Home | History | Annotate | Line # | Download | only in btree
bt_open.c revision 1.7
      1 /*	$NetBSD: bt_open.c,v 1.7 1995/02/27 13:20:27 cgd 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. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *	This product includes software developed by the University of
     21  *	California, Berkeley and its contributors.
     22  * 4. Neither the name of the University nor the names of its contributors
     23  *    may be used to endorse or promote products derived from this software
     24  *    without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     36  * SUCH DAMAGE.
     37  */
     38 
     39 #if defined(LIBC_SCCS) && !defined(lint)
     40 #if 0
     41 static char sccsid[] = "@(#)bt_open.c	8.7 (Berkeley) 6/16/94";
     42 #else
     43 static char rcsid[] = "$NetBSD: bt_open.c,v 1.7 1995/02/27 13:20:27 cgd Exp $";
     44 #endif
     45 #endif /* LIBC_SCCS and not lint */
     46 
     47 /*
     48  * Implementation of btree access method for 4.4BSD.
     49  *
     50  * The design here was originally based on that of the btree access method
     51  * used in the Postgres database system at UC Berkeley.  This implementation
     52  * is wholly independent of the Postgres code.
     53  */
     54 
     55 #include <sys/param.h>
     56 #include <sys/stat.h>
     57 
     58 #include <errno.h>
     59 #include <fcntl.h>
     60 #include <limits.h>
     61 #include <signal.h>
     62 #include <stdio.h>
     63 #include <stdlib.h>
     64 #include <string.h>
     65 #include <unistd.h>
     66 
     67 #include <db.h>
     68 #include "btree.h"
     69 
     70 static int byteorder __P((void));
     71 static int nroot __P((BTREE *));
     72 static int tmp __P((void));
     73 
     74 /*
     75  * __BT_OPEN -- Open a btree.
     76  *
     77  * Creates and fills a DB struct, and calls the routine that actually
     78  * opens the btree.
     79  *
     80  * Parameters:
     81  *	fname:	filename (NULL for in-memory trees)
     82  *	flags:	open flag bits
     83  *	mode:	open permission bits
     84  *	b:	BTREEINFO pointer
     85  *
     86  * Returns:
     87  *	NULL on failure, pointer to DB on success.
     88  *
     89  */
     90 DB *
     91 __bt_open(fname, flags, mode, openinfo, dflags)
     92 	const char *fname;
     93 	int flags, mode, dflags;
     94 	const BTREEINFO *openinfo;
     95 {
     96 	struct stat sb;
     97 	BTMETA m;
     98 	BTREE *t;
     99 	BTREEINFO b;
    100 	DB *dbp;
    101 	pgno_t ncache;
    102 	ssize_t nr;
    103 	int machine_lorder;
    104 
    105 	t = NULL;
    106 
    107 	/*
    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
    112 	 * to change.
    113 	 */
    114 	machine_lorder = byteorder();
    115 	if (openinfo) {
    116 		b = *openinfo;
    117 
    118 		/* Flags: R_DUP. */
    119 		if (b.flags & ~(R_DUP))
    120 			goto einval;
    121 
    122 		/*
    123 		 * Page size must be indx_t aligned and >= MINPSIZE.  Default
    124 		 * page size is set farther on, based on the underlying file
    125 		 * transfer size.
    126 		 */
    127 		if (b.psize &&
    128 		    (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 ||
    129 		    b.psize & sizeof(indx_t) - 1))
    130 			goto einval;
    131 
    132 		/* Minimum number of keys per page; absolute minimum is 2. */
    133 		if (b.minkeypage) {
    134 			if (b.minkeypage < 2)
    135 				goto einval;
    136 		} else
    137 			b.minkeypage = DEFMINKEYPAGE;
    138 
    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;
    144 		}
    145 
    146 		if (b.lorder == 0)
    147 			b.lorder = machine_lorder;
    148 	} else {
    149 		b.compare = __bt_defcmp;
    150 		b.cachesize = 0;
    151 		b.flags = 0;
    152 		b.lorder = machine_lorder;
    153 		b.minkeypage = DEFMINKEYPAGE;
    154 		b.prefix = __bt_defpfx;
    155 		b.psize = 0;
    156 	}
    157 
    158 	/* Check for the ubiquitous PDP-11. */
    159 	if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN)
    160 		goto einval;
    161 
    162 	/* Allocate and initialize DB and BTREE structures. */
    163 	if ((t = (BTREE *)malloc(sizeof(BTREE))) == NULL)
    164 		goto err;
    165 	memset(t, 0, sizeof(BTREE));
    166 	t->bt_bcursor.pgno = P_INVALID;
    167 	t->bt_fd = -1;			/* Don't close unopened fd on error. */
    168 	t->bt_lorder = b.lorder;
    169 	t->bt_order = NOT;
    170 	t->bt_cmp = b.compare;
    171 	t->bt_pfx = b.prefix;
    172 	t->bt_rfd = -1;
    173 
    174 	if ((t->bt_dbp = dbp = (DB *)malloc(sizeof(DB))) == NULL)
    175 		goto err;
    176 	t->bt_flags = 0;
    177 	if (t->bt_lorder != machine_lorder)
    178 		SET(t, B_NEEDSWAP);
    179 
    180 	dbp->type = DB_BTREE;
    181 	dbp->internal = t;
    182 	dbp->close = __bt_close;
    183 	dbp->del = __bt_delete;
    184 	dbp->fd = __bt_fd;
    185 	dbp->get = __bt_get;
    186 	dbp->put = __bt_put;
    187 	dbp->seq = __bt_seq;
    188 	dbp->sync = __bt_sync;
    189 
    190 	/*
    191 	 * If no file name was supplied, this is an in-memory btree and we
    192 	 * open a backing temporary file.  Otherwise, it's a disk-based tree.
    193 	 */
    194 	if (fname) {
    195 		switch(flags & O_ACCMODE) {
    196 		case O_RDONLY:
    197 			SET(t, B_RDONLY);
    198 			break;
    199 		case O_RDWR:
    200 			break;
    201 		case O_WRONLY:
    202 		default:
    203 			goto einval;
    204 		}
    205 
    206 		if ((t->bt_fd = open(fname, flags, mode)) < 0)
    207 			goto err;
    208 
    209 	} else {
    210 		if ((flags & O_ACCMODE) != O_RDWR)
    211 			goto einval;
    212 		if ((t->bt_fd = tmp()) == -1)
    213 			goto err;
    214 		SET(t, B_INMEM);
    215 	}
    216 
    217 	if (fcntl(t->bt_fd, F_SETFD, 1) == -1)
    218 		goto err;
    219 
    220 	if (fstat(t->bt_fd, &sb))
    221 		goto err;
    222 	if (sb.st_size) {
    223 		if ((nr = read(t->bt_fd, &m, sizeof(BTMETA))) < 0)
    224 			goto err;
    225 		if (nr != sizeof(BTMETA))
    226 			goto eftype;
    227 
    228 		/*
    229 		 * Read in the meta-data.  This can change the notion of what
    230 		 * the lorder, page size and flags are, and, when the page size
    231 		 * changes, the cachesize value can change too.  If the user
    232 		 * specified the wrong byte order for an existing database, we
    233 		 * don't bother to return an error, we just clear the NEEDSWAP
    234 		 * bit.
    235 		 */
    236 		if (m.m_magic == BTREEMAGIC)
    237 			CLR(t, B_NEEDSWAP);
    238 		else {
    239 			SET(t, B_NEEDSWAP);
    240 			M_32_SWAP(m.m_magic);
    241 			M_32_SWAP(m.m_version);
    242 			M_32_SWAP(m.m_psize);
    243 			M_32_SWAP(m.m_free);
    244 			M_32_SWAP(m.m_nrecs);
    245 			M_32_SWAP(m.m_flags);
    246 		}
    247 		if (m.m_magic != BTREEMAGIC || m.m_version != BTREEVERSION)
    248 			goto eftype;
    249 		if (m.m_psize < MINPSIZE || m.m_psize > MAX_PAGE_OFFSET + 1 ||
    250 		    m.m_psize & sizeof(indx_t) - 1)
    251 			goto eftype;
    252 		if (m.m_flags & ~SAVEMETA)
    253 			goto eftype;
    254 		b.psize = m.m_psize;
    255 		t->bt_flags |= m.m_flags;
    256 		t->bt_free = m.m_free;
    257 		t->bt_nrecs = m.m_nrecs;
    258 	} else {
    259 		/*
    260 		 * Set the page size to the best value for I/O to this file.
    261 		 * Don't overflow the page offset type.
    262 		 */
    263 		if (b.psize == 0) {
    264 			b.psize = sb.st_blksize;
    265 			if (b.psize < MINPSIZE)
    266 				b.psize = MINPSIZE;
    267 			if (b.psize > MAX_PAGE_OFFSET + 1)
    268 				b.psize = MAX_PAGE_OFFSET + 1;
    269 		}
    270 
    271 		/* Set flag if duplicates permitted. */
    272 		if (!(b.flags & R_DUP))
    273 			SET(t, B_NODUPS);
    274 
    275 		t->bt_free = P_INVALID;
    276 		t->bt_nrecs = 0;
    277 		SET(t, B_METADIRTY);
    278 	}
    279 
    280 	t->bt_psize = b.psize;
    281 
    282 	/* Set the cache size; must be a multiple of the page size. */
    283 	if (b.cachesize && b.cachesize & b.psize - 1)
    284 		b.cachesize += (~b.cachesize & b.psize - 1) + 1;
    285 	if (b.cachesize < b.psize * MINCACHE)
    286 		b.cachesize = b.psize * MINCACHE;
    287 
    288 	/* Calculate number of pages to cache. */
    289 	ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize;
    290 
    291 	/*
    292 	 * The btree data structure requires that at least two keys can fit on
    293 	 * a page, but other than that there's no fixed requirement.  The user
    294 	 * specified a minimum number per page, and we translated that into the
    295 	 * number of bytes a key/data pair can use before being placed on an
    296 	 * overflow page.  This calculation includes the page header, the size
    297 	 * of the index referencing the leaf item and the size of the leaf item
    298 	 * structure.  Also, don't let the user specify a minkeypage such that
    299 	 * a key/data pair won't fit even if both key and data are on overflow
    300 	 * pages.
    301 	 */
    302 	t->bt_ovflsize = (t->bt_psize - BTDATAOFF) / b.minkeypage -
    303 	    (sizeof(indx_t) + NBLEAFDBT(0, 0));
    304 	if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t))
    305 		t->bt_ovflsize =
    306 		    NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t);
    307 
    308 	/* Initialize the buffer pool. */
    309 	if ((t->bt_mp =
    310 	    mpool_open(NULL, t->bt_fd, t->bt_psize, ncache)) == NULL)
    311 		goto err;
    312 	if (!ISSET(t, B_INMEM))
    313 		mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t);
    314 
    315 	/* Create a root page if new tree. */
    316 	if (nroot(t) == RET_ERROR)
    317 		goto err;
    318 
    319 	/* Global flags. */
    320 	if (dflags & DB_LOCK)
    321 		SET(t, B_DB_LOCK);
    322 	if (dflags & DB_SHMEM)
    323 		SET(t, B_DB_SHMEM);
    324 	if (dflags & DB_TXN)
    325 		SET(t, B_DB_TXN);
    326 
    327 	return (dbp);
    328 
    329 einval:	errno = EINVAL;
    330 	goto err;
    331 
    332 eftype:	errno = EFTYPE;
    333 	goto err;
    334 
    335 err:	if (t) {
    336 		if (t->bt_dbp)
    337 			free(t->bt_dbp);
    338 		if (t->bt_fd != -1)
    339 			(void)close(t->bt_fd);
    340 		free(t);
    341 	}
    342 	return (NULL);
    343 }
    344 
    345 /*
    346  * NROOT -- Create the root of a new tree.
    347  *
    348  * Parameters:
    349  *	t:	tree
    350  *
    351  * Returns:
    352  *	RET_ERROR, RET_SUCCESS
    353  */
    354 static int
    355 nroot(t)
    356 	BTREE *t;
    357 {
    358 	PAGE *meta, *root;
    359 	pgno_t npg;
    360 
    361 	if ((meta = mpool_get(t->bt_mp, 0, 0)) != NULL) {
    362 		mpool_put(t->bt_mp, meta, 0);
    363 		return (RET_SUCCESS);
    364 	}
    365 	if (errno != EINVAL)		/* It's OK to not exist. */
    366 		return (RET_ERROR);
    367 	errno = 0;
    368 
    369 	if ((meta = mpool_new(t->bt_mp, &npg)) == NULL)
    370 		return (RET_ERROR);
    371 
    372 	if ((root = mpool_new(t->bt_mp, &npg)) == NULL)
    373 		return (RET_ERROR);
    374 
    375 	if (npg != P_ROOT)
    376 		return (RET_ERROR);
    377 	root->pgno = npg;
    378 	root->prevpg = root->nextpg = P_INVALID;
    379 	root->lower = BTDATAOFF;
    380 	root->upper = t->bt_psize;
    381 	root->flags = P_BLEAF;
    382 	memset(meta, 0, t->bt_psize);
    383 	mpool_put(t->bt_mp, meta, MPOOL_DIRTY);
    384 	mpool_put(t->bt_mp, root, MPOOL_DIRTY);
    385 	return (RET_SUCCESS);
    386 }
    387 
    388 static int
    389 tmp()
    390 {
    391 	sigset_t set, oset;
    392 	int fd;
    393 	char *envtmp;
    394 	char path[MAXPATHLEN];
    395 
    396 	envtmp = getenv("TMPDIR");
    397 	(void)snprintf(path,
    398 	    sizeof(path), "%s/bt.XXXXXX", envtmp ? envtmp : "/tmp");
    399 
    400 	(void)sigfillset(&set);
    401 	(void)sigprocmask(SIG_BLOCK, &set, &oset);
    402 	if ((fd = mkstemp(path)) != -1)
    403 		(void)unlink(path);
    404 	(void)sigprocmask(SIG_SETMASK, &oset, NULL);
    405 	return(fd);
    406 }
    407 
    408 static int
    409 byteorder()
    410 {
    411 	u_int32_t x;
    412 	u_char *p;
    413 
    414 	x = 0x01020304;
    415 	p = (u_char *)&x;
    416 	switch (*p) {
    417 	case 1:
    418 		return (BIG_ENDIAN);
    419 	case 4:
    420 		return (LITTLE_ENDIAN);
    421 	default:
    422 		return (0);
    423 	}
    424 }
    425 
    426 int
    427 __bt_fd(dbp)
    428         const DB *dbp;
    429 {
    430 	BTREE *t;
    431 
    432 	t = dbp->internal;
    433 
    434 	/* Toss any page pinned across calls. */
    435 	if (t->bt_pinned != NULL) {
    436 		mpool_put(t->bt_mp, t->bt_pinned, 0);
    437 		t->bt_pinned = NULL;
    438 	}
    439 
    440 	/* In-memory database can't have a file descriptor. */
    441 	if (ISSET(t, B_INMEM)) {
    442 		errno = ENOENT;
    443 		return (-1);
    444 	}
    445 	return (t->bt_fd);
    446 }
    447