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
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"
29 #include "xfs_inode_item.h"
32 #include "xfs_dir2_priv.h"
33 #include "xfs_error.h"
34 #include "xfs_trace.h"
36 struct xfs_name xfs_name_dotdot
= { (unsigned char *)"..", 2, XFS_DIR3_FT_DIR
};
39 * @mode, if set, indicates that the type field needs to be set up.
40 * This uses the transformation from file mode to DT_* as defined in linux/fs.h
41 * for file type specification. This will be propagated into the directory
42 * structure if appropriate for the given operation and filesystem config.
44 const unsigned char xfs_mode_to_ftype
[S_IFMT
>> S_SHIFT
] = {
45 [0] = XFS_DIR3_FT_UNKNOWN
,
46 [S_IFREG
>> S_SHIFT
] = XFS_DIR3_FT_REG_FILE
,
47 [S_IFDIR
>> S_SHIFT
] = XFS_DIR3_FT_DIR
,
48 [S_IFCHR
>> S_SHIFT
] = XFS_DIR3_FT_CHRDEV
,
49 [S_IFBLK
>> S_SHIFT
] = XFS_DIR3_FT_BLKDEV
,
50 [S_IFIFO
>> S_SHIFT
] = XFS_DIR3_FT_FIFO
,
51 [S_IFSOCK
>> S_SHIFT
] = XFS_DIR3_FT_SOCK
,
52 [S_IFLNK
>> S_SHIFT
] = XFS_DIR3_FT_SYMLINK
,
56 * ASCII case-insensitive (ie. A-Z) support for directories that was
60 xfs_ascii_ci_hashname(
61 struct xfs_name
*name
)
66 for (i
= 0, hash
= 0; i
< name
->len
; i
++)
67 hash
= tolower(name
->name
[i
]) ^ rol32(hash
, 7);
73 xfs_ascii_ci_compname(
74 struct xfs_da_args
*args
,
75 const unsigned char *name
,
78 enum xfs_dacmp result
;
81 if (args
->namelen
!= len
)
82 return XFS_CMP_DIFFERENT
;
84 result
= XFS_CMP_EXACT
;
85 for (i
= 0; i
< len
; i
++) {
86 if (args
->name
[i
] == name
[i
])
88 if (tolower(args
->name
[i
]) != tolower(name
[i
]))
89 return XFS_CMP_DIFFERENT
;
90 result
= XFS_CMP_CASE
;
96 static struct xfs_nameops xfs_ascii_ci_nameops
= {
97 .hashname
= xfs_ascii_ci_hashname
,
98 .compname
= xfs_ascii_ci_compname
,
103 struct xfs_mount
*mp
)
105 struct xfs_da_geometry
*dageo
;
109 ASSERT(mp
->m_sb
.sb_versionnum
& XFS_SB_VERSION_DIRV2BIT
);
110 ASSERT((1 << (mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
)) <=
113 mp
->m_dir_inode_ops
= xfs_dir_get_ops(mp
, NULL
);
114 mp
->m_nondir_inode_ops
= xfs_nondir_get_ops(mp
, NULL
);
116 nodehdr_size
= mp
->m_dir_inode_ops
->node_hdr_size
;
117 mp
->m_dir_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
118 KM_SLEEP
| KM_MAYFAIL
);
119 mp
->m_attr_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
120 KM_SLEEP
| KM_MAYFAIL
);
121 if (!mp
->m_dir_geo
|| !mp
->m_attr_geo
) {
122 kmem_free(mp
->m_dir_geo
);
123 kmem_free(mp
->m_attr_geo
);
127 /* set up directory geometry */
128 dageo
= mp
->m_dir_geo
;
129 dageo
->blklog
= mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
;
130 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
131 dageo
->blksize
= 1 << dageo
->blklog
;
132 dageo
->fsbcount
= 1 << mp
->m_sb
.sb_dirblklog
;
135 * Now we've set up the block conversion variables, we can calculate the
136 * segment block constants using the geometry structure.
138 dageo
->datablk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_DATA_OFFSET
);
139 dageo
->leafblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_LEAF_OFFSET
);
140 dageo
->freeblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_FREE_OFFSET
);
141 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
142 (uint
)sizeof(xfs_da_node_entry_t
);
143 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
145 /* set up attribute geometry - single fsb only */
146 dageo
= mp
->m_attr_geo
;
147 dageo
->blklog
= mp
->m_sb
.sb_blocklog
;
148 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
149 dageo
->blksize
= 1 << dageo
->blklog
;
151 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
152 (uint
)sizeof(xfs_da_node_entry_t
);
153 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
155 if (xfs_sb_version_hasasciici(&mp
->m_sb
))
156 mp
->m_dirnameops
= &xfs_ascii_ci_nameops
;
158 mp
->m_dirnameops
= &xfs_default_nameops
;
165 struct xfs_mount
*mp
)
167 kmem_free(mp
->m_dir_geo
);
168 kmem_free(mp
->m_attr_geo
);
172 * Return 1 if directory contains only "." and "..".
178 xfs_dir2_sf_hdr_t
*sfp
;
180 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
181 if (dp
->i_d
.di_size
== 0) /* might happen during shutdown. */
183 if (dp
->i_d
.di_size
> XFS_IFORK_DSIZE(dp
))
185 sfp
= (xfs_dir2_sf_hdr_t
*)dp
->i_df
.if_u1
.if_data
;
190 * Validate a given inode number.
193 xfs_dir_ino_validate(
197 xfs_agblock_t agblkno
;
203 agno
= XFS_INO_TO_AGNO(mp
, ino
);
204 agblkno
= XFS_INO_TO_AGBNO(mp
, ino
);
205 ioff
= XFS_INO_TO_OFFSET(mp
, ino
);
206 agino
= XFS_OFFBNO_TO_AGINO(mp
, agblkno
, ioff
);
208 agno
< mp
->m_sb
.sb_agcount
&&
209 agblkno
< mp
->m_sb
.sb_agblocks
&&
211 ioff
< (1 << mp
->m_sb
.sb_inopblog
) &&
212 XFS_AGINO_TO_INO(mp
, agno
, agino
) == ino
;
213 if (unlikely(XFS_TEST_ERROR(!ino_ok
, mp
, XFS_ERRTAG_DIR_INO_VALIDATE
,
214 XFS_RANDOM_DIR_INO_VALIDATE
))) {
215 xfs_warn(mp
, "Invalid inode number 0x%Lx",
216 (unsigned long long) ino
);
217 XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW
, mp
);
218 return -EFSCORRUPTED
;
224 * Initialize a directory with its "." and ".." entries.
232 struct xfs_da_args
*args
;
235 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
236 error
= xfs_dir_ino_validate(tp
->t_mountp
, pdp
->i_ino
);
240 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
244 args
->geo
= dp
->i_mount
->m_dir_geo
;
247 error
= xfs_dir2_sf_create(args
, pdp
->i_ino
);
253 * Enter a name in a directory, or check for available space.
254 * If inum is 0, only the available space test is performed.
260 struct xfs_name
*name
,
261 xfs_ino_t inum
, /* new entry inode number */
262 xfs_fsblock_t
*first
, /* bmap's firstblock */
263 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
264 xfs_extlen_t total
) /* bmap's total block count */
266 struct xfs_da_args
*args
;
268 int v
; /* type-checking value */
270 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
272 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
275 XFS_STATS_INC(dp
->i_mount
, xs_dir_create
);
278 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
282 args
->geo
= dp
->i_mount
->m_dir_geo
;
283 args
->name
= name
->name
;
284 args
->namelen
= name
->len
;
285 args
->filetype
= name
->type
;
286 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
287 args
->inumber
= inum
;
289 args
->firstblock
= first
;
292 args
->whichfork
= XFS_DATA_FORK
;
294 args
->op_flags
= XFS_DA_OP_ADDNAME
| XFS_DA_OP_OKNOENT
;
296 args
->op_flags
|= XFS_DA_OP_JUSTCHECK
;
298 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
299 rval
= xfs_dir2_sf_addname(args
);
303 rval
= xfs_dir2_isblock(args
, &v
);
307 rval
= xfs_dir2_block_addname(args
);
311 rval
= xfs_dir2_isleaf(args
, &v
);
315 rval
= xfs_dir2_leaf_addname(args
);
317 rval
= xfs_dir2_node_addname(args
);
325 * If doing a CI lookup and case-insensitive match, dup actual name into
326 * args.value. Return EEXIST for success (ie. name found) or an error.
329 xfs_dir_cilookup_result(
330 struct xfs_da_args
*args
,
331 const unsigned char *name
,
334 if (args
->cmpresult
== XFS_CMP_DIFFERENT
)
336 if (args
->cmpresult
!= XFS_CMP_CASE
||
337 !(args
->op_flags
& XFS_DA_OP_CILOOKUP
))
340 args
->value
= kmem_alloc(len
, KM_NOFS
| KM_MAYFAIL
);
344 memcpy(args
->value
, name
, len
);
345 args
->valuelen
= len
;
350 * Lookup a name in a directory, give back the inode number.
351 * If ci_name is not NULL, returns the actual name in ci_name if it differs
352 * to name, or ci_name->name is set to NULL for an exact match.
359 struct xfs_name
*name
,
360 xfs_ino_t
*inum
, /* out: inode number */
361 struct xfs_name
*ci_name
) /* out: actual name if CI match */
363 struct xfs_da_args
*args
;
365 int v
; /* type-checking value */
368 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
369 XFS_STATS_INC(dp
->i_mount
, xs_dir_lookup
);
372 * We need to use KM_NOFS here so that lockdep will not throw false
373 * positive deadlock warnings on a non-transactional lookup path. It is
374 * safe to recurse into inode recalim in that case, but lockdep can't
375 * easily be taught about it. Hence KM_NOFS avoids having to add more
376 * lockdep Doing this avoids having to add a bunch of lockdep class
377 * annotations into the reclaim path for the ilock.
379 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
380 args
->geo
= dp
->i_mount
->m_dir_geo
;
381 args
->name
= name
->name
;
382 args
->namelen
= name
->len
;
383 args
->filetype
= name
->type
;
384 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
386 args
->whichfork
= XFS_DATA_FORK
;
388 args
->op_flags
= XFS_DA_OP_OKNOENT
;
390 args
->op_flags
|= XFS_DA_OP_CILOOKUP
;
392 lock_mode
= xfs_ilock_data_map_shared(dp
);
393 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
394 rval
= xfs_dir2_sf_lookup(args
);
398 rval
= xfs_dir2_isblock(args
, &v
);
402 rval
= xfs_dir2_block_lookup(args
);
406 rval
= xfs_dir2_isleaf(args
, &v
);
410 rval
= xfs_dir2_leaf_lookup(args
);
412 rval
= xfs_dir2_node_lookup(args
);
418 *inum
= args
->inumber
;
420 ci_name
->name
= args
->value
;
421 ci_name
->len
= args
->valuelen
;
425 xfs_iunlock(dp
, lock_mode
);
431 * Remove an entry from a directory.
437 struct xfs_name
*name
,
439 xfs_fsblock_t
*first
, /* bmap's firstblock */
440 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
441 xfs_extlen_t total
) /* bmap's total block count */
443 struct xfs_da_args
*args
;
445 int v
; /* type-checking value */
447 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
448 XFS_STATS_INC(dp
->i_mount
, xs_dir_remove
);
450 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
454 args
->geo
= dp
->i_mount
->m_dir_geo
;
455 args
->name
= name
->name
;
456 args
->namelen
= name
->len
;
457 args
->filetype
= name
->type
;
458 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
461 args
->firstblock
= first
;
464 args
->whichfork
= XFS_DATA_FORK
;
467 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
468 rval
= xfs_dir2_sf_removename(args
);
472 rval
= xfs_dir2_isblock(args
, &v
);
476 rval
= xfs_dir2_block_removename(args
);
480 rval
= xfs_dir2_isleaf(args
, &v
);
484 rval
= xfs_dir2_leaf_removename(args
);
486 rval
= xfs_dir2_node_removename(args
);
493 * Replace the inode number of a directory entry.
499 struct xfs_name
*name
, /* name of entry to replace */
500 xfs_ino_t inum
, /* new inode number */
501 xfs_fsblock_t
*first
, /* bmap's firstblock */
502 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
503 xfs_extlen_t total
) /* bmap's total block count */
505 struct xfs_da_args
*args
;
507 int v
; /* type-checking value */
509 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
511 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
515 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
519 args
->geo
= dp
->i_mount
->m_dir_geo
;
520 args
->name
= name
->name
;
521 args
->namelen
= name
->len
;
522 args
->filetype
= name
->type
;
523 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
524 args
->inumber
= inum
;
526 args
->firstblock
= first
;
529 args
->whichfork
= XFS_DATA_FORK
;
532 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
533 rval
= xfs_dir2_sf_replace(args
);
537 rval
= xfs_dir2_isblock(args
, &v
);
541 rval
= xfs_dir2_block_replace(args
);
545 rval
= xfs_dir2_isleaf(args
, &v
);
549 rval
= xfs_dir2_leaf_replace(args
);
551 rval
= xfs_dir2_node_replace(args
);
558 * See if this entry can be added to the directory without allocating space.
564 struct xfs_name
*name
) /* name of entry to add */
566 return xfs_dir_createname(tp
, dp
, name
, 0, NULL
, NULL
, 0);
574 * Add a block to the directory.
576 * This routine is for data and free blocks, not leaf/node blocks which are
577 * handled by xfs_da_grow_inode.
581 struct xfs_da_args
*args
,
582 int space
, /* v2 dir's space XFS_DIR2_xxx_SPACE */
583 xfs_dir2_db_t
*dbp
) /* out: block number added */
585 struct xfs_inode
*dp
= args
->dp
;
586 struct xfs_mount
*mp
= dp
->i_mount
;
587 xfs_fileoff_t bno
; /* directory offset of new block */
588 int count
; /* count of filesystem blocks */
591 trace_xfs_dir2_grow_inode(args
, space
);
594 * Set lowest possible block in the space requested.
596 bno
= XFS_B_TO_FSBT(mp
, space
* XFS_DIR2_SPACE_SIZE
);
597 count
= args
->geo
->fsbcount
;
599 error
= xfs_da_grow_inode_int(args
, &bno
, count
);
603 *dbp
= xfs_dir2_da_to_db(args
->geo
, (xfs_dablk_t
)bno
);
606 * Update file's size if this is the data space and it grew.
608 if (space
== XFS_DIR2_DATA_SPACE
) {
609 xfs_fsize_t size
; /* directory file (data) size */
611 size
= XFS_FSB_TO_B(mp
, bno
+ count
);
612 if (size
> dp
->i_d
.di_size
) {
613 dp
->i_d
.di_size
= size
;
614 xfs_trans_log_inode(args
->trans
, dp
, XFS_ILOG_CORE
);
621 * See if the directory is a single-block form directory.
625 struct xfs_da_args
*args
,
626 int *vp
) /* out: 1 is block, 0 is not block */
628 xfs_fileoff_t last
; /* last file offset */
631 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
633 rval
= XFS_FSB_TO_B(args
->dp
->i_mount
, last
) == args
->geo
->blksize
;
634 ASSERT(rval
== 0 || args
->dp
->i_d
.di_size
== args
->geo
->blksize
);
640 * See if the directory is a single-leaf form directory.
644 struct xfs_da_args
*args
,
645 int *vp
) /* out: 1 is block, 0 is not block */
647 xfs_fileoff_t last
; /* last file offset */
650 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
652 *vp
= last
== args
->geo
->leafblk
+ args
->geo
->fsbcount
;
657 * Remove the given block from the directory.
658 * This routine is used for data and free blocks, leaf/node are done
659 * by xfs_da_shrink_inode.
662 xfs_dir2_shrink_inode(
667 xfs_fileoff_t bno
; /* directory file offset */
668 xfs_dablk_t da
; /* directory file offset */
669 int done
; /* bunmap is finished */
675 trace_xfs_dir2_shrink_inode(args
, db
);
680 da
= xfs_dir2_db_to_da(args
->geo
, db
);
682 /* Unmap the fsblock(s). */
683 error
= xfs_bunmapi(tp
, dp
, da
, args
->geo
->fsbcount
, 0, 0,
684 args
->firstblock
, args
->dfops
, &done
);
687 * ENOSPC actually can happen if we're in a removename with no
688 * space reservation, and the resulting block removal would
689 * cause a bmap btree split or conversion from extents to btree.
690 * This can only happen for un-fragmented directory blocks,
691 * since you need to be punching out the middle of an extent.
692 * In this case we need to leave the block in the file, and not
693 * binval it. So the block has to be in a consistent empty
694 * state and appropriately logged. We don't free up the buffer,
695 * the caller can tell it hasn't happened since it got an error
702 * Invalidate the buffer from the transaction.
704 xfs_trans_binval(tp
, bp
);
706 * If it's not a data block, we're done.
708 if (db
>= xfs_dir2_byte_to_db(args
->geo
, XFS_DIR2_LEAF_OFFSET
))
711 * If the block isn't the last one in the directory, we're done.
713 if (dp
->i_d
.di_size
> xfs_dir2_db_off_to_byte(args
->geo
, db
+ 1, 0))
716 if ((error
= xfs_bmap_last_before(tp
, dp
, &bno
, XFS_DATA_FORK
))) {
718 * This can't really happen unless there's kernel corruption.
722 if (db
== args
->geo
->datablk
)
727 * Set the size to the new last block.
729 dp
->i_d
.di_size
= XFS_FSB_TO_B(mp
, bno
);
730 xfs_trans_log_inode(tp
, dp
, XFS_ILOG_CORE
);