dmu.h revision 44cd46cadd9aab751dae6a4023c1cb5bf316d274
1fa9e406ahrens/*
2fa9e406ahrens * CDDL HEADER START
3fa9e406ahrens *
4fa9e406ahrens * The contents of this file are subject to the terms of the
5ea8dc4beschrock * Common Development and Distribution License (the "License").
6ea8dc4beschrock * You may not use this file except in compliance with the License.
7fa9e406ahrens *
8fa9e406ahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e406ahrens * or http://www.opensolaris.org/os/licensing.
10fa9e406ahrens * See the License for the specific language governing permissions
11fa9e406ahrens * and limitations under the License.
12fa9e406ahrens *
13fa9e406ahrens * When distributing Covered Code, include this CDDL HEADER in each
14fa9e406ahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e406ahrens * If applicable, add the following below this CDDL HEADER, with the
16fa9e406ahrens * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e406ahrens * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e406ahrens *
19fa9e406ahrens * CDDL HEADER END
20fa9e406ahrens */
21fa9e406ahrens/*
227f7322feschrock * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23fa9e406ahrens * Use is subject to license terms.
24fa9e406ahrens */
25fa9e406ahrens
26fa9e406ahrens#ifndef	_SYS_DMU_H
27fa9e406ahrens#define	_SYS_DMU_H
28fa9e406ahrens
29fa9e406ahrens#pragma ident	"%Z%%M%	%I%	%E% SMI"
30fa9e406ahrens
31fa9e406ahrens/*
32fa9e406ahrens * This file describes the interface that the DMU provides for its
33fa9e406ahrens * consumers.
34fa9e406ahrens *
35fa9e406ahrens * The DMU also interacts with the SPA.  That interface is described in
36fa9e406ahrens * dmu_spa.h.
37fa9e406ahrens */
38fa9e406ahrens
39fa9e406ahrens#include <sys/inttypes.h>
40fa9e406ahrens#include <sys/types.h>
41fa9e406ahrens#include <sys/param.h>
42fa9e406ahrens
43fa9e406ahrens#ifdef	__cplusplus
44fa9e406ahrensextern "C" {
45fa9e406ahrens#endif
46fa9e406ahrens
47fa9e406ahrensstruct uio;
48fa9e406ahrensstruct vnode;
49fa9e406ahrensstruct spa;
50fa9e406ahrensstruct zilog;
51fa9e406ahrensstruct zio;
52fa9e406ahrensstruct blkptr;
53fa9e406ahrensstruct zap_cursor;
54fa9e406ahrensstruct dsl_dataset;
55fa9e406ahrensstruct dsl_pool;
56fa9e406ahrensstruct dnode;
57fa9e406ahrensstruct drr_begin;
58fa9e406ahrensstruct drr_end;
5944cd46cbillmstruct zbookmark;
6044cd46cbillmstruct spa;
61fa9e406ahrens
62fa9e406ahrenstypedef struct objset objset_t;
63fa9e406ahrenstypedef struct dmu_tx dmu_tx_t;
64fa9e406ahrenstypedef struct dsl_dir dsl_dir_t;
65fa9e406ahrens
66fa9e406ahrenstypedef enum dmu_object_type {
67fa9e406ahrens	DMU_OT_NONE,
68fa9e406ahrens	/* general: */
69fa9e406ahrens	DMU_OT_OBJECT_DIRECTORY,	/* ZAP */
70fa9e406ahrens	DMU_OT_OBJECT_ARRAY,		/* UINT64 */
71fa9e406ahrens	DMU_OT_PACKED_NVLIST,		/* UINT8 (XDR by nvlist_pack/unpack) */
72fa9e406ahrens	DMU_OT_PACKED_NVLIST_SIZE,	/* UINT64 */
73fa9e406ahrens	DMU_OT_BPLIST,			/* UINT64 */
74fa9e406ahrens	DMU_OT_BPLIST_HDR,		/* UINT64 */
75fa9e406ahrens	/* spa: */
76fa9e406ahrens	DMU_OT_SPACE_MAP_HEADER,	/* UINT64 */
77fa9e406ahrens	DMU_OT_SPACE_MAP,		/* UINT64 */
78fa9e406ahrens	/* zil: */
79fa9e406ahrens	DMU_OT_INTENT_LOG,		/* UINT64 */
80fa9e406ahrens	/* dmu: */
81fa9e406ahrens	DMU_OT_DNODE,			/* DNODE */
82fa9e406ahrens	DMU_OT_OBJSET,			/* OBJSET */
83fa9e406ahrens	/* dsl: */
841649cd4tabriz	DMU_OT_DSL_DIR,			/* UINT64 */
8587e5029ahrens	DMU_OT_DSL_DIR_CHILD_MAP,	/* ZAP */
8687e5029ahrens	DMU_OT_DSL_DS_SNAP_MAP,		/* ZAP */
87fa9e406ahrens	DMU_OT_DSL_PROPS,		/* ZAP */
881649cd4tabriz	DMU_OT_DSL_DATASET,		/* UINT64 */
89fa9e406ahrens	/* zpl: */
90fa9e406ahrens	DMU_OT_ZNODE,			/* ZNODE */
91fa9e406ahrens	DMU_OT_ACL,			/* ACL */
92fa9e406ahrens	DMU_OT_PLAIN_FILE_CONTENTS,	/* UINT8 */
93fa9e406ahrens	DMU_OT_DIRECTORY_CONTENTS,	/* ZAP */
94fa9e406ahrens	DMU_OT_MASTER_NODE,		/* ZAP */
95fa9e406ahrens	DMU_OT_DELETE_QUEUE,		/* ZAP */
96fa9e406ahrens	/* zvol: */
97fa9e406ahrens	DMU_OT_ZVOL,			/* UINT8 */
98fa9e406ahrens	DMU_OT_ZVOL_PROP,		/* ZAP */
99fa9e406ahrens	/* other; for testing only! */
100fa9e406ahrens	DMU_OT_PLAIN_OTHER,		/* UINT8 */
101fa9e406ahrens	DMU_OT_UINT64_OTHER,		/* UINT64 */
102fa9e406ahrens	DMU_OT_ZAP_OTHER,		/* ZAP */
103ea8dc4beschrock	/* new object types: */
104ea8dc4beschrock	DMU_OT_ERROR_LOG,		/* ZAP */
105fa9e406ahrens
106fa9e406ahrens	DMU_OT_NUMTYPES
107fa9e406ahrens} dmu_object_type_t;
108fa9e406ahrens
109fa9e406ahrenstypedef enum dmu_objset_type {
110fa9e406ahrens	DMU_OST_NONE,
111fa9e406ahrens	DMU_OST_META,
112fa9e406ahrens	DMU_OST_ZFS,
113fa9e406ahrens	DMU_OST_ZVOL,
114fa9e406ahrens	DMU_OST_OTHER,			/* For testing only! */
115fa9e406ahrens	DMU_OST_ANY,			/* Be careful! */
116fa9e406ahrens	DMU_OST_NUMTYPES
117fa9e406ahrens} dmu_objset_type_t;
118fa9e406ahrens
119fa9e406ahrensvoid byteswap_uint64_array(void *buf, size_t size);
120fa9e406ahrensvoid byteswap_uint32_array(void *buf, size_t size);
121fa9e406ahrensvoid byteswap_uint16_array(void *buf, size_t size);
122fa9e406ahrensvoid byteswap_uint8_array(void *buf, size_t size);
123fa9e406ahrensvoid zap_byteswap(void *buf, size_t size);
124fa9e406ahrensvoid zfs_acl_byteswap(void *buf, size_t size);
125fa9e406ahrensvoid zfs_znode_byteswap(void *buf, size_t size);
126fa9e406ahrens
127fa9e406ahrens#define	DS_MODE_NONE		0	/* invalid, to aid debugging */
128fa9e406ahrens#define	DS_MODE_STANDARD	1	/* normal access, no special needs */
129fa9e406ahrens#define	DS_MODE_PRIMARY		2	/* the "main" access, e.g. a mount */
130fa9e406ahrens#define	DS_MODE_EXCLUSIVE	3	/* exclusive access, e.g. to destroy */
131fa9e406ahrens#define	DS_MODE_LEVELS		4
132fa9e406ahrens#define	DS_MODE_LEVEL(x)	((x) & (DS_MODE_LEVELS - 1))
133fa9e406ahrens#define	DS_MODE_READONLY	0x8
134fa9e406ahrens#define	DS_MODE_IS_READONLY(x)	((x) & DS_MODE_READONLY)
135e193023bonwick#define	DS_MODE_INCONSISTENT	0x10
136e193023bonwick#define	DS_MODE_IS_INCONSISTENT(x)	((x) & DS_MODE_INCONSISTENT)
137fa9e406ahrens
138fa9e406ahrens#define	DS_FIND_SNAPSHOTS	0x01
139fa9e406ahrens
140fa9e406ahrens/*
141fa9e406ahrens * The maximum number of bytes that can be accessed as part of one
142fa9e406ahrens * operation, including metadata.
143fa9e406ahrens */
144fa9e406ahrens#define	DMU_MAX_ACCESS (10<<20) /* 10MB */
145fa9e406ahrens
146fa9e406ahrens/*
147fa9e406ahrens * Public routines to create, destroy, open, and close objsets.
148fa9e406ahrens */
149fa9e406ahrensint dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
150fa9e406ahrens    objset_t **osp);
151fa9e406ahrensvoid dmu_objset_close(objset_t *os);
152436b295perrinint dmu_objset_evict_dbufs(objset_t *os, int try);
153fa9e406ahrensint dmu_objset_create(const char *name, dmu_objset_type_t type,
154fa9e406ahrens    objset_t *clone_parent,
155fa9e406ahrens    void (*func)(objset_t *os, void *arg, dmu_tx_t *tx), void *arg);
156fa9e406ahrensint dmu_objset_destroy(const char *name);
157fa9e406ahrensint dmu_objset_rollback(const char *name);
158fa9e406ahrensint dmu_objset_rename(const char *name, const char *newname);
159fa9e406ahrensvoid dmu_objset_set_quota(objset_t *os, uint64_t quota);
160fa9e406ahrensuint64_t dmu_objset_get_quota(objset_t *os);
161fa9e406ahrensint dmu_objset_request_reservation(objset_t *os, uint64_t reservation);
162fa9e406ahrensvoid dmu_objset_find(char *name, void func(char *, void *), void *arg,
163fa9e406ahrens    int flags);
164fa9e406ahrensvoid dmu_objset_byteswap(void *buf, size_t size);
165fa9e406ahrens
166fa9e406ahrenstypedef struct dmu_buf {
167fa9e406ahrens	uint64_t db_object;		/* object that this buffer is part of */
168fa9e406ahrens	uint64_t db_offset;		/* byte offset in this object */
169fa9e406ahrens	uint64_t db_size;		/* size of buffer in bytes */
170fa9e406ahrens	void *db_data;			/* data in buffer */
171fa9e406ahrens} dmu_buf_t;
172fa9e406ahrens
173fa9e406ahrenstypedef void dmu_buf_evict_func_t(struct dmu_buf *db, void *user_ptr);
174fa9e406ahrens
175fa9e406ahrens/*
176fa9e406ahrens * Callback function to perform byte swapping on a block.
177fa9e406ahrens */
178fa9e406ahrenstypedef void dmu_byteswap_func_t(void *buf, size_t size);
179fa9e406ahrens
180fa9e406ahrens#define	DMU_POOL_DIRECTORY_OBJECT	1
181fa9e406ahrens#define	DMU_POOL_CONFIG			"config"
182fa9e406ahrens#define	DMU_POOL_ROOT_DATASET		"root_dataset"
183fa9e406ahrens#define	DMU_POOL_SYNC_BPLIST		"sync_bplist"
184ea8dc4beschrock#define	DMU_POOL_ERRLOG_SCRUB		"errlog_scrub"
185ea8dc4beschrock#define	DMU_POOL_ERRLOG_LAST		"errlog_last"
186fa9e406ahrens
187fa9e406ahrens/*
188fa9e406ahrens * Allocate an object from this objset.  The range of object numbers
189fa9e406ahrens * available is (0, DN_MAX_OBJECT).  Object 0 is the meta-dnode.
190fa9e406ahrens *
191fa9e406ahrens * The transaction must be assigned to a txg.  The newly allocated
192fa9e406ahrens * object will be "held" in the transaction (ie. you can modify the
193fa9e406ahrens * newly allocated object in this transaction).
194fa9e406ahrens *
195fa9e406ahrens * dmu_object_alloc() chooses an object and returns it in *objectp.
196fa9e406ahrens *
197fa9e406ahrens * dmu_object_claim() allocates a specific object number.  If that
198fa9e406ahrens * number is already allocated, it fails and returns EEXIST.
199fa9e406ahrens *
200fa9e406ahrens * Return 0 on success, or ENOSPC or EEXIST as specified above.
201fa9e406ahrens */
202fa9e406ahrensuint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
203fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
204fa9e406ahrensint dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
205fa9e406ahrens    int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
206fa9e406ahrensint dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
207fa9e406ahrens    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
208fa9e406ahrens
209fa9e406ahrens/*
210fa9e406ahrens * Free an object from this objset.
211fa9e406ahrens *
212fa9e406ahrens * The object's data will be freed as well (ie. you don't need to call
213fa9e406ahrens * dmu_free(object, 0, -1, tx)).
214fa9e406ahrens *
215fa9e406ahrens * The object need not be held in the transaction.
216fa9e406ahrens *
217fa9e406ahrens * If there are any holds on this object's buffers (via dmu_buf_hold()),
218fa9e406ahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not
219fa9e406ahrens * free it; it fails and returns EBUSY.
220fa9e406ahrens *
221fa9e406ahrens * If the object is not allocated, it fails and returns ENOENT.
222fa9e406ahrens *
223fa9e406ahrens * Return 0 on success, or EBUSY or ENOENT as specified above.
224fa9e406ahrens */
225fa9e406ahrensint dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
226fa9e406ahrens
227fa9e406ahrens/*
228fa9e406ahrens * Find the next allocated or free object.
229fa9e406ahrens *
230fa9e406ahrens * The objectp parameter is in-out.  It will be updated to be the next
231fa9e406ahrens * object which is allocated.
232fa9e406ahrens *
233fa9e406ahrens * XXX Can only be called on a objset with no dirty data.
234fa9e406ahrens *
235fa9e406ahrens * Returns 0 on success, or ENOENT if there are no more objects.
236fa9e406ahrens */
237fa9e406ahrensint dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole);
238fa9e406ahrens
239fa9e406ahrens/*
240fa9e406ahrens * Set the data blocksize for an object.
241fa9e406ahrens *
242fa9e406ahrens * The object cannot have any blocks allcated beyond the first.  If
243fa9e406ahrens * the first block is allocated already, the new size must be greater
244fa9e406ahrens * than the current block size.  If these conditions are not met,
245fa9e406ahrens * ENOTSUP will be returned.
246fa9e406ahrens *
247fa9e406ahrens * Returns 0 on success, or EBUSY if there are any holds on the object
248fa9e406ahrens * contents, or ENOTSUP as described above.
249fa9e406ahrens */
250fa9e406ahrensint dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
251fa9e406ahrens    int ibs, dmu_tx_t *tx);
252fa9e406ahrens
253fa9e406ahrens/*
254fa9e406ahrens * Set the checksum property on a dnode.  The new checksum algorithm will
255fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
256fa9e406ahrens */
257fa9e406ahrensvoid dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
258fa9e406ahrens    dmu_tx_t *tx);
259fa9e406ahrens
260fa9e406ahrens/*
261fa9e406ahrens * Set the compress property on a dnode.  The new compression algorithm will
262fa9e406ahrens * apply to all newly written blocks; existing blocks will not be affected.
263fa9e406ahrens */
264fa9e406ahrensvoid dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
265fa9e406ahrens    dmu_tx_t *tx);
266fa9e406ahrens
267fa9e406ahrens/*
26844cd46cbillm * Decide how many copies of a given block we should make.  Can be from
26944cd46cbillm * 1 to SPA_DVAS_PER_BP.
27044cd46cbillm */
27144cd46cbillmint dmu_get_replication_level(struct spa *spa, struct zbookmark *zb,
27244cd46cbillm    dmu_object_type_t ot);
27344cd46cbillm/*
274fa9e406ahrens * The bonus data is accessed more or less like a regular buffer.
275fa9e406ahrens * You must dmu_bonus_hold() to get the buffer, which will give you a
276fa9e406ahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
277fa9e406ahrens * data.  As with any normal buffer, you must call dmu_buf_read() to
278fa9e406ahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the
279fa9e406ahrens * object must be held in an assigned transaction before calling
280fa9e406ahrens * dmu_buf_will_dirty.  You may use dmu_buf_set_user() on the bonus
281fa9e406ahrens * buffer as well.  You must release your hold with dmu_buf_rele().
282fa9e406ahrens */
283ea8dc4beschrockint dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **);
284fa9e406ahrensint dmu_bonus_max(void);
285fa9e406ahrens
286fa9e406ahrens/*
287fa9e406ahrens * Obtain the DMU buffer from the specified object which contains the
288fa9e406ahrens * specified offset.  dmu_buf_hold() puts a "hold" on the buffer, so
289fa9e406ahrens * that it will remain in memory.  You must release the hold with
290fa9e406ahrens * dmu_buf_rele().  You musn't access the dmu_buf_t after releasing your
291fa9e406ahrens * hold.  You must have a hold on any dmu_buf_t* you pass to the DMU.
292fa9e406ahrens *
293fa9e406ahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
294fa9e406ahrens * on the returned buffer before reading or writing the buffer's
295fa9e406ahrens * db_data.  The comments for those routines describe what particular
296fa9e406ahrens * operations are valid after calling them.
297fa9e406ahrens *
298fa9e406ahrens * The object number must be a valid, allocated object number.
299fa9e406ahrens */
300ea8dc4beschrockint dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
301ea8dc4beschrock    void *tag, dmu_buf_t **);
302fa9e406ahrensvoid dmu_buf_add_ref(dmu_buf_t *db, void* tag);
303ea8dc4beschrockvoid dmu_buf_rele(dmu_buf_t *db, void *tag);
304fa9e406ahrensuint64_t dmu_buf_refcount(dmu_buf_t *db);
305fa9e406ahrens
306fa9e406ahrens/*
307fa9e406ahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
308fa9e406ahrens * range of an object.  A pointer to an array of dmu_buf_t*'s is
309fa9e406ahrens * returned (in *dbpp).
310fa9e406ahrens *
311fa9e406ahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
312fa9e406ahrens * frees the array.  The hold on the array of buffers MUST be released
313fa9e406ahrens * with dmu_buf_rele_array.  You can NOT release the hold on each buffer
314fa9e406ahrens * individually with dmu_buf_rele.
315fa9e406ahrens */
316ea8dc4beschrockint dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
317ea8dc4beschrock    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
318ea8dc4beschrockvoid dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
319fa9e406ahrens
320fa9e406ahrens/*
321fa9e406ahrens * Returns NULL on success, or the existing user ptr if it's already
322fa9e406ahrens * been set.
323fa9e406ahrens *
324fa9e406ahrens * user_ptr is for use by the user and can be obtained via dmu_buf_get_user().
325fa9e406ahrens *
326fa9e406ahrens * user_data_ptr_ptr should be NULL, or a pointer to a pointer which
327fa9e406ahrens * will be set to db->db_data when you are allowed to access it.  Note
328fa9e406ahrens * that db->db_data (the pointer) can change when you do dmu_buf_read(),
329fa9e406ahrens * dmu_buf_tryupgrade(), dmu_buf_will_dirty(), or dmu_buf_will_fill().
330fa9e406ahrens * *user_data_ptr_ptr will be set to the new value when it changes.
331fa9e406ahrens *
332fa9e406ahrens * If non-NULL, pageout func will be called when this buffer is being
333fa9e406ahrens * excised from the cache, so that you can clean up the data structure
334fa9e406ahrens * pointed to by user_ptr.
335fa9e406ahrens *
336fa9e406ahrens * dmu_evict_user() will call the pageout func for all buffers in a
337fa9e406ahrens * objset with a given pageout func.
338fa9e406ahrens */
339fa9e406ahrensvoid *dmu_buf_set_user(dmu_buf_t *db, void *user_ptr, void *user_data_ptr_ptr,
340fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
341fa9e406ahrens/*
342fa9e406ahrens * set_user_ie is the same as set_user, but request immediate eviction
343fa9e406ahrens * when hold count goes to zero.
344fa9e406ahrens */
345fa9e406ahrensvoid *dmu_buf_set_user_ie(dmu_buf_t *db, void *user_ptr,
346fa9e406ahrens    void *user_data_ptr_ptr, dmu_buf_evict_func_t *pageout_func);
347fa9e406ahrensvoid *dmu_buf_update_user(dmu_buf_t *db_fake, void *old_user_ptr,
348fa9e406ahrens    void *user_ptr, void *user_data_ptr_ptr,
349fa9e406ahrens    dmu_buf_evict_func_t *pageout_func);
350fa9e406ahrensvoid dmu_evict_user(objset_t *os, dmu_buf_evict_func_t *func);
351fa9e406ahrens
352fa9e406ahrensvoid dmu_buf_hold_data(dmu_buf_t *db);
353fa9e406ahrensvoid dmu_buf_rele_data(dmu_buf_t *db);
354fa9e406ahrens
355fa9e406ahrens/*
356fa9e406ahrens * Returns the user_ptr set with dmu_buf_set_user(), or NULL if not set.
357fa9e406ahrens */
358fa9e406ahrensvoid *dmu_buf_get_user(dmu_buf_t *db);
359fa9e406ahrens
360fa9e406ahrens/*
361fa9e406ahrens * Indicate that you are going to modify the buffer's data (db_data).
362fa9e406ahrens *
363fa9e406ahrens * The transaction (tx) must be assigned to a txg (ie. you've called
364fa9e406ahrens * dmu_tx_assign()).  The buffer's object must be held in the tx
365fa9e406ahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
366fa9e406ahrens */
367fa9e406ahrensvoid dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
368fa9e406ahrens
369fa9e406ahrens/*
370fa9e406ahrens * You must create a transaction, then hold the objects which you will
371fa9e406ahrens * (or might) modify as part of this transaction.  Then you must assign
372fa9e406ahrens * the transaction to a transaction group.  Once the transaction has
373fa9e406ahrens * been assigned, you can modify buffers which belong to held objects as
374fa9e406ahrens * part of this transaction.  You can't modify buffers before the
375fa9e406ahrens * transaction has been assigned; you can't modify buffers which don't
376fa9e406ahrens * belong to objects which this transaction holds; you can't hold
377fa9e406ahrens * objects once the transaction has been assigned.  You may hold an
378fa9e406ahrens * object which you are going to free (with dmu_object_free()), but you
379fa9e406ahrens * don't have to.
380fa9e406ahrens *
381fa9e406ahrens * You can abort the transaction before it has been assigned.
382fa9e406ahrens *
383fa9e406ahrens * Note that you may hold buffers (with dmu_buf_hold) at any time,
384fa9e406ahrens * regardless of transaction state.
385fa9e406ahrens */
386fa9e406ahrens
387fa9e406ahrens#define	DMU_NEW_OBJECT	(-1ULL)
388fa9e406ahrens#define	DMU_OBJECT_END	(-1ULL)
389fa9e406ahrens
390fa9e406ahrensdmu_tx_t *dmu_tx_create(objset_t *os);
391fa9e406ahrensvoid dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
392fa9e406ahrensvoid dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
393fa9e406ahrens    uint64_t len);
394ea8dc4beschrockvoid dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, char *name);
395fa9e406ahrensvoid dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
396fa9e406ahrensvoid dmu_tx_abort(dmu_tx_t *tx);
397fa9e406ahrensint dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
398fa9e406ahrensvoid dmu_tx_commit(dmu_tx_t *tx);
399fa9e406ahrens
400fa9e406ahrens/*
401fa9e406ahrens * Free up the data blocks for a defined range of a file.  If size is
402fa9e406ahrens * zero, the range from offset to end-of-file is freed.
403fa9e406ahrens */
404ea8dc4beschrockint dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
405fa9e406ahrens	uint64_t size, dmu_tx_t *tx);
406fa9e406ahrens
407fa9e406ahrens/*
408fa9e406ahrens * Convenience functions.
409fa9e406ahrens *
410fa9e406ahrens * Canfail routines will return 0 on success, or an errno if there is a
411fa9e406ahrens * nonrecoverable I/O error.
412fa9e406ahrens */
413ea8dc4beschrockint dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
414fa9e406ahrens	void *buf);
415fa9e406ahrensvoid dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
416fa9e406ahrens	const void *buf, dmu_tx_t *tx);
417fa9e406ahrensint dmu_write_uio(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
418fa9e406ahrens    struct uio *uio, dmu_tx_t *tx);
419fa9e406ahrens
420fa9e406ahrens/*
421fa9e406ahrens * Asynchronously try to read in the data.
422fa9e406ahrens */
423fa9e406ahrensvoid dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset,
424fa9e406ahrens    uint64_t len);
425fa9e406ahrens
426fa9e406ahrenstypedef struct dmu_object_info {
427fa9e406ahrens	/* All sizes are in bytes. */
428fa9e406ahrens	uint32_t doi_data_block_size;
429fa9e406ahrens	uint32_t doi_metadata_block_size;
430fa9e406ahrens	uint64_t doi_bonus_size;
431fa9e406ahrens	dmu_object_type_t doi_type;
432fa9e406ahrens	dmu_object_type_t doi_bonus_type;
433fa9e406ahrens	uint8_t doi_indirection;		/* 2 = dnode->indirect->data */
434fa9e406ahrens	uint8_t doi_checksum;
435fa9e406ahrens	uint8_t doi_compress;
436fa9e406ahrens	uint8_t doi_pad[5];
437fa9e406ahrens	/* Values below are number of 512-byte blocks. */
438fa9e406ahrens	uint64_t doi_physical_blks;		/* data + metadata */
439fa9e406ahrens	uint64_t doi_max_block_offset;
440fa9e406ahrens} dmu_object_info_t;
441fa9e406ahrens
442fa9e406ahrenstypedef struct dmu_object_type_info {
443fa9e406ahrens	dmu_byteswap_func_t	*ot_byteswap;
444fa9e406ahrens	boolean_t		ot_metadata;
445fa9e406ahrens	char			*ot_name;
446fa9e406ahrens} dmu_object_type_info_t;
447fa9e406ahrens
448fa9e406ahrensextern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
449fa9e406ahrens
450fa9e406ahrens/*
451fa9e406ahrens * Get information on a DMU object.
452fa9e406ahrens *
453fa9e406ahrens * Return 0 on success or ENOENT if object is not allocated.
454fa9e406ahrens *
455fa9e406ahrens * If doi is NULL, just indicates whether the object exists.
456fa9e406ahrens */
457fa9e406ahrensint dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
458fa9e406ahrensvoid dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
459fa9e406ahrensvoid dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
460fa9e406ahrensvoid dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
461fa9e406ahrens    u_longlong_t *nblk512);
462fa9e406ahrens
463fa9e406ahrens/*
464fa9e406ahrens * Get the maximum nonzero offset in the object (ie. this offset and all
465fa9e406ahrens * offsets following are zero).
466fa9e406ahrens *
467fa9e406ahrens * XXX Perhaps integrate this with dmu_object_info(), although that
468fa9e406ahrens * would then have to bring in the indirect blocks.
469fa9e406ahrens */
470fa9e406ahrensuint64_t dmu_object_max_nonzero_offset(objset_t *os, uint64_t object);
471fa9e406ahrens
472fa9e406ahrenstypedef struct dmu_objset_stats {
473fa9e406ahrens	dmu_objset_type_t dds_type;
474fa9e406ahrens	uint8_t dds_is_snapshot;
47531fd60dahrens	uint8_t dds_inconsistent;
47631fd60dahrens	uint8_t dds_pad[2];
477fa9e406ahrens
478fa9e406ahrens	uint64_t dds_creation_time;
479fa9e406ahrens	uint64_t dds_creation_txg;
480fa9e406ahrens
481fa9e406ahrens	char dds_clone_of[MAXNAMELEN];
482fa9e406ahrens
483fa9e406ahrens	/* How much data is there in this objset? */
484fa9e406ahrens
485fa9e406ahrens	/*
486fa9e406ahrens	 * Space referenced, taking into account pending writes and
487fa9e406ahrens	 * frees.  Only relavent to filesystems and snapshots (not
488fa9e406ahrens	 * collections).
489fa9e406ahrens	 */
490fa9e406ahrens	uint64_t dds_space_refd;
491fa9e406ahrens
492fa9e406ahrens	/*
493fa9e406ahrens	 * Space "used", taking into account pending writes and frees, and
494fa9e406ahrens	 * children's reservations (in bytes).  This is the amount of
495fa9e406ahrens	 * space that will be freed if this and all dependent items are
496fa9e406ahrens	 * destroyed (eg. child datasets, objsets, and snapshots).  So
497fa9e406ahrens	 * for snapshots, this is the amount of space unique to this
498fa9e406ahrens	 * snapshot.
499fa9e406ahrens	 */
500fa9e406ahrens	uint64_t dds_space_used;
501fa9e406ahrens
502fa9e406ahrens	/*
503fa9e406ahrens	 * Compressed and uncompressed bytes consumed.  Does not take
504fa9e406ahrens	 * into account reservations.  Used for computing compression
505fa9e406ahrens	 * ratio.
506fa9e406ahrens	 */
507fa9e406ahrens	uint64_t dds_compressed_bytes;
508fa9e406ahrens	uint64_t dds_uncompressed_bytes;
509fa9e406ahrens
510fa9e406ahrens	/*
511fa9e406ahrens	 * The ds_fsid_guid is a 56-bit ID that can change to avoid
512fa9e406ahrens	 * collisions.  The ds_guid is a 64-bit ID that will never
513fa9e406ahrens	 * change, so there is a small probability that it will collide.
514fa9e406ahrens	 */
515fa9e406ahrens	uint64_t dds_fsid_guid;
516fa9e406ahrens
517fa9e406ahrens	uint64_t dds_objects_used;	/* number of objects used */
518fa9e406ahrens	uint64_t dds_objects_avail;	/* number of objects available */
519fa9e406ahrens
520fa9e406ahrens	uint64_t dds_num_clones; /* number of clones of this */
521fa9e406ahrens
522fa9e406ahrens	/* The dataset's administratively-set quota, in bytes. */
523fa9e406ahrens	uint64_t dds_quota;
524fa9e406ahrens
525fa9e406ahrens	/* The dataset's administratively-set reservation, in bytes */
526fa9e406ahrens	uint64_t dds_reserved;
527fa9e406ahrens
528fa9e406ahrens	/*
529fa9e406ahrens	 * The amount of additional space that this dataset can consume.
530fa9e406ahrens	 * Takes into account quotas & reservations.
531fa9e406ahrens	 * (Assuming that no other datasets consume it first.)
532fa9e406ahrens	 */
533fa9e406ahrens	uint64_t dds_available;
534fa9e406ahrens
535fa9e406ahrens	/*
536ea8dc4beschrock	 * Used for debugging purposes
537fa9e406ahrens	 */
538fa9e406ahrens	uint64_t dds_last_txg;
539fa9e406ahrens} dmu_objset_stats_t;
540fa9e406ahrens
541fa9e406ahrens/*
542fa9e406ahrens * Get stats on a dataset.
543fa9e406ahrens */
544fa9e406ahrensvoid dmu_objset_stats(objset_t *os, dmu_objset_stats_t *dds);
545fa9e406ahrens
546fa9e406ahrensint dmu_objset_is_snapshot(objset_t *os);
547fa9e406ahrens
548fa9e406ahrensextern struct spa *dmu_objset_spa(objset_t *os);
549fa9e406ahrensextern struct zilog *dmu_objset_zil(objset_t *os);
550fa9e406ahrensextern struct dsl_pool *dmu_objset_pool(objset_t *os);
551fa9e406ahrensextern struct dsl_dataset *dmu_objset_ds(objset_t *os);
552fa9e406ahrensextern void dmu_objset_name(objset_t *os, char *buf);
553fa9e406ahrensextern dmu_objset_type_t dmu_objset_type(objset_t *os);
554fa9e406ahrensextern uint64_t dmu_objset_id(objset_t *os);
555fa9e406ahrensextern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
556fa9e406ahrens    uint64_t *id, uint64_t *offp);
55787e5029ahrensextern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
55887e5029ahrens    uint64_t *idp, uint64_t *offp);
559fa9e406ahrens
560fa9e406ahrens/*
561fa9e406ahrens * Return the txg number for the given assigned transaction.
562fa9e406ahrens */
56387e5029ahrensuint64_t dmu_tx_get_txg(dmu_tx_t *tx);
564fa9e406ahrens
565fa9e406ahrens/*
566fa9e406ahrens * Synchronous write.
567fa9e406ahrens * On success returns 0 and fills in the blk pointed at by bp.
568fa9e406ahrens * Note that while the data covered by this function will be on stable
569fa9e406ahrens * storage when the function returns this new data does not become a
570fa9e406ahrens * permanent part of the file until the associated transaction commits.
571fa9e406ahrens */
572fa9e406ahrensint dmu_sync(objset_t *os, uint64_t object, uint64_t offset, uint64_t *blkoff,
573fa9e406ahrens    struct blkptr *bp, uint64_t txg);
574fa9e406ahrens
575fa9e406ahrens/*
576fa9e406ahrens * Find the next hole or data block in file starting at *off
577fa9e406ahrens * Return found offset in *off. Return ESRCH for end of file.
578fa9e406ahrens */
579fa9e406ahrensint dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
580fa9e406ahrens    uint64_t *off);
581fa9e406ahrens
582fa9e406ahrens/*
583fa9e406ahrens * Initial setup and final teardown.
584fa9e406ahrens */
585fa9e406ahrensextern void dmu_init(void);
586fa9e406ahrensextern void dmu_fini(void);
587fa9e406ahrens
588fa9e406ahrenstypedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
589fa9e406ahrens    uint64_t object, uint64_t offset, int len);
590fa9e406ahrensvoid dmu_traverse_objset(objset_t *os, uint64_t txg_start,
591fa9e406ahrens    dmu_traverse_cb_t cb, void *arg);
592fa9e406ahrens
593fa9e406ahrensint dmu_sendbackup(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp);
594ea8dc4beschrockint dmu_recvbackup(char *tosnap, struct drr_begin *drrb, uint64_t *sizep,
595fa9e406ahrens    struct vnode *vp, uint64_t voffset);
596fa9e406ahrens
597fa9e406ahrens/* CRC64 table */
598fa9e406ahrens#define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
599fa9e406ahrensextern uint64_t zfs_crc64_table[256];
600fa9e406ahrens
601fa9e406ahrens#ifdef	__cplusplus
602fa9e406ahrens}
603fa9e406ahrens#endif
604fa9e406ahrens
605fa9e406ahrens#endif	/* _SYS_DMU_H */
606