dmu.h revision 088f389458728c464569a5506b58070254fa4f7d
1fa9e406ahrens/*
2fa9e406ahrens * CDDL HEADER START
3fa9e406ahrens *
4fa9e406ahrens * The contents of this file are subject to the terms of the
5ea8dc4beschrock * Common Development and Distribution License (the "License").
6ea8dc4beschrock * You may not use this file except in compliance with the License.
7fa9e406ahrens *
8fa9e406ahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e406ahrens * or http://www.opensolaris.org/os/licensing.
10fa9e406ahrens * See the License for the specific language governing permissions
11fa9e406ahrens * and limitations under the License.
12fa9e406ahrens *
13fa9e406ahrens * When distributing Covered Code, include this CDDL HEADER in each
14fa9e406ahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e406ahrens * If applicable, add the following below this CDDL HEADER, with the
16fa9e406ahrens * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e406ahrens * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e406ahrens *
19fa9e406ahrens * CDDL HEADER END
20fa9e406ahrens */
21fa9e406ahrens/*
2247f263fek * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23fa9e406ahrens * Use is subject to license terms.
24fa9e406ahrens */
25fa9e406ahrens
26fa9e406ahrens#ifndef	_SYS_DMU_H
27fa9e406ahrens#define	_SYS_DMU_H
28fa9e406ahrens
29fa9e406ahrens#pragma ident	"%Z%%M%	%I%	%E% SMI"
30fa9e406ahrens
31fa9e406ahrens/*
32fa9e406ahrens * This file describes the interface that the DMU provides for its
33fa9e406ahrens * consumers.
34fa9e406ahrens *
35fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
36fa9e406ahrens * dmu_spa.h.
37fa9e406ahrens */
38fa9e406ahrens
39fa9e406ahrens#include <sys/inttypes.h>
40fa9e406ahrens#include <sys/types.h>
41fa9e406ahrens#include <sys/param.h>
42ecd6cf8marks#include <sys/cred.h>
43fa9e406ahrens
44fa9e406ahrens#ifdef	__cplusplus
45fa9e406ahrensextern "C" {
46fa9e406ahrens#endif
47fa9e406ahrens
48fa9e406ahrensstruct uio;
4944eda4dmaybeestruct page;
50fa9e406ahrensstruct vnode;
51fa9e406ahrensstruct spa;
52fa9e406ahrensstruct zilog;
53fa9e406ahrensstruct zio;
54fa9e406ahrensstruct blkptr;
55fa9e406ahrensstruct zap_cursor;
56fa9e406ahrensstruct dsl_dataset;
57fa9e406ahrensstruct dsl_pool;
58fa9e406ahrensstruct dnode;
59fa9e406ahrensstruct drr_begin;
60fa9e406ahrensstruct drr_end;
6144cd46cbillmstruct zbookmark;
6244cd46cbillmstruct spa;
63a2eea2eahrensstruct nvlist;
64d0ad202ahrensstruct objset_impl;
65fa9e406ahrens
66fa9e406ahrenstypedef struct objset objset_t;
67fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
68fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
69fa9e406ahrens
70fa9e406ahrenstypedef enum dmu_object_type {
71fa9e406ahrens	DMU_OT_NONE,
72fa9e406ahrens	/* general: */
73fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
74fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
75fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
76fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
77fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
78fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
79fa9e406ahrens	/* spa: */
80fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
81fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
82fa9e406ahrens	/* zil: */
83fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
84fa9e406ahrens	/* dmu: */
85fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
86fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
87fa9e406ahrens	/* dsl: */
881649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
8987e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
9087e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
91fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
921649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
93fa9e406ahrens	/* zpl: */
94fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
95da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
96fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
97fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
98fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
99893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
100fa9e406ahrens	/* zvol: */
101fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
102fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
103fa9e406ahrens	/* other; for testing only! */
104fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
105fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
106fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
107ea8dc4beschrock	/* new object types: */
108ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
10906eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
11006eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
111b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
112ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
113da6c28aamw	DMU_OT_ACL,			/* ACL */
114da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
115da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
116da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
117088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
118088f389ahrens	DMU_OT_SCRUB_QUEUE,		/* ZAP */
119fa9e406ahrens	DMU_OT_NUMTYPES
120fa9e406ahrens} dmu_object_type_t;
121fa9e406ahrens
122fa9e406ahrenstypedef enum dmu_objset_type {
123fa9e406ahrens	DMU_OST_NONE,
124fa9e406ahrens	DMU_OST_META,
125fa9e406ahrens	DMU_OST_ZFS,
126fa9e406ahrens	DMU_OST_ZVOL,
127fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
128fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
129fa9e406ahrens	DMU_OST_NUMTYPES
130fa9e406ahrens} dmu_objset_type_t;
131fa9e406ahrens
132fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
133fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
134fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
135fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
136fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
137da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
138fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
139fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
140fa9e406ahrens
141745cd3cmaybee#define	DS_MODE_NOHOLD		0	/* internal use only */
142745cd3cmaybee#define	DS_MODE_USER		1	/* simple access, no special needs */
143745cd3cmaybee#define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
144745cd3cmaybee#define	DS_MODE_TYPE_MASK	0x3
145745cd3cmaybee#define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
146fa9e406ahrens#define	DS_MODE_READONLY	0x8
147fa9e406ahrens#define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
148e193023bonwick#define	DS_MODE_INCONSISTENT	0x10
149e193023bonwick#define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
150fa9e406ahrens
1510b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
1520b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
153fa9e406ahrens
154fa9e406ahrens/*
155fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
156fa9e406ahrens * operation, including metadata.
157fa9e406ahrens */
158fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
159cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
160fa9e406ahrens
161fa9e406ahrens/*
162fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
163fa9e406ahrens */
164fa9e406ahrensint dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
165fa9e406ahrens    objset_t **osp);
1663cb34c6ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
1673cb34c6ahrens    objset_t **osp);
168fa9e406ahrensvoid dmu_objset_close(objset_t *os);
1691934e92maybeeint dmu_objset_evict_dbufs(objset_t *os);
170fa9e406ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type,
171ab04eb8timh    objset_t *clone_parent, uint64_t flags,
172ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
173fa9e406ahrensint dmu_objset_destroy(const char *name);
1741d452cfahrensint dmu_snapshots_destroy(char *fsname, char *snapname);
1754ccbb6eahrensint dmu_objset_rollback(objset_t *os);
1761d452cfahrensint dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
177cdf5b4cmmusanteint dmu_objset_rename(const char *name, const char *newname,
178cdf5b4cmmusante    boolean_t recursive);
1791d452cfahrensint dmu_objset_find(char *name, int func(char *, void *), void *arg,
180fa9e406ahrens    int flags);
181fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
182fa9e406ahrens
183fa9e406ahrenstypedef struct dmu_buf {
184fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
185fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
186fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
187fa9e406ahrens	void *db_data;			/* data in buffer */
188fa9e406ahrens} dmu_buf_t;
189fa9e406ahrens
190fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
191fa9e406ahrens
192fa9e406ahrens/*
19399653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
19499653d4eschrock */
195fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
196fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
197fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
198fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
199ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
200ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
20199653d4eschrock#define	DMU_POOL_SPARES			"spares"
20299653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
20306eeb2aek#define	DMU_POOL_HISTORY		"history"
204b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
205fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
206fa9e406ahrens
207088f389ahrens/* 4x8 zbookmark_t */
208088f389ahrens#define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
209088f389ahrens/* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
210088f389ahrens#define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
211088f389ahrens/* 1x8 txg */
212088f389ahrens#define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
213088f389ahrens/* 1x8 txg */
214088f389ahrens#define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
215088f389ahrens/* 1x4 enum scrub_func */
216088f389ahrens#define	DMU_POOL_SCRUB_FUNC		"scrub_func"
217088f389ahrens/* 1x8 count */
218088f389ahrens#define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
219088f389ahrens
220fa9e406ahrens/*
221fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
222fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
223fa9e406ahrens *
224fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
225fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
226fa9e406ahrens * newly allocated object in this transaction).
227fa9e406ahrens *
228fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
229fa9e406ahrens *
230fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
231fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
232fa9e406ahrens *
233fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
234fa9e406ahrens */
235fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
236fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
237fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
238fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
239fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
240fa9e406ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
241fa9e406ahrens
242fa9e406ahrens/*
243fa9e406ahrens * Free an object from this objset.
244fa9e406ahrens *
245fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
246fa9e406ahrens * dmu_free(object, 0, -1, tx)).
247fa9e406ahrens *
248fa9e406ahrens * The object need not be held in the transaction.
249fa9e406ahrens *
250fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
251fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
252fa9e406ahrens * free it; it fails and returns EBUSY.
253fa9e406ahrens *
254fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
255fa9e406ahrens *
256fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
257fa9e406ahrens */
258fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
259fa9e406ahrens
260fa9e406ahrens/*
261fa9e406ahrens * Find the next allocated or free object.
262fa9e406ahrens *
263fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
2646754306ahrens * object which is allocated.  Ignore objects which have not been
2656754306ahrens * modified since txg.
266fa9e406ahrens *
267fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
268fa9e406ahrens *
269fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
270fa9e406ahrens */
2716754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
2726754306ahrens    boolean_t hole, uint64_t txg);
273fa9e406ahrens
274fa9e406ahrens/*
275fa9e406ahrens * Set the data blocksize for an object.
276fa9e406ahrens *
277fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
278fa9e406ahrens * the first block is allocated already, the new size must be greater
279fa9e406ahrens * than the current block size.  If these conditions are not met,
280fa9e406ahrens * ENOTSUP will be returned.
281fa9e406ahrens *
282fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
283fa9e406ahrens * contents, or ENOTSUP as described above.
284fa9e406ahrens */
285fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
286fa9e406ahrens    int ibs, dmu_tx_t *tx);
287fa9e406ahrens
288fa9e406ahrens/*
289fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
290fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
291fa9e406ahrens */
292fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
293fa9e406ahrens    dmu_tx_t *tx);
294fa9e406ahrens
295fa9e406ahrens/*
296fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
297fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
298fa9e406ahrens */
299fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
300fa9e406ahrens    dmu_tx_t *tx);
301fa9e406ahrens
302fa9e406ahrens/*
30344cd46cbillm * Decide how many copies of a given block we should make.  Can be from
30444cd46cbillm * 1 to SPA_DVAS_PER_BP.
30544cd46cbillm */
306d0ad202ahrensint dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
30744cd46cbillm    dmu_object_type_t ot);
30844cd46cbillm/*
309fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
310fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
311fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
312fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
313fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
314fa9e406ahrens * object must be held in an assigned transaction before calling
315fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
316fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
317fa9e406ahrens */
318ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
319fa9e406ahrensint dmu_bonus_max(void);
3201934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
321fa9e406ahrens
322fa9e406ahrens/*
323fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
324fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
325fa9e406ahrens * that it will remain in memory.  You must release the hold with
326fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
327fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
328fa9e406ahrens *
329fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
330fa9e406ahrens * on the returned buffer before reading or writing the buffer's
331fa9e406ahrens * db_data.  The comments for those routines describe what particular
332fa9e406ahrens * operations are valid after calling them.
333fa9e406ahrens *
334fa9e406ahrens * The object number must be a valid, allocated object number.
335fa9e406ahrens */
336ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
337ea8dc4beschrock    void *tag, dmu_buf_t **);
338fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
339ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
340fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
341fa9e406ahrens
342fa9e406ahrens/*
343fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
344fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
345fa9e406ahrens * returned (in *dbpp).
346fa9e406ahrens *
347fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
348fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
349fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
350fa9e406ahrens * individually with dmu_buf_rele.
351fa9e406ahrens */
35213506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
35313506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
354ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
355fa9e406ahrens
356fa9e406ahrens/*
357fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
358fa9e406ahrens * been set.
359fa9e406ahrens *
360fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
361fa9e406ahrens *
362fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
363fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
364fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
365fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
366fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
367fa9e406ahrens *
368fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
369fa9e406ahrens * excised from the cache, so that you can clean up the data structure
370fa9e406ahrens * pointed to by user_ptr.
371fa9e406ahrens *
372fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
373fa9e406ahrens * objset with a given pageout func.
374fa9e406ahrens */
375fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
376fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
377fa9e406ahrens/*
378fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
379fa9e406ahrens * when hold count goes to zero.
380fa9e406ahrens */
381fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
382fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
383fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
384fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
385fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
386fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
387fa9e406ahrens
388fa9e406ahrens/*
389fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
390fa9e406ahrens */
391fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
392fa9e406ahrens
393fa9e406ahrens/*
394fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
395fa9e406ahrens *
396fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
397fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
398fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
399fa9e406ahrens */
400fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
401fa9e406ahrens
402fa9e406ahrens/*
403fa9e406ahrens * You must create a transaction, then hold the objects which you will
404fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
405fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
406fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
407fa9e406ahrens * part of this transaction.  You can't modify buffers before the
408fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
409fa9e406ahrens * belong to objects which this transaction holds; you can't hold
410fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
411fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
412fa9e406ahrens * don't have to.
413fa9e406ahrens *
414fa9e406ahrens * You can abort the transaction before it has been assigned.
415fa9e406ahrens *
416fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
417fa9e406ahrens * regardless of transaction state.
418fa9e406ahrens */
419fa9e406ahrens
420fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
421fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
422fa9e406ahrens
423fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
424fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
425fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
426fa9e406ahrens    uint64_t len);
427ea8dc4beschrockvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
428fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
429fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
430fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4318a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
432fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
433fa9e406ahrens
434fa9e406ahrens/*
435fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
436fa9e406ahrens * zero, the range from offset to end-of-file is freed.
437fa9e406ahrens */
438ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
439fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
440cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
441cdb0ab7maybee	uint64_t size);
442cdb0ab7maybeeint dmu_free_object(objset_t *os, uint64_t object);
443fa9e406ahrens
444fa9e406ahrens/*
445fa9e406ahrens * Convenience functions.
446fa9e406ahrens *
447fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
448fa9e406ahrens * nonrecoverable I/O error.
449fa9e406ahrens */
450ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
451fa9e406ahrens	void *buf);
452fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
453fa9e406ahrens	const void *buf, dmu_tx_t *tx);
454feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
455feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
456feb08c6billm    dmu_tx_t *tx);
45744eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
45844eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
459fa9e406ahrens
460416e0cdekextern int zfs_prefetch_disable;
461416e0cdek
462fa9e406ahrens/*
463fa9e406ahrens * Asynchronously try to read in the data.
464fa9e406ahrens */
465fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
466fa9e406ahrens    uint64_t len);
467fa9e406ahrens
468fa9e406ahrenstypedef struct dmu_object_info {
469fa9e406ahrens	/* All sizes are in bytes. */
470fa9e406ahrens	uint32_t doi_data_block_size;
471fa9e406ahrens	uint32_t doi_metadata_block_size;
472fa9e406ahrens	uint64_t doi_bonus_size;
473fa9e406ahrens	dmu_object_type_t doi_type;
474fa9e406ahrens	dmu_object_type_t doi_bonus_type;
475fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
476fa9e406ahrens	uint8_t doi_checksum;
477fa9e406ahrens	uint8_t doi_compress;
478fa9e406ahrens	uint8_t doi_pad[5];
479fa9e406ahrens	/* Values below are number of 512-byte blocks. */
480fa9e406ahrens	uint64_t doi_physical_blks;		/* data + metadata */
481fa9e406ahrens	uint64_t doi_max_block_offset;
482fa9e406ahrens} dmu_object_info_t;
483fa9e406ahrens
484e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
485e45ce72ahrens
486fa9e406ahrenstypedef struct dmu_object_type_info {
487e45ce72ahrens	arc_byteswap_func_t	*ot_byteswap;
488fa9e406ahrens	boolean_t		ot_metadata;
489fa9e406ahrens	char			*ot_name;
490fa9e406ahrens} dmu_object_type_info_t;
491fa9e406ahrens
492fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
493fa9e406ahrens
494fa9e406ahrens/*
495fa9e406ahrens * Get information on a DMU object.
496fa9e406ahrens *
497fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
498fa9e406ahrens *
499fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
500fa9e406ahrens */
501fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
502fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
503fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
504fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
505fa9e406ahrens    u_longlong_t *nblk512);
506fa9e406ahrens
507fa9e406ahrenstypedef struct dmu_objset_stats {
508a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
509a2eea2eahrens	uint64_t dds_creation_txg;
5103cb34c6ahrens	uint64_t dds_guid;
511fa9e406ahrens	dmu_objset_type_t dds_type;
512fa9e406ahrens	uint8_t dds_is_snapshot;
51331fd60dahrens	uint8_t dds_inconsistent;
5143cb34c6ahrens	char dds_origin[MAXNAMELEN];
515a2eea2eahrens} dmu_objset_stats_t;
516fa9e406ahrens
517a2eea2eahrens/*
518a2eea2eahrens * Get stats on a dataset.
519a2eea2eahrens */
520a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
521fa9e406ahrens
522a2eea2eahrens/*
523a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
524a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
525a2eea2eahrens */
526a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
527fa9e406ahrens
528a2eea2eahrens/*
529a2eea2eahrens * Get the space usage statistics for statvfs().
530a2eea2eahrens *
531a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
532a2eea2eahrens * availbytes is the amount of space available to this objset, taking
533a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
534a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
535a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
536a2eea2eahrens *
537a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
538a2eea2eahrens * and available.
539a2eea2eahrens */
540a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
541a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
542fa9e406ahrens
543fa9e406ahrens/*
544a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
545a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
546a2eea2eahrens * change, so there is a small probability that it will collide.)
547fa9e406ahrens */
548a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
549fa9e406ahrens
550fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
551fa9e406ahrens
552fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
553fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
554fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
555fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
556fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
557fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
558fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
559fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
560b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
561ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
562ab04eb8timh    int maxlen, boolean_t *conflict);
56387e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
56487e5029ahrens    uint64_t *idp, uint64_t *offp);
565f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
566f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
567fa9e406ahrens
568fa9e406ahrens/*
569fa9e406ahrens * Return the txg number for the given assigned transaction.
570fa9e406ahrens */
57187e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
572fa9e406ahrens
573fa9e406ahrens/*
574fa9e406ahrens * Synchronous write.
575c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
576c5c6ffamaybee * provided buffer as a child of the parent zio.
577da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
578c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
579fa9e406ahrens * Note that while the data covered by this function will be on stable
580c5c6ffamaybee * storage when the write completes this new data does not become a
581fa9e406ahrens * permanent part of the file until the associated transaction commits.
582fa9e406ahrens */
583c5c6ffamaybeetypedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
584c5c6ffamaybeeint dmu_sync(struct zio *zio, dmu_buf_t *db,
585c5c6ffamaybee    struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
586fa9e406ahrens
587fa9e406ahrens/*
588fa9e406ahrens * Find the next hole or data block in file starting at *off
589fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
590fa9e406ahrens */
591fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
592fa9e406ahrens    uint64_t *off);
593fa9e406ahrens
594fa9e406ahrens/*
595fa9e406ahrens * Initial setup and final teardown.
596fa9e406ahrens */
597fa9e406ahrensextern void dmu_init(void);
598fa9e406ahrensextern void dmu_fini(void);
599fa9e406ahrens
600fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
601fa9e406ahrens    uint64_t object, uint64_t offset, int len);
602fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
603fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
604fa9e406ahrens
6053cb34c6ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
6063cb34c6ahrens    struct vnode *vp, offset_t *off);
6073cb34c6ahrens
6083cb34c6ahrenstypedef struct dmu_recv_cookie {
6093cb34c6ahrens	/*
6103cb34c6ahrens	 * This structure is opaque!
6113cb34c6ahrens	 *
6123cb34c6ahrens	 * If logical and real are different, we are recving the stream
6133cb34c6ahrens	 * into the "real" temporary clone, and then switching it with
6143cb34c6ahrens	 * the "logical" target.
6153cb34c6ahrens	 */
6163cb34c6ahrens	struct dsl_dataset *drc_logical_ds;
6173cb34c6ahrens	struct dsl_dataset *drc_real_ds;
6183cb34c6ahrens	struct drr_begin *drc_drrb;
6193cb34c6ahrens	char *drc_tosnap;
6203cb34c6ahrens	boolean_t drc_newfs;
6213cb34c6ahrens	boolean_t drc_force;
6223cb34c6ahrens} dmu_recv_cookie_t;
6233cb34c6ahrens
6243cb34c6ahrensint dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
6253cb34c6ahrens    boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
6263cb34c6ahrensint dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
6273cb34c6ahrensint dmu_recv_end(dmu_recv_cookie_t *drc);
62847f263fekvoid dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
629fa9e406ahrens
630fa9e406ahrens/* CRC64 table */
631fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
632fa9e406ahrensextern uint64_t zfs_crc64_table[256];
633fa9e406ahrens
634fa9e406ahrens#ifdef	__cplusplus
635fa9e406ahrens}
636fa9e406ahrens#endif
637fa9e406ahrens
638fa9e406ahrens#endif	/* _SYS_DMU_H */
639