hash.c revision 1.2 1 /*-
2 * Copyright (c) 1990, 1993
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
37 #if defined(LIBC_SCCS) && !defined(lint)
38 /*static char sccsid[] = "from: @(#)hash.c 8.1 (Berkeley) 6/6/93";*/
39 static char rcsid[] = "$Id: hash.c,v 1.2 1993/08/01 18:43:32 mycroft Exp $";
40 #endif /* LIBC_SCCS and not lint */
41
42 #include <sys/param.h>
43 #include <sys/stat.h>
44
45 #include <errno.h>
46 #include <fcntl.h>
47 #include <stdio.h>
48 #include <stdlib.h>
49 #include <string.h>
50 #include <unistd.h>
51 #ifdef DEBUG
52 #include <assert.h>
53 #endif
54
55 #include <db.h>
56 #include "hash.h"
57 #include "page.h"
58 #include "extern.h"
59
60 static int alloc_segs __P((HTAB *, int));
61 static int flush_meta __P((HTAB *));
62 static int hash_access __P((HTAB *, ACTION, DBT *, DBT *));
63 static int hash_close __P((DB *));
64 static int hash_delete __P((const DB *, const DBT *, u_int));
65 static int hash_fd __P((const DB *));
66 static int hash_get __P((const DB *, const DBT *, DBT *, u_int));
67 static int hash_put __P((const DB *, DBT *, const DBT *, u_int));
68 static void *hash_realloc __P((SEGMENT **, int, int));
69 static int hash_seq __P((const DB *, DBT *, DBT *, u_int));
70 static int hash_sync __P((const DB *, u_int));
71 static int hdestroy __P((HTAB *));
72 static HTAB *init_hash __P((HTAB *, const char *, HASHINFO *));
73 static int init_htab __P((HTAB *, int));
74 #if BYTE_ORDER == LITTLE_ENDIAN
75 static void swap_header __P((HTAB *));
76 static void swap_header_copy __P((HASHHDR *, HASHHDR *));
77 #endif
78
79 /* Fast arithmetic, relying on powers of 2, */
80 #define MOD(x, y) ((x) & ((y) - 1))
81
82 #define RETURN_ERROR(ERR, LOC) { save_errno = ERR; goto LOC; }
83
84 /* Return values */
85 #define SUCCESS (0)
86 #define ERROR (-1)
87 #define ABNORMAL (1)
88
89 #ifdef HASH_STATISTICS
90 long hash_accesses, hash_collisions, hash_expansions, hash_overflows;
91 #endif
92
93 /************************** INTERFACE ROUTINES ***************************/
94 /* OPEN/CLOSE */
95
96 extern DB *
97 __hash_open(file, flags, mode, info)
98 const char *file;
99 int flags, mode;
100 const HASHINFO *info; /* Special directives for create */
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 = calloc(1, sizeof(HTAB))))
113 return (NULL);
114 hashp->fp = -1;
115 /*
116 * Select flags relevant to us. Even if user wants write only, we need
117 * to be able to read the actual file, so we need to open it read/write.
118 * But, the field in the hashp structure needs to be accurate so that
119 * we can check accesses.
120 */
121 hashp->flags = flags = flags & __USE_OPEN_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 (void)fcntl(hashp->fp, F_SETFD, 1);
134 }
135 if (new_table) {
136 if (!(hashp = init_hash(hashp, file, (HASHINFO *)info)))
137 RETURN_ERROR(errno, error1);
138 } else {
139 /* Table already exists */
140 if (info && info->hash)
141 hashp->hash = info->hash;
142 else
143 hashp->hash = __default_hash;
144
145 hdrsize = read(hashp->fp, &hashp->hdr, sizeof(HASHHDR));
146 #if BYTE_ORDER == LITTLE_ENDIAN
147 swap_header(hashp);
148 #endif
149 if (hdrsize == -1)
150 RETURN_ERROR(errno, error1);
151 if (hdrsize != sizeof(HASHHDR))
152 RETURN_ERROR(EFTYPE, error1);
153 /* Verify file type, versions and hash function */
154 if (hashp->MAGIC != HASHMAGIC)
155 RETURN_ERROR(EFTYPE, error1);
156 if (hashp->VERSION != HASHVERSION)
157 RETURN_ERROR(EFTYPE, error1);
158 if (hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY)
159 RETURN_ERROR(EFTYPE, error1);
160 /*
161 * Figure out how many segments we need. Max_Bucket is the
162 * maximum bucket number, so the number of buckets is
163 * max_bucket + 1.
164 */
165 nsegs = (hashp->MAX_BUCKET + 1 + hashp->SGSIZE - 1) /
166 hashp->SGSIZE;
167 hashp->nsegs = 0;
168 if (alloc_segs(hashp, nsegs))
169 /*
170 * If alloc_segs fails, table will have been destroyed
171 * and errno will have been set.
172 */
173 return (NULL);
174 /* Read in bitmaps */
175 bpages = (hashp->SPARES[hashp->OVFL_POINT] +
176 (hashp->BSIZE << BYTE_SHIFT) - 1) >>
177 (hashp->BSHIFT + BYTE_SHIFT);
178
179 hashp->nmaps = bpages;
180 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(u_long *));
181 }
182
183 /* Initialize Buffer Manager */
184 if (info && info->cachesize)
185 __buf_init(hashp, info->cachesize);
186 else
187 __buf_init(hashp, DEF_BUFSIZE);
188
189 hashp->new_file = new_table;
190 hashp->save_file = file && (hashp->flags & O_RDWR);
191 hashp->cbucket = -1;
192 if (!(dbp = malloc(sizeof(DB)))) {
193 save_errno = errno;
194 hdestroy(hashp);
195 errno = save_errno;
196 return (NULL);
197 }
198 dbp->internal = hashp;
199 dbp->close = hash_close;
200 dbp->del = hash_delete;
201 dbp->fd = hash_fd;
202 dbp->get = hash_get;
203 dbp->put = hash_put;
204 dbp->seq = hash_seq;
205 dbp->sync = hash_sync;
206 dbp->type = DB_HASH;
207
208 #ifdef DEBUG
209 (void)fprintf(stderr,
210 "%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",
211 "init_htab:",
212 "TABLE POINTER ", hashp,
213 "BUCKET SIZE ", hashp->BSIZE,
214 "BUCKET SHIFT ", hashp->BSHIFT,
215 "DIRECTORY SIZE ", hashp->DSIZE,
216 "SEGMENT SIZE ", hashp->SGSIZE,
217 "SEGMENT SHIFT ", hashp->SSHIFT,
218 "FILL FACTOR ", hashp->FFACTOR,
219 "MAX BUCKET ", hashp->MAX_BUCKET,
220 "OVFL POINT ", hashp->OVFL_POINT,
221 "LAST FREED ", hashp->LAST_FREED,
222 "HIGH MASK ", hashp->HIGH_MASK,
223 "LOW MASK ", hashp->LOW_MASK,
224 "NSEGS ", hashp->nsegs,
225 "NKEYS ", hashp->NKEYS);
226 #endif
227 #ifdef HASH_STATISTICS
228 hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0;
229 #endif
230 return (dbp);
231
232 error1:
233 if (hashp != NULL)
234 (void)close(hashp->fp);
235
236 error0:
237 free(hashp);
238 errno = save_errno;
239 return (NULL);
240 }
241
242 static int
243 hash_close(dbp)
244 DB *dbp;
245 {
246 HTAB *hashp;
247 int retval;
248
249 if (!dbp)
250 return (ERROR);
251
252 hashp = (HTAB *)dbp->internal;
253 retval = hdestroy(hashp);
254 free(dbp);
255 return (retval);
256 }
257
258 static int
259 hash_fd(dbp)
260 const DB *dbp;
261 {
262 HTAB *hashp;
263
264 if (!dbp)
265 return (ERROR);
266
267 hashp = (HTAB *)dbp->internal;
268 if (hashp->fp == -1) {
269 errno = ENOENT;
270 return (-1);
271 }
272 return (hashp->fp);
273 }
274
275 /************************** LOCAL CREATION ROUTINES **********************/
276 static HTAB *
277 init_hash(hashp, file, info)
278 HTAB *hashp;
279 const char *file;
280 HASHINFO *info;
281 {
282 struct stat statbuf;
283 int nelem;
284
285 nelem = 1;
286 hashp->NKEYS = 0;
287 hashp->LORDER = BYTE_ORDER;
288 hashp->BSIZE = DEF_BUCKET_SIZE;
289 hashp->BSHIFT = DEF_BUCKET_SHIFT;
290 hashp->SGSIZE = DEF_SEGSIZE;
291 hashp->SSHIFT = DEF_SEGSIZE_SHIFT;
292 hashp->DSIZE = DEF_DIRSIZE;
293 hashp->FFACTOR = DEF_FFACTOR;
294 hashp->hash = __default_hash;
295 memset(hashp->SPARES, 0, sizeof(hashp->SPARES));
296 memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS));
297
298 /* Fix bucket size to be optimal for file system */
299 if (file != NULL) {
300 if (stat(file, &statbuf))
301 return (NULL);
302 hashp->BSIZE = statbuf.st_blksize;
303 hashp->BSHIFT = __log2(hashp->BSIZE);
304 }
305
306 if (info) {
307 if (info->bsize) {
308 /* Round pagesize up to power of 2 */
309 hashp->BSHIFT = __log2(info->bsize);
310 hashp->BSIZE = 1 << hashp->BSHIFT;
311 if (hashp->BSIZE > MAX_BSIZE) {
312 errno = EINVAL;
313 return (NULL);
314 }
315 }
316 if (info->ffactor)
317 hashp->FFACTOR = info->ffactor;
318 if (info->hash)
319 hashp->hash = info->hash;
320 if (info->nelem)
321 nelem = info->nelem;
322 if (info->lorder) {
323 if (info->lorder != BIG_ENDIAN &&
324 info->lorder != LITTLE_ENDIAN) {
325 errno = EINVAL;
326 return (NULL);
327 }
328 hashp->LORDER = info->lorder;
329 }
330 }
331 /* init_htab should destroy the table and set errno if it fails */
332 if (init_htab(hashp, nelem))
333 return (NULL);
334 else
335 return (hashp);
336 }
337 /*
338 * This calls alloc_segs which may run out of memory. Alloc_segs will destroy
339 * the table and set errno, so we just pass the error information along.
340 *
341 * Returns 0 on No Error
342 */
343 static int
344 init_htab(hashp, nelem)
345 HTAB *hashp;
346 int nelem;
347 {
348 register int nbuckets, nsegs;
349 int l2;
350
351 /*
352 * Divide number of elements by the fill factor and determine a
353 * desired number of buckets. Allocate space for the next greater
354 * power of two number of buckets.
355 */
356 nelem = (nelem - 1) / hashp->FFACTOR + 1;
357
358 l2 = __log2(MAX(nelem, 2));
359 nbuckets = 1 << l2;
360
361 hashp->SPARES[l2] = l2 + 1;
362 hashp->SPARES[l2 + 1] = l2 + 1;
363 hashp->OVFL_POINT = l2;
364 hashp->LAST_FREED = 2;
365
366 /* First bitmap page is at: splitpoint l2 page offset 1 */
367 if (__init_bitmap(hashp, OADDR_OF(l2, 1), l2 + 1, 0))
368 return (-1);
369
370 hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1;
371 hashp->HIGH_MASK = (nbuckets << 1) - 1;
372 hashp->HDRPAGES = ((MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >>
373 hashp->BSHIFT) + 1;
374
375 nsegs = (nbuckets - 1) / hashp->SGSIZE + 1;
376 nsegs = 1 << __log2(nsegs);
377
378 if (nsegs > hashp->DSIZE)
379 hashp->DSIZE = nsegs;
380 return (alloc_segs(hashp, nsegs));
381 }
382
383 /********************** DESTROY/CLOSE ROUTINES ************************/
384
385 /*
386 * Flushes any changes to the file if necessary and destroys the hashp
387 * structure, freeing all allocated space.
388 */
389 static int
390 hdestroy(hashp)
391 HTAB *hashp;
392 {
393 int i, save_errno;
394
395 save_errno = 0;
396
397 #ifdef HASH_STATISTICS
398 (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n",
399 hash_accesses, hash_collisions);
400 (void)fprintf(stderr, "hdestroy: expansions %ld\n",
401 hash_expansions);
402 (void)fprintf(stderr, "hdestroy: overflows %ld\n",
403 hash_overflows);
404 (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n",
405 hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs);
406
407 for (i = 0; i < NCACHED; i++)
408 (void)fprintf(stderr,
409 "spares[%d] = %d\n", i, hashp->SPARES[i]);
410 #endif
411 /*
412 * Call on buffer manager to free buffers, and if required,
413 * write them to disk.
414 */
415 if (__buf_free(hashp, 1, hashp->save_file))
416 save_errno = errno;
417 if (hashp->dir) {
418 free(*hashp->dir); /* Free initial segments */
419 /* Free extra segments */
420 while (hashp->exsegs--)
421 free(hashp->dir[--hashp->nsegs]);
422 free(hashp->dir);
423 }
424 if (flush_meta(hashp) && !save_errno)
425 save_errno = errno;
426 /* Free Bigmaps */
427 for (i = 0; i < hashp->nmaps; i++)
428 if (hashp->mapp[i])
429 free(hashp->mapp[i]);
430
431 if (hashp->fp != -1)
432 (void)close(hashp->fp);
433
434 if (save_errno) {
435 errno = save_errno;
436 return (ERROR);
437 }
438 return (SUCCESS);
439 }
440 /*
441 * Write modified pages to disk
442 *
443 * Returns:
444 * 0 == OK
445 * -1 ERROR
446 */
447 static int
448 hash_sync(dbp, flags)
449 const DB *dbp;
450 u_int flags;
451 {
452 HTAB *hashp;
453
454 if (flags != 0) {
455 errno = EINVAL;
456 return (ERROR);
457 }
458
459 if (!dbp)
460 return (ERROR);
461
462 hashp = (HTAB *)dbp->internal;
463 if (!hashp->save_file)
464 return (0);
465 if (__buf_free(hashp, 0, 1) || flush_meta(hashp))
466 return (ERROR);
467 hashp->new_file = 0;
468 return (0);
469 }
470
471 /*
472 * Returns:
473 * 0 == OK
474 * -1 indicates that errno should be set
475 */
476 static int
477 flush_meta(hashp)
478 HTAB *hashp;
479 {
480 HASHHDR *whdrp;
481 #if BYTE_ORDER == LITTLE_ENDIAN
482 HASHHDR whdr;
483 #endif
484 int fp, i, wsize;
485
486 if (!hashp->save_file)
487 return (0);
488 hashp->MAGIC = HASHMAGIC;
489 hashp->VERSION = HASHVERSION;
490 hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY));
491
492 fp = hashp->fp;
493 whdrp = &hashp->hdr;
494 #if BYTE_ORDER == LITTLE_ENDIAN
495 whdrp = &whdr;
496 swap_header_copy(&hashp->hdr, whdrp);
497 #endif
498 if ((lseek(fp, (off_t)0, SEEK_SET) == -1) ||
499 ((wsize = write(fp, whdrp, sizeof(HASHHDR))) == -1))
500 return (-1);
501 else
502 if (wsize != sizeof(HASHHDR)) {
503 errno = EFTYPE;
504 hashp->errno = errno;
505 return (-1);
506 }
507 for (i = 0; i < NCACHED; i++)
508 if (hashp->mapp[i])
509 if (__put_page(hashp, (char *)hashp->mapp[i],
510 hashp->BITMAPS[i], 0, 1))
511 return (-1);
512 return (0);
513 }
514
515 /*******************************SEARCH ROUTINES *****************************/
516 /*
517 * All the access routines return
518 *
519 * Returns:
520 * 0 on SUCCESS
521 * 1 to indicate an external ERROR (i.e. key not found, etc)
522 * -1 to indicate an internal ERROR (i.e. out of memory, etc)
523 */
524 static int
525 hash_get(dbp, key, data, flag)
526 const DB *dbp;
527 const DBT *key;
528 DBT *data;
529 u_int flag;
530 {
531 HTAB *hashp;
532
533 hashp = (HTAB *)dbp->internal;
534 if (flag) {
535 hashp->errno = errno = EINVAL;
536 return (ERROR);
537 }
538 return (hash_access(hashp, HASH_GET, (DBT *)key, data));
539 }
540
541 static int
542 hash_put(dbp, key, data, flag)
543 const DB *dbp;
544 DBT *key;
545 const DBT *data;
546 u_int flag;
547 {
548 HTAB *hashp;
549
550 hashp = (HTAB *)dbp->internal;
551 if (flag && flag != R_NOOVERWRITE) {
552 hashp->errno = errno = EINVAL;
553 return (ERROR);
554 }
555 if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
556 hashp->errno = errno = EPERM;
557 return (ERROR);
558 }
559 return (hash_access(hashp, flag == R_NOOVERWRITE ?
560 HASH_PUTNEW : HASH_PUT, (DBT *)key, (DBT *)data));
561 }
562
563 static int
564 hash_delete(dbp, key, flag)
565 const DB *dbp;
566 const DBT *key;
567 u_int flag; /* Ignored */
568 {
569 HTAB *hashp;
570
571 hashp = (HTAB *)dbp->internal;
572 if (flag && flag != R_CURSOR) {
573 hashp->errno = errno = EINVAL;
574 return (ERROR);
575 }
576 if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
577 hashp->errno = errno = EPERM;
578 return (ERROR);
579 }
580 return (hash_access(hashp, HASH_DELETE, (DBT *)key, NULL));
581 }
582
583 /*
584 * Assume that hashp has been set in wrapper routine.
585 */
586 static int
587 hash_access(hashp, action, key, val)
588 HTAB *hashp;
589 ACTION action;
590 DBT *key, *val;
591 {
592 register BUFHEAD *rbufp;
593 BUFHEAD *bufp, *save_bufp;
594 register u_short *bp;
595 register int n, ndx, off, size;
596 register char *kp;
597 u_short pageno;
598
599 #ifdef HASH_STATISTICS
600 hash_accesses++;
601 #endif
602
603 off = hashp->BSIZE;
604 size = key->size;
605 kp = (char *)key->data;
606 rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0);
607 if (!rbufp)
608 return (ERROR);
609 save_bufp = rbufp;
610
611 /* Pin the bucket chain */
612 rbufp->flags |= BUF_PIN;
613 for (bp = (u_short *)rbufp->page, n = *bp++, ndx = 1; ndx < n;)
614 if (bp[1] >= REAL_KEY) {
615 /* Real key/data pair */
616 if (size == off - *bp &&
617 memcmp(kp, rbufp->page + *bp, size) == 0)
618 goto found;
619 off = bp[1];
620 #ifdef HASH_STATISTICS
621 hash_collisions++;
622 #endif
623 bp += 2;
624 ndx += 2;
625 } else if (bp[1] == OVFLPAGE) {
626 rbufp = __get_buf(hashp, *bp, rbufp, 0);
627 if (!rbufp) {
628 save_bufp->flags &= ~BUF_PIN;
629 return (ERROR);
630 }
631 /* FOR LOOP INIT */
632 bp = (u_short *)rbufp->page;
633 n = *bp++;
634 ndx = 1;
635 off = hashp->BSIZE;
636 } else if (bp[1] < REAL_KEY) {
637 if ((ndx =
638 __find_bigpair(hashp, rbufp, ndx, kp, size)) > 0)
639 goto found;
640 if (ndx == -2) {
641 bufp = rbufp;
642 if (!(pageno =
643 __find_last_page(hashp, &bufp))) {
644 ndx = 0;
645 rbufp = bufp;
646 break; /* FOR */
647 }
648 rbufp = __get_buf(hashp, pageno, bufp, 0);
649 if (!rbufp) {
650 save_bufp->flags &= ~BUF_PIN;
651 return (ERROR);
652 }
653 /* FOR LOOP INIT */
654 bp = (u_short *)rbufp->page;
655 n = *bp++;
656 ndx = 1;
657 off = hashp->BSIZE;
658 } else {
659 save_bufp->flags &= ~BUF_PIN;
660 return (ERROR);
661 }
662 }
663
664 /* Not found */
665 switch (action) {
666 case HASH_PUT:
667 case HASH_PUTNEW:
668 if (__addel(hashp, rbufp, key, val)) {
669 save_bufp->flags &= ~BUF_PIN;
670 return (ERROR);
671 } else {
672 save_bufp->flags &= ~BUF_PIN;
673 return (SUCCESS);
674 }
675 case HASH_GET:
676 case HASH_DELETE:
677 default:
678 save_bufp->flags &= ~BUF_PIN;
679 return (ABNORMAL);
680 }
681
682 found:
683 switch (action) {
684 case HASH_PUTNEW:
685 save_bufp->flags &= ~BUF_PIN;
686 return (ABNORMAL);
687 case HASH_GET:
688 bp = (u_short *)rbufp->page;
689 if (bp[ndx + 1] < REAL_KEY) {
690 if (__big_return(hashp, rbufp, ndx, val, 0))
691 return (ERROR);
692 } else {
693 val->data = (u_char *)rbufp->page + (int)bp[ndx + 1];
694 val->size = bp[ndx] - bp[ndx + 1];
695 }
696 break;
697 case HASH_PUT:
698 if ((__delpair(hashp, rbufp, ndx)) ||
699 (__addel(hashp, rbufp, key, val))) {
700 save_bufp->flags &= ~BUF_PIN;
701 return (ERROR);
702 }
703 break;
704 case HASH_DELETE:
705 if (__delpair(hashp, rbufp, ndx))
706 return (ERROR);
707 break;
708 default:
709 abort();
710 }
711 save_bufp->flags &= ~BUF_PIN;
712 return (SUCCESS);
713 }
714
715 static int
716 hash_seq(dbp, key, data, flag)
717 const DB *dbp;
718 DBT *key, *data;
719 u_int flag;
720 {
721 register u_int bucket;
722 register BUFHEAD *bufp;
723 HTAB *hashp;
724 u_short *bp, ndx;
725
726 hashp = (HTAB *)dbp->internal;
727 if (flag && flag != R_FIRST && flag != R_NEXT) {
728 hashp->errno = errno = EINVAL;
729 return (ERROR);
730 }
731 #ifdef HASH_STATISTICS
732 hash_accesses++;
733 #endif
734 if ((hashp->cbucket < 0) || (flag == R_FIRST)) {
735 hashp->cbucket = 0;
736 hashp->cndx = 1;
737 hashp->cpage = NULL;
738 }
739
740 for (bp = NULL; !bp || !bp[0]; ) {
741 if (!(bufp = hashp->cpage)) {
742 for (bucket = hashp->cbucket;
743 bucket <= hashp->MAX_BUCKET;
744 bucket++, hashp->cndx = 1) {
745 bufp = __get_buf(hashp, bucket, NULL, 0);
746 if (!bufp)
747 return (ERROR);
748 hashp->cpage = bufp;
749 bp = (u_short *)bufp->page;
750 if (bp[0])
751 break;
752 }
753 hashp->cbucket = bucket;
754 if (hashp->cbucket > hashp->MAX_BUCKET) {
755 hashp->cbucket = -1;
756 return (ABNORMAL);
757 }
758 } else
759 bp = (u_short *)hashp->cpage->page;
760
761 #ifdef DEBUG
762 assert(bp);
763 assert(bufp);
764 #endif
765 while (bp[hashp->cndx + 1] == OVFLPAGE) {
766 bufp = hashp->cpage =
767 __get_buf(hashp, bp[hashp->cndx], bufp, 0);
768 if (!bufp)
769 return (ERROR);
770 bp = (u_short *)(bufp->page);
771 hashp->cndx = 1;
772 }
773 if (!bp[0]) {
774 hashp->cpage = NULL;
775 ++hashp->cbucket;
776 }
777 }
778 ndx = hashp->cndx;
779 if (bp[ndx + 1] < REAL_KEY) {
780 if (__big_keydata(hashp, bufp, key, data, 1))
781 return (ERROR);
782 } else {
783 key->data = (u_char *)hashp->cpage->page + bp[ndx];
784 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->BSIZE) - bp[ndx];
785 data->data = (u_char *)hashp->cpage->page + bp[ndx + 1];
786 data->size = bp[ndx] - bp[ndx + 1];
787 ndx += 2;
788 if (ndx > bp[0]) {
789 hashp->cpage = NULL;
790 hashp->cbucket++;
791 hashp->cndx = 1;
792 } else
793 hashp->cndx = ndx;
794 }
795 return (SUCCESS);
796 }
797
798 /********************************* UTILITIES ************************/
799
800 /*
801 * Returns:
802 * 0 ==> OK
803 * -1 ==> Error
804 */
805 extern int
806 __expand_table(hashp)
807 HTAB *hashp;
808 {
809 u_int old_bucket, new_bucket;
810 int dirsize, new_segnum, spare_ndx;
811
812 #ifdef HASH_STATISTICS
813 hash_expansions++;
814 #endif
815 new_bucket = ++hashp->MAX_BUCKET;
816 old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK);
817
818 new_segnum = new_bucket >> hashp->SSHIFT;
819
820 /* Check if we need a new segment */
821 if (new_segnum >= hashp->nsegs) {
822 /* Check if we need to expand directory */
823 if (new_segnum >= hashp->DSIZE) {
824 /* Reallocate directory */
825 dirsize = hashp->DSIZE * sizeof(SEGMENT *);
826 if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1))
827 return (-1);
828 hashp->DSIZE = dirsize << 1;
829 }
830 if (!(hashp->dir[new_segnum] =
831 calloc(hashp->SGSIZE, sizeof(SEGMENT))))
832 return (-1);
833 hashp->exsegs++;
834 hashp->nsegs++;
835 }
836 /*
837 * If the split point is increasing (MAX_BUCKET's log base 2
838 * * increases), we need to copy the current contents of the spare
839 * split bucket to the next bucket.
840 */
841 spare_ndx = __log2(hashp->MAX_BUCKET + 1);
842 if (spare_ndx > hashp->OVFL_POINT) {
843 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT];
844 hashp->OVFL_POINT = spare_ndx;
845 }
846
847 if (new_bucket > hashp->HIGH_MASK) {
848 /* Starting a new doubling */
849 hashp->LOW_MASK = hashp->HIGH_MASK;
850 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK;
851 }
852 /* Relocate records to the new bucket */
853 return (__split_page(hashp, old_bucket, new_bucket));
854 }
855
856 /*
857 * If realloc guarantees that the pointer is not destroyed if the realloc
858 * fails, then this routine can go away.
859 */
860 static void *
861 hash_realloc(p_ptr, oldsize, newsize)
862 SEGMENT **p_ptr;
863 int oldsize, newsize;
864 {
865 register void *p;
866
867 if (p = malloc(newsize)) {
868 memmove(p, *p_ptr, oldsize);
869 memset(p + oldsize, 0, newsize - oldsize);
870 free(*p_ptr);
871 *p_ptr = p;
872 }
873 return (p);
874 }
875
876 extern u_int
877 __call_hash(hashp, k, len)
878 HTAB *hashp;
879 char *k;
880 int len;
881 {
882 int n, bucket;
883
884 n = hashp->hash(k, len);
885 bucket = n & hashp->HIGH_MASK;
886 if (bucket > hashp->MAX_BUCKET)
887 bucket = bucket & hashp->LOW_MASK;
888 return (bucket);
889 }
890
891 /*
892 * Allocate segment table. On error, destroy the table and set errno.
893 *
894 * Returns 0 on success
895 */
896 static int
897 alloc_segs(hashp, nsegs)
898 HTAB *hashp;
899 int nsegs;
900 {
901 register int i;
902 register SEGMENT store;
903
904 int save_errno;
905
906 if (!(hashp->dir = calloc(hashp->DSIZE, sizeof(SEGMENT *)))) {
907 save_errno = errno;
908 (void)hdestroy(hashp);
909 errno = save_errno;
910 return (-1);
911 }
912 /* Allocate segments */
913 store = calloc(nsegs << hashp->SSHIFT, sizeof(SEGMENT));
914 if (!store) {
915 save_errno = errno;
916 (void)hdestroy(hashp);
917 errno = save_errno;
918 return (-1);
919 }
920 for (i = 0; i < nsegs; i++, hashp->nsegs++)
921 hashp->dir[i] = &store[i << hashp->SSHIFT];
922 return (0);
923 }
924
925 #if BYTE_ORDER == LITTLE_ENDIAN
926 /*
927 * Hashp->hdr needs to be byteswapped.
928 */
929 static void
930 swap_header_copy(srcp, destp)
931 HASHHDR *srcp, *destp;
932 {
933 int i;
934
935 BLSWAP_COPY(srcp->magic, destp->magic);
936 BLSWAP_COPY(srcp->version, destp->version);
937 BLSWAP_COPY(srcp->lorder, destp->lorder);
938 BLSWAP_COPY(srcp->bsize, destp->bsize);
939 BLSWAP_COPY(srcp->bshift, destp->bshift);
940 BLSWAP_COPY(srcp->dsize, destp->dsize);
941 BLSWAP_COPY(srcp->ssize, destp->ssize);
942 BLSWAP_COPY(srcp->sshift, destp->sshift);
943 BLSWAP_COPY(srcp->ovfl_point, destp->ovfl_point);
944 BLSWAP_COPY(srcp->last_freed, destp->last_freed);
945 BLSWAP_COPY(srcp->max_bucket, destp->max_bucket);
946 BLSWAP_COPY(srcp->high_mask, destp->high_mask);
947 BLSWAP_COPY(srcp->low_mask, destp->low_mask);
948 BLSWAP_COPY(srcp->ffactor, destp->ffactor);
949 BLSWAP_COPY(srcp->nkeys, destp->nkeys);
950 BLSWAP_COPY(srcp->hdrpages, destp->hdrpages);
951 BLSWAP_COPY(srcp->h_charkey, destp->h_charkey);
952 for (i = 0; i < NCACHED; i++) {
953 BLSWAP_COPY(srcp->spares[i], destp->spares[i]);
954 BSSWAP_COPY(srcp->bitmaps[i], destp->bitmaps[i]);
955 }
956 }
957
958 static void
959 swap_header(hashp)
960 HTAB *hashp;
961 {
962 HASHHDR *hdrp;
963 int i;
964
965 hdrp = &hashp->hdr;
966
967 BLSWAP(hdrp->magic);
968 BLSWAP(hdrp->version);
969 BLSWAP(hdrp->lorder);
970 BLSWAP(hdrp->bsize);
971 BLSWAP(hdrp->bshift);
972 BLSWAP(hdrp->dsize);
973 BLSWAP(hdrp->ssize);
974 BLSWAP(hdrp->sshift);
975 BLSWAP(hdrp->ovfl_point);
976 BLSWAP(hdrp->last_freed);
977 BLSWAP(hdrp->max_bucket);
978 BLSWAP(hdrp->high_mask);
979 BLSWAP(hdrp->low_mask);
980 BLSWAP(hdrp->ffactor);
981 BLSWAP(hdrp->nkeys);
982 BLSWAP(hdrp->hdrpages);
983 BLSWAP(hdrp->h_charkey);
984 for (i = 0; i < NCACHED; i++) {
985 BLSWAP(hdrp->spares[i]);
986 BSSWAP(hdrp->bitmaps[i]);
987 }
988 }
989 #endif
990