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 /*
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, Joyent, Inc. All rights reserved.
25  * Copyright (c) 2013 Steven Hartland. All rights reserved.
26  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
27  * Copyright (c) 2014 Integros [integros.com]
28  */
29 
30 #ifndef	_SYS_DSL_DATASET_H
31 #define	_SYS_DSL_DATASET_H
32 
33 #include <sys/dmu.h>
34 #include <sys/spa.h>
35 #include <sys/txg.h>
36 #include <sys/zio.h>
37 #include <sys/bplist.h>
38 #include <sys/dsl_synctask.h>
39 #include <sys/zfs_context.h>
40 #include <sys/dsl_deadlist.h>
41 #include <sys/refcount.h>
42 #include <sys/rrwlock.h>
43 #include <zfeature_common.h>
44 
45 #ifdef	__cplusplus
46 extern "C" {
47 #endif
48 
49 struct dsl_dataset;
50 struct dsl_dir;
51 struct dsl_pool;
52 
53 #define	DS_FLAG_INCONSISTENT	(1ULL<<0)
54 #define	DS_IS_INCONSISTENT(ds)	\
55 	(dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT)
56 
57 /*
58  * Do not allow this dataset to be promoted.
59  */
60 #define	DS_FLAG_NOPROMOTE	(1ULL<<1)
61 
62 /*
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).
66  */
67 #define	DS_FLAG_UNIQUE_ACCURATE	(1ULL<<2)
68 
69 /*
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'.
72  */
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)
76 
77 /*
78  * DS_FIELD_* are strings that are used in the "extensified" dataset zap object.
79  * They should be of the format <reverse-dns>:<field>.
80  */
81 
82 /*
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.
86  */
87 #define	DS_FIELD_BOOKMARK_NAMES "com.delphix:bookmarks"
88 
89 /*
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.
93  */
94 #define	DS_FIELD_LARGE_DNODE "org.zfsonlinux:large_dnode"
95 
96 /*
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.
99  */
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 
110 /*
111  * This field is set to the object number of the remap deadlist if one exists.
112  */
113 #define	DS_FIELD_REMAP_DEADLIST	"com.delphix:remap_deadlist"
114 
115 /*
116  * DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
117  * name lookups should be performed case-insensitively.
118  */
119 #define	DS_FLAG_CI_DATASET	(1ULL<<16)
120 
121 #define	DS_CREATE_FLAG_NODIRTY	(1ULL<<24)
122 
123 typedef struct dsl_dataset_phys {
124 	uint64_t ds_dir_obj;		/* DMU_OT_DSL_DIR */
125 	uint64_t ds_prev_snap_obj;	/* DMU_OT_DSL_DATASET */
126 	uint64_t ds_prev_snap_txg;
127 	uint64_t ds_next_snap_obj;	/* DMU_OT_DSL_DATASET */
128 	uint64_t ds_snapnames_zapobj;	/* DMU_OT_DSL_DS_SNAP_MAP 0 for snaps */
129 	uint64_t ds_num_children;	/* clone/snap children; ==0 for head */
130 	uint64_t ds_creation_time;	/* seconds since 1970 */
131 	uint64_t ds_creation_txg;
132 	uint64_t ds_deadlist_obj;	/* DMU_OT_DEADLIST */
133 	/*
134 	 * ds_referenced_bytes, ds_compressed_bytes, and ds_uncompressed_bytes
135 	 * include all blocks referenced by this dataset, including those
136 	 * shared with any other datasets.
137 	 */
138 	uint64_t ds_referenced_bytes;
139 	uint64_t ds_compressed_bytes;
140 	uint64_t ds_uncompressed_bytes;
141 	uint64_t ds_unique_bytes;	/* only relevant to snapshots */
142 	/*
143 	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
144 	 * collisions.  The ds_guid is a 64-bit ID that will never
145 	 * change, so there is a small probability that it will collide.
146 	 */
147 	uint64_t ds_fsid_guid;
148 	uint64_t ds_guid;
149 	uint64_t ds_flags;		/* DS_FLAG_* */
150 	blkptr_t ds_bp;
151 	uint64_t ds_next_clones_obj;	/* DMU_OT_DSL_CLONES */
152 	uint64_t ds_props_obj;		/* DMU_OT_DSL_PROPS for snaps */
153 	uint64_t ds_userrefs_obj;	/* DMU_OT_USERREFS */
154 	uint64_t ds_pad[5]; /* pad out to 320 bytes for good measure */
155 } dsl_dataset_phys_t;
156 
157 typedef struct dsl_dataset {
158 	dmu_buf_user_t ds_dbu;
159 	rrwlock_t ds_bp_rwlock; /* Protects ds_phys->ds_bp */
160 
161 	/* Immutable: */
162 	struct dsl_dir *ds_dir;
163 	dmu_buf_t *ds_dbuf;
164 	uint64_t ds_object;
165 	uint64_t ds_fsid_guid;
166 	boolean_t ds_is_snapshot;
167 
168 	/* only used in syncing context, only valid for non-snapshots: */
169 	struct dsl_dataset *ds_prev;
170 	uint64_t ds_bookmarks;  /* DMU_OTN_ZAP_METADATA */
171 
172 	/* has internal locking: */
173 	dsl_deadlist_t ds_deadlist;
174 	bplist_t ds_pending_deadlist;
175 
176 	/*
177 	 * The remap deadlist contains blocks (DVA's, really) that are
178 	 * referenced by the previous snapshot and point to indirect vdevs,
179 	 * but in this dataset they have been remapped to point to concrete
180 	 * (or at least, less-indirect) vdevs.  In other words, the
181 	 * physical DVA is referenced by the previous snapshot but not by
182 	 * this dataset.  Logically, the DVA continues to be referenced,
183 	 * but we are using a different (less indirect) physical DVA.
184 	 * This deadlist is used to determine when physical DVAs that
185 	 * point to indirect vdevs are no longer referenced anywhere,
186 	 * and thus should be marked obsolete.
187 	 *
188 	 * This is only used if SPA_FEATURE_OBSOLETE_COUNTS is enabled.
189 	 */
190 	dsl_deadlist_t ds_remap_deadlist;
191 	/* protects creation of the ds_remap_deadlist */
192 	kmutex_t ds_remap_deadlist_lock;
193 
194 	/* protected by lock on pool's dp_dirty_datasets list */
195 	txg_node_t ds_dirty_link;
196 	list_node_t ds_synced_link;
197 
198 	/*
199 	 * ds_phys->ds_<accounting> is also protected by ds_lock.
200 	 * Protected by ds_lock:
201 	 */
202 	kmutex_t ds_lock;
203 	objset_t *ds_objset;
204 	uint64_t ds_userrefs;
205 	void *ds_owner;
206 
207 	/*
208 	 * Long holds prevent the ds from being destroyed; they allow the
209 	 * ds to remain held even after dropping the dp_config_rwlock.
210 	 * Owning counts as a long hold.  See the comments above
211 	 * dsl_pool_hold() for details.
212 	 */
213 	zfs_refcount_t ds_longholds;
214 
215 	/* no locking; only for making guesses */
216 	uint64_t ds_trysnap_txg;
217 
218 	/* for objset_open() */
219 	kmutex_t ds_opening_lock;
220 
221 	uint64_t ds_reserved;	/* cached refreservation */
222 	uint64_t ds_quota;	/* cached refquota */
223 
224 	kmutex_t ds_sendstream_lock;
225 	list_t ds_sendstreams;
226 
227 	/*
228 	 * When in the middle of a resumable receive, tracks how much
229 	 * progress we have made.
230 	 */
231 	uint64_t ds_resume_object[TXG_SIZE];
232 	uint64_t ds_resume_offset[TXG_SIZE];
233 	uint64_t ds_resume_bytes[TXG_SIZE];
234 
235 	/* Protected by our dsl_dir's dd_lock */
236 	list_t ds_prop_cbs;
237 
238 	/*
239 	 * For ZFEATURE_FLAG_PER_DATASET features, set if this dataset
240 	 * uses this feature.
241 	 */
242 	uint8_t ds_feature_inuse[SPA_FEATURES];
243 
244 	/*
245 	 * Set if we need to activate the feature on this dataset this txg
246 	 * (used only in syncing context).
247 	 */
248 	uint8_t ds_feature_activation_needed[SPA_FEATURES];
249 
250 	/* Protected by ds_lock; keep at end of struct for better locality */
251 	char ds_snapname[ZFS_MAX_DATASET_NAME_LEN];
252 } dsl_dataset_t;
253 
254 inline dsl_dataset_phys_t *
255 dsl_dataset_phys(dsl_dataset_t *ds)
256 {
257 	return (ds->ds_dbuf->db_data);
258 }
259 
260 typedef struct dsl_dataset_promote_arg {
261 	const char *ddpa_clonename;
262 	dsl_dataset_t *ddpa_clone;
263 	list_t shared_snaps, origin_snaps, clone_snaps;
264 	dsl_dataset_t *origin_origin; /* origin of the origin */
265 	uint64_t used, comp, uncomp, unique, cloneusedsnap, originusedsnap;
266 	nvlist_t *err_ds;
267 	cred_t *cr;
268 } dsl_dataset_promote_arg_t;
269 
270 typedef struct dsl_dataset_rollback_arg {
271 	const char *ddra_fsname;
272 	const char *ddra_tosnap;
273 	void *ddra_owner;
274 	nvlist_t *ddra_result;
275 } dsl_dataset_rollback_arg_t;
276 
277 typedef struct dsl_dataset_snapshot_arg {
278 	nvlist_t *ddsa_snaps;
279 	nvlist_t *ddsa_props;
280 	nvlist_t *ddsa_errors;
281 	cred_t *ddsa_cr;
282 } dsl_dataset_snapshot_arg_t;
283 
284 /*
285  * The max length of a temporary tag prefix is the number of hex digits
286  * required to express UINT64_MAX plus one for the hyphen.
287  */
288 #define	MAX_TAG_PREFIX_LEN	17
289 
290 #define	dsl_dataset_is_snapshot(ds) \
291 	(dsl_dataset_phys(ds)->ds_num_children != 0)
292 
293 #define	DS_UNIQUE_IS_ACCURATE(ds)	\
294 	((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_UNIQUE_ACCURATE) != 0)
295 
296 int dsl_dataset_hold(struct dsl_pool *dp, const char *name, void *tag,
297     dsl_dataset_t **dsp);
298 boolean_t dsl_dataset_try_add_ref(struct dsl_pool *dp, dsl_dataset_t *ds,
299     void *tag);
300 int dsl_dataset_hold_obj(struct dsl_pool *dp, uint64_t dsobj, void *tag,
301     dsl_dataset_t **);
302 void dsl_dataset_rele(dsl_dataset_t *ds, void *tag);
303 int dsl_dataset_own(struct dsl_pool *dp, const char *name,
304     void *tag, dsl_dataset_t **dsp);
305 int dsl_dataset_own_obj(struct dsl_pool *dp, uint64_t dsobj,
306     void *tag, dsl_dataset_t **dsp);
307 void dsl_dataset_disown(dsl_dataset_t *ds, void *tag);
308 void dsl_dataset_name(dsl_dataset_t *ds, char *name);
309 boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, void *tag);
310 int dsl_dataset_namelen(dsl_dataset_t *ds);
311 boolean_t dsl_dataset_has_owner(dsl_dataset_t *ds);
312 uint64_t dsl_dataset_create_sync(dsl_dir_t *pds, const char *lastname,
313     dsl_dataset_t *origin, uint64_t flags, cred_t *, dmu_tx_t *);
314 uint64_t dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
315     uint64_t flags, dmu_tx_t *tx);
316 void dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx);
317 int dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx);
318 int dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors);
319 void dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx);
320 int dsl_dataset_promote_check(void *arg, dmu_tx_t *tx);
321 int dsl_dataset_promote(const char *name, char *conflsnap);
322 int dsl_dataset_clone_swap(dsl_dataset_t *clone, dsl_dataset_t *origin_head,
323     boolean_t force);
324 int dsl_dataset_rename_snapshot(const char *fsname,
325     const char *oldsnapname, const char *newsnapname, boolean_t recursive);
326 int dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
327     minor_t cleanup_minor, const char *htag);
328 
329 blkptr_t *dsl_dataset_get_blkptr(dsl_dataset_t *ds);
330 
331 spa_t *dsl_dataset_get_spa(dsl_dataset_t *ds);
332 
333 boolean_t dsl_dataset_modified_since_snap(dsl_dataset_t *ds,
334     dsl_dataset_t *snap);
335 
336 void dsl_dataset_sync(dsl_dataset_t *os, zio_t *zio, dmu_tx_t *tx);
337 void dsl_dataset_sync_done(dsl_dataset_t *os, dmu_tx_t *tx);
338 
339 void dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp,
340     dmu_tx_t *tx);
341 int dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp,
342     dmu_tx_t *tx, boolean_t async);
343 void dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev,
344     uint64_t offset, uint64_t size, uint64_t birth, dmu_tx_t *tx);
345 
346 void dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx);
347 
348 int get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val);
349 char *get_receive_resume_stats_impl(dsl_dataset_t *ds);
350 char *get_child_receive_stats(dsl_dataset_t *ds);
351 uint64_t dsl_get_refratio(dsl_dataset_t *ds);
352 uint64_t dsl_get_logicalreferenced(dsl_dataset_t *ds);
353 uint64_t dsl_get_compressratio(dsl_dataset_t *ds);
354 uint64_t dsl_get_used(dsl_dataset_t *ds);
355 uint64_t dsl_get_creation(dsl_dataset_t *ds);
356 uint64_t dsl_get_creationtxg(dsl_dataset_t *ds);
357 uint64_t dsl_get_refquota(dsl_dataset_t *ds);
358 uint64_t dsl_get_refreservation(dsl_dataset_t *ds);
359 uint64_t dsl_get_guid(dsl_dataset_t *ds);
360 uint64_t dsl_get_unique(dsl_dataset_t *ds);
361 uint64_t dsl_get_objsetid(dsl_dataset_t *ds);
362 uint64_t dsl_get_userrefs(dsl_dataset_t *ds);
363 uint64_t dsl_get_defer_destroy(dsl_dataset_t *ds);
364 uint64_t dsl_get_referenced(dsl_dataset_t *ds);
365 uint64_t dsl_get_numclones(dsl_dataset_t *ds);
366 uint64_t dsl_get_inconsistent(dsl_dataset_t *ds);
367 uint64_t dsl_get_available(dsl_dataset_t *ds);
368 int dsl_get_written(dsl_dataset_t *ds, uint64_t *written);
369 int dsl_get_prev_snap(dsl_dataset_t *ds, char *snap);
370 int dsl_get_mountpoint(dsl_dataset_t *ds, const char *dsname, char *value,
371     char *source);
372 
373 void get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv);
374 
375 void dsl_dataset_stats(dsl_dataset_t *os, nvlist_t *nv);
376 
377 void dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat);
378 void dsl_dataset_space(dsl_dataset_t *ds,
379     uint64_t *refdbytesp, uint64_t *availbytesp,
380     uint64_t *usedobjsp, uint64_t *availobjsp);
381 uint64_t dsl_dataset_fsid_guid(dsl_dataset_t *ds);
382 int dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new,
383     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
384 int dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, dsl_dataset_t *last,
385     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
386 boolean_t dsl_dataset_is_dirty(dsl_dataset_t *ds);
387 
388 int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf);
389 
390 int dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
391     uint64_t asize, uint64_t inflight, uint64_t *used,
392     uint64_t *ref_rsrv);
393 int dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
394     uint64_t quota);
395 int dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
396     uint64_t reservation);
397 
398 boolean_t dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier,
399     uint64_t earlier_txg);
400 void dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag);
401 void dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag);
402 boolean_t dsl_dataset_long_held(dsl_dataset_t *ds);
403 
404 int dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone,
405     dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx);
406 void dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone,
407     dsl_dataset_t *origin_head, dmu_tx_t *tx);
408 int dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname,
409     dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr);
410 void dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname,
411     dmu_tx_t *tx);
412 
413 void dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj,
414     dmu_tx_t *tx);
415 void dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds);
416 int dsl_dataset_get_snapname(dsl_dataset_t *ds);
417 int dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name,
418     uint64_t *value);
419 int dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx,
420     boolean_t adj_cnt);
421 void dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
422     zprop_source_t source, uint64_t value, dmu_tx_t *tx);
423 void dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx);
424 boolean_t dsl_dataset_is_zapified(dsl_dataset_t *ds);
425 boolean_t dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds);
426 
427 int dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx);
428 void dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx);
429 int dsl_dataset_rollback(const char *fsname, const char *tosnap, void *owner,
430     nvlist_t *result);
431 
432 uint64_t dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds);
433 void dsl_dataset_create_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx);
434 boolean_t dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds);
435 void dsl_dataset_destroy_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx);
436 
437 void dsl_dataset_deactivate_feature(uint64_t dsobj,
438     spa_feature_t f, dmu_tx_t *tx);
439 
440 #ifdef ZFS_DEBUG
441 #define	dprintf_ds(ds, fmt, ...) do { \
442 	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
443 	char *__ds_name = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); \
444 	dsl_dataset_name(ds, __ds_name); \
445 	dprintf("ds=%s " fmt, __ds_name, __VA_ARGS__); \
446 	kmem_free(__ds_name, ZFS_MAX_DATASET_NAME_LEN); \
447 	} \
448 _NOTE(CONSTCOND) } while (0)
449 #else
450 #define	dprintf_ds(dd, fmt, ...)
451 #endif
452 
453 #ifdef	__cplusplus
454 }
455 #endif
456 
457 #endif /* _SYS_DSL_DATASET_H */
458