1/*-
2 * Copyright (c) 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Mike Olson.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
23 *
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
34 * SUCH DAMAGE.
35 */
36
37#if defined(LIBC_SCCS) && !defined(lint)
38static char sccsid[] = "@(#)bt_open.c 8.10 (Berkeley) 8/17/94";
39#endif /* LIBC_SCCS and not lint */
40
41/*
42 * Implementation of btree access method for 4.4BSD.
43 *
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.
47 */
48
49#include <sys/param.h>
50
51#include <errno.h>
52#include <limits.h>
53#include <stdio.h>
54#include <stdlib.h>
55#include <string.h>
56#include <stdint.h>
57
58#include <db.h>
59#include "btree.h"
60
61#ifdef DEBUG
62#undef MINPSIZE
63#define MINPSIZE 128
64#endif
65
66static int byteorder __P((void));
67static int nroot __P((BTREE *));
68
69#ifdef BTREE_POSIX
70/* Default vmethods and vtable to work with POSIX fd's. */
71static ssize_t bt_read(void *fd, void *buf, size_t size)
72{
73//printf("read: %p(%lx)\n", buf, size);
74 return read((int)(uintptr_t)fd, buf, size);
75}
76
77static ssize_t bt_write(void *fd, const void *buf, size_t size)
78{
79//printf("write: %p(%lx)\n", buf, size);
80 return write((int)(uintptr_t)fd, buf, size);
81}
82
83static off_t bt_lseek(void *fd, off_t offset, int whence)
84{
85//printf("lseek: %lx(%d)\n", offset, whence);
86 return lseek((int)(uintptr_t)fd, offset, whence);
87}
88
89static int bt_fsync(void *fd)
90{
91 return fsync((int)(uintptr_t)fd);
92}
93
94static FILEVTABLE bt_fd_fvtable = {
95 bt_read,
96 bt_write,
97 bt_lseek,
98 bt_fsync
99};
100#endif
101
102/*
103 * __BT_OPEN -- Open a btree.
104 *
105 * Creates and fills a DB struct, and calls the routine that actually
106 * opens the btree.
107 *
108 * Parameters:
109 * fname: filename (NULL for in-memory trees)
110 * b: BTREEINFO pointer
111 *
112 * Returns:
113 * NULL on failure, pointer to DB on success.
114 *
115 */
116DB *
117__bt_open(file, vtable, openinfo, dflags)
118 virt_fd_t file;
119 const FILEVTABLE *vtable;
120 int dflags;
121 const BTREEINFO *openinfo;
122{
123 BTMETA m;
124 BTREE *t;
125 BTREEINFO b;
126 DB *dbp;
127 pgno_t ncache;
128 ssize_t nr;
129 int machine_lorder;
130
131 t = NULL;
132#ifdef BTREE_POSIX
133 if (vtable == NULL)
134 vtable = &bt_fd_fvtable;
135#endif
136
137 /*
138 * Intention is to make sure all of the user's selections are okay
139 * here and then use them without checking. Can't be complete, since
140 * we don't know the right page size, lorder or flags until the backing
141 * file is opened. Also, the file's page size can cause the cachesize
142 * to change.
143 */
144 machine_lorder = byteorder();
145 if (openinfo) {
146 b = *openinfo;
147
148 /* Flags: R_DUP. */
149 if (b.flags & ~(R_DUP))
150 goto einval;
151
152 /*
153 * Page size must be indx_t aligned and >= MINPSIZE. Default
154 * page size is set farther on, based on the underlying file
155 * transfer size.
156 */
157 if (b.psize &&
158 (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 ||
159 b.psize & (sizeof(indx_t) - 1)))
160 goto einval;
161
162 /* Minimum number of keys per page; absolute minimum is 2. */
163 if (b.minkeypage) {
164 if (b.minkeypage < 2)
165 goto einval;
166 } else
167 b.minkeypage = DEFMINKEYPAGE;
168
169 /* If no comparison, use default comparison and prefix. */
170 if (b.compare == NULL) {
171 b.compare = __bt_defcmp;
172 if (b.prefix == NULL)
173 b.prefix = __bt_defpfx;
174 }
175
176 if (b.lorder == 0)
177 b.lorder = machine_lorder;
178 } else {
179 b.compare = __bt_defcmp;
180 b.cachesize = 0;
181 b.flags = 0;
182 b.lorder = machine_lorder;
183 b.minkeypage = DEFMINKEYPAGE;
184 b.prefix = __bt_defpfx;
185 b.psize = 0;
186 }
187
188 /* Check for the ubiquitous PDP-11. */
189 if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN)
190 goto einval;
191
192 /* Allocate and initialize DB and BTREE structures. */
193 if ((t = (BTREE *)malloc(sizeof(BTREE))) == NULL)
194 goto err;
195 memset(t, 0, sizeof(BTREE));
196 t->bt_fd = file;
197 t->bt_lorder = b.lorder;
198 t->bt_order = NOT;
199 t->bt_cmp = b.compare;
200 t->bt_pfx = b.prefix;
201 t->bt_rfd = -1;
202
203 if ((t->bt_dbp = dbp = (DB *)malloc(sizeof(DB))) == NULL)
204 goto err;
205 memset(t->bt_dbp, 0, sizeof(DB));
206 if (t->bt_lorder != machine_lorder)
207 F_SET(t, B_NEEDSWAP);
208
209 dbp->type = DB_BTREE;
210 dbp->internal = t;
211 dbp->close = __bt_close;
212 dbp->del = __bt_delete;
213 dbp->fd = __bt_fd;
214 dbp->get = __bt_get;
215 dbp->put = __bt_put;
216 dbp->seq = __bt_seq;
217 dbp->sync = __bt_sync;
218
219 if ((nr = vtable->read(t->bt_fd, &m, sizeof(BTMETA))) < 0)
220 goto err;
221 if (nr != 0) {
222 if (nr != sizeof(BTMETA))
223 goto eftype;
224
225 /*
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
231 * bit.
232 */
233 if (m.magic == BTREEMAGIC)
234 F_CLR(t, B_NEEDSWAP);
235 else {
236 F_SET(t, B_NEEDSWAP);
237 M_32_SWAP(m.magic);
238 M_32_SWAP(m.version);
239 M_32_SWAP(m.psize);
240 M_32_SWAP(m.free);
241 M_32_SWAP(m.nrecs);
242 M_32_SWAP(m.flags);
243 }
244 if (m.magic != BTREEMAGIC || m.version != BTREEVERSION)
245 goto eftype;
246 if (m.psize < MINPSIZE || m.psize > MAX_PAGE_OFFSET + 1 ||
247 m.psize & (sizeof(indx_t) - 1))
248 goto eftype;
249 if (m.flags & ~SAVEMETA)
250 goto eftype;
251 b.psize = m.psize;
252 F_SET(t, m.flags);
253 t->bt_free = m.free;
254 t->bt_nrecs = m.nrecs;
255 } else {
256 /*
257 * Set the page size to the best value for I/O to this file.
258 * Don't overflow the page offset type.
259 */
260 if (b.psize == 0) {
261 b.psize = DEFPSIZE;
262 }
263
264 /* Set flag if duplicates permitted. */
265 if (!(b.flags & R_DUP))
266 F_SET(t, B_NODUPS);
267
268 t->bt_free = P_INVALID;
269 t->bt_nrecs = 0;
270 F_SET(t, B_METADIRTY);
271 }
272
273 t->bt_psize = b.psize;
274
275 /* Set the cache size; must be a multiple of the page size. */
276 if (b.cachesize && b.cachesize & (b.psize - 1))
277 b.cachesize += (~b.cachesize & (b.psize - 1)) + 1;
278 if (b.cachesize < b.psize * MINCACHE)
279 b.cachesize = b.psize * MINCACHE;
280
281 /* Calculate number of pages to cache. */
282 ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize;
283
284 /*
285 * The btree data structure requires that at least two keys can fit on
286 * a page, but other than that there's no fixed requirement. The user
287 * specified a minimum number per page, and we translated that into the
288 * number of bytes a key/data pair can use before being placed on an
289 * overflow page. This calculation includes the page header, the size
290 * of the index referencing the leaf item and the size of the leaf item
291 * structure. Also, don't let the user specify a minkeypage such that
292 * a key/data pair won't fit even if both key and data are on overflow
293 * pages.
294 */
295 t->bt_ovflsize = (t->bt_psize - BTDATAOFF) / b.minkeypage -
296 (sizeof(indx_t) + NBLEAFDBT(0, 0));
297 if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t))
298 t->bt_ovflsize =
299 NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t);
300
301 /* Initialize the buffer pool. */
302 if ((t->bt_mp =
303 mpool_open(NULL, t->bt_fd, vtable, t->bt_psize, ncache)) == NULL)
304 goto err;
305 if (!F_ISSET(t, B_INMEM))
306 mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t);
307
308 /* Create a root page if new tree. */
309 if (nroot(t) == RET_ERROR)
310 goto err;
311
312 /* Global flags. */
313 if (dflags & DB_LOCK)
314 F_SET(t, B_DB_LOCK);
315 if (dflags & DB_SHMEM)
316 F_SET(t, B_DB_SHMEM);
317 if (dflags & DB_TXN)
318 F_SET(t, B_DB_TXN);
319
320 return (dbp);
321
322einval: errno = EINVAL;
323 goto err;
324
325eftype: errno = EFTYPE;
326 goto err;
327
328err: if (t) {
329 if (t->bt_dbp)
330 free(t->bt_dbp);
331 free(t);
332 }
333 return (NULL);
334}
335
336/*
337 * NROOT -- Create the root of a new tree.
338 *
339 * Parameters:
340 * t: tree
341 *
342 * Returns:
343 * RET_ERROR, RET_SUCCESS
344 */
345static int
346nroot(t)
347 BTREE *t;
348{
349 PAGE *meta, *root;
350 pgno_t npg;
351
352 if ((meta = mpool_get(t->bt_mp, 0, 0)) != NULL) {
353 mpool_put(t->bt_mp, meta, 0);
354 return (RET_SUCCESS);
355 }
356 if (errno != EINVAL) /* It's OK to not exist. */
357 return (RET_ERROR);
358 errno = 0;
359
360 if ((meta = mpool_new(t->bt_mp, &npg)) == NULL)
361 return (RET_ERROR);
362
363 if ((root = mpool_new(t->bt_mp, &npg)) == NULL)
364 return (RET_ERROR);
365
366 if (npg != P_ROOT)
367 return (RET_ERROR);
368 root->pgno = npg;
369 root->prevpg = root->nextpg = P_INVALID;
370 root->lower = BTDATAOFF;
371 root->upper = t->bt_psize;
372 root->flags = P_BLEAF;
373 memset(meta, 0, t->bt_psize);
374 mpool_put(t->bt_mp, meta, MPOOL_DIRTY);
375 mpool_put(t->bt_mp, root, MPOOL_DIRTY);
376 return (RET_SUCCESS);
377}
378
379static int
380byteorder()
381{
382 u_int32_t x;
383 u_char *p;
384
385 x = 0x01020304;
386 p = (u_char *)&x;
387 switch (*p) {
388 case 1:
389 return (BIG_ENDIAN);
390 case 4:
391 return (LITTLE_ENDIAN);
392 default:
393 return (0);
394 }
395}
396
397int
398__bt_fd(dbp)
399 const DB *dbp;
400{
401 BTREE *t;
402
403 t = dbp->internal;
404
405 /* Toss any page pinned across calls. */
406 if (t->bt_pinned != NULL) {
407 mpool_put(t->bt_mp, t->bt_pinned, 0);
408 t->bt_pinned = NULL;
409 }
410
411 return -1;
412}
413