dmu.h revision 8dfe5547fbf0979fc1065a8b6fddc1e940a7cf4f
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
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
2394d1a21Tim Haley * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24adaec86Matthew Ahrens * Copyright (c) 2011, 2017 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 []
32ec5cf9dAlexander Stetsenko */
3455da60bMark J Musante/* Portions Copyright 2010 Robert Milkowski */
3555da60bMark J Musante
36fa9e406ahrens#ifndef	_SYS_DMU_H
37fa9e406ahrens#define	_SYS_DMU_H
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 */
47bc9014eJustin Gibbs#include <sys/zfs_context.h>
48fa9e406ahrens#include <sys/inttypes.h>
49ecd6cf8marks#include <sys/cred.h>
504445fffMatthew Ahrens#include <sys/fs/zfs.h>
515602294Dan Kimmel#include <sys/zio_compress.h>
52a2cdcddPaul Dagnelie#include <sys/zio_priority.h>
54fa9e406ahrens#ifdef	__cplusplus
55fa9e406ahrensextern "C" {
58fa9e406ahrensstruct uio;
59c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct xuio;
6044eda4dmaybeestruct page;
61fa9e406ahrensstruct vnode;
62fa9e406ahrensstruct spa;
63fa9e406ahrensstruct zilog;
64fa9e406ahrensstruct zio;
65fa9e406ahrensstruct blkptr;
66fa9e406ahrensstruct zap_cursor;
67fa9e406ahrensstruct dsl_dataset;
68fa9e406ahrensstruct dsl_pool;
69fa9e406ahrensstruct dnode;
70fa9e406ahrensstruct drr_begin;
71fa9e406ahrensstruct drr_end;
727802d7bMatthew Ahrensstruct zbookmark_phys;
7344cd46cbillmstruct spa;
74a2eea2eahrensstruct nvlist;
752fdbea2Aleksandr Guzovskiystruct arc_buf;
76b24ab67Jeff Bonwickstruct zio_prop;
770a586ceMark Shellenbaumstruct sa_handle;
79fa9e406ahrenstypedef struct objset objset_t;
80fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
81fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
8279d7283Matthew Ahrenstypedef struct dnode dnode_t;
84ad135b5Christopher Sidentypedef enum dmu_object_byteswap {
85ad135b5Christopher Siden	DMU_BSWAP_UINT8,
86ad135b5Christopher Siden	DMU_BSWAP_UINT16,
87ad135b5Christopher Siden	DMU_BSWAP_UINT32,
88ad135b5Christopher Siden	DMU_BSWAP_UINT64,
89ad135b5Christopher Siden	DMU_BSWAP_ZAP,
90ad135b5Christopher Siden	DMU_BSWAP_DNODE,
91ad135b5Christopher Siden	DMU_BSWAP_OBJSET,
92ad135b5Christopher Siden	DMU_BSWAP_ZNODE,
93ad135b5Christopher Siden	DMU_BSWAP_OLDACL,
94ad135b5Christopher Siden	DMU_BSWAP_ACL,
95ad135b5Christopher Siden	/*
96ad135b5Christopher Siden	 * Allocating a new byteswap type number makes the on-disk format
97ad135b5Christopher Siden	 * incompatible with any other format that uses the same number.
98ad135b5Christopher Siden	 *
99ad135b5Christopher Siden	 * Data can usually be structured to work with one of the
100ad135b5Christopher Siden	 * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
101ad135b5Christopher Siden	 */
102ad135b5Christopher Siden	DMU_BSWAP_NUMFUNCS
103ad135b5Christopher Siden} dmu_object_byteswap_t;
104ad135b5Christopher Siden
105ad135b5Christopher Siden#define	DMU_OT_NEWTYPE 0x80
106ad135b5Christopher Siden#define	DMU_OT_METADATA 0x40
107ad135b5Christopher Siden#define	DMU_OT_BYTESWAP_MASK 0x3f
108ad135b5Christopher Siden
109ad135b5Christopher Siden/*
110ad135b5Christopher Siden * Defines a uint8_t object type. Object types specify if the data
111ad135b5Christopher Siden * in the object is metadata (boolean) and how to byteswap the data
112adb52d9Matthew Ahrens * (dmu_object_byteswap_t). All of the types created by this method
113adb52d9Matthew Ahrens * are cached in the dbuf metadata cache.
114ad135b5Christopher Siden */
115ad135b5Christopher Siden#define	DMU_OT(byteswap, metadata) \
116ad135b5Christopher Siden	(DMU_OT_NEWTYPE | \
117ad135b5Christopher Siden	((metadata) ? DMU_OT_METADATA : 0) | \
118ad135b5Christopher Siden	((byteswap) & DMU_OT_BYTESWAP_MASK))
119ad135b5Christopher Siden
120ad135b5Christopher Siden#define	DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
121ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
122ad135b5Christopher Siden	(ot) < DMU_OT_NUMTYPES)
123ad135b5Christopher Siden
124ad135b5Christopher Siden#define	DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
125ad135b5Christopher Siden	((ot) & DMU_OT_METADATA) : \
126ad135b5Christopher Siden	dmu_ot[(ot)].ot_metadata)
127ad135b5Christopher Siden
128adb52d9Matthew Ahrens#define	DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
129adb52d9Matthew Ahrens	B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache)
130adb52d9Matthew Ahrens
1315d7b4d4Matthew Ahrens/*
1325d7b4d4Matthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
1335d7b4d4Matthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
1345d7b4d4Matthew Ahrens * is repurposed for embedded BPs.
1355d7b4d4Matthew Ahrens */
1365d7b4d4Matthew Ahrens#define	DMU_OT_HAS_FILL(ot) \
1375d7b4d4Matthew Ahrens	((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
1385d7b4d4Matthew Ahrens
139ad135b5Christopher Siden#define	DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
140ad135b5Christopher Siden	((ot) & DMU_OT_BYTESWAP_MASK) : \
141ad135b5Christopher Siden	dmu_ot[(ot)].ot_byteswap)
142ad135b5Christopher Siden
143fa9e406ahrenstypedef enum dmu_object_type {
144fa9e406ahrens	DMU_OT_NONE,
145fa9e406ahrens	/* general: */
146fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
147fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
148fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
149fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
150cde58dbMatthew Ahrens	DMU_OT_BPOBJ,			/* UINT64 */
151cde58dbMatthew Ahrens	DMU_OT_BPOBJ_HDR,		/* UINT64 */
152fa9e406ahrens	/* spa: */
153fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
154fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
155fa9e406ahrens	/* zil: */
156fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
157fa9e406ahrens	/* dmu: */
158fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
159fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
160fa9e406ahrens	/* dsl: */
1611649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
16287e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
16387e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
164fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
1651649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
166fa9e406ahrens	/* zpl: */
167fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
168da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
169fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
170fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
171fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
172893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
173fa9e406ahrens	/* zvol: */
174fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
175fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
176fa9e406ahrens	/* other; for testing only! */
177fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
178fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
179fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
180ea8dc4beschrock	/* new object types: */
181ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
18206eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
18306eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
184b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
185ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
186da6c28aamw	DMU_OT_ACL,			/* ACL */
187da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
188da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
189da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
190088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
1913f9d6adLin Ling	DMU_OT_SCAN_QUEUE,		/* ZAP */
1921484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1931484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
194842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
195b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
196b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
1970a586ceMark Shellenbaum	DMU_OT_SA,			/* System attr */
1980a586ceMark Shellenbaum	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1990a586ceMark Shellenbaum	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
2000a586ceMark Shellenbaum	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
2013f9d6adLin Ling	DMU_OT_SCAN_XLATE,		/* ZAP */
2023f9d6adLin Ling	DMU_OT_DEDUP,			/* fake dedup BP from ddt_bp_create() */
203cde58dbMatthew Ahrens	DMU_OT_DEADLIST,		/* ZAP */
204cde58dbMatthew Ahrens	DMU_OT_DEADLIST_HDR,		/* UINT64 */
205cde58dbMatthew Ahrens	DMU_OT_DSL_CLONES,		/* ZAP */
206cde58dbMatthew Ahrens	DMU_OT_BPOBJ_SUBOBJ,		/* UINT64 */
207ad135b5Christopher Siden	/*
208ad135b5Christopher Siden	 * Do not allocate new object types here. Doing so makes the on-disk
209ad135b5Christopher Siden	 * format incompatible with any other format that uses the same object
210ad135b5Christopher Siden	 * type number.
211ad135b5Christopher Siden	 *
212ad135b5Christopher Siden	 * When creating an object which does not have one of the above types
213ad135b5Christopher Siden	 * use the DMU_OTN_* type with the correct byteswap and metadata
214ad135b5Christopher Siden	 * values.
215ad135b5Christopher Siden	 *
216ad135b5Christopher Siden	 * The DMU_OTN_* types do not have entries in the dmu_ot table,
217ad135b5Christopher Siden	 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
218ad135b5Christopher Siden	 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
219ad135b5Christopher Siden	 * and DMU_OTN_* types).
220ad135b5Christopher Siden	 */
221ad135b5Christopher Siden	DMU_OT_NUMTYPES,
222ad135b5Christopher Siden
223ad135b5Christopher Siden	/*
224ad135b5Christopher Siden	 * Names for valid types declared with DMU_OT().
225ad135b5Christopher Siden	 */
226ad135b5Christopher Siden	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
227ad135b5Christopher Siden	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
228ad135b5Christopher Siden	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
229ad135b5Christopher Siden	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
230ad135b5Christopher Siden	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
231ad135b5Christopher Siden	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
232ad135b5Christopher Siden	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
233ad135b5Christopher Siden	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
234ad135b5Christopher Siden	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
235ad135b5Christopher Siden	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
236fa9e406ahrens} dmu_object_type_t;
238f864f99Prakash Surya/*
239f864f99Prakash Surya * These flags are intended to be used to specify the "txg_how"
240f864f99Prakash Surya * parameter when calling the dmu_tx_assign() function. See the comment
241f864f99Prakash Surya * above dmu_tx_assign() for more details on the meaning of these flags.
242f864f99Prakash Surya */
243f864f99Prakash Surya#define	TXG_NOWAIT	(0ULL)
244f864f99Prakash Surya#define	TXG_WAIT	(1ULL<<0)
245f864f99Prakash Surya#define	TXG_NOTHROTTLE	(1ULL<<1)
2463b2aab1Matthew Ahrens
247fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
248fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
249fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
250fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
251fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
252da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
253fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
254fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
2560b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
2570b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
25812380e1Arne Jansen#define	DS_FIND_SERIALIZE	(1<<2)
261fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
262fa9e406ahrens * operation, including metadata.
263fa9e406ahrens */
264b515258Matthew Ahrens#define	DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */
265cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
2671484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
2681484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
2691484342Matthew Ahrens
2710a586ceMark Shellenbaum * artificial blkids for bonus buffer and spill blocks
2720a586ceMark Shellenbaum */
2730a586ceMark Shellenbaum#define	DMU_BONUS_BLKID		(-1ULL)
2740a586ceMark Shellenbaum#define	DMU_SPILL_BLKID		(-2ULL)
2750a586ceMark Shellenbaum/*
276fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
277fa9e406ahrens */
278503ad85Matthew Ahrensint dmu_objset_hold(const char *name, void *tag, objset_t **osp);
279503ad85Matthew Ahrensint dmu_objset_own(const char *name, dmu_objset_type_t type,
280503ad85Matthew Ahrens    boolean_t readonly, void *tag, objset_t **osp);
281503ad85Matthew Ahrensvoid dmu_objset_rele(objset_t *os, void *tag);
282503ad85Matthew Ahrensvoid dmu_objset_disown(objset_t *os, void *tag);
283503ad85Matthew Ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
284503ad85Matthew Ahrens
2853b2aab1Matthew Ahrensvoid dmu_objset_evict_dbufs(objset_t *os);
286ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
287ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
2883b2aab1Matthew Ahrensint dmu_objset_clone(const char *name, const char *origin);
2893b2aab1Matthew Ahrensint dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
2904445fffMatthew Ahrens    struct nvlist *errlist);
2914445fffMatthew Ahrensint dmu_objset_snapshot_one(const char *fsname, const char *snapname);
2924445fffMatthew Ahrensint dmu_objset_snapshot_tmp(const char *, const char *, int);
293fd13687Matthew Ahrensint dmu_objset_find(char *name, int func(const char *, void *), void *arg,
294fa9e406ahrens    int flags);
295fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
2963b2aab1Matthew Ahrensint dsl_dataset_rename_snapshot(const char *fsname,
2973b2aab1Matthew Ahrens    const char *oldsnapname, const char *newsnapname, boolean_t recursive);
2985cabbc6Prashanth Sreenivasaint dmu_objset_remap_indirects(const char *fsname);
300fa9e406ahrenstypedef struct dmu_buf {
301fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
302fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
303fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
304fa9e406ahrens	void *db_data;			/* data in buffer */
305fa9e406ahrens} dmu_buf_t;
30899653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
30999653d4eschrock */
310fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
311fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
312ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_WRITE	"features_for_write"
313ad135b5Christopher Siden#define	DMU_POOL_FEATURES_FOR_READ	"features_for_read"
314ad135b5Christopher Siden#define	DMU_POOL_FEATURE_DESCRIPTIONS	"feature_descriptions"
31543466aaMax Grossman#define	DMU_POOL_FEATURE_ENABLED_TXG	"feature_enabled_txg"
316fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
317cde58dbMatthew Ahrens#define	DMU_POOL_SYNC_BPOBJ		"sync_bplist"
318ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
319ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
32099653d4eschrock#define	DMU_POOL_SPARES			"spares"
32199653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
32206eeb2aek#define	DMU_POOL_HISTORY		"history"
323b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
324fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
325ca45db4Chris Kirby#define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
326b24ab67Jeff Bonwick#define	DMU_POOL_DDT			"DDT-%s-%s-%s"
327b24ab67Jeff Bonwick#define	DMU_POOL_DDT_STATS		"DDT-statistics"
3283f9d6adLin Ling#define	DMU_POOL_CREATION_VERSION	"creation_version"
3293f9d6adLin Ling#define	DMU_POOL_SCAN			"scan"
330cde58dbMatthew Ahrens#define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
331ad135b5Christopher Siden#define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
332f174573Matthew Ahrens#define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"
33345818eeMatthew Ahrens#define	DMU_POOL_CHECKSUM_SALT		"org.illumos:checksum_salt"
334215198aJoe Stein#define	DMU_POOL_VDEV_ZAP_MAP		"com.delphix:vdev_zap_map"
3355cabbc6Prashanth Sreenivasa#define	DMU_POOL_REMOVING		"com.delphix:removing"
3365cabbc6Prashanth Sreenivasa#define	DMU_POOL_OBSOLETE_BPOBJ		"com.delphix:obsolete_bpobj"
3375cabbc6Prashanth Sreenivasa#define	DMU_POOL_CONDENSING_INDIRECT	"com.delphix:condensing_indirect"
3388671400Serapheim Dimitropoulos#define	DMU_POOL_ZPOOL_CHECKPOINT	"com.delphix:zpool_checkpoint"
341fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
342fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
343fa9e406ahrens *
344fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
345fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
346fa9e406ahrens * newly allocated object in this transaction).
347fa9e406ahrens *
348fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
349fa9e406ahrens *
350fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
351fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
352fa9e406ahrens *
353fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
354fa9e406ahrens */
355fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
356fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
357221813cMatthew Ahrensuint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
358221813cMatthew Ahrens    int indirect_blockshift,
359221813cMatthew Ahrens    dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
360fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
361fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
362fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
363e77d42eMatthew Ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
366fa9e406ahrens * Free an object from this objset.
367fa9e406ahrens *
368fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
369fa9e406ahrens * dmu_free(object, 0, -1, tx)).
370fa9e406ahrens *
371fa9e406ahrens * The object need not be held in the transaction.
372fa9e406ahrens *
373fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
374fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
375fa9e406ahrens * free it; it fails and returns EBUSY.
376fa9e406ahrens *
377fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
378fa9e406ahrens *
379fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
380fa9e406ahrens */
381fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
384fa9e406ahrens * Find the next allocated or free object.
385fa9e406ahrens *
386fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
3876754306ahrens * object which is allocated.  Ignore objects which have not been
3886754306ahrens * modified since txg.
389fa9e406ahrens *
390fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
391fa9e406ahrens *
392fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
393fa9e406ahrens */
3946754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
3956754306ahrens    boolean_t hole, uint64_t txg);
398fa9e406ahrens * Set the data blocksize for an object.
399fa9e406ahrens *
400fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
401fa9e406ahrens * the first block is allocated already, the new size must be greater
402fa9e406ahrens * than the current block size.  If these conditions are not met,
403fa9e406ahrens * ENOTSUP will be returned.
404fa9e406ahrens *
405fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
406fa9e406ahrens * contents, or ENOTSUP as described above.
407fa9e406ahrens */
408fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
409fa9e406ahrens    int ibs, dmu_tx_t *tx);
412fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
413fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
414fa9e406ahrens */
415fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
416fa9e406ahrens    dmu_tx_t *tx);
419fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
420fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
421fa9e406ahrens */
422fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
423fa9e406ahrens    dmu_tx_t *tx);
4255cabbc6Prashanth Sreenivasaint dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg);
4265cabbc6Prashanth Sreenivasa
4275d7b4d4Matthew Ahrensvoid
4285d7b4d4Matthew Ahrensdmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
4295d7b4d4Matthew Ahrens    void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
4305d7b4d4Matthew Ahrens    int compressed_size, int byteorder, dmu_tx_t *tx);
4315d7b4d4Matthew Ahrens
433b24ab67Jeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc.
43444cd46cbillm */
435b24ab67Jeff Bonwick#define	WP_NOFILL	0x1
436b24ab67Jeff Bonwick#define	WP_DMU_SYNC	0x2
4370a586ceMark Shellenbaum#define	WP_SPILL	0x4
438b24ab67Jeff Bonwick
43979d7283Matthew Ahrensvoid dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp,
440adaec86Matthew Ahrens    struct zio_prop *zp);
442fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
443fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
444fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
4455cabbc6Prashanth Sreenivasa * data.  As with any normal buffer, you must call dmu_buf_will_dirty()
4465cabbc6Prashanth Sreenivasa * before modifying it, and the
447fa9e406ahrens * object must be held in an assigned transaction before calling
448fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
449fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
4503e30c24Will Andrews *
4513e30c24Will Andrews * Returns ENOENT, EIO, or 0.
452fa9e406ahrens */
453ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
454fa9e406ahrensint dmu_bonus_max(void);
4551934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
4560a586ceMark Shellenbaumint dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
457744947dTom Ericksondmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
4580a586ceMark Shellenbaumint dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
4590a586ceMark Shellenbaum
4600a586ceMark Shellenbaum/*
4610a586ceMark Shellenbaum * Special spill buffer support used by "SA" framework
4620a586ceMark Shellenbaum */
4630a586ceMark Shellenbaum
4640a586ceMark Shellenbaumint dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
46579d7283Matthew Ahrensint dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags,
4660a586ceMark Shellenbaum    void *tag, dmu_buf_t **dbp);
4670a586ceMark Shellenbaumint dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
470fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
471fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
472fa9e406ahrens * that it will remain in memory.  You must release the hold with
473fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
474fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
475fa9e406ahrens *
476fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
477fa9e406ahrens * on the returned buffer before reading or writing the buffer's
478fa9e406ahrens * db_data.  The comments for those routines describe what particular
479fa9e406ahrens * operations are valid after calling them.
480fa9e406ahrens *
481fa9e406ahrens * The object number must be a valid, allocated object number.
482fa9e406ahrens */
483ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
48447cb52dJeff Bonwick    void *tag, dmu_buf_t **, int flags);
48579d7283Matthew Ahrensint dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
48679d7283Matthew Ahrens    void *tag, dmu_buf_t **dbp, int flags);
487e57a022Justin T. Gibbs
488e57a022Justin T. Gibbs/*
489e57a022Justin T. Gibbs * Add a reference to a dmu buffer that has already been held via
490e57a022Justin T. Gibbs * dmu_buf_hold() in the current context.
491e57a022Justin T. Gibbs */
492fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
493e57a022Justin T. Gibbs
494e57a022Justin T. Gibbs/*
495e57a022Justin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state,
496e57a022Justin T. Gibbs * using a pointer that may have been invalidated by eviction processing.
497e57a022Justin T. Gibbs * The request will succeed if the passed in dbuf still represents the
498e57a022Justin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least
499e57a022Justin T. Gibbs * one hold by a user other than the syncer.
500e57a022Justin T. Gibbs */
501e57a022Justin T. Gibbsboolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object,
502e57a022Justin T. Gibbs    uint64_t blkid, void *tag);
503e57a022Justin T. Gibbs
504ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
505fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
508fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
509fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
510fa9e406ahrens * returned (in *dbpp).
511fa9e406ahrens *
512fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
513fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
514fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
515fa9e406ahrens * individually with dmu_buf_rele.
516fa9e406ahrens */
51713506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
518cf6106cMatthew Ahrens    uint64_t length, boolean_t read, void *tag,
519cf6106cMatthew Ahrens    int *numbufsp, dmu_buf_t ***dbpp);
5208dfe554Richard Yaoint dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
5218dfe554Richard Yao    boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp,
5228dfe554Richard Yao    uint32_t flags);
523ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
525bc9014eJustin Gibbstypedef void dmu_buf_evict_func_t(void *user_ptr);
526bc9014eJustin Gibbs
528bc9014eJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the
529bc9014eJustin Gibbs * duration of its lifetime.  This allows the user of a dbuf (client)
530bc9014eJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a
531bc9014eJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
532bc9014eJustin Gibbs * when that dbuf has been evicted.  Clients typically respond to the
533bc9014eJustin Gibbs * eviction notification by freeing their private data, thus ensuring
534bc9014eJustin Gibbs * the same lifetime for both dbuf and private data.
535fa9e406ahrens *
536bc9014eJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the
537bc9014eJustin Gibbs * client's responsibility.  All current consumers of the API with private
538bc9014eJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for
539bc9014eJustin Gibbs * their private data.  This allows conversions between the two types
540bc9014eJustin Gibbs * with a simple cast.  Since the DMU buf user API never needs access
541bc9014eJustin Gibbs * to the private data, other strategies can be employed if necessary
542bc9014eJustin Gibbs * or convenient for the client (e.g. using container_of() to do the
543bc9014eJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as
544bc9014eJustin Gibbs * its first member).
545fa9e406ahrens *
546bc9014eJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any
547bc9014eJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available.
548bc9014eJustin Gibbs * For this reason, users must assume the dbuf has already been freed
549bc9014eJustin Gibbs * and not reference the dbuf from the callback context.
550fa9e406ahrens *
551bc9014eJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf
552bc9014eJustin Gibbs * is only referenced by dirty records (dirties == holds).  Otherwise the
553bc9014eJustin Gibbs * notification occurs after eviction processing for the dbuf begins.
554fa9e406ahrens */
555bc9014eJustin Gibbstypedef struct dmu_buf_user {
556bc9014eJustin Gibbs	/*
557bc9014eJustin Gibbs	 * Asynchronous user eviction callback state.
558bc9014eJustin Gibbs	 */
559bc9014eJustin Gibbs	taskq_ent_t	dbu_tqent;
560bc9014eJustin Gibbs
56140510e8Josef 'Jeff' Sipek	/*
56240510e8Josef 'Jeff' Sipek	 * This instance's eviction function pointers.
56340510e8Josef 'Jeff' Sipek	 *
56440510e8Josef 'Jeff' Sipek	 * dbu_evict_func_sync is called synchronously and then
56540510e8Josef 'Jeff' Sipek	 * dbu_evict_func_async is executed asynchronously on a taskq.
56640510e8Josef 'Jeff' Sipek	 */
56740510e8Josef 'Jeff' Sipek	dmu_buf_evict_func_t *dbu_evict_func_sync;
56840510e8Josef 'Jeff' Sipek	dmu_buf_evict_func_t *dbu_evict_func_async;
569bc9014eJustin Gibbs#ifdef ZFS_DEBUG
570bc9014eJustin Gibbs	/*
571bc9014eJustin Gibbs	 * Pointer to user's dbuf pointer.  NULL for clients that do
572bc9014eJustin Gibbs	 * not associate a dbuf with their user data.
573bc9014eJustin Gibbs	 *
574bc9014eJustin Gibbs	 * The dbuf pointer is cleared upon eviction so as to catch
575bc9014eJustin Gibbs	 * use-after-evict bugs in clients.
576bc9014eJustin Gibbs	 */
577bc9014eJustin Gibbs	dmu_buf_t **dbu_clear_on_evict_dbufp;
578bc9014eJustin Gibbs#endif
579bc9014eJustin Gibbs} dmu_buf_user_t;
580bc9014eJustin Gibbs
582bc9014eJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function
583bc9014eJustin Gibbs * evict_func, to be called when the user is evicted.
584bc9014eJustin Gibbs *
585bc9014eJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t.
586bc9014eJustin Gibbs *       To allow enforcement of this, dbu must already be zeroed on entry.
587fa9e406ahrens */
5884ba5b96Dan Kimmel/*ARGSUSED*/
589bc9014eJustin Gibbsinline void
59040510e8Josef 'Jeff' Sipekdmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync,
59140510e8Josef 'Jeff' Sipek    dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp)
592bc9014eJustin Gibbs{
59340510e8Josef 'Jeff' Sipek	ASSERT(dbu->dbu_evict_func_sync == NULL);
59440510e8Josef 'Jeff' Sipek	ASSERT(dbu->dbu_evict_func_async == NULL);
59540510e8Josef 'Jeff' Sipek
59640510e8Josef 'Jeff' Sipek	/* must have at least one evict func */
59740510e8Josef 'Jeff' Sipek	IMPLY(evict_func_sync == NULL, evict_func_async != NULL);
59840510e8Josef 'Jeff' Sipek	dbu->dbu_evict_func_sync = evict_func_sync;
59940510e8Josef 'Jeff' Sipek	dbu->dbu_evict_func_async = evict_func_async;
600bc9014eJustin Gibbs#ifdef ZFS_DEBUG
601bc9014eJustin Gibbs	dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
602bc9014eJustin Gibbs#endif
603bc9014eJustin Gibbs}
606bc9014eJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's
607bc9014eJustin Gibbs * data is cleared or its reference count goes to zero) eviction processing.
608bc9014eJustin Gibbs *
609bc9014eJustin Gibbs * Returns NULL on success, or the existing user if another user currently
610bc9014eJustin Gibbs * owns the buffer.
611bc9014eJustin Gibbs */
612bc9014eJustin Gibbsvoid *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user);
613bc9014eJustin Gibbs
614bc9014eJustin Gibbs/*
615bc9014eJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and
616bc9014eJustin Gibbs * reference counts are equal) eviction processing.
617bc9014eJustin Gibbs *
618bc9014eJustin Gibbs * Returns NULL on success, or the existing user if another user currently
619bc9014eJustin Gibbs * owns the buffer.
620bc9014eJustin Gibbs */
621bc9014eJustin Gibbsvoid *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user);
622bc9014eJustin Gibbs
623bc9014eJustin Gibbs/*
624bc9014eJustin Gibbs * Replace the current user of a dbuf.
625bc9014eJustin Gibbs *
626bc9014eJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with
627bc9014eJustin Gibbs * "new_user" and returns the user data pointer that was replaced.
628bc9014eJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer.
629bc9014eJustin Gibbs */
630bc9014eJustin Gibbsvoid *dmu_buf_replace_user(dmu_buf_t *db,
631bc9014eJustin Gibbs    dmu_buf_user_t *old_user, dmu_buf_user_t *new_user);
632bc9014eJustin Gibbs
633bc9014eJustin Gibbs/*
634bc9014eJustin Gibbs * Remove the specified user data for a DMU buffer.
635bc9014eJustin Gibbs *
636bc9014eJustin Gibbs * Returns the user that was removed on success, or the current user if
637bc9014eJustin Gibbs * another user currently owns the buffer.
638bc9014eJustin Gibbs */
639bc9014eJustin Gibbsvoid *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user);
640bc9014eJustin Gibbs
641bc9014eJustin Gibbs/*
642bc9014eJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf.
643fa9e406ahrens */
644fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
646ae97279Matthew Ahrensobjset_t *dmu_buf_get_objset(dmu_buf_t *db);
64779d7283Matthew Ahrensdnode_t *dmu_buf_dnode_enter(dmu_buf_t *db);
64879d7283Matthew Ahrensvoid dmu_buf_dnode_exit(dmu_buf_t *db);
649ae97279Matthew Ahrens
650bc9014eJustin Gibbs/* Block until any in-progress dmu buf user evictions complete. */
651bc9014eJustin Gibbsvoid dmu_buf_user_evict_wait(void);
652bc9014eJustin Gibbs
65480901aeGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set.
65580901aeGeorge Wilson */
65680901aeGeorge Wilsonstruct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
65780901aeGeorge Wilson
65880901aeGeorge Wilson/*
659fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
660fa9e406ahrens *
661fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
662fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
663fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
664fa9e406ahrens */
665fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
668fa9e406ahrens * You must create a transaction, then hold the objects which you will
669fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
670fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
671fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
672fa9e406ahrens * part of this transaction.  You can't modify buffers before the
673fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
674fa9e406ahrens * belong to objects which this transaction holds; you can't hold
675fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
676fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
677fa9e406ahrens * don't have to.
678fa9e406ahrens *
679fa9e406ahrens * You can abort the transaction before it has been assigned.
680fa9e406ahrens *
681fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
682fa9e406ahrens * regardless of transaction state.
683fa9e406ahrens */
685fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
686fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
688fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
689fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
690b0c42cdbzzzvoid dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
691b0c42cdbzzz    int len);
692fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
693fa9e406ahrens    uint64_t len);
694b0c42cdbzzzvoid dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
695b0c42cdbzzz    uint64_t len);
6965cabbc6Prashanth Sreenivasavoid dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object);
6971484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
698b0c42cdbzzzvoid dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add,
699b0c42cdbzzz    const char *name);
700fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
701b0c42cdbzzzvoid dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn);
7020a586ceMark Shellenbaumvoid dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
7030a586ceMark Shellenbaumvoid dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
7040a586ceMark Shellenbaumvoid dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
705fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
706f864f99Prakash Suryaint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
7078a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
708fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
7094bb7380Matthew Ahrensvoid dmu_tx_mark_netfree(dmu_tx_t *tx);
712d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
713d20e665Ricardo M. Correia *
714d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
715d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
716d20e665Ricardo M. Correia * freeing it.
717d20e665Ricardo M. Correia *
718d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
719d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
720d20e665Ricardo M. Correia *
721d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
722d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
723d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
724d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
725d20e665Ricardo M. Correia */
726d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
727d20e665Ricardo M. Correia
728d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
729d20e665Ricardo M. Correia    void *dcb_data);
730d20e665Ricardo M. Correia
731d20e665Ricardo M. Correia/*
732fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
733ad135b5Christopher Siden * -1, the range from offset to end-of-file is freed.
734fa9e406ahrens */
735ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
736fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
737cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
738cdb0ab7maybee	uint64_t size);
739713d6c2Matthew Ahrensint dmu_free_long_object(objset_t *os, uint64_t object);
742fa9e406ahrens * Convenience functions.
743fa9e406ahrens *
744fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
745fa9e406ahrens * nonrecoverable I/O error.
746fa9e406ahrens */
7477bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
7487bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
749ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
7507bfdf01Neil Perrin	void *buf, uint32_t flags);
751b0c42cdbzzzint dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
752b0c42cdbzzz    uint32_t flags);
753fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
754fa9e406ahrens	const void *buf, dmu_tx_t *tx);
755b0c42cdbzzzvoid dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
756b0c42cdbzzz    const void *buf, dmu_tx_t *tx);
75782c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
75882c9918Tim Haley	dmu_tx_t *tx);
759feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
760f8554bbMatthew Ahrensint dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size);
7618dfe554Richard Yaoint dmu_read_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size);
762feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
763feb08c6billm    dmu_tx_t *tx);
76494d1a21Tim Haleyint dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
76594d1a21Tim Haley    dmu_tx_t *tx);
7668dfe554Richard Yaoint dmu_write_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size,
7678dfe554Richard Yao    dmu_tx_t *tx);
76844eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
76944eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
7702fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
7712fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
7728dfe554Richard Yaovoid dmu_assign_arcbuf_dnode(dnode_t *handle, uint64_t offset,
7738dfe554Richard Yao    struct arc_buf *buf, dmu_tx_t *tx);
7742fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
7752fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
776c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_init(struct xuio *uio, int niov);
777c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_fini(struct xuio *uio);
778c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
779c242f9achunli zhang - Sun Microsystems - Irvine United States    size_t n);
780c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_cnt(struct xuio *uio);
781c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
782c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_clear(struct xuio *uio, int i);
78399aa8b5Prashanth Sreenivasavoid xuio_stat_wbuf_copied(void);
78499aa8b5Prashanth Sreenivasavoid xuio_stat_wbuf_nocopy(void);
786cf6106cMatthew Ahrensextern boolean_t zfs_prefetch_disable;
787b515258Matthew Ahrensextern int zfs_max_recordsize;
790fa9e406ahrens * Asynchronously try to read in the data.
791fa9e406ahrens */
792a2cdcddPaul Dagnelievoid dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
793a2cdcddPaul Dagnelie    uint64_t len, enum zio_priority pri);
795fa9e406ahrenstypedef struct dmu_object_info {
796b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
797fa9e406ahrens	uint32_t doi_data_block_size;
798fa9e406ahrens	uint32_t doi_metadata_block_size;
799fa9e406ahrens	dmu_object_type_t doi_type;
800fa9e406ahrens	dmu_object_type_t doi_bonus_type;
801b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
802fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
803fa9e406ahrens	uint8_t doi_checksum;
804fa9e406ahrens	uint8_t doi_compress;
805e77d42eMatthew Ahrens	uint8_t doi_nblkptr;
806e77d42eMatthew Ahrens	uint8_t doi_pad[4];
807b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
808b24ab67Jeff Bonwick	uint64_t doi_max_offset;
809b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
810fa9e406ahrens} dmu_object_info_t;
812e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
814fa9e406ahrenstypedef struct dmu_object_type_info {
815ad135b5Christopher Siden	dmu_object_byteswap_t	ot_byteswap;
816fa9e406ahrens	boolean_t		ot_metadata;
817adb52d9Matthew Ahrens	boolean_t		ot_dbuf_metadata_cache;
818fa9e406ahrens	char			*ot_name;
819fa9e406ahrens} dmu_object_type_info_t;
821ad135b5Christopher Sidentypedef struct dmu_object_byteswap_info {
822ad135b5Christopher Siden	arc_byteswap_func_t	*ob_func;
823ad135b5Christopher Siden	char			*ob_name;
824ad135b5Christopher Siden} dmu_object_byteswap_info_t;
825ad135b5Christopher Siden
826fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
827ad135b5Christopher Sidenextern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
830fa9e406ahrens * Get information on a DMU object.
831fa9e406ahrens *
832fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
833fa9e406ahrens *
834fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
835fa9e406ahrens */
836fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
8373e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dnode in hand. */
83879d7283Matthew Ahrensvoid dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi);
8393e30c24Will Andrews/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
840fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
8413e30c24Will Andrews/*
8423e30c24Will Andrews * Like dmu_object_info_from_db, but faster still when you only care about
8433e30c24Will Andrews * the size.  This is specifically optimized for zfs_getattr().
8443e30c24Will Andrews */
845fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
846fa9e406ahrens    u_longlong_t *nblk512);
848fa9e406ahrenstypedef struct dmu_objset_stats {
849a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
850a2eea2eahrens	uint64_t dds_creation_txg;
8513cb34c6ahrens	uint64_t dds_guid;
852fa9e406ahrens	dmu_objset_type_t dds_type;
853fa9e406ahrens	uint8_t dds_is_snapshot;
85431fd60dahrens	uint8_t dds_inconsistent;
8559adfa60Matthew Ahrens	char dds_origin[ZFS_MAX_DATASET_NAME_LEN];
856a2eea2eahrens} dmu_objset_stats_t;
859a2eea2eahrens * Get stats on a dataset.
860a2eea2eahrens */
861a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
864a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
865a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
866a2eea2eahrens */