xref: /titanic_50/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 7ff836697c120cb94bd30d5c2204eb9b74718e4c)
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_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_MODE_NOHOLD		0	/* internal use only */
144 #define	DS_MODE_USER		1	/* simple access, no special needs */
145 #define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
146 #define	DS_MODE_TYPE_MASK	0x3
147 #define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
148 #define	DS_MODE_READONLY	0x8
149 #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
150 #define	DS_MODE_INCONSISTENT	0x10
151 #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
152 
153 #define	DS_FIND_SNAPSHOTS	(1<<0)
154 #define	DS_FIND_CHILDREN	(1<<1)
155 
156 /*
157  * The maximum number of bytes that can be accessed as part of one
158  * operation, including metadata.
159  */
160 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
161 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
162 
163 #define	DMU_USERUSED_OBJECT	(-1ULL)
164 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
165 
166 /*
167  * Public routines to create, destroy, open, and close objsets.
168  */
169 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
170     objset_t **osp);
171 int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
172     objset_t **osp);
173 void dmu_objset_close(objset_t *os);
174 int dmu_objset_evict_dbufs(objset_t *os);
175 int dmu_objset_create(const char *name, dmu_objset_type_t type,
176     objset_t *clone_parent, uint64_t flags,
177     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
178 int dmu_objset_destroy(const char *name, boolean_t defer);
179 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
180 int dmu_objset_rollback(objset_t *os);
181 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
182     boolean_t recursive);
183 int dmu_objset_rename(const char *name, const char *newname,
184     boolean_t recursive);
185 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
186     int flags);
187 void dmu_objset_byteswap(void *buf, size_t size);
188 
189 typedef struct dmu_buf {
190 	uint64_t db_object;		/* object that this buffer is part of */
191 	uint64_t db_offset;		/* byte offset in this object */
192 	uint64_t db_size;		/* size of buffer in bytes */
193 	void *db_data;			/* data in buffer */
194 } dmu_buf_t;
195 
196 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
197 
198 /*
199  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
200  */
201 #define	DMU_POOL_DIRECTORY_OBJECT	1
202 #define	DMU_POOL_CONFIG			"config"
203 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
204 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
205 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
206 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
207 #define	DMU_POOL_SPARES			"spares"
208 #define	DMU_POOL_DEFLATE		"deflate"
209 #define	DMU_POOL_HISTORY		"history"
210 #define	DMU_POOL_PROPS			"pool_props"
211 #define	DMU_POOL_L2CACHE		"l2cache"
212 
213 /* 4x8 zbookmark_t */
214 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
215 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
216 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
217 /* 1x8 txg */
218 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
219 /* 1x8 txg */
220 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
221 /* 1x4 enum scrub_func */
222 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
223 /* 1x8 count */
224 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
225 
226 /*
227  * Allocate an object from this objset.  The range of object numbers
228  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
229  *
230  * The transaction must be assigned to a txg.  The newly allocated
231  * object will be "held" in the transaction (ie. you can modify the
232  * newly allocated object in this transaction).
233  *
234  * dmu_object_alloc() chooses an object and returns it in *objectp.
235  *
236  * dmu_object_claim() allocates a specific object number.  If that
237  * number is already allocated, it fails and returns EEXIST.
238  *
239  * Return 0 on success, or ENOSPC or EEXIST as specified above.
240  */
241 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
242     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
243 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
244     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
245 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
246     int blocksize, dmu_object_type_t bonustype, int bonuslen);
247 
248 /*
249  * Free an object from this objset.
250  *
251  * The object's data will be freed as well (ie. you don't need to call
252  * dmu_free(object, 0, -1, tx)).
253  *
254  * The object need not be held in the transaction.
255  *
256  * If there are any holds on this object's buffers (via dmu_buf_hold()),
257  * or tx holds on the object (via dmu_tx_hold_object()), you can not
258  * free it; it fails and returns EBUSY.
259  *
260  * If the object is not allocated, it fails and returns ENOENT.
261  *
262  * Return 0 on success, or EBUSY or ENOENT as specified above.
263  */
264 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
265 
266 /*
267  * Find the next allocated or free object.
268  *
269  * The objectp parameter is in-out.  It will be updated to be the next
270  * object which is allocated.  Ignore objects which have not been
271  * modified since txg.
272  *
273  * XXX Can only be called on a objset with no dirty data.
274  *
275  * Returns 0 on success, or ENOENT if there are no more objects.
276  */
277 int dmu_object_next(objset_t *os, uint64_t *objectp,
278     boolean_t hole, uint64_t txg);
279 
280 /*
281  * Set the data blocksize for an object.
282  *
283  * The object cannot have any blocks allcated beyond the first.  If
284  * the first block is allocated already, the new size must be greater
285  * than the current block size.  If these conditions are not met,
286  * ENOTSUP will be returned.
287  *
288  * Returns 0 on success, or EBUSY if there are any holds on the object
289  * contents, or ENOTSUP as described above.
290  */
291 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
292     int ibs, dmu_tx_t *tx);
293 
294 /*
295  * Set the checksum property on a dnode.  The new checksum algorithm will
296  * apply to all newly written blocks; existing blocks will not be affected.
297  */
298 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
299     dmu_tx_t *tx);
300 
301 /*
302  * Set the compress property on a dnode.  The new compression algorithm will
303  * apply to all newly written blocks; existing blocks will not be affected.
304  */
305 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
306     dmu_tx_t *tx);
307 
308 /*
309  * Decide how many copies of a given block we should make.  Can be from
310  * 1 to SPA_DVAS_PER_BP.
311  */
312 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
313     dmu_object_type_t ot);
314 /*
315  * The bonus data is accessed more or less like a regular buffer.
316  * You must dmu_bonus_hold() to get the buffer, which will give you a
317  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
318  * data.  As with any normal buffer, you must call dmu_buf_read() to
319  * read db_data, dmu_buf_will_dirty() before modifying it, and the
320  * object must be held in an assigned transaction before calling
321  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
322  * buffer as well.  You must release your hold with dmu_buf_rele().
323  */
324 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
325 int dmu_bonus_max(void);
326 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
327 
328 /*
329  * Obtain the DMU buffer from the specified object which contains the
330  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
331  * that it will remain in memory.  You must release the hold with
332  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
333  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
334  *
335  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
336  * on the returned buffer before reading or writing the buffer's
337  * db_data.  The comments for those routines describe what particular
338  * operations are valid after calling them.
339  *
340  * The object number must be a valid, allocated object number.
341  */
342 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
343     void *tag, dmu_buf_t **);
344 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
345 void dmu_buf_rele(dmu_buf_t *db, void *tag);
346 uint64_t dmu_buf_refcount(dmu_buf_t *db);
347 
348 /*
349  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
350  * range of an object.  A pointer to an array of dmu_buf_t*'s is
351  * returned (in *dbpp).
352  *
353  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
354  * frees the array.  The hold on the array of buffers MUST be released
355  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
356  * individually with dmu_buf_rele.
357  */
358 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
359     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
360 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
361 
362 /*
363  * Returns NULL on success, or the existing user ptr if it's already
364  * been set.
365  *
366  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
367  *
368  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
369  * will be set to db->db_data when you are allowed to access it.  Note
370  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
371  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
372  * *user_data_ptr_ptr will be set to the new value when it changes.
373  *
374  * If non-NULL, pageout func will be called when this buffer is being
375  * excised from the cache, so that you can clean up the data structure
376  * pointed to by user_ptr.
377  *
378  * dmu_evict_user() will call the pageout func for all buffers in a
379  * objset with a given pageout func.
380  */
381 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
382     dmu_buf_evict_func_t *pageout_func);
383 /*
384  * set_user_ie is the same as set_user, but request immediate eviction
385  * when hold count goes to zero.
386  */
387 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
388     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
389 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
390     void *user_ptr, void *user_data_ptr_ptr,
391     dmu_buf_evict_func_t *pageout_func);
392 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
393 
394 /*
395  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
396  */
397 void *dmu_buf_get_user(dmu_buf_t *db);
398 
399 /*
400  * Indicate that you are going to modify the buffer's data (db_data).
401  *
402  * The transaction (tx) must be assigned to a txg (ie. you've called
403  * dmu_tx_assign()).  The buffer's object must be held in the tx
404  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
405  */
406 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
407 
408 /*
409  * Tells if the given dbuf is freeable.
410  */
411 boolean_t dmu_buf_freeable(dmu_buf_t *);
412 
413 /*
414  * You must create a transaction, then hold the objects which you will
415  * (or might) modify as part of this transaction.  Then you must assign
416  * the transaction to a transaction group.  Once the transaction has
417  * been assigned, you can modify buffers which belong to held objects as
418  * part of this transaction.  You can't modify buffers before the
419  * transaction has been assigned; you can't modify buffers which don't
420  * belong to objects which this transaction holds; you can't hold
421  * objects once the transaction has been assigned.  You may hold an
422  * object which you are going to free (with dmu_object_free()), but you
423  * don't have to.
424  *
425  * You can abort the transaction before it has been assigned.
426  *
427  * Note that you may hold buffers (with dmu_buf_hold) at any time,
428  * regardless of transaction state.
429  */
430 
431 #define	DMU_NEW_OBJECT	(-1ULL)
432 #define	DMU_OBJECT_END	(-1ULL)
433 
434 dmu_tx_t *dmu_tx_create(objset_t *os);
435 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
436 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
437     uint64_t len);
438 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
439 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
440 void dmu_tx_abort(dmu_tx_t *tx);
441 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
442 void dmu_tx_wait(dmu_tx_t *tx);
443 void dmu_tx_commit(dmu_tx_t *tx);
444 
445 /*
446  * Free up the data blocks for a defined range of a file.  If size is
447  * zero, the range from offset to end-of-file is freed.
448  */
449 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
450 	uint64_t size, dmu_tx_t *tx);
451 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
452 	uint64_t size);
453 int dmu_free_object(objset_t *os, uint64_t object);
454 
455 /*
456  * Convenience functions.
457  *
458  * Canfail routines will return 0 on success, or an errno if there is a
459  * nonrecoverable I/O error.
460  */
461 #define	DMU_READ_PREFETCH	0 /* prefetch */
462 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
463 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
464 	void *buf, uint32_t flags);
465 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
466 	const void *buf, dmu_tx_t *tx);
467 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
468 	dmu_tx_t *tx);
469 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
470 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
471     dmu_tx_t *tx);
472 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
473     uint64_t size, struct page *pp, dmu_tx_t *tx);
474 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
475 void dmu_return_arcbuf(struct arc_buf *buf);
476 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
477     dmu_tx_t *tx);
478 
479 extern int zfs_prefetch_disable;
480 
481 /*
482  * Asynchronously try to read in the data.
483  */
484 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
485     uint64_t len);
486 
487 typedef struct dmu_object_info {
488 	/* All sizes are in bytes. */
489 	uint32_t doi_data_block_size;
490 	uint32_t doi_metadata_block_size;
491 	uint64_t doi_bonus_size;
492 	dmu_object_type_t doi_type;
493 	dmu_object_type_t doi_bonus_type;
494 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
495 	uint8_t doi_checksum;
496 	uint8_t doi_compress;
497 	uint8_t doi_pad[5];
498 	/* Values below are number of 512-byte blocks. */
499 	uint64_t doi_physical_blks;		/* data + metadata */
500 	uint64_t doi_max_block_offset;
501 } dmu_object_info_t;
502 
503 typedef void arc_byteswap_func_t(void *buf, size_t size);
504 
505 typedef struct dmu_object_type_info {
506 	arc_byteswap_func_t	*ot_byteswap;
507 	boolean_t		ot_metadata;
508 	char			*ot_name;
509 } dmu_object_type_info_t;
510 
511 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
512 
513 /*
514  * Get information on a DMU object.
515  *
516  * Return 0 on success or ENOENT if object is not allocated.
517  *
518  * If doi is NULL, just indicates whether the object exists.
519  */
520 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
521 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
522 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
523 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
524     u_longlong_t *nblk512);
525 
526 typedef struct dmu_objset_stats {
527 	uint64_t dds_num_clones; /* number of clones of this */
528 	uint64_t dds_creation_txg;
529 	uint64_t dds_guid;
530 	dmu_objset_type_t dds_type;
531 	uint8_t dds_is_snapshot;
532 	uint8_t dds_inconsistent;
533 	char dds_origin[MAXNAMELEN];
534 } dmu_objset_stats_t;
535 
536 /*
537  * Get stats on a dataset.
538  */
539 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
540 
541 /*
542  * Add entries to the nvlist for all the objset's properties.  See
543  * zfs_prop_table[] and zfs(1m) for details on the properties.
544  */
545 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
546 
547 /*
548  * Get the space usage statistics for statvfs().
549  *
550  * refdbytes is the amount of space "referenced" by this objset.
551  * availbytes is the amount of space available to this objset, taking
552  * into account quotas & reservations, assuming that no other objsets
553  * use the space first.  These values correspond to the 'referenced' and
554  * 'available' properties, described in the zfs(1m) manpage.
555  *
556  * usedobjs and availobjs are the number of objects currently allocated,
557  * and available.
558  */
559 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
560     uint64_t *usedobjsp, uint64_t *availobjsp);
561 
562 /*
563  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
564  * (Contrast with the ds_guid which is a 64-bit ID that will never
565  * change, so there is a small probability that it will collide.)
566  */
567 uint64_t dmu_objset_fsid_guid(objset_t *os);
568 
569 int dmu_objset_is_snapshot(objset_t *os);
570 
571 extern struct spa *dmu_objset_spa(objset_t *os);
572 extern struct zilog *dmu_objset_zil(objset_t *os);
573 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
574 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
575 extern void dmu_objset_name(objset_t *os, char *buf);
576 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
577 extern uint64_t dmu_objset_id(objset_t *os);
578 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
579     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
580 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
581     int maxlen, boolean_t *conflict);
582 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
583     uint64_t *idp, uint64_t *offp);
584 
585 typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
586     void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
587     dmu_tx_t *tx);
588 extern void dmu_objset_register_type(dmu_objset_type_t ost,
589     objset_used_cb_t *cb);
590 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
591 extern void *dmu_objset_get_user(objset_t *os);
592 
593 /*
594  * Return the txg number for the given assigned transaction.
595  */
596 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
597 
598 /*
599  * Synchronous write.
600  * If a parent zio is provided this function initiates a write on the
601  * provided buffer as a child of the parent zio.
602  * In the absence of a parent zio, the write is completed synchronously.
603  * At write completion, blk is filled with the bp of the written block.
604  * Note that while the data covered by this function will be on stable
605  * storage when the write completes this new data does not become a
606  * permanent part of the file until the associated transaction commits.
607  */
608 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
609 int dmu_sync(struct zio *zio, dmu_buf_t *db,
610     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
611 
612 /*
613  * Find the next hole or data block in file starting at *off
614  * Return found offset in *off. Return ESRCH for end of file.
615  */
616 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
617     uint64_t *off);
618 
619 /*
620  * Initial setup and final teardown.
621  */
622 extern void dmu_init(void);
623 extern void dmu_fini(void);
624 
625 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
626     uint64_t object, uint64_t offset, int len);
627 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
628     dmu_traverse_cb_t cb, void *arg);
629 
630 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
631     struct vnode *vp, offset_t *off);
632 
633 typedef struct dmu_recv_cookie {
634 	/*
635 	 * This structure is opaque!
636 	 *
637 	 * If logical and real are different, we are recving the stream
638 	 * into the "real" temporary clone, and then switching it with
639 	 * the "logical" target.
640 	 */
641 	struct dsl_dataset *drc_logical_ds;
642 	struct dsl_dataset *drc_real_ds;
643 	struct drr_begin *drc_drrb;
644 	char *drc_tosnap;
645 	boolean_t drc_newfs;
646 	boolean_t drc_force;
647 } dmu_recv_cookie_t;
648 
649 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
650     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
651 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
652 int dmu_recv_end(dmu_recv_cookie_t *drc);
653 
654 /* CRC64 table */
655 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
656 extern uint64_t zfs_crc64_table[256];
657 
658 #ifdef	__cplusplus
659 }
660 #endif
661 
662 #endif	/* _SYS_DMU_H */
663