xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zap_leaf.h (revision efd4c9b63ad77503c101fc6c2ed8ba96c9d52964)
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  */
24 
25 #ifndef	_SYS_ZAP_LEAF_H
26 #define	_SYS_ZAP_LEAF_H
27 
28 #include <sys/zap.h>
29 
30 #ifdef	__cplusplus
31 extern "C" {
32 #endif
33 
34 struct zap;
35 struct zap_name;
36 struct zap_stats;
37 
38 #define	ZAP_LEAF_MAGIC 0x2AB1EAF
39 
40 /* chunk size = 24 bytes */
41 #define	ZAP_LEAF_CHUNKSIZE 24
42 
43 /*
44  * The amount of space available for chunks is:
45  * block size (1<<l->l_bs) - hash entry size (2) * number of hash
46  * entries - header space (2*chunksize)
47  */
48 #define	ZAP_LEAF_NUMCHUNKS(l) \
49 	(((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \
50 	ZAP_LEAF_CHUNKSIZE - 2)
51 
52 /*
53  * The amount of space within the chunk available for the array is:
54  * chunk size - space for type (1) - space for next pointer (2)
55  */
56 #define	ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3)
57 
58 #define	ZAP_LEAF_ARRAY_NCHUNKS(bytes) \
59 	(((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES)
60 
61 /*
62  * Low water mark:  when there are only this many chunks free, start
63  * growing the ptrtbl.  Ideally, this should be larger than a
64  * "reasonably-sized" entry.  20 chunks is more than enough for the
65  * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value),
66  * while still being only around 3% for 16k blocks.
67  */
68 #define	ZAP_LEAF_LOW_WATER (20)
69 
70 /*
71  * The leaf hash table has block size / 2^5 (32) number of entries,
72  * which should be more than enough for the maximum number of entries,
73  * which is less than block size / CHUNKSIZE (24) / minimum number of
74  * chunks per entry (3).
75  */
76 #define	ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5)
77 #define	ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l))
78 
79 /*
80  * The chunks start immediately after the hash table.  The end of the
81  * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a
82  * chunk_t.
83  */
84 #define	ZAP_LEAF_CHUNK(l, idx) \
85 	((zap_leaf_chunk_t *) \
86 	((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx]
87 #define	ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry)
88 
89 typedef enum zap_chunk_type {
90 	ZAP_CHUNK_FREE = 253,
91 	ZAP_CHUNK_ENTRY = 252,
92 	ZAP_CHUNK_ARRAY = 251,
93 	ZAP_CHUNK_TYPE_MAX = 250
94 } zap_chunk_type_t;
95 
96 #define	ZLF_ENTRIES_CDSORTED (1<<0)
97 
98 /*
99  * TAKE NOTE:
100  * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified.
101  */
102 typedef struct zap_leaf_phys {
103 	struct zap_leaf_header {
104 		uint64_t lh_block_type;		/* ZBT_LEAF */
105 		uint64_t lh_pad1;
106 		uint64_t lh_prefix;		/* hash prefix of this leaf */
107 		uint32_t lh_magic;		/* ZAP_LEAF_MAGIC */
108 		uint16_t lh_nfree;		/* number free chunks */
109 		uint16_t lh_nentries;		/* number of entries */
110 		uint16_t lh_prefix_len;		/* num bits used to id this */
111 
112 /* above is accessable to zap, below is zap_leaf private */
113 
114 		uint16_t lh_freelist;		/* chunk head of free list */
115 		uint8_t lh_flags;		/* ZLF_* flags */
116 		uint8_t lh_pad2[11];
117 	} l_hdr; /* 2 24-byte chunks */
118 
119 	/*
120 	 * The header is followed by a hash table with
121 	 * ZAP_LEAF_HASH_NUMENTRIES(zap) entries.  The hash table is
122 	 * followed by an array of ZAP_LEAF_NUMCHUNKS(zap)
123 	 * zap_leaf_chunk structures.  These structures are accessed
124 	 * with the ZAP_LEAF_CHUNK() macro.
125 	 */
126 
127 	uint16_t l_hash[1];
128 } zap_leaf_phys_t;
129 
130 typedef union zap_leaf_chunk {
131 	struct zap_leaf_entry {
132 		uint8_t le_type; 		/* always ZAP_CHUNK_ENTRY */
133 		uint8_t le_value_intlen;	/* size of value's ints */
134 		uint16_t le_next;		/* next entry in hash chain */
135 		uint16_t le_name_chunk;		/* first chunk of the name */
136 		uint16_t le_name_numints;	/* ints in name (incl null) */
137 		uint16_t le_value_chunk;	/* first chunk of the value */
138 		uint16_t le_value_numints;	/* value length in ints */
139 		uint32_t le_cd;			/* collision differentiator */
140 		uint64_t le_hash;		/* hash value of the name */
141 	} l_entry;
142 	struct zap_leaf_array {
143 		uint8_t la_type;		/* always ZAP_CHUNK_ARRAY */
144 		uint8_t la_array[ZAP_LEAF_ARRAY_BYTES];
145 		uint16_t la_next;		/* next blk or CHAIN_END */
146 	} l_array;
147 	struct zap_leaf_free {
148 		uint8_t lf_type;		/* always ZAP_CHUNK_FREE */
149 		uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES];
150 		uint16_t lf_next;	/* next in free list, or CHAIN_END */
151 	} l_free;
152 } zap_leaf_chunk_t;
153 
154 typedef struct zap_leaf {
155 	krwlock_t l_rwlock;
156 	uint64_t l_blkid;		/* 1<<ZAP_BLOCK_SHIFT byte block off */
157 	int l_bs;			/* block size shift */
158 	dmu_buf_t *l_dbuf;
159 	zap_leaf_phys_t *l_phys;
160 } zap_leaf_t;
161 
162 
163 typedef struct zap_entry_handle {
164 	/* below is set by zap_leaf.c and is public to zap.c */
165 	uint64_t zeh_num_integers;
166 	uint64_t zeh_hash;
167 	uint32_t zeh_cd;
168 	uint8_t zeh_integer_size;
169 
170 	/* below is private to zap_leaf.c */
171 	uint16_t zeh_fakechunk;
172 	uint16_t *zeh_chunkp;
173 	zap_leaf_t *zeh_leaf;
174 } zap_entry_handle_t;
175 
176 /*
177  * Return a handle to the named entry, or ENOENT if not found.  The hash
178  * value must equal zap_hash(name).
179  */
180 extern int zap_leaf_lookup(zap_leaf_t *l,
181     struct zap_name *zn, zap_entry_handle_t *zeh);
182 
183 /*
184  * Return a handle to the entry with this hash+cd, or the entry with the
185  * next closest hash+cd.
186  */
187 extern int zap_leaf_lookup_closest(zap_leaf_t *l,
188     uint64_t hash, uint32_t cd, zap_entry_handle_t *zeh);
189 
190 /*
191  * Read the first num_integers in the attribute.  Integer size
192  * conversion will be done without sign extension.  Return EINVAL if
193  * integer_size is too small.  Return EOVERFLOW if there are more than
194  * num_integers in the attribute.
195  */
196 extern int zap_entry_read(const zap_entry_handle_t *zeh,
197     uint8_t integer_size, uint64_t num_integers, void *buf);
198 
199 extern int zap_entry_read_name(struct zap *zap, const zap_entry_handle_t *zeh,
200     uint16_t buflen, char *buf);
201 
202 /*
203  * Replace the value of an existing entry.
204  *
205  * zap_entry_update may fail if it runs out of space (ENOSPC).
206  */
207 extern int zap_entry_update(zap_entry_handle_t *zeh,
208     uint8_t integer_size, uint64_t num_integers, const void *buf);
209 
210 /*
211  * Remove an entry.
212  */
213 extern void zap_entry_remove(zap_entry_handle_t *zeh);
214 
215 /*
216  * Create an entry. An equal entry must not exist, and this entry must
217  * belong in this leaf (according to its hash value).  Fills in the
218  * entry handle on success.  Returns 0 on success or ENOSPC on failure.
219  */
220 extern int zap_entry_create(zap_leaf_t *l, struct zap_name *zn, uint32_t cd,
221     uint8_t integer_size, uint64_t num_integers, const void *buf,
222     zap_entry_handle_t *zeh);
223 
224 /*
225  * Return true if there are additional entries with the same normalized
226  * form.
227  */
228 extern boolean_t zap_entry_normalization_conflict(zap_entry_handle_t *zeh,
229     struct zap_name *zn, const char *name, struct zap *zap);
230 
231 /*
232  * Other stuff.
233  */
234 
235 extern void zap_leaf_init(zap_leaf_t *l, boolean_t sort);
236 extern void zap_leaf_byteswap(zap_leaf_phys_t *buf, int len);
237 extern void zap_leaf_split(zap_leaf_t *l, zap_leaf_t *nl, boolean_t sort);
238 extern void zap_leaf_stats(struct zap *zap, zap_leaf_t *l,
239     struct zap_stats *zs);
240 
241 #ifdef	__cplusplus
242 }
243 #endif
244 
245 #endif /* _SYS_ZAP_LEAF_H */
246