xref: /original-bsd/lib/libc/db/hash/hash_bigkey.c (revision e4165910)
1 /*-
2  * Copyright (c) 1990 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Margo Seltzer.
7  *
8  * %sccs.include.redist.c%
9  */
10 
11 #if defined(LIBC_SCCS) && !defined(lint)
12 static char sccsid[] = "@(#)hash_bigkey.c	5.5 (Berkeley) 03/12/91";
13 #endif /* LIBC_SCCS and not lint */
14 
15 /******************************************************************************
16 
17 PACKAGE: hash
18 
19 DESCRIPTION:
20 	Big key/data handling for the hashing package.
21 
22 ROUTINES:
23     External
24 	__big_keydata
25 	__big_split
26 	__big_insert
27 	__big_return
28 	__big_delete
29 	__find_last_page
30     Internal
31 	collect_key
32 	collect_data
33 ******************************************************************************/
34 /* Includes */
35 #include <sys/param.h>
36 #include <assert.h>
37 #include <errno.h>
38 #include <db.h>
39 #include <stdio.h>
40 #include <stdlib.h>
41 #include <string.h>
42 #include "hash.h"
43 #include "page.h"
44 
45 /* Externals */
46 /* buf.c */
47 extern BUFHEAD *__get_buf();
48 
49 /* dynahash.c */
50 extern	u_int call_hash();
51 
52 /* page.c */
53 extern BUFHEAD *__add_ovflpage();
54 
55 /* My externals */
56 extern int __big_keydata();
57 extern int __big_split();
58 extern int __big_insert();
59 extern int __big_return();
60 extern int __big_delete();
61 extern u_short __find_last_page();
62 extern int __find_bigpair();
63 
64 /* My internals */
65 static int collect_key();
66 static int collect_data();
67 
68 #ifdef HASH_STATISTICS
69 extern long hash_accesses, hash_collisions, hash_expansions, hash_overflows;
70 #endif
71 /*
72 Big_insert
73 
74 You need to do an insert and the key/data pair is too big
75 0 ==> OK
76 -1 ==> ERROR
77 */
78 extern int
79 __big_insert ( bufp, key, val )
80 BUFHEAD *bufp;
81 DBT	*key, *val;
82 {
83     char	*cp = bufp->page;	/* Character pointer of p */
84     register u_short	*p = (u_short *)cp;
85     char	*key_data, *val_data;
86     int		key_size, val_size;
87     int		n;
88     u_short	space, move_bytes, off;
89 
90     key_data = (char *)key->data;
91     key_size = key->size;
92     val_data = (char *)val->data;
93     val_size = val->size;
94 
95     /* First move the Key */
96     for ( space = FREESPACE(p) - BIGOVERHEAD;
97 	  key_size;
98 	  space = FREESPACE(p) - BIGOVERHEAD ) {
99 	move_bytes = MIN(space, key_size);
100 	off = OFFSET(p) - move_bytes;
101 	bcopy (key_data, cp+off, move_bytes );
102 	key_size -= move_bytes;
103 	key_data += move_bytes;
104 	n = p[0];
105 	p[++n] = off;
106 	p[0] = ++n;
107 	FREESPACE(p) = off - PAGE_META(n);
108 	OFFSET(p) = off;
109 	p[n] = PARTIAL_KEY;
110 	bufp = __add_ovflpage(bufp);
111 	if ( !bufp ) {
112 	    return(-1);
113 	}
114 	n = p[0];
115 	if ( !key_size ) {
116 	    if ( FREESPACE(p) ) {
117 		move_bytes = MIN (FREESPACE(p), val_size);
118 		off = OFFSET(p) - move_bytes;
119 		p[n] = off;
120 		bcopy ( val_data, cp + off, move_bytes );
121 		val_data += move_bytes;
122 		val_size -= move_bytes;
123 		p[n-2] = FULL_KEY_DATA;
124 		FREESPACE(p) = FREESPACE(p) - move_bytes;
125 		OFFSET(p) = off;
126 	    }
127 	    else p[n-2] = FULL_KEY;
128 	}
129 	p = (u_short *)bufp->page;
130 	cp = bufp->page;
131 	bufp->flags |= BUF_MOD;
132     }
133 
134     /* Now move the data */
135     for ( space = FREESPACE(p) - BIGOVERHEAD;
136 	  val_size;
137 	  space = FREESPACE(p) - BIGOVERHEAD ) {
138 	move_bytes = MIN(space, val_size);
139 	/*
140 	    Here's the hack to make sure that if the data ends
141 	    on the same page as the key ends, FREESPACE is
142 	    at least one
143 	*/
144 	if ( space == val_size && val_size == val->size ) {
145 	    move_bytes--;
146 	}
147 	off = OFFSET(p) - move_bytes;
148 	bcopy (val_data, cp+off, move_bytes );
149 	val_size -= move_bytes;
150 	val_data += move_bytes;
151 	n = p[0];
152 	p[++n] = off;
153 	p[0] = ++n;
154 	FREESPACE(p) = off - PAGE_META(n);
155 	OFFSET(p) = off;
156 	if ( val_size ) {
157 	    p[n] = FULL_KEY;
158 	    bufp = __add_ovflpage (bufp);
159 	    if ( !bufp ) {
160 		return(-1);
161 	    }
162 	    cp = bufp->page;
163 	    p = (u_short *)cp;
164 	} else {
165 	    p[n] = FULL_KEY_DATA;
166 	}
167 	bufp->flags |= BUF_MOD;
168     }
169     return(0);
170 }
171 
172 /*
173     Called when bufp's page  contains a partial key (index should be 1)
174 
175     All pages in the big key/data pair except bufp are freed.  We cannot
176     free bufp because the page pointing to it is lost and we can't
177     get rid of its pointer.
178 
179     Returns 0 => OK
180 	    -1 => ERROR
181 */
182 extern int
183 __big_delete (bufp, ndx)
184 BUFHEAD	*bufp;
185 int	ndx;
186 {
187 	register	BUFHEAD		*rbufp = bufp;
188 	register	BUFHEAD		*last_bfp = NULL;
189 	char	*cp;
190 	u_short	*bp = (u_short *)bufp->page;
191 	u_short	*xbp;
192 	u_short	pageno = 0;
193 	u_short	off, free_sp;
194 	int	key_done = 0;
195 	int	n;
196 
197 	while (!key_done || (bp[2] != FULL_KEY_DATA)) {
198 	    if ( bp[2] == FULL_KEY || bp[2] == FULL_KEY_DATA ) key_done = 1;
199 
200 	    /*
201 		If there is freespace left on a FULL_KEY_DATA page,
202 		then the data is short and fits entirely on this
203 		page, and this is the last page.
204 	    */
205 	    if ( bp[2] == FULL_KEY_DATA && FREESPACE(bp) ) break;
206 	    pageno = bp[bp[0]-1];
207 	    rbufp->flags |= BUF_MOD;
208 	    rbufp = __get_buf ( pageno, rbufp, 0 );
209 	    if ( last_bfp ) __free_ovflpage(last_bfp);
210 	    last_bfp = rbufp;
211 	    if ( !rbufp ) return(-1);			/* Error */
212 	    bp = (u_short *)rbufp->page;
213 	}
214 
215 	/*
216 	    If we get here then rbufp points to the last page of
217 	    the big key/data pair.  Bufp points to the first
218 	    one -- it should now be empty pointing to the next
219 	    page after this pair.  Can't free it because we don't
220 	    have the page pointing to it.
221 	*/
222 
223 	/* This is information from the last page of the pair */
224 	n = bp[0];
225 	pageno = bp[n-1];
226 
227 	/* Now, bp is the first page of the pair */
228 	bp = (u_short *)bufp->page;
229 	if ( n > 2 ) {
230 	    /* There is an overflow page */
231 	    bp[1] = pageno;
232 	    bp[2] = OVFLPAGE;
233 	    bufp->ovfl = rbufp->ovfl;
234 	} else {
235 	    /* This is the last page */
236 	    bufp->ovfl = NULL;
237 	}
238 	n -= 2;
239 	bp[0] = n;
240 	FREESPACE(bp) = hashp->BSIZE - PAGE_META(n);
241 	OFFSET(bp) = hashp->BSIZE - 1;
242 
243 	bufp->flags |= BUF_MOD;
244 	if ( rbufp ) __free_ovflpage(rbufp);
245 	if ( last_bfp != rbufp ) __free_ovflpage(last_bfp);
246 
247 	hashp->NKEYS--;
248 	return(0);
249 }
250 
251 /*
252     0 = key not found
253     -1 = get next overflow page
254     -2 means key not found and this is big key/data
255     -3 error
256 */
257 extern int
258 __find_bigpair(bufp, ndx, key, size )
259 BUFHEAD	*bufp;
260 int	ndx;
261 char	*key;
262 int	size;
263 {
264     register	u_short	*bp = (u_short *)bufp->page;
265     register	char	*p = bufp->page;
266     int		ksize = size;
267     char	*kkey = key;
268     u_short	bytes;
269 
270 
271     for ( bytes = hashp->BSIZE - bp[ndx];
272 	  bytes <= size && bp[ndx+1] == PARTIAL_KEY;
273 	  bytes = hashp->BSIZE - bp[ndx] ) {
274 
275 	if ( bcmp ( p+bp[ndx], kkey, bytes ))return(-2);
276 	kkey += bytes;
277 	ksize -= bytes;
278 	bufp = __get_buf ( bp[ndx+2], bufp, 0 );
279 	if ( !bufp ) {
280 	    return(-3);
281 	}
282 	p = bufp->page;
283 	bp = (u_short *)p;
284 	ndx = 1;
285     }
286 
287     if ( (bytes != ksize) || bcmp ( p+bp[ndx], kkey, bytes )) {
288 #ifdef HASH_STATISTICS
289 	hash_collisions++;
290 #endif
291 	return(-2);
292     }
293     else return (ndx);
294 }
295 
296 
297 /*
298     Given the buffer pointer of the first overflow page of a big pair,
299     find the end of the big pair
300 
301     This will set bpp to the buffer header of the last page of the big pair.
302     It will return the pageno of the overflow page following the last page of
303     the pair; 0 if there isn't any (i.e. big pair is the last key in the
304     bucket)
305 */
306 extern u_short
307 __find_last_page ( bpp )
308 BUFHEAD	**bpp;
309 {
310 	int	n;
311 	u_short	pageno;
312 	BUFHEAD	*bufp = *bpp;
313 	u_short	*bp = (u_short *)bufp->page;
314 
315 	while ( 1 ) {
316 	    n = bp[0];
317 
318 	    /*
319 		This is the last page if:
320 			the tag is FULL_KEY_DATA and either
321 				only 2 entries
322 				OVFLPAGE marker is explicit
323 				there is freespace on the page
324 	    */
325 	    if ( bp[2] == FULL_KEY_DATA &&
326 		 ((n == 2) ||  (bp[n] == OVFLPAGE) || (FREESPACE(bp)) ) ) break;
327 
328 	    pageno = bp[n-1];
329 	    bufp = __get_buf ( pageno, bufp, 0 );
330 	    if ( !bufp ) return (0);		/* Need to indicate an error! */
331 	    bp = (u_short *)bufp->page;
332 	}
333 
334 	*bpp = bufp;
335 	if ( bp[0] > 2 ) return ( bp[3] );
336 	else return(0);
337 }
338 
339 
340 /*
341     Return the data for the key/data pair
342     that begins on this page at this index
343     (index should always be 1)
344 */
345 extern int
346 __big_return ( bufp, ndx, val, set_current )
347 BUFHEAD	*bufp;
348 int	ndx;
349 DBT	*val;
350 int	set_current;
351 {
352     BUFHEAD	*save_p;
353     u_short	save_addr;
354     u_short	*bp = (u_short *)bufp->page;
355     u_short	off, len;
356     char	*cp, *tp;
357 
358     while ( bp[ndx+1] == PARTIAL_KEY ) {
359 	bufp = __get_buf ( bp[bp[0]-1], bufp, 0 );
360 	if ( !bufp ) return(-1);
361 	bp = (u_short *)bufp->page;
362 	ndx = 1;
363     }
364 
365     if ( bp[ndx+1] == FULL_KEY ) {
366 	bufp = __get_buf ( bp[bp[0]-1], bufp, 0 );
367 	if ( !bufp ) return(-1);
368 	bp = (u_short *)bufp->page;
369 	save_p = bufp;
370 	save_addr = save_p->addr;
371 	off = bp[1];
372 	len = 0;
373     } else if (!FREESPACE(bp)) {
374 	/*
375 	    This is a hack.  We can't distinguish between
376 	    FULL_KEY_DATA that contains complete data or
377 	    incomplete data, so we require that if the
378 	    data  is complete, there is at least 1 byte
379 	    of free space left.
380 	*/
381 	off = bp[bp[0]];
382 	len = bp[1] - off;
383 	save_p = bufp;
384 	save_addr = bufp->addr;
385 	bufp = __get_buf ( bp[bp[0]-1], bufp, 0 );
386 	if ( !bufp ) return(-1);
387 	bp = (u_short *)bufp->page;
388     } else {
389 	/* The data is all on one page */
390 	tp = (char *)bp;
391 	off = bp[bp[0]];
392 	val->data = (u_char *)tp + off;
393 	val->size = bp[1] - off;
394 	if ( set_current ) {
395 	    if ( bp[0] == 2 ) {		/* No more buckets in chain */
396 		hashp->cpage = NULL;
397 		hashp->cbucket++;
398 		hashp->cndx=1;
399 	    } else  {
400 		hashp->cpage = __get_buf ( bp[bp[0]-1], bufp, 0 );
401 		if ( !hashp->cpage )return(-1);
402 		hashp->cndx = 1;
403 		if ( !((u_short *)hashp->cpage->page)[0] ) {
404 		    hashp->cbucket++;
405 		    hashp->cpage = NULL;
406 		}
407 	    }
408 	}
409 	return(0);
410     }
411 
412     val->size = collect_data ( bufp, len, set_current );
413     if ( val->size == -1 ) {
414 	return(-1);
415     }
416     if ( save_p->addr != save_addr ) {
417 	/* We are pretty short on buffers */
418 	errno = EINVAL;		/* OUT OF BUFFERS */
419 	return(-1);
420     }
421     bcopy ( (save_p->page)+off, hashp->tmp_buf, len );
422     val->data = (u_char *)hashp->tmp_buf;
423     return(0);
424 }
425 
426 /*
427     Count how big the total datasize is by
428     recursing through the pages.  Then allocate
429     a buffer and copy the data as you recurse up.
430 */
431 static int
432 collect_data ( bufp, len, set )
433 BUFHEAD	*bufp;
434 int	len;
435 int	set;
436 {
437     register	char	*p = bufp->page;
438     register	u_short	*bp = (u_short *)p;
439     u_short	save_addr;
440     int	mylen, totlen;
441     BUFHEAD	*xbp;
442 
443     mylen = hashp->BSIZE - bp[1];
444     save_addr = bufp->addr;
445 
446     if ( bp[2] == FULL_KEY_DATA ) {	/* End of Data */
447 	totlen = len + mylen;
448 	if ( hashp->tmp_buf ) free (hashp->tmp_buf);
449 	hashp->tmp_buf = (char *)malloc ( totlen );
450 	if ( !hashp->tmp_buf ) {
451 	    return(-1);
452 	}
453 	if ( set ) {
454 	    hashp->cndx = 1;
455 	    if ( bp[0] == 2 ) {		/* No more buckets in chain */
456 		hashp->cpage = NULL;
457 		hashp->cbucket++;
458 	    } else  {
459 		hashp->cpage = __get_buf ( bp[bp[0]-1], bufp, 0 );
460 		if (!hashp->cpage) {
461 		    return(-1);
462 		} else if ( !((u_short *)hashp->cpage->page)[0] ) {
463 		    hashp->cbucket++;
464 		    hashp->cpage = NULL;
465 		}
466 	    }
467 	}
468     } else {
469 	xbp = __get_buf ( bp[bp[0]-1], bufp, 0 );
470 	if ( !xbp || ((totlen = collect_data ( xbp, len + mylen, set )) < 1) ) {
471 	    return(-1);
472 	}
473     }
474     if ( bufp->addr != save_addr ) {
475 	errno = EINVAL;		/* Out of buffers */
476 	return(-1);
477     }
478     bcopy ( (bufp->page) + bp[1], &hashp->tmp_buf[len], mylen );
479     return ( totlen );
480 }
481 
482 /*
483 	Fill in the key and data
484 	for this big pair
485 */
486 extern int
487 __big_keydata ( bufp, ndx, key, val, set )
488 BUFHEAD	*bufp;
489 int	ndx;
490 DBT	*key, *val;
491 int	set;
492 {
493     key->size = collect_key ( bufp, 0, val, set );
494     if ( key->size == -1 ) {
495 	return (-1);
496     }
497     key->data = (u_char *)hashp->tmp_key;
498     return(0);
499 }
500 
501 /*
502     Count how big the total key size is by
503     recursing through the pages.  Then collect
504     the data, allocate a buffer and copy the key as
505     you recurse up.
506 */
507 static int
508 collect_key ( bufp, len, val, set )
509 BUFHEAD	*bufp;
510 int	len;
511 DBT	*val;
512 int	set;
513 {
514     char	*p = bufp->page;
515     u_short	*bp = (u_short *)p;
516     u_short	save_addr;
517     int	mylen, totlen;
518     BUFHEAD	*xbp;
519 
520     mylen = hashp->BSIZE - bp[1];
521 
522     save_addr = bufp->addr;
523     totlen = len + mylen;
524     if ( bp[2] == FULL_KEY || bp[2] == FULL_KEY_DATA ) {/* End of Key */
525 	if ( hashp->tmp_key ) free (hashp->tmp_key);
526 	hashp->tmp_key = (char *)malloc ( totlen );
527 	if ( !hashp->tmp_key ) {
528 	    return(-1);
529 	}
530 	__big_return ( bufp, 1, val, set );
531     } else {
532 	xbp = __get_buf (bp[bp[0]-1], bufp, 0);
533 	if ( !xbp || ((totlen = collect_key (xbp, totlen, val, set)) < 1 ) ) {
534 	    return(-1);
535 	}
536     }
537     if ( bufp->addr != save_addr ) {
538 	errno = EINVAL;		/* MIS -- OUT OF BUFFERS */
539 	return (-1);
540     }
541     bcopy ( (bufp->page) + bp[1], &hashp->tmp_key[len], mylen );
542     return ( totlen );
543 }
544 
545 
546 /*
547     return 0 => OK
548 	   -1 => error
549 */
550 extern int
551 __big_split ( op, np, big_keyp, addr, obucket, ret )
552 BUFHEAD	*op;		/* Pointer to where to put keys that go in old bucket */
553 BUFHEAD	*np;		/* Pointer to new bucket page */
554 BUFHEAD	*big_keyp;	/* Pointer to first page containing the big key/data */
555 u_short	addr;		/* Address of big_keyp */
556 u_int	obucket;	/* Old Bucket */
557 SPLIT_RETURN	*ret;
558 {
559     register	u_short	*prev_pagep;
560     register	BUFHEAD	*tmpp;
561     register	u_short 	*tp;
562     BUFHEAD	*bp = big_keyp;
563     u_short	off, free_space;
564     u_short	n;
565 
566     DBT		key, val;
567 
568     u_int	change;
569 
570     /* Now figure out where the big key/data goes */
571     if (__big_keydata ( big_keyp, 1, &key, &val, 0 )) {
572 	return(-1);
573     }
574     change = (__call_hash ( key.data, key.size ) != obucket );
575 
576     if ( ret->next_addr = __find_last_page ( &big_keyp ) ) {
577 	if (!(ret->nextp = __get_buf ( ret->next_addr, big_keyp, 0 ))) {
578 	    return(-1);;
579 	}
580     } else {
581 	ret->nextp = NULL;
582     }
583 
584     /* Now make one of np/op point to the big key/data pair */
585     assert(np->ovfl == NULL);
586     if ( change ) tmpp = np;
587     else tmpp = op;
588 
589     tmpp->flags |= BUF_MOD;
590 #ifdef DEBUG1
591     fprintf ( stderr, "BIG_SPLIT: %d->ovfl was %d is now %d\n", tmpp->addr,
592 		(tmpp->ovfl?tmpp->ovfl->addr:0),
593 		(bp?bp->addr:0) );
594 #endif
595     tmpp->ovfl = bp;		/* one of op/np point to big_keyp */
596     tp = (u_short *)tmpp->page;
597     assert ( FREESPACE(tp) >= OVFLSIZE);
598     n = tp[0];
599     off = OFFSET(tp);
600     free_space = FREESPACE(tp);
601     tp[++n] = addr;
602     tp[++n] = OVFLPAGE;
603     tp[0] = n;
604     OFFSET(tp) = off;
605     FREESPACE(tp) = free_space - OVFLSIZE;
606 
607     /*
608 	Finally, set the new and old return values.
609 	BIG_KEYP contains a pointer to the last page of the big key_data pair.
610 	Make sure that big_keyp has no following page (2 elements) or create
611 	an empty following page.
612     */
613 
614     ret->newp = np;
615     ret->oldp = op;
616 
617     tp = (u_short *)big_keyp->page;
618     big_keyp->flags |= BUF_MOD;
619     if ( tp[0] > 2 ) {
620 	/*
621 	    There may be either one or two offsets on this page
622 	    If there is one, then the overflow page is linked on
623 	    normally and tp[4] is OVFLPAGE.  If there are two, tp[4]
624 	    contains the second offset and needs to get stuffed in
625 	    after the next overflow page is added
626 	*/
627 	n = tp[4];
628 	free_space = FREESPACE(tp);
629 	off = OFFSET(tp);
630 	tp[0] -= 2;
631 	FREESPACE(tp) = free_space + OVFLSIZE;
632 	OFFSET(tp) = off;
633 	tmpp = __add_ovflpage ( big_keyp );
634 	if ( !tmpp ) {
635 	    return(-1);
636 	}
637 	tp[4] = n;
638     } else {
639 	tmpp = big_keyp;
640     }
641 
642     if ( change ) ret->newp = tmpp;
643     else ret->oldp = tmpp;
644 
645     return(0);
646 }
647