xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 3cb34c601f3ef3016f638574f5982e80c3735c71)
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 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_SYS_DMU_H
27 #define	_SYS_DMU_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 /*
32  * This file describes the interface that the DMU provides for its
33  * consumers.
34  *
35  * The DMU also interacts with the SPA.  That interface is described in
36  * dmu_spa.h.
37  */
38 
39 #include <sys/inttypes.h>
40 #include <sys/types.h>
41 #include <sys/param.h>
42 #include <sys/cred.h>
43 
44 #ifdef	__cplusplus
45 extern "C" {
46 #endif
47 
48 struct uio;
49 struct page;
50 struct vnode;
51 struct spa;
52 struct zilog;
53 struct zio;
54 struct blkptr;
55 struct zap_cursor;
56 struct dsl_dataset;
57 struct dsl_pool;
58 struct dnode;
59 struct drr_begin;
60 struct drr_end;
61 struct zbookmark;
62 struct spa;
63 struct nvlist;
64 struct objset_impl;
65 
66 typedef struct objset objset_t;
67 typedef struct dmu_tx dmu_tx_t;
68 typedef struct dsl_dir dsl_dir_t;
69 
70 typedef enum dmu_object_type {
71 	DMU_OT_NONE,
72 	/* general: */
73 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
74 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
75 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
76 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
77 	DMU_OT_BPLIST,			/* UINT64 */
78 	DMU_OT_BPLIST_HDR,		/* UINT64 */
79 	/* spa: */
80 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
81 	DMU_OT_SPACE_MAP,		/* UINT64 */
82 	/* zil: */
83 	DMU_OT_INTENT_LOG,		/* UINT64 */
84 	/* dmu: */
85 	DMU_OT_DNODE,			/* DNODE */
86 	DMU_OT_OBJSET,			/* OBJSET */
87 	/* dsl: */
88 	DMU_OT_DSL_DIR,			/* UINT64 */
89 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
90 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
91 	DMU_OT_DSL_PROPS,		/* ZAP */
92 	DMU_OT_DSL_DATASET,		/* UINT64 */
93 	/* zpl: */
94 	DMU_OT_ZNODE,			/* ZNODE */
95 	DMU_OT_OLDACL,			/* Old ACL */
96 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
97 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
98 	DMU_OT_MASTER_NODE,		/* ZAP */
99 	DMU_OT_UNLINKED_SET,		/* ZAP */
100 	/* zvol: */
101 	DMU_OT_ZVOL,			/* UINT8 */
102 	DMU_OT_ZVOL_PROP,		/* ZAP */
103 	/* other; for testing only! */
104 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
105 	DMU_OT_UINT64_OTHER,		/* UINT64 */
106 	DMU_OT_ZAP_OTHER,		/* ZAP */
107 	/* new object types: */
108 	DMU_OT_ERROR_LOG,		/* ZAP */
109 	DMU_OT_SPA_HISTORY,		/* UINT8 */
110 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
111 	DMU_OT_POOL_PROPS,		/* ZAP */
112 	DMU_OT_DSL_PERMS,		/* ZAP */
113 	DMU_OT_ACL,			/* ACL */
114 	DMU_OT_SYSACL,			/* SYSACL */
115 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
116 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
117 	DMU_OT_NUMTYPES
118 } dmu_object_type_t;
119 
120 typedef enum dmu_objset_type {
121 	DMU_OST_NONE,
122 	DMU_OST_META,
123 	DMU_OST_ZFS,
124 	DMU_OST_ZVOL,
125 	DMU_OST_OTHER,			/* For testing only! */
126 	DMU_OST_ANY,			/* Be careful! */
127 	DMU_OST_NUMTYPES
128 } dmu_objset_type_t;
129 
130 void byteswap_uint64_array(void *buf, size_t size);
131 void byteswap_uint32_array(void *buf, size_t size);
132 void byteswap_uint16_array(void *buf, size_t size);
133 void byteswap_uint8_array(void *buf, size_t size);
134 void zap_byteswap(void *buf, size_t size);
135 void zfs_oldacl_byteswap(void *buf, size_t size);
136 void zfs_acl_byteswap(void *buf, size_t size);
137 void zfs_znode_byteswap(void *buf, size_t size);
138 
139 #define	DS_MODE_NONE		0	/* invalid, to aid debugging */
140 #define	DS_MODE_STANDARD	1	/* normal access, no special needs */
141 #define	DS_MODE_PRIMARY		2	/* the "main" access, e.g. a mount */
142 #define	DS_MODE_EXCLUSIVE	3	/* exclusive access, e.g. to destroy */
143 #define	DS_MODE_LEVELS		4
144 #define	DS_MODE_LEVEL(x)	((x) & (DS_MODE_LEVELS - 1))
145 #define	DS_MODE_READONLY	0x8
146 #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
147 #define	DS_MODE_INCONSISTENT	0x10
148 #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
149 
150 #define	DS_FIND_SNAPSHOTS	(1<<0)
151 #define	DS_FIND_CHILDREN	(1<<1)
152 
153 /*
154  * The maximum number of bytes that can be accessed as part of one
155  * operation, including metadata.
156  */
157 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
158 
159 /*
160  * Public routines to create, destroy, open, and close objsets.
161  */
162 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
163     objset_t **osp);
164 int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
165     objset_t **osp);
166 void dmu_objset_close(objset_t *os);
167 int dmu_objset_evict_dbufs(objset_t *os);
168 int dmu_objset_create(const char *name, dmu_objset_type_t type,
169     objset_t *clone_parent,
170     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
171 int dmu_objset_destroy(const char *name);
172 int dmu_snapshots_destroy(char *fsname, char *snapname);
173 int dmu_objset_rollback(const char *name);
174 int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
175 int dmu_objset_rename(const char *name, const char *newname,
176     boolean_t recursive);
177 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
178     int flags);
179 void dmu_objset_byteswap(void *buf, size_t size);
180 
181 typedef struct dmu_buf {
182 	uint64_t db_object;		/* object that this buffer is part of */
183 	uint64_t db_offset;		/* byte offset in this object */
184 	uint64_t db_size;		/* size of buffer in bytes */
185 	void *db_data;			/* data in buffer */
186 } dmu_buf_t;
187 
188 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
189 
190 /*
191  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
192  */
193 #define	DMU_POOL_DIRECTORY_OBJECT	1
194 #define	DMU_POOL_CONFIG			"config"
195 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
196 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
197 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
198 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
199 #define	DMU_POOL_SPARES			"spares"
200 #define	DMU_POOL_DEFLATE		"deflate"
201 #define	DMU_POOL_HISTORY		"history"
202 #define	DMU_POOL_PROPS			"pool_props"
203 
204 /*
205  * Allocate an object from this objset.  The range of object numbers
206  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
207  *
208  * The transaction must be assigned to a txg.  The newly allocated
209  * object will be "held" in the transaction (ie. you can modify the
210  * newly allocated object in this transaction).
211  *
212  * dmu_object_alloc() chooses an object and returns it in *objectp.
213  *
214  * dmu_object_claim() allocates a specific object number.  If that
215  * number is already allocated, it fails and returns EEXIST.
216  *
217  * Return 0 on success, or ENOSPC or EEXIST as specified above.
218  */
219 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
220     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
221 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
222     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
223 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
224     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
225 
226 /*
227  * Free an object from this objset.
228  *
229  * The object's data will be freed as well (ie. you don't need to call
230  * dmu_free(object, 0, -1, tx)).
231  *
232  * The object need not be held in the transaction.
233  *
234  * If there are any holds on this object's buffers (via dmu_buf_hold()),
235  * or tx holds on the object (via dmu_tx_hold_object()), you can not
236  * free it; it fails and returns EBUSY.
237  *
238  * If the object is not allocated, it fails and returns ENOENT.
239  *
240  * Return 0 on success, or EBUSY or ENOENT as specified above.
241  */
242 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
243 
244 /*
245  * Find the next allocated or free object.
246  *
247  * The objectp parameter is in-out.  It will be updated to be the next
248  * object which is allocated.  Ignore objects which have not been
249  * modified since txg.
250  *
251  * XXX Can only be called on a objset with no dirty data.
252  *
253  * Returns 0 on success, or ENOENT if there are no more objects.
254  */
255 int dmu_object_next(objset_t *os, uint64_t *objectp,
256     boolean_t hole, uint64_t txg);
257 
258 /*
259  * Set the data blocksize for an object.
260  *
261  * The object cannot have any blocks allcated beyond the first.  If
262  * the first block is allocated already, the new size must be greater
263  * than the current block size.  If these conditions are not met,
264  * ENOTSUP will be returned.
265  *
266  * Returns 0 on success, or EBUSY if there are any holds on the object
267  * contents, or ENOTSUP as described above.
268  */
269 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
270     int ibs, dmu_tx_t *tx);
271 
272 /*
273  * Set the checksum property on a dnode.  The new checksum algorithm will
274  * apply to all newly written blocks; existing blocks will not be affected.
275  */
276 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
277     dmu_tx_t *tx);
278 
279 /*
280  * Set the compress property on a dnode.  The new compression algorithm will
281  * apply to all newly written blocks; existing blocks will not be affected.
282  */
283 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
284     dmu_tx_t *tx);
285 
286 /*
287  * Decide how many copies of a given block we should make.  Can be from
288  * 1 to SPA_DVAS_PER_BP.
289  */
290 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
291     dmu_object_type_t ot);
292 /*
293  * The bonus data is accessed more or less like a regular buffer.
294  * You must dmu_bonus_hold() to get the buffer, which will give you a
295  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
296  * data.  As with any normal buffer, you must call dmu_buf_read() to
297  * read db_data, dmu_buf_will_dirty() before modifying it, and the
298  * object must be held in an assigned transaction before calling
299  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
300  * buffer as well.  You must release your hold with dmu_buf_rele().
301  */
302 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
303 int dmu_bonus_max(void);
304 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
305 
306 /*
307  * Obtain the DMU buffer from the specified object which contains the
308  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
309  * that it will remain in memory.  You must release the hold with
310  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
311  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
312  *
313  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
314  * on the returned buffer before reading or writing the buffer's
315  * db_data.  The comments for those routines describe what particular
316  * operations are valid after calling them.
317  *
318  * The object number must be a valid, allocated object number.
319  */
320 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
321     void *tag, dmu_buf_t **);
322 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
323 void dmu_buf_rele(dmu_buf_t *db, void *tag);
324 uint64_t dmu_buf_refcount(dmu_buf_t *db);
325 
326 /*
327  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
328  * range of an object.  A pointer to an array of dmu_buf_t*'s is
329  * returned (in *dbpp).
330  *
331  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
332  * frees the array.  The hold on the array of buffers MUST be released
333  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
334  * individually with dmu_buf_rele.
335  */
336 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
337     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
338 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
339 
340 /*
341  * Returns NULL on success, or the existing user ptr if it's already
342  * been set.
343  *
344  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
345  *
346  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
347  * will be set to db->db_data when you are allowed to access it.  Note
348  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
349  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
350  * *user_data_ptr_ptr will be set to the new value when it changes.
351  *
352  * If non-NULL, pageout func will be called when this buffer is being
353  * excised from the cache, so that you can clean up the data structure
354  * pointed to by user_ptr.
355  *
356  * dmu_evict_user() will call the pageout func for all buffers in a
357  * objset with a given pageout func.
358  */
359 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
360     dmu_buf_evict_func_t *pageout_func);
361 /*
362  * set_user_ie is the same as set_user, but request immediate eviction
363  * when hold count goes to zero.
364  */
365 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
366     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
367 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
368     void *user_ptr, void *user_data_ptr_ptr,
369     dmu_buf_evict_func_t *pageout_func);
370 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
371 
372 /*
373  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
374  */
375 void *dmu_buf_get_user(dmu_buf_t *db);
376 
377 /*
378  * Indicate that you are going to modify the buffer's data (db_data).
379  *
380  * The transaction (tx) must be assigned to a txg (ie. you've called
381  * dmu_tx_assign()).  The buffer's object must be held in the tx
382  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
383  */
384 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
385 
386 /*
387  * You must create a transaction, then hold the objects which you will
388  * (or might) modify as part of this transaction.  Then you must assign
389  * the transaction to a transaction group.  Once the transaction has
390  * been assigned, you can modify buffers which belong to held objects as
391  * part of this transaction.  You can't modify buffers before the
392  * transaction has been assigned; you can't modify buffers which don't
393  * belong to objects which this transaction holds; you can't hold
394  * objects once the transaction has been assigned.  You may hold an
395  * object which you are going to free (with dmu_object_free()), but you
396  * don't have to.
397  *
398  * You can abort the transaction before it has been assigned.
399  *
400  * Note that you may hold buffers (with dmu_buf_hold) at any time,
401  * regardless of transaction state.
402  */
403 
404 #define	DMU_NEW_OBJECT	(-1ULL)
405 #define	DMU_OBJECT_END	(-1ULL)
406 
407 dmu_tx_t *dmu_tx_create(objset_t *os);
408 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
409 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
410     uint64_t len);
411 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
412 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
413 void dmu_tx_abort(dmu_tx_t *tx);
414 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
415 void dmu_tx_wait(dmu_tx_t *tx);
416 void dmu_tx_commit(dmu_tx_t *tx);
417 
418 /*
419  * Free up the data blocks for a defined range of a file.  If size is
420  * zero, the range from offset to end-of-file is freed.
421  */
422 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
423 	uint64_t size, dmu_tx_t *tx);
424 
425 /*
426  * Convenience functions.
427  *
428  * Canfail routines will return 0 on success, or an errno if there is a
429  * nonrecoverable I/O error.
430  */
431 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
432 	void *buf);
433 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
434 	const void *buf, dmu_tx_t *tx);
435 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
436 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
437     dmu_tx_t *tx);
438 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
439     uint64_t size, struct page *pp, dmu_tx_t *tx);
440 
441 extern int zfs_prefetch_disable;
442 
443 /*
444  * Asynchronously try to read in the data.
445  */
446 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
447     uint64_t len);
448 
449 typedef struct dmu_object_info {
450 	/* All sizes are in bytes. */
451 	uint32_t doi_data_block_size;
452 	uint32_t doi_metadata_block_size;
453 	uint64_t doi_bonus_size;
454 	dmu_object_type_t doi_type;
455 	dmu_object_type_t doi_bonus_type;
456 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
457 	uint8_t doi_checksum;
458 	uint8_t doi_compress;
459 	uint8_t doi_pad[5];
460 	/* Values below are number of 512-byte blocks. */
461 	uint64_t doi_physical_blks;		/* data + metadata */
462 	uint64_t doi_max_block_offset;
463 } dmu_object_info_t;
464 
465 typedef void arc_byteswap_func_t(void *buf, size_t size);
466 
467 typedef struct dmu_object_type_info {
468 	arc_byteswap_func_t	*ot_byteswap;
469 	boolean_t		ot_metadata;
470 	char			*ot_name;
471 } dmu_object_type_info_t;
472 
473 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
474 
475 /*
476  * Get information on a DMU object.
477  *
478  * Return 0 on success or ENOENT if object is not allocated.
479  *
480  * If doi is NULL, just indicates whether the object exists.
481  */
482 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
483 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
484 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
485 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
486     u_longlong_t *nblk512);
487 
488 typedef struct dmu_objset_stats {
489 	uint64_t dds_num_clones; /* number of clones of this */
490 	uint64_t dds_creation_txg;
491 	uint64_t dds_guid;
492 	dmu_objset_type_t dds_type;
493 	uint8_t dds_is_snapshot;
494 	uint8_t dds_inconsistent;
495 	char dds_origin[MAXNAMELEN];
496 } dmu_objset_stats_t;
497 
498 /*
499  * Get stats on a dataset.
500  */
501 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
502 
503 /*
504  * Add entries to the nvlist for all the objset's properties.  See
505  * zfs_prop_table[] and zfs(1m) for details on the properties.
506  */
507 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
508 
509 /*
510  * Get the space usage statistics for statvfs().
511  *
512  * refdbytes is the amount of space "referenced" by this objset.
513  * availbytes is the amount of space available to this objset, taking
514  * into account quotas & reservations, assuming that no other objsets
515  * use the space first.  These values correspond to the 'referenced' and
516  * 'available' properties, described in the zfs(1m) manpage.
517  *
518  * usedobjs and availobjs are the number of objects currently allocated,
519  * and available.
520  */
521 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
522     uint64_t *usedobjsp, uint64_t *availobjsp);
523 
524 /*
525  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
526  * (Contrast with the ds_guid which is a 64-bit ID that will never
527  * change, so there is a small probability that it will collide.)
528  */
529 uint64_t dmu_objset_fsid_guid(objset_t *os);
530 
531 int dmu_objset_is_snapshot(objset_t *os);
532 
533 extern struct spa *dmu_objset_spa(objset_t *os);
534 extern struct zilog *dmu_objset_zil(objset_t *os);
535 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
536 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
537 extern void dmu_objset_name(objset_t *os, char *buf);
538 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
539 extern uint64_t dmu_objset_id(objset_t *os);
540 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
541     uint64_t *id, uint64_t *offp);
542 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
543     uint64_t *idp, uint64_t *offp);
544 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
545 extern void *dmu_objset_get_user(objset_t *os);
546 
547 /*
548  * Return the txg number for the given assigned transaction.
549  */
550 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
551 
552 /*
553  * Synchronous write.
554  * If a parent zio is provided this function initiates a write on the
555  * provided buffer as a child of the parent zio.
556  * In the absence of a parent zio, the write is completed synchronously.
557  * At write completion, blk is filled with the bp of the written block.
558  * Note that while the data covered by this function will be on stable
559  * storage when the write completes this new data does not become a
560  * permanent part of the file until the associated transaction commits.
561  */
562 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
563 int dmu_sync(struct zio *zio, dmu_buf_t *db,
564     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
565 
566 /*
567  * Find the next hole or data block in file starting at *off
568  * Return found offset in *off. Return ESRCH for end of file.
569  */
570 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
571     uint64_t *off);
572 
573 /*
574  * Initial setup and final teardown.
575  */
576 extern void dmu_init(void);
577 extern void dmu_fini(void);
578 
579 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
580     uint64_t object, uint64_t offset, int len);
581 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
582     dmu_traverse_cb_t cb, void *arg);
583 
584 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
585     struct vnode *vp, offset_t *off);
586 
587 typedef struct dmu_recv_cookie {
588 	/*
589 	 * This structure is opaque!
590 	 *
591 	 * If logical and real are different, we are recving the stream
592 	 * into the "real" temporary clone, and then switching it with
593 	 * the "logical" target.
594 	 */
595 	struct dsl_dataset *drc_logical_ds;
596 	struct dsl_dataset *drc_real_ds;
597 	struct drr_begin *drc_drrb;
598 	char *drc_tosnap;
599 	boolean_t drc_newfs;
600 	boolean_t drc_force;
601 } dmu_recv_cookie_t;
602 
603 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
604     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
605 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
606 int dmu_recv_end(dmu_recv_cookie_t *drc);
607 
608 /* CRC64 table */
609 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
610 extern uint64_t zfs_crc64_table[256];
611 
612 #ifdef	__cplusplus
613 }
614 #endif
615 
616 #endif	/* _SYS_DMU_H */
617