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 * 4. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #if defined(LIBC_SCCS) && !defined(lint) 34 static char sccsid[] = "@(#)hash.c 8.9 (Berkeley) 6/16/94"; 35 #endif /* LIBC_SCCS and not lint */ 36 #include <sys/cdefs.h> 37 __FBSDID("$FreeBSD$"); 38 39 #include "namespace.h" 40 #include <sys/param.h> 41 #include <sys/stat.h> 42 43 #include <errno.h> 44 #include <fcntl.h> 45 #include <stdio.h> 46 #include <stdlib.h> 47 #include <string.h> 48 #include <unistd.h> 49 #ifdef DEBUG 50 #include <assert.h> 51 #endif 52 #include "un-namespace.h" 53 54 #include <db.h> 55 #include "hash.h" 56 #include "page.h" 57 #include "extern.h" 58 59 static int alloc_segs(HTAB *, int); 60 static int flush_meta(HTAB *); 61 static int hash_access(HTAB *, ACTION, DBT *, DBT *); 62 static int hash_close(DB *); 63 static int hash_delete(const DB *, const DBT *, u_int32_t); 64 static int hash_fd(const DB *); 65 static int hash_get(const DB *, const DBT *, DBT *, u_int32_t); 66 static int hash_put(const DB *, DBT *, const DBT *, u_int32_t); 67 static void *hash_realloc(SEGMENT **, int, int); 68 static int hash_seq(const DB *, DBT *, DBT *, u_int32_t); 69 static int hash_sync(const DB *, u_int32_t); 70 static int hdestroy(HTAB *); 71 static HTAB *init_hash(HTAB *, const char *, const HASHINFO *); 72 static int init_htab(HTAB *, int); 73 #if BYTE_ORDER == LITTLE_ENDIAN 74 static void swap_header(HTAB *); 75 static void swap_header_copy(HASHHDR *, HASHHDR *); 76 #endif 77 78 /* Fast arithmetic, relying on powers of 2, */ 79 #define MOD(x, y) ((x) & ((y) - 1)) 80 81 #define RETURN_ERROR(ERR, LOC) { save_errno = ERR; goto LOC; } 82 83 /* Return values */ 84 #define SUCCESS (0) 85 #define ERROR (-1) 86 #define ABNORMAL (1) 87 88 #ifdef HASH_STATISTICS 89 int hash_accesses, hash_collisions, hash_expansions, hash_overflows; 90 #endif 91 92 /************************** INTERFACE ROUTINES ***************************/ 93 /* OPEN/CLOSE */ 94 95 extern DB * 96 __hash_open(file, flags, mode, info, dflags) 97 const char *file; 98 int flags, mode, dflags; 99 const HASHINFO *info; /* Special directives for create */ 100 { 101 HTAB *hashp; 102 struct stat statbuf; 103 DB *dbp; 104 int bpages, hdrsize, new_table, nsegs, save_errno; 105 106 if ((flags & O_ACCMODE) == O_WRONLY) { 107 errno = EINVAL; 108 return (NULL); 109 } 110 111 if (!(hashp = (HTAB *)calloc(1, sizeof(HTAB)))) 112 return (NULL); 113 hashp->fp = -1; 114 115 /* 116 * Even if user wants write only, we need to be able to read 117 * the actual file, so we need to open it read/write. But, the 118 * field in the hashp structure needs to be accurate so that 119 * we can check accesses. 120 */ 121 hashp->flags = flags; 122 123 new_table = 0; 124 if (!file || (flags & O_TRUNC) || 125 (stat(file, &statbuf) && (errno == ENOENT))) { 126 if (errno == ENOENT) 127 errno = 0; /* Just in case someone looks at errno */ 128 new_table = 1; 129 } 130 if (file) { 131 if ((hashp->fp = _open(file, flags, mode)) == -1) 132 RETURN_ERROR(errno, error0); 133 134 /* if the .db file is empty, and we had permission to create 135 a new .db file, then reinitialize the database */ 136 if ((flags & O_CREAT) && 137 _fstat(hashp->fp, &statbuf) == 0 && statbuf.st_size == 0) 138 new_table = 1; 139 140 (void)_fcntl(hashp->fp, F_SETFD, 1); 141 } 142 if (new_table) { 143 if (!(hashp = init_hash(hashp, file, info))) 144 RETURN_ERROR(errno, error1); 145 } else { 146 /* Table already exists */ 147 if (info && info->hash) 148 hashp->hash = info->hash; 149 else 150 hashp->hash = __default_hash; 151 152 hdrsize = _read(hashp->fp, &hashp->hdr, sizeof(HASHHDR)); 153 #if BYTE_ORDER == LITTLE_ENDIAN 154 swap_header(hashp); 155 #endif 156 if (hdrsize == -1) 157 RETURN_ERROR(errno, error1); 158 if (hdrsize != sizeof(HASHHDR)) 159 RETURN_ERROR(EFTYPE, error1); 160 /* Verify file type, versions and hash function */ 161 if (hashp->MAGIC != HASHMAGIC) 162 RETURN_ERROR(EFTYPE, error1); 163 #define OLDHASHVERSION 1 164 if (hashp->VERSION != HASHVERSION && 165 hashp->VERSION != OLDHASHVERSION) 166 RETURN_ERROR(EFTYPE, error1); 167 if (hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY) 168 RETURN_ERROR(EFTYPE, error1); 169 /* 170 * Figure out how many segments we need. Max_Bucket is the 171 * maximum bucket number, so the number of buckets is 172 * max_bucket + 1. 173 */ 174 nsegs = (hashp->MAX_BUCKET + 1 + hashp->SGSIZE - 1) / 175 hashp->SGSIZE; 176 hashp->nsegs = 0; 177 if (alloc_segs(hashp, nsegs)) 178 /* 179 * If alloc_segs fails, table will have been destroyed 180 * and errno will have been set. 181 */ 182 return (NULL); 183 /* Read in bitmaps */ 184 bpages = (hashp->SPARES[hashp->OVFL_POINT] + 185 (hashp->BSIZE << BYTE_SHIFT) - 1) >> 186 (hashp->BSHIFT + BYTE_SHIFT); 187 188 hashp->nmaps = bpages; 189 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(u_int32_t *)); 190 } 191 192 /* Initialize Buffer Manager */ 193 if (info && info->cachesize) 194 __buf_init(hashp, info->cachesize); 195 else 196 __buf_init(hashp, DEF_BUFSIZE); 197 198 hashp->new_file = new_table; 199 hashp->save_file = file && (hashp->flags & O_RDWR); 200 hashp->cbucket = -1; 201 if (!(dbp = (DB *)malloc(sizeof(DB)))) { 202 save_errno = errno; 203 hdestroy(hashp); 204 errno = save_errno; 205 return (NULL); 206 } 207 dbp->internal = hashp; 208 dbp->close = hash_close; 209 dbp->del = hash_delete; 210 dbp->fd = hash_fd; 211 dbp->get = hash_get; 212 dbp->put = hash_put; 213 dbp->seq = hash_seq; 214 dbp->sync = hash_sync; 215 dbp->type = DB_HASH; 216 217 #ifdef DEBUG 218 (void)fprintf(stderr, 219 "%s\n%s%p\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", 220 "init_htab:", 221 "TABLE POINTER ", hashp, 222 "BUCKET SIZE ", hashp->BSIZE, 223 "BUCKET SHIFT ", hashp->BSHIFT, 224 "DIRECTORY SIZE ", hashp->DSIZE, 225 "SEGMENT SIZE ", hashp->SGSIZE, 226 "SEGMENT SHIFT ", hashp->SSHIFT, 227 "FILL FACTOR ", hashp->FFACTOR, 228 "MAX BUCKET ", hashp->MAX_BUCKET, 229 "OVFL POINT ", hashp->OVFL_POINT, 230 "LAST FREED ", hashp->LAST_FREED, 231 "HIGH MASK ", hashp->HIGH_MASK, 232 "LOW MASK ", hashp->LOW_MASK, 233 "NSEGS ", hashp->nsegs, 234 "NKEYS ", hashp->NKEYS); 235 #endif 236 #ifdef HASH_STATISTICS 237 hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0; 238 #endif 239 return (dbp); 240 241 error1: 242 if (hashp != NULL) 243 (void)_close(hashp->fp); 244 245 error0: 246 free(hashp); 247 errno = save_errno; 248 return (NULL); 249 } 250 251 static int 252 hash_close(dbp) 253 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(dbp) 269 const DB *dbp; 270 { 271 HTAB *hashp; 272 273 if (!dbp) 274 return (ERROR); 275 276 hashp = (HTAB *)dbp->internal; 277 if (hashp->fp == -1) { 278 errno = ENOENT; 279 return (-1); 280 } 281 return (hashp->fp); 282 } 283 284 /************************** LOCAL CREATION ROUTINES **********************/ 285 static HTAB * 286 init_hash(hashp, file, info) 287 HTAB *hashp; 288 const char *file; 289 const HASHINFO *info; 290 { 291 struct stat statbuf; 292 int nelem; 293 294 nelem = 1; 295 hashp->NKEYS = 0; 296 hashp->LORDER = BYTE_ORDER; 297 hashp->BSIZE = DEF_BUCKET_SIZE; 298 hashp->BSHIFT = DEF_BUCKET_SHIFT; 299 hashp->SGSIZE = DEF_SEGSIZE; 300 hashp->SSHIFT = DEF_SEGSIZE_SHIFT; 301 hashp->DSIZE = DEF_DIRSIZE; 302 hashp->FFACTOR = DEF_FFACTOR; 303 hashp->hash = __default_hash; 304 memset(hashp->SPARES, 0, sizeof(hashp->SPARES)); 305 memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS)); 306 307 /* Fix bucket size to be optimal for file system */ 308 if (file != NULL) { 309 if (stat(file, &statbuf)) 310 return (NULL); 311 hashp->BSIZE = statbuf.st_blksize; 312 hashp->BSHIFT = __log2(hashp->BSIZE); 313 } 314 315 if (info) { 316 if (info->bsize) { 317 /* Round pagesize up to power of 2 */ 318 hashp->BSHIFT = __log2(info->bsize); 319 hashp->BSIZE = 1 << hashp->BSHIFT; 320 if (hashp->BSIZE > MAX_BSIZE) { 321 errno = EINVAL; 322 return (NULL); 323 } 324 } 325 if (info->ffactor) 326 hashp->FFACTOR = info->ffactor; 327 if (info->hash) 328 hashp->hash = info->hash; 329 if (info->nelem) 330 nelem = info->nelem; 331 if (info->lorder) { 332 if (info->lorder != BIG_ENDIAN && 333 info->lorder != LITTLE_ENDIAN) { 334 errno = EINVAL; 335 return (NULL); 336 } 337 hashp->LORDER = info->lorder; 338 } 339 } 340 /* init_htab should destroy the table and set errno if it fails */ 341 if (init_htab(hashp, nelem)) 342 return (NULL); 343 else 344 return (hashp); 345 } 346 /* 347 * This calls alloc_segs which may run out of memory. Alloc_segs will destroy 348 * the table and set errno, so we just pass the error information along. 349 * 350 * Returns 0 on No Error 351 */ 352 static int 353 init_htab(hashp, nelem) 354 HTAB *hashp; 355 int nelem; 356 { 357 int nbuckets, nsegs; 358 int l2; 359 360 /* 361 * Divide number of elements by the fill factor and determine a 362 * desired number of buckets. Allocate space for the next greater 363 * power of two number of buckets. 364 */ 365 nelem = (nelem - 1) / hashp->FFACTOR + 1; 366 367 l2 = __log2(MAX(nelem, 2)); 368 nbuckets = 1 << l2; 369 370 hashp->SPARES[l2] = l2 + 1; 371 hashp->SPARES[l2 + 1] = l2 + 1; 372 hashp->OVFL_POINT = l2; 373 hashp->LAST_FREED = 2; 374 375 /* First bitmap page is at: splitpoint l2 page offset 1 */ 376 if (__ibitmap(hashp, OADDR_OF(l2, 1), l2 + 1, 0)) 377 return (-1); 378 379 hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1; 380 hashp->HIGH_MASK = (nbuckets << 1) - 1; 381 hashp->HDRPAGES = ((MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >> 382 hashp->BSHIFT) + 1; 383 384 nsegs = (nbuckets - 1) / hashp->SGSIZE + 1; 385 nsegs = 1 << __log2(nsegs); 386 387 if (nsegs > hashp->DSIZE) 388 hashp->DSIZE = nsegs; 389 return (alloc_segs(hashp, nsegs)); 390 } 391 392 /********************** DESTROY/CLOSE ROUTINES ************************/ 393 394 /* 395 * Flushes any changes to the file if necessary and destroys the hashp 396 * structure, freeing all allocated space. 397 */ 398 static int 399 hdestroy(hashp) 400 HTAB *hashp; 401 { 402 int i, save_errno; 403 404 save_errno = 0; 405 406 #ifdef HASH_STATISTICS 407 (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n", 408 hash_accesses, hash_collisions); 409 (void)fprintf(stderr, "hdestroy: expansions %ld\n", 410 hash_expansions); 411 (void)fprintf(stderr, "hdestroy: overflows %ld\n", 412 hash_overflows); 413 (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n", 414 hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs); 415 416 for (i = 0; i < NCACHED; i++) 417 (void)fprintf(stderr, 418 "spares[%d] = %d\n", i, hashp->SPARES[i]); 419 #endif 420 /* 421 * Call on buffer manager to free buffers, and if required, 422 * write them to disk. 423 */ 424 if (__buf_free(hashp, 1, hashp->save_file)) 425 save_errno = errno; 426 if (hashp->dir) { 427 free(*hashp->dir); /* Free initial segments */ 428 /* Free extra segments */ 429 while (hashp->exsegs--) 430 free(hashp->dir[--hashp->nsegs]); 431 free(hashp->dir); 432 } 433 if (flush_meta(hashp) && !save_errno) 434 save_errno = errno; 435 /* Free Bigmaps */ 436 for (i = 0; i < hashp->nmaps; i++) 437 if (hashp->mapp[i]) 438 free(hashp->mapp[i]); 439 440 if (hashp->fp != -1) 441 (void)_close(hashp->fp); 442 443 free(hashp); 444 445 if (save_errno) { 446 errno = save_errno; 447 return (ERROR); 448 } 449 return (SUCCESS); 450 } 451 /* 452 * Write modified pages to disk 453 * 454 * Returns: 455 * 0 == OK 456 * -1 ERROR 457 */ 458 static int 459 hash_sync(dbp, flags) 460 const DB *dbp; 461 u_int32_t flags; 462 { 463 HTAB *hashp; 464 465 if (flags != 0) { 466 errno = EINVAL; 467 return (ERROR); 468 } 469 470 if (!dbp) 471 return (ERROR); 472 473 hashp = (HTAB *)dbp->internal; 474 if (!hashp->save_file) 475 return (0); 476 if (__buf_free(hashp, 0, 1) || flush_meta(hashp)) 477 return (ERROR); 478 hashp->new_file = 0; 479 return (0); 480 } 481 482 /* 483 * Returns: 484 * 0 == OK 485 * -1 indicates that errno should be set 486 */ 487 static int 488 flush_meta(hashp) 489 HTAB *hashp; 490 { 491 HASHHDR *whdrp; 492 #if BYTE_ORDER == LITTLE_ENDIAN 493 HASHHDR whdr; 494 #endif 495 int fp, i, wsize; 496 497 if (!hashp->save_file) 498 return (0); 499 hashp->MAGIC = HASHMAGIC; 500 hashp->VERSION = HASHVERSION; 501 hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY)); 502 503 fp = hashp->fp; 504 whdrp = &hashp->hdr; 505 #if BYTE_ORDER == LITTLE_ENDIAN 506 whdrp = &whdr; 507 swap_header_copy(&hashp->hdr, whdrp); 508 #endif 509 if ((lseek(fp, (off_t)0, SEEK_SET) == -1) || 510 ((wsize = _write(fp, whdrp, sizeof(HASHHDR))) == -1)) 511 return (-1); 512 else 513 if (wsize != sizeof(HASHHDR)) { 514 errno = EFTYPE; 515 hashp->error = errno; 516 return (-1); 517 } 518 for (i = 0; i < NCACHED; i++) 519 if (hashp->mapp[i]) 520 if (__put_page(hashp, (char *)hashp->mapp[i], 521 hashp->BITMAPS[i], 0, 1)) 522 return (-1); 523 return (0); 524 } 525 526 /*******************************SEARCH ROUTINES *****************************/ 527 /* 528 * All the access routines return 529 * 530 * Returns: 531 * 0 on SUCCESS 532 * 1 to indicate an external ERROR (i.e. key not found, etc) 533 * -1 to indicate an internal ERROR (i.e. out of memory, etc) 534 */ 535 static int 536 hash_get(dbp, key, data, flag) 537 const DB *dbp; 538 const DBT *key; 539 DBT *data; 540 u_int32_t flag; 541 { 542 HTAB *hashp; 543 544 hashp = (HTAB *)dbp->internal; 545 if (flag) { 546 hashp->error = errno = EINVAL; 547 return (ERROR); 548 } 549 return (hash_access(hashp, HASH_GET, (DBT *)key, data)); 550 } 551 552 static int 553 hash_put(dbp, key, data, flag) 554 const DB *dbp; 555 DBT *key; 556 const DBT *data; 557 u_int32_t flag; 558 { 559 HTAB *hashp; 560 561 hashp = (HTAB *)dbp->internal; 562 if (flag && flag != R_NOOVERWRITE) { 563 hashp->error = EINVAL; 564 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, flag == R_NOOVERWRITE ? 572 HASH_PUTNEW : HASH_PUT, (DBT *)key, (DBT *)data)); 573 } 574 575 static int 576 hash_delete(dbp, key, flag) 577 const DB *dbp; 578 const DBT *key; 579 u_int32_t flag; /* Ignored */ 580 { 581 HTAB *hashp; 582 583 hashp = (HTAB *)dbp->internal; 584 if (flag && flag != R_CURSOR) { 585 hashp->error = errno = EINVAL; 586 return (ERROR); 587 } 588 if ((hashp->flags & O_ACCMODE) == O_RDONLY) { 589 hashp->error = errno = EPERM; 590 return (ERROR); 591 } 592 return (hash_access(hashp, HASH_DELETE, (DBT *)key, NULL)); 593 } 594 595 /* 596 * Assume that hashp has been set in wrapper routine. 597 */ 598 static int 599 hash_access(hashp, action, key, val) 600 HTAB *hashp; 601 ACTION action; 602 DBT *key, *val; 603 { 604 BUFHEAD *rbufp; 605 BUFHEAD *bufp, *save_bufp; 606 u_int16_t *bp; 607 int n, ndx, off, size; 608 char *kp; 609 u_int16_t pageno; 610 611 #ifdef HASH_STATISTICS 612 hash_accesses++; 613 #endif 614 615 off = hashp->BSIZE; 616 size = key->size; 617 kp = (char *)key->data; 618 rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0); 619 if (!rbufp) 620 return (ERROR); 621 save_bufp = rbufp; 622 623 /* Pin the bucket chain */ 624 rbufp->flags |= BUF_PIN; 625 for (bp = (u_int16_t *)rbufp->page, n = *bp++, ndx = 1; ndx < n;) 626 if (bp[1] >= REAL_KEY) { 627 /* Real key/data pair */ 628 if (size == off - *bp && 629 memcmp(kp, rbufp->page + *bp, size) == 0) 630 goto found; 631 off = bp[1]; 632 #ifdef HASH_STATISTICS 633 hash_collisions++; 634 #endif 635 bp += 2; 636 ndx += 2; 637 } else if (bp[1] == OVFLPAGE) { 638 rbufp = __get_buf(hashp, *bp, rbufp, 0); 639 if (!rbufp) { 640 save_bufp->flags &= ~BUF_PIN; 641 return (ERROR); 642 } 643 /* FOR LOOP INIT */ 644 bp = (u_int16_t *)rbufp->page; 645 n = *bp++; 646 ndx = 1; 647 off = hashp->BSIZE; 648 } else if (bp[1] < REAL_KEY) { 649 if ((ndx = 650 __find_bigpair(hashp, rbufp, ndx, kp, size)) > 0) 651 goto found; 652 if (ndx == -2) { 653 bufp = rbufp; 654 if (!(pageno = 655 __find_last_page(hashp, &bufp))) { 656 ndx = 0; 657 rbufp = bufp; 658 break; /* FOR */ 659 } 660 rbufp = __get_buf(hashp, pageno, bufp, 0); 661 if (!rbufp) { 662 save_bufp->flags &= ~BUF_PIN; 663 return (ERROR); 664 } 665 /* FOR LOOP INIT */ 666 bp = (u_int16_t *)rbufp->page; 667 n = *bp++; 668 ndx = 1; 669 off = hashp->BSIZE; 670 } else { 671 save_bufp->flags &= ~BUF_PIN; 672 return (ERROR); 673 } 674 } 675 676 /* Not found */ 677 switch (action) { 678 case HASH_PUT: 679 case HASH_PUTNEW: 680 if (__addel(hashp, rbufp, key, val)) { 681 save_bufp->flags &= ~BUF_PIN; 682 return (ERROR); 683 } else { 684 save_bufp->flags &= ~BUF_PIN; 685 return (SUCCESS); 686 } 687 case HASH_GET: 688 case HASH_DELETE: 689 default: 690 save_bufp->flags &= ~BUF_PIN; 691 return (ABNORMAL); 692 } 693 694 found: 695 switch (action) { 696 case HASH_PUTNEW: 697 save_bufp->flags &= ~BUF_PIN; 698 return (ABNORMAL); 699 case HASH_GET: 700 bp = (u_int16_t *)rbufp->page; 701 if (bp[ndx + 1] < REAL_KEY) { 702 if (__big_return(hashp, rbufp, ndx, val, 0)) 703 return (ERROR); 704 } else { 705 val->data = (u_char *)rbufp->page + (int)bp[ndx + 1]; 706 val->size = bp[ndx] - bp[ndx + 1]; 707 } 708 break; 709 case HASH_PUT: 710 if ((__delpair(hashp, rbufp, ndx)) || 711 (__addel(hashp, rbufp, key, val))) { 712 save_bufp->flags &= ~BUF_PIN; 713 return (ERROR); 714 } 715 break; 716 case HASH_DELETE: 717 if (__delpair(hashp, rbufp, ndx)) 718 return (ERROR); 719 break; 720 default: 721 abort(); 722 } 723 save_bufp->flags &= ~BUF_PIN; 724 return (SUCCESS); 725 } 726 727 static int 728 hash_seq(dbp, key, data, flag) 729 const DB *dbp; 730 DBT *key, *data; 731 u_int32_t flag; 732 { 733 u_int32_t bucket; 734 BUFHEAD *bufp; 735 HTAB *hashp; 736 u_int16_t *bp, ndx; 737 738 hashp = (HTAB *)dbp->internal; 739 if (flag && flag != R_FIRST && flag != R_NEXT) { 740 hashp->error = errno = EINVAL; 741 return (ERROR); 742 } 743 #ifdef HASH_STATISTICS 744 hash_accesses++; 745 #endif 746 if ((hashp->cbucket < 0) || (flag == R_FIRST)) { 747 hashp->cbucket = 0; 748 hashp->cndx = 1; 749 hashp->cpage = NULL; 750 } 751 752 for (bp = NULL; !bp || !bp[0]; ) { 753 if (!(bufp = hashp->cpage)) { 754 for (bucket = hashp->cbucket; 755 bucket <= hashp->MAX_BUCKET; 756 bucket++, hashp->cndx = 1) { 757 bufp = __get_buf(hashp, bucket, NULL, 0); 758 if (!bufp) 759 return (ERROR); 760 hashp->cpage = bufp; 761 bp = (u_int16_t *)bufp->page; 762 if (bp[0]) 763 break; 764 } 765 hashp->cbucket = bucket; 766 if (hashp->cbucket > hashp->MAX_BUCKET) { 767 hashp->cbucket = -1; 768 return (ABNORMAL); 769 } 770 } else 771 bp = (u_int16_t *)hashp->cpage->page; 772 773 #ifdef DEBUG 774 assert(bp); 775 assert(bufp); 776 #endif 777 while (bp[hashp->cndx + 1] == OVFLPAGE) { 778 bufp = hashp->cpage = 779 __get_buf(hashp, bp[hashp->cndx], bufp, 0); 780 if (!bufp) 781 return (ERROR); 782 bp = (u_int16_t *)(bufp->page); 783 hashp->cndx = 1; 784 } 785 if (!bp[0]) { 786 hashp->cpage = NULL; 787 ++hashp->cbucket; 788 } 789 } 790 ndx = hashp->cndx; 791 if (bp[ndx + 1] < REAL_KEY) { 792 if (__big_keydata(hashp, bufp, key, data, 1)) 793 return (ERROR); 794 } else { 795 key->data = (u_char *)hashp->cpage->page + bp[ndx]; 796 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->BSIZE) - bp[ndx]; 797 data->data = (u_char *)hashp->cpage->page + bp[ndx + 1]; 798 data->size = bp[ndx] - bp[ndx + 1]; 799 ndx += 2; 800 if (ndx > bp[0]) { 801 hashp->cpage = NULL; 802 hashp->cbucket++; 803 hashp->cndx = 1; 804 } else 805 hashp->cndx = ndx; 806 } 807 return (SUCCESS); 808 } 809 810 /********************************* UTILITIES ************************/ 811 812 /* 813 * Returns: 814 * 0 ==> OK 815 * -1 ==> Error 816 */ 817 extern int 818 __expand_table(hashp) 819 HTAB *hashp; 820 { 821 u_int32_t old_bucket, new_bucket; 822 int dirsize, new_segnum, spare_ndx; 823 824 #ifdef HASH_STATISTICS 825 hash_expansions++; 826 #endif 827 new_bucket = ++hashp->MAX_BUCKET; 828 old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK); 829 830 new_segnum = new_bucket >> hashp->SSHIFT; 831 832 /* Check if we need a new segment */ 833 if (new_segnum >= hashp->nsegs) { 834 /* Check if we need to expand directory */ 835 if (new_segnum >= hashp->DSIZE) { 836 /* Reallocate directory */ 837 dirsize = hashp->DSIZE * sizeof(SEGMENT *); 838 if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1)) 839 return (-1); 840 hashp->DSIZE = dirsize << 1; 841 } 842 if ((hashp->dir[new_segnum] = 843 (SEGMENT)calloc(hashp->SGSIZE, sizeof(SEGMENT))) == NULL) 844 return (-1); 845 hashp->exsegs++; 846 hashp->nsegs++; 847 } 848 /* 849 * If the split point is increasing (MAX_BUCKET's log base 2 850 * * increases), we need to copy the current contents of the spare 851 * split bucket to the next bucket. 852 */ 853 spare_ndx = __log2(hashp->MAX_BUCKET + 1); 854 if (spare_ndx > hashp->OVFL_POINT) { 855 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT]; 856 hashp->OVFL_POINT = spare_ndx; 857 } 858 859 if (new_bucket > hashp->HIGH_MASK) { 860 /* Starting a new doubling */ 861 hashp->LOW_MASK = hashp->HIGH_MASK; 862 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK; 863 } 864 /* Relocate records to the new bucket */ 865 return (__split_page(hashp, old_bucket, new_bucket)); 866 } 867 868 /* 869 * If realloc guarantees that the pointer is not destroyed if the realloc 870 * fails, then this routine can go away. 871 */ 872 static void * 873 hash_realloc(p_ptr, oldsize, newsize) 874 SEGMENT **p_ptr; 875 int oldsize, newsize; 876 { 877 void *p; 878 879 if ( (p = malloc(newsize)) ) { 880 memmove(p, *p_ptr, oldsize); 881 memset((char *)p + oldsize, 0, newsize - oldsize); 882 free(*p_ptr); 883 *p_ptr = p; 884 } 885 return (p); 886 } 887 888 extern u_int32_t 889 __call_hash(hashp, k, len) 890 HTAB *hashp; 891 char *k; 892 int len; 893 { 894 int n, bucket; 895 896 n = hashp->hash(k, len); 897 bucket = n & hashp->HIGH_MASK; 898 if (bucket > hashp->MAX_BUCKET) 899 bucket = bucket & hashp->LOW_MASK; 900 return (bucket); 901 } 902 903 /* 904 * Allocate segment table. On error, destroy the table and set errno. 905 * 906 * Returns 0 on success 907 */ 908 static int 909 alloc_segs(hashp, nsegs) 910 HTAB *hashp; 911 int nsegs; 912 { 913 int i; 914 SEGMENT store; 915 916 int save_errno; 917 918 if ((hashp->dir = 919 (SEGMENT *)calloc(hashp->DSIZE, sizeof(SEGMENT *))) == NULL) { 920 save_errno = errno; 921 (void)hdestroy(hashp); 922 errno = save_errno; 923 return (-1); 924 } 925 /* Allocate segments */ 926 if ((store = 927 (SEGMENT)calloc(nsegs << hashp->SSHIFT, sizeof(SEGMENT))) == NULL) { 928 save_errno = errno; 929 (void)hdestroy(hashp); 930 errno = save_errno; 931 return (-1); 932 } 933 for (i = 0; i < nsegs; i++, hashp->nsegs++) 934 hashp->dir[i] = &store[i << hashp->SSHIFT]; 935 return (0); 936 } 937 938 #if BYTE_ORDER == LITTLE_ENDIAN 939 /* 940 * Hashp->hdr needs to be byteswapped. 941 */ 942 static void 943 swap_header_copy(srcp, destp) 944 HASHHDR *srcp, *destp; 945 { 946 int i; 947 948 P_32_COPY(srcp->magic, destp->magic); 949 P_32_COPY(srcp->version, destp->version); 950 P_32_COPY(srcp->lorder, destp->lorder); 951 P_32_COPY(srcp->bsize, destp->bsize); 952 P_32_COPY(srcp->bshift, destp->bshift); 953 P_32_COPY(srcp->dsize, destp->dsize); 954 P_32_COPY(srcp->ssize, destp->ssize); 955 P_32_COPY(srcp->sshift, destp->sshift); 956 P_32_COPY(srcp->ovfl_point, destp->ovfl_point); 957 P_32_COPY(srcp->last_freed, destp->last_freed); 958 P_32_COPY(srcp->max_bucket, destp->max_bucket); 959 P_32_COPY(srcp->high_mask, destp->high_mask); 960 P_32_COPY(srcp->low_mask, destp->low_mask); 961 P_32_COPY(srcp->ffactor, destp->ffactor); 962 P_32_COPY(srcp->nkeys, destp->nkeys); 963 P_32_COPY(srcp->hdrpages, destp->hdrpages); 964 P_32_COPY(srcp->h_charkey, destp->h_charkey); 965 for (i = 0; i < NCACHED; i++) { 966 P_32_COPY(srcp->spares[i], destp->spares[i]); 967 P_16_COPY(srcp->bitmaps[i], destp->bitmaps[i]); 968 } 969 } 970 971 static void 972 swap_header(hashp) 973 HTAB *hashp; 974 { 975 HASHHDR *hdrp; 976 int i; 977 978 hdrp = &hashp->hdr; 979 980 M_32_SWAP(hdrp->magic); 981 M_32_SWAP(hdrp->version); 982 M_32_SWAP(hdrp->lorder); 983 M_32_SWAP(hdrp->bsize); 984 M_32_SWAP(hdrp->bshift); 985 M_32_SWAP(hdrp->dsize); 986 M_32_SWAP(hdrp->ssize); 987 M_32_SWAP(hdrp->sshift); 988 M_32_SWAP(hdrp->ovfl_point); 989 M_32_SWAP(hdrp->last_freed); 990 M_32_SWAP(hdrp->max_bucket); 991 M_32_SWAP(hdrp->high_mask); 992 M_32_SWAP(hdrp->low_mask); 993 M_32_SWAP(hdrp->ffactor); 994 M_32_SWAP(hdrp->nkeys); 995 M_32_SWAP(hdrp->hdrpages); 996 M_32_SWAP(hdrp->h_charkey); 997 for (i = 0; i < NCACHED; i++) { 998 M_32_SWAP(hdrp->spares[i]); 999 M_16_SWAP(hdrp->bitmaps[i]); 1000 } 1001 } 1002 #endif 1003