1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26#ifndef _SYS_DDT_H
27#define	_SYS_DDT_H
28
29#include <sys/sysmacros.h>
30#include <sys/types.h>
31#include <sys/fs/zfs.h>
32#include <sys/zio.h>
33#include <sys/dmu.h>
34
35#ifdef	__cplusplus
36extern "C" {
37#endif
38
39/*
40 * On-disk DDT formats, in the desired search order (newest version first).
41 */
42enum ddt_type {
43	DDT_TYPE_ZAP = 0,
44	DDT_TYPES
45};
46
47/*
48 * DDT classes, in the desired search order (highest replication level first).
49 */
50enum ddt_class {
51	DDT_CLASS_DITTO = 0,
52	DDT_CLASS_DUPLICATE,
53	DDT_CLASS_UNIQUE,
54	DDT_CLASSES
55};
56
57#define	DDT_TYPE_CURRENT		0
58
59#define	DDT_COMPRESS_BYTEORDER_MASK	0x80
60#define	DDT_COMPRESS_FUNCTION_MASK	0x7f
61
62/*
63 * On-disk ddt entry:  key (name) and physical storage (value).
64 */
65typedef struct ddt_key {
66	zio_cksum_t	ddk_cksum;	/* 256-bit block checksum */
67	uint64_t	ddk_prop;	/* LSIZE, PSIZE, compression */
68} ddt_key_t;
69
70/*
71 * ddk_prop layout:
72 *
73 *	+-------+-------+-------+-------+-------+-------+-------+-------+
74 *	|   0	|   0	|   0	| comp	|     PSIZE	|     LSIZE	|
75 *	+-------+-------+-------+-------+-------+-------+-------+-------+
76 */
77#define	DDK_GET_LSIZE(ddk)	\
78	BF64_GET_SB((ddk)->ddk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1)
79#define	DDK_SET_LSIZE(ddk, x)	\
80	BF64_SET_SB((ddk)->ddk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
81
82#define	DDK_GET_PSIZE(ddk)	\
83	BF64_GET_SB((ddk)->ddk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1)
84#define	DDK_SET_PSIZE(ddk, x)	\
85	BF64_SET_SB((ddk)->ddk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x)
86
87#define	DDK_GET_COMPRESS(ddk)		BF64_GET((ddk)->ddk_prop, 32, 8)
88#define	DDK_SET_COMPRESS(ddk, x)	BF64_SET((ddk)->ddk_prop, 32, 8, x)
89
90#define	DDT_KEY_WORDS	(sizeof (ddt_key_t) / sizeof (uint64_t))
91
92typedef struct ddt_phys {
93	dva_t		ddp_dva[SPA_DVAS_PER_BP];
94	uint64_t	ddp_refcnt;
95	uint64_t	ddp_phys_birth;
96} ddt_phys_t;
97
98enum ddt_phys_type {
99	DDT_PHYS_DITTO = 0,
100	DDT_PHYS_SINGLE = 1,
101	DDT_PHYS_DOUBLE = 2,
102	DDT_PHYS_TRIPLE = 3,
103	DDT_PHYS_TYPES
104} ddt_phys_type_t;
105
106/*
107 * In-core ddt entry
108 */
109struct ddt_entry {
110	ddt_key_t	dde_key;
111	ddt_phys_t	dde_phys[DDT_PHYS_TYPES];
112	zio_t		*dde_lead_zio[DDT_PHYS_TYPES];
113	void		*dde_repair_data;
114	enum ddt_type	dde_type;
115	enum ddt_class	dde_class;
116	uint8_t		dde_loading;
117	uint8_t		dde_loaded;
118	kcondvar_t	dde_cv;
119	avl_node_t	dde_node;
120};
121
122/*
123 * In-core ddt
124 */
125struct ddt {
126	kmutex_t	ddt_lock;
127	avl_tree_t	ddt_tree;
128	avl_tree_t	ddt_repair_tree;
129	enum zio_checksum ddt_checksum;
130	spa_t		*ddt_spa;
131	objset_t	*ddt_os;
132	uint64_t	ddt_stat_object;
133	uint64_t	ddt_object[DDT_TYPES][DDT_CLASSES];
134	ddt_histogram_t	ddt_histogram[DDT_TYPES][DDT_CLASSES];
135	avl_node_t	ddt_node;
136};
137
138/*
139 * In-core and on-disk bookmark for DDT walks
140 */
141typedef struct ddt_bookmark {
142	uint64_t	ddb_class;
143	uint64_t	ddb_type;
144	uint64_t	ddb_checksum;
145	uint64_t	ddb_cursor;
146} ddt_bookmark_t;
147
148/*
149 * Ops vector to access a specific DDT object type.
150 */
151typedef struct ddt_ops {
152	char ddt_op_name[32];
153	int (*ddt_op_create)(objset_t *os, uint64_t *object, dmu_tx_t *tx,
154	    boolean_t prehash);
155	int (*ddt_op_destroy)(objset_t *os, uint64_t object, dmu_tx_t *tx);
156	int (*ddt_op_lookup)(objset_t *os, uint64_t object, ddt_entry_t *dde);
157	int (*ddt_op_update)(objset_t *os, uint64_t object, ddt_entry_t *dde,
158	    dmu_tx_t *tx);
159	int (*ddt_op_remove)(objset_t *os, uint64_t object, ddt_entry_t *dde,
160	    dmu_tx_t *tx);
161	int (*ddt_op_walk)(objset_t *os, uint64_t object, ddt_entry_t *dde,
162	    uint64_t *walk);
163	uint64_t (*ddt_op_count)(objset_t *os, uint64_t object);
164} ddt_ops_t;
165
166#define	DDT_NAMELEN	80
167
168extern void ddt_object_name(ddt_t *ddt, enum ddt_type type,
169    enum ddt_class class, char *name);
170extern int ddt_object_walk(ddt_t *ddt, enum ddt_type type,
171    enum ddt_class class, uint64_t *walk, ddt_entry_t *dde);
172extern uint64_t ddt_object_count(ddt_t *ddt, enum ddt_type type,
173    enum ddt_class class);
174extern int ddt_object_info(ddt_t *ddt, enum ddt_type type,
175    enum ddt_class class, dmu_object_info_t *);
176extern boolean_t ddt_object_exists(ddt_t *ddt, enum ddt_type type,
177    enum ddt_class class);
178
179extern void ddt_bp_fill(const ddt_phys_t *ddp, blkptr_t *bp,
180    uint64_t txg);
181extern void ddt_bp_create(enum zio_checksum checksum, const ddt_key_t *ddk,
182    const ddt_phys_t *ddp, blkptr_t *bp);
183
184extern void ddt_key_fill(ddt_key_t *ddk, const blkptr_t *bp);
185
186extern void ddt_phys_fill(ddt_phys_t *ddp, const blkptr_t *bp);
187extern void ddt_phys_clear(ddt_phys_t *ddp);
188extern void ddt_phys_addref(ddt_phys_t *ddp);
189extern void ddt_phys_decref(ddt_phys_t *ddp);
190extern void ddt_phys_free(ddt_t *ddt, ddt_key_t *ddk, ddt_phys_t *ddp,
191    uint64_t txg);
192extern ddt_phys_t *ddt_phys_select(const ddt_entry_t *dde, const blkptr_t *bp);
193extern uint64_t ddt_phys_total_refcnt(const ddt_entry_t *dde);
194
195extern void ddt_stat_add(ddt_stat_t *dst, const ddt_stat_t *src, uint64_t neg);
196
197extern void ddt_histogram_add(ddt_histogram_t *dst, const ddt_histogram_t *src);
198extern void ddt_histogram_stat(ddt_stat_t *dds, const ddt_histogram_t *ddh);
199extern boolean_t ddt_histogram_empty(const ddt_histogram_t *ddh);
200extern void ddt_get_dedup_object_stats(spa_t *spa, ddt_object_t *ddo);
201extern void ddt_get_dedup_histogram(spa_t *spa, ddt_histogram_t *ddh);
202extern void ddt_get_dedup_stats(spa_t *spa, ddt_stat_t *dds_total);
203
204extern uint64_t ddt_get_dedup_dspace(spa_t *spa);
205extern uint64_t ddt_get_pool_dedup_ratio(spa_t *spa);
206
207extern int ddt_ditto_copies_needed(ddt_t *ddt, ddt_entry_t *dde,
208    ddt_phys_t *ddp_willref);
209extern int ddt_ditto_copies_present(ddt_entry_t *dde);
210
211extern size_t ddt_compress(void *src, uchar_t *dst, size_t s_len, size_t d_len);
212extern void ddt_decompress(uchar_t *src, void *dst, size_t s_len, size_t d_len);
213
214extern ddt_t *ddt_select(spa_t *spa, const blkptr_t *bp);
215extern void ddt_enter(ddt_t *ddt);
216extern void ddt_exit(ddt_t *ddt);
217extern ddt_entry_t *ddt_lookup(ddt_t *ddt, const blkptr_t *bp, boolean_t add);
218extern void ddt_remove(ddt_t *ddt, ddt_entry_t *dde);
219
220extern boolean_t ddt_class_contains(spa_t *spa, enum ddt_class max_class,
221    const blkptr_t *bp);
222
223extern ddt_entry_t *ddt_repair_start(ddt_t *ddt, const blkptr_t *bp);
224extern void ddt_repair_done(ddt_t *ddt, ddt_entry_t *dde);
225
226extern int ddt_entry_compare(const void *x1, const void *x2);
227
228extern void ddt_create(spa_t *spa);
229extern int ddt_load(spa_t *spa);
230extern void ddt_unload(spa_t *spa);
231extern void ddt_sync(spa_t *spa, uint64_t txg);
232extern int ddt_walk(spa_t *spa, ddt_bookmark_t *ddb, ddt_entry_t *dde);
233
234extern const ddt_ops_t ddt_zap_ops;
235
236#ifdef	__cplusplus
237}
238#endif
239
240#endif	/* _SYS_DDT_H */
241