xref: /freebsd/lib/libc/db/hash/hash.h (revision c98323078dede7579020518ec84cdcb478e5c142)
1 /*-
2  * Copyright (c) 1990, 1993, 1994
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Margo Seltzer.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)hash.h	8.3 (Berkeley) 5/31/94
37  * $FreeBSD$
38  */
39 
40 /* Operations */
41 typedef enum {
42 	HASH_GET, HASH_PUT, HASH_PUTNEW, HASH_DELETE, HASH_FIRST, HASH_NEXT
43 } ACTION;
44 
45 /* Buffer Management structures */
46 typedef struct _bufhead BUFHEAD;
47 
48 struct _bufhead {
49 	BUFHEAD		*prev;		/* LRU links */
50 	BUFHEAD		*next;		/* LRU links */
51 	BUFHEAD		*ovfl;		/* Overflow page buffer header */
52 	u_int32_t	 addr;		/* Address of this page */
53 	char		*page;		/* Actual page data */
54 	char	 	flags;
55 #define	BUF_MOD		0x0001
56 #define BUF_DISK	0x0002
57 #define	BUF_BUCKET	0x0004
58 #define	BUF_PIN		0x0008
59 };
60 
61 #define IS_BUCKET(X)	((X) & BUF_BUCKET)
62 
63 typedef BUFHEAD **SEGMENT;
64 
65 /* Hash Table Information */
66 typedef struct hashhdr {		/* Disk resident portion */
67 	int		magic;		/* Magic NO for hash tables */
68 	int		version;	/* Version ID */
69 	u_int32_t	lorder;		/* Byte Order */
70 	int		bsize;		/* Bucket/Page Size */
71 	int		bshift;		/* Bucket shift */
72 	int		dsize;		/* Directory Size */
73 	int		ssize;		/* Segment Size */
74 	int		sshift;		/* Segment shift */
75 	int		ovfl_point;	/* Where overflow pages are being
76 					 * allocated */
77 	int		last_freed;	/* Last overflow page freed */
78 	int		max_bucket;	/* ID of Maximum bucket in use */
79 	int		high_mask;	/* Mask to modulo into entire table */
80 	int		low_mask;	/* Mask to modulo into lower half of
81 					 * table */
82 	int		ffactor;	/* Fill factor */
83 	int		nkeys;		/* Number of keys in hash table */
84 	int		hdrpages;	/* Size of table header */
85 	int		h_charkey;	/* value of hash(CHARKEY) */
86 #define NCACHED	32			/* number of bit maps and spare
87 					 * points */
88 	int		spares[NCACHED];/* spare pages for overflow */
89 	u_int16_t	bitmaps[NCACHED];	/* address of overflow page
90 						 * bitmaps */
91 } HASHHDR;
92 
93 typedef struct htab	 {		/* Memory resident data structure */
94 	HASHHDR 	hdr;		/* Header */
95 	int		nsegs;		/* Number of allocated segments */
96 	int		exsegs;		/* Number of extra allocated
97 					 * segments */
98 	u_int32_t			/* Hash function */
99 	    (*hash)(const void *, size_t);
100 	int		flags;		/* Flag values */
101 	int		fp;		/* File pointer */
102 	char		*tmp_buf;	/* Temporary Buffer for BIG data */
103 	char		*tmp_key;	/* Temporary Buffer for BIG keys */
104 	BUFHEAD 	*cpage;		/* Current page */
105 	int		cbucket;	/* Current bucket */
106 	int		cndx;		/* Index of next item on cpage */
107 	int		error;		/* Error Number -- for DBM
108 					 * compatibility */
109 	int		new_file;	/* Indicates if fd is backing store
110 					 * or no */
111 	int		save_file;	/* Indicates whether we need to flush
112 					 * file at
113 					 * exit */
114 	u_int32_t	*mapp[NCACHED];	/* Pointers to page maps */
115 	int		nmaps;		/* Initial number of bitmaps */
116 	int		nbufs;		/* Number of buffers left to
117 					 * allocate */
118 	BUFHEAD 	bufhead;	/* Header of buffer lru list */
119 	SEGMENT 	*dir;		/* Hash Bucket directory */
120 } HTAB;
121 
122 /*
123  * Constants
124  */
125 #define	MAX_BSIZE		65536		/* 2^16 */
126 #define MIN_BUFFERS		6
127 #define MINHDRSIZE		512
128 #define DEF_BUFSIZE		65536		/* 64 K */
129 #define DEF_BUCKET_SIZE		4096
130 #define DEF_BUCKET_SHIFT	12		/* log2(BUCKET) */
131 #define DEF_SEGSIZE		256
132 #define DEF_SEGSIZE_SHIFT	8		/* log2(SEGSIZE)	 */
133 #define DEF_DIRSIZE		256
134 #define DEF_FFACTOR		65536
135 #define MIN_FFACTOR		4
136 #define SPLTMAX			8
137 #define CHARKEY			"%$sniglet^&"
138 #define NUMKEY			1038583
139 #define BYTE_SHIFT		3
140 #define INT_TO_BYTE		2
141 #define INT_BYTE_SHIFT		5
142 #define ALL_SET			((u_int32_t)0xFFFFFFFF)
143 #define ALL_CLEAR		0
144 
145 #define PTROF(X)	((BUFHEAD *)((ptrdiff_t)(X)&~0x3))
146 #define ISMOD(X)	((u_int32_t)(ptrdiff_t)(X)&0x1)
147 #define DOMOD(X)	((X) = (char *)((ptrdiff_t)(X)|0x1))
148 #define ISDISK(X)	((u_int32_t)(ptrdiff_t)(X)&0x2)
149 #define DODISK(X)	((X) = (char *)((ptrdiff_t)(X)|0x2))
150 
151 #define BITS_PER_MAP	32
152 
153 /* Given the address of the beginning of a big map, clear/set the nth bit */
154 #define CLRBIT(A, N)	((A)[(N)/BITS_PER_MAP] &= ~(1<<((N)%BITS_PER_MAP)))
155 #define SETBIT(A, N)	((A)[(N)/BITS_PER_MAP] |= (1<<((N)%BITS_PER_MAP)))
156 #define ISSET(A, N)	((A)[(N)/BITS_PER_MAP] & (1<<((N)%BITS_PER_MAP)))
157 
158 /* Overflow management */
159 /*
160  * Overflow page numbers are allocated per split point.  At each doubling of
161  * the table, we can allocate extra pages.  So, an overflow page number has
162  * the top 5 bits indicate which split point and the lower 11 bits indicate
163  * which page at that split point is indicated (pages within split points are
164  * numberered starting with 1).
165  */
166 
167 #define SPLITSHIFT	11
168 #define SPLITMASK	0x7FF
169 #define SPLITNUM(N)	(((u_int32_t)(N)) >> SPLITSHIFT)
170 #define OPAGENUM(N)	((N) & SPLITMASK)
171 #define	OADDR_OF(S,O)	((u_int32_t)((u_int32_t)(S) << SPLITSHIFT) + (O))
172 
173 #define BUCKET_TO_PAGE(B) \
174 	(B) + hashp->HDRPAGES + ((B) ? hashp->SPARES[__log2((B)+1)-1] : 0)
175 #define OADDR_TO_PAGE(B) 	\
176 	BUCKET_TO_PAGE ( (1 << SPLITNUM((B))) -1 ) + OPAGENUM((B));
177 
178 /*
179  * page.h contains a detailed description of the page format.
180  *
181  * Normally, keys and data are accessed from offset tables in the top of
182  * each page which point to the beginning of the key and data.  There are
183  * four flag values which may be stored in these offset tables which indicate
184  * the following:
185  *
186  *
187  * OVFLPAGE	Rather than a key data pair, this pair contains
188  *		the address of an overflow page.  The format of
189  *		the pair is:
190  *		    OVERFLOW_PAGE_NUMBER OVFLPAGE
191  *
192  * PARTIAL_KEY	This must be the first key/data pair on a page
193  *		and implies that page contains only a partial key.
194  *		That is, the key is too big to fit on a single page
195  *		so it starts on this page and continues on the next.
196  *		The format of the page is:
197  *		    KEY_OFF PARTIAL_KEY OVFL_PAGENO OVFLPAGE
198  *
199  *		    KEY_OFF -- offset of the beginning of the key
200  *		    PARTIAL_KEY -- 1
201  *		    OVFL_PAGENO - page number of the next overflow page
202  *		    OVFLPAGE -- 0
203  *
204  * FULL_KEY	This must be the first key/data pair on the page.  It
205  *		is used in two cases.
206  *
207  *		Case 1:
208  *		    There is a complete key on the page but no data
209  *		    (because it wouldn't fit).  The next page contains
210  *		    the data.
211  *
212  *		    Page format it:
213  *		    KEY_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
214  *
215  *		    KEY_OFF -- offset of the beginning of the key
216  *		    FULL_KEY -- 2
217  *		    OVFL_PAGENO - page number of the next overflow page
218  *		    OVFLPAGE -- 0
219  *
220  *		Case 2:
221  *		    This page contains no key, but part of a large
222  *		    data field, which is continued on the next page.
223  *
224  *		    Page format it:
225  *		    DATA_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
226  *
227  *		    KEY_OFF -- offset of the beginning of the data on
228  *				this page
229  *		    FULL_KEY -- 2
230  *		    OVFL_PAGENO - page number of the next overflow page
231  *		    OVFLPAGE -- 0
232  *
233  * FULL_KEY_DATA
234  *		This must be the first key/data pair on the page.
235  *		There are two cases:
236  *
237  *		Case 1:
238  *		    This page contains a key and the beginning of the
239  *		    data field, but the data field is continued on the
240  *		    next page.
241  *
242  *		    Page format is:
243  *		    KEY_OFF FULL_KEY_DATA OVFL_PAGENO DATA_OFF
244  *
245  *		    KEY_OFF -- offset of the beginning of the key
246  *		    FULL_KEY_DATA -- 3
247  *		    OVFL_PAGENO - page number of the next overflow page
248  *		    DATA_OFF -- offset of the beginning of the data
249  *
250  *		Case 2:
251  *		    This page contains the last page of a big data pair.
252  *		    There is no key, only the  tail end of the data
253  *		    on this page.
254  *
255  *		    Page format is:
256  *		    DATA_OFF FULL_KEY_DATA <OVFL_PAGENO> <OVFLPAGE>
257  *
258  *		    DATA_OFF -- offset of the beginning of the data on
259  *				this page
260  *		    FULL_KEY_DATA -- 3
261  *		    OVFL_PAGENO - page number of the next overflow page
262  *		    OVFLPAGE -- 0
263  *
264  *		    OVFL_PAGENO and OVFLPAGE are optional (they are
265  *		    not present if there is no next page).
266  */
267 
268 #define OVFLPAGE	0
269 #define PARTIAL_KEY	1
270 #define FULL_KEY	2
271 #define FULL_KEY_DATA	3
272 #define	REAL_KEY	4
273 
274 /* Short hands for accessing structure */
275 #define BSIZE		hdr.bsize
276 #define BSHIFT		hdr.bshift
277 #define DSIZE		hdr.dsize
278 #define SGSIZE		hdr.ssize
279 #define SSHIFT		hdr.sshift
280 #define LORDER		hdr.lorder
281 #define OVFL_POINT	hdr.ovfl_point
282 #define	LAST_FREED	hdr.last_freed
283 #define MAX_BUCKET	hdr.max_bucket
284 #define FFACTOR		hdr.ffactor
285 #define HIGH_MASK	hdr.high_mask
286 #define LOW_MASK	hdr.low_mask
287 #define NKEYS		hdr.nkeys
288 #define HDRPAGES	hdr.hdrpages
289 #define SPARES		hdr.spares
290 #define BITMAPS		hdr.bitmaps
291 #define VERSION		hdr.version
292 #define MAGIC		hdr.magic
293 #define NEXT_FREE	hdr.next_free
294 #define H_CHARKEY	hdr.h_charkey
295