xref: /titanic_41/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 75ce41a57ff334bd8fe2cb9ed51eea835892f944)
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 
42 #ifdef	__cplusplus
43 extern "C" {
44 #endif
45 
46 struct uio;
47 struct page;
48 struct vnode;
49 struct spa;
50 struct zilog;
51 struct zio;
52 struct blkptr;
53 struct zap_cursor;
54 struct dsl_dataset;
55 struct dsl_pool;
56 struct dnode;
57 struct drr_begin;
58 struct drr_end;
59 struct zbookmark;
60 struct spa;
61 struct nvlist;
62 struct arc_buf;
63 
64 typedef struct objset objset_t;
65 typedef struct dmu_tx dmu_tx_t;
66 typedef struct dsl_dir dsl_dir_t;
67 
68 typedef enum dmu_object_type {
69 	DMU_OT_NONE,
70 	/* general: */
71 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
72 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
73 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
74 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
75 	DMU_OT_BPLIST,			/* UINT64 */
76 	DMU_OT_BPLIST_HDR,		/* UINT64 */
77 	/* spa: */
78 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
79 	DMU_OT_SPACE_MAP,		/* UINT64 */
80 	/* zil: */
81 	DMU_OT_INTENT_LOG,		/* UINT64 */
82 	/* dmu: */
83 	DMU_OT_DNODE,			/* DNODE */
84 	DMU_OT_OBJSET,			/* OBJSET */
85 	/* dsl: */
86 	DMU_OT_DSL_DIR,			/* UINT64 */
87 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
88 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
89 	DMU_OT_DSL_PROPS,		/* ZAP */
90 	DMU_OT_DSL_DATASET,		/* UINT64 */
91 	/* zpl: */
92 	DMU_OT_ZNODE,			/* ZNODE */
93 	DMU_OT_OLDACL,			/* Old ACL */
94 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
95 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
96 	DMU_OT_MASTER_NODE,		/* ZAP */
97 	DMU_OT_UNLINKED_SET,		/* ZAP */
98 	/* zvol: */
99 	DMU_OT_ZVOL,			/* UINT8 */
100 	DMU_OT_ZVOL_PROP,		/* ZAP */
101 	/* other; for testing only! */
102 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
103 	DMU_OT_UINT64_OTHER,		/* UINT64 */
104 	DMU_OT_ZAP_OTHER,		/* ZAP */
105 	/* new object types: */
106 	DMU_OT_ERROR_LOG,		/* ZAP */
107 	DMU_OT_SPA_HISTORY,		/* UINT8 */
108 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
109 	DMU_OT_POOL_PROPS,		/* ZAP */
110 	DMU_OT_DSL_PERMS,		/* ZAP */
111 	DMU_OT_ACL,			/* ACL */
112 	DMU_OT_SYSACL,			/* SYSACL */
113 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
114 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
115 	DMU_OT_NEXT_CLONES,		/* ZAP */
116 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
117 	DMU_OT_USERGROUP_USED,		/* ZAP */
118 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
119 	DMU_OT_USERREFS,		/* ZAP */
120 	DMU_OT_NUMTYPES
121 } dmu_object_type_t;
122 
123 typedef enum dmu_objset_type {
124 	DMU_OST_NONE,
125 	DMU_OST_META,
126 	DMU_OST_ZFS,
127 	DMU_OST_ZVOL,
128 	DMU_OST_OTHER,			/* For testing only! */
129 	DMU_OST_ANY,			/* Be careful! */
130 	DMU_OST_NUMTYPES
131 } dmu_objset_type_t;
132 
133 void byteswap_uint64_array(void *buf, size_t size);
134 void byteswap_uint32_array(void *buf, size_t size);
135 void byteswap_uint16_array(void *buf, size_t size);
136 void byteswap_uint8_array(void *buf, size_t size);
137 void zap_byteswap(void *buf, size_t size);
138 void zfs_oldacl_byteswap(void *buf, size_t size);
139 void zfs_acl_byteswap(void *buf, size_t size);
140 void zfs_znode_byteswap(void *buf, size_t size);
141 
142 #define	DS_FIND_SNAPSHOTS	(1<<0)
143 #define	DS_FIND_CHILDREN	(1<<1)
144 
145 /*
146  * The maximum number of bytes that can be accessed as part of one
147  * operation, including metadata.
148  */
149 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
150 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
151 
152 #define	DMU_USERUSED_OBJECT	(-1ULL)
153 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
154 
155 /*
156  * Public routines to create, destroy, open, and close objsets.
157  */
158 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
159 int dmu_objset_own(const char *name, dmu_objset_type_t type,
160     boolean_t readonly, void *tag, objset_t **osp);
161 void dmu_objset_rele(objset_t *os, void *tag);
162 void dmu_objset_disown(objset_t *os, void *tag);
163 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
164 
165 int dmu_objset_evict_dbufs(objset_t *os);
166 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
167     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
168 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
169     uint64_t flags);
170 int dmu_objset_destroy(const char *name, boolean_t defer);
171 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
172 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
173     boolean_t recursive);
174 int dmu_objset_rename(const char *name, const char *newname,
175     boolean_t recursive);
176 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
177     int flags);
178 void dmu_objset_byteswap(void *buf, size_t size);
179 
180 typedef struct dmu_buf {
181 	uint64_t db_object;		/* object that this buffer is part of */
182 	uint64_t db_offset;		/* byte offset in this object */
183 	uint64_t db_size;		/* size of buffer in bytes */
184 	void *db_data;			/* data in buffer */
185 } dmu_buf_t;
186 
187 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
188 
189 /*
190  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
191  */
192 #define	DMU_POOL_DIRECTORY_OBJECT	1
193 #define	DMU_POOL_CONFIG			"config"
194 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
195 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
196 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
197 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
198 #define	DMU_POOL_SPARES			"spares"
199 #define	DMU_POOL_DEFLATE		"deflate"
200 #define	DMU_POOL_HISTORY		"history"
201 #define	DMU_POOL_PROPS			"pool_props"
202 #define	DMU_POOL_L2CACHE		"l2cache"
203 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
204 
205 /* 4x8 zbookmark_t */
206 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
207 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
208 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
209 /* 1x8 txg */
210 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
211 /* 1x8 txg */
212 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
213 /* 1x4 enum scrub_func */
214 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
215 /* 1x8 count */
216 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
217 
218 /*
219  * Allocate an object from this objset.  The range of object numbers
220  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
221  *
222  * The transaction must be assigned to a txg.  The newly allocated
223  * object will be "held" in the transaction (ie. you can modify the
224  * newly allocated object in this transaction).
225  *
226  * dmu_object_alloc() chooses an object and returns it in *objectp.
227  *
228  * dmu_object_claim() allocates a specific object number.  If that
229  * number is already allocated, it fails and returns EEXIST.
230  *
231  * Return 0 on success, or ENOSPC or EEXIST as specified above.
232  */
233 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
234     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
235 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
236     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
237 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
238     int blocksize, dmu_object_type_t bonustype, int bonuslen);
239 
240 /*
241  * Free an object from this objset.
242  *
243  * The object's data will be freed as well (ie. you don't need to call
244  * dmu_free(object, 0, -1, tx)).
245  *
246  * The object need not be held in the transaction.
247  *
248  * If there are any holds on this object's buffers (via dmu_buf_hold()),
249  * or tx holds on the object (via dmu_tx_hold_object()), you can not
250  * free it; it fails and returns EBUSY.
251  *
252  * If the object is not allocated, it fails and returns ENOENT.
253  *
254  * Return 0 on success, or EBUSY or ENOENT as specified above.
255  */
256 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
257 
258 /*
259  * Find the next allocated or free object.
260  *
261  * The objectp parameter is in-out.  It will be updated to be the next
262  * object which is allocated.  Ignore objects which have not been
263  * modified since txg.
264  *
265  * XXX Can only be called on a objset with no dirty data.
266  *
267  * Returns 0 on success, or ENOENT if there are no more objects.
268  */
269 int dmu_object_next(objset_t *os, uint64_t *objectp,
270     boolean_t hole, uint64_t txg);
271 
272 /*
273  * Set the data blocksize for an object.
274  *
275  * The object cannot have any blocks allcated beyond the first.  If
276  * the first block is allocated already, the new size must be greater
277  * than the current block size.  If these conditions are not met,
278  * ENOTSUP will be returned.
279  *
280  * Returns 0 on success, or EBUSY if there are any holds on the object
281  * contents, or ENOTSUP as described above.
282  */
283 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
284     int ibs, dmu_tx_t *tx);
285 
286 /*
287  * Set the checksum property on a dnode.  The new checksum algorithm will
288  * apply to all newly written blocks; existing blocks will not be affected.
289  */
290 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
291     dmu_tx_t *tx);
292 
293 /*
294  * Set the compress property on a dnode.  The new compression algorithm will
295  * apply to all newly written blocks; existing blocks will not be affected.
296  */
297 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
298     dmu_tx_t *tx);
299 
300 /*
301  * Decide how many copies of a given block we should make.  Can be from
302  * 1 to SPA_DVAS_PER_BP.
303  */
304 int dmu_get_replication_level(objset_t *os, struct zbookmark *zb,
305     dmu_object_type_t ot);
306 /*
307  * The bonus data is accessed more or less like a regular buffer.
308  * You must dmu_bonus_hold() to get the buffer, which will give you a
309  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
310  * data.  As with any normal buffer, you must call dmu_buf_read() to
311  * read db_data, dmu_buf_will_dirty() before modifying it, and the
312  * object must be held in an assigned transaction before calling
313  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
314  * buffer as well.  You must release your hold with dmu_buf_rele().
315  */
316 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
317 int dmu_bonus_max(void);
318 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
319 
320 /*
321  * Obtain the DMU buffer from the specified object which contains the
322  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
323  * that it will remain in memory.  You must release the hold with
324  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
325  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
326  *
327  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
328  * on the returned buffer before reading or writing the buffer's
329  * db_data.  The comments for those routines describe what particular
330  * operations are valid after calling them.
331  *
332  * The object number must be a valid, allocated object number.
333  */
334 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
335     void *tag, dmu_buf_t **);
336 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
337 void dmu_buf_rele(dmu_buf_t *db, void *tag);
338 uint64_t dmu_buf_refcount(dmu_buf_t *db);
339 
340 /*
341  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
342  * range of an object.  A pointer to an array of dmu_buf_t*'s is
343  * returned (in *dbpp).
344  *
345  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
346  * frees the array.  The hold on the array of buffers MUST be released
347  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
348  * individually with dmu_buf_rele.
349  */
350 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
351     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
352 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
353 
354 /*
355  * Returns NULL on success, or the existing user ptr if it's already
356  * been set.
357  *
358  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
359  *
360  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
361  * will be set to db->db_data when you are allowed to access it.  Note
362  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
363  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
364  * *user_data_ptr_ptr will be set to the new value when it changes.
365  *
366  * If non-NULL, pageout func will be called when this buffer is being
367  * excised from the cache, so that you can clean up the data structure
368  * pointed to by user_ptr.
369  *
370  * dmu_evict_user() will call the pageout func for all buffers in a
371  * objset with a given pageout func.
372  */
373 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
374     dmu_buf_evict_func_t *pageout_func);
375 /*
376  * set_user_ie is the same as set_user, but request immediate eviction
377  * when hold count goes to zero.
378  */
379 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
380     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
381 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
382     void *user_ptr, void *user_data_ptr_ptr,
383     dmu_buf_evict_func_t *pageout_func);
384 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
385 
386 /*
387  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
388  */
389 void *dmu_buf_get_user(dmu_buf_t *db);
390 
391 /*
392  * Indicate that you are going to modify the buffer's data (db_data).
393  *
394  * The transaction (tx) must be assigned to a txg (ie. you've called
395  * dmu_tx_assign()).  The buffer's object must be held in the tx
396  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
397  */
398 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
399 
400 /*
401  * Tells if the given dbuf is freeable.
402  */
403 boolean_t dmu_buf_freeable(dmu_buf_t *);
404 
405 /*
406  * You must create a transaction, then hold the objects which you will
407  * (or might) modify as part of this transaction.  Then you must assign
408  * the transaction to a transaction group.  Once the transaction has
409  * been assigned, you can modify buffers which belong to held objects as
410  * part of this transaction.  You can't modify buffers before the
411  * transaction has been assigned; you can't modify buffers which don't
412  * belong to objects which this transaction holds; you can't hold
413  * objects once the transaction has been assigned.  You may hold an
414  * object which you are going to free (with dmu_object_free()), but you
415  * don't have to.
416  *
417  * You can abort the transaction before it has been assigned.
418  *
419  * Note that you may hold buffers (with dmu_buf_hold) at any time,
420  * regardless of transaction state.
421  */
422 
423 #define	DMU_NEW_OBJECT	(-1ULL)
424 #define	DMU_OBJECT_END	(-1ULL)
425 
426 dmu_tx_t *dmu_tx_create(objset_t *os);
427 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
428 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
429     uint64_t len);
430 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
431 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
432 void dmu_tx_abort(dmu_tx_t *tx);
433 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
434 void dmu_tx_wait(dmu_tx_t *tx);
435 void dmu_tx_commit(dmu_tx_t *tx);
436 
437 /*
438  * Free up the data blocks for a defined range of a file.  If size is
439  * zero, the range from offset to end-of-file is freed.
440  */
441 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
442 	uint64_t size, dmu_tx_t *tx);
443 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
444 	uint64_t size);
445 int dmu_free_object(objset_t *os, uint64_t object);
446 
447 /*
448  * Convenience functions.
449  *
450  * Canfail routines will return 0 on success, or an errno if there is a
451  * nonrecoverable I/O error.
452  */
453 #define	DMU_READ_PREFETCH	0 /* prefetch */
454 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
455 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
456 	void *buf, uint32_t flags);
457 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
458 	const void *buf, dmu_tx_t *tx);
459 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
460 	dmu_tx_t *tx);
461 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
462 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
463     dmu_tx_t *tx);
464 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
465     uint64_t size, struct page *pp, dmu_tx_t *tx);
466 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
467 void dmu_return_arcbuf(struct arc_buf *buf);
468 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
469     dmu_tx_t *tx);
470 
471 extern int zfs_prefetch_disable;
472 
473 /*
474  * Asynchronously try to read in the data.
475  */
476 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
477     uint64_t len);
478 
479 typedef struct dmu_object_info {
480 	/* All sizes are in bytes. */
481 	uint32_t doi_data_block_size;
482 	uint32_t doi_metadata_block_size;
483 	uint64_t doi_bonus_size;
484 	dmu_object_type_t doi_type;
485 	dmu_object_type_t doi_bonus_type;
486 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
487 	uint8_t doi_checksum;
488 	uint8_t doi_compress;
489 	uint8_t doi_pad[5];
490 	/* Values below are number of 512-byte blocks. */
491 	uint64_t doi_physical_blks;		/* data + metadata */
492 	uint64_t doi_max_block_offset;
493 } dmu_object_info_t;
494 
495 typedef void arc_byteswap_func_t(void *buf, size_t size);
496 
497 typedef struct dmu_object_type_info {
498 	arc_byteswap_func_t	*ot_byteswap;
499 	boolean_t		ot_metadata;
500 	char			*ot_name;
501 } dmu_object_type_info_t;
502 
503 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
504 
505 /*
506  * Get information on a DMU object.
507  *
508  * Return 0 on success or ENOENT if object is not allocated.
509  *
510  * If doi is NULL, just indicates whether the object exists.
511  */
512 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
513 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
514 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
515 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
516     u_longlong_t *nblk512);
517 
518 typedef struct dmu_objset_stats {
519 	uint64_t dds_num_clones; /* number of clones of this */
520 	uint64_t dds_creation_txg;
521 	uint64_t dds_guid;
522 	dmu_objset_type_t dds_type;
523 	uint8_t dds_is_snapshot;
524 	uint8_t dds_inconsistent;
525 	char dds_origin[MAXNAMELEN];
526 } dmu_objset_stats_t;
527 
528 /*
529  * Get stats on a dataset.
530  */
531 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
532 
533 /*
534  * Add entries to the nvlist for all the objset's properties.  See
535  * zfs_prop_table[] and zfs(1m) for details on the properties.
536  */
537 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
538 
539 /*
540  * Get the space usage statistics for statvfs().
541  *
542  * refdbytes is the amount of space "referenced" by this objset.
543  * availbytes is the amount of space available to this objset, taking
544  * into account quotas & reservations, assuming that no other objsets
545  * use the space first.  These values correspond to the 'referenced' and
546  * 'available' properties, described in the zfs(1m) manpage.
547  *
548  * usedobjs and availobjs are the number of objects currently allocated,
549  * and available.
550  */
551 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
552     uint64_t *usedobjsp, uint64_t *availobjsp);
553 
554 /*
555  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
556  * (Contrast with the ds_guid which is a 64-bit ID that will never
557  * change, so there is a small probability that it will collide.)
558  */
559 uint64_t dmu_objset_fsid_guid(objset_t *os);
560 
561 int dmu_objset_is_snapshot(objset_t *os);
562 
563 extern struct spa *dmu_objset_spa(objset_t *os);
564 extern struct zilog *dmu_objset_zil(objset_t *os);
565 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
566 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
567 extern void dmu_objset_name(objset_t *os, char *buf);
568 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
569 extern uint64_t dmu_objset_id(objset_t *os);
570 extern uint64_t dmu_objset_logbias(objset_t *os);
571 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
572     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
573 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
574     int maxlen, boolean_t *conflict);
575 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
576     uint64_t *idp, uint64_t *offp);
577 
578 typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
579     void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
580     dmu_tx_t *tx);
581 extern void dmu_objset_register_type(dmu_objset_type_t ost,
582     objset_used_cb_t *cb);
583 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
584 extern void *dmu_objset_get_user(objset_t *os);
585 
586 /*
587  * Return the txg number for the given assigned transaction.
588  */
589 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
590 
591 /*
592  * Synchronous write.
593  * If a parent zio is provided this function initiates a write on the
594  * provided buffer as a child of the parent zio.
595  * In the absence of a parent zio, the write is completed synchronously.
596  * At write completion, blk is filled with the bp of the written block.
597  * Note that while the data covered by this function will be on stable
598  * storage when the write completes this new data does not become a
599  * permanent part of the file until the associated transaction commits.
600  */
601 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
602 int dmu_sync(struct zio *zio, dmu_buf_t *db,
603     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
604 
605 /*
606  * Find the next hole or data block in file starting at *off
607  * Return found offset in *off. Return ESRCH for end of file.
608  */
609 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
610     uint64_t *off);
611 
612 /*
613  * Initial setup and final teardown.
614  */
615 extern void dmu_init(void);
616 extern void dmu_fini(void);
617 
618 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
619     uint64_t object, uint64_t offset, int len);
620 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
621     dmu_traverse_cb_t cb, void *arg);
622 
623 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
624     struct vnode *vp, offset_t *off);
625 
626 typedef struct dmu_recv_cookie {
627 	/*
628 	 * This structure is opaque!
629 	 *
630 	 * If logical and real are different, we are recving the stream
631 	 * into the "real" temporary clone, and then switching it with
632 	 * the "logical" target.
633 	 */
634 	struct dsl_dataset *drc_logical_ds;
635 	struct dsl_dataset *drc_real_ds;
636 	struct drr_begin *drc_drrb;
637 	char *drc_tosnap;
638 	boolean_t drc_newfs;
639 	boolean_t drc_force;
640 } dmu_recv_cookie_t;
641 
642 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
643     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
644 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
645 int dmu_recv_end(dmu_recv_cookie_t *drc);
646 
647 /* CRC64 table */
648 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
649 extern uint64_t zfs_crc64_table[256];
650 
651 #ifdef	__cplusplus
652 }
653 #endif
654 
655 #endif	/* _SYS_DMU_H */
656