bt_delete.c revision 1.15
1/*	$NetBSD: bt_delete.c,v 1.15 2008/09/10 17:52:35 joerg Exp $	*/
2
3/*-
4 * Copyright (c) 1990, 1993, 1994
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Mike Olson.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#include <sys/cdefs.h>
36__RCSID("$NetBSD: bt_delete.c,v 1.15 2008/09/10 17:52:35 joerg Exp $");
37
38#include "namespace.h"
39#include <sys/types.h>
40
41#include <assert.h>
42#include <errno.h>
43#include <stdio.h>
44#include <string.h>
45
46#include <db.h>
47#include "btree.h"
48
49static int __bt_bdelete(BTREE *, const DBT *);
50static int __bt_curdel(BTREE *, const DBT *, PAGE *, u_int);
51static int __bt_pdelete(BTREE *, PAGE *);
52static int __bt_relink(BTREE *, PAGE *);
53static int __bt_stkacq(BTREE *, PAGE **, CURSOR *);
54
55/*
56 * __bt_delete
57 *	Delete the item(s) referenced by a key.
58 *
59 * Return RET_SPECIAL if the key is not found.
60 */
61int
62__bt_delete(const DB *dbp, const DBT *key, u_int flags)
63{
64	BTREE *t;
65	CURSOR *c;
66	PAGE *h;
67	int status;
68
69	t = dbp->internal;
70
71	/* Toss any page pinned across calls. */
72	if (t->bt_pinned != NULL) {
73		mpool_put(t->bt_mp, t->bt_pinned, 0);
74		t->bt_pinned = NULL;
75	}
76
77	/* Check for change to a read-only tree. */
78	if (F_ISSET(t, B_RDONLY)) {
79		errno = EPERM;
80		return (RET_ERROR);
81	}
82
83	switch (flags) {
84	case 0:
85		status = __bt_bdelete(t, key);
86		break;
87	case R_CURSOR:
88		/*
89		 * If flags is R_CURSOR, delete the cursor.  Must already
90		 * have started a scan and not have already deleted it.
91		 */
92		c = &t->bt_cursor;
93		if (F_ISSET(c, CURS_INIT)) {
94			if (F_ISSET(c, CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
95				return (RET_SPECIAL);
96			if ((h = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL)
97				return (RET_ERROR);
98
99			/*
100			 * If the page is about to be emptied, we'll need to
101			 * delete it, which means we have to acquire a stack.
102			 */
103			if (NEXTINDEX(h) == 1)
104				if (__bt_stkacq(t, &h, &t->bt_cursor))
105					return (RET_ERROR);
106
107			status = __bt_dleaf(t, NULL, h, (u_int)c->pg.index);
108
109			if (NEXTINDEX(h) == 0 && status == RET_SUCCESS) {
110				if (__bt_pdelete(t, h))
111					return (RET_ERROR);
112			} else
113				mpool_put(t->bt_mp, h,
114				    (u_int)(status == RET_SUCCESS ?
115				    MPOOL_DIRTY : 0));
116			break;
117		}
118		/* FALLTHROUGH */
119	default:
120		errno = EINVAL;
121		return (RET_ERROR);
122	}
123	if (status == RET_SUCCESS)
124		F_SET(t, B_MODIFIED);
125	return (status);
126}
127
128/*
129 * __bt_stkacq --
130 *	Acquire a stack so we can delete a cursor entry.
131 *
132 * Parameters:
133 *	  t:	tree
134 *	 hp:	pointer to current, pinned PAGE pointer
135 *	  c:	pointer to the cursor
136 *
137 * Returns:
138 *	0 on success, 1 on failure
139 */
140static int
141__bt_stkacq(BTREE *t, PAGE **hp, CURSOR *c)
142{
143	BINTERNAL *bi;
144	EPG *e;
145	EPGNO *parent;
146	PAGE *h;
147	indx_t idx = 0;	/* Pacify gcc */
148	pgno_t pgno;
149	recno_t nextpg, prevpg;
150	int exact, level;
151
152	/*
153	 * Find the first occurrence of the key in the tree.  Toss the
154	 * currently locked page so we don't hit an already-locked page.
155	 */
156	h = *hp;
157	mpool_put(t->bt_mp, h, 0);
158	if ((e = __bt_search(t, &c->key, &exact)) == NULL)
159		return (1);
160	h = e->page;
161
162	/* See if we got it in one shot. */
163	if (h->pgno == c->pg.pgno)
164		goto ret;
165
166	/*
167	 * Move right, looking for the page.  At each move we have to move
168	 * up the stack until we don't have to move to the next page.  If
169	 * we have to change pages at an internal level, we have to fix the
170	 * stack back up.
171	 */
172	while (h->pgno != c->pg.pgno) {
173		if ((nextpg = h->nextpg) == P_INVALID)
174			break;
175		mpool_put(t->bt_mp, h, 0);
176
177		/* Move up the stack. */
178		for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
179			/* Get the parent page. */
180			if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
181				return (1);
182
183			/* Move to the next index. */
184			if (parent->index != NEXTINDEX(h) - 1) {
185				idx = parent->index + 1;
186				BT_PUSH(t, h->pgno, idx);
187				break;
188			}
189			mpool_put(t->bt_mp, h, 0);
190		}
191
192		/* Restore the stack. */
193		while (level--) {
194			/* Push the next level down onto the stack. */
195			bi = GETBINTERNAL(h, idx);
196			pgno = bi->pgno;
197			BT_PUSH(t, pgno, 0);
198
199			/* Lose the currently pinned page. */
200			mpool_put(t->bt_mp, h, 0);
201
202			/* Get the next level down. */
203			if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
204				return (1);
205			idx = 0;
206		}
207		mpool_put(t->bt_mp, h, 0);
208		if ((h = mpool_get(t->bt_mp, nextpg, 0)) == NULL)
209			return (1);
210	}
211
212	if (h->pgno == c->pg.pgno)
213		goto ret;
214
215	/* Reacquire the original stack. */
216	mpool_put(t->bt_mp, h, 0);
217	if ((e = __bt_search(t, &c->key, &exact)) == NULL)
218		return (1);
219	h = e->page;
220
221	/*
222	 * Move left, looking for the page.  At each move we have to move
223	 * up the stack until we don't have to change pages to move to the
224	 * next page.  If we have to change pages at an internal level, we
225	 * have to fix the stack back up.
226	 */
227	while (h->pgno != c->pg.pgno) {
228		if ((prevpg = h->prevpg) == P_INVALID)
229			break;
230		mpool_put(t->bt_mp, h, 0);
231
232		/* Move up the stack. */
233		for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
234			/* Get the parent page. */
235			if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
236				return (1);
237
238			/* Move to the next index. */
239			if (parent->index != 0) {
240				idx = parent->index - 1;
241				BT_PUSH(t, h->pgno, idx);
242				break;
243			}
244			mpool_put(t->bt_mp, h, 0);
245		}
246
247		/* Restore the stack. */
248		while (level--) {
249			/* Push the next level down onto the stack. */
250			bi = GETBINTERNAL(h, idx);
251			pgno = bi->pgno;
252
253			/* Lose the currently pinned page. */
254			mpool_put(t->bt_mp, h, 0);
255
256			/* Get the next level down. */
257			if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
258				return (1);
259
260			idx = NEXTINDEX(h) - 1;
261			BT_PUSH(t, pgno, idx);
262		}
263		mpool_put(t->bt_mp, h, 0);
264		if ((h = mpool_get(t->bt_mp, prevpg, 0)) == NULL)
265			return (1);
266	}
267
268
269ret:	mpool_put(t->bt_mp, h, 0);
270	return ((*hp = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL);
271}
272
273/*
274 * __bt_bdelete --
275 *	Delete all key/data pairs matching the specified key.
276 *
277 * Parameters:
278 *	  t:	tree
279 *	key:	key to delete
280 *
281 * Returns:
282 *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key not found.
283 */
284static int
285__bt_bdelete(BTREE *t, const DBT *key)
286{
287	EPG *e;
288	PAGE *h;
289	int deleted, exact, redo;
290
291	deleted = 0;
292
293	/* Find any matching record; __bt_search pins the page. */
294loop:	if ((e = __bt_search(t, key, &exact)) == NULL)
295		return (deleted ? RET_SUCCESS : RET_ERROR);
296	if (!exact) {
297		mpool_put(t->bt_mp, e->page, 0);
298		return (deleted ? RET_SUCCESS : RET_SPECIAL);
299	}
300
301	/*
302	 * Delete forward, then delete backward, from the found key.  If
303	 * there are duplicates and we reach either side of the page, do
304	 * the key search again, so that we get them all.
305	 */
306	redo = 0;
307	h = e->page;
308	do {
309		if (__bt_dleaf(t, key, h, (u_int)e->index)) {
310			mpool_put(t->bt_mp, h, 0);
311			return (RET_ERROR);
312		}
313		if (F_ISSET(t, B_NODUPS)) {
314			if (NEXTINDEX(h) == 0) {
315				if (__bt_pdelete(t, h))
316					return (RET_ERROR);
317			} else
318				mpool_put(t->bt_mp, h, MPOOL_DIRTY);
319			return (RET_SUCCESS);
320		}
321		deleted = 1;
322	} while (e->index < NEXTINDEX(h) && __bt_cmp(t, key, e) == 0);
323
324	/* Check for right-hand edge of the page. */
325	if (e->index == NEXTINDEX(h))
326		redo = 1;
327
328	/* Delete from the key to the beginning of the page. */
329	while (e->index-- > 0) {
330		if (__bt_cmp(t, key, e) != 0)
331			break;
332		if (__bt_dleaf(t, key, h, (u_int)e->index) == RET_ERROR) {
333			mpool_put(t->bt_mp, h, 0);
334			return (RET_ERROR);
335		}
336		if (e->index == 0)
337			redo = 1;
338	}
339
340	/* Check for an empty page. */
341	if (NEXTINDEX(h) == 0) {
342		if (__bt_pdelete(t, h))
343			return (RET_ERROR);
344		goto loop;
345	}
346
347	/* Put the page. */
348	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
349
350	if (redo)
351		goto loop;
352	return (RET_SUCCESS);
353}
354
355/*
356 * __bt_pdelete --
357 *	Delete a single page from the tree.
358 *
359 * Parameters:
360 *	t:	tree
361 *	h:	leaf page
362 *
363 * Returns:
364 *	RET_SUCCESS, RET_ERROR.
365 *
366 * Side-effects:
367 *	mpool_put's the page
368 */
369static int
370__bt_pdelete(BTREE *t, PAGE *h)
371{
372	BINTERNAL *bi;
373	PAGE *pg;
374	EPGNO *parent;
375	indx_t cnt, idx, *ip, offset;
376	uint32_t nksize;
377	char *from;
378
379	/*
380	 * Walk the parent page stack -- a LIFO stack of the pages that were
381	 * traversed when we searched for the page where the delete occurred.
382	 * Each stack entry is a page number and a page index offset.  The
383	 * offset is for the page traversed on the search.  We've just deleted
384	 * a page, so we have to delete the key from the parent page.
385	 *
386	 * If the delete from the parent page makes it empty, this process may
387	 * continue all the way up the tree.  We stop if we reach the root page
388	 * (which is never deleted, it's just not worth the effort) or if the
389	 * delete does not empty the page.
390	 */
391	while ((parent = BT_POP(t)) != NULL) {
392		/* Get the parent page. */
393		if ((pg = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
394			return (RET_ERROR);
395
396		idx = parent->index;
397		bi = GETBINTERNAL(pg, idx);
398
399		/* Free any overflow pages. */
400		if (bi->flags & P_BIGKEY &&
401		    __ovfl_delete(t, bi->bytes) == RET_ERROR) {
402			mpool_put(t->bt_mp, pg, 0);
403			return (RET_ERROR);
404		}
405
406		/*
407		 * Free the parent if it has only the one key and it's not the
408		 * root page. If it's the rootpage, turn it back into an empty
409		 * leaf page.
410		 */
411		if (NEXTINDEX(pg) == 1) {
412			if (pg->pgno == P_ROOT) {
413				pg->lower = BTDATAOFF;
414				pg->upper = t->bt_psize;
415				pg->flags = P_BLEAF;
416			} else {
417				if (__bt_relink(t, pg) || __bt_free(t, pg))
418					return (RET_ERROR);
419				continue;
420			}
421		} else {
422			/* Pack remaining key items at the end of the page. */
423			nksize = NBINTERNAL(bi->ksize);
424			from = (char *)(void *)pg + pg->upper;
425			memmove(from + nksize, from,
426			(size_t)((char *)(void *)bi - from));
427			pg->upper += nksize;
428
429			/* Adjust indices' offsets, shift the indices down. */
430			offset = pg->linp[idx];
431			for (cnt = idx, ip = &pg->linp[0]; cnt--; ++ip)
432				if (ip[0] < offset)
433					ip[0] += nksize;
434			for (cnt = NEXTINDEX(pg) - idx; --cnt; ++ip)
435				ip[0] = ip[1] < offset ? ip[1] + nksize : ip[1];
436			pg->lower -= sizeof(indx_t);
437		}
438
439		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
440		break;
441	}
442
443	/* Free the leaf page, as long as it wasn't the root. */
444	if (h->pgno == P_ROOT) {
445		mpool_put(t->bt_mp, h, MPOOL_DIRTY);
446		return (RET_SUCCESS);
447	}
448	return (__bt_relink(t, h) || __bt_free(t, h));
449}
450
451/*
452 * __bt_dleaf --
453 *	Delete a single record from a leaf page.
454 *
455 * Parameters:
456 *	t:	tree
457 *    key:	referenced key
458 *	h:	page
459 *	idx:	index on page to delete
460 *
461 * Returns:
462 *	RET_SUCCESS, RET_ERROR.
463 */
464int
465__bt_dleaf(BTREE *t, const DBT *key, PAGE *h, u_int idx)
466{
467	BLEAF *bl;
468	indx_t cnt, *ip, offset;
469	uint32_t nbytes;
470	void *to;
471	char *from;
472
473	/* If this record is referenced by the cursor, delete the cursor. */
474	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
475	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
476	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index == idx &&
477	    __bt_curdel(t, key, h, idx))
478		return (RET_ERROR);
479
480	/* If the entry uses overflow pages, make them available for reuse. */
481	to = bl = GETBLEAF(h, idx);
482	if (bl->flags & P_BIGKEY && __ovfl_delete(t, bl->bytes) == RET_ERROR)
483		return (RET_ERROR);
484	if (bl->flags & P_BIGDATA &&
485	    __ovfl_delete(t, bl->bytes + bl->ksize) == RET_ERROR)
486		return (RET_ERROR);
487
488	/* Pack the remaining key/data items at the end of the page. */
489	nbytes = NBLEAF(bl);
490	from = (char *)(void *)h + h->upper;
491	memmove(from + nbytes, from, (size_t)((char *)(void *)to - from));
492	h->upper += nbytes;
493
494	/* Adjust the indices' offsets, shift the indices down. */
495	offset = h->linp[idx];
496	for (cnt = idx, ip = &h->linp[0]; cnt--; ++ip)
497		if (ip[0] < offset)
498			ip[0] += nbytes;
499	for (cnt = NEXTINDEX(h) - idx; --cnt; ++ip)
500		ip[0] = ip[1] < offset ? ip[1] + nbytes : ip[1];
501	h->lower -= sizeof(indx_t);
502
503	/* If the cursor is on this page, adjust it as necessary. */
504	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
505	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
506	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index > idx)
507		--t->bt_cursor.pg.index;
508
509	return (RET_SUCCESS);
510}
511
512/*
513 * __bt_curdel --
514 *	Delete the cursor.
515 *
516 * Parameters:
517 *	t:	tree
518 *    key:	referenced key (or NULL)
519 *	h:	page
520 *  idx:	index on page to delete
521 *
522 * Returns:
523 *	RET_SUCCESS, RET_ERROR.
524 */
525static int
526__bt_curdel(BTREE *t, const DBT *key, PAGE *h, u_int idx)
527{
528	CURSOR *c;
529	EPG e;
530	PAGE *pg;
531	int curcopy, status;
532
533	/*
534	 * If there are duplicates, move forward or backward to one.
535	 * Otherwise, copy the key into the cursor area.
536	 */
537	c = &t->bt_cursor;
538	F_CLR(c, CURS_AFTER | CURS_BEFORE | CURS_ACQUIRE);
539
540	curcopy = 0;
541	if (!F_ISSET(t, B_NODUPS)) {
542		/*
543		 * We're going to have to do comparisons.  If we weren't
544		 * provided a copy of the key, i.e. the user is deleting
545		 * the current cursor position, get one.
546		 */
547		if (key == NULL) {
548			e.page = h;
549			e.index = idx;
550			if ((status = __bt_ret(t, &e,
551			    &c->key, &c->key, NULL, NULL, 1)) != RET_SUCCESS)
552				return (status);
553			curcopy = 1;
554			key = &c->key;
555		}
556		/* Check previous key, if not at the beginning of the page. */
557		if (idx > 0) {
558			e.page = h;
559			e.index = idx - 1;
560			if (__bt_cmp(t, key, &e) == 0) {
561				F_SET(c, CURS_BEFORE);
562				goto dup2;
563			}
564		}
565		/* Check next key, if not at the end of the page. */
566		if (idx < NEXTINDEX(h) - 1) {
567			e.page = h;
568			e.index = idx + 1;
569			if (__bt_cmp(t, key, &e) == 0) {
570				F_SET(c, CURS_AFTER);
571				goto dup2;
572			}
573		}
574		/* Check previous key if at the beginning of the page. */
575		if (idx == 0 && h->prevpg != P_INVALID) {
576			if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
577				return (RET_ERROR);
578			e.page = pg;
579			e.index = NEXTINDEX(pg) - 1;
580			if (__bt_cmp(t, key, &e) == 0) {
581				F_SET(c, CURS_BEFORE);
582				goto dup1;
583			}
584			mpool_put(t->bt_mp, pg, 0);
585		}
586		/* Check next key if at the end of the page. */
587		if (idx == NEXTINDEX(h) - 1 && h->nextpg != P_INVALID) {
588			if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
589				return (RET_ERROR);
590			e.page = pg;
591			e.index = 0;
592			if (__bt_cmp(t, key, &e) == 0) {
593				F_SET(c, CURS_AFTER);
594dup1:				mpool_put(t->bt_mp, pg, 0);
595dup2:				c->pg.pgno = e.page->pgno;
596				c->pg.index = e.index;
597				return (RET_SUCCESS);
598			}
599			mpool_put(t->bt_mp, pg, 0);
600		}
601	}
602	e.page = h;
603	e.index = idx;
604	if (curcopy || (status =
605	    __bt_ret(t, &e, &c->key, &c->key, NULL, NULL, 1)) == RET_SUCCESS) {
606		F_SET(c, CURS_ACQUIRE);
607		return (RET_SUCCESS);
608	}
609	return (status);
610}
611
612/*
613 * __bt_relink --
614 *	Link around a deleted page.
615 *
616 * Parameters:
617 *	t:	tree
618 *	h:	page to be deleted
619 */
620static int
621__bt_relink(BTREE *t, PAGE *h)
622{
623	PAGE *pg;
624
625	if (h->nextpg != P_INVALID) {
626		if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
627			return (RET_ERROR);
628		pg->prevpg = h->prevpg;
629		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
630	}
631	if (h->prevpg != P_INVALID) {
632		if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
633			return (RET_ERROR);
634		pg->nextpg = h->nextpg;
635		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
636	}
637	return (0);
638}
639