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"
47 #include "xfs_buf_item.h"
48 #include "xfs_utils.h"
49 #include "xfs_vnodeops.h"
50 #include "xfs_trace.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 timestamps in the XFS inode uptodate.
63 * Used before writing the inode to disk.
66 xfs_synchronize_times(
69 struct inode
*inode
= VFS_I(ip
);
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
;
73 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)inode
->i_ctime
.tv_sec
;
74 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)inode
->i_ctime
.tv_nsec
;
75 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)inode
->i_mtime
.tv_sec
;
76 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)inode
->i_mtime
.tv_nsec
;
80 * If the linux inode is valid, mark it dirty.
81 * Used when commiting a dirty inode into a transaction so that
82 * the inode will get written back by the linux code
85 xfs_mark_inode_dirty_sync(
88 struct inode
*inode
= VFS_I(ip
);
90 if (!(inode
->i_state
& (I_WILL_FREE
|I_FREEING
|I_CLEAR
)))
91 mark_inode_dirty_sync(inode
);
95 * Change the requested timestamp in the given inode.
96 * We don't lock across timestamp updates, and we don't log them but
97 * we do record the fact that there is dirty information in core.
104 struct inode
*inode
= VFS_I(ip
);
108 tv
= current_fs_time(inode
->i_sb
);
110 if ((flags
& XFS_ICHGTIME_MOD
) &&
111 !timespec_equal(&inode
->i_mtime
, &tv
)) {
115 if ((flags
& XFS_ICHGTIME_CHG
) &&
116 !timespec_equal(&inode
->i_ctime
, &tv
)) {
122 * Update complete - now make sure everyone knows that the inode
126 xfs_mark_inode_dirty_sync(ip
);
130 * Hook in SELinux. This is not quite correct yet, what we really need
131 * here (as we do for default ACLs) is a mechanism by which creation of
132 * these attrs can be journalled at inode creation time (along with the
133 * inode, of course, such that log replay can't cause these to be lost).
140 struct xfs_inode
*ip
= XFS_I(inode
);
146 error
= security_inode_init_security(inode
, dir
, (char **)&name
,
149 if (error
== -EOPNOTSUPP
)
154 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
163 struct xfs_name
*namep
,
164 struct dentry
*dentry
)
166 namep
->name
= dentry
->d_name
.name
;
167 namep
->len
= dentry
->d_name
.len
;
174 struct dentry
*dentry
)
176 struct xfs_name teardown
;
179 * If we can't add the ACL or we fail in
180 * xfs_init_security we must back out.
181 * ENOSPC can hit here, among other things.
183 xfs_dentry_to_name(&teardown
, dentry
);
185 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
192 struct dentry
*dentry
,
197 struct xfs_inode
*ip
= NULL
;
198 struct posix_acl
*default_acl
= NULL
;
199 struct xfs_name name
;
203 * Irix uses Missed'em'V split, but doesn't want to see
204 * the upper 5 bits of (14bit) major.
206 if (S_ISCHR(mode
) || S_ISBLK(mode
)) {
207 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
209 rdev
= sysv_encode_dev(rdev
);
214 if (IS_POSIXACL(dir
)) {
215 default_acl
= xfs_get_acl(dir
, ACL_TYPE_DEFAULT
);
216 if (IS_ERR(default_acl
))
217 return -PTR_ERR(default_acl
);
220 mode
&= ~current_umask();
223 xfs_dentry_to_name(&name
, dentry
);
224 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
230 error
= xfs_init_security(inode
, dir
);
232 goto out_cleanup_inode
;
235 error
= -xfs_inherit_acl(inode
, default_acl
);
237 goto out_cleanup_inode
;
238 posix_acl_release(default_acl
);
242 d_instantiate(dentry
, inode
);
246 xfs_cleanup_inode(dir
, inode
, dentry
);
248 posix_acl_release(default_acl
);
255 struct dentry
*dentry
,
257 struct nameidata
*nd
)
259 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
265 struct dentry
*dentry
,
268 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
271 STATIC
struct dentry
*
274 struct dentry
*dentry
,
275 struct nameidata
*nd
)
277 struct xfs_inode
*cip
;
278 struct xfs_name name
;
281 if (dentry
->d_name
.len
>= MAXNAMELEN
)
282 return ERR_PTR(-ENAMETOOLONG
);
284 xfs_dentry_to_name(&name
, dentry
);
285 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
286 if (unlikely(error
)) {
287 if (unlikely(error
!= ENOENT
))
288 return ERR_PTR(-error
);
293 return d_splice_alias(VFS_I(cip
), dentry
);
296 STATIC
struct dentry
*
299 struct dentry
*dentry
,
300 struct nameidata
*nd
)
302 struct xfs_inode
*ip
;
303 struct xfs_name xname
;
304 struct xfs_name ci_name
;
308 if (dentry
->d_name
.len
>= MAXNAMELEN
)
309 return ERR_PTR(-ENAMETOOLONG
);
311 xfs_dentry_to_name(&xname
, dentry
);
312 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
313 if (unlikely(error
)) {
314 if (unlikely(error
!= ENOENT
))
315 return ERR_PTR(-error
);
317 * call d_add(dentry, NULL) here when d_drop_negative_children
318 * is called in xfs_vn_mknod (ie. allow negative dentries
319 * with CI filesystems).
324 /* if exact match, just splice and exit */
326 return d_splice_alias(VFS_I(ip
), dentry
);
328 /* else case-insensitive match... */
329 dname
.name
= ci_name
.name
;
330 dname
.len
= ci_name
.len
;
331 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
332 kmem_free(ci_name
.name
);
338 struct dentry
*old_dentry
,
340 struct dentry
*dentry
)
342 struct inode
*inode
= old_dentry
->d_inode
;
343 struct xfs_name name
;
346 xfs_dentry_to_name(&name
, dentry
);
348 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
352 atomic_inc(&inode
->i_count
);
353 d_instantiate(dentry
, inode
);
360 struct dentry
*dentry
)
362 struct xfs_name name
;
365 xfs_dentry_to_name(&name
, dentry
);
367 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
372 * With unlink, the VFS makes the dentry "negative": no inode,
373 * but still hashed. This is incompatible with case-insensitive
374 * mode, so invalidate (unhash) the dentry in CI-mode.
376 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
377 d_invalidate(dentry
);
384 struct dentry
*dentry
,
388 struct xfs_inode
*cip
= NULL
;
389 struct xfs_name name
;
394 (irix_symlink_mode
? 0777 & ~current_umask() : S_IRWXUGO
);
395 xfs_dentry_to_name(&name
, dentry
);
397 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
403 error
= xfs_init_security(inode
, dir
);
405 goto out_cleanup_inode
;
407 d_instantiate(dentry
, inode
);
411 xfs_cleanup_inode(dir
, inode
, dentry
);
419 struct dentry
*odentry
,
421 struct dentry
*ndentry
)
423 struct inode
*new_inode
= ndentry
->d_inode
;
424 struct xfs_name oname
;
425 struct xfs_name nname
;
427 xfs_dentry_to_name(&oname
, odentry
);
428 xfs_dentry_to_name(&nname
, ndentry
);
430 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
431 XFS_I(ndir
), &nname
, new_inode
?
432 XFS_I(new_inode
) : NULL
);
436 * careful here - this function can get called recursively, so
437 * we need to be very careful about how much stack we use.
438 * uio is kmalloced for this reason...
442 struct dentry
*dentry
,
443 struct nameidata
*nd
)
448 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
452 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
456 nd_set_link(nd
, link
);
462 nd_set_link(nd
, ERR_PTR(error
));
468 struct dentry
*dentry
,
469 struct nameidata
*nd
,
472 char *s
= nd_get_link(nd
);
480 struct vfsmount
*mnt
,
481 struct dentry
*dentry
,
484 struct inode
*inode
= dentry
->d_inode
;
485 struct xfs_inode
*ip
= XFS_I(inode
);
486 struct xfs_mount
*mp
= ip
->i_mount
;
488 xfs_itrace_entry(ip
);
490 if (XFS_FORCED_SHUTDOWN(mp
))
491 return XFS_ERROR(EIO
);
493 stat
->size
= XFS_ISIZE(ip
);
494 stat
->dev
= inode
->i_sb
->s_dev
;
495 stat
->mode
= ip
->i_d
.di_mode
;
496 stat
->nlink
= ip
->i_d
.di_nlink
;
497 stat
->uid
= ip
->i_d
.di_uid
;
498 stat
->gid
= ip
->i_d
.di_gid
;
499 stat
->ino
= ip
->i_ino
;
500 stat
->atime
= inode
->i_atime
;
501 stat
->mtime
= inode
->i_mtime
;
502 stat
->ctime
= inode
->i_ctime
;
504 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
507 switch (inode
->i_mode
& S_IFMT
) {
510 stat
->blksize
= BLKDEV_IOSIZE
;
511 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
512 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
515 if (XFS_IS_REALTIME_INODE(ip
)) {
517 * If the file blocks are being allocated from a
518 * realtime volume, then return the inode's realtime
519 * extent size or the realtime volume's extent size.
522 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
524 stat
->blksize
= xfs_preferred_iosize(mp
);
534 struct dentry
*dentry
,
537 return -xfs_setattr(XFS_I(dentry
->d_inode
), iattr
, 0);
541 * block_truncate_page can return an error, but we can't propagate it
542 * at all here. Leave a complaint + stack trace in the syslog because
543 * this could be bad. If it is bad, we need to propagate the error further.
550 error
= block_truncate_page(inode
->i_mapping
, inode
->i_size
,
565 xfs_inode_t
*ip
= XFS_I(inode
);
567 /* preallocation on directories not yet supported */
569 if (S_ISDIR(inode
->i_mode
))
576 xfs_ilock(ip
, XFS_IOLOCK_EXCL
);
577 error
= -xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
579 if (!error
&& !(mode
& FALLOC_FL_KEEP_SIZE
) &&
580 offset
+ len
> i_size_read(inode
))
581 new_size
= offset
+ len
;
583 /* Change file size if needed */
587 iattr
.ia_valid
= ATTR_SIZE
;
588 iattr
.ia_size
= new_size
;
589 error
= -xfs_setattr(ip
, &iattr
, XFS_ATTR_NOLOCK
);
592 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
597 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
600 * Call fiemap helper to fill in user data.
601 * Returns positive errors to xfs_getbmap.
606 struct getbmapx
*bmv
,
610 struct fiemap_extent_info
*fieinfo
= *arg
;
611 u32 fiemap_flags
= 0;
612 u64 logical
, physical
, length
;
614 /* Do nothing for a hole */
615 if (bmv
->bmv_block
== -1LL)
618 logical
= BBTOB(bmv
->bmv_offset
);
619 physical
= BBTOB(bmv
->bmv_block
);
620 length
= BBTOB(bmv
->bmv_length
);
622 if (bmv
->bmv_oflags
& BMV_OF_PREALLOC
)
623 fiemap_flags
|= FIEMAP_EXTENT_UNWRITTEN
;
624 else if (bmv
->bmv_oflags
& BMV_OF_DELALLOC
) {
625 fiemap_flags
|= FIEMAP_EXTENT_DELALLOC
;
626 physical
= 0; /* no block yet */
628 if (bmv
->bmv_oflags
& BMV_OF_LAST
)
629 fiemap_flags
|= FIEMAP_EXTENT_LAST
;
631 error
= fiemap_fill_next_extent(fieinfo
, logical
, physical
,
632 length
, fiemap_flags
);
635 *full
= 1; /* user array now full */
644 struct fiemap_extent_info
*fieinfo
,
648 xfs_inode_t
*ip
= XFS_I(inode
);
652 error
= fiemap_check_flags(fieinfo
, XFS_FIEMAP_FLAGS
);
656 /* Set up bmap header for xfs internal routine */
657 bm
.bmv_offset
= BTOBB(start
);
658 /* Special case for whole file */
659 if (length
== FIEMAP_MAX_OFFSET
)
660 bm
.bmv_length
= -1LL;
662 bm
.bmv_length
= BTOBB(length
);
664 /* We add one because in getbmap world count includes the header */
665 bm
.bmv_count
= fieinfo
->fi_extents_max
+ 1;
666 bm
.bmv_iflags
= BMV_IF_PREALLOC
;
667 if (fieinfo
->fi_flags
& FIEMAP_FLAG_XATTR
)
668 bm
.bmv_iflags
|= BMV_IF_ATTRFORK
;
669 if (!(fieinfo
->fi_flags
& FIEMAP_FLAG_SYNC
))
670 bm
.bmv_iflags
|= BMV_IF_DELALLOC
;
672 error
= xfs_getbmap(ip
, &bm
, xfs_fiemap_format
, fieinfo
);
679 static const struct inode_operations xfs_inode_operations
= {
680 .check_acl
= xfs_check_acl
,
681 .truncate
= xfs_vn_truncate
,
682 .getattr
= xfs_vn_getattr
,
683 .setattr
= xfs_vn_setattr
,
684 .setxattr
= generic_setxattr
,
685 .getxattr
= generic_getxattr
,
686 .removexattr
= generic_removexattr
,
687 .listxattr
= xfs_vn_listxattr
,
688 .fallocate
= xfs_vn_fallocate
,
689 .fiemap
= xfs_vn_fiemap
,
692 static const struct inode_operations xfs_dir_inode_operations
= {
693 .create
= xfs_vn_create
,
694 .lookup
= xfs_vn_lookup
,
696 .unlink
= xfs_vn_unlink
,
697 .symlink
= xfs_vn_symlink
,
698 .mkdir
= xfs_vn_mkdir
,
700 * Yes, XFS uses the same method for rmdir and unlink.
702 * There are some subtile differences deeper in the code,
703 * but we use S_ISDIR to check for those.
705 .rmdir
= xfs_vn_unlink
,
706 .mknod
= xfs_vn_mknod
,
707 .rename
= xfs_vn_rename
,
708 .check_acl
= xfs_check_acl
,
709 .getattr
= xfs_vn_getattr
,
710 .setattr
= xfs_vn_setattr
,
711 .setxattr
= generic_setxattr
,
712 .getxattr
= generic_getxattr
,
713 .removexattr
= generic_removexattr
,
714 .listxattr
= xfs_vn_listxattr
,
717 static const struct inode_operations xfs_dir_ci_inode_operations
= {
718 .create
= xfs_vn_create
,
719 .lookup
= xfs_vn_ci_lookup
,
721 .unlink
= xfs_vn_unlink
,
722 .symlink
= xfs_vn_symlink
,
723 .mkdir
= xfs_vn_mkdir
,
725 * Yes, XFS uses the same method for rmdir and unlink.
727 * There are some subtile differences deeper in the code,
728 * but we use S_ISDIR to check for those.
730 .rmdir
= xfs_vn_unlink
,
731 .mknod
= xfs_vn_mknod
,
732 .rename
= xfs_vn_rename
,
733 .check_acl
= xfs_check_acl
,
734 .getattr
= xfs_vn_getattr
,
735 .setattr
= xfs_vn_setattr
,
736 .setxattr
= generic_setxattr
,
737 .getxattr
= generic_getxattr
,
738 .removexattr
= generic_removexattr
,
739 .listxattr
= xfs_vn_listxattr
,
742 static const struct inode_operations xfs_symlink_inode_operations
= {
743 .readlink
= generic_readlink
,
744 .follow_link
= xfs_vn_follow_link
,
745 .put_link
= xfs_vn_put_link
,
746 .check_acl
= xfs_check_acl
,
747 .getattr
= xfs_vn_getattr
,
748 .setattr
= xfs_vn_setattr
,
749 .setxattr
= generic_setxattr
,
750 .getxattr
= generic_getxattr
,
751 .removexattr
= generic_removexattr
,
752 .listxattr
= xfs_vn_listxattr
,
756 xfs_diflags_to_iflags(
758 struct xfs_inode
*ip
)
760 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
761 inode
->i_flags
|= S_IMMUTABLE
;
763 inode
->i_flags
&= ~S_IMMUTABLE
;
764 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
765 inode
->i_flags
|= S_APPEND
;
767 inode
->i_flags
&= ~S_APPEND
;
768 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
769 inode
->i_flags
|= S_SYNC
;
771 inode
->i_flags
&= ~S_SYNC
;
772 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
773 inode
->i_flags
|= S_NOATIME
;
775 inode
->i_flags
&= ~S_NOATIME
;
779 * Initialize the Linux inode, set up the operation vectors and
782 * When reading existing inodes from disk this is called directly
783 * from xfs_iget, when creating a new inode it is called from
784 * xfs_ialloc after setting up the inode.
786 * We are always called with an uninitialised linux inode here.
787 * We need to initialise the necessary fields and take a reference
792 struct xfs_inode
*ip
)
794 struct inode
*inode
= &ip
->i_vnode
;
796 inode
->i_ino
= ip
->i_ino
;
797 inode
->i_state
= I_NEW
;
798 inode_add_to_lists(ip
->i_mount
->m_super
, inode
);
800 inode
->i_mode
= ip
->i_d
.di_mode
;
801 inode
->i_nlink
= ip
->i_d
.di_nlink
;
802 inode
->i_uid
= ip
->i_d
.di_uid
;
803 inode
->i_gid
= ip
->i_d
.di_gid
;
805 switch (inode
->i_mode
& S_IFMT
) {
809 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
810 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
817 inode
->i_generation
= ip
->i_d
.di_gen
;
818 i_size_write(inode
, ip
->i_d
.di_size
);
819 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
820 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
821 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
822 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
823 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
824 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
825 xfs_diflags_to_iflags(inode
, ip
);
827 switch (inode
->i_mode
& S_IFMT
) {
829 inode
->i_op
= &xfs_inode_operations
;
830 inode
->i_fop
= &xfs_file_operations
;
831 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
834 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
835 inode
->i_op
= &xfs_dir_ci_inode_operations
;
837 inode
->i_op
= &xfs_dir_inode_operations
;
838 inode
->i_fop
= &xfs_dir_file_operations
;
841 inode
->i_op
= &xfs_symlink_inode_operations
;
842 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
843 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
846 inode
->i_op
= &xfs_inode_operations
;
847 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
851 xfs_iflags_clear(ip
, XFS_INEW
);
854 unlock_new_inode(inode
);