2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_da_btree.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_attr_sf.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_dinode.h"
35 #include "xfs_inode.h"
37 #include "xfs_dir2_data.h"
38 #include "xfs_dir2_leaf.h"
39 #include "xfs_dir2_block.h"
40 #include "xfs_dir2_node.h"
41 #include "xfs_dir2_trace.h"
42 #include "xfs_error.h"
45 * Local function declarations.
48 static void xfs_dir2_leaf_check(xfs_inode_t *dp, xfs_dabuf_t *bp);
50 #define xfs_dir2_leaf_check(dp, bp)
52 static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, xfs_dabuf_t **lbpp,
53 int *indexp, xfs_dabuf_t **dbpp);
54 static void xfs_dir2_leaf_log_bests(struct xfs_trans *tp, struct xfs_dabuf *bp,
56 static void xfs_dir2_leaf_log_tail(struct xfs_trans *tp, struct xfs_dabuf *bp);
60 * Convert a block form directory to a leaf form directory.
63 xfs_dir2_block_to_leaf(
64 xfs_da_args_t *args, /* operation arguments */
65 xfs_dabuf_t *dbp) /* input block's buffer */
67 __be16 *bestsp; /* leaf's bestsp entries */
68 xfs_dablk_t blkno; /* leaf block's bno */
69 xfs_dir2_block_t *block; /* block structure */
70 xfs_dir2_leaf_entry_t *blp; /* block's leaf entries */
71 xfs_dir2_block_tail_t *btp; /* block's tail */
72 xfs_inode_t *dp; /* incore directory inode */
73 int error; /* error return code */
74 xfs_dabuf_t *lbp; /* leaf block's buffer */
75 xfs_dir2_db_t ldb; /* leaf block's bno */
76 xfs_dir2_leaf_t *leaf; /* leaf structure */
77 xfs_dir2_leaf_tail_t *ltp; /* leaf's tail */
78 xfs_mount_t *mp; /* filesystem mount point */
79 int needlog; /* need to log block header */
80 int needscan; /* need to rescan bestfree */
81 xfs_trans_t *tp; /* transaction pointer */
83 xfs_dir2_trace_args_b("block_to_leaf", args, dbp);
88 * Add the leaf block to the inode.
89 * This interface will only put blocks in the leaf/node range.
90 * Since that's empty now, we'll get the root (block 0 in range).
92 if ((error = xfs_da_grow_inode(args, &blkno))) {
95 ldb = XFS_DIR2_DA_TO_DB(mp, blkno);
96 ASSERT(ldb == XFS_DIR2_LEAF_FIRSTDB(mp));
98 * Initialize the leaf block, get a buffer for it.
100 if ((error = xfs_dir2_leaf_init(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC))) {
106 xfs_dir2_data_check(dp, dbp);
107 btp = XFS_DIR2_BLOCK_TAIL_P(mp, block);
108 blp = XFS_DIR2_BLOCK_LEAF_P(btp);
110 * Set the counts in the leaf header.
112 leaf->hdr.count = cpu_to_be16(be32_to_cpu(btp->count));
113 leaf->hdr.stale = cpu_to_be16(be32_to_cpu(btp->stale));
115 * Could compact these but I think we always do the conversion
116 * after squeezing out stale entries.
118 memcpy(leaf->ents, blp, be32_to_cpu(btp->count) * sizeof(xfs_dir2_leaf_entry_t));
119 xfs_dir2_leaf_log_ents(tp, lbp, 0, be16_to_cpu(leaf->hdr.count) - 1);
123 * Make the space formerly occupied by the leaf entries and block
126 xfs_dir2_data_make_free(tp, dbp,
127 (xfs_dir2_data_aoff_t)((char *)blp - (char *)block),
128 (xfs_dir2_data_aoff_t)((char *)block + mp->m_dirblksize -
130 &needlog, &needscan);
132 * Fix up the block header, make it a data block.
134 block->hdr.magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
136 xfs_dir2_data_freescan(mp, (xfs_dir2_data_t *)block, &needlog);
138 * Set up leaf tail and bests table.
140 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
141 ltp->bestcount = cpu_to_be32(1);
142 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
143 bestsp[0] = block->hdr.bestfree[0].length;
145 * Log the data header and leaf bests table.
148 xfs_dir2_data_log_header(tp, dbp);
149 xfs_dir2_leaf_check(dp, lbp);
150 xfs_dir2_data_check(dp, dbp);
151 xfs_dir2_leaf_log_bests(tp, lbp, 0, 0);
152 xfs_da_buf_done(lbp);
157 * Add an entry to a leaf form directory.
160 xfs_dir2_leaf_addname(
161 xfs_da_args_t *args) /* operation arguments */
163 __be16 *bestsp; /* freespace table in leaf */
164 int compact; /* need to compact leaves */
165 xfs_dir2_data_t *data; /* data block structure */
166 xfs_dabuf_t *dbp; /* data block buffer */
167 xfs_dir2_data_entry_t *dep; /* data block entry */
168 xfs_inode_t *dp; /* incore directory inode */
169 xfs_dir2_data_unused_t *dup; /* data unused entry */
170 int error; /* error return value */
171 int grown; /* allocated new data block */
172 int highstale; /* index of next stale leaf */
173 int i; /* temporary, index */
174 int index; /* leaf table position */
175 xfs_dabuf_t *lbp; /* leaf's buffer */
176 xfs_dir2_leaf_t *leaf; /* leaf structure */
177 int length; /* length of new entry */
178 xfs_dir2_leaf_entry_t *lep; /* leaf entry table pointer */
179 int lfloglow; /* low leaf logging index */
180 int lfloghigh; /* high leaf logging index */
181 int lowstale; /* index of prev stale leaf */
182 xfs_dir2_leaf_tail_t *ltp; /* leaf tail pointer */
183 xfs_mount_t *mp; /* filesystem mount point */
184 int needbytes; /* leaf block bytes needed */
185 int needlog; /* need to log data header */
186 int needscan; /* need to rescan data free */
187 __be16 *tagp; /* end of data entry */
188 xfs_trans_t *tp; /* transaction pointer */
189 xfs_dir2_db_t use_block; /* data block number */
191 xfs_dir2_trace_args("leaf_addname", args);
196 * Read the leaf block.
198 error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
205 * Look up the entry by hash value and name.
206 * We know it's not there, our caller has already done a lookup.
207 * So the index is of the entry to insert in front of.
208 * But if there are dup hash values the index is of the first of those.
210 index = xfs_dir2_leaf_search_hash(args, lbp);
212 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
213 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
214 length = XFS_DIR2_DATA_ENTSIZE(args->namelen);
216 * See if there are any entries with the same hash value
217 * and space in their block for the new entry.
218 * This is good because it puts multiple same-hash value entries
219 * in a data block, improving the lookup of those entries.
221 for (use_block = -1, lep = &leaf->ents[index];
222 index < be16_to_cpu(leaf->hdr.count) && be32_to_cpu(lep->hashval) == args->hashval;
224 if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
226 i = XFS_DIR2_DATAPTR_TO_DB(mp, be32_to_cpu(lep->address));
227 ASSERT(i < be32_to_cpu(ltp->bestcount));
228 ASSERT(be16_to_cpu(bestsp[i]) != NULLDATAOFF);
229 if (be16_to_cpu(bestsp[i]) >= length) {
235 * Didn't find a block yet, linear search all the data blocks.
237 if (use_block == -1) {
238 for (i = 0; i < be32_to_cpu(ltp->bestcount); i++) {
240 * Remember a block we see that's missing.
242 if (be16_to_cpu(bestsp[i]) == NULLDATAOFF && use_block == -1)
244 else if (be16_to_cpu(bestsp[i]) >= length) {
251 * How many bytes do we need in the leaf block?
254 (leaf->hdr.stale ? 0 : (uint)sizeof(leaf->ents[0])) +
255 (use_block != -1 ? 0 : (uint)sizeof(leaf->bests[0]));
257 * Now kill use_block if it refers to a missing block, so we
258 * can use it as an indication of allocation needed.
260 if (use_block != -1 && be16_to_cpu(bestsp[use_block]) == NULLDATAOFF)
263 * If we don't have enough free bytes but we can make enough
264 * by compacting out stale entries, we'll do that.
266 if ((char *)bestsp - (char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] < needbytes &&
267 be16_to_cpu(leaf->hdr.stale) > 1) {
271 * Otherwise if we don't have enough free bytes we need to
272 * convert to node form.
274 else if ((char *)bestsp - (char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] <
277 * Just checking or no space reservation, give up.
279 if (args->justcheck || args->total == 0) {
280 xfs_da_brelse(tp, lbp);
281 return XFS_ERROR(ENOSPC);
284 * Convert to node form.
286 error = xfs_dir2_leaf_to_node(args, lbp);
287 xfs_da_buf_done(lbp);
291 * Then add the new entry.
293 return xfs_dir2_node_addname(args);
296 * Otherwise it will fit without compaction.
301 * If just checking, then it will fit unless we needed to allocate
304 if (args->justcheck) {
305 xfs_da_brelse(tp, lbp);
306 return use_block == -1 ? XFS_ERROR(ENOSPC) : 0;
309 * If no allocations are allowed, return now before we've
312 if (args->total == 0 && use_block == -1) {
313 xfs_da_brelse(tp, lbp);
314 return XFS_ERROR(ENOSPC);
317 * Need to compact the leaf entries, removing stale ones.
318 * Leave one stale entry behind - the one closest to our
319 * insertion index - and we'll shift that one to our insertion
323 xfs_dir2_leaf_compact_x1(lbp, &index, &lowstale, &highstale,
324 &lfloglow, &lfloghigh);
327 * There are stale entries, so we'll need log-low and log-high
328 * impossibly bad values later.
330 else if (be16_to_cpu(leaf->hdr.stale)) {
331 lfloglow = be16_to_cpu(leaf->hdr.count);
335 * If there was no data block space found, we need to allocate
338 if (use_block == -1) {
340 * Add the new data block.
342 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
344 xfs_da_brelse(tp, lbp);
348 * Initialize the block.
350 if ((error = xfs_dir2_data_init(args, use_block, &dbp))) {
351 xfs_da_brelse(tp, lbp);
355 * If we're adding a new data block on the end we need to
356 * extend the bests table. Copy it up one entry.
358 if (use_block >= be32_to_cpu(ltp->bestcount)) {
360 memmove(&bestsp[0], &bestsp[1],
361 be32_to_cpu(ltp->bestcount) * sizeof(bestsp[0]));
362 be32_add(<p->bestcount, 1);
363 xfs_dir2_leaf_log_tail(tp, lbp);
364 xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
367 * If we're filling in a previously empty block just log it.
370 xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
372 bestsp[use_block] = data->hdr.bestfree[0].length;
376 * Already had space in some data block.
377 * Just read that one in.
381 xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, use_block),
382 -1, &dbp, XFS_DATA_FORK))) {
383 xfs_da_brelse(tp, lbp);
389 xfs_dir2_data_check(dp, dbp);
391 * Point to the biggest freespace in our data block.
393 dup = (xfs_dir2_data_unused_t *)
394 ((char *)data + be16_to_cpu(data->hdr.bestfree[0].offset));
395 ASSERT(be16_to_cpu(dup->length) >= length);
396 needscan = needlog = 0;
398 * Mark the initial part of our freespace in use for the new entry.
400 xfs_dir2_data_use_free(tp, dbp, dup,
401 (xfs_dir2_data_aoff_t)((char *)dup - (char *)data), length,
402 &needlog, &needscan);
404 * Initialize our new entry (at last).
406 dep = (xfs_dir2_data_entry_t *)dup;
407 dep->inumber = cpu_to_be64(args->inumber);
408 dep->namelen = args->namelen;
409 memcpy(dep->name, args->name, dep->namelen);
410 tagp = XFS_DIR2_DATA_ENTRY_TAG_P(dep);
411 *tagp = cpu_to_be16((char *)dep - (char *)data);
413 * Need to scan fix up the bestfree table.
416 xfs_dir2_data_freescan(mp, data, &needlog);
418 * Need to log the data block's header.
421 xfs_dir2_data_log_header(tp, dbp);
422 xfs_dir2_data_log_entry(tp, dbp, dep);
424 * If the bests table needs to be changed, do it.
425 * Log the change unless we've already done that.
427 if (be16_to_cpu(bestsp[use_block]) != be16_to_cpu(data->hdr.bestfree[0].length)) {
428 bestsp[use_block] = data->hdr.bestfree[0].length;
430 xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
433 * Now we need to make room to insert the leaf entry.
434 * If there are no stale entries, we just insert a hole at index.
436 if (!leaf->hdr.stale) {
438 * lep is still good as the index leaf entry.
440 if (index < be16_to_cpu(leaf->hdr.count))
441 memmove(lep + 1, lep,
442 (be16_to_cpu(leaf->hdr.count) - index) * sizeof(*lep));
444 * Record low and high logging indices for the leaf.
447 lfloghigh = be16_to_cpu(leaf->hdr.count);
448 be16_add(&leaf->hdr.count, 1);
451 * There are stale entries.
452 * We will use one of them for the new entry.
453 * It's probably not at the right location, so we'll have to
454 * shift some up or down first.
458 * If we didn't compact before, we need to find the nearest
459 * stale entries before and after our insertion point.
463 * Find the first stale entry before the insertion
466 for (lowstale = index - 1;
468 be32_to_cpu(leaf->ents[lowstale].address) !=
469 XFS_DIR2_NULL_DATAPTR;
473 * Find the next stale entry at or after the insertion
474 * point, if any. Stop if we go so far that the
475 * lowstale entry would be better.
477 for (highstale = index;
478 highstale < be16_to_cpu(leaf->hdr.count) &&
479 be32_to_cpu(leaf->ents[highstale].address) !=
480 XFS_DIR2_NULL_DATAPTR &&
482 index - lowstale - 1 >= highstale - index);
487 * If the low one is better, use it.
490 (highstale == be16_to_cpu(leaf->hdr.count) ||
491 index - lowstale - 1 < highstale - index)) {
492 ASSERT(index - lowstale - 1 >= 0);
493 ASSERT(be32_to_cpu(leaf->ents[lowstale].address) ==
494 XFS_DIR2_NULL_DATAPTR);
496 * Copy entries up to cover the stale entry
497 * and make room for the new entry.
499 if (index - lowstale - 1 > 0)
500 memmove(&leaf->ents[lowstale],
501 &leaf->ents[lowstale + 1],
502 (index - lowstale - 1) * sizeof(*lep));
503 lep = &leaf->ents[index - 1];
504 lfloglow = MIN(lowstale, lfloglow);
505 lfloghigh = MAX(index - 1, lfloghigh);
508 * The high one is better, so use that one.
511 ASSERT(highstale - index >= 0);
512 ASSERT(be32_to_cpu(leaf->ents[highstale].address) ==
513 XFS_DIR2_NULL_DATAPTR);
515 * Copy entries down to cover the stale entry
516 * and make room for the new entry.
518 if (highstale - index > 0)
519 memmove(&leaf->ents[index + 1],
521 (highstale - index) * sizeof(*lep));
522 lep = &leaf->ents[index];
523 lfloglow = MIN(index, lfloglow);
524 lfloghigh = MAX(highstale, lfloghigh);
526 be16_add(&leaf->hdr.stale, -1);
529 * Fill in the new leaf entry.
531 lep->hashval = cpu_to_be32(args->hashval);
532 lep->address = cpu_to_be32(XFS_DIR2_DB_OFF_TO_DATAPTR(mp, use_block,
533 be16_to_cpu(*tagp)));
535 * Log the leaf fields and give up the buffers.
537 xfs_dir2_leaf_log_header(tp, lbp);
538 xfs_dir2_leaf_log_ents(tp, lbp, lfloglow, lfloghigh);
539 xfs_dir2_leaf_check(dp, lbp);
540 xfs_da_buf_done(lbp);
541 xfs_dir2_data_check(dp, dbp);
542 xfs_da_buf_done(dbp);
548 * Check the internal consistency of a leaf1 block.
549 * Pop an assert if something is wrong.
553 xfs_inode_t *dp, /* incore directory inode */
554 xfs_dabuf_t *bp) /* leaf's buffer */
556 int i; /* leaf index */
557 xfs_dir2_leaf_t *leaf; /* leaf structure */
558 xfs_dir2_leaf_tail_t *ltp; /* leaf tail pointer */
559 xfs_mount_t *mp; /* filesystem mount point */
560 int stale; /* count of stale leaves */
564 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAF1_MAGIC);
566 * This value is not restrictive enough.
567 * Should factor in the size of the bests table as well.
568 * We can deduce a value for that from di_size.
570 ASSERT(be16_to_cpu(leaf->hdr.count) <= XFS_DIR2_MAX_LEAF_ENTS(mp));
571 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
573 * Leaves and bests don't overlap.
575 ASSERT((char *)&leaf->ents[be16_to_cpu(leaf->hdr.count)] <=
576 (char *)XFS_DIR2_LEAF_BESTS_P(ltp));
578 * Check hash value order, count stale entries.
580 for (i = stale = 0; i < be16_to_cpu(leaf->hdr.count); i++) {
581 if (i + 1 < be16_to_cpu(leaf->hdr.count))
582 ASSERT(be32_to_cpu(leaf->ents[i].hashval) <=
583 be32_to_cpu(leaf->ents[i + 1].hashval));
584 if (be32_to_cpu(leaf->ents[i].address) == XFS_DIR2_NULL_DATAPTR)
587 ASSERT(be16_to_cpu(leaf->hdr.stale) == stale);
592 * Compact out any stale entries in the leaf.
593 * Log the header and changed leaf entries, if any.
596 xfs_dir2_leaf_compact(
597 xfs_da_args_t *args, /* operation arguments */
598 xfs_dabuf_t *bp) /* leaf buffer */
600 int from; /* source leaf index */
601 xfs_dir2_leaf_t *leaf; /* leaf structure */
602 int loglow; /* first leaf entry to log */
603 int to; /* target leaf index */
606 if (!leaf->hdr.stale) {
610 * Compress out the stale entries in place.
612 for (from = to = 0, loglow = -1; from < be16_to_cpu(leaf->hdr.count); from++) {
613 if (be32_to_cpu(leaf->ents[from].address) == XFS_DIR2_NULL_DATAPTR)
616 * Only actually copy the entries that are different.
621 leaf->ents[to] = leaf->ents[from];
626 * Update and log the header, log the leaf entries.
628 ASSERT(be16_to_cpu(leaf->hdr.stale) == from - to);
629 be16_add(&leaf->hdr.count, -(be16_to_cpu(leaf->hdr.stale)));
631 xfs_dir2_leaf_log_header(args->trans, bp);
633 xfs_dir2_leaf_log_ents(args->trans, bp, loglow, to - 1);
637 * Compact the leaf entries, removing stale ones.
638 * Leave one stale entry behind - the one closest to our
639 * insertion index - and the caller will shift that one to our insertion
641 * Return new insertion index, where the remaining stale entry is,
642 * and leaf logging indices.
645 xfs_dir2_leaf_compact_x1(
646 xfs_dabuf_t *bp, /* leaf buffer */
647 int *indexp, /* insertion index */
648 int *lowstalep, /* out: stale entry before us */
649 int *highstalep, /* out: stale entry after us */
650 int *lowlogp, /* out: low log index */
651 int *highlogp) /* out: high log index */
653 int from; /* source copy index */
654 int highstale; /* stale entry at/after index */
655 int index; /* insertion index */
656 int keepstale; /* source index of kept stale */
657 xfs_dir2_leaf_t *leaf; /* leaf structure */
658 int lowstale; /* stale entry before index */
659 int newindex=0; /* new insertion index */
660 int to; /* destination copy index */
663 ASSERT(be16_to_cpu(leaf->hdr.stale) > 1);
666 * Find the first stale entry before our index, if any.
668 for (lowstale = index - 1;
670 be32_to_cpu(leaf->ents[lowstale].address) != XFS_DIR2_NULL_DATAPTR;
674 * Find the first stale entry at or after our index, if any.
675 * Stop if the answer would be worse than lowstale.
677 for (highstale = index;
678 highstale < be16_to_cpu(leaf->hdr.count) &&
679 be32_to_cpu(leaf->ents[highstale].address) != XFS_DIR2_NULL_DATAPTR &&
680 (lowstale < 0 || index - lowstale > highstale - index);
684 * Pick the better of lowstale and highstale.
687 (highstale == be16_to_cpu(leaf->hdr.count) ||
688 index - lowstale <= highstale - index))
689 keepstale = lowstale;
691 keepstale = highstale;
693 * Copy the entries in place, removing all the stale entries
696 for (from = to = 0; from < be16_to_cpu(leaf->hdr.count); from++) {
698 * Notice the new value of index.
702 if (from != keepstale &&
703 be32_to_cpu(leaf->ents[from].address) == XFS_DIR2_NULL_DATAPTR) {
709 * Record the new keepstale value for the insertion.
711 if (from == keepstale)
712 lowstale = highstale = to;
714 * Copy only the entries that have moved.
717 leaf->ents[to] = leaf->ents[from];
722 * If the insertion point was past the last entry,
723 * set the new insertion point accordingly.
729 * Adjust the leaf header values.
731 be16_add(&leaf->hdr.count, -(from - to));
732 leaf->hdr.stale = cpu_to_be16(1);
734 * Remember the low/high stale value only in the "right"
737 if (lowstale >= newindex)
740 highstale = be16_to_cpu(leaf->hdr.count);
741 *highlogp = be16_to_cpu(leaf->hdr.count) - 1;
742 *lowstalep = lowstale;
743 *highstalep = highstale;
747 * Getdents (readdir) for leaf and node directories.
748 * This reads the data blocks only, so is the same for both forms.
751 xfs_dir2_leaf_getdents(
752 xfs_trans_t *tp, /* transaction pointer */
753 xfs_inode_t *dp, /* incore directory inode */
754 uio_t *uio, /* I/O control & vectors */
755 int *eofp, /* out: reached end of dir */
756 xfs_dirent_t *dbp, /* caller's buffer */
757 xfs_dir2_put_t put) /* ABI formatting routine */
759 xfs_dabuf_t *bp; /* data block buffer */
760 int byteoff; /* offset in current block */
761 xfs_dir2_db_t curdb; /* db for current block */
762 xfs_dir2_off_t curoff; /* current overall offset */
763 xfs_dir2_data_t *data; /* data block structure */
764 xfs_dir2_data_entry_t *dep; /* data entry */
765 xfs_dir2_data_unused_t *dup; /* unused entry */
766 int eof; /* reached end of directory */
767 int error = 0; /* error return value */
768 int i; /* temporary loop index */
769 int j; /* temporary loop index */
770 int length; /* temporary length value */
771 xfs_bmbt_irec_t *map; /* map vector for blocks */
772 xfs_extlen_t map_blocks; /* number of fsbs in map */
773 xfs_dablk_t map_off; /* last mapped file offset */
774 int map_size; /* total entries in *map */
775 int map_valid; /* valid entries in *map */
776 xfs_mount_t *mp; /* filesystem mount point */
777 xfs_dir2_off_t newoff; /* new curoff after new blk */
778 int nmap; /* mappings to ask xfs_bmapi */
779 xfs_dir2_put_args_t *p; /* formatting arg bundle */
780 char *ptr = NULL; /* pointer to current data */
781 int ra_current; /* number of read-ahead blks */
782 int ra_index; /* *map index for read-ahead */
783 int ra_offset; /* map entry offset for ra */
784 int ra_want; /* readahead count wanted */
787 * If the offset is at or past the largest allowed value,
788 * give up right away, return eof.
790 if (uio->uio_offset >= XFS_DIR2_MAX_DATAPTR) {
796 * Setup formatting arguments.
798 p = kmem_alloc(sizeof(*p), KM_SLEEP);
803 * Set up to bmap a number of blocks based on the caller's
804 * buffer size, the directory block size, and the filesystem
808 howmany(uio->uio_resid + mp->m_dirblksize,
809 mp->m_sb.sb_blocksize);
810 map = kmem_alloc(map_size * sizeof(*map), KM_SLEEP);
811 map_valid = ra_index = ra_offset = ra_current = map_blocks = 0;
815 * Inside the loop we keep the main offset value as a byte offset
816 * in the directory file.
818 curoff = XFS_DIR2_DATAPTR_TO_BYTE(mp, uio->uio_offset);
820 * Force this conversion through db so we truncate the offset
821 * down to get the start of the data block.
823 map_off = XFS_DIR2_DB_TO_DA(mp, XFS_DIR2_BYTE_TO_DB(mp, curoff));
825 * Loop over directory entries until we reach the end offset.
826 * Get more blocks and readahead as necessary.
828 while (curoff < XFS_DIR2_LEAF_OFFSET) {
830 * If we have no buffer, or we're off the end of the
831 * current buffer, need to get another one.
833 if (!bp || ptr >= (char *)bp->data + mp->m_dirblksize) {
835 * If we have a buffer, we need to release it and
836 * take it out of the mapping.
839 xfs_da_brelse(tp, bp);
841 map_blocks -= mp->m_dirblkfsbs;
843 * Loop to get rid of the extents for the
846 for (i = mp->m_dirblkfsbs; i > 0; ) {
847 j = MIN((int)map->br_blockcount, i);
848 map->br_blockcount -= j;
849 map->br_startblock += j;
850 map->br_startoff += j;
852 * If mapping is done, pitch it from
855 if (!map->br_blockcount && --map_valid)
856 memmove(&map[0], &map[1],
863 * Recalculate the readahead blocks wanted.
865 ra_want = howmany(uio->uio_resid + mp->m_dirblksize,
866 mp->m_sb.sb_blocksize) - 1;
868 * If we don't have as many as we want, and we haven't
869 * run out of data blocks, get some more mappings.
871 if (1 + ra_want > map_blocks &&
873 XFS_DIR2_BYTE_TO_DA(mp, XFS_DIR2_LEAF_OFFSET)) {
875 * Get more bmaps, fill in after the ones
876 * we already have in the table.
878 nmap = map_size - map_valid;
879 error = xfs_bmapi(tp, dp,
881 XFS_DIR2_BYTE_TO_DA(mp,
882 XFS_DIR2_LEAF_OFFSET) - map_off,
883 XFS_BMAPI_METADATA, NULL, 0,
884 &map[map_valid], &nmap, NULL, NULL);
886 * Don't know if we should ignore this or
887 * try to return an error.
888 * The trouble with returning errors
889 * is that readdir will just stop without
890 * actually passing the error through.
895 * If we got all the mappings we asked for,
896 * set the final map offset based on the
897 * last bmap value received.
898 * Otherwise, we've reached the end.
900 if (nmap == map_size - map_valid)
902 map[map_valid + nmap - 1].br_startoff +
903 map[map_valid + nmap - 1].br_blockcount;
906 XFS_DIR2_BYTE_TO_DA(mp,
907 XFS_DIR2_LEAF_OFFSET);
909 * Look for holes in the mapping, and
910 * eliminate them. Count up the valid blocks.
912 for (i = map_valid; i < map_valid + nmap; ) {
913 if (map[i].br_startblock ==
916 length = map_valid + nmap - i;
924 map[i].br_blockcount;
931 * No valid mappings, so no more data blocks.
934 curoff = XFS_DIR2_DA_TO_BYTE(mp, map_off);
938 * Read the directory block starting at the first
941 curdb = XFS_DIR2_DA_TO_DB(mp, map->br_startoff);
942 error = xfs_da_read_buf(tp, dp, map->br_startoff,
943 map->br_blockcount >= mp->m_dirblkfsbs ?
944 XFS_FSB_TO_DADDR(mp, map->br_startblock) :
948 * Should just skip over the data block instead
954 * Adjust the current amount of read-ahead: we just
955 * read a block that was previously ra.
958 ra_current -= mp->m_dirblkfsbs;
960 * Do we need more readahead?
962 for (ra_index = ra_offset = i = 0;
963 ra_want > ra_current && i < map_blocks;
964 i += mp->m_dirblkfsbs) {
965 ASSERT(ra_index < map_valid);
967 * Read-ahead a contiguous directory block.
969 if (i > ra_current &&
970 map[ra_index].br_blockcount >=
972 xfs_baread(mp->m_ddev_targp,
974 map[ra_index].br_startblock +
976 (int)BTOBB(mp->m_dirblksize));
980 * Read-ahead a non-contiguous directory block.
981 * This doesn't use our mapping, but this
982 * is a very rare case.
984 else if (i > ra_current) {
985 (void)xfs_da_reada_buf(tp, dp,
986 map[ra_index].br_startoff +
987 ra_offset, XFS_DATA_FORK);
991 * Advance offset through the mapping table.
993 for (j = 0; j < mp->m_dirblkfsbs; j++) {
995 * The rest of this extent but not
996 * more than a dir block.
998 length = MIN(mp->m_dirblkfsbs,
999 (int)(map[ra_index].br_blockcount -
1002 ra_offset += length;
1004 * Advance to the next mapping if
1005 * this one is used up.
1008 map[ra_index].br_blockcount) {
1015 * Having done a read, we need to set a new offset.
1017 newoff = XFS_DIR2_DB_OFF_TO_BYTE(mp, curdb, 0);
1019 * Start of the current block.
1021 if (curoff < newoff)
1024 * Make sure we're in the right block.
1026 else if (curoff > newoff)
1027 ASSERT(XFS_DIR2_BYTE_TO_DB(mp, curoff) ==
1030 xfs_dir2_data_check(dp, bp);
1032 * Find our position in the block.
1034 ptr = (char *)&data->u;
1035 byteoff = XFS_DIR2_BYTE_TO_OFF(mp, curoff);
1037 * Skip past the header.
1040 curoff += (uint)sizeof(data->hdr);
1042 * Skip past entries until we reach our offset.
1045 while ((char *)ptr - (char *)data < byteoff) {
1046 dup = (xfs_dir2_data_unused_t *)ptr;
1048 if (be16_to_cpu(dup->freetag)
1049 == XFS_DIR2_DATA_FREE_TAG) {
1051 length = be16_to_cpu(dup->length);
1055 dep = (xfs_dir2_data_entry_t *)ptr;
1057 XFS_DIR2_DATA_ENTSIZE(dep->namelen);
1061 * Now set our real offset.
1064 XFS_DIR2_DB_OFF_TO_BYTE(mp,
1065 XFS_DIR2_BYTE_TO_DB(mp, curoff),
1066 (char *)ptr - (char *)data);
1067 if (ptr >= (char *)data + mp->m_dirblksize) {
1073 * We have a pointer to an entry.
1076 dup = (xfs_dir2_data_unused_t *)ptr;
1078 * No, it's unused, skip over it.
1080 if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
1081 length = be16_to_cpu(dup->length);
1088 * Copy the entry into the putargs, and try formatting it.
1090 dep = (xfs_dir2_data_entry_t *)ptr;
1092 p->namelen = dep->namelen;
1094 length = XFS_DIR2_DATA_ENTSIZE(p->namelen);
1096 p->cook = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff + length);
1098 p->ino = be64_to_cpu(dep->inumber);
1100 p->ino += mp->m_inoadd;
1102 p->name = (char *)dep->name;
1107 * Won't fit. Return to caller.
1114 * Advance to next entry in the block.
1121 * All done. Set output offset value to current offset.
1124 if (curoff > XFS_DIR2_DATAPTR_TO_BYTE(mp, XFS_DIR2_MAX_DATAPTR))
1125 uio->uio_offset = XFS_DIR2_MAX_DATAPTR;
1127 uio->uio_offset = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff);
1128 kmem_free(map, map_size * sizeof(*map));
1129 kmem_free(p, sizeof(*p));
1131 xfs_da_brelse(tp, bp);
1136 * Initialize a new leaf block, leaf1 or leafn magic accepted.
1140 xfs_da_args_t *args, /* operation arguments */
1141 xfs_dir2_db_t bno, /* directory block number */
1142 xfs_dabuf_t **bpp, /* out: leaf buffer */
1143 int magic) /* magic number for block */
1145 xfs_dabuf_t *bp; /* leaf buffer */
1146 xfs_inode_t *dp; /* incore directory inode */
1147 int error; /* error return code */
1148 xfs_dir2_leaf_t *leaf; /* leaf structure */
1149 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1150 xfs_mount_t *mp; /* filesystem mount point */
1151 xfs_trans_t *tp; /* transaction pointer */
1157 ASSERT(bno >= XFS_DIR2_LEAF_FIRSTDB(mp) &&
1158 bno < XFS_DIR2_FREE_FIRSTDB(mp));
1160 * Get the buffer for the block.
1162 error = xfs_da_get_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, bno), -1, &bp,
1170 * Initialize the header.
1172 leaf->hdr.info.magic = cpu_to_be16(magic);
1173 leaf->hdr.info.forw = 0;
1174 leaf->hdr.info.back = 0;
1175 leaf->hdr.count = 0;
1176 leaf->hdr.stale = 0;
1177 xfs_dir2_leaf_log_header(tp, bp);
1179 * If it's a leaf-format directory initialize the tail.
1180 * In this case our caller has the real bests table to copy into
1183 if (magic == XFS_DIR2_LEAF1_MAGIC) {
1184 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1186 xfs_dir2_leaf_log_tail(tp, bp);
1193 * Log the bests entries indicated from a leaf1 block.
1196 xfs_dir2_leaf_log_bests(
1197 xfs_trans_t *tp, /* transaction pointer */
1198 xfs_dabuf_t *bp, /* leaf buffer */
1199 int first, /* first entry to log */
1200 int last) /* last entry to log */
1202 __be16 *firstb; /* pointer to first entry */
1203 __be16 *lastb; /* pointer to last entry */
1204 xfs_dir2_leaf_t *leaf; /* leaf structure */
1205 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1208 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAF1_MAGIC);
1209 ltp = XFS_DIR2_LEAF_TAIL_P(tp->t_mountp, leaf);
1210 firstb = XFS_DIR2_LEAF_BESTS_P(ltp) + first;
1211 lastb = XFS_DIR2_LEAF_BESTS_P(ltp) + last;
1212 xfs_da_log_buf(tp, bp, (uint)((char *)firstb - (char *)leaf),
1213 (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1217 * Log the leaf entries indicated from a leaf1 or leafn block.
1220 xfs_dir2_leaf_log_ents(
1221 xfs_trans_t *tp, /* transaction pointer */
1222 xfs_dabuf_t *bp, /* leaf buffer */
1223 int first, /* first entry to log */
1224 int last) /* last entry to log */
1226 xfs_dir2_leaf_entry_t *firstlep; /* pointer to first entry */
1227 xfs_dir2_leaf_entry_t *lastlep; /* pointer to last entry */
1228 xfs_dir2_leaf_t *leaf; /* leaf structure */
1231 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAF1_MAGIC ||
1232 be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1233 firstlep = &leaf->ents[first];
1234 lastlep = &leaf->ents[last];
1235 xfs_da_log_buf(tp, bp, (uint)((char *)firstlep - (char *)leaf),
1236 (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1240 * Log the header of the leaf1 or leafn block.
1243 xfs_dir2_leaf_log_header(
1244 xfs_trans_t *tp, /* transaction pointer */
1245 xfs_dabuf_t *bp) /* leaf buffer */
1247 xfs_dir2_leaf_t *leaf; /* leaf structure */
1250 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAF1_MAGIC ||
1251 be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1252 xfs_da_log_buf(tp, bp, (uint)((char *)&leaf->hdr - (char *)leaf),
1253 (uint)(sizeof(leaf->hdr) - 1));
1257 * Log the tail of the leaf1 block.
1260 xfs_dir2_leaf_log_tail(
1261 xfs_trans_t *tp, /* transaction pointer */
1262 xfs_dabuf_t *bp) /* leaf buffer */
1264 xfs_dir2_leaf_t *leaf; /* leaf structure */
1265 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1266 xfs_mount_t *mp; /* filesystem mount point */
1270 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAF1_MAGIC);
1271 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1272 xfs_da_log_buf(tp, bp, (uint)((char *)ltp - (char *)leaf),
1273 (uint)(mp->m_dirblksize - 1));
1277 * Look up the entry referred to by args in the leaf format directory.
1278 * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1279 * is also used by the node-format code.
1282 xfs_dir2_leaf_lookup(
1283 xfs_da_args_t *args) /* operation arguments */
1285 xfs_dabuf_t *dbp; /* data block buffer */
1286 xfs_dir2_data_entry_t *dep; /* data block entry */
1287 xfs_inode_t *dp; /* incore directory inode */
1288 int error; /* error return code */
1289 int index; /* found entry index */
1290 xfs_dabuf_t *lbp; /* leaf buffer */
1291 xfs_dir2_leaf_t *leaf; /* leaf structure */
1292 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1293 xfs_trans_t *tp; /* transaction pointer */
1295 xfs_dir2_trace_args("leaf_lookup", args);
1297 * Look up name in the leaf block, returning both buffers and index.
1299 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1304 xfs_dir2_leaf_check(dp, lbp);
1307 * Get to the leaf entry and contained data entry address.
1309 lep = &leaf->ents[index];
1311 * Point to the data entry.
1313 dep = (xfs_dir2_data_entry_t *)
1314 ((char *)dbp->data +
1315 XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, be32_to_cpu(lep->address)));
1317 * Return the found inode number.
1319 args->inumber = be64_to_cpu(dep->inumber);
1320 xfs_da_brelse(tp, dbp);
1321 xfs_da_brelse(tp, lbp);
1322 return XFS_ERROR(EEXIST);
1326 * Look up name/hash in the leaf block.
1327 * Fill in indexp with the found index, and dbpp with the data buffer.
1328 * If not found dbpp will be NULL, and ENOENT comes back.
1329 * lbpp will always be filled in with the leaf buffer unless there's an error.
1331 static int /* error */
1332 xfs_dir2_leaf_lookup_int(
1333 xfs_da_args_t *args, /* operation arguments */
1334 xfs_dabuf_t **lbpp, /* out: leaf buffer */
1335 int *indexp, /* out: index in leaf block */
1336 xfs_dabuf_t **dbpp) /* out: data buffer */
1338 xfs_dir2_db_t curdb; /* current data block number */
1339 xfs_dabuf_t *dbp; /* data buffer */
1340 xfs_dir2_data_entry_t *dep; /* data entry */
1341 xfs_inode_t *dp; /* incore directory inode */
1342 int error; /* error return code */
1343 int index; /* index in leaf block */
1344 xfs_dabuf_t *lbp; /* leaf buffer */
1345 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1346 xfs_dir2_leaf_t *leaf; /* leaf structure */
1347 xfs_mount_t *mp; /* filesystem mount point */
1348 xfs_dir2_db_t newdb; /* new data block number */
1349 xfs_trans_t *tp; /* transaction pointer */
1355 * Read the leaf block into the buffer.
1358 xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
1364 xfs_dir2_leaf_check(dp, lbp);
1366 * Look for the first leaf entry with our hash value.
1368 index = xfs_dir2_leaf_search_hash(args, lbp);
1370 * Loop over all the entries with the right hash value
1371 * looking to match the name.
1373 for (lep = &leaf->ents[index], dbp = NULL, curdb = -1;
1374 index < be16_to_cpu(leaf->hdr.count) && be32_to_cpu(lep->hashval) == args->hashval;
1377 * Skip over stale leaf entries.
1379 if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
1382 * Get the new data block number.
1384 newdb = XFS_DIR2_DATAPTR_TO_DB(mp, be32_to_cpu(lep->address));
1386 * If it's not the same as the old data block number,
1387 * need to pitch the old one and read the new one.
1389 if (newdb != curdb) {
1391 xfs_da_brelse(tp, dbp);
1393 xfs_da_read_buf(tp, dp,
1394 XFS_DIR2_DB_TO_DA(mp, newdb), -1, &dbp,
1396 xfs_da_brelse(tp, lbp);
1399 xfs_dir2_data_check(dp, dbp);
1403 * Point to the data entry.
1405 dep = (xfs_dir2_data_entry_t *)
1406 ((char *)dbp->data +
1407 XFS_DIR2_DATAPTR_TO_OFF(mp, be32_to_cpu(lep->address)));
1409 * If it matches then return it.
1411 if (dep->namelen == args->namelen &&
1412 dep->name[0] == args->name[0] &&
1413 memcmp(dep->name, args->name, args->namelen) == 0) {
1420 * No match found, return ENOENT.
1422 ASSERT(args->oknoent);
1424 xfs_da_brelse(tp, dbp);
1425 xfs_da_brelse(tp, lbp);
1426 return XFS_ERROR(ENOENT);
1430 * Remove an entry from a leaf format directory.
1433 xfs_dir2_leaf_removename(
1434 xfs_da_args_t *args) /* operation arguments */
1436 __be16 *bestsp; /* leaf block best freespace */
1437 xfs_dir2_data_t *data; /* data block structure */
1438 xfs_dir2_db_t db; /* data block number */
1439 xfs_dabuf_t *dbp; /* data block buffer */
1440 xfs_dir2_data_entry_t *dep; /* data entry structure */
1441 xfs_inode_t *dp; /* incore directory inode */
1442 int error; /* error return code */
1443 xfs_dir2_db_t i; /* temporary data block # */
1444 int index; /* index into leaf entries */
1445 xfs_dabuf_t *lbp; /* leaf buffer */
1446 xfs_dir2_leaf_t *leaf; /* leaf structure */
1447 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1448 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1449 xfs_mount_t *mp; /* filesystem mount point */
1450 int needlog; /* need to log data header */
1451 int needscan; /* need to rescan data frees */
1452 xfs_dir2_data_off_t oldbest; /* old value of best free */
1453 xfs_trans_t *tp; /* transaction pointer */
1455 xfs_dir2_trace_args("leaf_removename", args);
1457 * Lookup the leaf entry, get the leaf and data blocks read in.
1459 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1467 xfs_dir2_data_check(dp, dbp);
1469 * Point to the leaf entry, use that to point to the data entry.
1471 lep = &leaf->ents[index];
1472 db = XFS_DIR2_DATAPTR_TO_DB(mp, be32_to_cpu(lep->address));
1473 dep = (xfs_dir2_data_entry_t *)
1474 ((char *)data + XFS_DIR2_DATAPTR_TO_OFF(mp, be32_to_cpu(lep->address)));
1475 needscan = needlog = 0;
1476 oldbest = be16_to_cpu(data->hdr.bestfree[0].length);
1477 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1478 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1479 ASSERT(be16_to_cpu(bestsp[db]) == oldbest);
1481 * Mark the former data entry unused.
1483 xfs_dir2_data_make_free(tp, dbp,
1484 (xfs_dir2_data_aoff_t)((char *)dep - (char *)data),
1485 XFS_DIR2_DATA_ENTSIZE(dep->namelen), &needlog, &needscan);
1487 * We just mark the leaf entry stale by putting a null in it.
1489 be16_add(&leaf->hdr.stale, 1);
1490 xfs_dir2_leaf_log_header(tp, lbp);
1491 lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
1492 xfs_dir2_leaf_log_ents(tp, lbp, index, index);
1494 * Scan the freespace in the data block again if necessary,
1495 * log the data block header if necessary.
1498 xfs_dir2_data_freescan(mp, data, &needlog);
1500 xfs_dir2_data_log_header(tp, dbp);
1502 * If the longest freespace in the data block has changed,
1503 * put the new value in the bests table and log that.
1505 if (be16_to_cpu(data->hdr.bestfree[0].length) != oldbest) {
1506 bestsp[db] = data->hdr.bestfree[0].length;
1507 xfs_dir2_leaf_log_bests(tp, lbp, db, db);
1509 xfs_dir2_data_check(dp, dbp);
1511 * If the data block is now empty then get rid of the data block.
1513 if (be16_to_cpu(data->hdr.bestfree[0].length) ==
1514 mp->m_dirblksize - (uint)sizeof(data->hdr)) {
1515 ASSERT(db != mp->m_dirdatablk);
1516 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1518 * Nope, can't get rid of it because it caused
1519 * allocation of a bmap btree block to do so.
1520 * Just go on, returning success, leaving the
1521 * empty block in place.
1523 if (error == ENOSPC && args->total == 0) {
1524 xfs_da_buf_done(dbp);
1527 xfs_dir2_leaf_check(dp, lbp);
1528 xfs_da_buf_done(lbp);
1533 * If this is the last data block then compact the
1534 * bests table by getting rid of entries.
1536 if (db == be32_to_cpu(ltp->bestcount) - 1) {
1538 * Look for the last active entry (i).
1540 for (i = db - 1; i > 0; i--) {
1541 if (be16_to_cpu(bestsp[i]) != NULLDATAOFF)
1545 * Copy the table down so inactive entries at the
1548 memmove(&bestsp[db - i], bestsp,
1549 (be32_to_cpu(ltp->bestcount) - (db - i)) * sizeof(*bestsp));
1550 be32_add(<p->bestcount, -(db - i));
1551 xfs_dir2_leaf_log_tail(tp, lbp);
1552 xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1554 bestsp[db] = cpu_to_be16(NULLDATAOFF);
1557 * If the data block was not the first one, drop it.
1559 else if (db != mp->m_dirdatablk && dbp != NULL) {
1560 xfs_da_buf_done(dbp);
1563 xfs_dir2_leaf_check(dp, lbp);
1565 * See if we can convert to block form.
1567 return xfs_dir2_leaf_to_block(args, lbp, dbp);
1571 * Replace the inode number in a leaf format directory entry.
1574 xfs_dir2_leaf_replace(
1575 xfs_da_args_t *args) /* operation arguments */
1577 xfs_dabuf_t *dbp; /* data block buffer */
1578 xfs_dir2_data_entry_t *dep; /* data block entry */
1579 xfs_inode_t *dp; /* incore directory inode */
1580 int error; /* error return code */
1581 int index; /* index of leaf entry */
1582 xfs_dabuf_t *lbp; /* leaf buffer */
1583 xfs_dir2_leaf_t *leaf; /* leaf structure */
1584 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1585 xfs_trans_t *tp; /* transaction pointer */
1587 xfs_dir2_trace_args("leaf_replace", args);
1589 * Look up the entry.
1591 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1597 * Point to the leaf entry, get data address from it.
1599 lep = &leaf->ents[index];
1601 * Point to the data entry.
1603 dep = (xfs_dir2_data_entry_t *)
1604 ((char *)dbp->data +
1605 XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, be32_to_cpu(lep->address)));
1606 ASSERT(args->inumber != be64_to_cpu(dep->inumber));
1608 * Put the new inode number in, log it.
1610 dep->inumber = cpu_to_be64(args->inumber);
1612 xfs_dir2_data_log_entry(tp, dbp, dep);
1613 xfs_da_buf_done(dbp);
1614 xfs_dir2_leaf_check(dp, lbp);
1615 xfs_da_brelse(tp, lbp);
1620 * Return index in the leaf block (lbp) which is either the first
1621 * one with this hash value, or if there are none, the insert point
1622 * for that hash value.
1624 int /* index value */
1625 xfs_dir2_leaf_search_hash(
1626 xfs_da_args_t *args, /* operation arguments */
1627 xfs_dabuf_t *lbp) /* leaf buffer */
1629 xfs_dahash_t hash=0; /* hash from this entry */
1630 xfs_dahash_t hashwant; /* hash value looking for */
1631 int high; /* high leaf index */
1632 int low; /* low leaf index */
1633 xfs_dir2_leaf_t *leaf; /* leaf structure */
1634 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1635 int mid=0; /* current leaf index */
1639 if (!leaf->hdr.count)
1643 * Note, the table cannot be empty, so we have to go through the loop.
1644 * Binary search the leaf entries looking for our hash value.
1646 for (lep = leaf->ents, low = 0, high = be16_to_cpu(leaf->hdr.count) - 1,
1647 hashwant = args->hashval;
1649 mid = (low + high) >> 1;
1650 if ((hash = be32_to_cpu(lep[mid].hashval)) == hashwant)
1652 if (hash < hashwant)
1658 * Found one, back up through all the equal hash values.
1660 if (hash == hashwant) {
1661 while (mid > 0 && be32_to_cpu(lep[mid - 1].hashval) == hashwant) {
1666 * Need to point to an entry higher than ours.
1668 else if (hash < hashwant)
1674 * Trim off a trailing data block. We know it's empty since the leaf
1675 * freespace table says so.
1678 xfs_dir2_leaf_trim_data(
1679 xfs_da_args_t *args, /* operation arguments */
1680 xfs_dabuf_t *lbp, /* leaf buffer */
1681 xfs_dir2_db_t db) /* data block number */
1683 __be16 *bestsp; /* leaf bests table */
1685 xfs_dir2_data_t *data; /* data block structure */
1687 xfs_dabuf_t *dbp; /* data block buffer */
1688 xfs_inode_t *dp; /* incore directory inode */
1689 int error; /* error return value */
1690 xfs_dir2_leaf_t *leaf; /* leaf structure */
1691 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1692 xfs_mount_t *mp; /* filesystem mount point */
1693 xfs_trans_t *tp; /* transaction pointer */
1699 * Read the offending data block. We need its buffer.
1701 if ((error = xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, db), -1, &dbp,
1707 ASSERT(be32_to_cpu(data->hdr.magic) == XFS_DIR2_DATA_MAGIC);
1709 /* this seems to be an error
1710 * data is only valid if DEBUG is defined?
1715 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1716 ASSERT(be16_to_cpu(data->hdr.bestfree[0].length) ==
1717 mp->m_dirblksize - (uint)sizeof(data->hdr));
1718 ASSERT(db == be32_to_cpu(ltp->bestcount) - 1);
1720 * Get rid of the data block.
1722 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1723 ASSERT(error != ENOSPC);
1724 xfs_da_brelse(tp, dbp);
1728 * Eliminate the last bests entry from the table.
1730 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1731 be32_add(<p->bestcount, -1);
1732 memmove(&bestsp[1], &bestsp[0], be32_to_cpu(ltp->bestcount) * sizeof(*bestsp));
1733 xfs_dir2_leaf_log_tail(tp, lbp);
1734 xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1739 * Convert node form directory to leaf form directory.
1740 * The root of the node form dir needs to already be a LEAFN block.
1741 * Just return if we can't do anything.
1744 xfs_dir2_node_to_leaf(
1745 xfs_da_state_t *state) /* directory operation state */
1747 xfs_da_args_t *args; /* operation arguments */
1748 xfs_inode_t *dp; /* incore directory inode */
1749 int error; /* error return code */
1750 xfs_dabuf_t *fbp; /* buffer for freespace block */
1751 xfs_fileoff_t fo; /* freespace file offset */
1752 xfs_dir2_free_t *free; /* freespace structure */
1753 xfs_dabuf_t *lbp; /* buffer for leaf block */
1754 xfs_dir2_leaf_tail_t *ltp; /* tail of leaf structure */
1755 xfs_dir2_leaf_t *leaf; /* leaf structure */
1756 xfs_mount_t *mp; /* filesystem mount point */
1757 int rval; /* successful free trim? */
1758 xfs_trans_t *tp; /* transaction pointer */
1761 * There's more than a leaf level in the btree, so there must
1762 * be multiple leafn blocks. Give up.
1764 if (state->path.active > 1)
1767 xfs_dir2_trace_args("node_to_leaf", args);
1772 * Get the last offset in the file.
1774 if ((error = xfs_bmap_last_offset(tp, dp, &fo, XFS_DATA_FORK))) {
1777 fo -= mp->m_dirblkfsbs;
1779 * If there are freespace blocks other than the first one,
1780 * take this opportunity to remove trailing empty freespace blocks
1781 * that may have been left behind during no-space-reservation
1784 while (fo > mp->m_dirfreeblk) {
1785 if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1789 fo -= mp->m_dirblkfsbs;
1794 * Now find the block just before the freespace block.
1796 if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1800 * If it's not the single leaf block, give up.
1802 if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + mp->m_dirblksize)
1804 lbp = state->path.blk[0].bp;
1806 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1808 * Read the freespace block.
1810 if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
1815 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
1816 ASSERT(!free->hdr.firstdb);
1818 * Now see if the leafn and free data will fit in a leaf1.
1819 * If not, release the buffer and give up.
1821 if ((uint)sizeof(leaf->hdr) +
1822 (be16_to_cpu(leaf->hdr.count) - be16_to_cpu(leaf->hdr.stale)) * (uint)sizeof(leaf->ents[0]) +
1823 be32_to_cpu(free->hdr.nvalid) * (uint)sizeof(leaf->bests[0]) +
1824 (uint)sizeof(leaf->tail) >
1826 xfs_da_brelse(tp, fbp);
1830 * If the leaf has any stale entries in it, compress them out.
1831 * The compact routine will log the header.
1833 if (be16_to_cpu(leaf->hdr.stale))
1834 xfs_dir2_leaf_compact(args, lbp);
1836 xfs_dir2_leaf_log_header(tp, lbp);
1837 leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAF1_MAGIC);
1839 * Set up the leaf tail from the freespace block.
1841 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1842 ltp->bestcount = free->hdr.nvalid;
1844 * Set up the leaf bests table.
1846 memcpy(XFS_DIR2_LEAF_BESTS_P(ltp), free->bests,
1847 be32_to_cpu(ltp->bestcount) * sizeof(leaf->bests[0]));
1848 xfs_dir2_leaf_log_bests(tp, lbp, 0, be32_to_cpu(ltp->bestcount) - 1);
1849 xfs_dir2_leaf_log_tail(tp, lbp);
1850 xfs_dir2_leaf_check(dp, lbp);
1852 * Get rid of the freespace block.
1854 error = xfs_dir2_shrink_inode(args, XFS_DIR2_FREE_FIRSTDB(mp), fbp);
1857 * This can't fail here because it can only happen when
1858 * punching out the middle of an extent, and this is an
1861 ASSERT(error != ENOSPC);
1866 * Now see if we can convert the single-leaf directory
1867 * down to a block form directory.
1868 * This routine always kills the dabuf for the leaf, so
1869 * eliminate it from the path.
1871 error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1872 state->path.blk[0].bp = NULL;