1 /* $NetBSD: bt_split.c,v 1.22 2016/09/24 21:31:25 christos 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 #if HAVE_NBTOOL_CONFIG_H
36 #include "nbtool_config.h"
37 #endif
38
39 #include <sys/cdefs.h>
40 __RCSID("$NetBSD: bt_split.c,v 1.22 2016/09/24 21:31:25 christos Exp $");
41
42 #include "namespace.h"
43 #include <sys/types.h>
44
45 #include <assert.h>
46 #include <limits.h>
47 #include <stdio.h>
48 #include <stdlib.h>
49 #include <string.h>
50
51 #include <db.h>
52 #include "btree.h"
53
54 static int bt_broot(BTREE *, PAGE *, PAGE *, PAGE *);
55 static PAGE *bt_page(BTREE *, PAGE *, PAGE **, PAGE **, indx_t *, size_t);
56 static int bt_preserve(BTREE *, pgno_t);
57 static PAGE *bt_psplit(BTREE *, PAGE *, PAGE *, PAGE *, indx_t *, size_t);
58 static PAGE *bt_root(BTREE *, PAGE *, PAGE **, PAGE **, indx_t *, size_t);
59 static int bt_rroot(BTREE *, PAGE *, PAGE *, PAGE *);
60 static recno_t rec_total(PAGE *);
61
62 #ifdef STATISTICS
63 unsigned long bt_rootsplit, bt_split, bt_sortsplit, bt_pfxsaved;
64 #endif
65
66 /*
67 * __BT_SPLIT -- Split the tree.
68 *
69 * Parameters:
70 * t: tree
71 * sp: page to split
72 * key: key to insert
73 * data: data to insert
74 * flags: BIGKEY/BIGDATA flags
75 * ilen: insert length
76 * skip: index to leave open
77 *
78 * Returns:
79 * RET_ERROR, RET_SUCCESS
80 */
81 int
__bt_split(BTREE * t,PAGE * sp,const DBT * key,const DBT * data,int flags,size_t ilen,uint32_t argskip)82 __bt_split(BTREE *t, PAGE *sp, const DBT *key, const DBT *data, int flags,
83 size_t ilen, uint32_t argskip)
84 {
85 BINTERNAL *bi = NULL; /* pacify gcc */
86 BLEAF *bl = NULL, *tbl; /* pacify gcc */
87 DBT a, b;
88 EPGNO *parent;
89 PAGE *h, *l, *r, *lchild, *rchild;
90 indx_t nxtindex;
91 uint16_t skip;
92 uint32_t n, nbytes, nksize = 0; /* pacify gcc */
93 int parentsplit;
94 char *dest;
95
96 /*
97 * Split the page into two pages, l and r. The split routines return
98 * a pointer to the page into which the key should be inserted and with
99 * skip set to the offset which should be used. Additionally, l and r
100 * are pinned.
101 */
102 skip = argskip;
103 h = sp->pgno == P_ROOT ?
104 bt_root(t, sp, &l, &r, &skip, ilen) :
105 bt_page(t, sp, &l, &r, &skip, ilen);
106 if (h == NULL)
107 return (RET_ERROR);
108
109 /*
110 * Insert the new key/data pair into the leaf page. (Key inserts
111 * always cause a leaf page to split first.)
112 */
113 _DBFIT(ilen, indx_t);
114 h->upper -= (indx_t)ilen;
115 h->linp[skip] = h->upper;
116 dest = (char *)(void *)h + h->upper;
117 if (F_ISSET(t, R_RECNO))
118 WR_RLEAF(dest, data, flags);
119 else
120 WR_BLEAF(dest, key, data, flags);
121
122 /* If the root page was split, make it look right. */
123 if (sp->pgno == P_ROOT &&
124 (F_ISSET(t, R_RECNO) ?
125 bt_rroot(t, sp, l, r) : bt_broot(t, sp, l, r)) == RET_ERROR)
126 goto err2;
127
128 /*
129 * Now we walk the parent page stack -- a LIFO stack of the pages that
130 * were traversed when we searched for the page that split. Each stack
131 * entry is a page number and a page index offset. The offset is for
132 * the page traversed on the search. We've just split a page, so we
133 * have to insert a new key into the parent page.
134 *
135 * If the insert into the parent page causes it to split, may have to
136 * continue splitting all the way up the tree. We stop if the root
137 * splits or the page inserted into didn't have to split to hold the
138 * new key. Some algorithms replace the key for the old page as well
139 * as the new page. We don't, as there's no reason to believe that the
140 * first key on the old page is any better than the key we have, and,
141 * in the case of a key being placed at index 0 causing the split, the
142 * key is unavailable.
143 *
144 * There are a maximum of 5 pages pinned at any time. We keep the left
145 * and right pages pinned while working on the parent. The 5 are the
146 * two children, left parent and right parent (when the parent splits)
147 * and the root page or the overflow key page when calling bt_preserve.
148 * This code must make sure that all pins are released other than the
149 * root page or overflow page which is unlocked elsewhere.
150 */
151 while ((parent = BT_POP(t)) != NULL) {
152 lchild = l;
153 rchild = r;
154
155 /* Get the parent page. */
156 if ((h = mpool_get(t->bt_mp, parent->pgno, 0)) == NULL)
157 goto err2;
158
159 /*
160 * The new key goes ONE AFTER the index, because the split
161 * was to the right.
162 */
163 skip = parent->index + 1;
164
165 /*
166 * Calculate the space needed on the parent page.
167 *
168 * Prefix trees: space hack when inserting into BINTERNAL
169 * pages. Retain only what's needed to distinguish between
170 * the new entry and the LAST entry on the page to its left.
171 * If the keys compare equal, retain the entire key. Note,
172 * we don't touch overflow keys, and the entire key must be
173 * retained for the next-to-left most key on the leftmost
174 * page of each level, or the search will fail. Applicable
175 * ONLY to internal pages that have leaf pages as children.
176 * Further reduction of the key between pairs of internal
177 * pages loses too much information.
178 */
179 switch (rchild->flags & P_TYPE) {
180 case P_BINTERNAL:
181 bi = GETBINTERNAL(rchild, 0);
182 nbytes = NBINTERNAL(bi->ksize);
183 break;
184 case P_BLEAF:
185 bl = GETBLEAF(rchild, 0);
186 nbytes = NBINTERNAL(bl->ksize);
187 if (t->bt_pfx && !(bl->flags & P_BIGKEY) &&
188 (h->prevpg != P_INVALID || skip > 1)) {
189 size_t temp;
190 tbl = GETBLEAF(lchild, NEXTINDEX(lchild) - 1);
191 a.size = tbl->ksize;
192 a.data = tbl->bytes;
193 b.size = bl->ksize;
194 b.data = bl->bytes;
195 temp = t->bt_pfx(&a, &b);
196 _DBFIT(temp, uint32_t);
197 nksize = (uint32_t)temp;
198 n = NBINTERNAL(nksize);
199 if (n < nbytes) {
200 #ifdef STATISTICS
201 bt_pfxsaved += nbytes - n;
202 #endif
203 nbytes = n;
204 } else
205 nksize = 0;
206 } else
207 nksize = 0;
208 break;
209 case P_RINTERNAL:
210 case P_RLEAF:
211 nbytes = NRINTERNAL;
212 break;
213 default:
214 abort();
215 }
216
217 /* Split the parent page if necessary or shift the indices. */
218 if ((uint32_t)h->upper - (uint32_t)h->lower < nbytes + sizeof(indx_t)) {
219 sp = h;
220 h = h->pgno == P_ROOT ?
221 bt_root(t, h, &l, &r, &skip, nbytes) :
222 bt_page(t, h, &l, &r, &skip, nbytes);
223 if (h == NULL)
224 goto err1;
225 parentsplit = 1;
226 } else {
227 if (skip < (nxtindex = NEXTINDEX(h)))
228 memmove(h->linp + skip + 1, h->linp + skip,
229 (nxtindex - skip) * sizeof(indx_t));
230 h->lower += sizeof(indx_t);
231 parentsplit = 0;
232 }
233
234 /* Insert the key into the parent page. */
235 switch (rchild->flags & P_TYPE) {
236 case P_BINTERNAL:
237 h->linp[skip] = h->upper -= nbytes;
238 dest = (char *)(void *)h + h->linp[skip];
239 memmove(dest, bi, nbytes);
240 ((BINTERNAL *)(void *)dest)->pgno = rchild->pgno;
241 break;
242 case P_BLEAF:
243 h->linp[skip] = h->upper -= nbytes;
244 dest = (char *)(void *)h + h->linp[skip];
245 WR_BINTERNAL(dest, nksize ? nksize : bl->ksize,
246 rchild->pgno, bl->flags & P_BIGKEY);
247 memmove(dest, bl->bytes, nksize ? nksize : bl->ksize);
248 if (bl->flags & P_BIGKEY) {
249 pgno_t pgno;
250 memcpy(&pgno, bl->bytes, sizeof(pgno));
251 if (bt_preserve(t, pgno) == RET_ERROR)
252 goto err1;
253 }
254 break;
255 case P_RINTERNAL:
256 /*
257 * Update the left page count. If split
258 * added at index 0, fix the correct page.
259 */
260 if (skip > 0)
261 dest = (char *)(void *)h + h->linp[skip - 1];
262 else
263 dest = (char *)(void *)l + l->linp[NEXTINDEX(l) - 1];
264 ((RINTERNAL *)(void *)dest)->nrecs = rec_total(lchild);
265 ((RINTERNAL *)(void *)dest)->pgno = lchild->pgno;
266
267 /* Update the right page count. */
268 h->linp[skip] = h->upper -= nbytes;
269 dest = (char *)(void *)h + h->linp[skip];
270 ((RINTERNAL *)(void *)dest)->nrecs = rec_total(rchild);
271 ((RINTERNAL *)(void *)dest)->pgno = rchild->pgno;
272 break;
273 case P_RLEAF:
274 /*
275 * Update the left page count. If split
276 * added at index 0, fix the correct page.
277 */
278 if (skip > 0)
279 dest = (char *)(void *)h + h->linp[skip - 1];
280 else
281 dest = (char *)(void *)l + l->linp[NEXTINDEX(l) - 1];
282 ((RINTERNAL *)(void *)dest)->nrecs = NEXTINDEX(lchild);
283 ((RINTERNAL *)(void *)dest)->pgno = lchild->pgno;
284
285 /* Update the right page count. */
286 h->linp[skip] = h->upper -= nbytes;
287 dest = (char *)(void *)h + h->linp[skip];
288 ((RINTERNAL *)(void *)dest)->nrecs = NEXTINDEX(rchild);
289 ((RINTERNAL *)(void *)dest)->pgno = rchild->pgno;
290 break;
291 default:
292 abort();
293 }
294
295 /* Unpin the held pages. */
296 if (!parentsplit) {
297 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
298 break;
299 }
300
301 /* If the root page was split, make it look right. */
302 if (sp->pgno == P_ROOT &&
303 (F_ISSET(t, R_RECNO) ?
304 bt_rroot(t, sp, l, r) : bt_broot(t, sp, l, r)) == RET_ERROR)
305 goto err1;
306
307 mpool_put(t->bt_mp, lchild, MPOOL_DIRTY);
308 mpool_put(t->bt_mp, rchild, MPOOL_DIRTY);
309 }
310
311 /* Unpin the held pages. */
312 mpool_put(t->bt_mp, l, MPOOL_DIRTY);
313 mpool_put(t->bt_mp, r, MPOOL_DIRTY);
314
315 /* Clear any pages left on the stack. */
316 return (RET_SUCCESS);
317
318 /*
319 * If something fails in the above loop we were already walking back
320 * up the tree and the tree is now inconsistent. Nothing much we can
321 * do about it but release any memory we're holding.
322 */
323 err1: mpool_put(t->bt_mp, lchild, MPOOL_DIRTY);
324 mpool_put(t->bt_mp, rchild, MPOOL_DIRTY);
325
326 err2: mpool_put(t->bt_mp, l, 0);
327 mpool_put(t->bt_mp, r, 0);
328 __dbpanic(t->bt_dbp);
329 return (RET_ERROR);
330 }
331
332 /*
333 * BT_PAGE -- Split a non-root page of a btree.
334 *
335 * Parameters:
336 * t: tree
337 * h: root page
338 * lp: pointer to left page pointer
339 * rp: pointer to right page pointer
340 * skip: pointer to index to leave open
341 * ilen: insert length
342 *
343 * Returns:
344 * Pointer to page in which to insert or NULL on error.
345 */
346 static PAGE *
bt_page(BTREE * t,PAGE * h,PAGE ** lp,PAGE ** rp,indx_t * skip,size_t ilen)347 bt_page(BTREE *t, PAGE *h, PAGE **lp, PAGE **rp, indx_t *skip, size_t ilen)
348 {
349 PAGE *l, *r, *tp;
350 pgno_t npg;
351
352 #ifdef STATISTICS
353 ++bt_split;
354 #endif
355 /* Put the new right page for the split into place. */
356 if ((r = __bt_new(t, &npg)) == NULL)
357 return (NULL);
358 r->pgno = npg;
359 r->lower = BTDATAOFF;
360 r->upper = t->bt_psize;
361 r->nextpg = h->nextpg;
362 r->prevpg = h->pgno;
363 r->flags = h->flags & P_TYPE;
364
365 /*
366 * If we're splitting the last page on a level because we're appending
367 * a key to it (skip is NEXTINDEX()), it's likely that the data is
368 * sorted. Adding an empty page on the side of the level is less work
369 * and can push the fill factor much higher than normal. If we're
370 * wrong it's no big deal, we'll just do the split the right way next
371 * time. It may look like it's equally easy to do a similar hack for
372 * reverse sorted data, that is, split the tree left, but it's not.
373 * Don't even try.
374 */
375 if (h->nextpg == P_INVALID && *skip == NEXTINDEX(h)) {
376 #ifdef STATISTICS
377 ++bt_sortsplit;
378 #endif
379 h->nextpg = r->pgno;
380 r->lower = BTDATAOFF + sizeof(indx_t);
381 *skip = 0;
382 *lp = h;
383 *rp = r;
384 return (r);
385 }
386
387 /* Put the new left page for the split into place. */
388 if ((l = calloc(1, t->bt_psize)) == NULL) {
389 mpool_put(t->bt_mp, r, 0);
390 return (NULL);
391 }
392 #ifdef PURIFY
393 memset(l, 0xff, t->bt_psize);
394 #endif
395 l->pgno = h->pgno;
396 l->nextpg = r->pgno;
397 l->prevpg = h->prevpg;
398 l->lower = BTDATAOFF;
399 l->upper = t->bt_psize;
400 l->flags = h->flags & P_TYPE;
401
402 /* Fix up the previous pointer of the page after the split page. */
403 if (h->nextpg != P_INVALID) {
404 if ((tp = mpool_get(t->bt_mp, h->nextpg, 0)) == NULL) {
405 free(l);
406 /* XXX mpool_free(t->bt_mp, r->pgno); */
407 return (NULL);
408 }
409 tp->prevpg = r->pgno;
410 mpool_put(t->bt_mp, tp, MPOOL_DIRTY);
411 }
412
413 /*
414 * Split right. The key/data pairs aren't sorted in the btree page so
415 * it's simpler to copy the data from the split page onto two new pages
416 * instead of copying half the data to the right page and compacting
417 * the left page in place. Since the left page can't change, we have
418 * to swap the original and the allocated left page after the split.
419 */
420 tp = bt_psplit(t, h, l, r, skip, ilen);
421
422 /* Move the new left page onto the old left page. */
423 memmove(h, l, t->bt_psize);
424 if (tp == l)
425 tp = h;
426 free(l);
427
428 *lp = h;
429 *rp = r;
430 return (tp);
431 }
432
433 /*
434 * BT_ROOT -- Split the root page of a btree.
435 *
436 * Parameters:
437 * t: tree
438 * h: root page
439 * lp: pointer to left page pointer
440 * rp: pointer to right page pointer
441 * skip: pointer to index to leave open
442 * ilen: insert length
443 *
444 * Returns:
445 * Pointer to page in which to insert or NULL on error.
446 */
447 static PAGE *
bt_root(BTREE * t,PAGE * h,PAGE ** lp,PAGE ** rp,indx_t * skip,size_t ilen)448 bt_root(BTREE *t, PAGE *h, PAGE **lp, PAGE **rp, indx_t *skip, size_t ilen)
449 {
450 PAGE *l, *r, *tp;
451 pgno_t lnpg, rnpg;
452
453 #ifdef STATISTICS
454 ++bt_split;
455 ++bt_rootsplit;
456 #endif
457 /* Put the new left and right pages for the split into place. */
458 if ((l = __bt_new(t, &lnpg)) == NULL ||
459 (r = __bt_new(t, &rnpg)) == NULL)
460 return (NULL);
461 l->pgno = lnpg;
462 r->pgno = rnpg;
463 l->nextpg = r->pgno;
464 r->prevpg = l->pgno;
465 l->prevpg = r->nextpg = P_INVALID;
466 l->lower = r->lower = BTDATAOFF;
467 l->upper = r->upper = t->bt_psize;
468 l->flags = r->flags = h->flags & P_TYPE;
469
470 /* Split the root page. */
471 tp = bt_psplit(t, h, l, r, skip, ilen);
472
473 *lp = l;
474 *rp = r;
475 return (tp);
476 }
477
478 /*
479 * BT_RROOT -- Fix up the recno root page after it has been split.
480 *
481 * Parameters:
482 * t: tree
483 * h: root page
484 * l: left page
485 * r: right page
486 *
487 * Returns:
488 * RET_ERROR, RET_SUCCESS
489 */
490 static int
bt_rroot(BTREE * t,PAGE * h,PAGE * l,PAGE * r)491 bt_rroot(BTREE *t, PAGE *h, PAGE *l, PAGE *r)
492 {
493 char *dest;
494 uint32_t sz;
495 size_t temp;
496
497 temp = t->bt_psize - NRINTERNAL;
498 _DBFIT(temp, uint32_t);
499 sz = (uint32_t)temp;
500
501 /* Insert the left and right keys, set the header information. */
502 _DBFIT(sz, indx_t);
503 h->linp[0] = h->upper = (indx_t)sz;
504 dest = (char *)(void *)h + h->upper;
505 WR_RINTERNAL(dest,
506 l->flags & P_RLEAF ? NEXTINDEX(l) : rec_total(l), l->pgno);
507
508 h->linp[1] = h->upper -= NRINTERNAL;
509 dest = (char *)(void *)h + h->upper;
510 WR_RINTERNAL(dest,
511 r->flags & P_RLEAF ? NEXTINDEX(r) : rec_total(r), r->pgno);
512
513 h->lower = BTDATAOFF + 2 * sizeof(indx_t);
514
515 /* Unpin the root page, set to recno internal page. */
516 h->flags &= ~P_TYPE;
517 h->flags |= P_RINTERNAL;
518 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
519
520 return (RET_SUCCESS);
521 }
522
523 /*
524 * BT_BROOT -- Fix up the btree root page after it has been split.
525 *
526 * Parameters:
527 * t: tree
528 * h: root page
529 * l: left page
530 * r: right page
531 *
532 * Returns:
533 * RET_ERROR, RET_SUCCESS
534 */
535 static int
bt_broot(BTREE * t,PAGE * h,PAGE * l,PAGE * r)536 bt_broot(BTREE *t, PAGE *h, PAGE *l, PAGE *r)
537 {
538 BINTERNAL *bi = NULL; /* pacify gcc */
539 BLEAF *bl;
540 uint32_t nbytes;
541 char *dest;
542
543 /*
544 * If the root page was a leaf page, change it into an internal page.
545 * We copy the key we split on (but not the key's data, in the case of
546 * a leaf page) to the new root page.
547 *
548 * The btree comparison code guarantees that the left-most key on any
549 * level of the tree is never used, so it doesn't need to be filled in.
550 */
551 nbytes = NBINTERNAL(0);
552 h->linp[0] = h->upper = t->bt_psize - nbytes;
553 dest = (char *)(void *)h + h->upper;
554 WR_BINTERNAL(dest, 0, l->pgno, 0);
555
556 switch (h->flags & P_TYPE) {
557 case P_BLEAF:
558 bl = GETBLEAF(r, 0);
559 nbytes = NBINTERNAL(bl->ksize);
560 h->linp[1] = h->upper -= nbytes;
561 dest = (char *)(void *)h + h->upper;
562 WR_BINTERNAL(dest, bl->ksize, r->pgno, 0);
563 memmove(dest, bl->bytes, bl->ksize);
564
565 /*
566 * If the key is on an overflow page, mark the overflow chain
567 * so it isn't deleted when the leaf copy of the key is deleted.
568 */
569 if (bl->flags & P_BIGKEY) {
570 pgno_t pgno;
571 memcpy(&pgno, bl->bytes, sizeof(pgno));
572 if (bt_preserve(t, pgno) == RET_ERROR)
573 return (RET_ERROR);
574 }
575 break;
576 case P_BINTERNAL:
577 bi = GETBINTERNAL(r, 0);
578 nbytes = NBINTERNAL(bi->ksize);
579 h->linp[1] = h->upper -= nbytes;
580 dest = (char *)(void *)h + h->upper;
581 memmove(dest, bi, nbytes);
582 ((BINTERNAL *)(void *)dest)->pgno = r->pgno;
583 break;
584 default:
585 abort();
586 }
587
588 /* There are two keys on the page. */
589 h->lower = BTDATAOFF + 2 * sizeof(indx_t);
590
591 /* Unpin the root page, set to btree internal page. */
592 h->flags &= ~P_TYPE;
593 h->flags |= P_BINTERNAL;
594 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
595
596 return (RET_SUCCESS);
597 }
598
599 /*
600 * BT_PSPLIT -- Do the real work of splitting the page.
601 *
602 * Parameters:
603 * t: tree
604 * h: page to be split
605 * l: page to put lower half of data
606 * r: page to put upper half of data
607 * pskip: pointer to index to leave open
608 * ilen: insert length
609 *
610 * Returns:
611 * Pointer to page in which to insert.
612 */
613 static PAGE *
bt_psplit(BTREE * t,PAGE * h,PAGE * l,PAGE * r,indx_t * pskip,size_t ilen)614 bt_psplit(BTREE *t, PAGE *h, PAGE *l, PAGE *r, indx_t *pskip, size_t ilen)
615 {
616 BINTERNAL *bi;
617 BLEAF *bl;
618 CURSOR *c;
619 RLEAF *rl;
620 PAGE *rval;
621 void *src = NULL; /* pacify gcc */
622 indx_t full, half, nxt, off, skip, top, used;
623 uint32_t nbytes;
624 size_t temp;
625 int bigkeycnt, isbigkey;
626
627 /*
628 * Split the data to the left and right pages. Leave the skip index
629 * open. Additionally, make some effort not to split on an overflow
630 * key. This makes internal page processing faster and can save
631 * space as overflow keys used by internal pages are never deleted.
632 */
633 bigkeycnt = 0;
634 skip = *pskip;
635 temp = t->bt_psize - BTDATAOFF;
636 _DBFIT(temp, indx_t);
637 full = (indx_t)temp;
638 half = full / 2;
639 used = 0;
640 for (nxt = off = 0, top = NEXTINDEX(h); nxt < top; ++off) {
641 if (skip == off) {
642 _DBFIT(ilen, uint32_t);
643 nbytes = (uint32_t)ilen;
644 isbigkey = 0; /* XXX: not really known. */
645 } else
646 switch (h->flags & P_TYPE) {
647 case P_BINTERNAL:
648 src = bi = GETBINTERNAL(h, nxt);
649 nbytes = NBINTERNAL(bi->ksize);
650 isbigkey = bi->flags & P_BIGKEY;
651 break;
652 case P_BLEAF:
653 src = bl = GETBLEAF(h, nxt);
654 nbytes = NBLEAF(bl);
655 isbigkey = bl->flags & P_BIGKEY;
656 break;
657 case P_RINTERNAL:
658 src = GETRINTERNAL(h, nxt);
659 nbytes = NRINTERNAL;
660 isbigkey = 0;
661 break;
662 case P_RLEAF:
663 src = rl = GETRLEAF(h, nxt);
664 nbytes = NRLEAF(rl);
665 isbigkey = 0;
666 break;
667 default:
668 abort();
669 }
670
671 /*
672 * If the key/data pairs are substantial fractions of the max
673 * possible size for the page, it's possible to get situations
674 * where we decide to try and copy too much onto the left page.
675 * Make sure that doesn't happen.
676 */
677 if ((skip <= off && used + nbytes + sizeof(indx_t) >= full) ||
678 nxt == top - 1) {
679 --off;
680 break;
681 }
682
683 /* Copy the key/data pair, if not the skipped index. */
684 if (skip != off) {
685 ++nxt;
686
687 l->linp[off] = l->upper -= nbytes;
688 memmove((char *)(void *)l + l->upper, src, nbytes);
689 }
690
691 temp = nbytes + sizeof(indx_t);
692 _DBFIT(temp, indx_t);
693 used += (indx_t)temp;
694 if (used >= half) {
695 if (!isbigkey || bigkeycnt == 3)
696 break;
697 else
698 ++bigkeycnt;
699 }
700 }
701
702 /*
703 * Off is the last offset that's valid for the left page.
704 * Nxt is the first offset to be placed on the right page.
705 */
706 temp = (off + 1) * sizeof(indx_t);
707 _DBFIT(temp, indx_t);
708 l->lower += (indx_t)temp;
709
710 /*
711 * If splitting the page that the cursor was on, the cursor has to be
712 * adjusted to point to the same record as before the split. If the
713 * cursor is at or past the skipped slot, the cursor is incremented by
714 * one. If the cursor is on the right page, it is decremented by the
715 * number of records split to the left page.
716 */
717 c = &t->bt_cursor;
718 if (F_ISSET(c, CURS_INIT) && c->pg.pgno == h->pgno) {
719 if (c->pg.index >= skip)
720 ++c->pg.index;
721 if (c->pg.index < nxt) /* Left page. */
722 c->pg.pgno = l->pgno;
723 else { /* Right page. */
724 c->pg.pgno = r->pgno;
725 c->pg.index -= nxt;
726 }
727 }
728
729 /*
730 * If the skipped index was on the left page, just return that page.
731 * Otherwise, adjust the skip index to reflect the new position on
732 * the right page.
733 */
734 if (skip <= off) {
735 skip = MAX_PAGE_OFFSET;
736 rval = l;
737 } else {
738 rval = r;
739 *pskip -= nxt;
740 }
741
742 for (off = 0; nxt < top; ++off) {
743 if (skip == nxt) {
744 ++off;
745 skip = MAX_PAGE_OFFSET;
746 }
747 switch (h->flags & P_TYPE) {
748 case P_BINTERNAL:
749 src = bi = GETBINTERNAL(h, nxt);
750 nbytes = NBINTERNAL(bi->ksize);
751 break;
752 case P_BLEAF:
753 src = bl = GETBLEAF(h, nxt);
754 nbytes = NBLEAF(bl);
755 break;
756 case P_RINTERNAL:
757 src = GETRINTERNAL(h, nxt);
758 nbytes = NRINTERNAL;
759 break;
760 case P_RLEAF:
761 src = rl = GETRLEAF(h, nxt);
762 nbytes = NRLEAF(rl);
763 break;
764 default:
765 abort();
766 }
767 ++nxt;
768 r->linp[off] = r->upper -= nbytes;
769 memmove((char *)(void *)r + r->upper, src, nbytes);
770 }
771 temp = off * sizeof(indx_t);
772 _DBFIT(temp, indx_t);
773 r->lower += (indx_t)temp;
774
775 /* If the key is being appended to the page, adjust the index. */
776 if (skip == top)
777 r->lower += sizeof(indx_t);
778
779 return (rval);
780 }
781
782 /*
783 * BT_PRESERVE -- Mark a chain of pages as used by an internal node.
784 *
785 * Chains of indirect blocks pointed to by leaf nodes get reclaimed when the
786 * record that references them gets deleted. Chains pointed to by internal
787 * pages never get deleted. This routine marks a chain as pointed to by an
788 * internal page.
789 *
790 * Parameters:
791 * t: tree
792 * pg: page number of first page in the chain.
793 *
794 * Returns:
795 * RET_SUCCESS, RET_ERROR.
796 */
797 static int
bt_preserve(BTREE * t,pgno_t pg)798 bt_preserve(BTREE *t, pgno_t pg)
799 {
800 PAGE *h;
801
802 if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL)
803 return (RET_ERROR);
804 h->flags |= P_PRESERVE;
805 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
806 return (RET_SUCCESS);
807 }
808
809 /*
810 * REC_TOTAL -- Return the number of recno entries below a page.
811 *
812 * Parameters:
813 * h: page
814 *
815 * Returns:
816 * The number of recno entries below a page.
817 *
818 * XXX
819 * These values could be set by the bt_psplit routine. The problem is that the
820 * entry has to be popped off of the stack etc. or the values have to be passed
821 * all the way back to bt_split/bt_rroot and it's not very clean.
822 */
823 static recno_t
rec_total(PAGE * h)824 rec_total(PAGE *h)
825 {
826 recno_t recs;
827 indx_t nxt, top;
828
829 for (recs = 0, nxt = 0, top = NEXTINDEX(h); nxt < top; ++nxt)
830 recs += GETRINTERNAL(h, nxt)->nrecs;
831 return (recs);
832 }
833