Loading...
Note: File does not exist in v3.15.
1/*
2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * Copyright (c) 2013 Red Hat, Inc.
4 * All Rights Reserved.
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as
8 * published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it would be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 */
19#include "xfs.h"
20#include "xfs_fs.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_mount.h"
25#include "xfs_da_format.h"
26#include "xfs_da_btree.h"
27#include "xfs_inode.h"
28#include "xfs_dir2.h"
29#include "xfs_dir2_priv.h"
30#include "xfs_error.h"
31#include "xfs_trans.h"
32#include "xfs_buf_item.h"
33#include "xfs_cksum.h"
34#include "xfs_log.h"
35
36/*
37 * Check the consistency of the data block.
38 * The input can also be a block-format directory.
39 * Return 0 is the buffer is good, otherwise an error.
40 */
41int
42__xfs_dir3_data_check(
43 struct xfs_inode *dp, /* incore inode pointer */
44 struct xfs_buf *bp) /* data block's buffer */
45{
46 xfs_dir2_dataptr_t addr; /* addr for leaf lookup */
47 xfs_dir2_data_free_t *bf; /* bestfree table */
48 xfs_dir2_block_tail_t *btp=NULL; /* block tail */
49 int count; /* count of entries found */
50 xfs_dir2_data_hdr_t *hdr; /* data block header */
51 xfs_dir2_data_entry_t *dep; /* data entry */
52 xfs_dir2_data_free_t *dfp; /* bestfree entry */
53 xfs_dir2_data_unused_t *dup; /* unused entry */
54 char *endp; /* end of useful data */
55 int freeseen; /* mask of bestfrees seen */
56 xfs_dahash_t hash; /* hash of current name */
57 int i; /* leaf index */
58 int lastfree; /* last entry was unused */
59 xfs_dir2_leaf_entry_t *lep=NULL; /* block leaf entries */
60 xfs_mount_t *mp; /* filesystem mount point */
61 char *p; /* current data position */
62 int stale; /* count of stale leaves */
63 struct xfs_name name;
64 const struct xfs_dir_ops *ops;
65 struct xfs_da_geometry *geo;
66
67 mp = bp->b_target->bt_mount;
68 geo = mp->m_dir_geo;
69
70 /*
71 * We can be passed a null dp here from a verifier, so we need to go the
72 * hard way to get them.
73 */
74 ops = xfs_dir_get_ops(mp, dp);
75
76 hdr = bp->b_addr;
77 p = (char *)ops->data_entry_p(hdr);
78
79 switch (hdr->magic) {
80 case cpu_to_be32(XFS_DIR3_BLOCK_MAGIC):
81 case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
82 btp = xfs_dir2_block_tail_p(geo, hdr);
83 lep = xfs_dir2_block_leaf_p(btp);
84 endp = (char *)lep;
85
86 /*
87 * The number of leaf entries is limited by the size of the
88 * block and the amount of space used by the data entries.
89 * We don't know how much space is used by the data entries yet,
90 * so just ensure that the count falls somewhere inside the
91 * block right now.
92 */
93 XFS_WANT_CORRUPTED_RETURN(mp, be32_to_cpu(btp->count) <
94 ((char *)btp - p) / sizeof(struct xfs_dir2_leaf_entry));
95 break;
96 case cpu_to_be32(XFS_DIR3_DATA_MAGIC):
97 case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
98 endp = (char *)hdr + geo->blksize;
99 break;
100 default:
101 XFS_ERROR_REPORT("Bad Magic", XFS_ERRLEVEL_LOW, mp);
102 return -EFSCORRUPTED;
103 }
104
105 /*
106 * Account for zero bestfree entries.
107 */
108 bf = ops->data_bestfree_p(hdr);
109 count = lastfree = freeseen = 0;
110 if (!bf[0].length) {
111 XFS_WANT_CORRUPTED_RETURN(mp, !bf[0].offset);
112 freeseen |= 1 << 0;
113 }
114 if (!bf[1].length) {
115 XFS_WANT_CORRUPTED_RETURN(mp, !bf[1].offset);
116 freeseen |= 1 << 1;
117 }
118 if (!bf[2].length) {
119 XFS_WANT_CORRUPTED_RETURN(mp, !bf[2].offset);
120 freeseen |= 1 << 2;
121 }
122
123 XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(bf[0].length) >=
124 be16_to_cpu(bf[1].length));
125 XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(bf[1].length) >=
126 be16_to_cpu(bf[2].length));
127 /*
128 * Loop over the data/unused entries.
129 */
130 while (p < endp) {
131 dup = (xfs_dir2_data_unused_t *)p;
132 /*
133 * If it's unused, look for the space in the bestfree table.
134 * If we find it, account for that, else make sure it
135 * doesn't need to be there.
136 */
137 if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
138 XFS_WANT_CORRUPTED_RETURN(mp, lastfree == 0);
139 XFS_WANT_CORRUPTED_RETURN(mp,
140 be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) ==
141 (char *)dup - (char *)hdr);
142 dfp = xfs_dir2_data_freefind(hdr, bf, dup);
143 if (dfp) {
144 i = (int)(dfp - bf);
145 XFS_WANT_CORRUPTED_RETURN(mp,
146 (freeseen & (1 << i)) == 0);
147 freeseen |= 1 << i;
148 } else {
149 XFS_WANT_CORRUPTED_RETURN(mp,
150 be16_to_cpu(dup->length) <=
151 be16_to_cpu(bf[2].length));
152 }
153 p += be16_to_cpu(dup->length);
154 lastfree = 1;
155 continue;
156 }
157 /*
158 * It's a real entry. Validate the fields.
159 * If this is a block directory then make sure it's
160 * in the leaf section of the block.
161 * The linear search is crude but this is DEBUG code.
162 */
163 dep = (xfs_dir2_data_entry_t *)p;
164 XFS_WANT_CORRUPTED_RETURN(mp, dep->namelen != 0);
165 XFS_WANT_CORRUPTED_RETURN(mp,
166 !xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)));
167 XFS_WANT_CORRUPTED_RETURN(mp,
168 be16_to_cpu(*ops->data_entry_tag_p(dep)) ==
169 (char *)dep - (char *)hdr);
170 XFS_WANT_CORRUPTED_RETURN(mp,
171 ops->data_get_ftype(dep) < XFS_DIR3_FT_MAX);
172 count++;
173 lastfree = 0;
174 if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
175 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC)) {
176 addr = xfs_dir2_db_off_to_dataptr(geo, geo->datablk,
177 (xfs_dir2_data_aoff_t)
178 ((char *)dep - (char *)hdr));
179 name.name = dep->name;
180 name.len = dep->namelen;
181 hash = mp->m_dirnameops->hashname(&name);
182 for (i = 0; i < be32_to_cpu(btp->count); i++) {
183 if (be32_to_cpu(lep[i].address) == addr &&
184 be32_to_cpu(lep[i].hashval) == hash)
185 break;
186 }
187 XFS_WANT_CORRUPTED_RETURN(mp,
188 i < be32_to_cpu(btp->count));
189 }
190 p += ops->data_entsize(dep->namelen);
191 }
192 /*
193 * Need to have seen all the entries and all the bestfree slots.
194 */
195 XFS_WANT_CORRUPTED_RETURN(mp, freeseen == 7);
196 if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
197 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC)) {
198 for (i = stale = 0; i < be32_to_cpu(btp->count); i++) {
199 if (lep[i].address ==
200 cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
201 stale++;
202 if (i > 0)
203 XFS_WANT_CORRUPTED_RETURN(mp,
204 be32_to_cpu(lep[i].hashval) >=
205 be32_to_cpu(lep[i - 1].hashval));
206 }
207 XFS_WANT_CORRUPTED_RETURN(mp, count ==
208 be32_to_cpu(btp->count) - be32_to_cpu(btp->stale));
209 XFS_WANT_CORRUPTED_RETURN(mp, stale == be32_to_cpu(btp->stale));
210 }
211 return 0;
212}
213
214static bool
215xfs_dir3_data_verify(
216 struct xfs_buf *bp)
217{
218 struct xfs_mount *mp = bp->b_target->bt_mount;
219 struct xfs_dir3_blk_hdr *hdr3 = bp->b_addr;
220
221 if (xfs_sb_version_hascrc(&mp->m_sb)) {
222 if (hdr3->magic != cpu_to_be32(XFS_DIR3_DATA_MAGIC))
223 return false;
224 if (!uuid_equal(&hdr3->uuid, &mp->m_sb.sb_meta_uuid))
225 return false;
226 if (be64_to_cpu(hdr3->blkno) != bp->b_bn)
227 return false;
228 if (!xfs_log_check_lsn(mp, be64_to_cpu(hdr3->lsn)))
229 return false;
230 } else {
231 if (hdr3->magic != cpu_to_be32(XFS_DIR2_DATA_MAGIC))
232 return false;
233 }
234 if (__xfs_dir3_data_check(NULL, bp))
235 return false;
236 return true;
237}
238
239/*
240 * Readahead of the first block of the directory when it is opened is completely
241 * oblivious to the format of the directory. Hence we can either get a block
242 * format buffer or a data format buffer on readahead.
243 */
244static void
245xfs_dir3_data_reada_verify(
246 struct xfs_buf *bp)
247{
248 struct xfs_dir2_data_hdr *hdr = bp->b_addr;
249
250 switch (hdr->magic) {
251 case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
252 case cpu_to_be32(XFS_DIR3_BLOCK_MAGIC):
253 bp->b_ops = &xfs_dir3_block_buf_ops;
254 bp->b_ops->verify_read(bp);
255 return;
256 case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
257 case cpu_to_be32(XFS_DIR3_DATA_MAGIC):
258 bp->b_ops = &xfs_dir3_data_buf_ops;
259 bp->b_ops->verify_read(bp);
260 return;
261 default:
262 xfs_buf_ioerror(bp, -EFSCORRUPTED);
263 xfs_verifier_error(bp);
264 break;
265 }
266}
267
268static void
269xfs_dir3_data_read_verify(
270 struct xfs_buf *bp)
271{
272 struct xfs_mount *mp = bp->b_target->bt_mount;
273
274 if (xfs_sb_version_hascrc(&mp->m_sb) &&
275 !xfs_buf_verify_cksum(bp, XFS_DIR3_DATA_CRC_OFF))
276 xfs_buf_ioerror(bp, -EFSBADCRC);
277 else if (!xfs_dir3_data_verify(bp))
278 xfs_buf_ioerror(bp, -EFSCORRUPTED);
279
280 if (bp->b_error)
281 xfs_verifier_error(bp);
282}
283
284static void
285xfs_dir3_data_write_verify(
286 struct xfs_buf *bp)
287{
288 struct xfs_mount *mp = bp->b_target->bt_mount;
289 struct xfs_buf_log_item *bip = bp->b_fspriv;
290 struct xfs_dir3_blk_hdr *hdr3 = bp->b_addr;
291
292 if (!xfs_dir3_data_verify(bp)) {
293 xfs_buf_ioerror(bp, -EFSCORRUPTED);
294 xfs_verifier_error(bp);
295 return;
296 }
297
298 if (!xfs_sb_version_hascrc(&mp->m_sb))
299 return;
300
301 if (bip)
302 hdr3->lsn = cpu_to_be64(bip->bli_item.li_lsn);
303
304 xfs_buf_update_cksum(bp, XFS_DIR3_DATA_CRC_OFF);
305}
306
307const struct xfs_buf_ops xfs_dir3_data_buf_ops = {
308 .name = "xfs_dir3_data",
309 .verify_read = xfs_dir3_data_read_verify,
310 .verify_write = xfs_dir3_data_write_verify,
311};
312
313static const struct xfs_buf_ops xfs_dir3_data_reada_buf_ops = {
314 .name = "xfs_dir3_data_reada",
315 .verify_read = xfs_dir3_data_reada_verify,
316 .verify_write = xfs_dir3_data_write_verify,
317};
318
319
320int
321xfs_dir3_data_read(
322 struct xfs_trans *tp,
323 struct xfs_inode *dp,
324 xfs_dablk_t bno,
325 xfs_daddr_t mapped_bno,
326 struct xfs_buf **bpp)
327{
328 int err;
329
330 err = xfs_da_read_buf(tp, dp, bno, mapped_bno, bpp,
331 XFS_DATA_FORK, &xfs_dir3_data_buf_ops);
332 if (!err && tp)
333 xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_DATA_BUF);
334 return err;
335}
336
337int
338xfs_dir3_data_readahead(
339 struct xfs_inode *dp,
340 xfs_dablk_t bno,
341 xfs_daddr_t mapped_bno)
342{
343 return xfs_da_reada_buf(dp, bno, mapped_bno,
344 XFS_DATA_FORK, &xfs_dir3_data_reada_buf_ops);
345}
346
347/*
348 * Given a data block and an unused entry from that block,
349 * return the bestfree entry if any that corresponds to it.
350 */
351xfs_dir2_data_free_t *
352xfs_dir2_data_freefind(
353 struct xfs_dir2_data_hdr *hdr, /* data block header */
354 struct xfs_dir2_data_free *bf, /* bestfree table pointer */
355 struct xfs_dir2_data_unused *dup) /* unused space */
356{
357 xfs_dir2_data_free_t *dfp; /* bestfree entry */
358 xfs_dir2_data_aoff_t off; /* offset value needed */
359#ifdef DEBUG
360 int matched; /* matched the value */
361 int seenzero; /* saw a 0 bestfree entry */
362#endif
363
364 off = (xfs_dir2_data_aoff_t)((char *)dup - (char *)hdr);
365
366#ifdef DEBUG
367 /*
368 * Validate some consistency in the bestfree table.
369 * Check order, non-overlapping entries, and if we find the
370 * one we're looking for it has to be exact.
371 */
372 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
373 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
374 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
375 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
376 for (dfp = &bf[0], seenzero = matched = 0;
377 dfp < &bf[XFS_DIR2_DATA_FD_COUNT];
378 dfp++) {
379 if (!dfp->offset) {
380 ASSERT(!dfp->length);
381 seenzero = 1;
382 continue;
383 }
384 ASSERT(seenzero == 0);
385 if (be16_to_cpu(dfp->offset) == off) {
386 matched = 1;
387 ASSERT(dfp->length == dup->length);
388 } else if (off < be16_to_cpu(dfp->offset))
389 ASSERT(off + be16_to_cpu(dup->length) <= be16_to_cpu(dfp->offset));
390 else
391 ASSERT(be16_to_cpu(dfp->offset) + be16_to_cpu(dfp->length) <= off);
392 ASSERT(matched || be16_to_cpu(dfp->length) >= be16_to_cpu(dup->length));
393 if (dfp > &bf[0])
394 ASSERT(be16_to_cpu(dfp[-1].length) >= be16_to_cpu(dfp[0].length));
395 }
396#endif
397 /*
398 * If this is smaller than the smallest bestfree entry,
399 * it can't be there since they're sorted.
400 */
401 if (be16_to_cpu(dup->length) <
402 be16_to_cpu(bf[XFS_DIR2_DATA_FD_COUNT - 1].length))
403 return NULL;
404 /*
405 * Look at the three bestfree entries for our guy.
406 */
407 for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
408 if (!dfp->offset)
409 return NULL;
410 if (be16_to_cpu(dfp->offset) == off)
411 return dfp;
412 }
413 /*
414 * Didn't find it. This only happens if there are duplicate lengths.
415 */
416 return NULL;
417}
418
419/*
420 * Insert an unused-space entry into the bestfree table.
421 */
422xfs_dir2_data_free_t * /* entry inserted */
423xfs_dir2_data_freeinsert(
424 struct xfs_dir2_data_hdr *hdr, /* data block pointer */
425 struct xfs_dir2_data_free *dfp, /* bestfree table pointer */
426 struct xfs_dir2_data_unused *dup, /* unused space */
427 int *loghead) /* log the data header (out) */
428{
429 xfs_dir2_data_free_t new; /* new bestfree entry */
430
431 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
432 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
433 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
434 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
435
436 new.length = dup->length;
437 new.offset = cpu_to_be16((char *)dup - (char *)hdr);
438
439 /*
440 * Insert at position 0, 1, or 2; or not at all.
441 */
442 if (be16_to_cpu(new.length) > be16_to_cpu(dfp[0].length)) {
443 dfp[2] = dfp[1];
444 dfp[1] = dfp[0];
445 dfp[0] = new;
446 *loghead = 1;
447 return &dfp[0];
448 }
449 if (be16_to_cpu(new.length) > be16_to_cpu(dfp[1].length)) {
450 dfp[2] = dfp[1];
451 dfp[1] = new;
452 *loghead = 1;
453 return &dfp[1];
454 }
455 if (be16_to_cpu(new.length) > be16_to_cpu(dfp[2].length)) {
456 dfp[2] = new;
457 *loghead = 1;
458 return &dfp[2];
459 }
460 return NULL;
461}
462
463/*
464 * Remove a bestfree entry from the table.
465 */
466STATIC void
467xfs_dir2_data_freeremove(
468 struct xfs_dir2_data_hdr *hdr, /* data block header */
469 struct xfs_dir2_data_free *bf, /* bestfree table pointer */
470 struct xfs_dir2_data_free *dfp, /* bestfree entry pointer */
471 int *loghead) /* out: log data header */
472{
473
474 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
475 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
476 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
477 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
478
479 /*
480 * It's the first entry, slide the next 2 up.
481 */
482 if (dfp == &bf[0]) {
483 bf[0] = bf[1];
484 bf[1] = bf[2];
485 }
486 /*
487 * It's the second entry, slide the 3rd entry up.
488 */
489 else if (dfp == &bf[1])
490 bf[1] = bf[2];
491 /*
492 * Must be the last entry.
493 */
494 else
495 ASSERT(dfp == &bf[2]);
496 /*
497 * Clear the 3rd entry, must be zero now.
498 */
499 bf[2].length = 0;
500 bf[2].offset = 0;
501 *loghead = 1;
502}
503
504/*
505 * Given a data block, reconstruct its bestfree map.
506 */
507void
508xfs_dir2_data_freescan(
509 struct xfs_inode *dp,
510 struct xfs_dir2_data_hdr *hdr,
511 int *loghead)
512{
513 xfs_dir2_block_tail_t *btp; /* block tail */
514 xfs_dir2_data_entry_t *dep; /* active data entry */
515 xfs_dir2_data_unused_t *dup; /* unused data entry */
516 struct xfs_dir2_data_free *bf;
517 char *endp; /* end of block's data */
518 char *p; /* current entry pointer */
519 struct xfs_da_geometry *geo = dp->i_mount->m_dir_geo;
520
521 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
522 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
523 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
524 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
525
526 /*
527 * Start by clearing the table.
528 */
529 bf = dp->d_ops->data_bestfree_p(hdr);
530 memset(bf, 0, sizeof(*bf) * XFS_DIR2_DATA_FD_COUNT);
531 *loghead = 1;
532 /*
533 * Set up pointers.
534 */
535 p = (char *)dp->d_ops->data_entry_p(hdr);
536 if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
537 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC)) {
538 btp = xfs_dir2_block_tail_p(geo, hdr);
539 endp = (char *)xfs_dir2_block_leaf_p(btp);
540 } else
541 endp = (char *)hdr + geo->blksize;
542 /*
543 * Loop over the block's entries.
544 */
545 while (p < endp) {
546 dup = (xfs_dir2_data_unused_t *)p;
547 /*
548 * If it's a free entry, insert it.
549 */
550 if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
551 ASSERT((char *)dup - (char *)hdr ==
552 be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)));
553 xfs_dir2_data_freeinsert(hdr, bf, dup, loghead);
554 p += be16_to_cpu(dup->length);
555 }
556 /*
557 * For active entries, check their tags and skip them.
558 */
559 else {
560 dep = (xfs_dir2_data_entry_t *)p;
561 ASSERT((char *)dep - (char *)hdr ==
562 be16_to_cpu(*dp->d_ops->data_entry_tag_p(dep)));
563 p += dp->d_ops->data_entsize(dep->namelen);
564 }
565 }
566}
567
568/*
569 * Initialize a data block at the given block number in the directory.
570 * Give back the buffer for the created block.
571 */
572int /* error */
573xfs_dir3_data_init(
574 xfs_da_args_t *args, /* directory operation args */
575 xfs_dir2_db_t blkno, /* logical dir block number */
576 struct xfs_buf **bpp) /* output block buffer */
577{
578 struct xfs_buf *bp; /* block buffer */
579 xfs_dir2_data_hdr_t *hdr; /* data block header */
580 xfs_inode_t *dp; /* incore directory inode */
581 xfs_dir2_data_unused_t *dup; /* unused entry pointer */
582 struct xfs_dir2_data_free *bf;
583 int error; /* error return value */
584 int i; /* bestfree index */
585 xfs_mount_t *mp; /* filesystem mount point */
586 xfs_trans_t *tp; /* transaction pointer */
587 int t; /* temp */
588
589 dp = args->dp;
590 mp = dp->i_mount;
591 tp = args->trans;
592 /*
593 * Get the buffer set up for the block.
594 */
595 error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(args->geo, blkno),
596 -1, &bp, XFS_DATA_FORK);
597 if (error)
598 return error;
599 bp->b_ops = &xfs_dir3_data_buf_ops;
600 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_DATA_BUF);
601
602 /*
603 * Initialize the header.
604 */
605 hdr = bp->b_addr;
606 if (xfs_sb_version_hascrc(&mp->m_sb)) {
607 struct xfs_dir3_blk_hdr *hdr3 = bp->b_addr;
608
609 memset(hdr3, 0, sizeof(*hdr3));
610 hdr3->magic = cpu_to_be32(XFS_DIR3_DATA_MAGIC);
611 hdr3->blkno = cpu_to_be64(bp->b_bn);
612 hdr3->owner = cpu_to_be64(dp->i_ino);
613 uuid_copy(&hdr3->uuid, &mp->m_sb.sb_meta_uuid);
614
615 } else
616 hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
617
618 bf = dp->d_ops->data_bestfree_p(hdr);
619 bf[0].offset = cpu_to_be16(dp->d_ops->data_entry_offset);
620 for (i = 1; i < XFS_DIR2_DATA_FD_COUNT; i++) {
621 bf[i].length = 0;
622 bf[i].offset = 0;
623 }
624
625 /*
626 * Set up an unused entry for the block's body.
627 */
628 dup = dp->d_ops->data_unused_p(hdr);
629 dup->freetag = cpu_to_be16(XFS_DIR2_DATA_FREE_TAG);
630
631 t = args->geo->blksize - (uint)dp->d_ops->data_entry_offset;
632 bf[0].length = cpu_to_be16(t);
633 dup->length = cpu_to_be16(t);
634 *xfs_dir2_data_unused_tag_p(dup) = cpu_to_be16((char *)dup - (char *)hdr);
635 /*
636 * Log it and return it.
637 */
638 xfs_dir2_data_log_header(args, bp);
639 xfs_dir2_data_log_unused(args, bp, dup);
640 *bpp = bp;
641 return 0;
642}
643
644/*
645 * Log an active data entry from the block.
646 */
647void
648xfs_dir2_data_log_entry(
649 struct xfs_da_args *args,
650 struct xfs_buf *bp,
651 xfs_dir2_data_entry_t *dep) /* data entry pointer */
652{
653 struct xfs_dir2_data_hdr *hdr = bp->b_addr;
654
655 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
656 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
657 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
658 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
659
660 xfs_trans_log_buf(args->trans, bp, (uint)((char *)dep - (char *)hdr),
661 (uint)((char *)(args->dp->d_ops->data_entry_tag_p(dep) + 1) -
662 (char *)hdr - 1));
663}
664
665/*
666 * Log a data block header.
667 */
668void
669xfs_dir2_data_log_header(
670 struct xfs_da_args *args,
671 struct xfs_buf *bp)
672{
673#ifdef DEBUG
674 struct xfs_dir2_data_hdr *hdr = bp->b_addr;
675
676 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
677 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
678 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
679 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
680#endif
681
682 xfs_trans_log_buf(args->trans, bp, 0,
683 args->dp->d_ops->data_entry_offset - 1);
684}
685
686/*
687 * Log a data unused entry.
688 */
689void
690xfs_dir2_data_log_unused(
691 struct xfs_da_args *args,
692 struct xfs_buf *bp,
693 xfs_dir2_data_unused_t *dup) /* data unused pointer */
694{
695 xfs_dir2_data_hdr_t *hdr = bp->b_addr;
696
697 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
698 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
699 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
700 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
701
702 /*
703 * Log the first part of the unused entry.
704 */
705 xfs_trans_log_buf(args->trans, bp, (uint)((char *)dup - (char *)hdr),
706 (uint)((char *)&dup->length + sizeof(dup->length) -
707 1 - (char *)hdr));
708 /*
709 * Log the end (tag) of the unused entry.
710 */
711 xfs_trans_log_buf(args->trans, bp,
712 (uint)((char *)xfs_dir2_data_unused_tag_p(dup) - (char *)hdr),
713 (uint)((char *)xfs_dir2_data_unused_tag_p(dup) - (char *)hdr +
714 sizeof(xfs_dir2_data_off_t) - 1));
715}
716
717/*
718 * Make a byte range in the data block unused.
719 * Its current contents are unimportant.
720 */
721void
722xfs_dir2_data_make_free(
723 struct xfs_da_args *args,
724 struct xfs_buf *bp,
725 xfs_dir2_data_aoff_t offset, /* starting byte offset */
726 xfs_dir2_data_aoff_t len, /* length in bytes */
727 int *needlogp, /* out: log header */
728 int *needscanp) /* out: regen bestfree */
729{
730 xfs_dir2_data_hdr_t *hdr; /* data block pointer */
731 xfs_dir2_data_free_t *dfp; /* bestfree pointer */
732 char *endptr; /* end of data area */
733 int needscan; /* need to regen bestfree */
734 xfs_dir2_data_unused_t *newdup; /* new unused entry */
735 xfs_dir2_data_unused_t *postdup; /* unused entry after us */
736 xfs_dir2_data_unused_t *prevdup; /* unused entry before us */
737 struct xfs_dir2_data_free *bf;
738
739 hdr = bp->b_addr;
740
741 /*
742 * Figure out where the end of the data area is.
743 */
744 if (hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
745 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC))
746 endptr = (char *)hdr + args->geo->blksize;
747 else {
748 xfs_dir2_block_tail_t *btp; /* block tail */
749
750 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
751 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
752 btp = xfs_dir2_block_tail_p(args->geo, hdr);
753 endptr = (char *)xfs_dir2_block_leaf_p(btp);
754 }
755 /*
756 * If this isn't the start of the block, then back up to
757 * the previous entry and see if it's free.
758 */
759 if (offset > args->dp->d_ops->data_entry_offset) {
760 __be16 *tagp; /* tag just before us */
761
762 tagp = (__be16 *)((char *)hdr + offset) - 1;
763 prevdup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
764 if (be16_to_cpu(prevdup->freetag) != XFS_DIR2_DATA_FREE_TAG)
765 prevdup = NULL;
766 } else
767 prevdup = NULL;
768 /*
769 * If this isn't the end of the block, see if the entry after
770 * us is free.
771 */
772 if ((char *)hdr + offset + len < endptr) {
773 postdup =
774 (xfs_dir2_data_unused_t *)((char *)hdr + offset + len);
775 if (be16_to_cpu(postdup->freetag) != XFS_DIR2_DATA_FREE_TAG)
776 postdup = NULL;
777 } else
778 postdup = NULL;
779 ASSERT(*needscanp == 0);
780 needscan = 0;
781 /*
782 * Previous and following entries are both free,
783 * merge everything into a single free entry.
784 */
785 bf = args->dp->d_ops->data_bestfree_p(hdr);
786 if (prevdup && postdup) {
787 xfs_dir2_data_free_t *dfp2; /* another bestfree pointer */
788
789 /*
790 * See if prevdup and/or postdup are in bestfree table.
791 */
792 dfp = xfs_dir2_data_freefind(hdr, bf, prevdup);
793 dfp2 = xfs_dir2_data_freefind(hdr, bf, postdup);
794 /*
795 * We need a rescan unless there are exactly 2 free entries
796 * namely our two. Then we know what's happening, otherwise
797 * since the third bestfree is there, there might be more
798 * entries.
799 */
800 needscan = (bf[2].length != 0);
801 /*
802 * Fix up the new big freespace.
803 */
804 be16_add_cpu(&prevdup->length, len + be16_to_cpu(postdup->length));
805 *xfs_dir2_data_unused_tag_p(prevdup) =
806 cpu_to_be16((char *)prevdup - (char *)hdr);
807 xfs_dir2_data_log_unused(args, bp, prevdup);
808 if (!needscan) {
809 /*
810 * Has to be the case that entries 0 and 1 are
811 * dfp and dfp2 (don't know which is which), and
812 * entry 2 is empty.
813 * Remove entry 1 first then entry 0.
814 */
815 ASSERT(dfp && dfp2);
816 if (dfp == &bf[1]) {
817 dfp = &bf[0];
818 ASSERT(dfp2 == dfp);
819 dfp2 = &bf[1];
820 }
821 xfs_dir2_data_freeremove(hdr, bf, dfp2, needlogp);
822 xfs_dir2_data_freeremove(hdr, bf, dfp, needlogp);
823 /*
824 * Now insert the new entry.
825 */
826 dfp = xfs_dir2_data_freeinsert(hdr, bf, prevdup,
827 needlogp);
828 ASSERT(dfp == &bf[0]);
829 ASSERT(dfp->length == prevdup->length);
830 ASSERT(!dfp[1].length);
831 ASSERT(!dfp[2].length);
832 }
833 }
834 /*
835 * The entry before us is free, merge with it.
836 */
837 else if (prevdup) {
838 dfp = xfs_dir2_data_freefind(hdr, bf, prevdup);
839 be16_add_cpu(&prevdup->length, len);
840 *xfs_dir2_data_unused_tag_p(prevdup) =
841 cpu_to_be16((char *)prevdup - (char *)hdr);
842 xfs_dir2_data_log_unused(args, bp, prevdup);
843 /*
844 * If the previous entry was in the table, the new entry
845 * is longer, so it will be in the table too. Remove
846 * the old one and add the new one.
847 */
848 if (dfp) {
849 xfs_dir2_data_freeremove(hdr, bf, dfp, needlogp);
850 xfs_dir2_data_freeinsert(hdr, bf, prevdup, needlogp);
851 }
852 /*
853 * Otherwise we need a scan if the new entry is big enough.
854 */
855 else {
856 needscan = be16_to_cpu(prevdup->length) >
857 be16_to_cpu(bf[2].length);
858 }
859 }
860 /*
861 * The following entry is free, merge with it.
862 */
863 else if (postdup) {
864 dfp = xfs_dir2_data_freefind(hdr, bf, postdup);
865 newdup = (xfs_dir2_data_unused_t *)((char *)hdr + offset);
866 newdup->freetag = cpu_to_be16(XFS_DIR2_DATA_FREE_TAG);
867 newdup->length = cpu_to_be16(len + be16_to_cpu(postdup->length));
868 *xfs_dir2_data_unused_tag_p(newdup) =
869 cpu_to_be16((char *)newdup - (char *)hdr);
870 xfs_dir2_data_log_unused(args, bp, newdup);
871 /*
872 * If the following entry was in the table, the new entry
873 * is longer, so it will be in the table too. Remove
874 * the old one and add the new one.
875 */
876 if (dfp) {
877 xfs_dir2_data_freeremove(hdr, bf, dfp, needlogp);
878 xfs_dir2_data_freeinsert(hdr, bf, newdup, needlogp);
879 }
880 /*
881 * Otherwise we need a scan if the new entry is big enough.
882 */
883 else {
884 needscan = be16_to_cpu(newdup->length) >
885 be16_to_cpu(bf[2].length);
886 }
887 }
888 /*
889 * Neither neighbor is free. Make a new entry.
890 */
891 else {
892 newdup = (xfs_dir2_data_unused_t *)((char *)hdr + offset);
893 newdup->freetag = cpu_to_be16(XFS_DIR2_DATA_FREE_TAG);
894 newdup->length = cpu_to_be16(len);
895 *xfs_dir2_data_unused_tag_p(newdup) =
896 cpu_to_be16((char *)newdup - (char *)hdr);
897 xfs_dir2_data_log_unused(args, bp, newdup);
898 xfs_dir2_data_freeinsert(hdr, bf, newdup, needlogp);
899 }
900 *needscanp = needscan;
901}
902
903/*
904 * Take a byte range out of an existing unused space and make it un-free.
905 */
906void
907xfs_dir2_data_use_free(
908 struct xfs_da_args *args,
909 struct xfs_buf *bp,
910 xfs_dir2_data_unused_t *dup, /* unused entry */
911 xfs_dir2_data_aoff_t offset, /* starting offset to use */
912 xfs_dir2_data_aoff_t len, /* length to use */
913 int *needlogp, /* out: need to log header */
914 int *needscanp) /* out: need regen bestfree */
915{
916 xfs_dir2_data_hdr_t *hdr; /* data block header */
917 xfs_dir2_data_free_t *dfp; /* bestfree pointer */
918 int matchback; /* matches end of freespace */
919 int matchfront; /* matches start of freespace */
920 int needscan; /* need to regen bestfree */
921 xfs_dir2_data_unused_t *newdup; /* new unused entry */
922 xfs_dir2_data_unused_t *newdup2; /* another new unused entry */
923 int oldlen; /* old unused entry's length */
924 struct xfs_dir2_data_free *bf;
925
926 hdr = bp->b_addr;
927 ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC) ||
928 hdr->magic == cpu_to_be32(XFS_DIR3_DATA_MAGIC) ||
929 hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) ||
930 hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC));
931 ASSERT(be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG);
932 ASSERT(offset >= (char *)dup - (char *)hdr);
933 ASSERT(offset + len <= (char *)dup + be16_to_cpu(dup->length) - (char *)hdr);
934 ASSERT((char *)dup - (char *)hdr == be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)));
935 /*
936 * Look up the entry in the bestfree table.
937 */
938 oldlen = be16_to_cpu(dup->length);
939 bf = args->dp->d_ops->data_bestfree_p(hdr);
940 dfp = xfs_dir2_data_freefind(hdr, bf, dup);
941 ASSERT(dfp || oldlen <= be16_to_cpu(bf[2].length));
942 /*
943 * Check for alignment with front and back of the entry.
944 */
945 matchfront = (char *)dup - (char *)hdr == offset;
946 matchback = (char *)dup + oldlen - (char *)hdr == offset + len;
947 ASSERT(*needscanp == 0);
948 needscan = 0;
949 /*
950 * If we matched it exactly we just need to get rid of it from
951 * the bestfree table.
952 */
953 if (matchfront && matchback) {
954 if (dfp) {
955 needscan = (bf[2].offset != 0);
956 if (!needscan)
957 xfs_dir2_data_freeremove(hdr, bf, dfp,
958 needlogp);
959 }
960 }
961 /*
962 * We match the first part of the entry.
963 * Make a new entry with the remaining freespace.
964 */
965 else if (matchfront) {
966 newdup = (xfs_dir2_data_unused_t *)((char *)hdr + offset + len);
967 newdup->freetag = cpu_to_be16(XFS_DIR2_DATA_FREE_TAG);
968 newdup->length = cpu_to_be16(oldlen - len);
969 *xfs_dir2_data_unused_tag_p(newdup) =
970 cpu_to_be16((char *)newdup - (char *)hdr);
971 xfs_dir2_data_log_unused(args, bp, newdup);
972 /*
973 * If it was in the table, remove it and add the new one.
974 */
975 if (dfp) {
976 xfs_dir2_data_freeremove(hdr, bf, dfp, needlogp);
977 dfp = xfs_dir2_data_freeinsert(hdr, bf, newdup,
978 needlogp);
979 ASSERT(dfp != NULL);
980 ASSERT(dfp->length == newdup->length);
981 ASSERT(be16_to_cpu(dfp->offset) == (char *)newdup - (char *)hdr);
982 /*
983 * If we got inserted at the last slot,
984 * that means we don't know if there was a better
985 * choice for the last slot, or not. Rescan.
986 */
987 needscan = dfp == &bf[2];
988 }
989 }
990 /*
991 * We match the last part of the entry.
992 * Trim the allocated space off the tail of the entry.
993 */
994 else if (matchback) {
995 newdup = dup;
996 newdup->length = cpu_to_be16(((char *)hdr + offset) - (char *)newdup);
997 *xfs_dir2_data_unused_tag_p(newdup) =
998 cpu_to_be16((char *)newdup - (char *)hdr);
999 xfs_dir2_data_log_unused(args, bp, newdup);
1000 /*
1001 * If it was in the table, remove it and add the new one.
1002 */
1003 if (dfp) {
1004 xfs_dir2_data_freeremove(hdr, bf, dfp, needlogp);
1005 dfp = xfs_dir2_data_freeinsert(hdr, bf, newdup,
1006 needlogp);
1007 ASSERT(dfp != NULL);
1008 ASSERT(dfp->length == newdup->length);
1009 ASSERT(be16_to_cpu(dfp->offset) == (char *)newdup - (char *)hdr);
1010 /*
1011 * If we got inserted at the last slot,
1012 * that means we don't know if there was a better
1013 * choice for the last slot, or not. Rescan.
1014 */
1015 needscan = dfp == &bf[2];
1016 }
1017 }
1018 /*
1019 * Poking out the middle of an entry.
1020 * Make two new entries.
1021 */
1022 else {
1023 newdup = dup;
1024 newdup->length = cpu_to_be16(((char *)hdr + offset) - (char *)newdup);
1025 *xfs_dir2_data_unused_tag_p(newdup) =
1026 cpu_to_be16((char *)newdup - (char *)hdr);
1027 xfs_dir2_data_log_unused(args, bp, newdup);
1028 newdup2 = (xfs_dir2_data_unused_t *)((char *)hdr + offset + len);
1029 newdup2->freetag = cpu_to_be16(XFS_DIR2_DATA_FREE_TAG);
1030 newdup2->length = cpu_to_be16(oldlen - len - be16_to_cpu(newdup->length));
1031 *xfs_dir2_data_unused_tag_p(newdup2) =
1032 cpu_to_be16((char *)newdup2 - (char *)hdr);
1033 xfs_dir2_data_log_unused(args, bp, newdup2);
1034 /*
1035 * If the old entry was in the table, we need to scan
1036 * if the 3rd entry was valid, since these entries
1037 * are smaller than the old one.
1038 * If we don't need to scan that means there were 1 or 2
1039 * entries in the table, and removing the old and adding
1040 * the 2 new will work.
1041 */
1042 if (dfp) {
1043 needscan = (bf[2].length != 0);
1044 if (!needscan) {
1045 xfs_dir2_data_freeremove(hdr, bf, dfp,
1046 needlogp);
1047 xfs_dir2_data_freeinsert(hdr, bf, newdup,
1048 needlogp);
1049 xfs_dir2_data_freeinsert(hdr, bf, newdup2,
1050 needlogp);
1051 }
1052 }
1053 }
1054 *needscanp = needscan;
1055}