Linux-2.6.12-rc2
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / xfs / xfs_dir2_leaf.c
1 /*
2 * Copyright (c) 2000-2003 Silicon Graphics, Inc. All Rights Reserved.
3 *
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms of version 2 of the GNU General Public License as
6 * published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it would be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
11 *
12 * Further, this software is distributed without any warranty that it is
13 * free of the rightful claim of any third person regarding infringement
14 * or the like. Any license provided herein, whether implied or
15 * otherwise, applies only to this software file. Patent licenses, if
16 * any, provided herein do not apply to combinations of this program with
17 * other software, or any other product whatsoever.
18 *
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write the Free Software Foundation, Inc., 59
21 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
22 *
23 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
24 * Mountain View, CA 94043, or:
25 *
26 * http://www.sgi.com
27 *
28 * For further information regarding this notice, see:
29 *
30 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
31 */
32
33 /*
34 * xfs_dir2_leaf.c
35 * XFS directory version 2 implementation - single leaf form
36 * see xfs_dir2_leaf.h for data structures.
37 * These directories have multiple XFS_DIR2_DATA blocks and one
38 * XFS_DIR2_LEAF1 block containing the hash table and freespace map.
39 */
40
41 #include "xfs.h"
42
43 #include "xfs_macros.h"
44 #include "xfs_types.h"
45 #include "xfs_inum.h"
46 #include "xfs_log.h"
47 #include "xfs_trans.h"
48 #include "xfs_sb.h"
49 #include "xfs_ag.h"
50 #include "xfs_dir.h"
51 #include "xfs_dir2.h"
52 #include "xfs_dmapi.h"
53 #include "xfs_mount.h"
54 #include "xfs_bmap_btree.h"
55 #include "xfs_attr_sf.h"
56 #include "xfs_dir_sf.h"
57 #include "xfs_dir2_sf.h"
58 #include "xfs_dinode.h"
59 #include "xfs_inode.h"
60 #include "xfs_bmap.h"
61 #include "xfs_da_btree.h"
62 #include "xfs_dir2_data.h"
63 #include "xfs_dir2_leaf.h"
64 #include "xfs_dir2_block.h"
65 #include "xfs_dir2_node.h"
66 #include "xfs_dir2_trace.h"
67 #include "xfs_error.h"
68 #include "xfs_bit.h"
69
70 /*
71 * Local function declarations.
72 */
73 #ifdef DEBUG
74 static void xfs_dir2_leaf_check(xfs_inode_t *dp, xfs_dabuf_t *bp);
75 #else
76 #define xfs_dir2_leaf_check(dp, bp)
77 #endif
78 static int xfs_dir2_leaf_lookup_int(xfs_da_args_t *args, xfs_dabuf_t **lbpp,
79 int *indexp, xfs_dabuf_t **dbpp);
80
81 /*
82 * Convert a block form directory to a leaf form directory.
83 */
84 int /* error */
85 xfs_dir2_block_to_leaf(
86 xfs_da_args_t *args, /* operation arguments */
87 xfs_dabuf_t *dbp) /* input block's buffer */
88 {
89 xfs_dir2_data_off_t *bestsp; /* leaf's bestsp entries */
90 xfs_dablk_t blkno; /* leaf block's bno */
91 xfs_dir2_block_t *block; /* block structure */
92 xfs_dir2_leaf_entry_t *blp; /* block's leaf entries */
93 xfs_dir2_block_tail_t *btp; /* block's tail */
94 xfs_inode_t *dp; /* incore directory inode */
95 int error; /* error return code */
96 xfs_dabuf_t *lbp; /* leaf block's buffer */
97 xfs_dir2_db_t ldb; /* leaf block's bno */
98 xfs_dir2_leaf_t *leaf; /* leaf structure */
99 xfs_dir2_leaf_tail_t *ltp; /* leaf's tail */
100 xfs_mount_t *mp; /* filesystem mount point */
101 int needlog; /* need to log block header */
102 int needscan; /* need to rescan bestfree */
103 xfs_trans_t *tp; /* transaction pointer */
104
105 xfs_dir2_trace_args_b("block_to_leaf", args, dbp);
106 dp = args->dp;
107 mp = dp->i_mount;
108 tp = args->trans;
109 /*
110 * Add the leaf block to the inode.
111 * This interface will only put blocks in the leaf/node range.
112 * Since that's empty now, we'll get the root (block 0 in range).
113 */
114 if ((error = xfs_da_grow_inode(args, &blkno))) {
115 return error;
116 }
117 ldb = XFS_DIR2_DA_TO_DB(mp, blkno);
118 ASSERT(ldb == XFS_DIR2_LEAF_FIRSTDB(mp));
119 /*
120 * Initialize the leaf block, get a buffer for it.
121 */
122 if ((error = xfs_dir2_leaf_init(args, ldb, &lbp, XFS_DIR2_LEAF1_MAGIC))) {
123 return error;
124 }
125 ASSERT(lbp != NULL);
126 leaf = lbp->data;
127 block = dbp->data;
128 xfs_dir2_data_check(dp, dbp);
129 btp = XFS_DIR2_BLOCK_TAIL_P(mp, block);
130 blp = XFS_DIR2_BLOCK_LEAF_P(btp);
131 /*
132 * Set the counts in the leaf header.
133 */
134 INT_COPY(leaf->hdr.count, btp->count, ARCH_CONVERT); /* INT_: type change */
135 INT_COPY(leaf->hdr.stale, btp->stale, ARCH_CONVERT); /* INT_: type change */
136 /*
137 * Could compact these but I think we always do the conversion
138 * after squeezing out stale entries.
139 */
140 memcpy(leaf->ents, blp, INT_GET(btp->count, ARCH_CONVERT) * sizeof(xfs_dir2_leaf_entry_t));
141 xfs_dir2_leaf_log_ents(tp, lbp, 0, INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1);
142 needscan = 0;
143 needlog = 1;
144 /*
145 * Make the space formerly occupied by the leaf entries and block
146 * tail be free.
147 */
148 xfs_dir2_data_make_free(tp, dbp,
149 (xfs_dir2_data_aoff_t)((char *)blp - (char *)block),
150 (xfs_dir2_data_aoff_t)((char *)block + mp->m_dirblksize -
151 (char *)blp),
152 &needlog, &needscan);
153 /*
154 * Fix up the block header, make it a data block.
155 */
156 INT_SET(block->hdr.magic, ARCH_CONVERT, XFS_DIR2_DATA_MAGIC);
157 if (needscan)
158 xfs_dir2_data_freescan(mp, (xfs_dir2_data_t *)block, &needlog,
159 NULL);
160 /*
161 * Set up leaf tail and bests table.
162 */
163 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
164 INT_SET(ltp->bestcount, ARCH_CONVERT, 1);
165 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
166 INT_COPY(bestsp[0], block->hdr.bestfree[0].length, ARCH_CONVERT);
167 /*
168 * Log the data header and leaf bests table.
169 */
170 if (needlog)
171 xfs_dir2_data_log_header(tp, dbp);
172 xfs_dir2_leaf_check(dp, lbp);
173 xfs_dir2_data_check(dp, dbp);
174 xfs_dir2_leaf_log_bests(tp, lbp, 0, 0);
175 xfs_da_buf_done(lbp);
176 return 0;
177 }
178
179 /*
180 * Add an entry to a leaf form directory.
181 */
182 int /* error */
183 xfs_dir2_leaf_addname(
184 xfs_da_args_t *args) /* operation arguments */
185 {
186 xfs_dir2_data_off_t *bestsp; /* freespace table in leaf */
187 int compact; /* need to compact leaves */
188 xfs_dir2_data_t *data; /* data block structure */
189 xfs_dabuf_t *dbp; /* data block buffer */
190 xfs_dir2_data_entry_t *dep; /* data block entry */
191 xfs_inode_t *dp; /* incore directory inode */
192 xfs_dir2_data_unused_t *dup; /* data unused entry */
193 int error; /* error return value */
194 int grown; /* allocated new data block */
195 int highstale; /* index of next stale leaf */
196 int i; /* temporary, index */
197 int index; /* leaf table position */
198 xfs_dabuf_t *lbp; /* leaf's buffer */
199 xfs_dir2_leaf_t *leaf; /* leaf structure */
200 int length; /* length of new entry */
201 xfs_dir2_leaf_entry_t *lep; /* leaf entry table pointer */
202 int lfloglow; /* low leaf logging index */
203 int lfloghigh; /* high leaf logging index */
204 int lowstale; /* index of prev stale leaf */
205 xfs_dir2_leaf_tail_t *ltp; /* leaf tail pointer */
206 xfs_mount_t *mp; /* filesystem mount point */
207 int needbytes; /* leaf block bytes needed */
208 int needlog; /* need to log data header */
209 int needscan; /* need to rescan data free */
210 xfs_dir2_data_off_t *tagp; /* end of data entry */
211 xfs_trans_t *tp; /* transaction pointer */
212 xfs_dir2_db_t use_block; /* data block number */
213
214 xfs_dir2_trace_args("leaf_addname", args);
215 dp = args->dp;
216 tp = args->trans;
217 mp = dp->i_mount;
218 /*
219 * Read the leaf block.
220 */
221 error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
222 XFS_DATA_FORK);
223 if (error) {
224 return error;
225 }
226 ASSERT(lbp != NULL);
227 /*
228 * Look up the entry by hash value and name.
229 * We know it's not there, our caller has already done a lookup.
230 * So the index is of the entry to insert in front of.
231 * But if there are dup hash values the index is of the first of those.
232 */
233 index = xfs_dir2_leaf_search_hash(args, lbp);
234 leaf = lbp->data;
235 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
236 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
237 length = XFS_DIR2_DATA_ENTSIZE(args->namelen);
238 /*
239 * See if there are any entries with the same hash value
240 * and space in their block for the new entry.
241 * This is good because it puts multiple same-hash value entries
242 * in a data block, improving the lookup of those entries.
243 */
244 for (use_block = -1, lep = &leaf->ents[index];
245 index < INT_GET(leaf->hdr.count, ARCH_CONVERT) && INT_GET(lep->hashval, ARCH_CONVERT) == args->hashval;
246 index++, lep++) {
247 if (INT_GET(lep->address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
248 continue;
249 i = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
250 ASSERT(i < INT_GET(ltp->bestcount, ARCH_CONVERT));
251 ASSERT(INT_GET(bestsp[i], ARCH_CONVERT) != NULLDATAOFF);
252 if (INT_GET(bestsp[i], ARCH_CONVERT) >= length) {
253 use_block = i;
254 break;
255 }
256 }
257 /*
258 * Didn't find a block yet, linear search all the data blocks.
259 */
260 if (use_block == -1) {
261 for (i = 0; i < INT_GET(ltp->bestcount, ARCH_CONVERT); i++) {
262 /*
263 * Remember a block we see that's missing.
264 */
265 if (INT_GET(bestsp[i], ARCH_CONVERT) == NULLDATAOFF && use_block == -1)
266 use_block = i;
267 else if (INT_GET(bestsp[i], ARCH_CONVERT) >= length) {
268 use_block = i;
269 break;
270 }
271 }
272 }
273 /*
274 * How many bytes do we need in the leaf block?
275 */
276 needbytes =
277 (leaf->hdr.stale ? 0 : (uint)sizeof(leaf->ents[0])) +
278 (use_block != -1 ? 0 : (uint)sizeof(leaf->bests[0]));
279 /*
280 * Now kill use_block if it refers to a missing block, so we
281 * can use it as an indication of allocation needed.
282 */
283 if (use_block != -1 && INT_GET(bestsp[use_block], ARCH_CONVERT) == NULLDATAOFF)
284 use_block = -1;
285 /*
286 * If we don't have enough free bytes but we can make enough
287 * by compacting out stale entries, we'll do that.
288 */
289 if ((char *)bestsp - (char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] < needbytes &&
290 INT_GET(leaf->hdr.stale, ARCH_CONVERT) > 1) {
291 compact = 1;
292 }
293 /*
294 * Otherwise if we don't have enough free bytes we need to
295 * convert to node form.
296 */
297 else if ((char *)bestsp - (char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] <
298 needbytes) {
299 /*
300 * Just checking or no space reservation, give up.
301 */
302 if (args->justcheck || args->total == 0) {
303 xfs_da_brelse(tp, lbp);
304 return XFS_ERROR(ENOSPC);
305 }
306 /*
307 * Convert to node form.
308 */
309 error = xfs_dir2_leaf_to_node(args, lbp);
310 xfs_da_buf_done(lbp);
311 if (error)
312 return error;
313 /*
314 * Then add the new entry.
315 */
316 return xfs_dir2_node_addname(args);
317 }
318 /*
319 * Otherwise it will fit without compaction.
320 */
321 else
322 compact = 0;
323 /*
324 * If just checking, then it will fit unless we needed to allocate
325 * a new data block.
326 */
327 if (args->justcheck) {
328 xfs_da_brelse(tp, lbp);
329 return use_block == -1 ? XFS_ERROR(ENOSPC) : 0;
330 }
331 /*
332 * If no allocations are allowed, return now before we've
333 * changed anything.
334 */
335 if (args->total == 0 && use_block == -1) {
336 xfs_da_brelse(tp, lbp);
337 return XFS_ERROR(ENOSPC);
338 }
339 /*
340 * Need to compact the leaf entries, removing stale ones.
341 * Leave one stale entry behind - the one closest to our
342 * insertion index - and we'll shift that one to our insertion
343 * point later.
344 */
345 if (compact) {
346 xfs_dir2_leaf_compact_x1(lbp, &index, &lowstale, &highstale,
347 &lfloglow, &lfloghigh);
348 }
349 /*
350 * There are stale entries, so we'll need log-low and log-high
351 * impossibly bad values later.
352 */
353 else if (INT_GET(leaf->hdr.stale, ARCH_CONVERT)) {
354 lfloglow = INT_GET(leaf->hdr.count, ARCH_CONVERT);
355 lfloghigh = -1;
356 }
357 /*
358 * If there was no data block space found, we need to allocate
359 * a new one.
360 */
361 if (use_block == -1) {
362 /*
363 * Add the new data block.
364 */
365 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_DATA_SPACE,
366 &use_block))) {
367 xfs_da_brelse(tp, lbp);
368 return error;
369 }
370 /*
371 * Initialize the block.
372 */
373 if ((error = xfs_dir2_data_init(args, use_block, &dbp))) {
374 xfs_da_brelse(tp, lbp);
375 return error;
376 }
377 /*
378 * If we're adding a new data block on the end we need to
379 * extend the bests table. Copy it up one entry.
380 */
381 if (use_block >= INT_GET(ltp->bestcount, ARCH_CONVERT)) {
382 bestsp--;
383 memmove(&bestsp[0], &bestsp[1],
384 INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(bestsp[0]));
385 INT_MOD(ltp->bestcount, ARCH_CONVERT, +1);
386 xfs_dir2_leaf_log_tail(tp, lbp);
387 xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
388 }
389 /*
390 * If we're filling in a previously empty block just log it.
391 */
392 else
393 xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
394 data = dbp->data;
395 INT_COPY(bestsp[use_block], data->hdr.bestfree[0].length, ARCH_CONVERT);
396 grown = 1;
397 }
398 /*
399 * Already had space in some data block.
400 * Just read that one in.
401 */
402 else {
403 if ((error =
404 xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, use_block),
405 -1, &dbp, XFS_DATA_FORK))) {
406 xfs_da_brelse(tp, lbp);
407 return error;
408 }
409 data = dbp->data;
410 grown = 0;
411 }
412 xfs_dir2_data_check(dp, dbp);
413 /*
414 * Point to the biggest freespace in our data block.
415 */
416 dup = (xfs_dir2_data_unused_t *)
417 ((char *)data + INT_GET(data->hdr.bestfree[0].offset, ARCH_CONVERT));
418 ASSERT(INT_GET(dup->length, ARCH_CONVERT) >= length);
419 needscan = needlog = 0;
420 /*
421 * Mark the initial part of our freespace in use for the new entry.
422 */
423 xfs_dir2_data_use_free(tp, dbp, dup,
424 (xfs_dir2_data_aoff_t)((char *)dup - (char *)data), length,
425 &needlog, &needscan);
426 /*
427 * Initialize our new entry (at last).
428 */
429 dep = (xfs_dir2_data_entry_t *)dup;
430 INT_SET(dep->inumber, ARCH_CONVERT, args->inumber);
431 dep->namelen = args->namelen;
432 memcpy(dep->name, args->name, dep->namelen);
433 tagp = XFS_DIR2_DATA_ENTRY_TAG_P(dep);
434 INT_SET(*tagp, ARCH_CONVERT, (xfs_dir2_data_off_t)((char *)dep - (char *)data));
435 /*
436 * Need to scan fix up the bestfree table.
437 */
438 if (needscan)
439 xfs_dir2_data_freescan(mp, data, &needlog, NULL);
440 /*
441 * Need to log the data block's header.
442 */
443 if (needlog)
444 xfs_dir2_data_log_header(tp, dbp);
445 xfs_dir2_data_log_entry(tp, dbp, dep);
446 /*
447 * If the bests table needs to be changed, do it.
448 * Log the change unless we've already done that.
449 */
450 if (INT_GET(bestsp[use_block], ARCH_CONVERT) != INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT)) {
451 INT_COPY(bestsp[use_block], data->hdr.bestfree[0].length, ARCH_CONVERT);
452 if (!grown)
453 xfs_dir2_leaf_log_bests(tp, lbp, use_block, use_block);
454 }
455 /*
456 * Now we need to make room to insert the leaf entry.
457 * If there are no stale entries, we just insert a hole at index.
458 */
459 if (!leaf->hdr.stale) {
460 /*
461 * lep is still good as the index leaf entry.
462 */
463 if (index < INT_GET(leaf->hdr.count, ARCH_CONVERT))
464 memmove(lep + 1, lep,
465 (INT_GET(leaf->hdr.count, ARCH_CONVERT) - index) * sizeof(*lep));
466 /*
467 * Record low and high logging indices for the leaf.
468 */
469 lfloglow = index;
470 lfloghigh = INT_GET(leaf->hdr.count, ARCH_CONVERT);
471 INT_MOD(leaf->hdr.count, ARCH_CONVERT, +1);
472 }
473 /*
474 * There are stale entries.
475 * We will use one of them for the new entry.
476 * It's probably not at the right location, so we'll have to
477 * shift some up or down first.
478 */
479 else {
480 /*
481 * If we didn't compact before, we need to find the nearest
482 * stale entries before and after our insertion point.
483 */
484 if (compact == 0) {
485 /*
486 * Find the first stale entry before the insertion
487 * point, if any.
488 */
489 for (lowstale = index - 1;
490 lowstale >= 0 &&
491 INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) !=
492 XFS_DIR2_NULL_DATAPTR;
493 lowstale--)
494 continue;
495 /*
496 * Find the next stale entry at or after the insertion
497 * point, if any. Stop if we go so far that the
498 * lowstale entry would be better.
499 */
500 for (highstale = index;
501 highstale < INT_GET(leaf->hdr.count, ARCH_CONVERT) &&
502 INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) !=
503 XFS_DIR2_NULL_DATAPTR &&
504 (lowstale < 0 ||
505 index - lowstale - 1 >= highstale - index);
506 highstale++)
507 continue;
508 }
509 /*
510 * If the low one is better, use it.
511 */
512 if (lowstale >= 0 &&
513 (highstale == INT_GET(leaf->hdr.count, ARCH_CONVERT) ||
514 index - lowstale - 1 < highstale - index)) {
515 ASSERT(index - lowstale - 1 >= 0);
516 ASSERT(INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) ==
517 XFS_DIR2_NULL_DATAPTR);
518 /*
519 * Copy entries up to cover the stale entry
520 * and make room for the new entry.
521 */
522 if (index - lowstale - 1 > 0)
523 memmove(&leaf->ents[lowstale],
524 &leaf->ents[lowstale + 1],
525 (index - lowstale - 1) * sizeof(*lep));
526 lep = &leaf->ents[index - 1];
527 lfloglow = MIN(lowstale, lfloglow);
528 lfloghigh = MAX(index - 1, lfloghigh);
529 }
530 /*
531 * The high one is better, so use that one.
532 */
533 else {
534 ASSERT(highstale - index >= 0);
535 ASSERT(INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) ==
536 XFS_DIR2_NULL_DATAPTR);
537 /*
538 * Copy entries down to copver the stale entry
539 * and make room for the new entry.
540 */
541 if (highstale - index > 0)
542 memmove(&leaf->ents[index + 1],
543 &leaf->ents[index],
544 (highstale - index) * sizeof(*lep));
545 lep = &leaf->ents[index];
546 lfloglow = MIN(index, lfloglow);
547 lfloghigh = MAX(highstale, lfloghigh);
548 }
549 INT_MOD(leaf->hdr.stale, ARCH_CONVERT, -1);
550 }
551 /*
552 * Fill in the new leaf entry.
553 */
554 INT_SET(lep->hashval, ARCH_CONVERT, args->hashval);
555 INT_SET(lep->address, ARCH_CONVERT, XFS_DIR2_DB_OFF_TO_DATAPTR(mp, use_block, INT_GET(*tagp, ARCH_CONVERT)));
556 /*
557 * Log the leaf fields and give up the buffers.
558 */
559 xfs_dir2_leaf_log_header(tp, lbp);
560 xfs_dir2_leaf_log_ents(tp, lbp, lfloglow, lfloghigh);
561 xfs_dir2_leaf_check(dp, lbp);
562 xfs_da_buf_done(lbp);
563 xfs_dir2_data_check(dp, dbp);
564 xfs_da_buf_done(dbp);
565 return 0;
566 }
567
568 #ifdef DEBUG
569 /*
570 * Check the internal consistency of a leaf1 block.
571 * Pop an assert if something is wrong.
572 */
573 void
574 xfs_dir2_leaf_check(
575 xfs_inode_t *dp, /* incore directory inode */
576 xfs_dabuf_t *bp) /* leaf's buffer */
577 {
578 int i; /* leaf index */
579 xfs_dir2_leaf_t *leaf; /* leaf structure */
580 xfs_dir2_leaf_tail_t *ltp; /* leaf tail pointer */
581 xfs_mount_t *mp; /* filesystem mount point */
582 int stale; /* count of stale leaves */
583
584 leaf = bp->data;
585 mp = dp->i_mount;
586 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
587 /*
588 * This value is not restrictive enough.
589 * Should factor in the size of the bests table as well.
590 * We can deduce a value for that from di_size.
591 */
592 ASSERT(INT_GET(leaf->hdr.count, ARCH_CONVERT) <= XFS_DIR2_MAX_LEAF_ENTS(mp));
593 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
594 /*
595 * Leaves and bests don't overlap.
596 */
597 ASSERT((char *)&leaf->ents[INT_GET(leaf->hdr.count, ARCH_CONVERT)] <=
598 (char *)XFS_DIR2_LEAF_BESTS_P(ltp));
599 /*
600 * Check hash value order, count stale entries.
601 */
602 for (i = stale = 0; i < INT_GET(leaf->hdr.count, ARCH_CONVERT); i++) {
603 if (i + 1 < INT_GET(leaf->hdr.count, ARCH_CONVERT))
604 ASSERT(INT_GET(leaf->ents[i].hashval, ARCH_CONVERT) <=
605 INT_GET(leaf->ents[i + 1].hashval, ARCH_CONVERT));
606 if (INT_GET(leaf->ents[i].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
607 stale++;
608 }
609 ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) == stale);
610 }
611 #endif /* DEBUG */
612
613 /*
614 * Compact out any stale entries in the leaf.
615 * Log the header and changed leaf entries, if any.
616 */
617 void
618 xfs_dir2_leaf_compact(
619 xfs_da_args_t *args, /* operation arguments */
620 xfs_dabuf_t *bp) /* leaf buffer */
621 {
622 int from; /* source leaf index */
623 xfs_dir2_leaf_t *leaf; /* leaf structure */
624 int loglow; /* first leaf entry to log */
625 int to; /* target leaf index */
626
627 leaf = bp->data;
628 if (!leaf->hdr.stale) {
629 return;
630 }
631 /*
632 * Compress out the stale entries in place.
633 */
634 for (from = to = 0, loglow = -1; from < INT_GET(leaf->hdr.count, ARCH_CONVERT); from++) {
635 if (INT_GET(leaf->ents[from].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
636 continue;
637 /*
638 * Only actually copy the entries that are different.
639 */
640 if (from > to) {
641 if (loglow == -1)
642 loglow = to;
643 leaf->ents[to] = leaf->ents[from];
644 }
645 to++;
646 }
647 /*
648 * Update and log the header, log the leaf entries.
649 */
650 ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) == from - to);
651 INT_MOD(leaf->hdr.count, ARCH_CONVERT, -(INT_GET(leaf->hdr.stale, ARCH_CONVERT)));
652 leaf->hdr.stale = 0;
653 xfs_dir2_leaf_log_header(args->trans, bp);
654 if (loglow != -1)
655 xfs_dir2_leaf_log_ents(args->trans, bp, loglow, to - 1);
656 }
657
658 /*
659 * Compact the leaf entries, removing stale ones.
660 * Leave one stale entry behind - the one closest to our
661 * insertion index - and the caller will shift that one to our insertion
662 * point later.
663 * Return new insertion index, where the remaining stale entry is,
664 * and leaf logging indices.
665 */
666 void
667 xfs_dir2_leaf_compact_x1(
668 xfs_dabuf_t *bp, /* leaf buffer */
669 int *indexp, /* insertion index */
670 int *lowstalep, /* out: stale entry before us */
671 int *highstalep, /* out: stale entry after us */
672 int *lowlogp, /* out: low log index */
673 int *highlogp) /* out: high log index */
674 {
675 int from; /* source copy index */
676 int highstale; /* stale entry at/after index */
677 int index; /* insertion index */
678 int keepstale; /* source index of kept stale */
679 xfs_dir2_leaf_t *leaf; /* leaf structure */
680 int lowstale; /* stale entry before index */
681 int newindex=0; /* new insertion index */
682 int to; /* destination copy index */
683
684 leaf = bp->data;
685 ASSERT(INT_GET(leaf->hdr.stale, ARCH_CONVERT) > 1);
686 index = *indexp;
687 /*
688 * Find the first stale entry before our index, if any.
689 */
690 for (lowstale = index - 1;
691 lowstale >= 0 &&
692 INT_GET(leaf->ents[lowstale].address, ARCH_CONVERT) != XFS_DIR2_NULL_DATAPTR;
693 lowstale--)
694 continue;
695 /*
696 * Find the first stale entry at or after our index, if any.
697 * Stop if the answer would be worse than lowstale.
698 */
699 for (highstale = index;
700 highstale < INT_GET(leaf->hdr.count, ARCH_CONVERT) &&
701 INT_GET(leaf->ents[highstale].address, ARCH_CONVERT) != XFS_DIR2_NULL_DATAPTR &&
702 (lowstale < 0 || index - lowstale > highstale - index);
703 highstale++)
704 continue;
705 /*
706 * Pick the better of lowstale and highstale.
707 */
708 if (lowstale >= 0 &&
709 (highstale == INT_GET(leaf->hdr.count, ARCH_CONVERT) ||
710 index - lowstale <= highstale - index))
711 keepstale = lowstale;
712 else
713 keepstale = highstale;
714 /*
715 * Copy the entries in place, removing all the stale entries
716 * except keepstale.
717 */
718 for (from = to = 0; from < INT_GET(leaf->hdr.count, ARCH_CONVERT); from++) {
719 /*
720 * Notice the new value of index.
721 */
722 if (index == from)
723 newindex = to;
724 if (from != keepstale &&
725 INT_GET(leaf->ents[from].address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR) {
726 if (from == to)
727 *lowlogp = to;
728 continue;
729 }
730 /*
731 * Record the new keepstale value for the insertion.
732 */
733 if (from == keepstale)
734 lowstale = highstale = to;
735 /*
736 * Copy only the entries that have moved.
737 */
738 if (from > to)
739 leaf->ents[to] = leaf->ents[from];
740 to++;
741 }
742 ASSERT(from > to);
743 /*
744 * If the insertion point was past the last entry,
745 * set the new insertion point accordingly.
746 */
747 if (index == from)
748 newindex = to;
749 *indexp = newindex;
750 /*
751 * Adjust the leaf header values.
752 */
753 INT_MOD(leaf->hdr.count, ARCH_CONVERT, -(from - to));
754 INT_SET(leaf->hdr.stale, ARCH_CONVERT, 1);
755 /*
756 * Remember the low/high stale value only in the "right"
757 * direction.
758 */
759 if (lowstale >= newindex)
760 lowstale = -1;
761 else
762 highstale = INT_GET(leaf->hdr.count, ARCH_CONVERT);
763 *highlogp = INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1;
764 *lowstalep = lowstale;
765 *highstalep = highstale;
766 }
767
768 /*
769 * Getdents (readdir) for leaf and node directories.
770 * This reads the data blocks only, so is the same for both forms.
771 */
772 int /* error */
773 xfs_dir2_leaf_getdents(
774 xfs_trans_t *tp, /* transaction pointer */
775 xfs_inode_t *dp, /* incore directory inode */
776 uio_t *uio, /* I/O control & vectors */
777 int *eofp, /* out: reached end of dir */
778 xfs_dirent_t *dbp, /* caller's buffer */
779 xfs_dir2_put_t put) /* ABI formatting routine */
780 {
781 xfs_dabuf_t *bp; /* data block buffer */
782 int byteoff; /* offset in current block */
783 xfs_dir2_db_t curdb; /* db for current block */
784 xfs_dir2_off_t curoff; /* current overall offset */
785 xfs_dir2_data_t *data; /* data block structure */
786 xfs_dir2_data_entry_t *dep; /* data entry */
787 xfs_dir2_data_unused_t *dup; /* unused entry */
788 int eof; /* reached end of directory */
789 int error=0; /* error return value */
790 int i; /* temporary loop index */
791 int j; /* temporary loop index */
792 int length; /* temporary length value */
793 xfs_bmbt_irec_t *map; /* map vector for blocks */
794 xfs_extlen_t map_blocks; /* number of fsbs in map */
795 xfs_dablk_t map_off; /* last mapped file offset */
796 int map_size; /* total entries in *map */
797 int map_valid; /* valid entries in *map */
798 xfs_mount_t *mp; /* filesystem mount point */
799 xfs_dir2_off_t newoff; /* new curoff after new blk */
800 int nmap; /* mappings to ask xfs_bmapi */
801 xfs_dir2_put_args_t p; /* formatting arg bundle */
802 char *ptr=NULL; /* pointer to current data */
803 int ra_current; /* number of read-ahead blks */
804 int ra_index; /* *map index for read-ahead */
805 int ra_offset; /* map entry offset for ra */
806 int ra_want; /* readahead count wanted */
807
808 /*
809 * If the offset is at or past the largest allowed value,
810 * give up right away, return eof.
811 */
812 if (uio->uio_offset >= XFS_DIR2_MAX_DATAPTR) {
813 *eofp = 1;
814 return 0;
815 }
816 mp = dp->i_mount;
817 /*
818 * Setup formatting arguments.
819 */
820 p.dbp = dbp;
821 p.put = put;
822 p.uio = uio;
823 /*
824 * Set up to bmap a number of blocks based on the caller's
825 * buffer size, the directory block size, and the filesystem
826 * block size.
827 */
828 map_size =
829 howmany(uio->uio_resid + mp->m_dirblksize,
830 mp->m_sb.sb_blocksize);
831 map = kmem_alloc(map_size * sizeof(*map), KM_SLEEP);
832 map_valid = ra_index = ra_offset = ra_current = map_blocks = 0;
833 bp = NULL;
834 eof = 1;
835 /*
836 * Inside the loop we keep the main offset value as a byte offset
837 * in the directory file.
838 */
839 curoff = XFS_DIR2_DATAPTR_TO_BYTE(mp, uio->uio_offset);
840 /*
841 * Force this conversion through db so we truncate the offset
842 * down to get the start of the data block.
843 */
844 map_off = XFS_DIR2_DB_TO_DA(mp, XFS_DIR2_BYTE_TO_DB(mp, curoff));
845 /*
846 * Loop over directory entries until we reach the end offset.
847 * Get more blocks and readahead as necessary.
848 */
849 while (curoff < XFS_DIR2_LEAF_OFFSET) {
850 /*
851 * If we have no buffer, or we're off the end of the
852 * current buffer, need to get another one.
853 */
854 if (!bp || ptr >= (char *)bp->data + mp->m_dirblksize) {
855 /*
856 * If we have a buffer, we need to release it and
857 * take it out of the mapping.
858 */
859 if (bp) {
860 xfs_da_brelse(tp, bp);
861 bp = NULL;
862 map_blocks -= mp->m_dirblkfsbs;
863 /*
864 * Loop to get rid of the extents for the
865 * directory block.
866 */
867 for (i = mp->m_dirblkfsbs; i > 0; ) {
868 j = MIN((int)map->br_blockcount, i);
869 map->br_blockcount -= j;
870 map->br_startblock += j;
871 map->br_startoff += j;
872 /*
873 * If mapping is done, pitch it from
874 * the table.
875 */
876 if (!map->br_blockcount && --map_valid)
877 memmove(&map[0], &map[1],
878 sizeof(map[0]) *
879 map_valid);
880 i -= j;
881 }
882 }
883 /*
884 * Recalculate the readahead blocks wanted.
885 */
886 ra_want = howmany(uio->uio_resid + mp->m_dirblksize,
887 mp->m_sb.sb_blocksize) - 1;
888 /*
889 * If we don't have as many as we want, and we haven't
890 * run out of data blocks, get some more mappings.
891 */
892 if (1 + ra_want > map_blocks &&
893 map_off <
894 XFS_DIR2_BYTE_TO_DA(mp, XFS_DIR2_LEAF_OFFSET)) {
895 /*
896 * Get more bmaps, fill in after the ones
897 * we already have in the table.
898 */
899 nmap = map_size - map_valid;
900 error = xfs_bmapi(tp, dp,
901 map_off,
902 XFS_DIR2_BYTE_TO_DA(mp,
903 XFS_DIR2_LEAF_OFFSET) - map_off,
904 XFS_BMAPI_METADATA, NULL, 0,
905 &map[map_valid], &nmap, NULL);
906 /*
907 * Don't know if we should ignore this or
908 * try to return an error.
909 * The trouble with returning errors
910 * is that readdir will just stop without
911 * actually passing the error through.
912 */
913 if (error)
914 break; /* XXX */
915 /*
916 * If we got all the mappings we asked for,
917 * set the final map offset based on the
918 * last bmap value received.
919 * Otherwise, we've reached the end.
920 */
921 if (nmap == map_size - map_valid)
922 map_off =
923 map[map_valid + nmap - 1].br_startoff +
924 map[map_valid + nmap - 1].br_blockcount;
925 else
926 map_off =
927 XFS_DIR2_BYTE_TO_DA(mp,
928 XFS_DIR2_LEAF_OFFSET);
929 /*
930 * Look for holes in the mapping, and
931 * eliminate them. Count up the valid blocks.
932 */
933 for (i = map_valid; i < map_valid + nmap; ) {
934 if (map[i].br_startblock ==
935 HOLESTARTBLOCK) {
936 nmap--;
937 length = map_valid + nmap - i;
938 if (length)
939 memmove(&map[i],
940 &map[i + 1],
941 sizeof(map[i]) *
942 length);
943 } else {
944 map_blocks +=
945 map[i].br_blockcount;
946 i++;
947 }
948 }
949 map_valid += nmap;
950 }
951 /*
952 * No valid mappings, so no more data blocks.
953 */
954 if (!map_valid) {
955 curoff = XFS_DIR2_DA_TO_BYTE(mp, map_off);
956 break;
957 }
958 /*
959 * Read the directory block starting at the first
960 * mapping.
961 */
962 curdb = XFS_DIR2_DA_TO_DB(mp, map->br_startoff);
963 error = xfs_da_read_buf(tp, dp, map->br_startoff,
964 map->br_blockcount >= mp->m_dirblkfsbs ?
965 XFS_FSB_TO_DADDR(mp, map->br_startblock) :
966 -1,
967 &bp, XFS_DATA_FORK);
968 /*
969 * Should just skip over the data block instead
970 * of giving up.
971 */
972 if (error)
973 break; /* XXX */
974 /*
975 * Adjust the current amount of read-ahead: we just
976 * read a block that was previously ra.
977 */
978 if (ra_current)
979 ra_current -= mp->m_dirblkfsbs;
980 /*
981 * Do we need more readahead?
982 */
983 for (ra_index = ra_offset = i = 0;
984 ra_want > ra_current && i < map_blocks;
985 i += mp->m_dirblkfsbs) {
986 ASSERT(ra_index < map_valid);
987 /*
988 * Read-ahead a contiguous directory block.
989 */
990 if (i > ra_current &&
991 map[ra_index].br_blockcount >=
992 mp->m_dirblkfsbs) {
993 xfs_baread(mp->m_ddev_targp,
994 XFS_FSB_TO_DADDR(mp,
995 map[ra_index].br_startblock +
996 ra_offset),
997 (int)BTOBB(mp->m_dirblksize));
998 ra_current = i;
999 }
1000 /*
1001 * Read-ahead a non-contiguous directory block.
1002 * This doesn't use our mapping, but this
1003 * is a very rare case.
1004 */
1005 else if (i > ra_current) {
1006 (void)xfs_da_reada_buf(tp, dp,
1007 map[ra_index].br_startoff +
1008 ra_offset, XFS_DATA_FORK);
1009 ra_current = i;
1010 }
1011 /*
1012 * Advance offset through the mapping table.
1013 */
1014 for (j = 0; j < mp->m_dirblkfsbs; j++) {
1015 /*
1016 * The rest of this extent but not
1017 * more than a dir block.
1018 */
1019 length = MIN(mp->m_dirblkfsbs,
1020 (int)(map[ra_index].br_blockcount -
1021 ra_offset));
1022 j += length;
1023 ra_offset += length;
1024 /*
1025 * Advance to the next mapping if
1026 * this one is used up.
1027 */
1028 if (ra_offset ==
1029 map[ra_index].br_blockcount) {
1030 ra_offset = 0;
1031 ra_index++;
1032 }
1033 }
1034 }
1035 /*
1036 * Having done a read, we need to set a new offset.
1037 */
1038 newoff = XFS_DIR2_DB_OFF_TO_BYTE(mp, curdb, 0);
1039 /*
1040 * Start of the current block.
1041 */
1042 if (curoff < newoff)
1043 curoff = newoff;
1044 /*
1045 * Make sure we're in the right block.
1046 */
1047 else if (curoff > newoff)
1048 ASSERT(XFS_DIR2_BYTE_TO_DB(mp, curoff) ==
1049 curdb);
1050 data = bp->data;
1051 xfs_dir2_data_check(dp, bp);
1052 /*
1053 * Find our position in the block.
1054 */
1055 ptr = (char *)&data->u;
1056 byteoff = XFS_DIR2_BYTE_TO_OFF(mp, curoff);
1057 /*
1058 * Skip past the header.
1059 */
1060 if (byteoff == 0)
1061 curoff += (uint)sizeof(data->hdr);
1062 /*
1063 * Skip past entries until we reach our offset.
1064 */
1065 else {
1066 while ((char *)ptr - (char *)data < byteoff) {
1067 dup = (xfs_dir2_data_unused_t *)ptr;
1068
1069 if (INT_GET(dup->freetag, ARCH_CONVERT)
1070 == XFS_DIR2_DATA_FREE_TAG) {
1071
1072 length = INT_GET(dup->length,
1073 ARCH_CONVERT);
1074 ptr += length;
1075 continue;
1076 }
1077 dep = (xfs_dir2_data_entry_t *)ptr;
1078 length =
1079 XFS_DIR2_DATA_ENTSIZE(dep->namelen);
1080 ptr += length;
1081 }
1082 /*
1083 * Now set our real offset.
1084 */
1085 curoff =
1086 XFS_DIR2_DB_OFF_TO_BYTE(mp,
1087 XFS_DIR2_BYTE_TO_DB(mp, curoff),
1088 (char *)ptr - (char *)data);
1089 if (ptr >= (char *)data + mp->m_dirblksize) {
1090 continue;
1091 }
1092 }
1093 }
1094 /*
1095 * We have a pointer to an entry.
1096 * Is it a live one?
1097 */
1098 dup = (xfs_dir2_data_unused_t *)ptr;
1099 /*
1100 * No, it's unused, skip over it.
1101 */
1102 if (INT_GET(dup->freetag, ARCH_CONVERT)
1103 == XFS_DIR2_DATA_FREE_TAG) {
1104 length = INT_GET(dup->length, ARCH_CONVERT);
1105 ptr += length;
1106 curoff += length;
1107 continue;
1108 }
1109
1110 /*
1111 * Copy the entry into the putargs, and try formatting it.
1112 */
1113 dep = (xfs_dir2_data_entry_t *)ptr;
1114
1115 p.namelen = dep->namelen;
1116
1117 length = XFS_DIR2_DATA_ENTSIZE(p.namelen);
1118
1119 p.cook = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff + length);
1120
1121 p.ino = INT_GET(dep->inumber, ARCH_CONVERT);
1122 #if XFS_BIG_INUMS
1123 p.ino += mp->m_inoadd;
1124 #endif
1125 p.name = (char *)dep->name;
1126
1127 error = p.put(&p);
1128
1129 /*
1130 * Won't fit. Return to caller.
1131 */
1132 if (!p.done) {
1133 eof = 0;
1134 break;
1135 }
1136 /*
1137 * Advance to next entry in the block.
1138 */
1139 ptr += length;
1140 curoff += length;
1141 }
1142
1143 /*
1144 * All done. Set output offset value to current offset.
1145 */
1146 *eofp = eof;
1147 if (curoff > XFS_DIR2_DATAPTR_TO_BYTE(mp, XFS_DIR2_MAX_DATAPTR))
1148 uio->uio_offset = XFS_DIR2_MAX_DATAPTR;
1149 else
1150 uio->uio_offset = XFS_DIR2_BYTE_TO_DATAPTR(mp, curoff);
1151 kmem_free(map, map_size * sizeof(*map));
1152 if (bp)
1153 xfs_da_brelse(tp, bp);
1154 return error;
1155 }
1156
1157 /*
1158 * Initialize a new leaf block, leaf1 or leafn magic accepted.
1159 */
1160 int
1161 xfs_dir2_leaf_init(
1162 xfs_da_args_t *args, /* operation arguments */
1163 xfs_dir2_db_t bno, /* directory block number */
1164 xfs_dabuf_t **bpp, /* out: leaf buffer */
1165 int magic) /* magic number for block */
1166 {
1167 xfs_dabuf_t *bp; /* leaf buffer */
1168 xfs_inode_t *dp; /* incore directory inode */
1169 int error; /* error return code */
1170 xfs_dir2_leaf_t *leaf; /* leaf structure */
1171 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1172 xfs_mount_t *mp; /* filesystem mount point */
1173 xfs_trans_t *tp; /* transaction pointer */
1174
1175 dp = args->dp;
1176 ASSERT(dp != NULL);
1177 tp = args->trans;
1178 mp = dp->i_mount;
1179 ASSERT(bno >= XFS_DIR2_LEAF_FIRSTDB(mp) &&
1180 bno < XFS_DIR2_FREE_FIRSTDB(mp));
1181 /*
1182 * Get the buffer for the block.
1183 */
1184 error = xfs_da_get_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, bno), -1, &bp,
1185 XFS_DATA_FORK);
1186 if (error) {
1187 return error;
1188 }
1189 ASSERT(bp != NULL);
1190 leaf = bp->data;
1191 /*
1192 * Initialize the header.
1193 */
1194 INT_SET(leaf->hdr.info.magic, ARCH_CONVERT, magic);
1195 leaf->hdr.info.forw = 0;
1196 leaf->hdr.info.back = 0;
1197 leaf->hdr.count = 0;
1198 leaf->hdr.stale = 0;
1199 xfs_dir2_leaf_log_header(tp, bp);
1200 /*
1201 * If it's a leaf-format directory initialize the tail.
1202 * In this case our caller has the real bests table to copy into
1203 * the block.
1204 */
1205 if (magic == XFS_DIR2_LEAF1_MAGIC) {
1206 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1207 ltp->bestcount = 0;
1208 xfs_dir2_leaf_log_tail(tp, bp);
1209 }
1210 *bpp = bp;
1211 return 0;
1212 }
1213
1214 /*
1215 * Log the bests entries indicated from a leaf1 block.
1216 */
1217 void
1218 xfs_dir2_leaf_log_bests(
1219 xfs_trans_t *tp, /* transaction pointer */
1220 xfs_dabuf_t *bp, /* leaf buffer */
1221 int first, /* first entry to log */
1222 int last) /* last entry to log */
1223 {
1224 xfs_dir2_data_off_t *firstb; /* pointer to first entry */
1225 xfs_dir2_data_off_t *lastb; /* pointer to last entry */
1226 xfs_dir2_leaf_t *leaf; /* leaf structure */
1227 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1228
1229 leaf = bp->data;
1230 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
1231 ltp = XFS_DIR2_LEAF_TAIL_P(tp->t_mountp, leaf);
1232 firstb = XFS_DIR2_LEAF_BESTS_P(ltp) + first;
1233 lastb = XFS_DIR2_LEAF_BESTS_P(ltp) + last;
1234 xfs_da_log_buf(tp, bp, (uint)((char *)firstb - (char *)leaf),
1235 (uint)((char *)lastb - (char *)leaf + sizeof(*lastb) - 1));
1236 }
1237
1238 /*
1239 * Log the leaf entries indicated from a leaf1 or leafn block.
1240 */
1241 void
1242 xfs_dir2_leaf_log_ents(
1243 xfs_trans_t *tp, /* transaction pointer */
1244 xfs_dabuf_t *bp, /* leaf buffer */
1245 int first, /* first entry to log */
1246 int last) /* last entry to log */
1247 {
1248 xfs_dir2_leaf_entry_t *firstlep; /* pointer to first entry */
1249 xfs_dir2_leaf_entry_t *lastlep; /* pointer to last entry */
1250 xfs_dir2_leaf_t *leaf; /* leaf structure */
1251
1252 leaf = bp->data;
1253 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC ||
1254 INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1255 firstlep = &leaf->ents[first];
1256 lastlep = &leaf->ents[last];
1257 xfs_da_log_buf(tp, bp, (uint)((char *)firstlep - (char *)leaf),
1258 (uint)((char *)lastlep - (char *)leaf + sizeof(*lastlep) - 1));
1259 }
1260
1261 /*
1262 * Log the header of the leaf1 or leafn block.
1263 */
1264 void
1265 xfs_dir2_leaf_log_header(
1266 xfs_trans_t *tp, /* transaction pointer */
1267 xfs_dabuf_t *bp) /* leaf buffer */
1268 {
1269 xfs_dir2_leaf_t *leaf; /* leaf structure */
1270
1271 leaf = bp->data;
1272 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC ||
1273 INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1274 xfs_da_log_buf(tp, bp, (uint)((char *)&leaf->hdr - (char *)leaf),
1275 (uint)(sizeof(leaf->hdr) - 1));
1276 }
1277
1278 /*
1279 * Log the tail of the leaf1 block.
1280 */
1281 void
1282 xfs_dir2_leaf_log_tail(
1283 xfs_trans_t *tp, /* transaction pointer */
1284 xfs_dabuf_t *bp) /* leaf buffer */
1285 {
1286 xfs_dir2_leaf_t *leaf; /* leaf structure */
1287 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1288 xfs_mount_t *mp; /* filesystem mount point */
1289
1290 mp = tp->t_mountp;
1291 leaf = bp->data;
1292 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAF1_MAGIC);
1293 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1294 xfs_da_log_buf(tp, bp, (uint)((char *)ltp - (char *)leaf),
1295 (uint)(mp->m_dirblksize - 1));
1296 }
1297
1298 /*
1299 * Look up the entry referred to by args in the leaf format directory.
1300 * Most of the work is done by the xfs_dir2_leaf_lookup_int routine which
1301 * is also used by the node-format code.
1302 */
1303 int
1304 xfs_dir2_leaf_lookup(
1305 xfs_da_args_t *args) /* operation arguments */
1306 {
1307 xfs_dabuf_t *dbp; /* data block buffer */
1308 xfs_dir2_data_entry_t *dep; /* data block entry */
1309 xfs_inode_t *dp; /* incore directory inode */
1310 int error; /* error return code */
1311 int index; /* found entry index */
1312 xfs_dabuf_t *lbp; /* leaf buffer */
1313 xfs_dir2_leaf_t *leaf; /* leaf structure */
1314 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1315 xfs_trans_t *tp; /* transaction pointer */
1316
1317 xfs_dir2_trace_args("leaf_lookup", args);
1318 /*
1319 * Look up name in the leaf block, returning both buffers and index.
1320 */
1321 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1322 return error;
1323 }
1324 tp = args->trans;
1325 dp = args->dp;
1326 xfs_dir2_leaf_check(dp, lbp);
1327 leaf = lbp->data;
1328 /*
1329 * Get to the leaf entry and contained data entry address.
1330 */
1331 lep = &leaf->ents[index];
1332 /*
1333 * Point to the data entry.
1334 */
1335 dep = (xfs_dir2_data_entry_t *)
1336 ((char *)dbp->data +
1337 XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, INT_GET(lep->address, ARCH_CONVERT)));
1338 /*
1339 * Return the found inode number.
1340 */
1341 args->inumber = INT_GET(dep->inumber, ARCH_CONVERT);
1342 xfs_da_brelse(tp, dbp);
1343 xfs_da_brelse(tp, lbp);
1344 return XFS_ERROR(EEXIST);
1345 }
1346
1347 /*
1348 * Look up name/hash in the leaf block.
1349 * Fill in indexp with the found index, and dbpp with the data buffer.
1350 * If not found dbpp will be NULL, and ENOENT comes back.
1351 * lbpp will always be filled in with the leaf buffer unless there's an error.
1352 */
1353 static int /* error */
1354 xfs_dir2_leaf_lookup_int(
1355 xfs_da_args_t *args, /* operation arguments */
1356 xfs_dabuf_t **lbpp, /* out: leaf buffer */
1357 int *indexp, /* out: index in leaf block */
1358 xfs_dabuf_t **dbpp) /* out: data buffer */
1359 {
1360 xfs_dir2_db_t curdb; /* current data block number */
1361 xfs_dabuf_t *dbp; /* data buffer */
1362 xfs_dir2_data_entry_t *dep; /* data entry */
1363 xfs_inode_t *dp; /* incore directory inode */
1364 int error; /* error return code */
1365 int index; /* index in leaf block */
1366 xfs_dabuf_t *lbp; /* leaf buffer */
1367 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1368 xfs_dir2_leaf_t *leaf; /* leaf structure */
1369 xfs_mount_t *mp; /* filesystem mount point */
1370 xfs_dir2_db_t newdb; /* new data block number */
1371 xfs_trans_t *tp; /* transaction pointer */
1372
1373 dp = args->dp;
1374 tp = args->trans;
1375 mp = dp->i_mount;
1376 /*
1377 * Read the leaf block into the buffer.
1378 */
1379 if ((error =
1380 xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
1381 XFS_DATA_FORK))) {
1382 return error;
1383 }
1384 *lbpp = lbp;
1385 leaf = lbp->data;
1386 xfs_dir2_leaf_check(dp, lbp);
1387 /*
1388 * Look for the first leaf entry with our hash value.
1389 */
1390 index = xfs_dir2_leaf_search_hash(args, lbp);
1391 /*
1392 * Loop over all the entries with the right hash value
1393 * looking to match the name.
1394 */
1395 for (lep = &leaf->ents[index], dbp = NULL, curdb = -1;
1396 index < INT_GET(leaf->hdr.count, ARCH_CONVERT) && INT_GET(lep->hashval, ARCH_CONVERT) == args->hashval;
1397 lep++, index++) {
1398 /*
1399 * Skip over stale leaf entries.
1400 */
1401 if (INT_GET(lep->address, ARCH_CONVERT) == XFS_DIR2_NULL_DATAPTR)
1402 continue;
1403 /*
1404 * Get the new data block number.
1405 */
1406 newdb = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
1407 /*
1408 * If it's not the same as the old data block number,
1409 * need to pitch the old one and read the new one.
1410 */
1411 if (newdb != curdb) {
1412 if (dbp)
1413 xfs_da_brelse(tp, dbp);
1414 if ((error =
1415 xfs_da_read_buf(tp, dp,
1416 XFS_DIR2_DB_TO_DA(mp, newdb), -1, &dbp,
1417 XFS_DATA_FORK))) {
1418 xfs_da_brelse(tp, lbp);
1419 return error;
1420 }
1421 xfs_dir2_data_check(dp, dbp);
1422 curdb = newdb;
1423 }
1424 /*
1425 * Point to the data entry.
1426 */
1427 dep = (xfs_dir2_data_entry_t *)
1428 ((char *)dbp->data +
1429 XFS_DIR2_DATAPTR_TO_OFF(mp, INT_GET(lep->address, ARCH_CONVERT)));
1430 /*
1431 * If it matches then return it.
1432 */
1433 if (dep->namelen == args->namelen &&
1434 dep->name[0] == args->name[0] &&
1435 memcmp(dep->name, args->name, args->namelen) == 0) {
1436 *dbpp = dbp;
1437 *indexp = index;
1438 return 0;
1439 }
1440 }
1441 /*
1442 * No match found, return ENOENT.
1443 */
1444 ASSERT(args->oknoent);
1445 if (dbp)
1446 xfs_da_brelse(tp, dbp);
1447 xfs_da_brelse(tp, lbp);
1448 return XFS_ERROR(ENOENT);
1449 }
1450
1451 /*
1452 * Remove an entry from a leaf format directory.
1453 */
1454 int /* error */
1455 xfs_dir2_leaf_removename(
1456 xfs_da_args_t *args) /* operation arguments */
1457 {
1458 xfs_dir2_data_off_t *bestsp; /* leaf block best freespace */
1459 xfs_dir2_data_t *data; /* data block structure */
1460 xfs_dir2_db_t db; /* data block number */
1461 xfs_dabuf_t *dbp; /* data block buffer */
1462 xfs_dir2_data_entry_t *dep; /* data entry structure */
1463 xfs_inode_t *dp; /* incore directory inode */
1464 int error; /* error return code */
1465 xfs_dir2_db_t i; /* temporary data block # */
1466 int index; /* index into leaf entries */
1467 xfs_dabuf_t *lbp; /* leaf buffer */
1468 xfs_dir2_leaf_t *leaf; /* leaf structure */
1469 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1470 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1471 xfs_mount_t *mp; /* filesystem mount point */
1472 int needlog; /* need to log data header */
1473 int needscan; /* need to rescan data frees */
1474 xfs_dir2_data_off_t oldbest; /* old value of best free */
1475 xfs_trans_t *tp; /* transaction pointer */
1476
1477 xfs_dir2_trace_args("leaf_removename", args);
1478 /*
1479 * Lookup the leaf entry, get the leaf and data blocks read in.
1480 */
1481 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1482 return error;
1483 }
1484 dp = args->dp;
1485 tp = args->trans;
1486 mp = dp->i_mount;
1487 leaf = lbp->data;
1488 data = dbp->data;
1489 xfs_dir2_data_check(dp, dbp);
1490 /*
1491 * Point to the leaf entry, use that to point to the data entry.
1492 */
1493 lep = &leaf->ents[index];
1494 db = XFS_DIR2_DATAPTR_TO_DB(mp, INT_GET(lep->address, ARCH_CONVERT));
1495 dep = (xfs_dir2_data_entry_t *)
1496 ((char *)data + XFS_DIR2_DATAPTR_TO_OFF(mp, INT_GET(lep->address, ARCH_CONVERT)));
1497 needscan = needlog = 0;
1498 oldbest = INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT);
1499 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1500 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1501 ASSERT(INT_GET(bestsp[db], ARCH_CONVERT) == oldbest);
1502 /*
1503 * Mark the former data entry unused.
1504 */
1505 xfs_dir2_data_make_free(tp, dbp,
1506 (xfs_dir2_data_aoff_t)((char *)dep - (char *)data),
1507 XFS_DIR2_DATA_ENTSIZE(dep->namelen), &needlog, &needscan);
1508 /*
1509 * We just mark the leaf entry stale by putting a null in it.
1510 */
1511 INT_MOD(leaf->hdr.stale, ARCH_CONVERT, +1);
1512 xfs_dir2_leaf_log_header(tp, lbp);
1513 INT_SET(lep->address, ARCH_CONVERT, XFS_DIR2_NULL_DATAPTR);
1514 xfs_dir2_leaf_log_ents(tp, lbp, index, index);
1515 /*
1516 * Scan the freespace in the data block again if necessary,
1517 * log the data block header if necessary.
1518 */
1519 if (needscan)
1520 xfs_dir2_data_freescan(mp, data, &needlog, NULL);
1521 if (needlog)
1522 xfs_dir2_data_log_header(tp, dbp);
1523 /*
1524 * If the longest freespace in the data block has changed,
1525 * put the new value in the bests table and log that.
1526 */
1527 if (INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) != oldbest) {
1528 INT_COPY(bestsp[db], data->hdr.bestfree[0].length, ARCH_CONVERT);
1529 xfs_dir2_leaf_log_bests(tp, lbp, db, db);
1530 }
1531 xfs_dir2_data_check(dp, dbp);
1532 /*
1533 * If the data block is now empty then get rid of the data block.
1534 */
1535 if (INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) ==
1536 mp->m_dirblksize - (uint)sizeof(data->hdr)) {
1537 ASSERT(db != mp->m_dirdatablk);
1538 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1539 /*
1540 * Nope, can't get rid of it because it caused
1541 * allocation of a bmap btree block to do so.
1542 * Just go on, returning success, leaving the
1543 * empty block in place.
1544 */
1545 if (error == ENOSPC && args->total == 0) {
1546 xfs_da_buf_done(dbp);
1547 error = 0;
1548 }
1549 xfs_dir2_leaf_check(dp, lbp);
1550 xfs_da_buf_done(lbp);
1551 return error;
1552 }
1553 dbp = NULL;
1554 /*
1555 * If this is the last data block then compact the
1556 * bests table by getting rid of entries.
1557 */
1558 if (db == INT_GET(ltp->bestcount, ARCH_CONVERT) - 1) {
1559 /*
1560 * Look for the last active entry (i).
1561 */
1562 for (i = db - 1; i > 0; i--) {
1563 if (INT_GET(bestsp[i], ARCH_CONVERT) != NULLDATAOFF)
1564 break;
1565 }
1566 /*
1567 * Copy the table down so inactive entries at the
1568 * end are removed.
1569 */
1570 memmove(&bestsp[db - i], bestsp,
1571 (INT_GET(ltp->bestcount, ARCH_CONVERT) - (db - i)) * sizeof(*bestsp));
1572 INT_MOD(ltp->bestcount, ARCH_CONVERT, -(db - i));
1573 xfs_dir2_leaf_log_tail(tp, lbp);
1574 xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1575 } else
1576 INT_SET(bestsp[db], ARCH_CONVERT, NULLDATAOFF);
1577 }
1578 /*
1579 * If the data block was not the first one, drop it.
1580 */
1581 else if (db != mp->m_dirdatablk && dbp != NULL) {
1582 xfs_da_buf_done(dbp);
1583 dbp = NULL;
1584 }
1585 xfs_dir2_leaf_check(dp, lbp);
1586 /*
1587 * See if we can convert to block form.
1588 */
1589 return xfs_dir2_leaf_to_block(args, lbp, dbp);
1590 }
1591
1592 /*
1593 * Replace the inode number in a leaf format directory entry.
1594 */
1595 int /* error */
1596 xfs_dir2_leaf_replace(
1597 xfs_da_args_t *args) /* operation arguments */
1598 {
1599 xfs_dabuf_t *dbp; /* data block buffer */
1600 xfs_dir2_data_entry_t *dep; /* data block entry */
1601 xfs_inode_t *dp; /* incore directory inode */
1602 int error; /* error return code */
1603 int index; /* index of leaf entry */
1604 xfs_dabuf_t *lbp; /* leaf buffer */
1605 xfs_dir2_leaf_t *leaf; /* leaf structure */
1606 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1607 xfs_trans_t *tp; /* transaction pointer */
1608
1609 xfs_dir2_trace_args("leaf_replace", args);
1610 /*
1611 * Look up the entry.
1612 */
1613 if ((error = xfs_dir2_leaf_lookup_int(args, &lbp, &index, &dbp))) {
1614 return error;
1615 }
1616 dp = args->dp;
1617 leaf = lbp->data;
1618 /*
1619 * Point to the leaf entry, get data address from it.
1620 */
1621 lep = &leaf->ents[index];
1622 /*
1623 * Point to the data entry.
1624 */
1625 dep = (xfs_dir2_data_entry_t *)
1626 ((char *)dbp->data +
1627 XFS_DIR2_DATAPTR_TO_OFF(dp->i_mount, INT_GET(lep->address, ARCH_CONVERT)));
1628 ASSERT(args->inumber != INT_GET(dep->inumber, ARCH_CONVERT));
1629 /*
1630 * Put the new inode number in, log it.
1631 */
1632 INT_SET(dep->inumber, ARCH_CONVERT, args->inumber);
1633 tp = args->trans;
1634 xfs_dir2_data_log_entry(tp, dbp, dep);
1635 xfs_da_buf_done(dbp);
1636 xfs_dir2_leaf_check(dp, lbp);
1637 xfs_da_brelse(tp, lbp);
1638 return 0;
1639 }
1640
1641 /*
1642 * Return index in the leaf block (lbp) which is either the first
1643 * one with this hash value, or if there are none, the insert point
1644 * for that hash value.
1645 */
1646 int /* index value */
1647 xfs_dir2_leaf_search_hash(
1648 xfs_da_args_t *args, /* operation arguments */
1649 xfs_dabuf_t *lbp) /* leaf buffer */
1650 {
1651 xfs_dahash_t hash=0; /* hash from this entry */
1652 xfs_dahash_t hashwant; /* hash value looking for */
1653 int high; /* high leaf index */
1654 int low; /* low leaf index */
1655 xfs_dir2_leaf_t *leaf; /* leaf structure */
1656 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
1657 int mid=0; /* current leaf index */
1658
1659 leaf = lbp->data;
1660 #ifndef __KERNEL__
1661 if (!leaf->hdr.count)
1662 return 0;
1663 #endif
1664 /*
1665 * Note, the table cannot be empty, so we have to go through the loop.
1666 * Binary search the leaf entries looking for our hash value.
1667 */
1668 for (lep = leaf->ents, low = 0, high = INT_GET(leaf->hdr.count, ARCH_CONVERT) - 1,
1669 hashwant = args->hashval;
1670 low <= high; ) {
1671 mid = (low + high) >> 1;
1672 if ((hash = INT_GET(lep[mid].hashval, ARCH_CONVERT)) == hashwant)
1673 break;
1674 if (hash < hashwant)
1675 low = mid + 1;
1676 else
1677 high = mid - 1;
1678 }
1679 /*
1680 * Found one, back up through all the equal hash values.
1681 */
1682 if (hash == hashwant) {
1683 while (mid > 0 && INT_GET(lep[mid - 1].hashval, ARCH_CONVERT) == hashwant) {
1684 mid--;
1685 }
1686 }
1687 /*
1688 * Need to point to an entry higher than ours.
1689 */
1690 else if (hash < hashwant)
1691 mid++;
1692 return mid;
1693 }
1694
1695 /*
1696 * Trim off a trailing data block. We know it's empty since the leaf
1697 * freespace table says so.
1698 */
1699 int /* error */
1700 xfs_dir2_leaf_trim_data(
1701 xfs_da_args_t *args, /* operation arguments */
1702 xfs_dabuf_t *lbp, /* leaf buffer */
1703 xfs_dir2_db_t db) /* data block number */
1704 {
1705 xfs_dir2_data_off_t *bestsp; /* leaf bests table */
1706 #ifdef DEBUG
1707 xfs_dir2_data_t *data; /* data block structure */
1708 #endif
1709 xfs_dabuf_t *dbp; /* data block buffer */
1710 xfs_inode_t *dp; /* incore directory inode */
1711 int error; /* error return value */
1712 xfs_dir2_leaf_t *leaf; /* leaf structure */
1713 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
1714 xfs_mount_t *mp; /* filesystem mount point */
1715 xfs_trans_t *tp; /* transaction pointer */
1716
1717 dp = args->dp;
1718 mp = dp->i_mount;
1719 tp = args->trans;
1720 /*
1721 * Read the offending data block. We need its buffer.
1722 */
1723 if ((error = xfs_da_read_buf(tp, dp, XFS_DIR2_DB_TO_DA(mp, db), -1, &dbp,
1724 XFS_DATA_FORK))) {
1725 return error;
1726 }
1727 #ifdef DEBUG
1728 data = dbp->data;
1729 ASSERT(INT_GET(data->hdr.magic, ARCH_CONVERT) == XFS_DIR2_DATA_MAGIC);
1730 #endif
1731 /* this seems to be an error
1732 * data is only valid if DEBUG is defined?
1733 * RMC 09/08/1999
1734 */
1735
1736 leaf = lbp->data;
1737 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1738 ASSERT(INT_GET(data->hdr.bestfree[0].length, ARCH_CONVERT) ==
1739 mp->m_dirblksize - (uint)sizeof(data->hdr));
1740 ASSERT(db == INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1741 /*
1742 * Get rid of the data block.
1743 */
1744 if ((error = xfs_dir2_shrink_inode(args, db, dbp))) {
1745 ASSERT(error != ENOSPC);
1746 xfs_da_brelse(tp, dbp);
1747 return error;
1748 }
1749 /*
1750 * Eliminate the last bests entry from the table.
1751 */
1752 bestsp = XFS_DIR2_LEAF_BESTS_P(ltp);
1753 INT_MOD(ltp->bestcount, ARCH_CONVERT, -1);
1754 memmove(&bestsp[1], &bestsp[0], INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(*bestsp));
1755 xfs_dir2_leaf_log_tail(tp, lbp);
1756 xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1757 return 0;
1758 }
1759
1760 /*
1761 * Convert node form directory to leaf form directory.
1762 * The root of the node form dir needs to already be a LEAFN block.
1763 * Just return if we can't do anything.
1764 */
1765 int /* error */
1766 xfs_dir2_node_to_leaf(
1767 xfs_da_state_t *state) /* directory operation state */
1768 {
1769 xfs_da_args_t *args; /* operation arguments */
1770 xfs_inode_t *dp; /* incore directory inode */
1771 int error; /* error return code */
1772 xfs_dabuf_t *fbp; /* buffer for freespace block */
1773 xfs_fileoff_t fo; /* freespace file offset */
1774 xfs_dir2_free_t *free; /* freespace structure */
1775 xfs_dabuf_t *lbp; /* buffer for leaf block */
1776 xfs_dir2_leaf_tail_t *ltp; /* tail of leaf structure */
1777 xfs_dir2_leaf_t *leaf; /* leaf structure */
1778 xfs_mount_t *mp; /* filesystem mount point */
1779 int rval; /* successful free trim? */
1780 xfs_trans_t *tp; /* transaction pointer */
1781
1782 /*
1783 * There's more than a leaf level in the btree, so there must
1784 * be multiple leafn blocks. Give up.
1785 */
1786 if (state->path.active > 1)
1787 return 0;
1788 args = state->args;
1789 xfs_dir2_trace_args("node_to_leaf", args);
1790 mp = state->mp;
1791 dp = args->dp;
1792 tp = args->trans;
1793 /*
1794 * Get the last offset in the file.
1795 */
1796 if ((error = xfs_bmap_last_offset(tp, dp, &fo, XFS_DATA_FORK))) {
1797 return error;
1798 }
1799 fo -= mp->m_dirblkfsbs;
1800 /*
1801 * If there are freespace blocks other than the first one,
1802 * take this opportunity to remove trailing empty freespace blocks
1803 * that may have been left behind during no-space-reservation
1804 * operations.
1805 */
1806 while (fo > mp->m_dirfreeblk) {
1807 if ((error = xfs_dir2_node_trim_free(args, fo, &rval))) {
1808 return error;
1809 }
1810 if (rval)
1811 fo -= mp->m_dirblkfsbs;
1812 else
1813 return 0;
1814 }
1815 /*
1816 * Now find the block just before the freespace block.
1817 */
1818 if ((error = xfs_bmap_last_before(tp, dp, &fo, XFS_DATA_FORK))) {
1819 return error;
1820 }
1821 /*
1822 * If it's not the single leaf block, give up.
1823 */
1824 if (XFS_FSB_TO_B(mp, fo) > XFS_DIR2_LEAF_OFFSET + mp->m_dirblksize)
1825 return 0;
1826 lbp = state->path.blk[0].bp;
1827 leaf = lbp->data;
1828 ASSERT(INT_GET(leaf->hdr.info.magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC);
1829 /*
1830 * Read the freespace block.
1831 */
1832 if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
1833 XFS_DATA_FORK))) {
1834 return error;
1835 }
1836 free = fbp->data;
1837 ASSERT(INT_GET(free->hdr.magic, ARCH_CONVERT) == XFS_DIR2_FREE_MAGIC);
1838 ASSERT(!free->hdr.firstdb);
1839 /*
1840 * Now see if the leafn and free data will fit in a leaf1.
1841 * If not, release the buffer and give up.
1842 */
1843 if ((uint)sizeof(leaf->hdr) +
1844 (INT_GET(leaf->hdr.count, ARCH_CONVERT) - INT_GET(leaf->hdr.stale, ARCH_CONVERT)) * (uint)sizeof(leaf->ents[0]) +
1845 INT_GET(free->hdr.nvalid, ARCH_CONVERT) * (uint)sizeof(leaf->bests[0]) +
1846 (uint)sizeof(leaf->tail) >
1847 mp->m_dirblksize) {
1848 xfs_da_brelse(tp, fbp);
1849 return 0;
1850 }
1851 /*
1852 * If the leaf has any stale entries in it, compress them out.
1853 * The compact routine will log the header.
1854 */
1855 if (INT_GET(leaf->hdr.stale, ARCH_CONVERT))
1856 xfs_dir2_leaf_compact(args, lbp);
1857 else
1858 xfs_dir2_leaf_log_header(tp, lbp);
1859 INT_SET(leaf->hdr.info.magic, ARCH_CONVERT, XFS_DIR2_LEAF1_MAGIC);
1860 /*
1861 * Set up the leaf tail from the freespace block.
1862 */
1863 ltp = XFS_DIR2_LEAF_TAIL_P(mp, leaf);
1864 INT_COPY(ltp->bestcount, free->hdr.nvalid, ARCH_CONVERT);
1865 /*
1866 * Set up the leaf bests table.
1867 */
1868 memcpy(XFS_DIR2_LEAF_BESTS_P(ltp), free->bests,
1869 INT_GET(ltp->bestcount, ARCH_CONVERT) * sizeof(leaf->bests[0]));
1870 xfs_dir2_leaf_log_bests(tp, lbp, 0, INT_GET(ltp->bestcount, ARCH_CONVERT) - 1);
1871 xfs_dir2_leaf_log_tail(tp, lbp);
1872 xfs_dir2_leaf_check(dp, lbp);
1873 /*
1874 * Get rid of the freespace block.
1875 */
1876 error = xfs_dir2_shrink_inode(args, XFS_DIR2_FREE_FIRSTDB(mp), fbp);
1877 if (error) {
1878 /*
1879 * This can't fail here because it can only happen when
1880 * punching out the middle of an extent, and this is an
1881 * isolated block.
1882 */
1883 ASSERT(error != ENOSPC);
1884 return error;
1885 }
1886 fbp = NULL;
1887 /*
1888 * Now see if we can convert the single-leaf directory
1889 * down to a block form directory.
1890 * This routine always kills the dabuf for the leaf, so
1891 * eliminate it from the path.
1892 */
1893 error = xfs_dir2_leaf_to_block(args, lbp, NULL);
1894 state->path.blk[0].bp = NULL;
1895 return error;
1896 }