dmu.h revision e77d42eaa49fe55bfae1e0e0065c6e99affc001b
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.
28e77d42eMatthew Ahrens * Copyright 2014 HybridCluster. All rights reserved.
29ec5cf9dAlexander Stetsenko */
30fa9e406ahrens
3155da60bMark J Musante/* Portions Copyright 2010 Robert Milkowski */
3255da60bMark J Musante
33fa9e406ahrens#ifndef	_SYS_DMU_H
34fa9e406ahrens#define	_SYS_DMU_H
35fa9e406ahrens
36fa9e406ahrens/*
37fa9e406ahrens * This file describes the interface that the DMU provides for its
38fa9e406ahrens * consumers.
39fa9e406ahrens *
40fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
41fa9e406ahrens * dmu_spa.h.
42fa9e406ahrens */
43fa9e406ahrens
44fa9e406ahrens#include <sys/inttypes.h>
45fa9e406ahrens#include <sys/types.h>
46fa9e406ahrens#include <sys/param.h>
47ecd6cf8marks#include <sys/cred.h>
4871eb053Chris Kirby#include <sys/time.h>
494445fffMatthew Ahrens#include <sys/fs/zfs.h>
50fa9e406ahrens
51fa9e406ahrens#ifdef	__cplusplus
52fa9e406ahrensextern "C" {
53fa9e406ahrens#endif
54fa9e406ahrens
55fa9e406ahrensstruct uio;
56c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct xuio;
5744eda4dmaybeestruct page;
58fa9e406ahrensstruct vnode;
59fa9e406ahrensstruct spa;
60fa9e406ahrensstruct zilog;
61fa9e406ahrensstruct zio;
62fa9e406ahrensstruct blkptr;
63fa9e406ahrensstruct zap_cursor;
64fa9e406ahrensstruct dsl_dataset;
65fa9e406ahrensstruct dsl_pool;
66fa9e406ahrensstruct dnode;
67fa9e406ahrensstruct drr_begin;
68fa9e406ahrensstruct drr_end;
697802d7bMatthew Ahrensstruct zbookmark_phys;
7044cd46cbillmstruct spa;
71a2eea2eahrensstruct nvlist;
722fdbea2Aleksandr Guzovskiystruct arc_buf;
73b24ab67Jeff Bonwickstruct zio_prop;
740a586ceMark Shellenbaumstruct sa_handle;
75fa9e406ahrens
76fa9e406ahrenstypedef struct objset objset_t;
77fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
78fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
79fa9e406ahrens
80ad135b5Christopher Sidentypedef enum dmu_object_byteswap {
81ad135b5Christopher Siden	DMU_BSWAP_UINT8,
82ad135b5Christopher Siden	DMU_BSWAP_UINT16,
83ad135b5Christopher Siden	DMU_BSWAP_UINT32,
84ad135b5Christopher Siden	DMU_BSWAP_UINT64,
85ad135b5Christopher Siden	DMU_BSWAP_ZAP,
86ad135b5Christopher Siden	DMU_BSWAP_DNODE,
87ad135b5Christopher Siden	DMU_BSWAP_OBJSET,
88ad135b5Christopher Siden	DMU_BSWAP_ZNODE,
89ad135b5Christopher Siden	DMU_BSWAP_OLDACL,
90ad135b5Christopher Siden	DMU_BSWAP_ACL,
91ad135b5Christopher Siden	/*
92ad135b5Christopher Siden	 * Allocating a new byteswap type number makes the on-disk format
93ad135b5Christopher Siden	 * incompatible with any other format that uses the same number.
94ad135b5Christopher Siden	 *
95ad135b5Christopher Siden	 * Data can usually be structured to work with one of the
96ad135b5Christopher Siden	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
97ad135b5Christopher Siden	 */
98ad135b5Christopher Siden	DMU_BSWAP_NUMFUNCS
99ad135b5Christopher Siden} dmu_object_byteswap_t;
100ad135b5Christopher Siden
101ad135b5Christopher Siden#define	DMU_OT_NEWTYPE 0x80
102ad135b5Christopher Siden#define	DMU_OT_METADATA 0x40
103ad135b5Christopher Siden#define	DMU_OT_BYTESWAP_MASK 0x3f
104ad135b5Christopher Siden
105ad135b5Christopher Siden/*
106ad135b5Christopher Siden * Defines a uint8_t object type. Object types specify if the data
107ad135b5Christopher Siden * in the object is metadata (boolean) and how to byteswap the data
108ad135b5Christopher Siden * (dmu_object_byteswap_t).
109ad135b5Christopher Siden */
110ad135b5Christopher Siden#define	DMU_OT(byteswap, metadata) \
111ad135b5Christopher Siden	(DMU_OT_NEWTYPE | \
112ad135b5Christopher Siden	((metadata) ? DMU_OT_METADATA : 0) | \
113ad135b5Christopher Siden	((byteswap) & DMU_OT_BYTESWAP_MASK))
114ad135b5Christopher Siden
115ad135b5Christopher Siden#define	DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
116ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
117ad135b5Christopher Siden	(ot) < DMU_OT_NUMTYPES)
118ad135b5Christopher Siden
119ad135b5Christopher Siden#define	DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
120ad135b5Christopher Siden	((ot) & DMU_OT_METADATA) : \
121ad135b5Christopher Siden	dmu_ot[(ot)].ot_metadata)
122ad135b5Christopher Siden
1235d7b4d4Matthew Ahrens/*
1245d7b4d4Matthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
1255d7b4d4Matthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
1265d7b4d4Matthew Ahrens * is repurposed for embedded BPs.
1275d7b4d4Matthew Ahrens */
1285d7b4d4Matthew Ahrens#define	DMU_OT_HAS_FILL(ot) \
1295d7b4d4Matthew Ahrens	((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
1305d7b4d4Matthew Ahrens
131ad135b5Christopher Siden#define	DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
132ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) : \
133ad135b5Christopher Siden	dmu_ot[(ot)].ot_byteswap)
134ad135b5Christopher Siden
135fa9e406ahrenstypedef enum dmu_object_type {
136fa9e406ahrens	DMU_OT_NONE,
137fa9e406ahrens	/* general: */
138fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
139fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
140fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
141fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
142cde58dbMatthew Ahrens	DMU_OT_BPOBJ,			/* UINT64 */
143cde58dbMatthew Ahrens	DMU_OT_BPOBJ_HDR,		/* UINT64 */
144fa9e406ahrens	/* spa: */
145fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
146fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
147fa9e406ahrens	/* zil: */
148fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
149fa9e406ahrens	/* dmu: */
150fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
151fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
152fa9e406ahrens	/* dsl: */
1531649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
15487e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
15587e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
156fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
1571649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
158fa9e406ahrens	/* zpl: */
159fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
160da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
161fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
162fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
163fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
164893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
165fa9e406ahrens	/* zvol: */
166fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
167fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
168fa9e406ahrens	/* other; for testing only! */
169fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
170fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
171fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
172ea8dc4beschrock	/* new object types: */
173ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
17406eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
17506eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
176b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
177ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
178da6c28aamw	DMU_OT_ACL,			/* ACL */
179da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
180da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
181da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
182088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
1833f9d6adLin Ling	DMU_OT_SCAN_QUEUE,		/* ZAP */
1841484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1851484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
186842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
187b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
188b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
1890a586ceMark Shellenbaum	DMU_OT_SA,			/* System attr */
1900a586ceMark Shellenbaum	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1910a586ceMark Shellenbaum	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1920a586ceMark Shellenbaum	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
1933f9d6adLin Ling	DMU_OT_SCAN_XLATE,		/* ZAP */
1943f9d6adLin Ling	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
195cde58dbMatthew Ahrens	DMU_OT_DEADLIST,		/* ZAP */
196cde58dbMatthew Ahrens	DMU_OT_DEADLIST_HDR,		/* UINT64 */
197cde58dbMatthew Ahrens	DMU_OT_DSL_CLONES,		/* ZAP */
198cde58dbMatthew Ahrens	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
199ad135b5Christopher Siden	/*
200ad135b5Christopher Siden	 * Do not allocate new object types here. Doing so makes the on-disk
201ad135b5Christopher Siden	 * format incompatible with any other format that uses the same object
202ad135b5Christopher Siden	 * type number.
203ad135b5Christopher Siden	 *
204ad135b5Christopher Siden	 * When creating an object which does not have one of the above types
205ad135b5Christopher Siden	 * use the DMU_OTN_* type with the correct byteswap and metadata
206ad135b5Christopher Siden	 * values.
207ad135b5Christopher Siden	 *
208ad135b5Christopher Siden	 * The DMU_OTN_* types do not have entries in the dmu_ot table,
209ad135b5Christopher Siden	 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
210ad135b5Christopher Siden	 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
211ad135b5Christopher Siden	 * and DMU_OTN_* types).
212ad135b5Christopher Siden	 */
213ad135b5Christopher Siden	DMU_OT_NUMTYPES,
214ad135b5Christopher Siden
215ad135b5Christopher Siden	/*
216ad135b5Christopher Siden	 * Names for valid types declared with DMU_OT().
217ad135b5Christopher Siden	 */
218ad135b5Christopher Siden	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
219ad135b5Christopher Siden	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
220ad135b5Christopher Siden	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
221ad135b5Christopher Siden	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
222ad135b5Christopher Siden	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
223ad135b5Christopher Siden	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
224ad135b5Christopher Siden	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
225ad135b5Christopher Siden	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
226ad135b5Christopher Siden	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
227ad135b5Christopher Siden	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
228fa9e406ahrens} dmu_object_type_t;
229fa9e406ahrens
2303b2aab1Matthew Ahrenstypedef enum txg_how {
2313b2aab1Matthew Ahrens	TXG_WAIT = 1,
2323b2aab1Matthew Ahrens	TXG_NOWAIT,
23369962b5Matthew Ahrens	TXG_WAITED,
2343b2aab1Matthew Ahrens} txg_how_t;
2353b2aab1Matthew Ahrens
236fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
237fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
238fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
239fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
240fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
241da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
242fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
243fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
244fa9e406ahrens
2450b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
2460b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
247fa9e406ahrens
248fa9e406ahrens/*
249fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
250fa9e406ahrens * operation, including metadata.
251fa9e406ahrens */
252fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
253cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
254fa9e406ahrens
2551484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
2561484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
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,
343e77d42eMatthew Ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
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);
5714bb7380Matthew Ahrensvoid dmu_tx_mark_netfree(dmu_tx_t *tx);
572fa9e406ahrens
573fa9e406ahrens/*
574d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
575d20e665Ricardo M. Correia *
576d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
577d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
578d20e665Ricardo M. Correia * freeing it.
579d20e665Ricardo M. Correia *
580d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
581d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
582d20e665Ricardo M. Correia *
583d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
584d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
585d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
586d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
587d20e665Ricardo M. Correia */
588d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
589d20e665Ricardo M. Correia
590d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
591d20e665Ricardo M. Correia    void *dcb_data);
592d20e665Ricardo M. Correia
593d20e665Ricardo M. Correia/*
594fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
595ad135b5Christopher Siden * -1, the range from offset to end-of-file is freed.
596fa9e406ahrens */
597ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
598fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
599cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
600cdb0ab7maybee	uint64_t size);
601713d6c2Matthew Ahrensint dmu_free_long_object(objset_t *os, uint64_t object);
602fa9e406ahrens
603fa9e406ahrens/*
604fa9e406ahrens * Convenience functions.
605fa9e406ahrens *
606fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
607fa9e406ahrens * nonrecoverable I/O error.
608fa9e406ahrens */
6097bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
6107bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
611ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
6127bfdf01Neil Perrin	void *buf, uint32_t flags);
613fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
614fa9e406ahrens	const void *buf, dmu_tx_t *tx);
61582c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
61682c9918Tim Haley	dmu_tx_t *tx);
617feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
618f8554bbMatthew Ahrensint dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
619feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
620feb08c6billm    dmu_tx_t *tx);
62194d1a21Tim Haleyint dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
62294d1a21Tim Haley    dmu_tx_t *tx);
62344eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
62444eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
6252fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
6262fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
6272fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
6282fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
629c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_init(struct xuio *uio, int niov);
630c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_fini(struct xuio *uio);
631c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
632c242f9achunli zhang - Sun Microsystems - Irvine United States    size_t n);
633c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_cnt(struct xuio *uio);
634c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
635c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_clear(struct xuio *uio, int i);
636c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_copied();
637c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_nocopy();
638fa9e406ahrens
639416e0cdekextern int zfs_prefetch_disable;
640416e0cdek
641fa9e406ahrens/*
642fa9e406ahrens * Asynchronously try to read in the data.
643fa9e406ahrens */
644fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
645fa9e406ahrens    uint64_t len);
646fa9e406ahrens
647fa9e406ahrenstypedef struct dmu_object_info {
648b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
649fa9e406ahrens	uint32_t doi_data_block_size;
650fa9e406ahrens	uint32_t doi_metadata_block_size;
651fa9e406ahrens	dmu_object_type_t doi_type;
652fa9e406ahrens	dmu_object_type_t doi_bonus_type;
653b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
654fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
655fa9e406ahrens	uint8_t doi_checksum;
656fa9e406ahrens	uint8_t doi_compress;
657e77d42eMatthew Ahrens	uint8_t doi_nblkptr;
658e77d42eMatthew Ahrens	uint8_t doi_pad[4];
659b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
660b24ab67Jeff Bonwick	uint64_t doi_max_offset;
661b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
662fa9e406ahrens} dmu_object_info_t;
663fa9e406ahrens
664e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
665e45ce72ahrens
666fa9e406ahrenstypedef struct dmu_object_type_info {
667ad135b5Christopher Siden	dmu_object_byteswap_t	ot_byteswap;
668fa9e406ahrens	boolean_t		ot_metadata;
669fa9e406ahrens	char			*ot_name;
670fa9e406ahrens} dmu_object_type_info_t;
671fa9e406ahrens
672ad135b5Christopher Sidentypedef struct dmu_object_byteswap_info {
673ad135b5Christopher Siden	arc_byteswap_func_t	*ob_func;
674ad135b5Christopher Siden	char			*ob_name;
675ad135b5Christopher Siden} dmu_object_byteswap_info_t;
676ad135b5Christopher Siden
677fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
678ad135b5Christopher Sidenextern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
679fa9e406ahrens
680fa9e406ahrens/*
681fa9e406ahrens * Get information on a DMU object.
682fa9e406ahrens *
683fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
684fa9e406ahrens *
685fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
686fa9e406ahrens */
687fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
6883e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dnode in hand. */
689fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
6903e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
691fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
6923e30c24Will Andrews/*
6933e30c24Will Andrews * Like dmu_object_info_from_db, but faster still when you only care about
6943e30c24Will Andrews * the size.  This is specifically optimized for zfs_getattr().
6953e30c24Will Andrews */
696fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
697fa9e406ahrens    u_longlong_t *nblk512);
698fa9e406ahrens
699fa9e406ahrenstypedef struct dmu_objset_stats {
700a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
701a2eea2eahrens	uint64_t dds_creation_txg;
7023cb34c6ahrens	uint64_t dds_guid;
703fa9e406ahrens	dmu_objset_type_t dds_type;
704fa9e406ahrens	uint8_t dds_is_snapshot;
70531fd60dahrens	uint8_t dds_inconsistent;
7063cb34c6ahrens	char dds_origin[MAXNAMELEN];
707a2eea2eahrens} dmu_objset_stats_t;
708fa9e406ahrens
709a2eea2eahrens/*
710a2eea2eahrens * Get stats on a dataset.
711a2eea2eahrens */
712a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
713fa9e406ahrens
714a2eea2eahrens/*
715a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
716a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
717a2eea2eahrens */
718a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
719fa9e406ahrens
720a2eea2eahrens/*
721a2eea2eahrens * Get the space usage statistics for statvfs().
722a2eea2eahrens *
723a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
724a2eea2eahrens * availbytes is the amount of space available to this objset, taking
725a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
726a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
727a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
728a2eea2eahrens *
729a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
730a2eea2eahrens * and available.
731a2eea2eahrens */
732a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
733a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
734fa9e406ahrens
735fa9e406ahrens/*
736a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
737a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
738a2eea2eahrens * change, so there is a small probability that it will collide.)
739fa9e406ahrens */
740a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
741fa9e406ahrens
74271eb053Chris Kirby/*
74371eb053Chris Kirby * Get the [cm]time for an objset's snapshot dir
74471eb053Chris Kirby */
74571eb053Chris Kirbytimestruc_t dmu_objset_snap_cmtime(objset_t *os);
74671eb053Chris Kirby
747fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
748fa9e406ahrens
749fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
750fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
751fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
752fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
753fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
754fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
755fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
756edf345eMatthew Ahrensextern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
757edf345eMatthew Ahrensextern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
758fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
759b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
760ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
761ab04eb8timh    int maxlen, boolean_t *conflict);
76287e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
76387e5029ahrens    uint64_t *idp, uint64_t *offp);
7641484342Matthew Ahrens
7659966ca1Matthew Ahrenstypedef int objset_used_cb_t(dmu_object_type_t bonustype,
7669966ca1Matthew Ahrens    void *bonus, uint64_t *userp, uint64_t *groupp);
7671484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
7681484342Matthew Ahrens    objset_used_cb_t *cb);
769f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
770f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
771fa9e406ahrens
772fa9e406ahrens/*
773fa9e406ahrens * Return the txg number for the given assigned transaction.
774fa9e406ahrens */
77587e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
776fa9e406ahrens
777fa9e406ahrens/*
778fa9e406ahrens * Synchronous write.
779c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
780c5c6ffamaybee * provided buffer as a child of the parent zio.
781da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
782c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
783fa9e406ahrens * Note that while the data covered by this function will be on stable
784c5c6ffamaybee * storage when the write completes this new data does not become a
785fa9e406ahrens * permanent part of the file until the associated transaction commits.
786fa9e406ahrens */
787b24ab67Jeff Bonwick
788b24ab67Jeff Bonwick/*
789b24ab67Jeff Bonwick * {zfs,zvol,ztest}_get_done() args
790b24ab67Jeff Bonwick */
791b24ab67Jeff Bonwicktypedef struct zgd {
792b24ab67Jeff Bonwick	struct zilog	*zgd_zilog;
793b24ab67Jeff Bonwick	struct blkptr	*zgd_bp;
794b24ab67Jeff Bonwick	dmu_buf_t	*zgd_db;
795b24ab67Jeff Bonwick	struct rl	*zgd_rl;
796b24ab67Jeff Bonwick	void		*zgd_private;
797b24ab67Jeff Bonwick} zgd_t;
798b24ab67Jeff Bonwick
799b24ab67Jeff Bonwicktypedef void dmu_sync_cb_t(zgd_t *arg, int error);
800b24ab67Jeff Bonwickint dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
801fa9e406ahrens
802fa9e406ahrens/*
803fa9e406ahrens * Find the next hole or data block in file starting at *off
804fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
805fa9e406ahrens */
806fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
807fa9e406ahrens    uint64_t *off);
808fa9e406ahrens
809fa9e406ahrens/*
810fa9e406ahrens * Initial setup and final teardown.
811fa9e406ahrens */
812fa9e406ahrensextern void dmu_init(void);
813fa9e406ahrensextern void dmu_fini(void);
814fa9e406ahrens
815fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
816fa9e406ahrens    uint64_t object, uint64_t offset, int len);
817fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
818fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
819fa9e406ahrens
8203b2aab1Matthew Ahrensint dmu_diff(const char *tosnap_name, const char *fromsnap_name,
8213b2aab1Matthew Ahrens    struct vnode *vp, offset_t *offp);
82299d5e17Tim Haley
823fa9e406ahrens/* CRC64 table */
824fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
825fa9e406ahrensextern uint64_t zfs_crc64_table[256];
826fa9e406ahrens
827e4be62aSaso Kiselkovextern int zfs_mdcomp_disable;
828e4be62aSaso Kiselkov
829fa9e406ahrens#ifdef	__cplusplus
830fa9e406ahrens}
831fa9e406ahrens#endif
832fa9e406ahrens
833fa9e406ahrens#endif	/* _SYS_DMU_H */
834