4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
24 * Copyright (c) 2013 Steven Hartland. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
28 #ifndef _SYS_DSL_DATASET_H
29 #define _SYS_DSL_DATASET_H
35 #include <sys/bplist.h>
36 #include <sys/dsl_synctask.h>
37 #include <sys/zfs_context.h>
38 #include <sys/dsl_deadlist.h>
39 #include <sys/refcount.h>
40 #include <sys/rrwlock.h>
41 #include <sys/dsl_crypt.h>
42 #include <zfeature_common.h>
51 struct dsl_crypto_params
;
53 #define DS_FLAG_INCONSISTENT (1ULL<<0)
54 #define DS_IS_INCONSISTENT(ds) \
55 (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT)
58 * Do not allow this dataset to be promoted.
60 #define DS_FLAG_NOPROMOTE (1ULL<<1)
63 * DS_FLAG_UNIQUE_ACCURATE is set if ds_unique_bytes has been correctly
64 * calculated for head datasets (starting with SPA_VERSION_UNIQUE_ACCURATE,
65 * refquota/refreservations).
67 #define DS_FLAG_UNIQUE_ACCURATE (1ULL<<2)
70 * DS_FLAG_DEFER_DESTROY is set after 'zfs destroy -d' has been called
71 * on a dataset. This allows the dataset to be destroyed using 'zfs release'.
73 #define DS_FLAG_DEFER_DESTROY (1ULL<<3)
74 #define DS_IS_DEFER_DESTROY(ds) \
75 (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_DEFER_DESTROY)
78 * DS_FIELD_* are strings that are used in the "extensified" dataset zap object.
79 * They should be of the format <reverse-dns>:<field>.
83 * This field's value is the object ID of a zap object which contains the
84 * bookmarks of this dataset. If it is present, then this dataset is counted
85 * in the refcount of the SPA_FEATURES_BOOKMARKS feature.
87 #define DS_FIELD_BOOKMARK_NAMES "com.delphix:bookmarks"
90 * This field is present (with value=0) if this dataset may contain large
91 * dnodes (>512B). If it is present, then this dataset is counted in the
92 * refcount of the SPA_FEATURE_LARGE_DNODE feature.
94 #define DS_FIELD_LARGE_DNODE "org.zfsonlinux:large_dnode"
97 * These fields are set on datasets that are in the middle of a resumable
98 * receive, and allow the sender to resume the send if it is interrupted.
100 #define DS_FIELD_RESUME_FROMGUID "com.delphix:resume_fromguid"
101 #define DS_FIELD_RESUME_TONAME "com.delphix:resume_toname"
102 #define DS_FIELD_RESUME_TOGUID "com.delphix:resume_toguid"
103 #define DS_FIELD_RESUME_OBJECT "com.delphix:resume_object"
104 #define DS_FIELD_RESUME_OFFSET "com.delphix:resume_offset"
105 #define DS_FIELD_RESUME_BYTES "com.delphix:resume_bytes"
106 #define DS_FIELD_RESUME_LARGEBLOCK "com.delphix:resume_largeblockok"
107 #define DS_FIELD_RESUME_EMBEDOK "com.delphix:resume_embedok"
108 #define DS_FIELD_RESUME_COMPRESSOK "com.delphix:resume_compressok"
109 #define DS_FIELD_RESUME_RAWOK "com.datto:resume_rawok"
112 * This field is set to the object number of the remap deadlist if one exists.
114 #define DS_FIELD_REMAP_DEADLIST "com.delphix:remap_deadlist"
117 * DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
118 * name lookups should be performed case-insensitively.
120 #define DS_FLAG_CI_DATASET (1ULL<<16)
122 #define DS_CREATE_FLAG_NODIRTY (1ULL<<24)
124 typedef struct dsl_dataset_phys
{
125 uint64_t ds_dir_obj
; /* DMU_OT_DSL_DIR */
126 uint64_t ds_prev_snap_obj
; /* DMU_OT_DSL_DATASET */
127 uint64_t ds_prev_snap_txg
;
128 uint64_t ds_next_snap_obj
; /* DMU_OT_DSL_DATASET */
129 uint64_t ds_snapnames_zapobj
; /* DMU_OT_DSL_DS_SNAP_MAP 0 for snaps */
130 uint64_t ds_num_children
; /* clone/snap children; ==0 for head */
131 uint64_t ds_creation_time
; /* seconds since 1970 */
132 uint64_t ds_creation_txg
;
133 uint64_t ds_deadlist_obj
; /* DMU_OT_DEADLIST */
135 * ds_referenced_bytes, ds_compressed_bytes, and ds_uncompressed_bytes
136 * include all blocks referenced by this dataset, including those
137 * shared with any other datasets.
139 uint64_t ds_referenced_bytes
;
140 uint64_t ds_compressed_bytes
;
141 uint64_t ds_uncompressed_bytes
;
142 uint64_t ds_unique_bytes
; /* only relevant to snapshots */
144 * The ds_fsid_guid is a 56-bit ID that can change to avoid
145 * collisions. The ds_guid is a 64-bit ID that will never
146 * change, so there is a small probability that it will collide.
148 uint64_t ds_fsid_guid
;
150 uint64_t ds_flags
; /* DS_FLAG_* */
152 uint64_t ds_next_clones_obj
; /* DMU_OT_DSL_CLONES */
153 uint64_t ds_props_obj
; /* DMU_OT_DSL_PROPS for snaps */
154 uint64_t ds_userrefs_obj
; /* DMU_OT_USERREFS */
155 uint64_t ds_pad
[5]; /* pad out to 320 bytes for good measure */
156 } dsl_dataset_phys_t
;
158 typedef struct dsl_dataset
{
159 dmu_buf_user_t ds_dbu
;
160 rrwlock_t ds_bp_rwlock
; /* Protects ds_phys->ds_bp */
163 struct dsl_dir
*ds_dir
;
166 uint64_t ds_fsid_guid
;
167 boolean_t ds_is_snapshot
;
169 /* only used in syncing context, only valid for non-snapshots: */
170 struct dsl_dataset
*ds_prev
;
171 uint64_t ds_bookmarks
; /* DMU_OTN_ZAP_METADATA */
173 /* has internal locking: */
174 dsl_deadlist_t ds_deadlist
;
175 bplist_t ds_pending_deadlist
;
178 * The remap deadlist contains blocks (DVA's, really) that are
179 * referenced by the previous snapshot and point to indirect vdevs,
180 * but in this dataset they have been remapped to point to concrete
181 * (or at least, less-indirect) vdevs. In other words, the
182 * physical DVA is referenced by the previous snapshot but not by
183 * this dataset. Logically, the DVA continues to be referenced,
184 * but we are using a different (less indirect) physical DVA.
185 * This deadlist is used to determine when physical DVAs that
186 * point to indirect vdevs are no longer referenced anywhere,
187 * and thus should be marked obsolete.
189 * This is only used if SPA_FEATURE_OBSOLETE_COUNTS is enabled.
191 dsl_deadlist_t ds_remap_deadlist
;
192 /* protects creation of the ds_remap_deadlist */
193 kmutex_t ds_remap_deadlist_lock
;
195 /* protected by lock on pool's dp_dirty_datasets list */
196 txg_node_t ds_dirty_link
;
197 list_node_t ds_synced_link
;
200 * ds_phys->ds_<accounting> is also protected by ds_lock.
201 * Protected by ds_lock:
205 uint64_t ds_userrefs
;
209 * Long holds prevent the ds from being destroyed; they allow the
210 * ds to remain held even after dropping the dp_config_rwlock.
211 * Owning counts as a long hold. See the comments above
212 * dsl_pool_hold() for details.
214 refcount_t ds_longholds
;
216 /* no locking; only for making guesses */
217 uint64_t ds_trysnap_txg
;
219 /* for objset_open() */
220 kmutex_t ds_opening_lock
;
222 uint64_t ds_reserved
; /* cached refreservation */
223 uint64_t ds_quota
; /* cached refquota */
225 kmutex_t ds_sendstream_lock
;
226 list_t ds_sendstreams
;
229 * When in the middle of a resumable receive, tracks how much
230 * progress we have made.
232 uint64_t ds_resume_object
[TXG_SIZE
];
233 uint64_t ds_resume_offset
[TXG_SIZE
];
234 uint64_t ds_resume_bytes
[TXG_SIZE
];
236 /* Protected by our dsl_dir's dd_lock */
240 * For ZFEATURE_FLAG_PER_DATASET features, set if this dataset
243 uint8_t ds_feature_inuse
[SPA_FEATURES
];
246 * Set if we need to activate the feature on this dataset this txg
247 * (used only in syncing context).
249 uint8_t ds_feature_activation_needed
[SPA_FEATURES
];
251 /* Protected by ds_lock; keep at end of struct for better locality */
252 char ds_snapname
[ZFS_MAX_DATASET_NAME_LEN
];
255 static inline dsl_dataset_phys_t
*
256 dsl_dataset_phys(dsl_dataset_t
*ds
)
258 return (ds
->ds_dbuf
->db_data
);
261 typedef struct dsl_dataset_promote_arg
{
262 const char *ddpa_clonename
;
263 dsl_dataset_t
*ddpa_clone
;
264 list_t shared_snaps
, origin_snaps
, clone_snaps
;
265 dsl_dataset_t
*origin_origin
; /* origin of the origin */
266 uint64_t used
, comp
, uncomp
, unique
, cloneusedsnap
, originusedsnap
;
269 } dsl_dataset_promote_arg_t
;
271 typedef struct dsl_dataset_rollback_arg
{
272 const char *ddra_fsname
;
273 const char *ddra_tosnap
;
275 nvlist_t
*ddra_result
;
276 } dsl_dataset_rollback_arg_t
;
278 typedef struct dsl_dataset_snapshot_arg
{
279 nvlist_t
*ddsa_snaps
;
280 nvlist_t
*ddsa_props
;
281 nvlist_t
*ddsa_errors
;
283 } dsl_dataset_snapshot_arg_t
;
286 * The max length of a temporary tag prefix is the number of hex digits
287 * required to express UINT64_MAX plus one for the hyphen.
289 #define MAX_TAG_PREFIX_LEN 17
291 #define dsl_dataset_is_snapshot(ds) \
292 (dsl_dataset_phys(ds)->ds_num_children != 0)
294 #define DS_UNIQUE_IS_ACCURATE(ds) \
295 ((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_UNIQUE_ACCURATE) != 0)
297 /* flags for holding the dataset */
298 typedef enum ds_hold_flags
{
299 DS_HOLD_FLAG_DECRYPT
= 1 << 0 /* needs access to encrypted data */
302 int dsl_dataset_hold(struct dsl_pool
*dp
, const char *name
, void *tag
,
303 dsl_dataset_t
**dsp
);
304 int dsl_dataset_hold_flags(struct dsl_pool
*dp
, const char *name
,
305 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
);
306 boolean_t
dsl_dataset_try_add_ref(struct dsl_pool
*dp
, dsl_dataset_t
*ds
,
308 int dsl_dataset_hold_obj(struct dsl_pool
*dp
, uint64_t dsobj
, void *tag
,
310 int dsl_dataset_hold_obj_flags(struct dsl_pool
*dp
, uint64_t dsobj
,
311 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**);
312 void dsl_dataset_rele(dsl_dataset_t
*ds
, void *tag
);
313 void dsl_dataset_rele_flags(dsl_dataset_t
*ds
, ds_hold_flags_t flags
,
315 int dsl_dataset_own(struct dsl_pool
*dp
, const char *name
,
316 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
);
317 int dsl_dataset_own_obj(struct dsl_pool
*dp
, uint64_t dsobj
,
318 ds_hold_flags_t flags
, void *tag
, dsl_dataset_t
**dsp
);
319 void dsl_dataset_disown(dsl_dataset_t
*ds
, ds_hold_flags_t flags
, void *tag
);
320 void dsl_dataset_name(dsl_dataset_t
*ds
, char *name
);
321 int dsl_dataset_namelen(dsl_dataset_t
*ds
);
322 boolean_t
dsl_dataset_has_owner(dsl_dataset_t
*ds
);
323 boolean_t
dsl_dataset_tryown(dsl_dataset_t
*ds
, void *tag
);
324 uint64_t dsl_dataset_create_sync(dsl_dir_t
*pds
, const char *lastname
,
325 dsl_dataset_t
*origin
, uint64_t flags
, cred_t
*,
326 struct dsl_crypto_params
*, dmu_tx_t
*);
327 uint64_t dsl_dataset_create_sync_dd(dsl_dir_t
*dd
, dsl_dataset_t
*origin
,
328 struct dsl_crypto_params
*dcp
, uint64_t flags
, dmu_tx_t
*tx
);
329 void dsl_dataset_snapshot_sync(void *arg
, dmu_tx_t
*tx
);
330 int dsl_dataset_snapshot_check(void *arg
, dmu_tx_t
*tx
);
331 int dsl_dataset_snapshot(nvlist_t
*snaps
, nvlist_t
*props
, nvlist_t
*errors
);
332 void dsl_dataset_promote_sync(void *arg
, dmu_tx_t
*tx
);
333 int dsl_dataset_promote_check(void *arg
, dmu_tx_t
*tx
);
334 int dsl_dataset_promote(const char *name
, char *conflsnap
);
335 int dsl_dataset_rename_snapshot(const char *fsname
,
336 const char *oldsnapname
, const char *newsnapname
, boolean_t recursive
);
337 int dsl_dataset_snapshot_tmp(const char *fsname
, const char *snapname
,
338 minor_t cleanup_minor
, const char *htag
);
340 blkptr_t
*dsl_dataset_get_blkptr(dsl_dataset_t
*ds
);
342 spa_t
*dsl_dataset_get_spa(dsl_dataset_t
*ds
);
344 boolean_t
dsl_dataset_modified_since_snap(dsl_dataset_t
*ds
,
345 dsl_dataset_t
*snap
);
347 void dsl_dataset_sync(dsl_dataset_t
*ds
, zio_t
*zio
, dmu_tx_t
*tx
);
348 void dsl_dataset_sync_done(dsl_dataset_t
*ds
, dmu_tx_t
*tx
);
350 void dsl_dataset_block_born(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
352 int dsl_dataset_block_kill(dsl_dataset_t
*ds
, const blkptr_t
*bp
,
353 dmu_tx_t
*tx
, boolean_t async
);
354 void dsl_dataset_block_remapped(dsl_dataset_t
*ds
, uint64_t vdev
,
355 uint64_t offset
, uint64_t size
, uint64_t birth
, dmu_tx_t
*tx
);
356 int dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
,
359 void dsl_dataset_dirty(dsl_dataset_t
*ds
, dmu_tx_t
*tx
);
361 int get_clones_stat_impl(dsl_dataset_t
*ds
, nvlist_t
*val
);
362 char *get_receive_resume_stats_impl(dsl_dataset_t
*ds
);
363 char *get_child_receive_stats(dsl_dataset_t
*ds
);
364 uint64_t dsl_get_refratio(dsl_dataset_t
*ds
);
365 uint64_t dsl_get_logicalreferenced(dsl_dataset_t
*ds
);
366 uint64_t dsl_get_compressratio(dsl_dataset_t
*ds
);
367 uint64_t dsl_get_used(dsl_dataset_t
*ds
);
368 uint64_t dsl_get_creation(dsl_dataset_t
*ds
);
369 uint64_t dsl_get_creationtxg(dsl_dataset_t
*ds
);
370 uint64_t dsl_get_refquota(dsl_dataset_t
*ds
);
371 uint64_t dsl_get_refreservation(dsl_dataset_t
*ds
);
372 uint64_t dsl_get_guid(dsl_dataset_t
*ds
);
373 uint64_t dsl_get_unique(dsl_dataset_t
*ds
);
374 uint64_t dsl_get_objsetid(dsl_dataset_t
*ds
);
375 uint64_t dsl_get_userrefs(dsl_dataset_t
*ds
);
376 uint64_t dsl_get_defer_destroy(dsl_dataset_t
*ds
);
377 uint64_t dsl_get_referenced(dsl_dataset_t
*ds
);
378 uint64_t dsl_get_numclones(dsl_dataset_t
*ds
);
379 uint64_t dsl_get_inconsistent(dsl_dataset_t
*ds
);
380 uint64_t dsl_get_available(dsl_dataset_t
*ds
);
381 int dsl_get_written(dsl_dataset_t
*ds
, uint64_t *written
);
382 int dsl_get_prev_snap(dsl_dataset_t
*ds
, char *snap
);
383 int dsl_get_mountpoint(dsl_dataset_t
*ds
, const char *dsname
, char *value
,
386 void get_clones_stat(dsl_dataset_t
*ds
, nvlist_t
*nv
);
387 void dsl_dataset_stats(dsl_dataset_t
*os
, nvlist_t
*nv
);
389 void dsl_dataset_fast_stat(dsl_dataset_t
*ds
, dmu_objset_stats_t
*stat
);
390 void dsl_dataset_space(dsl_dataset_t
*ds
,
391 uint64_t *refdbytesp
, uint64_t *availbytesp
,
392 uint64_t *usedobjsp
, uint64_t *availobjsp
);
393 uint64_t dsl_dataset_fsid_guid(dsl_dataset_t
*ds
);
394 int dsl_dataset_space_written(dsl_dataset_t
*oldsnap
, dsl_dataset_t
*new,
395 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
);
396 int dsl_dataset_space_wouldfree(dsl_dataset_t
*firstsnap
, dsl_dataset_t
*last
,
397 uint64_t *usedp
, uint64_t *compp
, uint64_t *uncompp
);
398 boolean_t
dsl_dataset_is_dirty(dsl_dataset_t
*ds
);
400 int dsl_dsobj_to_dsname(char *pname
, uint64_t obj
, char *buf
);
402 int dsl_dataset_check_quota(dsl_dataset_t
*ds
, boolean_t check_quota
,
403 uint64_t asize
, uint64_t inflight
, uint64_t *used
,
405 int dsl_dataset_set_refquota(const char *dsname
, zprop_source_t source
,
407 int dsl_dataset_set_refreservation(const char *dsname
, zprop_source_t source
,
408 uint64_t reservation
);
410 boolean_t
dsl_dataset_is_before(dsl_dataset_t
*later
, dsl_dataset_t
*earlier
,
411 uint64_t earlier_txg
);
412 void dsl_dataset_long_hold(dsl_dataset_t
*ds
, void *tag
);
413 void dsl_dataset_long_rele(dsl_dataset_t
*ds
, void *tag
);
414 boolean_t
dsl_dataset_long_held(dsl_dataset_t
*ds
);
416 int dsl_dataset_clone_swap_check_impl(dsl_dataset_t
*clone
,
417 dsl_dataset_t
*origin_head
, boolean_t force
, void *owner
, dmu_tx_t
*tx
);
418 void dsl_dataset_clone_swap_sync_impl(dsl_dataset_t
*clone
,
419 dsl_dataset_t
*origin_head
, dmu_tx_t
*tx
);
420 int dsl_dataset_snapshot_check_impl(dsl_dataset_t
*ds
, const char *snapname
,
421 dmu_tx_t
*tx
, boolean_t recv
, uint64_t cnt
, cred_t
*cr
);
422 void dsl_dataset_snapshot_sync_impl(dsl_dataset_t
*ds
, const char *snapname
,
425 void dsl_dataset_remove_from_next_clones(dsl_dataset_t
*ds
, uint64_t obj
,
427 void dsl_dataset_recalc_head_uniq(dsl_dataset_t
*ds
);
428 int dsl_dataset_get_snapname(dsl_dataset_t
*ds
);
429 int dsl_dataset_snap_lookup(dsl_dataset_t
*ds
, const char *name
,
431 int dsl_dataset_snap_remove(dsl_dataset_t
*ds
, const char *name
, dmu_tx_t
*tx
,
433 void dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t
*ds
,
434 zprop_source_t source
, uint64_t value
, dmu_tx_t
*tx
);
435 void dsl_dataset_zapify(dsl_dataset_t
*ds
, dmu_tx_t
*tx
);
436 boolean_t
dsl_dataset_is_zapified(dsl_dataset_t
*ds
);
437 boolean_t
dsl_dataset_has_resume_receive_state(dsl_dataset_t
*ds
);
439 int dsl_dataset_rollback_check(void *arg
, dmu_tx_t
*tx
);
440 void dsl_dataset_rollback_sync(void *arg
, dmu_tx_t
*tx
);
441 int dsl_dataset_rollback(const char *fsname
, const char *tosnap
, void *owner
,
444 uint64_t dsl_dataset_get_remap_deadlist_object(dsl_dataset_t
*ds
);
445 void dsl_dataset_create_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
);
446 boolean_t
dsl_dataset_remap_deadlist_exists(dsl_dataset_t
*ds
);
447 void dsl_dataset_destroy_remap_deadlist(dsl_dataset_t
*ds
, dmu_tx_t
*tx
);
449 void dsl_dataset_activate_feature(uint64_t dsobj
,
450 spa_feature_t f
, dmu_tx_t
*tx
);
451 void dsl_dataset_deactivate_feature(uint64_t dsobj
,
452 spa_feature_t f
, dmu_tx_t
*tx
);
455 #define dprintf_ds(ds, fmt, ...) do { \
456 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
457 char *__ds_name = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); \
458 dsl_dataset_name(ds, __ds_name); \
459 dprintf("ds=%s " fmt, __ds_name, __VA_ARGS__); \
460 kmem_free(__ds_name, ZFS_MAX_DATASET_NAME_LEN); \
462 _NOTE(CONSTCOND) } while (0)
464 #define dprintf_ds(dd, fmt, ...)
471 #endif /* _SYS_DSL_DATASET_H */