summary refs log tree commit diff
path: root/db2/btree/bt_delete.c
blob: baa8a254019015c2a23bf90d438e0eea8339d19e (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
/*-
 * See the file LICENSE for redistribution information.
 *
 * Copyright (c) 1996, 1997
 *	Sleepycat Software.  All rights reserved.
 */
/*
 * Copyright (c) 1990, 1993, 1994, 1995, 1996
 *	Keith Bostic.  All rights reserved.
 */
/*
 * Copyright (c) 1990, 1993, 1994, 1995
 *	The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * Mike Olson.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

#include "config.h"

#ifndef lint
static const char sccsid[] = "@(#)bt_delete.c	10.25 (Sleepycat) 1/8/98";
#endif /* not lint */

#ifndef NO_SYSTEM_INCLUDES
#include <sys/types.h>

#include <stdio.h>
#include <string.h>
#endif

#include "db_int.h"
#include "db_page.h"
#include "btree.h"

static int __bam_dpages __P((DB *, BTREE *));

/*
 * __bam_delete --
 *	Delete the items referenced by a key.
 *
 * PUBLIC: int __bam_delete __P((DB *, DB_TXN *, DBT *, int));
 */
int
__bam_delete(argdbp, txn, key, flags)
	DB *argdbp;
	DB_TXN *txn;
	DBT *key;
	int flags;
{
	BTREE *t;
	DB *dbp;
	PAGE *h;
	db_indx_t cnt, i, indx;
	int dpage, exact, ret, stack;

	DEBUG_LWRITE(argdbp, txn, "bam_delete", key, NULL, flags);

	stack = 0;

	/* Check for invalid flags. */
	if ((ret =
	    __db_delchk(argdbp, flags, F_ISSET(argdbp, DB_AM_RDONLY))) != 0)
		return (ret);

	GETHANDLE(argdbp, txn, &dbp, ret);
	t = dbp->internal;

	/* Search the tree for the key; delete only deletes exact matches. */
	if ((ret = __bam_search(dbp, key, S_DELETE, 1, NULL, &exact)) != 0)
		goto err;
	stack = 1;
	h = t->bt_csp->page;
	indx = t->bt_csp->indx;

	/* Delete the key/data pair, including any on-or-off page duplicates. */
	for (cnt = 1, i = indx;; ++cnt)
		if ((i += P_INDX) >= NUM_ENT(h) || h->inp[i] != h->inp[indx])
			break;
	for (; cnt > 0; --cnt, ++t->lstat.bt_deleted)
		if (__bam_ca_delete(dbp, h->pgno, indx, NULL, 1) == 0) {
			if ((ret = __bam_ditem(dbp, h, indx)) != 0)
				goto err;
			if ((ret = __bam_ditem(dbp, h, indx)) != 0)
				goto err;
		} else {
			B_DSET(GET_BKEYDATA(h, indx + O_INDX)->type);
			indx += P_INDX;
		}

	/* If we're using record numbers, update internal page record counts. */
	if (F_ISSET(dbp, DB_BT_RECNUM) && (ret = __bam_adjust(dbp, t, -1)) != 0)
		goto err;

	/* If the page is now empty, delete it. */
	dpage = NUM_ENT(h) == 0 && h->pgno != PGNO_ROOT;

	__bam_stkrel(dbp);
	stack = 0;

	ret = dpage ? __bam_dpage(dbp, key) : 0;

err:	if (stack)
		__bam_stkrel(dbp);
	PUTHANDLE(dbp);
	return (ret);
}

/*
 * __ram_delete --
 *	Delete the items referenced by a key.
 *
 * PUBLIC: int __ram_delete __P((DB *, DB_TXN *, DBT *, int));
 */
int
__ram_delete(argdbp, txn, key, flags)
	DB *argdbp;
	DB_TXN *txn;
	DBT *key;
	int flags;
{
	BKEYDATA bk;
	BTREE *t;
	DB *dbp;
	DBT hdr, data;
	PAGE *h;
	db_indx_t indx;
	db_recno_t recno;
	int exact, ret, stack;

	stack = 0;

	/* Check for invalid flags. */
	if ((ret =
	    __db_delchk(argdbp, flags, F_ISSET(argdbp, DB_AM_RDONLY))) != 0)
		return (ret);

	GETHANDLE(argdbp, txn, &dbp, ret);
	t = dbp->internal;

	/* Check the user's record number and fill in as necessary. */
	if ((ret = __ram_getno(argdbp, key, &recno, 0)) != 0)
		goto err;

	/* Search the tree for the key; delete only deletes exact matches. */
	if ((ret = __bam_rsearch(dbp, &recno, S_DELETE, 1, &exact)) != 0)
		goto err;
	if (!exact) {
		ret = DB_NOTFOUND;
		goto err;
	}

	h = t->bt_csp->page;
	indx = t->bt_csp->indx;
	stack = 1;

	/* If the record has already been deleted, we couldn't have found it. */
	if (B_DISSET(GET_BKEYDATA(h, indx)->type)) {
		ret = DB_KEYEMPTY;
		goto done;
	}

	/*
	 * If we're not renumbering records, replace the record with a marker
	 * and return.
	 */
	if (!F_ISSET(dbp, DB_RE_RENUMBER)) {
		if ((ret = __bam_ditem(dbp, h, indx)) != 0)
			goto err;

		B_TSET(bk.type, B_KEYDATA, 1);
		bk.len = 0;
		memset(&hdr, 0, sizeof(hdr));
		hdr.data = &bk;
		hdr.size = SSZA(BKEYDATA, data);
		memset(&data, 0, sizeof(data));
		data.data = (char *)"";
		data.size = 0;
		if ((ret = __db_pitem(dbp,
		    h, indx, BKEYDATA_SIZE(0), &hdr, &data)) != 0)
			goto err;

		++t->lstat.bt_deleted;
		goto done;
	}

	/* Delete the item. */
	if ((ret = __bam_ditem(dbp, h, indx)) != 0)
		goto err;

	++t->lstat.bt_deleted;
	if (t->bt_recno != NULL)
		F_SET(t->bt_recno, RECNO_MODIFIED);

	/* Adjust the counts. */
	__bam_adjust(dbp, t, -1);

	/* Adjust the cursors. */
	__ram_ca(dbp, recno, CA_DELETE);

	/*
	 * If the page is now empty, delete it -- we have the whole tree
	 * locked, so there are no preparations to make.  Else, release
	 * the pages.
	 */
	if (NUM_ENT(h) == 0 && h->pgno != PGNO_ROOT) {
		stack = 0;
		ret = __bam_dpages(dbp, t);
	}

done:
err:	if (stack)
		__bam_stkrel(dbp);

	PUTHANDLE(dbp);
	return (ret);
}

/*
 * __bam_ditem --
 *	Delete one or more entries from a page.
 *
 * PUBLIC: int __bam_ditem __P((DB *, PAGE *, u_int32_t));
 */
int
__bam_ditem(dbp, h, indx)
	DB *dbp;
	PAGE *h;
	u_int32_t indx;
{
	BINTERNAL *bi;
	BKEYDATA *bk;
	BOVERFLOW *bo;
	u_int32_t nbytes;
	int ret;

	switch (TYPE(h)) {
	case P_IBTREE:
		bi = GET_BINTERNAL(h, indx);
		switch (B_TYPE(bi->type)) {
		case B_DUPLICATE:
		case B_OVERFLOW:
			nbytes = BINTERNAL_SIZE(bi->len);
			bo = (BOVERFLOW *)bi->data;
			goto offpage;
		case B_KEYDATA:
			nbytes = BINTERNAL_SIZE(bi->len);
			break;
		default:
			return (__db_pgfmt(dbp, h->pgno));
		}
		break;
	case P_IRECNO:
		nbytes = RINTERNAL_SIZE;
		break;
	case P_LBTREE:
		/*
		 * If it's a duplicate key, discard the index and don't touch
		 * the actual page item.  This works because no data item can
		 * have an index that matches any other index so even if the
		 * data item is in an index "slot", it won't match any other
		 * index.
		 */
		if (!(indx % 2)) {
			if (indx > 0 && h->inp[indx] == h->inp[indx - P_INDX])
				return (__bam_adjindx(dbp,
				    h, indx, indx - P_INDX, 0));
			if (indx + P_INDX < (u_int32_t)NUM_ENT(h) &&
			    h->inp[indx] == h->inp[indx + P_INDX])
				return (__bam_adjindx(dbp,
				    h, indx, indx + O_INDX, 0));
		}
		/* FALLTHROUGH */
	case P_LRECNO:
		bk = GET_BKEYDATA(h, indx);
		switch (B_TYPE(bk->type)) {
		case B_DUPLICATE:
		case B_OVERFLOW:
			nbytes = BOVERFLOW_SIZE;
			bo = GET_BOVERFLOW(h, indx);

offpage:		/* Delete duplicate/offpage chains. */
			if (B_TYPE(bo->type) == B_DUPLICATE) {
				if ((ret =
				    __db_ddup(dbp, bo->pgno, __bam_free)) != 0)
					return (ret);
			} else
				if ((ret =
				    __db_doff(dbp, bo->pgno, __bam_free)) != 0)
					return (ret);
			break;
		case B_KEYDATA:
			nbytes = BKEYDATA_SIZE(bk->len);
			break;
		default:
			return (__db_pgfmt(dbp, h->pgno));
		}
		break;
	default:
		return (__db_pgfmt(dbp, h->pgno));
	}

	/* Delete the item. */
	if ((ret = __db_ditem(dbp, h, indx, nbytes)) != 0)
		return (ret);

	/* Mark the page dirty. */
	return (memp_fset(dbp->mpf, h, DB_MPOOL_DIRTY));
}

/*
 * __bam_adjindx --
 *	Adjust an index on the page.
 *
 * PUBLIC: int __bam_adjindx __P((DB *, PAGE *, u_int32_t, u_int32_t, int));
 */
int
__bam_adjindx(dbp, h, indx, indx_copy, is_insert)
	DB *dbp;
	PAGE *h;
	u_int32_t indx, indx_copy;
	int is_insert;
{
	db_indx_t copy;
	int ret;

	/* Log the change. */
	if (DB_LOGGING(dbp) &&
	    (ret = __bam_adj_log(dbp->dbenv->lg_info, dbp->txn, &LSN(h),
	    0, dbp->log_fileid, PGNO(h), &LSN(h), indx, indx_copy,
	    (u_int32_t)is_insert)) != 0)
		return (ret);

	if (is_insert) {
		copy = h->inp[indx_copy];
		if (indx != NUM_ENT(h))
			memmove(&h->inp[indx + O_INDX], &h->inp[indx],
			    sizeof(db_indx_t) * (NUM_ENT(h) - indx));
		h->inp[indx] = copy;
		++NUM_ENT(h);
	} else {
		--NUM_ENT(h);
		if (indx != NUM_ENT(h))
			memmove(&h->inp[indx], &h->inp[indx + O_INDX],
			    sizeof(db_indx_t) * (NUM_ENT(h) - indx));
	}

	/* Mark the page dirty. */
	ret = memp_fset(dbp->mpf, h, DB_MPOOL_DIRTY);

	/* Adjust the cursors. */
	__bam_ca_di(dbp, h->pgno, indx, is_insert ? 1 : -1);
	return (0);
}

/*
 * __bam_dpage --
 *	Delete a page from the tree.
 *
 * PUBLIC: int __bam_dpage __P((DB *, const DBT *));
 */
int
__bam_dpage(dbp, key)
	DB *dbp;
	const DBT *key;
{
	BTREE *t;
	DB_LOCK lock;
	PAGE *h;
	db_pgno_t pgno;
	int exact, level, ret;

	ret = 0;
	t = dbp->internal;

	/*
	 * The locking protocol is that we acquire locks by walking down the
	 * tree, to avoid the obvious deadlocks.
	 *
	 * Call __bam_search to reacquire the empty leaf page, but this time
	 * get both the leaf page and it's parent, locked.  Walk back up the
	 * tree, until we have the top pair of pages that we want to delete.
	 * Once we have the top page that we want to delete locked, lock the
	 * underlying pages and check to make sure they're still empty.  If
	 * they are, delete them.
	 */
	for (level = LEAFLEVEL;; ++level) {
		/* Acquire a page and its parent, locked. */
		if ((ret =
		    __bam_search(dbp, key, S_WRPAIR, level, NULL, &exact)) != 0)
			return (ret);

		/*
		 * If we reach the root or the page isn't going to be empty
		 * when we delete one record, quit.
		 */
		h = t->bt_csp[-1].page;
		if (h->pgno == PGNO_ROOT || NUM_ENT(h) != 1)
			break;

		/* Release the two locked pages. */
		(void)memp_fput(dbp->mpf, t->bt_csp[-1].page, 0);
		(void)__BT_TLPUT(dbp, t->bt_csp[-1].lock);
		(void)memp_fput(dbp->mpf, t->bt_csp[0].page, 0);
		(void)__BT_TLPUT(dbp, t->bt_csp[0].lock);
	}

	/*
	 * Leave the stack pointer one after the last entry, we may be about
	 * to push more items on the stack.
	 */
	++t->bt_csp;

	/*
	 * t->bt_csp[-2].page is the top page, which we're not going to delete,
	 * and t->bt_csp[-1].page is the first page we are going to delete.
	 *
	 * Walk down the chain, acquiring the rest of the pages until we've
	 * retrieved the leaf page.  If we find any pages that aren't going
	 * to be emptied by the delete, someone else added something while we
	 * were walking the tree, and we discontinue the delete.
	 */
	for (h = t->bt_csp[-1].page;;) {
		if (ISLEAF(h)) {
			if (NUM_ENT(h) != 0)
				goto release;
			break;
		} else
			if (NUM_ENT(h) != 1)
				goto release;

		/*
		 * Get the next page, write lock it and push it onto the stack.
		 * We know it's index 0, because it can only have one element.
		 */
		pgno = TYPE(h) == P_IBTREE ?
		    GET_BINTERNAL(h, 0)->pgno : GET_RINTERNAL(h, 0)->pgno;

		if ((ret = __bam_lget(dbp, 0, pgno, DB_LOCK_WRITE, &lock)) != 0)
			goto release;
		if ((ret = __bam_pget(dbp, &h, &pgno, 0)) != 0)
			goto release;
		BT_STK_PUSH(t, h, 0, lock, ret);
		if (ret != 0)
			goto release;
	}

	BT_STK_POP(t);
	return (__bam_dpages(dbp, t));

release:
	/* Discard any locked pages and return. */
	BT_STK_POP(t);
	__bam_stkrel(dbp);
	return (ret);
}

/*
 * __bam_dpages --
 *	Delete a set of locked pages.
 */
static int
__bam_dpages(dbp, t)
	DB *dbp;
	BTREE *t;
{
	DBT a, b;
	DB_LOCK lock;
	EPG *epg;
	PAGE *h;
	db_pgno_t pgno;
	db_recno_t rcnt;
	int ret;

	COMPQUIET(rcnt, 0);

	epg = t->bt_sp;

	/*
	 * !!!
	 * There is an interesting deadlock situation here.  We have to relink
	 * the leaf page chain around the leaf page being deleted.  Consider
	 * a cursor walking through the leaf pages, that has the previous page
	 * read-locked and is waiting on a lock for the page we're deleting.
	 * It will deadlock here.  This is a problem, because if our process is
	 * selected to resolve the deadlock, we'll leave an empty leaf page
	 * that we can never again access by walking down the tree.  So, before
	 * we unlink the subtree, we relink the leaf page chain.
	 */
	if ((ret = __db_relink(dbp, t->bt_csp->page, NULL, 1)) != 0)
		goto release;

	/*
	 * We have the entire stack of deletable pages locked.  Start from the
	 * top of the tree and move to the bottom, as it's better to release
	 * the inner pages as soon as possible.
	 */
	if ((ret = __bam_ditem(dbp, epg->page, epg->indx)) != 0)
		goto release;

	/*
	 * If we deleted the next-to-last item from the root page, the tree
	 * can collapse a level.  Try and write lock the remaining root + 1
	 * page and copy it onto the root page.  If we can't get the lock,
	 * that's okay, the tree just stays a level deeper than we'd like.
	 */
	h = epg->page;
	if (h->pgno == PGNO_ROOT && NUM_ENT(h) == 1) {
		pgno = TYPE(epg->page) == P_IBTREE ?
		    GET_BINTERNAL(epg->page, 0)->pgno :
		    GET_RINTERNAL(epg->page, 0)->pgno;
		if ((ret = __bam_lget(dbp, 0, pgno, DB_LOCK_WRITE, &lock)) != 0)
			goto release;
		if ((ret = __bam_pget(dbp, &h, &pgno, 0)) != 0)
			goto release;

		/* Log the change. */
		if (DB_LOGGING(dbp)) {
			memset(&a, 0, sizeof(a));
			a.data = h;
			a.size = dbp->pgsize;
			memset(&b, 0, sizeof(b));
			b.data = P_ENTRY(epg->page, 0);
			b.size = BINTERNAL_SIZE(((BINTERNAL *)b.data)->len);
			__bam_rsplit_log(dbp->dbenv->lg_info, dbp->txn,
			   &h->lsn, 0, dbp->log_fileid, h->pgno, &a,
			   RE_NREC(epg->page), &b, &epg->page->lsn);
		}

		/*
		 * Make the switch.
		 *
		 * One fixup -- if the tree has record numbers and we're not
		 * converting to a leaf page, we have to preserve the total
		 * record count.
		 */
		if (TYPE(h) == P_IRECNO ||
		    (TYPE(h) == P_IBTREE && F_ISSET(dbp, DB_BT_RECNUM)))
			rcnt = RE_NREC(epg->page);
		memcpy(epg->page, h, dbp->pgsize);
		epg->page->pgno = PGNO_ROOT;
		if (TYPE(h) == P_IRECNO ||
		    (TYPE(h) == P_IBTREE && F_ISSET(dbp, DB_BT_RECNUM)))
			RE_NREC_SET(epg->page, rcnt);
		(void)memp_fset(dbp->mpf, epg->page, DB_MPOOL_DIRTY);

		/*
		 * Free the last page in that level of the btree and discard
		 * the lock.  (The call to __bam_free discards our reference
		 * to the page.)
		 */
		(void)__bam_free(dbp, h);
		(void)__BT_TLPUT(dbp, lock);
		++t->lstat.bt_freed;

		/* Adjust the cursors. */
		__bam_ca_move(dbp, h->pgno, PGNO_ROOT);
	}

	/* Release the top page in the subtree. */
	(void)memp_fput(dbp->mpf, epg->page, 0);
	(void)__BT_TLPUT(dbp, epg->lock);

	/*
	 * Free the rest of the pages.
	 *
	 * XXX
	 * Don't bother checking for errors.  We've unlinked the subtree from
	 * the tree, and there's no possibility of recovery.
	 */
	for (; ++epg <= t->bt_csp; ++t->lstat.bt_freed) {
		if (NUM_ENT(epg->page) != 0)
			(void)__bam_ditem(dbp, epg->page, epg->indx);

		(void)__bam_free(dbp, epg->page);
		(void)__BT_TLPUT(dbp, epg->lock);
	}
	return (0);

release:
	/* Discard any remaining pages and return. */
	for (; epg <= t->bt_csp; ++epg) {
		(void)memp_fput(dbp->mpf, epg->page, 0);
		(void)__BT_TLPUT(dbp, epg->lock);
	}
	return (ret);
}