]>
git.saurik.com Git - apple/libc.git/blob - db/hash/FreeBSD/hash_bigkey.c
487df5b36d0280fe8e3d35ab4bab6b9decd4734c
2 * Copyright (c) 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
37 #if defined(LIBC_SCCS) && !defined(lint)
38 static char sccsid
[] = "@(#)hash_bigkey.c 8.3 (Berkeley) 5/31/94";
39 #endif /* LIBC_SCCS and not lint */
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD: src/lib/libc/db/hash/hash_bigkey.c,v 1.5 2003/02/16 17:29:09 nectar Exp $");
46 * Big key/data handling for the hashing package.
61 #include <sys/param.h>
77 static int collect_key(HTAB
*, BUFHEAD
*, int, DBT
*, int);
78 static int collect_data(HTAB
*, BUFHEAD
*, int, int);
83 * You need to do an insert and the key/data pair is too big
90 __big_insert(hashp
, bufp
, key
, val
)
96 int key_size
, n
, val_size
;
97 u_int16_t space
, move_bytes
, off
;
98 char *cp
, *key_data
, *val_data
;
100 cp
= bufp
->page
; /* Character pointer of p. */
103 key_data
= (char *)key
->data
;
104 key_size
= key
->size
;
105 val_data
= (char *)val
->data
;
106 val_size
= val
->size
;
108 /* First move the Key */
109 for (space
= FREESPACE(p
) - BIGOVERHEAD
; key_size
;
110 space
= FREESPACE(p
) - BIGOVERHEAD
) {
111 move_bytes
= MIN(space
, key_size
);
112 off
= OFFSET(p
) - move_bytes
;
113 memmove(cp
+ off
, key_data
, move_bytes
);
114 key_size
-= move_bytes
;
115 key_data
+= move_bytes
;
119 FREESPACE(p
) = off
- PAGE_META(n
);
122 bufp
= __add_ovflpage(hashp
, bufp
);
128 move_bytes
= MIN(FREESPACE(p
), val_size
);
129 off
= OFFSET(p
) - move_bytes
;
131 memmove(cp
+ off
, val_data
, move_bytes
);
132 val_data
+= move_bytes
;
133 val_size
-= move_bytes
;
134 p
[n
- 2] = FULL_KEY_DATA
;
135 FREESPACE(p
) = FREESPACE(p
) - move_bytes
;
140 p
= (u_int16_t
*)bufp
->page
;
142 bufp
->flags
|= BUF_MOD
;
145 /* Now move the data */
146 for (space
= FREESPACE(p
) - BIGOVERHEAD
; val_size
;
147 space
= FREESPACE(p
) - BIGOVERHEAD
) {
148 move_bytes
= MIN(space
, val_size
);
150 * Here's the hack to make sure that if the data ends on the
151 * same page as the key ends, FREESPACE is at least one.
153 if (space
== val_size
&& val_size
== val
->size
)
155 off
= OFFSET(p
) - move_bytes
;
156 memmove(cp
+ off
, val_data
, move_bytes
);
157 val_size
-= move_bytes
;
158 val_data
+= move_bytes
;
162 FREESPACE(p
) = off
- PAGE_META(n
);
166 bufp
= __add_ovflpage(hashp
, bufp
);
172 p
[n
] = FULL_KEY_DATA
;
173 bufp
->flags
|= BUF_MOD
;
179 * Called when bufp's page contains a partial key (index should be 1)
181 * All pages in the big key/data pair except bufp are freed. We cannot
182 * free bufp because the page pointing to it is lost and we can't get rid
190 __big_delete(hashp
, bufp
)
194 BUFHEAD
*last_bfp
, *rbufp
;
195 u_int16_t
*bp
, pageno
;
200 bp
= (u_int16_t
*)bufp
->page
;
204 while (!key_done
|| (bp
[2] != FULL_KEY_DATA
)) {
205 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
)
209 * If there is freespace left on a FULL_KEY_DATA page, then
210 * the data is short and fits entirely on this page, and this
213 if (bp
[2] == FULL_KEY_DATA
&& FREESPACE(bp
))
215 pageno
= bp
[bp
[0] - 1];
216 rbufp
->flags
|= BUF_MOD
;
217 rbufp
= __get_buf(hashp
, pageno
, rbufp
, 0);
219 __free_ovflpage(hashp
, last_bfp
);
222 return (-1); /* Error. */
223 bp
= (u_int16_t
*)rbufp
->page
;
227 * If we get here then rbufp points to the last page of the big
228 * key/data pair. Bufp points to the first one -- it should now be
229 * empty pointing to the next page after this pair. Can't free it
230 * because we don't have the page pointing to it.
233 /* This is information from the last page of the pair. */
237 /* Now, bp is the first page of the pair. */
238 bp
= (u_int16_t
*)bufp
->page
;
240 /* There is an overflow page. */
243 bufp
->ovfl
= rbufp
->ovfl
;
245 /* This is the last page. */
249 FREESPACE(bp
) = hashp
->BSIZE
- PAGE_META(n
);
250 OFFSET(bp
) = hashp
->BSIZE
- 1;
252 bufp
->flags
|= BUF_MOD
;
254 __free_ovflpage(hashp
, rbufp
);
255 if (last_bfp
!= rbufp
)
256 __free_ovflpage(hashp
, last_bfp
);
264 * -1 = get next overflow page
265 * -2 means key not found and this is big key/data
269 __find_bigpair(hashp
, bufp
, ndx
, key
, size
)
282 bp
= (u_int16_t
*)bufp
->page
;
287 for (bytes
= hashp
->BSIZE
- bp
[ndx
];
288 bytes
<= size
&& bp
[ndx
+ 1] == PARTIAL_KEY
;
289 bytes
= hashp
->BSIZE
- bp
[ndx
]) {
290 if (memcmp(p
+ bp
[ndx
], kkey
, bytes
))
294 bufp
= __get_buf(hashp
, bp
[ndx
+ 2], bufp
, 0);
302 if (bytes
!= ksize
|| memcmp(p
+ bp
[ndx
], kkey
, bytes
)) {
303 #ifdef HASH_STATISTICS
312 * Given the buffer pointer of the first overflow page of a big pair,
313 * find the end of the big pair
315 * This will set bpp to the buffer header of the last page of the big pair.
316 * It will return the pageno of the overflow page following the last page
317 * of the pair; 0 if there isn't any (i.e. big pair is the last key in the
321 __find_last_page(hashp
, bpp
)
326 u_int16_t
*bp
, pageno
;
330 bp
= (u_int16_t
*)bufp
->page
;
335 * This is the last page if: the tag is FULL_KEY_DATA and
336 * either only 2 entries OVFLPAGE marker is explicit there
337 * is freespace on the page.
339 if (bp
[2] == FULL_KEY_DATA
&&
340 ((n
== 2) || (bp
[n
] == OVFLPAGE
) || (FREESPACE(bp
))))
344 bufp
= __get_buf(hashp
, pageno
, bufp
, 0);
346 return (0); /* Need to indicate an error! */
347 bp
= (u_int16_t
*)bufp
->page
;
358 * Return the data for the key/data pair that begins on this page at this
359 * index (index should always be 1).
362 __big_return(hashp
, bufp
, ndx
, val
, set_current
)
370 u_int16_t
*bp
, len
, off
, save_addr
;
373 bp
= (u_int16_t
*)bufp
->page
;
374 while (bp
[ndx
+ 1] == PARTIAL_KEY
) {
375 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
378 bp
= (u_int16_t
*)bufp
->page
;
382 if (bp
[ndx
+ 1] == FULL_KEY
) {
383 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
386 bp
= (u_int16_t
*)bufp
->page
;
388 save_addr
= save_p
->addr
;
392 if (!FREESPACE(bp
)) {
394 * This is a hack. We can't distinguish between
395 * FULL_KEY_DATA that contains complete data or
396 * incomplete data, so we require that if the data
397 * is complete, there is at least 1 byte of free
403 save_addr
= bufp
->addr
;
404 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
407 bp
= (u_int16_t
*)bufp
->page
;
409 /* The data is all on one page. */
412 val
->data
= (u_char
*)tp
+ off
;
413 val
->size
= bp
[1] - off
;
415 if (bp
[0] == 2) { /* No more buckets in
421 hashp
->cpage
= __get_buf(hashp
,
422 bp
[bp
[0] - 1], bufp
, 0);
427 hashp
->cpage
->page
)[0]) {
436 val
->size
= collect_data(hashp
, bufp
, (int)len
, set_current
);
439 if (save_p
->addr
!= save_addr
) {
440 /* We are pretty short on buffers. */
441 errno
= EINVAL
; /* OUT OF BUFFERS */
444 memmove(hashp
->tmp_buf
, (save_p
->page
) + off
, len
);
445 val
->data
= (u_char
*)hashp
->tmp_buf
;
449 * Count how big the total datasize is by recursing through the pages. Then
450 * allocate a buffer and copy the data as you recurse up.
453 collect_data(hashp
, bufp
, len
, set
)
466 mylen
= hashp
->BSIZE
- bp
[1];
467 save_addr
= bufp
->addr
;
469 if (bp
[2] == FULL_KEY_DATA
) { /* End of Data */
470 totlen
= len
+ mylen
;
472 free(hashp
->tmp_buf
);
473 if ((hashp
->tmp_buf
= (char *)malloc(totlen
)) == NULL
)
477 if (bp
[0] == 2) { /* No more buckets in chain */
482 __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
485 else if (!((u_int16_t
*)hashp
->cpage
->page
)[0]) {
492 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
493 if (!xbp
|| ((totlen
=
494 collect_data(hashp
, xbp
, len
+ mylen
, set
)) < 1))
497 if (bufp
->addr
!= save_addr
) {
498 errno
= EINVAL
; /* Out of buffers. */
501 memmove(&hashp
->tmp_buf
[len
], (bufp
->page
) + bp
[1], mylen
);
506 * Fill in the key and data for this big pair.
509 __big_keydata(hashp
, bufp
, key
, val
, set
)
515 key
->size
= collect_key(hashp
, bufp
, 0, val
, set
);
518 key
->data
= (u_char
*)hashp
->tmp_key
;
523 * Count how big the total key size is by recursing through the pages. Then
524 * collect the data, allocate a buffer and copy the key as you recurse up.
527 collect_key(hashp
, bufp
, len
, val
, set
)
537 u_int16_t
*bp
, save_addr
;
541 mylen
= hashp
->BSIZE
- bp
[1];
543 save_addr
= bufp
->addr
;
544 totlen
= len
+ mylen
;
545 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
) { /* End of Key. */
546 if (hashp
->tmp_key
!= NULL
)
547 free(hashp
->tmp_key
);
548 if ((hashp
->tmp_key
= (char *)malloc(totlen
)) == NULL
)
550 if (__big_return(hashp
, bufp
, 1, val
, set
))
553 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
554 if (!xbp
|| ((totlen
=
555 collect_key(hashp
, xbp
, totlen
, val
, set
)) < 1))
558 if (bufp
->addr
!= save_addr
) {
559 errno
= EINVAL
; /* MIS -- OUT OF BUFFERS */
562 memmove(&hashp
->tmp_key
[len
], (bufp
->page
) + bp
[1], mylen
);
572 __big_split(hashp
, op
, np
, big_keyp
, addr
, obucket
, ret
)
574 BUFHEAD
*op
; /* Pointer to where to put keys that go in old bucket */
575 BUFHEAD
*np
; /* Pointer to new bucket page */
576 /* Pointer to first page containing the big key/data */
578 int addr
; /* Address of big_keyp */
579 u_int32_t obucket
;/* Old Bucket */
587 u_int16_t free_space
, n
, off
;
591 /* Now figure out where the big key/data goes */
592 if (__big_keydata(hashp
, big_keyp
, &key
, &val
, 0))
594 change
= (__call_hash(hashp
, key
.data
, key
.size
) != obucket
);
596 if ( (ret
->next_addr
= __find_last_page(hashp
, &big_keyp
)) ) {
598 __get_buf(hashp
, ret
->next_addr
, big_keyp
, 0)))
603 /* Now make one of np/op point to the big key/data pair */
605 assert(np
->ovfl
== NULL
);
612 tmpp
->flags
|= BUF_MOD
;
614 (void)fprintf(stderr
,
615 "BIG_SPLIT: %d->ovfl was %d is now %d\n", tmpp
->addr
,
616 (tmpp
->ovfl
? tmpp
->ovfl
->addr
: 0), (bp
? bp
->addr
: 0));
618 tmpp
->ovfl
= bp
; /* one of op/np point to big_keyp */
619 tp
= (u_int16_t
*)tmpp
->page
;
621 assert(FREESPACE(tp
) >= OVFLSIZE
);
625 free_space
= FREESPACE(tp
);
626 tp
[++n
] = (u_int16_t
)addr
;
630 FREESPACE(tp
) = free_space
- OVFLSIZE
;
633 * Finally, set the new and old return values. BIG_KEYP contains a
634 * pointer to the last page of the big key_data pair. Make sure that
635 * big_keyp has no following page (2 elements) or create an empty
642 tp
= (u_int16_t
*)big_keyp
->page
;
643 big_keyp
->flags
|= BUF_MOD
;
646 * There may be either one or two offsets on this page. If
647 * there is one, then the overflow page is linked on normally
648 * and tp[4] is OVFLPAGE. If there are two, tp[4] contains
649 * the second offset and needs to get stuffed in after the
650 * next overflow page is added.
653 free_space
= FREESPACE(tp
);
656 FREESPACE(tp
) = free_space
+ OVFLSIZE
;
658 tmpp
= __add_ovflpage(hashp
, big_keyp
);