xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 148434217c040ea38dc844384f6ba68d9b325906)
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 
64 typedef struct objset objset_t;
65 typedef struct dmu_tx dmu_tx_t;
66 typedef struct dsl_dir dsl_dir_t;
67 
68 typedef enum dmu_object_type {
69 	DMU_OT_NONE,
70 	/* general: */
71 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
72 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
73 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
74 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
75 	DMU_OT_BPLIST,			/* UINT64 */
76 	DMU_OT_BPLIST_HDR,		/* UINT64 */
77 	/* spa: */
78 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
79 	DMU_OT_SPACE_MAP,		/* UINT64 */
80 	/* zil: */
81 	DMU_OT_INTENT_LOG,		/* UINT64 */
82 	/* dmu: */
83 	DMU_OT_DNODE,			/* DNODE */
84 	DMU_OT_OBJSET,			/* OBJSET */
85 	/* dsl: */
86 	DMU_OT_DSL_DIR,			/* UINT64 */
87 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
88 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
89 	DMU_OT_DSL_PROPS,		/* ZAP */
90 	DMU_OT_DSL_DATASET,		/* UINT64 */
91 	/* zpl: */
92 	DMU_OT_ZNODE,			/* ZNODE */
93 	DMU_OT_OLDACL,			/* Old ACL */
94 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
95 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
96 	DMU_OT_MASTER_NODE,		/* ZAP */
97 	DMU_OT_UNLINKED_SET,		/* ZAP */
98 	/* zvol: */
99 	DMU_OT_ZVOL,			/* UINT8 */
100 	DMU_OT_ZVOL_PROP,		/* ZAP */
101 	/* other; for testing only! */
102 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
103 	DMU_OT_UINT64_OTHER,		/* UINT64 */
104 	DMU_OT_ZAP_OTHER,		/* ZAP */
105 	/* new object types: */
106 	DMU_OT_ERROR_LOG,		/* ZAP */
107 	DMU_OT_SPA_HISTORY,		/* UINT8 */
108 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
109 	DMU_OT_POOL_PROPS,		/* ZAP */
110 	DMU_OT_DSL_PERMS,		/* ZAP */
111 	DMU_OT_ACL,			/* ACL */
112 	DMU_OT_SYSACL,			/* SYSACL */
113 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
114 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
115 	DMU_OT_NEXT_CLONES,		/* ZAP */
116 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
117 	DMU_OT_USERGROUP_USED,		/* ZAP */
118 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
119 	DMU_OT_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 #define	DMU_USERUSED_OBJECT	(-1ULL)
162 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
163 
164 /*
165  * Public routines to create, destroy, open, and close objsets.
166  */
167 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
168     objset_t **osp);
169 int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
170     objset_t **osp);
171 void dmu_objset_close(objset_t *os);
172 int dmu_objset_evict_dbufs(objset_t *os);
173 int dmu_objset_create(const char *name, dmu_objset_type_t type,
174     objset_t *clone_parent, uint64_t flags,
175     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
176 int dmu_objset_destroy(const char *name);
177 int dmu_snapshots_destroy(char *fsname, char *snapname);
178 int dmu_objset_rollback(objset_t *os);
179 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
180     boolean_t recursive);
181 int dmu_objset_rename(const char *name, const char *newname,
182     boolean_t recursive);
183 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
184     int flags);
185 void dmu_objset_byteswap(void *buf, size_t size);
186 
187 typedef struct dmu_buf {
188 	uint64_t db_object;		/* object that this buffer is part of */
189 	uint64_t db_offset;		/* byte offset in this object */
190 	uint64_t db_size;		/* size of buffer in bytes */
191 	void *db_data;			/* data in buffer */
192 } dmu_buf_t;
193 
194 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
195 
196 /*
197  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
198  */
199 #define	DMU_POOL_DIRECTORY_OBJECT	1
200 #define	DMU_POOL_CONFIG			"config"
201 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
202 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
203 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
204 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
205 #define	DMU_POOL_SPARES			"spares"
206 #define	DMU_POOL_DEFLATE		"deflate"
207 #define	DMU_POOL_HISTORY		"history"
208 #define	DMU_POOL_PROPS			"pool_props"
209 #define	DMU_POOL_L2CACHE		"l2cache"
210 
211 /* 4x8 zbookmark_t */
212 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
213 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
214 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
215 /* 1x8 txg */
216 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
217 /* 1x8 txg */
218 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
219 /* 1x4 enum scrub_func */
220 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
221 /* 1x8 count */
222 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
223 
224 /*
225  * Allocate an object from this objset.  The range of object numbers
226  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
227  *
228  * The transaction must be assigned to a txg.  The newly allocated
229  * object will be "held" in the transaction (ie. you can modify the
230  * newly allocated object in this transaction).
231  *
232  * dmu_object_alloc() chooses an object and returns it in *objectp.
233  *
234  * dmu_object_claim() allocates a specific object number.  If that
235  * number is already allocated, it fails and returns EEXIST.
236  *
237  * Return 0 on success, or ENOSPC or EEXIST as specified above.
238  */
239 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
240     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
241 int dmu_object_claim(objset_t *os, uint64_t object, 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_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
244     int blocksize, dmu_object_type_t bonustype, int bonuslen);
245 
246 /*
247  * Free an object from this objset.
248  *
249  * The object's data will be freed as well (ie. you don't need to call
250  * dmu_free(object, 0, -1, tx)).
251  *
252  * The object need not be held in the transaction.
253  *
254  * If there are any holds on this object's buffers (via dmu_buf_hold()),
255  * or tx holds on the object (via dmu_tx_hold_object()), you can not
256  * free it; it fails and returns EBUSY.
257  *
258  * If the object is not allocated, it fails and returns ENOENT.
259  *
260  * Return 0 on success, or EBUSY or ENOENT as specified above.
261  */
262 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
263 
264 /*
265  * Find the next allocated or free object.
266  *
267  * The objectp parameter is in-out.  It will be updated to be the next
268  * object which is allocated.  Ignore objects which have not been
269  * modified since txg.
270  *
271  * XXX Can only be called on a objset with no dirty data.
272  *
273  * Returns 0 on success, or ENOENT if there are no more objects.
274  */
275 int dmu_object_next(objset_t *os, uint64_t *objectp,
276     boolean_t hole, uint64_t txg);
277 
278 /*
279  * Set the data blocksize for an object.
280  *
281  * The object cannot have any blocks allcated beyond the first.  If
282  * the first block is allocated already, the new size must be greater
283  * than the current block size.  If these conditions are not met,
284  * ENOTSUP will be returned.
285  *
286  * Returns 0 on success, or EBUSY if there are any holds on the object
287  * contents, or ENOTSUP as described above.
288  */
289 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
290     int ibs, dmu_tx_t *tx);
291 
292 /*
293  * Set the checksum property on a dnode.  The new checksum algorithm will
294  * apply to all newly written blocks; existing blocks will not be affected.
295  */
296 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
297     dmu_tx_t *tx);
298 
299 /*
300  * Set the compress property on a dnode.  The new compression algorithm will
301  * apply to all newly written blocks; existing blocks will not be affected.
302  */
303 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
304     dmu_tx_t *tx);
305 
306 /*
307  * Decide how many copies of a given block we should make.  Can be from
308  * 1 to SPA_DVAS_PER_BP.
309  */
310 int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
311     dmu_object_type_t ot);
312 /*
313  * The bonus data is accessed more or less like a regular buffer.
314  * You must dmu_bonus_hold() to get the buffer, which will give you a
315  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
316  * data.  As with any normal buffer, you must call dmu_buf_read() to
317  * read db_data, dmu_buf_will_dirty() before modifying it, and the
318  * object must be held in an assigned transaction before calling
319  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
320  * buffer as well.  You must release your hold with dmu_buf_rele().
321  */
322 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
323 int dmu_bonus_max(void);
324 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
325 
326 /*
327  * Obtain the DMU buffer from the specified object which contains the
328  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
329  * that it will remain in memory.  You must release the hold with
330  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
331  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
332  *
333  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
334  * on the returned buffer before reading or writing the buffer's
335  * db_data.  The comments for those routines describe what particular
336  * operations are valid after calling them.
337  *
338  * The object number must be a valid, allocated object number.
339  */
340 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
341     void *tag, dmu_buf_t **);
342 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
343 void dmu_buf_rele(dmu_buf_t *db, void *tag);
344 uint64_t dmu_buf_refcount(dmu_buf_t *db);
345 
346 /*
347  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
348  * range of an object.  A pointer to an array of dmu_buf_t*'s is
349  * returned (in *dbpp).
350  *
351  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
352  * frees the array.  The hold on the array of buffers MUST be released
353  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
354  * individually with dmu_buf_rele.
355  */
356 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
357     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
358 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
359 
360 /*
361  * Returns NULL on success, or the existing user ptr if it's already
362  * been set.
363  *
364  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
365  *
366  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
367  * will be set to db->db_data when you are allowed to access it.  Note
368  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
369  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
370  * *user_data_ptr_ptr will be set to the new value when it changes.
371  *
372  * If non-NULL, pageout func will be called when this buffer is being
373  * excised from the cache, so that you can clean up the data structure
374  * pointed to by user_ptr.
375  *
376  * dmu_evict_user() will call the pageout func for all buffers in a
377  * objset with a given pageout func.
378  */
379 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
380     dmu_buf_evict_func_t *pageout_func);
381 /*
382  * set_user_ie is the same as set_user, but request immediate eviction
383  * when hold count goes to zero.
384  */
385 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
386     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
387 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
388     void *user_ptr, void *user_data_ptr_ptr,
389     dmu_buf_evict_func_t *pageout_func);
390 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
391 
392 /*
393  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
394  */
395 void *dmu_buf_get_user(dmu_buf_t *db);
396 
397 /*
398  * Indicate that you are going to modify the buffer's data (db_data).
399  *
400  * The transaction (tx) must be assigned to a txg (ie. you've called
401  * dmu_tx_assign()).  The buffer's object must be held in the tx
402  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
403  */
404 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
405 
406 /*
407  * You must create a transaction, then hold the objects which you will
408  * (or might) modify as part of this transaction.  Then you must assign
409  * the transaction to a transaction group.  Once the transaction has
410  * been assigned, you can modify buffers which belong to held objects as
411  * part of this transaction.  You can't modify buffers before the
412  * transaction has been assigned; you can't modify buffers which don't
413  * belong to objects which this transaction holds; you can't hold
414  * objects once the transaction has been assigned.  You may hold an
415  * object which you are going to free (with dmu_object_free()), but you
416  * don't have to.
417  *
418  * You can abort the transaction before it has been assigned.
419  *
420  * Note that you may hold buffers (with dmu_buf_hold) at any time,
421  * regardless of transaction state.
422  */
423 
424 #define	DMU_NEW_OBJECT	(-1ULL)
425 #define	DMU_OBJECT_END	(-1ULL)
426 
427 dmu_tx_t *dmu_tx_create(objset_t *os);
428 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
429 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
430     uint64_t len);
431 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
432 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
433 void dmu_tx_abort(dmu_tx_t *tx);
434 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
435 void dmu_tx_wait(dmu_tx_t *tx);
436 void dmu_tx_commit(dmu_tx_t *tx);
437 
438 /*
439  * Free up the data blocks for a defined range of a file.  If size is
440  * zero, the range from offset to end-of-file is freed.
441  */
442 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
443 	uint64_t size, dmu_tx_t *tx);
444 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
445 	uint64_t size);
446 int dmu_free_object(objset_t *os, uint64_t object);
447 
448 /*
449  * Convenience functions.
450  *
451  * Canfail routines will return 0 on success, or an errno if there is a
452  * nonrecoverable I/O error.
453  */
454 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
455 	void *buf);
456 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
457 	const void *buf, dmu_tx_t *tx);
458 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
459 	dmu_tx_t *tx);
460 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
461 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
462     dmu_tx_t *tx);
463 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
464     uint64_t size, struct page *pp, dmu_tx_t *tx);
465 
466 extern int zfs_prefetch_disable;
467 
468 /*
469  * Asynchronously try to read in the data.
470  */
471 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
472     uint64_t len);
473 
474 typedef struct dmu_object_info {
475 	/* All sizes are in bytes. */
476 	uint32_t doi_data_block_size;
477 	uint32_t doi_metadata_block_size;
478 	uint64_t doi_bonus_size;
479 	dmu_object_type_t doi_type;
480 	dmu_object_type_t doi_bonus_type;
481 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
482 	uint8_t doi_checksum;
483 	uint8_t doi_compress;
484 	uint8_t doi_pad[5];
485 	/* Values below are number of 512-byte blocks. */
486 	uint64_t doi_physical_blks;		/* data + metadata */
487 	uint64_t doi_max_block_offset;
488 } dmu_object_info_t;
489 
490 typedef void arc_byteswap_func_t(void *buf, size_t size);
491 
492 typedef struct dmu_object_type_info {
493 	arc_byteswap_func_t	*ot_byteswap;
494 	boolean_t		ot_metadata;
495 	char			*ot_name;
496 } dmu_object_type_info_t;
497 
498 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
499 
500 /*
501  * Get information on a DMU object.
502  *
503  * Return 0 on success or ENOENT if object is not allocated.
504  *
505  * If doi is NULL, just indicates whether the object exists.
506  */
507 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
508 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
509 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
510 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
511     u_longlong_t *nblk512);
512 
513 typedef struct dmu_objset_stats {
514 	uint64_t dds_num_clones; /* number of clones of this */
515 	uint64_t dds_creation_txg;
516 	uint64_t dds_guid;
517 	dmu_objset_type_t dds_type;
518 	uint8_t dds_is_snapshot;
519 	uint8_t dds_inconsistent;
520 	char dds_origin[MAXNAMELEN];
521 } dmu_objset_stats_t;
522 
523 /*
524  * Get stats on a dataset.
525  */
526 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
527 
528 /*
529  * Add entries to the nvlist for all the objset's properties.  See
530  * zfs_prop_table[] and zfs(1m) for details on the properties.
531  */
532 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
533 
534 /*
535  * Get the space usage statistics for statvfs().
536  *
537  * refdbytes is the amount of space "referenced" by this objset.
538  * availbytes is the amount of space available to this objset, taking
539  * into account quotas & reservations, assuming that no other objsets
540  * use the space first.  These values correspond to the 'referenced' and
541  * 'available' properties, described in the zfs(1m) manpage.
542  *
543  * usedobjs and availobjs are the number of objects currently allocated,
544  * and available.
545  */
546 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
547     uint64_t *usedobjsp, uint64_t *availobjsp);
548 
549 /*
550  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
551  * (Contrast with the ds_guid which is a 64-bit ID that will never
552  * change, so there is a small probability that it will collide.)
553  */
554 uint64_t dmu_objset_fsid_guid(objset_t *os);
555 
556 int dmu_objset_is_snapshot(objset_t *os);
557 
558 extern struct spa *dmu_objset_spa(objset_t *os);
559 extern struct zilog *dmu_objset_zil(objset_t *os);
560 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
561 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
562 extern void dmu_objset_name(objset_t *os, char *buf);
563 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
564 extern uint64_t dmu_objset_id(objset_t *os);
565 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
566     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
567 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
568     int maxlen, boolean_t *conflict);
569 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
570     uint64_t *idp, uint64_t *offp);
571 
572 typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
573     void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
574     dmu_tx_t *tx);
575 extern void dmu_objset_register_type(dmu_objset_type_t ost,
576     objset_used_cb_t *cb);
577 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
578 extern void *dmu_objset_get_user(objset_t *os);
579 
580 /*
581  * Return the txg number for the given assigned transaction.
582  */
583 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
584 
585 /*
586  * Synchronous write.
587  * If a parent zio is provided this function initiates a write on the
588  * provided buffer as a child of the parent zio.
589  * In the absence of a parent zio, the write is completed synchronously.
590  * At write completion, blk is filled with the bp of the written block.
591  * Note that while the data covered by this function will be on stable
592  * storage when the write completes this new data does not become a
593  * permanent part of the file until the associated transaction commits.
594  */
595 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
596 int dmu_sync(struct zio *zio, dmu_buf_t *db,
597     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
598 
599 /*
600  * Find the next hole or data block in file starting at *off
601  * Return found offset in *off. Return ESRCH for end of file.
602  */
603 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
604     uint64_t *off);
605 
606 /*
607  * Initial setup and final teardown.
608  */
609 extern void dmu_init(void);
610 extern void dmu_fini(void);
611 
612 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
613     uint64_t object, uint64_t offset, int len);
614 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
615     dmu_traverse_cb_t cb, void *arg);
616 
617 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
618     struct vnode *vp, offset_t *off);
619 
620 typedef struct dmu_recv_cookie {
621 	/*
622 	 * This structure is opaque!
623 	 *
624 	 * If logical and real are different, we are recving the stream
625 	 * into the "real" temporary clone, and then switching it with
626 	 * the "logical" target.
627 	 */
628 	struct dsl_dataset *drc_logical_ds;
629 	struct dsl_dataset *drc_real_ds;
630 	struct drr_begin *drc_drrb;
631 	char *drc_tosnap;
632 	boolean_t drc_newfs;
633 	boolean_t drc_force;
634 } dmu_recv_cookie_t;
635 
636 int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
637     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
638 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
639 int dmu_recv_end(dmu_recv_cookie_t *drc);
640 void dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
641 
642 /* CRC64 table */
643 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
644 extern uint64_t zfs_crc64_table[256];
645 
646 #ifdef	__cplusplus
647 }
648 #endif
649 
650 #endif	/* _SYS_DMU_H */
651