds
->ds_fsid_guid
);
311 dsl_dataset_evict_async(void *dbu
)
313 dsl_dataset_t
*ds
= dbu
;
315 ASSERT(ds
->ds_owner
== NULL
);
319 if (ds
->ds_objset
!= NULL
)
320 dmu_objset_evict(ds
->ds_objset
);
323 dsl_dataset_rele(ds
->ds_prev
, ds
);
327 bplist_destroy(&ds
->ds_pending_deadlist
);
328 if (dsl_deadlist_is_open(&ds
->ds_deadlist
))
329 dsl_deadlist_close(&ds
->ds_deadlist
);
330 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
331 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
333 dsl_dir_async_rele(ds
->ds_dir
, ds
);
335 ASSERT(!list_link_active(&ds
->ds_synced_link
));
337 list_destroy(&ds
->ds_prop_cbs
);
338 mutex_destroy(&ds
->ds_lock
);
339 mutex_destroy(&ds
->ds_opening_lock
);
340 mutex_destroy(&ds
->ds_sendstream_lock
);
341 mutex_destroy(&ds
->ds_remap_deadlist_lock
);
342 zfs_refcount_destroy(&ds
->ds_longholds
);
343 rrw_destroy(&ds
->ds_bp_rwlock
);
345 kmem_free(ds
, sizeof (dsl_dataset_t
));
349 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
351 dsl_dataset_phys_t
*headphys
;
354 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
355 objset_t
*mos
= dp
->dp_meta_objset
;
357 if (ds
->ds_snapname
[0])
359 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
362 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
366 headphys
= headdbuf
->db_data
;
367 err
= zap_value_search(dp
->dp_meta_objset
,
368 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
369 if (err
!= 0 && zfs_recover
== B_TRUE
) {
371 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
372 "SNAPOBJ=%llu-ERR=%d",
373 (unsigned long long)ds
->ds_object
, err
);
375 dmu_buf_rele(headdbuf
, FTAG
);
380 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
382 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
383 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
387 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
390 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
391 value
, mt
, NULL
, 0, NULL
);
392 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
393 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
398 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
401 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
402 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
406 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
408 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
411 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
412 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
413 err
= zap_remove(mos
, snapobj
, name
, tx
);
415 if (err
== 0 && adj_cnt
)
416 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
417 DD_FIELD_SNAPSHOT_COUNT
, tx
);
423 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
425 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
426 boolean_t result
= B_FALSE
;
428 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
429 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
431 if (ds
== dmu_buf_get_user(dbuf
))
434 dmu_buf_rele(dbuf
, tag
);
441 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
444 objset_t
*mos
= dp
->dp_meta_objset
;
448 dmu_object_info_t doi
;
450 ASSERT(dsl_pool_config_held(dp
));
452 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
456 /* Make sure dsobj has the correct object type. */
457 dmu_object_info_from_db(dbuf
, &doi
);
458 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
459 dmu_buf_rele(dbuf
, tag
);
460 return (SET_ERROR(EINVAL
));
463 ds
= dmu_buf_get_user(dbuf
);
465 dsl_dataset_t
*winner
= NULL
;
467 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
469 ds
->ds_object
= dsobj
;
470 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
471 list_link_init(&ds
->ds_synced_link
);
473 err
= dsl_dir_hold_obj(dp
, dsl_dataset_phys(ds
)->ds_dir_obj
,
474 NULL
, ds
, &ds
->ds_dir
);
476 kmem_free(ds
, sizeof (dsl_dataset_t
));
477 dmu_buf_rele(dbuf
, tag
);
481 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
482 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
483 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
484 mutex_init(&ds
->ds_remap_deadlist_lock
,
485 NULL
, MUTEX_DEFAULT
, NULL
);
486 rrw_init(&ds
->ds_bp_rwlock
, B_FALSE
);
487 zfs_refcount_create(&ds
->ds_longholds
);
489 bplist_create(&ds
->ds_pending_deadlist
);
491 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
492 offsetof(dmu_sendarg_t
, dsa_link
));
494 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
495 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
497 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
500 for (f
= 0; f
< SPA_FEATURES
; f
++) {
501 if (!(spa_feature_table
[f
].fi_flags
&
502 ZFEATURE_FLAG_PER_DATASET
))
504 err
= zap_contains(mos
, dsobj
,
505 spa_feature_table
[f
].fi_guid
);
507 ds
->ds_feature_inuse
[f
] = B_TRUE
;
509 ASSERT3U(err
, ==, ENOENT
);
515 if (!ds
->ds_is_snapshot
) {
516 ds
->ds_snapname
[0] = '\0';
517 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
518 err
= dsl_dataset_hold_obj(dp
,
519 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
522 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
523 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
524 DS_FIELD_BOOKMARK_NAMES
,
525 sizeof (ds
->ds_bookmarks
), 1,
527 if (zaperr
!= ENOENT
)
531 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
532 err
= dsl_dataset_get_snapname(ds
);
534 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
536 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
537 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
542 if (err
== 0 && !ds
->ds_is_snapshot
) {
543 err
= dsl_prop_get_int_ds(ds
,
544 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
547 err
= dsl_prop_get_int_ds(ds
,
548 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
552 ds
->ds_reserved
= ds
->ds_quota
= 0;
555 dsl_deadlist_open(&ds
->ds_deadlist
,
556 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
557 uint64_t remap_deadlist_obj
=
558 dsl_dataset_get_remap_deadlist_object(ds
);
559 if (remap_deadlist_obj
!= 0) {
560 dsl_deadlist_open(&ds
->ds_remap_deadlist
, mos
,
564 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict_sync
,
565 dsl_dataset_evict_async
, &ds
->ds_dbuf
);
567 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
569 if (err
!= 0 || winner
!= NULL
) {
570 bplist_destroy(&ds
->ds_pending_deadlist
);
571 dsl_deadlist_close(&ds
->ds_deadlist
);
572 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
573 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
575 dsl_dataset_rele(ds
->ds_prev
, ds
);
576 dsl_dir_rele(ds
->ds_dir
, ds
);
577 mutex_destroy(&ds
->ds_lock
);
578 mutex_destroy(&ds
->ds_opening_lock
);
579 mutex_destroy(&ds
->ds_sendstream_lock
);
580 zfs_refcount_destroy(&ds
->ds_longholds
);
581 kmem_free(ds
, sizeof (dsl_dataset_t
));
583 dmu_buf_rele(dbuf
, tag
);
589 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
590 if (ds
->ds_fsid_guid
!=
591 dsl_dataset_phys(ds
)->ds_fsid_guid
) {
592 zfs_dbgmsg("ds_fsid_guid changed from "
593 "%llx to %llx for pool %s dataset id %llu",
595 dsl_dataset_phys(ds
)->ds_fsid_guid
,
596 (long long)ds
->ds_fsid_guid
,
597 spa_name(dp
->dp_spa
),
603 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
604 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
605 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
606 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
607 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
614 dsl_dataset_create_key_mapping(dsl_dataset_t
*ds
)
616 dsl_dir_t
*dd
= ds
->ds_dir
;
618 if (dd
->dd_crypto_obj
== 0)
621 return (spa_keystore_create_mapping(dd
->dd_pool
->dp_spa
,
622 ds
, ds
, &ds
->ds_key_mapping
));
626 dsl_dataset_hold_obj_flags(dsl_pool_t
*dp
, uint64_t dsobj
,
627 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
)
631 err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
635 ASSERT3P(*dsp
, !=, NULL
);
637 if (flags
& DS_HOLD_FLAG_DECRYPT
) {
638 err
= dsl_dataset_create_key_mapping(*dsp
);
640 dsl_dataset_rele(*dsp
, tag
);
647 dsl_dataset_hold_flags(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
648 void *tag
, dsl_dataset_t
**dsp
)
651 const char *snapname
;
656 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
660 ASSERT(dsl_pool_config_held(dp
));
661 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
663 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
, &ds
);
665 err
= SET_ERROR(ENOENT
);
667 /* we may be looking for a snapshot */
668 if (err
== 0 && snapname
!= NULL
) {
669 dsl_dataset_t
*snap_ds
;
671 if (*snapname
++ != '@') {
672 dsl_dataset_rele_flags(ds
, flags
, tag
);
673 dsl_dir_rele(dd
, FTAG
);
674 return (SET_ERROR(ENOENT
));
677 dprintf("looking for snapshot '%s'\n", snapname
);
678 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
680 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
,
683 dsl_dataset_rele_flags(ds
, flags
, tag
);
686 mutex_enter(&snap_ds
->ds_lock
);
687 if (snap_ds
->ds_snapname
[0] == 0)
688 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
689 sizeof (snap_ds
->ds_snapname
));
690 mutex_exit(&snap_ds
->ds_lock
);
696 dsl_dir_rele(dd
, FTAG
);
701 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
, void *tag
,
704 return (dsl_dataset_hold_flags(dp
, name
, 0, tag
, dsp
));
708 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
709 void *tag
, dsl_dataset_t
**dsp
)
711 int err
= dsl_dataset_hold_obj_flags(dp
, dsobj
, flags
, tag
, dsp
);
714 if (!dsl_dataset_tryown(*dsp
, tag
)) {
715 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
717 return (SET_ERROR(EBUSY
));
723 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
724 void *tag
, dsl_dataset_t
**dsp
)
726 int err
= dsl_dataset_hold_flags(dp
, name
, flags
, tag
, dsp
);
729 if (!dsl_dataset_tryown(*dsp
, tag
)) {
730 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
731 return (SET_ERROR(EBUSY
));
737 * See the comment above dsl_pool_hold() for details. In summary, a long
738 * hold is used to prevent destruction of a dataset while the pool hold
739 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
741 * The dataset and pool must be held when this function is called. After it
742 * is called, the pool hold may be released while the dataset is still held
746 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
748 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
749 (void) zfs_refcount_add(&ds
->ds_longholds
, tag
);
753 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
755 (void) zfs_refcount_remove(&ds
->ds_longholds
, tag
);
758 /* Return B_TRUE if there are any long holds on this dataset. */
760 dsl_dataset_long_held(dsl_dataset_t
*ds
)
762 return (!zfs_refcount_is_zero(&ds
->ds_longholds
));
766 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
769 (void) strcpy(name
, "mos");
771 dsl_dir_name(ds
->ds_dir
, name
);
772 VERIFY0(dsl_dataset_get_snapname(ds
));
773 if (ds
->ds_snapname
[0]) {
774 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
775 <, ZFS_MAX_DATASET_NAME_LEN
);
777 * We use a "recursive" mutex so that we
778 * can call dprintf_ds() with ds_lock held.
780 if (!MUTEX_HELD(&ds
->ds_lock
)) {
781 mutex_enter(&ds
->ds_lock
);
782 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
783 ZFS_MAX_DATASET_NAME_LEN
), <,
784 ZFS_MAX_DATASET_NAME_LEN
);
785 mutex_exit(&ds
->ds_lock
);
787 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
788 ZFS_MAX_DATASET_NAME_LEN
), <,
789 ZFS_MAX_DATASET_NAME_LEN
);
796 dsl_dataset_namelen(dsl_dataset_t
*ds
)
798 VERIFY0(dsl_dataset_get_snapname(ds
));
799 mutex_enter(&ds
->ds_lock
);
800 int len
= strlen(ds
->ds_snapname
);
801 /* add '@' if ds is a snap */
804 len
+= dsl_dir_namelen(ds
->ds_dir
);
805 mutex_exit(&ds
->ds_lock
);
810 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
812 dmu_buf_rele(ds
->ds_dbuf
, tag
);
816 dsl_dataset_remove_key_mapping(dsl_dataset_t
*ds
)
818 dsl_dir_t
*dd
= ds
->ds_dir
;
820 if (dd
== NULL
|| dd
->dd_crypto_obj
== 0)
823 (void) spa_keystore_remove_mapping(dd
->dd_pool
->dp_spa
,
828 dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
830 if (flags
& DS_HOLD_FLAG_DECRYPT
)
831 dsl_dataset_remove_key_mapping(ds
);
833 dsl_dataset_rele(ds
, tag
);
837 dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
839 ASSERT3P(ds
->ds_owner
, ==, tag
);
840 ASSERT(ds
->ds_dbuf
!= NULL
);
842 mutex_enter(&ds
->ds_lock
);
844 mutex_exit(&ds
->ds_lock
);
845 dsl_dataset_long_rele(ds
, tag
);
846 dsl_dataset_rele_flags(ds
, flags
, tag
);
850 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
852 boolean_t gotit
= FALSE
;
854 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
855 mutex_enter(&ds
->ds_lock
);
856 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
858 dsl_dataset_long_hold(ds
, tag
);
861 mutex_exit(&ds
->ds_lock
);
866 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
869 mutex_enter(&ds
->ds_lock
);
870 rv
= (ds
->ds_owner
!= NULL
);
871 mutex_exit(&ds
->ds_lock
);
876 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
878 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
879 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
882 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
884 spa_feature_incr(spa
, f
, tx
);
885 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
887 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
888 sizeof (zero
), 1, &zero
, tx
));
892 dsl_dataset_deactivate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
894 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
895 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
897 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
899 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
900 spa_feature_decr(spa
, f
, tx
);
904 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
905 dsl_crypto_params_t
*dcp
, uint64_t flags
, dmu_tx_t
*tx
)
907 dsl_pool_t
*dp
= dd
->dd_pool
;
909 dsl_dataset_phys_t
*dsphys
;
911 objset_t
*mos
= dp
->dp_meta_objset
;
914 origin
= dp
->dp_origin_snap
;
916 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
917 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
918 ASSERT(dmu_tx_is_syncing(tx
));
919 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
921 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
922 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
923 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
924 dmu_buf_will_dirty(dbuf
, tx
);
925 dsphys
= dbuf
->db_data
;
926 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
927 dsphys
->ds_dir_obj
= dd
->dd_object
;
928 dsphys
->ds_flags
= flags
;
929 dsphys
->ds_fsid_guid
= unique_create();
930 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
931 sizeof (dsphys
->ds_guid
));
932 dsphys
->ds_snapnames_zapobj
=
933 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
935 dsphys
->ds_creation_time
= gethrestime_sec();
936 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
938 if (origin
== NULL
) {
939 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
941 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
943 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
944 dsphys
->ds_prev_snap_txg
=
945 dsl_dataset_phys(origin
)->ds_creation_txg
;
946 dsphys
->ds_referenced_bytes
=
947 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
948 dsphys
->ds_compressed_bytes
=
949 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
950 dsphys
->ds_uncompressed_bytes
=
951 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
952 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
953 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
954 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
957 * Inherit flags that describe the dataset's contents
958 * (INCONSISTENT) or properties (Case Insensitive).
960 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
961 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
963 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
964 if (origin
->ds_feature_inuse
[f
])
965 dsl_dataset_activate_feature(dsobj
, f
, tx
);
968 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
969 dsl_dataset_phys(origin
)->ds_num_children
++;
971 VERIFY0(dsl_dataset_hold_obj(dp
,
972 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
974 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
975 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
976 dsl_dataset_rele(ohds
, FTAG
);
978 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
979 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
980 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
982 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
984 VERIFY0(zap_add_int(mos
,
985 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
989 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
990 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
991 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
992 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
993 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
994 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
996 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
998 VERIFY0(zap_add_int(mos
,
999 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
1004 /* handle encryption */
1005 dsl_dataset_create_crypt_sync(dsobj
, dd
, origin
, dcp
, tx
);
1007 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1008 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1010 dmu_buf_rele(dbuf
, FTAG
);
1012 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
1013 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
1019 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1023 VERIFY0(dmu_objset_from_ds(ds
, &os
));
1024 if (bcmp(&os
->os_zil_header
, &zero_zil
, sizeof (zero_zil
)) != 0) {
1025 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1028 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
1029 if (os
->os_encrypted
)
1030 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_TRUE
;
1032 zio
= zio_root(dp
->dp_spa
, NULL
, NULL
, ZIO_FLAG_MUSTSUCCEED
);
1033 dsl_dataset_sync(ds
, zio
, tx
);
1034 VERIFY0(zio_wait(zio
));
1036 /* dsl_dataset_sync_done will drop this reference. */
1037 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1038 dsl_dataset_sync_done(ds
, tx
);
1043 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
1044 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
,
1045 dsl_crypto_params_t
*dcp
, dmu_tx_t
*tx
)
1047 dsl_pool_t
*dp
= pdd
->dd_pool
;
1048 uint64_t dsobj
, ddobj
;
1051 ASSERT(dmu_tx_is_syncing(tx
));
1052 ASSERT(lastname
[0] != '@');
1054 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
1055 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
1057 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
, dcp
,
1058 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
1060 dsl_deleg_set_create_perms(dd
, tx
, cr
);
1063 * Since we're creating a new node we know it's a leaf, so we can
1064 * initialize the counts if the limit feature is active.
1066 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
1068 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
1070 dsl_dir_zapify(dd
, tx
);
1071 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
1072 sizeof (cnt
), 1, &cnt
, tx
));
1073 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
1074 sizeof (cnt
), 1, &cnt
, tx
));
1077 dsl_dir_rele(dd
, FTAG
);
1080 * If we are creating a clone, make sure we zero out any stale
1081 * data from the origin snapshots zil header.
1083 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
1086 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
1087 dsl_dataset_zero_zil(ds
, tx
);
1088 dsl_dataset_rele(ds
, FTAG
);
1095 * The unique space in the head dataset can be calculated by subtracting
1096 * the space used in the most recent snapshot, that is still being used
1097 * in this file system, from the space currently in use. To figure out
1098 * the space in the most recent snapshot still in use, we need to take
1099 * the total space used in the snapshot and subtract out the space that
1100 * has been freed up since the snapshot was taken.
1103 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1106 uint64_t dlused
, dlcomp
, dluncomp
;
1108 ASSERT(!ds
->ds_is_snapshot
);
1110 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1111 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1115 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1117 ASSERT3U(dlused
, <=, mrs_used
);
1118 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1119 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1121 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1122 SPA_VERSION_UNIQUE_ACCURATE
)
1123 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1127 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1130 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1131 ASSERTV(uint64_t count
);
1134 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1135 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1138 * The err should not be ENOENT, but a bug in a previous version
1139 * of the code could cause upgrade_clones_cb() to not set
1140 * ds_next_snap_obj when it should, leading to a missing entry.
1141 * If we knew that the pool was created after
1142 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1143 * ENOENT. However, at least we can check that we don't have
1144 * too many entries in the next_clones_obj even after failing to
1149 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1151 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1156 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1158 return (&dsl_dataset_phys(ds
)->ds_bp
);
1162 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1164 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1168 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1172 if (ds
== NULL
) /* this is the meta-objset */
1175 ASSERT(ds
->ds_objset
!= NULL
);
1177 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1178 panic("dirtying snapshot!");
1180 /* Must not dirty a dataset in the same txg where it got snapshotted. */
1181 ASSERT3U(tx
->tx_txg
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
1183 dp
= ds
->ds_dir
->dd_pool
;
1184 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1185 objset_t
*os
= ds
->ds_objset
;
1187 /* up the hold count until we can be written out */
1188 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1190 /* if this dataset is encrypted, grab a reference to the DCK */
1191 if (ds
->ds_dir
->dd_crypto_obj
!= 0 &&
1192 !os
->os_raw_receive
&&
1193 !os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]) {
1194 ASSERT3P(ds
->ds_key_mapping
, !=, NULL
);
1195 key_mapping_add_ref(ds
->ds_key_mapping
, ds
);
1201 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1205 if (!dmu_tx_is_syncing(tx
))
1209 * If there's an fs-only reservation, any blocks that might become
1210 * owned by the snapshot dataset must be accommodated by space
1211 * outside of the reservation.
1213 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1214 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1215 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1216 return (SET_ERROR(ENOSPC
));
1219 * Propagate any reserved space for this snapshot to other
1220 * snapshot checks in this sync group.
1223 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1229 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1230 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1235 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1237 if (!dmu_tx_is_syncing(tx
))
1241 * We don't allow multiple snapshots of the same txg. If there
1242 * is already one, try again.
1244 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1245 return (SET_ERROR(EAGAIN
));
1248 * Check for conflicting snapshot name.
1250 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1252 return (SET_ERROR(EEXIST
));
1253 if (error
!= ENOENT
)
1257 * We don't allow taking snapshots of inconsistent datasets, such as
1258 * those into which we are currently receiving. However, if we are
1259 * creating this snapshot as part of a receive, this check will be
1260 * executed atomically with respect to the completion of the receive
1261 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1262 * case we ignore this, knowing it will be fixed up for us shortly in
1263 * dmu_recv_end_sync().
1265 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1266 return (SET_ERROR(EBUSY
));
1269 * Skip the check for temporary snapshots or if we have already checked
1270 * the counts in dsl_dataset_snapshot_check. This means we really only
1271 * check the count here when we're receiving a stream.
1273 if (cnt
!= 0 && cr
!= NULL
) {
1274 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1275 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1280 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1288 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1290 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1291 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1296 * Pre-compute how many total new snapshots will be created for each
1297 * level in the tree and below. This is needed for validating the
1298 * snapshot limit when either taking a recursive snapshot or when
1299 * taking multiple snapshots.
1301 * The problem is that the counts are not actually adjusted when
1302 * we are checking, only when we finally sync. For a single snapshot,
1303 * this is easy, the count will increase by 1 at each node up the tree,
1304 * but its more complicated for the recursive/multiple snapshot case.
1306 * The dsl_fs_ss_limit_check function does recursively check the count
1307 * at each level up the tree but since it is validating each snapshot
1308 * independently we need to be sure that we are validating the complete
1309 * count for the entire set of snapshots. We do this by rolling up the
1310 * counts for each component of the name into an nvlist and then
1311 * checking each of those cases with the aggregated count.
1313 * This approach properly handles not only the recursive snapshot
1314 * case (where we get all of those on the ddsa_snaps list) but also
1315 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1316 * validate the limit on 'a' using a count of 2).
1318 * We validate the snapshot names in the third loop and only report
1321 if (dmu_tx_is_syncing(tx
)) {
1323 nvlist_t
*cnt_track
= NULL
;
1324 cnt_track
= fnvlist_alloc();
1326 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1328 /* Rollup aggregated counts into the cnt_track list */
1329 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1331 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1335 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1336 pdelim
= strchr(nm
, '@');
1342 if (nvlist_lookup_uint64(cnt_track
, nm
,
1344 /* update existing entry */
1345 fnvlist_add_uint64(cnt_track
, nm
,
1349 fnvlist_add_uint64(cnt_track
, nm
, 1);
1352 pdelim
= strrchr(nm
, '/');
1355 } while (pdelim
!= NULL
);
1358 kmem_free(nm
, MAXPATHLEN
);
1360 /* Check aggregated counts at each level */
1361 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1362 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1368 name
= nvpair_name(pair
);
1369 cnt
= fnvpair_value_uint64(pair
);
1372 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1374 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1375 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1377 dsl_dataset_rele(ds
, FTAG
);
1381 if (ddsa
->ddsa_errors
!= NULL
)
1382 fnvlist_add_int32(ddsa
->ddsa_errors
,
1385 /* only report one error for this check */
1389 nvlist_free(cnt_track
);
1392 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1393 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1396 char *name
, *atp
= NULL
;
1397 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1399 name
= nvpair_name(pair
);
1400 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1401 error
= SET_ERROR(ENAMETOOLONG
);
1403 atp
= strchr(name
, '@');
1405 error
= SET_ERROR(EINVAL
);
1407 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1410 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1412 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1413 error
= dsl_dataset_snapshot_check_impl(ds
,
1414 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1415 dsl_dataset_rele(ds
, FTAG
);
1419 if (ddsa
->ddsa_errors
!= NULL
) {
1420 fnvlist_add_int32(ddsa
->ddsa_errors
,
1431 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1434 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1436 dsl_dataset_phys_t
*dsphys
;
1437 uint64_t dsobj
, crtxg
;
1438 objset_t
*mos
= dp
->dp_meta_objset
;
1439 ASSERTV(static zil_header_t zero_zil
);
1440 ASSERTV(objset_t
*os
);
1442 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1445 * If we are on an old pool, the zil must not be active, in which
1446 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1448 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1449 dmu_objset_from_ds(ds
, &os
) != 0 ||
1450 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1451 sizeof (zero_zil
)) == 0);
1453 /* Should not snapshot a dirty dataset. */
1454 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1457 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1460 * The origin's ds_creation_txg has to be < TXG_INITIAL
1462 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1467 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1468 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1469 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1470 dmu_buf_will_dirty(dbuf
, tx
);
1471 dsphys
= dbuf
->db_data
;
1472 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1473 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1474 dsphys
->ds_fsid_guid
= unique_create();
1475 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1476 sizeof (dsphys
->ds_guid
));
1477 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1478 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1479 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1480 dsphys
->ds_num_children
= 1;
1481 dsphys
->ds_creation_time
= gethrestime_sec();
1482 dsphys
->ds_creation_txg
= crtxg
;
1483 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1484 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1485 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1486 dsphys
->ds_uncompressed_bytes
=
1487 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1488 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1489 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1490 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1491 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1492 dmu_buf_rele(dbuf
, FTAG
);
1494 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1495 if (ds
->ds_feature_inuse
[f
])
1496 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1499 ASSERT3U(ds
->ds_prev
!= 0, ==,
1500 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1502 uint64_t next_clones_obj
=
1503 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1504 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1506 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1507 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1509 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1510 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1511 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1512 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1513 } else if (next_clones_obj
!= 0) {
1514 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1515 dsphys
->ds_next_snap_obj
, tx
);
1516 VERIFY0(zap_add_int(mos
,
1517 next_clones_obj
, dsobj
, tx
));
1522 * If we have a reference-reservation on this dataset, we will
1523 * need to increase the amount of refreservation being charged
1524 * since our unique space is going to zero.
1526 if (ds
->ds_reserved
) {
1528 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1529 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1531 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1535 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1536 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1537 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1538 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1539 dsl_deadlist_close(&ds
->ds_deadlist
);
1540 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1541 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1542 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1543 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1545 if (dsl_dataset_remap_deadlist_exists(ds
)) {
1546 uint64_t remap_deadlist_obj
=
1547 dsl_dataset_get_remap_deadlist_object(ds
);
1549 * Move the remap_deadlist to the snapshot. The head
1550 * will create a new remap deadlist on demand, from
1551 * dsl_dataset_block_remapped().
1553 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
1554 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
1556 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1557 VERIFY0(zap_add(mos
, dsobj
, DS_FIELD_REMAP_DEADLIST
,
1558 sizeof (remap_deadlist_obj
), 1, &remap_deadlist_obj
, tx
));
1561 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1562 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1563 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1564 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1566 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1567 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1569 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1570 snapname
, 8, 1, &dsobj
, tx
));
1573 dsl_dataset_rele(ds
->ds_prev
, ds
);
1574 VERIFY0(dsl_dataset_hold_obj(dp
,
1575 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1577 dsl_scan_ds_snapshotted(ds
, tx
);
1579 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1581 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1585 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1587 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1588 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1591 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1592 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1595 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1597 name
= nvpair_name(pair
);
1598 atp
= strchr(name
, '@');
1599 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1600 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1602 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1603 if (ddsa
->ddsa_props
!= NULL
) {
1604 dsl_props_set_sync_impl(ds
->ds_prev
,
1605 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1607 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1608 dsl_dataset_rele(ds
, FTAG
);
1613 * The snapshots must all be in the same pool.
1614 * All-or-nothing: if there are any failures, nothing will be modified.
1617 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1619 dsl_dataset_snapshot_arg_t ddsa
;
1621 boolean_t needsuspend
;
1625 nvlist_t
*suspended
= NULL
;
1627 pair
= nvlist_next_nvpair(snaps
, NULL
);
1630 firstname
= nvpair_name(pair
);
1632 error
= spa_open(firstname
, &spa
, FTAG
);
1635 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1636 spa_close(spa
, FTAG
);
1639 suspended
= fnvlist_alloc();
1640 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1641 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1642 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1643 char *snapname
= nvpair_name(pair
);
1647 atp
= strchr(snapname
, '@');
1649 error
= SET_ERROR(EINVAL
);
1652 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1654 error
= zil_suspend(fsname
, &cookie
);
1657 fnvlist_add_uint64(suspended
, fsname
,
1662 ddsa
.ddsa_snaps
= snaps
;
1663 ddsa
.ddsa_props
= props
;
1664 ddsa
.ddsa_errors
= errors
;
1665 ddsa
.ddsa_cr
= CRED();
1668 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1669 dsl_dataset_snapshot_sync
, &ddsa
,
1670 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1673 if (suspended
!= NULL
) {
1674 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1675 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1676 zil_resume((void *)(uintptr_t)
1677 fnvpair_value_uint64(pair
));
1679 fnvlist_free(suspended
);
1685 typedef struct dsl_dataset_snapshot_tmp_arg
{
1686 const char *ddsta_fsname
;
1687 const char *ddsta_snapname
;
1688 minor_t ddsta_cleanup_minor
;
1689 const char *ddsta_htag
;
1690 } dsl_dataset_snapshot_tmp_arg_t
;
1693 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1695 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1696 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1700 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1704 /* NULL cred means no limit check for tmp snapshot */
1705 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1706 tx
, B_FALSE
, 0, NULL
);
1708 dsl_dataset_rele(ds
, FTAG
);
1712 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1713 dsl_dataset_rele(ds
, FTAG
);
1714 return (SET_ERROR(ENOTSUP
));
1716 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1719 dsl_dataset_rele(ds
, FTAG
);
1723 dsl_dataset_rele(ds
, FTAG
);
1728 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1730 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1731 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1732 dsl_dataset_t
*ds
= NULL
;
1734 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1736 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1737 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1738 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1739 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1741 dsl_dataset_rele(ds
, FTAG
);
1745 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1746 minor_t cleanup_minor
, const char *htag
)
1748 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1751 boolean_t needsuspend
;
1754 ddsta
.ddsta_fsname
= fsname
;
1755 ddsta
.ddsta_snapname
= snapname
;
1756 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1757 ddsta
.ddsta_htag
= htag
;
1759 error
= spa_open(fsname
, &spa
, FTAG
);
1762 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1763 spa_close(spa
, FTAG
);
1766 error
= zil_suspend(fsname
, &cookie
);
1771 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1772 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1780 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1782 ASSERT(dmu_tx_is_syncing(tx
));
1783 ASSERT(ds
->ds_objset
!= NULL
);
1784 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1787 * in case we had to change ds_fsid_guid when we opened it,
1790 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1791 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1793 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1794 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1795 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1796 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1797 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1798 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1799 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1800 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1801 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1802 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1803 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1804 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1805 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1808 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1810 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1811 if (ds
->ds_feature_activation_needed
[f
]) {
1812 if (ds
->ds_feature_inuse
[f
])
1814 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1815 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1821 deadlist_enqueue_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
1823 dsl_deadlist_t
*dl
= arg
;
1824 dsl_deadlist_insert(dl
, bp
, tx
);
1829 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1831 objset_t
*os
= ds
->ds_objset
;
1833 bplist_iterate(&ds
->ds_pending_deadlist
,
1834 deadlist_enqueue_cb
, &ds
->ds_deadlist
, tx
);
1836 if (os
->os_synced_dnodes
!= NULL
) {
1837 multilist_destroy(os
->os_synced_dnodes
);
1838 os
->os_synced_dnodes
= NULL
;
1841 if (os
->os_encrypted
)
1842 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_FALSE
;
1844 ASSERT0(os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]);
1846 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
1848 dmu_buf_rele(ds
->ds_dbuf
, ds
);
1852 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
1855 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1859 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1862 * There may be missing entries in ds_next_clones_obj
1863 * due to a bug in a previous version of the code.
1864 * Only trust it if it has the right number of entries.
1866 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1867 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1870 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
1873 for (zap_cursor_init(&zc
, mos
,
1874 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1875 zap_cursor_retrieve(&zc
, &za
) == 0;
1876 zap_cursor_advance(&zc
)) {
1877 dsl_dataset_t
*clone
;
1878 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1879 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1880 za
.za_first_integer
, FTAG
, &clone
));
1881 dsl_dir_name(clone
->ds_dir
, buf
);
1882 fnvlist_add_boolean(val
, buf
);
1883 dsl_dataset_rele(clone
, FTAG
);
1885 zap_cursor_fini(&zc
);
1890 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1892 nvlist_t
*propval
= fnvlist_alloc();
1896 * We use nvlist_alloc() instead of fnvlist_alloc() because the
1897 * latter would allocate the list with NV_UNIQUE_NAME flag.
1898 * As a result, every time a clone name is appended to the list
1899 * it would be (linearly) searched for for a duplicate name.
1900 * We already know that all clone names must be unique and we
1901 * want avoid the quadratic complexity of double-checking that
1902 * because we can have a large number of clones.
1904 VERIFY0(nvlist_alloc(&val
, 0, KM_SLEEP
));
1906 if (get_clones_stat_impl(ds
, val
) == 0) {
1907 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1908 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
1913 nvlist_free(propval
);
1917 * Returns a string that represents the receive resume stats token. It should
1918 * be freed with strfree().
1921 get_receive_resume_stats_impl(dsl_dataset_t
*ds
)
1923 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1925 if (dsl_dataset_has_resume_receive_state(ds
)) {
1928 uint8_t *compressed
;
1930 nvlist_t
*token_nv
= fnvlist_alloc();
1931 size_t packed_size
, compressed_size
;
1933 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1934 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1935 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1937 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1938 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1939 fnvlist_add_uint64(token_nv
, "object", val
);
1941 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1942 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1943 fnvlist_add_uint64(token_nv
, "offset", val
);
1945 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1946 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1947 fnvlist_add_uint64(token_nv
, "bytes", val
);
1949 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1950 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1951 fnvlist_add_uint64(token_nv
, "toguid", val
);
1953 char buf
[MAXNAMELEN
];
1954 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1955 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1956 fnvlist_add_string(token_nv
, "toname", buf
);
1958 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1959 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1960 fnvlist_add_boolean(token_nv
, "largeblockok");
1962 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1963 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1964 fnvlist_add_boolean(token_nv
, "embedok");
1966 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1967 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1968 fnvlist_add_boolean(token_nv
, "compressok");
1970 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1971 DS_FIELD_RESUME_RAWOK
) == 0) {
1972 fnvlist_add_boolean(token_nv
, "rawok");
1974 packed
= fnvlist_pack(token_nv
, &packed_size
);
1975 fnvlist_free(token_nv
);
1976 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1978 compressed_size
= gzip_compress(packed
, compressed
,
1979 packed_size
, packed_size
, 6);
1982 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1984 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1985 for (int i
= 0; i
< compressed_size
; i
++) {
1986 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1988 str
[compressed_size
* 2] = '\0';
1989 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
1990 ZFS_SEND_RESUME_TOKEN_VERSION
,
1991 (longlong_t
)cksum
.zc_word
[0],
1992 (longlong_t
)packed_size
, str
);
1993 kmem_free(packed
, packed_size
);
1994 kmem_free(str
, compressed_size
* 2 + 1);
1995 kmem_free(compressed
, packed_size
);
1998 return (strdup(""));
2002 * Returns a string that represents the receive resume stats token of the
2003 * dataset's child. It should be freed with strfree().
2006 get_child_receive_stats(dsl_dataset_t
*ds
)
2008 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2009 dsl_dataset_t
*recv_ds
;
2010 dsl_dataset_name(ds
, recvname
);
2011 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2012 sizeof (recvname
) &&
2013 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2014 sizeof (recvname
) &&
2015 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, recvname
, FTAG
,
2017 char *propval
= get_receive_resume_stats_impl(recv_ds
);
2018 dsl_dataset_rele(recv_ds
, FTAG
);
2021 return (strdup(""));
2025 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2027 char *propval
= get_receive_resume_stats_impl(ds
);
2028 if (strcmp(propval
, "") != 0) {
2029 dsl_prop_nvlist_add_string(nv
,
2030 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
2032 char *childval
= get_child_receive_stats(ds
);
2033 if (strcmp(childval
, "") != 0) {
2034 dsl_prop_nvlist_add_string(nv
,
2035 ZFS_PROP_RECEIVE_RESUME_TOKEN
, childval
);
2043 dsl_get_refratio(dsl_dataset_t
*ds
)
2045 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
2046 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
2047 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
2052 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
2054 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
2058 dsl_get_compressratio(dsl_dataset_t
*ds
)
2060 if (ds
->ds_is_snapshot
) {
2061 return (dsl_get_refratio(ds
));
2063 dsl_dir_t
*dd
= ds
->ds_dir
;
2064 mutex_enter(&dd
->dd_lock
);
2065 uint64_t val
= dsl_dir_get_compressratio(dd
);
2066 mutex_exit(&dd
->dd_lock
);
2072 dsl_get_used(dsl_dataset_t
*ds
)
2074 if (ds
->ds_is_snapshot
) {
2075 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2077 dsl_dir_t
*dd
= ds
->ds_dir
;
2078 mutex_enter(&dd
->dd_lock
);
2079 uint64_t val
= dsl_dir_get_used(dd
);
2080 mutex_exit(&dd
->dd_lock
);
2086 dsl_get_creation(dsl_dataset_t
*ds
)
2088 return (dsl_dataset_phys(ds
)->ds_creation_time
);
2092 dsl_get_creationtxg(dsl_dataset_t
*ds
)
2094 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
2098 dsl_get_refquota(dsl_dataset_t
*ds
)
2100 return (ds
->ds_quota
);
2104 dsl_get_refreservation(dsl_dataset_t
*ds
)
2106 return (ds
->ds_reserved
);
2110 dsl_get_guid(dsl_dataset_t
*ds
)
2112 return (dsl_dataset_phys(ds
)->ds_guid
);
2116 dsl_get_unique(dsl_dataset_t
*ds
)
2118 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2122 dsl_get_objsetid(dsl_dataset_t
*ds
)
2124 return (ds
->ds_object
);
2128 dsl_get_userrefs(dsl_dataset_t
*ds
)
2130 return (ds
->ds_userrefs
);
2134 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2136 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2140 dsl_get_referenced(dsl_dataset_t
*ds
)
2142 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2146 dsl_get_numclones(dsl_dataset_t
*ds
)
2148 ASSERT(ds
->ds_is_snapshot
);
2149 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2153 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2155 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2160 dsl_get_available(dsl_dataset_t
*ds
)
2162 uint64_t refdbytes
= dsl_get_referenced(ds
);
2163 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2165 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2167 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2169 if (ds
->ds_quota
!= 0) {
2171 * Adjust available bytes according to refquota
2173 if (refdbytes
< ds
->ds_quota
) {
2174 availbytes
= MIN(availbytes
,
2175 ds
->ds_quota
- refdbytes
);
2180 return (availbytes
);
2184 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2186 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2187 dsl_dataset_t
*prev
;
2188 int err
= dsl_dataset_hold_obj(dp
,
2189 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2191 uint64_t comp
, uncomp
;
2192 err
= dsl_dataset_space_written(prev
, ds
, written
,
2194 dsl_dataset_rele(prev
, FTAG
);
2200 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2203 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2205 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2206 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2207 dsl_dataset_name(ds
->ds_prev
, snap
);
2215 * Returns the mountpoint property and source for the given dataset in the value
2216 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2217 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2218 * Returns 0 on success and an error on failure.
2221 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2225 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2227 /* Retrieve the mountpoint value stored in the zap opbject */
2228 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2229 ZAP_MAXVALUELEN
, value
, source
);
2235 * Process the dsname and source to find the full mountpoint string.
2236 * Can be skipped for 'legacy' or 'none'.
2238 if (value
[0] == '/') {
2239 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2241 const char *relpath
;
2244 * If we inherit the mountpoint, even from a dataset
2245 * with a received value, the source will be the path of
2246 * the dataset we inherit from. If source is
2247 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2250 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2253 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2254 relpath
= dsname
+ strlen(source
);
2255 if (relpath
[0] == '/')
2259 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2262 * Special case an alternate root of '/'. This will
2263 * avoid having multiple leading slashes in the
2266 if (strcmp(root
, "/") == 0)
2270 * If the mountpoint is '/' then skip over this
2271 * if we are obtaining either an alternate root or
2272 * an inherited mountpoint.
2275 if (value
[1] == '\0' && (root
[0] != '\0' ||
2276 relpath
[0] != '\0'))
2279 if (relpath
[0] == '\0') {
2280 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2283 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2284 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2287 kmem_free(buf
, ZAP_MAXVALUELEN
);
2294 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2296 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2298 ASSERT(dsl_pool_config_held(dp
));
2300 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2301 dsl_get_refratio(ds
));
2302 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2303 dsl_get_logicalreferenced(ds
));
2304 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2305 dsl_get_compressratio(ds
));
2306 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2309 if (ds
->ds_is_snapshot
) {
2310 get_clones_stat(ds
, nv
);
2312 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2313 if (dsl_get_prev_snap(ds
, buf
) == 0)
2314 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2316 dsl_dir_stats(ds
->ds_dir
, nv
);
2319 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2320 dsl_get_available(ds
));
2321 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2322 dsl_get_referenced(ds
));
2323 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2324 dsl_get_creation(ds
));
2325 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2326 dsl_get_creationtxg(ds
));
2327 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2328 dsl_get_refquota(ds
));
2329 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2330 dsl_get_refreservation(ds
));
2331 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2333 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2334 dsl_get_unique(ds
));
2335 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2336 dsl_get_objsetid(ds
));
2337 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2338 dsl_get_userrefs(ds
));
2339 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2340 dsl_get_defer_destroy(ds
));
2341 dsl_dataset_crypt_stats(ds
, nv
);
2343 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2345 if (dsl_get_written(ds
, &written
) == 0) {
2346 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2351 if (!dsl_dataset_is_snapshot(ds
)) {
2353 * A failed "newfs" (e.g. full) resumable receive leaves
2354 * the stats set on this dataset. Check here for the prop.
2356 get_receive_resume_stats(ds
, nv
);
2359 * A failed incremental resumable receive leaves the
2360 * stats set on our child named "%recv". Check the child
2363 /* 6 extra bytes for /%recv */
2364 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2365 dsl_dataset_t
*recv_ds
;
2366 dsl_dataset_name(ds
, recvname
);
2367 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2368 sizeof (recvname
) &&
2369 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2370 sizeof (recvname
) &&
2371 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
2372 get_receive_resume_stats(recv_ds
, nv
);
2373 dsl_dataset_rele(recv_ds
, FTAG
);
2379 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2381 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2382 ASSERT(dsl_pool_config_held(dp
));
2384 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2385 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2386 stat
->dds_guid
= dsl_get_guid(ds
);
2387 stat
->dds_origin
[0] = '\0';
2388 if (ds
->ds_is_snapshot
) {
2389 stat
->dds_is_snapshot
= B_TRUE
;
2390 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2392 stat
->dds_is_snapshot
= B_FALSE
;
2393 stat
->dds_num_clones
= 0;
2395 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2396 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2402 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2404 return (ds
->ds_fsid_guid
);
2408 dsl_dataset_space(dsl_dataset_t
*ds
,
2409 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2410 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2412 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2413 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2414 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2416 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2417 if (ds
->ds_quota
!= 0) {
2419 * Adjust available bytes according to refquota
2421 if (*refdbytesp
< ds
->ds_quota
)
2422 *availbytesp
= MIN(*availbytesp
,
2423 ds
->ds_quota
- *refdbytesp
);
2427 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2428 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2429 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2430 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2434 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2436 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2439 ASSERT(dsl_pool_config_held(dp
));
2442 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2443 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2444 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2445 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2446 objset_t
*os
, *os_snap
;
2448 * It may be that only the ZIL differs, because it was
2449 * reset in the head. Don't count that as being
2452 if (dmu_objset_from_ds(ds
, &os
) != 0)
2454 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2456 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2457 &os_snap
->os_phys
->os_meta_dnode
,
2458 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2463 typedef struct dsl_dataset_rename_snapshot_arg
{
2464 const char *ddrsa_fsname
;
2465 const char *ddrsa_oldsnapname
;
2466 const char *ddrsa_newsnapname
;
2467 boolean_t ddrsa_recursive
;
2469 } dsl_dataset_rename_snapshot_arg_t
;
2473 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2474 dsl_dataset_t
*hds
, void *arg
)
2476 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2480 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2482 /* ignore nonexistent snapshots */
2483 return (error
== ENOENT
? 0 : error
);
2486 /* new name should not exist */
2487 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2489 error
= SET_ERROR(EEXIST
);
2490 else if (error
== ENOENT
)
2493 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2494 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2495 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2496 error
= SET_ERROR(ENAMETOOLONG
);
2502 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2504 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2505 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2509 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2513 if (ddrsa
->ddrsa_recursive
) {
2514 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2515 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2518 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2520 dsl_dataset_rele(hds
, FTAG
);
2525 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2526 dsl_dataset_t
*hds
, void *arg
)
2528 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2531 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2534 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2535 ASSERT(error
== 0 || error
== ENOENT
);
2536 if (error
== ENOENT
) {
2537 /* ignore nonexistent snapshots */
2541 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2543 /* log before we change the name */
2544 spa_history_log_internal_ds(ds
, "rename", tx
,
2545 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2547 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2549 mutex_enter(&ds
->ds_lock
);
2550 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2551 sizeof (ds
->ds_snapname
));
2552 mutex_exit(&ds
->ds_lock
);
2553 VERIFY0(zap_add(dp
->dp_meta_objset
,
2554 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2555 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2556 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2557 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2559 dsl_dataset_rele(ds
, FTAG
);
2564 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2566 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2567 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2568 dsl_dataset_t
*hds
= NULL
;
2570 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2571 ddrsa
->ddrsa_tx
= tx
;
2572 if (ddrsa
->ddrsa_recursive
) {
2573 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2574 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2577 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2579 dsl_dataset_rele(hds
, FTAG
);
2583 dsl_dataset_rename_snapshot(const char *fsname
,
2584 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2586 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2588 ddrsa
.ddrsa_fsname
= fsname
;
2589 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2590 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2591 ddrsa
.ddrsa_recursive
= recursive
;
2593 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2594 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2595 1, ZFS_SPACE_CHECK_RESERVED
));
2599 * If we're doing an ownership handoff, we need to make sure that there is
2600 * only one long hold on the dataset. We're not allowed to change anything here
2601 * so we don't permanently release the long hold or regular hold here. We want
2602 * to do this only when syncing to avoid the dataset unexpectedly going away
2603 * when we release the long hold.
2606 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2610 if (!dmu_tx_is_syncing(tx
))
2613 if (owner
!= NULL
) {
2614 VERIFY3P(ds
->ds_owner
, ==, owner
);
2615 dsl_dataset_long_rele(ds
, owner
);
2618 held
= dsl_dataset_long_held(ds
);
2621 dsl_dataset_long_hold(ds
, owner
);
2624 return (SET_ERROR(EBUSY
));
2630 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2632 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2633 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2635 int64_t unused_refres_delta
;
2638 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2642 /* must not be a snapshot */
2643 if (ds
->ds_is_snapshot
) {
2644 dsl_dataset_rele(ds
, FTAG
);
2645 return (SET_ERROR(EINVAL
));
2648 /* must have a most recent snapshot */
2649 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2650 dsl_dataset_rele(ds
, FTAG
);
2651 return (SET_ERROR(ESRCH
));
2655 * No rollback to a snapshot created in the current txg, because
2656 * the rollback may dirty the dataset and create blocks that are
2657 * not reachable from the rootbp while having a birth txg that
2658 * falls into the snapshot's range.
2660 if (dmu_tx_is_syncing(tx
) &&
2661 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
2662 dsl_dataset_rele(ds
, FTAG
);
2663 return (SET_ERROR(EAGAIN
));
2667 * If the expected target snapshot is specified, then check that
2668 * the latest snapshot is it.
2670 if (ddra
->ddra_tosnap
!= NULL
) {
2671 dsl_dataset_t
*snapds
;
2673 /* Check if the target snapshot exists at all. */
2674 error
= dsl_dataset_hold(dp
, ddra
->ddra_tosnap
, FTAG
, &snapds
);
2677 * ESRCH is used to signal that the target snapshot does
2678 * not exist, while ENOENT is used to report that
2679 * the rolled back dataset does not exist.
2680 * ESRCH is also used to cover other cases where the
2681 * target snapshot is not related to the dataset being
2682 * rolled back such as being in a different pool.
2684 if (error
== ENOENT
|| error
== EXDEV
)
2685 error
= SET_ERROR(ESRCH
);
2686 dsl_dataset_rele(ds
, FTAG
);
2689 ASSERT(snapds
->ds_is_snapshot
);
2691 /* Check if the snapshot is the latest snapshot indeed. */
2692 if (snapds
!= ds
->ds_prev
) {
2694 * Distinguish between the case where the only problem
2695 * is intervening snapshots (EEXIST) vs the snapshot
2696 * not being a valid target for rollback (ESRCH).
2698 if (snapds
->ds_dir
== ds
->ds_dir
||
2699 (dsl_dir_is_clone(ds
->ds_dir
) &&
2700 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
==
2701 snapds
->ds_object
)) {
2702 error
= SET_ERROR(EEXIST
);
2704 error
= SET_ERROR(ESRCH
);
2706 dsl_dataset_rele(snapds
, FTAG
);
2707 dsl_dataset_rele(ds
, FTAG
);
2710 dsl_dataset_rele(snapds
, FTAG
);
2713 /* must not have any bookmarks after the most recent snapshot */
2714 nvlist_t
*proprequest
= fnvlist_alloc();
2715 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2716 nvlist_t
*bookmarks
= fnvlist_alloc();
2717 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2718 fnvlist_free(proprequest
);
2720 dsl_dataset_rele(ds
, FTAG
);
2723 for (nvpair_t
*pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2724 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2726 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2727 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2728 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2729 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2730 fnvlist_free(bookmarks
);
2731 dsl_dataset_rele(ds
, FTAG
);
2732 return (SET_ERROR(EEXIST
));
2735 fnvlist_free(bookmarks
);
2737 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2739 dsl_dataset_rele(ds
, FTAG
);
2744 * Check if the snap we are rolling back to uses more than
2747 if (ds
->ds_quota
!= 0 &&
2748 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2749 dsl_dataset_rele(ds
, FTAG
);
2750 return (SET_ERROR(EDQUOT
));
2754 * When we do the clone swap, we will temporarily use more space
2755 * due to the refreservation (the head will no longer have any
2756 * unique space, so the entire amount of the refreservation will need
2757 * to be free). We will immediately destroy the clone, freeing
2758 * this space, but the freeing happens over many txg's.
2760 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2761 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2763 if (unused_refres_delta
> 0 &&
2764 unused_refres_delta
>
2765 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2766 dsl_dataset_rele(ds
, FTAG
);
2767 return (SET_ERROR(ENOSPC
));
2770 dsl_dataset_rele(ds
, FTAG
);
2775 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2777 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2778 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2779 dsl_dataset_t
*ds
, *clone
;
2781 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2783 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2785 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2786 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2788 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2789 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
2791 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2793 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2794 dsl_dataset_zero_zil(ds
, tx
);
2796 dsl_destroy_head_sync_impl(clone
, tx
);
2798 dsl_dataset_rele(clone
, FTAG
);
2799 dsl_dataset_rele(ds
, FTAG
);
2803 * Rolls back the given filesystem or volume to the most recent snapshot.
2804 * The name of the most recent snapshot will be returned under key "target"
2805 * in the result nvlist.
2808 * - The existing dataset MUST be owned by the specified owner at entry
2809 * - Upon return, dataset will still be held by the same owner, whether we
2812 * This mode is required any time the existing filesystem is mounted. See
2813 * notes above zfs_suspend_fs() for further details.
2816 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
2819 dsl_dataset_rollback_arg_t ddra
;
2821 ddra
.ddra_fsname
= fsname
;
2822 ddra
.ddra_tosnap
= tosnap
;
2823 ddra
.ddra_owner
= owner
;
2824 ddra
.ddra_result
= result
;
2826 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2827 dsl_dataset_rollback_sync
, &ddra
,
2828 1, ZFS_SPACE_CHECK_RESERVED
));
2831 struct promotenode
{
2836 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2837 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2839 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2842 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2844 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2845 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2847 struct promotenode
*snap
;
2848 dsl_dataset_t
*origin_ds
;
2852 size_t max_snap_len
;
2853 boolean_t conflicting_snaps
;
2855 err
= promote_hold(ddpa
, dp
, FTAG
);
2859 hds
= ddpa
->ddpa_clone
;
2860 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2862 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2863 promote_rele(ddpa
, FTAG
);
2864 return (SET_ERROR(EXDEV
));
2867 snap
= list_head(&ddpa
->shared_snaps
);
2869 err
= SET_ERROR(ENOENT
);
2872 origin_ds
= snap
->ds
;
2875 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
2876 * When doing a promote we must make sure the encryption root for
2877 * both the target and the target's origin does not change to avoid
2878 * needing to rewrap encryption keys
2880 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
2885 * Compute and check the amount of space to transfer. Since this is
2886 * so expensive, don't do the preliminary check.
2888 if (!dmu_tx_is_syncing(tx
)) {
2889 promote_rele(ddpa
, FTAG
);
2893 /* compute origin's new unique space */
2894 snap
= list_tail(&ddpa
->clone_snaps
);
2895 ASSERT(snap
!= NULL
);
2896 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2897 origin_ds
->ds_object
);
2898 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2899 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2900 &ddpa
->unique
, &unused
, &unused
);
2903 * Walk the snapshots that we are moving
2905 * Compute space to transfer. Consider the incremental changes
2906 * to used by each snapshot:
2907 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2908 * So each snapshot gave birth to:
2909 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2910 * So a sequence would look like:
2911 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2912 * Which simplifies to:
2913 * uN + kN + kN-1 + ... + k1 + k0
2914 * Note however, if we stop before we reach the ORIGIN we get:
2915 * uN + kN + kN-1 + ... + kM - uM-1
2917 conflicting_snaps
= B_FALSE
;
2919 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2920 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2921 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2922 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2923 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2924 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2925 dsl_dataset_t
*ds
= snap
->ds
;
2930 * If there are long holds, we won't be able to evict
2933 if (dsl_dataset_long_held(ds
)) {
2934 err
= SET_ERROR(EBUSY
);
2938 /* Check that the snapshot name does not conflict */
2939 VERIFY0(dsl_dataset_get_snapname(ds
));
2940 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2941 err
= SET_ERROR(ENAMETOOLONG
);
2944 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2946 fnvlist_add_boolean(ddpa
->err_ds
,
2947 snap
->ds
->ds_snapname
);
2948 conflicting_snaps
= B_TRUE
;
2949 } else if (err
!= ENOENT
) {
2953 /* The very first snapshot does not have a deadlist */
2954 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2957 dsl_deadlist_space(&ds
->ds_deadlist
,
2958 &dlused
, &dlcomp
, &dluncomp
);
2959 ddpa
->used
+= dlused
;
2960 ddpa
->comp
+= dlcomp
;
2961 ddpa
->uncomp
+= dluncomp
;
2965 * In order to return the full list of conflicting snapshots, we check
2966 * whether there was a conflict after traversing all of them.
2968 if (conflicting_snaps
) {
2969 err
= SET_ERROR(EEXIST
);
2974 * If we are a clone of a clone then we never reached ORIGIN,
2975 * so we need to subtract out the clone origin's used space.
2977 if (ddpa
->origin_origin
) {
2979 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2981 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2983 dsl_dataset_phys(ddpa
->origin_origin
)->
2984 ds_uncompressed_bytes
;
2987 /* Check that there is enough space and limit headroom here */
2988 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2989 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2994 * Compute the amounts of space that will be used by snapshots
2995 * after the promotion (for both origin and clone). For each,
2996 * it is the amount of space that will be on all of their
2997 * deadlists (that was not born before their new origin).
2999 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
3003 * Note, typically this will not be a clone of a clone,
3004 * so dd_origin_txg will be < TXG_INITIAL, so
3005 * these snaplist_space() -> dsl_deadlist_space_range()
3006 * calls will be fast because they do not have to
3007 * iterate over all bps.
3009 snap
= list_head(&ddpa
->origin_snaps
);
3011 err
= SET_ERROR(ENOENT
);
3014 err
= snaplist_space(&ddpa
->shared_snaps
,
3015 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
3019 err
= snaplist_space(&ddpa
->clone_snaps
,
3020 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
3023 ddpa
->cloneusedsnap
+= space
;
3025 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
3026 DD_FLAG_USED_BREAKDOWN
) {
3027 err
= snaplist_space(&ddpa
->origin_snaps
,
3028 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
3029 &ddpa
->originusedsnap
);
3035 promote_rele(ddpa
, FTAG
);
3040 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
3042 dsl_dataset_promote_arg_t
*ddpa
= arg
;
3043 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3045 struct promotenode
*snap
;
3046 dsl_dataset_t
*origin_ds
;
3047 dsl_dataset_t
*origin_head
;
3049 dsl_dir_t
*odd
= NULL
;
3050 uint64_t oldnext_obj
;
3053 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
3054 hds
= ddpa
->ddpa_clone
;
3056 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
3058 snap
= list_head(&ddpa
->shared_snaps
);
3059 origin_ds
= snap
->ds
;
3062 snap
= list_head(&ddpa
->origin_snaps
);
3063 origin_head
= snap
->ds
;
3066 * We need to explicitly open odd, since origin_ds's dd will be
3069 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
3072 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
3074 /* change origin's next snap */
3075 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
3076 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
3077 snap
= list_tail(&ddpa
->clone_snaps
);
3078 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3079 origin_ds
->ds_object
);
3080 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
3082 /* change the origin's next clone */
3083 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
3084 dsl_dataset_remove_from_next_clones(origin_ds
,
3085 snap
->ds
->ds_object
, tx
);
3086 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3087 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
3092 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
3093 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
3094 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
3095 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
3096 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
3097 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
3098 origin_head
->ds_dir
->dd_origin_txg
=
3099 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3101 /* change dd_clone entries */
3102 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3103 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3104 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
3105 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3106 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3107 hds
->ds_object
, tx
));
3109 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3110 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3111 origin_head
->ds_object
, tx
));
3112 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
3113 dsl_dir_phys(dd
)->dd_clones
=
3114 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
3115 DMU_OT_NONE
, 0, tx
);
3117 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3118 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
3121 /* move snapshots to this dir */
3122 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3123 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3124 dsl_dataset_t
*ds
= snap
->ds
;
3127 * Property callbacks are registered to a particular
3128 * dsl_dir. Since ours is changing, evict the objset
3129 * so that they will be unregistered from the old dsl_dir.
3131 if (ds
->ds_objset
) {
3132 dmu_objset_evict(ds
->ds_objset
);
3133 ds
->ds_objset
= NULL
;
3136 /* move snap name entry */
3137 VERIFY0(dsl_dataset_get_snapname(ds
));
3138 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3139 ds
->ds_snapname
, tx
, B_TRUE
));
3140 VERIFY0(zap_add(dp
->dp_meta_objset
,
3141 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3142 8, 1, &ds
->ds_object
, tx
));
3143 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3144 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3146 /* change containing dsl_dir */
3147 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3148 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3149 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3150 ASSERT3P(ds
->ds_dir
, ==, odd
);
3151 dsl_dir_rele(ds
->ds_dir
, ds
);
3152 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3153 NULL
, ds
, &ds
->ds_dir
));
3155 /* move any clone references */
3156 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3157 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3161 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3162 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3163 zap_cursor_retrieve(&zc
, &za
) == 0;
3164 zap_cursor_advance(&zc
)) {
3165 dsl_dataset_t
*cnds
;
3168 if (za
.za_first_integer
== oldnext_obj
) {
3170 * We've already moved the
3171 * origin's reference.
3176 VERIFY0(dsl_dataset_hold_obj(dp
,
3177 za
.za_first_integer
, FTAG
, &cnds
));
3178 o
= dsl_dir_phys(cnds
->ds_dir
)->
3179 dd_head_dataset_obj
;
3181 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3182 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3183 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3184 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3185 dsl_dataset_rele(cnds
, FTAG
);
3187 zap_cursor_fini(&zc
);
3190 ASSERT(!dsl_prop_hascb(ds
));
3194 * Change space accounting.
3195 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3196 * both be valid, or both be 0 (resulting in delta == 0). This
3197 * is true for each of {clone,origin} independently.
3200 delta
= ddpa
->cloneusedsnap
-
3201 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3202 ASSERT3S(delta
, >=, 0);
3203 ASSERT3U(ddpa
->used
, >=, delta
);
3204 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3205 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3206 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3208 delta
= ddpa
->originusedsnap
-
3209 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3210 ASSERT3S(delta
, <=, 0);
3211 ASSERT3U(ddpa
->used
, >=, -delta
);
3212 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3213 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3214 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3216 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3218 /* log history record */
3219 spa_history_log_internal_ds(hds
, "promote", tx
, "");
3221 dsl_dir_rele(odd
, FTAG
);
3222 promote_rele(ddpa
, FTAG
);
3226 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3227 * (exclusive) and last_obj (inclusive). The list will be in reverse
3228 * order (last_obj will be the list_head()). If first_obj == 0, do all
3229 * snapshots back to this dataset's origin.
3232 snaplist_make(dsl_pool_t
*dp
,
3233 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
3235 uint64_t obj
= last_obj
;
3237 list_create(l
, sizeof (struct promotenode
),
3238 offsetof(struct promotenode
, link
));
3240 while (obj
!= first_obj
) {
3242 struct promotenode
*snap
;
3245 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3246 ASSERT(err
!= ENOENT
);
3251 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3253 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3255 list_insert_tail(l
, snap
);
3256 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3263 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3265 struct promotenode
*snap
;
3268 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3269 uint64_t used
, comp
, uncomp
;
3270 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3271 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3278 snaplist_destroy(list_t
*l
, void *tag
)
3280 struct promotenode
*snap
;
3282 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3285 while ((snap
= list_tail(l
)) != NULL
) {
3286 list_remove(l
, snap
);
3287 dsl_dataset_rele(snap
->ds
, tag
);
3288 kmem_free(snap
, sizeof (*snap
));
3294 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
3298 struct promotenode
*snap
;
3300 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3304 dd
= ddpa
->ddpa_clone
->ds_dir
;
3306 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3307 !dsl_dir_is_clone(dd
)) {
3308 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3309 return (SET_ERROR(EINVAL
));
3312 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3313 &ddpa
->shared_snaps
, tag
);
3317 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3318 &ddpa
->clone_snaps
, tag
);
3322 snap
= list_head(&ddpa
->shared_snaps
);
3323 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3324 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3325 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3326 &ddpa
->origin_snaps
, tag
);
3330 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3331 error
= dsl_dataset_hold_obj(dp
,
3332 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3333 tag
, &ddpa
->origin_origin
);
3339 promote_rele(ddpa
, tag
);
3344 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
3346 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3347 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3348 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3349 if (ddpa
->origin_origin
!= NULL
)
3350 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3351 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3357 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3358 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3361 dsl_dataset_promote(const char *name
, char *conflsnap
)
3363 dsl_dataset_promote_arg_t ddpa
= { 0 };
3366 nvpair_t
*snap_pair
;
3370 * We will modify space proportional to the number of
3371 * snapshots. Compute numsnaps.
3373 error
= dmu_objset_hold(name
, FTAG
, &os
);
3376 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3377 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3379 dmu_objset_rele(os
, FTAG
);
3383 ddpa
.ddpa_clonename
= name
;
3384 ddpa
.err_ds
= fnvlist_alloc();
3387 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3388 dsl_dataset_promote_sync
, &ddpa
,
3389 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3392 * Return the first conflicting snapshot found.
3394 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3395 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3396 (void) strcpy(conflsnap
, nvpair_name(snap_pair
));
3398 fnvlist_free(ddpa
.err_ds
);
3403 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3404 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3407 * "slack" factor for received datasets with refquota set on them.
3408 * See the bottom of this function for details on its use.
3410 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3411 spa_asize_inflation
;
3412 int64_t unused_refres_delta
;
3414 /* they should both be heads */
3415 if (clone
->ds_is_snapshot
||
3416 origin_head
->ds_is_snapshot
)
3417 return (SET_ERROR(EINVAL
));
3419 /* if we are not forcing, the branch point should be just before them */
3420 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3421 return (SET_ERROR(EINVAL
));
3423 /* clone should be the clone (unless they are unrelated) */
3424 if (clone
->ds_prev
!= NULL
&&
3425 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3426 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3427 return (SET_ERROR(EINVAL
));
3429 /* the clone should be a child of the origin */
3430 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3431 return (SET_ERROR(EINVAL
));
3433 /* origin_head shouldn't be modified unless 'force' */
3435 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3436 return (SET_ERROR(ETXTBSY
));
3438 /* origin_head should have no long holds (e.g. is not mounted) */
3439 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3440 return (SET_ERROR(EBUSY
));
3442 /* check amount of any unconsumed refreservation */
3443 unused_refres_delta
=
3444 (int64_t)MIN(origin_head
->ds_reserved
,
3445 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3446 (int64_t)MIN(origin_head
->ds_reserved
,
3447 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3449 if (unused_refres_delta
> 0 &&
3450 unused_refres_delta
>
3451 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3452 return (SET_ERROR(ENOSPC
));
3455 * The clone can't be too much over the head's refquota.
3457 * To ensure that the entire refquota can be used, we allow one
3458 * transaction to exceed the the refquota. Therefore, this check
3459 * needs to also allow for the space referenced to be more than the
3460 * refquota. The maximum amount of space that one transaction can use
3461 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3462 * overage ensures that we are able to receive a filesystem that
3463 * exceeds the refquota on the source system.
3465 * So that overage is the refquota_slack we use below.
3467 if (origin_head
->ds_quota
!= 0 &&
3468 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3469 origin_head
->ds_quota
+ refquota_slack
)
3470 return (SET_ERROR(EDQUOT
));
3476 dsl_dataset_swap_remap_deadlists(dsl_dataset_t
*clone
,
3477 dsl_dataset_t
*origin
, dmu_tx_t
*tx
)
3479 uint64_t clone_remap_dl_obj
, origin_remap_dl_obj
;
3480 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3482 ASSERT(dsl_pool_sync_context(dp
));
3484 clone_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(clone
);
3485 origin_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(origin
);
3487 if (clone_remap_dl_obj
!= 0) {
3488 dsl_deadlist_close(&clone
->ds_remap_deadlist
);
3489 dsl_dataset_unset_remap_deadlist_object(clone
, tx
);
3491 if (origin_remap_dl_obj
!= 0) {
3492 dsl_deadlist_close(&origin
->ds_remap_deadlist
);
3493 dsl_dataset_unset_remap_deadlist_object(origin
, tx
);
3496 if (clone_remap_dl_obj
!= 0) {
3497 dsl_dataset_set_remap_deadlist_object(origin
,
3498 clone_remap_dl_obj
, tx
);
3499 dsl_deadlist_open(&origin
->ds_remap_deadlist
,
3500 dp
->dp_meta_objset
, clone_remap_dl_obj
);
3502 if (origin_remap_dl_obj
!= 0) {
3503 dsl_dataset_set_remap_deadlist_object(clone
,
3504 origin_remap_dl_obj
, tx
);
3505 dsl_deadlist_open(&clone
->ds_remap_deadlist
,
3506 dp
->dp_meta_objset
, origin_remap_dl_obj
);
3511 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
3512 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
3514 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3515 int64_t unused_refres_delta
;
3517 ASSERT(clone
->ds_reserved
== 0);
3519 * NOTE: On DEBUG kernels there could be a race between this and
3520 * the check function if spa_asize_inflation is adjusted...
3522 ASSERT(origin_head
->ds_quota
== 0 ||
3523 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
3524 DMU_MAX_ACCESS
* spa_asize_inflation
);
3525 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
3528 * Swap per-dataset feature flags.
3530 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
3531 if (!(spa_feature_table
[f
].fi_flags
&
3532 ZFEATURE_FLAG_PER_DATASET
)) {
3533 ASSERT(!clone
->ds_feature_inuse
[f
]);
3534 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
3538 boolean_t clone_inuse
= clone
->ds_feature_inuse
[f
];
3539 boolean_t origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
3542 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
3543 clone
->ds_feature_inuse
[f
] = B_FALSE
;
3545 if (origin_head_inuse
) {
3546 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
3548 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
3551 dsl_dataset_activate_feature(origin_head
->ds_object
,
3553 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
3555 if (origin_head_inuse
) {
3556 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
3557 clone
->ds_feature_inuse
[f
] = B_TRUE
;
3561 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3562 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3564 if (clone
->ds_objset
!= NULL
) {
3565 dmu_objset_evict(clone
->ds_objset
);
3566 clone
->ds_objset
= NULL
;
3569 if (origin_head
->ds_objset
!= NULL
) {
3570 dmu_objset_evict(origin_head
->ds_objset
);
3571 origin_head
->ds_objset
= NULL
;
3574 unused_refres_delta
=
3575 (int64_t)MIN(origin_head
->ds_reserved
,
3576 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3577 (int64_t)MIN(origin_head
->ds_reserved
,
3578 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3581 * Reset origin's unique bytes, if it exists.
3583 if (clone
->ds_prev
) {
3584 dsl_dataset_t
*origin
= clone
->ds_prev
;
3585 uint64_t comp
, uncomp
;
3587 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3588 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3589 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3590 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3595 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3596 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3598 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3599 dsl_dataset_phys(origin_head
)->ds_bp
=
3600 dsl_dataset_phys(clone
)->ds_bp
;
3601 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3602 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3603 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3606 /* set dd_*_bytes */
3608 int64_t dused
, dcomp
, duncomp
;
3609 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3610 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3612 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3613 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3615 dsl_deadlist_space(&clone
->ds_deadlist
,
3616 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3617 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3618 &odl_used
, &odl_comp
, &odl_uncomp
);
3620 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3622 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3624 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3626 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3628 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3630 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3633 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3634 dused
, dcomp
, duncomp
, tx
);
3635 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3636 -dused
, -dcomp
, -duncomp
, tx
);
3639 * The difference in the space used by snapshots is the
3640 * difference in snapshot space due to the head's
3641 * deadlist (since that's the only thing that's
3642 * changing that affects the snapused).
3644 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3645 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3646 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3647 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3648 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3649 &odl_used
, &odl_comp
, &odl_uncomp
);
3650 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3651 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3654 /* swap ds_*_bytes */
3655 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3656 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3657 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3658 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3659 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3660 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3661 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3662 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3664 /* apply any parent delta for change in unconsumed refreservation */
3665 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3666 unused_refres_delta
, 0, 0, tx
);
3671 dsl_deadlist_close(&clone
->ds_deadlist
);
3672 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3673 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3674 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3675 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3676 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3677 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3678 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3679 dsl_dataset_swap_remap_deadlists(clone
, origin_head
, tx
);
3681 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3683 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3684 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3688 * Given a pool name and a dataset object number in that pool,
3689 * return the name of that dataset.
3692 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3698 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3702 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3704 dsl_dataset_name(ds
, buf
);
3705 dsl_dataset_rele(ds
, FTAG
);
3707 dsl_pool_rele(dp
, FTAG
);
3713 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3714 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3718 ASSERT3S(asize
, >, 0);
3721 * *ref_rsrv is the portion of asize that will come from any
3722 * unconsumed refreservation space.
3726 mutex_enter(&ds
->ds_lock
);
3728 * Make a space adjustment for reserved bytes.
3730 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3732 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3734 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3736 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3739 if (!check_quota
|| ds
->ds_quota
== 0) {
3740 mutex_exit(&ds
->ds_lock
);
3744 * If they are requesting more space, and our current estimate
3745 * is over quota, they get to try again unless the actual
3746 * on-disk is over quota and there are no pending changes (which
3747 * may free up space for us).
3749 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3752 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3753 error
= SET_ERROR(ERESTART
);
3755 error
= SET_ERROR(EDQUOT
);
3757 mutex_exit(&ds
->ds_lock
);
3762 typedef struct dsl_dataset_set_qr_arg
{
3763 const char *ddsqra_name
;
3764 zprop_source_t ddsqra_source
;
3765 uint64_t ddsqra_value
;
3766 } dsl_dataset_set_qr_arg_t
;
3771 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3773 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3774 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3779 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3780 return (SET_ERROR(ENOTSUP
));
3782 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3786 if (ds
->ds_is_snapshot
) {
3787 dsl_dataset_rele(ds
, FTAG
);
3788 return (SET_ERROR(EINVAL
));
3791 error
= dsl_prop_predict(ds
->ds_dir
,
3792 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3793 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3795 dsl_dataset_rele(ds
, FTAG
);
3800 dsl_dataset_rele(ds
, FTAG
);
3804 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3805 newval
< ds
->ds_reserved
) {
3806 dsl_dataset_rele(ds
, FTAG
);
3807 return (SET_ERROR(ENOSPC
));
3810 dsl_dataset_rele(ds
, FTAG
);
3815 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3817 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3818 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3819 dsl_dataset_t
*ds
= NULL
;
3822 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3824 dsl_prop_set_sync_impl(ds
,
3825 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3826 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3827 &ddsqra
->ddsqra_value
, tx
);
3829 VERIFY0(dsl_prop_get_int_ds(ds
,
3830 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3832 if (ds
->ds_quota
!= newval
) {
3833 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3834 ds
->ds_quota
= newval
;
3836 dsl_dataset_rele(ds
, FTAG
);
3840 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3843 dsl_dataset_set_qr_arg_t ddsqra
;
3845 ddsqra
.ddsqra_name
= dsname
;
3846 ddsqra
.ddsqra_source
= source
;
3847 ddsqra
.ddsqra_value
= refquota
;
3849 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3850 dsl_dataset_set_refquota_sync
, &ddsqra
, 0,
3851 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
3855 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3857 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3858 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3861 uint64_t newval
, unique
;
3863 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3864 return (SET_ERROR(ENOTSUP
));
3866 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3870 if (ds
->ds_is_snapshot
) {
3871 dsl_dataset_rele(ds
, FTAG
);
3872 return (SET_ERROR(EINVAL
));
3875 error
= dsl_prop_predict(ds
->ds_dir
,
3876 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3877 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3879 dsl_dataset_rele(ds
, FTAG
);
3884 * If we are doing the preliminary check in open context, the
3885 * space estimates may be inaccurate.
3887 if (!dmu_tx_is_syncing(tx
)) {
3888 dsl_dataset_rele(ds
, FTAG
);
3892 mutex_enter(&ds
->ds_lock
);
3893 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3894 dsl_dataset_recalc_head_uniq(ds
);
3895 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3896 mutex_exit(&ds
->ds_lock
);
3898 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3899 uint64_t delta
= MAX(unique
, newval
) -
3900 MAX(unique
, ds
->ds_reserved
);
3903 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3904 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3905 dsl_dataset_rele(ds
, FTAG
);
3906 return (SET_ERROR(ENOSPC
));
3910 dsl_dataset_rele(ds
, FTAG
);
3915 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3916 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3922 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3923 source
, sizeof (value
), 1, &value
, tx
);
3925 VERIFY0(dsl_prop_get_int_ds(ds
,
3926 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3928 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3929 mutex_enter(&ds
->ds_dir
->dd_lock
);
3930 mutex_enter(&ds
->ds_lock
);
3931 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3932 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3933 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3934 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3935 ds
->ds_reserved
= newval
;
3936 mutex_exit(&ds
->ds_lock
);
3938 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3939 mutex_exit(&ds
->ds_dir
->dd_lock
);
3943 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3945 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3946 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3947 dsl_dataset_t
*ds
= NULL
;
3949 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3950 dsl_dataset_set_refreservation_sync_impl(ds
,
3951 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3952 dsl_dataset_rele(ds
, FTAG
);
3956 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3957 uint64_t refreservation
)
3959 dsl_dataset_set_qr_arg_t ddsqra
;
3961 ddsqra
.ddsqra_name
= dsname
;
3962 ddsqra
.ddsqra_source
= source
;
3963 ddsqra
.ddsqra_value
= refreservation
;
3965 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3966 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0,
3967 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
3971 * Return (in *usedp) the amount of space written in new that is not
3972 * present in oldsnap. New may be a snapshot or the head. Old must be
3973 * a snapshot before new, in new's filesystem (or its origin). If not then
3974 * fail and return EINVAL.
3976 * The written space is calculated by considering two components: First, we
3977 * ignore any freed space, and calculate the written as new's used space
3978 * minus old's used space. Next, we add in the amount of space that was freed
3979 * between the two snapshots, thus reducing new's used space relative to old's.
3980 * Specifically, this is the space that was born before old->ds_creation_txg,
3981 * and freed before new (ie. on new's deadlist or a previous deadlist).
3983 * space freed [---------------------]
3984 * snapshots ---O-------O--------O-------O------
3988 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3989 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3993 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3995 ASSERT(dsl_pool_config_held(dp
));
3998 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3999 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
4002 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
4003 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
4006 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
4007 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
4009 snapobj
= new->ds_object
;
4010 while (snapobj
!= oldsnap
->ds_object
) {
4011 dsl_dataset_t
*snap
;
4012 uint64_t used
, comp
, uncomp
;
4014 if (snapobj
== new->ds_object
) {
4017 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
4022 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
4023 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
4025 * The blocks in the deadlist can not be born after
4026 * ds_prev_snap_txg, so get the whole deadlist space,
4027 * which is more efficient (especially for old-format
4028 * deadlists). Unfortunately the deadlist code
4029 * doesn't have enough information to make this
4030 * optimization itself.
4032 dsl_deadlist_space(&snap
->ds_deadlist
,
4033 &used
, &comp
, &uncomp
);
4035 dsl_deadlist_space_range(&snap
->ds_deadlist
,
4036 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
4037 &used
, &comp
, &uncomp
);
4044 * If we get to the beginning of the chain of snapshots
4045 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
4046 * was not a snapshot of/before new.
4048 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
4050 dsl_dataset_rele(snap
, FTAG
);
4052 err
= SET_ERROR(EINVAL
);
4061 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
4062 * lastsnap, and all snapshots in between are deleted.
4064 * blocks that would be freed [---------------------------]
4065 * snapshots ---O-------O--------O-------O--------O
4066 * firstsnap lastsnap
4068 * This is the set of blocks that were born after the snap before firstsnap,
4069 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
4070 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
4071 * We calculate this by iterating over the relevant deadlists (from the snap
4072 * after lastsnap, backward to the snap after firstsnap), summing up the
4073 * space on the deadlist that was born after the snap before firstsnap.
4076 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
4077 dsl_dataset_t
*lastsnap
,
4078 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4082 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
4084 ASSERT(firstsnap
->ds_is_snapshot
);
4085 ASSERT(lastsnap
->ds_is_snapshot
);
4088 * Check that the snapshots are in the same dsl_dir, and firstsnap
4089 * is before lastsnap.
4091 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
4092 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
4093 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
4094 return (SET_ERROR(EINVAL
));
4096 *usedp
= *compp
= *uncompp
= 0;
4098 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
4099 while (snapobj
!= firstsnap
->ds_object
) {
4101 uint64_t used
, comp
, uncomp
;
4103 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
4107 dsl_deadlist_space_range(&ds
->ds_deadlist
,
4108 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
4109 &used
, &comp
, &uncomp
);
4114 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4115 ASSERT3U(snapobj
, !=, 0);
4116 dsl_dataset_rele(ds
, FTAG
);
4122 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
4123 * For example, they could both be snapshots of the same filesystem, and
4124 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
4125 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
4126 * filesystem. Or 'earlier' could be the origin's origin.
4128 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
4131 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
4132 uint64_t earlier_txg
)
4134 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
4138 ASSERT(dsl_pool_config_held(dp
));
4139 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
4141 if (earlier_txg
== 0)
4142 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
4144 if (later
->ds_is_snapshot
&&
4145 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
4148 if (later
->ds_dir
== earlier
->ds_dir
)
4150 if (!dsl_dir_is_clone(later
->ds_dir
))
4153 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
4155 dsl_dataset_t
*origin
;
4156 error
= dsl_dataset_hold_obj(dp
,
4157 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
4160 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
4161 dsl_dataset_rele(origin
, FTAG
);
4166 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4168 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4169 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4173 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4175 dmu_object_info_t doi
;
4177 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4178 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4182 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4184 return (dsl_dataset_is_zapified(ds
) &&
4185 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4186 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4190 dsl_dataset_get_remap_deadlist_object(dsl_dataset_t
*ds
)
4192 uint64_t remap_deadlist_obj
;
4195 if (!dsl_dataset_is_zapified(ds
))
4198 err
= zap_lookup(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4199 DS_FIELD_REMAP_DEADLIST
, sizeof (remap_deadlist_obj
), 1,
4200 &remap_deadlist_obj
);
4203 VERIFY3S(err
, ==, ENOENT
);
4207 ASSERT(remap_deadlist_obj
!= 0);
4208 return (remap_deadlist_obj
);
4212 dsl_dataset_remap_deadlist_exists(dsl_dataset_t
*ds
)
4214 EQUIV(dsl_deadlist_is_open(&ds
->ds_remap_deadlist
),
4215 dsl_dataset_get_remap_deadlist_object(ds
) != 0);
4216 return (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
));
4220 dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
, uint64_t obj
,
4224 dsl_dataset_zapify(ds
, tx
);
4225 VERIFY0(zap_add(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4226 DS_FIELD_REMAP_DEADLIST
, sizeof (obj
), 1, &obj
, tx
));
4230 dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4232 VERIFY0(zap_remove(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4233 ds
->ds_object
, DS_FIELD_REMAP_DEADLIST
, tx
));
4237 dsl_dataset_destroy_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4239 uint64_t remap_deadlist_object
;
4240 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4242 ASSERT(dmu_tx_is_syncing(tx
));
4243 ASSERT(dsl_dataset_remap_deadlist_exists(ds
));
4245 remap_deadlist_object
= ds
->ds_remap_deadlist
.dl_object
;
4246 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
4247 dsl_deadlist_free(spa_meta_objset(spa
), remap_deadlist_object
, tx
);
4248 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
4249 spa_feature_decr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4253 dsl_dataset_create_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4255 uint64_t remap_deadlist_obj
;
4256 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4258 ASSERT(dmu_tx_is_syncing(tx
));
4259 ASSERT(MUTEX_HELD(&ds
->ds_remap_deadlist_lock
));
4261 * Currently we only create remap deadlists when there are indirect
4262 * vdevs with referenced mappings.
4264 ASSERT(spa_feature_is_active(spa
, SPA_FEATURE_DEVICE_REMOVAL
));
4266 remap_deadlist_obj
= dsl_deadlist_clone(
4267 &ds
->ds_deadlist
, UINT64_MAX
,
4268 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
4269 dsl_dataset_set_remap_deadlist_object(ds
,
4270 remap_deadlist_obj
, tx
);
4271 dsl_deadlist_open(&ds
->ds_remap_deadlist
, spa_meta_objset(spa
),
4272 remap_deadlist_obj
);
4273 spa_feature_incr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4276 #if defined(_KERNEL)
4278 module_param(zfs_max_recordsize
, int, 0644);
4279 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4281 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
4282 module_param(zfs_max_recordsize
, int, 0444);
4283 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4286 EXPORT_SYMBOL(dsl_dataset_hold
);
4287 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4288 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4289 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4290 EXPORT_SYMBOL(dsl_dataset_own
);
4291 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4292 EXPORT_SYMBOL(dsl_dataset_name
);
4293 EXPORT_SYMBOL(dsl_dataset_rele
);
4294 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4295 EXPORT_SYMBOL(dsl_dataset_disown
);
4296 EXPORT_SYMBOL(dsl_dataset_tryown
);
4297 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4298 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4299 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4300 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4301 EXPORT_SYMBOL(dsl_dataset_promote
);
4302 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4303 EXPORT_SYMBOL(dsl_dataset_user_release
);
4304 EXPORT_SYMBOL(dsl_dataset_get_holds
);
4305 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
4306 EXPORT_SYMBOL(dsl_dataset_get_spa
);
4307 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
4308 EXPORT_SYMBOL(dsl_dataset_space_written
);
4309 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
4310 EXPORT_SYMBOL(dsl_dataset_sync
);
4311 EXPORT_SYMBOL(dsl_dataset_block_born
);
4312 EXPORT_SYMBOL(dsl_dataset_block_kill
);
4313 EXPORT_SYMBOL(dsl_dataset_dirty
);
4314 EXPORT_SYMBOL(dsl_dataset_stats
);
4315 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
4316 EXPORT_SYMBOL(dsl_dataset_space
);
4317 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
4318 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
4319 EXPORT_SYMBOL(dsl_dataset_check_quota
);
4320 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
4321 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);