bt_overflow.c revision 1.11 1 1.11 simonb /* $NetBSD: bt_overflow.c,v 1.11 2003/01/20 07:12:12 simonb Exp $ */
2 1.5 cgd
3 1.1 cgd /*-
4 1.4 cgd * Copyright (c) 1990, 1993, 1994
5 1.1 cgd * The Regents of the University of California. All rights reserved.
6 1.1 cgd *
7 1.1 cgd * This code is derived from software contributed to Berkeley by
8 1.1 cgd * Mike Olson.
9 1.1 cgd *
10 1.1 cgd * Redistribution and use in source and binary forms, with or without
11 1.1 cgd * modification, are permitted provided that the following conditions
12 1.1 cgd * are met:
13 1.1 cgd * 1. Redistributions of source code must retain the above copyright
14 1.1 cgd * notice, this list of conditions and the following disclaimer.
15 1.1 cgd * 2. Redistributions in binary form must reproduce the above copyright
16 1.1 cgd * notice, this list of conditions and the following disclaimer in the
17 1.1 cgd * documentation and/or other materials provided with the distribution.
18 1.1 cgd * 3. All advertising materials mentioning features or use of this software
19 1.1 cgd * must display the following acknowledgement:
20 1.1 cgd * This product includes software developed by the University of
21 1.1 cgd * California, Berkeley and its contributors.
22 1.1 cgd * 4. Neither the name of the University nor the names of its contributors
23 1.1 cgd * may be used to endorse or promote products derived from this software
24 1.1 cgd * without specific prior written permission.
25 1.1 cgd *
26 1.1 cgd * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 1.1 cgd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 1.1 cgd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 1.1 cgd * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 1.1 cgd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 1.1 cgd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 1.1 cgd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 1.1 cgd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 1.1 cgd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 1.1 cgd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 1.1 cgd * SUCH DAMAGE.
37 1.1 cgd */
38 1.1 cgd
39 1.7 christos #include <sys/cdefs.h>
40 1.1 cgd #if defined(LIBC_SCCS) && !defined(lint)
41 1.5 cgd #if 0
42 1.6 cgd static char sccsid[] = "@(#)bt_overflow.c 8.5 (Berkeley) 7/16/94";
43 1.5 cgd #else
44 1.11 simonb __RCSID("$NetBSD: bt_overflow.c,v 1.11 2003/01/20 07:12:12 simonb Exp $");
45 1.5 cgd #endif
46 1.1 cgd #endif /* LIBC_SCCS and not lint */
47 1.1 cgd
48 1.8 jtc #include "namespace.h"
49 1.1 cgd #include <sys/param.h>
50 1.1 cgd
51 1.1 cgd #include <stdio.h>
52 1.1 cgd #include <stdlib.h>
53 1.1 cgd #include <string.h>
54 1.1 cgd
55 1.1 cgd #include <db.h>
56 1.1 cgd #include "btree.h"
57 1.1 cgd
58 1.1 cgd /*
59 1.1 cgd * Big key/data code.
60 1.1 cgd *
61 1.1 cgd * Big key and data entries are stored on linked lists of pages. The initial
62 1.1 cgd * reference is byte string stored with the key or data and is the page number
63 1.1 cgd * and size. The actual record is stored in a chain of pages linked by the
64 1.1 cgd * nextpg field of the PAGE header.
65 1.1 cgd *
66 1.1 cgd * The first page of the chain has a special property. If the record is used
67 1.1 cgd * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
68 1.1 cgd * in the header.
69 1.1 cgd *
70 1.1 cgd * XXX
71 1.1 cgd * A single DBT is written to each chain, so a lot of space on the last page
72 1.1 cgd * is wasted. This is a fairly major bug for some data sets.
73 1.1 cgd */
74 1.1 cgd
75 1.1 cgd /*
76 1.1 cgd * __OVFL_GET -- Get an overflow key/data item.
77 1.1 cgd *
78 1.1 cgd * Parameters:
79 1.1 cgd * t: tree
80 1.4 cgd * p: pointer to { pgno_t, u_int32_t }
81 1.1 cgd * buf: storage address
82 1.1 cgd * bufsz: storage size
83 1.1 cgd *
84 1.1 cgd * Returns:
85 1.1 cgd * RET_ERROR, RET_SUCCESS
86 1.1 cgd */
87 1.1 cgd int
88 1.1 cgd __ovfl_get(t, p, ssz, buf, bufsz)
89 1.1 cgd BTREE *t;
90 1.1 cgd void *p;
91 1.1 cgd size_t *ssz;
92 1.6 cgd void **buf;
93 1.1 cgd size_t *bufsz;
94 1.1 cgd {
95 1.1 cgd PAGE *h;
96 1.1 cgd pgno_t pg;
97 1.4 cgd size_t nb, plen;
98 1.4 cgd u_int32_t sz;
99 1.1 cgd
100 1.1 cgd memmove(&pg, p, sizeof(pgno_t));
101 1.4 cgd memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
102 1.1 cgd *ssz = sz;
103 1.1 cgd
104 1.1 cgd #ifdef DEBUG
105 1.1 cgd if (pg == P_INVALID || sz == 0)
106 1.1 cgd abort();
107 1.1 cgd #endif
108 1.1 cgd /* Make the buffer bigger as necessary. */
109 1.1 cgd if (*bufsz < sz) {
110 1.4 cgd *buf = (char *)(*buf == NULL ? malloc(sz) : realloc(*buf, sz));
111 1.4 cgd if (*buf == NULL)
112 1.1 cgd return (RET_ERROR);
113 1.1 cgd *bufsz = sz;
114 1.1 cgd }
115 1.1 cgd
116 1.1 cgd /*
117 1.1 cgd * Step through the linked list of pages, copying the data on each one
118 1.1 cgd * into the buffer. Never copy more than the data's length.
119 1.1 cgd */
120 1.1 cgd plen = t->bt_psize - BTDATAOFF;
121 1.1 cgd for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) {
122 1.1 cgd if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
123 1.1 cgd return (RET_ERROR);
124 1.1 cgd
125 1.1 cgd nb = MIN(sz, plen);
126 1.9 christos memmove(p, (char *)(void *)h + BTDATAOFF, nb);
127 1.1 cgd mpool_put(t->bt_mp, h, 0);
128 1.1 cgd
129 1.1 cgd if ((sz -= nb) == 0)
130 1.1 cgd break;
131 1.1 cgd }
132 1.1 cgd return (RET_SUCCESS);
133 1.1 cgd }
134 1.1 cgd
135 1.1 cgd /*
136 1.1 cgd * __OVFL_PUT -- Store an overflow key/data item.
137 1.1 cgd *
138 1.1 cgd * Parameters:
139 1.1 cgd * t: tree
140 1.1 cgd * data: DBT to store
141 1.1 cgd * pgno: storage page number
142 1.1 cgd *
143 1.1 cgd * Returns:
144 1.1 cgd * RET_ERROR, RET_SUCCESS
145 1.1 cgd */
146 1.1 cgd int
147 1.1 cgd __ovfl_put(t, dbt, pg)
148 1.1 cgd BTREE *t;
149 1.1 cgd const DBT *dbt;
150 1.1 cgd pgno_t *pg;
151 1.1 cgd {
152 1.1 cgd PAGE *h, *last;
153 1.1 cgd void *p;
154 1.1 cgd pgno_t npg;
155 1.4 cgd size_t nb, plen;
156 1.4 cgd u_int32_t sz;
157 1.1 cgd
158 1.1 cgd /*
159 1.1 cgd * Allocate pages and copy the key/data record into them. Store the
160 1.1 cgd * number of the first page in the chain.
161 1.1 cgd */
162 1.1 cgd plen = t->bt_psize - BTDATAOFF;
163 1.11 simonb for (last = NULL, p = dbt->data, sz = dbt->size;;
164 1.1 cgd p = (char *)p + plen, last = h) {
165 1.1 cgd if ((h = __bt_new(t, &npg)) == NULL)
166 1.1 cgd return (RET_ERROR);
167 1.1 cgd
168 1.1 cgd h->pgno = npg;
169 1.1 cgd h->nextpg = h->prevpg = P_INVALID;
170 1.1 cgd h->flags = P_OVERFLOW;
171 1.1 cgd h->lower = h->upper = 0;
172 1.1 cgd
173 1.1 cgd nb = MIN(sz, plen);
174 1.9 christos memmove((char *)(void *)h + BTDATAOFF, p, nb);
175 1.1 cgd
176 1.1 cgd if (last) {
177 1.1 cgd last->nextpg = h->pgno;
178 1.1 cgd mpool_put(t->bt_mp, last, MPOOL_DIRTY);
179 1.1 cgd } else
180 1.1 cgd *pg = h->pgno;
181 1.1 cgd
182 1.1 cgd if ((sz -= nb) == 0) {
183 1.1 cgd mpool_put(t->bt_mp, h, MPOOL_DIRTY);
184 1.1 cgd break;
185 1.1 cgd }
186 1.1 cgd }
187 1.1 cgd return (RET_SUCCESS);
188 1.1 cgd }
189 1.1 cgd
190 1.1 cgd /*
191 1.1 cgd * __OVFL_DELETE -- Delete an overflow chain.
192 1.1 cgd *
193 1.1 cgd * Parameters:
194 1.1 cgd * t: tree
195 1.4 cgd * p: pointer to { pgno_t, u_int32_t }
196 1.1 cgd *
197 1.1 cgd * Returns:
198 1.1 cgd * RET_ERROR, RET_SUCCESS
199 1.1 cgd */
200 1.1 cgd int
201 1.1 cgd __ovfl_delete(t, p)
202 1.1 cgd BTREE *t;
203 1.1 cgd void *p;
204 1.1 cgd {
205 1.1 cgd PAGE *h;
206 1.1 cgd pgno_t pg;
207 1.4 cgd size_t plen;
208 1.4 cgd u_int32_t sz;
209 1.1 cgd
210 1.1 cgd memmove(&pg, p, sizeof(pgno_t));
211 1.4 cgd memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
212 1.1 cgd
213 1.1 cgd #ifdef DEBUG
214 1.1 cgd if (pg == P_INVALID || sz == 0)
215 1.1 cgd abort();
216 1.1 cgd #endif
217 1.1 cgd if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
218 1.1 cgd return (RET_ERROR);
219 1.1 cgd
220 1.1 cgd /* Don't delete chains used by internal pages. */
221 1.1 cgd if (h->flags & P_PRESERVE) {
222 1.1 cgd mpool_put(t->bt_mp, h, 0);
223 1.1 cgd return (RET_SUCCESS);
224 1.1 cgd }
225 1.1 cgd
226 1.1 cgd /* Step through the chain, calling the free routine for each page. */
227 1.1 cgd for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
228 1.1 cgd pg = h->nextpg;
229 1.1 cgd __bt_free(t, h);
230 1.1 cgd if (sz <= plen)
231 1.1 cgd break;
232 1.1 cgd if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
233 1.1 cgd return (RET_ERROR);
234 1.1 cgd }
235 1.1 cgd return (RET_SUCCESS);
236 1.1 cgd }
237