xref: /titanic_52/usr/src/uts/common/fs/zfs/sys/dmu.h (revision c0dd49bdd68c0d758a67d56f07826f3b45cfc664)
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 #ifndef	_SYS_DMU_H
26 #define	_SYS_DMU_H
27 
28 /*
29  * This file describes the interface that the DMU provides for its
30  * consumers.
31  *
32  * The DMU also interacts with the SPA.  That interface is described in
33  * dmu_spa.h.
34  */
35 
36 #include <sys/inttypes.h>
37 #include <sys/types.h>
38 #include <sys/param.h>
39 #include <sys/cred.h>
40 #include <sys/time.h>
41 
42 #ifdef	__cplusplus
43 extern "C" {
44 #endif
45 
46 struct uio;
47 struct xuio;
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 struct sa_handle;
66 
67 typedef struct objset objset_t;
68 typedef struct dmu_tx dmu_tx_t;
69 typedef struct dsl_dir dsl_dir_t;
70 
71 typedef enum dmu_object_type {
72 	DMU_OT_NONE,
73 	/* general: */
74 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
75 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
76 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
77 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
78 	DMU_OT_BPLIST,			/* UINT64 */
79 	DMU_OT_BPLIST_HDR,		/* UINT64 */
80 	/* spa: */
81 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
82 	DMU_OT_SPACE_MAP,		/* UINT64 */
83 	/* zil: */
84 	DMU_OT_INTENT_LOG,		/* UINT64 */
85 	/* dmu: */
86 	DMU_OT_DNODE,			/* DNODE */
87 	DMU_OT_OBJSET,			/* OBJSET */
88 	/* dsl: */
89 	DMU_OT_DSL_DIR,			/* UINT64 */
90 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
91 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
92 	DMU_OT_DSL_PROPS,		/* ZAP */
93 	DMU_OT_DSL_DATASET,		/* UINT64 */
94 	/* zpl: */
95 	DMU_OT_ZNODE,			/* ZNODE */
96 	DMU_OT_OLDACL,			/* Old ACL */
97 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
98 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
99 	DMU_OT_MASTER_NODE,		/* ZAP */
100 	DMU_OT_UNLINKED_SET,		/* ZAP */
101 	/* zvol: */
102 	DMU_OT_ZVOL,			/* UINT8 */
103 	DMU_OT_ZVOL_PROP,		/* ZAP */
104 	/* other; for testing only! */
105 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
106 	DMU_OT_UINT64_OTHER,		/* UINT64 */
107 	DMU_OT_ZAP_OTHER,		/* ZAP */
108 	/* new object types: */
109 	DMU_OT_ERROR_LOG,		/* ZAP */
110 	DMU_OT_SPA_HISTORY,		/* UINT8 */
111 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
112 	DMU_OT_POOL_PROPS,		/* ZAP */
113 	DMU_OT_DSL_PERMS,		/* ZAP */
114 	DMU_OT_ACL,			/* ACL */
115 	DMU_OT_SYSACL,			/* SYSACL */
116 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
117 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
118 	DMU_OT_NEXT_CLONES,		/* ZAP */
119 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
120 	DMU_OT_USERGROUP_USED,		/* ZAP */
121 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
122 	DMU_OT_USERREFS,		/* ZAP */
123 	DMU_OT_DDT_ZAP,			/* ZAP */
124 	DMU_OT_DDT_STATS,		/* ZAP */
125 	DMU_OT_SA,			/* System attr */
126 	DMU_OT_SA_MASTER_NODE,		/* ZAP */
127 	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
128 	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
129 	DMU_OT_NUMTYPES
130 } dmu_object_type_t;
131 
132 typedef enum dmu_objset_type {
133 	DMU_OST_NONE,
134 	DMU_OST_META,
135 	DMU_OST_ZFS,
136 	DMU_OST_ZVOL,
137 	DMU_OST_OTHER,			/* For testing only! */
138 	DMU_OST_ANY,			/* Be careful! */
139 	DMU_OST_NUMTYPES
140 } dmu_objset_type_t;
141 
142 void byteswap_uint64_array(void *buf, size_t size);
143 void byteswap_uint32_array(void *buf, size_t size);
144 void byteswap_uint16_array(void *buf, size_t size);
145 void byteswap_uint8_array(void *buf, size_t size);
146 void zap_byteswap(void *buf, size_t size);
147 void zfs_oldacl_byteswap(void *buf, size_t size);
148 void zfs_acl_byteswap(void *buf, size_t size);
149 void zfs_znode_byteswap(void *buf, size_t size);
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 #define	DMU_DEADLIST_OBJECT	(-3ULL)
164 
165 /*
166  * artificial blkids for bonus buffer and spill blocks
167  */
168 #define	DMU_BONUS_BLKID		(-1ULL)
169 #define	DMU_SPILL_BLKID		(-2ULL)
170 /*
171  * Public routines to create, destroy, open, and close objsets.
172  */
173 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
174 int dmu_objset_own(const char *name, dmu_objset_type_t type,
175     boolean_t readonly, void *tag, objset_t **osp);
176 void dmu_objset_rele(objset_t *os, void *tag);
177 void dmu_objset_disown(objset_t *os, void *tag);
178 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
179 
180 int dmu_objset_evict_dbufs(objset_t *os);
181 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
182     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
183 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
184     uint64_t flags);
185 int dmu_objset_destroy(const char *name, boolean_t defer);
186 int dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
187 int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
188     boolean_t recursive);
189 int dmu_objset_rename(const char *name, const char *newname,
190     boolean_t recursive);
191 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
192     int flags);
193 void dmu_objset_byteswap(void *buf, size_t size);
194 
195 typedef struct dmu_buf {
196 	uint64_t db_object;		/* object that this buffer is part of */
197 	uint64_t db_offset;		/* byte offset in this object */
198 	uint64_t db_size;		/* size of buffer in bytes */
199 	void *db_data;			/* data in buffer */
200 } dmu_buf_t;
201 
202 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
203 
204 /*
205  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
206  */
207 #define	DMU_POOL_DIRECTORY_OBJECT	1
208 #define	DMU_POOL_CONFIG			"config"
209 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
210 #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
211 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
212 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
213 #define	DMU_POOL_SPARES			"spares"
214 #define	DMU_POOL_DEFLATE		"deflate"
215 #define	DMU_POOL_HISTORY		"history"
216 #define	DMU_POOL_PROPS			"pool_props"
217 #define	DMU_POOL_L2CACHE		"l2cache"
218 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
219 #define	DMU_POOL_DDT			"DDT-%s-%s-%s"
220 #define	DMU_POOL_DDT_STATS		"DDT-statistics"
221 
222 /* 4x8 zbookmark_t */
223 #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
224 /* 4x8 ddt_bookmark_t */
225 #define	DMU_POOL_SCRUB_DDT_BOOKMARK	"scrub_ddt_bookmark"
226 /* 1x8 max_class */
227 #define	DMU_POOL_SCRUB_DDT_CLASS_MAX	"scrub_ddt_class_max"
228 /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
229 #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
230 /* 1x8 txg */
231 #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
232 /* 1x8 txg */
233 #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
234 /* 1x4 enum scrub_func */
235 #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
236 /* 1x8 count */
237 #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
238 
239 /*
240  * Allocate an object from this objset.  The range of object numbers
241  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
242  *
243  * The transaction must be assigned to a txg.  The newly allocated
244  * object will be "held" in the transaction (ie. you can modify the
245  * newly allocated object in this transaction).
246  *
247  * dmu_object_alloc() chooses an object and returns it in *objectp.
248  *
249  * dmu_object_claim() allocates a specific object number.  If that
250  * number is already allocated, it fails and returns EEXIST.
251  *
252  * Return 0 on success, or ENOSPC or EEXIST as specified above.
253  */
254 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
255     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
256 int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
257     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
258 int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
259     int blocksize, dmu_object_type_t bonustype, int bonuslen);
260 
261 /*
262  * Free an object from this objset.
263  *
264  * The object's data will be freed as well (ie. you don't need to call
265  * dmu_free(object, 0, -1, tx)).
266  *
267  * The object need not be held in the transaction.
268  *
269  * If there are any holds on this object's buffers (via dmu_buf_hold()),
270  * or tx holds on the object (via dmu_tx_hold_object()), you can not
271  * free it; it fails and returns EBUSY.
272  *
273  * If the object is not allocated, it fails and returns ENOENT.
274  *
275  * Return 0 on success, or EBUSY or ENOENT as specified above.
276  */
277 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
278 
279 /*
280  * Find the next allocated or free object.
281  *
282  * The objectp parameter is in-out.  It will be updated to be the next
283  * object which is allocated.  Ignore objects which have not been
284  * modified since txg.
285  *
286  * XXX Can only be called on a objset with no dirty data.
287  *
288  * Returns 0 on success, or ENOENT if there are no more objects.
289  */
290 int dmu_object_next(objset_t *os, uint64_t *objectp,
291     boolean_t hole, uint64_t txg);
292 
293 /*
294  * Set the data blocksize for an object.
295  *
296  * The object cannot have any blocks allcated beyond the first.  If
297  * the first block is allocated already, the new size must be greater
298  * than the current block size.  If these conditions are not met,
299  * ENOTSUP will be returned.
300  *
301  * Returns 0 on success, or EBUSY if there are any holds on the object
302  * contents, or ENOTSUP as described above.
303  */
304 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
305     int ibs, dmu_tx_t *tx);
306 
307 /*
308  * Set the checksum property on a dnode.  The new checksum algorithm will
309  * apply to all newly written blocks; existing blocks will not be affected.
310  */
311 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
312     dmu_tx_t *tx);
313 
314 /*
315  * Set the compress property on a dnode.  The new compression algorithm will
316  * apply to all newly written blocks; existing blocks will not be affected.
317  */
318 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
319     dmu_tx_t *tx);
320 
321 /*
322  * Decide how to write a block: checksum, compression, number of copies, etc.
323  */
324 #define	WP_NOFILL	0x1
325 #define	WP_DMU_SYNC	0x2
326 #define	WP_SPILL	0x4
327 
328 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
329     struct zio_prop *zp);
330 /*
331  * The bonus data is accessed more or less like a regular buffer.
332  * You must dmu_bonus_hold() to get the buffer, which will give you a
333  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
334  * data.  As with any normal buffer, you must call dmu_buf_read() to
335  * read db_data, dmu_buf_will_dirty() before modifying it, and the
336  * object must be held in an assigned transaction before calling
337  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
338  * buffer as well.  You must release your hold with dmu_buf_rele().
339  */
340 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
341 int dmu_bonus_max(void);
342 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
343 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
344 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
345 
346 /*
347  * Special spill buffer support used by "SA" framework
348  */
349 
350 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
351 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
352     void *tag, dmu_buf_t **dbp);
353 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
354 
355 /*
356  * Obtain the DMU buffer from the specified object which contains the
357  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
358  * that it will remain in memory.  You must release the hold with
359  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
360  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
361  *
362  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
363  * on the returned buffer before reading or writing the buffer's
364  * db_data.  The comments for those routines describe what particular
365  * operations are valid after calling them.
366  *
367  * The object number must be a valid, allocated object number.
368  */
369 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
370     void *tag, dmu_buf_t **);
371 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
372 void dmu_buf_rele(dmu_buf_t *db, void *tag);
373 uint64_t dmu_buf_refcount(dmu_buf_t *db);
374 
375 /*
376  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
377  * range of an object.  A pointer to an array of dmu_buf_t*'s is
378  * returned (in *dbpp).
379  *
380  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
381  * frees the array.  The hold on the array of buffers MUST be released
382  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
383  * individually with dmu_buf_rele.
384  */
385 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
386     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
387 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
388 
389 /*
390  * Returns NULL on success, or the existing user ptr if it's already
391  * been set.
392  *
393  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
394  *
395  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
396  * will be set to db->db_data when you are allowed to access it.  Note
397  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
398  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
399  * *user_data_ptr_ptr will be set to the new value when it changes.
400  *
401  * If non-NULL, pageout func will be called when this buffer is being
402  * excised from the cache, so that you can clean up the data structure
403  * pointed to by user_ptr.
404  *
405  * dmu_evict_user() will call the pageout func for all buffers in a
406  * objset with a given pageout func.
407  */
408 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
409     dmu_buf_evict_func_t *pageout_func);
410 /*
411  * set_user_ie is the same as set_user, but request immediate eviction
412  * when hold count goes to zero.
413  */
414 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
415     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
416 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
417     void *user_ptr, void *user_data_ptr_ptr,
418     dmu_buf_evict_func_t *pageout_func);
419 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
420 
421 /*
422  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
423  */
424 void *dmu_buf_get_user(dmu_buf_t *db);
425 
426 /*
427  * Indicate that you are going to modify the buffer's data (db_data).
428  *
429  * The transaction (tx) must be assigned to a txg (ie. you've called
430  * dmu_tx_assign()).  The buffer's object must be held in the tx
431  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
432  */
433 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
434 
435 /*
436  * Tells if the given dbuf is freeable.
437  */
438 boolean_t dmu_buf_freeable(dmu_buf_t *);
439 
440 /*
441  * You must create a transaction, then hold the objects which you will
442  * (or might) modify as part of this transaction.  Then you must assign
443  * the transaction to a transaction group.  Once the transaction has
444  * been assigned, you can modify buffers which belong to held objects as
445  * part of this transaction.  You can't modify buffers before the
446  * transaction has been assigned; you can't modify buffers which don't
447  * belong to objects which this transaction holds; you can't hold
448  * objects once the transaction has been assigned.  You may hold an
449  * object which you are going to free (with dmu_object_free()), but you
450  * don't have to.
451  *
452  * You can abort the transaction before it has been assigned.
453  *
454  * Note that you may hold buffers (with dmu_buf_hold) at any time,
455  * regardless of transaction state.
456  */
457 
458 #define	DMU_NEW_OBJECT	(-1ULL)
459 #define	DMU_OBJECT_END	(-1ULL)
460 
461 dmu_tx_t *dmu_tx_create(objset_t *os);
462 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
463 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
464     uint64_t len);
465 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
466 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
467 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
468 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
469 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
470 void dmu_tx_abort(dmu_tx_t *tx);
471 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
472 void dmu_tx_wait(dmu_tx_t *tx);
473 void dmu_tx_commit(dmu_tx_t *tx);
474 
475 /*
476  * To register a commit callback, dmu_tx_callback_register() must be called.
477  *
478  * dcb_data is a pointer to caller private data that is passed on as a
479  * callback parameter. The caller is responsible for properly allocating and
480  * freeing it.
481  *
482  * When registering a callback, the transaction must be already created, but
483  * it cannot be committed or aborted. It can be assigned to a txg or not.
484  *
485  * The callback will be called after the transaction has been safely written
486  * to stable storage and will also be called if the dmu_tx is aborted.
487  * If there is any error which prevents the transaction from being committed to
488  * disk, the callback will be called with a value of error != 0.
489  */
490 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
491 
492 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
493     void *dcb_data);
494 
495 /*
496  * Free up the data blocks for a defined range of a file.  If size is
497  * zero, the range from offset to end-of-file is freed.
498  */
499 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
500 	uint64_t size, dmu_tx_t *tx);
501 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
502 	uint64_t size);
503 int dmu_free_object(objset_t *os, uint64_t object);
504 
505 /*
506  * Convenience functions.
507  *
508  * Canfail routines will return 0 on success, or an errno if there is a
509  * nonrecoverable I/O error.
510  */
511 #define	DMU_READ_PREFETCH	0 /* prefetch */
512 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
513 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
514 	void *buf, uint32_t flags);
515 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
516 	const void *buf, dmu_tx_t *tx);
517 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
518 	dmu_tx_t *tx);
519 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
520 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
521     dmu_tx_t *tx);
522 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
523     dmu_tx_t *tx);
524 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
525     uint64_t size, struct page *pp, dmu_tx_t *tx);
526 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
527 void dmu_return_arcbuf(struct arc_buf *buf);
528 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
529     dmu_tx_t *tx);
530 int dmu_xuio_init(struct xuio *uio, int niov);
531 void dmu_xuio_fini(struct xuio *uio);
532 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
533     size_t n);
534 int dmu_xuio_cnt(struct xuio *uio);
535 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
536 void dmu_xuio_clear(struct xuio *uio, int i);
537 void xuio_stat_wbuf_copied();
538 void xuio_stat_wbuf_nocopy();
539 
540 extern int zfs_prefetch_disable;
541 
542 /*
543  * Asynchronously try to read in the data.
544  */
545 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
546     uint64_t len);
547 
548 typedef struct dmu_object_info {
549 	/* All sizes are in bytes unless otherwise indicated. */
550 	uint32_t doi_data_block_size;
551 	uint32_t doi_metadata_block_size;
552 	dmu_object_type_t doi_type;
553 	dmu_object_type_t doi_bonus_type;
554 	uint64_t doi_bonus_size;
555 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
556 	uint8_t doi_checksum;
557 	uint8_t doi_compress;
558 	uint8_t doi_pad[5];
559 	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
560 	uint64_t doi_max_offset;
561 	uint64_t doi_fill_count;		/* number of non-empty blocks */
562 } dmu_object_info_t;
563 
564 typedef void arc_byteswap_func_t(void *buf, size_t size);
565 
566 typedef struct dmu_object_type_info {
567 	arc_byteswap_func_t	*ot_byteswap;
568 	boolean_t		ot_metadata;
569 	char			*ot_name;
570 } dmu_object_type_info_t;
571 
572 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
573 
574 /*
575  * Get information on a DMU object.
576  *
577  * Return 0 on success or ENOENT if object is not allocated.
578  *
579  * If doi is NULL, just indicates whether the object exists.
580  */
581 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
582 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
583 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
584 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
585     u_longlong_t *nblk512);
586 
587 typedef struct dmu_objset_stats {
588 	uint64_t dds_num_clones; /* number of clones of this */
589 	uint64_t dds_creation_txg;
590 	uint64_t dds_guid;
591 	dmu_objset_type_t dds_type;
592 	uint8_t dds_is_snapshot;
593 	uint8_t dds_inconsistent;
594 	char dds_origin[MAXNAMELEN];
595 } dmu_objset_stats_t;
596 
597 /*
598  * Get stats on a dataset.
599  */
600 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
601 
602 /*
603  * Add entries to the nvlist for all the objset's properties.  See
604  * zfs_prop_table[] and zfs(1m) for details on the properties.
605  */
606 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
607 
608 /*
609  * Get the space usage statistics for statvfs().
610  *
611  * refdbytes is the amount of space "referenced" by this objset.
612  * availbytes is the amount of space available to this objset, taking
613  * into account quotas & reservations, assuming that no other objsets
614  * use the space first.  These values correspond to the 'referenced' and
615  * 'available' properties, described in the zfs(1m) manpage.
616  *
617  * usedobjs and availobjs are the number of objects currently allocated,
618  * and available.
619  */
620 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
621     uint64_t *usedobjsp, uint64_t *availobjsp);
622 
623 /*
624  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
625  * (Contrast with the ds_guid which is a 64-bit ID that will never
626  * change, so there is a small probability that it will collide.)
627  */
628 uint64_t dmu_objset_fsid_guid(objset_t *os);
629 
630 /*
631  * Get the [cm]time for an objset's snapshot dir
632  */
633 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
634 
635 int dmu_objset_is_snapshot(objset_t *os);
636 
637 extern struct spa *dmu_objset_spa(objset_t *os);
638 extern struct zilog *dmu_objset_zil(objset_t *os);
639 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
640 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
641 extern void dmu_objset_name(objset_t *os, char *buf);
642 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
643 extern uint64_t dmu_objset_id(objset_t *os);
644 extern uint64_t dmu_objset_logbias(objset_t *os);
645 extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
646     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
647 extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
648     int maxlen, boolean_t *conflict);
649 extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
650     uint64_t *idp, uint64_t *offp);
651 
652 typedef int objset_used_cb_t(dmu_object_type_t bonustype,
653     void *bonus, uint64_t *userp, uint64_t *groupp);
654 extern void dmu_objset_register_type(dmu_objset_type_t ost,
655     objset_used_cb_t *cb);
656 extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
657 extern void *dmu_objset_get_user(objset_t *os);
658 
659 /*
660  * Return the txg number for the given assigned transaction.
661  */
662 uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
663 
664 /*
665  * Synchronous write.
666  * If a parent zio is provided this function initiates a write on the
667  * provided buffer as a child of the parent zio.
668  * In the absence of a parent zio, the write is completed synchronously.
669  * At write completion, blk is filled with the bp of the written block.
670  * Note that while the data covered by this function will be on stable
671  * storage when the write completes this new data does not become a
672  * permanent part of the file until the associated transaction commits.
673  */
674 
675 /*
676  * {zfs,zvol,ztest}_get_done() args
677  */
678 typedef struct zgd {
679 	struct zilog	*zgd_zilog;
680 	struct blkptr	*zgd_bp;
681 	dmu_buf_t	*zgd_db;
682 	struct rl	*zgd_rl;
683 	void		*zgd_private;
684 } zgd_t;
685 
686 typedef void dmu_sync_cb_t(zgd_t *arg, int error);
687 int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
688 
689 /*
690  * Find the next hole or data block in file starting at *off
691  * Return found offset in *off. Return ESRCH for end of file.
692  */
693 int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
694     uint64_t *off);
695 
696 /*
697  * Initial setup and final teardown.
698  */
699 extern void dmu_init(void);
700 extern void dmu_fini(void);
701 
702 typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
703     uint64_t object, uint64_t offset, int len);
704 void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
705     dmu_traverse_cb_t cb, void *arg);
706 
707 int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
708     struct vnode *vp, offset_t *off);
709 
710 typedef struct dmu_recv_cookie {
711 	/*
712 	 * This structure is opaque!
713 	 *
714 	 * If logical and real are different, we are recving the stream
715 	 * into the "real" temporary clone, and then switching it with
716 	 * the "logical" target.
717 	 */
718 	struct dsl_dataset *drc_logical_ds;
719 	struct dsl_dataset *drc_real_ds;
720 	struct drr_begin *drc_drrb;
721 	char *drc_tosnap;
722 	char *drc_top_ds;
723 	boolean_t drc_newfs;
724 	boolean_t drc_force;
725 } dmu_recv_cookie_t;
726 
727 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
728     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
729 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
730 int dmu_recv_end(dmu_recv_cookie_t *drc);
731 
732 /* CRC64 table */
733 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
734 extern uint64_t zfs_crc64_table[256];
735 
736 #ifdef	__cplusplus
737 }
738 #endif
739 
740 #endif	/* _SYS_DMU_H */
741