xref: /openbmc/linux/fs/jfs/jfs_dtree.c (revision d0e22329)
1 /*
2  *   Copyright (C) International Business Machines Corp., 2000-2004
3  *
4  *   This program is free software;  you can redistribute it and/or modify
5  *   it under the terms of the GNU General Public License as published by
6  *   the Free Software Foundation; either version 2 of the License, or
7  *   (at your option) any later version.
8  *
9  *   This program is distributed in the hope that it will be useful,
10  *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
11  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
12  *   the GNU General Public License for more details.
13  *
14  *   You should have received a copy of the GNU General Public License
15  *   along with this program;  if not, write to the Free Software
16  *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17  */
18 
19 /*
20  *	jfs_dtree.c: directory B+-tree manager
21  *
22  * B+-tree with variable length key directory:
23  *
24  * each directory page is structured as an array of 32-byte
25  * directory entry slots initialized as a freelist
26  * to avoid search/compaction of free space at insertion.
27  * when an entry is inserted, a number of slots are allocated
28  * from the freelist as required to store variable length data
29  * of the entry; when the entry is deleted, slots of the entry
30  * are returned to freelist.
31  *
32  * leaf entry stores full name as key and file serial number
33  * (aka inode number) as data.
34  * internal/router entry stores sufffix compressed name
35  * as key and simple extent descriptor as data.
36  *
37  * each directory page maintains a sorted entry index table
38  * which stores the start slot index of sorted entries
39  * to allow binary search on the table.
40  *
41  * directory starts as a root/leaf page in on-disk inode
42  * inline data area.
43  * when it becomes full, it starts a leaf of a external extent
44  * of length of 1 block. each time the first leaf becomes full,
45  * it is extended rather than split (its size is doubled),
46  * until its length becoms 4 KBytes, from then the extent is split
47  * with new 4 Kbyte extent when it becomes full
48  * to reduce external fragmentation of small directories.
49  *
50  * blah, blah, blah, for linear scan of directory in pieces by
51  * readdir().
52  *
53  *
54  *	case-insensitive directory file system
55  *
56  * names are stored in case-sensitive way in leaf entry.
57  * but stored, searched and compared in case-insensitive (uppercase) order
58  * (i.e., both search key and entry key are folded for search/compare):
59  * (note that case-sensitive order is BROKEN in storage, e.g.,
60  *  sensitive: Ad, aB, aC, aD -> insensitive: aB, aC, aD, Ad
61  *
62  *  entries which folds to the same key makes up a equivalent class
63  *  whose members are stored as contiguous cluster (may cross page boundary)
64  *  but whose order is arbitrary and acts as duplicate, e.g.,
65  *  abc, Abc, aBc, abC)
66  *
67  * once match is found at leaf, requires scan forward/backward
68  * either for, in case-insensitive search, duplicate
69  * or for, in case-sensitive search, for exact match
70  *
71  * router entry must be created/stored in case-insensitive way
72  * in internal entry:
73  * (right most key of left page and left most key of right page
74  * are folded, and its suffix compression is propagated as router
75  * key in parent)
76  * (e.g., if split occurs <abc> and <aBd>, <ABD> trather than <aB>
77  * should be made the router key for the split)
78  *
79  * case-insensitive search:
80  *
81  *	fold search key;
82  *
83  *	case-insensitive search of B-tree:
84  *	for internal entry, router key is already folded;
85  *	for leaf entry, fold the entry key before comparison.
86  *
87  *	if (leaf entry case-insensitive match found)
88  *		if (next entry satisfies case-insensitive match)
89  *			return EDUPLICATE;
90  *		if (prev entry satisfies case-insensitive match)
91  *			return EDUPLICATE;
92  *		return match;
93  *	else
94  *		return no match;
95  *
96  *	serialization:
97  * target directory inode lock is being held on entry/exit
98  * of all main directory service routines.
99  *
100  *	log based recovery:
101  */
102 
103 #include <linux/fs.h>
104 #include <linux/quotaops.h>
105 #include <linux/slab.h>
106 #include "jfs_incore.h"
107 #include "jfs_superblock.h"
108 #include "jfs_filsys.h"
109 #include "jfs_metapage.h"
110 #include "jfs_dmap.h"
111 #include "jfs_unicode.h"
112 #include "jfs_debug.h"
113 
114 /* dtree split parameter */
115 struct dtsplit {
116 	struct metapage *mp;
117 	s16 index;
118 	s16 nslot;
119 	struct component_name *key;
120 	ddata_t *data;
121 	struct pxdlist *pxdlist;
122 };
123 
124 #define DT_PAGE(IP, MP) BT_PAGE(IP, MP, dtpage_t, i_dtroot)
125 
126 /* get page buffer for specified block address */
127 #define DT_GETPAGE(IP, BN, MP, SIZE, P, RC)				\
128 do {									\
129 	BT_GETPAGE(IP, BN, MP, dtpage_t, SIZE, P, RC, i_dtroot);	\
130 	if (!(RC)) {							\
131 		if (((P)->header.nextindex >				\
132 		     (((BN) == 0) ? DTROOTMAXSLOT : (P)->header.maxslot)) || \
133 		    ((BN) && ((P)->header.maxslot > DTPAGEMAXSLOT))) {	\
134 			BT_PUTPAGE(MP);					\
135 			jfs_error((IP)->i_sb,				\
136 				  "DT_GETPAGE: dtree page corrupt\n");	\
137 			MP = NULL;					\
138 			RC = -EIO;					\
139 		}							\
140 	}								\
141 } while (0)
142 
143 /* for consistency */
144 #define DT_PUTPAGE(MP) BT_PUTPAGE(MP)
145 
146 #define DT_GETSEARCH(IP, LEAF, BN, MP, P, INDEX) \
147 	BT_GETSEARCH(IP, LEAF, BN, MP, dtpage_t, P, INDEX, i_dtroot)
148 
149 /*
150  * forward references
151  */
152 static int dtSplitUp(tid_t tid, struct inode *ip,
153 		     struct dtsplit * split, struct btstack * btstack);
154 
155 static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
156 		       struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rxdp);
157 
158 static int dtExtendPage(tid_t tid, struct inode *ip,
159 			struct dtsplit * split, struct btstack * btstack);
160 
161 static int dtSplitRoot(tid_t tid, struct inode *ip,
162 		       struct dtsplit * split, struct metapage ** rmpp);
163 
164 static int dtDeleteUp(tid_t tid, struct inode *ip, struct metapage * fmp,
165 		      dtpage_t * fp, struct btstack * btstack);
166 
167 static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p);
168 
169 static int dtReadFirst(struct inode *ip, struct btstack * btstack);
170 
171 static int dtReadNext(struct inode *ip,
172 		      loff_t * offset, struct btstack * btstack);
173 
174 static int dtCompare(struct component_name * key, dtpage_t * p, int si);
175 
176 static int ciCompare(struct component_name * key, dtpage_t * p, int si,
177 		     int flag);
178 
179 static void dtGetKey(dtpage_t * p, int i, struct component_name * key,
180 		     int flag);
181 
182 static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
183 			      int ri, struct component_name * key, int flag);
184 
185 static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
186 			  ddata_t * data, struct dt_lock **);
187 
188 static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
189 			struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
190 			int do_index);
191 
192 static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock);
193 
194 static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock);
195 
196 static void dtLinelockFreelist(dtpage_t * p, int m, struct dt_lock ** dtlock);
197 
198 #define ciToUpper(c)	UniStrupr((c)->name)
199 
200 /*
201  *	read_index_page()
202  *
203  *	Reads a page of a directory's index table.
204  *	Having metadata mapped into the directory inode's address space
205  *	presents a multitude of problems.  We avoid this by mapping to
206  *	the absolute address space outside of the *_metapage routines
207  */
208 static struct metapage *read_index_page(struct inode *inode, s64 blkno)
209 {
210 	int rc;
211 	s64 xaddr;
212 	int xflag;
213 	s32 xlen;
214 
215 	rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
216 	if (rc || (xaddr == 0))
217 		return NULL;
218 
219 	return read_metapage(inode, xaddr, PSIZE, 1);
220 }
221 
222 /*
223  *	get_index_page()
224  *
225  *	Same as get_index_page(), but get's a new page without reading
226  */
227 static struct metapage *get_index_page(struct inode *inode, s64 blkno)
228 {
229 	int rc;
230 	s64 xaddr;
231 	int xflag;
232 	s32 xlen;
233 
234 	rc = xtLookup(inode, blkno, 1, &xflag, &xaddr, &xlen, 1);
235 	if (rc || (xaddr == 0))
236 		return NULL;
237 
238 	return get_metapage(inode, xaddr, PSIZE, 1);
239 }
240 
241 /*
242  *	find_index()
243  *
244  *	Returns dtree page containing directory table entry for specified
245  *	index and pointer to its entry.
246  *
247  *	mp must be released by caller.
248  */
249 static struct dir_table_slot *find_index(struct inode *ip, u32 index,
250 					 struct metapage ** mp, s64 *lblock)
251 {
252 	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
253 	s64 blkno;
254 	s64 offset;
255 	int page_offset;
256 	struct dir_table_slot *slot;
257 	static int maxWarnings = 10;
258 
259 	if (index < 2) {
260 		if (maxWarnings) {
261 			jfs_warn("find_entry called with index = %d", index);
262 			maxWarnings--;
263 		}
264 		return NULL;
265 	}
266 
267 	if (index >= jfs_ip->next_index) {
268 		jfs_warn("find_entry called with index >= next_index");
269 		return NULL;
270 	}
271 
272 	if (jfs_dirtable_inline(ip)) {
273 		/*
274 		 * Inline directory table
275 		 */
276 		*mp = NULL;
277 		slot = &jfs_ip->i_dirtable[index - 2];
278 	} else {
279 		offset = (index - 2) * sizeof(struct dir_table_slot);
280 		page_offset = offset & (PSIZE - 1);
281 		blkno = ((offset + 1) >> L2PSIZE) <<
282 		    JFS_SBI(ip->i_sb)->l2nbperpage;
283 
284 		if (*mp && (*lblock != blkno)) {
285 			release_metapage(*mp);
286 			*mp = NULL;
287 		}
288 		if (!(*mp)) {
289 			*lblock = blkno;
290 			*mp = read_index_page(ip, blkno);
291 		}
292 		if (!(*mp)) {
293 			jfs_err("free_index: error reading directory table");
294 			return NULL;
295 		}
296 
297 		slot =
298 		    (struct dir_table_slot *) ((char *) (*mp)->data +
299 					       page_offset);
300 	}
301 	return slot;
302 }
303 
304 static inline void lock_index(tid_t tid, struct inode *ip, struct metapage * mp,
305 			      u32 index)
306 {
307 	struct tlock *tlck;
308 	struct linelock *llck;
309 	struct lv *lv;
310 
311 	tlck = txLock(tid, ip, mp, tlckDATA);
312 	llck = (struct linelock *) tlck->lock;
313 
314 	if (llck->index >= llck->maxcnt)
315 		llck = txLinelock(llck);
316 	lv = &llck->lv[llck->index];
317 
318 	/*
319 	 *	Linelock slot size is twice the size of directory table
320 	 *	slot size.  512 entries per page.
321 	 */
322 	lv->offset = ((index - 2) & 511) >> 1;
323 	lv->length = 1;
324 	llck->index++;
325 }
326 
327 /*
328  *	add_index()
329  *
330  *	Adds an entry to the directory index table.  This is used to provide
331  *	each directory entry with a persistent index in which to resume
332  *	directory traversals
333  */
334 static u32 add_index(tid_t tid, struct inode *ip, s64 bn, int slot)
335 {
336 	struct super_block *sb = ip->i_sb;
337 	struct jfs_sb_info *sbi = JFS_SBI(sb);
338 	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
339 	u64 blkno;
340 	struct dir_table_slot *dirtab_slot;
341 	u32 index;
342 	struct linelock *llck;
343 	struct lv *lv;
344 	struct metapage *mp;
345 	s64 offset;
346 	uint page_offset;
347 	struct tlock *tlck;
348 	s64 xaddr;
349 
350 	ASSERT(DO_INDEX(ip));
351 
352 	if (jfs_ip->next_index < 2) {
353 		jfs_warn("add_index: next_index = %d.  Resetting!",
354 			   jfs_ip->next_index);
355 		jfs_ip->next_index = 2;
356 	}
357 
358 	index = jfs_ip->next_index++;
359 
360 	if (index <= MAX_INLINE_DIRTABLE_ENTRY) {
361 		/*
362 		 * i_size reflects size of index table, or 8 bytes per entry.
363 		 */
364 		ip->i_size = (loff_t) (index - 1) << 3;
365 
366 		/*
367 		 * dir table fits inline within inode
368 		 */
369 		dirtab_slot = &jfs_ip->i_dirtable[index-2];
370 		dirtab_slot->flag = DIR_INDEX_VALID;
371 		dirtab_slot->slot = slot;
372 		DTSaddress(dirtab_slot, bn);
373 
374 		set_cflag(COMMIT_Dirtable, ip);
375 
376 		return index;
377 	}
378 	if (index == (MAX_INLINE_DIRTABLE_ENTRY + 1)) {
379 		struct dir_table_slot temp_table[12];
380 
381 		/*
382 		 * It's time to move the inline table to an external
383 		 * page and begin to build the xtree
384 		 */
385 		if (dquot_alloc_block(ip, sbi->nbperpage))
386 			goto clean_up;
387 		if (dbAlloc(ip, 0, sbi->nbperpage, &xaddr)) {
388 			dquot_free_block(ip, sbi->nbperpage);
389 			goto clean_up;
390 		}
391 
392 		/*
393 		 * Save the table, we're going to overwrite it with the
394 		 * xtree root
395 		 */
396 		memcpy(temp_table, &jfs_ip->i_dirtable, sizeof(temp_table));
397 
398 		/*
399 		 * Initialize empty x-tree
400 		 */
401 		xtInitRoot(tid, ip);
402 
403 		/*
404 		 * Add the first block to the xtree
405 		 */
406 		if (xtInsert(tid, ip, 0, 0, sbi->nbperpage, &xaddr, 0)) {
407 			/* This really shouldn't fail */
408 			jfs_warn("add_index: xtInsert failed!");
409 			memcpy(&jfs_ip->i_dirtable, temp_table,
410 			       sizeof (temp_table));
411 			dbFree(ip, xaddr, sbi->nbperpage);
412 			dquot_free_block(ip, sbi->nbperpage);
413 			goto clean_up;
414 		}
415 		ip->i_size = PSIZE;
416 
417 		mp = get_index_page(ip, 0);
418 		if (!mp) {
419 			jfs_err("add_index: get_metapage failed!");
420 			xtTruncate(tid, ip, 0, COMMIT_PWMAP);
421 			memcpy(&jfs_ip->i_dirtable, temp_table,
422 			       sizeof (temp_table));
423 			goto clean_up;
424 		}
425 		tlck = txLock(tid, ip, mp, tlckDATA);
426 		llck = (struct linelock *) & tlck->lock;
427 		ASSERT(llck->index == 0);
428 		lv = &llck->lv[0];
429 
430 		lv->offset = 0;
431 		lv->length = 6;	/* tlckDATA slot size is 16 bytes */
432 		llck->index++;
433 
434 		memcpy(mp->data, temp_table, sizeof(temp_table));
435 
436 		mark_metapage_dirty(mp);
437 		release_metapage(mp);
438 
439 		/*
440 		 * Logging is now directed by xtree tlocks
441 		 */
442 		clear_cflag(COMMIT_Dirtable, ip);
443 	}
444 
445 	offset = (index - 2) * sizeof(struct dir_table_slot);
446 	page_offset = offset & (PSIZE - 1);
447 	blkno = ((offset + 1) >> L2PSIZE) << sbi->l2nbperpage;
448 	if (page_offset == 0) {
449 		/*
450 		 * This will be the beginning of a new page
451 		 */
452 		xaddr = 0;
453 		if (xtInsert(tid, ip, 0, blkno, sbi->nbperpage, &xaddr, 0)) {
454 			jfs_warn("add_index: xtInsert failed!");
455 			goto clean_up;
456 		}
457 		ip->i_size += PSIZE;
458 
459 		if ((mp = get_index_page(ip, blkno)))
460 			memset(mp->data, 0, PSIZE);	/* Just looks better */
461 		else
462 			xtTruncate(tid, ip, offset, COMMIT_PWMAP);
463 	} else
464 		mp = read_index_page(ip, blkno);
465 
466 	if (!mp) {
467 		jfs_err("add_index: get/read_metapage failed!");
468 		goto clean_up;
469 	}
470 
471 	lock_index(tid, ip, mp, index);
472 
473 	dirtab_slot =
474 	    (struct dir_table_slot *) ((char *) mp->data + page_offset);
475 	dirtab_slot->flag = DIR_INDEX_VALID;
476 	dirtab_slot->slot = slot;
477 	DTSaddress(dirtab_slot, bn);
478 
479 	mark_metapage_dirty(mp);
480 	release_metapage(mp);
481 
482 	return index;
483 
484       clean_up:
485 
486 	jfs_ip->next_index--;
487 
488 	return 0;
489 }
490 
491 /*
492  *	free_index()
493  *
494  *	Marks an entry to the directory index table as free.
495  */
496 static void free_index(tid_t tid, struct inode *ip, u32 index, u32 next)
497 {
498 	struct dir_table_slot *dirtab_slot;
499 	s64 lblock;
500 	struct metapage *mp = NULL;
501 
502 	dirtab_slot = find_index(ip, index, &mp, &lblock);
503 
504 	if (!dirtab_slot)
505 		return;
506 
507 	dirtab_slot->flag = DIR_INDEX_FREE;
508 	dirtab_slot->slot = dirtab_slot->addr1 = 0;
509 	dirtab_slot->addr2 = cpu_to_le32(next);
510 
511 	if (mp) {
512 		lock_index(tid, ip, mp, index);
513 		mark_metapage_dirty(mp);
514 		release_metapage(mp);
515 	} else
516 		set_cflag(COMMIT_Dirtable, ip);
517 }
518 
519 /*
520  *	modify_index()
521  *
522  *	Changes an entry in the directory index table
523  */
524 static void modify_index(tid_t tid, struct inode *ip, u32 index, s64 bn,
525 			 int slot, struct metapage ** mp, s64 *lblock)
526 {
527 	struct dir_table_slot *dirtab_slot;
528 
529 	dirtab_slot = find_index(ip, index, mp, lblock);
530 
531 	if (!dirtab_slot)
532 		return;
533 
534 	DTSaddress(dirtab_slot, bn);
535 	dirtab_slot->slot = slot;
536 
537 	if (*mp) {
538 		lock_index(tid, ip, *mp, index);
539 		mark_metapage_dirty(*mp);
540 	} else
541 		set_cflag(COMMIT_Dirtable, ip);
542 }
543 
544 /*
545  *	read_index()
546  *
547  *	reads a directory table slot
548  */
549 static int read_index(struct inode *ip, u32 index,
550 		     struct dir_table_slot * dirtab_slot)
551 {
552 	s64 lblock;
553 	struct metapage *mp = NULL;
554 	struct dir_table_slot *slot;
555 
556 	slot = find_index(ip, index, &mp, &lblock);
557 	if (!slot) {
558 		return -EIO;
559 	}
560 
561 	memcpy(dirtab_slot, slot, sizeof(struct dir_table_slot));
562 
563 	if (mp)
564 		release_metapage(mp);
565 
566 	return 0;
567 }
568 
569 /*
570  *	dtSearch()
571  *
572  * function:
573  *	Search for the entry with specified key
574  *
575  * parameter:
576  *
577  * return: 0 - search result on stack, leaf page pinned;
578  *	   errno - I/O error
579  */
580 int dtSearch(struct inode *ip, struct component_name * key, ino_t * data,
581 	     struct btstack * btstack, int flag)
582 {
583 	int rc = 0;
584 	int cmp = 1;		/* init for empty page */
585 	s64 bn;
586 	struct metapage *mp;
587 	dtpage_t *p;
588 	s8 *stbl;
589 	int base, index, lim;
590 	struct btframe *btsp;
591 	pxd_t *pxd;
592 	int psize = 288;	/* initial in-line directory */
593 	ino_t inumber;
594 	struct component_name ciKey;
595 	struct super_block *sb = ip->i_sb;
596 
597 	ciKey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
598 				   GFP_NOFS);
599 	if (!ciKey.name) {
600 		rc = -ENOMEM;
601 		goto dtSearch_Exit2;
602 	}
603 
604 
605 	/* uppercase search key for c-i directory */
606 	UniStrcpy(ciKey.name, key->name);
607 	ciKey.namlen = key->namlen;
608 
609 	/* only uppercase if case-insensitive support is on */
610 	if ((JFS_SBI(sb)->mntflag & JFS_OS2) == JFS_OS2) {
611 		ciToUpper(&ciKey);
612 	}
613 	BT_CLR(btstack);	/* reset stack */
614 
615 	/* init level count for max pages to split */
616 	btstack->nsplit = 1;
617 
618 	/*
619 	 *	search down tree from root:
620 	 *
621 	 * between two consecutive entries of <Ki, Pi> and <Kj, Pj> of
622 	 * internal page, child page Pi contains entry with k, Ki <= K < Kj.
623 	 *
624 	 * if entry with search key K is not found
625 	 * internal page search find the entry with largest key Ki
626 	 * less than K which point to the child page to search;
627 	 * leaf page search find the entry with smallest key Kj
628 	 * greater than K so that the returned index is the position of
629 	 * the entry to be shifted right for insertion of new entry.
630 	 * for empty tree, search key is greater than any key of the tree.
631 	 *
632 	 * by convention, root bn = 0.
633 	 */
634 	for (bn = 0;;) {
635 		/* get/pin the page to search */
636 		DT_GETPAGE(ip, bn, mp, psize, p, rc);
637 		if (rc)
638 			goto dtSearch_Exit1;
639 
640 		/* get sorted entry table of the page */
641 		stbl = DT_GETSTBL(p);
642 
643 		/*
644 		 * binary search with search key K on the current page.
645 		 */
646 		for (base = 0, lim = p->header.nextindex; lim; lim >>= 1) {
647 			index = base + (lim >> 1);
648 
649 			if (p->header.flag & BT_LEAF) {
650 				/* uppercase leaf name to compare */
651 				cmp =
652 				    ciCompare(&ciKey, p, stbl[index],
653 					      JFS_SBI(sb)->mntflag);
654 			} else {
655 				/* router key is in uppercase */
656 
657 				cmp = dtCompare(&ciKey, p, stbl[index]);
658 
659 
660 			}
661 			if (cmp == 0) {
662 				/*
663 				 *	search hit
664 				 */
665 				/* search hit - leaf page:
666 				 * return the entry found
667 				 */
668 				if (p->header.flag & BT_LEAF) {
669 					inumber = le32_to_cpu(
670 			((struct ldtentry *) & p->slot[stbl[index]])->inumber);
671 
672 					/*
673 					 * search for JFS_LOOKUP
674 					 */
675 					if (flag == JFS_LOOKUP) {
676 						*data = inumber;
677 						rc = 0;
678 						goto out;
679 					}
680 
681 					/*
682 					 * search for JFS_CREATE
683 					 */
684 					if (flag == JFS_CREATE) {
685 						*data = inumber;
686 						rc = -EEXIST;
687 						goto out;
688 					}
689 
690 					/*
691 					 * search for JFS_REMOVE or JFS_RENAME
692 					 */
693 					if ((flag == JFS_REMOVE ||
694 					     flag == JFS_RENAME) &&
695 					    *data != inumber) {
696 						rc = -ESTALE;
697 						goto out;
698 					}
699 
700 					/*
701 					 * JFS_REMOVE|JFS_FINDDIR|JFS_RENAME
702 					 */
703 					/* save search result */
704 					*data = inumber;
705 					btsp = btstack->top;
706 					btsp->bn = bn;
707 					btsp->index = index;
708 					btsp->mp = mp;
709 
710 					rc = 0;
711 					goto dtSearch_Exit1;
712 				}
713 
714 				/* search hit - internal page:
715 				 * descend/search its child page
716 				 */
717 				goto getChild;
718 			}
719 
720 			if (cmp > 0) {
721 				base = index + 1;
722 				--lim;
723 			}
724 		}
725 
726 		/*
727 		 *	search miss
728 		 *
729 		 * base is the smallest index with key (Kj) greater than
730 		 * search key (K) and may be zero or (maxindex + 1) index.
731 		 */
732 		/*
733 		 * search miss - leaf page
734 		 *
735 		 * return location of entry (base) where new entry with
736 		 * search key K is to be inserted.
737 		 */
738 		if (p->header.flag & BT_LEAF) {
739 			/*
740 			 * search for JFS_LOOKUP, JFS_REMOVE, or JFS_RENAME
741 			 */
742 			if (flag == JFS_LOOKUP || flag == JFS_REMOVE ||
743 			    flag == JFS_RENAME) {
744 				rc = -ENOENT;
745 				goto out;
746 			}
747 
748 			/*
749 			 * search for JFS_CREATE|JFS_FINDDIR:
750 			 *
751 			 * save search result
752 			 */
753 			*data = 0;
754 			btsp = btstack->top;
755 			btsp->bn = bn;
756 			btsp->index = base;
757 			btsp->mp = mp;
758 
759 			rc = 0;
760 			goto dtSearch_Exit1;
761 		}
762 
763 		/*
764 		 * search miss - internal page
765 		 *
766 		 * if base is non-zero, decrement base by one to get the parent
767 		 * entry of the child page to search.
768 		 */
769 		index = base ? base - 1 : base;
770 
771 		/*
772 		 * go down to child page
773 		 */
774 	      getChild:
775 		/* update max. number of pages to split */
776 		if (BT_STACK_FULL(btstack)) {
777 			/* Something's corrupted, mark filesystem dirty so
778 			 * chkdsk will fix it.
779 			 */
780 			jfs_error(sb, "stack overrun!\n");
781 			BT_STACK_DUMP(btstack);
782 			rc = -EIO;
783 			goto out;
784 		}
785 		btstack->nsplit++;
786 
787 		/* push (bn, index) of the parent page/entry */
788 		BT_PUSH(btstack, bn, index);
789 
790 		/* get the child page block number */
791 		pxd = (pxd_t *) & p->slot[stbl[index]];
792 		bn = addressPXD(pxd);
793 		psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;
794 
795 		/* unpin the parent page */
796 		DT_PUTPAGE(mp);
797 	}
798 
799       out:
800 	DT_PUTPAGE(mp);
801 
802       dtSearch_Exit1:
803 
804 	kfree(ciKey.name);
805 
806       dtSearch_Exit2:
807 
808 	return rc;
809 }
810 
811 
812 /*
813  *	dtInsert()
814  *
815  * function: insert an entry to directory tree
816  *
817  * parameter:
818  *
819  * return: 0 - success;
820  *	   errno - failure;
821  */
822 int dtInsert(tid_t tid, struct inode *ip,
823 	 struct component_name * name, ino_t * fsn, struct btstack * btstack)
824 {
825 	int rc = 0;
826 	struct metapage *mp;	/* meta-page buffer */
827 	dtpage_t *p;		/* base B+-tree index page */
828 	s64 bn;
829 	int index;
830 	struct dtsplit split;	/* split information */
831 	ddata_t data;
832 	struct dt_lock *dtlck;
833 	int n;
834 	struct tlock *tlck;
835 	struct lv *lv;
836 
837 	/*
838 	 *	retrieve search result
839 	 *
840 	 * dtSearch() returns (leaf page pinned, index at which to insert).
841 	 * n.b. dtSearch() may return index of (maxindex + 1) of
842 	 * the full page.
843 	 */
844 	DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
845 
846 	/*
847 	 *	insert entry for new key
848 	 */
849 	if (DO_INDEX(ip)) {
850 		if (JFS_IP(ip)->next_index == DIREND) {
851 			DT_PUTPAGE(mp);
852 			return -EMLINK;
853 		}
854 		n = NDTLEAF(name->namlen);
855 		data.leaf.tid = tid;
856 		data.leaf.ip = ip;
857 	} else {
858 		n = NDTLEAF_LEGACY(name->namlen);
859 		data.leaf.ip = NULL;	/* signifies legacy directory format */
860 	}
861 	data.leaf.ino = *fsn;
862 
863 	/*
864 	 *	leaf page does not have enough room for new entry:
865 	 *
866 	 *	extend/split the leaf page;
867 	 *
868 	 * dtSplitUp() will insert the entry and unpin the leaf page.
869 	 */
870 	if (n > p->header.freecnt) {
871 		split.mp = mp;
872 		split.index = index;
873 		split.nslot = n;
874 		split.key = name;
875 		split.data = &data;
876 		rc = dtSplitUp(tid, ip, &split, btstack);
877 		return rc;
878 	}
879 
880 	/*
881 	 *	leaf page does have enough room for new entry:
882 	 *
883 	 *	insert the new data entry into the leaf page;
884 	 */
885 	BT_MARK_DIRTY(mp, ip);
886 	/*
887 	 * acquire a transaction lock on the leaf page
888 	 */
889 	tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
890 	dtlck = (struct dt_lock *) & tlck->lock;
891 	ASSERT(dtlck->index == 0);
892 	lv = & dtlck->lv[0];
893 
894 	/* linelock header */
895 	lv->offset = 0;
896 	lv->length = 1;
897 	dtlck->index++;
898 
899 	dtInsertEntry(p, index, name, &data, &dtlck);
900 
901 	/* linelock stbl of non-root leaf page */
902 	if (!(p->header.flag & BT_ROOT)) {
903 		if (dtlck->index >= dtlck->maxcnt)
904 			dtlck = (struct dt_lock *) txLinelock(dtlck);
905 		lv = & dtlck->lv[dtlck->index];
906 		n = index >> L2DTSLOTSIZE;
907 		lv->offset = p->header.stblindex + n;
908 		lv->length =
909 		    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
910 		dtlck->index++;
911 	}
912 
913 	/* unpin the leaf page */
914 	DT_PUTPAGE(mp);
915 
916 	return 0;
917 }
918 
919 
920 /*
921  *	dtSplitUp()
922  *
923  * function: propagate insertion bottom up;
924  *
925  * parameter:
926  *
927  * return: 0 - success;
928  *	   errno - failure;
929  *	leaf page unpinned;
930  */
931 static int dtSplitUp(tid_t tid,
932 	  struct inode *ip, struct dtsplit * split, struct btstack * btstack)
933 {
934 	struct jfs_sb_info *sbi = JFS_SBI(ip->i_sb);
935 	int rc = 0;
936 	struct metapage *smp;
937 	dtpage_t *sp;		/* split page */
938 	struct metapage *rmp;
939 	dtpage_t *rp;		/* new right page split from sp */
940 	pxd_t rpxd;		/* new right page extent descriptor */
941 	struct metapage *lmp;
942 	dtpage_t *lp;		/* left child page */
943 	int skip;		/* index of entry of insertion */
944 	struct btframe *parent;	/* parent page entry on traverse stack */
945 	s64 xaddr, nxaddr;
946 	int xlen, xsize;
947 	struct pxdlist pxdlist;
948 	pxd_t *pxd;
949 	struct component_name key = { 0, NULL };
950 	ddata_t *data = split->data;
951 	int n;
952 	struct dt_lock *dtlck;
953 	struct tlock *tlck;
954 	struct lv *lv;
955 	int quota_allocation = 0;
956 
957 	/* get split page */
958 	smp = split->mp;
959 	sp = DT_PAGE(ip, smp);
960 
961 	key.name = kmalloc_array(JFS_NAME_MAX + 2, sizeof(wchar_t), GFP_NOFS);
962 	if (!key.name) {
963 		DT_PUTPAGE(smp);
964 		rc = -ENOMEM;
965 		goto dtSplitUp_Exit;
966 	}
967 
968 	/*
969 	 *	split leaf page
970 	 *
971 	 * The split routines insert the new entry, and
972 	 * acquire txLock as appropriate.
973 	 */
974 	/*
975 	 *	split root leaf page:
976 	 */
977 	if (sp->header.flag & BT_ROOT) {
978 		/*
979 		 * allocate a single extent child page
980 		 */
981 		xlen = 1;
982 		n = sbi->bsize >> L2DTSLOTSIZE;
983 		n -= (n + 31) >> L2DTSLOTSIZE;	/* stbl size */
984 		n -= DTROOTMAXSLOT - sp->header.freecnt; /* header + entries */
985 		if (n <= split->nslot)
986 			xlen++;
987 		if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr))) {
988 			DT_PUTPAGE(smp);
989 			goto freeKeyName;
990 		}
991 
992 		pxdlist.maxnpxd = 1;
993 		pxdlist.npxd = 0;
994 		pxd = &pxdlist.pxd[0];
995 		PXDaddress(pxd, xaddr);
996 		PXDlength(pxd, xlen);
997 		split->pxdlist = &pxdlist;
998 		rc = dtSplitRoot(tid, ip, split, &rmp);
999 
1000 		if (rc)
1001 			dbFree(ip, xaddr, xlen);
1002 		else
1003 			DT_PUTPAGE(rmp);
1004 
1005 		DT_PUTPAGE(smp);
1006 
1007 		if (!DO_INDEX(ip))
1008 			ip->i_size = xlen << sbi->l2bsize;
1009 
1010 		goto freeKeyName;
1011 	}
1012 
1013 	/*
1014 	 *	extend first leaf page
1015 	 *
1016 	 * extend the 1st extent if less than buffer page size
1017 	 * (dtExtendPage() reurns leaf page unpinned)
1018 	 */
1019 	pxd = &sp->header.self;
1020 	xlen = lengthPXD(pxd);
1021 	xsize = xlen << sbi->l2bsize;
1022 	if (xsize < PSIZE) {
1023 		xaddr = addressPXD(pxd);
1024 		n = xsize >> L2DTSLOTSIZE;
1025 		n -= (n + 31) >> L2DTSLOTSIZE;	/* stbl size */
1026 		if ((n + sp->header.freecnt) <= split->nslot)
1027 			n = xlen + (xlen << 1);
1028 		else
1029 			n = xlen;
1030 
1031 		/* Allocate blocks to quota. */
1032 		rc = dquot_alloc_block(ip, n);
1033 		if (rc)
1034 			goto extendOut;
1035 		quota_allocation += n;
1036 
1037 		if ((rc = dbReAlloc(sbi->ipbmap, xaddr, (s64) xlen,
1038 				    (s64) n, &nxaddr)))
1039 			goto extendOut;
1040 
1041 		pxdlist.maxnpxd = 1;
1042 		pxdlist.npxd = 0;
1043 		pxd = &pxdlist.pxd[0];
1044 		PXDaddress(pxd, nxaddr);
1045 		PXDlength(pxd, xlen + n);
1046 		split->pxdlist = &pxdlist;
1047 		if ((rc = dtExtendPage(tid, ip, split, btstack))) {
1048 			nxaddr = addressPXD(pxd);
1049 			if (xaddr != nxaddr) {
1050 				/* free relocated extent */
1051 				xlen = lengthPXD(pxd);
1052 				dbFree(ip, nxaddr, (s64) xlen);
1053 			} else {
1054 				/* free extended delta */
1055 				xlen = lengthPXD(pxd) - n;
1056 				xaddr = addressPXD(pxd) + xlen;
1057 				dbFree(ip, xaddr, (s64) n);
1058 			}
1059 		} else if (!DO_INDEX(ip))
1060 			ip->i_size = lengthPXD(pxd) << sbi->l2bsize;
1061 
1062 
1063 	      extendOut:
1064 		DT_PUTPAGE(smp);
1065 		goto freeKeyName;
1066 	}
1067 
1068 	/*
1069 	 *	split leaf page <sp> into <sp> and a new right page <rp>.
1070 	 *
1071 	 * return <rp> pinned and its extent descriptor <rpxd>
1072 	 */
1073 	/*
1074 	 * allocate new directory page extent and
1075 	 * new index page(s) to cover page split(s)
1076 	 *
1077 	 * allocation hint: ?
1078 	 */
1079 	n = btstack->nsplit;
1080 	pxdlist.maxnpxd = pxdlist.npxd = 0;
1081 	xlen = sbi->nbperpage;
1082 	for (pxd = pxdlist.pxd; n > 0; n--, pxd++) {
1083 		if ((rc = dbAlloc(ip, 0, (s64) xlen, &xaddr)) == 0) {
1084 			PXDaddress(pxd, xaddr);
1085 			PXDlength(pxd, xlen);
1086 			pxdlist.maxnpxd++;
1087 			continue;
1088 		}
1089 
1090 		DT_PUTPAGE(smp);
1091 
1092 		/* undo allocation */
1093 		goto splitOut;
1094 	}
1095 
1096 	split->pxdlist = &pxdlist;
1097 	if ((rc = dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd))) {
1098 		DT_PUTPAGE(smp);
1099 
1100 		/* undo allocation */
1101 		goto splitOut;
1102 	}
1103 
1104 	if (!DO_INDEX(ip))
1105 		ip->i_size += PSIZE;
1106 
1107 	/*
1108 	 * propagate up the router entry for the leaf page just split
1109 	 *
1110 	 * insert a router entry for the new page into the parent page,
1111 	 * propagate the insert/split up the tree by walking back the stack
1112 	 * of (bn of parent page, index of child page entry in parent page)
1113 	 * that were traversed during the search for the page that split.
1114 	 *
1115 	 * the propagation of insert/split up the tree stops if the root
1116 	 * splits or the page inserted into doesn't have to split to hold
1117 	 * the new entry.
1118 	 *
1119 	 * the parent entry for the split page remains the same, and
1120 	 * a new entry is inserted at its right with the first key and
1121 	 * block number of the new right page.
1122 	 *
1123 	 * There are a maximum of 4 pages pinned at any time:
1124 	 * two children, left parent and right parent (when the parent splits).
1125 	 * keep the child pages pinned while working on the parent.
1126 	 * make sure that all pins are released at exit.
1127 	 */
1128 	while ((parent = BT_POP(btstack)) != NULL) {
1129 		/* parent page specified by stack frame <parent> */
1130 
1131 		/* keep current child pages (<lp>, <rp>) pinned */
1132 		lmp = smp;
1133 		lp = sp;
1134 
1135 		/*
1136 		 * insert router entry in parent for new right child page <rp>
1137 		 */
1138 		/* get the parent page <sp> */
1139 		DT_GETPAGE(ip, parent->bn, smp, PSIZE, sp, rc);
1140 		if (rc) {
1141 			DT_PUTPAGE(lmp);
1142 			DT_PUTPAGE(rmp);
1143 			goto splitOut;
1144 		}
1145 
1146 		/*
1147 		 * The new key entry goes ONE AFTER the index of parent entry,
1148 		 * because the split was to the right.
1149 		 */
1150 		skip = parent->index + 1;
1151 
1152 		/*
1153 		 * compute the key for the router entry
1154 		 *
1155 		 * key suffix compression:
1156 		 * for internal pages that have leaf pages as children,
1157 		 * retain only what's needed to distinguish between
1158 		 * the new entry and the entry on the page to its left.
1159 		 * If the keys compare equal, retain the entire key.
1160 		 *
1161 		 * note that compression is performed only at computing
1162 		 * router key at the lowest internal level.
1163 		 * further compression of the key between pairs of higher
1164 		 * level internal pages loses too much information and
1165 		 * the search may fail.
1166 		 * (e.g., two adjacent leaf pages of {a, ..., x} {xx, ...,}
1167 		 * results in two adjacent parent entries (a)(xx).
1168 		 * if split occurs between these two entries, and
1169 		 * if compression is applied, the router key of parent entry
1170 		 * of right page (x) will divert search for x into right
1171 		 * subtree and miss x in the left subtree.)
1172 		 *
1173 		 * the entire key must be retained for the next-to-leftmost
1174 		 * internal key at any level of the tree, or search may fail
1175 		 * (e.g., ?)
1176 		 */
1177 		switch (rp->header.flag & BT_TYPE) {
1178 		case BT_LEAF:
1179 			/*
1180 			 * compute the length of prefix for suffix compression
1181 			 * between last entry of left page and first entry
1182 			 * of right page
1183 			 */
1184 			if ((sp->header.flag & BT_ROOT && skip > 1) ||
1185 			    sp->header.prev != 0 || skip > 1) {
1186 				/* compute uppercase router prefix key */
1187 				rc = ciGetLeafPrefixKey(lp,
1188 							lp->header.nextindex-1,
1189 							rp, 0, &key,
1190 							sbi->mntflag);
1191 				if (rc) {
1192 					DT_PUTPAGE(lmp);
1193 					DT_PUTPAGE(rmp);
1194 					DT_PUTPAGE(smp);
1195 					goto splitOut;
1196 				}
1197 			} else {
1198 				/* next to leftmost entry of
1199 				   lowest internal level */
1200 
1201 				/* compute uppercase router key */
1202 				dtGetKey(rp, 0, &key, sbi->mntflag);
1203 				key.name[key.namlen] = 0;
1204 
1205 				if ((sbi->mntflag & JFS_OS2) == JFS_OS2)
1206 					ciToUpper(&key);
1207 			}
1208 
1209 			n = NDTINTERNAL(key.namlen);
1210 			break;
1211 
1212 		case BT_INTERNAL:
1213 			dtGetKey(rp, 0, &key, sbi->mntflag);
1214 			n = NDTINTERNAL(key.namlen);
1215 			break;
1216 
1217 		default:
1218 			jfs_err("dtSplitUp(): UFO!");
1219 			break;
1220 		}
1221 
1222 		/* unpin left child page */
1223 		DT_PUTPAGE(lmp);
1224 
1225 		/*
1226 		 * compute the data for the router entry
1227 		 */
1228 		data->xd = rpxd;	/* child page xd */
1229 
1230 		/*
1231 		 * parent page is full - split the parent page
1232 		 */
1233 		if (n > sp->header.freecnt) {
1234 			/* init for parent page split */
1235 			split->mp = smp;
1236 			split->index = skip;	/* index at insert */
1237 			split->nslot = n;
1238 			split->key = &key;
1239 			/* split->data = data; */
1240 
1241 			/* unpin right child page */
1242 			DT_PUTPAGE(rmp);
1243 
1244 			/* The split routines insert the new entry,
1245 			 * acquire txLock as appropriate.
1246 			 * return <rp> pinned and its block number <rbn>.
1247 			 */
1248 			rc = (sp->header.flag & BT_ROOT) ?
1249 			    dtSplitRoot(tid, ip, split, &rmp) :
1250 			    dtSplitPage(tid, ip, split, &rmp, &rp, &rpxd);
1251 			if (rc) {
1252 				DT_PUTPAGE(smp);
1253 				goto splitOut;
1254 			}
1255 
1256 			/* smp and rmp are pinned */
1257 		}
1258 		/*
1259 		 * parent page is not full - insert router entry in parent page
1260 		 */
1261 		else {
1262 			BT_MARK_DIRTY(smp, ip);
1263 			/*
1264 			 * acquire a transaction lock on the parent page
1265 			 */
1266 			tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1267 			dtlck = (struct dt_lock *) & tlck->lock;
1268 			ASSERT(dtlck->index == 0);
1269 			lv = & dtlck->lv[0];
1270 
1271 			/* linelock header */
1272 			lv->offset = 0;
1273 			lv->length = 1;
1274 			dtlck->index++;
1275 
1276 			/* linelock stbl of non-root parent page */
1277 			if (!(sp->header.flag & BT_ROOT)) {
1278 				lv++;
1279 				n = skip >> L2DTSLOTSIZE;
1280 				lv->offset = sp->header.stblindex + n;
1281 				lv->length =
1282 				    ((sp->header.nextindex -
1283 				      1) >> L2DTSLOTSIZE) - n + 1;
1284 				dtlck->index++;
1285 			}
1286 
1287 			dtInsertEntry(sp, skip, &key, data, &dtlck);
1288 
1289 			/* exit propagate up */
1290 			break;
1291 		}
1292 	}
1293 
1294 	/* unpin current split and its right page */
1295 	DT_PUTPAGE(smp);
1296 	DT_PUTPAGE(rmp);
1297 
1298 	/*
1299 	 * free remaining extents allocated for split
1300 	 */
1301       splitOut:
1302 	n = pxdlist.npxd;
1303 	pxd = &pxdlist.pxd[n];
1304 	for (; n < pxdlist.maxnpxd; n++, pxd++)
1305 		dbFree(ip, addressPXD(pxd), (s64) lengthPXD(pxd));
1306 
1307       freeKeyName:
1308 	kfree(key.name);
1309 
1310 	/* Rollback quota allocation */
1311 	if (rc && quota_allocation)
1312 		dquot_free_block(ip, quota_allocation);
1313 
1314       dtSplitUp_Exit:
1315 
1316 	return rc;
1317 }
1318 
1319 
1320 /*
1321  *	dtSplitPage()
1322  *
1323  * function: Split a non-root page of a btree.
1324  *
1325  * parameter:
1326  *
1327  * return: 0 - success;
1328  *	   errno - failure;
1329  *	return split and new page pinned;
1330  */
1331 static int dtSplitPage(tid_t tid, struct inode *ip, struct dtsplit * split,
1332 	    struct metapage ** rmpp, dtpage_t ** rpp, pxd_t * rpxdp)
1333 {
1334 	int rc = 0;
1335 	struct metapage *smp;
1336 	dtpage_t *sp;
1337 	struct metapage *rmp;
1338 	dtpage_t *rp;		/* new right page allocated */
1339 	s64 rbn;		/* new right page block number */
1340 	struct metapage *mp;
1341 	dtpage_t *p;
1342 	s64 nextbn;
1343 	struct pxdlist *pxdlist;
1344 	pxd_t *pxd;
1345 	int skip, nextindex, half, left, nxt, off, si;
1346 	struct ldtentry *ldtentry;
1347 	struct idtentry *idtentry;
1348 	u8 *stbl;
1349 	struct dtslot *f;
1350 	int fsi, stblsize;
1351 	int n;
1352 	struct dt_lock *sdtlck, *rdtlck;
1353 	struct tlock *tlck;
1354 	struct dt_lock *dtlck;
1355 	struct lv *slv, *rlv, *lv;
1356 
1357 	/* get split page */
1358 	smp = split->mp;
1359 	sp = DT_PAGE(ip, smp);
1360 
1361 	/*
1362 	 * allocate the new right page for the split
1363 	 */
1364 	pxdlist = split->pxdlist;
1365 	pxd = &pxdlist->pxd[pxdlist->npxd];
1366 	pxdlist->npxd++;
1367 	rbn = addressPXD(pxd);
1368 	rmp = get_metapage(ip, rbn, PSIZE, 1);
1369 	if (rmp == NULL)
1370 		return -EIO;
1371 
1372 	/* Allocate blocks to quota. */
1373 	rc = dquot_alloc_block(ip, lengthPXD(pxd));
1374 	if (rc) {
1375 		release_metapage(rmp);
1376 		return rc;
1377 	}
1378 
1379 	jfs_info("dtSplitPage: ip:0x%p smp:0x%p rmp:0x%p", ip, smp, rmp);
1380 
1381 	BT_MARK_DIRTY(rmp, ip);
1382 	/*
1383 	 * acquire a transaction lock on the new right page
1384 	 */
1385 	tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1386 	rdtlck = (struct dt_lock *) & tlck->lock;
1387 
1388 	rp = (dtpage_t *) rmp->data;
1389 	*rpp = rp;
1390 	rp->header.self = *pxd;
1391 
1392 	BT_MARK_DIRTY(smp, ip);
1393 	/*
1394 	 * acquire a transaction lock on the split page
1395 	 *
1396 	 * action:
1397 	 */
1398 	tlck = txLock(tid, ip, smp, tlckDTREE | tlckENTRY);
1399 	sdtlck = (struct dt_lock *) & tlck->lock;
1400 
1401 	/* linelock header of split page */
1402 	ASSERT(sdtlck->index == 0);
1403 	slv = & sdtlck->lv[0];
1404 	slv->offset = 0;
1405 	slv->length = 1;
1406 	sdtlck->index++;
1407 
1408 	/*
1409 	 * initialize/update sibling pointers between sp and rp
1410 	 */
1411 	nextbn = le64_to_cpu(sp->header.next);
1412 	rp->header.next = cpu_to_le64(nextbn);
1413 	rp->header.prev = cpu_to_le64(addressPXD(&sp->header.self));
1414 	sp->header.next = cpu_to_le64(rbn);
1415 
1416 	/*
1417 	 * initialize new right page
1418 	 */
1419 	rp->header.flag = sp->header.flag;
1420 
1421 	/* compute sorted entry table at start of extent data area */
1422 	rp->header.nextindex = 0;
1423 	rp->header.stblindex = 1;
1424 
1425 	n = PSIZE >> L2DTSLOTSIZE;
1426 	rp->header.maxslot = n;
1427 	stblsize = (n + 31) >> L2DTSLOTSIZE;	/* in unit of slot */
1428 
1429 	/* init freelist */
1430 	fsi = rp->header.stblindex + stblsize;
1431 	rp->header.freelist = fsi;
1432 	rp->header.freecnt = rp->header.maxslot - fsi;
1433 
1434 	/*
1435 	 *	sequential append at tail: append without split
1436 	 *
1437 	 * If splitting the last page on a level because of appending
1438 	 * a entry to it (skip is maxentry), it's likely that the access is
1439 	 * sequential. Adding an empty page on the side of the level is less
1440 	 * work and can push the fill factor much higher than normal.
1441 	 * If we're wrong it's no big deal, we'll just do the split the right
1442 	 * way next time.
1443 	 * (It may look like it's equally easy to do a similar hack for
1444 	 * reverse sorted data, that is, split the tree left,
1445 	 * but it's not. Be my guest.)
1446 	 */
1447 	if (nextbn == 0 && split->index == sp->header.nextindex) {
1448 		/* linelock header + stbl (first slot) of new page */
1449 		rlv = & rdtlck->lv[rdtlck->index];
1450 		rlv->offset = 0;
1451 		rlv->length = 2;
1452 		rdtlck->index++;
1453 
1454 		/*
1455 		 * initialize freelist of new right page
1456 		 */
1457 		f = &rp->slot[fsi];
1458 		for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1459 			f->next = fsi;
1460 		f->next = -1;
1461 
1462 		/* insert entry at the first entry of the new right page */
1463 		dtInsertEntry(rp, 0, split->key, split->data, &rdtlck);
1464 
1465 		goto out;
1466 	}
1467 
1468 	/*
1469 	 *	non-sequential insert (at possibly middle page)
1470 	 */
1471 
1472 	/*
1473 	 * update prev pointer of previous right sibling page;
1474 	 */
1475 	if (nextbn != 0) {
1476 		DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
1477 		if (rc) {
1478 			discard_metapage(rmp);
1479 			return rc;
1480 		}
1481 
1482 		BT_MARK_DIRTY(mp, ip);
1483 		/*
1484 		 * acquire a transaction lock on the next page
1485 		 */
1486 		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
1487 		jfs_info("dtSplitPage: tlck = 0x%p, ip = 0x%p, mp=0x%p",
1488 			tlck, ip, mp);
1489 		dtlck = (struct dt_lock *) & tlck->lock;
1490 
1491 		/* linelock header of previous right sibling page */
1492 		lv = & dtlck->lv[dtlck->index];
1493 		lv->offset = 0;
1494 		lv->length = 1;
1495 		dtlck->index++;
1496 
1497 		p->header.prev = cpu_to_le64(rbn);
1498 
1499 		DT_PUTPAGE(mp);
1500 	}
1501 
1502 	/*
1503 	 * split the data between the split and right pages.
1504 	 */
1505 	skip = split->index;
1506 	half = (PSIZE >> L2DTSLOTSIZE) >> 1;	/* swag */
1507 	left = 0;
1508 
1509 	/*
1510 	 *	compute fill factor for split pages
1511 	 *
1512 	 * <nxt> traces the next entry to move to rp
1513 	 * <off> traces the next entry to stay in sp
1514 	 */
1515 	stbl = (u8 *) & sp->slot[sp->header.stblindex];
1516 	nextindex = sp->header.nextindex;
1517 	for (nxt = off = 0; nxt < nextindex; ++off) {
1518 		if (off == skip)
1519 			/* check for fill factor with new entry size */
1520 			n = split->nslot;
1521 		else {
1522 			si = stbl[nxt];
1523 			switch (sp->header.flag & BT_TYPE) {
1524 			case BT_LEAF:
1525 				ldtentry = (struct ldtentry *) & sp->slot[si];
1526 				if (DO_INDEX(ip))
1527 					n = NDTLEAF(ldtentry->namlen);
1528 				else
1529 					n = NDTLEAF_LEGACY(ldtentry->
1530 							   namlen);
1531 				break;
1532 
1533 			case BT_INTERNAL:
1534 				idtentry = (struct idtentry *) & sp->slot[si];
1535 				n = NDTINTERNAL(idtentry->namlen);
1536 				break;
1537 
1538 			default:
1539 				break;
1540 			}
1541 
1542 			++nxt;	/* advance to next entry to move in sp */
1543 		}
1544 
1545 		left += n;
1546 		if (left >= half)
1547 			break;
1548 	}
1549 
1550 	/* <nxt> poins to the 1st entry to move */
1551 
1552 	/*
1553 	 *	move entries to right page
1554 	 *
1555 	 * dtMoveEntry() initializes rp and reserves entry for insertion
1556 	 *
1557 	 * split page moved out entries are linelocked;
1558 	 * new/right page moved in entries are linelocked;
1559 	 */
1560 	/* linelock header + stbl of new right page */
1561 	rlv = & rdtlck->lv[rdtlck->index];
1562 	rlv->offset = 0;
1563 	rlv->length = 5;
1564 	rdtlck->index++;
1565 
1566 	dtMoveEntry(sp, nxt, rp, &sdtlck, &rdtlck, DO_INDEX(ip));
1567 
1568 	sp->header.nextindex = nxt;
1569 
1570 	/*
1571 	 * finalize freelist of new right page
1572 	 */
1573 	fsi = rp->header.freelist;
1574 	f = &rp->slot[fsi];
1575 	for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1576 		f->next = fsi;
1577 	f->next = -1;
1578 
1579 	/*
1580 	 * Update directory index table for entries now in right page
1581 	 */
1582 	if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1583 		s64 lblock;
1584 
1585 		mp = NULL;
1586 		stbl = DT_GETSTBL(rp);
1587 		for (n = 0; n < rp->header.nextindex; n++) {
1588 			ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
1589 			modify_index(tid, ip, le32_to_cpu(ldtentry->index),
1590 				     rbn, n, &mp, &lblock);
1591 		}
1592 		if (mp)
1593 			release_metapage(mp);
1594 	}
1595 
1596 	/*
1597 	 * the skipped index was on the left page,
1598 	 */
1599 	if (skip <= off) {
1600 		/* insert the new entry in the split page */
1601 		dtInsertEntry(sp, skip, split->key, split->data, &sdtlck);
1602 
1603 		/* linelock stbl of split page */
1604 		if (sdtlck->index >= sdtlck->maxcnt)
1605 			sdtlck = (struct dt_lock *) txLinelock(sdtlck);
1606 		slv = & sdtlck->lv[sdtlck->index];
1607 		n = skip >> L2DTSLOTSIZE;
1608 		slv->offset = sp->header.stblindex + n;
1609 		slv->length =
1610 		    ((sp->header.nextindex - 1) >> L2DTSLOTSIZE) - n + 1;
1611 		sdtlck->index++;
1612 	}
1613 	/*
1614 	 * the skipped index was on the right page,
1615 	 */
1616 	else {
1617 		/* adjust the skip index to reflect the new position */
1618 		skip -= nxt;
1619 
1620 		/* insert the new entry in the right page */
1621 		dtInsertEntry(rp, skip, split->key, split->data, &rdtlck);
1622 	}
1623 
1624       out:
1625 	*rmpp = rmp;
1626 	*rpxdp = *pxd;
1627 
1628 	return rc;
1629 }
1630 
1631 
1632 /*
1633  *	dtExtendPage()
1634  *
1635  * function: extend 1st/only directory leaf page
1636  *
1637  * parameter:
1638  *
1639  * return: 0 - success;
1640  *	   errno - failure;
1641  *	return extended page pinned;
1642  */
1643 static int dtExtendPage(tid_t tid,
1644 	     struct inode *ip, struct dtsplit * split, struct btstack * btstack)
1645 {
1646 	struct super_block *sb = ip->i_sb;
1647 	int rc;
1648 	struct metapage *smp, *pmp, *mp;
1649 	dtpage_t *sp, *pp;
1650 	struct pxdlist *pxdlist;
1651 	pxd_t *pxd, *tpxd;
1652 	int xlen, xsize;
1653 	int newstblindex, newstblsize;
1654 	int oldstblindex, oldstblsize;
1655 	int fsi, last;
1656 	struct dtslot *f;
1657 	struct btframe *parent;
1658 	int n;
1659 	struct dt_lock *dtlck;
1660 	s64 xaddr, txaddr;
1661 	struct tlock *tlck;
1662 	struct pxd_lock *pxdlock;
1663 	struct lv *lv;
1664 	uint type;
1665 	struct ldtentry *ldtentry;
1666 	u8 *stbl;
1667 
1668 	/* get page to extend */
1669 	smp = split->mp;
1670 	sp = DT_PAGE(ip, smp);
1671 
1672 	/* get parent/root page */
1673 	parent = BT_POP(btstack);
1674 	DT_GETPAGE(ip, parent->bn, pmp, PSIZE, pp, rc);
1675 	if (rc)
1676 		return (rc);
1677 
1678 	/*
1679 	 *	extend the extent
1680 	 */
1681 	pxdlist = split->pxdlist;
1682 	pxd = &pxdlist->pxd[pxdlist->npxd];
1683 	pxdlist->npxd++;
1684 
1685 	xaddr = addressPXD(pxd);
1686 	tpxd = &sp->header.self;
1687 	txaddr = addressPXD(tpxd);
1688 	/* in-place extension */
1689 	if (xaddr == txaddr) {
1690 		type = tlckEXTEND;
1691 	}
1692 	/* relocation */
1693 	else {
1694 		type = tlckNEW;
1695 
1696 		/* save moved extent descriptor for later free */
1697 		tlck = txMaplock(tid, ip, tlckDTREE | tlckRELOCATE);
1698 		pxdlock = (struct pxd_lock *) & tlck->lock;
1699 		pxdlock->flag = mlckFREEPXD;
1700 		pxdlock->pxd = sp->header.self;
1701 		pxdlock->index = 1;
1702 
1703 		/*
1704 		 * Update directory index table to reflect new page address
1705 		 */
1706 		if (DO_INDEX(ip)) {
1707 			s64 lblock;
1708 
1709 			mp = NULL;
1710 			stbl = DT_GETSTBL(sp);
1711 			for (n = 0; n < sp->header.nextindex; n++) {
1712 				ldtentry =
1713 				    (struct ldtentry *) & sp->slot[stbl[n]];
1714 				modify_index(tid, ip,
1715 					     le32_to_cpu(ldtentry->index),
1716 					     xaddr, n, &mp, &lblock);
1717 			}
1718 			if (mp)
1719 				release_metapage(mp);
1720 		}
1721 	}
1722 
1723 	/*
1724 	 *	extend the page
1725 	 */
1726 	sp->header.self = *pxd;
1727 
1728 	jfs_info("dtExtendPage: ip:0x%p smp:0x%p sp:0x%p", ip, smp, sp);
1729 
1730 	BT_MARK_DIRTY(smp, ip);
1731 	/*
1732 	 * acquire a transaction lock on the extended/leaf page
1733 	 */
1734 	tlck = txLock(tid, ip, smp, tlckDTREE | type);
1735 	dtlck = (struct dt_lock *) & tlck->lock;
1736 	lv = & dtlck->lv[0];
1737 
1738 	/* update buffer extent descriptor of extended page */
1739 	xlen = lengthPXD(pxd);
1740 	xsize = xlen << JFS_SBI(sb)->l2bsize;
1741 
1742 	/*
1743 	 * copy old stbl to new stbl at start of extended area
1744 	 */
1745 	oldstblindex = sp->header.stblindex;
1746 	oldstblsize = (sp->header.maxslot + 31) >> L2DTSLOTSIZE;
1747 	newstblindex = sp->header.maxslot;
1748 	n = xsize >> L2DTSLOTSIZE;
1749 	newstblsize = (n + 31) >> L2DTSLOTSIZE;
1750 	memcpy(&sp->slot[newstblindex], &sp->slot[oldstblindex],
1751 	       sp->header.nextindex);
1752 
1753 	/*
1754 	 * in-line extension: linelock old area of extended page
1755 	 */
1756 	if (type == tlckEXTEND) {
1757 		/* linelock header */
1758 		lv->offset = 0;
1759 		lv->length = 1;
1760 		dtlck->index++;
1761 		lv++;
1762 
1763 		/* linelock new stbl of extended page */
1764 		lv->offset = newstblindex;
1765 		lv->length = newstblsize;
1766 	}
1767 	/*
1768 	 * relocation: linelock whole relocated area
1769 	 */
1770 	else {
1771 		lv->offset = 0;
1772 		lv->length = sp->header.maxslot + newstblsize;
1773 	}
1774 
1775 	dtlck->index++;
1776 
1777 	sp->header.maxslot = n;
1778 	sp->header.stblindex = newstblindex;
1779 	/* sp->header.nextindex remains the same */
1780 
1781 	/*
1782 	 * add old stbl region at head of freelist
1783 	 */
1784 	fsi = oldstblindex;
1785 	f = &sp->slot[fsi];
1786 	last = sp->header.freelist;
1787 	for (n = 0; n < oldstblsize; n++, fsi++, f++) {
1788 		f->next = last;
1789 		last = fsi;
1790 	}
1791 	sp->header.freelist = last;
1792 	sp->header.freecnt += oldstblsize;
1793 
1794 	/*
1795 	 * append free region of newly extended area at tail of freelist
1796 	 */
1797 	/* init free region of newly extended area */
1798 	fsi = n = newstblindex + newstblsize;
1799 	f = &sp->slot[fsi];
1800 	for (fsi++; fsi < sp->header.maxslot; f++, fsi++)
1801 		f->next = fsi;
1802 	f->next = -1;
1803 
1804 	/* append new free region at tail of old freelist */
1805 	fsi = sp->header.freelist;
1806 	if (fsi == -1)
1807 		sp->header.freelist = n;
1808 	else {
1809 		do {
1810 			f = &sp->slot[fsi];
1811 			fsi = f->next;
1812 		} while (fsi != -1);
1813 
1814 		f->next = n;
1815 	}
1816 
1817 	sp->header.freecnt += sp->header.maxslot - n;
1818 
1819 	/*
1820 	 * insert the new entry
1821 	 */
1822 	dtInsertEntry(sp, split->index, split->key, split->data, &dtlck);
1823 
1824 	BT_MARK_DIRTY(pmp, ip);
1825 	/*
1826 	 * linelock any freeslots residing in old extent
1827 	 */
1828 	if (type == tlckEXTEND) {
1829 		n = sp->header.maxslot >> 2;
1830 		if (sp->header.freelist < n)
1831 			dtLinelockFreelist(sp, n, &dtlck);
1832 	}
1833 
1834 	/*
1835 	 *	update parent entry on the parent/root page
1836 	 */
1837 	/*
1838 	 * acquire a transaction lock on the parent/root page
1839 	 */
1840 	tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
1841 	dtlck = (struct dt_lock *) & tlck->lock;
1842 	lv = & dtlck->lv[dtlck->index];
1843 
1844 	/* linelock parent entry - 1st slot */
1845 	lv->offset = 1;
1846 	lv->length = 1;
1847 	dtlck->index++;
1848 
1849 	/* update the parent pxd for page extension */
1850 	tpxd = (pxd_t *) & pp->slot[1];
1851 	*tpxd = *pxd;
1852 
1853 	DT_PUTPAGE(pmp);
1854 	return 0;
1855 }
1856 
1857 
1858 /*
1859  *	dtSplitRoot()
1860  *
1861  * function:
1862  *	split the full root page into
1863  *	original/root/split page and new right page
1864  *	i.e., root remains fixed in tree anchor (inode) and
1865  *	the root is copied to a single new right child page
1866  *	since root page << non-root page, and
1867  *	the split root page contains a single entry for the
1868  *	new right child page.
1869  *
1870  * parameter:
1871  *
1872  * return: 0 - success;
1873  *	   errno - failure;
1874  *	return new page pinned;
1875  */
1876 static int dtSplitRoot(tid_t tid,
1877 	    struct inode *ip, struct dtsplit * split, struct metapage ** rmpp)
1878 {
1879 	struct super_block *sb = ip->i_sb;
1880 	struct metapage *smp;
1881 	dtroot_t *sp;
1882 	struct metapage *rmp;
1883 	dtpage_t *rp;
1884 	s64 rbn;
1885 	int xlen;
1886 	int xsize;
1887 	struct dtslot *f;
1888 	s8 *stbl;
1889 	int fsi, stblsize, n;
1890 	struct idtentry *s;
1891 	pxd_t *ppxd;
1892 	struct pxdlist *pxdlist;
1893 	pxd_t *pxd;
1894 	struct dt_lock *dtlck;
1895 	struct tlock *tlck;
1896 	struct lv *lv;
1897 	int rc;
1898 
1899 	/* get split root page */
1900 	smp = split->mp;
1901 	sp = &JFS_IP(ip)->i_dtroot;
1902 
1903 	/*
1904 	 *	allocate/initialize a single (right) child page
1905 	 *
1906 	 * N.B. at first split, a one (or two) block to fit new entry
1907 	 * is allocated; at subsequent split, a full page is allocated;
1908 	 */
1909 	pxdlist = split->pxdlist;
1910 	pxd = &pxdlist->pxd[pxdlist->npxd];
1911 	pxdlist->npxd++;
1912 	rbn = addressPXD(pxd);
1913 	xlen = lengthPXD(pxd);
1914 	xsize = xlen << JFS_SBI(sb)->l2bsize;
1915 	rmp = get_metapage(ip, rbn, xsize, 1);
1916 	if (!rmp)
1917 		return -EIO;
1918 
1919 	rp = rmp->data;
1920 
1921 	/* Allocate blocks to quota. */
1922 	rc = dquot_alloc_block(ip, lengthPXD(pxd));
1923 	if (rc) {
1924 		release_metapage(rmp);
1925 		return rc;
1926 	}
1927 
1928 	BT_MARK_DIRTY(rmp, ip);
1929 	/*
1930 	 * acquire a transaction lock on the new right page
1931 	 */
1932 	tlck = txLock(tid, ip, rmp, tlckDTREE | tlckNEW);
1933 	dtlck = (struct dt_lock *) & tlck->lock;
1934 
1935 	rp->header.flag =
1936 	    (sp->header.flag & BT_LEAF) ? BT_LEAF : BT_INTERNAL;
1937 	rp->header.self = *pxd;
1938 
1939 	/* initialize sibling pointers */
1940 	rp->header.next = 0;
1941 	rp->header.prev = 0;
1942 
1943 	/*
1944 	 *	move in-line root page into new right page extent
1945 	 */
1946 	/* linelock header + copied entries + new stbl (1st slot) in new page */
1947 	ASSERT(dtlck->index == 0);
1948 	lv = & dtlck->lv[0];
1949 	lv->offset = 0;
1950 	lv->length = 10;	/* 1 + 8 + 1 */
1951 	dtlck->index++;
1952 
1953 	n = xsize >> L2DTSLOTSIZE;
1954 	rp->header.maxslot = n;
1955 	stblsize = (n + 31) >> L2DTSLOTSIZE;
1956 
1957 	/* copy old stbl to new stbl at start of extended area */
1958 	rp->header.stblindex = DTROOTMAXSLOT;
1959 	stbl = (s8 *) & rp->slot[DTROOTMAXSLOT];
1960 	memcpy(stbl, sp->header.stbl, sp->header.nextindex);
1961 	rp->header.nextindex = sp->header.nextindex;
1962 
1963 	/* copy old data area to start of new data area */
1964 	memcpy(&rp->slot[1], &sp->slot[1], IDATASIZE);
1965 
1966 	/*
1967 	 * append free region of newly extended area at tail of freelist
1968 	 */
1969 	/* init free region of newly extended area */
1970 	fsi = n = DTROOTMAXSLOT + stblsize;
1971 	f = &rp->slot[fsi];
1972 	for (fsi++; fsi < rp->header.maxslot; f++, fsi++)
1973 		f->next = fsi;
1974 	f->next = -1;
1975 
1976 	/* append new free region at tail of old freelist */
1977 	fsi = sp->header.freelist;
1978 	if (fsi == -1)
1979 		rp->header.freelist = n;
1980 	else {
1981 		rp->header.freelist = fsi;
1982 
1983 		do {
1984 			f = &rp->slot[fsi];
1985 			fsi = f->next;
1986 		} while (fsi != -1);
1987 
1988 		f->next = n;
1989 	}
1990 
1991 	rp->header.freecnt = sp->header.freecnt + rp->header.maxslot - n;
1992 
1993 	/*
1994 	 * Update directory index table for entries now in right page
1995 	 */
1996 	if ((rp->header.flag & BT_LEAF) && DO_INDEX(ip)) {
1997 		s64 lblock;
1998 		struct metapage *mp = NULL;
1999 		struct ldtentry *ldtentry;
2000 
2001 		stbl = DT_GETSTBL(rp);
2002 		for (n = 0; n < rp->header.nextindex; n++) {
2003 			ldtentry = (struct ldtentry *) & rp->slot[stbl[n]];
2004 			modify_index(tid, ip, le32_to_cpu(ldtentry->index),
2005 				     rbn, n, &mp, &lblock);
2006 		}
2007 		if (mp)
2008 			release_metapage(mp);
2009 	}
2010 	/*
2011 	 * insert the new entry into the new right/child page
2012 	 * (skip index in the new right page will not change)
2013 	 */
2014 	dtInsertEntry(rp, split->index, split->key, split->data, &dtlck);
2015 
2016 	/*
2017 	 *	reset parent/root page
2018 	 *
2019 	 * set the 1st entry offset to 0, which force the left-most key
2020 	 * at any level of the tree to be less than any search key.
2021 	 *
2022 	 * The btree comparison code guarantees that the left-most key on any
2023 	 * level of the tree is never used, so it doesn't need to be filled in.
2024 	 */
2025 	BT_MARK_DIRTY(smp, ip);
2026 	/*
2027 	 * acquire a transaction lock on the root page (in-memory inode)
2028 	 */
2029 	tlck = txLock(tid, ip, smp, tlckDTREE | tlckNEW | tlckBTROOT);
2030 	dtlck = (struct dt_lock *) & tlck->lock;
2031 
2032 	/* linelock root */
2033 	ASSERT(dtlck->index == 0);
2034 	lv = & dtlck->lv[0];
2035 	lv->offset = 0;
2036 	lv->length = DTROOTMAXSLOT;
2037 	dtlck->index++;
2038 
2039 	/* update page header of root */
2040 	if (sp->header.flag & BT_LEAF) {
2041 		sp->header.flag &= ~BT_LEAF;
2042 		sp->header.flag |= BT_INTERNAL;
2043 	}
2044 
2045 	/* init the first entry */
2046 	s = (struct idtentry *) & sp->slot[DTENTRYSTART];
2047 	ppxd = (pxd_t *) s;
2048 	*ppxd = *pxd;
2049 	s->next = -1;
2050 	s->namlen = 0;
2051 
2052 	stbl = sp->header.stbl;
2053 	stbl[0] = DTENTRYSTART;
2054 	sp->header.nextindex = 1;
2055 
2056 	/* init freelist */
2057 	fsi = DTENTRYSTART + 1;
2058 	f = &sp->slot[fsi];
2059 
2060 	/* init free region of remaining area */
2061 	for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2062 		f->next = fsi;
2063 	f->next = -1;
2064 
2065 	sp->header.freelist = DTENTRYSTART + 1;
2066 	sp->header.freecnt = DTROOTMAXSLOT - (DTENTRYSTART + 1);
2067 
2068 	*rmpp = rmp;
2069 
2070 	return 0;
2071 }
2072 
2073 
2074 /*
2075  *	dtDelete()
2076  *
2077  * function: delete the entry(s) referenced by a key.
2078  *
2079  * parameter:
2080  *
2081  * return:
2082  */
2083 int dtDelete(tid_t tid,
2084 	 struct inode *ip, struct component_name * key, ino_t * ino, int flag)
2085 {
2086 	int rc = 0;
2087 	s64 bn;
2088 	struct metapage *mp, *imp;
2089 	dtpage_t *p;
2090 	int index;
2091 	struct btstack btstack;
2092 	struct dt_lock *dtlck;
2093 	struct tlock *tlck;
2094 	struct lv *lv;
2095 	int i;
2096 	struct ldtentry *ldtentry;
2097 	u8 *stbl;
2098 	u32 table_index, next_index;
2099 	struct metapage *nmp;
2100 	dtpage_t *np;
2101 
2102 	/*
2103 	 *	search for the entry to delete:
2104 	 *
2105 	 * dtSearch() returns (leaf page pinned, index at which to delete).
2106 	 */
2107 	if ((rc = dtSearch(ip, key, ino, &btstack, flag)))
2108 		return rc;
2109 
2110 	/* retrieve search result */
2111 	DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
2112 
2113 	/*
2114 	 * We need to find put the index of the next entry into the
2115 	 * directory index table in order to resume a readdir from this
2116 	 * entry.
2117 	 */
2118 	if (DO_INDEX(ip)) {
2119 		stbl = DT_GETSTBL(p);
2120 		ldtentry = (struct ldtentry *) & p->slot[stbl[index]];
2121 		table_index = le32_to_cpu(ldtentry->index);
2122 		if (index == (p->header.nextindex - 1)) {
2123 			/*
2124 			 * Last entry in this leaf page
2125 			 */
2126 			if ((p->header.flag & BT_ROOT)
2127 			    || (p->header.next == 0))
2128 				next_index = -1;
2129 			else {
2130 				/* Read next leaf page */
2131 				DT_GETPAGE(ip, le64_to_cpu(p->header.next),
2132 					   nmp, PSIZE, np, rc);
2133 				if (rc)
2134 					next_index = -1;
2135 				else {
2136 					stbl = DT_GETSTBL(np);
2137 					ldtentry =
2138 					    (struct ldtentry *) & np->
2139 					    slot[stbl[0]];
2140 					next_index =
2141 					    le32_to_cpu(ldtentry->index);
2142 					DT_PUTPAGE(nmp);
2143 				}
2144 			}
2145 		} else {
2146 			ldtentry =
2147 			    (struct ldtentry *) & p->slot[stbl[index + 1]];
2148 			next_index = le32_to_cpu(ldtentry->index);
2149 		}
2150 		free_index(tid, ip, table_index, next_index);
2151 	}
2152 	/*
2153 	 * the leaf page becomes empty, delete the page
2154 	 */
2155 	if (p->header.nextindex == 1) {
2156 		/* delete empty page */
2157 		rc = dtDeleteUp(tid, ip, mp, p, &btstack);
2158 	}
2159 	/*
2160 	 * the leaf page has other entries remaining:
2161 	 *
2162 	 * delete the entry from the leaf page.
2163 	 */
2164 	else {
2165 		BT_MARK_DIRTY(mp, ip);
2166 		/*
2167 		 * acquire a transaction lock on the leaf page
2168 		 */
2169 		tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2170 		dtlck = (struct dt_lock *) & tlck->lock;
2171 
2172 		/*
2173 		 * Do not assume that dtlck->index will be zero.  During a
2174 		 * rename within a directory, this transaction may have
2175 		 * modified this page already when adding the new entry.
2176 		 */
2177 
2178 		/* linelock header */
2179 		if (dtlck->index >= dtlck->maxcnt)
2180 			dtlck = (struct dt_lock *) txLinelock(dtlck);
2181 		lv = & dtlck->lv[dtlck->index];
2182 		lv->offset = 0;
2183 		lv->length = 1;
2184 		dtlck->index++;
2185 
2186 		/* linelock stbl of non-root leaf page */
2187 		if (!(p->header.flag & BT_ROOT)) {
2188 			if (dtlck->index >= dtlck->maxcnt)
2189 				dtlck = (struct dt_lock *) txLinelock(dtlck);
2190 			lv = & dtlck->lv[dtlck->index];
2191 			i = index >> L2DTSLOTSIZE;
2192 			lv->offset = p->header.stblindex + i;
2193 			lv->length =
2194 			    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2195 			    i + 1;
2196 			dtlck->index++;
2197 		}
2198 
2199 		/* free the leaf entry */
2200 		dtDeleteEntry(p, index, &dtlck);
2201 
2202 		/*
2203 		 * Update directory index table for entries moved in stbl
2204 		 */
2205 		if (DO_INDEX(ip) && index < p->header.nextindex) {
2206 			s64 lblock;
2207 
2208 			imp = NULL;
2209 			stbl = DT_GETSTBL(p);
2210 			for (i = index; i < p->header.nextindex; i++) {
2211 				ldtentry =
2212 				    (struct ldtentry *) & p->slot[stbl[i]];
2213 				modify_index(tid, ip,
2214 					     le32_to_cpu(ldtentry->index),
2215 					     bn, i, &imp, &lblock);
2216 			}
2217 			if (imp)
2218 				release_metapage(imp);
2219 		}
2220 
2221 		DT_PUTPAGE(mp);
2222 	}
2223 
2224 	return rc;
2225 }
2226 
2227 
2228 /*
2229  *	dtDeleteUp()
2230  *
2231  * function:
2232  *	free empty pages as propagating deletion up the tree
2233  *
2234  * parameter:
2235  *
2236  * return:
2237  */
2238 static int dtDeleteUp(tid_t tid, struct inode *ip,
2239 	   struct metapage * fmp, dtpage_t * fp, struct btstack * btstack)
2240 {
2241 	int rc = 0;
2242 	struct metapage *mp;
2243 	dtpage_t *p;
2244 	int index, nextindex;
2245 	int xlen;
2246 	struct btframe *parent;
2247 	struct dt_lock *dtlck;
2248 	struct tlock *tlck;
2249 	struct lv *lv;
2250 	struct pxd_lock *pxdlock;
2251 	int i;
2252 
2253 	/*
2254 	 *	keep the root leaf page which has become empty
2255 	 */
2256 	if (BT_IS_ROOT(fmp)) {
2257 		/*
2258 		 * reset the root
2259 		 *
2260 		 * dtInitRoot() acquires txlock on the root
2261 		 */
2262 		dtInitRoot(tid, ip, PARENT(ip));
2263 
2264 		DT_PUTPAGE(fmp);
2265 
2266 		return 0;
2267 	}
2268 
2269 	/*
2270 	 *	free the non-root leaf page
2271 	 */
2272 	/*
2273 	 * acquire a transaction lock on the page
2274 	 *
2275 	 * write FREEXTENT|NOREDOPAGE log record
2276 	 * N.B. linelock is overlaid as freed extent descriptor, and
2277 	 * the buffer page is freed;
2278 	 */
2279 	tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2280 	pxdlock = (struct pxd_lock *) & tlck->lock;
2281 	pxdlock->flag = mlckFREEPXD;
2282 	pxdlock->pxd = fp->header.self;
2283 	pxdlock->index = 1;
2284 
2285 	/* update sibling pointers */
2286 	if ((rc = dtRelink(tid, ip, fp))) {
2287 		BT_PUTPAGE(fmp);
2288 		return rc;
2289 	}
2290 
2291 	xlen = lengthPXD(&fp->header.self);
2292 
2293 	/* Free quota allocation. */
2294 	dquot_free_block(ip, xlen);
2295 
2296 	/* free/invalidate its buffer page */
2297 	discard_metapage(fmp);
2298 
2299 	/*
2300 	 *	propagate page deletion up the directory tree
2301 	 *
2302 	 * If the delete from the parent page makes it empty,
2303 	 * continue all the way up the tree.
2304 	 * stop if the root page is reached (which is never deleted) or
2305 	 * if the entry deletion does not empty the page.
2306 	 */
2307 	while ((parent = BT_POP(btstack)) != NULL) {
2308 		/* pin the parent page <sp> */
2309 		DT_GETPAGE(ip, parent->bn, mp, PSIZE, p, rc);
2310 		if (rc)
2311 			return rc;
2312 
2313 		/*
2314 		 * free the extent of the child page deleted
2315 		 */
2316 		index = parent->index;
2317 
2318 		/*
2319 		 * delete the entry for the child page from parent
2320 		 */
2321 		nextindex = p->header.nextindex;
2322 
2323 		/*
2324 		 * the parent has the single entry being deleted:
2325 		 *
2326 		 * free the parent page which has become empty.
2327 		 */
2328 		if (nextindex == 1) {
2329 			/*
2330 			 * keep the root internal page which has become empty
2331 			 */
2332 			if (p->header.flag & BT_ROOT) {
2333 				/*
2334 				 * reset the root
2335 				 *
2336 				 * dtInitRoot() acquires txlock on the root
2337 				 */
2338 				dtInitRoot(tid, ip, PARENT(ip));
2339 
2340 				DT_PUTPAGE(mp);
2341 
2342 				return 0;
2343 			}
2344 			/*
2345 			 * free the parent page
2346 			 */
2347 			else {
2348 				/*
2349 				 * acquire a transaction lock on the page
2350 				 *
2351 				 * write FREEXTENT|NOREDOPAGE log record
2352 				 */
2353 				tlck =
2354 				    txMaplock(tid, ip,
2355 					      tlckDTREE | tlckFREE);
2356 				pxdlock = (struct pxd_lock *) & tlck->lock;
2357 				pxdlock->flag = mlckFREEPXD;
2358 				pxdlock->pxd = p->header.self;
2359 				pxdlock->index = 1;
2360 
2361 				/* update sibling pointers */
2362 				if ((rc = dtRelink(tid, ip, p))) {
2363 					DT_PUTPAGE(mp);
2364 					return rc;
2365 				}
2366 
2367 				xlen = lengthPXD(&p->header.self);
2368 
2369 				/* Free quota allocation */
2370 				dquot_free_block(ip, xlen);
2371 
2372 				/* free/invalidate its buffer page */
2373 				discard_metapage(mp);
2374 
2375 				/* propagate up */
2376 				continue;
2377 			}
2378 		}
2379 
2380 		/*
2381 		 * the parent has other entries remaining:
2382 		 *
2383 		 * delete the router entry from the parent page.
2384 		 */
2385 		BT_MARK_DIRTY(mp, ip);
2386 		/*
2387 		 * acquire a transaction lock on the page
2388 		 *
2389 		 * action: router entry deletion
2390 		 */
2391 		tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
2392 		dtlck = (struct dt_lock *) & tlck->lock;
2393 
2394 		/* linelock header */
2395 		if (dtlck->index >= dtlck->maxcnt)
2396 			dtlck = (struct dt_lock *) txLinelock(dtlck);
2397 		lv = & dtlck->lv[dtlck->index];
2398 		lv->offset = 0;
2399 		lv->length = 1;
2400 		dtlck->index++;
2401 
2402 		/* linelock stbl of non-root leaf page */
2403 		if (!(p->header.flag & BT_ROOT)) {
2404 			if (dtlck->index < dtlck->maxcnt)
2405 				lv++;
2406 			else {
2407 				dtlck = (struct dt_lock *) txLinelock(dtlck);
2408 				lv = & dtlck->lv[0];
2409 			}
2410 			i = index >> L2DTSLOTSIZE;
2411 			lv->offset = p->header.stblindex + i;
2412 			lv->length =
2413 			    ((p->header.nextindex - 1) >> L2DTSLOTSIZE) -
2414 			    i + 1;
2415 			dtlck->index++;
2416 		}
2417 
2418 		/* free the router entry */
2419 		dtDeleteEntry(p, index, &dtlck);
2420 
2421 		/* reset key of new leftmost entry of level (for consistency) */
2422 		if (index == 0 &&
2423 		    ((p->header.flag & BT_ROOT) || p->header.prev == 0))
2424 			dtTruncateEntry(p, 0, &dtlck);
2425 
2426 		/* unpin the parent page */
2427 		DT_PUTPAGE(mp);
2428 
2429 		/* exit propagation up */
2430 		break;
2431 	}
2432 
2433 	if (!DO_INDEX(ip))
2434 		ip->i_size -= PSIZE;
2435 
2436 	return 0;
2437 }
2438 
2439 #ifdef _NOTYET
2440 /*
2441  * NAME:	dtRelocate()
2442  *
2443  * FUNCTION:	relocate dtpage (internal or leaf) of directory;
2444  *		This function is mainly used by defragfs utility.
2445  */
2446 int dtRelocate(tid_t tid, struct inode *ip, s64 lmxaddr, pxd_t * opxd,
2447 	       s64 nxaddr)
2448 {
2449 	int rc = 0;
2450 	struct metapage *mp, *pmp, *lmp, *rmp;
2451 	dtpage_t *p, *pp, *rp = 0, *lp= 0;
2452 	s64 bn;
2453 	int index;
2454 	struct btstack btstack;
2455 	pxd_t *pxd;
2456 	s64 oxaddr, nextbn, prevbn;
2457 	int xlen, xsize;
2458 	struct tlock *tlck;
2459 	struct dt_lock *dtlck;
2460 	struct pxd_lock *pxdlock;
2461 	s8 *stbl;
2462 	struct lv *lv;
2463 
2464 	oxaddr = addressPXD(opxd);
2465 	xlen = lengthPXD(opxd);
2466 
2467 	jfs_info("dtRelocate: lmxaddr:%Ld xaddr:%Ld:%Ld xlen:%d",
2468 		   (long long)lmxaddr, (long long)oxaddr, (long long)nxaddr,
2469 		   xlen);
2470 
2471 	/*
2472 	 *	1. get the internal parent dtpage covering
2473 	 *	router entry for the tartget page to be relocated;
2474 	 */
2475 	rc = dtSearchNode(ip, lmxaddr, opxd, &btstack);
2476 	if (rc)
2477 		return rc;
2478 
2479 	/* retrieve search result */
2480 	DT_GETSEARCH(ip, btstack.top, bn, pmp, pp, index);
2481 	jfs_info("dtRelocate: parent router entry validated.");
2482 
2483 	/*
2484 	 *	2. relocate the target dtpage
2485 	 */
2486 	/* read in the target page from src extent */
2487 	DT_GETPAGE(ip, oxaddr, mp, PSIZE, p, rc);
2488 	if (rc) {
2489 		/* release the pinned parent page */
2490 		DT_PUTPAGE(pmp);
2491 		return rc;
2492 	}
2493 
2494 	/*
2495 	 * read in sibling pages if any to update sibling pointers;
2496 	 */
2497 	rmp = NULL;
2498 	if (p->header.next) {
2499 		nextbn = le64_to_cpu(p->header.next);
2500 		DT_GETPAGE(ip, nextbn, rmp, PSIZE, rp, rc);
2501 		if (rc) {
2502 			DT_PUTPAGE(mp);
2503 			DT_PUTPAGE(pmp);
2504 			return (rc);
2505 		}
2506 	}
2507 
2508 	lmp = NULL;
2509 	if (p->header.prev) {
2510 		prevbn = le64_to_cpu(p->header.prev);
2511 		DT_GETPAGE(ip, prevbn, lmp, PSIZE, lp, rc);
2512 		if (rc) {
2513 			DT_PUTPAGE(mp);
2514 			DT_PUTPAGE(pmp);
2515 			if (rmp)
2516 				DT_PUTPAGE(rmp);
2517 			return (rc);
2518 		}
2519 	}
2520 
2521 	/* at this point, all xtpages to be updated are in memory */
2522 
2523 	/*
2524 	 * update sibling pointers of sibling dtpages if any;
2525 	 */
2526 	if (lmp) {
2527 		tlck = txLock(tid, ip, lmp, tlckDTREE | tlckRELINK);
2528 		dtlck = (struct dt_lock *) & tlck->lock;
2529 		/* linelock header */
2530 		ASSERT(dtlck->index == 0);
2531 		lv = & dtlck->lv[0];
2532 		lv->offset = 0;
2533 		lv->length = 1;
2534 		dtlck->index++;
2535 
2536 		lp->header.next = cpu_to_le64(nxaddr);
2537 		DT_PUTPAGE(lmp);
2538 	}
2539 
2540 	if (rmp) {
2541 		tlck = txLock(tid, ip, rmp, tlckDTREE | tlckRELINK);
2542 		dtlck = (struct dt_lock *) & tlck->lock;
2543 		/* linelock header */
2544 		ASSERT(dtlck->index == 0);
2545 		lv = & dtlck->lv[0];
2546 		lv->offset = 0;
2547 		lv->length = 1;
2548 		dtlck->index++;
2549 
2550 		rp->header.prev = cpu_to_le64(nxaddr);
2551 		DT_PUTPAGE(rmp);
2552 	}
2553 
2554 	/*
2555 	 * update the target dtpage to be relocated
2556 	 *
2557 	 * write LOG_REDOPAGE of LOG_NEW type for dst page
2558 	 * for the whole target page (logredo() will apply
2559 	 * after image and update bmap for allocation of the
2560 	 * dst extent), and update bmap for allocation of
2561 	 * the dst extent;
2562 	 */
2563 	tlck = txLock(tid, ip, mp, tlckDTREE | tlckNEW);
2564 	dtlck = (struct dt_lock *) & tlck->lock;
2565 	/* linelock header */
2566 	ASSERT(dtlck->index == 0);
2567 	lv = & dtlck->lv[0];
2568 
2569 	/* update the self address in the dtpage header */
2570 	pxd = &p->header.self;
2571 	PXDaddress(pxd, nxaddr);
2572 
2573 	/* the dst page is the same as the src page, i.e.,
2574 	 * linelock for afterimage of the whole page;
2575 	 */
2576 	lv->offset = 0;
2577 	lv->length = p->header.maxslot;
2578 	dtlck->index++;
2579 
2580 	/* update the buffer extent descriptor of the dtpage */
2581 	xsize = xlen << JFS_SBI(ip->i_sb)->l2bsize;
2582 
2583 	/* unpin the relocated page */
2584 	DT_PUTPAGE(mp);
2585 	jfs_info("dtRelocate: target dtpage relocated.");
2586 
2587 	/* the moved extent is dtpage, then a LOG_NOREDOPAGE log rec
2588 	 * needs to be written (in logredo(), the LOG_NOREDOPAGE log rec
2589 	 * will also force a bmap update ).
2590 	 */
2591 
2592 	/*
2593 	 *	3. acquire maplock for the source extent to be freed;
2594 	 */
2595 	/* for dtpage relocation, write a LOG_NOREDOPAGE record
2596 	 * for the source dtpage (logredo() will init NoRedoPage
2597 	 * filter and will also update bmap for free of the source
2598 	 * dtpage), and upadte bmap for free of the source dtpage;
2599 	 */
2600 	tlck = txMaplock(tid, ip, tlckDTREE | tlckFREE);
2601 	pxdlock = (struct pxd_lock *) & tlck->lock;
2602 	pxdlock->flag = mlckFREEPXD;
2603 	PXDaddress(&pxdlock->pxd, oxaddr);
2604 	PXDlength(&pxdlock->pxd, xlen);
2605 	pxdlock->index = 1;
2606 
2607 	/*
2608 	 *	4. update the parent router entry for relocation;
2609 	 *
2610 	 * acquire tlck for the parent entry covering the target dtpage;
2611 	 * write LOG_REDOPAGE to apply after image only;
2612 	 */
2613 	jfs_info("dtRelocate: update parent router entry.");
2614 	tlck = txLock(tid, ip, pmp, tlckDTREE | tlckENTRY);
2615 	dtlck = (struct dt_lock *) & tlck->lock;
2616 	lv = & dtlck->lv[dtlck->index];
2617 
2618 	/* update the PXD with the new address */
2619 	stbl = DT_GETSTBL(pp);
2620 	pxd = (pxd_t *) & pp->slot[stbl[index]];
2621 	PXDaddress(pxd, nxaddr);
2622 	lv->offset = stbl[index];
2623 	lv->length = 1;
2624 	dtlck->index++;
2625 
2626 	/* unpin the parent dtpage */
2627 	DT_PUTPAGE(pmp);
2628 
2629 	return rc;
2630 }
2631 
2632 /*
2633  * NAME:	dtSearchNode()
2634  *
2635  * FUNCTION:	Search for an dtpage containing a specified address
2636  *		This function is mainly used by defragfs utility.
2637  *
2638  * NOTE:	Search result on stack, the found page is pinned at exit.
2639  *		The result page must be an internal dtpage.
2640  *		lmxaddr give the address of the left most page of the
2641  *		dtree level, in which the required dtpage resides.
2642  */
2643 static int dtSearchNode(struct inode *ip, s64 lmxaddr, pxd_t * kpxd,
2644 			struct btstack * btstack)
2645 {
2646 	int rc = 0;
2647 	s64 bn;
2648 	struct metapage *mp;
2649 	dtpage_t *p;
2650 	int psize = 288;	/* initial in-line directory */
2651 	s8 *stbl;
2652 	int i;
2653 	pxd_t *pxd;
2654 	struct btframe *btsp;
2655 
2656 	BT_CLR(btstack);	/* reset stack */
2657 
2658 	/*
2659 	 *	descend tree to the level with specified leftmost page
2660 	 *
2661 	 *  by convention, root bn = 0.
2662 	 */
2663 	for (bn = 0;;) {
2664 		/* get/pin the page to search */
2665 		DT_GETPAGE(ip, bn, mp, psize, p, rc);
2666 		if (rc)
2667 			return rc;
2668 
2669 		/* does the xaddr of leftmost page of the levevl
2670 		 * matches levevl search key ?
2671 		 */
2672 		if (p->header.flag & BT_ROOT) {
2673 			if (lmxaddr == 0)
2674 				break;
2675 		} else if (addressPXD(&p->header.self) == lmxaddr)
2676 			break;
2677 
2678 		/*
2679 		 * descend down to leftmost child page
2680 		 */
2681 		if (p->header.flag & BT_LEAF) {
2682 			DT_PUTPAGE(mp);
2683 			return -ESTALE;
2684 		}
2685 
2686 		/* get the leftmost entry */
2687 		stbl = DT_GETSTBL(p);
2688 		pxd = (pxd_t *) & p->slot[stbl[0]];
2689 
2690 		/* get the child page block address */
2691 		bn = addressPXD(pxd);
2692 		psize = lengthPXD(pxd) << JFS_SBI(ip->i_sb)->l2bsize;
2693 		/* unpin the parent page */
2694 		DT_PUTPAGE(mp);
2695 	}
2696 
2697 	/*
2698 	 *	search each page at the current levevl
2699 	 */
2700       loop:
2701 	stbl = DT_GETSTBL(p);
2702 	for (i = 0; i < p->header.nextindex; i++) {
2703 		pxd = (pxd_t *) & p->slot[stbl[i]];
2704 
2705 		/* found the specified router entry */
2706 		if (addressPXD(pxd) == addressPXD(kpxd) &&
2707 		    lengthPXD(pxd) == lengthPXD(kpxd)) {
2708 			btsp = btstack->top;
2709 			btsp->bn = bn;
2710 			btsp->index = i;
2711 			btsp->mp = mp;
2712 
2713 			return 0;
2714 		}
2715 	}
2716 
2717 	/* get the right sibling page if any */
2718 	if (p->header.next)
2719 		bn = le64_to_cpu(p->header.next);
2720 	else {
2721 		DT_PUTPAGE(mp);
2722 		return -ESTALE;
2723 	}
2724 
2725 	/* unpin current page */
2726 	DT_PUTPAGE(mp);
2727 
2728 	/* get the right sibling page */
2729 	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
2730 	if (rc)
2731 		return rc;
2732 
2733 	goto loop;
2734 }
2735 #endif /* _NOTYET */
2736 
2737 /*
2738  *	dtRelink()
2739  *
2740  * function:
2741  *	link around a freed page.
2742  *
2743  * parameter:
2744  *	fp:	page to be freed
2745  *
2746  * return:
2747  */
2748 static int dtRelink(tid_t tid, struct inode *ip, dtpage_t * p)
2749 {
2750 	int rc;
2751 	struct metapage *mp;
2752 	s64 nextbn, prevbn;
2753 	struct tlock *tlck;
2754 	struct dt_lock *dtlck;
2755 	struct lv *lv;
2756 
2757 	nextbn = le64_to_cpu(p->header.next);
2758 	prevbn = le64_to_cpu(p->header.prev);
2759 
2760 	/* update prev pointer of the next page */
2761 	if (nextbn != 0) {
2762 		DT_GETPAGE(ip, nextbn, mp, PSIZE, p, rc);
2763 		if (rc)
2764 			return rc;
2765 
2766 		BT_MARK_DIRTY(mp, ip);
2767 		/*
2768 		 * acquire a transaction lock on the next page
2769 		 *
2770 		 * action: update prev pointer;
2771 		 */
2772 		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2773 		jfs_info("dtRelink nextbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2774 			tlck, ip, mp);
2775 		dtlck = (struct dt_lock *) & tlck->lock;
2776 
2777 		/* linelock header */
2778 		if (dtlck->index >= dtlck->maxcnt)
2779 			dtlck = (struct dt_lock *) txLinelock(dtlck);
2780 		lv = & dtlck->lv[dtlck->index];
2781 		lv->offset = 0;
2782 		lv->length = 1;
2783 		dtlck->index++;
2784 
2785 		p->header.prev = cpu_to_le64(prevbn);
2786 		DT_PUTPAGE(mp);
2787 	}
2788 
2789 	/* update next pointer of the previous page */
2790 	if (prevbn != 0) {
2791 		DT_GETPAGE(ip, prevbn, mp, PSIZE, p, rc);
2792 		if (rc)
2793 			return rc;
2794 
2795 		BT_MARK_DIRTY(mp, ip);
2796 		/*
2797 		 * acquire a transaction lock on the prev page
2798 		 *
2799 		 * action: update next pointer;
2800 		 */
2801 		tlck = txLock(tid, ip, mp, tlckDTREE | tlckRELINK);
2802 		jfs_info("dtRelink prevbn: tlck = 0x%p, ip = 0x%p, mp=0x%p",
2803 			tlck, ip, mp);
2804 		dtlck = (struct dt_lock *) & tlck->lock;
2805 
2806 		/* linelock header */
2807 		if (dtlck->index >= dtlck->maxcnt)
2808 			dtlck = (struct dt_lock *) txLinelock(dtlck);
2809 		lv = & dtlck->lv[dtlck->index];
2810 		lv->offset = 0;
2811 		lv->length = 1;
2812 		dtlck->index++;
2813 
2814 		p->header.next = cpu_to_le64(nextbn);
2815 		DT_PUTPAGE(mp);
2816 	}
2817 
2818 	return 0;
2819 }
2820 
2821 
2822 /*
2823  *	dtInitRoot()
2824  *
2825  * initialize directory root (inline in inode)
2826  */
2827 void dtInitRoot(tid_t tid, struct inode *ip, u32 idotdot)
2828 {
2829 	struct jfs_inode_info *jfs_ip = JFS_IP(ip);
2830 	dtroot_t *p;
2831 	int fsi;
2832 	struct dtslot *f;
2833 	struct tlock *tlck;
2834 	struct dt_lock *dtlck;
2835 	struct lv *lv;
2836 	u16 xflag_save;
2837 
2838 	/*
2839 	 * If this was previously an non-empty directory, we need to remove
2840 	 * the old directory table.
2841 	 */
2842 	if (DO_INDEX(ip)) {
2843 		if (!jfs_dirtable_inline(ip)) {
2844 			struct tblock *tblk = tid_to_tblock(tid);
2845 			/*
2846 			 * We're playing games with the tid's xflag.  If
2847 			 * we're removing a regular file, the file's xtree
2848 			 * is committed with COMMIT_PMAP, but we always
2849 			 * commit the directories xtree with COMMIT_PWMAP.
2850 			 */
2851 			xflag_save = tblk->xflag;
2852 			tblk->xflag = 0;
2853 			/*
2854 			 * xtTruncate isn't guaranteed to fully truncate
2855 			 * the xtree.  The caller needs to check i_size
2856 			 * after committing the transaction to see if
2857 			 * additional truncation is needed.  The
2858 			 * COMMIT_Stale flag tells caller that we
2859 			 * initiated the truncation.
2860 			 */
2861 			xtTruncate(tid, ip, 0, COMMIT_PWMAP);
2862 			set_cflag(COMMIT_Stale, ip);
2863 
2864 			tblk->xflag = xflag_save;
2865 		} else
2866 			ip->i_size = 1;
2867 
2868 		jfs_ip->next_index = 2;
2869 	} else
2870 		ip->i_size = IDATASIZE;
2871 
2872 	/*
2873 	 * acquire a transaction lock on the root
2874 	 *
2875 	 * action: directory initialization;
2876 	 */
2877 	tlck = txLock(tid, ip, (struct metapage *) & jfs_ip->bxflag,
2878 		      tlckDTREE | tlckENTRY | tlckBTROOT);
2879 	dtlck = (struct dt_lock *) & tlck->lock;
2880 
2881 	/* linelock root */
2882 	ASSERT(dtlck->index == 0);
2883 	lv = & dtlck->lv[0];
2884 	lv->offset = 0;
2885 	lv->length = DTROOTMAXSLOT;
2886 	dtlck->index++;
2887 
2888 	p = &jfs_ip->i_dtroot;
2889 
2890 	p->header.flag = DXD_INDEX | BT_ROOT | BT_LEAF;
2891 
2892 	p->header.nextindex = 0;
2893 
2894 	/* init freelist */
2895 	fsi = 1;
2896 	f = &p->slot[fsi];
2897 
2898 	/* init data area of root */
2899 	for (fsi++; fsi < DTROOTMAXSLOT; f++, fsi++)
2900 		f->next = fsi;
2901 	f->next = -1;
2902 
2903 	p->header.freelist = 1;
2904 	p->header.freecnt = 8;
2905 
2906 	/* init '..' entry */
2907 	p->header.idotdot = cpu_to_le32(idotdot);
2908 
2909 	return;
2910 }
2911 
2912 /*
2913  *	add_missing_indices()
2914  *
2915  * function: Fix dtree page in which one or more entries has an invalid index.
2916  *	     fsck.jfs should really fix this, but it currently does not.
2917  *	     Called from jfs_readdir when bad index is detected.
2918  */
2919 static void add_missing_indices(struct inode *inode, s64 bn)
2920 {
2921 	struct ldtentry *d;
2922 	struct dt_lock *dtlck;
2923 	int i;
2924 	uint index;
2925 	struct lv *lv;
2926 	struct metapage *mp;
2927 	dtpage_t *p;
2928 	int rc;
2929 	s8 *stbl;
2930 	tid_t tid;
2931 	struct tlock *tlck;
2932 
2933 	tid = txBegin(inode->i_sb, 0);
2934 
2935 	DT_GETPAGE(inode, bn, mp, PSIZE, p, rc);
2936 
2937 	if (rc) {
2938 		printk(KERN_ERR "DT_GETPAGE failed!\n");
2939 		goto end;
2940 	}
2941 	BT_MARK_DIRTY(mp, inode);
2942 
2943 	ASSERT(p->header.flag & BT_LEAF);
2944 
2945 	tlck = txLock(tid, inode, mp, tlckDTREE | tlckENTRY);
2946 	if (BT_IS_ROOT(mp))
2947 		tlck->type |= tlckBTROOT;
2948 
2949 	dtlck = (struct dt_lock *) &tlck->lock;
2950 
2951 	stbl = DT_GETSTBL(p);
2952 	for (i = 0; i < p->header.nextindex; i++) {
2953 		d = (struct ldtentry *) &p->slot[stbl[i]];
2954 		index = le32_to_cpu(d->index);
2955 		if ((index < 2) || (index >= JFS_IP(inode)->next_index)) {
2956 			d->index = cpu_to_le32(add_index(tid, inode, bn, i));
2957 			if (dtlck->index >= dtlck->maxcnt)
2958 				dtlck = (struct dt_lock *) txLinelock(dtlck);
2959 			lv = &dtlck->lv[dtlck->index];
2960 			lv->offset = stbl[i];
2961 			lv->length = 1;
2962 			dtlck->index++;
2963 		}
2964 	}
2965 
2966 	DT_PUTPAGE(mp);
2967 	(void) txCommit(tid, 1, &inode, 0);
2968 end:
2969 	txEnd(tid);
2970 }
2971 
2972 /*
2973  * Buffer to hold directory entry info while traversing a dtree page
2974  * before being fed to the filldir function
2975  */
2976 struct jfs_dirent {
2977 	loff_t position;
2978 	int ino;
2979 	u16 name_len;
2980 	char name[0];
2981 };
2982 
2983 /*
2984  * function to determine next variable-sized jfs_dirent in buffer
2985  */
2986 static inline struct jfs_dirent *next_jfs_dirent(struct jfs_dirent *dirent)
2987 {
2988 	return (struct jfs_dirent *)
2989 		((char *)dirent +
2990 		 ((sizeof (struct jfs_dirent) + dirent->name_len + 1 +
2991 		   sizeof (loff_t) - 1) &
2992 		  ~(sizeof (loff_t) - 1)));
2993 }
2994 
2995 /*
2996  *	jfs_readdir()
2997  *
2998  * function: read directory entries sequentially
2999  *	from the specified entry offset
3000  *
3001  * parameter:
3002  *
3003  * return: offset = (pn, index) of start entry
3004  *	of next jfs_readdir()/dtRead()
3005  */
3006 int jfs_readdir(struct file *file, struct dir_context *ctx)
3007 {
3008 	struct inode *ip = file_inode(file);
3009 	struct nls_table *codepage = JFS_SBI(ip->i_sb)->nls_tab;
3010 	int rc = 0;
3011 	loff_t dtpos;	/* legacy OS/2 style position */
3012 	struct dtoffset {
3013 		s16 pn;
3014 		s16 index;
3015 		s32 unused;
3016 	} *dtoffset = (struct dtoffset *) &dtpos;
3017 	s64 bn;
3018 	struct metapage *mp;
3019 	dtpage_t *p;
3020 	int index;
3021 	s8 *stbl;
3022 	struct btstack btstack;
3023 	int i, next;
3024 	struct ldtentry *d;
3025 	struct dtslot *t;
3026 	int d_namleft, len, outlen;
3027 	unsigned long dirent_buf;
3028 	char *name_ptr;
3029 	u32 dir_index;
3030 	int do_index = 0;
3031 	uint loop_count = 0;
3032 	struct jfs_dirent *jfs_dirent;
3033 	int jfs_dirents;
3034 	int overflow, fix_page, page_fixed = 0;
3035 	static int unique_pos = 2;	/* If we can't fix broken index */
3036 
3037 	if (ctx->pos == DIREND)
3038 		return 0;
3039 
3040 	if (DO_INDEX(ip)) {
3041 		/*
3042 		 * persistent index is stored in directory entries.
3043 		 * Special cases:	 0 = .
3044 		 *			 1 = ..
3045 		 *			-1 = End of directory
3046 		 */
3047 		do_index = 1;
3048 
3049 		dir_index = (u32) ctx->pos;
3050 
3051 		/*
3052 		 * NFSv4 reserves cookies 1 and 2 for . and .. so the value
3053 		 * we return to the vfs is one greater than the one we use
3054 		 * internally.
3055 		 */
3056 		if (dir_index)
3057 			dir_index--;
3058 
3059 		if (dir_index > 1) {
3060 			struct dir_table_slot dirtab_slot;
3061 
3062 			if (dtEmpty(ip) ||
3063 			    (dir_index >= JFS_IP(ip)->next_index)) {
3064 				/* Stale position.  Directory has shrunk */
3065 				ctx->pos = DIREND;
3066 				return 0;
3067 			}
3068 		      repeat:
3069 			rc = read_index(ip, dir_index, &dirtab_slot);
3070 			if (rc) {
3071 				ctx->pos = DIREND;
3072 				return rc;
3073 			}
3074 			if (dirtab_slot.flag == DIR_INDEX_FREE) {
3075 				if (loop_count++ > JFS_IP(ip)->next_index) {
3076 					jfs_err("jfs_readdir detected infinite loop!");
3077 					ctx->pos = DIREND;
3078 					return 0;
3079 				}
3080 				dir_index = le32_to_cpu(dirtab_slot.addr2);
3081 				if (dir_index == -1) {
3082 					ctx->pos = DIREND;
3083 					return 0;
3084 				}
3085 				goto repeat;
3086 			}
3087 			bn = addressDTS(&dirtab_slot);
3088 			index = dirtab_slot.slot;
3089 			DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3090 			if (rc) {
3091 				ctx->pos = DIREND;
3092 				return 0;
3093 			}
3094 			if (p->header.flag & BT_INTERNAL) {
3095 				jfs_err("jfs_readdir: bad index table");
3096 				DT_PUTPAGE(mp);
3097 				ctx->pos = DIREND;
3098 				return 0;
3099 			}
3100 		} else {
3101 			if (dir_index == 0) {
3102 				/*
3103 				 * self "."
3104 				 */
3105 				ctx->pos = 1;
3106 				if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
3107 					return 0;
3108 			}
3109 			/*
3110 			 * parent ".."
3111 			 */
3112 			ctx->pos = 2;
3113 			if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
3114 				return 0;
3115 
3116 			/*
3117 			 * Find first entry of left-most leaf
3118 			 */
3119 			if (dtEmpty(ip)) {
3120 				ctx->pos = DIREND;
3121 				return 0;
3122 			}
3123 
3124 			if ((rc = dtReadFirst(ip, &btstack)))
3125 				return rc;
3126 
3127 			DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
3128 		}
3129 	} else {
3130 		/*
3131 		 * Legacy filesystem - OS/2 & Linux JFS < 0.3.6
3132 		 *
3133 		 * pn = 0; index = 1:	First entry "."
3134 		 * pn = 0; index = 2:	Second entry ".."
3135 		 * pn > 0:		Real entries, pn=1 -> leftmost page
3136 		 * pn = index = -1:	No more entries
3137 		 */
3138 		dtpos = ctx->pos;
3139 		if (dtpos < 2) {
3140 			/* build "." entry */
3141 			ctx->pos = 1;
3142 			if (!dir_emit(ctx, ".", 1, ip->i_ino, DT_DIR))
3143 				return 0;
3144 			dtoffset->index = 2;
3145 			ctx->pos = dtpos;
3146 		}
3147 
3148 		if (dtoffset->pn == 0) {
3149 			if (dtoffset->index == 2) {
3150 				/* build ".." entry */
3151 				if (!dir_emit(ctx, "..", 2, PARENT(ip), DT_DIR))
3152 					return 0;
3153 			} else {
3154 				jfs_err("jfs_readdir called with invalid offset!");
3155 			}
3156 			dtoffset->pn = 1;
3157 			dtoffset->index = 0;
3158 			ctx->pos = dtpos;
3159 		}
3160 
3161 		if (dtEmpty(ip)) {
3162 			ctx->pos = DIREND;
3163 			return 0;
3164 		}
3165 
3166 		if ((rc = dtReadNext(ip, &ctx->pos, &btstack))) {
3167 			jfs_err("jfs_readdir: unexpected rc = %d from dtReadNext",
3168 				rc);
3169 			ctx->pos = DIREND;
3170 			return 0;
3171 		}
3172 		/* get start leaf page and index */
3173 		DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
3174 
3175 		/* offset beyond directory eof ? */
3176 		if (bn < 0) {
3177 			ctx->pos = DIREND;
3178 			return 0;
3179 		}
3180 	}
3181 
3182 	dirent_buf = __get_free_page(GFP_KERNEL);
3183 	if (dirent_buf == 0) {
3184 		DT_PUTPAGE(mp);
3185 		jfs_warn("jfs_readdir: __get_free_page failed!");
3186 		ctx->pos = DIREND;
3187 		return -ENOMEM;
3188 	}
3189 
3190 	while (1) {
3191 		jfs_dirent = (struct jfs_dirent *) dirent_buf;
3192 		jfs_dirents = 0;
3193 		overflow = fix_page = 0;
3194 
3195 		stbl = DT_GETSTBL(p);
3196 
3197 		for (i = index; i < p->header.nextindex; i++) {
3198 			d = (struct ldtentry *) & p->slot[stbl[i]];
3199 
3200 			if (((long) jfs_dirent + d->namlen + 1) >
3201 			    (dirent_buf + PAGE_SIZE)) {
3202 				/* DBCS codepages could overrun dirent_buf */
3203 				index = i;
3204 				overflow = 1;
3205 				break;
3206 			}
3207 
3208 			d_namleft = d->namlen;
3209 			name_ptr = jfs_dirent->name;
3210 			jfs_dirent->ino = le32_to_cpu(d->inumber);
3211 
3212 			if (do_index) {
3213 				len = min(d_namleft, DTLHDRDATALEN);
3214 				jfs_dirent->position = le32_to_cpu(d->index);
3215 				/*
3216 				 * d->index should always be valid, but it
3217 				 * isn't.  fsck.jfs doesn't create the
3218 				 * directory index for the lost+found
3219 				 * directory.  Rather than let it go,
3220 				 * we can try to fix it.
3221 				 */
3222 				if ((jfs_dirent->position < 2) ||
3223 				    (jfs_dirent->position >=
3224 				     JFS_IP(ip)->next_index)) {
3225 					if (!page_fixed && !isReadOnly(ip)) {
3226 						fix_page = 1;
3227 						/*
3228 						 * setting overflow and setting
3229 						 * index to i will cause the
3230 						 * same page to be processed
3231 						 * again starting here
3232 						 */
3233 						overflow = 1;
3234 						index = i;
3235 						break;
3236 					}
3237 					jfs_dirent->position = unique_pos++;
3238 				}
3239 				/*
3240 				 * We add 1 to the index because we may
3241 				 * use a value of 2 internally, and NFSv4
3242 				 * doesn't like that.
3243 				 */
3244 				jfs_dirent->position++;
3245 			} else {
3246 				jfs_dirent->position = dtpos;
3247 				len = min(d_namleft, DTLHDRDATALEN_LEGACY);
3248 			}
3249 
3250 			/* copy the name of head/only segment */
3251 			outlen = jfs_strfromUCS_le(name_ptr, d->name, len,
3252 						   codepage);
3253 			jfs_dirent->name_len = outlen;
3254 
3255 			/* copy name in the additional segment(s) */
3256 			next = d->next;
3257 			while (next >= 0) {
3258 				t = (struct dtslot *) & p->slot[next];
3259 				name_ptr += outlen;
3260 				d_namleft -= len;
3261 				/* Sanity Check */
3262 				if (d_namleft == 0) {
3263 					jfs_error(ip->i_sb,
3264 						  "JFS:Dtree error: ino = %ld, bn=%lld, index = %d\n",
3265 						  (long)ip->i_ino,
3266 						  (long long)bn,
3267 						  i);
3268 					goto skip_one;
3269 				}
3270 				len = min(d_namleft, DTSLOTDATALEN);
3271 				outlen = jfs_strfromUCS_le(name_ptr, t->name,
3272 							   len, codepage);
3273 				jfs_dirent->name_len += outlen;
3274 
3275 				next = t->next;
3276 			}
3277 
3278 			jfs_dirents++;
3279 			jfs_dirent = next_jfs_dirent(jfs_dirent);
3280 skip_one:
3281 			if (!do_index)
3282 				dtoffset->index++;
3283 		}
3284 
3285 		if (!overflow) {
3286 			/* Point to next leaf page */
3287 			if (p->header.flag & BT_ROOT)
3288 				bn = 0;
3289 			else {
3290 				bn = le64_to_cpu(p->header.next);
3291 				index = 0;
3292 				/* update offset (pn:index) for new page */
3293 				if (!do_index) {
3294 					dtoffset->pn++;
3295 					dtoffset->index = 0;
3296 				}
3297 			}
3298 			page_fixed = 0;
3299 		}
3300 
3301 		/* unpin previous leaf page */
3302 		DT_PUTPAGE(mp);
3303 
3304 		jfs_dirent = (struct jfs_dirent *) dirent_buf;
3305 		while (jfs_dirents--) {
3306 			ctx->pos = jfs_dirent->position;
3307 			if (!dir_emit(ctx, jfs_dirent->name,
3308 				    jfs_dirent->name_len,
3309 				    jfs_dirent->ino, DT_UNKNOWN))
3310 				goto out;
3311 			jfs_dirent = next_jfs_dirent(jfs_dirent);
3312 		}
3313 
3314 		if (fix_page) {
3315 			add_missing_indices(ip, bn);
3316 			page_fixed = 1;
3317 		}
3318 
3319 		if (!overflow && (bn == 0)) {
3320 			ctx->pos = DIREND;
3321 			break;
3322 		}
3323 
3324 		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3325 		if (rc) {
3326 			free_page(dirent_buf);
3327 			return rc;
3328 		}
3329 	}
3330 
3331       out:
3332 	free_page(dirent_buf);
3333 
3334 	return rc;
3335 }
3336 
3337 
3338 /*
3339  *	dtReadFirst()
3340  *
3341  * function: get the leftmost page of the directory
3342  */
3343 static int dtReadFirst(struct inode *ip, struct btstack * btstack)
3344 {
3345 	int rc = 0;
3346 	s64 bn;
3347 	int psize = 288;	/* initial in-line directory */
3348 	struct metapage *mp;
3349 	dtpage_t *p;
3350 	s8 *stbl;
3351 	struct btframe *btsp;
3352 	pxd_t *xd;
3353 
3354 	BT_CLR(btstack);	/* reset stack */
3355 
3356 	/*
3357 	 *	descend leftmost path of the tree
3358 	 *
3359 	 * by convention, root bn = 0.
3360 	 */
3361 	for (bn = 0;;) {
3362 		DT_GETPAGE(ip, bn, mp, psize, p, rc);
3363 		if (rc)
3364 			return rc;
3365 
3366 		/*
3367 		 * leftmost leaf page
3368 		 */
3369 		if (p->header.flag & BT_LEAF) {
3370 			/* return leftmost entry */
3371 			btsp = btstack->top;
3372 			btsp->bn = bn;
3373 			btsp->index = 0;
3374 			btsp->mp = mp;
3375 
3376 			return 0;
3377 		}
3378 
3379 		/*
3380 		 * descend down to leftmost child page
3381 		 */
3382 		if (BT_STACK_FULL(btstack)) {
3383 			DT_PUTPAGE(mp);
3384 			jfs_error(ip->i_sb, "btstack overrun\n");
3385 			BT_STACK_DUMP(btstack);
3386 			return -EIO;
3387 		}
3388 		/* push (bn, index) of the parent page/entry */
3389 		BT_PUSH(btstack, bn, 0);
3390 
3391 		/* get the leftmost entry */
3392 		stbl = DT_GETSTBL(p);
3393 		xd = (pxd_t *) & p->slot[stbl[0]];
3394 
3395 		/* get the child page block address */
3396 		bn = addressPXD(xd);
3397 		psize = lengthPXD(xd) << JFS_SBI(ip->i_sb)->l2bsize;
3398 
3399 		/* unpin the parent page */
3400 		DT_PUTPAGE(mp);
3401 	}
3402 }
3403 
3404 
3405 /*
3406  *	dtReadNext()
3407  *
3408  * function: get the page of the specified offset (pn:index)
3409  *
3410  * return: if (offset > eof), bn = -1;
3411  *
3412  * note: if index > nextindex of the target leaf page,
3413  * start with 1st entry of next leaf page;
3414  */
3415 static int dtReadNext(struct inode *ip, loff_t * offset,
3416 		      struct btstack * btstack)
3417 {
3418 	int rc = 0;
3419 	struct dtoffset {
3420 		s16 pn;
3421 		s16 index;
3422 		s32 unused;
3423 	} *dtoffset = (struct dtoffset *) offset;
3424 	s64 bn;
3425 	struct metapage *mp;
3426 	dtpage_t *p;
3427 	int index;
3428 	int pn;
3429 	s8 *stbl;
3430 	struct btframe *btsp, *parent;
3431 	pxd_t *xd;
3432 
3433 	/*
3434 	 * get leftmost leaf page pinned
3435 	 */
3436 	if ((rc = dtReadFirst(ip, btstack)))
3437 		return rc;
3438 
3439 	/* get leaf page */
3440 	DT_GETSEARCH(ip, btstack->top, bn, mp, p, index);
3441 
3442 	/* get the start offset (pn:index) */
3443 	pn = dtoffset->pn - 1;	/* Now pn = 0 represents leftmost leaf */
3444 	index = dtoffset->index;
3445 
3446 	/* start at leftmost page ? */
3447 	if (pn == 0) {
3448 		/* offset beyond eof ? */
3449 		if (index < p->header.nextindex)
3450 			goto out;
3451 
3452 		if (p->header.flag & BT_ROOT) {
3453 			bn = -1;
3454 			goto out;
3455 		}
3456 
3457 		/* start with 1st entry of next leaf page */
3458 		dtoffset->pn++;
3459 		dtoffset->index = index = 0;
3460 		goto a;
3461 	}
3462 
3463 	/* start at non-leftmost page: scan parent pages for large pn */
3464 	if (p->header.flag & BT_ROOT) {
3465 		bn = -1;
3466 		goto out;
3467 	}
3468 
3469 	/* start after next leaf page ? */
3470 	if (pn > 1)
3471 		goto b;
3472 
3473 	/* get leaf page pn = 1 */
3474       a:
3475 	bn = le64_to_cpu(p->header.next);
3476 
3477 	/* unpin leaf page */
3478 	DT_PUTPAGE(mp);
3479 
3480 	/* offset beyond eof ? */
3481 	if (bn == 0) {
3482 		bn = -1;
3483 		goto out;
3484 	}
3485 
3486 	goto c;
3487 
3488 	/*
3489 	 * scan last internal page level to get target leaf page
3490 	 */
3491       b:
3492 	/* unpin leftmost leaf page */
3493 	DT_PUTPAGE(mp);
3494 
3495 	/* get left most parent page */
3496 	btsp = btstack->top;
3497 	parent = btsp - 1;
3498 	bn = parent->bn;
3499 	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3500 	if (rc)
3501 		return rc;
3502 
3503 	/* scan parent pages at last internal page level */
3504 	while (pn >= p->header.nextindex) {
3505 		pn -= p->header.nextindex;
3506 
3507 		/* get next parent page address */
3508 		bn = le64_to_cpu(p->header.next);
3509 
3510 		/* unpin current parent page */
3511 		DT_PUTPAGE(mp);
3512 
3513 		/* offset beyond eof ? */
3514 		if (bn == 0) {
3515 			bn = -1;
3516 			goto out;
3517 		}
3518 
3519 		/* get next parent page */
3520 		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3521 		if (rc)
3522 			return rc;
3523 
3524 		/* update parent page stack frame */
3525 		parent->bn = bn;
3526 	}
3527 
3528 	/* get leaf page address */
3529 	stbl = DT_GETSTBL(p);
3530 	xd = (pxd_t *) & p->slot[stbl[pn]];
3531 	bn = addressPXD(xd);
3532 
3533 	/* unpin parent page */
3534 	DT_PUTPAGE(mp);
3535 
3536 	/*
3537 	 * get target leaf page
3538 	 */
3539       c:
3540 	DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3541 	if (rc)
3542 		return rc;
3543 
3544 	/*
3545 	 * leaf page has been completed:
3546 	 * start with 1st entry of next leaf page
3547 	 */
3548 	if (index >= p->header.nextindex) {
3549 		bn = le64_to_cpu(p->header.next);
3550 
3551 		/* unpin leaf page */
3552 		DT_PUTPAGE(mp);
3553 
3554 		/* offset beyond eof ? */
3555 		if (bn == 0) {
3556 			bn = -1;
3557 			goto out;
3558 		}
3559 
3560 		/* get next leaf page */
3561 		DT_GETPAGE(ip, bn, mp, PSIZE, p, rc);
3562 		if (rc)
3563 			return rc;
3564 
3565 		/* start with 1st entry of next leaf page */
3566 		dtoffset->pn++;
3567 		dtoffset->index = 0;
3568 	}
3569 
3570       out:
3571 	/* return target leaf page pinned */
3572 	btsp = btstack->top;
3573 	btsp->bn = bn;
3574 	btsp->index = dtoffset->index;
3575 	btsp->mp = mp;
3576 
3577 	return 0;
3578 }
3579 
3580 
3581 /*
3582  *	dtCompare()
3583  *
3584  * function: compare search key with an internal entry
3585  *
3586  * return:
3587  *	< 0 if k is < record
3588  *	= 0 if k is = record
3589  *	> 0 if k is > record
3590  */
3591 static int dtCompare(struct component_name * key,	/* search key */
3592 		     dtpage_t * p,	/* directory page */
3593 		     int si)
3594 {				/* entry slot index */
3595 	wchar_t *kname;
3596 	__le16 *name;
3597 	int klen, namlen, len, rc;
3598 	struct idtentry *ih;
3599 	struct dtslot *t;
3600 
3601 	/*
3602 	 * force the left-most key on internal pages, at any level of
3603 	 * the tree, to be less than any search key.
3604 	 * this obviates having to update the leftmost key on an internal
3605 	 * page when the user inserts a new key in the tree smaller than
3606 	 * anything that has been stored.
3607 	 *
3608 	 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3609 	 * at any internal page at any level of the tree,
3610 	 * it descends to child of the entry anyway -
3611 	 * ? make the entry as min size dummy entry)
3612 	 *
3613 	 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3614 	 * return (1);
3615 	 */
3616 
3617 	kname = key->name;
3618 	klen = key->namlen;
3619 
3620 	ih = (struct idtentry *) & p->slot[si];
3621 	si = ih->next;
3622 	name = ih->name;
3623 	namlen = ih->namlen;
3624 	len = min(namlen, DTIHDRDATALEN);
3625 
3626 	/* compare with head/only segment */
3627 	len = min(klen, len);
3628 	if ((rc = UniStrncmp_le(kname, name, len)))
3629 		return rc;
3630 
3631 	klen -= len;
3632 	namlen -= len;
3633 
3634 	/* compare with additional segment(s) */
3635 	kname += len;
3636 	while (klen > 0 && namlen > 0) {
3637 		/* compare with next name segment */
3638 		t = (struct dtslot *) & p->slot[si];
3639 		len = min(namlen, DTSLOTDATALEN);
3640 		len = min(klen, len);
3641 		name = t->name;
3642 		if ((rc = UniStrncmp_le(kname, name, len)))
3643 			return rc;
3644 
3645 		klen -= len;
3646 		namlen -= len;
3647 		kname += len;
3648 		si = t->next;
3649 	}
3650 
3651 	return (klen - namlen);
3652 }
3653 
3654 
3655 
3656 
3657 /*
3658  *	ciCompare()
3659  *
3660  * function: compare search key with an (leaf/internal) entry
3661  *
3662  * return:
3663  *	< 0 if k is < record
3664  *	= 0 if k is = record
3665  *	> 0 if k is > record
3666  */
3667 static int ciCompare(struct component_name * key,	/* search key */
3668 		     dtpage_t * p,	/* directory page */
3669 		     int si,	/* entry slot index */
3670 		     int flag)
3671 {
3672 	wchar_t *kname, x;
3673 	__le16 *name;
3674 	int klen, namlen, len, rc;
3675 	struct ldtentry *lh;
3676 	struct idtentry *ih;
3677 	struct dtslot *t;
3678 	int i;
3679 
3680 	/*
3681 	 * force the left-most key on internal pages, at any level of
3682 	 * the tree, to be less than any search key.
3683 	 * this obviates having to update the leftmost key on an internal
3684 	 * page when the user inserts a new key in the tree smaller than
3685 	 * anything that has been stored.
3686 	 *
3687 	 * (? if/when dtSearch() narrows down to 1st entry (index = 0),
3688 	 * at any internal page at any level of the tree,
3689 	 * it descends to child of the entry anyway -
3690 	 * ? make the entry as min size dummy entry)
3691 	 *
3692 	 * if (e->index == 0 && h->prevpg == P_INVALID && !(h->flags & BT_LEAF))
3693 	 * return (1);
3694 	 */
3695 
3696 	kname = key->name;
3697 	klen = key->namlen;
3698 
3699 	/*
3700 	 * leaf page entry
3701 	 */
3702 	if (p->header.flag & BT_LEAF) {
3703 		lh = (struct ldtentry *) & p->slot[si];
3704 		si = lh->next;
3705 		name = lh->name;
3706 		namlen = lh->namlen;
3707 		if (flag & JFS_DIR_INDEX)
3708 			len = min(namlen, DTLHDRDATALEN);
3709 		else
3710 			len = min(namlen, DTLHDRDATALEN_LEGACY);
3711 	}
3712 	/*
3713 	 * internal page entry
3714 	 */
3715 	else {
3716 		ih = (struct idtentry *) & p->slot[si];
3717 		si = ih->next;
3718 		name = ih->name;
3719 		namlen = ih->namlen;
3720 		len = min(namlen, DTIHDRDATALEN);
3721 	}
3722 
3723 	/* compare with head/only segment */
3724 	len = min(klen, len);
3725 	for (i = 0; i < len; i++, kname++, name++) {
3726 		/* only uppercase if case-insensitive support is on */
3727 		if ((flag & JFS_OS2) == JFS_OS2)
3728 			x = UniToupper(le16_to_cpu(*name));
3729 		else
3730 			x = le16_to_cpu(*name);
3731 		if ((rc = *kname - x))
3732 			return rc;
3733 	}
3734 
3735 	klen -= len;
3736 	namlen -= len;
3737 
3738 	/* compare with additional segment(s) */
3739 	while (klen > 0 && namlen > 0) {
3740 		/* compare with next name segment */
3741 		t = (struct dtslot *) & p->slot[si];
3742 		len = min(namlen, DTSLOTDATALEN);
3743 		len = min(klen, len);
3744 		name = t->name;
3745 		for (i = 0; i < len; i++, kname++, name++) {
3746 			/* only uppercase if case-insensitive support is on */
3747 			if ((flag & JFS_OS2) == JFS_OS2)
3748 				x = UniToupper(le16_to_cpu(*name));
3749 			else
3750 				x = le16_to_cpu(*name);
3751 
3752 			if ((rc = *kname - x))
3753 				return rc;
3754 		}
3755 
3756 		klen -= len;
3757 		namlen -= len;
3758 		si = t->next;
3759 	}
3760 
3761 	return (klen - namlen);
3762 }
3763 
3764 
3765 /*
3766  *	ciGetLeafPrefixKey()
3767  *
3768  * function: compute prefix of suffix compression
3769  *	     from two adjacent leaf entries
3770  *	     across page boundary
3771  *
3772  * return: non-zero on error
3773  *
3774  */
3775 static int ciGetLeafPrefixKey(dtpage_t * lp, int li, dtpage_t * rp,
3776 			       int ri, struct component_name * key, int flag)
3777 {
3778 	int klen, namlen;
3779 	wchar_t *pl, *pr, *kname;
3780 	struct component_name lkey;
3781 	struct component_name rkey;
3782 
3783 	lkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3784 					GFP_KERNEL);
3785 	if (lkey.name == NULL)
3786 		return -ENOMEM;
3787 
3788 	rkey.name = kmalloc_array(JFS_NAME_MAX + 1, sizeof(wchar_t),
3789 					GFP_KERNEL);
3790 	if (rkey.name == NULL) {
3791 		kfree(lkey.name);
3792 		return -ENOMEM;
3793 	}
3794 
3795 	/* get left and right key */
3796 	dtGetKey(lp, li, &lkey, flag);
3797 	lkey.name[lkey.namlen] = 0;
3798 
3799 	if ((flag & JFS_OS2) == JFS_OS2)
3800 		ciToUpper(&lkey);
3801 
3802 	dtGetKey(rp, ri, &rkey, flag);
3803 	rkey.name[rkey.namlen] = 0;
3804 
3805 
3806 	if ((flag & JFS_OS2) == JFS_OS2)
3807 		ciToUpper(&rkey);
3808 
3809 	/* compute prefix */
3810 	klen = 0;
3811 	kname = key->name;
3812 	namlen = min(lkey.namlen, rkey.namlen);
3813 	for (pl = lkey.name, pr = rkey.name;
3814 	     namlen; pl++, pr++, namlen--, klen++, kname++) {
3815 		*kname = *pr;
3816 		if (*pl != *pr) {
3817 			key->namlen = klen + 1;
3818 			goto free_names;
3819 		}
3820 	}
3821 
3822 	/* l->namlen <= r->namlen since l <= r */
3823 	if (lkey.namlen < rkey.namlen) {
3824 		*kname = *pr;
3825 		key->namlen = klen + 1;
3826 	} else			/* l->namelen == r->namelen */
3827 		key->namlen = klen;
3828 
3829 free_names:
3830 	kfree(lkey.name);
3831 	kfree(rkey.name);
3832 	return 0;
3833 }
3834 
3835 
3836 
3837 /*
3838  *	dtGetKey()
3839  *
3840  * function: get key of the entry
3841  */
3842 static void dtGetKey(dtpage_t * p, int i,	/* entry index */
3843 		     struct component_name * key, int flag)
3844 {
3845 	int si;
3846 	s8 *stbl;
3847 	struct ldtentry *lh;
3848 	struct idtentry *ih;
3849 	struct dtslot *t;
3850 	int namlen, len;
3851 	wchar_t *kname;
3852 	__le16 *name;
3853 
3854 	/* get entry */
3855 	stbl = DT_GETSTBL(p);
3856 	si = stbl[i];
3857 	if (p->header.flag & BT_LEAF) {
3858 		lh = (struct ldtentry *) & p->slot[si];
3859 		si = lh->next;
3860 		namlen = lh->namlen;
3861 		name = lh->name;
3862 		if (flag & JFS_DIR_INDEX)
3863 			len = min(namlen, DTLHDRDATALEN);
3864 		else
3865 			len = min(namlen, DTLHDRDATALEN_LEGACY);
3866 	} else {
3867 		ih = (struct idtentry *) & p->slot[si];
3868 		si = ih->next;
3869 		namlen = ih->namlen;
3870 		name = ih->name;
3871 		len = min(namlen, DTIHDRDATALEN);
3872 	}
3873 
3874 	key->namlen = namlen;
3875 	kname = key->name;
3876 
3877 	/*
3878 	 * move head/only segment
3879 	 */
3880 	UniStrncpy_from_le(kname, name, len);
3881 
3882 	/*
3883 	 * move additional segment(s)
3884 	 */
3885 	while (si >= 0) {
3886 		/* get next segment */
3887 		t = &p->slot[si];
3888 		kname += len;
3889 		namlen -= len;
3890 		len = min(namlen, DTSLOTDATALEN);
3891 		UniStrncpy_from_le(kname, t->name, len);
3892 
3893 		si = t->next;
3894 	}
3895 }
3896 
3897 
3898 /*
3899  *	dtInsertEntry()
3900  *
3901  * function: allocate free slot(s) and
3902  *	     write a leaf/internal entry
3903  *
3904  * return: entry slot index
3905  */
3906 static void dtInsertEntry(dtpage_t * p, int index, struct component_name * key,
3907 			  ddata_t * data, struct dt_lock ** dtlock)
3908 {
3909 	struct dtslot *h, *t;
3910 	struct ldtentry *lh = NULL;
3911 	struct idtentry *ih = NULL;
3912 	int hsi, fsi, klen, len, nextindex;
3913 	wchar_t *kname;
3914 	__le16 *name;
3915 	s8 *stbl;
3916 	pxd_t *xd;
3917 	struct dt_lock *dtlck = *dtlock;
3918 	struct lv *lv;
3919 	int xsi, n;
3920 	s64 bn = 0;
3921 	struct metapage *mp = NULL;
3922 
3923 	klen = key->namlen;
3924 	kname = key->name;
3925 
3926 	/* allocate a free slot */
3927 	hsi = fsi = p->header.freelist;
3928 	h = &p->slot[fsi];
3929 	p->header.freelist = h->next;
3930 	--p->header.freecnt;
3931 
3932 	/* open new linelock */
3933 	if (dtlck->index >= dtlck->maxcnt)
3934 		dtlck = (struct dt_lock *) txLinelock(dtlck);
3935 
3936 	lv = & dtlck->lv[dtlck->index];
3937 	lv->offset = hsi;
3938 
3939 	/* write head/only segment */
3940 	if (p->header.flag & BT_LEAF) {
3941 		lh = (struct ldtentry *) h;
3942 		lh->next = h->next;
3943 		lh->inumber = cpu_to_le32(data->leaf.ino);
3944 		lh->namlen = klen;
3945 		name = lh->name;
3946 		if (data->leaf.ip) {
3947 			len = min(klen, DTLHDRDATALEN);
3948 			if (!(p->header.flag & BT_ROOT))
3949 				bn = addressPXD(&p->header.self);
3950 			lh->index = cpu_to_le32(add_index(data->leaf.tid,
3951 							  data->leaf.ip,
3952 							  bn, index));
3953 		} else
3954 			len = min(klen, DTLHDRDATALEN_LEGACY);
3955 	} else {
3956 		ih = (struct idtentry *) h;
3957 		ih->next = h->next;
3958 		xd = (pxd_t *) ih;
3959 		*xd = data->xd;
3960 		ih->namlen = klen;
3961 		name = ih->name;
3962 		len = min(klen, DTIHDRDATALEN);
3963 	}
3964 
3965 	UniStrncpy_to_le(name, kname, len);
3966 
3967 	n = 1;
3968 	xsi = hsi;
3969 
3970 	/* write additional segment(s) */
3971 	t = h;
3972 	klen -= len;
3973 	while (klen) {
3974 		/* get free slot */
3975 		fsi = p->header.freelist;
3976 		t = &p->slot[fsi];
3977 		p->header.freelist = t->next;
3978 		--p->header.freecnt;
3979 
3980 		/* is next slot contiguous ? */
3981 		if (fsi != xsi + 1) {
3982 			/* close current linelock */
3983 			lv->length = n;
3984 			dtlck->index++;
3985 
3986 			/* open new linelock */
3987 			if (dtlck->index < dtlck->maxcnt)
3988 				lv++;
3989 			else {
3990 				dtlck = (struct dt_lock *) txLinelock(dtlck);
3991 				lv = & dtlck->lv[0];
3992 			}
3993 
3994 			lv->offset = fsi;
3995 			n = 0;
3996 		}
3997 
3998 		kname += len;
3999 		len = min(klen, DTSLOTDATALEN);
4000 		UniStrncpy_to_le(t->name, kname, len);
4001 
4002 		n++;
4003 		xsi = fsi;
4004 		klen -= len;
4005 	}
4006 
4007 	/* close current linelock */
4008 	lv->length = n;
4009 	dtlck->index++;
4010 
4011 	*dtlock = dtlck;
4012 
4013 	/* terminate last/only segment */
4014 	if (h == t) {
4015 		/* single segment entry */
4016 		if (p->header.flag & BT_LEAF)
4017 			lh->next = -1;
4018 		else
4019 			ih->next = -1;
4020 	} else
4021 		/* multi-segment entry */
4022 		t->next = -1;
4023 
4024 	/* if insert into middle, shift right succeeding entries in stbl */
4025 	stbl = DT_GETSTBL(p);
4026 	nextindex = p->header.nextindex;
4027 	if (index < nextindex) {
4028 		memmove(stbl + index + 1, stbl + index, nextindex - index);
4029 
4030 		if ((p->header.flag & BT_LEAF) && data->leaf.ip) {
4031 			s64 lblock;
4032 
4033 			/*
4034 			 * Need to update slot number for entries that moved
4035 			 * in the stbl
4036 			 */
4037 			mp = NULL;
4038 			for (n = index + 1; n <= nextindex; n++) {
4039 				lh = (struct ldtentry *) & (p->slot[stbl[n]]);
4040 				modify_index(data->leaf.tid, data->leaf.ip,
4041 					     le32_to_cpu(lh->index), bn, n,
4042 					     &mp, &lblock);
4043 			}
4044 			if (mp)
4045 				release_metapage(mp);
4046 		}
4047 	}
4048 
4049 	stbl[index] = hsi;
4050 
4051 	/* advance next available entry index of stbl */
4052 	++p->header.nextindex;
4053 }
4054 
4055 
4056 /*
4057  *	dtMoveEntry()
4058  *
4059  * function: move entries from split/left page to new/right page
4060  *
4061  *	nextindex of dst page and freelist/freecnt of both pages
4062  *	are updated.
4063  */
4064 static void dtMoveEntry(dtpage_t * sp, int si, dtpage_t * dp,
4065 			struct dt_lock ** sdtlock, struct dt_lock ** ddtlock,
4066 			int do_index)
4067 {
4068 	int ssi, next;		/* src slot index */
4069 	int di;			/* dst entry index */
4070 	int dsi;		/* dst slot index */
4071 	s8 *sstbl, *dstbl;	/* sorted entry table */
4072 	int snamlen, len;
4073 	struct ldtentry *slh, *dlh = NULL;
4074 	struct idtentry *sih, *dih = NULL;
4075 	struct dtslot *h, *s, *d;
4076 	struct dt_lock *sdtlck = *sdtlock, *ddtlck = *ddtlock;
4077 	struct lv *slv, *dlv;
4078 	int xssi, ns, nd;
4079 	int sfsi;
4080 
4081 	sstbl = (s8 *) & sp->slot[sp->header.stblindex];
4082 	dstbl = (s8 *) & dp->slot[dp->header.stblindex];
4083 
4084 	dsi = dp->header.freelist;	/* first (whole page) free slot */
4085 	sfsi = sp->header.freelist;
4086 
4087 	/* linelock destination entry slot */
4088 	dlv = & ddtlck->lv[ddtlck->index];
4089 	dlv->offset = dsi;
4090 
4091 	/* linelock source entry slot */
4092 	slv = & sdtlck->lv[sdtlck->index];
4093 	slv->offset = sstbl[si];
4094 	xssi = slv->offset - 1;
4095 
4096 	/*
4097 	 * move entries
4098 	 */
4099 	ns = nd = 0;
4100 	for (di = 0; si < sp->header.nextindex; si++, di++) {
4101 		ssi = sstbl[si];
4102 		dstbl[di] = dsi;
4103 
4104 		/* is next slot contiguous ? */
4105 		if (ssi != xssi + 1) {
4106 			/* close current linelock */
4107 			slv->length = ns;
4108 			sdtlck->index++;
4109 
4110 			/* open new linelock */
4111 			if (sdtlck->index < sdtlck->maxcnt)
4112 				slv++;
4113 			else {
4114 				sdtlck = (struct dt_lock *) txLinelock(sdtlck);
4115 				slv = & sdtlck->lv[0];
4116 			}
4117 
4118 			slv->offset = ssi;
4119 			ns = 0;
4120 		}
4121 
4122 		/*
4123 		 * move head/only segment of an entry
4124 		 */
4125 		/* get dst slot */
4126 		h = d = &dp->slot[dsi];
4127 
4128 		/* get src slot and move */
4129 		s = &sp->slot[ssi];
4130 		if (sp->header.flag & BT_LEAF) {
4131 			/* get source entry */
4132 			slh = (struct ldtentry *) s;
4133 			dlh = (struct ldtentry *) h;
4134 			snamlen = slh->namlen;
4135 
4136 			if (do_index) {
4137 				len = min(snamlen, DTLHDRDATALEN);
4138 				dlh->index = slh->index; /* little-endian */
4139 			} else
4140 				len = min(snamlen, DTLHDRDATALEN_LEGACY);
4141 
4142 			memcpy(dlh, slh, 6 + len * 2);
4143 
4144 			next = slh->next;
4145 
4146 			/* update dst head/only segment next field */
4147 			dsi++;
4148 			dlh->next = dsi;
4149 		} else {
4150 			sih = (struct idtentry *) s;
4151 			snamlen = sih->namlen;
4152 
4153 			len = min(snamlen, DTIHDRDATALEN);
4154 			dih = (struct idtentry *) h;
4155 			memcpy(dih, sih, 10 + len * 2);
4156 			next = sih->next;
4157 
4158 			dsi++;
4159 			dih->next = dsi;
4160 		}
4161 
4162 		/* free src head/only segment */
4163 		s->next = sfsi;
4164 		s->cnt = 1;
4165 		sfsi = ssi;
4166 
4167 		ns++;
4168 		nd++;
4169 		xssi = ssi;
4170 
4171 		/*
4172 		 * move additional segment(s) of the entry
4173 		 */
4174 		snamlen -= len;
4175 		while ((ssi = next) >= 0) {
4176 			/* is next slot contiguous ? */
4177 			if (ssi != xssi + 1) {
4178 				/* close current linelock */
4179 				slv->length = ns;
4180 				sdtlck->index++;
4181 
4182 				/* open new linelock */
4183 				if (sdtlck->index < sdtlck->maxcnt)
4184 					slv++;
4185 				else {
4186 					sdtlck =
4187 					    (struct dt_lock *)
4188 					    txLinelock(sdtlck);
4189 					slv = & sdtlck->lv[0];
4190 				}
4191 
4192 				slv->offset = ssi;
4193 				ns = 0;
4194 			}
4195 
4196 			/* get next source segment */
4197 			s = &sp->slot[ssi];
4198 
4199 			/* get next destination free slot */
4200 			d++;
4201 
4202 			len = min(snamlen, DTSLOTDATALEN);
4203 			UniStrncpy_le(d->name, s->name, len);
4204 
4205 			ns++;
4206 			nd++;
4207 			xssi = ssi;
4208 
4209 			dsi++;
4210 			d->next = dsi;
4211 
4212 			/* free source segment */
4213 			next = s->next;
4214 			s->next = sfsi;
4215 			s->cnt = 1;
4216 			sfsi = ssi;
4217 
4218 			snamlen -= len;
4219 		}		/* end while */
4220 
4221 		/* terminate dst last/only segment */
4222 		if (h == d) {
4223 			/* single segment entry */
4224 			if (dp->header.flag & BT_LEAF)
4225 				dlh->next = -1;
4226 			else
4227 				dih->next = -1;
4228 		} else
4229 			/* multi-segment entry */
4230 			d->next = -1;
4231 	}			/* end for */
4232 
4233 	/* close current linelock */
4234 	slv->length = ns;
4235 	sdtlck->index++;
4236 	*sdtlock = sdtlck;
4237 
4238 	dlv->length = nd;
4239 	ddtlck->index++;
4240 	*ddtlock = ddtlck;
4241 
4242 	/* update source header */
4243 	sp->header.freelist = sfsi;
4244 	sp->header.freecnt += nd;
4245 
4246 	/* update destination header */
4247 	dp->header.nextindex = di;
4248 
4249 	dp->header.freelist = dsi;
4250 	dp->header.freecnt -= nd;
4251 }
4252 
4253 
4254 /*
4255  *	dtDeleteEntry()
4256  *
4257  * function: free a (leaf/internal) entry
4258  *
4259  * log freelist header, stbl, and each segment slot of entry
4260  * (even though last/only segment next field is modified,
4261  * physical image logging requires all segment slots of
4262  * the entry logged to avoid applying previous updates
4263  * to the same slots)
4264  */
4265 static void dtDeleteEntry(dtpage_t * p, int fi, struct dt_lock ** dtlock)
4266 {
4267 	int fsi;		/* free entry slot index */
4268 	s8 *stbl;
4269 	struct dtslot *t;
4270 	int si, freecnt;
4271 	struct dt_lock *dtlck = *dtlock;
4272 	struct lv *lv;
4273 	int xsi, n;
4274 
4275 	/* get free entry slot index */
4276 	stbl = DT_GETSTBL(p);
4277 	fsi = stbl[fi];
4278 
4279 	/* open new linelock */
4280 	if (dtlck->index >= dtlck->maxcnt)
4281 		dtlck = (struct dt_lock *) txLinelock(dtlck);
4282 	lv = & dtlck->lv[dtlck->index];
4283 
4284 	lv->offset = fsi;
4285 
4286 	/* get the head/only segment */
4287 	t = &p->slot[fsi];
4288 	if (p->header.flag & BT_LEAF)
4289 		si = ((struct ldtentry *) t)->next;
4290 	else
4291 		si = ((struct idtentry *) t)->next;
4292 	t->next = si;
4293 	t->cnt = 1;
4294 
4295 	n = freecnt = 1;
4296 	xsi = fsi;
4297 
4298 	/* find the last/only segment */
4299 	while (si >= 0) {
4300 		/* is next slot contiguous ? */
4301 		if (si != xsi + 1) {
4302 			/* close current linelock */
4303 			lv->length = n;
4304 			dtlck->index++;
4305 
4306 			/* open new linelock */
4307 			if (dtlck->index < dtlck->maxcnt)
4308 				lv++;
4309 			else {
4310 				dtlck = (struct dt_lock *) txLinelock(dtlck);
4311 				lv = & dtlck->lv[0];
4312 			}
4313 
4314 			lv->offset = si;
4315 			n = 0;
4316 		}
4317 
4318 		n++;
4319 		xsi = si;
4320 		freecnt++;
4321 
4322 		t = &p->slot[si];
4323 		t->cnt = 1;
4324 		si = t->next;
4325 	}
4326 
4327 	/* close current linelock */
4328 	lv->length = n;
4329 	dtlck->index++;
4330 
4331 	*dtlock = dtlck;
4332 
4333 	/* update freelist */
4334 	t->next = p->header.freelist;
4335 	p->header.freelist = fsi;
4336 	p->header.freecnt += freecnt;
4337 
4338 	/* if delete from middle,
4339 	 * shift left the succedding entries in the stbl
4340 	 */
4341 	si = p->header.nextindex;
4342 	if (fi < si - 1)
4343 		memmove(&stbl[fi], &stbl[fi + 1], si - fi - 1);
4344 
4345 	p->header.nextindex--;
4346 }
4347 
4348 
4349 /*
4350  *	dtTruncateEntry()
4351  *
4352  * function: truncate a (leaf/internal) entry
4353  *
4354  * log freelist header, stbl, and each segment slot of entry
4355  * (even though last/only segment next field is modified,
4356  * physical image logging requires all segment slots of
4357  * the entry logged to avoid applying previous updates
4358  * to the same slots)
4359  */
4360 static void dtTruncateEntry(dtpage_t * p, int ti, struct dt_lock ** dtlock)
4361 {
4362 	int tsi;		/* truncate entry slot index */
4363 	s8 *stbl;
4364 	struct dtslot *t;
4365 	int si, freecnt;
4366 	struct dt_lock *dtlck = *dtlock;
4367 	struct lv *lv;
4368 	int fsi, xsi, n;
4369 
4370 	/* get free entry slot index */
4371 	stbl = DT_GETSTBL(p);
4372 	tsi = stbl[ti];
4373 
4374 	/* open new linelock */
4375 	if (dtlck->index >= dtlck->maxcnt)
4376 		dtlck = (struct dt_lock *) txLinelock(dtlck);
4377 	lv = & dtlck->lv[dtlck->index];
4378 
4379 	lv->offset = tsi;
4380 
4381 	/* get the head/only segment */
4382 	t = &p->slot[tsi];
4383 	ASSERT(p->header.flag & BT_INTERNAL);
4384 	((struct idtentry *) t)->namlen = 0;
4385 	si = ((struct idtentry *) t)->next;
4386 	((struct idtentry *) t)->next = -1;
4387 
4388 	n = 1;
4389 	freecnt = 0;
4390 	fsi = si;
4391 	xsi = tsi;
4392 
4393 	/* find the last/only segment */
4394 	while (si >= 0) {
4395 		/* is next slot contiguous ? */
4396 		if (si != xsi + 1) {
4397 			/* close current linelock */
4398 			lv->length = n;
4399 			dtlck->index++;
4400 
4401 			/* open new linelock */
4402 			if (dtlck->index < dtlck->maxcnt)
4403 				lv++;
4404 			else {
4405 				dtlck = (struct dt_lock *) txLinelock(dtlck);
4406 				lv = & dtlck->lv[0];
4407 			}
4408 
4409 			lv->offset = si;
4410 			n = 0;
4411 		}
4412 
4413 		n++;
4414 		xsi = si;
4415 		freecnt++;
4416 
4417 		t = &p->slot[si];
4418 		t->cnt = 1;
4419 		si = t->next;
4420 	}
4421 
4422 	/* close current linelock */
4423 	lv->length = n;
4424 	dtlck->index++;
4425 
4426 	*dtlock = dtlck;
4427 
4428 	/* update freelist */
4429 	if (freecnt == 0)
4430 		return;
4431 	t->next = p->header.freelist;
4432 	p->header.freelist = fsi;
4433 	p->header.freecnt += freecnt;
4434 }
4435 
4436 
4437 /*
4438  *	dtLinelockFreelist()
4439  */
4440 static void dtLinelockFreelist(dtpage_t * p,	/* directory page */
4441 			       int m,	/* max slot index */
4442 			       struct dt_lock ** dtlock)
4443 {
4444 	int fsi;		/* free entry slot index */
4445 	struct dtslot *t;
4446 	int si;
4447 	struct dt_lock *dtlck = *dtlock;
4448 	struct lv *lv;
4449 	int xsi, n;
4450 
4451 	/* get free entry slot index */
4452 	fsi = p->header.freelist;
4453 
4454 	/* open new linelock */
4455 	if (dtlck->index >= dtlck->maxcnt)
4456 		dtlck = (struct dt_lock *) txLinelock(dtlck);
4457 	lv = & dtlck->lv[dtlck->index];
4458 
4459 	lv->offset = fsi;
4460 
4461 	n = 1;
4462 	xsi = fsi;
4463 
4464 	t = &p->slot[fsi];
4465 	si = t->next;
4466 
4467 	/* find the last/only segment */
4468 	while (si < m && si >= 0) {
4469 		/* is next slot contiguous ? */
4470 		if (si != xsi + 1) {
4471 			/* close current linelock */
4472 			lv->length = n;
4473 			dtlck->index++;
4474 
4475 			/* open new linelock */
4476 			if (dtlck->index < dtlck->maxcnt)
4477 				lv++;
4478 			else {
4479 				dtlck = (struct dt_lock *) txLinelock(dtlck);
4480 				lv = & dtlck->lv[0];
4481 			}
4482 
4483 			lv->offset = si;
4484 			n = 0;
4485 		}
4486 
4487 		n++;
4488 		xsi = si;
4489 
4490 		t = &p->slot[si];
4491 		si = t->next;
4492 	}
4493 
4494 	/* close current linelock */
4495 	lv->length = n;
4496 	dtlck->index++;
4497 
4498 	*dtlock = dtlck;
4499 }
4500 
4501 
4502 /*
4503  * NAME: dtModify
4504  *
4505  * FUNCTION: Modify the inode number part of a directory entry
4506  *
4507  * PARAMETERS:
4508  *	tid	- Transaction id
4509  *	ip	- Inode of parent directory
4510  *	key	- Name of entry to be modified
4511  *	orig_ino	- Original inode number expected in entry
4512  *	new_ino	- New inode number to put into entry
4513  *	flag	- JFS_RENAME
4514  *
4515  * RETURNS:
4516  *	-ESTALE	- If entry found does not match orig_ino passed in
4517  *	-ENOENT	- If no entry can be found to match key
4518  *	0	- If successfully modified entry
4519  */
4520 int dtModify(tid_t tid, struct inode *ip,
4521 	 struct component_name * key, ino_t * orig_ino, ino_t new_ino, int flag)
4522 {
4523 	int rc;
4524 	s64 bn;
4525 	struct metapage *mp;
4526 	dtpage_t *p;
4527 	int index;
4528 	struct btstack btstack;
4529 	struct tlock *tlck;
4530 	struct dt_lock *dtlck;
4531 	struct lv *lv;
4532 	s8 *stbl;
4533 	int entry_si;		/* entry slot index */
4534 	struct ldtentry *entry;
4535 
4536 	/*
4537 	 *	search for the entry to modify:
4538 	 *
4539 	 * dtSearch() returns (leaf page pinned, index at which to modify).
4540 	 */
4541 	if ((rc = dtSearch(ip, key, orig_ino, &btstack, flag)))
4542 		return rc;
4543 
4544 	/* retrieve search result */
4545 	DT_GETSEARCH(ip, btstack.top, bn, mp, p, index);
4546 
4547 	BT_MARK_DIRTY(mp, ip);
4548 	/*
4549 	 * acquire a transaction lock on the leaf page of named entry
4550 	 */
4551 	tlck = txLock(tid, ip, mp, tlckDTREE | tlckENTRY);
4552 	dtlck = (struct dt_lock *) & tlck->lock;
4553 
4554 	/* get slot index of the entry */
4555 	stbl = DT_GETSTBL(p);
4556 	entry_si = stbl[index];
4557 
4558 	/* linelock entry */
4559 	ASSERT(dtlck->index == 0);
4560 	lv = & dtlck->lv[0];
4561 	lv->offset = entry_si;
4562 	lv->length = 1;
4563 	dtlck->index++;
4564 
4565 	/* get the head/only segment */
4566 	entry = (struct ldtentry *) & p->slot[entry_si];
4567 
4568 	/* substitute the inode number of the entry */
4569 	entry->inumber = cpu_to_le32(new_ino);
4570 
4571 	/* unpin the leaf page */
4572 	DT_PUTPAGE(mp);
4573 
4574 	return 0;
4575 }
4576