dmu.h revision b24ab6762772a3f6a89393947930c7fa61306783
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 */
21fa9e406ahrens/*
222bf405aMark Maybee * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23fa9e406ahrens * Use is subject to license terms.
24fa9e406ahrens */
25fa9e406ahrens
26fa9e406ahrens#ifndef	_SYS_DMU_H
27fa9e406ahrens#define	_SYS_DMU_H
28fa9e406ahrens
29fa9e406ahrens/*
30fa9e406ahrens * This file describes the interface that the DMU provides for its
31fa9e406ahrens * consumers.
32fa9e406ahrens *
33fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
34fa9e406ahrens * dmu_spa.h.
35fa9e406ahrens */
36fa9e406ahrens
37fa9e406ahrens#include <sys/inttypes.h>
38fa9e406ahrens#include <sys/types.h>
39fa9e406ahrens#include <sys/param.h>
40ecd6cf8marks#include <sys/cred.h>
4171eb053Chris Kirby#include <sys/time.h>
42fa9e406ahrens
43fa9e406ahrens#ifdef	__cplusplus
44fa9e406ahrensextern "C" {
45fa9e406ahrens#endif
46fa9e406ahrens
47fa9e406ahrensstruct uio;
4844eda4dmaybeestruct page;
49fa9e406ahrensstruct vnode;
50fa9e406ahrensstruct spa;
51fa9e406ahrensstruct zilog;
52fa9e406ahrensstruct zio;
53fa9e406ahrensstruct blkptr;
54fa9e406ahrensstruct zap_cursor;
55fa9e406ahrensstruct dsl_dataset;
56fa9e406ahrensstruct dsl_pool;
57fa9e406ahrensstruct dnode;
58fa9e406ahrensstruct drr_begin;
59fa9e406ahrensstruct drr_end;
6044cd46cbillmstruct zbookmark;
6144cd46cbillmstruct spa;
62a2eea2eahrensstruct nvlist;
632fdbea2Aleksandr Guzovskiystruct arc_buf;
64b24ab67Jeff Bonwickstruct zio_prop;
65fa9e406ahrens
66fa9e406ahrenstypedef struct objset objset_t;
67fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
68fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
69fa9e406ahrens
70fa9e406ahrenstypedef enum dmu_object_type {
71fa9e406ahrens	DMU_OT_NONE,
72fa9e406ahrens	/* general: */
73fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
74fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
75fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
76fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
77fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
78fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
79fa9e406ahrens	/* spa: */
80fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
81fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
82fa9e406ahrens	/* zil: */
83fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
84fa9e406ahrens	/* dmu: */
85fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
86fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
87fa9e406ahrens	/* dsl: */
881649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
8987e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
9087e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
91fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
921649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
93fa9e406ahrens	/* zpl: */
94fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
95da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
96fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
97fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
98fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
99893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
100fa9e406ahrens	/* zvol: */
101fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
102fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
103fa9e406ahrens	/* other; for testing only! */
104fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
105fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
106fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
107ea8dc4beschrock	/* new object types: */
108ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
10906eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
11006eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
111b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
112ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
113da6c28aamw	DMU_OT_ACL,			/* ACL */
114da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
115da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
116da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
117088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
118088f389ahrens	DMU_OT_SCRUB_QUEUE,		/* ZAP */
1191484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1201484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
121842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
122b24ab67Jeff Bonwick	DMU_OT_DDT_ZAP,			/* ZAP */
123b24ab67Jeff Bonwick	DMU_OT_DDT_STATS,		/* ZAP */
124fa9e406ahrens	DMU_OT_NUMTYPES
125fa9e406ahrens} dmu_object_type_t;
126fa9e406ahrens
127fa9e406ahrenstypedef enum dmu_objset_type {
128fa9e406ahrens	DMU_OST_NONE,
129fa9e406ahrens	DMU_OST_META,
130fa9e406ahrens	DMU_OST_ZFS,
131fa9e406ahrens	DMU_OST_ZVOL,
132fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
133fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
134fa9e406ahrens	DMU_OST_NUMTYPES
135fa9e406ahrens} dmu_objset_type_t;
136fa9e406ahrens
137fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
138fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
139fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
140fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
141fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
142da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
143fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
144fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
145fa9e406ahrens
1460b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
1470b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
148fa9e406ahrens
149fa9e406ahrens/*
150fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
151fa9e406ahrens * operation, including metadata.
152fa9e406ahrens */
153fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
154cdb0ab7maybee#define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
155fa9e406ahrens
1561484342Matthew Ahrens#define	DMU_USERUSED_OBJECT	(-1ULL)
1571484342Matthew Ahrens#define	DMU_GROUPUSED_OBJECT	(-2ULL)
158b24ab67Jeff Bonwick#define	DMU_DEADLIST_OBJECT	(-3ULL)
1591484342Matthew Ahrens
160fa9e406ahrens/*
161fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
162fa9e406ahrens */
163503ad85Matthew Ahrensint dmu_objset_hold(const char *name, void *tag, objset_t **osp);
164503ad85Matthew Ahrensint dmu_objset_own(const char *name, dmu_objset_type_t type,
165503ad85Matthew Ahrens    boolean_t readonly, void *tag, objset_t **osp);
166503ad85Matthew Ahrensvoid dmu_objset_rele(objset_t *os, void *tag);
167503ad85Matthew Ahrensvoid dmu_objset_disown(objset_t *os, void *tag);
168503ad85Matthew Ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
169503ad85Matthew Ahrens
1701934e92maybeeint dmu_objset_evict_dbufs(objset_t *os);
171ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
172ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
173ae46e4cMatthew Ahrensint dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
174ae46e4cMatthew Ahrens    uint64_t flags);
175842727cChris Kirbyint dmu_objset_destroy(const char *name, boolean_t defer);
176842727cChris Kirbyint dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
177ea2f5b9Matthew Ahrensint dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
178ea2f5b9Matthew Ahrens    boolean_t recursive);
179cdf5b4cmmusanteint dmu_objset_rename(const char *name, const char *newname,
180cdf5b4cmmusante    boolean_t recursive);
1811d452cfahrensint dmu_objset_find(char *name, int func(char *, void *), void *arg,
182fa9e406ahrens    int flags);
183fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
184fa9e406ahrens
185fa9e406ahrenstypedef struct dmu_buf {
186fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
187fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
188fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
189fa9e406ahrens	void *db_data;			/* data in buffer */
190fa9e406ahrens} dmu_buf_t;
191fa9e406ahrens
192fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
193fa9e406ahrens
194fa9e406ahrens/*
19599653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
19699653d4eschrock */
197fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
198fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
199fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
200fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
201ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
202ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
20399653d4eschrock#define	DMU_POOL_SPARES			"spares"
20499653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
20506eeb2aek#define	DMU_POOL_HISTORY		"history"
206b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
207fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
208ca45db4Chris Kirby#define	DMU_POOL_TMP_USERREFS		"tmp_userrefs"
209b24ab67Jeff Bonwick#define	DMU_POOL_DDT			"DDT-%s-%s-%s"
210b24ab67Jeff Bonwick#define	DMU_POOL_DDT_STATS		"DDT-statistics"
211fa9e406ahrens
212088f389ahrens/* 4x8 zbookmark_t */
213088f389ahrens#define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
214088f389ahrens/* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
215088f389ahrens#define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
216088f389ahrens/* 1x8 txg */
217088f389ahrens#define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
218088f389ahrens/* 1x8 txg */
219088f389ahrens#define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
220088f389ahrens/* 1x4 enum scrub_func */
221088f389ahrens#define	DMU_POOL_SCRUB_FUNC		"scrub_func"
222088f389ahrens/* 1x8 count */
223088f389ahrens#define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
224088f389ahrens
225fa9e406ahrens/*
226fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
227fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
228fa9e406ahrens *
229fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
230fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
231fa9e406ahrens * newly allocated object in this transaction).
232fa9e406ahrens *
233fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
234fa9e406ahrens *
235fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
236fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
237fa9e406ahrens *
238fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
239fa9e406ahrens */
240fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
241fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
242fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
243fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
244fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
2452bf405aMark Maybee    int blocksize, dmu_object_type_t bonustype, int bonuslen);
246fa9e406ahrens
247fa9e406ahrens/*
248fa9e406ahrens * Free an object from this objset.
249fa9e406ahrens *
250fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
251fa9e406ahrens * dmu_free(object, 0, -1, tx)).
252fa9e406ahrens *
253fa9e406ahrens * The object need not be held in the transaction.
254fa9e406ahrens *
255fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
256fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
257fa9e406ahrens * free it; it fails and returns EBUSY.
258fa9e406ahrens *
259fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
260fa9e406ahrens *
261fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
262fa9e406ahrens */
263fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
264fa9e406ahrens
265fa9e406ahrens/*
266fa9e406ahrens * Find the next allocated or free object.
267fa9e406ahrens *
268fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
2696754306ahrens * object which is allocated.  Ignore objects which have not been
2706754306ahrens * modified since txg.
271fa9e406ahrens *
272fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
273fa9e406ahrens *
274fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
275fa9e406ahrens */
2766754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
2776754306ahrens    boolean_t hole, uint64_t txg);
278fa9e406ahrens
279fa9e406ahrens/*
280fa9e406ahrens * Set the data blocksize for an object.
281fa9e406ahrens *
282fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
283fa9e406ahrens * the first block is allocated already, the new size must be greater
284fa9e406ahrens * than the current block size.  If these conditions are not met,
285fa9e406ahrens * ENOTSUP will be returned.
286fa9e406ahrens *
287fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
288fa9e406ahrens * contents, or ENOTSUP as described above.
289fa9e406ahrens */
290fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
291fa9e406ahrens    int ibs, dmu_tx_t *tx);
292fa9e406ahrens
293fa9e406ahrens/*
294fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
295fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
296fa9e406ahrens */
297fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
298fa9e406ahrens    dmu_tx_t *tx);
299fa9e406ahrens
300fa9e406ahrens/*
301fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
302fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
303fa9e406ahrens */
304fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
305fa9e406ahrens    dmu_tx_t *tx);
306fa9e406ahrens
307fa9e406ahrens/*
308b24ab67Jeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc.
30944cd46cbillm */
310b24ab67Jeff Bonwick#define	WP_NOFILL	0x1
311b24ab67Jeff Bonwick#define	WP_DMU_SYNC	0x2
312b24ab67Jeff Bonwick
313b24ab67Jeff Bonwickvoid dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp,
314b24ab67Jeff Bonwick    struct zio_prop *zp);
31544cd46cbillm/*
316fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
317fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
318fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
319fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
320fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
321fa9e406ahrens * object must be held in an assigned transaction before calling
322fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
323fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
324fa9e406ahrens */
325ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
326fa9e406ahrensint dmu_bonus_max(void);
3271934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
328fa9e406ahrens
329fa9e406ahrens/*
330fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
331fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
332fa9e406ahrens * that it will remain in memory.  You must release the hold with
333fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
334fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
335fa9e406ahrens *
336fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
337fa9e406ahrens * on the returned buffer before reading or writing the buffer's
338fa9e406ahrens * db_data.  The comments for those routines describe what particular
339fa9e406ahrens * operations are valid after calling them.
340fa9e406ahrens *
341fa9e406ahrens * The object number must be a valid, allocated object number.
342fa9e406ahrens */
343ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
344ea8dc4beschrock    void *tag, dmu_buf_t **);
345fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
346ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
347fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
348fa9e406ahrens
349fa9e406ahrens/*
350fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
351fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
352fa9e406ahrens * returned (in *dbpp).
353fa9e406ahrens *
354fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
355fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
356fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
357fa9e406ahrens * individually with dmu_buf_rele.
358fa9e406ahrens */
35913506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
36013506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
361ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
362fa9e406ahrens
363fa9e406ahrens/*
364fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
365fa9e406ahrens * been set.
366fa9e406ahrens *
367fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
368fa9e406ahrens *
369fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
370fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
371fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
372fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
373fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
374fa9e406ahrens *
375fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
376fa9e406ahrens * excised from the cache, so that you can clean up the data structure
377fa9e406ahrens * pointed to by user_ptr.
378fa9e406ahrens *
379fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
380fa9e406ahrens * objset with a given pageout func.
381fa9e406ahrens */
382fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
383fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
384fa9e406ahrens/*
385fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
386fa9e406ahrens * when hold count goes to zero.
387fa9e406ahrens */
388fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
389fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
390fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
391fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
392fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
393fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
394fa9e406ahrens
395fa9e406ahrens/*
396fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
397fa9e406ahrens */
398fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
399fa9e406ahrens
400fa9e406ahrens/*
401fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
402fa9e406ahrens *
403fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
404fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
405fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
406fa9e406ahrens */
407fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
408fa9e406ahrens
409fa9e406ahrens/*
4103d69262Sanjeev Bagewadi * Tells if the given dbuf is freeable.
4113d69262Sanjeev Bagewadi */
4123d69262Sanjeev Bagewadiboolean_t dmu_buf_freeable(dmu_buf_t *);
4133d69262Sanjeev Bagewadi
4143d69262Sanjeev Bagewadi/*
415fa9e406ahrens * You must create a transaction, then hold the objects which you will
416fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
417fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
418fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
419fa9e406ahrens * part of this transaction.  You can't modify buffers before the
420fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
421fa9e406ahrens * belong to objects which this transaction holds; you can't hold
422fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
423fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
424fa9e406ahrens * don't have to.
425fa9e406ahrens *
426fa9e406ahrens * You can abort the transaction before it has been assigned.
427fa9e406ahrens *
428fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
429fa9e406ahrens * regardless of transaction state.
430fa9e406ahrens */
431fa9e406ahrens
432fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
433fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
434fa9e406ahrens
435fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
436fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
437fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
438fa9e406ahrens    uint64_t len);
4391484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
440fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
441fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
442fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4438a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
444fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
445fa9e406ahrens
446fa9e406ahrens/*
447d20e665Ricardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called.
448d20e665Ricardo M. Correia *
449d20e665Ricardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a
450d20e665Ricardo M. Correia * callback parameter. The caller is responsible for properly allocating and
451d20e665Ricardo M. Correia * freeing it.
452d20e665Ricardo M. Correia *
453d20e665Ricardo M. Correia * When registering a callback, the transaction must be already created, but
454d20e665Ricardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not.
455d20e665Ricardo M. Correia *
456d20e665Ricardo M. Correia * The callback will be called after the transaction has been safely written
457d20e665Ricardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted.
458d20e665Ricardo M. Correia * If there is any error which prevents the transaction from being committed to
459d20e665Ricardo M. Correia * disk, the callback will be called with a value of error != 0.
460d20e665Ricardo M. Correia */
461d20e665Ricardo M. Correiatypedef void dmu_tx_callback_func_t(void *dcb_data, int error);
462d20e665Ricardo M. Correia
463d20e665Ricardo M. Correiavoid dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
464d20e665Ricardo M. Correia    void *dcb_data);
465d20e665Ricardo M. Correia
466d20e665Ricardo M. Correia/*
467fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
468fa9e406ahrens * zero, the range from offset to end-of-file is freed.
469fa9e406ahrens */
470ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
471fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
472cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
473cdb0ab7maybee	uint64_t size);
474cdb0ab7maybeeint dmu_free_object(objset_t *os, uint64_t object);
475fa9e406ahrens
476fa9e406ahrens/*
477fa9e406ahrens * Convenience functions.
478fa9e406ahrens *
479fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
480fa9e406ahrens * nonrecoverable I/O error.
481fa9e406ahrens */
4827bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
4837bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
484ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
4857bfdf01Neil Perrin	void *buf, uint32_t flags);
486fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
487fa9e406ahrens	const void *buf, dmu_tx_t *tx);
48882c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
48982c9918Tim Haley	dmu_tx_t *tx);
490feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
491feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
492feb08c6billm    dmu_tx_t *tx);
49344eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
49444eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
4952fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
4962fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
4972fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
4982fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
499fa9e406ahrens
500416e0cdekextern int zfs_prefetch_disable;
501416e0cdek
502fa9e406ahrens/*
503fa9e406ahrens * Asynchronously try to read in the data.
504fa9e406ahrens */
505fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
506fa9e406ahrens    uint64_t len);
507fa9e406ahrens
508fa9e406ahrenstypedef struct dmu_object_info {
509b24ab67Jeff Bonwick	/* All sizes are in bytes unless otherwise indicated. */
510fa9e406ahrens	uint32_t doi_data_block_size;
511fa9e406ahrens	uint32_t doi_metadata_block_size;
512fa9e406ahrens	dmu_object_type_t doi_type;
513fa9e406ahrens	dmu_object_type_t doi_bonus_type;
514b24ab67Jeff Bonwick	uint64_t doi_bonus_size;
515fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
516fa9e406ahrens	uint8_t doi_checksum;
517fa9e406ahrens	uint8_t doi_compress;
518fa9e406ahrens	uint8_t doi_pad[5];
519b24ab67Jeff Bonwick	uint64_t doi_physical_blocks_512;	/* data + metadata, 512b blks */
520b24ab67Jeff Bonwick	uint64_t doi_max_offset;
521b24ab67Jeff Bonwick	uint64_t doi_fill_count;		/* number of non-empty blocks */
522fa9e406ahrens} dmu_object_info_t;
523fa9e406ahrens
524e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
525e45ce72ahrens
526fa9e406ahrenstypedef struct dmu_object_type_info {
527e45ce72ahrens	arc_byteswap_func_t	*ot_byteswap;
528fa9e406ahrens	boolean_t		ot_metadata;
529fa9e406ahrens	char			*ot_name;
530fa9e406ahrens} dmu_object_type_info_t;
531fa9e406ahrens
532fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
533fa9e406ahrens
534fa9e406ahrens/*
535fa9e406ahrens * Get information on a DMU object.
536fa9e406ahrens *
537fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
538fa9e406ahrens *
539fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
540fa9e406ahrens */
541fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
542fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
543fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
544fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
545fa9e406ahrens    u_longlong_t *nblk512);
546fa9e406ahrens
547fa9e406ahrenstypedef struct dmu_objset_stats {
548a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
549a2eea2eahrens	uint64_t dds_creation_txg;
5503cb34c6ahrens	uint64_t dds_guid;
551fa9e406ahrens	dmu_objset_type_t dds_type;
552fa9e406ahrens	uint8_t dds_is_snapshot;
55331fd60dahrens	uint8_t dds_inconsistent;
5543cb34c6ahrens	char dds_origin[MAXNAMELEN];
555a2eea2eahrens} dmu_objset_stats_t;
556fa9e406ahrens
557a2eea2eahrens/*
558a2eea2eahrens * Get stats on a dataset.
559a2eea2eahrens */
560a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
561fa9e406ahrens
562a2eea2eahrens/*
563a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
564a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
565a2eea2eahrens */
566a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
567fa9e406ahrens
568a2eea2eahrens/*
569a2eea2eahrens * Get the space usage statistics for statvfs().
570a2eea2eahrens *
571a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
572a2eea2eahrens * availbytes is the amount of space available to this objset, taking
573a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
574a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
575a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
576a2eea2eahrens *
577a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
578a2eea2eahrens * and available.
579a2eea2eahrens */
580a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
581a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
582fa9e406ahrens
583fa9e406ahrens/*
584a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
585a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
586a2eea2eahrens * change, so there is a small probability that it will collide.)
587fa9e406ahrens */
588a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
589fa9e406ahrens
59071eb053Chris Kirby/*
59171eb053Chris Kirby * Get the [cm]time for an objset's snapshot dir
59271eb053Chris Kirby */
59371eb053Chris Kirbytimestruc_t dmu_objset_snap_cmtime(objset_t *os);
59471eb053Chris Kirby
595fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
596fa9e406ahrens
597fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
598fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
599fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
600fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
601fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
602fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
603fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
604e09fa4dNeil Perrinextern uint64_t dmu_objset_logbias(objset_t *os);
605fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
606b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
607ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
608ab04eb8timh    int maxlen, boolean_t *conflict);
60987e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
61087e5029ahrens    uint64_t *idp, uint64_t *offp);
6111484342Matthew Ahrens
6129966ca1Matthew Ahrenstypedef int objset_used_cb_t(dmu_object_type_t bonustype,
6139966ca1Matthew Ahrens    void *bonus, uint64_t *userp, uint64_t *groupp);
6141484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
6151484342Matthew Ahrens    objset_used_cb_t *cb);
616f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
617f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
618fa9e406ahrens
619fa9e406ahrens/*
620fa9e406ahrens * Return the txg number for the given assigned transaction.
621fa9e406ahrens */
62287e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
623fa9e406ahrens
624fa9e406ahrens/*
625fa9e406ahrens * Synchronous write.
626c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
627c5c6ffamaybee * provided buffer as a child of the parent zio.
628da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
629c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
630fa9e406ahrens * Note that while the data covered by this function will be on stable
631c5c6ffamaybee * storage when the write completes this new data does not become a
632fa9e406ahrens * permanent part of the file until the associated transaction commits.
633fa9e406ahrens */
634b24ab67Jeff Bonwick
635b24ab67Jeff Bonwick/*
636b24ab67Jeff Bonwick * {zfs,zvol,ztest}_get_done() args
637b24ab67Jeff Bonwick */
638b24ab67Jeff Bonwicktypedef struct zgd {
639b24ab67Jeff Bonwick	struct zilog	*zgd_zilog;
640b24ab67Jeff Bonwick	struct blkptr	*zgd_bp;
641b24ab67Jeff Bonwick	dmu_buf_t	*zgd_db;
642b24ab67Jeff Bonwick	struct rl	*zgd_rl;
643b24ab67Jeff Bonwick	void		*zgd_private;
644b24ab67Jeff Bonwick} zgd_t;
645b24ab67Jeff Bonwick
646b24ab67Jeff Bonwicktypedef void dmu_sync_cb_t(zgd_t *arg, int error);
647b24ab67Jeff Bonwickint dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
648fa9e406ahrens
649fa9e406ahrens/*
650fa9e406ahrens * Find the next hole or data block in file starting at *off
651fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
652fa9e406ahrens */
653fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
654fa9e406ahrens    uint64_t *off);
655fa9e406ahrens
656fa9e406ahrens/*
657fa9e406ahrens * Initial setup and final teardown.
658fa9e406ahrens */
659fa9e406ahrensextern void dmu_init(void);
660fa9e406ahrensextern void dmu_fini(void);
661fa9e406ahrens
662fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
663fa9e406ahrens    uint64_t object, uint64_t offset, int len);
664fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
665fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
666fa9e406ahrens
6673cb34c6ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
6683cb34c6ahrens    struct vnode *vp, offset_t *off);
6693cb34c6ahrens
6703cb34c6ahrenstypedef struct dmu_recv_cookie {
6713cb34c6ahrens	/*
6723cb34c6ahrens	 * This structure is opaque!
6733cb34c6ahrens	 *
6743cb34c6ahrens	 * If logical and real are different, we are recving the stream
6753cb34c6ahrens	 * into the "real" temporary clone, and then switching it with
6763cb34c6ahrens	 * the "logical" target.
6773cb34c6ahrens	 */
6783cb34c6ahrens	struct dsl_dataset *drc_logical_ds;
6793cb34c6ahrens	struct dsl_dataset *drc_real_ds;
6803cb34c6ahrens	struct drr_begin *drc_drrb;
6813cb34c6ahrens	char *drc_tosnap;
6823cb34c6ahrens	boolean_t drc_newfs;
6833cb34c6ahrens	boolean_t drc_force;
6843cb34c6ahrens} dmu_recv_cookie_t;
6853cb34c6ahrens
6863cb34c6ahrensint dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
687f4b94bdMatthew Ahrens    boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
6883cb34c6ahrensint dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
6893cb34c6ahrensint dmu_recv_end(dmu_recv_cookie_t *drc);
690fa9e406ahrens
691fa9e406ahrens/* CRC64 table */
692fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
693fa9e406ahrensextern uint64_t zfs_crc64_table[256];
694fa9e406ahrens
695fa9e406ahrens#ifdef	__cplusplus
696fa9e406ahrens}
697fa9e406ahrens#endif
698fa9e406ahrens
699fa9e406ahrens#endif	/* _SYS_DMU_H */
700