2 * Copyright (c) 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
36 * $FreeBSD: src/lib/libc/db/btree/bt_open.c,v 1.7.2.1 2000/11/02 10:30:07 kris Exp $
37 * $DragonFly: src/lib/libc/db/btree/bt_open.c,v 1.2 2003/06/17 04:26:41 dillon Exp $
39 * @(#)bt_open.c 8.10 (Berkeley) 8/17/94
43 * Implementation of btree access method for 4.4BSD.
45 * The design here was originally based on that of the btree access method
46 * used in the Postgres database system at UC Berkeley. This implementation
47 * is wholly independent of the Postgres code.
50 #include <sys/param.h>
70 static int byteorder __P((void));
71 static int nroot __P((BTREE *));
72 static int tmp __P((void));
75 * __BT_OPEN -- Open a btree.
77 * Creates and fills a DB struct, and calls the routine that actually
81 * fname: filename (NULL for in-memory trees)
82 * flags: open flag bits
83 * mode: open permission bits
84 * b: BTREEINFO pointer
87 * NULL on failure, pointer to DB on success.
91 __bt_open(fname, flags, mode, openinfo, dflags)
93 int flags, mode, dflags;
94 const BTREEINFO *openinfo;
108 * Intention is to make sure all of the user's selections are okay
109 * here and then use them without checking. Can't be complete, since
110 * we don't know the right page size, lorder or flags until the backing
111 * file is opened. Also, the file's page size can cause the cachesize
114 machine_lorder = byteorder();
119 if (b.flags & ~(R_DUP))
123 * Page size must be indx_t aligned and >= MINPSIZE. Default
124 * page size is set farther on, based on the underlying file
128 (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 ||
129 b.psize & (sizeof(indx_t) - 1) ))
132 /* Minimum number of keys per page; absolute minimum is 2. */
134 if (b.minkeypage < 2)
137 b.minkeypage = DEFMINKEYPAGE;
139 /* If no comparison, use default comparison and prefix. */
140 if (b.compare == NULL) {
141 b.compare = __bt_defcmp;
142 if (b.prefix == NULL)
143 b.prefix = __bt_defpfx;
147 b.lorder = machine_lorder;
149 b.compare = __bt_defcmp;
152 b.lorder = machine_lorder;
153 b.minkeypage = DEFMINKEYPAGE;
154 b.prefix = __bt_defpfx;
158 /* Check for the ubiquitous PDP-11. */
159 if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN)
162 /* Allocate and initialize DB and BTREE structures. */
163 if ((t = (BTREE *)malloc(sizeof(BTREE))) == NULL)
165 memset(t, 0, sizeof(BTREE));
166 t->bt_fd = -1; /* Don't close unopened fd on error. */
167 t->bt_lorder = b.lorder;
169 t->bt_cmp = b.compare;
170 t->bt_pfx = b.prefix;
173 if ((t->bt_dbp = dbp = (DB *)malloc(sizeof(DB))) == NULL)
175 memset(t->bt_dbp, 0, sizeof(DB));
176 if (t->bt_lorder != machine_lorder)
177 F_SET(t, B_NEEDSWAP);
179 dbp->type = DB_BTREE;
181 dbp->close = __bt_close;
182 dbp->del = __bt_delete;
187 dbp->sync = __bt_sync;
190 * If no file name was supplied, this is an in-memory btree and we
191 * open a backing temporary file. Otherwise, it's a disk-based tree.
194 switch (flags & O_ACCMODE) {
205 if ((t->bt_fd = _open(fname, flags, mode)) < 0)
209 if ((flags & O_ACCMODE) != O_RDWR)
211 if ((t->bt_fd = tmp()) == -1)
216 if (_fcntl(t->bt_fd, F_SETFD, 1) == -1)
219 if (fstat(t->bt_fd, &sb))
222 if ((nr = _read(t->bt_fd, &m, sizeof(BTMETA))) < 0)
224 if (nr != sizeof(BTMETA))
228 * Read in the meta-data. This can change the notion of what
229 * the lorder, page size and flags are, and, when the page size
230 * changes, the cachesize value can change too. If the user
231 * specified the wrong byte order for an existing database, we
232 * don't bother to return an error, we just clear the NEEDSWAP
235 if (m.magic == BTREEMAGIC)
236 F_CLR(t, B_NEEDSWAP);
238 F_SET(t, B_NEEDSWAP);
240 M_32_SWAP(m.version);
246 if (m.magic != BTREEMAGIC || m.version != BTREEVERSION)
248 if (m.psize < MINPSIZE || m.psize > MAX_PAGE_OFFSET + 1 ||
249 m.psize & (sizeof(indx_t) - 1) )
251 if (m.flags & ~SAVEMETA)
256 t->bt_nrecs = m.nrecs;
259 * Set the page size to the best value for I/O to this file.
260 * Don't overflow the page offset type.
263 b.psize = sb.st_blksize;
264 if (b.psize < MINPSIZE)
266 if (b.psize > MAX_PAGE_OFFSET + 1)
267 b.psize = MAX_PAGE_OFFSET + 1;
270 /* Set flag if duplicates permitted. */
271 if (!(b.flags & R_DUP))
274 t->bt_free = P_INVALID;
276 F_SET(t, B_METADIRTY);
279 t->bt_psize = b.psize;
281 /* Set the cache size; must be a multiple of the page size. */
282 if (b.cachesize && b.cachesize & (b.psize - 1) )
283 b.cachesize += (~b.cachesize & (b.psize - 1) ) + 1;
284 if (b.cachesize < b.psize * MINCACHE)
285 b.cachesize = b.psize * MINCACHE;
287 /* Calculate number of pages to cache. */
288 ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize;
291 * The btree data structure requires that at least two keys can fit on
292 * a page, but other than that there's no fixed requirement. The user
293 * specified a minimum number per page, and we translated that into the
294 * number of bytes a key/data pair can use before being placed on an
295 * overflow page. This calculation includes the page header, the size
296 * of the index referencing the leaf item and the size of the leaf item
297 * structure. Also, don't let the user specify a minkeypage such that
298 * a key/data pair won't fit even if both key and data are on overflow
301 t->bt_ovflsize = (t->bt_psize - BTDATAOFF) / b.minkeypage -
302 (sizeof(indx_t) + NBLEAFDBT(0, 0));
303 if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t))
305 NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t);
307 /* Initialize the buffer pool. */
309 mpool_open(NULL, t->bt_fd, t->bt_psize, ncache)) == NULL)
311 if (!F_ISSET(t, B_INMEM))
312 mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t);
314 /* Create a root page if new tree. */
315 if (nroot(t) == RET_ERROR)
319 if (dflags & DB_LOCK)
321 if (dflags & DB_SHMEM)
322 F_SET(t, B_DB_SHMEM);
328 einval: errno = EINVAL;
331 eftype: errno = EFTYPE;
338 (void)_close(t->bt_fd);
345 * NROOT -- Create the root of a new tree.
351 * RET_ERROR, RET_SUCCESS
360 if ((meta = mpool_get(t->bt_mp, 0, 0)) != NULL) {
361 mpool_put(t->bt_mp, meta, 0);
362 return (RET_SUCCESS);
364 if (errno != EINVAL) /* It's OK to not exist. */
368 if ((meta = mpool_new(t->bt_mp, &npg)) == NULL)
371 if ((root = mpool_new(t->bt_mp, &npg)) == NULL)
377 root->prevpg = root->nextpg = P_INVALID;
378 root->lower = BTDATAOFF;
379 root->upper = t->bt_psize;
380 root->flags = P_BLEAF;
381 memset(meta, 0, t->bt_psize);
382 mpool_put(t->bt_mp, meta, MPOOL_DIRTY);
383 mpool_put(t->bt_mp, root, MPOOL_DIRTY);
384 return (RET_SUCCESS);
393 char path[MAXPATHLEN];
395 if (issetugid() == 0)
396 envtmp = getenv("TMPDIR");
398 sizeof(path), "%s/bt.XXXXXXXXXX", envtmp ? envtmp : "/tmp");
400 (void)sigfillset(&set);
401 (void)sigprocmask(SIG_BLOCK, &set, &oset);
402 if ((fd = mkstemp(path)) != -1)
404 (void)sigprocmask(SIG_SETMASK, &oset, NULL);
420 return (LITTLE_ENDIAN);
434 /* Toss any page pinned across calls. */
435 if (t->bt_pinned != NULL) {
436 mpool_put(t->bt_mp, t->bt_pinned, 0);
440 /* In-memory database can't have a file descriptor. */
441 if (F_ISSET(t, B_INMEM)) {