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