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