]>
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> | |
d2734cce | 49 | #include <sys/spa_impl.h> |
a1d477c2 | 50 | #include <sys/vdev.h> |
b128c09f | 51 | #include <sys/zfs_znode.h> |
572e2857 | 52 | #include <sys/zfs_onexit.h> |
45d1cae3 | 53 | #include <sys/zvol.h> |
428870ff BB |
54 | #include <sys/dsl_scan.h> |
55 | #include <sys/dsl_deadlist.h> | |
13fe0198 MA |
56 | #include <sys/dsl_destroy.h> |
57 | #include <sys/dsl_userhold.h> | |
da536844 | 58 | #include <sys/dsl_bookmark.h> |
f74b821a | 59 | #include <sys/policy.h> |
47dfff3b MA |
60 | #include <sys/dmu_send.h> |
61 | #include <sys/zio_compress.h> | |
62 | #include <zfs_fletcher.h> | |
3c67d83a | 63 | #include <sys/zio_checksum.h> |
34dc7c2f | 64 | |
f1512ee6 MA |
65 | /* |
66 | * The SPA supports block sizes up to 16MB. However, very large blocks | |
67 | * can have an impact on i/o latency (e.g. tying up a spinning disk for | |
68 | * ~300ms), and also potentially on the memory allocator. Therefore, | |
69 | * we do not allow the recordsize to be set larger than zfs_max_recordsize | |
70 | * (default 1MB). Larger blocks can be created by changing this tunable, | |
71 | * and pools with larger blocks can always be imported and used, regardless | |
72 | * of this setting. | |
73 | */ | |
74 | int zfs_max_recordsize = 1 * 1024 * 1024; | |
75 | ||
428870ff BB |
76 | #define SWITCH64(x, y) \ |
77 | { \ | |
78 | uint64_t __tmp = (x); \ | |
79 | (x) = (y); \ | |
80 | (y) = __tmp; \ | |
81 | } | |
82 | ||
34dc7c2f BB |
83 | #define DS_REF_MAX (1ULL << 62) |
84 | ||
d683ddbb | 85 | extern inline dsl_dataset_phys_t *dsl_dataset_phys(dsl_dataset_t *ds); |
d683ddbb | 86 | |
a1d477c2 MA |
87 | static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, |
88 | uint64_t obj, dmu_tx_t *tx); | |
89 | static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds, | |
90 | dmu_tx_t *tx); | |
91 | ||
8c62a0d0 DM |
92 | extern int spa_asize_inflation; |
93 | ||
0efd9791 AG |
94 | static zil_header_t zero_zil; |
95 | ||
34dc7c2f | 96 | /* |
4e33ba4c | 97 | * Figure out how much of this delta should be propagated to the dsl_dir |
34dc7c2f BB |
98 | * layer. If there's a refreservation, that space has already been |
99 | * partially accounted for in our ancestors. | |
100 | */ | |
101 | static int64_t | |
102 | parent_delta(dsl_dataset_t *ds, int64_t delta) | |
103 | { | |
d683ddbb | 104 | dsl_dataset_phys_t *ds_phys; |
34dc7c2f BB |
105 | uint64_t old_bytes, new_bytes; |
106 | ||
107 | if (ds->ds_reserved == 0) | |
108 | return (delta); | |
109 | ||
d683ddbb JG |
110 | ds_phys = dsl_dataset_phys(ds); |
111 | old_bytes = MAX(ds_phys->ds_unique_bytes, ds->ds_reserved); | |
112 | new_bytes = MAX(ds_phys->ds_unique_bytes + delta, ds->ds_reserved); | |
34dc7c2f BB |
113 | |
114 | ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta)); | |
115 | return (new_bytes - old_bytes); | |
116 | } | |
117 | ||
118 | void | |
428870ff | 119 | dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) |
34dc7c2f | 120 | { |
d6320ddb | 121 | int used, compressed, uncompressed; |
34dc7c2f BB |
122 | int64_t delta; |
123 | ||
d6320ddb BB |
124 | used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp); |
125 | compressed = BP_GET_PSIZE(bp); | |
126 | uncompressed = BP_GET_UCSIZE(bp); | |
127 | ||
428870ff | 128 | dprintf_bp(bp, "ds=%p", ds); |
34dc7c2f BB |
129 | |
130 | ASSERT(dmu_tx_is_syncing(tx)); | |
131 | /* It could have been compressed away to nothing */ | |
132 | if (BP_IS_HOLE(bp)) | |
133 | return; | |
134 | ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE); | |
9ae529ec | 135 | ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp))); |
34dc7c2f | 136 | if (ds == NULL) { |
29809a6c MA |
137 | dsl_pool_mos_diduse_space(tx->tx_pool, |
138 | used, compressed, uncompressed); | |
34dc7c2f BB |
139 | return; |
140 | } | |
428870ff | 141 | |
0efd9791 | 142 | ASSERT3U(bp->blk_birth, >, dsl_dataset_phys(ds)->ds_prev_snap_txg); |
a169a625 | 143 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
34dc7c2f BB |
144 | mutex_enter(&ds->ds_lock); |
145 | delta = parent_delta(ds, used); | |
d683ddbb JG |
146 | dsl_dataset_phys(ds)->ds_referenced_bytes += used; |
147 | dsl_dataset_phys(ds)->ds_compressed_bytes += compressed; | |
148 | dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed; | |
149 | dsl_dataset_phys(ds)->ds_unique_bytes += used; | |
3c67d83a | 150 | |
241b5415 MA |
151 | if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE) { |
152 | ds->ds_feature_activation_needed[SPA_FEATURE_LARGE_BLOCKS] = | |
153 | B_TRUE; | |
154 | } | |
3c67d83a | 155 | |
1c27024e | 156 | spa_feature_t f = zio_checksum_to_feature(BP_GET_CHECKSUM(bp)); |
3c67d83a TH |
157 | if (f != SPA_FEATURE_NONE) |
158 | ds->ds_feature_activation_needed[f] = B_TRUE; | |
159 | ||
34dc7c2f | 160 | mutex_exit(&ds->ds_lock); |
b128c09f BB |
161 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta, |
162 | compressed, uncompressed, tx); | |
163 | dsl_dir_transfer_space(ds->ds_dir, used - delta, | |
164 | DD_USED_REFRSRV, DD_USED_HEAD, tx); | |
34dc7c2f BB |
165 | } |
166 | ||
a1d477c2 MA |
167 | /* |
168 | * Called when the specified segment has been remapped, and is thus no | |
169 | * longer referenced in the head dataset. The vdev must be indirect. | |
170 | * | |
171 | * If the segment is referenced by a snapshot, put it on the remap deadlist. | |
172 | * Otherwise, add this segment to the obsolete spacemap. | |
173 | */ | |
174 | void | |
175 | dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev, uint64_t offset, | |
176 | uint64_t size, uint64_t birth, dmu_tx_t *tx) | |
177 | { | |
178 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
179 | ||
180 | ASSERT(dmu_tx_is_syncing(tx)); | |
181 | ASSERT(birth <= tx->tx_txg); | |
182 | ASSERT(!ds->ds_is_snapshot); | |
183 | ||
184 | if (birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) { | |
185 | spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx); | |
186 | } else { | |
187 | blkptr_t fakebp; | |
188 | dva_t *dva = &fakebp.blk_dva[0]; | |
189 | ||
190 | ASSERT(ds != NULL); | |
191 | ||
192 | mutex_enter(&ds->ds_remap_deadlist_lock); | |
193 | if (!dsl_dataset_remap_deadlist_exists(ds)) { | |
194 | dsl_dataset_create_remap_deadlist(ds, tx); | |
195 | } | |
196 | mutex_exit(&ds->ds_remap_deadlist_lock); | |
197 | ||
198 | BP_ZERO(&fakebp); | |
199 | fakebp.blk_birth = birth; | |
200 | DVA_SET_VDEV(dva, vdev); | |
201 | DVA_SET_OFFSET(dva, offset); | |
202 | DVA_SET_ASIZE(dva, size); | |
203 | ||
204 | dsl_deadlist_insert(&ds->ds_remap_deadlist, &fakebp, tx); | |
205 | } | |
206 | } | |
207 | ||
b128c09f | 208 | int |
428870ff BB |
209 | dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, |
210 | boolean_t async) | |
34dc7c2f | 211 | { |
d2734cce SD |
212 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; |
213 | ||
214 | int used = bp_get_dsize_sync(spa, bp); | |
b0bc7a84 MG |
215 | int compressed = BP_GET_PSIZE(bp); |
216 | int uncompressed = BP_GET_UCSIZE(bp); | |
d6320ddb | 217 | |
34dc7c2f | 218 | if (BP_IS_HOLE(bp)) |
b128c09f | 219 | return (0); |
34dc7c2f | 220 | |
428870ff BB |
221 | ASSERT(dmu_tx_is_syncing(tx)); |
222 | ASSERT(bp->blk_birth <= tx->tx_txg); | |
223 | ||
34dc7c2f | 224 | if (ds == NULL) { |
428870ff | 225 | dsl_free(tx->tx_pool, tx->tx_txg, bp); |
29809a6c MA |
226 | dsl_pool_mos_diduse_space(tx->tx_pool, |
227 | -used, -compressed, -uncompressed); | |
b128c09f | 228 | return (used); |
34dc7c2f BB |
229 | } |
230 | ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool); | |
231 | ||
0c66c32d | 232 | ASSERT(!ds->ds_is_snapshot); |
34dc7c2f BB |
233 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
234 | ||
d683ddbb | 235 | if (bp->blk_birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) { |
34dc7c2f BB |
236 | int64_t delta; |
237 | ||
428870ff BB |
238 | dprintf_bp(bp, "freeing ds=%llu", ds->ds_object); |
239 | dsl_free(tx->tx_pool, tx->tx_txg, bp); | |
34dc7c2f BB |
240 | |
241 | mutex_enter(&ds->ds_lock); | |
d683ddbb | 242 | ASSERT(dsl_dataset_phys(ds)->ds_unique_bytes >= used || |
34dc7c2f BB |
243 | !DS_UNIQUE_IS_ACCURATE(ds)); |
244 | delta = parent_delta(ds, -used); | |
d683ddbb | 245 | dsl_dataset_phys(ds)->ds_unique_bytes -= used; |
34dc7c2f | 246 | mutex_exit(&ds->ds_lock); |
b128c09f | 247 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, |
34dc7c2f | 248 | delta, -compressed, -uncompressed, tx); |
b128c09f BB |
249 | dsl_dir_transfer_space(ds->ds_dir, -used - delta, |
250 | DD_USED_REFRSRV, DD_USED_HEAD, tx); | |
34dc7c2f BB |
251 | } else { |
252 | dprintf_bp(bp, "putting on dead list: %s", ""); | |
428870ff BB |
253 | if (async) { |
254 | /* | |
255 | * We are here as part of zio's write done callback, | |
256 | * which means we're a zio interrupt thread. We can't | |
257 | * call dsl_deadlist_insert() now because it may block | |
258 | * waiting for I/O. Instead, put bp on the deferred | |
259 | * queue and let dsl_pool_sync() finish the job. | |
260 | */ | |
261 | bplist_append(&ds->ds_pending_deadlist, bp); | |
262 | } else { | |
263 | dsl_deadlist_insert(&ds->ds_deadlist, bp, tx); | |
264 | } | |
34dc7c2f | 265 | ASSERT3U(ds->ds_prev->ds_object, ==, |
d683ddbb JG |
266 | dsl_dataset_phys(ds)->ds_prev_snap_obj); |
267 | ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0); | |
34dc7c2f | 268 | /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */ |
d683ddbb | 269 | if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == |
34dc7c2f | 270 | ds->ds_object && bp->blk_birth > |
d683ddbb | 271 | dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) { |
34dc7c2f BB |
272 | dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); |
273 | mutex_enter(&ds->ds_prev->ds_lock); | |
d683ddbb | 274 | dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used; |
34dc7c2f BB |
275 | mutex_exit(&ds->ds_prev->ds_lock); |
276 | } | |
428870ff | 277 | if (bp->blk_birth > ds->ds_dir->dd_origin_txg) { |
b128c09f BB |
278 | dsl_dir_transfer_space(ds->ds_dir, used, |
279 | DD_USED_HEAD, DD_USED_SNAP, tx); | |
280 | } | |
34dc7c2f BB |
281 | } |
282 | mutex_enter(&ds->ds_lock); | |
d683ddbb JG |
283 | ASSERT3U(dsl_dataset_phys(ds)->ds_referenced_bytes, >=, used); |
284 | dsl_dataset_phys(ds)->ds_referenced_bytes -= used; | |
285 | ASSERT3U(dsl_dataset_phys(ds)->ds_compressed_bytes, >=, compressed); | |
286 | dsl_dataset_phys(ds)->ds_compressed_bytes -= compressed; | |
287 | ASSERT3U(dsl_dataset_phys(ds)->ds_uncompressed_bytes, >=, uncompressed); | |
288 | dsl_dataset_phys(ds)->ds_uncompressed_bytes -= uncompressed; | |
34dc7c2f | 289 | mutex_exit(&ds->ds_lock); |
b128c09f BB |
290 | |
291 | return (used); | |
34dc7c2f BB |
292 | } |
293 | ||
39efbde7 GM |
294 | /* |
295 | * We have to release the fsid syncronously or we risk that a subsequent | |
296 | * mount of the same dataset will fail to unique_insert the fsid. This | |
297 | * failure would manifest itself as the fsid of this dataset changing | |
298 | * between mounts which makes NFS clients quite unhappy. | |
299 | */ | |
34dc7c2f | 300 | static void |
39efbde7 | 301 | dsl_dataset_evict_sync(void *dbu) |
34dc7c2f | 302 | { |
0c66c32d | 303 | dsl_dataset_t *ds = dbu; |
34dc7c2f | 304 | |
13fe0198 | 305 | ASSERT(ds->ds_owner == NULL); |
34dc7c2f | 306 | |
34dc7c2f | 307 | unique_remove(ds->ds_fsid_guid); |
39efbde7 GM |
308 | } |
309 | ||
310 | static void | |
311 | dsl_dataset_evict_async(void *dbu) | |
312 | { | |
313 | dsl_dataset_t *ds = dbu; | |
314 | ||
315 | ASSERT(ds->ds_owner == NULL); | |
316 | ||
317 | ds->ds_dbuf = NULL; | |
34dc7c2f | 318 | |
428870ff BB |
319 | if (ds->ds_objset != NULL) |
320 | dmu_objset_evict(ds->ds_objset); | |
34dc7c2f BB |
321 | |
322 | if (ds->ds_prev) { | |
13fe0198 | 323 | dsl_dataset_rele(ds->ds_prev, ds); |
34dc7c2f BB |
324 | ds->ds_prev = NULL; |
325 | } | |
326 | ||
428870ff | 327 | bplist_destroy(&ds->ds_pending_deadlist); |
a1d477c2 | 328 | if (dsl_deadlist_is_open(&ds->ds_deadlist)) |
428870ff | 329 | dsl_deadlist_close(&ds->ds_deadlist); |
a1d477c2 MA |
330 | if (dsl_deadlist_is_open(&ds->ds_remap_deadlist)) |
331 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
b128c09f | 332 | if (ds->ds_dir) |
0c66c32d | 333 | dsl_dir_async_rele(ds->ds_dir, ds); |
34dc7c2f BB |
334 | |
335 | ASSERT(!list_link_active(&ds->ds_synced_link)); | |
336 | ||
0eb21616 | 337 | list_destroy(&ds->ds_prop_cbs); |
34dc7c2f BB |
338 | mutex_destroy(&ds->ds_lock); |
339 | mutex_destroy(&ds->ds_opening_lock); | |
58c4aa00 | 340 | mutex_destroy(&ds->ds_sendstream_lock); |
a1d477c2 | 341 | mutex_destroy(&ds->ds_remap_deadlist_lock); |
13fe0198 | 342 | refcount_destroy(&ds->ds_longholds); |
cc9bb3e5 | 343 | rrw_destroy(&ds->ds_bp_rwlock); |
34dc7c2f BB |
344 | |
345 | kmem_free(ds, sizeof (dsl_dataset_t)); | |
346 | } | |
347 | ||
13fe0198 | 348 | int |
34dc7c2f BB |
349 | dsl_dataset_get_snapname(dsl_dataset_t *ds) |
350 | { | |
351 | dsl_dataset_phys_t *headphys; | |
352 | int err; | |
353 | dmu_buf_t *headdbuf; | |
354 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
355 | objset_t *mos = dp->dp_meta_objset; | |
356 | ||
357 | if (ds->ds_snapname[0]) | |
358 | return (0); | |
d683ddbb | 359 | if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) |
34dc7c2f BB |
360 | return (0); |
361 | ||
d683ddbb | 362 | err = dmu_bonus_hold(mos, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj, |
34dc7c2f | 363 | FTAG, &headdbuf); |
13fe0198 | 364 | if (err != 0) |
34dc7c2f BB |
365 | return (err); |
366 | headphys = headdbuf->db_data; | |
367 | err = zap_value_search(dp->dp_meta_objset, | |
368 | headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname); | |
d439f63f TC |
369 | if (err != 0 && zfs_recover == B_TRUE) { |
370 | err = 0; | |
371 | (void) snprintf(ds->ds_snapname, sizeof (ds->ds_snapname), | |
372 | "SNAPOBJ=%llu-ERR=%d", | |
373 | (unsigned long long)ds->ds_object, err); | |
374 | } | |
34dc7c2f BB |
375 | dmu_buf_rele(headdbuf, FTAG); |
376 | return (err); | |
377 | } | |
378 | ||
6772fb67 | 379 | int |
b128c09f | 380 | dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value) |
34dc7c2f | 381 | { |
b128c09f | 382 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
d683ddbb | 383 | uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; |
9b7b9cd3 | 384 | matchtype_t mt = 0; |
34dc7c2f BB |
385 | int err; |
386 | ||
d683ddbb | 387 | if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) |
9b7b9cd3 | 388 | mt = MT_NORMALIZE; |
34dc7c2f | 389 | |
b128c09f | 390 | err = zap_lookup_norm(mos, snapobj, name, 8, 1, |
34dc7c2f | 391 | value, mt, NULL, 0, NULL); |
9b7b9cd3 | 392 | if (err == ENOTSUP && (mt & MT_NORMALIZE)) |
b128c09f | 393 | err = zap_lookup(mos, snapobj, name, 8, 1, value); |
34dc7c2f BB |
394 | return (err); |
395 | } | |
396 | ||
13fe0198 | 397 | int |
788eb90c JJ |
398 | dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx, |
399 | boolean_t adj_cnt) | |
34dc7c2f | 400 | { |
b128c09f | 401 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
d683ddbb | 402 | uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; |
9b7b9cd3 | 403 | matchtype_t mt = 0; |
34dc7c2f BB |
404 | int err; |
405 | ||
428870ff BB |
406 | dsl_dir_snap_cmtime_update(ds->ds_dir); |
407 | ||
d683ddbb | 408 | if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET) |
9b7b9cd3 | 409 | mt = MT_NORMALIZE; |
34dc7c2f | 410 | |
b128c09f | 411 | err = zap_remove_norm(mos, snapobj, name, mt, tx); |
9b7b9cd3 | 412 | if (err == ENOTSUP && (mt & MT_NORMALIZE)) |
b128c09f | 413 | err = zap_remove(mos, snapobj, name, tx); |
788eb90c JJ |
414 | |
415 | if (err == 0 && adj_cnt) | |
416 | dsl_fs_ss_count_adjust(ds->ds_dir, -1, | |
417 | DD_FIELD_SNAPSHOT_COUNT, tx); | |
418 | ||
34dc7c2f BB |
419 | return (err); |
420 | } | |
421 | ||
6ebebace JG |
422 | boolean_t |
423 | dsl_dataset_try_add_ref(dsl_pool_t *dp, dsl_dataset_t *ds, void *tag) | |
424 | { | |
6186e297 JG |
425 | dmu_buf_t *dbuf = ds->ds_dbuf; |
426 | boolean_t result = B_FALSE; | |
427 | ||
428 | if (dbuf != NULL && dmu_buf_try_add_ref(dbuf, dp->dp_meta_objset, | |
429 | ds->ds_object, DMU_BONUS_BLKID, tag)) { | |
430 | ||
431 | if (ds == dmu_buf_get_user(dbuf)) | |
432 | result = B_TRUE; | |
433 | else | |
434 | dmu_buf_rele(dbuf, tag); | |
435 | } | |
436 | ||
437 | return (result); | |
6ebebace JG |
438 | } |
439 | ||
13fe0198 | 440 | int |
b5256303 TC |
441 | dsl_dataset_hold_obj_flags(dsl_pool_t *dp, uint64_t dsobj, |
442 | ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp) | |
34dc7c2f | 443 | { |
34dc7c2f BB |
444 | objset_t *mos = dp->dp_meta_objset; |
445 | dmu_buf_t *dbuf; | |
446 | dsl_dataset_t *ds; | |
447 | int err; | |
572e2857 | 448 | dmu_object_info_t doi; |
34dc7c2f | 449 | |
13fe0198 | 450 | ASSERT(dsl_pool_config_held(dp)); |
34dc7c2f BB |
451 | |
452 | err = dmu_bonus_hold(mos, dsobj, tag, &dbuf); | |
13fe0198 | 453 | if (err != 0) |
34dc7c2f | 454 | return (err); |
572e2857 BB |
455 | |
456 | /* Make sure dsobj has the correct object type. */ | |
457 | dmu_object_info_from_db(dbuf, &doi); | |
fa86b5db | 458 | if (doi.doi_bonus_type != DMU_OT_DSL_DATASET) { |
3a84951d | 459 | dmu_buf_rele(dbuf, tag); |
2e528b49 | 460 | return (SET_ERROR(EINVAL)); |
3a84951d | 461 | } |
572e2857 | 462 | |
34dc7c2f BB |
463 | ds = dmu_buf_get_user(dbuf); |
464 | if (ds == NULL) { | |
d4ed6673 | 465 | dsl_dataset_t *winner = NULL; |
34dc7c2f | 466 | |
79c76d5b | 467 | ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP); |
34dc7c2f BB |
468 | ds->ds_dbuf = dbuf; |
469 | ds->ds_object = dsobj; | |
0c66c32d | 470 | ds->ds_is_snapshot = dsl_dataset_phys(ds)->ds_num_children != 0; |
98f72a53 | 471 | list_link_init(&ds->ds_synced_link); |
34dc7c2f | 472 | |
a1d477c2 MA |
473 | err = dsl_dir_hold_obj(dp, dsl_dataset_phys(ds)->ds_dir_obj, |
474 | NULL, ds, &ds->ds_dir); | |
475 | if (err != 0) { | |
476 | kmem_free(ds, sizeof (dsl_dataset_t)); | |
477 | dmu_buf_rele(dbuf, tag); | |
478 | return (err); | |
479 | } | |
480 | ||
34dc7c2f BB |
481 | mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL); |
482 | mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL); | |
37abac6d | 483 | mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL); |
a1d477c2 MA |
484 | mutex_init(&ds->ds_remap_deadlist_lock, |
485 | NULL, MUTEX_DEFAULT, NULL); | |
cc9bb3e5 | 486 | rrw_init(&ds->ds_bp_rwlock, B_FALSE); |
13fe0198 | 487 | refcount_create(&ds->ds_longholds); |
34dc7c2f | 488 | |
428870ff | 489 | bplist_create(&ds->ds_pending_deadlist); |
428870ff | 490 | |
37abac6d BP |
491 | list_create(&ds->ds_sendstreams, sizeof (dmu_sendarg_t), |
492 | offsetof(dmu_sendarg_t, dsa_link)); | |
493 | ||
0eb21616 JG |
494 | list_create(&ds->ds_prop_cbs, sizeof (dsl_prop_cb_record_t), |
495 | offsetof(dsl_prop_cb_record_t, cbr_ds_node)); | |
496 | ||
f1512ee6 | 497 | if (doi.doi_type == DMU_OTN_ZAP_METADATA) { |
241b5415 MA |
498 | spa_feature_t f; |
499 | ||
500 | for (f = 0; f < SPA_FEATURES; f++) { | |
501 | if (!(spa_feature_table[f].fi_flags & | |
502 | ZFEATURE_FLAG_PER_DATASET)) | |
503 | continue; | |
504 | err = zap_contains(mos, dsobj, | |
505 | spa_feature_table[f].fi_guid); | |
506 | if (err == 0) { | |
507 | ds->ds_feature_inuse[f] = B_TRUE; | |
508 | } else { | |
509 | ASSERT3U(err, ==, ENOENT); | |
510 | err = 0; | |
511 | } | |
19b3b1d2 | 512 | } |
f1512ee6 MA |
513 | } |
514 | ||
0c66c32d | 515 | if (!ds->ds_is_snapshot) { |
34dc7c2f | 516 | ds->ds_snapname[0] = '\0'; |
d683ddbb | 517 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { |
13fe0198 | 518 | err = dsl_dataset_hold_obj(dp, |
d683ddbb | 519 | dsl_dataset_phys(ds)->ds_prev_snap_obj, |
b128c09f | 520 | ds, &ds->ds_prev); |
34dc7c2f | 521 | } |
da536844 MA |
522 | if (doi.doi_type == DMU_OTN_ZAP_METADATA) { |
523 | int zaperr = zap_lookup(mos, ds->ds_object, | |
524 | DS_FIELD_BOOKMARK_NAMES, | |
525 | sizeof (ds->ds_bookmarks), 1, | |
526 | &ds->ds_bookmarks); | |
527 | if (zaperr != ENOENT) | |
528 | VERIFY0(zaperr); | |
529 | } | |
45d1cae3 BB |
530 | } else { |
531 | if (zfs_flags & ZFS_DEBUG_SNAPNAMES) | |
532 | err = dsl_dataset_get_snapname(ds); | |
d683ddbb JG |
533 | if (err == 0 && |
534 | dsl_dataset_phys(ds)->ds_userrefs_obj != 0) { | |
45d1cae3 BB |
535 | err = zap_count( |
536 | ds->ds_dir->dd_pool->dp_meta_objset, | |
d683ddbb | 537 | dsl_dataset_phys(ds)->ds_userrefs_obj, |
45d1cae3 BB |
538 | &ds->ds_userrefs); |
539 | } | |
34dc7c2f BB |
540 | } |
541 | ||
0c66c32d | 542 | if (err == 0 && !ds->ds_is_snapshot) { |
13fe0198 MA |
543 | err = dsl_prop_get_int_ds(ds, |
544 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), | |
545 | &ds->ds_reserved); | |
34dc7c2f | 546 | if (err == 0) { |
13fe0198 MA |
547 | err = dsl_prop_get_int_ds(ds, |
548 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
549 | &ds->ds_quota); | |
34dc7c2f | 550 | } |
34dc7c2f BB |
551 | } else { |
552 | ds->ds_reserved = ds->ds_quota = 0; | |
553 | } | |
554 | ||
a1d477c2 MA |
555 | dsl_deadlist_open(&ds->ds_deadlist, |
556 | mos, dsl_dataset_phys(ds)->ds_deadlist_obj); | |
557 | uint64_t remap_deadlist_obj = | |
558 | dsl_dataset_get_remap_deadlist_object(ds); | |
559 | if (remap_deadlist_obj != 0) { | |
560 | dsl_deadlist_open(&ds->ds_remap_deadlist, mos, | |
561 | remap_deadlist_obj); | |
562 | } | |
563 | ||
39efbde7 GM |
564 | dmu_buf_init_user(&ds->ds_dbu, dsl_dataset_evict_sync, |
565 | dsl_dataset_evict_async, &ds->ds_dbuf); | |
0c66c32d JG |
566 | if (err == 0) |
567 | winner = dmu_buf_set_user_ie(dbuf, &ds->ds_dbu); | |
568 | ||
569 | if (err != 0 || winner != NULL) { | |
428870ff BB |
570 | bplist_destroy(&ds->ds_pending_deadlist); |
571 | dsl_deadlist_close(&ds->ds_deadlist); | |
a1d477c2 MA |
572 | if (dsl_deadlist_is_open(&ds->ds_remap_deadlist)) |
573 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
b128c09f | 574 | if (ds->ds_prev) |
13fe0198 MA |
575 | dsl_dataset_rele(ds->ds_prev, ds); |
576 | dsl_dir_rele(ds->ds_dir, ds); | |
34dc7c2f BB |
577 | mutex_destroy(&ds->ds_lock); |
578 | mutex_destroy(&ds->ds_opening_lock); | |
58c4aa00 | 579 | mutex_destroy(&ds->ds_sendstream_lock); |
13fe0198 | 580 | refcount_destroy(&ds->ds_longholds); |
34dc7c2f | 581 | kmem_free(ds, sizeof (dsl_dataset_t)); |
13fe0198 | 582 | if (err != 0) { |
34dc7c2f BB |
583 | dmu_buf_rele(dbuf, tag); |
584 | return (err); | |
585 | } | |
586 | ds = winner; | |
587 | } else { | |
588 | ds->ds_fsid_guid = | |
d683ddbb | 589 | unique_insert(dsl_dataset_phys(ds)->ds_fsid_guid); |
39efbde7 GM |
590 | if (ds->ds_fsid_guid != |
591 | dsl_dataset_phys(ds)->ds_fsid_guid) { | |
592 | zfs_dbgmsg("ds_fsid_guid changed from " | |
593 | "%llx to %llx for pool %s dataset id %llu", | |
594 | (long long) | |
595 | dsl_dataset_phys(ds)->ds_fsid_guid, | |
596 | (long long)ds->ds_fsid_guid, | |
597 | spa_name(dp->dp_spa), | |
598 | dsobj); | |
599 | } | |
34dc7c2f BB |
600 | } |
601 | } | |
602 | ASSERT3P(ds->ds_dbuf, ==, dbuf); | |
d683ddbb JG |
603 | ASSERT3P(dsl_dataset_phys(ds), ==, dbuf->db_data); |
604 | ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0 || | |
b128c09f BB |
605 | spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN || |
606 | dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap); | |
34dc7c2f | 607 | *dsp = ds; |
b5256303 TC |
608 | |
609 | if ((flags & DS_HOLD_FLAG_DECRYPT) && ds->ds_dir->dd_crypto_obj != 0) { | |
610 | err = spa_keystore_create_mapping(dp->dp_spa, ds, ds); | |
611 | if (err != 0) { | |
612 | dsl_dataset_rele(ds, tag); | |
613 | return (SET_ERROR(EACCES)); | |
614 | } | |
615 | } | |
616 | ||
34dc7c2f BB |
617 | return (0); |
618 | } | |
619 | ||
b128c09f | 620 | int |
b5256303 TC |
621 | dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag, |
622 | dsl_dataset_t **dsp) | |
623 | { | |
624 | return (dsl_dataset_hold_obj_flags(dp, dsobj, 0, tag, dsp)); | |
625 | } | |
626 | ||
627 | int | |
628 | dsl_dataset_hold_flags(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags, | |
428870ff | 629 | void *tag, dsl_dataset_t **dsp) |
34dc7c2f BB |
630 | { |
631 | dsl_dir_t *dd; | |
b128c09f | 632 | const char *snapname; |
34dc7c2f | 633 | uint64_t obj; |
34dc7c2f | 634 | int err = 0; |
fcff0f35 | 635 | dsl_dataset_t *ds; |
34dc7c2f | 636 | |
13fe0198 MA |
637 | err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname); |
638 | if (err != 0) | |
34dc7c2f BB |
639 | return (err); |
640 | ||
13fe0198 | 641 | ASSERT(dsl_pool_config_held(dp)); |
d683ddbb | 642 | obj = dsl_dir_phys(dd)->dd_head_dataset_obj; |
13fe0198 | 643 | if (obj != 0) |
b5256303 | 644 | err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, &ds); |
b128c09f | 645 | else |
2e528b49 | 646 | err = SET_ERROR(ENOENT); |
34dc7c2f | 647 | |
b128c09f BB |
648 | /* we may be looking for a snapshot */ |
649 | if (err == 0 && snapname != NULL) { | |
fcff0f35 | 650 | dsl_dataset_t *snap_ds; |
34dc7c2f | 651 | |
b128c09f | 652 | if (*snapname++ != '@') { |
b5256303 | 653 | dsl_dataset_rele_flags(ds, flags, tag); |
13fe0198 | 654 | dsl_dir_rele(dd, FTAG); |
2e528b49 | 655 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 656 | } |
34dc7c2f | 657 | |
b128c09f | 658 | dprintf("looking for snapshot '%s'\n", snapname); |
fcff0f35 | 659 | err = dsl_dataset_snap_lookup(ds, snapname, &obj); |
b5256303 TC |
660 | if (err == 0) { |
661 | err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, | |
662 | &snap_ds); | |
663 | } | |
664 | dsl_dataset_rele_flags(ds, flags, tag); | |
b128c09f | 665 | |
13fe0198 | 666 | if (err == 0) { |
fcff0f35 PD |
667 | mutex_enter(&snap_ds->ds_lock); |
668 | if (snap_ds->ds_snapname[0] == 0) | |
669 | (void) strlcpy(snap_ds->ds_snapname, snapname, | |
670 | sizeof (snap_ds->ds_snapname)); | |
671 | mutex_exit(&snap_ds->ds_lock); | |
672 | ds = snap_ds; | |
34dc7c2f | 673 | } |
34dc7c2f | 674 | } |
fcff0f35 PD |
675 | if (err == 0) |
676 | *dsp = ds; | |
13fe0198 | 677 | dsl_dir_rele(dd, FTAG); |
34dc7c2f BB |
678 | return (err); |
679 | } | |
680 | ||
681 | int | |
b5256303 TC |
682 | dsl_dataset_hold(dsl_pool_t *dp, const char *name, void *tag, |
683 | dsl_dataset_t **dsp) | |
684 | { | |
685 | return (dsl_dataset_hold_flags(dp, name, 0, tag, dsp)); | |
686 | } | |
687 | ||
688 | int | |
689 | dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags, | |
428870ff | 690 | void *tag, dsl_dataset_t **dsp) |
34dc7c2f | 691 | { |
b5256303 | 692 | int err = dsl_dataset_hold_obj_flags(dp, dsobj, flags, tag, dsp); |
13fe0198 MA |
693 | if (err != 0) |
694 | return (err); | |
695 | if (!dsl_dataset_tryown(*dsp, tag)) { | |
b5256303 | 696 | dsl_dataset_rele_flags(*dsp, flags, tag); |
13fe0198 | 697 | *dsp = NULL; |
2e528b49 | 698 | return (SET_ERROR(EBUSY)); |
13fe0198 MA |
699 | } |
700 | return (0); | |
701 | } | |
702 | ||
703 | int | |
b5256303 | 704 | dsl_dataset_own(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags, |
13fe0198 MA |
705 | void *tag, dsl_dataset_t **dsp) |
706 | { | |
b5256303 | 707 | int err = dsl_dataset_hold_flags(dp, name, flags, tag, dsp); |
13fe0198 | 708 | if (err != 0) |
b128c09f | 709 | return (err); |
13fe0198 | 710 | if (!dsl_dataset_tryown(*dsp, tag)) { |
b5256303 | 711 | dsl_dataset_rele_flags(*dsp, flags, tag); |
2e528b49 | 712 | return (SET_ERROR(EBUSY)); |
b128c09f BB |
713 | } |
714 | return (0); | |
34dc7c2f BB |
715 | } |
716 | ||
13fe0198 MA |
717 | /* |
718 | * See the comment above dsl_pool_hold() for details. In summary, a long | |
719 | * hold is used to prevent destruction of a dataset while the pool hold | |
720 | * is dropped, allowing other concurrent operations (e.g. spa_sync()). | |
721 | * | |
722 | * The dataset and pool must be held when this function is called. After it | |
723 | * is called, the pool hold may be released while the dataset is still held | |
724 | * and accessed. | |
725 | */ | |
726 | void | |
727 | dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag) | |
728 | { | |
729 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
730 | (void) refcount_add(&ds->ds_longholds, tag); | |
731 | } | |
732 | ||
733 | void | |
734 | dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag) | |
735 | { | |
736 | (void) refcount_remove(&ds->ds_longholds, tag); | |
737 | } | |
738 | ||
739 | /* Return B_TRUE if there are any long holds on this dataset. */ | |
740 | boolean_t | |
741 | dsl_dataset_long_held(dsl_dataset_t *ds) | |
742 | { | |
743 | return (!refcount_is_zero(&ds->ds_longholds)); | |
744 | } | |
745 | ||
34dc7c2f BB |
746 | void |
747 | dsl_dataset_name(dsl_dataset_t *ds, char *name) | |
748 | { | |
749 | if (ds == NULL) { | |
750 | (void) strcpy(name, "mos"); | |
751 | } else { | |
752 | dsl_dir_name(ds->ds_dir, name); | |
13fe0198 | 753 | VERIFY0(dsl_dataset_get_snapname(ds)); |
34dc7c2f | 754 | if (ds->ds_snapname[0]) { |
eca7b760 IK |
755 | VERIFY3U(strlcat(name, "@", ZFS_MAX_DATASET_NAME_LEN), |
756 | <, ZFS_MAX_DATASET_NAME_LEN); | |
b128c09f BB |
757 | /* |
758 | * We use a "recursive" mutex so that we | |
759 | * can call dprintf_ds() with ds_lock held. | |
760 | */ | |
34dc7c2f | 761 | if (!MUTEX_HELD(&ds->ds_lock)) { |
34dc7c2f | 762 | mutex_enter(&ds->ds_lock); |
eca7b760 IK |
763 | VERIFY3U(strlcat(name, ds->ds_snapname, |
764 | ZFS_MAX_DATASET_NAME_LEN), <, | |
765 | ZFS_MAX_DATASET_NAME_LEN); | |
34dc7c2f BB |
766 | mutex_exit(&ds->ds_lock); |
767 | } else { | |
eca7b760 IK |
768 | VERIFY3U(strlcat(name, ds->ds_snapname, |
769 | ZFS_MAX_DATASET_NAME_LEN), <, | |
770 | ZFS_MAX_DATASET_NAME_LEN); | |
34dc7c2f BB |
771 | } |
772 | } | |
773 | } | |
774 | } | |
775 | ||
eca7b760 IK |
776 | int |
777 | dsl_dataset_namelen(dsl_dataset_t *ds) | |
778 | { | |
eca7b760 IK |
779 | VERIFY0(dsl_dataset_get_snapname(ds)); |
780 | mutex_enter(&ds->ds_lock); | |
1c27024e | 781 | int len = strlen(ds->ds_snapname); |
a806cb6a CC |
782 | /* add '@' if ds is a snap */ |
783 | if (len > 0) | |
784 | len++; | |
785 | len += dsl_dir_namelen(ds->ds_dir); | |
eca7b760 IK |
786 | mutex_exit(&ds->ds_lock); |
787 | return (len); | |
788 | } | |
789 | ||
34dc7c2f | 790 | void |
b5256303 | 791 | dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag) |
b128c09f | 792 | { |
b5256303 TC |
793 | if (ds->ds_dir != NULL && ds->ds_dir->dd_crypto_obj != 0 && |
794 | (flags & DS_HOLD_FLAG_DECRYPT)) { | |
795 | (void) spa_keystore_remove_mapping(ds->ds_dir->dd_pool->dp_spa, | |
796 | ds->ds_object, ds); | |
797 | } | |
798 | ||
13fe0198 | 799 | dmu_buf_rele(ds->ds_dbuf, tag); |
b128c09f BB |
800 | } |
801 | ||
802 | void | |
b5256303 TC |
803 | dsl_dataset_rele(dsl_dataset_t *ds, void *tag) |
804 | { | |
805 | dsl_dataset_rele_flags(ds, 0, tag); | |
806 | } | |
807 | ||
808 | void | |
809 | dsl_dataset_disown(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag) | |
34dc7c2f | 810 | { |
945dd935 JG |
811 | ASSERT3P(ds->ds_owner, ==, tag); |
812 | ASSERT(ds->ds_dbuf != NULL); | |
b128c09f | 813 | |
34dc7c2f | 814 | mutex_enter(&ds->ds_lock); |
b128c09f | 815 | ds->ds_owner = NULL; |
34dc7c2f | 816 | mutex_exit(&ds->ds_lock); |
13fe0198 | 817 | dsl_dataset_long_rele(ds, tag); |
b5256303 | 818 | dsl_dataset_rele_flags(ds, flags, tag); |
34dc7c2f BB |
819 | } |
820 | ||
821 | boolean_t | |
13fe0198 | 822 | dsl_dataset_tryown(dsl_dataset_t *ds, void *tag) |
34dc7c2f | 823 | { |
b128c09f BB |
824 | boolean_t gotit = FALSE; |
825 | ||
47dfff3b | 826 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); |
34dc7c2f | 827 | mutex_enter(&ds->ds_lock); |
13fe0198 | 828 | if (ds->ds_owner == NULL && !DS_IS_INCONSISTENT(ds)) { |
428870ff | 829 | ds->ds_owner = tag; |
13fe0198 | 830 | dsl_dataset_long_hold(ds, tag); |
b128c09f | 831 | gotit = TRUE; |
34dc7c2f BB |
832 | } |
833 | mutex_exit(&ds->ds_lock); | |
b128c09f | 834 | return (gotit); |
34dc7c2f BB |
835 | } |
836 | ||
47dfff3b MA |
837 | boolean_t |
838 | dsl_dataset_has_owner(dsl_dataset_t *ds) | |
839 | { | |
840 | boolean_t rv; | |
841 | mutex_enter(&ds->ds_lock); | |
842 | rv = (ds->ds_owner != NULL); | |
843 | mutex_exit(&ds->ds_lock); | |
844 | return (rv); | |
845 | } | |
846 | ||
b5256303 | 847 | void |
241b5415 MA |
848 | dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) |
849 | { | |
850 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
851 | objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; | |
852 | uint64_t zero = 0; | |
853 | ||
854 | VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); | |
855 | ||
856 | spa_feature_incr(spa, f, tx); | |
857 | dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx); | |
858 | ||
859 | VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid, | |
860 | sizeof (zero), 1, &zero, tx)); | |
861 | } | |
862 | ||
863 | void | |
864 | dsl_dataset_deactivate_feature(uint64_t dsobj, spa_feature_t f, dmu_tx_t *tx) | |
865 | { | |
866 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
867 | objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset; | |
868 | ||
869 | VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET); | |
870 | ||
871 | VERIFY0(zap_remove(mos, dsobj, spa_feature_table[f].fi_guid, tx)); | |
872 | spa_feature_decr(spa, f, tx); | |
873 | } | |
874 | ||
34dc7c2f | 875 | uint64_t |
b128c09f | 876 | dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin, |
b5256303 | 877 | dsl_crypto_params_t *dcp, uint64_t flags, dmu_tx_t *tx) |
34dc7c2f BB |
878 | { |
879 | dsl_pool_t *dp = dd->dd_pool; | |
880 | dmu_buf_t *dbuf; | |
881 | dsl_dataset_phys_t *dsphys; | |
882 | uint64_t dsobj; | |
883 | objset_t *mos = dp->dp_meta_objset; | |
884 | ||
b128c09f BB |
885 | if (origin == NULL) |
886 | origin = dp->dp_origin_snap; | |
887 | ||
34dc7c2f | 888 | ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp); |
d683ddbb | 889 | ASSERT(origin == NULL || dsl_dataset_phys(origin)->ds_num_children > 0); |
34dc7c2f | 890 | ASSERT(dmu_tx_is_syncing(tx)); |
d683ddbb | 891 | ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0); |
34dc7c2f BB |
892 | |
893 | dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, | |
894 | DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); | |
13fe0198 | 895 | VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); |
34dc7c2f BB |
896 | dmu_buf_will_dirty(dbuf, tx); |
897 | dsphys = dbuf->db_data; | |
b128c09f | 898 | bzero(dsphys, sizeof (dsl_dataset_phys_t)); |
34dc7c2f BB |
899 | dsphys->ds_dir_obj = dd->dd_object; |
900 | dsphys->ds_flags = flags; | |
901 | dsphys->ds_fsid_guid = unique_create(); | |
902 | (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, | |
903 | sizeof (dsphys->ds_guid)); | |
904 | dsphys->ds_snapnames_zapobj = | |
905 | zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP, | |
906 | DMU_OT_NONE, 0, tx); | |
907 | dsphys->ds_creation_time = gethrestime_sec(); | |
b128c09f | 908 | dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg; |
34dc7c2f | 909 | |
428870ff BB |
910 | if (origin == NULL) { |
911 | dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx); | |
912 | } else { | |
13fe0198 | 913 | dsl_dataset_t *ohds; /* head of the origin snapshot */ |
428870ff | 914 | |
34dc7c2f BB |
915 | dsphys->ds_prev_snap_obj = origin->ds_object; |
916 | dsphys->ds_prev_snap_txg = | |
d683ddbb | 917 | dsl_dataset_phys(origin)->ds_creation_txg; |
9ae529ec | 918 | dsphys->ds_referenced_bytes = |
d683ddbb | 919 | dsl_dataset_phys(origin)->ds_referenced_bytes; |
34dc7c2f | 920 | dsphys->ds_compressed_bytes = |
d683ddbb | 921 | dsl_dataset_phys(origin)->ds_compressed_bytes; |
34dc7c2f | 922 | dsphys->ds_uncompressed_bytes = |
d683ddbb | 923 | dsl_dataset_phys(origin)->ds_uncompressed_bytes; |
cc9bb3e5 | 924 | rrw_enter(&origin->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 925 | dsphys->ds_bp = dsl_dataset_phys(origin)->ds_bp; |
cc9bb3e5 | 926 | rrw_exit(&origin->ds_bp_rwlock, FTAG); |
4b20a6f5 MA |
927 | |
928 | /* | |
929 | * Inherit flags that describe the dataset's contents | |
930 | * (INCONSISTENT) or properties (Case Insensitive). | |
931 | */ | |
d683ddbb | 932 | dsphys->ds_flags |= dsl_dataset_phys(origin)->ds_flags & |
4b20a6f5 | 933 | (DS_FLAG_INCONSISTENT | DS_FLAG_CI_DATASET); |
34dc7c2f | 934 | |
1c27024e | 935 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
936 | if (origin->ds_feature_inuse[f]) |
937 | dsl_dataset_activate_feature(dsobj, f, tx); | |
938 | } | |
f1512ee6 | 939 | |
34dc7c2f | 940 | dmu_buf_will_dirty(origin->ds_dbuf, tx); |
d683ddbb | 941 | dsl_dataset_phys(origin)->ds_num_children++; |
34dc7c2f | 942 | |
13fe0198 | 943 | VERIFY0(dsl_dataset_hold_obj(dp, |
d683ddbb JG |
944 | dsl_dir_phys(origin->ds_dir)->dd_head_dataset_obj, |
945 | FTAG, &ohds)); | |
428870ff BB |
946 | dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist, |
947 | dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx); | |
948 | dsl_dataset_rele(ohds, FTAG); | |
949 | ||
b128c09f | 950 | if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) { |
d683ddbb JG |
951 | if (dsl_dataset_phys(origin)->ds_next_clones_obj == 0) { |
952 | dsl_dataset_phys(origin)->ds_next_clones_obj = | |
b128c09f BB |
953 | zap_create(mos, |
954 | DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx); | |
955 | } | |
13fe0198 | 956 | VERIFY0(zap_add_int(mos, |
d683ddbb JG |
957 | dsl_dataset_phys(origin)->ds_next_clones_obj, |
958 | dsobj, tx)); | |
b128c09f BB |
959 | } |
960 | ||
34dc7c2f | 961 | dmu_buf_will_dirty(dd->dd_dbuf, tx); |
d683ddbb | 962 | dsl_dir_phys(dd)->dd_origin_obj = origin->ds_object; |
428870ff | 963 | if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { |
d683ddbb | 964 | if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) { |
428870ff | 965 | dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx); |
d683ddbb | 966 | dsl_dir_phys(origin->ds_dir)->dd_clones = |
428870ff BB |
967 | zap_create(mos, |
968 | DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx); | |
969 | } | |
13fe0198 | 970 | VERIFY0(zap_add_int(mos, |
d683ddbb JG |
971 | dsl_dir_phys(origin->ds_dir)->dd_clones, |
972 | dsobj, tx)); | |
428870ff | 973 | } |
34dc7c2f BB |
974 | } |
975 | ||
b5256303 TC |
976 | /* handle encryption */ |
977 | dsl_dataset_create_crypt_sync(dsobj, dd, origin, dcp, tx); | |
978 | ||
34dc7c2f BB |
979 | if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) |
980 | dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; | |
981 | ||
982 | dmu_buf_rele(dbuf, FTAG); | |
983 | ||
984 | dmu_buf_will_dirty(dd->dd_dbuf, tx); | |
d683ddbb | 985 | dsl_dir_phys(dd)->dd_head_dataset_obj = dsobj; |
34dc7c2f BB |
986 | |
987 | return (dsobj); | |
988 | } | |
989 | ||
13fe0198 MA |
990 | static void |
991 | dsl_dataset_zero_zil(dsl_dataset_t *ds, dmu_tx_t *tx) | |
992 | { | |
993 | objset_t *os; | |
994 | ||
995 | VERIFY0(dmu_objset_from_ds(ds, &os)); | |
0efd9791 AG |
996 | if (bcmp(&os->os_zil_header, &zero_zil, sizeof (zero_zil)) != 0) { |
997 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
998 | zio_t *zio; | |
999 | ||
1000 | bzero(&os->os_zil_header, sizeof (os->os_zil_header)); | |
b5256303 | 1001 | if (os->os_encrypted) |
1b66810b | 1002 | os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE; |
0efd9791 AG |
1003 | |
1004 | zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); | |
1005 | dsl_dataset_sync(ds, zio, tx); | |
1006 | VERIFY0(zio_wait(zio)); | |
1007 | ||
1008 | /* dsl_dataset_sync_done will drop this reference. */ | |
1009 | dmu_buf_add_ref(ds->ds_dbuf, ds); | |
1010 | dsl_dataset_sync_done(ds, tx); | |
1011 | } | |
13fe0198 MA |
1012 | } |
1013 | ||
34dc7c2f BB |
1014 | uint64_t |
1015 | dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname, | |
b5256303 TC |
1016 | dsl_dataset_t *origin, uint64_t flags, cred_t *cr, |
1017 | dsl_crypto_params_t *dcp, dmu_tx_t *tx) | |
34dc7c2f BB |
1018 | { |
1019 | dsl_pool_t *dp = pdd->dd_pool; | |
1020 | uint64_t dsobj, ddobj; | |
1021 | dsl_dir_t *dd; | |
1022 | ||
13fe0198 | 1023 | ASSERT(dmu_tx_is_syncing(tx)); |
34dc7c2f BB |
1024 | ASSERT(lastname[0] != '@'); |
1025 | ||
b128c09f | 1026 | ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx); |
13fe0198 | 1027 | VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd)); |
34dc7c2f | 1028 | |
b5256303 | 1029 | dsobj = dsl_dataset_create_sync_dd(dd, origin, dcp, |
13fe0198 | 1030 | flags & ~DS_CREATE_FLAG_NODIRTY, tx); |
34dc7c2f BB |
1031 | |
1032 | dsl_deleg_set_create_perms(dd, tx, cr); | |
1033 | ||
788eb90c JJ |
1034 | /* |
1035 | * Since we're creating a new node we know it's a leaf, so we can | |
1036 | * initialize the counts if the limit feature is active. | |
1037 | */ | |
1038 | if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) { | |
1039 | uint64_t cnt = 0; | |
1040 | objset_t *os = dd->dd_pool->dp_meta_objset; | |
1041 | ||
1042 | dsl_dir_zapify(dd, tx); | |
1043 | VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT, | |
1044 | sizeof (cnt), 1, &cnt, tx)); | |
1045 | VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT, | |
1046 | sizeof (cnt), 1, &cnt, tx)); | |
1047 | } | |
1048 | ||
13fe0198 | 1049 | dsl_dir_rele(dd, FTAG); |
34dc7c2f | 1050 | |
572e2857 BB |
1051 | /* |
1052 | * If we are creating a clone, make sure we zero out any stale | |
1053 | * data from the origin snapshots zil header. | |
1054 | */ | |
13fe0198 | 1055 | if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) { |
572e2857 | 1056 | dsl_dataset_t *ds; |
572e2857 | 1057 | |
13fe0198 MA |
1058 | VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds)); |
1059 | dsl_dataset_zero_zil(ds, tx); | |
572e2857 BB |
1060 | dsl_dataset_rele(ds, FTAG); |
1061 | } | |
1062 | ||
34dc7c2f BB |
1063 | return (dsobj); |
1064 | } | |
1065 | ||
34dc7c2f | 1066 | /* |
13fe0198 MA |
1067 | * The unique space in the head dataset can be calculated by subtracting |
1068 | * the space used in the most recent snapshot, that is still being used | |
1069 | * in this file system, from the space currently in use. To figure out | |
1070 | * the space in the most recent snapshot still in use, we need to take | |
1071 | * the total space used in the snapshot and subtract out the space that | |
1072 | * has been freed up since the snapshot was taken. | |
34dc7c2f | 1073 | */ |
13fe0198 MA |
1074 | void |
1075 | dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds) | |
34dc7c2f | 1076 | { |
13fe0198 MA |
1077 | uint64_t mrs_used; |
1078 | uint64_t dlused, dlcomp, dluncomp; | |
34dc7c2f | 1079 | |
0c66c32d | 1080 | ASSERT(!ds->ds_is_snapshot); |
34dc7c2f | 1081 | |
d683ddbb JG |
1082 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) |
1083 | mrs_used = dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes; | |
13fe0198 MA |
1084 | else |
1085 | mrs_used = 0; | |
45d1cae3 | 1086 | |
13fe0198 | 1087 | dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp); |
34dc7c2f | 1088 | |
13fe0198 | 1089 | ASSERT3U(dlused, <=, mrs_used); |
d683ddbb JG |
1090 | dsl_dataset_phys(ds)->ds_unique_bytes = |
1091 | dsl_dataset_phys(ds)->ds_referenced_bytes - (mrs_used - dlused); | |
330d06f9 | 1092 | |
13fe0198 MA |
1093 | if (spa_version(ds->ds_dir->dd_pool->dp_spa) >= |
1094 | SPA_VERSION_UNIQUE_ACCURATE) | |
d683ddbb | 1095 | dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; |
34dc7c2f BB |
1096 | } |
1097 | ||
13fe0198 MA |
1098 | void |
1099 | dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj, | |
1100 | dmu_tx_t *tx) | |
45d1cae3 | 1101 | { |
13fe0198 | 1102 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; |
13fe0198 | 1103 | ASSERTV(uint64_t count); |
1c27024e | 1104 | int err; |
13fe0198 | 1105 | |
d683ddbb JG |
1106 | ASSERT(dsl_dataset_phys(ds)->ds_num_children >= 2); |
1107 | err = zap_remove_int(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, | |
1108 | obj, tx); | |
13fe0198 MA |
1109 | /* |
1110 | * The err should not be ENOENT, but a bug in a previous version | |
1111 | * of the code could cause upgrade_clones_cb() to not set | |
1112 | * ds_next_snap_obj when it should, leading to a missing entry. | |
1113 | * If we knew that the pool was created after | |
1114 | * SPA_VERSION_NEXT_CLONES, we could assert that it isn't | |
1115 | * ENOENT. However, at least we can check that we don't have | |
1116 | * too many entries in the next_clones_obj even after failing to | |
1117 | * remove this one. | |
1118 | */ | |
1119 | if (err != ENOENT) | |
1120 | VERIFY0(err); | |
d683ddbb | 1121 | ASSERT0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, |
13fe0198 | 1122 | &count)); |
d683ddbb | 1123 | ASSERT3U(count, <=, dsl_dataset_phys(ds)->ds_num_children - 2); |
13fe0198 | 1124 | } |
45d1cae3 | 1125 | |
45d1cae3 | 1126 | |
13fe0198 MA |
1127 | blkptr_t * |
1128 | dsl_dataset_get_blkptr(dsl_dataset_t *ds) | |
1129 | { | |
d683ddbb | 1130 | return (&dsl_dataset_phys(ds)->ds_bp); |
45d1cae3 BB |
1131 | } |
1132 | ||
13fe0198 MA |
1133 | spa_t * |
1134 | dsl_dataset_get_spa(dsl_dataset_t *ds) | |
1135 | { | |
1136 | return (ds->ds_dir->dd_pool->dp_spa); | |
45d1cae3 BB |
1137 | } |
1138 | ||
13fe0198 MA |
1139 | void |
1140 | dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx) | |
34dc7c2f | 1141 | { |
13fe0198 | 1142 | dsl_pool_t *dp; |
45d1cae3 | 1143 | |
13fe0198 MA |
1144 | if (ds == NULL) /* this is the meta-objset */ |
1145 | return; | |
34dc7c2f | 1146 | |
428870ff | 1147 | ASSERT(ds->ds_objset != NULL); |
34dc7c2f | 1148 | |
d683ddbb | 1149 | if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) |
34dc7c2f BB |
1150 | panic("dirtying snapshot!"); |
1151 | ||
0efd9791 AG |
1152 | /* Must not dirty a dataset in the same txg where it got snapshotted. */ |
1153 | ASSERT3U(tx->tx_txg, >, dsl_dataset_phys(ds)->ds_prev_snap_txg); | |
34dc7c2f | 1154 | |
0efd9791 | 1155 | dp = ds->ds_dir->dd_pool; |
13fe0198 | 1156 | if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) { |
34dc7c2f BB |
1157 | /* up the hold count until we can be written out */ |
1158 | dmu_buf_add_ref(ds->ds_dbuf, ds); | |
1159 | } | |
1160 | } | |
1161 | ||
34dc7c2f | 1162 | static int |
13fe0198 | 1163 | dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx) |
34dc7c2f | 1164 | { |
13fe0198 | 1165 | uint64_t asize; |
34dc7c2f | 1166 | |
13fe0198 | 1167 | if (!dmu_tx_is_syncing(tx)) |
b128c09f BB |
1168 | return (0); |
1169 | ||
34dc7c2f | 1170 | /* |
13fe0198 MA |
1171 | * If there's an fs-only reservation, any blocks that might become |
1172 | * owned by the snapshot dataset must be accommodated by space | |
1173 | * outside of the reservation. | |
34dc7c2f | 1174 | */ |
13fe0198 | 1175 | ASSERT(ds->ds_reserved == 0 || DS_UNIQUE_IS_ACCURATE(ds)); |
d683ddbb | 1176 | asize = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved); |
13fe0198 | 1177 | if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) |
2e528b49 | 1178 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
1179 | |
1180 | /* | |
13fe0198 MA |
1181 | * Propagate any reserved space for this snapshot to other |
1182 | * snapshot checks in this sync group. | |
34dc7c2f | 1183 | */ |
13fe0198 MA |
1184 | if (asize > 0) |
1185 | dsl_dir_willuse_space(ds->ds_dir, asize, tx); | |
34dc7c2f BB |
1186 | |
1187 | return (0); | |
1188 | } | |
1189 | ||
34dc7c2f | 1190 | int |
13fe0198 | 1191 | dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname, |
788eb90c | 1192 | dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr) |
34dc7c2f | 1193 | { |
13fe0198 MA |
1194 | int error; |
1195 | uint64_t value; | |
34dc7c2f | 1196 | |
13fe0198 | 1197 | ds->ds_trysnap_txg = tx->tx_txg; |
b128c09f | 1198 | |
13fe0198 | 1199 | if (!dmu_tx_is_syncing(tx)) |
45d1cae3 | 1200 | return (0); |
34dc7c2f BB |
1201 | |
1202 | /* | |
13fe0198 MA |
1203 | * We don't allow multiple snapshots of the same txg. If there |
1204 | * is already one, try again. | |
34dc7c2f | 1205 | */ |
d683ddbb | 1206 | if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) |
2e528b49 | 1207 | return (SET_ERROR(EAGAIN)); |
34dc7c2f BB |
1208 | |
1209 | /* | |
13fe0198 | 1210 | * Check for conflicting snapshot name. |
34dc7c2f | 1211 | */ |
13fe0198 MA |
1212 | error = dsl_dataset_snap_lookup(ds, snapname, &value); |
1213 | if (error == 0) | |
2e528b49 | 1214 | return (SET_ERROR(EEXIST)); |
13fe0198 MA |
1215 | if (error != ENOENT) |
1216 | return (error); | |
45d1cae3 | 1217 | |
96c2e961 KW |
1218 | /* |
1219 | * We don't allow taking snapshots of inconsistent datasets, such as | |
1220 | * those into which we are currently receiving. However, if we are | |
1221 | * creating this snapshot as part of a receive, this check will be | |
1222 | * executed atomically with respect to the completion of the receive | |
1223 | * itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this | |
1224 | * case we ignore this, knowing it will be fixed up for us shortly in | |
1225 | * dmu_recv_end_sync(). | |
1226 | */ | |
1227 | if (!recv && DS_IS_INCONSISTENT(ds)) | |
1228 | return (SET_ERROR(EBUSY)); | |
1229 | ||
788eb90c JJ |
1230 | /* |
1231 | * Skip the check for temporary snapshots or if we have already checked | |
1232 | * the counts in dsl_dataset_snapshot_check. This means we really only | |
1233 | * check the count here when we're receiving a stream. | |
1234 | */ | |
1235 | if (cnt != 0 && cr != NULL) { | |
1236 | error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, | |
1237 | ZFS_PROP_SNAPSHOT_LIMIT, NULL, cr); | |
1238 | if (error != 0) | |
1239 | return (error); | |
1240 | } | |
1241 | ||
13fe0198 MA |
1242 | error = dsl_dataset_snapshot_reserve_space(ds, tx); |
1243 | if (error != 0) | |
1244 | return (error); | |
45d1cae3 | 1245 | |
34dc7c2f BB |
1246 | return (0); |
1247 | } | |
1248 | ||
234c91c5 | 1249 | int |
13fe0198 | 1250 | dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx) |
b128c09f | 1251 | { |
13fe0198 MA |
1252 | dsl_dataset_snapshot_arg_t *ddsa = arg; |
1253 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1254 | nvpair_t *pair; | |
1255 | int rv = 0; | |
b128c09f | 1256 | |
788eb90c JJ |
1257 | /* |
1258 | * Pre-compute how many total new snapshots will be created for each | |
1259 | * level in the tree and below. This is needed for validating the | |
1260 | * snapshot limit when either taking a recursive snapshot or when | |
1261 | * taking multiple snapshots. | |
1262 | * | |
1263 | * The problem is that the counts are not actually adjusted when | |
1264 | * we are checking, only when we finally sync. For a single snapshot, | |
1265 | * this is easy, the count will increase by 1 at each node up the tree, | |
1266 | * but its more complicated for the recursive/multiple snapshot case. | |
1267 | * | |
1268 | * The dsl_fs_ss_limit_check function does recursively check the count | |
1269 | * at each level up the tree but since it is validating each snapshot | |
1270 | * independently we need to be sure that we are validating the complete | |
1271 | * count for the entire set of snapshots. We do this by rolling up the | |
1272 | * counts for each component of the name into an nvlist and then | |
1273 | * checking each of those cases with the aggregated count. | |
1274 | * | |
1275 | * This approach properly handles not only the recursive snapshot | |
1276 | * case (where we get all of those on the ddsa_snaps list) but also | |
1277 | * the sibling case (e.g. snapshot a/b and a/c so that we will also | |
1278 | * validate the limit on 'a' using a count of 2). | |
1279 | * | |
1280 | * We validate the snapshot names in the third loop and only report | |
1281 | * name errors once. | |
1282 | */ | |
1283 | if (dmu_tx_is_syncing(tx)) { | |
1284 | char *nm; | |
1285 | nvlist_t *cnt_track = NULL; | |
1286 | cnt_track = fnvlist_alloc(); | |
1287 | ||
1288 | nm = kmem_alloc(MAXPATHLEN, KM_SLEEP); | |
1289 | ||
1290 | /* Rollup aggregated counts into the cnt_track list */ | |
1291 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); | |
1292 | pair != NULL; | |
1293 | pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1294 | char *pdelim; | |
1295 | uint64_t val; | |
1296 | ||
1297 | (void) strlcpy(nm, nvpair_name(pair), MAXPATHLEN); | |
1298 | pdelim = strchr(nm, '@'); | |
1299 | if (pdelim == NULL) | |
1300 | continue; | |
1301 | *pdelim = '\0'; | |
1302 | ||
1303 | do { | |
1304 | if (nvlist_lookup_uint64(cnt_track, nm, | |
1305 | &val) == 0) { | |
1306 | /* update existing entry */ | |
1307 | fnvlist_add_uint64(cnt_track, nm, | |
1308 | val + 1); | |
1309 | } else { | |
1310 | /* add to list */ | |
1311 | fnvlist_add_uint64(cnt_track, nm, 1); | |
1312 | } | |
1313 | ||
1314 | pdelim = strrchr(nm, '/'); | |
1315 | if (pdelim != NULL) | |
1316 | *pdelim = '\0'; | |
1317 | } while (pdelim != NULL); | |
1318 | } | |
1319 | ||
1320 | kmem_free(nm, MAXPATHLEN); | |
1321 | ||
1322 | /* Check aggregated counts at each level */ | |
1323 | for (pair = nvlist_next_nvpair(cnt_track, NULL); | |
1324 | pair != NULL; pair = nvlist_next_nvpair(cnt_track, pair)) { | |
1325 | int error = 0; | |
1326 | char *name; | |
1327 | uint64_t cnt = 0; | |
1328 | dsl_dataset_t *ds; | |
1329 | ||
1330 | name = nvpair_name(pair); | |
1331 | cnt = fnvpair_value_uint64(pair); | |
1332 | ASSERT(cnt > 0); | |
1333 | ||
1334 | error = dsl_dataset_hold(dp, name, FTAG, &ds); | |
1335 | if (error == 0) { | |
1336 | error = dsl_fs_ss_limit_check(ds->ds_dir, cnt, | |
1337 | ZFS_PROP_SNAPSHOT_LIMIT, NULL, | |
1338 | ddsa->ddsa_cr); | |
1339 | dsl_dataset_rele(ds, FTAG); | |
1340 | } | |
1341 | ||
1342 | if (error != 0) { | |
1343 | if (ddsa->ddsa_errors != NULL) | |
1344 | fnvlist_add_int32(ddsa->ddsa_errors, | |
1345 | name, error); | |
1346 | rv = error; | |
1347 | /* only report one error for this check */ | |
1348 | break; | |
1349 | } | |
1350 | } | |
1351 | nvlist_free(cnt_track); | |
1352 | } | |
1353 | ||
13fe0198 MA |
1354 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); |
1355 | pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1356 | int error = 0; | |
1357 | dsl_dataset_t *ds; | |
689f093e | 1358 | char *name, *atp = NULL; |
eca7b760 | 1359 | char dsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1360 | |
1361 | name = nvpair_name(pair); | |
eca7b760 | 1362 | if (strlen(name) >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 1363 | error = SET_ERROR(ENAMETOOLONG); |
13fe0198 MA |
1364 | if (error == 0) { |
1365 | atp = strchr(name, '@'); | |
1366 | if (atp == NULL) | |
2e528b49 | 1367 | error = SET_ERROR(EINVAL); |
13fe0198 MA |
1368 | if (error == 0) |
1369 | (void) strlcpy(dsname, name, atp - name + 1); | |
1370 | } | |
1371 | if (error == 0) | |
1372 | error = dsl_dataset_hold(dp, dsname, FTAG, &ds); | |
1373 | if (error == 0) { | |
788eb90c | 1374 | /* passing 0/NULL skips dsl_fs_ss_limit_check */ |
13fe0198 | 1375 | error = dsl_dataset_snapshot_check_impl(ds, |
788eb90c | 1376 | atp + 1, tx, B_FALSE, 0, NULL); |
13fe0198 MA |
1377 | dsl_dataset_rele(ds, FTAG); |
1378 | } | |
b128c09f | 1379 | |
13fe0198 MA |
1380 | if (error != 0) { |
1381 | if (ddsa->ddsa_errors != NULL) { | |
1382 | fnvlist_add_int32(ddsa->ddsa_errors, | |
1383 | name, error); | |
1384 | } | |
1385 | rv = error; | |
1386 | } | |
1387 | } | |
788eb90c | 1388 | |
13fe0198 | 1389 | return (rv); |
b128c09f BB |
1390 | } |
1391 | ||
13fe0198 MA |
1392 | void |
1393 | dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname, | |
1394 | dmu_tx_t *tx) | |
428870ff | 1395 | { |
13fe0198 MA |
1396 | dsl_pool_t *dp = ds->ds_dir->dd_pool; |
1397 | dmu_buf_t *dbuf; | |
1398 | dsl_dataset_phys_t *dsphys; | |
1399 | uint64_t dsobj, crtxg; | |
1400 | objset_t *mos = dp->dp_meta_objset; | |
1401 | ASSERTV(static zil_header_t zero_zil); | |
1402 | ASSERTV(objset_t *os); | |
1403 | ||
1404 | ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock)); | |
428870ff | 1405 | |
428870ff | 1406 | /* |
13fe0198 MA |
1407 | * If we are on an old pool, the zil must not be active, in which |
1408 | * case it will be zeroed. Usually zil_suspend() accomplishes this. | |
428870ff | 1409 | */ |
13fe0198 MA |
1410 | ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP || |
1411 | dmu_objset_from_ds(ds, &os) != 0 || | |
1412 | bcmp(&os->os_phys->os_zil_header, &zero_zil, | |
1413 | sizeof (zero_zil)) == 0); | |
428870ff | 1414 | |
0efd9791 AG |
1415 | /* Should not snapshot a dirty dataset. */ |
1416 | ASSERT(!txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets, | |
1417 | ds, tx->tx_txg)); | |
1418 | ||
788eb90c | 1419 | dsl_fs_ss_count_adjust(ds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx); |
428870ff BB |
1420 | |
1421 | /* | |
13fe0198 | 1422 | * The origin's ds_creation_txg has to be < TXG_INITIAL |
b128c09f BB |
1423 | */ |
1424 | if (strcmp(snapname, ORIGIN_DIR_NAME) == 0) | |
1425 | crtxg = 1; | |
1426 | else | |
1427 | crtxg = tx->tx_txg; | |
1428 | ||
34dc7c2f BB |
1429 | dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, |
1430 | DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); | |
13fe0198 | 1431 | VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf)); |
34dc7c2f BB |
1432 | dmu_buf_will_dirty(dbuf, tx); |
1433 | dsphys = dbuf->db_data; | |
b128c09f | 1434 | bzero(dsphys, sizeof (dsl_dataset_phys_t)); |
34dc7c2f BB |
1435 | dsphys->ds_dir_obj = ds->ds_dir->dd_object; |
1436 | dsphys->ds_fsid_guid = unique_create(); | |
1437 | (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid, | |
1438 | sizeof (dsphys->ds_guid)); | |
d683ddbb JG |
1439 | dsphys->ds_prev_snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
1440 | dsphys->ds_prev_snap_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg; | |
34dc7c2f BB |
1441 | dsphys->ds_next_snap_obj = ds->ds_object; |
1442 | dsphys->ds_num_children = 1; | |
1443 | dsphys->ds_creation_time = gethrestime_sec(); | |
b128c09f | 1444 | dsphys->ds_creation_txg = crtxg; |
d683ddbb JG |
1445 | dsphys->ds_deadlist_obj = dsl_dataset_phys(ds)->ds_deadlist_obj; |
1446 | dsphys->ds_referenced_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes; | |
1447 | dsphys->ds_compressed_bytes = dsl_dataset_phys(ds)->ds_compressed_bytes; | |
1448 | dsphys->ds_uncompressed_bytes = | |
1449 | dsl_dataset_phys(ds)->ds_uncompressed_bytes; | |
1450 | dsphys->ds_flags = dsl_dataset_phys(ds)->ds_flags; | |
cc9bb3e5 | 1451 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 1452 | dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp; |
cc9bb3e5 | 1453 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f BB |
1454 | dmu_buf_rele(dbuf, FTAG); |
1455 | ||
1c27024e | 1456 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
1457 | if (ds->ds_feature_inuse[f]) |
1458 | dsl_dataset_activate_feature(dsobj, f, tx); | |
1459 | } | |
f1512ee6 | 1460 | |
d683ddbb JG |
1461 | ASSERT3U(ds->ds_prev != 0, ==, |
1462 | dsl_dataset_phys(ds)->ds_prev_snap_obj != 0); | |
34dc7c2f | 1463 | if (ds->ds_prev) { |
b128c09f | 1464 | uint64_t next_clones_obj = |
d683ddbb JG |
1465 | dsl_dataset_phys(ds->ds_prev)->ds_next_clones_obj; |
1466 | ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == | |
34dc7c2f | 1467 | ds->ds_object || |
d683ddbb JG |
1468 | dsl_dataset_phys(ds->ds_prev)->ds_num_children > 1); |
1469 | if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == | |
1470 | ds->ds_object) { | |
34dc7c2f | 1471 | dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); |
d683ddbb JG |
1472 | ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, ==, |
1473 | dsl_dataset_phys(ds->ds_prev)->ds_creation_txg); | |
1474 | dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj = dsobj; | |
b128c09f | 1475 | } else if (next_clones_obj != 0) { |
13fe0198 | 1476 | dsl_dataset_remove_from_next_clones(ds->ds_prev, |
428870ff | 1477 | dsphys->ds_next_snap_obj, tx); |
13fe0198 | 1478 | VERIFY0(zap_add_int(mos, |
b128c09f | 1479 | next_clones_obj, dsobj, tx)); |
34dc7c2f BB |
1480 | } |
1481 | } | |
1482 | ||
1483 | /* | |
1484 | * If we have a reference-reservation on this dataset, we will | |
1485 | * need to increase the amount of refreservation being charged | |
1486 | * since our unique space is going to zero. | |
1487 | */ | |
1488 | if (ds->ds_reserved) { | |
428870ff BB |
1489 | int64_t delta; |
1490 | ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); | |
d683ddbb JG |
1491 | delta = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, |
1492 | ds->ds_reserved); | |
b128c09f | 1493 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, |
428870ff | 1494 | delta, 0, 0, tx); |
34dc7c2f BB |
1495 | } |
1496 | ||
34dc7c2f | 1497 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
d683ddbb JG |
1498 | dsl_dataset_phys(ds)->ds_deadlist_obj = |
1499 | dsl_deadlist_clone(&ds->ds_deadlist, UINT64_MAX, | |
1500 | dsl_dataset_phys(ds)->ds_prev_snap_obj, tx); | |
428870ff | 1501 | dsl_deadlist_close(&ds->ds_deadlist); |
d683ddbb JG |
1502 | dsl_deadlist_open(&ds->ds_deadlist, mos, |
1503 | dsl_dataset_phys(ds)->ds_deadlist_obj); | |
428870ff | 1504 | dsl_deadlist_add_key(&ds->ds_deadlist, |
d683ddbb | 1505 | dsl_dataset_phys(ds)->ds_prev_snap_txg, tx); |
428870ff | 1506 | |
a1d477c2 MA |
1507 | if (dsl_dataset_remap_deadlist_exists(ds)) { |
1508 | uint64_t remap_deadlist_obj = | |
1509 | dsl_dataset_get_remap_deadlist_object(ds); | |
1510 | /* | |
1511 | * Move the remap_deadlist to the snapshot. The head | |
1512 | * will create a new remap deadlist on demand, from | |
1513 | * dsl_dataset_block_remapped(). | |
1514 | */ | |
1515 | dsl_dataset_unset_remap_deadlist_object(ds, tx); | |
1516 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
1517 | ||
1518 | dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx); | |
1519 | VERIFY0(zap_add(mos, dsobj, DS_FIELD_REMAP_DEADLIST, | |
1520 | sizeof (remap_deadlist_obj), 1, &remap_deadlist_obj, tx)); | |
1521 | } | |
1522 | ||
d683ddbb JG |
1523 | ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg); |
1524 | dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj; | |
1525 | dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg; | |
1526 | dsl_dataset_phys(ds)->ds_unique_bytes = 0; | |
a1d477c2 | 1527 | |
34dc7c2f | 1528 | if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE) |
d683ddbb | 1529 | dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE; |
34dc7c2f | 1530 | |
d683ddbb | 1531 | VERIFY0(zap_add(mos, dsl_dataset_phys(ds)->ds_snapnames_zapobj, |
13fe0198 | 1532 | snapname, 8, 1, &dsobj, tx)); |
34dc7c2f BB |
1533 | |
1534 | if (ds->ds_prev) | |
13fe0198 MA |
1535 | dsl_dataset_rele(ds->ds_prev, ds); |
1536 | VERIFY0(dsl_dataset_hold_obj(dp, | |
d683ddbb | 1537 | dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev)); |
b128c09f | 1538 | |
428870ff BB |
1539 | dsl_scan_ds_snapshotted(ds, tx); |
1540 | ||
1541 | dsl_dir_snap_cmtime_update(ds->ds_dir); | |
34dc7c2f | 1542 | |
6f1ffb06 | 1543 | spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, ""); |
34dc7c2f BB |
1544 | } |
1545 | ||
234c91c5 | 1546 | void |
13fe0198 | 1547 | dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 1548 | { |
13fe0198 MA |
1549 | dsl_dataset_snapshot_arg_t *ddsa = arg; |
1550 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1551 | nvpair_t *pair; | |
34dc7c2f | 1552 | |
13fe0198 MA |
1553 | for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL); |
1554 | pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) { | |
1555 | dsl_dataset_t *ds; | |
1556 | char *name, *atp; | |
eca7b760 | 1557 | char dsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1558 | |
1559 | name = nvpair_name(pair); | |
1560 | atp = strchr(name, '@'); | |
1561 | (void) strlcpy(dsname, name, atp - name + 1); | |
1562 | VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds)); | |
1563 | ||
1564 | dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx); | |
1565 | if (ddsa->ddsa_props != NULL) { | |
1566 | dsl_props_set_sync_impl(ds->ds_prev, | |
1567 | ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx); | |
1568 | } | |
a0bd735a | 1569 | zvol_create_minors(dp->dp_spa, nvpair_name(pair), B_TRUE); |
13fe0198 MA |
1570 | dsl_dataset_rele(ds, FTAG); |
1571 | } | |
34dc7c2f BB |
1572 | } |
1573 | ||
13fe0198 MA |
1574 | /* |
1575 | * The snapshots must all be in the same pool. | |
1576 | * All-or-nothing: if there are any failures, nothing will be modified. | |
1577 | */ | |
1578 | int | |
1579 | dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors) | |
330d06f9 | 1580 | { |
13fe0198 MA |
1581 | dsl_dataset_snapshot_arg_t ddsa; |
1582 | nvpair_t *pair; | |
1583 | boolean_t needsuspend; | |
1584 | int error; | |
1585 | spa_t *spa; | |
1586 | char *firstname; | |
1587 | nvlist_t *suspended = NULL; | |
330d06f9 | 1588 | |
13fe0198 MA |
1589 | pair = nvlist_next_nvpair(snaps, NULL); |
1590 | if (pair == NULL) | |
1591 | return (0); | |
1592 | firstname = nvpair_name(pair); | |
1593 | ||
1594 | error = spa_open(firstname, &spa, FTAG); | |
1595 | if (error != 0) | |
1596 | return (error); | |
1597 | needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); | |
1598 | spa_close(spa, FTAG); | |
1599 | ||
1600 | if (needsuspend) { | |
1601 | suspended = fnvlist_alloc(); | |
1602 | for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; | |
1603 | pair = nvlist_next_nvpair(snaps, pair)) { | |
eca7b760 | 1604 | char fsname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1605 | char *snapname = nvpair_name(pair); |
1606 | char *atp; | |
1607 | void *cookie; | |
1608 | ||
1609 | atp = strchr(snapname, '@'); | |
1610 | if (atp == NULL) { | |
2e528b49 | 1611 | error = SET_ERROR(EINVAL); |
13fe0198 MA |
1612 | break; |
1613 | } | |
1614 | (void) strlcpy(fsname, snapname, atp - snapname + 1); | |
1615 | ||
1616 | error = zil_suspend(fsname, &cookie); | |
1617 | if (error != 0) | |
1618 | break; | |
1619 | fnvlist_add_uint64(suspended, fsname, | |
1620 | (uintptr_t)cookie); | |
1621 | } | |
1622 | } | |
1623 | ||
1624 | ddsa.ddsa_snaps = snaps; | |
1625 | ddsa.ddsa_props = props; | |
1626 | ddsa.ddsa_errors = errors; | |
788eb90c | 1627 | ddsa.ddsa_cr = CRED(); |
13fe0198 MA |
1628 | |
1629 | if (error == 0) { | |
1630 | error = dsl_sync_task(firstname, dsl_dataset_snapshot_check, | |
1631 | dsl_dataset_snapshot_sync, &ddsa, | |
3d45fdd6 | 1632 | fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL); |
13fe0198 MA |
1633 | } |
1634 | ||
1635 | if (suspended != NULL) { | |
1636 | for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL; | |
1637 | pair = nvlist_next_nvpair(suspended, pair)) { | |
1638 | zil_resume((void *)(uintptr_t) | |
1639 | fnvpair_value_uint64(pair)); | |
1640 | } | |
1641 | fnvlist_free(suspended); | |
1642 | } | |
1643 | ||
1644 | return (error); | |
1645 | } | |
1646 | ||
1647 | typedef struct dsl_dataset_snapshot_tmp_arg { | |
1648 | const char *ddsta_fsname; | |
1649 | const char *ddsta_snapname; | |
1650 | minor_t ddsta_cleanup_minor; | |
1651 | const char *ddsta_htag; | |
1652 | } dsl_dataset_snapshot_tmp_arg_t; | |
1653 | ||
1654 | static int | |
1655 | dsl_dataset_snapshot_tmp_check(void *arg, dmu_tx_t *tx) | |
1656 | { | |
1657 | dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; | |
1658 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
1659 | dsl_dataset_t *ds; | |
1660 | int error; | |
1661 | ||
1662 | error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds); | |
1663 | if (error != 0) | |
1664 | return (error); | |
1665 | ||
788eb90c | 1666 | /* NULL cred means no limit check for tmp snapshot */ |
96c2e961 | 1667 | error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname, |
788eb90c | 1668 | tx, B_FALSE, 0, NULL); |
13fe0198 MA |
1669 | if (error != 0) { |
1670 | dsl_dataset_rele(ds, FTAG); | |
1671 | return (error); | |
1672 | } | |
1673 | ||
1674 | if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) { | |
1675 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 1676 | return (SET_ERROR(ENOTSUP)); |
13fe0198 MA |
1677 | } |
1678 | error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag, | |
1679 | B_TRUE, tx); | |
1680 | if (error != 0) { | |
1681 | dsl_dataset_rele(ds, FTAG); | |
1682 | return (error); | |
1683 | } | |
1684 | ||
1685 | dsl_dataset_rele(ds, FTAG); | |
1686 | return (0); | |
1687 | } | |
1688 | ||
1689 | static void | |
1690 | dsl_dataset_snapshot_tmp_sync(void *arg, dmu_tx_t *tx) | |
1691 | { | |
1692 | dsl_dataset_snapshot_tmp_arg_t *ddsta = arg; | |
1693 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 1694 | dsl_dataset_t *ds = NULL; |
13fe0198 MA |
1695 | |
1696 | VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds)); | |
1697 | ||
1698 | dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx); | |
1699 | dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag, | |
1700 | ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx); | |
1701 | dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx); | |
1702 | ||
1703 | dsl_dataset_rele(ds, FTAG); | |
1704 | } | |
1705 | ||
1706 | int | |
1707 | dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname, | |
1708 | minor_t cleanup_minor, const char *htag) | |
1709 | { | |
1710 | dsl_dataset_snapshot_tmp_arg_t ddsta; | |
1711 | int error; | |
1712 | spa_t *spa; | |
1713 | boolean_t needsuspend; | |
1714 | void *cookie; | |
1715 | ||
1716 | ddsta.ddsta_fsname = fsname; | |
1717 | ddsta.ddsta_snapname = snapname; | |
1718 | ddsta.ddsta_cleanup_minor = cleanup_minor; | |
1719 | ddsta.ddsta_htag = htag; | |
1720 | ||
1721 | error = spa_open(fsname, &spa, FTAG); | |
1722 | if (error != 0) | |
1723 | return (error); | |
1724 | needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP); | |
1725 | spa_close(spa, FTAG); | |
1726 | ||
1727 | if (needsuspend) { | |
1728 | error = zil_suspend(fsname, &cookie); | |
1729 | if (error != 0) | |
1730 | return (error); | |
1731 | } | |
1732 | ||
1733 | error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check, | |
3d45fdd6 | 1734 | dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED); |
13fe0198 MA |
1735 | |
1736 | if (needsuspend) | |
1737 | zil_resume(cookie); | |
1738 | return (error); | |
1739 | } | |
1740 | ||
13fe0198 MA |
1741 | void |
1742 | dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx) | |
1743 | { | |
1744 | ASSERT(dmu_tx_is_syncing(tx)); | |
1745 | ASSERT(ds->ds_objset != NULL); | |
d683ddbb | 1746 | ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0); |
13fe0198 MA |
1747 | |
1748 | /* | |
1749 | * in case we had to change ds_fsid_guid when we opened it, | |
1750 | * sync it out now. | |
1751 | */ | |
1752 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
d683ddbb | 1753 | dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid; |
13fe0198 | 1754 | |
47dfff3b MA |
1755 | if (ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] != 0) { |
1756 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1757 | ds->ds_object, DS_FIELD_RESUME_OBJECT, 8, 1, | |
1758 | &ds->ds_resume_object[tx->tx_txg & TXG_MASK], tx)); | |
1759 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1760 | ds->ds_object, DS_FIELD_RESUME_OFFSET, 8, 1, | |
1761 | &ds->ds_resume_offset[tx->tx_txg & TXG_MASK], tx)); | |
1762 | VERIFY0(zap_update(tx->tx_pool->dp_meta_objset, | |
1763 | ds->ds_object, DS_FIELD_RESUME_BYTES, 8, 1, | |
1764 | &ds->ds_resume_bytes[tx->tx_txg & TXG_MASK], tx)); | |
1765 | ds->ds_resume_object[tx->tx_txg & TXG_MASK] = 0; | |
1766 | ds->ds_resume_offset[tx->tx_txg & TXG_MASK] = 0; | |
1767 | ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] = 0; | |
1768 | } | |
1769 | ||
13fe0198 | 1770 | dmu_objset_sync(ds->ds_objset, zio, tx); |
f1512ee6 | 1771 | |
1c27024e | 1772 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
1773 | if (ds->ds_feature_activation_needed[f]) { |
1774 | if (ds->ds_feature_inuse[f]) | |
1775 | continue; | |
1776 | dsl_dataset_activate_feature(ds->ds_object, f, tx); | |
1777 | ds->ds_feature_inuse[f] = B_TRUE; | |
1778 | } | |
f1512ee6 | 1779 | } |
13fe0198 MA |
1780 | } |
1781 | ||
0efd9791 AG |
1782 | static int |
1783 | deadlist_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx) | |
1784 | { | |
1785 | dsl_deadlist_t *dl = arg; | |
1786 | dsl_deadlist_insert(dl, bp, tx); | |
1787 | return (0); | |
1788 | } | |
1789 | ||
1790 | void | |
1791 | dsl_dataset_sync_done(dsl_dataset_t *ds, dmu_tx_t *tx) | |
1792 | { | |
64fc7762 | 1793 | objset_t *os = ds->ds_objset; |
0efd9791 AG |
1794 | |
1795 | bplist_iterate(&ds->ds_pending_deadlist, | |
1796 | deadlist_enqueue_cb, &ds->ds_deadlist, tx); | |
1797 | ||
64fc7762 MA |
1798 | if (os->os_synced_dnodes != NULL) { |
1799 | multilist_destroy(os->os_synced_dnodes); | |
1800 | os->os_synced_dnodes = NULL; | |
1801 | } | |
1802 | ||
0efd9791 AG |
1803 | ASSERT(!dmu_objset_is_dirty(os, dmu_tx_get_txg(tx))); |
1804 | ||
1805 | dmu_buf_rele(ds->ds_dbuf, ds); | |
1806 | } | |
1807 | ||
d99a0153 CW |
1808 | int |
1809 | get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val) | |
13fe0198 MA |
1810 | { |
1811 | uint64_t count = 0; | |
1812 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; | |
1813 | zap_cursor_t zc; | |
1814 | zap_attribute_t za; | |
13fe0198 MA |
1815 | |
1816 | ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); | |
330d06f9 MA |
1817 | |
1818 | /* | |
13fe0198 | 1819 | * There may be missing entries in ds_next_clones_obj |
330d06f9 MA |
1820 | * due to a bug in a previous version of the code. |
1821 | * Only trust it if it has the right number of entries. | |
1822 | */ | |
d683ddbb JG |
1823 | if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) { |
1824 | VERIFY0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj, | |
330d06f9 MA |
1825 | &count)); |
1826 | } | |
d99a0153 CW |
1827 | if (count != dsl_dataset_phys(ds)->ds_num_children - 1) { |
1828 | return (ENOENT); | |
1829 | } | |
d683ddbb JG |
1830 | for (zap_cursor_init(&zc, mos, |
1831 | dsl_dataset_phys(ds)->ds_next_clones_obj); | |
330d06f9 MA |
1832 | zap_cursor_retrieve(&zc, &za) == 0; |
1833 | zap_cursor_advance(&zc)) { | |
1834 | dsl_dataset_t *clone; | |
eca7b760 | 1835 | char buf[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
1836 | VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool, |
1837 | za.za_first_integer, FTAG, &clone)); | |
330d06f9 | 1838 | dsl_dir_name(clone->ds_dir, buf); |
13fe0198 | 1839 | fnvlist_add_boolean(val, buf); |
330d06f9 MA |
1840 | dsl_dataset_rele(clone, FTAG); |
1841 | } | |
1842 | zap_cursor_fini(&zc); | |
d99a0153 CW |
1843 | return (0); |
1844 | } | |
1845 | ||
1846 | void | |
1847 | get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv) | |
1848 | { | |
1849 | nvlist_t *propval = fnvlist_alloc(); | |
1850 | nvlist_t *val; | |
1851 | ||
1852 | /* | |
1853 | * We use nvlist_alloc() instead of fnvlist_alloc() because the | |
1854 | * latter would allocate the list with NV_UNIQUE_NAME flag. | |
1855 | * As a result, every time a clone name is appended to the list | |
1856 | * it would be (linearly) searched for for a duplicate name. | |
1857 | * We already know that all clone names must be unique and we | |
1858 | * want avoid the quadratic complexity of double-checking that | |
1859 | * because we can have a large number of clones. | |
1860 | */ | |
1861 | VERIFY0(nvlist_alloc(&val, 0, KM_SLEEP)); | |
1862 | ||
1863 | if (get_clones_stat_impl(ds, val) == 0) { | |
1864 | fnvlist_add_nvlist(propval, ZPROP_VALUE, val); | |
1865 | fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES), | |
1866 | propval); | |
1867 | } | |
1868 | ||
330d06f9 MA |
1869 | nvlist_free(val); |
1870 | nvlist_free(propval); | |
330d06f9 MA |
1871 | } |
1872 | ||
d99a0153 CW |
1873 | /* |
1874 | * Returns a string that represents the receive resume stats token. It should | |
1875 | * be freed with strfree(). | |
1876 | */ | |
1877 | char * | |
1878 | get_receive_resume_stats_impl(dsl_dataset_t *ds) | |
47dfff3b MA |
1879 | { |
1880 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
1881 | ||
1882 | if (dsl_dataset_has_resume_receive_state(ds)) { | |
1883 | char *str; | |
1884 | void *packed; | |
1885 | uint8_t *compressed; | |
1886 | uint64_t val; | |
1887 | nvlist_t *token_nv = fnvlist_alloc(); | |
1888 | size_t packed_size, compressed_size; | |
47dfff3b MA |
1889 | |
1890 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1891 | DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val) == 0) { | |
1892 | fnvlist_add_uint64(token_nv, "fromguid", val); | |
1893 | } | |
1894 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1895 | DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val) == 0) { | |
1896 | fnvlist_add_uint64(token_nv, "object", val); | |
1897 | } | |
1898 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1899 | DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val) == 0) { | |
1900 | fnvlist_add_uint64(token_nv, "offset", val); | |
1901 | } | |
1902 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1903 | DS_FIELD_RESUME_BYTES, sizeof (val), 1, &val) == 0) { | |
1904 | fnvlist_add_uint64(token_nv, "bytes", val); | |
1905 | } | |
1906 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, | |
1907 | DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val) == 0) { | |
1908 | fnvlist_add_uint64(token_nv, "toguid", val); | |
1909 | } | |
1c27024e | 1910 | char buf[MAXNAMELEN]; |
47dfff3b MA |
1911 | if (zap_lookup(dp->dp_meta_objset, ds->ds_object, |
1912 | DS_FIELD_RESUME_TONAME, 1, sizeof (buf), buf) == 0) { | |
1913 | fnvlist_add_string(token_nv, "toname", buf); | |
1914 | } | |
2aa34383 DK |
1915 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1916 | DS_FIELD_RESUME_LARGEBLOCK) == 0) { | |
1917 | fnvlist_add_boolean(token_nv, "largeblockok"); | |
1918 | } | |
47dfff3b MA |
1919 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1920 | DS_FIELD_RESUME_EMBEDOK) == 0) { | |
1921 | fnvlist_add_boolean(token_nv, "embedok"); | |
1922 | } | |
2aa34383 DK |
1923 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1924 | DS_FIELD_RESUME_COMPRESSOK) == 0) { | |
1925 | fnvlist_add_boolean(token_nv, "compressok"); | |
1926 | } | |
b5256303 TC |
1927 | if (zap_contains(dp->dp_meta_objset, ds->ds_object, |
1928 | DS_FIELD_RESUME_RAWOK) == 0) { | |
1929 | fnvlist_add_boolean(token_nv, "rawok"); | |
1930 | } | |
47dfff3b MA |
1931 | packed = fnvlist_pack(token_nv, &packed_size); |
1932 | fnvlist_free(token_nv); | |
1933 | compressed = kmem_alloc(packed_size, KM_SLEEP); | |
1934 | ||
1935 | compressed_size = gzip_compress(packed, compressed, | |
1936 | packed_size, packed_size, 6); | |
1937 | ||
1c27024e | 1938 | zio_cksum_t cksum; |
fc897b24 | 1939 | fletcher_4_native_varsize(compressed, compressed_size, &cksum); |
47dfff3b MA |
1940 | |
1941 | str = kmem_alloc(compressed_size * 2 + 1, KM_SLEEP); | |
1c27024e | 1942 | for (int i = 0; i < compressed_size; i++) { |
47dfff3b MA |
1943 | (void) sprintf(str + i * 2, "%02x", compressed[i]); |
1944 | } | |
1945 | str[compressed_size * 2] = '\0'; | |
1c27024e | 1946 | char *propval = kmem_asprintf("%u-%llx-%llx-%s", |
47dfff3b MA |
1947 | ZFS_SEND_RESUME_TOKEN_VERSION, |
1948 | (longlong_t)cksum.zc_word[0], | |
1949 | (longlong_t)packed_size, str); | |
47dfff3b MA |
1950 | kmem_free(packed, packed_size); |
1951 | kmem_free(str, compressed_size * 2 + 1); | |
1952 | kmem_free(compressed, packed_size); | |
d99a0153 CW |
1953 | return (propval); |
1954 | } | |
1955 | return (strdup("")); | |
1956 | } | |
1957 | ||
1958 | /* | |
1959 | * Returns a string that represents the receive resume stats token of the | |
1960 | * dataset's child. It should be freed with strfree(). | |
1961 | */ | |
1962 | char * | |
1963 | get_child_receive_stats(dsl_dataset_t *ds) | |
1964 | { | |
1965 | char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; | |
1966 | dsl_dataset_t *recv_ds; | |
1967 | dsl_dataset_name(ds, recvname); | |
1968 | if (strlcat(recvname, "/", sizeof (recvname)) < | |
1969 | sizeof (recvname) && | |
1970 | strlcat(recvname, recv_clone_name, sizeof (recvname)) < | |
1971 | sizeof (recvname) && | |
1972 | dsl_dataset_hold(ds->ds_dir->dd_pool, recvname, FTAG, | |
1973 | &recv_ds) == 0) { | |
1974 | char *propval = get_receive_resume_stats_impl(recv_ds); | |
1975 | dsl_dataset_rele(recv_ds, FTAG); | |
1976 | return (propval); | |
47dfff3b | 1977 | } |
d99a0153 CW |
1978 | return (strdup("")); |
1979 | } | |
1980 | ||
1981 | static void | |
1982 | get_receive_resume_stats(dsl_dataset_t *ds, nvlist_t *nv) | |
1983 | { | |
1984 | char *propval = get_receive_resume_stats_impl(ds); | |
1985 | if (strcmp(propval, "") != 0) { | |
1986 | dsl_prop_nvlist_add_string(nv, | |
1987 | ZFS_PROP_RECEIVE_RESUME_TOKEN, propval); | |
1988 | } else { | |
1989 | char *childval = get_child_receive_stats(ds); | |
1990 | if (strcmp(childval, "") != 0) { | |
1991 | dsl_prop_nvlist_add_string(nv, | |
1992 | ZFS_PROP_RECEIVE_RESUME_TOKEN, childval); | |
1993 | } | |
1994 | strfree(childval); | |
1995 | } | |
1996 | strfree(propval); | |
1997 | } | |
1998 | ||
1999 | uint64_t | |
2000 | dsl_get_refratio(dsl_dataset_t *ds) | |
2001 | { | |
2002 | uint64_t ratio = dsl_dataset_phys(ds)->ds_compressed_bytes == 0 ? 100 : | |
2003 | (dsl_dataset_phys(ds)->ds_uncompressed_bytes * 100 / | |
2004 | dsl_dataset_phys(ds)->ds_compressed_bytes); | |
2005 | return (ratio); | |
2006 | } | |
2007 | ||
2008 | uint64_t | |
2009 | dsl_get_logicalreferenced(dsl_dataset_t *ds) | |
2010 | { | |
2011 | return (dsl_dataset_phys(ds)->ds_uncompressed_bytes); | |
2012 | } | |
2013 | ||
2014 | uint64_t | |
2015 | dsl_get_compressratio(dsl_dataset_t *ds) | |
2016 | { | |
2017 | if (ds->ds_is_snapshot) { | |
2018 | return (dsl_get_refratio(ds)); | |
2019 | } else { | |
2020 | dsl_dir_t *dd = ds->ds_dir; | |
2021 | mutex_enter(&dd->dd_lock); | |
2022 | uint64_t val = dsl_dir_get_compressratio(dd); | |
2023 | mutex_exit(&dd->dd_lock); | |
2024 | return (val); | |
2025 | } | |
2026 | } | |
2027 | ||
2028 | uint64_t | |
2029 | dsl_get_used(dsl_dataset_t *ds) | |
2030 | { | |
2031 | if (ds->ds_is_snapshot) { | |
2032 | return (dsl_dataset_phys(ds)->ds_unique_bytes); | |
2033 | } else { | |
2034 | dsl_dir_t *dd = ds->ds_dir; | |
2035 | mutex_enter(&dd->dd_lock); | |
2036 | uint64_t val = dsl_dir_get_used(dd); | |
2037 | mutex_exit(&dd->dd_lock); | |
2038 | return (val); | |
2039 | } | |
2040 | } | |
2041 | ||
2042 | uint64_t | |
2043 | dsl_get_creation(dsl_dataset_t *ds) | |
2044 | { | |
2045 | return (dsl_dataset_phys(ds)->ds_creation_time); | |
2046 | } | |
2047 | ||
2048 | uint64_t | |
2049 | dsl_get_creationtxg(dsl_dataset_t *ds) | |
2050 | { | |
2051 | return (dsl_dataset_phys(ds)->ds_creation_txg); | |
2052 | } | |
2053 | ||
2054 | uint64_t | |
2055 | dsl_get_refquota(dsl_dataset_t *ds) | |
2056 | { | |
2057 | return (ds->ds_quota); | |
2058 | } | |
2059 | ||
2060 | uint64_t | |
2061 | dsl_get_refreservation(dsl_dataset_t *ds) | |
2062 | { | |
2063 | return (ds->ds_reserved); | |
2064 | } | |
2065 | ||
2066 | uint64_t | |
2067 | dsl_get_guid(dsl_dataset_t *ds) | |
2068 | { | |
2069 | return (dsl_dataset_phys(ds)->ds_guid); | |
2070 | } | |
2071 | ||
2072 | uint64_t | |
2073 | dsl_get_unique(dsl_dataset_t *ds) | |
2074 | { | |
2075 | return (dsl_dataset_phys(ds)->ds_unique_bytes); | |
2076 | } | |
2077 | ||
2078 | uint64_t | |
2079 | dsl_get_objsetid(dsl_dataset_t *ds) | |
2080 | { | |
2081 | return (ds->ds_object); | |
2082 | } | |
2083 | ||
2084 | uint64_t | |
2085 | dsl_get_userrefs(dsl_dataset_t *ds) | |
2086 | { | |
2087 | return (ds->ds_userrefs); | |
2088 | } | |
2089 | ||
2090 | uint64_t | |
2091 | dsl_get_defer_destroy(dsl_dataset_t *ds) | |
2092 | { | |
2093 | return (DS_IS_DEFER_DESTROY(ds) ? 1 : 0); | |
2094 | } | |
2095 | ||
2096 | uint64_t | |
2097 | dsl_get_referenced(dsl_dataset_t *ds) | |
2098 | { | |
2099 | return (dsl_dataset_phys(ds)->ds_referenced_bytes); | |
2100 | } | |
2101 | ||
2102 | uint64_t | |
2103 | dsl_get_numclones(dsl_dataset_t *ds) | |
2104 | { | |
2105 | ASSERT(ds->ds_is_snapshot); | |
2106 | return (dsl_dataset_phys(ds)->ds_num_children - 1); | |
2107 | } | |
2108 | ||
2109 | uint64_t | |
2110 | dsl_get_inconsistent(dsl_dataset_t *ds) | |
2111 | { | |
2112 | return ((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT) ? | |
2113 | 1 : 0); | |
2114 | } | |
2115 | ||
2116 | uint64_t | |
2117 | dsl_get_available(dsl_dataset_t *ds) | |
2118 | { | |
2119 | uint64_t refdbytes = dsl_get_referenced(ds); | |
2120 | uint64_t availbytes = dsl_dir_space_available(ds->ds_dir, | |
2121 | NULL, 0, TRUE); | |
2122 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) { | |
2123 | availbytes += | |
2124 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes; | |
2125 | } | |
2126 | if (ds->ds_quota != 0) { | |
2127 | /* | |
2128 | * Adjust available bytes according to refquota | |
2129 | */ | |
2130 | if (refdbytes < ds->ds_quota) { | |
2131 | availbytes = MIN(availbytes, | |
2132 | ds->ds_quota - refdbytes); | |
2133 | } else { | |
2134 | availbytes = 0; | |
2135 | } | |
2136 | } | |
2137 | return (availbytes); | |
2138 | } | |
2139 | ||
2140 | int | |
2141 | dsl_get_written(dsl_dataset_t *ds, uint64_t *written) | |
2142 | { | |
2143 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2144 | dsl_dataset_t *prev; | |
2145 | int err = dsl_dataset_hold_obj(dp, | |
2146 | dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev); | |
2147 | if (err == 0) { | |
2148 | uint64_t comp, uncomp; | |
2149 | err = dsl_dataset_space_written(prev, ds, written, | |
2150 | &comp, &uncomp); | |
2151 | dsl_dataset_rele(prev, FTAG); | |
2152 | } | |
2153 | return (err); | |
2154 | } | |
2155 | ||
2156 | /* | |
2157 | * 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN. | |
2158 | */ | |
2159 | int | |
2160 | dsl_get_prev_snap(dsl_dataset_t *ds, char *snap) | |
2161 | { | |
2162 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2163 | if (ds->ds_prev != NULL && ds->ds_prev != dp->dp_origin_snap) { | |
2164 | dsl_dataset_name(ds->ds_prev, snap); | |
2165 | return (0); | |
2166 | } else { | |
2167 | return (ENOENT); | |
2168 | } | |
2169 | } | |
2170 | ||
2171 | /* | |
2172 | * Returns the mountpoint property and source for the given dataset in the value | |
2173 | * and source buffers. The value buffer must be at least as large as MAXPATHLEN | |
2174 | * and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN. | |
2175 | * Returns 0 on success and an error on failure. | |
2176 | */ | |
2177 | int | |
2178 | dsl_get_mountpoint(dsl_dataset_t *ds, const char *dsname, char *value, | |
2179 | char *source) | |
2180 | { | |
2181 | int error; | |
2182 | dsl_pool_t *dp = ds->ds_dir->dd_pool; | |
2183 | ||
2184 | /* Retrieve the mountpoint value stored in the zap opbject */ | |
2185 | error = dsl_prop_get_ds(ds, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 1, | |
2186 | ZAP_MAXVALUELEN, value, source); | |
2187 | if (error != 0) { | |
2188 | return (error); | |
2189 | } | |
2190 | ||
2191 | /* | |
2192 | * Process the dsname and source to find the full mountpoint string. | |
2193 | * Can be skipped for 'legacy' or 'none'. | |
2194 | */ | |
2195 | if (value[0] == '/') { | |
2196 | char *buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP); | |
2197 | char *root = buf; | |
2198 | const char *relpath; | |
2199 | ||
2200 | /* | |
2201 | * If we inherit the mountpoint, even from a dataset | |
2202 | * with a received value, the source will be the path of | |
2203 | * the dataset we inherit from. If source is | |
2204 | * ZPROP_SOURCE_VAL_RECVD, the received value is not | |
2205 | * inherited. | |
2206 | */ | |
2207 | if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) { | |
2208 | relpath = ""; | |
2209 | } else { | |
2210 | ASSERT0(strncmp(dsname, source, strlen(source))); | |
2211 | relpath = dsname + strlen(source); | |
2212 | if (relpath[0] == '/') | |
2213 | relpath++; | |
2214 | } | |
2215 | ||
2216 | spa_altroot(dp->dp_spa, root, ZAP_MAXVALUELEN); | |
2217 | ||
2218 | /* | |
2219 | * Special case an alternate root of '/'. This will | |
2220 | * avoid having multiple leading slashes in the | |
2221 | * mountpoint path. | |
2222 | */ | |
2223 | if (strcmp(root, "/") == 0) | |
2224 | root++; | |
2225 | ||
2226 | /* | |
2227 | * If the mountpoint is '/' then skip over this | |
2228 | * if we are obtaining either an alternate root or | |
2229 | * an inherited mountpoint. | |
2230 | */ | |
2231 | char *mnt = value; | |
2232 | if (value[1] == '\0' && (root[0] != '\0' || | |
2233 | relpath[0] != '\0')) | |
2234 | mnt = value + 1; | |
2235 | ||
2236 | if (relpath[0] == '\0') { | |
2237 | (void) snprintf(value, ZAP_MAXVALUELEN, "%s%s", | |
2238 | root, mnt); | |
2239 | } else { | |
2240 | (void) snprintf(value, ZAP_MAXVALUELEN, "%s%s%s%s", | |
2241 | root, mnt, relpath[0] == '@' ? "" : "/", | |
2242 | relpath); | |
2243 | } | |
2244 | kmem_free(buf, ZAP_MAXVALUELEN); | |
2245 | } | |
2246 | ||
2247 | return (0); | |
47dfff3b MA |
2248 | } |
2249 | ||
34dc7c2f BB |
2250 | void |
2251 | dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv) | |
2252 | { | |
b858767a | 2253 | dsl_pool_t *dp = ds->ds_dir->dd_pool; |
13fe0198 MA |
2254 | |
2255 | ASSERT(dsl_pool_config_held(dp)); | |
34dc7c2f | 2256 | |
d99a0153 CW |
2257 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO, |
2258 | dsl_get_refratio(ds)); | |
24a64651 | 2259 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED, |
d99a0153 CW |
2260 | dsl_get_logicalreferenced(ds)); |
2261 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO, | |
2262 | dsl_get_compressratio(ds)); | |
2263 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED, | |
2264 | dsl_get_used(ds)); | |
6f1ffb06 | 2265 | |
0c66c32d | 2266 | if (ds->ds_is_snapshot) { |
6f1ffb06 MA |
2267 | get_clones_stat(ds, nv); |
2268 | } else { | |
d99a0153 CW |
2269 | char buf[ZFS_MAX_DATASET_NAME_LEN]; |
2270 | if (dsl_get_prev_snap(ds, buf) == 0) | |
2271 | dsl_prop_nvlist_add_string(nv, ZFS_PROP_PREV_SNAP, | |
2272 | buf); | |
6f1ffb06 MA |
2273 | dsl_dir_stats(ds->ds_dir, nv); |
2274 | } | |
34dc7c2f | 2275 | |
d99a0153 CW |
2276 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE, |
2277 | dsl_get_available(ds)); | |
2278 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED, | |
2279 | dsl_get_referenced(ds)); | |
34dc7c2f | 2280 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION, |
d99a0153 | 2281 | dsl_get_creation(ds)); |
34dc7c2f | 2282 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG, |
d99a0153 | 2283 | dsl_get_creationtxg(ds)); |
34dc7c2f | 2284 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA, |
d99a0153 | 2285 | dsl_get_refquota(ds)); |
34dc7c2f | 2286 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION, |
d99a0153 | 2287 | dsl_get_refreservation(ds)); |
b128c09f | 2288 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID, |
d99a0153 | 2289 | dsl_get_guid(ds)); |
428870ff | 2290 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE, |
d99a0153 | 2291 | dsl_get_unique(ds)); |
428870ff | 2292 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID, |
d99a0153 | 2293 | dsl_get_objsetid(ds)); |
428870ff | 2294 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS, |
d99a0153 | 2295 | dsl_get_userrefs(ds)); |
45d1cae3 | 2296 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY, |
d99a0153 | 2297 | dsl_get_defer_destroy(ds)); |
b5256303 | 2298 | dsl_dataset_crypt_stats(ds, nv); |
34dc7c2f | 2299 | |
d683ddbb | 2300 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { |
d99a0153 CW |
2301 | uint64_t written; |
2302 | if (dsl_get_written(ds, &written) == 0) { | |
2303 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN, | |
2304 | written); | |
330d06f9 MA |
2305 | } |
2306 | } | |
2307 | ||
47dfff3b | 2308 | if (!dsl_dataset_is_snapshot(ds)) { |
47dfff3b MA |
2309 | /* |
2310 | * A failed "newfs" (e.g. full) resumable receive leaves | |
2311 | * the stats set on this dataset. Check here for the prop. | |
2312 | */ | |
2313 | get_receive_resume_stats(ds, nv); | |
2314 | ||
2315 | /* | |
2316 | * A failed incremental resumable receive leaves the | |
2317 | * stats set on our child named "%recv". Check the child | |
2318 | * for the prop. | |
2319 | */ | |
1c27024e DB |
2320 | /* 6 extra bytes for /%recv */ |
2321 | char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; | |
2322 | dsl_dataset_t *recv_ds; | |
47dfff3b | 2323 | dsl_dataset_name(ds, recvname); |
eca7b760 IK |
2324 | if (strlcat(recvname, "/", sizeof (recvname)) < |
2325 | sizeof (recvname) && | |
2326 | strlcat(recvname, recv_clone_name, sizeof (recvname)) < | |
2327 | sizeof (recvname) && | |
2328 | dsl_dataset_hold(dp, recvname, FTAG, &recv_ds) == 0) { | |
47dfff3b MA |
2329 | get_receive_resume_stats(recv_ds, nv); |
2330 | dsl_dataset_rele(recv_ds, FTAG); | |
2331 | } | |
2332 | } | |
34dc7c2f BB |
2333 | } |
2334 | ||
2335 | void | |
2336 | dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat) | |
2337 | { | |
d99a0153 | 2338 | ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool); |
13fe0198 MA |
2339 | ASSERT(dsl_pool_config_held(dp)); |
2340 | ||
d99a0153 CW |
2341 | stat->dds_creation_txg = dsl_get_creationtxg(ds); |
2342 | stat->dds_inconsistent = dsl_get_inconsistent(ds); | |
2343 | stat->dds_guid = dsl_get_guid(ds); | |
6f1ffb06 | 2344 | stat->dds_origin[0] = '\0'; |
0c66c32d | 2345 | if (ds->ds_is_snapshot) { |
34dc7c2f | 2346 | stat->dds_is_snapshot = B_TRUE; |
d99a0153 | 2347 | stat->dds_num_clones = dsl_get_numclones(ds); |
fb5f0bc8 BB |
2348 | } else { |
2349 | stat->dds_is_snapshot = B_FALSE; | |
2350 | stat->dds_num_clones = 0; | |
34dc7c2f | 2351 | |
6f1ffb06 | 2352 | if (dsl_dir_is_clone(ds->ds_dir)) { |
d99a0153 | 2353 | dsl_dir_get_origin(ds->ds_dir, stat->dds_origin); |
6f1ffb06 | 2354 | } |
34dc7c2f | 2355 | } |
34dc7c2f BB |
2356 | } |
2357 | ||
2358 | uint64_t | |
2359 | dsl_dataset_fsid_guid(dsl_dataset_t *ds) | |
2360 | { | |
2361 | return (ds->ds_fsid_guid); | |
2362 | } | |
2363 | ||
2364 | void | |
2365 | dsl_dataset_space(dsl_dataset_t *ds, | |
2366 | uint64_t *refdbytesp, uint64_t *availbytesp, | |
2367 | uint64_t *usedobjsp, uint64_t *availobjsp) | |
2368 | { | |
d683ddbb | 2369 | *refdbytesp = dsl_dataset_phys(ds)->ds_referenced_bytes; |
34dc7c2f | 2370 | *availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE); |
d683ddbb JG |
2371 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) |
2372 | *availbytesp += | |
2373 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes; | |
34dc7c2f BB |
2374 | if (ds->ds_quota != 0) { |
2375 | /* | |
2376 | * Adjust available bytes according to refquota | |
2377 | */ | |
2378 | if (*refdbytesp < ds->ds_quota) | |
2379 | *availbytesp = MIN(*availbytesp, | |
2380 | ds->ds_quota - *refdbytesp); | |
2381 | else | |
2382 | *availbytesp = 0; | |
2383 | } | |
cc9bb3e5 | 2384 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
d683ddbb | 2385 | *usedobjsp = BP_GET_FILL(&dsl_dataset_phys(ds)->ds_bp); |
cc9bb3e5 | 2386 | rrw_exit(&ds->ds_bp_rwlock, FTAG); |
34dc7c2f BB |
2387 | *availobjsp = DN_MAX_OBJECT - *usedobjsp; |
2388 | } | |
2389 | ||
2390 | boolean_t | |
19580676 | 2391 | dsl_dataset_modified_since_snap(dsl_dataset_t *ds, dsl_dataset_t *snap) |
34dc7c2f | 2392 | { |
cc9bb3e5 GM |
2393 | ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool); |
2394 | uint64_t birth; | |
2395 | ||
2396 | ASSERT(dsl_pool_config_held(dp)); | |
19580676 | 2397 | if (snap == NULL) |
34dc7c2f | 2398 | return (B_FALSE); |
cc9bb3e5 GM |
2399 | rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); |
2400 | birth = dsl_dataset_get_blkptr(ds)->blk_birth; | |
2401 | rrw_exit(&ds->ds_bp_rwlock, FTAG); | |
2402 | if (birth > dsl_dataset_phys(snap)->ds_creation_txg) { | |
19580676 | 2403 | objset_t *os, *os_snap; |
572e2857 BB |
2404 | /* |
2405 | * It may be that only the ZIL differs, because it was | |
2406 | * reset in the head. Don't count that as being | |
2407 | * modified. | |
2408 | */ | |
2409 | if (dmu_objset_from_ds(ds, &os) != 0) | |
2410 | return (B_TRUE); | |
19580676 | 2411 | if (dmu_objset_from_ds(snap, &os_snap) != 0) |
572e2857 BB |
2412 | return (B_TRUE); |
2413 | return (bcmp(&os->os_phys->os_meta_dnode, | |
19580676 | 2414 | &os_snap->os_phys->os_meta_dnode, |
572e2857 BB |
2415 | sizeof (os->os_phys->os_meta_dnode)) != 0); |
2416 | } | |
34dc7c2f BB |
2417 | return (B_FALSE); |
2418 | } | |
2419 | ||
13fe0198 MA |
2420 | typedef struct dsl_dataset_rename_snapshot_arg { |
2421 | const char *ddrsa_fsname; | |
2422 | const char *ddrsa_oldsnapname; | |
2423 | const char *ddrsa_newsnapname; | |
2424 | boolean_t ddrsa_recursive; | |
2425 | dmu_tx_t *ddrsa_tx; | |
2426 | } dsl_dataset_rename_snapshot_arg_t; | |
2427 | ||
34dc7c2f BB |
2428 | /* ARGSUSED */ |
2429 | static int | |
13fe0198 MA |
2430 | dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp, |
2431 | dsl_dataset_t *hds, void *arg) | |
34dc7c2f | 2432 | { |
13fe0198 MA |
2433 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2434 | int error; | |
34dc7c2f | 2435 | uint64_t val; |
34dc7c2f | 2436 | |
13fe0198 MA |
2437 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); |
2438 | if (error != 0) { | |
2439 | /* ignore nonexistent snapshots */ | |
2440 | return (error == ENOENT ? 0 : error); | |
2441 | } | |
34dc7c2f | 2442 | |
13fe0198 MA |
2443 | /* new name should not exist */ |
2444 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val); | |
2445 | if (error == 0) | |
2e528b49 | 2446 | error = SET_ERROR(EEXIST); |
13fe0198 MA |
2447 | else if (error == ENOENT) |
2448 | error = 0; | |
34dc7c2f BB |
2449 | |
2450 | /* dataset name + 1 for the "@" + the new snapshot name must fit */ | |
13fe0198 | 2451 | if (dsl_dir_namelen(hds->ds_dir) + 1 + |
eca7b760 | 2452 | strlen(ddrsa->ddrsa_newsnapname) >= ZFS_MAX_DATASET_NAME_LEN) |
2e528b49 | 2453 | error = SET_ERROR(ENAMETOOLONG); |
34dc7c2f | 2454 | |
13fe0198 | 2455 | return (error); |
34dc7c2f BB |
2456 | } |
2457 | ||
13fe0198 MA |
2458 | static int |
2459 | dsl_dataset_rename_snapshot_check(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 2460 | { |
13fe0198 MA |
2461 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2462 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
34dc7c2f | 2463 | dsl_dataset_t *hds; |
13fe0198 | 2464 | int error; |
34dc7c2f | 2465 | |
13fe0198 MA |
2466 | error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds); |
2467 | if (error != 0) | |
2468 | return (error); | |
34dc7c2f | 2469 | |
13fe0198 MA |
2470 | if (ddrsa->ddrsa_recursive) { |
2471 | error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object, | |
2472 | dsl_dataset_rename_snapshot_check_impl, ddrsa, | |
2473 | DS_FIND_CHILDREN); | |
2474 | } else { | |
2475 | error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa); | |
2476 | } | |
b128c09f | 2477 | dsl_dataset_rele(hds, FTAG); |
13fe0198 | 2478 | return (error); |
34dc7c2f BB |
2479 | } |
2480 | ||
34dc7c2f | 2481 | static int |
13fe0198 MA |
2482 | dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp, |
2483 | dsl_dataset_t *hds, void *arg) | |
34dc7c2f | 2484 | { |
13fe0198 MA |
2485 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2486 | dsl_dataset_t *ds; | |
2487 | uint64_t val; | |
2488 | dmu_tx_t *tx = ddrsa->ddrsa_tx; | |
2489 | int error; | |
34dc7c2f | 2490 | |
13fe0198 MA |
2491 | error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val); |
2492 | ASSERT(error == 0 || error == ENOENT); | |
2493 | if (error == ENOENT) { | |
2494 | /* ignore nonexistent snapshots */ | |
2495 | return (0); | |
34dc7c2f BB |
2496 | } |
2497 | ||
13fe0198 MA |
2498 | VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds)); |
2499 | ||
2500 | /* log before we change the name */ | |
2501 | spa_history_log_internal_ds(ds, "rename", tx, | |
2502 | "-> @%s", ddrsa->ddrsa_newsnapname); | |
34dc7c2f | 2503 | |
788eb90c JJ |
2504 | VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx, |
2505 | B_FALSE)); | |
13fe0198 | 2506 | mutex_enter(&ds->ds_lock); |
c9d61adb | 2507 | (void) strlcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname, |
2508 | sizeof (ds->ds_snapname)); | |
13fe0198 | 2509 | mutex_exit(&ds->ds_lock); |
d683ddbb JG |
2510 | VERIFY0(zap_add(dp->dp_meta_objset, |
2511 | dsl_dataset_phys(hds)->ds_snapnames_zapobj, | |
13fe0198 | 2512 | ds->ds_snapname, 8, 1, &ds->ds_object, tx)); |
a0bd735a BP |
2513 | zvol_rename_minors(dp->dp_spa, ddrsa->ddrsa_oldsnapname, |
2514 | ddrsa->ddrsa_newsnapname, B_TRUE); | |
34dc7c2f | 2515 | |
13fe0198 | 2516 | dsl_dataset_rele(ds, FTAG); |
34dc7c2f BB |
2517 | return (0); |
2518 | } | |
2519 | ||
13fe0198 MA |
2520 | static void |
2521 | dsl_dataset_rename_snapshot_sync(void *arg, dmu_tx_t *tx) | |
34dc7c2f | 2522 | { |
13fe0198 MA |
2523 | dsl_dataset_rename_snapshot_arg_t *ddrsa = arg; |
2524 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 2525 | dsl_dataset_t *hds = NULL; |
34dc7c2f | 2526 | |
13fe0198 MA |
2527 | VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds)); |
2528 | ddrsa->ddrsa_tx = tx; | |
2529 | if (ddrsa->ddrsa_recursive) { | |
2530 | VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object, | |
2531 | dsl_dataset_rename_snapshot_sync_impl, ddrsa, | |
2532 | DS_FIND_CHILDREN)); | |
2533 | } else { | |
2534 | VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa)); | |
34dc7c2f | 2535 | } |
13fe0198 | 2536 | dsl_dataset_rele(hds, FTAG); |
34dc7c2f BB |
2537 | } |
2538 | ||
13fe0198 MA |
2539 | int |
2540 | dsl_dataset_rename_snapshot(const char *fsname, | |
2541 | const char *oldsnapname, const char *newsnapname, boolean_t recursive) | |
34dc7c2f | 2542 | { |
13fe0198 | 2543 | dsl_dataset_rename_snapshot_arg_t ddrsa; |
34dc7c2f | 2544 | |
13fe0198 MA |
2545 | ddrsa.ddrsa_fsname = fsname; |
2546 | ddrsa.ddrsa_oldsnapname = oldsnapname; | |
2547 | ddrsa.ddrsa_newsnapname = newsnapname; | |
2548 | ddrsa.ddrsa_recursive = recursive; | |
34dc7c2f | 2549 | |
a0bd735a | 2550 | return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check, |
3d45fdd6 | 2551 | dsl_dataset_rename_snapshot_sync, &ddrsa, |
a0bd735a | 2552 | 1, ZFS_SPACE_CHECK_RESERVED)); |
34dc7c2f BB |
2553 | } |
2554 | ||
831baf06 KW |
2555 | /* |
2556 | * If we're doing an ownership handoff, we need to make sure that there is | |
2557 | * only one long hold on the dataset. We're not allowed to change anything here | |
2558 | * so we don't permanently release the long hold or regular hold here. We want | |
2559 | * to do this only when syncing to avoid the dataset unexpectedly going away | |
040dab99 | 2560 | * when we release the long hold. |
831baf06 KW |
2561 | */ |
2562 | static int | |
2563 | dsl_dataset_handoff_check(dsl_dataset_t *ds, void *owner, dmu_tx_t *tx) | |
2564 | { | |
2565 | boolean_t held; | |
2566 | ||
2567 | if (!dmu_tx_is_syncing(tx)) | |
2568 | return (0); | |
2569 | ||
2570 | if (owner != NULL) { | |
2571 | VERIFY3P(ds->ds_owner, ==, owner); | |
2572 | dsl_dataset_long_rele(ds, owner); | |
2573 | } | |
2574 | ||
040dab99 | 2575 | held = dsl_dataset_long_held(ds); |
831baf06 KW |
2576 | |
2577 | if (owner != NULL) | |
2578 | dsl_dataset_long_hold(ds, owner); | |
2579 | ||
2580 | if (held) | |
2581 | return (SET_ERROR(EBUSY)); | |
2582 | ||
2583 | return (0); | |
2584 | } | |
2585 | ||
af073689 | 2586 | int |
13fe0198 | 2587 | dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 2588 | { |
831baf06 | 2589 | dsl_dataset_rollback_arg_t *ddra = arg; |
13fe0198 | 2590 | dsl_pool_t *dp = dmu_tx_pool(tx); |
34dc7c2f | 2591 | dsl_dataset_t *ds; |
13fe0198 MA |
2592 | int64_t unused_refres_delta; |
2593 | int error; | |
34dc7c2f | 2594 | |
831baf06 | 2595 | error = dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds); |
13fe0198 MA |
2596 | if (error != 0) |
2597 | return (error); | |
428870ff | 2598 | |
13fe0198 | 2599 | /* must not be a snapshot */ |
0c66c32d | 2600 | if (ds->ds_is_snapshot) { |
13fe0198 | 2601 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 2602 | return (SET_ERROR(EINVAL)); |
13fe0198 | 2603 | } |
34dc7c2f | 2604 | |
13fe0198 | 2605 | /* must have a most recent snapshot */ |
d683ddbb | 2606 | if (dsl_dataset_phys(ds)->ds_prev_snap_txg < TXG_INITIAL) { |
13fe0198 | 2607 | dsl_dataset_rele(ds, FTAG); |
13342832 | 2608 | return (SET_ERROR(ESRCH)); |
13fe0198 | 2609 | } |
34dc7c2f | 2610 | |
0efd9791 AG |
2611 | /* |
2612 | * No rollback to a snapshot created in the current txg, because | |
2613 | * the rollback may dirty the dataset and create blocks that are | |
2614 | * not reachable from the rootbp while having a birth txg that | |
2615 | * falls into the snapshot's range. | |
2616 | */ | |
2617 | if (dmu_tx_is_syncing(tx) && | |
2618 | dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) { | |
2619 | dsl_dataset_rele(ds, FTAG); | |
2620 | return (SET_ERROR(EAGAIN)); | |
2621 | } | |
2622 | ||
8ca78ab0 AG |
2623 | /* |
2624 | * If the expected target snapshot is specified, then check that | |
2625 | * the latest snapshot is it. | |
2626 | */ | |
2627 | if (ddra->ddra_tosnap != NULL) { | |
13342832 AG |
2628 | dsl_dataset_t *snapds; |
2629 | ||
2630 | /* Check if the target snapshot exists at all. */ | |
2631 | error = dsl_dataset_hold(dp, ddra->ddra_tosnap, FTAG, &snapds); | |
2632 | if (error != 0) { | |
2633 | /* | |
2634 | * ESRCH is used to signal that the target snapshot does | |
2635 | * not exist, while ENOENT is used to report that | |
2636 | * the rolled back dataset does not exist. | |
2637 | * ESRCH is also used to cover other cases where the | |
2638 | * target snapshot is not related to the dataset being | |
2639 | * rolled back such as being in a different pool. | |
2640 | */ | |
2641 | if (error == ENOENT || error == EXDEV) | |
2642 | error = SET_ERROR(ESRCH); | |
2643 | dsl_dataset_rele(ds, FTAG); | |
2644 | return (error); | |
2645 | } | |
2646 | ASSERT(snapds->ds_is_snapshot); | |
8ca78ab0 | 2647 | |
13342832 AG |
2648 | /* Check if the snapshot is the latest snapshot indeed. */ |
2649 | if (snapds != ds->ds_prev) { | |
2650 | /* | |
2651 | * Distinguish between the case where the only problem | |
2652 | * is intervening snapshots (EEXIST) vs the snapshot | |
2653 | * not being a valid target for rollback (ESRCH). | |
2654 | */ | |
2655 | if (snapds->ds_dir == ds->ds_dir || | |
2656 | (dsl_dir_is_clone(ds->ds_dir) && | |
2657 | dsl_dir_phys(ds->ds_dir)->dd_origin_obj == | |
2658 | snapds->ds_object)) { | |
2659 | error = SET_ERROR(EEXIST); | |
2660 | } else { | |
2661 | error = SET_ERROR(ESRCH); | |
2662 | } | |
2663 | dsl_dataset_rele(snapds, FTAG); | |
2664 | dsl_dataset_rele(ds, FTAG); | |
2665 | return (error); | |
2666 | } | |
2667 | dsl_dataset_rele(snapds, FTAG); | |
8ca78ab0 AG |
2668 | } |
2669 | ||
da536844 | 2670 | /* must not have any bookmarks after the most recent snapshot */ |
1c27024e | 2671 | nvlist_t *proprequest = fnvlist_alloc(); |
da536844 | 2672 | fnvlist_add_boolean(proprequest, zfs_prop_to_name(ZFS_PROP_CREATETXG)); |
1c27024e | 2673 | nvlist_t *bookmarks = fnvlist_alloc(); |
da536844 MA |
2674 | error = dsl_get_bookmarks_impl(ds, proprequest, bookmarks); |
2675 | fnvlist_free(proprequest); | |
13342832 AG |
2676 | if (error != 0) { |
2677 | dsl_dataset_rele(ds, FTAG); | |
da536844 | 2678 | return (error); |
13342832 | 2679 | } |
1c27024e | 2680 | for (nvpair_t *pair = nvlist_next_nvpair(bookmarks, NULL); |
da536844 MA |
2681 | pair != NULL; pair = nvlist_next_nvpair(bookmarks, pair)) { |
2682 | nvlist_t *valuenv = | |
2683 | fnvlist_lookup_nvlist(fnvpair_value_nvlist(pair), | |
2684 | zfs_prop_to_name(ZFS_PROP_CREATETXG)); | |
2685 | uint64_t createtxg = fnvlist_lookup_uint64(valuenv, "value"); | |
d683ddbb | 2686 | if (createtxg > dsl_dataset_phys(ds)->ds_prev_snap_txg) { |
da536844 MA |
2687 | fnvlist_free(bookmarks); |
2688 | dsl_dataset_rele(ds, FTAG); | |
2689 | return (SET_ERROR(EEXIST)); | |
2690 | } | |
2691 | } | |
2692 | fnvlist_free(bookmarks); | |
2693 | ||
831baf06 KW |
2694 | error = dsl_dataset_handoff_check(ds, ddra->ddra_owner, tx); |
2695 | if (error != 0) { | |
13fe0198 | 2696 | dsl_dataset_rele(ds, FTAG); |
831baf06 | 2697 | return (error); |
34dc7c2f | 2698 | } |
428870ff | 2699 | |
13fe0198 MA |
2700 | /* |
2701 | * Check if the snap we are rolling back to uses more than | |
2702 | * the refquota. | |
2703 | */ | |
2704 | if (ds->ds_quota != 0 && | |
d683ddbb | 2705 | dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes > ds->ds_quota) { |
13fe0198 | 2706 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 2707 | return (SET_ERROR(EDQUOT)); |
34dc7c2f BB |
2708 | } |
2709 | ||
13fe0198 MA |
2710 | /* |
2711 | * When we do the clone swap, we will temporarily use more space | |
2712 | * due to the refreservation (the head will no longer have any | |
2713 | * unique space, so the entire amount of the refreservation will need | |
2714 | * to be free). We will immediately destroy the clone, freeing | |
2715 | * this space, but the freeing happens over many txg's. | |
2716 | */ | |
2717 | unused_refres_delta = (int64_t)MIN(ds->ds_reserved, | |
d683ddbb | 2718 | dsl_dataset_phys(ds)->ds_unique_bytes); |
34dc7c2f | 2719 | |
13fe0198 MA |
2720 | if (unused_refres_delta > 0 && |
2721 | unused_refres_delta > | |
2722 | dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) { | |
2723 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 2724 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 2725 | } |
34dc7c2f | 2726 | |
13fe0198 MA |
2727 | dsl_dataset_rele(ds, FTAG); |
2728 | return (0); | |
2729 | } | |
34dc7c2f | 2730 | |
af073689 | 2731 | void |
13fe0198 MA |
2732 | dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx) |
2733 | { | |
831baf06 | 2734 | dsl_dataset_rollback_arg_t *ddra = arg; |
13fe0198 MA |
2735 | dsl_pool_t *dp = dmu_tx_pool(tx); |
2736 | dsl_dataset_t *ds, *clone; | |
2737 | uint64_t cloneobj; | |
eca7b760 | 2738 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 2739 | |
831baf06 | 2740 | VERIFY0(dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds)); |
34dc7c2f | 2741 | |
46ba1e59 MA |
2742 | dsl_dataset_name(ds->ds_prev, namebuf); |
2743 | fnvlist_add_string(ddra->ddra_result, "target", namebuf); | |
2744 | ||
13fe0198 | 2745 | cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback", |
b5256303 | 2746 | ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, NULL, tx); |
13fe0198 MA |
2747 | |
2748 | VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone)); | |
2749 | ||
2750 | dsl_dataset_clone_swap_sync_impl(clone, ds, tx); | |
2751 | dsl_dataset_zero_zil(ds, tx); | |
2752 | ||
2753 | dsl_destroy_head_sync_impl(clone, tx); | |
2754 | ||
2755 | dsl_dataset_rele(clone, FTAG); | |
2756 | dsl_dataset_rele(ds, FTAG); | |
2757 | } | |
2758 | ||
831baf06 | 2759 | /* |
46ba1e59 MA |
2760 | * Rolls back the given filesystem or volume to the most recent snapshot. |
2761 | * The name of the most recent snapshot will be returned under key "target" | |
2762 | * in the result nvlist. | |
831baf06 | 2763 | * |
46ba1e59 | 2764 | * If owner != NULL: |
831baf06 KW |
2765 | * - The existing dataset MUST be owned by the specified owner at entry |
2766 | * - Upon return, dataset will still be held by the same owner, whether we | |
2767 | * succeed or not. | |
2768 | * | |
2769 | * This mode is required any time the existing filesystem is mounted. See | |
2770 | * notes above zfs_suspend_fs() for further details. | |
2771 | */ | |
13fe0198 | 2772 | int |
8ca78ab0 AG |
2773 | dsl_dataset_rollback(const char *fsname, const char *tosnap, void *owner, |
2774 | nvlist_t *result) | |
13fe0198 | 2775 | { |
831baf06 KW |
2776 | dsl_dataset_rollback_arg_t ddra; |
2777 | ||
2778 | ddra.ddra_fsname = fsname; | |
8ca78ab0 | 2779 | ddra.ddra_tosnap = tosnap; |
831baf06 | 2780 | ddra.ddra_owner = owner; |
46ba1e59 | 2781 | ddra.ddra_result = result; |
831baf06 | 2782 | |
13fe0198 | 2783 | return (dsl_sync_task(fsname, dsl_dataset_rollback_check, |
3d45fdd6 MA |
2784 | dsl_dataset_rollback_sync, &ddra, |
2785 | 1, ZFS_SPACE_CHECK_RESERVED)); | |
34dc7c2f BB |
2786 | } |
2787 | ||
b128c09f BB |
2788 | struct promotenode { |
2789 | list_node_t link; | |
2790 | dsl_dataset_t *ds; | |
2791 | }; | |
2792 | ||
b128c09f | 2793 | static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep); |
13fe0198 MA |
2794 | static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, |
2795 | void *tag); | |
2796 | static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag); | |
b128c09f | 2797 | |
d99a0153 | 2798 | int |
13fe0198 | 2799 | dsl_dataset_promote_check(void *arg, dmu_tx_t *tx) |
34dc7c2f | 2800 | { |
13fe0198 MA |
2801 | dsl_dataset_promote_arg_t *ddpa = arg; |
2802 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
2803 | dsl_dataset_t *hds; | |
2804 | struct promotenode *snap; | |
2805 | dsl_dataset_t *origin_ds; | |
34dc7c2f | 2806 | int err; |
428870ff | 2807 | uint64_t unused; |
788eb90c | 2808 | uint64_t ss_mv_cnt; |
fec41709 | 2809 | size_t max_snap_len; |
d99a0153 | 2810 | boolean_t conflicting_snaps; |
34dc7c2f | 2811 | |
13fe0198 MA |
2812 | err = promote_hold(ddpa, dp, FTAG); |
2813 | if (err != 0) | |
2814 | return (err); | |
34dc7c2f | 2815 | |
13fe0198 | 2816 | hds = ddpa->ddpa_clone; |
fec41709 | 2817 | max_snap_len = MAXNAMELEN - strlen(ddpa->ddpa_clonename) - 1; |
34dc7c2f | 2818 | |
d683ddbb | 2819 | if (dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE) { |
13fe0198 | 2820 | promote_rele(ddpa, FTAG); |
2e528b49 | 2821 | return (SET_ERROR(EXDEV)); |
13fe0198 MA |
2822 | } |
2823 | ||
b5256303 TC |
2824 | snap = list_head(&ddpa->shared_snaps); |
2825 | if (snap == NULL) { | |
2826 | err = SET_ERROR(ENOENT); | |
2827 | goto out; | |
2828 | } | |
2829 | origin_ds = snap->ds; | |
2830 | ||
2831 | /* | |
2832 | * Encrypted clones share a DSL Crypto Key with their origin's dsl dir. | |
2833 | * When doing a promote we must make sure the encryption root for | |
2834 | * both the target and the target's origin does not change to avoid | |
2835 | * needing to rewrap encryption keys | |
2836 | */ | |
2837 | err = dsl_dataset_promote_crypt_check(hds->ds_dir, origin_ds->ds_dir); | |
2838 | if (err != 0) | |
2839 | goto out; | |
2840 | ||
13fe0198 MA |
2841 | /* |
2842 | * Compute and check the amount of space to transfer. Since this is | |
2843 | * so expensive, don't do the preliminary check. | |
2844 | */ | |
2845 | if (!dmu_tx_is_syncing(tx)) { | |
2846 | promote_rele(ddpa, FTAG); | |
2847 | return (0); | |
2848 | } | |
2849 | ||
34dc7c2f | 2850 | /* compute origin's new unique space */ |
13fe0198 | 2851 | snap = list_tail(&ddpa->clone_snaps); |
cbce5813 | 2852 | ASSERT(snap != NULL); |
d683ddbb JG |
2853 | ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, |
2854 | origin_ds->ds_object); | |
428870ff | 2855 | dsl_deadlist_space_range(&snap->ds->ds_deadlist, |
d683ddbb | 2856 | dsl_dataset_phys(origin_ds)->ds_prev_snap_txg, UINT64_MAX, |
13fe0198 | 2857 | &ddpa->unique, &unused, &unused); |
34dc7c2f | 2858 | |
b128c09f BB |
2859 | /* |
2860 | * Walk the snapshots that we are moving | |
2861 | * | |
2862 | * Compute space to transfer. Consider the incremental changes | |
13fe0198 | 2863 | * to used by each snapshot: |
b128c09f BB |
2864 | * (my used) = (prev's used) + (blocks born) - (blocks killed) |
2865 | * So each snapshot gave birth to: | |
2866 | * (blocks born) = (my used) - (prev's used) + (blocks killed) | |
2867 | * So a sequence would look like: | |
2868 | * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0) | |
2869 | * Which simplifies to: | |
2870 | * uN + kN + kN-1 + ... + k1 + k0 | |
2871 | * Note however, if we stop before we reach the ORIGIN we get: | |
2872 | * uN + kN + kN-1 + ... + kM - uM-1 | |
2873 | */ | |
d99a0153 | 2874 | conflicting_snaps = B_FALSE; |
788eb90c | 2875 | ss_mv_cnt = 0; |
d683ddbb JG |
2876 | ddpa->used = dsl_dataset_phys(origin_ds)->ds_referenced_bytes; |
2877 | ddpa->comp = dsl_dataset_phys(origin_ds)->ds_compressed_bytes; | |
2878 | ddpa->uncomp = dsl_dataset_phys(origin_ds)->ds_uncompressed_bytes; | |
13fe0198 MA |
2879 | for (snap = list_head(&ddpa->shared_snaps); snap; |
2880 | snap = list_next(&ddpa->shared_snaps, snap)) { | |
34dc7c2f | 2881 | uint64_t val, dlused, dlcomp, dluncomp; |
b128c09f | 2882 | dsl_dataset_t *ds = snap->ds; |
34dc7c2f | 2883 | |
788eb90c JJ |
2884 | ss_mv_cnt++; |
2885 | ||
13fe0198 MA |
2886 | /* |
2887 | * If there are long holds, we won't be able to evict | |
2888 | * the objset. | |
2889 | */ | |
2890 | if (dsl_dataset_long_held(ds)) { | |
2e528b49 | 2891 | err = SET_ERROR(EBUSY); |
13fe0198 MA |
2892 | goto out; |
2893 | } | |
2894 | ||
34dc7c2f | 2895 | /* Check that the snapshot name does not conflict */ |
13fe0198 | 2896 | VERIFY0(dsl_dataset_get_snapname(ds)); |
fec41709 AG |
2897 | if (strlen(ds->ds_snapname) >= max_snap_len) { |
2898 | err = SET_ERROR(ENAMETOOLONG); | |
2899 | goto out; | |
2900 | } | |
b128c09f | 2901 | err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val); |
428870ff | 2902 | if (err == 0) { |
d99a0153 CW |
2903 | fnvlist_add_boolean(ddpa->err_ds, |
2904 | snap->ds->ds_snapname); | |
2905 | conflicting_snaps = B_TRUE; | |
2906 | } else if (err != ENOENT) { | |
428870ff BB |
2907 | goto out; |
2908 | } | |
34dc7c2f | 2909 | |
b128c09f | 2910 | /* The very first snapshot does not have a deadlist */ |
d683ddbb | 2911 | if (dsl_dataset_phys(ds)->ds_prev_snap_obj == 0) |
b128c09f | 2912 | continue; |
34dc7c2f | 2913 | |
428870ff BB |
2914 | dsl_deadlist_space(&ds->ds_deadlist, |
2915 | &dlused, &dlcomp, &dluncomp); | |
13fe0198 MA |
2916 | ddpa->used += dlused; |
2917 | ddpa->comp += dlcomp; | |
2918 | ddpa->uncomp += dluncomp; | |
b128c09f | 2919 | } |
34dc7c2f | 2920 | |
d99a0153 CW |
2921 | /* |
2922 | * In order to return the full list of conflicting snapshots, we check | |
2923 | * whether there was a conflict after traversing all of them. | |
2924 | */ | |
2925 | if (conflicting_snaps) { | |
2926 | err = SET_ERROR(EEXIST); | |
2927 | goto out; | |
2928 | } | |
2929 | ||
b128c09f BB |
2930 | /* |
2931 | * If we are a clone of a clone then we never reached ORIGIN, | |
2932 | * so we need to subtract out the clone origin's used space. | |
2933 | */ | |
13fe0198 | 2934 | if (ddpa->origin_origin) { |
d683ddbb JG |
2935 | ddpa->used -= |
2936 | dsl_dataset_phys(ddpa->origin_origin)->ds_referenced_bytes; | |
2937 | ddpa->comp -= | |
2938 | dsl_dataset_phys(ddpa->origin_origin)->ds_compressed_bytes; | |
13fe0198 | 2939 | ddpa->uncomp -= |
d683ddbb JG |
2940 | dsl_dataset_phys(ddpa->origin_origin)-> |
2941 | ds_uncompressed_bytes; | |
34dc7c2f BB |
2942 | } |
2943 | ||
788eb90c | 2944 | /* Check that there is enough space and limit headroom here */ |
b128c09f | 2945 | err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir, |
788eb90c | 2946 | 0, ss_mv_cnt, ddpa->used, ddpa->cr); |
13fe0198 MA |
2947 | if (err != 0) |
2948 | goto out; | |
34dc7c2f | 2949 | |
b128c09f BB |
2950 | /* |
2951 | * Compute the amounts of space that will be used by snapshots | |
2952 | * after the promotion (for both origin and clone). For each, | |
2953 | * it is the amount of space that will be on all of their | |
2954 | * deadlists (that was not born before their new origin). | |
2955 | */ | |
d683ddbb | 2956 | if (dsl_dir_phys(hds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) { |
b128c09f BB |
2957 | uint64_t space; |
2958 | ||
2959 | /* | |
2960 | * Note, typically this will not be a clone of a clone, | |
428870ff BB |
2961 | * so dd_origin_txg will be < TXG_INITIAL, so |
2962 | * these snaplist_space() -> dsl_deadlist_space_range() | |
b128c09f BB |
2963 | * calls will be fast because they do not have to |
2964 | * iterate over all bps. | |
2965 | */ | |
13fe0198 | 2966 | snap = list_head(&ddpa->origin_snaps); |
0a8f18f9 | 2967 | if (snap == NULL) { |
2968 | err = SET_ERROR(ENOENT); | |
2969 | goto out; | |
2970 | } | |
13fe0198 MA |
2971 | err = snaplist_space(&ddpa->shared_snaps, |
2972 | snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap); | |
2973 | if (err != 0) | |
2974 | goto out; | |
b128c09f | 2975 | |
13fe0198 | 2976 | err = snaplist_space(&ddpa->clone_snaps, |
428870ff | 2977 | snap->ds->ds_dir->dd_origin_txg, &space); |
13fe0198 MA |
2978 | if (err != 0) |
2979 | goto out; | |
2980 | ddpa->cloneusedsnap += space; | |
b128c09f | 2981 | } |
d683ddbb JG |
2982 | if (dsl_dir_phys(origin_ds->ds_dir)->dd_flags & |
2983 | DD_FLAG_USED_BREAKDOWN) { | |
13fe0198 | 2984 | err = snaplist_space(&ddpa->origin_snaps, |
d683ddbb JG |
2985 | dsl_dataset_phys(origin_ds)->ds_creation_txg, |
2986 | &ddpa->originusedsnap); | |
13fe0198 MA |
2987 | if (err != 0) |
2988 | goto out; | |
b128c09f BB |
2989 | } |
2990 | ||
428870ff | 2991 | out: |
13fe0198 | 2992 | promote_rele(ddpa, FTAG); |
428870ff | 2993 | return (err); |
34dc7c2f BB |
2994 | } |
2995 | ||
d99a0153 | 2996 | void |
13fe0198 | 2997 | dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 2998 | { |
13fe0198 MA |
2999 | dsl_dataset_promote_arg_t *ddpa = arg; |
3000 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3001 | dsl_dataset_t *hds; | |
3002 | struct promotenode *snap; | |
3003 | dsl_dataset_t *origin_ds; | |
b128c09f | 3004 | dsl_dataset_t *origin_head; |
13fe0198 | 3005 | dsl_dir_t *dd; |
34dc7c2f | 3006 | dsl_dir_t *odd = NULL; |
b128c09f BB |
3007 | uint64_t oldnext_obj; |
3008 | int64_t delta; | |
34dc7c2f | 3009 | |
13fe0198 MA |
3010 | VERIFY0(promote_hold(ddpa, dp, FTAG)); |
3011 | hds = ddpa->ddpa_clone; | |
3012 | ||
d683ddbb | 3013 | ASSERT0(dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE); |
13fe0198 MA |
3014 | |
3015 | snap = list_head(&ddpa->shared_snaps); | |
3016 | origin_ds = snap->ds; | |
3017 | dd = hds->ds_dir; | |
34dc7c2f | 3018 | |
13fe0198 | 3019 | snap = list_head(&ddpa->origin_snaps); |
b128c09f BB |
3020 | origin_head = snap->ds; |
3021 | ||
34dc7c2f BB |
3022 | /* |
3023 | * We need to explicitly open odd, since origin_ds's dd will be | |
3024 | * changing. | |
3025 | */ | |
13fe0198 | 3026 | VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object, |
34dc7c2f BB |
3027 | NULL, FTAG, &odd)); |
3028 | ||
b5256303 TC |
3029 | dsl_dataset_promote_crypt_sync(hds->ds_dir, odd, tx); |
3030 | ||
b128c09f BB |
3031 | /* change origin's next snap */ |
3032 | dmu_buf_will_dirty(origin_ds->ds_dbuf, tx); | |
d683ddbb | 3033 | oldnext_obj = dsl_dataset_phys(origin_ds)->ds_next_snap_obj; |
13fe0198 | 3034 | snap = list_tail(&ddpa->clone_snaps); |
d683ddbb JG |
3035 | ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==, |
3036 | origin_ds->ds_object); | |
3037 | dsl_dataset_phys(origin_ds)->ds_next_snap_obj = snap->ds->ds_object; | |
b128c09f BB |
3038 | |
3039 | /* change the origin's next clone */ | |
d683ddbb | 3040 | if (dsl_dataset_phys(origin_ds)->ds_next_clones_obj) { |
13fe0198 MA |
3041 | dsl_dataset_remove_from_next_clones(origin_ds, |
3042 | snap->ds->ds_object, tx); | |
3043 | VERIFY0(zap_add_int(dp->dp_meta_objset, | |
d683ddbb | 3044 | dsl_dataset_phys(origin_ds)->ds_next_clones_obj, |
b128c09f BB |
3045 | oldnext_obj, tx)); |
3046 | } | |
3047 | ||
3048 | /* change origin */ | |
3049 | dmu_buf_will_dirty(dd->dd_dbuf, tx); | |
d683ddbb JG |
3050 | ASSERT3U(dsl_dir_phys(dd)->dd_origin_obj, ==, origin_ds->ds_object); |
3051 | dsl_dir_phys(dd)->dd_origin_obj = dsl_dir_phys(odd)->dd_origin_obj; | |
428870ff | 3052 | dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg; |
b128c09f | 3053 | dmu_buf_will_dirty(odd->dd_dbuf, tx); |
d683ddbb | 3054 | dsl_dir_phys(odd)->dd_origin_obj = origin_ds->ds_object; |
428870ff | 3055 | origin_head->ds_dir->dd_origin_txg = |
d683ddbb | 3056 | dsl_dataset_phys(origin_ds)->ds_creation_txg; |
428870ff BB |
3057 | |
3058 | /* change dd_clone entries */ | |
3059 | if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { | |
13fe0198 | 3060 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3061 | dsl_dir_phys(odd)->dd_clones, hds->ds_object, tx)); |
13fe0198 | 3062 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3063 | dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, |
428870ff BB |
3064 | hds->ds_object, tx)); |
3065 | ||
13fe0198 | 3066 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3067 | dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones, |
428870ff | 3068 | origin_head->ds_object, tx)); |
d683ddbb JG |
3069 | if (dsl_dir_phys(dd)->dd_clones == 0) { |
3070 | dsl_dir_phys(dd)->dd_clones = | |
3071 | zap_create(dp->dp_meta_objset, DMU_OT_DSL_CLONES, | |
3072 | DMU_OT_NONE, 0, tx); | |
428870ff | 3073 | } |
13fe0198 | 3074 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3075 | dsl_dir_phys(dd)->dd_clones, origin_head->ds_object, tx)); |
428870ff | 3076 | } |
34dc7c2f | 3077 | |
b128c09f | 3078 | /* move snapshots to this dir */ |
13fe0198 MA |
3079 | for (snap = list_head(&ddpa->shared_snaps); snap; |
3080 | snap = list_next(&ddpa->shared_snaps, snap)) { | |
b128c09f BB |
3081 | dsl_dataset_t *ds = snap->ds; |
3082 | ||
13fe0198 MA |
3083 | /* |
3084 | * Property callbacks are registered to a particular | |
3085 | * dsl_dir. Since ours is changing, evict the objset | |
3086 | * so that they will be unregistered from the old dsl_dir. | |
3087 | */ | |
428870ff BB |
3088 | if (ds->ds_objset) { |
3089 | dmu_objset_evict(ds->ds_objset); | |
3090 | ds->ds_objset = NULL; | |
b128c09f | 3091 | } |
13fe0198 | 3092 | |
34dc7c2f | 3093 | /* move snap name entry */ |
13fe0198 MA |
3094 | VERIFY0(dsl_dataset_get_snapname(ds)); |
3095 | VERIFY0(dsl_dataset_snap_remove(origin_head, | |
788eb90c | 3096 | ds->ds_snapname, tx, B_TRUE)); |
13fe0198 | 3097 | VERIFY0(zap_add(dp->dp_meta_objset, |
d683ddbb | 3098 | dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname, |
34dc7c2f | 3099 | 8, 1, &ds->ds_object, tx)); |
788eb90c JJ |
3100 | dsl_fs_ss_count_adjust(hds->ds_dir, 1, |
3101 | DD_FIELD_SNAPSHOT_COUNT, tx); | |
428870ff | 3102 | |
34dc7c2f BB |
3103 | /* change containing dsl_dir */ |
3104 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
d683ddbb JG |
3105 | ASSERT3U(dsl_dataset_phys(ds)->ds_dir_obj, ==, odd->dd_object); |
3106 | dsl_dataset_phys(ds)->ds_dir_obj = dd->dd_object; | |
34dc7c2f | 3107 | ASSERT3P(ds->ds_dir, ==, odd); |
13fe0198 MA |
3108 | dsl_dir_rele(ds->ds_dir, ds); |
3109 | VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object, | |
34dc7c2f BB |
3110 | NULL, ds, &ds->ds_dir)); |
3111 | ||
428870ff | 3112 | /* move any clone references */ |
d683ddbb | 3113 | if (dsl_dataset_phys(ds)->ds_next_clones_obj && |
428870ff BB |
3114 | spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { |
3115 | zap_cursor_t zc; | |
3116 | zap_attribute_t za; | |
3117 | ||
3118 | for (zap_cursor_init(&zc, dp->dp_meta_objset, | |
d683ddbb | 3119 | dsl_dataset_phys(ds)->ds_next_clones_obj); |
428870ff | 3120 | zap_cursor_retrieve(&zc, &za) == 0; |
13fe0198 MA |
3121 | zap_cursor_advance(&zc)) { |
3122 | dsl_dataset_t *cnds; | |
3123 | uint64_t o; | |
34dc7c2f | 3124 | |
13fe0198 MA |
3125 | if (za.za_first_integer == oldnext_obj) { |
3126 | /* | |
3127 | * We've already moved the | |
3128 | * origin's reference. | |
3129 | */ | |
3130 | continue; | |
3131 | } | |
34dc7c2f | 3132 | |
13fe0198 MA |
3133 | VERIFY0(dsl_dataset_hold_obj(dp, |
3134 | za.za_first_integer, FTAG, &cnds)); | |
d683ddbb JG |
3135 | o = dsl_dir_phys(cnds->ds_dir)-> |
3136 | dd_head_dataset_obj; | |
34dc7c2f | 3137 | |
13fe0198 | 3138 | VERIFY0(zap_remove_int(dp->dp_meta_objset, |
d683ddbb | 3139 | dsl_dir_phys(odd)->dd_clones, o, tx)); |
13fe0198 | 3140 | VERIFY0(zap_add_int(dp->dp_meta_objset, |
d683ddbb | 3141 | dsl_dir_phys(dd)->dd_clones, o, tx)); |
13fe0198 MA |
3142 | dsl_dataset_rele(cnds, FTAG); |
3143 | } | |
3144 | zap_cursor_fini(&zc); | |
3145 | } | |
34dc7c2f | 3146 | |
13fe0198 | 3147 | ASSERT(!dsl_prop_hascb(ds)); |
34dc7c2f BB |
3148 | } |
3149 | ||
34dc7c2f | 3150 | /* |
13fe0198 MA |
3151 | * Change space accounting. |
3152 | * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either | |
3153 | * both be valid, or both be 0 (resulting in delta == 0). This | |
3154 | * is true for each of {clone,origin} independently. | |
34dc7c2f | 3155 | */ |
570827e1 | 3156 | |
13fe0198 | 3157 | delta = ddpa->cloneusedsnap - |
d683ddbb | 3158 | dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP]; |
13fe0198 MA |
3159 | ASSERT3S(delta, >=, 0); |
3160 | ASSERT3U(ddpa->used, >=, delta); | |
3161 | dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx); | |
3162 | dsl_dir_diduse_space(dd, DD_USED_HEAD, | |
3163 | ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx); | |
34dc7c2f | 3164 | |
13fe0198 | 3165 | delta = ddpa->originusedsnap - |
d683ddbb | 3166 | dsl_dir_phys(odd)->dd_used_breakdown[DD_USED_SNAP]; |
13fe0198 MA |
3167 | ASSERT3S(delta, <=, 0); |
3168 | ASSERT3U(ddpa->used, >=, -delta); | |
3169 | dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx); | |
3170 | dsl_dir_diduse_space(odd, DD_USED_HEAD, | |
3171 | -ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx); | |
3172 | ||
d683ddbb | 3173 | dsl_dataset_phys(origin_ds)->ds_unique_bytes = ddpa->unique; |
13fe0198 MA |
3174 | |
3175 | /* log history record */ | |
3176 | spa_history_log_internal_ds(hds, "promote", tx, ""); | |
3177 | ||
3178 | dsl_dir_rele(odd, FTAG); | |
3179 | promote_rele(ddpa, FTAG); | |
34dc7c2f BB |
3180 | } |
3181 | ||
13fe0198 MA |
3182 | /* |
3183 | * Make a list of dsl_dataset_t's for the snapshots between first_obj | |
3184 | * (exclusive) and last_obj (inclusive). The list will be in reverse | |
3185 | * order (last_obj will be the list_head()). If first_obj == 0, do all | |
3186 | * snapshots back to this dataset's origin. | |
3187 | */ | |
34dc7c2f | 3188 | static int |
13fe0198 MA |
3189 | snaplist_make(dsl_pool_t *dp, |
3190 | uint64_t first_obj, uint64_t last_obj, list_t *l, void *tag) | |
34dc7c2f | 3191 | { |
13fe0198 | 3192 | uint64_t obj = last_obj; |
34dc7c2f | 3193 | |
13fe0198 MA |
3194 | list_create(l, sizeof (struct promotenode), |
3195 | offsetof(struct promotenode, link)); | |
34dc7c2f | 3196 | |
13fe0198 MA |
3197 | while (obj != first_obj) { |
3198 | dsl_dataset_t *ds; | |
3199 | struct promotenode *snap; | |
3200 | int err; | |
428870ff | 3201 | |
13fe0198 MA |
3202 | err = dsl_dataset_hold_obj(dp, obj, tag, &ds); |
3203 | ASSERT(err != ENOENT); | |
3204 | if (err != 0) | |
3205 | return (err); | |
34dc7c2f | 3206 | |
13fe0198 | 3207 | if (first_obj == 0) |
d683ddbb | 3208 | first_obj = dsl_dir_phys(ds->ds_dir)->dd_origin_obj; |
13fe0198 | 3209 | |
79c76d5b | 3210 | snap = kmem_alloc(sizeof (*snap), KM_SLEEP); |
13fe0198 MA |
3211 | snap->ds = ds; |
3212 | list_insert_tail(l, snap); | |
d683ddbb | 3213 | obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
13fe0198 | 3214 | } |
34dc7c2f BB |
3215 | |
3216 | return (0); | |
3217 | } | |
3218 | ||
13fe0198 MA |
3219 | static int |
3220 | snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep) | |
34dc7c2f | 3221 | { |
13fe0198 | 3222 | struct promotenode *snap; |
6f1ffb06 | 3223 | |
13fe0198 MA |
3224 | *spacep = 0; |
3225 | for (snap = list_head(l); snap; snap = list_next(l, snap)) { | |
3226 | uint64_t used, comp, uncomp; | |
3227 | dsl_deadlist_space_range(&snap->ds->ds_deadlist, | |
3228 | mintxg, UINT64_MAX, &used, &comp, &uncomp); | |
3229 | *spacep += used; | |
428870ff | 3230 | } |
13fe0198 | 3231 | return (0); |
34dc7c2f BB |
3232 | } |
3233 | ||
13fe0198 MA |
3234 | static void |
3235 | snaplist_destroy(list_t *l, void *tag) | |
34dc7c2f | 3236 | { |
13fe0198 | 3237 | struct promotenode *snap; |
428870ff | 3238 | |
13fe0198 MA |
3239 | if (l == NULL || !list_link_active(&l->list_head)) |
3240 | return; | |
34dc7c2f | 3241 | |
13fe0198 MA |
3242 | while ((snap = list_tail(l)) != NULL) { |
3243 | list_remove(l, snap); | |
3244 | dsl_dataset_rele(snap->ds, tag); | |
3245 | kmem_free(snap, sizeof (*snap)); | |
3246 | } | |
3247 | list_destroy(l); | |
34dc7c2f BB |
3248 | } |
3249 | ||
3250 | static int | |
13fe0198 | 3251 | promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag) |
34dc7c2f | 3252 | { |
13fe0198 MA |
3253 | int error; |
3254 | dsl_dir_t *dd; | |
3255 | struct promotenode *snap; | |
34dc7c2f | 3256 | |
13fe0198 MA |
3257 | error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag, |
3258 | &ddpa->ddpa_clone); | |
3259 | if (error != 0) | |
3260 | return (error); | |
3261 | dd = ddpa->ddpa_clone->ds_dir; | |
34dc7c2f | 3262 | |
0c66c32d | 3263 | if (ddpa->ddpa_clone->ds_is_snapshot || |
13fe0198 MA |
3264 | !dsl_dir_is_clone(dd)) { |
3265 | dsl_dataset_rele(ddpa->ddpa_clone, tag); | |
2e528b49 | 3266 | return (SET_ERROR(EINVAL)); |
13fe0198 | 3267 | } |
34dc7c2f | 3268 | |
d683ddbb | 3269 | error = snaplist_make(dp, 0, dsl_dir_phys(dd)->dd_origin_obj, |
13fe0198 MA |
3270 | &ddpa->shared_snaps, tag); |
3271 | if (error != 0) | |
3272 | goto out; | |
34dc7c2f | 3273 | |
13fe0198 MA |
3274 | error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object, |
3275 | &ddpa->clone_snaps, tag); | |
3276 | if (error != 0) | |
3277 | goto out; | |
34dc7c2f | 3278 | |
13fe0198 | 3279 | snap = list_head(&ddpa->shared_snaps); |
d683ddbb JG |
3280 | ASSERT3U(snap->ds->ds_object, ==, dsl_dir_phys(dd)->dd_origin_obj); |
3281 | error = snaplist_make(dp, dsl_dir_phys(dd)->dd_origin_obj, | |
3282 | dsl_dir_phys(snap->ds->ds_dir)->dd_head_dataset_obj, | |
13fe0198 MA |
3283 | &ddpa->origin_snaps, tag); |
3284 | if (error != 0) | |
3285 | goto out; | |
d164b209 | 3286 | |
d683ddbb | 3287 | if (dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj != 0) { |
13fe0198 | 3288 | error = dsl_dataset_hold_obj(dp, |
d683ddbb | 3289 | dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj, |
13fe0198 MA |
3290 | tag, &ddpa->origin_origin); |
3291 | if (error != 0) | |
3292 | goto out; | |
d164b209 | 3293 | } |
13fe0198 MA |
3294 | out: |
3295 | if (error != 0) | |
3296 | promote_rele(ddpa, tag); | |
3297 | return (error); | |
34dc7c2f BB |
3298 | } |
3299 | ||
34dc7c2f | 3300 | static void |
13fe0198 | 3301 | promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag) |
34dc7c2f | 3302 | { |
13fe0198 MA |
3303 | snaplist_destroy(&ddpa->shared_snaps, tag); |
3304 | snaplist_destroy(&ddpa->clone_snaps, tag); | |
3305 | snaplist_destroy(&ddpa->origin_snaps, tag); | |
3306 | if (ddpa->origin_origin != NULL) | |
3307 | dsl_dataset_rele(ddpa->origin_origin, tag); | |
3308 | dsl_dataset_rele(ddpa->ddpa_clone, tag); | |
3309 | } | |
428870ff | 3310 | |
13fe0198 MA |
3311 | /* |
3312 | * Promote a clone. | |
3313 | * | |
3314 | * If it fails due to a conflicting snapshot name, "conflsnap" will be filled | |
eca7b760 | 3315 | * in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.) |
13fe0198 MA |
3316 | */ |
3317 | int | |
3318 | dsl_dataset_promote(const char *name, char *conflsnap) | |
3319 | { | |
3320 | dsl_dataset_promote_arg_t ddpa = { 0 }; | |
3321 | uint64_t numsnaps; | |
3322 | int error; | |
d99a0153 | 3323 | nvpair_t *snap_pair; |
13fe0198 | 3324 | objset_t *os; |
34dc7c2f | 3325 | |
13fe0198 MA |
3326 | /* |
3327 | * We will modify space proportional to the number of | |
3328 | * snapshots. Compute numsnaps. | |
3329 | */ | |
3330 | error = dmu_objset_hold(name, FTAG, &os); | |
3331 | if (error != 0) | |
3332 | return (error); | |
3333 | error = zap_count(dmu_objset_pool(os)->dp_meta_objset, | |
d683ddbb JG |
3334 | dsl_dataset_phys(dmu_objset_ds(os))->ds_snapnames_zapobj, |
3335 | &numsnaps); | |
13fe0198 MA |
3336 | dmu_objset_rele(os, FTAG); |
3337 | if (error != 0) | |
3338 | return (error); | |
34dc7c2f | 3339 | |
13fe0198 | 3340 | ddpa.ddpa_clonename = name; |
d99a0153 | 3341 | ddpa.err_ds = fnvlist_alloc(); |
788eb90c | 3342 | ddpa.cr = CRED(); |
6f1ffb06 | 3343 | |
d99a0153 | 3344 | error = dsl_sync_task(name, dsl_dataset_promote_check, |
3d45fdd6 | 3345 | dsl_dataset_promote_sync, &ddpa, |
d99a0153 CW |
3346 | 2 + numsnaps, ZFS_SPACE_CHECK_RESERVED); |
3347 | ||
3348 | /* | |
3349 | * Return the first conflicting snapshot found. | |
3350 | */ | |
3351 | snap_pair = nvlist_next_nvpair(ddpa.err_ds, NULL); | |
3352 | if (snap_pair != NULL && conflsnap != NULL) | |
3353 | (void) strcpy(conflsnap, nvpair_name(snap_pair)); | |
3354 | ||
3355 | fnvlist_free(ddpa.err_ds); | |
3356 | return (error); | |
34dc7c2f BB |
3357 | } |
3358 | ||
3359 | int | |
13fe0198 | 3360 | dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone, |
831baf06 | 3361 | dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx) |
34dc7c2f | 3362 | { |
8c62a0d0 DM |
3363 | /* |
3364 | * "slack" factor for received datasets with refquota set on them. | |
3365 | * See the bottom of this function for details on its use. | |
3366 | */ | |
bf18fd89 CIK |
3367 | uint64_t refquota_slack = (uint64_t)DMU_MAX_ACCESS * |
3368 | spa_asize_inflation; | |
13fe0198 | 3369 | int64_t unused_refres_delta; |
34dc7c2f | 3370 | |
13fe0198 | 3371 | /* they should both be heads */ |
0c66c32d JG |
3372 | if (clone->ds_is_snapshot || |
3373 | origin_head->ds_is_snapshot) | |
2e528b49 | 3374 | return (SET_ERROR(EINVAL)); |
428870ff | 3375 | |
19580676 MA |
3376 | /* if we are not forcing, the branch point should be just before them */ |
3377 | if (!force && clone->ds_prev != origin_head->ds_prev) | |
2e528b49 | 3378 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3379 | |
13fe0198 MA |
3380 | /* clone should be the clone (unless they are unrelated) */ |
3381 | if (clone->ds_prev != NULL && | |
3382 | clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap && | |
19580676 | 3383 | origin_head->ds_dir != clone->ds_prev->ds_dir) |
2e528b49 | 3384 | return (SET_ERROR(EINVAL)); |
428870ff | 3385 | |
13fe0198 MA |
3386 | /* the clone should be a child of the origin */ |
3387 | if (clone->ds_dir->dd_parent != origin_head->ds_dir) | |
2e528b49 | 3388 | return (SET_ERROR(EINVAL)); |
45d1cae3 | 3389 | |
13fe0198 | 3390 | /* origin_head shouldn't be modified unless 'force' */ |
19580676 MA |
3391 | if (!force && |
3392 | dsl_dataset_modified_since_snap(origin_head, origin_head->ds_prev)) | |
2e528b49 | 3393 | return (SET_ERROR(ETXTBSY)); |
572e2857 | 3394 | |
13fe0198 | 3395 | /* origin_head should have no long holds (e.g. is not mounted) */ |
831baf06 | 3396 | if (dsl_dataset_handoff_check(origin_head, owner, tx)) |
2e528b49 | 3397 | return (SET_ERROR(EBUSY)); |
13fe0198 MA |
3398 | |
3399 | /* check amount of any unconsumed refreservation */ | |
3400 | unused_refres_delta = | |
3401 | (int64_t)MIN(origin_head->ds_reserved, | |
d683ddbb | 3402 | dsl_dataset_phys(origin_head)->ds_unique_bytes) - |
13fe0198 | 3403 | (int64_t)MIN(origin_head->ds_reserved, |
d683ddbb | 3404 | dsl_dataset_phys(clone)->ds_unique_bytes); |
13fe0198 MA |
3405 | |
3406 | if (unused_refres_delta > 0 && | |
3407 | unused_refres_delta > | |
3408 | dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE)) | |
2e528b49 | 3409 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3410 | |
8c62a0d0 DM |
3411 | /* |
3412 | * The clone can't be too much over the head's refquota. | |
3413 | * | |
3414 | * To ensure that the entire refquota can be used, we allow one | |
3415 | * transaction to exceed the the refquota. Therefore, this check | |
3416 | * needs to also allow for the space referenced to be more than the | |
3417 | * refquota. The maximum amount of space that one transaction can use | |
3418 | * on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this | |
3419 | * overage ensures that we are able to receive a filesystem that | |
3420 | * exceeds the refquota on the source system. | |
3421 | * | |
3422 | * So that overage is the refquota_slack we use below. | |
3423 | */ | |
13fe0198 | 3424 | if (origin_head->ds_quota != 0 && |
d683ddbb | 3425 | dsl_dataset_phys(clone)->ds_referenced_bytes > |
8c62a0d0 | 3426 | origin_head->ds_quota + refquota_slack) |
2e528b49 | 3427 | return (SET_ERROR(EDQUOT)); |
572e2857 | 3428 | |
13fe0198 | 3429 | return (0); |
572e2857 BB |
3430 | } |
3431 | ||
a1d477c2 MA |
3432 | static void |
3433 | dsl_dataset_swap_remap_deadlists(dsl_dataset_t *clone, | |
3434 | dsl_dataset_t *origin, dmu_tx_t *tx) | |
3435 | { | |
3436 | uint64_t clone_remap_dl_obj, origin_remap_dl_obj; | |
3437 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3438 | ||
3439 | ASSERT(dsl_pool_sync_context(dp)); | |
3440 | ||
3441 | clone_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(clone); | |
3442 | origin_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(origin); | |
3443 | ||
3444 | if (clone_remap_dl_obj != 0) { | |
3445 | dsl_deadlist_close(&clone->ds_remap_deadlist); | |
3446 | dsl_dataset_unset_remap_deadlist_object(clone, tx); | |
3447 | } | |
3448 | if (origin_remap_dl_obj != 0) { | |
3449 | dsl_deadlist_close(&origin->ds_remap_deadlist); | |
3450 | dsl_dataset_unset_remap_deadlist_object(origin, tx); | |
3451 | } | |
3452 | ||
3453 | if (clone_remap_dl_obj != 0) { | |
3454 | dsl_dataset_set_remap_deadlist_object(origin, | |
3455 | clone_remap_dl_obj, tx); | |
3456 | dsl_deadlist_open(&origin->ds_remap_deadlist, | |
3457 | dp->dp_meta_objset, clone_remap_dl_obj); | |
3458 | } | |
3459 | if (origin_remap_dl_obj != 0) { | |
3460 | dsl_dataset_set_remap_deadlist_object(clone, | |
3461 | origin_remap_dl_obj, tx); | |
3462 | dsl_deadlist_open(&clone->ds_remap_deadlist, | |
3463 | dp->dp_meta_objset, origin_remap_dl_obj); | |
3464 | } | |
3465 | } | |
3466 | ||
572e2857 | 3467 | void |
13fe0198 MA |
3468 | dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone, |
3469 | dsl_dataset_t *origin_head, dmu_tx_t *tx) | |
572e2857 | 3470 | { |
13fe0198 MA |
3471 | dsl_pool_t *dp = dmu_tx_pool(tx); |
3472 | int64_t unused_refres_delta; | |
428870ff | 3473 | |
13fe0198 | 3474 | ASSERT(clone->ds_reserved == 0); |
8c62a0d0 DM |
3475 | /* |
3476 | * NOTE: On DEBUG kernels there could be a race between this and | |
3477 | * the check function if spa_asize_inflation is adjusted... | |
3478 | */ | |
13fe0198 | 3479 | ASSERT(origin_head->ds_quota == 0 || |
8c62a0d0 DM |
3480 | dsl_dataset_phys(clone)->ds_unique_bytes <= origin_head->ds_quota + |
3481 | DMU_MAX_ACCESS * spa_asize_inflation); | |
19580676 | 3482 | ASSERT3P(clone->ds_prev, ==, origin_head->ds_prev); |
45d1cae3 | 3483 | |
241b5415 MA |
3484 | /* |
3485 | * Swap per-dataset feature flags. | |
3486 | */ | |
4a5d7f82 | 3487 | for (spa_feature_t f = 0; f < SPA_FEATURES; f++) { |
241b5415 MA |
3488 | if (!(spa_feature_table[f].fi_flags & |
3489 | ZFEATURE_FLAG_PER_DATASET)) { | |
3490 | ASSERT(!clone->ds_feature_inuse[f]); | |
3491 | ASSERT(!origin_head->ds_feature_inuse[f]); | |
3492 | continue; | |
3493 | } | |
3494 | ||
1c27024e DB |
3495 | boolean_t clone_inuse = clone->ds_feature_inuse[f]; |
3496 | boolean_t origin_head_inuse = origin_head->ds_feature_inuse[f]; | |
241b5415 MA |
3497 | |
3498 | if (clone_inuse) { | |
3499 | dsl_dataset_deactivate_feature(clone->ds_object, f, tx); | |
3500 | clone->ds_feature_inuse[f] = B_FALSE; | |
3501 | } | |
3502 | if (origin_head_inuse) { | |
3503 | dsl_dataset_deactivate_feature(origin_head->ds_object, | |
3504 | f, tx); | |
3505 | origin_head->ds_feature_inuse[f] = B_FALSE; | |
3506 | } | |
3507 | if (clone_inuse) { | |
3508 | dsl_dataset_activate_feature(origin_head->ds_object, | |
3509 | f, tx); | |
3510 | origin_head->ds_feature_inuse[f] = B_TRUE; | |
3511 | } | |
3512 | if (origin_head_inuse) { | |
3513 | dsl_dataset_activate_feature(clone->ds_object, f, tx); | |
3514 | clone->ds_feature_inuse[f] = B_TRUE; | |
3515 | } | |
3516 | } | |
3517 | ||
13fe0198 MA |
3518 | dmu_buf_will_dirty(clone->ds_dbuf, tx); |
3519 | dmu_buf_will_dirty(origin_head->ds_dbuf, tx); | |
45d1cae3 | 3520 | |
13fe0198 MA |
3521 | if (clone->ds_objset != NULL) { |
3522 | dmu_objset_evict(clone->ds_objset); | |
3523 | clone->ds_objset = NULL; | |
3524 | } | |
45d1cae3 | 3525 | |
13fe0198 MA |
3526 | if (origin_head->ds_objset != NULL) { |
3527 | dmu_objset_evict(origin_head->ds_objset); | |
3528 | origin_head->ds_objset = NULL; | |
45d1cae3 | 3529 | } |
45d1cae3 | 3530 | |
13fe0198 MA |
3531 | unused_refres_delta = |
3532 | (int64_t)MIN(origin_head->ds_reserved, | |
d683ddbb | 3533 | dsl_dataset_phys(origin_head)->ds_unique_bytes) - |
13fe0198 | 3534 | (int64_t)MIN(origin_head->ds_reserved, |
d683ddbb | 3535 | dsl_dataset_phys(clone)->ds_unique_bytes); |
13fe0198 MA |
3536 | |
3537 | /* | |
3538 | * Reset origin's unique bytes, if it exists. | |
3539 | */ | |
3540 | if (clone->ds_prev) { | |
3541 | dsl_dataset_t *origin = clone->ds_prev; | |
3542 | uint64_t comp, uncomp; | |
3543 | ||
3544 | dmu_buf_will_dirty(origin->ds_dbuf, tx); | |
3545 | dsl_deadlist_space_range(&clone->ds_deadlist, | |
d683ddbb JG |
3546 | dsl_dataset_phys(origin)->ds_prev_snap_txg, UINT64_MAX, |
3547 | &dsl_dataset_phys(origin)->ds_unique_bytes, &comp, &uncomp); | |
13fe0198 MA |
3548 | } |
3549 | ||
3550 | /* swap blkptrs */ | |
3551 | { | |
cc9bb3e5 GM |
3552 | rrw_enter(&clone->ds_bp_rwlock, RW_WRITER, FTAG); |
3553 | rrw_enter(&origin_head->ds_bp_rwlock, RW_WRITER, FTAG); | |
1c27024e | 3554 | blkptr_t tmp; |
d683ddbb JG |
3555 | tmp = dsl_dataset_phys(origin_head)->ds_bp; |
3556 | dsl_dataset_phys(origin_head)->ds_bp = | |
3557 | dsl_dataset_phys(clone)->ds_bp; | |
3558 | dsl_dataset_phys(clone)->ds_bp = tmp; | |
cc9bb3e5 GM |
3559 | rrw_exit(&origin_head->ds_bp_rwlock, FTAG); |
3560 | rrw_exit(&clone->ds_bp_rwlock, FTAG); | |
13fe0198 MA |
3561 | } |
3562 | ||
3563 | /* set dd_*_bytes */ | |
3564 | { | |
3565 | int64_t dused, dcomp, duncomp; | |
3566 | uint64_t cdl_used, cdl_comp, cdl_uncomp; | |
3567 | uint64_t odl_used, odl_comp, odl_uncomp; | |
3568 | ||
d683ddbb | 3569 | ASSERT3U(dsl_dir_phys(clone->ds_dir)-> |
13fe0198 MA |
3570 | dd_used_breakdown[DD_USED_SNAP], ==, 0); |
3571 | ||
3572 | dsl_deadlist_space(&clone->ds_deadlist, | |
3573 | &cdl_used, &cdl_comp, &cdl_uncomp); | |
3574 | dsl_deadlist_space(&origin_head->ds_deadlist, | |
3575 | &odl_used, &odl_comp, &odl_uncomp); | |
428870ff | 3576 | |
d683ddbb JG |
3577 | dused = dsl_dataset_phys(clone)->ds_referenced_bytes + |
3578 | cdl_used - | |
3579 | (dsl_dataset_phys(origin_head)->ds_referenced_bytes + | |
3580 | odl_used); | |
3581 | dcomp = dsl_dataset_phys(clone)->ds_compressed_bytes + | |
3582 | cdl_comp - | |
3583 | (dsl_dataset_phys(origin_head)->ds_compressed_bytes + | |
3584 | odl_comp); | |
3585 | duncomp = dsl_dataset_phys(clone)->ds_uncompressed_bytes + | |
13fe0198 | 3586 | cdl_uncomp - |
d683ddbb JG |
3587 | (dsl_dataset_phys(origin_head)->ds_uncompressed_bytes + |
3588 | odl_uncomp); | |
45d1cae3 | 3589 | |
13fe0198 MA |
3590 | dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD, |
3591 | dused, dcomp, duncomp, tx); | |
3592 | dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD, | |
3593 | -dused, -dcomp, -duncomp, tx); | |
45d1cae3 | 3594 | |
45d1cae3 | 3595 | /* |
13fe0198 MA |
3596 | * The difference in the space used by snapshots is the |
3597 | * difference in snapshot space due to the head's | |
3598 | * deadlist (since that's the only thing that's | |
3599 | * changing that affects the snapused). | |
45d1cae3 | 3600 | */ |
13fe0198 MA |
3601 | dsl_deadlist_space_range(&clone->ds_deadlist, |
3602 | origin_head->ds_dir->dd_origin_txg, UINT64_MAX, | |
3603 | &cdl_used, &cdl_comp, &cdl_uncomp); | |
3604 | dsl_deadlist_space_range(&origin_head->ds_deadlist, | |
3605 | origin_head->ds_dir->dd_origin_txg, UINT64_MAX, | |
3606 | &odl_used, &odl_comp, &odl_uncomp); | |
3607 | dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used, | |
3608 | DD_USED_HEAD, DD_USED_SNAP, tx); | |
45d1cae3 | 3609 | } |
45d1cae3 | 3610 | |
13fe0198 | 3611 | /* swap ds_*_bytes */ |
d683ddbb JG |
3612 | SWITCH64(dsl_dataset_phys(origin_head)->ds_referenced_bytes, |
3613 | dsl_dataset_phys(clone)->ds_referenced_bytes); | |
3614 | SWITCH64(dsl_dataset_phys(origin_head)->ds_compressed_bytes, | |
3615 | dsl_dataset_phys(clone)->ds_compressed_bytes); | |
3616 | SWITCH64(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes, | |
3617 | dsl_dataset_phys(clone)->ds_uncompressed_bytes); | |
3618 | SWITCH64(dsl_dataset_phys(origin_head)->ds_unique_bytes, | |
3619 | dsl_dataset_phys(clone)->ds_unique_bytes); | |
45d1cae3 | 3620 | |
13fe0198 MA |
3621 | /* apply any parent delta for change in unconsumed refreservation */ |
3622 | dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV, | |
3623 | unused_refres_delta, 0, 0, tx); | |
45d1cae3 | 3624 | |
13fe0198 MA |
3625 | /* |
3626 | * Swap deadlists. | |
3627 | */ | |
3628 | dsl_deadlist_close(&clone->ds_deadlist); | |
3629 | dsl_deadlist_close(&origin_head->ds_deadlist); | |
d683ddbb JG |
3630 | SWITCH64(dsl_dataset_phys(origin_head)->ds_deadlist_obj, |
3631 | dsl_dataset_phys(clone)->ds_deadlist_obj); | |
13fe0198 | 3632 | dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset, |
d683ddbb | 3633 | dsl_dataset_phys(clone)->ds_deadlist_obj); |
13fe0198 | 3634 | dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset, |
d683ddbb | 3635 | dsl_dataset_phys(origin_head)->ds_deadlist_obj); |
a1d477c2 | 3636 | dsl_dataset_swap_remap_deadlists(clone, origin_head, tx); |
45d1cae3 | 3637 | |
13fe0198 | 3638 | dsl_scan_ds_clone_swapped(origin_head, clone, tx); |
45d1cae3 | 3639 | |
13fe0198 MA |
3640 | spa_history_log_internal_ds(clone, "clone swap", tx, |
3641 | "parent=%s", origin_head->ds_dir->dd_myname); | |
45d1cae3 BB |
3642 | } |
3643 | ||
13fe0198 MA |
3644 | /* |
3645 | * Given a pool name and a dataset object number in that pool, | |
3646 | * return the name of that dataset. | |
3647 | */ | |
572e2857 | 3648 | int |
13fe0198 | 3649 | dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf) |
572e2857 | 3650 | { |
13fe0198 MA |
3651 | dsl_pool_t *dp; |
3652 | dsl_dataset_t *ds; | |
572e2857 BB |
3653 | int error; |
3654 | ||
13fe0198 MA |
3655 | error = dsl_pool_hold(pname, FTAG, &dp); |
3656 | if (error != 0) | |
3657 | return (error); | |
3658 | ||
3659 | error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds); | |
3660 | if (error == 0) { | |
3661 | dsl_dataset_name(ds, buf); | |
3662 | dsl_dataset_rele(ds, FTAG); | |
3663 | } | |
3664 | dsl_pool_rele(dp, FTAG); | |
572e2857 BB |
3665 | |
3666 | return (error); | |
3667 | } | |
3668 | ||
45d1cae3 | 3669 | int |
13fe0198 MA |
3670 | dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota, |
3671 | uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv) | |
45d1cae3 | 3672 | { |
13fe0198 | 3673 | int error = 0; |
45d1cae3 | 3674 | |
13fe0198 | 3675 | ASSERT3S(asize, >, 0); |
45d1cae3 | 3676 | |
13fe0198 MA |
3677 | /* |
3678 | * *ref_rsrv is the portion of asize that will come from any | |
3679 | * unconsumed refreservation space. | |
3680 | */ | |
3681 | *ref_rsrv = 0; | |
45d1cae3 | 3682 | |
13fe0198 MA |
3683 | mutex_enter(&ds->ds_lock); |
3684 | /* | |
3685 | * Make a space adjustment for reserved bytes. | |
3686 | */ | |
d683ddbb | 3687 | if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) { |
13fe0198 | 3688 | ASSERT3U(*used, >=, |
d683ddbb JG |
3689 | ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); |
3690 | *used -= | |
3691 | (ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes); | |
13fe0198 MA |
3692 | *ref_rsrv = |
3693 | asize - MIN(asize, parent_delta(ds, asize + inflight)); | |
45d1cae3 BB |
3694 | } |
3695 | ||
13fe0198 MA |
3696 | if (!check_quota || ds->ds_quota == 0) { |
3697 | mutex_exit(&ds->ds_lock); | |
3698 | return (0); | |
45d1cae3 | 3699 | } |
13fe0198 MA |
3700 | /* |
3701 | * If they are requesting more space, and our current estimate | |
3702 | * is over quota, they get to try again unless the actual | |
3703 | * on-disk is over quota and there are no pending changes (which | |
3704 | * may free up space for us). | |
3705 | */ | |
d683ddbb JG |
3706 | if (dsl_dataset_phys(ds)->ds_referenced_bytes + inflight >= |
3707 | ds->ds_quota) { | |
13fe0198 | 3708 | if (inflight > 0 || |
d683ddbb | 3709 | dsl_dataset_phys(ds)->ds_referenced_bytes < ds->ds_quota) |
2e528b49 | 3710 | error = SET_ERROR(ERESTART); |
13fe0198 | 3711 | else |
2e528b49 | 3712 | error = SET_ERROR(EDQUOT); |
45d1cae3 | 3713 | } |
13fe0198 | 3714 | mutex_exit(&ds->ds_lock); |
45d1cae3 | 3715 | |
45d1cae3 BB |
3716 | return (error); |
3717 | } | |
3718 | ||
13fe0198 MA |
3719 | typedef struct dsl_dataset_set_qr_arg { |
3720 | const char *ddsqra_name; | |
3721 | zprop_source_t ddsqra_source; | |
3722 | uint64_t ddsqra_value; | |
3723 | } dsl_dataset_set_qr_arg_t; | |
45d1cae3 | 3724 | |
13fe0198 MA |
3725 | |
3726 | /* ARGSUSED */ | |
45d1cae3 | 3727 | static int |
13fe0198 | 3728 | dsl_dataset_set_refquota_check(void *arg, dmu_tx_t *tx) |
45d1cae3 | 3729 | { |
13fe0198 MA |
3730 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3731 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
3732 | dsl_dataset_t *ds; | |
45d1cae3 | 3733 | int error; |
13fe0198 | 3734 | uint64_t newval; |
45d1cae3 | 3735 | |
13fe0198 | 3736 | if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA) |
2e528b49 | 3737 | return (SET_ERROR(ENOTSUP)); |
45d1cae3 | 3738 | |
13fe0198 MA |
3739 | error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); |
3740 | if (error != 0) | |
3741 | return (error); | |
3742 | ||
0c66c32d | 3743 | if (ds->ds_is_snapshot) { |
13fe0198 | 3744 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 3745 | return (SET_ERROR(EINVAL)); |
45d1cae3 BB |
3746 | } |
3747 | ||
13fe0198 MA |
3748 | error = dsl_prop_predict(ds->ds_dir, |
3749 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
3750 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); | |
3751 | if (error != 0) { | |
3752 | dsl_dataset_rele(ds, FTAG); | |
45d1cae3 BB |
3753 | return (error); |
3754 | } | |
3755 | ||
13fe0198 MA |
3756 | if (newval == 0) { |
3757 | dsl_dataset_rele(ds, FTAG); | |
3758 | return (0); | |
3759 | } | |
3760 | ||
d683ddbb | 3761 | if (newval < dsl_dataset_phys(ds)->ds_referenced_bytes || |
13fe0198 MA |
3762 | newval < ds->ds_reserved) { |
3763 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 3764 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3765 | } |
45d1cae3 | 3766 | |
13fe0198 | 3767 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 BB |
3768 | return (0); |
3769 | } | |
3770 | ||
13fe0198 MA |
3771 | static void |
3772 | dsl_dataset_set_refquota_sync(void *arg, dmu_tx_t *tx) | |
45d1cae3 | 3773 | { |
13fe0198 MA |
3774 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3775 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 3776 | dsl_dataset_t *ds = NULL; |
13fe0198 | 3777 | uint64_t newval; |
45d1cae3 | 3778 | |
13fe0198 | 3779 | VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); |
45d1cae3 | 3780 | |
13fe0198 MA |
3781 | dsl_prop_set_sync_impl(ds, |
3782 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), | |
3783 | ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1, | |
3784 | &ddsqra->ddsqra_value, tx); | |
45d1cae3 | 3785 | |
13fe0198 MA |
3786 | VERIFY0(dsl_prop_get_int_ds(ds, |
3787 | zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval)); | |
45d1cae3 | 3788 | |
13fe0198 MA |
3789 | if (ds->ds_quota != newval) { |
3790 | dmu_buf_will_dirty(ds->ds_dbuf, tx); | |
3791 | ds->ds_quota = newval; | |
45d1cae3 | 3792 | } |
13fe0198 | 3793 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 BB |
3794 | } |
3795 | ||
13fe0198 MA |
3796 | int |
3797 | dsl_dataset_set_refquota(const char *dsname, zprop_source_t source, | |
3798 | uint64_t refquota) | |
45d1cae3 | 3799 | { |
13fe0198 | 3800 | dsl_dataset_set_qr_arg_t ddsqra; |
428870ff | 3801 | |
13fe0198 MA |
3802 | ddsqra.ddsqra_name = dsname; |
3803 | ddsqra.ddsqra_source = source; | |
3804 | ddsqra.ddsqra_value = refquota; | |
3805 | ||
3806 | return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check, | |
d2734cce SD |
3807 | dsl_dataset_set_refquota_sync, &ddsqra, 0, |
3808 | ZFS_SPACE_CHECK_EXTRA_RESERVED)); | |
45d1cae3 BB |
3809 | } |
3810 | ||
3811 | static int | |
13fe0198 | 3812 | dsl_dataset_set_refreservation_check(void *arg, dmu_tx_t *tx) |
45d1cae3 | 3813 | { |
13fe0198 MA |
3814 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3815 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
45d1cae3 BB |
3816 | dsl_dataset_t *ds; |
3817 | int error; | |
13fe0198 | 3818 | uint64_t newval, unique; |
428870ff | 3819 | |
13fe0198 | 3820 | if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION) |
2e528b49 | 3821 | return (SET_ERROR(ENOTSUP)); |
45d1cae3 | 3822 | |
13fe0198 MA |
3823 | error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds); |
3824 | if (error != 0) | |
45d1cae3 | 3825 | return (error); |
45d1cae3 | 3826 | |
0c66c32d | 3827 | if (ds->ds_is_snapshot) { |
13fe0198 | 3828 | dsl_dataset_rele(ds, FTAG); |
2e528b49 | 3829 | return (SET_ERROR(EINVAL)); |
45d1cae3 BB |
3830 | } |
3831 | ||
13fe0198 MA |
3832 | error = dsl_prop_predict(ds->ds_dir, |
3833 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), | |
3834 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval); | |
3835 | if (error != 0) { | |
3836 | dsl_dataset_rele(ds, FTAG); | |
45d1cae3 BB |
3837 | return (error); |
3838 | } | |
3839 | ||
13fe0198 MA |
3840 | /* |
3841 | * If we are doing the preliminary check in open context, the | |
3842 | * space estimates may be inaccurate. | |
3843 | */ | |
3844 | if (!dmu_tx_is_syncing(tx)) { | |
3845 | dsl_dataset_rele(ds, FTAG); | |
3846 | return (0); | |
45d1cae3 | 3847 | } |
45d1cae3 | 3848 | |
13fe0198 MA |
3849 | mutex_enter(&ds->ds_lock); |
3850 | if (!DS_UNIQUE_IS_ACCURATE(ds)) | |
3851 | dsl_dataset_recalc_head_uniq(ds); | |
d683ddbb | 3852 | unique = dsl_dataset_phys(ds)->ds_unique_bytes; |
13fe0198 | 3853 | mutex_exit(&ds->ds_lock); |
45d1cae3 | 3854 | |
13fe0198 MA |
3855 | if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) { |
3856 | uint64_t delta = MAX(unique, newval) - | |
3857 | MAX(unique, ds->ds_reserved); | |
45d1cae3 | 3858 | |
13fe0198 MA |
3859 | if (delta > |
3860 | dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) || | |
3861 | (ds->ds_quota > 0 && newval > ds->ds_quota)) { | |
3862 | dsl_dataset_rele(ds, FTAG); | |
2e528b49 | 3863 | return (SET_ERROR(ENOSPC)); |
13fe0198 | 3864 | } |
45d1cae3 BB |
3865 | } |
3866 | ||
13fe0198 MA |
3867 | dsl_dataset_rele(ds, FTAG); |
3868 | return (0); | |
45d1cae3 BB |
3869 | } |
3870 | ||
13fe0198 MA |
3871 | void |
3872 | dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds, | |
3873 | zprop_source_t source, uint64_t value, dmu_tx_t *tx) | |
428870ff | 3874 | { |
13fe0198 MA |
3875 | uint64_t newval; |
3876 | uint64_t unique; | |
3877 | int64_t delta; | |
428870ff | 3878 | |
13fe0198 MA |
3879 | dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), |
3880 | source, sizeof (value), 1, &value, tx); | |
428870ff | 3881 | |
13fe0198 MA |
3882 | VERIFY0(dsl_prop_get_int_ds(ds, |
3883 | zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval)); | |
572e2857 | 3884 | |
13fe0198 MA |
3885 | dmu_buf_will_dirty(ds->ds_dbuf, tx); |
3886 | mutex_enter(&ds->ds_dir->dd_lock); | |
3887 | mutex_enter(&ds->ds_lock); | |
3888 | ASSERT(DS_UNIQUE_IS_ACCURATE(ds)); | |
d683ddbb | 3889 | unique = dsl_dataset_phys(ds)->ds_unique_bytes; |
13fe0198 MA |
3890 | delta = MAX(0, (int64_t)(newval - unique)) - |
3891 | MAX(0, (int64_t)(ds->ds_reserved - unique)); | |
3892 | ds->ds_reserved = newval; | |
3893 | mutex_exit(&ds->ds_lock); | |
572e2857 | 3894 | |
13fe0198 MA |
3895 | dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx); |
3896 | mutex_exit(&ds->ds_dir->dd_lock); | |
428870ff BB |
3897 | } |
3898 | ||
13fe0198 MA |
3899 | static void |
3900 | dsl_dataset_set_refreservation_sync(void *arg, dmu_tx_t *tx) | |
45d1cae3 | 3901 | { |
13fe0198 MA |
3902 | dsl_dataset_set_qr_arg_t *ddsqra = arg; |
3903 | dsl_pool_t *dp = dmu_tx_pool(tx); | |
689f093e | 3904 | dsl_dataset_t *ds = NULL; |
45d1cae3 | 3905 | |
13fe0198 MA |
3906 | VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds)); |
3907 | dsl_dataset_set_refreservation_sync_impl(ds, | |
3908 | ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx); | |
45d1cae3 | 3909 | dsl_dataset_rele(ds, FTAG); |
45d1cae3 | 3910 | } |
428870ff | 3911 | |
428870ff | 3912 | int |
13fe0198 MA |
3913 | dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source, |
3914 | uint64_t refreservation) | |
428870ff | 3915 | { |
13fe0198 | 3916 | dsl_dataset_set_qr_arg_t ddsqra; |
428870ff | 3917 | |
13fe0198 MA |
3918 | ddsqra.ddsqra_name = dsname; |
3919 | ddsqra.ddsqra_source = source; | |
3920 | ddsqra.ddsqra_value = refreservation; | |
c28b2279 | 3921 | |
13fe0198 | 3922 | return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check, |
d2734cce SD |
3923 | dsl_dataset_set_refreservation_sync, &ddsqra, 0, |
3924 | ZFS_SPACE_CHECK_EXTRA_RESERVED)); | |
13fe0198 | 3925 | } |
330d06f9 MA |
3926 | |
3927 | /* | |
3928 | * Return (in *usedp) the amount of space written in new that is not | |
3929 | * present in oldsnap. New may be a snapshot or the head. Old must be | |
3930 | * a snapshot before new, in new's filesystem (or its origin). If not then | |
3931 | * fail and return EINVAL. | |
3932 | * | |
3933 | * The written space is calculated by considering two components: First, we | |
3934 | * ignore any freed space, and calculate the written as new's used space | |
3935 | * minus old's used space. Next, we add in the amount of space that was freed | |
3936 | * between the two snapshots, thus reducing new's used space relative to old's. | |
3937 | * Specifically, this is the space that was born before old->ds_creation_txg, | |
3938 | * and freed before new (ie. on new's deadlist or a previous deadlist). | |
3939 | * | |
3940 | * space freed [---------------------] | |
3941 | * snapshots ---O-------O--------O-------O------ | |
3942 | * oldsnap new | |
3943 | */ | |
3944 | int | |
3945 | dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new, | |
3946 | uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) | |
3947 | { | |
3948 | int err = 0; | |
3949 | uint64_t snapobj; | |
3950 | dsl_pool_t *dp = new->ds_dir->dd_pool; | |
3951 | ||
13fe0198 MA |
3952 | ASSERT(dsl_pool_config_held(dp)); |
3953 | ||
330d06f9 | 3954 | *usedp = 0; |
d683ddbb JG |
3955 | *usedp += dsl_dataset_phys(new)->ds_referenced_bytes; |
3956 | *usedp -= dsl_dataset_phys(oldsnap)->ds_referenced_bytes; | |
330d06f9 MA |
3957 | |
3958 | *compp = 0; | |
d683ddbb JG |
3959 | *compp += dsl_dataset_phys(new)->ds_compressed_bytes; |
3960 | *compp -= dsl_dataset_phys(oldsnap)->ds_compressed_bytes; | |
330d06f9 MA |
3961 | |
3962 | *uncompp = 0; | |
d683ddbb JG |
3963 | *uncompp += dsl_dataset_phys(new)->ds_uncompressed_bytes; |
3964 | *uncompp -= dsl_dataset_phys(oldsnap)->ds_uncompressed_bytes; | |
330d06f9 | 3965 | |
330d06f9 MA |
3966 | snapobj = new->ds_object; |
3967 | while (snapobj != oldsnap->ds_object) { | |
3968 | dsl_dataset_t *snap; | |
3969 | uint64_t used, comp, uncomp; | |
3970 | ||
9ae529ec CS |
3971 | if (snapobj == new->ds_object) { |
3972 | snap = new; | |
3973 | } else { | |
3974 | err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap); | |
3975 | if (err != 0) | |
3976 | break; | |
3977 | } | |
330d06f9 | 3978 | |
d683ddbb JG |
3979 | if (dsl_dataset_phys(snap)->ds_prev_snap_txg == |
3980 | dsl_dataset_phys(oldsnap)->ds_creation_txg) { | |
330d06f9 MA |
3981 | /* |
3982 | * The blocks in the deadlist can not be born after | |
3983 | * ds_prev_snap_txg, so get the whole deadlist space, | |
3984 | * which is more efficient (especially for old-format | |
3985 | * deadlists). Unfortunately the deadlist code | |
3986 | * doesn't have enough information to make this | |
3987 | * optimization itself. | |
3988 | */ | |
3989 | dsl_deadlist_space(&snap->ds_deadlist, | |
3990 | &used, &comp, &uncomp); | |
3991 | } else { | |
3992 | dsl_deadlist_space_range(&snap->ds_deadlist, | |
d683ddbb | 3993 | 0, dsl_dataset_phys(oldsnap)->ds_creation_txg, |
330d06f9 MA |
3994 | &used, &comp, &uncomp); |
3995 | } | |
3996 | *usedp += used; | |
3997 | *compp += comp; | |
3998 | *uncompp += uncomp; | |
3999 | ||
4000 | /* | |
4001 | * If we get to the beginning of the chain of snapshots | |
4002 | * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap | |
4003 | * was not a snapshot of/before new. | |
4004 | */ | |
d683ddbb | 4005 | snapobj = dsl_dataset_phys(snap)->ds_prev_snap_obj; |
9ae529ec CS |
4006 | if (snap != new) |
4007 | dsl_dataset_rele(snap, FTAG); | |
330d06f9 | 4008 | if (snapobj == 0) { |
2e528b49 | 4009 | err = SET_ERROR(EINVAL); |
330d06f9 MA |
4010 | break; |
4011 | } | |
4012 | ||
4013 | } | |
330d06f9 MA |
4014 | return (err); |
4015 | } | |
4016 | ||
4017 | /* | |
4018 | * Return (in *usedp) the amount of space that will be reclaimed if firstsnap, | |
4019 | * lastsnap, and all snapshots in between are deleted. | |
4020 | * | |
4021 | * blocks that would be freed [---------------------------] | |
4022 | * snapshots ---O-------O--------O-------O--------O | |
4023 | * firstsnap lastsnap | |
4024 | * | |
4025 | * This is the set of blocks that were born after the snap before firstsnap, | |
4026 | * (birth > firstsnap->prev_snap_txg) and died before the snap after the | |
4027 | * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist). | |
4028 | * We calculate this by iterating over the relevant deadlists (from the snap | |
4029 | * after lastsnap, backward to the snap after firstsnap), summing up the | |
4030 | * space on the deadlist that was born after the snap before firstsnap. | |
4031 | */ | |
4032 | int | |
4033 | dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, | |
4034 | dsl_dataset_t *lastsnap, | |
4035 | uint64_t *usedp, uint64_t *compp, uint64_t *uncompp) | |
4036 | { | |
4037 | int err = 0; | |
4038 | uint64_t snapobj; | |
4039 | dsl_pool_t *dp = firstsnap->ds_dir->dd_pool; | |
4040 | ||
0c66c32d JG |
4041 | ASSERT(firstsnap->ds_is_snapshot); |
4042 | ASSERT(lastsnap->ds_is_snapshot); | |
330d06f9 MA |
4043 | |
4044 | /* | |
4045 | * Check that the snapshots are in the same dsl_dir, and firstsnap | |
4046 | * is before lastsnap. | |
4047 | */ | |
4048 | if (firstsnap->ds_dir != lastsnap->ds_dir || | |
d683ddbb JG |
4049 | dsl_dataset_phys(firstsnap)->ds_creation_txg > |
4050 | dsl_dataset_phys(lastsnap)->ds_creation_txg) | |
2e528b49 | 4051 | return (SET_ERROR(EINVAL)); |
330d06f9 MA |
4052 | |
4053 | *usedp = *compp = *uncompp = 0; | |
4054 | ||
d683ddbb | 4055 | snapobj = dsl_dataset_phys(lastsnap)->ds_next_snap_obj; |
330d06f9 MA |
4056 | while (snapobj != firstsnap->ds_object) { |
4057 | dsl_dataset_t *ds; | |
4058 | uint64_t used, comp, uncomp; | |
4059 | ||
4060 | err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds); | |
4061 | if (err != 0) | |
4062 | break; | |
4063 | ||
4064 | dsl_deadlist_space_range(&ds->ds_deadlist, | |
d683ddbb | 4065 | dsl_dataset_phys(firstsnap)->ds_prev_snap_txg, UINT64_MAX, |
330d06f9 MA |
4066 | &used, &comp, &uncomp); |
4067 | *usedp += used; | |
4068 | *compp += comp; | |
4069 | *uncompp += uncomp; | |
4070 | ||
d683ddbb | 4071 | snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj; |
330d06f9 MA |
4072 | ASSERT3U(snapobj, !=, 0); |
4073 | dsl_dataset_rele(ds, FTAG); | |
4074 | } | |
330d06f9 MA |
4075 | return (err); |
4076 | } | |
4077 | ||
13fe0198 MA |
4078 | /* |
4079 | * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline. | |
4080 | * For example, they could both be snapshots of the same filesystem, and | |
4081 | * 'earlier' is before 'later'. Or 'earlier' could be the origin of | |
4082 | * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's | |
4083 | * filesystem. Or 'earlier' could be the origin's origin. | |
da536844 MA |
4084 | * |
4085 | * If non-zero, earlier_txg is used instead of earlier's ds_creation_txg. | |
13fe0198 MA |
4086 | */ |
4087 | boolean_t | |
da536844 | 4088 | dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier, |
e9aa730c | 4089 | uint64_t earlier_txg) |
13fe0198 MA |
4090 | { |
4091 | dsl_pool_t *dp = later->ds_dir->dd_pool; | |
4092 | int error; | |
4093 | boolean_t ret; | |
13fe0198 MA |
4094 | |
4095 | ASSERT(dsl_pool_config_held(dp)); | |
0c66c32d | 4096 | ASSERT(earlier->ds_is_snapshot || earlier_txg != 0); |
da536844 MA |
4097 | |
4098 | if (earlier_txg == 0) | |
d683ddbb | 4099 | earlier_txg = dsl_dataset_phys(earlier)->ds_creation_txg; |
13fe0198 | 4100 | |
0c66c32d | 4101 | if (later->ds_is_snapshot && |
d683ddbb | 4102 | earlier_txg >= dsl_dataset_phys(later)->ds_creation_txg) |
13fe0198 MA |
4103 | return (B_FALSE); |
4104 | ||
4105 | if (later->ds_dir == earlier->ds_dir) | |
4106 | return (B_TRUE); | |
4107 | if (!dsl_dir_is_clone(later->ds_dir)) | |
4108 | return (B_FALSE); | |
4109 | ||
d683ddbb | 4110 | if (dsl_dir_phys(later->ds_dir)->dd_origin_obj == earlier->ds_object) |
13fe0198 | 4111 | return (B_TRUE); |
1c27024e | 4112 | dsl_dataset_t *origin; |
13fe0198 | 4113 | error = dsl_dataset_hold_obj(dp, |
d683ddbb | 4114 | dsl_dir_phys(later->ds_dir)->dd_origin_obj, FTAG, &origin); |
13fe0198 MA |
4115 | if (error != 0) |
4116 | return (B_FALSE); | |
da536844 | 4117 | ret = dsl_dataset_is_before(origin, earlier, earlier_txg); |
13fe0198 MA |
4118 | dsl_dataset_rele(origin, FTAG); |
4119 | return (ret); | |
4120 | } | |
4121 | ||
fa86b5db MA |
4122 | void |
4123 | dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4124 | { | |
4125 | objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; | |
4126 | dmu_object_zapify(mos, ds->ds_object, DMU_OT_DSL_DATASET, tx); | |
4127 | } | |
4128 | ||
47dfff3b MA |
4129 | boolean_t |
4130 | dsl_dataset_is_zapified(dsl_dataset_t *ds) | |
4131 | { | |
4132 | dmu_object_info_t doi; | |
4133 | ||
4134 | dmu_object_info_from_db(ds->ds_dbuf, &doi); | |
4135 | return (doi.doi_type == DMU_OTN_ZAP_METADATA); | |
4136 | } | |
4137 | ||
4138 | boolean_t | |
4139 | dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds) | |
4140 | { | |
4141 | return (dsl_dataset_is_zapified(ds) && | |
4142 | zap_contains(ds->ds_dir->dd_pool->dp_meta_objset, | |
4143 | ds->ds_object, DS_FIELD_RESUME_TOGUID) == 0); | |
4144 | } | |
4145 | ||
a1d477c2 MA |
4146 | uint64_t |
4147 | dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds) | |
4148 | { | |
4149 | uint64_t remap_deadlist_obj; | |
4150 | int err; | |
4151 | ||
4152 | if (!dsl_dataset_is_zapified(ds)) | |
4153 | return (0); | |
4154 | ||
4155 | err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object, | |
4156 | DS_FIELD_REMAP_DEADLIST, sizeof (remap_deadlist_obj), 1, | |
4157 | &remap_deadlist_obj); | |
4158 | ||
4159 | if (err != 0) { | |
4160 | VERIFY3S(err, ==, ENOENT); | |
4161 | return (0); | |
4162 | } | |
4163 | ||
4164 | ASSERT(remap_deadlist_obj != 0); | |
4165 | return (remap_deadlist_obj); | |
4166 | } | |
4167 | ||
4168 | boolean_t | |
4169 | dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds) | |
4170 | { | |
4171 | EQUIV(dsl_deadlist_is_open(&ds->ds_remap_deadlist), | |
4172 | dsl_dataset_get_remap_deadlist_object(ds) != 0); | |
4173 | return (dsl_deadlist_is_open(&ds->ds_remap_deadlist)); | |
4174 | } | |
4175 | ||
4176 | static void | |
4177 | dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, uint64_t obj, | |
4178 | dmu_tx_t *tx) | |
4179 | { | |
4180 | ASSERT(obj != 0); | |
4181 | dsl_dataset_zapify(ds, tx); | |
4182 | VERIFY0(zap_add(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object, | |
4183 | DS_FIELD_REMAP_DEADLIST, sizeof (obj), 1, &obj, tx)); | |
4184 | } | |
4185 | ||
4186 | static void | |
4187 | dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4188 | { | |
4189 | VERIFY0(zap_remove(ds->ds_dir->dd_pool->dp_meta_objset, | |
4190 | ds->ds_object, DS_FIELD_REMAP_DEADLIST, tx)); | |
4191 | } | |
4192 | ||
4193 | void | |
4194 | dsl_dataset_destroy_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4195 | { | |
4196 | uint64_t remap_deadlist_object; | |
4197 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
4198 | ||
4199 | ASSERT(dmu_tx_is_syncing(tx)); | |
4200 | ASSERT(dsl_dataset_remap_deadlist_exists(ds)); | |
4201 | ||
4202 | remap_deadlist_object = ds->ds_remap_deadlist.dl_object; | |
4203 | dsl_deadlist_close(&ds->ds_remap_deadlist); | |
4204 | dsl_deadlist_free(spa_meta_objset(spa), remap_deadlist_object, tx); | |
4205 | dsl_dataset_unset_remap_deadlist_object(ds, tx); | |
4206 | spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx); | |
4207 | } | |
4208 | ||
4209 | void | |
4210 | dsl_dataset_create_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx) | |
4211 | { | |
4212 | uint64_t remap_deadlist_obj; | |
4213 | spa_t *spa = ds->ds_dir->dd_pool->dp_spa; | |
4214 | ||
4215 | ASSERT(dmu_tx_is_syncing(tx)); | |
4216 | ASSERT(MUTEX_HELD(&ds->ds_remap_deadlist_lock)); | |
4217 | /* | |
4218 | * Currently we only create remap deadlists when there are indirect | |
4219 | * vdevs with referenced mappings. | |
4220 | */ | |
4221 | ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); | |
4222 | ||
4223 | remap_deadlist_obj = dsl_deadlist_clone( | |
4224 | &ds->ds_deadlist, UINT64_MAX, | |
4225 | dsl_dataset_phys(ds)->ds_prev_snap_obj, tx); | |
4226 | dsl_dataset_set_remap_deadlist_object(ds, | |
4227 | remap_deadlist_obj, tx); | |
4228 | dsl_deadlist_open(&ds->ds_remap_deadlist, spa_meta_objset(spa), | |
4229 | remap_deadlist_obj); | |
4230 | spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx); | |
4231 | } | |
4232 | ||
93ce2b4c | 4233 | #if defined(_KERNEL) |
f1512ee6 MA |
4234 | #if defined(_LP64) |
4235 | module_param(zfs_max_recordsize, int, 0644); | |
4236 | MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size"); | |
4237 | #else | |
4238 | /* Limited to 1M on 32-bit platforms due to lack of virtual address space */ | |
4239 | module_param(zfs_max_recordsize, int, 0444); | |
4240 | MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size"); | |
4241 | #endif | |
4242 | ||
c28b2279 | 4243 | EXPORT_SYMBOL(dsl_dataset_hold); |
b5256303 | 4244 | EXPORT_SYMBOL(dsl_dataset_hold_flags); |
c28b2279 | 4245 | EXPORT_SYMBOL(dsl_dataset_hold_obj); |
b5256303 | 4246 | EXPORT_SYMBOL(dsl_dataset_hold_obj_flags); |
c28b2279 BB |
4247 | EXPORT_SYMBOL(dsl_dataset_own); |
4248 | EXPORT_SYMBOL(dsl_dataset_own_obj); | |
4249 | EXPORT_SYMBOL(dsl_dataset_name); | |
4250 | EXPORT_SYMBOL(dsl_dataset_rele); | |
b5256303 | 4251 | EXPORT_SYMBOL(dsl_dataset_rele_flags); |
c28b2279 | 4252 | EXPORT_SYMBOL(dsl_dataset_disown); |
c28b2279 | 4253 | EXPORT_SYMBOL(dsl_dataset_tryown); |
c28b2279 BB |
4254 | EXPORT_SYMBOL(dsl_dataset_create_sync); |
4255 | EXPORT_SYMBOL(dsl_dataset_create_sync_dd); | |
c28b2279 BB |
4256 | EXPORT_SYMBOL(dsl_dataset_snapshot_check); |
4257 | EXPORT_SYMBOL(dsl_dataset_snapshot_sync); | |
c28b2279 | 4258 | EXPORT_SYMBOL(dsl_dataset_promote); |
c28b2279 BB |
4259 | EXPORT_SYMBOL(dsl_dataset_user_hold); |
4260 | EXPORT_SYMBOL(dsl_dataset_user_release); | |
c28b2279 BB |
4261 | EXPORT_SYMBOL(dsl_dataset_get_holds); |
4262 | EXPORT_SYMBOL(dsl_dataset_get_blkptr); | |
c28b2279 | 4263 | EXPORT_SYMBOL(dsl_dataset_get_spa); |
19580676 | 4264 | EXPORT_SYMBOL(dsl_dataset_modified_since_snap); |
330d06f9 MA |
4265 | EXPORT_SYMBOL(dsl_dataset_space_written); |
4266 | EXPORT_SYMBOL(dsl_dataset_space_wouldfree); | |
c28b2279 BB |
4267 | EXPORT_SYMBOL(dsl_dataset_sync); |
4268 | EXPORT_SYMBOL(dsl_dataset_block_born); | |
4269 | EXPORT_SYMBOL(dsl_dataset_block_kill); | |
c28b2279 BB |
4270 | EXPORT_SYMBOL(dsl_dataset_dirty); |
4271 | EXPORT_SYMBOL(dsl_dataset_stats); | |
4272 | EXPORT_SYMBOL(dsl_dataset_fast_stat); | |
4273 | EXPORT_SYMBOL(dsl_dataset_space); | |
4274 | EXPORT_SYMBOL(dsl_dataset_fsid_guid); | |
4275 | EXPORT_SYMBOL(dsl_dsobj_to_dsname); | |
4276 | EXPORT_SYMBOL(dsl_dataset_check_quota); | |
13fe0198 MA |
4277 | EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl); |
4278 | EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl); | |
c28b2279 | 4279 | #endif |