xref: /titanic_41/usr/src/cmd/sendmail/db/btree/bt_put.c (revision bdb9230ac765cb7af3fc1f4119caf2c5720dceb3)
1 /*-
2  * See the file LICENSE for redistribution information.
3  *
4  * Copyright (c) 1996, 1997, 1998
5  *	Sleepycat Software.  All rights reserved.
6  */
7 /*
8  * Copyright (c) 1990, 1993, 1994, 1995, 1996
9  *	Keith Bostic.  All rights reserved.
10  */
11 /*
12  * Copyright (c) 1990, 1993, 1994, 1995
13  *	The Regents of the University of California.  All rights reserved.
14  *
15  * This code is derived from software contributed to Berkeley by
16  * Mike Olson.
17  *
18  * Redistribution and use in source and binary forms, with or without
19  * modification, are permitted provided that the following conditions
20  * are met:
21  * 1. Redistributions of source code must retain the above copyright
22  *    notice, this list of conditions and the following disclaimer.
23  * 2. Redistributions in binary form must reproduce the above copyright
24  *    notice, this list of conditions and the following disclaimer in the
25  *    documentation and/or other materials provided with the distribution.
26  * 3. All advertising materials mentioning features or use of this software
27  *    must display the following acknowledgement:
28  *	This product includes software developed by the University of
29  *	California, Berkeley and its contributors.
30  * 4. Neither the name of the University nor the names of its contributors
31  *    may be used to endorse or promote products derived from this software
32  *    without specific prior written permission.
33  *
34  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
35  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
36  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
37  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
38  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
39  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
40  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
41  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
42  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
43  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
44  * SUCH DAMAGE.
45  */
46 
47 #include "config.h"
48 
49 #ifndef lint
50 static const char sccsid[] = "@(#)bt_put.c	10.54 (Sleepycat) 12/6/98";
51 #endif /* not lint */
52 
53 #ifndef NO_SYSTEM_INCLUDES
54 #include <sys/types.h>
55 
56 #include <errno.h>
57 #include <string.h>
58 #endif
59 
60 #include "db_int.h"
61 #include "db_page.h"
62 #include "btree.h"
63 
64 static int __bam_fixed __P((DBC *, DBT *));
65 static int __bam_ndup __P((DBC *, PAGE *, u_int32_t));
66 static int __bam_ovput __P((DBC *, PAGE *, u_int32_t, DBT *));
67 static int __bam_partial __P((DBC *,
68     DBT *, PAGE *, u_int32_t, u_int32_t, u_int32_t));
69 static u_int32_t __bam_partsize __P((DBT *, PAGE *, u_int32_t));
70 
71 /*
72  * __bam_iitem --
73  *	Insert an item into the tree.
74  *
75  * PUBLIC: int __bam_iitem __P((DBC *,
76  * PUBLIC:    PAGE **, db_indx_t *, DBT *, DBT *, u_int32_t, u_int32_t));
77  */
78 int
79 __bam_iitem(dbc, hp, indxp, key, data, op, flags)
80 	DBC *dbc;
81 	PAGE **hp;
82 	db_indx_t *indxp;
83 	DBT *key, *data;
84 	u_int32_t op, flags;
85 {
86 	BTREE *t;
87 	BKEYDATA *bk;
88 	DB *dbp;
89 	DBT tdbt;
90 	PAGE *h;
91 	db_indx_t indx, nbytes;
92 	u_int32_t data_size, have_bytes, need_bytes, needed;
93 	int bigkey, bigdata, dupadjust, replace, ret;
94 
95 	COMPQUIET(bk, NULL);
96 
97 	dbp = dbc->dbp;
98 	t = dbp->internal;
99 	h = *hp;
100 	indx = *indxp;
101 	dupadjust = replace = 0;
102 
103 	/*
104 	 * If it's a page of duplicates, call the common code to do the work.
105 	 *
106 	 * !!!
107 	 * Here's where the hp and indxp are important.  The duplicate code
108 	 * may decide to rework/rearrange the pages and indices we're using,
109 	 * so the caller must understand that the page stack may change.
110 	 */
111 	if (TYPE(h) == P_DUPLICATE) {
112 		/* Adjust the index for the new item if it's a DB_AFTER op. */
113 		if (op == DB_AFTER)
114 			++*indxp;
115 
116 		/* Remove the current item if it's a DB_CURRENT op. */
117 		if (op == DB_CURRENT) {
118 			bk = GET_BKEYDATA(*hp, *indxp);
119 			switch (B_TYPE(bk->type)) {
120 			case B_KEYDATA:
121 				nbytes = BKEYDATA_SIZE(bk->len);
122 				break;
123 			case B_OVERFLOW:
124 				nbytes = BOVERFLOW_SIZE;
125 				break;
126 			default:
127 				return (__db_pgfmt(dbp, h->pgno));
128 			}
129 			if ((ret = __db_ditem(dbc, *hp, *indxp, nbytes)) != 0)
130 				return (ret);
131 		}
132 
133 		/* Put the new/replacement item onto the page. */
134 		if ((ret = __db_dput(dbc, data, hp, indxp, __bam_new)) != 0)
135 			return (ret);
136 
137 		goto done;
138 	}
139 
140 	/* Handle fixed-length records: build the real record. */
141 	if (F_ISSET(dbp, DB_RE_FIXEDLEN) && data->size != t->recno->re_len) {
142 		tdbt = *data;
143 		if ((ret = __bam_fixed(dbc, &tdbt)) != 0)
144 			return (ret);
145 		data = &tdbt;
146 	}
147 
148 	/*
149 	 * Figure out how much space the data will take, including if it's a
150 	 * partial record.  If either of the key or data items won't fit on
151 	 * a page, we'll have to store them on overflow pages.
152 	 */
153 	bigkey = LF_ISSET(BI_NEWKEY) && key->size > t->bt_ovflsize;
154 	data_size = F_ISSET(data, DB_DBT_PARTIAL) ?
155 	    __bam_partsize(data, h, indx) : data->size;
156 	bigdata = data_size > t->bt_ovflsize;
157 
158 	needed = 0;
159 	if (LF_ISSET(BI_NEWKEY)) {
160 		/* If BI_NEWKEY is set we're adding a new key and data pair. */
161 		if (bigkey)
162 			needed += BOVERFLOW_PSIZE;
163 		else
164 			needed += BKEYDATA_PSIZE(key->size);
165 		if (bigdata)
166 			needed += BOVERFLOW_PSIZE;
167 		else
168 			needed += BKEYDATA_PSIZE(data_size);
169 	} else {
170 		/*
171 		 * We're either overwriting the data item of a key/data pair
172 		 * or we're adding the data item only, i.e. a new duplicate.
173 		 */
174 		if (op == DB_CURRENT) {
175 			bk = GET_BKEYDATA(h,
176 			    indx + (TYPE(h) == P_LBTREE ? O_INDX : 0));
177 			if (B_TYPE(bk->type) == B_KEYDATA)
178 				have_bytes = BKEYDATA_PSIZE(bk->len);
179 			else
180 				have_bytes = BOVERFLOW_PSIZE;
181 			need_bytes = 0;
182 		} else {
183 			have_bytes = 0;
184 			need_bytes = sizeof(db_indx_t);
185 		}
186 		if (bigdata)
187 			need_bytes += BOVERFLOW_PSIZE;
188 		else
189 			need_bytes += BKEYDATA_PSIZE(data_size);
190 
191 		if (have_bytes < need_bytes)
192 			needed += need_bytes - have_bytes;
193 	}
194 
195 	/*
196 	 * If there's not enough room, or the user has put a ceiling on the
197 	 * number of keys permitted in the page, split the page.
198 	 *
199 	 * XXX
200 	 * The t->bt_maxkey test here may be insufficient -- do we have to
201 	 * check in the btree split code, so we don't undo it there!?!?
202 	 */
203 	if (P_FREESPACE(h) < needed ||
204 	    (t->bt_maxkey != 0 && NUM_ENT(h) > t->bt_maxkey))
205 		return (DB_NEEDSPLIT);
206 
207 	/* Handle partial puts: build the real record. */
208 	if (F_ISSET(data, DB_DBT_PARTIAL)) {
209 		tdbt = *data;
210 		if ((ret = __bam_partial(dbc,
211 		    &tdbt, h, indx, data_size, flags)) != 0)
212 			return (ret);
213 		data = &tdbt;
214 	}
215 
216 	/*
217 	 * The code breaks it up into six cases:
218 	 *
219 	 * 1. Append a new key/data pair.
220 	 * 2. Insert a new key/data pair.
221 	 * 3. Append a new data item (a new duplicate).
222 	 * 4. Insert a new data item (a new duplicate).
223 	 * 5. Overflow item: delete and re-add the data item.
224 	 * 6. Replace the data item.
225 	 */
226 	if (LF_ISSET(BI_NEWKEY)) {
227 		switch (op) {
228 		case DB_AFTER:		/* 1. Append a new key/data pair. */
229 			indx += 2;
230 			*indxp += 2;
231 			break;
232 		case DB_BEFORE:		/* 2. Insert a new key/data pair. */
233 			break;
234 		default:
235 			return (EINVAL);
236 		}
237 
238 		/* Add the key. */
239 		if (bigkey) {
240 			if ((ret = __bam_ovput(dbc, h, indx, key)) != 0)
241 				return (ret);
242 		} else
243 			if ((ret = __db_pitem(dbc, h, indx,
244 			    BKEYDATA_SIZE(key->size), NULL, key)) != 0)
245 				return (ret);
246 		++indx;
247 	} else {
248 		switch (op) {
249 		case DB_AFTER:		/* 3. Append a new data item. */
250 			if (TYPE(h) == P_LBTREE) {
251 				/*
252 				 * Adjust the cursor and copy in the key for
253 				 * the duplicate.
254 				 */
255 				if ((ret = __bam_adjindx(dbc,
256 				    h, indx + P_INDX, indx, 1)) != 0)
257 					return (ret);
258 
259 				indx += 3;
260 				dupadjust = 1;
261 
262 				*indxp += 2;
263 			} else {
264 				++indx;
265 				__bam_ca_di(dbp, h->pgno, indx, 1);
266 
267 				*indxp += 1;
268 			}
269 			break;
270 		case DB_BEFORE:		/* 4. Insert a new data item. */
271 			if (TYPE(h) == P_LBTREE) {
272 				/*
273 				 * Adjust the cursor and copy in the key for
274 				 * the duplicate.
275 				 */
276 				if ((ret =
277 				    __bam_adjindx(dbc, h, indx, indx, 1)) != 0)
278 					return (ret);
279 
280 				++indx;
281 				dupadjust = 1;
282 			} else
283 				__bam_ca_di(dbp, h->pgno, indx, 1);
284 			break;
285 		case DB_CURRENT:
286 			if (TYPE(h) == P_LBTREE)
287 				++indx;
288 
289 			/*
290 			 * 5. Delete/re-add the data item.
291 			 *
292 			 * If we're dealing with offpage items, we have to
293 			 * delete and then re-add the item.
294 			 */
295 			if (bigdata || B_TYPE(bk->type) != B_KEYDATA) {
296 				if ((ret = __bam_ditem(dbc, h, indx)) != 0)
297 					return (ret);
298 				break;
299 			}
300 
301 			/* 6. Replace the data item. */
302 			replace = 1;
303 			break;
304 		default:
305 			return (EINVAL);
306 		}
307 	}
308 
309 	/* Add the data. */
310 	if (bigdata) {
311 		if ((ret = __bam_ovput(dbc, h, indx, data)) != 0)
312 			return (ret);
313 	} else {
314 		BKEYDATA __bk;
315 		DBT __hdr;
316 
317 		if (LF_ISSET(BI_DELETED)) {
318 			B_TSET(__bk.type, B_KEYDATA, 1);
319 			__bk.len = data->size;
320 			__hdr.data = &__bk;
321 			__hdr.size = SSZA(BKEYDATA, data);
322 			ret = __db_pitem(dbc, h, indx,
323 			    BKEYDATA_SIZE(data->size), &__hdr, data);
324 		} else if (replace)
325 			ret = __bam_ritem(dbc, h, indx, data);
326 		else
327 			ret = __db_pitem(dbc, h, indx,
328 			    BKEYDATA_SIZE(data->size), NULL, data);
329 		if (ret != 0)
330 			return (ret);
331 	}
332 
333 	if ((ret = memp_fset(dbp->mpf, h, DB_MPOOL_DIRTY)) != 0)
334 		return (ret);
335 
336 	/*
337 	 * If the page is at least 50% full, and we added a duplicate, see if
338 	 * that set of duplicates takes up at least 25% of the space.  If it
339 	 * does, move it off onto its own page.
340 	 */
341 	if (dupadjust && P_FREESPACE(h) <= dbp->pgsize / 2) {
342 		--indx;
343 		if ((ret = __bam_ndup(dbc, h, indx)) != 0)
344 			return (ret);
345 	}
346 
347 	/*
348 	 * If we've changed the record count, update the tree.  Record counts
349 	 * need to be updated in recno databases and in btree databases where
350 	 * we are supporting records.  In both cases, adjust the count if the
351 	 * operation wasn't performed on the current record or when the caller
352 	 * overrides and wants the adjustment made regardless.
353 	 */
354 done:	if (LF_ISSET(BI_DOINCR) ||
355 	    (op != DB_CURRENT &&
356 	    (F_ISSET(dbp, DB_BT_RECNUM) || dbp->type == DB_RECNO)))
357 		if ((ret = __bam_adjust(dbc, 1)) != 0)
358 			return (ret);
359 
360 	/* If we've modified a recno file, set the flag */
361 	if (t->recno != NULL)
362 		F_SET(t->recno, RECNO_MODIFIED);
363 
364 	return (ret);
365 }
366 
367 /*
368  * __bam_partsize --
369  *	Figure out how much space a partial data item is in total.
370  */
371 static u_int32_t
372 __bam_partsize(data, h, indx)
373 	DBT *data;
374 	PAGE *h;
375 	u_int32_t indx;
376 {
377 	BKEYDATA *bk;
378 	u_int32_t nbytes;
379 
380 	/*
381 	 * Figure out how much total space we'll need.  If the record doesn't
382 	 * already exist, it's simply the data we're provided.
383 	 */
384 	if (indx >= NUM_ENT(h))
385 		return (data->doff + data->size);
386 
387 	/*
388 	 * Otherwise, it's the data provided plus any already existing data
389 	 * that we're not replacing.
390 	 */
391 	bk = GET_BKEYDATA(h, indx + (TYPE(h) == P_LBTREE ? O_INDX : 0));
392 	nbytes =
393 	    B_TYPE(bk->type) == B_OVERFLOW ? ((BOVERFLOW *)bk)->tlen : bk->len;
394 
395 	/*
396 	 * There are really two cases here:
397 	 *
398 	 * Case 1: We are replacing some bytes that do not exist (i.e., they
399 	 * are past the end of the record).  In this case the number of bytes
400 	 * we are replacing is irrelevant and all we care about is how many
401 	 * bytes we are going to add from offset.  So, the new record length
402 	 * is going to be the size of the new bytes (size) plus wherever those
403 	 * new bytes begin (doff).
404 	 *
405 	 * Case 2: All the bytes we are replacing exist.  Therefore, the new
406 	 * size is the oldsize (nbytes) minus the bytes we are replacing (dlen)
407 	 * plus the bytes we are adding (size).
408 	 */
409 	if (nbytes < data->doff + data->dlen)		/* Case 1 */
410 		return (data->doff + data->size);
411 
412 	return (nbytes + data->size - data->dlen);	/* Case 2 */
413 }
414 
415 /*
416  * OVPUT --
417  *	Copy an overflow item onto a page.
418  */
419 #undef	OVPUT
420 #define	OVPUT(h, indx, bo) do {						\
421 	DBT __hdr;							\
422 	memset(&__hdr, 0, sizeof(__hdr));				\
423 	__hdr.data = &bo;						\
424 	__hdr.size = BOVERFLOW_SIZE;					\
425 	if ((ret = __db_pitem(dbc,					\
426 	    h, indx, BOVERFLOW_SIZE, &__hdr, NULL)) != 0)		\
427 		return (ret);						\
428 } while (0)
429 
430 /*
431  * __bam_ovput --
432  *	Build an overflow item and put it on the page.
433  */
434 static int
435 __bam_ovput(dbc, h, indx, item)
436 	DBC *dbc;
437 	PAGE *h;
438 	u_int32_t indx;
439 	DBT *item;
440 {
441 	BOVERFLOW bo;
442 	int ret;
443 
444 	UMRW(bo.unused1);
445 	B_TSET(bo.type, B_OVERFLOW, 0);
446 	UMRW(bo.unused2);
447 	if ((ret = __db_poff(dbc, item, &bo.pgno, __bam_new)) != 0)
448 		return (ret);
449 	bo.tlen = item->size;
450 
451 	OVPUT(h, indx, bo);
452 
453 	return (0);
454 }
455 
456 /*
457  * __bam_ritem --
458  *	Replace an item on a page.
459  *
460  * PUBLIC: int __bam_ritem __P((DBC *, PAGE *, u_int32_t, DBT *));
461  */
462 int
463 __bam_ritem(dbc, h, indx, data)
464 	DBC *dbc;
465 	PAGE *h;
466 	u_int32_t indx;
467 	DBT *data;
468 {
469 	BKEYDATA *bk;
470 	DB *dbp;
471 	DBT orig, repl;
472 	db_indx_t cnt, lo, ln, min, off, prefix, suffix;
473 	int32_t nbytes;
474 	int ret;
475 	u_int8_t *p, *t;
476 
477 	dbp = dbc->dbp;
478 
479 	/*
480 	 * Replace a single item onto a page.  The logic figuring out where
481 	 * to insert and whether it fits is handled in the caller.  All we do
482 	 * here is manage the page shuffling.
483 	 */
484 	bk = GET_BKEYDATA(h, indx);
485 
486 	/* Log the change. */
487 	if (DB_LOGGING(dbc)) {
488 		/*
489 		 * We might as well check to see if the two data items share
490 		 * a common prefix and suffix -- it can save us a lot of log
491 		 * message if they're large.
492 		 */
493 		min = data->size < bk->len ? data->size : bk->len;
494 		for (prefix = 0,
495 		    p = bk->data, t = data->data;
496 		    prefix < min && *p == *t; ++prefix, ++p, ++t)
497 			;
498 
499 		min -= prefix;
500 		for (suffix = 0,
501 		    p = (u_int8_t *)bk->data + bk->len - 1,
502 		    t = (u_int8_t *)data->data + data->size - 1;
503 		    suffix < min && *p == *t; ++suffix, --p, --t)
504 			;
505 
506 		/* We only log the parts of the keys that have changed. */
507 		orig.data = (u_int8_t *)bk->data + prefix;
508 		orig.size = bk->len - (prefix + suffix);
509 		repl.data = (u_int8_t *)data->data + prefix;
510 		repl.size = data->size - (prefix + suffix);
511 		if ((ret = __bam_repl_log(dbp->dbenv->lg_info, dbc->txn,
512 		    &LSN(h), 0, dbp->log_fileid, PGNO(h), &LSN(h),
513 		    (u_int32_t)indx, (u_int32_t)B_DISSET(bk->type),
514 		    &orig, &repl, (u_int32_t)prefix, (u_int32_t)suffix)) != 0)
515 			return (ret);
516 	}
517 
518 	/*
519 	 * Set references to the first in-use byte on the page and the
520 	 * first byte of the item being replaced.
521 	 */
522 	p = (u_int8_t *)h + HOFFSET(h);
523 	t = (u_int8_t *)bk;
524 
525 	/*
526 	 * If the entry is growing in size, shift the beginning of the data
527 	 * part of the page down.  If the entry is shrinking in size, shift
528 	 * the beginning of the data part of the page up.  Use memmove(3),
529 	 * the regions overlap.
530 	 */
531 	lo = BKEYDATA_SIZE(bk->len);
532 	ln = BKEYDATA_SIZE(data->size);
533 	if (lo != ln) {
534 		nbytes = lo - ln;		/* Signed difference. */
535 		if (p == t)			/* First index is fast. */
536 			h->inp[indx] += nbytes;
537 		else {				/* Else, shift the page. */
538 			memmove(p + nbytes, p, t - p);
539 
540 			/* Adjust the indices' offsets. */
541 			off = h->inp[indx];
542 			for (cnt = 0; cnt < NUM_ENT(h); ++cnt)
543 				if (h->inp[cnt] <= off)
544 					h->inp[cnt] += nbytes;
545 		}
546 
547 		/* Clean up the page and adjust the item's reference. */
548 		HOFFSET(h) += nbytes;
549 		t += nbytes;
550 	}
551 
552 	/* Copy the new item onto the page. */
553 	bk = (BKEYDATA *)t;
554 	B_TSET(bk->type, B_KEYDATA, 0);
555 	bk->len = data->size;
556 	memcpy(bk->data, data->data, data->size);
557 
558 	return (0);
559 }
560 
561 /*
562  * __bam_ndup --
563  *	Check to see if the duplicate set at indx should have its own page.
564  *	If it should, create it.
565  */
566 static int
567 __bam_ndup(dbc, h, indx)
568 	DBC *dbc;
569 	PAGE *h;
570 	u_int32_t indx;
571 {
572 	BKEYDATA *bk;
573 	BOVERFLOW bo;
574 	DB *dbp;
575 	DBT hdr;
576 	PAGE *cp;
577 	db_indx_t cnt, cpindx, first, sz;
578 	int ret;
579 
580 	dbp = dbc->dbp;
581 
582 	while (indx > 0 && h->inp[indx] == h->inp[indx - P_INDX])
583 		indx -= P_INDX;
584 	for (cnt = 0, sz = 0, first = indx;; ++cnt, indx += P_INDX) {
585 		if (indx >= NUM_ENT(h) || h->inp[first] != h->inp[indx])
586 			break;
587 		bk = GET_BKEYDATA(h, indx);
588 		sz += B_TYPE(bk->type) == B_KEYDATA ?
589 		    BKEYDATA_PSIZE(bk->len) : BOVERFLOW_PSIZE;
590 		bk = GET_BKEYDATA(h, indx + O_INDX);
591 		sz += B_TYPE(bk->type) == B_KEYDATA ?
592 		    BKEYDATA_PSIZE(bk->len) : BOVERFLOW_PSIZE;
593 	}
594 
595 	/*
596 	 * If this set of duplicates is using more than 25% of the page, move
597 	 * them off.  The choice of 25% is a WAG, but it has to be small enough
598 	 * that we can always split regardless of the presence of duplicates.
599 	 */
600 	if (sz < dbp->pgsize / 4)
601 		return (0);
602 
603 	/* Get a new page. */
604 	if ((ret = __bam_new(dbc, P_DUPLICATE, &cp)) != 0)
605 		return (ret);
606 
607 	/*
608 	 * Move this set of duplicates off the page.  First points to the first
609 	 * key of the first duplicate key/data pair, cnt is the number of pairs
610 	 * we're dealing with.
611 	 */
612 	memset(&hdr, 0, sizeof(hdr));
613 	for (indx = first + O_INDX, cpindx = 0;; ++cpindx) {
614 		/* Copy the entry to the new page. */
615 		bk = GET_BKEYDATA(h, indx);
616 		hdr.data = bk;
617 		hdr.size = B_TYPE(bk->type) == B_KEYDATA ?
618 		    BKEYDATA_SIZE(bk->len) : BOVERFLOW_SIZE;
619 		if ((ret =
620 		    __db_pitem(dbc, cp, cpindx, hdr.size, &hdr, NULL)) != 0)
621 			goto err;
622 
623 		/*
624 		 * Move cursors referencing the old entry to the new entry.
625 		 * Done after the page put because __db_pitem() adjusts
626 		 * cursors on the new page, and before the delete because
627 		 * __db_ditem adjusts cursors on the old page.
628 		 */
629 		__bam_ca_dup(dbp,
630 		    PGNO(h), first, indx - O_INDX, PGNO(cp), cpindx);
631 
632 		/* Delete the data item. */
633 		if ((ret = __db_ditem(dbc, h, indx, hdr.size)) != 0)
634 			goto err;
635 
636 		/* Delete all but the first reference to the key. */
637 		if (--cnt == 0)
638 			break;
639 		if ((ret = __bam_adjindx(dbc, h, indx, first, 0)) != 0)
640 			goto err;
641 	}
642 
643 	/* Put in a new data item that points to the duplicates page. */
644 	UMRW(bo.unused1);
645 	B_TSET(bo.type, B_DUPLICATE, 0);
646 	UMRW(bo.unused2);
647 	bo.pgno = cp->pgno;
648 	bo.tlen = 0;
649 
650 	OVPUT(h, indx, bo);
651 
652 	return (memp_fput(dbp->mpf, cp, DB_MPOOL_DIRTY));
653 
654 err:	(void)__bam_free(dbc, cp);
655 	return (ret);
656 }
657 
658 /*
659  * __bam_fixed --
660  *	Build the real record for a fixed length put.
661  */
662 static int
663 __bam_fixed(dbc, dbt)
664 	DBC *dbc;
665 	DBT *dbt;
666 {
667 	DB *dbp;
668 	RECNO *rp;
669 	int ret;
670 
671 	dbp = dbc->dbp;
672 	rp = ((BTREE *)dbp->internal)->recno;
673 
674 	/*
675 	 * If database contains fixed-length records, and the record is long,
676 	 * return EINVAL.
677 	 */
678 	if (dbt->size > rp->re_len)
679 		return (EINVAL);
680 
681 	/*
682 	 * The caller checked to see if it was just right, so we know it's
683 	 * short.  Pad it out.  We use the record data return memory, it's
684 	 * only a short-term use.
685 	 */
686 	if (dbc->rdata.ulen < rp->re_len) {
687 		 if ((ret = __os_realloc(&dbc->rdata.data, rp->re_len)) != 0) {
688 			dbc->rdata.ulen = 0;
689 			dbc->rdata.data = NULL;
690 			return (ret);
691 		}
692 		dbc->rdata.ulen = rp->re_len;
693 	}
694 	memcpy(dbc->rdata.data, dbt->data, dbt->size);
695 	memset((u_int8_t *)dbc->rdata.data + dbt->size,
696 	    rp->re_pad, rp->re_len - dbt->size);
697 
698 	/*
699 	 * Clean up our flags and other information just in case, and
700 	 * change the caller's DBT to reference our created record.
701 	 */
702 	dbc->rdata.size = rp->re_len;
703 	dbc->rdata.dlen = 0;
704 	dbc->rdata.doff = 0;
705 	dbc->rdata.flags = 0;
706 	*dbt = dbc->rdata;
707 
708 	return (0);
709 }
710 
711 /*
712  * __bam_partial --
713  *	Build the real record for a partial put.
714  */
715 static int
716 __bam_partial(dbc, dbt, h, indx, nbytes, flags)
717 	DBC *dbc;
718 	DBT *dbt;
719 	PAGE *h;
720 	u_int32_t indx, nbytes, flags;
721 {
722 	BKEYDATA *bk, tbk;
723 	BOVERFLOW *bo;
724 	DB *dbp;
725 	DBT copy;
726 	u_int32_t len, tlen;
727 	u_int8_t *p;
728 	int ret;
729 
730 	COMPQUIET(bo, NULL);
731 
732 	dbp = dbc->dbp;
733 
734 	/* We use the record data return memory, it's only a short-term use. */
735 	if (dbc->rdata.ulen < nbytes) {
736 		 if ((ret = __os_realloc(&dbc->rdata.data, nbytes)) != 0) {
737 			dbc->rdata.ulen = 0;
738 			dbc->rdata.data = NULL;
739 			return (ret);
740 		}
741 		dbc->rdata.ulen = nbytes;
742 	}
743 
744 	/*
745 	 * We use nul bytes for any part of the record that isn't specified;
746 	 * get it over with.
747 	 */
748 	memset(dbc->rdata.data, 0, nbytes);
749 
750 	/*
751 	 * In the next clauses, we need to do three things: a) set p to point
752 	 * to the place at which to copy the user's data, b) set tlen to the
753 	 * total length of the record, not including the bytes contributed by
754 	 * the user, and c) copy any valid data from an existing record.
755 	 */
756 	if (LF_ISSET(BI_NEWKEY)) {
757 		tlen = dbt->doff;
758 		p = (u_int8_t *)dbc->rdata.data + dbt->doff;
759 		goto ucopy;
760 	}
761 
762 	/* Find the current record. */
763 	if (indx < NUM_ENT(h)) {
764 		bk = GET_BKEYDATA(h, indx + (TYPE(h) == P_LBTREE ? O_INDX : 0));
765 		bo = (BOVERFLOW *)bk;
766 	} else {
767 		bk = &tbk;
768 		B_TSET(bk->type, B_KEYDATA, 0);
769 		bk->len = 0;
770 	}
771 	if (B_TYPE(bk->type) == B_OVERFLOW) {
772 		/*
773 		 * In the case of an overflow record, we shift things around
774 		 * in the current record rather than allocate a separate copy.
775 		 */
776 		memset(&copy, 0, sizeof(copy));
777 		if ((ret = __db_goff(dbp, &copy, bo->tlen,
778 		    bo->pgno, &dbc->rdata.data, &dbc->rdata.ulen)) != 0)
779 			return (ret);
780 
781 		/* Skip any leading data from the original record. */
782 		tlen = dbt->doff;
783 		p = (u_int8_t *)dbc->rdata.data + dbt->doff;
784 
785 		/*
786 		 * Copy in any trailing data from the original record.
787 		 *
788 		 * If the original record was larger than the original offset
789 		 * plus the bytes being deleted, there is trailing data in the
790 		 * original record we need to preserve.  If we aren't deleting
791 		 * the same number of bytes as we're inserting, copy it up or
792 		 * down, into place.
793 		 *
794 		 * Use memmove(), the regions may overlap.
795 		 */
796 		if (bo->tlen > dbt->doff + dbt->dlen) {
797 			len = bo->tlen - (dbt->doff + dbt->dlen);
798 			if (dbt->dlen != dbt->size)
799 				memmove(p + dbt->size, p + dbt->dlen, len);
800 			tlen += len;
801 		}
802 	} else {
803 		/* Copy in any leading data from the original record. */
804 		memcpy(dbc->rdata.data,
805 		    bk->data, dbt->doff > bk->len ? bk->len : dbt->doff);
806 		tlen = dbt->doff;
807 		p = (u_int8_t *)dbc->rdata.data + dbt->doff;
808 
809 		/* Copy in any trailing data from the original record. */
810 		len = dbt->doff + dbt->dlen;
811 		if (bk->len > len) {
812 			memcpy(p + dbt->size, bk->data + len, bk->len - len);
813 			tlen += bk->len - len;
814 		}
815 	}
816 
817 ucopy:	/*
818 	 * Copy in the application provided data -- p and tlen must have been
819 	 * initialized above.
820 	 */
821 	memcpy(p, dbt->data, dbt->size);
822 	tlen += dbt->size;
823 
824 	/* Set the DBT to reference our new record. */
825 	dbc->rdata.size = tlen;
826 	dbc->rdata.dlen = 0;
827 	dbc->rdata.doff = 0;
828 	dbc->rdata.flags = 0;
829 	*dbt = dbc->rdata;
830 	return (0);
831 }
832