Root/fs/jfs/jfs_dtree.c

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

Archive Download this file



interactive