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
50static 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
64static int __bam_fixed __P((DBC *, DBT *));
65static int __bam_ndup __P((DBC *, PAGE *, u_int32_t));
66static int __bam_ovput __P((DBC *, PAGE *, u_int32_t, DBT *));
67static int __bam_partial __P((DBC *,
68    DBT *, PAGE *, u_int32_t, u_int32_t, u_int32_t));
69static 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 */
78int
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	 */
354done:	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 */
371static 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 */
434static 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 */
462int
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 */
566static 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
654err:	(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 */
662static 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 */
715static 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
817ucopy:	/*
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