1 // SPDX-License-Identifier: CDDL-1.0
2 /*
3 * CDDL HEADER START
4 *
5 * The contents of this file are subject to the terms of the
6 * Common Development and Distribution License (the "License").
7 * You may not use this file except in compliance with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or https://opensource.org/licenses/CDDL-1.0.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22
23 /*
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26 * Copyright (c) 2013, 2016 by Delphix. All rights reserved.
27 * Copyright 2017 Nexenta Systems, Inc.
28 * Copyright (c) 2024, Klara, Inc.
29 * Copyright (c) 2026, TrueNAS.
30 */
31
32 #ifndef _SYS_ZAP_IMPL_H
33 #define _SYS_ZAP_IMPL_H
34
35 #include <sys/zap.h>
36 #include <sys/zfs_context.h>
37
38 #ifdef __cplusplus
39 extern "C" {
40 #endif
41
42 extern int fzap_default_block_shift;
43
44 #define ZAP_MAGIC 0x2F52AB2ABULL
45
46 #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_f.zap_block_shift)
47
48 #define MZAP_ENT_LEN 64
49 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
50
51 #define ZAP_NEED_CD (-1U)
52
53 typedef struct mzap_ent_phys {
54 uint64_t mze_value;
55 uint32_t mze_cd;
56 uint16_t mze_pad; /* in case we want to chain them someday */
57 char mze_name[MZAP_NAME_LEN];
58 } mzap_ent_phys_t;
59
60 typedef struct mzap_phys {
61 uint64_t mz_block_type; /* ZBT_MICRO */
62 uint64_t mz_salt;
63 uint64_t mz_normflags;
64 uint64_t mz_pad[5];
65
66 /* actually variable size depending on block size */
67 mzap_ent_phys_t mz_chunk[];
68 } mzap_phys_t;
69
70 typedef struct mzap_ent {
71 uint32_t mze_hash;
72 uint16_t mze_cd; /* copy from mze_phys->mze_cd */
73 uint16_t mze_chunkid;
74 } mzap_ent_t;
75
76 #define MZE_PHYS(zap, mze) \
77 (&zap_m_phys(zap)->mz_chunk[(mze)->mze_chunkid])
78
79 /*
80 * The (fat) zap is stored in one object. It is an array of
81 * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
82 *
83 * ptrtbl fits in first block:
84 * [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
85 *
86 * ptrtbl too big for first block:
87 * [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
88 *
89 */
90
91 struct dmu_buf;
92 struct zap_leaf;
93
94 #define ZBT_LEAF ((1ULL << 63) + 0)
95 #define ZBT_HEADER ((1ULL << 63) + 1)
96 #define ZBT_MICRO ((1ULL << 63) + 3)
97 /* any other values are ptrtbl blocks */
98
99 /*
100 * the embedded pointer table takes up half a block:
101 * block size / entry size (2^3) / 2
102 */
103 #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
104
105 /*
106 * The embedded pointer table starts half-way through the block. Since
107 * the pointer table itself is half the block, it starts at (64-bit)
108 * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
109 */
110 #define ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
111 ((uint64_t *)zap_f_phys(zap)) \
112 [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
113
114 /*
115 * TAKE NOTE:
116 * If zap_phys_t is modified, zap_byteswap() must be modified.
117 */
118 typedef struct zap_phys {
119 uint64_t zap_block_type; /* ZBT_HEADER */
120 uint64_t zap_magic; /* ZAP_MAGIC */
121
122 struct zap_table_phys {
123 uint64_t zt_blk; /* starting block number */
124 uint64_t zt_numblks; /* number of blocks */
125 uint64_t zt_shift; /* bits to index it */
126 uint64_t zt_nextblk; /* next (larger) copy start block */
127 uint64_t zt_blks_copied; /* number source blocks copied */
128 } zap_ptrtbl;
129
130 uint64_t zap_freeblk; /* the next free block */
131 uint64_t zap_num_leafs; /* number of leafs */
132 uint64_t zap_num_entries; /* number of entries */
133 uint64_t zap_salt; /* salt to stir into hash function */
134 uint64_t zap_normflags; /* flags for u8_textprep_str() */
135 uint64_t zap_flags; /* zap_flags_t */
136 /*
137 * This structure is followed by padding, and then the embedded
138 * pointer table. The embedded pointer table takes up second
139 * half of the block. It is accessed using the
140 * ZAP_EMBEDDED_PTRTBL_ENT() macro.
141 */
142 } zap_phys_t;
143
144 typedef struct zap_table_phys zap_table_phys_t;
145
146 typedef struct zap {
147 dmu_buf_user_t zap_dbu;
148 objset_t *zap_objset;
149 uint64_t zap_object;
150 dnode_t *zap_dnode;
151 struct dmu_buf *zap_dbuf;
152 krwlock_t zap_rwlock;
153 boolean_t zap_ismicro;
154 int zap_normflags;
155 uint64_t zap_salt;
156 union {
157 struct {
158 /*
159 * zap_num_entries_mtx protects
160 * zap_num_entries
161 */
162 kmutex_t zap_num_entries_mtx;
163 int zap_block_shift;
164 } zap_fat;
165 struct {
166 int16_t zap_num_entries;
167 int16_t zap_num_chunks;
168 int16_t zap_alloc_next;
169 zfs_btree_t zap_tree;
170 } zap_micro;
171 } zap_u;
172 } zap_t;
173
174 #define zap_f zap_u.zap_fat
175 #define zap_m zap_u.zap_micro
176
177 static inline zap_phys_t *
zap_f_phys(zap_t * zap)178 zap_f_phys(zap_t *zap)
179 {
180 return (zap->zap_dbuf->db_data);
181 }
182
183 static inline mzap_phys_t *
zap_m_phys(zap_t * zap)184 zap_m_phys(zap_t *zap)
185 {
186 return (zap->zap_dbuf->db_data);
187 }
188
189 /*
190 * zap_name_t carries the original key and whatever we've derived from it
191 * (normalised form, hash, etc) as we work through completing the operation.
192 */
193 typedef struct zap_name {
194 zap_t *zn_zap;
195 int zn_key_intlen;
196 const void *zn_key_orig;
197 int zn_key_orig_numints;
198 const void *zn_key_norm;
199 int zn_key_norm_numints;
200 uint64_t zn_hash;
201 matchtype_t zn_matchtype;
202 int zn_normflags;
203 int zn_normbuf_len;
204 char zn_normbuf[];
205 } zap_name_t;
206
207 /*
208 * Allocate a zap_name_t. The longname flag ensures there is enough room to
209 * hold a long filename when the 'longname' pool feature is active.
210 */
211 zap_name_t *zap_name_alloc(zap_t *zap, boolean_t longname);
212
213 /*
214 * Allocate a zap_name_t for the given key. zap_name_init_str() will be
215 * called to normalise the key and initialise the struct.
216 */
217 zap_name_t *zap_name_alloc_str(zap_t *zap, const char *key, matchtype_t mt);
218
219 /*
220 * Allocate a zap_name_t for a uint64 array key.
221 */
222 zap_name_t *zap_name_alloc_uint64(zap_t *zap, const uint64_t *key, int numints);
223
224 /*
225 * Free a zap_name_t.
226 */
227 void zap_name_free(zap_name_t *zn);
228
229 /*
230 * Initialise an existing zap_name_t with the normalised form of the key,
231 * computed according to the given matchtype.
232 */
233 int zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt);
234
235 /*
236 * Compare 'matchname' with the name represented by the zap_name_t, applying
237 * the same normalisation method first. Returns true if the normalised forms
238 * match, false otherwise.
239 */
240 boolean_t zap_match(zap_name_t *zn, const char *matchname);
241
242 /*
243 * Compute and return the 64-bit hash for the name, according to the name
244 * type and hash flags.
245 */
246 uint64_t zap_hash(zap_name_t *zn);
247
248 /*
249 * Return a zap_t for the given on-disk object, locked and ready for use.
250 * The zap_t will be allocated and loaded from disk if its not already loaded.
251 */
252 int zap_lock(objset_t *os, uint64_t obj, dmu_tx_t *tx,
253 krw_t lti, boolean_t fatreader, boolean_t adding, const void *tag,
254 zap_t **zapp);
255 int zap_lock_by_dnode(dnode_t *dn, dmu_tx_t *tx,
256 krw_t lti, boolean_t fatreader, boolean_t adding, const void *tag,
257 zap_t **zapp);
258
259 /* Unlock and release a zap_t. */
260 void zap_unlock(zap_t *zap, const void *tag);
261
262 /*
263 * Try to upgrade a zap lock from READER to WRITER. If the upgrade is not
264 * possible without blocking, returns 0. If the upgrade happened, returns 1.
265 */
266 int zap_lock_try_upgrade(zap_t *zap, dmu_tx_t *tx);
267
268 /*
269 * Upgrade a zap lock from READER to WRITER. If it can't be upgraded
270 * immediately it will block.
271 */
272 void zap_lock_upgrade(zap_t *zap, dmu_tx_t *tx);
273
274 /* zap_t release function for when associated dbuf is evicted. */
275 void zap_evict_sync(void *dbu);
276
277 /* Misc internal state & config. */
278 int zap_hashbits(zap_t *zap);
279 uint32_t zap_maxcd(zap_t *zap);
280 uint64_t zap_getflags(zap_t *zap);
281
282 /* Microzap implementation. */
283 zap_t *mzap_open(dmu_buf_t *db);
284 int mzap_upgrade(zap_t **zapp, dmu_tx_t *tx, zap_flags_t flags);
285 mzap_ent_t *mze_find(zap_name_t *zn, zfs_btree_index_t *idx);
286 boolean_t mze_canfit_fzap_leaf(zap_name_t *zn, uint64_t hash);
287 void mze_destroy(zap_t *zap);
288 boolean_t mzap_normalization_conflict(zap_t *zap, zap_name_t *zn,
289 mzap_ent_t *mze, zfs_btree_index_t *idx);
290 void mzap_addent(zap_name_t *zn, uint64_t value);
291 void mzap_byteswap(mzap_phys_t *buf, size_t size);
292 uint64_t zap_get_micro_max_size(spa_t *spa);
293
294 /* Fatzap implementation. */
295 void fzap_byteswap(void *buf, size_t size);
296 int fzap_count(zap_t *zap, uint64_t *count);
297 int fzap_lookup(zap_name_t *zn,
298 uint64_t integer_size, uint64_t num_integers, void *buf,
299 char *realname, int rn_len, boolean_t *normalization_conflictp,
300 uint64_t *actual_num_integers);
301 void fzap_prefetch(zap_name_t *zn);
302 int fzap_add(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers,
303 const void *val, dmu_tx_t *tx);
304 int fzap_update(zap_name_t *zn, int integer_size, uint64_t num_integers,
305 const void *val, dmu_tx_t *tx);
306 int fzap_length(zap_name_t *zn,
307 uint64_t *integer_size, uint64_t *num_integers);
308 int fzap_remove(zap_name_t *zn, dmu_tx_t *tx);
309 int fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za);
310 void fzap_get_stats(zap_t *zap, zap_stats_t *zs);
311 void zap_put_leaf(struct zap_leaf *l);
312 int fzap_add_cd(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers,
313 const void *val, uint32_t cd, dmu_tx_t *tx);
314 void fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags);
315
316 #ifdef __cplusplus
317 }
318 #endif
319
320 #endif /* _SYS_ZAP_IMPL_H */
321