xref: /titanic_50/usr/src/uts/common/fs/zfs/sys/dmu.h (revision a3c4695861e3f0a8d3706f77ccd53683cca48d67)
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(objset_t *os);
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 #define	DMU_POOL_L2CACHE		"l2cache"
204 
205 /*
206  * Allocate an object from this objset.  The range of object numbers
207  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
208  *
209  * The transaction must be assigned to a txg.  The newly allocated
210  * object will be "held" in the transaction (ie. you can modify the
211  * newly allocated object in this transaction).
212  *
213  * dmu_object_alloc() chooses an object and returns it in *objectp.
214  *
215  * dmu_object_claim() allocates a specific object number.  If that
216  * number is already allocated, it fails and returns EEXIST.
217  *
218  * Return 0 on success, or ENOSPC or EEXIST as specified above.
219  */
220 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
221     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
222 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
223     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
224 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
225     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
226 
227 /*
228  * Free an object from this objset.
229  *
230  * The object's data will be freed as well (ie. you don't need to call
231  * dmu_free(object, 0, -1, tx)).
232  *
233  * The object need not be held in the transaction.
234  *
235  * If there are any holds on this object's buffers (via dmu_buf_hold()),
236  * or tx holds on the object (via dmu_tx_hold_object()), you can not
237  * free it; it fails and returns EBUSY.
238  *
239  * If the object is not allocated, it fails and returns ENOENT.
240  *
241  * Return 0 on success, or EBUSY or ENOENT as specified above.
242  */
243 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
244 
245 /*
246  * Find the next allocated or free object.
247  *
248  * The objectp parameter is in-out.  It will be updated to be the next
249  * object which is allocated.  Ignore objects which have not been
250  * modified since txg.
251  *
252  * XXX Can only be called on a objset with no dirty data.
253  *
254  * Returns 0 on success, or ENOENT if there are no more objects.
255  */
256 int dmu_object_next(objset_t *os, uint64_t *objectp,
257     boolean_t hole, uint64_t txg);
258 
259 /*
260  * Set the data blocksize for an object.
261  *
262  * The object cannot have any blocks allcated beyond the first.  If
263  * the first block is allocated already, the new size must be greater
264  * than the current block size.  If these conditions are not met,
265  * ENOTSUP will be returned.
266  *
267  * Returns 0 on success, or EBUSY if there are any holds on the object
268  * contents, or ENOTSUP as described above.
269  */
270 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
271     int ibs, dmu_tx_t *tx);
272 
273 /*
274  * Set the checksum property on a dnode.  The new checksum algorithm will
275  * apply to all newly written blocks; existing blocks will not be affected.
276  */
277 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
278     dmu_tx_t *tx);
279 
280 /*
281  * Set the compress property on a dnode.  The new compression algorithm will
282  * apply to all newly written blocks; existing blocks will not be affected.
283  */
284 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
285     dmu_tx_t *tx);
286 
287 /*
288  * Decide how many copies of a given block we should make.  Can be from
289  * 1 to SPA_DVAS_PER_BP.
290  */
291 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
292     dmu_object_type_t ot);
293 /*
294  * The bonus data is accessed more or less like a regular buffer.
295  * You must dmu_bonus_hold() to get the buffer, which will give you a
296  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
297  * data.  As with any normal buffer, you must call dmu_buf_read() to
298  * read db_data, dmu_buf_will_dirty() before modifying it, and the
299  * object must be held in an assigned transaction before calling
300  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
301  * buffer as well.  You must release your hold with dmu_buf_rele().
302  */
303 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
304 int dmu_bonus_max(void);
305 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
306 
307 /*
308  * Obtain the DMU buffer from the specified object which contains the
309  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
310  * that it will remain in memory.  You must release the hold with
311  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
312  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
313  *
314  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
315  * on the returned buffer before reading or writing the buffer's
316  * db_data.  The comments for those routines describe what particular
317  * operations are valid after calling them.
318  *
319  * The object number must be a valid, allocated object number.
320  */
321 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
322     void *tag, dmu_buf_t **);
323 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
324 void dmu_buf_rele(dmu_buf_t *db, void *tag);
325 uint64_t dmu_buf_refcount(dmu_buf_t *db);
326 
327 /*
328  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
329  * range of an object.  A pointer to an array of dmu_buf_t*'s is
330  * returned (in *dbpp).
331  *
332  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
333  * frees the array.  The hold on the array of buffers MUST be released
334  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
335  * individually with dmu_buf_rele.
336  */
337 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
338     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
339 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
340 
341 /*
342  * Returns NULL on success, or the existing user ptr if it's already
343  * been set.
344  *
345  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
346  *
347  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
348  * will be set to db->db_data when you are allowed to access it.  Note
349  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
350  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
351  * *user_data_ptr_ptr will be set to the new value when it changes.
352  *
353  * If non-NULL, pageout func will be called when this buffer is being
354  * excised from the cache, so that you can clean up the data structure
355  * pointed to by user_ptr.
356  *
357  * dmu_evict_user() will call the pageout func for all buffers in a
358  * objset with a given pageout func.
359  */
360 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
361     dmu_buf_evict_func_t *pageout_func);
362 /*
363  * set_user_ie is the same as set_user, but request immediate eviction
364  * when hold count goes to zero.
365  */
366 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
367     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
368 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
369     void *user_ptr, void *user_data_ptr_ptr,
370     dmu_buf_evict_func_t *pageout_func);
371 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
372 
373 /*
374  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
375  */
376 void *dmu_buf_get_user(dmu_buf_t *db);
377 
378 /*
379  * Indicate that you are going to modify the buffer's data (db_data).
380  *
381  * The transaction (tx) must be assigned to a txg (ie. you've called
382  * dmu_tx_assign()).  The buffer's object must be held in the tx
383  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
384  */
385 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
386 
387 /*
388  * You must create a transaction, then hold the objects which you will
389  * (or might) modify as part of this transaction.  Then you must assign
390  * the transaction to a transaction group.  Once the transaction has
391  * been assigned, you can modify buffers which belong to held objects as
392  * part of this transaction.  You can't modify buffers before the
393  * transaction has been assigned; you can't modify buffers which don't
394  * belong to objects which this transaction holds; you can't hold
395  * objects once the transaction has been assigned.  You may hold an
396  * object which you are going to free (with dmu_object_free()), but you
397  * don't have to.
398  *
399  * You can abort the transaction before it has been assigned.
400  *
401  * Note that you may hold buffers (with dmu_buf_hold) at any time,
402  * regardless of transaction state.
403  */
404 
405 #define	DMU_NEW_OBJECT	(-1ULL)
406 #define	DMU_OBJECT_END	(-1ULL)
407 
408 dmu_tx_t *dmu_tx_create(objset_t *os);
409 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
410 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
411     uint64_t len);
412 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
413 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
414 void dmu_tx_abort(dmu_tx_t *tx);
415 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
416 void dmu_tx_wait(dmu_tx_t *tx);
417 void dmu_tx_commit(dmu_tx_t *tx);
418 
419 /*
420  * Free up the data blocks for a defined range of a file.  If size is
421  * zero, the range from offset to end-of-file is freed.
422  */
423 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
424 	uint64_t size, dmu_tx_t *tx);
425 
426 /*
427  * Convenience functions.
428  *
429  * Canfail routines will return 0 on success, or an errno if there is a
430  * nonrecoverable I/O error.
431  */
432 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
433 	void *buf);
434 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
435 	const void *buf, dmu_tx_t *tx);
436 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
437 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
438     dmu_tx_t *tx);
439 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
440     uint64_t size, struct page *pp, dmu_tx_t *tx);
441 
442 extern int zfs_prefetch_disable;
443 
444 /*
445  * Asynchronously try to read in the data.
446  */
447 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
448     uint64_t len);
449 
450 typedef struct dmu_object_info {
451 	/* All sizes are in bytes. */
452 	uint32_t doi_data_block_size;
453 	uint32_t doi_metadata_block_size;
454 	uint64_t doi_bonus_size;
455 	dmu_object_type_t doi_type;
456 	dmu_object_type_t doi_bonus_type;
457 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
458 	uint8_t doi_checksum;
459 	uint8_t doi_compress;
460 	uint8_t doi_pad[5];
461 	/* Values below are number of 512-byte blocks. */
462 	uint64_t doi_physical_blks;		/* data + metadata */
463 	uint64_t doi_max_block_offset;
464 } dmu_object_info_t;
465 
466 typedef void arc_byteswap_func_t(void *buf, size_t size);
467 
468 typedef struct dmu_object_type_info {
469 	arc_byteswap_func_t	*ot_byteswap;
470 	boolean_t		ot_metadata;
471 	char			*ot_name;
472 } dmu_object_type_info_t;
473 
474 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
475 
476 /*
477  * Get information on a DMU object.
478  *
479  * Return 0 on success or ENOENT if object is not allocated.
480  *
481  * If doi is NULL, just indicates whether the object exists.
482  */
483 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
484 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
485 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
486 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
487     u_longlong_t *nblk512);
488 
489 typedef struct dmu_objset_stats {
490 	uint64_t dds_num_clones; /* number of clones of this */
491 	uint64_t dds_creation_txg;
492 	uint64_t dds_guid;
493 	dmu_objset_type_t dds_type;
494 	uint8_t dds_is_snapshot;
495 	uint8_t dds_inconsistent;
496 	char dds_origin[MAXNAMELEN];
497 } dmu_objset_stats_t;
498 
499 /*
500  * Get stats on a dataset.
501  */
502 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
503 
504 /*
505  * Add entries to the nvlist for all the objset's properties.  See
506  * zfs_prop_table[] and zfs(1m) for details on the properties.
507  */
508 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
509 
510 /*
511  * Get the space usage statistics for statvfs().
512  *
513  * refdbytes is the amount of space "referenced" by this objset.
514  * availbytes is the amount of space available to this objset, taking
515  * into account quotas & reservations, assuming that no other objsets
516  * use the space first.  These values correspond to the 'referenced' and
517  * 'available' properties, described in the zfs(1m) manpage.
518  *
519  * usedobjs and availobjs are the number of objects currently allocated,
520  * and available.
521  */
522 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
523     uint64_t *usedobjsp, uint64_t *availobjsp);
524 
525 /*
526  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
527  * (Contrast with the ds_guid which is a 64-bit ID that will never
528  * change, so there is a small probability that it will collide.)
529  */
530 uint64_t dmu_objset_fsid_guid(objset_t *os);
531 
532 int dmu_objset_is_snapshot(objset_t *os);
533 
534 extern struct spa *dmu_objset_spa(objset_t *os);
535 extern struct zilog *dmu_objset_zil(objset_t *os);
536 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
537 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
538 extern void dmu_objset_name(objset_t *os, char *buf);
539 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
540 extern uint64_t dmu_objset_id(objset_t *os);
541 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
542     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
543 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
544     uint64_t *idp, uint64_t *offp);
545 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
546 extern void *dmu_objset_get_user(objset_t *os);
547 
548 /*
549  * Return the txg number for the given assigned transaction.
550  */
551 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
552 
553 /*
554  * Synchronous write.
555  * If a parent zio is provided this function initiates a write on the
556  * provided buffer as a child of the parent zio.
557  * In the absence of a parent zio, the write is completed synchronously.
558  * At write completion, blk is filled with the bp of the written block.
559  * Note that while the data covered by this function will be on stable
560  * storage when the write completes this new data does not become a
561  * permanent part of the file until the associated transaction commits.
562  */
563 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
564 int dmu_sync(struct zio *zio, dmu_buf_t *db,
565     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
566 
567 /*
568  * Find the next hole or data block in file starting at *off
569  * Return found offset in *off. Return ESRCH for end of file.
570  */
571 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
572     uint64_t *off);
573 
574 /*
575  * Initial setup and final teardown.
576  */
577 extern void dmu_init(void);
578 extern void dmu_fini(void);
579 
580 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
581     uint64_t object, uint64_t offset, int len);
582 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
583     dmu_traverse_cb_t cb, void *arg);
584 
585 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
586     struct vnode *vp, offset_t *off);
587 
588 typedef struct dmu_recv_cookie {
589 	/*
590 	 * This structure is opaque!
591 	 *
592 	 * If logical and real are different, we are recving the stream
593 	 * into the "real" temporary clone, and then switching it with
594 	 * the "logical" target.
595 	 */
596 	struct dsl_dataset *drc_logical_ds;
597 	struct dsl_dataset *drc_real_ds;
598 	struct drr_begin *drc_drrb;
599 	char *drc_tosnap;
600 	boolean_t drc_newfs;
601 	boolean_t drc_force;
602 } dmu_recv_cookie_t;
603 
604 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
605     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
606 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
607 int dmu_recv_end(dmu_recv_cookie_t *drc);
608 
609 /* CRC64 table */
610 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
611 extern uint64_t zfs_crc64_table[256];
612 
613 #ifdef	__cplusplus
614 }
615 #endif
616 
617 #endif	/* _SYS_DMU_H */
618