xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 9966ca11f4a1481acce85f690fa59e4084050627)
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 2009 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 /*
30  * This file describes the interface that the DMU provides for its
31  * consumers.
32  *
33  * The DMU also interacts with the SPA.  That interface is described in
34  * dmu_spa.h.
35  */
36 
37 #include <sys/inttypes.h>
38 #include <sys/types.h>
39 #include <sys/param.h>
40 #include <sys/cred.h>
41 #include <sys/time.h>
42 
43 #ifdef	__cplusplus
44 extern "C" {
45 #endif
46 
47 struct uio;
48 struct page;
49 struct vnode;
50 struct spa;
51 struct zilog;
52 struct zio;
53 struct blkptr;
54 struct zap_cursor;
55 struct dsl_dataset;
56 struct dsl_pool;
57 struct dnode;
58 struct drr_begin;
59 struct drr_end;
60 struct zbookmark;
61 struct spa;
62 struct nvlist;
63 struct arc_buf;
64 
65 typedef struct objset objset_t;
66 typedef struct dmu_tx dmu_tx_t;
67 typedef struct dsl_dir dsl_dir_t;
68 
69 typedef enum dmu_object_type {
70 	DMU_OT_NONE,
71 	/* general: */
72 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
73 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
74 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
75 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
76 	DMU_OT_BPLIST,			/* UINT64 */
77 	DMU_OT_BPLIST_HDR,		/* UINT64 */
78 	/* spa: */
79 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
80 	DMU_OT_SPACE_MAP,		/* UINT64 */
81 	/* zil: */
82 	DMU_OT_INTENT_LOG,		/* UINT64 */
83 	/* dmu: */
84 	DMU_OT_DNODE,			/* DNODE */
85 	DMU_OT_OBJSET,			/* OBJSET */
86 	/* dsl: */
87 	DMU_OT_DSL_DIR,			/* UINT64 */
88 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
89 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
90 	DMU_OT_DSL_PROPS,		/* ZAP */
91 	DMU_OT_DSL_DATASET,		/* UINT64 */
92 	/* zpl: */
93 	DMU_OT_ZNODE,			/* ZNODE */
94 	DMU_OT_OLDACL,			/* Old ACL */
95 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
96 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
97 	DMU_OT_MASTER_NODE,		/* ZAP */
98 	DMU_OT_UNLINKED_SET,		/* ZAP */
99 	/* zvol: */
100 	DMU_OT_ZVOL,			/* UINT8 */
101 	DMU_OT_ZVOL_PROP,		/* ZAP */
102 	/* other; for testing only! */
103 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
104 	DMU_OT_UINT64_OTHER,		/* UINT64 */
105 	DMU_OT_ZAP_OTHER,		/* ZAP */
106 	/* new object types: */
107 	DMU_OT_ERROR_LOG,		/* ZAP */
108 	DMU_OT_SPA_HISTORY,		/* UINT8 */
109 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
110 	DMU_OT_POOL_PROPS,		/* ZAP */
111 	DMU_OT_DSL_PERMS,		/* ZAP */
112 	DMU_OT_ACL,			/* ACL */
113 	DMU_OT_SYSACL,			/* SYSACL */
114 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
115 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
116 	DMU_OT_NEXT_CLONES,		/* ZAP */
117 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
118 	DMU_OT_USERGROUP_USED,		/* ZAP */
119 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
120 	DMU_OT_USERREFS,		/* ZAP */
121 	DMU_OT_NUMTYPES
122 } dmu_object_type_t;
123 
124 typedef enum dmu_objset_type {
125 	DMU_OST_NONE,
126 	DMU_OST_META,
127 	DMU_OST_ZFS,
128 	DMU_OST_ZVOL,
129 	DMU_OST_OTHER,			/* For testing only! */
130 	DMU_OST_ANY,			/* Be careful! */
131 	DMU_OST_NUMTYPES
132 } dmu_objset_type_t;
133 
134 void byteswap_uint64_array(void *buf, size_t size);
135 void byteswap_uint32_array(void *buf, size_t size);
136 void byteswap_uint16_array(void *buf, size_t size);
137 void byteswap_uint8_array(void *buf, size_t size);
138 void zap_byteswap(void *buf, size_t size);
139 void zfs_oldacl_byteswap(void *buf, size_t size);
140 void zfs_acl_byteswap(void *buf, size_t size);
141 void zfs_znode_byteswap(void *buf, size_t size);
142 
143 #define	DS_FIND_SNAPSHOTS	(1<<0)
144 #define	DS_FIND_CHILDREN	(1<<1)
145 
146 /*
147  * The maximum number of bytes that can be accessed as part of one
148  * operation, including metadata.
149  */
150 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
151 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
152 
153 #define	DMU_USERUSED_OBJECT	(-1ULL)
154 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
155 
156 /*
157  * Public routines to create, destroy, open, and close objsets.
158  */
159 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
160 int dmu_objset_own(const char *name, dmu_objset_type_t type,
161     boolean_t readonly, void *tag, objset_t **osp);
162 void dmu_objset_rele(objset_t *os, void *tag);
163 void dmu_objset_disown(objset_t *os, void *tag);
164 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
165 
166 int dmu_objset_evict_dbufs(objset_t *os);
167 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
168     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
169 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
170     uint64_t flags);
171 int dmu_objset_destroy(const char *name, boolean_t defer);
172 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
173 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
174     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 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
205 
206 /* 4x8 zbookmark_t */
207 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
208 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
209 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
210 /* 1x8 txg */
211 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
212 /* 1x8 txg */
213 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
214 /* 1x4 enum scrub_func */
215 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
216 /* 1x8 count */
217 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
218 
219 /*
220  * Allocate an object from this objset.  The range of object numbers
221  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
222  *
223  * The transaction must be assigned to a txg.  The newly allocated
224  * object will be "held" in the transaction (ie. you can modify the
225  * newly allocated object in this transaction).
226  *
227  * dmu_object_alloc() chooses an object and returns it in *objectp.
228  *
229  * dmu_object_claim() allocates a specific object number.  If that
230  * number is already allocated, it fails and returns EEXIST.
231  *
232  * Return 0 on success, or ENOSPC or EEXIST as specified above.
233  */
234 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
235     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
236 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
237     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
238 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
239     int blocksize, dmu_object_type_t bonustype, int bonuslen);
240 
241 /*
242  * Free an object from this objset.
243  *
244  * The object's data will be freed as well (ie. you don't need to call
245  * dmu_free(object, 0, -1, tx)).
246  *
247  * The object need not be held in the transaction.
248  *
249  * If there are any holds on this object's buffers (via dmu_buf_hold()),
250  * or tx holds on the object (via dmu_tx_hold_object()), you can not
251  * free it; it fails and returns EBUSY.
252  *
253  * If the object is not allocated, it fails and returns ENOENT.
254  *
255  * Return 0 on success, or EBUSY or ENOENT as specified above.
256  */
257 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
258 
259 /*
260  * Find the next allocated or free object.
261  *
262  * The objectp parameter is in-out.  It will be updated to be the next
263  * object which is allocated.  Ignore objects which have not been
264  * modified since txg.
265  *
266  * XXX Can only be called on a objset with no dirty data.
267  *
268  * Returns 0 on success, or ENOENT if there are no more objects.
269  */
270 int dmu_object_next(objset_t *os, uint64_t *objectp,
271     boolean_t hole, uint64_t txg);
272 
273 /*
274  * Set the data blocksize for an object.
275  *
276  * The object cannot have any blocks allcated beyond the first.  If
277  * the first block is allocated already, the new size must be greater
278  * than the current block size.  If these conditions are not met,
279  * ENOTSUP will be returned.
280  *
281  * Returns 0 on success, or EBUSY if there are any holds on the object
282  * contents, or ENOTSUP as described above.
283  */
284 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
285     int ibs, dmu_tx_t *tx);
286 
287 /*
288  * Set the checksum property on a dnode.  The new checksum algorithm will
289  * apply to all newly written blocks; existing blocks will not be affected.
290  */
291 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
292     dmu_tx_t *tx);
293 
294 /*
295  * Set the compress property on a dnode.  The new compression algorithm will
296  * apply to all newly written blocks; existing blocks will not be affected.
297  */
298 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
299     dmu_tx_t *tx);
300 
301 /*
302  * Decide how many copies of a given block we should make.  Can be from
303  * 1 to SPA_DVAS_PER_BP.
304  */
305 int dmu_get_replication_level(objset_t *os, struct zbookmark *zb,
306     dmu_object_type_t ot);
307 /*
308  * The bonus data is accessed more or less like a regular buffer.
309  * You must dmu_bonus_hold() to get the buffer, which will give you a
310  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
311  * data.  As with any normal buffer, you must call dmu_buf_read() to
312  * read db_data, dmu_buf_will_dirty() before modifying it, and the
313  * object must be held in an assigned transaction before calling
314  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
315  * buffer as well.  You must release your hold with dmu_buf_rele().
316  */
317 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
318 int dmu_bonus_max(void);
319 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
320 
321 /*
322  * Obtain the DMU buffer from the specified object which contains the
323  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
324  * that it will remain in memory.  You must release the hold with
325  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
326  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
327  *
328  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
329  * on the returned buffer before reading or writing the buffer's
330  * db_data.  The comments for those routines describe what particular
331  * operations are valid after calling them.
332  *
333  * The object number must be a valid, allocated object number.
334  */
335 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
336     void *tag, dmu_buf_t **);
337 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
338 void dmu_buf_rele(dmu_buf_t *db, void *tag);
339 uint64_t dmu_buf_refcount(dmu_buf_t *db);
340 
341 /*
342  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
343  * range of an object.  A pointer to an array of dmu_buf_t*'s is
344  * returned (in *dbpp).
345  *
346  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
347  * frees the array.  The hold on the array of buffers MUST be released
348  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
349  * individually with dmu_buf_rele.
350  */
351 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
352     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
353 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
354 
355 /*
356  * Returns NULL on success, or the existing user ptr if it's already
357  * been set.
358  *
359  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
360  *
361  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
362  * will be set to db->db_data when you are allowed to access it.  Note
363  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
364  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
365  * *user_data_ptr_ptr will be set to the new value when it changes.
366  *
367  * If non-NULL, pageout func will be called when this buffer is being
368  * excised from the cache, so that you can clean up the data structure
369  * pointed to by user_ptr.
370  *
371  * dmu_evict_user() will call the pageout func for all buffers in a
372  * objset with a given pageout func.
373  */
374 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
375     dmu_buf_evict_func_t *pageout_func);
376 /*
377  * set_user_ie is the same as set_user, but request immediate eviction
378  * when hold count goes to zero.
379  */
380 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
381     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
382 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
383     void *user_ptr, void *user_data_ptr_ptr,
384     dmu_buf_evict_func_t *pageout_func);
385 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
386 
387 /*
388  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
389  */
390 void *dmu_buf_get_user(dmu_buf_t *db);
391 
392 /*
393  * Indicate that you are going to modify the buffer's data (db_data).
394  *
395  * The transaction (tx) must be assigned to a txg (ie. you've called
396  * dmu_tx_assign()).  The buffer's object must be held in the tx
397  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
398  */
399 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
400 
401 /*
402  * Tells if the given dbuf is freeable.
403  */
404 boolean_t dmu_buf_freeable(dmu_buf_t *);
405 
406 /*
407  * You must create a transaction, then hold the objects which you will
408  * (or might) modify as part of this transaction.  Then you must assign
409  * the transaction to a transaction group.  Once the transaction has
410  * been assigned, you can modify buffers which belong to held objects as
411  * part of this transaction.  You can't modify buffers before the
412  * transaction has been assigned; you can't modify buffers which don't
413  * belong to objects which this transaction holds; you can't hold
414  * objects once the transaction has been assigned.  You may hold an
415  * object which you are going to free (with dmu_object_free()), but you
416  * don't have to.
417  *
418  * You can abort the transaction before it has been assigned.
419  *
420  * Note that you may hold buffers (with dmu_buf_hold) at any time,
421  * regardless of transaction state.
422  */
423 
424 #define	DMU_NEW_OBJECT	(-1ULL)
425 #define	DMU_OBJECT_END	(-1ULL)
426 
427 dmu_tx_t *dmu_tx_create(objset_t *os);
428 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
429 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
430     uint64_t len);
431 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
432 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
433 void dmu_tx_abort(dmu_tx_t *tx);
434 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
435 void dmu_tx_wait(dmu_tx_t *tx);
436 void dmu_tx_commit(dmu_tx_t *tx);
437 
438 /*
439  * Free up the data blocks for a defined range of a file.  If size is
440  * zero, the range from offset to end-of-file is freed.
441  */
442 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
443 	uint64_t size, dmu_tx_t *tx);
444 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
445 	uint64_t size);
446 int dmu_free_object(objset_t *os, uint64_t object);
447 
448 /*
449  * Convenience functions.
450  *
451  * Canfail routines will return 0 on success, or an errno if there is a
452  * nonrecoverable I/O error.
453  */
454 #define	DMU_READ_PREFETCH	0 /* prefetch */
455 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
456 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
457 	void *buf, uint32_t flags);
458 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
459 	const void *buf, dmu_tx_t *tx);
460 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
461 	dmu_tx_t *tx);
462 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
463 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
464     dmu_tx_t *tx);
465 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
466     uint64_t size, struct page *pp, dmu_tx_t *tx);
467 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
468 void dmu_return_arcbuf(struct arc_buf *buf);
469 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
470     dmu_tx_t *tx);
471 
472 extern int zfs_prefetch_disable;
473 
474 /*
475  * Asynchronously try to read in the data.
476  */
477 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
478     uint64_t len);
479 
480 typedef struct dmu_object_info {
481 	/* All sizes are in bytes. */
482 	uint32_t doi_data_block_size;
483 	uint32_t doi_metadata_block_size;
484 	uint64_t doi_bonus_size;
485 	dmu_object_type_t doi_type;
486 	dmu_object_type_t doi_bonus_type;
487 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
488 	uint8_t doi_checksum;
489 	uint8_t doi_compress;
490 	uint8_t doi_pad[5];
491 	/* Values below are number of 512-byte blocks. */
492 	uint64_t doi_physical_blks;		/* data + metadata */
493 	uint64_t doi_max_block_offset;
494 } dmu_object_info_t;
495 
496 typedef void arc_byteswap_func_t(void *buf, size_t size);
497 
498 typedef struct dmu_object_type_info {
499 	arc_byteswap_func_t	*ot_byteswap;
500 	boolean_t		ot_metadata;
501 	char			*ot_name;
502 } dmu_object_type_info_t;
503 
504 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
505 
506 /*
507  * Get information on a DMU object.
508  *
509  * Return 0 on success or ENOENT if object is not allocated.
510  *
511  * If doi is NULL, just indicates whether the object exists.
512  */
513 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
514 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
515 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
516 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
517     u_longlong_t *nblk512);
518 
519 typedef struct dmu_objset_stats {
520 	uint64_t dds_num_clones; /* number of clones of this */
521 	uint64_t dds_creation_txg;
522 	uint64_t dds_guid;
523 	dmu_objset_type_t dds_type;
524 	uint8_t dds_is_snapshot;
525 	uint8_t dds_inconsistent;
526 	char dds_origin[MAXNAMELEN];
527 } dmu_objset_stats_t;
528 
529 /*
530  * Get stats on a dataset.
531  */
532 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
533 
534 /*
535  * Add entries to the nvlist for all the objset's properties.  See
536  * zfs_prop_table[] and zfs(1m) for details on the properties.
537  */
538 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
539 
540 /*
541  * Get the space usage statistics for statvfs().
542  *
543  * refdbytes is the amount of space "referenced" by this objset.
544  * availbytes is the amount of space available to this objset, taking
545  * into account quotas & reservations, assuming that no other objsets
546  * use the space first.  These values correspond to the 'referenced' and
547  * 'available' properties, described in the zfs(1m) manpage.
548  *
549  * usedobjs and availobjs are the number of objects currently allocated,
550  * and available.
551  */
552 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
553     uint64_t *usedobjsp, uint64_t *availobjsp);
554 
555 /*
556  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
557  * (Contrast with the ds_guid which is a 64-bit ID that will never
558  * change, so there is a small probability that it will collide.)
559  */
560 uint64_t dmu_objset_fsid_guid(objset_t *os);
561 
562 /*
563  * Get the [cm]time for an objset's snapshot dir
564  */
565 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
566 
567 int dmu_objset_is_snapshot(objset_t *os);
568 
569 extern struct spa *dmu_objset_spa(objset_t *os);
570 extern struct zilog *dmu_objset_zil(objset_t *os);
571 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
572 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
573 extern void dmu_objset_name(objset_t *os, char *buf);
574 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
575 extern uint64_t dmu_objset_id(objset_t *os);
576 extern uint64_t dmu_objset_logbias(objset_t *os);
577 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
578     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
579 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
580     int maxlen, boolean_t *conflict);
581 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
582     uint64_t *idp, uint64_t *offp);
583 
584 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
585     void *bonus, uint64_t *userp, uint64_t *groupp);
586 extern void dmu_objset_register_type(dmu_objset_type_t ost,
587     objset_used_cb_t *cb);
588 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
589 extern void *dmu_objset_get_user(objset_t *os);
590 
591 /*
592  * Return the txg number for the given assigned transaction.
593  */
594 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
595 
596 /*
597  * Synchronous write.
598  * If a parent zio is provided this function initiates a write on the
599  * provided buffer as a child of the parent zio.
600  * In the absence of a parent zio, the write is completed synchronously.
601  * At write completion, blk is filled with the bp of the written block.
602  * Note that while the data covered by this function will be on stable
603  * storage when the write completes this new data does not become a
604  * permanent part of the file until the associated transaction commits.
605  */
606 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
607 int dmu_sync(struct zio *zio, dmu_buf_t *db,
608     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
609 
610 /*
611  * Find the next hole or data block in file starting at *off
612  * Return found offset in *off. Return ESRCH for end of file.
613  */
614 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
615     uint64_t *off);
616 
617 /*
618  * Initial setup and final teardown.
619  */
620 extern void dmu_init(void);
621 extern void dmu_fini(void);
622 
623 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
624     uint64_t object, uint64_t offset, int len);
625 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
626     dmu_traverse_cb_t cb, void *arg);
627 
628 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
629     struct vnode *vp, offset_t *off);
630 
631 typedef struct dmu_recv_cookie {
632 	/*
633 	 * This structure is opaque!
634 	 *
635 	 * If logical and real are different, we are recving the stream
636 	 * into the "real" temporary clone, and then switching it with
637 	 * the "logical" target.
638 	 */
639 	struct dsl_dataset *drc_logical_ds;
640 	struct dsl_dataset *drc_real_ds;
641 	struct drr_begin *drc_drrb;
642 	char *drc_tosnap;
643 	boolean_t drc_newfs;
644 	boolean_t drc_force;
645 } dmu_recv_cookie_t;
646 
647 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
648     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
649 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
650 int dmu_recv_end(dmu_recv_cookie_t *drc);
651 
652 /* CRC64 table */
653 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
654 extern uint64_t zfs_crc64_table[256];
655 
656 #ifdef	__cplusplus
657 }
658 #endif
659 
660 #endif	/* _SYS_DMU_H */
661