xref: /openbsd/lib/libc/db/btree/bt_overflow.c (revision 09467b48)
1 /*	$OpenBSD: bt_overflow.c,v 1.11 2015/01/16 16:48:51 deraadt 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 #include <stdio.h>
36 #include <stdlib.h>
37 #include <string.h>
38 
39 #include <db.h>
40 #include "btree.h"
41 
42 #define MINIMUM(a, b)	(((a) < (b)) ? (a) : (b))
43 
44 /*
45  * Big key/data code.
46  *
47  * Big key and data entries are stored on linked lists of pages.  The initial
48  * reference is byte string stored with the key or data and is the page number
49  * and size.  The actual record is stored in a chain of pages linked by the
50  * nextpg field of the PAGE header.
51  *
52  * The first page of the chain has a special property.  If the record is used
53  * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
54  * in the header.
55  *
56  * XXX
57  * A single DBT is written to each chain, so a lot of space on the last page
58  * is wasted.  This is a fairly major bug for some data sets.
59  */
60 
61 /*
62  * __OVFL_GET -- Get an overflow key/data item.
63  *
64  * Parameters:
65  *	t:	tree
66  *	p:	pointer to { pgno_t, u_int32_t }
67  *	buf:	storage address
68  *	bufsz:	storage size
69  *
70  * Returns:
71  *	RET_ERROR, RET_SUCCESS
72  */
73 int
74 __ovfl_get(BTREE *t, void *p, size_t *ssz, void **buf, size_t *bufsz)
75 {
76 	PAGE *h;
77 	pgno_t pg;
78 	size_t nb, plen;
79 	u_int32_t sz;
80 	void *tp;
81 
82 	memmove(&pg, p, sizeof(pgno_t));
83 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
84 	*ssz = sz;
85 
86 #ifdef DEBUG
87 	if (pg == P_INVALID || sz == 0)
88 		abort();
89 #endif
90 	/* Make the buffer bigger as necessary. */
91 	if (*bufsz < sz) {
92 		tp = realloc(*buf, sz);
93 		if (tp == NULL)
94 			return (RET_ERROR);
95 		*buf = tp;
96 		*bufsz = sz;
97 	}
98 
99 	/*
100 	 * Step through the linked list of pages, copying the data on each one
101 	 * into the buffer.  Never copy more than the data's length.
102 	 */
103 	plen = t->bt_psize - BTDATAOFF;
104 	for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) {
105 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
106 			return (RET_ERROR);
107 
108 		nb = MINIMUM(sz, plen);
109 		memmove(p, (char *)h + BTDATAOFF, nb);
110 		mpool_put(t->bt_mp, h, 0);
111 
112 		if ((sz -= nb) == 0)
113 			break;
114 	}
115 	return (RET_SUCCESS);
116 }
117 
118 /*
119  * __OVFL_PUT -- Store an overflow key/data item.
120  *
121  * Parameters:
122  *	t:	tree
123  *	data:	DBT to store
124  *	pgno:	storage page number
125  *
126  * Returns:
127  *	RET_ERROR, RET_SUCCESS
128  */
129 int
130 __ovfl_put(BTREE *t, const DBT *dbt, pgno_t *pg)
131 {
132 	PAGE *h, *last;
133 	void *p;
134 	pgno_t npg;
135 	size_t nb, plen;
136 	u_int32_t sz;
137 
138 	/*
139 	 * Allocate pages and copy the key/data record into them.  Store the
140 	 * number of the first page in the chain.
141 	 */
142 	plen = t->bt_psize - BTDATAOFF;
143 	for (last = NULL, p = dbt->data, sz = dbt->size;;
144 	    p = (char *)p + plen, last = h) {
145 		if ((h = __bt_new(t, &npg)) == NULL)
146 			return (RET_ERROR);
147 
148 		h->pgno = npg;
149 		h->nextpg = h->prevpg = P_INVALID;
150 		h->flags = P_OVERFLOW;
151 		h->lower = h->upper = 0;
152 
153 		nb = MINIMUM(sz, plen);
154 		memmove((char *)h + BTDATAOFF, p, nb);
155 
156 		if (last) {
157 			last->nextpg = h->pgno;
158 			mpool_put(t->bt_mp, last, MPOOL_DIRTY);
159 		} else
160 			*pg = h->pgno;
161 
162 		if ((sz -= nb) == 0) {
163 			mpool_put(t->bt_mp, h, MPOOL_DIRTY);
164 			break;
165 		}
166 	}
167 	return (RET_SUCCESS);
168 }
169 
170 /*
171  * __OVFL_DELETE -- Delete an overflow chain.
172  *
173  * Parameters:
174  *	t:	tree
175  *	p:	pointer to { pgno_t, u_int32_t }
176  *
177  * Returns:
178  *	RET_ERROR, RET_SUCCESS
179  */
180 int
181 __ovfl_delete(BTREE *t, void *p)
182 {
183 	PAGE *h;
184 	pgno_t pg;
185 	size_t plen;
186 	u_int32_t sz;
187 
188 	memmove(&pg, p, sizeof(pgno_t));
189 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
190 
191 #ifdef DEBUG
192 	if (pg == P_INVALID || sz == 0)
193 		abort();
194 #endif
195 	if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
196 		return (RET_ERROR);
197 
198 	/* Don't delete chains used by internal pages. */
199 	if (h->flags & P_PRESERVE) {
200 		mpool_put(t->bt_mp, h, 0);
201 		return (RET_SUCCESS);
202 	}
203 
204 	/* Step through the chain, calling the free routine for each page. */
205 	for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
206 		pg = h->nextpg;
207 		__bt_free(t, h);
208 		if (sz <= plen)
209 			break;
210 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
211 			return (RET_ERROR);
212 	}
213 	return (RET_SUCCESS);
214 }
215