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 by Delphix. All rights reserved.
24 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 /* Portions Copyright 2010 Robert Milkowski */
33 * This file describes the interface that the DMU provides for its
36 * The DMU also interacts with the SPA. That interface is described in
40 #include <sys/inttypes.h>
41 #include <sys/types.h>
42 #include <sys/param.h>
70 typedef struct objset objset_t;
71 typedef struct dmu_tx dmu_tx_t;
72 typedef struct dsl_dir dsl_dir_t;
74 typedef enum dmu_object_type {
77 DMU_OT_OBJECT_DIRECTORY, /* ZAP */
78 DMU_OT_OBJECT_ARRAY, /* UINT64 */
79 DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
80 DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
81 DMU_OT_BPOBJ, /* UINT64 */
82 DMU_OT_BPOBJ_HDR, /* UINT64 */
84 DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
85 DMU_OT_SPACE_MAP, /* UINT64 */
87 DMU_OT_INTENT_LOG, /* UINT64 */
89 DMU_OT_DNODE, /* DNODE */
90 DMU_OT_OBJSET, /* OBJSET */
92 DMU_OT_DSL_DIR, /* UINT64 */
93 DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
94 DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
95 DMU_OT_DSL_PROPS, /* ZAP */
96 DMU_OT_DSL_DATASET, /* UINT64 */
98 DMU_OT_ZNODE, /* ZNODE */
99 DMU_OT_OLDACL, /* Old ACL */
100 DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
101 DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
102 DMU_OT_MASTER_NODE, /* ZAP */
103 DMU_OT_UNLINKED_SET, /* ZAP */
105 DMU_OT_ZVOL, /* UINT8 */
106 DMU_OT_ZVOL_PROP, /* ZAP */
107 /* other; for testing only! */
108 DMU_OT_PLAIN_OTHER, /* UINT8 */
109 DMU_OT_UINT64_OTHER, /* UINT64 */
110 DMU_OT_ZAP_OTHER, /* ZAP */
111 /* new object types: */
112 DMU_OT_ERROR_LOG, /* ZAP */
113 DMU_OT_SPA_HISTORY, /* UINT8 */
114 DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
115 DMU_OT_POOL_PROPS, /* ZAP */
116 DMU_OT_DSL_PERMS, /* ZAP */
117 DMU_OT_ACL, /* ACL */
118 DMU_OT_SYSACL, /* SYSACL */
119 DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
120 DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
121 DMU_OT_NEXT_CLONES, /* ZAP */
122 DMU_OT_SCAN_QUEUE, /* ZAP */
123 DMU_OT_USERGROUP_USED, /* ZAP */
124 DMU_OT_USERGROUP_QUOTA, /* ZAP */
125 DMU_OT_USERREFS, /* ZAP */
126 DMU_OT_DDT_ZAP, /* ZAP */
127 DMU_OT_DDT_STATS, /* ZAP */
128 DMU_OT_SA, /* System attr */
129 DMU_OT_SA_MASTER_NODE, /* ZAP */
130 DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
131 DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
132 DMU_OT_SCAN_XLATE, /* ZAP */
133 DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
134 DMU_OT_DEADLIST, /* ZAP */
135 DMU_OT_DEADLIST_HDR, /* UINT64 */
136 DMU_OT_DSL_CLONES, /* ZAP */
137 DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
141 typedef enum dmu_objset_type {
146 DMU_OST_OTHER, /* For testing only! */
147 DMU_OST_ANY, /* Be careful! */
151 void byteswap_uint64_array(void *buf, size_t size);
152 void byteswap_uint32_array(void *buf, size_t size);
153 void byteswap_uint16_array(void *buf, size_t size);
154 void byteswap_uint8_array(void *buf, size_t size);
155 void zap_byteswap(void *buf, size_t size);
156 void zfs_oldacl_byteswap(void *buf, size_t size);
157 void zfs_acl_byteswap(void *buf, size_t size);
158 void zfs_znode_byteswap(void *buf, size_t size);
160 #define DS_FIND_SNAPSHOTS (1<<0)
161 #define DS_FIND_CHILDREN (1<<1)
164 * The maximum number of bytes that can be accessed as part of one
165 * operation, including metadata.
167 #define DMU_MAX_ACCESS (10<<20) /* 10MB */
168 #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
170 #define DMU_USERUSED_OBJECT (-1ULL)
171 #define DMU_GROUPUSED_OBJECT (-2ULL)
172 #define DMU_DEADLIST_OBJECT (-3ULL)
175 * artificial blkids for bonus buffer and spill blocks
177 #define DMU_BONUS_BLKID (-1ULL)
178 #define DMU_SPILL_BLKID (-2ULL)
180 * Public routines to create, destroy, open, and close objsets.
182 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
183 int dmu_objset_own(const char *name, dmu_objset_type_t type,
184 boolean_t readonly, void *tag, objset_t **osp);
185 void dmu_objset_rele(objset_t *os, void *tag);
186 void dmu_objset_disown(objset_t *os, void *tag);
187 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
189 int dmu_objset_evict_dbufs(objset_t *os);
190 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
191 void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
192 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
194 int dmu_objset_destroy(const char *name, boolean_t defer);
195 int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer, char *);
196 int dmu_objset_snapshot(char *fsname, char *snapname, char *tag,
197 struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd);
198 int dmu_objset_rename(const char *name, const char *newname,
199 boolean_t recursive);
200 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
202 void dmu_objset_byteswap(void *buf, size_t size);
204 typedef struct dmu_buf {
205 uint64_t db_object; /* object that this buffer is part of */
206 uint64_t db_offset; /* byte offset in this object */
207 uint64_t db_size; /* size of buffer in bytes */
208 void *db_data; /* data in buffer */
211 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
214 * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
216 #define DMU_POOL_DIRECTORY_OBJECT 1
217 #define DMU_POOL_CONFIG "config"
218 #define DMU_POOL_ROOT_DATASET "root_dataset"
219 #define DMU_POOL_SYNC_BPOBJ "sync_bplist"
220 #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
221 #define DMU_POOL_ERRLOG_LAST "errlog_last"
222 #define DMU_POOL_SPARES "spares"
223 #define DMU_POOL_DEFLATE "deflate"
224 #define DMU_POOL_HISTORY "history"
225 #define DMU_POOL_PROPS "pool_props"
226 #define DMU_POOL_L2CACHE "l2cache"
227 #define DMU_POOL_TMP_USERREFS "tmp_userrefs"
228 #define DMU_POOL_DDT "DDT-%s-%s-%s"
229 #define DMU_POOL_DDT_STATS "DDT-statistics"
230 #define DMU_POOL_CREATION_VERSION "creation_version"
231 #define DMU_POOL_SCAN "scan"
232 #define DMU_POOL_FREE_BPOBJ "free_bpobj"
235 * Allocate an object from this objset. The range of object numbers
236 * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
238 * The transaction must be assigned to a txg. The newly allocated
239 * object will be "held" in the transaction (ie. you can modify the
240 * newly allocated object in this transaction).
242 * dmu_object_alloc() chooses an object and returns it in *objectp.
244 * dmu_object_claim() allocates a specific object number. If that
245 * number is already allocated, it fails and returns EEXIST.
247 * Return 0 on success, or ENOSPC or EEXIST as specified above.
249 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
250 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
251 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
252 int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
253 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
254 int blocksize, dmu_object_type_t bonustype, int bonuslen);
257 * Free an object from this objset.
259 * The object's data will be freed as well (ie. you don't need to call
260 * dmu_free(object, 0, -1, tx)).
262 * The object need not be held in the transaction.
264 * If there are any holds on this object's buffers (via dmu_buf_hold()),
265 * or tx holds on the object (via dmu_tx_hold_object()), you can not
266 * free it; it fails and returns EBUSY.
268 * If the object is not allocated, it fails and returns ENOENT.
270 * Return 0 on success, or EBUSY or ENOENT as specified above.
272 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
275 * Find the next allocated or free object.
277 * The objectp parameter is in-out. It will be updated to be the next
278 * object which is allocated. Ignore objects which have not been
279 * modified since txg.
281 * XXX Can only be called on a objset with no dirty data.
283 * Returns 0 on success, or ENOENT if there are no more objects.
285 int dmu_object_next(objset_t *os, uint64_t *objectp,
286 boolean_t hole, uint64_t txg);
289 * Set the data blocksize for an object.
291 * The object cannot have any blocks allcated beyond the first. If
292 * the first block is allocated already, the new size must be greater
293 * than the current block size. If these conditions are not met,
294 * ENOTSUP will be returned.
296 * Returns 0 on success, or EBUSY if there are any holds on the object
297 * contents, or ENOTSUP as described above.
299 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
300 int ibs, dmu_tx_t *tx);
303 * Set the checksum property on a dnode. The new checksum algorithm will
304 * apply to all newly written blocks; existing blocks will not be affected.
306 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
310 * Set the compress property on a dnode. The new compression algorithm will
311 * apply to all newly written blocks; existing blocks will not be affected.
313 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
317 * Decide how to write a block: checksum, compression, number of copies, etc.
319 #define WP_NOFILL 0x1
320 #define WP_DMU_SYNC 0x2
323 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
324 struct zio_prop *zp);
326 * The bonus data is accessed more or less like a regular buffer.
327 * You must dmu_bonus_hold() to get the buffer, which will give you a
328 * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
329 * data. As with any normal buffer, you must call dmu_buf_read() to
330 * read db_data, dmu_buf_will_dirty() before modifying it, and the
331 * object must be held in an assigned transaction before calling
332 * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
333 * buffer as well. You must release your hold with dmu_buf_rele().
335 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
336 int dmu_bonus_max(void);
337 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
338 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
339 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
340 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
343 * Special spill buffer support used by "SA" framework
346 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
347 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
348 void *tag, dmu_buf_t **dbp);
349 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
352 * Obtain the DMU buffer from the specified object which contains the
353 * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
354 * that it will remain in memory. You must release the hold with
355 * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your
356 * hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
358 * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
359 * on the returned buffer before reading or writing the buffer's
360 * db_data. The comments for those routines describe what particular
361 * operations are valid after calling them.
363 * The object number must be a valid, allocated object number.
365 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
366 void *tag, dmu_buf_t **, int flags);
367 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
368 void dmu_buf_rele(dmu_buf_t *db, void *tag);
369 uint64_t dmu_buf_refcount(dmu_buf_t *db);
372 * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
373 * range of an object. A pointer to an array of dmu_buf_t*'s is
374 * returned (in *dbpp).
376 * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
377 * frees the array. The hold on the array of buffers MUST be released
378 * with dmu_buf_rele_array. You can NOT release the hold on each buffer
379 * individually with dmu_buf_rele.
381 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
382 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
383 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
386 * Returns NULL on success, or the existing user ptr if it's already
389 * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
391 * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
392 * will be set to db->db_data when you are allowed to access it. Note
393 * that db->db_data (the pointer) can change when you do dmu_buf_read(),
394 * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
395 * *user_data_ptr_ptr will be set to the new value when it changes.
397 * If non-NULL, pageout func will be called when this buffer is being
398 * excised from the cache, so that you can clean up the data structure
399 * pointed to by user_ptr.
401 * dmu_evict_user() will call the pageout func for all buffers in a
402 * objset with a given pageout func.
404 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
405 dmu_buf_evict_func_t *pageout_func);
407 * set_user_ie is the same as set_user, but request immediate eviction
408 * when hold count goes to zero.
410 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
411 void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
412 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
413 void *user_ptr, void *user_data_ptr_ptr,
414 dmu_buf_evict_func_t *pageout_func);
415 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
418 * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
420 void *dmu_buf_get_user(dmu_buf_t *db);
423 * Indicate that you are going to modify the buffer's data (db_data).
425 * The transaction (tx) must be assigned to a txg (ie. you've called
426 * dmu_tx_assign()). The buffer's object must be held in the tx
427 * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
429 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
432 * Tells if the given dbuf is freeable.
434 boolean_t dmu_buf_freeable(dmu_buf_t *);
437 * You must create a transaction, then hold the objects which you will
438 * (or might) modify as part of this transaction. Then you must assign
439 * the transaction to a transaction group. Once the transaction has
440 * been assigned, you can modify buffers which belong to held objects as
441 * part of this transaction. You can't modify buffers before the
442 * transaction has been assigned; you can't modify buffers which don't
443 * belong to objects which this transaction holds; you can't hold
444 * objects once the transaction has been assigned. You may hold an
445 * object which you are going to free (with dmu_object_free()), but you
448 * You can abort the transaction before it has been assigned.
450 * Note that you may hold buffers (with dmu_buf_hold) at any time,
451 * regardless of transaction state.
454 #define DMU_NEW_OBJECT (-1ULL)
455 #define DMU_OBJECT_END (-1ULL)
457 dmu_tx_t *dmu_tx_create(objset_t *os);
458 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
459 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
461 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
462 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
463 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
464 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
465 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
466 void dmu_tx_abort(dmu_tx_t *tx);
467 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
468 void dmu_tx_wait(dmu_tx_t *tx);
469 void dmu_tx_commit(dmu_tx_t *tx);
472 * To register a commit callback, dmu_tx_callback_register() must be called.
474 * dcb_data is a pointer to caller private data that is passed on as a
475 * callback parameter. The caller is responsible for properly allocating and
478 * When registering a callback, the transaction must be already created, but
479 * it cannot be committed or aborted. It can be assigned to a txg or not.
481 * The callback will be called after the transaction has been safely written
482 * to stable storage and will also be called if the dmu_tx is aborted.
483 * If there is any error which prevents the transaction from being committed to
484 * disk, the callback will be called with a value of error != 0.
486 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
488 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
492 * Free up the data blocks for a defined range of a file. If size is
493 * zero, the range from offset to end-of-file is freed.
495 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
496 uint64_t size, dmu_tx_t *tx);
497 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
499 int dmu_free_object(objset_t *os, uint64_t object);
502 * Convenience functions.
504 * Canfail routines will return 0 on success, or an errno if there is a
505 * nonrecoverable I/O error.
507 #define DMU_READ_PREFETCH 0 /* prefetch */
508 #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
509 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
510 void *buf, uint32_t flags);
511 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
512 const void *buf, dmu_tx_t *tx);
513 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
516 #include <linux/blkdev_compat.h>
517 int dmu_read_req(objset_t *os, uint64_t object, struct request *req);
518 int dmu_write_req(objset_t *os, uint64_t object, struct request *req,
520 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
521 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
523 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
526 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
527 void dmu_return_arcbuf(struct arc_buf *buf);
528 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
530 int dmu_xuio_init(struct xuio *uio, int niov);
531 void dmu_xuio_fini(struct xuio *uio);
532 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
534 int dmu_xuio_cnt(struct xuio *uio);
535 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
536 void dmu_xuio_clear(struct xuio *uio, int i);
537 void xuio_stat_wbuf_copied(void);
538 void xuio_stat_wbuf_nocopy(void);
540 extern int zfs_prefetch_disable;
543 * Asynchronously try to read in the data.
545 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
548 typedef struct dmu_object_info {
549 /* All sizes are in bytes unless otherwise indicated. */
550 uint32_t doi_data_block_size;
551 uint32_t doi_metadata_block_size;
552 dmu_object_type_t doi_type;
553 dmu_object_type_t doi_bonus_type;
554 uint64_t doi_bonus_size;
555 uint8_t doi_indirection; /* 2 = dnode->indirect->data */
556 uint8_t doi_checksum;
557 uint8_t doi_compress;
559 uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
560 uint64_t doi_max_offset;
561 uint64_t doi_fill_count; /* number of non-empty blocks */
564 typedef void arc_byteswap_func_t(void *buf, size_t size);
566 typedef struct dmu_object_type_info {
567 arc_byteswap_func_t *ot_byteswap;
568 boolean_t ot_metadata;
570 } dmu_object_type_info_t;
572 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
575 * Get information on a DMU object.
577 * Return 0 on success or ENOENT if object is not allocated.
579 * If doi is NULL, just indicates whether the object exists.
581 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
582 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
583 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
584 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
585 u_longlong_t *nblk512);
587 typedef struct dmu_objset_stats {
588 uint64_t dds_num_clones; /* number of clones of this */
589 uint64_t dds_creation_txg;
591 dmu_objset_type_t dds_type;
592 uint8_t dds_is_snapshot;
593 uint8_t dds_inconsistent;
594 char dds_origin[MAXNAMELEN];
595 } dmu_objset_stats_t;
598 * Get stats on a dataset.
600 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
603 * Add entries to the nvlist for all the objset's properties. See
604 * zfs_prop_table[] and zfs(1m) for details on the properties.
606 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
609 * Get the space usage statistics for statvfs().
611 * refdbytes is the amount of space "referenced" by this objset.
612 * availbytes is the amount of space available to this objset, taking
613 * into account quotas & reservations, assuming that no other objsets
614 * use the space first. These values correspond to the 'referenced' and
615 * 'available' properties, described in the zfs(1m) manpage.
617 * usedobjs and availobjs are the number of objects currently allocated,
620 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
621 uint64_t *usedobjsp, uint64_t *availobjsp);
624 * The fsid_guid is a 56-bit ID that can change to avoid collisions.
625 * (Contrast with the ds_guid which is a 64-bit ID that will never
626 * change, so there is a small probability that it will collide.)
628 uint64_t dmu_objset_fsid_guid(objset_t *os);
631 * Get the [cm]time for an objset's snapshot dir
633 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
635 int dmu_objset_is_snapshot(objset_t *os);
637 extern struct spa *dmu_objset_spa(objset_t *os);
638 extern struct zilog *dmu_objset_zil(objset_t *os);
639 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
640 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
641 extern void dmu_objset_name(objset_t *os, char *buf);
642 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
643 extern uint64_t dmu_objset_id(objset_t *os);
644 extern uint64_t dmu_objset_syncprop(objset_t *os);
645 extern uint64_t dmu_objset_logbias(objset_t *os);
646 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
647 uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
648 extern int dmu_snapshot_id(objset_t *os, const char *snapname, uint64_t *idp);
649 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
650 int maxlen, boolean_t *conflict);
651 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
652 uint64_t *idp, uint64_t *offp);
654 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
655 void *bonus, uint64_t *userp, uint64_t *groupp);
656 extern void dmu_objset_register_type(dmu_objset_type_t ost,
657 objset_used_cb_t *cb);
658 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
659 extern void *dmu_objset_get_user(objset_t *os);
662 * Return the txg number for the given assigned transaction.
664 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
668 * If a parent zio is provided this function initiates a write on the
669 * provided buffer as a child of the parent zio.
670 * In the absence of a parent zio, the write is completed synchronously.
671 * At write completion, blk is filled with the bp of the written block.
672 * Note that while the data covered by this function will be on stable
673 * storage when the write completes this new data does not become a
674 * permanent part of the file until the associated transaction commits.
678 * {zfs,zvol,ztest}_get_done() args
681 struct zilog *zgd_zilog;
682 struct blkptr *zgd_bp;
688 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
689 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
692 * Find the next hole or data block in file starting at *off
693 * Return found offset in *off. Return ESRCH for end of file.
695 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
699 * Initial setup and final teardown.
701 extern void dmu_init(void);
702 extern void dmu_fini(void);
704 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
705 uint64_t object, uint64_t offset, int len);
706 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
707 dmu_traverse_cb_t cb, void *arg);
709 int dmu_send(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
710 int outfd, struct vnode *vp, offset_t *off);
711 int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorign,
714 typedef struct dmu_recv_cookie {
716 * This structure is opaque!
718 * If logical and real are different, we are recving the stream
719 * into the "real" temporary clone, and then switching it with
720 * the "logical" target.
722 struct dsl_dataset *drc_logical_ds;
723 struct dsl_dataset *drc_real_ds;
724 struct drr_begin *drc_drrb;
729 struct avl_tree *drc_guid_to_ds_map;
732 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
733 boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
734 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
735 int cleanup_fd, uint64_t *action_handlep);
736 int dmu_recv_end(dmu_recv_cookie_t *drc);
738 int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp,
742 #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
743 extern uint64_t zfs_crc64_table[256];
749 #endif /* _SYS_DMU_H */