dmu.h revision ae46e4c775f2becc5343ff90b60a95acb79735f9
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 */
222bf405aMark Maybee * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23fa9e406ahrens * Use is subject to license terms.
24fa9e406ahrens */
26fa9e406ahrens#ifndef	_SYS_DMU_H
27fa9e406ahrens#define	_SYS_DMU_H
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 */
37fa9e406ahrens#include <sys/inttypes.h>
38fa9e406ahrens#include <sys/types.h>
39fa9e406ahrens#include <sys/param.h>
40ecd6cf8marks#include <sys/cred.h>
42fa9e406ahrens#ifdef	__cplusplus
43fa9e406ahrensextern "C" {
46fa9e406ahrensstruct uio;
4744eda4dmaybeestruct page;
48fa9e406ahrensstruct vnode;
49fa9e406ahrensstruct spa;
50fa9e406ahrensstruct zilog;
51fa9e406ahrensstruct zio;
52fa9e406ahrensstruct blkptr;
53fa9e406ahrensstruct zap_cursor;
54fa9e406ahrensstruct dsl_dataset;
55fa9e406ahrensstruct dsl_pool;
56fa9e406ahrensstruct dnode;
57fa9e406ahrensstruct drr_begin;
58fa9e406ahrensstruct drr_end;
5944cd46cbillmstruct zbookmark;
6044cd46cbillmstruct spa;
61a2eea2eahrensstruct nvlist;
62d0ad202ahrensstruct objset_impl;
632fdbea2Aleksandr Guzovskiystruct arc_buf;
65fa9e406ahrenstypedef struct objset objset_t;
66fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
67fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
69fa9e406ahrenstypedef enum dmu_object_type {
70fa9e406ahrens	DMU_OT_NONE,
71fa9e406ahrens	/* general: */
72fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
73fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
74fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
75fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
76fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
77fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
78fa9e406ahrens	/* spa: */
79fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
80fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
81fa9e406ahrens	/* zil: */
82fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
83fa9e406ahrens	/* dmu: */
84fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
85fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
86fa9e406ahrens	/* dsl: */
871649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
8887e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
8987e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
90fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
911649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
92fa9e406ahrens	/* zpl: */
93fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
94da6c28aamw	DMU_OT_OLDACL,			/* Old ACL */
95fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
96fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
97fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
98893a6d3ahrens	DMU_OT_UNLINKED_SET,		/* ZAP */
99fa9e406ahrens	/* zvol: */
100fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
101fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
102fa9e406ahrens	/* other; for testing only! */
103fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
104fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
105fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
106ea8dc4beschrock	/* new object types: */
107ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
10806eeb2aek	DMU_OT_SPA_HISTORY,		/* UINT8 */
10906eeb2aek	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
110b1b8ab3lling	DMU_OT_POOL_PROPS,		/* ZAP */
111ecd6cf8marks	DMU_OT_DSL_PERMS,		/* ZAP */
112da6c28aamw	DMU_OT_ACL,			/* ACL */
113da6c28aamw	DMU_OT_SYSACL,			/* SYSACL */
114da6c28aamw	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
115da6c28aamw	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
116088f389ahrens	DMU_OT_NEXT_CLONES,		/* ZAP */
117088f389ahrens	DMU_OT_SCRUB_QUEUE,		/* ZAP */
1181484342Matthew Ahrens	DMU_OT_USERGROUP_USED,		/* ZAP */
1191484342Matthew Ahrens	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
120842727cChris Kirby	DMU_OT_USERREFS,		/* ZAP */
121fa9e406ahrens	DMU_OT_NUMTYPES
122fa9e406ahrens} dmu_object_type_t;
124fa9e406ahrenstypedef enum dmu_objset_type {
125fa9e406ahrens	DMU_OST_NONE,
126fa9e406ahrens	DMU_OST_META,
127fa9e406ahrens	DMU_OST_ZFS,
128fa9e406ahrens	DMU_OST_ZVOL,
129fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
130fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
131fa9e406ahrens	DMU_OST_NUMTYPES
132fa9e406ahrens} dmu_objset_type_t;
134fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
135fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
136fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
137fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
138fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
139da6c28aamwvoid zfs_oldacl_byteswap(void *buf, size_t size);
140fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
141fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
143745cd3cmaybee#define	DS_MODE_NOHOLD		0	/* internal use only */
144745cd3cmaybee#define	DS_MODE_USER		1	/* simple access, no special needs */
145745cd3cmaybee#define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
146745cd3cmaybee#define	DS_MODE_TYPE_MASK	0x3
147745cd3cmaybee#define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
148fa9e406ahrens#define	DS_MODE_READONLY	0x8
149fa9e406ahrens#define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
150e193023bonwick#define	DS_MODE_INCONSISTENT	0x10
151e193023bonwick#define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
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)
1651484342Matthew Ahrens
167fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
168fa9e406ahrens */
169fa9e406ahrensint dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
170fa9e406ahrens    objset_t **osp);
1713cb34c6ahrensint dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
1723cb34c6ahrens    objset_t **osp);
173fa9e406ahrensvoid dmu_objset_close(objset_t *os);
1741934e92maybeeint dmu_objset_evict_dbufs(objset_t *os);
175ae46e4cMatthew Ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
176ecd6cf8marks    void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
177ae46e4cMatthew Ahrensint dmu_objset_clone(const char *name, struct dsl_dataset *clone_origin,
178ae46e4cMatthew Ahrens    uint64_t flags);
179842727cChris Kirbyint dmu_objset_destroy(const char *name, boolean_t defer);
180842727cChris Kirbyint dmu_snapshots_destroy(char *fsname, char *snapname, boolean_t defer);
181ea2f5b9Matthew Ahrensint dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
182ea2f5b9Matthew Ahrens    boolean_t recursive);
183cdf5b4cmmusanteint dmu_objset_rename(const char *name, const char *newname,
184cdf5b4cmmusante    boolean_t recursive);
1851d452cfahrensint dmu_objset_find(char *name, int func(char *, void *), void *arg,
186fa9e406ahrens    int flags);
187fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
189fa9e406ahrenstypedef struct dmu_buf {
190fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
191fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
192fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
193fa9e406ahrens	void *db_data;			/* data in buffer */
194fa9e406ahrens} dmu_buf_t;
196fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
19999653d4eschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
20099653d4eschrock */
201fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
202fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
203fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
204fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
205ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
206ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
20799653d4eschrock#define	DMU_POOL_SPARES			"spares"
20899653d4eschrock#define	DMU_POOL_DEFLATE		"deflate"
20906eeb2aek#define	DMU_POOL_HISTORY		"history"
210b1b8ab3lling#define	DMU_POOL_PROPS			"pool_props"
211fa94a07brendan#define	DMU_POOL_L2CACHE		"l2cache"
213088f389ahrens/* 4x8 zbookmark_t */
214088f389ahrens#define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
215088f389ahrens/* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
216088f389ahrens#define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
217088f389ahrens/* 1x8 txg */
218088f389ahrens#define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
219088f389ahrens/* 1x8 txg */
220088f389ahrens#define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
221088f389ahrens/* 1x4 enum scrub_func */
222088f389ahrens#define	DMU_POOL_SCRUB_FUNC		"scrub_func"
223088f389ahrens/* 1x8 count */
224088f389ahrens#define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
227fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
228fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
229fa9e406ahrens *
230fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
231fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
232fa9e406ahrens * newly allocated object in this transaction).
233fa9e406ahrens *
234fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
235fa9e406ahrens *
236fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
237fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
238fa9e406ahrens *
239fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
240fa9e406ahrens */
241fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
242fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
243fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
244fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
245fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
2462bf405aMark Maybee    int blocksize, dmu_object_type_t bonustype, int bonuslen);
249fa9e406ahrens * Free an object from this objset.
250fa9e406ahrens *
251fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
252fa9e406ahrens * dmu_free(object, 0, -1, tx)).
253fa9e406ahrens *
254fa9e406ahrens * The object need not be held in the transaction.
255fa9e406ahrens *
256fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
257fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
258fa9e406ahrens * free it; it fails and returns EBUSY.
259fa9e406ahrens *
260fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
261fa9e406ahrens *
262fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
263fa9e406ahrens */
264fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
267fa9e406ahrens * Find the next allocated or free object.
268fa9e406ahrens *
269fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
2706754306ahrens * object which is allocated.  Ignore objects which have not been
2716754306ahrens * modified since txg.
272fa9e406ahrens *
273fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
274fa9e406ahrens *
275fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
276fa9e406ahrens */
2776754306ahrensint dmu_object_next(objset_t *os, uint64_t *objectp,
2786754306ahrens    boolean_t hole, uint64_t txg);
281fa9e406ahrens * Set the data blocksize for an object.
282fa9e406ahrens *
283fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
284fa9e406ahrens * the first block is allocated already, the new size must be greater
285fa9e406ahrens * than the current block size.  If these conditions are not met,
286fa9e406ahrens * ENOTSUP will be returned.
287fa9e406ahrens *
288fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
289fa9e406ahrens * contents, or ENOTSUP as described above.
290fa9e406ahrens */
291fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
292fa9e406ahrens    int ibs, dmu_tx_t *tx);
295fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
296fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
297fa9e406ahrens */
298fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
299fa9e406ahrens    dmu_tx_t *tx);
302fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
303fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
304fa9e406ahrens */
305fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
306fa9e406ahrens    dmu_tx_t *tx);
30944cd46cbillm * Decide how many copies of a given block we should make.  Can be from
31044cd46cbillm * 1 to SPA_DVAS_PER_BP.
31144cd46cbillm */
312d0ad202ahrensint dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
31344cd46cbillm    dmu_object_type_t ot);
315fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
316fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
317fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
318fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
319fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
320fa9e406ahrens * object must be held in an assigned transaction before calling
321fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
322fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
323fa9e406ahrens */
324ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
325fa9e406ahrensint dmu_bonus_max(void);
3261934e92maybeeint dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
329fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
330fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
331fa9e406ahrens * that it will remain in memory.  You must release the hold with
332fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
333fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
334fa9e406ahrens *
335fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
336fa9e406ahrens * on the returned buffer before reading or writing the buffer's
337fa9e406ahrens * db_data.  The comments for those routines describe what particular
338fa9e406ahrens * operations are valid after calling them.
339fa9e406ahrens *
340fa9e406ahrens * The object number must be a valid, allocated object number.
341fa9e406ahrens */
342ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
343ea8dc4beschrock    void *tag, dmu_buf_t **);
344fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
345ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
346fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
349fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
350fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
351fa9e406ahrens * returned (in *dbpp).
352fa9e406ahrens *
353fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
354fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
355fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
356fa9e406ahrens * individually with dmu_buf_rele.
357fa9e406ahrens */
35813506d1maybeeint dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
35913506d1maybee    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
360ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
363fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
364fa9e406ahrens * been set.
365fa9e406ahrens *
366fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
367fa9e406ahrens *
368fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
369fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
370fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
371fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
372fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
373fa9e406ahrens *
374fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
375fa9e406ahrens * excised from the cache, so that you can clean up the data structure
376fa9e406ahrens * pointed to by user_ptr.
377fa9e406ahrens *
378fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
379fa9e406ahrens * objset with a given pageout func.
380fa9e406ahrens */
381fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
382fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
384fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
385fa9e406ahrens * when hold count goes to zero.
386fa9e406ahrens */
387fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
388fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
389fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
390fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
391fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
392fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
395fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
396fa9e406ahrens */
397fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
400fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
401fa9e406ahrens *
402fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
403fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
404fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
405fa9e406ahrens */
406fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
4093d69262Sanjeev Bagewadi * Tells if the given dbuf is freeable.
4103d69262Sanjeev Bagewadi */
4113d69262Sanjeev Bagewadiboolean_t dmu_buf_freeable(dmu_buf_t *);
4123d69262Sanjeev Bagewadi
4133d69262Sanjeev Bagewadi/*
414fa9e406ahrens * You must create a transaction, then hold the objects which you will
415fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
416fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
417fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
418fa9e406ahrens * part of this transaction.  You can't modify buffers before the
419fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
420fa9e406ahrens * belong to objects which this transaction holds; you can't hold
421fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
422fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
423fa9e406ahrens * don't have to.
424fa9e406ahrens *
425fa9e406ahrens * You can abort the transaction before it has been assigned.
426fa9e406ahrens *
427fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
428fa9e406ahrens * regardless of transaction state.
429fa9e406ahrens */
431fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
432fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
434fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
435fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
436fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
437fa9e406ahrens    uint64_t len);
4381484342Matthew Ahrensvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
439fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
440fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
441fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4428a2f1b9ahrensvoid dmu_tx_wait(dmu_tx_t *tx);
443fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
446fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
447fa9e406ahrens * zero, the range from offset to end-of-file is freed.
448fa9e406ahrens */
449ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
450fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
451cdb0ab7maybeeint dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
452cdb0ab7maybee	uint64_t size);
453cdb0ab7maybeeint dmu_free_object(objset_t *os, uint64_t object);
456fa9e406ahrens * Convenience functions.
457fa9e406ahrens *
458fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
459fa9e406ahrens * nonrecoverable I/O error.
460fa9e406ahrens */
4617bfdf01Neil Perrin#define	DMU_READ_PREFETCH	0 /* prefetch */
4627bfdf01Neil Perrin#define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
463ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
4647bfdf01Neil Perrin	void *buf, uint32_t flags);
465fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
466fa9e406ahrens	const void *buf, dmu_tx_t *tx);
46782c9918Tim Haleyvoid dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
46882c9918Tim Haley	dmu_tx_t *tx);
469feb08c6billmint dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
470feb08c6billmint dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
471feb08c6billm    dmu_tx_t *tx);
47244eda4dmaybeeint dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
47344eda4dmaybee    uint64_t size, struct page *pp, dmu_tx_t *tx);
4742fdbea2Aleksandr Guzovskiystruct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
4752fdbea2Aleksandr Guzovskiyvoid dmu_return_arcbuf(struct arc_buf *buf);
4762fdbea2Aleksandr Guzovskiyvoid dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
4772fdbea2Aleksandr Guzovskiy    dmu_tx_t *tx);
479416e0cdekextern int zfs_prefetch_disable;
482fa9e406ahrens * Asynchronously try to read in the data.
483fa9e406ahrens */
484fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
485fa9e406ahrens    uint64_t len);
487fa9e406ahrenstypedef struct dmu_object_info {
488fa9e406ahrens	/* All sizes are in bytes. */
489fa9e406ahrens	uint32_t doi_data_block_size;
490fa9e406ahrens	uint32_t doi_metadata_block_size;
491fa9e406ahrens	uint64_t doi_bonus_size;
492fa9e406ahrens	dmu_object_type_t doi_type;
493fa9e406ahrens	dmu_object_type_t doi_bonus_type;
494fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
495fa9e406ahrens	uint8_t doi_checksum;
496fa9e406ahrens	uint8_t doi_compress;
497fa9e406ahrens	uint8_t doi_pad[5];
498fa9e406ahrens	/* Values below are number of 512-byte blocks. */
499fa9e406ahrens	uint64_t doi_physical_blks;		/* data + metadata */
500fa9e406ahrens	uint64_t doi_max_block_offset;
501fa9e406ahrens} dmu_object_info_t;
503e45ce72ahrenstypedef void arc_byteswap_func_t(void *buf, size_t size);
505fa9e406ahrenstypedef struct dmu_object_type_info {
506e45ce72ahrens	arc_byteswap_func_t	*ot_byteswap;
507fa9e406ahrens	boolean_t		ot_metadata;
508fa9e406ahrens	char			*ot_name;
509fa9e406ahrens} dmu_object_type_info_t;
511fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
514fa9e406ahrens * Get information on a DMU object.
515fa9e406ahrens *
516fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
517fa9e406ahrens *
518fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
519fa9e406ahrens */
520fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
521fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
522fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
523fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
524fa9e406ahrens    u_longlong_t *nblk512);
526fa9e406ahrenstypedef struct dmu_objset_stats {
527a2eea2eahrens	uint64_t dds_num_clones; /* number of clones of this */
528a2eea2eahrens	uint64_t dds_creation_txg;
5293cb34c6ahrens	uint64_t dds_guid;
530fa9e406ahrens	dmu_objset_type_t dds_type;
531fa9e406ahrens	uint8_t dds_is_snapshot;
53231fd60dahrens	uint8_t dds_inconsistent;
5333cb34c6ahrens	char dds_origin[MAXNAMELEN];
534a2eea2eahrens} dmu_objset_stats_t;
537a2eea2eahrens * Get stats on a dataset.
538a2eea2eahrens */
539a2eea2eahrensvoid dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
542a2eea2eahrens * Add entries to the nvlist for all the objset's properties.  See
543a2eea2eahrens * zfs_prop_table[] and zfs(1m) for details on the properties.
544a2eea2eahrens */
545a2eea2eahrensvoid dmu_objset_stats(objset_t *os, struct nvlist *nv);
548a2eea2eahrens * Get the space usage statistics for statvfs().
549a2eea2eahrens *
550a2eea2eahrens * refdbytes is the amount of space "referenced" by this objset.
551a2eea2eahrens * availbytes is the amount of space available to this objset, taking
552a2eea2eahrens * into account quotas & reservations, assuming that no other objsets
553a2eea2eahrens * use the space first.  These values correspond to the 'referenced' and
554a2eea2eahrens * 'available' properties, described in the zfs(1m) manpage.
555a2eea2eahrens *
556a2eea2eahrens * usedobjs and availobjs are the number of objects currently allocated,
557a2eea2eahrens * and available.
558a2eea2eahrens */
559a2eea2eahrensvoid dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
560a2eea2eahrens    uint64_t *usedobjsp, uint64_t *availobjsp);
563a2eea2eahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions.
564a2eea2eahrens * (Contrast with the ds_guid which is a 64-bit ID that will never
565a2eea2eahrens * change, so there is a small probability that it will collide.)
566fa9e406ahrens */
567a2eea2eahrensuint64_t dmu_objset_fsid_guid(objset_t *os);
569fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
571fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
572fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
573fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
574fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
575fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
576fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
577fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
578fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
579b38f097ck    uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
580ab04eb8timhextern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
581ab04eb8timh    int maxlen, boolean_t *conflict);
58287e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
58387e5029ahrens    uint64_t *idp, uint64_t *offp);
5841484342Matthew Ahrens
5851484342Matthew Ahrenstypedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
5861484342Matthew Ahrens    void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
5871484342Matthew Ahrens    dmu_tx_t *tx);
5881484342Matthew Ahrensextern void dmu_objset_register_type(dmu_objset_type_t ost,
5891484342Matthew Ahrens    objset_used_cb_t *cb);
590f18faf3ekextern void dmu_objset_set_user(objset_t *os, void *user_ptr);
591f18faf3ekextern void *dmu_objset_get_user(objset_t *os);
594fa9e406ahrens * Return the txg number for the given assigned transaction.
595fa9e406ahrens */
59687e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
599fa9e406ahrens * Synchronous write.
600c5c6ffamaybee * If a parent zio is provided this function initiates a write on the
601c5c6ffamaybee * provided buffer as a child of the parent zio.
602da6c28aamw * In the absence of a parent zio, the write is completed synchronously.
603c5c6ffamaybee * At write completion, blk is filled with the bp of the written block.
604fa9e406ahrens * Note that while the data covered by this function will be on stable
605c5c6ffamaybee * storage when the write completes this new data does not become a
606fa9e406ahrens * permanent part of the file until the associated transaction commits.
607fa9e406ahrens */
608c5c6ffamaybeetypedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
609c5c6ffamaybeeint dmu_sync(struct zio *zio, dmu_buf_t *db,
610c5c6ffamaybee    struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
613fa9e406ahrens * Find the next hole or data block in file starting at *off
614fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
615fa9e406ahrens */
616fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
617fa9e406ahrens    uint64_t *off);
620fa9e406ahrens * Initial setup and final teardown.
621fa9e406ahrens */
622fa9e406ahrensextern void dmu_init(void);
623fa9e406ahrensextern void dmu_fini(void);
625fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
626fa9e406ahrens    uint64_t object, uint64_t offset, int len);
627fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
628fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
6303cb34c6ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
6313cb34c6ahrens    struct vnode *vp, offset_t *off);
6333cb34c6ahrenstypedef struct dmu_recv_cookie {
6343cb34c6ahrens	/*
6353cb34c6ahrens	 * This structure is opaque!
6363cb34c6ahrens	 *
6373cb34c6ahrens	 * If logical and real are different, we are recving the stream
6383cb34c6ahrens	 * into the "real" temporary clone, and then switching it with
6393cb34c6ahrens	 * the "logical" target.
6403cb34c6ahrens	 */
6413cb34c6ahrens	struct dsl_dataset *drc_logical_ds;
6423cb34c6ahrens	struct dsl_dataset *drc_real_ds;
6433cb34c6ahrens	struct drr_begin *drc_drrb;
6443cb34c6ahrens	char *drc_tosnap;
6453cb34c6ahrens	boolean_t drc_newfs;
6463cb34c6ahrens	boolean_t drc_force;
6473cb34c6ahrens} dmu_recv_cookie_t;
6493cb34c6ahrensint dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
650f4b94bdMatthew Ahrens    boolean_t force, objset_t *origin, dmu_recv_cookie_t *);
6513cb34c6ahrensint dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
6523cb34c6ahrensint dmu_recv_end(dmu_recv_cookie_t *drc);
654fa9e406ahrens/* CRC64 table */
655fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
656fa9e406ahrensextern uint64_t zfs_crc64_table[256];
658fa9e406ahrens#ifdef	__cplusplus
662fa9e406ahrens#endif	/* _SYS_DMU_H */