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. 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.
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
32 * $FreeBSD: src/lib/libc/db/btree/bt_open.c,v 1.7.2.1 2000/11/02 10:30:07 kris Exp $
33 * $DragonFly: src/lib/libc/db/btree/bt_open.c,v 1.5 2005/09/19 09:20:37 asmodai Exp $
35 * @(#)bt_open.c 8.10 (Berkeley) 8/17/94
39 * Implementation of btree access method for 4.4BSD.
41 * The design here was originally based on that of the btree access method
42 * used in the Postgres database system at UC Berkeley. This implementation
43 * is wholly independent of the Postgres code.
46 #include "namespace.h"
47 #include <sys/param.h>
58 #include "un-namespace.h"
68 static int byteorder (void);
69 static int nroot (BTREE *);
70 static int tmp (void);
73 * __BT_OPEN -- Open a btree.
75 * Creates and fills a DB struct, and calls the routine that actually
79 * fname: filename (NULL for in-memory trees)
80 * flags: open flag bits
81 * mode: open permission bits
82 * b: BTREEINFO pointer
85 * NULL on failure, pointer to DB on success.
89 __bt_open(fname, flags, mode, openinfo, dflags)
91 int flags, mode, dflags;
92 const BTREEINFO *openinfo;
106 * Intention is to make sure all of the user's selections are okay
107 * here and then use them without checking. Can't be complete, since
108 * we don't know the right page size, lorder or flags until the backing
109 * file is opened. Also, the file's page size can cause the cachesize
112 machine_lorder = byteorder();
117 if (b.flags & ~(R_DUP))
121 * Page size must be indx_t aligned and >= MINPSIZE. Default
122 * page size is set farther on, based on the underlying file
126 (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 ||
127 b.psize & (sizeof(indx_t) - 1) ))
130 /* Minimum number of keys per page; absolute minimum is 2. */
132 if (b.minkeypage < 2)
135 b.minkeypage = DEFMINKEYPAGE;
137 /* If no comparison, use default comparison and prefix. */
138 if (b.compare == NULL) {
139 b.compare = __bt_defcmp;
140 if (b.prefix == NULL)
141 b.prefix = __bt_defpfx;
145 b.lorder = machine_lorder;
147 b.compare = __bt_defcmp;
150 b.lorder = machine_lorder;
151 b.minkeypage = DEFMINKEYPAGE;
152 b.prefix = __bt_defpfx;
156 /* Check for the ubiquitous PDP-11. */
157 if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN)
160 /* Allocate and initialize DB and BTREE structures. */
161 if ((t = (BTREE *)malloc(sizeof(BTREE))) == NULL)
163 memset(t, 0, sizeof(BTREE));
164 t->bt_fd = -1; /* Don't close unopened fd on error. */
165 t->bt_lorder = b.lorder;
167 t->bt_cmp = b.compare;
168 t->bt_pfx = b.prefix;
171 if ((t->bt_dbp = dbp = (DB *)malloc(sizeof(DB))) == NULL)
173 memset(t->bt_dbp, 0, sizeof(DB));
174 if (t->bt_lorder != machine_lorder)
175 F_SET(t, B_NEEDSWAP);
177 dbp->type = DB_BTREE;
179 dbp->close = __bt_close;
180 dbp->del = __bt_delete;
185 dbp->sync = __bt_sync;
188 * If no file name was supplied, this is an in-memory btree and we
189 * open a backing temporary file. Otherwise, it's a disk-based tree.
192 switch (flags & O_ACCMODE) {
203 if ((t->bt_fd = _open(fname, flags, mode)) < 0)
207 if ((flags & O_ACCMODE) != O_RDWR)
209 if ((t->bt_fd = tmp()) == -1)
214 if (_fcntl(t->bt_fd, F_SETFD, 1) == -1)
217 if (_fstat(t->bt_fd, &sb))
220 if ((nr = _read(t->bt_fd, &m, sizeof(BTMETA))) < 0)
222 if (nr != sizeof(BTMETA))
226 * Read in the meta-data. This can change the notion of what
227 * the lorder, page size and flags are, and, when the page size
228 * changes, the cachesize value can change too. If the user
229 * specified the wrong byte order for an existing database, we
230 * don't bother to return an error, we just clear the NEEDSWAP
233 if (m.magic == BTREEMAGIC)
234 F_CLR(t, B_NEEDSWAP);
236 F_SET(t, B_NEEDSWAP);
238 M_32_SWAP(m.version);
244 if (m.magic != BTREEMAGIC || m.version != BTREEVERSION)
246 if (m.psize < MINPSIZE || m.psize > MAX_PAGE_OFFSET + 1 ||
247 m.psize & (sizeof(indx_t) - 1) )
249 if (m.flags & ~SAVEMETA)
254 t->bt_nrecs = m.nrecs;
257 * Set the page size to the best value for I/O to this file.
258 * Don't overflow the page offset type.
261 b.psize = sb.st_blksize;
262 if (b.psize < MINPSIZE)
264 if (b.psize > MAX_PAGE_OFFSET + 1)
265 b.psize = MAX_PAGE_OFFSET + 1;
268 /* Set flag if duplicates permitted. */
269 if (!(b.flags & R_DUP))
272 t->bt_free = P_INVALID;
274 F_SET(t, B_METADIRTY);
277 t->bt_psize = b.psize;
279 /* Set the cache size; must be a multiple of the page size. */
280 if (b.cachesize && b.cachesize & (b.psize - 1) )
281 b.cachesize += (~b.cachesize & (b.psize - 1) ) + 1;
282 if (b.cachesize < b.psize * MINCACHE)
283 b.cachesize = b.psize * MINCACHE;
285 /* Calculate number of pages to cache. */
286 ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize;
289 * The btree data structure requires that at least two keys can fit on
290 * a page, but other than that there's no fixed requirement. The user
291 * specified a minimum number per page, and we translated that into the
292 * number of bytes a key/data pair can use before being placed on an
293 * overflow page. This calculation includes the page header, the size
294 * of the index referencing the leaf item and the size of the leaf item
295 * structure. Also, don't let the user specify a minkeypage such that
296 * a key/data pair won't fit even if both key and data are on overflow
299 t->bt_ovflsize = (t->bt_psize - BTDATAOFF) / b.minkeypage -
300 (sizeof(indx_t) + NBLEAFDBT(0, 0));
301 if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t))
303 NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t);
305 /* Initialize the buffer pool. */
307 mpool_open(NULL, t->bt_fd, t->bt_psize, ncache)) == NULL)
309 if (!F_ISSET(t, B_INMEM))
310 mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t);
312 /* Create a root page if new tree. */
313 if (nroot(t) == RET_ERROR)
317 if (dflags & DB_LOCK)
319 if (dflags & DB_SHMEM)
320 F_SET(t, B_DB_SHMEM);
326 einval: errno = EINVAL;
329 eftype: errno = EFTYPE;
336 (void)_close(t->bt_fd);
343 * NROOT -- Create the root of a new tree.
349 * RET_ERROR, RET_SUCCESS
358 if ((meta = mpool_get(t->bt_mp, 0, 0)) != NULL) {
359 mpool_put(t->bt_mp, meta, 0);
360 return (RET_SUCCESS);
362 if (errno != EINVAL) /* It's OK to not exist. */
366 if ((meta = mpool_new(t->bt_mp, &npg)) == NULL)
369 if ((root = mpool_new(t->bt_mp, &npg)) == NULL)
375 root->prevpg = root->nextpg = P_INVALID;
376 root->lower = BTDATAOFF;
377 root->upper = t->bt_psize;
378 root->flags = P_BLEAF;
379 memset(meta, 0, t->bt_psize);
380 mpool_put(t->bt_mp, meta, MPOOL_DIRTY);
381 mpool_put(t->bt_mp, root, MPOOL_DIRTY);
382 return (RET_SUCCESS);
391 char path[MAXPATHLEN];
393 if (issetugid() == 0)
394 envtmp = getenv("TMPDIR");
396 sizeof(path), "%s/bt.XXXXXXXXXX", envtmp ? envtmp : "/tmp");
398 (void)sigfillset(&set);
399 (void)_sigprocmask(SIG_BLOCK, &set, &oset);
400 if ((fd = mkstemp(path)) != -1)
402 (void)_sigprocmask(SIG_SETMASK, &oset, NULL);
418 return (LITTLE_ENDIAN);
432 /* Toss any page pinned across calls. */
433 if (t->bt_pinned != NULL) {
434 mpool_put(t->bt_mp, t->bt_pinned, 0);
438 /* In-memory database can't have a file descriptor. */
439 if (F_ISSET(t, B_INMEM)) {