xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 9e69d7d0feb2a0394435ca1d9746c4c3d7bf9b22)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_SYS_DMU_H
27 #define	_SYS_DMU_H
28 
29 /*
30  * This file describes the interface that the DMU provides for its
31  * consumers.
32  *
33  * The DMU also interacts with the SPA.  That interface is described in
34  * dmu_spa.h.
35  */
36 
37 #include <sys/inttypes.h>
38 #include <sys/types.h>
39 #include <sys/param.h>
40 #include <sys/cred.h>
41 #include <sys/time.h>
42 
43 #ifdef	__cplusplus
44 extern "C" {
45 #endif
46 
47 struct uio;
48 struct page;
49 struct vnode;
50 struct spa;
51 struct zilog;
52 struct zio;
53 struct blkptr;
54 struct zap_cursor;
55 struct dsl_dataset;
56 struct dsl_pool;
57 struct dnode;
58 struct drr_begin;
59 struct drr_end;
60 struct zbookmark;
61 struct spa;
62 struct nvlist;
63 struct arc_buf;
64 struct zio_prop;
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_USERGROUP_USED,		/* ZAP */
120 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
121 	DMU_OT_USERREFS,		/* ZAP */
122 	DMU_OT_DDT_ZAP,			/* ZAP */
123 	DMU_OT_DDT_STATS,		/* ZAP */
124 	DMU_OT_NUMTYPES
125 } dmu_object_type_t;
126 
127 typedef enum dmu_objset_type {
128 	DMU_OST_NONE,
129 	DMU_OST_META,
130 	DMU_OST_ZFS,
131 	DMU_OST_ZVOL,
132 	DMU_OST_OTHER,			/* For testing only! */
133 	DMU_OST_ANY,			/* Be careful! */
134 	DMU_OST_NUMTYPES
135 } dmu_objset_type_t;
136 
137 void byteswap_uint64_array(void *buf, size_t size);
138 void byteswap_uint32_array(void *buf, size_t size);
139 void byteswap_uint16_array(void *buf, size_t size);
140 void byteswap_uint8_array(void *buf, size_t size);
141 void zap_byteswap(void *buf, size_t size);
142 void zfs_oldacl_byteswap(void *buf, size_t size);
143 void zfs_acl_byteswap(void *buf, size_t size);
144 void zfs_znode_byteswap(void *buf, size_t size);
145 
146 #define	DS_FIND_SNAPSHOTS	(1<<0)
147 #define	DS_FIND_CHILDREN	(1<<1)
148 
149 /*
150  * The maximum number of bytes that can be accessed as part of one
151  * operation, including metadata.
152  */
153 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
154 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
155 
156 #define	DMU_USERUSED_OBJECT	(-1ULL)
157 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
158 #define	DMU_DEADLIST_OBJECT	(-3ULL)
159 
160 /*
161  * Public routines to create, destroy, open, and close objsets.
162  */
163 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
164 int dmu_objset_own(const char *name, dmu_objset_type_t type,
165     boolean_t readonly, void *tag, objset_t **osp);
166 void dmu_objset_rele(objset_t *os, void *tag);
167 void dmu_objset_disown(objset_t *os, void *tag);
168 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
169 
170 int dmu_objset_evict_dbufs(objset_t *os);
171 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
172     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
173 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
174     uint64_t flags);
175 int dmu_objset_destroy(const char *name, boolean_t defer);
176 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
177 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
178     boolean_t recursive);
179 int dmu_objset_rename(const char *name, const char *newname,
180     boolean_t recursive);
181 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
182     int flags);
183 void dmu_objset_byteswap(void *buf, size_t size);
184 
185 typedef struct dmu_buf {
186 	uint64_t db_object;		/* object that this buffer is part of */
187 	uint64_t db_offset;		/* byte offset in this object */
188 	uint64_t db_size;		/* size of buffer in bytes */
189 	void *db_data;			/* data in buffer */
190 } dmu_buf_t;
191 
192 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
193 
194 /*
195  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
196  */
197 #define	DMU_POOL_DIRECTORY_OBJECT	1
198 #define	DMU_POOL_CONFIG			"config"
199 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
200 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
201 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
202 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
203 #define	DMU_POOL_SPARES			"spares"
204 #define	DMU_POOL_DEFLATE		"deflate"
205 #define	DMU_POOL_HISTORY		"history"
206 #define	DMU_POOL_PROPS			"pool_props"
207 #define	DMU_POOL_L2CACHE		"l2cache"
208 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
209 #define	DMU_POOL_DDT			"DDT-%s-%s-%s"
210 #define	DMU_POOL_DDT_STATS		"DDT-statistics"
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 to write a block: checksum, compression, number of copies, etc.
309  */
310 #define	WP_NOFILL	0x1
311 #define	WP_DMU_SYNC	0x2
312 
313 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
314     struct zio_prop *zp);
315 /*
316  * The bonus data is accessed more or less like a regular buffer.
317  * You must dmu_bonus_hold() to get the buffer, which will give you a
318  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
319  * data.  As with any normal buffer, you must call dmu_buf_read() to
320  * read db_data, dmu_buf_will_dirty() before modifying it, and the
321  * object must be held in an assigned transaction before calling
322  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
323  * buffer as well.  You must release your hold with dmu_buf_rele().
324  */
325 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
326 int dmu_bonus_max(void);
327 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
328 
329 /*
330  * Obtain the DMU buffer from the specified object which contains the
331  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
332  * that it will remain in memory.  You must release the hold with
333  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
334  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
335  *
336  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
337  * on the returned buffer before reading or writing the buffer's
338  * db_data.  The comments for those routines describe what particular
339  * operations are valid after calling them.
340  *
341  * The object number must be a valid, allocated object number.
342  */
343 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
344     void *tag, dmu_buf_t **);
345 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
346 void dmu_buf_rele(dmu_buf_t *db, void *tag);
347 uint64_t dmu_buf_refcount(dmu_buf_t *db);
348 
349 /*
350  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
351  * range of an object.  A pointer to an array of dmu_buf_t*'s is
352  * returned (in *dbpp).
353  *
354  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
355  * frees the array.  The hold on the array of buffers MUST be released
356  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
357  * individually with dmu_buf_rele.
358  */
359 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
360     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
361 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
362 
363 /*
364  * Returns NULL on success, or the existing user ptr if it's already
365  * been set.
366  *
367  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
368  *
369  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
370  * will be set to db->db_data when you are allowed to access it.  Note
371  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
372  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
373  * *user_data_ptr_ptr will be set to the new value when it changes.
374  *
375  * If non-NULL, pageout func will be called when this buffer is being
376  * excised from the cache, so that you can clean up the data structure
377  * pointed to by user_ptr.
378  *
379  * dmu_evict_user() will call the pageout func for all buffers in a
380  * objset with a given pageout func.
381  */
382 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
383     dmu_buf_evict_func_t *pageout_func);
384 /*
385  * set_user_ie is the same as set_user, but request immediate eviction
386  * when hold count goes to zero.
387  */
388 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
389     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
390 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
391     void *user_ptr, void *user_data_ptr_ptr,
392     dmu_buf_evict_func_t *pageout_func);
393 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
394 
395 /*
396  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
397  */
398 void *dmu_buf_get_user(dmu_buf_t *db);
399 
400 /*
401  * Indicate that you are going to modify the buffer's data (db_data).
402  *
403  * The transaction (tx) must be assigned to a txg (ie. you've called
404  * dmu_tx_assign()).  The buffer's object must be held in the tx
405  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
406  */
407 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
408 
409 /*
410  * Tells if the given dbuf is freeable.
411  */
412 boolean_t dmu_buf_freeable(dmu_buf_t *);
413 
414 /*
415  * You must create a transaction, then hold the objects which you will
416  * (or might) modify as part of this transaction.  Then you must assign
417  * the transaction to a transaction group.  Once the transaction has
418  * been assigned, you can modify buffers which belong to held objects as
419  * part of this transaction.  You can't modify buffers before the
420  * transaction has been assigned; you can't modify buffers which don't
421  * belong to objects which this transaction holds; you can't hold
422  * objects once the transaction has been assigned.  You may hold an
423  * object which you are going to free (with dmu_object_free()), but you
424  * don't have to.
425  *
426  * You can abort the transaction before it has been assigned.
427  *
428  * Note that you may hold buffers (with dmu_buf_hold) at any time,
429  * regardless of transaction state.
430  */
431 
432 #define	DMU_NEW_OBJECT	(-1ULL)
433 #define	DMU_OBJECT_END	(-1ULL)
434 
435 dmu_tx_t *dmu_tx_create(objset_t *os);
436 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
437 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
438     uint64_t len);
439 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
440 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
441 void dmu_tx_abort(dmu_tx_t *tx);
442 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
443 void dmu_tx_wait(dmu_tx_t *tx);
444 void dmu_tx_commit(dmu_tx_t *tx);
445 
446 /*
447  * To register a commit callback, dmu_tx_callback_register() must be called.
448  *
449  * dcb_data is a pointer to caller private data that is passed on as a
450  * callback parameter. The caller is responsible for properly allocating and
451  * freeing it.
452  *
453  * When registering a callback, the transaction must be already created, but
454  * it cannot be committed or aborted. It can be assigned to a txg or not.
455  *
456  * The callback will be called after the transaction has been safely written
457  * to stable storage and will also be called if the dmu_tx is aborted.
458  * If there is any error which prevents the transaction from being committed to
459  * disk, the callback will be called with a value of error != 0.
460  */
461 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
462 
463 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
464     void *dcb_data);
465 
466 /*
467  * Free up the data blocks for a defined range of a file.  If size is
468  * zero, the range from offset to end-of-file is freed.
469  */
470 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
471 	uint64_t size, dmu_tx_t *tx);
472 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
473 	uint64_t size);
474 int dmu_free_object(objset_t *os, uint64_t object);
475 
476 /*
477  * Convenience functions.
478  *
479  * Canfail routines will return 0 on success, or an errno if there is a
480  * nonrecoverable I/O error.
481  */
482 #define	DMU_READ_PREFETCH	0 /* prefetch */
483 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
484 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
485 	void *buf, uint32_t flags);
486 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
487 	const void *buf, dmu_tx_t *tx);
488 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
489 	dmu_tx_t *tx);
490 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
491 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
492     dmu_tx_t *tx);
493 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
494     uint64_t size, struct page *pp, dmu_tx_t *tx);
495 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
496 void dmu_return_arcbuf(struct arc_buf *buf);
497 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
498     dmu_tx_t *tx);
499 
500 extern int zfs_prefetch_disable;
501 
502 /*
503  * Asynchronously try to read in the data.
504  */
505 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
506     uint64_t len);
507 
508 typedef struct dmu_object_info {
509 	/* All sizes are in bytes unless otherwise indicated. */
510 	uint32_t doi_data_block_size;
511 	uint32_t doi_metadata_block_size;
512 	dmu_object_type_t doi_type;
513 	dmu_object_type_t doi_bonus_type;
514 	uint64_t doi_bonus_size;
515 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
516 	uint8_t doi_checksum;
517 	uint8_t doi_compress;
518 	uint8_t doi_pad[5];
519 	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
520 	uint64_t doi_max_offset;
521 	uint64_t doi_fill_count;		/* number of non-empty blocks */
522 } dmu_object_info_t;
523 
524 typedef void arc_byteswap_func_t(void *buf, size_t size);
525 
526 typedef struct dmu_object_type_info {
527 	arc_byteswap_func_t	*ot_byteswap;
528 	boolean_t		ot_metadata;
529 	char			*ot_name;
530 } dmu_object_type_info_t;
531 
532 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
533 
534 /*
535  * Get information on a DMU object.
536  *
537  * Return 0 on success or ENOENT if object is not allocated.
538  *
539  * If doi is NULL, just indicates whether the object exists.
540  */
541 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
542 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
543 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
544 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
545     u_longlong_t *nblk512);
546 
547 typedef struct dmu_objset_stats {
548 	uint64_t dds_num_clones; /* number of clones of this */
549 	uint64_t dds_creation_txg;
550 	uint64_t dds_guid;
551 	dmu_objset_type_t dds_type;
552 	uint8_t dds_is_snapshot;
553 	uint8_t dds_inconsistent;
554 	char dds_origin[MAXNAMELEN];
555 } dmu_objset_stats_t;
556 
557 /*
558  * Get stats on a dataset.
559  */
560 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
561 
562 /*
563  * Add entries to the nvlist for all the objset's properties.  See
564  * zfs_prop_table[] and zfs(1m) for details on the properties.
565  */
566 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
567 
568 /*
569  * Get the space usage statistics for statvfs().
570  *
571  * refdbytes is the amount of space "referenced" by this objset.
572  * availbytes is the amount of space available to this objset, taking
573  * into account quotas & reservations, assuming that no other objsets
574  * use the space first.  These values correspond to the 'referenced' and
575  * 'available' properties, described in the zfs(1m) manpage.
576  *
577  * usedobjs and availobjs are the number of objects currently allocated,
578  * and available.
579  */
580 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
581     uint64_t *usedobjsp, uint64_t *availobjsp);
582 
583 /*
584  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
585  * (Contrast with the ds_guid which is a 64-bit ID that will never
586  * change, so there is a small probability that it will collide.)
587  */
588 uint64_t dmu_objset_fsid_guid(objset_t *os);
589 
590 /*
591  * Get the [cm]time for an objset's snapshot dir
592  */
593 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
594 
595 int dmu_objset_is_snapshot(objset_t *os);
596 
597 extern struct spa *dmu_objset_spa(objset_t *os);
598 extern struct zilog *dmu_objset_zil(objset_t *os);
599 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
600 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
601 extern void dmu_objset_name(objset_t *os, char *buf);
602 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
603 extern uint64_t dmu_objset_id(objset_t *os);
604 extern uint64_t dmu_objset_logbias(objset_t *os);
605 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
606     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
607 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
608     int maxlen, boolean_t *conflict);
609 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
610     uint64_t *idp, uint64_t *offp);
611 
612 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
613     void *bonus, uint64_t *userp, uint64_t *groupp);
614 extern void dmu_objset_register_type(dmu_objset_type_t ost,
615     objset_used_cb_t *cb);
616 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
617 extern void *dmu_objset_get_user(objset_t *os);
618 
619 /*
620  * Return the txg number for the given assigned transaction.
621  */
622 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
623 
624 /*
625  * Synchronous write.
626  * If a parent zio is provided this function initiates a write on the
627  * provided buffer as a child of the parent zio.
628  * In the absence of a parent zio, the write is completed synchronously.
629  * At write completion, blk is filled with the bp of the written block.
630  * Note that while the data covered by this function will be on stable
631  * storage when the write completes this new data does not become a
632  * permanent part of the file until the associated transaction commits.
633  */
634 
635 /*
636  * {zfs,zvol,ztest}_get_done() args
637  */
638 typedef struct zgd {
639 	struct zilog	*zgd_zilog;
640 	struct blkptr	*zgd_bp;
641 	dmu_buf_t	*zgd_db;
642 	struct rl	*zgd_rl;
643 	void		*zgd_private;
644 } zgd_t;
645 
646 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
647 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
648 
649 /*
650  * Find the next hole or data block in file starting at *off
651  * Return found offset in *off. Return ESRCH for end of file.
652  */
653 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
654     uint64_t *off);
655 
656 /*
657  * Initial setup and final teardown.
658  */
659 extern void dmu_init(void);
660 extern void dmu_fini(void);
661 
662 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
663     uint64_t object, uint64_t offset, int len);
664 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
665     dmu_traverse_cb_t cb, void *arg);
666 
667 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
668     struct vnode *vp, offset_t *off);
669 
670 typedef struct dmu_recv_cookie {
671 	/*
672 	 * This structure is opaque!
673 	 *
674 	 * If logical and real are different, we are recving the stream
675 	 * into the "real" temporary clone, and then switching it with
676 	 * the "logical" target.
677 	 */
678 	struct dsl_dataset *drc_logical_ds;
679 	struct dsl_dataset *drc_real_ds;
680 	struct drr_begin *drc_drrb;
681 	char *drc_tosnap;
682 	char *drc_top_ds;
683 	boolean_t drc_newfs;
684 	boolean_t drc_force;
685 } dmu_recv_cookie_t;
686 
687 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
688     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
689 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
690 int dmu_recv_end(dmu_recv_cookie_t *drc);
691 
692 /* CRC64 table */
693 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
694 extern uint64_t zfs_crc64_table[256];
695 
696 #ifdef	__cplusplus
697 }
698 #endif
699 
700 #endif	/* _SYS_DMU_H */
701