4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
25 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
26 * Copyright (c) 2014 RackTop Systems.
27 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
28 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
29 * Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
30 * Copyright 2017 Nexenta Systems, Inc.
33 #include <sys/dmu_objset.h>
34 #include <sys/dsl_dataset.h>
35 #include <sys/dsl_dir.h>
36 #include <sys/dsl_prop.h>
37 #include <sys/dsl_synctask.h>
38 #include <sys/dmu_traverse.h>
39 #include <sys/dmu_impl.h>
40 #include <sys/dmu_tx.h>
44 #include <sys/zfeature.h>
45 #include <sys/unique.h>
46 #include <sys/zfs_context.h>
47 #include <sys/zfs_ioctl.h>
49 #include <sys/spa_impl.h>
51 #include <sys/zfs_znode.h>
52 #include <sys/zfs_onexit.h>
54 #include <sys/dsl_scan.h>
55 #include <sys/dsl_deadlist.h>
56 #include <sys/dsl_destroy.h>
57 #include <sys/dsl_userhold.h>
58 #include <sys/dsl_bookmark.h>
59 #include <sys/policy.h>
60 #include <sys/dmu_recv.h>
61 #include <sys/zio_compress.h>
62 #include <zfs_fletcher.h>
63 #include <sys/zio_checksum.h>
66 * The SPA supports block sizes up to 16MB. However, very large blocks
67 * can have an impact on i/o latency (e.g. tying up a spinning disk for
68 * ~300ms), and also potentially on the memory allocator. Therefore,
69 * we do not allow the recordsize to be set larger than zfs_max_recordsize
70 * (default 1MB). Larger blocks can be created by changing this tunable,
71 * and pools with larger blocks can always be imported and used, regardless
74 int zfs_max_recordsize
= 1 * 1024 * 1024;
76 #define SWITCH64(x, y) \
78 uint64_t __tmp = (x); \
83 #define DS_REF_MAX (1ULL << 62)
85 extern inline dsl_dataset_phys_t
*dsl_dataset_phys(dsl_dataset_t
*ds
);
87 static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
,
88 uint64_t obj
, dmu_tx_t
*tx
);
89 static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
,
92 static void unload_zfeature(dsl_dataset_t
*ds
, spa_feature_t f
);
94 extern int spa_asize_inflation
;
96 static zil_header_t zero_zil
;
99 * Figure out how much of this delta should be propagated to the dsl_dir
100 * layer. If there's a refreservation, that space has already been
101 * partially accounted for in our ancestors.
104 parent_delta(dsl_dataset_t
*ds
, int64_t delta
)
106 dsl_dataset_phys_t
*ds_phys
;
107 uint64_t old_bytes
, new_bytes
;
109 if (ds
->ds_reserved
== 0)
112 ds_phys
= dsl_dataset_phys(ds
);
113 old_bytes
= MAX(ds_phys
->ds_unique_bytes
, ds
->ds_reserved
);
114 new_bytes
= MAX(ds_phys
->ds_unique_bytes
+ delta
, ds
->ds_reserved
);
116 ASSERT3U(ABS((int64_t)(new_bytes
- old_bytes
)), <=, ABS(delta
));
117 return (new_bytes
- old_bytes
);
121 dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
123 int used
, compressed
, uncompressed
;
126 used
= bp_get_dsize_sync(tx
->tx_pool
->dp_spa
, bp
);
127 compressed
= BP_GET_PSIZE(bp
);
128 uncompressed
= BP_GET_UCSIZE(bp
);
130 dprintf_bp(bp
, "ds=%p", ds
);
132 ASSERT(dmu_tx_is_syncing(tx
));
133 /* It could have been compressed away to nothing */
136 ASSERT(BP_GET_TYPE(bp
) != DMU_OT_NONE
);
137 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp
)));
139 dsl_pool_mos_diduse_space(tx
->tx_pool
,
140 used
, compressed
, uncompressed
);
144 ASSERT3U(bp
->blk_birth
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
145 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
146 mutex_enter(&ds
->ds_lock
);
147 delta
= parent_delta(ds
, used
);
148 dsl_dataset_phys(ds
)->ds_referenced_bytes
+= used
;
149 dsl_dataset_phys(ds
)->ds_compressed_bytes
+= compressed
;
150 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
+= uncompressed
;
151 dsl_dataset_phys(ds
)->ds_unique_bytes
+= used
;
153 if (BP_GET_LSIZE(bp
) > SPA_OLD_MAXBLOCKSIZE
) {
154 ds
->ds_feature_activation
[SPA_FEATURE_LARGE_BLOCKS
] =
158 spa_feature_t f
= zio_checksum_to_feature(BP_GET_CHECKSUM(bp
));
159 if (f
!= SPA_FEATURE_NONE
) {
160 ASSERT3S(spa_feature_table
[f
].fi_type
, ==,
161 ZFEATURE_TYPE_BOOLEAN
);
162 ds
->ds_feature_activation
[f
] = (void *)B_TRUE
;
165 mutex_exit(&ds
->ds_lock
);
166 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
, delta
,
167 compressed
, uncompressed
, tx
);
168 dsl_dir_transfer_space(ds
->ds_dir
, used
- delta
,
169 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
173 * Called when the specified segment has been remapped, and is thus no
174 * longer referenced in the head dataset. The vdev must be indirect.
176 * If the segment is referenced by a snapshot, put it on the remap deadlist.
177 * Otherwise, add this segment to the obsolete spacemap.
180 dsl_dataset_block_remapped(dsl_dataset_t
*ds
, uint64_t vdev
, uint64_t offset
,
181 uint64_t size
, uint64_t birth
, dmu_tx_t
*tx
)
183 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
185 ASSERT(dmu_tx_is_syncing(tx
));
186 ASSERT(birth
<= tx
->tx_txg
);
187 ASSERT(!ds
->ds_is_snapshot
);
189 if (birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
190 spa_vdev_indirect_mark_obsolete(spa
, vdev
, offset
, size
, tx
);
193 dva_t
*dva
= &fakebp
.blk_dva
[0];
197 mutex_enter(&ds
->ds_remap_deadlist_lock
);
198 if (!dsl_dataset_remap_deadlist_exists(ds
)) {
199 dsl_dataset_create_remap_deadlist(ds
, tx
);
201 mutex_exit(&ds
->ds_remap_deadlist_lock
);
204 fakebp
.blk_birth
= birth
;
205 DVA_SET_VDEV(dva
, vdev
);
206 DVA_SET_OFFSET(dva
, offset
);
207 DVA_SET_ASIZE(dva
, size
);
209 dsl_deadlist_insert(&ds
->ds_remap_deadlist
, &fakebp
, tx
);
214 dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
, dmu_tx_t
*tx
,
217 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
219 int used
= bp_get_dsize_sync(spa
, bp
);
220 int compressed
= BP_GET_PSIZE(bp
);
221 int uncompressed
= BP_GET_UCSIZE(bp
);
226 ASSERT(dmu_tx_is_syncing(tx
));
227 ASSERT(bp
->blk_birth
<= tx
->tx_txg
);
230 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
231 dsl_pool_mos_diduse_space(tx
->tx_pool
,
232 -used
, -compressed
, -uncompressed
);
235 ASSERT3P(tx
->tx_pool
, ==, ds
->ds_dir
->dd_pool
);
237 ASSERT(!ds
->ds_is_snapshot
);
238 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
240 if (bp
->blk_birth
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
243 dprintf_bp(bp
, "freeing ds=%llu", ds
->ds_object
);
244 dsl_free(tx
->tx_pool
, tx
->tx_txg
, bp
);
246 mutex_enter(&ds
->ds_lock
);
247 ASSERT(dsl_dataset_phys(ds
)->ds_unique_bytes
>= used
||
248 !DS_UNIQUE_IS_ACCURATE(ds
));
249 delta
= parent_delta(ds
, -used
);
250 dsl_dataset_phys(ds
)->ds_unique_bytes
-= used
;
251 mutex_exit(&ds
->ds_lock
);
252 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_HEAD
,
253 delta
, -compressed
, -uncompressed
, tx
);
254 dsl_dir_transfer_space(ds
->ds_dir
, -used
- delta
,
255 DD_USED_REFRSRV
, DD_USED_HEAD
, tx
);
257 dprintf_bp(bp
, "putting on dead list: %s", "");
260 * We are here as part of zio's write done callback,
261 * which means we're a zio interrupt thread. We can't
262 * call dsl_deadlist_insert() now because it may block
263 * waiting for I/O. Instead, put bp on the deferred
264 * queue and let dsl_pool_sync() finish the job.
266 bplist_append(&ds
->ds_pending_deadlist
, bp
);
268 dsl_deadlist_insert(&ds
->ds_deadlist
, bp
, tx
);
270 ASSERT3U(ds
->ds_prev
->ds_object
, ==,
271 dsl_dataset_phys(ds
)->ds_prev_snap_obj
);
272 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 0);
273 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
274 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
275 ds
->ds_object
&& bp
->blk_birth
>
276 dsl_dataset_phys(ds
->ds_prev
)->ds_prev_snap_txg
) {
277 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
278 mutex_enter(&ds
->ds_prev
->ds_lock
);
279 dsl_dataset_phys(ds
->ds_prev
)->ds_unique_bytes
+= used
;
280 mutex_exit(&ds
->ds_prev
->ds_lock
);
282 if (bp
->blk_birth
> ds
->ds_dir
->dd_origin_txg
) {
283 dsl_dir_transfer_space(ds
->ds_dir
, used
,
284 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
287 mutex_enter(&ds
->ds_lock
);
288 ASSERT3U(dsl_dataset_phys(ds
)->ds_referenced_bytes
, >=, used
);
289 dsl_dataset_phys(ds
)->ds_referenced_bytes
-= used
;
290 ASSERT3U(dsl_dataset_phys(ds
)->ds_compressed_bytes
, >=, compressed
);
291 dsl_dataset_phys(ds
)->ds_compressed_bytes
-= compressed
;
292 ASSERT3U(dsl_dataset_phys(ds
)->ds_uncompressed_bytes
, >=, uncompressed
);
293 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
-= uncompressed
;
294 mutex_exit(&ds
->ds_lock
);
299 struct feature_type_uint64_array_arg
{
305 unload_zfeature(dsl_dataset_t
*ds
, spa_feature_t f
)
307 switch (spa_feature_table
[f
].fi_type
) {
308 case ZFEATURE_TYPE_BOOLEAN
:
310 case ZFEATURE_TYPE_UINT64_ARRAY
:
312 struct feature_type_uint64_array_arg
*ftuaa
= ds
->ds_feature
[f
];
313 kmem_free(ftuaa
->array
, ftuaa
->length
* sizeof (uint64_t));
314 kmem_free(ftuaa
, sizeof (*ftuaa
));
318 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
323 load_zfeature(objset_t
*mos
, dsl_dataset_t
*ds
, spa_feature_t f
)
326 switch (spa_feature_table
[f
].fi_type
) {
327 case ZFEATURE_TYPE_BOOLEAN
:
328 err
= zap_contains(mos
, ds
->ds_object
,
329 spa_feature_table
[f
].fi_guid
);
331 ds
->ds_feature
[f
] = (void *)B_TRUE
;
333 ASSERT3U(err
, ==, ENOENT
);
337 case ZFEATURE_TYPE_UINT64_ARRAY
:
339 uint64_t int_size
, num_int
;
341 err
= zap_length(mos
, ds
->ds_object
,
342 spa_feature_table
[f
].fi_guid
, &int_size
, &num_int
);
344 ASSERT3U(err
, ==, ENOENT
);
348 ASSERT3U(int_size
, ==, sizeof (uint64_t));
349 data
= kmem_alloc(int_size
* num_int
, KM_SLEEP
);
350 VERIFY0(zap_lookup(mos
, ds
->ds_object
,
351 spa_feature_table
[f
].fi_guid
, int_size
, num_int
, data
));
352 struct feature_type_uint64_array_arg
*ftuaa
=
353 kmem_alloc(sizeof (*ftuaa
), KM_SLEEP
);
354 ftuaa
->length
= num_int
;
356 ds
->ds_feature
[f
] = ftuaa
;
360 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
366 * We have to release the fsid syncronously or we risk that a subsequent
367 * mount of the same dataset will fail to unique_insert the fsid. This
368 * failure would manifest itself as the fsid of this dataset changing
369 * between mounts which makes NFS clients quite unhappy.
372 dsl_dataset_evict_sync(void *dbu
)
374 dsl_dataset_t
*ds
= dbu
;
376 ASSERT(ds
->ds_owner
== NULL
);
378 unique_remove(ds
->ds_fsid_guid
);
382 dsl_dataset_evict_async(void *dbu
)
384 dsl_dataset_t
*ds
= dbu
;
386 ASSERT(ds
->ds_owner
== NULL
);
390 if (ds
->ds_objset
!= NULL
)
391 dmu_objset_evict(ds
->ds_objset
);
394 dsl_dataset_rele(ds
->ds_prev
, ds
);
398 bplist_destroy(&ds
->ds_pending_deadlist
);
399 if (dsl_deadlist_is_open(&ds
->ds_deadlist
))
400 dsl_deadlist_close(&ds
->ds_deadlist
);
401 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
402 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
404 dsl_dir_async_rele(ds
->ds_dir
, ds
);
406 ASSERT(!list_link_active(&ds
->ds_synced_link
));
408 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
409 if (dsl_dataset_feature_is_active(ds
, f
))
410 unload_zfeature(ds
, f
);
413 list_destroy(&ds
->ds_prop_cbs
);
414 mutex_destroy(&ds
->ds_lock
);
415 mutex_destroy(&ds
->ds_opening_lock
);
416 mutex_destroy(&ds
->ds_sendstream_lock
);
417 mutex_destroy(&ds
->ds_remap_deadlist_lock
);
418 zfs_refcount_destroy(&ds
->ds_longholds
);
419 rrw_destroy(&ds
->ds_bp_rwlock
);
421 kmem_free(ds
, sizeof (dsl_dataset_t
));
425 dsl_dataset_get_snapname(dsl_dataset_t
*ds
)
427 dsl_dataset_phys_t
*headphys
;
430 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
431 objset_t
*mos
= dp
->dp_meta_objset
;
433 if (ds
->ds_snapname
[0])
435 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0)
438 err
= dmu_bonus_hold(mos
, dsl_dir_phys(ds
->ds_dir
)->dd_head_dataset_obj
,
442 headphys
= headdbuf
->db_data
;
443 err
= zap_value_search(dp
->dp_meta_objset
,
444 headphys
->ds_snapnames_zapobj
, ds
->ds_object
, 0, ds
->ds_snapname
);
445 if (err
!= 0 && zfs_recover
== B_TRUE
) {
447 (void) snprintf(ds
->ds_snapname
, sizeof (ds
->ds_snapname
),
448 "SNAPOBJ=%llu-ERR=%d",
449 (unsigned long long)ds
->ds_object
, err
);
451 dmu_buf_rele(headdbuf
, FTAG
);
456 dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
, uint64_t *value
)
458 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
459 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
463 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
466 err
= zap_lookup_norm(mos
, snapobj
, name
, 8, 1,
467 value
, mt
, NULL
, 0, NULL
);
468 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
469 err
= zap_lookup(mos
, snapobj
, name
, 8, 1, value
);
474 dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
477 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
478 uint64_t snapobj
= dsl_dataset_phys(ds
)->ds_snapnames_zapobj
;
482 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
484 if (dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_CI_DATASET
)
487 err
= zap_remove_norm(mos
, snapobj
, name
, mt
, tx
);
488 if (err
== ENOTSUP
&& (mt
& MT_NORMALIZE
))
489 err
= zap_remove(mos
, snapobj
, name
, tx
);
491 if (err
== 0 && adj_cnt
)
492 dsl_fs_ss_count_adjust(ds
->ds_dir
, -1,
493 DD_FIELD_SNAPSHOT_COUNT
, tx
);
499 dsl_dataset_try_add_ref(dsl_pool_t
*dp
, dsl_dataset_t
*ds
, void *tag
)
501 dmu_buf_t
*dbuf
= ds
->ds_dbuf
;
502 boolean_t result
= B_FALSE
;
504 if (dbuf
!= NULL
&& dmu_buf_try_add_ref(dbuf
, dp
->dp_meta_objset
,
505 ds
->ds_object
, DMU_BONUS_BLKID
, tag
)) {
507 if (ds
== dmu_buf_get_user(dbuf
))
510 dmu_buf_rele(dbuf
, tag
);
517 dsl_dataset_hold_obj(dsl_pool_t
*dp
, uint64_t dsobj
, void *tag
,
520 objset_t
*mos
= dp
->dp_meta_objset
;
524 dmu_object_info_t doi
;
526 ASSERT(dsl_pool_config_held(dp
));
528 err
= dmu_bonus_hold(mos
, dsobj
, tag
, &dbuf
);
532 /* Make sure dsobj has the correct object type. */
533 dmu_object_info_from_db(dbuf
, &doi
);
534 if (doi
.doi_bonus_type
!= DMU_OT_DSL_DATASET
) {
535 dmu_buf_rele(dbuf
, tag
);
536 return (SET_ERROR(EINVAL
));
539 ds
= dmu_buf_get_user(dbuf
);
541 dsl_dataset_t
*winner
= NULL
;
543 ds
= kmem_zalloc(sizeof (dsl_dataset_t
), KM_SLEEP
);
545 ds
->ds_object
= dsobj
;
546 ds
->ds_is_snapshot
= dsl_dataset_phys(ds
)->ds_num_children
!= 0;
547 list_link_init(&ds
->ds_synced_link
);
549 err
= dsl_dir_hold_obj(dp
, dsl_dataset_phys(ds
)->ds_dir_obj
,
550 NULL
, ds
, &ds
->ds_dir
);
552 kmem_free(ds
, sizeof (dsl_dataset_t
));
553 dmu_buf_rele(dbuf
, tag
);
557 mutex_init(&ds
->ds_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
558 mutex_init(&ds
->ds_opening_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
559 mutex_init(&ds
->ds_sendstream_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
560 mutex_init(&ds
->ds_remap_deadlist_lock
,
561 NULL
, MUTEX_DEFAULT
, NULL
);
562 rrw_init(&ds
->ds_bp_rwlock
, B_FALSE
);
563 zfs_refcount_create(&ds
->ds_longholds
);
565 bplist_create(&ds
->ds_pending_deadlist
);
567 list_create(&ds
->ds_sendstreams
, sizeof (dmu_sendarg_t
),
568 offsetof(dmu_sendarg_t
, dsa_link
));
570 list_create(&ds
->ds_prop_cbs
, sizeof (dsl_prop_cb_record_t
),
571 offsetof(dsl_prop_cb_record_t
, cbr_ds_node
));
573 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
576 for (f
= 0; f
< SPA_FEATURES
; f
++) {
577 if (!(spa_feature_table
[f
].fi_flags
&
578 ZFEATURE_FLAG_PER_DATASET
))
580 err
= load_zfeature(mos
, ds
, f
);
584 if (!ds
->ds_is_snapshot
) {
585 ds
->ds_snapname
[0] = '\0';
586 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
587 err
= dsl_dataset_hold_obj(dp
,
588 dsl_dataset_phys(ds
)->ds_prev_snap_obj
,
591 if (doi
.doi_type
== DMU_OTN_ZAP_METADATA
) {
592 int zaperr
= zap_lookup(mos
, ds
->ds_object
,
593 DS_FIELD_BOOKMARK_NAMES
,
594 sizeof (ds
->ds_bookmarks
), 1,
596 if (zaperr
!= ENOENT
)
600 if (zfs_flags
& ZFS_DEBUG_SNAPNAMES
)
601 err
= dsl_dataset_get_snapname(ds
);
603 dsl_dataset_phys(ds
)->ds_userrefs_obj
!= 0) {
605 ds
->ds_dir
->dd_pool
->dp_meta_objset
,
606 dsl_dataset_phys(ds
)->ds_userrefs_obj
,
611 if (err
== 0 && !ds
->ds_is_snapshot
) {
612 err
= dsl_prop_get_int_ds(ds
,
613 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
616 err
= dsl_prop_get_int_ds(ds
,
617 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
621 ds
->ds_reserved
= ds
->ds_quota
= 0;
624 dsl_deadlist_open(&ds
->ds_deadlist
,
625 mos
, dsl_dataset_phys(ds
)->ds_deadlist_obj
);
626 uint64_t remap_deadlist_obj
=
627 dsl_dataset_get_remap_deadlist_object(ds
);
628 if (remap_deadlist_obj
!= 0) {
629 dsl_deadlist_open(&ds
->ds_remap_deadlist
, mos
,
633 dmu_buf_init_user(&ds
->ds_dbu
, dsl_dataset_evict_sync
,
634 dsl_dataset_evict_async
, &ds
->ds_dbuf
);
636 winner
= dmu_buf_set_user_ie(dbuf
, &ds
->ds_dbu
);
638 if (err
!= 0 || winner
!= NULL
) {
639 bplist_destroy(&ds
->ds_pending_deadlist
);
640 dsl_deadlist_close(&ds
->ds_deadlist
);
641 if (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
))
642 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
644 dsl_dataset_rele(ds
->ds_prev
, ds
);
645 dsl_dir_rele(ds
->ds_dir
, ds
);
646 mutex_destroy(&ds
->ds_lock
);
647 mutex_destroy(&ds
->ds_opening_lock
);
648 mutex_destroy(&ds
->ds_sendstream_lock
);
649 zfs_refcount_destroy(&ds
->ds_longholds
);
650 kmem_free(ds
, sizeof (dsl_dataset_t
));
652 dmu_buf_rele(dbuf
, tag
);
658 unique_insert(dsl_dataset_phys(ds
)->ds_fsid_guid
);
659 if (ds
->ds_fsid_guid
!=
660 dsl_dataset_phys(ds
)->ds_fsid_guid
) {
661 zfs_dbgmsg("ds_fsid_guid changed from "
662 "%llx to %llx for pool %s dataset id %llu",
664 dsl_dataset_phys(ds
)->ds_fsid_guid
,
665 (long long)ds
->ds_fsid_guid
,
666 spa_name(dp
->dp_spa
),
672 ASSERT3P(ds
->ds_dbuf
, ==, dbuf
);
673 ASSERT3P(dsl_dataset_phys(ds
), ==, dbuf
->db_data
);
674 ASSERT(dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0 ||
675 spa_version(dp
->dp_spa
) < SPA_VERSION_ORIGIN
||
676 dp
->dp_origin_snap
== NULL
|| ds
== dp
->dp_origin_snap
);
683 dsl_dataset_create_key_mapping(dsl_dataset_t
*ds
)
685 dsl_dir_t
*dd
= ds
->ds_dir
;
687 if (dd
->dd_crypto_obj
== 0)
690 return (spa_keystore_create_mapping(dd
->dd_pool
->dp_spa
,
691 ds
, ds
, &ds
->ds_key_mapping
));
695 dsl_dataset_hold_obj_flags(dsl_pool_t
*dp
, uint64_t dsobj
,
696 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
)
700 err
= dsl_dataset_hold_obj(dp
, dsobj
, tag
, dsp
);
704 ASSERT3P(*dsp
, !=, NULL
);
706 if (flags
& DS_HOLD_FLAG_DECRYPT
) {
707 err
= dsl_dataset_create_key_mapping(*dsp
);
709 dsl_dataset_rele(*dsp
, tag
);
716 dsl_dataset_hold_flags(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
717 void *tag
, dsl_dataset_t
**dsp
)
720 const char *snapname
;
725 err
= dsl_dir_hold(dp
, name
, FTAG
, &dd
, &snapname
);
729 ASSERT(dsl_pool_config_held(dp
));
730 obj
= dsl_dir_phys(dd
)->dd_head_dataset_obj
;
732 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
, &ds
);
734 err
= SET_ERROR(ENOENT
);
736 /* we may be looking for a snapshot */
737 if (err
== 0 && snapname
!= NULL
) {
738 dsl_dataset_t
*snap_ds
;
740 if (*snapname
++ != '@') {
741 dsl_dataset_rele_flags(ds
, flags
, tag
);
742 dsl_dir_rele(dd
, FTAG
);
743 return (SET_ERROR(ENOENT
));
746 dprintf("looking for snapshot '%s'\n", snapname
);
747 err
= dsl_dataset_snap_lookup(ds
, snapname
, &obj
);
749 err
= dsl_dataset_hold_obj_flags(dp
, obj
, flags
, tag
,
752 dsl_dataset_rele_flags(ds
, flags
, tag
);
755 mutex_enter(&snap_ds
->ds_lock
);
756 if (snap_ds
->ds_snapname
[0] == 0)
757 (void) strlcpy(snap_ds
->ds_snapname
, snapname
,
758 sizeof (snap_ds
->ds_snapname
));
759 mutex_exit(&snap_ds
->ds_lock
);
765 dsl_dir_rele(dd
, FTAG
);
770 dsl_dataset_hold(dsl_pool_t
*dp
, const char *name
, void *tag
,
773 return (dsl_dataset_hold_flags(dp
, name
, 0, tag
, dsp
));
777 dsl_dataset_own_obj(dsl_pool_t
*dp
, uint64_t dsobj
, ds_hold_flags_t flags
,
778 void *tag
, dsl_dataset_t
**dsp
)
780 int err
= dsl_dataset_hold_obj_flags(dp
, dsobj
, flags
, tag
, dsp
);
783 if (!dsl_dataset_tryown(*dsp
, tag
)) {
784 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
786 return (SET_ERROR(EBUSY
));
792 dsl_dataset_own(dsl_pool_t
*dp
, const char *name
, ds_hold_flags_t flags
,
793 void *tag
, dsl_dataset_t
**dsp
)
795 int err
= dsl_dataset_hold_flags(dp
, name
, flags
, tag
, dsp
);
798 if (!dsl_dataset_tryown(*dsp
, tag
)) {
799 dsl_dataset_rele_flags(*dsp
, flags
, tag
);
800 return (SET_ERROR(EBUSY
));
806 * See the comment above dsl_pool_hold() for details. In summary, a long
807 * hold is used to prevent destruction of a dataset while the pool hold
808 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
810 * The dataset and pool must be held when this function is called. After it
811 * is called, the pool hold may be released while the dataset is still held
815 dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
)
817 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
818 (void) zfs_refcount_add(&ds
->ds_longholds
, tag
);
822 dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
)
824 (void) zfs_refcount_remove(&ds
->ds_longholds
, tag
);
827 /* Return B_TRUE if there are any long holds on this dataset. */
829 dsl_dataset_long_held(dsl_dataset_t
*ds
)
831 return (!zfs_refcount_is_zero(&ds
->ds_longholds
));
835 dsl_dataset_name(dsl_dataset_t
*ds
, char *name
)
838 (void) strcpy(name
, "mos");
840 dsl_dir_name(ds
->ds_dir
, name
);
841 VERIFY0(dsl_dataset_get_snapname(ds
));
842 if (ds
->ds_snapname
[0]) {
843 VERIFY3U(strlcat(name
, "@", ZFS_MAX_DATASET_NAME_LEN
),
844 <, ZFS_MAX_DATASET_NAME_LEN
);
846 * We use a "recursive" mutex so that we
847 * can call dprintf_ds() with ds_lock held.
849 if (!MUTEX_HELD(&ds
->ds_lock
)) {
850 mutex_enter(&ds
->ds_lock
);
851 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
852 ZFS_MAX_DATASET_NAME_LEN
), <,
853 ZFS_MAX_DATASET_NAME_LEN
);
854 mutex_exit(&ds
->ds_lock
);
856 VERIFY3U(strlcat(name
, ds
->ds_snapname
,
857 ZFS_MAX_DATASET_NAME_LEN
), <,
858 ZFS_MAX_DATASET_NAME_LEN
);
865 dsl_dataset_namelen(dsl_dataset_t
*ds
)
867 VERIFY0(dsl_dataset_get_snapname(ds
));
868 mutex_enter(&ds
->ds_lock
);
869 int len
= strlen(ds
->ds_snapname
);
870 /* add '@' if ds is a snap */
873 len
+= dsl_dir_namelen(ds
->ds_dir
);
874 mutex_exit(&ds
->ds_lock
);
879 dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
)
881 dmu_buf_rele(ds
->ds_dbuf
, tag
);
885 dsl_dataset_remove_key_mapping(dsl_dataset_t
*ds
)
887 dsl_dir_t
*dd
= ds
->ds_dir
;
889 if (dd
== NULL
|| dd
->dd_crypto_obj
== 0)
892 (void) spa_keystore_remove_mapping(dd
->dd_pool
->dp_spa
,
897 dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
899 if (flags
& DS_HOLD_FLAG_DECRYPT
)
900 dsl_dataset_remove_key_mapping(ds
);
902 dsl_dataset_rele(ds
, tag
);
906 dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
)
908 ASSERT3P(ds
->ds_owner
, ==, tag
);
909 ASSERT(ds
->ds_dbuf
!= NULL
);
911 mutex_enter(&ds
->ds_lock
);
913 mutex_exit(&ds
->ds_lock
);
914 dsl_dataset_long_rele(ds
, tag
);
915 dsl_dataset_rele_flags(ds
, flags
, tag
);
919 dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
)
921 boolean_t gotit
= FALSE
;
923 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
924 mutex_enter(&ds
->ds_lock
);
925 if (ds
->ds_owner
== NULL
&& !DS_IS_INCONSISTENT(ds
)) {
927 dsl_dataset_long_hold(ds
, tag
);
930 mutex_exit(&ds
->ds_lock
);
935 dsl_dataset_has_owner(dsl_dataset_t
*ds
)
938 mutex_enter(&ds
->ds_lock
);
939 rv
= (ds
->ds_owner
!= NULL
);
940 mutex_exit(&ds
->ds_lock
);
945 zfeature_active(spa_feature_t f
, void *arg
)
947 switch (spa_feature_table
[f
].fi_type
) {
948 case ZFEATURE_TYPE_BOOLEAN
: {
949 boolean_t val
= (boolean_t
)arg
;
950 ASSERT(val
== B_FALSE
|| val
== B_TRUE
);
953 case ZFEATURE_TYPE_UINT64_ARRAY
:
955 * In this case, arg is a uint64_t array. The feature is active
956 * if the array is non-null.
958 return (arg
!= NULL
);
960 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
966 dsl_dataset_feature_is_active(dsl_dataset_t
*ds
, spa_feature_t f
)
968 return (zfeature_active(f
, ds
->ds_feature
[f
]));
972 * The buffers passed out by this function are references to internal buffers;
973 * they should not be freed by callers of this function, and they should not be
974 * used after the dataset has been released.
977 dsl_dataset_get_uint64_array_feature(dsl_dataset_t
*ds
, spa_feature_t f
,
978 uint64_t *outlength
, uint64_t **outp
)
980 VERIFY(spa_feature_table
[f
].fi_type
& ZFEATURE_TYPE_UINT64_ARRAY
);
981 if (!dsl_dataset_feature_is_active(ds
, f
)) {
984 struct feature_type_uint64_array_arg
*ftuaa
= ds
->ds_feature
[f
];
985 *outp
= ftuaa
->array
;
986 *outlength
= ftuaa
->length
;
991 dsl_dataset_activate_feature(uint64_t dsobj
, spa_feature_t f
, void *arg
,
994 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
995 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
998 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
1000 spa_feature_incr(spa
, f
, tx
);
1001 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1003 switch (spa_feature_table
[f
].fi_type
) {
1004 case ZFEATURE_TYPE_BOOLEAN
:
1005 ASSERT3S((boolean_t
)arg
, ==, B_TRUE
);
1006 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
1007 sizeof (zero
), 1, &zero
, tx
));
1009 case ZFEATURE_TYPE_UINT64_ARRAY
:
1011 struct feature_type_uint64_array_arg
*ftuaa
= arg
;
1012 VERIFY0(zap_add(mos
, dsobj
, spa_feature_table
[f
].fi_guid
,
1013 sizeof (uint64_t), ftuaa
->length
, ftuaa
->array
, tx
));
1017 panic("Invalid zfeature type %d", spa_feature_table
[f
].fi_type
);
1022 dsl_dataset_deactivate_feature_impl(dsl_dataset_t
*ds
, spa_feature_t f
,
1025 spa_t
*spa
= dmu_tx_pool(tx
)->dp_spa
;
1026 objset_t
*mos
= dmu_tx_pool(tx
)->dp_meta_objset
;
1027 uint64_t dsobj
= ds
->ds_object
;
1029 VERIFY(spa_feature_table
[f
].fi_flags
& ZFEATURE_FLAG_PER_DATASET
);
1031 VERIFY0(zap_remove(mos
, dsobj
, spa_feature_table
[f
].fi_guid
, tx
));
1032 spa_feature_decr(spa
, f
, tx
);
1033 ds
->ds_feature
[f
] = NULL
;
1037 dsl_dataset_deactivate_feature(dsl_dataset_t
*ds
, spa_feature_t f
, dmu_tx_t
*tx
)
1039 unload_zfeature(ds
, f
);
1040 dsl_dataset_deactivate_feature_impl(ds
, f
, tx
);
1044 dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
1045 dsl_crypto_params_t
*dcp
, uint64_t flags
, dmu_tx_t
*tx
)
1047 dsl_pool_t
*dp
= dd
->dd_pool
;
1049 dsl_dataset_phys_t
*dsphys
;
1051 objset_t
*mos
= dp
->dp_meta_objset
;
1054 origin
= dp
->dp_origin_snap
;
1056 ASSERT(origin
== NULL
|| origin
->ds_dir
->dd_pool
== dp
);
1057 ASSERT(origin
== NULL
|| dsl_dataset_phys(origin
)->ds_num_children
> 0);
1058 ASSERT(dmu_tx_is_syncing(tx
));
1059 ASSERT(dsl_dir_phys(dd
)->dd_head_dataset_obj
== 0);
1061 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1062 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1063 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1064 dmu_buf_will_dirty(dbuf
, tx
);
1065 dsphys
= dbuf
->db_data
;
1066 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1067 dsphys
->ds_dir_obj
= dd
->dd_object
;
1068 dsphys
->ds_flags
= flags
;
1069 dsphys
->ds_fsid_guid
= unique_create();
1070 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1071 sizeof (dsphys
->ds_guid
));
1072 dsphys
->ds_snapnames_zapobj
=
1073 zap_create_norm(mos
, U8_TEXTPREP_TOUPPER
, DMU_OT_DSL_DS_SNAP_MAP
,
1074 DMU_OT_NONE
, 0, tx
);
1075 dsphys
->ds_creation_time
= gethrestime_sec();
1076 dsphys
->ds_creation_txg
= tx
->tx_txg
== TXG_INITIAL
? 1 : tx
->tx_txg
;
1078 if (origin
== NULL
) {
1079 dsphys
->ds_deadlist_obj
= dsl_deadlist_alloc(mos
, tx
);
1081 dsl_dataset_t
*ohds
; /* head of the origin snapshot */
1083 dsphys
->ds_prev_snap_obj
= origin
->ds_object
;
1084 dsphys
->ds_prev_snap_txg
=
1085 dsl_dataset_phys(origin
)->ds_creation_txg
;
1086 dsphys
->ds_referenced_bytes
=
1087 dsl_dataset_phys(origin
)->ds_referenced_bytes
;
1088 dsphys
->ds_compressed_bytes
=
1089 dsl_dataset_phys(origin
)->ds_compressed_bytes
;
1090 dsphys
->ds_uncompressed_bytes
=
1091 dsl_dataset_phys(origin
)->ds_uncompressed_bytes
;
1092 rrw_enter(&origin
->ds_bp_rwlock
, RW_READER
, FTAG
);
1093 dsphys
->ds_bp
= dsl_dataset_phys(origin
)->ds_bp
;
1094 rrw_exit(&origin
->ds_bp_rwlock
, FTAG
);
1097 * Inherit flags that describe the dataset's contents
1098 * (INCONSISTENT) or properties (Case Insensitive).
1100 dsphys
->ds_flags
|= dsl_dataset_phys(origin
)->ds_flags
&
1101 (DS_FLAG_INCONSISTENT
| DS_FLAG_CI_DATASET
);
1103 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1104 if (zfeature_active(f
, origin
->ds_feature
[f
])) {
1105 dsl_dataset_activate_feature(dsobj
, f
,
1106 origin
->ds_feature
[f
], tx
);
1110 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
1111 dsl_dataset_phys(origin
)->ds_num_children
++;
1113 VERIFY0(dsl_dataset_hold_obj(dp
,
1114 dsl_dir_phys(origin
->ds_dir
)->dd_head_dataset_obj
,
1116 dsphys
->ds_deadlist_obj
= dsl_deadlist_clone(&ohds
->ds_deadlist
,
1117 dsphys
->ds_prev_snap_txg
, dsphys
->ds_prev_snap_obj
, tx
);
1118 dsl_dataset_rele(ohds
, FTAG
);
1120 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_NEXT_CLONES
) {
1121 if (dsl_dataset_phys(origin
)->ds_next_clones_obj
== 0) {
1122 dsl_dataset_phys(origin
)->ds_next_clones_obj
=
1124 DMU_OT_NEXT_CLONES
, DMU_OT_NONE
, 0, tx
);
1126 VERIFY0(zap_add_int(mos
,
1127 dsl_dataset_phys(origin
)->ds_next_clones_obj
,
1131 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
1132 dsl_dir_phys(dd
)->dd_origin_obj
= origin
->ds_object
;
1133 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
1134 if (dsl_dir_phys(origin
->ds_dir
)->dd_clones
== 0) {
1135 dmu_buf_will_dirty(origin
->ds_dir
->dd_dbuf
, tx
);
1136 dsl_dir_phys(origin
->ds_dir
)->dd_clones
=
1138 DMU_OT_DSL_CLONES
, DMU_OT_NONE
, 0, tx
);
1140 VERIFY0(zap_add_int(mos
,
1141 dsl_dir_phys(origin
->ds_dir
)->dd_clones
,
1146 /* handle encryption */
1147 dsl_dataset_create_crypt_sync(dsobj
, dd
, origin
, dcp
, tx
);
1149 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1150 dsphys
->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1152 dmu_buf_rele(dbuf
, FTAG
);
1154 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
1155 dsl_dir_phys(dd
)->dd_head_dataset_obj
= dsobj
;
1161 dsl_dataset_zero_zil(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1165 VERIFY0(dmu_objset_from_ds(ds
, &os
));
1166 if (bcmp(&os
->os_zil_header
, &zero_zil
, sizeof (zero_zil
)) != 0) {
1167 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1170 bzero(&os
->os_zil_header
, sizeof (os
->os_zil_header
));
1171 if (os
->os_encrypted
)
1172 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_TRUE
;
1174 zio
= zio_root(dp
->dp_spa
, NULL
, NULL
, ZIO_FLAG_MUSTSUCCEED
);
1175 dsl_dataset_sync(ds
, zio
, tx
);
1176 VERIFY0(zio_wait(zio
));
1178 /* dsl_dataset_sync_done will drop this reference. */
1179 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1180 dsl_dataset_sync_done(ds
, tx
);
1185 dsl_dataset_create_sync(dsl_dir_t
*pdd
, const char *lastname
,
1186 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*cr
,
1187 dsl_crypto_params_t
*dcp
, dmu_tx_t
*tx
)
1189 dsl_pool_t
*dp
= pdd
->dd_pool
;
1190 uint64_t dsobj
, ddobj
;
1193 ASSERT(dmu_tx_is_syncing(tx
));
1194 ASSERT(lastname
[0] != '@');
1196 ddobj
= dsl_dir_create_sync(dp
, pdd
, lastname
, tx
);
1197 VERIFY0(dsl_dir_hold_obj(dp
, ddobj
, lastname
, FTAG
, &dd
));
1199 dsobj
= dsl_dataset_create_sync_dd(dd
, origin
, dcp
,
1200 flags
& ~DS_CREATE_FLAG_NODIRTY
, tx
);
1202 dsl_deleg_set_create_perms(dd
, tx
, cr
);
1205 * Since we're creating a new node we know it's a leaf, so we can
1206 * initialize the counts if the limit feature is active.
1208 if (spa_feature_is_active(dp
->dp_spa
, SPA_FEATURE_FS_SS_LIMIT
)) {
1210 objset_t
*os
= dd
->dd_pool
->dp_meta_objset
;
1212 dsl_dir_zapify(dd
, tx
);
1213 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_FILESYSTEM_COUNT
,
1214 sizeof (cnt
), 1, &cnt
, tx
));
1215 VERIFY0(zap_add(os
, dd
->dd_object
, DD_FIELD_SNAPSHOT_COUNT
,
1216 sizeof (cnt
), 1, &cnt
, tx
));
1219 dsl_dir_rele(dd
, FTAG
);
1222 * If we are creating a clone, make sure we zero out any stale
1223 * data from the origin snapshots zil header.
1225 if (origin
!= NULL
&& !(flags
& DS_CREATE_FLAG_NODIRTY
)) {
1228 VERIFY0(dsl_dataset_hold_obj(dp
, dsobj
, FTAG
, &ds
));
1229 dsl_dataset_zero_zil(ds
, tx
);
1230 dsl_dataset_rele(ds
, FTAG
);
1237 * The unique space in the head dataset can be calculated by subtracting
1238 * the space used in the most recent snapshot, that is still being used
1239 * in this file system, from the space currently in use. To figure out
1240 * the space in the most recent snapshot still in use, we need to take
1241 * the total space used in the snapshot and subtract out the space that
1242 * has been freed up since the snapshot was taken.
1245 dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
)
1248 uint64_t dlused
, dlcomp
, dluncomp
;
1250 ASSERT(!ds
->ds_is_snapshot
);
1252 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0)
1253 mrs_used
= dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
;
1257 dsl_deadlist_space(&ds
->ds_deadlist
, &dlused
, &dlcomp
, &dluncomp
);
1259 ASSERT3U(dlused
, <=, mrs_used
);
1260 dsl_dataset_phys(ds
)->ds_unique_bytes
=
1261 dsl_dataset_phys(ds
)->ds_referenced_bytes
- (mrs_used
- dlused
);
1263 if (spa_version(ds
->ds_dir
->dd_pool
->dp_spa
) >=
1264 SPA_VERSION_UNIQUE_ACCURATE
)
1265 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1269 dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
1272 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
1273 ASSERTV(uint64_t count
);
1276 ASSERT(dsl_dataset_phys(ds
)->ds_num_children
>= 2);
1277 err
= zap_remove_int(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1280 * The err should not be ENOENT, but a bug in a previous version
1281 * of the code could cause upgrade_clones_cb() to not set
1282 * ds_next_snap_obj when it should, leading to a missing entry.
1283 * If we knew that the pool was created after
1284 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
1285 * ENOENT. However, at least we can check that we don't have
1286 * too many entries in the next_clones_obj even after failing to
1291 ASSERT0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
1293 ASSERT3U(count
, <=, dsl_dataset_phys(ds
)->ds_num_children
- 2);
1298 dsl_dataset_get_blkptr(dsl_dataset_t
*ds
)
1300 return (&dsl_dataset_phys(ds
)->ds_bp
);
1304 dsl_dataset_get_spa(dsl_dataset_t
*ds
)
1306 return (ds
->ds_dir
->dd_pool
->dp_spa
);
1310 dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1314 if (ds
== NULL
) /* this is the meta-objset */
1317 ASSERT(ds
->ds_objset
!= NULL
);
1319 if (dsl_dataset_phys(ds
)->ds_next_snap_obj
!= 0)
1320 panic("dirtying snapshot!");
1322 /* Must not dirty a dataset in the same txg where it got snapshotted. */
1323 ASSERT3U(tx
->tx_txg
, >, dsl_dataset_phys(ds
)->ds_prev_snap_txg
);
1325 dp
= ds
->ds_dir
->dd_pool
;
1326 if (txg_list_add(&dp
->dp_dirty_datasets
, ds
, tx
->tx_txg
)) {
1327 objset_t
*os
= ds
->ds_objset
;
1329 /* up the hold count until we can be written out */
1330 dmu_buf_add_ref(ds
->ds_dbuf
, ds
);
1332 /* if this dataset is encrypted, grab a reference to the DCK */
1333 if (ds
->ds_dir
->dd_crypto_obj
!= 0 &&
1334 !os
->os_raw_receive
&&
1335 !os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]) {
1336 ASSERT3P(ds
->ds_key_mapping
, !=, NULL
);
1337 key_mapping_add_ref(ds
->ds_key_mapping
, ds
);
1343 dsl_dataset_snapshot_reserve_space(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1347 if (!dmu_tx_is_syncing(tx
))
1351 * If there's an fs-only reservation, any blocks that might become
1352 * owned by the snapshot dataset must be accommodated by space
1353 * outside of the reservation.
1355 ASSERT(ds
->ds_reserved
== 0 || DS_UNIQUE_IS_ACCURATE(ds
));
1356 asize
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
, ds
->ds_reserved
);
1357 if (asize
> dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
))
1358 return (SET_ERROR(ENOSPC
));
1361 * Propagate any reserved space for this snapshot to other
1362 * snapshot checks in this sync group.
1365 dsl_dir_willuse_space(ds
->ds_dir
, asize
, tx
);
1371 dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
1372 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
)
1377 ds
->ds_trysnap_txg
= tx
->tx_txg
;
1379 if (!dmu_tx_is_syncing(tx
))
1383 * We don't allow multiple snapshots of the same txg. If there
1384 * is already one, try again.
1386 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
)
1387 return (SET_ERROR(EAGAIN
));
1390 * Check for conflicting snapshot name.
1392 error
= dsl_dataset_snap_lookup(ds
, snapname
, &value
);
1394 return (SET_ERROR(EEXIST
));
1395 if (error
!= ENOENT
)
1399 * We don't allow taking snapshots of inconsistent datasets, such as
1400 * those into which we are currently receiving. However, if we are
1401 * creating this snapshot as part of a receive, this check will be
1402 * executed atomically with respect to the completion of the receive
1403 * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
1404 * case we ignore this, knowing it will be fixed up for us shortly in
1405 * dmu_recv_end_sync().
1407 if (!recv
&& DS_IS_INCONSISTENT(ds
))
1408 return (SET_ERROR(EBUSY
));
1411 * Skip the check for temporary snapshots or if we have already checked
1412 * the counts in dsl_dataset_snapshot_check. This means we really only
1413 * check the count here when we're receiving a stream.
1415 if (cnt
!= 0 && cr
!= NULL
) {
1416 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1417 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
, cr
);
1422 error
= dsl_dataset_snapshot_reserve_space(ds
, tx
);
1430 dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
)
1432 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1433 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1438 * Pre-compute how many total new snapshots will be created for each
1439 * level in the tree and below. This is needed for validating the
1440 * snapshot limit when either taking a recursive snapshot or when
1441 * taking multiple snapshots.
1443 * The problem is that the counts are not actually adjusted when
1444 * we are checking, only when we finally sync. For a single snapshot,
1445 * this is easy, the count will increase by 1 at each node up the tree,
1446 * but its more complicated for the recursive/multiple snapshot case.
1448 * The dsl_fs_ss_limit_check function does recursively check the count
1449 * at each level up the tree but since it is validating each snapshot
1450 * independently we need to be sure that we are validating the complete
1451 * count for the entire set of snapshots. We do this by rolling up the
1452 * counts for each component of the name into an nvlist and then
1453 * checking each of those cases with the aggregated count.
1455 * This approach properly handles not only the recursive snapshot
1456 * case (where we get all of those on the ddsa_snaps list) but also
1457 * the sibling case (e.g. snapshot a/b and a/c so that we will also
1458 * validate the limit on 'a' using a count of 2).
1460 * We validate the snapshot names in the third loop and only report
1463 if (dmu_tx_is_syncing(tx
)) {
1465 nvlist_t
*cnt_track
= NULL
;
1466 cnt_track
= fnvlist_alloc();
1468 nm
= kmem_alloc(MAXPATHLEN
, KM_SLEEP
);
1470 /* Rollup aggregated counts into the cnt_track list */
1471 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1473 pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1477 (void) strlcpy(nm
, nvpair_name(pair
), MAXPATHLEN
);
1478 pdelim
= strchr(nm
, '@');
1484 if (nvlist_lookup_uint64(cnt_track
, nm
,
1486 /* update existing entry */
1487 fnvlist_add_uint64(cnt_track
, nm
,
1491 fnvlist_add_uint64(cnt_track
, nm
, 1);
1494 pdelim
= strrchr(nm
, '/');
1497 } while (pdelim
!= NULL
);
1500 kmem_free(nm
, MAXPATHLEN
);
1502 /* Check aggregated counts at each level */
1503 for (pair
= nvlist_next_nvpair(cnt_track
, NULL
);
1504 pair
!= NULL
; pair
= nvlist_next_nvpair(cnt_track
, pair
)) {
1510 name
= nvpair_name(pair
);
1511 cnt
= fnvpair_value_uint64(pair
);
1514 error
= dsl_dataset_hold(dp
, name
, FTAG
, &ds
);
1516 error
= dsl_fs_ss_limit_check(ds
->ds_dir
, cnt
,
1517 ZFS_PROP_SNAPSHOT_LIMIT
, NULL
,
1519 dsl_dataset_rele(ds
, FTAG
);
1523 if (ddsa
->ddsa_errors
!= NULL
)
1524 fnvlist_add_int32(ddsa
->ddsa_errors
,
1527 /* only report one error for this check */
1531 nvlist_free(cnt_track
);
1534 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1535 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1538 char *name
, *atp
= NULL
;
1539 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1541 name
= nvpair_name(pair
);
1542 if (strlen(name
) >= ZFS_MAX_DATASET_NAME_LEN
)
1543 error
= SET_ERROR(ENAMETOOLONG
);
1545 atp
= strchr(name
, '@');
1547 error
= SET_ERROR(EINVAL
);
1549 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1552 error
= dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
);
1554 /* passing 0/NULL skips dsl_fs_ss_limit_check */
1555 error
= dsl_dataset_snapshot_check_impl(ds
,
1556 atp
+ 1, tx
, B_FALSE
, 0, NULL
);
1557 dsl_dataset_rele(ds
, FTAG
);
1561 if (ddsa
->ddsa_errors
!= NULL
) {
1562 fnvlist_add_int32(ddsa
->ddsa_errors
,
1573 dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
1576 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
1578 dsl_dataset_phys_t
*dsphys
;
1579 uint64_t dsobj
, crtxg
;
1580 objset_t
*mos
= dp
->dp_meta_objset
;
1581 ASSERTV(static zil_header_t zero_zil
);
1582 ASSERTV(objset_t
*os
);
1584 ASSERT(RRW_WRITE_HELD(&dp
->dp_config_rwlock
));
1587 * If we are on an old pool, the zil must not be active, in which
1588 * case it will be zeroed. Usually zil_suspend() accomplishes this.
1590 ASSERT(spa_version(dmu_tx_pool(tx
)->dp_spa
) >= SPA_VERSION_FAST_SNAP
||
1591 dmu_objset_from_ds(ds
, &os
) != 0 ||
1592 bcmp(&os
->os_phys
->os_zil_header
, &zero_zil
,
1593 sizeof (zero_zil
)) == 0);
1595 /* Should not snapshot a dirty dataset. */
1596 ASSERT(!txg_list_member(&ds
->ds_dir
->dd_pool
->dp_dirty_datasets
,
1599 dsl_fs_ss_count_adjust(ds
->ds_dir
, 1, DD_FIELD_SNAPSHOT_COUNT
, tx
);
1602 * The origin's ds_creation_txg has to be < TXG_INITIAL
1604 if (strcmp(snapname
, ORIGIN_DIR_NAME
) == 0)
1609 dsobj
= dmu_object_alloc(mos
, DMU_OT_DSL_DATASET
, 0,
1610 DMU_OT_DSL_DATASET
, sizeof (dsl_dataset_phys_t
), tx
);
1611 VERIFY0(dmu_bonus_hold(mos
, dsobj
, FTAG
, &dbuf
));
1612 dmu_buf_will_dirty(dbuf
, tx
);
1613 dsphys
= dbuf
->db_data
;
1614 bzero(dsphys
, sizeof (dsl_dataset_phys_t
));
1615 dsphys
->ds_dir_obj
= ds
->ds_dir
->dd_object
;
1616 dsphys
->ds_fsid_guid
= unique_create();
1617 (void) random_get_pseudo_bytes((void*)&dsphys
->ds_guid
,
1618 sizeof (dsphys
->ds_guid
));
1619 dsphys
->ds_prev_snap_obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
1620 dsphys
->ds_prev_snap_txg
= dsl_dataset_phys(ds
)->ds_prev_snap_txg
;
1621 dsphys
->ds_next_snap_obj
= ds
->ds_object
;
1622 dsphys
->ds_num_children
= 1;
1623 dsphys
->ds_creation_time
= gethrestime_sec();
1624 dsphys
->ds_creation_txg
= crtxg
;
1625 dsphys
->ds_deadlist_obj
= dsl_dataset_phys(ds
)->ds_deadlist_obj
;
1626 dsphys
->ds_referenced_bytes
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
1627 dsphys
->ds_compressed_bytes
= dsl_dataset_phys(ds
)->ds_compressed_bytes
;
1628 dsphys
->ds_uncompressed_bytes
=
1629 dsl_dataset_phys(ds
)->ds_uncompressed_bytes
;
1630 dsphys
->ds_flags
= dsl_dataset_phys(ds
)->ds_flags
;
1631 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
1632 dsphys
->ds_bp
= dsl_dataset_phys(ds
)->ds_bp
;
1633 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
1634 dmu_buf_rele(dbuf
, FTAG
);
1636 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1637 if (zfeature_active(f
, ds
->ds_feature
[f
])) {
1638 dsl_dataset_activate_feature(dsobj
, f
,
1639 ds
->ds_feature
[f
], tx
);
1643 ASSERT3U(ds
->ds_prev
!= 0, ==,
1644 dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0);
1646 uint64_t next_clones_obj
=
1647 dsl_dataset_phys(ds
->ds_prev
)->ds_next_clones_obj
;
1648 ASSERT(dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1650 dsl_dataset_phys(ds
->ds_prev
)->ds_num_children
> 1);
1651 if (dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
==
1653 dmu_buf_will_dirty(ds
->ds_prev
->ds_dbuf
, tx
);
1654 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, ==,
1655 dsl_dataset_phys(ds
->ds_prev
)->ds_creation_txg
);
1656 dsl_dataset_phys(ds
->ds_prev
)->ds_next_snap_obj
= dsobj
;
1657 } else if (next_clones_obj
!= 0) {
1658 dsl_dataset_remove_from_next_clones(ds
->ds_prev
,
1659 dsphys
->ds_next_snap_obj
, tx
);
1660 VERIFY0(zap_add_int(mos
,
1661 next_clones_obj
, dsobj
, tx
));
1666 * If we have a reference-reservation on this dataset, we will
1667 * need to increase the amount of refreservation being charged
1668 * since our unique space is going to zero.
1670 if (ds
->ds_reserved
) {
1672 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
1673 delta
= MIN(dsl_dataset_phys(ds
)->ds_unique_bytes
,
1675 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
,
1679 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1680 dsl_dataset_phys(ds
)->ds_deadlist_obj
=
1681 dsl_deadlist_clone(&ds
->ds_deadlist
, UINT64_MAX
,
1682 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
1683 dsl_deadlist_close(&ds
->ds_deadlist
);
1684 dsl_deadlist_open(&ds
->ds_deadlist
, mos
,
1685 dsl_dataset_phys(ds
)->ds_deadlist_obj
);
1686 dsl_deadlist_add_key(&ds
->ds_deadlist
,
1687 dsl_dataset_phys(ds
)->ds_prev_snap_txg
, tx
);
1689 if (dsl_dataset_remap_deadlist_exists(ds
)) {
1690 uint64_t remap_deadlist_obj
=
1691 dsl_dataset_get_remap_deadlist_object(ds
);
1693 * Move the remap_deadlist to the snapshot. The head
1694 * will create a new remap deadlist on demand, from
1695 * dsl_dataset_block_remapped().
1697 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
1698 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
1700 dmu_object_zapify(mos
, dsobj
, DMU_OT_DSL_DATASET
, tx
);
1701 VERIFY0(zap_add(mos
, dsobj
, DS_FIELD_REMAP_DEADLIST
,
1702 sizeof (remap_deadlist_obj
), 1, &remap_deadlist_obj
, tx
));
1705 ASSERT3U(dsl_dataset_phys(ds
)->ds_prev_snap_txg
, <, tx
->tx_txg
);
1706 dsl_dataset_phys(ds
)->ds_prev_snap_obj
= dsobj
;
1707 dsl_dataset_phys(ds
)->ds_prev_snap_txg
= crtxg
;
1708 dsl_dataset_phys(ds
)->ds_unique_bytes
= 0;
1710 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_UNIQUE_ACCURATE
)
1711 dsl_dataset_phys(ds
)->ds_flags
|= DS_FLAG_UNIQUE_ACCURATE
;
1713 VERIFY0(zap_add(mos
, dsl_dataset_phys(ds
)->ds_snapnames_zapobj
,
1714 snapname
, 8, 1, &dsobj
, tx
));
1717 dsl_dataset_rele(ds
->ds_prev
, ds
);
1718 VERIFY0(dsl_dataset_hold_obj(dp
,
1719 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, ds
, &ds
->ds_prev
));
1721 dsl_scan_ds_snapshotted(ds
, tx
);
1723 dsl_dir_snap_cmtime_update(ds
->ds_dir
);
1725 spa_history_log_internal_ds(ds
->ds_prev
, "snapshot", tx
, "");
1729 dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
1731 dsl_dataset_snapshot_arg_t
*ddsa
= arg
;
1732 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1735 for (pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, NULL
);
1736 pair
!= NULL
; pair
= nvlist_next_nvpair(ddsa
->ddsa_snaps
, pair
)) {
1739 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
1741 name
= nvpair_name(pair
);
1742 atp
= strchr(name
, '@');
1743 (void) strlcpy(dsname
, name
, atp
- name
+ 1);
1744 VERIFY0(dsl_dataset_hold(dp
, dsname
, FTAG
, &ds
));
1746 dsl_dataset_snapshot_sync_impl(ds
, atp
+ 1, tx
);
1747 if (ddsa
->ddsa_props
!= NULL
) {
1748 dsl_props_set_sync_impl(ds
->ds_prev
,
1749 ZPROP_SRC_LOCAL
, ddsa
->ddsa_props
, tx
);
1751 zvol_create_minors(dp
->dp_spa
, nvpair_name(pair
), B_TRUE
);
1752 dsl_dataset_rele(ds
, FTAG
);
1757 * The snapshots must all be in the same pool.
1758 * All-or-nothing: if there are any failures, nothing will be modified.
1761 dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
)
1763 dsl_dataset_snapshot_arg_t ddsa
;
1765 boolean_t needsuspend
;
1769 nvlist_t
*suspended
= NULL
;
1771 pair
= nvlist_next_nvpair(snaps
, NULL
);
1774 firstname
= nvpair_name(pair
);
1776 error
= spa_open(firstname
, &spa
, FTAG
);
1779 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1780 spa_close(spa
, FTAG
);
1783 suspended
= fnvlist_alloc();
1784 for (pair
= nvlist_next_nvpair(snaps
, NULL
); pair
!= NULL
;
1785 pair
= nvlist_next_nvpair(snaps
, pair
)) {
1786 char fsname
[ZFS_MAX_DATASET_NAME_LEN
];
1787 char *snapname
= nvpair_name(pair
);
1791 atp
= strchr(snapname
, '@');
1793 error
= SET_ERROR(EINVAL
);
1796 (void) strlcpy(fsname
, snapname
, atp
- snapname
+ 1);
1798 error
= zil_suspend(fsname
, &cookie
);
1801 fnvlist_add_uint64(suspended
, fsname
,
1806 ddsa
.ddsa_snaps
= snaps
;
1807 ddsa
.ddsa_props
= props
;
1808 ddsa
.ddsa_errors
= errors
;
1809 ddsa
.ddsa_cr
= CRED();
1812 error
= dsl_sync_task(firstname
, dsl_dataset_snapshot_check
,
1813 dsl_dataset_snapshot_sync
, &ddsa
,
1814 fnvlist_num_pairs(snaps
) * 3, ZFS_SPACE_CHECK_NORMAL
);
1817 if (suspended
!= NULL
) {
1818 for (pair
= nvlist_next_nvpair(suspended
, NULL
); pair
!= NULL
;
1819 pair
= nvlist_next_nvpair(suspended
, pair
)) {
1820 zil_resume((void *)(uintptr_t)
1821 fnvpair_value_uint64(pair
));
1823 fnvlist_free(suspended
);
1829 typedef struct dsl_dataset_snapshot_tmp_arg
{
1830 const char *ddsta_fsname
;
1831 const char *ddsta_snapname
;
1832 minor_t ddsta_cleanup_minor
;
1833 const char *ddsta_htag
;
1834 } dsl_dataset_snapshot_tmp_arg_t
;
1837 dsl_dataset_snapshot_tmp_check(void *arg
, dmu_tx_t
*tx
)
1839 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1840 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1844 error
= dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
);
1848 /* NULL cred means no limit check for tmp snapshot */
1849 error
= dsl_dataset_snapshot_check_impl(ds
, ddsta
->ddsta_snapname
,
1850 tx
, B_FALSE
, 0, NULL
);
1852 dsl_dataset_rele(ds
, FTAG
);
1856 if (spa_version(dp
->dp_spa
) < SPA_VERSION_USERREFS
) {
1857 dsl_dataset_rele(ds
, FTAG
);
1858 return (SET_ERROR(ENOTSUP
));
1860 error
= dsl_dataset_user_hold_check_one(NULL
, ddsta
->ddsta_htag
,
1863 dsl_dataset_rele(ds
, FTAG
);
1867 dsl_dataset_rele(ds
, FTAG
);
1872 dsl_dataset_snapshot_tmp_sync(void *arg
, dmu_tx_t
*tx
)
1874 dsl_dataset_snapshot_tmp_arg_t
*ddsta
= arg
;
1875 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
1876 dsl_dataset_t
*ds
= NULL
;
1878 VERIFY0(dsl_dataset_hold(dp
, ddsta
->ddsta_fsname
, FTAG
, &ds
));
1880 dsl_dataset_snapshot_sync_impl(ds
, ddsta
->ddsta_snapname
, tx
);
1881 dsl_dataset_user_hold_sync_one(ds
->ds_prev
, ddsta
->ddsta_htag
,
1882 ddsta
->ddsta_cleanup_minor
, gethrestime_sec(), tx
);
1883 dsl_destroy_snapshot_sync_impl(ds
->ds_prev
, B_TRUE
, tx
);
1885 dsl_dataset_rele(ds
, FTAG
);
1889 dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
1890 minor_t cleanup_minor
, const char *htag
)
1892 dsl_dataset_snapshot_tmp_arg_t ddsta
;
1895 boolean_t needsuspend
;
1898 ddsta
.ddsta_fsname
= fsname
;
1899 ddsta
.ddsta_snapname
= snapname
;
1900 ddsta
.ddsta_cleanup_minor
= cleanup_minor
;
1901 ddsta
.ddsta_htag
= htag
;
1903 error
= spa_open(fsname
, &spa
, FTAG
);
1906 needsuspend
= (spa_version(spa
) < SPA_VERSION_FAST_SNAP
);
1907 spa_close(spa
, FTAG
);
1910 error
= zil_suspend(fsname
, &cookie
);
1915 error
= dsl_sync_task(fsname
, dsl_dataset_snapshot_tmp_check
,
1916 dsl_dataset_snapshot_tmp_sync
, &ddsta
, 3, ZFS_SPACE_CHECK_RESERVED
);
1924 dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
)
1926 ASSERT(dmu_tx_is_syncing(tx
));
1927 ASSERT(ds
->ds_objset
!= NULL
);
1928 ASSERT(dsl_dataset_phys(ds
)->ds_next_snap_obj
== 0);
1931 * in case we had to change ds_fsid_guid when we opened it,
1934 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
1935 dsl_dataset_phys(ds
)->ds_fsid_guid
= ds
->ds_fsid_guid
;
1937 if (ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] != 0) {
1938 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1939 ds
->ds_object
, DS_FIELD_RESUME_OBJECT
, 8, 1,
1940 &ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
], tx
));
1941 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1942 ds
->ds_object
, DS_FIELD_RESUME_OFFSET
, 8, 1,
1943 &ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
], tx
));
1944 VERIFY0(zap_update(tx
->tx_pool
->dp_meta_objset
,
1945 ds
->ds_object
, DS_FIELD_RESUME_BYTES
, 8, 1,
1946 &ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
], tx
));
1947 ds
->ds_resume_object
[tx
->tx_txg
& TXG_MASK
] = 0;
1948 ds
->ds_resume_offset
[tx
->tx_txg
& TXG_MASK
] = 0;
1949 ds
->ds_resume_bytes
[tx
->tx_txg
& TXG_MASK
] = 0;
1952 dmu_objset_sync(ds
->ds_objset
, zio
, tx
);
1954 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
1955 if (zfeature_active(f
, ds
->ds_feature_activation
[f
])) {
1956 if (zfeature_active(f
, ds
->ds_feature
[f
]))
1958 dsl_dataset_activate_feature(ds
->ds_object
, f
,
1959 ds
->ds_feature_activation
[f
], tx
);
1960 ds
->ds_feature
[f
] = ds
->ds_feature_activation
[f
];
1966 deadlist_enqueue_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
1968 dsl_deadlist_t
*dl
= arg
;
1969 dsl_deadlist_insert(dl
, bp
, tx
);
1974 dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
1976 objset_t
*os
= ds
->ds_objset
;
1978 bplist_iterate(&ds
->ds_pending_deadlist
,
1979 deadlist_enqueue_cb
, &ds
->ds_deadlist
, tx
);
1981 if (os
->os_synced_dnodes
!= NULL
) {
1982 multilist_destroy(os
->os_synced_dnodes
);
1983 os
->os_synced_dnodes
= NULL
;
1986 if (os
->os_encrypted
)
1987 os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
] = B_FALSE
;
1989 ASSERT0(os
->os_next_write_raw
[tx
->tx_txg
& TXG_MASK
]);
1991 ASSERT(!dmu_objset_is_dirty(os
, dmu_tx_get_txg(tx
)));
1993 dmu_buf_rele(ds
->ds_dbuf
, ds
);
1997 get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
)
2000 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
2004 ASSERT(dsl_pool_config_held(ds
->ds_dir
->dd_pool
));
2007 * There may be missing entries in ds_next_clones_obj
2008 * due to a bug in a previous version of the code.
2009 * Only trust it if it has the right number of entries.
2011 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
!= 0) {
2012 VERIFY0(zap_count(mos
, dsl_dataset_phys(ds
)->ds_next_clones_obj
,
2015 if (count
!= dsl_dataset_phys(ds
)->ds_num_children
- 1) {
2018 for (zap_cursor_init(&zc
, mos
,
2019 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
2020 zap_cursor_retrieve(&zc
, &za
) == 0;
2021 zap_cursor_advance(&zc
)) {
2022 dsl_dataset_t
*clone
;
2023 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2024 VERIFY0(dsl_dataset_hold_obj(ds
->ds_dir
->dd_pool
,
2025 za
.za_first_integer
, FTAG
, &clone
));
2026 dsl_dir_name(clone
->ds_dir
, buf
);
2027 fnvlist_add_boolean(val
, buf
);
2028 dsl_dataset_rele(clone
, FTAG
);
2030 zap_cursor_fini(&zc
);
2035 get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2037 nvlist_t
*propval
= fnvlist_alloc();
2041 * We use nvlist_alloc() instead of fnvlist_alloc() because the
2042 * latter would allocate the list with NV_UNIQUE_NAME flag.
2043 * As a result, every time a clone name is appended to the list
2044 * it would be (linearly) searched for for a duplicate name.
2045 * We already know that all clone names must be unique and we
2046 * want avoid the quadratic complexity of double-checking that
2047 * because we can have a large number of clones.
2049 VERIFY0(nvlist_alloc(&val
, 0, KM_SLEEP
));
2051 if (get_clones_stat_impl(ds
, val
) == 0) {
2052 fnvlist_add_nvlist(propval
, ZPROP_VALUE
, val
);
2053 fnvlist_add_nvlist(nv
, zfs_prop_to_name(ZFS_PROP_CLONES
),
2058 nvlist_free(propval
);
2062 * Returns a string that represents the receive resume stats token. It should
2063 * be freed with strfree().
2066 get_receive_resume_stats_impl(dsl_dataset_t
*ds
)
2068 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2070 if (dsl_dataset_has_resume_receive_state(ds
)) {
2073 uint8_t *compressed
;
2075 nvlist_t
*token_nv
= fnvlist_alloc();
2076 size_t packed_size
, compressed_size
;
2078 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2079 DS_FIELD_RESUME_FROMGUID
, sizeof (val
), 1, &val
) == 0) {
2080 fnvlist_add_uint64(token_nv
, "fromguid", val
);
2082 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2083 DS_FIELD_RESUME_OBJECT
, sizeof (val
), 1, &val
) == 0) {
2084 fnvlist_add_uint64(token_nv
, "object", val
);
2086 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2087 DS_FIELD_RESUME_OFFSET
, sizeof (val
), 1, &val
) == 0) {
2088 fnvlist_add_uint64(token_nv
, "offset", val
);
2090 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2091 DS_FIELD_RESUME_BYTES
, sizeof (val
), 1, &val
) == 0) {
2092 fnvlist_add_uint64(token_nv
, "bytes", val
);
2094 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2095 DS_FIELD_RESUME_TOGUID
, sizeof (val
), 1, &val
) == 0) {
2096 fnvlist_add_uint64(token_nv
, "toguid", val
);
2098 char buf
[MAXNAMELEN
];
2099 if (zap_lookup(dp
->dp_meta_objset
, ds
->ds_object
,
2100 DS_FIELD_RESUME_TONAME
, 1, sizeof (buf
), buf
) == 0) {
2101 fnvlist_add_string(token_nv
, "toname", buf
);
2103 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2104 DS_FIELD_RESUME_LARGEBLOCK
) == 0) {
2105 fnvlist_add_boolean(token_nv
, "largeblockok");
2107 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2108 DS_FIELD_RESUME_EMBEDOK
) == 0) {
2109 fnvlist_add_boolean(token_nv
, "embedok");
2111 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2112 DS_FIELD_RESUME_COMPRESSOK
) == 0) {
2113 fnvlist_add_boolean(token_nv
, "compressok");
2115 if (zap_contains(dp
->dp_meta_objset
, ds
->ds_object
,
2116 DS_FIELD_RESUME_RAWOK
) == 0) {
2117 fnvlist_add_boolean(token_nv
, "rawok");
2119 packed
= fnvlist_pack(token_nv
, &packed_size
);
2120 fnvlist_free(token_nv
);
2121 compressed
= kmem_alloc(packed_size
, KM_SLEEP
);
2123 compressed_size
= gzip_compress(packed
, compressed
,
2124 packed_size
, packed_size
, 6);
2127 fletcher_4_native_varsize(compressed
, compressed_size
, &cksum
);
2129 str
= kmem_alloc(compressed_size
* 2 + 1, KM_SLEEP
);
2130 for (int i
= 0; i
< compressed_size
; i
++) {
2131 (void) sprintf(str
+ i
* 2, "%02x", compressed
[i
]);
2133 str
[compressed_size
* 2] = '\0';
2134 char *propval
= kmem_asprintf("%u-%llx-%llx-%s",
2135 ZFS_SEND_RESUME_TOKEN_VERSION
,
2136 (longlong_t
)cksum
.zc_word
[0],
2137 (longlong_t
)packed_size
, str
);
2138 kmem_free(packed
, packed_size
);
2139 kmem_free(str
, compressed_size
* 2 + 1);
2140 kmem_free(compressed
, packed_size
);
2143 return (strdup(""));
2147 * Returns a string that represents the receive resume stats token of the
2148 * dataset's child. It should be freed with strfree().
2151 get_child_receive_stats(dsl_dataset_t
*ds
)
2153 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2154 dsl_dataset_t
*recv_ds
;
2155 dsl_dataset_name(ds
, recvname
);
2156 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2157 sizeof (recvname
) &&
2158 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2159 sizeof (recvname
) &&
2160 dsl_dataset_hold(ds
->ds_dir
->dd_pool
, recvname
, FTAG
,
2162 char *propval
= get_receive_resume_stats_impl(recv_ds
);
2163 dsl_dataset_rele(recv_ds
, FTAG
);
2166 return (strdup(""));
2170 get_receive_resume_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2172 char *propval
= get_receive_resume_stats_impl(ds
);
2173 if (strcmp(propval
, "") != 0) {
2174 dsl_prop_nvlist_add_string(nv
,
2175 ZFS_PROP_RECEIVE_RESUME_TOKEN
, propval
);
2177 char *childval
= get_child_receive_stats(ds
);
2178 if (strcmp(childval
, "") != 0) {
2179 dsl_prop_nvlist_add_string(nv
,
2180 ZFS_PROP_RECEIVE_RESUME_TOKEN
, childval
);
2188 dsl_get_refratio(dsl_dataset_t
*ds
)
2190 uint64_t ratio
= dsl_dataset_phys(ds
)->ds_compressed_bytes
== 0 ? 100 :
2191 (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
* 100 /
2192 dsl_dataset_phys(ds
)->ds_compressed_bytes
);
2197 dsl_get_logicalreferenced(dsl_dataset_t
*ds
)
2199 return (dsl_dataset_phys(ds
)->ds_uncompressed_bytes
);
2203 dsl_get_compressratio(dsl_dataset_t
*ds
)
2205 if (ds
->ds_is_snapshot
) {
2206 return (dsl_get_refratio(ds
));
2208 dsl_dir_t
*dd
= ds
->ds_dir
;
2209 mutex_enter(&dd
->dd_lock
);
2210 uint64_t val
= dsl_dir_get_compressratio(dd
);
2211 mutex_exit(&dd
->dd_lock
);
2217 dsl_get_used(dsl_dataset_t
*ds
)
2219 if (ds
->ds_is_snapshot
) {
2220 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2222 dsl_dir_t
*dd
= ds
->ds_dir
;
2223 mutex_enter(&dd
->dd_lock
);
2224 uint64_t val
= dsl_dir_get_used(dd
);
2225 mutex_exit(&dd
->dd_lock
);
2231 dsl_get_creation(dsl_dataset_t
*ds
)
2233 return (dsl_dataset_phys(ds
)->ds_creation_time
);
2237 dsl_get_creationtxg(dsl_dataset_t
*ds
)
2239 return (dsl_dataset_phys(ds
)->ds_creation_txg
);
2243 dsl_get_refquota(dsl_dataset_t
*ds
)
2245 return (ds
->ds_quota
);
2249 dsl_get_refreservation(dsl_dataset_t
*ds
)
2251 return (ds
->ds_reserved
);
2255 dsl_get_guid(dsl_dataset_t
*ds
)
2257 return (dsl_dataset_phys(ds
)->ds_guid
);
2261 dsl_get_unique(dsl_dataset_t
*ds
)
2263 return (dsl_dataset_phys(ds
)->ds_unique_bytes
);
2267 dsl_get_objsetid(dsl_dataset_t
*ds
)
2269 return (ds
->ds_object
);
2273 dsl_get_userrefs(dsl_dataset_t
*ds
)
2275 return (ds
->ds_userrefs
);
2279 dsl_get_defer_destroy(dsl_dataset_t
*ds
)
2281 return (DS_IS_DEFER_DESTROY(ds
) ? 1 : 0);
2285 dsl_get_referenced(dsl_dataset_t
*ds
)
2287 return (dsl_dataset_phys(ds
)->ds_referenced_bytes
);
2291 dsl_get_numclones(dsl_dataset_t
*ds
)
2293 ASSERT(ds
->ds_is_snapshot
);
2294 return (dsl_dataset_phys(ds
)->ds_num_children
- 1);
2298 dsl_get_inconsistent(dsl_dataset_t
*ds
)
2300 return ((dsl_dataset_phys(ds
)->ds_flags
& DS_FLAG_INCONSISTENT
) ?
2305 dsl_get_available(dsl_dataset_t
*ds
)
2307 uint64_t refdbytes
= dsl_get_referenced(ds
);
2308 uint64_t availbytes
= dsl_dir_space_available(ds
->ds_dir
,
2310 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
2312 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2314 if (ds
->ds_quota
!= 0) {
2316 * Adjust available bytes according to refquota
2318 if (refdbytes
< ds
->ds_quota
) {
2319 availbytes
= MIN(availbytes
,
2320 ds
->ds_quota
- refdbytes
);
2325 return (availbytes
);
2329 dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
)
2331 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2332 dsl_dataset_t
*prev
;
2333 int err
= dsl_dataset_hold_obj(dp
,
2334 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, FTAG
, &prev
);
2336 uint64_t comp
, uncomp
;
2337 err
= dsl_dataset_space_written(prev
, ds
, written
,
2339 dsl_dataset_rele(prev
, FTAG
);
2345 * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
2348 dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
)
2350 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2351 if (ds
->ds_prev
!= NULL
&& ds
->ds_prev
!= dp
->dp_origin_snap
) {
2352 dsl_dataset_name(ds
->ds_prev
, snap
);
2360 * Returns the mountpoint property and source for the given dataset in the value
2361 * and source buffers. The value buffer must be at least as large as MAXPATHLEN
2362 * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
2363 * Returns 0 on success and an error on failure.
2366 dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
2370 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2372 /* Retrieve the mountpoint value stored in the zap opbject */
2373 error
= dsl_prop_get_ds(ds
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
), 1,
2374 ZAP_MAXVALUELEN
, value
, source
);
2380 * Process the dsname and source to find the full mountpoint string.
2381 * Can be skipped for 'legacy' or 'none'.
2383 if (value
[0] == '/') {
2384 char *buf
= kmem_alloc(ZAP_MAXVALUELEN
, KM_SLEEP
);
2386 const char *relpath
;
2389 * If we inherit the mountpoint, even from a dataset
2390 * with a received value, the source will be the path of
2391 * the dataset we inherit from. If source is
2392 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2395 if (strcmp(source
, ZPROP_SOURCE_VAL_RECVD
) == 0) {
2398 ASSERT0(strncmp(dsname
, source
, strlen(source
)));
2399 relpath
= dsname
+ strlen(source
);
2400 if (relpath
[0] == '/')
2404 spa_altroot(dp
->dp_spa
, root
, ZAP_MAXVALUELEN
);
2407 * Special case an alternate root of '/'. This will
2408 * avoid having multiple leading slashes in the
2411 if (strcmp(root
, "/") == 0)
2415 * If the mountpoint is '/' then skip over this
2416 * if we are obtaining either an alternate root or
2417 * an inherited mountpoint.
2420 if (value
[1] == '\0' && (root
[0] != '\0' ||
2421 relpath
[0] != '\0'))
2424 if (relpath
[0] == '\0') {
2425 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s",
2428 (void) snprintf(value
, ZAP_MAXVALUELEN
, "%s%s%s%s",
2429 root
, mnt
, relpath
[0] == '@' ? "" : "/",
2432 kmem_free(buf
, ZAP_MAXVALUELEN
);
2439 dsl_dataset_stats(dsl_dataset_t
*ds
, nvlist_t
*nv
)
2441 dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
;
2443 ASSERT(dsl_pool_config_held(dp
));
2445 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRATIO
,
2446 dsl_get_refratio(ds
));
2447 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_LOGICALREFERENCED
,
2448 dsl_get_logicalreferenced(ds
));
2449 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_COMPRESSRATIO
,
2450 dsl_get_compressratio(ds
));
2451 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USED
,
2454 if (ds
->ds_is_snapshot
) {
2455 get_clones_stat(ds
, nv
);
2457 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2458 if (dsl_get_prev_snap(ds
, buf
) == 0)
2459 dsl_prop_nvlist_add_string(nv
, ZFS_PROP_PREV_SNAP
,
2461 dsl_dir_stats(ds
->ds_dir
, nv
);
2464 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_AVAILABLE
,
2465 dsl_get_available(ds
));
2466 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFERENCED
,
2467 dsl_get_referenced(ds
));
2468 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATION
,
2469 dsl_get_creation(ds
));
2470 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_CREATETXG
,
2471 dsl_get_creationtxg(ds
));
2472 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFQUOTA
,
2473 dsl_get_refquota(ds
));
2474 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_REFRESERVATION
,
2475 dsl_get_refreservation(ds
));
2476 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_GUID
,
2478 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_UNIQUE
,
2479 dsl_get_unique(ds
));
2480 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_OBJSETID
,
2481 dsl_get_objsetid(ds
));
2482 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_USERREFS
,
2483 dsl_get_userrefs(ds
));
2484 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_DEFER_DESTROY
,
2485 dsl_get_defer_destroy(ds
));
2486 dsl_dataset_crypt_stats(ds
, nv
);
2488 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
!= 0) {
2490 if (dsl_get_written(ds
, &written
) == 0) {
2491 dsl_prop_nvlist_add_uint64(nv
, ZFS_PROP_WRITTEN
,
2496 if (!dsl_dataset_is_snapshot(ds
)) {
2498 * A failed "newfs" (e.g. full) resumable receive leaves
2499 * the stats set on this dataset. Check here for the prop.
2501 get_receive_resume_stats(ds
, nv
);
2504 * A failed incremental resumable receive leaves the
2505 * stats set on our child named "%recv". Check the child
2508 /* 6 extra bytes for /%recv */
2509 char recvname
[ZFS_MAX_DATASET_NAME_LEN
+ 6];
2510 dsl_dataset_t
*recv_ds
;
2511 dsl_dataset_name(ds
, recvname
);
2512 if (strlcat(recvname
, "/", sizeof (recvname
)) <
2513 sizeof (recvname
) &&
2514 strlcat(recvname
, recv_clone_name
, sizeof (recvname
)) <
2515 sizeof (recvname
) &&
2516 dsl_dataset_hold(dp
, recvname
, FTAG
, &recv_ds
) == 0) {
2517 get_receive_resume_stats(recv_ds
, nv
);
2518 dsl_dataset_rele(recv_ds
, FTAG
);
2524 dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
)
2526 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2527 ASSERT(dsl_pool_config_held(dp
));
2529 stat
->dds_creation_txg
= dsl_get_creationtxg(ds
);
2530 stat
->dds_inconsistent
= dsl_get_inconsistent(ds
);
2531 stat
->dds_guid
= dsl_get_guid(ds
);
2532 stat
->dds_origin
[0] = '\0';
2533 if (ds
->ds_is_snapshot
) {
2534 stat
->dds_is_snapshot
= B_TRUE
;
2535 stat
->dds_num_clones
= dsl_get_numclones(ds
);
2537 stat
->dds_is_snapshot
= B_FALSE
;
2538 stat
->dds_num_clones
= 0;
2540 if (dsl_dir_is_clone(ds
->ds_dir
)) {
2541 dsl_dir_get_origin(ds
->ds_dir
, stat
->dds_origin
);
2547 dsl_dataset_fsid_guid(dsl_dataset_t
*ds
)
2549 return (ds
->ds_fsid_guid
);
2553 dsl_dataset_space(dsl_dataset_t
*ds
,
2554 uint64_t *refdbytesp
, uint64_t *availbytesp
,
2555 uint64_t *usedobjsp
, uint64_t *availobjsp
)
2557 *refdbytesp
= dsl_dataset_phys(ds
)->ds_referenced_bytes
;
2558 *availbytesp
= dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
);
2559 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
)
2561 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
;
2562 if (ds
->ds_quota
!= 0) {
2564 * Adjust available bytes according to refquota
2566 if (*refdbytesp
< ds
->ds_quota
)
2567 *availbytesp
= MIN(*availbytesp
,
2568 ds
->ds_quota
- *refdbytesp
);
2572 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2573 *usedobjsp
= BP_GET_FILL(&dsl_dataset_phys(ds
)->ds_bp
);
2574 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2575 *availobjsp
= DN_MAX_OBJECT
- *usedobjsp
;
2579 dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
, dsl_dataset_t
*snap
)
2581 ASSERTV(dsl_pool_t
*dp
= ds
->ds_dir
->dd_pool
);
2584 ASSERT(dsl_pool_config_held(dp
));
2587 rrw_enter(&ds
->ds_bp_rwlock
, RW_READER
, FTAG
);
2588 birth
= dsl_dataset_get_blkptr(ds
)->blk_birth
;
2589 rrw_exit(&ds
->ds_bp_rwlock
, FTAG
);
2590 if (birth
> dsl_dataset_phys(snap
)->ds_creation_txg
) {
2591 objset_t
*os
, *os_snap
;
2593 * It may be that only the ZIL differs, because it was
2594 * reset in the head. Don't count that as being
2597 if (dmu_objset_from_ds(ds
, &os
) != 0)
2599 if (dmu_objset_from_ds(snap
, &os_snap
) != 0)
2601 return (bcmp(&os
->os_phys
->os_meta_dnode
,
2602 &os_snap
->os_phys
->os_meta_dnode
,
2603 sizeof (os
->os_phys
->os_meta_dnode
)) != 0);
2608 typedef struct dsl_dataset_rename_snapshot_arg
{
2609 const char *ddrsa_fsname
;
2610 const char *ddrsa_oldsnapname
;
2611 const char *ddrsa_newsnapname
;
2612 boolean_t ddrsa_recursive
;
2614 } dsl_dataset_rename_snapshot_arg_t
;
2618 dsl_dataset_rename_snapshot_check_impl(dsl_pool_t
*dp
,
2619 dsl_dataset_t
*hds
, void *arg
)
2621 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2625 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2627 /* ignore nonexistent snapshots */
2628 return (error
== ENOENT
? 0 : error
);
2631 /* new name should not exist */
2632 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_newsnapname
, &val
);
2634 error
= SET_ERROR(EEXIST
);
2635 else if (error
== ENOENT
)
2638 /* dataset name + 1 for the "@" + the new snapshot name must fit */
2639 if (dsl_dir_namelen(hds
->ds_dir
) + 1 +
2640 strlen(ddrsa
->ddrsa_newsnapname
) >= ZFS_MAX_DATASET_NAME_LEN
)
2641 error
= SET_ERROR(ENAMETOOLONG
);
2647 dsl_dataset_rename_snapshot_check(void *arg
, dmu_tx_t
*tx
)
2649 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2650 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2654 error
= dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
);
2658 if (ddrsa
->ddrsa_recursive
) {
2659 error
= dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2660 dsl_dataset_rename_snapshot_check_impl
, ddrsa
,
2663 error
= dsl_dataset_rename_snapshot_check_impl(dp
, hds
, ddrsa
);
2665 dsl_dataset_rele(hds
, FTAG
);
2670 dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t
*dp
,
2671 dsl_dataset_t
*hds
, void *arg
)
2673 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2676 dmu_tx_t
*tx
= ddrsa
->ddrsa_tx
;
2679 error
= dsl_dataset_snap_lookup(hds
, ddrsa
->ddrsa_oldsnapname
, &val
);
2680 ASSERT(error
== 0 || error
== ENOENT
);
2681 if (error
== ENOENT
) {
2682 /* ignore nonexistent snapshots */
2686 VERIFY0(dsl_dataset_hold_obj(dp
, val
, FTAG
, &ds
));
2688 /* log before we change the name */
2689 spa_history_log_internal_ds(ds
, "rename", tx
,
2690 "-> @%s", ddrsa
->ddrsa_newsnapname
);
2692 VERIFY0(dsl_dataset_snap_remove(hds
, ddrsa
->ddrsa_oldsnapname
, tx
,
2694 mutex_enter(&ds
->ds_lock
);
2695 (void) strlcpy(ds
->ds_snapname
, ddrsa
->ddrsa_newsnapname
,
2696 sizeof (ds
->ds_snapname
));
2697 mutex_exit(&ds
->ds_lock
);
2698 VERIFY0(zap_add(dp
->dp_meta_objset
,
2699 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
,
2700 ds
->ds_snapname
, 8, 1, &ds
->ds_object
, tx
));
2701 zvol_rename_minors(dp
->dp_spa
, ddrsa
->ddrsa_oldsnapname
,
2702 ddrsa
->ddrsa_newsnapname
, B_TRUE
);
2704 dsl_dataset_rele(ds
, FTAG
);
2709 dsl_dataset_rename_snapshot_sync(void *arg
, dmu_tx_t
*tx
)
2711 dsl_dataset_rename_snapshot_arg_t
*ddrsa
= arg
;
2712 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2713 dsl_dataset_t
*hds
= NULL
;
2715 VERIFY0(dsl_dataset_hold(dp
, ddrsa
->ddrsa_fsname
, FTAG
, &hds
));
2716 ddrsa
->ddrsa_tx
= tx
;
2717 if (ddrsa
->ddrsa_recursive
) {
2718 VERIFY0(dmu_objset_find_dp(dp
, hds
->ds_dir
->dd_object
,
2719 dsl_dataset_rename_snapshot_sync_impl
, ddrsa
,
2722 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp
, hds
, ddrsa
));
2724 dsl_dataset_rele(hds
, FTAG
);
2728 dsl_dataset_rename_snapshot(const char *fsname
,
2729 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
)
2731 dsl_dataset_rename_snapshot_arg_t ddrsa
;
2733 ddrsa
.ddrsa_fsname
= fsname
;
2734 ddrsa
.ddrsa_oldsnapname
= oldsnapname
;
2735 ddrsa
.ddrsa_newsnapname
= newsnapname
;
2736 ddrsa
.ddrsa_recursive
= recursive
;
2738 return (dsl_sync_task(fsname
, dsl_dataset_rename_snapshot_check
,
2739 dsl_dataset_rename_snapshot_sync
, &ddrsa
,
2740 1, ZFS_SPACE_CHECK_RESERVED
));
2744 * If we're doing an ownership handoff, we need to make sure that there is
2745 * only one long hold on the dataset. We're not allowed to change anything here
2746 * so we don't permanently release the long hold or regular hold here. We want
2747 * to do this only when syncing to avoid the dataset unexpectedly going away
2748 * when we release the long hold.
2751 dsl_dataset_handoff_check(dsl_dataset_t
*ds
, void *owner
, dmu_tx_t
*tx
)
2755 if (!dmu_tx_is_syncing(tx
))
2758 if (owner
!= NULL
) {
2759 VERIFY3P(ds
->ds_owner
, ==, owner
);
2760 dsl_dataset_long_rele(ds
, owner
);
2763 held
= dsl_dataset_long_held(ds
);
2766 dsl_dataset_long_hold(ds
, owner
);
2769 return (SET_ERROR(EBUSY
));
2775 dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
)
2777 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2778 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2780 int64_t unused_refres_delta
;
2783 error
= dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
);
2787 /* must not be a snapshot */
2788 if (ds
->ds_is_snapshot
) {
2789 dsl_dataset_rele(ds
, FTAG
);
2790 return (SET_ERROR(EINVAL
));
2793 /* must have a most recent snapshot */
2794 if (dsl_dataset_phys(ds
)->ds_prev_snap_txg
< TXG_INITIAL
) {
2795 dsl_dataset_rele(ds
, FTAG
);
2796 return (SET_ERROR(ESRCH
));
2800 * No rollback to a snapshot created in the current txg, because
2801 * the rollback may dirty the dataset and create blocks that are
2802 * not reachable from the rootbp while having a birth txg that
2803 * falls into the snapshot's range.
2805 if (dmu_tx_is_syncing(tx
) &&
2806 dsl_dataset_phys(ds
)->ds_prev_snap_txg
>= tx
->tx_txg
) {
2807 dsl_dataset_rele(ds
, FTAG
);
2808 return (SET_ERROR(EAGAIN
));
2812 * If the expected target snapshot is specified, then check that
2813 * the latest snapshot is it.
2815 if (ddra
->ddra_tosnap
!= NULL
) {
2816 dsl_dataset_t
*snapds
;
2818 /* Check if the target snapshot exists at all. */
2819 error
= dsl_dataset_hold(dp
, ddra
->ddra_tosnap
, FTAG
, &snapds
);
2822 * ESRCH is used to signal that the target snapshot does
2823 * not exist, while ENOENT is used to report that
2824 * the rolled back dataset does not exist.
2825 * ESRCH is also used to cover other cases where the
2826 * target snapshot is not related to the dataset being
2827 * rolled back such as being in a different pool.
2829 if (error
== ENOENT
|| error
== EXDEV
)
2830 error
= SET_ERROR(ESRCH
);
2831 dsl_dataset_rele(ds
, FTAG
);
2834 ASSERT(snapds
->ds_is_snapshot
);
2836 /* Check if the snapshot is the latest snapshot indeed. */
2837 if (snapds
!= ds
->ds_prev
) {
2839 * Distinguish between the case where the only problem
2840 * is intervening snapshots (EEXIST) vs the snapshot
2841 * not being a valid target for rollback (ESRCH).
2843 if (snapds
->ds_dir
== ds
->ds_dir
||
2844 (dsl_dir_is_clone(ds
->ds_dir
) &&
2845 dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
==
2846 snapds
->ds_object
)) {
2847 error
= SET_ERROR(EEXIST
);
2849 error
= SET_ERROR(ESRCH
);
2851 dsl_dataset_rele(snapds
, FTAG
);
2852 dsl_dataset_rele(ds
, FTAG
);
2855 dsl_dataset_rele(snapds
, FTAG
);
2858 /* must not have any bookmarks after the most recent snapshot */
2859 nvlist_t
*proprequest
= fnvlist_alloc();
2860 fnvlist_add_boolean(proprequest
, zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2861 nvlist_t
*bookmarks
= fnvlist_alloc();
2862 error
= dsl_get_bookmarks_impl(ds
, proprequest
, bookmarks
);
2863 fnvlist_free(proprequest
);
2865 dsl_dataset_rele(ds
, FTAG
);
2868 for (nvpair_t
*pair
= nvlist_next_nvpair(bookmarks
, NULL
);
2869 pair
!= NULL
; pair
= nvlist_next_nvpair(bookmarks
, pair
)) {
2871 fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair
),
2872 zfs_prop_to_name(ZFS_PROP_CREATETXG
));
2873 uint64_t createtxg
= fnvlist_lookup_uint64(valuenv
, "value");
2874 if (createtxg
> dsl_dataset_phys(ds
)->ds_prev_snap_txg
) {
2875 fnvlist_free(bookmarks
);
2876 dsl_dataset_rele(ds
, FTAG
);
2877 return (SET_ERROR(EEXIST
));
2880 fnvlist_free(bookmarks
);
2882 error
= dsl_dataset_handoff_check(ds
, ddra
->ddra_owner
, tx
);
2884 dsl_dataset_rele(ds
, FTAG
);
2889 * Check if the snap we are rolling back to uses more than
2892 if (ds
->ds_quota
!= 0 &&
2893 dsl_dataset_phys(ds
->ds_prev
)->ds_referenced_bytes
> ds
->ds_quota
) {
2894 dsl_dataset_rele(ds
, FTAG
);
2895 return (SET_ERROR(EDQUOT
));
2899 * When we do the clone swap, we will temporarily use more space
2900 * due to the refreservation (the head will no longer have any
2901 * unique space, so the entire amount of the refreservation will need
2902 * to be free). We will immediately destroy the clone, freeing
2903 * this space, but the freeing happens over many txg's.
2905 unused_refres_delta
= (int64_t)MIN(ds
->ds_reserved
,
2906 dsl_dataset_phys(ds
)->ds_unique_bytes
);
2908 if (unused_refres_delta
> 0 &&
2909 unused_refres_delta
>
2910 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, TRUE
)) {
2911 dsl_dataset_rele(ds
, FTAG
);
2912 return (SET_ERROR(ENOSPC
));
2915 dsl_dataset_rele(ds
, FTAG
);
2920 dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
)
2922 dsl_dataset_rollback_arg_t
*ddra
= arg
;
2923 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2924 dsl_dataset_t
*ds
, *clone
;
2926 char namebuf
[ZFS_MAX_DATASET_NAME_LEN
];
2928 VERIFY0(dsl_dataset_hold(dp
, ddra
->ddra_fsname
, FTAG
, &ds
));
2930 dsl_dataset_name(ds
->ds_prev
, namebuf
);
2931 fnvlist_add_string(ddra
->ddra_result
, "target", namebuf
);
2933 cloneobj
= dsl_dataset_create_sync(ds
->ds_dir
, "%rollback",
2934 ds
->ds_prev
, DS_CREATE_FLAG_NODIRTY
, kcred
, NULL
, tx
);
2936 VERIFY0(dsl_dataset_hold_obj(dp
, cloneobj
, FTAG
, &clone
));
2938 dsl_dataset_clone_swap_sync_impl(clone
, ds
, tx
);
2939 dsl_dataset_zero_zil(ds
, tx
);
2941 dsl_destroy_head_sync_impl(clone
, tx
);
2943 dsl_dataset_rele(clone
, FTAG
);
2944 dsl_dataset_rele(ds
, FTAG
);
2948 * Rolls back the given filesystem or volume to the most recent snapshot.
2949 * The name of the most recent snapshot will be returned under key "target"
2950 * in the result nvlist.
2953 * - The existing dataset MUST be owned by the specified owner at entry
2954 * - Upon return, dataset will still be held by the same owner, whether we
2957 * This mode is required any time the existing filesystem is mounted. See
2958 * notes above zfs_suspend_fs() for further details.
2961 dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
2964 dsl_dataset_rollback_arg_t ddra
;
2966 ddra
.ddra_fsname
= fsname
;
2967 ddra
.ddra_tosnap
= tosnap
;
2968 ddra
.ddra_owner
= owner
;
2969 ddra
.ddra_result
= result
;
2971 return (dsl_sync_task(fsname
, dsl_dataset_rollback_check
,
2972 dsl_dataset_rollback_sync
, &ddra
,
2973 1, ZFS_SPACE_CHECK_RESERVED
));
2976 struct promotenode
{
2981 static int snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
);
2982 static int promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
,
2984 static void promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
);
2987 dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
)
2989 dsl_dataset_promote_arg_t
*ddpa
= arg
;
2990 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
2992 struct promotenode
*snap
;
2993 dsl_dataset_t
*origin_ds
;
2997 size_t max_snap_len
;
2998 boolean_t conflicting_snaps
;
3000 err
= promote_hold(ddpa
, dp
, FTAG
);
3004 hds
= ddpa
->ddpa_clone
;
3005 max_snap_len
= MAXNAMELEN
- strlen(ddpa
->ddpa_clonename
) - 1;
3007 if (dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
) {
3008 promote_rele(ddpa
, FTAG
);
3009 return (SET_ERROR(EXDEV
));
3012 snap
= list_head(&ddpa
->shared_snaps
);
3014 err
= SET_ERROR(ENOENT
);
3017 origin_ds
= snap
->ds
;
3020 * Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
3021 * When doing a promote we must make sure the encryption root for
3022 * both the target and the target's origin does not change to avoid
3023 * needing to rewrap encryption keys
3025 err
= dsl_dataset_promote_crypt_check(hds
->ds_dir
, origin_ds
->ds_dir
);
3030 * Compute and check the amount of space to transfer. Since this is
3031 * so expensive, don't do the preliminary check.
3033 if (!dmu_tx_is_syncing(tx
)) {
3034 promote_rele(ddpa
, FTAG
);
3038 /* compute origin's new unique space */
3039 snap
= list_tail(&ddpa
->clone_snaps
);
3040 ASSERT(snap
!= NULL
);
3041 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3042 origin_ds
->ds_object
);
3043 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3044 dsl_dataset_phys(origin_ds
)->ds_prev_snap_txg
, UINT64_MAX
,
3045 &ddpa
->unique
, &unused
, &unused
);
3048 * Walk the snapshots that we are moving
3050 * Compute space to transfer. Consider the incremental changes
3051 * to used by each snapshot:
3052 * (my used) = (prev's used) + (blocks born) - (blocks killed)
3053 * So each snapshot gave birth to:
3054 * (blocks born) = (my used) - (prev's used) + (blocks killed)
3055 * So a sequence would look like:
3056 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
3057 * Which simplifies to:
3058 * uN + kN + kN-1 + ... + k1 + k0
3059 * Note however, if we stop before we reach the ORIGIN we get:
3060 * uN + kN + kN-1 + ... + kM - uM-1
3062 conflicting_snaps
= B_FALSE
;
3064 ddpa
->used
= dsl_dataset_phys(origin_ds
)->ds_referenced_bytes
;
3065 ddpa
->comp
= dsl_dataset_phys(origin_ds
)->ds_compressed_bytes
;
3066 ddpa
->uncomp
= dsl_dataset_phys(origin_ds
)->ds_uncompressed_bytes
;
3067 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3068 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3069 uint64_t val
, dlused
, dlcomp
, dluncomp
;
3070 dsl_dataset_t
*ds
= snap
->ds
;
3075 * If there are long holds, we won't be able to evict
3078 if (dsl_dataset_long_held(ds
)) {
3079 err
= SET_ERROR(EBUSY
);
3083 /* Check that the snapshot name does not conflict */
3084 VERIFY0(dsl_dataset_get_snapname(ds
));
3085 if (strlen(ds
->ds_snapname
) >= max_snap_len
) {
3086 err
= SET_ERROR(ENAMETOOLONG
);
3089 err
= dsl_dataset_snap_lookup(hds
, ds
->ds_snapname
, &val
);
3091 fnvlist_add_boolean(ddpa
->err_ds
,
3092 snap
->ds
->ds_snapname
);
3093 conflicting_snaps
= B_TRUE
;
3094 } else if (err
!= ENOENT
) {
3098 /* The very first snapshot does not have a deadlist */
3099 if (dsl_dataset_phys(ds
)->ds_prev_snap_obj
== 0)
3102 dsl_deadlist_space(&ds
->ds_deadlist
,
3103 &dlused
, &dlcomp
, &dluncomp
);
3104 ddpa
->used
+= dlused
;
3105 ddpa
->comp
+= dlcomp
;
3106 ddpa
->uncomp
+= dluncomp
;
3110 * In order to return the full list of conflicting snapshots, we check
3111 * whether there was a conflict after traversing all of them.
3113 if (conflicting_snaps
) {
3114 err
= SET_ERROR(EEXIST
);
3119 * If we are a clone of a clone then we never reached ORIGIN,
3120 * so we need to subtract out the clone origin's used space.
3122 if (ddpa
->origin_origin
) {
3124 dsl_dataset_phys(ddpa
->origin_origin
)->ds_referenced_bytes
;
3126 dsl_dataset_phys(ddpa
->origin_origin
)->ds_compressed_bytes
;
3128 dsl_dataset_phys(ddpa
->origin_origin
)->
3129 ds_uncompressed_bytes
;
3132 /* Check that there is enough space and limit headroom here */
3133 err
= dsl_dir_transfer_possible(origin_ds
->ds_dir
, hds
->ds_dir
,
3134 0, ss_mv_cnt
, ddpa
->used
, ddpa
->cr
);
3139 * Compute the amounts of space that will be used by snapshots
3140 * after the promotion (for both origin and clone). For each,
3141 * it is the amount of space that will be on all of their
3142 * deadlists (that was not born before their new origin).
3144 if (dsl_dir_phys(hds
->ds_dir
)->dd_flags
& DD_FLAG_USED_BREAKDOWN
) {
3148 * Note, typically this will not be a clone of a clone,
3149 * so dd_origin_txg will be < TXG_INITIAL, so
3150 * these snaplist_space() -> dsl_deadlist_space_range()
3151 * calls will be fast because they do not have to
3152 * iterate over all bps.
3154 snap
= list_head(&ddpa
->origin_snaps
);
3156 err
= SET_ERROR(ENOENT
);
3159 err
= snaplist_space(&ddpa
->shared_snaps
,
3160 snap
->ds
->ds_dir
->dd_origin_txg
, &ddpa
->cloneusedsnap
);
3164 err
= snaplist_space(&ddpa
->clone_snaps
,
3165 snap
->ds
->ds_dir
->dd_origin_txg
, &space
);
3168 ddpa
->cloneusedsnap
+= space
;
3170 if (dsl_dir_phys(origin_ds
->ds_dir
)->dd_flags
&
3171 DD_FLAG_USED_BREAKDOWN
) {
3172 err
= snaplist_space(&ddpa
->origin_snaps
,
3173 dsl_dataset_phys(origin_ds
)->ds_creation_txg
,
3174 &ddpa
->originusedsnap
);
3180 promote_rele(ddpa
, FTAG
);
3185 dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
)
3187 dsl_dataset_promote_arg_t
*ddpa
= arg
;
3188 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3190 struct promotenode
*snap
;
3191 dsl_dataset_t
*origin_ds
;
3192 dsl_dataset_t
*origin_head
;
3194 dsl_dir_t
*odd
= NULL
;
3195 uint64_t oldnext_obj
;
3198 VERIFY0(promote_hold(ddpa
, dp
, FTAG
));
3199 hds
= ddpa
->ddpa_clone
;
3201 ASSERT0(dsl_dataset_phys(hds
)->ds_flags
& DS_FLAG_NOPROMOTE
);
3203 snap
= list_head(&ddpa
->shared_snaps
);
3204 origin_ds
= snap
->ds
;
3207 snap
= list_head(&ddpa
->origin_snaps
);
3208 origin_head
= snap
->ds
;
3211 * We need to explicitly open odd, since origin_ds's dd will be
3214 VERIFY0(dsl_dir_hold_obj(dp
, origin_ds
->ds_dir
->dd_object
,
3217 dsl_dataset_promote_crypt_sync(hds
->ds_dir
, odd
, tx
);
3219 /* change origin's next snap */
3220 dmu_buf_will_dirty(origin_ds
->ds_dbuf
, tx
);
3221 oldnext_obj
= dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
;
3222 snap
= list_tail(&ddpa
->clone_snaps
);
3223 ASSERT3U(dsl_dataset_phys(snap
->ds
)->ds_prev_snap_obj
, ==,
3224 origin_ds
->ds_object
);
3225 dsl_dataset_phys(origin_ds
)->ds_next_snap_obj
= snap
->ds
->ds_object
;
3227 /* change the origin's next clone */
3228 if (dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
) {
3229 dsl_dataset_remove_from_next_clones(origin_ds
,
3230 snap
->ds
->ds_object
, tx
);
3231 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3232 dsl_dataset_phys(origin_ds
)->ds_next_clones_obj
,
3237 dmu_buf_will_dirty(dd
->dd_dbuf
, tx
);
3238 ASSERT3U(dsl_dir_phys(dd
)->dd_origin_obj
, ==, origin_ds
->ds_object
);
3239 dsl_dir_phys(dd
)->dd_origin_obj
= dsl_dir_phys(odd
)->dd_origin_obj
;
3240 dd
->dd_origin_txg
= origin_head
->ds_dir
->dd_origin_txg
;
3241 dmu_buf_will_dirty(odd
->dd_dbuf
, tx
);
3242 dsl_dir_phys(odd
)->dd_origin_obj
= origin_ds
->ds_object
;
3243 origin_head
->ds_dir
->dd_origin_txg
=
3244 dsl_dataset_phys(origin_ds
)->ds_creation_txg
;
3246 /* change dd_clone entries */
3247 if (spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3248 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3249 dsl_dir_phys(odd
)->dd_clones
, hds
->ds_object
, tx
));
3250 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3251 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3252 hds
->ds_object
, tx
));
3254 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3255 dsl_dir_phys(ddpa
->origin_origin
->ds_dir
)->dd_clones
,
3256 origin_head
->ds_object
, tx
));
3257 if (dsl_dir_phys(dd
)->dd_clones
== 0) {
3258 dsl_dir_phys(dd
)->dd_clones
=
3259 zap_create(dp
->dp_meta_objset
, DMU_OT_DSL_CLONES
,
3260 DMU_OT_NONE
, 0, tx
);
3262 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3263 dsl_dir_phys(dd
)->dd_clones
, origin_head
->ds_object
, tx
));
3266 /* move snapshots to this dir */
3267 for (snap
= list_head(&ddpa
->shared_snaps
); snap
;
3268 snap
= list_next(&ddpa
->shared_snaps
, snap
)) {
3269 dsl_dataset_t
*ds
= snap
->ds
;
3272 * Property callbacks are registered to a particular
3273 * dsl_dir. Since ours is changing, evict the objset
3274 * so that they will be unregistered from the old dsl_dir.
3276 if (ds
->ds_objset
) {
3277 dmu_objset_evict(ds
->ds_objset
);
3278 ds
->ds_objset
= NULL
;
3281 /* move snap name entry */
3282 VERIFY0(dsl_dataset_get_snapname(ds
));
3283 VERIFY0(dsl_dataset_snap_remove(origin_head
,
3284 ds
->ds_snapname
, tx
, B_TRUE
));
3285 VERIFY0(zap_add(dp
->dp_meta_objset
,
3286 dsl_dataset_phys(hds
)->ds_snapnames_zapobj
, ds
->ds_snapname
,
3287 8, 1, &ds
->ds_object
, tx
));
3288 dsl_fs_ss_count_adjust(hds
->ds_dir
, 1,
3289 DD_FIELD_SNAPSHOT_COUNT
, tx
);
3291 /* change containing dsl_dir */
3292 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3293 ASSERT3U(dsl_dataset_phys(ds
)->ds_dir_obj
, ==, odd
->dd_object
);
3294 dsl_dataset_phys(ds
)->ds_dir_obj
= dd
->dd_object
;
3295 ASSERT3P(ds
->ds_dir
, ==, odd
);
3296 dsl_dir_rele(ds
->ds_dir
, ds
);
3297 VERIFY0(dsl_dir_hold_obj(dp
, dd
->dd_object
,
3298 NULL
, ds
, &ds
->ds_dir
));
3300 /* move any clone references */
3301 if (dsl_dataset_phys(ds
)->ds_next_clones_obj
&&
3302 spa_version(dp
->dp_spa
) >= SPA_VERSION_DIR_CLONES
) {
3306 for (zap_cursor_init(&zc
, dp
->dp_meta_objset
,
3307 dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3308 zap_cursor_retrieve(&zc
, &za
) == 0;
3309 zap_cursor_advance(&zc
)) {
3310 dsl_dataset_t
*cnds
;
3313 if (za
.za_first_integer
== oldnext_obj
) {
3315 * We've already moved the
3316 * origin's reference.
3321 VERIFY0(dsl_dataset_hold_obj(dp
,
3322 za
.za_first_integer
, FTAG
, &cnds
));
3323 o
= dsl_dir_phys(cnds
->ds_dir
)->
3324 dd_head_dataset_obj
;
3326 VERIFY0(zap_remove_int(dp
->dp_meta_objset
,
3327 dsl_dir_phys(odd
)->dd_clones
, o
, tx
));
3328 VERIFY0(zap_add_int(dp
->dp_meta_objset
,
3329 dsl_dir_phys(dd
)->dd_clones
, o
, tx
));
3330 dsl_dataset_rele(cnds
, FTAG
);
3332 zap_cursor_fini(&zc
);
3335 ASSERT(!dsl_prop_hascb(ds
));
3339 * Change space accounting.
3340 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
3341 * both be valid, or both be 0 (resulting in delta == 0). This
3342 * is true for each of {clone,origin} independently.
3345 delta
= ddpa
->cloneusedsnap
-
3346 dsl_dir_phys(dd
)->dd_used_breakdown
[DD_USED_SNAP
];
3347 ASSERT3S(delta
, >=, 0);
3348 ASSERT3U(ddpa
->used
, >=, delta
);
3349 dsl_dir_diduse_space(dd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3350 dsl_dir_diduse_space(dd
, DD_USED_HEAD
,
3351 ddpa
->used
- delta
, ddpa
->comp
, ddpa
->uncomp
, tx
);
3353 delta
= ddpa
->originusedsnap
-
3354 dsl_dir_phys(odd
)->dd_used_breakdown
[DD_USED_SNAP
];
3355 ASSERT3S(delta
, <=, 0);
3356 ASSERT3U(ddpa
->used
, >=, -delta
);
3357 dsl_dir_diduse_space(odd
, DD_USED_SNAP
, delta
, 0, 0, tx
);
3358 dsl_dir_diduse_space(odd
, DD_USED_HEAD
,
3359 -ddpa
->used
- delta
, -ddpa
->comp
, -ddpa
->uncomp
, tx
);
3361 dsl_dataset_phys(origin_ds
)->ds_unique_bytes
= ddpa
->unique
;
3363 /* log history record */
3364 spa_history_log_internal_ds(hds
, "promote", tx
, "");
3366 dsl_dir_rele(odd
, FTAG
);
3367 promote_rele(ddpa
, FTAG
);
3371 * Make a list of dsl_dataset_t's for the snapshots between first_obj
3372 * (exclusive) and last_obj (inclusive). The list will be in reverse
3373 * order (last_obj will be the list_head()). If first_obj == 0, do all
3374 * snapshots back to this dataset's origin.
3377 snaplist_make(dsl_pool_t
*dp
,
3378 uint64_t first_obj
, uint64_t last_obj
, list_t
*l
, void *tag
)
3380 uint64_t obj
= last_obj
;
3382 list_create(l
, sizeof (struct promotenode
),
3383 offsetof(struct promotenode
, link
));
3385 while (obj
!= first_obj
) {
3387 struct promotenode
*snap
;
3390 err
= dsl_dataset_hold_obj(dp
, obj
, tag
, &ds
);
3391 ASSERT(err
!= ENOENT
);
3396 first_obj
= dsl_dir_phys(ds
->ds_dir
)->dd_origin_obj
;
3398 snap
= kmem_alloc(sizeof (*snap
), KM_SLEEP
);
3400 list_insert_tail(l
, snap
);
3401 obj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
3408 snaplist_space(list_t
*l
, uint64_t mintxg
, uint64_t *spacep
)
3410 struct promotenode
*snap
;
3413 for (snap
= list_head(l
); snap
; snap
= list_next(l
, snap
)) {
3414 uint64_t used
, comp
, uncomp
;
3415 dsl_deadlist_space_range(&snap
->ds
->ds_deadlist
,
3416 mintxg
, UINT64_MAX
, &used
, &comp
, &uncomp
);
3423 snaplist_destroy(list_t
*l
, void *tag
)
3425 struct promotenode
*snap
;
3427 if (l
== NULL
|| !list_link_active(&l
->list_head
))
3430 while ((snap
= list_tail(l
)) != NULL
) {
3431 list_remove(l
, snap
);
3432 dsl_dataset_rele(snap
->ds
, tag
);
3433 kmem_free(snap
, sizeof (*snap
));
3439 promote_hold(dsl_dataset_promote_arg_t
*ddpa
, dsl_pool_t
*dp
, void *tag
)
3443 struct promotenode
*snap
;
3445 error
= dsl_dataset_hold(dp
, ddpa
->ddpa_clonename
, tag
,
3449 dd
= ddpa
->ddpa_clone
->ds_dir
;
3451 if (ddpa
->ddpa_clone
->ds_is_snapshot
||
3452 !dsl_dir_is_clone(dd
)) {
3453 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3454 return (SET_ERROR(EINVAL
));
3457 error
= snaplist_make(dp
, 0, dsl_dir_phys(dd
)->dd_origin_obj
,
3458 &ddpa
->shared_snaps
, tag
);
3462 error
= snaplist_make(dp
, 0, ddpa
->ddpa_clone
->ds_object
,
3463 &ddpa
->clone_snaps
, tag
);
3467 snap
= list_head(&ddpa
->shared_snaps
);
3468 ASSERT3U(snap
->ds
->ds_object
, ==, dsl_dir_phys(dd
)->dd_origin_obj
);
3469 error
= snaplist_make(dp
, dsl_dir_phys(dd
)->dd_origin_obj
,
3470 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_head_dataset_obj
,
3471 &ddpa
->origin_snaps
, tag
);
3475 if (dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
!= 0) {
3476 error
= dsl_dataset_hold_obj(dp
,
3477 dsl_dir_phys(snap
->ds
->ds_dir
)->dd_origin_obj
,
3478 tag
, &ddpa
->origin_origin
);
3484 promote_rele(ddpa
, tag
);
3489 promote_rele(dsl_dataset_promote_arg_t
*ddpa
, void *tag
)
3491 snaplist_destroy(&ddpa
->shared_snaps
, tag
);
3492 snaplist_destroy(&ddpa
->clone_snaps
, tag
);
3493 snaplist_destroy(&ddpa
->origin_snaps
, tag
);
3494 if (ddpa
->origin_origin
!= NULL
)
3495 dsl_dataset_rele(ddpa
->origin_origin
, tag
);
3496 dsl_dataset_rele(ddpa
->ddpa_clone
, tag
);
3502 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
3503 * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
3506 dsl_dataset_promote(const char *name
, char *conflsnap
)
3508 dsl_dataset_promote_arg_t ddpa
= { 0 };
3511 nvpair_t
*snap_pair
;
3515 * We will modify space proportional to the number of
3516 * snapshots. Compute numsnaps.
3518 error
= dmu_objset_hold(name
, FTAG
, &os
);
3521 error
= zap_count(dmu_objset_pool(os
)->dp_meta_objset
,
3522 dsl_dataset_phys(dmu_objset_ds(os
))->ds_snapnames_zapobj
,
3524 dmu_objset_rele(os
, FTAG
);
3528 ddpa
.ddpa_clonename
= name
;
3529 ddpa
.err_ds
= fnvlist_alloc();
3532 error
= dsl_sync_task(name
, dsl_dataset_promote_check
,
3533 dsl_dataset_promote_sync
, &ddpa
,
3534 2 + numsnaps
, ZFS_SPACE_CHECK_RESERVED
);
3537 * Return the first conflicting snapshot found.
3539 snap_pair
= nvlist_next_nvpair(ddpa
.err_ds
, NULL
);
3540 if (snap_pair
!= NULL
&& conflsnap
!= NULL
)
3541 (void) strcpy(conflsnap
, nvpair_name(snap_pair
));
3543 fnvlist_free(ddpa
.err_ds
);
3548 dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
3549 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
)
3552 * "slack" factor for received datasets with refquota set on them.
3553 * See the bottom of this function for details on its use.
3555 uint64_t refquota_slack
= (uint64_t)DMU_MAX_ACCESS
*
3556 spa_asize_inflation
;
3557 int64_t unused_refres_delta
;
3559 /* they should both be heads */
3560 if (clone
->ds_is_snapshot
||
3561 origin_head
->ds_is_snapshot
)
3562 return (SET_ERROR(EINVAL
));
3564 /* if we are not forcing, the branch point should be just before them */
3565 if (!force
&& clone
->ds_prev
!= origin_head
->ds_prev
)
3566 return (SET_ERROR(EINVAL
));
3568 /* clone should be the clone (unless they are unrelated) */
3569 if (clone
->ds_prev
!= NULL
&&
3570 clone
->ds_prev
!= clone
->ds_dir
->dd_pool
->dp_origin_snap
&&
3571 origin_head
->ds_dir
!= clone
->ds_prev
->ds_dir
)
3572 return (SET_ERROR(EINVAL
));
3574 /* the clone should be a child of the origin */
3575 if (clone
->ds_dir
->dd_parent
!= origin_head
->ds_dir
)
3576 return (SET_ERROR(EINVAL
));
3578 /* origin_head shouldn't be modified unless 'force' */
3580 dsl_dataset_modified_since_snap(origin_head
, origin_head
->ds_prev
))
3581 return (SET_ERROR(ETXTBSY
));
3583 /* origin_head should have no long holds (e.g. is not mounted) */
3584 if (dsl_dataset_handoff_check(origin_head
, owner
, tx
))
3585 return (SET_ERROR(EBUSY
));
3587 /* check amount of any unconsumed refreservation */
3588 unused_refres_delta
=
3589 (int64_t)MIN(origin_head
->ds_reserved
,
3590 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3591 (int64_t)MIN(origin_head
->ds_reserved
,
3592 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3594 if (unused_refres_delta
> 0 &&
3595 unused_refres_delta
>
3596 dsl_dir_space_available(origin_head
->ds_dir
, NULL
, 0, TRUE
))
3597 return (SET_ERROR(ENOSPC
));
3600 * The clone can't be too much over the head's refquota.
3602 * To ensure that the entire refquota can be used, we allow one
3603 * transaction to exceed the the refquota. Therefore, this check
3604 * needs to also allow for the space referenced to be more than the
3605 * refquota. The maximum amount of space that one transaction can use
3606 * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
3607 * overage ensures that we are able to receive a filesystem that
3608 * exceeds the refquota on the source system.
3610 * So that overage is the refquota_slack we use below.
3612 if (origin_head
->ds_quota
!= 0 &&
3613 dsl_dataset_phys(clone
)->ds_referenced_bytes
>
3614 origin_head
->ds_quota
+ refquota_slack
)
3615 return (SET_ERROR(EDQUOT
));
3621 dsl_dataset_swap_remap_deadlists(dsl_dataset_t
*clone
,
3622 dsl_dataset_t
*origin
, dmu_tx_t
*tx
)
3624 uint64_t clone_remap_dl_obj
, origin_remap_dl_obj
;
3625 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3627 ASSERT(dsl_pool_sync_context(dp
));
3629 clone_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(clone
);
3630 origin_remap_dl_obj
= dsl_dataset_get_remap_deadlist_object(origin
);
3632 if (clone_remap_dl_obj
!= 0) {
3633 dsl_deadlist_close(&clone
->ds_remap_deadlist
);
3634 dsl_dataset_unset_remap_deadlist_object(clone
, tx
);
3636 if (origin_remap_dl_obj
!= 0) {
3637 dsl_deadlist_close(&origin
->ds_remap_deadlist
);
3638 dsl_dataset_unset_remap_deadlist_object(origin
, tx
);
3641 if (clone_remap_dl_obj
!= 0) {
3642 dsl_dataset_set_remap_deadlist_object(origin
,
3643 clone_remap_dl_obj
, tx
);
3644 dsl_deadlist_open(&origin
->ds_remap_deadlist
,
3645 dp
->dp_meta_objset
, clone_remap_dl_obj
);
3647 if (origin_remap_dl_obj
!= 0) {
3648 dsl_dataset_set_remap_deadlist_object(clone
,
3649 origin_remap_dl_obj
, tx
);
3650 dsl_deadlist_open(&clone
->ds_remap_deadlist
,
3651 dp
->dp_meta_objset
, origin_remap_dl_obj
);
3656 dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
3657 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
)
3659 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3660 int64_t unused_refres_delta
;
3662 ASSERT(clone
->ds_reserved
== 0);
3664 * NOTE: On DEBUG kernels there could be a race between this and
3665 * the check function if spa_asize_inflation is adjusted...
3667 ASSERT(origin_head
->ds_quota
== 0 ||
3668 dsl_dataset_phys(clone
)->ds_unique_bytes
<= origin_head
->ds_quota
+
3669 DMU_MAX_ACCESS
* spa_asize_inflation
);
3670 ASSERT3P(clone
->ds_prev
, ==, origin_head
->ds_prev
);
3673 * Swap per-dataset feature flags.
3675 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++) {
3676 if (!(spa_feature_table
[f
].fi_flags
&
3677 ZFEATURE_FLAG_PER_DATASET
)) {
3678 ASSERT(!dsl_dataset_feature_is_active(clone
, f
));
3679 ASSERT(!dsl_dataset_feature_is_active(origin_head
, f
));
3683 boolean_t clone_inuse
= dsl_dataset_feature_is_active(clone
, f
);
3684 void *clone_feature
= clone
->ds_feature
[f
];
3685 boolean_t origin_head_inuse
=
3686 dsl_dataset_feature_is_active(origin_head
, f
);
3687 void *origin_head_feature
= origin_head
->ds_feature
[f
];
3690 dsl_dataset_deactivate_feature_impl(clone
, f
, tx
);
3691 if (origin_head_inuse
)
3692 dsl_dataset_deactivate_feature_impl(origin_head
, f
, tx
);
3695 dsl_dataset_activate_feature(origin_head
->ds_object
, f
,
3697 origin_head
->ds_feature
[f
] = clone_feature
;
3699 if (origin_head_inuse
) {
3700 dsl_dataset_activate_feature(clone
->ds_object
, f
,
3701 origin_head_feature
, tx
);
3702 clone
->ds_feature
[f
] = origin_head_feature
;
3706 dmu_buf_will_dirty(clone
->ds_dbuf
, tx
);
3707 dmu_buf_will_dirty(origin_head
->ds_dbuf
, tx
);
3709 if (clone
->ds_objset
!= NULL
) {
3710 dmu_objset_evict(clone
->ds_objset
);
3711 clone
->ds_objset
= NULL
;
3714 if (origin_head
->ds_objset
!= NULL
) {
3715 dmu_objset_evict(origin_head
->ds_objset
);
3716 origin_head
->ds_objset
= NULL
;
3719 unused_refres_delta
=
3720 (int64_t)MIN(origin_head
->ds_reserved
,
3721 dsl_dataset_phys(origin_head
)->ds_unique_bytes
) -
3722 (int64_t)MIN(origin_head
->ds_reserved
,
3723 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3726 * Reset origin's unique bytes, if it exists.
3728 if (clone
->ds_prev
) {
3729 dsl_dataset_t
*origin
= clone
->ds_prev
;
3730 uint64_t comp
, uncomp
;
3732 dmu_buf_will_dirty(origin
->ds_dbuf
, tx
);
3733 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3734 dsl_dataset_phys(origin
)->ds_prev_snap_txg
, UINT64_MAX
,
3735 &dsl_dataset_phys(origin
)->ds_unique_bytes
, &comp
, &uncomp
);
3740 rrw_enter(&clone
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3741 rrw_enter(&origin_head
->ds_bp_rwlock
, RW_WRITER
, FTAG
);
3743 tmp
= dsl_dataset_phys(origin_head
)->ds_bp
;
3744 dsl_dataset_phys(origin_head
)->ds_bp
=
3745 dsl_dataset_phys(clone
)->ds_bp
;
3746 dsl_dataset_phys(clone
)->ds_bp
= tmp
;
3747 rrw_exit(&origin_head
->ds_bp_rwlock
, FTAG
);
3748 rrw_exit(&clone
->ds_bp_rwlock
, FTAG
);
3751 /* set dd_*_bytes */
3753 int64_t dused
, dcomp
, duncomp
;
3754 uint64_t cdl_used
, cdl_comp
, cdl_uncomp
;
3755 uint64_t odl_used
, odl_comp
, odl_uncomp
;
3757 ASSERT3U(dsl_dir_phys(clone
->ds_dir
)->
3758 dd_used_breakdown
[DD_USED_SNAP
], ==, 0);
3760 dsl_deadlist_space(&clone
->ds_deadlist
,
3761 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3762 dsl_deadlist_space(&origin_head
->ds_deadlist
,
3763 &odl_used
, &odl_comp
, &odl_uncomp
);
3765 dused
= dsl_dataset_phys(clone
)->ds_referenced_bytes
+
3767 (dsl_dataset_phys(origin_head
)->ds_referenced_bytes
+
3769 dcomp
= dsl_dataset_phys(clone
)->ds_compressed_bytes
+
3771 (dsl_dataset_phys(origin_head
)->ds_compressed_bytes
+
3773 duncomp
= dsl_dataset_phys(clone
)->ds_uncompressed_bytes
+
3775 (dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
+
3778 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_HEAD
,
3779 dused
, dcomp
, duncomp
, tx
);
3780 dsl_dir_diduse_space(clone
->ds_dir
, DD_USED_HEAD
,
3781 -dused
, -dcomp
, -duncomp
, tx
);
3784 * The difference in the space used by snapshots is the
3785 * difference in snapshot space due to the head's
3786 * deadlist (since that's the only thing that's
3787 * changing that affects the snapused).
3789 dsl_deadlist_space_range(&clone
->ds_deadlist
,
3790 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3791 &cdl_used
, &cdl_comp
, &cdl_uncomp
);
3792 dsl_deadlist_space_range(&origin_head
->ds_deadlist
,
3793 origin_head
->ds_dir
->dd_origin_txg
, UINT64_MAX
,
3794 &odl_used
, &odl_comp
, &odl_uncomp
);
3795 dsl_dir_transfer_space(origin_head
->ds_dir
, cdl_used
- odl_used
,
3796 DD_USED_HEAD
, DD_USED_SNAP
, tx
);
3799 /* swap ds_*_bytes */
3800 SWITCH64(dsl_dataset_phys(origin_head
)->ds_referenced_bytes
,
3801 dsl_dataset_phys(clone
)->ds_referenced_bytes
);
3802 SWITCH64(dsl_dataset_phys(origin_head
)->ds_compressed_bytes
,
3803 dsl_dataset_phys(clone
)->ds_compressed_bytes
);
3804 SWITCH64(dsl_dataset_phys(origin_head
)->ds_uncompressed_bytes
,
3805 dsl_dataset_phys(clone
)->ds_uncompressed_bytes
);
3806 SWITCH64(dsl_dataset_phys(origin_head
)->ds_unique_bytes
,
3807 dsl_dataset_phys(clone
)->ds_unique_bytes
);
3809 /* apply any parent delta for change in unconsumed refreservation */
3810 dsl_dir_diduse_space(origin_head
->ds_dir
, DD_USED_REFRSRV
,
3811 unused_refres_delta
, 0, 0, tx
);
3816 dsl_deadlist_close(&clone
->ds_deadlist
);
3817 dsl_deadlist_close(&origin_head
->ds_deadlist
);
3818 SWITCH64(dsl_dataset_phys(origin_head
)->ds_deadlist_obj
,
3819 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3820 dsl_deadlist_open(&clone
->ds_deadlist
, dp
->dp_meta_objset
,
3821 dsl_dataset_phys(clone
)->ds_deadlist_obj
);
3822 dsl_deadlist_open(&origin_head
->ds_deadlist
, dp
->dp_meta_objset
,
3823 dsl_dataset_phys(origin_head
)->ds_deadlist_obj
);
3824 dsl_dataset_swap_remap_deadlists(clone
, origin_head
, tx
);
3826 dsl_scan_ds_clone_swapped(origin_head
, clone
, tx
);
3828 spa_history_log_internal_ds(clone
, "clone swap", tx
,
3829 "parent=%s", origin_head
->ds_dir
->dd_myname
);
3833 * Given a pool name and a dataset object number in that pool,
3834 * return the name of that dataset.
3837 dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
)
3843 error
= dsl_pool_hold(pname
, FTAG
, &dp
);
3847 error
= dsl_dataset_hold_obj(dp
, obj
, FTAG
, &ds
);
3849 dsl_dataset_name(ds
, buf
);
3850 dsl_dataset_rele(ds
, FTAG
);
3852 dsl_pool_rele(dp
, FTAG
);
3858 dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
3859 uint64_t asize
, uint64_t inflight
, uint64_t *used
, uint64_t *ref_rsrv
)
3863 ASSERT3S(asize
, >, 0);
3866 * *ref_rsrv is the portion of asize that will come from any
3867 * unconsumed refreservation space.
3871 mutex_enter(&ds
->ds_lock
);
3873 * Make a space adjustment for reserved bytes.
3875 if (ds
->ds_reserved
> dsl_dataset_phys(ds
)->ds_unique_bytes
) {
3877 ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3879 (ds
->ds_reserved
- dsl_dataset_phys(ds
)->ds_unique_bytes
);
3881 asize
- MIN(asize
, parent_delta(ds
, asize
+ inflight
));
3884 if (!check_quota
|| ds
->ds_quota
== 0) {
3885 mutex_exit(&ds
->ds_lock
);
3889 * If they are requesting more space, and our current estimate
3890 * is over quota, they get to try again unless the actual
3891 * on-disk is over quota and there are no pending changes (which
3892 * may free up space for us).
3894 if (dsl_dataset_phys(ds
)->ds_referenced_bytes
+ inflight
>=
3897 dsl_dataset_phys(ds
)->ds_referenced_bytes
< ds
->ds_quota
)
3898 error
= SET_ERROR(ERESTART
);
3900 error
= SET_ERROR(EDQUOT
);
3902 mutex_exit(&ds
->ds_lock
);
3907 typedef struct dsl_dataset_set_qr_arg
{
3908 const char *ddsqra_name
;
3909 zprop_source_t ddsqra_source
;
3910 uint64_t ddsqra_value
;
3911 } dsl_dataset_set_qr_arg_t
;
3916 dsl_dataset_set_refquota_check(void *arg
, dmu_tx_t
*tx
)
3918 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3919 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3924 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFQUOTA
)
3925 return (SET_ERROR(ENOTSUP
));
3927 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
3931 if (ds
->ds_is_snapshot
) {
3932 dsl_dataset_rele(ds
, FTAG
);
3933 return (SET_ERROR(EINVAL
));
3936 error
= dsl_prop_predict(ds
->ds_dir
,
3937 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3938 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
3940 dsl_dataset_rele(ds
, FTAG
);
3945 dsl_dataset_rele(ds
, FTAG
);
3949 if (newval
< dsl_dataset_phys(ds
)->ds_referenced_bytes
||
3950 newval
< ds
->ds_reserved
) {
3951 dsl_dataset_rele(ds
, FTAG
);
3952 return (SET_ERROR(ENOSPC
));
3955 dsl_dataset_rele(ds
, FTAG
);
3960 dsl_dataset_set_refquota_sync(void *arg
, dmu_tx_t
*tx
)
3962 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
3963 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
3964 dsl_dataset_t
*ds
= NULL
;
3967 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
3969 dsl_prop_set_sync_impl(ds
,
3970 zfs_prop_to_name(ZFS_PROP_REFQUOTA
),
3971 ddsqra
->ddsqra_source
, sizeof (ddsqra
->ddsqra_value
), 1,
3972 &ddsqra
->ddsqra_value
, tx
);
3974 VERIFY0(dsl_prop_get_int_ds(ds
,
3975 zfs_prop_to_name(ZFS_PROP_REFQUOTA
), &newval
));
3977 if (ds
->ds_quota
!= newval
) {
3978 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
3979 ds
->ds_quota
= newval
;
3981 dsl_dataset_rele(ds
, FTAG
);
3985 dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
3988 dsl_dataset_set_qr_arg_t ddsqra
;
3990 ddsqra
.ddsqra_name
= dsname
;
3991 ddsqra
.ddsqra_source
= source
;
3992 ddsqra
.ddsqra_value
= refquota
;
3994 return (dsl_sync_task(dsname
, dsl_dataset_set_refquota_check
,
3995 dsl_dataset_set_refquota_sync
, &ddsqra
, 0,
3996 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
4000 dsl_dataset_set_refreservation_check(void *arg
, dmu_tx_t
*tx
)
4002 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4003 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4006 uint64_t newval
, unique
;
4008 if (spa_version(dp
->dp_spa
) < SPA_VERSION_REFRESERVATION
)
4009 return (SET_ERROR(ENOTSUP
));
4011 error
= dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
);
4015 if (ds
->ds_is_snapshot
) {
4016 dsl_dataset_rele(ds
, FTAG
);
4017 return (SET_ERROR(EINVAL
));
4020 error
= dsl_prop_predict(ds
->ds_dir
,
4021 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
4022 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, &newval
);
4024 dsl_dataset_rele(ds
, FTAG
);
4029 * If we are doing the preliminary check in open context, the
4030 * space estimates may be inaccurate.
4032 if (!dmu_tx_is_syncing(tx
)) {
4033 dsl_dataset_rele(ds
, FTAG
);
4037 mutex_enter(&ds
->ds_lock
);
4038 if (!DS_UNIQUE_IS_ACCURATE(ds
))
4039 dsl_dataset_recalc_head_uniq(ds
);
4040 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
4041 mutex_exit(&ds
->ds_lock
);
4043 if (MAX(unique
, newval
) > MAX(unique
, ds
->ds_reserved
)) {
4044 uint64_t delta
= MAX(unique
, newval
) -
4045 MAX(unique
, ds
->ds_reserved
);
4048 dsl_dir_space_available(ds
->ds_dir
, NULL
, 0, B_TRUE
) ||
4049 (ds
->ds_quota
> 0 && newval
> ds
->ds_quota
)) {
4050 dsl_dataset_rele(ds
, FTAG
);
4051 return (SET_ERROR(ENOSPC
));
4055 dsl_dataset_rele(ds
, FTAG
);
4060 dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
4061 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
)
4067 dsl_prop_set_sync_impl(ds
, zfs_prop_to_name(ZFS_PROP_REFRESERVATION
),
4068 source
, sizeof (value
), 1, &value
, tx
);
4070 VERIFY0(dsl_prop_get_int_ds(ds
,
4071 zfs_prop_to_name(ZFS_PROP_REFRESERVATION
), &newval
));
4073 dmu_buf_will_dirty(ds
->ds_dbuf
, tx
);
4074 mutex_enter(&ds
->ds_dir
->dd_lock
);
4075 mutex_enter(&ds
->ds_lock
);
4076 ASSERT(DS_UNIQUE_IS_ACCURATE(ds
));
4077 unique
= dsl_dataset_phys(ds
)->ds_unique_bytes
;
4078 delta
= MAX(0, (int64_t)(newval
- unique
)) -
4079 MAX(0, (int64_t)(ds
->ds_reserved
- unique
));
4080 ds
->ds_reserved
= newval
;
4081 mutex_exit(&ds
->ds_lock
);
4083 dsl_dir_diduse_space(ds
->ds_dir
, DD_USED_REFRSRV
, delta
, 0, 0, tx
);
4084 mutex_exit(&ds
->ds_dir
->dd_lock
);
4088 dsl_dataset_set_refreservation_sync(void *arg
, dmu_tx_t
*tx
)
4090 dsl_dataset_set_qr_arg_t
*ddsqra
= arg
;
4091 dsl_pool_t
*dp
= dmu_tx_pool(tx
);
4092 dsl_dataset_t
*ds
= NULL
;
4094 VERIFY0(dsl_dataset_hold(dp
, ddsqra
->ddsqra_name
, FTAG
, &ds
));
4095 dsl_dataset_set_refreservation_sync_impl(ds
,
4096 ddsqra
->ddsqra_source
, ddsqra
->ddsqra_value
, tx
);
4097 dsl_dataset_rele(ds
, FTAG
);
4101 dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
4102 uint64_t refreservation
)
4104 dsl_dataset_set_qr_arg_t ddsqra
;
4106 ddsqra
.ddsqra_name
= dsname
;
4107 ddsqra
.ddsqra_source
= source
;
4108 ddsqra
.ddsqra_value
= refreservation
;
4110 return (dsl_sync_task(dsname
, dsl_dataset_set_refreservation_check
,
4111 dsl_dataset_set_refreservation_sync
, &ddsqra
, 0,
4112 ZFS_SPACE_CHECK_EXTRA_RESERVED
));
4116 * Return (in *usedp) the amount of space written in new that is not
4117 * present in oldsnap. New may be a snapshot or the head. Old must be
4118 * a snapshot before new, in new's filesystem (or its origin). If not then
4119 * fail and return EINVAL.
4121 * The written space is calculated by considering two components: First, we
4122 * ignore any freed space, and calculate the written as new's used space
4123 * minus old's used space. Next, we add in the amount of space that was freed
4124 * between the two snapshots, thus reducing new's used space relative to old's.
4125 * Specifically, this is the space that was born before old->ds_creation_txg,
4126 * and freed before new (ie. on new's deadlist or a previous deadlist).
4128 * space freed [---------------------]
4129 * snapshots ---O-------O--------O-------O------
4133 dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
4134 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4138 dsl_pool_t
*dp
= new->ds_dir
->dd_pool
;
4140 ASSERT(dsl_pool_config_held(dp
));
4143 *usedp
+= dsl_dataset_phys(new)->ds_referenced_bytes
;
4144 *usedp
-= dsl_dataset_phys(oldsnap
)->ds_referenced_bytes
;
4147 *compp
+= dsl_dataset_phys(new)->ds_compressed_bytes
;
4148 *compp
-= dsl_dataset_phys(oldsnap
)->ds_compressed_bytes
;
4151 *uncompp
+= dsl_dataset_phys(new)->ds_uncompressed_bytes
;
4152 *uncompp
-= dsl_dataset_phys(oldsnap
)->ds_uncompressed_bytes
;
4154 snapobj
= new->ds_object
;
4155 while (snapobj
!= oldsnap
->ds_object
) {
4156 dsl_dataset_t
*snap
;
4157 uint64_t used
, comp
, uncomp
;
4159 if (snapobj
== new->ds_object
) {
4162 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &snap
);
4167 if (dsl_dataset_phys(snap
)->ds_prev_snap_txg
==
4168 dsl_dataset_phys(oldsnap
)->ds_creation_txg
) {
4170 * The blocks in the deadlist can not be born after
4171 * ds_prev_snap_txg, so get the whole deadlist space,
4172 * which is more efficient (especially for old-format
4173 * deadlists). Unfortunately the deadlist code
4174 * doesn't have enough information to make this
4175 * optimization itself.
4177 dsl_deadlist_space(&snap
->ds_deadlist
,
4178 &used
, &comp
, &uncomp
);
4180 dsl_deadlist_space_range(&snap
->ds_deadlist
,
4181 0, dsl_dataset_phys(oldsnap
)->ds_creation_txg
,
4182 &used
, &comp
, &uncomp
);
4189 * If we get to the beginning of the chain of snapshots
4190 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
4191 * was not a snapshot of/before new.
4193 snapobj
= dsl_dataset_phys(snap
)->ds_prev_snap_obj
;
4195 dsl_dataset_rele(snap
, FTAG
);
4197 err
= SET_ERROR(EINVAL
);
4206 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
4207 * lastsnap, and all snapshots in between are deleted.
4209 * blocks that would be freed [---------------------------]
4210 * snapshots ---O-------O--------O-------O--------O
4211 * firstsnap lastsnap
4213 * This is the set of blocks that were born after the snap before firstsnap,
4214 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
4215 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
4216 * We calculate this by iterating over the relevant deadlists (from the snap
4217 * after lastsnap, backward to the snap after firstsnap), summing up the
4218 * space on the deadlist that was born after the snap before firstsnap.
4221 dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
,
4222 dsl_dataset_t
*lastsnap
,
4223 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
)
4227 dsl_pool_t
*dp
= firstsnap
->ds_dir
->dd_pool
;
4229 ASSERT(firstsnap
->ds_is_snapshot
);
4230 ASSERT(lastsnap
->ds_is_snapshot
);
4233 * Check that the snapshots are in the same dsl_dir, and firstsnap
4234 * is before lastsnap.
4236 if (firstsnap
->ds_dir
!= lastsnap
->ds_dir
||
4237 dsl_dataset_phys(firstsnap
)->ds_creation_txg
>
4238 dsl_dataset_phys(lastsnap
)->ds_creation_txg
)
4239 return (SET_ERROR(EINVAL
));
4241 *usedp
= *compp
= *uncompp
= 0;
4243 snapobj
= dsl_dataset_phys(lastsnap
)->ds_next_snap_obj
;
4244 while (snapobj
!= firstsnap
->ds_object
) {
4246 uint64_t used
, comp
, uncomp
;
4248 err
= dsl_dataset_hold_obj(dp
, snapobj
, FTAG
, &ds
);
4252 dsl_deadlist_space_range(&ds
->ds_deadlist
,
4253 dsl_dataset_phys(firstsnap
)->ds_prev_snap_txg
, UINT64_MAX
,
4254 &used
, &comp
, &uncomp
);
4259 snapobj
= dsl_dataset_phys(ds
)->ds_prev_snap_obj
;
4260 ASSERT3U(snapobj
, !=, 0);
4261 dsl_dataset_rele(ds
, FTAG
);
4267 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
4268 * For example, they could both be snapshots of the same filesystem, and
4269 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
4270 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
4271 * filesystem. Or 'earlier' could be the origin's origin.
4273 * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
4276 dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
4277 uint64_t earlier_txg
)
4279 dsl_pool_t
*dp
= later
->ds_dir
->dd_pool
;
4283 ASSERT(dsl_pool_config_held(dp
));
4284 ASSERT(earlier
->ds_is_snapshot
|| earlier_txg
!= 0);
4286 if (earlier_txg
== 0)
4287 earlier_txg
= dsl_dataset_phys(earlier
)->ds_creation_txg
;
4289 if (later
->ds_is_snapshot
&&
4290 earlier_txg
>= dsl_dataset_phys(later
)->ds_creation_txg
)
4293 if (later
->ds_dir
== earlier
->ds_dir
)
4295 if (!dsl_dir_is_clone(later
->ds_dir
))
4298 if (dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
== earlier
->ds_object
)
4300 dsl_dataset_t
*origin
;
4301 error
= dsl_dataset_hold_obj(dp
,
4302 dsl_dir_phys(later
->ds_dir
)->dd_origin_obj
, FTAG
, &origin
);
4305 ret
= dsl_dataset_is_before(origin
, earlier
, earlier_txg
);
4306 dsl_dataset_rele(origin
, FTAG
);
4311 dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4313 objset_t
*mos
= ds
->ds_dir
->dd_pool
->dp_meta_objset
;
4314 dmu_object_zapify(mos
, ds
->ds_object
, DMU_OT_DSL_DATASET
, tx
);
4318 dsl_dataset_is_zapified(dsl_dataset_t
*ds
)
4320 dmu_object_info_t doi
;
4322 dmu_object_info_from_db(ds
->ds_dbuf
, &doi
);
4323 return (doi
.doi_type
== DMU_OTN_ZAP_METADATA
);
4327 dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
)
4329 return (dsl_dataset_is_zapified(ds
) &&
4330 zap_contains(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4331 ds
->ds_object
, DS_FIELD_RESUME_TOGUID
) == 0);
4335 dsl_dataset_get_remap_deadlist_object(dsl_dataset_t
*ds
)
4337 uint64_t remap_deadlist_obj
;
4340 if (!dsl_dataset_is_zapified(ds
))
4343 err
= zap_lookup(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4344 DS_FIELD_REMAP_DEADLIST
, sizeof (remap_deadlist_obj
), 1,
4345 &remap_deadlist_obj
);
4348 VERIFY3S(err
, ==, ENOENT
);
4352 ASSERT(remap_deadlist_obj
!= 0);
4353 return (remap_deadlist_obj
);
4357 dsl_dataset_remap_deadlist_exists(dsl_dataset_t
*ds
)
4359 EQUIV(dsl_deadlist_is_open(&ds
->ds_remap_deadlist
),
4360 dsl_dataset_get_remap_deadlist_object(ds
) != 0);
4361 return (dsl_deadlist_is_open(&ds
->ds_remap_deadlist
));
4365 dsl_dataset_set_remap_deadlist_object(dsl_dataset_t
*ds
, uint64_t obj
,
4369 dsl_dataset_zapify(ds
, tx
);
4370 VERIFY0(zap_add(ds
->ds_dir
->dd_pool
->dp_meta_objset
, ds
->ds_object
,
4371 DS_FIELD_REMAP_DEADLIST
, sizeof (obj
), 1, &obj
, tx
));
4375 dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4377 VERIFY0(zap_remove(ds
->ds_dir
->dd_pool
->dp_meta_objset
,
4378 ds
->ds_object
, DS_FIELD_REMAP_DEADLIST
, tx
));
4382 dsl_dataset_destroy_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4384 uint64_t remap_deadlist_object
;
4385 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4387 ASSERT(dmu_tx_is_syncing(tx
));
4388 ASSERT(dsl_dataset_remap_deadlist_exists(ds
));
4390 remap_deadlist_object
= ds
->ds_remap_deadlist
.dl_object
;
4391 dsl_deadlist_close(&ds
->ds_remap_deadlist
);
4392 dsl_deadlist_free(spa_meta_objset(spa
), remap_deadlist_object
, tx
);
4393 dsl_dataset_unset_remap_deadlist_object(ds
, tx
);
4394 spa_feature_decr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4398 dsl_dataset_create_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
)
4400 uint64_t remap_deadlist_obj
;
4401 spa_t
*spa
= ds
->ds_dir
->dd_pool
->dp_spa
;
4403 ASSERT(dmu_tx_is_syncing(tx
));
4404 ASSERT(MUTEX_HELD(&ds
->ds_remap_deadlist_lock
));
4406 * Currently we only create remap deadlists when there are indirect
4407 * vdevs with referenced mappings.
4409 ASSERT(spa_feature_is_active(spa
, SPA_FEATURE_DEVICE_REMOVAL
));
4411 remap_deadlist_obj
= dsl_deadlist_clone(
4412 &ds
->ds_deadlist
, UINT64_MAX
,
4413 dsl_dataset_phys(ds
)->ds_prev_snap_obj
, tx
);
4414 dsl_dataset_set_remap_deadlist_object(ds
,
4415 remap_deadlist_obj
, tx
);
4416 dsl_deadlist_open(&ds
->ds_remap_deadlist
, spa_meta_objset(spa
),
4417 remap_deadlist_obj
);
4418 spa_feature_incr(spa
, SPA_FEATURE_OBSOLETE_COUNTS
, tx
);
4421 #if defined(_KERNEL)
4423 module_param(zfs_max_recordsize
, int, 0644);
4424 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4426 /* Limited to 1M on 32-bit platforms due to lack of virtual address space */
4427 module_param(zfs_max_recordsize
, int, 0444);
4428 MODULE_PARM_DESC(zfs_max_recordsize
, "Max allowed record size");
4431 EXPORT_SYMBOL(dsl_dataset_hold
);
4432 EXPORT_SYMBOL(dsl_dataset_hold_flags
);
4433 EXPORT_SYMBOL(dsl_dataset_hold_obj
);
4434 EXPORT_SYMBOL(dsl_dataset_hold_obj_flags
);
4435 EXPORT_SYMBOL(dsl_dataset_own
);
4436 EXPORT_SYMBOL(dsl_dataset_own_obj
);
4437 EXPORT_SYMBOL(dsl_dataset_name
);
4438 EXPORT_SYMBOL(dsl_dataset_rele
);
4439 EXPORT_SYMBOL(dsl_dataset_rele_flags
);
4440 EXPORT_SYMBOL(dsl_dataset_disown
);
4441 EXPORT_SYMBOL(dsl_dataset_tryown
);
4442 EXPORT_SYMBOL(dsl_dataset_create_sync
);
4443 EXPORT_SYMBOL(dsl_dataset_create_sync_dd
);
4444 EXPORT_SYMBOL(dsl_dataset_snapshot_check
);
4445 EXPORT_SYMBOL(dsl_dataset_snapshot_sync
);
4446 EXPORT_SYMBOL(dsl_dataset_promote
);
4447 EXPORT_SYMBOL(dsl_dataset_user_hold
);
4448 EXPORT_SYMBOL(dsl_dataset_user_release
);
4449 EXPORT_SYMBOL(dsl_dataset_get_holds
);
4450 EXPORT_SYMBOL(dsl_dataset_get_blkptr
);
4451 EXPORT_SYMBOL(dsl_dataset_get_spa
);
4452 EXPORT_SYMBOL(dsl_dataset_modified_since_snap
);
4453 EXPORT_SYMBOL(dsl_dataset_space_written
);
4454 EXPORT_SYMBOL(dsl_dataset_space_wouldfree
);
4455 EXPORT_SYMBOL(dsl_dataset_sync
);
4456 EXPORT_SYMBOL(dsl_dataset_block_born
);
4457 EXPORT_SYMBOL(dsl_dataset_block_kill
);
4458 EXPORT_SYMBOL(dsl_dataset_dirty
);
4459 EXPORT_SYMBOL(dsl_dataset_stats
);
4460 EXPORT_SYMBOL(dsl_dataset_fast_stat
);
4461 EXPORT_SYMBOL(dsl_dataset_space
);
4462 EXPORT_SYMBOL(dsl_dataset_fsid_guid
);
4463 EXPORT_SYMBOL(dsl_dsobj_to_dsname
);
4464 EXPORT_SYMBOL(dsl_dataset_check_quota
);
4465 EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl
);
4466 EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl
);