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 https://opensource.org/licenses/CDDL-1.0.
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 * Copyright (c) 2023, Klara Inc.
25 */
26
27 #ifndef _SYS_DDT_IMPL_H
28 #define _SYS_DDT_IMPL_H
29
30 #include <sys/ddt.h>
31 #include <sys/bitops.h>
32
33 #ifdef __cplusplus
34 extern "C" {
35 #endif
36
37 /* DDT version numbers */
38 #define DDT_VERSION_LEGACY (0)
39 #define DDT_VERSION_FDT (1)
40
41 /* Dummy version to signal that configure is still necessary */
42 #define DDT_VERSION_UNCONFIGURED (UINT64_MAX)
43
44 /* Names of interesting objects in the DDT root dir */
45 #define DDT_DIR_VERSION "version"
46 #define DDT_DIR_FLAGS "flags"
47
48 /* Fill a lightweight entry from a live entry. */
49 #define DDT_ENTRY_TO_LIGHTWEIGHT(ddt, dde, ddlwe) do { \
50 memset((ddlwe), 0, sizeof (*ddlwe)); \
51 (ddlwe)->ddlwe_key = (dde)->dde_key; \
52 (ddlwe)->ddlwe_type = (dde)->dde_type; \
53 (ddlwe)->ddlwe_class = (dde)->dde_class; \
54 memcpy(&(ddlwe)->ddlwe_phys, (dde)->dde_phys, DDT_PHYS_SIZE(ddt)); \
55 } while (0)
56
57 #define DDT_LOG_ENTRY_TO_LIGHTWEIGHT(ddt, ddle, ddlwe) do { \
58 memset((ddlwe), 0, sizeof (*ddlwe)); \
59 (ddlwe)->ddlwe_key = (ddle)->ddle_key; \
60 (ddlwe)->ddlwe_type = (ddle)->ddle_type; \
61 (ddlwe)->ddlwe_class = (ddle)->ddle_class; \
62 memcpy(&(ddlwe)->ddlwe_phys, (ddle)->ddle_phys, DDT_PHYS_SIZE(ddt)); \
63 } while (0)
64
65 /*
66 * An entry on the log tree. These are "frozen", and a record of what's in
67 * the on-disk log. They can't be used in place, but can be "loaded" back into
68 * the live tree.
69 */
70 typedef struct {
71 ddt_key_t ddle_key; /* ddt_log_tree key */
72 avl_node_t ddle_node; /* ddt_log_tree node */
73
74 ddt_type_t ddle_type; /* storage type */
75 ddt_class_t ddle_class; /* storage class */
76
77 /* extra allocation for flat/trad phys */
78 ddt_univ_phys_t ddle_phys[];
79 } ddt_log_entry_t;
80
81 /* On-disk log record types. */
82 typedef enum {
83 DLR_INVALID = 0, /* end of block marker */
84 DLR_ENTRY = 1, /* an entry to add or replace in the log tree */
85 } ddt_log_record_type_t;
86
87 /* On-disk log record header. */
88 typedef struct {
89 /*
90 * dlr_info is a packed u64, use the DLR_GET/DLR_SET macros below to
91 * access it.
92 *
93 * bits 0-7: record type (ddt_log_record_type_t)
94 * bits 8-15: length of record header+payload
95 * bits 16-47: reserved, all zero
96 * bits 48-55: if type==DLR_ENTRY, storage type (ddt_type)
97 * otherwise all zero
98 * bits 56-63: if type==DLR_ENTRY, storage class (ddt_class)
99 * otherwise all zero
100 */
101 uint64_t dlr_info;
102 uint8_t dlr_payload[];
103 } ddt_log_record_t;
104
105 #define DLR_GET_TYPE(dlr) BF64_GET((dlr)->dlr_info, 0, 8)
106 #define DLR_SET_TYPE(dlr, v) BF64_SET((dlr)->dlr_info, 0, 8, v)
107 #define DLR_GET_RECLEN(dlr) BF64_GET((dlr)->dlr_info, 8, 16)
108 #define DLR_SET_RECLEN(dlr, v) BF64_SET((dlr)->dlr_info, 8, 16, v)
109 #define DLR_GET_ENTRY_TYPE(dlr) BF64_GET((dlr)->dlr_info, 48, 8)
110 #define DLR_SET_ENTRY_TYPE(dlr, v) BF64_SET((dlr)->dlr_info, 48, 8, v)
111 #define DLR_GET_ENTRY_CLASS(dlr) BF64_GET((dlr)->dlr_info, 56, 8)
112 #define DLR_SET_ENTRY_CLASS(dlr, v) BF64_SET((dlr)->dlr_info, 56, 8, v)
113
114 /* Payload for DLR_ENTRY. */
115 typedef struct {
116 ddt_key_t dlre_key;
117 ddt_univ_phys_t dlre_phys[];
118 } ddt_log_record_entry_t;
119
120 /* Log flags (ddl_flags, dlh_flags) */
121 #define DDL_FLAG_FLUSHING (1 << 0) /* this log is being flushed */
122 #define DDL_FLAG_CHECKPOINT (1 << 1) /* header has a checkpoint */
123
124 /* On-disk log header, stored in the bonus buffer. */
125 typedef struct {
126 /*
127 * dlh_info is a packed u64, use the DLH_GET/DLH_SET macros below to
128 * access it.
129 *
130 * bits 0-7: log version
131 * bits 8-15: log flags
132 * bits 16-63: reserved, all zero
133 */
134 uint64_t dlh_info;
135
136 uint64_t dlh_length; /* log size in bytes */
137 uint64_t dlh_first_txg; /* txg this log went active */
138 ddt_key_t dlh_checkpoint; /* last checkpoint */
139 } ddt_log_header_t;
140
141 #define DLH_GET_VERSION(dlh) BF64_GET((dlh)->dlh_info, 0, 8)
142 #define DLH_SET_VERSION(dlh, v) BF64_SET((dlh)->dlh_info, 0, 8, v)
143 #define DLH_GET_FLAGS(dlh) BF64_GET((dlh)->dlh_info, 8, 8)
144 #define DLH_SET_FLAGS(dlh, v) BF64_SET((dlh)->dlh_info, 8, 8, v)
145
146 /* DDT log update state */
147 typedef struct {
148 dmu_tx_t *dlu_tx; /* tx the update is being applied to */
149 dnode_t *dlu_dn; /* log object dnode */
150 dmu_buf_t **dlu_dbp; /* array of block buffer pointers */
151 int dlu_ndbp; /* number of block buffer pointers */
152 uint16_t dlu_reclen; /* cached length of record */
153 uint64_t dlu_block; /* block for next entry */
154 uint64_t dlu_offset; /* offset for next entry */
155 } ddt_log_update_t;
156
157 /*
158 * Ops vector to access a specific DDT object type.
159 */
160 typedef struct {
161 char ddt_op_name[32];
162 int (*ddt_op_create)(objset_t *os, uint64_t *object, dmu_tx_t *tx,
163 boolean_t prehash);
164 int (*ddt_op_destroy)(objset_t *os, uint64_t object, dmu_tx_t *tx);
165 int (*ddt_op_lookup)(objset_t *os, uint64_t object,
166 const ddt_key_t *ddk, void *phys, size_t psize);
167 int (*ddt_op_contains)(objset_t *os, uint64_t object,
168 const ddt_key_t *ddk);
169 void (*ddt_op_prefetch)(objset_t *os, uint64_t object,
170 const ddt_key_t *ddk);
171 void (*ddt_op_prefetch_all)(objset_t *os, uint64_t object);
172 int (*ddt_op_update)(objset_t *os, uint64_t object,
173 const ddt_key_t *ddk, const void *phys, size_t psize,
174 dmu_tx_t *tx);
175 int (*ddt_op_remove)(objset_t *os, uint64_t object,
176 const ddt_key_t *ddk, dmu_tx_t *tx);
177 int (*ddt_op_walk)(objset_t *os, uint64_t object, uint64_t *walk,
178 ddt_key_t *ddk, void *phys, size_t psize);
179 int (*ddt_op_count)(objset_t *os, uint64_t object, uint64_t *count);
180 } ddt_ops_t;
181
182 extern const ddt_ops_t ddt_zap_ops;
183
184 /* Dedup log API */
185 extern void ddt_log_begin(ddt_t *ddt, size_t nentries, dmu_tx_t *tx,
186 ddt_log_update_t *dlu);
187 extern void ddt_log_entry(ddt_t *ddt, ddt_lightweight_entry_t *dde,
188 ddt_log_update_t *dlu);
189 extern void ddt_log_commit(ddt_t *ddt, ddt_log_update_t *dlu);
190
191 extern boolean_t ddt_log_take_first(ddt_t *ddt, ddt_log_t *ddl,
192 ddt_lightweight_entry_t *ddlwe);
193
194 extern boolean_t ddt_log_find_key(ddt_t *ddt, const ddt_key_t *ddk,
195 ddt_lightweight_entry_t *ddlwe);
196 extern boolean_t ddt_log_remove_key(ddt_t *ddt, ddt_log_t *ddl,
197 const ddt_key_t *ddk);
198
199 extern void ddt_log_checkpoint(ddt_t *ddt, ddt_lightweight_entry_t *ddlwe,
200 dmu_tx_t *tx);
201 extern void ddt_log_truncate(ddt_t *ddt, dmu_tx_t *tx);
202
203 extern boolean_t ddt_log_swap(ddt_t *ddt, dmu_tx_t *tx);
204
205 extern void ddt_log_destroy(ddt_t *ddt, dmu_tx_t *tx);
206
207 extern int ddt_log_load(ddt_t *ddt);
208 extern void ddt_log_alloc(ddt_t *ddt);
209 extern void ddt_log_free(ddt_t *ddt);
210
211 extern void ddt_log_init(void);
212 extern void ddt_log_fini(void);
213
214 /*
215 * These are only exposed so that zdb can access them. Try not to use them
216 * outside of the DDT implementation proper, and if you do, consider moving
217 * them up.
218 */
219
220 /*
221 * We use a histogram to convert a percentage request into a
222 * cutoff value where entries older than the cutoff get pruned.
223 *
224 * The histogram bins represent hours in power-of-two increments.
225 * 16 bins covers up to four years.
226 */
227 #define HIST_BINS 16
228
229 typedef struct ddt_age_histo {
230 uint64_t dah_entries;
231 uint64_t dah_age_histo[HIST_BINS];
232 } ddt_age_histo_t;
233
234 void ddt_prune_walk(spa_t *spa, uint64_t cutoff, ddt_age_histo_t *histogram);
235
236 #if defined(_KERNEL) || !defined(ZFS_DEBUG)
237 #define ddt_dump_age_histogram(histo, cutoff) ((void)0)
238 #else
239 static inline void
ddt_dump_age_histogram(ddt_age_histo_t * histogram,uint64_t cutoff)240 ddt_dump_age_histogram(ddt_age_histo_t *histogram, uint64_t cutoff)
241 {
242 if (histogram->dah_entries == 0)
243 return;
244
245 (void) printf("DDT prune unique class age, %llu hour cutoff\n",
246 (u_longlong_t)(gethrestime_sec() - cutoff)/3600);
247 (void) printf("%5s %9s %4s\n", "age", "blocks", "amnt");
248 (void) printf("%5s %9s %4s\n", "-----", "---------", "----");
249 for (int i = 0; i < HIST_BINS; i++) {
250 (void) printf("%5d %9llu %4d%%\n", 1<<i,
251 (u_longlong_t)histogram->dah_age_histo[i],
252 (int)((histogram->dah_age_histo[i] * 100) /
253 histogram->dah_entries));
254 }
255 }
256 #endif
257
258 /*
259 * Enough room to expand DMU_POOL_DDT format for all possible DDT
260 * checksum/class/type combinations.
261 */
262 #define DDT_NAMELEN 32
263
264 extern uint64_t ddt_phys_total_refcnt(const ddt_t *ddt,
265 const ddt_univ_phys_t *ddp);
266
267 extern void ddt_key_fill(ddt_key_t *ddk, const blkptr_t *bp);
268
269 extern void ddt_object_name(ddt_t *ddt, ddt_type_t type, ddt_class_t clazz,
270 char *name);
271 extern int ddt_object_walk(ddt_t *ddt, ddt_type_t type, ddt_class_t clazz,
272 uint64_t *walk, ddt_lightweight_entry_t *ddlwe);
273 extern int ddt_object_count(ddt_t *ddt, ddt_type_t type, ddt_class_t clazz,
274 uint64_t *count);
275 extern int ddt_object_info(ddt_t *ddt, ddt_type_t type, ddt_class_t clazz,
276 dmu_object_info_t *);
277
278 #ifdef __cplusplus
279 }
280 #endif
281
282 #endif /* _SYS_DDT_H */
283