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 * Margo Seltzer. 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 * $FreeBSD: src/lib/libc/db/hash/hash.c,v 1.8 2000/01/27 23:06:08 jasone Exp $ 37 * $DragonFly: src/lib/libc/db/hash/hash.c,v 1.7 2005/04/25 19:36:57 joerg Exp $ 38 * 39 * @(#)hash.c 8.9 (Berkeley) 6/16/94 40 */ 41 42 #include "namespace.h" 43 #include <sys/param.h> 44 #include <sys/stat.h> 45 46 #include <errno.h> 47 #include <fcntl.h> 48 #include <stdio.h> 49 #include <stdlib.h> 50 #include <string.h> 51 #include <unistd.h> 52 #ifdef DEBUG 53 #include <assert.h> 54 #endif 55 #include "un-namespace.h" 56 57 #include <db.h> 58 #include "hash.h" 59 #include "page.h" 60 #include "extern.h" 61 62 static int alloc_segs(HTAB *, int); 63 static int flush_meta(HTAB *); 64 static int hash_access(HTAB *, ACTION, const DBT *, DBT *); 65 static int hash_close(DB *); 66 static int hash_delete(const DB *, const DBT *, u_int32_t); 67 static int hash_fd(const DB *); 68 static int hash_get(const DB *, const DBT *, DBT *, u_int32_t); 69 static int hash_put(const DB *, DBT *, const DBT *, u_int32_t); 70 static void *hash_realloc(SEGMENT **, int, int); 71 static int hash_seq(const DB *, DBT *, DBT *, u_int32_t); 72 static int hash_sync(const DB *, u_int32_t); 73 static int hdestroy(HTAB *); 74 static HTAB *init_hash(HTAB *, const char *, const HASHINFO *); 75 static int init_htab(HTAB *, int); 76 #if BYTE_ORDER == LITTLE_ENDIAN 77 static void swap_header(HTAB *); 78 static void swap_header_copy(HASHHDR *, HASHHDR *); 79 #endif 80 81 /* Fast arithmetic, relying on powers of 2, */ 82 #define MOD(x, y) ((x) & ((y) - 1)) 83 84 #define RETURN_ERROR(ERR, LOC) { save_errno = ERR; goto LOC; } 85 86 /* Return values */ 87 #define SUCCESS (0) 88 #define ERROR (-1) 89 #define ABNORMAL (1) 90 91 #ifdef HASH_STATISTICS 92 int hash_accesses, hash_collisions, hash_expansions, hash_overflows; 93 #endif 94 95 /************************** INTERFACE ROUTINES ***************************/ 96 /* OPEN/CLOSE */ 97 98 extern DB * 99 __hash_open(const char *file, int flags, int mode, const HASHINFO * info, 100 int dflags __unused) 101 { 102 HTAB *hashp; 103 struct stat statbuf; 104 DB *dbp; 105 int bpages, hdrsize, new_table, nsegs, save_errno; 106 107 if ((flags & O_ACCMODE) == O_WRONLY) { 108 errno = EINVAL; 109 return (NULL); 110 } 111 112 if (!(hashp = (HTAB *)calloc(1, sizeof(HTAB)))) 113 return (NULL); 114 hashp->fp = -1; 115 116 /* 117 * Even if user wants write only, we need to be able to read 118 * the actual file, so we need to open it read/write. But, the 119 * field in the hashp structure needs to be accurate so that 120 * we can check accesses. 121 */ 122 hashp->flags = flags; 123 124 new_table = 0; 125 if (!file || (flags & O_TRUNC) || 126 (stat(file, &statbuf) && (errno == ENOENT))) { 127 if (errno == ENOENT) 128 errno = 0; /* Just in case someone looks at errno */ 129 new_table = 1; 130 } 131 if (file) { 132 if ((hashp->fp = _open(file, flags, mode)) == -1) 133 RETURN_ERROR(errno, error0); 134 135 /* if the .db file is empty, and we had permission to create 136 a new .db file, then reinitialize the database */ 137 if ((flags & O_CREAT) && 138 _fstat(hashp->fp, &statbuf) == 0 && statbuf.st_size == 0) 139 new_table = 1; 140 141 (void)_fcntl(hashp->fp, F_SETFD, 1); 142 } 143 if (new_table) { 144 if (!(hashp = init_hash(hashp, file, info))) 145 RETURN_ERROR(errno, error1); 146 } else { 147 /* Table already exists */ 148 if (info && info->hash) 149 hashp->hash = info->hash; 150 else 151 hashp->hash = __default_hash; 152 153 hdrsize = _read(hashp->fp, &hashp->hdr, sizeof(HASHHDR)); 154 #if BYTE_ORDER == LITTLE_ENDIAN 155 swap_header(hashp); 156 #endif 157 if (hdrsize == -1) 158 RETURN_ERROR(errno, error1); 159 if (hdrsize != sizeof(HASHHDR)) 160 RETURN_ERROR(EFTYPE, error1); 161 /* Verify file type, versions and hash function */ 162 if (hashp->MAGIC != HASHMAGIC) 163 RETURN_ERROR(EFTYPE, error1); 164 #define OLDHASHVERSION 1 165 if (hashp->VERSION != HASHVERSION && 166 hashp->VERSION != OLDHASHVERSION) 167 RETURN_ERROR(EFTYPE, error1); 168 if (hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY) 169 RETURN_ERROR(EFTYPE, error1); 170 /* 171 * Figure out how many segments we need. Max_Bucket is the 172 * maximum bucket number, so the number of buckets is 173 * max_bucket + 1. 174 */ 175 nsegs = (hashp->MAX_BUCKET + 1 + hashp->SGSIZE - 1) / 176 hashp->SGSIZE; 177 hashp->nsegs = 0; 178 if (alloc_segs(hashp, nsegs)) 179 /* 180 * If alloc_segs fails, table will have been destroyed 181 * and errno will have been set. 182 */ 183 return (NULL); 184 /* Read in bitmaps */ 185 bpages = (hashp->SPARES[hashp->OVFL_POINT] + 186 (hashp->BSIZE << BYTE_SHIFT) - 1) >> 187 (hashp->BSHIFT + BYTE_SHIFT); 188 189 hashp->nmaps = bpages; 190 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(u_int32_t *)); 191 } 192 193 /* Initialize Buffer Manager */ 194 if (info && info->cachesize) 195 __buf_init(hashp, info->cachesize); 196 else 197 __buf_init(hashp, DEF_BUFSIZE); 198 199 hashp->new_file = new_table; 200 hashp->save_file = file && (hashp->flags & O_RDWR); 201 hashp->cbucket = -1; 202 if (!(dbp = (DB *)malloc(sizeof(DB)))) { 203 save_errno = errno; 204 hdestroy(hashp); 205 errno = save_errno; 206 return (NULL); 207 } 208 dbp->internal = hashp; 209 dbp->close = hash_close; 210 dbp->del = hash_delete; 211 dbp->fd = hash_fd; 212 dbp->get = hash_get; 213 dbp->put = hash_put; 214 dbp->seq = hash_seq; 215 dbp->sync = hash_sync; 216 dbp->type = DB_HASH; 217 218 #ifdef DEBUG 219 (void)fprintf(stderr, 220 "%s\n%s%x\n%s%d\n%s%d\n%s%d\n%s%d\n%s%d\n%s%d\n%s%d\n%s%d\n%s%d\n%s%x\n%s%x\n%s%d\n%s%d\n", 221 "init_htab:", 222 "TABLE POINTER ", hashp, 223 "BUCKET SIZE ", hashp->BSIZE, 224 "BUCKET SHIFT ", hashp->BSHIFT, 225 "DIRECTORY SIZE ", hashp->DSIZE, 226 "SEGMENT SIZE ", hashp->SGSIZE, 227 "SEGMENT SHIFT ", hashp->SSHIFT, 228 "FILL FACTOR ", hashp->FFACTOR, 229 "MAX BUCKET ", hashp->MAX_BUCKET, 230 "OVFL POINT ", hashp->OVFL_POINT, 231 "LAST FREED ", hashp->LAST_FREED, 232 "HIGH MASK ", hashp->HIGH_MASK, 233 "LOW MASK ", hashp->LOW_MASK, 234 "NSEGS ", hashp->nsegs, 235 "NKEYS ", hashp->NKEYS); 236 #endif 237 #ifdef HASH_STATISTICS 238 hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0; 239 #endif 240 return (dbp); 241 242 error1: 243 if (hashp != NULL) 244 (void)_close(hashp->fp); 245 246 error0: 247 free(hashp); 248 errno = save_errno; 249 return (NULL); 250 } 251 252 static int 253 hash_close(DB *dbp) 254 { 255 HTAB *hashp; 256 int retval; 257 258 if (!dbp) 259 return (ERROR); 260 261 hashp = (HTAB *)dbp->internal; 262 retval = hdestroy(hashp); 263 free(dbp); 264 return (retval); 265 } 266 267 static int 268 hash_fd(const DB *dbp) 269 { 270 HTAB *hashp; 271 272 if (!dbp) 273 return (ERROR); 274 275 hashp = (HTAB *)dbp->internal; 276 if (hashp->fp == -1) { 277 errno = ENOENT; 278 return (-1); 279 } 280 return (hashp->fp); 281 } 282 283 /************************** LOCAL CREATION ROUTINES **********************/ 284 static HTAB * 285 init_hash(HTAB *hashp, const char *file, const HASHINFO *info) 286 { 287 struct stat statbuf; 288 int nelem; 289 290 nelem = 1; 291 hashp->NKEYS = 0; 292 hashp->LORDER = BYTE_ORDER; 293 hashp->BSIZE = DEF_BUCKET_SIZE; 294 hashp->BSHIFT = DEF_BUCKET_SHIFT; 295 hashp->SGSIZE = DEF_SEGSIZE; 296 hashp->SSHIFT = DEF_SEGSIZE_SHIFT; 297 hashp->DSIZE = DEF_DIRSIZE; 298 hashp->FFACTOR = DEF_FFACTOR; 299 hashp->hash = __default_hash; 300 memset(hashp->SPARES, 0, sizeof(hashp->SPARES)); 301 memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS)); 302 303 /* Fix bucket size to be optimal for file system */ 304 if (file != NULL) { 305 if (stat(file, &statbuf)) 306 return (NULL); 307 hashp->BSIZE = statbuf.st_blksize; 308 hashp->BSHIFT = __log2(hashp->BSIZE); 309 } 310 311 if (info) { 312 if (info->bsize) { 313 /* Round pagesize up to power of 2 */ 314 hashp->BSHIFT = __log2(info->bsize); 315 hashp->BSIZE = 1 << hashp->BSHIFT; 316 if (hashp->BSIZE > MAX_BSIZE) { 317 errno = EINVAL; 318 return (NULL); 319 } 320 } 321 if (info->ffactor) 322 hashp->FFACTOR = info->ffactor; 323 if (info->hash) 324 hashp->hash = info->hash; 325 if (info->nelem) 326 nelem = info->nelem; 327 if (info->lorder) { 328 if (info->lorder != BIG_ENDIAN && 329 info->lorder != LITTLE_ENDIAN) { 330 errno = EINVAL; 331 return (NULL); 332 } 333 hashp->LORDER = info->lorder; 334 } 335 } 336 /* init_htab should destroy the table and set errno if it fails */ 337 if (init_htab(hashp, nelem)) 338 return (NULL); 339 else 340 return (hashp); 341 } 342 /* 343 * This calls alloc_segs which may run out of memory. Alloc_segs will destroy 344 * the table and set errno, so we just pass the error information along. 345 * 346 * Returns 0 on No Error 347 */ 348 static int 349 init_htab(HTAB *hashp, int nelem) 350 { 351 int nbuckets, nsegs; 352 int l2; 353 354 /* 355 * Divide number of elements by the fill factor and determine a 356 * desired number of buckets. Allocate space for the next greater 357 * power of two number of buckets. 358 */ 359 nelem = (nelem - 1) / hashp->FFACTOR + 1; 360 361 l2 = __log2(MAX(nelem, 2)); 362 nbuckets = 1 << l2; 363 364 hashp->SPARES[l2] = l2 + 1; 365 hashp->SPARES[l2 + 1] = l2 + 1; 366 hashp->OVFL_POINT = l2; 367 hashp->LAST_FREED = 2; 368 369 /* First bitmap page is at: splitpoint l2 page offset 1 */ 370 if (__ibitmap(hashp, OADDR_OF(l2, 1), l2 + 1, 0)) 371 return (-1); 372 373 hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1; 374 hashp->HIGH_MASK = (nbuckets << 1) - 1; 375 hashp->HDRPAGES = ((MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >> 376 hashp->BSHIFT) + 1; 377 378 nsegs = (nbuckets - 1) / hashp->SGSIZE + 1; 379 nsegs = 1 << __log2(nsegs); 380 381 if (nsegs > hashp->DSIZE) 382 hashp->DSIZE = nsegs; 383 return (alloc_segs(hashp, nsegs)); 384 } 385 386 /********************** DESTROY/CLOSE ROUTINES ************************/ 387 388 /* 389 * Flushes any changes to the file if necessary and destroys the hashp 390 * structure, freeing all allocated space. 391 */ 392 static int 393 hdestroy(HTAB *hashp) 394 { 395 int i, save_errno; 396 397 save_errno = 0; 398 399 #ifdef HASH_STATISTICS 400 (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n", 401 hash_accesses, hash_collisions); 402 (void)fprintf(stderr, "hdestroy: expansions %ld\n", 403 hash_expansions); 404 (void)fprintf(stderr, "hdestroy: overflows %ld\n", 405 hash_overflows); 406 (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n", 407 hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs); 408 409 for (i = 0; i < NCACHED; i++) 410 (void)fprintf(stderr, 411 "spares[%d] = %d\n", i, hashp->SPARES[i]); 412 #endif 413 /* 414 * Call on buffer manager to free buffers, and if required, 415 * write them to disk. 416 */ 417 if (__buf_free(hashp, 1, hashp->save_file)) 418 save_errno = errno; 419 if (hashp->dir) { 420 free(*hashp->dir); /* Free initial segments */ 421 /* Free extra segments */ 422 while (hashp->exsegs--) 423 free(hashp->dir[--hashp->nsegs]); 424 free(hashp->dir); 425 } 426 if (flush_meta(hashp) && !save_errno) 427 save_errno = errno; 428 /* Free Bigmaps */ 429 for (i = 0; i < hashp->nmaps; i++) 430 if (hashp->mapp[i]) 431 free(hashp->mapp[i]); 432 433 if (hashp->fp != -1) 434 (void)_close(hashp->fp); 435 436 free(hashp); 437 438 if (save_errno) { 439 errno = save_errno; 440 return (ERROR); 441 } 442 return (SUCCESS); 443 } 444 /* 445 * Write modified pages to disk 446 * 447 * Returns: 448 * 0 == OK 449 * -1 ERROR 450 */ 451 static int 452 hash_sync(const DB *dbp, u_int32_t flags) 453 { 454 HTAB *hashp; 455 456 if (flags != 0) { 457 errno = EINVAL; 458 return (ERROR); 459 } 460 461 if (!dbp) 462 return (ERROR); 463 464 hashp = (HTAB *)dbp->internal; 465 if (!hashp->save_file) 466 return (0); 467 if (__buf_free(hashp, 0, 1) || flush_meta(hashp)) 468 return (ERROR); 469 hashp->new_file = 0; 470 return (0); 471 } 472 473 /* 474 * Returns: 475 * 0 == OK 476 * -1 indicates that errno should be set 477 */ 478 static int 479 flush_meta(HTAB *hashp) 480 { 481 HASHHDR *whdrp; 482 #if BYTE_ORDER == LITTLE_ENDIAN 483 HASHHDR whdr; 484 #endif 485 int fp, i, wsize; 486 487 if (!hashp->save_file) 488 return (0); 489 hashp->MAGIC = HASHMAGIC; 490 hashp->VERSION = HASHVERSION; 491 hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY)); 492 493 fp = hashp->fp; 494 whdrp = &hashp->hdr; 495 #if BYTE_ORDER == LITTLE_ENDIAN 496 whdrp = &whdr; 497 swap_header_copy(&hashp->hdr, whdrp); 498 #endif 499 if ((lseek(fp, (off_t)0, SEEK_SET) == -1) || 500 ((wsize = _write(fp, whdrp, sizeof(HASHHDR))) == -1)) 501 return (-1); 502 else 503 if (wsize != sizeof(HASHHDR)) { 504 errno = EFTYPE; 505 hashp->error = errno; 506 return (-1); 507 } 508 for (i = 0; i < NCACHED; i++) 509 if (hashp->mapp[i]) 510 if (__put_page(hashp, (char *)hashp->mapp[i], 511 hashp->BITMAPS[i], 0, 1)) 512 return (-1); 513 return (0); 514 } 515 516 /*******************************SEARCH ROUTINES *****************************/ 517 /* 518 * All the access routines return 519 * 520 * Returns: 521 * 0 on SUCCESS 522 * 1 to indicate an external ERROR (i.e. key not found, etc) 523 * -1 to indicate an internal ERROR (i.e. out of memory, etc) 524 */ 525 static int 526 hash_get(const DB *dbp, const DBT *key, DBT *data, u_int32_t flag) 527 { 528 HTAB *hashp; 529 530 hashp = (HTAB *)dbp->internal; 531 if (flag) { 532 hashp->error = EINVAL; 533 errno = EINVAL; 534 return (ERROR); 535 } 536 return (hash_access(hashp, HASH_GET, key, data)); 537 } 538 539 static int 540 hash_put(const DB *dbp, DBT *key, const DBT *data, u_int32_t flag) 541 { 542 HTAB *hashp; 543 544 hashp = (HTAB *)dbp->internal; 545 if (flag && flag != R_NOOVERWRITE) { 546 hashp->error = errno = EINVAL; 547 return (ERROR); 548 } 549 if ((hashp->flags & O_ACCMODE) == O_RDONLY) { 550 hashp->error = errno = EPERM; 551 return (ERROR); 552 } 553 return (hash_access(hashp, flag == R_NOOVERWRITE ? 554 HASH_PUTNEW : HASH_PUT, key, __DECONST(DBT *, data))); 555 } 556 557 static int 558 hash_delete(const DB *dbp, const DBT *key, u_int32_t flag) 559 { 560 HTAB *hashp; 561 562 hashp = (HTAB *)dbp->internal; 563 if (flag && flag != R_CURSOR) { 564 hashp->error = errno = EINVAL; 565 return (ERROR); 566 } 567 if ((hashp->flags & O_ACCMODE) == O_RDONLY) { 568 hashp->error = errno = EPERM; 569 return (ERROR); 570 } 571 return (hash_access(hashp, HASH_DELETE, key, NULL)); 572 } 573 574 /* 575 * Assume that hashp has been set in wrapper routine. 576 */ 577 static int 578 hash_access(HTAB *hashp, ACTION action, const DBT *key, DBT *val) 579 { 580 BUFHEAD *rbufp; 581 BUFHEAD *bufp, *save_bufp; 582 u_int16_t *bp; 583 int n, ndx, off, size; 584 const char *kp; 585 u_int16_t pageno; 586 587 #ifdef HASH_STATISTICS 588 hash_accesses++; 589 #endif 590 591 off = hashp->BSIZE; 592 size = key->size; 593 kp = (char *)key->data; 594 rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0); 595 if (!rbufp) 596 return (ERROR); 597 save_bufp = rbufp; 598 599 /* Pin the bucket chain */ 600 rbufp->flags |= BUF_PIN; 601 for (bp = (u_int16_t *)rbufp->page, n = *bp++, ndx = 1; ndx < n;) 602 if (bp[1] >= REAL_KEY) { 603 /* Real key/data pair */ 604 if (size == off - *bp && 605 memcmp(kp, rbufp->page + *bp, size) == 0) 606 goto found; 607 off = bp[1]; 608 #ifdef HASH_STATISTICS 609 hash_collisions++; 610 #endif 611 bp += 2; 612 ndx += 2; 613 } else if (bp[1] == OVFLPAGE) { 614 rbufp = __get_buf(hashp, *bp, rbufp, 0); 615 if (!rbufp) { 616 save_bufp->flags &= ~BUF_PIN; 617 return (ERROR); 618 } 619 /* FOR LOOP INIT */ 620 bp = (u_int16_t *)rbufp->page; 621 n = *bp++; 622 ndx = 1; 623 off = hashp->BSIZE; 624 } else if (bp[1] < REAL_KEY) { 625 if ((ndx = 626 __find_bigpair(hashp, rbufp, ndx, kp, size)) > 0) 627 goto found; 628 if (ndx == -2) { 629 bufp = rbufp; 630 if (!(pageno = 631 __find_last_page(hashp, &bufp))) { 632 ndx = 0; 633 rbufp = bufp; 634 break; /* FOR */ 635 } 636 rbufp = __get_buf(hashp, pageno, bufp, 0); 637 if (!rbufp) { 638 save_bufp->flags &= ~BUF_PIN; 639 return (ERROR); 640 } 641 /* FOR LOOP INIT */ 642 bp = (u_int16_t *)rbufp->page; 643 n = *bp++; 644 ndx = 1; 645 off = hashp->BSIZE; 646 } else { 647 save_bufp->flags &= ~BUF_PIN; 648 return (ERROR); 649 } 650 } 651 652 /* Not found */ 653 switch (action) { 654 case HASH_PUT: 655 case HASH_PUTNEW: 656 if (__addel(hashp, rbufp, key, val)) { 657 save_bufp->flags &= ~BUF_PIN; 658 return (ERROR); 659 } else { 660 save_bufp->flags &= ~BUF_PIN; 661 return (SUCCESS); 662 } 663 case HASH_GET: 664 case HASH_DELETE: 665 default: 666 save_bufp->flags &= ~BUF_PIN; 667 return (ABNORMAL); 668 } 669 670 found: 671 switch (action) { 672 case HASH_PUTNEW: 673 save_bufp->flags &= ~BUF_PIN; 674 return (ABNORMAL); 675 case HASH_GET: 676 bp = (u_int16_t *)rbufp->page; 677 if (bp[ndx + 1] < REAL_KEY) { 678 if (__big_return(hashp, rbufp, ndx, val, 0)) 679 return (ERROR); 680 } else { 681 val->data = (u_char *)rbufp->page + (int)bp[ndx + 1]; 682 val->size = bp[ndx] - bp[ndx + 1]; 683 } 684 break; 685 case HASH_PUT: 686 if ((__delpair(hashp, rbufp, ndx)) || 687 (__addel(hashp, rbufp, key, val))) { 688 save_bufp->flags &= ~BUF_PIN; 689 return (ERROR); 690 } 691 break; 692 case HASH_DELETE: 693 if (__delpair(hashp, rbufp, ndx)) 694 return (ERROR); 695 break; 696 default: 697 abort(); 698 } 699 save_bufp->flags &= ~BUF_PIN; 700 return (SUCCESS); 701 } 702 703 static int 704 hash_seq(const DB *dbp, DBT *key, DBT *data, u_int32_t flag) 705 { 706 u_int32_t bucket; 707 BUFHEAD *bufp; 708 HTAB *hashp; 709 u_int16_t *bp, ndx; 710 711 hashp = (HTAB *)dbp->internal; 712 if (flag && flag != R_FIRST && flag != R_NEXT) { 713 hashp->error = errno = EINVAL; 714 return (ERROR); 715 } 716 #ifdef HASH_STATISTICS 717 hash_accesses++; 718 #endif 719 if ((hashp->cbucket < 0) || (flag == R_FIRST)) { 720 hashp->cbucket = 0; 721 hashp->cndx = 1; 722 hashp->cpage = NULL; 723 } 724 725 for (bp = NULL; !bp || !bp[0]; ) { 726 if (!(bufp = hashp->cpage)) { 727 for (bucket = hashp->cbucket; 728 bucket <= hashp->MAX_BUCKET; 729 bucket++, hashp->cndx = 1) { 730 bufp = __get_buf(hashp, bucket, NULL, 0); 731 if (!bufp) 732 return (ERROR); 733 hashp->cpage = bufp; 734 bp = (u_int16_t *)bufp->page; 735 if (bp[0]) 736 break; 737 } 738 hashp->cbucket = bucket; 739 if (hashp->cbucket > hashp->MAX_BUCKET) { 740 hashp->cbucket = -1; 741 return (ABNORMAL); 742 } 743 } else 744 bp = (u_int16_t *)hashp->cpage->page; 745 746 #ifdef DEBUG 747 assert(bp); 748 assert(bufp); 749 #endif 750 while (bp[hashp->cndx + 1] == OVFLPAGE) { 751 bufp = hashp->cpage = 752 __get_buf(hashp, bp[hashp->cndx], bufp, 0); 753 if (!bufp) 754 return (ERROR); 755 bp = (u_int16_t *)(bufp->page); 756 hashp->cndx = 1; 757 } 758 if (!bp[0]) { 759 hashp->cpage = NULL; 760 ++hashp->cbucket; 761 } 762 } 763 ndx = hashp->cndx; 764 if (bp[ndx + 1] < REAL_KEY) { 765 if (__big_keydata(hashp, bufp, key, data, 1)) 766 return (ERROR); 767 } else { 768 key->data = (u_char *)hashp->cpage->page + bp[ndx]; 769 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->BSIZE) - bp[ndx]; 770 data->data = (u_char *)hashp->cpage->page + bp[ndx + 1]; 771 data->size = bp[ndx] - bp[ndx + 1]; 772 ndx += 2; 773 if (ndx > bp[0]) { 774 hashp->cpage = NULL; 775 hashp->cbucket++; 776 hashp->cndx = 1; 777 } else 778 hashp->cndx = ndx; 779 } 780 return (SUCCESS); 781 } 782 783 /********************************* UTILITIES ************************/ 784 785 /* 786 * Returns: 787 * 0 ==> OK 788 * -1 ==> Error 789 */ 790 extern int 791 __expand_table(HTAB *hashp) 792 { 793 u_int32_t old_bucket, new_bucket; 794 int dirsize, new_segnum, spare_ndx; 795 796 #ifdef HASH_STATISTICS 797 hash_expansions++; 798 #endif 799 new_bucket = ++hashp->MAX_BUCKET; 800 old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK); 801 802 new_segnum = new_bucket >> hashp->SSHIFT; 803 804 /* Check if we need a new segment */ 805 if (new_segnum >= hashp->nsegs) { 806 /* Check if we need to expand directory */ 807 if (new_segnum >= hashp->DSIZE) { 808 /* Reallocate directory */ 809 dirsize = hashp->DSIZE * sizeof(SEGMENT *); 810 if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1)) 811 return (-1); 812 hashp->DSIZE = dirsize << 1; 813 } 814 if ((hashp->dir[new_segnum] = 815 (SEGMENT)calloc(hashp->SGSIZE, sizeof(SEGMENT))) == NULL) 816 return (-1); 817 hashp->exsegs++; 818 hashp->nsegs++; 819 } 820 /* 821 * If the split point is increasing (MAX_BUCKET's log base 2 822 * * increases), we need to copy the current contents of the spare 823 * split bucket to the next bucket. 824 */ 825 spare_ndx = __log2(hashp->MAX_BUCKET + 1); 826 if (spare_ndx > hashp->OVFL_POINT) { 827 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT]; 828 hashp->OVFL_POINT = spare_ndx; 829 } 830 831 if (new_bucket > hashp->HIGH_MASK) { 832 /* Starting a new doubling */ 833 hashp->LOW_MASK = hashp->HIGH_MASK; 834 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK; 835 } 836 /* Relocate records to the new bucket */ 837 return (__split_page(hashp, old_bucket, new_bucket)); 838 } 839 840 /* 841 * If realloc guarantees that the pointer is not destroyed if the realloc 842 * fails, then this routine can go away. 843 */ 844 static void * 845 hash_realloc(SEGMENT **p_ptr, int oldsize, int newsize) 846 { 847 void *p; 848 849 if ( (p = malloc(newsize)) ) { 850 memmove(p, *p_ptr, oldsize); 851 memset((char *)p + oldsize, 0, newsize - oldsize); 852 free(*p_ptr); 853 *p_ptr = p; 854 } 855 return (p); 856 } 857 858 extern u_int32_t 859 __call_hash(HTAB *hashp, const char *k, int len) 860 { 861 int n, bucket; 862 863 n = hashp->hash(k, len); 864 bucket = n & hashp->HIGH_MASK; 865 if (bucket > hashp->MAX_BUCKET) 866 bucket = bucket & hashp->LOW_MASK; 867 return (bucket); 868 } 869 870 /* 871 * Allocate segment table. On error, destroy the table and set errno. 872 * 873 * Returns 0 on success 874 */ 875 static int 876 alloc_segs(HTAB *hashp, int nsegs) 877 { 878 int i; 879 SEGMENT store; 880 881 int save_errno; 882 883 if ((hashp->dir = 884 (SEGMENT *)calloc(hashp->DSIZE, sizeof(SEGMENT *))) == NULL) { 885 save_errno = errno; 886 (void)hdestroy(hashp); 887 errno = save_errno; 888 return (-1); 889 } 890 /* Allocate segments */ 891 if ((store = 892 (SEGMENT)calloc(nsegs << hashp->SSHIFT, sizeof(SEGMENT))) == NULL) { 893 save_errno = errno; 894 (void)hdestroy(hashp); 895 errno = save_errno; 896 return (-1); 897 } 898 for (i = 0; i < nsegs; i++, hashp->nsegs++) 899 hashp->dir[i] = &store[i << hashp->SSHIFT]; 900 return (0); 901 } 902 903 #if BYTE_ORDER == LITTLE_ENDIAN 904 /* 905 * Hashp->hdr needs to be byteswapped. 906 */ 907 static void 908 swap_header_copy(HASHHDR *srcp, HASHHDR *destp) 909 { 910 int i; 911 912 P_32_COPY(srcp->magic, destp->magic); 913 P_32_COPY(srcp->version, destp->version); 914 P_32_COPY(srcp->lorder, destp->lorder); 915 P_32_COPY(srcp->bsize, destp->bsize); 916 P_32_COPY(srcp->bshift, destp->bshift); 917 P_32_COPY(srcp->dsize, destp->dsize); 918 P_32_COPY(srcp->ssize, destp->ssize); 919 P_32_COPY(srcp->sshift, destp->sshift); 920 P_32_COPY(srcp->ovfl_point, destp->ovfl_point); 921 P_32_COPY(srcp->last_freed, destp->last_freed); 922 P_32_COPY(srcp->max_bucket, destp->max_bucket); 923 P_32_COPY(srcp->high_mask, destp->high_mask); 924 P_32_COPY(srcp->low_mask, destp->low_mask); 925 P_32_COPY(srcp->ffactor, destp->ffactor); 926 P_32_COPY(srcp->nkeys, destp->nkeys); 927 P_32_COPY(srcp->hdrpages, destp->hdrpages); 928 P_32_COPY(srcp->h_charkey, destp->h_charkey); 929 for (i = 0; i < NCACHED; i++) { 930 P_32_COPY(srcp->spares[i], destp->spares[i]); 931 P_16_COPY(srcp->bitmaps[i], destp->bitmaps[i]); 932 } 933 } 934 935 static void 936 swap_header(HTAB *hashp) 937 { 938 HASHHDR *hdrp; 939 int i; 940 941 hdrp = &hashp->hdr; 942 943 M_32_SWAP(hdrp->magic); 944 M_32_SWAP(hdrp->version); 945 M_32_SWAP(hdrp->lorder); 946 M_32_SWAP(hdrp->bsize); 947 M_32_SWAP(hdrp->bshift); 948 M_32_SWAP(hdrp->dsize); 949 M_32_SWAP(hdrp->ssize); 950 M_32_SWAP(hdrp->sshift); 951 M_32_SWAP(hdrp->ovfl_point); 952 M_32_SWAP(hdrp->last_freed); 953 M_32_SWAP(hdrp->max_bucket); 954 M_32_SWAP(hdrp->high_mask); 955 M_32_SWAP(hdrp->low_mask); 956 M_32_SWAP(hdrp->ffactor); 957 M_32_SWAP(hdrp->nkeys); 958 M_32_SWAP(hdrp->hdrpages); 959 M_32_SWAP(hdrp->h_charkey); 960 for (i = 0; i < NCACHED; i++) { 961 M_32_SWAP(hdrp->spares[i]); 962 M_16_SWAP(hdrp->bitmaps[i]); 963 } 964 } 965 #endif 966