1 /*
2 * Copyright (C) International Business Machines Corp., 2000-2004
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
12 * the GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 */
18 /*
19 * jfs_xtree.c: extent allocation descriptor B+-tree manager
20 */
21
22 #include <linux/fs.h>
23 #include "jfs_incore.h"
24 #include "jfs_filsys.h"
25 #include "jfs_metapage.h"
26 #include "jfs_dmap.h"
27 #include "jfs_dinode.h"
28 #include "jfs_superblock.h"
29 #include "jfs_debug.h"
30
31 /*
32 * xtree local flag
33 */
34 #define XT_INSERT 0x00000001
35
36 /*
37 * xtree key/entry comparison: extent offset
38 *
39 * return:
40 * -1: k < start of extent
41 * 0: start_of_extent <= k <= end_of_extent
42 * 1: k > end_of_extent
43 */
44 #define XT_CMP(CMP, K, X, OFFSET64)\
45 {\
46 OFFSET64 = offsetXAD(X);\
47 (CMP) = ((K) >= OFFSET64 + lengthXAD(X)) ? 1 :\
48 ((K) < OFFSET64) ? -1 : 0;\
49 }
50
51 /* write a xad entry */
52 #define XT_PUTENTRY(XAD, FLAG, OFF, LEN, ADDR)\
53 {\
54 (XAD)->flag = (FLAG);\
55 XADoffset((XAD), (OFF));\
56 XADlength((XAD), (LEN));\
57 XADaddress((XAD), (ADDR));\
58 }
59
60 #define XT_PAGE(IP, MP) BT_PAGE(IP, MP, xtpage_t, i_xtroot)
61
62 /* get page buffer for specified block address */
63 /* ToDo: Replace this ugly macro with a function */
64 #define XT_GETPAGE(IP, BN, MP, SIZE, P, RC)\
65 {\
66 BT_GETPAGE(IP, BN, MP, xtpage_t, SIZE, P, RC, i_xtroot)\
67 if (!(RC))\
68 {\
69 if ((le16_to_cpu((P)->header.nextindex) < XTENTRYSTART) ||\
70 (le16_to_cpu((P)->header.nextindex) > le16_to_cpu((P)->header.maxentry)) ||\
71 (le16_to_cpu((P)->header.maxentry) > (((BN)==0)?XTROOTMAXSLOT:PSIZE>>L2XTSLOTSIZE)))\
72 {\
73 jfs_error((IP)->i_sb, "XT_GETPAGE: xtree page corrupt");\
74 BT_PUTPAGE(MP);\
75 MP = NULL;\
76 RC = -EIO;\
77 }\
78 }\
79 }
80
81 /* for consistency */
82 #define XT_PUTPAGE(MP) BT_PUTPAGE(MP)
83
84 #define XT_GETSEARCH(IP, LEAF, BN, MP, P, INDEX) \
85 BT_GETSEARCH(IP, LEAF, BN, MP, xtpage_t, P, INDEX, i_xtroot)
86 /* xtree entry parameter descriptor */
87 struct xtsplit {
88 struct metapage *mp;
89 s16 index;
90 u8 flag;
91 s64 off;
92 s64 addr;
93 int len;
94 struct pxdlist *pxdlist;
95 };
96
97
98 /*
99 * statistics
100 */
101 #ifdef CONFIG_JFS_STATISTICS
102 static struct {
103 uint search;
104 uint fastSearch;
105 uint split;
106 } xtStat;
107 #endif
108
109
110 /*
111 * forward references
112 */
113 static int xtSearch(struct inode *ip,
114 s64 xoff, int *cmpp, struct btstack * btstack, int flag);
115
116 static int xtSplitUp(tid_t tid,
117 struct inode *ip,
118 struct xtsplit * split, struct btstack * btstack);
119
120 static int xtSplitPage(tid_t tid, struct inode *ip, struct xtsplit * split,
121 struct metapage ** rmpp, s64 * rbnp);
122
123 static int xtSplitRoot(tid_t tid, struct inode *ip,
124 struct xtsplit * split, struct metapage ** rmpp);
125
126 #ifdef _STILL_TO_PORT
127 static int xtDeleteUp(tid_t tid, struct inode *ip, struct metapage * fmp,
128 xtpage_t * fp, struct btstack * btstack);
129
130 static int xtSearchNode(struct inode *ip,
131 xad_t * xad,
132 int *cmpp, struct btstack * btstack, int flag);
133
134 static int xtRelink(tid_t tid, struct inode *ip, xtpage_t * fp);
135 #endif /* _STILL_TO_PORT */
136
137 /* External references */
138
139 /*
140 * debug control
141 */
142 /* #define _JFS_DEBUG_XTREE 1 */
143
144
145 /*
146 * xtLookup()
147 *
148 * function: map a single page into a physical extent;
149 */
xtLookup(struct inode * ip,s64 lstart,s64 llen,int * pflag,s64 * paddr,s32 * plen,int no_check)150 int xtLookup(struct inode *ip, s64 lstart,
151 s64 llen, int *pflag, s64 * paddr, s32 * plen, int no_check)
152 {
153 int rc = 0;
154 struct btstack btstack;
155 int cmp;
156 s64 bn;
157 struct metapage *mp;
158 xtpage_t *p;
159 int index;
160 xad_t *xad;
161 s64 size, xoff, xend;
162 int xlen;
163 s64 xaddr;
164
165 *plen = 0;
166
167 if (!no_check) {
168 /* is lookup offset beyond eof ? */
169 size = ((u64) ip->i_size + (JFS_SBI(ip->i_sb)->bsize - 1)) >>
170 JFS_SBI(ip->i_sb)->l2bsize;
171 if (lstart >= size) {
172 jfs_err("xtLookup: lstart (0x%lx) >= size (0x%lx)",
173 (ulong) lstart, (ulong) size);
174 return 0;
175 }
176 }
177
178 /*
179 * search for the xad entry covering the logical extent
180 */
181 //search:
182 if ((rc = xtSearch(ip, lstart, &cmp, &btstack, 0))) {
183 jfs_err("xtLookup: xtSearch returned %d", rc);
184 return rc;
185 }
186
187 /*
188 * compute the physical extent covering logical extent
189 *
190 * N.B. search may have failed (e.g., hole in sparse file),
191 * and returned the index of the next entry.
192 */
193 /* retrieve search result */
194 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
195
196 /* is xad found covering start of logical extent ?
197 * lstart is a page start address,
198 * i.e., lstart cannot start in a hole;
199 */
200 if (cmp)
201 goto out;
202
203 /*
204 * lxd covered by xad
205 */
206 xad = &p->xad[index];
207 xoff = offsetXAD(xad);
208 xlen = lengthXAD(xad);
209 xend = xoff + xlen;
210 xaddr = addressXAD(xad);
211
212 /* initialize new pxd */
213 *pflag = xad->flag;
214 *paddr = xaddr + (lstart - xoff);
215 /* a page must be fully covered by an xad */
216 *plen = min(xend - lstart, llen);
217
218 out:
219 XT_PUTPAGE(mp);
220
221 return rc;
222 }
223
224
225 /*
226 * xtLookupList()
227 *
228 * function: map a single logical extent into a list of physical extent;
229 *
230 * parameter:
231 * struct inode *ip,
232 * struct lxdlist *lxdlist, lxd list (in)
233 * struct xadlist *xadlist, xad list (in/out)
234 * int flag)
235 *
236 * coverage of lxd by xad under assumption of
237 * . lxd's are ordered and disjoint.
238 * . xad's are ordered and disjoint.
239 *
240 * return:
241 * 0: success
242 *
243 * note: a page being written (even a single byte) is backed fully,
244 * except the last page which is only backed with blocks
245 * required to cover the last byte;
246 * the extent backing a page is fully contained within an xad;
247 */
xtLookupList(struct inode * ip,struct lxdlist * lxdlist,struct xadlist * xadlist,int flag)248 int xtLookupList(struct inode *ip, struct lxdlist * lxdlist,
249 struct xadlist * xadlist, int flag)
250 {
251 int rc = 0;
252 struct btstack btstack;
253 int cmp;
254 s64 bn;
255 struct metapage *mp;
256 xtpage_t *p;
257 int index;
258 lxd_t *lxd;
259 xad_t *xad, *pxd;
260 s64 size, lstart, lend, xstart, xend, pstart;
261 s64 llen, xlen, plen;
262 s64 xaddr, paddr;
263 int nlxd, npxd, maxnpxd;
264
265 npxd = xadlist->nxad = 0;
266 maxnpxd = xadlist->maxnxad;
267 pxd = xadlist->xad;
268
269 nlxd = lxdlist->nlxd;
270 lxd = lxdlist->lxd;
271
272 lstart = offsetLXD(lxd);
273 llen = lengthLXD(lxd);
274 lend = lstart + llen;
275
276 size = (ip->i_size + (JFS_SBI(ip->i_sb)->bsize - 1)) >>
277 JFS_SBI(ip->i_sb)->l2bsize;
278
279 /*
280 * search for the xad entry covering the logical extent
281 */
282 search:
283 if (lstart >= size)
284 return 0;
285
286 if ((rc = xtSearch(ip, lstart, &cmp, &btstack, 0)))
287 return rc;
288
289 /*
290 * compute the physical extent covering logical extent
291 *
292 * N.B. search may have failed (e.g., hole in sparse file),
293 * and returned the index of the next entry.
294 */
295 //map:
296 /* retrieve search result */
297 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
298
299 /* is xad on the next sibling page ? */
300 if (index == le16_to_cpu(p->header.nextindex)) {
301 if (p->header.flag & BT_ROOT)
302 goto mapend;
303
304 if ((bn = le64_to_cpu(p->header.next)) == 0)
305 goto mapend;
306
307 XT_PUTPAGE(mp);
308
309 /* get next sibling page */
310 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
311 if (rc)
312 return rc;
313
314 index = XTENTRYSTART;
315 }
316
317 xad = &p->xad[index];
318
319 /*
320 * is lxd covered by xad ?
321 */
322 compare:
323 xstart = offsetXAD(xad);
324 xlen = lengthXAD(xad);
325 xend = xstart + xlen;
326 xaddr = addressXAD(xad);
327
328 compare1:
329 if (xstart < lstart)
330 goto compare2;
331
332 /* (lstart <= xstart) */
333
334 /* lxd is NOT covered by xad */
335 if (lend <= xstart) {
336 /*
337 * get next lxd
338 */
339 if (--nlxd == 0)
340 goto mapend;
341 lxd++;
342
343 lstart = offsetLXD(lxd);
344 llen = lengthLXD(lxd);
345 lend = lstart + llen;
346 if (lstart >= size)
347 goto mapend;
348
349 /* compare with the current xad */
350 goto compare1;
351 }
352 /* lxd is covered by xad */
353 else { /* (xstart < lend) */
354
355 /* initialize new pxd */
356 pstart = xstart;
357 plen = min(lend - xstart, xlen);
358 paddr = xaddr;
359
360 goto cover;
361 }
362
363 /* (xstart < lstart) */
364 compare2:
365 /* lxd is covered by xad */
366 if (lstart < xend) {
367 /* initialize new pxd */
368 pstart = lstart;
369 plen = min(xend - lstart, llen);
370 paddr = xaddr + (lstart - xstart);
371
372 goto cover;
373 }
374 /* lxd is NOT covered by xad */
375 else { /* (xend <= lstart) */
376
377 /*
378 * get next xad
379 *
380 * linear search next xad covering lxd on
381 * the current xad page, and then tree search
382 */
383 if (index == le16_to_cpu(p->header.nextindex) - 1) {
384 if (p->header.flag & BT_ROOT)
385 goto mapend;
386
387 XT_PUTPAGE(mp);
388 goto search;
389 } else {
390 index++;
391 xad++;
392
393 /* compare with new xad */
394 goto compare;
395 }
396 }
397
398 /*
399 * lxd is covered by xad and a new pxd has been initialized
400 * (lstart <= xstart < lend) or (xstart < lstart < xend)
401 */
402 cover:
403 /* finalize pxd corresponding to current xad */
404 XT_PUTENTRY(pxd, xad->flag, pstart, plen, paddr);
405
406 if (++npxd >= maxnpxd)
407 goto mapend;
408 pxd++;
409
410 /*
411 * lxd is fully covered by xad
412 */
413 if (lend <= xend) {
414 /*
415 * get next lxd
416 */
417 if (--nlxd == 0)
418 goto mapend;
419 lxd++;
420
421 lstart = offsetLXD(lxd);
422 llen = lengthLXD(lxd);
423 lend = lstart + llen;
424 if (lstart >= size)
425 goto mapend;
426
427 /*
428 * test for old xad covering new lxd
429 * (old xstart < new lstart)
430 */
431 goto compare2;
432 }
433 /*
434 * lxd is partially covered by xad
435 */
436 else { /* (xend < lend) */
437
438 /*
439 * get next xad
440 *
441 * linear search next xad covering lxd on
442 * the current xad page, and then next xad page search
443 */
444 if (index == le16_to_cpu(p->header.nextindex) - 1) {
445 if (p->header.flag & BT_ROOT)
446 goto mapend;
447
448 if ((bn = le64_to_cpu(p->header.next)) == 0)
449 goto mapend;
450
451 XT_PUTPAGE(mp);
452
453 /* get next sibling page */
454 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
455 if (rc)
456 return rc;
457
458 index = XTENTRYSTART;
459 xad = &p->xad[index];
460 } else {
461 index++;
462 xad++;
463 }
464
465 /*
466 * test for new xad covering old lxd
467 * (old lstart < new xstart)
468 */
469 goto compare;
470 }
471
472 mapend:
473 xadlist->nxad = npxd;
474
475 //out:
476 XT_PUTPAGE(mp);
477
478 return rc;
479 }
480
481
482 /*
483 * xtSearch()
484 *
485 * function: search for the xad entry covering specified offset.
486 *
487 * parameters:
488 * ip - file object;
489 * xoff - extent offset;
490 * cmpp - comparison result:
491 * btstack - traverse stack;
492 * flag - search process flag (XT_INSERT);
493 *
494 * returns:
495 * btstack contains (bn, index) of search path traversed to the entry.
496 * *cmpp is set to result of comparison with the entry returned.
497 * the page containing the entry is pinned at exit.
498 */
xtSearch(struct inode * ip,s64 xoff,int * cmpp,struct btstack * btstack,int flag)499 static int xtSearch(struct inode *ip, s64 xoff, /* offset of extent */
500 int *cmpp, struct btstack * btstack, int flag)
501 {
502 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
503 int rc = 0;
504 int cmp = 1; /* init for empty page */
505 s64 bn; /* block number */
506 struct metapage *mp; /* page buffer */
507 xtpage_t *p; /* page */
508 xad_t *xad;
509 int base, index, lim, btindex;
510 struct btframe *btsp;
511 int nsplit = 0; /* number of pages to split */
512 s64 t64;
513
514 INCREMENT(xtStat.search);
515
516 BT_CLR(btstack);
517
518 btstack->nsplit = 0;
519
520 /*
521 * search down tree from root:
522 *
523 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
524 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
525 *
526 * if entry with search key K is not found
527 * internal page search find the entry with largest key Ki
528 * less than K which point to the child page to search;
529 * leaf page search find the entry with smallest key Kj
530 * greater than K so that the returned index is the position of
531 * the entry to be shifted right for insertion of new entry.
532 * for empty tree, search key is greater than any key of the tree.
533 *
534 * by convention, root bn = 0.
535 */
536 for (bn = 0;;) {
537 /* get/pin the page to search */
538 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
539 if (rc)
540 return rc;
541
542 /* try sequential access heuristics with the previous
543 * access entry in target leaf page:
544 * once search narrowed down into the target leaf,
545 * key must either match an entry in the leaf or
546 * key entry does not exist in the tree;
547 */
548 //fastSearch:
549 if ((jfs_ip->btorder & BT_SEQUENTIAL) &&
550 (p->header.flag & BT_LEAF) &&
551 (index = jfs_ip->btindex) <
552 le16_to_cpu(p->header.nextindex)) {
553 xad = &p->xad[index];
554 t64 = offsetXAD(xad);
555 if (xoff < t64 + lengthXAD(xad)) {
556 if (xoff >= t64) {
557 *cmpp = 0;
558 goto out;
559 }
560
561 /* stop sequential access heuristics */
562 goto binarySearch;
563 } else { /* (t64 + lengthXAD(xad)) <= xoff */
564
565 /* try next sequential entry */
566 index++;
567 if (index <
568 le16_to_cpu(p->header.nextindex)) {
569 xad++;
570 t64 = offsetXAD(xad);
571 if (xoff < t64 + lengthXAD(xad)) {
572 if (xoff >= t64) {
573 *cmpp = 0;
574 goto out;
575 }
576
577 /* miss: key falls between
578 * previous and this entry
579 */
580 *cmpp = 1;
581 goto out;
582 }
583
584 /* (xoff >= t64 + lengthXAD(xad));
585 * matching entry may be further out:
586 * stop heuristic search
587 */
588 /* stop sequential access heuristics */
589 goto binarySearch;
590 }
591
592 /* (index == p->header.nextindex);
593 * miss: key entry does not exist in
594 * the target leaf/tree
595 */
596 *cmpp = 1;
597 goto out;
598 }
599
600 /*
601 * if hit, return index of the entry found, and
602 * if miss, where new entry with search key is
603 * to be inserted;
604 */
605 out:
606 /* compute number of pages to split */
607 if (flag & XT_INSERT) {
608 if (p->header.nextindex == /* little-endian */
609 p->header.maxentry)
610 nsplit++;
611 else
612 nsplit = 0;
613 btstack->nsplit = nsplit;
614 }
615
616 /* save search result */
617 btsp = btstack->top;
618 btsp->bn = bn;
619 btsp->index = index;
620 btsp->mp = mp;
621
622 /* update sequential access heuristics */
623 jfs_ip->btindex = index;
624
625 INCREMENT(xtStat.fastSearch);
626 return 0;
627 }
628
629 /* well, ... full search now */
630 binarySearch:
631 lim = le16_to_cpu(p->header.nextindex) - XTENTRYSTART;
632
633 /*
634 * binary search with search key K on the current page
635 */
636 for (base = XTENTRYSTART; lim; lim >>= 1) {
637 index = base + (lim >> 1);
638
639 XT_CMP(cmp, xoff, &p->xad[index], t64);
640 if (cmp == 0) {
641 /*
642 * search hit
643 */
644 /* search hit - leaf page:
645 * return the entry found
646 */
647 if (p->header.flag & BT_LEAF) {
648 *cmpp = cmp;
649
650 /* compute number of pages to split */
651 if (flag & XT_INSERT) {
652 if (p->header.nextindex ==
653 p->header.maxentry)
654 nsplit++;
655 else
656 nsplit = 0;
657 btstack->nsplit = nsplit;
658 }
659
660 /* save search result */
661 btsp = btstack->top;
662 btsp->bn = bn;
663 btsp->index = index;
664 btsp->mp = mp;
665
666 /* init sequential access heuristics */
667 btindex = jfs_ip->btindex;
668 if (index == btindex ||
669 index == btindex + 1)
670 jfs_ip->btorder = BT_SEQUENTIAL;
671 else
672 jfs_ip->btorder = BT_RANDOM;
673 jfs_ip->btindex = index;
674
675 return 0;
676 }
677
678 /* search hit - internal page:
679 * descend/search its child page
680 */
681 goto next;
682 }
683
684 if (cmp > 0) {
685 base = index + 1;
686 --lim;
687 }
688 }
689
690 /*
691 * search miss
692 *
693 * base is the smallest index with key (Kj) greater than
694 * search key (K) and may be zero or maxentry index.
695 */
696 /*
697 * search miss - leaf page:
698 *
699 * return location of entry (base) where new entry with
700 * search key K is to be inserted.
701 */
702 if (p->header.flag & BT_LEAF) {
703 *cmpp = cmp;
704
705 /* compute number of pages to split */
706 if (flag & XT_INSERT) {
707 if (p->header.nextindex ==
708 p->header.maxentry)
709 nsplit++;
710 else
711 nsplit = 0;
712 btstack->nsplit = nsplit;
713 }
714
715 /* save search result */
716 btsp = btstack->top;
717 btsp->bn = bn;
718 btsp->index = base;
719 btsp->mp = mp;
720
721 /* init sequential access heuristics */
722 btindex = jfs_ip->btindex;
723 if (base == btindex || base == btindex + 1)
724 jfs_ip->btorder = BT_SEQUENTIAL;
725 else
726 jfs_ip->btorder = BT_RANDOM;
727 jfs_ip->btindex = base;
728
729 return 0;
730 }
731
732 /*
733 * search miss - non-leaf page:
734 *
735 * if base is non-zero, decrement base by one to get the parent
736 * entry of the child page to search.
737 */
738 index = base ? base - 1 : base;
739
740 /*
741 * go down to child page
742 */
743 next:
744 /* update number of pages to split */
745 if (p->header.nextindex == p->header.maxentry)
746 nsplit++;
747 else
748 nsplit = 0;
749
750 /* push (bn, index) of the parent page/entry */
751 BT_PUSH(btstack, bn, index);
752
753 /* get the child page block number */
754 bn = addressXAD(&p->xad[index]);
755
756 /* unpin the parent page */
757 XT_PUTPAGE(mp);
758 }
759 }
760
761 /*
762 * xtInsert()
763 *
764 * function:
765 *
766 * parameter:
767 * tid - transaction id;
768 * ip - file object;
769 * xflag - extent flag (XAD_NOTRECORDED):
770 * xoff - extent offset;
771 * xlen - extent length;
772 * xaddrp - extent address pointer (in/out):
773 * if (*xaddrp)
774 * caller allocated data extent at *xaddrp;
775 * else
776 * allocate data extent and return its xaddr;
777 * flag -
778 *
779 * return:
780 */
xtInsert(tid_t tid,struct inode * ip,int xflag,s64 xoff,s32 xlen,s64 * xaddrp,int flag)781 int xtInsert(tid_t tid, /* transaction id */
782 struct inode *ip, int xflag, s64 xoff, s32 xlen, s64 * xaddrp,
783 int flag)
784 {
785 int rc = 0;
786 s64 xaddr, hint;
787 struct metapage *mp; /* meta-page buffer */
788 xtpage_t *p; /* base B+-tree index page */
789 s64 bn;
790 int index, nextindex;
791 struct btstack btstack; /* traverse stack */
792 struct xtsplit split; /* split information */
793 xad_t *xad;
794 int cmp;
795 struct tlock *tlck;
796 struct xtlock *xtlck;
797
798 jfs_info("xtInsert: nxoff:0x%lx nxlen:0x%x", (ulong) xoff, xlen);
799
800 /*
801 * search for the entry location at which to insert:
802 *
803 * xtFastSearch() and xtSearch() both returns (leaf page
804 * pinned, index at which to insert).
805 * n.b. xtSearch() may return index of maxentry of
806 * the full page.
807 */
808 if ((rc = xtSearch(ip, xoff, &cmp, &btstack, XT_INSERT)))
809 return rc;
810
811 /* retrieve search result */
812 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
813
814 /* This test must follow XT_GETSEARCH since mp must be valid if
815 * we branch to out: */
816 if (cmp == 0) {
817 rc = -EEXIST;
818 goto out;
819 }
820
821 /*
822 * allocate data extent requested
823 *
824 * allocation hint: last xad
825 */
826 if ((xaddr = *xaddrp) == 0) {
827 if (index > XTENTRYSTART) {
828 xad = &p->xad[index - 1];
829 hint = addressXAD(xad) + lengthXAD(xad) - 1;
830 } else
831 hint = 0;
832 if ((rc = dbAlloc(ip, hint, (s64) xlen, &xaddr)))
833 goto out;
834 }
835
836 /*
837 * insert entry for new extent
838 */
839 xflag |= XAD_NEW;
840
841 /*
842 * if the leaf page is full, split the page and
843 * propagate up the router entry for the new page from split
844 *
845 * The xtSplitUp() will insert the entry and unpin the leaf page.
846 */
847 nextindex = le16_to_cpu(p->header.nextindex);
848 if (nextindex == le16_to_cpu(p->header.maxentry)) {
849 split.mp = mp;
850 split.index = index;
851 split.flag = xflag;
852 split.off = xoff;
853 split.len = xlen;
854 split.addr = xaddr;
855 split.pxdlist = NULL;
856 if ((rc = xtSplitUp(tid, ip, &split, &btstack))) {
857 /* undo data extent allocation */
858 if (*xaddrp == 0)
859 dbFree(ip, xaddr, (s64) xlen);
860 return rc;
861 }
862
863 *xaddrp = xaddr;
864 return 0;
865 }
866
867 /*
868 * insert the new entry into the leaf page
869 */
870 /*
871 * acquire a transaction lock on the leaf page;
872 *
873 * action: xad insertion/extension;
874 */
875 BT_MARK_DIRTY(mp, ip);
876
877 /* if insert into middle, shift right remaining entries. */
878 if (index < nextindex)
879 memmove(&p->xad[index + 1], &p->xad[index],
880 (nextindex - index) * sizeof(xad_t));
881
882 /* insert the new entry: mark the entry NEW */
883 xad = &p->xad[index];
884 XT_PUTENTRY(xad, xflag, xoff, xlen, xaddr);
885
886 /* advance next available entry index */
887 p->header.nextindex =
888 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
889
890 /* Don't log it if there are no links to the file */
891 if (!test_cflag(COMMIT_Nolink, ip)) {
892 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
893 xtlck = (struct xtlock *) & tlck->lock;
894 xtlck->lwm.offset =
895 (xtlck->lwm.offset) ? min(index,
896 (int)xtlck->lwm.offset) : index;
897 xtlck->lwm.length =
898 le16_to_cpu(p->header.nextindex) - xtlck->lwm.offset;
899 }
900
901 *xaddrp = xaddr;
902
903 out:
904 /* unpin the leaf page */
905 XT_PUTPAGE(mp);
906
907 return rc;
908 }
909
910
911 /*
912 * xtSplitUp()
913 *
914 * function:
915 * split full pages as propagating insertion up the tree
916 *
917 * parameter:
918 * tid - transaction id;
919 * ip - file object;
920 * split - entry parameter descriptor;
921 * btstack - traverse stack from xtSearch()
922 *
923 * return:
924 */
925 static int
xtSplitUp(tid_t tid,struct inode * ip,struct xtsplit * split,struct btstack * btstack)926 xtSplitUp(tid_t tid,
927 struct inode *ip, struct xtsplit * split, struct btstack * btstack)
928 {
929 int rc = 0;
930 struct metapage *smp;
931 xtpage_t *sp; /* split page */
932 struct metapage *rmp;
933 s64 rbn; /* new right page block number */
934 struct metapage *rcmp;
935 xtpage_t *rcp; /* right child page */
936 s64 rcbn; /* right child page block number */
937 int skip; /* index of entry of insertion */
938 int nextindex; /* next available entry index of p */
939 struct btframe *parent; /* parent page entry on traverse stack */
940 xad_t *xad;
941 s64 xaddr;
942 int xlen;
943 int nsplit; /* number of pages split */
944 struct pxdlist pxdlist;
945 pxd_t *pxd;
946 struct tlock *tlck;
947 struct xtlock *xtlck;
948
949 smp = split->mp;
950 sp = XT_PAGE(ip, smp);
951
952 /* is inode xtree root extension/inline EA area free ? */
953 if ((sp->header.flag & BT_ROOT) && (!S_ISDIR(ip->i_mode)) &&
954 (le16_to_cpu(sp->header.maxentry) < XTROOTMAXSLOT) &&
955 (JFS_IP(ip)->mode2 & INLINEEA)) {
956 sp->header.maxentry = cpu_to_le16(XTROOTMAXSLOT);
957 JFS_IP(ip)->mode2 &= ~INLINEEA;
958
959 BT_MARK_DIRTY(smp, ip);
960 /*
961 * acquire a transaction lock on the leaf page;
962 *
963 * action: xad insertion/extension;
964 */
965
966 /* if insert into middle, shift right remaining entries. */
967 skip = split->index;
968 nextindex = le16_to_cpu(sp->header.nextindex);
969 if (skip < nextindex)
970 memmove(&sp->xad[skip + 1], &sp->xad[skip],
971 (nextindex - skip) * sizeof(xad_t));
972
973 /* insert the new entry: mark the entry NEW */
974 xad = &sp->xad[skip];
975 XT_PUTENTRY(xad, split->flag, split->off, split->len,
976 split->addr);
977
978 /* advance next available entry index */
979 sp->header.nextindex =
980 cpu_to_le16(le16_to_cpu(sp->header.nextindex) + 1);
981
982 /* Don't log it if there are no links to the file */
983 if (!test_cflag(COMMIT_Nolink, ip)) {
984 tlck = txLock(tid, ip, smp, tlckXTREE | tlckGROW);
985 xtlck = (struct xtlock *) & tlck->lock;
986 xtlck->lwm.offset = (xtlck->lwm.offset) ?
987 min(skip, (int)xtlck->lwm.offset) : skip;
988 xtlck->lwm.length =
989 le16_to_cpu(sp->header.nextindex) -
990 xtlck->lwm.offset;
991 }
992
993 return 0;
994 }
995
996 /*
997 * allocate new index blocks to cover index page split(s)
998 *
999 * allocation hint: ?
1000 */
1001 if (split->pxdlist == NULL) {
1002 nsplit = btstack->nsplit;
1003 split->pxdlist = &pxdlist;
1004 pxdlist.maxnpxd = pxdlist.npxd = 0;
1005 pxd = &pxdlist.pxd[0];
1006 xlen = JFS_SBI(ip->i_sb)->nbperpage;
1007 for (; nsplit > 0; nsplit--, pxd++) {
1008 if ((rc = dbAlloc(ip, (s64) 0, (s64) xlen, &xaddr))
1009 == 0) {
1010 PXDaddress(pxd, xaddr);
1011 PXDlength(pxd, xlen);
1012
1013 pxdlist.maxnpxd++;
1014
1015 continue;
1016 }
1017
1018 /* undo allocation */
1019
1020 XT_PUTPAGE(smp);
1021 return rc;
1022 }
1023 }
1024
1025 /*
1026 * Split leaf page <sp> into <sp> and a new right page <rp>.
1027 *
1028 * The split routines insert the new entry into the leaf page,
1029 * and acquire txLock as appropriate.
1030 * return <rp> pinned and its block number <rpbn>.
1031 */
1032 rc = (sp->header.flag & BT_ROOT) ?
1033 xtSplitRoot(tid, ip, split, &rmp) :
1034 xtSplitPage(tid, ip, split, &rmp, &rbn);
1035
1036 XT_PUTPAGE(smp);
1037
1038 if (rc)
1039 return -EIO;
1040 /*
1041 * propagate up the router entry for the leaf page just split
1042 *
1043 * insert a router entry for the new page into the parent page,
1044 * propagate the insert/split up the tree by walking back the stack
1045 * of (bn of parent page, index of child page entry in parent page)
1046 * that were traversed during the search for the page that split.
1047 *
1048 * the propagation of insert/split up the tree stops if the root
1049 * splits or the page inserted into doesn't have to split to hold
1050 * the new entry.
1051 *
1052 * the parent entry for the split page remains the same, and
1053 * a new entry is inserted at its right with the first key and
1054 * block number of the new right page.
1055 *
1056 * There are a maximum of 3 pages pinned at any time:
1057 * right child, left parent and right parent (when the parent splits)
1058 * to keep the child page pinned while working on the parent.
1059 * make sure that all pins are released at exit.
1060 */
1061 while ((parent = BT_POP(btstack)) != NULL) {
1062 /* parent page specified by stack frame <parent> */
1063
1064 /* keep current child pages <rcp> pinned */
1065 rcmp = rmp;
1066 rcbn = rbn;
1067 rcp = XT_PAGE(ip, rcmp);
1068
1069 /*
1070 * insert router entry in parent for new right child page <rp>
1071 */
1072 /* get/pin the parent page <sp> */
1073 XT_GETPAGE(ip, parent->bn, smp, PSIZE, sp, rc);
1074 if (rc) {
1075 XT_PUTPAGE(rcmp);
1076 return rc;
1077 }
1078
1079 /*
1080 * The new key entry goes ONE AFTER the index of parent entry,
1081 * because the split was to the right.
1082 */
1083 skip = parent->index + 1;
1084
1085 /*
1086 * split or shift right remaining entries of the parent page
1087 */
1088 nextindex = le16_to_cpu(sp->header.nextindex);
1089 /*
1090 * parent page is full - split the parent page
1091 */
1092 if (nextindex == le16_to_cpu(sp->header.maxentry)) {
1093 /* init for parent page split */
1094 split->mp = smp;
1095 split->index = skip; /* index at insert */
1096 split->flag = XAD_NEW;
1097 split->off = offsetXAD(&rcp->xad[XTENTRYSTART]);
1098 split->len = JFS_SBI(ip->i_sb)->nbperpage;
1099 split->addr = rcbn;
1100
1101 /* unpin previous right child page */
1102 XT_PUTPAGE(rcmp);
1103
1104 /* The split routines insert the new entry,
1105 * and acquire txLock as appropriate.
1106 * return <rp> pinned and its block number <rpbn>.
1107 */
1108 rc = (sp->header.flag & BT_ROOT) ?
1109 xtSplitRoot(tid, ip, split, &rmp) :
1110 xtSplitPage(tid, ip, split, &rmp, &rbn);
1111 if (rc) {
1112 XT_PUTPAGE(smp);
1113 return rc;
1114 }
1115
1116 XT_PUTPAGE(smp);
1117 /* keep new child page <rp> pinned */
1118 }
1119 /*
1120 * parent page is not full - insert in parent page
1121 */
1122 else {
1123 /*
1124 * insert router entry in parent for the right child
1125 * page from the first entry of the right child page:
1126 */
1127 /*
1128 * acquire a transaction lock on the parent page;
1129 *
1130 * action: router xad insertion;
1131 */
1132 BT_MARK_DIRTY(smp, ip);
1133
1134 /*
1135 * if insert into middle, shift right remaining entries
1136 */
1137 if (skip < nextindex)
1138 memmove(&sp->xad[skip + 1], &sp->xad[skip],
1139 (nextindex -
1140 skip) << L2XTSLOTSIZE);
1141
1142 /* insert the router entry */
1143 xad = &sp->xad[skip];
1144 XT_PUTENTRY(xad, XAD_NEW,
1145 offsetXAD(&rcp->xad[XTENTRYSTART]),
1146 JFS_SBI(ip->i_sb)->nbperpage, rcbn);
1147
1148 /* advance next available entry index. */
1149 sp->header.nextindex =
1150 cpu_to_le16(le16_to_cpu(sp->header.nextindex) +
1151 1);
1152
1153 /* Don't log it if there are no links to the file */
1154 if (!test_cflag(COMMIT_Nolink, ip)) {
1155 tlck = txLock(tid, ip, smp,
1156 tlckXTREE | tlckGROW);
1157 xtlck = (struct xtlock *) & tlck->lock;
1158 xtlck->lwm.offset = (xtlck->lwm.offset) ?
1159 min(skip, (int)xtlck->lwm.offset) : skip;
1160 xtlck->lwm.length =
1161 le16_to_cpu(sp->header.nextindex) -
1162 xtlck->lwm.offset;
1163 }
1164
1165 /* unpin parent page */
1166 XT_PUTPAGE(smp);
1167
1168 /* exit propagate up */
1169 break;
1170 }
1171 }
1172
1173 /* unpin current right page */
1174 XT_PUTPAGE(rmp);
1175
1176 return 0;
1177 }
1178
1179
1180 /*
1181 * xtSplitPage()
1182 *
1183 * function:
1184 * split a full non-root page into
1185 * original/split/left page and new right page
1186 * i.e., the original/split page remains as left page.
1187 *
1188 * parameter:
1189 * int tid,
1190 * struct inode *ip,
1191 * struct xtsplit *split,
1192 * struct metapage **rmpp,
1193 * u64 *rbnp,
1194 *
1195 * return:
1196 * Pointer to page in which to insert or NULL on error.
1197 */
1198 static int
xtSplitPage(tid_t tid,struct inode * ip,struct xtsplit * split,struct metapage ** rmpp,s64 * rbnp)1199 xtSplitPage(tid_t tid, struct inode *ip,
1200 struct xtsplit * split, struct metapage ** rmpp, s64 * rbnp)
1201 {
1202 int rc = 0;
1203 struct metapage *smp;
1204 xtpage_t *sp;
1205 struct metapage *rmp;
1206 xtpage_t *rp; /* new right page allocated */
1207 s64 rbn; /* new right page block number */
1208 struct metapage *mp;
1209 xtpage_t *p;
1210 s64 nextbn;
1211 int skip, maxentry, middle, righthalf, n;
1212 xad_t *xad;
1213 struct pxdlist *pxdlist;
1214 pxd_t *pxd;
1215 struct tlock *tlck;
1216 struct xtlock *sxtlck = 0, *rxtlck = 0;
1217
1218 smp = split->mp;
1219 sp = XT_PAGE(ip, smp);
1220
1221 INCREMENT(xtStat.split);
1222
1223 /*
1224 * allocate the new right page for the split
1225 */
1226 pxdlist = split->pxdlist;
1227 pxd = &pxdlist->pxd[pxdlist->npxd];
1228 pxdlist->npxd++;
1229 rbn = addressPXD(pxd);
1230 rmp = get_metapage(ip, rbn, PSIZE, 1);
1231 if (rmp == NULL)
1232 return -EIO;
1233
1234 jfs_info("xtSplitPage: ip:0x%p smp:0x%p rmp:0x%p", ip, smp, rmp);
1235
1236 BT_MARK_DIRTY(rmp, ip);
1237 /*
1238 * action: new page;
1239 */
1240
1241 rp = (xtpage_t *) rmp->data;
1242 rp->header.self = *pxd;
1243 rp->header.flag = sp->header.flag & BT_TYPE;
1244 rp->header.maxentry = sp->header.maxentry; /* little-endian */
1245 rp->header.nextindex = cpu_to_le16(XTENTRYSTART);
1246
1247 BT_MARK_DIRTY(smp, ip);
1248 /* Don't log it if there are no links to the file */
1249 if (!test_cflag(COMMIT_Nolink, ip)) {
1250 /*
1251 * acquire a transaction lock on the new right page;
1252 */
1253 tlck = txLock(tid, ip, rmp, tlckXTREE | tlckNEW);
1254 rxtlck = (struct xtlock *) & tlck->lock;
1255 rxtlck->lwm.offset = XTENTRYSTART;
1256 /*
1257 * acquire a transaction lock on the split page
1258 */
1259 tlck = txLock(tid, ip, smp, tlckXTREE | tlckGROW);
1260 sxtlck = (struct xtlock *) & tlck->lock;
1261 }
1262
1263 /*
1264 * initialize/update sibling pointers of <sp> and <rp>
1265 */
1266 nextbn = le64_to_cpu(sp->header.next);
1267 rp->header.next = cpu_to_le64(nextbn);
1268 rp->header.prev = cpu_to_le64(addressPXD(&sp->header.self));
1269 sp->header.next = cpu_to_le64(rbn);
1270
1271 skip = split->index;
1272
1273 /*
1274 * sequential append at tail (after last entry of last page)
1275 *
1276 * if splitting the last page on a level because of appending
1277 * a entry to it (skip is maxentry), it's likely that the access is
1278 * sequential. adding an empty page on the side of the level is less
1279 * work and can push the fill factor much higher than normal.
1280 * if we're wrong it's no big deal - we will do the split the right
1281 * way next time.
1282 * (it may look like it's equally easy to do a similar hack for
1283 * reverse sorted data, that is, split the tree left, but it's not.
1284 * Be my guest.)
1285 */
1286 if (nextbn == 0 && skip == le16_to_cpu(sp->header.maxentry)) {
1287 /*
1288 * acquire a transaction lock on the new/right page;
1289 *
1290 * action: xad insertion;
1291 */
1292 /* insert entry at the first entry of the new right page */
1293 xad = &rp->xad[XTENTRYSTART];
1294 XT_PUTENTRY(xad, split->flag, split->off, split->len,
1295 split->addr);
1296
1297 rp->header.nextindex = cpu_to_le16(XTENTRYSTART + 1);
1298
1299 if (!test_cflag(COMMIT_Nolink, ip)) {
1300 /* rxtlck->lwm.offset = XTENTRYSTART; */
1301 rxtlck->lwm.length = 1;
1302 }
1303
1304 *rmpp = rmp;
1305 *rbnp = rbn;
1306
1307 ip->i_blocks += LBLK2PBLK(ip->i_sb, lengthPXD(pxd));
1308
1309 jfs_info("xtSplitPage: sp:0x%p rp:0x%p", sp, rp);
1310 return 0;
1311 }
1312
1313 /*
1314 * non-sequential insert (at possibly middle page)
1315 */
1316
1317 /*
1318 * update previous pointer of old next/right page of <sp>
1319 */
1320 if (nextbn != 0) {
1321 XT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
1322 if (rc) {
1323 XT_PUTPAGE(rmp);
1324 return rc;
1325 }
1326
1327 BT_MARK_DIRTY(mp, ip);
1328 /*
1329 * acquire a transaction lock on the next page;
1330 *
1331 * action:sibling pointer update;
1332 */
1333 if (!test_cflag(COMMIT_Nolink, ip))
1334 tlck = txLock(tid, ip, mp, tlckXTREE | tlckRELINK);
1335
1336 p->header.prev = cpu_to_le64(rbn);
1337
1338 /* sibling page may have been updated previously, or
1339 * it may be updated later;
1340 */
1341
1342 XT_PUTPAGE(mp);
1343 }
1344
1345 /*
1346 * split the data between the split and new/right pages
1347 */
1348 maxentry = le16_to_cpu(sp->header.maxentry);
1349 middle = maxentry >> 1;
1350 righthalf = maxentry - middle;
1351
1352 /*
1353 * skip index in old split/left page - insert into left page:
1354 */
1355 if (skip <= middle) {
1356 /* move right half of split page to the new right page */
1357 memmove(&rp->xad[XTENTRYSTART], &sp->xad[middle],
1358 righthalf << L2XTSLOTSIZE);
1359
1360 /* shift right tail of left half to make room for new entry */
1361 if (skip < middle)
1362 memmove(&sp->xad[skip + 1], &sp->xad[skip],
1363 (middle - skip) << L2XTSLOTSIZE);
1364
1365 /* insert new entry */
1366 xad = &sp->xad[skip];
1367 XT_PUTENTRY(xad, split->flag, split->off, split->len,
1368 split->addr);
1369
1370 /* update page header */
1371 sp->header.nextindex = cpu_to_le16(middle + 1);
1372 if (!test_cflag(COMMIT_Nolink, ip)) {
1373 sxtlck->lwm.offset = (sxtlck->lwm.offset) ?
1374 min(skip, (int)sxtlck->lwm.offset) : skip;
1375 }
1376
1377 rp->header.nextindex =
1378 cpu_to_le16(XTENTRYSTART + righthalf);
1379 }
1380 /*
1381 * skip index in new right page - insert into right page:
1382 */
1383 else {
1384 /* move left head of right half to right page */
1385 n = skip - middle;
1386 memmove(&rp->xad[XTENTRYSTART], &sp->xad[middle],
1387 n << L2XTSLOTSIZE);
1388
1389 /* insert new entry */
1390 n += XTENTRYSTART;
1391 xad = &rp->xad[n];
1392 XT_PUTENTRY(xad, split->flag, split->off, split->len,
1393 split->addr);
1394
1395 /* move right tail of right half to right page */
1396 if (skip < maxentry)
1397 memmove(&rp->xad[n + 1], &sp->xad[skip],
1398 (maxentry - skip) << L2XTSLOTSIZE);
1399
1400 /* update page header */
1401 sp->header.nextindex = cpu_to_le16(middle);
1402 if (!test_cflag(COMMIT_Nolink, ip)) {
1403 sxtlck->lwm.offset = (sxtlck->lwm.offset) ?
1404 min(middle, (int)sxtlck->lwm.offset) : middle;
1405 }
1406
1407 rp->header.nextindex = cpu_to_le16(XTENTRYSTART +
1408 righthalf + 1);
1409 }
1410
1411 if (!test_cflag(COMMIT_Nolink, ip)) {
1412 sxtlck->lwm.length = le16_to_cpu(sp->header.nextindex) -
1413 sxtlck->lwm.offset;
1414
1415 /* rxtlck->lwm.offset = XTENTRYSTART; */
1416 rxtlck->lwm.length = le16_to_cpu(rp->header.nextindex) -
1417 XTENTRYSTART;
1418 }
1419
1420 *rmpp = rmp;
1421 *rbnp = rbn;
1422
1423 ip->i_blocks += LBLK2PBLK(ip->i_sb, lengthPXD(pxd));
1424
1425 jfs_info("xtSplitPage: sp:0x%p rp:0x%p", sp, rp);
1426 return rc;
1427 }
1428
1429
1430 /*
1431 * xtSplitRoot()
1432 *
1433 * function:
1434 * split the full root page into
1435 * original/root/split page and new right page
1436 * i.e., root remains fixed in tree anchor (inode) and
1437 * the root is copied to a single new right child page
1438 * since root page << non-root page, and
1439 * the split root page contains a single entry for the
1440 * new right child page.
1441 *
1442 * parameter:
1443 * int tid,
1444 * struct inode *ip,
1445 * struct xtsplit *split,
1446 * struct metapage **rmpp)
1447 *
1448 * return:
1449 * Pointer to page in which to insert or NULL on error.
1450 */
1451 static int
xtSplitRoot(tid_t tid,struct inode * ip,struct xtsplit * split,struct metapage ** rmpp)1452 xtSplitRoot(tid_t tid,
1453 struct inode *ip, struct xtsplit * split, struct metapage ** rmpp)
1454 {
1455 xtpage_t *sp;
1456 struct metapage *rmp;
1457 xtpage_t *rp;
1458 s64 rbn;
1459 int skip, nextindex;
1460 xad_t *xad;
1461 pxd_t *pxd;
1462 struct pxdlist *pxdlist;
1463 struct tlock *tlck;
1464 struct xtlock *xtlck;
1465
1466 sp = &JFS_IP(ip)->i_xtroot;
1467
1468 INCREMENT(xtStat.split);
1469
1470 /*
1471 * allocate a single (right) child page
1472 */
1473 pxdlist = split->pxdlist;
1474 pxd = &pxdlist->pxd[pxdlist->npxd];
1475 pxdlist->npxd++;
1476 rbn = addressPXD(pxd);
1477 rmp = get_metapage(ip, rbn, PSIZE, 1);
1478 if (rmp == NULL)
1479 return -EIO;
1480
1481 jfs_info("xtSplitRoot: ip:0x%p rmp:0x%p", ip, rmp);
1482
1483 /*
1484 * acquire a transaction lock on the new right page;
1485 *
1486 * action: new page;
1487 */
1488 BT_MARK_DIRTY(rmp, ip);
1489
1490 rp = (xtpage_t *) rmp->data;
1491 rp->header.flag =
1492 (sp->header.flag & BT_LEAF) ? BT_LEAF : BT_INTERNAL;
1493 rp->header.self = *pxd;
1494 rp->header.nextindex = cpu_to_le16(XTENTRYSTART);
1495 rp->header.maxentry = cpu_to_le16(PSIZE >> L2XTSLOTSIZE);
1496
1497 /* initialize sibling pointers */
1498 rp->header.next = 0;
1499 rp->header.prev = 0;
1500
1501 /*
1502 * copy the in-line root page into new right page extent
1503 */
1504 nextindex = le16_to_cpu(sp->header.maxentry);
1505 memmove(&rp->xad[XTENTRYSTART], &sp->xad[XTENTRYSTART],
1506 (nextindex - XTENTRYSTART) << L2XTSLOTSIZE);
1507
1508 /*
1509 * insert the new entry into the new right/child page
1510 * (skip index in the new right page will not change)
1511 */
1512 skip = split->index;
1513 /* if insert into middle, shift right remaining entries */
1514 if (skip != nextindex)
1515 memmove(&rp->xad[skip + 1], &rp->xad[skip],
1516 (nextindex - skip) * sizeof(xad_t));
1517
1518 xad = &rp->xad[skip];
1519 XT_PUTENTRY(xad, split->flag, split->off, split->len, split->addr);
1520
1521 /* update page header */
1522 rp->header.nextindex = cpu_to_le16(nextindex + 1);
1523
1524 if (!test_cflag(COMMIT_Nolink, ip)) {
1525 tlck = txLock(tid, ip, rmp, tlckXTREE | tlckNEW);
1526 xtlck = (struct xtlock *) & tlck->lock;
1527 xtlck->lwm.offset = XTENTRYSTART;
1528 xtlck->lwm.length = le16_to_cpu(rp->header.nextindex) -
1529 XTENTRYSTART;
1530 }
1531
1532 /*
1533 * reset the root
1534 *
1535 * init root with the single entry for the new right page
1536 * set the 1st entry offset to 0, which force the left-most key
1537 * at any level of the tree to be less than any search key.
1538 */
1539 /*
1540 * acquire a transaction lock on the root page (in-memory inode);
1541 *
1542 * action: root split;
1543 */
1544 BT_MARK_DIRTY(split->mp, ip);
1545
1546 xad = &sp->xad[XTENTRYSTART];
1547 XT_PUTENTRY(xad, XAD_NEW, 0, JFS_SBI(ip->i_sb)->nbperpage, rbn);
1548
1549 /* update page header of root */
1550 sp->header.flag &= ~BT_LEAF;
1551 sp->header.flag |= BT_INTERNAL;
1552
1553 sp->header.nextindex = cpu_to_le16(XTENTRYSTART + 1);
1554
1555 if (!test_cflag(COMMIT_Nolink, ip)) {
1556 tlck = txLock(tid, ip, split->mp, tlckXTREE | tlckGROW);
1557 xtlck = (struct xtlock *) & tlck->lock;
1558 xtlck->lwm.offset = XTENTRYSTART;
1559 xtlck->lwm.length = 1;
1560 }
1561
1562 *rmpp = rmp;
1563
1564 ip->i_blocks += LBLK2PBLK(ip->i_sb, lengthPXD(pxd));
1565
1566 jfs_info("xtSplitRoot: sp:0x%p rp:0x%p", sp, rp);
1567 return 0;
1568 }
1569
1570
1571 /*
1572 * xtExtend()
1573 *
1574 * function: extend in-place;
1575 *
1576 * note: existing extent may or may not have been committed.
1577 * caller is responsible for pager buffer cache update, and
1578 * working block allocation map update;
1579 * update pmap: alloc whole extended extent;
1580 */
xtExtend(tid_t tid,struct inode * ip,s64 xoff,s32 xlen,int flag)1581 int xtExtend(tid_t tid, /* transaction id */
1582 struct inode *ip, s64 xoff, /* delta extent offset */
1583 s32 xlen, /* delta extent length */
1584 int flag)
1585 {
1586 int rc = 0;
1587 int cmp;
1588 struct metapage *mp; /* meta-page buffer */
1589 xtpage_t *p; /* base B+-tree index page */
1590 s64 bn;
1591 int index, nextindex, len;
1592 struct btstack btstack; /* traverse stack */
1593 struct xtsplit split; /* split information */
1594 xad_t *xad;
1595 s64 xaddr;
1596 struct tlock *tlck;
1597 struct xtlock *xtlck = 0;
1598
1599 jfs_info("xtExtend: nxoff:0x%lx nxlen:0x%x", (ulong) xoff, xlen);
1600
1601 /* there must exist extent to be extended */
1602 if ((rc = xtSearch(ip, xoff - 1, &cmp, &btstack, XT_INSERT)))
1603 return rc;
1604
1605 /* retrieve search result */
1606 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
1607
1608 if (cmp != 0) {
1609 XT_PUTPAGE(mp);
1610 jfs_error(ip->i_sb, "xtExtend: xtSearch did not find extent");
1611 return -EIO;
1612 }
1613
1614 /* extension must be contiguous */
1615 xad = &p->xad[index];
1616 if ((offsetXAD(xad) + lengthXAD(xad)) != xoff) {
1617 XT_PUTPAGE(mp);
1618 jfs_error(ip->i_sb, "xtExtend: extension is not contiguous");
1619 return -EIO;
1620 }
1621
1622 /*
1623 * acquire a transaction lock on the leaf page;
1624 *
1625 * action: xad insertion/extension;
1626 */
1627 BT_MARK_DIRTY(mp, ip);
1628 if (!test_cflag(COMMIT_Nolink, ip)) {
1629 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
1630 xtlck = (struct xtlock *) & tlck->lock;
1631 }
1632
1633 /* extend will overflow extent ? */
1634 xlen = lengthXAD(xad) + xlen;
1635 if ((len = xlen - MAXXLEN) <= 0)
1636 goto extendOld;
1637
1638 /*
1639 * extent overflow: insert entry for new extent
1640 */
1641 //insertNew:
1642 xoff = offsetXAD(xad) + MAXXLEN;
1643 xaddr = addressXAD(xad) + MAXXLEN;
1644 nextindex = le16_to_cpu(p->header.nextindex);
1645
1646 /*
1647 * if the leaf page is full, insert the new entry and
1648 * propagate up the router entry for the new page from split
1649 *
1650 * The xtSplitUp() will insert the entry and unpin the leaf page.
1651 */
1652 if (nextindex == le16_to_cpu(p->header.maxentry)) {
1653 /* xtSpliUp() unpins leaf pages */
1654 split.mp = mp;
1655 split.index = index + 1;
1656 split.flag = XAD_NEW;
1657 split.off = xoff; /* split offset */
1658 split.len = len;
1659 split.addr = xaddr;
1660 split.pxdlist = NULL;
1661 if ((rc = xtSplitUp(tid, ip, &split, &btstack)))
1662 return rc;
1663
1664 /* get back old page */
1665 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
1666 if (rc)
1667 return rc;
1668 /*
1669 * if leaf root has been split, original root has been
1670 * copied to new child page, i.e., original entry now
1671 * resides on the new child page;
1672 */
1673 if (p->header.flag & BT_INTERNAL) {
1674 ASSERT(p->header.nextindex ==
1675 cpu_to_le16(XTENTRYSTART + 1));
1676 xad = &p->xad[XTENTRYSTART];
1677 bn = addressXAD(xad);
1678 XT_PUTPAGE(mp);
1679
1680 /* get new child page */
1681 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
1682 if (rc)
1683 return rc;
1684
1685 BT_MARK_DIRTY(mp, ip);
1686 if (!test_cflag(COMMIT_Nolink, ip)) {
1687 tlck = txLock(tid, ip, mp, tlckXTREE|tlckGROW);
1688 xtlck = (struct xtlock *) & tlck->lock;
1689 }
1690 }
1691 }
1692 /*
1693 * insert the new entry into the leaf page
1694 */
1695 else {
1696 /* insert the new entry: mark the entry NEW */
1697 xad = &p->xad[index + 1];
1698 XT_PUTENTRY(xad, XAD_NEW, xoff, len, xaddr);
1699
1700 /* advance next available entry index */
1701 p->header.nextindex =
1702 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
1703 }
1704
1705 /* get back old entry */
1706 xad = &p->xad[index];
1707 xlen = MAXXLEN;
1708
1709 /*
1710 * extend old extent
1711 */
1712 extendOld:
1713 XADlength(xad, xlen);
1714 if (!(xad->flag & XAD_NEW))
1715 xad->flag |= XAD_EXTENDED;
1716
1717 if (!test_cflag(COMMIT_Nolink, ip)) {
1718 xtlck->lwm.offset =
1719 (xtlck->lwm.offset) ? min(index,
1720 (int)xtlck->lwm.offset) : index;
1721 xtlck->lwm.length =
1722 le16_to_cpu(p->header.nextindex) - xtlck->lwm.offset;
1723 }
1724
1725 /* unpin the leaf page */
1726 XT_PUTPAGE(mp);
1727
1728 return rc;
1729 }
1730
1731 #ifdef _NOTYET
1732 /*
1733 * xtTailgate()
1734 *
1735 * function: split existing 'tail' extent
1736 * (split offset >= start offset of tail extent), and
1737 * relocate and extend the split tail half;
1738 *
1739 * note: existing extent may or may not have been committed.
1740 * caller is responsible for pager buffer cache update, and
1741 * working block allocation map update;
1742 * update pmap: free old split tail extent, alloc new extent;
1743 */
xtTailgate(tid_t tid,struct inode * ip,s64 xoff,s32 xlen,s64 xaddr,int flag)1744 int xtTailgate(tid_t tid, /* transaction id */
1745 struct inode *ip, s64 xoff, /* split/new extent offset */
1746 s32 xlen, /* new extent length */
1747 s64 xaddr, /* new extent address */
1748 int flag)
1749 {
1750 int rc = 0;
1751 int cmp;
1752 struct metapage *mp; /* meta-page buffer */
1753 xtpage_t *p; /* base B+-tree index page */
1754 s64 bn;
1755 int index, nextindex, llen, rlen;
1756 struct btstack btstack; /* traverse stack */
1757 struct xtsplit split; /* split information */
1758 xad_t *xad;
1759 struct tlock *tlck;
1760 struct xtlock *xtlck = 0;
1761 struct tlock *mtlck;
1762 struct maplock *pxdlock;
1763
1764 /*
1765 printf("xtTailgate: nxoff:0x%lx nxlen:0x%x nxaddr:0x%lx\n",
1766 (ulong)xoff, xlen, (ulong)xaddr);
1767 */
1768
1769 /* there must exist extent to be tailgated */
1770 if ((rc = xtSearch(ip, xoff, &cmp, &btstack, XT_INSERT)))
1771 return rc;
1772
1773 /* retrieve search result */
1774 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
1775
1776 if (cmp != 0) {
1777 XT_PUTPAGE(mp);
1778 jfs_error(ip->i_sb, "xtTailgate: couldn't find extent");
1779 return -EIO;
1780 }
1781
1782 /* entry found must be last entry */
1783 nextindex = le16_to_cpu(p->header.nextindex);
1784 if (index != nextindex - 1) {
1785 XT_PUTPAGE(mp);
1786 jfs_error(ip->i_sb,
1787 "xtTailgate: the entry found is not the last entry");
1788 return -EIO;
1789 }
1790
1791 BT_MARK_DIRTY(mp, ip);
1792 /*
1793 * acquire tlock of the leaf page containing original entry
1794 */
1795 if (!test_cflag(COMMIT_Nolink, ip)) {
1796 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
1797 xtlck = (struct xtlock *) & tlck->lock;
1798 }
1799
1800 /* completely replace extent ? */
1801 xad = &p->xad[index];
1802 /*
1803 printf("xtTailgate: xoff:0x%lx xlen:0x%x xaddr:0x%lx\n",
1804 (ulong)offsetXAD(xad), lengthXAD(xad), (ulong)addressXAD(xad));
1805 */
1806 if ((llen = xoff - offsetXAD(xad)) == 0)
1807 goto updateOld;
1808
1809 /*
1810 * partially replace extent: insert entry for new extent
1811 */
1812 //insertNew:
1813 /*
1814 * if the leaf page is full, insert the new entry and
1815 * propagate up the router entry for the new page from split
1816 *
1817 * The xtSplitUp() will insert the entry and unpin the leaf page.
1818 */
1819 if (nextindex == le16_to_cpu(p->header.maxentry)) {
1820 /* xtSpliUp() unpins leaf pages */
1821 split.mp = mp;
1822 split.index = index + 1;
1823 split.flag = XAD_NEW;
1824 split.off = xoff; /* split offset */
1825 split.len = xlen;
1826 split.addr = xaddr;
1827 split.pxdlist = NULL;
1828 if ((rc = xtSplitUp(tid, ip, &split, &btstack)))
1829 return rc;
1830
1831 /* get back old page */
1832 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
1833 if (rc)
1834 return rc;
1835 /*
1836 * if leaf root has been split, original root has been
1837 * copied to new child page, i.e., original entry now
1838 * resides on the new child page;
1839 */
1840 if (p->header.flag & BT_INTERNAL) {
1841 ASSERT(p->header.nextindex ==
1842 cpu_to_le16(XTENTRYSTART + 1));
1843 xad = &p->xad[XTENTRYSTART];
1844 bn = addressXAD(xad);
1845 XT_PUTPAGE(mp);
1846
1847 /* get new child page */
1848 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
1849 if (rc)
1850 return rc;
1851
1852 BT_MARK_DIRTY(mp, ip);
1853 if (!test_cflag(COMMIT_Nolink, ip)) {
1854 tlck = txLock(tid, ip, mp, tlckXTREE|tlckGROW);
1855 xtlck = (struct xtlock *) & tlck->lock;
1856 }
1857 }
1858 }
1859 /*
1860 * insert the new entry into the leaf page
1861 */
1862 else {
1863 /* insert the new entry: mark the entry NEW */
1864 xad = &p->xad[index + 1];
1865 XT_PUTENTRY(xad, XAD_NEW, xoff, xlen, xaddr);
1866
1867 /* advance next available entry index */
1868 p->header.nextindex =
1869 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
1870 }
1871
1872 /* get back old XAD */
1873 xad = &p->xad[index];
1874
1875 /*
1876 * truncate/relocate old extent at split offset
1877 */
1878 updateOld:
1879 /* update dmap for old/committed/truncated extent */
1880 rlen = lengthXAD(xad) - llen;
1881 if (!(xad->flag & XAD_NEW)) {
1882 /* free from PWMAP at commit */
1883 if (!test_cflag(COMMIT_Nolink, ip)) {
1884 mtlck = txMaplock(tid, ip, tlckMAP);
1885 pxdlock = (struct maplock *) & mtlck->lock;
1886 pxdlock->flag = mlckFREEPXD;
1887 PXDaddress(&pxdlock->pxd, addressXAD(xad) + llen);
1888 PXDlength(&pxdlock->pxd, rlen);
1889 pxdlock->index = 1;
1890 }
1891 } else
1892 /* free from WMAP */
1893 dbFree(ip, addressXAD(xad) + llen, (s64) rlen);
1894
1895 if (llen)
1896 /* truncate */
1897 XADlength(xad, llen);
1898 else
1899 /* replace */
1900 XT_PUTENTRY(xad, XAD_NEW, xoff, xlen, xaddr);
1901
1902 if (!test_cflag(COMMIT_Nolink, ip)) {
1903 xtlck->lwm.offset = (xtlck->lwm.offset) ?
1904 min(index, (int)xtlck->lwm.offset) : index;
1905 xtlck->lwm.length = le16_to_cpu(p->header.nextindex) -
1906 xtlck->lwm.offset;
1907 }
1908
1909 /* unpin the leaf page */
1910 XT_PUTPAGE(mp);
1911
1912 return rc;
1913 }
1914 #endif /* _NOTYET */
1915
1916 /*
1917 * xtUpdate()
1918 *
1919 * function: update XAD;
1920 *
1921 * update extent for allocated_but_not_recorded or
1922 * compressed extent;
1923 *
1924 * parameter:
1925 * nxad - new XAD;
1926 * logical extent of the specified XAD must be completely
1927 * contained by an existing XAD;
1928 */
xtUpdate(tid_t tid,struct inode * ip,xad_t * nxad)1929 int xtUpdate(tid_t tid, struct inode *ip, xad_t * nxad)
1930 { /* new XAD */
1931 int rc = 0;
1932 int cmp;
1933 struct metapage *mp; /* meta-page buffer */
1934 xtpage_t *p; /* base B+-tree index page */
1935 s64 bn;
1936 int index0, index, newindex, nextindex;
1937 struct btstack btstack; /* traverse stack */
1938 struct xtsplit split; /* split information */
1939 xad_t *xad, *lxad, *rxad;
1940 int xflag;
1941 s64 nxoff, xoff;
1942 int nxlen, xlen, lxlen, rxlen;
1943 s64 nxaddr, xaddr;
1944 struct tlock *tlck;
1945 struct xtlock *xtlck = 0;
1946 int newpage = 0;
1947
1948 /* there must exist extent to be tailgated */
1949 nxoff = offsetXAD(nxad);
1950 nxlen = lengthXAD(nxad);
1951 nxaddr = addressXAD(nxad);
1952
1953 if ((rc = xtSearch(ip, nxoff, &cmp, &btstack, XT_INSERT)))
1954 return rc;
1955
1956 /* retrieve search result */
1957 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index0);
1958
1959 if (cmp != 0) {
1960 XT_PUTPAGE(mp);
1961 jfs_error(ip->i_sb, "xtUpdate: Could not find extent");
1962 return -EIO;
1963 }
1964
1965 BT_MARK_DIRTY(mp, ip);
1966 /*
1967 * acquire tlock of the leaf page containing original entry
1968 */
1969 if (!test_cflag(COMMIT_Nolink, ip)) {
1970 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
1971 xtlck = (struct xtlock *) & tlck->lock;
1972 }
1973
1974 xad = &p->xad[index0];
1975 xflag = xad->flag;
1976 xoff = offsetXAD(xad);
1977 xlen = lengthXAD(xad);
1978 xaddr = addressXAD(xad);
1979
1980 /* nXAD must be completely contained within XAD */
1981 if ((xoff > nxoff) ||
1982 (nxoff + nxlen > xoff + xlen)) {
1983 XT_PUTPAGE(mp);
1984 jfs_error(ip->i_sb,
1985 "xtUpdate: nXAD in not completely contained within XAD");
1986 return -EIO;
1987 }
1988
1989 index = index0;
1990 newindex = index + 1;
1991 nextindex = le16_to_cpu(p->header.nextindex);
1992
1993 #ifdef _JFS_WIP_NOCOALESCE
1994 if (xoff < nxoff)
1995 goto updateRight;
1996
1997 /*
1998 * replace XAD with nXAD
1999 */
2000 replace: /* (nxoff == xoff) */
2001 if (nxlen == xlen) {
2002 /* replace XAD with nXAD:recorded */
2003 *xad = *nxad;
2004 xad->flag = xflag & ~XAD_NOTRECORDED;
2005
2006 goto out;
2007 } else /* (nxlen < xlen) */
2008 goto updateLeft;
2009 #endif /* _JFS_WIP_NOCOALESCE */
2010
2011 /* #ifdef _JFS_WIP_COALESCE */
2012 if (xoff < nxoff)
2013 goto coalesceRight;
2014
2015 /*
2016 * coalesce with left XAD
2017 */
2018 //coalesceLeft: /* (xoff == nxoff) */
2019 /* is XAD first entry of page ? */
2020 if (index == XTENTRYSTART)
2021 goto replace;
2022
2023 /* is nXAD logically and physically contiguous with lXAD ? */
2024 lxad = &p->xad[index - 1];
2025 lxlen = lengthXAD(lxad);
2026 if (!(lxad->flag & XAD_NOTRECORDED) &&
2027 (nxoff == offsetXAD(lxad) + lxlen) &&
2028 (nxaddr == addressXAD(lxad) + lxlen) &&
2029 (lxlen + nxlen < MAXXLEN)) {
2030 /* extend right lXAD */
2031 index0 = index - 1;
2032 XADlength(lxad, lxlen + nxlen);
2033
2034 /* If we just merged two extents together, need to make sure the
2035 * right extent gets logged. If the left one is marked XAD_NEW,
2036 * then we know it will be logged. Otherwise, mark as
2037 * XAD_EXTENDED
2038 */
2039 if (!(lxad->flag & XAD_NEW))
2040 lxad->flag |= XAD_EXTENDED;
2041
2042 if (xlen > nxlen) {
2043 /* truncate XAD */
2044 XADoffset(xad, xoff + nxlen);
2045 XADlength(xad, xlen - nxlen);
2046 XADaddress(xad, xaddr + nxlen);
2047 goto out;
2048 } else { /* (xlen == nxlen) */
2049
2050 /* remove XAD */
2051 if (index < nextindex - 1)
2052 memmove(&p->xad[index], &p->xad[index + 1],
2053 (nextindex - index -
2054 1) << L2XTSLOTSIZE);
2055
2056 p->header.nextindex =
2057 cpu_to_le16(le16_to_cpu(p->header.nextindex) -
2058 1);
2059
2060 index = index0;
2061 newindex = index + 1;
2062 nextindex = le16_to_cpu(p->header.nextindex);
2063 xoff = nxoff = offsetXAD(lxad);
2064 xlen = nxlen = lxlen + nxlen;
2065 xaddr = nxaddr = addressXAD(lxad);
2066 goto coalesceRight;
2067 }
2068 }
2069
2070 /*
2071 * replace XAD with nXAD
2072 */
2073 replace: /* (nxoff == xoff) */
2074 if (nxlen == xlen) {
2075 /* replace XAD with nXAD:recorded */
2076 *xad = *nxad;
2077 xad->flag = xflag & ~XAD_NOTRECORDED;
2078
2079 goto coalesceRight;
2080 } else /* (nxlen < xlen) */
2081 goto updateLeft;
2082
2083 /*
2084 * coalesce with right XAD
2085 */
2086 coalesceRight: /* (xoff <= nxoff) */
2087 /* is XAD last entry of page ? */
2088 if (newindex == nextindex) {
2089 if (xoff == nxoff)
2090 goto out;
2091 goto updateRight;
2092 }
2093
2094 /* is nXAD logically and physically contiguous with rXAD ? */
2095 rxad = &p->xad[index + 1];
2096 rxlen = lengthXAD(rxad);
2097 if (!(rxad->flag & XAD_NOTRECORDED) &&
2098 (nxoff + nxlen == offsetXAD(rxad)) &&
2099 (nxaddr + nxlen == addressXAD(rxad)) &&
2100 (rxlen + nxlen < MAXXLEN)) {
2101 /* extend left rXAD */
2102 XADoffset(rxad, nxoff);
2103 XADlength(rxad, rxlen + nxlen);
2104 XADaddress(rxad, nxaddr);
2105
2106 /* If we just merged two extents together, need to make sure
2107 * the left extent gets logged. If the right one is marked
2108 * XAD_NEW, then we know it will be logged. Otherwise, mark as
2109 * XAD_EXTENDED
2110 */
2111 if (!(rxad->flag & XAD_NEW))
2112 rxad->flag |= XAD_EXTENDED;
2113
2114 if (xlen > nxlen)
2115 /* truncate XAD */
2116 XADlength(xad, xlen - nxlen);
2117 else { /* (xlen == nxlen) */
2118
2119 /* remove XAD */
2120 memmove(&p->xad[index], &p->xad[index + 1],
2121 (nextindex - index - 1) << L2XTSLOTSIZE);
2122
2123 p->header.nextindex =
2124 cpu_to_le16(le16_to_cpu(p->header.nextindex) -
2125 1);
2126 }
2127
2128 goto out;
2129 } else if (xoff == nxoff)
2130 goto out;
2131
2132 if (xoff >= nxoff) {
2133 XT_PUTPAGE(mp);
2134 jfs_error(ip->i_sb, "xtUpdate: xoff >= nxoff");
2135 return -EIO;
2136 }
2137 /* #endif _JFS_WIP_COALESCE */
2138
2139 /*
2140 * split XAD into (lXAD, nXAD):
2141 *
2142 * |---nXAD--->
2143 * --|----------XAD----------|--
2144 * |-lXAD-|
2145 */
2146 updateRight: /* (xoff < nxoff) */
2147 /* truncate old XAD as lXAD:not_recorded */
2148 xad = &p->xad[index];
2149 XADlength(xad, nxoff - xoff);
2150
2151 /* insert nXAD:recorded */
2152 if (nextindex == le16_to_cpu(p->header.maxentry)) {
2153
2154 /* xtSpliUp() unpins leaf pages */
2155 split.mp = mp;
2156 split.index = newindex;
2157 split.flag = xflag & ~XAD_NOTRECORDED;
2158 split.off = nxoff;
2159 split.len = nxlen;
2160 split.addr = nxaddr;
2161 split.pxdlist = NULL;
2162 if ((rc = xtSplitUp(tid, ip, &split, &btstack)))
2163 return rc;
2164
2165 /* get back old page */
2166 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2167 if (rc)
2168 return rc;
2169 /*
2170 * if leaf root has been split, original root has been
2171 * copied to new child page, i.e., original entry now
2172 * resides on the new child page;
2173 */
2174 if (p->header.flag & BT_INTERNAL) {
2175 ASSERT(p->header.nextindex ==
2176 cpu_to_le16(XTENTRYSTART + 1));
2177 xad = &p->xad[XTENTRYSTART];
2178 bn = addressXAD(xad);
2179 XT_PUTPAGE(mp);
2180
2181 /* get new child page */
2182 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2183 if (rc)
2184 return rc;
2185
2186 BT_MARK_DIRTY(mp, ip);
2187 if (!test_cflag(COMMIT_Nolink, ip)) {
2188 tlck = txLock(tid, ip, mp, tlckXTREE|tlckGROW);
2189 xtlck = (struct xtlock *) & tlck->lock;
2190 }
2191 } else {
2192 /* is nXAD on new page ? */
2193 if (newindex >
2194 (le16_to_cpu(p->header.maxentry) >> 1)) {
2195 newindex =
2196 newindex -
2197 le16_to_cpu(p->header.nextindex) +
2198 XTENTRYSTART;
2199 newpage = 1;
2200 }
2201 }
2202 } else {
2203 /* if insert into middle, shift right remaining entries */
2204 if (newindex < nextindex)
2205 memmove(&p->xad[newindex + 1], &p->xad[newindex],
2206 (nextindex - newindex) << L2XTSLOTSIZE);
2207
2208 /* insert the entry */
2209 xad = &p->xad[newindex];
2210 *xad = *nxad;
2211 xad->flag = xflag & ~XAD_NOTRECORDED;
2212
2213 /* advance next available entry index. */
2214 p->header.nextindex =
2215 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
2216 }
2217
2218 /*
2219 * does nXAD force 3-way split ?
2220 *
2221 * |---nXAD--->|
2222 * --|----------XAD-------------|--
2223 * |-lXAD-| |-rXAD -|
2224 */
2225 if (nxoff + nxlen == xoff + xlen)
2226 goto out;
2227
2228 /* reorient nXAD as XAD for further split XAD into (nXAD, rXAD) */
2229 if (newpage) {
2230 /* close out old page */
2231 if (!test_cflag(COMMIT_Nolink, ip)) {
2232 xtlck->lwm.offset = (xtlck->lwm.offset) ?
2233 min(index0, (int)xtlck->lwm.offset) : index0;
2234 xtlck->lwm.length =
2235 le16_to_cpu(p->header.nextindex) -
2236 xtlck->lwm.offset;
2237 }
2238
2239 bn = le64_to_cpu(p->header.next);
2240 XT_PUTPAGE(mp);
2241
2242 /* get new right page */
2243 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2244 if (rc)
2245 return rc;
2246
2247 BT_MARK_DIRTY(mp, ip);
2248 if (!test_cflag(COMMIT_Nolink, ip)) {
2249 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
2250 xtlck = (struct xtlock *) & tlck->lock;
2251 }
2252
2253 index0 = index = newindex;
2254 } else
2255 index++;
2256
2257 newindex = index + 1;
2258 nextindex = le16_to_cpu(p->header.nextindex);
2259 xlen = xlen - (nxoff - xoff);
2260 xoff = nxoff;
2261 xaddr = nxaddr;
2262
2263 /* recompute split pages */
2264 if (nextindex == le16_to_cpu(p->header.maxentry)) {
2265 XT_PUTPAGE(mp);
2266
2267 if ((rc = xtSearch(ip, nxoff, &cmp, &btstack, XT_INSERT)))
2268 return rc;
2269
2270 /* retrieve search result */
2271 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index0);
2272
2273 if (cmp != 0) {
2274 XT_PUTPAGE(mp);
2275 jfs_error(ip->i_sb, "xtUpdate: xtSearch failed");
2276 return -EIO;
2277 }
2278
2279 if (index0 != index) {
2280 XT_PUTPAGE(mp);
2281 jfs_error(ip->i_sb,
2282 "xtUpdate: unexpected value of index");
2283 return -EIO;
2284 }
2285 }
2286
2287 /*
2288 * split XAD into (nXAD, rXAD)
2289 *
2290 * ---nXAD---|
2291 * --|----------XAD----------|--
2292 * |-rXAD-|
2293 */
2294 updateLeft: /* (nxoff == xoff) && (nxlen < xlen) */
2295 /* update old XAD with nXAD:recorded */
2296 xad = &p->xad[index];
2297 *xad = *nxad;
2298 xad->flag = xflag & ~XAD_NOTRECORDED;
2299
2300 /* insert rXAD:not_recorded */
2301 xoff = xoff + nxlen;
2302 xlen = xlen - nxlen;
2303 xaddr = xaddr + nxlen;
2304 if (nextindex == le16_to_cpu(p->header.maxentry)) {
2305 /*
2306 printf("xtUpdate.updateLeft.split p:0x%p\n", p);
2307 */
2308 /* xtSpliUp() unpins leaf pages */
2309 split.mp = mp;
2310 split.index = newindex;
2311 split.flag = xflag;
2312 split.off = xoff;
2313 split.len = xlen;
2314 split.addr = xaddr;
2315 split.pxdlist = NULL;
2316 if ((rc = xtSplitUp(tid, ip, &split, &btstack)))
2317 return rc;
2318
2319 /* get back old page */
2320 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2321 if (rc)
2322 return rc;
2323
2324 /*
2325 * if leaf root has been split, original root has been
2326 * copied to new child page, i.e., original entry now
2327 * resides on the new child page;
2328 */
2329 if (p->header.flag & BT_INTERNAL) {
2330 ASSERT(p->header.nextindex ==
2331 cpu_to_le16(XTENTRYSTART + 1));
2332 xad = &p->xad[XTENTRYSTART];
2333 bn = addressXAD(xad);
2334 XT_PUTPAGE(mp);
2335
2336 /* get new child page */
2337 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2338 if (rc)
2339 return rc;
2340
2341 BT_MARK_DIRTY(mp, ip);
2342 if (!test_cflag(COMMIT_Nolink, ip)) {
2343 tlck = txLock(tid, ip, mp, tlckXTREE|tlckGROW);
2344 xtlck = (struct xtlock *) & tlck->lock;
2345 }
2346 }
2347 } else {
2348 /* if insert into middle, shift right remaining entries */
2349 if (newindex < nextindex)
2350 memmove(&p->xad[newindex + 1], &p->xad[newindex],
2351 (nextindex - newindex) << L2XTSLOTSIZE);
2352
2353 /* insert the entry */
2354 xad = &p->xad[newindex];
2355 XT_PUTENTRY(xad, xflag, xoff, xlen, xaddr);
2356
2357 /* advance next available entry index. */
2358 p->header.nextindex =
2359 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
2360 }
2361
2362 out:
2363 if (!test_cflag(COMMIT_Nolink, ip)) {
2364 xtlck->lwm.offset = (xtlck->lwm.offset) ?
2365 min(index0, (int)xtlck->lwm.offset) : index0;
2366 xtlck->lwm.length = le16_to_cpu(p->header.nextindex) -
2367 xtlck->lwm.offset;
2368 }
2369
2370 /* unpin the leaf page */
2371 XT_PUTPAGE(mp);
2372
2373 return rc;
2374 }
2375
2376
2377 /*
2378 * xtAppend()
2379 *
2380 * function: grow in append mode from contiguous region specified ;
2381 *
2382 * parameter:
2383 * tid - transaction id;
2384 * ip - file object;
2385 * xflag - extent flag:
2386 * xoff - extent offset;
2387 * maxblocks - max extent length;
2388 * xlen - extent length (in/out);
2389 * xaddrp - extent address pointer (in/out):
2390 * flag -
2391 *
2392 * return:
2393 */
xtAppend(tid_t tid,struct inode * ip,int xflag,s64 xoff,s32 maxblocks,s32 * xlenp,s64 * xaddrp,int flag)2394 int xtAppend(tid_t tid, /* transaction id */
2395 struct inode *ip, int xflag, s64 xoff, s32 maxblocks,
2396 s32 * xlenp, /* (in/out) */
2397 s64 * xaddrp, /* (in/out) */
2398 int flag)
2399 {
2400 int rc = 0;
2401 struct metapage *mp; /* meta-page buffer */
2402 xtpage_t *p; /* base B+-tree index page */
2403 s64 bn, xaddr;
2404 int index, nextindex;
2405 struct btstack btstack; /* traverse stack */
2406 struct xtsplit split; /* split information */
2407 xad_t *xad;
2408 int cmp;
2409 struct tlock *tlck;
2410 struct xtlock *xtlck;
2411 int nsplit, nblocks, xlen;
2412 struct pxdlist pxdlist;
2413 pxd_t *pxd;
2414
2415 xaddr = *xaddrp;
2416 xlen = *xlenp;
2417 jfs_info("xtAppend: xoff:0x%lx maxblocks:%d xlen:%d xaddr:0x%lx",
2418 (ulong) xoff, maxblocks, xlen, (ulong) xaddr);
2419
2420 /*
2421 * search for the entry location at which to insert:
2422 *
2423 * xtFastSearch() and xtSearch() both returns (leaf page
2424 * pinned, index at which to insert).
2425 * n.b. xtSearch() may return index of maxentry of
2426 * the full page.
2427 */
2428 if ((rc = xtSearch(ip, xoff, &cmp, &btstack, XT_INSERT)))
2429 return rc;
2430
2431 /* retrieve search result */
2432 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2433
2434 if (cmp == 0) {
2435 rc = -EEXIST;
2436 goto out;
2437 }
2438 //insert:
2439 /*
2440 * insert entry for new extent
2441 */
2442 xflag |= XAD_NEW;
2443
2444 /*
2445 * if the leaf page is full, split the page and
2446 * propagate up the router entry for the new page from split
2447 *
2448 * The xtSplitUp() will insert the entry and unpin the leaf page.
2449 */
2450 nextindex = le16_to_cpu(p->header.nextindex);
2451 if (nextindex < le16_to_cpu(p->header.maxentry))
2452 goto insertLeaf;
2453
2454 /*
2455 * allocate new index blocks to cover index page split(s)
2456 */
2457 nsplit = btstack.nsplit;
2458 split.pxdlist = &pxdlist;
2459 pxdlist.maxnpxd = pxdlist.npxd = 0;
2460 pxd = &pxdlist.pxd[0];
2461 nblocks = JFS_SBI(ip->i_sb)->nbperpage;
2462 for (; nsplit > 0; nsplit--, pxd++, xaddr += nblocks, maxblocks -= nblocks) {
2463 if ((rc = dbAllocBottomUp(ip, xaddr, (s64) nblocks)) == 0) {
2464 PXDaddress(pxd, xaddr);
2465 PXDlength(pxd, nblocks);
2466
2467 pxdlist.maxnpxd++;
2468
2469 continue;
2470 }
2471
2472 /* undo allocation */
2473
2474 goto out;
2475 }
2476
2477 xlen = min(xlen, maxblocks);
2478
2479 /*
2480 * allocate data extent requested
2481 */
2482 if ((rc = dbAllocBottomUp(ip, xaddr, (s64) xlen)))
2483 goto out;
2484
2485 split.mp = mp;
2486 split.index = index;
2487 split.flag = xflag;
2488 split.off = xoff;
2489 split.len = xlen;
2490 split.addr = xaddr;
2491 if ((rc = xtSplitUp(tid, ip, &split, &btstack))) {
2492 /* undo data extent allocation */
2493 dbFree(ip, *xaddrp, (s64) * xlenp);
2494
2495 return rc;
2496 }
2497
2498 *xaddrp = xaddr;
2499 *xlenp = xlen;
2500 return 0;
2501
2502 /*
2503 * insert the new entry into the leaf page
2504 */
2505 insertLeaf:
2506 /*
2507 * allocate data extent requested
2508 */
2509 if ((rc = dbAllocBottomUp(ip, xaddr, (s64) xlen)))
2510 goto out;
2511
2512 BT_MARK_DIRTY(mp, ip);
2513 /*
2514 * acquire a transaction lock on the leaf page;
2515 *
2516 * action: xad insertion/extension;
2517 */
2518 tlck = txLock(tid, ip, mp, tlckXTREE | tlckGROW);
2519 xtlck = (struct xtlock *) & tlck->lock;
2520
2521 /* insert the new entry: mark the entry NEW */
2522 xad = &p->xad[index];
2523 XT_PUTENTRY(xad, xflag, xoff, xlen, xaddr);
2524
2525 /* advance next available entry index */
2526 p->header.nextindex =
2527 cpu_to_le16(le16_to_cpu(p->header.nextindex) + 1);
2528
2529 xtlck->lwm.offset =
2530 (xtlck->lwm.offset) ? min(index,(int) xtlck->lwm.offset) : index;
2531 xtlck->lwm.length = le16_to_cpu(p->header.nextindex) -
2532 xtlck->lwm.offset;
2533
2534 *xaddrp = xaddr;
2535 *xlenp = xlen;
2536
2537 out:
2538 /* unpin the leaf page */
2539 XT_PUTPAGE(mp);
2540
2541 return rc;
2542 }
2543 #ifdef _STILL_TO_PORT
2544
2545 /* - TBD for defragmentaion/reorganization -
2546 *
2547 * xtDelete()
2548 *
2549 * function:
2550 * delete the entry with the specified key.
2551 *
2552 * N.B.: whole extent of the entry is assumed to be deleted.
2553 *
2554 * parameter:
2555 *
2556 * return:
2557 * ENOENT: if the entry is not found.
2558 *
2559 * exception:
2560 */
xtDelete(tid_t tid,struct inode * ip,s64 xoff,s32 xlen,int flag)2561 int xtDelete(tid_t tid, struct inode *ip, s64 xoff, s32 xlen, int flag)
2562 {
2563 int rc = 0;
2564 struct btstack btstack;
2565 int cmp;
2566 s64 bn;
2567 struct metapage *mp;
2568 xtpage_t *p;
2569 int index, nextindex;
2570 struct tlock *tlck;
2571 struct xtlock *xtlck;
2572
2573 /*
2574 * find the matching entry; xtSearch() pins the page
2575 */
2576 if ((rc = xtSearch(ip, xoff, &cmp, &btstack, 0)))
2577 return rc;
2578
2579 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2580 if (cmp) {
2581 /* unpin the leaf page */
2582 XT_PUTPAGE(mp);
2583 return -ENOENT;
2584 }
2585
2586 /*
2587 * delete the entry from the leaf page
2588 */
2589 nextindex = le16_to_cpu(p->header.nextindex);
2590 p->header.nextindex =
2591 cpu_to_le16(le16_to_cpu(p->header.nextindex) - 1);
2592
2593 /*
2594 * if the leaf page bocome empty, free the page
2595 */
2596 if (p->header.nextindex == cpu_to_le16(XTENTRYSTART))
2597 return (xtDeleteUp(tid, ip, mp, p, &btstack));
2598
2599 BT_MARK_DIRTY(mp, ip);
2600 /*
2601 * acquire a transaction lock on the leaf page;
2602 *
2603 * action:xad deletion;
2604 */
2605 tlck = txLock(tid, ip, mp, tlckXTREE);
2606 xtlck = (struct xtlock *) & tlck->lock;
2607 xtlck->lwm.offset =
2608 (xtlck->lwm.offset) ? min(index, xtlck->lwm.offset) : index;
2609
2610 /* if delete from middle, shift left/compact the remaining entries */
2611 if (index < nextindex - 1)
2612 memmove(&p->xad[index], &p->xad[index + 1],
2613 (nextindex - index - 1) * sizeof(xad_t));
2614
2615 XT_PUTPAGE(mp);
2616
2617 return 0;
2618 }
2619
2620
2621 /* - TBD for defragmentaion/reorganization -
2622 *
2623 * xtDeleteUp()
2624 *
2625 * function:
2626 * free empty pages as propagating deletion up the tree
2627 *
2628 * parameter:
2629 *
2630 * return:
2631 */
2632 static int
xtDeleteUp(tid_t tid,struct inode * ip,struct metapage * fmp,xtpage_t * fp,struct btstack * btstack)2633 xtDeleteUp(tid_t tid, struct inode *ip,
2634 struct metapage * fmp, xtpage_t * fp, struct btstack * btstack)
2635 {
2636 int rc = 0;
2637 struct metapage *mp;
2638 xtpage_t *p;
2639 int index, nextindex;
2640 s64 xaddr;
2641 int xlen;
2642 struct btframe *parent;
2643 struct tlock *tlck;
2644 struct xtlock *xtlck;
2645
2646 /*
2647 * keep root leaf page which has become empty
2648 */
2649 if (fp->header.flag & BT_ROOT) {
2650 /* keep the root page */
2651 fp->header.flag &= ~BT_INTERNAL;
2652 fp->header.flag |= BT_LEAF;
2653 fp->header.nextindex = cpu_to_le16(XTENTRYSTART);
2654
2655 /* XT_PUTPAGE(fmp); */
2656
2657 return 0;
2658 }
2659
2660 /*
2661 * free non-root leaf page
2662 */
2663 if ((rc = xtRelink(tid, ip, fp))) {
2664 XT_PUTPAGE(fmp);
2665 return rc;
2666 }
2667
2668 xaddr = addressPXD(&fp->header.self);
2669 xlen = lengthPXD(&fp->header.self);
2670 /* free the page extent */
2671 dbFree(ip, xaddr, (s64) xlen);
2672
2673 /* free the buffer page */
2674 discard_metapage(fmp);
2675
2676 /*
2677 * propagate page deletion up the index tree
2678 *
2679 * If the delete from the parent page makes it empty,
2680 * continue all the way up the tree.
2681 * stop if the root page is reached (which is never deleted) or
2682 * if the entry deletion does not empty the page.
2683 */
2684 while ((parent = BT_POP(btstack)) != NULL) {
2685 /* get/pin the parent page <sp> */
2686 XT_GETPAGE(ip, parent->bn, mp, PSIZE, p, rc);
2687 if (rc)
2688 return rc;
2689
2690 index = parent->index;
2691
2692 /* delete the entry for the freed child page from parent.
2693 */
2694 nextindex = le16_to_cpu(p->header.nextindex);
2695
2696 /*
2697 * the parent has the single entry being deleted:
2698 * free the parent page which has become empty.
2699 */
2700 if (nextindex == 1) {
2701 if (p->header.flag & BT_ROOT) {
2702 /* keep the root page */
2703 p->header.flag &= ~BT_INTERNAL;
2704 p->header.flag |= BT_LEAF;
2705 p->header.nextindex =
2706 cpu_to_le16(XTENTRYSTART);
2707
2708 /* XT_PUTPAGE(mp); */
2709
2710 break;
2711 } else {
2712 /* free the parent page */
2713 if ((rc = xtRelink(tid, ip, p)))
2714 return rc;
2715
2716 xaddr = addressPXD(&p->header.self);
2717 /* free the page extent */
2718 dbFree(ip, xaddr,
2719 (s64) JFS_SBI(ip->i_sb)->nbperpage);
2720
2721 /* unpin/free the buffer page */
2722 discard_metapage(mp);
2723
2724 /* propagate up */
2725 continue;
2726 }
2727 }
2728 /*
2729 * the parent has other entries remaining:
2730 * delete the router entry from the parent page.
2731 */
2732 else {
2733 BT_MARK_DIRTY(mp, ip);
2734 /*
2735 * acquire a transaction lock on the leaf page;
2736 *
2737 * action:xad deletion;
2738 */
2739 tlck = txLock(tid, ip, mp, tlckXTREE);
2740 xtlck = (struct xtlock *) & tlck->lock;
2741 xtlck->lwm.offset =
2742 (xtlck->lwm.offset) ? min(index,
2743 xtlck->lwm.
2744 offset) : index;
2745
2746 /* if delete from middle,
2747 * shift left/compact the remaining entries in the page
2748 */
2749 if (index < nextindex - 1)
2750 memmove(&p->xad[index], &p->xad[index + 1],
2751 (nextindex - index -
2752 1) << L2XTSLOTSIZE);
2753
2754 p->header.nextindex =
2755 cpu_to_le16(le16_to_cpu(p->header.nextindex) -
2756 1);
2757 jfs_info("xtDeleteUp(entry): 0x%lx[%d]",
2758 (ulong) parent->bn, index);
2759 }
2760
2761 /* unpin the parent page */
2762 XT_PUTPAGE(mp);
2763
2764 /* exit propagation up */
2765 break;
2766 }
2767
2768 return 0;
2769 }
2770
2771
2772 /*
2773 * NAME: xtRelocate()
2774 *
2775 * FUNCTION: relocate xtpage or data extent of regular file;
2776 * This function is mainly used by defragfs utility.
2777 *
2778 * NOTE: This routine does not have the logic to handle
2779 * uncommitted allocated extent. The caller should call
2780 * txCommit() to commit all the allocation before call
2781 * this routine.
2782 */
2783 int
xtRelocate(tid_t tid,struct inode * ip,xad_t * oxad,s64 nxaddr,int xtype)2784 xtRelocate(tid_t tid, struct inode * ip, xad_t * oxad, /* old XAD */
2785 s64 nxaddr, /* new xaddr */
2786 int xtype)
2787 { /* extent type: XTPAGE or DATAEXT */
2788 int rc = 0;
2789 struct tblock *tblk;
2790 struct tlock *tlck;
2791 struct xtlock *xtlck;
2792 struct metapage *mp, *pmp, *lmp, *rmp; /* meta-page buffer */
2793 xtpage_t *p, *pp, *rp, *lp; /* base B+-tree index page */
2794 xad_t *xad;
2795 pxd_t *pxd;
2796 s64 xoff, xsize;
2797 int xlen;
2798 s64 oxaddr, sxaddr, dxaddr, nextbn, prevbn;
2799 cbuf_t *cp;
2800 s64 offset, nbytes, nbrd, pno;
2801 int nb, npages, nblks;
2802 s64 bn;
2803 int cmp;
2804 int index;
2805 struct pxd_lock *pxdlock;
2806 struct btstack btstack; /* traverse stack */
2807
2808 xtype = xtype & EXTENT_TYPE;
2809
2810 xoff = offsetXAD(oxad);
2811 oxaddr = addressXAD(oxad);
2812 xlen = lengthXAD(oxad);
2813
2814 /* validate extent offset */
2815 offset = xoff << JFS_SBI(ip->i_sb)->l2bsize;
2816 if (offset >= ip->i_size)
2817 return -ESTALE; /* stale extent */
2818
2819 jfs_info("xtRelocate: xtype:%d xoff:0x%lx xlen:0x%x xaddr:0x%lx:0x%lx",
2820 xtype, (ulong) xoff, xlen, (ulong) oxaddr, (ulong) nxaddr);
2821
2822 /*
2823 * 1. get and validate the parent xtpage/xad entry
2824 * covering the source extent to be relocated;
2825 */
2826 if (xtype == DATAEXT) {
2827 /* search in leaf entry */
2828 rc = xtSearch(ip, xoff, &cmp, &btstack, 0);
2829 if (rc)
2830 return rc;
2831
2832 /* retrieve search result */
2833 XT_GETSEARCH(ip, btstack.top, bn, pmp, pp, index);
2834
2835 if (cmp) {
2836 XT_PUTPAGE(pmp);
2837 return -ESTALE;
2838 }
2839
2840 /* validate for exact match with a single entry */
2841 xad = &pp->xad[index];
2842 if (addressXAD(xad) != oxaddr || lengthXAD(xad) != xlen) {
2843 XT_PUTPAGE(pmp);
2844 return -ESTALE;
2845 }
2846 } else { /* (xtype == XTPAGE) */
2847
2848 /* search in internal entry */
2849 rc = xtSearchNode(ip, oxad, &cmp, &btstack, 0);
2850 if (rc)
2851 return rc;
2852
2853 /* retrieve search result */
2854 XT_GETSEARCH(ip, btstack.top, bn, pmp, pp, index);
2855
2856 if (cmp) {
2857 XT_PUTPAGE(pmp);
2858 return -ESTALE;
2859 }
2860
2861 /* xtSearchNode() validated for exact match with a single entry
2862 */
2863 xad = &pp->xad[index];
2864 }
2865 jfs_info("xtRelocate: parent xad entry validated.");
2866
2867 /*
2868 * 2. relocate the extent
2869 */
2870 if (xtype == DATAEXT) {
2871 /* if the extent is allocated-but-not-recorded
2872 * there is no real data to be moved in this extent,
2873 */
2874 if (xad->flag & XAD_NOTRECORDED)
2875 goto out;
2876 else
2877 /* release xtpage for cmRead()/xtLookup() */
2878 XT_PUTPAGE(pmp);
2879
2880 /*
2881 * cmRelocate()
2882 *
2883 * copy target data pages to be relocated;
2884 *
2885 * data extent must start at page boundary and
2886 * multiple of page size (except the last data extent);
2887 * read in each page of the source data extent into cbuf,
2888 * update the cbuf extent descriptor of the page to be
2889 * homeward bound to new dst data extent
2890 * copy the data from the old extent to new extent.
2891 * copy is essential for compressed files to avoid problems
2892 * that can arise if there was a change in compression
2893 * algorithms.
2894 * it is a good strategy because it may disrupt cache
2895 * policy to keep the pages in memory afterwards.
2896 */
2897 offset = xoff << JFS_SBI(ip->i_sb)->l2bsize;
2898 assert((offset & CM_OFFSET) == 0);
2899 nbytes = xlen << JFS_SBI(ip->i_sb)->l2bsize;
2900 pno = offset >> CM_L2BSIZE;
2901 npages = (nbytes + (CM_BSIZE - 1)) >> CM_L2BSIZE;
2902 /*
2903 npages = ((offset + nbytes - 1) >> CM_L2BSIZE) -
2904 (offset >> CM_L2BSIZE) + 1;
2905 */
2906 sxaddr = oxaddr;
2907 dxaddr = nxaddr;
2908
2909 /* process the request one cache buffer at a time */
2910 for (nbrd = 0; nbrd < nbytes; nbrd += nb,
2911 offset += nb, pno++, npages--) {
2912 /* compute page size */
2913 nb = min(nbytes - nbrd, CM_BSIZE);
2914
2915 /* get the cache buffer of the page */
2916 if (rc = cmRead(ip, offset, npages, &cp))
2917 break;
2918
2919 assert(addressPXD(&cp->cm_pxd) == sxaddr);
2920 assert(!cp->cm_modified);
2921
2922 /* bind buffer with the new extent address */
2923 nblks = nb >> JFS_IP(ip->i_sb)->l2bsize;
2924 cmSetXD(ip, cp, pno, dxaddr, nblks);
2925
2926 /* release the cbuf, mark it as modified */
2927 cmPut(cp, TRUE);
2928
2929 dxaddr += nblks;
2930 sxaddr += nblks;
2931 }
2932
2933 /* get back parent page */
2934 if ((rc = xtSearch(ip, xoff, &cmp, &btstack, 0)))
2935 return rc;
2936
2937 XT_GETSEARCH(ip, btstack.top, bn, pmp, pp, index);
2938 jfs_info("xtRelocate: target data extent relocated.");
2939 } else { /* (xtype == XTPAGE) */
2940
2941 /*
2942 * read in the target xtpage from the source extent;
2943 */
2944 XT_GETPAGE(ip, oxaddr, mp, PSIZE, p, rc);
2945 if (rc) {
2946 XT_PUTPAGE(pmp);
2947 return rc;
2948 }
2949
2950 /*
2951 * read in sibling pages if any to update sibling pointers;
2952 */
2953 rmp = NULL;
2954 if (p->header.next) {
2955 nextbn = le64_to_cpu(p->header.next);
2956 XT_GETPAGE(ip, nextbn, rmp, PSIZE, rp, rc);
2957 if (rc) {
2958 XT_PUTPAGE(pmp);
2959 XT_PUTPAGE(mp);
2960 return (rc);
2961 }
2962 }
2963
2964 lmp = NULL;
2965 if (p->header.prev) {
2966 prevbn = le64_to_cpu(p->header.prev);
2967 XT_GETPAGE(ip, prevbn, lmp, PSIZE, lp, rc);
2968 if (rc) {
2969 XT_PUTPAGE(pmp);
2970 XT_PUTPAGE(mp);
2971 if (rmp)
2972 XT_PUTPAGE(rmp);
2973 return (rc);
2974 }
2975 }
2976
2977 /* at this point, all xtpages to be updated are in memory */
2978
2979 /*
2980 * update sibling pointers of sibling xtpages if any;
2981 */
2982 if (lmp) {
2983 BT_MARK_DIRTY(lmp, ip);
2984 tlck =
2985 txLock(tid, ip, lmp, tlckXTREE | tlckRELINK);
2986 lp->header.next = cpu_to_le64(nxaddr);
2987 XT_PUTPAGE(lmp);
2988 }
2989
2990 if (rmp) {
2991 BT_MARK_DIRTY(rmp, ip);
2992 tlck =
2993 txLock(tid, ip, rmp, tlckXTREE | tlckRELINK);
2994 rp->header.prev = cpu_to_le64(nxaddr);
2995 XT_PUTPAGE(rmp);
2996 }
2997
2998 /*
2999 * update the target xtpage to be relocated
3000 *
3001 * update the self address of the target page
3002 * and write to destination extent;
3003 * redo image covers the whole xtpage since it is new page
3004 * to the destination extent;
3005 * update of bmap for the free of source extent
3006 * of the target xtpage itself:
3007 * update of bmap for the allocation of destination extent
3008 * of the target xtpage itself:
3009 * update of bmap for the extents covered by xad entries in
3010 * the target xtpage is not necessary since they are not
3011 * updated;
3012 * if not committed before this relocation,
3013 * target page may contain XAD_NEW entries which must
3014 * be scanned for bmap update (logredo() always
3015 * scan xtpage REDOPAGE image for bmap update);
3016 * if committed before this relocation (tlckRELOCATE),
3017 * scan may be skipped by commit() and logredo();
3018 */
3019 BT_MARK_DIRTY(mp, ip);
3020 /* tlckNEW init xtlck->lwm.offset = XTENTRYSTART; */
3021 tlck = txLock(tid, ip, mp, tlckXTREE | tlckNEW);
3022 xtlck = (struct xtlock *) & tlck->lock;
3023
3024 /* update the self address in the xtpage header */
3025 pxd = &p->header.self;
3026 PXDaddress(pxd, nxaddr);
3027
3028 /* linelock for the after image of the whole page */
3029 xtlck->lwm.length =
3030 le16_to_cpu(p->header.nextindex) - xtlck->lwm.offset;
3031
3032 /* update the buffer extent descriptor of target xtpage */
3033 xsize = xlen << JFS_SBI(ip->i_sb)->l2bsize;
3034 bmSetXD(mp, nxaddr, xsize);
3035
3036 /* unpin the target page to new homeward bound */
3037 XT_PUTPAGE(mp);
3038 jfs_info("xtRelocate: target xtpage relocated.");
3039 }
3040
3041 /*
3042 * 3. acquire maplock for the source extent to be freed;
3043 *
3044 * acquire a maplock saving the src relocated extent address;
3045 * to free of the extent at commit time;
3046 */
3047 out:
3048 /* if DATAEXT relocation, write a LOG_UPDATEMAP record for
3049 * free PXD of the source data extent (logredo() will update
3050 * bmap for free of source data extent), and update bmap for
3051 * free of the source data extent;
3052 */
3053 if (xtype == DATAEXT)
3054 tlck = txMaplock(tid, ip, tlckMAP);
3055 /* if XTPAGE relocation, write a LOG_NOREDOPAGE record
3056 * for the source xtpage (logredo() will init NoRedoPage
3057 * filter and will also update bmap for free of the source
3058 * xtpage), and update bmap for free of the source xtpage;
3059 * N.B. We use tlckMAP instead of tlkcXTREE because there
3060 * is no buffer associated with this lock since the buffer
3061 * has been redirected to the target location.
3062 */
3063 else /* (xtype == XTPAGE) */
3064 tlck = txMaplock(tid, ip, tlckMAP | tlckRELOCATE);
3065
3066 pxdlock = (struct pxd_lock *) & tlck->lock;
3067 pxdlock->flag = mlckFREEPXD;
3068 PXDaddress(&pxdlock->pxd, oxaddr);
3069 PXDlength(&pxdlock->pxd, xlen);
3070 pxdlock->index = 1;
3071
3072 /*
3073 * 4. update the parent xad entry for relocation;
3074 *
3075 * acquire tlck for the parent entry with XAD_NEW as entry
3076 * update which will write LOG_REDOPAGE and update bmap for
3077 * allocation of XAD_NEW destination extent;
3078 */
3079 jfs_info("xtRelocate: update parent xad entry.");
3080 BT_MARK_DIRTY(pmp, ip);
3081 tlck = txLock(tid, ip, pmp, tlckXTREE | tlckGROW);
3082 xtlck = (struct xtlock *) & tlck->lock;
3083
3084 /* update the XAD with the new destination extent; */
3085 xad = &pp->xad[index];
3086 xad->flag |= XAD_NEW;
3087 XADaddress(xad, nxaddr);
3088
3089 xtlck->lwm.offset = min(index, xtlck->lwm.offset);
3090 xtlck->lwm.length = le16_to_cpu(pp->header.nextindex) -
3091 xtlck->lwm.offset;
3092
3093 /* unpin the parent xtpage */
3094 XT_PUTPAGE(pmp);
3095
3096 return rc;
3097 }
3098
3099
3100 /*
3101 * xtSearchNode()
3102 *
3103 * function: search for the internal xad entry covering specified extent.
3104 * This function is mainly used by defragfs utility.
3105 *
3106 * parameters:
3107 * ip - file object;
3108 * xad - extent to find;
3109 * cmpp - comparison result:
3110 * btstack - traverse stack;
3111 * flag - search process flag;
3112 *
3113 * returns:
3114 * btstack contains (bn, index) of search path traversed to the entry.
3115 * *cmpp is set to result of comparison with the entry returned.
3116 * the page containing the entry is pinned at exit.
3117 */
xtSearchNode(struct inode * ip,xad_t * xad,int * cmpp,struct btstack * btstack,int flag)3118 static int xtSearchNode(struct inode *ip, xad_t * xad, /* required XAD entry */
3119 int *cmpp, struct btstack * btstack, int flag)
3120 {
3121 int rc = 0;
3122 s64 xoff, xaddr;
3123 int xlen;
3124 int cmp = 1; /* init for empty page */
3125 s64 bn; /* block number */
3126 struct metapage *mp; /* meta-page buffer */
3127 xtpage_t *p; /* page */
3128 int base, index, lim;
3129 struct btframe *btsp;
3130 s64 t64;
3131
3132 BT_CLR(btstack);
3133
3134 xoff = offsetXAD(xad);
3135 xlen = lengthXAD(xad);
3136 xaddr = addressXAD(xad);
3137
3138 /*
3139 * search down tree from root:
3140 *
3141 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
3142 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
3143 *
3144 * if entry with search key K is not found
3145 * internal page search find the entry with largest key Ki
3146 * less than K which point to the child page to search;
3147 * leaf page search find the entry with smallest key Kj
3148 * greater than K so that the returned index is the position of
3149 * the entry to be shifted right for insertion of new entry.
3150 * for empty tree, search key is greater than any key of the tree.
3151 *
3152 * by convention, root bn = 0.
3153 */
3154 for (bn = 0;;) {
3155 /* get/pin the page to search */
3156 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3157 if (rc)
3158 return rc;
3159 if (p->header.flag & BT_LEAF) {
3160 XT_PUTPAGE(mp);
3161 return -ESTALE;
3162 }
3163
3164 lim = le16_to_cpu(p->header.nextindex) - XTENTRYSTART;
3165
3166 /*
3167 * binary search with search key K on the current page
3168 */
3169 for (base = XTENTRYSTART; lim; lim >>= 1) {
3170 index = base + (lim >> 1);
3171
3172 XT_CMP(cmp, xoff, &p->xad[index], t64);
3173 if (cmp == 0) {
3174 /*
3175 * search hit
3176 *
3177 * verify for exact match;
3178 */
3179 if (xaddr == addressXAD(&p->xad[index]) &&
3180 xoff == offsetXAD(&p->xad[index])) {
3181 *cmpp = cmp;
3182
3183 /* save search result */
3184 btsp = btstack->top;
3185 btsp->bn = bn;
3186 btsp->index = index;
3187 btsp->mp = mp;
3188
3189 return 0;
3190 }
3191
3192 /* descend/search its child page */
3193 goto next;
3194 }
3195
3196 if (cmp > 0) {
3197 base = index + 1;
3198 --lim;
3199 }
3200 }
3201
3202 /*
3203 * search miss - non-leaf page:
3204 *
3205 * base is the smallest index with key (Kj) greater than
3206 * search key (K) and may be zero or maxentry index.
3207 * if base is non-zero, decrement base by one to get the parent
3208 * entry of the child page to search.
3209 */
3210 index = base ? base - 1 : base;
3211
3212 /*
3213 * go down to child page
3214 */
3215 next:
3216 /* get the child page block number */
3217 bn = addressXAD(&p->xad[index]);
3218
3219 /* unpin the parent page */
3220 XT_PUTPAGE(mp);
3221 }
3222 }
3223
3224
3225 /*
3226 * xtRelink()
3227 *
3228 * function:
3229 * link around a freed page.
3230 *
3231 * Parameter:
3232 * int tid,
3233 * struct inode *ip,
3234 * xtpage_t *p)
3235 *
3236 * returns:
3237 */
xtRelink(tid_t tid,struct inode * ip,xtpage_t * p)3238 static int xtRelink(tid_t tid, struct inode *ip, xtpage_t * p)
3239 {
3240 int rc = 0;
3241 struct metapage *mp;
3242 s64 nextbn, prevbn;
3243 struct tlock *tlck;
3244
3245 nextbn = le64_to_cpu(p->header.next);
3246 prevbn = le64_to_cpu(p->header.prev);
3247
3248 /* update prev pointer of the next page */
3249 if (nextbn != 0) {
3250 XT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
3251 if (rc)
3252 return rc;
3253
3254 /*
3255 * acquire a transaction lock on the page;
3256 *
3257 * action: update prev pointer;
3258 */
3259 BT_MARK_DIRTY(mp, ip);
3260 tlck = txLock(tid, ip, mp, tlckXTREE | tlckRELINK);
3261
3262 /* the page may already have been tlock'd */
3263
3264 p->header.prev = cpu_to_le64(prevbn);
3265
3266 XT_PUTPAGE(mp);
3267 }
3268
3269 /* update next pointer of the previous page */
3270 if (prevbn != 0) {
3271 XT_GETPAGE(ip, prevbn, mp, PSIZE, p, rc);
3272 if (rc)
3273 return rc;
3274
3275 /*
3276 * acquire a transaction lock on the page;
3277 *
3278 * action: update next pointer;
3279 */
3280 BT_MARK_DIRTY(mp, ip);
3281 tlck = txLock(tid, ip, mp, tlckXTREE | tlckRELINK);
3282
3283 /* the page may already have been tlock'd */
3284
3285 p->header.next = le64_to_cpu(nextbn);
3286
3287 XT_PUTPAGE(mp);
3288 }
3289
3290 return 0;
3291 }
3292 #endif /* _STILL_TO_PORT */
3293
3294
3295 /*
3296 * xtInitRoot()
3297 *
3298 * initialize file root (inline in inode)
3299 */
xtInitRoot(tid_t tid,struct inode * ip)3300 void xtInitRoot(tid_t tid, struct inode *ip)
3301 {
3302 xtpage_t *p;
3303
3304 /*
3305 * acquire a transaction lock on the root
3306 *
3307 * action:
3308 */
3309 txLock(tid, ip, (struct metapage *) &JFS_IP(ip)->bxflag,
3310 tlckXTREE | tlckNEW);
3311 p = &JFS_IP(ip)->i_xtroot;
3312
3313 p->header.flag = DXD_INDEX | BT_ROOT | BT_LEAF;
3314 p->header.nextindex = cpu_to_le16(XTENTRYSTART);
3315
3316 if (S_ISDIR(ip->i_mode))
3317 p->header.maxentry = cpu_to_le16(XTROOTINITSLOT_DIR);
3318 else {
3319 p->header.maxentry = cpu_to_le16(XTROOTINITSLOT);
3320 ip->i_size = 0;
3321 }
3322
3323
3324 return;
3325 }
3326
3327
3328 /*
3329 * We can run into a deadlock truncating a file with a large number of
3330 * xtree pages (large fragmented file). A robust fix would entail a
3331 * reservation system where we would reserve a number of metadata pages
3332 * and tlocks which we would be guaranteed without a deadlock. Without
3333 * this, a partial fix is to limit number of metadata pages we will lock
3334 * in a single transaction. Currently we will truncate the file so that
3335 * no more than 50 leaf pages will be locked. The caller of xtTruncate
3336 * will be responsible for ensuring that the current transaction gets
3337 * committed, and that subsequent transactions are created to truncate
3338 * the file further if needed.
3339 */
3340 #define MAX_TRUNCATE_LEAVES 50
3341
3342 /*
3343 * xtTruncate()
3344 *
3345 * function:
3346 * traverse for truncation logging backward bottom up;
3347 * terminate at the last extent entry at the current subtree
3348 * root page covering new down size.
3349 * truncation may occur within the last extent entry.
3350 *
3351 * parameter:
3352 * int tid,
3353 * struct inode *ip,
3354 * s64 newsize,
3355 * int type) {PWMAP, PMAP, WMAP; DELETE, TRUNCATE}
3356 *
3357 * return:
3358 *
3359 * note:
3360 * PWMAP:
3361 * 1. truncate (non-COMMIT_NOLINK file)
3362 * by jfs_truncate() or jfs_open(O_TRUNC):
3363 * xtree is updated;
3364 * 2. truncate index table of directory when last entry removed
3365 * map update via tlock at commit time;
3366 * PMAP:
3367 * Call xtTruncate_pmap instead
3368 * WMAP:
3369 * 1. remove (free zero link count) on last reference release
3370 * (pmap has been freed at commit zero link count);
3371 * 2. truncate (COMMIT_NOLINK file, i.e., tmp file):
3372 * xtree is updated;
3373 * map update directly at truncation time;
3374 *
3375 * if (DELETE)
3376 * no LOG_NOREDOPAGE is required (NOREDOFILE is sufficient);
3377 * else if (TRUNCATE)
3378 * must write LOG_NOREDOPAGE for deleted index page;
3379 *
3380 * pages may already have been tlocked by anonymous transactions
3381 * during file growth (i.e., write) before truncation;
3382 *
3383 * except last truncated entry, deleted entries remains as is
3384 * in the page (nextindex is updated) for other use
3385 * (e.g., log/update allocation map): this avoid copying the page
3386 * info but delay free of pages;
3387 *
3388 */
xtTruncate(tid_t tid,struct inode * ip,s64 newsize,int flag)3389 s64 xtTruncate(tid_t tid, struct inode *ip, s64 newsize, int flag)
3390 {
3391 int rc = 0;
3392 s64 teof;
3393 struct metapage *mp;
3394 xtpage_t *p;
3395 s64 bn;
3396 int index, nextindex;
3397 xad_t *xad;
3398 s64 xoff, xaddr;
3399 int xlen, len, freexlen;
3400 struct btstack btstack;
3401 struct btframe *parent;
3402 struct tblock *tblk = 0;
3403 struct tlock *tlck = 0;
3404 struct xtlock *xtlck = 0;
3405 struct xdlistlock xadlock; /* maplock for COMMIT_WMAP */
3406 struct pxd_lock *pxdlock; /* maplock for COMMIT_WMAP */
3407 s64 nfreed;
3408 int freed, log;
3409 int locked_leaves = 0;
3410
3411 /* save object truncation type */
3412 if (tid) {
3413 tblk = tid_to_tblock(tid);
3414 tblk->xflag |= flag;
3415 }
3416
3417 nfreed = 0;
3418
3419 flag &= COMMIT_MAP;
3420 assert(flag != COMMIT_PMAP);
3421
3422 if (flag == COMMIT_PWMAP)
3423 log = 1;
3424 else {
3425 log = 0;
3426 xadlock.flag = mlckFREEXADLIST;
3427 xadlock.index = 1;
3428 }
3429
3430 /*
3431 * if the newsize is not an integral number of pages,
3432 * the file between newsize and next page boundary will
3433 * be cleared.
3434 * if truncating into a file hole, it will cause
3435 * a full block to be allocated for the logical block.
3436 */
3437
3438 /*
3439 * release page blocks of truncated region <teof, eof>
3440 *
3441 * free the data blocks from the leaf index blocks.
3442 * delete the parent index entries corresponding to
3443 * the freed child data/index blocks.
3444 * free the index blocks themselves which aren't needed
3445 * in new sized file.
3446 *
3447 * index blocks are updated only if the blocks are to be
3448 * retained in the new sized file.
3449 * if type is PMAP, the data and index pages are NOT
3450 * freed, and the data and index blocks are NOT freed
3451 * from working map.
3452 * (this will allow continued access of data/index of
3453 * temporary file (zerolink count file truncated to zero-length)).
3454 */
3455 teof = (newsize + (JFS_SBI(ip->i_sb)->bsize - 1)) >>
3456 JFS_SBI(ip->i_sb)->l2bsize;
3457
3458 /* clear stack */
3459 BT_CLR(&btstack);
3460
3461 /*
3462 * start with root
3463 *
3464 * root resides in the inode
3465 */
3466 bn = 0;
3467
3468 /*
3469 * first access of each page:
3470 */
3471 getPage:
3472 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3473 if (rc)
3474 return rc;
3475
3476 /* process entries backward from last index */
3477 index = le16_to_cpu(p->header.nextindex) - 1;
3478
3479 if (p->header.flag & BT_INTERNAL)
3480 goto getChild;
3481
3482 /*
3483 * leaf page
3484 */
3485
3486 /* Since this is the rightmost leaf, and we may have already freed
3487 * a page that was formerly to the right, let's make sure that the
3488 * next pointer is zero.
3489 */
3490 if (p->header.next) {
3491 if (log)
3492 /*
3493 * Make sure this change to the header is logged.
3494 * If we really truncate this leaf, the flag
3495 * will be changed to tlckTRUNCATE
3496 */
3497 tlck = txLock(tid, ip, mp, tlckXTREE|tlckGROW);
3498 BT_MARK_DIRTY(mp, ip);
3499 p->header.next = 0;
3500 }
3501
3502 freed = 0;
3503
3504 /* does region covered by leaf page precede Teof ? */
3505 xad = &p->xad[index];
3506 xoff = offsetXAD(xad);
3507 xlen = lengthXAD(xad);
3508 if (teof >= xoff + xlen) {
3509 XT_PUTPAGE(mp);
3510 goto getParent;
3511 }
3512
3513 /* (re)acquire tlock of the leaf page */
3514 if (log) {
3515 if (++locked_leaves > MAX_TRUNCATE_LEAVES) {
3516 /*
3517 * We need to limit the size of the transaction
3518 * to avoid exhausting pagecache & tlocks
3519 */
3520 XT_PUTPAGE(mp);
3521 newsize = (xoff + xlen) << JFS_SBI(ip->i_sb)->l2bsize;
3522 goto getParent;
3523 }
3524 tlck = txLock(tid, ip, mp, tlckXTREE);
3525 tlck->type = tlckXTREE | tlckTRUNCATE;
3526 xtlck = (struct xtlock *) & tlck->lock;
3527 xtlck->hwm.offset = le16_to_cpu(p->header.nextindex) - 1;
3528 }
3529 BT_MARK_DIRTY(mp, ip);
3530
3531 /*
3532 * scan backward leaf page entries
3533 */
3534 for (; index >= XTENTRYSTART; index--) {
3535 xad = &p->xad[index];
3536 xoff = offsetXAD(xad);
3537 xlen = lengthXAD(xad);
3538 xaddr = addressXAD(xad);
3539
3540 /*
3541 * The "data" for a directory is indexed by the block
3542 * device's address space. This metadata must be invalidated
3543 * here
3544 */
3545 if (S_ISDIR(ip->i_mode) && (teof == 0))
3546 invalidate_xad_metapages(ip, *xad);
3547 /*
3548 * entry beyond eof: continue scan of current page
3549 * xad
3550 * ---|---=======------->
3551 * eof
3552 */
3553 if (teof < xoff) {
3554 nfreed += xlen;
3555 continue;
3556 }
3557
3558 /*
3559 * (xoff <= teof): last entry to be deleted from page;
3560 * If other entries remain in page: keep and update the page.
3561 */
3562
3563 /*
3564 * eof == entry_start: delete the entry
3565 * xad
3566 * -------|=======------->
3567 * eof
3568 *
3569 */
3570 if (teof == xoff) {
3571 nfreed += xlen;
3572
3573 if (index == XTENTRYSTART)
3574 break;
3575
3576 nextindex = index;
3577 }
3578 /*
3579 * eof within the entry: truncate the entry.
3580 * xad
3581 * -------===|===------->
3582 * eof
3583 */
3584 else if (teof < xoff + xlen) {
3585 /* update truncated entry */
3586 len = teof - xoff;
3587 freexlen = xlen - len;
3588 XADlength(xad, len);
3589
3590 /* save pxd of truncated extent in tlck */
3591 xaddr += len;
3592 if (log) { /* COMMIT_PWMAP */
3593 xtlck->lwm.offset = (xtlck->lwm.offset) ?
3594 min(index, (int)xtlck->lwm.offset) : index;
3595 xtlck->lwm.length = index + 1 -
3596 xtlck->lwm.offset;
3597 xtlck->twm.offset = index;
3598 pxdlock = (struct pxd_lock *) & xtlck->pxdlock;
3599 pxdlock->flag = mlckFREEPXD;
3600 PXDaddress(&pxdlock->pxd, xaddr);
3601 PXDlength(&pxdlock->pxd, freexlen);
3602 }
3603 /* free truncated extent */
3604 else { /* COMMIT_WMAP */
3605
3606 pxdlock = (struct pxd_lock *) & xadlock;
3607 pxdlock->flag = mlckFREEPXD;
3608 PXDaddress(&pxdlock->pxd, xaddr);
3609 PXDlength(&pxdlock->pxd, freexlen);
3610 txFreeMap(ip, pxdlock, 0, COMMIT_WMAP);
3611
3612 /* reset map lock */
3613 xadlock.flag = mlckFREEXADLIST;
3614 }
3615
3616 /* current entry is new last entry; */
3617 nextindex = index + 1;
3618
3619 nfreed += freexlen;
3620 }
3621 /*
3622 * eof beyond the entry:
3623 * xad
3624 * -------=======---|--->
3625 * eof
3626 */
3627 else { /* (xoff + xlen < teof) */
3628
3629 nextindex = index + 1;
3630 }
3631
3632 if (nextindex < le16_to_cpu(p->header.nextindex)) {
3633 if (!log) { /* COMMIT_WAMP */
3634 xadlock.xdlist = &p->xad[nextindex];
3635 xadlock.count =
3636 le16_to_cpu(p->header.nextindex) -
3637 nextindex;
3638 txFreeMap(ip, (struct maplock *) & xadlock, 0,
3639 COMMIT_WMAP);
3640 }
3641 p->header.nextindex = cpu_to_le16(nextindex);
3642 }
3643
3644 XT_PUTPAGE(mp);
3645
3646 /* assert(freed == 0); */
3647 goto getParent;
3648 } /* end scan of leaf page entries */
3649
3650 freed = 1;
3651
3652 /*
3653 * leaf page become empty: free the page if type != PMAP
3654 */
3655 if (log) { /* COMMIT_PWMAP */
3656 /* txCommit() with tlckFREE:
3657 * free data extents covered by leaf [XTENTRYSTART:hwm);
3658 * invalidate leaf if COMMIT_PWMAP;
3659 * if (TRUNCATE), will write LOG_NOREDOPAGE;
3660 */
3661 tlck->type = tlckXTREE | tlckFREE;
3662 } else { /* COMMIT_WAMP */
3663
3664 /* free data extents covered by leaf */
3665 xadlock.xdlist = &p->xad[XTENTRYSTART];
3666 xadlock.count =
3667 le16_to_cpu(p->header.nextindex) - XTENTRYSTART;
3668 txFreeMap(ip, (struct maplock *) & xadlock, 0, COMMIT_WMAP);
3669 }
3670
3671 if (p->header.flag & BT_ROOT) {
3672 p->header.flag &= ~BT_INTERNAL;
3673 p->header.flag |= BT_LEAF;
3674 p->header.nextindex = cpu_to_le16(XTENTRYSTART);
3675
3676 XT_PUTPAGE(mp); /* debug */
3677 goto out;
3678 } else {
3679 if (log) { /* COMMIT_PWMAP */
3680 /* page will be invalidated at tx completion
3681 */
3682 XT_PUTPAGE(mp);
3683 } else { /* COMMIT_WMAP */
3684
3685 if (mp->lid)
3686 lid_to_tlock(mp->lid)->flag |= tlckFREELOCK;
3687
3688 /* invalidate empty leaf page */
3689 discard_metapage(mp);
3690 }
3691 }
3692
3693 /*
3694 * the leaf page become empty: delete the parent entry
3695 * for the leaf page if the parent page is to be kept
3696 * in the new sized file.
3697 */
3698
3699 /*
3700 * go back up to the parent page
3701 */
3702 getParent:
3703 /* pop/restore parent entry for the current child page */
3704 if ((parent = BT_POP(&btstack)) == NULL)
3705 /* current page must have been root */
3706 goto out;
3707
3708 /* get back the parent page */
3709 bn = parent->bn;
3710 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3711 if (rc)
3712 return rc;
3713
3714 index = parent->index;
3715
3716 /*
3717 * child page was not empty:
3718 */
3719 if (freed == 0) {
3720 /* has any entry deleted from parent ? */
3721 if (index < le16_to_cpu(p->header.nextindex) - 1) {
3722 /* (re)acquire tlock on the parent page */
3723 if (log) { /* COMMIT_PWMAP */
3724 /* txCommit() with tlckTRUNCATE:
3725 * free child extents covered by parent [);
3726 */
3727 tlck = txLock(tid, ip, mp, tlckXTREE);
3728 xtlck = (struct xtlock *) & tlck->lock;
3729 if (!(tlck->type & tlckTRUNCATE)) {
3730 xtlck->hwm.offset =
3731 le16_to_cpu(p->header.
3732 nextindex) - 1;
3733 tlck->type =
3734 tlckXTREE | tlckTRUNCATE;
3735 }
3736 } else { /* COMMIT_WMAP */
3737
3738 /* free child extents covered by parent */
3739 xadlock.xdlist = &p->xad[index + 1];
3740 xadlock.count =
3741 le16_to_cpu(p->header.nextindex) -
3742 index - 1;
3743 txFreeMap(ip, (struct maplock *) & xadlock, 0,
3744 COMMIT_WMAP);
3745 }
3746 BT_MARK_DIRTY(mp, ip);
3747
3748 p->header.nextindex = cpu_to_le16(index + 1);
3749 }
3750 XT_PUTPAGE(mp);
3751 goto getParent;
3752 }
3753
3754 /*
3755 * child page was empty:
3756 */
3757 nfreed += lengthXAD(&p->xad[index]);
3758
3759 /*
3760 * During working map update, child page's tlock must be handled
3761 * before parent's. This is because the parent's tlock will cause
3762 * the child's disk space to be marked available in the wmap, so
3763 * it's important that the child page be released by that time.
3764 *
3765 * ToDo: tlocks should be on doubly-linked list, so we can
3766 * quickly remove it and add it to the end.
3767 */
3768
3769 /*
3770 * Move parent page's tlock to the end of the tid's tlock list
3771 */
3772 if (log && mp->lid && (tblk->last != mp->lid) &&
3773 lid_to_tlock(mp->lid)->tid) {
3774 lid_t lid = mp->lid;
3775 struct tlock *prev;
3776
3777 tlck = lid_to_tlock(lid);
3778
3779 if (tblk->next == lid)
3780 tblk->next = tlck->next;
3781 else {
3782 for (prev = lid_to_tlock(tblk->next);
3783 prev->next != lid;
3784 prev = lid_to_tlock(prev->next)) {
3785 assert(prev->next);
3786 }
3787 prev->next = tlck->next;
3788 }
3789 lid_to_tlock(tblk->last)->next = lid;
3790 tlck->next = 0;
3791 tblk->last = lid;
3792 }
3793
3794 /*
3795 * parent page become empty: free the page
3796 */
3797 if (index == XTENTRYSTART) {
3798 if (log) { /* COMMIT_PWMAP */
3799 /* txCommit() with tlckFREE:
3800 * free child extents covered by parent;
3801 * invalidate parent if COMMIT_PWMAP;
3802 */
3803 tlck = txLock(tid, ip, mp, tlckXTREE);
3804 xtlck = (struct xtlock *) & tlck->lock;
3805 xtlck->hwm.offset =
3806 le16_to_cpu(p->header.nextindex) - 1;
3807 tlck->type = tlckXTREE | tlckFREE;
3808 } else { /* COMMIT_WMAP */
3809
3810 /* free child extents covered by parent */
3811 xadlock.xdlist = &p->xad[XTENTRYSTART];
3812 xadlock.count =
3813 le16_to_cpu(p->header.nextindex) -
3814 XTENTRYSTART;
3815 txFreeMap(ip, (struct maplock *) & xadlock, 0,
3816 COMMIT_WMAP);
3817 }
3818 BT_MARK_DIRTY(mp, ip);
3819
3820 if (p->header.flag & BT_ROOT) {
3821 p->header.flag &= ~BT_INTERNAL;
3822 p->header.flag |= BT_LEAF;
3823 p->header.nextindex = cpu_to_le16(XTENTRYSTART);
3824 if (le16_to_cpu(p->header.maxentry) == XTROOTMAXSLOT) {
3825 /*
3826 * Shrink root down to allow inline
3827 * EA (otherwise fsck complains)
3828 */
3829 p->header.maxentry =
3830 cpu_to_le16(XTROOTINITSLOT);
3831 JFS_IP(ip)->mode2 |= INLINEEA;
3832 }
3833
3834 XT_PUTPAGE(mp); /* debug */
3835 goto out;
3836 } else {
3837 if (log) { /* COMMIT_PWMAP */
3838 /* page will be invalidated at tx completion
3839 */
3840 XT_PUTPAGE(mp);
3841 } else { /* COMMIT_WMAP */
3842
3843 if (mp->lid)
3844 lid_to_tlock(mp->lid)->flag |=
3845 tlckFREELOCK;
3846
3847 /* invalidate parent page */
3848 discard_metapage(mp);
3849 }
3850
3851 /* parent has become empty and freed:
3852 * go back up to its parent page
3853 */
3854 /* freed = 1; */
3855 goto getParent;
3856 }
3857 }
3858 /*
3859 * parent page still has entries for front region;
3860 */
3861 else {
3862 /* try truncate region covered by preceding entry
3863 * (process backward)
3864 */
3865 index--;
3866
3867 /* go back down to the child page corresponding
3868 * to the entry
3869 */
3870 goto getChild;
3871 }
3872
3873 /*
3874 * internal page: go down to child page of current entry
3875 */
3876 getChild:
3877 /* save current parent entry for the child page */
3878 BT_PUSH(&btstack, bn, index);
3879
3880 /* get child page */
3881 xad = &p->xad[index];
3882 bn = addressXAD(xad);
3883
3884 /*
3885 * first access of each internal entry:
3886 */
3887 /* release parent page */
3888 XT_PUTPAGE(mp);
3889
3890 /* process the child page */
3891 goto getPage;
3892
3893 out:
3894 /*
3895 * update file resource stat
3896 */
3897 /* set size
3898 */
3899 if (S_ISDIR(ip->i_mode) && !newsize)
3900 ip->i_size = 1; /* fsck hates zero-length directories */
3901 else
3902 ip->i_size = newsize;
3903
3904 /* update nblocks to reflect freed blocks */
3905 ip->i_blocks -= LBLK2PBLK(ip->i_sb, nfreed);
3906
3907 /*
3908 * free tlock of invalidated pages
3909 */
3910 if (flag == COMMIT_WMAP)
3911 txFreelock(ip);
3912
3913 return newsize;
3914 }
3915
3916
3917 /*
3918 * xtTruncate_pmap()
3919 *
3920 * function:
3921 * Perform truncate to zero lenghth for deleted file, leaving the
3922 * the xtree and working map untouched. This allows the file to
3923 * be accessed via open file handles, while the delete of the file
3924 * is committed to disk.
3925 *
3926 * parameter:
3927 * tid_t tid,
3928 * struct inode *ip,
3929 * s64 committed_size)
3930 *
3931 * return: new committed size
3932 *
3933 * note:
3934 *
3935 * To avoid deadlock by holding too many transaction locks, the
3936 * truncation may be broken up into multiple transactions.
3937 * The committed_size keeps track of part of the file has been
3938 * freed from the pmaps.
3939 */
xtTruncate_pmap(tid_t tid,struct inode * ip,s64 committed_size)3940 s64 xtTruncate_pmap(tid_t tid, struct inode *ip, s64 committed_size)
3941 {
3942 s64 bn;
3943 struct btstack btstack;
3944 int cmp;
3945 int index;
3946 int locked_leaves = 0;
3947 struct metapage *mp;
3948 xtpage_t *p;
3949 struct btframe *parent;
3950 int rc;
3951 struct tblock *tblk;
3952 struct tlock *tlck = 0;
3953 xad_t *xad;
3954 int xlen;
3955 s64 xoff;
3956 struct xtlock *xtlck = 0;
3957
3958 /* save object truncation type */
3959 tblk = tid_to_tblock(tid);
3960 tblk->xflag |= COMMIT_PMAP;
3961
3962 /* clear stack */
3963 BT_CLR(&btstack);
3964
3965 if (committed_size) {
3966 xoff = (committed_size >> JFS_SBI(ip->i_sb)->l2bsize) - 1;
3967 rc = xtSearch(ip, xoff, &cmp, &btstack, 0);
3968 if (rc)
3969 return rc;
3970
3971 XT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
3972
3973 if (cmp != 0) {
3974 XT_PUTPAGE(mp);
3975 jfs_error(ip->i_sb,
3976 "xtTruncate_pmap: did not find extent");
3977 return -EIO;
3978 }
3979 } else {
3980 /*
3981 * start with root
3982 *
3983 * root resides in the inode
3984 */
3985 bn = 0;
3986
3987 /*
3988 * first access of each page:
3989 */
3990 getPage:
3991 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3992 if (rc)
3993 return rc;
3994
3995 /* process entries backward from last index */
3996 index = le16_to_cpu(p->header.nextindex) - 1;
3997
3998 if (p->header.flag & BT_INTERNAL)
3999 goto getChild;
4000 }
4001
4002 /*
4003 * leaf page
4004 */
4005
4006 if (++locked_leaves > MAX_TRUNCATE_LEAVES) {
4007 /*
4008 * We need to limit the size of the transaction
4009 * to avoid exhausting pagecache & tlocks
4010 */
4011 xad = &p->xad[index];
4012 xoff = offsetXAD(xad);
4013 xlen = lengthXAD(xad);
4014 XT_PUTPAGE(mp);
4015 return (xoff + xlen) << JFS_SBI(ip->i_sb)->l2bsize;
4016 }
4017 tlck = txLock(tid, ip, mp, tlckXTREE);
4018 tlck->type = tlckXTREE | tlckFREE;
4019 xtlck = (struct xtlock *) & tlck->lock;
4020 xtlck->hwm.offset = index;
4021
4022
4023 XT_PUTPAGE(mp);
4024
4025 /*
4026 * go back up to the parent page
4027 */
4028 getParent:
4029 /* pop/restore parent entry for the current child page */
4030 if ((parent = BT_POP(&btstack)) == NULL)
4031 /* current page must have been root */
4032 goto out;
4033
4034 /* get back the parent page */
4035 bn = parent->bn;
4036 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4037 if (rc)
4038 return rc;
4039
4040 index = parent->index;
4041
4042 /*
4043 * parent page become empty: free the page
4044 */
4045 if (index == XTENTRYSTART) {
4046 /* txCommit() with tlckFREE:
4047 * free child extents covered by parent;
4048 * invalidate parent if COMMIT_PWMAP;
4049 */
4050 tlck = txLock(tid, ip, mp, tlckXTREE);
4051 xtlck = (struct xtlock *) & tlck->lock;
4052 xtlck->hwm.offset =
4053 le16_to_cpu(p->header.nextindex) - 1;
4054 tlck->type = tlckXTREE | tlckFREE;
4055
4056 XT_PUTPAGE(mp);
4057
4058 if (p->header.flag & BT_ROOT) {
4059
4060 goto out;
4061 } else {
4062 goto getParent;
4063 }
4064 }
4065 /*
4066 * parent page still has entries for front region;
4067 */
4068 else
4069 index--;
4070 /*
4071 * internal page: go down to child page of current entry
4072 */
4073 getChild:
4074 /* save current parent entry for the child page */
4075 BT_PUSH(&btstack, bn, index);
4076
4077 /* get child page */
4078 xad = &p->xad[index];
4079 bn = addressXAD(xad);
4080
4081 /*
4082 * first access of each internal entry:
4083 */
4084 /* release parent page */
4085 XT_PUTPAGE(mp);
4086
4087 /* process the child page */
4088 goto getPage;
4089
4090 out:
4091
4092 return 0;
4093 }
4094
4095
4096 #ifdef _JFS_DEBUG_XTREE
4097 /*
4098 * xtDisplayTree()
4099 *
4100 * function: traverse forward
4101 */
xtDisplayTree(struct inode * ip)4102 int xtDisplayTree(struct inode *ip)
4103 {
4104 int rc = 0;
4105 struct metapage *mp;
4106 xtpage_t *p;
4107 s64 bn, pbn;
4108 int index, lastindex, v, h;
4109 xad_t *xad;
4110 struct btstack btstack;
4111 struct btframe *btsp;
4112 struct btframe *parent;
4113
4114 printk("display B+-tree.\n");
4115
4116 /* clear stack */
4117 btsp = btstack.stack;
4118
4119 /*
4120 * start with root
4121 *
4122 * root resides in the inode
4123 */
4124 bn = 0;
4125 v = h = 0;
4126
4127 /*
4128 * first access of each page:
4129 */
4130 getPage:
4131 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4132 if (rc)
4133 return rc;
4134
4135 /* process entries forward from first index */
4136 index = XTENTRYSTART;
4137 lastindex = le16_to_cpu(p->header.nextindex) - 1;
4138
4139 if (p->header.flag & BT_INTERNAL) {
4140 /*
4141 * first access of each internal page
4142 */
4143 goto getChild;
4144 } else { /* (p->header.flag & BT_LEAF) */
4145
4146 /*
4147 * first access of each leaf page
4148 */
4149 printf("leaf page ");
4150 xtDisplayPage(ip, bn, p);
4151
4152 /* unpin the leaf page */
4153 XT_PUTPAGE(mp);
4154 }
4155
4156 /*
4157 * go back up to the parent page
4158 */
4159 getParent:
4160 /* pop/restore parent entry for the current child page */
4161 if ((parent = (btsp == btstack.stack ? NULL : --btsp)) == NULL)
4162 /* current page must have been root */
4163 return;
4164
4165 /*
4166 * parent page scan completed
4167 */
4168 if ((index = parent->index) == (lastindex = parent->lastindex)) {
4169 /* go back up to the parent page */
4170 goto getParent;
4171 }
4172
4173 /*
4174 * parent page has entries remaining
4175 */
4176 /* get back the parent page */
4177 bn = parent->bn;
4178 /* v = parent->level; */
4179 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4180 if (rc)
4181 return rc;
4182
4183 /* get next parent entry */
4184 index++;
4185
4186 /*
4187 * internal page: go down to child page of current entry
4188 */
4189 getChild:
4190 /* push/save current parent entry for the child page */
4191 btsp->bn = pbn = bn;
4192 btsp->index = index;
4193 btsp->lastindex = lastindex;
4194 /* btsp->level = v; */
4195 /* btsp->node = h; */
4196 ++btsp;
4197
4198 /* get child page */
4199 xad = &p->xad[index];
4200 bn = addressXAD(xad);
4201
4202 /*
4203 * first access of each internal entry:
4204 */
4205 /* release parent page */
4206 XT_PUTPAGE(mp);
4207
4208 printk("traverse down 0x%lx[%d]->0x%lx\n", (ulong) pbn, index,
4209 (ulong) bn);
4210 v++;
4211 h = index;
4212
4213 /* process the child page */
4214 goto getPage;
4215 }
4216
4217
4218 /*
4219 * xtDisplayPage()
4220 *
4221 * function: display page
4222 */
xtDisplayPage(struct inode * ip,s64 bn,xtpage_t * p)4223 int xtDisplayPage(struct inode *ip, s64 bn, xtpage_t * p)
4224 {
4225 int rc = 0;
4226 xad_t *xad;
4227 s64 xaddr, xoff;
4228 int xlen, i, j;
4229
4230 /* display page control */
4231 printf("bn:0x%lx flag:0x%x nextindex:%d\n",
4232 (ulong) bn, p->header.flag,
4233 le16_to_cpu(p->header.nextindex));
4234
4235 /* display entries */
4236 xad = &p->xad[XTENTRYSTART];
4237 for (i = XTENTRYSTART, j = 1; i < le16_to_cpu(p->header.nextindex);
4238 i++, xad++, j++) {
4239 xoff = offsetXAD(xad);
4240 xaddr = addressXAD(xad);
4241 xlen = lengthXAD(xad);
4242 printf("\t[%d] 0x%lx:0x%lx(0x%x)", i, (ulong) xoff,
4243 (ulong) xaddr, xlen);
4244
4245 if (j == 4) {
4246 printf("\n");
4247 j = 0;
4248 }
4249 }
4250
4251 printf("\n");
4252 }
4253 #endif /* _JFS_DEBUG_XTREE */
4254
4255
4256 #ifdef _JFS_WIP
4257 /*
4258 * xtGather()
4259 *
4260 * function:
4261 * traverse for allocation acquiring tlock at commit time
4262 * (vs at the time of update) logging backward top down
4263 *
4264 * note:
4265 * problem - establishing that all new allocation have been
4266 * processed both for append and random write in sparse file
4267 * at the current entry at the current subtree root page
4268 *
4269 */
xtGather(btree_t * t)4270 int xtGather(btree_t *t)
4271 {
4272 int rc = 0;
4273 xtpage_t *p;
4274 u64 bn;
4275 int index;
4276 btentry_t *e;
4277 struct btstack btstack;
4278 struct btsf *parent;
4279
4280 /* clear stack */
4281 BT_CLR(&btstack);
4282
4283 /*
4284 * start with root
4285 *
4286 * root resides in the inode
4287 */
4288 bn = 0;
4289 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4290 if (rc)
4291 return rc;
4292
4293 /* new root is NOT pointed by a new entry
4294 if (p->header.flag & NEW)
4295 allocate new page lock;
4296 write a NEWPAGE log;
4297 */
4298
4299 dopage:
4300 /*
4301 * first access of each page:
4302 */
4303 /* process entries backward from last index */
4304 index = le16_to_cpu(p->header.nextindex) - 1;
4305
4306 if (p->header.flag & BT_LEAF) {
4307 /*
4308 * first access of each leaf page
4309 */
4310 /* process leaf page entries backward */
4311 for (; index >= XTENTRYSTART; index--) {
4312 e = &p->xad[index];
4313 /*
4314 * if newpage, log NEWPAGE.
4315 *
4316 if (e->flag & XAD_NEW) {
4317 nfound =+ entry->length;
4318 update current page lock for the entry;
4319 newpage(entry);
4320 *
4321 * if moved, log move.
4322 *
4323 } else if (e->flag & XAD_MOVED) {
4324 reset flag;
4325 update current page lock for the entry;
4326 }
4327 */
4328 }
4329
4330 /* unpin the leaf page */
4331 XT_PUTPAGE(mp);
4332
4333 /*
4334 * go back up to the parent page
4335 */
4336 getParent:
4337 /* restore parent entry for the current child page */
4338 if ((parent = BT_POP(&btstack)) == NULL)
4339 /* current page must have been root */
4340 return 0;
4341
4342 if ((index = parent->index) == XTENTRYSTART) {
4343 /*
4344 * parent page scan completed
4345 */
4346 /* go back up to the parent page */
4347 goto getParent;
4348 } else {
4349 /*
4350 * parent page has entries remaining
4351 */
4352 /* get back the parent page */
4353 bn = parent->bn;
4354 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4355 if (rc)
4356 return -EIO;
4357
4358 /* first subroot page which
4359 * covers all new allocated blocks
4360 * itself not new/modified.
4361 * (if modified from split of descendent,
4362 * go down path of split page)
4363
4364 if (nfound == nnew &&
4365 !(p->header.flag & (NEW | MOD)))
4366 exit scan;
4367 */
4368
4369 /* process parent page entries backward */
4370 index--;
4371 }
4372 } else {
4373 /*
4374 * first access of each internal page
4375 */
4376 }
4377
4378 /*
4379 * internal page: go down to child page of current entry
4380 */
4381
4382 /* save current parent entry for the child page */
4383 BT_PUSH(&btstack, bn, index);
4384
4385 /* get current entry for the child page */
4386 e = &p->xad[index];
4387
4388 /*
4389 * first access of each internal entry:
4390 */
4391 /*
4392 * if new entry, log btree_tnewentry.
4393 *
4394 if (e->flag & XAD_NEW)
4395 update parent page lock for the entry;
4396 */
4397
4398 /* release parent page */
4399 XT_PUTPAGE(mp);
4400
4401 /* get child page */
4402 bn = e->bn;
4403 XT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
4404 if (rc)
4405 return rc;
4406
4407 /*
4408 * first access of each non-root page:
4409 */
4410 /*
4411 * if new, log btree_newpage.
4412 *
4413 if (p->header.flag & NEW)
4414 allocate new page lock;
4415 write a NEWPAGE log (next, prev);
4416 */
4417
4418 /* process the child page */
4419 goto dopage;
4420
4421 out:
4422 return 0;
4423 }
4424 #endif /* _JFS_WIP */
4425
4426
4427 #ifdef CONFIG_JFS_STATISTICS
jfs_xtstat_read(char * buffer,char ** start,off_t offset,int length,int * eof,void * data)4428 int jfs_xtstat_read(char *buffer, char **start, off_t offset, int length,
4429 int *eof, void *data)
4430 {
4431 int len = 0;
4432 off_t begin;
4433
4434 len += sprintf(buffer,
4435 "JFS Xtree statistics\n"
4436 "====================\n"
4437 "searches = %d\n"
4438 "fast searches = %d\n"
4439 "splits = %d\n",
4440 xtStat.search,
4441 xtStat.fastSearch,
4442 xtStat.split);
4443
4444 begin = offset;
4445 *start = buffer + begin;
4446 len -= begin;
4447
4448 if (len > length)
4449 len = length;
4450 else
4451 *eof = 1;
4452
4453 if (len < 0)
4454 len = 0;
4455
4456 return len;
4457 }
4458 #endif
4459