xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 744947dc83c634d985ed3ad79ac9c5e28d1865fd)
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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23  */
24 
25 /* Portions Copyright 2010 Robert Milkowski */
26 
27 #ifndef	_SYS_DMU_H
28 #define	_SYS_DMU_H
29 
30 /*
31  * This file describes the interface that the DMU provides for its
32  * consumers.
33  *
34  * The DMU also interacts with the SPA.  That interface is described in
35  * dmu_spa.h.
36  */
37 
38 #include <sys/inttypes.h>
39 #include <sys/types.h>
40 #include <sys/param.h>
41 #include <sys/cred.h>
42 #include <sys/time.h>
43 
44 #ifdef	__cplusplus
45 extern "C" {
46 #endif
47 
48 struct uio;
49 struct xuio;
50 struct page;
51 struct vnode;
52 struct spa;
53 struct zilog;
54 struct zio;
55 struct blkptr;
56 struct zap_cursor;
57 struct dsl_dataset;
58 struct dsl_pool;
59 struct dnode;
60 struct drr_begin;
61 struct drr_end;
62 struct zbookmark;
63 struct spa;
64 struct nvlist;
65 struct arc_buf;
66 struct zio_prop;
67 struct sa_handle;
68 
69 typedef struct objset objset_t;
70 typedef struct dmu_tx dmu_tx_t;
71 typedef struct dsl_dir dsl_dir_t;
72 
73 typedef enum dmu_object_type {
74 	DMU_OT_NONE,
75 	/* general: */
76 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
77 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
78 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
79 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
80 	DMU_OT_BPOBJ,			/* UINT64 */
81 	DMU_OT_BPOBJ_HDR,		/* UINT64 */
82 	/* spa: */
83 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
84 	DMU_OT_SPACE_MAP,		/* UINT64 */
85 	/* zil: */
86 	DMU_OT_INTENT_LOG,		/* UINT64 */
87 	/* dmu: */
88 	DMU_OT_DNODE,			/* DNODE */
89 	DMU_OT_OBJSET,			/* OBJSET */
90 	/* dsl: */
91 	DMU_OT_DSL_DIR,			/* UINT64 */
92 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
93 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
94 	DMU_OT_DSL_PROPS,		/* ZAP */
95 	DMU_OT_DSL_DATASET,		/* UINT64 */
96 	/* zpl: */
97 	DMU_OT_ZNODE,			/* ZNODE */
98 	DMU_OT_OLDACL,			/* Old ACL */
99 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
100 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
101 	DMU_OT_MASTER_NODE,		/* ZAP */
102 	DMU_OT_UNLINKED_SET,		/* ZAP */
103 	/* zvol: */
104 	DMU_OT_ZVOL,			/* UINT8 */
105 	DMU_OT_ZVOL_PROP,		/* ZAP */
106 	/* other; for testing only! */
107 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
108 	DMU_OT_UINT64_OTHER,		/* UINT64 */
109 	DMU_OT_ZAP_OTHER,		/* ZAP */
110 	/* new object types: */
111 	DMU_OT_ERROR_LOG,		/* ZAP */
112 	DMU_OT_SPA_HISTORY,		/* UINT8 */
113 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
114 	DMU_OT_POOL_PROPS,		/* ZAP */
115 	DMU_OT_DSL_PERMS,		/* ZAP */
116 	DMU_OT_ACL,			/* ACL */
117 	DMU_OT_SYSACL,			/* SYSACL */
118 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
119 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
120 	DMU_OT_NEXT_CLONES,		/* ZAP */
121 	DMU_OT_SCAN_QUEUE,		/* ZAP */
122 	DMU_OT_USERGROUP_USED,		/* ZAP */
123 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
124 	DMU_OT_USERREFS,		/* ZAP */
125 	DMU_OT_DDT_ZAP,			/* ZAP */
126 	DMU_OT_DDT_STATS,		/* ZAP */
127 	DMU_OT_SA,			/* System attr */
128 	DMU_OT_SA_MASTER_NODE,		/* ZAP */
129 	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
130 	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
131 	DMU_OT_SCAN_XLATE,		/* ZAP */
132 	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
133 	DMU_OT_DEADLIST,		/* ZAP */
134 	DMU_OT_DEADLIST_HDR,		/* UINT64 */
135 	DMU_OT_DSL_CLONES,		/* ZAP */
136 	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
137 	DMU_OT_NUMTYPES
138 } dmu_object_type_t;
139 
140 typedef enum dmu_objset_type {
141 	DMU_OST_NONE,
142 	DMU_OST_META,
143 	DMU_OST_ZFS,
144 	DMU_OST_ZVOL,
145 	DMU_OST_OTHER,			/* For testing only! */
146 	DMU_OST_ANY,			/* Be careful! */
147 	DMU_OST_NUMTYPES
148 } dmu_objset_type_t;
149 
150 void byteswap_uint64_array(void *buf, size_t size);
151 void byteswap_uint32_array(void *buf, size_t size);
152 void byteswap_uint16_array(void *buf, size_t size);
153 void byteswap_uint8_array(void *buf, size_t size);
154 void zap_byteswap(void *buf, size_t size);
155 void zfs_oldacl_byteswap(void *buf, size_t size);
156 void zfs_acl_byteswap(void *buf, size_t size);
157 void zfs_znode_byteswap(void *buf, size_t size);
158 
159 #define	DS_FIND_SNAPSHOTS	(1<<0)
160 #define	DS_FIND_CHILDREN	(1<<1)
161 
162 /*
163  * The maximum number of bytes that can be accessed as part of one
164  * operation, including metadata.
165  */
166 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
167 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
168 
169 #define	DMU_USERUSED_OBJECT	(-1ULL)
170 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
171 #define	DMU_DEADLIST_OBJECT	(-3ULL)
172 
173 /*
174  * artificial blkids for bonus buffer and spill blocks
175  */
176 #define	DMU_BONUS_BLKID		(-1ULL)
177 #define	DMU_SPILL_BLKID		(-2ULL)
178 /*
179  * Public routines to create, destroy, open, and close objsets.
180  */
181 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
182 int dmu_objset_own(const char *name, dmu_objset_type_t type,
183     boolean_t readonly, void *tag, objset_t **osp);
184 void dmu_objset_rele(objset_t *os, void *tag);
185 void dmu_objset_disown(objset_t *os, void *tag);
186 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
187 
188 int dmu_objset_evict_dbufs(objset_t *os);
189 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
190     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
191 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
192     uint64_t flags);
193 int dmu_objset_destroy(const char *name, boolean_t defer);
194 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
195 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
196     boolean_t recursive);
197 int dmu_objset_rename(const char *name, const char *newname,
198     boolean_t recursive);
199 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
200     int flags);
201 void dmu_objset_byteswap(void *buf, size_t size);
202 
203 typedef struct dmu_buf {
204 	uint64_t db_object;		/* object that this buffer is part of */
205 	uint64_t db_offset;		/* byte offset in this object */
206 	uint64_t db_size;		/* size of buffer in bytes */
207 	void *db_data;			/* data in buffer */
208 } dmu_buf_t;
209 
210 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
211 
212 /*
213  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
214  */
215 #define	DMU_POOL_DIRECTORY_OBJECT	1
216 #define	DMU_POOL_CONFIG			"config"
217 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
218 #define	DMU_POOL_SYNC_BPOBJ		"sync_bplist"
219 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
220 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
221 #define	DMU_POOL_SPARES			"spares"
222 #define	DMU_POOL_DEFLATE		"deflate"
223 #define	DMU_POOL_HISTORY		"history"
224 #define	DMU_POOL_PROPS			"pool_props"
225 #define	DMU_POOL_L2CACHE		"l2cache"
226 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
227 #define	DMU_POOL_DDT			"DDT-%s-%s-%s"
228 #define	DMU_POOL_DDT_STATS		"DDT-statistics"
229 #define	DMU_POOL_CREATION_VERSION	"creation_version"
230 #define	DMU_POOL_SCAN			"scan"
231 #define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
232 
233 /*
234  * Allocate an object from this objset.  The range of object numbers
235  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
236  *
237  * The transaction must be assigned to a txg.  The newly allocated
238  * object will be "held" in the transaction (ie. you can modify the
239  * newly allocated object in this transaction).
240  *
241  * dmu_object_alloc() chooses an object and returns it in *objectp.
242  *
243  * dmu_object_claim() allocates a specific object number.  If that
244  * number is already allocated, it fails and returns EEXIST.
245  *
246  * Return 0 on success, or ENOSPC or EEXIST as specified above.
247  */
248 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
249     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
250 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
251     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
252 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
253     int blocksize, dmu_object_type_t bonustype, int bonuslen);
254 
255 /*
256  * Free an object from this objset.
257  *
258  * The object's data will be freed as well (ie. you don't need to call
259  * dmu_free(object, 0, -1, tx)).
260  *
261  * The object need not be held in the transaction.
262  *
263  * If there are any holds on this object's buffers (via dmu_buf_hold()),
264  * or tx holds on the object (via dmu_tx_hold_object()), you can not
265  * free it; it fails and returns EBUSY.
266  *
267  * If the object is not allocated, it fails and returns ENOENT.
268  *
269  * Return 0 on success, or EBUSY or ENOENT as specified above.
270  */
271 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
272 
273 /*
274  * Find the next allocated or free object.
275  *
276  * The objectp parameter is in-out.  It will be updated to be the next
277  * object which is allocated.  Ignore objects which have not been
278  * modified since txg.
279  *
280  * XXX Can only be called on a objset with no dirty data.
281  *
282  * Returns 0 on success, or ENOENT if there are no more objects.
283  */
284 int dmu_object_next(objset_t *os, uint64_t *objectp,
285     boolean_t hole, uint64_t txg);
286 
287 /*
288  * Set the data blocksize for an object.
289  *
290  * The object cannot have any blocks allcated beyond the first.  If
291  * the first block is allocated already, the new size must be greater
292  * than the current block size.  If these conditions are not met,
293  * ENOTSUP will be returned.
294  *
295  * Returns 0 on success, or EBUSY if there are any holds on the object
296  * contents, or ENOTSUP as described above.
297  */
298 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
299     int ibs, dmu_tx_t *tx);
300 
301 /*
302  * Set the checksum property on a dnode.  The new checksum algorithm will
303  * apply to all newly written blocks; existing blocks will not be affected.
304  */
305 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
306     dmu_tx_t *tx);
307 
308 /*
309  * Set the compress property on a dnode.  The new compression algorithm will
310  * apply to all newly written blocks; existing blocks will not be affected.
311  */
312 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
313     dmu_tx_t *tx);
314 
315 /*
316  * Decide how to write a block: checksum, compression, number of copies, etc.
317  */
318 #define	WP_NOFILL	0x1
319 #define	WP_DMU_SYNC	0x2
320 #define	WP_SPILL	0x4
321 
322 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
323     struct zio_prop *zp);
324 /*
325  * The bonus data is accessed more or less like a regular buffer.
326  * You must dmu_bonus_hold() to get the buffer, which will give you a
327  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
328  * data.  As with any normal buffer, you must call dmu_buf_read() to
329  * read db_data, dmu_buf_will_dirty() before modifying it, and the
330  * object must be held in an assigned transaction before calling
331  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
332  * buffer as well.  You must release your hold with dmu_buf_rele().
333  */
334 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
335 int dmu_bonus_max(void);
336 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
337 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
338 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
339 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
340 
341 /*
342  * Special spill buffer support used by "SA" framework
343  */
344 
345 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
346 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
347     void *tag, dmu_buf_t **dbp);
348 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
349 
350 /*
351  * Obtain the DMU buffer from the specified object which contains the
352  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
353  * that it will remain in memory.  You must release the hold with
354  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
355  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
356  *
357  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
358  * on the returned buffer before reading or writing the buffer's
359  * db_data.  The comments for those routines describe what particular
360  * operations are valid after calling them.
361  *
362  * The object number must be a valid, allocated object number.
363  */
364 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
365     void *tag, dmu_buf_t **, int flags);
366 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
367 void dmu_buf_rele(dmu_buf_t *db, void *tag);
368 uint64_t dmu_buf_refcount(dmu_buf_t *db);
369 
370 /*
371  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
372  * range of an object.  A pointer to an array of dmu_buf_t*'s is
373  * returned (in *dbpp).
374  *
375  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
376  * frees the array.  The hold on the array of buffers MUST be released
377  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
378  * individually with dmu_buf_rele.
379  */
380 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
381     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
382 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
383 
384 /*
385  * Returns NULL on success, or the existing user ptr if it's already
386  * been set.
387  *
388  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
389  *
390  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
391  * will be set to db->db_data when you are allowed to access it.  Note
392  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
393  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
394  * *user_data_ptr_ptr will be set to the new value when it changes.
395  *
396  * If non-NULL, pageout func will be called when this buffer is being
397  * excised from the cache, so that you can clean up the data structure
398  * pointed to by user_ptr.
399  *
400  * dmu_evict_user() will call the pageout func for all buffers in a
401  * objset with a given pageout func.
402  */
403 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
404     dmu_buf_evict_func_t *pageout_func);
405 /*
406  * set_user_ie is the same as set_user, but request immediate eviction
407  * when hold count goes to zero.
408  */
409 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
410     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
411 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
412     void *user_ptr, void *user_data_ptr_ptr,
413     dmu_buf_evict_func_t *pageout_func);
414 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
415 
416 /*
417  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
418  */
419 void *dmu_buf_get_user(dmu_buf_t *db);
420 
421 /*
422  * Indicate that you are going to modify the buffer's data (db_data).
423  *
424  * The transaction (tx) must be assigned to a txg (ie. you've called
425  * dmu_tx_assign()).  The buffer's object must be held in the tx
426  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
427  */
428 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
429 
430 /*
431  * Tells if the given dbuf is freeable.
432  */
433 boolean_t dmu_buf_freeable(dmu_buf_t *);
434 
435 /*
436  * You must create a transaction, then hold the objects which you will
437  * (or might) modify as part of this transaction.  Then you must assign
438  * the transaction to a transaction group.  Once the transaction has
439  * been assigned, you can modify buffers which belong to held objects as
440  * part of this transaction.  You can't modify buffers before the
441  * transaction has been assigned; you can't modify buffers which don't
442  * belong to objects which this transaction holds; you can't hold
443  * objects once the transaction has been assigned.  You may hold an
444  * object which you are going to free (with dmu_object_free()), but you
445  * don't have to.
446  *
447  * You can abort the transaction before it has been assigned.
448  *
449  * Note that you may hold buffers (with dmu_buf_hold) at any time,
450  * regardless of transaction state.
451  */
452 
453 #define	DMU_NEW_OBJECT	(-1ULL)
454 #define	DMU_OBJECT_END	(-1ULL)
455 
456 dmu_tx_t *dmu_tx_create(objset_t *os);
457 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
458 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
459     uint64_t len);
460 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
461 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
462 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
463 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
464 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
465 void dmu_tx_abort(dmu_tx_t *tx);
466 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
467 void dmu_tx_wait(dmu_tx_t *tx);
468 void dmu_tx_commit(dmu_tx_t *tx);
469 
470 /*
471  * To register a commit callback, dmu_tx_callback_register() must be called.
472  *
473  * dcb_data is a pointer to caller private data that is passed on as a
474  * callback parameter. The caller is responsible for properly allocating and
475  * freeing it.
476  *
477  * When registering a callback, the transaction must be already created, but
478  * it cannot be committed or aborted. It can be assigned to a txg or not.
479  *
480  * The callback will be called after the transaction has been safely written
481  * to stable storage and will also be called if the dmu_tx is aborted.
482  * If there is any error which prevents the transaction from being committed to
483  * disk, the callback will be called with a value of error != 0.
484  */
485 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
486 
487 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
488     void *dcb_data);
489 
490 /*
491  * Free up the data blocks for a defined range of a file.  If size is
492  * zero, the range from offset to end-of-file is freed.
493  */
494 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
495 	uint64_t size, dmu_tx_t *tx);
496 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
497 	uint64_t size);
498 int dmu_free_object(objset_t *os, uint64_t object);
499 
500 /*
501  * Convenience functions.
502  *
503  * Canfail routines will return 0 on success, or an errno if there is a
504  * nonrecoverable I/O error.
505  */
506 #define	DMU_READ_PREFETCH	0 /* prefetch */
507 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
508 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
509 	void *buf, uint32_t flags);
510 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
511 	const void *buf, dmu_tx_t *tx);
512 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
513 	dmu_tx_t *tx);
514 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
515 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
516     dmu_tx_t *tx);
517 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
518     dmu_tx_t *tx);
519 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
520     uint64_t size, struct page *pp, dmu_tx_t *tx);
521 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
522 void dmu_return_arcbuf(struct arc_buf *buf);
523 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
524     dmu_tx_t *tx);
525 int dmu_xuio_init(struct xuio *uio, int niov);
526 void dmu_xuio_fini(struct xuio *uio);
527 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
528     size_t n);
529 int dmu_xuio_cnt(struct xuio *uio);
530 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
531 void dmu_xuio_clear(struct xuio *uio, int i);
532 void xuio_stat_wbuf_copied();
533 void xuio_stat_wbuf_nocopy();
534 
535 extern int zfs_prefetch_disable;
536 
537 /*
538  * Asynchronously try to read in the data.
539  */
540 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
541     uint64_t len);
542 
543 typedef struct dmu_object_info {
544 	/* All sizes are in bytes unless otherwise indicated. */
545 	uint32_t doi_data_block_size;
546 	uint32_t doi_metadata_block_size;
547 	dmu_object_type_t doi_type;
548 	dmu_object_type_t doi_bonus_type;
549 	uint64_t doi_bonus_size;
550 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
551 	uint8_t doi_checksum;
552 	uint8_t doi_compress;
553 	uint8_t doi_pad[5];
554 	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
555 	uint64_t doi_max_offset;
556 	uint64_t doi_fill_count;		/* number of non-empty blocks */
557 } dmu_object_info_t;
558 
559 typedef void arc_byteswap_func_t(void *buf, size_t size);
560 
561 typedef struct dmu_object_type_info {
562 	arc_byteswap_func_t	*ot_byteswap;
563 	boolean_t		ot_metadata;
564 	char			*ot_name;
565 } dmu_object_type_info_t;
566 
567 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
568 
569 /*
570  * Get information on a DMU object.
571  *
572  * Return 0 on success or ENOENT if object is not allocated.
573  *
574  * If doi is NULL, just indicates whether the object exists.
575  */
576 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
577 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
578 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
579 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
580     u_longlong_t *nblk512);
581 
582 typedef struct dmu_objset_stats {
583 	uint64_t dds_num_clones; /* number of clones of this */
584 	uint64_t dds_creation_txg;
585 	uint64_t dds_guid;
586 	dmu_objset_type_t dds_type;
587 	uint8_t dds_is_snapshot;
588 	uint8_t dds_inconsistent;
589 	char dds_origin[MAXNAMELEN];
590 } dmu_objset_stats_t;
591 
592 /*
593  * Get stats on a dataset.
594  */
595 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
596 
597 /*
598  * Add entries to the nvlist for all the objset's properties.  See
599  * zfs_prop_table[] and zfs(1m) for details on the properties.
600  */
601 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
602 
603 /*
604  * Get the space usage statistics for statvfs().
605  *
606  * refdbytes is the amount of space "referenced" by this objset.
607  * availbytes is the amount of space available to this objset, taking
608  * into account quotas & reservations, assuming that no other objsets
609  * use the space first.  These values correspond to the 'referenced' and
610  * 'available' properties, described in the zfs(1m) manpage.
611  *
612  * usedobjs and availobjs are the number of objects currently allocated,
613  * and available.
614  */
615 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
616     uint64_t *usedobjsp, uint64_t *availobjsp);
617 
618 /*
619  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
620  * (Contrast with the ds_guid which is a 64-bit ID that will never
621  * change, so there is a small probability that it will collide.)
622  */
623 uint64_t dmu_objset_fsid_guid(objset_t *os);
624 
625 /*
626  * Get the [cm]time for an objset's snapshot dir
627  */
628 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
629 
630 int dmu_objset_is_snapshot(objset_t *os);
631 
632 extern struct spa *dmu_objset_spa(objset_t *os);
633 extern struct zilog *dmu_objset_zil(objset_t *os);
634 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
635 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
636 extern void dmu_objset_name(objset_t *os, char *buf);
637 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
638 extern uint64_t dmu_objset_id(objset_t *os);
639 extern uint64_t dmu_objset_syncprop(objset_t *os);
640 extern uint64_t dmu_objset_logbias(objset_t *os);
641 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
642     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
643 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
644     int maxlen, boolean_t *conflict);
645 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
646     uint64_t *idp, uint64_t *offp);
647 
648 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
649     void *bonus, uint64_t *userp, uint64_t *groupp);
650 extern void dmu_objset_register_type(dmu_objset_type_t ost,
651     objset_used_cb_t *cb);
652 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
653 extern void *dmu_objset_get_user(objset_t *os);
654 
655 /*
656  * Return the txg number for the given assigned transaction.
657  */
658 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
659 
660 /*
661  * Synchronous write.
662  * If a parent zio is provided this function initiates a write on the
663  * provided buffer as a child of the parent zio.
664  * In the absence of a parent zio, the write is completed synchronously.
665  * At write completion, blk is filled with the bp of the written block.
666  * Note that while the data covered by this function will be on stable
667  * storage when the write completes this new data does not become a
668  * permanent part of the file until the associated transaction commits.
669  */
670 
671 /*
672  * {zfs,zvol,ztest}_get_done() args
673  */
674 typedef struct zgd {
675 	struct zilog	*zgd_zilog;
676 	struct blkptr	*zgd_bp;
677 	dmu_buf_t	*zgd_db;
678 	struct rl	*zgd_rl;
679 	void		*zgd_private;
680 } zgd_t;
681 
682 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
683 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
684 
685 /*
686  * Find the next hole or data block in file starting at *off
687  * Return found offset in *off. Return ESRCH for end of file.
688  */
689 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
690     uint64_t *off);
691 
692 /*
693  * Initial setup and final teardown.
694  */
695 extern void dmu_init(void);
696 extern void dmu_fini(void);
697 
698 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
699     uint64_t object, uint64_t offset, int len);
700 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
701     dmu_traverse_cb_t cb, void *arg);
702 
703 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
704     struct vnode *vp, offset_t *off);
705 
706 typedef struct dmu_recv_cookie {
707 	/*
708 	 * This structure is opaque!
709 	 *
710 	 * If logical and real are different, we are recving the stream
711 	 * into the "real" temporary clone, and then switching it with
712 	 * the "logical" target.
713 	 */
714 	struct dsl_dataset *drc_logical_ds;
715 	struct dsl_dataset *drc_real_ds;
716 	struct drr_begin *drc_drrb;
717 	char *drc_tosnap;
718 	char *drc_top_ds;
719 	boolean_t drc_newfs;
720 	boolean_t drc_force;
721 } dmu_recv_cookie_t;
722 
723 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
724     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
725 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
726     int cleanup_fd, uint64_t *action_handlep);
727 int dmu_recv_end(dmu_recv_cookie_t *drc);
728 
729 /* CRC64 table */
730 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
731 extern uint64_t zfs_crc64_table[256];
732 
733 #ifdef	__cplusplus
734 }
735 #endif
736 
737 #endif	/* _SYS_DMU_H */
738