xref: /freebsd/lib/libc/db/hash/hash.h (revision dc36d6f9)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1990, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Margo Seltzer.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /* Operations */
36 typedef enum {
37 	HASH_GET, HASH_PUT, HASH_PUTNEW, HASH_DELETE, HASH_FIRST, HASH_NEXT
38 } ACTION;
39 
40 /* Buffer Management structures */
41 typedef struct _bufhead BUFHEAD;
42 
43 struct _bufhead {
44 	BUFHEAD		*prev;		/* LRU links */
45 	BUFHEAD		*next;		/* LRU links */
46 	BUFHEAD		*ovfl;		/* Overflow page buffer header */
47 	u_int32_t	 addr;		/* Address of this page */
48 	char		*page;		/* Actual page data */
49 	char	 	flags;
50 #define	BUF_MOD		0x0001
51 #define BUF_DISK	0x0002
52 #define	BUF_BUCKET	0x0004
53 #define	BUF_PIN		0x0008
54 };
55 
56 #define IS_BUCKET(X)	((X) & BUF_BUCKET)
57 
58 typedef BUFHEAD **SEGMENT;
59 
60 /* Hash Table Information */
61 typedef struct hashhdr {		/* Disk resident portion */
62 	int32_t		magic;		/* Magic NO for hash tables */
63 	int32_t		version;	/* Version ID */
64 	u_int32_t	lorder;		/* Byte Order */
65 	int32_t		bsize;		/* Bucket/Page Size */
66 	int32_t		bshift;		/* Bucket shift */
67 	int32_t		dsize;		/* Directory Size */
68 	int32_t		ssize;		/* Segment Size */
69 	int32_t		sshift;		/* Segment shift */
70 	int32_t		ovfl_point;	/* Where overflow pages are being
71 					 * allocated */
72 	int32_t		last_freed;	/* Last overflow page freed */
73 	u_int32_t	max_bucket;	/* ID of Maximum bucket in use */
74 	u_int32_t	high_mask;	/* Mask to modulo into entire table */
75 	u_int32_t	low_mask;	/* Mask to modulo into lower half of
76 					 * table */
77 	u_int32_t	ffactor;	/* Fill factor */
78 	int32_t		nkeys;		/* Number of keys in hash table */
79 	int32_t		hdrpages;	/* Size of table header */
80 	int32_t		h_charkey;	/* value of hash(CHARKEY) */
81 #define NCACHED	32			/* number of bit maps and spare
82 					 * points */
83 	int32_t		spares[NCACHED];/* spare pages for overflow */
84 	u_int16_t	bitmaps[NCACHED];	/* address of overflow page
85 						 * bitmaps */
86 } HASHHDR;
87 
88 typedef struct htab	 {		/* Memory resident data structure */
89 	HASHHDR 	hdr;		/* Header */
90 	int		nsegs;		/* Number of allocated segments */
91 	int		exsegs;		/* Number of extra allocated
92 					 * segments */
93 	u_int32_t			/* Hash function */
94 	    (*hash)(const void *, size_t);
95 	int		flags;		/* Flag values */
96 	int		fp;		/* File pointer */
97 	char		*tmp_buf;	/* Temporary Buffer for BIG data */
98 	char		*tmp_key;	/* Temporary Buffer for BIG keys */
99 	BUFHEAD 	*cpage;		/* Current page */
100 	int		cbucket;	/* Current bucket */
101 	int		cndx;		/* Index of next item on cpage */
102 	int		error;		/* Error Number -- for DBM
103 					 * compatibility */
104 	int		new_file;	/* Indicates if fd is backing store
105 					 * or no */
106 	int		save_file;	/* Indicates whether we need to flush
107 					 * file at
108 					 * exit */
109 	u_int32_t	*mapp[NCACHED];	/* Pointers to page maps */
110 	int		nmaps;		/* Initial number of bitmaps */
111 	int		nbufs;		/* Number of buffers left to
112 					 * allocate */
113 	BUFHEAD 	bufhead;	/* Header of buffer lru list */
114 	SEGMENT 	*dir;		/* Hash Bucket directory */
115 } HTAB;
116 
117 /*
118  * Constants
119  */
120 #define	MAX_BSIZE		32768		/* 2^15 but should be 65536 */
121 #define MIN_BUFFERS		6
122 #define MINHDRSIZE		512
123 #define DEF_BUFSIZE		65536		/* 64 K */
124 #define DEF_BUCKET_SIZE		4096
125 #define DEF_BUCKET_SHIFT	12		/* log2(BUCKET) */
126 #define DEF_SEGSIZE		256
127 #define DEF_SEGSIZE_SHIFT	8		/* log2(SEGSIZE)	 */
128 #define DEF_DIRSIZE		256
129 #define DEF_FFACTOR		65536
130 #define MIN_FFACTOR		4
131 #define SPLTMAX			8
132 #define CHARKEY			"%$sniglet^&"
133 #define NUMKEY			1038583
134 #define BYTE_SHIFT		3
135 #define INT_TO_BYTE		2
136 #define INT_BYTE_SHIFT		5
137 #define ALL_SET			((u_int32_t)0xFFFFFFFF)
138 #define ALL_CLEAR		0
139 
140 #define PTROF(X)	((BUFHEAD *)((intptr_t)(X)&~0x3))
141 #define ISMOD(X)	((u_int32_t)(intptr_t)(X)&0x1)
142 #define DOMOD(X)	((X) = (char *)((intptr_t)(X)|0x1))
143 #define ISDISK(X)	((u_int32_t)(intptr_t)(X)&0x2)
144 #define DODISK(X)	((X) = (char *)((intptr_t)(X)|0x2))
145 
146 #define BITS_PER_MAP	32
147 
148 /* Given the address of the beginning of a big map, clear/set the nth bit */
149 #define CLRBIT(A, N)	((A)[(N)/BITS_PER_MAP] &= ~(1<<((N)%BITS_PER_MAP)))
150 #define SETBIT(A, N)	((A)[(N)/BITS_PER_MAP] |= (1<<((N)%BITS_PER_MAP)))
151 #define ISSET(A, N)	((A)[(N)/BITS_PER_MAP] & (1<<((N)%BITS_PER_MAP)))
152 
153 /* Overflow management */
154 /*
155  * Overflow page numbers are allocated per split point.  At each doubling of
156  * the table, we can allocate extra pages.  So, an overflow page number has
157  * the top 5 bits indicate which split point and the lower 11 bits indicate
158  * which page at that split point is indicated (pages within split points are
159  * numberered starting with 1).
160  */
161 
162 #define SPLITSHIFT	11
163 #define SPLITMASK	0x7FF
164 #define SPLITNUM(N)	(((u_int32_t)(N)) >> SPLITSHIFT)
165 #define OPAGENUM(N)	((N) & SPLITMASK)
166 #define	OADDR_OF(S,O)	((u_int32_t)((u_int32_t)(S) << SPLITSHIFT) + (O))
167 
168 #define BUCKET_TO_PAGE(B) \
169 	(B) + hashp->HDRPAGES + ((B) ? hashp->SPARES[__log2((B)+1)-1] : 0)
170 #define OADDR_TO_PAGE(B) 	\
171 	BUCKET_TO_PAGE ( (1 << SPLITNUM((B))) -1 ) + OPAGENUM((B));
172 
173 /*
174  * page.h contains a detailed description of the page format.
175  *
176  * Normally, keys and data are accessed from offset tables in the top of
177  * each page which point to the beginning of the key and data.  There are
178  * four flag values which may be stored in these offset tables which indicate
179  * the following:
180  *
181  *
182  * OVFLPAGE	Rather than a key data pair, this pair contains
183  *		the address of an overflow page.  The format of
184  *		the pair is:
185  *		    OVERFLOW_PAGE_NUMBER OVFLPAGE
186  *
187  * PARTIAL_KEY	This must be the first key/data pair on a page
188  *		and implies that page contains only a partial key.
189  *		That is, the key is too big to fit on a single page
190  *		so it starts on this page and continues on the next.
191  *		The format of the page is:
192  *		    KEY_OFF PARTIAL_KEY OVFL_PAGENO OVFLPAGE
193  *
194  *		    KEY_OFF -- offset of the beginning of the key
195  *		    PARTIAL_KEY -- 1
196  *		    OVFL_PAGENO - page number of the next overflow page
197  *		    OVFLPAGE -- 0
198  *
199  * FULL_KEY	This must be the first key/data pair on the page.  It
200  *		is used in two cases.
201  *
202  *		Case 1:
203  *		    There is a complete key on the page but no data
204  *		    (because it wouldn't fit).  The next page contains
205  *		    the data.
206  *
207  *		    Page format it:
208  *		    KEY_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
209  *
210  *		    KEY_OFF -- offset of the beginning of the key
211  *		    FULL_KEY -- 2
212  *		    OVFL_PAGENO - page number of the next overflow page
213  *		    OVFLPAGE -- 0
214  *
215  *		Case 2:
216  *		    This page contains no key, but part of a large
217  *		    data field, which is continued on the next page.
218  *
219  *		    Page format it:
220  *		    DATA_OFF FULL_KEY OVFL_PAGENO OVFL_PAGE
221  *
222  *		    KEY_OFF -- offset of the beginning of the data on
223  *				this page
224  *		    FULL_KEY -- 2
225  *		    OVFL_PAGENO - page number of the next overflow page
226  *		    OVFLPAGE -- 0
227  *
228  * FULL_KEY_DATA
229  *		This must be the first key/data pair on the page.
230  *		There are two cases:
231  *
232  *		Case 1:
233  *		    This page contains a key and the beginning of the
234  *		    data field, but the data field is continued on the
235  *		    next page.
236  *
237  *		    Page format is:
238  *		    KEY_OFF FULL_KEY_DATA OVFL_PAGENO DATA_OFF
239  *
240  *		    KEY_OFF -- offset of the beginning of the key
241  *		    FULL_KEY_DATA -- 3
242  *		    OVFL_PAGENO - page number of the next overflow page
243  *		    DATA_OFF -- offset of the beginning of the data
244  *
245  *		Case 2:
246  *		    This page contains the last page of a big data pair.
247  *		    There is no key, only the  tail end of the data
248  *		    on this page.
249  *
250  *		    Page format is:
251  *		    DATA_OFF FULL_KEY_DATA <OVFL_PAGENO> <OVFLPAGE>
252  *
253  *		    DATA_OFF -- offset of the beginning of the data on
254  *				this page
255  *		    FULL_KEY_DATA -- 3
256  *		    OVFL_PAGENO - page number of the next overflow page
257  *		    OVFLPAGE -- 0
258  *
259  *		    OVFL_PAGENO and OVFLPAGE are optional (they are
260  *		    not present if there is no next page).
261  */
262 
263 #define OVFLPAGE	0
264 #define PARTIAL_KEY	1
265 #define FULL_KEY	2
266 #define FULL_KEY_DATA	3
267 #define	REAL_KEY	4
268 
269 /* Short hands for accessing structure */
270 #define BSIZE		hdr.bsize
271 #define BSHIFT		hdr.bshift
272 #define DSIZE		hdr.dsize
273 #define SGSIZE		hdr.ssize
274 #define SSHIFT		hdr.sshift
275 #define LORDER		hdr.lorder
276 #define OVFL_POINT	hdr.ovfl_point
277 #define	LAST_FREED	hdr.last_freed
278 #define MAX_BUCKET	hdr.max_bucket
279 #define FFACTOR		hdr.ffactor
280 #define HIGH_MASK	hdr.high_mask
281 #define LOW_MASK	hdr.low_mask
282 #define NKEYS		hdr.nkeys
283 #define HDRPAGES	hdr.hdrpages
284 #define SPARES		hdr.spares
285 #define BITMAPS		hdr.bitmaps
286 #define VERSION		hdr.version
287 #define MAGIC		hdr.magic
288 #define NEXT_FREE	hdr.next_free
289 #define H_CHARKEY	hdr.h_charkey
290