1 /* $NetBSD: bt_open.c,v 1.30 2025/12/16 12:39:01 nia 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. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 #if HAVE_NBTOOL_CONFIG_H 36 #include "nbtool_config.h" 37 #endif 38 39 #include <sys/cdefs.h> 40 __RCSID("$NetBSD: bt_open.c,v 1.30 2025/12/16 12:39:01 nia Exp $"); 41 42 /* 43 * Implementation of btree access method for 4.4BSD. 44 * 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. 48 */ 49 50 #include "namespace.h" 51 #include <sys/stat.h> 52 #include <sys/endian.h> 53 54 #include <assert.h> 55 #include <errno.h> 56 #include <fcntl.h> 57 #include <limits.h> 58 #include <signal.h> 59 #include <stdio.h> 60 #include <stdlib.h> 61 #include <string.h> 62 #include <unistd.h> 63 #include <paths.h> 64 65 #include <db.h> 66 #include "btree.h" 67 68 #ifdef DEBUG 69 #undef MINPSIZE 70 #define MINPSIZE 128 71 #endif 72 73 static int byteorder(void); 74 static int nroot(BTREE *); 75 76 /* 77 * __BT_OPEN -- Open a btree. 78 * 79 * Creates and fills a DB struct, and calls the routine that actually 80 * opens the btree. 81 * 82 * Parameters: 83 * fname: filename (NULL for in-memory trees) 84 * flags: open flag bits 85 * mode: open permission bits 86 * b: BTREEINFO pointer 87 * 88 * Returns: 89 * NULL on failure, pointer to DB on success. 90 * 91 */ 92 DB * 93 __bt_open(const char *fname, int flags, mode_t mode, const BTREEINFO *openinfo, 94 int dflags) 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 size_t temp; 104 int machine_lorder; 105 106 t = NULL; 107 108 /* 109 * Intention is to make sure all of the user's selections are okay 110 * here and then use them without checking. Can't be complete, since 111 * we don't know the right page size, lorder or flags until the backing 112 * file is opened. Also, the file's page size can cause the cachesize 113 * to change. 114 */ 115 machine_lorder = byteorder(); 116 if (openinfo) { 117 b = *openinfo; 118 119 /* Flags: R_DUP. */ 120 if (b.flags & ~(R_DUP)) 121 goto einval; 122 123 /* 124 * Page size must be indx_t aligned and >= MINPSIZE. Default 125 * page size is set farther on, based on the underlying file 126 * transfer size. 127 */ 128 if (b.psize && 129 (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 || 130 b.psize & (sizeof(indx_t) - 1))) 131 goto einval; 132 133 /* Minimum number of keys per page; absolute minimum is 2. */ 134 if (b.minkeypage) { 135 if (b.minkeypage < 2) 136 goto einval; 137 } else 138 b.minkeypage = DEFMINKEYPAGE; 139 140 /* If no comparison, use default comparison and prefix. */ 141 if (b.compare == NULL) { 142 b.compare = __bt_defcmp; 143 if (b.prefix == NULL) 144 b.prefix = __bt_defpfx; 145 } 146 147 if (b.lorder == 0) 148 b.lorder = machine_lorder; 149 } else { 150 b.compare = __bt_defcmp; 151 b.cachesize = 0; 152 b.flags = 0; 153 b.lorder = machine_lorder; 154 b.minkeypage = DEFMINKEYPAGE; 155 b.prefix = __bt_defpfx; 156 b.psize = 0; 157 } 158 159 /* Check for the ubiquitous PDP-11. */ 160 if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN) 161 goto einval; 162 163 /* Allocate and initialize DB and BTREE structures. */ 164 if ((t = malloc(sizeof(*t))) == NULL) 165 goto err; 166 memset(t, 0, sizeof(BTREE)); 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 = malloc(sizeof(*dbp))) == NULL) 175 goto err; 176 memset(t->bt_dbp, 0, sizeof(DB)); 177 if (t->bt_lorder != machine_lorder) 178 F_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 F_SET(t, B_RDONLY); 198 break; 199 case O_RDWR: 200 break; 201 case O_WRONLY: 202 default: 203 goto einval; 204 } 205 if ((t->bt_fd = __dbopen(fname, flags, mode, &sb)) == -1) 206 goto err; 207 } else { 208 if ((flags & O_ACCMODE) != O_RDWR) 209 goto einval; 210 if ((t->bt_fd = __dbtemp("bt.", &sb)) == -1) 211 goto err; 212 F_SET(t, B_INMEM); 213 } 214 215 216 if (sb.st_size) { 217 if ((nr = read(t->bt_fd, &m, sizeof(BTMETA))) < 0) 218 goto err; 219 if (nr != sizeof(BTMETA)) 220 goto eftype; 221 222 /* 223 * Read in the meta-data. This can change the notion of what 224 * the lorder, page size and flags are, and, when the page size 225 * changes, the cachesize value can change too. If the user 226 * specified the wrong byte order for an existing database, we 227 * don't bother to return an error, we just clear the NEEDSWAP 228 * bit. 229 */ 230 if (m.magic == BTREEMAGIC) 231 F_CLR(t, B_NEEDSWAP); 232 else { 233 F_SET(t, B_NEEDSWAP); 234 M_32_SWAP(m.magic); 235 M_32_SWAP(m.version); 236 M_32_SWAP(m.psize); 237 M_32_SWAP(m.free); 238 M_32_SWAP(m.nrecs); 239 M_32_SWAP(m.flags); 240 } 241 if (m.magic != BTREEMAGIC || m.version != BTREEVERSION) 242 goto eftype; 243 if (m.psize < MINPSIZE || m.psize > MAX_PAGE_OFFSET + 1 || 244 m.psize & (sizeof(indx_t) - 1)) 245 goto eftype; 246 if (m.flags & ~SAVEMETA) 247 goto eftype; 248 b.psize = m.psize; 249 F_SET(t, m.flags); 250 t->bt_free = m.free; 251 t->bt_nrecs = m.nrecs; 252 } else { 253 /* 254 * Set the page size to the best value for I/O to this file. 255 * Don't overflow the page offset type. 256 */ 257 if (b.psize == 0) { 258 b.psize = sb.st_blksize; 259 if (b.psize < MINPSIZE) 260 b.psize = MINPSIZE; 261 if (b.psize > MAX_PAGE_OFFSET + 1) 262 b.psize = MAX_PAGE_OFFSET + 1; 263 } 264 265 /* Set flag if duplicates permitted. */ 266 if (!(b.flags & R_DUP)) 267 F_SET(t, B_NODUPS); 268 269 t->bt_free = P_INVALID; 270 t->bt_nrecs = 0; 271 F_SET(t, B_METADIRTY); 272 } 273 274 t->bt_psize = b.psize; 275 276 /* Set the cache size; must be a multiple of the page size. */ 277 if (b.cachesize && b.cachesize & (b.psize - 1)) 278 b.cachesize += (~b.cachesize & (b.psize - 1)) + 1; 279 if (b.cachesize < b.psize * MINCACHE) 280 b.cachesize = b.psize * MINCACHE; 281 282 /* Calculate number of pages to cache. */ 283 ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize; 284 285 /* 286 * The btree data structure requires that at least two keys can fit on 287 * a page, but other than that there's no fixed requirement. The user 288 * specified a minimum number per page, and we translated that into the 289 * number of bytes a key/data pair can use before being placed on an 290 * overflow page. This calculation includes the page header, the size 291 * of the index referencing the leaf item and the size of the leaf item 292 * structure. Also, don't let the user specify a minkeypage such that 293 * a key/data pair won't fit even if both key and data are on overflow 294 * pages. 295 */ 296 temp = (t->bt_psize - BTDATAOFF) / b.minkeypage - 297 (sizeof(indx_t) + NBLEAFDBT(0, 0)); 298 _DBFIT(temp, indx_t); 299 t->bt_ovflsize = (indx_t)temp; 300 if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t)) { 301 size_t l = NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t); 302 _DBFIT(l, indx_t); 303 t->bt_ovflsize = (indx_t)l; 304 } 305 306 /* Initialize the buffer pool. */ 307 if ((t->bt_mp = 308 mpool_open(NULL, t->bt_fd, t->bt_psize, ncache)) == NULL) 309 goto err; 310 if (!F_ISSET(t, B_INMEM)) 311 mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t); 312 313 /* Create a root page if new tree. */ 314 if (nroot(t) == RET_ERROR) 315 goto err; 316 317 /* Global flags. */ 318 if (dflags & DB_LOCK) 319 F_SET(t, B_DB_LOCK); 320 if (dflags & DB_SHMEM) 321 F_SET(t, B_DB_SHMEM); 322 if (dflags & DB_TXN) 323 F_SET(t, B_DB_TXN); 324 325 return (dbp); 326 327 einval: errno = EINVAL; 328 goto err; 329 330 eftype: errno = EFTYPE; 331 goto err; 332 333 err: if (t) { 334 if (t->bt_dbp) 335 free(t->bt_dbp); 336 if (t->bt_fd != -1) 337 (void)close(t->bt_fd); 338 free(t); 339 } 340 return (NULL); 341 } 342 343 /* 344 * NROOT -- Create the root of a new tree. 345 * 346 * Parameters: 347 * t: tree 348 * 349 * Returns: 350 * RET_ERROR, RET_SUCCESS 351 */ 352 static int 353 nroot(BTREE *t) 354 { 355 PAGE *meta, *root; 356 pgno_t npg; 357 358 if ((root = mpool_get(t->bt_mp, 1, 0)) != NULL) { 359 if (root->lower == 0 && 360 root->pgno == 0 && 361 root->linp[0] == 0) { 362 mpool_delete(t->bt_mp, root); 363 errno = EINVAL; 364 } else { 365 mpool_put(t->bt_mp, root, 0); 366 return RET_SUCCESS; 367 } 368 } 369 if (errno != EINVAL) /* It's OK to not exist. */ 370 return (RET_ERROR); 371 errno = 0; 372 373 if ((meta = mpool_newf(t->bt_mp, &npg, MPOOL_PAGE_NEXT)) == NULL) 374 return (RET_ERROR); 375 376 if ((root = mpool_newf(t->bt_mp, &npg, MPOOL_PAGE_NEXT)) == NULL) 377 return (RET_ERROR); 378 379 if (npg != P_ROOT) 380 return (RET_ERROR); 381 root->pgno = npg; 382 root->prevpg = root->nextpg = P_INVALID; 383 root->lower = BTDATAOFF; 384 root->upper = t->bt_psize; 385 root->flags = P_BLEAF; 386 memset(meta, 0, t->bt_psize); 387 mpool_put(t->bt_mp, meta, MPOOL_DIRTY); 388 mpool_put(t->bt_mp, root, MPOOL_DIRTY); 389 return (RET_SUCCESS); 390 } 391 392 static int 393 byteorder(void) 394 { 395 uint32_t x; 396 uint8_t *p; 397 398 x = 0x01020304; 399 p = (uint8_t *)(void *)&x; 400 switch (*p) { 401 case 1: 402 return (BIG_ENDIAN); 403 case 4: 404 return (LITTLE_ENDIAN); 405 default: 406 return (0); 407 } 408 } 409 410 int 411 __bt_fd(const DB *dbp) 412 { 413 BTREE *t; 414 415 t = dbp->internal; 416 417 /* Toss any page pinned across calls. */ 418 if (t->bt_pinned != NULL) { 419 mpool_put(t->bt_mp, t->bt_pinned, 0); 420 t->bt_pinned = NULL; 421 } 422 423 /* In-memory database can't have a file descriptor. */ 424 if (F_ISSET(t, B_INMEM)) { 425 errno = ENOENT; 426 return (-1); 427 } 428 return (t->bt_fd); 429 } 430
Indexes created Tue Feb 24 19:07:35 UTC 2026