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 * Convert inode mode to directory entry filetype
39 unsigned char xfs_mode_to_ftype(int mode
)
41 switch (mode
& S_IFMT
) {
43 return XFS_DIR3_FT_REG_FILE
;
45 return XFS_DIR3_FT_DIR
;
47 return XFS_DIR3_FT_CHRDEV
;
49 return XFS_DIR3_FT_BLKDEV
;
51 return XFS_DIR3_FT_FIFO
;
53 return XFS_DIR3_FT_SOCK
;
55 return XFS_DIR3_FT_SYMLINK
;
57 return XFS_DIR3_FT_UNKNOWN
;
62 * ASCII case-insensitive (ie. A-Z) support for directories that was
66 xfs_ascii_ci_hashname(
67 struct xfs_name
*name
)
72 for (i
= 0, hash
= 0; i
< name
->len
; i
++)
73 hash
= tolower(name
->name
[i
]) ^ rol32(hash
, 7);
79 xfs_ascii_ci_compname(
80 struct xfs_da_args
*args
,
81 const unsigned char *name
,
84 enum xfs_dacmp result
;
87 if (args
->namelen
!= len
)
88 return XFS_CMP_DIFFERENT
;
90 result
= XFS_CMP_EXACT
;
91 for (i
= 0; i
< len
; i
++) {
92 if (args
->name
[i
] == name
[i
])
94 if (tolower(args
->name
[i
]) != tolower(name
[i
]))
95 return XFS_CMP_DIFFERENT
;
96 result
= XFS_CMP_CASE
;
102 static const struct xfs_nameops xfs_ascii_ci_nameops
= {
103 .hashname
= xfs_ascii_ci_hashname
,
104 .compname
= xfs_ascii_ci_compname
,
109 struct xfs_mount
*mp
)
111 struct xfs_da_geometry
*dageo
;
115 ASSERT(mp
->m_sb
.sb_versionnum
& XFS_SB_VERSION_DIRV2BIT
);
116 ASSERT((1 << (mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
)) <=
119 mp
->m_dir_inode_ops
= xfs_dir_get_ops(mp
, NULL
);
120 mp
->m_nondir_inode_ops
= xfs_nondir_get_ops(mp
, NULL
);
122 nodehdr_size
= mp
->m_dir_inode_ops
->node_hdr_size
;
123 mp
->m_dir_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
124 KM_SLEEP
| KM_MAYFAIL
);
125 mp
->m_attr_geo
= kmem_zalloc(sizeof(struct xfs_da_geometry
),
126 KM_SLEEP
| KM_MAYFAIL
);
127 if (!mp
->m_dir_geo
|| !mp
->m_attr_geo
) {
128 kmem_free(mp
->m_dir_geo
);
129 kmem_free(mp
->m_attr_geo
);
133 /* set up directory geometry */
134 dageo
= mp
->m_dir_geo
;
135 dageo
->blklog
= mp
->m_sb
.sb_blocklog
+ mp
->m_sb
.sb_dirblklog
;
136 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
137 dageo
->blksize
= 1 << dageo
->blklog
;
138 dageo
->fsbcount
= 1 << mp
->m_sb
.sb_dirblklog
;
141 * Now we've set up the block conversion variables, we can calculate the
142 * segment block constants using the geometry structure.
144 dageo
->datablk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_DATA_OFFSET
);
145 dageo
->leafblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_LEAF_OFFSET
);
146 dageo
->freeblk
= xfs_dir2_byte_to_da(dageo
, XFS_DIR2_FREE_OFFSET
);
147 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
148 (uint
)sizeof(xfs_da_node_entry_t
);
149 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
151 /* set up attribute geometry - single fsb only */
152 dageo
= mp
->m_attr_geo
;
153 dageo
->blklog
= mp
->m_sb
.sb_blocklog
;
154 dageo
->fsblog
= mp
->m_sb
.sb_blocklog
;
155 dageo
->blksize
= 1 << dageo
->blklog
;
157 dageo
->node_ents
= (dageo
->blksize
- nodehdr_size
) /
158 (uint
)sizeof(xfs_da_node_entry_t
);
159 dageo
->magicpct
= (dageo
->blksize
* 37) / 100;
161 if (xfs_sb_version_hasasciici(&mp
->m_sb
))
162 mp
->m_dirnameops
= &xfs_ascii_ci_nameops
;
164 mp
->m_dirnameops
= &xfs_default_nameops
;
171 struct xfs_mount
*mp
)
173 kmem_free(mp
->m_dir_geo
);
174 kmem_free(mp
->m_attr_geo
);
178 * Return 1 if directory contains only "." and "..".
184 xfs_dir2_sf_hdr_t
*sfp
;
186 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
187 if (dp
->i_d
.di_size
== 0) /* might happen during shutdown. */
189 if (dp
->i_d
.di_size
> XFS_IFORK_DSIZE(dp
))
191 sfp
= (xfs_dir2_sf_hdr_t
*)dp
->i_df
.if_u1
.if_data
;
196 * Validate a given inode number.
199 xfs_dir_ino_validate(
203 xfs_agblock_t agblkno
;
209 agno
= XFS_INO_TO_AGNO(mp
, ino
);
210 agblkno
= XFS_INO_TO_AGBNO(mp
, ino
);
211 ioff
= XFS_INO_TO_OFFSET(mp
, ino
);
212 agino
= XFS_OFFBNO_TO_AGINO(mp
, agblkno
, ioff
);
214 agno
< mp
->m_sb
.sb_agcount
&&
215 agblkno
< mp
->m_sb
.sb_agblocks
&&
217 ioff
< (1 << mp
->m_sb
.sb_inopblog
) &&
218 XFS_AGINO_TO_INO(mp
, agno
, agino
) == ino
;
219 if (unlikely(XFS_TEST_ERROR(!ino_ok
, mp
, XFS_ERRTAG_DIR_INO_VALIDATE
))) {
220 xfs_warn(mp
, "Invalid inode number 0x%Lx",
221 (unsigned long long) ino
);
222 XFS_ERROR_REPORT("xfs_dir_ino_validate", XFS_ERRLEVEL_LOW
, mp
);
223 return -EFSCORRUPTED
;
229 * Initialize a directory with its "." and ".." entries.
237 struct xfs_da_args
*args
;
240 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
241 error
= xfs_dir_ino_validate(tp
->t_mountp
, pdp
->i_ino
);
245 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
249 args
->geo
= dp
->i_mount
->m_dir_geo
;
252 error
= xfs_dir2_sf_create(args
, pdp
->i_ino
);
258 * Enter a name in a directory, or check for available space.
259 * If inum is 0, only the available space test is performed.
265 struct xfs_name
*name
,
266 xfs_ino_t inum
, /* new entry inode number */
267 xfs_fsblock_t
*first
, /* bmap's firstblock */
268 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
269 xfs_extlen_t total
) /* bmap's total block count */
271 struct xfs_da_args
*args
;
273 int v
; /* type-checking value */
275 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
277 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
280 XFS_STATS_INC(dp
->i_mount
, xs_dir_create
);
283 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
287 args
->geo
= dp
->i_mount
->m_dir_geo
;
288 args
->name
= name
->name
;
289 args
->namelen
= name
->len
;
290 args
->filetype
= name
->type
;
291 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
292 args
->inumber
= inum
;
294 args
->firstblock
= first
;
297 args
->whichfork
= XFS_DATA_FORK
;
299 args
->op_flags
= XFS_DA_OP_ADDNAME
| XFS_DA_OP_OKNOENT
;
301 args
->op_flags
|= XFS_DA_OP_JUSTCHECK
;
303 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
304 rval
= xfs_dir2_sf_addname(args
);
308 rval
= xfs_dir2_isblock(args
, &v
);
312 rval
= xfs_dir2_block_addname(args
);
316 rval
= xfs_dir2_isleaf(args
, &v
);
320 rval
= xfs_dir2_leaf_addname(args
);
322 rval
= xfs_dir2_node_addname(args
);
330 * If doing a CI lookup and case-insensitive match, dup actual name into
331 * args.value. Return EEXIST for success (ie. name found) or an error.
334 xfs_dir_cilookup_result(
335 struct xfs_da_args
*args
,
336 const unsigned char *name
,
339 if (args
->cmpresult
== XFS_CMP_DIFFERENT
)
341 if (args
->cmpresult
!= XFS_CMP_CASE
||
342 !(args
->op_flags
& XFS_DA_OP_CILOOKUP
))
345 args
->value
= kmem_alloc(len
, KM_NOFS
| KM_MAYFAIL
);
349 memcpy(args
->value
, name
, len
);
350 args
->valuelen
= len
;
355 * Lookup a name in a directory, give back the inode number.
356 * If ci_name is not NULL, returns the actual name in ci_name if it differs
357 * to name, or ci_name->name is set to NULL for an exact match.
364 struct xfs_name
*name
,
365 xfs_ino_t
*inum
, /* out: inode number */
366 struct xfs_name
*ci_name
) /* out: actual name if CI match */
368 struct xfs_da_args
*args
;
370 int v
; /* type-checking value */
373 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
374 XFS_STATS_INC(dp
->i_mount
, xs_dir_lookup
);
377 * We need to use KM_NOFS here so that lockdep will not throw false
378 * positive deadlock warnings on a non-transactional lookup path. It is
379 * safe to recurse into inode recalim in that case, but lockdep can't
380 * easily be taught about it. Hence KM_NOFS avoids having to add more
381 * lockdep Doing this avoids having to add a bunch of lockdep class
382 * annotations into the reclaim path for the ilock.
384 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
385 args
->geo
= dp
->i_mount
->m_dir_geo
;
386 args
->name
= name
->name
;
387 args
->namelen
= name
->len
;
388 args
->filetype
= name
->type
;
389 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
391 args
->whichfork
= XFS_DATA_FORK
;
393 args
->op_flags
= XFS_DA_OP_OKNOENT
;
395 args
->op_flags
|= XFS_DA_OP_CILOOKUP
;
397 lock_mode
= xfs_ilock_data_map_shared(dp
);
398 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
399 rval
= xfs_dir2_sf_lookup(args
);
403 rval
= xfs_dir2_isblock(args
, &v
);
407 rval
= xfs_dir2_block_lookup(args
);
411 rval
= xfs_dir2_isleaf(args
, &v
);
415 rval
= xfs_dir2_leaf_lookup(args
);
417 rval
= xfs_dir2_node_lookup(args
);
423 *inum
= args
->inumber
;
425 ci_name
->name
= args
->value
;
426 ci_name
->len
= args
->valuelen
;
430 xfs_iunlock(dp
, lock_mode
);
436 * Remove an entry from a directory.
442 struct xfs_name
*name
,
444 xfs_fsblock_t
*first
, /* bmap's firstblock */
445 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
446 xfs_extlen_t total
) /* bmap's total block count */
448 struct xfs_da_args
*args
;
450 int v
; /* type-checking value */
452 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
453 XFS_STATS_INC(dp
->i_mount
, xs_dir_remove
);
455 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
459 args
->geo
= dp
->i_mount
->m_dir_geo
;
460 args
->name
= name
->name
;
461 args
->namelen
= name
->len
;
462 args
->filetype
= name
->type
;
463 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
466 args
->firstblock
= first
;
469 args
->whichfork
= XFS_DATA_FORK
;
472 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
473 rval
= xfs_dir2_sf_removename(args
);
477 rval
= xfs_dir2_isblock(args
, &v
);
481 rval
= xfs_dir2_block_removename(args
);
485 rval
= xfs_dir2_isleaf(args
, &v
);
489 rval
= xfs_dir2_leaf_removename(args
);
491 rval
= xfs_dir2_node_removename(args
);
498 * Replace the inode number of a directory entry.
504 struct xfs_name
*name
, /* name of entry to replace */
505 xfs_ino_t inum
, /* new inode number */
506 xfs_fsblock_t
*first
, /* bmap's firstblock */
507 struct xfs_defer_ops
*dfops
, /* bmap's freeblock list */
508 xfs_extlen_t total
) /* bmap's total block count */
510 struct xfs_da_args
*args
;
512 int v
; /* type-checking value */
514 ASSERT(S_ISDIR(VFS_I(dp
)->i_mode
));
516 rval
= xfs_dir_ino_validate(tp
->t_mountp
, inum
);
520 args
= kmem_zalloc(sizeof(*args
), KM_SLEEP
| KM_NOFS
);
524 args
->geo
= dp
->i_mount
->m_dir_geo
;
525 args
->name
= name
->name
;
526 args
->namelen
= name
->len
;
527 args
->filetype
= name
->type
;
528 args
->hashval
= dp
->i_mount
->m_dirnameops
->hashname(name
);
529 args
->inumber
= inum
;
531 args
->firstblock
= first
;
534 args
->whichfork
= XFS_DATA_FORK
;
537 if (dp
->i_d
.di_format
== XFS_DINODE_FMT_LOCAL
) {
538 rval
= xfs_dir2_sf_replace(args
);
542 rval
= xfs_dir2_isblock(args
, &v
);
546 rval
= xfs_dir2_block_replace(args
);
550 rval
= xfs_dir2_isleaf(args
, &v
);
554 rval
= xfs_dir2_leaf_replace(args
);
556 rval
= xfs_dir2_node_replace(args
);
563 * See if this entry can be added to the directory without allocating space.
569 struct xfs_name
*name
) /* name of entry to add */
571 return xfs_dir_createname(tp
, dp
, name
, 0, NULL
, NULL
, 0);
579 * Add a block to the directory.
581 * This routine is for data and free blocks, not leaf/node blocks which are
582 * handled by xfs_da_grow_inode.
586 struct xfs_da_args
*args
,
587 int space
, /* v2 dir's space XFS_DIR2_xxx_SPACE */
588 xfs_dir2_db_t
*dbp
) /* out: block number added */
590 struct xfs_inode
*dp
= args
->dp
;
591 struct xfs_mount
*mp
= dp
->i_mount
;
592 xfs_fileoff_t bno
; /* directory offset of new block */
593 int count
; /* count of filesystem blocks */
596 trace_xfs_dir2_grow_inode(args
, space
);
599 * Set lowest possible block in the space requested.
601 bno
= XFS_B_TO_FSBT(mp
, space
* XFS_DIR2_SPACE_SIZE
);
602 count
= args
->geo
->fsbcount
;
604 error
= xfs_da_grow_inode_int(args
, &bno
, count
);
608 *dbp
= xfs_dir2_da_to_db(args
->geo
, (xfs_dablk_t
)bno
);
611 * Update file's size if this is the data space and it grew.
613 if (space
== XFS_DIR2_DATA_SPACE
) {
614 xfs_fsize_t size
; /* directory file (data) size */
616 size
= XFS_FSB_TO_B(mp
, bno
+ count
);
617 if (size
> dp
->i_d
.di_size
) {
618 dp
->i_d
.di_size
= size
;
619 xfs_trans_log_inode(args
->trans
, dp
, XFS_ILOG_CORE
);
626 * See if the directory is a single-block form directory.
630 struct xfs_da_args
*args
,
631 int *vp
) /* out: 1 is block, 0 is not block */
633 xfs_fileoff_t last
; /* last file offset */
636 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
638 rval
= XFS_FSB_TO_B(args
->dp
->i_mount
, last
) == args
->geo
->blksize
;
639 if (rval
!= 0 && args
->dp
->i_d
.di_size
!= args
->geo
->blksize
)
640 return -EFSCORRUPTED
;
646 * See if the directory is a single-leaf form directory.
650 struct xfs_da_args
*args
,
651 int *vp
) /* out: 1 is block, 0 is not block */
653 xfs_fileoff_t last
; /* last file offset */
656 if ((rval
= xfs_bmap_last_offset(args
->dp
, &last
, XFS_DATA_FORK
)))
658 *vp
= last
== args
->geo
->leafblk
+ args
->geo
->fsbcount
;
663 * Remove the given block from the directory.
664 * This routine is used for data and free blocks, leaf/node are done
665 * by xfs_da_shrink_inode.
668 xfs_dir2_shrink_inode(
673 xfs_fileoff_t bno
; /* directory file offset */
674 xfs_dablk_t da
; /* directory file offset */
675 int done
; /* bunmap is finished */
681 trace_xfs_dir2_shrink_inode(args
, db
);
686 da
= xfs_dir2_db_to_da(args
->geo
, db
);
688 /* Unmap the fsblock(s). */
689 error
= xfs_bunmapi(tp
, dp
, da
, args
->geo
->fsbcount
, 0, 0,
690 args
->firstblock
, args
->dfops
, &done
);
693 * ENOSPC actually can happen if we're in a removename with no
694 * space reservation, and the resulting block removal would
695 * cause a bmap btree split or conversion from extents to btree.
696 * This can only happen for un-fragmented directory blocks,
697 * since you need to be punching out the middle of an extent.
698 * In this case we need to leave the block in the file, and not
699 * binval it. So the block has to be in a consistent empty
700 * state and appropriately logged. We don't free up the buffer,
701 * the caller can tell it hasn't happened since it got an error
708 * Invalidate the buffer from the transaction.
710 xfs_trans_binval(tp
, bp
);
712 * If it's not a data block, we're done.
714 if (db
>= xfs_dir2_byte_to_db(args
->geo
, XFS_DIR2_LEAF_OFFSET
))
717 * If the block isn't the last one in the directory, we're done.
719 if (dp
->i_d
.di_size
> xfs_dir2_db_off_to_byte(args
->geo
, db
+ 1, 0))
722 if ((error
= xfs_bmap_last_before(tp
, dp
, &bno
, XFS_DATA_FORK
))) {
724 * This can't really happen unless there's kernel corruption.
728 if (db
== args
->geo
->datablk
)
733 * Set the size to the new last block.
735 dp
->i_d
.di_size
= XFS_FSB_TO_B(mp
, bno
);
736 xfs_trans_log_inode(tp
, dp
, XFS_ILOG_CORE
);