1 /* $OpenBSD: bt_delete.c,v 1.11 2005/08/05 13:02:59 espie 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/types.h>
36
37 #include <errno.h>
38 #include <stdio.h>
39 #include <string.h>
40
41 #include <db.h>
42 #include "btree.h"
43
44 static int __bt_bdelete(BTREE *, const DBT *);
45 static int __bt_curdel(BTREE *, const DBT *, PAGE *, u_int);
46 static int __bt_pdelete(BTREE *, PAGE *);
47 static int __bt_relink(BTREE *, PAGE *);
48 static int __bt_stkacq(BTREE *, PAGE **, CURSOR *);
49
50 /*
51 * __bt_delete
52 * Delete the item(s) referenced by a key.
53 *
54 * Return RET_SPECIAL if the key is not found.
55 */
56 int
__bt_delete(const DB * dbp,const DBT * key,u_int flags)57 __bt_delete(const DB *dbp, const DBT *key, u_int flags)
58 {
59 BTREE *t;
60 CURSOR *c;
61 PAGE *h;
62 int status;
63
64 t = dbp->internal;
65
66 /* Toss any page pinned across calls. */
67 if (t->bt_pinned != NULL) {
68 mpool_put(t->bt_mp, t->bt_pinned, 0);
69 t->bt_pinned = NULL;
70 }
71
72 /* Check for change to a read-only tree. */
73 if (F_ISSET(t, B_RDONLY)) {
74 errno = EPERM;
75 return (RET_ERROR);
76 }
77
78 switch (flags) {
79 case 0:
80 status = __bt_bdelete(t, key);
81 break;
82 case R_CURSOR:
83 /*
84 * If flags is R_CURSOR, delete the cursor. Must already
85 * have started a scan and not have already deleted it.
86 */
87 c = &t->bt_cursor;
88 if (F_ISSET(c, CURS_INIT)) {
89 if (F_ISSET(c, CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
90 return (RET_SPECIAL);
91 if ((h = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL)
92 return (RET_ERROR);
93
94 /*
95 * If the page is about to be emptied, we'll need to
96 * delete it, which means we have to acquire a stack.
97 */
98 if (NEXTINDEX(h) == 1)
99 if (__bt_stkacq(t, &h, &t->bt_cursor))
100 return (RET_ERROR);
101
102 status = __bt_dleaf(t, NULL, h, c->pg.index);
103
104 if (NEXTINDEX(h) == 0 && status == RET_SUCCESS) {
105 if (__bt_pdelete(t, h))
106 return (RET_ERROR);
107 } else
108 mpool_put(t->bt_mp,
109 h, status == RET_SUCCESS ? MPOOL_DIRTY : 0);
110 break;
111 }
112 /* FALLTHROUGH */
113 default:
114 errno = EINVAL;
115 return (RET_ERROR);
116 }
117 if (status == RET_SUCCESS)
118 F_SET(t, B_MODIFIED);
119 return (status);
120 }
121
122 /*
123 * __bt_stkacq --
124 * Acquire a stack so we can delete a cursor entry.
125 *
126 * Parameters:
127 * t: tree
128 * hp: pointer to current, pinned PAGE pointer
129 * c: pointer to the cursor
130 *
131 * Returns:
132 * 0 on success, 1 on failure
133 */
134 static int
__bt_stkacq(BTREE * t,PAGE ** hp,CURSOR * c)135 __bt_stkacq(BTREE *t, PAGE **hp, CURSOR *c)
136 {
137 BINTERNAL *bi;
138 EPG *e;
139 EPGNO *parent;
140 PAGE *h;
141 indx_t idx;
142 pgno_t pgno;
143 recno_t nextpg, prevpg;
144 int exact, level;
145
146 /*
147 * Find the first occurrence of the key in the tree. Toss the
148 * currently locked page so we don't hit an already-locked page.
149 */
150 h = *hp;
151 mpool_put(t->bt_mp, h, 0);
152 if ((e = __bt_search(t, &c->key, &exact)) == NULL)
153 return (1);
154 h = e->page;
155
156 /* See if we got it in one shot. */
157 if (h->pgno == c->pg.pgno)
158 goto ret;
159
160 /*
161 * Move right, looking for the page. At each move we have to move
162 * up the stack until we don't have to move to the next page. If
163 * we have to change pages at an internal level, we have to fix the
164 * stack back up.
165 */
166 while (h->pgno != c->pg.pgno) {
167 if ((nextpg = h->nextpg) == P_INVALID)
168 break;
169 mpool_put(t->bt_mp, h, 0);
170
171 /* Move up the stack. */
172 for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
173 /* Get the parent page. */
174 if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
175 return (1);
176
177 /* Move to the next index. */
178 if (parent->index != NEXTINDEX(h) - 1) {
179 idx = parent->index + 1;
180 BT_PUSH(t, h->pgno, idx);
181 break;
182 }
183 mpool_put(t->bt_mp, h, 0);
184 }
185
186 /* Restore the stack. */
187 while (level--) {
188 /* Push the next level down onto the stack. */
189 bi = GETBINTERNAL(h, idx);
190 pgno = bi->pgno;
191 BT_PUSH(t, pgno, 0);
192
193 /* Lose the currently pinned page. */
194 mpool_put(t->bt_mp, h, 0);
195
196 /* Get the next level down. */
197 if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
198 return (1);
199 idx = 0;
200 }
201 mpool_put(t->bt_mp, h, 0);
202 if ((h = mpool_get(t->bt_mp, nextpg, 0)) == NULL)
203 return (1);
204 }
205
206 if (h->pgno == c->pg.pgno)
207 goto ret;
208
209 /* Reacquire the original stack. */
210 mpool_put(t->bt_mp, h, 0);
211 if ((e = __bt_search(t, &c->key, &exact)) == NULL)
212 return (1);
213 h = e->page;
214
215 /*
216 * Move left, looking for the page. At each move we have to move
217 * up the stack until we don't have to change pages to move to the
218 * next page. If we have to change pages at an internal level, we
219 * have to fix the stack back up.
220 */
221 while (h->pgno != c->pg.pgno) {
222 if ((prevpg = h->prevpg) == P_INVALID)
223 break;
224 mpool_put(t->bt_mp, h, 0);
225
226 /* Move up the stack. */
227 for (level = 0; (parent = BT_POP(t)) != NULL; ++level) {
228 /* Get the parent page. */
229 if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
230 return (1);
231
232 /* Move to the next index. */
233 if (parent->index != 0) {
234 idx = parent->index - 1;
235 BT_PUSH(t, h->pgno, idx);
236 break;
237 }
238 mpool_put(t->bt_mp, h, 0);
239 }
240
241 /* Restore the stack. */
242 while (level--) {
243 /* Push the next level down onto the stack. */
244 bi = GETBINTERNAL(h, idx);
245 pgno = bi->pgno;
246
247 /* Lose the currently pinned page. */
248 mpool_put(t->bt_mp, h, 0);
249
250 /* Get the next level down. */
251 if ((h = mpool_get(t->bt_mp, pgno, 0)) == NULL)
252 return (1);
253
254 idx = NEXTINDEX(h) - 1;
255 BT_PUSH(t, pgno, idx);
256 }
257 mpool_put(t->bt_mp, h, 0);
258 if ((h = mpool_get(t->bt_mp, prevpg, 0)) == NULL)
259 return (1);
260 }
261
262
263 ret: mpool_put(t->bt_mp, h, 0);
264 return ((*hp = mpool_get(t->bt_mp, c->pg.pgno, 0)) == NULL);
265 }
266
267 /*
268 * __bt_bdelete --
269 * Delete all key/data pairs matching the specified key.
270 *
271 * Parameters:
272 * t: tree
273 * key: key to delete
274 *
275 * Returns:
276 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key not found.
277 */
278 static int
__bt_bdelete(BTREE * t,const DBT * key)279 __bt_bdelete(BTREE *t, const DBT *key)
280 {
281 EPG *e;
282 PAGE *h;
283 int deleted, exact, redo;
284
285 deleted = 0;
286
287 /* Find any matching record; __bt_search pins the page. */
288 loop: if ((e = __bt_search(t, key, &exact)) == NULL)
289 return (deleted ? RET_SUCCESS : RET_ERROR);
290 if (!exact) {
291 mpool_put(t->bt_mp, e->page, 0);
292 return (deleted ? RET_SUCCESS : RET_SPECIAL);
293 }
294
295 /*
296 * Delete forward, then delete backward, from the found key. If
297 * there are duplicates and we reach either side of the page, do
298 * the key search again, so that we get them all.
299 */
300 redo = 0;
301 h = e->page;
302 do {
303 if (__bt_dleaf(t, key, h, e->index)) {
304 mpool_put(t->bt_mp, h, 0);
305 return (RET_ERROR);
306 }
307 if (F_ISSET(t, B_NODUPS)) {
308 if (NEXTINDEX(h) == 0) {
309 if (__bt_pdelete(t, h))
310 return (RET_ERROR);
311 } else
312 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
313 return (RET_SUCCESS);
314 }
315 deleted = 1;
316 } while (e->index < NEXTINDEX(h) && __bt_cmp(t, key, e) == 0);
317
318 /* Check for right-hand edge of the page. */
319 if (e->index == NEXTINDEX(h))
320 redo = 1;
321
322 /* Delete from the key to the beginning of the page. */
323 while (e->index-- > 0) {
324 if (__bt_cmp(t, key, e) != 0)
325 break;
326 if (__bt_dleaf(t, key, h, e->index) == RET_ERROR) {
327 mpool_put(t->bt_mp, h, 0);
328 return (RET_ERROR);
329 }
330 if (e->index == 0)
331 redo = 1;
332 }
333
334 /* Check for an empty page. */
335 if (NEXTINDEX(h) == 0) {
336 if (__bt_pdelete(t, h))
337 return (RET_ERROR);
338 goto loop;
339 }
340
341 /* Put the page. */
342 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
343
344 if (redo)
345 goto loop;
346 return (RET_SUCCESS);
347 }
348
349 /*
350 * __bt_pdelete --
351 * Delete a single page from the tree.
352 *
353 * Parameters:
354 * t: tree
355 * h: leaf page
356 *
357 * Returns:
358 * RET_SUCCESS, RET_ERROR.
359 *
360 * Side-effects:
361 * mpool_put's the page
362 */
363 static int
__bt_pdelete(BTREE * t,PAGE * h)364 __bt_pdelete(BTREE *t, PAGE *h)
365 {
366 BINTERNAL *bi;
367 PAGE *pg;
368 EPGNO *parent;
369 indx_t cnt, idx, *ip, offset;
370 u_int32_t nksize;
371 char *from;
372
373 /*
374 * Walk the parent page stack -- a LIFO stack of the pages that were
375 * traversed when we searched for the page where the delete occurred.
376 * Each stack entry is a page number and a page index offset. The
377 * offset is for the page traversed on the search. We've just deleted
378 * a page, so we have to delete the key from the parent page.
379 *
380 * If the delete from the parent page makes it empty, this process may
381 * continue all the way up the tree. We stop if we reach the root page
382 * (which is never deleted, it's just not worth the effort) or if the
383 * delete does not empty the page.
384 */
385 while ((parent = BT_POP(t)) != NULL) {
386 /* Get the parent page. */
387 if ((pg = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
388 return (RET_ERROR);
389
390 idx = parent->index;
391 bi = GETBINTERNAL(pg, idx);
392
393 /* Free any overflow pages. */
394 if (bi->flags & P_BIGKEY &&
395 __ovfl_delete(t, bi->bytes) == RET_ERROR) {
396 mpool_put(t->bt_mp, pg, 0);
397 return (RET_ERROR);
398 }
399
400 /*
401 * Free the parent if it has only the one key and it's not the
402 * root page. If it's the rootpage, turn it back into an empty
403 * leaf page.
404 */
405 if (NEXTINDEX(pg) == 1) {
406 if (pg->pgno == P_ROOT) {
407 pg->lower = BTDATAOFF;
408 pg->upper = t->bt_psize;
409 pg->flags = P_BLEAF;
410 } else {
411 if (__bt_relink(t, pg) || __bt_free(t, pg))
412 return (RET_ERROR);
413 continue;
414 }
415 } else {
416 /* Pack remaining key items at the end of the page. */
417 nksize = NBINTERNAL(bi->ksize);
418 from = (char *)pg + pg->upper;
419 memmove(from + nksize, from, (char *)bi - from);
420 pg->upper += nksize;
421
422 /* Adjust indices' offsets, shift the indices down. */
423 offset = pg->linp[idx];
424 for (cnt = idx, ip = &pg->linp[0]; cnt--; ++ip)
425 if (ip[0] < offset)
426 ip[0] += nksize;
427 for (cnt = NEXTINDEX(pg) - idx; --cnt; ++ip)
428 ip[0] = ip[1] < offset ? ip[1] + nksize : ip[1];
429 pg->lower -= sizeof(indx_t);
430 }
431
432 mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
433 break;
434 }
435
436 /* Free the leaf page, as long as it wasn't the root. */
437 if (h->pgno == P_ROOT) {
438 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
439 return (RET_SUCCESS);
440 }
441 return (__bt_relink(t, h) || __bt_free(t, h));
442 }
443
444 /*
445 * __bt_dleaf --
446 * Delete a single record from a leaf page.
447 *
448 * Parameters:
449 * t: tree
450 * key: referenced key
451 * h: page
452 * idx: index on page to delete
453 *
454 * Returns:
455 * RET_SUCCESS, RET_ERROR.
456 */
457 int
__bt_dleaf(BTREE * t,const DBT * key,PAGE * h,u_int idx)458 __bt_dleaf(BTREE *t, const DBT *key, PAGE *h, u_int idx)
459 {
460 BLEAF *bl;
461 indx_t cnt, *ip, offset;
462 u_int32_t nbytes;
463 void *to;
464 char *from;
465
466 /* If this record is referenced by the cursor, delete the cursor. */
467 if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
468 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
469 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index == idx &&
470 __bt_curdel(t, key, h, idx))
471 return (RET_ERROR);
472
473 /* If the entry uses overflow pages, make them available for reuse. */
474 to = bl = GETBLEAF(h, idx);
475 if (bl->flags & P_BIGKEY && __ovfl_delete(t, bl->bytes) == RET_ERROR)
476 return (RET_ERROR);
477 if (bl->flags & P_BIGDATA &&
478 __ovfl_delete(t, bl->bytes + bl->ksize) == RET_ERROR)
479 return (RET_ERROR);
480
481 /* Pack the remaining key/data items at the end of the page. */
482 nbytes = NBLEAF(bl);
483 from = (char *)h + h->upper;
484 memmove(from + nbytes, from, (char *)to - from);
485 h->upper += nbytes;
486
487 /* Adjust the indices' offsets, shift the indices down. */
488 offset = h->linp[idx];
489 for (cnt = idx, ip = &h->linp[0]; cnt--; ++ip)
490 if (ip[0] < offset)
491 ip[0] += nbytes;
492 for (cnt = NEXTINDEX(h) - idx; --cnt; ++ip)
493 ip[0] = ip[1] < offset ? ip[1] + nbytes : ip[1];
494 h->lower -= sizeof(indx_t);
495
496 /* If the cursor is on this page, adjust it as necessary. */
497 if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
498 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
499 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index > idx)
500 --t->bt_cursor.pg.index;
501
502 return (RET_SUCCESS);
503 }
504
505 /*
506 * __bt_curdel --
507 * Delete the cursor.
508 *
509 * Parameters:
510 * t: tree
511 * key: referenced key (or NULL)
512 * h: page
513 * idx: index on page to delete
514 *
515 * Returns:
516 * RET_SUCCESS, RET_ERROR.
517 */
518 static int
__bt_curdel(BTREE * t,const DBT * key,PAGE * h,u_int idx)519 __bt_curdel(BTREE *t, const DBT *key, PAGE *h, u_int idx)
520 {
521 CURSOR *c;
522 EPG e;
523 PAGE *pg;
524 int curcopy, status;
525
526 /*
527 * If there are duplicates, move forward or backward to one.
528 * Otherwise, copy the key into the cursor area.
529 */
530 c = &t->bt_cursor;
531 F_CLR(c, CURS_AFTER | CURS_BEFORE | CURS_ACQUIRE);
532
533 curcopy = 0;
534 if (!F_ISSET(t, B_NODUPS)) {
535 /*
536 * We're going to have to do comparisons. If we weren't
537 * provided a copy of the key, i.e. the user is deleting
538 * the current cursor position, get one.
539 */
540 if (key == NULL) {
541 e.page = h;
542 e.index = idx;
543 if ((status = __bt_ret(t, &e,
544 &c->key, &c->key, NULL, NULL, 1)) != RET_SUCCESS)
545 return (status);
546 curcopy = 1;
547 key = &c->key;
548 }
549 /* Check previous key, if not at the beginning of the page. */
550 if (idx > 0) {
551 e.page = h;
552 e.index = idx - 1;
553 if (__bt_cmp(t, key, &e) == 0) {
554 F_SET(c, CURS_BEFORE);
555 goto dup2;
556 }
557 }
558 /* Check next key, if not at the end of the page. */
559 if (idx < NEXTINDEX(h) - 1) {
560 e.page = h;
561 e.index = idx + 1;
562 if (__bt_cmp(t, key, &e) == 0) {
563 F_SET(c, CURS_AFTER);
564 goto dup2;
565 }
566 }
567 /* Check previous key if at the beginning of the page. */
568 if (idx == 0 && h->prevpg != P_INVALID) {
569 if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
570 return (RET_ERROR);
571 e.page = pg;
572 e.index = NEXTINDEX(pg) - 1;
573 if (__bt_cmp(t, key, &e) == 0) {
574 F_SET(c, CURS_BEFORE);
575 goto dup1;
576 }
577 mpool_put(t->bt_mp, pg, 0);
578 }
579 /* Check next key if at the end of the page. */
580 if (idx == NEXTINDEX(h) - 1 && h->nextpg != P_INVALID) {
581 if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
582 return (RET_ERROR);
583 e.page = pg;
584 e.index = 0;
585 if (__bt_cmp(t, key, &e) == 0) {
586 F_SET(c, CURS_AFTER);
587 dup1: mpool_put(t->bt_mp, pg, 0);
588 dup2: c->pg.pgno = e.page->pgno;
589 c->pg.index = e.index;
590 return (RET_SUCCESS);
591 }
592 mpool_put(t->bt_mp, pg, 0);
593 }
594 }
595 e.page = h;
596 e.index = idx;
597 if (curcopy || (status =
598 __bt_ret(t, &e, &c->key, &c->key, NULL, NULL, 1)) == RET_SUCCESS) {
599 F_SET(c, CURS_ACQUIRE);
600 return (RET_SUCCESS);
601 }
602 return (status);
603 }
604
605 /*
606 * __bt_relink --
607 * Link around a deleted page.
608 *
609 * Parameters:
610 * t: tree
611 * h: page to be deleted
612 */
613 static int
__bt_relink(BTREE * t,PAGE * h)614 __bt_relink(BTREE *t, PAGE *h)
615 {
616 PAGE *pg;
617
618 if (h->nextpg != P_INVALID) {
619 if ((pg = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL)
620 return (RET_ERROR);
621 pg->prevpg = h->prevpg;
622 mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
623 }
624 if (h->prevpg != P_INVALID) {
625 if ((pg = mpool_get(t->bt_mp, h->prevpg, 0)) == NULL)
626 return (RET_ERROR);
627 pg->nextpg = h->nextpg;
628 mpool_put(t->bt_mp, pg, MPOOL_DIRTY);
629 }
630 return (0);
631 }
632