xref: /titanic_50/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 0767cb98df68c7ec270b1b364b2796ea5e2ff28c)
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 objset_impl;
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_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_MODE_NOHOLD		0	/* internal use only */
143 #define	DS_MODE_USER		1	/* simple access, no special needs */
144 #define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
145 #define	DS_MODE_TYPE_MASK	0x3
146 #define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
147 #define	DS_MODE_READONLY	0x8
148 #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
149 #define	DS_MODE_INCONSISTENT	0x10
150 #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
151 
152 #define	DS_FIND_SNAPSHOTS	(1<<0)
153 #define	DS_FIND_CHILDREN	(1<<1)
154 
155 /*
156  * The maximum number of bytes that can be accessed as part of one
157  * operation, including metadata.
158  */
159 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
160 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
161 
162 #define	DMU_USERUSED_OBJECT	(-1ULL)
163 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
164 
165 /*
166  * Public routines to create, destroy, open, and close objsets.
167  */
168 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
169     objset_t **osp);
170 int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
171     objset_t **osp);
172 void dmu_objset_close(objset_t *os);
173 int dmu_objset_evict_dbufs(objset_t *os);
174 int dmu_objset_create(const char *name, dmu_objset_type_t type,
175     objset_t *clone_parent, uint64_t flags,
176     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
177 int dmu_objset_destroy(const char *name);
178 int dmu_snapshots_destroy(char *fsname, char *snapname);
179 int dmu_objset_rollback(objset_t *os);
180 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
181     boolean_t recursive);
182 int dmu_objset_rename(const char *name, const char *newname,
183     boolean_t recursive);
184 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
185     int flags);
186 void dmu_objset_byteswap(void *buf, size_t size);
187 
188 typedef struct dmu_buf {
189 	uint64_t db_object;		/* object that this buffer is part of */
190 	uint64_t db_offset;		/* byte offset in this object */
191 	uint64_t db_size;		/* size of buffer in bytes */
192 	void *db_data;			/* data in buffer */
193 } dmu_buf_t;
194 
195 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
196 
197 /*
198  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
199  */
200 #define	DMU_POOL_DIRECTORY_OBJECT	1
201 #define	DMU_POOL_CONFIG			"config"
202 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
203 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
204 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
205 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
206 #define	DMU_POOL_SPARES			"spares"
207 #define	DMU_POOL_DEFLATE		"deflate"
208 #define	DMU_POOL_HISTORY		"history"
209 #define	DMU_POOL_PROPS			"pool_props"
210 #define	DMU_POOL_L2CACHE		"l2cache"
211 
212 /* 4x8 zbookmark_t */
213 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
214 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
215 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
216 /* 1x8 txg */
217 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
218 /* 1x8 txg */
219 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
220 /* 1x4 enum scrub_func */
221 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
222 /* 1x8 count */
223 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
224 
225 /*
226  * Allocate an object from this objset.  The range of object numbers
227  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
228  *
229  * The transaction must be assigned to a txg.  The newly allocated
230  * object will be "held" in the transaction (ie. you can modify the
231  * newly allocated object in this transaction).
232  *
233  * dmu_object_alloc() chooses an object and returns it in *objectp.
234  *
235  * dmu_object_claim() allocates a specific object number.  If that
236  * number is already allocated, it fails and returns EEXIST.
237  *
238  * Return 0 on success, or ENOSPC or EEXIST as specified above.
239  */
240 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
241     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
242 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
243     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
244 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
245     int blocksize, dmu_object_type_t bonustype, int bonuslen);
246 
247 /*
248  * Free an object from this objset.
249  *
250  * The object's data will be freed as well (ie. you don't need to call
251  * dmu_free(object, 0, -1, tx)).
252  *
253  * The object need not be held in the transaction.
254  *
255  * If there are any holds on this object's buffers (via dmu_buf_hold()),
256  * or tx holds on the object (via dmu_tx_hold_object()), you can not
257  * free it; it fails and returns EBUSY.
258  *
259  * If the object is not allocated, it fails and returns ENOENT.
260  *
261  * Return 0 on success, or EBUSY or ENOENT as specified above.
262  */
263 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
264 
265 /*
266  * Find the next allocated or free object.
267  *
268  * The objectp parameter is in-out.  It will be updated to be the next
269  * object which is allocated.  Ignore objects which have not been
270  * modified since txg.
271  *
272  * XXX Can only be called on a objset with no dirty data.
273  *
274  * Returns 0 on success, or ENOENT if there are no more objects.
275  */
276 int dmu_object_next(objset_t *os, uint64_t *objectp,
277     boolean_t hole, uint64_t txg);
278 
279 /*
280  * Set the data blocksize for an object.
281  *
282  * The object cannot have any blocks allcated beyond the first.  If
283  * the first block is allocated already, the new size must be greater
284  * than the current block size.  If these conditions are not met,
285  * ENOTSUP will be returned.
286  *
287  * Returns 0 on success, or EBUSY if there are any holds on the object
288  * contents, or ENOTSUP as described above.
289  */
290 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
291     int ibs, dmu_tx_t *tx);
292 
293 /*
294  * Set the checksum property on a dnode.  The new checksum algorithm will
295  * apply to all newly written blocks; existing blocks will not be affected.
296  */
297 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
298     dmu_tx_t *tx);
299 
300 /*
301  * Set the compress property on a dnode.  The new compression algorithm will
302  * apply to all newly written blocks; existing blocks will not be affected.
303  */
304 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
305     dmu_tx_t *tx);
306 
307 /*
308  * Decide how many copies of a given block we should make.  Can be from
309  * 1 to SPA_DVAS_PER_BP.
310  */
311 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
312     dmu_object_type_t ot);
313 /*
314  * The bonus data is accessed more or less like a regular buffer.
315  * You must dmu_bonus_hold() to get the buffer, which will give you a
316  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
317  * data.  As with any normal buffer, you must call dmu_buf_read() to
318  * read db_data, dmu_buf_will_dirty() before modifying it, and the
319  * object must be held in an assigned transaction before calling
320  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
321  * buffer as well.  You must release your hold with dmu_buf_rele().
322  */
323 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
324 int dmu_bonus_max(void);
325 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
326 
327 /*
328  * Obtain the DMU buffer from the specified object which contains the
329  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
330  * that it will remain in memory.  You must release the hold with
331  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
332  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
333  *
334  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
335  * on the returned buffer before reading or writing the buffer's
336  * db_data.  The comments for those routines describe what particular
337  * operations are valid after calling them.
338  *
339  * The object number must be a valid, allocated object number.
340  */
341 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
342     void *tag, dmu_buf_t **);
343 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
344 void dmu_buf_rele(dmu_buf_t *db, void *tag);
345 uint64_t dmu_buf_refcount(dmu_buf_t *db);
346 
347 /*
348  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
349  * range of an object.  A pointer to an array of dmu_buf_t*'s is
350  * returned (in *dbpp).
351  *
352  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
353  * frees the array.  The hold on the array of buffers MUST be released
354  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
355  * individually with dmu_buf_rele.
356  */
357 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
358     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
359 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
360 
361 /*
362  * Returns NULL on success, or the existing user ptr if it's already
363  * been set.
364  *
365  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
366  *
367  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
368  * will be set to db->db_data when you are allowed to access it.  Note
369  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
370  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
371  * *user_data_ptr_ptr will be set to the new value when it changes.
372  *
373  * If non-NULL, pageout func will be called when this buffer is being
374  * excised from the cache, so that you can clean up the data structure
375  * pointed to by user_ptr.
376  *
377  * dmu_evict_user() will call the pageout func for all buffers in a
378  * objset with a given pageout func.
379  */
380 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
381     dmu_buf_evict_func_t *pageout_func);
382 /*
383  * set_user_ie is the same as set_user, but request immediate eviction
384  * when hold count goes to zero.
385  */
386 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
387     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
388 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
389     void *user_ptr, void *user_data_ptr_ptr,
390     dmu_buf_evict_func_t *pageout_func);
391 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
392 
393 /*
394  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
395  */
396 void *dmu_buf_get_user(dmu_buf_t *db);
397 
398 /*
399  * Indicate that you are going to modify the buffer's data (db_data).
400  *
401  * The transaction (tx) must be assigned to a txg (ie. you've called
402  * dmu_tx_assign()).  The buffer's object must be held in the tx
403  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
404  */
405 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
406 
407 /*
408  * You must create a transaction, then hold the objects which you will
409  * (or might) modify as part of this transaction.  Then you must assign
410  * the transaction to a transaction group.  Once the transaction has
411  * been assigned, you can modify buffers which belong to held objects as
412  * part of this transaction.  You can't modify buffers before the
413  * transaction has been assigned; you can't modify buffers which don't
414  * belong to objects which this transaction holds; you can't hold
415  * objects once the transaction has been assigned.  You may hold an
416  * object which you are going to free (with dmu_object_free()), but you
417  * don't have to.
418  *
419  * You can abort the transaction before it has been assigned.
420  *
421  * Note that you may hold buffers (with dmu_buf_hold) at any time,
422  * regardless of transaction state.
423  */
424 
425 #define	DMU_NEW_OBJECT	(-1ULL)
426 #define	DMU_OBJECT_END	(-1ULL)
427 
428 dmu_tx_t *dmu_tx_create(objset_t *os);
429 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
430 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
431     uint64_t len);
432 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
433 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
434 void dmu_tx_abort(dmu_tx_t *tx);
435 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
436 void dmu_tx_wait(dmu_tx_t *tx);
437 void dmu_tx_commit(dmu_tx_t *tx);
438 
439 /*
440  * Free up the data blocks for a defined range of a file.  If size is
441  * zero, the range from offset to end-of-file is freed.
442  */
443 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
444 	uint64_t size, dmu_tx_t *tx);
445 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
446 	uint64_t size);
447 int dmu_free_object(objset_t *os, uint64_t object);
448 
449 /*
450  * Convenience functions.
451  *
452  * Canfail routines will return 0 on success, or an errno if there is a
453  * nonrecoverable I/O error.
454  */
455 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
456 	void *buf);
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 int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
571     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
572 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
573     int maxlen, boolean_t *conflict);
574 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
575     uint64_t *idp, uint64_t *offp);
576 
577 typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
578     void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
579     dmu_tx_t *tx);
580 extern void dmu_objset_register_type(dmu_objset_type_t ost,
581     objset_used_cb_t *cb);
582 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
583 extern void *dmu_objset_get_user(objset_t *os);
584 
585 /*
586  * Return the txg number for the given assigned transaction.
587  */
588 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
589 
590 /*
591  * Synchronous write.
592  * If a parent zio is provided this function initiates a write on the
593  * provided buffer as a child of the parent zio.
594  * In the absence of a parent zio, the write is completed synchronously.
595  * At write completion, blk is filled with the bp of the written block.
596  * Note that while the data covered by this function will be on stable
597  * storage when the write completes this new data does not become a
598  * permanent part of the file until the associated transaction commits.
599  */
600 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
601 int dmu_sync(struct zio *zio, dmu_buf_t *db,
602     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
603 
604 /*
605  * Find the next hole or data block in file starting at *off
606  * Return found offset in *off. Return ESRCH for end of file.
607  */
608 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
609     uint64_t *off);
610 
611 /*
612  * Initial setup and final teardown.
613  */
614 extern void dmu_init(void);
615 extern void dmu_fini(void);
616 
617 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
618     uint64_t object, uint64_t offset, int len);
619 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
620     dmu_traverse_cb_t cb, void *arg);
621 
622 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
623     struct vnode *vp, offset_t *off);
624 
625 typedef struct dmu_recv_cookie {
626 	/*
627 	 * This structure is opaque!
628 	 *
629 	 * If logical and real are different, we are recving the stream
630 	 * into the "real" temporary clone, and then switching it with
631 	 * the "logical" target.
632 	 */
633 	struct dsl_dataset *drc_logical_ds;
634 	struct dsl_dataset *drc_real_ds;
635 	struct drr_begin *drc_drrb;
636 	char *drc_tosnap;
637 	boolean_t drc_newfs;
638 	boolean_t drc_force;
639 } dmu_recv_cookie_t;
640 
641 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
642     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
643 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
644 int dmu_recv_end(dmu_recv_cookie_t *drc);
645 void dmu_recv_abort_cleanup(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