xref: /freebsd/lib/libc/db/btree/bt_overflow.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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  * 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  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #if defined(LIBC_SCCS) && !defined(lint)
34 static char sccsid[] = "@(#)bt_overflow.c	8.5 (Berkeley) 7/16/94";
35 #endif /* LIBC_SCCS and not lint */
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 #include <sys/param.h>
40 
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <string.h>
44 
45 #include <db.h>
46 #include "btree.h"
47 
48 /*
49  * Big key/data code.
50  *
51  * Big key and data entries are stored on linked lists of pages.  The initial
52  * reference is byte string stored with the key or data and is the page number
53  * and size.  The actual record is stored in a chain of pages linked by the
54  * nextpg field of the PAGE header.
55  *
56  * The first page of the chain has a special property.  If the record is used
57  * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
58  * in the header.
59  *
60  * XXX
61  * A single DBT is written to each chain, so a lot of space on the last page
62  * is wasted.  This is a fairly major bug for some data sets.
63  */
64 
65 /*
66  * __OVFL_GET -- Get an overflow key/data item.
67  *
68  * Parameters:
69  *	t:	tree
70  *	p:	pointer to { pgno_t, u_int32_t }
71  *	buf:	storage address
72  *	bufsz:	storage size
73  *
74  * Returns:
75  *	RET_ERROR, RET_SUCCESS
76  */
77 int
78 __ovfl_get(t, p, ssz, buf, bufsz)
79 	BTREE *t;
80 	void *p;
81 	size_t *ssz;
82 	void **buf;
83 	size_t *bufsz;
84 {
85 	PAGE *h;
86 	pgno_t pg;
87 	size_t nb, plen;
88 	u_int32_t sz;
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 		*buf = (char *)(*buf == NULL ? malloc(sz) : reallocf(*buf, sz));
101 		if (*buf == 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
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;
146 	u_int32_t sz;
147 
148 	/*
149 	 * Allocate pages and copy the key/data record into them.  Store the
150 	 * number of the first page in the chain.
151 	 */
152 	plen = t->bt_psize - BTDATAOFF;
153 	for (last = NULL, p = dbt->data, sz = dbt->size;;
154 	    p = (char *)p + plen, last = h) {
155 		if ((h = __bt_new(t, &npg)) == NULL)
156 			return (RET_ERROR);
157 
158 		h->pgno = npg;
159 		h->nextpg = h->prevpg = P_INVALID;
160 		h->flags = P_OVERFLOW;
161 		h->lower = h->upper = 0;
162 
163 		nb = MIN(sz, plen);
164 		memmove((char *)h + BTDATAOFF, p, nb);
165 
166 		if (last) {
167 			last->nextpg = h->pgno;
168 			mpool_put(t->bt_mp, last, MPOOL_DIRTY);
169 		} else
170 			*pg = h->pgno;
171 
172 		if ((sz -= nb) == 0) {
173 			mpool_put(t->bt_mp, h, MPOOL_DIRTY);
174 			break;
175 		}
176 	}
177 	return (RET_SUCCESS);
178 }
179 
180 /*
181  * __OVFL_DELETE -- Delete an overflow chain.
182  *
183  * Parameters:
184  *	t:	tree
185  *	p:	pointer to { pgno_t, u_int32_t }
186  *
187  * Returns:
188  *	RET_ERROR, RET_SUCCESS
189  */
190 int
191 __ovfl_delete(t, p)
192 	BTREE *t;
193 	void *p;
194 {
195 	PAGE *h;
196 	pgno_t pg;
197 	size_t plen;
198 	u_int32_t sz;
199 
200 	memmove(&pg, p, sizeof(pgno_t));
201 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
202 
203 #ifdef DEBUG
204 	if (pg == P_INVALID || sz == 0)
205 		abort();
206 #endif
207 	if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
208 		return (RET_ERROR);
209 
210 	/* Don't delete chains used by internal pages. */
211 	if (h->flags & P_PRESERVE) {
212 		mpool_put(t->bt_mp, h, 0);
213 		return (RET_SUCCESS);
214 	}
215 
216 	/* Step through the chain, calling the free routine for each page. */
217 	for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
218 		pg = h->nextpg;
219 		__bt_free(t, h);
220 		if (sz <= plen)
221 			break;
222 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
223 			return (RET_ERROR);
224 	}
225 	return (RET_SUCCESS);
226 }
227