xref: /386bsd/usr/src/lib/libc/db/btree/bt_overflow.c (revision a2142627)
1 /*-
2  * Copyright (c) 1990 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Mike Olson.
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 
37 #if defined(LIBC_SCCS) && !defined(lint)
38 static char sccsid[] = "@(#)bt_overflow.c	5.8 (Berkeley) 2/16/93";
39 #endif /* LIBC_SCCS and not lint */
40 
41 #include <sys/param.h>
42 
43 #include <stdio.h>
44 #include <stdlib.h>
45 #include <string.h>
46 
47 #include <db.h>
48 #include "btree.h"
49 
50 /*
51  * Big key/data code.
52  *
53  * Big key and data entries are stored on linked lists of pages.  The initial
54  * reference is byte string stored with the key or data and is the page number
55  * and size.  The actual record is stored in a chain of pages linked by the
56  * nextpg field of the PAGE header.
57  *
58  * The first page of the chain has a special property.  If the record is used
59  * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
60  * in the header.
61  *
62  * XXX
63  * A single DBT is written to each chain, so a lot of space on the last page
64  * is wasted.  This is a fairly major bug for some data sets.
65  */
66 
67 /*
68  * __OVFL_GET -- Get an overflow key/data item.
69  *
70  * Parameters:
71  *	t:	tree
72  *	p:	pointer to { pgno_t, size_t }
73  *	buf:	storage address
74  *	bufsz:	storage size
75  *
76  * Returns:
77  *	RET_ERROR, RET_SUCCESS
78  */
79 int
__ovfl_get(t,p,ssz,buf,bufsz)80 __ovfl_get(t, p, ssz, buf, bufsz)
81 	BTREE *t;
82 	void *p;
83 	size_t *ssz;
84 	char **buf;
85 	size_t *bufsz;
86 {
87 	PAGE *h;
88 	pgno_t pg;
89 	size_t nb, plen, sz;
90 
91 	memmove(&pg, p, sizeof(pgno_t));
92 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(size_t));
93 	*ssz = sz;
94 
95 #ifdef DEBUG
96 	if (pg == P_INVALID || sz == 0)
97 		abort();
98 #endif
99 	/* Make the buffer bigger as necessary. */
100 	if (*bufsz < sz) {
101 		if ((*buf = realloc(*buf, sz)) == NULL)
102 			return (RET_ERROR);
103 		*bufsz = sz;
104 	}
105 
106 	/*
107 	 * Step through the linked list of pages, copying the data on each one
108 	 * into the buffer.  Never copy more than the data's length.
109 	 */
110 	plen = t->bt_psize - BTDATAOFF;
111 	for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) {
112 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
113 			return (RET_ERROR);
114 
115 		nb = MIN(sz, plen);
116 		memmove(p, (char *)h + BTDATAOFF, nb);
117 		mpool_put(t->bt_mp, h, 0);
118 
119 		if ((sz -= nb) == 0)
120 			break;
121 	}
122 	return (RET_SUCCESS);
123 }
124 
125 /*
126  * __OVFL_PUT -- Store an overflow key/data item.
127  *
128  * Parameters:
129  *	t:	tree
130  *	data:	DBT to store
131  *	pgno:	storage page number
132  *
133  * Returns:
134  *	RET_ERROR, RET_SUCCESS
135  */
136 int
__ovfl_put(t,dbt,pg)137 __ovfl_put(t, dbt, pg)
138 	BTREE *t;
139 	const DBT *dbt;
140 	pgno_t *pg;
141 {
142 	PAGE *h, *last;
143 	void *p;
144 	pgno_t npg;
145 	size_t nb, plen, sz;
146 
147 	/*
148 	 * Allocate pages and copy the key/data record into them.  Store the
149 	 * number of the first page in the chain.
150 	 */
151 	plen = t->bt_psize - BTDATAOFF;
152 	for (last = NULL, p = dbt->data, sz = dbt->size;;
153 	    p = (char *)p + plen, last = h) {
154 		if ((h = __bt_new(t, &npg)) == NULL)
155 			return (RET_ERROR);
156 
157 		h->pgno = npg;
158 		h->nextpg = h->prevpg = P_INVALID;
159 		h->flags = P_OVERFLOW;
160 		h->lower = h->upper = 0;
161 
162 		nb = MIN(sz, plen);
163 		memmove((char *)h + BTDATAOFF, p, nb);
164 
165 		if (last) {
166 			last->nextpg = h->pgno;
167 			mpool_put(t->bt_mp, last, MPOOL_DIRTY);
168 		} else
169 			*pg = h->pgno;
170 
171 		if ((sz -= nb) == 0) {
172 			mpool_put(t->bt_mp, h, MPOOL_DIRTY);
173 			break;
174 		}
175 	}
176 	return (RET_SUCCESS);
177 }
178 
179 /*
180  * __OVFL_DELETE -- Delete an overflow chain.
181  *
182  * Parameters:
183  *	t:	tree
184  *	p:	pointer to { pgno_t, size_t }
185  *
186  * Returns:
187  *	RET_ERROR, RET_SUCCESS
188  */
189 int
__ovfl_delete(t,p)190 __ovfl_delete(t, p)
191 	BTREE *t;
192 	void *p;
193 {
194 	PAGE *h;
195 	pgno_t pg;
196 	size_t plen, sz;
197 
198 	memmove(&pg, p, sizeof(pgno_t));
199 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(size_t));
200 
201 #ifdef DEBUG
202 	if (pg == P_INVALID || sz == 0)
203 		abort();
204 #endif
205 	if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
206 		return (RET_ERROR);
207 
208 	/* Don't delete chains used by internal pages. */
209 	if (h->flags & P_PRESERVE) {
210 		mpool_put(t->bt_mp, h, 0);
211 		return (RET_SUCCESS);
212 	}
213 
214 	/* Step through the chain, calling the free routine for each page. */
215 	for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
216 		pg = h->nextpg;
217 		__bt_free(t, h);
218 		if (sz <= plen)
219 			break;
220 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
221 			return (RET_ERROR);
222 	}
223 	return (RET_SUCCESS);
224 }
225