2 * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 #include "libxfs_priv.h"
20 #include "xfs_format.h"
21 #include "xfs_log_format.h"
22 #include "xfs_trans_resv.h"
23 #include "xfs_mount.h"
24 #include "xfs_defer.h"
25 #include "xfs_da_format.h"
26 #include "xfs_da_btree.h"
27 #include "xfs_inode.h"
28 #include "xfs_trans.h"
31 #include "xfs_dir2_priv.h"
32 #include "xfs_trace.h"
34 struct xfs_name xfs_name_dotdot
= { (unsigned char *)"..", 2, XFS_DIR3_FT_DIR
};
37 * @mode, if set, indicates that the type field needs to be set up.
38 * This uses the transformation from file mode to DT_* as defined in linux/fs.h
39 * for file type specification. This will be propagated into the directory
40 * structure if appropriate for the given operation and filesystem config.
42 const unsigned char xfs_mode_to_ftype
[S_IFMT
>> S_SHIFT
] = {
43 [0] = XFS_DIR3_FT_UNKNOWN
,
44 [S_IFREG
>> S_SHIFT
] = XFS_DIR3_FT_REG_FILE
,
45 [S_IFDIR
>> S_SHIFT
] = XFS_DIR3_FT_DIR
,
46 [S_IFCHR
>> S_SHIFT
] = XFS_DIR3_FT_CHRDEV
,
47 [S_IFBLK
>> S_SHIFT
] = XFS_DIR3_FT_BLKDEV
,
48 [S_IFIFO
>> S_SHIFT
] = XFS_DIR3_FT_FIFO
,
49 [S_IFSOCK
>> S_SHIFT
] = XFS_DIR3_FT_SOCK
,
50 [S_IFLNK
>> S_SHIFT
] = XFS_DIR3_FT_SYMLINK
,
54 * ASCII case-insensitive (ie. A-Z) support for directories that was
58 xfs_ascii_ci_hashname(
59 struct xfs_name
*name
)
64 for (i
= 0, hash
= 0; i
< name
->len
; i
++)
65 hash
= tolower(name
->name
[i
]) ^ rol32(hash
, 7);
71 xfs_ascii_ci_compname(
72 struct xfs_da_args
*args
,
73 const unsigned char *name
,
76 enum xfs_dacmp result
;
79 if (args
->namelen
!= len
)
80 return XFS_CMP_DIFFERENT
;
82 result
= XFS_CMP_EXACT
;
83 for (i
= 0; i
< len
; i
++) {
84 if (args
->name
[i
] == name
[i
])
86 if (tolower(args
->name
[i
]) != tolower(name
[i
]))
87 return XFS_CMP_DIFFERENT
;
88 result
= XFS_CMP_CASE
;
94 static const struct xfs_nameops xfs_ascii_ci_nameops
= {
95 .hashname
= xfs_ascii_ci_hashname
,
96 .compname
= xfs_ascii_ci_compname
,
101 struct xfs_mount
*mp
)
103 struct xfs_da_geometry
*dageo
;
107 ASSERT(mp
->m_sb
.sb_versionnum
& XFS_SB_VERSION_DIRV2BIT
);
108 ASSERT((1 << (mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
)) <=
111 mp
->m_dir_inode_ops
= xfs_dir_get_ops(mp
, NULL
);
112 mp
->m_nondir_inode_ops
= xfs_nondir_get_ops(mp
, NULL
);
114 nodehdr_size
= mp
->m_dir_inode_ops
->node_hdr_size
;
115 mp
->m_dir_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
116 KM_SLEEP
| KM_MAYFAIL
);
117 mp
->m_attr_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
118 KM_SLEEP
| KM_MAYFAIL
);
119 if (!mp
->m_dir_geo
|| !mp
->m_attr_geo
) {
120 kmem_free(mp
->m_dir_geo
);
121 kmem_free(mp
->m_attr_geo
);
125 /* set up directory geometry */
126 dageo
= mp
->m_dir_geo
;
127 dageo
->blklog
= mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
;
128 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
129 dageo
->blksize
= 1 << dageo
->blklog
;
130 dageo
->fsbcount
= 1 << mp
->m_sb
.sb_dirblklog
;
133 * Now we've set up the block conversion variables, we can calculate the
134 * segment block constants using the geometry structure.
136 dageo
->datablk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_DATA_OFFSET
);
137 dageo
->leafblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_LEAF_OFFSET
);
138 dageo
->freeblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_FREE_OFFSET
);
139 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
140 (uint
)sizeof(xfs_da_node_entry_t
);
141 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
143 /* set up attribute geometry - single fsb only */
144 dageo
= mp
->m_attr_geo
;
145 dageo
->blklog
= mp
->m_sb
.sb_blocklog
;
146 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
147 dageo
->blksize
= 1 << dageo
->blklog
;
149 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
150 (uint
)sizeof(xfs_da_node_entry_t
);
151 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
153 if (xfs_sb_version_hasasciici(&mp
->m_sb
))
154 mp
->m_dirnameops
= &xfs_ascii_ci_nameops
;
156 mp
->m_dirnameops
= &xfs_default_nameops
;
163 struct xfs_mount
*mp
)
165 kmem_free(mp
->m_dir_geo
);
166 kmem_free(mp
->m_attr_geo
);
170 * Return 1 if directory contains only "." and "..".
176 xfs_dir2_sf_hdr_t
*sfp
;
178 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
179 if (dp
->i_d
.di_size
== 0) /* might happen during shutdown. */
181 if (dp
->i_d
.di_size
> XFS_IFORK_DSIZE(dp
))
183 sfp
= (xfs_dir2_sf_hdr_t
*)dp
->i_df
.if_u1
.if_data
;
188 * Validate a given inode number.
191 xfs_dir_ino_validate(
195 xfs_agblock_t agblkno
;
201 agno
= XFS_INO_TO_AGNO(mp
, ino
);
202 agblkno
= XFS_INO_TO_AGBNO(mp
, ino
);
203 ioff
= XFS_INO_TO_OFFSET(mp
, ino
);
204 agino
= XFS_OFFBNO_TO_AGINO(mp
, agblkno
, ioff
);
206 agno
< mp
->m_sb
.sb_agcount
&&
207 agblkno
< mp
->m_sb
.sb_agblocks
&&
209 ioff
< (1 << mp
->m_sb
.sb_inopblog
) &&
210 XFS_AGINO_TO_INO(mp
, agno
, agino
) == ino
;
211 if (unlikely(XFS_TEST_ERROR(!ino_ok
, mp
, XFS_ERRTAG_DIR_INO_VALIDATE
,
212 XFS_RANDOM_DIR_INO_VALIDATE
))) {
213 xfs_warn(mp
, "Invalid inode number 0x%Lx",
214 (unsigned long long) ino
);
215 XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW
, mp
);
216 return -EFSCORRUPTED
;
222 * Initialize a directory with its "." and ".." entries.
230 struct xfs_da_args
*args
;
233 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
234 error
= xfs_dir_ino_validate(tp
->t_mountp
, pdp
->i_ino
);
238 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
242 args
->geo
= dp
->i_mount
->m_dir_geo
;
245 error
= xfs_dir2_sf_create(args
, pdp
->i_ino
);
251 * Enter a name in a directory, or check for available space.
252 * If inum is 0, only the available space test is performed.
258 struct xfs_name
*name
,
259 xfs_ino_t inum
, /* new entry inode number */
260 xfs_fsblock_t
*first
, /* bmap's firstblock */
261 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
262 xfs_extlen_t total
) /* bmap's total block count */
264 struct xfs_da_args
*args
;
266 int v
; /* type-checking value */
268 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
270 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
273 XFS_STATS_INC(dp
->i_mount
, xs_dir_create
);
276 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
280 args
->geo
= dp
->i_mount
->m_dir_geo
;
281 args
->name
= name
->name
;
282 args
->namelen
= name
->len
;
283 args
->filetype
= name
->type
;
284 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
285 args
->inumber
= inum
;
287 args
->firstblock
= first
;
290 args
->whichfork
= XFS_DATA_FORK
;
292 args
->op_flags
= XFS_DA_OP_ADDNAME
| XFS_DA_OP_OKNOENT
;
294 args
->op_flags
|= XFS_DA_OP_JUSTCHECK
;
296 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
297 rval
= xfs_dir2_sf_addname(args
);
301 rval
= xfs_dir2_isblock(args
, &v
);
305 rval
= xfs_dir2_block_addname(args
);
309 rval
= xfs_dir2_isleaf(args
, &v
);
313 rval
= xfs_dir2_leaf_addname(args
);
315 rval
= xfs_dir2_node_addname(args
);
323 * If doing a CI lookup and case-insensitive match, dup actual name into
324 * args.value. Return EEXIST for success (ie. name found) or an error.
327 xfs_dir_cilookup_result(
328 struct xfs_da_args
*args
,
329 const unsigned char *name
,
332 if (args
->cmpresult
== XFS_CMP_DIFFERENT
)
334 if (args
->cmpresult
!= XFS_CMP_CASE
||
335 !(args
->op_flags
& XFS_DA_OP_CILOOKUP
))
338 args
->value
= kmem_alloc(len
, KM_NOFS
| KM_MAYFAIL
);
342 memcpy(args
->value
, name
, len
);
343 args
->valuelen
= len
;
348 * Lookup a name in a directory, give back the inode number.
349 * If ci_name is not NULL, returns the actual name in ci_name if it differs
350 * to name, or ci_name->name is set to NULL for an exact match.
357 struct xfs_name
*name
,
358 xfs_ino_t
*inum
, /* out: inode number */
359 struct xfs_name
*ci_name
) /* out: actual name if CI match */
361 struct xfs_da_args
*args
;
363 int v
; /* type-checking value */
366 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
367 XFS_STATS_INC(dp
->i_mount
, xs_dir_lookup
);
370 * We need to use KM_NOFS here so that lockdep will not throw false
371 * positive deadlock warnings on a non-transactional lookup path. It is
372 * safe to recurse into inode recalim in that case, but lockdep can't
373 * easily be taught about it. Hence KM_NOFS avoids having to add more
374 * lockdep Doing this avoids having to add a bunch of lockdep class
375 * annotations into the reclaim path for the ilock.
377 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
378 args
->geo
= dp
->i_mount
->m_dir_geo
;
379 args
->name
= name
->name
;
380 args
->namelen
= name
->len
;
381 args
->filetype
= name
->type
;
382 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
384 args
->whichfork
= XFS_DATA_FORK
;
386 args
->op_flags
= XFS_DA_OP_OKNOENT
;
388 args
->op_flags
|= XFS_DA_OP_CILOOKUP
;
390 lock_mode
= xfs_ilock_data_map_shared(dp
);
391 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
392 rval
= xfs_dir2_sf_lookup(args
);
396 rval
= xfs_dir2_isblock(args
, &v
);
400 rval
= xfs_dir2_block_lookup(args
);
404 rval
= xfs_dir2_isleaf(args
, &v
);
408 rval
= xfs_dir2_leaf_lookup(args
);
410 rval
= xfs_dir2_node_lookup(args
);
416 *inum
= args
->inumber
;
418 ci_name
->name
= args
->value
;
419 ci_name
->len
= args
->valuelen
;
423 xfs_iunlock(dp
, lock_mode
);
429 * Remove an entry from a directory.
435 struct xfs_name
*name
,
437 xfs_fsblock_t
*first
, /* bmap's firstblock */
438 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
439 xfs_extlen_t total
) /* bmap's total block count */
441 struct xfs_da_args
*args
;
443 int v
; /* type-checking value */
445 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
446 XFS_STATS_INC(dp
->i_mount
, xs_dir_remove
);
448 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
452 args
->geo
= dp
->i_mount
->m_dir_geo
;
453 args
->name
= name
->name
;
454 args
->namelen
= name
->len
;
455 args
->filetype
= name
->type
;
456 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
459 args
->firstblock
= first
;
462 args
->whichfork
= XFS_DATA_FORK
;
465 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
466 rval
= xfs_dir2_sf_removename(args
);
470 rval
= xfs_dir2_isblock(args
, &v
);
474 rval
= xfs_dir2_block_removename(args
);
478 rval
= xfs_dir2_isleaf(args
, &v
);
482 rval
= xfs_dir2_leaf_removename(args
);
484 rval
= xfs_dir2_node_removename(args
);
491 * Replace the inode number of a directory entry.
497 struct xfs_name
*name
, /* name of entry to replace */
498 xfs_ino_t inum
, /* new inode number */
499 xfs_fsblock_t
*first
, /* bmap's firstblock */
500 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
501 xfs_extlen_t total
) /* bmap's total block count */
503 struct xfs_da_args
*args
;
505 int v
; /* type-checking value */
507 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
509 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
513 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
517 args
->geo
= dp
->i_mount
->m_dir_geo
;
518 args
->name
= name
->name
;
519 args
->namelen
= name
->len
;
520 args
->filetype
= name
->type
;
521 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
522 args
->inumber
= inum
;
524 args
->firstblock
= first
;
527 args
->whichfork
= XFS_DATA_FORK
;
530 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
531 rval
= xfs_dir2_sf_replace(args
);
535 rval
= xfs_dir2_isblock(args
, &v
);
539 rval
= xfs_dir2_block_replace(args
);
543 rval
= xfs_dir2_isleaf(args
, &v
);
547 rval
= xfs_dir2_leaf_replace(args
);
549 rval
= xfs_dir2_node_replace(args
);
556 * See if this entry can be added to the directory without allocating space.
562 struct xfs_name
*name
) /* name of entry to add */
564 return xfs_dir_createname(tp
, dp
, name
, 0, NULL
, NULL
, 0);
572 * Add a block to the directory.
574 * This routine is for data and free blocks, not leaf/node blocks which are
575 * handled by xfs_da_grow_inode.
579 struct xfs_da_args
*args
,
580 int space
, /* v2 dir's space XFS_DIR2_xxx_SPACE */
581 xfs_dir2_db_t
*dbp
) /* out: block number added */
583 struct xfs_inode
*dp
= args
->dp
;
584 struct xfs_mount
*mp
= dp
->i_mount
;
585 xfs_fileoff_t bno
; /* directory offset of new block */
586 int count
; /* count of filesystem blocks */
589 trace_xfs_dir2_grow_inode(args
, space
);
592 * Set lowest possible block in the space requested.
594 bno
= XFS_B_TO_FSBT(mp
, space
* XFS_DIR2_SPACE_SIZE
);
595 count
= args
->geo
->fsbcount
;
597 error
= xfs_da_grow_inode_int(args
, &bno
, count
);
601 *dbp
= xfs_dir2_da_to_db(args
->geo
, (xfs_dablk_t
)bno
);
604 * Update file's size if this is the data space and it grew.
606 if (space
== XFS_DIR2_DATA_SPACE
) {
607 xfs_fsize_t size
; /* directory file (data) size */
609 size
= XFS_FSB_TO_B(mp
, bno
+ count
);
610 if (size
> dp
->i_d
.di_size
) {
611 dp
->i_d
.di_size
= size
;
612 xfs_trans_log_inode(args
->trans
, dp
, XFS_ILOG_CORE
);
619 * See if the directory is a single-block form directory.
623 struct xfs_da_args
*args
,
624 int *vp
) /* out: 1 is block, 0 is not block */
626 xfs_fileoff_t last
; /* last file offset */
629 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
631 rval
= XFS_FSB_TO_B(args
->dp
->i_mount
, last
) == args
->geo
->blksize
;
632 ASSERT(rval
== 0 || args
->dp
->i_d
.di_size
== args
->geo
->blksize
);
638 * See if the directory is a single-leaf form directory.
642 struct xfs_da_args
*args
,
643 int *vp
) /* out: 1 is block, 0 is not block */
645 xfs_fileoff_t last
; /* last file offset */
648 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
650 *vp
= last
== args
->geo
->leafblk
+ args
->geo
->fsbcount
;
655 * Remove the given block from the directory.
656 * This routine is used for data and free blocks, leaf/node are done
657 * by xfs_da_shrink_inode.
660 xfs_dir2_shrink_inode(
665 xfs_fileoff_t bno
; /* directory file offset */
666 xfs_dablk_t da
; /* directory file offset */
667 int done
; /* bunmap is finished */
673 trace_xfs_dir2_shrink_inode(args
, db
);
678 da
= xfs_dir2_db_to_da(args
->geo
, db
);
680 /* Unmap the fsblock(s). */
681 error
= xfs_bunmapi(tp
, dp
, da
, args
->geo
->fsbcount
, 0, 0,
682 args
->firstblock
, args
->dfops
, &done
);
685 * ENOSPC actually can happen if we're in a removename with no
686 * space reservation, and the resulting block removal would
687 * cause a bmap btree split or conversion from extents to btree.
688 * This can only happen for un-fragmented directory blocks,
689 * since you need to be punching out the middle of an extent.
690 * In this case we need to leave the block in the file, and not
691 * binval it. So the block has to be in a consistent empty
692 * state and appropriately logged. We don't free up the buffer,
693 * the caller can tell it hasn't happened since it got an error
700 * Invalidate the buffer from the transaction.
702 xfs_trans_binval(tp
, bp
);
704 * If it's not a data block, we're done.
706 if (db
>= xfs_dir2_byte_to_db(args
->geo
, XFS_DIR2_LEAF_OFFSET
))
709 * If the block isn't the last one in the directory, we're done.
711 if (dp
->i_d
.di_size
> xfs_dir2_db_off_to_byte(args
->geo
, db
+ 1, 0))
714 if ((error
= xfs_bmap_last_before(tp
, dp
, &bno
, XFS_DATA_FORK
))) {
716 * This can't really happen unless there's kernel corruption.
720 if (db
== args
->geo
->datablk
)
725 * Set the size to the new last block.
727 dp
->i_d
.di_size
= XFS_FSB_TO_B(mp
, bno
);
728 xfs_trans_log_inode(tp
, dp
, XFS_ILOG_CORE
);