]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
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. | |
7 | * | |
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. | |
12 | * | |
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] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
af073689 | 24 | * Copyright (c) 2011, 2017 by Delphix. All rights reserved. |
788eb90c | 25 | * Copyright (c) 2014, Joyent, Inc. All rights reserved. |
36f92e93 | 26 | * Copyright (c) 2014 RackTop Systems. |
0c66c32d | 27 | * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. |
a0bd735a | 28 | * Copyright (c) 2016 Actifio, Inc. All rights reserved. |
8c62a0d0 | 29 | * Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved. |
9b7b9cd3 | 30 | * Copyright 2017 Nexenta Systems, Inc. |
34dc7c2f BB |
31 | */ |
32 | ||
34dc7c2f BB |
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> | |
37abac6d | 39 | #include <sys/dmu_impl.h> |
34dc7c2f BB |
40 | #include <sys/dmu_tx.h> |
41 | #include <sys/arc.h> | |
42 | #include <sys/zio.h> | |
43 | #include <sys/zap.h> | |
9ae529ec | 44 | #include <sys/zfeature.h> |
34dc7c2f BB |
45 | #include <sys/unique.h> |
46 | #include <sys/zfs_context.h> | |
47 | #include <sys/zfs_ioctl.h> | |
48 | #include <sys/spa.h> | |
a1d477c2 | 49 | #include <sys/vdev.h> |
b128c09f | 50 | #include <sys/zfs_znode.h> |
572e2857 | 51 | #include <sys/zfs_onexit.h> |
45d1cae3 | 52 | #include <sys/zvol.h> |
428870ff BB |
53 | #include <sys/dsl_scan.h> |
54 | #include <sys/dsl_deadlist.h> | |
13fe0198 MA |
55 | #include <sys/dsl_destroy.h> |
56 | #include <sys/dsl_userhold.h> | |
da536844 | 57 | #include <sys/dsl_bookmark.h> |
f74b821a | 58 | #include <sys/policy.h> |
47dfff3b MA |
59 | #include <sys/dmu_send.h> |
60 | #include <sys/zio_compress.h> | |
61 | #include <zfs_fletcher.h> | |
3c67d83a | 62 | #include <sys/zio_checksum.h> |
34dc7c2f | 63 | |
f1512ee6 MA |
64 | /* |
65 | * The SPA supports block sizes up to 16MB. However, very large blocks | |
66 | * can have an impact on i/o latency (e.g. tying up a spinning disk for | |
67 | * ~300ms), and also potentially on the memory allocator. Therefore, | |
68 | * we do not allow the recordsize to be set larger than zfs_max_recordsize | |
69 | * (default 1MB). Larger blocks can be created by changing this tunable, | |
70 | * and pools with larger blocks can always be imported and used, regardless | |
71 | * of this setting. | |
72 | */ | |
73 | int zfs_max_recordsize = 1 * 1024 * 1024; | |
74 | ||
428870ff BB |
75 | #define SWITCH64(x, y) \ |
76 | { \ | |
77 | uint64_t __tmp = (x); \ | |
78 | (x) = (y); \ | |
79 | (y) = __tmp; \ | |
80 | } | |
81 | ||
34dc7c2f BB |
82 | #define DS_REF_MAX (1ULL << 62) |
83 | ||
d683ddbb | 84 | extern inline dsl_dataset_phys_t *dsl_dataset_phys(dsl_dataset_t *ds); |
d683ddbb | 85 | |
a1d477c2 MA |
86 | static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, |
87 | uint64_t obj, dmu_tx_t *tx); | |
88 | static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds, | |
89 | dmu_tx_t *tx); | |
90 | ||
8c62a0d0 DM |
91 | extern int spa_asize_inflation; |
92 | ||
0efd9791 AG |
93 | static zil_header_t zero_zil; |
94 | ||
34dc7c2f | 95 | /* |
4e33ba4c | 96 | * Figure out how much of this delta should be propagated to the dsl_dir |
34dc7c2f BB |
97 | * layer. If there's a refreservation, that space has already been |
98 | * partially accounted for in our ancestors. | |
99 | */ | |
100 | static int64_t | |
101 | parent_delta(dsl_dataset_t *ds, int64_t delta) | |
102 | { | |
d683ddbb | 103 | dsl_dataset_phys_t *ds_phys; |
34dc7c2f BB |
104 | uint64_t old_bytes, new_bytes; |
105 | ||
106 | if (ds->ds_reserved == 0) | |
107 | return (delta); | |
108 | ||
d683ddbb JG |
109 | ds_phys = dsl_dataset_phys(ds); |
110 | old_bytes = MAX(ds_phys->ds_unique_bytes, ds->ds_reserved); | |
111 | new_bytes = MAX(ds_phys->ds_unique_bytes + delta, ds->ds_reserved); | |
34dc7c2f BB |
112 | |
113 | ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta)); | |
114 | return (new_bytes - old_bytes); | |
115 | } | |
116 | ||
117 | void | |
428870ff | 118 | dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) |
34dc7c2f | 119 | { |
d6320ddb | 120 | int used, compressed, uncompressed; |
34dc7c2f BB |
121 | int64_t delta; |
122 | ||
d6320ddb BB |
123 | used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp); |
124 | compressed = BP_GET_PSIZE(bp); | |
125 | uncompressed = BP_GET_UCSIZE(bp); | |
126 | ||
428870ff | 127 | dprintf_bp(bp, "ds=%p", ds); |
34dc7c2f BB |
128 | |
129 | ASSERT(dmu_tx_is_syncing(tx)); | |
130 | /* It could have been compressed away to nothing */ | |
131 | if (BP_IS_HOLE(bp)) | |
132 | return; | |
133 | ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE); | |
9ae529ec | 134 | ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp))); |
34dc7c2f | 135 | if (ds == NULL) { |
29809a6c MA |
136 | dsl_pool_mos_diduse_space(tx->tx_pool, |
137 | used, compressed, uncompressed); | |
34dc7c2f BB |
138 | return; |
139 | } | |
428870ff | 140 | |
0efd9791 | 141 | ASSERT3U(bp->blk_birth, >, dsl_dataset_phys(ds)->ds_prev_snap_txg); |
a169a625 | 142 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
34dc7c2f BB |
143 | mutex_enter(&ds->ds_lock); |
144 | delta = parent_delta(ds, used); | |
d683ddbb JG |
145 | dsl_dataset_phys(ds)->ds_referenced_bytes += used; |
146 | dsl_dataset_phys(ds)->ds_compressed_bytes += compressed; | |
147 | dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed; | |
148 | dsl_dataset_phys(ds)->ds_unique_bytes += used; | |
3c67d83a | 149 | |
241b5415 MA |
150 | if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE) { |
151 | ds->ds_feature_activation_needed[SPA_FEATURE_LARGE_BLOCKS] = | |
152 | B_TRUE; | |
153 | } | |
3c67d83a | 154 | |
1c27024e | 155 | spa_feature_t f = zio_checksum_to_feature(BP_GET_CHECKSUM(bp)); |
3c67d83a TH |
156 | if (f != SPA_FEATURE_NONE) |
157 | ds->ds_feature_activation_needed[f] = B_TRUE; | |
158 | ||
34dc7c2f | 159 | mutex_exit(&ds->ds_lock); |
b128c09f BB |
160 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta, |
161 | compressed, uncompressed, tx); | |
162 | dsl_dir_transfer_space(ds->ds_dir, used - delta, | |
163 | DD_USED_REFRSRV, DD_USED_HEAD, tx); | |
34dc7c2f BB |
164 | } |
165 | ||
a1d477c2 MA |
166 | /* |
167 | * Called when the specified segment has been remapped, and is thus no | |
168 | * longer referenced in the head dataset. The vdev must be indirect. | |
169 | * | |
170 | * If the segment is referenced by a snapshot, put it on the remap deadlist. | |
171 | * Otherwise, add this segment to the obsolete spacemap. | |
172 | */ | |
173 | void | |
174 | dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev, uint64_t offset, | |
175 | uint64_t size, uint64_t birth, dmu_tx_t *tx) | |
176 | { | |
177 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
178 | ||
179 | ASSERT(dmu_tx_is_syncing(tx)); | |
180 | ASSERT(birth <= tx->tx_txg); | |
181 | ASSERT(!ds->ds_is_snapshot); | |
182 | ||
183 | if (birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) { | |
184 | spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx); | |
185 | } else { | |
186 | blkptr_t fakebp; | |
187 | dva_t *dva = &fakebp.blk_dva[0]; | |
188 | ||
189 | ASSERT(ds != NULL); | |
190 | ||
191 | mutex_enter(&ds->ds_remap_deadlist_lock); | |
192 | if (!dsl_dataset_remap_deadlist_exists(ds)) { | |
193 | dsl_dataset_create_remap_deadlist(ds, tx); | |
194 | } | |
195 | mutex_exit(&ds->ds_remap_deadlist_lock); | |
196 | ||
197 | BP_ZERO(&fakebp); | |
198 | fakebp.blk_birth = birth; | |
199 | DVA_SET_VDEV(dva, vdev); | |
200 | DVA_SET_OFFSET(dva, offset); | |
201 | DVA_SET_ASIZE(dva, size); | |
202 | ||
203 | dsl_deadlist_insert(&ds->ds_remap_deadlist, &fakebp, tx); | |
204 | } | |
205 | } | |
206 | ||
b128c09f | 207 | int |
428870ff BB |
208 | dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, |
209 | boolean_t async) | |
34dc7c2f | 210 | { |
b0bc7a84 MG |
211 | int used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp); |
212 | int compressed = BP_GET_PSIZE(bp); | |
213 | int uncompressed = BP_GET_UCSIZE(bp); | |
d6320ddb | 214 | |
34dc7c2f | 215 | if (BP_IS_HOLE(bp)) |
b128c09f | 216 | return (0); |
34dc7c2f | 217 | |
428870ff BB |
218 | ASSERT(dmu_tx_is_syncing(tx)); |
219 | ASSERT(bp->blk_birth <= tx->tx_txg); | |
220 | ||
34dc7c2f | 221 | if (ds == NULL) { |
428870ff | 222 | dsl_free(tx->tx_pool, tx->tx_txg, bp); |
29809a6c MA |
223 | dsl_pool_mos_diduse_space(tx->tx_pool, |
224 | -used, -compressed, -uncompressed); | |
b128c09f | 225 | return (used); |
34dc7c2f BB |
226 | } |
227 | ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool); | |
228 | ||
0c66c32d | 229 | ASSERT(!ds->ds_is_snapshot); |
34dc7c2f BB |
230 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
231 | ||
d683ddbb | 232 | if (bp->blk_birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) { |
34dc7c2f BB |
233 | int64_t delta; |
234 | ||
428870ff BB |
235 | dprintf_bp(bp, "freeing ds=%llu", ds->ds_object); |
236 | dsl_free(tx->tx_pool, tx->tx_txg, bp); | |
34dc7c2f BB |
237 | |
238 | mutex_enter(&ds->ds_lock); | |
d683ddbb | 239 | ASSERT(dsl_dataset_phys(ds)->ds_unique_bytes >= used || |
34dc7c2f BB |
240 | !DS_UNIQUE_IS_ACCURATE(ds)); |
241 | delta = parent_delta(ds, -used); | |
d683ddbb | 242 | dsl_dataset_phys(ds)->ds_unique_bytes -= used; |
34dc7c2f | 243 | mutex_exit(&ds->ds_lock); |
b128c09f | 244 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, |
34dc7c2f | 245 | delta, -compressed, -uncompressed, tx); |
b128c09f BB |
246 | dsl_dir_transfer_space(ds->ds_dir, -used - delta, |
247 | DD_USED_REFRSRV, DD_USED_HEAD, tx); | |
34dc7c2f BB |
248 | } else { |
249 | dprintf_bp(bp, "putting on dead list: %s", ""); | |
428870ff BB |
250 | if (async) { |
251 | /* | |
252 | * We are here as part of zio's write done callback, | |
253 | * which means we're a zio interrupt thread. We can't | |
254 | * call dsl_deadlist_insert() now because it may block | |
255 | * waiting for I/O. Instead, put bp on the deferred | |
256 | * queue and let dsl_pool_sync() finish the job. | |
257 | */ | |
258 | bplist_append(&ds->ds_pending_deadlist, bp); | |
259 | } else { | |
260 | dsl_deadlist_insert(&ds->ds_deadlist, bp, tx); | |
261 | } | |
34dc7c2f | 262 | ASSERT3U(ds->ds_prev->ds_object, ==, |
d683ddbb JG |
263 | dsl_dataset_phys(ds)->ds_prev_snap_obj); |
264 | ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0); | |
34dc7c2f | 265 | /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */ |
d683ddbb | 266 | if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == |
34dc7c2f | 267 | ds->ds_object && bp->blk_birth > |
d683ddbb | 268 | dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) { |
34dc7c2f BB |
269 | dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); |
270 | mutex_enter(&ds->ds_prev->ds_lock); | |
d683ddbb | 271 | dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used; |
34dc7c2f BB |
272 | mutex_exit(&ds->ds_prev->ds_lock); |
273 | } | |
428870ff | 274 | if (bp->blk_birth > ds->ds_dir->dd_origin_txg) { |
b128c09f BB |
275 | dsl_dir_transfer_space(ds->ds_dir, used, |
276 | DD_USED_HEAD, DD_USED_SNAP, tx); | |
277 | } | |
34dc7c2f BB |
278 | } |
279 | mutex_enter(&ds->ds_lock); | |
d683ddbb JG |
280 | ASSERT3U(dsl_dataset_phys(ds)->ds_referenced_bytes, >=, used); |
281 | dsl_dataset_phys(ds)->ds_referenced_bytes -= used; | |
282 | ASSERT3U(dsl_dataset_phys(ds)->ds_compressed_bytes, >=, compressed); | |
283 | dsl_dataset_phys(ds)->ds_compressed_bytes -= compressed; | |
284 | ASSERT3U(dsl_dataset_phys(ds)->ds_uncompressed_bytes, >=, uncompressed); | |
285 | dsl_dataset_phys(ds)->ds_uncompressed_bytes -= uncompressed; | |
34dc7c2f | 286 | mutex_exit(&ds->ds_lock); |
b128c09f BB |
287 | |
288 | return (used); | |
34dc7c2f BB |
289 | } |
290 | ||
39efbde7 GM |
291 | /* |
292 | * We have to release the fsid syncronously or we risk that a subsequent | |
293 | * mount of the same dataset will fail to unique_insert the fsid. This | |
294 | * failure would manifest itself as the fsid of this dataset changing | |
295 | * between mounts which makes NFS clients quite unhappy. | |
296 | */ | |
34dc7c2f | 297 | static void |
39efbde7 | 298 | dsl_dataset_evict_sync(void *dbu) |
34dc7c2f | 299 | { |
0c66c32d | 300 | dsl_dataset_t *ds = dbu; |
34dc7c2f | 301 | |
13fe0198 | 302 | ASSERT(ds->ds_owner == NULL); |
34dc7c2f | 303 | |
34dc7c2f | 304 | unique_remove(ds->ds_fsid_guid); |
39efbde7 GM |
305 | } |
306 | ||
307 | static void | |
308 | dsl_dataset_evict_async(void *dbu) | |
309 | { | |
310 | dsl_dataset_t *ds = dbu; | |
311 | ||
312 | ASSERT(ds->ds_owner == NULL); | |
313 | ||
314 | ds->ds_dbuf = NULL; | |
34dc7c2f | 315 | |
428870ff BB |
316 | if (ds->ds_objset != NULL) |
317 | dmu_objset_evict(ds->ds_objset); | |
34dc7c2f BB |
318 | |
319 | if (ds->ds_prev) { | |
13fe0198 | 320 | dsl_dataset_rele(ds->ds_prev, ds); |
34dc7c2f BB |
321 | ds->ds_prev = NULL; |
322 | } | |
323 | ||
428870ff | 324 | bplist_destroy(&ds->ds_pending_deadlist); |
a1d477c2 | 325 | if (dsl_deadlist_is_open(&ds->ds_deadlist)) |
428870ff | 326 | dsl_deadlist_close(&ds->ds_deadlist); |
a1d477c2 MA |
327 | if (dsl_deadlist_is_open(&ds->ds_remap_deadlist)) |
328 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
b128c09f | 329 | if (ds->ds_dir) |
0c66c32d | 330 | dsl_dir_async_rele(ds->ds_dir, ds); |
34dc7c2f BB |
331 | |
332 | ASSERT(!list_link_active(&ds->ds_synced_link)); | |
333 | ||
0eb21616 | 334 | list_destroy(&ds->ds_prop_cbs); |
34dc7c2f BB |
335 | mutex_destroy(&ds->ds_lock); |
336 | mutex_destroy(&ds->ds_opening_lock); | |
58c4aa00 | 337 | mutex_destroy(&ds->ds_sendstream_lock); |
a1d477c2 | 338 | mutex_destroy(&ds->ds_remap_deadlist_lock); |
13fe0198 | 339 | refcount_destroy(&ds->ds_longholds); |
cc9bb3e5 | 340 | rrw_destroy(&ds->ds_bp_rwlock); |
34dc7c2f BB |
341 | |
342 | kmem_free(ds, sizeof (dsl_dataset_t)); | |
343 | } | |
344 | ||
13fe0198 | 345 | int |
34dc7c2f BB |
346 | dsl_dataset_get_snapname(dsl_dataset_t *ds) |
347 | { | |
348 | dsl_dataset_phys_t *headphys; | |
349 | int err; | |
350 | dmu_buf_t *headdbuf; | |
351 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
352 | objset_t *mos = dp->dp_meta_objset; | |
353 | ||
354 | if (ds->ds_snapname[0]) | |
355 | return (0); | |
d683ddbb | 356 | if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) |
34dc7c2f BB |
357 | return (0); |
358 | ||
d683ddbb | 359 | err = dmu_bonus_hold(mos, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj, |
34dc7c2f | 360 | FTAG, &headdbuf); |
13fe0198 | 361 | if (err != 0) |
34dc7c2f BB |
362 | return (err); |
363 | headphys = headdbuf->db_data; | |
364 | err = zap_value_search(dp->dp_meta_objset, | |
365 | headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname); | |
d439f63f TC |
366 | if (err != 0 && zfs_recover == B_TRUE) { |
367 | err = 0; | |
368 | (void) snprintf(ds->ds_snapname, sizeof (ds->ds_snapname), | |
369 | "SNAPOBJ=%llu-ERR=%d", | |
370 | (unsigned long long)ds->ds_object, err); | |
371 | } | |
34dc7c2f BB |
372 | dmu_buf_rele(headdbuf, FTAG); |
373 | return (err); | |
374 | } | |
375 | ||
6772fb67 | 376 | int |
b128c09f | 377 | dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value) |
34dc7c2f | 378 | { |
b128c09f | 379 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
d683ddbb | 380 | uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; |
9b7b9cd3 | 381 | matchtype_t mt = 0; |
34dc7c2f BB |
382 | int err; |
383 | ||
d683ddbb | 384 | if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) |
9b7b9cd3 | 385 | mt = MT_NORMALIZE; |
34dc7c2f | 386 | |
b128c09f | 387 | err = zap_lookup_norm(mos, snapobj, name, 8, 1, |
34dc7c2f | 388 | value, mt, NULL, 0, NULL); |
9b7b9cd3 | 389 | if (err == ENOTSUP && (mt & MT_NORMALIZE)) |
b128c09f | 390 | err = zap_lookup(mos, snapobj, name, 8, 1, value); |
34dc7c2f BB |
391 | return (err); |
392 | } | |
393 | ||
13fe0198 | 394 | int |
788eb90c JJ |
395 | dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx, |
396 | boolean_t adj_cnt) | |
34dc7c2f | 397 | { |
b128c09f | 398 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
d683ddbb | 399 | uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; |
9b7b9cd3 | 400 | matchtype_t mt = 0; |
34dc7c2f BB |
401 | int err; |
402 | ||
428870ff BB |
403 | dsl_dir_snap_cmtime_update(ds->ds_dir); |
404 | ||
d683ddbb | 405 | if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) |
9b7b9cd3 | 406 | mt = MT_NORMALIZE; |
34dc7c2f | 407 | |
b128c09f | 408 | err = zap_remove_norm(mos, snapobj, name, mt, tx); |
9b7b9cd3 | 409 | if (err == ENOTSUP && (mt & MT_NORMALIZE)) |
b128c09f | 410 | err = zap_remove(mos, snapobj, name, tx); |
788eb90c JJ |
411 | |
412 | if (err == 0 && adj_cnt) | |
413 | dsl_fs_ss_count_adjust(ds->ds_dir, -1, | |
414 | DD_FIELD_SNAPSHOT_COUNT, tx); | |
415 | ||
34dc7c2f BB |
416 | return (err); |
417 | } | |
418 | ||
6ebebace JG |
419 | boolean_t |
420 | dsl_dataset_try_add_ref(dsl_pool_t *dp, dsl_dataset_t *ds, void *tag) | |
421 | { | |
6186e297 JG |
422 | dmu_buf_t *dbuf = ds->ds_dbuf; |
423 | boolean_t result = B_FALSE; | |
424 | ||
425 | if (dbuf != NULL && dmu_buf_try_add_ref(dbuf, dp->dp_meta_objset, | |
426 | ds->ds_object, DMU_BONUS_BLKID, tag)) { | |
427 | ||
428 | if (ds == dmu_buf_get_user(dbuf)) | |
429 | result = B_TRUE; | |
430 | else | |
431 | dmu_buf_rele(dbuf, tag); | |
432 | } | |
433 | ||
434 | return (result); | |
6ebebace JG |
435 | } |
436 | ||
13fe0198 | 437 | int |
b5256303 TC |
438 | dsl_dataset_hold_obj_flags(dsl_pool_t *dp, uint64_t dsobj, |
439 | ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp) | |
34dc7c2f | 440 | { |
34dc7c2f BB |
441 | objset_t *mos = dp->dp_meta_objset; |
442 | dmu_buf_t *dbuf; | |
443 | dsl_dataset_t *ds; | |
444 | int err; | |
572e2857 | 445 | dmu_object_info_t doi; |
34dc7c2f | 446 | |
13fe0198 | 447 | ASSERT(dsl_pool_config_held(dp)); |
34dc7c2f BB |
448 | |
449 | err = dmu_bonus_hold(mos, dsobj, tag, &dbuf); | |
13fe0198 | 450 | if (err != 0) |
34dc7c2f | 451 | return (err); |
572e2857 BB |
452 | |
453 | /* Make sure dsobj has the correct object type. */ | |
454 | dmu_object_info_from_db(dbuf, &doi); | |
fa86b5db | 455 | if (doi.doi_bonus_type != DMU_OT_DSL_DATASET) { |
3a84951d | 456 | dmu_buf_rele(dbuf, tag); |
2e528b49 | 457 | return (SET_ERROR(EINVAL)); |
3a84951d | 458 | } |
572e2857 | 459 | |
34dc7c2f BB |
460 | ds = dmu_buf_get_user(dbuf); |
461 | if (ds == NULL) { | |
d4ed6673 | 462 | dsl_dataset_t *winner = NULL; |
34dc7c2f | 463 | |
79c76d5b | 464 | ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP); |
34dc7c2f BB |
465 | ds->ds_dbuf = dbuf; |
466 | ds->ds_object = dsobj; | |
0c66c32d | 467 | ds->ds_is_snapshot = dsl_dataset_phys(ds)->ds_num_children != 0; |
98f72a53 | 468 | list_link_init(&ds->ds_synced_link); |
34dc7c2f | 469 | |
a1d477c2 MA |
470 | err = dsl_dir_hold_obj(dp, dsl_dataset_phys(ds)->ds_dir_obj, |
471 | NULL, ds, &ds->ds_dir); | |
472 | if (err != 0) { | |
473 | kmem_free(ds, sizeof (dsl_dataset_t)); | |
474 | dmu_buf_rele(dbuf, tag); | |
475 | return (err); | |
476 | } | |
477 | ||
34dc7c2f BB |
478 | mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL); |
479 | mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL); | |
37abac6d | 480 | mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL); |
a1d477c2 MA |
481 | mutex_init(&ds->ds_remap_deadlist_lock, |
482 | NULL, MUTEX_DEFAULT, NULL); | |
cc9bb3e5 | 483 | rrw_init(&ds->ds_bp_rwlock, B_FALSE); |
13fe0198 | 484 | refcount_create(&ds->ds_longholds); |
34dc7c2f | 485 | |
428870ff | 486 | bplist_create(&ds->ds_pending_deadlist); |
428870ff | 487 | |
37abac6d BP |
488 | list_create(&ds->ds_sendstreams, sizeof (dmu_sendarg_t), |
489 | offsetof(dmu_sendarg_t, dsa_link)); | |
490 | ||
0eb21616 JG |
491 | list_create(&ds->ds_prop_cbs, sizeof (dsl_prop_cb_record_t), |
492 | offsetof(dsl_prop_cb_record_t, cbr_ds_node)); | |
493 | ||
f1512ee6 | 494 | if (doi.doi_type == DMU_OTN_ZAP_METADATA) { |
241b5415 MA |
495 | spa_feature_t f; |
496 | ||
497 | for (f = 0; f < SPA_FEATURES; f++) { | |
498 | if (!(spa_feature_table[f].fi_flags & | |
499 | ZFEATURE_FLAG_PER_DATASET)) | |
500 | continue; | |
501 | err = zap_contains(mos, dsobj, | |
502 | spa_feature_table[f].fi_guid); | |
503 | if (err == 0) { | |
504 | ds->ds_feature_inuse[f] = B_TRUE; | |
505 | } else { | |
506 | ASSERT3U(err, ==, ENOENT); | |
507 | err = 0; | |
508 | } | |
19b3b1d2 | 509 | } |
f1512ee6 MA |
510 | } |
511 | ||
0c66c32d | 512 | if (!ds->ds_is_snapshot) { |
34dc7c2f | 513 | ds->ds_snapname[0] = '\0'; |
d683ddbb | 514 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { |
13fe0198 | 515 | err = dsl_dataset_hold_obj(dp, |
d683ddbb | 516 | dsl_dataset_phys(ds)->ds_prev_snap_obj, |
b128c09f | 517 | ds, &ds->ds_prev); |
34dc7c2f | 518 | } |
da536844 MA |
519 | if (doi.doi_type == DMU_OTN_ZAP_METADATA) { |
520 | int zaperr = zap_lookup(mos, ds->ds_object, | |
521 | DS_FIELD_BOOKMARK_NAMES, | |
522 | sizeof (ds->ds_bookmarks), 1, | |
523 | &ds->ds_bookmarks); | |
524 | if (zaperr != ENOENT) | |
525 | VERIFY0(zaperr); | |
526 | } | |
45d1cae3 BB |
527 | } else { |
528 | if (zfs_flags & ZFS_DEBUG_SNAPNAMES) | |
529 | err = dsl_dataset_get_snapname(ds); | |
d683ddbb JG |
530 | if (err == 0 && |
531 | dsl_dataset_phys(ds)->ds_userrefs_obj != 0) { | |
45d1cae3 BB |
532 | err = zap_count( |
533 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 534 | dsl_dataset_phys(ds)->ds_userrefs_obj, |
45d1cae3 BB |
535 | &ds->ds_userrefs); |
536 | } | |
34dc7c2f BB |
537 | } |
538 | ||
0c66c32d | 539 | if (err == 0 && !ds->ds_is_snapshot) { |
13fe0198 MA |
540 | err = dsl_prop_get_int_ds(ds, |
541 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), | |
542 | &ds->ds_reserved); | |
34dc7c2f | 543 | if (err == 0) { |
13fe0198 MA |
544 | err = dsl_prop_get_int_ds(ds, |
545 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
546 | &ds->ds_quota); | |
34dc7c2f | 547 | } |
34dc7c2f BB |
548 | } else { |
549 | ds->ds_reserved = ds->ds_quota = 0; | |
550 | } | |
551 | ||
a1d477c2 MA |
552 | dsl_deadlist_open(&ds->ds_deadlist, |
553 | mos, dsl_dataset_phys(ds)->ds_deadlist_obj); | |
554 | uint64_t remap_deadlist_obj = | |
555 | dsl_dataset_get_remap_deadlist_object(ds); | |
556 | if (remap_deadlist_obj != 0) { | |
557 | dsl_deadlist_open(&ds->ds_remap_deadlist, mos, | |
558 | remap_deadlist_obj); | |
559 | } | |
560 | ||
39efbde7 GM |
561 | dmu_buf_init_user(&ds->ds_dbu, dsl_dataset_evict_sync, |
562 | dsl_dataset_evict_async, &ds->ds_dbuf); | |
0c66c32d JG |
563 | if (err == 0) |
564 | winner = dmu_buf_set_user_ie(dbuf, &ds->ds_dbu); | |
565 | ||
566 | if (err != 0 || winner != NULL) { | |
428870ff BB |
567 | bplist_destroy(&ds->ds_pending_deadlist); |
568 | dsl_deadlist_close(&ds->ds_deadlist); | |
a1d477c2 MA |
569 | if (dsl_deadlist_is_open(&ds->ds_remap_deadlist)) |
570 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
b128c09f | 571 | if (ds->ds_prev) |
13fe0198 MA |
572 | dsl_dataset_rele(ds->ds_prev, ds); |
573 | dsl_dir_rele(ds->ds_dir, ds); | |
34dc7c2f BB |
574 | mutex_destroy(&ds->ds_lock); |
575 | mutex_destroy(&ds->ds_opening_lock); | |
58c4aa00 | 576 | mutex_destroy(&ds->ds_sendstream_lock); |
13fe0198 | 577 | refcount_destroy(&ds->ds_longholds); |
34dc7c2f | 578 | kmem_free(ds, sizeof (dsl_dataset_t)); |
13fe0198 | 579 | if (err != 0) { |
34dc7c2f BB |
580 | dmu_buf_rele(dbuf, tag); |
581 | return (err); | |
582 | } | |
583 | ds = winner; | |
584 | } else { | |
585 | ds->ds_fsid_guid = | |
d683ddbb | 586 | unique_insert(dsl_dataset_phys(ds)->ds_fsid_guid); |
39efbde7 GM |
587 | if (ds->ds_fsid_guid != |
588 | dsl_dataset_phys(ds)->ds_fsid_guid) { | |
589 | zfs_dbgmsg("ds_fsid_guid changed from " | |
590 | "%llx to %llx for pool %s dataset id %llu", | |
591 | (long long) | |
592 | dsl_dataset_phys(ds)->ds_fsid_guid, | |
593 | (long long)ds->ds_fsid_guid, | |
594 | spa_name(dp->dp_spa), | |
595 | dsobj); | |
596 | } | |
34dc7c2f BB |
597 | } |
598 | } | |
599 | ASSERT3P(ds->ds_dbuf, ==, dbuf); | |
d683ddbb JG |
600 | ASSERT3P(dsl_dataset_phys(ds), ==, dbuf->db_data); |
601 | ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0 || | |
b128c09f BB |
602 | spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN || |
603 | dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap); | |
34dc7c2f | 604 | *dsp = ds; |
b5256303 TC |
605 | |
606 | if ((flags & DS_HOLD_FLAG_DECRYPT) && ds->ds_dir->dd_crypto_obj != 0) { | |
607 | err = spa_keystore_create_mapping(dp->dp_spa, ds, ds); | |
608 | if (err != 0) { | |
609 | dsl_dataset_rele(ds, tag); | |
610 | return (SET_ERROR(EACCES)); | |
611 | } | |
612 | } | |
613 | ||
34dc7c2f BB |
614 | return (0); |
615 | } | |
616 | ||
b128c09f | 617 | int |
b5256303 TC |
618 | dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag, |
619 | dsl_dataset_t **dsp) | |
620 | { | |
621 | return (dsl_dataset_hold_obj_flags(dp, dsobj, 0, tag, dsp)); | |
622 | } | |
623 | ||
624 | int | |
625 | dsl_dataset_hold_flags(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags, | |
428870ff | 626 | void *tag, dsl_dataset_t **dsp) |
34dc7c2f BB |
627 | { |
628 | dsl_dir_t *dd; | |
b128c09f | 629 | const char *snapname; |
34dc7c2f | 630 | uint64_t obj; |
34dc7c2f | 631 | int err = 0; |
fcff0f35 | 632 | dsl_dataset_t *ds; |
34dc7c2f | 633 | |
13fe0198 MA |
634 | err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname); |
635 | if (err != 0) | |
34dc7c2f BB |
636 | return (err); |
637 | ||
13fe0198 | 638 | ASSERT(dsl_pool_config_held(dp)); |
d683ddbb | 639 | obj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
13fe0198 | 640 | if (obj != 0) |
b5256303 | 641 | err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, &ds); |
b128c09f | 642 | else |
2e528b49 | 643 | err = SET_ERROR(ENOENT); |
34dc7c2f | 644 | |
b128c09f BB |
645 | /* we may be looking for a snapshot */ |
646 | if (err == 0 && snapname != NULL) { | |
fcff0f35 | 647 | dsl_dataset_t *snap_ds; |
34dc7c2f | 648 | |
b128c09f | 649 | if (*snapname++ != '@') { |
b5256303 | 650 | dsl_dataset_rele_flags(ds, flags, tag); |
13fe0198 | 651 | dsl_dir_rele(dd, FTAG); |
2e528b49 | 652 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 653 | } |
34dc7c2f | 654 | |
b128c09f | 655 | dprintf("looking for snapshot '%s'\n", snapname); |
fcff0f35 | 656 | err = dsl_dataset_snap_lookup(ds, snapname, &obj); |
b5256303 TC |
657 | if (err == 0) { |
658 | err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, | |
659 | &snap_ds); | |
660 | } | |
661 | dsl_dataset_rele_flags(ds, flags, tag); | |
b128c09f | 662 | |
13fe0198 | 663 | if (err == 0) { |
fcff0f35 PD |
664 | mutex_enter(&snap_ds->ds_lock); |
665 | if (snap_ds->ds_snapname[0] == 0) | |
666 | (void) strlcpy(snap_ds->ds_snapname, snapname, | |
667 | sizeof (snap_ds->ds_snapname)); | |
668 | mutex_exit(&snap_ds->ds_lock); | |
669 | ds = snap_ds; | |
34dc7c2f | 670 | } |
34dc7c2f | 671 | } |
fcff0f35 PD |
672 | if (err == 0) |
673 | *dsp = ds; | |
13fe0198 | 674 | dsl_dir_rele(dd, FTAG); |
34dc7c2f BB |
675 | return (err); |
676 | } | |
677 | ||
678 | int | |
b5256303 TC |
679 | dsl_dataset_hold(dsl_pool_t *dp, const char *name, void *tag, |
680 | dsl_dataset_t **dsp) | |
681 | { | |
682 | return (dsl_dataset_hold_flags(dp, name, 0, tag, dsp)); | |
683 | } | |
684 | ||
685 | int | |
686 | dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags, | |
428870ff | 687 | void *tag, dsl_dataset_t **dsp) |
34dc7c2f | 688 | { |
b5256303 | 689 | int err = dsl_dataset_hold_obj_flags(dp, dsobj, flags, tag, dsp); |
13fe0198 MA |
690 | if (err != 0) |
691 | return (err); | |
692 | if (!dsl_dataset_tryown(*dsp, tag)) { | |
b5256303 | 693 | dsl_dataset_rele_flags(*dsp, flags, tag); |
13fe0198 | 694 | *dsp = NULL; |
2e528b49 | 695 | return (SET_ERROR(EBUSY)); |
13fe0198 MA |
696 | } |
697 | return (0); | |
698 | } | |
699 | ||
700 | int | |
b5256303 | 701 | dsl_dataset_own(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags, |
13fe0198 MA |
702 | void *tag, dsl_dataset_t **dsp) |
703 | { | |
b5256303 | 704 | int err = dsl_dataset_hold_flags(dp, name, flags, tag, dsp); |
13fe0198 | 705 | if (err != 0) |
b128c09f | 706 | return (err); |
13fe0198 | 707 | if (!dsl_dataset_tryown(*dsp, tag)) { |
b5256303 | 708 | dsl_dataset_rele_flags(*dsp, flags, tag); |
2e528b49 | 709 | return (SET_ERROR(EBUSY)); |
b128c09f BB |
710 | } |
711 | return (0); | |
34dc7c2f BB |
712 | } |
713 | ||
13fe0198 MA |
714 | /* |
715 | * See the comment above dsl_pool_hold() for details. In summary, a long | |
716 | * hold is used to prevent destruction of a dataset while the pool hold | |
717 | * is dropped, allowing other concurrent operations (e.g. spa_sync()). | |
718 | * | |
719 | * The dataset and pool must be held when this function is called. After it | |
720 | * is called, the pool hold may be released while the dataset is still held | |
721 | * and accessed. | |
722 | */ | |
723 | void | |
724 | dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag) | |
725 | { | |
726 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
727 | (void) refcount_add(&ds->ds_longholds, tag); | |
728 | } | |
729 | ||
730 | void | |
731 | dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag) | |
732 | { | |
733 | (void) refcount_remove(&ds->ds_longholds, tag); | |
734 | } | |
735 | ||
736 | /* Return B_TRUE if there are any long holds on this dataset. */ | |
737 | boolean_t | |
738 | dsl_dataset_long_held(dsl_dataset_t *ds) | |
739 | { | |
740 | return (!refcount_is_zero(&ds->ds_longholds)); | |
741 | } | |
742 | ||
34dc7c2f BB |
743 | void |
744 | dsl_dataset_name(dsl_dataset_t *ds, char *name) | |
745 | { | |
746 | if (ds == NULL) { | |
747 | (void) strcpy(name, "mos"); | |
748 | } else { | |
749 | dsl_dir_name(ds->ds_dir, name); | |
13fe0198 | 750 | VERIFY0(dsl_dataset_get_snapname(ds)); |
34dc7c2f | 751 | if (ds->ds_snapname[0]) { |
eca7b760 IK |
752 | VERIFY3U(strlcat(name, "@", ZFS_MAX_DATASET_NAME_LEN), |
753 | <, ZFS_MAX_DATASET_NAME_LEN); | |
b128c09f BB |
754 | /* |
755 | * We use a "recursive" mutex so that we | |
756 | * can call dprintf_ds() with ds_lock held. | |
757 | */ | |
34dc7c2f | 758 | if (!MUTEX_HELD(&ds->ds_lock)) { |
34dc7c2f | 759 | mutex_enter(&ds->ds_lock); |
eca7b760 IK |
760 | VERIFY3U(strlcat(name, ds->ds_snapname, |
761 | ZFS_MAX_DATASET_NAME_LEN), <, | |
762 | ZFS_MAX_DATASET_NAME_LEN); | |
34dc7c2f BB |
763 | mutex_exit(&ds->ds_lock); |
764 | } else { | |
eca7b760 IK |
765 | VERIFY3U(strlcat(name, ds->ds_snapname, |
766 | ZFS_MAX_DATASET_NAME_LEN), <, | |
767 | ZFS_MAX_DATASET_NAME_LEN); | |
34dc7c2f BB |
768 | } |
769 | } | |
770 | } | |
771 | } | |
772 | ||
eca7b760 IK |
773 | int |
774 | dsl_dataset_namelen(dsl_dataset_t *ds) | |
775 | { | |
eca7b760 IK |
776 | VERIFY0(dsl_dataset_get_snapname(ds)); |
777 | mutex_enter(&ds->ds_lock); | |
1c27024e | 778 | int len = strlen(ds->ds_snapname); |
a806cb6a CC |
779 | /* add '@' if ds is a snap */ |
780 | if (len > 0) | |
781 | len++; | |
782 | len += dsl_dir_namelen(ds->ds_dir); | |
eca7b760 IK |
783 | mutex_exit(&ds->ds_lock); |
784 | return (len); | |
785 | } | |
786 | ||
34dc7c2f | 787 | void |
b5256303 | 788 | dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag) |
b128c09f | 789 | { |
b5256303 TC |
790 | if (ds->ds_dir != NULL && ds->ds_dir->dd_crypto_obj != 0 && |
791 | (flags & DS_HOLD_FLAG_DECRYPT)) { | |
792 | (void) spa_keystore_remove_mapping(ds->ds_dir->dd_pool->dp_spa, | |
793 | ds->ds_object, ds); | |
cd32e5db TC |
794 | |
795 | /* | |
796 | * Encrypted datasets require that users only release their | |
797 | * decrypting reference after the dirty data has actually | |
798 | * been written out. This ensures that the mapping exists | |
799 | * when it is needed to write out dirty data. | |
800 | */ | |
801 | ASSERT(dmu_buf_user_refcount(ds->ds_dbuf) != 0 || | |
802 | !dsl_dataset_is_dirty(ds)); | |
b5256303 TC |
803 | } |
804 | ||
13fe0198 | 805 | dmu_buf_rele(ds->ds_dbuf, tag); |
b128c09f BB |
806 | } |
807 | ||
808 | void | |
b5256303 TC |
809 | dsl_dataset_rele(dsl_dataset_t *ds, void *tag) |
810 | { | |
811 | dsl_dataset_rele_flags(ds, 0, tag); | |
812 | } | |
813 | ||
814 | void | |
815 | dsl_dataset_disown(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag) | |
34dc7c2f | 816 | { |
945dd935 JG |
817 | ASSERT3P(ds->ds_owner, ==, tag); |
818 | ASSERT(ds->ds_dbuf != NULL); | |
b128c09f | 819 | |
34dc7c2f | 820 | mutex_enter(&ds->ds_lock); |
b128c09f | 821 | ds->ds_owner = NULL; |
34dc7c2f | 822 | mutex_exit(&ds->ds_lock); |
13fe0198 | 823 | dsl_dataset_long_rele(ds, tag); |
b5256303 | 824 | dsl_dataset_rele_flags(ds, flags, tag); |
34dc7c2f BB |
825 | } |
826 | ||
827 | boolean_t | |
13fe0198 | 828 | dsl_dataset_tryown(dsl_dataset_t *ds, void *tag) |
34dc7c2f | 829 | { |
b128c09f BB |
830 | boolean_t gotit = FALSE; |
831 | ||
47dfff3b | 832 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); |
34dc7c2f | 833 | mutex_enter(&ds->ds_lock); |
13fe0198 | 834 | if (ds->ds_owner == NULL && !DS_IS_INCONSISTENT(ds)) { |
428870ff | 835 | ds->ds_owner = tag; |
13fe0198 | 836 | dsl_dataset_long_hold(ds, tag); |
b128c09f | 837 | gotit = TRUE; |
34dc7c2f BB |
838 | } |
839 | mutex_exit(&ds->ds_lock); | |
b128c09f | 840 | return (gotit); |
34dc7c2f BB |
841 | } |
842 | ||
47dfff3b MA |
843 | boolean_t |
844 | dsl_dataset_has_owner(dsl_dataset_t *ds) | |
845 | { | |
846 | boolean_t rv; | |
847 | mutex_enter(&ds->ds_lock); | |
848 | rv = (ds->ds_owner != NULL); | |
849 | mutex_exit(&ds->ds_lock); | |
850 | return (rv); | |
851 | } | |
852 | ||
b5256303 | 853 | void |
241b5415 MA |
854 | dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) |
855 | { | |
856 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
857 | objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; | |
858 | uint64_t zero = 0; | |
859 | ||
860 | VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); | |
861 | ||
862 | spa_feature_incr(spa, f, tx); | |
863 | dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx); | |
864 | ||
865 | VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid, | |
866 | sizeof (zero), 1, &zero, tx)); | |
867 | } | |
868 | ||
869 | void | |
870 | dsl_dataset_deactivate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) | |
871 | { | |
872 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
873 | objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; | |
874 | ||
875 | VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); | |
876 | ||
877 | VERIFY0(zap_remove(mos, dsobj, spa_feature_table[f].fi_guid, tx)); | |
878 | spa_feature_decr(spa, f, tx); | |
879 | } | |
880 | ||
34dc7c2f | 881 | uint64_t |
b128c09f | 882 | dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin, |
b5256303 | 883 | dsl_crypto_params_t *dcp, uint64_t flags, dmu_tx_t *tx) |
34dc7c2f BB |
884 | { |
885 | dsl_pool_t *dp = dd->dd_pool; | |
886 | dmu_buf_t *dbuf; | |
887 | dsl_dataset_phys_t *dsphys; | |
888 | uint64_t dsobj; | |
889 | objset_t *mos = dp->dp_meta_objset; | |
890 | ||
b128c09f BB |
891 | if (origin == NULL) |
892 | origin = dp->dp_origin_snap; | |
893 | ||
34dc7c2f | 894 | ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp); |
d683ddbb | 895 | ASSERT(origin == NULL || dsl_dataset_phys(origin)->ds_num_children > 0); |
34dc7c2f | 896 | ASSERT(dmu_tx_is_syncing(tx)); |
d683ddbb | 897 | ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0); |
34dc7c2f BB |
898 | |
899 | dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, | |
900 | DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); | |
13fe0198 | 901 | VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); |
34dc7c2f BB |
902 | dmu_buf_will_dirty(dbuf, tx); |
903 | dsphys = dbuf->db_data; | |
b128c09f | 904 | bzero(dsphys, sizeof (dsl_dataset_phys_t)); |
34dc7c2f BB |
905 | dsphys->ds_dir_obj = dd->dd_object; |
906 | dsphys->ds_flags = flags; | |
907 | dsphys->ds_fsid_guid = unique_create(); | |
908 | (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, | |
909 | sizeof (dsphys->ds_guid)); | |
910 | dsphys->ds_snapnames_zapobj = | |
911 | zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP, | |
912 | DMU_OT_NONE, 0, tx); | |
913 | dsphys->ds_creation_time = gethrestime_sec(); | |
b128c09f | 914 | dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg; |
34dc7c2f | 915 | |
428870ff BB |
916 | if (origin == NULL) { |
917 | dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx); | |
918 | } else { | |
13fe0198 | 919 | dsl_dataset_t *ohds; /* head of the origin snapshot */ |
428870ff | 920 | |
34dc7c2f BB |
921 | dsphys->ds_prev_snap_obj = origin->ds_object; |
922 | dsphys->ds_prev_snap_txg = | |
d683ddbb | 923 | dsl_dataset_phys(origin)->ds_creation_txg; |
9ae529ec | 924 | dsphys->ds_referenced_bytes = |
d683ddbb | 925 | dsl_dataset_phys(origin)->ds_referenced_bytes; |
34dc7c2f | 926 | dsphys->ds_compressed_bytes = |
d683ddbb | 927 | dsl_dataset_phys(origin)->ds_compressed_bytes; |
34dc7c2f | 928 | dsphys->ds_uncompressed_bytes = |
d683ddbb | 929 | dsl_dataset_phys(origin)->ds_uncompressed_bytes; |
cc9bb3e5 | 930 | rrw_enter(&origin->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 931 | dsphys->ds_bp = dsl_dataset_phys(origin)->ds_bp; |
cc9bb3e5 | 932 | rrw_exit(&origin->ds_bp_rwlock, FTAG); |
4b20a6f5 MA |
933 | |
934 | /* | |
935 | * Inherit flags that describe the dataset's contents | |
936 | * (INCONSISTENT) or properties (Case Insensitive). | |
937 | */ | |
d683ddbb | 938 | dsphys->ds_flags |= dsl_dataset_phys(origin)->ds_flags & |
4b20a6f5 | 939 | (DS_FLAG_INCONSISTENT | DS_FLAG_CI_DATASET); |
34dc7c2f | 940 | |
1c27024e | 941 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
942 | if (origin->ds_feature_inuse[f]) |
943 | dsl_dataset_activate_feature(dsobj, f, tx); | |
944 | } | |
f1512ee6 | 945 | |
34dc7c2f | 946 | dmu_buf_will_dirty(origin->ds_dbuf, tx); |
d683ddbb | 947 | dsl_dataset_phys(origin)->ds_num_children++; |
34dc7c2f | 948 | |
13fe0198 | 949 | VERIFY0(dsl_dataset_hold_obj(dp, |
d683ddbb JG |
950 | dsl_dir_phys(origin->ds_dir)->dd_head_dataset_obj, |
951 | FTAG, &ohds)); | |
428870ff BB |
952 | dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist, |
953 | dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx); | |
954 | dsl_dataset_rele(ohds, FTAG); | |
955 | ||
b128c09f | 956 | if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) { |
d683ddbb JG |
957 | if (dsl_dataset_phys(origin)->ds_next_clones_obj == 0) { |
958 | dsl_dataset_phys(origin)->ds_next_clones_obj = | |
b128c09f BB |
959 | zap_create(mos, |
960 | DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx); | |
961 | } | |
13fe0198 | 962 | VERIFY0(zap_add_int(mos, |
d683ddbb JG |
963 | dsl_dataset_phys(origin)->ds_next_clones_obj, |
964 | dsobj, tx)); | |
b128c09f BB |
965 | } |
966 | ||
34dc7c2f | 967 | dmu_buf_will_dirty(dd->dd_dbuf, tx); |
d683ddbb | 968 | dsl_dir_phys(dd)->dd_origin_obj = origin->ds_object; |
428870ff | 969 | if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { |
d683ddbb | 970 | if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) { |
428870ff | 971 | dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx); |
d683ddbb | 972 | dsl_dir_phys(origin->ds_dir)->dd_clones = |
428870ff BB |
973 | zap_create(mos, |
974 | DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx); | |
975 | } | |
13fe0198 | 976 | VERIFY0(zap_add_int(mos, |
d683ddbb JG |
977 | dsl_dir_phys(origin->ds_dir)->dd_clones, |
978 | dsobj, tx)); | |
428870ff | 979 | } |
34dc7c2f BB |
980 | } |
981 | ||
b5256303 TC |
982 | /* handle encryption */ |
983 | dsl_dataset_create_crypt_sync(dsobj, dd, origin, dcp, tx); | |
984 | ||
34dc7c2f BB |
985 | if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) |
986 | dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; | |
987 | ||
988 | dmu_buf_rele(dbuf, FTAG); | |
989 | ||
990 | dmu_buf_will_dirty(dd->dd_dbuf, tx); | |
d683ddbb | 991 | dsl_dir_phys(dd)->dd_head_dataset_obj = dsobj; |
34dc7c2f BB |
992 | |
993 | return (dsobj); | |
994 | } | |
995 | ||
13fe0198 MA |
996 | static void |
997 | dsl_dataset_zero_zil(dsl_dataset_t *ds, dmu_tx_t *tx) | |
998 | { | |
999 | objset_t *os; | |
1000 | ||
1001 | VERIFY0(dmu_objset_from_ds(ds, &os)); | |
0efd9791 AG |
1002 | if (bcmp(&os->os_zil_header, &zero_zil, sizeof (zero_zil)) != 0) { |
1003 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
1004 | zio_t *zio; | |
1005 | ||
1006 | bzero(&os->os_zil_header, sizeof (os->os_zil_header)); | |
b5256303 | 1007 | if (os->os_encrypted) |
1b66810b | 1008 | os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE; |
0efd9791 AG |
1009 | |
1010 | zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); | |
1011 | dsl_dataset_sync(ds, zio, tx); | |
1012 | VERIFY0(zio_wait(zio)); | |
1013 | ||
1014 | /* dsl_dataset_sync_done will drop this reference. */ | |
1015 | dmu_buf_add_ref(ds->ds_dbuf, ds); | |
1016 | dsl_dataset_sync_done(ds, tx); | |
1017 | } | |
13fe0198 MA |
1018 | } |
1019 | ||
34dc7c2f BB |
1020 | uint64_t |
1021 | dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname, | |
b5256303 TC |
1022 | dsl_dataset_t *origin, uint64_t flags, cred_t *cr, |
1023 | dsl_crypto_params_t *dcp, dmu_tx_t *tx) | |
34dc7c2f BB |
1024 | { |
1025 | dsl_pool_t *dp = pdd->dd_pool; | |
1026 | uint64_t dsobj, ddobj; | |
1027 | dsl_dir_t *dd; | |
1028 | ||
13fe0198 | 1029 | ASSERT(dmu_tx_is_syncing(tx)); |
34dc7c2f BB |
1030 | ASSERT(lastname[0] != '@'); |
1031 | ||
b128c09f | 1032 | ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx); |
13fe0198 | 1033 | VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd)); |
34dc7c2f | 1034 | |
b5256303 | 1035 | dsobj = dsl_dataset_create_sync_dd(dd, origin, dcp, |
13fe0198 | 1036 | flags & ~DS_CREATE_FLAG_NODIRTY, tx); |
34dc7c2f BB |
1037 | |
1038 | dsl_deleg_set_create_perms(dd, tx, cr); | |
1039 | ||
788eb90c JJ |
1040 | /* |
1041 | * Since we're creating a new node we know it's a leaf, so we can | |
1042 | * initialize the counts if the limit feature is active. | |
1043 | */ | |
1044 | if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) { | |
1045 | uint64_t cnt = 0; | |
1046 | objset_t *os = dd->dd_pool->dp_meta_objset; | |
1047 | ||
1048 | dsl_dir_zapify(dd, tx); | |
1049 | VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT, | |
1050 | sizeof (cnt), 1, &cnt, tx)); | |
1051 | VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT, | |
1052 | sizeof (cnt), 1, &cnt, tx)); | |
1053 | } | |
1054 | ||
13fe0198 | 1055 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1056 | |
572e2857 BB |
1057 | /* |
1058 | * If we are creating a clone, make sure we zero out any stale | |
1059 | * data from the origin snapshots zil header. | |
1060 | */ | |
13fe0198 | 1061 | if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) { |
572e2857 | 1062 | dsl_dataset_t *ds; |
572e2857 | 1063 | |
13fe0198 MA |
1064 | VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds)); |
1065 | dsl_dataset_zero_zil(ds, tx); | |
572e2857 BB |
1066 | dsl_dataset_rele(ds, FTAG); |
1067 | } | |
1068 | ||
34dc7c2f BB |
1069 | return (dsobj); |
1070 | } | |
1071 | ||
34dc7c2f | 1072 | /* |
13fe0198 MA |
1073 | * The unique space in the head dataset can be calculated by subtracting |
1074 | * the space used in the most recent snapshot, that is still being used | |
1075 | * in this file system, from the space currently in use. To figure out | |
1076 | * the space in the most recent snapshot still in use, we need to take | |
1077 | * the total space used in the snapshot and subtract out the space that | |
1078 | * has been freed up since the snapshot was taken. | |
34dc7c2f | 1079 | */ |
13fe0198 MA |
1080 | void |
1081 | dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds) | |
34dc7c2f | 1082 | { |
13fe0198 MA |
1083 | uint64_t mrs_used; |
1084 | uint64_t dlused, dlcomp, dluncomp; | |
34dc7c2f | 1085 | |
0c66c32d | 1086 | ASSERT(!ds->ds_is_snapshot); |
34dc7c2f | 1087 | |
d683ddbb JG |
1088 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) |
1089 | mrs_used = dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes; | |
13fe0198 MA |
1090 | else |
1091 | mrs_used = 0; | |
45d1cae3 | 1092 | |
13fe0198 | 1093 | dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp); |
34dc7c2f | 1094 | |
13fe0198 | 1095 | ASSERT3U(dlused, <=, mrs_used); |
d683ddbb JG |
1096 | dsl_dataset_phys(ds)->ds_unique_bytes = |
1097 | dsl_dataset_phys(ds)->ds_referenced_bytes - (mrs_used - dlused); | |
330d06f9 | 1098 | |
13fe0198 MA |
1099 | if (spa_version(ds->ds_dir->dd_pool->dp_spa) >= |
1100 | SPA_VERSION_UNIQUE_ACCURATE) | |
d683ddbb | 1101 | dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; |
34dc7c2f BB |
1102 | } |
1103 | ||
13fe0198 MA |
1104 | void |
1105 | dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj, | |
1106 | dmu_tx_t *tx) | |
45d1cae3 | 1107 | { |
13fe0198 | 1108 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
13fe0198 | 1109 | ASSERTV(uint64_t count); |
1c27024e | 1110 | int err; |
13fe0198 | 1111 | |
d683ddbb JG |
1112 | ASSERT(dsl_dataset_phys(ds)->ds_num_children >= 2); |
1113 | err = zap_remove_int(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, | |
1114 | obj, tx); | |
13fe0198 MA |
1115 | /* |
1116 | * The err should not be ENOENT, but a bug in a previous version | |
1117 | * of the code could cause upgrade_clones_cb() to not set | |
1118 | * ds_next_snap_obj when it should, leading to a missing entry. | |
1119 | * If we knew that the pool was created after | |
1120 | * SPA_VERSION_NEXT_CLONES, we could assert that it isn't | |
1121 | * ENOENT. However, at least we can check that we don't have | |
1122 | * too many entries in the next_clones_obj even after failing to | |
1123 | * remove this one. | |
1124 | */ | |
1125 | if (err != ENOENT) | |
1126 | VERIFY0(err); | |
d683ddbb | 1127 | ASSERT0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, |
13fe0198 | 1128 | &count)); |
d683ddbb | 1129 | ASSERT3U(count, <=, dsl_dataset_phys(ds)->ds_num_children - 2); |
13fe0198 | 1130 | } |
45d1cae3 | 1131 | |
45d1cae3 | 1132 | |
13fe0198 MA |
1133 | blkptr_t * |
1134 | dsl_dataset_get_blkptr(dsl_dataset_t *ds) | |
1135 | { | |
d683ddbb | 1136 | return (&dsl_dataset_phys(ds)->ds_bp); |
45d1cae3 BB |
1137 | } |
1138 | ||
13fe0198 MA |
1139 | spa_t * |
1140 | dsl_dataset_get_spa(dsl_dataset_t *ds) | |
1141 | { | |
1142 | return (ds->ds_dir->dd_pool->dp_spa); | |
45d1cae3 BB |
1143 | } |
1144 | ||
13fe0198 MA |
1145 | void |
1146 | dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx) | |
34dc7c2f | 1147 | { |
13fe0198 | 1148 | dsl_pool_t *dp; |
45d1cae3 | 1149 | |
13fe0198 MA |
1150 | if (ds == NULL) /* this is the meta-objset */ |
1151 | return; | |
34dc7c2f | 1152 | |
428870ff | 1153 | ASSERT(ds->ds_objset != NULL); |
34dc7c2f | 1154 | |
d683ddbb | 1155 | if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) |
34dc7c2f BB |
1156 | panic("dirtying snapshot!"); |
1157 | ||
0efd9791 AG |
1158 | /* Must not dirty a dataset in the same txg where it got snapshotted. */ |
1159 | ASSERT3U(tx->tx_txg, >, dsl_dataset_phys(ds)->ds_prev_snap_txg); | |
34dc7c2f | 1160 | |
0efd9791 | 1161 | dp = ds->ds_dir->dd_pool; |
13fe0198 | 1162 | if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) { |
34dc7c2f BB |
1163 | /* up the hold count until we can be written out */ |
1164 | dmu_buf_add_ref(ds->ds_dbuf, ds); | |
1165 | } | |
1166 | } | |
1167 | ||
04434775 MA |
1168 | boolean_t |
1169 | dsl_dataset_is_dirty(dsl_dataset_t *ds) | |
1170 | { | |
1c27024e | 1171 | for (int t = 0; t < TXG_SIZE; t++) { |
04434775 MA |
1172 | if (txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets, |
1173 | ds, t)) | |
1174 | return (B_TRUE); | |
1175 | } | |
1176 | return (B_FALSE); | |
1177 | } | |
1178 | ||
34dc7c2f | 1179 | static int |
13fe0198 | 1180 | dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx) |
34dc7c2f | 1181 | { |
13fe0198 | 1182 | uint64_t asize; |
34dc7c2f | 1183 | |
13fe0198 | 1184 | if (!dmu_tx_is_syncing(tx)) |
b128c09f BB |
1185 | return (0); |
1186 | ||
34dc7c2f | 1187 | /* |
13fe0198 MA |
1188 | * If there's an fs-only reservation, any blocks that might become |
1189 | * owned by the snapshot dataset must be accommodated by space | |
1190 | * outside of the reservation. | |
34dc7c2f | 1191 | */ |
13fe0198 | 1192 | ASSERT(ds->ds_reserved == 0 || DS_UNIQUE_IS_ACCURATE(ds)); |
d683ddbb | 1193 | asize = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved); |
13fe0198 | 1194 | if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) |
2e528b49 | 1195 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
1196 | |
1197 | /* | |
13fe0198 MA |
1198 | * Propagate any reserved space for this snapshot to other |
1199 | * snapshot checks in this sync group. | |
34dc7c2f | 1200 | */ |
13fe0198 MA |
1201 | if (asize > 0) |
1202 | dsl_dir_willuse_space(ds->ds_dir, asize, tx); | |
34dc7c2f BB |
1203 | |
1204 | return (0); | |
1205 | } | |
1206 | ||
34dc7c2f | 1207 | int |
13fe0198 | 1208 | dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname, |
788eb90c | 1209 | dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr) |
34dc7c2f | 1210 | { |
13fe0198 MA |
1211 | int error; |
1212 | uint64_t value; | |
34dc7c2f | 1213 | |
13fe0198 | 1214 | ds->ds_trysnap_txg = tx->tx_txg; |
b128c09f | 1215 | |
13fe0198 | 1216 | if (!dmu_tx_is_syncing(tx)) |
45d1cae3 | 1217 | return (0); |
34dc7c2f BB |
1218 | |
1219 | /* | |
13fe0198 MA |
1220 | * We don't allow multiple snapshots of the same txg. If there |
1221 | * is already one, try again. | |
34dc7c2f | 1222 | */ |
d683ddbb | 1223 | if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) |
2e528b49 | 1224 | return (SET_ERROR(EAGAIN)); |
34dc7c2f BB |
1225 | |
1226 | /* | |
13fe0198 | 1227 | * Check for conflicting snapshot name. |
34dc7c2f | 1228 | */ |
13fe0198 MA |
1229 | error = dsl_dataset_snap_lookup(ds, snapname, &value); |
1230 | if (error == 0) | |
2e528b49 | 1231 | return (SET_ERROR(EEXIST)); |
13fe0198 MA |
1232 | if (error != ENOENT) |
1233 | return (error); | |
45d1cae3 | 1234 | |
96c2e961 KW |
1235 | /* |
1236 | * We don't allow taking snapshots of inconsistent datasets, such as | |
1237 | * those into which we are currently receiving. However, if we are | |
1238 | * creating this snapshot as part of a receive, this check will be | |
1239 | * executed atomically with respect to the completion of the receive | |
1240 | * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this | |
1241 | * case we ignore this, knowing it will be fixed up for us shortly in | |
1242 | * dmu_recv_end_sync(). | |
1243 | */ | |
1244 | if (!recv && DS_IS_INCONSISTENT(ds)) | |
1245 | return (SET_ERROR(EBUSY)); | |
1246 | ||
788eb90c JJ |
1247 | /* |
1248 | * Skip the check for temporary snapshots or if we have already checked | |
1249 | * the counts in dsl_dataset_snapshot_check. This means we really only | |
1250 | * check the count here when we're receiving a stream. | |
1251 | */ | |
1252 | if (cnt != 0 && cr != NULL) { | |
1253 | error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, | |
1254 | ZFS_PROP_SNAPSHOT_LIMIT, NULL, cr); | |
1255 | if (error != 0) | |
1256 | return (error); | |
1257 | } | |
1258 | ||
13fe0198 MA |
1259 | error = dsl_dataset_snapshot_reserve_space(ds, tx); |
1260 | if (error != 0) | |
1261 | return (error); | |
45d1cae3 | 1262 | |
34dc7c2f BB |
1263 | return (0); |
1264 | } | |
1265 | ||
234c91c5 | 1266 | int |
13fe0198 | 1267 | dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx) |
b128c09f | 1268 | { |
13fe0198 MA |
1269 | dsl_dataset_snapshot_arg_t *ddsa = arg; |
1270 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1271 | nvpair_t *pair; | |
1272 | int rv = 0; | |
b128c09f | 1273 | |
788eb90c JJ |
1274 | /* |
1275 | * Pre-compute how many total new snapshots will be created for each | |
1276 | * level in the tree and below. This is needed for validating the | |
1277 | * snapshot limit when either taking a recursive snapshot or when | |
1278 | * taking multiple snapshots. | |
1279 | * | |
1280 | * The problem is that the counts are not actually adjusted when | |
1281 | * we are checking, only when we finally sync. For a single snapshot, | |
1282 | * this is easy, the count will increase by 1 at each node up the tree, | |
1283 | * but its more complicated for the recursive/multiple snapshot case. | |
1284 | * | |
1285 | * The dsl_fs_ss_limit_check function does recursively check the count | |
1286 | * at each level up the tree but since it is validating each snapshot | |
1287 | * independently we need to be sure that we are validating the complete | |
1288 | * count for the entire set of snapshots. We do this by rolling up the | |
1289 | * counts for each component of the name into an nvlist and then | |
1290 | * checking each of those cases with the aggregated count. | |
1291 | * | |
1292 | * This approach properly handles not only the recursive snapshot | |
1293 | * case (where we get all of those on the ddsa_snaps list) but also | |
1294 | * the sibling case (e.g. snapshot a/b and a/c so that we will also | |
1295 | * validate the limit on 'a' using a count of 2). | |
1296 | * | |
1297 | * We validate the snapshot names in the third loop and only report | |
1298 | * name errors once. | |
1299 | */ | |
1300 | if (dmu_tx_is_syncing(tx)) { | |
1301 | char *nm; | |
1302 | nvlist_t *cnt_track = NULL; | |
1303 | cnt_track = fnvlist_alloc(); | |
1304 | ||
1305 | nm = kmem_alloc(MAXPATHLEN, KM_SLEEP); | |
1306 | ||
1307 | /* Rollup aggregated counts into the cnt_track list */ | |
1308 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); | |
1309 | pair != NULL; | |
1310 | pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1311 | char *pdelim; | |
1312 | uint64_t val; | |
1313 | ||
1314 | (void) strlcpy(nm, nvpair_name(pair), MAXPATHLEN); | |
1315 | pdelim = strchr(nm, '@'); | |
1316 | if (pdelim == NULL) | |
1317 | continue; | |
1318 | *pdelim = '\0'; | |
1319 | ||
1320 | do { | |
1321 | if (nvlist_lookup_uint64(cnt_track, nm, | |
1322 | &val) == 0) { | |
1323 | /* update existing entry */ | |
1324 | fnvlist_add_uint64(cnt_track, nm, | |
1325 | val + 1); | |
1326 | } else { | |
1327 | /* add to list */ | |
1328 | fnvlist_add_uint64(cnt_track, nm, 1); | |
1329 | } | |
1330 | ||
1331 | pdelim = strrchr(nm, '/'); | |
1332 | if (pdelim != NULL) | |
1333 | *pdelim = '\0'; | |
1334 | } while (pdelim != NULL); | |
1335 | } | |
1336 | ||
1337 | kmem_free(nm, MAXPATHLEN); | |
1338 | ||
1339 | /* Check aggregated counts at each level */ | |
1340 | for (pair = nvlist_next_nvpair(cnt_track, NULL); | |
1341 | pair != NULL; pair = nvlist_next_nvpair(cnt_track, pair)) { | |
1342 | int error = 0; | |
1343 | char *name; | |
1344 | uint64_t cnt = 0; | |
1345 | dsl_dataset_t *ds; | |
1346 | ||
1347 | name = nvpair_name(pair); | |
1348 | cnt = fnvpair_value_uint64(pair); | |
1349 | ASSERT(cnt > 0); | |
1350 | ||
1351 | error = dsl_dataset_hold(dp, name, FTAG, &ds); | |
1352 | if (error == 0) { | |
1353 | error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, | |
1354 | ZFS_PROP_SNAPSHOT_LIMIT, NULL, | |
1355 | ddsa->ddsa_cr); | |
1356 | dsl_dataset_rele(ds, FTAG); | |
1357 | } | |
1358 | ||
1359 | if (error != 0) { | |
1360 | if (ddsa->ddsa_errors != NULL) | |
1361 | fnvlist_add_int32(ddsa->ddsa_errors, | |
1362 | name, error); | |
1363 | rv = error; | |
1364 | /* only report one error for this check */ | |
1365 | break; | |
1366 | } | |
1367 | } | |
1368 | nvlist_free(cnt_track); | |
1369 | } | |
1370 | ||
13fe0198 MA |
1371 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); |
1372 | pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1373 | int error = 0; | |
1374 | dsl_dataset_t *ds; | |
689f093e | 1375 | char *name, *atp = NULL; |
eca7b760 | 1376 | char dsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1377 | |
1378 | name = nvpair_name(pair); | |
eca7b760 | 1379 | if (strlen(name) >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 1380 | error = SET_ERROR(ENAMETOOLONG); |
13fe0198 MA |
1381 | if (error == 0) { |
1382 | atp = strchr(name, '@'); | |
1383 | if (atp == NULL) | |
2e528b49 | 1384 | error = SET_ERROR(EINVAL); |
13fe0198 MA |
1385 | if (error == 0) |
1386 | (void) strlcpy(dsname, name, atp - name + 1); | |
1387 | } | |
1388 | if (error == 0) | |
1389 | error = dsl_dataset_hold(dp, dsname, FTAG, &ds); | |
1390 | if (error == 0) { | |
788eb90c | 1391 | /* passing 0/NULL skips dsl_fs_ss_limit_check */ |
13fe0198 | 1392 | error = dsl_dataset_snapshot_check_impl(ds, |
788eb90c | 1393 | atp + 1, tx, B_FALSE, 0, NULL); |
13fe0198 MA |
1394 | dsl_dataset_rele(ds, FTAG); |
1395 | } | |
b128c09f | 1396 | |
13fe0198 MA |
1397 | if (error != 0) { |
1398 | if (ddsa->ddsa_errors != NULL) { | |
1399 | fnvlist_add_int32(ddsa->ddsa_errors, | |
1400 | name, error); | |
1401 | } | |
1402 | rv = error; | |
1403 | } | |
1404 | } | |
788eb90c | 1405 | |
13fe0198 | 1406 | return (rv); |
b128c09f BB |
1407 | } |
1408 | ||
13fe0198 MA |
1409 | void |
1410 | dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname, | |
1411 | dmu_tx_t *tx) | |
428870ff | 1412 | { |
13fe0198 MA |
1413 | dsl_pool_t *dp = ds->ds_dir->dd_pool; |
1414 | dmu_buf_t *dbuf; | |
1415 | dsl_dataset_phys_t *dsphys; | |
1416 | uint64_t dsobj, crtxg; | |
1417 | objset_t *mos = dp->dp_meta_objset; | |
1418 | ASSERTV(static zil_header_t zero_zil); | |
1419 | ASSERTV(objset_t *os); | |
1420 | ||
1421 | ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock)); | |
428870ff | 1422 | |
428870ff | 1423 | /* |
13fe0198 MA |
1424 | * If we are on an old pool, the zil must not be active, in which |
1425 | * case it will be zeroed. Usually zil_suspend() accomplishes this. | |
428870ff | 1426 | */ |
13fe0198 MA |
1427 | ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP || |
1428 | dmu_objset_from_ds(ds, &os) != 0 || | |
1429 | bcmp(&os->os_phys->os_zil_header, &zero_zil, | |
1430 | sizeof (zero_zil)) == 0); | |
428870ff | 1431 | |
0efd9791 AG |
1432 | /* Should not snapshot a dirty dataset. */ |
1433 | ASSERT(!txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets, | |
1434 | ds, tx->tx_txg)); | |
1435 | ||
788eb90c | 1436 | dsl_fs_ss_count_adjust(ds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx); |
428870ff BB |
1437 | |
1438 | /* | |
13fe0198 | 1439 | * The origin's ds_creation_txg has to be < TXG_INITIAL |
b128c09f BB |
1440 | */ |
1441 | if (strcmp(snapname, ORIGIN_DIR_NAME) == 0) | |
1442 | crtxg = 1; | |
1443 | else | |
1444 | crtxg = tx->tx_txg; | |
1445 | ||
34dc7c2f BB |
1446 | dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, |
1447 | DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); | |
13fe0198 | 1448 | VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); |
34dc7c2f BB |
1449 | dmu_buf_will_dirty(dbuf, tx); |
1450 | dsphys = dbuf->db_data; | |
b128c09f | 1451 | bzero(dsphys, sizeof (dsl_dataset_phys_t)); |
34dc7c2f BB |
1452 | dsphys->ds_dir_obj = ds->ds_dir->dd_object; |
1453 | dsphys->ds_fsid_guid = unique_create(); | |
1454 | (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, | |
1455 | sizeof (dsphys->ds_guid)); | |
d683ddbb JG |
1456 | dsphys->ds_prev_snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
1457 | dsphys->ds_prev_snap_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg; | |
34dc7c2f BB |
1458 | dsphys->ds_next_snap_obj = ds->ds_object; |
1459 | dsphys->ds_num_children = 1; | |
1460 | dsphys->ds_creation_time = gethrestime_sec(); | |
b128c09f | 1461 | dsphys->ds_creation_txg = crtxg; |
d683ddbb JG |
1462 | dsphys->ds_deadlist_obj = dsl_dataset_phys(ds)->ds_deadlist_obj; |
1463 | dsphys->ds_referenced_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes; | |
1464 | dsphys->ds_compressed_bytes = dsl_dataset_phys(ds)->ds_compressed_bytes; | |
1465 | dsphys->ds_uncompressed_bytes = | |
1466 | dsl_dataset_phys(ds)->ds_uncompressed_bytes; | |
1467 | dsphys->ds_flags = dsl_dataset_phys(ds)->ds_flags; | |
cc9bb3e5 | 1468 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 1469 | dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp; |
cc9bb3e5 | 1470 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f BB |
1471 | dmu_buf_rele(dbuf, FTAG); |
1472 | ||
1c27024e | 1473 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
1474 | if (ds->ds_feature_inuse[f]) |
1475 | dsl_dataset_activate_feature(dsobj, f, tx); | |
1476 | } | |
f1512ee6 | 1477 | |
d683ddbb JG |
1478 | ASSERT3U(ds->ds_prev != 0, ==, |
1479 | dsl_dataset_phys(ds)->ds_prev_snap_obj != 0); | |
34dc7c2f | 1480 | if (ds->ds_prev) { |
b128c09f | 1481 | uint64_t next_clones_obj = |
d683ddbb JG |
1482 | dsl_dataset_phys(ds->ds_prev)->ds_next_clones_obj; |
1483 | ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == | |
34dc7c2f | 1484 | ds->ds_object || |
d683ddbb JG |
1485 | dsl_dataset_phys(ds->ds_prev)->ds_num_children > 1); |
1486 | if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == | |
1487 | ds->ds_object) { | |
34dc7c2f | 1488 | dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); |
d683ddbb JG |
1489 | ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, ==, |
1490 | dsl_dataset_phys(ds->ds_prev)->ds_creation_txg); | |
1491 | dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj = dsobj; | |
b128c09f | 1492 | } else if (next_clones_obj != 0) { |
13fe0198 | 1493 | dsl_dataset_remove_from_next_clones(ds->ds_prev, |
428870ff | 1494 | dsphys->ds_next_snap_obj, tx); |
13fe0198 | 1495 | VERIFY0(zap_add_int(mos, |
b128c09f | 1496 | next_clones_obj, dsobj, tx)); |
34dc7c2f BB |
1497 | } |
1498 | } | |
1499 | ||
1500 | /* | |
1501 | * If we have a reference-reservation on this dataset, we will | |
1502 | * need to increase the amount of refreservation being charged | |
1503 | * since our unique space is going to zero. | |
1504 | */ | |
1505 | if (ds->ds_reserved) { | |
428870ff BB |
1506 | int64_t delta; |
1507 | ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); | |
d683ddbb JG |
1508 | delta = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, |
1509 | ds->ds_reserved); | |
b128c09f | 1510 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, |
428870ff | 1511 | delta, 0, 0, tx); |
34dc7c2f BB |
1512 | } |
1513 | ||
34dc7c2f | 1514 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
d683ddbb JG |
1515 | dsl_dataset_phys(ds)->ds_deadlist_obj = |
1516 | dsl_deadlist_clone(&ds->ds_deadlist, UINT64_MAX, | |
1517 | dsl_dataset_phys(ds)->ds_prev_snap_obj, tx); | |
428870ff | 1518 | dsl_deadlist_close(&ds->ds_deadlist); |
d683ddbb JG |
1519 | dsl_deadlist_open(&ds->ds_deadlist, mos, |
1520 | dsl_dataset_phys(ds)->ds_deadlist_obj); | |
428870ff | 1521 | dsl_deadlist_add_key(&ds->ds_deadlist, |
d683ddbb | 1522 | dsl_dataset_phys(ds)->ds_prev_snap_txg, tx); |
428870ff | 1523 | |
a1d477c2 MA |
1524 | if (dsl_dataset_remap_deadlist_exists(ds)) { |
1525 | uint64_t remap_deadlist_obj = | |
1526 | dsl_dataset_get_remap_deadlist_object(ds); | |
1527 | /* | |
1528 | * Move the remap_deadlist to the snapshot. The head | |
1529 | * will create a new remap deadlist on demand, from | |
1530 | * dsl_dataset_block_remapped(). | |
1531 | */ | |
1532 | dsl_dataset_unset_remap_deadlist_object(ds, tx); | |
1533 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
1534 | ||
1535 | dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx); | |
1536 | VERIFY0(zap_add(mos, dsobj, DS_FIELD_REMAP_DEADLIST, | |
1537 | sizeof (remap_deadlist_obj), 1, &remap_deadlist_obj, tx)); | |
1538 | } | |
1539 | ||
d683ddbb JG |
1540 | ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg); |
1541 | dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj; | |
1542 | dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg; | |
1543 | dsl_dataset_phys(ds)->ds_unique_bytes = 0; | |
a1d477c2 | 1544 | |
34dc7c2f | 1545 | if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) |
d683ddbb | 1546 | dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; |
34dc7c2f | 1547 | |
d683ddbb | 1548 | VERIFY0(zap_add(mos, dsl_dataset_phys(ds)->ds_snapnames_zapobj, |
13fe0198 | 1549 | snapname, 8, 1, &dsobj, tx)); |
34dc7c2f BB |
1550 | |
1551 | if (ds->ds_prev) | |
13fe0198 MA |
1552 | dsl_dataset_rele(ds->ds_prev, ds); |
1553 | VERIFY0(dsl_dataset_hold_obj(dp, | |
d683ddbb | 1554 | dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev)); |
b128c09f | 1555 | |
428870ff BB |
1556 | dsl_scan_ds_snapshotted(ds, tx); |
1557 | ||
1558 | dsl_dir_snap_cmtime_update(ds->ds_dir); | |
34dc7c2f | 1559 | |
6f1ffb06 | 1560 | spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, ""); |
34dc7c2f BB |
1561 | } |
1562 | ||
234c91c5 | 1563 | void |
13fe0198 | 1564 | dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1565 | { |
13fe0198 MA |
1566 | dsl_dataset_snapshot_arg_t *ddsa = arg; |
1567 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1568 | nvpair_t *pair; | |
34dc7c2f | 1569 | |
13fe0198 MA |
1570 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); |
1571 | pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1572 | dsl_dataset_t *ds; | |
1573 | char *name, *atp; | |
eca7b760 | 1574 | char dsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1575 | |
1576 | name = nvpair_name(pair); | |
1577 | atp = strchr(name, '@'); | |
1578 | (void) strlcpy(dsname, name, atp - name + 1); | |
1579 | VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds)); | |
1580 | ||
1581 | dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx); | |
1582 | if (ddsa->ddsa_props != NULL) { | |
1583 | dsl_props_set_sync_impl(ds->ds_prev, | |
1584 | ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx); | |
1585 | } | |
a0bd735a | 1586 | zvol_create_minors(dp->dp_spa, nvpair_name(pair), B_TRUE); |
13fe0198 MA |
1587 | dsl_dataset_rele(ds, FTAG); |
1588 | } | |
34dc7c2f BB |
1589 | } |
1590 | ||
13fe0198 MA |
1591 | /* |
1592 | * The snapshots must all be in the same pool. | |
1593 | * All-or-nothing: if there are any failures, nothing will be modified. | |
1594 | */ | |
1595 | int | |
1596 | dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors) | |
330d06f9 | 1597 | { |
13fe0198 MA |
1598 | dsl_dataset_snapshot_arg_t ddsa; |
1599 | nvpair_t *pair; | |
1600 | boolean_t needsuspend; | |
1601 | int error; | |
1602 | spa_t *spa; | |
1603 | char *firstname; | |
1604 | nvlist_t *suspended = NULL; | |
330d06f9 | 1605 | |
13fe0198 MA |
1606 | pair = nvlist_next_nvpair(snaps, NULL); |
1607 | if (pair == NULL) | |
1608 | return (0); | |
1609 | firstname = nvpair_name(pair); | |
1610 | ||
1611 | error = spa_open(firstname, &spa, FTAG); | |
1612 | if (error != 0) | |
1613 | return (error); | |
1614 | needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); | |
1615 | spa_close(spa, FTAG); | |
1616 | ||
1617 | if (needsuspend) { | |
1618 | suspended = fnvlist_alloc(); | |
1619 | for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; | |
1620 | pair = nvlist_next_nvpair(snaps, pair)) { | |
eca7b760 | 1621 | char fsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1622 | char *snapname = nvpair_name(pair); |
1623 | char *atp; | |
1624 | void *cookie; | |
1625 | ||
1626 | atp = strchr(snapname, '@'); | |
1627 | if (atp == NULL) { | |
2e528b49 | 1628 | error = SET_ERROR(EINVAL); |
13fe0198 MA |
1629 | break; |
1630 | } | |
1631 | (void) strlcpy(fsname, snapname, atp - snapname + 1); | |
1632 | ||
1633 | error = zil_suspend(fsname, &cookie); | |
1634 | if (error != 0) | |
1635 | break; | |
1636 | fnvlist_add_uint64(suspended, fsname, | |
1637 | (uintptr_t)cookie); | |
1638 | } | |
1639 | } | |
1640 | ||
1641 | ddsa.ddsa_snaps = snaps; | |
1642 | ddsa.ddsa_props = props; | |
1643 | ddsa.ddsa_errors = errors; | |
788eb90c | 1644 | ddsa.ddsa_cr = CRED(); |
13fe0198 MA |
1645 | |
1646 | if (error == 0) { | |
1647 | error = dsl_sync_task(firstname, dsl_dataset_snapshot_check, | |
1648 | dsl_dataset_snapshot_sync, &ddsa, | |
3d45fdd6 | 1649 | fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL); |
13fe0198 MA |
1650 | } |
1651 | ||
1652 | if (suspended != NULL) { | |
1653 | for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL; | |
1654 | pair = nvlist_next_nvpair(suspended, pair)) { | |
1655 | zil_resume((void *)(uintptr_t) | |
1656 | fnvpair_value_uint64(pair)); | |
1657 | } | |
1658 | fnvlist_free(suspended); | |
1659 | } | |
1660 | ||
1661 | return (error); | |
1662 | } | |
1663 | ||
1664 | typedef struct dsl_dataset_snapshot_tmp_arg { | |
1665 | const char *ddsta_fsname; | |
1666 | const char *ddsta_snapname; | |
1667 | minor_t ddsta_cleanup_minor; | |
1668 | const char *ddsta_htag; | |
1669 | } dsl_dataset_snapshot_tmp_arg_t; | |
1670 | ||
1671 | static int | |
1672 | dsl_dataset_snapshot_tmp_check(void *arg, dmu_tx_t *tx) | |
1673 | { | |
1674 | dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; | |
1675 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1676 | dsl_dataset_t *ds; | |
1677 | int error; | |
1678 | ||
1679 | error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds); | |
1680 | if (error != 0) | |
1681 | return (error); | |
1682 | ||
788eb90c | 1683 | /* NULL cred means no limit check for tmp snapshot */ |
96c2e961 | 1684 | error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname, |
788eb90c | 1685 | tx, B_FALSE, 0, NULL); |
13fe0198 MA |
1686 | if (error != 0) { |
1687 | dsl_dataset_rele(ds, FTAG); | |
1688 | return (error); | |
1689 | } | |
1690 | ||
1691 | if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) { | |
1692 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 1693 | return (SET_ERROR(ENOTSUP)); |
13fe0198 MA |
1694 | } |
1695 | error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag, | |
1696 | B_TRUE, tx); | |
1697 | if (error != 0) { | |
1698 | dsl_dataset_rele(ds, FTAG); | |
1699 | return (error); | |
1700 | } | |
1701 | ||
1702 | dsl_dataset_rele(ds, FTAG); | |
1703 | return (0); | |
1704 | } | |
1705 | ||
1706 | static void | |
1707 | dsl_dataset_snapshot_tmp_sync(void *arg, dmu_tx_t *tx) | |
1708 | { | |
1709 | dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; | |
1710 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 1711 | dsl_dataset_t *ds = NULL; |
13fe0198 MA |
1712 | |
1713 | VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds)); | |
1714 | ||
1715 | dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx); | |
1716 | dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag, | |
1717 | ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx); | |
1718 | dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx); | |
1719 | ||
1720 | dsl_dataset_rele(ds, FTAG); | |
1721 | } | |
1722 | ||
1723 | int | |
1724 | dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname, | |
1725 | minor_t cleanup_minor, const char *htag) | |
1726 | { | |
1727 | dsl_dataset_snapshot_tmp_arg_t ddsta; | |
1728 | int error; | |
1729 | spa_t *spa; | |
1730 | boolean_t needsuspend; | |
1731 | void *cookie; | |
1732 | ||
1733 | ddsta.ddsta_fsname = fsname; | |
1734 | ddsta.ddsta_snapname = snapname; | |
1735 | ddsta.ddsta_cleanup_minor = cleanup_minor; | |
1736 | ddsta.ddsta_htag = htag; | |
1737 | ||
1738 | error = spa_open(fsname, &spa, FTAG); | |
1739 | if (error != 0) | |
1740 | return (error); | |
1741 | needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); | |
1742 | spa_close(spa, FTAG); | |
1743 | ||
1744 | if (needsuspend) { | |
1745 | error = zil_suspend(fsname, &cookie); | |
1746 | if (error != 0) | |
1747 | return (error); | |
1748 | } | |
1749 | ||
1750 | error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check, | |
3d45fdd6 | 1751 | dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED); |
13fe0198 MA |
1752 | |
1753 | if (needsuspend) | |
1754 | zil_resume(cookie); | |
1755 | return (error); | |
1756 | } | |
1757 | ||
13fe0198 MA |
1758 | void |
1759 | dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx) | |
1760 | { | |
1761 | ASSERT(dmu_tx_is_syncing(tx)); | |
1762 | ASSERT(ds->ds_objset != NULL); | |
d683ddbb | 1763 | ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0); |
13fe0198 MA |
1764 | |
1765 | /* | |
1766 | * in case we had to change ds_fsid_guid when we opened it, | |
1767 | * sync it out now. | |
1768 | */ | |
1769 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
d683ddbb | 1770 | dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid; |
13fe0198 | 1771 | |
47dfff3b MA |
1772 | if (ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] != 0) { |
1773 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1774 | ds->ds_object, DS_FIELD_RESUME_OBJECT, 8, 1, | |
1775 | &ds->ds_resume_object[tx->tx_txg & TXG_MASK], tx)); | |
1776 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1777 | ds->ds_object, DS_FIELD_RESUME_OFFSET, 8, 1, | |
1778 | &ds->ds_resume_offset[tx->tx_txg & TXG_MASK], tx)); | |
1779 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1780 | ds->ds_object, DS_FIELD_RESUME_BYTES, 8, 1, | |
1781 | &ds->ds_resume_bytes[tx->tx_txg & TXG_MASK], tx)); | |
1782 | ds->ds_resume_object[tx->tx_txg & TXG_MASK] = 0; | |
1783 | ds->ds_resume_offset[tx->tx_txg & TXG_MASK] = 0; | |
1784 | ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] = 0; | |
1785 | } | |
1786 | ||
13fe0198 | 1787 | dmu_objset_sync(ds->ds_objset, zio, tx); |
f1512ee6 | 1788 | |
1c27024e | 1789 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
1790 | if (ds->ds_feature_activation_needed[f]) { |
1791 | if (ds->ds_feature_inuse[f]) | |
1792 | continue; | |
1793 | dsl_dataset_activate_feature(ds->ds_object, f, tx); | |
1794 | ds->ds_feature_inuse[f] = B_TRUE; | |
1795 | } | |
f1512ee6 | 1796 | } |
13fe0198 MA |
1797 | } |
1798 | ||
0efd9791 AG |
1799 | static int |
1800 | deadlist_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) | |
1801 | { | |
1802 | dsl_deadlist_t *dl = arg; | |
1803 | dsl_deadlist_insert(dl, bp, tx); | |
1804 | return (0); | |
1805 | } | |
1806 | ||
1807 | void | |
1808 | dsl_dataset_sync_done(dsl_dataset_t *ds, dmu_tx_t *tx) | |
1809 | { | |
64fc7762 | 1810 | objset_t *os = ds->ds_objset; |
0efd9791 AG |
1811 | |
1812 | bplist_iterate(&ds->ds_pending_deadlist, | |
1813 | deadlist_enqueue_cb, &ds->ds_deadlist, tx); | |
1814 | ||
64fc7762 MA |
1815 | if (os->os_synced_dnodes != NULL) { |
1816 | multilist_destroy(os->os_synced_dnodes); | |
1817 | os->os_synced_dnodes = NULL; | |
1818 | } | |
1819 | ||
0efd9791 AG |
1820 | ASSERT(!dmu_objset_is_dirty(os, dmu_tx_get_txg(tx))); |
1821 | ||
1822 | dmu_buf_rele(ds->ds_dbuf, ds); | |
1823 | } | |
1824 | ||
d99a0153 CW |
1825 | int |
1826 | get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val) | |
13fe0198 MA |
1827 | { |
1828 | uint64_t count = 0; | |
1829 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; | |
1830 | zap_cursor_t zc; | |
1831 | zap_attribute_t za; | |
13fe0198 MA |
1832 | |
1833 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
330d06f9 MA |
1834 | |
1835 | /* | |
13fe0198 | 1836 | * There may be missing entries in ds_next_clones_obj |
330d06f9 MA |
1837 | * due to a bug in a previous version of the code. |
1838 | * Only trust it if it has the right number of entries. | |
1839 | */ | |
d683ddbb JG |
1840 | if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) { |
1841 | VERIFY0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, | |
330d06f9 MA |
1842 | &count)); |
1843 | } | |
d99a0153 CW |
1844 | if (count != dsl_dataset_phys(ds)->ds_num_children - 1) { |
1845 | return (ENOENT); | |
1846 | } | |
d683ddbb JG |
1847 | for (zap_cursor_init(&zc, mos, |
1848 | dsl_dataset_phys(ds)->ds_next_clones_obj); | |
330d06f9 MA |
1849 | zap_cursor_retrieve(&zc, &za) == 0; |
1850 | zap_cursor_advance(&zc)) { | |
1851 | dsl_dataset_t *clone; | |
eca7b760 | 1852 | char buf[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1853 | VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool, |
1854 | za.za_first_integer, FTAG, &clone)); | |
330d06f9 | 1855 | dsl_dir_name(clone->ds_dir, buf); |
13fe0198 | 1856 | fnvlist_add_boolean(val, buf); |
330d06f9 MA |
1857 | dsl_dataset_rele(clone, FTAG); |
1858 | } | |
1859 | zap_cursor_fini(&zc); | |
d99a0153 CW |
1860 | return (0); |
1861 | } | |
1862 | ||
1863 | void | |
1864 | get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv) | |
1865 | { | |
1866 | nvlist_t *propval = fnvlist_alloc(); | |
1867 | nvlist_t *val; | |
1868 | ||
1869 | /* | |
1870 | * We use nvlist_alloc() instead of fnvlist_alloc() because the | |
1871 | * latter would allocate the list with NV_UNIQUE_NAME flag. | |
1872 | * As a result, every time a clone name is appended to the list | |
1873 | * it would be (linearly) searched for for a duplicate name. | |
1874 | * We already know that all clone names must be unique and we | |
1875 | * want avoid the quadratic complexity of double-checking that | |
1876 | * because we can have a large number of clones. | |
1877 | */ | |
1878 | VERIFY0(nvlist_alloc(&val, 0, KM_SLEEP)); | |
1879 | ||
1880 | if (get_clones_stat_impl(ds, val) == 0) { | |
1881 | fnvlist_add_nvlist(propval, ZPROP_VALUE, val); | |
1882 | fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES), | |
1883 | propval); | |
1884 | } | |
1885 | ||
330d06f9 MA |
1886 | nvlist_free(val); |
1887 | nvlist_free(propval); | |
330d06f9 MA |
1888 | } |
1889 | ||
d99a0153 CW |
1890 | /* |
1891 | * Returns a string that represents the receive resume stats token. It should | |
1892 | * be freed with strfree(). | |
1893 | */ | |
1894 | char * | |
1895 | get_receive_resume_stats_impl(dsl_dataset_t *ds) | |
47dfff3b MA |
1896 | { |
1897 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
1898 | ||
1899 | if (dsl_dataset_has_resume_receive_state(ds)) { | |
1900 | char *str; | |
1901 | void *packed; | |
1902 | uint8_t *compressed; | |
1903 | uint64_t val; | |
1904 | nvlist_t *token_nv = fnvlist_alloc(); | |
1905 | size_t packed_size, compressed_size; | |
47dfff3b MA |
1906 | |
1907 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1908 | DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val) == 0) { | |
1909 | fnvlist_add_uint64(token_nv, "fromguid", val); | |
1910 | } | |
1911 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1912 | DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val) == 0) { | |
1913 | fnvlist_add_uint64(token_nv, "object", val); | |
1914 | } | |
1915 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1916 | DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val) == 0) { | |
1917 | fnvlist_add_uint64(token_nv, "offset", val); | |
1918 | } | |
1919 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1920 | DS_FIELD_RESUME_BYTES, sizeof (val), 1, &val) == 0) { | |
1921 | fnvlist_add_uint64(token_nv, "bytes", val); | |
1922 | } | |
1923 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1924 | DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val) == 0) { | |
1925 | fnvlist_add_uint64(token_nv, "toguid", val); | |
1926 | } | |
1c27024e | 1927 | char buf[MAXNAMELEN]; |
47dfff3b MA |
1928 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, |
1929 | DS_FIELD_RESUME_TONAME, 1, sizeof (buf), buf) == 0) { | |
1930 | fnvlist_add_string(token_nv, "toname", buf); | |
1931 | } | |
2aa34383 DK |
1932 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1933 | DS_FIELD_RESUME_LARGEBLOCK) == 0) { | |
1934 | fnvlist_add_boolean(token_nv, "largeblockok"); | |
1935 | } | |
47dfff3b MA |
1936 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1937 | DS_FIELD_RESUME_EMBEDOK) == 0) { | |
1938 | fnvlist_add_boolean(token_nv, "embedok"); | |
1939 | } | |
2aa34383 DK |
1940 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1941 | DS_FIELD_RESUME_COMPRESSOK) == 0) { | |
1942 | fnvlist_add_boolean(token_nv, "compressok"); | |
1943 | } | |
b5256303 TC |
1944 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1945 | DS_FIELD_RESUME_RAWOK) == 0) { | |
1946 | fnvlist_add_boolean(token_nv, "rawok"); | |
1947 | } | |
47dfff3b MA |
1948 | packed = fnvlist_pack(token_nv, &packed_size); |
1949 | fnvlist_free(token_nv); | |
1950 | compressed = kmem_alloc(packed_size, KM_SLEEP); | |
1951 | ||
1952 | compressed_size = gzip_compress(packed, compressed, | |
1953 | packed_size, packed_size, 6); | |
1954 | ||
1c27024e | 1955 | zio_cksum_t cksum; |
fc897b24 | 1956 | fletcher_4_native_varsize(compressed, compressed_size, &cksum); |
47dfff3b MA |
1957 | |
1958 | str = kmem_alloc(compressed_size * 2 + 1, KM_SLEEP); | |
1c27024e | 1959 | for (int i = 0; i < compressed_size; i++) { |
47dfff3b MA |
1960 | (void) sprintf(str + i * 2, "%02x", compressed[i]); |
1961 | } | |
1962 | str[compressed_size * 2] = '\0'; | |
1c27024e | 1963 | char *propval = kmem_asprintf("%u-%llx-%llx-%s", |
47dfff3b MA |
1964 | ZFS_SEND_RESUME_TOKEN_VERSION, |
1965 | (longlong_t)cksum.zc_word[0], | |
1966 | (longlong_t)packed_size, str); | |
47dfff3b MA |
1967 | kmem_free(packed, packed_size); |
1968 | kmem_free(str, compressed_size * 2 + 1); | |
1969 | kmem_free(compressed, packed_size); | |
d99a0153 CW |
1970 | return (propval); |
1971 | } | |
1972 | return (strdup("")); | |
1973 | } | |
1974 | ||
1975 | /* | |
1976 | * Returns a string that represents the receive resume stats token of the | |
1977 | * dataset's child. It should be freed with strfree(). | |
1978 | */ | |
1979 | char * | |
1980 | get_child_receive_stats(dsl_dataset_t *ds) | |
1981 | { | |
1982 | char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; | |
1983 | dsl_dataset_t *recv_ds; | |
1984 | dsl_dataset_name(ds, recvname); | |
1985 | if (strlcat(recvname, "/", sizeof (recvname)) < | |
1986 | sizeof (recvname) && | |
1987 | strlcat(recvname, recv_clone_name, sizeof (recvname)) < | |
1988 | sizeof (recvname) && | |
1989 | dsl_dataset_hold(ds->ds_dir->dd_pool, recvname, FTAG, | |
1990 | &recv_ds) == 0) { | |
1991 | char *propval = get_receive_resume_stats_impl(recv_ds); | |
1992 | dsl_dataset_rele(recv_ds, FTAG); | |
1993 | return (propval); | |
47dfff3b | 1994 | } |
d99a0153 CW |
1995 | return (strdup("")); |
1996 | } | |
1997 | ||
1998 | static void | |
1999 | get_receive_resume_stats(dsl_dataset_t *ds, nvlist_t *nv) | |
2000 | { | |
2001 | char *propval = get_receive_resume_stats_impl(ds); | |
2002 | if (strcmp(propval, "") != 0) { | |
2003 | dsl_prop_nvlist_add_string(nv, | |
2004 | ZFS_PROP_RECEIVE_RESUME_TOKEN, propval); | |
2005 | } else { | |
2006 | char *childval = get_child_receive_stats(ds); | |
2007 | if (strcmp(childval, "") != 0) { | |
2008 | dsl_prop_nvlist_add_string(nv, | |
2009 | ZFS_PROP_RECEIVE_RESUME_TOKEN, childval); | |
2010 | } | |
2011 | strfree(childval); | |
2012 | } | |
2013 | strfree(propval); | |
2014 | } | |
2015 | ||
2016 | uint64_t | |
2017 | dsl_get_refratio(dsl_dataset_t *ds) | |
2018 | { | |
2019 | uint64_t ratio = dsl_dataset_phys(ds)->ds_compressed_bytes == 0 ? 100 : | |
2020 | (dsl_dataset_phys(ds)->ds_uncompressed_bytes * 100 / | |
2021 | dsl_dataset_phys(ds)->ds_compressed_bytes); | |
2022 | return (ratio); | |
2023 | } | |
2024 | ||
2025 | uint64_t | |
2026 | dsl_get_logicalreferenced(dsl_dataset_t *ds) | |
2027 | { | |
2028 | return (dsl_dataset_phys(ds)->ds_uncompressed_bytes); | |
2029 | } | |
2030 | ||
2031 | uint64_t | |
2032 | dsl_get_compressratio(dsl_dataset_t *ds) | |
2033 | { | |
2034 | if (ds->ds_is_snapshot) { | |
2035 | return (dsl_get_refratio(ds)); | |
2036 | } else { | |
2037 | dsl_dir_t *dd = ds->ds_dir; | |
2038 | mutex_enter(&dd->dd_lock); | |
2039 | uint64_t val = dsl_dir_get_compressratio(dd); | |
2040 | mutex_exit(&dd->dd_lock); | |
2041 | return (val); | |
2042 | } | |
2043 | } | |
2044 | ||
2045 | uint64_t | |
2046 | dsl_get_used(dsl_dataset_t *ds) | |
2047 | { | |
2048 | if (ds->ds_is_snapshot) { | |
2049 | return (dsl_dataset_phys(ds)->ds_unique_bytes); | |
2050 | } else { | |
2051 | dsl_dir_t *dd = ds->ds_dir; | |
2052 | mutex_enter(&dd->dd_lock); | |
2053 | uint64_t val = dsl_dir_get_used(dd); | |
2054 | mutex_exit(&dd->dd_lock); | |
2055 | return (val); | |
2056 | } | |
2057 | } | |
2058 | ||
2059 | uint64_t | |
2060 | dsl_get_creation(dsl_dataset_t *ds) | |
2061 | { | |
2062 | return (dsl_dataset_phys(ds)->ds_creation_time); | |
2063 | } | |
2064 | ||
2065 | uint64_t | |
2066 | dsl_get_creationtxg(dsl_dataset_t *ds) | |
2067 | { | |
2068 | return (dsl_dataset_phys(ds)->ds_creation_txg); | |
2069 | } | |
2070 | ||
2071 | uint64_t | |
2072 | dsl_get_refquota(dsl_dataset_t *ds) | |
2073 | { | |
2074 | return (ds->ds_quota); | |
2075 | } | |
2076 | ||
2077 | uint64_t | |
2078 | dsl_get_refreservation(dsl_dataset_t *ds) | |
2079 | { | |
2080 | return (ds->ds_reserved); | |
2081 | } | |
2082 | ||
2083 | uint64_t | |
2084 | dsl_get_guid(dsl_dataset_t *ds) | |
2085 | { | |
2086 | return (dsl_dataset_phys(ds)->ds_guid); | |
2087 | } | |
2088 | ||
2089 | uint64_t | |
2090 | dsl_get_unique(dsl_dataset_t *ds) | |
2091 | { | |
2092 | return (dsl_dataset_phys(ds)->ds_unique_bytes); | |
2093 | } | |
2094 | ||
2095 | uint64_t | |
2096 | dsl_get_objsetid(dsl_dataset_t *ds) | |
2097 | { | |
2098 | return (ds->ds_object); | |
2099 | } | |
2100 | ||
2101 | uint64_t | |
2102 | dsl_get_userrefs(dsl_dataset_t *ds) | |
2103 | { | |
2104 | return (ds->ds_userrefs); | |
2105 | } | |
2106 | ||
2107 | uint64_t | |
2108 | dsl_get_defer_destroy(dsl_dataset_t *ds) | |
2109 | { | |
2110 | return (DS_IS_DEFER_DESTROY(ds) ? 1 : 0); | |
2111 | } | |
2112 | ||
2113 | uint64_t | |
2114 | dsl_get_referenced(dsl_dataset_t *ds) | |
2115 | { | |
2116 | return (dsl_dataset_phys(ds)->ds_referenced_bytes); | |
2117 | } | |
2118 | ||
2119 | uint64_t | |
2120 | dsl_get_numclones(dsl_dataset_t *ds) | |
2121 | { | |
2122 | ASSERT(ds->ds_is_snapshot); | |
2123 | return (dsl_dataset_phys(ds)->ds_num_children - 1); | |
2124 | } | |
2125 | ||
2126 | uint64_t | |
2127 | dsl_get_inconsistent(dsl_dataset_t *ds) | |
2128 | { | |
2129 | return ((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT) ? | |
2130 | 1 : 0); | |
2131 | } | |
2132 | ||
2133 | uint64_t | |
2134 | dsl_get_available(dsl_dataset_t *ds) | |
2135 | { | |
2136 | uint64_t refdbytes = dsl_get_referenced(ds); | |
2137 | uint64_t availbytes = dsl_dir_space_available(ds->ds_dir, | |
2138 | NULL, 0, TRUE); | |
2139 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) { | |
2140 | availbytes += | |
2141 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes; | |
2142 | } | |
2143 | if (ds->ds_quota != 0) { | |
2144 | /* | |
2145 | * Adjust available bytes according to refquota | |
2146 | */ | |
2147 | if (refdbytes < ds->ds_quota) { | |
2148 | availbytes = MIN(availbytes, | |
2149 | ds->ds_quota - refdbytes); | |
2150 | } else { | |
2151 | availbytes = 0; | |
2152 | } | |
2153 | } | |
2154 | return (availbytes); | |
2155 | } | |
2156 | ||
2157 | int | |
2158 | dsl_get_written(dsl_dataset_t *ds, uint64_t *written) | |
2159 | { | |
2160 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2161 | dsl_dataset_t *prev; | |
2162 | int err = dsl_dataset_hold_obj(dp, | |
2163 | dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev); | |
2164 | if (err == 0) { | |
2165 | uint64_t comp, uncomp; | |
2166 | err = dsl_dataset_space_written(prev, ds, written, | |
2167 | &comp, &uncomp); | |
2168 | dsl_dataset_rele(prev, FTAG); | |
2169 | } | |
2170 | return (err); | |
2171 | } | |
2172 | ||
2173 | /* | |
2174 | * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN. | |
2175 | */ | |
2176 | int | |
2177 | dsl_get_prev_snap(dsl_dataset_t *ds, char *snap) | |
2178 | { | |
2179 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2180 | if (ds->ds_prev != NULL && ds->ds_prev != dp->dp_origin_snap) { | |
2181 | dsl_dataset_name(ds->ds_prev, snap); | |
2182 | return (0); | |
2183 | } else { | |
2184 | return (ENOENT); | |
2185 | } | |
2186 | } | |
2187 | ||
2188 | /* | |
2189 | * Returns the mountpoint property and source for the given dataset in the value | |
2190 | * and source buffers. The value buffer must be at least as large as MAXPATHLEN | |
2191 | * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN. | |
2192 | * Returns 0 on success and an error on failure. | |
2193 | */ | |
2194 | int | |
2195 | dsl_get_mountpoint(dsl_dataset_t *ds, const char *dsname, char *value, | |
2196 | char *source) | |
2197 | { | |
2198 | int error; | |
2199 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2200 | ||
2201 | /* Retrieve the mountpoint value stored in the zap opbject */ | |
2202 | error = dsl_prop_get_ds(ds, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 1, | |
2203 | ZAP_MAXVALUELEN, value, source); | |
2204 | if (error != 0) { | |
2205 | return (error); | |
2206 | } | |
2207 | ||
2208 | /* | |
2209 | * Process the dsname and source to find the full mountpoint string. | |
2210 | * Can be skipped for 'legacy' or 'none'. | |
2211 | */ | |
2212 | if (value[0] == '/') { | |
2213 | char *buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP); | |
2214 | char *root = buf; | |
2215 | const char *relpath; | |
2216 | ||
2217 | /* | |
2218 | * If we inherit the mountpoint, even from a dataset | |
2219 | * with a received value, the source will be the path of | |
2220 | * the dataset we inherit from. If source is | |
2221 | * ZPROP_SOURCE_VAL_RECVD, the received value is not | |
2222 | * inherited. | |
2223 | */ | |
2224 | if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) { | |
2225 | relpath = ""; | |
2226 | } else { | |
2227 | ASSERT0(strncmp(dsname, source, strlen(source))); | |
2228 | relpath = dsname + strlen(source); | |
2229 | if (relpath[0] == '/') | |
2230 | relpath++; | |
2231 | } | |
2232 | ||
2233 | spa_altroot(dp->dp_spa, root, ZAP_MAXVALUELEN); | |
2234 | ||
2235 | /* | |
2236 | * Special case an alternate root of '/'. This will | |
2237 | * avoid having multiple leading slashes in the | |
2238 | * mountpoint path. | |
2239 | */ | |
2240 | if (strcmp(root, "/") == 0) | |
2241 | root++; | |
2242 | ||
2243 | /* | |
2244 | * If the mountpoint is '/' then skip over this | |
2245 | * if we are obtaining either an alternate root or | |
2246 | * an inherited mountpoint. | |
2247 | */ | |
2248 | char *mnt = value; | |
2249 | if (value[1] == '\0' && (root[0] != '\0' || | |
2250 | relpath[0] != '\0')) | |
2251 | mnt = value + 1; | |
2252 | ||
2253 | if (relpath[0] == '\0') { | |
2254 | (void) snprintf(value, ZAP_MAXVALUELEN, "%s%s", | |
2255 | root, mnt); | |
2256 | } else { | |
2257 | (void) snprintf(value, ZAP_MAXVALUELEN, "%s%s%s%s", | |
2258 | root, mnt, relpath[0] == '@' ? "" : "/", | |
2259 | relpath); | |
2260 | } | |
2261 | kmem_free(buf, ZAP_MAXVALUELEN); | |
2262 | } | |
2263 | ||
2264 | return (0); | |
47dfff3b MA |
2265 | } |
2266 | ||
34dc7c2f BB |
2267 | void |
2268 | dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv) | |
2269 | { | |
b858767a | 2270 | dsl_pool_t *dp = ds->ds_dir->dd_pool; |
13fe0198 MA |
2271 | |
2272 | ASSERT(dsl_pool_config_held(dp)); | |
34dc7c2f | 2273 | |
d99a0153 CW |
2274 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO, |
2275 | dsl_get_refratio(ds)); | |
24a64651 | 2276 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED, |
d99a0153 CW |
2277 | dsl_get_logicalreferenced(ds)); |
2278 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO, | |
2279 | dsl_get_compressratio(ds)); | |
2280 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED, | |
2281 | dsl_get_used(ds)); | |
6f1ffb06 | 2282 | |
0c66c32d | 2283 | if (ds->ds_is_snapshot) { |
6f1ffb06 MA |
2284 | get_clones_stat(ds, nv); |
2285 | } else { | |
d99a0153 CW |
2286 | char buf[ZFS_MAX_DATASET_NAME_LEN]; |
2287 | if (dsl_get_prev_snap(ds, buf) == 0) | |
2288 | dsl_prop_nvlist_add_string(nv, ZFS_PROP_PREV_SNAP, | |
2289 | buf); | |
6f1ffb06 MA |
2290 | dsl_dir_stats(ds->ds_dir, nv); |
2291 | } | |
34dc7c2f | 2292 | |
d99a0153 CW |
2293 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE, |
2294 | dsl_get_available(ds)); | |
2295 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED, | |
2296 | dsl_get_referenced(ds)); | |
34dc7c2f | 2297 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION, |
d99a0153 | 2298 | dsl_get_creation(ds)); |
34dc7c2f | 2299 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG, |
d99a0153 | 2300 | dsl_get_creationtxg(ds)); |
34dc7c2f | 2301 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA, |
d99a0153 | 2302 | dsl_get_refquota(ds)); |
34dc7c2f | 2303 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION, |
d99a0153 | 2304 | dsl_get_refreservation(ds)); |
b128c09f | 2305 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID, |
d99a0153 | 2306 | dsl_get_guid(ds)); |
428870ff | 2307 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE, |
d99a0153 | 2308 | dsl_get_unique(ds)); |
428870ff | 2309 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID, |
d99a0153 | 2310 | dsl_get_objsetid(ds)); |
428870ff | 2311 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS, |
d99a0153 | 2312 | dsl_get_userrefs(ds)); |
45d1cae3 | 2313 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY, |
d99a0153 | 2314 | dsl_get_defer_destroy(ds)); |
b5256303 | 2315 | dsl_dataset_crypt_stats(ds, nv); |
34dc7c2f | 2316 | |
d683ddbb | 2317 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { |
d99a0153 CW |
2318 | uint64_t written; |
2319 | if (dsl_get_written(ds, &written) == 0) { | |
2320 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN, | |
2321 | written); | |
330d06f9 MA |
2322 | } |
2323 | } | |
2324 | ||
47dfff3b | 2325 | if (!dsl_dataset_is_snapshot(ds)) { |
47dfff3b MA |
2326 | /* |
2327 | * A failed "newfs" (e.g. full) resumable receive leaves | |
2328 | * the stats set on this dataset. Check here for the prop. | |
2329 | */ | |
2330 | get_receive_resume_stats(ds, nv); | |
2331 | ||
2332 | /* | |
2333 | * A failed incremental resumable receive leaves the | |
2334 | * stats set on our child named "%recv". Check the child | |
2335 | * for the prop. | |
2336 | */ | |
1c27024e DB |
2337 | /* 6 extra bytes for /%recv */ |
2338 | char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; | |
2339 | dsl_dataset_t *recv_ds; | |
47dfff3b | 2340 | dsl_dataset_name(ds, recvname); |
eca7b760 IK |
2341 | if (strlcat(recvname, "/", sizeof (recvname)) < |
2342 | sizeof (recvname) && | |
2343 | strlcat(recvname, recv_clone_name, sizeof (recvname)) < | |
2344 | sizeof (recvname) && | |
2345 | dsl_dataset_hold(dp, recvname, FTAG, &recv_ds) == 0) { | |
47dfff3b MA |
2346 | get_receive_resume_stats(recv_ds, nv); |
2347 | dsl_dataset_rele(recv_ds, FTAG); | |
2348 | } | |
2349 | } | |
34dc7c2f BB |
2350 | } |
2351 | ||
2352 | void | |
2353 | dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat) | |
2354 | { | |
d99a0153 | 2355 | ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool); |
13fe0198 MA |
2356 | ASSERT(dsl_pool_config_held(dp)); |
2357 | ||
d99a0153 CW |
2358 | stat->dds_creation_txg = dsl_get_creationtxg(ds); |
2359 | stat->dds_inconsistent = dsl_get_inconsistent(ds); | |
2360 | stat->dds_guid = dsl_get_guid(ds); | |
6f1ffb06 | 2361 | stat->dds_origin[0] = '\0'; |
0c66c32d | 2362 | if (ds->ds_is_snapshot) { |
34dc7c2f | 2363 | stat->dds_is_snapshot = B_TRUE; |
d99a0153 | 2364 | stat->dds_num_clones = dsl_get_numclones(ds); |
fb5f0bc8 BB |
2365 | } else { |
2366 | stat->dds_is_snapshot = B_FALSE; | |
2367 | stat->dds_num_clones = 0; | |
34dc7c2f | 2368 | |
6f1ffb06 | 2369 | if (dsl_dir_is_clone(ds->ds_dir)) { |
d99a0153 | 2370 | dsl_dir_get_origin(ds->ds_dir, stat->dds_origin); |
6f1ffb06 | 2371 | } |
34dc7c2f | 2372 | } |
34dc7c2f BB |
2373 | } |
2374 | ||
2375 | uint64_t | |
2376 | dsl_dataset_fsid_guid(dsl_dataset_t *ds) | |
2377 | { | |
2378 | return (ds->ds_fsid_guid); | |
2379 | } | |
2380 | ||
2381 | void | |
2382 | dsl_dataset_space(dsl_dataset_t *ds, | |
2383 | uint64_t *refdbytesp, uint64_t *availbytesp, | |
2384 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
2385 | { | |
d683ddbb | 2386 | *refdbytesp = dsl_dataset_phys(ds)->ds_referenced_bytes; |
34dc7c2f | 2387 | *availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE); |
d683ddbb JG |
2388 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) |
2389 | *availbytesp += | |
2390 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes; | |
34dc7c2f BB |
2391 | if (ds->ds_quota != 0) { |
2392 | /* | |
2393 | * Adjust available bytes according to refquota | |
2394 | */ | |
2395 | if (*refdbytesp < ds->ds_quota) | |
2396 | *availbytesp = MIN(*availbytesp, | |
2397 | ds->ds_quota - *refdbytesp); | |
2398 | else | |
2399 | *availbytesp = 0; | |
2400 | } | |
cc9bb3e5 | 2401 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 2402 | *usedobjsp = BP_GET_FILL(&dsl_dataset_phys(ds)->ds_bp); |
cc9bb3e5 | 2403 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f BB |
2404 | *availobjsp = DN_MAX_OBJECT - *usedobjsp; |
2405 | } | |
2406 | ||
2407 | boolean_t | |
19580676 | 2408 | dsl_dataset_modified_since_snap(dsl_dataset_t *ds, dsl_dataset_t *snap) |
34dc7c2f | 2409 | { |
cc9bb3e5 GM |
2410 | ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool); |
2411 | uint64_t birth; | |
2412 | ||
2413 | ASSERT(dsl_pool_config_held(dp)); | |
19580676 | 2414 | if (snap == NULL) |
34dc7c2f | 2415 | return (B_FALSE); |
cc9bb3e5 GM |
2416 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
2417 | birth = dsl_dataset_get_blkptr(ds)->blk_birth; | |
2418 | rrw_exit(&ds->ds_bp_rwlock, FTAG); | |
2419 | if (birth > dsl_dataset_phys(snap)->ds_creation_txg) { | |
19580676 | 2420 | objset_t *os, *os_snap; |
572e2857 BB |
2421 | /* |
2422 | * It may be that only the ZIL differs, because it was | |
2423 | * reset in the head. Don't count that as being | |
2424 | * modified. | |
2425 | */ | |
2426 | if (dmu_objset_from_ds(ds, &os) != 0) | |
2427 | return (B_TRUE); | |
19580676 | 2428 | if (dmu_objset_from_ds(snap, &os_snap) != 0) |
572e2857 BB |
2429 | return (B_TRUE); |
2430 | return (bcmp(&os->os_phys->os_meta_dnode, | |
19580676 | 2431 | &os_snap->os_phys->os_meta_dnode, |
572e2857 BB |
2432 | sizeof (os->os_phys->os_meta_dnode)) != 0); |
2433 | } | |
34dc7c2f BB |
2434 | return (B_FALSE); |
2435 | } | |
2436 | ||
13fe0198 MA |
2437 | typedef struct dsl_dataset_rename_snapshot_arg { |
2438 | const char *ddrsa_fsname; | |
2439 | const char *ddrsa_oldsnapname; | |
2440 | const char *ddrsa_newsnapname; | |
2441 | boolean_t ddrsa_recursive; | |
2442 | dmu_tx_t *ddrsa_tx; | |
2443 | } dsl_dataset_rename_snapshot_arg_t; | |
2444 | ||
34dc7c2f BB |
2445 | /* ARGSUSED */ |
2446 | static int | |
13fe0198 MA |
2447 | dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp, |
2448 | dsl_dataset_t *hds, void *arg) | |
34dc7c2f | 2449 | { |
13fe0198 MA |
2450 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2451 | int error; | |
34dc7c2f | 2452 | uint64_t val; |
34dc7c2f | 2453 | |
13fe0198 MA |
2454 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); |
2455 | if (error != 0) { | |
2456 | /* ignore nonexistent snapshots */ | |
2457 | return (error == ENOENT ? 0 : error); | |
2458 | } | |
34dc7c2f | 2459 | |
13fe0198 MA |
2460 | /* new name should not exist */ |
2461 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val); | |
2462 | if (error == 0) | |
2e528b49 | 2463 | error = SET_ERROR(EEXIST); |
13fe0198 MA |
2464 | else if (error == ENOENT) |
2465 | error = 0; | |
34dc7c2f BB |
2466 | |
2467 | /* dataset name + 1 for the "@" + the new snapshot name must fit */ | |
13fe0198 | 2468 | if (dsl_dir_namelen(hds->ds_dir) + 1 + |
eca7b760 | 2469 | strlen(ddrsa->ddrsa_newsnapname) >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2470 | error = SET_ERROR(ENAMETOOLONG); |
34dc7c2f | 2471 | |
13fe0198 | 2472 | return (error); |
34dc7c2f BB |
2473 | } |
2474 | ||
13fe0198 MA |
2475 | static int |
2476 | dsl_dataset_rename_snapshot_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 2477 | { |
13fe0198 MA |
2478 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2479 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 2480 | dsl_dataset_t *hds; |
13fe0198 | 2481 | int error; |
34dc7c2f | 2482 | |
13fe0198 MA |
2483 | error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds); |
2484 | if (error != 0) | |
2485 | return (error); | |
34dc7c2f | 2486 | |
13fe0198 MA |
2487 | if (ddrsa->ddrsa_recursive) { |
2488 | error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object, | |
2489 | dsl_dataset_rename_snapshot_check_impl, ddrsa, | |
2490 | DS_FIND_CHILDREN); | |
2491 | } else { | |
2492 | error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa); | |
2493 | } | |
b128c09f | 2494 | dsl_dataset_rele(hds, FTAG); |
13fe0198 | 2495 | return (error); |
34dc7c2f BB |
2496 | } |
2497 | ||
34dc7c2f | 2498 | static int |
13fe0198 MA |
2499 | dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp, |
2500 | dsl_dataset_t *hds, void *arg) | |
34dc7c2f | 2501 | { |
13fe0198 MA |
2502 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2503 | dsl_dataset_t *ds; | |
2504 | uint64_t val; | |
2505 | dmu_tx_t *tx = ddrsa->ddrsa_tx; | |
2506 | int error; | |
34dc7c2f | 2507 | |
13fe0198 MA |
2508 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); |
2509 | ASSERT(error == 0 || error == ENOENT); | |
2510 | if (error == ENOENT) { | |
2511 | /* ignore nonexistent snapshots */ | |
2512 | return (0); | |
34dc7c2f BB |
2513 | } |
2514 | ||
13fe0198 MA |
2515 | VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds)); |
2516 | ||
2517 | /* log before we change the name */ | |
2518 | spa_history_log_internal_ds(ds, "rename", tx, | |
2519 | "-> @%s", ddrsa->ddrsa_newsnapname); | |
34dc7c2f | 2520 | |
788eb90c JJ |
2521 | VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx, |
2522 | B_FALSE)); | |
13fe0198 | 2523 | mutex_enter(&ds->ds_lock); |
c9d61adb | 2524 | (void) strlcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname, |
2525 | sizeof (ds->ds_snapname)); | |
13fe0198 | 2526 | mutex_exit(&ds->ds_lock); |
d683ddbb JG |
2527 | VERIFY0(zap_add(dp->dp_meta_objset, |
2528 | dsl_dataset_phys(hds)->ds_snapnames_zapobj, | |
13fe0198 | 2529 | ds->ds_snapname, 8, 1, &ds->ds_object, tx)); |
a0bd735a BP |
2530 | zvol_rename_minors(dp->dp_spa, ddrsa->ddrsa_oldsnapname, |
2531 | ddrsa->ddrsa_newsnapname, B_TRUE); | |
34dc7c2f | 2532 | |
13fe0198 | 2533 | dsl_dataset_rele(ds, FTAG); |
34dc7c2f BB |
2534 | return (0); |
2535 | } | |
2536 | ||
13fe0198 MA |
2537 | static void |
2538 | dsl_dataset_rename_snapshot_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 2539 | { |
13fe0198 MA |
2540 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2541 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 2542 | dsl_dataset_t *hds = NULL; |
34dc7c2f | 2543 | |
13fe0198 MA |
2544 | VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds)); |
2545 | ddrsa->ddrsa_tx = tx; | |
2546 | if (ddrsa->ddrsa_recursive) { | |
2547 | VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object, | |
2548 | dsl_dataset_rename_snapshot_sync_impl, ddrsa, | |
2549 | DS_FIND_CHILDREN)); | |
2550 | } else { | |
2551 | VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa)); | |
34dc7c2f | 2552 | } |
13fe0198 | 2553 | dsl_dataset_rele(hds, FTAG); |
34dc7c2f BB |
2554 | } |
2555 | ||
13fe0198 MA |
2556 | int |
2557 | dsl_dataset_rename_snapshot(const char *fsname, | |
2558 | const char *oldsnapname, const char *newsnapname, boolean_t recursive) | |
34dc7c2f | 2559 | { |
13fe0198 | 2560 | dsl_dataset_rename_snapshot_arg_t ddrsa; |
34dc7c2f | 2561 | |
13fe0198 MA |
2562 | ddrsa.ddrsa_fsname = fsname; |
2563 | ddrsa.ddrsa_oldsnapname = oldsnapname; | |
2564 | ddrsa.ddrsa_newsnapname = newsnapname; | |
2565 | ddrsa.ddrsa_recursive = recursive; | |
34dc7c2f | 2566 | |
a0bd735a | 2567 | return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check, |
3d45fdd6 | 2568 | dsl_dataset_rename_snapshot_sync, &ddrsa, |
a0bd735a | 2569 | 1, ZFS_SPACE_CHECK_RESERVED)); |
34dc7c2f BB |
2570 | } |
2571 | ||
831baf06 KW |
2572 | /* |
2573 | * If we're doing an ownership handoff, we need to make sure that there is | |
2574 | * only one long hold on the dataset. We're not allowed to change anything here | |
2575 | * so we don't permanently release the long hold or regular hold here. We want | |
2576 | * to do this only when syncing to avoid the dataset unexpectedly going away | |
040dab99 | 2577 | * when we release the long hold. |
831baf06 KW |
2578 | */ |
2579 | static int | |
2580 | dsl_dataset_handoff_check(dsl_dataset_t *ds, void *owner, dmu_tx_t *tx) | |
2581 | { | |
2582 | boolean_t held; | |
2583 | ||
2584 | if (!dmu_tx_is_syncing(tx)) | |
2585 | return (0); | |
2586 | ||
2587 | if (owner != NULL) { | |
2588 | VERIFY3P(ds->ds_owner, ==, owner); | |
2589 | dsl_dataset_long_rele(ds, owner); | |
2590 | } | |
2591 | ||
040dab99 | 2592 | held = dsl_dataset_long_held(ds); |
831baf06 KW |
2593 | |
2594 | if (owner != NULL) | |
2595 | dsl_dataset_long_hold(ds, owner); | |
2596 | ||
2597 | if (held) | |
2598 | return (SET_ERROR(EBUSY)); | |
2599 | ||
2600 | return (0); | |
2601 | } | |
2602 | ||
af073689 | 2603 | int |
13fe0198 | 2604 | dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 2605 | { |
831baf06 | 2606 | dsl_dataset_rollback_arg_t *ddra = arg; |
13fe0198 | 2607 | dsl_pool_t *dp = dmu_tx_pool(tx); |
34dc7c2f | 2608 | dsl_dataset_t *ds; |
13fe0198 MA |
2609 | int64_t unused_refres_delta; |
2610 | int error; | |
34dc7c2f | 2611 | |
831baf06 | 2612 | error = dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds); |
13fe0198 MA |
2613 | if (error != 0) |
2614 | return (error); | |
428870ff | 2615 | |
13fe0198 | 2616 | /* must not be a snapshot */ |
0c66c32d | 2617 | if (ds->ds_is_snapshot) { |
13fe0198 | 2618 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 2619 | return (SET_ERROR(EINVAL)); |
13fe0198 | 2620 | } |
34dc7c2f | 2621 | |
13fe0198 | 2622 | /* must have a most recent snapshot */ |
d683ddbb | 2623 | if (dsl_dataset_phys(ds)->ds_prev_snap_txg < TXG_INITIAL) { |
13fe0198 | 2624 | dsl_dataset_rele(ds, FTAG); |
13342832 | 2625 | return (SET_ERROR(ESRCH)); |
13fe0198 | 2626 | } |
34dc7c2f | 2627 | |
0efd9791 AG |
2628 | /* |
2629 | * No rollback to a snapshot created in the current txg, because | |
2630 | * the rollback may dirty the dataset and create blocks that are | |
2631 | * not reachable from the rootbp while having a birth txg that | |
2632 | * falls into the snapshot's range. | |
2633 | */ | |
2634 | if (dmu_tx_is_syncing(tx) && | |
2635 | dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) { | |
2636 | dsl_dataset_rele(ds, FTAG); | |
2637 | return (SET_ERROR(EAGAIN)); | |
2638 | } | |
2639 | ||
8ca78ab0 AG |
2640 | /* |
2641 | * If the expected target snapshot is specified, then check that | |
2642 | * the latest snapshot is it. | |
2643 | */ | |
2644 | if (ddra->ddra_tosnap != NULL) { | |
13342832 AG |
2645 | dsl_dataset_t *snapds; |
2646 | ||
2647 | /* Check if the target snapshot exists at all. */ | |
2648 | error = dsl_dataset_hold(dp, ddra->ddra_tosnap, FTAG, &snapds); | |
2649 | if (error != 0) { | |
2650 | /* | |
2651 | * ESRCH is used to signal that the target snapshot does | |
2652 | * not exist, while ENOENT is used to report that | |
2653 | * the rolled back dataset does not exist. | |
2654 | * ESRCH is also used to cover other cases where the | |
2655 | * target snapshot is not related to the dataset being | |
2656 | * rolled back such as being in a different pool. | |
2657 | */ | |
2658 | if (error == ENOENT || error == EXDEV) | |
2659 | error = SET_ERROR(ESRCH); | |
2660 | dsl_dataset_rele(ds, FTAG); | |
2661 | return (error); | |
2662 | } | |
2663 | ASSERT(snapds->ds_is_snapshot); | |
8ca78ab0 | 2664 | |
13342832 AG |
2665 | /* Check if the snapshot is the latest snapshot indeed. */ |
2666 | if (snapds != ds->ds_prev) { | |
2667 | /* | |
2668 | * Distinguish between the case where the only problem | |
2669 | * is intervening snapshots (EEXIST) vs the snapshot | |
2670 | * not being a valid target for rollback (ESRCH). | |
2671 | */ | |
2672 | if (snapds->ds_dir == ds->ds_dir || | |
2673 | (dsl_dir_is_clone(ds->ds_dir) && | |
2674 | dsl_dir_phys(ds->ds_dir)->dd_origin_obj == | |
2675 | snapds->ds_object)) { | |
2676 | error = SET_ERROR(EEXIST); | |
2677 | } else { | |
2678 | error = SET_ERROR(ESRCH); | |
2679 | } | |
2680 | dsl_dataset_rele(snapds, FTAG); | |
2681 | dsl_dataset_rele(ds, FTAG); | |
2682 | return (error); | |
2683 | } | |
2684 | dsl_dataset_rele(snapds, FTAG); | |
8ca78ab0 AG |
2685 | } |
2686 | ||
da536844 | 2687 | /* must not have any bookmarks after the most recent snapshot */ |
1c27024e | 2688 | nvlist_t *proprequest = fnvlist_alloc(); |
da536844 | 2689 | fnvlist_add_boolean(proprequest, zfs_prop_to_name(ZFS_PROP_CREATETXG)); |
1c27024e | 2690 | nvlist_t *bookmarks = fnvlist_alloc(); |
da536844 MA |
2691 | error = dsl_get_bookmarks_impl(ds, proprequest, bookmarks); |
2692 | fnvlist_free(proprequest); | |
13342832 AG |
2693 | if (error != 0) { |
2694 | dsl_dataset_rele(ds, FTAG); | |
da536844 | 2695 | return (error); |
13342832 | 2696 | } |
1c27024e | 2697 | for (nvpair_t *pair = nvlist_next_nvpair(bookmarks, NULL); |
da536844 MA |
2698 | pair != NULL; pair = nvlist_next_nvpair(bookmarks, pair)) { |
2699 | nvlist_t *valuenv = | |
2700 | fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair), | |
2701 | zfs_prop_to_name(ZFS_PROP_CREATETXG)); | |
2702 | uint64_t createtxg = fnvlist_lookup_uint64(valuenv, "value"); | |
d683ddbb | 2703 | if (createtxg > dsl_dataset_phys(ds)->ds_prev_snap_txg) { |
da536844 MA |
2704 | fnvlist_free(bookmarks); |
2705 | dsl_dataset_rele(ds, FTAG); | |
2706 | return (SET_ERROR(EEXIST)); | |
2707 | } | |
2708 | } | |
2709 | fnvlist_free(bookmarks); | |
2710 | ||
831baf06 KW |
2711 | error = dsl_dataset_handoff_check(ds, ddra->ddra_owner, tx); |
2712 | if (error != 0) { | |
13fe0198 | 2713 | dsl_dataset_rele(ds, FTAG); |
831baf06 | 2714 | return (error); |
34dc7c2f | 2715 | } |
428870ff | 2716 | |
13fe0198 MA |
2717 | /* |
2718 | * Check if the snap we are rolling back to uses more than | |
2719 | * the refquota. | |
2720 | */ | |
2721 | if (ds->ds_quota != 0 && | |
d683ddbb | 2722 | dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes > ds->ds_quota) { |
13fe0198 | 2723 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 2724 | return (SET_ERROR(EDQUOT)); |
34dc7c2f BB |
2725 | } |
2726 | ||
13fe0198 MA |
2727 | /* |
2728 | * When we do the clone swap, we will temporarily use more space | |
2729 | * due to the refreservation (the head will no longer have any | |
2730 | * unique space, so the entire amount of the refreservation will need | |
2731 | * to be free). We will immediately destroy the clone, freeing | |
2732 | * this space, but the freeing happens over many txg's. | |
2733 | */ | |
2734 | unused_refres_delta = (int64_t)MIN(ds->ds_reserved, | |
d683ddbb | 2735 | dsl_dataset_phys(ds)->ds_unique_bytes); |
34dc7c2f | 2736 | |
13fe0198 MA |
2737 | if (unused_refres_delta > 0 && |
2738 | unused_refres_delta > | |
2739 | dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) { | |
2740 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 2741 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 2742 | } |
34dc7c2f | 2743 | |
13fe0198 MA |
2744 | dsl_dataset_rele(ds, FTAG); |
2745 | return (0); | |
2746 | } | |
34dc7c2f | 2747 | |
af073689 | 2748 | void |
13fe0198 MA |
2749 | dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx) |
2750 | { | |
831baf06 | 2751 | dsl_dataset_rollback_arg_t *ddra = arg; |
13fe0198 MA |
2752 | dsl_pool_t *dp = dmu_tx_pool(tx); |
2753 | dsl_dataset_t *ds, *clone; | |
2754 | uint64_t cloneobj; | |
eca7b760 | 2755 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 2756 | |
831baf06 | 2757 | VERIFY0(dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds)); |
34dc7c2f | 2758 | |
46ba1e59 MA |
2759 | dsl_dataset_name(ds->ds_prev, namebuf); |
2760 | fnvlist_add_string(ddra->ddra_result, "target", namebuf); | |
2761 | ||
13fe0198 | 2762 | cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback", |
b5256303 | 2763 | ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, NULL, tx); |
13fe0198 MA |
2764 | |
2765 | VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone)); | |
2766 | ||
2767 | dsl_dataset_clone_swap_sync_impl(clone, ds, tx); | |
2768 | dsl_dataset_zero_zil(ds, tx); | |
2769 | ||
2770 | dsl_destroy_head_sync_impl(clone, tx); | |
2771 | ||
2772 | dsl_dataset_rele(clone, FTAG); | |
2773 | dsl_dataset_rele(ds, FTAG); | |
2774 | } | |
2775 | ||
831baf06 | 2776 | /* |
46ba1e59 MA |
2777 | * Rolls back the given filesystem or volume to the most recent snapshot. |
2778 | * The name of the most recent snapshot will be returned under key "target" | |
2779 | * in the result nvlist. | |
831baf06 | 2780 | * |
46ba1e59 | 2781 | * If owner != NULL: |
831baf06 KW |
2782 | * - The existing dataset MUST be owned by the specified owner at entry |
2783 | * - Upon return, dataset will still be held by the same owner, whether we | |
2784 | * succeed or not. | |
2785 | * | |
2786 | * This mode is required any time the existing filesystem is mounted. See | |
2787 | * notes above zfs_suspend_fs() for further details. | |
2788 | */ | |
13fe0198 | 2789 | int |
8ca78ab0 AG |
2790 | dsl_dataset_rollback(const char *fsname, const char *tosnap, void *owner, |
2791 | nvlist_t *result) | |
13fe0198 | 2792 | { |
831baf06 KW |
2793 | dsl_dataset_rollback_arg_t ddra; |
2794 | ||
2795 | ddra.ddra_fsname = fsname; | |
8ca78ab0 | 2796 | ddra.ddra_tosnap = tosnap; |
831baf06 | 2797 | ddra.ddra_owner = owner; |
46ba1e59 | 2798 | ddra.ddra_result = result; |
831baf06 | 2799 | |
13fe0198 | 2800 | return (dsl_sync_task(fsname, dsl_dataset_rollback_check, |
3d45fdd6 MA |
2801 | dsl_dataset_rollback_sync, &ddra, |
2802 | 1, ZFS_SPACE_CHECK_RESERVED)); | |
34dc7c2f BB |
2803 | } |
2804 | ||
b128c09f BB |
2805 | struct promotenode { |
2806 | list_node_t link; | |
2807 | dsl_dataset_t *ds; | |
2808 | }; | |
2809 | ||
b128c09f | 2810 | static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep); |
13fe0198 MA |
2811 | static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, |
2812 | void *tag); | |
2813 | static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag); | |
b128c09f | 2814 | |
d99a0153 | 2815 | int |
13fe0198 | 2816 | dsl_dataset_promote_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 2817 | { |
13fe0198 MA |
2818 | dsl_dataset_promote_arg_t *ddpa = arg; |
2819 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
2820 | dsl_dataset_t *hds; | |
2821 | struct promotenode *snap; | |
2822 | dsl_dataset_t *origin_ds; | |
34dc7c2f | 2823 | int err; |
428870ff | 2824 | uint64_t unused; |
788eb90c | 2825 | uint64_t ss_mv_cnt; |
fec41709 | 2826 | size_t max_snap_len; |
d99a0153 | 2827 | boolean_t conflicting_snaps; |
34dc7c2f | 2828 | |
13fe0198 MA |
2829 | err = promote_hold(ddpa, dp, FTAG); |
2830 | if (err != 0) | |
2831 | return (err); | |
34dc7c2f | 2832 | |
13fe0198 | 2833 | hds = ddpa->ddpa_clone; |
fec41709 | 2834 | max_snap_len = MAXNAMELEN - strlen(ddpa->ddpa_clonename) - 1; |
34dc7c2f | 2835 | |
d683ddbb | 2836 | if (dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE) { |
13fe0198 | 2837 | promote_rele(ddpa, FTAG); |
2e528b49 | 2838 | return (SET_ERROR(EXDEV)); |
13fe0198 MA |
2839 | } |
2840 | ||
b5256303 TC |
2841 | snap = list_head(&ddpa->shared_snaps); |
2842 | if (snap == NULL) { | |
2843 | err = SET_ERROR(ENOENT); | |
2844 | goto out; | |
2845 | } | |
2846 | origin_ds = snap->ds; | |
2847 | ||
2848 | /* | |
2849 | * Encrypted clones share a DSL Crypto Key with their origin's dsl dir. | |
2850 | * When doing a promote we must make sure the encryption root for | |
2851 | * both the target and the target's origin does not change to avoid | |
2852 | * needing to rewrap encryption keys | |
2853 | */ | |
2854 | err = dsl_dataset_promote_crypt_check(hds->ds_dir, origin_ds->ds_dir); | |
2855 | if (err != 0) | |
2856 | goto out; | |
2857 | ||
13fe0198 MA |
2858 | /* |
2859 | * Compute and check the amount of space to transfer. Since this is | |
2860 | * so expensive, don't do the preliminary check. | |
2861 | */ | |
2862 | if (!dmu_tx_is_syncing(tx)) { | |
2863 | promote_rele(ddpa, FTAG); | |
2864 | return (0); | |
2865 | } | |
2866 | ||
34dc7c2f | 2867 | /* compute origin's new unique space */ |
13fe0198 | 2868 | snap = list_tail(&ddpa->clone_snaps); |
cbce5813 | 2869 | ASSERT(snap != NULL); |
d683ddbb JG |
2870 | ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, |
2871 | origin_ds->ds_object); | |
428870ff | 2872 | dsl_deadlist_space_range(&snap->ds->ds_deadlist, |
d683ddbb | 2873 | dsl_dataset_phys(origin_ds)->ds_prev_snap_txg, UINT64_MAX, |
13fe0198 | 2874 | &ddpa->unique, &unused, &unused); |
34dc7c2f | 2875 | |
b128c09f BB |
2876 | /* |
2877 | * Walk the snapshots that we are moving | |
2878 | * | |
2879 | * Compute space to transfer. Consider the incremental changes | |
13fe0198 | 2880 | * to used by each snapshot: |
b128c09f BB |
2881 | * (my used) = (prev's used) + (blocks born) - (blocks killed) |
2882 | * So each snapshot gave birth to: | |
2883 | * (blocks born) = (my used) - (prev's used) + (blocks killed) | |
2884 | * So a sequence would look like: | |
2885 | * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0) | |
2886 | * Which simplifies to: | |
2887 | * uN + kN + kN-1 + ... + k1 + k0 | |
2888 | * Note however, if we stop before we reach the ORIGIN we get: | |
2889 | * uN + kN + kN-1 + ... + kM - uM-1 | |
2890 | */ | |
d99a0153 | 2891 | conflicting_snaps = B_FALSE; |
788eb90c | 2892 | ss_mv_cnt = 0; |
d683ddbb JG |
2893 | ddpa->used = dsl_dataset_phys(origin_ds)->ds_referenced_bytes; |
2894 | ddpa->comp = dsl_dataset_phys(origin_ds)->ds_compressed_bytes; | |
2895 | ddpa->uncomp = dsl_dataset_phys(origin_ds)->ds_uncompressed_bytes; | |
13fe0198 MA |
2896 | for (snap = list_head(&ddpa->shared_snaps); snap; |
2897 | snap = list_next(&ddpa->shared_snaps, snap)) { | |
34dc7c2f | 2898 | uint64_t val, dlused, dlcomp, dluncomp; |
b128c09f | 2899 | dsl_dataset_t *ds = snap->ds; |
34dc7c2f | 2900 | |
788eb90c JJ |
2901 | ss_mv_cnt++; |
2902 | ||
13fe0198 MA |
2903 | /* |
2904 | * If there are long holds, we won't be able to evict | |
2905 | * the objset. | |
2906 | */ | |
2907 | if (dsl_dataset_long_held(ds)) { | |
2e528b49 | 2908 | err = SET_ERROR(EBUSY); |
13fe0198 MA |
2909 | goto out; |
2910 | } | |
2911 | ||
34dc7c2f | 2912 | /* Check that the snapshot name does not conflict */ |
13fe0198 | 2913 | VERIFY0(dsl_dataset_get_snapname(ds)); |
fec41709 AG |
2914 | if (strlen(ds->ds_snapname) >= max_snap_len) { |
2915 | err = SET_ERROR(ENAMETOOLONG); | |
2916 | goto out; | |
2917 | } | |
b128c09f | 2918 | err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val); |
428870ff | 2919 | if (err == 0) { |
d99a0153 CW |
2920 | fnvlist_add_boolean(ddpa->err_ds, |
2921 | snap->ds->ds_snapname); | |
2922 | conflicting_snaps = B_TRUE; | |
2923 | } else if (err != ENOENT) { | |
428870ff BB |
2924 | goto out; |
2925 | } | |
34dc7c2f | 2926 | |
b128c09f | 2927 | /* The very first snapshot does not have a deadlist */ |
d683ddbb | 2928 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj == 0) |
b128c09f | 2929 | continue; |
34dc7c2f | 2930 | |
428870ff BB |
2931 | dsl_deadlist_space(&ds->ds_deadlist, |
2932 | &dlused, &dlcomp, &dluncomp); | |
13fe0198 MA |
2933 | ddpa->used += dlused; |
2934 | ddpa->comp += dlcomp; | |
2935 | ddpa->uncomp += dluncomp; | |
b128c09f | 2936 | } |
34dc7c2f | 2937 | |
d99a0153 CW |
2938 | /* |
2939 | * In order to return the full list of conflicting snapshots, we check | |
2940 | * whether there was a conflict after traversing all of them. | |
2941 | */ | |
2942 | if (conflicting_snaps) { | |
2943 | err = SET_ERROR(EEXIST); | |
2944 | goto out; | |
2945 | } | |
2946 | ||
b128c09f BB |
2947 | /* |
2948 | * If we are a clone of a clone then we never reached ORIGIN, | |
2949 | * so we need to subtract out the clone origin's used space. | |
2950 | */ | |
13fe0198 | 2951 | if (ddpa->origin_origin) { |
d683ddbb JG |
2952 | ddpa->used -= |
2953 | dsl_dataset_phys(ddpa->origin_origin)->ds_referenced_bytes; | |
2954 | ddpa->comp -= | |
2955 | dsl_dataset_phys(ddpa->origin_origin)->ds_compressed_bytes; | |
13fe0198 | 2956 | ddpa->uncomp -= |
d683ddbb JG |
2957 | dsl_dataset_phys(ddpa->origin_origin)-> |
2958 | ds_uncompressed_bytes; | |
34dc7c2f BB |
2959 | } |
2960 | ||
788eb90c | 2961 | /* Check that there is enough space and limit headroom here */ |
b128c09f | 2962 | err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir, |
788eb90c | 2963 | 0, ss_mv_cnt, ddpa->used, ddpa->cr); |
13fe0198 MA |
2964 | if (err != 0) |
2965 | goto out; | |
34dc7c2f | 2966 | |
b128c09f BB |
2967 | /* |
2968 | * Compute the amounts of space that will be used by snapshots | |
2969 | * after the promotion (for both origin and clone). For each, | |
2970 | * it is the amount of space that will be on all of their | |
2971 | * deadlists (that was not born before their new origin). | |
2972 | */ | |
d683ddbb | 2973 | if (dsl_dir_phys(hds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) { |
b128c09f BB |
2974 | uint64_t space; |
2975 | ||
2976 | /* | |
2977 | * Note, typically this will not be a clone of a clone, | |
428870ff BB |
2978 | * so dd_origin_txg will be < TXG_INITIAL, so |
2979 | * these snaplist_space() -> dsl_deadlist_space_range() | |
b128c09f BB |
2980 | * calls will be fast because they do not have to |
2981 | * iterate over all bps. | |
2982 | */ | |
13fe0198 | 2983 | snap = list_head(&ddpa->origin_snaps); |
0a8f18f9 | 2984 | if (snap == NULL) { |
2985 | err = SET_ERROR(ENOENT); | |
2986 | goto out; | |
2987 | } | |
13fe0198 MA |
2988 | err = snaplist_space(&ddpa->shared_snaps, |
2989 | snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap); | |
2990 | if (err != 0) | |
2991 | goto out; | |
b128c09f | 2992 | |
13fe0198 | 2993 | err = snaplist_space(&ddpa->clone_snaps, |
428870ff | 2994 | snap->ds->ds_dir->dd_origin_txg, &space); |
13fe0198 MA |
2995 | if (err != 0) |
2996 | goto out; | |
2997 | ddpa->cloneusedsnap += space; | |
b128c09f | 2998 | } |
d683ddbb JG |
2999 | if (dsl_dir_phys(origin_ds->ds_dir)->dd_flags & |
3000 | DD_FLAG_USED_BREAKDOWN) { | |
13fe0198 | 3001 | err = snaplist_space(&ddpa->origin_snaps, |
d683ddbb JG |
3002 | dsl_dataset_phys(origin_ds)->ds_creation_txg, |
3003 | &ddpa->originusedsnap); | |
13fe0198 MA |
3004 | if (err != 0) |
3005 | goto out; | |
b128c09f BB |
3006 | } |
3007 | ||
428870ff | 3008 | out: |
13fe0198 | 3009 | promote_rele(ddpa, FTAG); |
428870ff | 3010 | return (err); |
34dc7c2f BB |
3011 | } |
3012 | ||
d99a0153 | 3013 | void |
13fe0198 | 3014 | dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 3015 | { |
13fe0198 MA |
3016 | dsl_dataset_promote_arg_t *ddpa = arg; |
3017 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3018 | dsl_dataset_t *hds; | |
3019 | struct promotenode *snap; | |
3020 | dsl_dataset_t *origin_ds; | |
b128c09f | 3021 | dsl_dataset_t *origin_head; |
13fe0198 | 3022 | dsl_dir_t *dd; |
34dc7c2f | 3023 | dsl_dir_t *odd = NULL; |
b128c09f BB |
3024 | uint64_t oldnext_obj; |
3025 | int64_t delta; | |
34dc7c2f | 3026 | |
13fe0198 MA |
3027 | VERIFY0(promote_hold(ddpa, dp, FTAG)); |
3028 | hds = ddpa->ddpa_clone; | |
3029 | ||
d683ddbb | 3030 | ASSERT0(dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE); |
13fe0198 MA |
3031 | |
3032 | snap = list_head(&ddpa->shared_snaps); | |
3033 | origin_ds = snap->ds; | |
3034 | dd = hds->ds_dir; | |
34dc7c2f | 3035 | |
13fe0198 | 3036 | snap = list_head(&ddpa->origin_snaps); |
b128c09f BB |
3037 | origin_head = snap->ds; |
3038 | ||
34dc7c2f BB |
3039 | /* |
3040 | * We need to explicitly open odd, since origin_ds's dd will be | |
3041 | * changing. | |
3042 | */ | |
13fe0198 | 3043 | VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object, |
34dc7c2f BB |
3044 | NULL, FTAG, &odd)); |
3045 | ||
b5256303 TC |
3046 | dsl_dataset_promote_crypt_sync(hds->ds_dir, odd, tx); |
3047 | ||
b128c09f BB |
3048 | /* change origin's next snap */ |
3049 | dmu_buf_will_dirty(origin_ds->ds_dbuf, tx); | |
d683ddbb | 3050 | oldnext_obj = dsl_dataset_phys(origin_ds)->ds_next_snap_obj; |
13fe0198 | 3051 | snap = list_tail(&ddpa->clone_snaps); |
d683ddbb JG |
3052 | ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, |
3053 | origin_ds->ds_object); | |
3054 | dsl_dataset_phys(origin_ds)->ds_next_snap_obj = snap->ds->ds_object; | |
b128c09f BB |
3055 | |
3056 | /* change the origin's next clone */ | |
d683ddbb | 3057 | if (dsl_dataset_phys(origin_ds)->ds_next_clones_obj) { |
13fe0198 MA |
3058 | dsl_dataset_remove_from_next_clones(origin_ds, |
3059 | snap->ds->ds_object, tx); | |
3060 | VERIFY0(zap_add_int(dp->dp_meta_objset, | |
d683ddbb | 3061 | dsl_dataset_phys(origin_ds)->ds_next_clones_obj, |
b128c09f BB |
3062 | oldnext_obj, tx)); |
3063 | } | |
3064 | ||
3065 | /* change origin */ | |
3066 | dmu_buf_will_dirty(dd->dd_dbuf, tx); | |
d683ddbb JG |
3067 | ASSERT3U(dsl_dir_phys(dd)->dd_origin_obj, ==, origin_ds->ds_object); |
3068 | dsl_dir_phys(dd)->dd_origin_obj = dsl_dir_phys(odd)->dd_origin_obj; | |
428870ff | 3069 | dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg; |
b128c09f | 3070 | dmu_buf_will_dirty(odd->dd_dbuf, tx); |
d683ddbb | 3071 | dsl_dir_phys(odd)->dd_origin_obj = origin_ds->ds_object; |
428870ff | 3072 | origin_head->ds_dir->dd_origin_txg = |
d683ddbb | 3073 | dsl_dataset_phys(origin_ds)->ds_creation_txg; |
428870ff BB |
3074 | |
3075 | /* change dd_clone entries */ | |
3076 | if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { | |
13fe0198 | 3077 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3078 | dsl_dir_phys(odd)->dd_clones, hds->ds_object, tx)); |
13fe0198 | 3079 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3080 | dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, |
428870ff BB |
3081 | hds->ds_object, tx)); |
3082 | ||
13fe0198 | 3083 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3084 | dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, |
428870ff | 3085 | origin_head->ds_object, tx)); |
d683ddbb JG |
3086 | if (dsl_dir_phys(dd)->dd_clones == 0) { |
3087 | dsl_dir_phys(dd)->dd_clones = | |
3088 | zap_create(dp->dp_meta_objset, DMU_OT_DSL_CLONES, | |
3089 | DMU_OT_NONE, 0, tx); | |
428870ff | 3090 | } |
13fe0198 | 3091 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3092 | dsl_dir_phys(dd)->dd_clones, origin_head->ds_object, tx)); |
428870ff | 3093 | } |
34dc7c2f | 3094 | |
b128c09f | 3095 | /* move snapshots to this dir */ |
13fe0198 MA |
3096 | for (snap = list_head(&ddpa->shared_snaps); snap; |
3097 | snap = list_next(&ddpa->shared_snaps, snap)) { | |
b128c09f BB |
3098 | dsl_dataset_t *ds = snap->ds; |
3099 | ||
13fe0198 MA |
3100 | /* |
3101 | * Property callbacks are registered to a particular | |
3102 | * dsl_dir. Since ours is changing, evict the objset | |
3103 | * so that they will be unregistered from the old dsl_dir. | |
3104 | */ | |
428870ff BB |
3105 | if (ds->ds_objset) { |
3106 | dmu_objset_evict(ds->ds_objset); | |
3107 | ds->ds_objset = NULL; | |
b128c09f | 3108 | } |
13fe0198 | 3109 | |
34dc7c2f | 3110 | /* move snap name entry */ |
13fe0198 MA |
3111 | VERIFY0(dsl_dataset_get_snapname(ds)); |
3112 | VERIFY0(dsl_dataset_snap_remove(origin_head, | |
788eb90c | 3113 | ds->ds_snapname, tx, B_TRUE)); |
13fe0198 | 3114 | VERIFY0(zap_add(dp->dp_meta_objset, |
d683ddbb | 3115 | dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname, |
34dc7c2f | 3116 | 8, 1, &ds->ds_object, tx)); |
788eb90c JJ |
3117 | dsl_fs_ss_count_adjust(hds->ds_dir, 1, |
3118 | DD_FIELD_SNAPSHOT_COUNT, tx); | |
428870ff | 3119 | |
34dc7c2f BB |
3120 | /* change containing dsl_dir */ |
3121 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
d683ddbb JG |
3122 | ASSERT3U(dsl_dataset_phys(ds)->ds_dir_obj, ==, odd->dd_object); |
3123 | dsl_dataset_phys(ds)->ds_dir_obj = dd->dd_object; | |
34dc7c2f | 3124 | ASSERT3P(ds->ds_dir, ==, odd); |
13fe0198 MA |
3125 | dsl_dir_rele(ds->ds_dir, ds); |
3126 | VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object, | |
34dc7c2f BB |
3127 | NULL, ds, &ds->ds_dir)); |
3128 | ||
428870ff | 3129 | /* move any clone references */ |
d683ddbb | 3130 | if (dsl_dataset_phys(ds)->ds_next_clones_obj && |
428870ff BB |
3131 | spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { |
3132 | zap_cursor_t zc; | |
3133 | zap_attribute_t za; | |
3134 | ||
3135 | for (zap_cursor_init(&zc, dp->dp_meta_objset, | |
d683ddbb | 3136 | dsl_dataset_phys(ds)->ds_next_clones_obj); |
428870ff | 3137 | zap_cursor_retrieve(&zc, &za) == 0; |
13fe0198 MA |
3138 | zap_cursor_advance(&zc)) { |
3139 | dsl_dataset_t *cnds; | |
3140 | uint64_t o; | |
34dc7c2f | 3141 | |
13fe0198 MA |
3142 | if (za.za_first_integer == oldnext_obj) { |
3143 | /* | |
3144 | * We've already moved the | |
3145 | * origin's reference. | |
3146 | */ | |
3147 | continue; | |
3148 | } | |
34dc7c2f | 3149 | |
13fe0198 MA |
3150 | VERIFY0(dsl_dataset_hold_obj(dp, |
3151 | za.za_first_integer, FTAG, &cnds)); | |
d683ddbb JG |
3152 | o = dsl_dir_phys(cnds->ds_dir)-> |
3153 | dd_head_dataset_obj; | |
34dc7c2f | 3154 | |
13fe0198 | 3155 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3156 | dsl_dir_phys(odd)->dd_clones, o, tx)); |
13fe0198 | 3157 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3158 | dsl_dir_phys(dd)->dd_clones, o, tx)); |
13fe0198 MA |
3159 | dsl_dataset_rele(cnds, FTAG); |
3160 | } | |
3161 | zap_cursor_fini(&zc); | |
3162 | } | |
34dc7c2f | 3163 | |
13fe0198 | 3164 | ASSERT(!dsl_prop_hascb(ds)); |
34dc7c2f BB |
3165 | } |
3166 | ||
34dc7c2f | 3167 | /* |
13fe0198 MA |
3168 | * Change space accounting. |
3169 | * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either | |
3170 | * both be valid, or both be 0 (resulting in delta == 0). This | |
3171 | * is true for each of {clone,origin} independently. | |
34dc7c2f | 3172 | */ |
570827e1 | 3173 | |
13fe0198 | 3174 | delta = ddpa->cloneusedsnap - |
d683ddbb | 3175 | dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP]; |
13fe0198 MA |
3176 | ASSERT3S(delta, >=, 0); |
3177 | ASSERT3U(ddpa->used, >=, delta); | |
3178 | dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx); | |
3179 | dsl_dir_diduse_space(dd, DD_USED_HEAD, | |
3180 | ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx); | |
34dc7c2f | 3181 | |
13fe0198 | 3182 | delta = ddpa->originusedsnap - |
d683ddbb | 3183 | dsl_dir_phys(odd)->dd_used_breakdown[DD_USED_SNAP]; |
13fe0198 MA |
3184 | ASSERT3S(delta, <=, 0); |
3185 | ASSERT3U(ddpa->used, >=, -delta); | |
3186 | dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx); | |
3187 | dsl_dir_diduse_space(odd, DD_USED_HEAD, | |
3188 | -ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx); | |
3189 | ||
d683ddbb | 3190 | dsl_dataset_phys(origin_ds)->ds_unique_bytes = ddpa->unique; |
13fe0198 MA |
3191 | |
3192 | /* log history record */ | |
3193 | spa_history_log_internal_ds(hds, "promote", tx, ""); | |
3194 | ||
3195 | dsl_dir_rele(odd, FTAG); | |
3196 | promote_rele(ddpa, FTAG); | |
34dc7c2f BB |
3197 | } |
3198 | ||
13fe0198 MA |
3199 | /* |
3200 | * Make a list of dsl_dataset_t's for the snapshots between first_obj | |
3201 | * (exclusive) and last_obj (inclusive). The list will be in reverse | |
3202 | * order (last_obj will be the list_head()). If first_obj == 0, do all | |
3203 | * snapshots back to this dataset's origin. | |
3204 | */ | |
34dc7c2f | 3205 | static int |
13fe0198 MA |
3206 | snaplist_make(dsl_pool_t *dp, |
3207 | uint64_t first_obj, uint64_t last_obj, list_t *l, void *tag) | |
34dc7c2f | 3208 | { |
13fe0198 | 3209 | uint64_t obj = last_obj; |
34dc7c2f | 3210 | |
13fe0198 MA |
3211 | list_create(l, sizeof (struct promotenode), |
3212 | offsetof(struct promotenode, link)); | |
34dc7c2f | 3213 | |
13fe0198 MA |
3214 | while (obj != first_obj) { |
3215 | dsl_dataset_t *ds; | |
3216 | struct promotenode *snap; | |
3217 | int err; | |
428870ff | 3218 | |
13fe0198 MA |
3219 | err = dsl_dataset_hold_obj(dp, obj, tag, &ds); |
3220 | ASSERT(err != ENOENT); | |
3221 | if (err != 0) | |
3222 | return (err); | |
34dc7c2f | 3223 | |
13fe0198 | 3224 | if (first_obj == 0) |
d683ddbb | 3225 | first_obj = dsl_dir_phys(ds->ds_dir)->dd_origin_obj; |
13fe0198 | 3226 | |
79c76d5b | 3227 | snap = kmem_alloc(sizeof (*snap), KM_SLEEP); |
13fe0198 MA |
3228 | snap->ds = ds; |
3229 | list_insert_tail(l, snap); | |
d683ddbb | 3230 | obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
13fe0198 | 3231 | } |
34dc7c2f BB |
3232 | |
3233 | return (0); | |
3234 | } | |
3235 | ||
13fe0198 MA |
3236 | static int |
3237 | snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep) | |
34dc7c2f | 3238 | { |
13fe0198 | 3239 | struct promotenode *snap; |
6f1ffb06 | 3240 | |
13fe0198 MA |
3241 | *spacep = 0; |
3242 | for (snap = list_head(l); snap; snap = list_next(l, snap)) { | |
3243 | uint64_t used, comp, uncomp; | |
3244 | dsl_deadlist_space_range(&snap->ds->ds_deadlist, | |
3245 | mintxg, UINT64_MAX, &used, &comp, &uncomp); | |
3246 | *spacep += used; | |
428870ff | 3247 | } |
13fe0198 | 3248 | return (0); |
34dc7c2f BB |
3249 | } |
3250 | ||
13fe0198 MA |
3251 | static void |
3252 | snaplist_destroy(list_t *l, void *tag) | |
34dc7c2f | 3253 | { |
13fe0198 | 3254 | struct promotenode *snap; |
428870ff | 3255 | |
13fe0198 MA |
3256 | if (l == NULL || !list_link_active(&l->list_head)) |
3257 | return; | |
34dc7c2f | 3258 | |
13fe0198 MA |
3259 | while ((snap = list_tail(l)) != NULL) { |
3260 | list_remove(l, snap); | |
3261 | dsl_dataset_rele(snap->ds, tag); | |
3262 | kmem_free(snap, sizeof (*snap)); | |
3263 | } | |
3264 | list_destroy(l); | |
34dc7c2f BB |
3265 | } |
3266 | ||
3267 | static int | |
13fe0198 | 3268 | promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag) |
34dc7c2f | 3269 | { |
13fe0198 MA |
3270 | int error; |
3271 | dsl_dir_t *dd; | |
3272 | struct promotenode *snap; | |
34dc7c2f | 3273 | |
13fe0198 MA |
3274 | error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag, |
3275 | &ddpa->ddpa_clone); | |
3276 | if (error != 0) | |
3277 | return (error); | |
3278 | dd = ddpa->ddpa_clone->ds_dir; | |
34dc7c2f | 3279 | |
0c66c32d | 3280 | if (ddpa->ddpa_clone->ds_is_snapshot || |
13fe0198 MA |
3281 | !dsl_dir_is_clone(dd)) { |
3282 | dsl_dataset_rele(ddpa->ddpa_clone, tag); | |
2e528b49 | 3283 | return (SET_ERROR(EINVAL)); |
13fe0198 | 3284 | } |
34dc7c2f | 3285 | |
d683ddbb | 3286 | error = snaplist_make(dp, 0, dsl_dir_phys(dd)->dd_origin_obj, |
13fe0198 MA |
3287 | &ddpa->shared_snaps, tag); |
3288 | if (error != 0) | |
3289 | goto out; | |
34dc7c2f | 3290 | |
13fe0198 MA |
3291 | error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object, |
3292 | &ddpa->clone_snaps, tag); | |
3293 | if (error != 0) | |
3294 | goto out; | |
34dc7c2f | 3295 | |
13fe0198 | 3296 | snap = list_head(&ddpa->shared_snaps); |
d683ddbb JG |
3297 | ASSERT3U(snap->ds->ds_object, ==, dsl_dir_phys(dd)->dd_origin_obj); |
3298 | error = snaplist_make(dp, dsl_dir_phys(dd)->dd_origin_obj, | |
3299 | dsl_dir_phys(snap->ds->ds_dir)->dd_head_dataset_obj, | |
13fe0198 MA |
3300 | &ddpa->origin_snaps, tag); |
3301 | if (error != 0) | |
3302 | goto out; | |
d164b209 | 3303 | |
d683ddbb | 3304 | if (dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj != 0) { |
13fe0198 | 3305 | error = dsl_dataset_hold_obj(dp, |
d683ddbb | 3306 | dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj, |
13fe0198 MA |
3307 | tag, &ddpa->origin_origin); |
3308 | if (error != 0) | |
3309 | goto out; | |
d164b209 | 3310 | } |
13fe0198 MA |
3311 | out: |
3312 | if (error != 0) | |
3313 | promote_rele(ddpa, tag); | |
3314 | return (error); | |
34dc7c2f BB |
3315 | } |
3316 | ||
34dc7c2f | 3317 | static void |
13fe0198 | 3318 | promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag) |
34dc7c2f | 3319 | { |
13fe0198 MA |
3320 | snaplist_destroy(&ddpa->shared_snaps, tag); |
3321 | snaplist_destroy(&ddpa->clone_snaps, tag); | |
3322 | snaplist_destroy(&ddpa->origin_snaps, tag); | |
3323 | if (ddpa->origin_origin != NULL) | |
3324 | dsl_dataset_rele(ddpa->origin_origin, tag); | |
3325 | dsl_dataset_rele(ddpa->ddpa_clone, tag); | |
3326 | } | |
428870ff | 3327 | |
13fe0198 MA |
3328 | /* |
3329 | * Promote a clone. | |
3330 | * | |
3331 | * If it fails due to a conflicting snapshot name, "conflsnap" will be filled | |
eca7b760 | 3332 | * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.) |
13fe0198 MA |
3333 | */ |
3334 | int | |
3335 | dsl_dataset_promote(const char *name, char *conflsnap) | |
3336 | { | |
3337 | dsl_dataset_promote_arg_t ddpa = { 0 }; | |
3338 | uint64_t numsnaps; | |
3339 | int error; | |
d99a0153 | 3340 | nvpair_t *snap_pair; |
13fe0198 | 3341 | objset_t *os; |
34dc7c2f | 3342 | |
13fe0198 MA |
3343 | /* |
3344 | * We will modify space proportional to the number of | |
3345 | * snapshots. Compute numsnaps. | |
3346 | */ | |
3347 | error = dmu_objset_hold(name, FTAG, &os); | |
3348 | if (error != 0) | |
3349 | return (error); | |
3350 | error = zap_count(dmu_objset_pool(os)->dp_meta_objset, | |
d683ddbb JG |
3351 | dsl_dataset_phys(dmu_objset_ds(os))->ds_snapnames_zapobj, |
3352 | &numsnaps); | |
13fe0198 MA |
3353 | dmu_objset_rele(os, FTAG); |
3354 | if (error != 0) | |
3355 | return (error); | |
34dc7c2f | 3356 | |
13fe0198 | 3357 | ddpa.ddpa_clonename = name; |
d99a0153 | 3358 | ddpa.err_ds = fnvlist_alloc(); |
788eb90c | 3359 | ddpa.cr = CRED(); |
6f1ffb06 | 3360 | |
d99a0153 | 3361 | error = dsl_sync_task(name, dsl_dataset_promote_check, |
3d45fdd6 | 3362 | dsl_dataset_promote_sync, &ddpa, |
d99a0153 CW |
3363 | 2 + numsnaps, ZFS_SPACE_CHECK_RESERVED); |
3364 | ||
3365 | /* | |
3366 | * Return the first conflicting snapshot found. | |
3367 | */ | |
3368 | snap_pair = nvlist_next_nvpair(ddpa.err_ds, NULL); | |
3369 | if (snap_pair != NULL && conflsnap != NULL) | |
3370 | (void) strcpy(conflsnap, nvpair_name(snap_pair)); | |
3371 | ||
3372 | fnvlist_free(ddpa.err_ds); | |
3373 | return (error); | |
34dc7c2f BB |
3374 | } |
3375 | ||
3376 | int | |
13fe0198 | 3377 | dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone, |
831baf06 | 3378 | dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx) |
34dc7c2f | 3379 | { |
8c62a0d0 DM |
3380 | /* |
3381 | * "slack" factor for received datasets with refquota set on them. | |
3382 | * See the bottom of this function for details on its use. | |
3383 | */ | |
bf18fd89 CIK |
3384 | uint64_t refquota_slack = (uint64_t)DMU_MAX_ACCESS * |
3385 | spa_asize_inflation; | |
13fe0198 | 3386 | int64_t unused_refres_delta; |
34dc7c2f | 3387 | |
13fe0198 | 3388 | /* they should both be heads */ |
0c66c32d JG |
3389 | if (clone->ds_is_snapshot || |
3390 | origin_head->ds_is_snapshot) | |
2e528b49 | 3391 | return (SET_ERROR(EINVAL)); |
428870ff | 3392 | |
19580676 MA |
3393 | /* if we are not forcing, the branch point should be just before them */ |
3394 | if (!force && clone->ds_prev != origin_head->ds_prev) | |
2e528b49 | 3395 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3396 | |
13fe0198 MA |
3397 | /* clone should be the clone (unless they are unrelated) */ |
3398 | if (clone->ds_prev != NULL && | |
3399 | clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap && | |
19580676 | 3400 | origin_head->ds_dir != clone->ds_prev->ds_dir) |
2e528b49 | 3401 | return (SET_ERROR(EINVAL)); |
428870ff | 3402 | |
13fe0198 MA |
3403 | /* the clone should be a child of the origin */ |
3404 | if (clone->ds_dir->dd_parent != origin_head->ds_dir) | |
2e528b49 | 3405 | return (SET_ERROR(EINVAL)); |
45d1cae3 | 3406 | |
13fe0198 | 3407 | /* origin_head shouldn't be modified unless 'force' */ |
19580676 MA |
3408 | if (!force && |
3409 | dsl_dataset_modified_since_snap(origin_head, origin_head->ds_prev)) | |
2e528b49 | 3410 | return (SET_ERROR(ETXTBSY)); |
572e2857 | 3411 | |
13fe0198 | 3412 | /* origin_head should have no long holds (e.g. is not mounted) */ |
831baf06 | 3413 | if (dsl_dataset_handoff_check(origin_head, owner, tx)) |
2e528b49 | 3414 | return (SET_ERROR(EBUSY)); |
13fe0198 MA |
3415 | |
3416 | /* check amount of any unconsumed refreservation */ | |
3417 | unused_refres_delta = | |
3418 | (int64_t)MIN(origin_head->ds_reserved, | |
d683ddbb | 3419 | dsl_dataset_phys(origin_head)->ds_unique_bytes) - |
13fe0198 | 3420 | (int64_t)MIN(origin_head->ds_reserved, |
d683ddbb | 3421 | dsl_dataset_phys(clone)->ds_unique_bytes); |
13fe0198 MA |
3422 | |
3423 | if (unused_refres_delta > 0 && | |
3424 | unused_refres_delta > | |
3425 | dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE)) | |
2e528b49 | 3426 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3427 | |
8c62a0d0 DM |
3428 | /* |
3429 | * The clone can't be too much over the head's refquota. | |
3430 | * | |
3431 | * To ensure that the entire refquota can be used, we allow one | |
3432 | * transaction to exceed the the refquota. Therefore, this check | |
3433 | * needs to also allow for the space referenced to be more than the | |
3434 | * refquota. The maximum amount of space that one transaction can use | |
3435 | * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this | |
3436 | * overage ensures that we are able to receive a filesystem that | |
3437 | * exceeds the refquota on the source system. | |
3438 | * | |
3439 | * So that overage is the refquota_slack we use below. | |
3440 | */ | |
13fe0198 | 3441 | if (origin_head->ds_quota != 0 && |
d683ddbb | 3442 | dsl_dataset_phys(clone)->ds_referenced_bytes > |
8c62a0d0 | 3443 | origin_head->ds_quota + refquota_slack) |
2e528b49 | 3444 | return (SET_ERROR(EDQUOT)); |
572e2857 | 3445 | |
13fe0198 | 3446 | return (0); |
572e2857 BB |
3447 | } |
3448 | ||
a1d477c2 MA |
3449 | static void |
3450 | dsl_dataset_swap_remap_deadlists(dsl_dataset_t *clone, | |
3451 | dsl_dataset_t *origin, dmu_tx_t *tx) | |
3452 | { | |
3453 | uint64_t clone_remap_dl_obj, origin_remap_dl_obj; | |
3454 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3455 | ||
3456 | ASSERT(dsl_pool_sync_context(dp)); | |
3457 | ||
3458 | clone_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(clone); | |
3459 | origin_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(origin); | |
3460 | ||
3461 | if (clone_remap_dl_obj != 0) { | |
3462 | dsl_deadlist_close(&clone->ds_remap_deadlist); | |
3463 | dsl_dataset_unset_remap_deadlist_object(clone, tx); | |
3464 | } | |
3465 | if (origin_remap_dl_obj != 0) { | |
3466 | dsl_deadlist_close(&origin->ds_remap_deadlist); | |
3467 | dsl_dataset_unset_remap_deadlist_object(origin, tx); | |
3468 | } | |
3469 | ||
3470 | if (clone_remap_dl_obj != 0) { | |
3471 | dsl_dataset_set_remap_deadlist_object(origin, | |
3472 | clone_remap_dl_obj, tx); | |
3473 | dsl_deadlist_open(&origin->ds_remap_deadlist, | |
3474 | dp->dp_meta_objset, clone_remap_dl_obj); | |
3475 | } | |
3476 | if (origin_remap_dl_obj != 0) { | |
3477 | dsl_dataset_set_remap_deadlist_object(clone, | |
3478 | origin_remap_dl_obj, tx); | |
3479 | dsl_deadlist_open(&clone->ds_remap_deadlist, | |
3480 | dp->dp_meta_objset, origin_remap_dl_obj); | |
3481 | } | |
3482 | } | |
3483 | ||
572e2857 | 3484 | void |
13fe0198 MA |
3485 | dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone, |
3486 | dsl_dataset_t *origin_head, dmu_tx_t *tx) | |
572e2857 | 3487 | { |
13fe0198 MA |
3488 | dsl_pool_t *dp = dmu_tx_pool(tx); |
3489 | int64_t unused_refres_delta; | |
428870ff | 3490 | |
13fe0198 | 3491 | ASSERT(clone->ds_reserved == 0); |
8c62a0d0 DM |
3492 | /* |
3493 | * NOTE: On DEBUG kernels there could be a race between this and | |
3494 | * the check function if spa_asize_inflation is adjusted... | |
3495 | */ | |
13fe0198 | 3496 | ASSERT(origin_head->ds_quota == 0 || |
8c62a0d0 DM |
3497 | dsl_dataset_phys(clone)->ds_unique_bytes <= origin_head->ds_quota + |
3498 | DMU_MAX_ACCESS * spa_asize_inflation); | |
19580676 | 3499 | ASSERT3P(clone->ds_prev, ==, origin_head->ds_prev); |
45d1cae3 | 3500 | |
241b5415 MA |
3501 | /* |
3502 | * Swap per-dataset feature flags. | |
3503 | */ | |
4a5d7f82 | 3504 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
3505 | if (!(spa_feature_table[f].fi_flags & |
3506 | ZFEATURE_FLAG_PER_DATASET)) { | |
3507 | ASSERT(!clone->ds_feature_inuse[f]); | |
3508 | ASSERT(!origin_head->ds_feature_inuse[f]); | |
3509 | continue; | |
3510 | } | |
3511 | ||
1c27024e DB |
3512 | boolean_t clone_inuse = clone->ds_feature_inuse[f]; |
3513 | boolean_t origin_head_inuse = origin_head->ds_feature_inuse[f]; | |
241b5415 MA |
3514 | |
3515 | if (clone_inuse) { | |
3516 | dsl_dataset_deactivate_feature(clone->ds_object, f, tx); | |
3517 | clone->ds_feature_inuse[f] = B_FALSE; | |
3518 | } | |
3519 | if (origin_head_inuse) { | |
3520 | dsl_dataset_deactivate_feature(origin_head->ds_object, | |
3521 | f, tx); | |
3522 | origin_head->ds_feature_inuse[f] = B_FALSE; | |
3523 | } | |
3524 | if (clone_inuse) { | |
3525 | dsl_dataset_activate_feature(origin_head->ds_object, | |
3526 | f, tx); | |
3527 | origin_head->ds_feature_inuse[f] = B_TRUE; | |
3528 | } | |
3529 | if (origin_head_inuse) { | |
3530 | dsl_dataset_activate_feature(clone->ds_object, f, tx); | |
3531 | clone->ds_feature_inuse[f] = B_TRUE; | |
3532 | } | |
3533 | } | |
3534 | ||
13fe0198 MA |
3535 | dmu_buf_will_dirty(clone->ds_dbuf, tx); |
3536 | dmu_buf_will_dirty(origin_head->ds_dbuf, tx); | |
45d1cae3 | 3537 | |
13fe0198 MA |
3538 | if (clone->ds_objset != NULL) { |
3539 | dmu_objset_evict(clone->ds_objset); | |
3540 | clone->ds_objset = NULL; | |
3541 | } | |
45d1cae3 | 3542 | |
13fe0198 MA |
3543 | if (origin_head->ds_objset != NULL) { |
3544 | dmu_objset_evict(origin_head->ds_objset); | |
3545 | origin_head->ds_objset = NULL; | |
45d1cae3 | 3546 | } |
45d1cae3 | 3547 | |
13fe0198 MA |
3548 | unused_refres_delta = |
3549 | (int64_t)MIN(origin_head->ds_reserved, | |
d683ddbb | 3550 | dsl_dataset_phys(origin_head)->ds_unique_bytes) - |
13fe0198 | 3551 | (int64_t)MIN(origin_head->ds_reserved, |
d683ddbb | 3552 | dsl_dataset_phys(clone)->ds_unique_bytes); |
13fe0198 MA |
3553 | |
3554 | /* | |
3555 | * Reset origin's unique bytes, if it exists. | |
3556 | */ | |
3557 | if (clone->ds_prev) { | |
3558 | dsl_dataset_t *origin = clone->ds_prev; | |
3559 | uint64_t comp, uncomp; | |
3560 | ||
3561 | dmu_buf_will_dirty(origin->ds_dbuf, tx); | |
3562 | dsl_deadlist_space_range(&clone->ds_deadlist, | |
d683ddbb JG |
3563 | dsl_dataset_phys(origin)->ds_prev_snap_txg, UINT64_MAX, |
3564 | &dsl_dataset_phys(origin)->ds_unique_bytes, &comp, &uncomp); | |
13fe0198 MA |
3565 | } |
3566 | ||
3567 | /* swap blkptrs */ | |
3568 | { | |
cc9bb3e5 GM |
3569 | rrw_enter(&clone->ds_bp_rwlock, RW_WRITER, FTAG); |
3570 | rrw_enter(&origin_head->ds_bp_rwlock, RW_WRITER, FTAG); | |
1c27024e | 3571 | blkptr_t tmp; |
d683ddbb JG |
3572 | tmp = dsl_dataset_phys(origin_head)->ds_bp; |
3573 | dsl_dataset_phys(origin_head)->ds_bp = | |
3574 | dsl_dataset_phys(clone)->ds_bp; | |
3575 | dsl_dataset_phys(clone)->ds_bp = tmp; | |
cc9bb3e5 GM |
3576 | rrw_exit(&origin_head->ds_bp_rwlock, FTAG); |
3577 | rrw_exit(&clone->ds_bp_rwlock, FTAG); | |
13fe0198 MA |
3578 | } |
3579 | ||
3580 | /* set dd_*_bytes */ | |
3581 | { | |
3582 | int64_t dused, dcomp, duncomp; | |
3583 | uint64_t cdl_used, cdl_comp, cdl_uncomp; | |
3584 | uint64_t odl_used, odl_comp, odl_uncomp; | |
3585 | ||
d683ddbb | 3586 | ASSERT3U(dsl_dir_phys(clone->ds_dir)-> |
13fe0198 MA |
3587 | dd_used_breakdown[DD_USED_SNAP], ==, 0); |
3588 | ||
3589 | dsl_deadlist_space(&clone->ds_deadlist, | |
3590 | &cdl_used, &cdl_comp, &cdl_uncomp); | |
3591 | dsl_deadlist_space(&origin_head->ds_deadlist, | |
3592 | &odl_used, &odl_comp, &odl_uncomp); | |
428870ff | 3593 | |
d683ddbb JG |
3594 | dused = dsl_dataset_phys(clone)->ds_referenced_bytes + |
3595 | cdl_used - | |
3596 | (dsl_dataset_phys(origin_head)->ds_referenced_bytes + | |
3597 | odl_used); | |
3598 | dcomp = dsl_dataset_phys(clone)->ds_compressed_bytes + | |
3599 | cdl_comp - | |
3600 | (dsl_dataset_phys(origin_head)->ds_compressed_bytes + | |
3601 | odl_comp); | |
3602 | duncomp = dsl_dataset_phys(clone)->ds_uncompressed_bytes + | |
13fe0198 | 3603 | cdl_uncomp - |
d683ddbb JG |
3604 | (dsl_dataset_phys(origin_head)->ds_uncompressed_bytes + |
3605 | odl_uncomp); | |
45d1cae3 | 3606 | |
13fe0198 MA |
3607 | dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD, |
3608 | dused, dcomp, duncomp, tx); | |
3609 | dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD, | |
3610 | -dused, -dcomp, -duncomp, tx); | |
45d1cae3 | 3611 | |
45d1cae3 | 3612 | /* |
13fe0198 MA |
3613 | * The difference in the space used by snapshots is the |
3614 | * difference in snapshot space due to the head's | |
3615 | * deadlist (since that's the only thing that's | |
3616 | * changing that affects the snapused). | |
45d1cae3 | 3617 | */ |
13fe0198 MA |
3618 | dsl_deadlist_space_range(&clone->ds_deadlist, |
3619 | origin_head->ds_dir->dd_origin_txg, UINT64_MAX, | |
3620 | &cdl_used, &cdl_comp, &cdl_uncomp); | |
3621 | dsl_deadlist_space_range(&origin_head->ds_deadlist, | |
3622 | origin_head->ds_dir->dd_origin_txg, UINT64_MAX, | |
3623 | &odl_used, &odl_comp, &odl_uncomp); | |
3624 | dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used, | |
3625 | DD_USED_HEAD, DD_USED_SNAP, tx); | |
45d1cae3 | 3626 | } |
45d1cae3 | 3627 | |
13fe0198 | 3628 | /* swap ds_*_bytes */ |
d683ddbb JG |
3629 | SWITCH64(dsl_dataset_phys(origin_head)->ds_referenced_bytes, |
3630 | dsl_dataset_phys(clone)->ds_referenced_bytes); | |
3631 | SWITCH64(dsl_dataset_phys(origin_head)->ds_compressed_bytes, | |
3632 | dsl_dataset_phys(clone)->ds_compressed_bytes); | |
3633 | SWITCH64(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes, | |
3634 | dsl_dataset_phys(clone)->ds_uncompressed_bytes); | |
3635 | SWITCH64(dsl_dataset_phys(origin_head)->ds_unique_bytes, | |
3636 | dsl_dataset_phys(clone)->ds_unique_bytes); | |
45d1cae3 | 3637 | |
13fe0198 MA |
3638 | /* apply any parent delta for change in unconsumed refreservation */ |
3639 | dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV, | |
3640 | unused_refres_delta, 0, 0, tx); | |
45d1cae3 | 3641 | |
13fe0198 MA |
3642 | /* |
3643 | * Swap deadlists. | |
3644 | */ | |
3645 | dsl_deadlist_close(&clone->ds_deadlist); | |
3646 | dsl_deadlist_close(&origin_head->ds_deadlist); | |
d683ddbb JG |
3647 | SWITCH64(dsl_dataset_phys(origin_head)->ds_deadlist_obj, |
3648 | dsl_dataset_phys(clone)->ds_deadlist_obj); | |
13fe0198 | 3649 | dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset, |
d683ddbb | 3650 | dsl_dataset_phys(clone)->ds_deadlist_obj); |
13fe0198 | 3651 | dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset, |
d683ddbb | 3652 | dsl_dataset_phys(origin_head)->ds_deadlist_obj); |
a1d477c2 | 3653 | dsl_dataset_swap_remap_deadlists(clone, origin_head, tx); |
45d1cae3 | 3654 | |
13fe0198 | 3655 | dsl_scan_ds_clone_swapped(origin_head, clone, tx); |
45d1cae3 | 3656 | |
13fe0198 MA |
3657 | spa_history_log_internal_ds(clone, "clone swap", tx, |
3658 | "parent=%s", origin_head->ds_dir->dd_myname); | |
45d1cae3 BB |
3659 | } |
3660 | ||
13fe0198 MA |
3661 | /* |
3662 | * Given a pool name and a dataset object number in that pool, | |
3663 | * return the name of that dataset. | |
3664 | */ | |
572e2857 | 3665 | int |
13fe0198 | 3666 | dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf) |
572e2857 | 3667 | { |
13fe0198 MA |
3668 | dsl_pool_t *dp; |
3669 | dsl_dataset_t *ds; | |
572e2857 BB |
3670 | int error; |
3671 | ||
13fe0198 MA |
3672 | error = dsl_pool_hold(pname, FTAG, &dp); |
3673 | if (error != 0) | |
3674 | return (error); | |
3675 | ||
3676 | error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds); | |
3677 | if (error == 0) { | |
3678 | dsl_dataset_name(ds, buf); | |
3679 | dsl_dataset_rele(ds, FTAG); | |
3680 | } | |
3681 | dsl_pool_rele(dp, FTAG); | |
572e2857 BB |
3682 | |
3683 | return (error); | |
3684 | } | |
3685 | ||
45d1cae3 | 3686 | int |
13fe0198 MA |
3687 | dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota, |
3688 | uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv) | |
45d1cae3 | 3689 | { |
13fe0198 | 3690 | int error = 0; |
45d1cae3 | 3691 | |
13fe0198 | 3692 | ASSERT3S(asize, >, 0); |
45d1cae3 | 3693 | |
13fe0198 MA |
3694 | /* |
3695 | * *ref_rsrv is the portion of asize that will come from any | |
3696 | * unconsumed refreservation space. | |
3697 | */ | |
3698 | *ref_rsrv = 0; | |
45d1cae3 | 3699 | |
13fe0198 MA |
3700 | mutex_enter(&ds->ds_lock); |
3701 | /* | |
3702 | * Make a space adjustment for reserved bytes. | |
3703 | */ | |
d683ddbb | 3704 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) { |
13fe0198 | 3705 | ASSERT3U(*used, >=, |
d683ddbb JG |
3706 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); |
3707 | *used -= | |
3708 | (ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); | |
13fe0198 MA |
3709 | *ref_rsrv = |
3710 | asize - MIN(asize, parent_delta(ds, asize + inflight)); | |
45d1cae3 BB |
3711 | } |
3712 | ||
13fe0198 MA |
3713 | if (!check_quota || ds->ds_quota == 0) { |
3714 | mutex_exit(&ds->ds_lock); | |
3715 | return (0); | |
45d1cae3 | 3716 | } |
13fe0198 MA |
3717 | /* |
3718 | * If they are requesting more space, and our current estimate | |
3719 | * is over quota, they get to try again unless the actual | |
3720 | * on-disk is over quota and there are no pending changes (which | |
3721 | * may free up space for us). | |
3722 | */ | |
d683ddbb JG |
3723 | if (dsl_dataset_phys(ds)->ds_referenced_bytes + inflight >= |
3724 | ds->ds_quota) { | |
13fe0198 | 3725 | if (inflight > 0 || |
d683ddbb | 3726 | dsl_dataset_phys(ds)->ds_referenced_bytes < ds->ds_quota) |
2e528b49 | 3727 | error = SET_ERROR(ERESTART); |
13fe0198 | 3728 | else |
2e528b49 | 3729 | error = SET_ERROR(EDQUOT); |
45d1cae3 | 3730 | } |
13fe0198 | 3731 | mutex_exit(&ds->ds_lock); |
45d1cae3 | 3732 | |
45d1cae3 BB |
3733 | return (error); |
3734 | } | |
3735 | ||
13fe0198 MA |
3736 | typedef struct dsl_dataset_set_qr_arg { |
3737 | const char *ddsqra_name; | |
3738 | zprop_source_t ddsqra_source; | |
3739 | uint64_t ddsqra_value; | |
3740 | } dsl_dataset_set_qr_arg_t; | |
45d1cae3 | 3741 | |
13fe0198 MA |
3742 | |
3743 | /* ARGSUSED */ | |
45d1cae3 | 3744 | static int |
13fe0198 | 3745 | dsl_dataset_set_refquota_check(void *arg, dmu_tx_t *tx) |
45d1cae3 | 3746 | { |
13fe0198 MA |
3747 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3748 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3749 | dsl_dataset_t *ds; | |
45d1cae3 | 3750 | int error; |
13fe0198 | 3751 | uint64_t newval; |
45d1cae3 | 3752 | |
13fe0198 | 3753 | if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA) |
2e528b49 | 3754 | return (SET_ERROR(ENOTSUP)); |
45d1cae3 | 3755 | |
13fe0198 MA |
3756 | error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); |
3757 | if (error != 0) | |
3758 | return (error); | |
3759 | ||
0c66c32d | 3760 | if (ds->ds_is_snapshot) { |
13fe0198 | 3761 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 3762 | return (SET_ERROR(EINVAL)); |
45d1cae3 BB |
3763 | } |
3764 | ||
13fe0198 MA |
3765 | error = dsl_prop_predict(ds->ds_dir, |
3766 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
3767 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); | |
3768 | if (error != 0) { | |
3769 | dsl_dataset_rele(ds, FTAG); | |
45d1cae3 BB |
3770 | return (error); |
3771 | } | |
3772 | ||
13fe0198 MA |
3773 | if (newval == 0) { |
3774 | dsl_dataset_rele(ds, FTAG); | |
3775 | return (0); | |
3776 | } | |
3777 | ||
d683ddbb | 3778 | if (newval < dsl_dataset_phys(ds)->ds_referenced_bytes || |
13fe0198 MA |
3779 | newval < ds->ds_reserved) { |
3780 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 3781 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3782 | } |
45d1cae3 | 3783 | |
13fe0198 | 3784 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 BB |
3785 | return (0); |
3786 | } | |
3787 | ||
13fe0198 MA |
3788 | static void |
3789 | dsl_dataset_set_refquota_sync(void *arg, dmu_tx_t *tx) | |
45d1cae3 | 3790 | { |
13fe0198 MA |
3791 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3792 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 3793 | dsl_dataset_t *ds = NULL; |
13fe0198 | 3794 | uint64_t newval; |
45d1cae3 | 3795 | |
13fe0198 | 3796 | VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); |
45d1cae3 | 3797 | |
13fe0198 MA |
3798 | dsl_prop_set_sync_impl(ds, |
3799 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
3800 | ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1, | |
3801 | &ddsqra->ddsqra_value, tx); | |
45d1cae3 | 3802 | |
13fe0198 MA |
3803 | VERIFY0(dsl_prop_get_int_ds(ds, |
3804 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval)); | |
45d1cae3 | 3805 | |
13fe0198 MA |
3806 | if (ds->ds_quota != newval) { |
3807 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
3808 | ds->ds_quota = newval; | |
45d1cae3 | 3809 | } |
13fe0198 | 3810 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 BB |
3811 | } |
3812 | ||
13fe0198 MA |
3813 | int |
3814 | dsl_dataset_set_refquota(const char *dsname, zprop_source_t source, | |
3815 | uint64_t refquota) | |
45d1cae3 | 3816 | { |
13fe0198 | 3817 | dsl_dataset_set_qr_arg_t ddsqra; |
428870ff | 3818 | |
13fe0198 MA |
3819 | ddsqra.ddsqra_name = dsname; |
3820 | ddsqra.ddsqra_source = source; | |
3821 | ddsqra.ddsqra_value = refquota; | |
3822 | ||
3823 | return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check, | |
3d45fdd6 | 3824 | dsl_dataset_set_refquota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE)); |
45d1cae3 BB |
3825 | } |
3826 | ||
3827 | static int | |
13fe0198 | 3828 | dsl_dataset_set_refreservation_check(void *arg, dmu_tx_t *tx) |
45d1cae3 | 3829 | { |
13fe0198 MA |
3830 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3831 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
45d1cae3 BB |
3832 | dsl_dataset_t *ds; |
3833 | int error; | |
13fe0198 | 3834 | uint64_t newval, unique; |
428870ff | 3835 | |
13fe0198 | 3836 | if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION) |
2e528b49 | 3837 | return (SET_ERROR(ENOTSUP)); |
45d1cae3 | 3838 | |
13fe0198 MA |
3839 | error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); |
3840 | if (error != 0) | |
45d1cae3 | 3841 | return (error); |
45d1cae3 | 3842 | |
0c66c32d | 3843 | if (ds->ds_is_snapshot) { |
13fe0198 | 3844 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 3845 | return (SET_ERROR(EINVAL)); |
45d1cae3 BB |
3846 | } |
3847 | ||
13fe0198 MA |
3848 | error = dsl_prop_predict(ds->ds_dir, |
3849 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), | |
3850 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); | |
3851 | if (error != 0) { | |
3852 | dsl_dataset_rele(ds, FTAG); | |
45d1cae3 BB |
3853 | return (error); |
3854 | } | |
3855 | ||
13fe0198 MA |
3856 | /* |
3857 | * If we are doing the preliminary check in open context, the | |
3858 | * space estimates may be inaccurate. | |
3859 | */ | |
3860 | if (!dmu_tx_is_syncing(tx)) { | |
3861 | dsl_dataset_rele(ds, FTAG); | |
3862 | return (0); | |
45d1cae3 | 3863 | } |
45d1cae3 | 3864 | |
13fe0198 MA |
3865 | mutex_enter(&ds->ds_lock); |
3866 | if (!DS_UNIQUE_IS_ACCURATE(ds)) | |
3867 | dsl_dataset_recalc_head_uniq(ds); | |
d683ddbb | 3868 | unique = dsl_dataset_phys(ds)->ds_unique_bytes; |
13fe0198 | 3869 | mutex_exit(&ds->ds_lock); |
45d1cae3 | 3870 | |
13fe0198 MA |
3871 | if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) { |
3872 | uint64_t delta = MAX(unique, newval) - | |
3873 | MAX(unique, ds->ds_reserved); | |
45d1cae3 | 3874 | |
13fe0198 MA |
3875 | if (delta > |
3876 | dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) || | |
3877 | (ds->ds_quota > 0 && newval > ds->ds_quota)) { | |
3878 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 3879 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3880 | } |
45d1cae3 BB |
3881 | } |
3882 | ||
13fe0198 MA |
3883 | dsl_dataset_rele(ds, FTAG); |
3884 | return (0); | |
45d1cae3 BB |
3885 | } |
3886 | ||
13fe0198 MA |
3887 | void |
3888 | dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds, | |
3889 | zprop_source_t source, uint64_t value, dmu_tx_t *tx) | |
428870ff | 3890 | { |
13fe0198 MA |
3891 | uint64_t newval; |
3892 | uint64_t unique; | |
3893 | int64_t delta; | |
428870ff | 3894 | |
13fe0198 MA |
3895 | dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), |
3896 | source, sizeof (value), 1, &value, tx); | |
428870ff | 3897 | |
13fe0198 MA |
3898 | VERIFY0(dsl_prop_get_int_ds(ds, |
3899 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval)); | |
572e2857 | 3900 | |
13fe0198 MA |
3901 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
3902 | mutex_enter(&ds->ds_dir->dd_lock); | |
3903 | mutex_enter(&ds->ds_lock); | |
3904 | ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); | |
d683ddbb | 3905 | unique = dsl_dataset_phys(ds)->ds_unique_bytes; |
13fe0198 MA |
3906 | delta = MAX(0, (int64_t)(newval - unique)) - |
3907 | MAX(0, (int64_t)(ds->ds_reserved - unique)); | |
3908 | ds->ds_reserved = newval; | |
3909 | mutex_exit(&ds->ds_lock); | |
572e2857 | 3910 | |
13fe0198 MA |
3911 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx); |
3912 | mutex_exit(&ds->ds_dir->dd_lock); | |
428870ff BB |
3913 | } |
3914 | ||
13fe0198 MA |
3915 | static void |
3916 | dsl_dataset_set_refreservation_sync(void *arg, dmu_tx_t *tx) | |
45d1cae3 | 3917 | { |
13fe0198 MA |
3918 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3919 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 3920 | dsl_dataset_t *ds = NULL; |
45d1cae3 | 3921 | |
13fe0198 MA |
3922 | VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); |
3923 | dsl_dataset_set_refreservation_sync_impl(ds, | |
3924 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx); | |
45d1cae3 | 3925 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 | 3926 | } |
428870ff | 3927 | |
428870ff | 3928 | int |
13fe0198 MA |
3929 | dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source, |
3930 | uint64_t refreservation) | |
428870ff | 3931 | { |
13fe0198 | 3932 | dsl_dataset_set_qr_arg_t ddsqra; |
428870ff | 3933 | |
13fe0198 MA |
3934 | ddsqra.ddsqra_name = dsname; |
3935 | ddsqra.ddsqra_source = source; | |
3936 | ddsqra.ddsqra_value = refreservation; | |
c28b2279 | 3937 | |
13fe0198 | 3938 | return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check, |
3d45fdd6 MA |
3939 | dsl_dataset_set_refreservation_sync, &ddsqra, |
3940 | 0, ZFS_SPACE_CHECK_NONE)); | |
13fe0198 | 3941 | } |
330d06f9 MA |
3942 | |
3943 | /* | |
3944 | * Return (in *usedp) the amount of space written in new that is not | |
3945 | * present in oldsnap. New may be a snapshot or the head. Old must be | |
3946 | * a snapshot before new, in new's filesystem (or its origin). If not then | |
3947 | * fail and return EINVAL. | |
3948 | * | |
3949 | * The written space is calculated by considering two components: First, we | |
3950 | * ignore any freed space, and calculate the written as new's used space | |
3951 | * minus old's used space. Next, we add in the amount of space that was freed | |
3952 | * between the two snapshots, thus reducing new's used space relative to old's. | |
3953 | * Specifically, this is the space that was born before old->ds_creation_txg, | |
3954 | * and freed before new (ie. on new's deadlist or a previous deadlist). | |
3955 | * | |
3956 | * space freed [---------------------] | |
3957 | * snapshots ---O-------O--------O-------O------ | |
3958 | * oldsnap new | |
3959 | */ | |
3960 | int | |
3961 | dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new, | |
3962 | uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) | |
3963 | { | |
3964 | int err = 0; | |
3965 | uint64_t snapobj; | |
3966 | dsl_pool_t *dp = new->ds_dir->dd_pool; | |
3967 | ||
13fe0198 MA |
3968 | ASSERT(dsl_pool_config_held(dp)); |
3969 | ||
330d06f9 | 3970 | *usedp = 0; |
d683ddbb JG |
3971 | *usedp += dsl_dataset_phys(new)->ds_referenced_bytes; |
3972 | *usedp -= dsl_dataset_phys(oldsnap)->ds_referenced_bytes; | |
330d06f9 MA |
3973 | |
3974 | *compp = 0; | |
d683ddbb JG |
3975 | *compp += dsl_dataset_phys(new)->ds_compressed_bytes; |
3976 | *compp -= dsl_dataset_phys(oldsnap)->ds_compressed_bytes; | |
330d06f9 MA |
3977 | |
3978 | *uncompp = 0; | |
d683ddbb JG |
3979 | *uncompp += dsl_dataset_phys(new)->ds_uncompressed_bytes; |
3980 | *uncompp -= dsl_dataset_phys(oldsnap)->ds_uncompressed_bytes; | |
330d06f9 | 3981 | |
330d06f9 MA |
3982 | snapobj = new->ds_object; |
3983 | while (snapobj != oldsnap->ds_object) { | |
3984 | dsl_dataset_t *snap; | |
3985 | uint64_t used, comp, uncomp; | |
3986 | ||
9ae529ec CS |
3987 | if (snapobj == new->ds_object) { |
3988 | snap = new; | |
3989 | } else { | |
3990 | err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap); | |
3991 | if (err != 0) | |
3992 | break; | |
3993 | } | |
330d06f9 | 3994 | |
d683ddbb JG |
3995 | if (dsl_dataset_phys(snap)->ds_prev_snap_txg == |
3996 | dsl_dataset_phys(oldsnap)->ds_creation_txg) { | |
330d06f9 MA |
3997 | /* |
3998 | * The blocks in the deadlist can not be born after | |
3999 | * ds_prev_snap_txg, so get the whole deadlist space, | |
4000 | * which is more efficient (especially for old-format | |
4001 | * deadlists). Unfortunately the deadlist code | |
4002 | * doesn't have enough information to make this | |
4003 | * optimization itself. | |
4004 | */ | |
4005 | dsl_deadlist_space(&snap->ds_deadlist, | |
4006 | &used, &comp, &uncomp); | |
4007 | } else { | |
4008 | dsl_deadlist_space_range(&snap->ds_deadlist, | |
d683ddbb | 4009 | 0, dsl_dataset_phys(oldsnap)->ds_creation_txg, |
330d06f9 MA |
4010 | &used, &comp, &uncomp); |
4011 | } | |
4012 | *usedp += used; | |
4013 | *compp += comp; | |
4014 | *uncompp += uncomp; | |
4015 | ||
4016 | /* | |
4017 | * If we get to the beginning of the chain of snapshots | |
4018 | * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap | |
4019 | * was not a snapshot of/before new. | |
4020 | */ | |
d683ddbb | 4021 | snapobj = dsl_dataset_phys(snap)->ds_prev_snap_obj; |
9ae529ec CS |
4022 | if (snap != new) |
4023 | dsl_dataset_rele(snap, FTAG); | |
330d06f9 | 4024 | if (snapobj == 0) { |
2e528b49 | 4025 | err = SET_ERROR(EINVAL); |
330d06f9 MA |
4026 | break; |
4027 | } | |
4028 | ||
4029 | } | |
330d06f9 MA |
4030 | return (err); |
4031 | } | |
4032 | ||
4033 | /* | |
4034 | * Return (in *usedp) the amount of space that will be reclaimed if firstsnap, | |
4035 | * lastsnap, and all snapshots in between are deleted. | |
4036 | * | |
4037 | * blocks that would be freed [---------------------------] | |
4038 | * snapshots ---O-------O--------O-------O--------O | |
4039 | * firstsnap lastsnap | |
4040 | * | |
4041 | * This is the set of blocks that were born after the snap before firstsnap, | |
4042 | * (birth > firstsnap->prev_snap_txg) and died before the snap after the | |
4043 | * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist). | |
4044 | * We calculate this by iterating over the relevant deadlists (from the snap | |
4045 | * after lastsnap, backward to the snap after firstsnap), summing up the | |
4046 | * space on the deadlist that was born after the snap before firstsnap. | |
4047 | */ | |
4048 | int | |
4049 | dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, | |
4050 | dsl_dataset_t *lastsnap, | |
4051 | uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) | |
4052 | { | |
4053 | int err = 0; | |
4054 | uint64_t snapobj; | |
4055 | dsl_pool_t *dp = firstsnap->ds_dir->dd_pool; | |
4056 | ||
0c66c32d JG |
4057 | ASSERT(firstsnap->ds_is_snapshot); |
4058 | ASSERT(lastsnap->ds_is_snapshot); | |
330d06f9 MA |
4059 | |
4060 | /* | |
4061 | * Check that the snapshots are in the same dsl_dir, and firstsnap | |
4062 | * is before lastsnap. | |
4063 | */ | |
4064 | if (firstsnap->ds_dir != lastsnap->ds_dir || | |
d683ddbb JG |
4065 | dsl_dataset_phys(firstsnap)->ds_creation_txg > |
4066 | dsl_dataset_phys(lastsnap)->ds_creation_txg) | |
2e528b49 | 4067 | return (SET_ERROR(EINVAL)); |
330d06f9 MA |
4068 | |
4069 | *usedp = *compp = *uncompp = 0; | |
4070 | ||
d683ddbb | 4071 | snapobj = dsl_dataset_phys(lastsnap)->ds_next_snap_obj; |
330d06f9 MA |
4072 | while (snapobj != firstsnap->ds_object) { |
4073 | dsl_dataset_t *ds; | |
4074 | uint64_t used, comp, uncomp; | |
4075 | ||
4076 | err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds); | |
4077 | if (err != 0) | |
4078 | break; | |
4079 | ||
4080 | dsl_deadlist_space_range(&ds->ds_deadlist, | |
d683ddbb | 4081 | dsl_dataset_phys(firstsnap)->ds_prev_snap_txg, UINT64_MAX, |
330d06f9 MA |
4082 | &used, &comp, &uncomp); |
4083 | *usedp += used; | |
4084 | *compp += comp; | |
4085 | *uncompp += uncomp; | |
4086 | ||
d683ddbb | 4087 | snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
330d06f9 MA |
4088 | ASSERT3U(snapobj, !=, 0); |
4089 | dsl_dataset_rele(ds, FTAG); | |
4090 | } | |
330d06f9 MA |
4091 | return (err); |
4092 | } | |
4093 | ||
13fe0198 MA |
4094 | /* |
4095 | * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline. | |
4096 | * For example, they could both be snapshots of the same filesystem, and | |
4097 | * 'earlier' is before 'later'. Or 'earlier' could be the origin of | |
4098 | * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's | |
4099 | * filesystem. Or 'earlier' could be the origin's origin. | |
da536844 MA |
4100 | * |
4101 | * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg. | |
13fe0198 MA |
4102 | */ |
4103 | boolean_t | |
da536844 | 4104 | dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier, |
e9aa730c | 4105 | uint64_t earlier_txg) |
13fe0198 MA |
4106 | { |
4107 | dsl_pool_t *dp = later->ds_dir->dd_pool; | |
4108 | int error; | |
4109 | boolean_t ret; | |
13fe0198 MA |
4110 | |
4111 | ASSERT(dsl_pool_config_held(dp)); | |
0c66c32d | 4112 | ASSERT(earlier->ds_is_snapshot || earlier_txg != 0); |
da536844 MA |
4113 | |
4114 | if (earlier_txg == 0) | |
d683ddbb | 4115 | earlier_txg = dsl_dataset_phys(earlier)->ds_creation_txg; |
13fe0198 | 4116 | |
0c66c32d | 4117 | if (later->ds_is_snapshot && |
d683ddbb | 4118 | earlier_txg >= dsl_dataset_phys(later)->ds_creation_txg) |
13fe0198 MA |
4119 | return (B_FALSE); |
4120 | ||
4121 | if (later->ds_dir == earlier->ds_dir) | |
4122 | return (B_TRUE); | |
4123 | if (!dsl_dir_is_clone(later->ds_dir)) | |
4124 | return (B_FALSE); | |
4125 | ||
d683ddbb | 4126 | if (dsl_dir_phys(later->ds_dir)->dd_origin_obj == earlier->ds_object) |
13fe0198 | 4127 | return (B_TRUE); |
1c27024e | 4128 | dsl_dataset_t *origin; |
13fe0198 | 4129 | error = dsl_dataset_hold_obj(dp, |
d683ddbb | 4130 | dsl_dir_phys(later->ds_dir)->dd_origin_obj, FTAG, &origin); |
13fe0198 MA |
4131 | if (error != 0) |
4132 | return (B_FALSE); | |
da536844 | 4133 | ret = dsl_dataset_is_before(origin, earlier, earlier_txg); |
13fe0198 MA |
4134 | dsl_dataset_rele(origin, FTAG); |
4135 | return (ret); | |
4136 | } | |
4137 | ||
fa86b5db MA |
4138 | void |
4139 | dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4140 | { | |
4141 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; | |
4142 | dmu_object_zapify(mos, ds->ds_object, DMU_OT_DSL_DATASET, tx); | |
4143 | } | |
4144 | ||
47dfff3b MA |
4145 | boolean_t |
4146 | dsl_dataset_is_zapified(dsl_dataset_t *ds) | |
4147 | { | |
4148 | dmu_object_info_t doi; | |
4149 | ||
4150 | dmu_object_info_from_db(ds->ds_dbuf, &doi); | |
4151 | return (doi.doi_type == DMU_OTN_ZAP_METADATA); | |
4152 | } | |
4153 | ||
4154 | boolean_t | |
4155 | dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds) | |
4156 | { | |
4157 | return (dsl_dataset_is_zapified(ds) && | |
4158 | zap_contains(ds->ds_dir->dd_pool->dp_meta_objset, | |
4159 | ds->ds_object, DS_FIELD_RESUME_TOGUID) == 0); | |
4160 | } | |
4161 | ||
a1d477c2 MA |
4162 | uint64_t |
4163 | dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds) | |
4164 | { | |
4165 | uint64_t remap_deadlist_obj; | |
4166 | int err; | |
4167 | ||
4168 | if (!dsl_dataset_is_zapified(ds)) | |
4169 | return (0); | |
4170 | ||
4171 | err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object, | |
4172 | DS_FIELD_REMAP_DEADLIST, sizeof (remap_deadlist_obj), 1, | |
4173 | &remap_deadlist_obj); | |
4174 | ||
4175 | if (err != 0) { | |
4176 | VERIFY3S(err, ==, ENOENT); | |
4177 | return (0); | |
4178 | } | |
4179 | ||
4180 | ASSERT(remap_deadlist_obj != 0); | |
4181 | return (remap_deadlist_obj); | |
4182 | } | |
4183 | ||
4184 | boolean_t | |
4185 | dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds) | |
4186 | { | |
4187 | EQUIV(dsl_deadlist_is_open(&ds->ds_remap_deadlist), | |
4188 | dsl_dataset_get_remap_deadlist_object(ds) != 0); | |
4189 | return (dsl_deadlist_is_open(&ds->ds_remap_deadlist)); | |
4190 | } | |
4191 | ||
4192 | static void | |
4193 | dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, uint64_t obj, | |
4194 | dmu_tx_t *tx) | |
4195 | { | |
4196 | ASSERT(obj != 0); | |
4197 | dsl_dataset_zapify(ds, tx); | |
4198 | VERIFY0(zap_add(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object, | |
4199 | DS_FIELD_REMAP_DEADLIST, sizeof (obj), 1, &obj, tx)); | |
4200 | } | |
4201 | ||
4202 | static void | |
4203 | dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4204 | { | |
4205 | VERIFY0(zap_remove(ds->ds_dir->dd_pool->dp_meta_objset, | |
4206 | ds->ds_object, DS_FIELD_REMAP_DEADLIST, tx)); | |
4207 | } | |
4208 | ||
4209 | void | |
4210 | dsl_dataset_destroy_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4211 | { | |
4212 | uint64_t remap_deadlist_object; | |
4213 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
4214 | ||
4215 | ASSERT(dmu_tx_is_syncing(tx)); | |
4216 | ASSERT(dsl_dataset_remap_deadlist_exists(ds)); | |
4217 | ||
4218 | remap_deadlist_object = ds->ds_remap_deadlist.dl_object; | |
4219 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
4220 | dsl_deadlist_free(spa_meta_objset(spa), remap_deadlist_object, tx); | |
4221 | dsl_dataset_unset_remap_deadlist_object(ds, tx); | |
4222 | spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx); | |
4223 | } | |
4224 | ||
4225 | void | |
4226 | dsl_dataset_create_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4227 | { | |
4228 | uint64_t remap_deadlist_obj; | |
4229 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
4230 | ||
4231 | ASSERT(dmu_tx_is_syncing(tx)); | |
4232 | ASSERT(MUTEX_HELD(&ds->ds_remap_deadlist_lock)); | |
4233 | /* | |
4234 | * Currently we only create remap deadlists when there are indirect | |
4235 | * vdevs with referenced mappings. | |
4236 | */ | |
4237 | ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); | |
4238 | ||
4239 | remap_deadlist_obj = dsl_deadlist_clone( | |
4240 | &ds->ds_deadlist, UINT64_MAX, | |
4241 | dsl_dataset_phys(ds)->ds_prev_snap_obj, tx); | |
4242 | dsl_dataset_set_remap_deadlist_object(ds, | |
4243 | remap_deadlist_obj, tx); | |
4244 | dsl_deadlist_open(&ds->ds_remap_deadlist, spa_meta_objset(spa), | |
4245 | remap_deadlist_obj); | |
4246 | spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx); | |
4247 | } | |
4248 | ||
93ce2b4c | 4249 | #if defined(_KERNEL) |
f1512ee6 MA |
4250 | #if defined(_LP64) |
4251 | module_param(zfs_max_recordsize, int, 0644); | |
4252 | MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size"); | |
4253 | #else | |
4254 | /* Limited to 1M on 32-bit platforms due to lack of virtual address space */ | |
4255 | module_param(zfs_max_recordsize, int, 0444); | |
4256 | MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size"); | |
4257 | #endif | |
4258 | ||
c28b2279 | 4259 | EXPORT_SYMBOL(dsl_dataset_hold); |
b5256303 | 4260 | EXPORT_SYMBOL(dsl_dataset_hold_flags); |
c28b2279 | 4261 | EXPORT_SYMBOL(dsl_dataset_hold_obj); |
b5256303 | 4262 | EXPORT_SYMBOL(dsl_dataset_hold_obj_flags); |
c28b2279 BB |
4263 | EXPORT_SYMBOL(dsl_dataset_own); |
4264 | EXPORT_SYMBOL(dsl_dataset_own_obj); | |
4265 | EXPORT_SYMBOL(dsl_dataset_name); | |
4266 | EXPORT_SYMBOL(dsl_dataset_rele); | |
b5256303 | 4267 | EXPORT_SYMBOL(dsl_dataset_rele_flags); |
c28b2279 | 4268 | EXPORT_SYMBOL(dsl_dataset_disown); |
c28b2279 | 4269 | EXPORT_SYMBOL(dsl_dataset_tryown); |
c28b2279 BB |
4270 | EXPORT_SYMBOL(dsl_dataset_create_sync); |
4271 | EXPORT_SYMBOL(dsl_dataset_create_sync_dd); | |
c28b2279 BB |
4272 | EXPORT_SYMBOL(dsl_dataset_snapshot_check); |
4273 | EXPORT_SYMBOL(dsl_dataset_snapshot_sync); | |
c28b2279 | 4274 | EXPORT_SYMBOL(dsl_dataset_promote); |
c28b2279 BB |
4275 | EXPORT_SYMBOL(dsl_dataset_user_hold); |
4276 | EXPORT_SYMBOL(dsl_dataset_user_release); | |
c28b2279 BB |
4277 | EXPORT_SYMBOL(dsl_dataset_get_holds); |
4278 | EXPORT_SYMBOL(dsl_dataset_get_blkptr); | |
c28b2279 | 4279 | EXPORT_SYMBOL(dsl_dataset_get_spa); |
19580676 | 4280 | EXPORT_SYMBOL(dsl_dataset_modified_since_snap); |
330d06f9 MA |
4281 | EXPORT_SYMBOL(dsl_dataset_space_written); |
4282 | EXPORT_SYMBOL(dsl_dataset_space_wouldfree); | |
c28b2279 BB |
4283 | EXPORT_SYMBOL(dsl_dataset_sync); |
4284 | EXPORT_SYMBOL(dsl_dataset_block_born); | |
4285 | EXPORT_SYMBOL(dsl_dataset_block_kill); | |
c28b2279 BB |
4286 | EXPORT_SYMBOL(dsl_dataset_dirty); |
4287 | EXPORT_SYMBOL(dsl_dataset_stats); | |
4288 | EXPORT_SYMBOL(dsl_dataset_fast_stat); | |
4289 | EXPORT_SYMBOL(dsl_dataset_space); | |
4290 | EXPORT_SYMBOL(dsl_dataset_fsid_guid); | |
4291 | EXPORT_SYMBOL(dsl_dsobj_to_dsname); | |
4292 | EXPORT_SYMBOL(dsl_dataset_check_quota); | |
13fe0198 MA |
4293 | EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl); |
4294 | EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl); | |
c28b2279 | 4295 | #endif |