dmu.h revision 55da60b91d96984f12de050ce428373ea25c7f35
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 */
2294d1a21Tim Haley * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23fa9e406ahrens */
2555da60bMark J Musante/* Portions Copyright 2010 Robert Milkowski */
2655da60bMark J Musante
27fa9e406ahrens#ifndef	_SYS_DMU_H
28fa9e406ahrens#define	_SYS_DMU_H
31fa9e406ahrens * This file describes the interface that the DMU provides for its
32fa9e406ahrens * consumers.
33fa9e406ahrens *
34fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
35fa9e406ahrens * dmu_spa.h.
36fa9e406ahrens */
38fa9e406ahrens#include <sys/inttypes.h>
39fa9e406ahrens#include <sys/types.h>
40fa9e406ahrens#include <sys/param.h>
41ecd6cf8marks#include <sys/cred.h>
4271eb053Chris Kirby#include <sys/time.h>
44fa9e406ahrens#ifdef	__cplusplus
45fa9e406ahrensextern "C" {
48fa9e406ahrensstruct uio;
49c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct xuio;
5044eda4dmaybeestruct page;
51fa9e406ahrensstruct vnode;
52fa9e406ahrensstruct spa;
53fa9e406ahrensstruct zilog;
54fa9e406ahrensstruct zio;
55fa9e406ahrensstruct blkptr;
56fa9e406ahrensstruct zap_cursor;
57fa9e406ahrensstruct dsl_dataset;
58fa9e406ahrensstruct dsl_pool;
59fa9e406ahrensstruct dnode;
60fa9e406ahrensstruct drr_begin;
61fa9e406ahrensstruct drr_end;
6244cd46cbillmstruct zbookmark;
6344cd46cbillmstruct spa;
64a2eea2eahrensstruct nvlist;
652fdbea2Aleksandr Guzovskiystruct arc_buf;
66b24ab67Jeff Bonwickstruct zio_prop;
670a586ceMark Shellenbaumstruct sa_handle;
69fa9e406ahrenstypedef struct objset objset_t;
70fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
71fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
73fa9e406ahrenstypedef enum dmu_object_type {
74fa9e406ahrens	DMU_OT_NONE,
75fa9e406ahrens	/* general: */
76fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
77fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
78fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
79fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
80fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
81fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
82fa9e406ahrens	/* spa: */
83fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
84fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
85fa9e406ahrens	/* zil: */
86fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
87fa9e406ahrens	/* dmu: */
88fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
89fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
90fa9e406ahrens	/* dsl: */
911649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
9287e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
9387e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
94fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
951649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
96fa9e406ahrens	/* zpl: */
97fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
98da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
99fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
100fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
101fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
102893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
103fa9e406ahrens	/* zvol: */
104fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
105fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
106fa9e406ahrens	/* other; for testing only! */
107fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
108fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
109fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
110ea8dc4beschrock	/* new object types: */
111ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
11206eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
11306eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
114b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
115ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
116da6c28aamw	DMU_OT_ACL,			/* ACL */
117da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
118da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
119da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
120088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
121088f389ahrens	DMU_OT_SCRUB_QUEUE,		/* ZAP */
1221484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1231484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
124842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
125b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
126b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
1270a586ceMark Shellenbaum	DMU_OT_SA,			/* System attr */
1280a586ceMark Shellenbaum	DMU_OT_SA_MASTER_NODE,		/* ZAP */
1290a586ceMark Shellenbaum	DMU_OT_SA_ATTR_REGISTRATION,	/* ZAP */
1300a586ceMark Shellenbaum	DMU_OT_SA_ATTR_LAYOUTS,		/* ZAP */
131fa9e406ahrens	DMU_OT_NUMTYPES
132fa9e406ahrens} dmu_object_type_t;
134fa9e406ahrenstypedef enum dmu_objset_type {
135fa9e406ahrens	DMU_OST_NONE,
136fa9e406ahrens	DMU_OST_META,
137fa9e406ahrens	DMU_OST_ZFS,
138fa9e406ahrens	DMU_OST_ZVOL,
139fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
140fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
141fa9e406ahrens	DMU_OST_NUMTYPES
142fa9e406ahrens} dmu_objset_type_t;
144fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
145fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
146fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
147fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
148fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
149da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
150fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
151fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
1530b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
1540b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
157fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
158fa9e406ahrens * operation, including metadata.
159fa9e406ahrens */
160fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
161cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
1631484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
1641484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
165b24ab67Jeff Bonwick#define	DMU_DEADLIST_OBJECT	(-3ULL)
1661484342Matthew Ahrens
1680a586ceMark Shellenbaum * artificial blkids for bonus buffer and spill blocks
1690a586ceMark Shellenbaum */
1700a586ceMark Shellenbaum#define	DMU_BONUS_BLKID		(-1ULL)
1710a586ceMark Shellenbaum#define	DMU_SPILL_BLKID		(-2ULL)
1720a586ceMark Shellenbaum/*
173fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
174fa9e406ahrens */
175503ad85Matthew Ahrensint dmu_objset_hold(const char *name, void *tag, objset_t **osp);
176503ad85Matthew Ahrensint dmu_objset_own(const char *name, dmu_objset_type_t type,
177503ad85Matthew Ahrens    boolean_t readonly, void *tag, objset_t **osp);
178503ad85Matthew Ahrensvoid dmu_objset_rele(objset_t *os, void *tag);
179503ad85Matthew Ahrensvoid dmu_objset_disown(objset_t *os, void *tag);
180503ad85Matthew Ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
181503ad85Matthew Ahrens
1821934e92maybeeint dmu_objset_evict_dbufs(objset_t *os);
183ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
184ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
185ae46e4cMatthew Ahrensint dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
186ae46e4cMatthew Ahrens    uint64_t flags);
187842727cChris Kirbyint dmu_objset_destroy(const char *name, boolean_t defer);
188842727cChris Kirbyint dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
189ea2f5b9Matthew Ahrensint dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
190ea2f5b9Matthew Ahrens    boolean_t recursive);
191cdf5b4cmmusanteint dmu_objset_rename(const char *name, const char *newname,
192cdf5b4cmmusante    boolean_t recursive);
193fd13687Matthew Ahrensint dmu_objset_find(char *name, int func(const char *, void *), void *arg,
194fa9e406ahrens    int flags);
195fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
197fa9e406ahrenstypedef struct dmu_buf {
198fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
199fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
200fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
201fa9e406ahrens	void *db_data;			/* data in buffer */
202fa9e406ahrens} dmu_buf_t;
204fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
20799653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
20899653d4eschrock */
209fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
210fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
211fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
212fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
213ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
214ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
21599653d4eschrock#define	DMU_POOL_SPARES			"spares"
21699653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
21706eeb2aek#define	DMU_POOL_HISTORY		"history"
218b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
219fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
220ca45db4Chris Kirby#define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
221b24ab67Jeff Bonwick#define	DMU_POOL_DDT			"DDT-%s-%s-%s"
222b24ab67Jeff Bonwick#define	DMU_POOL_DDT_STATS		"DDT-statistics"
224088f389ahrens/* 4x8 zbookmark_t */
225088f389ahrens#define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
226bbfd46cJeff Bonwick/* 4x8 ddt_bookmark_t */
227bbfd46cJeff Bonwick#define	DMU_POOL_SCRUB_DDT_BOOKMARK	"scrub_ddt_bookmark"
228bbfd46cJeff Bonwick/* 1x8 max_class */
229bbfd46cJeff Bonwick#define	DMU_POOL_SCRUB_DDT_CLASS_MAX	"scrub_ddt_class_max"
230088f389ahrens/* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
231088f389ahrens#define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
232088f389ahrens/* 1x8 txg */
233088f389ahrens#define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
234088f389ahrens/* 1x8 txg */
235088f389ahrens#define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
236088f389ahrens/* 1x4 enum scrub_func */
237088f389ahrens#define	DMU_POOL_SCRUB_FUNC		"scrub_func"
238088f389ahrens/* 1x8 count */
239088f389ahrens#define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
242fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
243fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
244fa9e406ahrens *
245fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
246fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
247fa9e406ahrens * newly allocated object in this transaction).
248fa9e406ahrens *
249fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
250fa9e406ahrens *
251fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
252fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
253fa9e406ahrens *
254fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
255fa9e406ahrens */
256fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
257fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
258fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
259fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
260fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
2612bf405aMark Maybee    int blocksize, dmu_object_type_t bonustype, int bonuslen);
264fa9e406ahrens * Free an object from this objset.
265fa9e406ahrens *
266fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
267fa9e406ahrens * dmu_free(object, 0, -1, tx)).
268fa9e406ahrens *
269fa9e406ahrens * The object need not be held in the transaction.
270fa9e406ahrens *
271fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
272fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
273fa9e406ahrens * free it; it fails and returns EBUSY.
274fa9e406ahrens *
275fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
276fa9e406ahrens *
277fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
278fa9e406ahrens */
279fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
282fa9e406ahrens * Find the next allocated or free object.
283fa9e406ahrens *
284fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
2856754306ahrens * object which is allocated.  Ignore objects which have not been
2866754306ahrens * modified since txg.
287fa9e406ahrens *
288fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
289fa9e406ahrens *
290fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
291fa9e406ahrens */
2926754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
2936754306ahrens    boolean_t hole, uint64_t txg);
296fa9e406ahrens * Set the data blocksize for an object.
297fa9e406ahrens *
298fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
299fa9e406ahrens * the first block is allocated already, the new size must be greater
300fa9e406ahrens * than the current block size.  If these conditions are not met,
301fa9e406ahrens * ENOTSUP will be returned.
302fa9e406ahrens *
303fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
304fa9e406ahrens * contents, or ENOTSUP as described above.
305fa9e406ahrens */
306fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
307fa9e406ahrens    int ibs, dmu_tx_t *tx);
310fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
311fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
312fa9e406ahrens */
313fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
314fa9e406ahrens    dmu_tx_t *tx);
317fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
318fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
319fa9e406ahrens */
320fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
321fa9e406ahrens    dmu_tx_t *tx);
324b24ab67Jeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc.
32544cd46cbillm */
326b24ab67Jeff Bonwick#define	WP_NOFILL	0x1
327b24ab67Jeff Bonwick#define	WP_DMU_SYNC	0x2
3280a586ceMark Shellenbaum#define	WP_SPILL	0x4
329b24ab67Jeff Bonwick
330b24ab67Jeff Bonwickvoid dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
331b24ab67Jeff Bonwick    struct zio_prop *zp);
333fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
334fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
335fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
336fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
337fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
338fa9e406ahrens * object must be held in an assigned transaction before calling
339fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
340fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
341fa9e406ahrens */
342ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
343fa9e406ahrensint dmu_bonus_max(void);
3441934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
3450a586ceMark Shellenbaumint dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
3460a586ceMark Shellenbaumint dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
3470a586ceMark Shellenbaum
3480a586ceMark Shellenbaum/*
3490a586ceMark Shellenbaum * Special spill buffer support used by "SA" framework
3500a586ceMark Shellenbaum */
3510a586ceMark Shellenbaum
3520a586ceMark Shellenbaumint dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
3530a586ceMark Shellenbaumint dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags,
3540a586ceMark Shellenbaum    void *tag, dmu_buf_t **dbp);
3550a586ceMark Shellenbaumint dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
358fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
359fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
360fa9e406ahrens * that it will remain in memory.  You must release the hold with
361fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
362fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
363fa9e406ahrens *
364fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
365fa9e406ahrens * on the returned buffer before reading or writing the buffer's
366fa9e406ahrens * db_data.  The comments for those routines describe what particular
367fa9e406ahrens * operations are valid after calling them.
368fa9e406ahrens *
369fa9e406ahrens * The object number must be a valid, allocated object number.
370fa9e406ahrens */
371ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
37247cb52dJeff Bonwick    void *tag, dmu_buf_t **, int flags);
373fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
374ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
375fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
378fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
379fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
380fa9e406ahrens * returned (in *dbpp).
381fa9e406ahrens *
382fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
383fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
384fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
385fa9e406ahrens * individually with dmu_buf_rele.
386fa9e406ahrens */
38713506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
38813506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
389ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
392fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
393fa9e406ahrens * been set.
394fa9e406ahrens *
395fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
396fa9e406ahrens *
397fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
398fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
399fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
400fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
401fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
402fa9e406ahrens *
403fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
404fa9e406ahrens * excised from the cache, so that you can clean up the data structure
405fa9e406ahrens * pointed to by user_ptr.
406fa9e406ahrens *
407fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
408fa9e406ahrens * objset with a given pageout func.
409fa9e406ahrens */
410fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
411fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
413fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
414fa9e406ahrens * when hold count goes to zero.
415fa9e406ahrens */
416fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
417fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
418fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
419fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
420fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
421fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
424fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
425fa9e406ahrens */
426fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
429fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
430fa9e406ahrens *
431fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
432fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
433fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
434fa9e406ahrens */
435fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
4383d69262Sanjeev Bagewadi * Tells if the given dbuf is freeable.
4393d69262Sanjeev Bagewadi */
4403d69262Sanjeev Bagewadiboolean_t dmu_buf_freeable(dmu_buf_t *);
4413d69262Sanjeev Bagewadi
4423d69262Sanjeev Bagewadi/*
443fa9e406ahrens * You must create a transaction, then hold the objects which you will
444fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
445fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
446fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
447fa9e406ahrens * part of this transaction.  You can't modify buffers before the
448fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
449fa9e406ahrens * belong to objects which this transaction holds; you can't hold
450fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
451fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
452fa9e406ahrens * don't have to.
453fa9e406ahrens *
454fa9e406ahrens * You can abort the transaction before it has been assigned.
455fa9e406ahrens *
456fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
457fa9e406ahrens * regardless of transaction state.
458fa9e406ahrens */
460fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
461fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
463fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
464fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
465fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
466fa9e406ahrens    uint64_t len);
4671484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
468fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
4690a586ceMark Shellenbaumvoid dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
4700a586ceMark Shellenbaumvoid dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
4710a586ceMark Shellenbaumvoid dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
472fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
473fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4748a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
475fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
478d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
479d20e665Ricardo M. Correia *
480d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
481d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
482d20e665Ricardo M. Correia * freeing it.
483d20e665Ricardo M. Correia *
484d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
485d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
486d20e665Ricardo M. Correia *
487d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
488d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
489d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
490d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
491d20e665Ricardo M. Correia */
492d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
493d20e665Ricardo M. Correia
494d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
495d20e665Ricardo M. Correia    void *dcb_data);
496d20e665Ricardo M. Correia
497d20e665Ricardo M. Correia/*
498fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
499fa9e406ahrens * zero, the range from offset to end-of-file is freed.
500fa9e406ahrens */
501ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
502fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
503cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
504cdb0ab7maybee	uint64_t size);
505cdb0ab7maybeeint dmu_free_object(objset_t *os, uint64_t object);
508fa9e406ahrens * Convenience functions.
509fa9e406ahrens *
510fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
511fa9e406ahrens * nonrecoverable I/O error.
512fa9e406ahrens */
5137bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
5147bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
515ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
5167bfdf01Neil Perrin	void *buf, uint32_t flags);
517fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
518fa9e406ahrens	const void *buf, dmu_tx_t *tx);
51982c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
52082c9918Tim Haley	dmu_tx_t *tx);
521feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
522feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
523feb08c6billm    dmu_tx_t *tx);
52494d1a21Tim Haleyint dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size,
52594d1a21Tim Haley    dmu_tx_t *tx);
52644eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
52744eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
5282fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
5292fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
5302fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
5312fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
532c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_init(struct xuio *uio, int niov);
533c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_fini(struct xuio *uio);
534c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off,
535c242f9achunli zhang - Sun Microsystems - Irvine United States    size_t n);
536c242f9achunli zhang - Sun Microsystems - Irvine United Statesint dmu_xuio_cnt(struct xuio *uio);
537c242f9achunli zhang - Sun Microsystems - Irvine United Statesstruct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i);
538c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid dmu_xuio_clear(struct xuio *uio, int i);
539c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_copied();
540c242f9achunli zhang - Sun Microsystems - Irvine United Statesvoid xuio_stat_wbuf_nocopy();
542416e0cdekextern int zfs_prefetch_disable;
545fa9e406ahrens * Asynchronously try to read in the data.
546fa9e406ahrens */
547fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
548fa9e406ahrens    uint64_t len);
550fa9e406ahrenstypedef struct dmu_object_info {
551b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
552fa9e406ahrens	uint32_t doi_data_block_size;
553fa9e406ahrens	uint32_t doi_metadata_block_size;
554fa9e406ahrens	dmu_object_type_t doi_type;
555fa9e406ahrens	dmu_object_type_t doi_bonus_type;
556b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
557fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
558fa9e406ahrens	uint8_t doi_checksum;
559fa9e406ahrens	uint8_t doi_compress;
560fa9e406ahrens	uint8_t doi_pad[5];
561b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
562b24ab67Jeff Bonwick	uint64_t doi_max_offset;
563b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
564fa9e406ahrens} dmu_object_info_t;
566e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
568fa9e406ahrenstypedef struct dmu_object_type_info {
569e45ce72ahrens	arc_byteswap_func_t	*ot_byteswap;
570fa9e406ahrens	boolean_t		ot_metadata;
571fa9e406ahrens	char			*ot_name;
572fa9e406ahrens} dmu_object_type_info_t;
574fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
577fa9e406ahrens * Get information on a DMU object.
578fa9e406ahrens *
579fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
580fa9e406ahrens *
581fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
582fa9e406ahrens */
583fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
584fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
585fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
586fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
587fa9e406ahrens    u_longlong_t *nblk512);
589fa9e406ahrenstypedef struct dmu_objset_stats {
590a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
591a2eea2eahrens	uint64_t dds_creation_txg;
5923cb34c6ahrens	uint64_t dds_guid;
593fa9e406ahrens	dmu_objset_type_t dds_type;
594fa9e406ahrens	uint8_t dds_is_snapshot;
59531fd60dahrens	uint8_t dds_inconsistent;
5963cb34c6ahrens	char dds_origin[MAXNAMELEN];
597a2eea2eahrens} dmu_objset_stats_t;
600a2eea2eahrens * Get stats on a dataset.
601a2eea2eahrens */
602a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
605a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
606a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
607a2eea2eahrens */
608a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
611a2eea2eahrens * Get the space usage statistics for statvfs().
612a2eea2eahrens *
613a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
614a2eea2eahrens * availbytes is the amount of space available to this objset, taking
615a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
616a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
617a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
618a2eea2eahrens *
619a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
620a2eea2eahrens * and available.
621a2eea2eahrens */
622a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
623a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
626a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
627a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
628a2eea2eahrens * change, so there is a small probability that it will collide.)
629fa9e406ahrens */
630a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
63271eb053Chris Kirby/*
63371eb053Chris Kirby * Get the [cm]time for an objset's snapshot dir
63471eb053Chris Kirby */
63571eb053Chris Kirbytimestruc_t dmu_objset_snap_cmtime(objset_t *os);
63671eb053Chris Kirby
637fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
639fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
640fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
641fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
642fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
643fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
644fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
645fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
64655da60bMark J Musanteextern uint64_t dmu_objset_syncprop(objset_t *os);
647e09fa4dNeil Perrinextern uint64_t dmu_objset_logbias(objset_t *os);
648fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
649b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
650ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
651ab04eb8timh    int maxlen, boolean_t *conflict);
65287e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
65387e5029ahrens    uint64_t *idp, uint64_t *offp);
6541484342Matthew Ahrens
6559966ca1Matthew Ahrenstypedef int objset_used_cb_t(dmu_object_type_t bonustype,
6569966ca1Matthew Ahrens    void *bonus, uint64_t *userp, uint64_t *groupp);
6571484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
6581484342Matthew Ahrens    objset_used_cb_t *cb);
659f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
660f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
663fa9e406ahrens * Return the txg number for the given assigned transaction.
664fa9e406ahrens */
66587e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
668fa9e406ahrens * Synchronous write.
669c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
670c5c6ffamaybee * provided buffer as a child of the parent zio.
671da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
672c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
673fa9e406ahrens * Note that while the data covered by this function will be on stable
674c5c6ffamaybee * storage when the write completes this new data does not become a
675fa9e406ahrens * permanent part of the file until the associated transaction commits.
676fa9e406ahrens */
677b24ab67Jeff Bonwick
678b24ab67Jeff Bonwick/*
679b24ab67Jeff Bonwick * {zfs,zvol,ztest}_get_done() args
680b24ab67Jeff Bonwick */
681b24ab67Jeff Bonwicktypedef struct zgd {
682b24ab67Jeff Bonwick	struct zilog	*zgd_zilog;
683b24ab67Jeff Bonwick	struct blkptr	*zgd_bp;
684b24ab67Jeff Bonwick	dmu_buf_t	*zgd_db;
685b24ab67Jeff Bonwick	struct rl	*zgd_rl;
686b24ab67Jeff Bonwick	void		*zgd_private;
687b24ab67Jeff Bonwick} zgd_t;
688b24ab67Jeff Bonwick
689b24ab67Jeff Bonwicktypedef void dmu_sync_cb_t(zgd_t *arg, int error);
690b24ab67Jeff Bonwickint dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
693fa9e406ahrens * Find the next hole or data block in file starting at *off
694fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
695fa9e406ahrens */
696fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
697fa9e406ahrens    uint64_t *off);
700fa9e406ahrens * Initial setup and final teardown.
701fa9e406ahrens */
702fa9e406ahrensextern void dmu_init(void);
703fa9e406ahrensextern void dmu_fini(void);
705fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
706fa9e406ahrens    uint64_t object, uint64_t offset, int len);
707fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
708fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
7103cb34c6ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
7113cb34c6ahrens    struct vnode *vp, offset_t *off);
7133cb34c6ahrenstypedef struct dmu_recv_cookie {
7143cb34c6ahrens	/*
7153cb34c6ahrens	 * This structure is opaque!
7163cb34c6ahrens	 *
7173cb34c6ahrens	 * If logical and real are different, we are recving the stream
7183cb34c6ahrens	 * into the "real" temporary clone, and then switching it with
7193cb34c6ahrens	 * the "logical" target.
7203cb34c6ahrens	 */
7213cb34c6ahrens	struct dsl_dataset *drc_logical_ds;
7223cb34c6ahrens	struct dsl_dataset *drc_real_ds;
7233cb34c6ahrens	struct drr_begin *drc_drrb;
7243cb34c6ahrens	char *drc_tosnap;
7259e69d7dLori Alt	char *drc_top_ds;
7263cb34c6ahrens	boolean_t drc_newfs;
7273cb34c6ahrens	boolean_t drc_force;
7283cb34c6ahrens} dmu_recv_cookie_t;
7309e69d7dLori Altint dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,
731f4b94bdMatthew Ahrens    boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
7323cb34c6ahrensint dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
7333cb34c6ahrensint dmu_recv_end(dmu_recv_cookie_t *drc);
735fa9e406ahrens/* CRC64 table */
736fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
737fa9e406ahrensextern uint64_t zfs_crc64_table[256];
739fa9e406ahrens#ifdef	__cplusplus
743fa9e406ahrens#endif	/* _SYS_DMU_H */