Home | History | Annotate | Line # | Download | only in hash
hash_page.c revision 1.6
      1 /*-
      2  * Copyright (c) 1990, 1993, 1994
      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[] = "@(#)hash_page.c	8.6 (Berkeley) 6/16/94";
     39 #endif /* LIBC_SCCS and not lint */
     40 
     41 /*
     42  * PACKAGE:  hashing
     43  *
     44  * DESCRIPTION:
     45  *	Page manipulation for hashing package.
     46  *
     47  * ROUTINES:
     48  *
     49  * External
     50  *	__get_page
     51  *	__add_ovflpage
     52  * Internal
     53  *	overflow_page
     54  *	open_temp
     55  */
     56 
     57 #include <sys/types.h>
     58 
     59 #include <errno.h>
     60 #include <fcntl.h>
     61 #include <signal.h>
     62 #include <stdio.h>
     63 #include <stdlib.h>
     64 #include <string.h>
     65 #include <unistd.h>
     66 #ifdef DEBUG
     67 #include <assert.h>
     68 #endif
     69 
     70 #include <db.h>
     71 #include "hash.h"
     72 #include "page.h"
     73 #include "extern.h"
     74 
     75 static u_int32_t	*fetch_bitmap __P((HTAB *, int));
     76 static u_int32_t	 first_free __P((u_int32_t));
     77 static int	 open_temp __P((HTAB *));
     78 static u_int16_t	 overflow_page __P((HTAB *));
     79 static void	 putpair __P((char *, const DBT *, const DBT *));
     80 static void	 squeeze_key __P((u_int16_t *, const DBT *, const DBT *));
     81 static int	 ugly_split
     82 		    __P((HTAB *, u_int32_t, BUFHEAD *, BUFHEAD *, int, int));
     83 
     84 #define	PAGE_INIT(P) { \
     85 	((u_int16_t *)(P))[0] = 0; \
     86 	((u_int16_t *)(P))[1] = hashp->BSIZE - 3 * sizeof(u_int16_t); \
     87 	((u_int16_t *)(P))[2] = hashp->BSIZE; \
     88 }
     89 
     90 /*
     91  * This is called AFTER we have verified that there is room on the page for
     92  * the pair (PAIRFITS has returned true) so we go right ahead and start moving
     93  * stuff on.
     94  */
     95 static void
     96 putpair(p, key, val)
     97 	char *p;
     98 	const DBT *key, *val;
     99 {
    100 	register u_int16_t *bp, n, off;
    101 
    102 	bp = (u_int16_t *)p;
    103 
    104 	/* Enter the key first. */
    105 	n = bp[0];
    106 
    107 	off = OFFSET(bp) - key->size;
    108 	memmove(p + off, key->data, key->size);
    109 	bp[++n] = off;
    110 
    111 	/* Now the data. */
    112 	off -= val->size;
    113 	memmove(p + off, val->data, val->size);
    114 	bp[++n] = off;
    115 
    116 	/* Adjust page info. */
    117 	bp[0] = n;
    118 	bp[n + 1] = off - ((n + 3) * sizeof(u_int16_t));
    119 	bp[n + 2] = off;
    120 }
    121 
    122 /*
    123  * Returns:
    124  *	 0 OK
    125  *	-1 error
    126  */
    127 extern int
    128 __delpair(hashp, bufp, ndx)
    129 	HTAB *hashp;
    130 	BUFHEAD *bufp;
    131 	register int ndx;
    132 {
    133 	register u_int16_t *bp, newoff;
    134 	register int n;
    135 	u_int16_t pairlen;
    136 
    137 	bp = (u_int16_t *)bufp->page;
    138 	n = bp[0];
    139 
    140 	if (bp[ndx + 1] < REAL_KEY)
    141 		return (__big_delete(hashp, bufp));
    142 	if (ndx != 1)
    143 		newoff = bp[ndx - 1];
    144 	else
    145 		newoff = hashp->BSIZE;
    146 	pairlen = newoff - bp[ndx + 1];
    147 
    148 	if (ndx != (n - 1)) {
    149 		/* Hard Case -- need to shuffle keys */
    150 		register int i;
    151 		register char *src = bufp->page + (int)OFFSET(bp);
    152 		register char *dst = src + (int)pairlen;
    153 		memmove(dst, src, bp[ndx + 1] - OFFSET(bp));
    154 
    155 		/* Now adjust the pointers */
    156 		for (i = ndx + 2; i <= n; i += 2) {
    157 			if (bp[i + 1] == OVFLPAGE) {
    158 				bp[i - 2] = bp[i];
    159 				bp[i - 1] = bp[i + 1];
    160 			} else {
    161 				bp[i - 2] = bp[i] + pairlen;
    162 				bp[i - 1] = bp[i + 1] + pairlen;
    163 			}
    164 		}
    165 	}
    166 	/* Finally adjust the page data */
    167 	bp[n] = OFFSET(bp) + pairlen;
    168 	bp[n - 1] = bp[n + 1] + pairlen + 2 * sizeof(u_int16_t);
    169 	bp[0] = n - 2;
    170 	hashp->NKEYS--;
    171 
    172 	bufp->flags |= BUF_MOD;
    173 	return (0);
    174 }
    175 /*
    176  * Returns:
    177  *	 0 ==> OK
    178  *	-1 ==> Error
    179  */
    180 extern int
    181 __split_page(hashp, obucket, nbucket)
    182 	HTAB *hashp;
    183 	u_int32_t obucket, nbucket;
    184 {
    185 	register BUFHEAD *new_bufp, *old_bufp;
    186 	register u_int16_t *ino;
    187 	register char *np;
    188 	DBT key, val;
    189 	int n, ndx, retval;
    190 	u_int16_t copyto, diff, off, moved;
    191 	char *op;
    192 
    193 	copyto = (u_int16_t)hashp->BSIZE;
    194 	off = (u_int16_t)hashp->BSIZE;
    195 	old_bufp = __get_buf(hashp, obucket, NULL, 0);
    196 	if (old_bufp == NULL)
    197 		return (-1);
    198 	new_bufp = __get_buf(hashp, nbucket, NULL, 0);
    199 	if (new_bufp == NULL)
    200 		return (-1);
    201 
    202 	old_bufp->flags |= (BUF_MOD | BUF_PIN);
    203 	new_bufp->flags |= (BUF_MOD | BUF_PIN);
    204 
    205 	ino = (u_int16_t *)(op = old_bufp->page);
    206 	np = new_bufp->page;
    207 
    208 	moved = 0;
    209 
    210 	for (n = 1, ndx = 1; n < ino[0]; n += 2) {
    211 		if (ino[n + 1] < REAL_KEY) {
    212 			retval = ugly_split(hashp, obucket, old_bufp, new_bufp,
    213 			    (int)copyto, (int)moved);
    214 			old_bufp->flags &= ~BUF_PIN;
    215 			new_bufp->flags &= ~BUF_PIN;
    216 			return (retval);
    217 
    218 		}
    219 		key.data = (u_char *)op + ino[n];
    220 		key.size = off - ino[n];
    221 
    222 		if (__call_hash(hashp, key.data, key.size) == obucket) {
    223 			/* Don't switch page */
    224 			diff = copyto - off;
    225 			if (diff) {
    226 				copyto = ino[n + 1] + diff;
    227 				memmove(op + copyto, op + ino[n + 1],
    228 				    off - ino[n + 1]);
    229 				ino[ndx] = copyto + ino[n] - ino[n + 1];
    230 				ino[ndx + 1] = copyto;
    231 			} else
    232 				copyto = ino[n + 1];
    233 			ndx += 2;
    234 		} else {
    235 			/* Switch page */
    236 			val.data = (u_char *)op + ino[n + 1];
    237 			val.size = ino[n] - ino[n + 1];
    238 			putpair(np, &key, &val);
    239 			moved += 2;
    240 		}
    241 
    242 		off = ino[n + 1];
    243 	}
    244 
    245 	/* Now clean up the page */
    246 	ino[0] -= moved;
    247 	FREESPACE(ino) = copyto - sizeof(u_int16_t) * (ino[0] + 3);
    248 	OFFSET(ino) = copyto;
    249 
    250 #ifdef DEBUG3
    251 	(void)fprintf(stderr, "split %d/%d\n",
    252 	    ((u_int16_t *)np)[0] / 2,
    253 	    ((u_int16_t *)op)[0] / 2);
    254 #endif
    255 	/* unpin both pages */
    256 	old_bufp->flags &= ~BUF_PIN;
    257 	new_bufp->flags &= ~BUF_PIN;
    258 	return (0);
    259 }
    260 
    261 /*
    262  * Called when we encounter an overflow or big key/data page during split
    263  * handling.  This is special cased since we have to begin checking whether
    264  * the key/data pairs fit on their respective pages and because we may need
    265  * overflow pages for both the old and new pages.
    266  *
    267  * The first page might be a page with regular key/data pairs in which case
    268  * we have a regular overflow condition and just need to go on to the next
    269  * page or it might be a big key/data pair in which case we need to fix the
    270  * big key/data pair.
    271  *
    272  * Returns:
    273  *	 0 ==> success
    274  *	-1 ==> failure
    275  */
    276 static int
    277 ugly_split(hashp, obucket, old_bufp, new_bufp, copyto, moved)
    278 	HTAB *hashp;
    279 	u_int32_t obucket;	/* Same as __split_page. */
    280 	BUFHEAD *old_bufp, *new_bufp;
    281 	int copyto;	/* First byte on page which contains key/data values. */
    282 	int moved;	/* Number of pairs moved to new page. */
    283 {
    284 	register BUFHEAD *bufp;	/* Buffer header for ino */
    285 	register u_int16_t *ino;	/* Page keys come off of */
    286 	register u_int16_t *np;	/* New page */
    287 	register u_int16_t *op;	/* Page keys go on to if they aren't moving */
    288 
    289 	BUFHEAD *last_bfp;	/* Last buf header OVFL needing to be freed */
    290 	DBT key, val;
    291 	SPLIT_RETURN ret;
    292 	u_int16_t n, off, ov_addr, scopyto;
    293 	char *cino;		/* Character value of ino */
    294 
    295 	bufp = old_bufp;
    296 	ino = (u_int16_t *)old_bufp->page;
    297 	np = (u_int16_t *)new_bufp->page;
    298 	op = (u_int16_t *)old_bufp->page;
    299 	last_bfp = NULL;
    300 	scopyto = (u_int16_t)copyto;	/* ANSI */
    301 
    302 	n = ino[0] - 1;
    303 	while (n < ino[0]) {
    304 		if (ino[2] < REAL_KEY && ino[2] != OVFLPAGE) {
    305 			if (__big_split(hashp, old_bufp,
    306 			    new_bufp, bufp, bufp->addr, obucket, &ret))
    307 				return (-1);
    308 			old_bufp = ret.oldp;
    309 			if (!old_bufp)
    310 				return (-1);
    311 			op = (u_int16_t *)old_bufp->page;
    312 			new_bufp = ret.newp;
    313 			if (!new_bufp)
    314 				return (-1);
    315 			np = (u_int16_t *)new_bufp->page;
    316 			bufp = ret.nextp;
    317 			if (!bufp)
    318 				return (0);
    319 			cino = (char *)bufp->page;
    320 			ino = (u_int16_t *)cino;
    321 			last_bfp = ret.nextp;
    322 		} else if (ino[n + 1] == OVFLPAGE) {
    323 			ov_addr = ino[n];
    324 			/*
    325 			 * Fix up the old page -- the extra 2 are the fields
    326 			 * which contained the overflow information.
    327 			 */
    328 			ino[0] -= (moved + 2);
    329 			FREESPACE(ino) =
    330 			    scopyto - sizeof(u_int16_t) * (ino[0] + 3);
    331 			OFFSET(ino) = scopyto;
    332 
    333 			bufp = __get_buf(hashp, ov_addr, bufp, 0);
    334 			if (!bufp)
    335 				return (-1);
    336 
    337 			ino = (u_int16_t *)bufp->page;
    338 			n = 1;
    339 			scopyto = hashp->BSIZE;
    340 			moved = 0;
    341 
    342 			if (last_bfp)
    343 				__free_ovflpage(hashp, last_bfp);
    344 			last_bfp = bufp;
    345 		}
    346 		/* Move regular sized pairs of there are any */
    347 		off = hashp->BSIZE;
    348 		for (n = 1; (n < ino[0]) && (ino[n + 1] >= REAL_KEY); n += 2) {
    349 			cino = (char *)ino;
    350 			key.data = (u_char *)cino + ino[n];
    351 			key.size = off - ino[n];
    352 			val.data = (u_char *)cino + ino[n + 1];
    353 			val.size = ino[n] - ino[n + 1];
    354 			off = ino[n + 1];
    355 
    356 			if (__call_hash(hashp, key.data, key.size) == obucket) {
    357 				/* Keep on old page */
    358 				if (PAIRFITS(op, (&key), (&val)))
    359 					putpair((char *)op, &key, &val);
    360 				else {
    361 					old_bufp =
    362 					    __add_ovflpage(hashp, old_bufp);
    363 					if (!old_bufp)
    364 						return (-1);
    365 					op = (u_int16_t *)old_bufp->page;
    366 					putpair((char *)op, &key, &val);
    367 				}
    368 				old_bufp->flags |= BUF_MOD;
    369 			} else {
    370 				/* Move to new page */
    371 				if (PAIRFITS(np, (&key), (&val)))
    372 					putpair((char *)np, &key, &val);
    373 				else {
    374 					new_bufp =
    375 					    __add_ovflpage(hashp, new_bufp);
    376 					if (!new_bufp)
    377 						return (-1);
    378 					np = (u_int16_t *)new_bufp->page;
    379 					putpair((char *)np, &key, &val);
    380 				}
    381 				new_bufp->flags |= BUF_MOD;
    382 			}
    383 		}
    384 	}
    385 	if (last_bfp)
    386 		__free_ovflpage(hashp, last_bfp);
    387 	return (0);
    388 }
    389 
    390 /*
    391  * Add the given pair to the page
    392  *
    393  * Returns:
    394  *	0 ==> OK
    395  *	1 ==> failure
    396  */
    397 extern int
    398 __addel(hashp, bufp, key, val)
    399 	HTAB *hashp;
    400 	BUFHEAD *bufp;
    401 	const DBT *key, *val;
    402 {
    403 	register u_int16_t *bp, *sop;
    404 	int do_expand;
    405 
    406 	bp = (u_int16_t *)bufp->page;
    407 	do_expand = 0;
    408 	while (bp[0] && (bp[2] < REAL_KEY || bp[bp[0]] < REAL_KEY))
    409 		/* Exception case */
    410 		if (bp[2] == FULL_KEY_DATA && bp[0] == 2)
    411 			/* This is the last page of a big key/data pair
    412 			   and we need to add another page */
    413 			break;
    414 		else if (bp[2] < REAL_KEY && bp[bp[0]] != OVFLPAGE) {
    415 			bufp = __get_buf(hashp, bp[bp[0] - 1], bufp, 0);
    416 			if (!bufp)
    417 				return (-1);
    418 			bp = (u_int16_t *)bufp->page;
    419 		} else
    420 			/* Try to squeeze key on this page */
    421 			if (FREESPACE(bp) > PAIRSIZE(key, val)) {
    422 				squeeze_key(bp, key, val);
    423 				return (0);
    424 			} else {
    425 				bufp = __get_buf(hashp, bp[bp[0] - 1], bufp, 0);
    426 				if (!bufp)
    427 					return (-1);
    428 				bp = (u_int16_t *)bufp->page;
    429 			}
    430 
    431 	if (PAIRFITS(bp, key, val))
    432 		putpair(bufp->page, key, val);
    433 	else {
    434 		do_expand = 1;
    435 		bufp = __add_ovflpage(hashp, bufp);
    436 		if (!bufp)
    437 			return (-1);
    438 		sop = (u_int16_t *)bufp->page;
    439 
    440 		if (PAIRFITS(sop, key, val))
    441 			putpair((char *)sop, key, val);
    442 		else
    443 			if (__big_insert(hashp, bufp, key, val))
    444 				return (-1);
    445 	}
    446 	bufp->flags |= BUF_MOD;
    447 	/*
    448 	 * If the average number of keys per bucket exceeds the fill factor,
    449 	 * expand the table.
    450 	 */
    451 	hashp->NKEYS++;
    452 	if (do_expand ||
    453 	    (hashp->NKEYS / (hashp->MAX_BUCKET + 1) > hashp->FFACTOR))
    454 		return (__expand_table(hashp));
    455 	return (0);
    456 }
    457 
    458 /*
    459  *
    460  * Returns:
    461  *	pointer on success
    462  *	NULL on error
    463  */
    464 extern BUFHEAD *
    465 __add_ovflpage(hashp, bufp)
    466 	HTAB *hashp;
    467 	BUFHEAD *bufp;
    468 {
    469 	register u_int16_t *sp;
    470 	u_int16_t ndx, ovfl_num;
    471 #ifdef DEBUG1
    472 	int tmp1, tmp2;
    473 #endif
    474 	sp = (u_int16_t *)bufp->page;
    475 
    476 	/* Check if we are dynamically determining the fill factor */
    477 	if (hashp->FFACTOR == DEF_FFACTOR) {
    478 		hashp->FFACTOR = sp[0] >> 1;
    479 		if (hashp->FFACTOR < MIN_FFACTOR)
    480 			hashp->FFACTOR = MIN_FFACTOR;
    481 	}
    482 	bufp->flags |= BUF_MOD;
    483 	ovfl_num = overflow_page(hashp);
    484 #ifdef DEBUG1
    485 	tmp1 = bufp->addr;
    486 	tmp2 = bufp->ovfl ? bufp->ovfl->addr : 0;
    487 #endif
    488 	if (!ovfl_num || !(bufp->ovfl = __get_buf(hashp, ovfl_num, bufp, 1)))
    489 		return (NULL);
    490 	bufp->ovfl->flags |= BUF_MOD;
    491 #ifdef DEBUG1
    492 	(void)fprintf(stderr, "ADDOVFLPAGE: %d->ovfl was %d is now %d\n",
    493 	    tmp1, tmp2, bufp->ovfl->addr);
    494 #endif
    495 	ndx = sp[0];
    496 	/*
    497 	 * Since a pair is allocated on a page only if there's room to add
    498 	 * an overflow page, we know that the OVFL information will fit on
    499 	 * the page.
    500 	 */
    501 	sp[ndx + 4] = OFFSET(sp);
    502 	sp[ndx + 3] = FREESPACE(sp) - OVFLSIZE;
    503 	sp[ndx + 1] = ovfl_num;
    504 	sp[ndx + 2] = OVFLPAGE;
    505 	sp[0] = ndx + 2;
    506 #ifdef HASH_STATISTICS
    507 	hash_overflows++;
    508 #endif
    509 	return (bufp->ovfl);
    510 }
    511 
    512 /*
    513  * Returns:
    514  *	 0 indicates SUCCESS
    515  *	-1 indicates FAILURE
    516  */
    517 extern int
    518 __get_page(hashp, p, bucket, is_bucket, is_disk, is_bitmap)
    519 	HTAB *hashp;
    520 	char *p;
    521 	u_int32_t bucket;
    522 	int is_bucket, is_disk, is_bitmap;
    523 {
    524 	register int fd, page, size;
    525 	int rsize;
    526 	u_int16_t *bp;
    527 
    528 	fd = hashp->fp;
    529 	size = hashp->BSIZE;
    530 
    531 	if ((fd == -1) || !is_disk) {
    532 		PAGE_INIT(p);
    533 		return (0);
    534 	}
    535 	if (is_bucket)
    536 		page = BUCKET_TO_PAGE(bucket);
    537 	else
    538 		page = OADDR_TO_PAGE(bucket);
    539 	if ((lseek(fd, (off_t)page << hashp->BSHIFT, SEEK_SET) == -1) ||
    540 	    ((rsize = read(fd, p, size)) == -1))
    541 		return (-1);
    542 	bp = (u_int16_t *)p;
    543 	if (!rsize)
    544 		bp[0] = 0;	/* We hit the EOF, so initialize a new page */
    545 	else
    546 		if (rsize != size) {
    547 			errno = EFTYPE;
    548 			return (-1);
    549 		}
    550 	if (!is_bitmap && !bp[0]) {
    551 		PAGE_INIT(p);
    552 	} else
    553 		if (hashp->LORDER != BYTE_ORDER) {
    554 			register int i, max;
    555 
    556 			if (is_bitmap) {
    557 				max = hashp->BSIZE >> 2; /* divide by 4 */
    558 				for (i = 0; i < max; i++)
    559 					M_32_SWAP(((int *)p)[i]);
    560 			} else {
    561 				M_16_SWAP(bp[0]);
    562 				max = bp[0] + 2;
    563 				for (i = 1; i <= max; i++)
    564 					M_16_SWAP(bp[i]);
    565 			}
    566 		}
    567 	return (0);
    568 }
    569 
    570 /*
    571  * Write page p to disk
    572  *
    573  * Returns:
    574  *	 0 ==> OK
    575  *	-1 ==>failure
    576  */
    577 extern int
    578 __put_page(hashp, p, bucket, is_bucket, is_bitmap)
    579 	HTAB *hashp;
    580 	char *p;
    581 	u_int32_t bucket;
    582 	int is_bucket, is_bitmap;
    583 {
    584 	register int fd, page, size;
    585 	int wsize;
    586 
    587 	size = hashp->BSIZE;
    588 	if ((hashp->fp == -1) && open_temp(hashp))
    589 		return (-1);
    590 	fd = hashp->fp;
    591 
    592 	if (hashp->LORDER != BYTE_ORDER) {
    593 		register int i;
    594 		register int max;
    595 
    596 		if (is_bitmap) {
    597 			max = hashp->BSIZE >> 2;	/* divide by 4 */
    598 			for (i = 0; i < max; i++)
    599 				M_32_SWAP(((int *)p)[i]);
    600 		} else {
    601 			max = ((u_int16_t *)p)[0] + 2;
    602 			for (i = 0; i <= max; i++)
    603 				M_16_SWAP(((u_int16_t *)p)[i]);
    604 		}
    605 	}
    606 	if (is_bucket)
    607 		page = BUCKET_TO_PAGE(bucket);
    608 	else
    609 		page = OADDR_TO_PAGE(bucket);
    610 	if ((lseek(fd, (off_t)page << hashp->BSHIFT, SEEK_SET) == -1) ||
    611 	    ((wsize = write(fd, p, size)) == -1))
    612 		/* Errno is set */
    613 		return (-1);
    614 	if (wsize != size) {
    615 		errno = EFTYPE;
    616 		return (-1);
    617 	}
    618 	return (0);
    619 }
    620 
    621 #define BYTE_MASK	((1 << INT_BYTE_SHIFT) -1)
    622 /*
    623  * Initialize a new bitmap page.  Bitmap pages are left in memory
    624  * once they are read in.
    625  */
    626 extern int
    627 __ibitmap(hashp, pnum, nbits, ndx)
    628 	HTAB *hashp;
    629 	int pnum, nbits, ndx;
    630 {
    631 	u_int32_t *ip;
    632 	int clearbytes, clearints;
    633 
    634 	if ((ip = (u_int32_t *)malloc(hashp->BSIZE)) == NULL)
    635 		return (1);
    636 	hashp->nmaps++;
    637 	clearints = ((nbits - 1) >> INT_BYTE_SHIFT) + 1;
    638 	clearbytes = clearints << INT_TO_BYTE;
    639 	(void)memset((char *)ip, 0, clearbytes);
    640 	(void)memset(((char *)ip) + clearbytes, 0xFF,
    641 	    hashp->BSIZE - clearbytes);
    642 	ip[clearints - 1] = ALL_SET << (nbits & BYTE_MASK);
    643 	SETBIT(ip, 0);
    644 	hashp->BITMAPS[ndx] = (u_int16_t)pnum;
    645 	hashp->mapp[ndx] = ip;
    646 	return (0);
    647 }
    648 
    649 static u_int32_t
    650 first_free(map)
    651 	u_int32_t map;
    652 {
    653 	register u_int32_t i, mask;
    654 
    655 	mask = 0x1;
    656 	for (i = 0; i < BITS_PER_MAP; i++) {
    657 		if (!(mask & map))
    658 			return (i);
    659 		mask = mask << 1;
    660 	}
    661 	return (i);
    662 }
    663 
    664 static u_int16_t
    665 overflow_page(hashp)
    666 	HTAB *hashp;
    667 {
    668 	register u_int32_t *freep;
    669 	register int max_free, offset, splitnum;
    670 	u_int16_t addr;
    671 	int bit, first_page, free_bit, free_page, i, in_use_bits, j;
    672 #ifdef DEBUG2
    673 	int tmp1, tmp2;
    674 #endif
    675 	splitnum = hashp->OVFL_POINT;
    676 	max_free = hashp->SPARES[splitnum];
    677 
    678 	free_page = (max_free - 1) >> (hashp->BSHIFT + BYTE_SHIFT);
    679 	free_bit = (max_free - 1) & ((hashp->BSIZE << BYTE_SHIFT) - 1);
    680 
    681 	/* Look through all the free maps to find the first free block */
    682 	first_page = hashp->LAST_FREED >>(hashp->BSHIFT + BYTE_SHIFT);
    683 	for ( i = first_page; i <= free_page; i++ ) {
    684 		if (!(freep = (u_int32_t *)hashp->mapp[i]) &&
    685 		    !(freep = fetch_bitmap(hashp, i)))
    686 			return (NULL);
    687 		if (i == free_page)
    688 			in_use_bits = free_bit;
    689 		else
    690 			in_use_bits = (hashp->BSIZE << BYTE_SHIFT) - 1;
    691 
    692 		if (i == first_page) {
    693 			bit = hashp->LAST_FREED &
    694 			    ((hashp->BSIZE << BYTE_SHIFT) - 1);
    695 			j = bit / BITS_PER_MAP;
    696 			bit = bit & ~(BITS_PER_MAP - 1);
    697 		} else {
    698 			bit = 0;
    699 			j = 0;
    700 		}
    701 		for (; bit <= in_use_bits; j++, bit += BITS_PER_MAP)
    702 			if (freep[j] != ALL_SET)
    703 				goto found;
    704 	}
    705 
    706 	/* No Free Page Found */
    707 	hashp->LAST_FREED = hashp->SPARES[splitnum];
    708 	hashp->SPARES[splitnum]++;
    709 	offset = hashp->SPARES[splitnum] -
    710 	    (splitnum ? hashp->SPARES[splitnum - 1] : 0);
    711 
    712 #define	OVMSG	"HASH: Out of overflow pages.  Increase page size\n"
    713 	if (offset > SPLITMASK) {
    714 		if (++splitnum >= NCACHED) {
    715 			(void)write(STDERR_FILENO, OVMSG, sizeof(OVMSG) - 1);
    716 			return (NULL);
    717 		}
    718 		hashp->OVFL_POINT = splitnum;
    719 		hashp->SPARES[splitnum] = hashp->SPARES[splitnum-1];
    720 		hashp->SPARES[splitnum-1]--;
    721 		offset = 1;
    722 	}
    723 
    724 	/* Check if we need to allocate a new bitmap page */
    725 	if (free_bit == (hashp->BSIZE << BYTE_SHIFT) - 1) {
    726 		free_page++;
    727 		if (free_page >= NCACHED) {
    728 			(void)write(STDERR_FILENO, OVMSG, sizeof(OVMSG) - 1);
    729 			return (NULL);
    730 		}
    731 		/*
    732 		 * This is tricky.  The 1 indicates that you want the new page
    733 		 * allocated with 1 clear bit.  Actually, you are going to
    734 		 * allocate 2 pages from this map.  The first is going to be
    735 		 * the map page, the second is the overflow page we were
    736 		 * looking for.  The init_bitmap routine automatically, sets
    737 		 * the first bit of itself to indicate that the bitmap itself
    738 		 * is in use.  We would explicitly set the second bit, but
    739 		 * don't have to if we tell init_bitmap not to leave it clear
    740 		 * in the first place.
    741 		 */
    742 		if (__ibitmap(hashp, (int)OADDR_OF(splitnum, offset),
    743 		    1, free_page))
    744 			return (NULL);
    745 		hashp->SPARES[splitnum]++;
    746 #ifdef DEBUG2
    747 		free_bit = 2;
    748 #endif
    749 		offset++;
    750 		if (offset > SPLITMASK) {
    751 			if (++splitnum >= NCACHED) {
    752 				(void)write(STDERR_FILENO, OVMSG,
    753 				    sizeof(OVMSG) - 1);
    754 				return (NULL);
    755 			}
    756 			hashp->OVFL_POINT = splitnum;
    757 			hashp->SPARES[splitnum] = hashp->SPARES[splitnum-1];
    758 			hashp->SPARES[splitnum-1]--;
    759 			offset = 0;
    760 		}
    761 	} else {
    762 		/*
    763 		 * Free_bit addresses the last used bit.  Bump it to address
    764 		 * the first available bit.
    765 		 */
    766 		free_bit++;
    767 		SETBIT(freep, free_bit);
    768 	}
    769 
    770 	/* Calculate address of the new overflow page */
    771 	addr = OADDR_OF(splitnum, offset);
    772 #ifdef DEBUG2
    773 	(void)fprintf(stderr, "OVERFLOW_PAGE: ADDR: %d BIT: %d PAGE %d\n",
    774 	    addr, free_bit, free_page);
    775 #endif
    776 	return (addr);
    777 
    778 found:
    779 	bit = bit + first_free(freep[j]);
    780 	SETBIT(freep, bit);
    781 #ifdef DEBUG2
    782 	tmp1 = bit;
    783 	tmp2 = i;
    784 #endif
    785 	/*
    786 	 * Bits are addressed starting with 0, but overflow pages are addressed
    787 	 * beginning at 1. Bit is a bit addressnumber, so we need to increment
    788 	 * it to convert it to a page number.
    789 	 */
    790 	bit = 1 + bit + (i * (hashp->BSIZE << BYTE_SHIFT));
    791 	if (bit >= hashp->LAST_FREED)
    792 		hashp->LAST_FREED = bit - 1;
    793 
    794 	/* Calculate the split number for this page */
    795 	for (i = 0; (i < splitnum) && (bit > hashp->SPARES[i]); i++);
    796 	offset = (i ? bit - hashp->SPARES[i - 1] : bit);
    797 	if (offset >= SPLITMASK)
    798 		return (NULL);	/* Out of overflow pages */
    799 	addr = OADDR_OF(i, offset);
    800 #ifdef DEBUG2
    801 	(void)fprintf(stderr, "OVERFLOW_PAGE: ADDR: %d BIT: %d PAGE %d\n",
    802 	    addr, tmp1, tmp2);
    803 #endif
    804 
    805 	/* Allocate and return the overflow page */
    806 	return (addr);
    807 }
    808 
    809 /*
    810  * Mark this overflow page as free.
    811  */
    812 extern void
    813 __free_ovflpage(hashp, obufp)
    814 	HTAB *hashp;
    815 	BUFHEAD *obufp;
    816 {
    817 	register u_int16_t addr;
    818 	u_int32_t *freep;
    819 	int bit_address, free_page, free_bit;
    820 	u_int16_t ndx;
    821 
    822 	addr = obufp->addr;
    823 #ifdef DEBUG1
    824 	(void)fprintf(stderr, "Freeing %d\n", addr);
    825 #endif
    826 	ndx = (((u_int16_t)addr) >> SPLITSHIFT);
    827 	bit_address =
    828 	    (ndx ? hashp->SPARES[ndx - 1] : 0) + (addr & SPLITMASK) - 1;
    829 	 if (bit_address < hashp->LAST_FREED)
    830 		hashp->LAST_FREED = bit_address;
    831 	free_page = (bit_address >> (hashp->BSHIFT + BYTE_SHIFT));
    832 	free_bit = bit_address & ((hashp->BSIZE << BYTE_SHIFT) - 1);
    833 
    834 	if (!(freep = hashp->mapp[free_page]))
    835 		freep = fetch_bitmap(hashp, free_page);
    836 #ifdef DEBUG
    837 	/*
    838 	 * This had better never happen.  It means we tried to read a bitmap
    839 	 * that has already had overflow pages allocated off it, and we
    840 	 * failed to read it from the file.
    841 	 */
    842 	if (!freep)
    843 		assert(0);
    844 #endif
    845 	CLRBIT(freep, free_bit);
    846 #ifdef DEBUG2
    847 	(void)fprintf(stderr, "FREE_OVFLPAGE: ADDR: %d BIT: %d PAGE %d\n",
    848 	    obufp->addr, free_bit, free_page);
    849 #endif
    850 	__reclaim_buf(hashp, obufp);
    851 }
    852 
    853 /*
    854  * Returns:
    855  *	 0 success
    856  *	-1 failure
    857  */
    858 static int
    859 open_temp(hashp)
    860 	HTAB *hashp;
    861 {
    862 	sigset_t set, oset;
    863 	static char namestr[] = "_hashXXXXXX";
    864 
    865 	/* Block signals; make sure file goes away at process exit. */
    866 	(void)sigfillset(&set);
    867 	(void)sigprocmask(SIG_BLOCK, &set, &oset);
    868 	if ((hashp->fp = mkstemp(namestr)) != -1) {
    869 		(void)unlink(namestr);
    870 		(void)fcntl(hashp->fp, F_SETFD, 1);
    871 	}
    872 	(void)sigprocmask(SIG_SETMASK, &oset, (sigset_t *)NULL);
    873 	return (hashp->fp != -1 ? 0 : -1);
    874 }
    875 
    876 /*
    877  * We have to know that the key will fit, but the last entry on the page is
    878  * an overflow pair, so we need to shift things.
    879  */
    880 static void
    881 squeeze_key(sp, key, val)
    882 	u_int16_t *sp;
    883 	const DBT *key, *val;
    884 {
    885 	register char *p;
    886 	u_int16_t free_space, n, off, pageno;
    887 
    888 	p = (char *)sp;
    889 	n = sp[0];
    890 	free_space = FREESPACE(sp);
    891 	off = OFFSET(sp);
    892 
    893 	pageno = sp[n - 1];
    894 	off -= key->size;
    895 	sp[n - 1] = off;
    896 	memmove(p + off, key->data, key->size);
    897 	off -= val->size;
    898 	sp[n] = off;
    899 	memmove(p + off, val->data, val->size);
    900 	sp[0] = n + 2;
    901 	sp[n + 1] = pageno;
    902 	sp[n + 2] = OVFLPAGE;
    903 	FREESPACE(sp) = free_space - PAIRSIZE(key, val);
    904 	OFFSET(sp) = off;
    905 }
    906 
    907 static u_int32_t *
    908 fetch_bitmap(hashp, ndx)
    909 	HTAB *hashp;
    910 	int ndx;
    911 {
    912 	if (ndx >= hashp->nmaps)
    913 		return (NULL);
    914 	if ((hashp->mapp[ndx] = (u_int32_t *)malloc(hashp->BSIZE)) == NULL)
    915 		return (NULL);
    916 	if (__get_page(hashp,
    917 	    (char *)hashp->mapp[ndx], hashp->BITMAPS[ndx], 0, 1, 1)) {
    918 		free(hashp->mapp[ndx]);
    919 		return (NULL);
    920 	}
    921 	return (hashp->mapp[ndx]);
    922 }
    923 
    924 #ifdef DEBUG4
    925 int
    926 print_chain(addr)
    927 	int addr;
    928 {
    929 	BUFHEAD *bufp;
    930 	short *bp, oaddr;
    931 
    932 	(void)fprintf(stderr, "%d ", addr);
    933 	bufp = __get_buf(hashp, addr, NULL, 0);
    934 	bp = (short *)bufp->page;
    935 	while (bp[0] && ((bp[bp[0]] == OVFLPAGE) ||
    936 		((bp[0] > 2) && bp[2] < REAL_KEY))) {
    937 		oaddr = bp[bp[0] - 1];
    938 		(void)fprintf(stderr, "%d ", (int)oaddr);
    939 		bufp = __get_buf(hashp, (int)oaddr, bufp, 0);
    940 		bp = (short *)bufp->page;
    941 	}
    942 	(void)fprintf(stderr, "\n");
    943 }
    944 #endif
    945