xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/dmu.h (revision 87e5029a3226958edab1512d6182bc74d8d80c9a)
1fa9e4066Sahrens /*
2fa9e4066Sahrens  * CDDL HEADER START
3fa9e4066Sahrens  *
4fa9e4066Sahrens  * The contents of this file are subject to the terms of the
5fa9e4066Sahrens  * Common Development and Distribution License, Version 1.0 only
6fa9e4066Sahrens  * (the "License").  You may not use this file except in compliance
7fa9e4066Sahrens  * with the License.
8fa9e4066Sahrens  *
9fa9e4066Sahrens  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10fa9e4066Sahrens  * or http://www.opensolaris.org/os/licensing.
11fa9e4066Sahrens  * See the License for the specific language governing permissions
12fa9e4066Sahrens  * and limitations under the License.
13fa9e4066Sahrens  *
14fa9e4066Sahrens  * When distributing Covered Code, include this CDDL HEADER in each
15fa9e4066Sahrens  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16fa9e4066Sahrens  * If applicable, add the following below this CDDL HEADER, with the
17fa9e4066Sahrens  * fields enclosed by brackets "[]" replaced with your own identifying
18fa9e4066Sahrens  * information: Portions Copyright [yyyy] [name of copyright owner]
19fa9e4066Sahrens  *
20fa9e4066Sahrens  * CDDL HEADER END
21fa9e4066Sahrens  */
22fa9e4066Sahrens /*
23fa9e4066Sahrens  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24fa9e4066Sahrens  * Use is subject to license terms.
25fa9e4066Sahrens  */
26fa9e4066Sahrens 
27fa9e4066Sahrens #ifndef	_SYS_DMU_H
28fa9e4066Sahrens #define	_SYS_DMU_H
29fa9e4066Sahrens 
30fa9e4066Sahrens #pragma ident	"%Z%%M%	%I%	%E% SMI"
31fa9e4066Sahrens 
32fa9e4066Sahrens /*
33fa9e4066Sahrens  * This file describes the interface that the DMU provides for its
34fa9e4066Sahrens  * consumers.
35fa9e4066Sahrens  *
36fa9e4066Sahrens  * The DMU also interacts with the SPA.  That interface is described in
37fa9e4066Sahrens  * dmu_spa.h.
38fa9e4066Sahrens  */
39fa9e4066Sahrens 
40fa9e4066Sahrens #include <sys/inttypes.h>
41fa9e4066Sahrens #include <sys/types.h>
42fa9e4066Sahrens #include <sys/param.h>
43fa9e4066Sahrens 
44fa9e4066Sahrens #ifdef	__cplusplus
45fa9e4066Sahrens extern "C" {
46fa9e4066Sahrens #endif
47fa9e4066Sahrens 
48fa9e4066Sahrens struct uio;
49fa9e4066Sahrens struct vnode;
50fa9e4066Sahrens struct spa;
51fa9e4066Sahrens struct zilog;
52fa9e4066Sahrens struct zio;
53fa9e4066Sahrens struct blkptr;
54fa9e4066Sahrens struct zap_cursor;
55fa9e4066Sahrens struct dsl_dataset;
56fa9e4066Sahrens struct dsl_pool;
57fa9e4066Sahrens struct dnode;
58fa9e4066Sahrens struct drr_begin;
59fa9e4066Sahrens struct drr_end;
60fa9e4066Sahrens 
61fa9e4066Sahrens typedef struct objset objset_t;
62fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t;
63fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t;
64fa9e4066Sahrens 
65fa9e4066Sahrens typedef enum dmu_object_type {
66fa9e4066Sahrens 	DMU_OT_NONE,
67fa9e4066Sahrens 	/* general: */
68fa9e4066Sahrens 	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
69fa9e4066Sahrens 	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
70fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
71fa9e4066Sahrens 	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
72fa9e4066Sahrens 	DMU_OT_BPLIST,			/* UINT64 */
73fa9e4066Sahrens 	DMU_OT_BPLIST_HDR,		/* UINT64 */
74fa9e4066Sahrens 	/* spa: */
75fa9e4066Sahrens 	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
76fa9e4066Sahrens 	DMU_OT_SPACE_MAP,		/* UINT64 */
77fa9e4066Sahrens 	/* zil: */
78fa9e4066Sahrens 	DMU_OT_INTENT_LOG,		/* UINT64 */
79fa9e4066Sahrens 	/* dmu: */
80fa9e4066Sahrens 	DMU_OT_DNODE,			/* DNODE */
81fa9e4066Sahrens 	DMU_OT_OBJSET,			/* OBJSET */
82fa9e4066Sahrens 	/* dsl: */
83fa9e4066Sahrens 	DMU_OT_DSL_DATASET,		/* UINT64 */
84*87e5029aSahrens 	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
85*87e5029aSahrens 	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
86fa9e4066Sahrens 	DMU_OT_DSL_PROPS,		/* ZAP */
87fa9e4066Sahrens 	DMU_OT_DSL_OBJSET,		/* UINT64 */
88fa9e4066Sahrens 	/* zpl: */
89fa9e4066Sahrens 	DMU_OT_ZNODE,			/* ZNODE */
90fa9e4066Sahrens 	DMU_OT_ACL,			/* ACL */
91fa9e4066Sahrens 	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
92fa9e4066Sahrens 	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
93fa9e4066Sahrens 	DMU_OT_MASTER_NODE,		/* ZAP */
94fa9e4066Sahrens 	DMU_OT_DELETE_QUEUE,		/* ZAP */
95fa9e4066Sahrens 	/* zvol: */
96fa9e4066Sahrens 	DMU_OT_ZVOL,			/* UINT8 */
97fa9e4066Sahrens 	DMU_OT_ZVOL_PROP,		/* ZAP */
98fa9e4066Sahrens 	/* other; for testing only! */
99fa9e4066Sahrens 	DMU_OT_PLAIN_OTHER,		/* UINT8 */
100fa9e4066Sahrens 	DMU_OT_UINT64_OTHER,		/* UINT64 */
101fa9e4066Sahrens 	DMU_OT_ZAP_OTHER,		/* ZAP */
102fa9e4066Sahrens 
103fa9e4066Sahrens 	DMU_OT_NUMTYPES
104fa9e4066Sahrens } dmu_object_type_t;
105fa9e4066Sahrens 
106fa9e4066Sahrens typedef enum dmu_objset_type {
107fa9e4066Sahrens 	DMU_OST_NONE,
108fa9e4066Sahrens 	DMU_OST_META,
109fa9e4066Sahrens 	DMU_OST_ZFS,
110fa9e4066Sahrens 	DMU_OST_ZVOL,
111fa9e4066Sahrens 	DMU_OST_OTHER,			/* For testing only! */
112fa9e4066Sahrens 	DMU_OST_ANY,			/* Be careful! */
113fa9e4066Sahrens 	DMU_OST_NUMTYPES
114fa9e4066Sahrens } dmu_objset_type_t;
115fa9e4066Sahrens 
116fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size);
117fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size);
118fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size);
119fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size);
120fa9e4066Sahrens void zap_byteswap(void *buf, size_t size);
121fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size);
122fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size);
123fa9e4066Sahrens 
124fa9e4066Sahrens #define	DS_MODE_NONE		0	/* invalid, to aid debugging */
125fa9e4066Sahrens #define	DS_MODE_STANDARD	1	/* normal access, no special needs */
126fa9e4066Sahrens #define	DS_MODE_PRIMARY		2	/* the "main" access, e.g. a mount */
127fa9e4066Sahrens #define	DS_MODE_EXCLUSIVE	3	/* exclusive access, e.g. to destroy */
128fa9e4066Sahrens #define	DS_MODE_LEVELS		4
129fa9e4066Sahrens #define	DS_MODE_LEVEL(x)	((x) & (DS_MODE_LEVELS - 1))
130fa9e4066Sahrens #define	DS_MODE_READONLY	0x8
131fa9e4066Sahrens #define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
132fa9e4066Sahrens #define	DS_MODE_RESTORE		0x10
133fa9e4066Sahrens #define	DS_MODE_IS_RESTORE(x)	((x) & DS_MODE_RESTORE)
134fa9e4066Sahrens 
135fa9e4066Sahrens #define	DS_FIND_SNAPSHOTS	0x01
136fa9e4066Sahrens 
137fa9e4066Sahrens /*
138fa9e4066Sahrens  * The maximum number of bytes that can be accessed as part of one
139fa9e4066Sahrens  * operation, including metadata.
140fa9e4066Sahrens  */
141fa9e4066Sahrens #define	DMU_MAX_ACCESS (10<<20) /* 10MB */
142fa9e4066Sahrens 
143fa9e4066Sahrens /*
144fa9e4066Sahrens  * Public routines to create, destroy, open, and close objsets.
145fa9e4066Sahrens  */
146fa9e4066Sahrens int dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
147fa9e4066Sahrens     objset_t **osp);
148fa9e4066Sahrens void dmu_objset_close(objset_t *os);
149fa9e4066Sahrens int dmu_objset_create(const char *name, dmu_objset_type_t type,
150fa9e4066Sahrens     objset_t *clone_parent,
151fa9e4066Sahrens     void (*func)(objset_t *os, void *arg, dmu_tx_t *tx), void *arg);
152fa9e4066Sahrens int dmu_objset_destroy(const char *name);
153fa9e4066Sahrens int dmu_objset_rollback(const char *name);
154fa9e4066Sahrens int dmu_objset_rename(const char *name, const char *newname);
155fa9e4066Sahrens void dmu_objset_set_quota(objset_t *os, uint64_t quota);
156fa9e4066Sahrens uint64_t dmu_objset_get_quota(objset_t *os);
157fa9e4066Sahrens int dmu_objset_request_reservation(objset_t *os, uint64_t reservation);
158fa9e4066Sahrens void dmu_objset_find(char *name, void func(char *, void *), void *arg,
159fa9e4066Sahrens     int flags);
160fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size);
161fa9e4066Sahrens 
162fa9e4066Sahrens typedef struct dmu_buf {
163fa9e4066Sahrens 	uint64_t db_object;		/* object that this buffer is part of */
164fa9e4066Sahrens 	uint64_t db_offset;		/* byte offset in this object */
165fa9e4066Sahrens 	uint64_t db_size;		/* size of buffer in bytes */
166fa9e4066Sahrens 	void *db_data;			/* data in buffer */
167fa9e4066Sahrens } dmu_buf_t;
168fa9e4066Sahrens 
169fa9e4066Sahrens typedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
170fa9e4066Sahrens 
171fa9e4066Sahrens /*
172fa9e4066Sahrens  * Callback function to perform byte swapping on a block.
173fa9e4066Sahrens  */
174fa9e4066Sahrens typedef void dmu_byteswap_func_t(void *buf, size_t size);
175fa9e4066Sahrens 
176fa9e4066Sahrens #define	DMU_POOL_DIRECTORY_OBJECT	1
177fa9e4066Sahrens #define	DMU_POOL_CONFIG			"config"
178fa9e4066Sahrens #define	DMU_POOL_ROOT_DATASET		"root_dataset"
179fa9e4066Sahrens #define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
180fa9e4066Sahrens 
181fa9e4066Sahrens /*
182fa9e4066Sahrens  * Allocate an object from this objset.  The range of object numbers
183fa9e4066Sahrens  * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
184fa9e4066Sahrens  *
185fa9e4066Sahrens  * The transaction must be assigned to a txg.  The newly allocated
186fa9e4066Sahrens  * object will be "held" in the transaction (ie. you can modify the
187fa9e4066Sahrens  * newly allocated object in this transaction).
188fa9e4066Sahrens  *
189fa9e4066Sahrens  * dmu_object_alloc() chooses an object and returns it in *objectp.
190fa9e4066Sahrens  *
191fa9e4066Sahrens  * dmu_object_claim() allocates a specific object number.  If that
192fa9e4066Sahrens  * number is already allocated, it fails and returns EEXIST.
193fa9e4066Sahrens  *
194fa9e4066Sahrens  * Return 0 on success, or ENOSPC or EEXIST as specified above.
195fa9e4066Sahrens  */
196fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
197fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
198fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
199fa9e4066Sahrens     int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
200fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
201fa9e4066Sahrens     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
202fa9e4066Sahrens 
203fa9e4066Sahrens /*
204fa9e4066Sahrens  * Free an object from this objset.
205fa9e4066Sahrens  *
206fa9e4066Sahrens  * The object's data will be freed as well (ie. you don't need to call
207fa9e4066Sahrens  * dmu_free(object, 0, -1, tx)).
208fa9e4066Sahrens  *
209fa9e4066Sahrens  * The object need not be held in the transaction.
210fa9e4066Sahrens  *
211fa9e4066Sahrens  * If there are any holds on this object's buffers (via dmu_buf_hold()),
212fa9e4066Sahrens  * or tx holds on the object (via dmu_tx_hold_object()), you can not
213fa9e4066Sahrens  * free it; it fails and returns EBUSY.
214fa9e4066Sahrens  *
215fa9e4066Sahrens  * If the object is not allocated, it fails and returns ENOENT.
216fa9e4066Sahrens  *
217fa9e4066Sahrens  * Return 0 on success, or EBUSY or ENOENT as specified above.
218fa9e4066Sahrens  */
219fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
220fa9e4066Sahrens 
221fa9e4066Sahrens /*
222fa9e4066Sahrens  * Find the next allocated or free object.
223fa9e4066Sahrens  *
224fa9e4066Sahrens  * The objectp parameter is in-out.  It will be updated to be the next
225fa9e4066Sahrens  * object which is allocated.
226fa9e4066Sahrens  *
227fa9e4066Sahrens  * XXX Can only be called on a objset with no dirty data.
228fa9e4066Sahrens  *
229fa9e4066Sahrens  * Returns 0 on success, or ENOENT if there are no more objects.
230fa9e4066Sahrens  */
231fa9e4066Sahrens int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole);
232fa9e4066Sahrens 
233fa9e4066Sahrens /*
234fa9e4066Sahrens  * Set the data blocksize for an object.
235fa9e4066Sahrens  *
236fa9e4066Sahrens  * The object cannot have any blocks allcated beyond the first.  If
237fa9e4066Sahrens  * the first block is allocated already, the new size must be greater
238fa9e4066Sahrens  * than the current block size.  If these conditions are not met,
239fa9e4066Sahrens  * ENOTSUP will be returned.
240fa9e4066Sahrens  *
241fa9e4066Sahrens  * Returns 0 on success, or EBUSY if there are any holds on the object
242fa9e4066Sahrens  * contents, or ENOTSUP as described above.
243fa9e4066Sahrens  */
244fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
245fa9e4066Sahrens     int ibs, dmu_tx_t *tx);
246fa9e4066Sahrens 
247fa9e4066Sahrens /*
248fa9e4066Sahrens  * Set the checksum property on a dnode.  The new checksum algorithm will
249fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
250fa9e4066Sahrens  */
251fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
252fa9e4066Sahrens     dmu_tx_t *tx);
253fa9e4066Sahrens 
254fa9e4066Sahrens /*
255fa9e4066Sahrens  * Set the compress property on a dnode.  The new compression algorithm will
256fa9e4066Sahrens  * apply to all newly written blocks; existing blocks will not be affected.
257fa9e4066Sahrens  */
258fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
259fa9e4066Sahrens     dmu_tx_t *tx);
260fa9e4066Sahrens 
261fa9e4066Sahrens /*
262fa9e4066Sahrens  * The bonus data is accessed more or less like a regular buffer.
263fa9e4066Sahrens  * You must dmu_bonus_hold() to get the buffer, which will give you a
264fa9e4066Sahrens  * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
265fa9e4066Sahrens  * data.  As with any normal buffer, you must call dmu_buf_read() to
266fa9e4066Sahrens  * read db_data, dmu_buf_will_dirty() before modifying it, and the
267fa9e4066Sahrens  * object must be held in an assigned transaction before calling
268fa9e4066Sahrens  * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
269fa9e4066Sahrens  * buffer as well.  You must release your hold with dmu_buf_rele().
270fa9e4066Sahrens  */
271fa9e4066Sahrens dmu_buf_t *dmu_bonus_hold(objset_t *os, uint64_t object);
272fa9e4066Sahrens dmu_buf_t *dmu_bonus_hold_tag(objset_t *os, uint64_t object, void *tag);
273fa9e4066Sahrens int dmu_bonus_max(void);
274fa9e4066Sahrens 
275fa9e4066Sahrens /*
276fa9e4066Sahrens  * Obtain the DMU buffer from the specified object which contains the
277fa9e4066Sahrens  * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
278fa9e4066Sahrens  * that it will remain in memory.  You must release the hold with
279fa9e4066Sahrens  * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
280fa9e4066Sahrens  * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
281fa9e4066Sahrens  *
282fa9e4066Sahrens  * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
283fa9e4066Sahrens  * on the returned buffer before reading or writing the buffer's
284fa9e4066Sahrens  * db_data.  The comments for those routines describe what particular
285fa9e4066Sahrens  * operations are valid after calling them.
286fa9e4066Sahrens  *
287fa9e4066Sahrens  * The object number must be a valid, allocated object number.
288fa9e4066Sahrens  */
289fa9e4066Sahrens dmu_buf_t *dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset);
290fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
291fa9e4066Sahrens void dmu_buf_remove_ref(dmu_buf_t *db, void* tag);
292fa9e4066Sahrens void dmu_buf_rele(dmu_buf_t *db);
293fa9e4066Sahrens void dmu_buf_rele_tag(dmu_buf_t *db, void *tag);
294fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db);
295fa9e4066Sahrens 
296fa9e4066Sahrens /*
297fa9e4066Sahrens  * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
298fa9e4066Sahrens  * range of an object.  A pointer to an array of dmu_buf_t*'s is
299fa9e4066Sahrens  * returned (in *dbpp).
300fa9e4066Sahrens  *
301fa9e4066Sahrens  * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
302fa9e4066Sahrens  * frees the array.  The hold on the array of buffers MUST be released
303fa9e4066Sahrens  * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
304fa9e4066Sahrens  * individually with dmu_buf_rele.
305fa9e4066Sahrens  */
306fa9e4066Sahrens dmu_buf_t **dmu_buf_hold_array(objset_t *os, uint64_t object,
307fa9e4066Sahrens     uint64_t offset, uint64_t length, int *numbufs);
308fa9e4066Sahrens void dmu_buf_rele_array(dmu_buf_t **, int numbufs);
309fa9e4066Sahrens 
310fa9e4066Sahrens /*
311fa9e4066Sahrens  * Returns NULL on success, or the existing user ptr if it's already
312fa9e4066Sahrens  * been set.
313fa9e4066Sahrens  *
314fa9e4066Sahrens  * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
315fa9e4066Sahrens  *
316fa9e4066Sahrens  * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
317fa9e4066Sahrens  * will be set to db->db_data when you are allowed to access it.  Note
318fa9e4066Sahrens  * that db->db_data (the pointer) can change when you do dmu_buf_read(),
319fa9e4066Sahrens  * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
320fa9e4066Sahrens  * *user_data_ptr_ptr will be set to the new value when it changes.
321fa9e4066Sahrens  *
322fa9e4066Sahrens  * If non-NULL, pageout func will be called when this buffer is being
323fa9e4066Sahrens  * excised from the cache, so that you can clean up the data structure
324fa9e4066Sahrens  * pointed to by user_ptr.
325fa9e4066Sahrens  *
326fa9e4066Sahrens  * dmu_evict_user() will call the pageout func for all buffers in a
327fa9e4066Sahrens  * objset with a given pageout func.
328fa9e4066Sahrens  */
329fa9e4066Sahrens void *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
330fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
331fa9e4066Sahrens /*
332fa9e4066Sahrens  * set_user_ie is the same as set_user, but request immediate eviction
333fa9e4066Sahrens  * when hold count goes to zero.
334fa9e4066Sahrens  */
335fa9e4066Sahrens void *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
336fa9e4066Sahrens     void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
337fa9e4066Sahrens void *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
338fa9e4066Sahrens     void *user_ptr, void *user_data_ptr_ptr,
339fa9e4066Sahrens     dmu_buf_evict_func_t *pageout_func);
340fa9e4066Sahrens void dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
341fa9e4066Sahrens 
342fa9e4066Sahrens void dmu_buf_hold_data(dmu_buf_t *db);
343fa9e4066Sahrens void dmu_buf_rele_data(dmu_buf_t *db);
344fa9e4066Sahrens 
345fa9e4066Sahrens /*
346fa9e4066Sahrens  * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
347fa9e4066Sahrens  */
348fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db);
349fa9e4066Sahrens 
350fa9e4066Sahrens /*
351fa9e4066Sahrens  * Indicate that you are going to read the buffer's data (db_data).
352fa9e4066Sahrens  *
353fa9e4066Sahrens  * This routine will read the data from disk if necessary.
354fa9e4066Sahrens  *
355fa9e4066Sahrens  * These routines will return 0 on success, or an errno if there is a
356fa9e4066Sahrens  * nonrecoverable I/O error.
357fa9e4066Sahrens  */
358fa9e4066Sahrens void dmu_buf_read(dmu_buf_t *db);
359fa9e4066Sahrens int dmu_buf_read_canfail(dmu_buf_t *db);
360fa9e4066Sahrens void dmu_buf_read_array(dmu_buf_t **dbp, int numbufs);
361fa9e4066Sahrens int dmu_buf_read_array_canfail(dmu_buf_t **dbp, int numbufs);
362fa9e4066Sahrens 
363fa9e4066Sahrens /*
364fa9e4066Sahrens  * Indicate that you are going to modify the buffer's data (db_data).
365fa9e4066Sahrens  *
366fa9e4066Sahrens  * The transaction (tx) must be assigned to a txg (ie. you've called
367fa9e4066Sahrens  * dmu_tx_assign()).  The buffer's object must be held in the tx
368fa9e4066Sahrens  * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
369fa9e4066Sahrens  */
370fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
371fa9e4066Sahrens 
372fa9e4066Sahrens /*
373fa9e4066Sahrens  * Indicate that you are going to modify the entire contents of the
374fa9e4066Sahrens  * buffer's data ("fill" it).
375fa9e4066Sahrens  *
376fa9e4066Sahrens  * This routine is the same as dmu_buf_will_dirty, except that it won't
377fa9e4066Sahrens  * read the contents off the disk, so the contents may be uninitialized
378fa9e4066Sahrens  * and you must overwrite it.
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 (ie.
382fa9e4066Sahrens  * you've called dmu_tx_hold_object(tx, db->db_object)).
383fa9e4066Sahrens  */
384fa9e4066Sahrens /* void dmu_buf_will_fill(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);
411fa9e4066Sahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int ops);
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);
415fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx);
416fa9e4066Sahrens 
417fa9e4066Sahrens /*
418fa9e4066Sahrens  * Free up the data blocks for a defined range of a file.  If size is
419fa9e4066Sahrens  * zero, the range from offset to end-of-file is freed.
420fa9e4066Sahrens  */
421fa9e4066Sahrens void dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
422fa9e4066Sahrens 	uint64_t size, dmu_tx_t *tx);
423fa9e4066Sahrens 
424fa9e4066Sahrens /*
425fa9e4066Sahrens  * Convenience functions.
426fa9e4066Sahrens  *
427fa9e4066Sahrens  * Canfail routines will return 0 on success, or an errno if there is a
428fa9e4066Sahrens  * nonrecoverable I/O error.
429fa9e4066Sahrens  */
430fa9e4066Sahrens void dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
431fa9e4066Sahrens 	void *buf);
432fa9e4066Sahrens int dmu_read_canfail(objset_t *dd, uint64_t object, uint64_t offset,
433fa9e4066Sahrens 	uint64_t size, void *buf);
434fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
435fa9e4066Sahrens 	const void *buf, dmu_tx_t *tx);
436fa9e4066Sahrens int dmu_write_uio(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
437fa9e4066Sahrens     struct uio *uio, dmu_tx_t *tx);
438fa9e4066Sahrens 
439fa9e4066Sahrens /*
440fa9e4066Sahrens  * Asynchronously try to read in the data.
441fa9e4066Sahrens  */
442fa9e4066Sahrens void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
443fa9e4066Sahrens     uint64_t len);
444fa9e4066Sahrens 
445fa9e4066Sahrens typedef struct dmu_object_info {
446fa9e4066Sahrens 	/* All sizes are in bytes. */
447fa9e4066Sahrens 	uint32_t doi_data_block_size;
448fa9e4066Sahrens 	uint32_t doi_metadata_block_size;
449fa9e4066Sahrens 	uint64_t doi_bonus_size;
450fa9e4066Sahrens 	dmu_object_type_t doi_type;
451fa9e4066Sahrens 	dmu_object_type_t doi_bonus_type;
452fa9e4066Sahrens 	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
453fa9e4066Sahrens 	uint8_t doi_checksum;
454fa9e4066Sahrens 	uint8_t doi_compress;
455fa9e4066Sahrens 	uint8_t doi_pad[5];
456fa9e4066Sahrens 	/* Values below are number of 512-byte blocks. */
457fa9e4066Sahrens 	uint64_t doi_physical_blks;		/* data + metadata */
458fa9e4066Sahrens 	uint64_t doi_max_block_offset;
459fa9e4066Sahrens } dmu_object_info_t;
460fa9e4066Sahrens 
461fa9e4066Sahrens typedef struct dmu_object_type_info {
462fa9e4066Sahrens 	dmu_byteswap_func_t	*ot_byteswap;
463fa9e4066Sahrens 	boolean_t		ot_metadata;
464fa9e4066Sahrens 	char			*ot_name;
465fa9e4066Sahrens } dmu_object_type_info_t;
466fa9e4066Sahrens 
467fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
468fa9e4066Sahrens 
469fa9e4066Sahrens /*
470fa9e4066Sahrens  * Get information on a DMU object.
471fa9e4066Sahrens  *
472fa9e4066Sahrens  * Return 0 on success or ENOENT if object is not allocated.
473fa9e4066Sahrens  *
474fa9e4066Sahrens  * If doi is NULL, just indicates whether the object exists.
475fa9e4066Sahrens  */
476fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
477fa9e4066Sahrens void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
478fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
479fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
480fa9e4066Sahrens     u_longlong_t *nblk512);
481fa9e4066Sahrens 
482fa9e4066Sahrens /*
483fa9e4066Sahrens  * Get the maximum nonzero offset in the object (ie. this offset and all
484fa9e4066Sahrens  * offsets following are zero).
485fa9e4066Sahrens  *
486fa9e4066Sahrens  * XXX Perhaps integrate this with dmu_object_info(), although that
487fa9e4066Sahrens  * would then have to bring in the indirect blocks.
488fa9e4066Sahrens  */
489fa9e4066Sahrens uint64_t dmu_object_max_nonzero_offset(objset_t *os, uint64_t object);
490fa9e4066Sahrens 
491fa9e4066Sahrens typedef struct dmu_objset_stats {
492fa9e4066Sahrens 	dmu_objset_type_t dds_type;
493fa9e4066Sahrens 	uint8_t dds_is_snapshot;
494fa9e4066Sahrens 	uint8_t dds_is_placeholder;
495fa9e4066Sahrens 	uint8_t dds_pad[2];
496fa9e4066Sahrens 
497fa9e4066Sahrens 	uint64_t dds_creation_time;
498fa9e4066Sahrens 	uint64_t dds_creation_txg;
499fa9e4066Sahrens 
500fa9e4066Sahrens 	char dds_clone_of[MAXNAMELEN];
501fa9e4066Sahrens 
502fa9e4066Sahrens 	/* How much data is there in this objset? */
503fa9e4066Sahrens 
504fa9e4066Sahrens 	/*
505fa9e4066Sahrens 	 * Space referenced, taking into account pending writes and
506fa9e4066Sahrens 	 * frees.  Only relavent to filesystems and snapshots (not
507fa9e4066Sahrens 	 * collections).
508fa9e4066Sahrens 	 */
509fa9e4066Sahrens 	uint64_t dds_space_refd;
510fa9e4066Sahrens 
511fa9e4066Sahrens 	/*
512fa9e4066Sahrens 	 * Space "used", taking into account pending writes and frees, and
513fa9e4066Sahrens 	 * children's reservations (in bytes).  This is the amount of
514fa9e4066Sahrens 	 * space that will be freed if this and all dependent items are
515fa9e4066Sahrens 	 * destroyed (eg. child datasets, objsets, and snapshots).  So
516fa9e4066Sahrens 	 * for snapshots, this is the amount of space unique to this
517fa9e4066Sahrens 	 * snapshot.
518fa9e4066Sahrens 	 */
519fa9e4066Sahrens 	uint64_t dds_space_used;
520fa9e4066Sahrens 
521fa9e4066Sahrens 	/*
522fa9e4066Sahrens 	 * Compressed and uncompressed bytes consumed.  Does not take
523fa9e4066Sahrens 	 * into account reservations.  Used for computing compression
524fa9e4066Sahrens 	 * ratio.
525fa9e4066Sahrens 	 */
526fa9e4066Sahrens 	uint64_t dds_compressed_bytes;
527fa9e4066Sahrens 	uint64_t dds_uncompressed_bytes;
528fa9e4066Sahrens 
529fa9e4066Sahrens 	/*
530fa9e4066Sahrens 	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
531fa9e4066Sahrens 	 * collisions.  The ds_guid is a 64-bit ID that will never
532fa9e4066Sahrens 	 * change, so there is a small probability that it will collide.
533fa9e4066Sahrens 	 */
534fa9e4066Sahrens 	uint64_t dds_fsid_guid;
535fa9e4066Sahrens 	uint64_t dds_guid;
536fa9e4066Sahrens 
537fa9e4066Sahrens 	uint64_t dds_objects_used;	/* number of objects used */
538fa9e4066Sahrens 	uint64_t dds_objects_avail;	/* number of objects available */
539fa9e4066Sahrens 
540fa9e4066Sahrens 	uint64_t dds_num_clones; /* number of clones of this */
541fa9e4066Sahrens 
542fa9e4066Sahrens 	/* The dataset's administratively-set quota, in bytes. */
543fa9e4066Sahrens 	uint64_t dds_quota;
544fa9e4066Sahrens 
545fa9e4066Sahrens 	/* The dataset's administratively-set reservation, in bytes */
546fa9e4066Sahrens 	uint64_t dds_reserved;
547fa9e4066Sahrens 
548fa9e4066Sahrens 	/*
549fa9e4066Sahrens 	 * The amount of additional space that this dataset can consume.
550fa9e4066Sahrens 	 * Takes into account quotas & reservations.
551fa9e4066Sahrens 	 * (Assuming that no other datasets consume it first.)
552fa9e4066Sahrens 	 */
553fa9e4066Sahrens 	uint64_t dds_available;
554fa9e4066Sahrens 
555fa9e4066Sahrens 	/*
556fa9e4066Sahrens 	 * Various properties.
557fa9e4066Sahrens 	 */
558fa9e4066Sahrens 	uint64_t dds_compression;
559fa9e4066Sahrens 	uint64_t dds_checksum;
560fa9e4066Sahrens 	uint64_t dds_zoned;
561fa9e4066Sahrens 	char dds_compression_setpoint[MAXNAMELEN];
562fa9e4066Sahrens 	char dds_checksum_setpoint[MAXNAMELEN];
563fa9e4066Sahrens 	char dds_zoned_setpoint[MAXNAMELEN];
564fa9e4066Sahrens 	char dds_altroot[MAXPATHLEN];
565fa9e4066Sahrens 
566fa9e4066Sahrens 	/* The following are for debugging purposes only */
567fa9e4066Sahrens 	uint64_t dds_last_txg;
568fa9e4066Sahrens 	uint64_t dds_dir_obj;
569fa9e4066Sahrens 	uint64_t dds_objset_obj;
570fa9e4066Sahrens 	uint64_t dds_clone_of_obj;
571fa9e4066Sahrens } dmu_objset_stats_t;
572fa9e4066Sahrens 
573fa9e4066Sahrens /*
574fa9e4066Sahrens  * Get stats on a dataset.
575fa9e4066Sahrens  */
576fa9e4066Sahrens void dmu_objset_stats(objset_t *os, dmu_objset_stats_t *dds);
577fa9e4066Sahrens 
578fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os);
579fa9e4066Sahrens 
580fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os);
581fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os);
582fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os);
583fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
584fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf);
585fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os);
586fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os);
587fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
588fa9e4066Sahrens     uint64_t *id, uint64_t *offp);
589*87e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
590*87e5029aSahrens     uint64_t *idp, uint64_t *offp);
591fa9e4066Sahrens 
592fa9e4066Sahrens /*
593fa9e4066Sahrens  * Return the txg number for the given assigned transaction.
594fa9e4066Sahrens  */
595*87e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
596fa9e4066Sahrens 
597fa9e4066Sahrens /*
598fa9e4066Sahrens  * Synchronous write.
599fa9e4066Sahrens  * On success returns 0 and fills in the blk pointed at by bp.
600fa9e4066Sahrens  * Note that while the data covered by this function will be on stable
601fa9e4066Sahrens  * storage when the function returns this new data does not become a
602fa9e4066Sahrens  * permanent part of the file until the associated transaction commits.
603fa9e4066Sahrens  */
604fa9e4066Sahrens int dmu_sync(objset_t *os, uint64_t object, uint64_t offset, uint64_t *blkoff,
605fa9e4066Sahrens     struct blkptr *bp, uint64_t txg);
606fa9e4066Sahrens 
607fa9e4066Sahrens /*
608fa9e4066Sahrens  * Find the next hole or data block in file starting at *off
609fa9e4066Sahrens  * Return found offset in *off. Return ESRCH for end of file.
610fa9e4066Sahrens  */
611fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
612fa9e4066Sahrens     uint64_t *off);
613fa9e4066Sahrens 
614fa9e4066Sahrens /*
615fa9e4066Sahrens  * Initial setup and final teardown.
616fa9e4066Sahrens  */
617fa9e4066Sahrens extern void dmu_init(void);
618fa9e4066Sahrens extern void dmu_fini(void);
619fa9e4066Sahrens 
620fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
621fa9e4066Sahrens     uint64_t object, uint64_t offset, int len);
622fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
623fa9e4066Sahrens     dmu_traverse_cb_t cb, void *arg);
624fa9e4066Sahrens 
625fa9e4066Sahrens int dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp);
626fa9e4066Sahrens int dmu_recvbackup(struct drr_begin *drrb, uint64_t *sizep,
627fa9e4066Sahrens     struct vnode *vp, uint64_t voffset);
628fa9e4066Sahrens 
629fa9e4066Sahrens /* CRC64 table */
630fa9e4066Sahrens #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
631fa9e4066Sahrens extern uint64_t zfs_crc64_table[256];
632fa9e4066Sahrens 
633fa9e4066Sahrens #ifdef	__cplusplus
634fa9e4066Sahrens }
635fa9e4066Sahrens #endif
636fa9e4066Sahrens 
637fa9e4066Sahrens #endif	/* _SYS_DMU_H */
638