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 https://opensource.org/licenses/CDDL-1.0.
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, 2020 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.
31 * Copyright (c) 2019, Klara Inc.
32 * Copyright (c) 2019, Allan Jude
33 * Copyright (c) 2020 The FreeBSD Foundation [1]
35 * [1] Portions of this software were developed by Allan Jude
36 * under sponsorship from the FreeBSD Foundation.
39 #include <sys/dmu_objset.h>
40 #include <sys/dsl_dataset.h>
41 #include <sys/dsl_dir.h>
42 #include <sys/dsl_prop.h>
43 #include <sys/dsl_synctask.h>
44 #include <sys/dmu_traverse.h>
45 #include <sys/dmu_impl.h>
46 #include <sys/dmu_tx.h>
50 #include <sys/zfeature.h>
51 #include <sys/unique.h>
52 #include <sys/zfs_context.h>
53 #include <sys/zfs_ioctl.h>
55 #include <sys/spa_impl.h>
57 #include <sys/zfs_znode.h>
58 #include <sys/zfs_onexit.h>
60 #include <sys/dsl_scan.h>
61 #include <sys/dsl_deadlist.h>
62 #include <sys/dsl_destroy.h>
63 #include <sys/dsl_userhold.h>
64 #include <sys/dsl_bookmark.h>
65 #include <sys/policy.h>
66 #include <sys/dmu_send.h>
67 #include <sys/dmu_recv.h>
68 #include <sys/zio_compress.h>
69 #include <zfs_fletcher.h>
70 #include <sys/zio_checksum.h>
73 * The SPA supports block sizes up to 16MB. However, very large blocks
74 * can have an impact on i/o latency (e.g. tying up a spinning disk for
75 * ~300ms), and also potentially on the memory allocator. Therefore,
76 * we did not allow the recordsize to be set larger than zfs_max_recordsize
77 * (former default: 1MB). Larger blocks could be created by changing this
78 * tunable, and pools with larger blocks could always be imported and used,
79 * regardless of this setting.
81 * We do, however, still limit it by default to 1M on x86_32, because Linux's
82 * 3/1 memory split doesn't leave much room for 16M chunks.
85 uint_t zfs_max_recordsize
= 1 * 1024 * 1024;
87 uint_t zfs_max_recordsize
= 16 * 1024 * 1024;
89 static int zfs_allow_redacted_dataset_mount
= 0;
91 int zfs_snapshot_history_enabled
= 1;
93 #define SWITCH64(x, y) \
95 uint64_t __tmp = (x); \
100 #define DS_REF_MAX (1ULL << 62)
102 static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
,
103 uint64_t obj
, dmu_tx_t
*tx
);
104 static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
,
107 static void unload_zfeature(dsl_dataset_t
*ds
, spa_feature_t f
);
109 extern uint_t spa_asize_inflation
;
111 static zil_header_t zero_zil
;
114 * Figure out how much of this delta should be propagated to the dsl_dir
115 * layer. If there's a refreservation, that space has already been
116 * partially accounted for in our ancestors.
119 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
121 dsl_dataset_phys_t
*ds_phys
;
122 uint64_t old_bytes
, new_bytes
;
124 if (ds
->ds_reserved
== 0)
127 ds_phys
= dsl_dataset_phys(ds
);
128 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
129 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
131 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
132 return (new_bytes
- old_bytes
);
136 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
138 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
139 int used
= bp_get_dsize_sync(spa
, bp
);
140 int compressed
= BP_GET_PSIZE(bp
);
141 int uncompressed
= BP_GET_UCSIZE(bp
);
145 dprintf_bp(bp
, "ds=%p", ds
);
147 ASSERT(dmu_tx_is_syncing(tx
));
148 /* It could have been compressed away to nothing */
149 if (BP_IS_HOLE(bp
) || BP_IS_REDACTED(bp
))
151 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
152 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
154 dsl_pool_mos_diduse_space(tx
->tx_pool
,
155 used
, compressed
, uncompressed
);
159 ASSERT3U(BP_GET_LOGICAL_BIRTH(bp
), >,
160 dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
161 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
162 mutex_enter(&ds
->ds_lock
);
163 delta
= parent_delta(ds
, used
);
164 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
165 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
166 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
167 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
169 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
170 ds
->ds_feature_activation
[SPA_FEATURE_LARGE_BLOCKS
] =
175 f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
176 if (f
!= SPA_FEATURE_NONE
) {
177 ASSERT3S(spa_feature_table
[f
].fi_type
, ==,
178 ZFEATURE_TYPE_BOOLEAN
);
179 ds
->ds_feature_activation
[f
] = (void *)B_TRUE
;
182 f
= zio_compress_to_feature(BP_GET_COMPRESS(bp
));
183 if (f
!= SPA_FEATURE_NONE
) {
184 ASSERT3S(spa_feature_table
[f
].fi_type
, ==,
185 ZFEATURE_TYPE_BOOLEAN
);
186 ds
->ds_feature_activation
[f
] = (void *)B_TRUE
;
190 * Track block for livelist, but ignore embedded blocks because
191 * they do not need to be freed.
193 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
) &&
194 BP_GET_LOGICAL_BIRTH(bp
) > ds
->ds_dir
->dd_origin_txg
&&
195 !(BP_IS_EMBEDDED(bp
))) {
196 ASSERT(dsl_dir_is_clone(ds
->ds_dir
));
197 ASSERT(spa_feature_is_enabled(spa
,
198 SPA_FEATURE_LIVELIST
));
199 bplist_append(&ds
->ds_dir
->dd_pending_allocs
, bp
);
202 mutex_exit(&ds
->ds_lock
);
203 dsl_dir_diduse_transfer_space(ds
->ds_dir
, delta
,
204 compressed
, uncompressed
, used
,
205 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
209 * Called when the specified segment has been remapped, and is thus no
210 * longer referenced in the head dataset. The vdev must be indirect.
212 * If the segment is referenced by a snapshot, put it on the remap deadlist.
213 * Otherwise, add this segment to the obsolete spacemap.
216 dsl_dataset_block_remapped(dsl_dataset_t
*ds
, uint64_t vdev
, uint64_t offset
,
217 uint64_t size
, uint64_t birth
, dmu_tx_t
*tx
)
219 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
221 ASSERT(dmu_tx_is_syncing(tx
));
222 ASSERT(birth
<= tx
->tx_txg
);
223 ASSERT(!ds
->ds_is_snapshot
);
225 if (birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
226 spa_vdev_indirect_mark_obsolete(spa
, vdev
, offset
, size
, tx
);
229 dva_t
*dva
= &fakebp
.blk_dva
[0];
233 mutex_enter(&ds
->ds_remap_deadlist_lock
);
234 if (!dsl_dataset_remap_deadlist_exists(ds
)) {
235 dsl_dataset_create_remap_deadlist(ds
, tx
);
237 mutex_exit(&ds
->ds_remap_deadlist_lock
);
240 BP_SET_LOGICAL_BIRTH(&fakebp
, birth
);
241 DVA_SET_VDEV(dva
, vdev
);
242 DVA_SET_OFFSET(dva
, offset
);
243 DVA_SET_ASIZE(dva
, size
);
244 dsl_deadlist_insert(&ds
->ds_remap_deadlist
, &fakebp
, B_FALSE
,
250 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
253 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
255 int used
= bp_get_dsize_sync(spa
, bp
);
256 int compressed
= BP_GET_PSIZE(bp
);
257 int uncompressed
= BP_GET_UCSIZE(bp
);
259 if (BP_IS_HOLE(bp
) || BP_IS_REDACTED(bp
))
262 ASSERT(dmu_tx_is_syncing(tx
));
263 ASSERT(BP_GET_LOGICAL_BIRTH(bp
) <= tx
->tx_txg
);
266 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
267 dsl_pool_mos_diduse_space(tx
->tx_pool
,
268 -used
, -compressed
, -uncompressed
);
271 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
273 ASSERT(!ds
->ds_is_snapshot
);
274 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
277 * Track block for livelist, but ignore embedded blocks because
278 * they do not need to be freed.
280 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
) &&
281 BP_GET_LOGICAL_BIRTH(bp
) > ds
->ds_dir
->dd_origin_txg
&&
282 !(BP_IS_EMBEDDED(bp
))) {
283 ASSERT(dsl_dir_is_clone(ds
->ds_dir
));
284 ASSERT(spa_feature_is_enabled(spa
,
285 SPA_FEATURE_LIVELIST
));
286 bplist_append(&ds
->ds_dir
->dd_pending_frees
, bp
);
289 if (BP_GET_LOGICAL_BIRTH(bp
) > dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
292 dprintf_bp(bp
, "freeing ds=%llu", (u_longlong_t
)ds
->ds_object
);
293 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
295 mutex_enter(&ds
->ds_lock
);
296 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
297 !DS_UNIQUE_IS_ACCURATE(ds
));
298 delta
= parent_delta(ds
, -used
);
299 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
300 mutex_exit(&ds
->ds_lock
);
301 dsl_dir_diduse_transfer_space(ds
->ds_dir
,
302 delta
, -compressed
, -uncompressed
, -used
,
303 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
305 dprintf_bp(bp
, "putting on dead list: %s", "");
308 * We are here as part of zio's write done callback,
309 * which means we're a zio interrupt thread. We can't
310 * call dsl_deadlist_insert() now because it may block
311 * waiting for I/O. Instead, put bp on the deferred
312 * queue and let dsl_pool_sync() finish the job.
314 bplist_append(&ds
->ds_pending_deadlist
, bp
);
316 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, B_FALSE
, tx
);
318 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
319 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
320 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
321 /* if (logical birth > prev prev snap txg) prev unique += bs */
322 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
323 ds
->ds_object
&& BP_GET_LOGICAL_BIRTH(bp
) >
324 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
325 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
326 mutex_enter(&ds
->ds_prev
->ds_lock
);
327 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
328 mutex_exit(&ds
->ds_prev
->ds_lock
);
330 if (BP_GET_LOGICAL_BIRTH(bp
) > ds
->ds_dir
->dd_origin_txg
) {
331 dsl_dir_transfer_space(ds
->ds_dir
, used
,
332 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
336 dsl_bookmark_block_killed(ds
, bp
, tx
);
338 mutex_enter(&ds
->ds_lock
);
339 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
340 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
341 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
342 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
343 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
344 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
345 mutex_exit(&ds
->ds_lock
);
350 struct feature_type_uint64_array_arg
{
356 unload_zfeature(dsl_dataset_t
*ds
, spa_feature_t f
)
358 switch (spa_feature_table
[f
].fi_type
) {
359 case ZFEATURE_TYPE_BOOLEAN
:
361 case ZFEATURE_TYPE_UINT64_ARRAY
:
363 struct feature_type_uint64_array_arg
*ftuaa
= ds
->ds_feature
[f
];
364 kmem_free(ftuaa
->array
, ftuaa
->length
* sizeof (uint64_t));
365 kmem_free(ftuaa
, sizeof (*ftuaa
));
369 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
374 load_zfeature(objset_t
*mos
, dsl_dataset_t
*ds
, spa_feature_t f
)
377 switch (spa_feature_table
[f
].fi_type
) {
378 case ZFEATURE_TYPE_BOOLEAN
:
379 err
= zap_contains(mos
, ds
->ds_object
,
380 spa_feature_table
[f
].fi_guid
);
382 ds
->ds_feature
[f
] = (void *)B_TRUE
;
384 ASSERT3U(err
, ==, ENOENT
);
388 case ZFEATURE_TYPE_UINT64_ARRAY
:
390 uint64_t int_size
, num_int
;
392 err
= zap_length(mos
, ds
->ds_object
,
393 spa_feature_table
[f
].fi_guid
, &int_size
, &num_int
);
395 ASSERT3U(err
, ==, ENOENT
);
399 ASSERT3U(int_size
, ==, sizeof (uint64_t));
400 data
= kmem_alloc(int_size
* num_int
, KM_SLEEP
);
401 VERIFY0(zap_lookup(mos
, ds
->ds_object
,
402 spa_feature_table
[f
].fi_guid
, int_size
, num_int
, data
));
403 struct feature_type_uint64_array_arg
*ftuaa
=
404 kmem_alloc(sizeof (*ftuaa
), KM_SLEEP
);
405 ftuaa
->length
= num_int
;
407 ds
->ds_feature
[f
] = ftuaa
;
411 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
417 * We have to release the fsid synchronously or we risk that a subsequent
418 * mount of the same dataset will fail to unique_insert the fsid. This
419 * failure would manifest itself as the fsid of this dataset changing
420 * between mounts which makes NFS clients quite unhappy.
423 dsl_dataset_evict_sync(void *dbu
)
425 dsl_dataset_t
*ds
= dbu
;
427 ASSERT(ds
->ds_owner
== NULL
);
429 unique_remove(ds
->ds_fsid_guid
);
433 dsl_dataset_evict_async(void *dbu
)
435 dsl_dataset_t
*ds
= dbu
;
437 ASSERT(ds
->ds_owner
== NULL
);
441 if (ds
->ds_objset
!= NULL
)
442 dmu_objset_evict(ds
->ds_objset
);
445 dsl_dataset_rele(ds
->ds_prev
, ds
);
449 dsl_bookmark_fini_ds(ds
);
451 bplist_destroy(&ds
->ds_pending_deadlist
);
452 if (dsl_deadlist_is_open(&ds
->ds_deadlist
))
453 dsl_deadlist_close(&ds
->ds_deadlist
);
454 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
455 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
457 dsl_dir_async_rele(ds
->ds_dir
, ds
);
459 ASSERT(!list_link_active(&ds
->ds_synced_link
));
461 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
462 if (dsl_dataset_feature_is_active(ds
, f
))
463 unload_zfeature(ds
, f
);
466 list_destroy(&ds
->ds_prop_cbs
);
467 mutex_destroy(&ds
->ds_lock
);
468 mutex_destroy(&ds
->ds_opening_lock
);
469 mutex_destroy(&ds
->ds_sendstream_lock
);
470 mutex_destroy(&ds
->ds_remap_deadlist_lock
);
471 zfs_refcount_destroy(&ds
->ds_longholds
);
472 rrw_destroy(&ds
->ds_bp_rwlock
);
474 kmem_free(ds
, sizeof (dsl_dataset_t
));
478 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
480 dsl_dataset_phys_t
*headphys
;
483 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
484 objset_t
*mos
= dp
->dp_meta_objset
;
486 if (ds
->ds_snapname
[0])
488 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
491 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
495 headphys
= headdbuf
->db_data
;
496 err
= zap_value_search(dp
->dp_meta_objset
,
497 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
498 if (err
!= 0 && zfs_recover
== B_TRUE
) {
500 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
501 "SNAPOBJ=%llu-ERR=%d",
502 (unsigned long long)ds
->ds_object
, err
);
504 dmu_buf_rele(headdbuf
, FTAG
);
509 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
511 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
512 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
516 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
519 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
520 value
, mt
, NULL
, 0, NULL
);
521 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
522 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
527 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
530 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
531 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
535 dsl_dir_snap_cmtime_update(ds
->ds_dir
, tx
);
537 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
540 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
541 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
542 err
= zap_remove(mos
, snapobj
, name
, tx
);
544 if (err
== 0 && adj_cnt
)
545 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
546 DD_FIELD_SNAPSHOT_COUNT
, tx
);
552 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, const void *tag
)
554 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
555 boolean_t result
= B_FALSE
;
557 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
558 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
560 if (ds
== dmu_buf_get_user(dbuf
))
563 dmu_buf_rele(dbuf
, tag
);
570 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, const void *tag
,
573 objset_t
*mos
= dp
->dp_meta_objset
;
577 dmu_object_info_t doi
;
579 ASSERT(dsl_pool_config_held(dp
));
581 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
585 /* Make sure dsobj has the correct object type. */
586 dmu_object_info_from_db(dbuf
, &doi
);
587 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
588 dmu_buf_rele(dbuf
, tag
);
589 return (SET_ERROR(EINVAL
));
592 ds
= dmu_buf_get_user(dbuf
);
594 dsl_dataset_t
*winner
= NULL
;
596 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
598 ds
->ds_object
= dsobj
;
599 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
600 list_link_init(&ds
->ds_synced_link
);
602 err
= dsl_dir_hold_obj(dp
, dsl_dataset_phys(ds
)->ds_dir_obj
,
603 NULL
, ds
, &ds
->ds_dir
);
605 kmem_free(ds
, sizeof (dsl_dataset_t
));
606 dmu_buf_rele(dbuf
, tag
);
610 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
611 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
612 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
613 mutex_init(&ds
->ds_remap_deadlist_lock
,
614 NULL
, MUTEX_DEFAULT
, NULL
);
615 rrw_init(&ds
->ds_bp_rwlock
, B_FALSE
);
616 zfs_refcount_create(&ds
->ds_longholds
);
618 bplist_create(&ds
->ds_pending_deadlist
);
620 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendstatus_t
),
621 offsetof(dmu_sendstatus_t
, dss_link
));
623 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
624 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
626 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
629 for (f
= 0; f
< SPA_FEATURES
; f
++) {
630 if (!(spa_feature_table
[f
].fi_flags
&
631 ZFEATURE_FLAG_PER_DATASET
))
633 err
= load_zfeature(mos
, ds
, f
);
637 if (!ds
->ds_is_snapshot
) {
638 ds
->ds_snapname
[0] = '\0';
639 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
640 err
= dsl_dataset_hold_obj(dp
,
641 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
645 goto after_dsl_bookmark_fini
;
646 err
= dsl_bookmark_init_ds(ds
);
648 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
649 err
= dsl_dataset_get_snapname(ds
);
651 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
653 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
654 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
659 if (err
== 0 && !ds
->ds_is_snapshot
) {
660 err
= dsl_prop_get_int_ds(ds
,
661 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
664 err
= dsl_prop_get_int_ds(ds
,
665 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
669 ds
->ds_reserved
= ds
->ds_quota
= 0;
672 if (err
== 0 && ds
->ds_dir
->dd_crypto_obj
!= 0 &&
673 ds
->ds_is_snapshot
&&
674 zap_contains(mos
, dsobj
, DS_FIELD_IVSET_GUID
) != 0) {
675 dp
->dp_spa
->spa_errata
=
676 ZPOOL_ERRATA_ZOL_8308_ENCRYPTION
;
679 dsl_deadlist_open(&ds
->ds_deadlist
,
680 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
681 uint64_t remap_deadlist_obj
=
682 dsl_dataset_get_remap_deadlist_object(ds
);
683 if (remap_deadlist_obj
!= 0) {
684 dsl_deadlist_open(&ds
->ds_remap_deadlist
, mos
,
688 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict_sync
,
689 dsl_dataset_evict_async
, &ds
->ds_dbuf
);
691 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
693 if (err
!= 0 || winner
!= NULL
) {
694 dsl_deadlist_close(&ds
->ds_deadlist
);
695 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
696 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
697 dsl_bookmark_fini_ds(ds
);
698 after_dsl_bookmark_fini
:
700 dsl_dataset_rele(ds
->ds_prev
, ds
);
701 dsl_dir_rele(ds
->ds_dir
, ds
);
702 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
703 if (dsl_dataset_feature_is_active(ds
, f
))
704 unload_zfeature(ds
, f
);
707 list_destroy(&ds
->ds_prop_cbs
);
708 list_destroy(&ds
->ds_sendstreams
);
709 bplist_destroy(&ds
->ds_pending_deadlist
);
710 mutex_destroy(&ds
->ds_lock
);
711 mutex_destroy(&ds
->ds_opening_lock
);
712 mutex_destroy(&ds
->ds_sendstream_lock
);
713 mutex_destroy(&ds
->ds_remap_deadlist_lock
);
714 zfs_refcount_destroy(&ds
->ds_longholds
);
715 rrw_destroy(&ds
->ds_bp_rwlock
);
716 kmem_free(ds
, sizeof (dsl_dataset_t
));
718 dmu_buf_rele(dbuf
, tag
);
724 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
725 if (ds
->ds_fsid_guid
!=
726 dsl_dataset_phys(ds
)->ds_fsid_guid
) {
727 zfs_dbgmsg("ds_fsid_guid changed from "
728 "%llx to %llx for pool %s dataset id %llu",
730 dsl_dataset_phys(ds
)->ds_fsid_guid
,
731 (long long)ds
->ds_fsid_guid
,
732 spa_name(dp
->dp_spa
),
733 (u_longlong_t
)dsobj
);
738 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
739 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
740 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
741 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
742 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
749 dsl_dataset_create_key_mapping(dsl_dataset_t
*ds
)
751 dsl_dir_t
*dd
= ds
->ds_dir
;
753 if (dd
->dd_crypto_obj
== 0)
756 return (spa_keystore_create_mapping(dd
->dd_pool
->dp_spa
,
757 ds
, ds
, &ds
->ds_key_mapping
));
761 dsl_dataset_hold_obj_flags(dsl_pool_t
*dp
, uint64_t dsobj
,
762 ds_hold_flags_t flags
, const void *tag
, dsl_dataset_t
**dsp
)
766 err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
770 ASSERT3P(*dsp
, !=, NULL
);
772 if (flags
& DS_HOLD_FLAG_DECRYPT
) {
773 err
= dsl_dataset_create_key_mapping(*dsp
);
775 dsl_dataset_rele(*dsp
, tag
);
782 dsl_dataset_hold_flags(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
783 const void *tag
, dsl_dataset_t
**dsp
)
786 const char *snapname
;
791 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
795 ASSERT(dsl_pool_config_held(dp
));
796 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
798 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
, &ds
);
800 err
= SET_ERROR(ENOENT
);
802 /* we may be looking for a snapshot */
803 if (err
== 0 && snapname
!= NULL
) {
804 dsl_dataset_t
*snap_ds
;
806 if (*snapname
++ != '@') {
807 dsl_dataset_rele_flags(ds
, flags
, tag
);
808 dsl_dir_rele(dd
, FTAG
);
809 return (SET_ERROR(ENOENT
));
812 dprintf("looking for snapshot '%s'\n", snapname
);
813 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
815 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
,
818 dsl_dataset_rele_flags(ds
, flags
, tag
);
821 mutex_enter(&snap_ds
->ds_lock
);
822 if (snap_ds
->ds_snapname
[0] == 0)
823 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
824 sizeof (snap_ds
->ds_snapname
));
825 mutex_exit(&snap_ds
->ds_lock
);
831 dsl_dir_rele(dd
, FTAG
);
836 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
, const void *tag
,
839 return (dsl_dataset_hold_flags(dp
, name
, 0, tag
, dsp
));
843 dsl_dataset_own_obj_impl(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
844 const void *tag
, boolean_t override
, dsl_dataset_t
**dsp
)
846 int err
= dsl_dataset_hold_obj_flags(dp
, dsobj
, flags
, tag
, dsp
);
849 if (!dsl_dataset_tryown(*dsp
, tag
, override
)) {
850 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
852 return (SET_ERROR(EBUSY
));
859 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
860 const void *tag
, dsl_dataset_t
**dsp
)
862 return (dsl_dataset_own_obj_impl(dp
, dsobj
, flags
, tag
, B_FALSE
, dsp
));
866 dsl_dataset_own_obj_force(dsl_pool_t
*dp
, uint64_t dsobj
,
867 ds_hold_flags_t flags
, const void *tag
, dsl_dataset_t
**dsp
)
869 return (dsl_dataset_own_obj_impl(dp
, dsobj
, flags
, tag
, B_TRUE
, dsp
));
873 dsl_dataset_own_impl(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
874 const void *tag
, boolean_t override
, dsl_dataset_t
**dsp
)
876 int err
= dsl_dataset_hold_flags(dp
, name
, flags
, tag
, dsp
);
879 if (!dsl_dataset_tryown(*dsp
, tag
, override
)) {
880 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
881 return (SET_ERROR(EBUSY
));
887 dsl_dataset_own_force(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
888 const void *tag
, dsl_dataset_t
**dsp
)
890 return (dsl_dataset_own_impl(dp
, name
, flags
, tag
, B_TRUE
, dsp
));
894 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
895 const void *tag
, dsl_dataset_t
**dsp
)
897 return (dsl_dataset_own_impl(dp
, name
, flags
, tag
, B_FALSE
, dsp
));
901 * See the comment above dsl_pool_hold() for details. In summary, a long
902 * hold is used to prevent destruction of a dataset while the pool hold
903 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
905 * The dataset and pool must be held when this function is called. After it
906 * is called, the pool hold may be released while the dataset is still held
910 dsl_dataset_long_hold(dsl_dataset_t
*ds
, const void *tag
)
912 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
913 (void) zfs_refcount_add(&ds
->ds_longholds
, tag
);
917 dsl_dataset_long_rele(dsl_dataset_t
*ds
, const void *tag
)
919 (void) zfs_refcount_remove(&ds
->ds_longholds
, tag
);
922 /* Return B_TRUE if there are any long holds on this dataset. */
924 dsl_dataset_long_held(dsl_dataset_t
*ds
)
926 return (!zfs_refcount_is_zero(&ds
->ds_longholds
));
930 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
933 (void) strlcpy(name
, "mos", ZFS_MAX_DATASET_NAME_LEN
);
935 dsl_dir_name(ds
->ds_dir
, name
);
936 VERIFY0(dsl_dataset_get_snapname(ds
));
937 if (ds
->ds_snapname
[0]) {
938 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
939 <, ZFS_MAX_DATASET_NAME_LEN
);
941 * We use a "recursive" mutex so that we
942 * can call dprintf_ds() with ds_lock held.
944 if (!MUTEX_HELD(&ds
->ds_lock
)) {
945 mutex_enter(&ds
->ds_lock
);
946 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
947 ZFS_MAX_DATASET_NAME_LEN
), <,
948 ZFS_MAX_DATASET_NAME_LEN
);
949 mutex_exit(&ds
->ds_lock
);
951 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
952 ZFS_MAX_DATASET_NAME_LEN
), <,
953 ZFS_MAX_DATASET_NAME_LEN
);
960 dsl_dataset_namelen(dsl_dataset_t
*ds
)
962 VERIFY0(dsl_dataset_get_snapname(ds
));
963 mutex_enter(&ds
->ds_lock
);
964 int len
= strlen(ds
->ds_snapname
);
965 mutex_exit(&ds
->ds_lock
);
966 /* add '@' if ds is a snap */
969 len
+= dsl_dir_namelen(ds
->ds_dir
);
974 dsl_dataset_rele(dsl_dataset_t
*ds
, const void *tag
)
976 dmu_buf_rele(ds
->ds_dbuf
, tag
);
980 dsl_dataset_remove_key_mapping(dsl_dataset_t
*ds
)
982 dsl_dir_t
*dd
= ds
->ds_dir
;
984 if (dd
== NULL
|| dd
->dd_crypto_obj
== 0)
987 (void) spa_keystore_remove_mapping(dd
->dd_pool
->dp_spa
,
992 dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
,
995 if (flags
& DS_HOLD_FLAG_DECRYPT
)
996 dsl_dataset_remove_key_mapping(ds
);
998 dsl_dataset_rele(ds
, tag
);
1002 dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, const void *tag
)
1004 ASSERT3P(ds
->ds_owner
, ==, tag
);
1005 ASSERT(ds
->ds_dbuf
!= NULL
);
1007 mutex_enter(&ds
->ds_lock
);
1008 ds
->ds_owner
= NULL
;
1009 mutex_exit(&ds
->ds_lock
);
1010 dsl_dataset_long_rele(ds
, tag
);
1011 dsl_dataset_rele_flags(ds
, flags
, tag
);
1015 dsl_dataset_tryown(dsl_dataset_t
*ds
, const void *tag
, boolean_t override
)
1017 boolean_t gotit
= FALSE
;
1019 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
1020 mutex_enter(&ds
->ds_lock
);
1021 if (ds
->ds_owner
== NULL
&& (override
|| !(DS_IS_INCONSISTENT(ds
) ||
1022 (dsl_dataset_feature_is_active(ds
,
1023 SPA_FEATURE_REDACTED_DATASETS
) &&
1024 !zfs_allow_redacted_dataset_mount
)))) {
1026 dsl_dataset_long_hold(ds
, tag
);
1029 mutex_exit(&ds
->ds_lock
);
1034 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
1037 mutex_enter(&ds
->ds_lock
);
1038 rv
= (ds
->ds_owner
!= NULL
);
1039 mutex_exit(&ds
->ds_lock
);
1044 zfeature_active(spa_feature_t f
, void *arg
)
1046 switch (spa_feature_table
[f
].fi_type
) {
1047 case ZFEATURE_TYPE_BOOLEAN
: {
1048 boolean_t val
= (boolean_t
)(uintptr_t)arg
;
1049 ASSERT(val
== B_FALSE
|| val
== B_TRUE
);
1052 case ZFEATURE_TYPE_UINT64_ARRAY
:
1054 * In this case, arg is a uint64_t array. The feature is active
1055 * if the array is non-null.
1057 return (arg
!= NULL
);
1059 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
1065 dsl_dataset_feature_is_active(dsl_dataset_t
*ds
, spa_feature_t f
)
1067 return (zfeature_active(f
, ds
->ds_feature
[f
]));
1071 * The buffers passed out by this function are references to internal buffers;
1072 * they should not be freed by callers of this function, and they should not be
1073 * used after the dataset has been released.
1076 dsl_dataset_get_uint64_array_feature(dsl_dataset_t
*ds
, spa_feature_t f
,
1077 uint64_t *outlength
, uint64_t **outp
)
1079 VERIFY(spa_feature_table
[f
].fi_type
& ZFEATURE_TYPE_UINT64_ARRAY
);
1080 if (!dsl_dataset_feature_is_active(ds
, f
)) {
1083 struct feature_type_uint64_array_arg
*ftuaa
= ds
->ds_feature
[f
];
1084 *outp
= ftuaa
->array
;
1085 *outlength
= ftuaa
->length
;
1090 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, void *arg
,
1093 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
1094 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
1097 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
1099 spa_feature_incr(spa
, f
, tx
);
1100 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1102 switch (spa_feature_table
[f
].fi_type
) {
1103 case ZFEATURE_TYPE_BOOLEAN
:
1104 ASSERT3S((boolean_t
)(uintptr_t)arg
, ==, B_TRUE
);
1105 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
1106 sizeof (zero
), 1, &zero
, tx
));
1108 case ZFEATURE_TYPE_UINT64_ARRAY
:
1110 struct feature_type_uint64_array_arg
*ftuaa
= arg
;
1111 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
1112 sizeof (uint64_t), ftuaa
->length
, ftuaa
->array
, tx
));
1116 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
1121 dsl_dataset_deactivate_feature_impl(dsl_dataset_t
*ds
, spa_feature_t f
,
1124 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
1125 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
1126 uint64_t dsobj
= ds
->ds_object
;
1128 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
1130 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
1131 spa_feature_decr(spa
, f
, tx
);
1132 ds
->ds_feature
[f
] = NULL
;
1136 dsl_dataset_deactivate_feature(dsl_dataset_t
*ds
, spa_feature_t f
, dmu_tx_t
*tx
)
1138 unload_zfeature(ds
, f
);
1139 dsl_dataset_deactivate_feature_impl(ds
, f
, tx
);
1143 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
1144 dsl_crypto_params_t
*dcp
, uint64_t flags
, dmu_tx_t
*tx
)
1146 dsl_pool_t
*dp
= dd
->dd_pool
;
1148 dsl_dataset_phys_t
*dsphys
;
1150 objset_t
*mos
= dp
->dp_meta_objset
;
1153 origin
= dp
->dp_origin_snap
;
1155 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
1156 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
1157 ASSERT(dmu_tx_is_syncing(tx
));
1158 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
1160 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1161 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1162 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1163 dmu_buf_will_dirty(dbuf
, tx
);
1164 dsphys
= dbuf
->db_data
;
1165 memset(dsphys
, 0, sizeof (dsl_dataset_phys_t
));
1166 dsphys
->ds_dir_obj
= dd
->dd_object
;
1167 dsphys
->ds_flags
= flags
;
1168 dsphys
->ds_fsid_guid
= unique_create();
1169 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1170 sizeof (dsphys
->ds_guid
));
1171 dsphys
->ds_snapnames_zapobj
=
1172 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
1173 DMU_OT_NONE
, 0, tx
);
1174 dsphys
->ds_creation_time
= gethrestime_sec();
1175 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
1177 if (origin
== NULL
) {
1178 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
1180 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
1182 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
1183 dsphys
->ds_prev_snap_txg
=
1184 dsl_dataset_phys(origin
)->ds_creation_txg
;
1185 dsphys
->ds_referenced_bytes
=
1186 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
1187 dsphys
->ds_compressed_bytes
=
1188 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
1189 dsphys
->ds_uncompressed_bytes
=
1190 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
1191 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
1192 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
1193 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
1196 * Inherit flags that describe the dataset's contents
1197 * (INCONSISTENT) or properties (Case Insensitive).
1199 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
1200 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
1202 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1203 if (zfeature_active(f
, origin
->ds_feature
[f
])) {
1204 dsl_dataset_activate_feature(dsobj
, f
,
1205 origin
->ds_feature
[f
], tx
);
1209 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
1210 dsl_dataset_phys(origin
)->ds_num_children
++;
1212 VERIFY0(dsl_dataset_hold_obj(dp
,
1213 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
1215 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
1216 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
1217 dsl_dataset_rele(ohds
, FTAG
);
1219 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
1220 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
1221 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
1223 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
1225 VERIFY0(zap_add_int(mos
,
1226 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
1230 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
1231 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
1232 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
1233 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
1234 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
1235 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
1237 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
1239 VERIFY0(zap_add_int(mos
,
1240 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
1245 /* handle encryption */
1246 dsl_dataset_create_crypt_sync(dsobj
, dd
, origin
, dcp
, tx
);
1248 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1249 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1251 dmu_buf_rele(dbuf
, FTAG
);
1253 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
1254 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
1260 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1264 VERIFY0(dmu_objset_from_ds(ds
, &os
));
1265 if (memcmp(&os
->os_zil_header
, &zero_zil
, sizeof (zero_zil
)) != 0) {
1266 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1269 memset(&os
->os_zil_header
, 0, sizeof (os
->os_zil_header
));
1270 if (os
->os_encrypted
)
1271 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_TRUE
;
1273 zio
= zio_root(dp
->dp_spa
, NULL
, NULL
, ZIO_FLAG_MUSTSUCCEED
);
1274 dsl_dataset_sync(ds
, zio
, tx
);
1275 VERIFY0(zio_wait(zio
));
1276 dsl_dataset_sync_done(ds
, tx
);
1281 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
1282 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
,
1283 dsl_crypto_params_t
*dcp
, dmu_tx_t
*tx
)
1285 dsl_pool_t
*dp
= pdd
->dd_pool
;
1286 uint64_t dsobj
, ddobj
;
1289 ASSERT(dmu_tx_is_syncing(tx
));
1290 ASSERT(lastname
[0] != '@');
1292 * Filesystems will eventually have their origin set to dp_origin_snap,
1293 * but that's taken care of in dsl_dataset_create_sync_dd. When
1294 * creating a filesystem, this function is called with origin equal to
1298 ASSERT3P(origin
, !=, dp
->dp_origin_snap
);
1300 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
1301 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
1303 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
, dcp
,
1304 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
1306 dsl_deleg_set_create_perms(dd
, tx
, cr
);
1309 * If we are creating a clone and the livelist feature is enabled,
1310 * add the entry DD_FIELD_LIVELIST to ZAP.
1312 if (origin
!= NULL
&&
1313 spa_feature_is_enabled(dp
->dp_spa
, SPA_FEATURE_LIVELIST
)) {
1314 objset_t
*mos
= dd
->dd_pool
->dp_meta_objset
;
1315 dsl_dir_zapify(dd
, tx
);
1316 uint64_t obj
= dsl_deadlist_alloc(mos
, tx
);
1317 VERIFY0(zap_add(mos
, dd
->dd_object
, DD_FIELD_LIVELIST
,
1318 sizeof (uint64_t), 1, &obj
, tx
));
1319 spa_feature_incr(dp
->dp_spa
, SPA_FEATURE_LIVELIST
, tx
);
1323 * Since we're creating a new node we know it's a leaf, so we can
1324 * initialize the counts if the limit feature is active.
1326 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
1328 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
1330 dsl_dir_zapify(dd
, tx
);
1331 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
1332 sizeof (cnt
), 1, &cnt
, tx
));
1333 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
1334 sizeof (cnt
), 1, &cnt
, tx
));
1337 dsl_dir_rele(dd
, FTAG
);
1340 * If we are creating a clone, make sure we zero out any stale
1341 * data from the origin snapshots zil header.
1343 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
1346 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
1347 dsl_dataset_zero_zil(ds
, tx
);
1348 dsl_dataset_rele(ds
, FTAG
);
1355 * The unique space in the head dataset can be calculated by subtracting
1356 * the space used in the most recent snapshot, that is still being used
1357 * in this file system, from the space currently in use. To figure out
1358 * the space in the most recent snapshot still in use, we need to take
1359 * the total space used in the snapshot and subtract out the space that
1360 * has been freed up since the snapshot was taken.
1363 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1366 uint64_t dlused
, dlcomp
, dluncomp
;
1368 ASSERT(!ds
->ds_is_snapshot
);
1370 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1371 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1375 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1377 ASSERT3U(dlused
, <=, mrs_used
);
1378 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1379 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1381 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1382 SPA_VERSION_UNIQUE_ACCURATE
)
1383 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1387 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1390 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1391 uint64_t count __maybe_unused
;
1394 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1395 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1398 * The err should not be ENOENT, but a bug in a previous version
1399 * of the code could cause upgrade_clones_cb() to not set
1400 * ds_next_snap_obj when it should, leading to a missing entry.
1401 * If we knew that the pool was created after
1402 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1403 * ENOENT. However, at least we can check that we don't have
1404 * too many entries in the next_clones_obj even after failing to
1409 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1411 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1416 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1418 return (&dsl_dataset_phys(ds
)->ds_bp
);
1422 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1424 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1428 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1432 if (ds
== NULL
) /* this is the meta-objset */
1435 ASSERT(ds
->ds_objset
!= NULL
);
1437 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1438 panic("dirtying snapshot!");
1440 /* Must not dirty a dataset in the same txg where it got snapshotted. */
1441 ASSERT3U(tx
->tx_txg
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
1443 dp
= ds
->ds_dir
->dd_pool
;
1444 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1445 objset_t
*os
= ds
->ds_objset
;
1447 /* up the hold count until we can be written out */
1448 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1450 /* if this dataset is encrypted, grab a reference to the DCK */
1451 if (ds
->ds_dir
->dd_crypto_obj
!= 0 &&
1452 !os
->os_raw_receive
&&
1453 !os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]) {
1454 ASSERT3P(ds
->ds_key_mapping
, !=, NULL
);
1455 key_mapping_add_ref(ds
->ds_key_mapping
, ds
);
1461 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1465 if (!dmu_tx_is_syncing(tx
))
1469 * If there's an fs-only reservation, any blocks that might become
1470 * owned by the snapshot dataset must be accommodated by space
1471 * outside of the reservation.
1473 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1474 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1475 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1476 return (SET_ERROR(ENOSPC
));
1479 * Propagate any reserved space for this snapshot to other
1480 * snapshot checks in this sync group.
1483 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1489 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1490 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
, proc_t
*proc
)
1495 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1497 if (!dmu_tx_is_syncing(tx
))
1501 * We don't allow multiple snapshots of the same txg. If there
1502 * is already one, try again.
1504 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1505 return (SET_ERROR(EAGAIN
));
1508 * Check for conflicting snapshot name.
1510 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1512 return (SET_ERROR(EEXIST
));
1513 if (error
!= ENOENT
)
1517 * We don't allow taking snapshots of inconsistent datasets, such as
1518 * those into which we are currently receiving. However, if we are
1519 * creating this snapshot as part of a receive, this check will be
1520 * executed atomically with respect to the completion of the receive
1521 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1522 * case we ignore this, knowing it will be fixed up for us shortly in
1523 * dmu_recv_end_sync().
1525 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1526 return (SET_ERROR(EBUSY
));
1529 * Skip the check for temporary snapshots or if we have already checked
1530 * the counts in dsl_dataset_snapshot_check. This means we really only
1531 * check the count here when we're receiving a stream.
1533 if (cnt
!= 0 && cr
!= NULL
) {
1534 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1535 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
, proc
);
1540 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1548 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1550 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1551 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1556 * Pre-compute how many total new snapshots will be created for each
1557 * level in the tree and below. This is needed for validating the
1558 * snapshot limit when either taking a recursive snapshot or when
1559 * taking multiple snapshots.
1561 * The problem is that the counts are not actually adjusted when
1562 * we are checking, only when we finally sync. For a single snapshot,
1563 * this is easy, the count will increase by 1 at each node up the tree,
1564 * but its more complicated for the recursive/multiple snapshot case.
1566 * The dsl_fs_ss_limit_check function does recursively check the count
1567 * at each level up the tree but since it is validating each snapshot
1568 * independently we need to be sure that we are validating the complete
1569 * count for the entire set of snapshots. We do this by rolling up the
1570 * counts for each component of the name into an nvlist and then
1571 * checking each of those cases with the aggregated count.
1573 * This approach properly handles not only the recursive snapshot
1574 * case (where we get all of those on the ddsa_snaps list) but also
1575 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1576 * validate the limit on 'a' using a count of 2).
1578 * We validate the snapshot names in the third loop and only report
1581 if (dmu_tx_is_syncing(tx
)) {
1583 nvlist_t
*cnt_track
= NULL
;
1584 cnt_track
= fnvlist_alloc();
1586 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1588 /* Rollup aggregated counts into the cnt_track list */
1589 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1591 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1595 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1596 pdelim
= strchr(nm
, '@');
1602 if (nvlist_lookup_uint64(cnt_track
, nm
,
1604 /* update existing entry */
1605 fnvlist_add_uint64(cnt_track
, nm
,
1609 fnvlist_add_uint64(cnt_track
, nm
, 1);
1612 pdelim
= strrchr(nm
, '/');
1615 } while (pdelim
!= NULL
);
1618 kmem_free(nm
, MAXPATHLEN
);
1620 /* Check aggregated counts at each level */
1621 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1622 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1628 name
= nvpair_name(pair
);
1629 cnt
= fnvpair_value_uint64(pair
);
1632 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1634 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1635 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1636 ddsa
->ddsa_cr
, ddsa
->ddsa_proc
);
1637 dsl_dataset_rele(ds
, FTAG
);
1641 if (ddsa
->ddsa_errors
!= NULL
)
1642 fnvlist_add_int32(ddsa
->ddsa_errors
,
1645 /* only report one error for this check */
1649 nvlist_free(cnt_track
);
1652 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1653 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1656 const char *name
, *atp
= NULL
;
1657 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1659 name
= nvpair_name(pair
);
1660 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1661 error
= SET_ERROR(ENAMETOOLONG
);
1663 atp
= strchr(name
, '@');
1665 error
= SET_ERROR(EINVAL
);
1667 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1670 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1672 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1673 error
= dsl_dataset_snapshot_check_impl(ds
,
1674 atp
+ 1, tx
, B_FALSE
, 0, NULL
, NULL
);
1675 dsl_dataset_rele(ds
, FTAG
);
1679 if (ddsa
->ddsa_errors
!= NULL
) {
1680 fnvlist_add_int32(ddsa
->ddsa_errors
,
1691 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1694 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1696 dsl_dataset_phys_t
*dsphys
;
1697 uint64_t dsobj
, crtxg
;
1698 objset_t
*mos
= dp
->dp_meta_objset
;
1699 objset_t
*os __maybe_unused
;
1701 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1704 * If we are on an old pool, the zil must not be active, in which
1705 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1707 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1708 dmu_objset_from_ds(ds
, &os
) != 0 ||
1709 memcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1710 sizeof (zero_zil
)) == 0);
1712 /* Should not snapshot a dirty dataset. */
1713 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1716 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1719 * The origin's ds_creation_txg has to be < TXG_INITIAL
1721 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1726 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1727 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1728 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1729 dmu_buf_will_dirty(dbuf
, tx
);
1730 dsphys
= dbuf
->db_data
;
1731 memset(dsphys
, 0, sizeof (dsl_dataset_phys_t
));
1732 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1733 dsphys
->ds_fsid_guid
= unique_create();
1734 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1735 sizeof (dsphys
->ds_guid
));
1736 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1737 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1738 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1739 dsphys
->ds_num_children
= 1;
1740 dsphys
->ds_creation_time
= gethrestime_sec();
1741 dsphys
->ds_creation_txg
= crtxg
;
1742 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1743 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1744 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1745 dsphys
->ds_uncompressed_bytes
=
1746 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1747 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1748 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1749 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1750 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1751 dmu_buf_rele(dbuf
, FTAG
);
1753 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1754 if (zfeature_active(f
, ds
->ds_feature
[f
])) {
1755 dsl_dataset_activate_feature(dsobj
, f
,
1756 ds
->ds_feature
[f
], tx
);
1760 ASSERT3U(ds
->ds_prev
!= 0, ==,
1761 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1763 uint64_t next_clones_obj
=
1764 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1765 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1767 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1768 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1770 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1771 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1772 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1773 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1774 } else if (next_clones_obj
!= 0) {
1775 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1776 dsphys
->ds_next_snap_obj
, tx
);
1777 VERIFY0(zap_add_int(mos
,
1778 next_clones_obj
, dsobj
, tx
));
1783 * If we have a reference-reservation on this dataset, we will
1784 * need to increase the amount of refreservation being charged
1785 * since our unique space is going to zero.
1787 if (ds
->ds_reserved
) {
1789 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1790 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1792 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1796 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1797 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1798 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1799 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1800 dsl_deadlist_close(&ds
->ds_deadlist
);
1801 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1802 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1803 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1804 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1805 dsl_bookmark_snapshotted(ds
, tx
);
1807 if (dsl_dataset_remap_deadlist_exists(ds
)) {
1808 uint64_t remap_deadlist_obj
=
1809 dsl_dataset_get_remap_deadlist_object(ds
);
1811 * Move the remap_deadlist to the snapshot. The head
1812 * will create a new remap deadlist on demand, from
1813 * dsl_dataset_block_remapped().
1815 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
1816 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
1818 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1819 VERIFY0(zap_add(mos
, dsobj
, DS_FIELD_REMAP_DEADLIST
,
1820 sizeof (remap_deadlist_obj
), 1, &remap_deadlist_obj
, tx
));
1824 * Create a ivset guid for this snapshot if the dataset is
1825 * encrypted. This may be overridden by a raw receive. A
1826 * previous implementation of this code did not have this
1827 * field as part of the on-disk format for ZFS encryption
1828 * (see errata #4). As part of the remediation for this
1829 * issue, we ask the user to enable the bookmark_v2 feature
1830 * which is now a dependency of the encryption feature. We
1831 * use this as a heuristic to determine when the user has
1832 * elected to correct any datasets created with the old code.
1833 * As a result, we only do this step if the bookmark_v2
1834 * feature is enabled, which limits the number of states a
1835 * given pool / dataset can be in with regards to terms of
1836 * correcting the issue.
1838 if (ds
->ds_dir
->dd_crypto_obj
!= 0 &&
1839 spa_feature_is_enabled(dp
->dp_spa
, SPA_FEATURE_BOOKMARK_V2
)) {
1840 uint64_t ivset_guid
= unique_create();
1842 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1843 VERIFY0(zap_add(mos
, dsobj
, DS_FIELD_IVSET_GUID
,
1844 sizeof (ivset_guid
), 1, &ivset_guid
, tx
));
1847 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1848 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1849 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1850 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1852 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1853 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1855 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1856 snapname
, 8, 1, &dsobj
, tx
));
1859 dsl_dataset_rele(ds
->ds_prev
, ds
);
1860 VERIFY0(dsl_dataset_hold_obj(dp
,
1861 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1863 dsl_scan_ds_snapshotted(ds
, tx
);
1865 dsl_dir_snap_cmtime_update(ds
->ds_dir
, tx
);
1867 if (zfs_snapshot_history_enabled
)
1868 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, " ");
1872 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1874 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1875 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1878 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1879 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1881 const char *name
, *atp
;
1882 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1884 name
= nvpair_name(pair
);
1885 atp
= strchr(name
, '@');
1886 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1887 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1889 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1890 if (ddsa
->ddsa_props
!= NULL
) {
1891 dsl_props_set_sync_impl(ds
->ds_prev
,
1892 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1894 dsl_dataset_rele(ds
, FTAG
);
1899 * The snapshots must all be in the same pool.
1900 * All-or-nothing: if there are any failures, nothing will be modified.
1903 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1905 dsl_dataset_snapshot_arg_t ddsa
;
1907 boolean_t needsuspend
;
1910 const char *firstname
;
1911 nvlist_t
*suspended
= NULL
;
1913 pair
= nvlist_next_nvpair(snaps
, NULL
);
1916 firstname
= nvpair_name(pair
);
1918 error
= spa_open(firstname
, &spa
, FTAG
);
1921 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1922 spa_close(spa
, FTAG
);
1925 suspended
= fnvlist_alloc();
1926 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1927 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1928 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1929 const char *snapname
= nvpair_name(pair
);
1933 atp
= strchr(snapname
, '@');
1935 error
= SET_ERROR(EINVAL
);
1938 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1940 error
= zil_suspend(fsname
, &cookie
);
1943 fnvlist_add_uint64(suspended
, fsname
,
1948 ddsa
.ddsa_snaps
= snaps
;
1949 ddsa
.ddsa_props
= props
;
1950 ddsa
.ddsa_errors
= errors
;
1951 ddsa
.ddsa_cr
= CRED();
1952 ddsa
.ddsa_proc
= curproc
;
1955 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1956 dsl_dataset_snapshot_sync
, &ddsa
,
1957 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1960 if (suspended
!= NULL
) {
1961 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1962 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1963 zil_resume((void *)(uintptr_t)
1964 fnvpair_value_uint64(pair
));
1966 fnvlist_free(suspended
);
1970 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1971 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1972 zvol_create_minor(nvpair_name(pair
));
1979 typedef struct dsl_dataset_snapshot_tmp_arg
{
1980 const char *ddsta_fsname
;
1981 const char *ddsta_snapname
;
1982 minor_t ddsta_cleanup_minor
;
1983 const char *ddsta_htag
;
1984 } dsl_dataset_snapshot_tmp_arg_t
;
1987 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1989 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1990 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1994 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1998 /* NULL cred means no limit check for tmp snapshot */
1999 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
2000 tx
, B_FALSE
, 0, NULL
, NULL
);
2002 dsl_dataset_rele(ds
, FTAG
);
2006 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
2007 dsl_dataset_rele(ds
, FTAG
);
2008 return (SET_ERROR(ENOTSUP
));
2010 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
2013 dsl_dataset_rele(ds
, FTAG
);
2017 dsl_dataset_rele(ds
, FTAG
);
2022 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
2024 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
2025 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2026 dsl_dataset_t
*ds
= NULL
;
2028 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
2030 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
2031 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
2032 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
2033 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
2035 dsl_dataset_rele(ds
, FTAG
);
2039 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
2040 minor_t cleanup_minor
, const char *htag
)
2042 dsl_dataset_snapshot_tmp_arg_t ddsta
;
2045 boolean_t needsuspend
;
2048 ddsta
.ddsta_fsname
= fsname
;
2049 ddsta
.ddsta_snapname
= snapname
;
2050 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
2051 ddsta
.ddsta_htag
= htag
;
2053 error
= spa_open(fsname
, &spa
, FTAG
);
2056 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
2057 spa_close(spa
, FTAG
);
2060 error
= zil_suspend(fsname
, &cookie
);
2065 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
2066 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
2073 /* Nonblocking dataset sync. Assumes dataset:objset is always 1:1 */
2075 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*rio
, dmu_tx_t
*tx
)
2077 ASSERT(dmu_tx_is_syncing(tx
));
2078 ASSERT(ds
->ds_objset
!= NULL
);
2079 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
2082 * in case we had to change ds_fsid_guid when we opened it,
2085 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
2086 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
2088 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
2089 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
2090 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
2091 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
2092 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
2093 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
2094 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
2095 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
2096 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
2097 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
2098 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
2099 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
2100 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
2103 dmu_objset_sync(ds
->ds_objset
, rio
, tx
);
2107 * Check if the percentage of blocks shared between the clone and the
2108 * snapshot (as opposed to those that are clone only) is below a certain
2112 dsl_livelist_should_disable(dsl_dataset_t
*ds
)
2114 uint64_t used
, referenced
;
2117 used
= dsl_dir_get_usedds(ds
->ds_dir
);
2118 referenced
= dsl_get_referenced(ds
);
2119 if (referenced
== 0)
2121 percent_shared
= (100 * (referenced
- used
)) / referenced
;
2122 if (percent_shared
<= zfs_livelist_min_percent_shared
)
2128 * Check if it is possible to combine two livelist entries into one.
2129 * This is the case if the combined number of 'live' blkptrs (ALLOCs that
2130 * don't have a matching FREE) is under the maximum sublist size.
2131 * We check this by subtracting twice the total number of frees from the total
2132 * number of blkptrs. FREEs are counted twice because each FREE blkptr
2133 * will cancel out an ALLOC blkptr when the livelist is processed.
2136 dsl_livelist_should_condense(dsl_deadlist_entry_t
*first
,
2137 dsl_deadlist_entry_t
*next
)
2139 uint64_t total_free
= first
->dle_bpobj
.bpo_phys
->bpo_num_freed
+
2140 next
->dle_bpobj
.bpo_phys
->bpo_num_freed
;
2141 uint64_t total_entries
= first
->dle_bpobj
.bpo_phys
->bpo_num_blkptrs
+
2142 next
->dle_bpobj
.bpo_phys
->bpo_num_blkptrs
;
2143 if ((total_entries
- (2 * total_free
)) < zfs_livelist_max_entries
)
2148 typedef struct try_condense_arg
{
2151 } try_condense_arg_t
;
2154 * Iterate over the livelist entries, searching for a pair to condense.
2155 * A nonzero return value means stop, 0 means keep looking.
2158 dsl_livelist_try_condense(void *arg
, dsl_deadlist_entry_t
*first
)
2160 try_condense_arg_t
*tca
= arg
;
2161 spa_t
*spa
= tca
->spa
;
2162 dsl_dataset_t
*ds
= tca
->ds
;
2163 dsl_deadlist_t
*ll
= &ds
->ds_dir
->dd_livelist
;
2164 dsl_deadlist_entry_t
*next
;
2166 /* The condense thread has not yet been created at import */
2167 if (spa
->spa_livelist_condense_zthr
== NULL
)
2170 /* A condense is already in progress */
2171 if (spa
->spa_to_condense
.ds
!= NULL
)
2174 next
= AVL_NEXT(&ll
->dl_tree
, &first
->dle_node
);
2175 /* The livelist has only one entry - don't condense it */
2179 /* Next is the newest entry - don't condense it */
2180 if (AVL_NEXT(&ll
->dl_tree
, &next
->dle_node
) == NULL
)
2183 /* This pair is not ready to condense but keep looking */
2184 if (!dsl_livelist_should_condense(first
, next
))
2188 * Add a ref to prevent the dataset from being evicted while
2189 * the condense zthr or synctask are running. Ref will be
2190 * released at the end of the condense synctask
2192 dmu_buf_add_ref(ds
->ds_dbuf
, spa
);
2194 spa
->spa_to_condense
.ds
= ds
;
2195 spa
->spa_to_condense
.first
= first
;
2196 spa
->spa_to_condense
.next
= next
;
2197 spa
->spa_to_condense
.syncing
= B_FALSE
;
2198 spa
->spa_to_condense
.cancelled
= B_FALSE
;
2200 zthr_wakeup(spa
->spa_livelist_condense_zthr
);
2205 dsl_flush_pending_livelist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
2207 dsl_dir_t
*dd
= ds
->ds_dir
;
2208 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
2209 dsl_deadlist_entry_t
*last
= dsl_deadlist_last(&dd
->dd_livelist
);
2211 /* Check if we need to add a new sub-livelist */
2213 /* The livelist is empty */
2214 dsl_deadlist_add_key(&dd
->dd_livelist
,
2215 tx
->tx_txg
- 1, tx
);
2216 } else if (spa_sync_pass(spa
) == 1) {
2218 * Check if the newest entry is full. If it is, make a new one.
2219 * We only do this once per sync because we could overfill a
2220 * sublist in one sync pass and don't want to add another entry
2221 * for a txg that is already represented. This ensures that
2222 * blkptrs born in the same txg are stored in the same sublist.
2224 bpobj_t bpobj
= last
->dle_bpobj
;
2225 uint64_t all
= bpobj
.bpo_phys
->bpo_num_blkptrs
;
2226 uint64_t free
= bpobj
.bpo_phys
->bpo_num_freed
;
2227 uint64_t alloc
= all
- free
;
2228 if (alloc
> zfs_livelist_max_entries
) {
2229 dsl_deadlist_add_key(&dd
->dd_livelist
,
2230 tx
->tx_txg
- 1, tx
);
2234 /* Insert each entry into the on-disk livelist */
2235 bplist_iterate(&dd
->dd_pending_allocs
,
2236 dsl_deadlist_insert_alloc_cb
, &dd
->dd_livelist
, tx
);
2237 bplist_iterate(&dd
->dd_pending_frees
,
2238 dsl_deadlist_insert_free_cb
, &dd
->dd_livelist
, tx
);
2240 /* Attempt to condense every pair of adjacent entries */
2241 try_condense_arg_t arg
= {
2245 dsl_deadlist_iterate(&dd
->dd_livelist
, dsl_livelist_try_condense
,
2250 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
2252 objset_t
*os
= ds
->ds_objset
;
2254 bplist_iterate(&ds
->ds_pending_deadlist
,
2255 dsl_deadlist_insert_alloc_cb
, &ds
->ds_deadlist
, tx
);
2257 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
)) {
2258 dsl_flush_pending_livelist(ds
, tx
);
2259 if (dsl_livelist_should_disable(ds
)) {
2260 dsl_dir_remove_livelist(ds
->ds_dir
, tx
, B_TRUE
);
2264 dsl_bookmark_sync_done(ds
, tx
);
2266 multilist_destroy(&os
->os_synced_dnodes
);
2268 if (os
->os_encrypted
)
2269 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_FALSE
;
2271 ASSERT0(os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]);
2273 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
2274 if (zfeature_active(f
,
2275 ds
->ds_feature_activation
[f
])) {
2276 if (zfeature_active(f
, ds
->ds_feature
[f
]))
2278 dsl_dataset_activate_feature(ds
->ds_object
, f
,
2279 ds
->ds_feature_activation
[f
], tx
);
2280 ds
->ds_feature
[f
] = ds
->ds_feature_activation
[f
];
2284 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
2288 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
2291 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
2295 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
2298 * There may be missing entries in ds_next_clones_obj
2299 * due to a bug in a previous version of the code.
2300 * Only trust it if it has the right number of entries.
2302 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
2303 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
2306 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
2307 return (SET_ERROR(ENOENT
));
2309 for (zap_cursor_init(&zc
, mos
,
2310 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2311 zap_cursor_retrieve(&zc
, &za
) == 0;
2312 zap_cursor_advance(&zc
)) {
2313 dsl_dataset_t
*clone
;
2314 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2315 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
2316 za
.za_first_integer
, FTAG
, &clone
));
2317 dsl_dir_name(clone
->ds_dir
, buf
);
2318 fnvlist_add_boolean(val
, buf
);
2319 dsl_dataset_rele(clone
, FTAG
);
2321 zap_cursor_fini(&zc
);
2326 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2328 nvlist_t
*propval
= fnvlist_alloc();
2329 nvlist_t
*val
= fnvlist_alloc();
2331 if (get_clones_stat_impl(ds
, val
) == 0) {
2332 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
2333 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
2338 nvlist_free(propval
);
2342 get_receive_resume_token_impl(dsl_dataset_t
*ds
)
2344 if (!dsl_dataset_has_resume_receive_state(ds
))
2347 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2350 uint8_t *compressed
;
2352 nvlist_t
*token_nv
= fnvlist_alloc();
2353 size_t packed_size
, compressed_size
;
2355 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2356 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
2357 fnvlist_add_uint64(token_nv
, "fromguid", val
);
2359 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2360 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
2361 fnvlist_add_uint64(token_nv
, "object", val
);
2363 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2364 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
2365 fnvlist_add_uint64(token_nv
, "offset", val
);
2367 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2368 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
2369 fnvlist_add_uint64(token_nv
, "bytes", val
);
2371 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2372 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
2373 fnvlist_add_uint64(token_nv
, "toguid", val
);
2375 char buf
[MAXNAMELEN
];
2376 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2377 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
2378 fnvlist_add_string(token_nv
, "toname", buf
);
2380 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2381 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
2382 fnvlist_add_boolean(token_nv
, "largeblockok");
2384 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2385 DS_FIELD_RESUME_EMBEDOK
) == 0) {
2386 fnvlist_add_boolean(token_nv
, "embedok");
2388 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2389 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
2390 fnvlist_add_boolean(token_nv
, "compressok");
2392 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2393 DS_FIELD_RESUME_RAWOK
) == 0) {
2394 fnvlist_add_boolean(token_nv
, "rawok");
2396 if (dsl_dataset_feature_is_active(ds
,
2397 SPA_FEATURE_REDACTED_DATASETS
)) {
2398 uint64_t num_redact_snaps
= 0;
2399 uint64_t *redact_snaps
= NULL
;
2400 VERIFY3B(dsl_dataset_get_uint64_array_feature(ds
,
2401 SPA_FEATURE_REDACTED_DATASETS
, &num_redact_snaps
,
2402 &redact_snaps
), ==, B_TRUE
);
2403 fnvlist_add_uint64_array(token_nv
, "redact_snaps",
2404 redact_snaps
, num_redact_snaps
);
2406 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2407 DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS
) == 0) {
2408 uint64_t num_redact_snaps
= 0, int_size
= 0;
2409 uint64_t *redact_snaps
= NULL
;
2410 VERIFY0(zap_length(dp
->dp_meta_objset
, ds
->ds_object
,
2411 DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS
, &int_size
,
2412 &num_redact_snaps
));
2413 ASSERT3U(int_size
, ==, sizeof (uint64_t));
2415 redact_snaps
= kmem_alloc(int_size
* num_redact_snaps
,
2417 VERIFY0(zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2418 DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS
, int_size
,
2419 num_redact_snaps
, redact_snaps
));
2420 fnvlist_add_uint64_array(token_nv
, "book_redact_snaps",
2421 redact_snaps
, num_redact_snaps
);
2422 kmem_free(redact_snaps
, int_size
* num_redact_snaps
);
2424 packed
= fnvlist_pack(token_nv
, &packed_size
);
2425 fnvlist_free(token_nv
);
2426 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
2428 compressed_size
= gzip_compress(packed
, compressed
,
2429 packed_size
, packed_size
, 6);
2432 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
2434 size_t alloc_size
= compressed_size
* 2 + 1;
2435 str
= kmem_alloc(alloc_size
, KM_SLEEP
);
2436 for (int i
= 0; i
< compressed_size
; i
++) {
2437 size_t offset
= i
* 2;
2438 (void) snprintf(str
+ offset
, alloc_size
- offset
,
2439 "%02x", compressed
[i
]);
2441 str
[compressed_size
* 2] = '\0';
2442 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
2443 ZFS_SEND_RESUME_TOKEN_VERSION
,
2444 (longlong_t
)cksum
.zc_word
[0],
2445 (longlong_t
)packed_size
, str
);
2446 kmem_free(packed
, packed_size
);
2447 kmem_free(str
, alloc_size
);
2448 kmem_free(compressed
, packed_size
);
2453 * Returns a string that represents the receive resume state token. It should
2454 * be freed with strfree(). NULL is returned if no resume state is present.
2457 get_receive_resume_token(dsl_dataset_t
*ds
)
2460 * A failed "newfs" (e.g. full) resumable receive leaves
2461 * the stats set on this dataset. Check here for the prop.
2463 char *token
= get_receive_resume_token_impl(ds
);
2467 * A failed incremental resumable receive leaves the
2468 * stats set on our child named "%recv". Check the child
2471 /* 6 extra bytes for /%recv */
2472 char name
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2473 dsl_dataset_t
*recv_ds
;
2474 dsl_dataset_name(ds
, name
);
2475 if (strlcat(name
, "/", sizeof (name
)) < sizeof (name
) &&
2476 strlcat(name
, recv_clone_name
, sizeof (name
)) < sizeof (name
) &&
2477 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, name
, FTAG
, &recv_ds
) == 0) {
2478 token
= get_receive_resume_token_impl(recv_ds
);
2479 dsl_dataset_rele(recv_ds
, FTAG
);
2485 dsl_get_refratio(dsl_dataset_t
*ds
)
2487 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
2488 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
2489 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
2494 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
2496 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
2500 dsl_get_compressratio(dsl_dataset_t
*ds
)
2502 if (ds
->ds_is_snapshot
) {
2503 return (dsl_get_refratio(ds
));
2505 dsl_dir_t
*dd
= ds
->ds_dir
;
2506 mutex_enter(&dd
->dd_lock
);
2507 uint64_t val
= dsl_dir_get_compressratio(dd
);
2508 mutex_exit(&dd
->dd_lock
);
2514 dsl_get_used(dsl_dataset_t
*ds
)
2516 if (ds
->ds_is_snapshot
) {
2517 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2519 dsl_dir_t
*dd
= ds
->ds_dir
;
2520 mutex_enter(&dd
->dd_lock
);
2521 uint64_t val
= dsl_dir_get_used(dd
);
2522 mutex_exit(&dd
->dd_lock
);
2528 dsl_get_creation(dsl_dataset_t
*ds
)
2530 return (dsl_dataset_phys(ds
)->ds_creation_time
);
2534 dsl_get_creationtxg(dsl_dataset_t
*ds
)
2536 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
2540 dsl_get_refquota(dsl_dataset_t
*ds
)
2542 return (ds
->ds_quota
);
2546 dsl_get_refreservation(dsl_dataset_t
*ds
)
2548 return (ds
->ds_reserved
);
2552 dsl_get_guid(dsl_dataset_t
*ds
)
2554 return (dsl_dataset_phys(ds
)->ds_guid
);
2558 dsl_get_unique(dsl_dataset_t
*ds
)
2560 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2564 dsl_get_objsetid(dsl_dataset_t
*ds
)
2566 return (ds
->ds_object
);
2570 dsl_get_userrefs(dsl_dataset_t
*ds
)
2572 return (ds
->ds_userrefs
);
2576 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2578 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2582 dsl_get_referenced(dsl_dataset_t
*ds
)
2584 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2588 dsl_get_numclones(dsl_dataset_t
*ds
)
2590 ASSERT(ds
->ds_is_snapshot
);
2591 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2595 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2597 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2602 dsl_get_redacted(dsl_dataset_t
*ds
)
2604 return (dsl_dataset_feature_is_active(ds
,
2605 SPA_FEATURE_REDACTED_DATASETS
));
2609 dsl_get_available(dsl_dataset_t
*ds
)
2611 uint64_t refdbytes
= dsl_get_referenced(ds
);
2612 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2614 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2616 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2618 if (ds
->ds_quota
!= 0) {
2620 * Adjust available bytes according to refquota
2622 if (refdbytes
< ds
->ds_quota
) {
2623 availbytes
= MIN(availbytes
,
2624 ds
->ds_quota
- refdbytes
);
2629 return (availbytes
);
2633 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2635 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2636 dsl_dataset_t
*prev
;
2637 int err
= dsl_dataset_hold_obj(dp
,
2638 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2640 uint64_t comp
, uncomp
;
2641 err
= dsl_dataset_space_written(prev
, ds
, written
,
2643 dsl_dataset_rele(prev
, FTAG
);
2649 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2652 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2654 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2655 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2656 dsl_dataset_name(ds
->ds_prev
, snap
);
2659 return (SET_ERROR(ENOENT
));
2664 dsl_get_redact_snaps(dsl_dataset_t
*ds
, nvlist_t
*propval
)
2668 if (dsl_dataset_get_uint64_array_feature(ds
,
2669 SPA_FEATURE_REDACTED_DATASETS
, &nsnaps
, &snaps
)) {
2670 fnvlist_add_uint64_array(propval
, ZPROP_VALUE
, snaps
,
2676 * Returns the mountpoint property and source for the given dataset in the value
2677 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2678 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2679 * Returns 0 on success and an error on failure.
2682 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2686 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2688 /* Retrieve the mountpoint value stored in the zap object */
2689 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2690 ZAP_MAXVALUELEN
, value
, source
);
2696 * Process the dsname and source to find the full mountpoint string.
2697 * Can be skipped for 'legacy' or 'none'.
2699 if (value
[0] == '/') {
2700 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2702 const char *relpath
;
2705 * If we inherit the mountpoint, even from a dataset
2706 * with a received value, the source will be the path of
2707 * the dataset we inherit from. If source is
2708 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2711 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2714 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2715 relpath
= dsname
+ strlen(source
);
2716 if (relpath
[0] == '/')
2720 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2723 * Special case an alternate root of '/'. This will
2724 * avoid having multiple leading slashes in the
2727 if (strcmp(root
, "/") == 0)
2731 * If the mountpoint is '/' then skip over this
2732 * if we are obtaining either an alternate root or
2733 * an inherited mountpoint.
2736 if (value
[1] == '\0' && (root
[0] != '\0' ||
2737 relpath
[0] != '\0'))
2740 mnt
= kmem_strdup(mnt
);
2742 if (relpath
[0] == '\0') {
2743 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2746 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2747 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2750 kmem_free(buf
, ZAP_MAXVALUELEN
);
2758 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2760 dsl_pool_t
*dp __maybe_unused
= ds
->ds_dir
->dd_pool
;
2762 ASSERT(dsl_pool_config_held(dp
));
2764 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2765 dsl_get_refratio(ds
));
2766 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2767 dsl_get_logicalreferenced(ds
));
2768 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2769 dsl_get_compressratio(ds
));
2770 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2773 if (ds
->ds_is_snapshot
) {
2774 get_clones_stat(ds
, nv
);
2776 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2777 if (dsl_get_prev_snap(ds
, buf
) == 0)
2778 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2780 dsl_dir_stats(ds
->ds_dir
, nv
);
2783 nvlist_t
*propval
= fnvlist_alloc();
2784 dsl_get_redact_snaps(ds
, propval
);
2785 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS
),
2787 nvlist_free(propval
);
2789 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2790 dsl_get_available(ds
));
2791 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2792 dsl_get_referenced(ds
));
2793 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2794 dsl_get_creation(ds
));
2795 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2796 dsl_get_creationtxg(ds
));
2797 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2798 dsl_get_refquota(ds
));
2799 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2800 dsl_get_refreservation(ds
));
2801 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2803 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2804 dsl_get_unique(ds
));
2805 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2806 dsl_get_objsetid(ds
));
2807 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2808 dsl_get_userrefs(ds
));
2809 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2810 dsl_get_defer_destroy(ds
));
2811 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_SNAPSHOTS_CHANGED
,
2812 dsl_dir_snap_cmtime(ds
->ds_dir
).tv_sec
);
2813 dsl_dataset_crypt_stats(ds
, nv
);
2815 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2817 if (dsl_get_written(ds
, &written
) == 0) {
2818 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2823 if (!dsl_dataset_is_snapshot(ds
)) {
2824 char *token
= get_receive_resume_token(ds
);
2825 if (token
!= NULL
) {
2826 dsl_prop_nvlist_add_string(nv
,
2827 ZFS_PROP_RECEIVE_RESUME_TOKEN
, token
);
2828 kmem_strfree(token
);
2834 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2836 dsl_pool_t
*dp __maybe_unused
= ds
->ds_dir
->dd_pool
;
2837 ASSERT(dsl_pool_config_held(dp
));
2839 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2840 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2841 stat
->dds_guid
= dsl_get_guid(ds
);
2842 stat
->dds_redacted
= dsl_get_redacted(ds
);
2843 stat
->dds_origin
[0] = '\0';
2844 if (ds
->ds_is_snapshot
) {
2845 stat
->dds_is_snapshot
= B_TRUE
;
2846 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2848 stat
->dds_is_snapshot
= B_FALSE
;
2849 stat
->dds_num_clones
= 0;
2851 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2852 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2858 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2860 return (ds
->ds_fsid_guid
);
2864 dsl_dataset_space(dsl_dataset_t
*ds
,
2865 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2866 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2868 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2869 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2870 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2872 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2873 if (ds
->ds_quota
!= 0) {
2875 * Adjust available bytes according to refquota
2877 if (*refdbytesp
< ds
->ds_quota
)
2878 *availbytesp
= MIN(*availbytesp
,
2879 ds
->ds_quota
- *refdbytesp
);
2883 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2884 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2885 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2886 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2890 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2892 dsl_pool_t
*dp __maybe_unused
= ds
->ds_dir
->dd_pool
;
2895 ASSERT(dsl_pool_config_held(dp
));
2898 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2899 birth
= BP_GET_LOGICAL_BIRTH(dsl_dataset_get_blkptr(ds
));
2900 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2901 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2902 objset_t
*os
, *os_snap
;
2904 * It may be that only the ZIL differs, because it was
2905 * reset in the head. Don't count that as being
2908 if (dmu_objset_from_ds(ds
, &os
) != 0)
2910 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2912 return (memcmp(&os
->os_phys
->os_meta_dnode
,
2913 &os_snap
->os_phys
->os_meta_dnode
,
2914 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2920 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2921 dsl_dataset_t
*hds
, void *arg
)
2924 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2928 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2930 /* ignore nonexistent snapshots */
2931 return (error
== ENOENT
? 0 : error
);
2934 /* new name should not exist */
2935 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2937 error
= SET_ERROR(EEXIST
);
2938 else if (error
== ENOENT
)
2941 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2942 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2943 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2944 error
= SET_ERROR(ENAMETOOLONG
);
2950 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2952 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2953 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2957 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2961 if (ddrsa
->ddrsa_recursive
) {
2962 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2963 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2966 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2968 dsl_dataset_rele(hds
, FTAG
);
2973 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2974 dsl_dataset_t
*hds
, void *arg
)
2976 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2979 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2982 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2983 ASSERT(error
== 0 || error
== ENOENT
);
2984 if (error
== ENOENT
) {
2985 /* ignore nonexistent snapshots */
2989 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2991 /* log before we change the name */
2992 spa_history_log_internal_ds(ds
, "rename", tx
,
2993 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2995 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2997 mutex_enter(&ds
->ds_lock
);
2998 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2999 sizeof (ds
->ds_snapname
));
3000 mutex_exit(&ds
->ds_lock
);
3001 VERIFY0(zap_add(dp
->dp_meta_objset
,
3002 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
3003 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
3004 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
3005 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
3007 dsl_dataset_rele(ds
, FTAG
);
3012 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
3014 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
3015 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3016 dsl_dataset_t
*hds
= NULL
;
3018 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
3019 ddrsa
->ddrsa_tx
= tx
;
3020 if (ddrsa
->ddrsa_recursive
) {
3021 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
3022 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
3025 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
3027 dsl_dataset_rele(hds
, FTAG
);
3031 dsl_dataset_rename_snapshot(const char *fsname
,
3032 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
3034 dsl_dataset_rename_snapshot_arg_t ddrsa
;
3036 ddrsa
.ddrsa_fsname
= fsname
;
3037 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
3038 ddrsa
.ddrsa_newsnapname
= newsnapname
;
3039 ddrsa
.ddrsa_recursive
= recursive
;
3041 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
3042 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
3043 1, ZFS_SPACE_CHECK_RESERVED
));
3047 * If we're doing an ownership handoff, we need to make sure that there is
3048 * only one long hold on the dataset. We're not allowed to change anything here
3049 * so we don't permanently release the long hold or regular hold here. We want
3050 * to do this only when syncing to avoid the dataset unexpectedly going away
3051 * when we release the long hold.
3054 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
3056 boolean_t held
= B_FALSE
;
3058 if (!dmu_tx_is_syncing(tx
))
3061 dsl_dir_t
*dd
= ds
->ds_dir
;
3062 mutex_enter(&dd
->dd_activity_lock
);
3063 uint64_t holds
= zfs_refcount_count(&ds
->ds_longholds
) -
3064 (owner
!= NULL
? 1 : 0);
3066 * The value of dd_activity_waiters can chance as soon as we drop the
3067 * lock, but we're fine with that; new waiters coming in or old
3068 * waiters leaving doesn't cause problems, since we're going to cancel
3069 * waiters later anyway. The goal of this check is to verify that no
3070 * non-waiters have long-holds, and all new long-holds will be
3071 * prevented because we're holding the pool config as writer.
3073 if (holds
!= dd
->dd_activity_waiters
)
3075 mutex_exit(&dd
->dd_activity_lock
);
3078 return (SET_ERROR(EBUSY
));
3084 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
3086 dsl_dataset_rollback_arg_t
*ddra
= arg
;
3087 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3089 int64_t unused_refres_delta
;
3092 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
3096 /* must not be a snapshot */
3097 if (ds
->ds_is_snapshot
) {
3098 dsl_dataset_rele(ds
, FTAG
);
3099 return (SET_ERROR(EINVAL
));
3102 /* must have a most recent snapshot */
3103 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
3104 dsl_dataset_rele(ds
, FTAG
);
3105 return (SET_ERROR(ESRCH
));
3109 * No rollback to a snapshot created in the current txg, because
3110 * the rollback may dirty the dataset and create blocks that are
3111 * not reachable from the rootbp while having a birth txg that
3112 * falls into the snapshot's range.
3114 if (dmu_tx_is_syncing(tx
) &&
3115 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
3116 dsl_dataset_rele(ds
, FTAG
);
3117 return (SET_ERROR(EAGAIN
));
3121 * If the expected target snapshot is specified, then check that
3122 * the latest snapshot is it.
3124 if (ddra
->ddra_tosnap
!= NULL
) {
3125 dsl_dataset_t
*snapds
;
3127 /* Check if the target snapshot exists at all. */
3128 error
= dsl_dataset_hold(dp
, ddra
->ddra_tosnap
, FTAG
, &snapds
);
3131 * ESRCH is used to signal that the target snapshot does
3132 * not exist, while ENOENT is used to report that
3133 * the rolled back dataset does not exist.
3134 * ESRCH is also used to cover other cases where the
3135 * target snapshot is not related to the dataset being
3136 * rolled back such as being in a different pool.
3138 if (error
== ENOENT
|| error
== EXDEV
)
3139 error
= SET_ERROR(ESRCH
);
3140 dsl_dataset_rele(ds
, FTAG
);
3143 ASSERT(snapds
->ds_is_snapshot
);
3145 /* Check if the snapshot is the latest snapshot indeed. */
3146 if (snapds
!= ds
->ds_prev
) {
3148 * Distinguish between the case where the only problem
3149 * is intervening snapshots (EEXIST) vs the snapshot
3150 * not being a valid target for rollback (ESRCH).
3152 if (snapds
->ds_dir
== ds
->ds_dir
||
3153 (dsl_dir_is_clone(ds
->ds_dir
) &&
3154 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
==
3155 snapds
->ds_object
)) {
3156 error
= SET_ERROR(EEXIST
);
3158 error
= SET_ERROR(ESRCH
);
3160 dsl_dataset_rele(snapds
, FTAG
);
3161 dsl_dataset_rele(ds
, FTAG
);
3164 dsl_dataset_rele(snapds
, FTAG
);
3167 /* must not have any bookmarks after the most recent snapshot */
3168 if (dsl_bookmark_latest_txg(ds
) >
3169 dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
3170 dsl_dataset_rele(ds
, FTAG
);
3171 return (SET_ERROR(EEXIST
));
3174 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
3176 dsl_dataset_rele(ds
, FTAG
);
3181 * Check if the snap we are rolling back to uses more than
3184 if (ds
->ds_quota
!= 0 &&
3185 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
3186 dsl_dataset_rele(ds
, FTAG
);
3187 return (SET_ERROR(EDQUOT
));
3191 * When we do the clone swap, we will temporarily use more space
3192 * due to the refreservation (the head will no longer have any
3193 * unique space, so the entire amount of the refreservation will need
3194 * to be free). We will immediately destroy the clone, freeing
3195 * this space, but the freeing happens over many txg's.
3197 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
3198 dsl_dataset_phys(ds
)->ds_unique_bytes
);
3200 if (unused_refres_delta
> 0 &&
3201 unused_refres_delta
>
3202 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
3203 dsl_dataset_rele(ds
, FTAG
);
3204 return (SET_ERROR(ENOSPC
));
3207 dsl_dataset_rele(ds
, FTAG
);
3212 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
3214 dsl_dataset_rollback_arg_t
*ddra
= arg
;
3215 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3216 dsl_dataset_t
*ds
, *clone
;
3218 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
3220 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
3222 dsl_dataset_name(ds
->ds_prev
, namebuf
);
3223 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
3225 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
3226 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
3228 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
3230 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
3231 dsl_dataset_zero_zil(ds
, tx
);
3233 dsl_destroy_head_sync_impl(clone
, tx
);
3235 dsl_dataset_rele(clone
, FTAG
);
3236 dsl_dataset_rele(ds
, FTAG
);
3240 * Rolls back the given filesystem or volume to the most recent snapshot.
3241 * The name of the most recent snapshot will be returned under key "target"
3242 * in the result nvlist.
3245 * - The existing dataset MUST be owned by the specified owner at entry
3246 * - Upon return, dataset will still be held by the same owner, whether we
3249 * This mode is required any time the existing filesystem is mounted. See
3250 * notes above zfs_suspend_fs() for further details.
3253 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
3256 dsl_dataset_rollback_arg_t ddra
;
3258 ddra
.ddra_fsname
= fsname
;
3259 ddra
.ddra_tosnap
= tosnap
;
3260 ddra
.ddra_owner
= owner
;
3261 ddra
.ddra_result
= result
;
3263 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
3264 dsl_dataset_rollback_sync
, &ddra
,
3265 1, ZFS_SPACE_CHECK_RESERVED
));
3268 struct promotenode
{
3273 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
3274 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
3276 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, const void *tag
);
3279 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
3281 dsl_dataset_promote_arg_t
*ddpa
= arg
;
3282 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3284 struct promotenode
*snap
;
3288 size_t max_snap_len
;
3289 boolean_t conflicting_snaps
;
3291 err
= promote_hold(ddpa
, dp
, FTAG
);
3295 hds
= ddpa
->ddpa_clone
;
3296 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
3298 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
3299 promote_rele(ddpa
, FTAG
);
3300 return (SET_ERROR(EXDEV
));
3303 snap
= list_head(&ddpa
->shared_snaps
);
3305 err
= SET_ERROR(ENOENT
);
3308 dsl_dataset_t
*const origin_ds
= snap
->ds
;
3311 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
3312 * When doing a promote we must make sure the encryption root for
3313 * both the target and the target's origin does not change to avoid
3314 * needing to rewrap encryption keys
3316 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
3321 * Compute and check the amount of space to transfer. Since this is
3322 * so expensive, don't do the preliminary check.
3324 if (!dmu_tx_is_syncing(tx
)) {
3325 promote_rele(ddpa
, FTAG
);
3329 /* compute origin's new unique space */
3330 snap
= list_tail(&ddpa
->clone_snaps
);
3331 ASSERT(snap
!= NULL
);
3332 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3333 origin_ds
->ds_object
);
3334 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3335 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
3336 &ddpa
->unique
, &unused
, &unused
);
3339 * Walk the snapshots that we are moving
3341 * Compute space to transfer. Consider the incremental changes
3342 * to used by each snapshot:
3343 * (my used) = (prev's used) + (blocks born) - (blocks killed)
3344 * So each snapshot gave birth to:
3345 * (blocks born) = (my used) - (prev's used) + (blocks killed)
3346 * So a sequence would look like:
3347 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
3348 * Which simplifies to:
3349 * uN + kN + kN-1 + ... + k1 + k0
3350 * Note however, if we stop before we reach the ORIGIN we get:
3351 * uN + kN + kN-1 + ... + kM - uM-1
3353 conflicting_snaps
= B_FALSE
;
3355 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
3356 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
3357 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
3358 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3359 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3360 uint64_t val
, dlused
, dlcomp
, dluncomp
;
3361 dsl_dataset_t
*ds
= snap
->ds
;
3366 * If there are long holds, we won't be able to evict
3369 if (dsl_dataset_long_held(ds
)) {
3370 err
= SET_ERROR(EBUSY
);
3374 /* Check that the snapshot name does not conflict */
3375 VERIFY0(dsl_dataset_get_snapname(ds
));
3376 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
3377 err
= SET_ERROR(ENAMETOOLONG
);
3380 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
3382 fnvlist_add_boolean(ddpa
->err_ds
,
3383 snap
->ds
->ds_snapname
);
3384 conflicting_snaps
= B_TRUE
;
3385 } else if (err
!= ENOENT
) {
3389 /* The very first snapshot does not have a deadlist */
3390 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
3393 dsl_deadlist_space(&ds
->ds_deadlist
,
3394 &dlused
, &dlcomp
, &dluncomp
);
3395 ddpa
->used
+= dlused
;
3396 ddpa
->comp
+= dlcomp
;
3397 ddpa
->uncomp
+= dluncomp
;
3401 * Check that bookmarks that are being transferred don't have
3404 for (dsl_bookmark_node_t
*dbn
= avl_first(&origin_ds
->ds_bookmarks
);
3405 dbn
!= NULL
&& dbn
->dbn_phys
.zbm_creation_txg
<=
3406 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3407 dbn
= AVL_NEXT(&origin_ds
->ds_bookmarks
, dbn
)) {
3408 if (strlen(dbn
->dbn_name
) >= max_snap_len
) {
3409 err
= SET_ERROR(ENAMETOOLONG
);
3412 zfs_bookmark_phys_t bm
;
3413 err
= dsl_bookmark_lookup_impl(ddpa
->ddpa_clone
,
3414 dbn
->dbn_name
, &bm
);
3417 fnvlist_add_boolean(ddpa
->err_ds
, dbn
->dbn_name
);
3418 conflicting_snaps
= B_TRUE
;
3419 } else if (err
== ESRCH
) {
3428 * In order to return the full list of conflicting snapshots, we check
3429 * whether there was a conflict after traversing all of them.
3431 if (conflicting_snaps
) {
3432 err
= SET_ERROR(EEXIST
);
3437 * If we are a clone of a clone then we never reached ORIGIN,
3438 * so we need to subtract out the clone origin's used space.
3440 if (ddpa
->origin_origin
) {
3442 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
3444 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
3446 dsl_dataset_phys(ddpa
->origin_origin
)->
3447 ds_uncompressed_bytes
;
3450 /* Check that there is enough space and limit headroom here */
3451 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
3452 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
, ddpa
->proc
);
3457 * Compute the amounts of space that will be used by snapshots
3458 * after the promotion (for both origin and clone). For each,
3459 * it is the amount of space that will be on all of their
3460 * deadlists (that was not born before their new origin).
3462 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
3466 * Note, typically this will not be a clone of a clone,
3467 * so dd_origin_txg will be < TXG_INITIAL, so
3468 * these snaplist_space() -> dsl_deadlist_space_range()
3469 * calls will be fast because they do not have to
3470 * iterate over all bps.
3472 snap
= list_head(&ddpa
->origin_snaps
);
3474 err
= SET_ERROR(ENOENT
);
3477 err
= snaplist_space(&ddpa
->shared_snaps
,
3478 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
3482 err
= snaplist_space(&ddpa
->clone_snaps
,
3483 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
3486 ddpa
->cloneusedsnap
+= space
;
3488 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
3489 DD_FLAG_USED_BREAKDOWN
) {
3490 err
= snaplist_space(&ddpa
->origin_snaps
,
3491 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
3492 &ddpa
->originusedsnap
);
3498 promote_rele(ddpa
, FTAG
);
3503 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
3505 dsl_dataset_promote_arg_t
*ddpa
= arg
;
3506 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3508 struct promotenode
*snap
;
3509 dsl_dataset_t
*origin_ds
;
3510 dsl_dataset_t
*origin_head
;
3512 dsl_dir_t
*odd
= NULL
;
3513 uint64_t oldnext_obj
;
3516 ASSERT(nvlist_empty(ddpa
->err_ds
));
3518 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
3519 hds
= ddpa
->ddpa_clone
;
3521 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
3523 snap
= list_head(&ddpa
->shared_snaps
);
3524 origin_ds
= snap
->ds
;
3527 snap
= list_head(&ddpa
->origin_snaps
);
3528 origin_head
= snap
->ds
;
3531 * We need to explicitly open odd, since origin_ds's dd will be
3534 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
3537 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
3539 /* change origin's next snap */
3540 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
3541 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
3542 snap
= list_tail(&ddpa
->clone_snaps
);
3543 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3544 origin_ds
->ds_object
);
3545 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
3547 /* change the origin's next clone */
3548 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
3549 dsl_dataset_remove_from_next_clones(origin_ds
,
3550 snap
->ds
->ds_object
, tx
);
3551 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3552 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
3557 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
3558 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
3559 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
3560 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
3561 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
3562 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
3563 origin_head
->ds_dir
->dd_origin_txg
=
3564 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3566 /* change dd_clone entries */
3567 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3568 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3569 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
3570 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3571 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3572 hds
->ds_object
, tx
));
3574 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3575 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3576 origin_head
->ds_object
, tx
));
3577 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
3578 dsl_dir_phys(dd
)->dd_clones
=
3579 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
3580 DMU_OT_NONE
, 0, tx
);
3582 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3583 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
3587 * Move bookmarks to this dir.
3589 dsl_bookmark_node_t
*dbn_next
;
3590 for (dsl_bookmark_node_t
*dbn
= avl_first(&origin_head
->ds_bookmarks
);
3591 dbn
!= NULL
&& dbn
->dbn_phys
.zbm_creation_txg
<=
3592 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3594 dbn_next
= AVL_NEXT(&origin_head
->ds_bookmarks
, dbn
);
3596 avl_remove(&origin_head
->ds_bookmarks
, dbn
);
3597 VERIFY0(zap_remove(dp
->dp_meta_objset
,
3598 origin_head
->ds_bookmarks_obj
, dbn
->dbn_name
, tx
));
3600 dsl_bookmark_node_add(hds
, dbn
, tx
);
3603 dsl_bookmark_next_changed(hds
, origin_ds
, tx
);
3605 /* move snapshots to this dir */
3606 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3607 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3608 dsl_dataset_t
*ds
= snap
->ds
;
3611 * Property callbacks are registered to a particular
3612 * dsl_dir. Since ours is changing, evict the objset
3613 * so that they will be unregistered from the old dsl_dir.
3615 if (ds
->ds_objset
) {
3616 dmu_objset_evict(ds
->ds_objset
);
3617 ds
->ds_objset
= NULL
;
3620 /* move snap name entry */
3621 VERIFY0(dsl_dataset_get_snapname(ds
));
3622 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3623 ds
->ds_snapname
, tx
, B_TRUE
));
3624 VERIFY0(zap_add(dp
->dp_meta_objset
,
3625 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3626 8, 1, &ds
->ds_object
, tx
));
3627 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3628 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3630 /* change containing dsl_dir */
3631 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3632 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3633 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3634 ASSERT3P(ds
->ds_dir
, ==, odd
);
3635 dsl_dir_rele(ds
->ds_dir
, ds
);
3636 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3637 NULL
, ds
, &ds
->ds_dir
));
3639 /* move any clone references */
3640 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3641 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3645 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3646 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3647 zap_cursor_retrieve(&zc
, &za
) == 0;
3648 zap_cursor_advance(&zc
)) {
3649 dsl_dataset_t
*cnds
;
3652 if (za
.za_first_integer
== oldnext_obj
) {
3654 * We've already moved the
3655 * origin's reference.
3660 VERIFY0(dsl_dataset_hold_obj(dp
,
3661 za
.za_first_integer
, FTAG
, &cnds
));
3662 o
= dsl_dir_phys(cnds
->ds_dir
)->
3663 dd_head_dataset_obj
;
3665 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3666 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3667 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3668 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3669 dsl_dataset_rele(cnds
, FTAG
);
3671 zap_cursor_fini(&zc
);
3674 ASSERT(!dsl_prop_hascb(ds
));
3678 * Change space accounting.
3679 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3680 * both be valid, or both be 0 (resulting in delta == 0). This
3681 * is true for each of {clone,origin} independently.
3684 delta
= ddpa
->cloneusedsnap
-
3685 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3686 ASSERT3S(delta
, >=, 0);
3687 ASSERT3U(ddpa
->used
, >=, delta
);
3688 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3689 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3690 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3692 delta
= ddpa
->originusedsnap
-
3693 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3694 ASSERT3S(delta
, <=, 0);
3695 ASSERT3U(ddpa
->used
, >=, -delta
);
3696 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3697 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3698 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3700 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3703 * Since livelists are specific to a clone's origin txg, they
3704 * are no longer accurate. Destroy the livelist from the clone being
3705 * promoted. If the origin dataset is a clone, destroy its livelist
3708 dsl_dir_remove_livelist(dd
, tx
, B_TRUE
);
3709 dsl_dir_remove_livelist(odd
, tx
, B_TRUE
);
3711 /* log history record */
3712 spa_history_log_internal_ds(hds
, "promote", tx
, " ");
3714 dsl_dir_rele(odd
, FTAG
);
3715 promote_rele(ddpa
, FTAG
);
3718 * Transfer common error blocks from old head to new head.
3720 if (spa_feature_is_enabled(dp
->dp_spa
, SPA_FEATURE_HEAD_ERRLOG
)) {
3721 uint64_t old_head
= origin_head
->ds_object
;
3722 uint64_t new_head
= hds
->ds_object
;
3723 spa_swap_errlog(dp
->dp_spa
, new_head
, old_head
, tx
);
3728 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3729 * (exclusive) and last_obj (inclusive). The list will be in reverse
3730 * order (last_obj will be the list_head()). If first_obj == 0, do all
3731 * snapshots back to this dataset's origin.
3734 snaplist_make(dsl_pool_t
*dp
,
3735 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, const void *tag
)
3737 uint64_t obj
= last_obj
;
3739 list_create(l
, sizeof (struct promotenode
),
3740 offsetof(struct promotenode
, link
));
3742 while (obj
!= first_obj
) {
3744 struct promotenode
*snap
;
3747 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3748 ASSERT(err
!= ENOENT
);
3753 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3755 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3757 list_insert_tail(l
, snap
);
3758 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3765 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3767 struct promotenode
*snap
;
3770 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3771 uint64_t used
, comp
, uncomp
;
3772 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3773 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3780 snaplist_destroy(list_t
*l
, const void *tag
)
3782 struct promotenode
*snap
;
3784 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3787 while ((snap
= list_remove_tail(l
)) != NULL
) {
3788 dsl_dataset_rele(snap
->ds
, tag
);
3789 kmem_free(snap
, sizeof (*snap
));
3795 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, const void *tag
)
3799 struct promotenode
*snap
;
3801 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3805 dd
= ddpa
->ddpa_clone
->ds_dir
;
3807 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3808 !dsl_dir_is_clone(dd
)) {
3809 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3810 return (SET_ERROR(EINVAL
));
3813 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3814 &ddpa
->shared_snaps
, tag
);
3818 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3819 &ddpa
->clone_snaps
, tag
);
3823 snap
= list_head(&ddpa
->shared_snaps
);
3824 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3825 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3826 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3827 &ddpa
->origin_snaps
, tag
);
3831 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3832 error
= dsl_dataset_hold_obj(dp
,
3833 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3834 tag
, &ddpa
->origin_origin
);
3840 promote_rele(ddpa
, tag
);
3845 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, const void *tag
)
3847 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3848 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3849 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3850 if (ddpa
->origin_origin
!= NULL
)
3851 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3852 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3858 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3859 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3862 dsl_dataset_promote(const char *name
, char *conflsnap
)
3864 dsl_dataset_promote_arg_t ddpa
= { 0 };
3867 nvpair_t
*snap_pair
;
3871 * We will modify space proportional to the number of
3872 * snapshots. Compute numsnaps.
3874 error
= dmu_objset_hold(name
, FTAG
, &os
);
3877 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3878 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3880 dmu_objset_rele(os
, FTAG
);
3884 ddpa
.ddpa_clonename
= name
;
3885 ddpa
.err_ds
= fnvlist_alloc();
3887 ddpa
.proc
= curproc
;
3889 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3890 dsl_dataset_promote_sync
, &ddpa
,
3891 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3894 * Return the first conflicting snapshot found.
3896 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3897 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3898 (void) strlcpy(conflsnap
, nvpair_name(snap_pair
),
3899 ZFS_MAX_DATASET_NAME_LEN
);
3901 fnvlist_free(ddpa
.err_ds
);
3906 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3907 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3910 * "slack" factor for received datasets with refquota set on them.
3911 * See the bottom of this function for details on its use.
3913 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3914 spa_asize_inflation
;
3915 int64_t unused_refres_delta
;
3917 /* they should both be heads */
3918 if (clone
->ds_is_snapshot
||
3919 origin_head
->ds_is_snapshot
)
3920 return (SET_ERROR(EINVAL
));
3922 /* if we are not forcing, the branch point should be just before them */
3923 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3924 return (SET_ERROR(EINVAL
));
3926 /* clone should be the clone (unless they are unrelated) */
3927 if (clone
->ds_prev
!= NULL
&&
3928 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3929 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3930 return (SET_ERROR(EINVAL
));
3932 /* the clone should be a child of the origin */
3933 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3934 return (SET_ERROR(EINVAL
));
3936 /* origin_head shouldn't be modified unless 'force' */
3938 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3939 return (SET_ERROR(ETXTBSY
));
3941 /* origin_head should have no long holds (e.g. is not mounted) */
3942 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3943 return (SET_ERROR(EBUSY
));
3945 /* check amount of any unconsumed refreservation */
3946 unused_refres_delta
=
3947 (int64_t)MIN(origin_head
->ds_reserved
,
3948 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3949 (int64_t)MIN(origin_head
->ds_reserved
,
3950 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3952 if (unused_refres_delta
> 0 &&
3953 unused_refres_delta
>
3954 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3955 return (SET_ERROR(ENOSPC
));
3958 * The clone can't be too much over the head's refquota.
3960 * To ensure that the entire refquota can be used, we allow one
3961 * transaction to exceed the refquota. Therefore, this check
3962 * needs to also allow for the space referenced to be more than the
3963 * refquota. The maximum amount of space that one transaction can use
3964 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3965 * overage ensures that we are able to receive a filesystem that
3966 * exceeds the refquota on the source system.
3968 * So that overage is the refquota_slack we use below.
3970 if (origin_head
->ds_quota
!= 0 &&
3971 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3972 origin_head
->ds_quota
+ refquota_slack
)
3973 return (SET_ERROR(EDQUOT
));
3979 dsl_dataset_swap_remap_deadlists(dsl_dataset_t
*clone
,
3980 dsl_dataset_t
*origin
, dmu_tx_t
*tx
)
3982 uint64_t clone_remap_dl_obj
, origin_remap_dl_obj
;
3983 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3985 ASSERT(dsl_pool_sync_context(dp
));
3987 clone_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(clone
);
3988 origin_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(origin
);
3990 if (clone_remap_dl_obj
!= 0) {
3991 dsl_deadlist_close(&clone
->ds_remap_deadlist
);
3992 dsl_dataset_unset_remap_deadlist_object(clone
, tx
);
3994 if (origin_remap_dl_obj
!= 0) {
3995 dsl_deadlist_close(&origin
->ds_remap_deadlist
);
3996 dsl_dataset_unset_remap_deadlist_object(origin
, tx
);
3999 if (clone_remap_dl_obj
!= 0) {
4000 dsl_dataset_set_remap_deadlist_object(origin
,
4001 clone_remap_dl_obj
, tx
);
4002 dsl_deadlist_open(&origin
->ds_remap_deadlist
,
4003 dp
->dp_meta_objset
, clone_remap_dl_obj
);
4005 if (origin_remap_dl_obj
!= 0) {
4006 dsl_dataset_set_remap_deadlist_object(clone
,
4007 origin_remap_dl_obj
, tx
);
4008 dsl_deadlist_open(&clone
->ds_remap_deadlist
,
4009 dp
->dp_meta_objset
, origin_remap_dl_obj
);
4014 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
4015 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
4017 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4018 int64_t unused_refres_delta
;
4020 ASSERT(clone
->ds_reserved
== 0);
4022 * NOTE: On DEBUG kernels there could be a race between this and
4023 * the check function if spa_asize_inflation is adjusted...
4025 ASSERT(origin_head
->ds_quota
== 0 ||
4026 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
4027 DMU_MAX_ACCESS
* spa_asize_inflation
);
4028 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
4030 dsl_dir_cancel_waiters(origin_head
->ds_dir
);
4033 * Swap per-dataset feature flags.
4035 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
4036 if (!(spa_feature_table
[f
].fi_flags
&
4037 ZFEATURE_FLAG_PER_DATASET
)) {
4038 ASSERT(!dsl_dataset_feature_is_active(clone
, f
));
4039 ASSERT(!dsl_dataset_feature_is_active(origin_head
, f
));
4043 boolean_t clone_inuse
= dsl_dataset_feature_is_active(clone
, f
);
4044 void *clone_feature
= clone
->ds_feature
[f
];
4045 boolean_t origin_head_inuse
=
4046 dsl_dataset_feature_is_active(origin_head
, f
);
4047 void *origin_head_feature
= origin_head
->ds_feature
[f
];
4050 dsl_dataset_deactivate_feature_impl(clone
, f
, tx
);
4051 if (origin_head_inuse
)
4052 dsl_dataset_deactivate_feature_impl(origin_head
, f
, tx
);
4055 dsl_dataset_activate_feature(origin_head
->ds_object
, f
,
4057 origin_head
->ds_feature
[f
] = clone_feature
;
4059 if (origin_head_inuse
) {
4060 dsl_dataset_activate_feature(clone
->ds_object
, f
,
4061 origin_head_feature
, tx
);
4062 clone
->ds_feature
[f
] = origin_head_feature
;
4066 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
4067 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
4069 if (clone
->ds_objset
!= NULL
) {
4070 dmu_objset_evict(clone
->ds_objset
);
4071 clone
->ds_objset
= NULL
;
4074 if (origin_head
->ds_objset
!= NULL
) {
4075 dmu_objset_evict(origin_head
->ds_objset
);
4076 origin_head
->ds_objset
= NULL
;
4079 unused_refres_delta
=
4080 (int64_t)MIN(origin_head
->ds_reserved
,
4081 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
4082 (int64_t)MIN(origin_head
->ds_reserved
,
4083 dsl_dataset_phys(clone
)->ds_unique_bytes
);
4086 * Reset origin's unique bytes.
4089 dsl_dataset_t
*origin
= clone
->ds_prev
;
4090 uint64_t comp
, uncomp
;
4092 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
4093 dsl_deadlist_space_range(&clone
->ds_deadlist
,
4094 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
4095 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
4100 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
4101 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
4103 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
4104 dsl_dataset_phys(origin_head
)->ds_bp
=
4105 dsl_dataset_phys(clone
)->ds_bp
;
4106 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
4107 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
4108 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
4111 /* set dd_*_bytes */
4113 int64_t dused
, dcomp
, duncomp
;
4114 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
4115 uint64_t odl_used
, odl_comp
, odl_uncomp
;
4117 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
4118 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
4120 dsl_deadlist_space(&clone
->ds_deadlist
,
4121 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
4122 dsl_deadlist_space(&origin_head
->ds_deadlist
,
4123 &odl_used
, &odl_comp
, &odl_uncomp
);
4125 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
4127 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
4129 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
4131 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
4133 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
4135 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
4138 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
4139 dused
, dcomp
, duncomp
, tx
);
4140 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
4141 -dused
, -dcomp
, -duncomp
, tx
);
4144 * The difference in the space used by snapshots is the
4145 * difference in snapshot space due to the head's
4146 * deadlist (since that's the only thing that's
4147 * changing that affects the snapused).
4149 dsl_deadlist_space_range(&clone
->ds_deadlist
,
4150 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
4151 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
4152 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
4153 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
4154 &odl_used
, &odl_comp
, &odl_uncomp
);
4155 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
4156 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
4159 /* swap ds_*_bytes */
4160 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
4161 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
4162 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
4163 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
4164 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
4165 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
4166 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
4167 dsl_dataset_phys(clone
)->ds_unique_bytes
);
4169 /* apply any parent delta for change in unconsumed refreservation */
4170 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
4171 unused_refres_delta
, 0, 0, tx
);
4176 dsl_deadlist_close(&clone
->ds_deadlist
);
4177 dsl_deadlist_close(&origin_head
->ds_deadlist
);
4178 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
4179 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
4180 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
4181 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
4182 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
4183 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
4184 dsl_dataset_swap_remap_deadlists(clone
, origin_head
, tx
);
4187 * If there is a bookmark at the origin, its "next dataset" is
4188 * changing, so we need to reset its FBN.
4190 dsl_bookmark_next_changed(origin_head
, origin_head
->ds_prev
, tx
);
4192 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
4195 * Destroy any livelists associated with the clone or the origin,
4196 * since after the swap the corresponding livelists are no longer
4199 dsl_dir_remove_livelist(clone
->ds_dir
, tx
, B_TRUE
);
4200 dsl_dir_remove_livelist(origin_head
->ds_dir
, tx
, B_TRUE
);
4202 spa_history_log_internal_ds(clone
, "clone swap", tx
,
4203 "parent=%s", origin_head
->ds_dir
->dd_myname
);
4207 * Given a pool name and a dataset object number in that pool,
4208 * return the name of that dataset.
4211 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
4217 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
4221 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
4223 dsl_dataset_name(ds
, buf
);
4224 dsl_dataset_rele(ds
, FTAG
);
4226 dsl_pool_rele(dp
, FTAG
);
4232 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
4233 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
4237 ASSERT3S(asize
, >, 0);
4240 * *ref_rsrv is the portion of asize that will come from any
4241 * unconsumed refreservation space.
4245 mutex_enter(&ds
->ds_lock
);
4247 * Make a space adjustment for reserved bytes.
4249 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
4251 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
4253 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
4255 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
4258 if (!check_quota
|| ds
->ds_quota
== 0) {
4259 mutex_exit(&ds
->ds_lock
);
4263 * If they are requesting more space, and our current estimate
4264 * is over quota, they get to try again unless the actual
4265 * on-disk is over quota and there are no pending changes (which
4266 * may free up space for us).
4268 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
4271 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
4272 error
= SET_ERROR(ERESTART
);
4274 error
= SET_ERROR(EDQUOT
);
4276 mutex_exit(&ds
->ds_lock
);
4281 typedef struct dsl_dataset_set_qr_arg
{
4282 const char *ddsqra_name
;
4283 zprop_source_t ddsqra_source
;
4284 uint64_t ddsqra_value
;
4285 } dsl_dataset_set_qr_arg_t
;
4289 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
4291 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4292 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4297 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
4298 return (SET_ERROR(ENOTSUP
));
4300 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
4304 if (ds
->ds_is_snapshot
) {
4305 dsl_dataset_rele(ds
, FTAG
);
4306 return (SET_ERROR(EINVAL
));
4309 error
= dsl_prop_predict(ds
->ds_dir
,
4310 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
4311 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
4313 dsl_dataset_rele(ds
, FTAG
);
4318 dsl_dataset_rele(ds
, FTAG
);
4322 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
4323 newval
< ds
->ds_reserved
) {
4324 dsl_dataset_rele(ds
, FTAG
);
4325 return (SET_ERROR(ENOSPC
));
4328 dsl_dataset_rele(ds
, FTAG
);
4333 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
4335 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4336 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4337 dsl_dataset_t
*ds
= NULL
;
4340 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
4342 dsl_prop_set_sync_impl(ds
,
4343 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
4344 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
4345 &ddsqra
->ddsqra_value
, tx
);
4347 VERIFY0(dsl_prop_get_int_ds(ds
,
4348 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
4350 if (ds
->ds_quota
!= newval
) {
4351 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
4352 ds
->ds_quota
= newval
;
4354 dsl_dataset_rele(ds
, FTAG
);
4358 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
4361 dsl_dataset_set_qr_arg_t ddsqra
;
4363 ddsqra
.ddsqra_name
= dsname
;
4364 ddsqra
.ddsqra_source
= source
;
4365 ddsqra
.ddsqra_value
= refquota
;
4367 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
4368 dsl_dataset_set_refquota_sync
, &ddsqra
, 0,
4369 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
4373 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
4375 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4376 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4379 uint64_t newval
, unique
;
4381 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
4382 return (SET_ERROR(ENOTSUP
));
4384 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
4388 if (ds
->ds_is_snapshot
) {
4389 dsl_dataset_rele(ds
, FTAG
);
4390 return (SET_ERROR(EINVAL
));
4393 error
= dsl_prop_predict(ds
->ds_dir
,
4394 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
4395 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
4397 dsl_dataset_rele(ds
, FTAG
);
4402 * If we are doing the preliminary check in open context, the
4403 * space estimates may be inaccurate.
4405 if (!dmu_tx_is_syncing(tx
)) {
4406 dsl_dataset_rele(ds
, FTAG
);
4410 mutex_enter(&ds
->ds_lock
);
4411 if (!DS_UNIQUE_IS_ACCURATE(ds
))
4412 dsl_dataset_recalc_head_uniq(ds
);
4413 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
4414 mutex_exit(&ds
->ds_lock
);
4416 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
4417 uint64_t delta
= MAX(unique
, newval
) -
4418 MAX(unique
, ds
->ds_reserved
);
4421 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
4422 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
4423 dsl_dataset_rele(ds
, FTAG
);
4424 return (SET_ERROR(ENOSPC
));
4428 dsl_dataset_rele(ds
, FTAG
);
4433 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
4434 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
4440 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
4441 source
, sizeof (value
), 1, &value
, tx
);
4443 VERIFY0(dsl_prop_get_int_ds(ds
,
4444 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
4446 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
4447 mutex_enter(&ds
->ds_dir
->dd_lock
);
4448 mutex_enter(&ds
->ds_lock
);
4449 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
4450 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
4451 delta
= MAX(0, (int64_t)(newval
- unique
)) -
4452 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
4453 ds
->ds_reserved
= newval
;
4454 mutex_exit(&ds
->ds_lock
);
4456 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
4457 mutex_exit(&ds
->ds_dir
->dd_lock
);
4461 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
4463 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4464 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4465 dsl_dataset_t
*ds
= NULL
;
4467 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
4468 dsl_dataset_set_refreservation_sync_impl(ds
,
4469 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
4470 dsl_dataset_rele(ds
, FTAG
);
4474 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
4475 uint64_t refreservation
)
4477 dsl_dataset_set_qr_arg_t ddsqra
;
4479 ddsqra
.ddsqra_name
= dsname
;
4480 ddsqra
.ddsqra_source
= source
;
4481 ddsqra
.ddsqra_value
= refreservation
;
4483 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
4484 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0,
4485 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
4488 typedef struct dsl_dataset_set_compression_arg
{
4489 const char *ddsca_name
;
4490 zprop_source_t ddsca_source
;
4491 uint64_t ddsca_value
;
4492 } dsl_dataset_set_compression_arg_t
;
4495 dsl_dataset_set_compression_check(void *arg
, dmu_tx_t
*tx
)
4497 dsl_dataset_set_compression_arg_t
*ddsca
= arg
;
4498 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4500 uint64_t compval
= ZIO_COMPRESS_ALGO(ddsca
->ddsca_value
);
4501 spa_feature_t f
= zio_compress_to_feature(compval
);
4503 if (f
== SPA_FEATURE_NONE
)
4504 return (SET_ERROR(EINVAL
));
4506 if (!spa_feature_is_enabled(dp
->dp_spa
, f
))
4507 return (SET_ERROR(ENOTSUP
));
4513 dsl_dataset_set_compression_sync(void *arg
, dmu_tx_t
*tx
)
4515 dsl_dataset_set_compression_arg_t
*ddsca
= arg
;
4516 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4517 dsl_dataset_t
*ds
= NULL
;
4519 uint64_t compval
= ZIO_COMPRESS_ALGO(ddsca
->ddsca_value
);
4520 spa_feature_t f
= zio_compress_to_feature(compval
);
4521 ASSERT3S(f
, !=, SPA_FEATURE_NONE
);
4522 ASSERT3S(spa_feature_table
[f
].fi_type
, ==, ZFEATURE_TYPE_BOOLEAN
);
4524 VERIFY0(dsl_dataset_hold(dp
, ddsca
->ddsca_name
, FTAG
, &ds
));
4525 if (zfeature_active(f
, ds
->ds_feature
[f
]) != B_TRUE
) {
4526 ds
->ds_feature_activation
[f
] = (void *)B_TRUE
;
4527 dsl_dataset_activate_feature(ds
->ds_object
, f
,
4528 ds
->ds_feature_activation
[f
], tx
);
4529 ds
->ds_feature
[f
] = ds
->ds_feature_activation
[f
];
4531 dsl_dataset_rele(ds
, FTAG
);
4535 dsl_dataset_set_compression(const char *dsname
, zprop_source_t source
,
4536 uint64_t compression
)
4538 dsl_dataset_set_compression_arg_t ddsca
;
4541 * The sync task is only required for zstd in order to activate
4542 * the feature flag when the property is first set.
4544 if (ZIO_COMPRESS_ALGO(compression
) != ZIO_COMPRESS_ZSTD
)
4547 ddsca
.ddsca_name
= dsname
;
4548 ddsca
.ddsca_source
= source
;
4549 ddsca
.ddsca_value
= compression
;
4551 return (dsl_sync_task(dsname
, dsl_dataset_set_compression_check
,
4552 dsl_dataset_set_compression_sync
, &ddsca
, 0,
4553 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
4557 * Return (in *usedp) the amount of space referenced by "new" that was not
4558 * referenced at the time the bookmark corresponds to. "New" may be a
4559 * snapshot or a head. The bookmark must be before new, in
4560 * new's filesystem (or its origin) -- caller verifies this.
4562 * The written space is calculated by considering two components: First, we
4563 * ignore any freed space, and calculate the written as new's used space
4564 * minus old's used space. Next, we add in the amount of space that was freed
4565 * between the two time points, thus reducing new's used space relative to
4566 * old's. Specifically, this is the space that was born before
4567 * zbm_creation_txg, and freed before new (ie. on new's deadlist or a
4568 * previous deadlist).
4570 * space freed [---------------------]
4571 * snapshots ---O-------O--------O-------O------
4574 * Note, the bookmark's zbm_*_bytes_refd must be valid, but if the HAS_FBN
4575 * flag is not set, we will calculate the freed_before_next based on the
4576 * next snapshot's deadlist, rather than using zbm_*_freed_before_next_snap.
4579 dsl_dataset_space_written_impl(zfs_bookmark_phys_t
*bmp
,
4580 dsl_dataset_t
*new, uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4583 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
4585 ASSERT(dsl_pool_config_held(dp
));
4586 if (dsl_dataset_is_snapshot(new)) {
4587 ASSERT3U(bmp
->zbm_creation_txg
, <,
4588 dsl_dataset_phys(new)->ds_creation_txg
);
4592 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
4593 *usedp
-= bmp
->zbm_referenced_bytes_refd
;
4596 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
4597 *compp
-= bmp
->zbm_compressed_bytes_refd
;
4600 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
4601 *uncompp
-= bmp
->zbm_uncompressed_bytes_refd
;
4603 dsl_dataset_t
*snap
= new;
4605 while (dsl_dataset_phys(snap
)->ds_prev_snap_txg
>
4606 bmp
->zbm_creation_txg
) {
4607 uint64_t used
, comp
, uncomp
;
4609 dsl_deadlist_space_range(&snap
->ds_deadlist
,
4610 0, bmp
->zbm_creation_txg
,
4611 &used
, &comp
, &uncomp
);
4616 uint64_t snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
4618 dsl_dataset_rele(snap
, FTAG
);
4619 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
4625 * We might not have the FBN if we are calculating written from
4626 * a snapshot (because we didn't know the correct "next" snapshot
4629 if (bmp
->zbm_flags
& ZBM_FLAG_HAS_FBN
) {
4630 *usedp
+= bmp
->zbm_referenced_freed_before_next_snap
;
4631 *compp
+= bmp
->zbm_compressed_freed_before_next_snap
;
4632 *uncompp
+= bmp
->zbm_uncompressed_freed_before_next_snap
;
4634 ASSERT3U(dsl_dataset_phys(snap
)->ds_prev_snap_txg
, ==,
4635 bmp
->zbm_creation_txg
);
4636 uint64_t used
, comp
, uncomp
;
4637 dsl_deadlist_space(&snap
->ds_deadlist
, &used
, &comp
, &uncomp
);
4643 dsl_dataset_rele(snap
, FTAG
);
4648 * Return (in *usedp) the amount of space written in new that was not
4649 * present at the time the bookmark corresponds to. New may be a
4650 * snapshot or the head. Old must be a bookmark before new, in
4651 * new's filesystem (or its origin) -- caller verifies this.
4654 dsl_dataset_space_written_bookmark(zfs_bookmark_phys_t
*bmp
,
4655 dsl_dataset_t
*new, uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4657 if (!(bmp
->zbm_flags
& ZBM_FLAG_HAS_FBN
))
4658 return (SET_ERROR(ENOTSUP
));
4659 return (dsl_dataset_space_written_impl(bmp
, new,
4660 usedp
, compp
, uncompp
));
4664 * Return (in *usedp) the amount of space written in new that is not
4665 * present in oldsnap. New may be a snapshot or the head. Old must be
4666 * a snapshot before new, in new's filesystem (or its origin). If not then
4667 * fail and return EINVAL.
4670 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
4671 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4673 if (!dsl_dataset_is_before(new, oldsnap
, 0))
4674 return (SET_ERROR(EINVAL
));
4676 zfs_bookmark_phys_t zbm
= { 0 };
4677 dsl_dataset_phys_t
*dsp
= dsl_dataset_phys(oldsnap
);
4678 zbm
.zbm_guid
= dsp
->ds_guid
;
4679 zbm
.zbm_creation_txg
= dsp
->ds_creation_txg
;
4680 zbm
.zbm_creation_time
= dsp
->ds_creation_time
;
4681 zbm
.zbm_referenced_bytes_refd
= dsp
->ds_referenced_bytes
;
4682 zbm
.zbm_compressed_bytes_refd
= dsp
->ds_compressed_bytes
;
4683 zbm
.zbm_uncompressed_bytes_refd
= dsp
->ds_uncompressed_bytes
;
4686 * If oldsnap is the origin (or origin's origin, ...) of new,
4687 * we can't easily calculate the effective FBN. Therefore,
4688 * we do not set ZBM_FLAG_HAS_FBN, so that the _impl will calculate
4689 * it relative to the correct "next": the next snapshot towards "new",
4690 * rather than the next snapshot in oldsnap's dsl_dir.
4692 return (dsl_dataset_space_written_impl(&zbm
, new,
4693 usedp
, compp
, uncompp
));
4697 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
4698 * lastsnap, and all snapshots in between are deleted.
4700 * blocks that would be freed [---------------------------]
4701 * snapshots ---O-------O--------O-------O--------O
4702 * firstsnap lastsnap
4704 * This is the set of blocks that were born after the snap before firstsnap,
4705 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
4706 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
4707 * We calculate this by iterating over the relevant deadlists (from the snap
4708 * after lastsnap, backward to the snap after firstsnap), summing up the
4709 * space on the deadlist that was born after the snap before firstsnap.
4712 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
4713 dsl_dataset_t
*lastsnap
,
4714 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4718 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
4720 ASSERT(firstsnap
->ds_is_snapshot
);
4721 ASSERT(lastsnap
->ds_is_snapshot
);
4724 * Check that the snapshots are in the same dsl_dir, and firstsnap
4725 * is before lastsnap.
4727 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
4728 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
4729 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
4730 return (SET_ERROR(EINVAL
));
4732 *usedp
= *compp
= *uncompp
= 0;
4734 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
4735 while (snapobj
!= firstsnap
->ds_object
) {
4737 uint64_t used
, comp
, uncomp
;
4739 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
4743 dsl_deadlist_space_range(&ds
->ds_deadlist
,
4744 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
4745 &used
, &comp
, &uncomp
);
4750 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4751 ASSERT3U(snapobj
, !=, 0);
4752 dsl_dataset_rele(ds
, FTAG
);
4758 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
4759 * For example, they could both be snapshots of the same filesystem, and
4760 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
4761 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
4762 * filesystem. Or 'earlier' could be the origin's origin.
4764 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
4767 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
4768 uint64_t earlier_txg
)
4770 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
4774 ASSERT(dsl_pool_config_held(dp
));
4775 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
4777 if (earlier_txg
== 0)
4778 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
4780 if (later
->ds_is_snapshot
&&
4781 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
4784 if (later
->ds_dir
== earlier
->ds_dir
)
4788 * We check dd_origin_obj explicitly here rather than using
4789 * dsl_dir_is_clone() so that we will return TRUE if "earlier"
4790 * is $ORIGIN@$ORIGIN. dsl_dataset_space_written() depends on
4793 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== 0)
4796 dsl_dataset_t
*origin
;
4797 error
= dsl_dataset_hold_obj(dp
,
4798 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
4801 if (dsl_dataset_phys(origin
)->ds_creation_txg
== earlier_txg
&&
4802 origin
->ds_dir
== earlier
->ds_dir
) {
4803 dsl_dataset_rele(origin
, FTAG
);
4806 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
4807 dsl_dataset_rele(origin
, FTAG
);
4812 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4814 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4815 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4819 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4821 dmu_object_info_t doi
;
4823 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4824 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4828 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4830 return (dsl_dataset_is_zapified(ds
) &&
4831 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4832 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4836 dsl_dataset_get_remap_deadlist_object(dsl_dataset_t
*ds
)
4838 uint64_t remap_deadlist_obj
;
4841 if (!dsl_dataset_is_zapified(ds
))
4844 err
= zap_lookup(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4845 DS_FIELD_REMAP_DEADLIST
, sizeof (remap_deadlist_obj
), 1,
4846 &remap_deadlist_obj
);
4849 VERIFY3S(err
, ==, ENOENT
);
4853 ASSERT(remap_deadlist_obj
!= 0);
4854 return (remap_deadlist_obj
);
4858 dsl_dataset_remap_deadlist_exists(dsl_dataset_t
*ds
)
4860 EQUIV(dsl_deadlist_is_open(&ds
->ds_remap_deadlist
),
4861 dsl_dataset_get_remap_deadlist_object(ds
) != 0);
4862 return (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
));
4866 dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
, uint64_t obj
,
4870 dsl_dataset_zapify(ds
, tx
);
4871 VERIFY0(zap_add(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4872 DS_FIELD_REMAP_DEADLIST
, sizeof (obj
), 1, &obj
, tx
));
4876 dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4878 VERIFY0(zap_remove(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4879 ds
->ds_object
, DS_FIELD_REMAP_DEADLIST
, tx
));
4883 dsl_dataset_destroy_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4885 uint64_t remap_deadlist_object
;
4886 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4888 ASSERT(dmu_tx_is_syncing(tx
));
4889 ASSERT(dsl_dataset_remap_deadlist_exists(ds
));
4891 remap_deadlist_object
= ds
->ds_remap_deadlist
.dl_object
;
4892 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
4893 dsl_deadlist_free(spa_meta_objset(spa
), remap_deadlist_object
, tx
);
4894 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
4895 spa_feature_decr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4899 dsl_dataset_create_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4901 uint64_t remap_deadlist_obj
;
4902 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4904 ASSERT(dmu_tx_is_syncing(tx
));
4905 ASSERT(MUTEX_HELD(&ds
->ds_remap_deadlist_lock
));
4907 * Currently we only create remap deadlists when there are indirect
4908 * vdevs with referenced mappings.
4910 ASSERT(spa_feature_is_active(spa
, SPA_FEATURE_DEVICE_REMOVAL
));
4912 remap_deadlist_obj
= dsl_deadlist_clone(
4913 &ds
->ds_deadlist
, UINT64_MAX
,
4914 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
4915 dsl_dataset_set_remap_deadlist_object(ds
,
4916 remap_deadlist_obj
, tx
);
4917 dsl_deadlist_open(&ds
->ds_remap_deadlist
, spa_meta_objset(spa
),
4918 remap_deadlist_obj
);
4919 spa_feature_incr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4923 dsl_dataset_activate_redaction(dsl_dataset_t
*ds
, uint64_t *redact_snaps
,
4924 uint64_t num_redact_snaps
, dmu_tx_t
*tx
)
4926 uint64_t dsobj
= ds
->ds_object
;
4927 struct feature_type_uint64_array_arg
*ftuaa
=
4928 kmem_zalloc(sizeof (*ftuaa
), KM_SLEEP
);
4929 ftuaa
->length
= (int64_t)num_redact_snaps
;
4930 if (num_redact_snaps
> 0) {
4931 ftuaa
->array
= kmem_alloc(num_redact_snaps
* sizeof (uint64_t),
4933 memcpy(ftuaa
->array
, redact_snaps
, num_redact_snaps
*
4936 dsl_dataset_activate_feature(dsobj
, SPA_FEATURE_REDACTED_DATASETS
,
4938 ds
->ds_feature
[SPA_FEATURE_REDACTED_DATASETS
] = ftuaa
;
4942 * Find and return (in *oldest_dsobj) the oldest snapshot of the dsobj
4943 * dataset whose birth time is >= min_txg.
4946 dsl_dataset_oldest_snapshot(spa_t
*spa
, uint64_t head_ds
, uint64_t min_txg
,
4947 uint64_t *oldest_dsobj
)
4950 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
4952 int error
= dsl_dataset_hold_obj(dp
, head_ds
, FTAG
, &ds
);
4956 uint64_t prev_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4957 uint64_t prev_obj_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
4959 while (prev_obj
!= 0 && min_txg
< prev_obj_txg
) {
4960 dsl_dataset_rele(ds
, FTAG
);
4961 if ((error
= dsl_dataset_hold_obj(dp
, prev_obj
,
4964 prev_obj_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
4965 prev_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4967 *oldest_dsobj
= ds
->ds_object
;
4968 dsl_dataset_rele(ds
, FTAG
);
4972 ZFS_MODULE_PARAM(zfs
, zfs_
, max_recordsize
, UINT
, ZMOD_RW
,
4973 "Max allowed record size");
4975 ZFS_MODULE_PARAM(zfs
, zfs_
, allow_redacted_dataset_mount
, INT
, ZMOD_RW
,
4976 "Allow mounting of redacted datasets");
4978 ZFS_MODULE_PARAM(zfs
, zfs_
, snapshot_history_enabled
, INT
, ZMOD_RW
,
4979 "Include snapshot events in pool history/events");
4981 EXPORT_SYMBOL(dsl_dataset_hold
);
4982 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4983 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4984 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4985 EXPORT_SYMBOL(dsl_dataset_own
);
4986 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4987 EXPORT_SYMBOL(dsl_dataset_name
);
4988 EXPORT_SYMBOL(dsl_dataset_rele
);
4989 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4990 EXPORT_SYMBOL(dsl_dataset_disown
);
4991 EXPORT_SYMBOL(dsl_dataset_tryown
);
4992 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4993 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4994 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4995 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4996 EXPORT_SYMBOL(dsl_dataset_promote
);
4997 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4998 EXPORT_SYMBOL(dsl_dataset_user_release
);
4999 EXPORT_SYMBOL(dsl_dataset_get_holds
);
5000 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
5001 EXPORT_SYMBOL(dsl_dataset_get_spa
);
5002 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
5003 EXPORT_SYMBOL(dsl_dataset_space_written
);
5004 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
5005 EXPORT_SYMBOL(dsl_dataset_sync
);
5006 EXPORT_SYMBOL(dsl_dataset_block_born
);
5007 EXPORT_SYMBOL(dsl_dataset_block_kill
);
5008 EXPORT_SYMBOL(dsl_dataset_dirty
);
5009 EXPORT_SYMBOL(dsl_dataset_stats
);
5010 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
5011 EXPORT_SYMBOL(dsl_dataset_space
);
5012 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
5013 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
5014 EXPORT_SYMBOL(dsl_dataset_check_quota
);
5015 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
5016 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);