xref: /freebsd/lib/libc/db/btree/bt_overflow.c (revision 2008043f386721d58158e37e0d7e50df8095942d)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
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 static char sccsid[] = "@(#)bt_overflow.c	8.5 (Berkeley) 7/16/94";
37 #endif /* LIBC_SCCS and not lint */
38 #include <sys/param.h>
39 
40 #include <stdio.h>
41 #include <stdlib.h>
42 #include <string.h>
43 
44 #include <db.h>
45 #include "btree.h"
46 
47 /*
48  * Big key/data code.
49  *
50  * Big key and data entries are stored on linked lists of pages.  The initial
51  * reference is byte string stored with the key or data and is the page number
52  * and size.  The actual record is stored in a chain of pages linked by the
53  * nextpg field of the PAGE header.
54  *
55  * The first page of the chain has a special property.  If the record is used
56  * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set
57  * in the header.
58  *
59  * XXX
60  * A single DBT is written to each chain, so a lot of space on the last page
61  * is wasted.  This is a fairly major bug for some data sets.
62  */
63 
64 /*
65  * __OVFL_GET -- Get an overflow key/data item.
66  *
67  * Parameters:
68  *	t:	tree
69  *	p:	pointer to { pgno_t, u_int32_t }
70  *	buf:	storage address
71  *	bufsz:	storage size
72  *
73  * Returns:
74  *	RET_ERROR, RET_SUCCESS
75  */
76 int
77 __ovfl_get(BTREE *t, void *p, size_t *ssz, void **buf, size_t *bufsz)
78 {
79 	PAGE *h;
80 	pgno_t pg;
81 	size_t nb, plen;
82 	u_int32_t sz;
83 
84 	memmove(&pg, p, sizeof(pgno_t));
85 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
86 	*ssz = sz;
87 
88 #ifdef DEBUG
89 	if (pg == P_INVALID || sz == 0)
90 		abort();
91 #endif
92 	/* Make the buffer bigger as necessary. */
93 	if (*bufsz < sz) {
94 		*buf = reallocf(*buf, sz);
95 		if (*buf == NULL)
96 			return (RET_ERROR);
97 		*bufsz = sz;
98 	}
99 
100 	/*
101 	 * Step through the linked list of pages, copying the data on each one
102 	 * into the buffer.  Never copy more than the data's length.
103 	 */
104 	plen = t->bt_psize - BTDATAOFF;
105 	for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) {
106 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
107 			return (RET_ERROR);
108 
109 		nb = MIN(sz, plen);
110 		memmove(p, (char *)h + BTDATAOFF, nb);
111 		mpool_put(t->bt_mp, h, 0);
112 
113 		if ((sz -= nb) == 0)
114 			break;
115 	}
116 	return (RET_SUCCESS);
117 }
118 
119 /*
120  * __OVFL_PUT -- Store an overflow key/data item.
121  *
122  * Parameters:
123  *	t:	tree
124  *	data:	DBT to store
125  *	pgno:	storage page number
126  *
127  * Returns:
128  *	RET_ERROR, RET_SUCCESS
129  */
130 int
131 __ovfl_put(BTREE *t, const DBT *dbt, pgno_t *pg)
132 {
133 	PAGE *h, *last;
134 	void *p;
135 	pgno_t npg;
136 	size_t nb, plen;
137 	u_int32_t sz;
138 
139 	/*
140 	 * Allocate pages and copy the key/data record into them.  Store the
141 	 * number of the first page in the chain.
142 	 */
143 	plen = t->bt_psize - BTDATAOFF;
144 	for (last = NULL, p = dbt->data, sz = dbt->size;;
145 	    p = (char *)p + plen, last = h) {
146 		if ((h = __bt_new(t, &npg)) == NULL)
147 			return (RET_ERROR);
148 
149 		h->pgno = npg;
150 		h->nextpg = h->prevpg = P_INVALID;
151 		h->flags = P_OVERFLOW;
152 		h->lower = h->upper = 0;
153 
154 		nb = MIN(sz, plen);
155 		memmove((char *)h + BTDATAOFF, p, nb);
156 
157 		if (last) {
158 			last->nextpg = h->pgno;
159 			mpool_put(t->bt_mp, last, MPOOL_DIRTY);
160 		} else
161 			*pg = h->pgno;
162 
163 		if ((sz -= nb) == 0) {
164 			mpool_put(t->bt_mp, h, MPOOL_DIRTY);
165 			break;
166 		}
167 	}
168 	return (RET_SUCCESS);
169 }
170 
171 /*
172  * __OVFL_DELETE -- Delete an overflow chain.
173  *
174  * Parameters:
175  *	t:	tree
176  *	p:	pointer to { pgno_t, u_int32_t }
177  *
178  * Returns:
179  *	RET_ERROR, RET_SUCCESS
180  */
181 int
182 __ovfl_delete(BTREE *t, void *p)
183 {
184 	PAGE *h;
185 	pgno_t pg;
186 	size_t plen;
187 	u_int32_t sz;
188 
189 	memmove(&pg, p, sizeof(pgno_t));
190 	memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(u_int32_t));
191 
192 #ifdef DEBUG
193 	if (pg == P_INVALID || sz == 0)
194 		abort();
195 #endif
196 	if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
197 		return (RET_ERROR);
198 
199 	/* Don't delete chains used by internal pages. */
200 	if (h->flags & P_PRESERVE) {
201 		mpool_put(t->bt_mp, h, 0);
202 		return (RET_SUCCESS);
203 	}
204 
205 	/* Step through the chain, calling the free routine for each page. */
206 	for (plen = t->bt_psize - BTDATAOFF;; sz -= plen) {
207 		pg = h->nextpg;
208 		__bt_free(t, h);
209 		if (sz <= plen)
210 			break;
211 		if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
212 			return (RET_ERROR);
213 	}
214 	return (RET_SUCCESS);
215 }
216