1 /*$Id$*/
2 /*-
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Margo Seltzer.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 */
37
38 #if defined(LIBC_SCCS) && !defined(lint)
39 static char sccsid[] = "@(#)hash.c 8.1 (Berkeley) 6/6/93";
40 #endif /* LIBC_SCCS and not lint */
41
42 #include <sys/types.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 #ifndef IPSC
52 #include <unistd.h>
53 #endif
54 #ifdef DEBUG
55 #include <assert.h>
56 #endif
57
58 #include <db.h>
59 #include "hash.h"
60 #include "page.h"
61 #include "extern.h"
62
63 #if defined(NX)
64 extern int ipsc_open(const char *, int, int);
65 extern int ipsc_read(int, char *, int);
66 extern int ipsc_write(int, const char *, int);
67 extern int ipsc_lseek(int, off_t, int);
68
69 #define open(a,b,c) ipsc_open(a,b,c)
70 #define read(a,b,c) ipsc_read(a,b,c)
71 #define write(a,b,c) ipsc_write(a,b,c)
72 #define lseek(a,b,c) ipsc_lseek(a,b,c)
73 #endif
74
75 static int alloc_segs __P((HTAB *, int));
76 static int flush_meta __P((HTAB *));
77 static int hash_access __P((HTAB *, ACTION, DBT *, DBT *));
78 static int hash_close __P((DB *));
79 static int hash_delete __P((const DB *, const DBT *, u_int));
80 static int hash_fd __P((const DB *));
81 static int hash_get __P((const DB *, const DBT *, DBT *, u_int));
82 static int hash_put __P((const DB *, DBT *, const DBT *, u_int));
83 static void *hash_realloc __P((SEGMENT **, int, int));
84 static int hash_seq __P((const DB *, DBT *, DBT *, u_int));
85 static int hash_sync __P((const DB *, u_int));
86 static int hdestroy __P((HTAB *));
87 static HTAB *init_hash __P((HTAB *, const char *, HASHINFO *));
88 static int init_htab __P((HTAB *, int));
89 #if BYTE_ORDER == LITTLE_ENDIAN
90 static void swap_header __P((HTAB *));
91 static void swap_header_copy __P((HASHHDR *, HASHHDR *));
92 #endif
93
94 /* Fast arithmetic, relying on powers of 2, */
95 #define MOD(x, y) ((x) & ((y) - 1))
96
97 #define RETURN_ERROR(ERR, LOC) { save_errno = ERR; goto LOC; }
98
99 /* Return values */
100 #define SUCCESS (0)
101 #define ERROR (-1)
102 #define ABNORMAL (1)
103
104 #ifdef HASH_STATISTICS
105 long hash_accesses, hash_collisions, hash_expansions, hash_overflows;
106 #endif
107
108 /************************** INTERFACE ROUTINES ***************************/
109 /* OPEN/CLOSE */
110
111 extern DB *
112 __hash_open(file, flags, mode, info)
113 const char *file;
114 int flags, mode;
115 const HASHINFO *info; /* Special directives for create */
116 {
117 HTAB *hashp;
118 struct stat statbuf;
119 DB *dbp;
120 int bpages, new_table, nsegs, save_errno;
121 long hdrsize;
122
123 if ((flags & O_ACCMODE) == O_WRONLY) {
124 errno = EINVAL;
125 return (NULL);
126 }
127
128 if (!(hashp = (HTAB *) calloc(1, sizeof(HTAB))))
129 return (NULL);
130 hashp->fp = -1;
131 /*
132 * Select flags relevant to us. Even if user wants write only, we need
133 * to be able to read the actual file, so we need to open it read/write.
134 * But, the field in the hashp structure needs to be accurate so that
135 * we can check accesses.
136 */
137 hashp->flags = flags = flags & __USE_OPEN_FLAGS;
138
139 new_table = 0;
140 if (!file || (flags & O_TRUNC) ||
141 (stat(file, &statbuf) && (errno == ENOENT))) {
142 if (errno == ENOENT)
143 errno = 0; /* Just in case someone looks at errno */
144 new_table = 1;
145 }
146 if (file) {
147 if ((hashp->fp = open(file, flags, mode)) == -1)
148 RETURN_ERROR(errno, error0);
149 (void)fcntl(hashp->fp, F_SETFD, 1);
150 }
151 if (new_table) {
152 if (!(hashp = init_hash(hashp, file, (HASHINFO *)info)))
153 RETURN_ERROR(errno, error1);
154 } else {
155 /* Table already exists */
156 if (info && info->hash)
157 hashp->hash = info->hash;
158 else
159 hashp->hash = __default_hash;
160
161 hdrsize = read(hashp->fp, (char *) &hashp->hdr, sizeof(HASHHDR));
162 #if BYTE_ORDER == LITTLE_ENDIAN
163 swap_header(hashp);
164 #endif
165 if (hdrsize == -1)
166 RETURN_ERROR(errno, error1);
167 if (hdrsize != sizeof(HASHHDR))
168 RETURN_ERROR(EFTYPE, error1);
169 /* Verify file type, versions and hash function */
170 if (hashp->MAGIC != HASHMAGIC)
171 RETURN_ERROR(EFTYPE, error1);
172 if (hashp->VERSION != HASHVERSION)
173 RETURN_ERROR(EFTYPE, error1);
174 if (hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY)
175 RETURN_ERROR(EFTYPE, error1);
176 /*
177 * Figure out how many segments we need. Max_Bucket is the
178 * maximum bucket number, so the number of buckets is
179 * max_bucket + 1.
180 */
181 nsegs = (hashp->MAX_BUCKET + 1 + hashp->SGSIZE - 1) /
182 hashp->SGSIZE;
183 hashp->nsegs = 0;
184 if (alloc_segs(hashp, nsegs))
185 /*
186 * If alloc_segs fails, table will have been destroyed
187 * and errno will have been set.
188 */
189 return (NULL);
190 /* Read in bitmaps */
191 bpages = (hashp->SPARES[hashp->OVFL_POINT] +
192 (hashp->HBSIZE << BYTE_SHIFT) - 1) >>
193 (hashp->HBSHIFT + BYTE_SHIFT);
194
195 hashp->nmaps = bpages;
196 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(u_long *));
197 }
198
199 /* Initialize Buffer Manager */
200 if (info && info->cachesize)
201 __buf_init(hashp, info->cachesize);
202 else
203 __buf_init(hashp, DEF_BUFSIZE);
204
205 hashp->new_file = new_table;
206 hashp->save_file = file && (hashp->flags & O_RDWR);
207 hashp->cbucket = -1;
208 if (!(dbp = (DB *) calloc(sizeof(DB), 1U))) {
209 save_errno = errno;
210 hdestroy(hashp);
211 errno = save_errno;
212 return (NULL);
213 }
214 dbp->internal = hashp;
215 dbp->close = hash_close;
216 dbp->del = hash_delete;
217 dbp->fd = hash_fd;
218 dbp->get = hash_get;
219 dbp->put = hash_put;
220 dbp->seq = hash_seq;
221 dbp->sync = hash_sync;
222 dbp->type = DB_HASH;
223
224 #ifdef DEBUG
225 (void)fprintf(stderr,
226 "%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",
227 "init_htab:",
228 "TABLE POINTER ", hashp,
229 "BUCKET SIZE ", hashp->HBSIZE,
230 "BUCKET SHIFT ", hashp->HBSHIFT,
231 "DIRECTORY SIZE ", hashp->DSIZE,
232 "SEGMENT SIZE ", hashp->SGSIZE,
233 "SEGMENT SHIFT ", hashp->SSHIFT,
234 "FILL FACTOR ", hashp->FFACTOR,
235 "MAX BUCKET ", hashp->MAX_BUCKET,
236 "OVFL POINT ", hashp->OVFL_POINT,
237 "LAST FREED ", hashp->LAST_FREED,
238 "HIGH MASK ", hashp->HIGH_MASK,
239 "LOW MASK ", hashp->LOW_MASK,
240 "NSEGS ", hashp->nsegs,
241 "NKEYS ", hashp->NKEYS);
242 #endif
243 #ifdef HASH_STATISTICS
244 hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0;
245 #endif
246 return (dbp);
247
248 error1:
249 if (hashp != NULL)
250 (void)close(hashp->fp);
251
252 error0:
253 free(hashp);
254 errno = save_errno;
255 return (NULL);
256 }
257
258 static int
hash_close(dbp)259 hash_close(dbp)
260 DB *dbp;
261 {
262 HTAB *hashp;
263 int retval;
264
265 if (!dbp)
266 return (ERROR);
267
268 hashp = (HTAB *)dbp->internal;
269 retval = hdestroy(hashp);
270 free(dbp);
271 return (retval);
272 }
273
274 static int
hash_fd(dbp)275 hash_fd(dbp)
276 const DB *dbp;
277 {
278 HTAB *hashp;
279
280 if (!dbp)
281 return (ERROR);
282
283 hashp = (HTAB *)dbp->internal;
284 if (hashp->fp == -1) {
285 errno = ENOENT;
286 return (-1);
287 }
288 return (hashp->fp);
289 }
290
291 /************************** LOCAL CREATION ROUTINES **********************/
292 static HTAB *
init_hash(hashp,file,info)293 init_hash(hashp, file, info)
294 HTAB *hashp;
295 const char *file;
296 HASHINFO *info;
297 {
298 struct stat statbuf;
299 int nelem;
300
301 nelem = 1;
302 hashp->NKEYS = 0;
303 hashp->LORDER = BYTE_ORDER;
304 hashp->HBSIZE = DEF_BUCKET_SIZE;
305 hashp->HBSHIFT = DEF_BUCKET_SHIFT;
306 hashp->SGSIZE = DEF_SEGSIZE;
307 hashp->SSHIFT = DEF_SEGSIZE_SHIFT;
308 hashp->DSIZE = DEF_DIRSIZE;
309 hashp->FFACTOR = DEF_FFACTOR;
310 hashp->hash = __default_hash;
311 memset(hashp->SPARES, 0, sizeof(hashp->SPARES));
312 memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS));
313
314 /* Fix bucket size to be optimal for file system */
315 if (file != NULL) {
316 if (stat(file, &statbuf))
317 return (NULL);
318 #ifdef IPSC
319 hashp->HBSIZE = 1024;
320 #else
321 hashp->HBSIZE = (int) statbuf.st_blksize;
322 #endif
323 hashp->HBSHIFT = __log2(hashp->HBSIZE);
324 }
325
326 if (info) {
327 if (info->bsize) {
328 /* Round pagesize up to power of 2 */
329 hashp->HBSHIFT = __log2(info->bsize);
330 hashp->HBSIZE = 1 << hashp->HBSHIFT;
331 if (hashp->HBSIZE > MAX_BSIZE) {
332 errno = EINVAL;
333 return (NULL);
334 }
335 }
336 if (info->ffactor)
337 hashp->FFACTOR = info->ffactor;
338 if (info->hash)
339 hashp->hash = info->hash;
340 if (info->nelem)
341 nelem = info->nelem;
342 if (info->lorder) {
343 if (info->lorder != BIG_ENDIAN &&
344 info->lorder != LITTLE_ENDIAN) {
345 errno = EINVAL;
346 return (NULL);
347 }
348 hashp->LORDER = info->lorder;
349 }
350 }
351 /* init_htab should destroy the table and set errno if it fails */
352 if (init_htab(hashp, nelem))
353 return (NULL);
354 else
355 return (hashp);
356 }
357 /*
358 * This calls alloc_segs which may run out of memory. Alloc_segs will destroy
359 * the table and set errno, so we just pass the error information along.
360 *
361 * Returns 0 on No Error
362 */
363 static int
init_htab(hashp,nelem)364 init_htab(hashp, nelem)
365 HTAB *hashp;
366 int nelem;
367 {
368 register int nbuckets, nsegs;
369 int l2;
370
371 /*
372 * Divide number of elements by the fill factor and determine a
373 * desired number of buckets. Allocate space for the next greater
374 * power of two number of buckets.
375 */
376 nelem = (nelem - 1) / hashp->FFACTOR + 1;
377
378 l2 = __log2(MAX(nelem, 2));
379 nbuckets = 1 << l2;
380
381 hashp->SPARES[l2] = l2 + 1;
382 hashp->SPARES[l2 + 1] = l2 + 1;
383 hashp->OVFL_POINT = l2;
384 hashp->LAST_FREED = 2;
385
386 /* First bitmap page is at: splitpoint l2 page offset 1 */
387 if (__init_bitmap(hashp, OADDR_OF(l2, 1), l2 + 1, 0))
388 return (-1);
389
390 hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1;
391 hashp->HIGH_MASK = (nbuckets << 1) - 1;
392 hashp->HDRPAGES = (int) (((MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >>
393 hashp->HBSHIFT) + 1);
394
395 nsegs = (nbuckets - 1) / hashp->SGSIZE + 1;
396 nsegs = 1 << __log2(nsegs);
397
398 if (nsegs > hashp->DSIZE)
399 hashp->DSIZE = nsegs;
400 return (alloc_segs(hashp, nsegs));
401 }
402
403 /********************** DESTROY/CLOSE ROUTINES ************************/
404
405 /*
406 * Flushes any changes to the file if necessary and destroys the hashp
407 * structure, freeing all allocated space.
408 */
409 static int
hdestroy(hashp)410 hdestroy(hashp)
411 HTAB *hashp;
412 {
413 int i, save_errno;
414
415 save_errno = 0;
416
417 #ifdef HASH_STATISTICS
418 (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n",
419 hash_accesses, hash_collisions);
420 (void)fprintf(stderr, "hdestroy: expansions %ld\n",
421 hash_expansions);
422 (void)fprintf(stderr, "hdestroy: overflows %ld\n",
423 hash_overflows);
424 (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n",
425 hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs);
426
427 for (i = 0; i < NCACHED; i++)
428 (void)fprintf(stderr,
429 "spares[%d] = %d\n", i, hashp->SPARES[i]);
430 #endif
431 /*
432 * Call on buffer manager to free buffers, and if required,
433 * write them to disk.
434 */
435 if (__buf_free(hashp, 1, hashp->save_file))
436 save_errno = errno;
437 if (hashp->dir) {
438 free(*hashp->dir); /* Free initial segments */
439 /* Free extra segments */
440 while (hashp->exsegs--)
441 free(hashp->dir[--hashp->nsegs]);
442 free(hashp->dir);
443 }
444 if (flush_meta(hashp) && !save_errno)
445 save_errno = errno;
446 /* Free Bigmaps */
447 for (i = 0; i < hashp->nmaps; i++)
448 if (hashp->mapp[i])
449 free(hashp->mapp[i]);
450
451 if (hashp->fp != -1)
452 (void)close(hashp->fp);
453
454 if (save_errno) {
455 errno = save_errno;
456 return (ERROR);
457 }
458 return (SUCCESS);
459 }
460 /*
461 * Write modified pages to disk
462 *
463 * Returns:
464 * 0 == OK
465 * -1 ERROR
466 */
467 static int
hash_sync(dbp,flags)468 hash_sync(dbp, flags)
469 const DB *dbp;
470 u_int flags;
471 {
472 HTAB *hashp;
473
474 if (flags != 0) {
475 errno = EINVAL;
476 return (ERROR);
477 }
478
479 if (!dbp)
480 return (ERROR);
481
482 hashp = (HTAB *)dbp->internal;
483 if (!hashp->save_file)
484 return (0);
485 if (__buf_free(hashp, 0, 1) || flush_meta(hashp))
486 return (ERROR);
487 hashp->new_file = 0;
488 return (0);
489 }
490
491 /*
492 * Returns:
493 * 0 == OK
494 * -1 indicates that errno should be set
495 */
496 static int
flush_meta(hashp)497 flush_meta(hashp)
498 HTAB *hashp;
499 {
500 HASHHDR *whdrp;
501 #if BYTE_ORDER == LITTLE_ENDIAN
502 HASHHDR whdr;
503 #endif
504 int fp, i;
505 long wsize;
506
507 if (!hashp->save_file)
508 return (0);
509 hashp->MAGIC = HASHMAGIC;
510 hashp->VERSION = HASHVERSION;
511 hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY));
512
513 fp = hashp->fp;
514 whdrp = &hashp->hdr;
515 #if BYTE_ORDER == LITTLE_ENDIAN
516 whdrp = &whdr;
517 swap_header_copy(&hashp->hdr, whdrp);
518 #endif
519 if ((lseek(fp, (off_t)0, SEEK_SET) == -1) ||
520 ((wsize = write(fp, (char *) whdrp, sizeof(HASHHDR))) == -1))
521 return (-1);
522 else
523 if (wsize != sizeof(HASHHDR)) {
524 errno = EFTYPE;
525 hashp->errnumb = errno;
526 return (-1);
527 }
528 for (i = 0; i < NCACHED; i++)
529 if (hashp->mapp[i])
530 if (__put_page(hashp, (char *)hashp->mapp[i],
531 hashp->BITMAPS[i], 0, 1))
532 return (-1);
533 return (0);
534 }
535
536 /*******************************SEARCH ROUTINES *****************************/
537 /*
538 * All the access routines return
539 *
540 * Returns:
541 * 0 on SUCCESS
542 * 1 to indicate an external ERROR (i.e. key not found, etc)
543 * -1 to indicate an internal ERROR (i.e. out of memory, etc)
544 */
545 static int
hash_get(dbp,key,data,flag)546 hash_get(dbp, key, data, flag)
547 const DB *dbp;
548 const DBT *key;
549 DBT *data;
550 u_int flag;
551 {
552 HTAB *hashp;
553
554 hashp = (HTAB *)dbp->internal;
555 if (flag) {
556 hashp->errnumb = errno = EINVAL;
557 return (ERROR);
558 }
559 return (hash_access(hashp, HASH_GET, (DBT *)key, data));
560 }
561
562 static int
hash_put(dbp,key,data,flag)563 hash_put(dbp, key, data, flag)
564 const DB *dbp;
565 DBT *key;
566 const DBT *data;
567 u_int flag;
568 {
569 HTAB *hashp;
570
571 hashp = (HTAB *)dbp->internal;
572 if (flag && flag != R_NOOVERWRITE) {
573 hashp->errnumb = errno = EINVAL;
574 return (ERROR);
575 }
576 if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
577 hashp->errnumb = errno = EPERM;
578 return (ERROR);
579 }
580 return (hash_access(hashp, flag == R_NOOVERWRITE ?
581 HASH_PUTNEW : HASH_PUT, (DBT *)key, (DBT *)data));
582 }
583
584 static int
hash_delete(dbp,key,flag)585 hash_delete(dbp, key, flag)
586 const DB *dbp;
587 const DBT *key;
588 u_int flag; /* Ignored */
589 {
590 HTAB *hashp;
591
592 hashp = (HTAB *)dbp->internal;
593 if (flag && flag != R_CURSOR) {
594 hashp->errnumb = errno = EINVAL;
595 return (ERROR);
596 }
597 if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
598 hashp->errnumb = errno = EPERM;
599 return (ERROR);
600 }
601 return (hash_access(hashp, HASH_DELETE, (DBT *)key, NULL));
602 }
603
604 /*
605 * Assume that hashp has been set in wrapper routine.
606 */
607 static int
hash_access(hashp,action,key,val)608 hash_access(hashp, action, key, val)
609 HTAB *hashp;
610 ACTION action;
611 DBT *key, *val;
612 {
613 register BUFHEAD *rbufp;
614 BUFHEAD *bufp, *save_bufp;
615 register u_short *bp;
616 register int n, ndx, off, size;
617 register char *kp;
618 u_short pageno;
619
620 #ifdef HASH_STATISTICS
621 hash_accesses++;
622 #endif
623
624 off = hashp->HBSIZE;
625 size = (int) key->size;
626 kp = (char *)key->data;
627 rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0);
628 if (!rbufp) {
629 return (ERROR);
630 }
631 save_bufp = rbufp;
632
633 /* Pin the bucket chain */
634 rbufp->flags |= BUF_PIN;
635 for (bp = (u_short *)rbufp->page, n = *bp++, ndx = 1; ndx < n;)
636 if (bp[1] >= REAL_KEY) {
637 /* Real key/data pair */
638 if (size == off - *bp &&
639 memcmp(kp, rbufp->page + *bp, size) == 0)
640 goto found;
641 off = bp[1];
642 #ifdef HASH_STATISTICS
643 hash_collisions++;
644 #endif
645 bp += 2;
646 ndx += 2;
647 } else if (bp[1] == OVFLPAGE) {
648 rbufp = __get_buf(hashp, *bp, rbufp, 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->HBSIZE;
658 } else if (bp[1] < REAL_KEY) {
659 if ((ndx =
660 __find_bigpair(hashp, rbufp, ndx, kp, size)) > 0)
661 goto found;
662 if (ndx == -2) {
663 bufp = rbufp;
664 if (!(pageno =
665 __find_last_page(hashp, &bufp))) {
666 ndx = 0;
667 rbufp = bufp;
668 break; /* FOR */
669 }
670 rbufp = __get_buf(hashp, pageno, bufp, 0);
671 if (!rbufp) {
672 save_bufp->flags &= ~BUF_PIN;
673 return (ERROR);
674 }
675 /* FOR LOOP INIT */
676 bp = (u_short *)rbufp->page;
677 n = *bp++;
678 ndx = 1;
679 off = hashp->HBSIZE;
680 } else {
681 save_bufp->flags &= ~BUF_PIN;
682 return (ERROR);
683 }
684 }
685
686 /* Not found */
687 switch (action) {
688 case HASH_PUT:
689 case HASH_PUTNEW:
690 if (__addel(hashp, rbufp, key, val)) {
691 save_bufp->flags &= ~BUF_PIN;
692 return (ERROR);
693 } else {
694 save_bufp->flags &= ~BUF_PIN;
695 return (SUCCESS);
696 }
697 case HASH_GET:
698 case HASH_DELETE:
699 default:
700 save_bufp->flags &= ~BUF_PIN;
701 return (ABNORMAL);
702 }
703
704 found:
705 switch (action) {
706 case HASH_PUTNEW:
707 save_bufp->flags &= ~BUF_PIN;
708 return (ABNORMAL);
709 case HASH_GET:
710 bp = (u_short *)rbufp->page;
711 if (bp[ndx + 1] < REAL_KEY) {
712 if (__big_return(hashp, rbufp, ndx, val, 0))
713 return (ERROR);
714 } else {
715 val->data = (u_char *)rbufp->page + (int)bp[ndx + 1];
716 val->size = bp[ndx] - bp[ndx + 1];
717 }
718 break;
719 case HASH_PUT:
720 if ((__delpair(hashp, rbufp, ndx)) ||
721 (__addel(hashp, rbufp, key, val))) {
722 save_bufp->flags &= ~BUF_PIN;
723 return (ERROR);
724 }
725 break;
726 case HASH_DELETE:
727 if (__delpair(hashp, rbufp, ndx))
728 return (ERROR);
729 break;
730 default:
731 abort();
732 }
733 save_bufp->flags &= ~BUF_PIN;
734 return (SUCCESS);
735 }
736
737 static int
hash_seq(dbp,key,data,flag)738 hash_seq(dbp, key, data, flag)
739 const DB *dbp;
740 DBT *key, *data;
741 u_int flag;
742 {
743 register u_int bucket;
744 register BUFHEAD *bufp;
745 HTAB *hashp;
746 u_short *bp, ndx;
747
748 hashp = (HTAB *)dbp->internal;
749 if (flag && flag != R_FIRST && flag != R_NEXT) {
750 hashp->errnumb = errno = EINVAL;
751 return (ERROR);
752 }
753 #ifdef HASH_STATISTICS
754 hash_accesses++;
755 #endif
756 if ((hashp->cbucket < 0) || (flag == R_FIRST)) {
757 hashp->cbucket = 0;
758 hashp->cndx = 1;
759 hashp->cpage = NULL;
760 }
761
762 for (bp = NULL; !bp || !bp[0]; ) {
763 if (!(bufp = hashp->cpage)) {
764 for (bucket = hashp->cbucket;
765 bucket <= hashp->MAX_BUCKET;
766 bucket++, hashp->cndx = 1) {
767 bufp = __get_buf(hashp, bucket, NULL, 0);
768 if (!bufp)
769 return (ERROR);
770 hashp->cpage = bufp;
771 bp = (u_short *)bufp->page;
772 if (bp[0]) {
773 break;
774 }
775 }
776 hashp->cbucket = bucket;
777 if (hashp->cbucket > hashp->MAX_BUCKET) {
778 hashp->cbucket = -1;
779 return (ABNORMAL);
780 }
781 } else
782 bp = (u_short *)hashp->cpage->page;
783
784 #ifdef DEBUG
785 assert(bp);
786 assert(bufp);
787 #endif
788 while (bp[hashp->cndx + 1] == OVFLPAGE) {
789 bufp = hashp->cpage =
790 __get_buf(hashp, bp[hashp->cndx], bufp, 0);
791 if (!bufp)
792 return (ERROR);
793 bp = (u_short *)(bufp->page);
794 hashp->cndx = 1;
795 }
796 if (!bp[0]) {
797 hashp->cpage = NULL;
798 ++hashp->cbucket;
799 }
800 }
801 ndx = hashp->cndx;
802 if (bp[ndx + 1] < REAL_KEY) {
803 if (__big_keydata(hashp, bufp, key, data, 1))
804 return (ERROR);
805 } else {
806 key->data = (u_char *)hashp->cpage->page + bp[ndx];
807 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->HBSIZE) - bp[ndx];
808 data->data = (u_char *)hashp->cpage->page + bp[ndx + 1];
809 data->size = bp[ndx] - bp[ndx + 1];
810 ndx += 2;
811 if (ndx > bp[0]) {
812 hashp->cpage = NULL;
813 hashp->cbucket++;
814 hashp->cndx = 1;
815 } else
816 hashp->cndx = ndx;
817 }
818 return (SUCCESS);
819 }
820
821 /********************************* UTILITIES ************************/
822
823 /*
824 * Returns:
825 * 0 ==> OK
826 * -1 ==> Error
827 */
828 extern int
829 __expand_table(hashp)
830 HTAB *hashp;
831 {
832 u_int old_bucket, new_bucket;
833 int dirsize, new_segnum, spare_ndx;
834
835 #ifdef HASH_STATISTICS
836 hash_expansions++;
837 #endif
838 new_bucket = ++hashp->MAX_BUCKET;
839 old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK);
840
841 new_segnum = new_bucket >> hashp->SSHIFT;
842
843 /* Check if we need a new segment */
844 if (new_segnum >= hashp->nsegs) {
845 /* Check if we need to expand directory */
846 if (new_segnum >= hashp->DSIZE) {
847 /* Reallocate directory */
848 dirsize = (int) (hashp->DSIZE * sizeof(SEGMENT *));
849 if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1))
850 return (-1);
851 hashp->DSIZE = dirsize << 1;
852 }
853 if (!(hashp->dir[new_segnum] =
854 (SEGMENT) calloc(hashp->SGSIZE, sizeof(SEGMENT))))
855 return (-1);
856 hashp->exsegs++;
857 hashp->nsegs++;
858 }
859 /*
860 * If the split point is increasing (MAX_BUCKET's log base 2
861 * * increases), we need to copy the current contents of the spare
862 * split bucket to the next bucket.
863 */
864 spare_ndx = __log2(hashp->MAX_BUCKET + 1);
865 if (spare_ndx > hashp->OVFL_POINT) {
866 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT];
867 hashp->OVFL_POINT = spare_ndx;
868 }
869
870 if (new_bucket > hashp->HIGH_MASK) {
871 /* Starting a new doubling */
872 hashp->LOW_MASK = hashp->HIGH_MASK;
873 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK;
874 }
875 /* Relocate records to the new bucket */
876 return (__split_page(hashp, old_bucket, new_bucket));
877 }
878
879 /*
880 * If realloc guarantees that the pointer is not destroyed if the realloc
881 * fails, then this routine can go away.
882 */
883 static void *
hash_realloc(p_ptr,oldsize,newsize)884 hash_realloc(p_ptr, oldsize, newsize)
885 SEGMENT **p_ptr;
886 int oldsize, newsize;
887 {
888 register void *p;
889
890 if (p = calloc(newsize, 1U)) {
891 memmove(p, *p_ptr, oldsize);
892 memset((char *)p + oldsize, 0, newsize - oldsize);
893 free(*p_ptr);
894 *p_ptr = p;
895 }
896 return (p);
897 }
898
899 extern u_int
900 __call_hash(hashp, k, len)
901 HTAB *hashp;
902 char *k;
903 int len;
904 {
905 int n, bucket;
906
907 n = hashp->hash((const u_char *) k, len);
908 bucket = n & hashp->HIGH_MASK;
909 if (bucket > hashp->MAX_BUCKET)
910 bucket = bucket & hashp->LOW_MASK;
911 return (bucket);
912 }
913
914 /*
915 * Allocate segment table. On error, destroy the table and set errno.
916 *
917 * Returns 0 on success
918 */
919 static int
alloc_segs(hashp,nsegs)920 alloc_segs(hashp, nsegs)
921 HTAB *hashp;
922 int nsegs;
923 {
924 register int i;
925 register SEGMENT store;
926
927 int save_errno;
928
929 if (!(hashp->dir = (SEGMENT *) calloc(hashp->DSIZE, sizeof(SEGMENT)))) {
930 save_errno = errno;
931 (void)hdestroy(hashp);
932 errno = save_errno;
933 return (-1);
934 }
935 /* Allocate segments */
936 store = (SEGMENT) calloc(nsegs << hashp->SSHIFT, sizeof(SEGMENT));
937 if (!store) {
938 save_errno = errno;
939 (void)hdestroy(hashp);
940 errno = save_errno;
941 return (-1);
942 }
943 for (i = 0; i < nsegs; i++, hashp->nsegs++)
944 hashp->dir[i] = &store[i << hashp->SSHIFT];
945 return (0);
946 }
947
948 #if BYTE_ORDER == LITTLE_ENDIAN
949 /*
950 * Hashp->hdr needs to be byteswapped.
951 */
952 static void
swap_header_copy(srcp,destp)953 swap_header_copy(srcp, destp)
954 HASHHDR *srcp, *destp;
955 {
956 int i;
957
958 BLSWAP_COPY(srcp->magic, destp->magic);
959 BLSWAP_COPY(srcp->version, destp->version);
960 BLSWAP_COPY(srcp->lorder, destp->lorder);
961 BLSWAP_COPY(srcp->bsize, destp->bsize);
962 BLSWAP_COPY(srcp->bshift, destp->bshift);
963 BLSWAP_COPY(srcp->dsize, destp->dsize);
964 BLSWAP_COPY(srcp->ssize, destp->ssize);
965 BLSWAP_COPY(srcp->sshift, destp->sshift);
966 BLSWAP_COPY(srcp->ovfl_point, destp->ovfl_point);
967 BLSWAP_COPY(srcp->last_freed, destp->last_freed);
968 BLSWAP_COPY(srcp->max_bucket, destp->max_bucket);
969 BLSWAP_COPY(srcp->high_mask, destp->high_mask);
970 BLSWAP_COPY(srcp->low_mask, destp->low_mask);
971 BLSWAP_COPY(srcp->ffactor, destp->ffactor);
972 BLSWAP_COPY(srcp->nkeys, destp->nkeys);
973 BLSWAP_COPY(srcp->hdrpages, destp->hdrpages);
974 BLSWAP_COPY(srcp->h_charkey, destp->h_charkey);
975 for (i = 0; i < NCACHED; i++) {
976 BLSWAP_COPY(srcp->spares[i], destp->spares[i]);
977 BSSWAP_COPY(srcp->bitmaps[i], destp->bitmaps[i]);
978 }
979 }
980
981 static void
swap_header(hashp)982 swap_header(hashp)
983 HTAB *hashp;
984 {
985 HASHHDR *hdrp;
986 int i;
987
988 hdrp = &hashp->hdr;
989
990 BLSWAP(hdrp->magic);
991 BLSWAP(hdrp->version);
992 BLSWAP(hdrp->lorder);
993 BLSWAP(hdrp->bsize);
994 BLSWAP(hdrp->bshift);
995 BLSWAP(hdrp->dsize);
996 BLSWAP(hdrp->ssize);
997 BLSWAP(hdrp->sshift);
998 BLSWAP(hdrp->ovfl_point);
999 BLSWAP(hdrp->last_freed);
1000 BLSWAP(hdrp->max_bucket);
1001 BLSWAP(hdrp->high_mask);
1002 BLSWAP(hdrp->low_mask);
1003 BLSWAP(hdrp->ffactor);
1004 BLSWAP(hdrp->nkeys);
1005 BLSWAP(hdrp->hdrpages);
1006 BLSWAP(hdrp->h_charkey);
1007 for (i = 0; i < NCACHED; i++) {
1008 BLSWAP(hdrp->spares[i]);
1009 BSSWAP(hdrp->bitmaps[i]);
1010 }
1011 }
1012 #endif
1013