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, 2016 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
);
1143 typedef struct dsl_dataset_snapshot_arg
{
1144 nvlist_t
*ddsa_snaps
;
1145 nvlist_t
*ddsa_props
;
1146 nvlist_t
*ddsa_errors
;
1148 } dsl_dataset_snapshot_arg_t
;
1151 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1152 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1157 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1159 if (!dmu_tx_is_syncing(tx
))
1163 * We don't allow multiple snapshots of the same txg. If there
1164 * is already one, try again.
1166 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1167 return (SET_ERROR(EAGAIN
));
1170 * Check for conflicting snapshot name.
1172 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1174 return (SET_ERROR(EEXIST
));
1175 if (error
!= ENOENT
)
1179 * We don't allow taking snapshots of inconsistent datasets, such as
1180 * those into which we are currently receiving. However, if we are
1181 * creating this snapshot as part of a receive, this check will be
1182 * executed atomically with respect to the completion of the receive
1183 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1184 * case we ignore this, knowing it will be fixed up for us shortly in
1185 * dmu_recv_end_sync().
1187 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1188 return (SET_ERROR(EBUSY
));
1191 * Skip the check for temporary snapshots or if we have already checked
1192 * the counts in dsl_dataset_snapshot_check. This means we really only
1193 * check the count here when we're receiving a stream.
1195 if (cnt
!= 0 && cr
!= NULL
) {
1196 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1197 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1202 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1210 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1212 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1213 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1218 * Pre-compute how many total new snapshots will be created for each
1219 * level in the tree and below. This is needed for validating the
1220 * snapshot limit when either taking a recursive snapshot or when
1221 * taking multiple snapshots.
1223 * The problem is that the counts are not actually adjusted when
1224 * we are checking, only when we finally sync. For a single snapshot,
1225 * this is easy, the count will increase by 1 at each node up the tree,
1226 * but its more complicated for the recursive/multiple snapshot case.
1228 * The dsl_fs_ss_limit_check function does recursively check the count
1229 * at each level up the tree but since it is validating each snapshot
1230 * independently we need to be sure that we are validating the complete
1231 * count for the entire set of snapshots. We do this by rolling up the
1232 * counts for each component of the name into an nvlist and then
1233 * checking each of those cases with the aggregated count.
1235 * This approach properly handles not only the recursive snapshot
1236 * case (where we get all of those on the ddsa_snaps list) but also
1237 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1238 * validate the limit on 'a' using a count of 2).
1240 * We validate the snapshot names in the third loop and only report
1243 if (dmu_tx_is_syncing(tx
)) {
1245 nvlist_t
*cnt_track
= NULL
;
1246 cnt_track
= fnvlist_alloc();
1248 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1250 /* Rollup aggregated counts into the cnt_track list */
1251 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1253 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1257 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1258 pdelim
= strchr(nm
, '@');
1264 if (nvlist_lookup_uint64(cnt_track
, nm
,
1266 /* update existing entry */
1267 fnvlist_add_uint64(cnt_track
, nm
,
1271 fnvlist_add_uint64(cnt_track
, nm
, 1);
1274 pdelim
= strrchr(nm
, '/');
1277 } while (pdelim
!= NULL
);
1280 kmem_free(nm
, MAXPATHLEN
);
1282 /* Check aggregated counts at each level */
1283 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1284 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1290 name
= nvpair_name(pair
);
1291 cnt
= fnvpair_value_uint64(pair
);
1294 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1296 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1297 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1299 dsl_dataset_rele(ds
, FTAG
);
1303 if (ddsa
->ddsa_errors
!= NULL
)
1304 fnvlist_add_int32(ddsa
->ddsa_errors
,
1307 /* only report one error for this check */
1311 nvlist_free(cnt_track
);
1314 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1315 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1318 char *name
, *atp
= NULL
;
1319 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1321 name
= nvpair_name(pair
);
1322 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1323 error
= SET_ERROR(ENAMETOOLONG
);
1325 atp
= strchr(name
, '@');
1327 error
= SET_ERROR(EINVAL
);
1329 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1332 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1334 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1335 error
= dsl_dataset_snapshot_check_impl(ds
,
1336 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1337 dsl_dataset_rele(ds
, FTAG
);
1341 if (ddsa
->ddsa_errors
!= NULL
) {
1342 fnvlist_add_int32(ddsa
->ddsa_errors
,
1353 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1356 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1358 dsl_dataset_phys_t
*dsphys
;
1359 uint64_t dsobj
, crtxg
;
1360 objset_t
*mos
= dp
->dp_meta_objset
;
1361 ASSERTV(static zil_header_t zero_zil
);
1362 ASSERTV(objset_t
*os
);
1364 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1367 * If we are on an old pool, the zil must not be active, in which
1368 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1370 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1371 dmu_objset_from_ds(ds
, &os
) != 0 ||
1372 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1373 sizeof (zero_zil
)) == 0);
1375 /* Should not snapshot a dirty dataset. */
1376 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1379 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1382 * The origin's ds_creation_txg has to be < TXG_INITIAL
1384 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1389 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1390 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1391 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1392 dmu_buf_will_dirty(dbuf
, tx
);
1393 dsphys
= dbuf
->db_data
;
1394 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1395 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1396 dsphys
->ds_fsid_guid
= unique_create();
1397 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1398 sizeof (dsphys
->ds_guid
));
1399 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1400 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1401 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1402 dsphys
->ds_num_children
= 1;
1403 dsphys
->ds_creation_time
= gethrestime_sec();
1404 dsphys
->ds_creation_txg
= crtxg
;
1405 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1406 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1407 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1408 dsphys
->ds_uncompressed_bytes
=
1409 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1410 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1411 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1412 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1413 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1414 dmu_buf_rele(dbuf
, FTAG
);
1416 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1417 if (ds
->ds_feature_inuse
[f
])
1418 dsl_dataset_activate_feature(dsobj
, f
, tx
);
1421 ASSERT3U(ds
->ds_prev
!= 0, ==,
1422 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1424 uint64_t next_clones_obj
=
1425 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1426 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1428 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1429 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1431 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1432 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1433 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1434 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1435 } else if (next_clones_obj
!= 0) {
1436 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1437 dsphys
->ds_next_snap_obj
, tx
);
1438 VERIFY0(zap_add_int(mos
,
1439 next_clones_obj
, dsobj
, tx
));
1444 * If we have a reference-reservation on this dataset, we will
1445 * need to increase the amount of refreservation being charged
1446 * since our unique space is going to zero.
1448 if (ds
->ds_reserved
) {
1450 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1451 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1453 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1457 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1458 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1459 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1460 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1461 dsl_deadlist_close(&ds
->ds_deadlist
);
1462 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1463 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1464 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1465 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1467 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1468 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1469 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1470 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1471 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1472 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1474 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1475 snapname
, 8, 1, &dsobj
, tx
));
1478 dsl_dataset_rele(ds
->ds_prev
, ds
);
1479 VERIFY0(dsl_dataset_hold_obj(dp
,
1480 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1482 dsl_scan_ds_snapshotted(ds
, tx
);
1484 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1486 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1490 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1492 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1493 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1496 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1497 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1500 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1502 name
= nvpair_name(pair
);
1503 atp
= strchr(name
, '@');
1504 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1505 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1507 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1508 if (ddsa
->ddsa_props
!= NULL
) {
1509 dsl_props_set_sync_impl(ds
->ds_prev
,
1510 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1512 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1513 dsl_dataset_rele(ds
, FTAG
);
1518 * The snapshots must all be in the same pool.
1519 * All-or-nothing: if there are any failures, nothing will be modified.
1522 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1524 dsl_dataset_snapshot_arg_t ddsa
;
1526 boolean_t needsuspend
;
1530 nvlist_t
*suspended
= NULL
;
1532 pair
= nvlist_next_nvpair(snaps
, NULL
);
1535 firstname
= nvpair_name(pair
);
1537 error
= spa_open(firstname
, &spa
, FTAG
);
1540 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1541 spa_close(spa
, FTAG
);
1544 suspended
= fnvlist_alloc();
1545 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1546 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1547 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1548 char *snapname
= nvpair_name(pair
);
1552 atp
= strchr(snapname
, '@');
1554 error
= SET_ERROR(EINVAL
);
1557 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1559 error
= zil_suspend(fsname
, &cookie
);
1562 fnvlist_add_uint64(suspended
, fsname
,
1567 ddsa
.ddsa_snaps
= snaps
;
1568 ddsa
.ddsa_props
= props
;
1569 ddsa
.ddsa_errors
= errors
;
1570 ddsa
.ddsa_cr
= CRED();
1573 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1574 dsl_dataset_snapshot_sync
, &ddsa
,
1575 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1578 if (suspended
!= NULL
) {
1579 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1580 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1581 zil_resume((void *)(uintptr_t)
1582 fnvpair_value_uint64(pair
));
1584 fnvlist_free(suspended
);
1590 typedef struct dsl_dataset_snapshot_tmp_arg
{
1591 const char *ddsta_fsname
;
1592 const char *ddsta_snapname
;
1593 minor_t ddsta_cleanup_minor
;
1594 const char *ddsta_htag
;
1595 } dsl_dataset_snapshot_tmp_arg_t
;
1598 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1600 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1601 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1605 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1609 /* NULL cred means no limit check for tmp snapshot */
1610 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1611 tx
, B_FALSE
, 0, NULL
);
1613 dsl_dataset_rele(ds
, FTAG
);
1617 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1618 dsl_dataset_rele(ds
, FTAG
);
1619 return (SET_ERROR(ENOTSUP
));
1621 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1624 dsl_dataset_rele(ds
, FTAG
);
1628 dsl_dataset_rele(ds
, FTAG
);
1633 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1635 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1636 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1637 dsl_dataset_t
*ds
= NULL
;
1639 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1641 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1642 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1643 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1644 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1646 dsl_dataset_rele(ds
, FTAG
);
1650 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1651 minor_t cleanup_minor
, const char *htag
)
1653 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1656 boolean_t needsuspend
;
1659 ddsta
.ddsta_fsname
= fsname
;
1660 ddsta
.ddsta_snapname
= snapname
;
1661 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1662 ddsta
.ddsta_htag
= htag
;
1664 error
= spa_open(fsname
, &spa
, FTAG
);
1667 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1668 spa_close(spa
, FTAG
);
1671 error
= zil_suspend(fsname
, &cookie
);
1676 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1677 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1685 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1687 ASSERT(dmu_tx_is_syncing(tx
));
1688 ASSERT(ds
->ds_objset
!= NULL
);
1689 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1692 * in case we had to change ds_fsid_guid when we opened it,
1695 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1696 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1698 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1699 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1700 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1701 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1702 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1703 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1704 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1705 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1706 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1707 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1708 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1709 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1710 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1713 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1715 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1716 if (ds
->ds_feature_activation_needed
[f
]) {
1717 if (ds
->ds_feature_inuse
[f
])
1719 dsl_dataset_activate_feature(ds
->ds_object
, f
, tx
);
1720 ds
->ds_feature_inuse
[f
] = B_TRUE
;
1726 deadlist_enqueue_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
1728 dsl_deadlist_t
*dl
= arg
;
1729 dsl_deadlist_insert(dl
, bp
, tx
);
1734 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1736 objset_t
*os
= ds
->ds_objset
;
1738 bplist_iterate(&ds
->ds_pending_deadlist
,
1739 deadlist_enqueue_cb
, &ds
->ds_deadlist
, tx
);
1741 if (os
->os_synced_dnodes
!= NULL
) {
1742 multilist_destroy(os
->os_synced_dnodes
);
1743 os
->os_synced_dnodes
= NULL
;
1746 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
1748 dmu_buf_rele(ds
->ds_dbuf
, ds
);
1752 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
1755 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1759 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1762 * There may be missing entries in ds_next_clones_obj
1763 * due to a bug in a previous version of the code.
1764 * Only trust it if it has the right number of entries.
1766 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
1767 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1770 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
1773 for (zap_cursor_init(&zc
, mos
,
1774 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
1775 zap_cursor_retrieve(&zc
, &za
) == 0;
1776 zap_cursor_advance(&zc
)) {
1777 dsl_dataset_t
*clone
;
1778 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
1779 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
1780 za
.za_first_integer
, FTAG
, &clone
));
1781 dsl_dir_name(clone
->ds_dir
, buf
);
1782 fnvlist_add_boolean(val
, buf
);
1783 dsl_dataset_rele(clone
, FTAG
);
1785 zap_cursor_fini(&zc
);
1790 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1792 nvlist_t
*propval
= fnvlist_alloc();
1796 * We use nvlist_alloc() instead of fnvlist_alloc() because the
1797 * latter would allocate the list with NV_UNIQUE_NAME flag.
1798 * As a result, every time a clone name is appended to the list
1799 * it would be (linearly) searched for for a duplicate name.
1800 * We already know that all clone names must be unique and we
1801 * want avoid the quadratic complexity of double-checking that
1802 * because we can have a large number of clones.
1804 VERIFY0(nvlist_alloc(&val
, 0, KM_SLEEP
));
1806 if (get_clones_stat_impl(ds
, val
) == 0) {
1807 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
1808 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
1813 nvlist_free(propval
);
1817 * Returns a string that represents the receive resume stats token. It should
1818 * be freed with strfree().
1821 get_receive_resume_stats_impl(dsl_dataset_t
*ds
)
1823 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1825 if (dsl_dataset_has_resume_receive_state(ds
)) {
1828 uint8_t *compressed
;
1830 nvlist_t
*token_nv
= fnvlist_alloc();
1831 size_t packed_size
, compressed_size
;
1833 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1834 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
1835 fnvlist_add_uint64(token_nv
, "fromguid", val
);
1837 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1838 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
1839 fnvlist_add_uint64(token_nv
, "object", val
);
1841 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1842 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
1843 fnvlist_add_uint64(token_nv
, "offset", val
);
1845 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1846 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
1847 fnvlist_add_uint64(token_nv
, "bytes", val
);
1849 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1850 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
1851 fnvlist_add_uint64(token_nv
, "toguid", val
);
1853 char buf
[MAXNAMELEN
];
1854 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
1855 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
1856 fnvlist_add_string(token_nv
, "toname", buf
);
1858 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1859 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
1860 fnvlist_add_boolean(token_nv
, "largeblockok");
1862 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1863 DS_FIELD_RESUME_EMBEDOK
) == 0) {
1864 fnvlist_add_boolean(token_nv
, "embedok");
1866 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1867 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
1868 fnvlist_add_boolean(token_nv
, "compressok");
1870 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
1871 DS_FIELD_RESUME_RAWOK
) == 0) {
1872 fnvlist_add_boolean(token_nv
, "rawok");
1874 packed
= fnvlist_pack(token_nv
, &packed_size
);
1875 fnvlist_free(token_nv
);
1876 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
1878 compressed_size
= gzip_compress(packed
, compressed
,
1879 packed_size
, packed_size
, 6);
1882 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
1884 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
1885 for (int i
= 0; i
< compressed_size
; i
++) {
1886 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
1888 str
[compressed_size
* 2] = '\0';
1889 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
1890 ZFS_SEND_RESUME_TOKEN_VERSION
,
1891 (longlong_t
)cksum
.zc_word
[0],
1892 (longlong_t
)packed_size
, str
);
1893 kmem_free(packed
, packed_size
);
1894 kmem_free(str
, compressed_size
* 2 + 1);
1895 kmem_free(compressed
, packed_size
);
1898 return (strdup(""));
1902 * Returns a string that represents the receive resume stats token of the
1903 * dataset's child. It should be freed with strfree().
1906 get_child_receive_stats(dsl_dataset_t
*ds
)
1908 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
1909 dsl_dataset_t
*recv_ds
;
1910 dsl_dataset_name(ds
, recvname
);
1911 if (strlcat(recvname
, "/", sizeof (recvname
)) <
1912 sizeof (recvname
) &&
1913 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
1914 sizeof (recvname
) &&
1915 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, recvname
, FTAG
,
1917 char *propval
= get_receive_resume_stats_impl(recv_ds
);
1918 dsl_dataset_rele(recv_ds
, FTAG
);
1921 return (strdup(""));
1925 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
1927 char *propval
= get_receive_resume_stats_impl(ds
);
1928 if (strcmp(propval
, "") != 0) {
1929 dsl_prop_nvlist_add_string(nv
,
1930 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
1932 char *childval
= get_child_receive_stats(ds
);
1933 if (strcmp(childval
, "") != 0) {
1934 dsl_prop_nvlist_add_string(nv
,
1935 ZFS_PROP_RECEIVE_RESUME_TOKEN
, childval
);
1943 dsl_get_refratio(dsl_dataset_t
*ds
)
1945 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
1946 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
1947 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
1952 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
1954 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
1958 dsl_get_compressratio(dsl_dataset_t
*ds
)
1960 if (ds
->ds_is_snapshot
) {
1961 return (dsl_get_refratio(ds
));
1963 dsl_dir_t
*dd
= ds
->ds_dir
;
1964 mutex_enter(&dd
->dd_lock
);
1965 uint64_t val
= dsl_dir_get_compressratio(dd
);
1966 mutex_exit(&dd
->dd_lock
);
1972 dsl_get_used(dsl_dataset_t
*ds
)
1974 if (ds
->ds_is_snapshot
) {
1975 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
1977 dsl_dir_t
*dd
= ds
->ds_dir
;
1978 mutex_enter(&dd
->dd_lock
);
1979 uint64_t val
= dsl_dir_get_used(dd
);
1980 mutex_exit(&dd
->dd_lock
);
1986 dsl_get_creation(dsl_dataset_t
*ds
)
1988 return (dsl_dataset_phys(ds
)->ds_creation_time
);
1992 dsl_get_creationtxg(dsl_dataset_t
*ds
)
1994 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
1998 dsl_get_refquota(dsl_dataset_t
*ds
)
2000 return (ds
->ds_quota
);
2004 dsl_get_refreservation(dsl_dataset_t
*ds
)
2006 return (ds
->ds_reserved
);
2010 dsl_get_guid(dsl_dataset_t
*ds
)
2012 return (dsl_dataset_phys(ds
)->ds_guid
);
2016 dsl_get_unique(dsl_dataset_t
*ds
)
2018 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2022 dsl_get_objsetid(dsl_dataset_t
*ds
)
2024 return (ds
->ds_object
);
2028 dsl_get_userrefs(dsl_dataset_t
*ds
)
2030 return (ds
->ds_userrefs
);
2034 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2036 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2040 dsl_get_referenced(dsl_dataset_t
*ds
)
2042 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2046 dsl_get_numclones(dsl_dataset_t
*ds
)
2048 ASSERT(ds
->ds_is_snapshot
);
2049 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2053 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2055 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2060 dsl_get_available(dsl_dataset_t
*ds
)
2062 uint64_t refdbytes
= dsl_get_referenced(ds
);
2063 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2065 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2067 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2069 if (ds
->ds_quota
!= 0) {
2071 * Adjust available bytes according to refquota
2073 if (refdbytes
< ds
->ds_quota
) {
2074 availbytes
= MIN(availbytes
,
2075 ds
->ds_quota
- refdbytes
);
2080 return (availbytes
);
2084 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2086 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2087 dsl_dataset_t
*prev
;
2088 int err
= dsl_dataset_hold_obj(dp
,
2089 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2091 uint64_t comp
, uncomp
;
2092 err
= dsl_dataset_space_written(prev
, ds
, written
,
2094 dsl_dataset_rele(prev
, FTAG
);
2100 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2103 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2105 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2106 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2107 dsl_dataset_name(ds
->ds_prev
, snap
);
2115 * Returns the mountpoint property and source for the given dataset in the value
2116 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2117 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2118 * Returns 0 on success and an error on failure.
2121 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2125 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2127 /* Retrieve the mountpoint value stored in the zap opbject */
2128 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2129 ZAP_MAXVALUELEN
, value
, source
);
2135 * Process the dsname and source to find the full mountpoint string.
2136 * Can be skipped for 'legacy' or 'none'.
2138 if (value
[0] == '/') {
2139 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2141 const char *relpath
;
2144 * If we inherit the mountpoint, even from a dataset
2145 * with a received value, the source will be the path of
2146 * the dataset we inherit from. If source is
2147 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2150 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2153 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2154 relpath
= dsname
+ strlen(source
);
2155 if (relpath
[0] == '/')
2159 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2162 * Special case an alternate root of '/'. This will
2163 * avoid having multiple leading slashes in the
2166 if (strcmp(root
, "/") == 0)
2170 * If the mountpoint is '/' then skip over this
2171 * if we are obtaining either an alternate root or
2172 * an inherited mountpoint.
2175 if (value
[1] == '\0' && (root
[0] != '\0' ||
2176 relpath
[0] != '\0'))
2179 if (relpath
[0] == '\0') {
2180 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2183 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2184 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2187 kmem_free(buf
, ZAP_MAXVALUELEN
);
2194 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2196 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2198 ASSERT(dsl_pool_config_held(dp
));
2200 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2201 dsl_get_refratio(ds
));
2202 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2203 dsl_get_logicalreferenced(ds
));
2204 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2205 dsl_get_compressratio(ds
));
2206 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2209 if (ds
->ds_is_snapshot
) {
2210 get_clones_stat(ds
, nv
);
2212 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2213 if (dsl_get_prev_snap(ds
, buf
) == 0)
2214 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2216 dsl_dir_stats(ds
->ds_dir
, nv
);
2219 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2220 dsl_get_available(ds
));
2221 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2222 dsl_get_referenced(ds
));
2223 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2224 dsl_get_creation(ds
));
2225 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2226 dsl_get_creationtxg(ds
));
2227 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2228 dsl_get_refquota(ds
));
2229 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2230 dsl_get_refreservation(ds
));
2231 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2233 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2234 dsl_get_unique(ds
));
2235 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2236 dsl_get_objsetid(ds
));
2237 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2238 dsl_get_userrefs(ds
));
2239 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2240 dsl_get_defer_destroy(ds
));
2241 dsl_dataset_crypt_stats(ds
, nv
);
2243 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2245 if (dsl_get_written(ds
, &written
) == 0) {
2246 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2251 if (!dsl_dataset_is_snapshot(ds
)) {
2253 * A failed "newfs" (e.g. full) resumable receive leaves
2254 * the stats set on this dataset. Check here for the prop.
2256 get_receive_resume_stats(ds
, nv
);
2259 * A failed incremental resumable receive leaves the
2260 * stats set on our child named "%recv". Check the child
2263 /* 6 extra bytes for /%recv */
2264 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2265 dsl_dataset_t
*recv_ds
;
2266 dsl_dataset_name(ds
, recvname
);
2267 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2268 sizeof (recvname
) &&
2269 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2270 sizeof (recvname
) &&
2271 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
2272 get_receive_resume_stats(recv_ds
, nv
);
2273 dsl_dataset_rele(recv_ds
, FTAG
);
2279 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2281 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2282 ASSERT(dsl_pool_config_held(dp
));
2284 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2285 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2286 stat
->dds_guid
= dsl_get_guid(ds
);
2287 stat
->dds_origin
[0] = '\0';
2288 if (ds
->ds_is_snapshot
) {
2289 stat
->dds_is_snapshot
= B_TRUE
;
2290 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2292 stat
->dds_is_snapshot
= B_FALSE
;
2293 stat
->dds_num_clones
= 0;
2295 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2296 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2302 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2304 return (ds
->ds_fsid_guid
);
2308 dsl_dataset_space(dsl_dataset_t
*ds
,
2309 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2310 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2312 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2313 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2314 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2316 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2317 if (ds
->ds_quota
!= 0) {
2319 * Adjust available bytes according to refquota
2321 if (*refdbytesp
< ds
->ds_quota
)
2322 *availbytesp
= MIN(*availbytesp
,
2323 ds
->ds_quota
- *refdbytesp
);
2327 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2328 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2329 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2330 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2334 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2336 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2339 ASSERT(dsl_pool_config_held(dp
));
2342 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2343 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2344 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2345 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2346 objset_t
*os
, *os_snap
;
2348 * It may be that only the ZIL differs, because it was
2349 * reset in the head. Don't count that as being
2352 if (dmu_objset_from_ds(ds
, &os
) != 0)
2354 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2356 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2357 &os_snap
->os_phys
->os_meta_dnode
,
2358 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2363 typedef struct dsl_dataset_rename_snapshot_arg
{
2364 const char *ddrsa_fsname
;
2365 const char *ddrsa_oldsnapname
;
2366 const char *ddrsa_newsnapname
;
2367 boolean_t ddrsa_recursive
;
2369 } dsl_dataset_rename_snapshot_arg_t
;
2373 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2374 dsl_dataset_t
*hds
, void *arg
)
2376 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2380 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2382 /* ignore nonexistent snapshots */
2383 return (error
== ENOENT
? 0 : error
);
2386 /* new name should not exist */
2387 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2389 error
= SET_ERROR(EEXIST
);
2390 else if (error
== ENOENT
)
2393 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2394 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2395 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2396 error
= SET_ERROR(ENAMETOOLONG
);
2402 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2404 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2405 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2409 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2413 if (ddrsa
->ddrsa_recursive
) {
2414 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2415 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2418 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2420 dsl_dataset_rele(hds
, FTAG
);
2425 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2426 dsl_dataset_t
*hds
, void *arg
)
2428 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2431 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2434 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2435 ASSERT(error
== 0 || error
== ENOENT
);
2436 if (error
== ENOENT
) {
2437 /* ignore nonexistent snapshots */
2441 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2443 /* log before we change the name */
2444 spa_history_log_internal_ds(ds
, "rename", tx
,
2445 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2447 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2449 mutex_enter(&ds
->ds_lock
);
2450 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2451 sizeof (ds
->ds_snapname
));
2452 mutex_exit(&ds
->ds_lock
);
2453 VERIFY0(zap_add(dp
->dp_meta_objset
,
2454 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2455 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2456 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2457 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2459 dsl_dataset_rele(ds
, FTAG
);
2464 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2466 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2467 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2468 dsl_dataset_t
*hds
= NULL
;
2470 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2471 ddrsa
->ddrsa_tx
= tx
;
2472 if (ddrsa
->ddrsa_recursive
) {
2473 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2474 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2477 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2479 dsl_dataset_rele(hds
, FTAG
);
2483 dsl_dataset_rename_snapshot(const char *fsname
,
2484 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2486 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2488 ddrsa
.ddrsa_fsname
= fsname
;
2489 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2490 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2491 ddrsa
.ddrsa_recursive
= recursive
;
2493 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2494 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2495 1, ZFS_SPACE_CHECK_RESERVED
));
2499 * If we're doing an ownership handoff, we need to make sure that there is
2500 * only one long hold on the dataset. We're not allowed to change anything here
2501 * so we don't permanently release the long hold or regular hold here. We want
2502 * to do this only when syncing to avoid the dataset unexpectedly going away
2503 * when we release the long hold.
2506 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2510 if (!dmu_tx_is_syncing(tx
))
2513 if (owner
!= NULL
) {
2514 VERIFY3P(ds
->ds_owner
, ==, owner
);
2515 dsl_dataset_long_rele(ds
, owner
);
2518 held
= dsl_dataset_long_held(ds
);
2521 dsl_dataset_long_hold(ds
, owner
);
2524 return (SET_ERROR(EBUSY
));
2529 typedef struct dsl_dataset_rollback_arg
{
2530 const char *ddra_fsname
;
2531 const char *ddra_tosnap
;
2533 nvlist_t
*ddra_result
;
2534 } dsl_dataset_rollback_arg_t
;
2537 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2539 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2540 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2542 int64_t unused_refres_delta
;
2545 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2549 /* must not be a snapshot */
2550 if (ds
->ds_is_snapshot
) {
2551 dsl_dataset_rele(ds
, FTAG
);
2552 return (SET_ERROR(EINVAL
));
2555 /* must have a most recent snapshot */
2556 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2557 dsl_dataset_rele(ds
, FTAG
);
2558 return (SET_ERROR(EINVAL
));
2562 * No rollback to a snapshot created in the current txg, because
2563 * the rollback may dirty the dataset and create blocks that are
2564 * not reachable from the rootbp while having a birth txg that
2565 * falls into the snapshot's range.
2567 if (dmu_tx_is_syncing(tx
) &&
2568 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
2569 dsl_dataset_rele(ds
, FTAG
);
2570 return (SET_ERROR(EAGAIN
));
2574 * If the expected target snapshot is specified, then check that
2575 * the latest snapshot is it.
2577 if (ddra
->ddra_tosnap
!= NULL
) {
2578 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2580 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2581 if (strcmp(namebuf
, ddra
->ddra_tosnap
) != 0)
2582 return (SET_ERROR(EXDEV
));
2585 /* must not have any bookmarks after the most recent snapshot */
2586 nvlist_t
*proprequest
= fnvlist_alloc();
2587 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2588 nvlist_t
*bookmarks
= fnvlist_alloc();
2589 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2590 fnvlist_free(proprequest
);
2593 for (nvpair_t
*pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2594 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2596 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2597 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2598 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2599 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2600 fnvlist_free(bookmarks
);
2601 dsl_dataset_rele(ds
, FTAG
);
2602 return (SET_ERROR(EEXIST
));
2605 fnvlist_free(bookmarks
);
2607 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2609 dsl_dataset_rele(ds
, FTAG
);
2614 * Check if the snap we are rolling back to uses more than
2617 if (ds
->ds_quota
!= 0 &&
2618 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2619 dsl_dataset_rele(ds
, FTAG
);
2620 return (SET_ERROR(EDQUOT
));
2624 * When we do the clone swap, we will temporarily use more space
2625 * due to the refreservation (the head will no longer have any
2626 * unique space, so the entire amount of the refreservation will need
2627 * to be free). We will immediately destroy the clone, freeing
2628 * this space, but the freeing happens over many txg's.
2630 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2631 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2633 if (unused_refres_delta
> 0 &&
2634 unused_refres_delta
>
2635 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2636 dsl_dataset_rele(ds
, FTAG
);
2637 return (SET_ERROR(ENOSPC
));
2640 dsl_dataset_rele(ds
, FTAG
);
2645 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2647 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2648 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2649 dsl_dataset_t
*ds
, *clone
;
2651 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2653 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2655 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2656 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2658 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2659 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
2661 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2663 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2664 dsl_dataset_zero_zil(ds
, tx
);
2666 dsl_destroy_head_sync_impl(clone
, tx
);
2668 dsl_dataset_rele(clone
, FTAG
);
2669 dsl_dataset_rele(ds
, FTAG
);
2673 * Rolls back the given filesystem or volume to the most recent snapshot.
2674 * The name of the most recent snapshot will be returned under key "target"
2675 * in the result nvlist.
2678 * - The existing dataset MUST be owned by the specified owner at entry
2679 * - Upon return, dataset will still be held by the same owner, whether we
2682 * This mode is required any time the existing filesystem is mounted. See
2683 * notes above zfs_suspend_fs() for further details.
2686 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
2689 dsl_dataset_rollback_arg_t ddra
;
2691 ddra
.ddra_fsname
= fsname
;
2692 ddra
.ddra_tosnap
= tosnap
;
2693 ddra
.ddra_owner
= owner
;
2694 ddra
.ddra_result
= result
;
2696 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2697 dsl_dataset_rollback_sync
, &ddra
,
2698 1, ZFS_SPACE_CHECK_RESERVED
));
2701 struct promotenode
{
2706 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2707 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2709 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2712 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2714 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2715 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2717 struct promotenode
*snap
;
2718 dsl_dataset_t
*origin_ds
;
2722 size_t max_snap_len
;
2723 boolean_t conflicting_snaps
;
2725 err
= promote_hold(ddpa
, dp
, FTAG
);
2729 hds
= ddpa
->ddpa_clone
;
2730 snap
= list_head(&ddpa
->shared_snaps
);
2731 origin_ds
= snap
->ds
;
2732 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
2734 snap
= list_head(&ddpa
->origin_snaps
);
2736 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
2737 promote_rele(ddpa
, FTAG
);
2738 return (SET_ERROR(EXDEV
));
2741 snap
= list_head(&ddpa
->shared_snaps
);
2743 err
= SET_ERROR(ENOENT
);
2746 origin_ds
= snap
->ds
;
2749 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
2750 * When doing a promote we must make sure the encryption root for
2751 * both the target and the target's origin does not change to avoid
2752 * needing to rewrap encryption keys
2754 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
2759 * Compute and check the amount of space to transfer. Since this is
2760 * so expensive, don't do the preliminary check.
2762 if (!dmu_tx_is_syncing(tx
)) {
2763 promote_rele(ddpa
, FTAG
);
2767 /* compute origin's new unique space */
2768 snap
= list_tail(&ddpa
->clone_snaps
);
2769 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2770 origin_ds
->ds_object
);
2771 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
2772 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
2773 &ddpa
->unique
, &unused
, &unused
);
2776 * Walk the snapshots that we are moving
2778 * Compute space to transfer. Consider the incremental changes
2779 * to used by each snapshot:
2780 * (my used) = (prev's used) + (blocks born) - (blocks killed)
2781 * So each snapshot gave birth to:
2782 * (blocks born) = (my used) - (prev's used) + (blocks killed)
2783 * So a sequence would look like:
2784 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
2785 * Which simplifies to:
2786 * uN + kN + kN-1 + ... + k1 + k0
2787 * Note however, if we stop before we reach the ORIGIN we get:
2788 * uN + kN + kN-1 + ... + kM - uM-1
2790 conflicting_snaps
= B_FALSE
;
2792 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
2793 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
2794 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
2795 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2796 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2797 uint64_t val
, dlused
, dlcomp
, dluncomp
;
2798 dsl_dataset_t
*ds
= snap
->ds
;
2803 * If there are long holds, we won't be able to evict
2806 if (dsl_dataset_long_held(ds
)) {
2807 err
= SET_ERROR(EBUSY
);
2811 /* Check that the snapshot name does not conflict */
2812 VERIFY0(dsl_dataset_get_snapname(ds
));
2813 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
2814 err
= SET_ERROR(ENAMETOOLONG
);
2817 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
2819 fnvlist_add_boolean(ddpa
->err_ds
,
2820 snap
->ds
->ds_snapname
);
2821 conflicting_snaps
= B_TRUE
;
2822 } else if (err
!= ENOENT
) {
2826 /* The very first snapshot does not have a deadlist */
2827 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
2830 dsl_deadlist_space(&ds
->ds_deadlist
,
2831 &dlused
, &dlcomp
, &dluncomp
);
2832 ddpa
->used
+= dlused
;
2833 ddpa
->comp
+= dlcomp
;
2834 ddpa
->uncomp
+= dluncomp
;
2838 * In order to return the full list of conflicting snapshots, we check
2839 * whether there was a conflict after traversing all of them.
2841 if (conflicting_snaps
) {
2842 err
= SET_ERROR(EEXIST
);
2847 * If we are a clone of a clone then we never reached ORIGIN,
2848 * so we need to subtract out the clone origin's used space.
2850 if (ddpa
->origin_origin
) {
2852 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
2854 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
2856 dsl_dataset_phys(ddpa
->origin_origin
)->
2857 ds_uncompressed_bytes
;
2860 /* Check that there is enough space and limit headroom here */
2861 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
2862 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
2867 * Compute the amounts of space that will be used by snapshots
2868 * after the promotion (for both origin and clone). For each,
2869 * it is the amount of space that will be on all of their
2870 * deadlists (that was not born before their new origin).
2872 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
2876 * Note, typically this will not be a clone of a clone,
2877 * so dd_origin_txg will be < TXG_INITIAL, so
2878 * these snaplist_space() -> dsl_deadlist_space_range()
2879 * calls will be fast because they do not have to
2880 * iterate over all bps.
2882 snap
= list_head(&ddpa
->origin_snaps
);
2884 err
= SET_ERROR(ENOENT
);
2887 err
= snaplist_space(&ddpa
->shared_snaps
,
2888 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
2892 err
= snaplist_space(&ddpa
->clone_snaps
,
2893 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
2896 ddpa
->cloneusedsnap
+= space
;
2898 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
2899 DD_FLAG_USED_BREAKDOWN
) {
2900 err
= snaplist_space(&ddpa
->origin_snaps
,
2901 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
2902 &ddpa
->originusedsnap
);
2908 promote_rele(ddpa
, FTAG
);
2913 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
2915 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2916 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2918 struct promotenode
*snap
;
2919 dsl_dataset_t
*origin_ds
;
2920 dsl_dataset_t
*origin_head
;
2922 dsl_dir_t
*odd
= NULL
;
2923 uint64_t oldnext_obj
;
2926 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
2927 hds
= ddpa
->ddpa_clone
;
2929 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
2931 snap
= list_head(&ddpa
->shared_snaps
);
2932 origin_ds
= snap
->ds
;
2935 snap
= list_head(&ddpa
->origin_snaps
);
2936 origin_head
= snap
->ds
;
2939 * We need to explicitly open odd, since origin_ds's dd will be
2942 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
2945 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
2947 /* change origin's next snap */
2948 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
2949 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
2950 snap
= list_tail(&ddpa
->clone_snaps
);
2951 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
2952 origin_ds
->ds_object
);
2953 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
2955 /* change the origin's next clone */
2956 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
2957 dsl_dataset_remove_from_next_clones(origin_ds
,
2958 snap
->ds
->ds_object
, tx
);
2959 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2960 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
2965 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
2966 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
2967 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
2968 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
2969 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
2970 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
2971 origin_head
->ds_dir
->dd_origin_txg
=
2972 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
2974 /* change dd_clone entries */
2975 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
2976 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2977 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
2978 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2979 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2980 hds
->ds_object
, tx
));
2982 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
2983 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
2984 origin_head
->ds_object
, tx
));
2985 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
2986 dsl_dir_phys(dd
)->dd_clones
=
2987 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
2988 DMU_OT_NONE
, 0, tx
);
2990 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
2991 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
2994 /* move snapshots to this dir */
2995 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
2996 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
2997 dsl_dataset_t
*ds
= snap
->ds
;
3000 * Property callbacks are registered to a particular
3001 * dsl_dir. Since ours is changing, evict the objset
3002 * so that they will be unregistered from the old dsl_dir.
3004 if (ds
->ds_objset
) {
3005 dmu_objset_evict(ds
->ds_objset
);
3006 ds
->ds_objset
= NULL
;
3009 /* move snap name entry */
3010 VERIFY0(dsl_dataset_get_snapname(ds
));
3011 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3012 ds
->ds_snapname
, tx
, B_TRUE
));
3013 VERIFY0(zap_add(dp
->dp_meta_objset
,
3014 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3015 8, 1, &ds
->ds_object
, tx
));
3016 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3017 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3019 /* change containing dsl_dir */
3020 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3021 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3022 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3023 ASSERT3P(ds
->ds_dir
, ==, odd
);
3024 dsl_dir_rele(ds
->ds_dir
, ds
);
3025 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3026 NULL
, ds
, &ds
->ds_dir
));
3028 /* move any clone references */
3029 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3030 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3034 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3035 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3036 zap_cursor_retrieve(&zc
, &za
) == 0;
3037 zap_cursor_advance(&zc
)) {
3038 dsl_dataset_t
*cnds
;
3041 if (za
.za_first_integer
== oldnext_obj
) {
3043 * We've already moved the
3044 * origin's reference.
3049 VERIFY0(dsl_dataset_hold_obj(dp
,
3050 za
.za_first_integer
, FTAG
, &cnds
));
3051 o
= dsl_dir_phys(cnds
->ds_dir
)->
3052 dd_head_dataset_obj
;
3054 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3055 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3056 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3057 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3058 dsl_dataset_rele(cnds
, FTAG
);
3060 zap_cursor_fini(&zc
);
3063 ASSERT(!dsl_prop_hascb(ds
));
3067 * Change space accounting.
3068 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3069 * both be valid, or both be 0 (resulting in delta == 0). This
3070 * is true for each of {clone,origin} independently.
3073 delta
= ddpa
->cloneusedsnap
-
3074 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3075 ASSERT3S(delta
, >=, 0);
3076 ASSERT3U(ddpa
->used
, >=, delta
);
3077 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3078 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3079 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3081 delta
= ddpa
->originusedsnap
-
3082 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3083 ASSERT3S(delta
, <=, 0);
3084 ASSERT3U(ddpa
->used
, >=, -delta
);
3085 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3086 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3087 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3089 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3091 /* log history record */
3092 spa_history_log_internal_ds(hds
, "promote", tx
, "");
3094 dsl_dir_rele(odd
, FTAG
);
3095 promote_rele(ddpa
, FTAG
);
3099 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3100 * (exclusive) and last_obj (inclusive). The list will be in reverse
3101 * order (last_obj will be the list_head()). If first_obj == 0, do all
3102 * snapshots back to this dataset's origin.
3105 snaplist_make(dsl_pool_t
*dp
,
3106 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
3108 uint64_t obj
= last_obj
;
3110 list_create(l
, sizeof (struct promotenode
),
3111 offsetof(struct promotenode
, link
));
3113 while (obj
!= first_obj
) {
3115 struct promotenode
*snap
;
3118 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3119 ASSERT(err
!= ENOENT
);
3124 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3126 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3128 list_insert_tail(l
, snap
);
3129 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3136 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3138 struct promotenode
*snap
;
3141 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3142 uint64_t used
, comp
, uncomp
;
3143 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3144 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3151 snaplist_destroy(list_t
*l
, void *tag
)
3153 struct promotenode
*snap
;
3155 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3158 while ((snap
= list_tail(l
)) != NULL
) {
3159 list_remove(l
, snap
);
3160 dsl_dataset_rele(snap
->ds
, tag
);
3161 kmem_free(snap
, sizeof (*snap
));
3167 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
3171 struct promotenode
*snap
;
3173 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3177 dd
= ddpa
->ddpa_clone
->ds_dir
;
3179 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3180 !dsl_dir_is_clone(dd
)) {
3181 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3182 return (SET_ERROR(EINVAL
));
3185 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3186 &ddpa
->shared_snaps
, tag
);
3190 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3191 &ddpa
->clone_snaps
, tag
);
3195 snap
= list_head(&ddpa
->shared_snaps
);
3196 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3197 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3198 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3199 &ddpa
->origin_snaps
, tag
);
3203 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3204 error
= dsl_dataset_hold_obj(dp
,
3205 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3206 tag
, &ddpa
->origin_origin
);
3212 promote_rele(ddpa
, tag
);
3217 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
3219 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3220 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3221 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3222 if (ddpa
->origin_origin
!= NULL
)
3223 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3224 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3230 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3231 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3234 dsl_dataset_promote(const char *name
, char *conflsnap
)
3236 dsl_dataset_promote_arg_t ddpa
= { 0 };
3239 nvpair_t
*snap_pair
;
3243 * We will modify space proportional to the number of
3244 * snapshots. Compute numsnaps.
3246 error
= dmu_objset_hold(name
, FTAG
, &os
);
3249 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3250 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3252 dmu_objset_rele(os
, FTAG
);
3256 ddpa
.ddpa_clonename
= name
;
3257 ddpa
.err_ds
= fnvlist_alloc();
3260 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3261 dsl_dataset_promote_sync
, &ddpa
,
3262 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3265 * Return the first conflicting snapshot found.
3267 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3268 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3269 (void) strcpy(conflsnap
, nvpair_name(snap_pair
));
3271 fnvlist_free(ddpa
.err_ds
);
3276 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3277 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3280 * "slack" factor for received datasets with refquota set on them.
3281 * See the bottom of this function for details on its use.
3283 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3284 spa_asize_inflation
;
3285 int64_t unused_refres_delta
;
3287 /* they should both be heads */
3288 if (clone
->ds_is_snapshot
||
3289 origin_head
->ds_is_snapshot
)
3290 return (SET_ERROR(EINVAL
));
3292 /* if we are not forcing, the branch point should be just before them */
3293 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3294 return (SET_ERROR(EINVAL
));
3296 /* clone should be the clone (unless they are unrelated) */
3297 if (clone
->ds_prev
!= NULL
&&
3298 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3299 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3300 return (SET_ERROR(EINVAL
));
3302 /* the clone should be a child of the origin */
3303 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3304 return (SET_ERROR(EINVAL
));
3306 /* origin_head shouldn't be modified unless 'force' */
3308 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3309 return (SET_ERROR(ETXTBSY
));
3311 /* origin_head should have no long holds (e.g. is not mounted) */
3312 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3313 return (SET_ERROR(EBUSY
));
3315 /* check amount of any unconsumed refreservation */
3316 unused_refres_delta
=
3317 (int64_t)MIN(origin_head
->ds_reserved
,
3318 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3319 (int64_t)MIN(origin_head
->ds_reserved
,
3320 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3322 if (unused_refres_delta
> 0 &&
3323 unused_refres_delta
>
3324 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3325 return (SET_ERROR(ENOSPC
));
3328 * The clone can't be too much over the head's refquota.
3330 * To ensure that the entire refquota can be used, we allow one
3331 * transaction to exceed the the refquota. Therefore, this check
3332 * needs to also allow for the space referenced to be more than the
3333 * refquota. The maximum amount of space that one transaction can use
3334 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3335 * overage ensures that we are able to receive a filesystem that
3336 * exceeds the refquota on the source system.
3338 * So that overage is the refquota_slack we use below.
3340 if (origin_head
->ds_quota
!= 0 &&
3341 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3342 origin_head
->ds_quota
+ refquota_slack
)
3343 return (SET_ERROR(EDQUOT
));
3349 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
3350 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
3352 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3353 int64_t unused_refres_delta
;
3355 ASSERT(clone
->ds_reserved
== 0);
3357 * NOTE: On DEBUG kernels there could be a race between this and
3358 * the check function if spa_asize_inflation is adjusted...
3360 ASSERT(origin_head
->ds_quota
== 0 ||
3361 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
3362 DMU_MAX_ACCESS
* spa_asize_inflation
);
3363 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
3366 * Swap per-dataset feature flags.
3368 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
3369 if (!(spa_feature_table
[f
].fi_flags
&
3370 ZFEATURE_FLAG_PER_DATASET
)) {
3371 ASSERT(!clone
->ds_feature_inuse
[f
]);
3372 ASSERT(!origin_head
->ds_feature_inuse
[f
]);
3376 boolean_t clone_inuse
= clone
->ds_feature_inuse
[f
];
3377 boolean_t origin_head_inuse
= origin_head
->ds_feature_inuse
[f
];
3380 dsl_dataset_deactivate_feature(clone
->ds_object
, f
, tx
);
3381 clone
->ds_feature_inuse
[f
] = B_FALSE
;
3383 if (origin_head_inuse
) {
3384 dsl_dataset_deactivate_feature(origin_head
->ds_object
,
3386 origin_head
->ds_feature_inuse
[f
] = B_FALSE
;
3389 dsl_dataset_activate_feature(origin_head
->ds_object
,
3391 origin_head
->ds_feature_inuse
[f
] = B_TRUE
;
3393 if (origin_head_inuse
) {
3394 dsl_dataset_activate_feature(clone
->ds_object
, f
, tx
);
3395 clone
->ds_feature_inuse
[f
] = B_TRUE
;
3399 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3400 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3402 if (clone
->ds_objset
!= NULL
) {
3403 dmu_objset_evict(clone
->ds_objset
);
3404 clone
->ds_objset
= NULL
;
3407 if (origin_head
->ds_objset
!= NULL
) {
3408 dmu_objset_evict(origin_head
->ds_objset
);
3409 origin_head
->ds_objset
= NULL
;
3412 unused_refres_delta
=
3413 (int64_t)MIN(origin_head
->ds_reserved
,
3414 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3415 (int64_t)MIN(origin_head
->ds_reserved
,
3416 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3419 * Reset origin's unique bytes, if it exists.
3421 if (clone
->ds_prev
) {
3422 dsl_dataset_t
*origin
= clone
->ds_prev
;
3423 uint64_t comp
, uncomp
;
3425 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3426 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3427 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3428 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3433 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3434 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3436 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3437 dsl_dataset_phys(origin_head
)->ds_bp
=
3438 dsl_dataset_phys(clone
)->ds_bp
;
3439 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3440 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3441 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3444 /* set dd_*_bytes */
3446 int64_t dused
, dcomp
, duncomp
;
3447 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3448 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3450 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3451 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3453 dsl_deadlist_space(&clone
->ds_deadlist
,
3454 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3455 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3456 &odl_used
, &odl_comp
, &odl_uncomp
);
3458 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3460 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3462 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3464 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3466 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3468 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3471 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3472 dused
, dcomp
, duncomp
, tx
);
3473 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3474 -dused
, -dcomp
, -duncomp
, tx
);
3477 * The difference in the space used by snapshots is the
3478 * difference in snapshot space due to the head's
3479 * deadlist (since that's the only thing that's
3480 * changing that affects the snapused).
3482 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3483 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3484 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3485 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3486 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3487 &odl_used
, &odl_comp
, &odl_uncomp
);
3488 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3489 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3492 /* swap ds_*_bytes */
3493 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3494 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3495 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3496 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3497 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3498 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3499 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3500 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3502 /* apply any parent delta for change in unconsumed refreservation */
3503 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3504 unused_refres_delta
, 0, 0, tx
);
3509 dsl_deadlist_close(&clone
->ds_deadlist
);
3510 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3511 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3512 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3513 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3514 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3515 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3516 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3518 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3520 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3521 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3525 * Given a pool name and a dataset object number in that pool,
3526 * return the name of that dataset.
3529 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3535 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3539 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3541 dsl_dataset_name(ds
, buf
);
3542 dsl_dataset_rele(ds
, FTAG
);
3544 dsl_pool_rele(dp
, FTAG
);
3550 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3551 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3555 ASSERT3S(asize
, >, 0);
3558 * *ref_rsrv is the portion of asize that will come from any
3559 * unconsumed refreservation space.
3563 mutex_enter(&ds
->ds_lock
);
3565 * Make a space adjustment for reserved bytes.
3567 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3569 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3571 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3573 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3576 if (!check_quota
|| ds
->ds_quota
== 0) {
3577 mutex_exit(&ds
->ds_lock
);
3581 * If they are requesting more space, and our current estimate
3582 * is over quota, they get to try again unless the actual
3583 * on-disk is over quota and there are no pending changes (which
3584 * may free up space for us).
3586 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3589 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3590 error
= SET_ERROR(ERESTART
);
3592 error
= SET_ERROR(EDQUOT
);
3594 mutex_exit(&ds
->ds_lock
);
3599 typedef struct dsl_dataset_set_qr_arg
{
3600 const char *ddsqra_name
;
3601 zprop_source_t ddsqra_source
;
3602 uint64_t ddsqra_value
;
3603 } dsl_dataset_set_qr_arg_t
;
3608 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3610 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3611 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3616 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3617 return (SET_ERROR(ENOTSUP
));
3619 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3623 if (ds
->ds_is_snapshot
) {
3624 dsl_dataset_rele(ds
, FTAG
);
3625 return (SET_ERROR(EINVAL
));
3628 error
= dsl_prop_predict(ds
->ds_dir
,
3629 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3630 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3632 dsl_dataset_rele(ds
, FTAG
);
3637 dsl_dataset_rele(ds
, FTAG
);
3641 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3642 newval
< ds
->ds_reserved
) {
3643 dsl_dataset_rele(ds
, FTAG
);
3644 return (SET_ERROR(ENOSPC
));
3647 dsl_dataset_rele(ds
, FTAG
);
3652 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3654 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3655 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3656 dsl_dataset_t
*ds
= NULL
;
3659 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3661 dsl_prop_set_sync_impl(ds
,
3662 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3663 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3664 &ddsqra
->ddsqra_value
, tx
);
3666 VERIFY0(dsl_prop_get_int_ds(ds
,
3667 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3669 if (ds
->ds_quota
!= newval
) {
3670 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3671 ds
->ds_quota
= newval
;
3673 dsl_dataset_rele(ds
, FTAG
);
3677 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3680 dsl_dataset_set_qr_arg_t ddsqra
;
3682 ddsqra
.ddsqra_name
= dsname
;
3683 ddsqra
.ddsqra_source
= source
;
3684 ddsqra
.ddsqra_value
= refquota
;
3686 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3687 dsl_dataset_set_refquota_sync
, &ddsqra
, 0, ZFS_SPACE_CHECK_NONE
));
3691 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
3693 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3694 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3697 uint64_t newval
, unique
;
3699 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
3700 return (SET_ERROR(ENOTSUP
));
3702 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3706 if (ds
->ds_is_snapshot
) {
3707 dsl_dataset_rele(ds
, FTAG
);
3708 return (SET_ERROR(EINVAL
));
3711 error
= dsl_prop_predict(ds
->ds_dir
,
3712 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3713 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3715 dsl_dataset_rele(ds
, FTAG
);
3720 * If we are doing the preliminary check in open context, the
3721 * space estimates may be inaccurate.
3723 if (!dmu_tx_is_syncing(tx
)) {
3724 dsl_dataset_rele(ds
, FTAG
);
3728 mutex_enter(&ds
->ds_lock
);
3729 if (!DS_UNIQUE_IS_ACCURATE(ds
))
3730 dsl_dataset_recalc_head_uniq(ds
);
3731 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3732 mutex_exit(&ds
->ds_lock
);
3734 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
3735 uint64_t delta
= MAX(unique
, newval
) -
3736 MAX(unique
, ds
->ds_reserved
);
3739 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
3740 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
3741 dsl_dataset_rele(ds
, FTAG
);
3742 return (SET_ERROR(ENOSPC
));
3746 dsl_dataset_rele(ds
, FTAG
);
3751 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
3752 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
3758 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
3759 source
, sizeof (value
), 1, &value
, tx
);
3761 VERIFY0(dsl_prop_get_int_ds(ds
,
3762 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
3764 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3765 mutex_enter(&ds
->ds_dir
->dd_lock
);
3766 mutex_enter(&ds
->ds_lock
);
3767 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
3768 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
3769 delta
= MAX(0, (int64_t)(newval
- unique
)) -
3770 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
3771 ds
->ds_reserved
= newval
;
3772 mutex_exit(&ds
->ds_lock
);
3774 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
3775 mutex_exit(&ds
->ds_dir
->dd_lock
);
3779 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
3781 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3782 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3783 dsl_dataset_t
*ds
= NULL
;
3785 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3786 dsl_dataset_set_refreservation_sync_impl(ds
,
3787 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
3788 dsl_dataset_rele(ds
, FTAG
);
3792 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
3793 uint64_t refreservation
)
3795 dsl_dataset_set_qr_arg_t ddsqra
;
3797 ddsqra
.ddsqra_name
= dsname
;
3798 ddsqra
.ddsqra_source
= source
;
3799 ddsqra
.ddsqra_value
= refreservation
;
3801 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
3802 dsl_dataset_set_refreservation_sync
, &ddsqra
,
3803 0, ZFS_SPACE_CHECK_NONE
));
3807 * Return (in *usedp) the amount of space written in new that is not
3808 * present in oldsnap. New may be a snapshot or the head. Old must be
3809 * a snapshot before new, in new's filesystem (or its origin). If not then
3810 * fail and return EINVAL.
3812 * The written space is calculated by considering two components: First, we
3813 * ignore any freed space, and calculate the written as new's used space
3814 * minus old's used space. Next, we add in the amount of space that was freed
3815 * between the two snapshots, thus reducing new's used space relative to old's.
3816 * Specifically, this is the space that was born before old->ds_creation_txg,
3817 * and freed before new (ie. on new's deadlist or a previous deadlist).
3819 * space freed [---------------------]
3820 * snapshots ---O-------O--------O-------O------
3824 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
3825 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3829 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
3831 ASSERT(dsl_pool_config_held(dp
));
3834 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
3835 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
3838 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
3839 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
3842 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
3843 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
3845 snapobj
= new->ds_object
;
3846 while (snapobj
!= oldsnap
->ds_object
) {
3847 dsl_dataset_t
*snap
;
3848 uint64_t used
, comp
, uncomp
;
3850 if (snapobj
== new->ds_object
) {
3853 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
3858 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
3859 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
3861 * The blocks in the deadlist can not be born after
3862 * ds_prev_snap_txg, so get the whole deadlist space,
3863 * which is more efficient (especially for old-format
3864 * deadlists). Unfortunately the deadlist code
3865 * doesn't have enough information to make this
3866 * optimization itself.
3868 dsl_deadlist_space(&snap
->ds_deadlist
,
3869 &used
, &comp
, &uncomp
);
3871 dsl_deadlist_space_range(&snap
->ds_deadlist
,
3872 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
3873 &used
, &comp
, &uncomp
);
3880 * If we get to the beginning of the chain of snapshots
3881 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
3882 * was not a snapshot of/before new.
3884 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
3886 dsl_dataset_rele(snap
, FTAG
);
3888 err
= SET_ERROR(EINVAL
);
3897 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
3898 * lastsnap, and all snapshots in between are deleted.
3900 * blocks that would be freed [---------------------------]
3901 * snapshots ---O-------O--------O-------O--------O
3902 * firstsnap lastsnap
3904 * This is the set of blocks that were born after the snap before firstsnap,
3905 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
3906 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
3907 * We calculate this by iterating over the relevant deadlists (from the snap
3908 * after lastsnap, backward to the snap after firstsnap), summing up the
3909 * space on the deadlist that was born after the snap before firstsnap.
3912 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
3913 dsl_dataset_t
*lastsnap
,
3914 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
3918 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
3920 ASSERT(firstsnap
->ds_is_snapshot
);
3921 ASSERT(lastsnap
->ds_is_snapshot
);
3924 * Check that the snapshots are in the same dsl_dir, and firstsnap
3925 * is before lastsnap.
3927 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
3928 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
3929 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
3930 return (SET_ERROR(EINVAL
));
3932 *usedp
= *compp
= *uncompp
= 0;
3934 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
3935 while (snapobj
!= firstsnap
->ds_object
) {
3937 uint64_t used
, comp
, uncomp
;
3939 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
3943 dsl_deadlist_space_range(&ds
->ds_deadlist
,
3944 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
3945 &used
, &comp
, &uncomp
);
3950 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3951 ASSERT3U(snapobj
, !=, 0);
3952 dsl_dataset_rele(ds
, FTAG
);
3958 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
3959 * For example, they could both be snapshots of the same filesystem, and
3960 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
3961 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
3962 * filesystem. Or 'earlier' could be the origin's origin.
3964 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
3967 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
3968 uint64_t earlier_txg
)
3970 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
3974 ASSERT(dsl_pool_config_held(dp
));
3975 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
3977 if (earlier_txg
== 0)
3978 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
3980 if (later
->ds_is_snapshot
&&
3981 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
3984 if (later
->ds_dir
== earlier
->ds_dir
)
3986 if (!dsl_dir_is_clone(later
->ds_dir
))
3989 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
3991 dsl_dataset_t
*origin
;
3992 error
= dsl_dataset_hold_obj(dp
,
3993 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
3996 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
3997 dsl_dataset_rele(origin
, FTAG
);
4002 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4004 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4005 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4009 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4011 dmu_object_info_t doi
;
4013 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4014 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4018 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4020 return (dsl_dataset_is_zapified(ds
) &&
4021 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4022 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4025 #if defined(_KERNEL) && defined(HAVE_SPL)
4027 module_param(zfs_max_recordsize
, int, 0644);
4028 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4030 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
4031 module_param(zfs_max_recordsize
, int, 0444);
4032 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4035 EXPORT_SYMBOL(dsl_dataset_hold
);
4036 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4037 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4038 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4039 EXPORT_SYMBOL(dsl_dataset_own
);
4040 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4041 EXPORT_SYMBOL(dsl_dataset_name
);
4042 EXPORT_SYMBOL(dsl_dataset_rele
);
4043 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4044 EXPORT_SYMBOL(dsl_dataset_disown
);
4045 EXPORT_SYMBOL(dsl_dataset_tryown
);
4046 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4047 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4048 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4049 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4050 EXPORT_SYMBOL(dsl_dataset_promote
);
4051 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4052 EXPORT_SYMBOL(dsl_dataset_user_release
);
4053 EXPORT_SYMBOL(dsl_dataset_get_holds
);
4054 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
4055 EXPORT_SYMBOL(dsl_dataset_get_spa
);
4056 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
4057 EXPORT_SYMBOL(dsl_dataset_space_written
);
4058 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
4059 EXPORT_SYMBOL(dsl_dataset_sync
);
4060 EXPORT_SYMBOL(dsl_dataset_block_born
);
4061 EXPORT_SYMBOL(dsl_dataset_block_kill
);
4062 EXPORT_SYMBOL(dsl_dataset_dirty
);
4063 EXPORT_SYMBOL(dsl_dataset_stats
);
4064 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
4065 EXPORT_SYMBOL(dsl_dataset_space
);
4066 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
4067 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
4068 EXPORT_SYMBOL(dsl_dataset_check_quota
);
4069 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
4070 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);