xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 7bfdf011e081684f853a3242d0296695110d9d84)
1fa9e4066Sahrens /*
2fa9e4066Sahrens  * CDDL HEADER START
3fa9e4066Sahrens  *
4fa9e4066Sahrens  * The contents of this file are subject to the terms of the
5ea8dc4b6Seschrock  * Common Development and Distribution License (the "License").
6ea8dc4b6Seschrock  * You may not use this file except in compliance with the License.
7fa9e4066Sahrens  *
8fa9e4066Sahrens  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e4066Sahrens  * or http://www.opensolaris.org/os/licensing.
10fa9e4066Sahrens  * See the License for the specific language governing permissions
11fa9e4066Sahrens  * and limitations under the License.
12fa9e4066Sahrens  *
13fa9e4066Sahrens  * When distributing Covered Code, include this CDDL HEADER in each
14fa9e4066Sahrens  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e4066Sahrens  * If applicable, add the following below this CDDL HEADER, with the
16fa9e4066Sahrens  * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e4066Sahrens  * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e4066Sahrens  *
19fa9e4066Sahrens  * CDDL HEADER END
20fa9e4066Sahrens  */
21fa9e4066Sahrens /*
222bf405a2SMark Maybee  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23fa9e4066Sahrens  * Use is subject to license terms.
24fa9e4066Sahrens  */
25fa9e4066Sahrens 
26fa9e4066Sahrens #ifndef	_SYS_DMU_H
27fa9e4066Sahrens #define	_SYS_DMU_H
28fa9e4066Sahrens 
29fa9e4066Sahrens /*
30fa9e4066Sahrens  * This file describes the interface that the DMU provides for its
31fa9e4066Sahrens  * consumers.
32fa9e4066Sahrens  *
33fa9e4066Sahrens  * The DMU also interacts with the SPA.  That interface is described in
34fa9e4066Sahrens  * dmu_spa.h.
35fa9e4066Sahrens  */
36fa9e4066Sahrens 
37fa9e4066Sahrens #include <sys/inttypes.h>
38fa9e4066Sahrens #include <sys/types.h>
39fa9e4066Sahrens #include <sys/param.h>
40ecd6cf80Smarks #include <sys/cred.h>
41fa9e4066Sahrens 
42fa9e4066Sahrens #ifdef	__cplusplus
43fa9e4066Sahrens extern "C" {
44fa9e4066Sahrens #endif
45fa9e4066Sahrens 
46fa9e4066Sahrens struct uio;
4744eda4d7Smaybee struct page;
48fa9e4066Sahrens struct vnode;
49fa9e4066Sahrens struct spa;
50fa9e4066Sahrens struct zilog;
51fa9e4066Sahrens struct zio;
52fa9e4066Sahrens struct blkptr;
53fa9e4066Sahrens struct zap_cursor;
54fa9e4066Sahrens struct dsl_dataset;
55fa9e4066Sahrens struct dsl_pool;
56fa9e4066Sahrens struct dnode;
57fa9e4066Sahrens struct drr_begin;
58fa9e4066Sahrens struct drr_end;
5944cd46caSbillm struct zbookmark;
6044cd46caSbillm struct spa;
61a2eea2e1Sahrens struct nvlist;
62d0ad202dSahrens struct objset_impl;
632fdbea25SAleksandr Guzovskiy struct arc_buf;
64fa9e4066Sahrens 
65fa9e4066Sahrens typedef struct objset objset_t;
66fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t;
67fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t;
68fa9e4066Sahrens 
69fa9e4066Sahrens typedef enum dmu_object_type {
70fa9e4066Sahrens 	DMU_OT_NONE,
71fa9e4066Sahrens 	/* general: */
72fa9e4066Sahrens 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
73fa9e4066Sahrens 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
74fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
75fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
76fa9e4066Sahrens 	DMU_OT_BPLIST,			/* UINT64 */
77fa9e4066Sahrens 	DMU_OT_BPLIST_HDR,		/* UINT64 */
78fa9e4066Sahrens 	/* spa: */
79fa9e4066Sahrens 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
80fa9e4066Sahrens 	DMU_OT_SPACE_MAP,		/* UINT64 */
81fa9e4066Sahrens 	/* zil: */
82fa9e4066Sahrens 	DMU_OT_INTENT_LOG,		/* UINT64 */
83fa9e4066Sahrens 	/* dmu: */
84fa9e4066Sahrens 	DMU_OT_DNODE,			/* DNODE */
85fa9e4066Sahrens 	DMU_OT_OBJSET,			/* OBJSET */
86fa9e4066Sahrens 	/* dsl: */
871649cd4bStabriz 	DMU_OT_DSL_DIR,			/* UINT64 */
8887e5029aSahrens 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
8987e5029aSahrens 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
90fa9e4066Sahrens 	DMU_OT_DSL_PROPS,		/* ZAP */
911649cd4bStabriz 	DMU_OT_DSL_DATASET,		/* UINT64 */
92fa9e4066Sahrens 	/* zpl: */
93fa9e4066Sahrens 	DMU_OT_ZNODE,			/* ZNODE */
94da6c28aaSamw 	DMU_OT_OLDACL,			/* Old ACL */
95fa9e4066Sahrens 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
96fa9e4066Sahrens 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
97fa9e4066Sahrens 	DMU_OT_MASTER_NODE,		/* ZAP */
98893a6d32Sahrens 	DMU_OT_UNLINKED_SET,		/* ZAP */
99fa9e4066Sahrens 	/* zvol: */
100fa9e4066Sahrens 	DMU_OT_ZVOL,			/* UINT8 */
101fa9e4066Sahrens 	DMU_OT_ZVOL_PROP,		/* ZAP */
102fa9e4066Sahrens 	/* other; for testing only! */
103fa9e4066Sahrens 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
104fa9e4066Sahrens 	DMU_OT_UINT64_OTHER,		/* UINT64 */
105fa9e4066Sahrens 	DMU_OT_ZAP_OTHER,		/* ZAP */
106ea8dc4b6Seschrock 	/* new object types: */
107ea8dc4b6Seschrock 	DMU_OT_ERROR_LOG,		/* ZAP */
10806eeb2adSek 	DMU_OT_SPA_HISTORY,		/* UINT8 */
10906eeb2adSek 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
110b1b8ab34Slling 	DMU_OT_POOL_PROPS,		/* ZAP */
111ecd6cf80Smarks 	DMU_OT_DSL_PERMS,		/* ZAP */
112da6c28aaSamw 	DMU_OT_ACL,			/* ACL */
113da6c28aaSamw 	DMU_OT_SYSACL,			/* SYSACL */
114da6c28aaSamw 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
115da6c28aaSamw 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
116088f3894Sahrens 	DMU_OT_NEXT_CLONES,		/* ZAP */
117088f3894Sahrens 	DMU_OT_SCRUB_QUEUE,		/* ZAP */
11814843421SMatthew Ahrens 	DMU_OT_USERGROUP_USED,		/* ZAP */
11914843421SMatthew Ahrens 	DMU_OT_USERGROUP_QUOTA,		/* ZAP */
120fa9e4066Sahrens 	DMU_OT_NUMTYPES
121fa9e4066Sahrens } dmu_object_type_t;
122fa9e4066Sahrens 
123fa9e4066Sahrens typedef enum dmu_objset_type {
124fa9e4066Sahrens 	DMU_OST_NONE,
125fa9e4066Sahrens 	DMU_OST_META,
126fa9e4066Sahrens 	DMU_OST_ZFS,
127fa9e4066Sahrens 	DMU_OST_ZVOL,
128fa9e4066Sahrens 	DMU_OST_OTHER,			/* For testing only! */
129fa9e4066Sahrens 	DMU_OST_ANY,			/* Be careful! */
130fa9e4066Sahrens 	DMU_OST_NUMTYPES
131fa9e4066Sahrens } dmu_objset_type_t;
132fa9e4066Sahrens 
133fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size);
134fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size);
135fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size);
136fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size);
137fa9e4066Sahrens void zap_byteswap(void *buf, size_t size);
138da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size);
139fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size);
140fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size);
141fa9e4066Sahrens 
142745cd3c5Smaybee #define	DS_MODE_NOHOLD		0	/* internal use only */
143745cd3c5Smaybee #define	DS_MODE_USER		1	/* simple access, no special needs */
144745cd3c5Smaybee #define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
145745cd3c5Smaybee #define	DS_MODE_TYPE_MASK	0x3
146745cd3c5Smaybee #define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
147fa9e4066Sahrens #define	DS_MODE_READONLY	0x8
148fa9e4066Sahrens #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
149e1930233Sbonwick #define	DS_MODE_INCONSISTENT	0x10
150e1930233Sbonwick #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
151fa9e4066Sahrens 
1520b69c2f0Sahrens #define	DS_FIND_SNAPSHOTS	(1<<0)
1530b69c2f0Sahrens #define	DS_FIND_CHILDREN	(1<<1)
154fa9e4066Sahrens 
155fa9e4066Sahrens /*
156fa9e4066Sahrens  * The maximum number of bytes that can be accessed as part of one
157fa9e4066Sahrens  * operation, including metadata.
158fa9e4066Sahrens  */
159fa9e4066Sahrens #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
160cdb0ab79Smaybee #define	DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
161fa9e4066Sahrens 
16214843421SMatthew Ahrens #define	DMU_USERUSED_OBJECT	(-1ULL)
16314843421SMatthew Ahrens #define	DMU_GROUPUSED_OBJECT	(-2ULL)
16414843421SMatthew Ahrens 
165fa9e4066Sahrens /*
166fa9e4066Sahrens  * Public routines to create, destroy, open, and close objsets.
167fa9e4066Sahrens  */
168fa9e4066Sahrens int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
169fa9e4066Sahrens     objset_t **osp);
1703cb34c60Sahrens int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
1713cb34c60Sahrens     objset_t **osp);
172fa9e4066Sahrens void dmu_objset_close(objset_t *os);
1731934e92fSmaybee int dmu_objset_evict_dbufs(objset_t *os);
174fa9e4066Sahrens int dmu_objset_create(const char *name, dmu_objset_type_t type,
175ab04eb8eStimh     objset_t *clone_parent, uint64_t flags,
176ecd6cf80Smarks     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
177fa9e4066Sahrens int dmu_objset_destroy(const char *name);
1781d452cf5Sahrens int dmu_snapshots_destroy(char *fsname, char *snapname);
1794ccbb6e7Sahrens int dmu_objset_rollback(objset_t *os);
180ea2f5b9eSMatthew Ahrens int dmu_objset_snapshot(char *fsname, char *snapname, struct nvlist *props,
181ea2f5b9eSMatthew Ahrens     boolean_t recursive);
182cdf5b4caSmmusante int dmu_objset_rename(const char *name, const char *newname,
183cdf5b4caSmmusante     boolean_t recursive);
1841d452cf5Sahrens int dmu_objset_find(char *name, int func(char *, void *), void *arg,
185fa9e4066Sahrens     int flags);
186fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size);
187fa9e4066Sahrens 
188fa9e4066Sahrens typedef struct dmu_buf {
189fa9e4066Sahrens 	uint64_t db_object;		/* object that this buffer is part of */
190fa9e4066Sahrens 	uint64_t db_offset;		/* byte offset in this object */
191fa9e4066Sahrens 	uint64_t db_size;		/* size of buffer in bytes */
192fa9e4066Sahrens 	void *db_data;			/* data in buffer */
193fa9e4066Sahrens } dmu_buf_t;
194fa9e4066Sahrens 
195fa9e4066Sahrens typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
196fa9e4066Sahrens 
19799653d4eSeschrock /*
19899653d4eSeschrock  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
19999653d4eSeschrock  */
200fa9e4066Sahrens #define	DMU_POOL_DIRECTORY_OBJECT	1
201fa9e4066Sahrens #define	DMU_POOL_CONFIG			"config"
202fa9e4066Sahrens #define	DMU_POOL_ROOT_DATASET		"root_dataset"
203fa9e4066Sahrens #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
204ea8dc4b6Seschrock #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
205ea8dc4b6Seschrock #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
20699653d4eSeschrock #define	DMU_POOL_SPARES			"spares"
20799653d4eSeschrock #define	DMU_POOL_DEFLATE		"deflate"
20806eeb2adSek #define	DMU_POOL_HISTORY		"history"
209b1b8ab34Slling #define	DMU_POOL_PROPS			"pool_props"
210fa94a07fSbrendan #define	DMU_POOL_L2CACHE		"l2cache"
211fa9e4066Sahrens 
212088f3894Sahrens /* 4x8 zbookmark_t */
213088f3894Sahrens #define	DMU_POOL_SCRUB_BOOKMARK		"scrub_bookmark"
214088f3894Sahrens /* 1x8 zap obj DMU_OT_SCRUB_QUEUE */
215088f3894Sahrens #define	DMU_POOL_SCRUB_QUEUE		"scrub_queue"
216088f3894Sahrens /* 1x8 txg */
217088f3894Sahrens #define	DMU_POOL_SCRUB_MIN_TXG		"scrub_min_txg"
218088f3894Sahrens /* 1x8 txg */
219088f3894Sahrens #define	DMU_POOL_SCRUB_MAX_TXG		"scrub_max_txg"
220088f3894Sahrens /* 1x4 enum scrub_func */
221088f3894Sahrens #define	DMU_POOL_SCRUB_FUNC		"scrub_func"
222088f3894Sahrens /* 1x8 count */
223088f3894Sahrens #define	DMU_POOL_SCRUB_ERRORS		"scrub_errors"
224088f3894Sahrens 
225fa9e4066Sahrens /*
226fa9e4066Sahrens  * Allocate an object from this objset.  The range of object numbers
227fa9e4066Sahrens  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
228fa9e4066Sahrens  *
229fa9e4066Sahrens  * The transaction must be assigned to a txg.  The newly allocated
230fa9e4066Sahrens  * object will be "held" in the transaction (ie. you can modify the
231fa9e4066Sahrens  * newly allocated object in this transaction).
232fa9e4066Sahrens  *
233fa9e4066Sahrens  * dmu_object_alloc() chooses an object and returns it in *objectp.
234fa9e4066Sahrens  *
235fa9e4066Sahrens  * dmu_object_claim() allocates a specific object number.  If that
236fa9e4066Sahrens  * number is already allocated, it fails and returns EEXIST.
237fa9e4066Sahrens  *
238fa9e4066Sahrens  * Return 0 on success, or ENOSPC or EEXIST as specified above.
239fa9e4066Sahrens  */
240fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
241fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
242fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
243fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
244fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
2452bf405a2SMark Maybee     int blocksize, dmu_object_type_t bonustype, int bonuslen);
246fa9e4066Sahrens 
247fa9e4066Sahrens /*
248fa9e4066Sahrens  * Free an object from this objset.
249fa9e4066Sahrens  *
250fa9e4066Sahrens  * The object's data will be freed as well (ie. you don't need to call
251fa9e4066Sahrens  * dmu_free(object, 0, -1, tx)).
252fa9e4066Sahrens  *
253fa9e4066Sahrens  * The object need not be held in the transaction.
254fa9e4066Sahrens  *
255fa9e4066Sahrens  * If there are any holds on this object's buffers (via dmu_buf_hold()),
256fa9e4066Sahrens  * or tx holds on the object (via dmu_tx_hold_object()), you can not
257fa9e4066Sahrens  * free it; it fails and returns EBUSY.
258fa9e4066Sahrens  *
259fa9e4066Sahrens  * If the object is not allocated, it fails and returns ENOENT.
260fa9e4066Sahrens  *
261fa9e4066Sahrens  * Return 0 on success, or EBUSY or ENOENT as specified above.
262fa9e4066Sahrens  */
263fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
264fa9e4066Sahrens 
265fa9e4066Sahrens /*
266fa9e4066Sahrens  * Find the next allocated or free object.
267fa9e4066Sahrens  *
268fa9e4066Sahrens  * The objectp parameter is in-out.  It will be updated to be the next
2696754306eSahrens  * object which is allocated.  Ignore objects which have not been
2706754306eSahrens  * modified since txg.
271fa9e4066Sahrens  *
272fa9e4066Sahrens  * XXX Can only be called on a objset with no dirty data.
273fa9e4066Sahrens  *
274fa9e4066Sahrens  * Returns 0 on success, or ENOENT if there are no more objects.
275fa9e4066Sahrens  */
2766754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp,
2776754306eSahrens     boolean_t hole, uint64_t txg);
278fa9e4066Sahrens 
279fa9e4066Sahrens /*
280fa9e4066Sahrens  * Set the data blocksize for an object.
281fa9e4066Sahrens  *
282fa9e4066Sahrens  * The object cannot have any blocks allcated beyond the first.  If
283fa9e4066Sahrens  * the first block is allocated already, the new size must be greater
284fa9e4066Sahrens  * than the current block size.  If these conditions are not met,
285fa9e4066Sahrens  * ENOTSUP will be returned.
286fa9e4066Sahrens  *
287fa9e4066Sahrens  * Returns 0 on success, or EBUSY if there are any holds on the object
288fa9e4066Sahrens  * contents, or ENOTSUP as described above.
289fa9e4066Sahrens  */
290fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
291fa9e4066Sahrens     int ibs, dmu_tx_t *tx);
292fa9e4066Sahrens 
293fa9e4066Sahrens /*
294fa9e4066Sahrens  * Set the checksum property on a dnode.  The new checksum algorithm will
295fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
296fa9e4066Sahrens  */
297fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
298fa9e4066Sahrens     dmu_tx_t *tx);
299fa9e4066Sahrens 
300fa9e4066Sahrens /*
301fa9e4066Sahrens  * Set the compress property on a dnode.  The new compression algorithm will
302fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
303fa9e4066Sahrens  */
304fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
305fa9e4066Sahrens     dmu_tx_t *tx);
306fa9e4066Sahrens 
30744cd46caSbillm /*
30844cd46caSbillm  * Decide how many copies of a given block we should make.  Can be from
30944cd46caSbillm  * 1 to SPA_DVAS_PER_BP.
31044cd46caSbillm  */
311d0ad202dSahrens int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
31244cd46caSbillm     dmu_object_type_t ot);
313fa9e4066Sahrens /*
314fa9e4066Sahrens  * The bonus data is accessed more or less like a regular buffer.
315fa9e4066Sahrens  * You must dmu_bonus_hold() to get the buffer, which will give you a
316fa9e4066Sahrens  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
317fa9e4066Sahrens  * data.  As with any normal buffer, you must call dmu_buf_read() to
318fa9e4066Sahrens  * read db_data, dmu_buf_will_dirty() before modifying it, and the
319fa9e4066Sahrens  * object must be held in an assigned transaction before calling
320fa9e4066Sahrens  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
321fa9e4066Sahrens  * buffer as well.  You must release your hold with dmu_buf_rele().
322fa9e4066Sahrens  */
323ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
324fa9e4066Sahrens int dmu_bonus_max(void);
3251934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
326fa9e4066Sahrens 
327fa9e4066Sahrens /*
328fa9e4066Sahrens  * Obtain the DMU buffer from the specified object which contains the
329fa9e4066Sahrens  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
330fa9e4066Sahrens  * that it will remain in memory.  You must release the hold with
331fa9e4066Sahrens  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
332fa9e4066Sahrens  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
333fa9e4066Sahrens  *
334fa9e4066Sahrens  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
335fa9e4066Sahrens  * on the returned buffer before reading or writing the buffer's
336fa9e4066Sahrens  * db_data.  The comments for those routines describe what particular
337fa9e4066Sahrens  * operations are valid after calling them.
338fa9e4066Sahrens  *
339fa9e4066Sahrens  * The object number must be a valid, allocated object number.
340fa9e4066Sahrens  */
341ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
342ea8dc4b6Seschrock     void *tag, dmu_buf_t **);
343fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
344ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag);
345fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db);
346fa9e4066Sahrens 
347fa9e4066Sahrens /*
348fa9e4066Sahrens  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
349fa9e4066Sahrens  * range of an object.  A pointer to an array of dmu_buf_t*'s is
350fa9e4066Sahrens  * returned (in *dbpp).
351fa9e4066Sahrens  *
352fa9e4066Sahrens  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
353fa9e4066Sahrens  * frees the array.  The hold on the array of buffers MUST be released
354fa9e4066Sahrens  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
355fa9e4066Sahrens  * individually with dmu_buf_rele.
356fa9e4066Sahrens  */
35713506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
35813506d1eSmaybee     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
359ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
360fa9e4066Sahrens 
361fa9e4066Sahrens /*
362fa9e4066Sahrens  * Returns NULL on success, or the existing user ptr if it's already
363fa9e4066Sahrens  * been set.
364fa9e4066Sahrens  *
365fa9e4066Sahrens  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
366fa9e4066Sahrens  *
367fa9e4066Sahrens  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
368fa9e4066Sahrens  * will be set to db->db_data when you are allowed to access it.  Note
369fa9e4066Sahrens  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
370fa9e4066Sahrens  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
371fa9e4066Sahrens  * *user_data_ptr_ptr will be set to the new value when it changes.
372fa9e4066Sahrens  *
373fa9e4066Sahrens  * If non-NULL, pageout func will be called when this buffer is being
374fa9e4066Sahrens  * excised from the cache, so that you can clean up the data structure
375fa9e4066Sahrens  * pointed to by user_ptr.
376fa9e4066Sahrens  *
377fa9e4066Sahrens  * dmu_evict_user() will call the pageout func for all buffers in a
378fa9e4066Sahrens  * objset with a given pageout func.
379fa9e4066Sahrens  */
380fa9e4066Sahrens void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
381fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
382fa9e4066Sahrens /*
383fa9e4066Sahrens  * set_user_ie is the same as set_user, but request immediate eviction
384fa9e4066Sahrens  * when hold count goes to zero.
385fa9e4066Sahrens  */
386fa9e4066Sahrens void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
387fa9e4066Sahrens     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
388fa9e4066Sahrens void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
389fa9e4066Sahrens     void *user_ptr, void *user_data_ptr_ptr,
390fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
391fa9e4066Sahrens void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
392fa9e4066Sahrens 
393fa9e4066Sahrens /*
394fa9e4066Sahrens  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
395fa9e4066Sahrens  */
396fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db);
397fa9e4066Sahrens 
398fa9e4066Sahrens /*
399fa9e4066Sahrens  * Indicate that you are going to modify the buffer's data (db_data).
400fa9e4066Sahrens  *
401fa9e4066Sahrens  * The transaction (tx) must be assigned to a txg (ie. you've called
402fa9e4066Sahrens  * dmu_tx_assign()).  The buffer's object must be held in the tx
403fa9e4066Sahrens  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
404fa9e4066Sahrens  */
405fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
406fa9e4066Sahrens 
407fa9e4066Sahrens /*
408fa9e4066Sahrens  * You must create a transaction, then hold the objects which you will
409fa9e4066Sahrens  * (or might) modify as part of this transaction.  Then you must assign
410fa9e4066Sahrens  * the transaction to a transaction group.  Once the transaction has
411fa9e4066Sahrens  * been assigned, you can modify buffers which belong to held objects as
412fa9e4066Sahrens  * part of this transaction.  You can't modify buffers before the
413fa9e4066Sahrens  * transaction has been assigned; you can't modify buffers which don't
414fa9e4066Sahrens  * belong to objects which this transaction holds; you can't hold
415fa9e4066Sahrens  * objects once the transaction has been assigned.  You may hold an
416fa9e4066Sahrens  * object which you are going to free (with dmu_object_free()), but you
417fa9e4066Sahrens  * don't have to.
418fa9e4066Sahrens  *
419fa9e4066Sahrens  * You can abort the transaction before it has been assigned.
420fa9e4066Sahrens  *
421fa9e4066Sahrens  * Note that you may hold buffers (with dmu_buf_hold) at any time,
422fa9e4066Sahrens  * regardless of transaction state.
423fa9e4066Sahrens  */
424fa9e4066Sahrens 
425fa9e4066Sahrens #define	DMU_NEW_OBJECT	(-1ULL)
426fa9e4066Sahrens #define	DMU_OBJECT_END	(-1ULL)
427fa9e4066Sahrens 
428fa9e4066Sahrens dmu_tx_t *dmu_tx_create(objset_t *os);
429fa9e4066Sahrens void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
430fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
431fa9e4066Sahrens     uint64_t len);
43214843421SMatthew Ahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
433fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
434fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx);
435fa9e4066Sahrens int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4368a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx);
437fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx);
438fa9e4066Sahrens 
439fa9e4066Sahrens /*
440fa9e4066Sahrens  * Free up the data blocks for a defined range of a file.  If size is
441fa9e4066Sahrens  * zero, the range from offset to end-of-file is freed.
442fa9e4066Sahrens  */
443ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
444fa9e4066Sahrens 	uint64_t size, dmu_tx_t *tx);
445cdb0ab79Smaybee int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
446cdb0ab79Smaybee 	uint64_t size);
447cdb0ab79Smaybee int dmu_free_object(objset_t *os, uint64_t object);
448fa9e4066Sahrens 
449fa9e4066Sahrens /*
450fa9e4066Sahrens  * Convenience functions.
451fa9e4066Sahrens  *
452fa9e4066Sahrens  * Canfail routines will return 0 on success, or an errno if there is a
453fa9e4066Sahrens  * nonrecoverable I/O error.
454fa9e4066Sahrens  */
455*7bfdf011SNeil Perrin #define	DMU_READ_PREFETCH	0 /* prefetch */
456*7bfdf011SNeil Perrin #define	DMU_READ_NO_PREFETCH	1 /* don't prefetch */
457ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
458*7bfdf011SNeil Perrin 	void *buf, uint32_t flags);
459fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
460fa9e4066Sahrens 	const void *buf, dmu_tx_t *tx);
46182c9918fSTim Haley void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
46282c9918fSTim Haley 	dmu_tx_t *tx);
463feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
464feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
465feb08c6bSbillm     dmu_tx_t *tx);
46644eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
46744eda4d7Smaybee     uint64_t size, struct page *pp, dmu_tx_t *tx);
4682fdbea25SAleksandr Guzovskiy struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
4692fdbea25SAleksandr Guzovskiy void dmu_return_arcbuf(struct arc_buf *buf);
4702fdbea25SAleksandr Guzovskiy void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf,
4712fdbea25SAleksandr Guzovskiy     dmu_tx_t *tx);
472fa9e4066Sahrens 
473416e0cd8Sek extern int zfs_prefetch_disable;
474416e0cd8Sek 
475fa9e4066Sahrens /*
476fa9e4066Sahrens  * Asynchronously try to read in the data.
477fa9e4066Sahrens  */
478fa9e4066Sahrens void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
479fa9e4066Sahrens     uint64_t len);
480fa9e4066Sahrens 
481fa9e4066Sahrens typedef struct dmu_object_info {
482fa9e4066Sahrens 	/* All sizes are in bytes. */
483fa9e4066Sahrens 	uint32_t doi_data_block_size;
484fa9e4066Sahrens 	uint32_t doi_metadata_block_size;
485fa9e4066Sahrens 	uint64_t doi_bonus_size;
486fa9e4066Sahrens 	dmu_object_type_t doi_type;
487fa9e4066Sahrens 	dmu_object_type_t doi_bonus_type;
488fa9e4066Sahrens 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
489fa9e4066Sahrens 	uint8_t doi_checksum;
490fa9e4066Sahrens 	uint8_t doi_compress;
491fa9e4066Sahrens 	uint8_t doi_pad[5];
492fa9e4066Sahrens 	/* Values below are number of 512-byte blocks. */
493fa9e4066Sahrens 	uint64_t doi_physical_blks;		/* data + metadata */
494fa9e4066Sahrens 	uint64_t doi_max_block_offset;
495fa9e4066Sahrens } dmu_object_info_t;
496fa9e4066Sahrens 
497e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size);
498e45ce728Sahrens 
499fa9e4066Sahrens typedef struct dmu_object_type_info {
500e45ce728Sahrens 	arc_byteswap_func_t	*ot_byteswap;
501fa9e4066Sahrens 	boolean_t		ot_metadata;
502fa9e4066Sahrens 	char			*ot_name;
503fa9e4066Sahrens } dmu_object_type_info_t;
504fa9e4066Sahrens 
505fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
506fa9e4066Sahrens 
507fa9e4066Sahrens /*
508fa9e4066Sahrens  * Get information on a DMU object.
509fa9e4066Sahrens  *
510fa9e4066Sahrens  * Return 0 on success or ENOENT if object is not allocated.
511fa9e4066Sahrens  *
512fa9e4066Sahrens  * If doi is NULL, just indicates whether the object exists.
513fa9e4066Sahrens  */
514fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
515fa9e4066Sahrens void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
516fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
517fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
518fa9e4066Sahrens     u_longlong_t *nblk512);
519fa9e4066Sahrens 
520fa9e4066Sahrens typedef struct dmu_objset_stats {
521a2eea2e1Sahrens 	uint64_t dds_num_clones; /* number of clones of this */
522a2eea2e1Sahrens 	uint64_t dds_creation_txg;
5233cb34c60Sahrens 	uint64_t dds_guid;
524fa9e4066Sahrens 	dmu_objset_type_t dds_type;
525fa9e4066Sahrens 	uint8_t dds_is_snapshot;
52631fd60d3Sahrens 	uint8_t dds_inconsistent;
5273cb34c60Sahrens 	char dds_origin[MAXNAMELEN];
528a2eea2e1Sahrens } dmu_objset_stats_t;
529fa9e4066Sahrens 
530a2eea2e1Sahrens /*
531a2eea2e1Sahrens  * Get stats on a dataset.
532a2eea2e1Sahrens  */
533a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
534fa9e4066Sahrens 
535a2eea2e1Sahrens /*
536a2eea2e1Sahrens  * Add entries to the nvlist for all the objset's properties.  See
537a2eea2e1Sahrens  * zfs_prop_table[] and zfs(1m) for details on the properties.
538a2eea2e1Sahrens  */
539a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv);
540fa9e4066Sahrens 
541a2eea2e1Sahrens /*
542a2eea2e1Sahrens  * Get the space usage statistics for statvfs().
543a2eea2e1Sahrens  *
544a2eea2e1Sahrens  * refdbytes is the amount of space "referenced" by this objset.
545a2eea2e1Sahrens  * availbytes is the amount of space available to this objset, taking
546a2eea2e1Sahrens  * into account quotas & reservations, assuming that no other objsets
547a2eea2e1Sahrens  * use the space first.  These values correspond to the 'referenced' and
548a2eea2e1Sahrens  * 'available' properties, described in the zfs(1m) manpage.
549a2eea2e1Sahrens  *
550a2eea2e1Sahrens  * usedobjs and availobjs are the number of objects currently allocated,
551a2eea2e1Sahrens  * and available.
552a2eea2e1Sahrens  */
553a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
554a2eea2e1Sahrens     uint64_t *usedobjsp, uint64_t *availobjsp);
555fa9e4066Sahrens 
556fa9e4066Sahrens /*
557a2eea2e1Sahrens  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
558a2eea2e1Sahrens  * (Contrast with the ds_guid which is a 64-bit ID that will never
559a2eea2e1Sahrens  * change, so there is a small probability that it will collide.)
560fa9e4066Sahrens  */
561a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os);
562fa9e4066Sahrens 
563fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os);
564fa9e4066Sahrens 
565fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os);
566fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os);
567fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os);
568fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
569fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf);
570fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os);
571fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os);
572fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
573b38f0970Sck     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
574ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
575ab04eb8eStimh     int maxlen, boolean_t *conflict);
57687e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
57787e5029aSahrens     uint64_t *idp, uint64_t *offp);
57814843421SMatthew Ahrens 
57914843421SMatthew Ahrens typedef void objset_used_cb_t(objset_t *os, dmu_object_type_t bonustype,
58014843421SMatthew Ahrens     void *oldbonus, void *newbonus, uint64_t oldused, uint64_t newused,
58114843421SMatthew Ahrens     dmu_tx_t *tx);
58214843421SMatthew Ahrens extern void dmu_objset_register_type(dmu_objset_type_t ost,
58314843421SMatthew Ahrens     objset_used_cb_t *cb);
584f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
585f18faf3fSek extern void *dmu_objset_get_user(objset_t *os);
586fa9e4066Sahrens 
587fa9e4066Sahrens /*
588fa9e4066Sahrens  * Return the txg number for the given assigned transaction.
589fa9e4066Sahrens  */
59087e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
591fa9e4066Sahrens 
592fa9e4066Sahrens /*
593fa9e4066Sahrens  * Synchronous write.
594c5c6ffa0Smaybee  * If a parent zio is provided this function initiates a write on the
595c5c6ffa0Smaybee  * provided buffer as a child of the parent zio.
596da6c28aaSamw  * In the absence of a parent zio, the write is completed synchronously.
597c5c6ffa0Smaybee  * At write completion, blk is filled with the bp of the written block.
598fa9e4066Sahrens  * Note that while the data covered by this function will be on stable
599c5c6ffa0Smaybee  * storage when the write completes this new data does not become a
600fa9e4066Sahrens  * permanent part of the file until the associated transaction commits.
601fa9e4066Sahrens  */
602c5c6ffa0Smaybee typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
603c5c6ffa0Smaybee int dmu_sync(struct zio *zio, dmu_buf_t *db,
604c5c6ffa0Smaybee     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
605fa9e4066Sahrens 
606fa9e4066Sahrens /*
607fa9e4066Sahrens  * Find the next hole or data block in file starting at *off
608fa9e4066Sahrens  * Return found offset in *off. Return ESRCH for end of file.
609fa9e4066Sahrens  */
610fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
611fa9e4066Sahrens     uint64_t *off);
612fa9e4066Sahrens 
613fa9e4066Sahrens /*
614fa9e4066Sahrens  * Initial setup and final teardown.
615fa9e4066Sahrens  */
616fa9e4066Sahrens extern void dmu_init(void);
617fa9e4066Sahrens extern void dmu_fini(void);
618fa9e4066Sahrens 
619fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
620fa9e4066Sahrens     uint64_t object, uint64_t offset, int len);
621fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
622fa9e4066Sahrens     dmu_traverse_cb_t cb, void *arg);
623fa9e4066Sahrens 
6243cb34c60Sahrens int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
6253cb34c60Sahrens     struct vnode *vp, offset_t *off);
6263cb34c60Sahrens 
6273cb34c60Sahrens typedef struct dmu_recv_cookie {
6283cb34c60Sahrens 	/*
6293cb34c60Sahrens 	 * This structure is opaque!
6303cb34c60Sahrens 	 *
6313cb34c60Sahrens 	 * If logical and real are different, we are recving the stream
6323cb34c60Sahrens 	 * into the "real" temporary clone, and then switching it with
6333cb34c60Sahrens 	 * the "logical" target.
6343cb34c60Sahrens 	 */
6353cb34c60Sahrens 	struct dsl_dataset *drc_logical_ds;
6363cb34c60Sahrens 	struct dsl_dataset *drc_real_ds;
6373cb34c60Sahrens 	struct drr_begin *drc_drrb;
6383cb34c60Sahrens 	char *drc_tosnap;
6393cb34c60Sahrens 	boolean_t drc_newfs;
6403cb34c60Sahrens 	boolean_t drc_force;
6413cb34c60Sahrens } dmu_recv_cookie_t;
6423cb34c60Sahrens 
6433cb34c60Sahrens int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
6443cb34c60Sahrens     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
6453cb34c60Sahrens int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
6463cb34c60Sahrens int dmu_recv_end(dmu_recv_cookie_t *drc);
64747f263f4Sek void dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
648fa9e4066Sahrens 
649fa9e4066Sahrens /* CRC64 table */
650fa9e4066Sahrens #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
651fa9e4066Sahrens extern uint64_t zfs_crc64_table[256];
652fa9e4066Sahrens 
653fa9e4066Sahrens #ifdef	__cplusplus
654fa9e4066Sahrens }
655fa9e4066Sahrens #endif
656fa9e4066Sahrens 
657fa9e4066Sahrens #endif	/* _SYS_DMU_H */
658