dmu.h revision 6754306ec9a89fd28806908d10c76141e8fbba3f
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/*
227f7322feschrock * Copyright 2006 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#pragma ident	"%Z%%M%	%I%	%E% SMI"
30fa9e406ahrens
31fa9e406ahrens/*
32fa9e406ahrens * This file describes the interface that the DMU provides for its
33fa9e406ahrens * consumers.
34fa9e406ahrens *
35fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
36fa9e406ahrens * dmu_spa.h.
37fa9e406ahrens */
38fa9e406ahrens
39fa9e406ahrens#include <sys/inttypes.h>
40fa9e406ahrens#include <sys/types.h>
41fa9e406ahrens#include <sys/param.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;
63fa9e406ahrens
64fa9e406ahrenstypedef struct objset objset_t;
65fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
66fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
67fa9e406ahrens
68fa9e406ahrenstypedef enum dmu_object_type {
69fa9e406ahrens	DMU_OT_NONE,
70fa9e406ahrens	/* general: */
71fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
72fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
73fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
74fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
75fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
76fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
77fa9e406ahrens	/* spa: */
78fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
79fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
80fa9e406ahrens	/* zil: */
81fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
82fa9e406ahrens	/* dmu: */
83fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
84fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
85fa9e406ahrens	/* dsl: */
861649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
8787e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
8887e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
89fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
901649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
91fa9e406ahrens	/* zpl: */
92fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
93fa9e406ahrens	DMU_OT_ACL,			/* ACL */
94fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
95fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
96fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
97fa9e406ahrens	DMU_OT_DELETE_QUEUE,		/* ZAP */
98fa9e406ahrens	/* zvol: */
99fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
100fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
101fa9e406ahrens	/* other; for testing only! */
102fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
103fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
104fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
105ea8dc4beschrock	/* new object types: */
106ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
10706eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
10806eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
109fa9e406ahrens
110fa9e406ahrens	DMU_OT_NUMTYPES
111fa9e406ahrens} dmu_object_type_t;
112fa9e406ahrens
113fa9e406ahrenstypedef enum dmu_objset_type {
114fa9e406ahrens	DMU_OST_NONE,
115fa9e406ahrens	DMU_OST_META,
116fa9e406ahrens	DMU_OST_ZFS,
117fa9e406ahrens	DMU_OST_ZVOL,
118fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
119fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
120fa9e406ahrens	DMU_OST_NUMTYPES
121fa9e406ahrens} dmu_objset_type_t;
122fa9e406ahrens
123fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
124fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
125fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
126fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
127fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
128fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
129fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
130fa9e406ahrens
131fa9e406ahrens#define	DS_MODE_NONE		0	/* invalid, to aid debugging */
132fa9e406ahrens#define	DS_MODE_STANDARD	1	/* normal access, no special needs */
133fa9e406ahrens#define	DS_MODE_PRIMARY		2	/* the "main" access, e.g. a mount */
134fa9e406ahrens#define	DS_MODE_EXCLUSIVE	3	/* exclusive access, e.g. to destroy */
135fa9e406ahrens#define	DS_MODE_LEVELS		4
136fa9e406ahrens#define	DS_MODE_LEVEL(x)	((x) & (DS_MODE_LEVELS - 1))
137fa9e406ahrens#define	DS_MODE_READONLY	0x8
138fa9e406ahrens#define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
139e193023bonwick#define	DS_MODE_INCONSISTENT	0x10
140e193023bonwick#define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
141fa9e406ahrens
1420b69c2fahrens#define	DS_FIND_SNAPSHOTS	(1<<0)
1430b69c2fahrens#define	DS_FIND_CHILDREN	(1<<1)
144fa9e406ahrens
145fa9e406ahrens/*
146fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
147fa9e406ahrens * operation, including metadata.
148fa9e406ahrens */
149fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
150fa9e406ahrens
151fa9e406ahrens/*
152fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
153fa9e406ahrens */
154fa9e406ahrensint dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
155fa9e406ahrens    objset_t **osp);
156fa9e406ahrensvoid dmu_objset_close(objset_t *os);
157436b295perrinint dmu_objset_evict_dbufs(objset_t *os, int try);
158fa9e406ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type,
159fa9e406ahrens    objset_t *clone_parent,
160fa9e406ahrens    void (*func)(objset_t *os, void *arg, dmu_tx_t *tx), void *arg);
161fa9e406ahrensint dmu_objset_destroy(const char *name);
1621d452cfahrensint dmu_snapshots_destroy(char *fsname, char *snapname);
163fa9e406ahrensint dmu_objset_rollback(const char *name);
1641d452cfahrensint dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
165fa9e406ahrensint dmu_objset_rename(const char *name, const char *newname);
1661d452cfahrensint dmu_objset_find(char *name, int func(char *, void *), void *arg,
167fa9e406ahrens    int flags);
168fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
169fa9e406ahrens
170fa9e406ahrenstypedef struct dmu_buf {
171fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
172fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
173fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
174fa9e406ahrens	void *db_data;			/* data in buffer */
175fa9e406ahrens} dmu_buf_t;
176fa9e406ahrens
177fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
178fa9e406ahrens
179fa9e406ahrens/*
180fa9e406ahrens * Callback function to perform byte swapping on a block.
181fa9e406ahrens */
182fa9e406ahrenstypedef void dmu_byteswap_func_t(void *buf, size_t size);
183fa9e406ahrens
18499653d4eschrock/*
18599653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
18699653d4eschrock */
187fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
188fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
189fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
190fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
191ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
192ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
19399653d4eschrock#define	DMU_POOL_SPARES			"spares"
19499653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
19506eeb2aek#define	DMU_POOL_HISTORY		"history"
196fa9e406ahrens
197fa9e406ahrens/*
198fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
199fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
200fa9e406ahrens *
201fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
202fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
203fa9e406ahrens * newly allocated object in this transaction).
204fa9e406ahrens *
205fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
206fa9e406ahrens *
207fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
208fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
209fa9e406ahrens *
210fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
211fa9e406ahrens */
212fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
213fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
214fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
215fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
216fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
217fa9e406ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
218fa9e406ahrens
219fa9e406ahrens/*
220fa9e406ahrens * Free an object from this objset.
221fa9e406ahrens *
222fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
223fa9e406ahrens * dmu_free(object, 0, -1, tx)).
224fa9e406ahrens *
225fa9e406ahrens * The object need not be held in the transaction.
226fa9e406ahrens *
227fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
228fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
229fa9e406ahrens * free it; it fails and returns EBUSY.
230fa9e406ahrens *
231fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
232fa9e406ahrens *
233fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
234fa9e406ahrens */
235fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
236fa9e406ahrens
237fa9e406ahrens/*
238fa9e406ahrens * Find the next allocated or free object.
239fa9e406ahrens *
240fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
2416754306ahrens * object which is allocated.  Ignore objects which have not been
2426754306ahrens * modified since txg.
243fa9e406ahrens *
244fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
245fa9e406ahrens *
246fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
247fa9e406ahrens */
2486754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
2496754306ahrens    boolean_t hole, uint64_t txg);
250fa9e406ahrens
251fa9e406ahrens/*
252fa9e406ahrens * Set the data blocksize for an object.
253fa9e406ahrens *
254fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
255fa9e406ahrens * the first block is allocated already, the new size must be greater
256fa9e406ahrens * than the current block size.  If these conditions are not met,
257fa9e406ahrens * ENOTSUP will be returned.
258fa9e406ahrens *
259fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
260fa9e406ahrens * contents, or ENOTSUP as described above.
261fa9e406ahrens */
262fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
263fa9e406ahrens    int ibs, dmu_tx_t *tx);
264fa9e406ahrens
265fa9e406ahrens/*
266fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
267fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
268fa9e406ahrens */
269fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
270fa9e406ahrens    dmu_tx_t *tx);
271fa9e406ahrens
272fa9e406ahrens/*
273fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
274fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
275fa9e406ahrens */
276fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
277fa9e406ahrens    dmu_tx_t *tx);
278fa9e406ahrens
279fa9e406ahrens/*
28044cd46cbillm * Decide how many copies of a given block we should make.  Can be from
28144cd46cbillm * 1 to SPA_DVAS_PER_BP.
28244cd46cbillm */
28344cd46cbillmint dmu_get_replication_level(struct spa *spa, struct zbookmark *zb,
28444cd46cbillm    dmu_object_type_t ot);
28544cd46cbillm/*
286fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
287fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
288fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
289fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
290fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
291fa9e406ahrens * object must be held in an assigned transaction before calling
292fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
293fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
294fa9e406ahrens */
295ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
296fa9e406ahrensint dmu_bonus_max(void);
297fa9e406ahrens
298fa9e406ahrens/*
299fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
300fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
301fa9e406ahrens * that it will remain in memory.  You must release the hold with
302fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
303fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
304fa9e406ahrens *
305fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
306fa9e406ahrens * on the returned buffer before reading or writing the buffer's
307fa9e406ahrens * db_data.  The comments for those routines describe what particular
308fa9e406ahrens * operations are valid after calling them.
309fa9e406ahrens *
310fa9e406ahrens * The object number must be a valid, allocated object number.
311fa9e406ahrens */
312ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
313ea8dc4beschrock    void *tag, dmu_buf_t **);
314fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
315ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
316fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
317fa9e406ahrens
318fa9e406ahrens/*
319fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
320fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
321fa9e406ahrens * returned (in *dbpp).
322fa9e406ahrens *
323fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
324fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
325fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
326fa9e406ahrens * individually with dmu_buf_rele.
327fa9e406ahrens */
32813506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
32913506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
330ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
331fa9e406ahrens
332fa9e406ahrens/*
333fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
334fa9e406ahrens * been set.
335fa9e406ahrens *
336fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
337fa9e406ahrens *
338fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
339fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
340fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
341fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
342fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
343fa9e406ahrens *
344fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
345fa9e406ahrens * excised from the cache, so that you can clean up the data structure
346fa9e406ahrens * pointed to by user_ptr.
347fa9e406ahrens *
348fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
349fa9e406ahrens * objset with a given pageout func.
350fa9e406ahrens */
351fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
352fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
353fa9e406ahrens/*
354fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
355fa9e406ahrens * when hold count goes to zero.
356fa9e406ahrens */
357fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
358fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
359fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
360fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
361fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
362fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
363fa9e406ahrens
364fa9e406ahrens/*
365fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
366fa9e406ahrens */
367fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
368fa9e406ahrens
369fa9e406ahrens/*
370fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
371fa9e406ahrens *
372fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
373fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
374fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
375fa9e406ahrens */
376fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
377fa9e406ahrens
378fa9e406ahrens/*
379fa9e406ahrens * You must create a transaction, then hold the objects which you will
380fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
381fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
382fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
383fa9e406ahrens * part of this transaction.  You can't modify buffers before the
384fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
385fa9e406ahrens * belong to objects which this transaction holds; you can't hold
386fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
387fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
388fa9e406ahrens * don't have to.
389fa9e406ahrens *
390fa9e406ahrens * You can abort the transaction before it has been assigned.
391fa9e406ahrens *
392fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
393fa9e406ahrens * regardless of transaction state.
394fa9e406ahrens */
395fa9e406ahrens
396fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
397fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
398fa9e406ahrens
399fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
400fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
401fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
402fa9e406ahrens    uint64_t len);
403ea8dc4beschrockvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
404fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
405fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
406fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4078a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
408fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
409fa9e406ahrens
410fa9e406ahrens/*
411fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
412fa9e406ahrens * zero, the range from offset to end-of-file is freed.
413fa9e406ahrens */
414ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
415fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
416fa9e406ahrens
417fa9e406ahrens/*
418fa9e406ahrens * Convenience functions.
419fa9e406ahrens *
420fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
421fa9e406ahrens * nonrecoverable I/O error.
422fa9e406ahrens */
423ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
424fa9e406ahrens	void *buf);
425fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
426fa9e406ahrens	const void *buf, dmu_tx_t *tx);
427fa9e406ahrensint dmu_write_uio(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
428fa9e406ahrens    struct uio *uio, dmu_tx_t *tx);
42944eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
43044eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
431fa9e406ahrens
432416e0cdekextern int zfs_prefetch_disable;
433416e0cdek
434fa9e406ahrens/*
435fa9e406ahrens * Asynchronously try to read in the data.
436fa9e406ahrens */
437fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
438fa9e406ahrens    uint64_t len);
439fa9e406ahrens
440fa9e406ahrenstypedef struct dmu_object_info {
441fa9e406ahrens	/* All sizes are in bytes. */
442fa9e406ahrens	uint32_t doi_data_block_size;
443fa9e406ahrens	uint32_t doi_metadata_block_size;
444fa9e406ahrens	uint64_t doi_bonus_size;
445fa9e406ahrens	dmu_object_type_t doi_type;
446fa9e406ahrens	dmu_object_type_t doi_bonus_type;
447fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
448fa9e406ahrens	uint8_t doi_checksum;
449fa9e406ahrens	uint8_t doi_compress;
450fa9e406ahrens	uint8_t doi_pad[5];
451fa9e406ahrens	/* Values below are number of 512-byte blocks. */
452fa9e406ahrens	uint64_t doi_physical_blks;		/* data + metadata */
453fa9e406ahrens	uint64_t doi_max_block_offset;
454fa9e406ahrens} dmu_object_info_t;
455fa9e406ahrens
456fa9e406ahrenstypedef struct dmu_object_type_info {
457fa9e406ahrens	dmu_byteswap_func_t	*ot_byteswap;
458fa9e406ahrens	boolean_t		ot_metadata;
459fa9e406ahrens	char			*ot_name;
460fa9e406ahrens} dmu_object_type_info_t;
461fa9e406ahrens
462fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
463fa9e406ahrens
464fa9e406ahrens/*
465fa9e406ahrens * Get information on a DMU object.
466fa9e406ahrens *
467fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
468fa9e406ahrens *
469fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
470fa9e406ahrens */
471fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
472fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
473fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
474fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
475fa9e406ahrens    u_longlong_t *nblk512);
476fa9e406ahrens
477fa9e406ahrenstypedef struct dmu_objset_stats {
478a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
479a2eea2eahrens	uint64_t dds_creation_txg;
480fa9e406ahrens	dmu_objset_type_t dds_type;
481fa9e406ahrens	uint8_t dds_is_snapshot;
48231fd60dahrens	uint8_t dds_inconsistent;
483fa9e406ahrens	char dds_clone_of[MAXNAMELEN];
484a2eea2eahrens} dmu_objset_stats_t;
485fa9e406ahrens
486a2eea2eahrens/*
487a2eea2eahrens * Get stats on a dataset.
488a2eea2eahrens */
489a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
490fa9e406ahrens
491a2eea2eahrens/*
492a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
493a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
494a2eea2eahrens */
495a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
496fa9e406ahrens
497a2eea2eahrens/*
498a2eea2eahrens * Get the space usage statistics for statvfs().
499a2eea2eahrens *
500a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
501a2eea2eahrens * availbytes is the amount of space available to this objset, taking
502a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
503a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
504a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
505a2eea2eahrens *
506a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
507a2eea2eahrens * and available.
508a2eea2eahrens */
509a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
510a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
511fa9e406ahrens
512fa9e406ahrens/*
513a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
514a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
515a2eea2eahrens * change, so there is a small probability that it will collide.)
516fa9e406ahrens */
517a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
518fa9e406ahrens
519fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
520fa9e406ahrens
521fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
522fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
523fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
524fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
525fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
526fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
527fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
528fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
529fa9e406ahrens    uint64_t *id, uint64_t *offp);
53087e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
53187e5029ahrens    uint64_t *idp, uint64_t *offp);
532fa9e406ahrens
533fa9e406ahrens/*
534fa9e406ahrens * Return the txg number for the given assigned transaction.
535fa9e406ahrens */
53687e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
537fa9e406ahrens
538fa9e406ahrens/*
539fa9e406ahrens * Synchronous write.
540c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
541c5c6ffamaybee * provided buffer as a child of the parent zio.
542c5c6ffamaybee * In the absense of a parent zio, the write is completed synchronously.
543c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
544fa9e406ahrens * Note that while the data covered by this function will be on stable
545c5c6ffamaybee * storage when the write completes this new data does not become a
546fa9e406ahrens * permanent part of the file until the associated transaction commits.
547fa9e406ahrens */
548c5c6ffamaybeetypedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
549c5c6ffamaybeeint dmu_sync(struct zio *zio, dmu_buf_t *db,
550c5c6ffamaybee    struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
551fa9e406ahrens
552fa9e406ahrens/*
553fa9e406ahrens * Find the next hole or data block in file starting at *off
554fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
555fa9e406ahrens */
556fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
557fa9e406ahrens    uint64_t *off);
558fa9e406ahrens
559fa9e406ahrens/*
560fa9e406ahrens * Initial setup and final teardown.
561fa9e406ahrens */
562fa9e406ahrensextern void dmu_init(void);
563fa9e406ahrensextern void dmu_fini(void);
564fa9e406ahrens
565fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
566fa9e406ahrens    uint64_t object, uint64_t offset, int len);
567fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
568fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
569fa9e406ahrens
570fa9e406ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp);
571ea8dc4beschrockint dmu_recvbackup(char *tosnap, struct drr_begin *drrb, uint64_t *sizep,
57298579b2nd    boolean_t force, struct vnode *vp, uint64_t voffset);
573fa9e406ahrens
574fa9e406ahrens/* CRC64 table */
575fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
576fa9e406ahrensextern uint64_t zfs_crc64_table[256];
577fa9e406ahrens
578fa9e406ahrens#ifdef	__cplusplus
579fa9e406ahrens}
580fa9e406ahrens#endif
581fa9e406ahrens
582fa9e406ahrens#endif	/* _SYS_DMU_H */
583