Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (C) International Business Machines Corp., 2000-2004
4 */
5
6/*
7 * jfs_dtree.c: directory B+-tree manager
8 *
9 * B+-tree with variable length key directory:
10 *
11 * each directory page is structured as an array of 32-byte
12 * directory entry slots initialized as a freelist
13 * to avoid search/compaction of free space at insertion.
14 * when an entry is inserted, a number of slots are allocated
15 * from the freelist as required to store variable length data
16 * of the entry; when the entry is deleted, slots of the entry
17 * are returned to freelist.
18 *
19 * leaf entry stores full name as key and file serial number
20 * (aka inode number) as data.
21 * internal/router entry stores sufffix compressed name
22 * as key and simple extent descriptor as data.
23 *
24 * each directory page maintains a sorted entry index table
25 * which stores the start slot index of sorted entries
26 * to allow binary search on the table.
27 *
28 * directory starts as a root/leaf page in on-disk inode
29 * inline data area.
30 * when it becomes full, it starts a leaf of a external extent
31 * of length of 1 block. each time the first leaf becomes full,
32 * it is extended rather than split (its size is doubled),
33 * until its length becoms 4 KBytes, from then the extent is split
34 * with new 4 Kbyte extent when it becomes full
35 * to reduce external fragmentation of small directories.
36 *
37 * blah, blah, blah, for linear scan of directory in pieces by
38 * readdir().
39 *
40 *
41 * case-insensitive directory file system
42 *
43 * names are stored in case-sensitive way in leaf entry.
44 * but stored, searched and compared in case-insensitive (uppercase) order
45 * (i.e., both search key and entry key are folded for search/compare):
46 * (note that case-sensitive order is BROKEN in storage, e.g.,
47 * sensitive: Ad, aB, aC, aD -> insensitive: aB, aC, aD, Ad
48 *
49 * entries which folds to the same key makes up a equivalent class
50 * whose members are stored as contiguous cluster (may cross page boundary)
51 * but whose order is arbitrary and acts as duplicate, e.g.,
52 * abc, Abc, aBc, abC)
53 *
54 * once match is found at leaf, requires scan forward/backward
55 * either for, in case-insensitive search, duplicate
56 * or for, in case-sensitive search, for exact match
57 *
58 * router entry must be created/stored in case-insensitive way
59 * in internal entry:
60 * (right most key of left page and left most key of right page
61 * are folded, and its suffix compression is propagated as router
62 * key in parent)
63 * (e.g., if split occurs <abc> and <aBd>, <ABD> trather than <aB>
64 * should be made the router key for the split)
65 *
66 * case-insensitive search:
67 *
68 * fold search key;
69 *
70 * case-insensitive search of B-tree:
71 * for internal entry, router key is already folded;
72 * for leaf entry, fold the entry key before comparison.
73 *
74 * if (leaf entry case-insensitive match found)
75 * if (next entry satisfies case-insensitive match)
76 * return EDUPLICATE;
77 * if (prev entry satisfies case-insensitive match)
78 * return EDUPLICATE;
79 * return match;
80 * else
81 * return no match;
82 *
83 * serialization:
84 * target directory inode lock is being held on entry/exit
85 * of all main directory service routines.
86 *
87 * log based recovery:
88 */
89
90#include <linux/fs.h>
91#include <linux/quotaops.h>
92#include <linux/slab.h>
93#include "jfs_incore.h"
94#include "jfs_superblock.h"
95#include "jfs_filsys.h"
96#include "jfs_metapage.h"
97#include "jfs_dmap.h"
98#include "jfs_unicode.h"
99#include "jfs_debug.h"
100
101/* dtree split parameter */
102struct dtsplit {
103 struct metapage *mp;
104 s16 index;
105 s16 nslot;
106 struct component_name *key;
107 ddata_t *data;
108 struct pxdlist *pxdlist;
109};
110
111#define DT_PAGE(IP, MP) BT_PAGE(IP, MP, dtpage_t, i_dtroot)
112
113/* get page buffer for specified block address */
114#define DT_GETPAGE(IP, BN, MP, SIZE, P, RC) \
115do { \
116 BT_GETPAGE(IP, BN, MP, dtpage_t, SIZE, P, RC, i_dtroot); \
117 if (!(RC)) { \
118 if (((P)->header.nextindex > \
119 (((BN) == 0) ? DTROOTMAXSLOT : (P)->header.maxslot)) || \
120 ((BN) && ((P)->header.maxslot > DTPAGEMAXSLOT))) { \
121 BT_PUTPAGE(MP); \
122 jfs_error((IP)->i_sb, \
123 "DT_GETPAGE: dtree page corrupt\n"); \
124 MP = NULL; \
125 RC = -EIO; \
126 } \
127 } \
128} while (0)
129
130/* for consistency */
131#define DT_PUTPAGE(MP) BT_PUTPAGE(MP)
132
133#define DT_GETSEARCH(IP, LEAF, BN, MP, P, INDEX) \
134 BT_GETSEARCH(IP, LEAF, BN, MP, dtpage_t, P, INDEX, i_dtroot)
135
136/*
137 * forward references
138 */
139static int dtSplitUp(tid_t tid, struct inode *ip,
140 struct dtsplit * split, struct btstack * btstack);
141
142static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
143 struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rxdp);
144
145static int dtExtendPage(tid_t tid, struct inode *ip,
146 struct dtsplit * split, struct btstack * btstack);
147
148static int dtSplitRoot(tid_t tid, struct inode *ip,
149 struct dtsplit * split, struct metapage ** rmpp);
150
151static int dtDeleteUp(tid_t tid, struct inode *ip, struct metapage * fmp,
152 dtpage_t * fp, struct btstack * btstack);
153
154static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p);
155
156static int dtReadFirst(struct inode *ip, struct btstack * btstack);
157
158static int dtReadNext(struct inode *ip,
159 loff_t * offset, struct btstack * btstack);
160
161static int dtCompare(struct component_name * key, dtpage_t * p, int si);
162
163static int ciCompare(struct component_name * key, dtpage_t * p, int si,
164 int flag);
165
166static void dtGetKey(dtpage_t * p, int i, struct component_name * key,
167 int flag);
168
169static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
170 int ri, struct component_name * key, int flag);
171
172static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
173 ddata_t * data, struct dt_lock **);
174
175static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
176 struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
177 int do_index);
178
179static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock);
180
181static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock);
182
183static void dtLinelockFreelist(dtpage_t * p, int m, struct dt_lock ** dtlock);
184
185#define ciToUpper(c) UniStrupr((c)->name)
186
187/*
188 * read_index_page()
189 *
190 * Reads a page of a directory's index table.
191 * Having metadata mapped into the directory inode's address space
192 * presents a multitude of problems. We avoid this by mapping to
193 * the absolute address space outside of the *_metapage routines
194 */
195static struct metapage *read_index_page(struct inode *inode, s64 blkno)
196{
197 int rc;
198 s64 xaddr;
199 int xflag;
200 s32 xlen;
201
202 rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
203 if (rc || (xaddr == 0))
204 return NULL;
205
206 return read_metapage(inode, xaddr, PSIZE, 1);
207}
208
209/*
210 * get_index_page()
211 *
212 * Same as get_index_page(), but get's a new page without reading
213 */
214static struct metapage *get_index_page(struct inode *inode, s64 blkno)
215{
216 int rc;
217 s64 xaddr;
218 int xflag;
219 s32 xlen;
220
221 rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
222 if (rc || (xaddr == 0))
223 return NULL;
224
225 return get_metapage(inode, xaddr, PSIZE, 1);
226}
227
228/*
229 * find_index()
230 *
231 * Returns dtree page containing directory table entry for specified
232 * index and pointer to its entry.
233 *
234 * mp must be released by caller.
235 */
236static struct dir_table_slot *find_index(struct inode *ip, u32 index,
237 struct metapage ** mp, s64 *lblock)
238{
239 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
240 s64 blkno;
241 s64 offset;
242 int page_offset;
243 struct dir_table_slot *slot;
244 static int maxWarnings = 10;
245
246 if (index < 2) {
247 if (maxWarnings) {
248 jfs_warn("find_entry called with index = %d", index);
249 maxWarnings--;
250 }
251 return NULL;
252 }
253
254 if (index >= jfs_ip->next_index) {
255 jfs_warn("find_entry called with index >= next_index");
256 return NULL;
257 }
258
259 if (jfs_dirtable_inline(ip)) {
260 /*
261 * Inline directory table
262 */
263 *mp = NULL;
264 slot = &jfs_ip->i_dirtable[index - 2];
265 } else {
266 offset = (index - 2) * sizeof(struct dir_table_slot);
267 page_offset = offset & (PSIZE - 1);
268 blkno = ((offset + 1) >> L2PSIZE) <<
269 JFS_SBI(ip->i_sb)->l2nbperpage;
270
271 if (*mp && (*lblock != blkno)) {
272 release_metapage(*mp);
273 *mp = NULL;
274 }
275 if (!(*mp)) {
276 *lblock = blkno;
277 *mp = read_index_page(ip, blkno);
278 }
279 if (!(*mp)) {
280 jfs_err("free_index: error reading directory table");
281 return NULL;
282 }
283
284 slot =
285 (struct dir_table_slot *) ((char *) (*mp)->data +
286 page_offset);
287 }
288 return slot;
289}
290
291static inline void lock_index(tid_t tid, struct inode *ip, struct metapage * mp,
292 u32 index)
293{
294 struct tlock *tlck;
295 struct linelock *llck;
296 struct lv *lv;
297
298 tlck = txLock(tid, ip, mp, tlckDATA);
299 llck = (struct linelock *) tlck->lock;
300
301 if (llck->index >= llck->maxcnt)
302 llck = txLinelock(llck);
303 lv = &llck->lv[llck->index];
304
305 /*
306 * Linelock slot size is twice the size of directory table
307 * slot size. 512 entries per page.
308 */
309 lv->offset = ((index - 2) & 511) >> 1;
310 lv->length = 1;
311 llck->index++;
312}
313
314/*
315 * add_index()
316 *
317 * Adds an entry to the directory index table. This is used to provide
318 * each directory entry with a persistent index in which to resume
319 * directory traversals
320 */
321static u32 add_index(tid_t tid, struct inode *ip, s64 bn, int slot)
322{
323 struct super_block *sb = ip->i_sb;
324 struct jfs_sb_info *sbi = JFS_SBI(sb);
325 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
326 u64 blkno;
327 struct dir_table_slot *dirtab_slot;
328 u32 index;
329 struct linelock *llck;
330 struct lv *lv;
331 struct metapage *mp;
332 s64 offset;
333 uint page_offset;
334 struct tlock *tlck;
335 s64 xaddr;
336
337 ASSERT(DO_INDEX(ip));
338
339 if (jfs_ip->next_index < 2) {
340 jfs_warn("add_index: next_index = %d. Resetting!",
341 jfs_ip->next_index);
342 jfs_ip->next_index = 2;
343 }
344
345 index = jfs_ip->next_index++;
346
347 if (index <= MAX_INLINE_DIRTABLE_ENTRY) {
348 /*
349 * i_size reflects size of index table, or 8 bytes per entry.
350 */
351 ip->i_size = (loff_t) (index - 1) << 3;
352
353 /*
354 * dir table fits inline within inode
355 */
356 dirtab_slot = &jfs_ip->i_dirtable[index-2];
357 dirtab_slot->flag = DIR_INDEX_VALID;
358 dirtab_slot->slot = slot;
359 DTSaddress(dirtab_slot, bn);
360
361 set_cflag(COMMIT_Dirtable, ip);
362
363 return index;
364 }
365 if (index == (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
366 struct dir_table_slot temp_table[12];
367
368 /*
369 * It's time to move the inline table to an external
370 * page and begin to build the xtree
371 */
372 if (dquot_alloc_block(ip, sbi->nbperpage))
373 goto clean_up;
374 if (dbAlloc(ip, 0, sbi->nbperpage, &xaddr)) {
375 dquot_free_block(ip, sbi->nbperpage);
376 goto clean_up;
377 }
378
379 /*
380 * Save the table, we're going to overwrite it with the
381 * xtree root
382 */
383 memcpy(temp_table, &jfs_ip->i_dirtable, sizeof(temp_table));
384
385 /*
386 * Initialize empty x-tree
387 */
388 xtInitRoot(tid, ip);
389
390 /*
391 * Add the first block to the xtree
392 */
393 if (xtInsert(tid, ip, 0, 0, sbi->nbperpage, &xaddr, 0)) {
394 /* This really shouldn't fail */
395 jfs_warn("add_index: xtInsert failed!");
396 memcpy(&jfs_ip->i_dirtable, temp_table,
397 sizeof (temp_table));
398 dbFree(ip, xaddr, sbi->nbperpage);
399 dquot_free_block(ip, sbi->nbperpage);
400 goto clean_up;
401 }
402 ip->i_size = PSIZE;
403
404 mp = get_index_page(ip, 0);
405 if (!mp) {
406 jfs_err("add_index: get_metapage failed!");
407 xtTruncate(tid, ip, 0, COMMIT_PWMAP);
408 memcpy(&jfs_ip->i_dirtable, temp_table,
409 sizeof (temp_table));
410 goto clean_up;
411 }
412 tlck = txLock(tid, ip, mp, tlckDATA);
413 llck = (struct linelock *) & tlck->lock;
414 ASSERT(llck->index == 0);
415 lv = &llck->lv[0];
416
417 lv->offset = 0;
418 lv->length = 6; /* tlckDATA slot size is 16 bytes */
419 llck->index++;
420
421 memcpy(mp->data, temp_table, sizeof(temp_table));
422
423 mark_metapage_dirty(mp);
424 release_metapage(mp);
425
426 /*
427 * Logging is now directed by xtree tlocks
428 */
429 clear_cflag(COMMIT_Dirtable, ip);
430 }
431
432 offset = (index - 2) * sizeof(struct dir_table_slot);
433 page_offset = offset & (PSIZE - 1);
434 blkno = ((offset + 1) >> L2PSIZE) << sbi->l2nbperpage;
435 if (page_offset == 0) {
436 /*
437 * This will be the beginning of a new page
438 */
439 xaddr = 0;
440 if (xtInsert(tid, ip, 0, blkno, sbi->nbperpage, &xaddr, 0)) {
441 jfs_warn("add_index: xtInsert failed!");
442 goto clean_up;
443 }
444 ip->i_size += PSIZE;
445
446 if ((mp = get_index_page(ip, blkno)))
447 memset(mp->data, 0, PSIZE); /* Just looks better */
448 else
449 xtTruncate(tid, ip, offset, COMMIT_PWMAP);
450 } else
451 mp = read_index_page(ip, blkno);
452
453 if (!mp) {
454 jfs_err("add_index: get/read_metapage failed!");
455 goto clean_up;
456 }
457
458 lock_index(tid, ip, mp, index);
459
460 dirtab_slot =
461 (struct dir_table_slot *) ((char *) mp->data + page_offset);
462 dirtab_slot->flag = DIR_INDEX_VALID;
463 dirtab_slot->slot = slot;
464 DTSaddress(dirtab_slot, bn);
465
466 mark_metapage_dirty(mp);
467 release_metapage(mp);
468
469 return index;
470
471 clean_up:
472
473 jfs_ip->next_index--;
474
475 return 0;
476}
477
478/*
479 * free_index()
480 *
481 * Marks an entry to the directory index table as free.
482 */
483static void free_index(tid_t tid, struct inode *ip, u32 index, u32 next)
484{
485 struct dir_table_slot *dirtab_slot;
486 s64 lblock;
487 struct metapage *mp = NULL;
488
489 dirtab_slot = find_index(ip, index, &mp, &lblock);
490
491 if (!dirtab_slot)
492 return;
493
494 dirtab_slot->flag = DIR_INDEX_FREE;
495 dirtab_slot->slot = dirtab_slot->addr1 = 0;
496 dirtab_slot->addr2 = cpu_to_le32(next);
497
498 if (mp) {
499 lock_index(tid, ip, mp, index);
500 mark_metapage_dirty(mp);
501 release_metapage(mp);
502 } else
503 set_cflag(COMMIT_Dirtable, ip);
504}
505
506/*
507 * modify_index()
508 *
509 * Changes an entry in the directory index table
510 */
511static void modify_index(tid_t tid, struct inode *ip, u32 index, s64 bn,
512 int slot, struct metapage ** mp, s64 *lblock)
513{
514 struct dir_table_slot *dirtab_slot;
515
516 dirtab_slot = find_index(ip, index, mp, lblock);
517
518 if (!dirtab_slot)
519 return;
520
521 DTSaddress(dirtab_slot, bn);
522 dirtab_slot->slot = slot;
523
524 if (*mp) {
525 lock_index(tid, ip, *mp, index);
526 mark_metapage_dirty(*mp);
527 } else
528 set_cflag(COMMIT_Dirtable, ip);
529}
530
531/*
532 * read_index()
533 *
534 * reads a directory table slot
535 */
536static int read_index(struct inode *ip, u32 index,
537 struct dir_table_slot * dirtab_slot)
538{
539 s64 lblock;
540 struct metapage *mp = NULL;
541 struct dir_table_slot *slot;
542
543 slot = find_index(ip, index, &mp, &lblock);
544 if (!slot) {
545 return -EIO;
546 }
547
548 memcpy(dirtab_slot, slot, sizeof(struct dir_table_slot));
549
550 if (mp)
551 release_metapage(mp);
552
553 return 0;
554}
555
556/*
557 * dtSearch()
558 *
559 * function:
560 * Search for the entry with specified key
561 *
562 * parameter:
563 *
564 * return: 0 - search result on stack, leaf page pinned;
565 * errno - I/O error
566 */
567int dtSearch(struct inode *ip, struct component_name * key, ino_t * data,
568 struct btstack * btstack, int flag)
569{
570 int rc = 0;
571 int cmp = 1; /* init for empty page */
572 s64 bn;
573 struct metapage *mp;
574 dtpage_t *p;
575 s8 *stbl;
576 int base, index, lim;
577 struct btframe *btsp;
578 pxd_t *pxd;
579 int psize = 288; /* initial in-line directory */
580 ino_t inumber;
581 struct component_name ciKey;
582 struct super_block *sb = ip->i_sb;
583
584 ciKey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
585 GFP_NOFS);
586 if (!ciKey.name) {
587 rc = -ENOMEM;
588 goto dtSearch_Exit2;
589 }
590
591
592 /* uppercase search key for c-i directory */
593 UniStrcpy(ciKey.name, key->name);
594 ciKey.namlen = key->namlen;
595
596 /* only uppercase if case-insensitive support is on */
597 if ((JFS_SBI(sb)->mntflag & JFS_OS2) == JFS_OS2) {
598 ciToUpper(&ciKey);
599 }
600 BT_CLR(btstack); /* reset stack */
601
602 /* init level count for max pages to split */
603 btstack->nsplit = 1;
604
605 /*
606 * search down tree from root:
607 *
608 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
609 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
610 *
611 * if entry with search key K is not found
612 * internal page search find the entry with largest key Ki
613 * less than K which point to the child page to search;
614 * leaf page search find the entry with smallest key Kj
615 * greater than K so that the returned index is the position of
616 * the entry to be shifted right for insertion of new entry.
617 * for empty tree, search key is greater than any key of the tree.
618 *
619 * by convention, root bn = 0.
620 */
621 for (bn = 0;;) {
622 /* get/pin the page to search */
623 DT_GETPAGE(ip, bn, mp, psize, p, rc);
624 if (rc)
625 goto dtSearch_Exit1;
626
627 /* get sorted entry table of the page */
628 stbl = DT_GETSTBL(p);
629
630 /*
631 * binary search with search key K on the current page.
632 */
633 for (base = 0, lim = p->header.nextindex; lim; lim >>= 1) {
634 index = base + (lim >> 1);
635
636 if (stbl[index] < 0) {
637 rc = -EIO;
638 goto out;
639 }
640
641 if (p->header.flag & BT_LEAF) {
642 /* uppercase leaf name to compare */
643 cmp =
644 ciCompare(&ciKey, p, stbl[index],
645 JFS_SBI(sb)->mntflag);
646 } else {
647 /* router key is in uppercase */
648
649 cmp = dtCompare(&ciKey, p, stbl[index]);
650
651
652 }
653 if (cmp == 0) {
654 /*
655 * search hit
656 */
657 /* search hit - leaf page:
658 * return the entry found
659 */
660 if (p->header.flag & BT_LEAF) {
661 inumber = le32_to_cpu(
662 ((struct ldtentry *) & p->slot[stbl[index]])->inumber);
663
664 /*
665 * search for JFS_LOOKUP
666 */
667 if (flag == JFS_LOOKUP) {
668 *data = inumber;
669 rc = 0;
670 goto out;
671 }
672
673 /*
674 * search for JFS_CREATE
675 */
676 if (flag == JFS_CREATE) {
677 *data = inumber;
678 rc = -EEXIST;
679 goto out;
680 }
681
682 /*
683 * search for JFS_REMOVE or JFS_RENAME
684 */
685 if ((flag == JFS_REMOVE ||
686 flag == JFS_RENAME) &&
687 *data != inumber) {
688 rc = -ESTALE;
689 goto out;
690 }
691
692 /*
693 * JFS_REMOVE|JFS_FINDDIR|JFS_RENAME
694 */
695 /* save search result */
696 *data = inumber;
697 btsp = btstack->top;
698 btsp->bn = bn;
699 btsp->index = index;
700 btsp->mp = mp;
701
702 rc = 0;
703 goto dtSearch_Exit1;
704 }
705
706 /* search hit - internal page:
707 * descend/search its child page
708 */
709 goto getChild;
710 }
711
712 if (cmp > 0) {
713 base = index + 1;
714 --lim;
715 }
716 }
717
718 /*
719 * search miss
720 *
721 * base is the smallest index with key (Kj) greater than
722 * search key (K) and may be zero or (maxindex + 1) index.
723 */
724 /*
725 * search miss - leaf page
726 *
727 * return location of entry (base) where new entry with
728 * search key K is to be inserted.
729 */
730 if (p->header.flag & BT_LEAF) {
731 /*
732 * search for JFS_LOOKUP, JFS_REMOVE, or JFS_RENAME
733 */
734 if (flag == JFS_LOOKUP || flag == JFS_REMOVE ||
735 flag == JFS_RENAME) {
736 rc = -ENOENT;
737 goto out;
738 }
739
740 /*
741 * search for JFS_CREATE|JFS_FINDDIR:
742 *
743 * save search result
744 */
745 *data = 0;
746 btsp = btstack->top;
747 btsp->bn = bn;
748 btsp->index = base;
749 btsp->mp = mp;
750
751 rc = 0;
752 goto dtSearch_Exit1;
753 }
754
755 /*
756 * search miss - internal page
757 *
758 * if base is non-zero, decrement base by one to get the parent
759 * entry of the child page to search.
760 */
761 index = base ? base - 1 : base;
762
763 /*
764 * go down to child page
765 */
766 getChild:
767 /* update max. number of pages to split */
768 if (BT_STACK_FULL(btstack)) {
769 /* Something's corrupted, mark filesystem dirty so
770 * chkdsk will fix it.
771 */
772 jfs_error(sb, "stack overrun!\n");
773 BT_STACK_DUMP(btstack);
774 rc = -EIO;
775 goto out;
776 }
777 btstack->nsplit++;
778
779 /* push (bn, index) of the parent page/entry */
780 BT_PUSH(btstack, bn, index);
781
782 /* get the child page block number */
783 pxd = (pxd_t *) & p->slot[stbl[index]];
784 bn = addressPXD(pxd);
785 psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;
786
787 /* unpin the parent page */
788 DT_PUTPAGE(mp);
789 }
790
791 out:
792 DT_PUTPAGE(mp);
793
794 dtSearch_Exit1:
795
796 kfree(ciKey.name);
797
798 dtSearch_Exit2:
799
800 return rc;
801}
802
803
804/*
805 * dtInsert()
806 *
807 * function: insert an entry to directory tree
808 *
809 * parameter:
810 *
811 * return: 0 - success;
812 * errno - failure;
813 */
814int dtInsert(tid_t tid, struct inode *ip,
815 struct component_name * name, ino_t * fsn, struct btstack * btstack)
816{
817 int rc = 0;
818 struct metapage *mp; /* meta-page buffer */
819 dtpage_t *p; /* base B+-tree index page */
820 s64 bn;
821 int index;
822 struct dtsplit split; /* split information */
823 ddata_t data;
824 struct dt_lock *dtlck;
825 int n;
826 struct tlock *tlck;
827 struct lv *lv;
828
829 /*
830 * retrieve search result
831 *
832 * dtSearch() returns (leaf page pinned, index at which to insert).
833 * n.b. dtSearch() may return index of (maxindex + 1) of
834 * the full page.
835 */
836 DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
837
838 /*
839 * insert entry for new key
840 */
841 if (DO_INDEX(ip)) {
842 if (JFS_IP(ip)->next_index == DIREND) {
843 DT_PUTPAGE(mp);
844 return -EMLINK;
845 }
846 n = NDTLEAF(name->namlen);
847 data.leaf.tid = tid;
848 data.leaf.ip = ip;
849 } else {
850 n = NDTLEAF_LEGACY(name->namlen);
851 data.leaf.ip = NULL; /* signifies legacy directory format */
852 }
853 data.leaf.ino = *fsn;
854
855 /*
856 * leaf page does not have enough room for new entry:
857 *
858 * extend/split the leaf page;
859 *
860 * dtSplitUp() will insert the entry and unpin the leaf page.
861 */
862 if (n > p->header.freecnt) {
863 split.mp = mp;
864 split.index = index;
865 split.nslot = n;
866 split.key = name;
867 split.data = &data;
868 rc = dtSplitUp(tid, ip, &split, btstack);
869 return rc;
870 }
871
872 /*
873 * leaf page does have enough room for new entry:
874 *
875 * insert the new data entry into the leaf page;
876 */
877 BT_MARK_DIRTY(mp, ip);
878 /*
879 * acquire a transaction lock on the leaf page
880 */
881 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
882 dtlck = (struct dt_lock *) & tlck->lock;
883 ASSERT(dtlck->index == 0);
884 lv = & dtlck->lv[0];
885
886 /* linelock header */
887 lv->offset = 0;
888 lv->length = 1;
889 dtlck->index++;
890
891 dtInsertEntry(p, index, name, &data, &dtlck);
892
893 /* linelock stbl of non-root leaf page */
894 if (!(p->header.flag & BT_ROOT)) {
895 if (dtlck->index >= dtlck->maxcnt)
896 dtlck = (struct dt_lock *) txLinelock(dtlck);
897 lv = & dtlck->lv[dtlck->index];
898 n = index >> L2DTSLOTSIZE;
899 lv->offset = p->header.stblindex + n;
900 lv->length =
901 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
902 dtlck->index++;
903 }
904
905 /* unpin the leaf page */
906 DT_PUTPAGE(mp);
907
908 return 0;
909}
910
911
912/*
913 * dtSplitUp()
914 *
915 * function: propagate insertion bottom up;
916 *
917 * parameter:
918 *
919 * return: 0 - success;
920 * errno - failure;
921 * leaf page unpinned;
922 */
923static int dtSplitUp(tid_t tid,
924 struct inode *ip, struct dtsplit * split, struct btstack * btstack)
925{
926 struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
927 int rc = 0;
928 struct metapage *smp;
929 dtpage_t *sp; /* split page */
930 struct metapage *rmp;
931 dtpage_t *rp; /* new right page split from sp */
932 pxd_t rpxd; /* new right page extent descriptor */
933 struct metapage *lmp;
934 dtpage_t *lp; /* left child page */
935 int skip; /* index of entry of insertion */
936 struct btframe *parent; /* parent page entry on traverse stack */
937 s64 xaddr, nxaddr;
938 int xlen, xsize;
939 struct pxdlist pxdlist;
940 pxd_t *pxd;
941 struct component_name key = { 0, NULL };
942 ddata_t *data = split->data;
943 int n;
944 struct dt_lock *dtlck;
945 struct tlock *tlck;
946 struct lv *lv;
947 int quota_allocation = 0;
948
949 /* get split page */
950 smp = split->mp;
951 sp = DT_PAGE(ip, smp);
952
953 key.name = kmalloc_array(JFS_NAME_MAX + 2, sizeof(wchar_t), GFP_NOFS);
954 if (!key.name) {
955 DT_PUTPAGE(smp);
956 rc = -ENOMEM;
957 goto dtSplitUp_Exit;
958 }
959
960 /*
961 * split leaf page
962 *
963 * The split routines insert the new entry, and
964 * acquire txLock as appropriate.
965 */
966 /*
967 * split root leaf page:
968 */
969 if (sp->header.flag & BT_ROOT) {
970 /*
971 * allocate a single extent child page
972 */
973 xlen = 1;
974 n = sbi->bsize >> L2DTSLOTSIZE;
975 n -= (n + 31) >> L2DTSLOTSIZE; /* stbl size */
976 n -= DTROOTMAXSLOT - sp->header.freecnt; /* header + entries */
977 if (n <= split->nslot)
978 xlen++;
979 if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr))) {
980 DT_PUTPAGE(smp);
981 goto freeKeyName;
982 }
983
984 pxdlist.maxnpxd = 1;
985 pxdlist.npxd = 0;
986 pxd = &pxdlist.pxd[0];
987 PXDaddress(pxd, xaddr);
988 PXDlength(pxd, xlen);
989 split->pxdlist = &pxdlist;
990 rc = dtSplitRoot(tid, ip, split, &rmp);
991
992 if (rc)
993 dbFree(ip, xaddr, xlen);
994 else
995 DT_PUTPAGE(rmp);
996
997 DT_PUTPAGE(smp);
998
999 if (!DO_INDEX(ip))
1000 ip->i_size = xlen << sbi->l2bsize;
1001
1002 goto freeKeyName;
1003 }
1004
1005 /*
1006 * extend first leaf page
1007 *
1008 * extend the 1st extent if less than buffer page size
1009 * (dtExtendPage() reurns leaf page unpinned)
1010 */
1011 pxd = &sp->header.self;
1012 xlen = lengthPXD(pxd);
1013 xsize = xlen << sbi->l2bsize;
1014 if (xsize < PSIZE) {
1015 xaddr = addressPXD(pxd);
1016 n = xsize >> L2DTSLOTSIZE;
1017 n -= (n + 31) >> L2DTSLOTSIZE; /* stbl size */
1018 if ((n + sp->header.freecnt) <= split->nslot)
1019 n = xlen + (xlen << 1);
1020 else
1021 n = xlen;
1022
1023 /* Allocate blocks to quota. */
1024 rc = dquot_alloc_block(ip, n);
1025 if (rc)
1026 goto extendOut;
1027 quota_allocation += n;
1028
1029 if ((rc = dbReAlloc(sbi->ipbmap, xaddr, (s64) xlen,
1030 (s64) n, &nxaddr)))
1031 goto extendOut;
1032
1033 pxdlist.maxnpxd = 1;
1034 pxdlist.npxd = 0;
1035 pxd = &pxdlist.pxd[0];
1036 PXDaddress(pxd, nxaddr);
1037 PXDlength(pxd, xlen + n);
1038 split->pxdlist = &pxdlist;
1039 if ((rc = dtExtendPage(tid, ip, split, btstack))) {
1040 nxaddr = addressPXD(pxd);
1041 if (xaddr != nxaddr) {
1042 /* free relocated extent */
1043 xlen = lengthPXD(pxd);
1044 dbFree(ip, nxaddr, (s64) xlen);
1045 } else {
1046 /* free extended delta */
1047 xlen = lengthPXD(pxd) - n;
1048 xaddr = addressPXD(pxd) + xlen;
1049 dbFree(ip, xaddr, (s64) n);
1050 }
1051 } else if (!DO_INDEX(ip))
1052 ip->i_size = lengthPXD(pxd) << sbi->l2bsize;
1053
1054
1055 extendOut:
1056 DT_PUTPAGE(smp);
1057 goto freeKeyName;
1058 }
1059
1060 /*
1061 * split leaf page <sp> into <sp> and a new right page <rp>.
1062 *
1063 * return <rp> pinned and its extent descriptor <rpxd>
1064 */
1065 /*
1066 * allocate new directory page extent and
1067 * new index page(s) to cover page split(s)
1068 *
1069 * allocation hint: ?
1070 */
1071 n = btstack->nsplit;
1072 pxdlist.maxnpxd = pxdlist.npxd = 0;
1073 xlen = sbi->nbperpage;
1074 for (pxd = pxdlist.pxd; n > 0; n--, pxd++) {
1075 if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr)) == 0) {
1076 PXDaddress(pxd, xaddr);
1077 PXDlength(pxd, xlen);
1078 pxdlist.maxnpxd++;
1079 continue;
1080 }
1081
1082 DT_PUTPAGE(smp);
1083
1084 /* undo allocation */
1085 goto splitOut;
1086 }
1087
1088 split->pxdlist = &pxdlist;
1089 if ((rc = dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd))) {
1090 DT_PUTPAGE(smp);
1091
1092 /* undo allocation */
1093 goto splitOut;
1094 }
1095
1096 if (!DO_INDEX(ip))
1097 ip->i_size += PSIZE;
1098
1099 /*
1100 * propagate up the router entry for the leaf page just split
1101 *
1102 * insert a router entry for the new page into the parent page,
1103 * propagate the insert/split up the tree by walking back the stack
1104 * of (bn of parent page, index of child page entry in parent page)
1105 * that were traversed during the search for the page that split.
1106 *
1107 * the propagation of insert/split up the tree stops if the root
1108 * splits or the page inserted into doesn't have to split to hold
1109 * the new entry.
1110 *
1111 * the parent entry for the split page remains the same, and
1112 * a new entry is inserted at its right with the first key and
1113 * block number of the new right page.
1114 *
1115 * There are a maximum of 4 pages pinned at any time:
1116 * two children, left parent and right parent (when the parent splits).
1117 * keep the child pages pinned while working on the parent.
1118 * make sure that all pins are released at exit.
1119 */
1120 while ((parent = BT_POP(btstack)) != NULL) {
1121 /* parent page specified by stack frame <parent> */
1122
1123 /* keep current child pages (<lp>, <rp>) pinned */
1124 lmp = smp;
1125 lp = sp;
1126
1127 /*
1128 * insert router entry in parent for new right child page <rp>
1129 */
1130 /* get the parent page <sp> */
1131 DT_GETPAGE(ip, parent->bn, smp, PSIZE, sp, rc);
1132 if (rc) {
1133 DT_PUTPAGE(lmp);
1134 DT_PUTPAGE(rmp);
1135 goto splitOut;
1136 }
1137
1138 /*
1139 * The new key entry goes ONE AFTER the index of parent entry,
1140 * because the split was to the right.
1141 */
1142 skip = parent->index + 1;
1143
1144 /*
1145 * compute the key for the router entry
1146 *
1147 * key suffix compression:
1148 * for internal pages that have leaf pages as children,
1149 * retain only what's needed to distinguish between
1150 * the new entry and the entry on the page to its left.
1151 * If the keys compare equal, retain the entire key.
1152 *
1153 * note that compression is performed only at computing
1154 * router key at the lowest internal level.
1155 * further compression of the key between pairs of higher
1156 * level internal pages loses too much information and
1157 * the search may fail.
1158 * (e.g., two adjacent leaf pages of {a, ..., x} {xx, ...,}
1159 * results in two adjacent parent entries (a)(xx).
1160 * if split occurs between these two entries, and
1161 * if compression is applied, the router key of parent entry
1162 * of right page (x) will divert search for x into right
1163 * subtree and miss x in the left subtree.)
1164 *
1165 * the entire key must be retained for the next-to-leftmost
1166 * internal key at any level of the tree, or search may fail
1167 * (e.g., ?)
1168 */
1169 switch (rp->header.flag & BT_TYPE) {
1170 case BT_LEAF:
1171 /*
1172 * compute the length of prefix for suffix compression
1173 * between last entry of left page and first entry
1174 * of right page
1175 */
1176 if ((sp->header.flag & BT_ROOT && skip > 1) ||
1177 sp->header.prev != 0 || skip > 1) {
1178 /* compute uppercase router prefix key */
1179 rc = ciGetLeafPrefixKey(lp,
1180 lp->header.nextindex-1,
1181 rp, 0, &key,
1182 sbi->mntflag);
1183 if (rc) {
1184 DT_PUTPAGE(lmp);
1185 DT_PUTPAGE(rmp);
1186 DT_PUTPAGE(smp);
1187 goto splitOut;
1188 }
1189 } else {
1190 /* next to leftmost entry of
1191 lowest internal level */
1192
1193 /* compute uppercase router key */
1194 dtGetKey(rp, 0, &key, sbi->mntflag);
1195 key.name[key.namlen] = 0;
1196
1197 if ((sbi->mntflag & JFS_OS2) == JFS_OS2)
1198 ciToUpper(&key);
1199 }
1200
1201 n = NDTINTERNAL(key.namlen);
1202 break;
1203
1204 case BT_INTERNAL:
1205 dtGetKey(rp, 0, &key, sbi->mntflag);
1206 n = NDTINTERNAL(key.namlen);
1207 break;
1208
1209 default:
1210 jfs_err("dtSplitUp(): UFO!");
1211 break;
1212 }
1213
1214 /* unpin left child page */
1215 DT_PUTPAGE(lmp);
1216
1217 /*
1218 * compute the data for the router entry
1219 */
1220 data->xd = rpxd; /* child page xd */
1221
1222 /*
1223 * parent page is full - split the parent page
1224 */
1225 if (n > sp->header.freecnt) {
1226 /* init for parent page split */
1227 split->mp = smp;
1228 split->index = skip; /* index at insert */
1229 split->nslot = n;
1230 split->key = &key;
1231 /* split->data = data; */
1232
1233 /* unpin right child page */
1234 DT_PUTPAGE(rmp);
1235
1236 /* The split routines insert the new entry,
1237 * acquire txLock as appropriate.
1238 * return <rp> pinned and its block number <rbn>.
1239 */
1240 rc = (sp->header.flag & BT_ROOT) ?
1241 dtSplitRoot(tid, ip, split, &rmp) :
1242 dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd);
1243 if (rc) {
1244 DT_PUTPAGE(smp);
1245 goto splitOut;
1246 }
1247
1248 /* smp and rmp are pinned */
1249 }
1250 /*
1251 * parent page is not full - insert router entry in parent page
1252 */
1253 else {
1254 BT_MARK_DIRTY(smp, ip);
1255 /*
1256 * acquire a transaction lock on the parent page
1257 */
1258 tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1259 dtlck = (struct dt_lock *) & tlck->lock;
1260 ASSERT(dtlck->index == 0);
1261 lv = & dtlck->lv[0];
1262
1263 /* linelock header */
1264 lv->offset = 0;
1265 lv->length = 1;
1266 dtlck->index++;
1267
1268 /* linelock stbl of non-root parent page */
1269 if (!(sp->header.flag & BT_ROOT)) {
1270 lv++;
1271 n = skip >> L2DTSLOTSIZE;
1272 lv->offset = sp->header.stblindex + n;
1273 lv->length =
1274 ((sp->header.nextindex -
1275 1) >> L2DTSLOTSIZE) - n + 1;
1276 dtlck->index++;
1277 }
1278
1279 dtInsertEntry(sp, skip, &key, data, &dtlck);
1280
1281 /* exit propagate up */
1282 break;
1283 }
1284 }
1285
1286 /* unpin current split and its right page */
1287 DT_PUTPAGE(smp);
1288 DT_PUTPAGE(rmp);
1289
1290 /*
1291 * free remaining extents allocated for split
1292 */
1293 splitOut:
1294 n = pxdlist.npxd;
1295 pxd = &pxdlist.pxd[n];
1296 for (; n < pxdlist.maxnpxd; n++, pxd++)
1297 dbFree(ip, addressPXD(pxd), (s64) lengthPXD(pxd));
1298
1299 freeKeyName:
1300 kfree(key.name);
1301
1302 /* Rollback quota allocation */
1303 if (rc && quota_allocation)
1304 dquot_free_block(ip, quota_allocation);
1305
1306 dtSplitUp_Exit:
1307
1308 return rc;
1309}
1310
1311
1312/*
1313 * dtSplitPage()
1314 *
1315 * function: Split a non-root page of a btree.
1316 *
1317 * parameter:
1318 *
1319 * return: 0 - success;
1320 * errno - failure;
1321 * return split and new page pinned;
1322 */
1323static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
1324 struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rpxdp)
1325{
1326 int rc = 0;
1327 struct metapage *smp;
1328 dtpage_t *sp;
1329 struct metapage *rmp;
1330 dtpage_t *rp; /* new right page allocated */
1331 s64 rbn; /* new right page block number */
1332 struct metapage *mp;
1333 dtpage_t *p;
1334 s64 nextbn;
1335 struct pxdlist *pxdlist;
1336 pxd_t *pxd;
1337 int skip, nextindex, half, left, nxt, off, si;
1338 struct ldtentry *ldtentry;
1339 struct idtentry *idtentry;
1340 u8 *stbl;
1341 struct dtslot *f;
1342 int fsi, stblsize;
1343 int n;
1344 struct dt_lock *sdtlck, *rdtlck;
1345 struct tlock *tlck;
1346 struct dt_lock *dtlck;
1347 struct lv *slv, *rlv, *lv;
1348
1349 /* get split page */
1350 smp = split->mp;
1351 sp = DT_PAGE(ip, smp);
1352
1353 /*
1354 * allocate the new right page for the split
1355 */
1356 pxdlist = split->pxdlist;
1357 pxd = &pxdlist->pxd[pxdlist->npxd];
1358 pxdlist->npxd++;
1359 rbn = addressPXD(pxd);
1360 rmp = get_metapage(ip, rbn, PSIZE, 1);
1361 if (rmp == NULL)
1362 return -EIO;
1363
1364 /* Allocate blocks to quota. */
1365 rc = dquot_alloc_block(ip, lengthPXD(pxd));
1366 if (rc) {
1367 release_metapage(rmp);
1368 return rc;
1369 }
1370
1371 jfs_info("dtSplitPage: ip:0x%p smp:0x%p rmp:0x%p", ip, smp, rmp);
1372
1373 BT_MARK_DIRTY(rmp, ip);
1374 /*
1375 * acquire a transaction lock on the new right page
1376 */
1377 tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1378 rdtlck = (struct dt_lock *) & tlck->lock;
1379
1380 rp = (dtpage_t *) rmp->data;
1381 *rpp = rp;
1382 rp->header.self = *pxd;
1383
1384 BT_MARK_DIRTY(smp, ip);
1385 /*
1386 * acquire a transaction lock on the split page
1387 *
1388 * action:
1389 */
1390 tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1391 sdtlck = (struct dt_lock *) & tlck->lock;
1392
1393 /* linelock header of split page */
1394 ASSERT(sdtlck->index == 0);
1395 slv = & sdtlck->lv[0];
1396 slv->offset = 0;
1397 slv->length = 1;
1398 sdtlck->index++;
1399
1400 /*
1401 * initialize/update sibling pointers between sp and rp
1402 */
1403 nextbn = le64_to_cpu(sp->header.next);
1404 rp->header.next = cpu_to_le64(nextbn);
1405 rp->header.prev = cpu_to_le64(addressPXD(&sp->header.self));
1406 sp->header.next = cpu_to_le64(rbn);
1407
1408 /*
1409 * initialize new right page
1410 */
1411 rp->header.flag = sp->header.flag;
1412
1413 /* compute sorted entry table at start of extent data area */
1414 rp->header.nextindex = 0;
1415 rp->header.stblindex = 1;
1416
1417 n = PSIZE >> L2DTSLOTSIZE;
1418 rp->header.maxslot = n;
1419 stblsize = (n + 31) >> L2DTSLOTSIZE; /* in unit of slot */
1420
1421 /* init freelist */
1422 fsi = rp->header.stblindex + stblsize;
1423 rp->header.freelist = fsi;
1424 rp->header.freecnt = rp->header.maxslot - fsi;
1425
1426 /*
1427 * sequential append at tail: append without split
1428 *
1429 * If splitting the last page on a level because of appending
1430 * a entry to it (skip is maxentry), it's likely that the access is
1431 * sequential. Adding an empty page on the side of the level is less
1432 * work and can push the fill factor much higher than normal.
1433 * If we're wrong it's no big deal, we'll just do the split the right
1434 * way next time.
1435 * (It may look like it's equally easy to do a similar hack for
1436 * reverse sorted data, that is, split the tree left,
1437 * but it's not. Be my guest.)
1438 */
1439 if (nextbn == 0 && split->index == sp->header.nextindex) {
1440 /* linelock header + stbl (first slot) of new page */
1441 rlv = & rdtlck->lv[rdtlck->index];
1442 rlv->offset = 0;
1443 rlv->length = 2;
1444 rdtlck->index++;
1445
1446 /*
1447 * initialize freelist of new right page
1448 */
1449 f = &rp->slot[fsi];
1450 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1451 f->next = fsi;
1452 f->next = -1;
1453
1454 /* insert entry at the first entry of the new right page */
1455 dtInsertEntry(rp, 0, split->key, split->data, &rdtlck);
1456
1457 goto out;
1458 }
1459
1460 /*
1461 * non-sequential insert (at possibly middle page)
1462 */
1463
1464 /*
1465 * update prev pointer of previous right sibling page;
1466 */
1467 if (nextbn != 0) {
1468 DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
1469 if (rc) {
1470 discard_metapage(rmp);
1471 return rc;
1472 }
1473
1474 BT_MARK_DIRTY(mp, ip);
1475 /*
1476 * acquire a transaction lock on the next page
1477 */
1478 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
1479 jfs_info("dtSplitPage: tlck = 0x%p, ip = 0x%p, mp=0x%p",
1480 tlck, ip, mp);
1481 dtlck = (struct dt_lock *) & tlck->lock;
1482
1483 /* linelock header of previous right sibling page */
1484 lv = & dtlck->lv[dtlck->index];
1485 lv->offset = 0;
1486 lv->length = 1;
1487 dtlck->index++;
1488
1489 p->header.prev = cpu_to_le64(rbn);
1490
1491 DT_PUTPAGE(mp);
1492 }
1493
1494 /*
1495 * split the data between the split and right pages.
1496 */
1497 skip = split->index;
1498 half = (PSIZE >> L2DTSLOTSIZE) >> 1; /* swag */
1499 left = 0;
1500
1501 /*
1502 * compute fill factor for split pages
1503 *
1504 * <nxt> traces the next entry to move to rp
1505 * <off> traces the next entry to stay in sp
1506 */
1507 stbl = (u8 *) & sp->slot[sp->header.stblindex];
1508 nextindex = sp->header.nextindex;
1509 for (nxt = off = 0; nxt < nextindex; ++off) {
1510 if (off == skip)
1511 /* check for fill factor with new entry size */
1512 n = split->nslot;
1513 else {
1514 si = stbl[nxt];
1515 switch (sp->header.flag & BT_TYPE) {
1516 case BT_LEAF:
1517 ldtentry = (struct ldtentry *) & sp->slot[si];
1518 if (DO_INDEX(ip))
1519 n = NDTLEAF(ldtentry->namlen);
1520 else
1521 n = NDTLEAF_LEGACY(ldtentry->
1522 namlen);
1523 break;
1524
1525 case BT_INTERNAL:
1526 idtentry = (struct idtentry *) & sp->slot[si];
1527 n = NDTINTERNAL(idtentry->namlen);
1528 break;
1529
1530 default:
1531 break;
1532 }
1533
1534 ++nxt; /* advance to next entry to move in sp */
1535 }
1536
1537 left += n;
1538 if (left >= half)
1539 break;
1540 }
1541
1542 /* <nxt> poins to the 1st entry to move */
1543
1544 /*
1545 * move entries to right page
1546 *
1547 * dtMoveEntry() initializes rp and reserves entry for insertion
1548 *
1549 * split page moved out entries are linelocked;
1550 * new/right page moved in entries are linelocked;
1551 */
1552 /* linelock header + stbl of new right page */
1553 rlv = & rdtlck->lv[rdtlck->index];
1554 rlv->offset = 0;
1555 rlv->length = 5;
1556 rdtlck->index++;
1557
1558 dtMoveEntry(sp, nxt, rp, &sdtlck, &rdtlck, DO_INDEX(ip));
1559
1560 sp->header.nextindex = nxt;
1561
1562 /*
1563 * finalize freelist of new right page
1564 */
1565 fsi = rp->header.freelist;
1566 f = &rp->slot[fsi];
1567 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1568 f->next = fsi;
1569 f->next = -1;
1570
1571 /*
1572 * Update directory index table for entries now in right page
1573 */
1574 if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1575 s64 lblock;
1576
1577 mp = NULL;
1578 stbl = DT_GETSTBL(rp);
1579 for (n = 0; n < rp->header.nextindex; n++) {
1580 ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
1581 modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1582 rbn, n, &mp, &lblock);
1583 }
1584 if (mp)
1585 release_metapage(mp);
1586 }
1587
1588 /*
1589 * the skipped index was on the left page,
1590 */
1591 if (skip <= off) {
1592 /* insert the new entry in the split page */
1593 dtInsertEntry(sp, skip, split->key, split->data, &sdtlck);
1594
1595 /* linelock stbl of split page */
1596 if (sdtlck->index >= sdtlck->maxcnt)
1597 sdtlck = (struct dt_lock *) txLinelock(sdtlck);
1598 slv = & sdtlck->lv[sdtlck->index];
1599 n = skip >> L2DTSLOTSIZE;
1600 slv->offset = sp->header.stblindex + n;
1601 slv->length =
1602 ((sp->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
1603 sdtlck->index++;
1604 }
1605 /*
1606 * the skipped index was on the right page,
1607 */
1608 else {
1609 /* adjust the skip index to reflect the new position */
1610 skip -= nxt;
1611
1612 /* insert the new entry in the right page */
1613 dtInsertEntry(rp, skip, split->key, split->data, &rdtlck);
1614 }
1615
1616 out:
1617 *rmpp = rmp;
1618 *rpxdp = *pxd;
1619
1620 return rc;
1621}
1622
1623
1624/*
1625 * dtExtendPage()
1626 *
1627 * function: extend 1st/only directory leaf page
1628 *
1629 * parameter:
1630 *
1631 * return: 0 - success;
1632 * errno - failure;
1633 * return extended page pinned;
1634 */
1635static int dtExtendPage(tid_t tid,
1636 struct inode *ip, struct dtsplit * split, struct btstack * btstack)
1637{
1638 struct super_block *sb = ip->i_sb;
1639 int rc;
1640 struct metapage *smp, *pmp, *mp;
1641 dtpage_t *sp, *pp;
1642 struct pxdlist *pxdlist;
1643 pxd_t *pxd, *tpxd;
1644 int xlen, xsize;
1645 int newstblindex, newstblsize;
1646 int oldstblindex, oldstblsize;
1647 int fsi, last;
1648 struct dtslot *f;
1649 struct btframe *parent;
1650 int n;
1651 struct dt_lock *dtlck;
1652 s64 xaddr, txaddr;
1653 struct tlock *tlck;
1654 struct pxd_lock *pxdlock;
1655 struct lv *lv;
1656 uint type;
1657 struct ldtentry *ldtentry;
1658 u8 *stbl;
1659
1660 /* get page to extend */
1661 smp = split->mp;
1662 sp = DT_PAGE(ip, smp);
1663
1664 /* get parent/root page */
1665 parent = BT_POP(btstack);
1666 DT_GETPAGE(ip, parent->bn, pmp, PSIZE, pp, rc);
1667 if (rc)
1668 return (rc);
1669
1670 /*
1671 * extend the extent
1672 */
1673 pxdlist = split->pxdlist;
1674 pxd = &pxdlist->pxd[pxdlist->npxd];
1675 pxdlist->npxd++;
1676
1677 xaddr = addressPXD(pxd);
1678 tpxd = &sp->header.self;
1679 txaddr = addressPXD(tpxd);
1680 /* in-place extension */
1681 if (xaddr == txaddr) {
1682 type = tlckEXTEND;
1683 }
1684 /* relocation */
1685 else {
1686 type = tlckNEW;
1687
1688 /* save moved extent descriptor for later free */
1689 tlck = txMaplock(tid, ip, tlckDTREE | tlckRELOCATE);
1690 pxdlock = (struct pxd_lock *) & tlck->lock;
1691 pxdlock->flag = mlckFREEPXD;
1692 pxdlock->pxd = sp->header.self;
1693 pxdlock->index = 1;
1694
1695 /*
1696 * Update directory index table to reflect new page address
1697 */
1698 if (DO_INDEX(ip)) {
1699 s64 lblock;
1700
1701 mp = NULL;
1702 stbl = DT_GETSTBL(sp);
1703 for (n = 0; n < sp->header.nextindex; n++) {
1704 ldtentry =
1705 (struct ldtentry *) & sp->slot[stbl[n]];
1706 modify_index(tid, ip,
1707 le32_to_cpu(ldtentry->index),
1708 xaddr, n, &mp, &lblock);
1709 }
1710 if (mp)
1711 release_metapage(mp);
1712 }
1713 }
1714
1715 /*
1716 * extend the page
1717 */
1718 sp->header.self = *pxd;
1719
1720 jfs_info("dtExtendPage: ip:0x%p smp:0x%p sp:0x%p", ip, smp, sp);
1721
1722 BT_MARK_DIRTY(smp, ip);
1723 /*
1724 * acquire a transaction lock on the extended/leaf page
1725 */
1726 tlck = txLock(tid, ip, smp, tlckDTREE | type);
1727 dtlck = (struct dt_lock *) & tlck->lock;
1728 lv = & dtlck->lv[0];
1729
1730 /* update buffer extent descriptor of extended page */
1731 xlen = lengthPXD(pxd);
1732 xsize = xlen << JFS_SBI(sb)->l2bsize;
1733
1734 /*
1735 * copy old stbl to new stbl at start of extended area
1736 */
1737 oldstblindex = sp->header.stblindex;
1738 oldstblsize = (sp->header.maxslot + 31) >> L2DTSLOTSIZE;
1739 newstblindex = sp->header.maxslot;
1740 n = xsize >> L2DTSLOTSIZE;
1741 newstblsize = (n + 31) >> L2DTSLOTSIZE;
1742 memcpy(&sp->slot[newstblindex], &sp->slot[oldstblindex],
1743 sp->header.nextindex);
1744
1745 /*
1746 * in-line extension: linelock old area of extended page
1747 */
1748 if (type == tlckEXTEND) {
1749 /* linelock header */
1750 lv->offset = 0;
1751 lv->length = 1;
1752 dtlck->index++;
1753 lv++;
1754
1755 /* linelock new stbl of extended page */
1756 lv->offset = newstblindex;
1757 lv->length = newstblsize;
1758 }
1759 /*
1760 * relocation: linelock whole relocated area
1761 */
1762 else {
1763 lv->offset = 0;
1764 lv->length = sp->header.maxslot + newstblsize;
1765 }
1766
1767 dtlck->index++;
1768
1769 sp->header.maxslot = n;
1770 sp->header.stblindex = newstblindex;
1771 /* sp->header.nextindex remains the same */
1772
1773 /*
1774 * add old stbl region at head of freelist
1775 */
1776 fsi = oldstblindex;
1777 f = &sp->slot[fsi];
1778 last = sp->header.freelist;
1779 for (n = 0; n < oldstblsize; n++, fsi++, f++) {
1780 f->next = last;
1781 last = fsi;
1782 }
1783 sp->header.freelist = last;
1784 sp->header.freecnt += oldstblsize;
1785
1786 /*
1787 * append free region of newly extended area at tail of freelist
1788 */
1789 /* init free region of newly extended area */
1790 fsi = n = newstblindex + newstblsize;
1791 f = &sp->slot[fsi];
1792 for (fsi++; fsi < sp->header.maxslot; f++, fsi++)
1793 f->next = fsi;
1794 f->next = -1;
1795
1796 /* append new free region at tail of old freelist */
1797 fsi = sp->header.freelist;
1798 if (fsi == -1)
1799 sp->header.freelist = n;
1800 else {
1801 do {
1802 f = &sp->slot[fsi];
1803 fsi = f->next;
1804 } while (fsi != -1);
1805
1806 f->next = n;
1807 }
1808
1809 sp->header.freecnt += sp->header.maxslot - n;
1810
1811 /*
1812 * insert the new entry
1813 */
1814 dtInsertEntry(sp, split->index, split->key, split->data, &dtlck);
1815
1816 BT_MARK_DIRTY(pmp, ip);
1817 /*
1818 * linelock any freeslots residing in old extent
1819 */
1820 if (type == tlckEXTEND) {
1821 n = sp->header.maxslot >> 2;
1822 if (sp->header.freelist < n)
1823 dtLinelockFreelist(sp, n, &dtlck);
1824 }
1825
1826 /*
1827 * update parent entry on the parent/root page
1828 */
1829 /*
1830 * acquire a transaction lock on the parent/root page
1831 */
1832 tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
1833 dtlck = (struct dt_lock *) & tlck->lock;
1834 lv = & dtlck->lv[dtlck->index];
1835
1836 /* linelock parent entry - 1st slot */
1837 lv->offset = 1;
1838 lv->length = 1;
1839 dtlck->index++;
1840
1841 /* update the parent pxd for page extension */
1842 tpxd = (pxd_t *) & pp->slot[1];
1843 *tpxd = *pxd;
1844
1845 DT_PUTPAGE(pmp);
1846 return 0;
1847}
1848
1849
1850/*
1851 * dtSplitRoot()
1852 *
1853 * function:
1854 * split the full root page into
1855 * original/root/split page and new right page
1856 * i.e., root remains fixed in tree anchor (inode) and
1857 * the root is copied to a single new right child page
1858 * since root page << non-root page, and
1859 * the split root page contains a single entry for the
1860 * new right child page.
1861 *
1862 * parameter:
1863 *
1864 * return: 0 - success;
1865 * errno - failure;
1866 * return new page pinned;
1867 */
1868static int dtSplitRoot(tid_t tid,
1869 struct inode *ip, struct dtsplit * split, struct metapage ** rmpp)
1870{
1871 struct super_block *sb = ip->i_sb;
1872 struct metapage *smp;
1873 dtroot_t *sp;
1874 struct metapage *rmp;
1875 dtpage_t *rp;
1876 s64 rbn;
1877 int xlen;
1878 int xsize;
1879 struct dtslot *f;
1880 s8 *stbl;
1881 int fsi, stblsize, n;
1882 struct idtentry *s;
1883 pxd_t *ppxd;
1884 struct pxdlist *pxdlist;
1885 pxd_t *pxd;
1886 struct dt_lock *dtlck;
1887 struct tlock *tlck;
1888 struct lv *lv;
1889 int rc;
1890
1891 /* get split root page */
1892 smp = split->mp;
1893 sp = &JFS_IP(ip)->i_dtroot;
1894
1895 /*
1896 * allocate/initialize a single (right) child page
1897 *
1898 * N.B. at first split, a one (or two) block to fit new entry
1899 * is allocated; at subsequent split, a full page is allocated;
1900 */
1901 pxdlist = split->pxdlist;
1902 pxd = &pxdlist->pxd[pxdlist->npxd];
1903 pxdlist->npxd++;
1904 rbn = addressPXD(pxd);
1905 xlen = lengthPXD(pxd);
1906 xsize = xlen << JFS_SBI(sb)->l2bsize;
1907 rmp = get_metapage(ip, rbn, xsize, 1);
1908 if (!rmp)
1909 return -EIO;
1910
1911 rp = rmp->data;
1912
1913 /* Allocate blocks to quota. */
1914 rc = dquot_alloc_block(ip, lengthPXD(pxd));
1915 if (rc) {
1916 release_metapage(rmp);
1917 return rc;
1918 }
1919
1920 BT_MARK_DIRTY(rmp, ip);
1921 /*
1922 * acquire a transaction lock on the new right page
1923 */
1924 tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1925 dtlck = (struct dt_lock *) & tlck->lock;
1926
1927 rp->header.flag =
1928 (sp->header.flag & BT_LEAF) ? BT_LEAF : BT_INTERNAL;
1929 rp->header.self = *pxd;
1930
1931 /* initialize sibling pointers */
1932 rp->header.next = 0;
1933 rp->header.prev = 0;
1934
1935 /*
1936 * move in-line root page into new right page extent
1937 */
1938 /* linelock header + copied entries + new stbl (1st slot) in new page */
1939 ASSERT(dtlck->index == 0);
1940 lv = & dtlck->lv[0];
1941 lv->offset = 0;
1942 lv->length = 10; /* 1 + 8 + 1 */
1943 dtlck->index++;
1944
1945 n = xsize >> L2DTSLOTSIZE;
1946 rp->header.maxslot = n;
1947 stblsize = (n + 31) >> L2DTSLOTSIZE;
1948
1949 /* copy old stbl to new stbl at start of extended area */
1950 rp->header.stblindex = DTROOTMAXSLOT;
1951 stbl = (s8 *) & rp->slot[DTROOTMAXSLOT];
1952 memcpy(stbl, sp->header.stbl, sp->header.nextindex);
1953 rp->header.nextindex = sp->header.nextindex;
1954
1955 /* copy old data area to start of new data area */
1956 memcpy(&rp->slot[1], &sp->slot[1], IDATASIZE);
1957
1958 /*
1959 * append free region of newly extended area at tail of freelist
1960 */
1961 /* init free region of newly extended area */
1962 fsi = n = DTROOTMAXSLOT + stblsize;
1963 f = &rp->slot[fsi];
1964 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1965 f->next = fsi;
1966 f->next = -1;
1967
1968 /* append new free region at tail of old freelist */
1969 fsi = sp->header.freelist;
1970 if (fsi == -1)
1971 rp->header.freelist = n;
1972 else {
1973 rp->header.freelist = fsi;
1974
1975 do {
1976 f = &rp->slot[fsi];
1977 fsi = f->next;
1978 } while (fsi >= 0);
1979
1980 f->next = n;
1981 }
1982
1983 rp->header.freecnt = sp->header.freecnt + rp->header.maxslot - n;
1984
1985 /*
1986 * Update directory index table for entries now in right page
1987 */
1988 if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1989 s64 lblock;
1990 struct metapage *mp = NULL;
1991 struct ldtentry *ldtentry;
1992
1993 stbl = DT_GETSTBL(rp);
1994 for (n = 0; n < rp->header.nextindex; n++) {
1995 ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
1996 modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1997 rbn, n, &mp, &lblock);
1998 }
1999 if (mp)
2000 release_metapage(mp);
2001 }
2002 /*
2003 * insert the new entry into the new right/child page
2004 * (skip index in the new right page will not change)
2005 */
2006 dtInsertEntry(rp, split->index, split->key, split->data, &dtlck);
2007
2008 /*
2009 * reset parent/root page
2010 *
2011 * set the 1st entry offset to 0, which force the left-most key
2012 * at any level of the tree to be less than any search key.
2013 *
2014 * The btree comparison code guarantees that the left-most key on any
2015 * level of the tree is never used, so it doesn't need to be filled in.
2016 */
2017 BT_MARK_DIRTY(smp, ip);
2018 /*
2019 * acquire a transaction lock on the root page (in-memory inode)
2020 */
2021 tlck = txLock(tid, ip, smp, tlckDTREE | tlckNEW | tlckBTROOT);
2022 dtlck = (struct dt_lock *) & tlck->lock;
2023
2024 /* linelock root */
2025 ASSERT(dtlck->index == 0);
2026 lv = & dtlck->lv[0];
2027 lv->offset = 0;
2028 lv->length = DTROOTMAXSLOT;
2029 dtlck->index++;
2030
2031 /* update page header of root */
2032 if (sp->header.flag & BT_LEAF) {
2033 sp->header.flag &= ~BT_LEAF;
2034 sp->header.flag |= BT_INTERNAL;
2035 }
2036
2037 /* init the first entry */
2038 s = (struct idtentry *) & sp->slot[DTENTRYSTART];
2039 ppxd = (pxd_t *) s;
2040 *ppxd = *pxd;
2041 s->next = -1;
2042 s->namlen = 0;
2043
2044 stbl = sp->header.stbl;
2045 stbl[0] = DTENTRYSTART;
2046 sp->header.nextindex = 1;
2047
2048 /* init freelist */
2049 fsi = DTENTRYSTART + 1;
2050 f = &sp->slot[fsi];
2051
2052 /* init free region of remaining area */
2053 for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2054 f->next = fsi;
2055 f->next = -1;
2056
2057 sp->header.freelist = DTENTRYSTART + 1;
2058 sp->header.freecnt = DTROOTMAXSLOT - (DTENTRYSTART + 1);
2059
2060 *rmpp = rmp;
2061
2062 return 0;
2063}
2064
2065
2066/*
2067 * dtDelete()
2068 *
2069 * function: delete the entry(s) referenced by a key.
2070 *
2071 * parameter:
2072 *
2073 * return:
2074 */
2075int dtDelete(tid_t tid,
2076 struct inode *ip, struct component_name * key, ino_t * ino, int flag)
2077{
2078 int rc = 0;
2079 s64 bn;
2080 struct metapage *mp, *imp;
2081 dtpage_t *p;
2082 int index;
2083 struct btstack btstack;
2084 struct dt_lock *dtlck;
2085 struct tlock *tlck;
2086 struct lv *lv;
2087 int i;
2088 struct ldtentry *ldtentry;
2089 u8 *stbl;
2090 u32 table_index, next_index;
2091 struct metapage *nmp;
2092 dtpage_t *np;
2093
2094 /*
2095 * search for the entry to delete:
2096 *
2097 * dtSearch() returns (leaf page pinned, index at which to delete).
2098 */
2099 if ((rc = dtSearch(ip, key, ino, &btstack, flag)))
2100 return rc;
2101
2102 /* retrieve search result */
2103 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2104
2105 /*
2106 * We need to find put the index of the next entry into the
2107 * directory index table in order to resume a readdir from this
2108 * entry.
2109 */
2110 if (DO_INDEX(ip)) {
2111 stbl = DT_GETSTBL(p);
2112 ldtentry = (struct ldtentry *) & p->slot[stbl[index]];
2113 table_index = le32_to_cpu(ldtentry->index);
2114 if (index == (p->header.nextindex - 1)) {
2115 /*
2116 * Last entry in this leaf page
2117 */
2118 if ((p->header.flag & BT_ROOT)
2119 || (p->header.next == 0))
2120 next_index = -1;
2121 else {
2122 /* Read next leaf page */
2123 DT_GETPAGE(ip, le64_to_cpu(p->header.next),
2124 nmp, PSIZE, np, rc);
2125 if (rc)
2126 next_index = -1;
2127 else {
2128 stbl = DT_GETSTBL(np);
2129 ldtentry =
2130 (struct ldtentry *) & np->
2131 slot[stbl[0]];
2132 next_index =
2133 le32_to_cpu(ldtentry->index);
2134 DT_PUTPAGE(nmp);
2135 }
2136 }
2137 } else {
2138 ldtentry =
2139 (struct ldtentry *) & p->slot[stbl[index + 1]];
2140 next_index = le32_to_cpu(ldtentry->index);
2141 }
2142 free_index(tid, ip, table_index, next_index);
2143 }
2144 /*
2145 * the leaf page becomes empty, delete the page
2146 */
2147 if (p->header.nextindex == 1) {
2148 /* delete empty page */
2149 rc = dtDeleteUp(tid, ip, mp, p, &btstack);
2150 }
2151 /*
2152 * the leaf page has other entries remaining:
2153 *
2154 * delete the entry from the leaf page.
2155 */
2156 else {
2157 BT_MARK_DIRTY(mp, ip);
2158 /*
2159 * acquire a transaction lock on the leaf page
2160 */
2161 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2162 dtlck = (struct dt_lock *) & tlck->lock;
2163
2164 /*
2165 * Do not assume that dtlck->index will be zero. During a
2166 * rename within a directory, this transaction may have
2167 * modified this page already when adding the new entry.
2168 */
2169
2170 /* linelock header */
2171 if (dtlck->index >= dtlck->maxcnt)
2172 dtlck = (struct dt_lock *) txLinelock(dtlck);
2173 lv = & dtlck->lv[dtlck->index];
2174 lv->offset = 0;
2175 lv->length = 1;
2176 dtlck->index++;
2177
2178 /* linelock stbl of non-root leaf page */
2179 if (!(p->header.flag & BT_ROOT)) {
2180 if (dtlck->index >= dtlck->maxcnt)
2181 dtlck = (struct dt_lock *) txLinelock(dtlck);
2182 lv = & dtlck->lv[dtlck->index];
2183 i = index >> L2DTSLOTSIZE;
2184 lv->offset = p->header.stblindex + i;
2185 lv->length =
2186 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2187 i + 1;
2188 dtlck->index++;
2189 }
2190
2191 /* free the leaf entry */
2192 dtDeleteEntry(p, index, &dtlck);
2193
2194 /*
2195 * Update directory index table for entries moved in stbl
2196 */
2197 if (DO_INDEX(ip) && index < p->header.nextindex) {
2198 s64 lblock;
2199
2200 imp = NULL;
2201 stbl = DT_GETSTBL(p);
2202 for (i = index; i < p->header.nextindex; i++) {
2203 ldtentry =
2204 (struct ldtentry *) & p->slot[stbl[i]];
2205 modify_index(tid, ip,
2206 le32_to_cpu(ldtentry->index),
2207 bn, i, &imp, &lblock);
2208 }
2209 if (imp)
2210 release_metapage(imp);
2211 }
2212
2213 DT_PUTPAGE(mp);
2214 }
2215
2216 return rc;
2217}
2218
2219
2220/*
2221 * dtDeleteUp()
2222 *
2223 * function:
2224 * free empty pages as propagating deletion up the tree
2225 *
2226 * parameter:
2227 *
2228 * return:
2229 */
2230static int dtDeleteUp(tid_t tid, struct inode *ip,
2231 struct metapage * fmp, dtpage_t * fp, struct btstack * btstack)
2232{
2233 int rc = 0;
2234 struct metapage *mp;
2235 dtpage_t *p;
2236 int index, nextindex;
2237 int xlen;
2238 struct btframe *parent;
2239 struct dt_lock *dtlck;
2240 struct tlock *tlck;
2241 struct lv *lv;
2242 struct pxd_lock *pxdlock;
2243 int i;
2244
2245 /*
2246 * keep the root leaf page which has become empty
2247 */
2248 if (BT_IS_ROOT(fmp)) {
2249 /*
2250 * reset the root
2251 *
2252 * dtInitRoot() acquires txlock on the root
2253 */
2254 dtInitRoot(tid, ip, PARENT(ip));
2255
2256 DT_PUTPAGE(fmp);
2257
2258 return 0;
2259 }
2260
2261 /*
2262 * free the non-root leaf page
2263 */
2264 /*
2265 * acquire a transaction lock on the page
2266 *
2267 * write FREEXTENT|NOREDOPAGE log record
2268 * N.B. linelock is overlaid as freed extent descriptor, and
2269 * the buffer page is freed;
2270 */
2271 tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2272 pxdlock = (struct pxd_lock *) & tlck->lock;
2273 pxdlock->flag = mlckFREEPXD;
2274 pxdlock->pxd = fp->header.self;
2275 pxdlock->index = 1;
2276
2277 /* update sibling pointers */
2278 if ((rc = dtRelink(tid, ip, fp))) {
2279 BT_PUTPAGE(fmp);
2280 return rc;
2281 }
2282
2283 xlen = lengthPXD(&fp->header.self);
2284
2285 /* Free quota allocation. */
2286 dquot_free_block(ip, xlen);
2287
2288 /* free/invalidate its buffer page */
2289 discard_metapage(fmp);
2290
2291 /*
2292 * propagate page deletion up the directory tree
2293 *
2294 * If the delete from the parent page makes it empty,
2295 * continue all the way up the tree.
2296 * stop if the root page is reached (which is never deleted) or
2297 * if the entry deletion does not empty the page.
2298 */
2299 while ((parent = BT_POP(btstack)) != NULL) {
2300 /* pin the parent page <sp> */
2301 DT_GETPAGE(ip, parent->bn, mp, PSIZE, p, rc);
2302 if (rc)
2303 return rc;
2304
2305 /*
2306 * free the extent of the child page deleted
2307 */
2308 index = parent->index;
2309
2310 /*
2311 * delete the entry for the child page from parent
2312 */
2313 nextindex = p->header.nextindex;
2314
2315 /*
2316 * the parent has the single entry being deleted:
2317 *
2318 * free the parent page which has become empty.
2319 */
2320 if (nextindex == 1) {
2321 /*
2322 * keep the root internal page which has become empty
2323 */
2324 if (p->header.flag & BT_ROOT) {
2325 /*
2326 * reset the root
2327 *
2328 * dtInitRoot() acquires txlock on the root
2329 */
2330 dtInitRoot(tid, ip, PARENT(ip));
2331
2332 DT_PUTPAGE(mp);
2333
2334 return 0;
2335 }
2336 /*
2337 * free the parent page
2338 */
2339 else {
2340 /*
2341 * acquire a transaction lock on the page
2342 *
2343 * write FREEXTENT|NOREDOPAGE log record
2344 */
2345 tlck =
2346 txMaplock(tid, ip,
2347 tlckDTREE | tlckFREE);
2348 pxdlock = (struct pxd_lock *) & tlck->lock;
2349 pxdlock->flag = mlckFREEPXD;
2350 pxdlock->pxd = p->header.self;
2351 pxdlock->index = 1;
2352
2353 /* update sibling pointers */
2354 if ((rc = dtRelink(tid, ip, p))) {
2355 DT_PUTPAGE(mp);
2356 return rc;
2357 }
2358
2359 xlen = lengthPXD(&p->header.self);
2360
2361 /* Free quota allocation */
2362 dquot_free_block(ip, xlen);
2363
2364 /* free/invalidate its buffer page */
2365 discard_metapage(mp);
2366
2367 /* propagate up */
2368 continue;
2369 }
2370 }
2371
2372 /*
2373 * the parent has other entries remaining:
2374 *
2375 * delete the router entry from the parent page.
2376 */
2377 BT_MARK_DIRTY(mp, ip);
2378 /*
2379 * acquire a transaction lock on the page
2380 *
2381 * action: router entry deletion
2382 */
2383 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2384 dtlck = (struct dt_lock *) & tlck->lock;
2385
2386 /* linelock header */
2387 if (dtlck->index >= dtlck->maxcnt)
2388 dtlck = (struct dt_lock *) txLinelock(dtlck);
2389 lv = & dtlck->lv[dtlck->index];
2390 lv->offset = 0;
2391 lv->length = 1;
2392 dtlck->index++;
2393
2394 /* linelock stbl of non-root leaf page */
2395 if (!(p->header.flag & BT_ROOT)) {
2396 if (dtlck->index < dtlck->maxcnt)
2397 lv++;
2398 else {
2399 dtlck = (struct dt_lock *) txLinelock(dtlck);
2400 lv = & dtlck->lv[0];
2401 }
2402 i = index >> L2DTSLOTSIZE;
2403 lv->offset = p->header.stblindex + i;
2404 lv->length =
2405 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2406 i + 1;
2407 dtlck->index++;
2408 }
2409
2410 /* free the router entry */
2411 dtDeleteEntry(p, index, &dtlck);
2412
2413 /* reset key of new leftmost entry of level (for consistency) */
2414 if (index == 0 &&
2415 ((p->header.flag & BT_ROOT) || p->header.prev == 0))
2416 dtTruncateEntry(p, 0, &dtlck);
2417
2418 /* unpin the parent page */
2419 DT_PUTPAGE(mp);
2420
2421 /* exit propagation up */
2422 break;
2423 }
2424
2425 if (!DO_INDEX(ip))
2426 ip->i_size -= PSIZE;
2427
2428 return 0;
2429}
2430
2431/*
2432 * dtRelink()
2433 *
2434 * function:
2435 * link around a freed page.
2436 *
2437 * parameter:
2438 * fp: page to be freed
2439 *
2440 * return:
2441 */
2442static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p)
2443{
2444 int rc;
2445 struct metapage *mp;
2446 s64 nextbn, prevbn;
2447 struct tlock *tlck;
2448 struct dt_lock *dtlck;
2449 struct lv *lv;
2450
2451 nextbn = le64_to_cpu(p->header.next);
2452 prevbn = le64_to_cpu(p->header.prev);
2453
2454 /* update prev pointer of the next page */
2455 if (nextbn != 0) {
2456 DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
2457 if (rc)
2458 return rc;
2459
2460 BT_MARK_DIRTY(mp, ip);
2461 /*
2462 * acquire a transaction lock on the next page
2463 *
2464 * action: update prev pointer;
2465 */
2466 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2467 jfs_info("dtRelink nextbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2468 tlck, ip, mp);
2469 dtlck = (struct dt_lock *) & tlck->lock;
2470
2471 /* linelock header */
2472 if (dtlck->index >= dtlck->maxcnt)
2473 dtlck = (struct dt_lock *) txLinelock(dtlck);
2474 lv = & dtlck->lv[dtlck->index];
2475 lv->offset = 0;
2476 lv->length = 1;
2477 dtlck->index++;
2478
2479 p->header.prev = cpu_to_le64(prevbn);
2480 DT_PUTPAGE(mp);
2481 }
2482
2483 /* update next pointer of the previous page */
2484 if (prevbn != 0) {
2485 DT_GETPAGE(ip, prevbn, mp, PSIZE, p, rc);
2486 if (rc)
2487 return rc;
2488
2489 BT_MARK_DIRTY(mp, ip);
2490 /*
2491 * acquire a transaction lock on the prev page
2492 *
2493 * action: update next pointer;
2494 */
2495 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2496 jfs_info("dtRelink prevbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2497 tlck, ip, mp);
2498 dtlck = (struct dt_lock *) & tlck->lock;
2499
2500 /* linelock header */
2501 if (dtlck->index >= dtlck->maxcnt)
2502 dtlck = (struct dt_lock *) txLinelock(dtlck);
2503 lv = & dtlck->lv[dtlck->index];
2504 lv->offset = 0;
2505 lv->length = 1;
2506 dtlck->index++;
2507
2508 p->header.next = cpu_to_le64(nextbn);
2509 DT_PUTPAGE(mp);
2510 }
2511
2512 return 0;
2513}
2514
2515
2516/*
2517 * dtInitRoot()
2518 *
2519 * initialize directory root (inline in inode)
2520 */
2521void dtInitRoot(tid_t tid, struct inode *ip, u32 idotdot)
2522{
2523 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
2524 dtroot_t *p;
2525 int fsi;
2526 struct dtslot *f;
2527 struct tlock *tlck;
2528 struct dt_lock *dtlck;
2529 struct lv *lv;
2530 u16 xflag_save;
2531
2532 /*
2533 * If this was previously an non-empty directory, we need to remove
2534 * the old directory table.
2535 */
2536 if (DO_INDEX(ip)) {
2537 if (!jfs_dirtable_inline(ip)) {
2538 struct tblock *tblk = tid_to_tblock(tid);
2539 /*
2540 * We're playing games with the tid's xflag. If
2541 * we're removing a regular file, the file's xtree
2542 * is committed with COMMIT_PMAP, but we always
2543 * commit the directories xtree with COMMIT_PWMAP.
2544 */
2545 xflag_save = tblk->xflag;
2546 tblk->xflag = 0;
2547 /*
2548 * xtTruncate isn't guaranteed to fully truncate
2549 * the xtree. The caller needs to check i_size
2550 * after committing the transaction to see if
2551 * additional truncation is needed. The
2552 * COMMIT_Stale flag tells caller that we
2553 * initiated the truncation.
2554 */
2555 xtTruncate(tid, ip, 0, COMMIT_PWMAP);
2556 set_cflag(COMMIT_Stale, ip);
2557
2558 tblk->xflag = xflag_save;
2559 } else
2560 ip->i_size = 1;
2561
2562 jfs_ip->next_index = 2;
2563 } else
2564 ip->i_size = IDATASIZE;
2565
2566 /*
2567 * acquire a transaction lock on the root
2568 *
2569 * action: directory initialization;
2570 */
2571 tlck = txLock(tid, ip, (struct metapage *) & jfs_ip->bxflag,
2572 tlckDTREE | tlckENTRY | tlckBTROOT);
2573 dtlck = (struct dt_lock *) & tlck->lock;
2574
2575 /* linelock root */
2576 ASSERT(dtlck->index == 0);
2577 lv = & dtlck->lv[0];
2578 lv->offset = 0;
2579 lv->length = DTROOTMAXSLOT;
2580 dtlck->index++;
2581
2582 p = &jfs_ip->i_dtroot;
2583
2584 p->header.flag = DXD_INDEX | BT_ROOT | BT_LEAF;
2585
2586 p->header.nextindex = 0;
2587
2588 /* init freelist */
2589 fsi = 1;
2590 f = &p->slot[fsi];
2591
2592 /* init data area of root */
2593 for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2594 f->next = fsi;
2595 f->next = -1;
2596
2597 p->header.freelist = 1;
2598 p->header.freecnt = 8;
2599
2600 /* init '..' entry */
2601 p->header.idotdot = cpu_to_le32(idotdot);
2602
2603 return;
2604}
2605
2606/*
2607 * add_missing_indices()
2608 *
2609 * function: Fix dtree page in which one or more entries has an invalid index.
2610 * fsck.jfs should really fix this, but it currently does not.
2611 * Called from jfs_readdir when bad index is detected.
2612 */
2613static void add_missing_indices(struct inode *inode, s64 bn)
2614{
2615 struct ldtentry *d;
2616 struct dt_lock *dtlck;
2617 int i;
2618 uint index;
2619 struct lv *lv;
2620 struct metapage *mp;
2621 dtpage_t *p;
2622 int rc;
2623 s8 *stbl;
2624 tid_t tid;
2625 struct tlock *tlck;
2626
2627 tid = txBegin(inode->i_sb, 0);
2628
2629 DT_GETPAGE(inode, bn, mp, PSIZE, p, rc);
2630
2631 if (rc) {
2632 printk(KERN_ERR "DT_GETPAGE failed!\n");
2633 goto end;
2634 }
2635 BT_MARK_DIRTY(mp, inode);
2636
2637 ASSERT(p->header.flag & BT_LEAF);
2638
2639 tlck = txLock(tid, inode, mp, tlckDTREE | tlckENTRY);
2640 if (BT_IS_ROOT(mp))
2641 tlck->type |= tlckBTROOT;
2642
2643 dtlck = (struct dt_lock *) &tlck->lock;
2644
2645 stbl = DT_GETSTBL(p);
2646 for (i = 0; i < p->header.nextindex; i++) {
2647 d = (struct ldtentry *) &p->slot[stbl[i]];
2648 index = le32_to_cpu(d->index);
2649 if ((index < 2) || (index >= JFS_IP(inode)->next_index)) {
2650 d->index = cpu_to_le32(add_index(tid, inode, bn, i));
2651 if (dtlck->index >= dtlck->maxcnt)
2652 dtlck = (struct dt_lock *) txLinelock(dtlck);
2653 lv = &dtlck->lv[dtlck->index];
2654 lv->offset = stbl[i];
2655 lv->length = 1;
2656 dtlck->index++;
2657 }
2658 }
2659
2660 DT_PUTPAGE(mp);
2661 (void) txCommit(tid, 1, &inode, 0);
2662end:
2663 txEnd(tid);
2664}
2665
2666/*
2667 * Buffer to hold directory entry info while traversing a dtree page
2668 * before being fed to the filldir function
2669 */
2670struct jfs_dirent {
2671 loff_t position;
2672 int ino;
2673 u16 name_len;
2674 char name[];
2675};
2676
2677/*
2678 * function to determine next variable-sized jfs_dirent in buffer
2679 */
2680static inline struct jfs_dirent *next_jfs_dirent(struct jfs_dirent *dirent)
2681{
2682 return (struct jfs_dirent *)
2683 ((char *)dirent +
2684 ((sizeof (struct jfs_dirent) + dirent->name_len + 1 +
2685 sizeof (loff_t) - 1) &
2686 ~(sizeof (loff_t) - 1)));
2687}
2688
2689/*
2690 * jfs_readdir()
2691 *
2692 * function: read directory entries sequentially
2693 * from the specified entry offset
2694 *
2695 * parameter:
2696 *
2697 * return: offset = (pn, index) of start entry
2698 * of next jfs_readdir()/dtRead()
2699 */
2700int jfs_readdir(struct file *file, struct dir_context *ctx)
2701{
2702 struct inode *ip = file_inode(file);
2703 struct nls_table *codepage = JFS_SBI(ip->i_sb)->nls_tab;
2704 int rc = 0;
2705 loff_t dtpos; /* legacy OS/2 style position */
2706 struct dtoffset {
2707 s16 pn;
2708 s16 index;
2709 s32 unused;
2710 } *dtoffset = (struct dtoffset *) &dtpos;
2711 s64 bn;
2712 struct metapage *mp;
2713 dtpage_t *p;
2714 int index;
2715 s8 *stbl;
2716 struct btstack btstack;
2717 int i, next;
2718 struct ldtentry *d;
2719 struct dtslot *t;
2720 int d_namleft, len, outlen;
2721 unsigned long dirent_buf;
2722 char *name_ptr;
2723 u32 dir_index;
2724 int do_index = 0;
2725 uint loop_count = 0;
2726 struct jfs_dirent *jfs_dirent;
2727 int jfs_dirents;
2728 int overflow, fix_page, page_fixed = 0;
2729 static int unique_pos = 2; /* If we can't fix broken index */
2730
2731 if (ctx->pos == DIREND)
2732 return 0;
2733
2734 if (DO_INDEX(ip)) {
2735 /*
2736 * persistent index is stored in directory entries.
2737 * Special cases: 0 = .
2738 * 1 = ..
2739 * -1 = End of directory
2740 */
2741 do_index = 1;
2742
2743 dir_index = (u32) ctx->pos;
2744
2745 /*
2746 * NFSv4 reserves cookies 1 and 2 for . and .. so the value
2747 * we return to the vfs is one greater than the one we use
2748 * internally.
2749 */
2750 if (dir_index)
2751 dir_index--;
2752
2753 if (dir_index > 1) {
2754 struct dir_table_slot dirtab_slot;
2755
2756 if (dtEmpty(ip) ||
2757 (dir_index >= JFS_IP(ip)->next_index)) {
2758 /* Stale position. Directory has shrunk */
2759 ctx->pos = DIREND;
2760 return 0;
2761 }
2762 repeat:
2763 rc = read_index(ip, dir_index, &dirtab_slot);
2764 if (rc) {
2765 ctx->pos = DIREND;
2766 return rc;
2767 }
2768 if (dirtab_slot.flag == DIR_INDEX_FREE) {
2769 if (loop_count++ > JFS_IP(ip)->next_index) {
2770 jfs_err("jfs_readdir detected infinite loop!");
2771 ctx->pos = DIREND;
2772 return 0;
2773 }
2774 dir_index = le32_to_cpu(dirtab_slot.addr2);
2775 if (dir_index == -1) {
2776 ctx->pos = DIREND;
2777 return 0;
2778 }
2779 goto repeat;
2780 }
2781 bn = addressDTS(&dirtab_slot);
2782 index = dirtab_slot.slot;
2783 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2784 if (rc) {
2785 ctx->pos = DIREND;
2786 return 0;
2787 }
2788 if (p->header.flag & BT_INTERNAL) {
2789 jfs_err("jfs_readdir: bad index table");
2790 DT_PUTPAGE(mp);
2791 ctx->pos = DIREND;
2792 return 0;
2793 }
2794 } else {
2795 if (dir_index == 0) {
2796 /*
2797 * self "."
2798 */
2799 ctx->pos = 1;
2800 if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
2801 return 0;
2802 }
2803 /*
2804 * parent ".."
2805 */
2806 ctx->pos = 2;
2807 if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
2808 return 0;
2809
2810 /*
2811 * Find first entry of left-most leaf
2812 */
2813 if (dtEmpty(ip)) {
2814 ctx->pos = DIREND;
2815 return 0;
2816 }
2817
2818 if ((rc = dtReadFirst(ip, &btstack)))
2819 return rc;
2820
2821 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2822 }
2823 } else {
2824 /*
2825 * Legacy filesystem - OS/2 & Linux JFS < 0.3.6
2826 *
2827 * pn = 0; index = 1: First entry "."
2828 * pn = 0; index = 2: Second entry ".."
2829 * pn > 0: Real entries, pn=1 -> leftmost page
2830 * pn = index = -1: No more entries
2831 */
2832 dtpos = ctx->pos;
2833 if (dtpos < 2) {
2834 /* build "." entry */
2835 ctx->pos = 1;
2836 if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
2837 return 0;
2838 dtoffset->index = 2;
2839 ctx->pos = dtpos;
2840 }
2841
2842 if (dtoffset->pn == 0) {
2843 if (dtoffset->index == 2) {
2844 /* build ".." entry */
2845 if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
2846 return 0;
2847 } else {
2848 jfs_err("jfs_readdir called with invalid offset!");
2849 }
2850 dtoffset->pn = 1;
2851 dtoffset->index = 0;
2852 ctx->pos = dtpos;
2853 }
2854
2855 if (dtEmpty(ip)) {
2856 ctx->pos = DIREND;
2857 return 0;
2858 }
2859
2860 if ((rc = dtReadNext(ip, &ctx->pos, &btstack))) {
2861 jfs_err("jfs_readdir: unexpected rc = %d from dtReadNext",
2862 rc);
2863 ctx->pos = DIREND;
2864 return 0;
2865 }
2866 /* get start leaf page and index */
2867 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2868
2869 /* offset beyond directory eof ? */
2870 if (bn < 0) {
2871 ctx->pos = DIREND;
2872 return 0;
2873 }
2874 }
2875
2876 dirent_buf = __get_free_page(GFP_KERNEL);
2877 if (dirent_buf == 0) {
2878 DT_PUTPAGE(mp);
2879 jfs_warn("jfs_readdir: __get_free_page failed!");
2880 ctx->pos = DIREND;
2881 return -ENOMEM;
2882 }
2883
2884 while (1) {
2885 jfs_dirent = (struct jfs_dirent *) dirent_buf;
2886 jfs_dirents = 0;
2887 overflow = fix_page = 0;
2888
2889 stbl = DT_GETSTBL(p);
2890
2891 for (i = index; i < p->header.nextindex; i++) {
2892 d = (struct ldtentry *) & p->slot[stbl[i]];
2893
2894 if (((long) jfs_dirent + d->namlen + 1) >
2895 (dirent_buf + PAGE_SIZE)) {
2896 /* DBCS codepages could overrun dirent_buf */
2897 index = i;
2898 overflow = 1;
2899 break;
2900 }
2901
2902 d_namleft = d->namlen;
2903 name_ptr = jfs_dirent->name;
2904 jfs_dirent->ino = le32_to_cpu(d->inumber);
2905
2906 if (do_index) {
2907 len = min(d_namleft, DTLHDRDATALEN);
2908 jfs_dirent->position = le32_to_cpu(d->index);
2909 /*
2910 * d->index should always be valid, but it
2911 * isn't. fsck.jfs doesn't create the
2912 * directory index for the lost+found
2913 * directory. Rather than let it go,
2914 * we can try to fix it.
2915 */
2916 if ((jfs_dirent->position < 2) ||
2917 (jfs_dirent->position >=
2918 JFS_IP(ip)->next_index)) {
2919 if (!page_fixed && !isReadOnly(ip)) {
2920 fix_page = 1;
2921 /*
2922 * setting overflow and setting
2923 * index to i will cause the
2924 * same page to be processed
2925 * again starting here
2926 */
2927 overflow = 1;
2928 index = i;
2929 break;
2930 }
2931 jfs_dirent->position = unique_pos++;
2932 }
2933 /*
2934 * We add 1 to the index because we may
2935 * use a value of 2 internally, and NFSv4
2936 * doesn't like that.
2937 */
2938 jfs_dirent->position++;
2939 } else {
2940 jfs_dirent->position = dtpos;
2941 len = min(d_namleft, DTLHDRDATALEN_LEGACY);
2942 }
2943
2944 /* copy the name of head/only segment */
2945 outlen = jfs_strfromUCS_le(name_ptr, d->name, len,
2946 codepage);
2947 jfs_dirent->name_len = outlen;
2948
2949 /* copy name in the additional segment(s) */
2950 next = d->next;
2951 while (next >= 0) {
2952 t = (struct dtslot *) & p->slot[next];
2953 name_ptr += outlen;
2954 d_namleft -= len;
2955 /* Sanity Check */
2956 if (d_namleft == 0) {
2957 jfs_error(ip->i_sb,
2958 "JFS:Dtree error: ino = %ld, bn=%lld, index = %d\n",
2959 (long)ip->i_ino,
2960 (long long)bn,
2961 i);
2962 goto skip_one;
2963 }
2964 len = min(d_namleft, DTSLOTDATALEN);
2965 outlen = jfs_strfromUCS_le(name_ptr, t->name,
2966 len, codepage);
2967 jfs_dirent->name_len += outlen;
2968
2969 next = t->next;
2970 }
2971
2972 jfs_dirents++;
2973 jfs_dirent = next_jfs_dirent(jfs_dirent);
2974skip_one:
2975 if (!do_index)
2976 dtoffset->index++;
2977 }
2978
2979 if (!overflow) {
2980 /* Point to next leaf page */
2981 if (p->header.flag & BT_ROOT)
2982 bn = 0;
2983 else {
2984 bn = le64_to_cpu(p->header.next);
2985 index = 0;
2986 /* update offset (pn:index) for new page */
2987 if (!do_index) {
2988 dtoffset->pn++;
2989 dtoffset->index = 0;
2990 }
2991 }
2992 page_fixed = 0;
2993 }
2994
2995 /* unpin previous leaf page */
2996 DT_PUTPAGE(mp);
2997
2998 jfs_dirent = (struct jfs_dirent *) dirent_buf;
2999 while (jfs_dirents--) {
3000 ctx->pos = jfs_dirent->position;
3001 if (!dir_emit(ctx, jfs_dirent->name,
3002 jfs_dirent->name_len,
3003 jfs_dirent->ino, DT_UNKNOWN))
3004 goto out;
3005 jfs_dirent = next_jfs_dirent(jfs_dirent);
3006 }
3007
3008 if (fix_page) {
3009 add_missing_indices(ip, bn);
3010 page_fixed = 1;
3011 }
3012
3013 if (!overflow && (bn == 0)) {
3014 ctx->pos = DIREND;
3015 break;
3016 }
3017
3018 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3019 if (rc) {
3020 free_page(dirent_buf);
3021 return rc;
3022 }
3023 }
3024
3025 out:
3026 free_page(dirent_buf);
3027
3028 return rc;
3029}
3030
3031
3032/*
3033 * dtReadFirst()
3034 *
3035 * function: get the leftmost page of the directory
3036 */
3037static int dtReadFirst(struct inode *ip, struct btstack * btstack)
3038{
3039 int rc = 0;
3040 s64 bn;
3041 int psize = 288; /* initial in-line directory */
3042 struct metapage *mp;
3043 dtpage_t *p;
3044 s8 *stbl;
3045 struct btframe *btsp;
3046 pxd_t *xd;
3047
3048 BT_CLR(btstack); /* reset stack */
3049
3050 /*
3051 * descend leftmost path of the tree
3052 *
3053 * by convention, root bn = 0.
3054 */
3055 for (bn = 0;;) {
3056 DT_GETPAGE(ip, bn, mp, psize, p, rc);
3057 if (rc)
3058 return rc;
3059
3060 /*
3061 * leftmost leaf page
3062 */
3063 if (p->header.flag & BT_LEAF) {
3064 /* return leftmost entry */
3065 btsp = btstack->top;
3066 btsp->bn = bn;
3067 btsp->index = 0;
3068 btsp->mp = mp;
3069
3070 return 0;
3071 }
3072
3073 /*
3074 * descend down to leftmost child page
3075 */
3076 if (BT_STACK_FULL(btstack)) {
3077 DT_PUTPAGE(mp);
3078 jfs_error(ip->i_sb, "btstack overrun\n");
3079 BT_STACK_DUMP(btstack);
3080 return -EIO;
3081 }
3082 /* push (bn, index) of the parent page/entry */
3083 BT_PUSH(btstack, bn, 0);
3084
3085 /* get the leftmost entry */
3086 stbl = DT_GETSTBL(p);
3087 xd = (pxd_t *) & p->slot[stbl[0]];
3088
3089 /* get the child page block address */
3090 bn = addressPXD(xd);
3091 psize = lengthPXD(xd) << JFS_SBI(ip->i_sb)->l2bsize;
3092
3093 /* unpin the parent page */
3094 DT_PUTPAGE(mp);
3095 }
3096}
3097
3098
3099/*
3100 * dtReadNext()
3101 *
3102 * function: get the page of the specified offset (pn:index)
3103 *
3104 * return: if (offset > eof), bn = -1;
3105 *
3106 * note: if index > nextindex of the target leaf page,
3107 * start with 1st entry of next leaf page;
3108 */
3109static int dtReadNext(struct inode *ip, loff_t * offset,
3110 struct btstack * btstack)
3111{
3112 int rc = 0;
3113 struct dtoffset {
3114 s16 pn;
3115 s16 index;
3116 s32 unused;
3117 } *dtoffset = (struct dtoffset *) offset;
3118 s64 bn;
3119 struct metapage *mp;
3120 dtpage_t *p;
3121 int index;
3122 int pn;
3123 s8 *stbl;
3124 struct btframe *btsp, *parent;
3125 pxd_t *xd;
3126
3127 /*
3128 * get leftmost leaf page pinned
3129 */
3130 if ((rc = dtReadFirst(ip, btstack)))
3131 return rc;
3132
3133 /* get leaf page */
3134 DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
3135
3136 /* get the start offset (pn:index) */
3137 pn = dtoffset->pn - 1; /* Now pn = 0 represents leftmost leaf */
3138 index = dtoffset->index;
3139
3140 /* start at leftmost page ? */
3141 if (pn == 0) {
3142 /* offset beyond eof ? */
3143 if (index < p->header.nextindex)
3144 goto out;
3145
3146 if (p->header.flag & BT_ROOT) {
3147 bn = -1;
3148 goto out;
3149 }
3150
3151 /* start with 1st entry of next leaf page */
3152 dtoffset->pn++;
3153 dtoffset->index = index = 0;
3154 goto a;
3155 }
3156
3157 /* start at non-leftmost page: scan parent pages for large pn */
3158 if (p->header.flag & BT_ROOT) {
3159 bn = -1;
3160 goto out;
3161 }
3162
3163 /* start after next leaf page ? */
3164 if (pn > 1)
3165 goto b;
3166
3167 /* get leaf page pn = 1 */
3168 a:
3169 bn = le64_to_cpu(p->header.next);
3170
3171 /* unpin leaf page */
3172 DT_PUTPAGE(mp);
3173
3174 /* offset beyond eof ? */
3175 if (bn == 0) {
3176 bn = -1;
3177 goto out;
3178 }
3179
3180 goto c;
3181
3182 /*
3183 * scan last internal page level to get target leaf page
3184 */
3185 b:
3186 /* unpin leftmost leaf page */
3187 DT_PUTPAGE(mp);
3188
3189 /* get left most parent page */
3190 btsp = btstack->top;
3191 parent = btsp - 1;
3192 bn = parent->bn;
3193 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3194 if (rc)
3195 return rc;
3196
3197 /* scan parent pages at last internal page level */
3198 while (pn >= p->header.nextindex) {
3199 pn -= p->header.nextindex;
3200
3201 /* get next parent page address */
3202 bn = le64_to_cpu(p->header.next);
3203
3204 /* unpin current parent page */
3205 DT_PUTPAGE(mp);
3206
3207 /* offset beyond eof ? */
3208 if (bn == 0) {
3209 bn = -1;
3210 goto out;
3211 }
3212
3213 /* get next parent page */
3214 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3215 if (rc)
3216 return rc;
3217
3218 /* update parent page stack frame */
3219 parent->bn = bn;
3220 }
3221
3222 /* get leaf page address */
3223 stbl = DT_GETSTBL(p);
3224 xd = (pxd_t *) & p->slot[stbl[pn]];
3225 bn = addressPXD(xd);
3226
3227 /* unpin parent page */
3228 DT_PUTPAGE(mp);
3229
3230 /*
3231 * get target leaf page
3232 */
3233 c:
3234 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3235 if (rc)
3236 return rc;
3237
3238 /*
3239 * leaf page has been completed:
3240 * start with 1st entry of next leaf page
3241 */
3242 if (index >= p->header.nextindex) {
3243 bn = le64_to_cpu(p->header.next);
3244
3245 /* unpin leaf page */
3246 DT_PUTPAGE(mp);
3247
3248 /* offset beyond eof ? */
3249 if (bn == 0) {
3250 bn = -1;
3251 goto out;
3252 }
3253
3254 /* get next leaf page */
3255 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3256 if (rc)
3257 return rc;
3258
3259 /* start with 1st entry of next leaf page */
3260 dtoffset->pn++;
3261 dtoffset->index = 0;
3262 }
3263
3264 out:
3265 /* return target leaf page pinned */
3266 btsp = btstack->top;
3267 btsp->bn = bn;
3268 btsp->index = dtoffset->index;
3269 btsp->mp = mp;
3270
3271 return 0;
3272}
3273
3274
3275/*
3276 * dtCompare()
3277 *
3278 * function: compare search key with an internal entry
3279 *
3280 * return:
3281 * < 0 if k is < record
3282 * = 0 if k is = record
3283 * > 0 if k is > record
3284 */
3285static int dtCompare(struct component_name * key, /* search key */
3286 dtpage_t * p, /* directory page */
3287 int si)
3288{ /* entry slot index */
3289 wchar_t *kname;
3290 __le16 *name;
3291 int klen, namlen, len, rc;
3292 struct idtentry *ih;
3293 struct dtslot *t;
3294
3295 /*
3296 * force the left-most key on internal pages, at any level of
3297 * the tree, to be less than any search key.
3298 * this obviates having to update the leftmost key on an internal
3299 * page when the user inserts a new key in the tree smaller than
3300 * anything that has been stored.
3301 *
3302 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3303 * at any internal page at any level of the tree,
3304 * it descends to child of the entry anyway -
3305 * ? make the entry as min size dummy entry)
3306 *
3307 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3308 * return (1);
3309 */
3310
3311 kname = key->name;
3312 klen = key->namlen;
3313
3314 ih = (struct idtentry *) & p->slot[si];
3315 si = ih->next;
3316 name = ih->name;
3317 namlen = ih->namlen;
3318 len = min(namlen, DTIHDRDATALEN);
3319
3320 /* compare with head/only segment */
3321 len = min(klen, len);
3322 if ((rc = UniStrncmp_le(kname, name, len)))
3323 return rc;
3324
3325 klen -= len;
3326 namlen -= len;
3327
3328 /* compare with additional segment(s) */
3329 kname += len;
3330 while (klen > 0 && namlen > 0) {
3331 /* compare with next name segment */
3332 t = (struct dtslot *) & p->slot[si];
3333 len = min(namlen, DTSLOTDATALEN);
3334 len = min(klen, len);
3335 name = t->name;
3336 if ((rc = UniStrncmp_le(kname, name, len)))
3337 return rc;
3338
3339 klen -= len;
3340 namlen -= len;
3341 kname += len;
3342 si = t->next;
3343 }
3344
3345 return (klen - namlen);
3346}
3347
3348
3349
3350
3351/*
3352 * ciCompare()
3353 *
3354 * function: compare search key with an (leaf/internal) entry
3355 *
3356 * return:
3357 * < 0 if k is < record
3358 * = 0 if k is = record
3359 * > 0 if k is > record
3360 */
3361static int ciCompare(struct component_name * key, /* search key */
3362 dtpage_t * p, /* directory page */
3363 int si, /* entry slot index */
3364 int flag)
3365{
3366 wchar_t *kname, x;
3367 __le16 *name;
3368 int klen, namlen, len, rc;
3369 struct ldtentry *lh;
3370 struct idtentry *ih;
3371 struct dtslot *t;
3372 int i;
3373
3374 /*
3375 * force the left-most key on internal pages, at any level of
3376 * the tree, to be less than any search key.
3377 * this obviates having to update the leftmost key on an internal
3378 * page when the user inserts a new key in the tree smaller than
3379 * anything that has been stored.
3380 *
3381 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3382 * at any internal page at any level of the tree,
3383 * it descends to child of the entry anyway -
3384 * ? make the entry as min size dummy entry)
3385 *
3386 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3387 * return (1);
3388 */
3389
3390 kname = key->name;
3391 klen = key->namlen;
3392
3393 /*
3394 * leaf page entry
3395 */
3396 if (p->header.flag & BT_LEAF) {
3397 lh = (struct ldtentry *) & p->slot[si];
3398 si = lh->next;
3399 name = lh->name;
3400 namlen = lh->namlen;
3401 if (flag & JFS_DIR_INDEX)
3402 len = min(namlen, DTLHDRDATALEN);
3403 else
3404 len = min(namlen, DTLHDRDATALEN_LEGACY);
3405 }
3406 /*
3407 * internal page entry
3408 */
3409 else {
3410 ih = (struct idtentry *) & p->slot[si];
3411 si = ih->next;
3412 name = ih->name;
3413 namlen = ih->namlen;
3414 len = min(namlen, DTIHDRDATALEN);
3415 }
3416
3417 /* compare with head/only segment */
3418 len = min(klen, len);
3419 for (i = 0; i < len; i++, kname++, name++) {
3420 /* only uppercase if case-insensitive support is on */
3421 if ((flag & JFS_OS2) == JFS_OS2)
3422 x = UniToupper(le16_to_cpu(*name));
3423 else
3424 x = le16_to_cpu(*name);
3425 if ((rc = *kname - x))
3426 return rc;
3427 }
3428
3429 klen -= len;
3430 namlen -= len;
3431
3432 /* compare with additional segment(s) */
3433 while (klen > 0 && namlen > 0) {
3434 /* compare with next name segment */
3435 t = (struct dtslot *) & p->slot[si];
3436 len = min(namlen, DTSLOTDATALEN);
3437 len = min(klen, len);
3438 name = t->name;
3439 for (i = 0; i < len; i++, kname++, name++) {
3440 /* only uppercase if case-insensitive support is on */
3441 if ((flag & JFS_OS2) == JFS_OS2)
3442 x = UniToupper(le16_to_cpu(*name));
3443 else
3444 x = le16_to_cpu(*name);
3445
3446 if ((rc = *kname - x))
3447 return rc;
3448 }
3449
3450 klen -= len;
3451 namlen -= len;
3452 si = t->next;
3453 }
3454
3455 return (klen - namlen);
3456}
3457
3458
3459/*
3460 * ciGetLeafPrefixKey()
3461 *
3462 * function: compute prefix of suffix compression
3463 * from two adjacent leaf entries
3464 * across page boundary
3465 *
3466 * return: non-zero on error
3467 *
3468 */
3469static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
3470 int ri, struct component_name * key, int flag)
3471{
3472 int klen, namlen;
3473 wchar_t *pl, *pr, *kname;
3474 struct component_name lkey;
3475 struct component_name rkey;
3476
3477 lkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3478 GFP_KERNEL);
3479 if (lkey.name == NULL)
3480 return -ENOMEM;
3481
3482 rkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3483 GFP_KERNEL);
3484 if (rkey.name == NULL) {
3485 kfree(lkey.name);
3486 return -ENOMEM;
3487 }
3488
3489 /* get left and right key */
3490 dtGetKey(lp, li, &lkey, flag);
3491 lkey.name[lkey.namlen] = 0;
3492
3493 if ((flag & JFS_OS2) == JFS_OS2)
3494 ciToUpper(&lkey);
3495
3496 dtGetKey(rp, ri, &rkey, flag);
3497 rkey.name[rkey.namlen] = 0;
3498
3499
3500 if ((flag & JFS_OS2) == JFS_OS2)
3501 ciToUpper(&rkey);
3502
3503 /* compute prefix */
3504 klen = 0;
3505 kname = key->name;
3506 namlen = min(lkey.namlen, rkey.namlen);
3507 for (pl = lkey.name, pr = rkey.name;
3508 namlen; pl++, pr++, namlen--, klen++, kname++) {
3509 *kname = *pr;
3510 if (*pl != *pr) {
3511 key->namlen = klen + 1;
3512 goto free_names;
3513 }
3514 }
3515
3516 /* l->namlen <= r->namlen since l <= r */
3517 if (lkey.namlen < rkey.namlen) {
3518 *kname = *pr;
3519 key->namlen = klen + 1;
3520 } else /* l->namelen == r->namelen */
3521 key->namlen = klen;
3522
3523free_names:
3524 kfree(lkey.name);
3525 kfree(rkey.name);
3526 return 0;
3527}
3528
3529
3530
3531/*
3532 * dtGetKey()
3533 *
3534 * function: get key of the entry
3535 */
3536static void dtGetKey(dtpage_t * p, int i, /* entry index */
3537 struct component_name * key, int flag)
3538{
3539 int si;
3540 s8 *stbl;
3541 struct ldtentry *lh;
3542 struct idtentry *ih;
3543 struct dtslot *t;
3544 int namlen, len;
3545 wchar_t *kname;
3546 __le16 *name;
3547
3548 /* get entry */
3549 stbl = DT_GETSTBL(p);
3550 si = stbl[i];
3551 if (p->header.flag & BT_LEAF) {
3552 lh = (struct ldtentry *) & p->slot[si];
3553 si = lh->next;
3554 namlen = lh->namlen;
3555 name = lh->name;
3556 if (flag & JFS_DIR_INDEX)
3557 len = min(namlen, DTLHDRDATALEN);
3558 else
3559 len = min(namlen, DTLHDRDATALEN_LEGACY);
3560 } else {
3561 ih = (struct idtentry *) & p->slot[si];
3562 si = ih->next;
3563 namlen = ih->namlen;
3564 name = ih->name;
3565 len = min(namlen, DTIHDRDATALEN);
3566 }
3567
3568 key->namlen = namlen;
3569 kname = key->name;
3570
3571 /*
3572 * move head/only segment
3573 */
3574 UniStrncpy_from_le(kname, name, len);
3575
3576 /*
3577 * move additional segment(s)
3578 */
3579 while (si >= 0) {
3580 /* get next segment */
3581 t = &p->slot[si];
3582 kname += len;
3583 namlen -= len;
3584 len = min(namlen, DTSLOTDATALEN);
3585 UniStrncpy_from_le(kname, t->name, len);
3586
3587 si = t->next;
3588 }
3589}
3590
3591
3592/*
3593 * dtInsertEntry()
3594 *
3595 * function: allocate free slot(s) and
3596 * write a leaf/internal entry
3597 *
3598 * return: entry slot index
3599 */
3600static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
3601 ddata_t * data, struct dt_lock ** dtlock)
3602{
3603 struct dtslot *h, *t;
3604 struct ldtentry *lh = NULL;
3605 struct idtentry *ih = NULL;
3606 int hsi, fsi, klen, len, nextindex;
3607 wchar_t *kname;
3608 __le16 *name;
3609 s8 *stbl;
3610 pxd_t *xd;
3611 struct dt_lock *dtlck = *dtlock;
3612 struct lv *lv;
3613 int xsi, n;
3614 s64 bn = 0;
3615 struct metapage *mp = NULL;
3616
3617 klen = key->namlen;
3618 kname = key->name;
3619
3620 /* allocate a free slot */
3621 hsi = fsi = p->header.freelist;
3622 h = &p->slot[fsi];
3623 p->header.freelist = h->next;
3624 --p->header.freecnt;
3625
3626 /* open new linelock */
3627 if (dtlck->index >= dtlck->maxcnt)
3628 dtlck = (struct dt_lock *) txLinelock(dtlck);
3629
3630 lv = & dtlck->lv[dtlck->index];
3631 lv->offset = hsi;
3632
3633 /* write head/only segment */
3634 if (p->header.flag & BT_LEAF) {
3635 lh = (struct ldtentry *) h;
3636 lh->next = h->next;
3637 lh->inumber = cpu_to_le32(data->leaf.ino);
3638 lh->namlen = klen;
3639 name = lh->name;
3640 if (data->leaf.ip) {
3641 len = min(klen, DTLHDRDATALEN);
3642 if (!(p->header.flag & BT_ROOT))
3643 bn = addressPXD(&p->header.self);
3644 lh->index = cpu_to_le32(add_index(data->leaf.tid,
3645 data->leaf.ip,
3646 bn, index));
3647 } else
3648 len = min(klen, DTLHDRDATALEN_LEGACY);
3649 } else {
3650 ih = (struct idtentry *) h;
3651 ih->next = h->next;
3652 xd = (pxd_t *) ih;
3653 *xd = data->xd;
3654 ih->namlen = klen;
3655 name = ih->name;
3656 len = min(klen, DTIHDRDATALEN);
3657 }
3658
3659 UniStrncpy_to_le(name, kname, len);
3660
3661 n = 1;
3662 xsi = hsi;
3663
3664 /* write additional segment(s) */
3665 t = h;
3666 klen -= len;
3667 while (klen) {
3668 /* get free slot */
3669 fsi = p->header.freelist;
3670 t = &p->slot[fsi];
3671 p->header.freelist = t->next;
3672 --p->header.freecnt;
3673
3674 /* is next slot contiguous ? */
3675 if (fsi != xsi + 1) {
3676 /* close current linelock */
3677 lv->length = n;
3678 dtlck->index++;
3679
3680 /* open new linelock */
3681 if (dtlck->index < dtlck->maxcnt)
3682 lv++;
3683 else {
3684 dtlck = (struct dt_lock *) txLinelock(dtlck);
3685 lv = & dtlck->lv[0];
3686 }
3687
3688 lv->offset = fsi;
3689 n = 0;
3690 }
3691
3692 kname += len;
3693 len = min(klen, DTSLOTDATALEN);
3694 UniStrncpy_to_le(t->name, kname, len);
3695
3696 n++;
3697 xsi = fsi;
3698 klen -= len;
3699 }
3700
3701 /* close current linelock */
3702 lv->length = n;
3703 dtlck->index++;
3704
3705 *dtlock = dtlck;
3706
3707 /* terminate last/only segment */
3708 if (h == t) {
3709 /* single segment entry */
3710 if (p->header.flag & BT_LEAF)
3711 lh->next = -1;
3712 else
3713 ih->next = -1;
3714 } else
3715 /* multi-segment entry */
3716 t->next = -1;
3717
3718 /* if insert into middle, shift right succeeding entries in stbl */
3719 stbl = DT_GETSTBL(p);
3720 nextindex = p->header.nextindex;
3721 if (index < nextindex) {
3722 memmove(stbl + index + 1, stbl + index, nextindex - index);
3723
3724 if ((p->header.flag & BT_LEAF) && data->leaf.ip) {
3725 s64 lblock;
3726
3727 /*
3728 * Need to update slot number for entries that moved
3729 * in the stbl
3730 */
3731 mp = NULL;
3732 for (n = index + 1; n <= nextindex; n++) {
3733 lh = (struct ldtentry *) & (p->slot[stbl[n]]);
3734 modify_index(data->leaf.tid, data->leaf.ip,
3735 le32_to_cpu(lh->index), bn, n,
3736 &mp, &lblock);
3737 }
3738 if (mp)
3739 release_metapage(mp);
3740 }
3741 }
3742
3743 stbl[index] = hsi;
3744
3745 /* advance next available entry index of stbl */
3746 ++p->header.nextindex;
3747}
3748
3749
3750/*
3751 * dtMoveEntry()
3752 *
3753 * function: move entries from split/left page to new/right page
3754 *
3755 * nextindex of dst page and freelist/freecnt of both pages
3756 * are updated.
3757 */
3758static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
3759 struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
3760 int do_index)
3761{
3762 int ssi, next; /* src slot index */
3763 int di; /* dst entry index */
3764 int dsi; /* dst slot index */
3765 s8 *sstbl, *dstbl; /* sorted entry table */
3766 int snamlen, len;
3767 struct ldtentry *slh, *dlh = NULL;
3768 struct idtentry *sih, *dih = NULL;
3769 struct dtslot *h, *s, *d;
3770 struct dt_lock *sdtlck = *sdtlock, *ddtlck = *ddtlock;
3771 struct lv *slv, *dlv;
3772 int xssi, ns, nd;
3773 int sfsi;
3774
3775 sstbl = (s8 *) & sp->slot[sp->header.stblindex];
3776 dstbl = (s8 *) & dp->slot[dp->header.stblindex];
3777
3778 dsi = dp->header.freelist; /* first (whole page) free slot */
3779 sfsi = sp->header.freelist;
3780
3781 /* linelock destination entry slot */
3782 dlv = & ddtlck->lv[ddtlck->index];
3783 dlv->offset = dsi;
3784
3785 /* linelock source entry slot */
3786 slv = & sdtlck->lv[sdtlck->index];
3787 slv->offset = sstbl[si];
3788 xssi = slv->offset - 1;
3789
3790 /*
3791 * move entries
3792 */
3793 ns = nd = 0;
3794 for (di = 0; si < sp->header.nextindex; si++, di++) {
3795 ssi = sstbl[si];
3796 dstbl[di] = dsi;
3797
3798 /* is next slot contiguous ? */
3799 if (ssi != xssi + 1) {
3800 /* close current linelock */
3801 slv->length = ns;
3802 sdtlck->index++;
3803
3804 /* open new linelock */
3805 if (sdtlck->index < sdtlck->maxcnt)
3806 slv++;
3807 else {
3808 sdtlck = (struct dt_lock *) txLinelock(sdtlck);
3809 slv = & sdtlck->lv[0];
3810 }
3811
3812 slv->offset = ssi;
3813 ns = 0;
3814 }
3815
3816 /*
3817 * move head/only segment of an entry
3818 */
3819 /* get dst slot */
3820 h = d = &dp->slot[dsi];
3821
3822 /* get src slot and move */
3823 s = &sp->slot[ssi];
3824 if (sp->header.flag & BT_LEAF) {
3825 /* get source entry */
3826 slh = (struct ldtentry *) s;
3827 dlh = (struct ldtentry *) h;
3828 snamlen = slh->namlen;
3829
3830 if (do_index) {
3831 len = min(snamlen, DTLHDRDATALEN);
3832 dlh->index = slh->index; /* little-endian */
3833 } else
3834 len = min(snamlen, DTLHDRDATALEN_LEGACY);
3835
3836 memcpy(dlh, slh, 6 + len * 2);
3837
3838 next = slh->next;
3839
3840 /* update dst head/only segment next field */
3841 dsi++;
3842 dlh->next = dsi;
3843 } else {
3844 sih = (struct idtentry *) s;
3845 snamlen = sih->namlen;
3846
3847 len = min(snamlen, DTIHDRDATALEN);
3848 dih = (struct idtentry *) h;
3849 memcpy(dih, sih, 10 + len * 2);
3850 next = sih->next;
3851
3852 dsi++;
3853 dih->next = dsi;
3854 }
3855
3856 /* free src head/only segment */
3857 s->next = sfsi;
3858 s->cnt = 1;
3859 sfsi = ssi;
3860
3861 ns++;
3862 nd++;
3863 xssi = ssi;
3864
3865 /*
3866 * move additional segment(s) of the entry
3867 */
3868 snamlen -= len;
3869 while ((ssi = next) >= 0) {
3870 /* is next slot contiguous ? */
3871 if (ssi != xssi + 1) {
3872 /* close current linelock */
3873 slv->length = ns;
3874 sdtlck->index++;
3875
3876 /* open new linelock */
3877 if (sdtlck->index < sdtlck->maxcnt)
3878 slv++;
3879 else {
3880 sdtlck =
3881 (struct dt_lock *)
3882 txLinelock(sdtlck);
3883 slv = & sdtlck->lv[0];
3884 }
3885
3886 slv->offset = ssi;
3887 ns = 0;
3888 }
3889
3890 /* get next source segment */
3891 s = &sp->slot[ssi];
3892
3893 /* get next destination free slot */
3894 d++;
3895
3896 len = min(snamlen, DTSLOTDATALEN);
3897 UniStrncpy_le(d->name, s->name, len);
3898
3899 ns++;
3900 nd++;
3901 xssi = ssi;
3902
3903 dsi++;
3904 d->next = dsi;
3905
3906 /* free source segment */
3907 next = s->next;
3908 s->next = sfsi;
3909 s->cnt = 1;
3910 sfsi = ssi;
3911
3912 snamlen -= len;
3913 } /* end while */
3914
3915 /* terminate dst last/only segment */
3916 if (h == d) {
3917 /* single segment entry */
3918 if (dp->header.flag & BT_LEAF)
3919 dlh->next = -1;
3920 else
3921 dih->next = -1;
3922 } else
3923 /* multi-segment entry */
3924 d->next = -1;
3925 } /* end for */
3926
3927 /* close current linelock */
3928 slv->length = ns;
3929 sdtlck->index++;
3930 *sdtlock = sdtlck;
3931
3932 dlv->length = nd;
3933 ddtlck->index++;
3934 *ddtlock = ddtlck;
3935
3936 /* update source header */
3937 sp->header.freelist = sfsi;
3938 sp->header.freecnt += nd;
3939
3940 /* update destination header */
3941 dp->header.nextindex = di;
3942
3943 dp->header.freelist = dsi;
3944 dp->header.freecnt -= nd;
3945}
3946
3947
3948/*
3949 * dtDeleteEntry()
3950 *
3951 * function: free a (leaf/internal) entry
3952 *
3953 * log freelist header, stbl, and each segment slot of entry
3954 * (even though last/only segment next field is modified,
3955 * physical image logging requires all segment slots of
3956 * the entry logged to avoid applying previous updates
3957 * to the same slots)
3958 */
3959static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock)
3960{
3961 int fsi; /* free entry slot index */
3962 s8 *stbl;
3963 struct dtslot *t;
3964 int si, freecnt;
3965 struct dt_lock *dtlck = *dtlock;
3966 struct lv *lv;
3967 int xsi, n;
3968
3969 /* get free entry slot index */
3970 stbl = DT_GETSTBL(p);
3971 fsi = stbl[fi];
3972
3973 /* open new linelock */
3974 if (dtlck->index >= dtlck->maxcnt)
3975 dtlck = (struct dt_lock *) txLinelock(dtlck);
3976 lv = & dtlck->lv[dtlck->index];
3977
3978 lv->offset = fsi;
3979
3980 /* get the head/only segment */
3981 t = &p->slot[fsi];
3982 if (p->header.flag & BT_LEAF)
3983 si = ((struct ldtentry *) t)->next;
3984 else
3985 si = ((struct idtentry *) t)->next;
3986 t->next = si;
3987 t->cnt = 1;
3988
3989 n = freecnt = 1;
3990 xsi = fsi;
3991
3992 /* find the last/only segment */
3993 while (si >= 0) {
3994 /* is next slot contiguous ? */
3995 if (si != xsi + 1) {
3996 /* close current linelock */
3997 lv->length = n;
3998 dtlck->index++;
3999
4000 /* open new linelock */
4001 if (dtlck->index < dtlck->maxcnt)
4002 lv++;
4003 else {
4004 dtlck = (struct dt_lock *) txLinelock(dtlck);
4005 lv = & dtlck->lv[0];
4006 }
4007
4008 lv->offset = si;
4009 n = 0;
4010 }
4011
4012 n++;
4013 xsi = si;
4014 freecnt++;
4015
4016 t = &p->slot[si];
4017 t->cnt = 1;
4018 si = t->next;
4019 }
4020
4021 /* close current linelock */
4022 lv->length = n;
4023 dtlck->index++;
4024
4025 *dtlock = dtlck;
4026
4027 /* update freelist */
4028 t->next = p->header.freelist;
4029 p->header.freelist = fsi;
4030 p->header.freecnt += freecnt;
4031
4032 /* if delete from middle,
4033 * shift left the succedding entries in the stbl
4034 */
4035 si = p->header.nextindex;
4036 if (fi < si - 1)
4037 memmove(&stbl[fi], &stbl[fi + 1], si - fi - 1);
4038
4039 p->header.nextindex--;
4040}
4041
4042
4043/*
4044 * dtTruncateEntry()
4045 *
4046 * function: truncate a (leaf/internal) entry
4047 *
4048 * log freelist header, stbl, and each segment slot of entry
4049 * (even though last/only segment next field is modified,
4050 * physical image logging requires all segment slots of
4051 * the entry logged to avoid applying previous updates
4052 * to the same slots)
4053 */
4054static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock)
4055{
4056 int tsi; /* truncate entry slot index */
4057 s8 *stbl;
4058 struct dtslot *t;
4059 int si, freecnt;
4060 struct dt_lock *dtlck = *dtlock;
4061 struct lv *lv;
4062 int fsi, xsi, n;
4063
4064 /* get free entry slot index */
4065 stbl = DT_GETSTBL(p);
4066 tsi = stbl[ti];
4067
4068 /* open new linelock */
4069 if (dtlck->index >= dtlck->maxcnt)
4070 dtlck = (struct dt_lock *) txLinelock(dtlck);
4071 lv = & dtlck->lv[dtlck->index];
4072
4073 lv->offset = tsi;
4074
4075 /* get the head/only segment */
4076 t = &p->slot[tsi];
4077 ASSERT(p->header.flag & BT_INTERNAL);
4078 ((struct idtentry *) t)->namlen = 0;
4079 si = ((struct idtentry *) t)->next;
4080 ((struct idtentry *) t)->next = -1;
4081
4082 n = 1;
4083 freecnt = 0;
4084 fsi = si;
4085 xsi = tsi;
4086
4087 /* find the last/only segment */
4088 while (si >= 0) {
4089 /* is next slot contiguous ? */
4090 if (si != xsi + 1) {
4091 /* close current linelock */
4092 lv->length = n;
4093 dtlck->index++;
4094
4095 /* open new linelock */
4096 if (dtlck->index < dtlck->maxcnt)
4097 lv++;
4098 else {
4099 dtlck = (struct dt_lock *) txLinelock(dtlck);
4100 lv = & dtlck->lv[0];
4101 }
4102
4103 lv->offset = si;
4104 n = 0;
4105 }
4106
4107 n++;
4108 xsi = si;
4109 freecnt++;
4110
4111 t = &p->slot[si];
4112 t->cnt = 1;
4113 si = t->next;
4114 }
4115
4116 /* close current linelock */
4117 lv->length = n;
4118 dtlck->index++;
4119
4120 *dtlock = dtlck;
4121
4122 /* update freelist */
4123 if (freecnt == 0)
4124 return;
4125 t->next = p->header.freelist;
4126 p->header.freelist = fsi;
4127 p->header.freecnt += freecnt;
4128}
4129
4130
4131/*
4132 * dtLinelockFreelist()
4133 */
4134static void dtLinelockFreelist(dtpage_t * p, /* directory page */
4135 int m, /* max slot index */
4136 struct dt_lock ** dtlock)
4137{
4138 int fsi; /* free entry slot index */
4139 struct dtslot *t;
4140 int si;
4141 struct dt_lock *dtlck = *dtlock;
4142 struct lv *lv;
4143 int xsi, n;
4144
4145 /* get free entry slot index */
4146 fsi = p->header.freelist;
4147
4148 /* open new linelock */
4149 if (dtlck->index >= dtlck->maxcnt)
4150 dtlck = (struct dt_lock *) txLinelock(dtlck);
4151 lv = & dtlck->lv[dtlck->index];
4152
4153 lv->offset = fsi;
4154
4155 n = 1;
4156 xsi = fsi;
4157
4158 t = &p->slot[fsi];
4159 si = t->next;
4160
4161 /* find the last/only segment */
4162 while (si < m && si >= 0) {
4163 /* is next slot contiguous ? */
4164 if (si != xsi + 1) {
4165 /* close current linelock */
4166 lv->length = n;
4167 dtlck->index++;
4168
4169 /* open new linelock */
4170 if (dtlck->index < dtlck->maxcnt)
4171 lv++;
4172 else {
4173 dtlck = (struct dt_lock *) txLinelock(dtlck);
4174 lv = & dtlck->lv[0];
4175 }
4176
4177 lv->offset = si;
4178 n = 0;
4179 }
4180
4181 n++;
4182 xsi = si;
4183
4184 t = &p->slot[si];
4185 si = t->next;
4186 }
4187
4188 /* close current linelock */
4189 lv->length = n;
4190 dtlck->index++;
4191
4192 *dtlock = dtlck;
4193}
4194
4195
4196/*
4197 * NAME: dtModify
4198 *
4199 * FUNCTION: Modify the inode number part of a directory entry
4200 *
4201 * PARAMETERS:
4202 * tid - Transaction id
4203 * ip - Inode of parent directory
4204 * key - Name of entry to be modified
4205 * orig_ino - Original inode number expected in entry
4206 * new_ino - New inode number to put into entry
4207 * flag - JFS_RENAME
4208 *
4209 * RETURNS:
4210 * -ESTALE - If entry found does not match orig_ino passed in
4211 * -ENOENT - If no entry can be found to match key
4212 * 0 - If successfully modified entry
4213 */
4214int dtModify(tid_t tid, struct inode *ip,
4215 struct component_name * key, ino_t * orig_ino, ino_t new_ino, int flag)
4216{
4217 int rc;
4218 s64 bn;
4219 struct metapage *mp;
4220 dtpage_t *p;
4221 int index;
4222 struct btstack btstack;
4223 struct tlock *tlck;
4224 struct dt_lock *dtlck;
4225 struct lv *lv;
4226 s8 *stbl;
4227 int entry_si; /* entry slot index */
4228 struct ldtentry *entry;
4229
4230 /*
4231 * search for the entry to modify:
4232 *
4233 * dtSearch() returns (leaf page pinned, index at which to modify).
4234 */
4235 if ((rc = dtSearch(ip, key, orig_ino, &btstack, flag)))
4236 return rc;
4237
4238 /* retrieve search result */
4239 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
4240
4241 BT_MARK_DIRTY(mp, ip);
4242 /*
4243 * acquire a transaction lock on the leaf page of named entry
4244 */
4245 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
4246 dtlck = (struct dt_lock *) & tlck->lock;
4247
4248 /* get slot index of the entry */
4249 stbl = DT_GETSTBL(p);
4250 entry_si = stbl[index];
4251
4252 /* linelock entry */
4253 ASSERT(dtlck->index == 0);
4254 lv = & dtlck->lv[0];
4255 lv->offset = entry_si;
4256 lv->length = 1;
4257 dtlck->index++;
4258
4259 /* get the head/only segment */
4260 entry = (struct ldtentry *) & p->slot[entry_si];
4261
4262 /* substitute the inode number of the entry */
4263 entry->inumber = cpu_to_le32(new_ino);
4264
4265 /* unpin the leaf page */
4266 DT_PUTPAGE(mp);
4267
4268 return 0;
4269}
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (C) International Business Machines Corp., 2000-2004
4 */
5
6/*
7 * jfs_dtree.c: directory B+-tree manager
8 *
9 * B+-tree with variable length key directory:
10 *
11 * each directory page is structured as an array of 32-byte
12 * directory entry slots initialized as a freelist
13 * to avoid search/compaction of free space at insertion.
14 * when an entry is inserted, a number of slots are allocated
15 * from the freelist as required to store variable length data
16 * of the entry; when the entry is deleted, slots of the entry
17 * are returned to freelist.
18 *
19 * leaf entry stores full name as key and file serial number
20 * (aka inode number) as data.
21 * internal/router entry stores sufffix compressed name
22 * as key and simple extent descriptor as data.
23 *
24 * each directory page maintains a sorted entry index table
25 * which stores the start slot index of sorted entries
26 * to allow binary search on the table.
27 *
28 * directory starts as a root/leaf page in on-disk inode
29 * inline data area.
30 * when it becomes full, it starts a leaf of a external extent
31 * of length of 1 block. each time the first leaf becomes full,
32 * it is extended rather than split (its size is doubled),
33 * until its length becoms 4 KBytes, from then the extent is split
34 * with new 4 Kbyte extent when it becomes full
35 * to reduce external fragmentation of small directories.
36 *
37 * blah, blah, blah, for linear scan of directory in pieces by
38 * readdir().
39 *
40 *
41 * case-insensitive directory file system
42 *
43 * names are stored in case-sensitive way in leaf entry.
44 * but stored, searched and compared in case-insensitive (uppercase) order
45 * (i.e., both search key and entry key are folded for search/compare):
46 * (note that case-sensitive order is BROKEN in storage, e.g.,
47 * sensitive: Ad, aB, aC, aD -> insensitive: aB, aC, aD, Ad
48 *
49 * entries which folds to the same key makes up a equivalent class
50 * whose members are stored as contiguous cluster (may cross page boundary)
51 * but whose order is arbitrary and acts as duplicate, e.g.,
52 * abc, Abc, aBc, abC)
53 *
54 * once match is found at leaf, requires scan forward/backward
55 * either for, in case-insensitive search, duplicate
56 * or for, in case-sensitive search, for exact match
57 *
58 * router entry must be created/stored in case-insensitive way
59 * in internal entry:
60 * (right most key of left page and left most key of right page
61 * are folded, and its suffix compression is propagated as router
62 * key in parent)
63 * (e.g., if split occurs <abc> and <aBd>, <ABD> trather than <aB>
64 * should be made the router key for the split)
65 *
66 * case-insensitive search:
67 *
68 * fold search key;
69 *
70 * case-insensitive search of B-tree:
71 * for internal entry, router key is already folded;
72 * for leaf entry, fold the entry key before comparison.
73 *
74 * if (leaf entry case-insensitive match found)
75 * if (next entry satisfies case-insensitive match)
76 * return EDUPLICATE;
77 * if (prev entry satisfies case-insensitive match)
78 * return EDUPLICATE;
79 * return match;
80 * else
81 * return no match;
82 *
83 * serialization:
84 * target directory inode lock is being held on entry/exit
85 * of all main directory service routines.
86 *
87 * log based recovery:
88 */
89
90#include <linux/fs.h>
91#include <linux/quotaops.h>
92#include <linux/slab.h>
93#include "jfs_incore.h"
94#include "jfs_superblock.h"
95#include "jfs_filsys.h"
96#include "jfs_metapage.h"
97#include "jfs_dmap.h"
98#include "jfs_unicode.h"
99#include "jfs_debug.h"
100
101/* dtree split parameter */
102struct dtsplit {
103 struct metapage *mp;
104 s16 index;
105 s16 nslot;
106 struct component_name *key;
107 ddata_t *data;
108 struct pxdlist *pxdlist;
109};
110
111#define DT_PAGE(IP, MP) BT_PAGE(IP, MP, dtpage_t, i_dtroot)
112
113/* get page buffer for specified block address */
114#define DT_GETPAGE(IP, BN, MP, SIZE, P, RC) \
115do { \
116 BT_GETPAGE(IP, BN, MP, dtpage_t, SIZE, P, RC, i_dtroot); \
117 if (!(RC)) { \
118 if (((P)->header.nextindex > \
119 (((BN) == 0) ? DTROOTMAXSLOT : (P)->header.maxslot)) || \
120 ((BN) && ((P)->header.maxslot > DTPAGEMAXSLOT))) { \
121 BT_PUTPAGE(MP); \
122 jfs_error((IP)->i_sb, \
123 "DT_GETPAGE: dtree page corrupt\n"); \
124 MP = NULL; \
125 RC = -EIO; \
126 } \
127 } \
128} while (0)
129
130/* for consistency */
131#define DT_PUTPAGE(MP) BT_PUTPAGE(MP)
132
133#define DT_GETSEARCH(IP, LEAF, BN, MP, P, INDEX) \
134 BT_GETSEARCH(IP, LEAF, BN, MP, dtpage_t, P, INDEX, i_dtroot)
135
136/*
137 * forward references
138 */
139static int dtSplitUp(tid_t tid, struct inode *ip,
140 struct dtsplit * split, struct btstack * btstack);
141
142static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
143 struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rxdp);
144
145static int dtExtendPage(tid_t tid, struct inode *ip,
146 struct dtsplit * split, struct btstack * btstack);
147
148static int dtSplitRoot(tid_t tid, struct inode *ip,
149 struct dtsplit * split, struct metapage ** rmpp);
150
151static int dtDeleteUp(tid_t tid, struct inode *ip, struct metapage * fmp,
152 dtpage_t * fp, struct btstack * btstack);
153
154static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p);
155
156static int dtReadFirst(struct inode *ip, struct btstack * btstack);
157
158static int dtReadNext(struct inode *ip,
159 loff_t * offset, struct btstack * btstack);
160
161static int dtCompare(struct component_name * key, dtpage_t * p, int si);
162
163static int ciCompare(struct component_name * key, dtpage_t * p, int si,
164 int flag);
165
166static void dtGetKey(dtpage_t * p, int i, struct component_name * key,
167 int flag);
168
169static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
170 int ri, struct component_name * key, int flag);
171
172static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
173 ddata_t * data, struct dt_lock **);
174
175static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
176 struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
177 int do_index);
178
179static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock);
180
181static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock);
182
183static void dtLinelockFreelist(dtpage_t * p, int m, struct dt_lock ** dtlock);
184
185#define ciToUpper(c) UniStrupr((c)->name)
186
187/*
188 * read_index_page()
189 *
190 * Reads a page of a directory's index table.
191 * Having metadata mapped into the directory inode's address space
192 * presents a multitude of problems. We avoid this by mapping to
193 * the absolute address space outside of the *_metapage routines
194 */
195static struct metapage *read_index_page(struct inode *inode, s64 blkno)
196{
197 int rc;
198 s64 xaddr;
199 int xflag;
200 s32 xlen;
201
202 rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
203 if (rc || (xaddr == 0))
204 return NULL;
205
206 return read_metapage(inode, xaddr, PSIZE, 1);
207}
208
209/*
210 * get_index_page()
211 *
212 * Same as get_index_page(), but get's a new page without reading
213 */
214static struct metapage *get_index_page(struct inode *inode, s64 blkno)
215{
216 int rc;
217 s64 xaddr;
218 int xflag;
219 s32 xlen;
220
221 rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
222 if (rc || (xaddr == 0))
223 return NULL;
224
225 return get_metapage(inode, xaddr, PSIZE, 1);
226}
227
228/*
229 * find_index()
230 *
231 * Returns dtree page containing directory table entry for specified
232 * index and pointer to its entry.
233 *
234 * mp must be released by caller.
235 */
236static struct dir_table_slot *find_index(struct inode *ip, u32 index,
237 struct metapage ** mp, s64 *lblock)
238{
239 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
240 s64 blkno;
241 s64 offset;
242 int page_offset;
243 struct dir_table_slot *slot;
244 static int maxWarnings = 10;
245
246 if (index < 2) {
247 if (maxWarnings) {
248 jfs_warn("find_entry called with index = %d", index);
249 maxWarnings--;
250 }
251 return NULL;
252 }
253
254 if (index >= jfs_ip->next_index) {
255 jfs_warn("find_entry called with index >= next_index");
256 return NULL;
257 }
258
259 if (jfs_dirtable_inline(ip)) {
260 /*
261 * Inline directory table
262 */
263 *mp = NULL;
264 slot = &jfs_ip->i_dirtable[index - 2];
265 } else {
266 offset = (index - 2) * sizeof(struct dir_table_slot);
267 page_offset = offset & (PSIZE - 1);
268 blkno = ((offset + 1) >> L2PSIZE) <<
269 JFS_SBI(ip->i_sb)->l2nbperpage;
270
271 if (*mp && (*lblock != blkno)) {
272 release_metapage(*mp);
273 *mp = NULL;
274 }
275 if (!(*mp)) {
276 *lblock = blkno;
277 *mp = read_index_page(ip, blkno);
278 }
279 if (!(*mp)) {
280 jfs_err("free_index: error reading directory table");
281 return NULL;
282 }
283
284 slot =
285 (struct dir_table_slot *) ((char *) (*mp)->data +
286 page_offset);
287 }
288 return slot;
289}
290
291static inline void lock_index(tid_t tid, struct inode *ip, struct metapage * mp,
292 u32 index)
293{
294 struct tlock *tlck;
295 struct linelock *llck;
296 struct lv *lv;
297
298 tlck = txLock(tid, ip, mp, tlckDATA);
299 llck = (struct linelock *) tlck->lock;
300
301 if (llck->index >= llck->maxcnt)
302 llck = txLinelock(llck);
303 lv = &llck->lv[llck->index];
304
305 /*
306 * Linelock slot size is twice the size of directory table
307 * slot size. 512 entries per page.
308 */
309 lv->offset = ((index - 2) & 511) >> 1;
310 lv->length = 1;
311 llck->index++;
312}
313
314/*
315 * add_index()
316 *
317 * Adds an entry to the directory index table. This is used to provide
318 * each directory entry with a persistent index in which to resume
319 * directory traversals
320 */
321static u32 add_index(tid_t tid, struct inode *ip, s64 bn, int slot)
322{
323 struct super_block *sb = ip->i_sb;
324 struct jfs_sb_info *sbi = JFS_SBI(sb);
325 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
326 u64 blkno;
327 struct dir_table_slot *dirtab_slot;
328 u32 index;
329 struct linelock *llck;
330 struct lv *lv;
331 struct metapage *mp;
332 s64 offset;
333 uint page_offset;
334 struct tlock *tlck;
335 s64 xaddr;
336
337 ASSERT(DO_INDEX(ip));
338
339 if (jfs_ip->next_index < 2) {
340 jfs_warn("add_index: next_index = %d. Resetting!",
341 jfs_ip->next_index);
342 jfs_ip->next_index = 2;
343 }
344
345 index = jfs_ip->next_index++;
346
347 if (index <= MAX_INLINE_DIRTABLE_ENTRY) {
348 /*
349 * i_size reflects size of index table, or 8 bytes per entry.
350 */
351 ip->i_size = (loff_t) (index - 1) << 3;
352
353 /*
354 * dir table fits inline within inode
355 */
356 dirtab_slot = &jfs_ip->i_dirtable[index-2];
357 dirtab_slot->flag = DIR_INDEX_VALID;
358 dirtab_slot->slot = slot;
359 DTSaddress(dirtab_slot, bn);
360
361 set_cflag(COMMIT_Dirtable, ip);
362
363 return index;
364 }
365 if (index == (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
366 struct dir_table_slot temp_table[12];
367
368 /*
369 * It's time to move the inline table to an external
370 * page and begin to build the xtree
371 */
372 if (dquot_alloc_block(ip, sbi->nbperpage))
373 goto clean_up;
374 if (dbAlloc(ip, 0, sbi->nbperpage, &xaddr)) {
375 dquot_free_block(ip, sbi->nbperpage);
376 goto clean_up;
377 }
378
379 /*
380 * Save the table, we're going to overwrite it with the
381 * xtree root
382 */
383 memcpy(temp_table, &jfs_ip->i_dirtable, sizeof(temp_table));
384
385 /*
386 * Initialize empty x-tree
387 */
388 xtInitRoot(tid, ip);
389
390 /*
391 * Add the first block to the xtree
392 */
393 if (xtInsert(tid, ip, 0, 0, sbi->nbperpage, &xaddr, 0)) {
394 /* This really shouldn't fail */
395 jfs_warn("add_index: xtInsert failed!");
396 memcpy(&jfs_ip->i_dirtable, temp_table,
397 sizeof (temp_table));
398 dbFree(ip, xaddr, sbi->nbperpage);
399 dquot_free_block(ip, sbi->nbperpage);
400 goto clean_up;
401 }
402 ip->i_size = PSIZE;
403
404 mp = get_index_page(ip, 0);
405 if (!mp) {
406 jfs_err("add_index: get_metapage failed!");
407 xtTruncate(tid, ip, 0, COMMIT_PWMAP);
408 memcpy(&jfs_ip->i_dirtable, temp_table,
409 sizeof (temp_table));
410 goto clean_up;
411 }
412 tlck = txLock(tid, ip, mp, tlckDATA);
413 llck = (struct linelock *) & tlck->lock;
414 ASSERT(llck->index == 0);
415 lv = &llck->lv[0];
416
417 lv->offset = 0;
418 lv->length = 6; /* tlckDATA slot size is 16 bytes */
419 llck->index++;
420
421 memcpy(mp->data, temp_table, sizeof(temp_table));
422
423 mark_metapage_dirty(mp);
424 release_metapage(mp);
425
426 /*
427 * Logging is now directed by xtree tlocks
428 */
429 clear_cflag(COMMIT_Dirtable, ip);
430 }
431
432 offset = (index - 2) * sizeof(struct dir_table_slot);
433 page_offset = offset & (PSIZE - 1);
434 blkno = ((offset + 1) >> L2PSIZE) << sbi->l2nbperpage;
435 if (page_offset == 0) {
436 /*
437 * This will be the beginning of a new page
438 */
439 xaddr = 0;
440 if (xtInsert(tid, ip, 0, blkno, sbi->nbperpage, &xaddr, 0)) {
441 jfs_warn("add_index: xtInsert failed!");
442 goto clean_up;
443 }
444 ip->i_size += PSIZE;
445
446 if ((mp = get_index_page(ip, blkno)))
447 memset(mp->data, 0, PSIZE); /* Just looks better */
448 else
449 xtTruncate(tid, ip, offset, COMMIT_PWMAP);
450 } else
451 mp = read_index_page(ip, blkno);
452
453 if (!mp) {
454 jfs_err("add_index: get/read_metapage failed!");
455 goto clean_up;
456 }
457
458 lock_index(tid, ip, mp, index);
459
460 dirtab_slot =
461 (struct dir_table_slot *) ((char *) mp->data + page_offset);
462 dirtab_slot->flag = DIR_INDEX_VALID;
463 dirtab_slot->slot = slot;
464 DTSaddress(dirtab_slot, bn);
465
466 mark_metapage_dirty(mp);
467 release_metapage(mp);
468
469 return index;
470
471 clean_up:
472
473 jfs_ip->next_index--;
474
475 return 0;
476}
477
478/*
479 * free_index()
480 *
481 * Marks an entry to the directory index table as free.
482 */
483static void free_index(tid_t tid, struct inode *ip, u32 index, u32 next)
484{
485 struct dir_table_slot *dirtab_slot;
486 s64 lblock;
487 struct metapage *mp = NULL;
488
489 dirtab_slot = find_index(ip, index, &mp, &lblock);
490
491 if (!dirtab_slot)
492 return;
493
494 dirtab_slot->flag = DIR_INDEX_FREE;
495 dirtab_slot->slot = dirtab_slot->addr1 = 0;
496 dirtab_slot->addr2 = cpu_to_le32(next);
497
498 if (mp) {
499 lock_index(tid, ip, mp, index);
500 mark_metapage_dirty(mp);
501 release_metapage(mp);
502 } else
503 set_cflag(COMMIT_Dirtable, ip);
504}
505
506/*
507 * modify_index()
508 *
509 * Changes an entry in the directory index table
510 */
511static void modify_index(tid_t tid, struct inode *ip, u32 index, s64 bn,
512 int slot, struct metapage ** mp, s64 *lblock)
513{
514 struct dir_table_slot *dirtab_slot;
515
516 dirtab_slot = find_index(ip, index, mp, lblock);
517
518 if (!dirtab_slot)
519 return;
520
521 DTSaddress(dirtab_slot, bn);
522 dirtab_slot->slot = slot;
523
524 if (*mp) {
525 lock_index(tid, ip, *mp, index);
526 mark_metapage_dirty(*mp);
527 } else
528 set_cflag(COMMIT_Dirtable, ip);
529}
530
531/*
532 * read_index()
533 *
534 * reads a directory table slot
535 */
536static int read_index(struct inode *ip, u32 index,
537 struct dir_table_slot * dirtab_slot)
538{
539 s64 lblock;
540 struct metapage *mp = NULL;
541 struct dir_table_slot *slot;
542
543 slot = find_index(ip, index, &mp, &lblock);
544 if (!slot) {
545 return -EIO;
546 }
547
548 memcpy(dirtab_slot, slot, sizeof(struct dir_table_slot));
549
550 if (mp)
551 release_metapage(mp);
552
553 return 0;
554}
555
556/*
557 * dtSearch()
558 *
559 * function:
560 * Search for the entry with specified key
561 *
562 * parameter:
563 *
564 * return: 0 - search result on stack, leaf page pinned;
565 * errno - I/O error
566 */
567int dtSearch(struct inode *ip, struct component_name * key, ino_t * data,
568 struct btstack * btstack, int flag)
569{
570 int rc = 0;
571 int cmp = 1; /* init for empty page */
572 s64 bn;
573 struct metapage *mp;
574 dtpage_t *p;
575 s8 *stbl;
576 int base, index, lim;
577 struct btframe *btsp;
578 pxd_t *pxd;
579 int psize = 288; /* initial in-line directory */
580 ino_t inumber;
581 struct component_name ciKey;
582 struct super_block *sb = ip->i_sb;
583
584 ciKey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
585 GFP_NOFS);
586 if (!ciKey.name) {
587 rc = -ENOMEM;
588 goto dtSearch_Exit2;
589 }
590
591
592 /* uppercase search key for c-i directory */
593 UniStrcpy(ciKey.name, key->name);
594 ciKey.namlen = key->namlen;
595
596 /* only uppercase if case-insensitive support is on */
597 if ((JFS_SBI(sb)->mntflag & JFS_OS2) == JFS_OS2) {
598 ciToUpper(&ciKey);
599 }
600 BT_CLR(btstack); /* reset stack */
601
602 /* init level count for max pages to split */
603 btstack->nsplit = 1;
604
605 /*
606 * search down tree from root:
607 *
608 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
609 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
610 *
611 * if entry with search key K is not found
612 * internal page search find the entry with largest key Ki
613 * less than K which point to the child page to search;
614 * leaf page search find the entry with smallest key Kj
615 * greater than K so that the returned index is the position of
616 * the entry to be shifted right for insertion of new entry.
617 * for empty tree, search key is greater than any key of the tree.
618 *
619 * by convention, root bn = 0.
620 */
621 for (bn = 0;;) {
622 /* get/pin the page to search */
623 DT_GETPAGE(ip, bn, mp, psize, p, rc);
624 if (rc)
625 goto dtSearch_Exit1;
626
627 /* get sorted entry table of the page */
628 stbl = DT_GETSTBL(p);
629
630 /*
631 * binary search with search key K on the current page.
632 */
633 for (base = 0, lim = p->header.nextindex; lim; lim >>= 1) {
634 index = base + (lim >> 1);
635
636 if (stbl[index] < 0) {
637 rc = -EIO;
638 goto out;
639 }
640
641 if (p->header.flag & BT_LEAF) {
642 /* uppercase leaf name to compare */
643 cmp =
644 ciCompare(&ciKey, p, stbl[index],
645 JFS_SBI(sb)->mntflag);
646 } else {
647 /* router key is in uppercase */
648
649 cmp = dtCompare(&ciKey, p, stbl[index]);
650
651
652 }
653 if (cmp == 0) {
654 /*
655 * search hit
656 */
657 /* search hit - leaf page:
658 * return the entry found
659 */
660 if (p->header.flag & BT_LEAF) {
661 inumber = le32_to_cpu(
662 ((struct ldtentry *) & p->slot[stbl[index]])->inumber);
663
664 /*
665 * search for JFS_LOOKUP
666 */
667 if (flag == JFS_LOOKUP) {
668 *data = inumber;
669 rc = 0;
670 goto out;
671 }
672
673 /*
674 * search for JFS_CREATE
675 */
676 if (flag == JFS_CREATE) {
677 *data = inumber;
678 rc = -EEXIST;
679 goto out;
680 }
681
682 /*
683 * search for JFS_REMOVE or JFS_RENAME
684 */
685 if ((flag == JFS_REMOVE ||
686 flag == JFS_RENAME) &&
687 *data != inumber) {
688 rc = -ESTALE;
689 goto out;
690 }
691
692 /*
693 * JFS_REMOVE|JFS_FINDDIR|JFS_RENAME
694 */
695 /* save search result */
696 *data = inumber;
697 btsp = btstack->top;
698 btsp->bn = bn;
699 btsp->index = index;
700 btsp->mp = mp;
701
702 rc = 0;
703 goto dtSearch_Exit1;
704 }
705
706 /* search hit - internal page:
707 * descend/search its child page
708 */
709 goto getChild;
710 }
711
712 if (cmp > 0) {
713 base = index + 1;
714 --lim;
715 }
716 }
717
718 /*
719 * search miss
720 *
721 * base is the smallest index with key (Kj) greater than
722 * search key (K) and may be zero or (maxindex + 1) index.
723 */
724 /*
725 * search miss - leaf page
726 *
727 * return location of entry (base) where new entry with
728 * search key K is to be inserted.
729 */
730 if (p->header.flag & BT_LEAF) {
731 /*
732 * search for JFS_LOOKUP, JFS_REMOVE, or JFS_RENAME
733 */
734 if (flag == JFS_LOOKUP || flag == JFS_REMOVE ||
735 flag == JFS_RENAME) {
736 rc = -ENOENT;
737 goto out;
738 }
739
740 /*
741 * search for JFS_CREATE|JFS_FINDDIR:
742 *
743 * save search result
744 */
745 *data = 0;
746 btsp = btstack->top;
747 btsp->bn = bn;
748 btsp->index = base;
749 btsp->mp = mp;
750
751 rc = 0;
752 goto dtSearch_Exit1;
753 }
754
755 /*
756 * search miss - internal page
757 *
758 * if base is non-zero, decrement base by one to get the parent
759 * entry of the child page to search.
760 */
761 index = base ? base - 1 : base;
762
763 /*
764 * go down to child page
765 */
766 getChild:
767 /* update max. number of pages to split */
768 if (BT_STACK_FULL(btstack)) {
769 /* Something's corrupted, mark filesystem dirty so
770 * chkdsk will fix it.
771 */
772 jfs_error(sb, "stack overrun!\n");
773 BT_STACK_DUMP(btstack);
774 rc = -EIO;
775 goto out;
776 }
777 btstack->nsplit++;
778
779 /* push (bn, index) of the parent page/entry */
780 BT_PUSH(btstack, bn, index);
781
782 /* get the child page block number */
783 pxd = (pxd_t *) & p->slot[stbl[index]];
784 bn = addressPXD(pxd);
785 psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;
786
787 /* unpin the parent page */
788 DT_PUTPAGE(mp);
789 }
790
791 out:
792 DT_PUTPAGE(mp);
793
794 dtSearch_Exit1:
795
796 kfree(ciKey.name);
797
798 dtSearch_Exit2:
799
800 return rc;
801}
802
803
804/*
805 * dtInsert()
806 *
807 * function: insert an entry to directory tree
808 *
809 * parameter:
810 *
811 * return: 0 - success;
812 * errno - failure;
813 */
814int dtInsert(tid_t tid, struct inode *ip,
815 struct component_name * name, ino_t * fsn, struct btstack * btstack)
816{
817 int rc = 0;
818 struct metapage *mp; /* meta-page buffer */
819 dtpage_t *p; /* base B+-tree index page */
820 s64 bn;
821 int index;
822 struct dtsplit split; /* split information */
823 ddata_t data;
824 struct dt_lock *dtlck;
825 int n;
826 struct tlock *tlck;
827 struct lv *lv;
828
829 /*
830 * retrieve search result
831 *
832 * dtSearch() returns (leaf page pinned, index at which to insert).
833 * n.b. dtSearch() may return index of (maxindex + 1) of
834 * the full page.
835 */
836 DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
837 if (p->header.freelist == 0)
838 return -EINVAL;
839
840 /*
841 * insert entry for new key
842 */
843 if (DO_INDEX(ip)) {
844 if (JFS_IP(ip)->next_index == DIREND) {
845 DT_PUTPAGE(mp);
846 return -EMLINK;
847 }
848 n = NDTLEAF(name->namlen);
849 data.leaf.tid = tid;
850 data.leaf.ip = ip;
851 } else {
852 n = NDTLEAF_LEGACY(name->namlen);
853 data.leaf.ip = NULL; /* signifies legacy directory format */
854 }
855 data.leaf.ino = *fsn;
856
857 /*
858 * leaf page does not have enough room for new entry:
859 *
860 * extend/split the leaf page;
861 *
862 * dtSplitUp() will insert the entry and unpin the leaf page.
863 */
864 if (n > p->header.freecnt) {
865 split.mp = mp;
866 split.index = index;
867 split.nslot = n;
868 split.key = name;
869 split.data = &data;
870 rc = dtSplitUp(tid, ip, &split, btstack);
871 return rc;
872 }
873
874 /*
875 * leaf page does have enough room for new entry:
876 *
877 * insert the new data entry into the leaf page;
878 */
879 BT_MARK_DIRTY(mp, ip);
880 /*
881 * acquire a transaction lock on the leaf page
882 */
883 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
884 dtlck = (struct dt_lock *) & tlck->lock;
885 ASSERT(dtlck->index == 0);
886 lv = & dtlck->lv[0];
887
888 /* linelock header */
889 lv->offset = 0;
890 lv->length = 1;
891 dtlck->index++;
892
893 dtInsertEntry(p, index, name, &data, &dtlck);
894
895 /* linelock stbl of non-root leaf page */
896 if (!(p->header.flag & BT_ROOT)) {
897 if (dtlck->index >= dtlck->maxcnt)
898 dtlck = (struct dt_lock *) txLinelock(dtlck);
899 lv = & dtlck->lv[dtlck->index];
900 n = index >> L2DTSLOTSIZE;
901 lv->offset = p->header.stblindex + n;
902 lv->length =
903 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
904 dtlck->index++;
905 }
906
907 /* unpin the leaf page */
908 DT_PUTPAGE(mp);
909
910 return 0;
911}
912
913
914/*
915 * dtSplitUp()
916 *
917 * function: propagate insertion bottom up;
918 *
919 * parameter:
920 *
921 * return: 0 - success;
922 * errno - failure;
923 * leaf page unpinned;
924 */
925static int dtSplitUp(tid_t tid,
926 struct inode *ip, struct dtsplit * split, struct btstack * btstack)
927{
928 struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
929 int rc = 0;
930 struct metapage *smp;
931 dtpage_t *sp; /* split page */
932 struct metapage *rmp;
933 dtpage_t *rp; /* new right page split from sp */
934 pxd_t rpxd; /* new right page extent descriptor */
935 struct metapage *lmp;
936 dtpage_t *lp; /* left child page */
937 int skip; /* index of entry of insertion */
938 struct btframe *parent; /* parent page entry on traverse stack */
939 s64 xaddr, nxaddr;
940 int xlen, xsize;
941 struct pxdlist pxdlist;
942 pxd_t *pxd;
943 struct component_name key = { 0, NULL };
944 ddata_t *data = split->data;
945 int n;
946 struct dt_lock *dtlck;
947 struct tlock *tlck;
948 struct lv *lv;
949 int quota_allocation = 0;
950
951 /* get split page */
952 smp = split->mp;
953 sp = DT_PAGE(ip, smp);
954
955 key.name = kmalloc_array(JFS_NAME_MAX + 2, sizeof(wchar_t), GFP_NOFS);
956 if (!key.name) {
957 DT_PUTPAGE(smp);
958 rc = -ENOMEM;
959 goto dtSplitUp_Exit;
960 }
961
962 /*
963 * split leaf page
964 *
965 * The split routines insert the new entry, and
966 * acquire txLock as appropriate.
967 */
968 /*
969 * split root leaf page:
970 */
971 if (sp->header.flag & BT_ROOT) {
972 /*
973 * allocate a single extent child page
974 */
975 xlen = 1;
976 n = sbi->bsize >> L2DTSLOTSIZE;
977 n -= (n + 31) >> L2DTSLOTSIZE; /* stbl size */
978 n -= DTROOTMAXSLOT - sp->header.freecnt; /* header + entries */
979 if (n <= split->nslot)
980 xlen++;
981 if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr))) {
982 DT_PUTPAGE(smp);
983 goto freeKeyName;
984 }
985
986 pxdlist.maxnpxd = 1;
987 pxdlist.npxd = 0;
988 pxd = &pxdlist.pxd[0];
989 PXDaddress(pxd, xaddr);
990 PXDlength(pxd, xlen);
991 split->pxdlist = &pxdlist;
992 rc = dtSplitRoot(tid, ip, split, &rmp);
993
994 if (rc)
995 dbFree(ip, xaddr, xlen);
996 else
997 DT_PUTPAGE(rmp);
998
999 DT_PUTPAGE(smp);
1000
1001 if (!DO_INDEX(ip))
1002 ip->i_size = xlen << sbi->l2bsize;
1003
1004 goto freeKeyName;
1005 }
1006
1007 /*
1008 * extend first leaf page
1009 *
1010 * extend the 1st extent if less than buffer page size
1011 * (dtExtendPage() reurns leaf page unpinned)
1012 */
1013 pxd = &sp->header.self;
1014 xlen = lengthPXD(pxd);
1015 xsize = xlen << sbi->l2bsize;
1016 if (xsize < PSIZE) {
1017 xaddr = addressPXD(pxd);
1018 n = xsize >> L2DTSLOTSIZE;
1019 n -= (n + 31) >> L2DTSLOTSIZE; /* stbl size */
1020 if ((n + sp->header.freecnt) <= split->nslot)
1021 n = xlen + (xlen << 1);
1022 else
1023 n = xlen;
1024
1025 /* Allocate blocks to quota. */
1026 rc = dquot_alloc_block(ip, n);
1027 if (rc)
1028 goto extendOut;
1029 quota_allocation += n;
1030
1031 if ((rc = dbReAlloc(sbi->ipbmap, xaddr, (s64) xlen,
1032 (s64) n, &nxaddr)))
1033 goto extendOut;
1034
1035 pxdlist.maxnpxd = 1;
1036 pxdlist.npxd = 0;
1037 pxd = &pxdlist.pxd[0];
1038 PXDaddress(pxd, nxaddr);
1039 PXDlength(pxd, xlen + n);
1040 split->pxdlist = &pxdlist;
1041 if ((rc = dtExtendPage(tid, ip, split, btstack))) {
1042 nxaddr = addressPXD(pxd);
1043 if (xaddr != nxaddr) {
1044 /* free relocated extent */
1045 xlen = lengthPXD(pxd);
1046 dbFree(ip, nxaddr, (s64) xlen);
1047 } else {
1048 /* free extended delta */
1049 xlen = lengthPXD(pxd) - n;
1050 xaddr = addressPXD(pxd) + xlen;
1051 dbFree(ip, xaddr, (s64) n);
1052 }
1053 } else if (!DO_INDEX(ip))
1054 ip->i_size = lengthPXD(pxd) << sbi->l2bsize;
1055
1056
1057 extendOut:
1058 DT_PUTPAGE(smp);
1059 goto freeKeyName;
1060 }
1061
1062 /*
1063 * split leaf page <sp> into <sp> and a new right page <rp>.
1064 *
1065 * return <rp> pinned and its extent descriptor <rpxd>
1066 */
1067 /*
1068 * allocate new directory page extent and
1069 * new index page(s) to cover page split(s)
1070 *
1071 * allocation hint: ?
1072 */
1073 n = btstack->nsplit;
1074 pxdlist.maxnpxd = pxdlist.npxd = 0;
1075 xlen = sbi->nbperpage;
1076 for (pxd = pxdlist.pxd; n > 0; n--, pxd++) {
1077 if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr)) == 0) {
1078 PXDaddress(pxd, xaddr);
1079 PXDlength(pxd, xlen);
1080 pxdlist.maxnpxd++;
1081 continue;
1082 }
1083
1084 DT_PUTPAGE(smp);
1085
1086 /* undo allocation */
1087 goto splitOut;
1088 }
1089
1090 split->pxdlist = &pxdlist;
1091 if ((rc = dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd))) {
1092 DT_PUTPAGE(smp);
1093
1094 /* undo allocation */
1095 goto splitOut;
1096 }
1097
1098 if (!DO_INDEX(ip))
1099 ip->i_size += PSIZE;
1100
1101 /*
1102 * propagate up the router entry for the leaf page just split
1103 *
1104 * insert a router entry for the new page into the parent page,
1105 * propagate the insert/split up the tree by walking back the stack
1106 * of (bn of parent page, index of child page entry in parent page)
1107 * that were traversed during the search for the page that split.
1108 *
1109 * the propagation of insert/split up the tree stops if the root
1110 * splits or the page inserted into doesn't have to split to hold
1111 * the new entry.
1112 *
1113 * the parent entry for the split page remains the same, and
1114 * a new entry is inserted at its right with the first key and
1115 * block number of the new right page.
1116 *
1117 * There are a maximum of 4 pages pinned at any time:
1118 * two children, left parent and right parent (when the parent splits).
1119 * keep the child pages pinned while working on the parent.
1120 * make sure that all pins are released at exit.
1121 */
1122 while ((parent = BT_POP(btstack)) != NULL) {
1123 /* parent page specified by stack frame <parent> */
1124
1125 /* keep current child pages (<lp>, <rp>) pinned */
1126 lmp = smp;
1127 lp = sp;
1128
1129 /*
1130 * insert router entry in parent for new right child page <rp>
1131 */
1132 /* get the parent page <sp> */
1133 DT_GETPAGE(ip, parent->bn, smp, PSIZE, sp, rc);
1134 if (rc) {
1135 DT_PUTPAGE(lmp);
1136 DT_PUTPAGE(rmp);
1137 goto splitOut;
1138 }
1139
1140 /*
1141 * The new key entry goes ONE AFTER the index of parent entry,
1142 * because the split was to the right.
1143 */
1144 skip = parent->index + 1;
1145
1146 /*
1147 * compute the key for the router entry
1148 *
1149 * key suffix compression:
1150 * for internal pages that have leaf pages as children,
1151 * retain only what's needed to distinguish between
1152 * the new entry and the entry on the page to its left.
1153 * If the keys compare equal, retain the entire key.
1154 *
1155 * note that compression is performed only at computing
1156 * router key at the lowest internal level.
1157 * further compression of the key between pairs of higher
1158 * level internal pages loses too much information and
1159 * the search may fail.
1160 * (e.g., two adjacent leaf pages of {a, ..., x} {xx, ...,}
1161 * results in two adjacent parent entries (a)(xx).
1162 * if split occurs between these two entries, and
1163 * if compression is applied, the router key of parent entry
1164 * of right page (x) will divert search for x into right
1165 * subtree and miss x in the left subtree.)
1166 *
1167 * the entire key must be retained for the next-to-leftmost
1168 * internal key at any level of the tree, or search may fail
1169 * (e.g., ?)
1170 */
1171 switch (rp->header.flag & BT_TYPE) {
1172 case BT_LEAF:
1173 /*
1174 * compute the length of prefix for suffix compression
1175 * between last entry of left page and first entry
1176 * of right page
1177 */
1178 if ((sp->header.flag & BT_ROOT && skip > 1) ||
1179 sp->header.prev != 0 || skip > 1) {
1180 /* compute uppercase router prefix key */
1181 rc = ciGetLeafPrefixKey(lp,
1182 lp->header.nextindex-1,
1183 rp, 0, &key,
1184 sbi->mntflag);
1185 if (rc) {
1186 DT_PUTPAGE(lmp);
1187 DT_PUTPAGE(rmp);
1188 DT_PUTPAGE(smp);
1189 goto splitOut;
1190 }
1191 } else {
1192 /* next to leftmost entry of
1193 lowest internal level */
1194
1195 /* compute uppercase router key */
1196 dtGetKey(rp, 0, &key, sbi->mntflag);
1197 key.name[key.namlen] = 0;
1198
1199 if ((sbi->mntflag & JFS_OS2) == JFS_OS2)
1200 ciToUpper(&key);
1201 }
1202
1203 n = NDTINTERNAL(key.namlen);
1204 break;
1205
1206 case BT_INTERNAL:
1207 dtGetKey(rp, 0, &key, sbi->mntflag);
1208 n = NDTINTERNAL(key.namlen);
1209 break;
1210
1211 default:
1212 jfs_err("dtSplitUp(): UFO!");
1213 break;
1214 }
1215
1216 /* unpin left child page */
1217 DT_PUTPAGE(lmp);
1218
1219 /*
1220 * compute the data for the router entry
1221 */
1222 data->xd = rpxd; /* child page xd */
1223
1224 /*
1225 * parent page is full - split the parent page
1226 */
1227 if (n > sp->header.freecnt) {
1228 /* init for parent page split */
1229 split->mp = smp;
1230 split->index = skip; /* index at insert */
1231 split->nslot = n;
1232 split->key = &key;
1233 /* split->data = data; */
1234
1235 /* unpin right child page */
1236 DT_PUTPAGE(rmp);
1237
1238 /* The split routines insert the new entry,
1239 * acquire txLock as appropriate.
1240 * return <rp> pinned and its block number <rbn>.
1241 */
1242 rc = (sp->header.flag & BT_ROOT) ?
1243 dtSplitRoot(tid, ip, split, &rmp) :
1244 dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd);
1245 if (rc) {
1246 DT_PUTPAGE(smp);
1247 goto splitOut;
1248 }
1249
1250 /* smp and rmp are pinned */
1251 }
1252 /*
1253 * parent page is not full - insert router entry in parent page
1254 */
1255 else {
1256 BT_MARK_DIRTY(smp, ip);
1257 /*
1258 * acquire a transaction lock on the parent page
1259 */
1260 tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1261 dtlck = (struct dt_lock *) & tlck->lock;
1262 ASSERT(dtlck->index == 0);
1263 lv = & dtlck->lv[0];
1264
1265 /* linelock header */
1266 lv->offset = 0;
1267 lv->length = 1;
1268 dtlck->index++;
1269
1270 /* linelock stbl of non-root parent page */
1271 if (!(sp->header.flag & BT_ROOT)) {
1272 lv++;
1273 n = skip >> L2DTSLOTSIZE;
1274 lv->offset = sp->header.stblindex + n;
1275 lv->length =
1276 ((sp->header.nextindex -
1277 1) >> L2DTSLOTSIZE) - n + 1;
1278 dtlck->index++;
1279 }
1280
1281 dtInsertEntry(sp, skip, &key, data, &dtlck);
1282
1283 /* exit propagate up */
1284 break;
1285 }
1286 }
1287
1288 /* unpin current split and its right page */
1289 DT_PUTPAGE(smp);
1290 DT_PUTPAGE(rmp);
1291
1292 /*
1293 * free remaining extents allocated for split
1294 */
1295 splitOut:
1296 n = pxdlist.npxd;
1297 pxd = &pxdlist.pxd[n];
1298 for (; n < pxdlist.maxnpxd; n++, pxd++)
1299 dbFree(ip, addressPXD(pxd), (s64) lengthPXD(pxd));
1300
1301 freeKeyName:
1302 kfree(key.name);
1303
1304 /* Rollback quota allocation */
1305 if (rc && quota_allocation)
1306 dquot_free_block(ip, quota_allocation);
1307
1308 dtSplitUp_Exit:
1309
1310 return rc;
1311}
1312
1313
1314/*
1315 * dtSplitPage()
1316 *
1317 * function: Split a non-root page of a btree.
1318 *
1319 * parameter:
1320 *
1321 * return: 0 - success;
1322 * errno - failure;
1323 * return split and new page pinned;
1324 */
1325static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
1326 struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rpxdp)
1327{
1328 int rc = 0;
1329 struct metapage *smp;
1330 dtpage_t *sp;
1331 struct metapage *rmp;
1332 dtpage_t *rp; /* new right page allocated */
1333 s64 rbn; /* new right page block number */
1334 struct metapage *mp;
1335 dtpage_t *p;
1336 s64 nextbn;
1337 struct pxdlist *pxdlist;
1338 pxd_t *pxd;
1339 int skip, nextindex, half, left, nxt, off, si;
1340 struct ldtentry *ldtentry;
1341 struct idtentry *idtentry;
1342 u8 *stbl;
1343 struct dtslot *f;
1344 int fsi, stblsize;
1345 int n;
1346 struct dt_lock *sdtlck, *rdtlck;
1347 struct tlock *tlck;
1348 struct dt_lock *dtlck;
1349 struct lv *slv, *rlv, *lv;
1350
1351 /* get split page */
1352 smp = split->mp;
1353 sp = DT_PAGE(ip, smp);
1354
1355 /*
1356 * allocate the new right page for the split
1357 */
1358 pxdlist = split->pxdlist;
1359 pxd = &pxdlist->pxd[pxdlist->npxd];
1360 pxdlist->npxd++;
1361 rbn = addressPXD(pxd);
1362 rmp = get_metapage(ip, rbn, PSIZE, 1);
1363 if (rmp == NULL)
1364 return -EIO;
1365
1366 /* Allocate blocks to quota. */
1367 rc = dquot_alloc_block(ip, lengthPXD(pxd));
1368 if (rc) {
1369 release_metapage(rmp);
1370 return rc;
1371 }
1372
1373 jfs_info("dtSplitPage: ip:0x%p smp:0x%p rmp:0x%p", ip, smp, rmp);
1374
1375 BT_MARK_DIRTY(rmp, ip);
1376 /*
1377 * acquire a transaction lock on the new right page
1378 */
1379 tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1380 rdtlck = (struct dt_lock *) & tlck->lock;
1381
1382 rp = (dtpage_t *) rmp->data;
1383 *rpp = rp;
1384 rp->header.self = *pxd;
1385
1386 BT_MARK_DIRTY(smp, ip);
1387 /*
1388 * acquire a transaction lock on the split page
1389 *
1390 * action:
1391 */
1392 tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1393 sdtlck = (struct dt_lock *) & tlck->lock;
1394
1395 /* linelock header of split page */
1396 ASSERT(sdtlck->index == 0);
1397 slv = & sdtlck->lv[0];
1398 slv->offset = 0;
1399 slv->length = 1;
1400 sdtlck->index++;
1401
1402 /*
1403 * initialize/update sibling pointers between sp and rp
1404 */
1405 nextbn = le64_to_cpu(sp->header.next);
1406 rp->header.next = cpu_to_le64(nextbn);
1407 rp->header.prev = cpu_to_le64(addressPXD(&sp->header.self));
1408 sp->header.next = cpu_to_le64(rbn);
1409
1410 /*
1411 * initialize new right page
1412 */
1413 rp->header.flag = sp->header.flag;
1414
1415 /* compute sorted entry table at start of extent data area */
1416 rp->header.nextindex = 0;
1417 rp->header.stblindex = 1;
1418
1419 n = PSIZE >> L2DTSLOTSIZE;
1420 rp->header.maxslot = n;
1421 stblsize = (n + 31) >> L2DTSLOTSIZE; /* in unit of slot */
1422
1423 /* init freelist */
1424 fsi = rp->header.stblindex + stblsize;
1425 rp->header.freelist = fsi;
1426 rp->header.freecnt = rp->header.maxslot - fsi;
1427
1428 /*
1429 * sequential append at tail: append without split
1430 *
1431 * If splitting the last page on a level because of appending
1432 * a entry to it (skip is maxentry), it's likely that the access is
1433 * sequential. Adding an empty page on the side of the level is less
1434 * work and can push the fill factor much higher than normal.
1435 * If we're wrong it's no big deal, we'll just do the split the right
1436 * way next time.
1437 * (It may look like it's equally easy to do a similar hack for
1438 * reverse sorted data, that is, split the tree left,
1439 * but it's not. Be my guest.)
1440 */
1441 if (nextbn == 0 && split->index == sp->header.nextindex) {
1442 /* linelock header + stbl (first slot) of new page */
1443 rlv = & rdtlck->lv[rdtlck->index];
1444 rlv->offset = 0;
1445 rlv->length = 2;
1446 rdtlck->index++;
1447
1448 /*
1449 * initialize freelist of new right page
1450 */
1451 f = &rp->slot[fsi];
1452 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1453 f->next = fsi;
1454 f->next = -1;
1455
1456 /* insert entry at the first entry of the new right page */
1457 dtInsertEntry(rp, 0, split->key, split->data, &rdtlck);
1458
1459 goto out;
1460 }
1461
1462 /*
1463 * non-sequential insert (at possibly middle page)
1464 */
1465
1466 /*
1467 * update prev pointer of previous right sibling page;
1468 */
1469 if (nextbn != 0) {
1470 DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
1471 if (rc) {
1472 discard_metapage(rmp);
1473 return rc;
1474 }
1475
1476 BT_MARK_DIRTY(mp, ip);
1477 /*
1478 * acquire a transaction lock on the next page
1479 */
1480 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
1481 jfs_info("dtSplitPage: tlck = 0x%p, ip = 0x%p, mp=0x%p",
1482 tlck, ip, mp);
1483 dtlck = (struct dt_lock *) & tlck->lock;
1484
1485 /* linelock header of previous right sibling page */
1486 lv = & dtlck->lv[dtlck->index];
1487 lv->offset = 0;
1488 lv->length = 1;
1489 dtlck->index++;
1490
1491 p->header.prev = cpu_to_le64(rbn);
1492
1493 DT_PUTPAGE(mp);
1494 }
1495
1496 /*
1497 * split the data between the split and right pages.
1498 */
1499 skip = split->index;
1500 half = (PSIZE >> L2DTSLOTSIZE) >> 1; /* swag */
1501 left = 0;
1502
1503 /*
1504 * compute fill factor for split pages
1505 *
1506 * <nxt> traces the next entry to move to rp
1507 * <off> traces the next entry to stay in sp
1508 */
1509 stbl = (u8 *) & sp->slot[sp->header.stblindex];
1510 nextindex = sp->header.nextindex;
1511 for (nxt = off = 0; nxt < nextindex; ++off) {
1512 if (off == skip)
1513 /* check for fill factor with new entry size */
1514 n = split->nslot;
1515 else {
1516 si = stbl[nxt];
1517 switch (sp->header.flag & BT_TYPE) {
1518 case BT_LEAF:
1519 ldtentry = (struct ldtentry *) & sp->slot[si];
1520 if (DO_INDEX(ip))
1521 n = NDTLEAF(ldtentry->namlen);
1522 else
1523 n = NDTLEAF_LEGACY(ldtentry->
1524 namlen);
1525 break;
1526
1527 case BT_INTERNAL:
1528 idtentry = (struct idtentry *) & sp->slot[si];
1529 n = NDTINTERNAL(idtentry->namlen);
1530 break;
1531
1532 default:
1533 break;
1534 }
1535
1536 ++nxt; /* advance to next entry to move in sp */
1537 }
1538
1539 left += n;
1540 if (left >= half)
1541 break;
1542 }
1543
1544 /* <nxt> poins to the 1st entry to move */
1545
1546 /*
1547 * move entries to right page
1548 *
1549 * dtMoveEntry() initializes rp and reserves entry for insertion
1550 *
1551 * split page moved out entries are linelocked;
1552 * new/right page moved in entries are linelocked;
1553 */
1554 /* linelock header + stbl of new right page */
1555 rlv = & rdtlck->lv[rdtlck->index];
1556 rlv->offset = 0;
1557 rlv->length = 5;
1558 rdtlck->index++;
1559
1560 dtMoveEntry(sp, nxt, rp, &sdtlck, &rdtlck, DO_INDEX(ip));
1561
1562 sp->header.nextindex = nxt;
1563
1564 /*
1565 * finalize freelist of new right page
1566 */
1567 fsi = rp->header.freelist;
1568 f = &rp->slot[fsi];
1569 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1570 f->next = fsi;
1571 f->next = -1;
1572
1573 /*
1574 * Update directory index table for entries now in right page
1575 */
1576 if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1577 s64 lblock;
1578
1579 mp = NULL;
1580 stbl = DT_GETSTBL(rp);
1581 for (n = 0; n < rp->header.nextindex; n++) {
1582 ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
1583 modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1584 rbn, n, &mp, &lblock);
1585 }
1586 if (mp)
1587 release_metapage(mp);
1588 }
1589
1590 /*
1591 * the skipped index was on the left page,
1592 */
1593 if (skip <= off) {
1594 /* insert the new entry in the split page */
1595 dtInsertEntry(sp, skip, split->key, split->data, &sdtlck);
1596
1597 /* linelock stbl of split page */
1598 if (sdtlck->index >= sdtlck->maxcnt)
1599 sdtlck = (struct dt_lock *) txLinelock(sdtlck);
1600 slv = & sdtlck->lv[sdtlck->index];
1601 n = skip >> L2DTSLOTSIZE;
1602 slv->offset = sp->header.stblindex + n;
1603 slv->length =
1604 ((sp->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
1605 sdtlck->index++;
1606 }
1607 /*
1608 * the skipped index was on the right page,
1609 */
1610 else {
1611 /* adjust the skip index to reflect the new position */
1612 skip -= nxt;
1613
1614 /* insert the new entry in the right page */
1615 dtInsertEntry(rp, skip, split->key, split->data, &rdtlck);
1616 }
1617
1618 out:
1619 *rmpp = rmp;
1620 *rpxdp = *pxd;
1621
1622 return rc;
1623}
1624
1625
1626/*
1627 * dtExtendPage()
1628 *
1629 * function: extend 1st/only directory leaf page
1630 *
1631 * parameter:
1632 *
1633 * return: 0 - success;
1634 * errno - failure;
1635 * return extended page pinned;
1636 */
1637static int dtExtendPage(tid_t tid,
1638 struct inode *ip, struct dtsplit * split, struct btstack * btstack)
1639{
1640 struct super_block *sb = ip->i_sb;
1641 int rc;
1642 struct metapage *smp, *pmp, *mp;
1643 dtpage_t *sp, *pp;
1644 struct pxdlist *pxdlist;
1645 pxd_t *pxd, *tpxd;
1646 int xlen, xsize;
1647 int newstblindex, newstblsize;
1648 int oldstblindex, oldstblsize;
1649 int fsi, last;
1650 struct dtslot *f;
1651 struct btframe *parent;
1652 int n;
1653 struct dt_lock *dtlck;
1654 s64 xaddr, txaddr;
1655 struct tlock *tlck;
1656 struct pxd_lock *pxdlock;
1657 struct lv *lv;
1658 uint type;
1659 struct ldtentry *ldtentry;
1660 u8 *stbl;
1661
1662 /* get page to extend */
1663 smp = split->mp;
1664 sp = DT_PAGE(ip, smp);
1665
1666 /* get parent/root page */
1667 parent = BT_POP(btstack);
1668 DT_GETPAGE(ip, parent->bn, pmp, PSIZE, pp, rc);
1669 if (rc)
1670 return (rc);
1671
1672 /*
1673 * extend the extent
1674 */
1675 pxdlist = split->pxdlist;
1676 pxd = &pxdlist->pxd[pxdlist->npxd];
1677 pxdlist->npxd++;
1678
1679 xaddr = addressPXD(pxd);
1680 tpxd = &sp->header.self;
1681 txaddr = addressPXD(tpxd);
1682 /* in-place extension */
1683 if (xaddr == txaddr) {
1684 type = tlckEXTEND;
1685 }
1686 /* relocation */
1687 else {
1688 type = tlckNEW;
1689
1690 /* save moved extent descriptor for later free */
1691 tlck = txMaplock(tid, ip, tlckDTREE | tlckRELOCATE);
1692 pxdlock = (struct pxd_lock *) & tlck->lock;
1693 pxdlock->flag = mlckFREEPXD;
1694 pxdlock->pxd = sp->header.self;
1695 pxdlock->index = 1;
1696
1697 /*
1698 * Update directory index table to reflect new page address
1699 */
1700 if (DO_INDEX(ip)) {
1701 s64 lblock;
1702
1703 mp = NULL;
1704 stbl = DT_GETSTBL(sp);
1705 for (n = 0; n < sp->header.nextindex; n++) {
1706 ldtentry =
1707 (struct ldtentry *) & sp->slot[stbl[n]];
1708 modify_index(tid, ip,
1709 le32_to_cpu(ldtentry->index),
1710 xaddr, n, &mp, &lblock);
1711 }
1712 if (mp)
1713 release_metapage(mp);
1714 }
1715 }
1716
1717 /*
1718 * extend the page
1719 */
1720 sp->header.self = *pxd;
1721
1722 jfs_info("dtExtendPage: ip:0x%p smp:0x%p sp:0x%p", ip, smp, sp);
1723
1724 BT_MARK_DIRTY(smp, ip);
1725 /*
1726 * acquire a transaction lock on the extended/leaf page
1727 */
1728 tlck = txLock(tid, ip, smp, tlckDTREE | type);
1729 dtlck = (struct dt_lock *) & tlck->lock;
1730 lv = & dtlck->lv[0];
1731
1732 /* update buffer extent descriptor of extended page */
1733 xlen = lengthPXD(pxd);
1734 xsize = xlen << JFS_SBI(sb)->l2bsize;
1735
1736 /*
1737 * copy old stbl to new stbl at start of extended area
1738 */
1739 oldstblindex = sp->header.stblindex;
1740 oldstblsize = (sp->header.maxslot + 31) >> L2DTSLOTSIZE;
1741 newstblindex = sp->header.maxslot;
1742 n = xsize >> L2DTSLOTSIZE;
1743 newstblsize = (n + 31) >> L2DTSLOTSIZE;
1744 memcpy(&sp->slot[newstblindex], &sp->slot[oldstblindex],
1745 sp->header.nextindex);
1746
1747 /*
1748 * in-line extension: linelock old area of extended page
1749 */
1750 if (type == tlckEXTEND) {
1751 /* linelock header */
1752 lv->offset = 0;
1753 lv->length = 1;
1754 dtlck->index++;
1755 lv++;
1756
1757 /* linelock new stbl of extended page */
1758 lv->offset = newstblindex;
1759 lv->length = newstblsize;
1760 }
1761 /*
1762 * relocation: linelock whole relocated area
1763 */
1764 else {
1765 lv->offset = 0;
1766 lv->length = sp->header.maxslot + newstblsize;
1767 }
1768
1769 dtlck->index++;
1770
1771 sp->header.maxslot = n;
1772 sp->header.stblindex = newstblindex;
1773 /* sp->header.nextindex remains the same */
1774
1775 /*
1776 * add old stbl region at head of freelist
1777 */
1778 fsi = oldstblindex;
1779 f = &sp->slot[fsi];
1780 last = sp->header.freelist;
1781 for (n = 0; n < oldstblsize; n++, fsi++, f++) {
1782 f->next = last;
1783 last = fsi;
1784 }
1785 sp->header.freelist = last;
1786 sp->header.freecnt += oldstblsize;
1787
1788 /*
1789 * append free region of newly extended area at tail of freelist
1790 */
1791 /* init free region of newly extended area */
1792 fsi = n = newstblindex + newstblsize;
1793 f = &sp->slot[fsi];
1794 for (fsi++; fsi < sp->header.maxslot; f++, fsi++)
1795 f->next = fsi;
1796 f->next = -1;
1797
1798 /* append new free region at tail of old freelist */
1799 fsi = sp->header.freelist;
1800 if (fsi == -1)
1801 sp->header.freelist = n;
1802 else {
1803 do {
1804 f = &sp->slot[fsi];
1805 fsi = f->next;
1806 } while (fsi != -1);
1807
1808 f->next = n;
1809 }
1810
1811 sp->header.freecnt += sp->header.maxslot - n;
1812
1813 /*
1814 * insert the new entry
1815 */
1816 dtInsertEntry(sp, split->index, split->key, split->data, &dtlck);
1817
1818 BT_MARK_DIRTY(pmp, ip);
1819 /*
1820 * linelock any freeslots residing in old extent
1821 */
1822 if (type == tlckEXTEND) {
1823 n = sp->header.maxslot >> 2;
1824 if (sp->header.freelist < n)
1825 dtLinelockFreelist(sp, n, &dtlck);
1826 }
1827
1828 /*
1829 * update parent entry on the parent/root page
1830 */
1831 /*
1832 * acquire a transaction lock on the parent/root page
1833 */
1834 tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
1835 dtlck = (struct dt_lock *) & tlck->lock;
1836 lv = & dtlck->lv[dtlck->index];
1837
1838 /* linelock parent entry - 1st slot */
1839 lv->offset = 1;
1840 lv->length = 1;
1841 dtlck->index++;
1842
1843 /* update the parent pxd for page extension */
1844 tpxd = (pxd_t *) & pp->slot[1];
1845 *tpxd = *pxd;
1846
1847 DT_PUTPAGE(pmp);
1848 return 0;
1849}
1850
1851
1852/*
1853 * dtSplitRoot()
1854 *
1855 * function:
1856 * split the full root page into
1857 * original/root/split page and new right page
1858 * i.e., root remains fixed in tree anchor (inode) and
1859 * the root is copied to a single new right child page
1860 * since root page << non-root page, and
1861 * the split root page contains a single entry for the
1862 * new right child page.
1863 *
1864 * parameter:
1865 *
1866 * return: 0 - success;
1867 * errno - failure;
1868 * return new page pinned;
1869 */
1870static int dtSplitRoot(tid_t tid,
1871 struct inode *ip, struct dtsplit * split, struct metapage ** rmpp)
1872{
1873 struct super_block *sb = ip->i_sb;
1874 struct metapage *smp;
1875 dtroot_t *sp;
1876 struct metapage *rmp;
1877 dtpage_t *rp;
1878 s64 rbn;
1879 int xlen;
1880 int xsize;
1881 struct dtslot *f;
1882 s8 *stbl;
1883 int fsi, stblsize, n;
1884 struct idtentry *s;
1885 pxd_t *ppxd;
1886 struct pxdlist *pxdlist;
1887 pxd_t *pxd;
1888 struct dt_lock *dtlck;
1889 struct tlock *tlck;
1890 struct lv *lv;
1891 int rc;
1892
1893 /* get split root page */
1894 smp = split->mp;
1895 sp = &JFS_IP(ip)->i_dtroot;
1896
1897 /*
1898 * allocate/initialize a single (right) child page
1899 *
1900 * N.B. at first split, a one (or two) block to fit new entry
1901 * is allocated; at subsequent split, a full page is allocated;
1902 */
1903 pxdlist = split->pxdlist;
1904 pxd = &pxdlist->pxd[pxdlist->npxd];
1905 pxdlist->npxd++;
1906 rbn = addressPXD(pxd);
1907 xlen = lengthPXD(pxd);
1908 xsize = xlen << JFS_SBI(sb)->l2bsize;
1909 rmp = get_metapage(ip, rbn, xsize, 1);
1910 if (!rmp)
1911 return -EIO;
1912
1913 rp = rmp->data;
1914
1915 /* Allocate blocks to quota. */
1916 rc = dquot_alloc_block(ip, lengthPXD(pxd));
1917 if (rc) {
1918 release_metapage(rmp);
1919 return rc;
1920 }
1921
1922 BT_MARK_DIRTY(rmp, ip);
1923 /*
1924 * acquire a transaction lock on the new right page
1925 */
1926 tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1927 dtlck = (struct dt_lock *) & tlck->lock;
1928
1929 rp->header.flag =
1930 (sp->header.flag & BT_LEAF) ? BT_LEAF : BT_INTERNAL;
1931 rp->header.self = *pxd;
1932
1933 /* initialize sibling pointers */
1934 rp->header.next = 0;
1935 rp->header.prev = 0;
1936
1937 /*
1938 * move in-line root page into new right page extent
1939 */
1940 /* linelock header + copied entries + new stbl (1st slot) in new page */
1941 ASSERT(dtlck->index == 0);
1942 lv = & dtlck->lv[0];
1943 lv->offset = 0;
1944 lv->length = 10; /* 1 + 8 + 1 */
1945 dtlck->index++;
1946
1947 n = xsize >> L2DTSLOTSIZE;
1948 rp->header.maxslot = n;
1949 stblsize = (n + 31) >> L2DTSLOTSIZE;
1950
1951 /* copy old stbl to new stbl at start of extended area */
1952 rp->header.stblindex = DTROOTMAXSLOT;
1953 stbl = (s8 *) & rp->slot[DTROOTMAXSLOT];
1954 memcpy(stbl, sp->header.stbl, sp->header.nextindex);
1955 rp->header.nextindex = sp->header.nextindex;
1956
1957 /* copy old data area to start of new data area */
1958 memcpy(&rp->slot[1], &sp->slot[1], IDATASIZE);
1959
1960 /*
1961 * append free region of newly extended area at tail of freelist
1962 */
1963 /* init free region of newly extended area */
1964 fsi = n = DTROOTMAXSLOT + stblsize;
1965 f = &rp->slot[fsi];
1966 for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1967 f->next = fsi;
1968 f->next = -1;
1969
1970 /* append new free region at tail of old freelist */
1971 fsi = sp->header.freelist;
1972 if (fsi == -1)
1973 rp->header.freelist = n;
1974 else {
1975 rp->header.freelist = fsi;
1976
1977 do {
1978 f = &rp->slot[fsi];
1979 fsi = f->next;
1980 } while (fsi >= 0);
1981
1982 f->next = n;
1983 }
1984
1985 rp->header.freecnt = sp->header.freecnt + rp->header.maxslot - n;
1986
1987 /*
1988 * Update directory index table for entries now in right page
1989 */
1990 if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1991 s64 lblock;
1992 struct metapage *mp = NULL;
1993 struct ldtentry *ldtentry;
1994
1995 stbl = DT_GETSTBL(rp);
1996 for (n = 0; n < rp->header.nextindex; n++) {
1997 ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
1998 modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1999 rbn, n, &mp, &lblock);
2000 }
2001 if (mp)
2002 release_metapage(mp);
2003 }
2004 /*
2005 * insert the new entry into the new right/child page
2006 * (skip index in the new right page will not change)
2007 */
2008 dtInsertEntry(rp, split->index, split->key, split->data, &dtlck);
2009
2010 /*
2011 * reset parent/root page
2012 *
2013 * set the 1st entry offset to 0, which force the left-most key
2014 * at any level of the tree to be less than any search key.
2015 *
2016 * The btree comparison code guarantees that the left-most key on any
2017 * level of the tree is never used, so it doesn't need to be filled in.
2018 */
2019 BT_MARK_DIRTY(smp, ip);
2020 /*
2021 * acquire a transaction lock on the root page (in-memory inode)
2022 */
2023 tlck = txLock(tid, ip, smp, tlckDTREE | tlckNEW | tlckBTROOT);
2024 dtlck = (struct dt_lock *) & tlck->lock;
2025
2026 /* linelock root */
2027 ASSERT(dtlck->index == 0);
2028 lv = & dtlck->lv[0];
2029 lv->offset = 0;
2030 lv->length = DTROOTMAXSLOT;
2031 dtlck->index++;
2032
2033 /* update page header of root */
2034 if (sp->header.flag & BT_LEAF) {
2035 sp->header.flag &= ~BT_LEAF;
2036 sp->header.flag |= BT_INTERNAL;
2037 }
2038
2039 /* init the first entry */
2040 s = (struct idtentry *) & sp->slot[DTENTRYSTART];
2041 ppxd = (pxd_t *) s;
2042 *ppxd = *pxd;
2043 s->next = -1;
2044 s->namlen = 0;
2045
2046 stbl = sp->header.stbl;
2047 stbl[0] = DTENTRYSTART;
2048 sp->header.nextindex = 1;
2049
2050 /* init freelist */
2051 fsi = DTENTRYSTART + 1;
2052 f = &sp->slot[fsi];
2053
2054 /* init free region of remaining area */
2055 for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2056 f->next = fsi;
2057 f->next = -1;
2058
2059 sp->header.freelist = DTENTRYSTART + 1;
2060 sp->header.freecnt = DTROOTMAXSLOT - (DTENTRYSTART + 1);
2061
2062 *rmpp = rmp;
2063
2064 return 0;
2065}
2066
2067
2068/*
2069 * dtDelete()
2070 *
2071 * function: delete the entry(s) referenced by a key.
2072 *
2073 * parameter:
2074 *
2075 * return:
2076 */
2077int dtDelete(tid_t tid,
2078 struct inode *ip, struct component_name * key, ino_t * ino, int flag)
2079{
2080 int rc = 0;
2081 s64 bn;
2082 struct metapage *mp, *imp;
2083 dtpage_t *p;
2084 int index;
2085 struct btstack btstack;
2086 struct dt_lock *dtlck;
2087 struct tlock *tlck;
2088 struct lv *lv;
2089 int i;
2090 struct ldtentry *ldtentry;
2091 u8 *stbl;
2092 u32 table_index, next_index;
2093 struct metapage *nmp;
2094 dtpage_t *np;
2095
2096 /*
2097 * search for the entry to delete:
2098 *
2099 * dtSearch() returns (leaf page pinned, index at which to delete).
2100 */
2101 if ((rc = dtSearch(ip, key, ino, &btstack, flag)))
2102 return rc;
2103
2104 /* retrieve search result */
2105 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2106
2107 /*
2108 * We need to find put the index of the next entry into the
2109 * directory index table in order to resume a readdir from this
2110 * entry.
2111 */
2112 if (DO_INDEX(ip)) {
2113 stbl = DT_GETSTBL(p);
2114 ldtentry = (struct ldtentry *) & p->slot[stbl[index]];
2115 table_index = le32_to_cpu(ldtentry->index);
2116 if (index == (p->header.nextindex - 1)) {
2117 /*
2118 * Last entry in this leaf page
2119 */
2120 if ((p->header.flag & BT_ROOT)
2121 || (p->header.next == 0))
2122 next_index = -1;
2123 else {
2124 /* Read next leaf page */
2125 DT_GETPAGE(ip, le64_to_cpu(p->header.next),
2126 nmp, PSIZE, np, rc);
2127 if (rc)
2128 next_index = -1;
2129 else {
2130 stbl = DT_GETSTBL(np);
2131 ldtentry =
2132 (struct ldtentry *) & np->
2133 slot[stbl[0]];
2134 next_index =
2135 le32_to_cpu(ldtentry->index);
2136 DT_PUTPAGE(nmp);
2137 }
2138 }
2139 } else {
2140 ldtentry =
2141 (struct ldtentry *) & p->slot[stbl[index + 1]];
2142 next_index = le32_to_cpu(ldtentry->index);
2143 }
2144 free_index(tid, ip, table_index, next_index);
2145 }
2146 /*
2147 * the leaf page becomes empty, delete the page
2148 */
2149 if (p->header.nextindex == 1) {
2150 /* delete empty page */
2151 rc = dtDeleteUp(tid, ip, mp, p, &btstack);
2152 }
2153 /*
2154 * the leaf page has other entries remaining:
2155 *
2156 * delete the entry from the leaf page.
2157 */
2158 else {
2159 BT_MARK_DIRTY(mp, ip);
2160 /*
2161 * acquire a transaction lock on the leaf page
2162 */
2163 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2164 dtlck = (struct dt_lock *) & tlck->lock;
2165
2166 /*
2167 * Do not assume that dtlck->index will be zero. During a
2168 * rename within a directory, this transaction may have
2169 * modified this page already when adding the new entry.
2170 */
2171
2172 /* linelock header */
2173 if (dtlck->index >= dtlck->maxcnt)
2174 dtlck = (struct dt_lock *) txLinelock(dtlck);
2175 lv = & dtlck->lv[dtlck->index];
2176 lv->offset = 0;
2177 lv->length = 1;
2178 dtlck->index++;
2179
2180 /* linelock stbl of non-root leaf page */
2181 if (!(p->header.flag & BT_ROOT)) {
2182 if (dtlck->index >= dtlck->maxcnt)
2183 dtlck = (struct dt_lock *) txLinelock(dtlck);
2184 lv = & dtlck->lv[dtlck->index];
2185 i = index >> L2DTSLOTSIZE;
2186 lv->offset = p->header.stblindex + i;
2187 lv->length =
2188 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2189 i + 1;
2190 dtlck->index++;
2191 }
2192
2193 /* free the leaf entry */
2194 dtDeleteEntry(p, index, &dtlck);
2195
2196 /*
2197 * Update directory index table for entries moved in stbl
2198 */
2199 if (DO_INDEX(ip) && index < p->header.nextindex) {
2200 s64 lblock;
2201
2202 imp = NULL;
2203 stbl = DT_GETSTBL(p);
2204 for (i = index; i < p->header.nextindex; i++) {
2205 ldtentry =
2206 (struct ldtentry *) & p->slot[stbl[i]];
2207 modify_index(tid, ip,
2208 le32_to_cpu(ldtentry->index),
2209 bn, i, &imp, &lblock);
2210 }
2211 if (imp)
2212 release_metapage(imp);
2213 }
2214
2215 DT_PUTPAGE(mp);
2216 }
2217
2218 return rc;
2219}
2220
2221
2222/*
2223 * dtDeleteUp()
2224 *
2225 * function:
2226 * free empty pages as propagating deletion up the tree
2227 *
2228 * parameter:
2229 *
2230 * return:
2231 */
2232static int dtDeleteUp(tid_t tid, struct inode *ip,
2233 struct metapage * fmp, dtpage_t * fp, struct btstack * btstack)
2234{
2235 int rc = 0;
2236 struct metapage *mp;
2237 dtpage_t *p;
2238 int index, nextindex;
2239 int xlen;
2240 struct btframe *parent;
2241 struct dt_lock *dtlck;
2242 struct tlock *tlck;
2243 struct lv *lv;
2244 struct pxd_lock *pxdlock;
2245 int i;
2246
2247 /*
2248 * keep the root leaf page which has become empty
2249 */
2250 if (BT_IS_ROOT(fmp)) {
2251 /*
2252 * reset the root
2253 *
2254 * dtInitRoot() acquires txlock on the root
2255 */
2256 dtInitRoot(tid, ip, PARENT(ip));
2257
2258 DT_PUTPAGE(fmp);
2259
2260 return 0;
2261 }
2262
2263 /*
2264 * free the non-root leaf page
2265 */
2266 /*
2267 * acquire a transaction lock on the page
2268 *
2269 * write FREEXTENT|NOREDOPAGE log record
2270 * N.B. linelock is overlaid as freed extent descriptor, and
2271 * the buffer page is freed;
2272 */
2273 tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2274 pxdlock = (struct pxd_lock *) & tlck->lock;
2275 pxdlock->flag = mlckFREEPXD;
2276 pxdlock->pxd = fp->header.self;
2277 pxdlock->index = 1;
2278
2279 /* update sibling pointers */
2280 if ((rc = dtRelink(tid, ip, fp))) {
2281 BT_PUTPAGE(fmp);
2282 return rc;
2283 }
2284
2285 xlen = lengthPXD(&fp->header.self);
2286
2287 /* Free quota allocation. */
2288 dquot_free_block(ip, xlen);
2289
2290 /* free/invalidate its buffer page */
2291 discard_metapage(fmp);
2292
2293 /*
2294 * propagate page deletion up the directory tree
2295 *
2296 * If the delete from the parent page makes it empty,
2297 * continue all the way up the tree.
2298 * stop if the root page is reached (which is never deleted) or
2299 * if the entry deletion does not empty the page.
2300 */
2301 while ((parent = BT_POP(btstack)) != NULL) {
2302 /* pin the parent page <sp> */
2303 DT_GETPAGE(ip, parent->bn, mp, PSIZE, p, rc);
2304 if (rc)
2305 return rc;
2306
2307 /*
2308 * free the extent of the child page deleted
2309 */
2310 index = parent->index;
2311
2312 /*
2313 * delete the entry for the child page from parent
2314 */
2315 nextindex = p->header.nextindex;
2316
2317 /*
2318 * the parent has the single entry being deleted:
2319 *
2320 * free the parent page which has become empty.
2321 */
2322 if (nextindex == 1) {
2323 /*
2324 * keep the root internal page which has become empty
2325 */
2326 if (p->header.flag & BT_ROOT) {
2327 /*
2328 * reset the root
2329 *
2330 * dtInitRoot() acquires txlock on the root
2331 */
2332 dtInitRoot(tid, ip, PARENT(ip));
2333
2334 DT_PUTPAGE(mp);
2335
2336 return 0;
2337 }
2338 /*
2339 * free the parent page
2340 */
2341 else {
2342 /*
2343 * acquire a transaction lock on the page
2344 *
2345 * write FREEXTENT|NOREDOPAGE log record
2346 */
2347 tlck =
2348 txMaplock(tid, ip,
2349 tlckDTREE | tlckFREE);
2350 pxdlock = (struct pxd_lock *) & tlck->lock;
2351 pxdlock->flag = mlckFREEPXD;
2352 pxdlock->pxd = p->header.self;
2353 pxdlock->index = 1;
2354
2355 /* update sibling pointers */
2356 if ((rc = dtRelink(tid, ip, p))) {
2357 DT_PUTPAGE(mp);
2358 return rc;
2359 }
2360
2361 xlen = lengthPXD(&p->header.self);
2362
2363 /* Free quota allocation */
2364 dquot_free_block(ip, xlen);
2365
2366 /* free/invalidate its buffer page */
2367 discard_metapage(mp);
2368
2369 /* propagate up */
2370 continue;
2371 }
2372 }
2373
2374 /*
2375 * the parent has other entries remaining:
2376 *
2377 * delete the router entry from the parent page.
2378 */
2379 BT_MARK_DIRTY(mp, ip);
2380 /*
2381 * acquire a transaction lock on the page
2382 *
2383 * action: router entry deletion
2384 */
2385 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2386 dtlck = (struct dt_lock *) & tlck->lock;
2387
2388 /* linelock header */
2389 if (dtlck->index >= dtlck->maxcnt)
2390 dtlck = (struct dt_lock *) txLinelock(dtlck);
2391 lv = & dtlck->lv[dtlck->index];
2392 lv->offset = 0;
2393 lv->length = 1;
2394 dtlck->index++;
2395
2396 /* linelock stbl of non-root leaf page */
2397 if (!(p->header.flag & BT_ROOT)) {
2398 if (dtlck->index < dtlck->maxcnt)
2399 lv++;
2400 else {
2401 dtlck = (struct dt_lock *) txLinelock(dtlck);
2402 lv = & dtlck->lv[0];
2403 }
2404 i = index >> L2DTSLOTSIZE;
2405 lv->offset = p->header.stblindex + i;
2406 lv->length =
2407 ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2408 i + 1;
2409 dtlck->index++;
2410 }
2411
2412 /* free the router entry */
2413 dtDeleteEntry(p, index, &dtlck);
2414
2415 /* reset key of new leftmost entry of level (for consistency) */
2416 if (index == 0 &&
2417 ((p->header.flag & BT_ROOT) || p->header.prev == 0))
2418 dtTruncateEntry(p, 0, &dtlck);
2419
2420 /* unpin the parent page */
2421 DT_PUTPAGE(mp);
2422
2423 /* exit propagation up */
2424 break;
2425 }
2426
2427 if (!DO_INDEX(ip))
2428 ip->i_size -= PSIZE;
2429
2430 return 0;
2431}
2432
2433/*
2434 * dtRelink()
2435 *
2436 * function:
2437 * link around a freed page.
2438 *
2439 * parameter:
2440 * fp: page to be freed
2441 *
2442 * return:
2443 */
2444static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p)
2445{
2446 int rc;
2447 struct metapage *mp;
2448 s64 nextbn, prevbn;
2449 struct tlock *tlck;
2450 struct dt_lock *dtlck;
2451 struct lv *lv;
2452
2453 nextbn = le64_to_cpu(p->header.next);
2454 prevbn = le64_to_cpu(p->header.prev);
2455
2456 /* update prev pointer of the next page */
2457 if (nextbn != 0) {
2458 DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
2459 if (rc)
2460 return rc;
2461
2462 BT_MARK_DIRTY(mp, ip);
2463 /*
2464 * acquire a transaction lock on the next page
2465 *
2466 * action: update prev pointer;
2467 */
2468 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2469 jfs_info("dtRelink nextbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2470 tlck, ip, mp);
2471 dtlck = (struct dt_lock *) & tlck->lock;
2472
2473 /* linelock header */
2474 if (dtlck->index >= dtlck->maxcnt)
2475 dtlck = (struct dt_lock *) txLinelock(dtlck);
2476 lv = & dtlck->lv[dtlck->index];
2477 lv->offset = 0;
2478 lv->length = 1;
2479 dtlck->index++;
2480
2481 p->header.prev = cpu_to_le64(prevbn);
2482 DT_PUTPAGE(mp);
2483 }
2484
2485 /* update next pointer of the previous page */
2486 if (prevbn != 0) {
2487 DT_GETPAGE(ip, prevbn, mp, PSIZE, p, rc);
2488 if (rc)
2489 return rc;
2490
2491 BT_MARK_DIRTY(mp, ip);
2492 /*
2493 * acquire a transaction lock on the prev page
2494 *
2495 * action: update next pointer;
2496 */
2497 tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2498 jfs_info("dtRelink prevbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2499 tlck, ip, mp);
2500 dtlck = (struct dt_lock *) & tlck->lock;
2501
2502 /* linelock header */
2503 if (dtlck->index >= dtlck->maxcnt)
2504 dtlck = (struct dt_lock *) txLinelock(dtlck);
2505 lv = & dtlck->lv[dtlck->index];
2506 lv->offset = 0;
2507 lv->length = 1;
2508 dtlck->index++;
2509
2510 p->header.next = cpu_to_le64(nextbn);
2511 DT_PUTPAGE(mp);
2512 }
2513
2514 return 0;
2515}
2516
2517
2518/*
2519 * dtInitRoot()
2520 *
2521 * initialize directory root (inline in inode)
2522 */
2523void dtInitRoot(tid_t tid, struct inode *ip, u32 idotdot)
2524{
2525 struct jfs_inode_info *jfs_ip = JFS_IP(ip);
2526 dtroot_t *p;
2527 int fsi;
2528 struct dtslot *f;
2529 struct tlock *tlck;
2530 struct dt_lock *dtlck;
2531 struct lv *lv;
2532 u16 xflag_save;
2533
2534 /*
2535 * If this was previously an non-empty directory, we need to remove
2536 * the old directory table.
2537 */
2538 if (DO_INDEX(ip)) {
2539 if (!jfs_dirtable_inline(ip)) {
2540 struct tblock *tblk = tid_to_tblock(tid);
2541 /*
2542 * We're playing games with the tid's xflag. If
2543 * we're removing a regular file, the file's xtree
2544 * is committed with COMMIT_PMAP, but we always
2545 * commit the directories xtree with COMMIT_PWMAP.
2546 */
2547 xflag_save = tblk->xflag;
2548 tblk->xflag = 0;
2549 /*
2550 * xtTruncate isn't guaranteed to fully truncate
2551 * the xtree. The caller needs to check i_size
2552 * after committing the transaction to see if
2553 * additional truncation is needed. The
2554 * COMMIT_Stale flag tells caller that we
2555 * initiated the truncation.
2556 */
2557 xtTruncate(tid, ip, 0, COMMIT_PWMAP);
2558 set_cflag(COMMIT_Stale, ip);
2559
2560 tblk->xflag = xflag_save;
2561 } else
2562 ip->i_size = 1;
2563
2564 jfs_ip->next_index = 2;
2565 } else
2566 ip->i_size = IDATASIZE;
2567
2568 /*
2569 * acquire a transaction lock on the root
2570 *
2571 * action: directory initialization;
2572 */
2573 tlck = txLock(tid, ip, (struct metapage *) & jfs_ip->bxflag,
2574 tlckDTREE | tlckENTRY | tlckBTROOT);
2575 dtlck = (struct dt_lock *) & tlck->lock;
2576
2577 /* linelock root */
2578 ASSERT(dtlck->index == 0);
2579 lv = & dtlck->lv[0];
2580 lv->offset = 0;
2581 lv->length = DTROOTMAXSLOT;
2582 dtlck->index++;
2583
2584 p = &jfs_ip->i_dtroot;
2585
2586 p->header.flag = DXD_INDEX | BT_ROOT | BT_LEAF;
2587
2588 p->header.nextindex = 0;
2589
2590 /* init freelist */
2591 fsi = 1;
2592 f = &p->slot[fsi];
2593
2594 /* init data area of root */
2595 for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2596 f->next = fsi;
2597 f->next = -1;
2598
2599 p->header.freelist = 1;
2600 p->header.freecnt = 8;
2601
2602 /* init '..' entry */
2603 p->header.idotdot = cpu_to_le32(idotdot);
2604
2605 return;
2606}
2607
2608/*
2609 * add_missing_indices()
2610 *
2611 * function: Fix dtree page in which one or more entries has an invalid index.
2612 * fsck.jfs should really fix this, but it currently does not.
2613 * Called from jfs_readdir when bad index is detected.
2614 */
2615static void add_missing_indices(struct inode *inode, s64 bn)
2616{
2617 struct ldtentry *d;
2618 struct dt_lock *dtlck;
2619 int i;
2620 uint index;
2621 struct lv *lv;
2622 struct metapage *mp;
2623 dtpage_t *p;
2624 int rc;
2625 s8 *stbl;
2626 tid_t tid;
2627 struct tlock *tlck;
2628
2629 tid = txBegin(inode->i_sb, 0);
2630
2631 DT_GETPAGE(inode, bn, mp, PSIZE, p, rc);
2632
2633 if (rc) {
2634 printk(KERN_ERR "DT_GETPAGE failed!\n");
2635 goto end;
2636 }
2637 BT_MARK_DIRTY(mp, inode);
2638
2639 ASSERT(p->header.flag & BT_LEAF);
2640
2641 tlck = txLock(tid, inode, mp, tlckDTREE | tlckENTRY);
2642 if (BT_IS_ROOT(mp))
2643 tlck->type |= tlckBTROOT;
2644
2645 dtlck = (struct dt_lock *) &tlck->lock;
2646
2647 stbl = DT_GETSTBL(p);
2648 for (i = 0; i < p->header.nextindex; i++) {
2649 d = (struct ldtentry *) &p->slot[stbl[i]];
2650 index = le32_to_cpu(d->index);
2651 if ((index < 2) || (index >= JFS_IP(inode)->next_index)) {
2652 d->index = cpu_to_le32(add_index(tid, inode, bn, i));
2653 if (dtlck->index >= dtlck->maxcnt)
2654 dtlck = (struct dt_lock *) txLinelock(dtlck);
2655 lv = &dtlck->lv[dtlck->index];
2656 lv->offset = stbl[i];
2657 lv->length = 1;
2658 dtlck->index++;
2659 }
2660 }
2661
2662 DT_PUTPAGE(mp);
2663 (void) txCommit(tid, 1, &inode, 0);
2664end:
2665 txEnd(tid);
2666}
2667
2668/*
2669 * Buffer to hold directory entry info while traversing a dtree page
2670 * before being fed to the filldir function
2671 */
2672struct jfs_dirent {
2673 loff_t position;
2674 int ino;
2675 u16 name_len;
2676 char name[];
2677};
2678
2679/*
2680 * function to determine next variable-sized jfs_dirent in buffer
2681 */
2682static inline struct jfs_dirent *next_jfs_dirent(struct jfs_dirent *dirent)
2683{
2684 return (struct jfs_dirent *)
2685 ((char *)dirent +
2686 ((sizeof (struct jfs_dirent) + dirent->name_len + 1 +
2687 sizeof (loff_t) - 1) &
2688 ~(sizeof (loff_t) - 1)));
2689}
2690
2691/*
2692 * jfs_readdir()
2693 *
2694 * function: read directory entries sequentially
2695 * from the specified entry offset
2696 *
2697 * parameter:
2698 *
2699 * return: offset = (pn, index) of start entry
2700 * of next jfs_readdir()/dtRead()
2701 */
2702int jfs_readdir(struct file *file, struct dir_context *ctx)
2703{
2704 struct inode *ip = file_inode(file);
2705 struct nls_table *codepage = JFS_SBI(ip->i_sb)->nls_tab;
2706 int rc = 0;
2707 loff_t dtpos; /* legacy OS/2 style position */
2708 struct dtoffset {
2709 s16 pn;
2710 s16 index;
2711 s32 unused;
2712 } *dtoffset = (struct dtoffset *) &dtpos;
2713 s64 bn;
2714 struct metapage *mp;
2715 dtpage_t *p;
2716 int index;
2717 s8 *stbl;
2718 struct btstack btstack;
2719 int i, next;
2720 struct ldtentry *d;
2721 struct dtslot *t;
2722 int d_namleft, len, outlen;
2723 unsigned long dirent_buf;
2724 char *name_ptr;
2725 u32 dir_index;
2726 int do_index = 0;
2727 uint loop_count = 0;
2728 struct jfs_dirent *jfs_dirent;
2729 int jfs_dirents;
2730 int overflow, fix_page, page_fixed = 0;
2731 static int unique_pos = 2; /* If we can't fix broken index */
2732
2733 if (ctx->pos == DIREND)
2734 return 0;
2735
2736 if (DO_INDEX(ip)) {
2737 /*
2738 * persistent index is stored in directory entries.
2739 * Special cases: 0 = .
2740 * 1 = ..
2741 * -1 = End of directory
2742 */
2743 do_index = 1;
2744
2745 dir_index = (u32) ctx->pos;
2746
2747 /*
2748 * NFSv4 reserves cookies 1 and 2 for . and .. so the value
2749 * we return to the vfs is one greater than the one we use
2750 * internally.
2751 */
2752 if (dir_index)
2753 dir_index--;
2754
2755 if (dir_index > 1) {
2756 struct dir_table_slot dirtab_slot;
2757
2758 if (dtEmpty(ip) ||
2759 (dir_index >= JFS_IP(ip)->next_index)) {
2760 /* Stale position. Directory has shrunk */
2761 ctx->pos = DIREND;
2762 return 0;
2763 }
2764 repeat:
2765 rc = read_index(ip, dir_index, &dirtab_slot);
2766 if (rc) {
2767 ctx->pos = DIREND;
2768 return rc;
2769 }
2770 if (dirtab_slot.flag == DIR_INDEX_FREE) {
2771 if (loop_count++ > JFS_IP(ip)->next_index) {
2772 jfs_err("jfs_readdir detected infinite loop!");
2773 ctx->pos = DIREND;
2774 return 0;
2775 }
2776 dir_index = le32_to_cpu(dirtab_slot.addr2);
2777 if (dir_index == -1) {
2778 ctx->pos = DIREND;
2779 return 0;
2780 }
2781 goto repeat;
2782 }
2783 bn = addressDTS(&dirtab_slot);
2784 index = dirtab_slot.slot;
2785 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2786 if (rc) {
2787 ctx->pos = DIREND;
2788 return 0;
2789 }
2790 if (p->header.flag & BT_INTERNAL) {
2791 jfs_err("jfs_readdir: bad index table");
2792 DT_PUTPAGE(mp);
2793 ctx->pos = DIREND;
2794 return 0;
2795 }
2796 } else {
2797 if (dir_index == 0) {
2798 /*
2799 * self "."
2800 */
2801 ctx->pos = 1;
2802 if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
2803 return 0;
2804 }
2805 /*
2806 * parent ".."
2807 */
2808 ctx->pos = 2;
2809 if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
2810 return 0;
2811
2812 /*
2813 * Find first entry of left-most leaf
2814 */
2815 if (dtEmpty(ip)) {
2816 ctx->pos = DIREND;
2817 return 0;
2818 }
2819
2820 if ((rc = dtReadFirst(ip, &btstack)))
2821 return rc;
2822
2823 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2824 }
2825 } else {
2826 /*
2827 * Legacy filesystem - OS/2 & Linux JFS < 0.3.6
2828 *
2829 * pn = 0; index = 1: First entry "."
2830 * pn = 0; index = 2: Second entry ".."
2831 * pn > 0: Real entries, pn=1 -> leftmost page
2832 * pn = index = -1: No more entries
2833 */
2834 dtpos = ctx->pos;
2835 if (dtpos < 2) {
2836 /* build "." entry */
2837 ctx->pos = 1;
2838 if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
2839 return 0;
2840 dtoffset->index = 2;
2841 ctx->pos = dtpos;
2842 }
2843
2844 if (dtoffset->pn == 0) {
2845 if (dtoffset->index == 2) {
2846 /* build ".." entry */
2847 if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
2848 return 0;
2849 } else {
2850 jfs_err("jfs_readdir called with invalid offset!");
2851 }
2852 dtoffset->pn = 1;
2853 dtoffset->index = 0;
2854 ctx->pos = dtpos;
2855 }
2856
2857 if (dtEmpty(ip)) {
2858 ctx->pos = DIREND;
2859 return 0;
2860 }
2861
2862 if ((rc = dtReadNext(ip, &ctx->pos, &btstack))) {
2863 jfs_err("jfs_readdir: unexpected rc = %d from dtReadNext",
2864 rc);
2865 ctx->pos = DIREND;
2866 return 0;
2867 }
2868 /* get start leaf page and index */
2869 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2870
2871 /* offset beyond directory eof ? */
2872 if (bn < 0) {
2873 ctx->pos = DIREND;
2874 return 0;
2875 }
2876 }
2877
2878 dirent_buf = __get_free_page(GFP_KERNEL);
2879 if (dirent_buf == 0) {
2880 DT_PUTPAGE(mp);
2881 jfs_warn("jfs_readdir: __get_free_page failed!");
2882 ctx->pos = DIREND;
2883 return -ENOMEM;
2884 }
2885
2886 while (1) {
2887 jfs_dirent = (struct jfs_dirent *) dirent_buf;
2888 jfs_dirents = 0;
2889 overflow = fix_page = 0;
2890
2891 stbl = DT_GETSTBL(p);
2892
2893 for (i = index; i < p->header.nextindex; i++) {
2894 if (stbl[i] < 0 || stbl[i] > 127) {
2895 jfs_err("JFS: Invalid stbl[%d] = %d for inode %ld, block = %lld",
2896 i, stbl[i], (long)ip->i_ino, (long long)bn);
2897 free_page(dirent_buf);
2898 DT_PUTPAGE(mp);
2899 return -EIO;
2900 }
2901
2902 d = (struct ldtentry *) & p->slot[stbl[i]];
2903
2904 if (((long) jfs_dirent + d->namlen + 1) >
2905 (dirent_buf + PAGE_SIZE)) {
2906 /* DBCS codepages could overrun dirent_buf */
2907 index = i;
2908 overflow = 1;
2909 break;
2910 }
2911
2912 d_namleft = d->namlen;
2913 name_ptr = jfs_dirent->name;
2914 jfs_dirent->ino = le32_to_cpu(d->inumber);
2915
2916 if (do_index) {
2917 len = min(d_namleft, DTLHDRDATALEN);
2918 jfs_dirent->position = le32_to_cpu(d->index);
2919 /*
2920 * d->index should always be valid, but it
2921 * isn't. fsck.jfs doesn't create the
2922 * directory index for the lost+found
2923 * directory. Rather than let it go,
2924 * we can try to fix it.
2925 */
2926 if ((jfs_dirent->position < 2) ||
2927 (jfs_dirent->position >=
2928 JFS_IP(ip)->next_index)) {
2929 if (!page_fixed && !isReadOnly(ip)) {
2930 fix_page = 1;
2931 /*
2932 * setting overflow and setting
2933 * index to i will cause the
2934 * same page to be processed
2935 * again starting here
2936 */
2937 overflow = 1;
2938 index = i;
2939 break;
2940 }
2941 jfs_dirent->position = unique_pos++;
2942 }
2943 /*
2944 * We add 1 to the index because we may
2945 * use a value of 2 internally, and NFSv4
2946 * doesn't like that.
2947 */
2948 jfs_dirent->position++;
2949 } else {
2950 jfs_dirent->position = dtpos;
2951 len = min(d_namleft, DTLHDRDATALEN_LEGACY);
2952 }
2953
2954 /* copy the name of head/only segment */
2955 outlen = jfs_strfromUCS_le(name_ptr, d->name, len,
2956 codepage);
2957 jfs_dirent->name_len = outlen;
2958
2959 /* copy name in the additional segment(s) */
2960 next = d->next;
2961 while (next >= 0) {
2962 t = (struct dtslot *) & p->slot[next];
2963 name_ptr += outlen;
2964 d_namleft -= len;
2965 /* Sanity Check */
2966 if (d_namleft == 0) {
2967 jfs_error(ip->i_sb,
2968 "JFS:Dtree error: ino = %ld, bn=%lld, index = %d\n",
2969 (long)ip->i_ino,
2970 (long long)bn,
2971 i);
2972 goto skip_one;
2973 }
2974 len = min(d_namleft, DTSLOTDATALEN);
2975 outlen = jfs_strfromUCS_le(name_ptr, t->name,
2976 len, codepage);
2977 jfs_dirent->name_len += outlen;
2978
2979 next = t->next;
2980 }
2981
2982 jfs_dirents++;
2983 jfs_dirent = next_jfs_dirent(jfs_dirent);
2984skip_one:
2985 if (!do_index)
2986 dtoffset->index++;
2987 }
2988
2989 if (!overflow) {
2990 /* Point to next leaf page */
2991 if (p->header.flag & BT_ROOT)
2992 bn = 0;
2993 else {
2994 bn = le64_to_cpu(p->header.next);
2995 index = 0;
2996 /* update offset (pn:index) for new page */
2997 if (!do_index) {
2998 dtoffset->pn++;
2999 dtoffset->index = 0;
3000 }
3001 }
3002 page_fixed = 0;
3003 }
3004
3005 /* unpin previous leaf page */
3006 DT_PUTPAGE(mp);
3007
3008 jfs_dirent = (struct jfs_dirent *) dirent_buf;
3009 while (jfs_dirents--) {
3010 ctx->pos = jfs_dirent->position;
3011 if (!dir_emit(ctx, jfs_dirent->name,
3012 jfs_dirent->name_len,
3013 jfs_dirent->ino, DT_UNKNOWN))
3014 goto out;
3015 jfs_dirent = next_jfs_dirent(jfs_dirent);
3016 }
3017
3018 if (fix_page) {
3019 add_missing_indices(ip, bn);
3020 page_fixed = 1;
3021 }
3022
3023 if (!overflow && (bn == 0)) {
3024 ctx->pos = DIREND;
3025 break;
3026 }
3027
3028 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3029 if (rc) {
3030 free_page(dirent_buf);
3031 return rc;
3032 }
3033 }
3034
3035 out:
3036 free_page(dirent_buf);
3037
3038 return rc;
3039}
3040
3041
3042/*
3043 * dtReadFirst()
3044 *
3045 * function: get the leftmost page of the directory
3046 */
3047static int dtReadFirst(struct inode *ip, struct btstack * btstack)
3048{
3049 int rc = 0;
3050 s64 bn;
3051 int psize = 288; /* initial in-line directory */
3052 struct metapage *mp;
3053 dtpage_t *p;
3054 s8 *stbl;
3055 struct btframe *btsp;
3056 pxd_t *xd;
3057
3058 BT_CLR(btstack); /* reset stack */
3059
3060 /*
3061 * descend leftmost path of the tree
3062 *
3063 * by convention, root bn = 0.
3064 */
3065 for (bn = 0;;) {
3066 DT_GETPAGE(ip, bn, mp, psize, p, rc);
3067 if (rc)
3068 return rc;
3069
3070 /*
3071 * leftmost leaf page
3072 */
3073 if (p->header.flag & BT_LEAF) {
3074 /* return leftmost entry */
3075 btsp = btstack->top;
3076 btsp->bn = bn;
3077 btsp->index = 0;
3078 btsp->mp = mp;
3079
3080 return 0;
3081 }
3082
3083 /*
3084 * descend down to leftmost child page
3085 */
3086 if (BT_STACK_FULL(btstack)) {
3087 DT_PUTPAGE(mp);
3088 jfs_error(ip->i_sb, "btstack overrun\n");
3089 BT_STACK_DUMP(btstack);
3090 return -EIO;
3091 }
3092 /* push (bn, index) of the parent page/entry */
3093 BT_PUSH(btstack, bn, 0);
3094
3095 /* get the leftmost entry */
3096 stbl = DT_GETSTBL(p);
3097
3098 if (stbl[0] < 0 || stbl[0] > 127) {
3099 DT_PUTPAGE(mp);
3100 jfs_error(ip->i_sb, "stbl[0] out of bound\n");
3101 return -EIO;
3102 }
3103
3104 xd = (pxd_t *) & p->slot[stbl[0]];
3105
3106 /* get the child page block address */
3107 bn = addressPXD(xd);
3108 psize = lengthPXD(xd) << JFS_SBI(ip->i_sb)->l2bsize;
3109
3110 /* unpin the parent page */
3111 DT_PUTPAGE(mp);
3112 }
3113}
3114
3115
3116/*
3117 * dtReadNext()
3118 *
3119 * function: get the page of the specified offset (pn:index)
3120 *
3121 * return: if (offset > eof), bn = -1;
3122 *
3123 * note: if index > nextindex of the target leaf page,
3124 * start with 1st entry of next leaf page;
3125 */
3126static int dtReadNext(struct inode *ip, loff_t * offset,
3127 struct btstack * btstack)
3128{
3129 int rc = 0;
3130 struct dtoffset {
3131 s16 pn;
3132 s16 index;
3133 s32 unused;
3134 } *dtoffset = (struct dtoffset *) offset;
3135 s64 bn;
3136 struct metapage *mp;
3137 dtpage_t *p;
3138 int index;
3139 int pn;
3140 s8 *stbl;
3141 struct btframe *btsp, *parent;
3142 pxd_t *xd;
3143
3144 /*
3145 * get leftmost leaf page pinned
3146 */
3147 if ((rc = dtReadFirst(ip, btstack)))
3148 return rc;
3149
3150 /* get leaf page */
3151 DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
3152
3153 /* get the start offset (pn:index) */
3154 pn = dtoffset->pn - 1; /* Now pn = 0 represents leftmost leaf */
3155 index = dtoffset->index;
3156
3157 /* start at leftmost page ? */
3158 if (pn == 0) {
3159 /* offset beyond eof ? */
3160 if (index < p->header.nextindex)
3161 goto out;
3162
3163 if (p->header.flag & BT_ROOT) {
3164 bn = -1;
3165 goto out;
3166 }
3167
3168 /* start with 1st entry of next leaf page */
3169 dtoffset->pn++;
3170 dtoffset->index = index = 0;
3171 goto a;
3172 }
3173
3174 /* start at non-leftmost page: scan parent pages for large pn */
3175 if (p->header.flag & BT_ROOT) {
3176 bn = -1;
3177 goto out;
3178 }
3179
3180 /* start after next leaf page ? */
3181 if (pn > 1)
3182 goto b;
3183
3184 /* get leaf page pn = 1 */
3185 a:
3186 bn = le64_to_cpu(p->header.next);
3187
3188 /* unpin leaf page */
3189 DT_PUTPAGE(mp);
3190
3191 /* offset beyond eof ? */
3192 if (bn == 0) {
3193 bn = -1;
3194 goto out;
3195 }
3196
3197 goto c;
3198
3199 /*
3200 * scan last internal page level to get target leaf page
3201 */
3202 b:
3203 /* unpin leftmost leaf page */
3204 DT_PUTPAGE(mp);
3205
3206 /* get left most parent page */
3207 btsp = btstack->top;
3208 parent = btsp - 1;
3209 bn = parent->bn;
3210 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3211 if (rc)
3212 return rc;
3213
3214 /* scan parent pages at last internal page level */
3215 while (pn >= p->header.nextindex) {
3216 pn -= p->header.nextindex;
3217
3218 /* get next parent page address */
3219 bn = le64_to_cpu(p->header.next);
3220
3221 /* unpin current parent page */
3222 DT_PUTPAGE(mp);
3223
3224 /* offset beyond eof ? */
3225 if (bn == 0) {
3226 bn = -1;
3227 goto out;
3228 }
3229
3230 /* get next parent page */
3231 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3232 if (rc)
3233 return rc;
3234
3235 /* update parent page stack frame */
3236 parent->bn = bn;
3237 }
3238
3239 /* get leaf page address */
3240 stbl = DT_GETSTBL(p);
3241 xd = (pxd_t *) & p->slot[stbl[pn]];
3242 bn = addressPXD(xd);
3243
3244 /* unpin parent page */
3245 DT_PUTPAGE(mp);
3246
3247 /*
3248 * get target leaf page
3249 */
3250 c:
3251 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3252 if (rc)
3253 return rc;
3254
3255 /*
3256 * leaf page has been completed:
3257 * start with 1st entry of next leaf page
3258 */
3259 if (index >= p->header.nextindex) {
3260 bn = le64_to_cpu(p->header.next);
3261
3262 /* unpin leaf page */
3263 DT_PUTPAGE(mp);
3264
3265 /* offset beyond eof ? */
3266 if (bn == 0) {
3267 bn = -1;
3268 goto out;
3269 }
3270
3271 /* get next leaf page */
3272 DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3273 if (rc)
3274 return rc;
3275
3276 /* start with 1st entry of next leaf page */
3277 dtoffset->pn++;
3278 dtoffset->index = 0;
3279 }
3280
3281 out:
3282 /* return target leaf page pinned */
3283 btsp = btstack->top;
3284 btsp->bn = bn;
3285 btsp->index = dtoffset->index;
3286 btsp->mp = mp;
3287
3288 return 0;
3289}
3290
3291
3292/*
3293 * dtCompare()
3294 *
3295 * function: compare search key with an internal entry
3296 *
3297 * return:
3298 * < 0 if k is < record
3299 * = 0 if k is = record
3300 * > 0 if k is > record
3301 */
3302static int dtCompare(struct component_name * key, /* search key */
3303 dtpage_t * p, /* directory page */
3304 int si)
3305{ /* entry slot index */
3306 wchar_t *kname;
3307 __le16 *name;
3308 int klen, namlen, len, rc;
3309 struct idtentry *ih;
3310 struct dtslot *t;
3311
3312 /*
3313 * force the left-most key on internal pages, at any level of
3314 * the tree, to be less than any search key.
3315 * this obviates having to update the leftmost key on an internal
3316 * page when the user inserts a new key in the tree smaller than
3317 * anything that has been stored.
3318 *
3319 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3320 * at any internal page at any level of the tree,
3321 * it descends to child of the entry anyway -
3322 * ? make the entry as min size dummy entry)
3323 *
3324 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3325 * return (1);
3326 */
3327
3328 kname = key->name;
3329 klen = key->namlen;
3330
3331 ih = (struct idtentry *) & p->slot[si];
3332 si = ih->next;
3333 name = ih->name;
3334 namlen = ih->namlen;
3335 len = min(namlen, DTIHDRDATALEN);
3336
3337 /* compare with head/only segment */
3338 len = min(klen, len);
3339 if ((rc = UniStrncmp_le(kname, name, len)))
3340 return rc;
3341
3342 klen -= len;
3343 namlen -= len;
3344
3345 /* compare with additional segment(s) */
3346 kname += len;
3347 while (klen > 0 && namlen > 0) {
3348 /* compare with next name segment */
3349 t = (struct dtslot *) & p->slot[si];
3350 len = min(namlen, DTSLOTDATALEN);
3351 len = min(klen, len);
3352 name = t->name;
3353 if ((rc = UniStrncmp_le(kname, name, len)))
3354 return rc;
3355
3356 klen -= len;
3357 namlen -= len;
3358 kname += len;
3359 si = t->next;
3360 }
3361
3362 return (klen - namlen);
3363}
3364
3365
3366
3367
3368/*
3369 * ciCompare()
3370 *
3371 * function: compare search key with an (leaf/internal) entry
3372 *
3373 * return:
3374 * < 0 if k is < record
3375 * = 0 if k is = record
3376 * > 0 if k is > record
3377 */
3378static int ciCompare(struct component_name * key, /* search key */
3379 dtpage_t * p, /* directory page */
3380 int si, /* entry slot index */
3381 int flag)
3382{
3383 wchar_t *kname, x;
3384 __le16 *name;
3385 int klen, namlen, len, rc;
3386 struct ldtentry *lh;
3387 struct idtentry *ih;
3388 struct dtslot *t;
3389 int i;
3390
3391 /*
3392 * force the left-most key on internal pages, at any level of
3393 * the tree, to be less than any search key.
3394 * this obviates having to update the leftmost key on an internal
3395 * page when the user inserts a new key in the tree smaller than
3396 * anything that has been stored.
3397 *
3398 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3399 * at any internal page at any level of the tree,
3400 * it descends to child of the entry anyway -
3401 * ? make the entry as min size dummy entry)
3402 *
3403 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3404 * return (1);
3405 */
3406
3407 kname = key->name;
3408 klen = key->namlen;
3409
3410 /*
3411 * leaf page entry
3412 */
3413 if (p->header.flag & BT_LEAF) {
3414 lh = (struct ldtentry *) & p->slot[si];
3415 si = lh->next;
3416 name = lh->name;
3417 namlen = lh->namlen;
3418 if (flag & JFS_DIR_INDEX)
3419 len = min(namlen, DTLHDRDATALEN);
3420 else
3421 len = min(namlen, DTLHDRDATALEN_LEGACY);
3422 }
3423 /*
3424 * internal page entry
3425 */
3426 else {
3427 ih = (struct idtentry *) & p->slot[si];
3428 si = ih->next;
3429 name = ih->name;
3430 namlen = ih->namlen;
3431 len = min(namlen, DTIHDRDATALEN);
3432 }
3433
3434 /* compare with head/only segment */
3435 len = min(klen, len);
3436 for (i = 0; i < len; i++, kname++, name++) {
3437 /* only uppercase if case-insensitive support is on */
3438 if ((flag & JFS_OS2) == JFS_OS2)
3439 x = UniToupper(le16_to_cpu(*name));
3440 else
3441 x = le16_to_cpu(*name);
3442 if ((rc = *kname - x))
3443 return rc;
3444 }
3445
3446 klen -= len;
3447 namlen -= len;
3448
3449 /* compare with additional segment(s) */
3450 while (klen > 0 && namlen > 0) {
3451 /* compare with next name segment */
3452 t = (struct dtslot *) & p->slot[si];
3453 len = min(namlen, DTSLOTDATALEN);
3454 len = min(klen, len);
3455 name = t->name;
3456 for (i = 0; i < len; i++, kname++, name++) {
3457 /* only uppercase if case-insensitive support is on */
3458 if ((flag & JFS_OS2) == JFS_OS2)
3459 x = UniToupper(le16_to_cpu(*name));
3460 else
3461 x = le16_to_cpu(*name);
3462
3463 if ((rc = *kname - x))
3464 return rc;
3465 }
3466
3467 klen -= len;
3468 namlen -= len;
3469 si = t->next;
3470 }
3471
3472 return (klen - namlen);
3473}
3474
3475
3476/*
3477 * ciGetLeafPrefixKey()
3478 *
3479 * function: compute prefix of suffix compression
3480 * from two adjacent leaf entries
3481 * across page boundary
3482 *
3483 * return: non-zero on error
3484 *
3485 */
3486static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
3487 int ri, struct component_name * key, int flag)
3488{
3489 int klen, namlen;
3490 wchar_t *pl, *pr, *kname;
3491 struct component_name lkey;
3492 struct component_name rkey;
3493
3494 lkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3495 GFP_KERNEL);
3496 if (lkey.name == NULL)
3497 return -ENOMEM;
3498
3499 rkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3500 GFP_KERNEL);
3501 if (rkey.name == NULL) {
3502 kfree(lkey.name);
3503 return -ENOMEM;
3504 }
3505
3506 /* get left and right key */
3507 dtGetKey(lp, li, &lkey, flag);
3508 lkey.name[lkey.namlen] = 0;
3509
3510 if ((flag & JFS_OS2) == JFS_OS2)
3511 ciToUpper(&lkey);
3512
3513 dtGetKey(rp, ri, &rkey, flag);
3514 rkey.name[rkey.namlen] = 0;
3515
3516
3517 if ((flag & JFS_OS2) == JFS_OS2)
3518 ciToUpper(&rkey);
3519
3520 /* compute prefix */
3521 klen = 0;
3522 kname = key->name;
3523 namlen = min(lkey.namlen, rkey.namlen);
3524 for (pl = lkey.name, pr = rkey.name;
3525 namlen; pl++, pr++, namlen--, klen++, kname++) {
3526 *kname = *pr;
3527 if (*pl != *pr) {
3528 key->namlen = klen + 1;
3529 goto free_names;
3530 }
3531 }
3532
3533 /* l->namlen <= r->namlen since l <= r */
3534 if (lkey.namlen < rkey.namlen) {
3535 *kname = *pr;
3536 key->namlen = klen + 1;
3537 } else /* l->namelen == r->namelen */
3538 key->namlen = klen;
3539
3540free_names:
3541 kfree(lkey.name);
3542 kfree(rkey.name);
3543 return 0;
3544}
3545
3546
3547
3548/*
3549 * dtGetKey()
3550 *
3551 * function: get key of the entry
3552 */
3553static void dtGetKey(dtpage_t * p, int i, /* entry index */
3554 struct component_name * key, int flag)
3555{
3556 int si;
3557 s8 *stbl;
3558 struct ldtentry *lh;
3559 struct idtentry *ih;
3560 struct dtslot *t;
3561 int namlen, len;
3562 wchar_t *kname;
3563 __le16 *name;
3564
3565 /* get entry */
3566 stbl = DT_GETSTBL(p);
3567 si = stbl[i];
3568 if (p->header.flag & BT_LEAF) {
3569 lh = (struct ldtentry *) & p->slot[si];
3570 si = lh->next;
3571 namlen = lh->namlen;
3572 name = lh->name;
3573 if (flag & JFS_DIR_INDEX)
3574 len = min(namlen, DTLHDRDATALEN);
3575 else
3576 len = min(namlen, DTLHDRDATALEN_LEGACY);
3577 } else {
3578 ih = (struct idtentry *) & p->slot[si];
3579 si = ih->next;
3580 namlen = ih->namlen;
3581 name = ih->name;
3582 len = min(namlen, DTIHDRDATALEN);
3583 }
3584
3585 key->namlen = namlen;
3586 kname = key->name;
3587
3588 /*
3589 * move head/only segment
3590 */
3591 UniStrncpy_from_le(kname, name, len);
3592
3593 /*
3594 * move additional segment(s)
3595 */
3596 while (si >= 0) {
3597 /* get next segment */
3598 t = &p->slot[si];
3599 kname += len;
3600 namlen -= len;
3601 len = min(namlen, DTSLOTDATALEN);
3602 UniStrncpy_from_le(kname, t->name, len);
3603
3604 si = t->next;
3605 }
3606}
3607
3608
3609/*
3610 * dtInsertEntry()
3611 *
3612 * function: allocate free slot(s) and
3613 * write a leaf/internal entry
3614 *
3615 * return: entry slot index
3616 */
3617static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
3618 ddata_t * data, struct dt_lock ** dtlock)
3619{
3620 struct dtslot *h, *t;
3621 struct ldtentry *lh = NULL;
3622 struct idtentry *ih = NULL;
3623 int hsi, fsi, klen, len, nextindex;
3624 wchar_t *kname;
3625 __le16 *name;
3626 s8 *stbl;
3627 pxd_t *xd;
3628 struct dt_lock *dtlck = *dtlock;
3629 struct lv *lv;
3630 int xsi, n;
3631 s64 bn = 0;
3632 struct metapage *mp = NULL;
3633
3634 klen = key->namlen;
3635 kname = key->name;
3636
3637 /* allocate a free slot */
3638 hsi = fsi = p->header.freelist;
3639 h = &p->slot[fsi];
3640 p->header.freelist = h->next;
3641 --p->header.freecnt;
3642
3643 /* open new linelock */
3644 if (dtlck->index >= dtlck->maxcnt)
3645 dtlck = (struct dt_lock *) txLinelock(dtlck);
3646
3647 lv = & dtlck->lv[dtlck->index];
3648 lv->offset = hsi;
3649
3650 /* write head/only segment */
3651 if (p->header.flag & BT_LEAF) {
3652 lh = (struct ldtentry *) h;
3653 lh->next = h->next;
3654 lh->inumber = cpu_to_le32(data->leaf.ino);
3655 lh->namlen = klen;
3656 name = lh->name;
3657 if (data->leaf.ip) {
3658 len = min(klen, DTLHDRDATALEN);
3659 if (!(p->header.flag & BT_ROOT))
3660 bn = addressPXD(&p->header.self);
3661 lh->index = cpu_to_le32(add_index(data->leaf.tid,
3662 data->leaf.ip,
3663 bn, index));
3664 } else
3665 len = min(klen, DTLHDRDATALEN_LEGACY);
3666 } else {
3667 ih = (struct idtentry *) h;
3668 ih->next = h->next;
3669 xd = (pxd_t *) ih;
3670 *xd = data->xd;
3671 ih->namlen = klen;
3672 name = ih->name;
3673 len = min(klen, DTIHDRDATALEN);
3674 }
3675
3676 UniStrncpy_to_le(name, kname, len);
3677
3678 n = 1;
3679 xsi = hsi;
3680
3681 /* write additional segment(s) */
3682 t = h;
3683 klen -= len;
3684 while (klen) {
3685 /* get free slot */
3686 fsi = p->header.freelist;
3687 t = &p->slot[fsi];
3688 p->header.freelist = t->next;
3689 --p->header.freecnt;
3690
3691 /* is next slot contiguous ? */
3692 if (fsi != xsi + 1) {
3693 /* close current linelock */
3694 lv->length = n;
3695 dtlck->index++;
3696
3697 /* open new linelock */
3698 if (dtlck->index < dtlck->maxcnt)
3699 lv++;
3700 else {
3701 dtlck = (struct dt_lock *) txLinelock(dtlck);
3702 lv = & dtlck->lv[0];
3703 }
3704
3705 lv->offset = fsi;
3706 n = 0;
3707 }
3708
3709 kname += len;
3710 len = min(klen, DTSLOTDATALEN);
3711 UniStrncpy_to_le(t->name, kname, len);
3712
3713 n++;
3714 xsi = fsi;
3715 klen -= len;
3716 }
3717
3718 /* close current linelock */
3719 lv->length = n;
3720 dtlck->index++;
3721
3722 *dtlock = dtlck;
3723
3724 /* terminate last/only segment */
3725 if (h == t) {
3726 /* single segment entry */
3727 if (p->header.flag & BT_LEAF)
3728 lh->next = -1;
3729 else
3730 ih->next = -1;
3731 } else
3732 /* multi-segment entry */
3733 t->next = -1;
3734
3735 /* if insert into middle, shift right succeeding entries in stbl */
3736 stbl = DT_GETSTBL(p);
3737 nextindex = p->header.nextindex;
3738 if (index < nextindex) {
3739 memmove(stbl + index + 1, stbl + index, nextindex - index);
3740
3741 if ((p->header.flag & BT_LEAF) && data->leaf.ip) {
3742 s64 lblock;
3743
3744 /*
3745 * Need to update slot number for entries that moved
3746 * in the stbl
3747 */
3748 mp = NULL;
3749 for (n = index + 1; n <= nextindex; n++) {
3750 lh = (struct ldtentry *) & (p->slot[stbl[n]]);
3751 modify_index(data->leaf.tid, data->leaf.ip,
3752 le32_to_cpu(lh->index), bn, n,
3753 &mp, &lblock);
3754 }
3755 if (mp)
3756 release_metapage(mp);
3757 }
3758 }
3759
3760 stbl[index] = hsi;
3761
3762 /* advance next available entry index of stbl */
3763 ++p->header.nextindex;
3764}
3765
3766
3767/*
3768 * dtMoveEntry()
3769 *
3770 * function: move entries from split/left page to new/right page
3771 *
3772 * nextindex of dst page and freelist/freecnt of both pages
3773 * are updated.
3774 */
3775static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
3776 struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
3777 int do_index)
3778{
3779 int ssi, next; /* src slot index */
3780 int di; /* dst entry index */
3781 int dsi; /* dst slot index */
3782 s8 *sstbl, *dstbl; /* sorted entry table */
3783 int snamlen, len;
3784 struct ldtentry *slh, *dlh = NULL;
3785 struct idtentry *sih, *dih = NULL;
3786 struct dtslot *h, *s, *d;
3787 struct dt_lock *sdtlck = *sdtlock, *ddtlck = *ddtlock;
3788 struct lv *slv, *dlv;
3789 int xssi, ns, nd;
3790 int sfsi;
3791
3792 sstbl = (s8 *) & sp->slot[sp->header.stblindex];
3793 dstbl = (s8 *) & dp->slot[dp->header.stblindex];
3794
3795 dsi = dp->header.freelist; /* first (whole page) free slot */
3796 sfsi = sp->header.freelist;
3797
3798 /* linelock destination entry slot */
3799 dlv = & ddtlck->lv[ddtlck->index];
3800 dlv->offset = dsi;
3801
3802 /* linelock source entry slot */
3803 slv = & sdtlck->lv[sdtlck->index];
3804 slv->offset = sstbl[si];
3805 xssi = slv->offset - 1;
3806
3807 /*
3808 * move entries
3809 */
3810 ns = nd = 0;
3811 for (di = 0; si < sp->header.nextindex; si++, di++) {
3812 ssi = sstbl[si];
3813 dstbl[di] = dsi;
3814
3815 /* is next slot contiguous ? */
3816 if (ssi != xssi + 1) {
3817 /* close current linelock */
3818 slv->length = ns;
3819 sdtlck->index++;
3820
3821 /* open new linelock */
3822 if (sdtlck->index < sdtlck->maxcnt)
3823 slv++;
3824 else {
3825 sdtlck = (struct dt_lock *) txLinelock(sdtlck);
3826 slv = & sdtlck->lv[0];
3827 }
3828
3829 slv->offset = ssi;
3830 ns = 0;
3831 }
3832
3833 /*
3834 * move head/only segment of an entry
3835 */
3836 /* get dst slot */
3837 h = d = &dp->slot[dsi];
3838
3839 /* get src slot and move */
3840 s = &sp->slot[ssi];
3841 if (sp->header.flag & BT_LEAF) {
3842 /* get source entry */
3843 slh = (struct ldtentry *) s;
3844 dlh = (struct ldtentry *) h;
3845 snamlen = slh->namlen;
3846
3847 if (do_index) {
3848 len = min(snamlen, DTLHDRDATALEN);
3849 dlh->index = slh->index; /* little-endian */
3850 } else
3851 len = min(snamlen, DTLHDRDATALEN_LEGACY);
3852
3853 memcpy(dlh, slh, 6 + len * 2);
3854
3855 next = slh->next;
3856
3857 /* update dst head/only segment next field */
3858 dsi++;
3859 dlh->next = dsi;
3860 } else {
3861 sih = (struct idtentry *) s;
3862 snamlen = sih->namlen;
3863
3864 len = min(snamlen, DTIHDRDATALEN);
3865 dih = (struct idtentry *) h;
3866 memcpy(dih, sih, 10 + len * 2);
3867 next = sih->next;
3868
3869 dsi++;
3870 dih->next = dsi;
3871 }
3872
3873 /* free src head/only segment */
3874 s->next = sfsi;
3875 s->cnt = 1;
3876 sfsi = ssi;
3877
3878 ns++;
3879 nd++;
3880 xssi = ssi;
3881
3882 /*
3883 * move additional segment(s) of the entry
3884 */
3885 snamlen -= len;
3886 while ((ssi = next) >= 0) {
3887 /* is next slot contiguous ? */
3888 if (ssi != xssi + 1) {
3889 /* close current linelock */
3890 slv->length = ns;
3891 sdtlck->index++;
3892
3893 /* open new linelock */
3894 if (sdtlck->index < sdtlck->maxcnt)
3895 slv++;
3896 else {
3897 sdtlck =
3898 (struct dt_lock *)
3899 txLinelock(sdtlck);
3900 slv = & sdtlck->lv[0];
3901 }
3902
3903 slv->offset = ssi;
3904 ns = 0;
3905 }
3906
3907 /* get next source segment */
3908 s = &sp->slot[ssi];
3909
3910 /* get next destination free slot */
3911 d++;
3912
3913 len = min(snamlen, DTSLOTDATALEN);
3914 UniStrncpy_le(d->name, s->name, len);
3915
3916 ns++;
3917 nd++;
3918 xssi = ssi;
3919
3920 dsi++;
3921 d->next = dsi;
3922
3923 /* free source segment */
3924 next = s->next;
3925 s->next = sfsi;
3926 s->cnt = 1;
3927 sfsi = ssi;
3928
3929 snamlen -= len;
3930 } /* end while */
3931
3932 /* terminate dst last/only segment */
3933 if (h == d) {
3934 /* single segment entry */
3935 if (dp->header.flag & BT_LEAF)
3936 dlh->next = -1;
3937 else
3938 dih->next = -1;
3939 } else
3940 /* multi-segment entry */
3941 d->next = -1;
3942 } /* end for */
3943
3944 /* close current linelock */
3945 slv->length = ns;
3946 sdtlck->index++;
3947 *sdtlock = sdtlck;
3948
3949 dlv->length = nd;
3950 ddtlck->index++;
3951 *ddtlock = ddtlck;
3952
3953 /* update source header */
3954 sp->header.freelist = sfsi;
3955 sp->header.freecnt += nd;
3956
3957 /* update destination header */
3958 dp->header.nextindex = di;
3959
3960 dp->header.freelist = dsi;
3961 dp->header.freecnt -= nd;
3962}
3963
3964
3965/*
3966 * dtDeleteEntry()
3967 *
3968 * function: free a (leaf/internal) entry
3969 *
3970 * log freelist header, stbl, and each segment slot of entry
3971 * (even though last/only segment next field is modified,
3972 * physical image logging requires all segment slots of
3973 * the entry logged to avoid applying previous updates
3974 * to the same slots)
3975 */
3976static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock)
3977{
3978 int fsi; /* free entry slot index */
3979 s8 *stbl;
3980 struct dtslot *t;
3981 int si, freecnt;
3982 struct dt_lock *dtlck = *dtlock;
3983 struct lv *lv;
3984 int xsi, n;
3985
3986 /* get free entry slot index */
3987 stbl = DT_GETSTBL(p);
3988 fsi = stbl[fi];
3989
3990 /* open new linelock */
3991 if (dtlck->index >= dtlck->maxcnt)
3992 dtlck = (struct dt_lock *) txLinelock(dtlck);
3993 lv = & dtlck->lv[dtlck->index];
3994
3995 lv->offset = fsi;
3996
3997 /* get the head/only segment */
3998 t = &p->slot[fsi];
3999 if (p->header.flag & BT_LEAF)
4000 si = ((struct ldtentry *) t)->next;
4001 else
4002 si = ((struct idtentry *) t)->next;
4003 t->next = si;
4004 t->cnt = 1;
4005
4006 n = freecnt = 1;
4007 xsi = fsi;
4008
4009 /* find the last/only segment */
4010 while (si >= 0) {
4011 /* is next slot contiguous ? */
4012 if (si != xsi + 1) {
4013 /* close current linelock */
4014 lv->length = n;
4015 dtlck->index++;
4016
4017 /* open new linelock */
4018 if (dtlck->index < dtlck->maxcnt)
4019 lv++;
4020 else {
4021 dtlck = (struct dt_lock *) txLinelock(dtlck);
4022 lv = & dtlck->lv[0];
4023 }
4024
4025 lv->offset = si;
4026 n = 0;
4027 }
4028
4029 n++;
4030 xsi = si;
4031 freecnt++;
4032
4033 t = &p->slot[si];
4034 t->cnt = 1;
4035 si = t->next;
4036 }
4037
4038 /* close current linelock */
4039 lv->length = n;
4040 dtlck->index++;
4041
4042 *dtlock = dtlck;
4043
4044 /* update freelist */
4045 t->next = p->header.freelist;
4046 p->header.freelist = fsi;
4047 p->header.freecnt += freecnt;
4048
4049 /* if delete from middle,
4050 * shift left the succedding entries in the stbl
4051 */
4052 si = p->header.nextindex;
4053 if (fi < si - 1)
4054 memmove(&stbl[fi], &stbl[fi + 1], si - fi - 1);
4055
4056 p->header.nextindex--;
4057}
4058
4059
4060/*
4061 * dtTruncateEntry()
4062 *
4063 * function: truncate a (leaf/internal) entry
4064 *
4065 * log freelist header, stbl, and each segment slot of entry
4066 * (even though last/only segment next field is modified,
4067 * physical image logging requires all segment slots of
4068 * the entry logged to avoid applying previous updates
4069 * to the same slots)
4070 */
4071static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock)
4072{
4073 int tsi; /* truncate entry slot index */
4074 s8 *stbl;
4075 struct dtslot *t;
4076 int si, freecnt;
4077 struct dt_lock *dtlck = *dtlock;
4078 struct lv *lv;
4079 int fsi, xsi, n;
4080
4081 /* get free entry slot index */
4082 stbl = DT_GETSTBL(p);
4083 tsi = stbl[ti];
4084
4085 /* open new linelock */
4086 if (dtlck->index >= dtlck->maxcnt)
4087 dtlck = (struct dt_lock *) txLinelock(dtlck);
4088 lv = & dtlck->lv[dtlck->index];
4089
4090 lv->offset = tsi;
4091
4092 /* get the head/only segment */
4093 t = &p->slot[tsi];
4094 ASSERT(p->header.flag & BT_INTERNAL);
4095 ((struct idtentry *) t)->namlen = 0;
4096 si = ((struct idtentry *) t)->next;
4097 ((struct idtentry *) t)->next = -1;
4098
4099 n = 1;
4100 freecnt = 0;
4101 fsi = si;
4102 xsi = tsi;
4103
4104 /* find the last/only segment */
4105 while (si >= 0) {
4106 /* is next slot contiguous ? */
4107 if (si != xsi + 1) {
4108 /* close current linelock */
4109 lv->length = n;
4110 dtlck->index++;
4111
4112 /* open new linelock */
4113 if (dtlck->index < dtlck->maxcnt)
4114 lv++;
4115 else {
4116 dtlck = (struct dt_lock *) txLinelock(dtlck);
4117 lv = & dtlck->lv[0];
4118 }
4119
4120 lv->offset = si;
4121 n = 0;
4122 }
4123
4124 n++;
4125 xsi = si;
4126 freecnt++;
4127
4128 t = &p->slot[si];
4129 t->cnt = 1;
4130 si = t->next;
4131 }
4132
4133 /* close current linelock */
4134 lv->length = n;
4135 dtlck->index++;
4136
4137 *dtlock = dtlck;
4138
4139 /* update freelist */
4140 if (freecnt == 0)
4141 return;
4142 t->next = p->header.freelist;
4143 p->header.freelist = fsi;
4144 p->header.freecnt += freecnt;
4145}
4146
4147
4148/*
4149 * dtLinelockFreelist()
4150 */
4151static void dtLinelockFreelist(dtpage_t * p, /* directory page */
4152 int m, /* max slot index */
4153 struct dt_lock ** dtlock)
4154{
4155 int fsi; /* free entry slot index */
4156 struct dtslot *t;
4157 int si;
4158 struct dt_lock *dtlck = *dtlock;
4159 struct lv *lv;
4160 int xsi, n;
4161
4162 /* get free entry slot index */
4163 fsi = p->header.freelist;
4164
4165 /* open new linelock */
4166 if (dtlck->index >= dtlck->maxcnt)
4167 dtlck = (struct dt_lock *) txLinelock(dtlck);
4168 lv = & dtlck->lv[dtlck->index];
4169
4170 lv->offset = fsi;
4171
4172 n = 1;
4173 xsi = fsi;
4174
4175 t = &p->slot[fsi];
4176 si = t->next;
4177
4178 /* find the last/only segment */
4179 while (si < m && si >= 0) {
4180 /* is next slot contiguous ? */
4181 if (si != xsi + 1) {
4182 /* close current linelock */
4183 lv->length = n;
4184 dtlck->index++;
4185
4186 /* open new linelock */
4187 if (dtlck->index < dtlck->maxcnt)
4188 lv++;
4189 else {
4190 dtlck = (struct dt_lock *) txLinelock(dtlck);
4191 lv = & dtlck->lv[0];
4192 }
4193
4194 lv->offset = si;
4195 n = 0;
4196 }
4197
4198 n++;
4199 xsi = si;
4200
4201 t = &p->slot[si];
4202 si = t->next;
4203 }
4204
4205 /* close current linelock */
4206 lv->length = n;
4207 dtlck->index++;
4208
4209 *dtlock = dtlck;
4210}
4211
4212
4213/*
4214 * NAME: dtModify
4215 *
4216 * FUNCTION: Modify the inode number part of a directory entry
4217 *
4218 * PARAMETERS:
4219 * tid - Transaction id
4220 * ip - Inode of parent directory
4221 * key - Name of entry to be modified
4222 * orig_ino - Original inode number expected in entry
4223 * new_ino - New inode number to put into entry
4224 * flag - JFS_RENAME
4225 *
4226 * RETURNS:
4227 * -ESTALE - If entry found does not match orig_ino passed in
4228 * -ENOENT - If no entry can be found to match key
4229 * 0 - If successfully modified entry
4230 */
4231int dtModify(tid_t tid, struct inode *ip,
4232 struct component_name * key, ino_t * orig_ino, ino_t new_ino, int flag)
4233{
4234 int rc;
4235 s64 bn;
4236 struct metapage *mp;
4237 dtpage_t *p;
4238 int index;
4239 struct btstack btstack;
4240 struct tlock *tlck;
4241 struct dt_lock *dtlck;
4242 struct lv *lv;
4243 s8 *stbl;
4244 int entry_si; /* entry slot index */
4245 struct ldtentry *entry;
4246
4247 /*
4248 * search for the entry to modify:
4249 *
4250 * dtSearch() returns (leaf page pinned, index at which to modify).
4251 */
4252 if ((rc = dtSearch(ip, key, orig_ino, &btstack, flag)))
4253 return rc;
4254
4255 /* retrieve search result */
4256 DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
4257
4258 BT_MARK_DIRTY(mp, ip);
4259 /*
4260 * acquire a transaction lock on the leaf page of named entry
4261 */
4262 tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
4263 dtlck = (struct dt_lock *) & tlck->lock;
4264
4265 /* get slot index of the entry */
4266 stbl = DT_GETSTBL(p);
4267 entry_si = stbl[index];
4268
4269 /* linelock entry */
4270 ASSERT(dtlck->index == 0);
4271 lv = & dtlck->lv[0];
4272 lv->offset = entry_si;
4273 lv->length = 1;
4274 dtlck->index++;
4275
4276 /* get the head/only segment */
4277 entry = (struct ldtentry *) & p->slot[entry_si];
4278
4279 /* substitute the inode number of the entry */
4280 entry->inumber = cpu_to_le32(new_ino);
4281
4282 /* unpin the leaf page */
4283 DT_PUTPAGE(mp);
4284
4285 return 0;
4286}