kmemtrace: SLOB hooks.
[linux-2.6/kmemtrace.git] / fs / xfs / xfs_dir2_node.c
blob8dade711f0995c4cfe2d74f46abd41ff624d7953
1 /*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
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
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_types.h"
21 #include "xfs_log.h"
22 #include "xfs_inum.h"
23 #include "xfs_trans.h"
24 #include "xfs_sb.h"
25 #include "xfs_ag.h"
26 #include "xfs_dir2.h"
27 #include "xfs_dmapi.h"
28 #include "xfs_mount.h"
29 #include "xfs_da_btree.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_dir2_sf.h"
32 #include "xfs_attr_sf.h"
33 #include "xfs_dinode.h"
34 #include "xfs_inode.h"
35 #include "xfs_bmap.h"
36 #include "xfs_dir2_data.h"
37 #include "xfs_dir2_leaf.h"
38 #include "xfs_dir2_block.h"
39 #include "xfs_dir2_node.h"
40 #include "xfs_dir2_trace.h"
41 #include "xfs_error.h"
44 * Function declarations.
46 static void xfs_dir2_free_log_header(xfs_trans_t *tp, xfs_dabuf_t *bp);
47 static int xfs_dir2_leafn_add(xfs_dabuf_t *bp, xfs_da_args_t *args, int index);
48 #ifdef DEBUG
49 static void xfs_dir2_leafn_check(xfs_inode_t *dp, xfs_dabuf_t *bp);
50 #else
51 #define xfs_dir2_leafn_check(dp, bp)
52 #endif
53 static void xfs_dir2_leafn_moveents(xfs_da_args_t *args, xfs_dabuf_t *bp_s,
54 int start_s, xfs_dabuf_t *bp_d, int start_d,
55 int count);
56 static void xfs_dir2_leafn_rebalance(xfs_da_state_t *state,
57 xfs_da_state_blk_t *blk1,
58 xfs_da_state_blk_t *blk2);
59 static int xfs_dir2_leafn_remove(xfs_da_args_t *args, xfs_dabuf_t *bp,
60 int index, xfs_da_state_blk_t *dblk,
61 int *rval);
62 static int xfs_dir2_node_addname_int(xfs_da_args_t *args,
63 xfs_da_state_blk_t *fblk);
66 * Log entries from a freespace block.
68 void
69 xfs_dir2_free_log_bests(
70 xfs_trans_t *tp, /* transaction pointer */
71 xfs_dabuf_t *bp, /* freespace buffer */
72 int first, /* first entry to log */
73 int last) /* last entry to log */
75 xfs_dir2_free_t *free; /* freespace structure */
77 free = bp->data;
78 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
79 xfs_da_log_buf(tp, bp,
80 (uint)((char *)&free->bests[first] - (char *)free),
81 (uint)((char *)&free->bests[last] - (char *)free +
82 sizeof(free->bests[0]) - 1));
86 * Log header from a freespace block.
88 static void
89 xfs_dir2_free_log_header(
90 xfs_trans_t *tp, /* transaction pointer */
91 xfs_dabuf_t *bp) /* freespace buffer */
93 xfs_dir2_free_t *free; /* freespace structure */
95 free = bp->data;
96 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
97 xfs_da_log_buf(tp, bp, (uint)((char *)&free->hdr - (char *)free),
98 (uint)(sizeof(xfs_dir2_free_hdr_t) - 1));
102 * Convert a leaf-format directory to a node-format directory.
103 * We need to change the magic number of the leaf block, and copy
104 * the freespace table out of the leaf block into its own block.
106 int /* error */
107 xfs_dir2_leaf_to_node(
108 xfs_da_args_t *args, /* operation arguments */
109 xfs_dabuf_t *lbp) /* leaf buffer */
111 xfs_inode_t *dp; /* incore directory inode */
112 int error; /* error return value */
113 xfs_dabuf_t *fbp; /* freespace buffer */
114 xfs_dir2_db_t fdb; /* freespace block number */
115 xfs_dir2_free_t *free; /* freespace structure */
116 __be16 *from; /* pointer to freespace entry */
117 int i; /* leaf freespace index */
118 xfs_dir2_leaf_t *leaf; /* leaf structure */
119 xfs_dir2_leaf_tail_t *ltp; /* leaf tail structure */
120 xfs_mount_t *mp; /* filesystem mount point */
121 int n; /* count of live freespc ents */
122 xfs_dir2_data_off_t off; /* freespace entry value */
123 __be16 *to; /* pointer to freespace entry */
124 xfs_trans_t *tp; /* transaction pointer */
126 xfs_dir2_trace_args_b("leaf_to_node", args, lbp);
127 dp = args->dp;
128 mp = dp->i_mount;
129 tp = args->trans;
131 * Add a freespace block to the directory.
133 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_FREE_SPACE, &fdb))) {
134 return error;
136 ASSERT(fdb == XFS_DIR2_FREE_FIRSTDB(mp));
138 * Get the buffer for the new freespace block.
140 if ((error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb), -1, &fbp,
141 XFS_DATA_FORK))) {
142 return error;
144 ASSERT(fbp != NULL);
145 free = fbp->data;
146 leaf = lbp->data;
147 ltp = xfs_dir2_leaf_tail_p(mp, leaf);
149 * Initialize the freespace block header.
151 free->hdr.magic = cpu_to_be32(XFS_DIR2_FREE_MAGIC);
152 free->hdr.firstdb = 0;
153 ASSERT(be32_to_cpu(ltp->bestcount) <= (uint)dp->i_d.di_size / mp->m_dirblksize);
154 free->hdr.nvalid = ltp->bestcount;
156 * Copy freespace entries from the leaf block to the new block.
157 * Count active entries.
159 for (i = n = 0, from = xfs_dir2_leaf_bests_p(ltp), to = free->bests;
160 i < be32_to_cpu(ltp->bestcount); i++, from++, to++) {
161 if ((off = be16_to_cpu(*from)) != NULLDATAOFF)
162 n++;
163 *to = cpu_to_be16(off);
165 free->hdr.nused = cpu_to_be32(n);
166 leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAFN_MAGIC);
168 * Log everything.
170 xfs_dir2_leaf_log_header(tp, lbp);
171 xfs_dir2_free_log_header(tp, fbp);
172 xfs_dir2_free_log_bests(tp, fbp, 0, be32_to_cpu(free->hdr.nvalid) - 1);
173 xfs_da_buf_done(fbp);
174 xfs_dir2_leafn_check(dp, lbp);
175 return 0;
179 * Add a leaf entry to a leaf block in a node-form directory.
180 * The other work necessary is done from the caller.
182 static int /* error */
183 xfs_dir2_leafn_add(
184 xfs_dabuf_t *bp, /* leaf buffer */
185 xfs_da_args_t *args, /* operation arguments */
186 int index) /* insertion pt for new entry */
188 int compact; /* compacting stale leaves */
189 xfs_inode_t *dp; /* incore directory inode */
190 int highstale; /* next stale entry */
191 xfs_dir2_leaf_t *leaf; /* leaf structure */
192 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
193 int lfloghigh; /* high leaf entry logging */
194 int lfloglow; /* low leaf entry logging */
195 int lowstale; /* previous stale entry */
196 xfs_mount_t *mp; /* filesystem mount point */
197 xfs_trans_t *tp; /* transaction pointer */
199 xfs_dir2_trace_args_sb("leafn_add", args, index, bp);
200 dp = args->dp;
201 mp = dp->i_mount;
202 tp = args->trans;
203 leaf = bp->data;
206 * Quick check just to make sure we are not going to index
207 * into other peoples memory
209 if (index < 0)
210 return XFS_ERROR(EFSCORRUPTED);
213 * If there are already the maximum number of leaf entries in
214 * the block, if there are no stale entries it won't fit.
215 * Caller will do a split. If there are stale entries we'll do
216 * a compact.
219 if (be16_to_cpu(leaf->hdr.count) == xfs_dir2_max_leaf_ents(mp)) {
220 if (!leaf->hdr.stale)
221 return XFS_ERROR(ENOSPC);
222 compact = be16_to_cpu(leaf->hdr.stale) > 1;
223 } else
224 compact = 0;
225 ASSERT(index == 0 || be32_to_cpu(leaf->ents[index - 1].hashval) <= args->hashval);
226 ASSERT(index == be16_to_cpu(leaf->hdr.count) ||
227 be32_to_cpu(leaf->ents[index].hashval) >= args->hashval);
229 if (args->justcheck)
230 return 0;
233 * Compact out all but one stale leaf entry. Leaves behind
234 * the entry closest to index.
236 if (compact) {
237 xfs_dir2_leaf_compact_x1(bp, &index, &lowstale, &highstale,
238 &lfloglow, &lfloghigh);
241 * Set impossible logging indices for this case.
243 else if (leaf->hdr.stale) {
244 lfloglow = be16_to_cpu(leaf->hdr.count);
245 lfloghigh = -1;
248 * No stale entries, just insert a space for the new entry.
250 if (!leaf->hdr.stale) {
251 lep = &leaf->ents[index];
252 if (index < be16_to_cpu(leaf->hdr.count))
253 memmove(lep + 1, lep,
254 (be16_to_cpu(leaf->hdr.count) - index) * sizeof(*lep));
255 lfloglow = index;
256 lfloghigh = be16_to_cpu(leaf->hdr.count);
257 be16_add_cpu(&leaf->hdr.count, 1);
260 * There are stale entries. We'll use one for the new entry.
262 else {
264 * If we didn't do a compact then we need to figure out
265 * which stale entry will be used.
267 if (compact == 0) {
269 * Find first stale entry before our insertion point.
271 for (lowstale = index - 1;
272 lowstale >= 0 &&
273 be32_to_cpu(leaf->ents[lowstale].address) !=
274 XFS_DIR2_NULL_DATAPTR;
275 lowstale--)
276 continue;
278 * Find next stale entry after insertion point.
279 * Stop looking if the answer would be worse than
280 * lowstale already found.
282 for (highstale = index;
283 highstale < be16_to_cpu(leaf->hdr.count) &&
284 be32_to_cpu(leaf->ents[highstale].address) !=
285 XFS_DIR2_NULL_DATAPTR &&
286 (lowstale < 0 ||
287 index - lowstale - 1 >= highstale - index);
288 highstale++)
289 continue;
292 * Using the low stale entry.
293 * Shift entries up toward the stale slot.
295 if (lowstale >= 0 &&
296 (highstale == be16_to_cpu(leaf->hdr.count) ||
297 index - lowstale - 1 < highstale - index)) {
298 ASSERT(be32_to_cpu(leaf->ents[lowstale].address) ==
299 XFS_DIR2_NULL_DATAPTR);
300 ASSERT(index - lowstale - 1 >= 0);
301 if (index - lowstale - 1 > 0)
302 memmove(&leaf->ents[lowstale],
303 &leaf->ents[lowstale + 1],
304 (index - lowstale - 1) * sizeof(*lep));
305 lep = &leaf->ents[index - 1];
306 lfloglow = MIN(lowstale, lfloglow);
307 lfloghigh = MAX(index - 1, lfloghigh);
310 * Using the high stale entry.
311 * Shift entries down toward the stale slot.
313 else {
314 ASSERT(be32_to_cpu(leaf->ents[highstale].address) ==
315 XFS_DIR2_NULL_DATAPTR);
316 ASSERT(highstale - index >= 0);
317 if (highstale - index > 0)
318 memmove(&leaf->ents[index + 1],
319 &leaf->ents[index],
320 (highstale - index) * sizeof(*lep));
321 lep = &leaf->ents[index];
322 lfloglow = MIN(index, lfloglow);
323 lfloghigh = MAX(highstale, lfloghigh);
325 be16_add_cpu(&leaf->hdr.stale, -1);
328 * Insert the new entry, log everything.
330 lep->hashval = cpu_to_be32(args->hashval);
331 lep->address = cpu_to_be32(xfs_dir2_db_off_to_dataptr(mp,
332 args->blkno, args->index));
333 xfs_dir2_leaf_log_header(tp, bp);
334 xfs_dir2_leaf_log_ents(tp, bp, lfloglow, lfloghigh);
335 xfs_dir2_leafn_check(dp, bp);
336 return 0;
339 #ifdef DEBUG
341 * Check internal consistency of a leafn block.
343 void
344 xfs_dir2_leafn_check(
345 xfs_inode_t *dp, /* incore directory inode */
346 xfs_dabuf_t *bp) /* leaf buffer */
348 int i; /* leaf index */
349 xfs_dir2_leaf_t *leaf; /* leaf structure */
350 xfs_mount_t *mp; /* filesystem mount point */
351 int stale; /* count of stale leaves */
353 leaf = bp->data;
354 mp = dp->i_mount;
355 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
356 ASSERT(be16_to_cpu(leaf->hdr.count) <= xfs_dir2_max_leaf_ents(mp));
357 for (i = stale = 0; i < be16_to_cpu(leaf->hdr.count); i++) {
358 if (i + 1 < be16_to_cpu(leaf->hdr.count)) {
359 ASSERT(be32_to_cpu(leaf->ents[i].hashval) <=
360 be32_to_cpu(leaf->ents[i + 1].hashval));
362 if (be32_to_cpu(leaf->ents[i].address) == XFS_DIR2_NULL_DATAPTR)
363 stale++;
365 ASSERT(be16_to_cpu(leaf->hdr.stale) == stale);
367 #endif /* DEBUG */
370 * Return the last hash value in the leaf.
371 * Stale entries are ok.
373 xfs_dahash_t /* hash value */
374 xfs_dir2_leafn_lasthash(
375 xfs_dabuf_t *bp, /* leaf buffer */
376 int *count) /* count of entries in leaf */
378 xfs_dir2_leaf_t *leaf; /* leaf structure */
380 leaf = bp->data;
381 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
382 if (count)
383 *count = be16_to_cpu(leaf->hdr.count);
384 if (!leaf->hdr.count)
385 return 0;
386 return be32_to_cpu(leaf->ents[be16_to_cpu(leaf->hdr.count) - 1].hashval);
390 * Look up a leaf entry in a node-format leaf block.
391 * If this is an addname then the extrablk in state is a freespace block,
392 * otherwise it's a data block.
395 xfs_dir2_leafn_lookup_int(
396 xfs_dabuf_t *bp, /* leaf buffer */
397 xfs_da_args_t *args, /* operation arguments */
398 int *indexp, /* out: leaf entry index */
399 xfs_da_state_t *state) /* state to fill in */
401 xfs_dabuf_t *curbp; /* current data/free buffer */
402 xfs_dir2_db_t curdb; /* current data block number */
403 xfs_dir2_db_t curfdb; /* current free block number */
404 xfs_dir2_data_entry_t *dep; /* data block entry */
405 xfs_inode_t *dp; /* incore directory inode */
406 int error; /* error return value */
407 int fi; /* free entry index */
408 xfs_dir2_free_t *free=NULL; /* free block structure */
409 int index; /* leaf entry index */
410 xfs_dir2_leaf_t *leaf; /* leaf structure */
411 int length=0; /* length of new data entry */
412 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
413 xfs_mount_t *mp; /* filesystem mount point */
414 xfs_dir2_db_t newdb; /* new data block number */
415 xfs_dir2_db_t newfdb; /* new free block number */
416 xfs_trans_t *tp; /* transaction pointer */
418 dp = args->dp;
419 tp = args->trans;
420 mp = dp->i_mount;
421 leaf = bp->data;
422 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
423 #ifdef __KERNEL__
424 ASSERT(be16_to_cpu(leaf->hdr.count) > 0);
425 #endif
426 xfs_dir2_leafn_check(dp, bp);
428 * Look up the hash value in the leaf entries.
430 index = xfs_dir2_leaf_search_hash(args, bp);
432 * Do we have a buffer coming in?
434 if (state->extravalid)
435 curbp = state->extrablk.bp;
436 else
437 curbp = NULL;
439 * For addname, it's a free block buffer, get the block number.
441 if (args->addname) {
442 curfdb = curbp ? state->extrablk.blkno : -1;
443 curdb = -1;
444 length = xfs_dir2_data_entsize(args->namelen);
445 if ((free = (curbp ? curbp->data : NULL)))
446 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
449 * For others, it's a data block buffer, get the block number.
451 else {
452 curfdb = -1;
453 curdb = curbp ? state->extrablk.blkno : -1;
456 * Loop over leaf entries with the right hash value.
458 for (lep = &leaf->ents[index];
459 index < be16_to_cpu(leaf->hdr.count) && be32_to_cpu(lep->hashval) == args->hashval;
460 lep++, index++) {
462 * Skip stale leaf entries.
464 if (be32_to_cpu(lep->address) == XFS_DIR2_NULL_DATAPTR)
465 continue;
467 * Pull the data block number from the entry.
469 newdb = xfs_dir2_dataptr_to_db(mp, be32_to_cpu(lep->address));
471 * For addname, we're looking for a place to put the new entry.
472 * We want to use a data block with an entry of equal
473 * hash value to ours if there is one with room.
475 if (args->addname) {
477 * If this block isn't the data block we already have
478 * in hand, take a look at it.
480 if (newdb != curdb) {
481 curdb = newdb;
483 * Convert the data block to the free block
484 * holding its freespace information.
486 newfdb = xfs_dir2_db_to_fdb(mp, newdb);
488 * If it's not the one we have in hand,
489 * read it in.
491 if (newfdb != curfdb) {
493 * If we had one before, drop it.
495 if (curbp)
496 xfs_da_brelse(tp, curbp);
498 * Read the free block.
500 if ((error = xfs_da_read_buf(tp, dp,
501 xfs_dir2_db_to_da(mp,
502 newfdb),
503 -1, &curbp,
504 XFS_DATA_FORK))) {
505 return error;
507 free = curbp->data;
508 ASSERT(be32_to_cpu(free->hdr.magic) ==
509 XFS_DIR2_FREE_MAGIC);
510 ASSERT((be32_to_cpu(free->hdr.firstdb) %
511 XFS_DIR2_MAX_FREE_BESTS(mp)) ==
513 ASSERT(be32_to_cpu(free->hdr.firstdb) <= curdb);
514 ASSERT(curdb <
515 be32_to_cpu(free->hdr.firstdb) +
516 be32_to_cpu(free->hdr.nvalid));
519 * Get the index for our entry.
521 fi = xfs_dir2_db_to_fdindex(mp, curdb);
523 * If it has room, return it.
525 if (unlikely(be16_to_cpu(free->bests[fi]) == NULLDATAOFF)) {
526 XFS_ERROR_REPORT("xfs_dir2_leafn_lookup_int",
527 XFS_ERRLEVEL_LOW, mp);
528 if (curfdb != newfdb)
529 xfs_da_brelse(tp, curbp);
530 return XFS_ERROR(EFSCORRUPTED);
532 curfdb = newfdb;
533 if (be16_to_cpu(free->bests[fi]) >= length) {
534 *indexp = index;
535 state->extravalid = 1;
536 state->extrablk.bp = curbp;
537 state->extrablk.blkno = curfdb;
538 state->extrablk.index = fi;
539 state->extrablk.magic =
540 XFS_DIR2_FREE_MAGIC;
541 ASSERT(args->oknoent);
542 return XFS_ERROR(ENOENT);
547 * Not adding a new entry, so we really want to find
548 * the name given to us.
550 else {
552 * If it's a different data block, go get it.
554 if (newdb != curdb) {
556 * If we had a block before, drop it.
558 if (curbp)
559 xfs_da_brelse(tp, curbp);
561 * Read the data block.
563 if ((error =
564 xfs_da_read_buf(tp, dp,
565 xfs_dir2_db_to_da(mp, newdb), -1,
566 &curbp, XFS_DATA_FORK))) {
567 return error;
569 xfs_dir2_data_check(dp, curbp);
570 curdb = newdb;
573 * Point to the data entry.
575 dep = (xfs_dir2_data_entry_t *)
576 ((char *)curbp->data +
577 xfs_dir2_dataptr_to_off(mp, be32_to_cpu(lep->address)));
579 * Compare the entry, return it if it matches.
581 if (dep->namelen == args->namelen &&
582 dep->name[0] == args->name[0] &&
583 memcmp(dep->name, args->name, args->namelen) == 0) {
584 args->inumber = be64_to_cpu(dep->inumber);
585 *indexp = index;
586 state->extravalid = 1;
587 state->extrablk.bp = curbp;
588 state->extrablk.blkno = curdb;
589 state->extrablk.index =
590 (int)((char *)dep -
591 (char *)curbp->data);
592 state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
593 return XFS_ERROR(EEXIST);
598 * Didn't find a match.
599 * If we are holding a buffer, give it back in case our caller
600 * finds it useful.
602 if ((state->extravalid = (curbp != NULL))) {
603 state->extrablk.bp = curbp;
604 state->extrablk.index = -1;
606 * For addname, giving back a free block.
608 if (args->addname) {
609 state->extrablk.blkno = curfdb;
610 state->extrablk.magic = XFS_DIR2_FREE_MAGIC;
613 * For other callers, giving back a data block.
615 else {
616 state->extrablk.blkno = curdb;
617 state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
621 * Return the final index, that will be the insertion point.
623 *indexp = index;
624 ASSERT(index == be16_to_cpu(leaf->hdr.count) || args->oknoent);
625 return XFS_ERROR(ENOENT);
629 * Move count leaf entries from source to destination leaf.
630 * Log entries and headers. Stale entries are preserved.
632 static void
633 xfs_dir2_leafn_moveents(
634 xfs_da_args_t *args, /* operation arguments */
635 xfs_dabuf_t *bp_s, /* source leaf buffer */
636 int start_s, /* source leaf index */
637 xfs_dabuf_t *bp_d, /* destination leaf buffer */
638 int start_d, /* destination leaf index */
639 int count) /* count of leaves to copy */
641 xfs_dir2_leaf_t *leaf_d; /* destination leaf structure */
642 xfs_dir2_leaf_t *leaf_s; /* source leaf structure */
643 int stale; /* count stale leaves copied */
644 xfs_trans_t *tp; /* transaction pointer */
646 xfs_dir2_trace_args_bibii("leafn_moveents", args, bp_s, start_s, bp_d,
647 start_d, count);
649 * Silently return if nothing to do.
651 if (count == 0) {
652 return;
654 tp = args->trans;
655 leaf_s = bp_s->data;
656 leaf_d = bp_d->data;
658 * If the destination index is not the end of the current
659 * destination leaf entries, open up a hole in the destination
660 * to hold the new entries.
662 if (start_d < be16_to_cpu(leaf_d->hdr.count)) {
663 memmove(&leaf_d->ents[start_d + count], &leaf_d->ents[start_d],
664 (be16_to_cpu(leaf_d->hdr.count) - start_d) *
665 sizeof(xfs_dir2_leaf_entry_t));
666 xfs_dir2_leaf_log_ents(tp, bp_d, start_d + count,
667 count + be16_to_cpu(leaf_d->hdr.count) - 1);
670 * If the source has stale leaves, count the ones in the copy range
671 * so we can update the header correctly.
673 if (leaf_s->hdr.stale) {
674 int i; /* temp leaf index */
676 for (i = start_s, stale = 0; i < start_s + count; i++) {
677 if (be32_to_cpu(leaf_s->ents[i].address) == XFS_DIR2_NULL_DATAPTR)
678 stale++;
680 } else
681 stale = 0;
683 * Copy the leaf entries from source to destination.
685 memcpy(&leaf_d->ents[start_d], &leaf_s->ents[start_s],
686 count * sizeof(xfs_dir2_leaf_entry_t));
687 xfs_dir2_leaf_log_ents(tp, bp_d, start_d, start_d + count - 1);
689 * If there are source entries after the ones we copied,
690 * delete the ones we copied by sliding the next ones down.
692 if (start_s + count < be16_to_cpu(leaf_s->hdr.count)) {
693 memmove(&leaf_s->ents[start_s], &leaf_s->ents[start_s + count],
694 count * sizeof(xfs_dir2_leaf_entry_t));
695 xfs_dir2_leaf_log_ents(tp, bp_s, start_s, start_s + count - 1);
698 * Update the headers and log them.
700 be16_add_cpu(&leaf_s->hdr.count, -(count));
701 be16_add_cpu(&leaf_s->hdr.stale, -(stale));
702 be16_add_cpu(&leaf_d->hdr.count, count);
703 be16_add_cpu(&leaf_d->hdr.stale, stale);
704 xfs_dir2_leaf_log_header(tp, bp_s);
705 xfs_dir2_leaf_log_header(tp, bp_d);
706 xfs_dir2_leafn_check(args->dp, bp_s);
707 xfs_dir2_leafn_check(args->dp, bp_d);
711 * Determine the sort order of two leaf blocks.
712 * Returns 1 if both are valid and leaf2 should be before leaf1, else 0.
714 int /* sort order */
715 xfs_dir2_leafn_order(
716 xfs_dabuf_t *leaf1_bp, /* leaf1 buffer */
717 xfs_dabuf_t *leaf2_bp) /* leaf2 buffer */
719 xfs_dir2_leaf_t *leaf1; /* leaf1 structure */
720 xfs_dir2_leaf_t *leaf2; /* leaf2 structure */
722 leaf1 = leaf1_bp->data;
723 leaf2 = leaf2_bp->data;
724 ASSERT(be16_to_cpu(leaf1->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
725 ASSERT(be16_to_cpu(leaf2->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
726 if (be16_to_cpu(leaf1->hdr.count) > 0 &&
727 be16_to_cpu(leaf2->hdr.count) > 0 &&
728 (be32_to_cpu(leaf2->ents[0].hashval) < be32_to_cpu(leaf1->ents[0].hashval) ||
729 be32_to_cpu(leaf2->ents[be16_to_cpu(leaf2->hdr.count) - 1].hashval) <
730 be32_to_cpu(leaf1->ents[be16_to_cpu(leaf1->hdr.count) - 1].hashval)))
731 return 1;
732 return 0;
736 * Rebalance leaf entries between two leaf blocks.
737 * This is actually only called when the second block is new,
738 * though the code deals with the general case.
739 * A new entry will be inserted in one of the blocks, and that
740 * entry is taken into account when balancing.
742 static void
743 xfs_dir2_leafn_rebalance(
744 xfs_da_state_t *state, /* btree cursor */
745 xfs_da_state_blk_t *blk1, /* first btree block */
746 xfs_da_state_blk_t *blk2) /* second btree block */
748 xfs_da_args_t *args; /* operation arguments */
749 int count; /* count (& direction) leaves */
750 int isleft; /* new goes in left leaf */
751 xfs_dir2_leaf_t *leaf1; /* first leaf structure */
752 xfs_dir2_leaf_t *leaf2; /* second leaf structure */
753 int mid; /* midpoint leaf index */
754 #ifdef DEBUG
755 int oldstale; /* old count of stale leaves */
756 #endif
757 int oldsum; /* old total leaf count */
758 int swap; /* swapped leaf blocks */
760 args = state->args;
762 * If the block order is wrong, swap the arguments.
764 if ((swap = xfs_dir2_leafn_order(blk1->bp, blk2->bp))) {
765 xfs_da_state_blk_t *tmp; /* temp for block swap */
767 tmp = blk1;
768 blk1 = blk2;
769 blk2 = tmp;
771 leaf1 = blk1->bp->data;
772 leaf2 = blk2->bp->data;
773 oldsum = be16_to_cpu(leaf1->hdr.count) + be16_to_cpu(leaf2->hdr.count);
774 #ifdef DEBUG
775 oldstale = be16_to_cpu(leaf1->hdr.stale) + be16_to_cpu(leaf2->hdr.stale);
776 #endif
777 mid = oldsum >> 1;
779 * If the old leaf count was odd then the new one will be even,
780 * so we need to divide the new count evenly.
782 if (oldsum & 1) {
783 xfs_dahash_t midhash; /* middle entry hash value */
785 if (mid >= be16_to_cpu(leaf1->hdr.count))
786 midhash = be32_to_cpu(leaf2->ents[mid - be16_to_cpu(leaf1->hdr.count)].hashval);
787 else
788 midhash = be32_to_cpu(leaf1->ents[mid].hashval);
789 isleft = args->hashval <= midhash;
792 * If the old count is even then the new count is odd, so there's
793 * no preferred side for the new entry.
794 * Pick the left one.
796 else
797 isleft = 1;
799 * Calculate moved entry count. Positive means left-to-right,
800 * negative means right-to-left. Then move the entries.
802 count = be16_to_cpu(leaf1->hdr.count) - mid + (isleft == 0);
803 if (count > 0)
804 xfs_dir2_leafn_moveents(args, blk1->bp,
805 be16_to_cpu(leaf1->hdr.count) - count, blk2->bp, 0, count);
806 else if (count < 0)
807 xfs_dir2_leafn_moveents(args, blk2->bp, 0, blk1->bp,
808 be16_to_cpu(leaf1->hdr.count), count);
809 ASSERT(be16_to_cpu(leaf1->hdr.count) + be16_to_cpu(leaf2->hdr.count) == oldsum);
810 ASSERT(be16_to_cpu(leaf1->hdr.stale) + be16_to_cpu(leaf2->hdr.stale) == oldstale);
812 * Mark whether we're inserting into the old or new leaf.
814 if (be16_to_cpu(leaf1->hdr.count) < be16_to_cpu(leaf2->hdr.count))
815 state->inleaf = swap;
816 else if (be16_to_cpu(leaf1->hdr.count) > be16_to_cpu(leaf2->hdr.count))
817 state->inleaf = !swap;
818 else
819 state->inleaf =
820 swap ^ (blk1->index <= be16_to_cpu(leaf1->hdr.count));
822 * Adjust the expected index for insertion.
824 if (!state->inleaf)
825 blk2->index = blk1->index - be16_to_cpu(leaf1->hdr.count);
828 * Finally sanity check just to make sure we are not returning a negative index
830 if(blk2->index < 0) {
831 state->inleaf = 1;
832 blk2->index = 0;
833 cmn_err(CE_ALERT,
834 "xfs_dir2_leafn_rebalance: picked the wrong leaf? reverting original leaf: "
835 "blk1->index %d\n",
836 blk1->index);
841 * Remove an entry from a node directory.
842 * This removes the leaf entry and the data entry,
843 * and updates the free block if necessary.
845 static int /* error */
846 xfs_dir2_leafn_remove(
847 xfs_da_args_t *args, /* operation arguments */
848 xfs_dabuf_t *bp, /* leaf buffer */
849 int index, /* leaf entry index */
850 xfs_da_state_blk_t *dblk, /* data block */
851 int *rval) /* resulting block needs join */
853 xfs_dir2_data_t *data; /* data block structure */
854 xfs_dir2_db_t db; /* data block number */
855 xfs_dabuf_t *dbp; /* data block buffer */
856 xfs_dir2_data_entry_t *dep; /* data block entry */
857 xfs_inode_t *dp; /* incore directory inode */
858 xfs_dir2_leaf_t *leaf; /* leaf structure */
859 xfs_dir2_leaf_entry_t *lep; /* leaf entry */
860 int longest; /* longest data free entry */
861 int off; /* data block entry offset */
862 xfs_mount_t *mp; /* filesystem mount point */
863 int needlog; /* need to log data header */
864 int needscan; /* need to rescan data frees */
865 xfs_trans_t *tp; /* transaction pointer */
867 xfs_dir2_trace_args_sb("leafn_remove", args, index, bp);
868 dp = args->dp;
869 tp = args->trans;
870 mp = dp->i_mount;
871 leaf = bp->data;
872 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
874 * Point to the entry we're removing.
876 lep = &leaf->ents[index];
878 * Extract the data block and offset from the entry.
880 db = xfs_dir2_dataptr_to_db(mp, be32_to_cpu(lep->address));
881 ASSERT(dblk->blkno == db);
882 off = xfs_dir2_dataptr_to_off(mp, be32_to_cpu(lep->address));
883 ASSERT(dblk->index == off);
885 * Kill the leaf entry by marking it stale.
886 * Log the leaf block changes.
888 be16_add_cpu(&leaf->hdr.stale, 1);
889 xfs_dir2_leaf_log_header(tp, bp);
890 lep->address = cpu_to_be32(XFS_DIR2_NULL_DATAPTR);
891 xfs_dir2_leaf_log_ents(tp, bp, index, index);
893 * Make the data entry free. Keep track of the longest freespace
894 * in the data block in case it changes.
896 dbp = dblk->bp;
897 data = dbp->data;
898 dep = (xfs_dir2_data_entry_t *)((char *)data + off);
899 longest = be16_to_cpu(data->hdr.bestfree[0].length);
900 needlog = needscan = 0;
901 xfs_dir2_data_make_free(tp, dbp, off,
902 xfs_dir2_data_entsize(dep->namelen), &needlog, &needscan);
904 * Rescan the data block freespaces for bestfree.
905 * Log the data block header if needed.
907 if (needscan)
908 xfs_dir2_data_freescan(mp, data, &needlog);
909 if (needlog)
910 xfs_dir2_data_log_header(tp, dbp);
911 xfs_dir2_data_check(dp, dbp);
913 * If the longest data block freespace changes, need to update
914 * the corresponding freeblock entry.
916 if (longest < be16_to_cpu(data->hdr.bestfree[0].length)) {
917 int error; /* error return value */
918 xfs_dabuf_t *fbp; /* freeblock buffer */
919 xfs_dir2_db_t fdb; /* freeblock block number */
920 int findex; /* index in freeblock entries */
921 xfs_dir2_free_t *free; /* freeblock structure */
922 int logfree; /* need to log free entry */
925 * Convert the data block number to a free block,
926 * read in the free block.
928 fdb = xfs_dir2_db_to_fdb(mp, db);
929 if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb),
930 -1, &fbp, XFS_DATA_FORK))) {
931 return error;
933 free = fbp->data;
934 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
935 ASSERT(be32_to_cpu(free->hdr.firstdb) ==
936 XFS_DIR2_MAX_FREE_BESTS(mp) *
937 (fdb - XFS_DIR2_FREE_FIRSTDB(mp)));
939 * Calculate which entry we need to fix.
941 findex = xfs_dir2_db_to_fdindex(mp, db);
942 longest = be16_to_cpu(data->hdr.bestfree[0].length);
944 * If the data block is now empty we can get rid of it
945 * (usually).
947 if (longest == mp->m_dirblksize - (uint)sizeof(data->hdr)) {
949 * Try to punch out the data block.
951 error = xfs_dir2_shrink_inode(args, db, dbp);
952 if (error == 0) {
953 dblk->bp = NULL;
954 data = NULL;
957 * We can get ENOSPC if there's no space reservation.
958 * In this case just drop the buffer and some one else
959 * will eventually get rid of the empty block.
961 else if (error == ENOSPC && args->total == 0)
962 xfs_da_buf_done(dbp);
963 else
964 return error;
967 * If we got rid of the data block, we can eliminate that entry
968 * in the free block.
970 if (data == NULL) {
972 * One less used entry in the free table.
974 be32_add_cpu(&free->hdr.nused, -1);
975 xfs_dir2_free_log_header(tp, fbp);
977 * If this was the last entry in the table, we can
978 * trim the table size back. There might be other
979 * entries at the end referring to non-existent
980 * data blocks, get those too.
982 if (findex == be32_to_cpu(free->hdr.nvalid) - 1) {
983 int i; /* free entry index */
985 for (i = findex - 1;
986 i >= 0 && be16_to_cpu(free->bests[i]) == NULLDATAOFF;
987 i--)
988 continue;
989 free->hdr.nvalid = cpu_to_be32(i + 1);
990 logfree = 0;
993 * Not the last entry, just punch it out.
995 else {
996 free->bests[findex] = cpu_to_be16(NULLDATAOFF);
997 logfree = 1;
1000 * If there are no useful entries left in the block,
1001 * get rid of the block if we can.
1003 if (!free->hdr.nused) {
1004 error = xfs_dir2_shrink_inode(args, fdb, fbp);
1005 if (error == 0) {
1006 fbp = NULL;
1007 logfree = 0;
1008 } else if (error != ENOSPC || args->total != 0)
1009 return error;
1011 * It's possible to get ENOSPC if there is no
1012 * space reservation. In this case some one
1013 * else will eventually get rid of this block.
1018 * Data block is not empty, just set the free entry to
1019 * the new value.
1021 else {
1022 free->bests[findex] = cpu_to_be16(longest);
1023 logfree = 1;
1026 * Log the free entry that changed, unless we got rid of it.
1028 if (logfree)
1029 xfs_dir2_free_log_bests(tp, fbp, findex, findex);
1031 * Drop the buffer if we still have it.
1033 if (fbp)
1034 xfs_da_buf_done(fbp);
1036 xfs_dir2_leafn_check(dp, bp);
1038 * Return indication of whether this leaf block is emtpy enough
1039 * to justify trying to join it with a neighbor.
1041 *rval =
1042 ((uint)sizeof(leaf->hdr) +
1043 (uint)sizeof(leaf->ents[0]) *
1044 (be16_to_cpu(leaf->hdr.count) - be16_to_cpu(leaf->hdr.stale))) <
1045 mp->m_dir_magicpct;
1046 return 0;
1050 * Split the leaf entries in the old block into old and new blocks.
1052 int /* error */
1053 xfs_dir2_leafn_split(
1054 xfs_da_state_t *state, /* btree cursor */
1055 xfs_da_state_blk_t *oldblk, /* original block */
1056 xfs_da_state_blk_t *newblk) /* newly created block */
1058 xfs_da_args_t *args; /* operation arguments */
1059 xfs_dablk_t blkno; /* new leaf block number */
1060 int error; /* error return value */
1061 xfs_mount_t *mp; /* filesystem mount point */
1064 * Allocate space for a new leaf node.
1066 args = state->args;
1067 mp = args->dp->i_mount;
1068 ASSERT(args != NULL);
1069 ASSERT(oldblk->magic == XFS_DIR2_LEAFN_MAGIC);
1070 error = xfs_da_grow_inode(args, &blkno);
1071 if (error) {
1072 return error;
1075 * Initialize the new leaf block.
1077 error = xfs_dir2_leaf_init(args, xfs_dir2_da_to_db(mp, blkno),
1078 &newblk->bp, XFS_DIR2_LEAFN_MAGIC);
1079 if (error) {
1080 return error;
1082 newblk->blkno = blkno;
1083 newblk->magic = XFS_DIR2_LEAFN_MAGIC;
1085 * Rebalance the entries across the two leaves, link the new
1086 * block into the leaves.
1088 xfs_dir2_leafn_rebalance(state, oldblk, newblk);
1089 error = xfs_da_blk_link(state, oldblk, newblk);
1090 if (error) {
1091 return error;
1094 * Insert the new entry in the correct block.
1096 if (state->inleaf)
1097 error = xfs_dir2_leafn_add(oldblk->bp, args, oldblk->index);
1098 else
1099 error = xfs_dir2_leafn_add(newblk->bp, args, newblk->index);
1101 * Update last hashval in each block since we added the name.
1103 oldblk->hashval = xfs_dir2_leafn_lasthash(oldblk->bp, NULL);
1104 newblk->hashval = xfs_dir2_leafn_lasthash(newblk->bp, NULL);
1105 xfs_dir2_leafn_check(args->dp, oldblk->bp);
1106 xfs_dir2_leafn_check(args->dp, newblk->bp);
1107 return error;
1111 * Check a leaf block and its neighbors to see if the block should be
1112 * collapsed into one or the other neighbor. Always keep the block
1113 * with the smaller block number.
1114 * If the current block is over 50% full, don't try to join it, return 0.
1115 * If the block is empty, fill in the state structure and return 2.
1116 * If it can be collapsed, fill in the state structure and return 1.
1117 * If nothing can be done, return 0.
1119 int /* error */
1120 xfs_dir2_leafn_toosmall(
1121 xfs_da_state_t *state, /* btree cursor */
1122 int *action) /* resulting action to take */
1124 xfs_da_state_blk_t *blk; /* leaf block */
1125 xfs_dablk_t blkno; /* leaf block number */
1126 xfs_dabuf_t *bp; /* leaf buffer */
1127 int bytes; /* bytes in use */
1128 int count; /* leaf live entry count */
1129 int error; /* error return value */
1130 int forward; /* sibling block direction */
1131 int i; /* sibling counter */
1132 xfs_da_blkinfo_t *info; /* leaf block header */
1133 xfs_dir2_leaf_t *leaf; /* leaf structure */
1134 int rval; /* result from path_shift */
1137 * Check for the degenerate case of the block being over 50% full.
1138 * If so, it's not worth even looking to see if we might be able
1139 * to coalesce with a sibling.
1141 blk = &state->path.blk[state->path.active - 1];
1142 info = blk->bp->data;
1143 ASSERT(be16_to_cpu(info->magic) == XFS_DIR2_LEAFN_MAGIC);
1144 leaf = (xfs_dir2_leaf_t *)info;
1145 count = be16_to_cpu(leaf->hdr.count) - be16_to_cpu(leaf->hdr.stale);
1146 bytes = (uint)sizeof(leaf->hdr) + count * (uint)sizeof(leaf->ents[0]);
1147 if (bytes > (state->blocksize >> 1)) {
1149 * Blk over 50%, don't try to join.
1151 *action = 0;
1152 return 0;
1155 * Check for the degenerate case of the block being empty.
1156 * If the block is empty, we'll simply delete it, no need to
1157 * coalesce it with a sibling block. We choose (arbitrarily)
1158 * to merge with the forward block unless it is NULL.
1160 if (count == 0) {
1162 * Make altpath point to the block we want to keep and
1163 * path point to the block we want to drop (this one).
1165 forward = (info->forw != 0);
1166 memcpy(&state->altpath, &state->path, sizeof(state->path));
1167 error = xfs_da_path_shift(state, &state->altpath, forward, 0,
1168 &rval);
1169 if (error)
1170 return error;
1171 *action = rval ? 2 : 0;
1172 return 0;
1175 * Examine each sibling block to see if we can coalesce with
1176 * at least 25% free space to spare. We need to figure out
1177 * whether to merge with the forward or the backward block.
1178 * We prefer coalescing with the lower numbered sibling so as
1179 * to shrink a directory over time.
1181 forward = be32_to_cpu(info->forw) < be32_to_cpu(info->back);
1182 for (i = 0, bp = NULL; i < 2; forward = !forward, i++) {
1183 blkno = forward ? be32_to_cpu(info->forw) : be32_to_cpu(info->back);
1184 if (blkno == 0)
1185 continue;
1187 * Read the sibling leaf block.
1189 if ((error =
1190 xfs_da_read_buf(state->args->trans, state->args->dp, blkno,
1191 -1, &bp, XFS_DATA_FORK))) {
1192 return error;
1194 ASSERT(bp != NULL);
1196 * Count bytes in the two blocks combined.
1198 leaf = (xfs_dir2_leaf_t *)info;
1199 count = be16_to_cpu(leaf->hdr.count) - be16_to_cpu(leaf->hdr.stale);
1200 bytes = state->blocksize - (state->blocksize >> 2);
1201 leaf = bp->data;
1202 ASSERT(be16_to_cpu(leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1203 count += be16_to_cpu(leaf->hdr.count) - be16_to_cpu(leaf->hdr.stale);
1204 bytes -= count * (uint)sizeof(leaf->ents[0]);
1206 * Fits with at least 25% to spare.
1208 if (bytes >= 0)
1209 break;
1210 xfs_da_brelse(state->args->trans, bp);
1213 * Didn't like either block, give up.
1215 if (i >= 2) {
1216 *action = 0;
1217 return 0;
1220 * Done with the sibling leaf block here, drop the dabuf
1221 * so path_shift can get it.
1223 xfs_da_buf_done(bp);
1225 * Make altpath point to the block we want to keep (the lower
1226 * numbered block) and path point to the block we want to drop.
1228 memcpy(&state->altpath, &state->path, sizeof(state->path));
1229 if (blkno < blk->blkno)
1230 error = xfs_da_path_shift(state, &state->altpath, forward, 0,
1231 &rval);
1232 else
1233 error = xfs_da_path_shift(state, &state->path, forward, 0,
1234 &rval);
1235 if (error) {
1236 return error;
1238 *action = rval ? 0 : 1;
1239 return 0;
1243 * Move all the leaf entries from drop_blk to save_blk.
1244 * This is done as part of a join operation.
1246 void
1247 xfs_dir2_leafn_unbalance(
1248 xfs_da_state_t *state, /* cursor */
1249 xfs_da_state_blk_t *drop_blk, /* dead block */
1250 xfs_da_state_blk_t *save_blk) /* surviving block */
1252 xfs_da_args_t *args; /* operation arguments */
1253 xfs_dir2_leaf_t *drop_leaf; /* dead leaf structure */
1254 xfs_dir2_leaf_t *save_leaf; /* surviving leaf structure */
1256 args = state->args;
1257 ASSERT(drop_blk->magic == XFS_DIR2_LEAFN_MAGIC);
1258 ASSERT(save_blk->magic == XFS_DIR2_LEAFN_MAGIC);
1259 drop_leaf = drop_blk->bp->data;
1260 save_leaf = save_blk->bp->data;
1261 ASSERT(be16_to_cpu(drop_leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1262 ASSERT(be16_to_cpu(save_leaf->hdr.info.magic) == XFS_DIR2_LEAFN_MAGIC);
1264 * If there are any stale leaf entries, take this opportunity
1265 * to purge them.
1267 if (drop_leaf->hdr.stale)
1268 xfs_dir2_leaf_compact(args, drop_blk->bp);
1269 if (save_leaf->hdr.stale)
1270 xfs_dir2_leaf_compact(args, save_blk->bp);
1272 * Move the entries from drop to the appropriate end of save.
1274 drop_blk->hashval = be32_to_cpu(drop_leaf->ents[be16_to_cpu(drop_leaf->hdr.count) - 1].hashval);
1275 if (xfs_dir2_leafn_order(save_blk->bp, drop_blk->bp))
1276 xfs_dir2_leafn_moveents(args, drop_blk->bp, 0, save_blk->bp, 0,
1277 be16_to_cpu(drop_leaf->hdr.count));
1278 else
1279 xfs_dir2_leafn_moveents(args, drop_blk->bp, 0, save_blk->bp,
1280 be16_to_cpu(save_leaf->hdr.count), be16_to_cpu(drop_leaf->hdr.count));
1281 save_blk->hashval = be32_to_cpu(save_leaf->ents[be16_to_cpu(save_leaf->hdr.count) - 1].hashval);
1282 xfs_dir2_leafn_check(args->dp, save_blk->bp);
1286 * Top-level node form directory addname routine.
1288 int /* error */
1289 xfs_dir2_node_addname(
1290 xfs_da_args_t *args) /* operation arguments */
1292 xfs_da_state_blk_t *blk; /* leaf block for insert */
1293 int error; /* error return value */
1294 int rval; /* sub-return value */
1295 xfs_da_state_t *state; /* btree cursor */
1297 xfs_dir2_trace_args("node_addname", args);
1299 * Allocate and initialize the state (btree cursor).
1301 state = xfs_da_state_alloc();
1302 state->args = args;
1303 state->mp = args->dp->i_mount;
1304 state->blocksize = state->mp->m_dirblksize;
1305 state->node_ents = state->mp->m_dir_node_ents;
1307 * Look up the name. We're not supposed to find it, but
1308 * this gives us the insertion point.
1310 error = xfs_da_node_lookup_int(state, &rval);
1311 if (error)
1312 rval = error;
1313 if (rval != ENOENT) {
1314 goto done;
1317 * Add the data entry to a data block.
1318 * Extravalid is set to a freeblock found by lookup.
1320 rval = xfs_dir2_node_addname_int(args,
1321 state->extravalid ? &state->extrablk : NULL);
1322 if (rval) {
1323 goto done;
1325 blk = &state->path.blk[state->path.active - 1];
1326 ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
1328 * Add the new leaf entry.
1330 rval = xfs_dir2_leafn_add(blk->bp, args, blk->index);
1331 if (rval == 0) {
1333 * It worked, fix the hash values up the btree.
1335 if (!args->justcheck)
1336 xfs_da_fixhashpath(state, &state->path);
1337 } else {
1339 * It didn't work, we need to split the leaf block.
1341 if (args->total == 0) {
1342 ASSERT(rval == ENOSPC);
1343 goto done;
1346 * Split the leaf block and insert the new entry.
1348 rval = xfs_da_split(state);
1350 done:
1351 xfs_da_state_free(state);
1352 return rval;
1356 * Add the data entry for a node-format directory name addition.
1357 * The leaf entry is added in xfs_dir2_leafn_add.
1358 * We may enter with a freespace block that the lookup found.
1360 static int /* error */
1361 xfs_dir2_node_addname_int(
1362 xfs_da_args_t *args, /* operation arguments */
1363 xfs_da_state_blk_t *fblk) /* optional freespace block */
1365 xfs_dir2_data_t *data; /* data block structure */
1366 xfs_dir2_db_t dbno; /* data block number */
1367 xfs_dabuf_t *dbp; /* data block buffer */
1368 xfs_dir2_data_entry_t *dep; /* data entry pointer */
1369 xfs_inode_t *dp; /* incore directory inode */
1370 xfs_dir2_data_unused_t *dup; /* data unused entry pointer */
1371 int error; /* error return value */
1372 xfs_dir2_db_t fbno; /* freespace block number */
1373 xfs_dabuf_t *fbp; /* freespace buffer */
1374 int findex; /* freespace entry index */
1375 xfs_dir2_free_t *free=NULL; /* freespace block structure */
1376 xfs_dir2_db_t ifbno; /* initial freespace block no */
1377 xfs_dir2_db_t lastfbno=0; /* highest freespace block no */
1378 int length; /* length of the new entry */
1379 int logfree; /* need to log free entry */
1380 xfs_mount_t *mp; /* filesystem mount point */
1381 int needlog; /* need to log data header */
1382 int needscan; /* need to rescan data frees */
1383 __be16 *tagp; /* data entry tag pointer */
1384 xfs_trans_t *tp; /* transaction pointer */
1386 dp = args->dp;
1387 mp = dp->i_mount;
1388 tp = args->trans;
1389 length = xfs_dir2_data_entsize(args->namelen);
1391 * If we came in with a freespace block that means that lookup
1392 * found an entry with our hash value. This is the freespace
1393 * block for that data entry.
1395 if (fblk) {
1396 fbp = fblk->bp;
1398 * Remember initial freespace block number.
1400 ifbno = fblk->blkno;
1401 free = fbp->data;
1402 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
1403 findex = fblk->index;
1405 * This means the free entry showed that the data block had
1406 * space for our entry, so we remembered it.
1407 * Use that data block.
1409 if (findex >= 0) {
1410 ASSERT(findex < be32_to_cpu(free->hdr.nvalid));
1411 ASSERT(be16_to_cpu(free->bests[findex]) != NULLDATAOFF);
1412 ASSERT(be16_to_cpu(free->bests[findex]) >= length);
1413 dbno = be32_to_cpu(free->hdr.firstdb) + findex;
1416 * The data block looked at didn't have enough room.
1417 * We'll start at the beginning of the freespace entries.
1419 else {
1420 dbno = -1;
1421 findex = 0;
1425 * Didn't come in with a freespace block, so don't have a data block.
1427 else {
1428 ifbno = dbno = -1;
1429 fbp = NULL;
1430 findex = 0;
1433 * If we don't have a data block yet, we're going to scan the
1434 * freespace blocks looking for one. Figure out what the
1435 * highest freespace block number is.
1437 if (dbno == -1) {
1438 xfs_fileoff_t fo; /* freespace block number */
1440 if ((error = xfs_bmap_last_offset(tp, dp, &fo, XFS_DATA_FORK)))
1441 return error;
1442 lastfbno = xfs_dir2_da_to_db(mp, (xfs_dablk_t)fo);
1443 fbno = ifbno;
1446 * While we haven't identified a data block, search the freeblock
1447 * data for a good data block. If we find a null freeblock entry,
1448 * indicating a hole in the data blocks, remember that.
1450 while (dbno == -1) {
1452 * If we don't have a freeblock in hand, get the next one.
1454 if (fbp == NULL) {
1456 * Happens the first time through unless lookup gave
1457 * us a freespace block to start with.
1459 if (++fbno == 0)
1460 fbno = XFS_DIR2_FREE_FIRSTDB(mp);
1462 * If it's ifbno we already looked at it.
1464 if (fbno == ifbno)
1465 fbno++;
1467 * If it's off the end we're done.
1469 if (fbno >= lastfbno)
1470 break;
1472 * Read the block. There can be holes in the
1473 * freespace blocks, so this might not succeed.
1474 * This should be really rare, so there's no reason
1475 * to avoid it.
1477 if ((error = xfs_da_read_buf(tp, dp,
1478 xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
1479 XFS_DATA_FORK))) {
1480 return error;
1482 if (unlikely(fbp == NULL)) {
1483 continue;
1485 free = fbp->data;
1486 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
1487 findex = 0;
1490 * Look at the current free entry. Is it good enough?
1492 if (be16_to_cpu(free->bests[findex]) != NULLDATAOFF &&
1493 be16_to_cpu(free->bests[findex]) >= length)
1494 dbno = be32_to_cpu(free->hdr.firstdb) + findex;
1495 else {
1497 * Are we done with the freeblock?
1499 if (++findex == be32_to_cpu(free->hdr.nvalid)) {
1501 * Drop the block.
1503 xfs_da_brelse(tp, fbp);
1504 fbp = NULL;
1505 if (fblk && fblk->bp)
1506 fblk->bp = NULL;
1511 * If we don't have a data block, we need to allocate one and make
1512 * the freespace entries refer to it.
1514 if (unlikely(dbno == -1)) {
1516 * Not allowed to allocate, return failure.
1518 if (args->justcheck || args->total == 0) {
1520 * Drop the freespace buffer unless it came from our
1521 * caller.
1523 if ((fblk == NULL || fblk->bp == NULL) && fbp != NULL)
1524 xfs_da_buf_done(fbp);
1525 return XFS_ERROR(ENOSPC);
1528 * Allocate and initialize the new data block.
1530 if (unlikely((error = xfs_dir2_grow_inode(args,
1531 XFS_DIR2_DATA_SPACE,
1532 &dbno)) ||
1533 (error = xfs_dir2_data_init(args, dbno, &dbp)))) {
1535 * Drop the freespace buffer unless it came from our
1536 * caller.
1538 if ((fblk == NULL || fblk->bp == NULL) && fbp != NULL)
1539 xfs_da_buf_done(fbp);
1540 return error;
1543 * If (somehow) we have a freespace block, get rid of it.
1545 if (fbp)
1546 xfs_da_brelse(tp, fbp);
1547 if (fblk && fblk->bp)
1548 fblk->bp = NULL;
1551 * Get the freespace block corresponding to the data block
1552 * that was just allocated.
1554 fbno = xfs_dir2_db_to_fdb(mp, dbno);
1555 if (unlikely(error = xfs_da_read_buf(tp, dp,
1556 xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
1557 XFS_DATA_FORK))) {
1558 xfs_da_buf_done(dbp);
1559 return error;
1562 * If there wasn't a freespace block, the read will
1563 * return a NULL fbp. Allocate and initialize a new one.
1565 if( fbp == NULL ) {
1566 if ((error = xfs_dir2_grow_inode(args, XFS_DIR2_FREE_SPACE,
1567 &fbno))) {
1568 return error;
1571 if (unlikely(xfs_dir2_db_to_fdb(mp, dbno) != fbno)) {
1572 cmn_err(CE_ALERT,
1573 "xfs_dir2_node_addname_int: dir ino "
1574 "%llu needed freesp block %lld for\n"
1575 " data block %lld, got %lld\n"
1576 " ifbno %llu lastfbno %d\n",
1577 (unsigned long long)dp->i_ino,
1578 (long long)xfs_dir2_db_to_fdb(mp, dbno),
1579 (long long)dbno, (long long)fbno,
1580 (unsigned long long)ifbno, lastfbno);
1581 if (fblk) {
1582 cmn_err(CE_ALERT,
1583 " fblk 0x%p blkno %llu "
1584 "index %d magic 0x%x\n",
1585 fblk,
1586 (unsigned long long)fblk->blkno,
1587 fblk->index,
1588 fblk->magic);
1589 } else {
1590 cmn_err(CE_ALERT,
1591 " ... fblk is NULL\n");
1593 XFS_ERROR_REPORT("xfs_dir2_node_addname_int",
1594 XFS_ERRLEVEL_LOW, mp);
1595 return XFS_ERROR(EFSCORRUPTED);
1599 * Get a buffer for the new block.
1601 if ((error = xfs_da_get_buf(tp, dp,
1602 xfs_dir2_db_to_da(mp, fbno),
1603 -1, &fbp, XFS_DATA_FORK))) {
1604 return error;
1606 ASSERT(fbp != NULL);
1609 * Initialize the new block to be empty, and remember
1610 * its first slot as our empty slot.
1612 free = fbp->data;
1613 free->hdr.magic = cpu_to_be32(XFS_DIR2_FREE_MAGIC);
1614 free->hdr.firstdb = cpu_to_be32(
1615 (fbno - XFS_DIR2_FREE_FIRSTDB(mp)) *
1616 XFS_DIR2_MAX_FREE_BESTS(mp));
1617 free->hdr.nvalid = 0;
1618 free->hdr.nused = 0;
1619 } else {
1620 free = fbp->data;
1621 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
1625 * Set the freespace block index from the data block number.
1627 findex = xfs_dir2_db_to_fdindex(mp, dbno);
1629 * If it's after the end of the current entries in the
1630 * freespace block, extend that table.
1632 if (findex >= be32_to_cpu(free->hdr.nvalid)) {
1633 ASSERT(findex < XFS_DIR2_MAX_FREE_BESTS(mp));
1634 free->hdr.nvalid = cpu_to_be32(findex + 1);
1636 * Tag new entry so nused will go up.
1638 free->bests[findex] = cpu_to_be16(NULLDATAOFF);
1641 * If this entry was for an empty data block
1642 * (this should always be true) then update the header.
1644 if (be16_to_cpu(free->bests[findex]) == NULLDATAOFF) {
1645 be32_add_cpu(&free->hdr.nused, 1);
1646 xfs_dir2_free_log_header(tp, fbp);
1649 * Update the real value in the table.
1650 * We haven't allocated the data entry yet so this will
1651 * change again.
1653 data = dbp->data;
1654 free->bests[findex] = data->hdr.bestfree[0].length;
1655 logfree = 1;
1658 * We had a data block so we don't have to make a new one.
1660 else {
1662 * If just checking, we succeeded.
1664 if (args->justcheck) {
1665 if ((fblk == NULL || fblk->bp == NULL) && fbp != NULL)
1666 xfs_da_buf_done(fbp);
1667 return 0;
1670 * Read the data block in.
1672 if (unlikely(
1673 error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, dbno),
1674 -1, &dbp, XFS_DATA_FORK))) {
1675 if ((fblk == NULL || fblk->bp == NULL) && fbp != NULL)
1676 xfs_da_buf_done(fbp);
1677 return error;
1679 data = dbp->data;
1680 logfree = 0;
1682 ASSERT(be16_to_cpu(data->hdr.bestfree[0].length) >= length);
1684 * Point to the existing unused space.
1686 dup = (xfs_dir2_data_unused_t *)
1687 ((char *)data + be16_to_cpu(data->hdr.bestfree[0].offset));
1688 needscan = needlog = 0;
1690 * Mark the first part of the unused space, inuse for us.
1692 xfs_dir2_data_use_free(tp, dbp, dup,
1693 (xfs_dir2_data_aoff_t)((char *)dup - (char *)data), length,
1694 &needlog, &needscan);
1696 * Fill in the new entry and log it.
1698 dep = (xfs_dir2_data_entry_t *)dup;
1699 dep->inumber = cpu_to_be64(args->inumber);
1700 dep->namelen = args->namelen;
1701 memcpy(dep->name, args->name, dep->namelen);
1702 tagp = xfs_dir2_data_entry_tag_p(dep);
1703 *tagp = cpu_to_be16((char *)dep - (char *)data);
1704 xfs_dir2_data_log_entry(tp, dbp, dep);
1706 * Rescan the block for bestfree if needed.
1708 if (needscan)
1709 xfs_dir2_data_freescan(mp, data, &needlog);
1711 * Log the data block header if needed.
1713 if (needlog)
1714 xfs_dir2_data_log_header(tp, dbp);
1716 * If the freespace entry is now wrong, update it.
1718 if (be16_to_cpu(free->bests[findex]) != be16_to_cpu(data->hdr.bestfree[0].length)) {
1719 free->bests[findex] = data->hdr.bestfree[0].length;
1720 logfree = 1;
1723 * Log the freespace entry if needed.
1725 if (logfree)
1726 xfs_dir2_free_log_bests(tp, fbp, findex, findex);
1728 * If the caller didn't hand us the freespace block, drop it.
1730 if ((fblk == NULL || fblk->bp == NULL) && fbp != NULL)
1731 xfs_da_buf_done(fbp);
1733 * Return the data block and offset in args, then drop the data block.
1735 args->blkno = (xfs_dablk_t)dbno;
1736 args->index = be16_to_cpu(*tagp);
1737 xfs_da_buf_done(dbp);
1738 return 0;
1742 * Lookup an entry in a node-format directory.
1743 * All the real work happens in xfs_da_node_lookup_int.
1744 * The only real output is the inode number of the entry.
1746 int /* error */
1747 xfs_dir2_node_lookup(
1748 xfs_da_args_t *args) /* operation arguments */
1750 int error; /* error return value */
1751 int i; /* btree level */
1752 int rval; /* operation return value */
1753 xfs_da_state_t *state; /* btree cursor */
1755 xfs_dir2_trace_args("node_lookup", args);
1757 * Allocate and initialize the btree cursor.
1759 state = xfs_da_state_alloc();
1760 state->args = args;
1761 state->mp = args->dp->i_mount;
1762 state->blocksize = state->mp->m_dirblksize;
1763 state->node_ents = state->mp->m_dir_node_ents;
1765 * Fill in the path to the entry in the cursor.
1767 error = xfs_da_node_lookup_int(state, &rval);
1768 if (error)
1769 rval = error;
1771 * Release the btree blocks and leaf block.
1773 for (i = 0; i < state->path.active; i++) {
1774 xfs_da_brelse(args->trans, state->path.blk[i].bp);
1775 state->path.blk[i].bp = NULL;
1778 * Release the data block if we have it.
1780 if (state->extravalid && state->extrablk.bp) {
1781 xfs_da_brelse(args->trans, state->extrablk.bp);
1782 state->extrablk.bp = NULL;
1784 xfs_da_state_free(state);
1785 return rval;
1789 * Remove an entry from a node-format directory.
1791 int /* error */
1792 xfs_dir2_node_removename(
1793 xfs_da_args_t *args) /* operation arguments */
1795 xfs_da_state_blk_t *blk; /* leaf block */
1796 int error; /* error return value */
1797 int rval; /* operation return value */
1798 xfs_da_state_t *state; /* btree cursor */
1800 xfs_dir2_trace_args("node_removename", args);
1802 * Allocate and initialize the btree cursor.
1804 state = xfs_da_state_alloc();
1805 state->args = args;
1806 state->mp = args->dp->i_mount;
1807 state->blocksize = state->mp->m_dirblksize;
1808 state->node_ents = state->mp->m_dir_node_ents;
1810 * Look up the entry we're deleting, set up the cursor.
1812 error = xfs_da_node_lookup_int(state, &rval);
1813 if (error) {
1814 rval = error;
1817 * Didn't find it, upper layer screwed up.
1819 if (rval != EEXIST) {
1820 xfs_da_state_free(state);
1821 return rval;
1823 blk = &state->path.blk[state->path.active - 1];
1824 ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
1825 ASSERT(state->extravalid);
1827 * Remove the leaf and data entries.
1828 * Extrablk refers to the data block.
1830 error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
1831 &state->extrablk, &rval);
1832 if (error) {
1833 return error;
1836 * Fix the hash values up the btree.
1838 xfs_da_fixhashpath(state, &state->path);
1840 * If we need to join leaf blocks, do it.
1842 if (rval && state->path.active > 1)
1843 error = xfs_da_join(state);
1845 * If no errors so far, try conversion to leaf format.
1847 if (!error)
1848 error = xfs_dir2_node_to_leaf(state);
1849 xfs_da_state_free(state);
1850 return error;
1854 * Replace an entry's inode number in a node-format directory.
1856 int /* error */
1857 xfs_dir2_node_replace(
1858 xfs_da_args_t *args) /* operation arguments */
1860 xfs_da_state_blk_t *blk; /* leaf block */
1861 xfs_dir2_data_t *data; /* data block structure */
1862 xfs_dir2_data_entry_t *dep; /* data entry changed */
1863 int error; /* error return value */
1864 int i; /* btree level */
1865 xfs_ino_t inum; /* new inode number */
1866 xfs_dir2_leaf_t *leaf; /* leaf structure */
1867 xfs_dir2_leaf_entry_t *lep; /* leaf entry being changed */
1868 int rval; /* internal return value */
1869 xfs_da_state_t *state; /* btree cursor */
1871 xfs_dir2_trace_args("node_replace", args);
1873 * Allocate and initialize the btree cursor.
1875 state = xfs_da_state_alloc();
1876 state->args = args;
1877 state->mp = args->dp->i_mount;
1878 state->blocksize = state->mp->m_dirblksize;
1879 state->node_ents = state->mp->m_dir_node_ents;
1880 inum = args->inumber;
1882 * Lookup the entry to change in the btree.
1884 error = xfs_da_node_lookup_int(state, &rval);
1885 if (error) {
1886 rval = error;
1889 * It should be found, since the vnodeops layer has looked it up
1890 * and locked it. But paranoia is good.
1892 if (rval == EEXIST) {
1894 * Find the leaf entry.
1896 blk = &state->path.blk[state->path.active - 1];
1897 ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
1898 leaf = blk->bp->data;
1899 lep = &leaf->ents[blk->index];
1900 ASSERT(state->extravalid);
1902 * Point to the data entry.
1904 data = state->extrablk.bp->data;
1905 ASSERT(be32_to_cpu(data->hdr.magic) == XFS_DIR2_DATA_MAGIC);
1906 dep = (xfs_dir2_data_entry_t *)
1907 ((char *)data +
1908 xfs_dir2_dataptr_to_off(state->mp, be32_to_cpu(lep->address)));
1909 ASSERT(inum != be64_to_cpu(dep->inumber));
1911 * Fill in the new inode number and log the entry.
1913 dep->inumber = cpu_to_be64(inum);
1914 xfs_dir2_data_log_entry(args->trans, state->extrablk.bp, dep);
1915 rval = 0;
1918 * Didn't find it, and we're holding a data block. Drop it.
1920 else if (state->extravalid) {
1921 xfs_da_brelse(args->trans, state->extrablk.bp);
1922 state->extrablk.bp = NULL;
1925 * Release all the buffers in the cursor.
1927 for (i = 0; i < state->path.active; i++) {
1928 xfs_da_brelse(args->trans, state->path.blk[i].bp);
1929 state->path.blk[i].bp = NULL;
1931 xfs_da_state_free(state);
1932 return rval;
1936 * Trim off a trailing empty freespace block.
1937 * Return (in rvalp) 1 if we did it, 0 if not.
1939 int /* error */
1940 xfs_dir2_node_trim_free(
1941 xfs_da_args_t *args, /* operation arguments */
1942 xfs_fileoff_t fo, /* free block number */
1943 int *rvalp) /* out: did something */
1945 xfs_dabuf_t *bp; /* freespace buffer */
1946 xfs_inode_t *dp; /* incore directory inode */
1947 int error; /* error return code */
1948 xfs_dir2_free_t *free; /* freespace structure */
1949 xfs_mount_t *mp; /* filesystem mount point */
1950 xfs_trans_t *tp; /* transaction pointer */
1952 dp = args->dp;
1953 mp = dp->i_mount;
1954 tp = args->trans;
1956 * Read the freespace block.
1958 if (unlikely(error = xfs_da_read_buf(tp, dp, (xfs_dablk_t)fo, -2, &bp,
1959 XFS_DATA_FORK))) {
1960 return error;
1964 * There can be holes in freespace. If fo is a hole, there's
1965 * nothing to do.
1967 if (bp == NULL) {
1968 return 0;
1970 free = bp->data;
1971 ASSERT(be32_to_cpu(free->hdr.magic) == XFS_DIR2_FREE_MAGIC);
1973 * If there are used entries, there's nothing to do.
1975 if (be32_to_cpu(free->hdr.nused) > 0) {
1976 xfs_da_brelse(tp, bp);
1977 *rvalp = 0;
1978 return 0;
1981 * Blow the block away.
1983 if ((error =
1984 xfs_dir2_shrink_inode(args, xfs_dir2_da_to_db(mp, (xfs_dablk_t)fo),
1985 bp))) {
1987 * Can't fail with ENOSPC since that only happens with no
1988 * space reservation, when breaking up an extent into two
1989 * pieces. This is the last block of an extent.
1991 ASSERT(error != ENOSPC);
1992 xfs_da_brelse(tp, bp);
1993 return error;
1996 * Return that we succeeded.
1998 *rvalp = 1;
1999 return 0;