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 (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2016 by Delphix. All rights reserved.
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
39struct abd;
40
41/*
42 * On-disk DDT formats, in the desired search order (newest version first).
43 */
44enum ddt_type {
45	DDT_TYPE_ZAP = 0,
46	DDT_TYPES
47};
48
49/*
50 * DDT classes, in the desired search order (highest replication level first).
51 */
52enum ddt_class {
53	DDT_CLASS_DITTO = 0,
54	DDT_CLASS_DUPLICATE,
55	DDT_CLASS_UNIQUE,
56	DDT_CLASSES
57};
58
59#define	DDT_TYPE_CURRENT		0
60
61#define	DDT_COMPRESS_BYTEORDER_MASK	0x80
62#define	DDT_COMPRESS_FUNCTION_MASK	0x7f
63
64/*
65 * On-disk ddt entry:  key (name) and physical storage (value).
66 */
67typedef struct ddt_key {
68	zio_cksum_t	ddk_cksum;	/* 256-bit block checksum */
69	/*
70	 * Encoded with logical & physical size, encryption, and compression,
71	 * as follows:
72	 *   +-------+-------+-------+-------+-------+-------+-------+-------+
73	 *   |   0   |   0   |   0   |X| comp|     PSIZE     |     LSIZE     |
74	 *   +-------+-------+-------+-------+-------+-------+-------+-------+
75	 */
76	uint64_t	ddk_prop;
77} ddt_key_t;
78
79#define	DDK_GET_LSIZE(ddk)	\
80	BF64_GET_SB((ddk)->ddk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1)
81#define	DDK_SET_LSIZE(ddk, x)	\
82	BF64_SET_SB((ddk)->ddk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
83
84#define	DDK_GET_PSIZE(ddk)	\
85	BF64_GET_SB((ddk)->ddk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1)
86#define	DDK_SET_PSIZE(ddk, x)	\
87	BF64_SET_SB((ddk)->ddk_prop, 16, 16, SPA_MINBLOCKSHIFT, 1, x)
88
89#define	DDK_GET_COMPRESS(ddk)		BF64_GET((ddk)->ddk_prop, 32, 7)
90#define	DDK_SET_COMPRESS(ddk, x)	BF64_SET((ddk)->ddk_prop, 32, 7, x)
91
92#define	DDK_GET_CRYPT(ddk)		BF64_GET((ddk)->ddk_prop, 39, 1)
93#define	DDK_SET_CRYPT(ddk, x)	BF64_SET((ddk)->ddk_prop, 39, 1, x)
94
95#define	DDT_KEY_WORDS	(sizeof (ddt_key_t) / sizeof (uint64_t))
96
97#define	DDE_GET_NDVAS(dde) (DDK_GET_CRYPT(&dde->dde_key) \
98	? SPA_DVAS_PER_BP - 1 : SPA_DVAS_PER_BP)
99
100typedef struct ddt_phys {
101	dva_t		ddp_dva[SPA_DVAS_PER_BP];
102	uint64_t	ddp_refcnt;
103	uint64_t	ddp_phys_birth;
104} ddt_phys_t;
105
106enum ddt_phys_type {
107	DDT_PHYS_DITTO = 0,
108	DDT_PHYS_SINGLE = 1,
109	DDT_PHYS_DOUBLE = 2,
110	DDT_PHYS_TRIPLE = 3,
111	DDT_PHYS_TYPES
112};
113
114/*
115 * In-core ddt entry
116 */
117struct ddt_entry {
118	ddt_key_t	dde_key;
119	ddt_phys_t	dde_phys[DDT_PHYS_TYPES];
120	zio_t		*dde_lead_zio[DDT_PHYS_TYPES];
121	struct abd	*dde_repair_abd;
122	enum ddt_type	dde_type;
123	enum ddt_class	dde_class;
124	uint8_t		dde_loading;
125	uint8_t		dde_loaded;
126	kcondvar_t	dde_cv;
127	avl_node_t	dde_node;
128};
129
130/*
131 * In-core ddt
132 */
133struct ddt {
134	kmutex_t	ddt_lock;
135	avl_tree_t	ddt_tree;
136	avl_tree_t	ddt_repair_tree;
137	enum zio_checksum ddt_checksum;
138	spa_t		*ddt_spa;
139	objset_t	*ddt_os;
140	uint64_t	ddt_stat_object;
141	uint64_t	ddt_object[DDT_TYPES][DDT_CLASSES];
142	ddt_histogram_t	ddt_histogram[DDT_TYPES][DDT_CLASSES];
143	ddt_histogram_t	ddt_histogram_cache[DDT_TYPES][DDT_CLASSES];
144	ddt_object_t	ddt_object_stats[DDT_TYPES][DDT_CLASSES];
145	avl_node_t	ddt_node;
146};
147
148/*
149 * In-core and on-disk bookmark for DDT walks
150 */
151typedef struct ddt_bookmark {
152	uint64_t	ddb_class;
153	uint64_t	ddb_type;
154	uint64_t	ddb_checksum;
155	uint64_t	ddb_cursor;
156} ddt_bookmark_t;
157
158/*
159 * Ops vector to access a specific DDT object type.
160 */
161typedef struct ddt_ops {
162	char ddt_op_name[32];
163	int (*ddt_op_create)(objset_t *os, uint64_t *object, dmu_tx_t *tx,
164	    boolean_t prehash);
165	int (*ddt_op_destroy)(objset_t *os, uint64_t object, dmu_tx_t *tx);
166	int (*ddt_op_lookup)(objset_t *os, uint64_t object, ddt_entry_t *dde);
167	void (*ddt_op_prefetch)(objset_t *os, uint64_t object,
168	    ddt_entry_t *dde);
169	int (*ddt_op_update)(objset_t *os, uint64_t object, ddt_entry_t *dde,
170	    dmu_tx_t *tx);
171	int (*ddt_op_remove)(objset_t *os, uint64_t object, ddt_entry_t *dde,
172	    dmu_tx_t *tx);
173	int (*ddt_op_walk)(objset_t *os, uint64_t object, ddt_entry_t *dde,
174	    uint64_t *walk);
175	uint64_t (*ddt_op_count)(objset_t *os, uint64_t object);
176} ddt_ops_t;
177
178#define	DDT_NAMELEN	80
179
180extern void ddt_object_name(ddt_t *ddt, enum ddt_type type,
181    enum ddt_class class, char *name);
182extern int ddt_object_walk(ddt_t *ddt, enum ddt_type type,
183    enum ddt_class class, uint64_t *walk, ddt_entry_t *dde);
184extern uint64_t ddt_object_count(ddt_t *ddt, enum ddt_type type,
185    enum ddt_class class);
186extern int ddt_object_info(ddt_t *ddt, enum ddt_type type,
187    enum ddt_class class, dmu_object_info_t *);
188extern boolean_t ddt_object_exists(ddt_t *ddt, enum ddt_type type,
189    enum ddt_class class);
190
191extern void ddt_bp_fill(const ddt_phys_t *ddp, blkptr_t *bp,
192    uint64_t txg);
193extern void ddt_bp_create(enum zio_checksum checksum, const ddt_key_t *ddk,
194    const ddt_phys_t *ddp, blkptr_t *bp);
195
196extern void ddt_key_fill(ddt_key_t *ddk, const blkptr_t *bp);
197
198extern void ddt_phys_fill(ddt_phys_t *ddp, const blkptr_t *bp);
199extern void ddt_phys_clear(ddt_phys_t *ddp);
200extern void ddt_phys_addref(ddt_phys_t *ddp);
201extern void ddt_phys_decref(ddt_phys_t *ddp);
202extern void ddt_phys_free(ddt_t *ddt, ddt_key_t *ddk, ddt_phys_t *ddp,
203    uint64_t txg);
204extern ddt_phys_t *ddt_phys_select(const ddt_entry_t *dde, const blkptr_t *bp);
205extern uint64_t ddt_phys_total_refcnt(const ddt_entry_t *dde);
206
207extern void ddt_stat_add(ddt_stat_t *dst, const ddt_stat_t *src, uint64_t neg);
208
209extern void ddt_histogram_add(ddt_histogram_t *dst, const ddt_histogram_t *src);
210extern void ddt_histogram_stat(ddt_stat_t *dds, const ddt_histogram_t *ddh);
211extern boolean_t ddt_histogram_empty(const ddt_histogram_t *ddh);
212extern void ddt_get_dedup_object_stats(spa_t *spa, ddt_object_t *ddo);
213extern void ddt_get_dedup_histogram(spa_t *spa, ddt_histogram_t *ddh);
214extern void ddt_get_dedup_stats(spa_t *spa, ddt_stat_t *dds_total);
215
216extern uint64_t ddt_get_dedup_dspace(spa_t *spa);
217extern uint64_t ddt_get_pool_dedup_ratio(spa_t *spa);
218
219extern int ddt_ditto_copies_needed(ddt_t *ddt, ddt_entry_t *dde,
220    ddt_phys_t *ddp_willref);
221extern int ddt_ditto_copies_present(ddt_entry_t *dde);
222
223extern size_t ddt_compress(void *src, uchar_t *dst, size_t s_len, size_t d_len);
224extern void ddt_decompress(uchar_t *src, void *dst, size_t s_len, size_t d_len);
225
226extern ddt_t *ddt_select(spa_t *spa, const blkptr_t *bp);
227extern void ddt_enter(ddt_t *ddt);
228extern void ddt_exit(ddt_t *ddt);
229extern ddt_entry_t *ddt_lookup(ddt_t *ddt, const blkptr_t *bp, boolean_t add);
230extern void ddt_prefetch(spa_t *spa, const blkptr_t *bp);
231extern void ddt_remove(ddt_t *ddt, ddt_entry_t *dde);
232
233extern boolean_t ddt_class_contains(spa_t *spa, enum ddt_class max_class,
234    const blkptr_t *bp);
235
236extern ddt_entry_t *ddt_repair_start(ddt_t *ddt, const blkptr_t *bp);
237extern void ddt_repair_done(ddt_t *ddt, ddt_entry_t *dde);
238
239extern int ddt_entry_compare(const void *x1, const void *x2);
240
241extern void ddt_create(spa_t *spa);
242extern int ddt_load(spa_t *spa);
243extern void ddt_unload(spa_t *spa);
244extern void ddt_sync(spa_t *spa, uint64_t txg);
245extern int ddt_walk(spa_t *spa, ddt_bookmark_t *ddb, ddt_entry_t *dde);
246extern int ddt_object_update(ddt_t *ddt, enum ddt_type type,
247    enum ddt_class class, ddt_entry_t *dde, dmu_tx_t *tx);
248
249extern const ddt_ops_t ddt_zap_ops;
250
251#ifdef	__cplusplus
252}
253#endif
254
255#endif	/* _SYS_DDT_H */
256