xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 24b9abbad58fdd63dad716fd35a99a7944c4e3eb)
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 2008 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_NEXT_CLONES,		/* ZAP */
118 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
119 	DMU_OT_NUMTYPES
120 } dmu_object_type_t;
121 
122 typedef enum dmu_objset_type {
123 	DMU_OST_NONE,
124 	DMU_OST_META,
125 	DMU_OST_ZFS,
126 	DMU_OST_ZVOL,
127 	DMU_OST_OTHER,			/* For testing only! */
128 	DMU_OST_ANY,			/* Be careful! */
129 	DMU_OST_NUMTYPES
130 } dmu_objset_type_t;
131 
132 void byteswap_uint64_array(void *buf, size_t size);
133 void byteswap_uint32_array(void *buf, size_t size);
134 void byteswap_uint16_array(void *buf, size_t size);
135 void byteswap_uint8_array(void *buf, size_t size);
136 void zap_byteswap(void *buf, size_t size);
137 void zfs_oldacl_byteswap(void *buf, size_t size);
138 void zfs_acl_byteswap(void *buf, size_t size);
139 void zfs_znode_byteswap(void *buf, size_t size);
140 
141 #define	DS_MODE_NOHOLD		0	/* internal use only */
142 #define	DS_MODE_USER		1	/* simple access, no special needs */
143 #define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
144 #define	DS_MODE_TYPE_MASK	0x3
145 #define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
146 #define	DS_MODE_READONLY	0x8
147 #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
148 #define	DS_MODE_INCONSISTENT	0x10
149 #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
150 
151 #define	DS_FIND_SNAPSHOTS	(1<<0)
152 #define	DS_FIND_CHILDREN	(1<<1)
153 
154 /*
155  * The maximum number of bytes that can be accessed as part of one
156  * operation, including metadata.
157  */
158 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
159 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
160 
161 /*
162  * Public routines to create, destroy, open, and close objsets.
163  */
164 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
165     objset_t **osp);
166 int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
167     objset_t **osp);
168 void dmu_objset_close(objset_t *os);
169 int dmu_objset_evict_dbufs(objset_t *os);
170 int dmu_objset_create(const char *name, dmu_objset_type_t type,
171     objset_t *clone_parent, uint64_t flags,
172     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
173 int dmu_objset_destroy(const char *name);
174 int dmu_snapshots_destroy(char *fsname, char *snapname);
175 int dmu_objset_rollback(objset_t *os);
176 int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
177 int dmu_objset_rename(const char *name, const char *newname,
178     boolean_t recursive);
179 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
180     int flags);
181 void dmu_objset_byteswap(void *buf, size_t size);
182 
183 typedef struct dmu_buf {
184 	uint64_t db_object;		/* object that this buffer is part of */
185 	uint64_t db_offset;		/* byte offset in this object */
186 	uint64_t db_size;		/* size of buffer in bytes */
187 	void *db_data;			/* data in buffer */
188 } dmu_buf_t;
189 
190 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
191 
192 /*
193  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
194  */
195 #define	DMU_POOL_DIRECTORY_OBJECT	1
196 #define	DMU_POOL_CONFIG			"config"
197 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
198 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
199 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
200 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
201 #define	DMU_POOL_SPARES			"spares"
202 #define	DMU_POOL_DEFLATE		"deflate"
203 #define	DMU_POOL_HISTORY		"history"
204 #define	DMU_POOL_PROPS			"pool_props"
205 #define	DMU_POOL_L2CACHE		"l2cache"
206 
207 /* 4x8 zbookmark_t */
208 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
209 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
210 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
211 /* 1x8 txg */
212 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
213 /* 1x8 txg */
214 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
215 /* 1x4 enum scrub_func */
216 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
217 /* 1x8 count */
218 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
219 
220 /*
221  * Allocate an object from this objset.  The range of object numbers
222  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
223  *
224  * The transaction must be assigned to a txg.  The newly allocated
225  * object will be "held" in the transaction (ie. you can modify the
226  * newly allocated object in this transaction).
227  *
228  * dmu_object_alloc() chooses an object and returns it in *objectp.
229  *
230  * dmu_object_claim() allocates a specific object number.  If that
231  * number is already allocated, it fails and returns EEXIST.
232  *
233  * Return 0 on success, or ENOSPC or EEXIST as specified above.
234  */
235 uint64_t dmu_object_alloc(objset_t *os, 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_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
238     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
239 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
240     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
241 
242 /*
243  * Free an object from this objset.
244  *
245  * The object's data will be freed as well (ie. you don't need to call
246  * dmu_free(object, 0, -1, tx)).
247  *
248  * The object need not be held in the transaction.
249  *
250  * If there are any holds on this object's buffers (via dmu_buf_hold()),
251  * or tx holds on the object (via dmu_tx_hold_object()), you can not
252  * free it; it fails and returns EBUSY.
253  *
254  * If the object is not allocated, it fails and returns ENOENT.
255  *
256  * Return 0 on success, or EBUSY or ENOENT as specified above.
257  */
258 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
259 
260 /*
261  * Find the next allocated or free object.
262  *
263  * The objectp parameter is in-out.  It will be updated to be the next
264  * object which is allocated.  Ignore objects which have not been
265  * modified since txg.
266  *
267  * XXX Can only be called on a objset with no dirty data.
268  *
269  * Returns 0 on success, or ENOENT if there are no more objects.
270  */
271 int dmu_object_next(objset_t *os, uint64_t *objectp,
272     boolean_t hole, uint64_t txg);
273 
274 /*
275  * Set the data blocksize for an object.
276  *
277  * The object cannot have any blocks allcated beyond the first.  If
278  * the first block is allocated already, the new size must be greater
279  * than the current block size.  If these conditions are not met,
280  * ENOTSUP will be returned.
281  *
282  * Returns 0 on success, or EBUSY if there are any holds on the object
283  * contents, or ENOTSUP as described above.
284  */
285 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
286     int ibs, dmu_tx_t *tx);
287 
288 /*
289  * Set the checksum property on a dnode.  The new checksum algorithm will
290  * apply to all newly written blocks; existing blocks will not be affected.
291  */
292 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
293     dmu_tx_t *tx);
294 
295 /*
296  * Set the compress property on a dnode.  The new compression algorithm will
297  * apply to all newly written blocks; existing blocks will not be affected.
298  */
299 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
300     dmu_tx_t *tx);
301 
302 /*
303  * Decide how many copies of a given block we should make.  Can be from
304  * 1 to SPA_DVAS_PER_BP.
305  */
306 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
307     dmu_object_type_t ot);
308 /*
309  * The bonus data is accessed more or less like a regular buffer.
310  * You must dmu_bonus_hold() to get the buffer, which will give you a
311  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
312  * data.  As with any normal buffer, you must call dmu_buf_read() to
313  * read db_data, dmu_buf_will_dirty() before modifying it, and the
314  * object must be held in an assigned transaction before calling
315  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
316  * buffer as well.  You must release your hold with dmu_buf_rele().
317  */
318 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
319 int dmu_bonus_max(void);
320 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
321 
322 /*
323  * Obtain the DMU buffer from the specified object which contains the
324  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
325  * that it will remain in memory.  You must release the hold with
326  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
327  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
328  *
329  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
330  * on the returned buffer before reading or writing the buffer's
331  * db_data.  The comments for those routines describe what particular
332  * operations are valid after calling them.
333  *
334  * The object number must be a valid, allocated object number.
335  */
336 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
337     void *tag, dmu_buf_t **);
338 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
339 void dmu_buf_rele(dmu_buf_t *db, void *tag);
340 uint64_t dmu_buf_refcount(dmu_buf_t *db);
341 
342 /*
343  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
344  * range of an object.  A pointer to an array of dmu_buf_t*'s is
345  * returned (in *dbpp).
346  *
347  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
348  * frees the array.  The hold on the array of buffers MUST be released
349  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
350  * individually with dmu_buf_rele.
351  */
352 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
353     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
354 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
355 
356 /*
357  * Returns NULL on success, or the existing user ptr if it's already
358  * been set.
359  *
360  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
361  *
362  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
363  * will be set to db->db_data when you are allowed to access it.  Note
364  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
365  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
366  * *user_data_ptr_ptr will be set to the new value when it changes.
367  *
368  * If non-NULL, pageout func will be called when this buffer is being
369  * excised from the cache, so that you can clean up the data structure
370  * pointed to by user_ptr.
371  *
372  * dmu_evict_user() will call the pageout func for all buffers in a
373  * objset with a given pageout func.
374  */
375 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
376     dmu_buf_evict_func_t *pageout_func);
377 /*
378  * set_user_ie is the same as set_user, but request immediate eviction
379  * when hold count goes to zero.
380  */
381 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
382     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
383 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
384     void *user_ptr, void *user_data_ptr_ptr,
385     dmu_buf_evict_func_t *pageout_func);
386 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
387 
388 /*
389  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
390  */
391 void *dmu_buf_get_user(dmu_buf_t *db);
392 
393 /*
394  * Indicate that you are going to modify the buffer's data (db_data).
395  *
396  * The transaction (tx) must be assigned to a txg (ie. you've called
397  * dmu_tx_assign()).  The buffer's object must be held in the tx
398  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
399  */
400 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
401 
402 /*
403  * You must create a transaction, then hold the objects which you will
404  * (or might) modify as part of this transaction.  Then you must assign
405  * the transaction to a transaction group.  Once the transaction has
406  * been assigned, you can modify buffers which belong to held objects as
407  * part of this transaction.  You can't modify buffers before the
408  * transaction has been assigned; you can't modify buffers which don't
409  * belong to objects which this transaction holds; you can't hold
410  * objects once the transaction has been assigned.  You may hold an
411  * object which you are going to free (with dmu_object_free()), but you
412  * don't have to.
413  *
414  * You can abort the transaction before it has been assigned.
415  *
416  * Note that you may hold buffers (with dmu_buf_hold) at any time,
417  * regardless of transaction state.
418  */
419 
420 #define	DMU_NEW_OBJECT	(-1ULL)
421 #define	DMU_OBJECT_END	(-1ULL)
422 
423 dmu_tx_t *dmu_tx_create(objset_t *os);
424 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
425 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
426     uint64_t len);
427 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
428 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
429 void dmu_tx_abort(dmu_tx_t *tx);
430 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
431 void dmu_tx_wait(dmu_tx_t *tx);
432 void dmu_tx_commit(dmu_tx_t *tx);
433 
434 /*
435  * Free up the data blocks for a defined range of a file.  If size is
436  * zero, the range from offset to end-of-file is freed.
437  */
438 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
439 	uint64_t size, dmu_tx_t *tx);
440 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
441 	uint64_t size);
442 int dmu_free_object(objset_t *os, uint64_t object);
443 
444 /*
445  * Convenience functions.
446  *
447  * Canfail routines will return 0 on success, or an errno if there is a
448  * nonrecoverable I/O error.
449  */
450 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
451 	void *buf);
452 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
453 	const void *buf, dmu_tx_t *tx);
454 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
455 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
456     dmu_tx_t *tx);
457 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
458     uint64_t size, struct page *pp, dmu_tx_t *tx);
459 
460 extern int zfs_prefetch_disable;
461 
462 /*
463  * Asynchronously try to read in the data.
464  */
465 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
466     uint64_t len);
467 
468 typedef struct dmu_object_info {
469 	/* All sizes are in bytes. */
470 	uint32_t doi_data_block_size;
471 	uint32_t doi_metadata_block_size;
472 	uint64_t doi_bonus_size;
473 	dmu_object_type_t doi_type;
474 	dmu_object_type_t doi_bonus_type;
475 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
476 	uint8_t doi_checksum;
477 	uint8_t doi_compress;
478 	uint8_t doi_pad[5];
479 	/* Values below are number of 512-byte blocks. */
480 	uint64_t doi_physical_blks;		/* data + metadata */
481 	uint64_t doi_max_block_offset;
482 } dmu_object_info_t;
483 
484 typedef void arc_byteswap_func_t(void *buf, size_t size);
485 
486 typedef struct dmu_object_type_info {
487 	arc_byteswap_func_t	*ot_byteswap;
488 	boolean_t		ot_metadata;
489 	char			*ot_name;
490 } dmu_object_type_info_t;
491 
492 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
493 
494 /*
495  * Get information on a DMU object.
496  *
497  * Return 0 on success or ENOENT if object is not allocated.
498  *
499  * If doi is NULL, just indicates whether the object exists.
500  */
501 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
502 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
503 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
504 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
505     u_longlong_t *nblk512);
506 
507 typedef struct dmu_objset_stats {
508 	uint64_t dds_num_clones; /* number of clones of this */
509 	uint64_t dds_creation_txg;
510 	uint64_t dds_guid;
511 	dmu_objset_type_t dds_type;
512 	uint8_t dds_is_snapshot;
513 	uint8_t dds_inconsistent;
514 	char dds_origin[MAXNAMELEN];
515 } dmu_objset_stats_t;
516 
517 /*
518  * Get stats on a dataset.
519  */
520 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
521 
522 /*
523  * Add entries to the nvlist for all the objset's properties.  See
524  * zfs_prop_table[] and zfs(1m) for details on the properties.
525  */
526 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
527 
528 /*
529  * Get the space usage statistics for statvfs().
530  *
531  * refdbytes is the amount of space "referenced" by this objset.
532  * availbytes is the amount of space available to this objset, taking
533  * into account quotas & reservations, assuming that no other objsets
534  * use the space first.  These values correspond to the 'referenced' and
535  * 'available' properties, described in the zfs(1m) manpage.
536  *
537  * usedobjs and availobjs are the number of objects currently allocated,
538  * and available.
539  */
540 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
541     uint64_t *usedobjsp, uint64_t *availobjsp);
542 
543 /*
544  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
545  * (Contrast with the ds_guid which is a 64-bit ID that will never
546  * change, so there is a small probability that it will collide.)
547  */
548 uint64_t dmu_objset_fsid_guid(objset_t *os);
549 
550 int dmu_objset_is_snapshot(objset_t *os);
551 
552 extern struct spa *dmu_objset_spa(objset_t *os);
553 extern struct zilog *dmu_objset_zil(objset_t *os);
554 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
555 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
556 extern void dmu_objset_name(objset_t *os, char *buf);
557 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
558 extern uint64_t dmu_objset_id(objset_t *os);
559 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
560     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
561 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
562     int maxlen, boolean_t *conflict);
563 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
564     uint64_t *idp, uint64_t *offp);
565 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
566 extern void *dmu_objset_get_user(objset_t *os);
567 
568 /*
569  * Return the txg number for the given assigned transaction.
570  */
571 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
572 
573 /*
574  * Synchronous write.
575  * If a parent zio is provided this function initiates a write on the
576  * provided buffer as a child of the parent zio.
577  * In the absence of a parent zio, the write is completed synchronously.
578  * At write completion, blk is filled with the bp of the written block.
579  * Note that while the data covered by this function will be on stable
580  * storage when the write completes this new data does not become a
581  * permanent part of the file until the associated transaction commits.
582  */
583 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
584 int dmu_sync(struct zio *zio, dmu_buf_t *db,
585     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
586 
587 /*
588  * Find the next hole or data block in file starting at *off
589  * Return found offset in *off. Return ESRCH for end of file.
590  */
591 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
592     uint64_t *off);
593 
594 /*
595  * Initial setup and final teardown.
596  */
597 extern void dmu_init(void);
598 extern void dmu_fini(void);
599 
600 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
601     uint64_t object, uint64_t offset, int len);
602 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
603     dmu_traverse_cb_t cb, void *arg);
604 
605 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
606     struct vnode *vp, offset_t *off);
607 
608 typedef struct dmu_recv_cookie {
609 	/*
610 	 * This structure is opaque!
611 	 *
612 	 * If logical and real are different, we are recving the stream
613 	 * into the "real" temporary clone, and then switching it with
614 	 * the "logical" target.
615 	 */
616 	struct dsl_dataset *drc_logical_ds;
617 	struct dsl_dataset *drc_real_ds;
618 	struct drr_begin *drc_drrb;
619 	char *drc_tosnap;
620 	boolean_t drc_newfs;
621 	boolean_t drc_force;
622 } dmu_recv_cookie_t;
623 
624 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
625     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
626 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
627 int dmu_recv_end(dmu_recv_cookie_t *drc);
628 void dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
629 
630 /* CRC64 table */
631 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
632 extern uint64_t zfs_crc64_table[256];
633 
634 #ifdef	__cplusplus
635 }
636 #endif
637 
638 #endif	/* _SYS_DMU_H */
639