1/* $OpenBSD: bt_open.c,v 1.19 2015/12/28 22:08:18 mmcc 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/* 36 * Implementation of btree access method for 4.4BSD. 37 * 38 * The design here was originally based on that of the btree access method 39 * used in the Postgres database system at UC Berkeley. This implementation 40 * is wholly independent of the Postgres code. 41 */ 42 43#include <sys/stat.h> 44 45#include <errno.h> 46#include <fcntl.h> 47#include <limits.h> 48#include <signal.h> 49#include <stdio.h> 50#include <stdlib.h> 51#include <string.h> 52#include <unistd.h> 53 54#include <db.h> 55#include "btree.h" 56 57#ifdef DEBUG 58#undef MINPSIZE 59#define MINPSIZE 128 60#endif 61 62static int byteorder(void); 63static int nroot(BTREE *); 64static int tmp(void); 65 66/* 67 * __BT_OPEN -- Open a btree. 68 * 69 * Creates and fills a DB struct, and calls the routine that actually 70 * opens the btree. 71 * 72 * Parameters: 73 * fname: filename (NULL for in-memory trees) 74 * flags: open flag bits 75 * mode: open permission bits 76 * b: BTREEINFO pointer 77 * 78 * Returns: 79 * NULL on failure, pointer to DB on success. 80 * 81 */ 82DB * 83__bt_open(const char *fname, int flags, int mode, const BTREEINFO *openinfo, 84 int dflags) 85{ 86 struct stat sb; 87 BTMETA m; 88 BTREE *t; 89 BTREEINFO b; 90 DB *dbp; 91 pgno_t ncache; 92 ssize_t nr; 93 int machine_lorder, saved_errno; 94 95 t = NULL; 96 97 /* 98 * Intention is to make sure all of the user's selections are okay 99 * here and then use them without checking. Can't be complete, since 100 * we don't know the right page size, lorder or flags until the backing 101 * file is opened. Also, the file's page size can cause the cachesize 102 * to change. 103 */ 104 machine_lorder = byteorder(); 105 if (openinfo) { 106 b = *openinfo; 107 108 /* Flags: R_DUP. */ 109 if (b.flags & ~(R_DUP)) 110 goto einval; 111 112 /* 113 * Page size must be indx_t aligned and >= MINPSIZE. Default 114 * page size is set farther on, based on the underlying file 115 * transfer size. 116 */ 117 if (b.psize && 118 (b.psize < MINPSIZE || b.psize > MAX_PAGE_OFFSET + 1 || 119 b.psize & (sizeof(indx_t) - 1))) 120 goto einval; 121 122 /* Minimum number of keys per page; absolute minimum is 2. */ 123 if (b.minkeypage) { 124 if (b.minkeypage < 2) 125 goto einval; 126 } else 127 b.minkeypage = DEFMINKEYPAGE; 128 129 /* If no comparison, use default comparison and prefix. */ 130 if (b.compare == NULL) { 131 b.compare = __bt_defcmp; 132 if (b.prefix == NULL) 133 b.prefix = __bt_defpfx; 134 } 135 136 if (b.lorder == 0) 137 b.lorder = machine_lorder; 138 } else { 139 b.compare = __bt_defcmp; 140 b.cachesize = 0; 141 b.flags = 0; 142 b.lorder = machine_lorder; 143 b.minkeypage = DEFMINKEYPAGE; 144 b.prefix = __bt_defpfx; 145 b.psize = 0; 146 } 147 148 /* Check for the ubiquitous PDP-11. */ 149 if (b.lorder != BIG_ENDIAN && b.lorder != LITTLE_ENDIAN) 150 goto einval; 151 152 /* Allocate and initialize DB and BTREE structures. */ 153 if ((t = calloc(1, sizeof(BTREE))) == NULL) 154 goto err; 155 t->bt_fd = -1; /* Don't close unopened fd on error. */ 156 t->bt_lorder = b.lorder; 157 t->bt_order = NOT; 158 t->bt_cmp = b.compare; 159 t->bt_pfx = b.prefix; 160 t->bt_rfd = -1; 161 162 if ((t->bt_dbp = dbp = calloc(1, sizeof(DB))) == NULL) 163 goto err; 164 if (t->bt_lorder != machine_lorder) 165 F_SET(t, B_NEEDSWAP); 166 167 dbp->type = DB_BTREE; 168 dbp->internal = t; 169 dbp->close = __bt_close; 170 dbp->del = __bt_delete; 171 dbp->fd = __bt_fd; 172 dbp->get = __bt_get; 173 dbp->put = __bt_put; 174 dbp->seq = __bt_seq; 175 dbp->sync = __bt_sync; 176 177 /* 178 * If no file name was supplied, this is an in-memory btree and we 179 * open a backing temporary file. Otherwise, it's a disk-based tree. 180 */ 181 if (fname) { 182 switch (flags & O_ACCMODE) { 183 case O_RDONLY: 184 F_SET(t, B_RDONLY); 185 break; 186 case O_RDWR: 187 break; 188 case O_WRONLY: 189 default: 190 goto einval; 191 } 192 193 if ((t->bt_fd = open(fname, flags | O_CLOEXEC, mode)) < 0) 194 goto err; 195 196 } else { 197 if ((flags & O_ACCMODE) != O_RDWR) 198 goto einval; 199 if ((t->bt_fd = tmp()) == -1) 200 goto err; 201 F_SET(t, B_INMEM); 202 } 203 204 if (fstat(t->bt_fd, &sb)) 205 goto err; 206 if (sb.st_size) { 207 if ((nr = read(t->bt_fd, &m, sizeof(BTMETA))) < 0) 208 goto err; 209 if (nr != sizeof(BTMETA)) 210 goto eftype; 211 212 /* 213 * Read in the meta-data. This can change the notion of what 214 * the lorder, page size and flags are, and, when the page size 215 * changes, the cachesize value can change too. If the user 216 * specified the wrong byte order for an existing database, we 217 * don't bother to return an error, we just clear the NEEDSWAP 218 * bit. 219 */ 220 if (m.magic == BTREEMAGIC) 221 F_CLR(t, B_NEEDSWAP); 222 else { 223 F_SET(t, B_NEEDSWAP); 224 M_32_SWAP(m.magic); 225 M_32_SWAP(m.version); 226 M_32_SWAP(m.psize); 227 M_32_SWAP(m.free); 228 M_32_SWAP(m.nrecs); 229 M_32_SWAP(m.flags); 230 } 231 if (m.magic != BTREEMAGIC || m.version != BTREEVERSION) 232 goto eftype; 233 if (m.psize < MINPSIZE || m.psize > MAX_PAGE_OFFSET + 1 || 234 m.psize & (sizeof(indx_t) - 1)) 235 goto eftype; 236 if (m.flags & ~SAVEMETA) 237 goto eftype; 238 b.psize = m.psize; 239 F_SET(t, m.flags); 240 t->bt_free = m.free; 241 t->bt_nrecs = m.nrecs; 242 } else { 243 /* 244 * Set the page size to the best value for I/O to this file. 245 * Don't overflow the page offset type. 246 */ 247 if (b.psize == 0) { 248 b.psize = sb.st_blksize; 249 if (b.psize < MINPSIZE) 250 b.psize = MINPSIZE; 251 if (b.psize > MAX_PAGE_OFFSET + 1) 252 b.psize = MAX_PAGE_OFFSET + 1; 253 } 254 255 /* Set flag if duplicates permitted. */ 256 if (!(b.flags & R_DUP)) 257 F_SET(t, B_NODUPS); 258 259 t->bt_free = P_INVALID; 260 t->bt_nrecs = 0; 261 F_SET(t, B_METADIRTY); 262 } 263 264 t->bt_psize = b.psize; 265 266 /* Set the cache size; must be a multiple of the page size. */ 267 if (b.cachesize && b.cachesize & (b.psize - 1)) 268 b.cachesize += (~b.cachesize & (b.psize - 1)) + 1; 269 if (b.cachesize < b.psize * MINCACHE) 270 b.cachesize = b.psize * MINCACHE; 271 272 /* Calculate number of pages to cache. */ 273 ncache = (b.cachesize + t->bt_psize - 1) / t->bt_psize; 274 275 /* 276 * The btree data structure requires that at least two keys can fit on 277 * a page, but other than that there's no fixed requirement. The user 278 * specified a minimum number per page, and we translated that into the 279 * number of bytes a key/data pair can use before being placed on an 280 * overflow page. This calculation includes the page header, the size 281 * of the index referencing the leaf item and the size of the leaf item 282 * structure. Also, don't let the user specify a minkeypage such that 283 * a key/data pair won't fit even if both key and data are on overflow 284 * pages. 285 */ 286 t->bt_ovflsize = (t->bt_psize - BTDATAOFF) / b.minkeypage - 287 (sizeof(indx_t) + NBLEAFDBT(0, 0)); 288 if (t->bt_ovflsize < NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t)) 289 t->bt_ovflsize = 290 NBLEAFDBT(NOVFLSIZE, NOVFLSIZE) + sizeof(indx_t); 291 292 /* Initialize the buffer pool. */ 293 if ((t->bt_mp = 294 mpool_open(NULL, t->bt_fd, t->bt_psize, ncache)) == NULL) 295 goto err; 296 if (!F_ISSET(t, B_INMEM)) 297 mpool_filter(t->bt_mp, __bt_pgin, __bt_pgout, t); 298 299 /* Create a root page if new tree. */ 300 if (nroot(t) == RET_ERROR) 301 goto err; 302 303 /* Global flags. */ 304 if (dflags & DB_LOCK) 305 F_SET(t, B_DB_LOCK); 306 if (dflags & DB_SHMEM) 307 F_SET(t, B_DB_SHMEM); 308 if (dflags & DB_TXN) 309 F_SET(t, B_DB_TXN); 310 311 return (dbp); 312 313einval: errno = EINVAL; 314 goto err; 315 316eftype: errno = EFTYPE; 317 goto err; 318 319err: saved_errno = errno; 320 if (t) { 321 free(t->bt_dbp); 322 if (t->bt_fd != -1) 323 (void)close(t->bt_fd); 324 free(t); 325 } 326 errno = saved_errno; 327 return (NULL); 328} 329 330/* 331 * NROOT -- Create the root of a new tree. 332 * 333 * Parameters: 334 * t: tree 335 * 336 * Returns: 337 * RET_ERROR, RET_SUCCESS 338 */ 339static int 340nroot(BTREE *t) 341{ 342 PAGE *meta, *root; 343 pgno_t npg; 344 345 if ((root = mpool_get(t->bt_mp, 1, 0)) != NULL) { 346 if (root->lower == 0 && 347 root->pgno == 0 && 348 root->linp[0] == 0) { 349 mpool_delete(t->bt_mp, root); 350 errno = EINVAL; 351 } else { 352 mpool_put(t->bt_mp, root, 0); 353 return (RET_SUCCESS); 354 } 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, MPOOL_PAGE_NEXT)) == NULL) 361 return (RET_ERROR); 362 363 if ((root = mpool_new(t->bt_mp, &npg, MPOOL_PAGE_NEXT)) == 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 380tmp(void) 381{ 382 sigset_t set, oset; 383 int fd, len; 384 char *envtmp = NULL; 385 char path[PATH_MAX]; 386 387 if (issetugid() == 0) 388 envtmp = getenv("TMPDIR"); 389 len = snprintf(path, 390 sizeof(path), "%s/bt.XXXXXX", envtmp ? envtmp : "/tmp"); 391 if (len < 0 || len >= sizeof(path)) { 392 errno = ENAMETOOLONG; 393 return(-1); 394 } 395 396 (void)sigfillset(&set); 397 (void)sigprocmask(SIG_BLOCK, &set, &oset); 398 if ((fd = mkostemp(path, O_CLOEXEC)) != -1) 399 (void)unlink(path); 400 (void)sigprocmask(SIG_SETMASK, &oset, NULL); 401 return(fd); 402} 403 404static int 405byteorder(void) 406{ 407 u_int32_t x; 408 u_char *p; 409 410 x = 0x01020304; 411 p = (u_char *)&x; 412 switch (*p) { 413 case 1: 414 return (BIG_ENDIAN); 415 case 4: 416 return (LITTLE_ENDIAN); 417 default: 418 return (0); 419 } 420} 421 422int 423__bt_fd(const DB *dbp) 424{ 425 BTREE *t; 426 427 t = dbp->internal; 428 429 /* Toss any page pinned across calls. */ 430 if (t->bt_pinned != NULL) { 431 mpool_put(t->bt_mp, t->bt_pinned, 0); 432 t->bt_pinned = NULL; 433 } 434 435 /* In-memory database can't have a file descriptor. */ 436 if (F_ISSET(t, B_INMEM)) { 437 errno = ENOENT; 438 return (-1); 439 } 440 return (t->bt_fd); 441} 442