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) 2013 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
28 /* Portions Copyright 2010 Robert Milkowski */
34 * This file describes the interface that the DMU provides for its
37 * The DMU also interacts with the SPA. That interface is described in
41 #include <sys/inttypes.h>
42 #include <sys/types.h>
43 #include <sys/param.h>
46 #include <sys/fs/zfs.h>
72 typedef struct objset objset_t;
73 typedef struct dmu_tx dmu_tx_t;
74 typedef struct dsl_dir dsl_dir_t;
76 typedef enum dmu_object_byteswap {
88 * Allocating a new byteswap type number makes the on-disk format
89 * incompatible with any other format that uses the same number.
91 * Data can usually be structured to work with one of the
92 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
95 } dmu_object_byteswap_t;
97 #define DMU_OT_NEWTYPE 0x80
98 #define DMU_OT_METADATA 0x40
99 #define DMU_OT_BYTESWAP_MASK 0x3f
102 * Defines a uint8_t object type. Object types specify if the data
103 * in the object is metadata (boolean) and how to byteswap the data
104 * (dmu_object_byteswap_t).
106 #define DMU_OT(byteswap, metadata) \
108 ((metadata) ? DMU_OT_METADATA : 0) | \
109 ((byteswap) & DMU_OT_BYTESWAP_MASK))
111 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
112 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
113 (ot) < DMU_OT_NUMTYPES)
115 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
116 ((ot) & DMU_OT_METADATA) : \
117 dmu_ot[(int)(ot)].ot_metadata)
119 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
120 ((ot) & DMU_OT_BYTESWAP_MASK) : \
121 dmu_ot[(int)(ot)].ot_byteswap)
123 typedef enum dmu_object_type {
126 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
127 DMU_OT_OBJECT_ARRAY, /* UINT64 */
128 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
129 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
130 DMU_OT_BPOBJ, /* UINT64 */
131 DMU_OT_BPOBJ_HDR, /* UINT64 */
133 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
134 DMU_OT_SPACE_MAP, /* UINT64 */
136 DMU_OT_INTENT_LOG, /* UINT64 */
138 DMU_OT_DNODE, /* DNODE */
139 DMU_OT_OBJSET, /* OBJSET */
141 DMU_OT_DSL_DIR, /* UINT64 */
142 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
143 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
144 DMU_OT_DSL_PROPS, /* ZAP */
145 DMU_OT_DSL_DATASET, /* UINT64 */
147 DMU_OT_ZNODE, /* ZNODE */
148 DMU_OT_OLDACL, /* Old ACL */
149 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
150 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
151 DMU_OT_MASTER_NODE, /* ZAP */
152 DMU_OT_UNLINKED_SET, /* ZAP */
154 DMU_OT_ZVOL, /* UINT8 */
155 DMU_OT_ZVOL_PROP, /* ZAP */
156 /* other; for testing only! */
157 DMU_OT_PLAIN_OTHER, /* UINT8 */
158 DMU_OT_UINT64_OTHER, /* UINT64 */
159 DMU_OT_ZAP_OTHER, /* ZAP */
160 /* new object types: */
161 DMU_OT_ERROR_LOG, /* ZAP */
162 DMU_OT_SPA_HISTORY, /* UINT8 */
163 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
164 DMU_OT_POOL_PROPS, /* ZAP */
165 DMU_OT_DSL_PERMS, /* ZAP */
166 DMU_OT_ACL, /* ACL */
167 DMU_OT_SYSACL, /* SYSACL */
168 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
169 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
170 DMU_OT_NEXT_CLONES, /* ZAP */
171 DMU_OT_SCAN_QUEUE, /* ZAP */
172 DMU_OT_USERGROUP_USED, /* ZAP */
173 DMU_OT_USERGROUP_QUOTA, /* ZAP */
174 DMU_OT_USERREFS, /* ZAP */
175 DMU_OT_DDT_ZAP, /* ZAP */
176 DMU_OT_DDT_STATS, /* ZAP */
177 DMU_OT_SA, /* System attr */
178 DMU_OT_SA_MASTER_NODE, /* ZAP */
179 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
180 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
181 DMU_OT_SCAN_XLATE, /* ZAP */
182 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
183 DMU_OT_DEADLIST, /* ZAP */
184 DMU_OT_DEADLIST_HDR, /* UINT64 */
185 DMU_OT_DSL_CLONES, /* ZAP */
186 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
188 * Do not allocate new object types here. Doing so makes the on-disk
189 * format incompatible with any other format that uses the same object
192 * When creating an object which does not have one of the above types
193 * use the DMU_OTN_* type with the correct byteswap and metadata
196 * The DMU_OTN_* types do not have entries in the dmu_ot table,
197 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
198 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
199 * and DMU_OTN_* types).
204 * Names for valid types declared with DMU_OT().
206 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
207 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
208 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
209 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
210 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
211 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
212 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
213 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
214 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
215 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
218 typedef enum txg_how {
223 void byteswap_uint64_array(void *buf, size_t size);
224 void byteswap_uint32_array(void *buf, size_t size);
225 void byteswap_uint16_array(void *buf, size_t size);
226 void byteswap_uint8_array(void *buf, size_t size);
227 void zap_byteswap(void *buf, size_t size);
228 void zfs_oldacl_byteswap(void *buf, size_t size);
229 void zfs_acl_byteswap(void *buf, size_t size);
230 void zfs_znode_byteswap(void *buf, size_t size);
232 #define DS_FIND_SNAPSHOTS (1<<0)
233 #define DS_FIND_CHILDREN (1<<1)
236 * The maximum number of bytes that can be accessed as part of one
237 * operation, including metadata.
239 #define DMU_MAX_ACCESS (10<<20) /* 10MB */
240 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
242 #define DMU_USERUSED_OBJECT (-1ULL)
243 #define DMU_GROUPUSED_OBJECT (-2ULL)
244 #define DMU_DEADLIST_OBJECT (-3ULL)
247 * artificial blkids for bonus buffer and spill blocks
249 #define DMU_BONUS_BLKID (-1ULL)
250 #define DMU_SPILL_BLKID (-2ULL)
252 * Public routines to create, destroy, open, and close objsets.
254 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
255 int dmu_objset_own(const char *name, dmu_objset_type_t type,
256 boolean_t readonly, void *tag, objset_t **osp);
257 void dmu_objset_rele(objset_t *os, void *tag);
258 void dmu_objset_disown(objset_t *os, void *tag);
259 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
261 void dmu_objset_evict_dbufs(objset_t *os);
262 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
263 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
264 int dmu_objset_clone(const char *name, const char *origin);
265 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
266 struct nvlist *errlist);
267 int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
268 int dmu_objset_snapshot_tmp(const char *, const char *, int);
269 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
271 void dmu_objset_byteswap(void *buf, size_t size);
272 int dsl_dataset_rename_snapshot(const char *fsname,
273 const char *oldsnapname, const char *newsnapname, boolean_t recursive);
275 typedef struct dmu_buf {
276 uint64_t db_object; /* object that this buffer is part of */
277 uint64_t db_offset; /* byte offset in this object */
278 uint64_t db_size; /* size of buffer in bytes */
279 void *db_data; /* data in buffer */
282 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
285 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
287 #define DMU_POOL_DIRECTORY_OBJECT 1
288 #define DMU_POOL_CONFIG "config"
289 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
290 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
291 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
292 #define DMU_POOL_ROOT_DATASET "root_dataset"
293 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
294 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
295 #define DMU_POOL_ERRLOG_LAST "errlog_last"
296 #define DMU_POOL_SPARES "spares"
297 #define DMU_POOL_DEFLATE "deflate"
298 #define DMU_POOL_HISTORY "history"
299 #define DMU_POOL_PROPS "pool_props"
300 #define DMU_POOL_L2CACHE "l2cache"
301 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
302 #define DMU_POOL_DDT "DDT-%s-%s-%s"
303 #define DMU_POOL_DDT_STATS "DDT-statistics"
304 #define DMU_POOL_CREATION_VERSION "creation_version"
305 #define DMU_POOL_SCAN "scan"
306 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
307 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
308 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
311 * Allocate an object from this objset. The range of object numbers
312 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
314 * The transaction must be assigned to a txg. The newly allocated
315 * object will be "held" in the transaction (ie. you can modify the
316 * newly allocated object in this transaction).
318 * dmu_object_alloc() chooses an object and returns it in *objectp.
320 * dmu_object_claim() allocates a specific object number. If that
321 * number is already allocated, it fails and returns EEXIST.
323 * Return 0 on success, or ENOSPC or EEXIST as specified above.
325 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
326 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
327 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
328 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
329 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
330 int blocksize, dmu_object_type_t bonustype, int bonuslen);
333 * Free an object from this objset.
335 * The object's data will be freed as well (ie. you don't need to call
336 * dmu_free(object, 0, -1, tx)).
338 * The object need not be held in the transaction.
340 * If there are any holds on this object's buffers (via dmu_buf_hold()),
341 * or tx holds on the object (via dmu_tx_hold_object()), you can not
342 * free it; it fails and returns EBUSY.
344 * If the object is not allocated, it fails and returns ENOENT.
346 * Return 0 on success, or EBUSY or ENOENT as specified above.
348 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
351 * Find the next allocated or free object.
353 * The objectp parameter is in-out. It will be updated to be the next
354 * object which is allocated. Ignore objects which have not been
355 * modified since txg.
357 * XXX Can only be called on a objset with no dirty data.
359 * Returns 0 on success, or ENOENT if there are no more objects.
361 int dmu_object_next(objset_t *os, uint64_t *objectp,
362 boolean_t hole, uint64_t txg);
365 * Set the data blocksize for an object.
367 * The object cannot have any blocks allcated beyond the first. If
368 * the first block is allocated already, the new size must be greater
369 * than the current block size. If these conditions are not met,
370 * ENOTSUP will be returned.
372 * Returns 0 on success, or EBUSY if there are any holds on the object
373 * contents, or ENOTSUP as described above.
375 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
376 int ibs, dmu_tx_t *tx);
379 * Set the checksum property on a dnode. The new checksum algorithm will
380 * apply to all newly written blocks; existing blocks will not be affected.
382 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
386 * Set the compress property on a dnode. The new compression algorithm will
387 * apply to all newly written blocks; existing blocks will not be affected.
389 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
393 * Decide how to write a block: checksum, compression, number of copies, etc.
395 #define WP_NOFILL 0x1
396 #define WP_DMU_SYNC 0x2
399 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
400 struct zio_prop *zp);
402 * The bonus data is accessed more or less like a regular buffer.
403 * You must dmu_bonus_hold() to get the buffer, which will give you a
404 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
405 * data. As with any normal buffer, you must call dmu_buf_read() to
406 * read db_data, dmu_buf_will_dirty() before modifying it, and the
407 * object must be held in an assigned transaction before calling
408 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
409 * buffer as well. You must release what you hold with dmu_buf_rele().
411 * Returns ENOENT, EIO, or 0.
413 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
414 int dmu_bonus_max(void);
415 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
416 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
417 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
418 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
421 * Special spill buffer support used by "SA" framework
424 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
425 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
426 void *tag, dmu_buf_t **dbp);
427 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
430 * Obtain the DMU buffer from the specified object which contains the
431 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
432 * that it will remain in memory. You must release the hold with
433 * dmu_buf_rele(). You must not access the dmu_buf_t after releasing
434 * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
436 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
437 * on the returned buffer before reading or writing the buffer's
438 * db_data. The comments for those routines describe what particular
439 * operations are valid after calling them.
441 * The object number must be a valid, allocated object number.
443 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
444 void *tag, dmu_buf_t **, int flags);
445 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
446 void dmu_buf_rele(dmu_buf_t *db, void *tag);
447 uint64_t dmu_buf_refcount(dmu_buf_t *db);
450 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
451 * range of an object. A pointer to an array of dmu_buf_t*'s is
452 * returned (in *dbpp).
454 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
455 * frees the array. The hold on the array of buffers MUST be released
456 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
457 * individually with dmu_buf_rele.
459 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
460 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
461 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
464 * Returns NULL on success, or the existing user ptr if it's already
467 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
469 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
470 * will be set to db->db_data when you are allowed to access it. Note
471 * that db->db_data (the pointer) can change when you do dmu_buf_read(),
472 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
473 * *user_data_ptr_ptr will be set to the new value when it changes.
475 * If non-NULL, pageout func will be called when this buffer is being
476 * excised from the cache, so that you can clean up the data structure
477 * pointed to by user_ptr.
479 * dmu_evict_user() will call the pageout func for all buffers in a
480 * objset with a given pageout func.
482 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
483 dmu_buf_evict_func_t *pageout_func);
485 * set_user_ie is the same as set_user, but request immediate eviction
486 * when hold count goes to zero.
488 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
489 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
490 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
491 void *user_ptr, void *user_data_ptr_ptr,
492 dmu_buf_evict_func_t *pageout_func);
493 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
496 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
498 void *dmu_buf_get_user(dmu_buf_t *db);
501 * Returns the blkptr associated with this dbuf, or NULL if not set.
503 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
506 * Indicate that you are going to modify the buffer's data (db_data).
508 * The transaction (tx) must be assigned to a txg (ie. you've called
509 * dmu_tx_assign()). The buffer's object must be held in the tx
510 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
512 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
515 * Tells if the given dbuf is freeable.
517 boolean_t dmu_buf_freeable(dmu_buf_t *);
520 * You must create a transaction, then hold the objects which you will
521 * (or might) modify as part of this transaction. Then you must assign
522 * the transaction to a transaction group. Once the transaction has
523 * been assigned, you can modify buffers which belong to held objects as
524 * part of this transaction. You can't modify buffers before the
525 * transaction has been assigned; you can't modify buffers which don't
526 * belong to objects which this transaction holds; you can't hold
527 * objects once the transaction has been assigned. You may hold an
528 * object which you are going to free (with dmu_object_free()), but you
531 * You can abort the transaction before it has been assigned.
533 * Note that you may hold buffers (with dmu_buf_hold) at any time,
534 * regardless of transaction state.
537 #define DMU_NEW_OBJECT (-1ULL)
538 #define DMU_OBJECT_END (-1ULL)
540 dmu_tx_t *dmu_tx_create(objset_t *os);
541 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
542 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
544 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
545 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
546 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
547 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
548 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
549 void dmu_tx_abort(dmu_tx_t *tx);
550 int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
551 void dmu_tx_wait(dmu_tx_t *tx);
552 void dmu_tx_commit(dmu_tx_t *tx);
555 * To register a commit callback, dmu_tx_callback_register() must be called.
557 * dcb_data is a pointer to caller private data that is passed on as a
558 * callback parameter. The caller is responsible for properly allocating and
561 * When registering a callback, the transaction must be already created, but
562 * it cannot be committed or aborted. It can be assigned to a txg or not.
564 * The callback will be called after the transaction has been safely written
565 * to stable storage and will also be called if the dmu_tx is aborted.
566 * If there is any error which prevents the transaction from being committed to
567 * disk, the callback will be called with a value of error != 0.
569 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
571 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
575 * Free up the data blocks for a defined range of a file. If size is
576 * -1, the range from offset to end-of-file is freed.
578 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
579 uint64_t size, dmu_tx_t *tx);
580 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
582 int dmu_free_long_object(objset_t *os, uint64_t object);
585 * Convenience functions.
587 * Canfail routines will return 0 on success, or an errno if there is a
588 * nonrecoverable I/O error.
590 #define DMU_READ_PREFETCH 0 /* prefetch */
591 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
592 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
593 void *buf, uint32_t flags);
594 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
595 const void *buf, dmu_tx_t *tx);
596 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
599 #include <linux/blkdev_compat.h>
600 int dmu_read_req(objset_t *os, uint64_t object, struct request *req);
601 int dmu_write_req(objset_t *os, uint64_t object, struct request *req,
603 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
604 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
606 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
609 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
610 void dmu_return_arcbuf(struct arc_buf *buf);
611 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
613 int dmu_xuio_init(struct xuio *uio, int niov);
614 void dmu_xuio_fini(struct xuio *uio);
615 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
617 int dmu_xuio_cnt(struct xuio *uio);
618 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
619 void dmu_xuio_clear(struct xuio *uio, int i);
620 void xuio_stat_wbuf_copied(void);
621 void xuio_stat_wbuf_nocopy(void);
623 extern int zfs_prefetch_disable;
626 * Asynchronously try to read in the data.
628 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
631 typedef struct dmu_object_info {
632 /* All sizes are in bytes unless otherwise indicated. */
633 uint32_t doi_data_block_size;
634 uint32_t doi_metadata_block_size;
635 dmu_object_type_t doi_type;
636 dmu_object_type_t doi_bonus_type;
637 uint64_t doi_bonus_size;
638 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
639 uint8_t doi_checksum;
640 uint8_t doi_compress;
642 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
643 uint64_t doi_max_offset;
644 uint64_t doi_fill_count; /* number of non-empty blocks */
647 typedef void (*const arc_byteswap_func_t)(void *buf, size_t size);
649 typedef struct dmu_object_type_info {
650 dmu_object_byteswap_t ot_byteswap;
651 boolean_t ot_metadata;
653 } dmu_object_type_info_t;
655 typedef const struct dmu_object_byteswap_info {
656 arc_byteswap_func_t ob_func;
658 } dmu_object_byteswap_info_t;
660 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
661 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
664 * Get information on a DMU object.
666 * Return 0 on success or ENOENT if object is not allocated.
668 * If doi is NULL, just indicates whether the object exists.
670 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
671 void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
672 /* Like dmu_object_info, but faster if you have a held dnode in hand. */
673 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
674 /* Like dmu_object_info, but faster if you have a held dbuf in hand. */
675 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
677 * Like dmu_object_info_from_db, but faster still when you only care about
678 * the size. This is specifically optimized for zfs_getattr().
680 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
681 u_longlong_t *nblk512);
683 typedef struct dmu_objset_stats {
684 uint64_t dds_num_clones; /* number of clones of this */
685 uint64_t dds_creation_txg;
687 dmu_objset_type_t dds_type;
688 uint8_t dds_is_snapshot;
689 uint8_t dds_inconsistent;
690 char dds_origin[MAXNAMELEN];
691 } dmu_objset_stats_t;
694 * Get stats on a dataset.
696 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
699 * Add entries to the nvlist for all the objset's properties. See
700 * zfs_prop_table[] and zfs(1m) for details on the properties.
702 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
705 * Get the space usage statistics for statvfs().
707 * refdbytes is the amount of space "referenced" by this objset.
708 * availbytes is the amount of space available to this objset, taking
709 * into account quotas & reservations, assuming that no other objsets
710 * use the space first. These values correspond to the 'referenced' and
711 * 'available' properties, described in the zfs(1m) manpage.
713 * usedobjs and availobjs are the number of objects currently allocated,
716 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
717 uint64_t *usedobjsp, uint64_t *availobjsp);
720 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
721 * (Contrast with the ds_guid which is a 64-bit ID that will never
722 * change, so there is a small probability that it will collide.)
724 uint64_t dmu_objset_fsid_guid(objset_t *os);
727 * Get the [cm]time for an objset's snapshot dir
729 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
731 int dmu_objset_is_snapshot(objset_t *os);
733 extern struct spa *dmu_objset_spa(objset_t *os);
734 extern struct zilog *dmu_objset_zil(objset_t *os);
735 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
736 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
737 extern void dmu_objset_name(objset_t *os, char *buf);
738 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
739 extern uint64_t dmu_objset_id(objset_t *os);
740 extern uint64_t dmu_objset_syncprop(objset_t *os);
741 extern uint64_t dmu_objset_logbias(objset_t *os);
742 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
743 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
744 extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val);
745 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
746 int maxlen, boolean_t *conflict);
747 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
748 uint64_t *idp, uint64_t *offp);
750 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
751 void *bonus, uint64_t *userp, uint64_t *groupp);
752 extern void dmu_objset_register_type(dmu_objset_type_t ost,
753 objset_used_cb_t *cb);
754 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
755 extern void *dmu_objset_get_user(objset_t *os);
758 * Return the txg number for the given assigned transaction.
760 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
764 * If a parent zio is provided this function initiates a write on the
765 * provided buffer as a child of the parent zio.
766 * In the absence of a parent zio, the write is completed synchronously.
767 * At write completion, blk is filled with the bp of the written block.
768 * Note that while the data covered by this function will be on stable
769 * storage when the write completes this new data does not become a
770 * permanent part of the file until the associated transaction commits.
774 * {zfs,zvol,ztest}_get_done() args
777 struct zilog *zgd_zilog;
778 struct blkptr *zgd_bp;
784 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
785 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
788 * Find the next hole or data block in file starting at *off
789 * Return found offset in *off. Return ESRCH for end of file.
791 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
795 * Initial setup and final teardown.
797 extern void dmu_init(void);
798 extern void dmu_fini(void);
800 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
801 uint64_t object, uint64_t offset, int len);
802 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
803 dmu_traverse_cb_t cb, void *arg);
805 int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
806 struct vnode *vp, offset_t *offp);
809 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
810 extern uint64_t zfs_crc64_table[256];
812 extern int zfs_mdcomp_disable;
818 #endif /* _SYS_DMU_H */