xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 745cd3c5371d020efae7a911c58c526aa1fd0dba)
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 /*
2247f263f4Sek  * Copyright 2008 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 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30fa9e4066Sahrens 
31fa9e4066Sahrens /*
32fa9e4066Sahrens  * This file describes the interface that the DMU provides for its
33fa9e4066Sahrens  * consumers.
34fa9e4066Sahrens  *
35fa9e4066Sahrens  * The DMU also interacts with the SPA.  That interface is described in
36fa9e4066Sahrens  * dmu_spa.h.
37fa9e4066Sahrens  */
38fa9e4066Sahrens 
39fa9e4066Sahrens #include <sys/inttypes.h>
40fa9e4066Sahrens #include <sys/types.h>
41fa9e4066Sahrens #include <sys/param.h>
42ecd6cf80Smarks #include <sys/cred.h>
43fa9e4066Sahrens 
44fa9e4066Sahrens #ifdef	__cplusplus
45fa9e4066Sahrens extern "C" {
46fa9e4066Sahrens #endif
47fa9e4066Sahrens 
48fa9e4066Sahrens struct uio;
4944eda4d7Smaybee struct page;
50fa9e4066Sahrens struct vnode;
51fa9e4066Sahrens struct spa;
52fa9e4066Sahrens struct zilog;
53fa9e4066Sahrens struct zio;
54fa9e4066Sahrens struct blkptr;
55fa9e4066Sahrens struct zap_cursor;
56fa9e4066Sahrens struct dsl_dataset;
57fa9e4066Sahrens struct dsl_pool;
58fa9e4066Sahrens struct dnode;
59fa9e4066Sahrens struct drr_begin;
60fa9e4066Sahrens struct drr_end;
6144cd46caSbillm struct zbookmark;
6244cd46caSbillm struct spa;
63a2eea2e1Sahrens struct nvlist;
64d0ad202dSahrens struct objset_impl;
65fa9e4066Sahrens 
66fa9e4066Sahrens typedef struct objset objset_t;
67fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t;
68fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t;
69fa9e4066Sahrens 
70fa9e4066Sahrens typedef enum dmu_object_type {
71fa9e4066Sahrens 	DMU_OT_NONE,
72fa9e4066Sahrens 	/* general: */
73fa9e4066Sahrens 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
74fa9e4066Sahrens 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
75fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
76fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
77fa9e4066Sahrens 	DMU_OT_BPLIST,			/* UINT64 */
78fa9e4066Sahrens 	DMU_OT_BPLIST_HDR,		/* UINT64 */
79fa9e4066Sahrens 	/* spa: */
80fa9e4066Sahrens 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
81fa9e4066Sahrens 	DMU_OT_SPACE_MAP,		/* UINT64 */
82fa9e4066Sahrens 	/* zil: */
83fa9e4066Sahrens 	DMU_OT_INTENT_LOG,		/* UINT64 */
84fa9e4066Sahrens 	/* dmu: */
85fa9e4066Sahrens 	DMU_OT_DNODE,			/* DNODE */
86fa9e4066Sahrens 	DMU_OT_OBJSET,			/* OBJSET */
87fa9e4066Sahrens 	/* dsl: */
881649cd4bStabriz 	DMU_OT_DSL_DIR,			/* UINT64 */
8987e5029aSahrens 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
9087e5029aSahrens 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
91fa9e4066Sahrens 	DMU_OT_DSL_PROPS,		/* ZAP */
921649cd4bStabriz 	DMU_OT_DSL_DATASET,		/* UINT64 */
93fa9e4066Sahrens 	/* zpl: */
94fa9e4066Sahrens 	DMU_OT_ZNODE,			/* ZNODE */
95da6c28aaSamw 	DMU_OT_OLDACL,			/* Old ACL */
96fa9e4066Sahrens 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
97fa9e4066Sahrens 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
98fa9e4066Sahrens 	DMU_OT_MASTER_NODE,		/* ZAP */
99893a6d32Sahrens 	DMU_OT_UNLINKED_SET,		/* ZAP */
100fa9e4066Sahrens 	/* zvol: */
101fa9e4066Sahrens 	DMU_OT_ZVOL,			/* UINT8 */
102fa9e4066Sahrens 	DMU_OT_ZVOL_PROP,		/* ZAP */
103fa9e4066Sahrens 	/* other; for testing only! */
104fa9e4066Sahrens 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
105fa9e4066Sahrens 	DMU_OT_UINT64_OTHER,		/* UINT64 */
106fa9e4066Sahrens 	DMU_OT_ZAP_OTHER,		/* ZAP */
107ea8dc4b6Seschrock 	/* new object types: */
108ea8dc4b6Seschrock 	DMU_OT_ERROR_LOG,		/* ZAP */
10906eeb2adSek 	DMU_OT_SPA_HISTORY,		/* UINT8 */
11006eeb2adSek 	DMU_OT_SPA_HISTORY_OFFSETS,	/* spa_his_phys_t */
111b1b8ab34Slling 	DMU_OT_POOL_PROPS,		/* ZAP */
112ecd6cf80Smarks 	DMU_OT_DSL_PERMS,		/* ZAP */
113da6c28aaSamw 	DMU_OT_ACL,			/* ACL */
114da6c28aaSamw 	DMU_OT_SYSACL,			/* SYSACL */
115da6c28aaSamw 	DMU_OT_FUID,			/* FUID table (Packed NVLIST UINT8) */
116da6c28aaSamw 	DMU_OT_FUID_SIZE,		/* FUID table size UINT64 */
117fa9e4066Sahrens 	DMU_OT_NUMTYPES
118fa9e4066Sahrens } dmu_object_type_t;
119fa9e4066Sahrens 
120fa9e4066Sahrens typedef enum dmu_objset_type {
121fa9e4066Sahrens 	DMU_OST_NONE,
122fa9e4066Sahrens 	DMU_OST_META,
123fa9e4066Sahrens 	DMU_OST_ZFS,
124fa9e4066Sahrens 	DMU_OST_ZVOL,
125fa9e4066Sahrens 	DMU_OST_OTHER,			/* For testing only! */
126fa9e4066Sahrens 	DMU_OST_ANY,			/* Be careful! */
127fa9e4066Sahrens 	DMU_OST_NUMTYPES
128fa9e4066Sahrens } dmu_objset_type_t;
129fa9e4066Sahrens 
130fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size);
131fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size);
132fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size);
133fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size);
134fa9e4066Sahrens void zap_byteswap(void *buf, size_t size);
135da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size);
136fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size);
137fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size);
138fa9e4066Sahrens 
139*745cd3c5Smaybee #define	DS_MODE_NOHOLD		0	/* internal use only */
140*745cd3c5Smaybee #define	DS_MODE_USER		1	/* simple access, no special needs */
141*745cd3c5Smaybee #define	DS_MODE_OWNER		2	/* the "main" access, e.g. a mount */
142*745cd3c5Smaybee #define	DS_MODE_TYPE_MASK	0x3
143*745cd3c5Smaybee #define	DS_MODE_TYPE(x)		((x) & DS_MODE_TYPE_MASK)
144fa9e4066Sahrens #define	DS_MODE_READONLY	0x8
145fa9e4066Sahrens #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
146e1930233Sbonwick #define	DS_MODE_INCONSISTENT	0x10
147e1930233Sbonwick #define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
148fa9e4066Sahrens 
1490b69c2f0Sahrens #define	DS_FIND_SNAPSHOTS	(1<<0)
1500b69c2f0Sahrens #define	DS_FIND_CHILDREN	(1<<1)
151fa9e4066Sahrens 
152fa9e4066Sahrens /*
153fa9e4066Sahrens  * The maximum number of bytes that can be accessed as part of one
154fa9e4066Sahrens  * operation, including metadata.
155fa9e4066Sahrens  */
156fa9e4066Sahrens #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
157fa9e4066Sahrens 
158fa9e4066Sahrens /*
159fa9e4066Sahrens  * Public routines to create, destroy, open, and close objsets.
160fa9e4066Sahrens  */
161fa9e4066Sahrens int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
162fa9e4066Sahrens     objset_t **osp);
1633cb34c60Sahrens int dmu_objset_open_ds(struct dsl_dataset *ds, dmu_objset_type_t type,
1643cb34c60Sahrens     objset_t **osp);
165fa9e4066Sahrens void dmu_objset_close(objset_t *os);
1661934e92fSmaybee int dmu_objset_evict_dbufs(objset_t *os);
167fa9e4066Sahrens int dmu_objset_create(const char *name, dmu_objset_type_t type,
168ab04eb8eStimh     objset_t *clone_parent, uint64_t flags,
169ecd6cf80Smarks     void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg);
170fa9e4066Sahrens int dmu_objset_destroy(const char *name);
1711d452cf5Sahrens int dmu_snapshots_destroy(char *fsname, char *snapname);
1724ccbb6e7Sahrens int dmu_objset_rollback(objset_t *os);
1731d452cf5Sahrens int dmu_objset_snapshot(char *fsname, char *snapname, boolean_t recursive);
174cdf5b4caSmmusante int dmu_objset_rename(const char *name, const char *newname,
175cdf5b4caSmmusante     boolean_t recursive);
1761d452cf5Sahrens int dmu_objset_find(char *name, int func(char *, void *), void *arg,
177fa9e4066Sahrens     int flags);
178fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size);
179fa9e4066Sahrens 
180fa9e4066Sahrens typedef struct dmu_buf {
181fa9e4066Sahrens 	uint64_t db_object;		/* object that this buffer is part of */
182fa9e4066Sahrens 	uint64_t db_offset;		/* byte offset in this object */
183fa9e4066Sahrens 	uint64_t db_size;		/* size of buffer in bytes */
184fa9e4066Sahrens 	void *db_data;			/* data in buffer */
185fa9e4066Sahrens } dmu_buf_t;
186fa9e4066Sahrens 
187fa9e4066Sahrens typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
188fa9e4066Sahrens 
18999653d4eSeschrock /*
19099653d4eSeschrock  * The names of zap entries in the DIRECTORY_OBJECT of the MOS.
19199653d4eSeschrock  */
192fa9e4066Sahrens #define	DMU_POOL_DIRECTORY_OBJECT	1
193fa9e4066Sahrens #define	DMU_POOL_CONFIG			"config"
194fa9e4066Sahrens #define	DMU_POOL_ROOT_DATASET		"root_dataset"
195fa9e4066Sahrens #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
196ea8dc4b6Seschrock #define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
197ea8dc4b6Seschrock #define	DMU_POOL_ERRLOG_LAST		"errlog_last"
19899653d4eSeschrock #define	DMU_POOL_SPARES			"spares"
19999653d4eSeschrock #define	DMU_POOL_DEFLATE		"deflate"
20006eeb2adSek #define	DMU_POOL_HISTORY		"history"
201b1b8ab34Slling #define	DMU_POOL_PROPS			"pool_props"
202fa94a07fSbrendan #define	DMU_POOL_L2CACHE		"l2cache"
203fa9e4066Sahrens 
204fa9e4066Sahrens /*
205fa9e4066Sahrens  * Allocate an object from this objset.  The range of object numbers
206fa9e4066Sahrens  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
207fa9e4066Sahrens  *
208fa9e4066Sahrens  * The transaction must be assigned to a txg.  The newly allocated
209fa9e4066Sahrens  * object will be "held" in the transaction (ie. you can modify the
210fa9e4066Sahrens  * newly allocated object in this transaction).
211fa9e4066Sahrens  *
212fa9e4066Sahrens  * dmu_object_alloc() chooses an object and returns it in *objectp.
213fa9e4066Sahrens  *
214fa9e4066Sahrens  * dmu_object_claim() allocates a specific object number.  If that
215fa9e4066Sahrens  * number is already allocated, it fails and returns EEXIST.
216fa9e4066Sahrens  *
217fa9e4066Sahrens  * Return 0 on success, or ENOSPC or EEXIST as specified above.
218fa9e4066Sahrens  */
219fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
220fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
221fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
222fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
223fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
224fa9e4066Sahrens     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
225fa9e4066Sahrens 
226fa9e4066Sahrens /*
227fa9e4066Sahrens  * Free an object from this objset.
228fa9e4066Sahrens  *
229fa9e4066Sahrens  * The object's data will be freed as well (ie. you don't need to call
230fa9e4066Sahrens  * dmu_free(object, 0, -1, tx)).
231fa9e4066Sahrens  *
232fa9e4066Sahrens  * The object need not be held in the transaction.
233fa9e4066Sahrens  *
234fa9e4066Sahrens  * If there are any holds on this object's buffers (via dmu_buf_hold()),
235fa9e4066Sahrens  * or tx holds on the object (via dmu_tx_hold_object()), you can not
236fa9e4066Sahrens  * free it; it fails and returns EBUSY.
237fa9e4066Sahrens  *
238fa9e4066Sahrens  * If the object is not allocated, it fails and returns ENOENT.
239fa9e4066Sahrens  *
240fa9e4066Sahrens  * Return 0 on success, or EBUSY or ENOENT as specified above.
241fa9e4066Sahrens  */
242fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
243fa9e4066Sahrens 
244fa9e4066Sahrens /*
245fa9e4066Sahrens  * Find the next allocated or free object.
246fa9e4066Sahrens  *
247fa9e4066Sahrens  * The objectp parameter is in-out.  It will be updated to be the next
2486754306eSahrens  * object which is allocated.  Ignore objects which have not been
2496754306eSahrens  * modified since txg.
250fa9e4066Sahrens  *
251fa9e4066Sahrens  * XXX Can only be called on a objset with no dirty data.
252fa9e4066Sahrens  *
253fa9e4066Sahrens  * Returns 0 on success, or ENOENT if there are no more objects.
254fa9e4066Sahrens  */
2556754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp,
2566754306eSahrens     boolean_t hole, uint64_t txg);
257fa9e4066Sahrens 
258fa9e4066Sahrens /*
259fa9e4066Sahrens  * Set the data blocksize for an object.
260fa9e4066Sahrens  *
261fa9e4066Sahrens  * The object cannot have any blocks allcated beyond the first.  If
262fa9e4066Sahrens  * the first block is allocated already, the new size must be greater
263fa9e4066Sahrens  * than the current block size.  If these conditions are not met,
264fa9e4066Sahrens  * ENOTSUP will be returned.
265fa9e4066Sahrens  *
266fa9e4066Sahrens  * Returns 0 on success, or EBUSY if there are any holds on the object
267fa9e4066Sahrens  * contents, or ENOTSUP as described above.
268fa9e4066Sahrens  */
269fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
270fa9e4066Sahrens     int ibs, dmu_tx_t *tx);
271fa9e4066Sahrens 
272fa9e4066Sahrens /*
273fa9e4066Sahrens  * Set the checksum property on a dnode.  The new checksum algorithm will
274fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
275fa9e4066Sahrens  */
276fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
277fa9e4066Sahrens     dmu_tx_t *tx);
278fa9e4066Sahrens 
279fa9e4066Sahrens /*
280fa9e4066Sahrens  * Set the compress property on a dnode.  The new compression algorithm will
281fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
282fa9e4066Sahrens  */
283fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
284fa9e4066Sahrens     dmu_tx_t *tx);
285fa9e4066Sahrens 
28644cd46caSbillm /*
28744cd46caSbillm  * Decide how many copies of a given block we should make.  Can be from
28844cd46caSbillm  * 1 to SPA_DVAS_PER_BP.
28944cd46caSbillm  */
290d0ad202dSahrens int dmu_get_replication_level(struct objset_impl *, struct zbookmark *zb,
29144cd46caSbillm     dmu_object_type_t ot);
292fa9e4066Sahrens /*
293fa9e4066Sahrens  * The bonus data is accessed more or less like a regular buffer.
294fa9e4066Sahrens  * You must dmu_bonus_hold() to get the buffer, which will give you a
295fa9e4066Sahrens  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
296fa9e4066Sahrens  * data.  As with any normal buffer, you must call dmu_buf_read() to
297fa9e4066Sahrens  * read db_data, dmu_buf_will_dirty() before modifying it, and the
298fa9e4066Sahrens  * object must be held in an assigned transaction before calling
299fa9e4066Sahrens  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
300fa9e4066Sahrens  * buffer as well.  You must release your hold with dmu_buf_rele().
301fa9e4066Sahrens  */
302ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
303fa9e4066Sahrens int dmu_bonus_max(void);
3041934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
305fa9e4066Sahrens 
306fa9e4066Sahrens /*
307fa9e4066Sahrens  * Obtain the DMU buffer from the specified object which contains the
308fa9e4066Sahrens  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
309fa9e4066Sahrens  * that it will remain in memory.  You must release the hold with
310fa9e4066Sahrens  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
311fa9e4066Sahrens  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
312fa9e4066Sahrens  *
313fa9e4066Sahrens  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
314fa9e4066Sahrens  * on the returned buffer before reading or writing the buffer's
315fa9e4066Sahrens  * db_data.  The comments for those routines describe what particular
316fa9e4066Sahrens  * operations are valid after calling them.
317fa9e4066Sahrens  *
318fa9e4066Sahrens  * The object number must be a valid, allocated object number.
319fa9e4066Sahrens  */
320ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
321ea8dc4b6Seschrock     void *tag, dmu_buf_t **);
322fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
323ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag);
324fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db);
325fa9e4066Sahrens 
326fa9e4066Sahrens /*
327fa9e4066Sahrens  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
328fa9e4066Sahrens  * range of an object.  A pointer to an array of dmu_buf_t*'s is
329fa9e4066Sahrens  * returned (in *dbpp).
330fa9e4066Sahrens  *
331fa9e4066Sahrens  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
332fa9e4066Sahrens  * frees the array.  The hold on the array of buffers MUST be released
333fa9e4066Sahrens  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
334fa9e4066Sahrens  * individually with dmu_buf_rele.
335fa9e4066Sahrens  */
33613506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
33713506d1eSmaybee     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
338ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
339fa9e4066Sahrens 
340fa9e4066Sahrens /*
341fa9e4066Sahrens  * Returns NULL on success, or the existing user ptr if it's already
342fa9e4066Sahrens  * been set.
343fa9e4066Sahrens  *
344fa9e4066Sahrens  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
345fa9e4066Sahrens  *
346fa9e4066Sahrens  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
347fa9e4066Sahrens  * will be set to db->db_data when you are allowed to access it.  Note
348fa9e4066Sahrens  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
349fa9e4066Sahrens  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
350fa9e4066Sahrens  * *user_data_ptr_ptr will be set to the new value when it changes.
351fa9e4066Sahrens  *
352fa9e4066Sahrens  * If non-NULL, pageout func will be called when this buffer is being
353fa9e4066Sahrens  * excised from the cache, so that you can clean up the data structure
354fa9e4066Sahrens  * pointed to by user_ptr.
355fa9e4066Sahrens  *
356fa9e4066Sahrens  * dmu_evict_user() will call the pageout func for all buffers in a
357fa9e4066Sahrens  * objset with a given pageout func.
358fa9e4066Sahrens  */
359fa9e4066Sahrens void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
360fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
361fa9e4066Sahrens /*
362fa9e4066Sahrens  * set_user_ie is the same as set_user, but request immediate eviction
363fa9e4066Sahrens  * when hold count goes to zero.
364fa9e4066Sahrens  */
365fa9e4066Sahrens void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
366fa9e4066Sahrens     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
367fa9e4066Sahrens void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
368fa9e4066Sahrens     void *user_ptr, void *user_data_ptr_ptr,
369fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
370fa9e4066Sahrens void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
371fa9e4066Sahrens 
372fa9e4066Sahrens /*
373fa9e4066Sahrens  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
374fa9e4066Sahrens  */
375fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db);
376fa9e4066Sahrens 
377fa9e4066Sahrens /*
378fa9e4066Sahrens  * Indicate that you are going to modify the buffer's data (db_data).
379fa9e4066Sahrens  *
380fa9e4066Sahrens  * The transaction (tx) must be assigned to a txg (ie. you've called
381fa9e4066Sahrens  * dmu_tx_assign()).  The buffer's object must be held in the tx
382fa9e4066Sahrens  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
383fa9e4066Sahrens  */
384fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
385fa9e4066Sahrens 
386fa9e4066Sahrens /*
387fa9e4066Sahrens  * You must create a transaction, then hold the objects which you will
388fa9e4066Sahrens  * (or might) modify as part of this transaction.  Then you must assign
389fa9e4066Sahrens  * the transaction to a transaction group.  Once the transaction has
390fa9e4066Sahrens  * been assigned, you can modify buffers which belong to held objects as
391fa9e4066Sahrens  * part of this transaction.  You can't modify buffers before the
392fa9e4066Sahrens  * transaction has been assigned; you can't modify buffers which don't
393fa9e4066Sahrens  * belong to objects which this transaction holds; you can't hold
394fa9e4066Sahrens  * objects once the transaction has been assigned.  You may hold an
395fa9e4066Sahrens  * object which you are going to free (with dmu_object_free()), but you
396fa9e4066Sahrens  * don't have to.
397fa9e4066Sahrens  *
398fa9e4066Sahrens  * You can abort the transaction before it has been assigned.
399fa9e4066Sahrens  *
400fa9e4066Sahrens  * Note that you may hold buffers (with dmu_buf_hold) at any time,
401fa9e4066Sahrens  * regardless of transaction state.
402fa9e4066Sahrens  */
403fa9e4066Sahrens 
404fa9e4066Sahrens #define	DMU_NEW_OBJECT	(-1ULL)
405fa9e4066Sahrens #define	DMU_OBJECT_END	(-1ULL)
406fa9e4066Sahrens 
407fa9e4066Sahrens dmu_tx_t *dmu_tx_create(objset_t *os);
408fa9e4066Sahrens void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
409fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
410fa9e4066Sahrens     uint64_t len);
411ea8dc4b6Seschrock void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
412fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
413fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx);
414fa9e4066Sahrens int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
4158a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx);
416fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx);
417fa9e4066Sahrens 
418fa9e4066Sahrens /*
419fa9e4066Sahrens  * Free up the data blocks for a defined range of a file.  If size is
420fa9e4066Sahrens  * zero, the range from offset to end-of-file is freed.
421fa9e4066Sahrens  */
422ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
423fa9e4066Sahrens 	uint64_t size, dmu_tx_t *tx);
424fa9e4066Sahrens 
425fa9e4066Sahrens /*
426fa9e4066Sahrens  * Convenience functions.
427fa9e4066Sahrens  *
428fa9e4066Sahrens  * Canfail routines will return 0 on success, or an errno if there is a
429fa9e4066Sahrens  * nonrecoverable I/O error.
430fa9e4066Sahrens  */
431ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
432fa9e4066Sahrens 	void *buf);
433fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
434fa9e4066Sahrens 	const void *buf, dmu_tx_t *tx);
435feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size);
436feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size,
437feb08c6bSbillm     dmu_tx_t *tx);
43844eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset,
43944eda4d7Smaybee     uint64_t size, struct page *pp, dmu_tx_t *tx);
440fa9e4066Sahrens 
441416e0cd8Sek extern int zfs_prefetch_disable;
442416e0cd8Sek 
443fa9e4066Sahrens /*
444fa9e4066Sahrens  * Asynchronously try to read in the data.
445fa9e4066Sahrens  */
446fa9e4066Sahrens void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
447fa9e4066Sahrens     uint64_t len);
448fa9e4066Sahrens 
449fa9e4066Sahrens typedef struct dmu_object_info {
450fa9e4066Sahrens 	/* All sizes are in bytes. */
451fa9e4066Sahrens 	uint32_t doi_data_block_size;
452fa9e4066Sahrens 	uint32_t doi_metadata_block_size;
453fa9e4066Sahrens 	uint64_t doi_bonus_size;
454fa9e4066Sahrens 	dmu_object_type_t doi_type;
455fa9e4066Sahrens 	dmu_object_type_t doi_bonus_type;
456fa9e4066Sahrens 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
457fa9e4066Sahrens 	uint8_t doi_checksum;
458fa9e4066Sahrens 	uint8_t doi_compress;
459fa9e4066Sahrens 	uint8_t doi_pad[5];
460fa9e4066Sahrens 	/* Values below are number of 512-byte blocks. */
461fa9e4066Sahrens 	uint64_t doi_physical_blks;		/* data + metadata */
462fa9e4066Sahrens 	uint64_t doi_max_block_offset;
463fa9e4066Sahrens } dmu_object_info_t;
464fa9e4066Sahrens 
465e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size);
466e45ce728Sahrens 
467fa9e4066Sahrens typedef struct dmu_object_type_info {
468e45ce728Sahrens 	arc_byteswap_func_t	*ot_byteswap;
469fa9e4066Sahrens 	boolean_t		ot_metadata;
470fa9e4066Sahrens 	char			*ot_name;
471fa9e4066Sahrens } dmu_object_type_info_t;
472fa9e4066Sahrens 
473fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
474fa9e4066Sahrens 
475fa9e4066Sahrens /*
476fa9e4066Sahrens  * Get information on a DMU object.
477fa9e4066Sahrens  *
478fa9e4066Sahrens  * Return 0 on success or ENOENT if object is not allocated.
479fa9e4066Sahrens  *
480fa9e4066Sahrens  * If doi is NULL, just indicates whether the object exists.
481fa9e4066Sahrens  */
482fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
483fa9e4066Sahrens void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
484fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
485fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
486fa9e4066Sahrens     u_longlong_t *nblk512);
487fa9e4066Sahrens 
488fa9e4066Sahrens typedef struct dmu_objset_stats {
489a2eea2e1Sahrens 	uint64_t dds_num_clones; /* number of clones of this */
490a2eea2e1Sahrens 	uint64_t dds_creation_txg;
4913cb34c60Sahrens 	uint64_t dds_guid;
492fa9e4066Sahrens 	dmu_objset_type_t dds_type;
493fa9e4066Sahrens 	uint8_t dds_is_snapshot;
49431fd60d3Sahrens 	uint8_t dds_inconsistent;
4953cb34c60Sahrens 	char dds_origin[MAXNAMELEN];
496a2eea2e1Sahrens } dmu_objset_stats_t;
497fa9e4066Sahrens 
498a2eea2e1Sahrens /*
499a2eea2e1Sahrens  * Get stats on a dataset.
500a2eea2e1Sahrens  */
501a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
502fa9e4066Sahrens 
503a2eea2e1Sahrens /*
504a2eea2e1Sahrens  * Add entries to the nvlist for all the objset's properties.  See
505a2eea2e1Sahrens  * zfs_prop_table[] and zfs(1m) for details on the properties.
506a2eea2e1Sahrens  */
507a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv);
508fa9e4066Sahrens 
509a2eea2e1Sahrens /*
510a2eea2e1Sahrens  * Get the space usage statistics for statvfs().
511a2eea2e1Sahrens  *
512a2eea2e1Sahrens  * refdbytes is the amount of space "referenced" by this objset.
513a2eea2e1Sahrens  * availbytes is the amount of space available to this objset, taking
514a2eea2e1Sahrens  * into account quotas & reservations, assuming that no other objsets
515a2eea2e1Sahrens  * use the space first.  These values correspond to the 'referenced' and
516a2eea2e1Sahrens  * 'available' properties, described in the zfs(1m) manpage.
517a2eea2e1Sahrens  *
518a2eea2e1Sahrens  * usedobjs and availobjs are the number of objects currently allocated,
519a2eea2e1Sahrens  * and available.
520a2eea2e1Sahrens  */
521a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
522a2eea2e1Sahrens     uint64_t *usedobjsp, uint64_t *availobjsp);
523fa9e4066Sahrens 
524fa9e4066Sahrens /*
525a2eea2e1Sahrens  * The fsid_guid is a 56-bit ID that can change to avoid collisions.
526a2eea2e1Sahrens  * (Contrast with the ds_guid which is a 64-bit ID that will never
527a2eea2e1Sahrens  * change, so there is a small probability that it will collide.)
528fa9e4066Sahrens  */
529a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os);
530fa9e4066Sahrens 
531fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os);
532fa9e4066Sahrens 
533fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os);
534fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os);
535fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os);
536fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
537fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf);
538fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os);
539fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os);
540fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
541b38f0970Sck     uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
542ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,
543ab04eb8eStimh     int maxlen, boolean_t *conflict);
54487e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
54587e5029aSahrens     uint64_t *idp, uint64_t *offp);
546f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
547f18faf3fSek extern void *dmu_objset_get_user(objset_t *os);
548fa9e4066Sahrens 
549fa9e4066Sahrens /*
550fa9e4066Sahrens  * Return the txg number for the given assigned transaction.
551fa9e4066Sahrens  */
55287e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
553fa9e4066Sahrens 
554fa9e4066Sahrens /*
555fa9e4066Sahrens  * Synchronous write.
556c5c6ffa0Smaybee  * If a parent zio is provided this function initiates a write on the
557c5c6ffa0Smaybee  * provided buffer as a child of the parent zio.
558da6c28aaSamw  * In the absence of a parent zio, the write is completed synchronously.
559c5c6ffa0Smaybee  * At write completion, blk is filled with the bp of the written block.
560fa9e4066Sahrens  * Note that while the data covered by this function will be on stable
561c5c6ffa0Smaybee  * storage when the write completes this new data does not become a
562fa9e4066Sahrens  * permanent part of the file until the associated transaction commits.
563fa9e4066Sahrens  */
564c5c6ffa0Smaybee typedef void dmu_sync_cb_t(dmu_buf_t *db, void *arg);
565c5c6ffa0Smaybee int dmu_sync(struct zio *zio, dmu_buf_t *db,
566c5c6ffa0Smaybee     struct blkptr *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg);
567fa9e4066Sahrens 
568fa9e4066Sahrens /*
569fa9e4066Sahrens  * Find the next hole or data block in file starting at *off
570fa9e4066Sahrens  * Return found offset in *off. Return ESRCH for end of file.
571fa9e4066Sahrens  */
572fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
573fa9e4066Sahrens     uint64_t *off);
574fa9e4066Sahrens 
575fa9e4066Sahrens /*
576fa9e4066Sahrens  * Initial setup and final teardown.
577fa9e4066Sahrens  */
578fa9e4066Sahrens extern void dmu_init(void);
579fa9e4066Sahrens extern void dmu_fini(void);
580fa9e4066Sahrens 
581fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
582fa9e4066Sahrens     uint64_t object, uint64_t offset, int len);
583fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
584fa9e4066Sahrens     dmu_traverse_cb_t cb, void *arg);
585fa9e4066Sahrens 
5863cb34c60Sahrens int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, boolean_t fromorigin,
5873cb34c60Sahrens     struct vnode *vp, offset_t *off);
5883cb34c60Sahrens 
5893cb34c60Sahrens typedef struct dmu_recv_cookie {
5903cb34c60Sahrens 	/*
5913cb34c60Sahrens 	 * This structure is opaque!
5923cb34c60Sahrens 	 *
5933cb34c60Sahrens 	 * If logical and real are different, we are recving the stream
5943cb34c60Sahrens 	 * into the "real" temporary clone, and then switching it with
5953cb34c60Sahrens 	 * the "logical" target.
5963cb34c60Sahrens 	 */
5973cb34c60Sahrens 	struct dsl_dataset *drc_logical_ds;
5983cb34c60Sahrens 	struct dsl_dataset *drc_real_ds;
5993cb34c60Sahrens 	struct drr_begin *drc_drrb;
6003cb34c60Sahrens 	char *drc_tosnap;
6013cb34c60Sahrens 	boolean_t drc_newfs;
6023cb34c60Sahrens 	boolean_t drc_force;
6033cb34c60Sahrens } dmu_recv_cookie_t;
6043cb34c60Sahrens 
6053cb34c60Sahrens int dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *,
6063cb34c60Sahrens     boolean_t force, objset_t *origin, boolean_t online, dmu_recv_cookie_t *);
6073cb34c60Sahrens int dmu_recv_stream(dmu_recv_cookie_t *drc, struct vnode *vp, offset_t *voffp);
6083cb34c60Sahrens int dmu_recv_end(dmu_recv_cookie_t *drc);
60947f263f4Sek void dmu_recv_abort_cleanup(dmu_recv_cookie_t *drc);
610fa9e4066Sahrens 
611fa9e4066Sahrens /* CRC64 table */
612fa9e4066Sahrens #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
613fa9e4066Sahrens extern uint64_t zfs_crc64_table[256];
614fa9e4066Sahrens 
615fa9e4066Sahrens #ifdef	__cplusplus
616fa9e4066Sahrens }
617fa9e4066Sahrens #endif
618fa9e4066Sahrens 
619fa9e4066Sahrens #endif	/* _SYS_DMU_H */
620