xref: /freebsd/lib/libc/db/btree/bt_delete.c (revision a0ee8cc6)
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  * Mike Olson.
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  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #if defined(LIBC_SCCS) && !defined(lint)
34 static char sccsid[] = "@(#)bt_delete.c	8.13 (Berkeley) 7/28/94";
35 #endif /* LIBC_SCCS and not lint */
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 #include <sys/types.h>
40 
41 #include <errno.h>
42 #include <stdio.h>
43 #include <string.h>
44 
45 #include <db.h>
46 #include "btree.h"
47 
48 static int __bt_bdelete(BTREE *, const DBT *);
49 static int __bt_curdel(BTREE *, const DBT *, PAGE *, u_int);
50 static int __bt_pdelete(BTREE *, PAGE *);
51 static int __bt_relink(BTREE *, PAGE *);
52 static int __bt_stkacq(BTREE *, PAGE **, CURSOR *);
53 
54 /*
55  * __bt_delete
56  *	Delete the item(s) referenced by a key.
57  *
58  * Return RET_SPECIAL if the key is not found.
59  */
60 int
61 __bt_delete(const DB *dbp, const DBT *key, u_int flags)
62 {
63 	BTREE *t;
64 	CURSOR *c;
65 	PAGE *h;
66 	int status;
67 
68 	t = dbp->internal;
69 
70 	/* Toss any page pinned across calls. */
71 	if (t->bt_pinned != NULL) {
72 		mpool_put(t->bt_mp, t->bt_pinned, 0);
73 		t->bt_pinned = NULL;
74 	}
75 
76 	/* Check for change to a read-only tree. */
77 	if (F_ISSET(t, B_RDONLY)) {
78 		errno = EPERM;
79 		return (RET_ERROR);
80 	}
81 
82 	switch (flags) {
83 	case 0:
84 		status = __bt_bdelete(t, key);
85 		break;
86 	case R_CURSOR:
87 		/*
88 		 * If flags is R_CURSOR, delete the cursor.  Must already
89 		 * have started a scan and not have already deleted it.
90 		 */
91 		c = &t->bt_cursor;
92 		if (F_ISSET(c, CURS_INIT)) {
93 			if (F_ISSET(c, CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
94 				return (RET_SPECIAL);
95 			if ((h = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL)
96 				return (RET_ERROR);
97 
98 			/*
99 			 * If the page is about to be emptied, we'll need to
100 			 * delete it, which means we have to acquire a stack.
101 			 */
102 			if (NEXTINDEX(h) == 1)
103 				if (__bt_stkacq(t, &h, &t->bt_cursor))
104 					return (RET_ERROR);
105 
106 			status = __bt_dleaf(t, NULL, h, c->pg.index);
107 
108 			if (NEXTINDEX(h) == 0 && status == RET_SUCCESS) {
109 				if (__bt_pdelete(t, h))
110 					return (RET_ERROR);
111 			} else
112 				mpool_put(t->bt_mp,
113 				    h, status == RET_SUCCESS ? MPOOL_DIRTY : 0);
114 			break;
115 		}
116 		/* FALLTHROUGH */
117 	default:
118 		errno = EINVAL;
119 		return (RET_ERROR);
120 	}
121 	if (status == RET_SUCCESS)
122 		F_SET(t, B_MODIFIED);
123 	return (status);
124 }
125 
126 /*
127  * __bt_stkacq --
128  *	Acquire a stack so we can delete a cursor entry.
129  *
130  * Parameters:
131  *	  t:	tree
132  *	 hp:	pointer to current, pinned PAGE pointer
133  *	  c:	pointer to the cursor
134  *
135  * Returns:
136  *	0 on success, 1 on failure
137  */
138 static int
139 __bt_stkacq(BTREE *t, PAGE **hp, CURSOR *c)
140 {
141 	BINTERNAL *bi;
142 	EPG *e;
143 	EPGNO *parent;
144 	PAGE *h;
145 	indx_t idx;
146 	pgno_t pgno;
147 	recno_t nextpg, prevpg;
148 	int exact, level;
149 
150 	/*
151 	 * Find the first occurrence of the key in the tree.  Toss the
152 	 * currently locked page so we don't hit an already-locked page.
153 	 */
154 	h = *hp;
155 	mpool_put(t->bt_mp, h, 0);
156 	if ((e = __bt_search(t, &c->key, &exact)) == NULL)
157 		return (1);
158 	h = e->page;
159 
160 	/* See if we got it in one shot. */
161 	if (h->pgno == c->pg.pgno)
162 		goto ret;
163 
164 	/*
165 	 * Move right, looking for the page.  At each move we have to move
166 	 * up the stack until we don't have to move to the next page.  If
167 	 * we have to change pages at an internal level, we have to fix the
168 	 * stack back up.
169 	 */
170 	while (h->pgno != c->pg.pgno) {
171 		if ((nextpg = h->nextpg) == P_INVALID)
172 			break;
173 		mpool_put(t->bt_mp, h, 0);
174 
175 		/* Move up the stack. */
176 		for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
177 			/* Get the parent page. */
178 			if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
179 				return (1);
180 
181 			/* Move to the next index. */
182 			if (parent->index != NEXTINDEX(h) - 1) {
183 				idx = parent->index + 1;
184 				BT_PUSH(t, h->pgno, idx);
185 				break;
186 			}
187 			mpool_put(t->bt_mp, h, 0);
188 		}
189 
190 		/* Restore the stack. */
191 		while (level--) {
192 			/* Push the next level down onto the stack. */
193 			bi = GETBINTERNAL(h, idx);
194 			pgno = bi->pgno;
195 			BT_PUSH(t, pgno, 0);
196 
197 			/* Lose the currently pinned page. */
198 			mpool_put(t->bt_mp, h, 0);
199 
200 			/* Get the next level down. */
201 			if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
202 				return (1);
203 			idx = 0;
204 		}
205 		mpool_put(t->bt_mp, h, 0);
206 		if ((h = mpool_get(t->bt_mp, nextpg, 0)) == NULL)
207 			return (1);
208 	}
209 
210 	if (h->pgno == c->pg.pgno)
211 		goto ret;
212 
213 	/* Reacquire the original stack. */
214 	mpool_put(t->bt_mp, h, 0);
215 	if ((e = __bt_search(t, &c->key, &exact)) == NULL)
216 		return (1);
217 	h = e->page;
218 
219 	/*
220 	 * Move left, looking for the page.  At each move we have to move
221 	 * up the stack until we don't have to change pages to move to the
222 	 * next page.  If we have to change pages at an internal level, we
223 	 * have to fix the stack back up.
224 	 */
225 	while (h->pgno != c->pg.pgno) {
226 		if ((prevpg = h->prevpg) == P_INVALID)
227 			break;
228 		mpool_put(t->bt_mp, h, 0);
229 
230 		/* Move up the stack. */
231 		for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
232 			/* Get the parent page. */
233 			if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
234 				return (1);
235 
236 			/* Move to the next index. */
237 			if (parent->index != 0) {
238 				idx = parent->index - 1;
239 				BT_PUSH(t, h->pgno, idx);
240 				break;
241 			}
242 			mpool_put(t->bt_mp, h, 0);
243 		}
244 
245 		/* Restore the stack. */
246 		while (level--) {
247 			/* Push the next level down onto the stack. */
248 			bi = GETBINTERNAL(h, idx);
249 			pgno = bi->pgno;
250 
251 			/* Lose the currently pinned page. */
252 			mpool_put(t->bt_mp, h, 0);
253 
254 			/* Get the next level down. */
255 			if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
256 				return (1);
257 
258 			idx = NEXTINDEX(h) - 1;
259 			BT_PUSH(t, pgno, idx);
260 		}
261 		mpool_put(t->bt_mp, h, 0);
262 		if ((h = mpool_get(t->bt_mp, prevpg, 0)) == NULL)
263 			return (1);
264 	}
265 
266 
267 ret:	mpool_put(t->bt_mp, h, 0);
268 	return ((*hp = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL);
269 }
270 
271 /*
272  * __bt_bdelete --
273  *	Delete all key/data pairs matching the specified key.
274  *
275  * Parameters:
276  *	  t:	tree
277  *	key:	key to delete
278  *
279  * Returns:
280  *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key not found.
281  */
282 static int
283 __bt_bdelete(BTREE *t, const DBT *key)
284 {
285 	EPG *e;
286 	PAGE *h;
287 	int deleted, exact, redo;
288 
289 	deleted = 0;
290 
291 	/* Find any matching record; __bt_search pins the page. */
292 loop:	if ((e = __bt_search(t, key, &exact)) == NULL)
293 		return (deleted ? RET_SUCCESS : RET_ERROR);
294 	if (!exact) {
295 		mpool_put(t->bt_mp, e->page, 0);
296 		return (deleted ? RET_SUCCESS : RET_SPECIAL);
297 	}
298 
299 	/*
300 	 * Delete forward, then delete backward, from the found key.  If
301 	 * there are duplicates and we reach either side of the page, do
302 	 * the key search again, so that we get them all.
303 	 */
304 	redo = 0;
305 	h = e->page;
306 	do {
307 		if (__bt_dleaf(t, key, h, e->index)) {
308 			mpool_put(t->bt_mp, h, 0);
309 			return (RET_ERROR);
310 		}
311 		if (F_ISSET(t, B_NODUPS)) {
312 			if (NEXTINDEX(h) == 0) {
313 				if (__bt_pdelete(t, h))
314 					return (RET_ERROR);
315 			} else
316 				mpool_put(t->bt_mp, h, MPOOL_DIRTY);
317 			return (RET_SUCCESS);
318 		}
319 		deleted = 1;
320 	} while (e->index < NEXTINDEX(h) && __bt_cmp(t, key, e) == 0);
321 
322 	/* Check for right-hand edge of the page. */
323 	if (e->index == NEXTINDEX(h))
324 		redo = 1;
325 
326 	/* Delete from the key to the beginning of the page. */
327 	while (e->index-- > 0) {
328 		if (__bt_cmp(t, key, e) != 0)
329 			break;
330 		if (__bt_dleaf(t, key, h, e->index) == RET_ERROR) {
331 			mpool_put(t->bt_mp, h, 0);
332 			return (RET_ERROR);
333 		}
334 		if (e->index == 0)
335 			redo = 1;
336 	}
337 
338 	/* Check for an empty page. */
339 	if (NEXTINDEX(h) == 0) {
340 		if (__bt_pdelete(t, h))
341 			return (RET_ERROR);
342 		goto loop;
343 	}
344 
345 	/* Put the page. */
346 	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
347 
348 	if (redo)
349 		goto loop;
350 	return (RET_SUCCESS);
351 }
352 
353 /*
354  * __bt_pdelete --
355  *	Delete a single page from the tree.
356  *
357  * Parameters:
358  *	t:	tree
359  *	h:	leaf page
360  *
361  * Returns:
362  *	RET_SUCCESS, RET_ERROR.
363  *
364  * Side-effects:
365  *	mpool_put's the page
366  */
367 static int
368 __bt_pdelete(BTREE *t, PAGE *h)
369 {
370 	BINTERNAL *bi;
371 	PAGE *pg;
372 	EPGNO *parent;
373 	indx_t cnt, idx, *ip, offset;
374 	u_int32_t nksize;
375 	char *from;
376 
377 	/*
378 	 * Walk the parent page stack -- a LIFO stack of the pages that were
379 	 * traversed when we searched for the page where the delete occurred.
380 	 * Each stack entry is a page number and a page index offset.  The
381 	 * offset is for the page traversed on the search.  We've just deleted
382 	 * a page, so we have to delete the key from the parent page.
383 	 *
384 	 * If the delete from the parent page makes it empty, this process may
385 	 * continue all the way up the tree.  We stop if we reach the root page
386 	 * (which is never deleted, it's just not worth the effort) or if the
387 	 * delete does not empty the page.
388 	 */
389 	while ((parent = BT_POP(t)) != NULL) {
390 		/* Get the parent page. */
391 		if ((pg = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
392 			return (RET_ERROR);
393 
394 		idx = parent->index;
395 		bi = GETBINTERNAL(pg, idx);
396 
397 		/* Free any overflow pages. */
398 		if (bi->flags & P_BIGKEY &&
399 		    __ovfl_delete(t, bi->bytes) == RET_ERROR) {
400 			mpool_put(t->bt_mp, pg, 0);
401 			return (RET_ERROR);
402 		}
403 
404 		/*
405 		 * Free the parent if it has only the one key and it's not the
406 		 * root page. If it's the rootpage, turn it back into an empty
407 		 * leaf page.
408 		 */
409 		if (NEXTINDEX(pg) == 1) {
410 			if (pg->pgno == P_ROOT) {
411 				pg->lower = BTDATAOFF;
412 				pg->upper = t->bt_psize;
413 				pg->flags = P_BLEAF;
414 			} else {
415 				if (__bt_relink(t, pg) || __bt_free(t, pg))
416 					return (RET_ERROR);
417 				continue;
418 			}
419 		} else {
420 			/* Pack remaining key items at the end of the page. */
421 			nksize = NBINTERNAL(bi->ksize);
422 			from = (char *)pg + pg->upper;
423 			memmove(from + nksize, from, (char *)bi - from);
424 			pg->upper += nksize;
425 
426 			/* Adjust indices' offsets, shift the indices down. */
427 			offset = pg->linp[idx];
428 			for (cnt = idx, ip = &pg->linp[0]; cnt--; ++ip)
429 				if (ip[0] < offset)
430 					ip[0] += nksize;
431 			for (cnt = NEXTINDEX(pg) - idx; --cnt; ++ip)
432 				ip[0] = ip[1] < offset ? ip[1] + nksize : ip[1];
433 			pg->lower -= sizeof(indx_t);
434 		}
435 
436 		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
437 		break;
438 	}
439 
440 	/* Free the leaf page, as long as it wasn't the root. */
441 	if (h->pgno == P_ROOT) {
442 		mpool_put(t->bt_mp, h, MPOOL_DIRTY);
443 		return (RET_SUCCESS);
444 	}
445 	return (__bt_relink(t, h) || __bt_free(t, h));
446 }
447 
448 /*
449  * __bt_dleaf --
450  *	Delete a single record from a leaf page.
451  *
452  * Parameters:
453  *	t:	tree
454  *    key:	referenced key
455  *	h:	page
456  *	idx:	index on page to delete
457  *
458  * Returns:
459  *	RET_SUCCESS, RET_ERROR.
460  */
461 int
462 __bt_dleaf(BTREE *t, const DBT *key, PAGE *h, u_int idx)
463 {
464 	BLEAF *bl;
465 	indx_t cnt, *ip, offset;
466 	u_int32_t nbytes;
467 	void *to;
468 	char *from;
469 
470 	/* If this record is referenced by the cursor, delete the cursor. */
471 	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
472 	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
473 	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index == idx &&
474 	    __bt_curdel(t, key, h, idx))
475 		return (RET_ERROR);
476 
477 	/* If the entry uses overflow pages, make them available for reuse. */
478 	to = bl = GETBLEAF(h, idx);
479 	if (bl->flags & P_BIGKEY && __ovfl_delete(t, bl->bytes) == RET_ERROR)
480 		return (RET_ERROR);
481 	if (bl->flags & P_BIGDATA &&
482 	    __ovfl_delete(t, bl->bytes + bl->ksize) == RET_ERROR)
483 		return (RET_ERROR);
484 
485 	/* Pack the remaining key/data items at the end of the page. */
486 	nbytes = NBLEAF(bl);
487 	from = (char *)h + h->upper;
488 	memmove(from + nbytes, from, (char *)to - from);
489 	h->upper += nbytes;
490 
491 	/* Adjust the indices' offsets, shift the indices down. */
492 	offset = h->linp[idx];
493 	for (cnt = idx, ip = &h->linp[0]; cnt--; ++ip)
494 		if (ip[0] < offset)
495 			ip[0] += nbytes;
496 	for (cnt = NEXTINDEX(h) - idx; --cnt; ++ip)
497 		ip[0] = ip[1] < offset ? ip[1] + nbytes : ip[1];
498 	h->lower -= sizeof(indx_t);
499 
500 	/* If the cursor is on this page, adjust it as necessary. */
501 	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
502 	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
503 	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index > idx)
504 		--t->bt_cursor.pg.index;
505 
506 	return (RET_SUCCESS);
507 }
508 
509 /*
510  * __bt_curdel --
511  *	Delete the cursor.
512  *
513  * Parameters:
514  *	t:	tree
515  *    key:	referenced key (or NULL)
516  *	h:	page
517  *    idx:	index on page to delete
518  *
519  * Returns:
520  *	RET_SUCCESS, RET_ERROR.
521  */
522 static int
523 __bt_curdel(BTREE *t, const DBT *key, PAGE *h, u_int idx)
524 {
525 	CURSOR *c;
526 	EPG e;
527 	PAGE *pg;
528 	int curcopy, status;
529 
530 	/*
531 	 * If there are duplicates, move forward or backward to one.
532 	 * Otherwise, copy the key into the cursor area.
533 	 */
534 	c = &t->bt_cursor;
535 	F_CLR(c, CURS_AFTER | CURS_BEFORE | CURS_ACQUIRE);
536 
537 	curcopy = 0;
538 	if (!F_ISSET(t, B_NODUPS)) {
539 		/*
540 		 * We're going to have to do comparisons.  If we weren't
541 		 * provided a copy of the key, i.e. the user is deleting
542 		 * the current cursor position, get one.
543 		 */
544 		if (key == NULL) {
545 			e.page = h;
546 			e.index = idx;
547 			if ((status = __bt_ret(t, &e,
548 			    &c->key, &c->key, NULL, NULL, 1)) != RET_SUCCESS)
549 				return (status);
550 			curcopy = 1;
551 			key = &c->key;
552 		}
553 		/* Check previous key, if not at the beginning of the page. */
554 		if (idx > 0) {
555 			e.page = h;
556 			e.index = idx - 1;
557 			if (__bt_cmp(t, key, &e) == 0) {
558 				F_SET(c, CURS_BEFORE);
559 				goto dup2;
560 			}
561 		}
562 		/* Check next key, if not at the end of the page. */
563 		if (idx < NEXTINDEX(h) - 1) {
564 			e.page = h;
565 			e.index = idx + 1;
566 			if (__bt_cmp(t, key, &e) == 0) {
567 				F_SET(c, CURS_AFTER);
568 				goto dup2;
569 			}
570 		}
571 		/* Check previous key if at the beginning of the page. */
572 		if (idx == 0 && h->prevpg != P_INVALID) {
573 			if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
574 				return (RET_ERROR);
575 			e.page = pg;
576 			e.index = NEXTINDEX(pg) - 1;
577 			if (__bt_cmp(t, key, &e) == 0) {
578 				F_SET(c, CURS_BEFORE);
579 				goto dup1;
580 			}
581 			mpool_put(t->bt_mp, pg, 0);
582 		}
583 		/* Check next key if at the end of the page. */
584 		if (idx == NEXTINDEX(h) - 1 && h->nextpg != P_INVALID) {
585 			if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
586 				return (RET_ERROR);
587 			e.page = pg;
588 			e.index = 0;
589 			if (__bt_cmp(t, key, &e) == 0) {
590 				F_SET(c, CURS_AFTER);
591 dup1:				mpool_put(t->bt_mp, pg, 0);
592 dup2:				c->pg.pgno = e.page->pgno;
593 				c->pg.index = e.index;
594 				return (RET_SUCCESS);
595 			}
596 			mpool_put(t->bt_mp, pg, 0);
597 		}
598 	}
599 	e.page = h;
600 	e.index = idx;
601 	if (curcopy || (status =
602 	    __bt_ret(t, &e, &c->key, &c->key, NULL, NULL, 1)) == RET_SUCCESS) {
603 		F_SET(c, CURS_ACQUIRE);
604 		return (RET_SUCCESS);
605 	}
606 	return (status);
607 }
608 
609 /*
610  * __bt_relink --
611  *	Link around a deleted page.
612  *
613  * Parameters:
614  *	t:	tree
615  *	h:	page to be deleted
616  */
617 static int
618 __bt_relink(BTREE *t, PAGE *h)
619 {
620 	PAGE *pg;
621 
622 	if (h->nextpg != P_INVALID) {
623 		if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
624 			return (RET_ERROR);
625 		pg->prevpg = h->prevpg;
626 		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
627 	}
628 	if (h->prevpg != P_INVALID) {
629 		if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
630 			return (RET_ERROR);
631 		pg->nextpg = h->nextpg;
632 		mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
633 	}
634 	return (0);
635 }
636