]>
git.saurik.com Git - apple/libc.git/blob - db/hash/hash_bigkey.c
91ed98a789fca397bdd4d30e23b4e06e19a5c50e
2 * Copyright (c) 2003 Apple Computer, Inc. All rights reserved.
4 * @APPLE_LICENSE_HEADER_START@
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. Please obtain a copy of the License at
10 * http://www.opensource.apple.com/apsl/ and read it before using this
13 * The Original Code and all software distributed under the License are
14 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
15 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
16 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
18 * Please see the License for the specific language governing rights and
19 * limitations under the License.
21 * @APPLE_LICENSE_HEADER_END@
24 * Copyright (c) 1990, 1993, 1994
25 * The Regents of the University of California. All rights reserved.
27 * This code is derived from software contributed to Berkeley by
30 * Redistribution and use in source and binary forms, with or without
31 * modification, are permitted provided that the following conditions
33 * 1. Redistributions of source code must retain the above copyright
34 * notice, this list of conditions and the following disclaimer.
35 * 2. Redistributions in binary form must reproduce the above copyright
36 * notice, this list of conditions and the following disclaimer in the
37 * documentation and/or other materials provided with the distribution.
38 * 3. All advertising materials mentioning features or use of this software
39 * must display the following acknowledgement:
40 * This product includes software developed by the University of
41 * California, Berkeley and its contributors.
42 * 4. Neither the name of the University nor the names of its contributors
43 * may be used to endorse or promote products derived from this software
44 * without specific prior written permission.
46 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
47 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
48 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
49 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
50 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
51 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
52 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
53 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
54 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
55 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 #if defined(LIBC_SCCS) && !defined(lint)
60 static char sccsid
[] = "@(#)hash_bigkey.c 8.3 (Berkeley) 5/31/94";
61 #endif /* LIBC_SCCS and not lint */
62 #include <sys/cdefs.h>
67 * Big key/data handling for the hashing package.
82 #include <sys/param.h>
98 static int collect_key(HTAB
*, BUFHEAD
*, int, DBT
*, int);
99 static int collect_data(HTAB
*, BUFHEAD
*, int, int);
104 * You need to do an insert and the key/data pair is too big
111 __big_insert(hashp
, bufp
, key
, val
)
114 const DBT
*key
, *val
;
117 int key_size
, n
, val_size
;
118 u_int16_t space
, move_bytes
, off
;
119 char *cp
, *key_data
, *val_data
;
121 cp
= bufp
->page
; /* Character pointer of p. */
124 key_data
= (char *)key
->data
;
125 key_size
= key
->size
;
126 val_data
= (char *)val
->data
;
127 val_size
= val
->size
;
129 /* First move the Key */
130 for (space
= FREESPACE(p
) - BIGOVERHEAD
; key_size
;
131 space
= FREESPACE(p
) - BIGOVERHEAD
) {
132 move_bytes
= MIN(space
, key_size
);
133 off
= OFFSET(p
) - move_bytes
;
134 memmove(cp
+ off
, key_data
, move_bytes
);
135 key_size
-= move_bytes
;
136 key_data
+= move_bytes
;
140 FREESPACE(p
) = off
- PAGE_META(n
);
143 bufp
= __add_ovflpage(hashp
, bufp
);
149 move_bytes
= MIN(FREESPACE(p
), val_size
);
150 off
= OFFSET(p
) - move_bytes
;
152 memmove(cp
+ off
, val_data
, move_bytes
);
153 val_data
+= move_bytes
;
154 val_size
-= move_bytes
;
155 p
[n
- 2] = FULL_KEY_DATA
;
156 FREESPACE(p
) = FREESPACE(p
) - move_bytes
;
161 p
= (u_int16_t
*)bufp
->page
;
163 bufp
->flags
|= BUF_MOD
;
166 /* Now move the data */
167 for (space
= FREESPACE(p
) - BIGOVERHEAD
; val_size
;
168 space
= FREESPACE(p
) - BIGOVERHEAD
) {
169 move_bytes
= MIN(space
, val_size
);
171 * Here's the hack to make sure that if the data ends on the
172 * same page as the key ends, FREESPACE is at least one.
174 if (space
== val_size
&& val_size
== val
->size
)
176 off
= OFFSET(p
) - move_bytes
;
177 memmove(cp
+ off
, val_data
, move_bytes
);
178 val_size
-= move_bytes
;
179 val_data
+= move_bytes
;
183 FREESPACE(p
) = off
- PAGE_META(n
);
187 bufp
= __add_ovflpage(hashp
, bufp
);
193 p
[n
] = FULL_KEY_DATA
;
194 bufp
->flags
|= BUF_MOD
;
200 * Called when bufp's page contains a partial key (index should be 1)
202 * All pages in the big key/data pair except bufp are freed. We cannot
203 * free bufp because the page pointing to it is lost and we can't get rid
211 __big_delete(hashp
, bufp
)
215 BUFHEAD
*last_bfp
, *rbufp
;
216 u_int16_t
*bp
, pageno
;
221 bp
= (u_int16_t
*)bufp
->page
;
225 while (!key_done
|| (bp
[2] != FULL_KEY_DATA
)) {
226 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
)
230 * If there is freespace left on a FULL_KEY_DATA page, then
231 * the data is short and fits entirely on this page, and this
234 if (bp
[2] == FULL_KEY_DATA
&& FREESPACE(bp
))
236 pageno
= bp
[bp
[0] - 1];
237 rbufp
->flags
|= BUF_MOD
;
238 rbufp
= __get_buf(hashp
, pageno
, rbufp
, 0);
240 __free_ovflpage(hashp
, last_bfp
);
243 return (-1); /* Error. */
244 bp
= (u_int16_t
*)rbufp
->page
;
248 * If we get here then rbufp points to the last page of the big
249 * key/data pair. Bufp points to the first one -- it should now be
250 * empty pointing to the next page after this pair. Can't free it
251 * because we don't have the page pointing to it.
254 /* This is information from the last page of the pair. */
258 /* Now, bp is the first page of the pair. */
259 bp
= (u_int16_t
*)bufp
->page
;
261 /* There is an overflow page. */
264 bufp
->ovfl
= rbufp
->ovfl
;
266 /* This is the last page. */
270 FREESPACE(bp
) = hashp
->BSIZE
- PAGE_META(n
);
271 OFFSET(bp
) = hashp
->BSIZE
- 1;
273 bufp
->flags
|= BUF_MOD
;
275 __free_ovflpage(hashp
, rbufp
);
276 if (last_bfp
!= rbufp
)
277 __free_ovflpage(hashp
, last_bfp
);
285 * -1 = get next overflow page
286 * -2 means key not found and this is big key/data
290 __find_bigpair(hashp
, bufp
, ndx
, key
, size
)
303 bp
= (u_int16_t
*)bufp
->page
;
308 for (bytes
= hashp
->BSIZE
- bp
[ndx
];
309 bytes
<= size
&& bp
[ndx
+ 1] == PARTIAL_KEY
;
310 bytes
= hashp
->BSIZE
- bp
[ndx
]) {
311 if (memcmp(p
+ bp
[ndx
], kkey
, bytes
))
315 bufp
= __get_buf(hashp
, bp
[ndx
+ 2], bufp
, 0);
323 if (bytes
!= ksize
|| memcmp(p
+ bp
[ndx
], kkey
, bytes
)) {
324 #ifdef HASH_STATISTICS
333 * Given the buffer pointer of the first overflow page of a big pair,
334 * find the end of the big pair
336 * This will set bpp to the buffer header of the last page of the big pair.
337 * It will return the pageno of the overflow page following the last page
338 * of the pair; 0 if there isn't any (i.e. big pair is the last key in the
342 __find_last_page(hashp
, bpp
)
347 u_int16_t
*bp
, pageno
;
351 bp
= (u_int16_t
*)bufp
->page
;
356 * This is the last page if: the tag is FULL_KEY_DATA and
357 * either only 2 entries OVFLPAGE marker is explicit there
358 * is freespace on the page.
360 if (bp
[2] == FULL_KEY_DATA
&&
361 ((n
== 2) || (bp
[n
] == OVFLPAGE
) || (FREESPACE(bp
))))
365 bufp
= __get_buf(hashp
, pageno
, bufp
, 0);
367 return (0); /* Need to indicate an error! */
368 bp
= (u_int16_t
*)bufp
->page
;
379 * Return the data for the key/data pair that begins on this page at this
380 * index (index should always be 1).
383 __big_return(hashp
, bufp
, ndx
, val
, set_current
)
391 u_int16_t
*bp
, len
, off
, save_addr
;
394 bp
= (u_int16_t
*)bufp
->page
;
395 while (bp
[ndx
+ 1] == PARTIAL_KEY
) {
396 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
399 bp
= (u_int16_t
*)bufp
->page
;
403 if (bp
[ndx
+ 1] == FULL_KEY
) {
404 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
407 bp
= (u_int16_t
*)bufp
->page
;
409 save_addr
= save_p
->addr
;
413 if (!FREESPACE(bp
)) {
415 * This is a hack. We can't distinguish between
416 * FULL_KEY_DATA that contains complete data or
417 * incomplete data, so we require that if the data
418 * is complete, there is at least 1 byte of free
424 save_addr
= bufp
->addr
;
425 bufp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
428 bp
= (u_int16_t
*)bufp
->page
;
430 /* The data is all on one page. */
433 val
->data
= (u_char
*)tp
+ off
;
434 val
->size
= bp
[1] - off
;
436 if (bp
[0] == 2) { /* No more buckets in
442 hashp
->cpage
= __get_buf(hashp
,
443 bp
[bp
[0] - 1], bufp
, 0);
448 hashp
->cpage
->page
)[0]) {
457 val
->size
= collect_data(hashp
, bufp
, (int)len
, set_current
);
460 if (save_p
->addr
!= save_addr
) {
461 /* We are pretty short on buffers. */
462 errno
= EINVAL
; /* OUT OF BUFFERS */
465 memmove(hashp
->tmp_buf
, (save_p
->page
) + off
, len
);
466 val
->data
= (u_char
*)hashp
->tmp_buf
;
470 * Count how big the total datasize is by recursing through the pages. Then
471 * allocate a buffer and copy the data as you recurse up.
474 collect_data(hashp
, bufp
, len
, set
)
487 mylen
= hashp
->BSIZE
- bp
[1];
488 save_addr
= bufp
->addr
;
490 if (bp
[2] == FULL_KEY_DATA
) { /* End of Data */
491 totlen
= len
+ mylen
;
493 free(hashp
->tmp_buf
);
494 if ((hashp
->tmp_buf
= (char *)malloc(totlen
)) == NULL
)
498 if (bp
[0] == 2) { /* No more buckets in chain */
503 __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
506 else if (!((u_int16_t
*)hashp
->cpage
->page
)[0]) {
513 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
514 if (!xbp
|| ((totlen
=
515 collect_data(hashp
, xbp
, len
+ mylen
, set
)) < 1))
518 if (bufp
->addr
!= save_addr
) {
519 errno
= EINVAL
; /* Out of buffers. */
522 memmove(&hashp
->tmp_buf
[len
], (bufp
->page
) + bp
[1], mylen
);
527 * Fill in the key and data for this big pair.
530 __big_keydata(hashp
, bufp
, key
, val
, set
)
536 key
->size
= collect_key(hashp
, bufp
, 0, val
, set
);
539 key
->data
= (u_char
*)hashp
->tmp_key
;
544 * Count how big the total key size is by recursing through the pages. Then
545 * collect the data, allocate a buffer and copy the key as you recurse up.
548 collect_key(hashp
, bufp
, len
, val
, set
)
558 u_int16_t
*bp
, save_addr
;
562 mylen
= hashp
->BSIZE
- bp
[1];
564 save_addr
= bufp
->addr
;
565 totlen
= len
+ mylen
;
566 if (bp
[2] == FULL_KEY
|| bp
[2] == FULL_KEY_DATA
) { /* End of Key. */
567 if (hashp
->tmp_key
!= NULL
)
568 free(hashp
->tmp_key
);
569 if ((hashp
->tmp_key
= (char *)malloc(totlen
)) == NULL
)
571 if (__big_return(hashp
, bufp
, 1, val
, set
))
574 xbp
= __get_buf(hashp
, bp
[bp
[0] - 1], bufp
, 0);
575 if (!xbp
|| ((totlen
=
576 collect_key(hashp
, xbp
, totlen
, val
, set
)) < 1))
579 if (bufp
->addr
!= save_addr
) {
580 errno
= EINVAL
; /* MIS -- OUT OF BUFFERS */
583 memmove(&hashp
->tmp_key
[len
], (bufp
->page
) + bp
[1], mylen
);
593 __big_split(hashp
, op
, np
, big_keyp
, addr
, obucket
, ret
)
595 BUFHEAD
*op
; /* Pointer to where to put keys that go in old bucket */
596 BUFHEAD
*np
; /* Pointer to new bucket page */
597 /* Pointer to first page containing the big key/data */
599 int addr
; /* Address of big_keyp */
600 u_int32_t obucket
;/* Old Bucket */
608 u_int16_t free_space
, n
, off
;
612 /* Now figure out where the big key/data goes */
613 if (__big_keydata(hashp
, big_keyp
, &key
, &val
, 0))
615 change
= (__call_hash(hashp
, key
.data
, key
.size
) != obucket
);
617 if ( (ret
->next_addr
= __find_last_page(hashp
, &big_keyp
)) ) {
619 __get_buf(hashp
, ret
->next_addr
, big_keyp
, 0)))
624 /* Now make one of np/op point to the big key/data pair */
626 assert(np
->ovfl
== NULL
);
633 tmpp
->flags
|= BUF_MOD
;
635 (void)fprintf(stderr
,
636 "BIG_SPLIT: %d->ovfl was %d is now %d\n", tmpp
->addr
,
637 (tmpp
->ovfl
? tmpp
->ovfl
->addr
: 0), (bp
? bp
->addr
: 0));
639 tmpp
->ovfl
= bp
; /* one of op/np point to big_keyp */
640 tp
= (u_int16_t
*)tmpp
->page
;
642 assert(FREESPACE(tp
) >= OVFLSIZE
);
646 free_space
= FREESPACE(tp
);
647 tp
[++n
] = (u_int16_t
)addr
;
651 FREESPACE(tp
) = free_space
- OVFLSIZE
;
654 * Finally, set the new and old return values. BIG_KEYP contains a
655 * pointer to the last page of the big key_data pair. Make sure that
656 * big_keyp has no following page (2 elements) or create an empty
663 tp
= (u_int16_t
*)big_keyp
->page
;
664 big_keyp
->flags
|= BUF_MOD
;
667 * There may be either one or two offsets on this page. If
668 * there is one, then the overflow page is linked on normally
669 * and tp[4] is OVFLPAGE. If there are two, tp[4] contains
670 * the second offset and needs to get stuffed in after the
671 * next overflow page is added.
674 free_space
= FREESPACE(tp
);
677 FREESPACE(tp
) = free_space
+ OVFLSIZE
;
679 tmpp
= __add_ovflpage(hashp
, big_keyp
);