2 * Copyright (c) 2000-2005 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
24 #include "xfs_trans.h"
28 #include "xfs_alloc.h"
29 #include "xfs_dmapi.h"
30 #include "xfs_quota.h"
31 #include "xfs_mount.h"
32 #include "xfs_bmap_btree.h"
33 #include "xfs_alloc_btree.h"
34 #include "xfs_ialloc_btree.h"
35 #include "xfs_dir2_sf.h"
36 #include "xfs_attr_sf.h"
37 #include "xfs_dinode.h"
38 #include "xfs_inode.h"
40 #include "xfs_btree.h"
41 #include "xfs_ialloc.h"
42 #include "xfs_rtalloc.h"
43 #include "xfs_error.h"
44 #include "xfs_itable.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
52 #include <linux/capability.h>
53 #include <linux/xattr.h>
54 #include <linux/namei.h>
55 #include <linux/posix_acl.h>
56 #include <linux/security.h>
57 #include <linux/falloc.h>
58 #include <linux/fiemap.h>
61 * Bring the atime in the XFS inode uptodate.
62 * Used before logging the inode to disk or when the Linux inode goes away.
65 xfs_synchronize_atime(
68 struct inode
*inode
= VFS_I(ip
);
70 if (!(inode
->i_state
& I_CLEAR
)) {
71 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
72 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
77 * If the linux inode is valid, mark it dirty.
78 * Used when commiting a dirty inode into a transaction so that
79 * the inode will get written back by the linux code
82 xfs_mark_inode_dirty_sync(
85 struct inode
*inode
= VFS_I(ip
);
87 if (!(inode
->i_state
& (I_WILL_FREE
|I_FREEING
|I_CLEAR
)))
88 mark_inode_dirty_sync(inode
);
92 * Change the requested timestamp in the given inode.
93 * We don't lock across timestamp updates, and we don't log them but
94 * we do record the fact that there is dirty information in core.
101 struct inode
*inode
= VFS_I(ip
);
105 tv
= current_fs_time(inode
->i_sb
);
107 if ((flags
& XFS_ICHGTIME_MOD
) &&
108 !timespec_equal(&inode
->i_mtime
, &tv
)) {
110 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)tv
.tv_sec
;
111 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
114 if ((flags
& XFS_ICHGTIME_CHG
) &&
115 !timespec_equal(&inode
->i_ctime
, &tv
)) {
117 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tv
.tv_sec
;
118 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
123 * We update the i_update_core field _after_ changing
124 * the timestamps in order to coordinate properly with
125 * xfs_iflush() so that we don't lose timestamp updates.
126 * This keeps us from having to hold the inode lock
127 * while doing this. We use the SYNCHRONIZE macro to
128 * ensure that the compiler does not reorder the update
129 * of i_update_core above the timestamp updates above.
133 ip
->i_update_core
= 1;
134 xfs_mark_inode_dirty_sync(ip
);
139 * Hook in SELinux. This is not quite correct yet, what we really need
140 * here (as we do for default ACLs) is a mechanism by which creation of
141 * these attrs can be journalled at inode creation time (along with the
142 * inode, of course, such that log replay can't cause these to be lost).
149 struct xfs_inode
*ip
= XFS_I(inode
);
155 error
= security_inode_init_security(inode
, dir
, &name
,
158 if (error
== -EOPNOTSUPP
)
163 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
172 struct xfs_name
*namep
,
173 struct dentry
*dentry
)
175 namep
->name
= dentry
->d_name
.name
;
176 namep
->len
= dentry
->d_name
.len
;
183 struct dentry
*dentry
)
185 struct xfs_name teardown
;
188 * If we can't add the ACL or we fail in
189 * xfs_init_security we must back out.
190 * ENOSPC can hit here, among other things.
192 xfs_dentry_to_name(&teardown
, dentry
);
194 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
201 struct dentry
*dentry
,
206 struct xfs_inode
*ip
= NULL
;
207 struct posix_acl
*default_acl
= NULL
;
208 struct xfs_name name
;
212 * Irix uses Missed'em'V split, but doesn't want to see
213 * the upper 5 bits of (14bit) major.
215 if (S_ISCHR(mode
) || S_ISBLK(mode
)) {
216 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
218 rdev
= sysv_encode_dev(rdev
);
223 if (IS_POSIXACL(dir
)) {
224 default_acl
= xfs_get_acl(dir
, ACL_TYPE_DEFAULT
);
225 if (IS_ERR(default_acl
))
226 return -PTR_ERR(default_acl
);
229 mode
&= ~current_umask();
232 xfs_dentry_to_name(&name
, dentry
);
233 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
239 error
= xfs_init_security(inode
, dir
);
241 goto out_cleanup_inode
;
244 error
= -xfs_inherit_acl(inode
, default_acl
);
246 goto out_cleanup_inode
;
247 posix_acl_release(default_acl
);
251 d_instantiate(dentry
, inode
);
255 xfs_cleanup_inode(dir
, inode
, dentry
);
257 posix_acl_release(default_acl
);
264 struct dentry
*dentry
,
266 struct nameidata
*nd
)
268 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
274 struct dentry
*dentry
,
277 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
280 STATIC
struct dentry
*
283 struct dentry
*dentry
,
284 struct nameidata
*nd
)
286 struct xfs_inode
*cip
;
287 struct xfs_name name
;
290 if (dentry
->d_name
.len
>= MAXNAMELEN
)
291 return ERR_PTR(-ENAMETOOLONG
);
293 xfs_dentry_to_name(&name
, dentry
);
294 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
295 if (unlikely(error
)) {
296 if (unlikely(error
!= ENOENT
))
297 return ERR_PTR(-error
);
302 return d_splice_alias(VFS_I(cip
), dentry
);
305 STATIC
struct dentry
*
308 struct dentry
*dentry
,
309 struct nameidata
*nd
)
311 struct xfs_inode
*ip
;
312 struct xfs_name xname
;
313 struct xfs_name ci_name
;
317 if (dentry
->d_name
.len
>= MAXNAMELEN
)
318 return ERR_PTR(-ENAMETOOLONG
);
320 xfs_dentry_to_name(&xname
, dentry
);
321 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
322 if (unlikely(error
)) {
323 if (unlikely(error
!= ENOENT
))
324 return ERR_PTR(-error
);
326 * call d_add(dentry, NULL) here when d_drop_negative_children
327 * is called in xfs_vn_mknod (ie. allow negative dentries
328 * with CI filesystems).
333 /* if exact match, just splice and exit */
335 return d_splice_alias(VFS_I(ip
), dentry
);
337 /* else case-insensitive match... */
338 dname
.name
= ci_name
.name
;
339 dname
.len
= ci_name
.len
;
340 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
341 kmem_free(ci_name
.name
);
347 struct dentry
*old_dentry
,
349 struct dentry
*dentry
)
351 struct inode
*inode
= old_dentry
->d_inode
;
352 struct xfs_name name
;
355 xfs_dentry_to_name(&name
, dentry
);
357 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
361 atomic_inc(&inode
->i_count
);
362 d_instantiate(dentry
, inode
);
369 struct dentry
*dentry
)
371 struct xfs_name name
;
374 xfs_dentry_to_name(&name
, dentry
);
376 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
381 * With unlink, the VFS makes the dentry "negative": no inode,
382 * but still hashed. This is incompatible with case-insensitive
383 * mode, so invalidate (unhash) the dentry in CI-mode.
385 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
386 d_invalidate(dentry
);
393 struct dentry
*dentry
,
397 struct xfs_inode
*cip
= NULL
;
398 struct xfs_name name
;
403 (irix_symlink_mode
? 0777 & ~current_umask() : S_IRWXUGO
);
404 xfs_dentry_to_name(&name
, dentry
);
406 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
412 error
= xfs_init_security(inode
, dir
);
414 goto out_cleanup_inode
;
416 d_instantiate(dentry
, inode
);
420 xfs_cleanup_inode(dir
, inode
, dentry
);
428 struct dentry
*odentry
,
430 struct dentry
*ndentry
)
432 struct inode
*new_inode
= ndentry
->d_inode
;
433 struct xfs_name oname
;
434 struct xfs_name nname
;
436 xfs_dentry_to_name(&oname
, odentry
);
437 xfs_dentry_to_name(&nname
, ndentry
);
439 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
440 XFS_I(ndir
), &nname
, new_inode
?
441 XFS_I(new_inode
) : NULL
);
445 * careful here - this function can get called recursively, so
446 * we need to be very careful about how much stack we use.
447 * uio is kmalloced for this reason...
451 struct dentry
*dentry
,
452 struct nameidata
*nd
)
457 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
461 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
465 nd_set_link(nd
, link
);
471 nd_set_link(nd
, ERR_PTR(error
));
477 struct dentry
*dentry
,
478 struct nameidata
*nd
,
481 char *s
= nd_get_link(nd
);
489 struct vfsmount
*mnt
,
490 struct dentry
*dentry
,
493 struct inode
*inode
= dentry
->d_inode
;
494 struct xfs_inode
*ip
= XFS_I(inode
);
495 struct xfs_mount
*mp
= ip
->i_mount
;
497 xfs_itrace_entry(ip
);
499 if (XFS_FORCED_SHUTDOWN(mp
))
500 return XFS_ERROR(EIO
);
502 stat
->size
= XFS_ISIZE(ip
);
503 stat
->dev
= inode
->i_sb
->s_dev
;
504 stat
->mode
= ip
->i_d
.di_mode
;
505 stat
->nlink
= ip
->i_d
.di_nlink
;
506 stat
->uid
= ip
->i_d
.di_uid
;
507 stat
->gid
= ip
->i_d
.di_gid
;
508 stat
->ino
= ip
->i_ino
;
509 stat
->atime
= inode
->i_atime
;
510 stat
->mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
511 stat
->mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
512 stat
->ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
513 stat
->ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
515 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
518 switch (inode
->i_mode
& S_IFMT
) {
521 stat
->blksize
= BLKDEV_IOSIZE
;
522 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
523 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
526 if (XFS_IS_REALTIME_INODE(ip
)) {
528 * If the file blocks are being allocated from a
529 * realtime volume, then return the inode's realtime
530 * extent size or the realtime volume's extent size.
533 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
535 stat
->blksize
= xfs_preferred_iosize(mp
);
545 struct dentry
*dentry
,
548 return -xfs_setattr(XFS_I(dentry
->d_inode
), iattr
, 0);
552 * block_truncate_page can return an error, but we can't propagate it
553 * at all here. Leave a complaint + stack trace in the syslog because
554 * this could be bad. If it is bad, we need to propagate the error further.
561 error
= block_truncate_page(inode
->i_mapping
, inode
->i_size
,
576 xfs_inode_t
*ip
= XFS_I(inode
);
578 /* preallocation on directories not yet supported */
580 if (S_ISDIR(inode
->i_mode
))
587 xfs_ilock(ip
, XFS_IOLOCK_EXCL
);
588 error
= xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
590 if (!error
&& !(mode
& FALLOC_FL_KEEP_SIZE
) &&
591 offset
+ len
> i_size_read(inode
))
592 new_size
= offset
+ len
;
594 /* Change file size if needed */
598 iattr
.ia_valid
= ATTR_SIZE
;
599 iattr
.ia_size
= new_size
;
600 error
= xfs_setattr(ip
, &iattr
, XFS_ATTR_NOLOCK
);
603 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
608 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
611 * Call fiemap helper to fill in user data.
612 * Returns positive errors to xfs_getbmap.
617 struct getbmapx
*bmv
,
621 struct fiemap_extent_info
*fieinfo
= *arg
;
622 u32 fiemap_flags
= 0;
623 u64 logical
, physical
, length
;
625 /* Do nothing for a hole */
626 if (bmv
->bmv_block
== -1LL)
629 logical
= BBTOB(bmv
->bmv_offset
);
630 physical
= BBTOB(bmv
->bmv_block
);
631 length
= BBTOB(bmv
->bmv_length
);
633 if (bmv
->bmv_oflags
& BMV_OF_PREALLOC
)
634 fiemap_flags
|= FIEMAP_EXTENT_UNWRITTEN
;
635 else if (bmv
->bmv_oflags
& BMV_OF_DELALLOC
) {
636 fiemap_flags
|= FIEMAP_EXTENT_DELALLOC
;
637 physical
= 0; /* no block yet */
639 if (bmv
->bmv_oflags
& BMV_OF_LAST
)
640 fiemap_flags
|= FIEMAP_EXTENT_LAST
;
642 error
= fiemap_fill_next_extent(fieinfo
, logical
, physical
,
643 length
, fiemap_flags
);
646 *full
= 1; /* user array now full */
655 struct fiemap_extent_info
*fieinfo
,
659 xfs_inode_t
*ip
= XFS_I(inode
);
663 error
= fiemap_check_flags(fieinfo
, XFS_FIEMAP_FLAGS
);
667 /* Set up bmap header for xfs internal routine */
668 bm
.bmv_offset
= BTOBB(start
);
669 /* Special case for whole file */
670 if (length
== FIEMAP_MAX_OFFSET
)
671 bm
.bmv_length
= -1LL;
673 bm
.bmv_length
= BTOBB(length
);
675 /* We add one because in getbmap world count includes the header */
676 bm
.bmv_count
= fieinfo
->fi_extents_max
+ 1;
677 bm
.bmv_iflags
= BMV_IF_PREALLOC
;
678 if (fieinfo
->fi_flags
& FIEMAP_FLAG_XATTR
)
679 bm
.bmv_iflags
|= BMV_IF_ATTRFORK
;
680 if (!(fieinfo
->fi_flags
& FIEMAP_FLAG_SYNC
))
681 bm
.bmv_iflags
|= BMV_IF_DELALLOC
;
683 error
= xfs_getbmap(ip
, &bm
, xfs_fiemap_format
, fieinfo
);
690 static const struct inode_operations xfs_inode_operations
= {
691 .check_acl
= xfs_check_acl
,
692 .truncate
= xfs_vn_truncate
,
693 .getattr
= xfs_vn_getattr
,
694 .setattr
= xfs_vn_setattr
,
695 .setxattr
= generic_setxattr
,
696 .getxattr
= generic_getxattr
,
697 .removexattr
= generic_removexattr
,
698 .listxattr
= xfs_vn_listxattr
,
699 .fallocate
= xfs_vn_fallocate
,
700 .fiemap
= xfs_vn_fiemap
,
703 static const struct inode_operations xfs_dir_inode_operations
= {
704 .create
= xfs_vn_create
,
705 .lookup
= xfs_vn_lookup
,
707 .unlink
= xfs_vn_unlink
,
708 .symlink
= xfs_vn_symlink
,
709 .mkdir
= xfs_vn_mkdir
,
711 * Yes, XFS uses the same method for rmdir and unlink.
713 * There are some subtile differences deeper in the code,
714 * but we use S_ISDIR to check for those.
716 .rmdir
= xfs_vn_unlink
,
717 .mknod
= xfs_vn_mknod
,
718 .rename
= xfs_vn_rename
,
719 .check_acl
= xfs_check_acl
,
720 .getattr
= xfs_vn_getattr
,
721 .setattr
= xfs_vn_setattr
,
722 .setxattr
= generic_setxattr
,
723 .getxattr
= generic_getxattr
,
724 .removexattr
= generic_removexattr
,
725 .listxattr
= xfs_vn_listxattr
,
728 static const struct inode_operations xfs_dir_ci_inode_operations
= {
729 .create
= xfs_vn_create
,
730 .lookup
= xfs_vn_ci_lookup
,
732 .unlink
= xfs_vn_unlink
,
733 .symlink
= xfs_vn_symlink
,
734 .mkdir
= xfs_vn_mkdir
,
736 * Yes, XFS uses the same method for rmdir and unlink.
738 * There are some subtile differences deeper in the code,
739 * but we use S_ISDIR to check for those.
741 .rmdir
= xfs_vn_unlink
,
742 .mknod
= xfs_vn_mknod
,
743 .rename
= xfs_vn_rename
,
744 .check_acl
= xfs_check_acl
,
745 .getattr
= xfs_vn_getattr
,
746 .setattr
= xfs_vn_setattr
,
747 .setxattr
= generic_setxattr
,
748 .getxattr
= generic_getxattr
,
749 .removexattr
= generic_removexattr
,
750 .listxattr
= xfs_vn_listxattr
,
753 static const struct inode_operations xfs_symlink_inode_operations
= {
754 .readlink
= generic_readlink
,
755 .follow_link
= xfs_vn_follow_link
,
756 .put_link
= xfs_vn_put_link
,
757 .check_acl
= xfs_check_acl
,
758 .getattr
= xfs_vn_getattr
,
759 .setattr
= xfs_vn_setattr
,
760 .setxattr
= generic_setxattr
,
761 .getxattr
= generic_getxattr
,
762 .removexattr
= generic_removexattr
,
763 .listxattr
= xfs_vn_listxattr
,
767 xfs_diflags_to_iflags(
769 struct xfs_inode
*ip
)
771 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
772 inode
->i_flags
|= S_IMMUTABLE
;
774 inode
->i_flags
&= ~S_IMMUTABLE
;
775 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
776 inode
->i_flags
|= S_APPEND
;
778 inode
->i_flags
&= ~S_APPEND
;
779 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
780 inode
->i_flags
|= S_SYNC
;
782 inode
->i_flags
&= ~S_SYNC
;
783 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
784 inode
->i_flags
|= S_NOATIME
;
786 inode
->i_flags
&= ~S_NOATIME
;
790 * Initialize the Linux inode, set up the operation vectors and
793 * When reading existing inodes from disk this is called directly
794 * from xfs_iget, when creating a new inode it is called from
795 * xfs_ialloc after setting up the inode.
797 * We are always called with an uninitialised linux inode here.
798 * We need to initialise the necessary fields and take a reference
803 struct xfs_inode
*ip
)
805 struct inode
*inode
= &ip
->i_vnode
;
807 inode
->i_ino
= ip
->i_ino
;
808 inode
->i_state
= I_NEW
|I_LOCK
;
809 inode_add_to_lists(ip
->i_mount
->m_super
, inode
);
811 inode
->i_mode
= ip
->i_d
.di_mode
;
812 inode
->i_nlink
= ip
->i_d
.di_nlink
;
813 inode
->i_uid
= ip
->i_d
.di_uid
;
814 inode
->i_gid
= ip
->i_d
.di_gid
;
816 switch (inode
->i_mode
& S_IFMT
) {
820 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
821 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
828 inode
->i_generation
= ip
->i_d
.di_gen
;
829 i_size_write(inode
, ip
->i_d
.di_size
);
830 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
831 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
832 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
833 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
834 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
835 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
836 xfs_diflags_to_iflags(inode
, ip
);
838 switch (inode
->i_mode
& S_IFMT
) {
840 inode
->i_op
= &xfs_inode_operations
;
841 inode
->i_fop
= &xfs_file_operations
;
842 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
845 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
846 inode
->i_op
= &xfs_dir_ci_inode_operations
;
848 inode
->i_op
= &xfs_dir_inode_operations
;
849 inode
->i_fop
= &xfs_dir_file_operations
;
852 inode
->i_op
= &xfs_symlink_inode_operations
;
853 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
854 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
857 inode
->i_op
= &xfs_inode_operations
;
858 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
862 xfs_iflags_clear(ip
, XFS_INEW
);
865 unlock_new_inode(inode
);