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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 * Copyright 2013 DEY Storage Systems, Inc.
28 * Copyright 2014 HybridCluster. All rights reserved.
31 /* Portions Copyright 2010 Robert Milkowski */
37 * This file describes the interface that the DMU provides for its
40 * The DMU also interacts with the SPA. That interface is described in
44 #include <sys/types.h>
45 #include <sys/param.h>
48 #include <sys/fs/zfs.h>
68 struct zbookmark_phys;
76 typedef struct objset objset_t;
77 typedef struct dmu_tx dmu_tx_t;
78 typedef struct dsl_dir dsl_dir_t;
80 typedef enum dmu_object_byteswap {
92 * Allocating a new byteswap type number makes the on-disk format
93 * incompatible with any other format that uses the same number.
95 * Data can usually be structured to work with one of the
96 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
99 } dmu_object_byteswap_t;
101 #define DMU_OT_NEWTYPE 0x80
102 #define DMU_OT_METADATA 0x40
103 #define DMU_OT_BYTESWAP_MASK 0x3f
106 * Defines a uint8_t object type. Object types specify if the data
107 * in the object is metadata (boolean) and how to byteswap the data
108 * (dmu_object_byteswap_t).
110 #define DMU_OT(byteswap, metadata) \
112 ((metadata) ? DMU_OT_METADATA : 0) | \
113 ((byteswap) & DMU_OT_BYTESWAP_MASK))
115 #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
116 ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
117 (ot) < DMU_OT_NUMTYPES)
119 #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
120 ((ot) & DMU_OT_METADATA) : \
121 dmu_ot[(ot)].ot_metadata)
124 * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
125 * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
126 * is repurposed for embedded BPs.
128 #define DMU_OT_HAS_FILL(ot) \
129 ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
131 #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
132 ((ot) & DMU_OT_BYTESWAP_MASK) : \
133 dmu_ot[(ot)].ot_byteswap)
135 typedef enum dmu_object_type {
138 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
139 DMU_OT_OBJECT_ARRAY, /* UINT64 */
140 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
141 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
142 DMU_OT_BPOBJ, /* UINT64 */
143 DMU_OT_BPOBJ_HDR, /* UINT64 */
145 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
146 DMU_OT_SPACE_MAP, /* UINT64 */
148 DMU_OT_INTENT_LOG, /* UINT64 */
150 DMU_OT_DNODE, /* DNODE */
151 DMU_OT_OBJSET, /* OBJSET */
153 DMU_OT_DSL_DIR, /* UINT64 */
154 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
155 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
156 DMU_OT_DSL_PROPS, /* ZAP */
157 DMU_OT_DSL_DATASET, /* UINT64 */
159 DMU_OT_ZNODE, /* ZNODE */
160 DMU_OT_OLDACL, /* Old ACL */
161 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
162 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
163 DMU_OT_MASTER_NODE, /* ZAP */
164 DMU_OT_UNLINKED_SET, /* ZAP */
166 DMU_OT_ZVOL, /* UINT8 */
167 DMU_OT_ZVOL_PROP, /* ZAP */
168 /* other; for testing only! */
169 DMU_OT_PLAIN_OTHER, /* UINT8 */
170 DMU_OT_UINT64_OTHER, /* UINT64 */
171 DMU_OT_ZAP_OTHER, /* ZAP */
172 /* new object types: */
173 DMU_OT_ERROR_LOG, /* ZAP */
174 DMU_OT_SPA_HISTORY, /* UINT8 */
175 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
176 DMU_OT_POOL_PROPS, /* ZAP */
177 DMU_OT_DSL_PERMS, /* ZAP */
178 DMU_OT_ACL, /* ACL */
179 DMU_OT_SYSACL, /* SYSACL */
180 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
181 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
182 DMU_OT_NEXT_CLONES, /* ZAP */
183 DMU_OT_SCAN_QUEUE, /* ZAP */
184 DMU_OT_USERGROUP_USED, /* ZAP */
185 DMU_OT_USERGROUP_QUOTA, /* ZAP */
186 DMU_OT_USERREFS, /* ZAP */
187 DMU_OT_DDT_ZAP, /* ZAP */
188 DMU_OT_DDT_STATS, /* ZAP */
189 DMU_OT_SA, /* System attr */
190 DMU_OT_SA_MASTER_NODE, /* ZAP */
191 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
192 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
193 DMU_OT_SCAN_XLATE, /* ZAP */
194 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
195 DMU_OT_DEADLIST, /* ZAP */
196 DMU_OT_DEADLIST_HDR, /* UINT64 */
197 DMU_OT_DSL_CLONES, /* ZAP */
198 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
200 * Do not allocate new object types here. Doing so makes the on-disk
201 * format incompatible with any other format that uses the same object
204 * When creating an object which does not have one of the above types
205 * use the DMU_OTN_* type with the correct byteswap and metadata
208 * The DMU_OTN_* types do not have entries in the dmu_ot table,
209 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
210 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
211 * and DMU_OTN_* types).
216 * Names for valid types declared with DMU_OT().
218 DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
219 DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
220 DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
221 DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
222 DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
223 DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
224 DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
225 DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
226 DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
227 DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
230 typedef enum txg_how {
236 void byteswap_uint64_array(void *buf, size_t size);
237 void byteswap_uint32_array(void *buf, size_t size);
238 void byteswap_uint16_array(void *buf, size_t size);
239 void byteswap_uint8_array(void *buf, size_t size);
240 void zap_byteswap(void *buf, size_t size);
241 void zfs_oldacl_byteswap(void *buf, size_t size);
242 void zfs_acl_byteswap(void *buf, size_t size);
243 void zfs_znode_byteswap(void *buf, size_t size);
245 #define DS_FIND_SNAPSHOTS (1<<0)
246 #define DS_FIND_CHILDREN (1<<1)
249 * The maximum number of bytes that can be accessed as part of one
250 * operation, including metadata.
252 #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */
253 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
255 #define DMU_USERUSED_OBJECT (-1ULL)
256 #define DMU_GROUPUSED_OBJECT (-2ULL)
259 * artificial blkids for bonus buffer and spill blocks
261 #define DMU_BONUS_BLKID (-1ULL)
262 #define DMU_SPILL_BLKID (-2ULL)
264 * Public routines to create, destroy, open, and close objsets.
266 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
267 int dmu_objset_own(const char *name, dmu_objset_type_t type,
268 boolean_t readonly, void *tag, objset_t **osp);
269 void dmu_objset_rele(objset_t *os, void *tag);
270 void dmu_objset_disown(objset_t *os, void *tag);
271 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
273 void dmu_objset_evict_dbufs(objset_t *os);
274 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
275 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
276 int dmu_get_recursive_snaps_nvl(char *fsname, const char *snapname,
277 struct nvlist *snaps);
278 int dmu_objset_clone(const char *name, const char *origin);
279 int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
280 struct nvlist *errlist);
281 int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
282 int dmu_objset_snapshot_tmp(const char *, const char *, int);
283 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
285 void dmu_objset_byteswap(void *buf, size_t size);
286 int dsl_dataset_rename_snapshot(const char *fsname,
287 const char *oldsnapname, const char *newsnapname, boolean_t recursive);
289 typedef struct dmu_buf {
290 uint64_t db_object; /* object that this buffer is part of */
291 uint64_t db_offset; /* byte offset in this object */
292 uint64_t db_size; /* size of buffer in bytes */
293 void *db_data; /* data in buffer */
296 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
299 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
301 #define DMU_POOL_DIRECTORY_OBJECT 1
302 #define DMU_POOL_CONFIG "config"
303 #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
304 #define DMU_POOL_FEATURES_FOR_READ "features_for_read"
305 #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
306 #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg"
307 #define DMU_POOL_ROOT_DATASET "root_dataset"
308 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
309 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
310 #define DMU_POOL_ERRLOG_LAST "errlog_last"
311 #define DMU_POOL_SPARES "spares"
312 #define DMU_POOL_DEFLATE "deflate"
313 #define DMU_POOL_HISTORY "history"
314 #define DMU_POOL_PROPS "pool_props"
315 #define DMU_POOL_L2CACHE "l2cache"
316 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
317 #define DMU_POOL_DDT "DDT-%s-%s-%s"
318 #define DMU_POOL_DDT_STATS "DDT-statistics"
319 #define DMU_POOL_CREATION_VERSION "creation_version"
320 #define DMU_POOL_SCAN "scan"
321 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
322 #define DMU_POOL_BPTREE_OBJ "bptree_obj"
323 #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
326 * Allocate an object from this objset. The range of object numbers
327 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
329 * The transaction must be assigned to a txg. The newly allocated
330 * object will be "held" in the transaction (ie. you can modify the
331 * newly allocated object in this transaction).
333 * dmu_object_alloc() chooses an object and returns it in *objectp.
335 * dmu_object_claim() allocates a specific object number. If that
336 * number is already allocated, it fails and returns EEXIST.
338 * Return 0 on success, or ENOSPC or EEXIST as specified above.
340 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
341 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
342 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
343 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
344 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
345 int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
348 * Free an object from this objset.
350 * The object's data will be freed as well (ie. you don't need to call
351 * dmu_free(object, 0, -1, tx)).
353 * The object need not be held in the transaction.
355 * If there are any holds on this object's buffers (via dmu_buf_hold()),
356 * or tx holds on the object (via dmu_tx_hold_object()), you can not
357 * free it; it fails and returns EBUSY.
359 * If the object is not allocated, it fails and returns ENOENT.
361 * Return 0 on success, or EBUSY or ENOENT as specified above.
363 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
366 * Find the next allocated or free object.
368 * The objectp parameter is in-out. It will be updated to be the next
369 * object which is allocated. Ignore objects which have not been
370 * modified since txg.
372 * XXX Can only be called on a objset with no dirty data.
374 * Returns 0 on success, or ENOENT if there are no more objects.
376 int dmu_object_next(objset_t *os, uint64_t *objectp,
377 boolean_t hole, uint64_t txg);
380 * Set the data blocksize for an object.
382 * The object cannot have any blocks allcated beyond the first. If
383 * the first block is allocated already, the new size must be greater
384 * than the current block size. If these conditions are not met,
385 * ENOTSUP will be returned.
387 * Returns 0 on success, or EBUSY if there are any holds on the object
388 * contents, or ENOTSUP as described above.
390 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
391 int ibs, dmu_tx_t *tx);
394 * Set the checksum property on a dnode. The new checksum algorithm will
395 * apply to all newly written blocks; existing blocks will not be affected.
397 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
401 * Set the compress property on a dnode. The new compression algorithm will
402 * apply to all newly written blocks; existing blocks will not be affected.
404 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
408 dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
409 void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
410 int compressed_size, int byteorder, dmu_tx_t *tx);
413 * Decide how to write a block: checksum, compression, number of copies, etc.
415 #define WP_NOFILL 0x1
416 #define WP_DMU_SYNC 0x2
419 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
420 struct zio_prop *zp);
422 * The bonus data is accessed more or less like a regular buffer.
423 * You must dmu_bonus_hold() to get the buffer, which will give you a
424 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
425 * data. As with any normal buffer, you must call dmu_buf_read() to
426 * read db_data, dmu_buf_will_dirty() before modifying it, and the
427 * object must be held in an assigned transaction before calling
428 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
429 * buffer as well. You must release your hold with dmu_buf_rele().
431 * Returns ENOENT, EIO, or 0.
433 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
434 int dmu_bonus_max(void);
435 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
436 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
437 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
438 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
441 * Special spill buffer support used by "SA" framework
444 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
445 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
446 void *tag, dmu_buf_t **dbp);
447 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
450 * Obtain the DMU buffer from the specified object which contains the
451 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
452 * that it will remain in memory. You must release the hold with
453 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
454 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
456 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
457 * on the returned buffer before reading or writing the buffer's
458 * db_data. The comments for those routines describe what particular
459 * operations are valid after calling them.
461 * The object number must be a valid, allocated object number.
463 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
464 void *tag, dmu_buf_t **, int flags);
465 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
466 void dmu_buf_rele(dmu_buf_t *db, void *tag);
467 uint64_t dmu_buf_refcount(dmu_buf_t *db);
470 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
471 * range of an object. A pointer to an array of dmu_buf_t*'s is
472 * returned (in *dbpp).
474 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
475 * frees the array. The hold on the array of buffers MUST be released
476 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
477 * individually with dmu_buf_rele.
479 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
480 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
481 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
484 * Returns NULL on success, or the existing user ptr if it's already
487 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
489 * If non-NULL, pageout func will be called when this buffer is being
490 * excised from the cache, so that you can clean up the data structure
491 * pointed to by user_ptr.
493 * dmu_evict_user() will call the pageout func for all buffers in a
494 * objset with a given pageout func.
496 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr,
497 dmu_buf_evict_func_t *pageout_func);
499 * set_user_ie is the same as set_user, but request immediate eviction
500 * when hold count goes to zero.
502 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
503 dmu_buf_evict_func_t *pageout_func);
504 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
505 void *user_ptr, dmu_buf_evict_func_t *pageout_func);
506 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
509 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
511 void *dmu_buf_get_user(dmu_buf_t *db);
514 * Returns the blkptr associated with this dbuf, or NULL if not set.
516 struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
519 * Indicate that you are going to modify the buffer's data (db_data).
521 * The transaction (tx) must be assigned to a txg (ie. you've called
522 * dmu_tx_assign()). The buffer's object must be held in the tx
523 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
525 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
528 * Tells if the given dbuf is freeable.
530 boolean_t dmu_buf_freeable(dmu_buf_t *);
533 * You must create a transaction, then hold the objects which you will
534 * (or might) modify as part of this transaction. Then you must assign
535 * the transaction to a transaction group. Once the transaction has
536 * been assigned, you can modify buffers which belong to held objects as
537 * part of this transaction. You can't modify buffers before the
538 * transaction has been assigned; you can't modify buffers which don't
539 * belong to objects which this transaction holds; you can't hold
540 * objects once the transaction has been assigned. You may hold an
541 * object which you are going to free (with dmu_object_free()), but you
544 * You can abort the transaction before it has been assigned.
546 * Note that you may hold buffers (with dmu_buf_hold) at any time,
547 * regardless of transaction state.
550 #define DMU_NEW_OBJECT (-1ULL)
551 #define DMU_OBJECT_END (-1ULL)
553 dmu_tx_t *dmu_tx_create(objset_t *os);
554 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
555 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
557 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
558 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
559 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
560 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
561 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
562 void dmu_tx_abort(dmu_tx_t *tx);
563 int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
564 void dmu_tx_wait(dmu_tx_t *tx);
565 void dmu_tx_commit(dmu_tx_t *tx);
566 void dmu_tx_mark_netfree(dmu_tx_t *tx);
569 * To register a commit callback, dmu_tx_callback_register() must be called.
571 * dcb_data is a pointer to caller private data that is passed on as a
572 * callback parameter. The caller is responsible for properly allocating and
575 * When registering a callback, the transaction must be already created, but
576 * it cannot be committed or aborted. It can be assigned to a txg or not.
578 * The callback will be called after the transaction has been safely written
579 * to stable storage and will also be called if the dmu_tx is aborted.
580 * If there is any error which prevents the transaction from being committed to
581 * disk, the callback will be called with a value of error != 0.
583 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
585 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
589 * Free up the data blocks for a defined range of a file. If size is
590 * -1, the range from offset to end-of-file is freed.
592 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
593 uint64_t size, dmu_tx_t *tx);
594 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
596 int dmu_free_long_object(objset_t *os, uint64_t object);
599 * Convenience functions.
601 * Canfail routines will return 0 on success, or an errno if there is a
602 * nonrecoverable I/O error.
604 #define DMU_READ_PREFETCH 0 /* prefetch */
605 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
606 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
607 void *buf, uint32_t flags);
608 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
609 const void *buf, dmu_tx_t *tx);
610 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
612 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
613 int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
614 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
616 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
620 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
621 uint64_t size, struct page *pp, dmu_tx_t *tx);
623 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
624 uint64_t size, struct vm_page **ppa, dmu_tx_t *tx);
627 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
628 void dmu_return_arcbuf(struct arc_buf *buf);
629 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
631 int dmu_xuio_init(struct xuio *uio, int niov);
632 void dmu_xuio_fini(struct xuio *uio);
633 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
635 int dmu_xuio_cnt(struct xuio *uio);
636 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
637 void dmu_xuio_clear(struct xuio *uio, int i);
638 void xuio_stat_wbuf_copied();
639 void xuio_stat_wbuf_nocopy();
641 extern int zfs_prefetch_disable;
642 extern int zfs_max_recordsize;
645 * Asynchronously try to read in the data.
647 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
650 typedef struct dmu_object_info {
651 /* All sizes are in bytes unless otherwise indicated. */
652 uint32_t doi_data_block_size;
653 uint32_t doi_metadata_block_size;
654 dmu_object_type_t doi_type;
655 dmu_object_type_t doi_bonus_type;
656 uint64_t doi_bonus_size;
657 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
658 uint8_t doi_checksum;
659 uint8_t doi_compress;
662 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
663 uint64_t doi_max_offset;
664 uint64_t doi_fill_count; /* number of non-empty blocks */
667 typedef void arc_byteswap_func_t(void *buf, size_t size);
669 typedef struct dmu_object_type_info {
670 dmu_object_byteswap_t ot_byteswap;
671 boolean_t ot_metadata;
673 } dmu_object_type_info_t;
675 typedef struct dmu_object_byteswap_info {
676 arc_byteswap_func_t *ob_func;
678 } dmu_object_byteswap_info_t;
680 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
681 extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
684 * Get information on a DMU object.
686 * Return 0 on success or ENOENT if object is not allocated.
688 * If doi is NULL, just indicates whether the object exists.
690 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
691 /* Like dmu_object_info, but faster if you have a held dnode in hand. */
692 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
693 /* Like dmu_object_info, but faster if you have a held dbuf in hand. */
694 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
696 * Like dmu_object_info_from_db, but faster still when you only care about
697 * the size. This is specifically optimized for zfs_getattr().
699 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
700 u_longlong_t *nblk512);
702 typedef struct dmu_objset_stats {
703 uint64_t dds_num_clones; /* number of clones of this */
704 uint64_t dds_creation_txg;
706 dmu_objset_type_t dds_type;
707 uint8_t dds_is_snapshot;
708 uint8_t dds_inconsistent;
709 char dds_origin[MAXNAMELEN];
710 } dmu_objset_stats_t;
713 * Get stats on a dataset.
715 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
718 * Add entries to the nvlist for all the objset's properties. See
719 * zfs_prop_table[] and zfs(1m) for details on the properties.
721 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
724 * Get the space usage statistics for statvfs().
726 * refdbytes is the amount of space "referenced" by this objset.
727 * availbytes is the amount of space available to this objset, taking
728 * into account quotas & reservations, assuming that no other objsets
729 * use the space first. These values correspond to the 'referenced' and
730 * 'available' properties, described in the zfs(1m) manpage.
732 * usedobjs and availobjs are the number of objects currently allocated,
735 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
736 uint64_t *usedobjsp, uint64_t *availobjsp);
739 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
740 * (Contrast with the ds_guid which is a 64-bit ID that will never
741 * change, so there is a small probability that it will collide.)
743 uint64_t dmu_objset_fsid_guid(objset_t *os);
746 * Get the [cm]time for an objset's snapshot dir
748 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
750 int dmu_objset_is_snapshot(objset_t *os);
752 extern struct spa *dmu_objset_spa(objset_t *os);
753 extern struct zilog *dmu_objset_zil(objset_t *os);
754 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
755 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
756 extern void dmu_objset_name(objset_t *os, char *buf);
757 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
758 extern uint64_t dmu_objset_id(objset_t *os);
759 extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
760 extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
761 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
762 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
763 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
764 int maxlen, boolean_t *conflict);
765 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
766 uint64_t *idp, uint64_t *offp);
768 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
769 void *bonus, uint64_t *userp, uint64_t *groupp);
770 extern void dmu_objset_register_type(dmu_objset_type_t ost,
771 objset_used_cb_t *cb);
772 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
773 extern void *dmu_objset_get_user(objset_t *os);
776 * Return the txg number for the given assigned transaction.
778 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
782 * If a parent zio is provided this function initiates a write on the
783 * provided buffer as a child of the parent zio.
784 * In the absence of a parent zio, the write is completed synchronously.
785 * At write completion, blk is filled with the bp of the written block.
786 * Note that while the data covered by this function will be on stable
787 * storage when the write completes this new data does not become a
788 * permanent part of the file until the associated transaction commits.
792 * {zfs,zvol,ztest}_get_done() args
795 struct zilog *zgd_zilog;
796 struct blkptr *zgd_bp;
802 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
803 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
806 * Find the next hole or data block in file starting at *off
807 * Return found offset in *off. Return ESRCH for end of file.
809 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
813 * Initial setup and final teardown.
815 extern void dmu_init(void);
816 extern void dmu_fini(void);
818 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
819 uint64_t object, uint64_t offset, int len);
820 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
821 dmu_traverse_cb_t cb, void *arg);
822 int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
823 struct file *fp, offset_t *offp);
826 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
827 extern uint64_t zfs_crc64_table[256];
829 extern int zfs_mdcomp_disable;
835 #endif /* _SYS_DMU_H */