]> git.proxmox.com Git - mirror_zfs.git/blame - module/zfs/dsl_dataset.c
projects / mirror_zfs.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Illumos #3742
[mirror_zfs.git] / module / zfs / dsl_dataset.c
CommitLineData
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 */
21/*
428870ff 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
13fe0198 23 * Copyright (c) 2013 by Delphix. All rights reserved.
37abac6d 24 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
34dc7c2f
BB		 25 */
26
34dc7c2f
BB		 27#include <sys/dmu_objset.h>
28#include <sys/dsl_dataset.h>
29#include <sys/dsl_dir.h>
30#include <sys/dsl_prop.h>
31#include <sys/dsl_synctask.h>
32#include <sys/dmu_traverse.h>
37abac6d 33#include <sys/dmu_impl.h>
34dc7c2f
BB		 34#include <sys/dmu_tx.h>
35#include <sys/arc.h>
36#include <sys/zio.h>
37#include <sys/zap.h>
9ae529ec 38#include <sys/zfeature.h>
34dc7c2f
BB		 39#include <sys/unique.h>
40#include <sys/zfs_context.h>
41#include <sys/zfs_ioctl.h>
42#include <sys/spa.h>
b128c09f 43#include <sys/zfs_znode.h>
572e2857 44#include <sys/zfs_onexit.h>
45d1cae3 45#include <sys/zvol.h>
428870ff
BB		 46#include <sys/dsl_scan.h>
47#include <sys/dsl_deadlist.h>
13fe0198
MA		 48#include <sys/dsl_destroy.h>
49#include <sys/dsl_userhold.h>
34dc7c2f 50
428870ff
BB		 51#define SWITCH64(x, y) \
52 { \
53 uint64_t __tmp = (x); \
54 (x) = (y); \
55 (y) = __tmp; \
56 }
57
34dc7c2f
BB		 58#define DS_REF_MAX (1ULL << 62)
59
60#define DSL_DEADLIST_BLOCKSIZE SPA_MAXBLOCKSIZE
61
34dc7c2f
BB		 62/*
63 * Figure out how much of this delta should be propogated to the dsl_dir
64 * layer. If there's a refreservation, that space has already been
65 * partially accounted for in our ancestors.
66 */
67static int64_t
68parent_delta(dsl_dataset_t *ds, int64_t delta)
69{
70 uint64_t old_bytes, new_bytes;
71
72 if (ds->ds_reserved == 0)
73 return (delta);
74
75 old_bytes = MAX(ds->ds_phys->ds_unique_bytes, ds->ds_reserved);
76 new_bytes = MAX(ds->ds_phys->ds_unique_bytes + delta, ds->ds_reserved);
77
78 ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta));
79 return (new_bytes - old_bytes);
80}
81
82void
428870ff 83dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx)
34dc7c2f 84{
d6320ddb 85 int used, compressed, uncompressed;
34dc7c2f
BB		 86 int64_t delta;
87
d6320ddb
BB		 88 used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp);
89 compressed = BP_GET_PSIZE(bp);
90 uncompressed = BP_GET_UCSIZE(bp);
91
428870ff 92 dprintf_bp(bp, "ds=%p", ds);
34dc7c2f
BB		 93
94 ASSERT(dmu_tx_is_syncing(tx));
95 /* It could have been compressed away to nothing */
96 if (BP_IS_HOLE(bp))
97 return;
98 ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE);
9ae529ec 99 ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp)));
34dc7c2f 100 if (ds == NULL) {
29809a6c
MA		 101 dsl_pool_mos_diduse_space(tx->tx_pool,
102 used, compressed, uncompressed);
34dc7c2f
BB		 103 return;
104 }
105 dmu_buf_will_dirty(ds->ds_dbuf, tx);
428870ff 106
b128c09f 107 mutex_enter(&ds->ds_dir->dd_lock);
34dc7c2f
BB		 108 mutex_enter(&ds->ds_lock);
109 delta = parent_delta(ds, used);
9ae529ec 110 ds->ds_phys->ds_referenced_bytes += used;
34dc7c2f
BB		 111 ds->ds_phys->ds_compressed_bytes += compressed;
112 ds->ds_phys->ds_uncompressed_bytes += uncompressed;
113 ds->ds_phys->ds_unique_bytes += used;
114 mutex_exit(&ds->ds_lock);
b128c09f
BB		 115 dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta,
116 compressed, uncompressed, tx);
117 dsl_dir_transfer_space(ds->ds_dir, used - delta,
118 DD_USED_REFRSRV, DD_USED_HEAD, tx);
119 mutex_exit(&ds->ds_dir->dd_lock);
34dc7c2f
BB		 120}
121
b128c09f 122int
428870ff
BB		 123dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx,
124 boolean_t async)
34dc7c2f 125{
d6320ddb
BB		 126 int used, compressed, uncompressed;
127
34dc7c2f 128 if (BP_IS_HOLE(bp))
b128c09f 129 return (0);
34dc7c2f 130
428870ff
BB		 131 ASSERT(dmu_tx_is_syncing(tx));
132 ASSERT(bp->blk_birth <= tx->tx_txg);
133
d6320ddb
BB		 134 used = bp_get_dsize_sync(tx->tx_pool->dp_spa, bp);
135 compressed = BP_GET_PSIZE(bp);
136 uncompressed = BP_GET_UCSIZE(bp);
428870ff 137
34dc7c2f
BB		 138 ASSERT(used > 0);
139 if (ds == NULL) {
428870ff 140 dsl_free(tx->tx_pool, tx->tx_txg, bp);
29809a6c
MA		 141 dsl_pool_mos_diduse_space(tx->tx_pool,
142 -used, -compressed, -uncompressed);
b128c09f 143 return (used);
34dc7c2f
BB		 144 }
145 ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool);
146
b128c09f 147 ASSERT(!dsl_dataset_is_snapshot(ds));
34dc7c2f
BB		 148 dmu_buf_will_dirty(ds->ds_dbuf, tx);
149
150 if (bp->blk_birth > ds->ds_phys->ds_prev_snap_txg) {
34dc7c2f
BB		 151 int64_t delta;
152
428870ff
BB		 153 dprintf_bp(bp, "freeing ds=%llu", ds->ds_object);
154 dsl_free(tx->tx_pool, tx->tx_txg, bp);
34dc7c2f 155
b128c09f 156 mutex_enter(&ds->ds_dir->dd_lock);
34dc7c2f
BB		 157 mutex_enter(&ds->ds_lock);
158 ASSERT(ds->ds_phys->ds_unique_bytes >= used ||
159 !DS_UNIQUE_IS_ACCURATE(ds));
160 delta = parent_delta(ds, -used);
161 ds->ds_phys->ds_unique_bytes -= used;
162 mutex_exit(&ds->ds_lock);
b128c09f 163 dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD,
34dc7c2f 164 delta, -compressed, -uncompressed, tx);
b128c09f
BB		 165 dsl_dir_transfer_space(ds->ds_dir, -used - delta,
166 DD_USED_REFRSRV, DD_USED_HEAD, tx);
167 mutex_exit(&ds->ds_dir->dd_lock);
34dc7c2f
BB		 168 } else {
169 dprintf_bp(bp, "putting on dead list: %s", "");
428870ff
BB		 170 if (async) {
171 /*
172 * We are here as part of zio's write done callback,
173 * which means we're a zio interrupt thread. We can't
174 * call dsl_deadlist_insert() now because it may block
175 * waiting for I/O. Instead, put bp on the deferred
176 * queue and let dsl_pool_sync() finish the job.
177 */
178 bplist_append(&ds->ds_pending_deadlist, bp);
179 } else {
180 dsl_deadlist_insert(&ds->ds_deadlist, bp, tx);
181 }
34dc7c2f
BB		 182 ASSERT3U(ds->ds_prev->ds_object, ==,
183 ds->ds_phys->ds_prev_snap_obj);
184 ASSERT(ds->ds_prev->ds_phys->ds_num_children > 0);
185 /* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
186 if (ds->ds_prev->ds_phys->ds_next_snap_obj ==
187 ds->ds_object && bp->blk_birth >
188 ds->ds_prev->ds_phys->ds_prev_snap_txg) {
189 dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
190 mutex_enter(&ds->ds_prev->ds_lock);
191 ds->ds_prev->ds_phys->ds_unique_bytes += used;
192 mutex_exit(&ds->ds_prev->ds_lock);
193 }
428870ff 194 if (bp->blk_birth > ds->ds_dir->dd_origin_txg) {
b128c09f
BB		 195 dsl_dir_transfer_space(ds->ds_dir, used,
196 DD_USED_HEAD, DD_USED_SNAP, tx);
197 }
34dc7c2f
BB		 198 }
199 mutex_enter(&ds->ds_lock);
9ae529ec
CS		 200 ASSERT3U(ds->ds_phys->ds_referenced_bytes, >=, used);
201 ds->ds_phys->ds_referenced_bytes -= used;
34dc7c2f
BB		 202 ASSERT3U(ds->ds_phys->ds_compressed_bytes, >=, compressed);
203 ds->ds_phys->ds_compressed_bytes -= compressed;
204 ASSERT3U(ds->ds_phys->ds_uncompressed_bytes, >=, uncompressed);
205 ds->ds_phys->ds_uncompressed_bytes -= uncompressed;
206 mutex_exit(&ds->ds_lock);
b128c09f
BB		 207
208 return (used);
34dc7c2f
BB		 209}
210
211uint64_t
212dsl_dataset_prev_snap_txg(dsl_dataset_t *ds)
213{
214 uint64_t trysnap = 0;
215
216 if (ds == NULL)
217 return (0);
218 /*
219 * The snapshot creation could fail, but that would cause an
220 * incorrect FALSE return, which would only result in an
221 * overestimation of the amount of space that an operation would
222 * consume, which is OK.
223 *
224 * There's also a small window where we could miss a pending
225 * snapshot, because we could set the sync task in the quiescing
226 * phase. So this should only be used as a guess.
227 */
228 if (ds->ds_trysnap_txg >
229 spa_last_synced_txg(ds->ds_dir->dd_pool->dp_spa))
230 trysnap = ds->ds_trysnap_txg;
231 return (MAX(ds->ds_phys->ds_prev_snap_txg, trysnap));
232}
233
9babb374 234boolean_t
428870ff
BB		 235dsl_dataset_block_freeable(dsl_dataset_t *ds, const blkptr_t *bp,
236 uint64_t blk_birth)
34dc7c2f 237{
428870ff
BB		 238 if (blk_birth <= dsl_dataset_prev_snap_txg(ds))
239 return (B_FALSE);
240
572e2857 241 ddt_prefetch(dsl_dataset_get_spa(ds), bp);
428870ff
BB		 242
243 return (B_TRUE);
34dc7c2f
BB		 244}
245
246/* ARGSUSED */
247static void
248dsl_dataset_evict(dmu_buf_t *db, void *dsv)
249{
250 dsl_dataset_t *ds = dsv;
251
13fe0198 252 ASSERT(ds->ds_owner == NULL);
34dc7c2f 253
34dc7c2f
BB		 254 unique_remove(ds->ds_fsid_guid);
255
428870ff
BB		 256 if (ds->ds_objset != NULL)
257 dmu_objset_evict(ds->ds_objset);
34dc7c2f
BB		 258
259 if (ds->ds_prev) {
13fe0198 260 dsl_dataset_rele(ds->ds_prev, ds);
34dc7c2f
BB		 261 ds->ds_prev = NULL;
262 }
263
428870ff 264 bplist_destroy(&ds->ds_pending_deadlist);
13fe0198 265 if (ds->ds_phys->ds_deadlist_obj != 0)
428870ff 266 dsl_deadlist_close(&ds->ds_deadlist);
b128c09f 267 if (ds->ds_dir)
13fe0198 268 dsl_dir_rele(ds->ds_dir, ds);
34dc7c2f
BB		 269
270 ASSERT(!list_link_active(&ds->ds_synced_link));
271
272 mutex_destroy(&ds->ds_lock);
273 mutex_destroy(&ds->ds_opening_lock);
13fe0198 274 refcount_destroy(&ds->ds_longholds);
34dc7c2f
BB		 275
276 kmem_free(ds, sizeof (dsl_dataset_t));
277}
278
13fe0198 279int
34dc7c2f
BB		 280dsl_dataset_get_snapname(dsl_dataset_t *ds)
281{
282 dsl_dataset_phys_t *headphys;
283 int err;
284 dmu_buf_t *headdbuf;
285 dsl_pool_t *dp = ds->ds_dir->dd_pool;
286 objset_t *mos = dp->dp_meta_objset;
287
288 if (ds->ds_snapname[0])
289 return (0);
290 if (ds->ds_phys->ds_next_snap_obj == 0)
291 return (0);
292
293 err = dmu_bonus_hold(mos, ds->ds_dir->dd_phys->dd_head_dataset_obj,
294 FTAG, &headdbuf);
13fe0198 295 if (err != 0)
34dc7c2f
BB		 296 return (err);
297 headphys = headdbuf->db_data;
298 err = zap_value_search(dp->dp_meta_objset,
299 headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname);
300 dmu_buf_rele(headdbuf, FTAG);
301 return (err);
302}
303
6772fb67 304int
b128c09f 305dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value)
34dc7c2f 306{
b128c09f
BB		 307 objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
308 uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
34dc7c2f
BB		 309 matchtype_t mt;
310 int err;
311
b128c09f 312 if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
34dc7c2f
BB		 313 mt = MT_FIRST;
314 else
315 mt = MT_EXACT;
316
b128c09f 317 err = zap_lookup_norm(mos, snapobj, name, 8, 1,
34dc7c2f
BB		 318 value, mt, NULL, 0, NULL);
319 if (err == ENOTSUP && mt == MT_FIRST)
b128c09f 320 err = zap_lookup(mos, snapobj, name, 8, 1, value);
34dc7c2f
BB		 321 return (err);
322}
323
13fe0198
MA		 324int
325dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx)
34dc7c2f 326{
b128c09f
BB		 327 objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
328 uint64_t snapobj = ds->ds_phys->ds_snapnames_zapobj;
34dc7c2f
BB		 329 matchtype_t mt;
330 int err;
331
428870ff
BB		 332 dsl_dir_snap_cmtime_update(ds->ds_dir);
333
b128c09f 334 if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
34dc7c2f
BB		 335 mt = MT_FIRST;
336 else
337 mt = MT_EXACT;
338
b128c09f 339 err = zap_remove_norm(mos, snapobj, name, mt, tx);
34dc7c2f 340 if (err == ENOTSUP && mt == MT_FIRST)
b128c09f 341 err = zap_remove(mos, snapobj, name, tx);
34dc7c2f
BB		 342 return (err);
343}
344
13fe0198
MA		 345int
346dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag,
b128c09f 347 dsl_dataset_t **dsp)
34dc7c2f 348{
34dc7c2f
BB		 349 objset_t *mos = dp->dp_meta_objset;
350 dmu_buf_t *dbuf;
351 dsl_dataset_t *ds;
352 int err;
572e2857 353 dmu_object_info_t doi;
34dc7c2f 354
13fe0198 355 ASSERT(dsl_pool_config_held(dp));
34dc7c2f
BB		 356
357 err = dmu_bonus_hold(mos, dsobj, tag, &dbuf);
13fe0198 358 if (err != 0)
34dc7c2f 359 return (err);
572e2857
BB		 360
361 /* Make sure dsobj has the correct object type. */
362 dmu_object_info_from_db(dbuf, &doi);
363 if (doi.doi_type != DMU_OT_DSL_DATASET)
2e528b49 364 return (SET_ERROR(EINVAL));
572e2857 365
34dc7c2f
BB		 366 ds = dmu_buf_get_user(dbuf);
367 if (ds == NULL) {
d4ed6673 368 dsl_dataset_t *winner = NULL;
34dc7c2f 369
b8d06fca 370 ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_PUSHPAGE);
34dc7c2f
BB		 371 ds->ds_dbuf = dbuf;
372 ds->ds_object = dsobj;
373 ds->ds_phys = dbuf->db_data;
98f72a53 374 list_link_init(&ds->ds_synced_link);
34dc7c2f
BB		 375
376 mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL);
377 mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL);
37abac6d 378 mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL);
13fe0198 379 refcount_create(&ds->ds_longholds);
34dc7c2f 380
428870ff
BB		 381 bplist_create(&ds->ds_pending_deadlist);
382 dsl_deadlist_open(&ds->ds_deadlist,
34dc7c2f 383 mos, ds->ds_phys->ds_deadlist_obj);
428870ff 384
37abac6d
BP		 385 list_create(&ds->ds_sendstreams, sizeof (dmu_sendarg_t),
386 offsetof(dmu_sendarg_t, dsa_link));
387
34dc7c2f 388 if (err == 0) {
13fe0198 389 err = dsl_dir_hold_obj(dp,
34dc7c2f
BB		 390 ds->ds_phys->ds_dir_obj, NULL, ds, &ds->ds_dir);
391 }
13fe0198 392 if (err != 0) {
34dc7c2f
BB		 393 mutex_destroy(&ds->ds_lock);
394 mutex_destroy(&ds->ds_opening_lock);
13fe0198 395 refcount_destroy(&ds->ds_longholds);
428870ff
BB		 396 bplist_destroy(&ds->ds_pending_deadlist);
397 dsl_deadlist_close(&ds->ds_deadlist);
34dc7c2f
BB		 398 kmem_free(ds, sizeof (dsl_dataset_t));
399 dmu_buf_rele(dbuf, tag);
400 return (err);
401 }
402
b128c09f 403 if (!dsl_dataset_is_snapshot(ds)) {
34dc7c2f 404 ds->ds_snapname[0] = '\0';
13fe0198
MA		 405 if (ds->ds_phys->ds_prev_snap_obj != 0) {
406 err = dsl_dataset_hold_obj(dp,
b128c09f
BB		 407 ds->ds_phys->ds_prev_snap_obj,
408 ds, &ds->ds_prev);
34dc7c2f 409 }
45d1cae3
BB		 410 } else {
411 if (zfs_flags & ZFS_DEBUG_SNAPNAMES)
412 err = dsl_dataset_get_snapname(ds);
413 if (err == 0 && ds->ds_phys->ds_userrefs_obj != 0) {
414 err = zap_count(
415 ds->ds_dir->dd_pool->dp_meta_objset,
416 ds->ds_phys->ds_userrefs_obj,
417 &ds->ds_userrefs);
418 }
34dc7c2f
BB		 419 }
420
b128c09f 421 if (err == 0 && !dsl_dataset_is_snapshot(ds)) {
13fe0198
MA		 422 err = dsl_prop_get_int_ds(ds,
423 zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
424 &ds->ds_reserved);
34dc7c2f 425 if (err == 0) {
13fe0198
MA		 426 err = dsl_prop_get_int_ds(ds,
427 zfs_prop_to_name(ZFS_PROP_REFQUOTA),
428 &ds->ds_quota);
34dc7c2f 429 }
34dc7c2f
BB		 430 } else {
431 ds->ds_reserved = ds->ds_quota = 0;
432 }
433
a117a6d6
GW		 434 if (err != 0 || (winner = dmu_buf_set_user_ie(dbuf, ds,
435 &ds->ds_phys, dsl_dataset_evict)) != NULL) {
428870ff
BB		 436 bplist_destroy(&ds->ds_pending_deadlist);
437 dsl_deadlist_close(&ds->ds_deadlist);
b128c09f 438 if (ds->ds_prev)
13fe0198
MA		 439 dsl_dataset_rele(ds->ds_prev, ds);
440 dsl_dir_rele(ds->ds_dir, ds);
34dc7c2f
BB		 441 mutex_destroy(&ds->ds_lock);
442 mutex_destroy(&ds->ds_opening_lock);
13fe0198 443 refcount_destroy(&ds->ds_longholds);
34dc7c2f 444 kmem_free(ds, sizeof (dsl_dataset_t));
13fe0198 445 if (err != 0) {
34dc7c2f
BB		 446 dmu_buf_rele(dbuf, tag);
447 return (err);
448 }
449 ds = winner;
450 } else {
451 ds->ds_fsid_guid =
452 unique_insert(ds->ds_phys->ds_fsid_guid);
453 }
454 }
455 ASSERT3P(ds->ds_dbuf, ==, dbuf);
456 ASSERT3P(ds->ds_phys, ==, dbuf->db_data);
b128c09f
BB		 457 ASSERT(ds->ds_phys->ds_prev_snap_obj != 0 ||
458 spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN ||
459 dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap);
34dc7c2f
BB		 460 *dsp = ds;
461 return (0);
462}
463
b128c09f 464int
13fe0198 465dsl_dataset_hold(dsl_pool_t *dp, const char *name,
428870ff 466 void *tag, dsl_dataset_t **dsp)
34dc7c2f
BB		 467{
468 dsl_dir_t *dd;
b128c09f 469 const char *snapname;
34dc7c2f 470 uint64_t obj;
34dc7c2f
BB		 471 int err = 0;
472
13fe0198
MA		 473 err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname);
474 if (err != 0)
34dc7c2f
BB		 475 return (err);
476
13fe0198 477 ASSERT(dsl_pool_config_held(dp));
34dc7c2f 478 obj = dd->dd_phys->dd_head_dataset_obj;
13fe0198
MA		 479 if (obj != 0)
480 err = dsl_dataset_hold_obj(dp, obj, tag, dsp);
b128c09f 481 else
2e528b49 482 err = SET_ERROR(ENOENT);
34dc7c2f 483
b128c09f
BB		 484 /* we may be looking for a snapshot */
485 if (err == 0 && snapname != NULL) {
13fe0198 486 dsl_dataset_t *ds;
34dc7c2f 487
b128c09f
BB		 488 if (*snapname++ != '@') {
489 dsl_dataset_rele(*dsp, tag);
13fe0198 490 dsl_dir_rele(dd, FTAG);
2e528b49 491 return (SET_ERROR(ENOENT));
34dc7c2f 492 }
34dc7c2f 493
b128c09f
BB		 494 dprintf("looking for snapshot '%s'\n", snapname);
495 err = dsl_dataset_snap_lookup(*dsp, snapname, &obj);
496 if (err == 0)
13fe0198 497 err = dsl_dataset_hold_obj(dp, obj, tag, &ds);
b128c09f
BB		 498 dsl_dataset_rele(*dsp, tag);
499
13fe0198 500 if (err == 0) {
b128c09f
BB		 501 mutex_enter(&ds->ds_lock);
502 if (ds->ds_snapname[0] == 0)
503 (void) strlcpy(ds->ds_snapname, snapname,
504 sizeof (ds->ds_snapname));
505 mutex_exit(&ds->ds_lock);
13fe0198 506 *dsp = ds;
34dc7c2f 507 }
34dc7c2f 508 }
13fe0198
MA		 509
510 dsl_dir_rele(dd, FTAG);
34dc7c2f
BB		 511 return (err);
512}
513
514int
13fe0198 515dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj,
428870ff 516 void *tag, dsl_dataset_t **dsp)
34dc7c2f 517{
13fe0198
MA		 518 int err = dsl_dataset_hold_obj(dp, dsobj, tag, dsp);
519 if (err != 0)
520 return (err);
521 if (!dsl_dataset_tryown(*dsp, tag)) {
522 dsl_dataset_rele(*dsp, tag);
523 *dsp = NULL;
2e528b49 524 return (SET_ERROR(EBUSY));
13fe0198
MA		 525 }
526 return (0);
527}
528
529int
530dsl_dataset_own(dsl_pool_t *dp, const char *name,
531 void *tag, dsl_dataset_t **dsp)
532{
533 int err = dsl_dataset_hold(dp, name, tag, dsp);
534 if (err != 0)
b128c09f 535 return (err);
13fe0198 536 if (!dsl_dataset_tryown(*dsp, tag)) {
428870ff 537 dsl_dataset_rele(*dsp, tag);
2e528b49 538 return (SET_ERROR(EBUSY));
b128c09f
BB		 539 }
540 return (0);
34dc7c2f
BB		 541}
542
13fe0198
MA		 543/*
544 * See the comment above dsl_pool_hold() for details. In summary, a long
545 * hold is used to prevent destruction of a dataset while the pool hold
546 * is dropped, allowing other concurrent operations (e.g. spa_sync()).
547 *
548 * The dataset and pool must be held when this function is called. After it
549 * is called, the pool hold may be released while the dataset is still held
550 * and accessed.
551 */
552void
553dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag)
554{
555 ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
556 (void) refcount_add(&ds->ds_longholds, tag);
557}
558
559void
560dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag)
561{
562 (void) refcount_remove(&ds->ds_longholds, tag);
563}
564
565/* Return B_TRUE if there are any long holds on this dataset. */
566boolean_t
567dsl_dataset_long_held(dsl_dataset_t *ds)
568{
569 return (!refcount_is_zero(&ds->ds_longholds));
570}
571
34dc7c2f
BB		 572void
573dsl_dataset_name(dsl_dataset_t *ds, char *name)
574{
575 if (ds == NULL) {
576 (void) strcpy(name, "mos");
577 } else {
578 dsl_dir_name(ds->ds_dir, name);
13fe0198 579 VERIFY0(dsl_dataset_get_snapname(ds));
34dc7c2f
BB		 580 if (ds->ds_snapname[0]) {
581 (void) strcat(name, "@");
b128c09f
BB		 582 /*
583 * We use a "recursive" mutex so that we
584 * can call dprintf_ds() with ds_lock held.
585 */
34dc7c2f 586 if (!MUTEX_HELD(&ds->ds_lock)) {
34dc7c2f
BB		 587 mutex_enter(&ds->ds_lock);
588 (void) strcat(name, ds->ds_snapname);
589 mutex_exit(&ds->ds_lock);
590 } else {
591 (void) strcat(name, ds->ds_snapname);
592 }
593 }
594 }
595}
596
34dc7c2f 597void
b128c09f
BB		 598dsl_dataset_rele(dsl_dataset_t *ds, void *tag)
599{
13fe0198 600 dmu_buf_rele(ds->ds_dbuf, tag);
b128c09f
BB		 601}
602
603void
428870ff 604dsl_dataset_disown(dsl_dataset_t *ds, void *tag)
34dc7c2f 605{
13fe0198 606 ASSERT(ds->ds_owner == tag && ds->ds_dbuf != NULL);
b128c09f 607
34dc7c2f 608 mutex_enter(&ds->ds_lock);
b128c09f 609 ds->ds_owner = NULL;
34dc7c2f 610 mutex_exit(&ds->ds_lock);
13fe0198
MA		 611 dsl_dataset_long_rele(ds, tag);
612 if (ds->ds_dbuf != NULL)
613 dsl_dataset_rele(ds, tag);
b128c09f 614 else
428870ff 615 dsl_dataset_evict(NULL, ds);
34dc7c2f
BB		 616}
617
618boolean_t
13fe0198 619dsl_dataset_tryown(dsl_dataset_t *ds, void *tag)
34dc7c2f 620{
b128c09f
BB		 621 boolean_t gotit = FALSE;
622
34dc7c2f 623 mutex_enter(&ds->ds_lock);
13fe0198 624 if (ds->ds_owner == NULL && !DS_IS_INCONSISTENT(ds)) {
428870ff 625 ds->ds_owner = tag;
13fe0198 626 dsl_dataset_long_hold(ds, tag);
b128c09f 627 gotit = TRUE;
34dc7c2f
BB		 628 }
629 mutex_exit(&ds->ds_lock);
b128c09f 630 return (gotit);
34dc7c2f
BB		 631}
632
34dc7c2f 633uint64_t
b128c09f 634dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
34dc7c2f
BB		 635 uint64_t flags, dmu_tx_t *tx)
636{
637 dsl_pool_t *dp = dd->dd_pool;
638 dmu_buf_t *dbuf;
639 dsl_dataset_phys_t *dsphys;
640 uint64_t dsobj;
641 objset_t *mos = dp->dp_meta_objset;
642
b128c09f
BB		 643 if (origin == NULL)
644 origin = dp->dp_origin_snap;
645
34dc7c2f
BB		 646 ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp);
647 ASSERT(origin == NULL || origin->ds_phys->ds_num_children > 0);
648 ASSERT(dmu_tx_is_syncing(tx));
649 ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
650
651 dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
652 DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
13fe0198 653 VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
34dc7c2f
BB		 654 dmu_buf_will_dirty(dbuf, tx);
655 dsphys = dbuf->db_data;
b128c09f 656 bzero(dsphys, sizeof (dsl_dataset_phys_t));
34dc7c2f
BB		 657 dsphys->ds_dir_obj = dd->dd_object;
658 dsphys->ds_flags = flags;
659 dsphys->ds_fsid_guid = unique_create();
660 (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
661 sizeof (dsphys->ds_guid));
662 dsphys->ds_snapnames_zapobj =
663 zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP,
664 DMU_OT_NONE, 0, tx);
665 dsphys->ds_creation_time = gethrestime_sec();
b128c09f 666 dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg;
34dc7c2f 667
428870ff
BB		 668 if (origin == NULL) {
669 dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx);
670 } else {
13fe0198 671 dsl_dataset_t *ohds; /* head of the origin snapshot */
428870ff 672
34dc7c2f
BB		 673 dsphys->ds_prev_snap_obj = origin->ds_object;
674 dsphys->ds_prev_snap_txg =
675 origin->ds_phys->ds_creation_txg;
9ae529ec
CS		 676 dsphys->ds_referenced_bytes =
677 origin->ds_phys->ds_referenced_bytes;
34dc7c2f
BB		 678 dsphys->ds_compressed_bytes =
679 origin->ds_phys->ds_compressed_bytes;
680 dsphys->ds_uncompressed_bytes =
681 origin->ds_phys->ds_uncompressed_bytes;
682 dsphys->ds_bp = origin->ds_phys->ds_bp;
683 dsphys->ds_flags |= origin->ds_phys->ds_flags;
684
685 dmu_buf_will_dirty(origin->ds_dbuf, tx);
686 origin->ds_phys->ds_num_children++;
687
13fe0198 688 VERIFY0(dsl_dataset_hold_obj(dp,
428870ff
BB		 689 origin->ds_dir->dd_phys->dd_head_dataset_obj, FTAG, &ohds));
690 dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist,
691 dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx);
692 dsl_dataset_rele(ohds, FTAG);
693
b128c09f
BB		 694 if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) {
695 if (origin->ds_phys->ds_next_clones_obj == 0) {
696 origin->ds_phys->ds_next_clones_obj =
697 zap_create(mos,
698 DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
699 }
13fe0198
MA		 700 VERIFY0(zap_add_int(mos,
701 origin->ds_phys->ds_next_clones_obj, dsobj, tx));
b128c09f
BB		 702 }
703
34dc7c2f
BB		 704 dmu_buf_will_dirty(dd->dd_dbuf, tx);
705 dd->dd_phys->dd_origin_obj = origin->ds_object;
428870ff
BB		 706 if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
707 if (origin->ds_dir->dd_phys->dd_clones == 0) {
708 dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
709 origin->ds_dir->dd_phys->dd_clones =
710 zap_create(mos,
711 DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx);
712 }
13fe0198 713 VERIFY0(zap_add_int(mos,
428870ff
BB		 714 origin->ds_dir->dd_phys->dd_clones, dsobj, tx));
715 }
34dc7c2f
BB		 716 }
717
718 if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
719 dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
720
721 dmu_buf_rele(dbuf, FTAG);
722
723 dmu_buf_will_dirty(dd->dd_dbuf, tx);
724 dd->dd_phys->dd_head_dataset_obj = dsobj;
725
726 return (dsobj);
727}
728
13fe0198
MA		 729static void
730dsl_dataset_zero_zil(dsl_dataset_t *ds, dmu_tx_t *tx)
731{
732 objset_t *os;
733
734 VERIFY0(dmu_objset_from_ds(ds, &os));
735 bzero(&os->os_zil_header, sizeof (os->os_zil_header));
736 dsl_dataset_dirty(ds, tx);
737}
738
34dc7c2f
BB		 739uint64_t
740dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname,
741 dsl_dataset_t *origin, uint64_t flags, cred_t *cr, dmu_tx_t *tx)
742{
743 dsl_pool_t *dp = pdd->dd_pool;
744 uint64_t dsobj, ddobj;
745 dsl_dir_t *dd;
746
13fe0198 747 ASSERT(dmu_tx_is_syncing(tx));
34dc7c2f
BB		 748 ASSERT(lastname[0] != '@');
749
b128c09f 750 ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx);
13fe0198 751 VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd));
34dc7c2f 752
13fe0198
MA		 753 dsobj = dsl_dataset_create_sync_dd(dd, origin,
754 flags & ~DS_CREATE_FLAG_NODIRTY, tx);
34dc7c2f
BB		 755
756 dsl_deleg_set_create_perms(dd, tx, cr);
757
13fe0198 758 dsl_dir_rele(dd, FTAG);
34dc7c2f 759
572e2857
BB		 760 /*
761 * If we are creating a clone, make sure we zero out any stale
762 * data from the origin snapshots zil header.
763 */
13fe0198 764 if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) {
572e2857 765 dsl_dataset_t *ds;
572e2857 766
13fe0198
MA		 767 VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
768 dsl_dataset_zero_zil(ds, tx);
572e2857
BB		 769 dsl_dataset_rele(ds, FTAG);
770 }
771
34dc7c2f
BB		 772 return (dsobj);
773}
774
34dc7c2f 775/*
13fe0198
MA		 776 * The unique space in the head dataset can be calculated by subtracting
777 * the space used in the most recent snapshot, that is still being used
778 * in this file system, from the space currently in use. To figure out
779 * the space in the most recent snapshot still in use, we need to take
780 * the total space used in the snapshot and subtract out the space that
781 * has been freed up since the snapshot was taken.
34dc7c2f 782 */
13fe0198
MA		 783void
784dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds)
34dc7c2f 785{
13fe0198
MA		 786 uint64_t mrs_used;
787 uint64_t dlused, dlcomp, dluncomp;
34dc7c2f 788
13fe0198 789 ASSERT(!dsl_dataset_is_snapshot(ds));
34dc7c2f 790
13fe0198
MA		 791 if (ds->ds_phys->ds_prev_snap_obj != 0)
792 mrs_used = ds->ds_prev->ds_phys->ds_referenced_bytes;
793 else
794 mrs_used = 0;
45d1cae3 795
13fe0198 796 dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp);
34dc7c2f 797
13fe0198
MA		 798 ASSERT3U(dlused, <=, mrs_used);
799 ds->ds_phys->ds_unique_bytes =
800 ds->ds_phys->ds_referenced_bytes - (mrs_used - dlused);
330d06f9 801
13fe0198
MA		 802 if (spa_version(ds->ds_dir->dd_pool->dp_spa) >=
803 SPA_VERSION_UNIQUE_ACCURATE)
804 ds->ds_phys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
34dc7c2f
BB		 805}
806
13fe0198
MA		 807void
808dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj,
809 dmu_tx_t *tx)
45d1cae3 810{
13fe0198
MA		 811 objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
812 int err;
813 ASSERTV(uint64_t count);
814
815 ASSERT(ds->ds_phys->ds_num_children >= 2);
816 err = zap_remove_int(mos, ds->ds_phys->ds_next_clones_obj, obj, tx);
817 /*
818 * The err should not be ENOENT, but a bug in a previous version
819 * of the code could cause upgrade_clones_cb() to not set
820 * ds_next_snap_obj when it should, leading to a missing entry.
821 * If we knew that the pool was created after
822 * SPA_VERSION_NEXT_CLONES, we could assert that it isn't
823 * ENOENT. However, at least we can check that we don't have
824 * too many entries in the next_clones_obj even after failing to
825 * remove this one.
826 */
827 if (err != ENOENT)
828 VERIFY0(err);
829 ASSERT0(zap_count(mos, ds->ds_phys->ds_next_clones_obj,
830 &count));
831 ASSERT3U(count, <=, ds->ds_phys->ds_num_children - 2);
832}
45d1cae3 833
45d1cae3 834
13fe0198
MA		 835blkptr_t *
836dsl_dataset_get_blkptr(dsl_dataset_t *ds)
837{
838 return (&ds->ds_phys->ds_bp);
45d1cae3
BB		 839}
840
13fe0198
MA		 841void
842dsl_dataset_set_blkptr(dsl_dataset_t *ds, blkptr_t *bp, dmu_tx_t *tx)
45d1cae3 843{
13fe0198
MA		 844 ASSERT(dmu_tx_is_syncing(tx));
845 /* If it's the meta-objset, set dp_meta_rootbp */
846 if (ds == NULL) {
847 tx->tx_pool->dp_meta_rootbp = *bp;
848 } else {
849 dmu_buf_will_dirty(ds->ds_dbuf, tx);
850 ds->ds_phys->ds_bp = *bp;
45d1cae3 851 }
13fe0198 852}
45d1cae3 853
13fe0198
MA		 854spa_t *
855dsl_dataset_get_spa(dsl_dataset_t *ds)
856{
857 return (ds->ds_dir->dd_pool->dp_spa);
45d1cae3
BB		 858}
859
13fe0198
MA		 860void
861dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx)
34dc7c2f 862{
13fe0198 863 dsl_pool_t *dp;
45d1cae3 864
13fe0198
MA		 865 if (ds == NULL) /* this is the meta-objset */
866 return;
34dc7c2f 867
428870ff 868 ASSERT(ds->ds_objset != NULL);
34dc7c2f
BB		 869
870 if (ds->ds_phys->ds_next_snap_obj != 0)
871 panic("dirtying snapshot!");
872
873 dp = ds->ds_dir->dd_pool;
874
13fe0198 875 if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) {
34dc7c2f
BB		 876 /* up the hold count until we can be written out */
877 dmu_buf_add_ref(ds->ds_dbuf, ds);
878 }
879}
880
04434775
MA		 881boolean_t
882dsl_dataset_is_dirty(dsl_dataset_t *ds)
883{
884 int t;
885
886 for (t = 0; t < TXG_SIZE; t++) {
887 if (txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets,
888 ds, t))
889 return (B_TRUE);
890 }
891 return (B_FALSE);
892}
893
34dc7c2f 894static int
13fe0198 895dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx)
34dc7c2f 896{
13fe0198 897 uint64_t asize;
34dc7c2f 898
13fe0198 899 if (!dmu_tx_is_syncing(tx))
b128c09f
BB		 900 return (0);
901
34dc7c2f 902 /*
13fe0198
MA		 903 * If there's an fs-only reservation, any blocks that might become
904 * owned by the snapshot dataset must be accommodated by space
905 * outside of the reservation.
34dc7c2f 906 */
13fe0198
MA		 907 ASSERT(ds->ds_reserved == 0 || DS_UNIQUE_IS_ACCURATE(ds));
908 asize = MIN(ds->ds_phys->ds_unique_bytes, ds->ds_reserved);
909 if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE))
2e528b49 910 return (SET_ERROR(ENOSPC));
34dc7c2f
BB		 911
912 /*
13fe0198
MA		 913 * Propagate any reserved space for this snapshot to other
914 * snapshot checks in this sync group.
34dc7c2f 915 */
13fe0198
MA		 916 if (asize > 0)
917 dsl_dir_willuse_space(ds->ds_dir, asize, tx);
34dc7c2f
BB		 918
919 return (0);
920}
921
13fe0198
MA		 922typedef struct dsl_dataset_snapshot_arg {
923 nvlist_t *ddsa_snaps;
924 nvlist_t *ddsa_props;
925 nvlist_t *ddsa_errors;
926} dsl_dataset_snapshot_arg_t;
45d1cae3 927
34dc7c2f 928int
13fe0198
MA		 929dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname,
930 dmu_tx_t *tx)
34dc7c2f 931{
13fe0198
MA		 932 int error;
933 uint64_t value;
34dc7c2f 934
13fe0198 935 ds->ds_trysnap_txg = tx->tx_txg;
b128c09f 936
13fe0198 937 if (!dmu_tx_is_syncing(tx))
45d1cae3 938 return (0);
34dc7c2f
BB		 939
940 /*
13fe0198
MA		 941 * We don't allow multiple snapshots of the same txg. If there
942 * is already one, try again.
34dc7c2f 943 */
13fe0198 944 if (ds->ds_phys->ds_prev_snap_txg >= tx->tx_txg)
2e528b49 945 return (SET_ERROR(EAGAIN));
34dc7c2f
BB		 946
947 /*
13fe0198 948 * Check for conflicting snapshot name.
34dc7c2f 949 */
13fe0198
MA		 950 error = dsl_dataset_snap_lookup(ds, snapname, &value);
951 if (error == 0)
2e528b49 952 return (SET_ERROR(EEXIST));
13fe0198
MA		 953 if (error != ENOENT)
954 return (error);
45d1cae3 955
13fe0198
MA		 956 error = dsl_dataset_snapshot_reserve_space(ds, tx);
957 if (error != 0)
958 return (error);
45d1cae3 959
34dc7c2f
BB		 960 return (0);
961}
962
13fe0198
MA		 963static int
964dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx)
b128c09f 965{
13fe0198
MA		 966 dsl_dataset_snapshot_arg_t *ddsa = arg;
967 dsl_pool_t *dp = dmu_tx_pool(tx);
968 nvpair_t *pair;
969 int rv = 0;
b128c09f 970
13fe0198
MA		 971 for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
972 pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
973 int error = 0;
974 dsl_dataset_t *ds;
975 char *name, *atp;
976 char dsname[MAXNAMELEN];
977
978 name = nvpair_name(pair);
979 if (strlen(name) >= MAXNAMELEN)
2e528b49 980 error = SET_ERROR(ENAMETOOLONG);
13fe0198
MA		 981 if (error == 0) {
982 atp = strchr(name, '@');
983 if (atp == NULL)
2e528b49 984 error = SET_ERROR(EINVAL);
13fe0198
MA		 985 if (error == 0)
986 (void) strlcpy(dsname, name, atp - name + 1);
987 }
988 if (error == 0)
989 error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
990 if (error == 0) {
991 error = dsl_dataset_snapshot_check_impl(ds,
992 atp + 1, tx);
993 dsl_dataset_rele(ds, FTAG);
994 }
b128c09f 995
13fe0198
MA		 996 if (error != 0) {
997 if (ddsa->ddsa_errors != NULL) {
998 fnvlist_add_int32(ddsa->ddsa_errors,
999 name, error);
1000 }
1001 rv = error;
1002 }
1003 }
1004 return (rv);
b128c09f
BB		 1005}
1006
13fe0198
MA		 1007void
1008dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname,
1009 dmu_tx_t *tx)
428870ff 1010{
13fe0198
MA		 1011 dsl_pool_t *dp = ds->ds_dir->dd_pool;
1012 dmu_buf_t *dbuf;
1013 dsl_dataset_phys_t *dsphys;
1014 uint64_t dsobj, crtxg;
1015 objset_t *mos = dp->dp_meta_objset;
1016 ASSERTV(static zil_header_t zero_zil);
1017 ASSERTV(objset_t *os);
1018
1019 ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
428870ff 1020
428870ff 1021 /*
13fe0198
MA		 1022 * If we are on an old pool, the zil must not be active, in which
1023 * case it will be zeroed. Usually zil_suspend() accomplishes this.
428870ff 1024 */
13fe0198
MA		 1025 ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP ||
1026 dmu_objset_from_ds(ds, &os) != 0 ||
1027 bcmp(&os->os_phys->os_zil_header, &zero_zil,
1028 sizeof (zero_zil)) == 0);
428870ff 1029
428870ff
BB		 1030
1031 /*
13fe0198 1032 * The origin's ds_creation_txg has to be < TXG_INITIAL
b128c09f
BB		 1033 */
1034 if (strcmp(snapname, ORIGIN_DIR_NAME) == 0)
1035 crtxg = 1;
1036 else
1037 crtxg = tx->tx_txg;
1038
34dc7c2f
BB		 1039 dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
1040 DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
13fe0198 1041 VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
34dc7c2f
BB		 1042 dmu_buf_will_dirty(dbuf, tx);
1043 dsphys = dbuf->db_data;
b128c09f 1044 bzero(dsphys, sizeof (dsl_dataset_phys_t));
34dc7c2f
BB		 1045 dsphys->ds_dir_obj = ds->ds_dir->dd_object;
1046 dsphys->ds_fsid_guid = unique_create();
1047 (void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
1048 sizeof (dsphys->ds_guid));
1049 dsphys->ds_prev_snap_obj = ds->ds_phys->ds_prev_snap_obj;
1050 dsphys->ds_prev_snap_txg = ds->ds_phys->ds_prev_snap_txg;
1051 dsphys->ds_next_snap_obj = ds->ds_object;
1052 dsphys->ds_num_children = 1;
1053 dsphys->ds_creation_time = gethrestime_sec();
b128c09f 1054 dsphys->ds_creation_txg = crtxg;
34dc7c2f 1055 dsphys->ds_deadlist_obj = ds->ds_phys->ds_deadlist_obj;
9ae529ec 1056 dsphys->ds_referenced_bytes = ds->ds_phys->ds_referenced_bytes;
34dc7c2f
BB		 1057 dsphys->ds_compressed_bytes = ds->ds_phys->ds_compressed_bytes;
1058 dsphys->ds_uncompressed_bytes = ds->ds_phys->ds_uncompressed_bytes;
1059 dsphys->ds_flags = ds->ds_phys->ds_flags;
1060 dsphys->ds_bp = ds->ds_phys->ds_bp;
1061 dmu_buf_rele(dbuf, FTAG);
1062
1063 ASSERT3U(ds->ds_prev != 0, ==, ds->ds_phys->ds_prev_snap_obj != 0);
1064 if (ds->ds_prev) {
b128c09f
BB		 1065 uint64_t next_clones_obj =
1066 ds->ds_prev->ds_phys->ds_next_clones_obj;
34dc7c2f
BB		 1067 ASSERT(ds->ds_prev->ds_phys->ds_next_snap_obj ==
1068 ds->ds_object ||
1069 ds->ds_prev->ds_phys->ds_num_children > 1);
1070 if (ds->ds_prev->ds_phys->ds_next_snap_obj == ds->ds_object) {
1071 dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
1072 ASSERT3U(ds->ds_phys->ds_prev_snap_txg, ==,
1073 ds->ds_prev->ds_phys->ds_creation_txg);
1074 ds->ds_prev->ds_phys->ds_next_snap_obj = dsobj;
b128c09f 1075 } else if (next_clones_obj != 0) {
13fe0198 1076 dsl_dataset_remove_from_next_clones(ds->ds_prev,
428870ff 1077 dsphys->ds_next_snap_obj, tx);
13fe0198 1078 VERIFY0(zap_add_int(mos,
b128c09f 1079 next_clones_obj, dsobj, tx));
34dc7c2f
BB		 1080 }
1081 }
1082
1083 /*
1084 * If we have a reference-reservation on this dataset, we will
1085 * need to increase the amount of refreservation being charged
1086 * since our unique space is going to zero.
1087 */
1088 if (ds->ds_reserved) {
428870ff
BB		 1089 int64_t delta;
1090 ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
1091 delta = MIN(ds->ds_phys->ds_unique_bytes, ds->ds_reserved);
b128c09f 1092 dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV,
428870ff 1093 delta, 0, 0, tx);
34dc7c2f
BB		 1094 }
1095
34dc7c2f 1096 dmu_buf_will_dirty(ds->ds_dbuf, tx);
428870ff
BB		 1097 ds->ds_phys->ds_deadlist_obj = dsl_deadlist_clone(&ds->ds_deadlist,
1098 UINT64_MAX, ds->ds_phys->ds_prev_snap_obj, tx);
1099 dsl_deadlist_close(&ds->ds_deadlist);
1100 dsl_deadlist_open(&ds->ds_deadlist, mos, ds->ds_phys->ds_deadlist_obj);
1101 dsl_deadlist_add_key(&ds->ds_deadlist,
1102 ds->ds_phys->ds_prev_snap_txg, tx);
1103
34dc7c2f
BB		 1104 ASSERT3U(ds->ds_phys->ds_prev_snap_txg, <, tx->tx_txg);
1105 ds->ds_phys->ds_prev_snap_obj = dsobj;
b128c09f 1106 ds->ds_phys->ds_prev_snap_txg = crtxg;
34dc7c2f
BB		 1107 ds->ds_phys->ds_unique_bytes = 0;
1108 if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
1109 ds->ds_phys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
34dc7c2f 1110
13fe0198
MA		 1111 VERIFY0(zap_add(mos, ds->ds_phys->ds_snapnames_zapobj,
1112 snapname, 8, 1, &dsobj, tx));
34dc7c2f
BB		 1113
1114 if (ds->ds_prev)
13fe0198
MA		 1115 dsl_dataset_rele(ds->ds_prev, ds);
1116 VERIFY0(dsl_dataset_hold_obj(dp,
b128c09f
BB		 1117 ds->ds_phys->ds_prev_snap_obj, ds, &ds->ds_prev));
1118
428870ff
BB		 1119 dsl_scan_ds_snapshotted(ds, tx);
1120
1121 dsl_dir_snap_cmtime_update(ds->ds_dir);
34dc7c2f 1122
6f1ffb06 1123 spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, "");
34dc7c2f
BB		 1124}
1125
13fe0198
MA		 1126static void
1127dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx)
34dc7c2f 1128{
13fe0198
MA		 1129 dsl_dataset_snapshot_arg_t *ddsa = arg;
1130 dsl_pool_t *dp = dmu_tx_pool(tx);
1131 nvpair_t *pair;
34dc7c2f 1132
13fe0198
MA		 1133 for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
1134 pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
1135 dsl_dataset_t *ds;
1136 char *name, *atp;
1137 char dsname[MAXNAMELEN];
1138
1139 name = nvpair_name(pair);
1140 atp = strchr(name, '@');
1141 (void) strlcpy(dsname, name, atp - name + 1);
1142 VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds));
1143
1144 dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx);
1145 if (ddsa->ddsa_props != NULL) {
1146 dsl_props_set_sync_impl(ds->ds_prev,
1147 ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx);
1148 }
1149 dsl_dataset_rele(ds, FTAG);
1150 }
34dc7c2f
BB		 1151}
1152
13fe0198
MA		 1153/*
1154 * The snapshots must all be in the same pool.
1155 * All-or-nothing: if there are any failures, nothing will be modified.
1156 */
1157int
1158dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors)
330d06f9 1159{
13fe0198
MA		 1160 dsl_dataset_snapshot_arg_t ddsa;
1161 nvpair_t *pair;
1162 boolean_t needsuspend;
1163 int error;
1164 spa_t *spa;
1165 char *firstname;
1166 nvlist_t *suspended = NULL;
330d06f9 1167
13fe0198
MA		 1168 pair = nvlist_next_nvpair(snaps, NULL);
1169 if (pair == NULL)
1170 return (0);
1171 firstname = nvpair_name(pair);
1172
1173 error = spa_open(firstname, &spa, FTAG);
1174 if (error != 0)
1175 return (error);
1176 needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
1177 spa_close(spa, FTAG);
1178
1179 if (needsuspend) {
1180 suspended = fnvlist_alloc();
1181 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1182 pair = nvlist_next_nvpair(snaps, pair)) {
1183 char fsname[MAXNAMELEN];
1184 char *snapname = nvpair_name(pair);
1185 char *atp;
1186 void *cookie;
1187
1188 atp = strchr(snapname, '@');
1189 if (atp == NULL) {
2e528b49 1190 error = SET_ERROR(EINVAL);
13fe0198
MA		 1191 break;
1192 }
1193 (void) strlcpy(fsname, snapname, atp - snapname + 1);
1194
1195 error = zil_suspend(fsname, &cookie);
1196 if (error != 0)
1197 break;
1198 fnvlist_add_uint64(suspended, fsname,
1199 (uintptr_t)cookie);
1200 }
1201 }
1202
1203 ddsa.ddsa_snaps = snaps;
1204 ddsa.ddsa_props = props;
1205 ddsa.ddsa_errors = errors;
1206
1207 if (error == 0) {
1208 error = dsl_sync_task(firstname, dsl_dataset_snapshot_check,
1209 dsl_dataset_snapshot_sync, &ddsa,
1210 fnvlist_num_pairs(snaps) * 3);
1211 }
1212
1213 if (suspended != NULL) {
1214 for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL;
1215 pair = nvlist_next_nvpair(suspended, pair)) {
1216 zil_resume((void *)(uintptr_t)
1217 fnvpair_value_uint64(pair));
1218 }
1219 fnvlist_free(suspended);
1220 }
1221
1222 return (error);
1223}
1224
1225typedef struct dsl_dataset_snapshot_tmp_arg {
1226 const char *ddsta_fsname;
1227 const char *ddsta_snapname;
1228 minor_t ddsta_cleanup_minor;
1229 const char *ddsta_htag;
1230} dsl_dataset_snapshot_tmp_arg_t;
1231
1232static int
1233dsl_dataset_snapshot_tmp_check(void *arg, dmu_tx_t *tx)
1234{
1235 dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
1236 dsl_pool_t *dp = dmu_tx_pool(tx);
1237 dsl_dataset_t *ds;
1238 int error;
1239
1240 error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds);
1241 if (error != 0)
1242 return (error);
1243
1244 error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname, tx);
1245 if (error != 0) {
1246 dsl_dataset_rele(ds, FTAG);
1247 return (error);
1248 }
1249
1250 if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) {
1251 dsl_dataset_rele(ds, FTAG);
2e528b49 1252 return (SET_ERROR(ENOTSUP));
13fe0198
MA		 1253 }
1254 error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag,
1255 B_TRUE, tx);
1256 if (error != 0) {
1257 dsl_dataset_rele(ds, FTAG);
1258 return (error);
1259 }
1260
1261 dsl_dataset_rele(ds, FTAG);
1262 return (0);
1263}
1264
1265static void
1266dsl_dataset_snapshot_tmp_sync(void *arg, dmu_tx_t *tx)
1267{
1268 dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
1269 dsl_pool_t *dp = dmu_tx_pool(tx);
1270 dsl_dataset_t *ds;
1271
1272 VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds));
1273
1274 dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx);
1275 dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag,
1276 ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx);
1277 dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx);
1278
1279 dsl_dataset_rele(ds, FTAG);
1280}
1281
1282int
1283dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
1284 minor_t cleanup_minor, const char *htag)
1285{
1286 dsl_dataset_snapshot_tmp_arg_t ddsta;
1287 int error;
1288 spa_t *spa;
1289 boolean_t needsuspend;
1290 void *cookie;
1291
1292 ddsta.ddsta_fsname = fsname;
1293 ddsta.ddsta_snapname = snapname;
1294 ddsta.ddsta_cleanup_minor = cleanup_minor;
1295 ddsta.ddsta_htag = htag;
1296
1297 error = spa_open(fsname, &spa, FTAG);
1298 if (error != 0)
1299 return (error);
1300 needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
1301 spa_close(spa, FTAG);
1302
1303 if (needsuspend) {
1304 error = zil_suspend(fsname, &cookie);
1305 if (error != 0)
1306 return (error);
1307 }
1308
1309 error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check,
1310 dsl_dataset_snapshot_tmp_sync, &ddsta, 3);
1311
1312 if (needsuspend)
1313 zil_resume(cookie);
1314 return (error);
1315}
1316
1317
1318void
1319dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx)
1320{
1321 ASSERT(dmu_tx_is_syncing(tx));
1322 ASSERT(ds->ds_objset != NULL);
1323 ASSERT(ds->ds_phys->ds_next_snap_obj == 0);
1324
1325 /*
1326 * in case we had to change ds_fsid_guid when we opened it,
1327 * sync it out now.
1328 */
1329 dmu_buf_will_dirty(ds->ds_dbuf, tx);
1330 ds->ds_phys->ds_fsid_guid = ds->ds_fsid_guid;
1331
1332 dmu_objset_sync(ds->ds_objset, zio, tx);
1333}
1334
1335static void
1336get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv)
1337{
1338 uint64_t count = 0;
1339 objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
1340 zap_cursor_t zc;
1341 zap_attribute_t za;
1342 nvlist_t *propval = fnvlist_alloc();
1343 nvlist_t *val = fnvlist_alloc();
1344
1345 ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
330d06f9
MA		 1346
1347 /*
13fe0198 1348 * There may be missing entries in ds_next_clones_obj
330d06f9
MA		 1349 * due to a bug in a previous version of the code.
1350 * Only trust it if it has the right number of entries.
1351 */
1352 if (ds->ds_phys->ds_next_clones_obj != 0) {
13fe0198 1353 ASSERT0(zap_count(mos, ds->ds_phys->ds_next_clones_obj,
330d06f9
MA		 1354 &count));
1355 }
13fe0198 1356 if (count != ds->ds_phys->ds_num_children - 1)
330d06f9 1357 goto fail;
330d06f9
MA		 1358 for (zap_cursor_init(&zc, mos, ds->ds_phys->ds_next_clones_obj);
1359 zap_cursor_retrieve(&zc, &za) == 0;
1360 zap_cursor_advance(&zc)) {
1361 dsl_dataset_t *clone;
1362 char buf[ZFS_MAXNAMELEN];
13fe0198
MA		 1363 VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,
1364 za.za_first_integer, FTAG, &clone));
330d06f9 1365 dsl_dir_name(clone->ds_dir, buf);
13fe0198 1366 fnvlist_add_boolean(val, buf);
330d06f9
MA		 1367 dsl_dataset_rele(clone, FTAG);
1368 }
1369 zap_cursor_fini(&zc);
13fe0198
MA		 1370 fnvlist_add_nvlist(propval, ZPROP_VALUE, val);
1371 fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES), propval);
330d06f9
MA		 1372fail:
1373 nvlist_free(val);
1374 nvlist_free(propval);
330d06f9
MA		 1375}
1376
34dc7c2f
BB		 1377void
1378dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv)
1379{
f5fc4aca 1380 uint64_t refd, avail, uobjs, aobjs, ratio;
13fe0198
MA		 1381 ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool);
1382
1383 ASSERT(dsl_pool_config_held(dp));
34dc7c2f 1384
6f1ffb06
MA		 1385 ratio = ds->ds_phys->ds_compressed_bytes == 0 ? 100 :
1386 (ds->ds_phys->ds_uncompressed_bytes * 100 /
1387 ds->ds_phys->ds_compressed_bytes);
1388
1389 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO, ratio);
24a64651
MA		 1390 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED,
1391 ds->ds_phys->ds_uncompressed_bytes);
6f1ffb06
MA		 1392
1393 if (dsl_dataset_is_snapshot(ds)) {
1394 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO, ratio);
1395 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED,
1396 ds->ds_phys->ds_unique_bytes);
1397 get_clones_stat(ds, nv);
1398 } else {
1399 dsl_dir_stats(ds->ds_dir, nv);
1400 }
34dc7c2f
BB		 1401
1402 dsl_dataset_space(ds, &refd, &avail, &uobjs, &aobjs);
1403 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE, avail);
1404 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED, refd);
1405
1406 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION,
1407 ds->ds_phys->ds_creation_time);
1408 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG,
1409 ds->ds_phys->ds_creation_txg);
1410 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA,
1411 ds->ds_quota);
1412 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION,
1413 ds->ds_reserved);
b128c09f
BB		 1414 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID,
1415 ds->ds_phys->ds_guid);
428870ff
BB		 1416 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE,
1417 ds->ds_phys->ds_unique_bytes);
1418 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID,
1419 ds->ds_object);
1420 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS,
1421 ds->ds_userrefs);
45d1cae3
BB		 1422 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY,
1423 DS_IS_DEFER_DESTROY(ds) ? 1 : 0);
34dc7c2f 1424
330d06f9
MA		 1425 if (ds->ds_phys->ds_prev_snap_obj != 0) {
1426 uint64_t written, comp, uncomp;
1427 dsl_pool_t *dp = ds->ds_dir->dd_pool;
1428 dsl_dataset_t *prev;
1429 int err;
1430
330d06f9
MA		 1431 err = dsl_dataset_hold_obj(dp,
1432 ds->ds_phys->ds_prev_snap_obj, FTAG, &prev);
330d06f9
MA		 1433 if (err == 0) {
1434 err = dsl_dataset_space_written(prev, ds, &written,
1435 &comp, &uncomp);
1436 dsl_dataset_rele(prev, FTAG);
1437 if (err == 0) {
1438 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN,
1439 written);
1440 }
1441 }
1442 }
1443
34dc7c2f
BB		 1444}
1445
1446void
1447dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat)
1448{
13fe0198
MA		 1449 dsl_pool_t *dp = ds->ds_dir->dd_pool;
1450 ASSERT(dsl_pool_config_held(dp));
1451
34dc7c2f
BB		 1452 stat->dds_creation_txg = ds->ds_phys->ds_creation_txg;
1453 stat->dds_inconsistent = ds->ds_phys->ds_flags & DS_FLAG_INCONSISTENT;
1454 stat->dds_guid = ds->ds_phys->ds_guid;
6f1ffb06
MA		 1455 stat->dds_origin[0] = '\0';
1456 if (dsl_dataset_is_snapshot(ds)) {
34dc7c2f
BB		 1457 stat->dds_is_snapshot = B_TRUE;
1458 stat->dds_num_clones = ds->ds_phys->ds_num_children - 1;
fb5f0bc8
BB		 1459 } else {
1460 stat->dds_is_snapshot = B_FALSE;
1461 stat->dds_num_clones = 0;
34dc7c2f 1462
6f1ffb06
MA		 1463 if (dsl_dir_is_clone(ds->ds_dir)) {
1464 dsl_dataset_t *ods;
34dc7c2f 1465
13fe0198 1466 VERIFY0(dsl_dataset_hold_obj(dp,
6f1ffb06
MA		 1467 ds->ds_dir->dd_phys->dd_origin_obj, FTAG, &ods));
1468 dsl_dataset_name(ods, stat->dds_origin);
13fe0198 1469 dsl_dataset_rele(ods, FTAG);
6f1ffb06 1470 }
34dc7c2f 1471 }
34dc7c2f
BB		 1472}
1473
1474uint64_t
1475dsl_dataset_fsid_guid(dsl_dataset_t *ds)
1476{
1477 return (ds->ds_fsid_guid);
1478}
1479
1480void
1481dsl_dataset_space(dsl_dataset_t *ds,
1482 uint64_t *refdbytesp, uint64_t *availbytesp,
1483 uint64_t *usedobjsp, uint64_t *availobjsp)
1484{
9ae529ec 1485 *refdbytesp = ds->ds_phys->ds_referenced_bytes;
34dc7c2f
BB		 1486 *availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE);
1487 if (ds->ds_reserved > ds->ds_phys->ds_unique_bytes)
1488 *availbytesp += ds->ds_reserved - ds->ds_phys->ds_unique_bytes;
1489 if (ds->ds_quota != 0) {
1490 /*
1491 * Adjust available bytes according to refquota
1492 */
1493 if (*refdbytesp < ds->ds_quota)
1494 *availbytesp = MIN(*availbytesp,
1495 ds->ds_quota - *refdbytesp);
1496 else
1497 *availbytesp = 0;
1498 }
1499 *usedobjsp = ds->ds_phys->ds_bp.blk_fill;
1500 *availobjsp = DN_MAX_OBJECT - *usedobjsp;
1501}
1502
1503boolean_t
1504dsl_dataset_modified_since_lastsnap(dsl_dataset_t *ds)
1505{
1fde1e37 1506 ASSERTV(dsl_pool_t *dp = ds->ds_dir->dd_pool);
34dc7c2f 1507
13fe0198 1508 ASSERT(dsl_pool_config_held(dp));
34dc7c2f
BB		 1509 if (ds->ds_prev == NULL)
1510 return (B_FALSE);
1511 if (ds->ds_phys->ds_bp.blk_birth >
572e2857
BB		 1512 ds->ds_prev->ds_phys->ds_creation_txg) {
1513 objset_t *os, *os_prev;
1514 /*
1515 * It may be that only the ZIL differs, because it was
1516 * reset in the head. Don't count that as being
1517 * modified.
1518 */
1519 if (dmu_objset_from_ds(ds, &os) != 0)
1520 return (B_TRUE);
1521 if (dmu_objset_from_ds(ds->ds_prev, &os_prev) != 0)
1522 return (B_TRUE);
1523 return (bcmp(&os->os_phys->os_meta_dnode,
1524 &os_prev->os_phys->os_meta_dnode,
1525 sizeof (os->os_phys->os_meta_dnode)) != 0);
1526 }
34dc7c2f
BB		 1527 return (B_FALSE);
1528}
1529
13fe0198
MA		 1530typedef struct dsl_dataset_rename_snapshot_arg {
1531 const char *ddrsa_fsname;
1532 const char *ddrsa_oldsnapname;
1533 const char *ddrsa_newsnapname;
1534 boolean_t ddrsa_recursive;
1535 dmu_tx_t *ddrsa_tx;
1536} dsl_dataset_rename_snapshot_arg_t;
1537
34dc7c2f
BB		 1538/* ARGSUSED */
1539static int
13fe0198
MA		 1540dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp,
1541 dsl_dataset_t *hds, void *arg)
34dc7c2f 1542{
13fe0198
MA		 1543 dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
1544 int error;
34dc7c2f 1545 uint64_t val;
34dc7c2f 1546
13fe0198
MA		 1547 error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
1548 if (error != 0) {
1549 /* ignore nonexistent snapshots */
1550 return (error == ENOENT ? 0 : error);
1551 }
34dc7c2f 1552
13fe0198
MA		 1553 /* new name should not exist */
1554 error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val);
1555 if (error == 0)
2e528b49 1556 error = SET_ERROR(EEXIST);
13fe0198
MA		 1557 else if (error == ENOENT)
1558 error = 0;
34dc7c2f
BB		 1559
1560 /* dataset name + 1 for the "@" + the new snapshot name must fit */
13fe0198
MA		 1561 if (dsl_dir_namelen(hds->ds_dir) + 1 +
1562 strlen(ddrsa->ddrsa_newsnapname) >= MAXNAMELEN)
2e528b49 1563 error = SET_ERROR(ENAMETOOLONG);
34dc7c2f 1564
13fe0198 1565 return (error);
34dc7c2f
BB		 1566}
1567
13fe0198
MA		 1568static int
1569dsl_dataset_rename_snapshot_check(void *arg, dmu_tx_t *tx)
34dc7c2f 1570{
13fe0198
MA		 1571 dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
1572 dsl_pool_t *dp = dmu_tx_pool(tx);
34dc7c2f 1573 dsl_dataset_t *hds;
13fe0198 1574 int error;
34dc7c2f 1575
13fe0198
MA		 1576 error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds);
1577 if (error != 0)
1578 return (error);
34dc7c2f 1579
13fe0198
MA		 1580 if (ddrsa->ddrsa_recursive) {
1581 error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
1582 dsl_dataset_rename_snapshot_check_impl, ddrsa,
1583 DS_FIND_CHILDREN);
1584 } else {
1585 error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa);
1586 }
b128c09f 1587 dsl_dataset_rele(hds, FTAG);
13fe0198 1588 return (error);
34dc7c2f
BB		 1589}
1590
34dc7c2f 1591static int
13fe0198
MA		 1592dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp,
1593 dsl_dataset_t *hds, void *arg)
34dc7c2f 1594{
13fe0198
MA		 1595 dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
1596 dsl_dataset_t *ds;
1597 uint64_t val;
1598 dmu_tx_t *tx = ddrsa->ddrsa_tx;
1599 int error;
34dc7c2f 1600
13fe0198
MA		 1601 error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
1602 ASSERT(error == 0 || error == ENOENT);
1603 if (error == ENOENT) {
1604 /* ignore nonexistent snapshots */
1605 return (0);
34dc7c2f
BB		 1606 }
1607
13fe0198
MA		 1608 VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds));
1609
1610 /* log before we change the name */
1611 spa_history_log_internal_ds(ds, "rename", tx,
1612 "-> @%s", ddrsa->ddrsa_newsnapname);
34dc7c2f 1613
13fe0198
MA		 1614 VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx));
1615 mutex_enter(&ds->ds_lock);
1616 (void) strcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname);
1617 mutex_exit(&ds->ds_lock);
1618 VERIFY0(zap_add(dp->dp_meta_objset, hds->ds_phys->ds_snapnames_zapobj,
1619 ds->ds_snapname, 8, 1, &ds->ds_object, tx));
34dc7c2f 1620
13fe0198 1621 dsl_dataset_rele(ds, FTAG);
34dc7c2f
BB		 1622 return (0);
1623}
1624
13fe0198
MA		 1625static void
1626dsl_dataset_rename_snapshot_sync(void *arg, dmu_tx_t *tx)
34dc7c2f 1627{
13fe0198
MA		 1628 dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
1629 dsl_pool_t *dp = dmu_tx_pool(tx);
1630 dsl_dataset_t *hds;
34dc7c2f 1631
13fe0198
MA		 1632 VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds));
1633 ddrsa->ddrsa_tx = tx;
1634 if (ddrsa->ddrsa_recursive) {
1635 VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
1636 dsl_dataset_rename_snapshot_sync_impl, ddrsa,
1637 DS_FIND_CHILDREN));
1638 } else {
1639 VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa));
34dc7c2f 1640 }
13fe0198 1641 dsl_dataset_rele(hds, FTAG);
34dc7c2f
BB		 1642}
1643
13fe0198
MA		 1644int
1645dsl_dataset_rename_snapshot(const char *fsname,
1646 const char *oldsnapname, const char *newsnapname, boolean_t recursive)
34dc7c2f 1647{
13fe0198 1648 dsl_dataset_rename_snapshot_arg_t ddrsa;
34dc7c2f 1649
13fe0198
MA		 1650 ddrsa.ddrsa_fsname = fsname;
1651 ddrsa.ddrsa_oldsnapname = oldsnapname;
1652 ddrsa.ddrsa_newsnapname = newsnapname;
1653 ddrsa.ddrsa_recursive = recursive;
34dc7c2f 1654
13fe0198
MA		 1655 return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check,
1656 dsl_dataset_rename_snapshot_sync, &ddrsa, 1));
34dc7c2f
BB		 1657}
1658
13fe0198
MA		 1659static int
1660dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx)
34dc7c2f 1661{
13fe0198
MA		 1662 const char *fsname = arg;
1663 dsl_pool_t *dp = dmu_tx_pool(tx);
34dc7c2f 1664 dsl_dataset_t *ds;
13fe0198
MA		 1665 int64_t unused_refres_delta;
1666 int error;
34dc7c2f 1667
13fe0198
MA		 1668 error = dsl_dataset_hold(dp, fsname, FTAG, &ds);
1669 if (error != 0)
1670 return (error);
428870ff 1671
13fe0198
MA		 1672 /* must not be a snapshot */
1673 if (dsl_dataset_is_snapshot(ds)) {
1674 dsl_dataset_rele(ds, FTAG);
2e528b49 1675 return (SET_ERROR(EINVAL));
13fe0198 1676 }
34dc7c2f 1677
13fe0198
MA		 1678 /* must have a most recent snapshot */
1679 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
1680 dsl_dataset_rele(ds, FTAG);
2e528b49 1681 return (SET_ERROR(EINVAL));
13fe0198 1682 }
34dc7c2f 1683
13fe0198
MA		 1684 if (dsl_dataset_long_held(ds)) {
1685 dsl_dataset_rele(ds, FTAG);
2e528b49 1686 return (SET_ERROR(EBUSY));
34dc7c2f 1687 }
428870ff 1688
13fe0198
MA		 1689 /*
1690 * Check if the snap we are rolling back to uses more than
1691 * the refquota.
1692 */
1693 if (ds->ds_quota != 0 &&
1694 ds->ds_prev->ds_phys->ds_referenced_bytes > ds->ds_quota) {
1695 dsl_dataset_rele(ds, FTAG);
2e528b49 1696 return (SET_ERROR(EDQUOT));
34dc7c2f
BB		 1697 }
1698
13fe0198
MA		 1699 /*
1700 * When we do the clone swap, we will temporarily use more space
1701 * due to the refreservation (the head will no longer have any
1702 * unique space, so the entire amount of the refreservation will need
1703 * to be free). We will immediately destroy the clone, freeing
1704 * this space, but the freeing happens over many txg's.
1705 */
1706 unused_refres_delta = (int64_t)MIN(ds->ds_reserved,
1707 ds->ds_phys->ds_unique_bytes);
34dc7c2f 1708
13fe0198
MA		 1709 if (unused_refres_delta > 0 &&
1710 unused_refres_delta >
1711 dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) {
1712 dsl_dataset_rele(ds, FTAG);
2e528b49 1713 return (SET_ERROR(ENOSPC));
13fe0198 1714 }
34dc7c2f 1715
13fe0198
MA		 1716 dsl_dataset_rele(ds, FTAG);
1717 return (0);
1718}
34dc7c2f 1719
13fe0198
MA		 1720static void
1721dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx)
1722{
1723 const char *fsname = arg;
1724 dsl_pool_t *dp = dmu_tx_pool(tx);
1725 dsl_dataset_t *ds, *clone;
1726 uint64_t cloneobj;
34dc7c2f 1727
13fe0198 1728 VERIFY0(dsl_dataset_hold(dp, fsname, FTAG, &ds));
34dc7c2f 1729
13fe0198
MA		 1730 cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback",
1731 ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, tx);
1732
1733 VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone));
1734
1735 dsl_dataset_clone_swap_sync_impl(clone, ds, tx);
1736 dsl_dataset_zero_zil(ds, tx);
1737
1738 dsl_destroy_head_sync_impl(clone, tx);
1739
1740 dsl_dataset_rele(clone, FTAG);
1741 dsl_dataset_rele(ds, FTAG);
1742}
1743
1744int
1745dsl_dataset_rollback(const char *fsname)
1746{
1747 return (dsl_sync_task(fsname, dsl_dataset_rollback_check,
1748 dsl_dataset_rollback_sync, (void *)fsname, 1));
34dc7c2f
BB		 1749}
1750
b128c09f
BB		 1751struct promotenode {
1752 list_node_t link;
1753 dsl_dataset_t *ds;
1754};
1755
13fe0198
MA		 1756typedef struct dsl_dataset_promote_arg {
1757 const char *ddpa_clonename;
1758 dsl_dataset_t *ddpa_clone;
b128c09f 1759 list_t shared_snaps, origin_snaps, clone_snaps;
13fe0198 1760 dsl_dataset_t *origin_origin; /* origin of the origin */
b128c09f 1761 uint64_t used, comp, uncomp, unique, cloneusedsnap, originusedsnap;
428870ff 1762 char *err_ds;
13fe0198 1763} dsl_dataset_promote_arg_t;
34dc7c2f 1764
b128c09f 1765static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep);
13fe0198
MA		 1766static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp,
1767 void *tag);
1768static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag);
b128c09f 1769
34dc7c2f 1770static int
13fe0198 1771dsl_dataset_promote_check(void *arg, dmu_tx_t *tx)
34dc7c2f 1772{
13fe0198
MA		 1773 dsl_dataset_promote_arg_t *ddpa = arg;
1774 dsl_pool_t *dp = dmu_tx_pool(tx);
1775 dsl_dataset_t *hds;
1776 struct promotenode *snap;
1777 dsl_dataset_t *origin_ds;
34dc7c2f 1778 int err;
428870ff 1779 uint64_t unused;
34dc7c2f 1780
13fe0198
MA		 1781 err = promote_hold(ddpa, dp, FTAG);
1782 if (err != 0)
1783 return (err);
34dc7c2f 1784
13fe0198 1785 hds = ddpa->ddpa_clone;
34dc7c2f 1786
13fe0198
MA		 1787 if (hds->ds_phys->ds_flags & DS_FLAG_NOPROMOTE) {
1788 promote_rele(ddpa, FTAG);
2e528b49 1789 return (SET_ERROR(EXDEV));
13fe0198
MA		 1790 }
1791
1792 /*
1793 * Compute and check the amount of space to transfer. Since this is
1794 * so expensive, don't do the preliminary check.
1795 */
1796 if (!dmu_tx_is_syncing(tx)) {
1797 promote_rele(ddpa, FTAG);
1798 return (0);
1799 }
1800
1801 snap = list_head(&ddpa->shared_snaps);
1802 origin_ds = snap->ds;
34dc7c2f
BB		 1803
1804 /* compute origin's new unique space */
13fe0198 1805 snap = list_tail(&ddpa->clone_snaps);
b128c09f 1806 ASSERT3U(snap->ds->ds_phys->ds_prev_snap_obj, ==, origin_ds->ds_object);
428870ff
BB		 1807 dsl_deadlist_space_range(&snap->ds->ds_deadlist,
1808 origin_ds->ds_phys->ds_prev_snap_txg, UINT64_MAX,
13fe0198 1809 &ddpa->unique, &unused, &unused);
34dc7c2f 1810
b128c09f
BB		 1811 /*
1812 * Walk the snapshots that we are moving
1813 *
1814 * Compute space to transfer. Consider the incremental changes
13fe0198 1815 * to used by each snapshot:
b128c09f
BB		 1816 * (my used) = (prev's used) + (blocks born) - (blocks killed)
1817 * So each snapshot gave birth to:
1818 * (blocks born) = (my used) - (prev's used) + (blocks killed)
1819 * So a sequence would look like:
1820 * (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
1821 * Which simplifies to:
1822 * uN + kN + kN-1 + ... + k1 + k0
1823 * Note however, if we stop before we reach the ORIGIN we get:
1824 * uN + kN + kN-1 + ... + kM - uM-1
1825 */
13fe0198
MA		 1826 ddpa->used = origin_ds->ds_phys->ds_referenced_bytes;
1827 ddpa->comp = origin_ds->ds_phys->ds_compressed_bytes;
1828 ddpa->uncomp = origin_ds->ds_phys->ds_uncompressed_bytes;
1829 for (snap = list_head(&ddpa->shared_snaps); snap;
1830 snap = list_next(&ddpa->shared_snaps, snap)) {
34dc7c2f 1831 uint64_t val, dlused, dlcomp, dluncomp;
b128c09f 1832 dsl_dataset_t *ds = snap->ds;
34dc7c2f 1833
13fe0198
MA		 1834 /*
1835 * If there are long holds, we won't be able to evict
1836 * the objset.
1837 */
1838 if (dsl_dataset_long_held(ds)) {
2e528b49 1839 err = SET_ERROR(EBUSY);
13fe0198
MA		 1840 goto out;
1841 }
1842
34dc7c2f 1843 /* Check that the snapshot name does not conflict */
13fe0198 1844 VERIFY0(dsl_dataset_get_snapname(ds));
b128c09f 1845 err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val);
428870ff 1846 if (err == 0) {
13fe0198 1847 (void) strcpy(ddpa->err_ds, snap->ds->ds_snapname);
2e528b49 1848 err = SET_ERROR(EEXIST);
428870ff
BB		 1849 goto out;
1850 }
b128c09f 1851 if (err != ENOENT)
428870ff 1852 goto out;
34dc7c2f 1853
b128c09f 1854 /* The very first snapshot does not have a deadlist */
34dc7c2f 1855 if (ds->ds_phys->ds_prev_snap_obj == 0)
b128c09f 1856 continue;
34dc7c2f 1857
428870ff
BB		 1858 dsl_deadlist_space(&ds->ds_deadlist,
1859 &dlused, &dlcomp, &dluncomp);
13fe0198
MA		 1860 ddpa->used += dlused;
1861 ddpa->comp += dlcomp;
1862 ddpa->uncomp += dluncomp;
b128c09f 1863 }
34dc7c2f 1864
b128c09f
BB		 1865 /*
1866 * If we are a clone of a clone then we never reached ORIGIN,
1867 * so we need to subtract out the clone origin's used space.
1868 */
13fe0198
MA		 1869 if (ddpa->origin_origin) {
1870 ddpa->used -= ddpa->origin_origin->ds_phys->ds_referenced_bytes;
1871 ddpa->comp -= ddpa->origin_origin->ds_phys->ds_compressed_bytes;
1872 ddpa->uncomp -=
1873 ddpa->origin_origin->ds_phys->ds_uncompressed_bytes;
34dc7c2f
BB		 1874 }
1875
1876 /* Check that there is enough space here */
b128c09f 1877 err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir,
13fe0198
MA		 1878 ddpa->used);
1879 if (err != 0)
1880 goto out;
34dc7c2f 1881
b128c09f
BB		 1882 /*
1883 * Compute the amounts of space that will be used by snapshots
1884 * after the promotion (for both origin and clone). For each,
1885 * it is the amount of space that will be on all of their
1886 * deadlists (that was not born before their new origin).
1887 */
1888 if (hds->ds_dir->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
1889 uint64_t space;
1890
1891 /*
1892 * Note, typically this will not be a clone of a clone,
428870ff
BB		 1893 * so dd_origin_txg will be < TXG_INITIAL, so
1894 * these snaplist_space() -> dsl_deadlist_space_range()
b128c09f
BB		 1895 * calls will be fast because they do not have to
1896 * iterate over all bps.
1897 */
13fe0198
MA		 1898 snap = list_head(&ddpa->origin_snaps);
1899 err = snaplist_space(&ddpa->shared_snaps,
1900 snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap);
1901 if (err != 0)
1902 goto out;
b128c09f 1903
13fe0198 1904 err = snaplist_space(&ddpa->clone_snaps,
428870ff 1905 snap->ds->ds_dir->dd_origin_txg, &space);
13fe0198
MA		 1906 if (err != 0)
1907 goto out;
1908 ddpa->cloneusedsnap += space;
b128c09f
BB		 1909 }
1910 if (origin_ds->ds_dir->dd_phys->dd_flags & DD_FLAG_USED_BREAKDOWN) {
13fe0198
MA		 1911 err = snaplist_space(&ddpa->origin_snaps,
1912 origin_ds->ds_phys->ds_creation_txg, &ddpa->originusedsnap);
1913 if (err != 0)
1914 goto out;
b128c09f
BB		 1915 }
1916
428870ff 1917out:
13fe0198 1918 promote_rele(ddpa, FTAG);
428870ff 1919 return (err);
34dc7c2f
BB		 1920}
1921
1922static void
13fe0198 1923dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx)
34dc7c2f 1924{
13fe0198
MA		 1925 dsl_dataset_promote_arg_t *ddpa = arg;
1926 dsl_pool_t *dp = dmu_tx_pool(tx);
1927 dsl_dataset_t *hds;
1928 struct promotenode *snap;
1929 dsl_dataset_t *origin_ds;
b128c09f 1930 dsl_dataset_t *origin_head;
13fe0198 1931 dsl_dir_t *dd;
34dc7c2f 1932 dsl_dir_t *odd = NULL;
b128c09f
BB		 1933 uint64_t oldnext_obj;
1934 int64_t delta;
34dc7c2f 1935
13fe0198
MA		 1936 VERIFY0(promote_hold(ddpa, dp, FTAG));
1937 hds = ddpa->ddpa_clone;
1938
1939 ASSERT0(hds->ds_phys->ds_flags & DS_FLAG_NOPROMOTE);
1940
1941 snap = list_head(&ddpa->shared_snaps);
1942 origin_ds = snap->ds;
1943 dd = hds->ds_dir;
34dc7c2f 1944
13fe0198 1945 snap = list_head(&ddpa->origin_snaps);
b128c09f
BB		 1946 origin_head = snap->ds;
1947
34dc7c2f
BB		 1948 /*
1949 * We need to explicitly open odd, since origin_ds's dd will be
1950 * changing.
1951 */
13fe0198 1952 VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object,
34dc7c2f
BB		 1953 NULL, FTAG, &odd));
1954
b128c09f
BB		 1955 /* change origin's next snap */
1956 dmu_buf_will_dirty(origin_ds->ds_dbuf, tx);
1957 oldnext_obj = origin_ds->ds_phys->ds_next_snap_obj;
13fe0198 1958 snap = list_tail(&ddpa->clone_snaps);
b128c09f
BB		 1959 ASSERT3U(snap->ds->ds_phys->ds_prev_snap_obj, ==, origin_ds->ds_object);
1960 origin_ds->ds_phys->ds_next_snap_obj = snap->ds->ds_object;
1961
1962 /* change the origin's next clone */
1963 if (origin_ds->ds_phys->ds_next_clones_obj) {
13fe0198
MA		 1964 dsl_dataset_remove_from_next_clones(origin_ds,
1965 snap->ds->ds_object, tx);
1966 VERIFY0(zap_add_int(dp->dp_meta_objset,
b128c09f
BB		 1967 origin_ds->ds_phys->ds_next_clones_obj,
1968 oldnext_obj, tx));
1969 }
1970
1971 /* change origin */
1972 dmu_buf_will_dirty(dd->dd_dbuf, tx);
1973 ASSERT3U(dd->dd_phys->dd_origin_obj, ==, origin_ds->ds_object);
1974 dd->dd_phys->dd_origin_obj = odd->dd_phys->dd_origin_obj;
428870ff 1975 dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg;
b128c09f
BB		 1976 dmu_buf_will_dirty(odd->dd_dbuf, tx);
1977 odd->dd_phys->dd_origin_obj = origin_ds->ds_object;
428870ff
BB		 1978 origin_head->ds_dir->dd_origin_txg =
1979 origin_ds->ds_phys->ds_creation_txg;
1980
1981 /* change dd_clone entries */
1982 if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
13fe0198 1983 VERIFY0(zap_remove_int(dp->dp_meta_objset,
428870ff 1984 odd->dd_phys->dd_clones, hds->ds_object, tx));
13fe0198
MA		 1985 VERIFY0(zap_add_int(dp->dp_meta_objset,
1986 ddpa->origin_origin->ds_dir->dd_phys->dd_clones,
428870ff
BB		 1987 hds->ds_object, tx));
1988
13fe0198
MA		 1989 VERIFY0(zap_remove_int(dp->dp_meta_objset,
1990 ddpa->origin_origin->ds_dir->dd_phys->dd_clones,
428870ff
BB		 1991 origin_head->ds_object, tx));
1992 if (dd->dd_phys->dd_clones == 0) {
1993 dd->dd_phys->dd_clones = zap_create(dp->dp_meta_objset,
1994 DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx);
1995 }
13fe0198 1996 VERIFY0(zap_add_int(dp->dp_meta_objset,
428870ff 1997 dd->dd_phys->dd_clones, origin_head->ds_object, tx));
428870ff 1998 }
34dc7c2f 1999
b128c09f 2000 /* move snapshots to this dir */
13fe0198
MA		 2001 for (snap = list_head(&ddpa->shared_snaps); snap;
2002 snap = list_next(&ddpa->shared_snaps, snap)) {
b128c09f
BB		 2003 dsl_dataset_t *ds = snap->ds;
2004
13fe0198
MA		 2005 /*
2006 * Property callbacks are registered to a particular
2007 * dsl_dir. Since ours is changing, evict the objset
2008 * so that they will be unregistered from the old dsl_dir.
2009 */
428870ff
BB		 2010 if (ds->ds_objset) {
2011 dmu_objset_evict(ds->ds_objset);
2012 ds->ds_objset = NULL;
b128c09f 2013 }
13fe0198 2014
34dc7c2f 2015 /* move snap name entry */
13fe0198
MA		 2016 VERIFY0(dsl_dataset_get_snapname(ds));
2017 VERIFY0(dsl_dataset_snap_remove(origin_head,
b128c09f 2018 ds->ds_snapname, tx));
13fe0198 2019 VERIFY0(zap_add(dp->dp_meta_objset,
34dc7c2f
BB		 2020 hds->ds_phys->ds_snapnames_zapobj, ds->ds_snapname,
2021 8, 1, &ds->ds_object, tx));
428870ff 2022
34dc7c2f
BB		 2023 /* change containing dsl_dir */
2024 dmu_buf_will_dirty(ds->ds_dbuf, tx);
2025 ASSERT3U(ds->ds_phys->ds_dir_obj, ==, odd->dd_object);
2026 ds->ds_phys->ds_dir_obj = dd->dd_object;
2027 ASSERT3P(ds->ds_dir, ==, odd);
13fe0198
MA		 2028 dsl_dir_rele(ds->ds_dir, ds);
2029 VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object,
34dc7c2f
BB		 2030 NULL, ds, &ds->ds_dir));
2031
428870ff
BB		 2032 /* move any clone references */
2033 if (ds->ds_phys->ds_next_clones_obj &&
2034 spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
2035 zap_cursor_t zc;
2036 zap_attribute_t za;
2037
2038 for (zap_cursor_init(&zc, dp->dp_meta_objset,
2039 ds->ds_phys->ds_next_clones_obj);
2040 zap_cursor_retrieve(&zc, &za) == 0;
13fe0198
MA		 2041 zap_cursor_advance(&zc)) {
2042 dsl_dataset_t *cnds;
2043 uint64_t o;
34dc7c2f 2044
13fe0198
MA		 2045 if (za.za_first_integer == oldnext_obj) {
2046 /*
2047 * We've already moved the
2048 * origin's reference.
2049 */
2050 continue;
2051 }
34dc7c2f 2052
13fe0198
MA		 2053 VERIFY0(dsl_dataset_hold_obj(dp,
2054 za.za_first_integer, FTAG, &cnds));
2055 o = cnds->ds_dir->dd_phys->dd_head_dataset_obj;
34dc7c2f 2056
13fe0198
MA		 2057 VERIFY0(zap_remove_int(dp->dp_meta_objset,
2058 odd->dd_phys->dd_clones, o, tx));
2059 VERIFY0(zap_add_int(dp->dp_meta_objset,
2060 dd->dd_phys->dd_clones, o, tx));
2061 dsl_dataset_rele(cnds, FTAG);
2062 }
2063 zap_cursor_fini(&zc);
2064 }
34dc7c2f 2065
13fe0198 2066 ASSERT(!dsl_prop_hascb(ds));
34dc7c2f
BB		 2067 }
2068
34dc7c2f 2069 /*
13fe0198
MA		 2070 * Change space accounting.
2071 * Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
2072 * both be valid, or both be 0 (resulting in delta == 0). This
2073 * is true for each of {clone,origin} independently.
34dc7c2f 2074 */
570827e1 2075
13fe0198
MA		 2076 delta = ddpa->cloneusedsnap -
2077 dd->dd_phys->dd_used_breakdown[DD_USED_SNAP];
2078 ASSERT3S(delta, >=, 0);
2079 ASSERT3U(ddpa->used, >=, delta);
2080 dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx);
2081 dsl_dir_diduse_space(dd, DD_USED_HEAD,
2082 ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx);
34dc7c2f 2083
13fe0198
MA		 2084 delta = ddpa->originusedsnap -
2085 odd->dd_phys->dd_used_breakdown[DD_USED_SNAP];
2086 ASSERT3S(delta, <=, 0);
2087 ASSERT3U(ddpa->used, >=, -delta);
2088 dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx);
2089 dsl_dir_diduse_space(odd, DD_USED_HEAD,
2090 -ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx);
2091
2092 origin_ds->ds_phys->ds_unique_bytes = ddpa->unique;
2093
2094 /* log history record */
2095 spa_history_log_internal_ds(hds, "promote", tx, "");
2096
2097 dsl_dir_rele(odd, FTAG);
2098 promote_rele(ddpa, FTAG);
34dc7c2f
BB		 2099}
2100
13fe0198
MA		 2101/*
2102 * Make a list of dsl_dataset_t's for the snapshots between first_obj
2103 * (exclusive) and last_obj (inclusive). The list will be in reverse
2104 * order (last_obj will be the list_head()). If first_obj == 0, do all
2105 * snapshots back to this dataset's origin.
2106 */
34dc7c2f 2107static int
13fe0198
MA		 2108snaplist_make(dsl_pool_t *dp,
2109 uint64_t first_obj, uint64_t last_obj, list_t *l, void *tag)
34dc7c2f 2110{
13fe0198 2111 uint64_t obj = last_obj;
34dc7c2f 2112
13fe0198
MA		 2113 list_create(l, sizeof (struct promotenode),
2114 offsetof(struct promotenode, link));
34dc7c2f 2115
13fe0198
MA		 2116 while (obj != first_obj) {
2117 dsl_dataset_t *ds;
2118 struct promotenode *snap;
2119 int err;
428870ff 2120
13fe0198
MA		 2121 err = dsl_dataset_hold_obj(dp, obj, tag, &ds);
2122 ASSERT(err != ENOENT);
2123 if (err != 0)
2124 return (err);
34dc7c2f 2125
13fe0198
MA		 2126 if (first_obj == 0)
2127 first_obj = ds->ds_dir->dd_phys->dd_origin_obj;
2128
2129 snap = kmem_alloc(sizeof (*snap), KM_SLEEP);
2130 snap->ds = ds;
2131 list_insert_tail(l, snap);
2132 obj = ds->ds_phys->ds_prev_snap_obj;
2133 }
34dc7c2f
BB		 2134
2135 return (0);
2136}
2137
13fe0198
MA		 2138static int
2139snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep)
34dc7c2f 2140{
13fe0198 2141 struct promotenode *snap;
6f1ffb06 2142
13fe0198
MA		 2143 *spacep = 0;
2144 for (snap = list_head(l); snap; snap = list_next(l, snap)) {
2145 uint64_t used, comp, uncomp;
2146 dsl_deadlist_space_range(&snap->ds->ds_deadlist,
2147 mintxg, UINT64_MAX, &used, &comp, &uncomp);
2148 *spacep += used;
428870ff 2149 }
13fe0198 2150 return (0);
34dc7c2f
BB		 2151}
2152
13fe0198
MA		 2153static void
2154snaplist_destroy(list_t *l, void *tag)
34dc7c2f 2155{
13fe0198 2156 struct promotenode *snap;
428870ff 2157
13fe0198
MA		 2158 if (l == NULL || !list_link_active(&l->list_head))
2159 return;
34dc7c2f 2160
13fe0198
MA		 2161 while ((snap = list_tail(l)) != NULL) {
2162 list_remove(l, snap);
2163 dsl_dataset_rele(snap->ds, tag);
2164 kmem_free(snap, sizeof (*snap));
2165 }
2166 list_destroy(l);
34dc7c2f
BB		 2167}
2168
2169static int
13fe0198 2170promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag)
34dc7c2f 2171{
13fe0198
MA		 2172 int error;
2173 dsl_dir_t *dd;
2174 struct promotenode *snap;
34dc7c2f 2175
13fe0198
MA		 2176 error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag,
2177 &ddpa->ddpa_clone);
2178 if (error != 0)
2179 return (error);
2180 dd = ddpa->ddpa_clone->ds_dir;
34dc7c2f 2181
13fe0198
MA		 2182 if (dsl_dataset_is_snapshot(ddpa->ddpa_clone) ||
2183 !dsl_dir_is_clone(dd)) {
2184 dsl_dataset_rele(ddpa->ddpa_clone, tag);
2e528b49 2185 return (SET_ERROR(EINVAL));
13fe0198 2186 }
34dc7c2f 2187
13fe0198
MA		 2188 error = snaplist_make(dp, 0, dd->dd_phys->dd_origin_obj,
2189 &ddpa->shared_snaps, tag);
2190 if (error != 0)
2191 goto out;
34dc7c2f 2192
13fe0198
MA		 2193 error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object,
2194 &ddpa->clone_snaps, tag);
2195 if (error != 0)
2196 goto out;
34dc7c2f 2197
13fe0198
MA		 2198 snap = list_head(&ddpa->shared_snaps);
2199 ASSERT3U(snap->ds->ds_object, ==, dd->dd_phys->dd_origin_obj);
2200 error = snaplist_make(dp, dd->dd_phys->dd_origin_obj,
2201 snap->ds->ds_dir->dd_phys->dd_head_dataset_obj,
2202 &ddpa->origin_snaps, tag);
2203 if (error != 0)
2204 goto out;
d164b209 2205
13fe0198
MA		 2206 if (snap->ds->ds_dir->dd_phys->dd_origin_obj != 0) {
2207 error = dsl_dataset_hold_obj(dp,
2208 snap->ds->ds_dir->dd_phys->dd_origin_obj,
2209 tag, &ddpa->origin_origin);
2210 if (error != 0)
2211 goto out;
d164b209 2212 }
13fe0198
MA		 2213out:
2214 if (error != 0)
2215 promote_rele(ddpa, tag);
2216 return (error);
34dc7c2f
BB		 2217}
2218
34dc7c2f 2219static void
13fe0198 2220promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag)
34dc7c2f 2221{
13fe0198
MA		 2222 snaplist_destroy(&ddpa->shared_snaps, tag);
2223 snaplist_destroy(&ddpa->clone_snaps, tag);
2224 snaplist_destroy(&ddpa->origin_snaps, tag);
2225 if (ddpa->origin_origin != NULL)
2226 dsl_dataset_rele(ddpa->origin_origin, tag);
2227 dsl_dataset_rele(ddpa->ddpa_clone, tag);
2228}
428870ff 2229
13fe0198
MA		 2230/*
2231 * Promote a clone.
2232 *
2233 * If it fails due to a conflicting snapshot name, "conflsnap" will be filled
2234 * in with the name. (It must be at least MAXNAMELEN bytes long.)
2235 */
2236int
2237dsl_dataset_promote(const char *name, char *conflsnap)
2238{
2239 dsl_dataset_promote_arg_t ddpa = { 0 };
2240 uint64_t numsnaps;
2241 int error;
2242 objset_t *os;
34dc7c2f 2243
13fe0198
MA		 2244 /*
2245 * We will modify space proportional to the number of
2246 * snapshots. Compute numsnaps.
2247 */
2248 error = dmu_objset_hold(name, FTAG, &os);
2249 if (error != 0)
2250 return (error);
2251 error = zap_count(dmu_objset_pool(os)->dp_meta_objset,
2252 dmu_objset_ds(os)->ds_phys->ds_snapnames_zapobj, &numsnaps);
2253 dmu_objset_rele(os, FTAG);
2254 if (error != 0)
2255 return (error);
34dc7c2f 2256
13fe0198
MA		 2257 ddpa.ddpa_clonename = name;
2258 ddpa.err_ds = conflsnap;
6f1ffb06 2259
13fe0198
MA		 2260 return (dsl_sync_task(name, dsl_dataset_promote_check,
2261 dsl_dataset_promote_sync, &ddpa, 2 + numsnaps));
34dc7c2f
BB		 2262}
2263
2264int
13fe0198
MA		 2265dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone,
2266 dsl_dataset_t *origin_head, boolean_t force)
34dc7c2f 2267{
13fe0198 2268 int64_t unused_refres_delta;
34dc7c2f 2269
13fe0198
MA		 2270 /* they should both be heads */
2271 if (dsl_dataset_is_snapshot(clone) ||
2272 dsl_dataset_is_snapshot(origin_head))
2e528b49 2273 return (SET_ERROR(EINVAL));
428870ff 2274
13fe0198
MA		 2275 /* the branch point should be just before them */
2276 if (clone->ds_prev != origin_head->ds_prev)
2e528b49 2277 return (SET_ERROR(EINVAL));
34dc7c2f 2278
13fe0198
MA		 2279 /* clone should be the clone (unless they are unrelated) */
2280 if (clone->ds_prev != NULL &&
2281 clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap &&
2282 origin_head->ds_object !=
2283 clone->ds_prev->ds_phys->ds_next_snap_obj)
2e528b49 2284 return (SET_ERROR(EINVAL));
428870ff 2285
13fe0198
MA		 2286 /* the clone should be a child of the origin */
2287 if (clone->ds_dir->dd_parent != origin_head->ds_dir)
2e528b49 2288 return (SET_ERROR(EINVAL));
45d1cae3 2289
13fe0198
MA		 2290 /* origin_head shouldn't be modified unless 'force' */
2291 if (!force && dsl_dataset_modified_since_lastsnap(origin_head))
2e528b49 2292 return (SET_ERROR(ETXTBSY));
572e2857 2293
13fe0198
MA		 2294 /* origin_head should have no long holds (e.g. is not mounted) */
2295 if (dsl_dataset_long_held(origin_head))
2e528b49 2296 return (SET_ERROR(EBUSY));
13fe0198
MA		 2297
2298 /* check amount of any unconsumed refreservation */
2299 unused_refres_delta =
2300 (int64_t)MIN(origin_head->ds_reserved,
2301 origin_head->ds_phys->ds_unique_bytes) -
2302 (int64_t)MIN(origin_head->ds_reserved,
2303 clone->ds_phys->ds_unique_bytes);
2304
2305 if (unused_refres_delta > 0 &&
2306 unused_refres_delta >
2307 dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE))
2e528b49 2308 return (SET_ERROR(ENOSPC));
13fe0198
MA		 2309
2310 /* clone can't be over the head's refquota */
2311 if (origin_head->ds_quota != 0 &&
2312 clone->ds_phys->ds_referenced_bytes > origin_head->ds_quota)
2e528b49 2313 return (SET_ERROR(EDQUOT));
572e2857 2314
13fe0198 2315 return (0);
572e2857
BB		 2316}
2317
2318void
13fe0198
MA		 2319dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone,
2320 dsl_dataset_t *origin_head, dmu_tx_t *tx)
572e2857 2321{
13fe0198
MA		 2322 dsl_pool_t *dp = dmu_tx_pool(tx);
2323 int64_t unused_refres_delta;
428870ff 2324
13fe0198
MA		 2325 ASSERT(clone->ds_reserved == 0);
2326 ASSERT(origin_head->ds_quota == 0 ||
2327 clone->ds_phys->ds_unique_bytes <= origin_head->ds_quota);
45d1cae3 2328
13fe0198
MA		 2329 dmu_buf_will_dirty(clone->ds_dbuf, tx);
2330 dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
45d1cae3 2331
13fe0198
MA		 2332 if (clone->ds_objset != NULL) {
2333 dmu_objset_evict(clone->ds_objset);
2334 clone->ds_objset = NULL;
2335 }
45d1cae3 2336
13fe0198
MA		 2337 if (origin_head->ds_objset != NULL) {
2338 dmu_objset_evict(origin_head->ds_objset);
2339 origin_head->ds_objset = NULL;
45d1cae3 2340 }
45d1cae3 2341
13fe0198
MA		 2342 unused_refres_delta =
2343 (int64_t)MIN(origin_head->ds_reserved,
2344 origin_head->ds_phys->ds_unique_bytes) -
2345 (int64_t)MIN(origin_head->ds_reserved,
2346 clone->ds_phys->ds_unique_bytes);
2347
2348 /*
2349 * Reset origin's unique bytes, if it exists.
2350 */
2351 if (clone->ds_prev) {
2352 dsl_dataset_t *origin = clone->ds_prev;
2353 uint64_t comp, uncomp;
2354
2355 dmu_buf_will_dirty(origin->ds_dbuf, tx);
2356 dsl_deadlist_space_range(&clone->ds_deadlist,
2357 origin->ds_phys->ds_prev_snap_txg, UINT64_MAX,
2358 &origin->ds_phys->ds_unique_bytes, &comp, &uncomp);
2359 }
2360
2361 /* swap blkptrs */
2362 {
2363 blkptr_t tmp;
2364 tmp = origin_head->ds_phys->ds_bp;
2365 origin_head->ds_phys->ds_bp = clone->ds_phys->ds_bp;
2366 clone->ds_phys->ds_bp = tmp;
2367 }
2368
2369 /* set dd_*_bytes */
2370 {
2371 int64_t dused, dcomp, duncomp;
2372 uint64_t cdl_used, cdl_comp, cdl_uncomp;
2373 uint64_t odl_used, odl_comp, odl_uncomp;
2374
2375 ASSERT3U(clone->ds_dir->dd_phys->
2376 dd_used_breakdown[DD_USED_SNAP], ==, 0);
2377
2378 dsl_deadlist_space(&clone->ds_deadlist,
2379 &cdl_used, &cdl_comp, &cdl_uncomp);
2380 dsl_deadlist_space(&origin_head->ds_deadlist,
2381 &odl_used, &odl_comp, &odl_uncomp);
428870ff 2382
13fe0198
MA		 2383 dused = clone->ds_phys->ds_referenced_bytes + cdl_used -
2384 (origin_head->ds_phys->ds_referenced_bytes + odl_used);
2385 dcomp = clone->ds_phys->ds_compressed_bytes + cdl_comp -
2386 (origin_head->ds_phys->ds_compressed_bytes + odl_comp);
2387 duncomp = clone->ds_phys->ds_uncompressed_bytes +
2388 cdl_uncomp -
2389 (origin_head->ds_phys->ds_uncompressed_bytes + odl_uncomp);
45d1cae3 2390
13fe0198
MA		 2391 dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD,
2392 dused, dcomp, duncomp, tx);
2393 dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD,
2394 -dused, -dcomp, -duncomp, tx);
45d1cae3 2395
45d1cae3 2396 /*
13fe0198
MA		 2397 * The difference in the space used by snapshots is the
2398 * difference in snapshot space due to the head's
2399 * deadlist (since that's the only thing that's
2400 * changing that affects the snapused).
45d1cae3 2401 */
13fe0198
MA		 2402 dsl_deadlist_space_range(&clone->ds_deadlist,
2403 origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
2404 &cdl_used, &cdl_comp, &cdl_uncomp);
2405 dsl_deadlist_space_range(&origin_head->ds_deadlist,
2406 origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
2407 &odl_used, &odl_comp, &odl_uncomp);
2408 dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used,
2409 DD_USED_HEAD, DD_USED_SNAP, tx);
45d1cae3 2410 }
45d1cae3 2411
13fe0198
MA		 2412 /* swap ds_*_bytes */
2413 SWITCH64(origin_head->ds_phys->ds_referenced_bytes,
2414 clone->ds_phys->ds_referenced_bytes);
2415 SWITCH64(origin_head->ds_phys->ds_compressed_bytes,
2416 clone->ds_phys->ds_compressed_bytes);
2417 SWITCH64(origin_head->ds_phys->ds_uncompressed_bytes,
2418 clone->ds_phys->ds_uncompressed_bytes);
2419 SWITCH64(origin_head->ds_phys->ds_unique_bytes,
2420 clone->ds_phys->ds_unique_bytes);
45d1cae3 2421
13fe0198
MA		 2422 /* apply any parent delta for change in unconsumed refreservation */
2423 dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV,
2424 unused_refres_delta, 0, 0, tx);
45d1cae3 2425
13fe0198
MA		 2426 /*
2427 * Swap deadlists.
2428 */
2429 dsl_deadlist_close(&clone->ds_deadlist);
2430 dsl_deadlist_close(&origin_head->ds_deadlist);
2431 SWITCH64(origin_head->ds_phys->ds_deadlist_obj,
2432 clone->ds_phys->ds_deadlist_obj);
2433 dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset,
2434 clone->ds_phys->ds_deadlist_obj);
2435 dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset,
2436 origin_head->ds_phys->ds_deadlist_obj);
45d1cae3 2437
13fe0198 2438 dsl_scan_ds_clone_swapped(origin_head, clone, tx);
45d1cae3 2439
13fe0198
MA		 2440 spa_history_log_internal_ds(clone, "clone swap", tx,
2441 "parent=%s", origin_head->ds_dir->dd_myname);
45d1cae3
BB		 2442}
2443
13fe0198
MA		 2444/*
2445 * Given a pool name and a dataset object number in that pool,
2446 * return the name of that dataset.
2447 */
572e2857 2448int
13fe0198 2449dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf)
572e2857 2450{
13fe0198
MA		 2451 dsl_pool_t *dp;
2452 dsl_dataset_t *ds;
572e2857
BB		 2453 int error;
2454
13fe0198
MA		 2455 error = dsl_pool_hold(pname, FTAG, &dp);
2456 if (error != 0)
2457 return (error);
2458
2459 error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
2460 if (error == 0) {
2461 dsl_dataset_name(ds, buf);
2462 dsl_dataset_rele(ds, FTAG);
2463 }
2464 dsl_pool_rele(dp, FTAG);
572e2857
BB		 2465
2466 return (error);
2467}
2468
45d1cae3 2469int
13fe0198
MA		 2470dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
2471 uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv)
45d1cae3 2472{
13fe0198 2473 int error = 0;
45d1cae3 2474
13fe0198 2475 ASSERT3S(asize, >, 0);
45d1cae3 2476
13fe0198
MA		 2477 /*
2478 * *ref_rsrv is the portion of asize that will come from any
2479 * unconsumed refreservation space.
2480 */
2481 *ref_rsrv = 0;
45d1cae3 2482
13fe0198
MA		 2483 mutex_enter(&ds->ds_lock);
2484 /*
2485 * Make a space adjustment for reserved bytes.
2486 */
2487 if (ds->ds_reserved > ds->ds_phys->ds_unique_bytes) {
2488 ASSERT3U(*used, >=,
2489 ds->ds_reserved - ds->ds_phys->ds_unique_bytes);
2490 *used -= (ds->ds_reserved - ds->ds_phys->ds_unique_bytes);
2491 *ref_rsrv =
2492 asize - MIN(asize, parent_delta(ds, asize + inflight));
45d1cae3
BB		 2493 }
2494
13fe0198
MA		 2495 if (!check_quota || ds->ds_quota == 0) {
2496 mutex_exit(&ds->ds_lock);
2497 return (0);
45d1cae3 2498 }
13fe0198
MA		 2499 /*
2500 * If they are requesting more space, and our current estimate
2501 * is over quota, they get to try again unless the actual
2502 * on-disk is over quota and there are no pending changes (which
2503 * may free up space for us).
2504 */
2505 if (ds->ds_phys->ds_referenced_bytes + inflight >= ds->ds_quota) {
2506 if (inflight > 0 ||
2507 ds->ds_phys->ds_referenced_bytes < ds->ds_quota)
2e528b49 2508 error = SET_ERROR(ERESTART);
13fe0198 2509 else
2e528b49 2510 error = SET_ERROR(EDQUOT);
45d1cae3 2511 }
13fe0198 2512 mutex_exit(&ds->ds_lock);
45d1cae3 2513
45d1cae3
BB		 2514 return (error);
2515}
2516
13fe0198
MA		 2517typedef struct dsl_dataset_set_qr_arg {
2518 const char *ddsqra_name;
2519 zprop_source_t ddsqra_source;
2520 uint64_t ddsqra_value;
2521} dsl_dataset_set_qr_arg_t;
45d1cae3 2522
13fe0198
MA		 2523
2524/* ARGSUSED */
45d1cae3 2525static int
13fe0198 2526dsl_dataset_set_refquota_check(void *arg, dmu_tx_t *tx)
45d1cae3 2527{
13fe0198
MA		 2528 dsl_dataset_set_qr_arg_t *ddsqra = arg;
2529 dsl_pool_t *dp = dmu_tx_pool(tx);
2530 dsl_dataset_t *ds;
45d1cae3 2531 int error;
13fe0198 2532 uint64_t newval;
45d1cae3 2533
13fe0198 2534 if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA)
2e528b49 2535 return (SET_ERROR(ENOTSUP));
45d1cae3 2536
13fe0198
MA		 2537 error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
2538 if (error != 0)
2539 return (error);
2540
2541 if (dsl_dataset_is_snapshot(ds)) {
2542 dsl_dataset_rele(ds, FTAG);
2e528b49 2543 return (SET_ERROR(EINVAL));
45d1cae3
BB		 2544 }
2545
13fe0198
MA		 2546 error = dsl_prop_predict(ds->ds_dir,
2547 zfs_prop_to_name(ZFS_PROP_REFQUOTA),
2548 ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
2549 if (error != 0) {
2550 dsl_dataset_rele(ds, FTAG);
45d1cae3
BB		 2551 return (error);
2552 }
2553
13fe0198
MA		 2554 if (newval == 0) {
2555 dsl_dataset_rele(ds, FTAG);
2556 return (0);
2557 }
2558
2559 if (newval < ds->ds_phys->ds_referenced_bytes ||
2560 newval < ds->ds_reserved) {
2561 dsl_dataset_rele(ds, FTAG);
2e528b49 2562 return (SET_ERROR(ENOSPC));
13fe0198 2563 }
45d1cae3 2564
13fe0198 2565 dsl_dataset_rele(ds, FTAG);
45d1cae3
BB		 2566 return (0);
2567}
2568
13fe0198
MA		 2569static void
2570dsl_dataset_set_refquota_sync(void *arg, dmu_tx_t *tx)
45d1cae3 2571{
13fe0198
MA		 2572 dsl_dataset_set_qr_arg_t *ddsqra = arg;
2573 dsl_pool_t *dp = dmu_tx_pool(tx);
2574 dsl_dataset_t *ds;
2575 uint64_t newval;
45d1cae3 2576
13fe0198 2577 VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));
45d1cae3 2578
13fe0198
MA		 2579 dsl_prop_set_sync_impl(ds,
2580 zfs_prop_to_name(ZFS_PROP_REFQUOTA),
2581 ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1,
2582 &ddsqra->ddsqra_value, tx);
45d1cae3 2583
13fe0198
MA		 2584 VERIFY0(dsl_prop_get_int_ds(ds,
2585 zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval));
45d1cae3 2586
13fe0198
MA		 2587 if (ds->ds_quota != newval) {
2588 dmu_buf_will_dirty(ds->ds_dbuf, tx);
2589 ds->ds_quota = newval;
45d1cae3 2590 }
13fe0198 2591 dsl_dataset_rele(ds, FTAG);
45d1cae3
BB		 2592}
2593
13fe0198
MA		 2594int
2595dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
2596 uint64_t refquota)
45d1cae3 2597{
13fe0198 2598 dsl_dataset_set_qr_arg_t ddsqra;
428870ff 2599
13fe0198
MA		 2600 ddsqra.ddsqra_name = dsname;
2601 ddsqra.ddsqra_source = source;
2602 ddsqra.ddsqra_value = refquota;
2603
2604 return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check,
2605 dsl_dataset_set_refquota_sync, &ddsqra, 0));
45d1cae3
BB		 2606}
2607
2608static int
13fe0198 2609dsl_dataset_set_refreservation_check(void *arg, dmu_tx_t *tx)
45d1cae3 2610{
13fe0198
MA		 2611 dsl_dataset_set_qr_arg_t *ddsqra = arg;
2612 dsl_pool_t *dp = dmu_tx_pool(tx);
45d1cae3
BB		 2613 dsl_dataset_t *ds;
2614 int error;
13fe0198 2615 uint64_t newval, unique;
428870ff 2616
13fe0198 2617 if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION)
2e528b49 2618 return (SET_ERROR(ENOTSUP));
45d1cae3 2619
13fe0198
MA		 2620 error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
2621 if (error != 0)
45d1cae3 2622 return (error);
45d1cae3 2623
13fe0198
MA		 2624 if (dsl_dataset_is_snapshot(ds)) {
2625 dsl_dataset_rele(ds, FTAG);
2e528b49 2626 return (SET_ERROR(EINVAL));
45d1cae3
BB		 2627 }
2628
13fe0198
MA		 2629 error = dsl_prop_predict(ds->ds_dir,
2630 zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
2631 ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
2632 if (error != 0) {
2633 dsl_dataset_rele(ds, FTAG);
45d1cae3
BB		 2634 return (error);
2635 }
2636
13fe0198
MA		 2637 /*
2638 * If we are doing the preliminary check in open context, the
2639 * space estimates may be inaccurate.
2640 */
2641 if (!dmu_tx_is_syncing(tx)) {
2642 dsl_dataset_rele(ds, FTAG);
2643 return (0);
45d1cae3 2644 }
45d1cae3 2645
13fe0198
MA		 2646 mutex_enter(&ds->ds_lock);
2647 if (!DS_UNIQUE_IS_ACCURATE(ds))
2648 dsl_dataset_recalc_head_uniq(ds);
2649 unique = ds->ds_phys->ds_unique_bytes;
2650 mutex_exit(&ds->ds_lock);
45d1cae3 2651
13fe0198
MA		 2652 if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) {
2653 uint64_t delta = MAX(unique, newval) -
2654 MAX(unique, ds->ds_reserved);
45d1cae3 2655
13fe0198
MA		 2656 if (delta >
2657 dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) ||
2658 (ds->ds_quota > 0 && newval > ds->ds_quota)) {
2659 dsl_dataset_rele(ds, FTAG);
2e528b49 2660 return (SET_ERROR(ENOSPC));
13fe0198 2661 }
45d1cae3
BB		 2662 }
2663
13fe0198
MA		 2664 dsl_dataset_rele(ds, FTAG);
2665 return (0);
45d1cae3
BB		 2666}
2667
13fe0198
MA		 2668void
2669dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
2670 zprop_source_t source, uint64_t value, dmu_tx_t *tx)
428870ff 2671{
13fe0198
MA		 2672 uint64_t newval;
2673 uint64_t unique;
2674 int64_t delta;
428870ff 2675
13fe0198
MA		 2676 dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
2677 source, sizeof (value), 1, &value, tx);
428870ff 2678
13fe0198
MA		 2679 VERIFY0(dsl_prop_get_int_ds(ds,
2680 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval));
572e2857 2681
13fe0198
MA		 2682 dmu_buf_will_dirty(ds->ds_dbuf, tx);
2683 mutex_enter(&ds->ds_dir->dd_lock);
2684 mutex_enter(&ds->ds_lock);
2685 ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
2686 unique = ds->ds_phys->ds_unique_bytes;
2687 delta = MAX(0, (int64_t)(newval - unique)) -
2688 MAX(0, (int64_t)(ds->ds_reserved - unique));
2689 ds->ds_reserved = newval;
2690 mutex_exit(&ds->ds_lock);
572e2857 2691
13fe0198
MA		 2692 dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx);
2693 mutex_exit(&ds->ds_dir->dd_lock);
428870ff
BB		 2694}
2695
13fe0198
MA		 2696static void
2697dsl_dataset_set_refreservation_sync(void *arg, dmu_tx_t *tx)
45d1cae3 2698{
13fe0198
MA		 2699 dsl_dataset_set_qr_arg_t *ddsqra = arg;
2700 dsl_pool_t *dp = dmu_tx_pool(tx);
45d1cae3 2701 dsl_dataset_t *ds;
45d1cae3 2702
13fe0198
MA		 2703 VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));
2704 dsl_dataset_set_refreservation_sync_impl(ds,
2705 ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx);
45d1cae3 2706 dsl_dataset_rele(ds, FTAG);
45d1cae3 2707}
428870ff 2708
428870ff 2709int
13fe0198
MA		 2710dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
2711 uint64_t refreservation)
428870ff 2712{
13fe0198 2713 dsl_dataset_set_qr_arg_t ddsqra;
428870ff 2714
13fe0198
MA		 2715 ddsqra.ddsqra_name = dsname;
2716 ddsqra.ddsqra_source = source;
2717 ddsqra.ddsqra_value = refreservation;
c28b2279 2718
13fe0198
MA		 2719 return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check,
2720 dsl_dataset_set_refreservation_sync, &ddsqra, 0));
2721}
330d06f9
MA		 2722
2723/*
2724 * Return (in *usedp) the amount of space written in new that is not
2725 * present in oldsnap. New may be a snapshot or the head. Old must be
2726 * a snapshot before new, in new's filesystem (or its origin). If not then
2727 * fail and return EINVAL.
2728 *
2729 * The written space is calculated by considering two components: First, we
2730 * ignore any freed space, and calculate the written as new's used space
2731 * minus old's used space. Next, we add in the amount of space that was freed
2732 * between the two snapshots, thus reducing new's used space relative to old's.
2733 * Specifically, this is the space that was born before old->ds_creation_txg,
2734 * and freed before new (ie. on new's deadlist or a previous deadlist).
2735 *
2736 * space freed [---------------------]
2737 * snapshots ---O-------O--------O-------O------
2738 * oldsnap new
2739 */
2740int
2741dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new,
2742 uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
2743{
2744 int err = 0;
2745 uint64_t snapobj;
2746 dsl_pool_t *dp = new->ds_dir->dd_pool;
2747
13fe0198
MA		 2748 ASSERT(dsl_pool_config_held(dp));
2749
330d06f9 2750 *usedp = 0;
9ae529ec
CS		 2751 *usedp += new->ds_phys->ds_referenced_bytes;
2752 *usedp -= oldsnap->ds_phys->ds_referenced_bytes;
330d06f9
MA		 2753
2754 *compp = 0;
2755 *compp += new->ds_phys->ds_compressed_bytes;
2756 *compp -= oldsnap->ds_phys->ds_compressed_bytes;
2757
2758 *uncompp = 0;
2759 *uncompp += new->ds_phys->ds_uncompressed_bytes;
2760 *uncompp -= oldsnap->ds_phys->ds_uncompressed_bytes;
2761
330d06f9
MA		 2762 snapobj = new->ds_object;
2763 while (snapobj != oldsnap->ds_object) {
2764 dsl_dataset_t *snap;
2765 uint64_t used, comp, uncomp;
2766
9ae529ec
CS		 2767 if (snapobj == new->ds_object) {
2768 snap = new;
2769 } else {
2770 err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap);
2771 if (err != 0)
2772 break;
2773 }
330d06f9
MA		 2774
2775 if (snap->ds_phys->ds_prev_snap_txg ==
2776 oldsnap->ds_phys->ds_creation_txg) {
2777 /*
2778 * The blocks in the deadlist can not be born after
2779 * ds_prev_snap_txg, so get the whole deadlist space,
2780 * which is more efficient (especially for old-format
2781 * deadlists). Unfortunately the deadlist code
2782 * doesn't have enough information to make this
2783 * optimization itself.
2784 */
2785 dsl_deadlist_space(&snap->ds_deadlist,
2786 &used, &comp, &uncomp);
2787 } else {
2788 dsl_deadlist_space_range(&snap->ds_deadlist,
2789 0, oldsnap->ds_phys->ds_creation_txg,
2790 &used, &comp, &uncomp);
2791 }
2792 *usedp += used;
2793 *compp += comp;
2794 *uncompp += uncomp;
2795
2796 /*
2797 * If we get to the beginning of the chain of snapshots
2798 * (ds_prev_snap_obj == 0) before oldsnap, then oldsnap
2799 * was not a snapshot of/before new.
2800 */
2801 snapobj = snap->ds_phys->ds_prev_snap_obj;
9ae529ec
CS		 2802 if (snap != new)
2803 dsl_dataset_rele(snap, FTAG);
330d06f9 2804 if (snapobj == 0) {
2e528b49 2805 err = SET_ERROR(EINVAL);
330d06f9
MA		 2806 break;
2807 }
2808
2809 }
330d06f9
MA		 2810 return (err);
2811}
2812
2813/*
2814 * Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
2815 * lastsnap, and all snapshots in between are deleted.
2816 *
2817 * blocks that would be freed [---------------------------]
2818 * snapshots ---O-------O--------O-------O--------O
2819 * firstsnap lastsnap
2820 *
2821 * This is the set of blocks that were born after the snap before firstsnap,
2822 * (birth > firstsnap->prev_snap_txg) and died before the snap after the
2823 * last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
2824 * We calculate this by iterating over the relevant deadlists (from the snap
2825 * after lastsnap, backward to the snap after firstsnap), summing up the
2826 * space on the deadlist that was born after the snap before firstsnap.
2827 */
2828int
2829dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap,
2830 dsl_dataset_t *lastsnap,
2831 uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
2832{
2833 int err = 0;
2834 uint64_t snapobj;
2835 dsl_pool_t *dp = firstsnap->ds_dir->dd_pool;
2836
2837 ASSERT(dsl_dataset_is_snapshot(firstsnap));
2838 ASSERT(dsl_dataset_is_snapshot(lastsnap));
2839
2840 /*
2841 * Check that the snapshots are in the same dsl_dir, and firstsnap
2842 * is before lastsnap.
2843 */
2844 if (firstsnap->ds_dir != lastsnap->ds_dir ||
2845 firstsnap->ds_phys->ds_creation_txg >
2846 lastsnap->ds_phys->ds_creation_txg)
2e528b49 2847 return (SET_ERROR(EINVAL));
330d06f9
MA		 2848
2849 *usedp = *compp = *uncompp = 0;
2850
330d06f9
MA		 2851 snapobj = lastsnap->ds_phys->ds_next_snap_obj;
2852 while (snapobj != firstsnap->ds_object) {
2853 dsl_dataset_t *ds;
2854 uint64_t used, comp, uncomp;
2855
2856 err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds);
2857 if (err != 0)
2858 break;
2859
2860 dsl_deadlist_space_range(&ds->ds_deadlist,
2861 firstsnap->ds_phys->ds_prev_snap_txg, UINT64_MAX,
2862 &used, &comp, &uncomp);
2863 *usedp += used;
2864 *compp += comp;
2865 *uncompp += uncomp;
2866
2867 snapobj = ds->ds_phys->ds_prev_snap_obj;
2868 ASSERT3U(snapobj, !=, 0);
2869 dsl_dataset_rele(ds, FTAG);
2870 }
330d06f9
MA		 2871 return (err);
2872}
2873
13fe0198
MA		 2874/*
2875 * Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
2876 * For example, they could both be snapshots of the same filesystem, and
2877 * 'earlier' is before 'later'. Or 'earlier' could be the origin of
2878 * 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
2879 * filesystem. Or 'earlier' could be the origin's origin.
2880 */
2881boolean_t
2882dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier)
2883{
2884 dsl_pool_t *dp = later->ds_dir->dd_pool;
2885 int error;
2886 boolean_t ret;
2887 dsl_dataset_t *origin;
2888
2889 ASSERT(dsl_pool_config_held(dp));
2890
2891 if (earlier->ds_phys->ds_creation_txg >=
2892 later->ds_phys->ds_creation_txg)
2893 return (B_FALSE);
2894
2895 if (later->ds_dir == earlier->ds_dir)
2896 return (B_TRUE);
2897 if (!dsl_dir_is_clone(later->ds_dir))
2898 return (B_FALSE);
2899
2900 if (later->ds_dir->dd_phys->dd_origin_obj == earlier->ds_object)
2901 return (B_TRUE);
2902 error = dsl_dataset_hold_obj(dp,
2903 later->ds_dir->dd_phys->dd_origin_obj, FTAG, &origin);
2904 if (error != 0)
2905 return (B_FALSE);
2906 ret = dsl_dataset_is_before(origin, earlier);
2907 dsl_dataset_rele(origin, FTAG);
2908 return (ret);
2909}
2910
c28b2279
BB		 2911#if defined(_KERNEL) && defined(HAVE_SPL)
2912EXPORT_SYMBOL(dsl_dataset_hold);
2913EXPORT_SYMBOL(dsl_dataset_hold_obj);
2914EXPORT_SYMBOL(dsl_dataset_own);
2915EXPORT_SYMBOL(dsl_dataset_own_obj);
2916EXPORT_SYMBOL(dsl_dataset_name);
2917EXPORT_SYMBOL(dsl_dataset_rele);
2918EXPORT_SYMBOL(dsl_dataset_disown);
c28b2279 2919EXPORT_SYMBOL(dsl_dataset_tryown);
c28b2279
BB		 2920EXPORT_SYMBOL(dsl_dataset_create_sync);
2921EXPORT_SYMBOL(dsl_dataset_create_sync_dd);
c28b2279
BB		 2922EXPORT_SYMBOL(dsl_dataset_snapshot_check);
2923EXPORT_SYMBOL(dsl_dataset_snapshot_sync);
c28b2279 2924EXPORT_SYMBOL(dsl_dataset_promote);
c28b2279
BB		 2925EXPORT_SYMBOL(dsl_dataset_user_hold);
2926EXPORT_SYMBOL(dsl_dataset_user_release);
c28b2279
BB		 2927EXPORT_SYMBOL(dsl_dataset_get_holds);
2928EXPORT_SYMBOL(dsl_dataset_get_blkptr);
2929EXPORT_SYMBOL(dsl_dataset_set_blkptr);
2930EXPORT_SYMBOL(dsl_dataset_get_spa);
2931EXPORT_SYMBOL(dsl_dataset_modified_since_lastsnap);
330d06f9
MA		 2932EXPORT_SYMBOL(dsl_dataset_space_written);
2933EXPORT_SYMBOL(dsl_dataset_space_wouldfree);
c28b2279
BB		 2934EXPORT_SYMBOL(dsl_dataset_sync);
2935EXPORT_SYMBOL(dsl_dataset_block_born);
2936EXPORT_SYMBOL(dsl_dataset_block_kill);
2937EXPORT_SYMBOL(dsl_dataset_block_freeable);
2938EXPORT_SYMBOL(dsl_dataset_prev_snap_txg);
2939EXPORT_SYMBOL(dsl_dataset_dirty);
2940EXPORT_SYMBOL(dsl_dataset_stats);
2941EXPORT_SYMBOL(dsl_dataset_fast_stat);
2942EXPORT_SYMBOL(dsl_dataset_space);
2943EXPORT_SYMBOL(dsl_dataset_fsid_guid);
2944EXPORT_SYMBOL(dsl_dsobj_to_dsname);
2945EXPORT_SYMBOL(dsl_dataset_check_quota);
13fe0198
MA		 2946EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl);
2947EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl);
c28b2279 2948#endif
ZFS On Linux mirror for PVE