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_send.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 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_flags(dsl_pool_t
*dp
, uint64_t dsobj
,
442 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
)
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 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 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
),
602 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
603 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
604 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
605 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
606 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
609 if ((flags
& DS_HOLD_FLAG_DECRYPT
) && ds
->ds_dir
->dd_crypto_obj
!= 0) {
610 err
= spa_keystore_create_mapping(dp
->dp_spa
, ds
, ds
);
612 dsl_dataset_rele(ds
, tag
);
613 return (SET_ERROR(EACCES
));
621 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
624 return (dsl_dataset_hold_obj_flags(dp
, dsobj
, 0, tag
, dsp
));
628 dsl_dataset_hold_flags(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
629 void *tag
, dsl_dataset_t
**dsp
)
632 const char *snapname
;
637 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
641 ASSERT(dsl_pool_config_held(dp
));
642 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
644 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
, &ds
);
646 err
= SET_ERROR(ENOENT
);
648 /* we may be looking for a snapshot */
649 if (err
== 0 && snapname
!= NULL
) {
650 dsl_dataset_t
*snap_ds
;
652 if (*snapname
++ != '@') {
653 dsl_dataset_rele_flags(ds
, flags
, tag
);
654 dsl_dir_rele(dd
, FTAG
);
655 return (SET_ERROR(ENOENT
));
658 dprintf("looking for snapshot '%s'\n", snapname
);
659 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
661 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
,
664 dsl_dataset_rele_flags(ds
, flags
, tag
);
667 mutex_enter(&snap_ds
->ds_lock
);
668 if (snap_ds
->ds_snapname
[0] == 0)
669 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
670 sizeof (snap_ds
->ds_snapname
));
671 mutex_exit(&snap_ds
->ds_lock
);
677 dsl_dir_rele(dd
, FTAG
);
682 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
, void *tag
,
685 return (dsl_dataset_hold_flags(dp
, name
, 0, tag
, dsp
));
689 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
690 void *tag
, dsl_dataset_t
**dsp
)
692 int err
= dsl_dataset_hold_obj_flags(dp
, dsobj
, flags
, tag
, dsp
);
695 if (!dsl_dataset_tryown(*dsp
, tag
)) {
696 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
698 return (SET_ERROR(EBUSY
));
704 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
705 void *tag
, dsl_dataset_t
**dsp
)
707 int err
= dsl_dataset_hold_flags(dp
, name
, flags
, tag
, dsp
);
710 if (!dsl_dataset_tryown(*dsp
, tag
)) {
711 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
712 return (SET_ERROR(EBUSY
));
718 * See the comment above dsl_pool_hold() for details. In summary, a long
719 * hold is used to prevent destruction of a dataset while the pool hold
720 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
722 * The dataset and pool must be held when this function is called. After it
723 * is called, the pool hold may be released while the dataset is still held
727 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
729 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
730 (void) refcount_add(&ds
->ds_longholds
, tag
);
734 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
736 (void) refcount_remove(&ds
->ds_longholds
, tag
);
739 /* Return B_TRUE if there are any long holds on this dataset. */
741 dsl_dataset_long_held(dsl_dataset_t
*ds
)
743 return (!refcount_is_zero(&ds
->ds_longholds
));
747 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
750 (void) strcpy(name
, "mos");
752 dsl_dir_name(ds
->ds_dir
, name
);
753 VERIFY0(dsl_dataset_get_snapname(ds
));
754 if (ds
->ds_snapname
[0]) {
755 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
756 <, ZFS_MAX_DATASET_NAME_LEN
);
758 * We use a "recursive" mutex so that we
759 * can call dprintf_ds() with ds_lock held.
761 if (!MUTEX_HELD(&ds
->ds_lock
)) {
762 mutex_enter(&ds
->ds_lock
);
763 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
764 ZFS_MAX_DATASET_NAME_LEN
), <,
765 ZFS_MAX_DATASET_NAME_LEN
);
766 mutex_exit(&ds
->ds_lock
);
768 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
769 ZFS_MAX_DATASET_NAME_LEN
), <,
770 ZFS_MAX_DATASET_NAME_LEN
);
777 dsl_dataset_namelen(dsl_dataset_t
*ds
)
779 VERIFY0(dsl_dataset_get_snapname(ds
));
780 mutex_enter(&ds
->ds_lock
);
781 int len
= strlen(ds
->ds_snapname
);
782 /* add '@' if ds is a snap */
785 len
+= dsl_dir_namelen(ds
->ds_dir
);
786 mutex_exit(&ds
->ds_lock
);
791 dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
793 if (ds
->ds_dir
!= NULL
&& ds
->ds_dir
->dd_crypto_obj
!= 0 &&
794 (flags
& DS_HOLD_FLAG_DECRYPT
)) {
795 (void) spa_keystore_remove_mapping(ds
->ds_dir
->dd_pool
->dp_spa
,
799 * Encrypted datasets require that users only release their
800 * decrypting reference after the dirty data has actually
801 * been written out. This ensures that the mapping exists
802 * when it is needed to write out dirty data.
804 ASSERT(dmu_buf_user_refcount(ds
->ds_dbuf
) != 0 ||
805 !dsl_dataset_is_dirty(ds
));
808 dmu_buf_rele(ds
->ds_dbuf
, tag
);
812 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
814 dsl_dataset_rele_flags(ds
, 0, tag
);
818 dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
820 ASSERT3P(ds
->ds_owner
, ==, tag
);
821 ASSERT(ds
->ds_dbuf
!= NULL
);
823 mutex_enter(&ds
->ds_lock
);
825 mutex_exit(&ds
->ds_lock
);
826 dsl_dataset_long_rele(ds
, tag
);
827 dsl_dataset_rele_flags(ds
, flags
, tag
);
831 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
833 boolean_t gotit
= FALSE
;
835 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
836 mutex_enter(&ds
->ds_lock
);
837 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
839 dsl_dataset_long_hold(ds
, tag
);
842 mutex_exit(&ds
->ds_lock
);
847 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
850 mutex_enter(&ds
->ds_lock
);
851 rv
= (ds
->ds_owner
!= NULL
);
852 mutex_exit(&ds
->ds_lock
);
857 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
859 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
860 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
863 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
865 spa_feature_incr(spa
, f
, tx
);
866 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
868 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
869 sizeof (zero
), 1, &zero
, tx
));
873 dsl_dataset_deactivate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
875 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
876 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
878 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
880 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
881 spa_feature_decr(spa
, f
, tx
);
885 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
886 dsl_crypto_params_t
*dcp
, uint64_t flags
, dmu_tx_t
*tx
)
888 dsl_pool_t
*dp
= dd
->dd_pool
;
890 dsl_dataset_phys_t
*dsphys
;
892 objset_t
*mos
= dp
->dp_meta_objset
;
895 origin
= dp
->dp_origin_snap
;
897 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
898 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
899 ASSERT(dmu_tx_is_syncing(tx
));
900 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
902 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
903 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
904 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
905 dmu_buf_will_dirty(dbuf
, tx
);
906 dsphys
= dbuf
->db_data
;
907 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
908 dsphys
->ds_dir_obj
= dd
->dd_object
;
909 dsphys
->ds_flags
= flags
;
910 dsphys
->ds_fsid_guid
= unique_create();
911 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
912 sizeof (dsphys
->ds_guid
));
913 dsphys
->ds_snapnames_zapobj
=
914 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
916 dsphys
->ds_creation_time
= gethrestime_sec();
917 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
919 if (origin
== NULL
) {
920 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
922 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
924 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
925 dsphys
->ds_prev_snap_txg
=
926 dsl_dataset_phys(origin
)->ds_creation_txg
;
927 dsphys
->ds_referenced_bytes
=
928 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
929 dsphys
->ds_compressed_bytes
=
930 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
931 dsphys
->ds_uncompressed_bytes
=
932 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
933 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
934 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
935 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
938 * Inherit flags that describe the dataset's contents
939 * (INCONSISTENT) or properties (Case Insensitive).
941 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
942 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
944 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
945 if (origin
->ds_feature_inuse
[f
])
946 dsl_dataset_activate_feature(dsobj
, f
, tx
);
949 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
950 dsl_dataset_phys(origin
)->ds_num_children
++;
952 VERIFY0(dsl_dataset_hold_obj(dp
,
953 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
955 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
956 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
957 dsl_dataset_rele(ohds
, FTAG
);
959 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
960 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
961 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
963 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
965 VERIFY0(zap_add_int(mos
,
966 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
970 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
971 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
972 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
973 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
974 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
975 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
977 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
979 VERIFY0(zap_add_int(mos
,
980 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
985 /* handle encryption */
986 dsl_dataset_create_crypt_sync(dsobj
, dd
, origin
, dcp
, tx
);
988 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
989 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
991 dmu_buf_rele(dbuf
, FTAG
);
993 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
994 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
1000 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1004 VERIFY0(dmu_objset_from_ds(ds
, &os
));
1005 if (bcmp(&os
->os_zil_header
, &zero_zil
, sizeof (zero_zil
)) != 0) {
1006 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1009 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
1010 if (os
->os_encrypted
)
1011 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_TRUE
;
1013 zio
= zio_root(dp
->dp_spa
, NULL
, NULL
, ZIO_FLAG_MUSTSUCCEED
);
1014 dsl_dataset_sync(ds
, zio
, tx
);
1015 VERIFY0(zio_wait(zio
));
1017 /* dsl_dataset_sync_done will drop this reference. */
1018 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1019 dsl_dataset_sync_done(ds
, tx
);
1024 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
1025 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
,
1026 dsl_crypto_params_t
*dcp
, dmu_tx_t
*tx
)
1028 dsl_pool_t
*dp
= pdd
->dd_pool
;
1029 uint64_t dsobj
, ddobj
;
1032 ASSERT(dmu_tx_is_syncing(tx
));
1033 ASSERT(lastname
[0] != '@');
1035 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
1036 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
1038 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
, dcp
,
1039 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
1041 dsl_deleg_set_create_perms(dd
, tx
, cr
);
1044 * Since we're creating a new node we know it's a leaf, so we can
1045 * initialize the counts if the limit feature is active.
1047 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
1049 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
1051 dsl_dir_zapify(dd
, tx
);
1052 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
1053 sizeof (cnt
), 1, &cnt
, tx
));
1054 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
1055 sizeof (cnt
), 1, &cnt
, tx
));
1058 dsl_dir_rele(dd
, FTAG
);
1061 * If we are creating a clone, make sure we zero out any stale
1062 * data from the origin snapshots zil header.
1064 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
1067 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
1068 dsl_dataset_zero_zil(ds
, tx
);
1069 dsl_dataset_rele(ds
, FTAG
);
1076 * The unique space in the head dataset can be calculated by subtracting
1077 * the space used in the most recent snapshot, that is still being used
1078 * in this file system, from the space currently in use. To figure out
1079 * the space in the most recent snapshot still in use, we need to take
1080 * the total space used in the snapshot and subtract out the space that
1081 * has been freed up since the snapshot was taken.
1084 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1087 uint64_t dlused
, dlcomp
, dluncomp
;
1089 ASSERT(!ds
->ds_is_snapshot
);
1091 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1092 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1096 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1098 ASSERT3U(dlused
, <=, mrs_used
);
1099 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1100 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1102 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1103 SPA_VERSION_UNIQUE_ACCURATE
)
1104 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1108 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1111 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1112 ASSERTV(uint64_t count
);
1115 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1116 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1119 * The err should not be ENOENT, but a bug in a previous version
1120 * of the code could cause upgrade_clones_cb() to not set
1121 * ds_next_snap_obj when it should, leading to a missing entry.
1122 * If we knew that the pool was created after
1123 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1124 * ENOENT. However, at least we can check that we don't have
1125 * too many entries in the next_clones_obj even after failing to
1130 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1132 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1137 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1139 return (&dsl_dataset_phys(ds
)->ds_bp
);
1143 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1145 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1149 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1153 if (ds
== NULL
) /* this is the meta-objset */
1156 ASSERT(ds
->ds_objset
!= NULL
);
1158 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1159 panic("dirtying snapshot!");
1161 /* Must not dirty a dataset in the same txg where it got snapshotted. */
1162 ASSERT3U(tx
->tx_txg
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
1164 dp
= ds
->ds_dir
->dd_pool
;
1165 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1166 /* up the hold count until we can be written out */
1167 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1172 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
1174 for (int t
= 0; t
< TXG_SIZE
; t
++) {
1175 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1183 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1187 if (!dmu_tx_is_syncing(tx
))
1191 * If there's an fs-only reservation, any blocks that might become
1192 * owned by the snapshot dataset must be accommodated by space
1193 * outside of the reservation.
1195 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1196 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1197 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1198 return (SET_ERROR(ENOSPC
));
1201 * Propagate any reserved space for this snapshot to other
1202 * snapshot checks in this sync group.
1205 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1211 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1212 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1217 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1219 if (!dmu_tx_is_syncing(tx
))
1223 * We don't allow multiple snapshots of the same txg. If there
1224 * is already one, try again.
1226 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1227 return (SET_ERROR(EAGAIN
));
1230 * Check for conflicting snapshot name.
1232 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1234 return (SET_ERROR(EEXIST
));
1235 if (error
!= ENOENT
)
1239 * We don't allow taking snapshots of inconsistent datasets, such as
1240 * those into which we are currently receiving. However, if we are
1241 * creating this snapshot as part of a receive, this check will be
1242 * executed atomically with respect to the completion of the receive
1243 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1244 * case we ignore this, knowing it will be fixed up for us shortly in
1245 * dmu_recv_end_sync().
1247 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1248 return (SET_ERROR(EBUSY
));
1251 * Skip the check for temporary snapshots or if we have already checked
1252 * the counts in dsl_dataset_snapshot_check. This means we really only
1253 * check the count here when we're receiving a stream.
1255 if (cnt
!= 0 && cr
!= NULL
) {
1256 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1257 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1262 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1270 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1272 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1273 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1278 * Pre-compute how many total new snapshots will be created for each
1279 * level in the tree and below. This is needed for validating the
1280 * snapshot limit when either taking a recursive snapshot or when
1281 * taking multiple snapshots.
1283 * The problem is that the counts are not actually adjusted when
1284 * we are checking, only when we finally sync. For a single snapshot,
1285 * this is easy, the count will increase by 1 at each node up the tree,
1286 * but its more complicated for the recursive/multiple snapshot case.
1288 * The dsl_fs_ss_limit_check function does recursively check the count
1289 * at each level up the tree but since it is validating each snapshot
1290 * independently we need to be sure that we are validating the complete
1291 * count for the entire set of snapshots. We do this by rolling up the
1292 * counts for each component of the name into an nvlist and then
1293 * checking each of those cases with the aggregated count.
1295 * This approach properly handles not only the recursive snapshot
1296 * case (where we get all of those on the ddsa_snaps list) but also
1297 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1298 * validate the limit on 'a' using a count of 2).
1300 * We validate the snapshot names in the third loop and only report
1303 if (dmu_tx_is_syncing(tx
)) {
1305 nvlist_t
*cnt_track
= NULL
;
1306 cnt_track
= fnvlist_alloc();
1308 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1310 /* Rollup aggregated counts into the cnt_track list */
1311 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1313 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1317 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1318 pdelim
= strchr(nm
, '@');
1324 if (nvlist_lookup_uint64(cnt_track
, nm
,
1326 /* update existing entry */
1327 fnvlist_add_uint64(cnt_track
, nm
,
1331 fnvlist_add_uint64(cnt_track
, nm
, 1);
1334 pdelim
= strrchr(nm
, '/');
1337 } while (pdelim
!= NULL
);
1340 kmem_free(nm
, MAXPATHLEN
);
1342 /* Check aggregated counts at each level */
1343 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1344 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1350 name
= nvpair_name(pair
);
1351 cnt
= fnvpair_value_uint64(pair
);
1354 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1356 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1357 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1359 dsl_dataset_rele(ds
, FTAG
);
1363 if (ddsa
->ddsa_errors
!= NULL
)
1364 fnvlist_add_int32(ddsa
->ddsa_errors
,
1367 /* only report one error for this check */
1371 nvlist_free(cnt_track
);
1374 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1375 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1378 char *name
, *atp
= NULL
;
1379 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1381 name
= nvpair_name(pair
);
1382 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1383 error
= SET_ERROR(ENAMETOOLONG
);
1385 atp
= strchr(name
, '@');
1387 error
= SET_ERROR(EINVAL
);
1389 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1392 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1394 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1395 error
= dsl_dataset_snapshot_check_impl(ds
,
1396 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1397 dsl_dataset_rele(ds
, FTAG
);
1401 if (ddsa
->ddsa_errors
!= NULL
) {
1402 fnvlist_add_int32(ddsa
->ddsa_errors
,
1413 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1416 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1418 dsl_dataset_phys_t
*dsphys
;
1419 uint64_t dsobj
, crtxg
;
1420 objset_t
*mos
= dp
->dp_meta_objset
;
1421 ASSERTV(static zil_header_t zero_zil
);
1422 ASSERTV(objset_t
*os
);
1424 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1427 * If we are on an old pool, the zil must not be active, in which
1428 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1430 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1431 dmu_objset_from_ds(ds
, &os
) != 0 ||
1432 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1433 sizeof (zero_zil
)) == 0);
1435 /* Should not snapshot a dirty dataset. */
1436 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1439 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1442 * The origin's ds_creation_txg has to be < TXG_INITIAL
1444 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1449 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1450 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1451 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1452 dmu_buf_will_dirty(dbuf
, tx
);
1453 dsphys
= dbuf
->db_data
;
1454 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1455 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1456 dsphys
->ds_fsid_guid
= unique_create();
1457 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1458 sizeof (dsphys
->ds_guid
));
1459 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1460 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1461 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1462 dsphys
->ds_num_children
= 1;
1463 dsphys
->ds_creation_time
= gethrestime_sec();
1464 dsphys
->ds_creation_txg
= crtxg
;
1465 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1466 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1467 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1468 dsphys
->ds_uncompressed_bytes
=
1469 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1470 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1471 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1472 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1473 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1474 dmu_buf_rele(dbuf
, FTAG
);
1476 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1477 if (ds
->ds_feature_inuse
[f
])
1478 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1481 ASSERT3U(ds
->ds_prev
!= 0, ==,
1482 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1484 uint64_t next_clones_obj
=
1485 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1486 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1488 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1489 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1491 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1492 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1493 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1494 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1495 } else if (next_clones_obj
!= 0) {
1496 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1497 dsphys
->ds_next_snap_obj
, tx
);
1498 VERIFY0(zap_add_int(mos
,
1499 next_clones_obj
, dsobj
, tx
));
1504 * If we have a reference-reservation on this dataset, we will
1505 * need to increase the amount of refreservation being charged
1506 * since our unique space is going to zero.
1508 if (ds
->ds_reserved
) {
1510 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1511 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1513 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1517 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1518 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1519 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1520 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1521 dsl_deadlist_close(&ds
->ds_deadlist
);
1522 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1523 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1524 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1525 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1527 if (dsl_dataset_remap_deadlist_exists(ds
)) {
1528 uint64_t remap_deadlist_obj
=
1529 dsl_dataset_get_remap_deadlist_object(ds
);
1531 * Move the remap_deadlist to the snapshot. The head
1532 * will create a new remap deadlist on demand, from
1533 * dsl_dataset_block_remapped().
1535 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
1536 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
1538 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1539 VERIFY0(zap_add(mos
, dsobj
, DS_FIELD_REMAP_DEADLIST
,
1540 sizeof (remap_deadlist_obj
), 1, &remap_deadlist_obj
, tx
));
1543 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1544 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1545 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1546 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1548 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1549 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1551 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1552 snapname
, 8, 1, &dsobj
, tx
));
1555 dsl_dataset_rele(ds
->ds_prev
, ds
);
1556 VERIFY0(dsl_dataset_hold_obj(dp
,
1557 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1559 dsl_scan_ds_snapshotted(ds
, tx
);
1561 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1563 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1567 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1569 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1570 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1573 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1574 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1577 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1579 name
= nvpair_name(pair
);
1580 atp
= strchr(name
, '@');
1581 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1582 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1584 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1585 if (ddsa
->ddsa_props
!= NULL
) {
1586 dsl_props_set_sync_impl(ds
->ds_prev
,
1587 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1589 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1590 dsl_dataset_rele(ds
, FTAG
);
1595 * The snapshots must all be in the same pool.
1596 * All-or-nothing: if there are any failures, nothing will be modified.
1599 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1601 dsl_dataset_snapshot_arg_t ddsa
;
1603 boolean_t needsuspend
;
1607 nvlist_t
*suspended
= NULL
;
1609 pair
= nvlist_next_nvpair(snaps
, NULL
);
1612 firstname
= nvpair_name(pair
);
1614 error
= spa_open(firstname
, &spa
, FTAG
);
1617 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1618 spa_close(spa
, FTAG
);
1621 suspended
= fnvlist_alloc();
1622 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1623 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1624 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1625 char *snapname
= nvpair_name(pair
);
1629 atp
= strchr(snapname
, '@');
1631 error
= SET_ERROR(EINVAL
);
1634 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1636 error
= zil_suspend(fsname
, &cookie
);
1639 fnvlist_add_uint64(suspended
, fsname
,
1644 ddsa
.ddsa_snaps
= snaps
;
1645 ddsa
.ddsa_props
= props
;
1646 ddsa
.ddsa_errors
= errors
;
1647 ddsa
.ddsa_cr
= CRED();
1650 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1651 dsl_dataset_snapshot_sync
, &ddsa
,
1652 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1655 if (suspended
!= NULL
) {
1656 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1657 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1658 zil_resume((void *)(uintptr_t)
1659 fnvpair_value_uint64(pair
));
1661 fnvlist_free(suspended
);
1667 typedef struct dsl_dataset_snapshot_tmp_arg
{
1668 const char *ddsta_fsname
;
1669 const char *ddsta_snapname
;
1670 minor_t ddsta_cleanup_minor
;
1671 const char *ddsta_htag
;
1672 } dsl_dataset_snapshot_tmp_arg_t
;
1675 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1677 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1678 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1682 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1686 /* NULL cred means no limit check for tmp snapshot */
1687 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1688 tx
, B_FALSE
, 0, NULL
);
1690 dsl_dataset_rele(ds
, FTAG
);
1694 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1695 dsl_dataset_rele(ds
, FTAG
);
1696 return (SET_ERROR(ENOTSUP
));
1698 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1701 dsl_dataset_rele(ds
, FTAG
);
1705 dsl_dataset_rele(ds
, FTAG
);
1710 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1712 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1713 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1714 dsl_dataset_t
*ds
= NULL
;
1716 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1718 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1719 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1720 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1721 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1723 dsl_dataset_rele(ds
, FTAG
);
1727 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1728 minor_t cleanup_minor
, const char *htag
)
1730 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1733 boolean_t needsuspend
;
1736 ddsta
.ddsta_fsname
= fsname
;
1737 ddsta
.ddsta_snapname
= snapname
;
1738 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1739 ddsta
.ddsta_htag
= htag
;
1741 error
= spa_open(fsname
, &spa
, FTAG
);
1744 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1745 spa_close(spa
, FTAG
);
1748 error
= zil_suspend(fsname
, &cookie
);
1753 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1754 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1762 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1764 ASSERT(dmu_tx_is_syncing(tx
));
1765 ASSERT(ds
->ds_objset
!= NULL
);
1766 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1769 * in case we had to change ds_fsid_guid when we opened it,
1772 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1773 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1775 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1776 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1777 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1778 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1779 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1780 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1781 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1782 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1783 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1784 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1785 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1786 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1787 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1790 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1792 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1793 if (ds
->ds_feature_activation_needed
[f
]) {
1794 if (ds
->ds_feature_inuse
[f
])
1796 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1797 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1803 deadlist_enqueue_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
1805 dsl_deadlist_t
*dl
= arg
;
1806 dsl_deadlist_insert(dl
, bp
, tx
);
1811 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1813 objset_t
*os
= ds
->ds_objset
;
1815 bplist_iterate(&ds
->ds_pending_deadlist
,
1816 deadlist_enqueue_cb
, &ds
->ds_deadlist
, tx
);
1818 if (os
->os_synced_dnodes
!= NULL
) {
1819 multilist_destroy(os
->os_synced_dnodes
);
1820 os
->os_synced_dnodes
= NULL
;
1823 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
1825 dmu_buf_rele(ds
->ds_dbuf
, ds
);
1829 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
1832 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1836 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1839 * There may be missing entries in ds_next_clones_obj
1840 * due to a bug in a previous version of the code.
1841 * Only trust it if it has the right number of entries.
1843 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1844 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1847 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
1850 for (zap_cursor_init(&zc
, mos
,
1851 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1852 zap_cursor_retrieve(&zc
, &za
) == 0;
1853 zap_cursor_advance(&zc
)) {
1854 dsl_dataset_t
*clone
;
1855 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1856 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1857 za
.za_first_integer
, FTAG
, &clone
));
1858 dsl_dir_name(clone
->ds_dir
, buf
);
1859 fnvlist_add_boolean(val
, buf
);
1860 dsl_dataset_rele(clone
, FTAG
);
1862 zap_cursor_fini(&zc
);
1867 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1869 nvlist_t
*propval
= fnvlist_alloc();
1873 * We use nvlist_alloc() instead of fnvlist_alloc() because the
1874 * latter would allocate the list with NV_UNIQUE_NAME flag.
1875 * As a result, every time a clone name is appended to the list
1876 * it would be (linearly) searched for for a duplicate name.
1877 * We already know that all clone names must be unique and we
1878 * want avoid the quadratic complexity of double-checking that
1879 * because we can have a large number of clones.
1881 VERIFY0(nvlist_alloc(&val
, 0, KM_SLEEP
));
1883 if (get_clones_stat_impl(ds
, val
) == 0) {
1884 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1885 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
1890 nvlist_free(propval
);
1894 * Returns a string that represents the receive resume stats token. It should
1895 * be freed with strfree().
1898 get_receive_resume_stats_impl(dsl_dataset_t
*ds
)
1900 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1902 if (dsl_dataset_has_resume_receive_state(ds
)) {
1905 uint8_t *compressed
;
1907 nvlist_t
*token_nv
= fnvlist_alloc();
1908 size_t packed_size
, compressed_size
;
1910 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1911 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1912 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1914 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1915 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1916 fnvlist_add_uint64(token_nv
, "object", val
);
1918 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1919 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1920 fnvlist_add_uint64(token_nv
, "offset", val
);
1922 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1923 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1924 fnvlist_add_uint64(token_nv
, "bytes", val
);
1926 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1927 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1928 fnvlist_add_uint64(token_nv
, "toguid", val
);
1930 char buf
[MAXNAMELEN
];
1931 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1932 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1933 fnvlist_add_string(token_nv
, "toname", buf
);
1935 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1936 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1937 fnvlist_add_boolean(token_nv
, "largeblockok");
1939 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1940 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1941 fnvlist_add_boolean(token_nv
, "embedok");
1943 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1944 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1945 fnvlist_add_boolean(token_nv
, "compressok");
1947 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1948 DS_FIELD_RESUME_RAWOK
) == 0) {
1949 fnvlist_add_boolean(token_nv
, "rawok");
1951 packed
= fnvlist_pack(token_nv
, &packed_size
);
1952 fnvlist_free(token_nv
);
1953 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1955 compressed_size
= gzip_compress(packed
, compressed
,
1956 packed_size
, packed_size
, 6);
1959 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1961 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1962 for (int i
= 0; i
< compressed_size
; i
++) {
1963 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1965 str
[compressed_size
* 2] = '\0';
1966 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
1967 ZFS_SEND_RESUME_TOKEN_VERSION
,
1968 (longlong_t
)cksum
.zc_word
[0],
1969 (longlong_t
)packed_size
, str
);
1970 kmem_free(packed
, packed_size
);
1971 kmem_free(str
, compressed_size
* 2 + 1);
1972 kmem_free(compressed
, packed_size
);
1975 return (strdup(""));
1979 * Returns a string that represents the receive resume stats token of the
1980 * dataset's child. It should be freed with strfree().
1983 get_child_receive_stats(dsl_dataset_t
*ds
)
1985 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
1986 dsl_dataset_t
*recv_ds
;
1987 dsl_dataset_name(ds
, recvname
);
1988 if (strlcat(recvname
, "/", sizeof (recvname
)) <
1989 sizeof (recvname
) &&
1990 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
1991 sizeof (recvname
) &&
1992 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, recvname
, FTAG
,
1994 char *propval
= get_receive_resume_stats_impl(recv_ds
);
1995 dsl_dataset_rele(recv_ds
, FTAG
);
1998 return (strdup(""));
2002 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2004 char *propval
= get_receive_resume_stats_impl(ds
);
2005 if (strcmp(propval
, "") != 0) {
2006 dsl_prop_nvlist_add_string(nv
,
2007 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
2009 char *childval
= get_child_receive_stats(ds
);
2010 if (strcmp(childval
, "") != 0) {
2011 dsl_prop_nvlist_add_string(nv
,
2012 ZFS_PROP_RECEIVE_RESUME_TOKEN
, childval
);
2020 dsl_get_refratio(dsl_dataset_t
*ds
)
2022 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
2023 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
2024 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
2029 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
2031 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
2035 dsl_get_compressratio(dsl_dataset_t
*ds
)
2037 if (ds
->ds_is_snapshot
) {
2038 return (dsl_get_refratio(ds
));
2040 dsl_dir_t
*dd
= ds
->ds_dir
;
2041 mutex_enter(&dd
->dd_lock
);
2042 uint64_t val
= dsl_dir_get_compressratio(dd
);
2043 mutex_exit(&dd
->dd_lock
);
2049 dsl_get_used(dsl_dataset_t
*ds
)
2051 if (ds
->ds_is_snapshot
) {
2052 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2054 dsl_dir_t
*dd
= ds
->ds_dir
;
2055 mutex_enter(&dd
->dd_lock
);
2056 uint64_t val
= dsl_dir_get_used(dd
);
2057 mutex_exit(&dd
->dd_lock
);
2063 dsl_get_creation(dsl_dataset_t
*ds
)
2065 return (dsl_dataset_phys(ds
)->ds_creation_time
);
2069 dsl_get_creationtxg(dsl_dataset_t
*ds
)
2071 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
2075 dsl_get_refquota(dsl_dataset_t
*ds
)
2077 return (ds
->ds_quota
);
2081 dsl_get_refreservation(dsl_dataset_t
*ds
)
2083 return (ds
->ds_reserved
);
2087 dsl_get_guid(dsl_dataset_t
*ds
)
2089 return (dsl_dataset_phys(ds
)->ds_guid
);
2093 dsl_get_unique(dsl_dataset_t
*ds
)
2095 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2099 dsl_get_objsetid(dsl_dataset_t
*ds
)
2101 return (ds
->ds_object
);
2105 dsl_get_userrefs(dsl_dataset_t
*ds
)
2107 return (ds
->ds_userrefs
);
2111 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2113 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2117 dsl_get_referenced(dsl_dataset_t
*ds
)
2119 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2123 dsl_get_numclones(dsl_dataset_t
*ds
)
2125 ASSERT(ds
->ds_is_snapshot
);
2126 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2130 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2132 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2137 dsl_get_available(dsl_dataset_t
*ds
)
2139 uint64_t refdbytes
= dsl_get_referenced(ds
);
2140 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2142 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2144 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2146 if (ds
->ds_quota
!= 0) {
2148 * Adjust available bytes according to refquota
2150 if (refdbytes
< ds
->ds_quota
) {
2151 availbytes
= MIN(availbytes
,
2152 ds
->ds_quota
- refdbytes
);
2157 return (availbytes
);
2161 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2163 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2164 dsl_dataset_t
*prev
;
2165 int err
= dsl_dataset_hold_obj(dp
,
2166 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2168 uint64_t comp
, uncomp
;
2169 err
= dsl_dataset_space_written(prev
, ds
, written
,
2171 dsl_dataset_rele(prev
, FTAG
);
2177 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2180 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2182 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2183 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2184 dsl_dataset_name(ds
->ds_prev
, snap
);
2192 * Returns the mountpoint property and source for the given dataset in the value
2193 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2194 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2195 * Returns 0 on success and an error on failure.
2198 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2202 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2204 /* Retrieve the mountpoint value stored in the zap opbject */
2205 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2206 ZAP_MAXVALUELEN
, value
, source
);
2212 * Process the dsname and source to find the full mountpoint string.
2213 * Can be skipped for 'legacy' or 'none'.
2215 if (value
[0] == '/') {
2216 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2218 const char *relpath
;
2221 * If we inherit the mountpoint, even from a dataset
2222 * with a received value, the source will be the path of
2223 * the dataset we inherit from. If source is
2224 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2227 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2230 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2231 relpath
= dsname
+ strlen(source
);
2232 if (relpath
[0] == '/')
2236 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2239 * Special case an alternate root of '/'. This will
2240 * avoid having multiple leading slashes in the
2243 if (strcmp(root
, "/") == 0)
2247 * If the mountpoint is '/' then skip over this
2248 * if we are obtaining either an alternate root or
2249 * an inherited mountpoint.
2252 if (value
[1] == '\0' && (root
[0] != '\0' ||
2253 relpath
[0] != '\0'))
2256 if (relpath
[0] == '\0') {
2257 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2260 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2261 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2264 kmem_free(buf
, ZAP_MAXVALUELEN
);
2271 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2273 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2275 ASSERT(dsl_pool_config_held(dp
));
2277 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2278 dsl_get_refratio(ds
));
2279 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2280 dsl_get_logicalreferenced(ds
));
2281 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2282 dsl_get_compressratio(ds
));
2283 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2286 if (ds
->ds_is_snapshot
) {
2287 get_clones_stat(ds
, nv
);
2289 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2290 if (dsl_get_prev_snap(ds
, buf
) == 0)
2291 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2293 dsl_dir_stats(ds
->ds_dir
, nv
);
2296 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2297 dsl_get_available(ds
));
2298 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2299 dsl_get_referenced(ds
));
2300 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2301 dsl_get_creation(ds
));
2302 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2303 dsl_get_creationtxg(ds
));
2304 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2305 dsl_get_refquota(ds
));
2306 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2307 dsl_get_refreservation(ds
));
2308 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2310 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2311 dsl_get_unique(ds
));
2312 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2313 dsl_get_objsetid(ds
));
2314 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2315 dsl_get_userrefs(ds
));
2316 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2317 dsl_get_defer_destroy(ds
));
2318 dsl_dataset_crypt_stats(ds
, nv
);
2320 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2322 if (dsl_get_written(ds
, &written
) == 0) {
2323 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2328 if (!dsl_dataset_is_snapshot(ds
)) {
2330 * A failed "newfs" (e.g. full) resumable receive leaves
2331 * the stats set on this dataset. Check here for the prop.
2333 get_receive_resume_stats(ds
, nv
);
2336 * A failed incremental resumable receive leaves the
2337 * stats set on our child named "%recv". Check the child
2340 /* 6 extra bytes for /%recv */
2341 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2342 dsl_dataset_t
*recv_ds
;
2343 dsl_dataset_name(ds
, recvname
);
2344 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2345 sizeof (recvname
) &&
2346 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2347 sizeof (recvname
) &&
2348 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
2349 get_receive_resume_stats(recv_ds
, nv
);
2350 dsl_dataset_rele(recv_ds
, FTAG
);
2356 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2358 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2359 ASSERT(dsl_pool_config_held(dp
));
2361 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2362 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2363 stat
->dds_guid
= dsl_get_guid(ds
);
2364 stat
->dds_origin
[0] = '\0';
2365 if (ds
->ds_is_snapshot
) {
2366 stat
->dds_is_snapshot
= B_TRUE
;
2367 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2369 stat
->dds_is_snapshot
= B_FALSE
;
2370 stat
->dds_num_clones
= 0;
2372 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2373 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2379 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2381 return (ds
->ds_fsid_guid
);
2385 dsl_dataset_space(dsl_dataset_t
*ds
,
2386 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2387 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2389 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2390 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2391 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2393 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2394 if (ds
->ds_quota
!= 0) {
2396 * Adjust available bytes according to refquota
2398 if (*refdbytesp
< ds
->ds_quota
)
2399 *availbytesp
= MIN(*availbytesp
,
2400 ds
->ds_quota
- *refdbytesp
);
2404 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2405 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2406 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2407 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2411 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2413 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2416 ASSERT(dsl_pool_config_held(dp
));
2419 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2420 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2421 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2422 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2423 objset_t
*os
, *os_snap
;
2425 * It may be that only the ZIL differs, because it was
2426 * reset in the head. Don't count that as being
2429 if (dmu_objset_from_ds(ds
, &os
) != 0)
2431 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2433 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2434 &os_snap
->os_phys
->os_meta_dnode
,
2435 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2440 typedef struct dsl_dataset_rename_snapshot_arg
{
2441 const char *ddrsa_fsname
;
2442 const char *ddrsa_oldsnapname
;
2443 const char *ddrsa_newsnapname
;
2444 boolean_t ddrsa_recursive
;
2446 } dsl_dataset_rename_snapshot_arg_t
;
2450 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2451 dsl_dataset_t
*hds
, void *arg
)
2453 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2457 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2459 /* ignore nonexistent snapshots */
2460 return (error
== ENOENT
? 0 : error
);
2463 /* new name should not exist */
2464 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2466 error
= SET_ERROR(EEXIST
);
2467 else if (error
== ENOENT
)
2470 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2471 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2472 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2473 error
= SET_ERROR(ENAMETOOLONG
);
2479 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2481 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2482 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2486 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2490 if (ddrsa
->ddrsa_recursive
) {
2491 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2492 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2495 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2497 dsl_dataset_rele(hds
, FTAG
);
2502 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2503 dsl_dataset_t
*hds
, void *arg
)
2505 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2508 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2511 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2512 ASSERT(error
== 0 || error
== ENOENT
);
2513 if (error
== ENOENT
) {
2514 /* ignore nonexistent snapshots */
2518 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2520 /* log before we change the name */
2521 spa_history_log_internal_ds(ds
, "rename", tx
,
2522 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2524 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2526 mutex_enter(&ds
->ds_lock
);
2527 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2528 sizeof (ds
->ds_snapname
));
2529 mutex_exit(&ds
->ds_lock
);
2530 VERIFY0(zap_add(dp
->dp_meta_objset
,
2531 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2532 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2533 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2534 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2536 dsl_dataset_rele(ds
, FTAG
);
2541 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2543 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2544 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2545 dsl_dataset_t
*hds
= NULL
;
2547 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2548 ddrsa
->ddrsa_tx
= tx
;
2549 if (ddrsa
->ddrsa_recursive
) {
2550 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2551 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2554 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2556 dsl_dataset_rele(hds
, FTAG
);
2560 dsl_dataset_rename_snapshot(const char *fsname
,
2561 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2563 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2565 ddrsa
.ddrsa_fsname
= fsname
;
2566 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2567 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2568 ddrsa
.ddrsa_recursive
= recursive
;
2570 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2571 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2572 1, ZFS_SPACE_CHECK_RESERVED
));
2576 * If we're doing an ownership handoff, we need to make sure that there is
2577 * only one long hold on the dataset. We're not allowed to change anything here
2578 * so we don't permanently release the long hold or regular hold here. We want
2579 * to do this only when syncing to avoid the dataset unexpectedly going away
2580 * when we release the long hold.
2583 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2587 if (!dmu_tx_is_syncing(tx
))
2590 if (owner
!= NULL
) {
2591 VERIFY3P(ds
->ds_owner
, ==, owner
);
2592 dsl_dataset_long_rele(ds
, owner
);
2595 held
= dsl_dataset_long_held(ds
);
2598 dsl_dataset_long_hold(ds
, owner
);
2601 return (SET_ERROR(EBUSY
));
2607 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2609 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2610 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2612 int64_t unused_refres_delta
;
2615 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2619 /* must not be a snapshot */
2620 if (ds
->ds_is_snapshot
) {
2621 dsl_dataset_rele(ds
, FTAG
);
2622 return (SET_ERROR(EINVAL
));
2625 /* must have a most recent snapshot */
2626 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2627 dsl_dataset_rele(ds
, FTAG
);
2628 return (SET_ERROR(ESRCH
));
2632 * No rollback to a snapshot created in the current txg, because
2633 * the rollback may dirty the dataset and create blocks that are
2634 * not reachable from the rootbp while having a birth txg that
2635 * falls into the snapshot's range.
2637 if (dmu_tx_is_syncing(tx
) &&
2638 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
2639 dsl_dataset_rele(ds
, FTAG
);
2640 return (SET_ERROR(EAGAIN
));
2644 * If the expected target snapshot is specified, then check that
2645 * the latest snapshot is it.
2647 if (ddra
->ddra_tosnap
!= NULL
) {
2648 dsl_dataset_t
*snapds
;
2650 /* Check if the target snapshot exists at all. */
2651 error
= dsl_dataset_hold(dp
, ddra
->ddra_tosnap
, FTAG
, &snapds
);
2654 * ESRCH is used to signal that the target snapshot does
2655 * not exist, while ENOENT is used to report that
2656 * the rolled back dataset does not exist.
2657 * ESRCH is also used to cover other cases where the
2658 * target snapshot is not related to the dataset being
2659 * rolled back such as being in a different pool.
2661 if (error
== ENOENT
|| error
== EXDEV
)
2662 error
= SET_ERROR(ESRCH
);
2663 dsl_dataset_rele(ds
, FTAG
);
2666 ASSERT(snapds
->ds_is_snapshot
);
2668 /* Check if the snapshot is the latest snapshot indeed. */
2669 if (snapds
!= ds
->ds_prev
) {
2671 * Distinguish between the case where the only problem
2672 * is intervening snapshots (EEXIST) vs the snapshot
2673 * not being a valid target for rollback (ESRCH).
2675 if (snapds
->ds_dir
== ds
->ds_dir
||
2676 (dsl_dir_is_clone(ds
->ds_dir
) &&
2677 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
==
2678 snapds
->ds_object
)) {
2679 error
= SET_ERROR(EEXIST
);
2681 error
= SET_ERROR(ESRCH
);
2683 dsl_dataset_rele(snapds
, FTAG
);
2684 dsl_dataset_rele(ds
, FTAG
);
2687 dsl_dataset_rele(snapds
, FTAG
);
2690 /* must not have any bookmarks after the most recent snapshot */
2691 nvlist_t
*proprequest
= fnvlist_alloc();
2692 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2693 nvlist_t
*bookmarks
= fnvlist_alloc();
2694 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2695 fnvlist_free(proprequest
);
2697 dsl_dataset_rele(ds
, FTAG
);
2700 for (nvpair_t
*pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2701 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2703 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2704 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2705 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2706 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2707 fnvlist_free(bookmarks
);
2708 dsl_dataset_rele(ds
, FTAG
);
2709 return (SET_ERROR(EEXIST
));
2712 fnvlist_free(bookmarks
);
2714 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2716 dsl_dataset_rele(ds
, FTAG
);
2721 * Check if the snap we are rolling back to uses more than
2724 if (ds
->ds_quota
!= 0 &&
2725 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2726 dsl_dataset_rele(ds
, FTAG
);
2727 return (SET_ERROR(EDQUOT
));
2731 * When we do the clone swap, we will temporarily use more space
2732 * due to the refreservation (the head will no longer have any
2733 * unique space, so the entire amount of the refreservation will need
2734 * to be free). We will immediately destroy the clone, freeing
2735 * this space, but the freeing happens over many txg's.
2737 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2738 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2740 if (unused_refres_delta
> 0 &&
2741 unused_refres_delta
>
2742 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2743 dsl_dataset_rele(ds
, FTAG
);
2744 return (SET_ERROR(ENOSPC
));
2747 dsl_dataset_rele(ds
, FTAG
);
2752 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2754 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2755 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2756 dsl_dataset_t
*ds
, *clone
;
2758 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2760 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2762 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2763 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2765 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2766 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
2768 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2770 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2771 dsl_dataset_zero_zil(ds
, tx
);
2773 dsl_destroy_head_sync_impl(clone
, tx
);
2775 dsl_dataset_rele(clone
, FTAG
);
2776 dsl_dataset_rele(ds
, FTAG
);
2780 * Rolls back the given filesystem or volume to the most recent snapshot.
2781 * The name of the most recent snapshot will be returned under key "target"
2782 * in the result nvlist.
2785 * - The existing dataset MUST be owned by the specified owner at entry
2786 * - Upon return, dataset will still be held by the same owner, whether we
2789 * This mode is required any time the existing filesystem is mounted. See
2790 * notes above zfs_suspend_fs() for further details.
2793 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
2796 dsl_dataset_rollback_arg_t ddra
;
2798 ddra
.ddra_fsname
= fsname
;
2799 ddra
.ddra_tosnap
= tosnap
;
2800 ddra
.ddra_owner
= owner
;
2801 ddra
.ddra_result
= result
;
2803 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2804 dsl_dataset_rollback_sync
, &ddra
,
2805 1, ZFS_SPACE_CHECK_RESERVED
));
2808 struct promotenode
{
2813 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2814 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2816 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2819 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2821 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2822 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2824 struct promotenode
*snap
;
2825 dsl_dataset_t
*origin_ds
;
2829 size_t max_snap_len
;
2830 boolean_t conflicting_snaps
;
2832 err
= promote_hold(ddpa
, dp
, FTAG
);
2836 hds
= ddpa
->ddpa_clone
;
2837 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2839 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2840 promote_rele(ddpa
, FTAG
);
2841 return (SET_ERROR(EXDEV
));
2844 snap
= list_head(&ddpa
->shared_snaps
);
2846 err
= SET_ERROR(ENOENT
);
2849 origin_ds
= snap
->ds
;
2852 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
2853 * When doing a promote we must make sure the encryption root for
2854 * both the target and the target's origin does not change to avoid
2855 * needing to rewrap encryption keys
2857 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
2862 * Compute and check the amount of space to transfer. Since this is
2863 * so expensive, don't do the preliminary check.
2865 if (!dmu_tx_is_syncing(tx
)) {
2866 promote_rele(ddpa
, FTAG
);
2870 /* compute origin's new unique space */
2871 snap
= list_tail(&ddpa
->clone_snaps
);
2872 ASSERT(snap
!= NULL
);
2873 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2874 origin_ds
->ds_object
);
2875 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2876 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2877 &ddpa
->unique
, &unused
, &unused
);
2880 * Walk the snapshots that we are moving
2882 * Compute space to transfer. Consider the incremental changes
2883 * to used by each snapshot:
2884 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2885 * So each snapshot gave birth to:
2886 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2887 * So a sequence would look like:
2888 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2889 * Which simplifies to:
2890 * uN + kN + kN-1 + ... + k1 + k0
2891 * Note however, if we stop before we reach the ORIGIN we get:
2892 * uN + kN + kN-1 + ... + kM - uM-1
2894 conflicting_snaps
= B_FALSE
;
2896 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2897 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2898 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2899 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2900 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2901 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2902 dsl_dataset_t
*ds
= snap
->ds
;
2907 * If there are long holds, we won't be able to evict
2910 if (dsl_dataset_long_held(ds
)) {
2911 err
= SET_ERROR(EBUSY
);
2915 /* Check that the snapshot name does not conflict */
2916 VERIFY0(dsl_dataset_get_snapname(ds
));
2917 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2918 err
= SET_ERROR(ENAMETOOLONG
);
2921 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2923 fnvlist_add_boolean(ddpa
->err_ds
,
2924 snap
->ds
->ds_snapname
);
2925 conflicting_snaps
= B_TRUE
;
2926 } else if (err
!= ENOENT
) {
2930 /* The very first snapshot does not have a deadlist */
2931 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2934 dsl_deadlist_space(&ds
->ds_deadlist
,
2935 &dlused
, &dlcomp
, &dluncomp
);
2936 ddpa
->used
+= dlused
;
2937 ddpa
->comp
+= dlcomp
;
2938 ddpa
->uncomp
+= dluncomp
;
2942 * In order to return the full list of conflicting snapshots, we check
2943 * whether there was a conflict after traversing all of them.
2945 if (conflicting_snaps
) {
2946 err
= SET_ERROR(EEXIST
);
2951 * If we are a clone of a clone then we never reached ORIGIN,
2952 * so we need to subtract out the clone origin's used space.
2954 if (ddpa
->origin_origin
) {
2956 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2958 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2960 dsl_dataset_phys(ddpa
->origin_origin
)->
2961 ds_uncompressed_bytes
;
2964 /* Check that there is enough space and limit headroom here */
2965 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2966 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2971 * Compute the amounts of space that will be used by snapshots
2972 * after the promotion (for both origin and clone). For each,
2973 * it is the amount of space that will be on all of their
2974 * deadlists (that was not born before their new origin).
2976 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2980 * Note, typically this will not be a clone of a clone,
2981 * so dd_origin_txg will be < TXG_INITIAL, so
2982 * these snaplist_space() -> dsl_deadlist_space_range()
2983 * calls will be fast because they do not have to
2984 * iterate over all bps.
2986 snap
= list_head(&ddpa
->origin_snaps
);
2988 err
= SET_ERROR(ENOENT
);
2991 err
= snaplist_space(&ddpa
->shared_snaps
,
2992 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2996 err
= snaplist_space(&ddpa
->clone_snaps
,
2997 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
3000 ddpa
->cloneusedsnap
+= space
;
3002 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
3003 DD_FLAG_USED_BREAKDOWN
) {
3004 err
= snaplist_space(&ddpa
->origin_snaps
,
3005 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
3006 &ddpa
->originusedsnap
);
3012 promote_rele(ddpa
, FTAG
);
3017 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
3019 dsl_dataset_promote_arg_t
*ddpa
= arg
;
3020 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3022 struct promotenode
*snap
;
3023 dsl_dataset_t
*origin_ds
;
3024 dsl_dataset_t
*origin_head
;
3026 dsl_dir_t
*odd
= NULL
;
3027 uint64_t oldnext_obj
;
3030 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
3031 hds
= ddpa
->ddpa_clone
;
3033 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
3035 snap
= list_head(&ddpa
->shared_snaps
);
3036 origin_ds
= snap
->ds
;
3039 snap
= list_head(&ddpa
->origin_snaps
);
3040 origin_head
= snap
->ds
;
3043 * We need to explicitly open odd, since origin_ds's dd will be
3046 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
3049 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
3051 /* change origin's next snap */
3052 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
3053 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
3054 snap
= list_tail(&ddpa
->clone_snaps
);
3055 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3056 origin_ds
->ds_object
);
3057 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
3059 /* change the origin's next clone */
3060 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
3061 dsl_dataset_remove_from_next_clones(origin_ds
,
3062 snap
->ds
->ds_object
, tx
);
3063 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3064 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
3069 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
3070 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
3071 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
3072 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
3073 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
3074 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
3075 origin_head
->ds_dir
->dd_origin_txg
=
3076 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3078 /* change dd_clone entries */
3079 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3080 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3081 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
3082 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3083 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3084 hds
->ds_object
, tx
));
3086 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3087 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3088 origin_head
->ds_object
, tx
));
3089 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
3090 dsl_dir_phys(dd
)->dd_clones
=
3091 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
3092 DMU_OT_NONE
, 0, tx
);
3094 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3095 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
3098 /* move snapshots to this dir */
3099 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3100 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3101 dsl_dataset_t
*ds
= snap
->ds
;
3104 * Property callbacks are registered to a particular
3105 * dsl_dir. Since ours is changing, evict the objset
3106 * so that they will be unregistered from the old dsl_dir.
3108 if (ds
->ds_objset
) {
3109 dmu_objset_evict(ds
->ds_objset
);
3110 ds
->ds_objset
= NULL
;
3113 /* move snap name entry */
3114 VERIFY0(dsl_dataset_get_snapname(ds
));
3115 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3116 ds
->ds_snapname
, tx
, B_TRUE
));
3117 VERIFY0(zap_add(dp
->dp_meta_objset
,
3118 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3119 8, 1, &ds
->ds_object
, tx
));
3120 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3121 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3123 /* change containing dsl_dir */
3124 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3125 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3126 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3127 ASSERT3P(ds
->ds_dir
, ==, odd
);
3128 dsl_dir_rele(ds
->ds_dir
, ds
);
3129 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3130 NULL
, ds
, &ds
->ds_dir
));
3132 /* move any clone references */
3133 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3134 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3138 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3139 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3140 zap_cursor_retrieve(&zc
, &za
) == 0;
3141 zap_cursor_advance(&zc
)) {
3142 dsl_dataset_t
*cnds
;
3145 if (za
.za_first_integer
== oldnext_obj
) {
3147 * We've already moved the
3148 * origin's reference.
3153 VERIFY0(dsl_dataset_hold_obj(dp
,
3154 za
.za_first_integer
, FTAG
, &cnds
));
3155 o
= dsl_dir_phys(cnds
->ds_dir
)->
3156 dd_head_dataset_obj
;
3158 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3159 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3160 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3161 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3162 dsl_dataset_rele(cnds
, FTAG
);
3164 zap_cursor_fini(&zc
);
3167 ASSERT(!dsl_prop_hascb(ds
));
3171 * Change space accounting.
3172 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3173 * both be valid, or both be 0 (resulting in delta == 0). This
3174 * is true for each of {clone,origin} independently.
3177 delta
= ddpa
->cloneusedsnap
-
3178 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3179 ASSERT3S(delta
, >=, 0);
3180 ASSERT3U(ddpa
->used
, >=, delta
);
3181 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3182 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3183 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3185 delta
= ddpa
->originusedsnap
-
3186 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3187 ASSERT3S(delta
, <=, 0);
3188 ASSERT3U(ddpa
->used
, >=, -delta
);
3189 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3190 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3191 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3193 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3195 /* log history record */
3196 spa_history_log_internal_ds(hds
, "promote", tx
, "");
3198 dsl_dir_rele(odd
, FTAG
);
3199 promote_rele(ddpa
, FTAG
);
3203 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3204 * (exclusive) and last_obj (inclusive). The list will be in reverse
3205 * order (last_obj will be the list_head()). If first_obj == 0, do all
3206 * snapshots back to this dataset's origin.
3209 snaplist_make(dsl_pool_t
*dp
,
3210 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
3212 uint64_t obj
= last_obj
;
3214 list_create(l
, sizeof (struct promotenode
),
3215 offsetof(struct promotenode
, link
));
3217 while (obj
!= first_obj
) {
3219 struct promotenode
*snap
;
3222 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3223 ASSERT(err
!= ENOENT
);
3228 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3230 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3232 list_insert_tail(l
, snap
);
3233 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3240 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3242 struct promotenode
*snap
;
3245 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3246 uint64_t used
, comp
, uncomp
;
3247 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3248 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3255 snaplist_destroy(list_t
*l
, void *tag
)
3257 struct promotenode
*snap
;
3259 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3262 while ((snap
= list_tail(l
)) != NULL
) {
3263 list_remove(l
, snap
);
3264 dsl_dataset_rele(snap
->ds
, tag
);
3265 kmem_free(snap
, sizeof (*snap
));
3271 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
3275 struct promotenode
*snap
;
3277 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3281 dd
= ddpa
->ddpa_clone
->ds_dir
;
3283 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3284 !dsl_dir_is_clone(dd
)) {
3285 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3286 return (SET_ERROR(EINVAL
));
3289 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3290 &ddpa
->shared_snaps
, tag
);
3294 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3295 &ddpa
->clone_snaps
, tag
);
3299 snap
= list_head(&ddpa
->shared_snaps
);
3300 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3301 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3302 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3303 &ddpa
->origin_snaps
, tag
);
3307 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3308 error
= dsl_dataset_hold_obj(dp
,
3309 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3310 tag
, &ddpa
->origin_origin
);
3316 promote_rele(ddpa
, tag
);
3321 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
3323 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3324 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3325 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3326 if (ddpa
->origin_origin
!= NULL
)
3327 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3328 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3334 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3335 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3338 dsl_dataset_promote(const char *name
, char *conflsnap
)
3340 dsl_dataset_promote_arg_t ddpa
= { 0 };
3343 nvpair_t
*snap_pair
;
3347 * We will modify space proportional to the number of
3348 * snapshots. Compute numsnaps.
3350 error
= dmu_objset_hold(name
, FTAG
, &os
);
3353 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3354 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3356 dmu_objset_rele(os
, FTAG
);
3360 ddpa
.ddpa_clonename
= name
;
3361 ddpa
.err_ds
= fnvlist_alloc();
3364 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3365 dsl_dataset_promote_sync
, &ddpa
,
3366 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3369 * Return the first conflicting snapshot found.
3371 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3372 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3373 (void) strcpy(conflsnap
, nvpair_name(snap_pair
));
3375 fnvlist_free(ddpa
.err_ds
);
3380 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3381 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3384 * "slack" factor for received datasets with refquota set on them.
3385 * See the bottom of this function for details on its use.
3387 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3388 spa_asize_inflation
;
3389 int64_t unused_refres_delta
;
3391 /* they should both be heads */
3392 if (clone
->ds_is_snapshot
||
3393 origin_head
->ds_is_snapshot
)
3394 return (SET_ERROR(EINVAL
));
3396 /* if we are not forcing, the branch point should be just before them */
3397 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3398 return (SET_ERROR(EINVAL
));
3400 /* clone should be the clone (unless they are unrelated) */
3401 if (clone
->ds_prev
!= NULL
&&
3402 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3403 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3404 return (SET_ERROR(EINVAL
));
3406 /* the clone should be a child of the origin */
3407 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3408 return (SET_ERROR(EINVAL
));
3410 /* origin_head shouldn't be modified unless 'force' */
3412 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3413 return (SET_ERROR(ETXTBSY
));
3415 /* origin_head should have no long holds (e.g. is not mounted) */
3416 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3417 return (SET_ERROR(EBUSY
));
3419 /* check amount of any unconsumed refreservation */
3420 unused_refres_delta
=
3421 (int64_t)MIN(origin_head
->ds_reserved
,
3422 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3423 (int64_t)MIN(origin_head
->ds_reserved
,
3424 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3426 if (unused_refres_delta
> 0 &&
3427 unused_refres_delta
>
3428 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3429 return (SET_ERROR(ENOSPC
));
3432 * The clone can't be too much over the head's refquota.
3434 * To ensure that the entire refquota can be used, we allow one
3435 * transaction to exceed the the refquota. Therefore, this check
3436 * needs to also allow for the space referenced to be more than the
3437 * refquota. The maximum amount of space that one transaction can use
3438 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3439 * overage ensures that we are able to receive a filesystem that
3440 * exceeds the refquota on the source system.
3442 * So that overage is the refquota_slack we use below.
3444 if (origin_head
->ds_quota
!= 0 &&
3445 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3446 origin_head
->ds_quota
+ refquota_slack
)
3447 return (SET_ERROR(EDQUOT
));
3453 dsl_dataset_swap_remap_deadlists(dsl_dataset_t
*clone
,
3454 dsl_dataset_t
*origin
, dmu_tx_t
*tx
)
3456 uint64_t clone_remap_dl_obj
, origin_remap_dl_obj
;
3457 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3459 ASSERT(dsl_pool_sync_context(dp
));
3461 clone_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(clone
);
3462 origin_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(origin
);
3464 if (clone_remap_dl_obj
!= 0) {
3465 dsl_deadlist_close(&clone
->ds_remap_deadlist
);
3466 dsl_dataset_unset_remap_deadlist_object(clone
, tx
);
3468 if (origin_remap_dl_obj
!= 0) {
3469 dsl_deadlist_close(&origin
->ds_remap_deadlist
);
3470 dsl_dataset_unset_remap_deadlist_object(origin
, tx
);
3473 if (clone_remap_dl_obj
!= 0) {
3474 dsl_dataset_set_remap_deadlist_object(origin
,
3475 clone_remap_dl_obj
, tx
);
3476 dsl_deadlist_open(&origin
->ds_remap_deadlist
,
3477 dp
->dp_meta_objset
, clone_remap_dl_obj
);
3479 if (origin_remap_dl_obj
!= 0) {
3480 dsl_dataset_set_remap_deadlist_object(clone
,
3481 origin_remap_dl_obj
, tx
);
3482 dsl_deadlist_open(&clone
->ds_remap_deadlist
,
3483 dp
->dp_meta_objset
, origin_remap_dl_obj
);
3488 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
3489 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
3491 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3492 int64_t unused_refres_delta
;
3494 ASSERT(clone
->ds_reserved
== 0);
3496 * NOTE: On DEBUG kernels there could be a race between this and
3497 * the check function if spa_asize_inflation is adjusted...
3499 ASSERT(origin_head
->ds_quota
== 0 ||
3500 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
3501 DMU_MAX_ACCESS
* spa_asize_inflation
);
3502 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
3505 * Swap per-dataset feature flags.
3507 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
3508 if (!(spa_feature_table
[f
].fi_flags
&
3509 ZFEATURE_FLAG_PER_DATASET
)) {
3510 ASSERT(!clone
->ds_feature_inuse
[f
]);
3511 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
3515 boolean_t clone_inuse
= clone
->ds_feature_inuse
[f
];
3516 boolean_t origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
3519 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
3520 clone
->ds_feature_inuse
[f
] = B_FALSE
;
3522 if (origin_head_inuse
) {
3523 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
3525 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
3528 dsl_dataset_activate_feature(origin_head
->ds_object
,
3530 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
3532 if (origin_head_inuse
) {
3533 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
3534 clone
->ds_feature_inuse
[f
] = B_TRUE
;
3538 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3539 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3541 if (clone
->ds_objset
!= NULL
) {
3542 dmu_objset_evict(clone
->ds_objset
);
3543 clone
->ds_objset
= NULL
;
3546 if (origin_head
->ds_objset
!= NULL
) {
3547 dmu_objset_evict(origin_head
->ds_objset
);
3548 origin_head
->ds_objset
= NULL
;
3551 unused_refres_delta
=
3552 (int64_t)MIN(origin_head
->ds_reserved
,
3553 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3554 (int64_t)MIN(origin_head
->ds_reserved
,
3555 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3558 * Reset origin's unique bytes, if it exists.
3560 if (clone
->ds_prev
) {
3561 dsl_dataset_t
*origin
= clone
->ds_prev
;
3562 uint64_t comp
, uncomp
;
3564 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3565 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3566 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3567 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3572 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3573 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3575 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3576 dsl_dataset_phys(origin_head
)->ds_bp
=
3577 dsl_dataset_phys(clone
)->ds_bp
;
3578 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3579 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3580 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3583 /* set dd_*_bytes */
3585 int64_t dused
, dcomp
, duncomp
;
3586 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3587 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3589 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3590 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3592 dsl_deadlist_space(&clone
->ds_deadlist
,
3593 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3594 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3595 &odl_used
, &odl_comp
, &odl_uncomp
);
3597 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3599 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3601 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3603 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3605 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3607 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3610 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3611 dused
, dcomp
, duncomp
, tx
);
3612 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3613 -dused
, -dcomp
, -duncomp
, tx
);
3616 * The difference in the space used by snapshots is the
3617 * difference in snapshot space due to the head's
3618 * deadlist (since that's the only thing that's
3619 * changing that affects the snapused).
3621 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3622 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3623 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3624 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3625 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3626 &odl_used
, &odl_comp
, &odl_uncomp
);
3627 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3628 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3631 /* swap ds_*_bytes */
3632 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3633 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3634 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3635 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3636 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3637 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3638 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3639 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3641 /* apply any parent delta for change in unconsumed refreservation */
3642 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3643 unused_refres_delta
, 0, 0, tx
);
3648 dsl_deadlist_close(&clone
->ds_deadlist
);
3649 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3650 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3651 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3652 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3653 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3654 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3655 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3656 dsl_dataset_swap_remap_deadlists(clone
, origin_head
, tx
);
3658 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3660 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3661 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3665 * Given a pool name and a dataset object number in that pool,
3666 * return the name of that dataset.
3669 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3675 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3679 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3681 dsl_dataset_name(ds
, buf
);
3682 dsl_dataset_rele(ds
, FTAG
);
3684 dsl_pool_rele(dp
, FTAG
);
3690 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3691 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3695 ASSERT3S(asize
, >, 0);
3698 * *ref_rsrv is the portion of asize that will come from any
3699 * unconsumed refreservation space.
3703 mutex_enter(&ds
->ds_lock
);
3705 * Make a space adjustment for reserved bytes.
3707 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3709 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3711 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3713 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3716 if (!check_quota
|| ds
->ds_quota
== 0) {
3717 mutex_exit(&ds
->ds_lock
);
3721 * If they are requesting more space, and our current estimate
3722 * is over quota, they get to try again unless the actual
3723 * on-disk is over quota and there are no pending changes (which
3724 * may free up space for us).
3726 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3729 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3730 error
= SET_ERROR(ERESTART
);
3732 error
= SET_ERROR(EDQUOT
);
3734 mutex_exit(&ds
->ds_lock
);
3739 typedef struct dsl_dataset_set_qr_arg
{
3740 const char *ddsqra_name
;
3741 zprop_source_t ddsqra_source
;
3742 uint64_t ddsqra_value
;
3743 } dsl_dataset_set_qr_arg_t
;
3748 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3750 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3751 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3756 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3757 return (SET_ERROR(ENOTSUP
));
3759 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3763 if (ds
->ds_is_snapshot
) {
3764 dsl_dataset_rele(ds
, FTAG
);
3765 return (SET_ERROR(EINVAL
));
3768 error
= dsl_prop_predict(ds
->ds_dir
,
3769 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3770 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3772 dsl_dataset_rele(ds
, FTAG
);
3777 dsl_dataset_rele(ds
, FTAG
);
3781 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3782 newval
< ds
->ds_reserved
) {
3783 dsl_dataset_rele(ds
, FTAG
);
3784 return (SET_ERROR(ENOSPC
));
3787 dsl_dataset_rele(ds
, FTAG
);
3792 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3794 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3795 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3796 dsl_dataset_t
*ds
= NULL
;
3799 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3801 dsl_prop_set_sync_impl(ds
,
3802 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3803 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3804 &ddsqra
->ddsqra_value
, tx
);
3806 VERIFY0(dsl_prop_get_int_ds(ds
,
3807 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3809 if (ds
->ds_quota
!= newval
) {
3810 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3811 ds
->ds_quota
= newval
;
3813 dsl_dataset_rele(ds
, FTAG
);
3817 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3820 dsl_dataset_set_qr_arg_t ddsqra
;
3822 ddsqra
.ddsqra_name
= dsname
;
3823 ddsqra
.ddsqra_source
= source
;
3824 ddsqra
.ddsqra_value
= refquota
;
3826 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3827 dsl_dataset_set_refquota_sync
, &ddsqra
, 0,
3828 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
3832 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3834 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3835 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3838 uint64_t newval
, unique
;
3840 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3841 return (SET_ERROR(ENOTSUP
));
3843 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3847 if (ds
->ds_is_snapshot
) {
3848 dsl_dataset_rele(ds
, FTAG
);
3849 return (SET_ERROR(EINVAL
));
3852 error
= dsl_prop_predict(ds
->ds_dir
,
3853 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3854 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3856 dsl_dataset_rele(ds
, FTAG
);
3861 * If we are doing the preliminary check in open context, the
3862 * space estimates may be inaccurate.
3864 if (!dmu_tx_is_syncing(tx
)) {
3865 dsl_dataset_rele(ds
, FTAG
);
3869 mutex_enter(&ds
->ds_lock
);
3870 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3871 dsl_dataset_recalc_head_uniq(ds
);
3872 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3873 mutex_exit(&ds
->ds_lock
);
3875 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3876 uint64_t delta
= MAX(unique
, newval
) -
3877 MAX(unique
, ds
->ds_reserved
);
3880 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3881 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3882 dsl_dataset_rele(ds
, FTAG
);
3883 return (SET_ERROR(ENOSPC
));
3887 dsl_dataset_rele(ds
, FTAG
);
3892 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3893 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3899 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3900 source
, sizeof (value
), 1, &value
, tx
);
3902 VERIFY0(dsl_prop_get_int_ds(ds
,
3903 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3905 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3906 mutex_enter(&ds
->ds_dir
->dd_lock
);
3907 mutex_enter(&ds
->ds_lock
);
3908 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3909 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3910 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3911 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3912 ds
->ds_reserved
= newval
;
3913 mutex_exit(&ds
->ds_lock
);
3915 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3916 mutex_exit(&ds
->ds_dir
->dd_lock
);
3920 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3922 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3923 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3924 dsl_dataset_t
*ds
= NULL
;
3926 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3927 dsl_dataset_set_refreservation_sync_impl(ds
,
3928 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3929 dsl_dataset_rele(ds
, FTAG
);
3933 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3934 uint64_t refreservation
)
3936 dsl_dataset_set_qr_arg_t ddsqra
;
3938 ddsqra
.ddsqra_name
= dsname
;
3939 ddsqra
.ddsqra_source
= source
;
3940 ddsqra
.ddsqra_value
= refreservation
;
3942 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3943 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0,
3944 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
3948 * Return (in *usedp) the amount of space written in new that is not
3949 * present in oldsnap. New may be a snapshot or the head. Old must be
3950 * a snapshot before new, in new's filesystem (or its origin). If not then
3951 * fail and return EINVAL.
3953 * The written space is calculated by considering two components: First, we
3954 * ignore any freed space, and calculate the written as new's used space
3955 * minus old's used space. Next, we add in the amount of space that was freed
3956 * between the two snapshots, thus reducing new's used space relative to old's.
3957 * Specifically, this is the space that was born before old->ds_creation_txg,
3958 * and freed before new (ie. on new's deadlist or a previous deadlist).
3960 * space freed [---------------------]
3961 * snapshots ---O-------O--------O-------O------
3965 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3966 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3970 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3972 ASSERT(dsl_pool_config_held(dp
));
3975 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3976 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3979 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3980 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3983 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3984 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3986 snapobj
= new->ds_object
;
3987 while (snapobj
!= oldsnap
->ds_object
) {
3988 dsl_dataset_t
*snap
;
3989 uint64_t used
, comp
, uncomp
;
3991 if (snapobj
== new->ds_object
) {
3994 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3999 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
4000 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
4002 * The blocks in the deadlist can not be born after
4003 * ds_prev_snap_txg, so get the whole deadlist space,
4004 * which is more efficient (especially for old-format
4005 * deadlists). Unfortunately the deadlist code
4006 * doesn't have enough information to make this
4007 * optimization itself.
4009 dsl_deadlist_space(&snap
->ds_deadlist
,
4010 &used
, &comp
, &uncomp
);
4012 dsl_deadlist_space_range(&snap
->ds_deadlist
,
4013 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
4014 &used
, &comp
, &uncomp
);
4021 * If we get to the beginning of the chain of snapshots
4022 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
4023 * was not a snapshot of/before new.
4025 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
4027 dsl_dataset_rele(snap
, FTAG
);
4029 err
= SET_ERROR(EINVAL
);
4038 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
4039 * lastsnap, and all snapshots in between are deleted.
4041 * blocks that would be freed [---------------------------]
4042 * snapshots ---O-------O--------O-------O--------O
4043 * firstsnap lastsnap
4045 * This is the set of blocks that were born after the snap before firstsnap,
4046 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
4047 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
4048 * We calculate this by iterating over the relevant deadlists (from the snap
4049 * after lastsnap, backward to the snap after firstsnap), summing up the
4050 * space on the deadlist that was born after the snap before firstsnap.
4053 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
4054 dsl_dataset_t
*lastsnap
,
4055 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4059 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
4061 ASSERT(firstsnap
->ds_is_snapshot
);
4062 ASSERT(lastsnap
->ds_is_snapshot
);
4065 * Check that the snapshots are in the same dsl_dir, and firstsnap
4066 * is before lastsnap.
4068 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
4069 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
4070 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
4071 return (SET_ERROR(EINVAL
));
4073 *usedp
= *compp
= *uncompp
= 0;
4075 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
4076 while (snapobj
!= firstsnap
->ds_object
) {
4078 uint64_t used
, comp
, uncomp
;
4080 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
4084 dsl_deadlist_space_range(&ds
->ds_deadlist
,
4085 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
4086 &used
, &comp
, &uncomp
);
4091 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4092 ASSERT3U(snapobj
, !=, 0);
4093 dsl_dataset_rele(ds
, FTAG
);
4099 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
4100 * For example, they could both be snapshots of the same filesystem, and
4101 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
4102 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
4103 * filesystem. Or 'earlier' could be the origin's origin.
4105 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
4108 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
4109 uint64_t earlier_txg
)
4111 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
4115 ASSERT(dsl_pool_config_held(dp
));
4116 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
4118 if (earlier_txg
== 0)
4119 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
4121 if (later
->ds_is_snapshot
&&
4122 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
4125 if (later
->ds_dir
== earlier
->ds_dir
)
4127 if (!dsl_dir_is_clone(later
->ds_dir
))
4130 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
4132 dsl_dataset_t
*origin
;
4133 error
= dsl_dataset_hold_obj(dp
,
4134 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
4137 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
4138 dsl_dataset_rele(origin
, FTAG
);
4143 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4145 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4146 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4150 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4152 dmu_object_info_t doi
;
4154 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4155 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4159 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4161 return (dsl_dataset_is_zapified(ds
) &&
4162 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4163 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4167 dsl_dataset_get_remap_deadlist_object(dsl_dataset_t
*ds
)
4169 uint64_t remap_deadlist_obj
;
4172 if (!dsl_dataset_is_zapified(ds
))
4175 err
= zap_lookup(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4176 DS_FIELD_REMAP_DEADLIST
, sizeof (remap_deadlist_obj
), 1,
4177 &remap_deadlist_obj
);
4180 VERIFY3S(err
, ==, ENOENT
);
4184 ASSERT(remap_deadlist_obj
!= 0);
4185 return (remap_deadlist_obj
);
4189 dsl_dataset_remap_deadlist_exists(dsl_dataset_t
*ds
)
4191 EQUIV(dsl_deadlist_is_open(&ds
->ds_remap_deadlist
),
4192 dsl_dataset_get_remap_deadlist_object(ds
) != 0);
4193 return (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
));
4197 dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
, uint64_t obj
,
4201 dsl_dataset_zapify(ds
, tx
);
4202 VERIFY0(zap_add(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4203 DS_FIELD_REMAP_DEADLIST
, sizeof (obj
), 1, &obj
, tx
));
4207 dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4209 VERIFY0(zap_remove(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4210 ds
->ds_object
, DS_FIELD_REMAP_DEADLIST
, tx
));
4214 dsl_dataset_destroy_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4216 uint64_t remap_deadlist_object
;
4217 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4219 ASSERT(dmu_tx_is_syncing(tx
));
4220 ASSERT(dsl_dataset_remap_deadlist_exists(ds
));
4222 remap_deadlist_object
= ds
->ds_remap_deadlist
.dl_object
;
4223 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
4224 dsl_deadlist_free(spa_meta_objset(spa
), remap_deadlist_object
, tx
);
4225 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
4226 spa_feature_decr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4230 dsl_dataset_create_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4232 uint64_t remap_deadlist_obj
;
4233 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4235 ASSERT(dmu_tx_is_syncing(tx
));
4236 ASSERT(MUTEX_HELD(&ds
->ds_remap_deadlist_lock
));
4238 * Currently we only create remap deadlists when there are indirect
4239 * vdevs with referenced mappings.
4241 ASSERT(spa_feature_is_active(spa
, SPA_FEATURE_DEVICE_REMOVAL
));
4243 remap_deadlist_obj
= dsl_deadlist_clone(
4244 &ds
->ds_deadlist
, UINT64_MAX
,
4245 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
4246 dsl_dataset_set_remap_deadlist_object(ds
,
4247 remap_deadlist_obj
, tx
);
4248 dsl_deadlist_open(&ds
->ds_remap_deadlist
, spa_meta_objset(spa
),
4249 remap_deadlist_obj
);
4250 spa_feature_incr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4253 #if defined(_KERNEL)
4255 module_param(zfs_max_recordsize
, int, 0644);
4256 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4258 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
4259 module_param(zfs_max_recordsize
, int, 0444);
4260 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4263 EXPORT_SYMBOL(dsl_dataset_hold
);
4264 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4265 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4266 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4267 EXPORT_SYMBOL(dsl_dataset_own
);
4268 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4269 EXPORT_SYMBOL(dsl_dataset_name
);
4270 EXPORT_SYMBOL(dsl_dataset_rele
);
4271 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4272 EXPORT_SYMBOL(dsl_dataset_disown
);
4273 EXPORT_SYMBOL(dsl_dataset_tryown
);
4274 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4275 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4276 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4277 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4278 EXPORT_SYMBOL(dsl_dataset_promote
);
4279 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4280 EXPORT_SYMBOL(dsl_dataset_user_release
);
4281 EXPORT_SYMBOL(dsl_dataset_get_holds
);
4282 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
4283 EXPORT_SYMBOL(dsl_dataset_get_spa
);
4284 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
4285 EXPORT_SYMBOL(dsl_dataset_space_written
);
4286 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
4287 EXPORT_SYMBOL(dsl_dataset_sync
);
4288 EXPORT_SYMBOL(dsl_dataset_block_born
);
4289 EXPORT_SYMBOL(dsl_dataset_block_kill
);
4290 EXPORT_SYMBOL(dsl_dataset_dirty
);
4291 EXPORT_SYMBOL(dsl_dataset_stats
);
4292 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
4293 EXPORT_SYMBOL(dsl_dataset_space
);
4294 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
4295 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
4296 EXPORT_SYMBOL(dsl_dataset_check_quota
);
4297 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
4298 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);