dmu.h revision 5d7b4d438c4a51eccc95e77a83a437b4d48380eb
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 */
21ad135b5Christopher Siden
22fa9e406ahrens/*
2394d1a21Tim Haley * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24edf345eMatthew Ahrens * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
25ec5cf9dAlexander Stetsenko * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
264e3c9f4Bill Pijewski * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27e4be62aSaso Kiselkov * Copyright 2013 DEY Storage Systems, Inc.
28ec5cf9dAlexander Stetsenko */
29fa9e406ahrens
3055da60bMark J Musante/* Portions Copyright 2010 Robert Milkowski */
3155da60bMark J Musante
32fa9e406ahrens#ifndef	_SYS_DMU_H
33fa9e406ahrens#define	_SYS_DMU_H
34fa9e406ahrens
35fa9e406ahrens/*
36fa9e406ahrens * This file describes the interface that the DMU provides for its
37fa9e406ahrens * consumers.
38fa9e406ahrens *
39fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
40fa9e406ahrens * dmu_spa.h.
41fa9e406ahrens */
42fa9e406ahrens
43fa9e406ahrens#include <sys/inttypes.h>
44fa9e406ahrens#include <sys/types.h>
45fa9e406ahrens#include <sys/param.h>
46ecd6cf8marks#include <sys/cred.h>
4771eb053Chris Kirby#include <sys/time.h>
484445fffMatthew Ahrens#include <sys/fs/zfs.h>
49fa9e406ahrens
50fa9e406ahrens#ifdef	__cplusplus
51fa9e406ahrensextern "C" {
52fa9e406ahrens#endif
53fa9e406ahrens
54fa9e406ahrensstruct uio;
55c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct xuio;
5644eda4dmaybeestruct page;
57fa9e406ahrensstruct vnode;
58fa9e406ahrensstruct spa;
59fa9e406ahrensstruct zilog;
60fa9e406ahrensstruct zio;
61fa9e406ahrensstruct blkptr;
62fa9e406ahrensstruct zap_cursor;
63fa9e406ahrensstruct dsl_dataset;
64fa9e406ahrensstruct dsl_pool;
65fa9e406ahrensstruct dnode;
66fa9e406ahrensstruct drr_begin;
67fa9e406ahrensstruct drr_end;
6844cd46cbillmstruct zbookmark;
6944cd46cbillmstruct spa;
70a2eea2eahrensstruct nvlist;
712fdbea2Aleksandr Guzovskiystruct arc_buf;
72b24ab67Jeff Bonwickstruct zio_prop;
730a586ceMark Shellenbaumstruct sa_handle;
74fa9e406ahrens
75fa9e406ahrenstypedef struct objset objset_t;
76fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
77fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
78fa9e406ahrens
79ad135b5Christopher Sidentypedef enum dmu_object_byteswap {
80ad135b5Christopher Siden	DMU_BSWAP_UINT8,
81ad135b5Christopher Siden	DMU_BSWAP_UINT16,
82ad135b5Christopher Siden	DMU_BSWAP_UINT32,
83ad135b5Christopher Siden	DMU_BSWAP_UINT64,
84ad135b5Christopher Siden	DMU_BSWAP_ZAP,
85ad135b5Christopher Siden	DMU_BSWAP_DNODE,
86ad135b5Christopher Siden	DMU_BSWAP_OBJSET,
87ad135b5Christopher Siden	DMU_BSWAP_ZNODE,
88ad135b5Christopher Siden	DMU_BSWAP_OLDACL,
89ad135b5Christopher Siden	DMU_BSWAP_ACL,
90ad135b5Christopher Siden	/*
91ad135b5Christopher Siden	 * Allocating a new byteswap type number makes the on-disk format
92ad135b5Christopher Siden	 * incompatible with any other format that uses the same number.
93ad135b5Christopher Siden	 *
94ad135b5Christopher Siden	 * Data can usually be structured to work with one of the
95ad135b5Christopher Siden	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
96ad135b5Christopher Siden	 */
97ad135b5Christopher Siden	DMU_BSWAP_NUMFUNCS
98ad135b5Christopher Siden} dmu_object_byteswap_t;
99ad135b5Christopher Siden
100ad135b5Christopher Siden#define	DMU_OT_NEWTYPE 0x80
101ad135b5Christopher Siden#define	DMU_OT_METADATA 0x40
102ad135b5Christopher Siden#define	DMU_OT_BYTESWAP_MASK 0x3f
103ad135b5Christopher Siden
104ad135b5Christopher Siden/*
105ad135b5Christopher Siden * Defines a uint8_t object type. Object types specify if the data
106ad135b5Christopher Siden * in the object is metadata (boolean) and how to byteswap the data
107ad135b5Christopher Siden * (dmu_object_byteswap_t).
108ad135b5Christopher Siden */
109ad135b5Christopher Siden#define	DMU_OT(byteswap, metadata) \
110ad135b5Christopher Siden	(DMU_OT_NEWTYPE | \
111ad135b5Christopher Siden	((metadata) ? DMU_OT_METADATA : 0) | \
112ad135b5Christopher Siden	((byteswap) & DMU_OT_BYTESWAP_MASK))
113ad135b5Christopher Siden
114ad135b5Christopher Siden#define	DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
115ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
116ad135b5Christopher Siden	(ot) < DMU_OT_NUMTYPES)
117ad135b5Christopher Siden
118ad135b5Christopher Siden#define	DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
119ad135b5Christopher Siden	((ot) & DMU_OT_METADATA) : \
120ad135b5Christopher Siden	dmu_ot[(ot)].ot_metadata)
121ad135b5Christopher Siden
1225d7b4d4Matthew Ahrens/*
1235d7b4d4Matthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
1245d7b4d4Matthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
1255d7b4d4Matthew Ahrens * is repurposed for embedded BPs.
1265d7b4d4Matthew Ahrens */
1275d7b4d4Matthew Ahrens#define	DMU_OT_HAS_FILL(ot) \
1285d7b4d4Matthew Ahrens	((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
1295d7b4d4Matthew Ahrens
130ad135b5Christopher Siden#define	DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
131ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) : \
132ad135b5Christopher Siden	dmu_ot[(ot)].ot_byteswap)
133ad135b5Christopher Siden
134fa9e406ahrenstypedef enum dmu_object_type {
135fa9e406ahrens	DMU_OT_NONE,
136fa9e406ahrens	/* general: */
137fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
138fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
139fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
140fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
141cde58dbMatthew Ahrens	DMU_OT_BPOBJ,			/* UINT64 */
142cde58dbMatthew Ahrens	DMU_OT_BPOBJ_HDR,		/* UINT64 */
143fa9e406ahrens	/* spa: */
144fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
145fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
146fa9e406ahrens	/* zil: */
147fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
148fa9e406ahrens	/* dmu: */
149fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
150fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
151fa9e406ahrens	/* dsl: */
1521649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
15387e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
15487e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
155fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
1561649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
157fa9e406ahrens	/* zpl: */
158fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
159da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
160fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
161fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
162fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
163893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
164fa9e406ahrens	/* zvol: */
165fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
166fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
167fa9e406ahrens	/* other; for testing only! */
168fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
169fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
170fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
171ea8dc4beschrock	/* new object types: */
172ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
17306eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
17406eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
175b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
176ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
177da6c28aamw	DMU_OT_ACL,			/* ACL */
178da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
179da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
180da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
181088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
1823f9d6adLin Ling	DMU_OT_SCAN_QUEUE,		/* ZAP */
1831484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1841484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
185842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
186b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
187b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
1880a586ceMark Shellenbaum	DMU_OT_SA,			/* System attr */
1890a586ceMark Shellenbaum	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1900a586ceMark Shellenbaum	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1910a586ceMark Shellenbaum	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
1923f9d6adLin Ling	DMU_OT_SCAN_XLATE,		/* ZAP */
1933f9d6adLin Ling	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
194cde58dbMatthew Ahrens	DMU_OT_DEADLIST,		/* ZAP */
195cde58dbMatthew Ahrens	DMU_OT_DEADLIST_HDR,		/* UINT64 */
196cde58dbMatthew Ahrens	DMU_OT_DSL_CLONES,		/* ZAP */
197cde58dbMatthew Ahrens	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
198ad135b5Christopher Siden	/*
199ad135b5Christopher Siden	 * Do not allocate new object types here. Doing so makes the on-disk
200ad135b5Christopher Siden	 * format incompatible with any other format that uses the same object
201ad135b5Christopher Siden	 * type number.
202ad135b5Christopher Siden	 *
203ad135b5Christopher Siden	 * When creating an object which does not have one of the above types
204ad135b5Christopher Siden	 * use the DMU_OTN_* type with the correct byteswap and metadata
205ad135b5Christopher Siden	 * values.
206ad135b5Christopher Siden	 *
207ad135b5Christopher Siden	 * The DMU_OTN_* types do not have entries in the dmu_ot table,
208ad135b5Christopher Siden	 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
209ad135b5Christopher Siden	 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
210ad135b5Christopher Siden	 * and DMU_OTN_* types).
211ad135b5Christopher Siden	 */
212ad135b5Christopher Siden	DMU_OT_NUMTYPES,
213ad135b5Christopher Siden
214ad135b5Christopher Siden	/*
215ad135b5Christopher Siden	 * Names for valid types declared with DMU_OT().
216ad135b5Christopher Siden	 */
217ad135b5Christopher Siden	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
218ad135b5Christopher Siden	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
219ad135b5Christopher Siden	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
220ad135b5Christopher Siden	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
221ad135b5Christopher Siden	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
222ad135b5Christopher Siden	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
223ad135b5Christopher Siden	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
224ad135b5Christopher Siden	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
225ad135b5Christopher Siden	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
226ad135b5Christopher Siden	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
227fa9e406ahrens} dmu_object_type_t;
228fa9e406ahrens
2293b2aab1Matthew Ahrenstypedef enum txg_how {
2303b2aab1Matthew Ahrens	TXG_WAIT = 1,
2313b2aab1Matthew Ahrens	TXG_NOWAIT,
23269962b5Matthew Ahrens	TXG_WAITED,
2333b2aab1Matthew Ahrens} txg_how_t;
2343b2aab1Matthew Ahrens
235fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
236fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
237fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
238fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
239fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
240da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
241fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
242fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
243fa9e406ahrens
2440b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
2450b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
246fa9e406ahrens
247fa9e406ahrens/*
248fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
249fa9e406ahrens * operation, including metadata.
250fa9e406ahrens */
251fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
252cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
253fa9e406ahrens
2541484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
2551484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
256b24ab67Jeff Bonwick#define	DMU_DEADLIST_OBJECT	(-3ULL)
2571484342Matthew Ahrens
258fa9e406ahrens/*
2590a586ceMark Shellenbaum * artificial blkids for bonus buffer and spill blocks
2600a586ceMark Shellenbaum */
2610a586ceMark Shellenbaum#define	DMU_BONUS_BLKID		(-1ULL)
2620a586ceMark Shellenbaum#define	DMU_SPILL_BLKID		(-2ULL)
2630a586ceMark Shellenbaum/*
264fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
265fa9e406ahrens */
266503ad85Matthew Ahrensint dmu_objset_hold(const char *name, void *tag, objset_t **osp);
267503ad85Matthew Ahrensint dmu_objset_own(const char *name, dmu_objset_type_t type,
268503ad85Matthew Ahrens    boolean_t readonly, void *tag, objset_t **osp);
269503ad85Matthew Ahrensvoid dmu_objset_rele(objset_t *os, void *tag);
270503ad85Matthew Ahrensvoid dmu_objset_disown(objset_t *os, void *tag);
271503ad85Matthew Ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
272503ad85Matthew Ahrens
2733b2aab1Matthew Ahrensvoid dmu_objset_evict_dbufs(objset_t *os);
274ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
275ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
2763b2aab1Matthew Ahrensint dmu_objset_clone(const char *name, const char *origin);
2773b2aab1Matthew Ahrensint dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
2784445fffMatthew Ahrens    struct nvlist *errlist);
2794445fffMatthew Ahrensint dmu_objset_snapshot_one(const char *fsname, const char *snapname);
2804445fffMatthew Ahrensint dmu_objset_snapshot_tmp(const char *, const char *, int);
281fd13687Matthew Ahrensint dmu_objset_find(char *name, int func(const char *, void *), void *arg,
282fa9e406ahrens    int flags);
283fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
2843b2aab1Matthew Ahrensint dsl_dataset_rename_snapshot(const char *fsname,
2853b2aab1Matthew Ahrens    const char *oldsnapname, const char *newsnapname, boolean_t recursive);
286fa9e406ahrens
287fa9e406ahrenstypedef struct dmu_buf {
288fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
289fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
290fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
291fa9e406ahrens	void *db_data;			/* data in buffer */
292fa9e406ahrens} dmu_buf_t;
293fa9e406ahrens
294fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
295fa9e406ahrens
296fa9e406ahrens/*
29799653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
29899653d4eschrock */
299fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
300fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
301ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_WRITE	"features_for_write"
302ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_READ	"features_for_read"
303ad135b5Christopher Siden#define	DMU_POOL_FEATURE_DESCRIPTIONS	"feature_descriptions"
30443466aaMax Grossman#define	DMU_POOL_FEATURE_ENABLED_TXG	"feature_enabled_txg"
305fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
306cde58dbMatthew Ahrens#define	DMU_POOL_SYNC_BPOBJ		"sync_bplist"
307ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
308ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
30999653d4eschrock#define	DMU_POOL_SPARES			"spares"
31099653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
31106eeb2aek#define	DMU_POOL_HISTORY		"history"
312b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
313fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
314ca45db4Chris Kirby#define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
315b24ab67Jeff Bonwick#define	DMU_POOL_DDT			"DDT-%s-%s-%s"
316b24ab67Jeff Bonwick#define	DMU_POOL_DDT_STATS		"DDT-statistics"
3173f9d6adLin Ling#define	DMU_POOL_CREATION_VERSION	"creation_version"
3183f9d6adLin Ling#define	DMU_POOL_SCAN			"scan"
319cde58dbMatthew Ahrens#define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
320ad135b5Christopher Siden#define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
321f174573Matthew Ahrens#define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"
322088f389ahrens
323fa9e406ahrens/*
324fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
325fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
326fa9e406ahrens *
327fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
328fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
329fa9e406ahrens * newly allocated object in this transaction).
330fa9e406ahrens *
331fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
332fa9e406ahrens *
333fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
334fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
335fa9e406ahrens *
336fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
337fa9e406ahrens */
338fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
339fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
340fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
341fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
342fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
3432bf405aMark Maybee    int blocksize, dmu_object_type_t bonustype, int bonuslen);
344fa9e406ahrens
345fa9e406ahrens/*
346fa9e406ahrens * Free an object from this objset.
347fa9e406ahrens *
348fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
349fa9e406ahrens * dmu_free(object, 0, -1, tx)).
350fa9e406ahrens *
351fa9e406ahrens * The object need not be held in the transaction.
352fa9e406ahrens *
353fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
354fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
355fa9e406ahrens * free it; it fails and returns EBUSY.
356fa9e406ahrens *
357fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
358fa9e406ahrens *
359fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
360fa9e406ahrens */
361fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
362fa9e406ahrens
363fa9e406ahrens/*
364fa9e406ahrens * Find the next allocated or free object.
365fa9e406ahrens *
366fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
3676754306ahrens * object which is allocated.  Ignore objects which have not been
3686754306ahrens * modified since txg.
369fa9e406ahrens *
370fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
371fa9e406ahrens *
372fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
373fa9e406ahrens */
3746754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
3756754306ahrens    boolean_t hole, uint64_t txg);
376fa9e406ahrens
377fa9e406ahrens/*
378fa9e406ahrens * Set the data blocksize for an object.
379fa9e406ahrens *
380fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
381fa9e406ahrens * the first block is allocated already, the new size must be greater
382fa9e406ahrens * than the current block size.  If these conditions are not met,
383fa9e406ahrens * ENOTSUP will be returned.
384fa9e406ahrens *
385fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
386fa9e406ahrens * contents, or ENOTSUP as described above.
387fa9e406ahrens */
388fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
389fa9e406ahrens    int ibs, dmu_tx_t *tx);
390fa9e406ahrens
391fa9e406ahrens/*
392fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
393fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
394fa9e406ahrens */
395fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
396fa9e406ahrens    dmu_tx_t *tx);
397fa9e406ahrens
398fa9e406ahrens/*
399fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
400fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
401fa9e406ahrens */
402fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
403fa9e406ahrens    dmu_tx_t *tx);
404fa9e406ahrens
4055d7b4d4Matthew Ahrensvoid
4065d7b4d4Matthew Ahrensdmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
4075d7b4d4Matthew Ahrens    void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
4085d7b4d4Matthew Ahrens    int compressed_size, int byteorder, dmu_tx_t *tx);
4095d7b4d4Matthew Ahrens
410fa9e406ahrens/*
411b24ab67Jeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc.
41244cd46cbillm */
413b24ab67Jeff Bonwick#define	WP_NOFILL	0x1
414b24ab67Jeff Bonwick#define	WP_DMU_SYNC	0x2
4150a586ceMark Shellenbaum#define	WP_SPILL	0x4
416b24ab67Jeff Bonwick
417b24ab67Jeff Bonwickvoid dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
418b24ab67Jeff Bonwick    struct zio_prop *zp);
41944cd46cbillm/*
420fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
421fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
422fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
423fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
424fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
425fa9e406ahrens * object must be held in an assigned transaction before calling
426fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
427fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
4283e30c24Will Andrews *
4293e30c24Will Andrews * Returns ENOENT, EIO, or 0.
430fa9e406ahrens */
431ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
432fa9e406ahrensint dmu_bonus_max(void);
4331934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
4340a586ceMark Shellenbaumint dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
435744947dTom Ericksondmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
4360a586ceMark Shellenbaumint dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
4370a586ceMark Shellenbaum
4380a586ceMark Shellenbaum/*
4390a586ceMark Shellenbaum * Special spill buffer support used by "SA" framework
4400a586ceMark Shellenbaum */
4410a586ceMark Shellenbaum
4420a586ceMark Shellenbaumint dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
4430a586ceMark Shellenbaumint dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
4440a586ceMark Shellenbaum    void *tag, dmu_buf_t **dbp);
4450a586ceMark Shellenbaumint dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
446fa9e406ahrens
447fa9e406ahrens/*
448fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
449fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
450fa9e406ahrens * that it will remain in memory.  You must release the hold with
451fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
452fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
453fa9e406ahrens *
454fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
455fa9e406ahrens * on the returned buffer before reading or writing the buffer's
456fa9e406ahrens * db_data.  The comments for those routines describe what particular
457fa9e406ahrens * operations are valid after calling them.
458fa9e406ahrens *
459fa9e406ahrens * The object number must be a valid, allocated object number.
460fa9e406ahrens */
461ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
46247cb52dJeff Bonwick    void *tag, dmu_buf_t **, int flags);
463fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
464ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
465fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
466fa9e406ahrens
467fa9e406ahrens/*
468fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
469fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
470fa9e406ahrens * returned (in *dbpp).
471fa9e406ahrens *
472fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
473fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
474fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
475fa9e406ahrens * individually with dmu_buf_rele.
476fa9e406ahrens */
47713506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
47813506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
479ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
480fa9e406ahrens
481fa9e406ahrens/*
482fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
483fa9e406ahrens * been set.
484fa9e406ahrens *
485fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
486fa9e406ahrens *
487fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
488fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
489fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
490fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
491fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
492fa9e406ahrens *
493fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
494fa9e406ahrens * excised from the cache, so that you can clean up the data structure
495fa9e406ahrens * pointed to by user_ptr.
496fa9e406ahrens *
497fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
498fa9e406ahrens * objset with a given pageout func.
499fa9e406ahrens */
500fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
501fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
502fa9e406ahrens/*
503fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
504fa9e406ahrens * when hold count goes to zero.
505fa9e406ahrens */
506fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
507fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
508fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
509fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
510fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
511fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
512fa9e406ahrens
513fa9e406ahrens/*
514fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
515fa9e406ahrens */
516fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
517fa9e406ahrens
518fa9e406ahrens/*
51980901aeGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set.
52080901aeGeorge Wilson */
52180901aeGeorge Wilsonstruct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
52280901aeGeorge Wilson
52380901aeGeorge Wilson/*
524fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
525fa9e406ahrens *
526fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
527fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
528fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
529fa9e406ahrens */
530fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
531fa9e406ahrens
532fa9e406ahrens/*
5333d69262Sanjeev Bagewadi * Tells if the given dbuf is freeable.
5343d69262Sanjeev Bagewadi */
5353d69262Sanjeev Bagewadiboolean_t dmu_buf_freeable(dmu_buf_t *);
5363d69262Sanjeev Bagewadi
5373d69262Sanjeev Bagewadi/*
538fa9e406ahrens * You must create a transaction, then hold the objects which you will
539fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
540fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
541fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
542fa9e406ahrens * part of this transaction.  You can't modify buffers before the
543fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
544fa9e406ahrens * belong to objects which this transaction holds; you can't hold
545fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
546fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
547fa9e406ahrens * don't have to.
548fa9e406ahrens *
549fa9e406ahrens * You can abort the transaction before it has been assigned.
550fa9e406ahrens *
551fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
552fa9e406ahrens * regardless of transaction state.
553fa9e406ahrens */
554fa9e406ahrens
555fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
556fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
557fa9e406ahrens
558fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
559fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
560fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
561fa9e406ahrens    uint64_t len);
5621484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
563fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
5640a586ceMark Shellenbaumvoid dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
5650a586ceMark Shellenbaumvoid dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
5660a586ceMark Shellenbaumvoid dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
567fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
5683b2aab1Matthew Ahrensint dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
5698a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
570fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
571fa9e406ahrens
572fa9e406ahrens/*
573d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
574d20e665Ricardo M. Correia *
575d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
576d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
577d20e665Ricardo M. Correia * freeing it.
578d20e665Ricardo M. Correia *
579d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
580d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
581d20e665Ricardo M. Correia *
582d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
583d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
584d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
585d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
586d20e665Ricardo M. Correia */
587d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
588d20e665Ricardo M. Correia
589d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
590d20e665Ricardo M. Correia    void *dcb_data);
591d20e665Ricardo M. Correia
592d20e665Ricardo M. Correia/*
593fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
594ad135b5Christopher Siden * -1, the range from offset to end-of-file is freed.
595fa9e406ahrens */
596ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
597fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
598cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
599cdb0ab7maybee	uint64_t size);
600713d6c2Matthew Ahrensint dmu_free_long_object(objset_t *os, uint64_t object);
601fa9e406ahrens
602fa9e406ahrens/*
603fa9e406ahrens * Convenience functions.
604fa9e406ahrens *
605fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
606fa9e406ahrens * nonrecoverable I/O error.
607fa9e406ahrens */
6087bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
6097bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
610ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
6117bfdf01Neil Perrin	void *buf, uint32_t flags);
612fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
613fa9e406ahrens	const void *buf, dmu_tx_t *tx);
61482c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
61582c9918Tim Haley	dmu_tx_t *tx);
616feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
617feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
618feb08c6billm    dmu_tx_t *tx);
61994d1a21Tim Haleyint dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
62094d1a21Tim Haley    dmu_tx_t *tx);
62144eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
62244eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
6232fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
6242fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
6252fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
6262fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
627c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_init(struct xuio *uio, int niov);
628c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_fini(struct xuio *uio);
629c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
630c242f9achunli zhang - Sun Microsystems - Irvine United States    size_t n);
631c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_cnt(struct xuio *uio);
632c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
633c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_clear(struct xuio *uio, int i);
634c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_copied();
635c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_nocopy();
636fa9e406ahrens
637416e0cdekextern int zfs_prefetch_disable;
638416e0cdek
639fa9e406ahrens/*
640fa9e406ahrens * Asynchronously try to read in the data.
641fa9e406ahrens */
642fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
643fa9e406ahrens    uint64_t len);
644fa9e406ahrens
645fa9e406ahrenstypedef struct dmu_object_info {
646b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
647fa9e406ahrens	uint32_t doi_data_block_size;
648fa9e406ahrens	uint32_t doi_metadata_block_size;
649fa9e406ahrens	dmu_object_type_t doi_type;
650fa9e406ahrens	dmu_object_type_t doi_bonus_type;
651b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
652fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
653fa9e406ahrens	uint8_t doi_checksum;
654fa9e406ahrens	uint8_t doi_compress;
655fa9e406ahrens	uint8_t doi_pad[5];
656b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
657b24ab67Jeff Bonwick	uint64_t doi_max_offset;
658b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
659fa9e406ahrens} dmu_object_info_t;
660fa9e406ahrens
661e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
662e45ce72ahrens
663fa9e406ahrenstypedef struct dmu_object_type_info {
664ad135b5Christopher Siden	dmu_object_byteswap_t	ot_byteswap;
665fa9e406ahrens	boolean_t		ot_metadata;
666fa9e406ahrens	char			*ot_name;
667fa9e406ahrens} dmu_object_type_info_t;
668fa9e406ahrens
669ad135b5Christopher Sidentypedef struct dmu_object_byteswap_info {
670ad135b5Christopher Siden	arc_byteswap_func_t	*ob_func;
671ad135b5Christopher Siden	char			*ob_name;
672ad135b5Christopher Siden} dmu_object_byteswap_info_t;
673ad135b5Christopher Siden
674fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
675ad135b5Christopher Sidenextern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
676fa9e406ahrens
677fa9e406ahrens/*
678fa9e406ahrens * Get information on a DMU object.
679fa9e406ahrens *
680fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
681fa9e406ahrens *
682fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
683fa9e406ahrens */
684fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
6853e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dnode in hand. */
686fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
6873e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
688fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
6893e30c24Will Andrews/*
6903e30c24Will Andrews * Like dmu_object_info_from_db, but faster still when you only care about
6913e30c24Will Andrews * the size.  This is specifically optimized for zfs_getattr().
6923e30c24Will Andrews */
693fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
694fa9e406ahrens    u_longlong_t *nblk512);
695fa9e406ahrens
696fa9e406ahrenstypedef struct dmu_objset_stats {
697a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
698a2eea2eahrens	uint64_t dds_creation_txg;
6993cb34c6ahrens	uint64_t dds_guid;
700fa9e406ahrens	dmu_objset_type_t dds_type;
701fa9e406ahrens	uint8_t dds_is_snapshot;
70231fd60dahrens	uint8_t dds_inconsistent;
7033cb34c6ahrens	char dds_origin[MAXNAMELEN];
704a2eea2eahrens} dmu_objset_stats_t;
705fa9e406ahrens
706a2eea2eahrens/*
707a2eea2eahrens * Get stats on a dataset.
708a2eea2eahrens */
709a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
710fa9e406ahrens
711a2eea2eahrens/*
712a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
713a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
714a2eea2eahrens */
715a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
716fa9e406ahrens
717a2eea2eahrens/*
718a2eea2eahrens * Get the space usage statistics for statvfs().
719a2eea2eahrens *
720a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
721a2eea2eahrens * availbytes is the amount of space available to this objset, taking
722a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
723a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
724a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
725a2eea2eahrens *
726a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
727a2eea2eahrens * and available.
728a2eea2eahrens */
729a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
730a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
731fa9e406ahrens
732fa9e406ahrens/*
733a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
734a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
735a2eea2eahrens * change, so there is a small probability that it will collide.)
736fa9e406ahrens */
737a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
738fa9e406ahrens
73971eb053Chris Kirby/*
74071eb053Chris Kirby * Get the [cm]time for an objset's snapshot dir
74171eb053Chris Kirby */
74271eb053Chris Kirbytimestruc_t dmu_objset_snap_cmtime(objset_t *os);
74371eb053Chris Kirby
744fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
745fa9e406ahrens
746fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
747fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
748fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
749fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
750fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
751fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
752fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
753edf345eMatthew Ahrensextern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
754edf345eMatthew Ahrensextern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
755fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
756b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
757ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
758ab04eb8timh    int maxlen, boolean_t *conflict);
75987e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
76087e5029ahrens    uint64_t *idp, uint64_t *offp);
7611484342Matthew Ahrens
7629966ca1Matthew Ahrenstypedef int objset_used_cb_t(dmu_object_type_t bonustype,
7639966ca1Matthew Ahrens    void *bonus, uint64_t *userp, uint64_t *groupp);
7641484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
7651484342Matthew Ahrens    objset_used_cb_t *cb);
766f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
767f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
768fa9e406ahrens
769fa9e406ahrens/*
770fa9e406ahrens * Return the txg number for the given assigned transaction.
771fa9e406ahrens */
77287e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
773fa9e406ahrens
774fa9e406ahrens/*
775fa9e406ahrens * Synchronous write.
776c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
777c5c6ffamaybee * provided buffer as a child of the parent zio.
778da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
779c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
780fa9e406ahrens * Note that while the data covered by this function will be on stable
781c5c6ffamaybee * storage when the write completes this new data does not become a
782fa9e406ahrens * permanent part of the file until the associated transaction commits.
783fa9e406ahrens */
784b24ab67Jeff Bonwick
785b24ab67Jeff Bonwick/*
786b24ab67Jeff Bonwick * {zfs,zvol,ztest}_get_done() args
787b24ab67Jeff Bonwick */
788b24ab67Jeff Bonwicktypedef struct zgd {
789b24ab67Jeff Bonwick	struct zilog	*zgd_zilog;
790b24ab67Jeff Bonwick	struct blkptr	*zgd_bp;
791b24ab67Jeff Bonwick	dmu_buf_t	*zgd_db;
792b24ab67Jeff Bonwick	struct rl	*zgd_rl;
793b24ab67Jeff Bonwick	void		*zgd_private;
794b24ab67Jeff Bonwick} zgd_t;
795b24ab67Jeff Bonwick
796b24ab67Jeff Bonwicktypedef void dmu_sync_cb_t(zgd_t *arg, int error);
797b24ab67Jeff Bonwickint dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
798fa9e406ahrens
799fa9e406ahrens/*
800fa9e406ahrens * Find the next hole or data block in file starting at *off
801fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
802fa9e406ahrens */
803fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
804fa9e406ahrens    uint64_t *off);
805fa9e406ahrens
806fa9e406ahrens/*
807fa9e406ahrens * Initial setup and final teardown.
808fa9e406ahrens */
809fa9e406ahrensextern void dmu_init(void);
810fa9e406ahrensextern void dmu_fini(void);
811fa9e406ahrens
812fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
813fa9e406ahrens    uint64_t object, uint64_t offset, int len);
814fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
815fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
816fa9e406ahrens
8173b2aab1Matthew Ahrensint dmu_diff(const char *tosnap_name, const char *fromsnap_name,
8183b2aab1Matthew Ahrens    struct vnode *vp, offset_t *offp);
81999d5e17Tim Haley
820fa9e406ahrens/* CRC64 table */
821fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
822fa9e406ahrensextern uint64_t zfs_crc64_table[256];
823fa9e406ahrens
824e4be62aSaso Kiselkovextern int zfs_mdcomp_disable;
825e4be62aSaso Kiselkov
826fa9e406ahrens#ifdef	__cplusplus
827fa9e406ahrens}
828fa9e406ahrens#endif
829fa9e406ahrens
830fa9e406ahrens#endif	/* _SYS_DMU_H */
831