dmu.h revision c3d26abc9ee97b4f60233556aadeb57e0bd30bb9
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.
29bc9014eJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
3045818eeMatthew Ahrens * Copyright 2013 Saso Kiselkov. All rights reserved.
31c3d26abMatthew Ahrens * Copyright (c) 2014 Integros [integros.com]
32ec5cf9dAlexander Stetsenko */
33fa9e406ahrens
3455da60bMark J Musante/* Portions Copyright 2010 Robert Milkowski */
3555da60bMark J Musante
36fa9e406ahrens#ifndef	_SYS_DMU_H
37fa9e406ahrens#define	_SYS_DMU_H
38fa9e406ahrens
39fa9e406ahrens/*
40fa9e406ahrens * This file describes the interface that the DMU provides for its
41fa9e406ahrens * consumers.
42fa9e406ahrens *
43fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
44fa9e406ahrens * dmu_spa.h.
45fa9e406ahrens */
46fa9e406ahrens
47bc9014eJustin Gibbs#include <sys/zfs_context.h>
48fa9e406ahrens#include <sys/inttypes.h>
49ecd6cf8marks#include <sys/cred.h>
504445fffMatthew Ahrens#include <sys/fs/zfs.h>
51a2cdcddPaul Dagnelie#include <sys/zio_priority.h>
52fa9e406ahrens
53fa9e406ahrens#ifdef	__cplusplus
54fa9e406ahrensextern "C" {
55fa9e406ahrens#endif
56fa9e406ahrens
57fa9e406ahrensstruct uio;
58c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct xuio;
5944eda4dmaybeestruct page;
60fa9e406ahrensstruct vnode;
61fa9e406ahrensstruct spa;
62fa9e406ahrensstruct zilog;
63fa9e406ahrensstruct zio;
64fa9e406ahrensstruct blkptr;
65fa9e406ahrensstruct zap_cursor;
66fa9e406ahrensstruct dsl_dataset;
67fa9e406ahrensstruct dsl_pool;
68fa9e406ahrensstruct dnode;
69fa9e406ahrensstruct drr_begin;
70fa9e406ahrensstruct drr_end;
717802d7bMatthew Ahrensstruct zbookmark_phys;
7244cd46cbillmstruct spa;
73a2eea2eahrensstruct nvlist;
742fdbea2Aleksandr Guzovskiystruct arc_buf;
75b24ab67Jeff Bonwickstruct zio_prop;
760a586ceMark Shellenbaumstruct sa_handle;
77fa9e406ahrens
78fa9e406ahrenstypedef struct objset objset_t;
79fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
80fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
81fa9e406ahrens
82ad135b5Christopher Sidentypedef enum dmu_object_byteswap {
83ad135b5Christopher Siden	DMU_BSWAP_UINT8,
84ad135b5Christopher Siden	DMU_BSWAP_UINT16,
85ad135b5Christopher Siden	DMU_BSWAP_UINT32,
86ad135b5Christopher Siden	DMU_BSWAP_UINT64,
87ad135b5Christopher Siden	DMU_BSWAP_ZAP,
88ad135b5Christopher Siden	DMU_BSWAP_DNODE,
89ad135b5Christopher Siden	DMU_BSWAP_OBJSET,
90ad135b5Christopher Siden	DMU_BSWAP_ZNODE,
91ad135b5Christopher Siden	DMU_BSWAP_OLDACL,
92ad135b5Christopher Siden	DMU_BSWAP_ACL,
93ad135b5Christopher Siden	/*
94ad135b5Christopher Siden	 * Allocating a new byteswap type number makes the on-disk format
95ad135b5Christopher Siden	 * incompatible with any other format that uses the same number.
96ad135b5Christopher Siden	 *
97ad135b5Christopher Siden	 * Data can usually be structured to work with one of the
98ad135b5Christopher Siden	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
99ad135b5Christopher Siden	 */
100ad135b5Christopher Siden	DMU_BSWAP_NUMFUNCS
101ad135b5Christopher Siden} dmu_object_byteswap_t;
102ad135b5Christopher Siden
103ad135b5Christopher Siden#define	DMU_OT_NEWTYPE 0x80
104ad135b5Christopher Siden#define	DMU_OT_METADATA 0x40
105ad135b5Christopher Siden#define	DMU_OT_BYTESWAP_MASK 0x3f
106ad135b5Christopher Siden
107ad135b5Christopher Siden/*
108ad135b5Christopher Siden * Defines a uint8_t object type. Object types specify if the data
109ad135b5Christopher Siden * in the object is metadata (boolean) and how to byteswap the data
110ad135b5Christopher Siden * (dmu_object_byteswap_t).
111ad135b5Christopher Siden */
112ad135b5Christopher Siden#define	DMU_OT(byteswap, metadata) \
113ad135b5Christopher Siden	(DMU_OT_NEWTYPE | \
114ad135b5Christopher Siden	((metadata) ? DMU_OT_METADATA : 0) | \
115ad135b5Christopher Siden	((byteswap) & DMU_OT_BYTESWAP_MASK))
116ad135b5Christopher Siden
117ad135b5Christopher Siden#define	DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
118ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
119ad135b5Christopher Siden	(ot) < DMU_OT_NUMTYPES)
120ad135b5Christopher Siden
121ad135b5Christopher Siden#define	DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
122ad135b5Christopher Siden	((ot) & DMU_OT_METADATA) : \
123ad135b5Christopher Siden	dmu_ot[(ot)].ot_metadata)
124ad135b5Christopher Siden
1255d7b4d4Matthew Ahrens/*
1265d7b4d4Matthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
1275d7b4d4Matthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
1285d7b4d4Matthew Ahrens * is repurposed for embedded BPs.
1295d7b4d4Matthew Ahrens */
1305d7b4d4Matthew Ahrens#define	DMU_OT_HAS_FILL(ot) \
1315d7b4d4Matthew Ahrens	((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
1325d7b4d4Matthew Ahrens
133ad135b5Christopher Siden#define	DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
134ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) : \
135ad135b5Christopher Siden	dmu_ot[(ot)].ot_byteswap)
136ad135b5Christopher Siden
137fa9e406ahrenstypedef enum dmu_object_type {
138fa9e406ahrens	DMU_OT_NONE,
139fa9e406ahrens	/* general: */
140fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
141fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
142fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
143fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
144cde58dbMatthew Ahrens	DMU_OT_BPOBJ,			/* UINT64 */
145cde58dbMatthew Ahrens	DMU_OT_BPOBJ_HDR,		/* UINT64 */
146fa9e406ahrens	/* spa: */
147fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
148fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
149fa9e406ahrens	/* zil: */
150fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
151fa9e406ahrens	/* dmu: */
152fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
153fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
154fa9e406ahrens	/* dsl: */
1551649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
15687e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
15787e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
158fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
1591649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
160fa9e406ahrens	/* zpl: */
161fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
162da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
163fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
164fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
165fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
166893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
167fa9e406ahrens	/* zvol: */
168fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
169fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
170fa9e406ahrens	/* other; for testing only! */
171fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
172fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
173fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
174ea8dc4beschrock	/* new object types: */
175ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
17606eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
17706eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
178b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
179ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
180da6c28aamw	DMU_OT_ACL,			/* ACL */
181da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
182da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
183da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
184088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
1853f9d6adLin Ling	DMU_OT_SCAN_QUEUE,		/* ZAP */
1861484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1871484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
188842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
189b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
190b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
1910a586ceMark Shellenbaum	DMU_OT_SA,			/* System attr */
1920a586ceMark Shellenbaum	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1930a586ceMark Shellenbaum	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1940a586ceMark Shellenbaum	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
1953f9d6adLin Ling	DMU_OT_SCAN_XLATE,		/* ZAP */
1963f9d6adLin Ling	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
197cde58dbMatthew Ahrens	DMU_OT_DEADLIST,		/* ZAP */
198cde58dbMatthew Ahrens	DMU_OT_DEADLIST_HDR,		/* UINT64 */
199cde58dbMatthew Ahrens	DMU_OT_DSL_CLONES,		/* ZAP */
200cde58dbMatthew Ahrens	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
201ad135b5Christopher Siden	/*
202ad135b5Christopher Siden	 * Do not allocate new object types here. Doing so makes the on-disk
203ad135b5Christopher Siden	 * format incompatible with any other format that uses the same object
204ad135b5Christopher Siden	 * type number.
205ad135b5Christopher Siden	 *
206ad135b5Christopher Siden	 * When creating an object which does not have one of the above types
207ad135b5Christopher Siden	 * use the DMU_OTN_* type with the correct byteswap and metadata
208ad135b5Christopher Siden	 * values.
209ad135b5Christopher Siden	 *
210ad135b5Christopher Siden	 * The DMU_OTN_* types do not have entries in the dmu_ot table,
211ad135b5Christopher Siden	 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
212ad135b5Christopher Siden	 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
213ad135b5Christopher Siden	 * and DMU_OTN_* types).
214ad135b5Christopher Siden	 */
215ad135b5Christopher Siden	DMU_OT_NUMTYPES,
216ad135b5Christopher Siden
217ad135b5Christopher Siden	/*
218ad135b5Christopher Siden	 * Names for valid types declared with DMU_OT().
219ad135b5Christopher Siden	 */
220ad135b5Christopher Siden	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
221ad135b5Christopher Siden	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
222ad135b5Christopher Siden	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
223ad135b5Christopher Siden	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
224ad135b5Christopher Siden	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
225ad135b5Christopher Siden	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
226ad135b5Christopher Siden	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
227ad135b5Christopher Siden	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
228ad135b5Christopher Siden	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
229ad135b5Christopher Siden	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
230fa9e406ahrens} dmu_object_type_t;
231fa9e406ahrens
2323b2aab1Matthew Ahrenstypedef enum txg_how {
2333b2aab1Matthew Ahrens	TXG_WAIT = 1,
2343b2aab1Matthew Ahrens	TXG_NOWAIT,
23569962b5Matthew Ahrens	TXG_WAITED,
2363b2aab1Matthew Ahrens} txg_how_t;
2373b2aab1Matthew Ahrens
238fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
239fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
240fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
241fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
242fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
243da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
244fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
245fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
246fa9e406ahrens
2470b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
2480b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
24912380e1Arne Jansen#define	DS_FIND_SERIALIZE	(1<<2)
250fa9e406ahrens
251fa9e406ahrens/*
252fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
253fa9e406ahrens * operation, including metadata.
254fa9e406ahrens */
255b515258Matthew Ahrens#define	DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */
256cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
257fa9e406ahrens
2581484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
2591484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
2601484342Matthew Ahrens
261fa9e406ahrens/*
2620a586ceMark Shellenbaum * artificial blkids for bonus buffer and spill blocks
2630a586ceMark Shellenbaum */
2640a586ceMark Shellenbaum#define	DMU_BONUS_BLKID		(-1ULL)
2650a586ceMark Shellenbaum#define	DMU_SPILL_BLKID		(-2ULL)
2660a586ceMark Shellenbaum/*
267fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
268fa9e406ahrens */
269503ad85Matthew Ahrensint dmu_objset_hold(const char *name, void *tag, objset_t **osp);
270503ad85Matthew Ahrensint dmu_objset_own(const char *name, dmu_objset_type_t type,
271503ad85Matthew Ahrens    boolean_t readonly, void *tag, objset_t **osp);
272503ad85Matthew Ahrensvoid dmu_objset_rele(objset_t *os, void *tag);
273503ad85Matthew Ahrensvoid dmu_objset_disown(objset_t *os, void *tag);
274503ad85Matthew Ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
275503ad85Matthew Ahrens
2763b2aab1Matthew Ahrensvoid dmu_objset_evict_dbufs(objset_t *os);
277ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
278ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
2793b2aab1Matthew Ahrensint dmu_objset_clone(const char *name, const char *origin);
2803b2aab1Matthew Ahrensint dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
2814445fffMatthew Ahrens    struct nvlist *errlist);
2824445fffMatthew Ahrensint dmu_objset_snapshot_one(const char *fsname, const char *snapname);
2834445fffMatthew Ahrensint dmu_objset_snapshot_tmp(const char *, const char *, int);
284fd13687Matthew Ahrensint dmu_objset_find(char *name, int func(const char *, void *), void *arg,
285fa9e406ahrens    int flags);
286fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
2873b2aab1Matthew Ahrensint dsl_dataset_rename_snapshot(const char *fsname,
2883b2aab1Matthew Ahrens    const char *oldsnapname, const char *newsnapname, boolean_t recursive);
289fa9e406ahrens
290fa9e406ahrenstypedef struct dmu_buf {
291fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
292fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
293fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
294fa9e406ahrens	void *db_data;			/* data in buffer */
295fa9e406ahrens} dmu_buf_t;
296fa9e406ahrens
297fa9e406ahrens/*
29899653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
29999653d4eschrock */
300fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
301fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
302ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_WRITE	"features_for_write"
303ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_READ	"features_for_read"
304ad135b5Christopher Siden#define	DMU_POOL_FEATURE_DESCRIPTIONS	"feature_descriptions"
30543466aaMax Grossman#define	DMU_POOL_FEATURE_ENABLED_TXG	"feature_enabled_txg"
306fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
307cde58dbMatthew Ahrens#define	DMU_POOL_SYNC_BPOBJ		"sync_bplist"
308ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
309ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
31099653d4eschrock#define	DMU_POOL_SPARES			"spares"
31199653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
31206eeb2aek#define	DMU_POOL_HISTORY		"history"
313b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
314fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
315ca45db4Chris Kirby#define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
316b24ab67Jeff Bonwick#define	DMU_POOL_DDT			"DDT-%s-%s-%s"
317b24ab67Jeff Bonwick#define	DMU_POOL_DDT_STATS		"DDT-statistics"
3183f9d6adLin Ling#define	DMU_POOL_CREATION_VERSION	"creation_version"
3193f9d6adLin Ling#define	DMU_POOL_SCAN			"scan"
320cde58dbMatthew Ahrens#define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
321ad135b5Christopher Siden#define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
322f174573Matthew Ahrens#define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"
32345818eeMatthew Ahrens#define	DMU_POOL_CHECKSUM_SALT		"org.illumos:checksum_salt"
324088f389ahrens
325fa9e406ahrens/*
326fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
327fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
328fa9e406ahrens *
329fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
330fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
331fa9e406ahrens * newly allocated object in this transaction).
332fa9e406ahrens *
333fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
334fa9e406ahrens *
335fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
336fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
337fa9e406ahrens *
338fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
339fa9e406ahrens */
340fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
341fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
342fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
343fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
344fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
345e77d42eMatthew Ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
346fa9e406ahrens
347fa9e406ahrens/*
348fa9e406ahrens * Free an object from this objset.
349fa9e406ahrens *
350fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
351fa9e406ahrens * dmu_free(object, 0, -1, tx)).
352fa9e406ahrens *
353fa9e406ahrens * The object need not be held in the transaction.
354fa9e406ahrens *
355fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
356fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
357fa9e406ahrens * free it; it fails and returns EBUSY.
358fa9e406ahrens *
359fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
360fa9e406ahrens *
361fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
362fa9e406ahrens */
363fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
364fa9e406ahrens
365fa9e406ahrens/*
366fa9e406ahrens * Find the next allocated or free object.
367fa9e406ahrens *
368fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
3696754306ahrens * object which is allocated.  Ignore objects which have not been
3706754306ahrens * modified since txg.
371fa9e406ahrens *
372fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
373fa9e406ahrens *
374fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
375fa9e406ahrens */
3766754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
3776754306ahrens    boolean_t hole, uint64_t txg);
378fa9e406ahrens
379fa9e406ahrens/*
380fa9e406ahrens * Set the data blocksize for an object.
381fa9e406ahrens *
382fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
383fa9e406ahrens * the first block is allocated already, the new size must be greater
384fa9e406ahrens * than the current block size.  If these conditions are not met,
385fa9e406ahrens * ENOTSUP will be returned.
386fa9e406ahrens *
387fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
388fa9e406ahrens * contents, or ENOTSUP as described above.
389fa9e406ahrens */
390fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
391fa9e406ahrens    int ibs, dmu_tx_t *tx);
392fa9e406ahrens
393fa9e406ahrens/*
394fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
395fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
396fa9e406ahrens */
397fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
398fa9e406ahrens    dmu_tx_t *tx);
399fa9e406ahrens
400fa9e406ahrens/*
401fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
402fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
403fa9e406ahrens */
404fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
405fa9e406ahrens    dmu_tx_t *tx);
406fa9e406ahrens
4075d7b4d4Matthew Ahrensvoid
4085d7b4d4Matthew Ahrensdmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
4095d7b4d4Matthew Ahrens    void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
4105d7b4d4Matthew Ahrens    int compressed_size, int byteorder, dmu_tx_t *tx);
4115d7b4d4Matthew Ahrens
412fa9e406ahrens/*
413b24ab67Jeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc.
41444cd46cbillm */
415b24ab67Jeff Bonwick#define	WP_NOFILL	0x1
416b24ab67Jeff Bonwick#define	WP_DMU_SYNC	0x2
4170a586ceMark Shellenbaum#define	WP_SPILL	0x4
418b24ab67Jeff Bonwick
419b24ab67Jeff Bonwickvoid dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
420b24ab67Jeff Bonwick    struct zio_prop *zp);
42144cd46cbillm/*
422fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
423fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
424fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
425fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
426fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
427fa9e406ahrens * object must be held in an assigned transaction before calling
428fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
429fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
4303e30c24Will Andrews *
4313e30c24Will Andrews * Returns ENOENT, EIO, or 0.
432fa9e406ahrens */
433ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
434fa9e406ahrensint dmu_bonus_max(void);
4351934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
4360a586ceMark Shellenbaumint dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
437744947dTom Ericksondmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
4380a586ceMark Shellenbaumint dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
4390a586ceMark Shellenbaum
4400a586ceMark Shellenbaum/*
4410a586ceMark Shellenbaum * Special spill buffer support used by "SA" framework
4420a586ceMark Shellenbaum */
4430a586ceMark Shellenbaum
4440a586ceMark Shellenbaumint dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
4450a586ceMark Shellenbaumint dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
4460a586ceMark Shellenbaum    void *tag, dmu_buf_t **dbp);
4470a586ceMark Shellenbaumint dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
448fa9e406ahrens
449fa9e406ahrens/*
450fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
451fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
452fa9e406ahrens * that it will remain in memory.  You must release the hold with
453fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
454fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
455fa9e406ahrens *
456fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
457fa9e406ahrens * on the returned buffer before reading or writing the buffer's
458fa9e406ahrens * db_data.  The comments for those routines describe what particular
459fa9e406ahrens * operations are valid after calling them.
460fa9e406ahrens *
461fa9e406ahrens * The object number must be a valid, allocated object number.
462fa9e406ahrens */
463ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
46447cb52dJeff Bonwick    void *tag, dmu_buf_t **, int flags);
465e57a022Justin T. Gibbs
466e57a022Justin T. Gibbs/*
467e57a022Justin T. Gibbs * Add a reference to a dmu buffer that has already been held via
468e57a022Justin T. Gibbs * dmu_buf_hold() in the current context.
469e57a022Justin T. Gibbs */
470fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
471e57a022Justin T. Gibbs
472e57a022Justin T. Gibbs/*
473e57a022Justin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state,
474e57a022Justin T. Gibbs * using a pointer that may have been invalidated by eviction processing.
475e57a022Justin T. Gibbs * The request will succeed if the passed in dbuf still represents the
476e57a022Justin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least
477e57a022Justin T. Gibbs * one hold by a user other than the syncer.
478e57a022Justin T. Gibbs */
479e57a022Justin T. Gibbsboolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object,
480e57a022Justin T. Gibbs    uint64_t blkid, void *tag);
481e57a022Justin T. Gibbs
482ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
483fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
484fa9e406ahrens
485fa9e406ahrens/*
486fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
487fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
488fa9e406ahrens * returned (in *dbpp).
489fa9e406ahrens *
490fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
491fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
492fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
493fa9e406ahrens * individually with dmu_buf_rele.
494fa9e406ahrens */
49513506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
496cf6106cMatthew Ahrens    uint64_t length, boolean_t read, void *tag,
497cf6106cMatthew Ahrens    int *numbufsp, dmu_buf_t ***dbpp);
498ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
499fa9e406ahrens
500bc9014eJustin Gibbstypedef void dmu_buf_evict_func_t(void *user_ptr);
501bc9014eJustin Gibbs
502fa9e406ahrens/*
503bc9014eJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the
504bc9014eJustin Gibbs * duration of its lifetime.  This allows the user of a dbuf (client)
505bc9014eJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a
506bc9014eJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
507bc9014eJustin Gibbs * when that dbuf has been evicted.  Clients typically respond to the
508bc9014eJustin Gibbs * eviction notification by freeing their private data, thus ensuring
509bc9014eJustin Gibbs * the same lifetime for both dbuf and private data.
510fa9e406ahrens *
511bc9014eJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the
512bc9014eJustin Gibbs * client's responsibility.  All current consumers of the API with private
513bc9014eJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for
514bc9014eJustin Gibbs * their private data.  This allows conversions between the two types
515bc9014eJustin Gibbs * with a simple cast.  Since the DMU buf user API never needs access
516bc9014eJustin Gibbs * to the private data, other strategies can be employed if necessary
517bc9014eJustin Gibbs * or convenient for the client (e.g. using container_of() to do the
518bc9014eJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as
519bc9014eJustin Gibbs * its first member).
520fa9e406ahrens *
521bc9014eJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any
522bc9014eJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available.
523bc9014eJustin Gibbs * For this reason, users must assume the dbuf has already been freed
524bc9014eJustin Gibbs * and not reference the dbuf from the callback context.
525fa9e406ahrens *
526bc9014eJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf
527bc9014eJustin Gibbs * is only referenced by dirty records (dirties == holds).  Otherwise the
528bc9014eJustin Gibbs * notification occurs after eviction processing for the dbuf begins.
529fa9e406ahrens */
530bc9014eJustin Gibbstypedef struct dmu_buf_user {
531bc9014eJustin Gibbs	/*
532bc9014eJustin Gibbs	 * Asynchronous user eviction callback state.
533bc9014eJustin Gibbs	 */
534bc9014eJustin Gibbs	taskq_ent_t	dbu_tqent;
535bc9014eJustin Gibbs
536bc9014eJustin Gibbs	/* This instance's eviction function pointer. */
537bc9014eJustin Gibbs	dmu_buf_evict_func_t *dbu_evict_func;
538bc9014eJustin Gibbs#ifdef ZFS_DEBUG
539bc9014eJustin Gibbs	/*
540bc9014eJustin Gibbs	 * Pointer to user's dbuf pointer.  NULL for clients that do
541bc9014eJustin Gibbs	 * not associate a dbuf with their user data.
542bc9014eJustin Gibbs	 *
543bc9014eJustin Gibbs	 * The dbuf pointer is cleared upon eviction so as to catch
544bc9014eJustin Gibbs	 * use-after-evict bugs in clients.
545bc9014eJustin Gibbs	 */
546bc9014eJustin Gibbs	dmu_buf_t **dbu_clear_on_evict_dbufp;
547bc9014eJustin Gibbs#endif
548bc9014eJustin Gibbs} dmu_buf_user_t;
549bc9014eJustin Gibbs
550fa9e406ahrens/*
551bc9014eJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function
552bc9014eJustin Gibbs * evict_func, to be called when the user is evicted.
553bc9014eJustin Gibbs *
554bc9014eJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t.
555bc9014eJustin Gibbs *       To allow enforcement of this, dbu must already be zeroed on entry.
556fa9e406ahrens */
557bc9014eJustin Gibbs#ifdef __lint
558bc9014eJustin Gibbs/* Very ugly, but it beats issuing suppression directives in many Makefiles. */
559bc9014eJustin Gibbsextern void
560bc9014eJustin Gibbsdmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func,
561bc9014eJustin Gibbs    dmu_buf_t **clear_on_evict_dbufp);
562bc9014eJustin Gibbs#else /* __lint */
563bc9014eJustin Gibbsinline void
564bc9014eJustin Gibbsdmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func,
565bc9014eJustin Gibbs    dmu_buf_t **clear_on_evict_dbufp)
566bc9014eJustin Gibbs{
567bc9014eJustin Gibbs	ASSERT(dbu->dbu_evict_func == NULL);
568bc9014eJustin Gibbs	ASSERT(evict_func != NULL);
569bc9014eJustin Gibbs	dbu->dbu_evict_func = evict_func;
570bc9014eJustin Gibbs#ifdef ZFS_DEBUG
571bc9014eJustin Gibbs	dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
572bc9014eJustin Gibbs#endif
573bc9014eJustin Gibbs}
574bc9014eJustin Gibbs#endif /* __lint */
575fa9e406ahrens
576fa9e406ahrens/*
577bc9014eJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's
578bc9014eJustin Gibbs * data is cleared or its reference count goes to zero) eviction processing.
579bc9014eJustin Gibbs *
580bc9014eJustin Gibbs * Returns NULL on success, or the existing user if another user currently
581bc9014eJustin Gibbs * owns the buffer.
582bc9014eJustin Gibbs */
583bc9014eJustin Gibbsvoid *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user);
584bc9014eJustin Gibbs
585bc9014eJustin Gibbs/*
586bc9014eJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and
587bc9014eJustin Gibbs * reference counts are equal) eviction processing.
588bc9014eJustin Gibbs *
589bc9014eJustin Gibbs * Returns NULL on success, or the existing user if another user currently
590bc9014eJustin Gibbs * owns the buffer.
591bc9014eJustin Gibbs */
592bc9014eJustin Gibbsvoid *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user);
593bc9014eJustin Gibbs
594bc9014eJustin Gibbs/*
595bc9014eJustin Gibbs * Replace the current user of a dbuf.
596bc9014eJustin Gibbs *
597bc9014eJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with
598bc9014eJustin Gibbs * "new_user" and returns the user data pointer that was replaced.
599bc9014eJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer.
600bc9014eJustin Gibbs */
601bc9014eJustin Gibbsvoid *dmu_buf_replace_user(dmu_buf_t *db,
602bc9014eJustin Gibbs    dmu_buf_user_t *old_user, dmu_buf_user_t *new_user);
603bc9014eJustin Gibbs
604bc9014eJustin Gibbs/*
605bc9014eJustin Gibbs * Remove the specified user data for a DMU buffer.
606bc9014eJustin Gibbs *
607bc9014eJustin Gibbs * Returns the user that was removed on success, or the current user if
608bc9014eJustin Gibbs * another user currently owns the buffer.
609bc9014eJustin Gibbs */
610bc9014eJustin Gibbsvoid *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user);
611bc9014eJustin Gibbs
612bc9014eJustin Gibbs/*
613bc9014eJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf.
614fa9e406ahrens */
615fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
616fa9e406ahrens
617bc9014eJustin Gibbs/* Block until any in-progress dmu buf user evictions complete. */
618bc9014eJustin Gibbsvoid dmu_buf_user_evict_wait(void);
619bc9014eJustin Gibbs
620fa9e406ahrens/*
62180901aeGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set.
62280901aeGeorge Wilson */
62380901aeGeorge Wilsonstruct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
62480901aeGeorge Wilson
62580901aeGeorge Wilson/*
626fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
627fa9e406ahrens *
628fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
629fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
630fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
631fa9e406ahrens */
632fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
633fa9e406ahrens
634fa9e406ahrens/*
6353d69262Sanjeev Bagewadi * Tells if the given dbuf is freeable.
6363d69262Sanjeev Bagewadi */
6373d69262Sanjeev Bagewadiboolean_t dmu_buf_freeable(dmu_buf_t *);
6383d69262Sanjeev Bagewadi
6393d69262Sanjeev Bagewadi/*
640fa9e406ahrens * You must create a transaction, then hold the objects which you will
641fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
642fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
643fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
644fa9e406ahrens * part of this transaction.  You can't modify buffers before the
645fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
646fa9e406ahrens * belong to objects which this transaction holds; you can't hold
647fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
648fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
649fa9e406ahrens * don't have to.
650fa9e406ahrens *
651fa9e406ahrens * You can abort the transaction before it has been assigned.
652fa9e406ahrens *
653fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
654fa9e406ahrens * regardless of transaction state.
655fa9e406ahrens */
656fa9e406ahrens
657fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
658fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
659fa9e406ahrens
660fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
661fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
662fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
663fa9e406ahrens    uint64_t len);
6641484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
665fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
6660a586ceMark Shellenbaumvoid dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
6670a586ceMark Shellenbaumvoid dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
6680a586ceMark Shellenbaumvoid dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
669fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
6703b2aab1Matthew Ahrensint dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
6718a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
672fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
6734bb7380Matthew Ahrensvoid dmu_tx_mark_netfree(dmu_tx_t *tx);
674fa9e406ahrens
675fa9e406ahrens/*
676d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
677d20e665Ricardo M. Correia *
678d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
679d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
680d20e665Ricardo M. Correia * freeing it.
681d20e665Ricardo M. Correia *
682d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
683d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
684d20e665Ricardo M. Correia *
685d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
686d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
687d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
688d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
689d20e665Ricardo M. Correia */
690d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
691d20e665Ricardo M. Correia
692d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
693d20e665Ricardo M. Correia    void *dcb_data);
694d20e665Ricardo M. Correia
695d20e665Ricardo M. Correia/*
696fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
697ad135b5Christopher Siden * -1, the range from offset to end-of-file is freed.
698fa9e406ahrens */
699ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
700fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
701cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
702cdb0ab7maybee	uint64_t size);
703713d6c2Matthew Ahrensint dmu_free_long_object(objset_t *os, uint64_t object);
704fa9e406ahrens
705fa9e406ahrens/*
706fa9e406ahrens * Convenience functions.
707fa9e406ahrens *
708fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
709fa9e406ahrens * nonrecoverable I/O error.
710fa9e406ahrens */
7117bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
7127bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
713ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
7147bfdf01Neil Perrin	void *buf, uint32_t flags);
715fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
716fa9e406ahrens	const void *buf, dmu_tx_t *tx);
71782c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
71882c9918Tim Haley	dmu_tx_t *tx);
719feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
720f8554bbMatthew Ahrensint dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
721feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
722feb08c6billm    dmu_tx_t *tx);
72394d1a21Tim Haleyint dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
72494d1a21Tim Haley    dmu_tx_t *tx);
72544eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
72644eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
7272fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
7282fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
7292fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
7302fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
731c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_init(struct xuio *uio, int niov);
732c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_fini(struct xuio *uio);
733c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
734c242f9achunli zhang - Sun Microsystems - Irvine United States    size_t n);
735c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_cnt(struct xuio *uio);
736c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
737c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_clear(struct xuio *uio, int i);
738c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_copied();
739c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_nocopy();
740fa9e406ahrens
741cf6106cMatthew Ahrensextern boolean_t zfs_prefetch_disable;
742b515258Matthew Ahrensextern int zfs_max_recordsize;
743416e0cdek
744fa9e406ahrens/*
745fa9e406ahrens * Asynchronously try to read in the data.
746fa9e406ahrens */
747a2cdcddPaul Dagnelievoid dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
748a2cdcddPaul Dagnelie    uint64_t len, enum zio_priority pri);
749fa9e406ahrens
750fa9e406ahrenstypedef struct dmu_object_info {
751b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
752fa9e406ahrens	uint32_t doi_data_block_size;
753fa9e406ahrens	uint32_t doi_metadata_block_size;
754fa9e406ahrens	dmu_object_type_t doi_type;
755fa9e406ahrens	dmu_object_type_t doi_bonus_type;
756b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
757fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
758fa9e406ahrens	uint8_t doi_checksum;
759fa9e406ahrens	uint8_t doi_compress;
760e77d42eMatthew Ahrens	uint8_t doi_nblkptr;
761e77d42eMatthew Ahrens	uint8_t doi_pad[4];
762b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
763b24ab67Jeff Bonwick	uint64_t doi_max_offset;
764b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
765fa9e406ahrens} dmu_object_info_t;
766fa9e406ahrens
767e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
768e45ce72ahrens
769fa9e406ahrenstypedef struct dmu_object_type_info {
770ad135b5Christopher Siden	dmu_object_byteswap_t	ot_byteswap;
771fa9e406ahrens	boolean_t		ot_metadata;
772fa9e406ahrens	char			*ot_name;
773fa9e406ahrens} dmu_object_type_info_t;
774fa9e406ahrens
775ad135b5Christopher Sidentypedef struct dmu_object_byteswap_info {
776ad135b5Christopher Siden	arc_byteswap_func_t	*ob_func;
777ad135b5Christopher Siden	char			*ob_name;
778ad135b5Christopher Siden} dmu_object_byteswap_info_t;
779ad135b5Christopher Siden
780fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
781ad135b5Christopher Sidenextern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
782fa9e406ahrens
783fa9e406ahrens/*
784fa9e406ahrens * Get information on a DMU object.
785fa9e406ahrens *
786fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
787fa9e406ahrens *
788fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
789fa9e406ahrens */
790fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
7913e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dnode in hand. */
792fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
7933e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
794fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
7953e30c24Will Andrews/*
7963e30c24Will Andrews * Like dmu_object_info_from_db, but faster still when you only care about
7973e30c24Will Andrews * the size.  This is specifically optimized for zfs_getattr().
7983e30c24Will Andrews */
799fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
800fa9e406ahrens    u_longlong_t *nblk512);
801fa9e406ahrens
802fa9e406ahrenstypedef struct dmu_objset_stats {
803a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
804a2eea2eahrens	uint64_t dds_creation_txg;
8053cb34c6ahrens	uint64_t dds_guid;
806fa9e406ahrens	dmu_objset_type_t dds_type;
807fa9e406ahrens	uint8_t dds_is_snapshot;
80831fd60dahrens	uint8_t dds_inconsistent;
8093cb34c6ahrens	char dds_origin[MAXNAMELEN];
810a2eea2eahrens} dmu_objset_stats_t;
811fa9e406ahrens
812a2eea2eahrens/*
813a2eea2eahrens * Get stats on a dataset.
814a2eea2eahrens */
815a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
816fa9e406ahrens
817a2eea2eahrens/*
818a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
819a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
820a2eea2eahrens */
821a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
822fa9e406ahrens
823a2eea2eahrens/*
824a2eea2eahrens * Get the space usage statistics for statvfs().
825a2eea2eahrens *
826a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
827a2eea2eahrens * availbytes is the amount of space available to this objset, taking
828a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
829a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
830a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
831a2eea2eahrens *
832a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
833a2eea2eahrens * and available.
834a2eea2eahrens */
835a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
836a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
837fa9e406ahrens
838fa9e406ahrens/*
839a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
840a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
841a2eea2eahrens * change, so there is a small probability that it will collide.)
842fa9e406ahrens */
843a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
844fa9e406ahrens
84571eb053Chris Kirby/*
84671eb053Chris Kirby * Get the [cm]time for an objset's snapshot dir
84771eb053Chris Kirby */
84871eb053Chris Kirbytimestruc_t dmu_objset_snap_cmtime(objset_t *os);
84971eb053Chris Kirby
850fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
851fa9e406ahrens
852fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
853fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
854fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
855fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
856fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
857fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
858fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
859edf345eMatthew Ahrensextern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
860edf345eMatthew Ahrensextern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
861fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
862b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
863ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
864ab04eb8timh    int maxlen, boolean_t *conflict);
86587e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
86687e5029ahrens    uint64_t *idp, uint64_t *offp);
8671484342Matthew Ahrens
8689966ca1Matthew Ahrenstypedef int objset_used_cb_t(dmu_object_type_t bonustype,
8699966ca1Matthew Ahrens    void *bonus, uint64_t *userp, uint64_t *groupp);
8701484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
8711484342Matthew Ahrens    objset_used_cb_t *cb);
872f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
873f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
874fa9e406ahrens
875fa9e406ahrens/*
876fa9e406ahrens * Return the txg number for the given assigned transaction.
877fa9e406ahrens */
87887e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
879fa9e406ahrens
880