xref: /titanic_41/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 29fd2c1627adafbe0a83726c224e60b5514c7736)
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  * Copyright (c) 2011 by Delphix. All rights reserved.
24  */
25 /*
26  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
27  */
28 
29 /* Portions Copyright 2010 Robert Milkowski */
30 
31 #ifndef	_SYS_DMU_H
32 #define	_SYS_DMU_H
33 
34 /*
35  * This file describes the interface that the DMU provides for its
36  * consumers.
37  *
38  * The DMU also interacts with the SPA.  That interface is described in
39  * dmu_spa.h.
40  */
41 
42 #include <sys/inttypes.h>
43 #include <sys/types.h>
44 #include <sys/param.h>
45 #include <sys/cred.h>
46 #include <sys/time.h>
47 
48 #ifdef	__cplusplus
49 extern "C" {
50 #endif
51 
52 struct uio;
53 struct xuio;
54 struct page;
55 struct vnode;
56 struct spa;
57 struct zilog;
58 struct zio;
59 struct blkptr;
60 struct zap_cursor;
61 struct dsl_dataset;
62 struct dsl_pool;
63 struct dnode;
64 struct drr_begin;
65 struct drr_end;
66 struct zbookmark;
67 struct spa;
68 struct nvlist;
69 struct arc_buf;
70 struct zio_prop;
71 struct sa_handle;
72 
73 typedef struct objset objset_t;
74 typedef struct dmu_tx dmu_tx_t;
75 typedef struct dsl_dir dsl_dir_t;
76 
77 typedef enum dmu_object_type {
78 	DMU_OT_NONE,
79 	/* general: */
80 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
81 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
82 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
83 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
84 	DMU_OT_BPOBJ,			/* UINT64 */
85 	DMU_OT_BPOBJ_HDR,		/* UINT64 */
86 	/* spa: */
87 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
88 	DMU_OT_SPACE_MAP,		/* UINT64 */
89 	/* zil: */
90 	DMU_OT_INTENT_LOG,		/* UINT64 */
91 	/* dmu: */
92 	DMU_OT_DNODE,			/* DNODE */
93 	DMU_OT_OBJSET,			/* OBJSET */
94 	/* dsl: */
95 	DMU_OT_DSL_DIR,			/* UINT64 */
96 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
97 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
98 	DMU_OT_DSL_PROPS,		/* ZAP */
99 	DMU_OT_DSL_DATASET,		/* UINT64 */
100 	/* zpl: */
101 	DMU_OT_ZNODE,			/* ZNODE */
102 	DMU_OT_OLDACL,			/* Old ACL */
103 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
104 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
105 	DMU_OT_MASTER_NODE,		/* ZAP */
106 	DMU_OT_UNLINKED_SET,		/* ZAP */
107 	/* zvol: */
108 	DMU_OT_ZVOL,			/* UINT8 */
109 	DMU_OT_ZVOL_PROP,		/* ZAP */
110 	/* other; for testing only! */
111 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
112 	DMU_OT_UINT64_OTHER,		/* UINT64 */
113 	DMU_OT_ZAP_OTHER,		/* ZAP */
114 	/* new object types: */
115 	DMU_OT_ERROR_LOG,		/* ZAP */
116 	DMU_OT_SPA_HISTORY,		/* UINT8 */
117 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
118 	DMU_OT_POOL_PROPS,		/* ZAP */
119 	DMU_OT_DSL_PERMS,		/* ZAP */
120 	DMU_OT_ACL,			/* ACL */
121 	DMU_OT_SYSACL,			/* SYSACL */
122 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
123 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
124 	DMU_OT_NEXT_CLONES,		/* ZAP */
125 	DMU_OT_SCAN_QUEUE,		/* ZAP */
126 	DMU_OT_USERGROUP_USED,		/* ZAP */
127 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
128 	DMU_OT_USERREFS,		/* ZAP */
129 	DMU_OT_DDT_ZAP,			/* ZAP */
130 	DMU_OT_DDT_STATS,		/* ZAP */
131 	DMU_OT_SA,			/* System attr */
132 	DMU_OT_SA_MASTER_NODE,		/* ZAP */
133 	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
134 	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
135 	DMU_OT_SCAN_XLATE,		/* ZAP */
136 	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
137 	DMU_OT_DEADLIST,		/* ZAP */
138 	DMU_OT_DEADLIST_HDR,		/* UINT64 */
139 	DMU_OT_DSL_CLONES,		/* ZAP */
140 	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
141 	DMU_OT_NUMTYPES
142 } dmu_object_type_t;
143 
144 typedef enum dmu_objset_type {
145 	DMU_OST_NONE,
146 	DMU_OST_META,
147 	DMU_OST_ZFS,
148 	DMU_OST_ZVOL,
149 	DMU_OST_OTHER,			/* For testing only! */
150 	DMU_OST_ANY,			/* Be careful! */
151 	DMU_OST_NUMTYPES
152 } dmu_objset_type_t;
153 
154 void byteswap_uint64_array(void *buf, size_t size);
155 void byteswap_uint32_array(void *buf, size_t size);
156 void byteswap_uint16_array(void *buf, size_t size);
157 void byteswap_uint8_array(void *buf, size_t size);
158 void zap_byteswap(void *buf, size_t size);
159 void zfs_oldacl_byteswap(void *buf, size_t size);
160 void zfs_acl_byteswap(void *buf, size_t size);
161 void zfs_znode_byteswap(void *buf, size_t size);
162 
163 #define	DS_FIND_SNAPSHOTS	(1<<0)
164 #define	DS_FIND_CHILDREN	(1<<1)
165 
166 /*
167  * The maximum number of bytes that can be accessed as part of one
168  * operation, including metadata.
169  */
170 #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
171 #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
172 
173 #define	DMU_USERUSED_OBJECT	(-1ULL)
174 #define	DMU_GROUPUSED_OBJECT	(-2ULL)
175 #define	DMU_DEADLIST_OBJECT	(-3ULL)
176 
177 /*
178  * artificial blkids for bonus buffer and spill blocks
179  */
180 #define	DMU_BONUS_BLKID		(-1ULL)
181 #define	DMU_SPILL_BLKID		(-2ULL)
182 /*
183  * Public routines to create, destroy, open, and close objsets.
184  */
185 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
186 int dmu_objset_own(const char *name, dmu_objset_type_t type,
187     boolean_t readonly, void *tag, objset_t **osp);
188 void dmu_objset_rele(objset_t *os, void *tag);
189 void dmu_objset_disown(objset_t *os, void *tag);
190 int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
191 
192 int dmu_objset_evict_dbufs(objset_t *os);
193 int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
194     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
195 int dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
196     uint64_t flags);
197 int dmu_objset_destroy(const char *name, boolean_t defer);
198 int dmu_snapshots_destroy_nvl(struct nvlist *snaps, boolean_t defer, char *);
199 int dmu_objset_snapshot(char *fsname, char *snapname, char *tag,
200     struct nvlist *props, boolean_t recursive, boolean_t temporary, int fd);
201 int dmu_objset_rename(const char *name, const char *newname,
202     boolean_t recursive);
203 int dmu_objset_find(char *name, int func(const char *, void *), void *arg,
204     int flags);
205 void dmu_objset_byteswap(void *buf, size_t size);
206 
207 typedef struct dmu_buf {
208 	uint64_t db_object;		/* object that this buffer is part of */
209 	uint64_t db_offset;		/* byte offset in this object */
210 	uint64_t db_size;		/* size of buffer in bytes */
211 	void *db_data;			/* data in buffer */
212 } dmu_buf_t;
213 
214 typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
215 
216 /*
217  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
218  */
219 #define	DMU_POOL_DIRECTORY_OBJECT	1
220 #define	DMU_POOL_CONFIG			"config"
221 #define	DMU_POOL_ROOT_DATASET		"root_dataset"
222 #define	DMU_POOL_SYNC_BPOBJ		"sync_bplist"
223 #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
224 #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
225 #define	DMU_POOL_SPARES			"spares"
226 #define	DMU_POOL_DEFLATE		"deflate"
227 #define	DMU_POOL_HISTORY		"history"
228 #define	DMU_POOL_PROPS			"pool_props"
229 #define	DMU_POOL_L2CACHE		"l2cache"
230 #define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
231 #define	DMU_POOL_DDT			"DDT-%s-%s-%s"
232 #define	DMU_POOL_DDT_STATS		"DDT-statistics"
233 #define	DMU_POOL_CREATION_VERSION	"creation_version"
234 #define	DMU_POOL_SCAN			"scan"
235 #define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
236 
237 /*
238  * Allocate an object from this objset.  The range of object numbers
239  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
240  *
241  * The transaction must be assigned to a txg.  The newly allocated
242  * object will be "held" in the transaction (ie. you can modify the
243  * newly allocated object in this transaction).
244  *
245  * dmu_object_alloc() chooses an object and returns it in *objectp.
246  *
247  * dmu_object_claim() allocates a specific object number.  If that
248  * number is already allocated, it fails and returns EEXIST.
249  *
250  * Return 0 on success, or ENOSPC or EEXIST as specified above.
251  */
252 uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
253     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
254 int dmu_object_claim(objset_t *os, uint64_t object, 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_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
257     int blocksize, dmu_object_type_t bonustype, int bonuslen);
258 
259 /*
260  * Free an object from this objset.
261  *
262  * The object's data will be freed as well (ie. you don't need to call
263  * dmu_free(object, 0, -1, tx)).
264  *
265  * The object need not be held in the transaction.
266  *
267  * If there are any holds on this object's buffers (via dmu_buf_hold()),
268  * or tx holds on the object (via dmu_tx_hold_object()), you can not
269  * free it; it fails and returns EBUSY.
270  *
271  * If the object is not allocated, it fails and returns ENOENT.
272  *
273  * Return 0 on success, or EBUSY or ENOENT as specified above.
274  */
275 int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
276 
277 /*
278  * Find the next allocated or free object.
279  *
280  * The objectp parameter is in-out.  It will be updated to be the next
281  * object which is allocated.  Ignore objects which have not been
282  * modified since txg.
283  *
284  * XXX Can only be called on a objset with no dirty data.
285  *
286  * Returns 0 on success, or ENOENT if there are no more objects.
287  */
288 int dmu_object_next(objset_t *os, uint64_t *objectp,
289     boolean_t hole, uint64_t txg);
290 
291 /*
292  * Set the data blocksize for an object.
293  *
294  * The object cannot have any blocks allcated beyond the first.  If
295  * the first block is allocated already, the new size must be greater
296  * than the current block size.  If these conditions are not met,
297  * ENOTSUP will be returned.
298  *
299  * Returns 0 on success, or EBUSY if there are any holds on the object
300  * contents, or ENOTSUP as described above.
301  */
302 int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
303     int ibs, dmu_tx_t *tx);
304 
305 /*
306  * Set the checksum property on a dnode.  The new checksum algorithm will
307  * apply to all newly written blocks; existing blocks will not be affected.
308  */
309 void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
310     dmu_tx_t *tx);
311 
312 /*
313  * Set the compress property on a dnode.  The new compression algorithm will
314  * apply to all newly written blocks; existing blocks will not be affected.
315  */
316 void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
317     dmu_tx_t *tx);
318 
319 /*
320  * Decide how to write a block: checksum, compression, number of copies, etc.
321  */
322 #define	WP_NOFILL	0x1
323 #define	WP_DMU_SYNC	0x2
324 #define	WP_SPILL	0x4
325 
326 void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
327     struct zio_prop *zp);
328 /*
329  * The bonus data is accessed more or less like a regular buffer.
330  * You must dmu_bonus_hold() to get the buffer, which will give you a
331  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
332  * data.  As with any normal buffer, you must call dmu_buf_read() to
333  * read db_data, dmu_buf_will_dirty() before modifying it, and the
334  * object must be held in an assigned transaction before calling
335  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
336  * buffer as well.  You must release your hold with dmu_buf_rele().
337  */
338 int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
339 int dmu_bonus_max(void);
340 int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
341 int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
342 dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
343 int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
344 
345 /*
346  * Special spill buffer support used by "SA" framework
347  */
348 
349 int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
350 int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
351     void *tag, dmu_buf_t **dbp);
352 int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
353 
354 /*
355  * Obtain the DMU buffer from the specified object which contains the
356  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
357  * that it will remain in memory.  You must release the hold with
358  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
359  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
360  *
361  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
362  * on the returned buffer before reading or writing the buffer's
363  * db_data.  The comments for those routines describe what particular
364  * operations are valid after calling them.
365  *
366  * The object number must be a valid, allocated object number.
367  */
368 int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
369     void *tag, dmu_buf_t **, int flags);
370 void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
371 void dmu_buf_rele(dmu_buf_t *db, void *tag);
372 uint64_t dmu_buf_refcount(dmu_buf_t *db);
373 
374 /*
375  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
376  * range of an object.  A pointer to an array of dmu_buf_t*'s is
377  * returned (in *dbpp).
378  *
379  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
380  * frees the array.  The hold on the array of buffers MUST be released
381  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
382  * individually with dmu_buf_rele.
383  */
384 int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
385     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
386 void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
387 
388 /*
389  * Returns NULL on success, or the existing user ptr if it's already
390  * been set.
391  *
392  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
393  *
394  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
395  * will be set to db->db_data when you are allowed to access it.  Note
396  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
397  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
398  * *user_data_ptr_ptr will be set to the new value when it changes.
399  *
400  * If non-NULL, pageout func will be called when this buffer is being
401  * excised from the cache, so that you can clean up the data structure
402  * pointed to by user_ptr.
403  *
404  * dmu_evict_user() will call the pageout func for all buffers in a
405  * objset with a given pageout func.
406  */
407 void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
408     dmu_buf_evict_func_t *pageout_func);
409 /*
410  * set_user_ie is the same as set_user, but request immediate eviction
411  * when hold count goes to zero.
412  */
413 void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
414     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
415 void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
416     void *user_ptr, void *user_data_ptr_ptr,
417     dmu_buf_evict_func_t *pageout_func);
418 void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
419 
420 /*
421  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
422  */
423 void *dmu_buf_get_user(dmu_buf_t *db);
424 
425 /*
426  * Indicate that you are going to modify the buffer's data (db_data).
427  *
428  * The transaction (tx) must be assigned to a txg (ie. you've called
429  * dmu_tx_assign()).  The buffer's object must be held in the tx
430  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
431  */
432 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
433 
434 /*
435  * Tells if the given dbuf is freeable.
436  */
437 boolean_t dmu_buf_freeable(dmu_buf_t *);
438 
439 /*
440  * You must create a transaction, then hold the objects which you will
441  * (or might) modify as part of this transaction.  Then you must assign
442  * the transaction to a transaction group.  Once the transaction has
443  * been assigned, you can modify buffers which belong to held objects as
444  * part of this transaction.  You can't modify buffers before the
445  * transaction has been assigned; you can't modify buffers which don't
446  * belong to objects which this transaction holds; you can't hold
447  * objects once the transaction has been assigned.  You may hold an
448  * object which you are going to free (with dmu_object_free()), but you
449  * don't have to.
450  *
451  * You can abort the transaction before it has been assigned.
452  *
453  * Note that you may hold buffers (with dmu_buf_hold) at any time,
454  * regardless of transaction state.
455  */
456 
457 #define	DMU_NEW_OBJECT	(-1ULL)
458 #define	DMU_OBJECT_END	(-1ULL)
459 
460 dmu_tx_t *dmu_tx_create(objset_t *os);
461 void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
462 void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
463     uint64_t len);
464 void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
465 void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
466 void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
467 void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
468 void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
469 void dmu_tx_abort(dmu_tx_t *tx);
470 int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
471 void dmu_tx_wait(dmu_tx_t *tx);
472 void dmu_tx_commit(dmu_tx_t *tx);
473 
474 /*
475  * To register a commit callback, dmu_tx_callback_register() must be called.
476  *
477  * dcb_data is a pointer to caller private data that is passed on as a
478  * callback parameter. The caller is responsible for properly allocating and
479  * freeing it.
480  *
481  * When registering a callback, the transaction must be already created, but
482  * it cannot be committed or aborted. It can be assigned to a txg or not.
483  *
484  * The callback will be called after the transaction has been safely written
485  * to stable storage and will also be called if the dmu_tx is aborted.
486  * If there is any error which prevents the transaction from being committed to
487  * disk, the callback will be called with a value of error != 0.
488  */
489 typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
490 
491 void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
492     void *dcb_data);
493 
494 /*
495  * Free up the data blocks for a defined range of a file.  If size is
496  * zero, the range from offset to end-of-file is freed.
497  */
498 int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
499 	uint64_t size, dmu_tx_t *tx);
500 int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
501 	uint64_t size);
502 int dmu_free_object(objset_t *os, uint64_t object);
503 
504 /*
505  * Convenience functions.
506  *
507  * Canfail routines will return 0 on success, or an errno if there is a
508  * nonrecoverable I/O error.
509  */
510 #define	DMU_READ_PREFETCH	0 /* prefetch */
511 #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
512 int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
513 	void *buf, uint32_t flags);
514 void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
515 	const void *buf, dmu_tx_t *tx);
516 void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
517 	dmu_tx_t *tx);
518 int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
519 int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
520     dmu_tx_t *tx);
521 int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
522     dmu_tx_t *tx);
523 int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
524     uint64_t size, struct page *pp, dmu_tx_t *tx);
525 struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
526 void dmu_return_arcbuf(struct arc_buf *buf);
527 void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
528     dmu_tx_t *tx);
529 int dmu_xuio_init(struct xuio *uio, int niov);
530 void dmu_xuio_fini(struct xuio *uio);
531 int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
532     size_t n);
533 int dmu_xuio_cnt(struct xuio *uio);
534 struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
535 void dmu_xuio_clear(struct xuio *uio, int i);
536 void xuio_stat_wbuf_copied();
537 void xuio_stat_wbuf_nocopy();
538 
539 extern int zfs_prefetch_disable;
540 
541 /*
542  * Asynchronously try to read in the data.
543  */
544 void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
545     uint64_t len);
546 
547 typedef struct dmu_object_info {
548 	/* All sizes are in bytes unless otherwise indicated. */
549 	uint32_t doi_data_block_size;
550 	uint32_t doi_metadata_block_size;
551 	dmu_object_type_t doi_type;
552 	dmu_object_type_t doi_bonus_type;
553 	uint64_t doi_bonus_size;
554 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
555 	uint8_t doi_checksum;
556 	uint8_t doi_compress;
557 	uint8_t doi_pad[5];
558 	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
559 	uint64_t doi_max_offset;
560 	uint64_t doi_fill_count;		/* number of non-empty blocks */
561 } dmu_object_info_t;
562 
563 typedef void arc_byteswap_func_t(void *buf, size_t size);
564 
565 typedef struct dmu_object_type_info {
566 	arc_byteswap_func_t	*ot_byteswap;
567 	boolean_t		ot_metadata;
568 	char			*ot_name;
569 } dmu_object_type_info_t;
570 
571 extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
572 
573 /*
574  * Get information on a DMU object.
575  *
576  * Return 0 on success or ENOENT if object is not allocated.
577  *
578  * If doi is NULL, just indicates whether the object exists.
579  */
580 int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
581 void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
582 void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
583 void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
584     u_longlong_t *nblk512);
585 
586 typedef struct dmu_objset_stats {
587 	uint64_t dds_num_clones; /* number of clones of this */
588 	uint64_t dds_creation_txg;
589 	uint64_t dds_guid;
590 	dmu_objset_type_t dds_type;
591 	uint8_t dds_is_snapshot;
592 	uint8_t dds_inconsistent;
593 	char dds_origin[MAXNAMELEN];
594 } dmu_objset_stats_t;
595 
596 /*
597  * Get stats on a dataset.
598  */
599 void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
600 
601 /*
602  * Add entries to the nvlist for all the objset's properties.  See
603  * zfs_prop_table[] and zfs(1m) for details on the properties.
604  */
605 void dmu_objset_stats(objset_t *os, struct nvlist *nv);
606 
607 /*
608  * Get the space usage statistics for statvfs().
609  *
610  * refdbytes is the amount of space "referenced" by this objset.
611  * availbytes is the amount of space available to this objset, taking
612  * into account quotas & reservations, assuming that no other objsets
613  * use the space first.  These values correspond to the 'referenced' and
614  * 'available' properties, described in the zfs(1m) manpage.
615  *
616  * usedobjs and availobjs are the number of objects currently allocated,
617  * and available.
618  */
619 void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
620     uint64_t *usedobjsp, uint64_t *availobjsp);
621 
622 /*
623  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
624  * (Contrast with the ds_guid which is a 64-bit ID that will never
625  * change, so there is a small probability that it will collide.)
626  */
627 uint64_t dmu_objset_fsid_guid(objset_t *os);
628 
629 /*
630  * Get the [cm]time for an objset's snapshot dir
631  */
632 timestruc_t dmu_objset_snap_cmtime(objset_t *os);
633 
634 int dmu_objset_is_snapshot(objset_t *os);
635 
636 extern struct spa *dmu_objset_spa(objset_t *os);
637 extern struct zilog *dmu_objset_zil(objset_t *os);
638 extern struct dsl_pool *dmu_objset_pool(objset_t *os);
639 extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
640 extern void dmu_objset_name(objset_t *os, char *buf);
641 extern dmu_objset_type_t dmu_objset_type(objset_t *os);
642 extern uint64_t dmu_objset_id(objset_t *os);
643 extern uint64_t dmu_objset_syncprop(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 int dmu_send_estimate(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorign,
710     uint64_t *sizep);
711 
712 typedef struct dmu_recv_cookie {
713 	/*
714 	 * This structure is opaque!
715 	 *
716 	 * If logical and real are different, we are recving the stream
717 	 * into the "real" temporary clone, and then switching it with
718 	 * the "logical" target.
719 	 */
720 	struct dsl_dataset *drc_logical_ds;
721 	struct dsl_dataset *drc_real_ds;
722 	struct drr_begin *drc_drrb;
723 	char *drc_tosnap;
724 	char *drc_top_ds;
725 	boolean_t drc_newfs;
726 	boolean_t drc_force;
727 	struct avl_tree *drc_guid_to_ds_map;
728 } dmu_recv_cookie_t;
729 
730 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
731     boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
732 int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp,
733     int cleanup_fd, uint64_t *action_handlep);
734 int dmu_recv_end(dmu_recv_cookie_t *drc);
735 
736 int dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp,
737     offset_t *off);
738 
739 /* CRC64 table */
740 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
741 extern uint64_t zfs_crc64_table[256];
742 
743 #ifdef	__cplusplus
744 }
745 #endif
746 
747 #endif	/* _SYS_DMU_H */
748