xref: /openbsd-src/lib/libc/db/btree/bt_overflow.c (revision e20a56a53b18fb9479a580438ff1988d9a3762a7)
1 /*	$OpenBSD: bt_overflow.c,v 1.8 2005/03/23 19:34:58 otto Exp $	*/
2 
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  * Mike Olson.
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 #if defined(LIBC_SCCS) && !defined(lint)
36 #if 0
37 static char sccsid[] = "@(#)bt_overflow.c	8.5 (Berkeley) 7/16/94";
38 #else
39 static const char rcsid[] = "$OpenBSD: bt_overflow.c,v 1.8 2005/03/23 19:34:58 otto Exp $";
40 #endif
41 #endif /* LIBC_SCCS and not lint */
42 
43 #include <sys/param.h>
44 
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48 
49 #include <db.h>
50 #include "btree.h"
51 
52 /*
53  * Big key/data code.
54  *
55  * Big key and data entries are stored on linked lists of pages.  The initial
56  * reference is byte string stored with the key or data and is the page number
57  * and size.  The actual record is stored in a chain of pages linked by the
58  * nextpg field of the PAGE header.
59  *
60  * The first page of the chain has a special property.  If the record is used
61  * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
62  * in the header.
63  *
64  * XXX
65  * A single DBT is written to each chain, so a lot of space on the last page
66  * is wasted.  This is a fairly major bug for some data sets.
67  */
68 
69 /*
70  * __OVFL_GET -- Get an overflow key/data item.
71  *
72  * Parameters:
73  *	t:	tree
74  *	p:	pointer to { pgno_t, u_int32_t }
75  *	buf:	storage address
76  *	bufsz:	storage size
77  *
78  * Returns:
79  *	RET_ERROR, RET_SUCCESS
80  */
81 int
82 __ovfl_get(BTREE *t, void *p, size_t *ssz, void **buf, size_t *bufsz)
83 {
84 	PAGE *h;
85 	pgno_t pg;
86 	size_t nb, plen;
87 	u_int32_t sz;
88 	void *tp;
89 
90 	memmove(&pg, p, sizeof(pgno_t));
91 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
92 	*ssz = sz;
93 
94 #ifdef DEBUG
95 	if (pg == P_INVALID || sz == 0)
96 		abort();
97 #endif
98 	/* Make the buffer bigger as necessary. */
99 	if (*bufsz < sz) {
100 		tp = (char *)(*buf == NULL ? malloc(sz) : realloc(*buf, sz));
101 		if (tp == NULL)
102 			return (RET_ERROR);
103 		*buf = tp;
104 		*bufsz = sz;
105 	}
106 
107 	/*
108 	 * Step through the linked list of pages, copying the data on each one
109 	 * into the buffer.  Never copy more than the data's length.
110 	 */
111 	plen = t->bt_psize - BTDATAOFF;
112 	for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) {
113 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
114 			return (RET_ERROR);
115 
116 		nb = MIN(sz, plen);
117 		memmove(p, (char *)h + BTDATAOFF, nb);
118 		mpool_put(t->bt_mp, h, 0);
119 
120 		if ((sz -= nb) == 0)
121 			break;
122 	}
123 	return (RET_SUCCESS);
124 }
125 
126 /*
127  * __OVFL_PUT -- Store an overflow key/data item.
128  *
129  * Parameters:
130  *	t:	tree
131  *	data:	DBT to store
132  *	pgno:	storage page number
133  *
134  * Returns:
135  *	RET_ERROR, RET_SUCCESS
136  */
137 int
138 __ovfl_put(BTREE *t, const DBT *dbt, pgno_t *pg)
139 {
140 	PAGE *h, *last;
141 	void *p;
142 	pgno_t npg;
143 	size_t nb, plen;
144 	u_int32_t sz;
145 
146 	/*
147 	 * Allocate pages and copy the key/data record into them.  Store the
148 	 * number of the first page in the chain.
149 	 */
150 	plen = t->bt_psize - BTDATAOFF;
151 	for (last = NULL, p = dbt->data, sz = dbt->size;;
152 	    p = (char *)p + plen, last = h) {
153 		if ((h = __bt_new(t, &npg)) == NULL)
154 			return (RET_ERROR);
155 
156 		h->pgno = npg;
157 		h->nextpg = h->prevpg = P_INVALID;
158 		h->flags = P_OVERFLOW;
159 		h->lower = h->upper = 0;
160 
161 		nb = MIN(sz, plen);
162 		memmove((char *)h + BTDATAOFF, p, nb);
163 
164 		if (last) {
165 			last->nextpg = h->pgno;
166 			mpool_put(t->bt_mp, last, MPOOL_DIRTY);
167 		} else
168 			*pg = h->pgno;
169 
170 		if ((sz -= nb) == 0) {
171 			mpool_put(t->bt_mp, h, MPOOL_DIRTY);
172 			break;
173 		}
174 	}
175 	return (RET_SUCCESS);
176 }
177 
178 /*
179  * __OVFL_DELETE -- Delete an overflow chain.
180  *
181  * Parameters:
182  *	t:	tree
183  *	p:	pointer to { pgno_t, u_int32_t }
184  *
185  * Returns:
186  *	RET_ERROR, RET_SUCCESS
187  */
188 int
189 __ovfl_delete(BTREE *t, void *p)
190 {
191 	PAGE *h;
192 	pgno_t pg;
193 	size_t plen;
194 	u_int32_t sz;
195 
196 	memmove(&pg, p, sizeof(pgno_t));
197 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
198 
199 #ifdef DEBUG
200 	if (pg == P_INVALID || sz == 0)
201 		abort();
202 #endif
203 	if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
204 		return (RET_ERROR);
205 
206 	/* Don't delete chains used by internal pages. */
207 	if (h->flags & P_PRESERVE) {
208 		mpool_put(t->bt_mp, h, 0);
209 		return (RET_SUCCESS);
210 	}
211 
212 	/* Step through the chain, calling the free routine for each page. */
213 	for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
214 		pg = h->nextpg;
215 		__bt_free(t, h);
216 		if (sz <= plen)
217 			break;
218 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
219 			return (RET_ERROR);
220 	}
221 	return (RET_SUCCESS);
222 }
223