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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23  * Copyright (c) 2013, 2016 by Delphix. All rights reserved.
24  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
25  * Copyright (c) 2014 Integros [integros.com]
26  */
27 
28 #ifndef   _SYS_ZAP_IMPL_H
29 #define   _SYS_ZAP_IMPL_H
30 
31 #include <sys/zap.h>
32 #include <sys/zfs_context.h>
33 #include <sys/avl.h>
34 
35 #ifdef    __cplusplus
36 extern "C" {
37 #endif
38 
39 extern int fzap_default_block_shift;
40 
41 #define   ZAP_MAGIC 0x2F52AB2ABULL
42 
43 #define   FZAP_BLOCK_SHIFT(zap)         ((zap)->zap_f.zap_block_shift)
44 
45 #define   MZAP_ENT_LEN                  64
46 #define   MZAP_NAME_LEN                 (MZAP_ENT_LEN - 8 - 4 - 2)
47 #define   MZAP_MAX_BLKSZ                SPA_OLD_MAXBLOCKSIZE
48 
49 #define   ZAP_NEED_CD                   (-1U)
50 
51 typedef struct mzap_ent_phys {
52           uint64_t mze_value;
53           uint32_t mze_cd;
54           uint16_t mze_pad;   /* in case we want to chain them someday */
55           char mze_name[MZAP_NAME_LEN];
56 } mzap_ent_phys_t;
57 
58 typedef struct mzap_phys {
59           uint64_t mz_block_type;       /* ZBT_MICRO */
60           uint64_t mz_salt;
61           uint64_t mz_normflags;
62           uint64_t mz_pad[5];
63           mzap_ent_phys_t mz_chunk[1];
64           /* actually variable size depending on block size */
65 } mzap_phys_t;
66 
67 typedef struct mzap_ent {
68           avl_node_t mze_node;
69           int mze_chunkid;
70           uint64_t mze_hash;
71           uint32_t mze_cd; /* copy from mze_phys->mze_cd */
72 } mzap_ent_t;
73 
74 #define   MZE_PHYS(zap, mze) \
75           (&zap_m_phys(zap)->mz_chunk[(mze)->mze_chunkid])
76 
77 /*
78  * The (fat) zap is stored in one object. It is an array of
79  * 1<<FZAP_BLOCK_SHIFT byte blocks. The layout looks like one of:
80  *
81  * ptrtbl fits in first block:
82  *        [zap_phys_t zap_ptrtbl_shift < 6] [zap_leaf_t] ...
83  *
84  * ptrtbl too big for first block:
85  *        [zap_phys_t zap_ptrtbl_shift >= 6] [zap_leaf_t] [ptrtbl] ...
86  *
87  */
88 
89 struct dmu_buf;
90 struct zap_leaf;
91 
92 #define   ZBT_LEAF            ((1ULL << 63) + 0)
93 #define   ZBT_HEADER                    ((1ULL << 63) + 1)
94 #define   ZBT_MICRO           ((1ULL << 63) + 3)
95 /* any other values are ptrtbl blocks */
96 
97 /*
98  * the embedded pointer table takes up half a block:
99  * block size / entry size (2^3) / 2
100  */
101 #define   ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1)
102 
103 /*
104  * The embedded pointer table starts half-way through the block.  Since
105  * the pointer table itself is half the block, it starts at (64-bit)
106  * word number (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap)).
107  */
108 #define   ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) \
109           ((uint64_t *)zap_f_phys(zap)) \
110           [(idx) + (1<<ZAP_EMBEDDED_PTRTBL_SHIFT(zap))]
111 
112 /*
113  * TAKE NOTE:
114  * If zap_phys_t is modified, zap_byteswap() must be modified.
115  */
116 typedef struct zap_phys {
117           uint64_t zap_block_type;      /* ZBT_HEADER */
118           uint64_t zap_magic;           /* ZAP_MAGIC */
119 
120           struct zap_table_phys {
121                     uint64_t zt_blk;    /* starting block number */
122                     uint64_t zt_numblks;          /* number of blocks */
123                     uint64_t zt_shift;  /* bits to index it */
124                     uint64_t zt_nextblk;          /* next (larger) copy start block */
125                     uint64_t zt_blks_copied; /* number source blocks copied */
126           } zap_ptrtbl;
127 
128           uint64_t zap_freeblk;                   /* the next free block */
129           uint64_t zap_num_leafs;                 /* number of leafs */
130           uint64_t zap_num_entries;     /* number of entries */
131           uint64_t zap_salt;            /* salt to stir into hash function */
132           uint64_t zap_normflags;                 /* flags for u8_textprep_str() */
133           uint64_t zap_flags;           /* zap_flags_t */
134           /*
135            * This structure is followed by padding, and then the embedded
136            * pointer table.  The embedded pointer table takes up second
137            * half of the block.  It is accessed using the
138            * ZAP_EMBEDDED_PTRTBL_ENT() macro.
139            */
140 } zap_phys_t;
141 
142 typedef struct zap_table_phys zap_table_phys_t;
143 
144 typedef struct zap {
145           dmu_buf_user_t zap_dbu;
146           objset_t *zap_objset;
147           uint64_t zap_object;
148           struct dmu_buf *zap_dbuf;
149           krwlock_t zap_rwlock;
150           boolean_t zap_ismicro;
151           int zap_normflags;
152           uint64_t zap_salt;
153           union {
154                     struct {
155                               /*
156                                * zap_num_entries_mtx protects
157                                * zap_num_entries
158                                */
159                               kmutex_t zap_num_entries_mtx;
160                               int zap_block_shift;
161                     } zap_fat;
162                     struct {
163                               int16_t zap_num_entries;
164                               int16_t zap_num_chunks;
165                               int16_t zap_alloc_next;
166                               avl_tree_t zap_avl;
167                     } zap_micro;
168           } zap_u;
169 } zap_t;
170 
171 inline zap_phys_t *
zap_f_phys(zap_t * zap)172 zap_f_phys(zap_t *zap)
173 {
174           return (zap->zap_dbuf->db_data);
175 }
176 
177 inline mzap_phys_t *
zap_m_phys(zap_t * zap)178 zap_m_phys(zap_t *zap)
179 {
180           return (zap->zap_dbuf->db_data);
181 }
182 
183 typedef struct zap_name {
184           zap_t *zn_zap;
185           int zn_key_intlen;
186           const void *zn_key_orig;
187           int zn_key_orig_numints;
188           const void *zn_key_norm;
189           int zn_key_norm_numints;
190           uint64_t zn_hash;
191           matchtype_t zn_matchtype;
192           char zn_normbuf[ZAP_MAXNAMELEN];
193 } zap_name_t;
194 
195 #define   zap_f     zap_u.zap_fat
196 #define   zap_m     zap_u.zap_micro
197 
198 boolean_t zap_match(zap_name_t *zn, const char *matchname);
199 int zap_lockdir(objset_t *os, uint64_t obj, dmu_tx_t *tx,
200     krw_t lti, boolean_t fatreader, boolean_t adding, void *tag, zap_t **zapp);
201 void zap_unlockdir(zap_t *zap, void *tag);
202 void zap_evict_sync(void *dbu);
203 zap_name_t *zap_name_alloc(zap_t *zap, const char *key, matchtype_t mt);
204 void zap_name_free(zap_name_t *zn);
205 int zap_hashbits(zap_t *zap);
206 uint32_t zap_maxcd(zap_t *zap);
207 uint64_t zap_getflags(zap_t *zap);
208 
209 #define   ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n))))
210 
211 void fzap_byteswap(void *buf, size_t size);
212 int fzap_count(zap_t *zap, uint64_t *count);
213 int fzap_lookup(zap_name_t *zn,
214     uint64_t integer_size, uint64_t num_integers, void *buf,
215     char *realname, int rn_len, boolean_t *normalization_conflictp);
216 void fzap_prefetch(zap_name_t *zn);
217 int fzap_count_write(zap_name_t *zn, int add, refcount_t *towrite,
218     refcount_t *tooverwrite);
219 int fzap_add(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers,
220     const void *val, void *tag, dmu_tx_t *tx);
221 int fzap_update(zap_name_t *zn,
222     int integer_size, uint64_t num_integers, const void *val,
223     void *tag, dmu_tx_t *tx);
224 int fzap_length(zap_name_t *zn,
225     uint64_t *integer_size, uint64_t *num_integers);
226 int fzap_remove(zap_name_t *zn, dmu_tx_t *tx);
227 int fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za);
228 void fzap_get_stats(zap_t *zap, zap_stats_t *zs);
229 void zap_put_leaf(struct zap_leaf *l);
230 
231 int fzap_add_cd(zap_name_t *zn,
232     uint64_t integer_size, uint64_t num_integers,
233     const void *val, uint32_t cd, void *tag, dmu_tx_t *tx);
234 void fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags);
235 int fzap_cursor_move_to_key(zap_cursor_t *zc, zap_name_t *zn);
236 
237 #ifdef    __cplusplus
238 }
239 #endif
240 
241 #endif /* _SYS_ZAP_IMPL_H */
242