xref: /titanic_41/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 264a6e7478846334593be7663fb6b1a8f37784a0)
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 2006 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 
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 
63 typedef struct objset objset_t;
64 typedef struct dmu_tx dmu_tx_t;
65 typedef struct dsl_dir dsl_dir_t;
66 
67 typedef enum dmu_object_type {
68 	DMU_OT_NONE,
69 	/* general: */
70 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
71 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
72 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
73 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
74 	DMU_OT_BPLIST,			/* UINT64 */
75 	DMU_OT_BPLIST_HDR,		/* UINT64 */
76 	/* spa: */
77 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
78 	DMU_OT_SPACE_MAP,		/* UINT64 */
79 	/* zil: */
80 	DMU_OT_INTENT_LOG,		/* UINT64 */
81 	/* dmu: */
82 	DMU_OT_DNODE,			/* DNODE */
83 	DMU_OT_OBJSET,			/* OBJSET */
84 	/* dsl: */
85 	DMU_OT_DSL_DIR,			/* UINT64 */
86 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
87 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
88 	DMU_OT_DSL_PROPS,		/* ZAP */
89 	DMU_OT_DSL_DATASET,		/* UINT64 */
90 	/* zpl: */
91 	DMU_OT_ZNODE,			/* ZNODE */
92 	DMU_OT_ACL,			/* ACL */
93 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
94 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
95 	DMU_OT_MASTER_NODE,		/* ZAP */
96 	DMU_OT_DELETE_QUEUE,		/* ZAP */
97 	/* zvol: */
98 	DMU_OT_ZVOL,			/* UINT8 */
99 	DMU_OT_ZVOL_PROP,		/* ZAP */
100 	/* other; for testing only! */
101 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
102 	DMU_OT_UINT64_OTHER,		/* UINT64 */
103 	DMU_OT_ZAP_OTHER,		/* ZAP */
104 	/* new object types: */
105 	DMU_OT_ERROR_LOG,		/* ZAP */
106 
107 	DMU_OT_NUMTYPES
108 } dmu_object_type_t;
109 
110 typedef enum dmu_objset_type {
111 	DMU_OST_NONE,
112 	DMU_OST_META,
113 	DMU_OST_ZFS,
114 	DMU_OST_ZVOL,
115 	DMU_OST_OTHER,			/* For testing only! */
116 	DMU_OST_ANY,			/* Be careful! */
117 	DMU_OST_NUMTYPES
118 } dmu_objset_type_t;
119 
120 void byteswap_uint64_array(void *buf, size_t size);
121 void byteswap_uint32_array(void *buf, size_t size);
122 void byteswap_uint16_array(void *buf, size_t size);
123 void byteswap_uint8_array(void *buf, size_t size);
124 void zap_byteswap(void *buf, size_t size);
125 void zfs_acl_byteswap(void *buf, size_t size);
126 void zfs_znode_byteswap(void *buf, size_t size);
127 
128 #define	DS_MODE_NONE		0	/* invalid, to aid debugging */
129 #define	DS_MODE_STANDARD	1	/* normal access, no special needs */
130 #define	DS_MODE_PRIMARY		2	/* the "main" access, e.g. a mount */
131 #define	DS_MODE_EXCLUSIVE	3	/* exclusive access, e.g. to destroy */
132 #define	DS_MODE_LEVELS		4
133 #define	DS_MODE_LEVEL(x)	((x) & (DS_MODE_LEVELS - 1))
134 #define	DS_MODE_READONLY	0x8
135 #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
136 #define	DS_MODE_INCONSISTENT	0x10
137 #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
138 
139 #define	DS_FIND_SNAPSHOTS	(1<<0)
140 #define	DS_FIND_CHILDREN	(1<<1)
141 
142 /*
143  * The maximum number of bytes that can be accessed as part of one
144  * operation, including metadata.
145  */
146 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
147 
148 /*
149  * Public routines to create, destroy, open, and close objsets.
150  */
151 int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
152     objset_t **osp);
153 void dmu_objset_close(objset_t *os);
154 int dmu_objset_evict_dbufs(objset_t *os, int try);
155 int dmu_objset_create(const char *name, dmu_objset_type_t type,
156     objset_t *clone_parent,
157     void (*func)(objset_t *os, void *arg, dmu_tx_t *tx), void *arg);
158 int dmu_objset_destroy(const char *name);
159 int dmu_snapshots_destroy(char *fsname, char *snapname);
160 int dmu_objset_rollback(const char *name);
161 int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
162 int dmu_objset_rename(const char *name, const char *newname);
163 void dmu_objset_set_quota(objset_t *os, uint64_t quota);
164 uint64_t dmu_objset_get_quota(objset_t *os);
165 int dmu_objset_request_reservation(objset_t *os, uint64_t reservation);
166 int dmu_objset_find(char *name, int func(char *, void *), void *arg,
167     int flags);
168 void dmu_objset_byteswap(void *buf, size_t size);
169 
170 typedef struct dmu_buf {
171 	uint64_t db_object;		/* object that this buffer is part of */
172 	uint64_t db_offset;		/* byte offset in this object */
173 	uint64_t db_size;		/* size of buffer in bytes */
174 	void *db_data;			/* data in buffer */
175 } dmu_buf_t;
176 
177 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
178 
179 /*
180  * Callback function to perform byte swapping on a block.
181  */
182 typedef void dmu_byteswap_func_t(void *buf, size_t size);
183 
184 /*
185  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
186  */
187 #define	DMU_POOL_DIRECTORY_OBJECT	1
188 #define	DMU_POOL_CONFIG			"config"
189 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
190 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
191 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
192 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
193 #define	DMU_POOL_SPARES			"spares"
194 #define	DMU_POOL_DEFLATE		"deflate"
195 
196 /*
197  * Allocate an object from this objset.  The range of object numbers
198  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
199  *
200  * The transaction must be assigned to a txg.  The newly allocated
201  * object will be "held" in the transaction (ie. you can modify the
202  * newly allocated object in this transaction).
203  *
204  * dmu_object_alloc() chooses an object and returns it in *objectp.
205  *
206  * dmu_object_claim() allocates a specific object number.  If that
207  * number is already allocated, it fails and returns EEXIST.
208  *
209  * Return 0 on success, or ENOSPC or EEXIST as specified above.
210  */
211 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
212     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
213 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
214     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
215 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
216     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
217 
218 /*
219  * Free an object from this objset.
220  *
221  * The object's data will be freed as well (ie. you don't need to call
222  * dmu_free(object, 0, -1, tx)).
223  *
224  * The object need not be held in the transaction.
225  *
226  * If there are any holds on this object's buffers (via dmu_buf_hold()),
227  * or tx holds on the object (via dmu_tx_hold_object()), you can not
228  * free it; it fails and returns EBUSY.
229  *
230  * If the object is not allocated, it fails and returns ENOENT.
231  *
232  * Return 0 on success, or EBUSY or ENOENT as specified above.
233  */
234 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
235 
236 /*
237  * Find the next allocated or free object.
238  *
239  * The objectp parameter is in-out.  It will be updated to be the next
240  * object which is allocated.
241  *
242  * XXX Can only be called on a objset with no dirty data.
243  *
244  * Returns 0 on success, or ENOENT if there are no more objects.
245  */
246 int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole);
247 
248 /*
249  * Set the data blocksize for an object.
250  *
251  * The object cannot have any blocks allcated beyond the first.  If
252  * the first block is allocated already, the new size must be greater
253  * than the current block size.  If these conditions are not met,
254  * ENOTSUP will be returned.
255  *
256  * Returns 0 on success, or EBUSY if there are any holds on the object
257  * contents, or ENOTSUP as described above.
258  */
259 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
260     int ibs, dmu_tx_t *tx);
261 
262 /*
263  * Set the checksum property on a dnode.  The new checksum algorithm will
264  * apply to all newly written blocks; existing blocks will not be affected.
265  */
266 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
267     dmu_tx_t *tx);
268 
269 /*
270  * Set the compress property on a dnode.  The new compression algorithm will
271  * apply to all newly written blocks; existing blocks will not be affected.
272  */
273 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
274     dmu_tx_t *tx);
275 
276 /*
277  * Decide how many copies of a given block we should make.  Can be from
278  * 1 to SPA_DVAS_PER_BP.
279  */
280 int dmu_get_replication_level(struct spa *spa, struct zbookmark *zb,
281     dmu_object_type_t ot);
282 /*
283  * The bonus data is accessed more or less like a regular buffer.
284  * You must dmu_bonus_hold() to get the buffer, which will give you a
285  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
286  * data.  As with any normal buffer, you must call dmu_buf_read() to
287  * read db_data, dmu_buf_will_dirty() before modifying it, and the
288  * object must be held in an assigned transaction before calling
289  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
290  * buffer as well.  You must release your hold with dmu_buf_rele().
291  */
292 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
293 int dmu_bonus_max(void);
294 
295 /*
296  * Obtain the DMU buffer from the specified object which contains the
297  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
298  * that it will remain in memory.  You must release the hold with
299  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
300  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
301  *
302  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
303  * on the returned buffer before reading or writing the buffer's
304  * db_data.  The comments for those routines describe what particular
305  * operations are valid after calling them.
306  *
307  * The object number must be a valid, allocated object number.
308  */
309 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
310     void *tag, dmu_buf_t **);
311 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
312 void dmu_buf_rele(dmu_buf_t *db, void *tag);
313 uint64_t dmu_buf_refcount(dmu_buf_t *db);
314 
315 /*
316  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
317  * range of an object.  A pointer to an array of dmu_buf_t*'s is
318  * returned (in *dbpp).
319  *
320  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
321  * frees the array.  The hold on the array of buffers MUST be released
322  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
323  * individually with dmu_buf_rele.
324  */
325 int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
326     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
327 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
328     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
329 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
330 
331 /*
332  * Returns NULL on success, or the existing user ptr if it's already
333  * been set.
334  *
335  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
336  *
337  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
338  * will be set to db->db_data when you are allowed to access it.  Note
339  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
340  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
341  * *user_data_ptr_ptr will be set to the new value when it changes.
342  *
343  * If non-NULL, pageout func will be called when this buffer is being
344  * excised from the cache, so that you can clean up the data structure
345  * pointed to by user_ptr.
346  *
347  * dmu_evict_user() will call the pageout func for all buffers in a
348  * objset with a given pageout func.
349  */
350 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
351     dmu_buf_evict_func_t *pageout_func);
352 /*
353  * set_user_ie is the same as set_user, but request immediate eviction
354  * when hold count goes to zero.
355  */
356 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
357     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
358 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
359     void *user_ptr, void *user_data_ptr_ptr,
360     dmu_buf_evict_func_t *pageout_func);
361 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
362 
363 void dmu_buf_hold_data(dmu_buf_t *db);
364 void dmu_buf_rele_data(dmu_buf_t *db);
365 
366 /*
367  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
368  */
369 void *dmu_buf_get_user(dmu_buf_t *db);
370 
371 /*
372  * Indicate that you are going to modify the buffer's data (db_data).
373  *
374  * The transaction (tx) must be assigned to a txg (ie. you've called
375  * dmu_tx_assign()).  The buffer's object must be held in the tx
376  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
377  */
378 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
379 
380 /*
381  * You must create a transaction, then hold the objects which you will
382  * (or might) modify as part of this transaction.  Then you must assign
383  * the transaction to a transaction group.  Once the transaction has
384  * been assigned, you can modify buffers which belong to held objects as
385  * part of this transaction.  You can't modify buffers before the
386  * transaction has been assigned; you can't modify buffers which don't
387  * belong to objects which this transaction holds; you can't hold
388  * objects once the transaction has been assigned.  You may hold an
389  * object which you are going to free (with dmu_object_free()), but you
390  * don't have to.
391  *
392  * You can abort the transaction before it has been assigned.
393  *
394  * Note that you may hold buffers (with dmu_buf_hold) at any time,
395  * regardless of transaction state.
396  */
397 
398 #define	DMU_NEW_OBJECT	(-1ULL)
399 #define	DMU_OBJECT_END	(-1ULL)
400 
401 dmu_tx_t *dmu_tx_create(objset_t *os);
402 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
403 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
404     uint64_t len);
405 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
406 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
407 void dmu_tx_abort(dmu_tx_t *tx);
408 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
409 void dmu_tx_wait(dmu_tx_t *tx);
410 void dmu_tx_commit(dmu_tx_t *tx);
411 
412 /*
413  * Free up the data blocks for a defined range of a file.  If size is
414  * zero, the range from offset to end-of-file is freed.
415  */
416 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
417 	uint64_t size, dmu_tx_t *tx);
418 
419 /*
420  * Convenience functions.
421  *
422  * Canfail routines will return 0 on success, or an errno if there is a
423  * nonrecoverable I/O error.
424  */
425 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
426 	void *buf);
427 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
428 	const void *buf, dmu_tx_t *tx);
429 int dmu_write_uio(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
430     struct uio *uio, dmu_tx_t *tx);
431 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
432     uint64_t size, struct page *pp, dmu_tx_t *tx);
433 
434 /*
435  * Asynchronously try to read in the data.
436  */
437 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
438     uint64_t len);
439 
440 typedef struct dmu_object_info {
441 	/* All sizes are in bytes. */
442 	uint32_t doi_data_block_size;
443 	uint32_t doi_metadata_block_size;
444 	uint64_t doi_bonus_size;
445 	dmu_object_type_t doi_type;
446 	dmu_object_type_t doi_bonus_type;
447 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
448 	uint8_t doi_checksum;
449 	uint8_t doi_compress;
450 	uint8_t doi_pad[5];
451 	/* Values below are number of 512-byte blocks. */
452 	uint64_t doi_physical_blks;		/* data + metadata */
453 	uint64_t doi_max_block_offset;
454 } dmu_object_info_t;
455 
456 typedef struct dmu_object_type_info {
457 	dmu_byteswap_func_t	*ot_byteswap;
458 	boolean_t		ot_metadata;
459 	char			*ot_name;
460 } dmu_object_type_info_t;
461 
462 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
463 
464 /*
465  * Get information on a DMU object.
466  *
467  * Return 0 on success or ENOENT if object is not allocated.
468  *
469  * If doi is NULL, just indicates whether the object exists.
470  */
471 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
472 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
473 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
474 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
475     u_longlong_t *nblk512);
476 
477 /*
478  * Get the maximum nonzero offset in the object (ie. this offset and all
479  * offsets following are zero).
480  *
481  * XXX Perhaps integrate this with dmu_object_info(), although that
482  * would then have to bring in the indirect blocks.
483  */
484 uint64_t dmu_object_max_nonzero_offset(objset_t *os, uint64_t object);
485 
486 typedef struct dmu_objset_stats {
487 	dmu_objset_type_t dds_type;
488 	uint8_t dds_is_snapshot;
489 	uint8_t dds_inconsistent;
490 	uint8_t dds_pad[2];
491 
492 	uint64_t dds_creation_time;
493 	uint64_t dds_creation_txg;
494 
495 	char dds_clone_of[MAXNAMELEN];
496 
497 	/* How much data is there in this objset? */
498 
499 	/*
500 	 * Space referenced, taking into account pending writes and
501 	 * frees.  Only relavent to filesystems and snapshots (not
502 	 * collections).
503 	 */
504 	uint64_t dds_space_refd;
505 
506 	/*
507 	 * Space "used", taking into account pending writes and frees, and
508 	 * children's reservations (in bytes).  This is the amount of
509 	 * space that will be freed if this and all dependent items are
510 	 * destroyed (eg. child datasets, objsets, and snapshots).  So
511 	 * for snapshots, this is the amount of space unique to this
512 	 * snapshot.
513 	 */
514 	uint64_t dds_space_used;
515 
516 	/*
517 	 * Compressed and uncompressed bytes consumed.  Does not take
518 	 * into account reservations.  Used for computing compression
519 	 * ratio.
520 	 */
521 	uint64_t dds_compressed_bytes;
522 	uint64_t dds_uncompressed_bytes;
523 
524 	/*
525 	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
526 	 * collisions.  The ds_guid is a 64-bit ID that will never
527 	 * change, so there is a small probability that it will collide.
528 	 */
529 	uint64_t dds_fsid_guid;
530 
531 	uint64_t dds_objects_used;	/* number of objects used */
532 	uint64_t dds_objects_avail;	/* number of objects available */
533 
534 	uint64_t dds_num_clones; /* number of clones of this */
535 
536 	/* The dataset's administratively-set quota, in bytes. */
537 	uint64_t dds_quota;
538 
539 	/* The dataset's administratively-set reservation, in bytes */
540 	uint64_t dds_reserved;
541 
542 	/*
543 	 * The amount of additional space that this dataset can consume.
544 	 * Takes into account quotas & reservations.
545 	 * (Assuming that no other datasets consume it first.)
546 	 */
547 	uint64_t dds_available;
548 
549 	/*
550 	 * Used for debugging purposes
551 	 */
552 	uint64_t dds_last_txg;
553 } dmu_objset_stats_t;
554 
555 /*
556  * Get stats on a dataset.
557  */
558 void dmu_objset_stats(objset_t *os, dmu_objset_stats_t *dds);
559 
560 int dmu_objset_is_snapshot(objset_t *os);
561 
562 extern struct spa *dmu_objset_spa(objset_t *os);
563 extern struct zilog *dmu_objset_zil(objset_t *os);
564 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
565 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
566 extern void dmu_objset_name(objset_t *os, char *buf);
567 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
568 extern uint64_t dmu_objset_id(objset_t *os);
569 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
570     uint64_t *id, uint64_t *offp);
571 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
572     uint64_t *idp, uint64_t *offp);
573 
574 /*
575  * Return the txg number for the given assigned transaction.
576  */
577 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
578 
579 /*
580  * Synchronous write.
581  * If a parent zio is provided this function initiates a write on the
582  * provided buffer as a child of the parent zio.
583  * In the absense of a parent zio, the write is completed synchronously.
584  * At write completion, blk is filled with the bp of the written block.
585  * Note that while the data covered by this function will be on stable
586  * storage when the write completes this new data does not become a
587  * permanent part of the file until the associated transaction commits.
588  */
589 typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
590 int dmu_sync(struct zio *zio, dmu_buf_t *db,
591     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
592 
593 /*
594  * Find the next hole or data block in file starting at *off
595  * Return found offset in *off. Return ESRCH for end of file.
596  */
597 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
598     uint64_t *off);
599 
600 /*
601  * Initial setup and final teardown.
602  */
603 extern void dmu_init(void);
604 extern void dmu_fini(void);
605 
606 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
607     uint64_t object, uint64_t offset, int len);
608 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
609     dmu_traverse_cb_t cb, void *arg);
610 
611 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp);
612 int dmu_recvbackup(char *tosnap, struct drr_begin *drrb, uint64_t *sizep,
613     boolean_t force, struct vnode *vp, uint64_t voffset);
614 
615 /* CRC64 table */
616 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
617 extern uint64_t zfs_crc64_table[256];
618 
619 #ifdef	__cplusplus
620 }
621 #endif
622 
623 #endif	/* _SYS_DMU_H */
624