1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 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 * Margo Seltzer. 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 #include "namespace.h" 36 #include <sys/param.h> 37 #include <sys/stat.h> 38 39 #include <errno.h> 40 #include <fcntl.h> 41 #include <stdio.h> 42 #include <stdlib.h> 43 #include <string.h> 44 #include <unistd.h> 45 #ifdef DEBUG 46 #include <assert.h> 47 #endif 48 #include "un-namespace.h" 49 50 #include <db.h> 51 #include "hash.h" 52 #include "page.h" 53 #include "extern.h" 54 55 static int alloc_segs(HTAB *, int); 56 static int flush_meta(HTAB *); 57 static int hash_access(HTAB *, ACTION, DBT *, DBT *); 58 static int hash_close(DB *); 59 static int hash_delete(const DB *, const DBT *, u_int32_t); 60 static int hash_fd(const DB *); 61 static int hash_get(const DB *, const DBT *, DBT *, u_int32_t); 62 static int hash_put(const DB *, DBT *, const DBT *, u_int32_t); 63 static void *hash_realloc(SEGMENT **, int, int); 64 static int hash_seq(const DB *, DBT *, DBT *, u_int32_t); 65 static int hash_sync(const DB *, u_int32_t); 66 static int hdestroy(HTAB *); 67 static HTAB *init_hash(HTAB *, const char *, const HASHINFO *); 68 static int init_htab(HTAB *, int); 69 #if BYTE_ORDER == LITTLE_ENDIAN 70 static void swap_header(HTAB *); 71 static void swap_header_copy(HASHHDR *, HASHHDR *); 72 #endif 73 74 /* Fast arithmetic, relying on powers of 2, */ 75 #define MOD(x, y) ((x) & ((y) - 1)) 76 77 #define RETURN_ERROR(ERR, LOC) { save_errno = ERR; goto LOC; } 78 79 /* Return values */ 80 #define SUCCESS (0) 81 #define ERROR (-1) 82 #define ABNORMAL (1) 83 84 /* Cursor status */ 85 #define CURSOR_NOT_SET -1 86 #define CURSOR_OVERFLOW -2 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 /* ARGSUSED */ 96 DB * 97 __hash_open(const char *file, int flags, int mode, 98 const HASHINFO *info, /* Special directives for create */ 99 int dflags) 100 { 101 HTAB *hashp; 102 struct stat statbuf; 103 DB *dbp; 104 int bpages, hdrsize, new_table, nsegs, save_errno; 105 106 if (!(hashp = (HTAB *)calloc(1, sizeof(HTAB)))) 107 return (NULL); 108 hashp->fp = -1; 109 110 /* 111 * Even if user wants write only, we need to be able to read 112 * the actual file, so we need to open it read/write. But, the 113 * field in the hashp structure needs to be accurate so that 114 * we can check accesses. 115 */ 116 hashp->flags = flags; 117 if ((flags & O_ACCMODE) == O_WRONLY) { 118 flags &= ~O_WRONLY; 119 flags |= O_RDWR; 120 } 121 122 if (file) { 123 if ((hashp->fp = _open(file, flags | O_CLOEXEC, mode)) == -1) 124 RETURN_ERROR(errno, error0); 125 new_table = _fstat(hashp->fp, &statbuf) == 0 && 126 statbuf.st_size == 0 && 127 ((flags & O_ACCMODE) != O_RDONLY || (flags & O_CREAT) != 0); 128 } else 129 new_table = 1; 130 131 if (new_table) { 132 if (!(hashp = init_hash(hashp, file, info))) 133 RETURN_ERROR(errno, error1); 134 } else { 135 /* Table already exists */ 136 if (info && info->hash) 137 hashp->hash = info->hash; 138 else 139 hashp->hash = __default_hash; 140 141 hdrsize = _read(hashp->fp, &hashp->hdr, sizeof(HASHHDR)); 142 #if BYTE_ORDER == LITTLE_ENDIAN 143 swap_header(hashp); 144 #endif 145 if (hdrsize == -1) 146 RETURN_ERROR(errno, error1); 147 if (hdrsize != sizeof(HASHHDR)) 148 RETURN_ERROR(EFTYPE, error1); 149 /* Verify file type, versions and hash function */ 150 if (hashp->MAGIC != HASHMAGIC) 151 RETURN_ERROR(EFTYPE, error1); 152 #define OLDHASHVERSION 1 153 if (hashp->VERSION != HASHVERSION && 154 hashp->VERSION != OLDHASHVERSION) 155 RETURN_ERROR(EFTYPE, error1); 156 if ((int32_t)hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY) 157 RETURN_ERROR(EFTYPE, error1); 158 /* 159 * Figure out how many segments we need. Max_Bucket is the 160 * maximum bucket number, so the number of buckets is 161 * max_bucket + 1. 162 */ 163 nsegs = howmany(hashp->MAX_BUCKET + 1, hashp->SGSIZE); 164 if (alloc_segs(hashp, nsegs)) 165 /* 166 * If alloc_segs fails, table will have been destroyed 167 * and errno will have been set. 168 */ 169 return (NULL); 170 /* Read in bitmaps */ 171 bpages = (hashp->SPARES[hashp->OVFL_POINT] + 172 (hashp->BSIZE << BYTE_SHIFT) - 1) >> 173 (hashp->BSHIFT + BYTE_SHIFT); 174 175 hashp->nmaps = bpages; 176 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(u_int32_t *)); 177 } 178 179 /* Initialize Buffer Manager */ 180 if (info && info->cachesize) 181 __buf_init(hashp, info->cachesize); 182 else 183 __buf_init(hashp, DEF_BUFSIZE); 184 185 hashp->new_file = new_table; 186 hashp->save_file = file && (flags & O_RDWR); 187 hashp->cbucket = CURSOR_NOT_SET; 188 if (!(dbp = (DB *)malloc(sizeof(DB)))) { 189 save_errno = errno; 190 hdestroy(hashp); 191 errno = save_errno; 192 return (NULL); 193 } 194 dbp->internal = hashp; 195 dbp->close = hash_close; 196 dbp->del = hash_delete; 197 dbp->fd = hash_fd; 198 dbp->get = hash_get; 199 dbp->put = hash_put; 200 dbp->seq = hash_seq; 201 dbp->sync = hash_sync; 202 dbp->type = DB_HASH; 203 204 #ifdef DEBUG 205 (void)fprintf(stderr, 206 "%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", 207 "init_htab:", 208 "TABLE POINTER ", hashp, 209 "BUCKET SIZE ", hashp->BSIZE, 210 "BUCKET SHIFT ", hashp->BSHIFT, 211 "DIRECTORY SIZE ", hashp->DSIZE, 212 "SEGMENT SIZE ", hashp->SGSIZE, 213 "SEGMENT SHIFT ", hashp->SSHIFT, 214 "FILL FACTOR ", hashp->FFACTOR, 215 "MAX BUCKET ", hashp->MAX_BUCKET, 216 "OVFL POINT ", hashp->OVFL_POINT, 217 "LAST FREED ", hashp->LAST_FREED, 218 "HIGH MASK ", hashp->HIGH_MASK, 219 "LOW MASK ", hashp->LOW_MASK, 220 "NSEGS ", hashp->nsegs, 221 "NKEYS ", hashp->NKEYS); 222 #endif 223 #ifdef HASH_STATISTICS 224 hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0; 225 #endif 226 return (dbp); 227 228 error1: 229 if (hashp != NULL) 230 (void)_close(hashp->fp); 231 232 error0: 233 free(hashp); 234 errno = save_errno; 235 return (NULL); 236 } 237 238 static int 239 hash_close(DB *dbp) 240 { 241 HTAB *hashp; 242 int retval; 243 244 if (!dbp) 245 return (ERROR); 246 247 hashp = (HTAB *)dbp->internal; 248 retval = hdestroy(hashp); 249 free(dbp); 250 return (retval); 251 } 252 253 static int 254 hash_fd(const DB *dbp) 255 { 256 HTAB *hashp; 257 258 if (!dbp) 259 return (ERROR); 260 261 hashp = (HTAB *)dbp->internal; 262 if (hashp->fp == -1) { 263 errno = ENOENT; 264 return (-1); 265 } 266 return (hashp->fp); 267 } 268 269 /************************** LOCAL CREATION ROUTINES **********************/ 270 static HTAB * 271 init_hash(HTAB *hashp, const char *file, const HASHINFO *info) 272 { 273 struct stat statbuf; 274 int nelem; 275 276 nelem = 1; 277 hashp->NKEYS = 0; 278 hashp->LORDER = BYTE_ORDER; 279 hashp->BSIZE = DEF_BUCKET_SIZE; 280 hashp->BSHIFT = DEF_BUCKET_SHIFT; 281 hashp->SGSIZE = DEF_SEGSIZE; 282 hashp->SSHIFT = DEF_SEGSIZE_SHIFT; 283 hashp->DSIZE = DEF_DIRSIZE; 284 hashp->FFACTOR = DEF_FFACTOR; 285 hashp->hash = __default_hash; 286 memset(hashp->SPARES, 0, sizeof(hashp->SPARES)); 287 memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS)); 288 289 /* Fix bucket size to be optimal for file system */ 290 if (file != NULL) { 291 if (stat(file, &statbuf)) 292 return (NULL); 293 hashp->BSIZE = statbuf.st_blksize; 294 if (hashp->BSIZE > MAX_BSIZE) 295 hashp->BSIZE = MAX_BSIZE; 296 hashp->BSHIFT = __log2(hashp->BSIZE); 297 } 298 299 if (info) { 300 if (info->bsize) { 301 /* Round pagesize up to power of 2 */ 302 hashp->BSHIFT = __log2(info->bsize); 303 hashp->BSIZE = 1 << hashp->BSHIFT; 304 if (hashp->BSIZE > MAX_BSIZE) { 305 errno = EINVAL; 306 return (NULL); 307 } 308 } 309 if (info->ffactor) 310 hashp->FFACTOR = info->ffactor; 311 if (info->hash) 312 hashp->hash = info->hash; 313 if (info->nelem) 314 nelem = info->nelem; 315 if (info->lorder) { 316 if (info->lorder != BIG_ENDIAN && 317 info->lorder != LITTLE_ENDIAN) { 318 errno = EINVAL; 319 return (NULL); 320 } 321 hashp->LORDER = info->lorder; 322 } 323 } 324 /* init_htab should destroy the table and set errno if it fails */ 325 if (init_htab(hashp, nelem)) 326 return (NULL); 327 else 328 return (hashp); 329 } 330 /* 331 * This calls alloc_segs which may run out of memory. Alloc_segs will destroy 332 * the table and set errno, so we just pass the error information along. 333 * 334 * Returns 0 on No Error 335 */ 336 static int 337 init_htab(HTAB *hashp, int nelem) 338 { 339 int nbuckets, nsegs, l2; 340 341 /* 342 * Divide number of elements by the fill factor and determine a 343 * desired number of buckets. Allocate space for the next greater 344 * power of two number of buckets. 345 */ 346 nelem = (nelem - 1) / hashp->FFACTOR + 1; 347 348 l2 = __log2(MAX(nelem, 2)); 349 nbuckets = 1 << l2; 350 351 hashp->SPARES[l2] = l2 + 1; 352 hashp->SPARES[l2 + 1] = l2 + 1; 353 hashp->OVFL_POINT = l2; 354 hashp->LAST_FREED = 2; 355 356 /* First bitmap page is at: splitpoint l2 page offset 1 */ 357 if (__ibitmap(hashp, OADDR_OF(l2, 1), l2 + 1, 0)) 358 return (-1); 359 360 hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1; 361 hashp->HIGH_MASK = (nbuckets << 1) - 1; 362 hashp->HDRPAGES = ((MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >> 363 hashp->BSHIFT) + 1; 364 365 nsegs = (nbuckets - 1) / hashp->SGSIZE + 1; 366 nsegs = 1 << __log2(nsegs); 367 368 if (nsegs > hashp->DSIZE) 369 hashp->DSIZE = nsegs; 370 return (alloc_segs(hashp, nsegs)); 371 } 372 373 /********************** DESTROY/CLOSE ROUTINES ************************/ 374 375 /* 376 * Flushes any changes to the file if necessary and destroys the hashp 377 * structure, freeing all allocated space. 378 */ 379 static int 380 hdestroy(HTAB *hashp) 381 { 382 int i, save_errno; 383 384 save_errno = 0; 385 386 #ifdef HASH_STATISTICS 387 (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n", 388 hash_accesses, hash_collisions); 389 (void)fprintf(stderr, "hdestroy: expansions %ld\n", 390 hash_expansions); 391 (void)fprintf(stderr, "hdestroy: overflows %ld\n", 392 hash_overflows); 393 (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n", 394 hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs); 395 396 for (i = 0; i < NCACHED; i++) 397 (void)fprintf(stderr, 398 "spares[%d] = %d\n", i, hashp->SPARES[i]); 399 #endif 400 /* 401 * Call on buffer manager to free buffers, and if required, 402 * write them to disk. 403 */ 404 if (__buf_free(hashp, 1, hashp->save_file)) 405 save_errno = errno; 406 if (hashp->dir) { 407 free(*hashp->dir); /* Free initial segments */ 408 /* Free extra segments */ 409 while (hashp->exsegs--) 410 free(hashp->dir[--hashp->nsegs]); 411 free(hashp->dir); 412 } 413 if (flush_meta(hashp) && !save_errno) 414 save_errno = errno; 415 /* Free Bigmaps */ 416 for (i = 0; i < hashp->nmaps; i++) 417 if (hashp->mapp[i]) 418 free(hashp->mapp[i]); 419 if (hashp->tmp_key) 420 free(hashp->tmp_key); 421 if (hashp->tmp_buf) 422 free(hashp->tmp_buf); 423 424 if (hashp->fp != -1) { 425 if (hashp->save_file) 426 (void)_fsync(hashp->fp); 427 (void)_close(hashp->fp); 428 } 429 430 free(hashp); 431 432 if (save_errno) { 433 errno = save_errno; 434 return (ERROR); 435 } 436 return (SUCCESS); 437 } 438 /* 439 * Write modified pages to disk 440 * 441 * Returns: 442 * 0 == OK 443 * -1 ERROR 444 */ 445 static int 446 hash_sync(const DB *dbp, u_int32_t flags) 447 { 448 HTAB *hashp; 449 450 if (flags != 0) { 451 errno = EINVAL; 452 return (ERROR); 453 } 454 455 if (!dbp) 456 return (ERROR); 457 458 hashp = (HTAB *)dbp->internal; 459 if (!hashp->save_file) 460 return (0); 461 if (__buf_free(hashp, 0, 1) || flush_meta(hashp)) 462 return (ERROR); 463 if (hashp->fp != -1 && _fsync(hashp->fp) != 0) 464 return (ERROR); 465 hashp->new_file = 0; 466 return (0); 467 } 468 469 /* 470 * Returns: 471 * 0 == OK 472 * -1 indicates that errno should be set 473 */ 474 static int 475 flush_meta(HTAB *hashp) 476 { 477 HASHHDR *whdrp; 478 #if BYTE_ORDER == LITTLE_ENDIAN 479 HASHHDR whdr; 480 #endif 481 int fp, i, wsize; 482 483 if (!hashp->save_file) 484 return (0); 485 hashp->MAGIC = HASHMAGIC; 486 hashp->VERSION = HASHVERSION; 487 hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY)); 488 489 fp = hashp->fp; 490 whdrp = &hashp->hdr; 491 #if BYTE_ORDER == LITTLE_ENDIAN 492 whdrp = &whdr; 493 swap_header_copy(&hashp->hdr, whdrp); 494 #endif 495 if ((wsize = pwrite(fp, whdrp, sizeof(HASHHDR), (off_t)0)) == -1) 496 return (-1); 497 else 498 if (wsize != sizeof(HASHHDR)) { 499 errno = EFTYPE; 500 hashp->error = errno; 501 return (-1); 502 } 503 for (i = 0; i < NCACHED; i++) 504 if (hashp->mapp[i]) 505 if (__put_page(hashp, (char *)hashp->mapp[i], 506 hashp->BITMAPS[i], 0, 1)) 507 return (-1); 508 return (0); 509 } 510 511 /*******************************SEARCH ROUTINES *****************************/ 512 /* 513 * All the access routines return 514 * 515 * Returns: 516 * 0 on SUCCESS 517 * 1 to indicate an external ERROR (i.e. key not found, etc) 518 * -1 to indicate an internal ERROR (i.e. out of memory, etc) 519 */ 520 static int 521 hash_get(const DB *dbp, const DBT *key, DBT *data, u_int32_t flag) 522 { 523 HTAB *hashp; 524 525 hashp = (HTAB *)dbp->internal; 526 if (flag) { 527 hashp->error = errno = EINVAL; 528 return (ERROR); 529 } 530 if ((hashp->flags & O_ACCMODE) == O_WRONLY) { 531 hashp->error = errno = EPERM; 532 return (ERROR); 533 } 534 return (hash_access(hashp, HASH_GET, (DBT *)key, data)); 535 } 536 537 static int 538 hash_put(const DB *dbp, DBT *key, const DBT *data, u_int32_t flag) 539 { 540 HTAB *hashp; 541 542 hashp = (HTAB *)dbp->internal; 543 if (flag && flag != R_NOOVERWRITE) { 544 hashp->error = errno = EINVAL; 545 return (ERROR); 546 } 547 if ((hashp->flags & O_ACCMODE) == O_RDONLY) { 548 hashp->error = errno = EPERM; 549 return (ERROR); 550 } 551 return (hash_access(hashp, flag == R_NOOVERWRITE ? 552 HASH_PUTNEW : HASH_PUT, (DBT *)key, (DBT *)data)); 553 } 554 555 static int 556 hash_delete(const DB *dbp, const DBT *key, 557 u_int32_t flag) /* Ignored */ 558 { 559 HTAB *hashp; 560 561 hashp = (HTAB *)dbp->internal; 562 if (flag && flag != R_CURSOR) { 563 hashp->error = errno = EINVAL; 564 return (ERROR); 565 } 566 if ((hashp->flags & O_ACCMODE) == O_RDONLY) { 567 hashp->error = errno = EPERM; 568 return (ERROR); 569 } 570 return (hash_access(hashp, HASH_DELETE, (DBT *)key, NULL)); 571 } 572 573 /* 574 * Assume that hashp has been set in wrapper routine. 575 */ 576 static int 577 hash_access(HTAB *hashp, ACTION action, DBT *key, DBT *val) 578 { 579 BUFHEAD *rbufp; 580 BUFHEAD *bufp, *save_bufp; 581 u_int16_t *bp; 582 int n, ndx, off, size; 583 char *kp; 584 u_int16_t pageno; 585 586 #ifdef HASH_STATISTICS 587 hash_accesses++; 588 #endif 589 590 off = hashp->BSIZE; 591 size = key->size; 592 kp = (char *)key->data; 593 rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0); 594 if (!rbufp) 595 return (ERROR); 596 save_bufp = rbufp; 597 598 /* Pin the bucket chain */ 599 rbufp->flags |= BUF_PIN; 600 for (bp = (u_int16_t *)rbufp->page, n = *bp++, ndx = 1; ndx < n;) 601 if (bp[1] >= REAL_KEY) { 602 /* Real key/data pair */ 603 if (size == off - *bp && 604 memcmp(kp, rbufp->page + *bp, size) == 0) 605 goto found; 606 off = bp[1]; 607 #ifdef HASH_STATISTICS 608 hash_collisions++; 609 #endif 610 bp += 2; 611 ndx += 2; 612 } else if (bp[1] == OVFLPAGE) { 613 rbufp = __get_buf(hashp, *bp, rbufp, 0); 614 if (!rbufp) { 615 save_bufp->flags &= ~BUF_PIN; 616 return (ERROR); 617 } 618 /* FOR LOOP INIT */ 619 bp = (u_int16_t *)rbufp->page; 620 n = *bp++; 621 ndx = 1; 622 off = hashp->BSIZE; 623 } else if (bp[1] < REAL_KEY) { 624 if ((ndx = 625 __find_bigpair(hashp, rbufp, ndx, kp, size)) > 0) 626 goto found; 627 if (ndx == -2) { 628 bufp = rbufp; 629 if (!(pageno = 630 __find_last_page(hashp, &bufp))) { 631 ndx = 0; 632 rbufp = bufp; 633 break; /* FOR */ 634 } 635 rbufp = __get_buf(hashp, pageno, bufp, 0); 636 if (!rbufp) { 637 save_bufp->flags &= ~BUF_PIN; 638 return (ERROR); 639 } 640 /* FOR LOOP INIT */ 641 bp = (u_int16_t *)rbufp->page; 642 n = *bp++; 643 ndx = 1; 644 off = hashp->BSIZE; 645 } else { 646 save_bufp->flags &= ~BUF_PIN; 647 return (ERROR); 648 } 649 } 650 651 /* Not found */ 652 switch (action) { 653 case HASH_PUT: 654 case HASH_PUTNEW: 655 if (__addel(hashp, rbufp, key, val)) { 656 save_bufp->flags &= ~BUF_PIN; 657 return (ERROR); 658 } else { 659 save_bufp->flags &= ~BUF_PIN; 660 return (SUCCESS); 661 } 662 case HASH_GET: 663 case HASH_DELETE: 664 default: 665 save_bufp->flags &= ~BUF_PIN; 666 return (ABNORMAL); 667 } 668 669 found: 670 switch (action) { 671 case HASH_PUTNEW: 672 save_bufp->flags &= ~BUF_PIN; 673 return (ABNORMAL); 674 case HASH_GET: 675 bp = (u_int16_t *)rbufp->page; 676 if (bp[ndx + 1] < REAL_KEY) { 677 if (__big_return(hashp, rbufp, ndx, val, 0)) 678 return (ERROR); 679 } else { 680 val->data = (u_char *)rbufp->page + (int)bp[ndx + 1]; 681 val->size = bp[ndx] - bp[ndx + 1]; 682 } 683 break; 684 case HASH_PUT: 685 if ((__delpair(hashp, rbufp, ndx)) || 686 (__addel(hashp, rbufp, key, val))) { 687 save_bufp->flags &= ~BUF_PIN; 688 return (ERROR); 689 } 690 break; 691 case HASH_DELETE: 692 if (__delpair(hashp, rbufp, ndx)) 693 return (ERROR); 694 break; 695 default: 696 abort(); 697 } 698 save_bufp->flags &= ~BUF_PIN; 699 return (SUCCESS); 700 } 701 702 static int 703 hash_seq(const DB *dbp, DBT *key, DBT *data, u_int32_t flag) 704 { 705 u_int32_t bucket; 706 BUFHEAD *bufp; 707 HTAB *hashp; 708 u_int16_t *bp, ndx; 709 710 hashp = (HTAB *)dbp->internal; 711 if (flag != R_FIRST && flag != R_NEXT) { 712 hashp->error = errno = EINVAL; 713 return (ERROR); 714 } 715 #ifdef HASH_STATISTICS 716 hash_accesses++; 717 #endif 718 if (flag == R_FIRST || hashp->cbucket == CURSOR_NOT_SET) { 719 hashp->cbucket = 0; 720 hashp->cndx = 1; 721 hashp->cpage = NULL; 722 } else if (hashp->cbucket <= CURSOR_OVERFLOW) { 723 return (ABNORMAL); 724 } 725 next_bucket: 726 for (bp = NULL; !bp || !bp[0]; ) { 727 if (!(bufp = hashp->cpage)) { 728 for (bucket = hashp->cbucket; 729 bucket <= hashp->MAX_BUCKET; 730 bucket++, hashp->cndx = 1) { 731 bufp = __get_buf(hashp, bucket, NULL, 0); 732 if (!bufp) 733 return (ERROR); 734 hashp->cpage = bufp; 735 bp = (u_int16_t *)bufp->page; 736 if (bp[0]) 737 break; 738 } 739 hashp->cbucket = bucket; 740 if ((u_int32_t)hashp->cbucket > hashp->MAX_BUCKET) { 741 hashp->cbucket = CURSOR_OVERFLOW; 742 return (ABNORMAL); 743 } 744 } else { 745 bp = (u_int16_t *)hashp->cpage->page; 746 if (flag == R_NEXT || flag == 0) { 747 hashp->cndx += 2; 748 if (hashp->cndx > bp[0]) { 749 hashp->cpage = NULL; 750 hashp->cbucket++; 751 hashp->cndx = 1; 752 goto next_bucket; 753 } 754 } 755 } 756 757 #ifdef DEBUG 758 assert(bp); 759 assert(bufp); 760 #endif 761 while (bp[hashp->cndx + 1] == OVFLPAGE) { 762 bufp = hashp->cpage = 763 __get_buf(hashp, bp[hashp->cndx], bufp, 0); 764 if (!bufp) 765 return (ERROR); 766 bp = (u_int16_t *)(bufp->page); 767 hashp->cndx = 1; 768 } 769 if (!bp[0]) { 770 hashp->cpage = NULL; 771 ++hashp->cbucket; 772 } 773 } 774 ndx = hashp->cndx; 775 if (bp[ndx + 1] < REAL_KEY) { 776 if (__big_keydata(hashp, bufp, key, data, 1)) 777 return (ERROR); 778 } else { 779 if (hashp->cpage == NULL) 780 return (ERROR); 781 key->data = (u_char *)hashp->cpage->page + bp[ndx]; 782 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->BSIZE) - bp[ndx]; 783 data->data = (u_char *)hashp->cpage->page + bp[ndx + 1]; 784 data->size = bp[ndx] - bp[ndx + 1]; 785 } 786 return (SUCCESS); 787 } 788 789 /********************************* UTILITIES ************************/ 790 791 /* 792 * Returns: 793 * 0 ==> OK 794 * -1 ==> Error 795 */ 796 int 797 __expand_table(HTAB *hashp) 798 { 799 u_int32_t old_bucket, new_bucket; 800 int dirsize, new_segnum, spare_ndx; 801 802 #ifdef HASH_STATISTICS 803 hash_expansions++; 804 #endif 805 new_bucket = ++hashp->MAX_BUCKET; 806 old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK); 807 808 new_segnum = new_bucket >> hashp->SSHIFT; 809 810 /* Check if we need a new segment */ 811 if (new_segnum >= hashp->nsegs) { 812 /* Check if we need to expand directory */ 813 if (new_segnum >= hashp->DSIZE) { 814 /* Reallocate directory */ 815 dirsize = hashp->DSIZE * sizeof(SEGMENT *); 816 if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1)) 817 return (-1); 818 hashp->DSIZE = dirsize << 1; 819 } 820 if ((hashp->dir[new_segnum] = 821 calloc(hashp->SGSIZE, sizeof(SEGMENT))) == NULL) 822 return (-1); 823 hashp->exsegs++; 824 hashp->nsegs++; 825 } 826 /* 827 * If the split point is increasing (MAX_BUCKET's log base 2 828 * * increases), we need to copy the current contents of the spare 829 * split bucket to the next bucket. 830 */ 831 spare_ndx = __log2(hashp->MAX_BUCKET + 1); 832 if (spare_ndx > hashp->OVFL_POINT) { 833 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT]; 834 hashp->OVFL_POINT = spare_ndx; 835 } 836 837 if (new_bucket > hashp->HIGH_MASK) { 838 /* Starting a new doubling */ 839 hashp->LOW_MASK = hashp->HIGH_MASK; 840 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK; 841 } 842 /* Relocate records to the new bucket */ 843 return (__split_page(hashp, old_bucket, new_bucket)); 844 } 845 846 /* 847 * If realloc guarantees that the pointer is not destroyed if the realloc 848 * fails, then this routine can go away. 849 */ 850 static void * 851 hash_realloc(SEGMENT **p_ptr, int oldsize, int newsize) 852 { 853 void *p; 854 855 if ( (p = malloc(newsize)) ) { 856 memmove(p, *p_ptr, oldsize); 857 memset((char *)p + oldsize, 0, newsize - oldsize); 858 free(*p_ptr); 859 *p_ptr = p; 860 } 861 return (p); 862 } 863 864 u_int32_t 865 __call_hash(HTAB *hashp, char *k, int len) 866 { 867 unsigned int n, bucket; 868 869 n = hashp->hash(k, len); 870 bucket = n & hashp->HIGH_MASK; 871 if (bucket > hashp->MAX_BUCKET) 872 bucket = bucket & hashp->LOW_MASK; 873 return (bucket); 874 } 875 876 /* 877 * Allocate segment table. On error, destroy the table and set errno. 878 * 879 * Returns 0 on success 880 */ 881 static int 882 alloc_segs(HTAB *hashp, int nsegs) 883 { 884 int i; 885 SEGMENT store; 886 887 int save_errno; 888 889 if ((hashp->dir = 890 calloc(hashp->DSIZE, sizeof(SEGMENT *))) == NULL) { 891 save_errno = errno; 892 (void)hdestroy(hashp); 893 errno = save_errno; 894 return (-1); 895 } 896 hashp->nsegs = nsegs; 897 if (nsegs == 0) 898 return (0); 899 /* Allocate segments */ 900 if ((store = calloc(nsegs << hashp->SSHIFT, sizeof(SEGMENT))) == NULL) { 901 save_errno = errno; 902 (void)hdestroy(hashp); 903 errno = save_errno; 904 return (-1); 905 } 906 for (i = 0; i < nsegs; i++) 907 hashp->dir[i] = &store[i << hashp->SSHIFT]; 908 return (0); 909 } 910 911 #if BYTE_ORDER == LITTLE_ENDIAN 912 /* 913 * Hashp->hdr needs to be byteswapped. 914 */ 915 static void 916 swap_header_copy(HASHHDR *srcp, HASHHDR *destp) 917 { 918 int i; 919 920 P_32_COPY(srcp->magic, destp->magic); 921 P_32_COPY(srcp->version, destp->version); 922 P_32_COPY(srcp->lorder, destp->lorder); 923 P_32_COPY(srcp->bsize, destp->bsize); 924 P_32_COPY(srcp->bshift, destp->bshift); 925 P_32_COPY(srcp->dsize, destp->dsize); 926 P_32_COPY(srcp->ssize, destp->ssize); 927 P_32_COPY(srcp->sshift, destp->sshift); 928 P_32_COPY(srcp->ovfl_point, destp->ovfl_point); 929 P_32_COPY(srcp->last_freed, destp->last_freed); 930 P_32_COPY(srcp->max_bucket, destp->max_bucket); 931 P_32_COPY(srcp->high_mask, destp->high_mask); 932 P_32_COPY(srcp->low_mask, destp->low_mask); 933 P_32_COPY(srcp->ffactor, destp->ffactor); 934 P_32_COPY(srcp->nkeys, destp->nkeys); 935 P_32_COPY(srcp->hdrpages, destp->hdrpages); 936 P_32_COPY(srcp->h_charkey, destp->h_charkey); 937 for (i = 0; i < NCACHED; i++) { 938 P_32_COPY(srcp->spares[i], destp->spares[i]); 939 P_16_COPY(srcp->bitmaps[i], destp->bitmaps[i]); 940 } 941 } 942 943 static void 944 swap_header(HTAB *hashp) 945 { 946 HASHHDR *hdrp; 947 int i; 948 949 hdrp = &hashp->hdr; 950 951 M_32_SWAP(hdrp->magic); 952 M_32_SWAP(hdrp->version); 953 M_32_SWAP(hdrp->lorder); 954 M_32_SWAP(hdrp->bsize); 955 M_32_SWAP(hdrp->bshift); 956 M_32_SWAP(hdrp->dsize); 957 M_32_SWAP(hdrp->ssize); 958 M_32_SWAP(hdrp->sshift); 959 M_32_SWAP(hdrp->ovfl_point); 960 M_32_SWAP(hdrp->last_freed); 961 M_32_SWAP(hdrp->max_bucket); 962 M_32_SWAP(hdrp->high_mask); 963 M_32_SWAP(hdrp->low_mask); 964 M_32_SWAP(hdrp->ffactor); 965 M_32_SWAP(hdrp->nkeys); 966 M_32_SWAP(hdrp->hdrpages); 967 M_32_SWAP(hdrp->h_charkey); 968 for (i = 0; i < NCACHED; i++) { 969 M_32_SWAP(hdrp->spares[i]); 970 M_16_SWAP(hdrp->bitmaps[i]); 971 } 972 } 973 #endif 974