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
37 #if defined(LIBC_SCCS) && !defined(lint)
38 static char sccsid
[] = "@(#)bt_open.c 8.10 (Berkeley) 8/17/94";
39 #endif /* LIBC_SCCS and not lint */
42 * Implementation of btree access method for 4.4BSD.
44 * The design here was originally based on that of the btree access method
45 * used in the Postgres database system at UC Berkeley. This implementation
46 * is wholly independent of the Postgres code.
49 #include <sys/param.h>
69 static int byteorder
__P((void));
70 static int nroot
__P((BTREE
*));
71 static int tmp
__P((void));
74 * __BT_OPEN -- Open a btree.
76 * Creates and fills a DB struct, and calls the routine that actually
80 * fname: filename (NULL for in-memory trees)
81 * flags: open flag bits
82 * mode: open permission bits
83 * b: BTREEINFO pointer
86 * NULL on failure, pointer to DB on success.
90 __bt_open(fname
, flags
, mode
, openinfo
, dflags
)
92 int flags
, mode
, dflags
;
93 const BTREEINFO
*openinfo
;
107 * Intention is to make sure all of the user's selections are okay
108 * here and then use them without checking. Can't be complete, since
109 * we don't know the right page size, lorder or flags until the backing
110 * file is opened. Also, the file's page size can cause the cachesize
113 machine_lorder
= byteorder();
118 if (b
.flags
& ~(R_DUP
))
122 * Page size must be indx_t aligned and >= MINPSIZE. Default
123 * page size is set farther on, based on the underlying file
127 (b
.psize
< MINPSIZE
|| b
.psize
> MAX_PAGE_OFFSET
+ 1 ||
128 b
.psize
& (sizeof(indx_t
) - 1)))
131 /* Minimum number of keys per page; absolute minimum is 2. */
133 if (b
.minkeypage
< 2)
136 b
.minkeypage
= DEFMINKEYPAGE
;
138 /* If no comparison, use default comparison and prefix. */
139 if (b
.compare
== NULL
) {
140 b
.compare
= __bt_defcmp
;
141 if (b
.prefix
== NULL
)
142 b
.prefix
= __bt_defpfx
;
146 b
.lorder
= machine_lorder
;
148 b
.compare
= __bt_defcmp
;
151 b
.lorder
= machine_lorder
;
152 b
.minkeypage
= DEFMINKEYPAGE
;
153 b
.prefix
= __bt_defpfx
;
157 /* Check for the ubiquitous PDP-11. */
158 if (b
.lorder
!= BIG_ENDIAN
&& b
.lorder
!= LITTLE_ENDIAN
)
161 /* Allocate and initialize DB and BTREE structures. */
162 if ((t
= (BTREE
*)malloc(sizeof(BTREE
))) == NULL
)
164 memset(t
, 0, sizeof(BTREE
));
165 t
->bt_fd
= -1; /* Don't close unopened fd on error. */
166 t
->bt_lorder
= b
.lorder
;
168 t
->bt_cmp
= b
.compare
;
169 t
->bt_pfx
= b
.prefix
;
172 if ((t
->bt_dbp
= dbp
= (DB
*)malloc(sizeof(DB
))) == NULL
)
174 memset(t
->bt_dbp
, 0, sizeof(DB
));
175 if (t
->bt_lorder
!= machine_lorder
)
176 F_SET(t
, B_NEEDSWAP
);
178 dbp
->type
= DB_BTREE
;
180 dbp
->close
= __bt_close
;
181 dbp
->del
= __bt_delete
;
186 dbp
->sync
= __bt_sync
;
189 * If no file name was supplied, this is an in-memory btree and we
190 * open a backing temporary file. Otherwise, it's a disk-based tree.
193 switch (flags
& O_ACCMODE
) {
204 if ((t
->bt_fd
= open(fname
, flags
, mode
)) < 0)
208 if ((flags
& O_ACCMODE
) != O_RDWR
)
210 if ((t
->bt_fd
= tmp()) == -1)
215 if (fcntl(t
->bt_fd
, F_SETFD
, 1) == -1)
218 if (fstat(t
->bt_fd
, &sb
))
221 if ((nr
= read(t
->bt_fd
, &m
, sizeof(BTMETA
))) < 0)
223 if (nr
!= sizeof(BTMETA
))
227 * Read in the meta-data. This can change the notion of what
228 * the lorder, page size and flags are, and, when the page size
229 * changes, the cachesize value can change too. If the user
230 * specified the wrong byte order for an existing database, we
231 * don't bother to return an error, we just clear the NEEDSWAP
234 if (m
.magic
== BTREEMAGIC
)
235 F_CLR(t
, B_NEEDSWAP
);
237 F_SET(t
, B_NEEDSWAP
);
239 M_32_SWAP(m
.version
);
245 if (m
.magic
!= BTREEMAGIC
|| m
.version
!= BTREEVERSION
)
247 if (m
.psize
< MINPSIZE
|| m
.psize
> MAX_PAGE_OFFSET
+ 1 ||
248 m
.psize
& (sizeof(indx_t
) - 1))
250 if (m
.flags
& ~SAVEMETA
)
255 t
->bt_nrecs
= m
.nrecs
;
258 * Set the page size to the best value for I/O to this file.
259 * Don't overflow the page offset type.
262 #ifdef _STATBUF_ST_BLKSIZE
263 b
.psize
= sb
.st_blksize
;
265 if (b
.psize
< MINPSIZE
)
267 if (b
.psize
> MAX_PAGE_OFFSET
+ 1)
268 b
.psize
= MAX_PAGE_OFFSET
+ 1;
271 /* Set flag if duplicates permitted. */
272 if (!(b
.flags
& R_DUP
))
275 t
->bt_free
= P_INVALID
;
277 F_SET(t
, B_METADIRTY
);
280 t
->bt_psize
= b
.psize
;
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
;
288 /* Calculate number of pages to cache. */
289 ncache
= (b
.cachesize
+ t
->bt_psize
- 1) / t
->bt_psize
;
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
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
))
306 NBLEAFDBT(NOVFLSIZE
, NOVFLSIZE
) + sizeof(indx_t
);
308 /* Initialize the buffer pool. */
310 mpool_open(NULL
, t
->bt_fd
, t
->bt_psize
, ncache
)) == NULL
)
312 if (!F_ISSET(t
, B_INMEM
))
313 mpool_filter(t
->bt_mp
, __bt_pgin
, __bt_pgout
, t
);
315 /* Create a root page if new tree. */
316 if (nroot(t
) == RET_ERROR
)
320 if (dflags
& DB_LOCK
)
322 if (dflags
& DB_SHMEM
)
323 F_SET(t
, B_DB_SHMEM
);
329 einval
: errno
= EINVAL
;
332 eftype
: errno
= EFTYPE
;
339 (void)close(t
->bt_fd
);
346 * NROOT -- Create the root of a new tree.
352 * RET_ERROR, RET_SUCCESS
361 if ((meta
= mpool_get(t
->bt_mp
, 0, 0)) != NULL
) {
362 mpool_put(t
->bt_mp
, meta
, 0);
363 return (RET_SUCCESS
);
365 if (errno
!= EINVAL
) /* It's OK to not exist. */
369 if ((meta
= mpool_new(t
->bt_mp
, &npg
)) == NULL
)
372 if ((root
= mpool_new(t
->bt_mp
, &npg
)) == NULL
)
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
);
395 static const char fmt
[] = "%s/bt.XXXXXX";
398 envtmp
= getenv("TMPDIR");
401 n
= strlen (envtmp
) + sizeof fmt
;
403 path
= __builtin_alloca(n
);
407 (void)snprintf(path
, n
, fmt
, envtmp
? envtmp
: "/tmp");
409 (void)sigfillset(&set
);
410 (void)sigprocmask(SIG_BLOCK
, &set
, &oset
);
411 if ((fd
= mkstemp(path
)) != -1)
413 (void)sigprocmask(SIG_SETMASK
, &oset
, NULL
);
432 return (LITTLE_ENDIAN
);
446 /* Toss any page pinned across calls. */
447 if (t
->bt_pinned
!= NULL
) {
448 mpool_put(t
->bt_mp
, t
->bt_pinned
, 0);
452 /* In-memory database can't have a file descriptor. */
453 if (F_ISSET(t
, B_INMEM
)) {