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/zfs_znode.h>
50 #include <sys/zfs_onexit.h>
52 #include <sys/dsl_scan.h>
53 #include <sys/dsl_deadlist.h>
54 #include <sys/dsl_destroy.h>
55 #include <sys/dsl_userhold.h>
56 #include <sys/dsl_bookmark.h>
57 #include <sys/policy.h>
58 #include <sys/dmu_send.h>
59 #include <sys/zio_compress.h>
60 #include <zfs_fletcher.h>
61 #include <sys/zio_checksum.h>
64 * The SPA supports block sizes up to 16MB. However, very large blocks
65 * can have an impact on i/o latency (e.g. tying up a spinning disk for
66 * ~300ms), and also potentially on the memory allocator. Therefore,
67 * we do not allow the recordsize to be set larger than zfs_max_recordsize
68 * (default 1MB). Larger blocks can be created by changing this tunable,
69 * and pools with larger blocks can always be imported and used, regardless
72 int zfs_max_recordsize
= 1 * 1024 * 1024;
74 #define SWITCH64(x, y) \
76 uint64_t __tmp = (x); \
81 #define DS_REF_MAX (1ULL << 62)
83 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
85 extern int spa_asize_inflation
;
87 static zil_header_t zero_zil
;
90 * Figure out how much of this delta should be propagated to the dsl_dir
91 * layer. If there's a refreservation, that space has already been
92 * partially accounted for in our ancestors.
95 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
97 dsl_dataset_phys_t
*ds_phys
;
98 uint64_t old_bytes
, new_bytes
;
100 if (ds
->ds_reserved
== 0)
103 ds_phys
= dsl_dataset_phys(ds
);
104 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
105 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
107 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
108 return (new_bytes
- old_bytes
);
112 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
114 int used
, compressed
, uncompressed
;
117 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
118 compressed
= BP_GET_PSIZE(bp
);
119 uncompressed
= BP_GET_UCSIZE(bp
);
121 dprintf_bp(bp
, "ds=%p", ds
);
123 ASSERT(dmu_tx_is_syncing(tx
));
124 /* It could have been compressed away to nothing */
127 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
128 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
130 dsl_pool_mos_diduse_space(tx
->tx_pool
,
131 used
, compressed
, uncompressed
);
135 ASSERT3U(bp
->blk_birth
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
136 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
137 mutex_enter(&ds
->ds_lock
);
138 delta
= parent_delta(ds
, used
);
139 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
140 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
141 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
142 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
144 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
145 ds
->ds_feature_activation_needed
[SPA_FEATURE_LARGE_BLOCKS
] =
149 spa_feature_t f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
150 if (f
!= SPA_FEATURE_NONE
)
151 ds
->ds_feature_activation_needed
[f
] = B_TRUE
;
153 mutex_exit(&ds
->ds_lock
);
154 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
155 compressed
, uncompressed
, tx
);
156 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
157 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
161 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
164 int used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
165 int compressed
= BP_GET_PSIZE(bp
);
166 int uncompressed
= BP_GET_UCSIZE(bp
);
171 ASSERT(dmu_tx_is_syncing(tx
));
172 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
175 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
176 dsl_pool_mos_diduse_space(tx
->tx_pool
,
177 -used
, -compressed
, -uncompressed
);
180 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
182 ASSERT(!ds
->ds_is_snapshot
);
183 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
185 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
188 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
189 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
191 mutex_enter(&ds
->ds_lock
);
192 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
193 !DS_UNIQUE_IS_ACCURATE(ds
));
194 delta
= parent_delta(ds
, -used
);
195 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
196 mutex_exit(&ds
->ds_lock
);
197 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
198 delta
, -compressed
, -uncompressed
, tx
);
199 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
200 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
202 dprintf_bp(bp
, "putting on dead list: %s", "");
205 * We are here as part of zio's write done callback,
206 * which means we're a zio interrupt thread. We can't
207 * call dsl_deadlist_insert() now because it may block
208 * waiting for I/O. Instead, put bp on the deferred
209 * queue and let dsl_pool_sync() finish the job.
211 bplist_append(&ds
->ds_pending_deadlist
, bp
);
213 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
215 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
216 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
217 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
218 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
219 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
220 ds
->ds_object
&& bp
->blk_birth
>
221 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
222 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
223 mutex_enter(&ds
->ds_prev
->ds_lock
);
224 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
225 mutex_exit(&ds
->ds_prev
->ds_lock
);
227 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
228 dsl_dir_transfer_space(ds
->ds_dir
, used
,
229 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
232 mutex_enter(&ds
->ds_lock
);
233 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
234 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
235 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
236 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
237 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
238 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
239 mutex_exit(&ds
->ds_lock
);
245 * We have to release the fsid syncronously or we risk that a subsequent
246 * mount of the same dataset will fail to unique_insert the fsid. This
247 * failure would manifest itself as the fsid of this dataset changing
248 * between mounts which makes NFS clients quite unhappy.
251 dsl_dataset_evict_sync(void *dbu
)
253 dsl_dataset_t
*ds
= dbu
;
255 ASSERT(ds
->ds_owner
== NULL
);
257 unique_remove(ds
->ds_fsid_guid
);
261 dsl_dataset_evict_async(void *dbu
)
263 dsl_dataset_t
*ds
= dbu
;
265 ASSERT(ds
->ds_owner
== NULL
);
269 if (ds
->ds_objset
!= NULL
)
270 dmu_objset_evict(ds
->ds_objset
);
273 dsl_dataset_rele(ds
->ds_prev
, ds
);
277 bplist_destroy(&ds
->ds_pending_deadlist
);
278 if (ds
->ds_deadlist
.dl_os
!= NULL
)
279 dsl_deadlist_close(&ds
->ds_deadlist
);
281 dsl_dir_async_rele(ds
->ds_dir
, ds
);
283 ASSERT(!list_link_active(&ds
->ds_synced_link
));
285 list_destroy(&ds
->ds_prop_cbs
);
286 mutex_destroy(&ds
->ds_lock
);
287 mutex_destroy(&ds
->ds_opening_lock
);
288 mutex_destroy(&ds
->ds_sendstream_lock
);
289 refcount_destroy(&ds
->ds_longholds
);
290 rrw_destroy(&ds
->ds_bp_rwlock
);
292 kmem_free(ds
, sizeof (dsl_dataset_t
));
296 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
298 dsl_dataset_phys_t
*headphys
;
301 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
302 objset_t
*mos
= dp
->dp_meta_objset
;
304 if (ds
->ds_snapname
[0])
306 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
309 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
313 headphys
= headdbuf
->db_data
;
314 err
= zap_value_search(dp
->dp_meta_objset
,
315 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
316 if (err
!= 0 && zfs_recover
== B_TRUE
) {
318 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
319 "SNAPOBJ=%llu-ERR=%d",
320 (unsigned long long)ds
->ds_object
, err
);
322 dmu_buf_rele(headdbuf
, FTAG
);
327 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
329 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
330 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
334 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
337 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
338 value
, mt
, NULL
, 0, NULL
);
339 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
340 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
345 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
348 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
349 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
353 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
355 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
358 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
359 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
360 err
= zap_remove(mos
, snapobj
, name
, tx
);
362 if (err
== 0 && adj_cnt
)
363 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
364 DD_FIELD_SNAPSHOT_COUNT
, tx
);
370 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
372 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
373 boolean_t result
= B_FALSE
;
375 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
376 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
378 if (ds
== dmu_buf_get_user(dbuf
))
381 dmu_buf_rele(dbuf
, tag
);
388 dsl_dataset_hold_obj_flags(dsl_pool_t
*dp
, uint64_t dsobj
,
389 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
)
391 objset_t
*mos
= dp
->dp_meta_objset
;
395 dmu_object_info_t doi
;
397 ASSERT(dsl_pool_config_held(dp
));
399 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
403 /* Make sure dsobj has the correct object type. */
404 dmu_object_info_from_db(dbuf
, &doi
);
405 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
406 dmu_buf_rele(dbuf
, tag
);
407 return (SET_ERROR(EINVAL
));
410 ds
= dmu_buf_get_user(dbuf
);
412 dsl_dataset_t
*winner
= NULL
;
414 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
416 ds
->ds_object
= dsobj
;
417 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
418 list_link_init(&ds
->ds_synced_link
);
420 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
421 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
422 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
423 rrw_init(&ds
->ds_bp_rwlock
, B_FALSE
);
424 refcount_create(&ds
->ds_longholds
);
426 bplist_create(&ds
->ds_pending_deadlist
);
427 dsl_deadlist_open(&ds
->ds_deadlist
,
428 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
430 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
431 offsetof(dmu_sendarg_t
, dsa_link
));
433 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
434 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
436 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
439 for (f
= 0; f
< SPA_FEATURES
; f
++) {
440 if (!(spa_feature_table
[f
].fi_flags
&
441 ZFEATURE_FLAG_PER_DATASET
))
443 err
= zap_contains(mos
, dsobj
,
444 spa_feature_table
[f
].fi_guid
);
446 ds
->ds_feature_inuse
[f
] = B_TRUE
;
448 ASSERT3U(err
, ==, ENOENT
);
454 err
= dsl_dir_hold_obj(dp
,
455 dsl_dataset_phys(ds
)->ds_dir_obj
, NULL
, ds
, &ds
->ds_dir
);
457 mutex_destroy(&ds
->ds_lock
);
458 mutex_destroy(&ds
->ds_opening_lock
);
459 mutex_destroy(&ds
->ds_sendstream_lock
);
460 refcount_destroy(&ds
->ds_longholds
);
461 bplist_destroy(&ds
->ds_pending_deadlist
);
462 dsl_deadlist_close(&ds
->ds_deadlist
);
463 kmem_free(ds
, sizeof (dsl_dataset_t
));
464 dmu_buf_rele(dbuf
, tag
);
468 if (!ds
->ds_is_snapshot
) {
469 ds
->ds_snapname
[0] = '\0';
470 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
471 err
= dsl_dataset_hold_obj(dp
,
472 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
475 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
476 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
477 DS_FIELD_BOOKMARK_NAMES
,
478 sizeof (ds
->ds_bookmarks
), 1,
480 if (zaperr
!= ENOENT
)
484 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
485 err
= dsl_dataset_get_snapname(ds
);
487 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
489 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
490 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
495 if (err
== 0 && !ds
->ds_is_snapshot
) {
496 err
= dsl_prop_get_int_ds(ds
,
497 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
500 err
= dsl_prop_get_int_ds(ds
,
501 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
505 ds
->ds_reserved
= ds
->ds_quota
= 0;
508 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict_sync
,
509 dsl_dataset_evict_async
, &ds
->ds_dbuf
);
511 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
513 if (err
!= 0 || winner
!= NULL
) {
514 bplist_destroy(&ds
->ds_pending_deadlist
);
515 dsl_deadlist_close(&ds
->ds_deadlist
);
517 dsl_dataset_rele(ds
->ds_prev
, ds
);
518 dsl_dir_rele(ds
->ds_dir
, ds
);
519 mutex_destroy(&ds
->ds_lock
);
520 mutex_destroy(&ds
->ds_opening_lock
);
521 mutex_destroy(&ds
->ds_sendstream_lock
);
522 refcount_destroy(&ds
->ds_longholds
);
523 kmem_free(ds
, sizeof (dsl_dataset_t
));
525 dmu_buf_rele(dbuf
, tag
);
531 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
532 if (ds
->ds_fsid_guid
!=
533 dsl_dataset_phys(ds
)->ds_fsid_guid
) {
534 zfs_dbgmsg("ds_fsid_guid changed from "
535 "%llx to %llx for pool %s dataset id %llu",
537 dsl_dataset_phys(ds
)->ds_fsid_guid
,
538 (long long)ds
->ds_fsid_guid
,
539 spa_name(dp
->dp_spa
),
544 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
545 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
546 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
547 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
548 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
551 if ((flags
& DS_HOLD_FLAG_DECRYPT
) && ds
->ds_dir
->dd_crypto_obj
!= 0) {
552 err
= spa_keystore_create_mapping(dp
->dp_spa
, ds
, ds
);
554 dsl_dataset_rele(ds
, tag
);
555 return (SET_ERROR(EACCES
));
563 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
566 return (dsl_dataset_hold_obj_flags(dp
, dsobj
, 0, tag
, dsp
));
570 dsl_dataset_hold_flags(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
571 void *tag
, dsl_dataset_t
**dsp
)
574 const char *snapname
;
579 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
583 ASSERT(dsl_pool_config_held(dp
));
584 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
586 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
, &ds
);
588 err
= SET_ERROR(ENOENT
);
590 /* we may be looking for a snapshot */
591 if (err
== 0 && snapname
!= NULL
) {
592 dsl_dataset_t
*snap_ds
;
594 if (*snapname
++ != '@') {
595 dsl_dataset_rele_flags(ds
, flags
, tag
);
596 dsl_dir_rele(dd
, FTAG
);
597 return (SET_ERROR(ENOENT
));
600 dprintf("looking for snapshot '%s'\n", snapname
);
601 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
603 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
,
606 dsl_dataset_rele_flags(ds
, flags
, tag
);
609 mutex_enter(&snap_ds
->ds_lock
);
610 if (snap_ds
->ds_snapname
[0] == 0)
611 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
612 sizeof (snap_ds
->ds_snapname
));
613 mutex_exit(&snap_ds
->ds_lock
);
619 dsl_dir_rele(dd
, FTAG
);
624 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
, void *tag
,
627 return (dsl_dataset_hold_flags(dp
, name
, 0, tag
, dsp
));
631 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
632 void *tag
, dsl_dataset_t
**dsp
)
634 int err
= dsl_dataset_hold_obj_flags(dp
, dsobj
, flags
, tag
, dsp
);
637 if (!dsl_dataset_tryown(*dsp
, tag
)) {
638 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
640 return (SET_ERROR(EBUSY
));
646 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
647 void *tag
, dsl_dataset_t
**dsp
)
649 int err
= dsl_dataset_hold_flags(dp
, name
, flags
, tag
, dsp
);
652 if (!dsl_dataset_tryown(*dsp
, tag
)) {
653 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
654 return (SET_ERROR(EBUSY
));
660 * See the comment above dsl_pool_hold() for details. In summary, a long
661 * hold is used to prevent destruction of a dataset while the pool hold
662 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
664 * The dataset and pool must be held when this function is called. After it
665 * is called, the pool hold may be released while the dataset is still held
669 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
671 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
672 (void) refcount_add(&ds
->ds_longholds
, tag
);
676 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
678 (void) refcount_remove(&ds
->ds_longholds
, tag
);
681 /* Return B_TRUE if there are any long holds on this dataset. */
683 dsl_dataset_long_held(dsl_dataset_t
*ds
)
685 return (!refcount_is_zero(&ds
->ds_longholds
));
689 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
692 (void) strcpy(name
, "mos");
694 dsl_dir_name(ds
->ds_dir
, name
);
695 VERIFY0(dsl_dataset_get_snapname(ds
));
696 if (ds
->ds_snapname
[0]) {
697 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
698 <, ZFS_MAX_DATASET_NAME_LEN
);
700 * We use a "recursive" mutex so that we
701 * can call dprintf_ds() with ds_lock held.
703 if (!MUTEX_HELD(&ds
->ds_lock
)) {
704 mutex_enter(&ds
->ds_lock
);
705 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
706 ZFS_MAX_DATASET_NAME_LEN
), <,
707 ZFS_MAX_DATASET_NAME_LEN
);
708 mutex_exit(&ds
->ds_lock
);
710 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
711 ZFS_MAX_DATASET_NAME_LEN
), <,
712 ZFS_MAX_DATASET_NAME_LEN
);
719 dsl_dataset_namelen(dsl_dataset_t
*ds
)
721 VERIFY0(dsl_dataset_get_snapname(ds
));
722 mutex_enter(&ds
->ds_lock
);
723 int len
= strlen(ds
->ds_snapname
);
724 /* add '@' if ds is a snap */
727 len
+= dsl_dir_namelen(ds
->ds_dir
);
728 mutex_exit(&ds
->ds_lock
);
733 dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
735 if (ds
->ds_dir
!= NULL
&& ds
->ds_dir
->dd_crypto_obj
!= 0 &&
736 (flags
& DS_HOLD_FLAG_DECRYPT
)) {
737 (void) spa_keystore_remove_mapping(ds
->ds_dir
->dd_pool
->dp_spa
,
741 dmu_buf_rele(ds
->ds_dbuf
, tag
);
745 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
747 dsl_dataset_rele_flags(ds
, 0, tag
);
751 dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
753 ASSERT3P(ds
->ds_owner
, ==, tag
);
754 ASSERT(ds
->ds_dbuf
!= NULL
);
756 mutex_enter(&ds
->ds_lock
);
758 mutex_exit(&ds
->ds_lock
);
759 dsl_dataset_long_rele(ds
, tag
);
760 dsl_dataset_rele_flags(ds
, flags
, tag
);
764 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
766 boolean_t gotit
= FALSE
;
768 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
769 mutex_enter(&ds
->ds_lock
);
770 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
772 dsl_dataset_long_hold(ds
, tag
);
775 mutex_exit(&ds
->ds_lock
);
780 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
783 mutex_enter(&ds
->ds_lock
);
784 rv
= (ds
->ds_owner
!= NULL
);
785 mutex_exit(&ds
->ds_lock
);
790 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
792 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
793 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
796 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
798 spa_feature_incr(spa
, f
, tx
);
799 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
801 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
802 sizeof (zero
), 1, &zero
, tx
));
806 dsl_dataset_deactivate_feature(uint64_t dsobj
, spa_feature_t f
, dmu_tx_t
*tx
)
808 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
809 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
811 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
813 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
814 spa_feature_decr(spa
, f
, tx
);
818 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
819 dsl_crypto_params_t
*dcp
, uint64_t flags
, dmu_tx_t
*tx
)
821 dsl_pool_t
*dp
= dd
->dd_pool
;
823 dsl_dataset_phys_t
*dsphys
;
825 objset_t
*mos
= dp
->dp_meta_objset
;
828 origin
= dp
->dp_origin_snap
;
830 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
831 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
832 ASSERT(dmu_tx_is_syncing(tx
));
833 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
835 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
836 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
837 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
838 dmu_buf_will_dirty(dbuf
, tx
);
839 dsphys
= dbuf
->db_data
;
840 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
841 dsphys
->ds_dir_obj
= dd
->dd_object
;
842 dsphys
->ds_flags
= flags
;
843 dsphys
->ds_fsid_guid
= unique_create();
844 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
845 sizeof (dsphys
->ds_guid
));
846 dsphys
->ds_snapnames_zapobj
=
847 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
849 dsphys
->ds_creation_time
= gethrestime_sec();
850 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
852 if (origin
== NULL
) {
853 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
855 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
857 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
858 dsphys
->ds_prev_snap_txg
=
859 dsl_dataset_phys(origin
)->ds_creation_txg
;
860 dsphys
->ds_referenced_bytes
=
861 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
862 dsphys
->ds_compressed_bytes
=
863 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
864 dsphys
->ds_uncompressed_bytes
=
865 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
866 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
867 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
868 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
871 * Inherit flags that describe the dataset's contents
872 * (INCONSISTENT) or properties (Case Insensitive).
874 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
875 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
877 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
878 if (origin
->ds_feature_inuse
[f
])
879 dsl_dataset_activate_feature(dsobj
, f
, tx
);
882 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
883 dsl_dataset_phys(origin
)->ds_num_children
++;
885 VERIFY0(dsl_dataset_hold_obj(dp
,
886 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
888 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
889 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
890 dsl_dataset_rele(ohds
, FTAG
);
892 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
893 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
894 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
896 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
898 VERIFY0(zap_add_int(mos
,
899 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
903 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
904 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
905 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
906 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
907 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
908 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
910 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
912 VERIFY0(zap_add_int(mos
,
913 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
918 /* handle encryption */
919 dsl_dataset_create_crypt_sync(dsobj
, dd
, origin
, dcp
, tx
);
921 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
922 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
924 dmu_buf_rele(dbuf
, FTAG
);
926 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
927 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
933 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
937 VERIFY0(dmu_objset_from_ds(ds
, &os
));
938 if (bcmp(&os
->os_zil_header
, &zero_zil
, sizeof (zero_zil
)) != 0) {
939 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
942 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
943 if (os
->os_encrypted
)
944 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_TRUE
;
946 zio
= zio_root(dp
->dp_spa
, NULL
, NULL
, ZIO_FLAG_MUSTSUCCEED
);
947 dsl_dataset_sync(ds
, zio
, tx
);
948 VERIFY0(zio_wait(zio
));
950 /* dsl_dataset_sync_done will drop this reference. */
951 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
952 dsl_dataset_sync_done(ds
, tx
);
957 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
958 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
,
959 dsl_crypto_params_t
*dcp
, dmu_tx_t
*tx
)
961 dsl_pool_t
*dp
= pdd
->dd_pool
;
962 uint64_t dsobj
, ddobj
;
965 ASSERT(dmu_tx_is_syncing(tx
));
966 ASSERT(lastname
[0] != '@');
968 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
969 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
971 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
, dcp
,
972 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
974 dsl_deleg_set_create_perms(dd
, tx
, cr
);
977 * Since we're creating a new node we know it's a leaf, so we can
978 * initialize the counts if the limit feature is active.
980 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
982 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
984 dsl_dir_zapify(dd
, tx
);
985 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
986 sizeof (cnt
), 1, &cnt
, tx
));
987 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
988 sizeof (cnt
), 1, &cnt
, tx
));
991 dsl_dir_rele(dd
, FTAG
);
994 * If we are creating a clone, make sure we zero out any stale
995 * data from the origin snapshots zil header.
997 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
1000 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
1001 dsl_dataset_zero_zil(ds
, tx
);
1002 dsl_dataset_rele(ds
, FTAG
);
1009 * The unique space in the head dataset can be calculated by subtracting
1010 * the space used in the most recent snapshot, that is still being used
1011 * in this file system, from the space currently in use. To figure out
1012 * the space in the most recent snapshot still in use, we need to take
1013 * the total space used in the snapshot and subtract out the space that
1014 * has been freed up since the snapshot was taken.
1017 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1020 uint64_t dlused
, dlcomp
, dluncomp
;
1022 ASSERT(!ds
->ds_is_snapshot
);
1024 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1025 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1029 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1031 ASSERT3U(dlused
, <=, mrs_used
);
1032 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1033 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1035 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1036 SPA_VERSION_UNIQUE_ACCURATE
)
1037 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1041 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1044 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1045 ASSERTV(uint64_t count
);
1048 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1049 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1052 * The err should not be ENOENT, but a bug in a previous version
1053 * of the code could cause upgrade_clones_cb() to not set
1054 * ds_next_snap_obj when it should, leading to a missing entry.
1055 * If we knew that the pool was created after
1056 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1057 * ENOENT. However, at least we can check that we don't have
1058 * too many entries in the next_clones_obj even after failing to
1063 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1065 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1070 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1072 return (&dsl_dataset_phys(ds
)->ds_bp
);
1076 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1078 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1082 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1086 if (ds
== NULL
) /* this is the meta-objset */
1089 ASSERT(ds
->ds_objset
!= NULL
);
1091 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1092 panic("dirtying snapshot!");
1094 /* Must not dirty a dataset in the same txg where it got snapshotted. */
1095 ASSERT3U(tx
->tx_txg
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
1097 dp
= ds
->ds_dir
->dd_pool
;
1098 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1099 /* up the hold count until we can be written out */
1100 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1105 dsl_dataset_is_dirty(dsl_dataset_t
*ds
)
1107 for (int t
= 0; t
< TXG_SIZE
; t
++) {
1108 if (txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1116 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1120 if (!dmu_tx_is_syncing(tx
))
1124 * If there's an fs-only reservation, any blocks that might become
1125 * owned by the snapshot dataset must be accommodated by space
1126 * outside of the reservation.
1128 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1129 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1130 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1131 return (SET_ERROR(ENOSPC
));
1134 * Propagate any reserved space for this snapshot to other
1135 * snapshot checks in this sync group.
1138 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1144 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1145 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1150 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1152 if (!dmu_tx_is_syncing(tx
))
1156 * We don't allow multiple snapshots of the same txg. If there
1157 * is already one, try again.
1159 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1160 return (SET_ERROR(EAGAIN
));
1163 * Check for conflicting snapshot name.
1165 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1167 return (SET_ERROR(EEXIST
));
1168 if (error
!= ENOENT
)
1172 * We don't allow taking snapshots of inconsistent datasets, such as
1173 * those into which we are currently receiving. However, if we are
1174 * creating this snapshot as part of a receive, this check will be
1175 * executed atomically with respect to the completion of the receive
1176 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1177 * case we ignore this, knowing it will be fixed up for us shortly in
1178 * dmu_recv_end_sync().
1180 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1181 return (SET_ERROR(EBUSY
));
1184 * Skip the check for temporary snapshots or if we have already checked
1185 * the counts in dsl_dataset_snapshot_check. This means we really only
1186 * check the count here when we're receiving a stream.
1188 if (cnt
!= 0 && cr
!= NULL
) {
1189 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1190 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1195 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1203 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1205 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1206 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1211 * Pre-compute how many total new snapshots will be created for each
1212 * level in the tree and below. This is needed for validating the
1213 * snapshot limit when either taking a recursive snapshot or when
1214 * taking multiple snapshots.
1216 * The problem is that the counts are not actually adjusted when
1217 * we are checking, only when we finally sync. For a single snapshot,
1218 * this is easy, the count will increase by 1 at each node up the tree,
1219 * but its more complicated for the recursive/multiple snapshot case.
1221 * The dsl_fs_ss_limit_check function does recursively check the count
1222 * at each level up the tree but since it is validating each snapshot
1223 * independently we need to be sure that we are validating the complete
1224 * count for the entire set of snapshots. We do this by rolling up the
1225 * counts for each component of the name into an nvlist and then
1226 * checking each of those cases with the aggregated count.
1228 * This approach properly handles not only the recursive snapshot
1229 * case (where we get all of those on the ddsa_snaps list) but also
1230 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1231 * validate the limit on 'a' using a count of 2).
1233 * We validate the snapshot names in the third loop and only report
1236 if (dmu_tx_is_syncing(tx
)) {
1238 nvlist_t
*cnt_track
= NULL
;
1239 cnt_track
= fnvlist_alloc();
1241 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1243 /* Rollup aggregated counts into the cnt_track list */
1244 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1246 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1250 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1251 pdelim
= strchr(nm
, '@');
1257 if (nvlist_lookup_uint64(cnt_track
, nm
,
1259 /* update existing entry */
1260 fnvlist_add_uint64(cnt_track
, nm
,
1264 fnvlist_add_uint64(cnt_track
, nm
, 1);
1267 pdelim
= strrchr(nm
, '/');
1270 } while (pdelim
!= NULL
);
1273 kmem_free(nm
, MAXPATHLEN
);
1275 /* Check aggregated counts at each level */
1276 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1277 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1283 name
= nvpair_name(pair
);
1284 cnt
= fnvpair_value_uint64(pair
);
1287 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1289 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1290 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1292 dsl_dataset_rele(ds
, FTAG
);
1296 if (ddsa
->ddsa_errors
!= NULL
)
1297 fnvlist_add_int32(ddsa
->ddsa_errors
,
1300 /* only report one error for this check */
1304 nvlist_free(cnt_track
);
1307 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1308 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1311 char *name
, *atp
= NULL
;
1312 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1314 name
= nvpair_name(pair
);
1315 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1316 error
= SET_ERROR(ENAMETOOLONG
);
1318 atp
= strchr(name
, '@');
1320 error
= SET_ERROR(EINVAL
);
1322 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1325 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1327 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1328 error
= dsl_dataset_snapshot_check_impl(ds
,
1329 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1330 dsl_dataset_rele(ds
, FTAG
);
1334 if (ddsa
->ddsa_errors
!= NULL
) {
1335 fnvlist_add_int32(ddsa
->ddsa_errors
,
1346 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1349 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1351 dsl_dataset_phys_t
*dsphys
;
1352 uint64_t dsobj
, crtxg
;
1353 objset_t
*mos
= dp
->dp_meta_objset
;
1354 ASSERTV(static zil_header_t zero_zil
);
1355 ASSERTV(objset_t
*os
);
1357 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1360 * If we are on an old pool, the zil must not be active, in which
1361 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1363 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1364 dmu_objset_from_ds(ds
, &os
) != 0 ||
1365 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1366 sizeof (zero_zil
)) == 0);
1368 /* Should not snapshot a dirty dataset. */
1369 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1372 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1375 * The origin's ds_creation_txg has to be < TXG_INITIAL
1377 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1382 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1383 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1384 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1385 dmu_buf_will_dirty(dbuf
, tx
);
1386 dsphys
= dbuf
->db_data
;
1387 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1388 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1389 dsphys
->ds_fsid_guid
= unique_create();
1390 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1391 sizeof (dsphys
->ds_guid
));
1392 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1393 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1394 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1395 dsphys
->ds_num_children
= 1;
1396 dsphys
->ds_creation_time
= gethrestime_sec();
1397 dsphys
->ds_creation_txg
= crtxg
;
1398 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1399 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1400 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1401 dsphys
->ds_uncompressed_bytes
=
1402 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1403 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1404 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1405 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1406 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1407 dmu_buf_rele(dbuf
, FTAG
);
1409 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1410 if (ds
->ds_feature_inuse
[f
])
1411 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1414 ASSERT3U(ds
->ds_prev
!= 0, ==,
1415 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1417 uint64_t next_clones_obj
=
1418 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1419 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1421 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1422 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1424 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1425 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1426 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1427 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1428 } else if (next_clones_obj
!= 0) {
1429 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1430 dsphys
->ds_next_snap_obj
, tx
);
1431 VERIFY0(zap_add_int(mos
,
1432 next_clones_obj
, dsobj
, tx
));
1437 * If we have a reference-reservation on this dataset, we will
1438 * need to increase the amount of refreservation being charged
1439 * since our unique space is going to zero.
1441 if (ds
->ds_reserved
) {
1443 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1444 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1446 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1450 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1451 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1452 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1453 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1454 dsl_deadlist_close(&ds
->ds_deadlist
);
1455 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1456 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1457 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1458 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1460 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1461 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1462 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1463 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1464 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1465 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1467 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1468 snapname
, 8, 1, &dsobj
, tx
));
1471 dsl_dataset_rele(ds
->ds_prev
, ds
);
1472 VERIFY0(dsl_dataset_hold_obj(dp
,
1473 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1475 dsl_scan_ds_snapshotted(ds
, tx
);
1477 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1479 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1483 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1485 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1486 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1489 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1490 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1493 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1495 name
= nvpair_name(pair
);
1496 atp
= strchr(name
, '@');
1497 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1498 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1500 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1501 if (ddsa
->ddsa_props
!= NULL
) {
1502 dsl_props_set_sync_impl(ds
->ds_prev
,
1503 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1505 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1506 dsl_dataset_rele(ds
, FTAG
);
1511 * The snapshots must all be in the same pool.
1512 * All-or-nothing: if there are any failures, nothing will be modified.
1515 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1517 dsl_dataset_snapshot_arg_t ddsa
;
1519 boolean_t needsuspend
;
1523 nvlist_t
*suspended
= NULL
;
1525 pair
= nvlist_next_nvpair(snaps
, NULL
);
1528 firstname
= nvpair_name(pair
);
1530 error
= spa_open(firstname
, &spa
, FTAG
);
1533 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1534 spa_close(spa
, FTAG
);
1537 suspended
= fnvlist_alloc();
1538 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1539 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1540 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1541 char *snapname
= nvpair_name(pair
);
1545 atp
= strchr(snapname
, '@');
1547 error
= SET_ERROR(EINVAL
);
1550 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1552 error
= zil_suspend(fsname
, &cookie
);
1555 fnvlist_add_uint64(suspended
, fsname
,
1560 ddsa
.ddsa_snaps
= snaps
;
1561 ddsa
.ddsa_props
= props
;
1562 ddsa
.ddsa_errors
= errors
;
1563 ddsa
.ddsa_cr
= CRED();
1566 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1567 dsl_dataset_snapshot_sync
, &ddsa
,
1568 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1571 if (suspended
!= NULL
) {
1572 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1573 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1574 zil_resume((void *)(uintptr_t)
1575 fnvpair_value_uint64(pair
));
1577 fnvlist_free(suspended
);
1583 typedef struct dsl_dataset_snapshot_tmp_arg
{
1584 const char *ddsta_fsname
;
1585 const char *ddsta_snapname
;
1586 minor_t ddsta_cleanup_minor
;
1587 const char *ddsta_htag
;
1588 } dsl_dataset_snapshot_tmp_arg_t
;
1591 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1593 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1594 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1598 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1602 /* NULL cred means no limit check for tmp snapshot */
1603 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1604 tx
, B_FALSE
, 0, NULL
);
1606 dsl_dataset_rele(ds
, FTAG
);
1610 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1611 dsl_dataset_rele(ds
, FTAG
);
1612 return (SET_ERROR(ENOTSUP
));
1614 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1617 dsl_dataset_rele(ds
, FTAG
);
1621 dsl_dataset_rele(ds
, FTAG
);
1626 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1628 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1629 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1630 dsl_dataset_t
*ds
= NULL
;
1632 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1634 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1635 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1636 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1637 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1639 dsl_dataset_rele(ds
, FTAG
);
1643 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1644 minor_t cleanup_minor
, const char *htag
)
1646 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1649 boolean_t needsuspend
;
1652 ddsta
.ddsta_fsname
= fsname
;
1653 ddsta
.ddsta_snapname
= snapname
;
1654 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1655 ddsta
.ddsta_htag
= htag
;
1657 error
= spa_open(fsname
, &spa
, FTAG
);
1660 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1661 spa_close(spa
, FTAG
);
1664 error
= zil_suspend(fsname
, &cookie
);
1669 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1670 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1678 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1680 ASSERT(dmu_tx_is_syncing(tx
));
1681 ASSERT(ds
->ds_objset
!= NULL
);
1682 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1685 * in case we had to change ds_fsid_guid when we opened it,
1688 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1689 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1691 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1692 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1693 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1694 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1695 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1696 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1697 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1698 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1699 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1700 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1701 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1702 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1703 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1706 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1708 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1709 if (ds
->ds_feature_activation_needed
[f
]) {
1710 if (ds
->ds_feature_inuse
[f
])
1712 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1713 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1719 deadlist_enqueue_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
1721 dsl_deadlist_t
*dl
= arg
;
1722 dsl_deadlist_insert(dl
, bp
, tx
);
1727 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1729 objset_t
*os
= ds
->ds_objset
;
1731 bplist_iterate(&ds
->ds_pending_deadlist
,
1732 deadlist_enqueue_cb
, &ds
->ds_deadlist
, tx
);
1734 if (os
->os_synced_dnodes
!= NULL
) {
1735 multilist_destroy(os
->os_synced_dnodes
);
1736 os
->os_synced_dnodes
= NULL
;
1739 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
1741 dmu_buf_rele(ds
->ds_dbuf
, ds
);
1745 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
1748 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1752 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1755 * There may be missing entries in ds_next_clones_obj
1756 * due to a bug in a previous version of the code.
1757 * Only trust it if it has the right number of entries.
1759 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1760 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1763 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
1766 for (zap_cursor_init(&zc
, mos
,
1767 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1768 zap_cursor_retrieve(&zc
, &za
) == 0;
1769 zap_cursor_advance(&zc
)) {
1770 dsl_dataset_t
*clone
;
1771 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1772 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1773 za
.za_first_integer
, FTAG
, &clone
));
1774 dsl_dir_name(clone
->ds_dir
, buf
);
1775 fnvlist_add_boolean(val
, buf
);
1776 dsl_dataset_rele(clone
, FTAG
);
1778 zap_cursor_fini(&zc
);
1783 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1785 nvlist_t
*propval
= fnvlist_alloc();
1789 * We use nvlist_alloc() instead of fnvlist_alloc() because the
1790 * latter would allocate the list with NV_UNIQUE_NAME flag.
1791 * As a result, every time a clone name is appended to the list
1792 * it would be (linearly) searched for for a duplicate name.
1793 * We already know that all clone names must be unique and we
1794 * want avoid the quadratic complexity of double-checking that
1795 * because we can have a large number of clones.
1797 VERIFY0(nvlist_alloc(&val
, 0, KM_SLEEP
));
1799 if (get_clones_stat_impl(ds
, val
) == 0) {
1800 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1801 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
1806 nvlist_free(propval
);
1810 * Returns a string that represents the receive resume stats token. It should
1811 * be freed with strfree().
1814 get_receive_resume_stats_impl(dsl_dataset_t
*ds
)
1816 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1818 if (dsl_dataset_has_resume_receive_state(ds
)) {
1821 uint8_t *compressed
;
1823 nvlist_t
*token_nv
= fnvlist_alloc();
1824 size_t packed_size
, compressed_size
;
1826 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1827 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1828 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1830 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1831 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1832 fnvlist_add_uint64(token_nv
, "object", val
);
1834 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1835 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1836 fnvlist_add_uint64(token_nv
, "offset", val
);
1838 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1839 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1840 fnvlist_add_uint64(token_nv
, "bytes", val
);
1842 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1843 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1844 fnvlist_add_uint64(token_nv
, "toguid", val
);
1846 char buf
[MAXNAMELEN
];
1847 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1848 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1849 fnvlist_add_string(token_nv
, "toname", buf
);
1851 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1852 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1853 fnvlist_add_boolean(token_nv
, "largeblockok");
1855 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1856 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1857 fnvlist_add_boolean(token_nv
, "embedok");
1859 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1860 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1861 fnvlist_add_boolean(token_nv
, "compressok");
1863 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1864 DS_FIELD_RESUME_RAWOK
) == 0) {
1865 fnvlist_add_boolean(token_nv
, "rawok");
1867 packed
= fnvlist_pack(token_nv
, &packed_size
);
1868 fnvlist_free(token_nv
);
1869 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1871 compressed_size
= gzip_compress(packed
, compressed
,
1872 packed_size
, packed_size
, 6);
1875 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1877 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1878 for (int i
= 0; i
< compressed_size
; i
++) {
1879 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1881 str
[compressed_size
* 2] = '\0';
1882 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
1883 ZFS_SEND_RESUME_TOKEN_VERSION
,
1884 (longlong_t
)cksum
.zc_word
[0],
1885 (longlong_t
)packed_size
, str
);
1886 kmem_free(packed
, packed_size
);
1887 kmem_free(str
, compressed_size
* 2 + 1);
1888 kmem_free(compressed
, packed_size
);
1891 return (strdup(""));
1895 * Returns a string that represents the receive resume stats token of the
1896 * dataset's child. It should be freed with strfree().
1899 get_child_receive_stats(dsl_dataset_t
*ds
)
1901 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
1902 dsl_dataset_t
*recv_ds
;
1903 dsl_dataset_name(ds
, recvname
);
1904 if (strlcat(recvname
, "/", sizeof (recvname
)) <
1905 sizeof (recvname
) &&
1906 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
1907 sizeof (recvname
) &&
1908 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, recvname
, FTAG
,
1910 char *propval
= get_receive_resume_stats_impl(recv_ds
);
1911 dsl_dataset_rele(recv_ds
, FTAG
);
1914 return (strdup(""));
1918 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1920 char *propval
= get_receive_resume_stats_impl(ds
);
1921 if (strcmp(propval
, "") != 0) {
1922 dsl_prop_nvlist_add_string(nv
,
1923 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
1925 char *childval
= get_child_receive_stats(ds
);
1926 if (strcmp(childval
, "") != 0) {
1927 dsl_prop_nvlist_add_string(nv
,
1928 ZFS_PROP_RECEIVE_RESUME_TOKEN
, childval
);
1936 dsl_get_refratio(dsl_dataset_t
*ds
)
1938 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1939 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1940 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1945 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
1947 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1951 dsl_get_compressratio(dsl_dataset_t
*ds
)
1953 if (ds
->ds_is_snapshot
) {
1954 return (dsl_get_refratio(ds
));
1956 dsl_dir_t
*dd
= ds
->ds_dir
;
1957 mutex_enter(&dd
->dd_lock
);
1958 uint64_t val
= dsl_dir_get_compressratio(dd
);
1959 mutex_exit(&dd
->dd_lock
);
1965 dsl_get_used(dsl_dataset_t
*ds
)
1967 if (ds
->ds_is_snapshot
) {
1968 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
1970 dsl_dir_t
*dd
= ds
->ds_dir
;
1971 mutex_enter(&dd
->dd_lock
);
1972 uint64_t val
= dsl_dir_get_used(dd
);
1973 mutex_exit(&dd
->dd_lock
);
1979 dsl_get_creation(dsl_dataset_t
*ds
)
1981 return (dsl_dataset_phys(ds
)->ds_creation_time
);
1985 dsl_get_creationtxg(dsl_dataset_t
*ds
)
1987 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
1991 dsl_get_refquota(dsl_dataset_t
*ds
)
1993 return (ds
->ds_quota
);
1997 dsl_get_refreservation(dsl_dataset_t
*ds
)
1999 return (ds
->ds_reserved
);
2003 dsl_get_guid(dsl_dataset_t
*ds
)
2005 return (dsl_dataset_phys(ds
)->ds_guid
);
2009 dsl_get_unique(dsl_dataset_t
*ds
)
2011 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2015 dsl_get_objsetid(dsl_dataset_t
*ds
)
2017 return (ds
->ds_object
);
2021 dsl_get_userrefs(dsl_dataset_t
*ds
)
2023 return (ds
->ds_userrefs
);
2027 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2029 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2033 dsl_get_referenced(dsl_dataset_t
*ds
)
2035 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2039 dsl_get_numclones(dsl_dataset_t
*ds
)
2041 ASSERT(ds
->ds_is_snapshot
);
2042 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2046 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2048 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2053 dsl_get_available(dsl_dataset_t
*ds
)
2055 uint64_t refdbytes
= dsl_get_referenced(ds
);
2056 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2058 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2060 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2062 if (ds
->ds_quota
!= 0) {
2064 * Adjust available bytes according to refquota
2066 if (refdbytes
< ds
->ds_quota
) {
2067 availbytes
= MIN(availbytes
,
2068 ds
->ds_quota
- refdbytes
);
2073 return (availbytes
);
2077 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2079 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2080 dsl_dataset_t
*prev
;
2081 int err
= dsl_dataset_hold_obj(dp
,
2082 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2084 uint64_t comp
, uncomp
;
2085 err
= dsl_dataset_space_written(prev
, ds
, written
,
2087 dsl_dataset_rele(prev
, FTAG
);
2093 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2096 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2098 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2099 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2100 dsl_dataset_name(ds
->ds_prev
, snap
);
2108 * Returns the mountpoint property and source for the given dataset in the value
2109 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2110 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2111 * Returns 0 on success and an error on failure.
2114 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2118 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2120 /* Retrieve the mountpoint value stored in the zap opbject */
2121 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2122 ZAP_MAXVALUELEN
, value
, source
);
2128 * Process the dsname and source to find the full mountpoint string.
2129 * Can be skipped for 'legacy' or 'none'.
2131 if (value
[0] == '/') {
2132 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2134 const char *relpath
;
2137 * If we inherit the mountpoint, even from a dataset
2138 * with a received value, the source will be the path of
2139 * the dataset we inherit from. If source is
2140 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2143 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2146 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2147 relpath
= dsname
+ strlen(source
);
2148 if (relpath
[0] == '/')
2152 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2155 * Special case an alternate root of '/'. This will
2156 * avoid having multiple leading slashes in the
2159 if (strcmp(root
, "/") == 0)
2163 * If the mountpoint is '/' then skip over this
2164 * if we are obtaining either an alternate root or
2165 * an inherited mountpoint.
2168 if (value
[1] == '\0' && (root
[0] != '\0' ||
2169 relpath
[0] != '\0'))
2172 if (relpath
[0] == '\0') {
2173 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2176 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2177 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2180 kmem_free(buf
, ZAP_MAXVALUELEN
);
2187 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2189 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2191 ASSERT(dsl_pool_config_held(dp
));
2193 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2194 dsl_get_refratio(ds
));
2195 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2196 dsl_get_logicalreferenced(ds
));
2197 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2198 dsl_get_compressratio(ds
));
2199 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2202 if (ds
->ds_is_snapshot
) {
2203 get_clones_stat(ds
, nv
);
2205 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2206 if (dsl_get_prev_snap(ds
, buf
) == 0)
2207 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2209 dsl_dir_stats(ds
->ds_dir
, nv
);
2212 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2213 dsl_get_available(ds
));
2214 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2215 dsl_get_referenced(ds
));
2216 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2217 dsl_get_creation(ds
));
2218 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2219 dsl_get_creationtxg(ds
));
2220 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2221 dsl_get_refquota(ds
));
2222 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2223 dsl_get_refreservation(ds
));
2224 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2226 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2227 dsl_get_unique(ds
));
2228 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2229 dsl_get_objsetid(ds
));
2230 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2231 dsl_get_userrefs(ds
));
2232 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2233 dsl_get_defer_destroy(ds
));
2234 dsl_dataset_crypt_stats(ds
, nv
);
2236 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2238 if (dsl_get_written(ds
, &written
) == 0) {
2239 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2244 if (!dsl_dataset_is_snapshot(ds
)) {
2246 * A failed "newfs" (e.g. full) resumable receive leaves
2247 * the stats set on this dataset. Check here for the prop.
2249 get_receive_resume_stats(ds
, nv
);
2252 * A failed incremental resumable receive leaves the
2253 * stats set on our child named "%recv". Check the child
2256 /* 6 extra bytes for /%recv */
2257 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2258 dsl_dataset_t
*recv_ds
;
2259 dsl_dataset_name(ds
, recvname
);
2260 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2261 sizeof (recvname
) &&
2262 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2263 sizeof (recvname
) &&
2264 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
2265 get_receive_resume_stats(recv_ds
, nv
);
2266 dsl_dataset_rele(recv_ds
, FTAG
);
2272 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2274 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2275 ASSERT(dsl_pool_config_held(dp
));
2277 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2278 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2279 stat
->dds_guid
= dsl_get_guid(ds
);
2280 stat
->dds_origin
[0] = '\0';
2281 if (ds
->ds_is_snapshot
) {
2282 stat
->dds_is_snapshot
= B_TRUE
;
2283 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2285 stat
->dds_is_snapshot
= B_FALSE
;
2286 stat
->dds_num_clones
= 0;
2288 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2289 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2295 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2297 return (ds
->ds_fsid_guid
);
2301 dsl_dataset_space(dsl_dataset_t
*ds
,
2302 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2303 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2305 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2306 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2307 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2309 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2310 if (ds
->ds_quota
!= 0) {
2312 * Adjust available bytes according to refquota
2314 if (*refdbytesp
< ds
->ds_quota
)
2315 *availbytesp
= MIN(*availbytesp
,
2316 ds
->ds_quota
- *refdbytesp
);
2320 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2321 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2322 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2323 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2327 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2329 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2332 ASSERT(dsl_pool_config_held(dp
));
2335 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2336 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2337 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2338 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2339 objset_t
*os
, *os_snap
;
2341 * It may be that only the ZIL differs, because it was
2342 * reset in the head. Don't count that as being
2345 if (dmu_objset_from_ds(ds
, &os
) != 0)
2347 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2349 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2350 &os_snap
->os_phys
->os_meta_dnode
,
2351 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2356 typedef struct dsl_dataset_rename_snapshot_arg
{
2357 const char *ddrsa_fsname
;
2358 const char *ddrsa_oldsnapname
;
2359 const char *ddrsa_newsnapname
;
2360 boolean_t ddrsa_recursive
;
2362 } dsl_dataset_rename_snapshot_arg_t
;
2366 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2367 dsl_dataset_t
*hds
, void *arg
)
2369 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2373 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2375 /* ignore nonexistent snapshots */
2376 return (error
== ENOENT
? 0 : error
);
2379 /* new name should not exist */
2380 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2382 error
= SET_ERROR(EEXIST
);
2383 else if (error
== ENOENT
)
2386 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2387 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2388 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2389 error
= SET_ERROR(ENAMETOOLONG
);
2395 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2397 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2398 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2402 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2406 if (ddrsa
->ddrsa_recursive
) {
2407 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2408 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2411 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2413 dsl_dataset_rele(hds
, FTAG
);
2418 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2419 dsl_dataset_t
*hds
, void *arg
)
2421 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2424 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2427 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2428 ASSERT(error
== 0 || error
== ENOENT
);
2429 if (error
== ENOENT
) {
2430 /* ignore nonexistent snapshots */
2434 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2436 /* log before we change the name */
2437 spa_history_log_internal_ds(ds
, "rename", tx
,
2438 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2440 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2442 mutex_enter(&ds
->ds_lock
);
2443 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2444 sizeof (ds
->ds_snapname
));
2445 mutex_exit(&ds
->ds_lock
);
2446 VERIFY0(zap_add(dp
->dp_meta_objset
,
2447 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2448 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2449 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2450 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2452 dsl_dataset_rele(ds
, FTAG
);
2457 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2459 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2460 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2461 dsl_dataset_t
*hds
= NULL
;
2463 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2464 ddrsa
->ddrsa_tx
= tx
;
2465 if (ddrsa
->ddrsa_recursive
) {
2466 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2467 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2470 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2472 dsl_dataset_rele(hds
, FTAG
);
2476 dsl_dataset_rename_snapshot(const char *fsname
,
2477 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2479 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2481 ddrsa
.ddrsa_fsname
= fsname
;
2482 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2483 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2484 ddrsa
.ddrsa_recursive
= recursive
;
2486 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2487 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2488 1, ZFS_SPACE_CHECK_RESERVED
));
2492 * If we're doing an ownership handoff, we need to make sure that there is
2493 * only one long hold on the dataset. We're not allowed to change anything here
2494 * so we don't permanently release the long hold or regular hold here. We want
2495 * to do this only when syncing to avoid the dataset unexpectedly going away
2496 * when we release the long hold.
2499 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2503 if (!dmu_tx_is_syncing(tx
))
2506 if (owner
!= NULL
) {
2507 VERIFY3P(ds
->ds_owner
, ==, owner
);
2508 dsl_dataset_long_rele(ds
, owner
);
2511 held
= dsl_dataset_long_held(ds
);
2514 dsl_dataset_long_hold(ds
, owner
);
2517 return (SET_ERROR(EBUSY
));
2523 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2525 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2526 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2528 int64_t unused_refres_delta
;
2531 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2535 /* must not be a snapshot */
2536 if (ds
->ds_is_snapshot
) {
2537 dsl_dataset_rele(ds
, FTAG
);
2538 return (SET_ERROR(EINVAL
));
2541 /* must have a most recent snapshot */
2542 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2543 dsl_dataset_rele(ds
, FTAG
);
2544 return (SET_ERROR(ESRCH
));
2548 * No rollback to a snapshot created in the current txg, because
2549 * the rollback may dirty the dataset and create blocks that are
2550 * not reachable from the rootbp while having a birth txg that
2551 * falls into the snapshot's range.
2553 if (dmu_tx_is_syncing(tx
) &&
2554 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
2555 dsl_dataset_rele(ds
, FTAG
);
2556 return (SET_ERROR(EAGAIN
));
2560 * If the expected target snapshot is specified, then check that
2561 * the latest snapshot is it.
2563 if (ddra
->ddra_tosnap
!= NULL
) {
2564 dsl_dataset_t
*snapds
;
2566 /* Check if the target snapshot exists at all. */
2567 error
= dsl_dataset_hold(dp
, ddra
->ddra_tosnap
, FTAG
, &snapds
);
2570 * ESRCH is used to signal that the target snapshot does
2571 * not exist, while ENOENT is used to report that
2572 * the rolled back dataset does not exist.
2573 * ESRCH is also used to cover other cases where the
2574 * target snapshot is not related to the dataset being
2575 * rolled back such as being in a different pool.
2577 if (error
== ENOENT
|| error
== EXDEV
)
2578 error
= SET_ERROR(ESRCH
);
2579 dsl_dataset_rele(ds
, FTAG
);
2582 ASSERT(snapds
->ds_is_snapshot
);
2584 /* Check if the snapshot is the latest snapshot indeed. */
2585 if (snapds
!= ds
->ds_prev
) {
2587 * Distinguish between the case where the only problem
2588 * is intervening snapshots (EEXIST) vs the snapshot
2589 * not being a valid target for rollback (ESRCH).
2591 if (snapds
->ds_dir
== ds
->ds_dir
||
2592 (dsl_dir_is_clone(ds
->ds_dir
) &&
2593 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
==
2594 snapds
->ds_object
)) {
2595 error
= SET_ERROR(EEXIST
);
2597 error
= SET_ERROR(ESRCH
);
2599 dsl_dataset_rele(snapds
, FTAG
);
2600 dsl_dataset_rele(ds
, FTAG
);
2603 dsl_dataset_rele(snapds
, FTAG
);
2606 /* must not have any bookmarks after the most recent snapshot */
2607 nvlist_t
*proprequest
= fnvlist_alloc();
2608 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2609 nvlist_t
*bookmarks
= fnvlist_alloc();
2610 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2611 fnvlist_free(proprequest
);
2613 dsl_dataset_rele(ds
, FTAG
);
2616 for (nvpair_t
*pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2617 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2619 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2620 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2621 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2622 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2623 fnvlist_free(bookmarks
);
2624 dsl_dataset_rele(ds
, FTAG
);
2625 return (SET_ERROR(EEXIST
));
2628 fnvlist_free(bookmarks
);
2630 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2632 dsl_dataset_rele(ds
, FTAG
);
2637 * Check if the snap we are rolling back to uses more than
2640 if (ds
->ds_quota
!= 0 &&
2641 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2642 dsl_dataset_rele(ds
, FTAG
);
2643 return (SET_ERROR(EDQUOT
));
2647 * When we do the clone swap, we will temporarily use more space
2648 * due to the refreservation (the head will no longer have any
2649 * unique space, so the entire amount of the refreservation will need
2650 * to be free). We will immediately destroy the clone, freeing
2651 * this space, but the freeing happens over many txg's.
2653 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2654 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2656 if (unused_refres_delta
> 0 &&
2657 unused_refres_delta
>
2658 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2659 dsl_dataset_rele(ds
, FTAG
);
2660 return (SET_ERROR(ENOSPC
));
2663 dsl_dataset_rele(ds
, FTAG
);
2668 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2670 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2671 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2672 dsl_dataset_t
*ds
, *clone
;
2674 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2676 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2678 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2679 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2681 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2682 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
2684 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2686 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2687 dsl_dataset_zero_zil(ds
, tx
);
2689 dsl_destroy_head_sync_impl(clone
, tx
);
2691 dsl_dataset_rele(clone
, FTAG
);
2692 dsl_dataset_rele(ds
, FTAG
);
2696 * Rolls back the given filesystem or volume to the most recent snapshot.
2697 * The name of the most recent snapshot will be returned under key "target"
2698 * in the result nvlist.
2701 * - The existing dataset MUST be owned by the specified owner at entry
2702 * - Upon return, dataset will still be held by the same owner, whether we
2705 * This mode is required any time the existing filesystem is mounted. See
2706 * notes above zfs_suspend_fs() for further details.
2709 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
2712 dsl_dataset_rollback_arg_t ddra
;
2714 ddra
.ddra_fsname
= fsname
;
2715 ddra
.ddra_tosnap
= tosnap
;
2716 ddra
.ddra_owner
= owner
;
2717 ddra
.ddra_result
= result
;
2719 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2720 dsl_dataset_rollback_sync
, &ddra
,
2721 1, ZFS_SPACE_CHECK_RESERVED
));
2724 struct promotenode
{
2729 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2730 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2732 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2735 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2737 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2738 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2740 struct promotenode
*snap
;
2741 dsl_dataset_t
*origin_ds
;
2745 size_t max_snap_len
;
2746 boolean_t conflicting_snaps
;
2748 err
= promote_hold(ddpa
, dp
, FTAG
);
2752 hds
= ddpa
->ddpa_clone
;
2753 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2755 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2756 promote_rele(ddpa
, FTAG
);
2757 return (SET_ERROR(EXDEV
));
2760 snap
= list_head(&ddpa
->shared_snaps
);
2762 err
= SET_ERROR(ENOENT
);
2765 origin_ds
= snap
->ds
;
2768 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
2769 * When doing a promote we must make sure the encryption root for
2770 * both the target and the target's origin does not change to avoid
2771 * needing to rewrap encryption keys
2773 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
2778 * Compute and check the amount of space to transfer. Since this is
2779 * so expensive, don't do the preliminary check.
2781 if (!dmu_tx_is_syncing(tx
)) {
2782 promote_rele(ddpa
, FTAG
);
2786 /* compute origin's new unique space */
2787 snap
= list_tail(&ddpa
->clone_snaps
);
2788 ASSERT(snap
!= NULL
);
2789 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2790 origin_ds
->ds_object
);
2791 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2792 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2793 &ddpa
->unique
, &unused
, &unused
);
2796 * Walk the snapshots that we are moving
2798 * Compute space to transfer. Consider the incremental changes
2799 * to used by each snapshot:
2800 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2801 * So each snapshot gave birth to:
2802 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2803 * So a sequence would look like:
2804 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2805 * Which simplifies to:
2806 * uN + kN + kN-1 + ... + k1 + k0
2807 * Note however, if we stop before we reach the ORIGIN we get:
2808 * uN + kN + kN-1 + ... + kM - uM-1
2810 conflicting_snaps
= B_FALSE
;
2812 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2813 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2814 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2815 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2816 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2817 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2818 dsl_dataset_t
*ds
= snap
->ds
;
2823 * If there are long holds, we won't be able to evict
2826 if (dsl_dataset_long_held(ds
)) {
2827 err
= SET_ERROR(EBUSY
);
2831 /* Check that the snapshot name does not conflict */
2832 VERIFY0(dsl_dataset_get_snapname(ds
));
2833 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2834 err
= SET_ERROR(ENAMETOOLONG
);
2837 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2839 fnvlist_add_boolean(ddpa
->err_ds
,
2840 snap
->ds
->ds_snapname
);
2841 conflicting_snaps
= B_TRUE
;
2842 } else if (err
!= ENOENT
) {
2846 /* The very first snapshot does not have a deadlist */
2847 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2850 dsl_deadlist_space(&ds
->ds_deadlist
,
2851 &dlused
, &dlcomp
, &dluncomp
);
2852 ddpa
->used
+= dlused
;
2853 ddpa
->comp
+= dlcomp
;
2854 ddpa
->uncomp
+= dluncomp
;
2858 * In order to return the full list of conflicting snapshots, we check
2859 * whether there was a conflict after traversing all of them.
2861 if (conflicting_snaps
) {
2862 err
= SET_ERROR(EEXIST
);
2867 * If we are a clone of a clone then we never reached ORIGIN,
2868 * so we need to subtract out the clone origin's used space.
2870 if (ddpa
->origin_origin
) {
2872 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2874 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2876 dsl_dataset_phys(ddpa
->origin_origin
)->
2877 ds_uncompressed_bytes
;
2880 /* Check that there is enough space and limit headroom here */
2881 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2882 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2887 * Compute the amounts of space that will be used by snapshots
2888 * after the promotion (for both origin and clone). For each,
2889 * it is the amount of space that will be on all of their
2890 * deadlists (that was not born before their new origin).
2892 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2896 * Note, typically this will not be a clone of a clone,
2897 * so dd_origin_txg will be < TXG_INITIAL, so
2898 * these snaplist_space() -> dsl_deadlist_space_range()
2899 * calls will be fast because they do not have to
2900 * iterate over all bps.
2902 snap
= list_head(&ddpa
->origin_snaps
);
2904 err
= SET_ERROR(ENOENT
);
2907 err
= snaplist_space(&ddpa
->shared_snaps
,
2908 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2912 err
= snaplist_space(&ddpa
->clone_snaps
,
2913 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2916 ddpa
->cloneusedsnap
+= space
;
2918 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2919 DD_FLAG_USED_BREAKDOWN
) {
2920 err
= snaplist_space(&ddpa
->origin_snaps
,
2921 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2922 &ddpa
->originusedsnap
);
2928 promote_rele(ddpa
, FTAG
);
2933 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2935 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2936 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2938 struct promotenode
*snap
;
2939 dsl_dataset_t
*origin_ds
;
2940 dsl_dataset_t
*origin_head
;
2942 dsl_dir_t
*odd
= NULL
;
2943 uint64_t oldnext_obj
;
2946 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2947 hds
= ddpa
->ddpa_clone
;
2949 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2951 snap
= list_head(&ddpa
->shared_snaps
);
2952 origin_ds
= snap
->ds
;
2955 snap
= list_head(&ddpa
->origin_snaps
);
2956 origin_head
= snap
->ds
;
2959 * We need to explicitly open odd, since origin_ds's dd will be
2962 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2965 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
2967 /* change origin's next snap */
2968 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2969 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2970 snap
= list_tail(&ddpa
->clone_snaps
);
2971 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2972 origin_ds
->ds_object
);
2973 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2975 /* change the origin's next clone */
2976 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2977 dsl_dataset_remove_from_next_clones(origin_ds
,
2978 snap
->ds
->ds_object
, tx
);
2979 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2980 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2985 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2986 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2987 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2988 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2989 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2990 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2991 origin_head
->ds_dir
->dd_origin_txg
=
2992 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2994 /* change dd_clone entries */
2995 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2996 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2997 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2998 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2999 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3000 hds
->ds_object
, tx
));
3002 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3003 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3004 origin_head
->ds_object
, tx
));
3005 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
3006 dsl_dir_phys(dd
)->dd_clones
=
3007 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
3008 DMU_OT_NONE
, 0, tx
);
3010 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3011 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
3014 /* move snapshots to this dir */
3015 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3016 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3017 dsl_dataset_t
*ds
= snap
->ds
;
3020 * Property callbacks are registered to a particular
3021 * dsl_dir. Since ours is changing, evict the objset
3022 * so that they will be unregistered from the old dsl_dir.
3024 if (ds
->ds_objset
) {
3025 dmu_objset_evict(ds
->ds_objset
);
3026 ds
->ds_objset
= NULL
;
3029 /* move snap name entry */
3030 VERIFY0(dsl_dataset_get_snapname(ds
));
3031 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3032 ds
->ds_snapname
, tx
, B_TRUE
));
3033 VERIFY0(zap_add(dp
->dp_meta_objset
,
3034 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3035 8, 1, &ds
->ds_object
, tx
));
3036 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3037 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3039 /* change containing dsl_dir */
3040 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3041 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3042 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3043 ASSERT3P(ds
->ds_dir
, ==, odd
);
3044 dsl_dir_rele(ds
->ds_dir
, ds
);
3045 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3046 NULL
, ds
, &ds
->ds_dir
));
3048 /* move any clone references */
3049 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3050 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3054 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3055 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3056 zap_cursor_retrieve(&zc
, &za
) == 0;
3057 zap_cursor_advance(&zc
)) {
3058 dsl_dataset_t
*cnds
;
3061 if (za
.za_first_integer
== oldnext_obj
) {
3063 * We've already moved the
3064 * origin's reference.
3069 VERIFY0(dsl_dataset_hold_obj(dp
,
3070 za
.za_first_integer
, FTAG
, &cnds
));
3071 o
= dsl_dir_phys(cnds
->ds_dir
)->
3072 dd_head_dataset_obj
;
3074 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3075 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3076 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3077 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3078 dsl_dataset_rele(cnds
, FTAG
);
3080 zap_cursor_fini(&zc
);
3083 ASSERT(!dsl_prop_hascb(ds
));
3087 * Change space accounting.
3088 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3089 * both be valid, or both be 0 (resulting in delta == 0). This
3090 * is true for each of {clone,origin} independently.
3093 delta
= ddpa
->cloneusedsnap
-
3094 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3095 ASSERT3S(delta
, >=, 0);
3096 ASSERT3U(ddpa
->used
, >=, delta
);
3097 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3098 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3099 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3101 delta
= ddpa
->originusedsnap
-
3102 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3103 ASSERT3S(delta
, <=, 0);
3104 ASSERT3U(ddpa
->used
, >=, -delta
);
3105 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3106 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3107 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3109 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3111 /* log history record */
3112 spa_history_log_internal_ds(hds
, "promote", tx
, "");
3114 dsl_dir_rele(odd
, FTAG
);
3115 promote_rele(ddpa
, FTAG
);
3119 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3120 * (exclusive) and last_obj (inclusive). The list will be in reverse
3121 * order (last_obj will be the list_head()). If first_obj == 0, do all
3122 * snapshots back to this dataset's origin.
3125 snaplist_make(dsl_pool_t
*dp
,
3126 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
3128 uint64_t obj
= last_obj
;
3130 list_create(l
, sizeof (struct promotenode
),
3131 offsetof(struct promotenode
, link
));
3133 while (obj
!= first_obj
) {
3135 struct promotenode
*snap
;
3138 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3139 ASSERT(err
!= ENOENT
);
3144 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3146 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3148 list_insert_tail(l
, snap
);
3149 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3156 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3158 struct promotenode
*snap
;
3161 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3162 uint64_t used
, comp
, uncomp
;
3163 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3164 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3171 snaplist_destroy(list_t
*l
, void *tag
)
3173 struct promotenode
*snap
;
3175 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3178 while ((snap
= list_tail(l
)) != NULL
) {
3179 list_remove(l
, snap
);
3180 dsl_dataset_rele(snap
->ds
, tag
);
3181 kmem_free(snap
, sizeof (*snap
));
3187 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
3191 struct promotenode
*snap
;
3193 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3197 dd
= ddpa
->ddpa_clone
->ds_dir
;
3199 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3200 !dsl_dir_is_clone(dd
)) {
3201 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3202 return (SET_ERROR(EINVAL
));
3205 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3206 &ddpa
->shared_snaps
, tag
);
3210 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3211 &ddpa
->clone_snaps
, tag
);
3215 snap
= list_head(&ddpa
->shared_snaps
);
3216 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3217 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3218 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3219 &ddpa
->origin_snaps
, tag
);
3223 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3224 error
= dsl_dataset_hold_obj(dp
,
3225 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3226 tag
, &ddpa
->origin_origin
);
3232 promote_rele(ddpa
, tag
);
3237 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
3239 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3240 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3241 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3242 if (ddpa
->origin_origin
!= NULL
)
3243 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3244 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3250 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3251 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3254 dsl_dataset_promote(const char *name
, char *conflsnap
)
3256 dsl_dataset_promote_arg_t ddpa
= { 0 };
3259 nvpair_t
*snap_pair
;
3263 * We will modify space proportional to the number of
3264 * snapshots. Compute numsnaps.
3266 error
= dmu_objset_hold(name
, FTAG
, &os
);
3269 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3270 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3272 dmu_objset_rele(os
, FTAG
);
3276 ddpa
.ddpa_clonename
= name
;
3277 ddpa
.err_ds
= fnvlist_alloc();
3280 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3281 dsl_dataset_promote_sync
, &ddpa
,
3282 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3285 * Return the first conflicting snapshot found.
3287 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3288 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3289 (void) strcpy(conflsnap
, nvpair_name(snap_pair
));
3291 fnvlist_free(ddpa
.err_ds
);
3296 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3297 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3300 * "slack" factor for received datasets with refquota set on them.
3301 * See the bottom of this function for details on its use.
3303 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3304 spa_asize_inflation
;
3305 int64_t unused_refres_delta
;
3307 /* they should both be heads */
3308 if (clone
->ds_is_snapshot
||
3309 origin_head
->ds_is_snapshot
)
3310 return (SET_ERROR(EINVAL
));
3312 /* if we are not forcing, the branch point should be just before them */
3313 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3314 return (SET_ERROR(EINVAL
));
3316 /* clone should be the clone (unless they are unrelated) */
3317 if (clone
->ds_prev
!= NULL
&&
3318 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3319 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3320 return (SET_ERROR(EINVAL
));
3322 /* the clone should be a child of the origin */
3323 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3324 return (SET_ERROR(EINVAL
));
3326 /* origin_head shouldn't be modified unless 'force' */
3328 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3329 return (SET_ERROR(ETXTBSY
));
3331 /* origin_head should have no long holds (e.g. is not mounted) */
3332 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3333 return (SET_ERROR(EBUSY
));
3335 /* check amount of any unconsumed refreservation */
3336 unused_refres_delta
=
3337 (int64_t)MIN(origin_head
->ds_reserved
,
3338 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3339 (int64_t)MIN(origin_head
->ds_reserved
,
3340 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3342 if (unused_refres_delta
> 0 &&
3343 unused_refres_delta
>
3344 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3345 return (SET_ERROR(ENOSPC
));
3348 * The clone can't be too much over the head's refquota.
3350 * To ensure that the entire refquota can be used, we allow one
3351 * transaction to exceed the the refquota. Therefore, this check
3352 * needs to also allow for the space referenced to be more than the
3353 * refquota. The maximum amount of space that one transaction can use
3354 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3355 * overage ensures that we are able to receive a filesystem that
3356 * exceeds the refquota on the source system.
3358 * So that overage is the refquota_slack we use below.
3360 if (origin_head
->ds_quota
!= 0 &&
3361 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3362 origin_head
->ds_quota
+ refquota_slack
)
3363 return (SET_ERROR(EDQUOT
));
3369 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
3370 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
3372 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3373 int64_t unused_refres_delta
;
3375 ASSERT(clone
->ds_reserved
== 0);
3377 * NOTE: On DEBUG kernels there could be a race between this and
3378 * the check function if spa_asize_inflation is adjusted...
3380 ASSERT(origin_head
->ds_quota
== 0 ||
3381 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
3382 DMU_MAX_ACCESS
* spa_asize_inflation
);
3383 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
3386 * Swap per-dataset feature flags.
3388 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
3389 if (!(spa_feature_table
[f
].fi_flags
&
3390 ZFEATURE_FLAG_PER_DATASET
)) {
3391 ASSERT(!clone
->ds_feature_inuse
[f
]);
3392 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
3396 boolean_t clone_inuse
= clone
->ds_feature_inuse
[f
];
3397 boolean_t origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
3400 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
3401 clone
->ds_feature_inuse
[f
] = B_FALSE
;
3403 if (origin_head_inuse
) {
3404 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
3406 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
3409 dsl_dataset_activate_feature(origin_head
->ds_object
,
3411 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
3413 if (origin_head_inuse
) {
3414 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
3415 clone
->ds_feature_inuse
[f
] = B_TRUE
;
3419 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3420 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3422 if (clone
->ds_objset
!= NULL
) {
3423 dmu_objset_evict(clone
->ds_objset
);
3424 clone
->ds_objset
= NULL
;
3427 if (origin_head
->ds_objset
!= NULL
) {
3428 dmu_objset_evict(origin_head
->ds_objset
);
3429 origin_head
->ds_objset
= NULL
;
3432 unused_refres_delta
=
3433 (int64_t)MIN(origin_head
->ds_reserved
,
3434 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3435 (int64_t)MIN(origin_head
->ds_reserved
,
3436 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3439 * Reset origin's unique bytes, if it exists.
3441 if (clone
->ds_prev
) {
3442 dsl_dataset_t
*origin
= clone
->ds_prev
;
3443 uint64_t comp
, uncomp
;
3445 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3446 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3447 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3448 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3453 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3454 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3456 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3457 dsl_dataset_phys(origin_head
)->ds_bp
=
3458 dsl_dataset_phys(clone
)->ds_bp
;
3459 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3460 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3461 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3464 /* set dd_*_bytes */
3466 int64_t dused
, dcomp
, duncomp
;
3467 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3468 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3470 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3471 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3473 dsl_deadlist_space(&clone
->ds_deadlist
,
3474 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3475 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3476 &odl_used
, &odl_comp
, &odl_uncomp
);
3478 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3480 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3482 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3484 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3486 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3488 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3491 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3492 dused
, dcomp
, duncomp
, tx
);
3493 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3494 -dused
, -dcomp
, -duncomp
, tx
);
3497 * The difference in the space used by snapshots is the
3498 * difference in snapshot space due to the head's
3499 * deadlist (since that's the only thing that's
3500 * changing that affects the snapused).
3502 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3503 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3504 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3505 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3506 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3507 &odl_used
, &odl_comp
, &odl_uncomp
);
3508 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3509 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3512 /* swap ds_*_bytes */
3513 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3514 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3515 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3516 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3517 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3518 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3519 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3520 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3522 /* apply any parent delta for change in unconsumed refreservation */
3523 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3524 unused_refres_delta
, 0, 0, tx
);
3529 dsl_deadlist_close(&clone
->ds_deadlist
);
3530 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3531 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3532 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3533 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3534 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3535 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3536 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3538 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3540 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3541 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3545 * Given a pool name and a dataset object number in that pool,
3546 * return the name of that dataset.
3549 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3555 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3559 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3561 dsl_dataset_name(ds
, buf
);
3562 dsl_dataset_rele(ds
, FTAG
);
3564 dsl_pool_rele(dp
, FTAG
);
3570 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3571 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3575 ASSERT3S(asize
, >, 0);
3578 * *ref_rsrv is the portion of asize that will come from any
3579 * unconsumed refreservation space.
3583 mutex_enter(&ds
->ds_lock
);
3585 * Make a space adjustment for reserved bytes.
3587 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3589 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3591 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3593 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3596 if (!check_quota
|| ds
->ds_quota
== 0) {
3597 mutex_exit(&ds
->ds_lock
);
3601 * If they are requesting more space, and our current estimate
3602 * is over quota, they get to try again unless the actual
3603 * on-disk is over quota and there are no pending changes (which
3604 * may free up space for us).
3606 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3609 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3610 error
= SET_ERROR(ERESTART
);
3612 error
= SET_ERROR(EDQUOT
);
3614 mutex_exit(&ds
->ds_lock
);
3619 typedef struct dsl_dataset_set_qr_arg
{
3620 const char *ddsqra_name
;
3621 zprop_source_t ddsqra_source
;
3622 uint64_t ddsqra_value
;
3623 } dsl_dataset_set_qr_arg_t
;
3628 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3630 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3631 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3636 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3637 return (SET_ERROR(ENOTSUP
));
3639 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3643 if (ds
->ds_is_snapshot
) {
3644 dsl_dataset_rele(ds
, FTAG
);
3645 return (SET_ERROR(EINVAL
));
3648 error
= dsl_prop_predict(ds
->ds_dir
,
3649 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3650 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3652 dsl_dataset_rele(ds
, FTAG
);
3657 dsl_dataset_rele(ds
, FTAG
);
3661 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3662 newval
< ds
->ds_reserved
) {
3663 dsl_dataset_rele(ds
, FTAG
);
3664 return (SET_ERROR(ENOSPC
));
3667 dsl_dataset_rele(ds
, FTAG
);
3672 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3674 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3675 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3676 dsl_dataset_t
*ds
= NULL
;
3679 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3681 dsl_prop_set_sync_impl(ds
,
3682 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3683 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3684 &ddsqra
->ddsqra_value
, tx
);
3686 VERIFY0(dsl_prop_get_int_ds(ds
,
3687 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3689 if (ds
->ds_quota
!= newval
) {
3690 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3691 ds
->ds_quota
= newval
;
3693 dsl_dataset_rele(ds
, FTAG
);
3697 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3700 dsl_dataset_set_qr_arg_t ddsqra
;
3702 ddsqra
.ddsqra_name
= dsname
;
3703 ddsqra
.ddsqra_source
= source
;
3704 ddsqra
.ddsqra_value
= refquota
;
3706 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3707 dsl_dataset_set_refquota_sync
, &ddsqra
, 0, ZFS_SPACE_CHECK_NONE
));
3711 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3713 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3714 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3717 uint64_t newval
, unique
;
3719 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3720 return (SET_ERROR(ENOTSUP
));
3722 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3726 if (ds
->ds_is_snapshot
) {
3727 dsl_dataset_rele(ds
, FTAG
);
3728 return (SET_ERROR(EINVAL
));
3731 error
= dsl_prop_predict(ds
->ds_dir
,
3732 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3733 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3735 dsl_dataset_rele(ds
, FTAG
);
3740 * If we are doing the preliminary check in open context, the
3741 * space estimates may be inaccurate.
3743 if (!dmu_tx_is_syncing(tx
)) {
3744 dsl_dataset_rele(ds
, FTAG
);
3748 mutex_enter(&ds
->ds_lock
);
3749 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3750 dsl_dataset_recalc_head_uniq(ds
);
3751 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3752 mutex_exit(&ds
->ds_lock
);
3754 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3755 uint64_t delta
= MAX(unique
, newval
) -
3756 MAX(unique
, ds
->ds_reserved
);
3759 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3760 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3761 dsl_dataset_rele(ds
, FTAG
);
3762 return (SET_ERROR(ENOSPC
));
3766 dsl_dataset_rele(ds
, FTAG
);
3771 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3772 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3778 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3779 source
, sizeof (value
), 1, &value
, tx
);
3781 VERIFY0(dsl_prop_get_int_ds(ds
,
3782 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3784 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3785 mutex_enter(&ds
->ds_dir
->dd_lock
);
3786 mutex_enter(&ds
->ds_lock
);
3787 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3788 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3789 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3790 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3791 ds
->ds_reserved
= newval
;
3792 mutex_exit(&ds
->ds_lock
);
3794 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3795 mutex_exit(&ds
->ds_dir
->dd_lock
);
3799 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3801 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3802 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3803 dsl_dataset_t
*ds
= NULL
;
3805 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3806 dsl_dataset_set_refreservation_sync_impl(ds
,
3807 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3808 dsl_dataset_rele(ds
, FTAG
);
3812 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3813 uint64_t refreservation
)
3815 dsl_dataset_set_qr_arg_t ddsqra
;
3817 ddsqra
.ddsqra_name
= dsname
;
3818 ddsqra
.ddsqra_source
= source
;
3819 ddsqra
.ddsqra_value
= refreservation
;
3821 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3822 dsl_dataset_set_refreservation_sync
, &ddsqra
,
3823 0, ZFS_SPACE_CHECK_NONE
));
3827 * Return (in *usedp) the amount of space written in new that is not
3828 * present in oldsnap. New may be a snapshot or the head. Old must be
3829 * a snapshot before new, in new's filesystem (or its origin). If not then
3830 * fail and return EINVAL.
3832 * The written space is calculated by considering two components: First, we
3833 * ignore any freed space, and calculate the written as new's used space
3834 * minus old's used space. Next, we add in the amount of space that was freed
3835 * between the two snapshots, thus reducing new's used space relative to old's.
3836 * Specifically, this is the space that was born before old->ds_creation_txg,
3837 * and freed before new (ie. on new's deadlist or a previous deadlist).
3839 * space freed [---------------------]
3840 * snapshots ---O-------O--------O-------O------
3844 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3845 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3849 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3851 ASSERT(dsl_pool_config_held(dp
));
3854 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3855 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3858 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3859 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3862 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3863 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3865 snapobj
= new->ds_object
;
3866 while (snapobj
!= oldsnap
->ds_object
) {
3867 dsl_dataset_t
*snap
;
3868 uint64_t used
, comp
, uncomp
;
3870 if (snapobj
== new->ds_object
) {
3873 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3878 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3879 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3881 * The blocks in the deadlist can not be born after
3882 * ds_prev_snap_txg, so get the whole deadlist space,
3883 * which is more efficient (especially for old-format
3884 * deadlists). Unfortunately the deadlist code
3885 * doesn't have enough information to make this
3886 * optimization itself.
3888 dsl_deadlist_space(&snap
->ds_deadlist
,
3889 &used
, &comp
, &uncomp
);
3891 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3892 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3893 &used
, &comp
, &uncomp
);
3900 * If we get to the beginning of the chain of snapshots
3901 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3902 * was not a snapshot of/before new.
3904 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3906 dsl_dataset_rele(snap
, FTAG
);
3908 err
= SET_ERROR(EINVAL
);
3917 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3918 * lastsnap, and all snapshots in between are deleted.
3920 * blocks that would be freed [---------------------------]
3921 * snapshots ---O-------O--------O-------O--------O
3922 * firstsnap lastsnap
3924 * This is the set of blocks that were born after the snap before firstsnap,
3925 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3926 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3927 * We calculate this by iterating over the relevant deadlists (from the snap
3928 * after lastsnap, backward to the snap after firstsnap), summing up the
3929 * space on the deadlist that was born after the snap before firstsnap.
3932 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3933 dsl_dataset_t
*lastsnap
,
3934 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3938 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3940 ASSERT(firstsnap
->ds_is_snapshot
);
3941 ASSERT(lastsnap
->ds_is_snapshot
);
3944 * Check that the snapshots are in the same dsl_dir, and firstsnap
3945 * is before lastsnap.
3947 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3948 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3949 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3950 return (SET_ERROR(EINVAL
));
3952 *usedp
= *compp
= *uncompp
= 0;
3954 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3955 while (snapobj
!= firstsnap
->ds_object
) {
3957 uint64_t used
, comp
, uncomp
;
3959 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3963 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3964 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3965 &used
, &comp
, &uncomp
);
3970 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3971 ASSERT3U(snapobj
, !=, 0);
3972 dsl_dataset_rele(ds
, FTAG
);
3978 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3979 * For example, they could both be snapshots of the same filesystem, and
3980 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3981 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3982 * filesystem. Or 'earlier' could be the origin's origin.
3984 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3987 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3988 uint64_t earlier_txg
)
3990 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3994 ASSERT(dsl_pool_config_held(dp
));
3995 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3997 if (earlier_txg
== 0)
3998 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
4000 if (later
->ds_is_snapshot
&&
4001 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
4004 if (later
->ds_dir
== earlier
->ds_dir
)
4006 if (!dsl_dir_is_clone(later
->ds_dir
))
4009 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
4011 dsl_dataset_t
*origin
;
4012 error
= dsl_dataset_hold_obj(dp
,
4013 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
4016 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
4017 dsl_dataset_rele(origin
, FTAG
);
4022 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4024 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4025 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4029 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4031 dmu_object_info_t doi
;
4033 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4034 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4038 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4040 return (dsl_dataset_is_zapified(ds
) &&
4041 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4042 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4045 #if defined(_KERNEL) && defined(HAVE_SPL)
4047 module_param(zfs_max_recordsize
, int, 0644);
4048 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4050 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
4051 module_param(zfs_max_recordsize
, int, 0444);
4052 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4055 EXPORT_SYMBOL(dsl_dataset_hold
);
4056 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4057 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4058 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4059 EXPORT_SYMBOL(dsl_dataset_own
);
4060 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4061 EXPORT_SYMBOL(dsl_dataset_name
);
4062 EXPORT_SYMBOL(dsl_dataset_rele
);
4063 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4064 EXPORT_SYMBOL(dsl_dataset_disown
);
4065 EXPORT_SYMBOL(dsl_dataset_tryown
);
4066 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4067 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4068 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4069 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4070 EXPORT_SYMBOL(dsl_dataset_promote
);
4071 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4072 EXPORT_SYMBOL(dsl_dataset_user_release
);
4073 EXPORT_SYMBOL(dsl_dataset_get_holds
);
4074 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
4075 EXPORT_SYMBOL(dsl_dataset_get_spa
);
4076 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
4077 EXPORT_SYMBOL(dsl_dataset_space_written
);
4078 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
4079 EXPORT_SYMBOL(dsl_dataset_sync
);
4080 EXPORT_SYMBOL(dsl_dataset_block_born
);
4081 EXPORT_SYMBOL(dsl_dataset_block_kill
);
4082 EXPORT_SYMBOL(dsl_dataset_dirty
);
4083 EXPORT_SYMBOL(dsl_dataset_stats
);
4084 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
4085 EXPORT_SYMBOL(dsl_dataset_space
);
4086 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
4087 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
4088 EXPORT_SYMBOL(dsl_dataset_check_quota
);
4089 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
4090 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);