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>
59 #include <linux/slab.h>
62 * Bring the timestamps in the XFS inode uptodate.
64 * Used before writing the inode to disk.
67 xfs_synchronize_times(
70 struct inode
*inode
= VFS_I(ip
);
72 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
73 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
74 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)inode
->i_ctime
.tv_sec
;
75 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)inode
->i_ctime
.tv_nsec
;
76 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)inode
->i_mtime
.tv_sec
;
77 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)inode
->i_mtime
.tv_nsec
;
81 * If the linux inode is valid, mark it dirty.
82 * Used when commiting a dirty inode into a transaction so that
83 * the inode will get written back by the linux code
86 xfs_mark_inode_dirty_sync(
89 struct inode
*inode
= VFS_I(ip
);
91 if (!(inode
->i_state
& (I_WILL_FREE
|I_FREEING
|I_CLEAR
)))
92 mark_inode_dirty_sync(inode
);
99 struct inode
*inode
= VFS_I(ip
);
101 if (!(inode
->i_state
& (I_WILL_FREE
|I_FREEING
|I_CLEAR
)))
102 mark_inode_dirty(inode
);
106 * Change the requested timestamp in the given inode.
107 * We don't lock across timestamp updates, and we don't log them but
108 * we do record the fact that there is dirty information in core.
115 struct inode
*inode
= VFS_I(ip
);
119 tv
= current_fs_time(inode
->i_sb
);
121 if ((flags
& XFS_ICHGTIME_MOD
) &&
122 !timespec_equal(&inode
->i_mtime
, &tv
)) {
126 if ((flags
& XFS_ICHGTIME_CHG
) &&
127 !timespec_equal(&inode
->i_ctime
, &tv
)) {
133 * Update complete - now make sure everyone knows that the inode
137 xfs_mark_inode_dirty_sync(ip
);
141 * Hook in SELinux. This is not quite correct yet, what we really need
142 * here (as we do for default ACLs) is a mechanism by which creation of
143 * these attrs can be journalled at inode creation time (along with the
144 * inode, of course, such that log replay can't cause these to be lost).
151 struct xfs_inode
*ip
= XFS_I(inode
);
157 error
= security_inode_init_security(inode
, dir
, (char **)&name
,
160 if (error
== -EOPNOTSUPP
)
165 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
174 struct xfs_name
*namep
,
175 struct dentry
*dentry
)
177 namep
->name
= dentry
->d_name
.name
;
178 namep
->len
= dentry
->d_name
.len
;
185 struct dentry
*dentry
)
187 struct xfs_name teardown
;
190 * If we can't add the ACL or we fail in
191 * xfs_init_security we must back out.
192 * ENOSPC can hit here, among other things.
194 xfs_dentry_to_name(&teardown
, dentry
);
196 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
203 struct dentry
*dentry
,
208 struct xfs_inode
*ip
= NULL
;
209 struct posix_acl
*default_acl
= NULL
;
210 struct xfs_name name
;
214 * Irix uses Missed'em'V split, but doesn't want to see
215 * the upper 5 bits of (14bit) major.
217 if (S_ISCHR(mode
) || S_ISBLK(mode
)) {
218 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
220 rdev
= sysv_encode_dev(rdev
);
225 if (IS_POSIXACL(dir
)) {
226 default_acl
= xfs_get_acl(dir
, ACL_TYPE_DEFAULT
);
227 if (IS_ERR(default_acl
))
228 return -PTR_ERR(default_acl
);
231 mode
&= ~current_umask();
234 xfs_dentry_to_name(&name
, dentry
);
235 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
241 error
= xfs_init_security(inode
, dir
);
243 goto out_cleanup_inode
;
246 error
= -xfs_inherit_acl(inode
, default_acl
);
248 goto out_cleanup_inode
;
249 posix_acl_release(default_acl
);
253 d_instantiate(dentry
, inode
);
257 xfs_cleanup_inode(dir
, inode
, dentry
);
259 posix_acl_release(default_acl
);
266 struct dentry
*dentry
,
268 struct nameidata
*nd
)
270 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
276 struct dentry
*dentry
,
279 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
282 STATIC
struct dentry
*
285 struct dentry
*dentry
,
286 struct nameidata
*nd
)
288 struct xfs_inode
*cip
;
289 struct xfs_name name
;
292 if (dentry
->d_name
.len
>= MAXNAMELEN
)
293 return ERR_PTR(-ENAMETOOLONG
);
295 xfs_dentry_to_name(&name
, dentry
);
296 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
297 if (unlikely(error
)) {
298 if (unlikely(error
!= ENOENT
))
299 return ERR_PTR(-error
);
304 return d_splice_alias(VFS_I(cip
), dentry
);
307 STATIC
struct dentry
*
310 struct dentry
*dentry
,
311 struct nameidata
*nd
)
313 struct xfs_inode
*ip
;
314 struct xfs_name xname
;
315 struct xfs_name ci_name
;
319 if (dentry
->d_name
.len
>= MAXNAMELEN
)
320 return ERR_PTR(-ENAMETOOLONG
);
322 xfs_dentry_to_name(&xname
, dentry
);
323 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
324 if (unlikely(error
)) {
325 if (unlikely(error
!= ENOENT
))
326 return ERR_PTR(-error
);
328 * call d_add(dentry, NULL) here when d_drop_negative_children
329 * is called in xfs_vn_mknod (ie. allow negative dentries
330 * with CI filesystems).
335 /* if exact match, just splice and exit */
337 return d_splice_alias(VFS_I(ip
), dentry
);
339 /* else case-insensitive match... */
340 dname
.name
= ci_name
.name
;
341 dname
.len
= ci_name
.len
;
342 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
343 kmem_free(ci_name
.name
);
349 struct dentry
*old_dentry
,
351 struct dentry
*dentry
)
353 struct inode
*inode
= old_dentry
->d_inode
;
354 struct xfs_name name
;
357 xfs_dentry_to_name(&name
, dentry
);
359 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
363 atomic_inc(&inode
->i_count
);
364 d_instantiate(dentry
, inode
);
371 struct dentry
*dentry
)
373 struct xfs_name name
;
376 xfs_dentry_to_name(&name
, dentry
);
378 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
383 * With unlink, the VFS makes the dentry "negative": no inode,
384 * but still hashed. This is incompatible with case-insensitive
385 * mode, so invalidate (unhash) the dentry in CI-mode.
387 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
388 d_invalidate(dentry
);
395 struct dentry
*dentry
,
399 struct xfs_inode
*cip
= NULL
;
400 struct xfs_name name
;
405 (irix_symlink_mode
? 0777 & ~current_umask() : S_IRWXUGO
);
406 xfs_dentry_to_name(&name
, dentry
);
408 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
414 error
= xfs_init_security(inode
, dir
);
416 goto out_cleanup_inode
;
418 d_instantiate(dentry
, inode
);
422 xfs_cleanup_inode(dir
, inode
, dentry
);
430 struct dentry
*odentry
,
432 struct dentry
*ndentry
)
434 struct inode
*new_inode
= ndentry
->d_inode
;
435 struct xfs_name oname
;
436 struct xfs_name nname
;
438 xfs_dentry_to_name(&oname
, odentry
);
439 xfs_dentry_to_name(&nname
, ndentry
);
441 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
442 XFS_I(ndir
), &nname
, new_inode
?
443 XFS_I(new_inode
) : NULL
);
447 * careful here - this function can get called recursively, so
448 * we need to be very careful about how much stack we use.
449 * uio is kmalloced for this reason...
453 struct dentry
*dentry
,
454 struct nameidata
*nd
)
459 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
463 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
467 nd_set_link(nd
, link
);
473 nd_set_link(nd
, ERR_PTR(error
));
479 struct dentry
*dentry
,
480 struct nameidata
*nd
,
483 char *s
= nd_get_link(nd
);
491 struct vfsmount
*mnt
,
492 struct dentry
*dentry
,
495 struct inode
*inode
= dentry
->d_inode
;
496 struct xfs_inode
*ip
= XFS_I(inode
);
497 struct xfs_mount
*mp
= ip
->i_mount
;
499 xfs_itrace_entry(ip
);
501 if (XFS_FORCED_SHUTDOWN(mp
))
502 return XFS_ERROR(EIO
);
504 stat
->size
= XFS_ISIZE(ip
);
505 stat
->dev
= inode
->i_sb
->s_dev
;
506 stat
->mode
= ip
->i_d
.di_mode
;
507 stat
->nlink
= ip
->i_d
.di_nlink
;
508 stat
->uid
= ip
->i_d
.di_uid
;
509 stat
->gid
= ip
->i_d
.di_gid
;
510 stat
->ino
= ip
->i_ino
;
511 stat
->atime
= inode
->i_atime
;
512 stat
->mtime
= inode
->i_mtime
;
513 stat
->ctime
= inode
->i_ctime
;
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
);
589 /* check the new inode size is valid before allocating */
590 if (!(mode
& FALLOC_FL_KEEP_SIZE
) &&
591 offset
+ len
> i_size_read(inode
)) {
592 new_size
= offset
+ len
;
593 error
= inode_newsize_ok(inode
, new_size
);
598 error
= -xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
603 /* Change file size if needed */
607 iattr
.ia_valid
= ATTR_SIZE
;
608 iattr
.ia_size
= new_size
;
609 error
= -xfs_setattr(ip
, &iattr
, XFS_ATTR_NOLOCK
);
613 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
618 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
621 * Call fiemap helper to fill in user data.
622 * Returns positive errors to xfs_getbmap.
627 struct getbmapx
*bmv
,
631 struct fiemap_extent_info
*fieinfo
= *arg
;
632 u32 fiemap_flags
= 0;
633 u64 logical
, physical
, length
;
635 /* Do nothing for a hole */
636 if (bmv
->bmv_block
== -1LL)
639 logical
= BBTOB(bmv
->bmv_offset
);
640 physical
= BBTOB(bmv
->bmv_block
);
641 length
= BBTOB(bmv
->bmv_length
);
643 if (bmv
->bmv_oflags
& BMV_OF_PREALLOC
)
644 fiemap_flags
|= FIEMAP_EXTENT_UNWRITTEN
;
645 else if (bmv
->bmv_oflags
& BMV_OF_DELALLOC
) {
646 fiemap_flags
|= FIEMAP_EXTENT_DELALLOC
;
647 physical
= 0; /* no block yet */
649 if (bmv
->bmv_oflags
& BMV_OF_LAST
)
650 fiemap_flags
|= FIEMAP_EXTENT_LAST
;
652 error
= fiemap_fill_next_extent(fieinfo
, logical
, physical
,
653 length
, fiemap_flags
);
656 *full
= 1; /* user array now full */
665 struct fiemap_extent_info
*fieinfo
,
669 xfs_inode_t
*ip
= XFS_I(inode
);
673 error
= fiemap_check_flags(fieinfo
, XFS_FIEMAP_FLAGS
);
677 /* Set up bmap header for xfs internal routine */
678 bm
.bmv_offset
= BTOBB(start
);
679 /* Special case for whole file */
680 if (length
== FIEMAP_MAX_OFFSET
)
681 bm
.bmv_length
= -1LL;
683 bm
.bmv_length
= BTOBB(length
);
685 /* We add one because in getbmap world count includes the header */
686 bm
.bmv_count
= !fieinfo
->fi_extents_max
? MAXEXTNUM
:
687 fieinfo
->fi_extents_max
+ 1;
688 bm
.bmv_count
= min_t(__s32
, bm
.bmv_count
,
689 (PAGE_SIZE
* 16 / sizeof(struct getbmapx
)));
690 bm
.bmv_iflags
= BMV_IF_PREALLOC
;
691 if (fieinfo
->fi_flags
& FIEMAP_FLAG_XATTR
)
692 bm
.bmv_iflags
|= BMV_IF_ATTRFORK
;
693 if (!(fieinfo
->fi_flags
& FIEMAP_FLAG_SYNC
))
694 bm
.bmv_iflags
|= BMV_IF_DELALLOC
;
696 error
= xfs_getbmap(ip
, &bm
, xfs_fiemap_format
, fieinfo
);
703 static const struct inode_operations xfs_inode_operations
= {
704 .check_acl
= xfs_check_acl
,
705 .truncate
= xfs_vn_truncate
,
706 .getattr
= xfs_vn_getattr
,
707 .setattr
= xfs_vn_setattr
,
708 .setxattr
= generic_setxattr
,
709 .getxattr
= generic_getxattr
,
710 .removexattr
= generic_removexattr
,
711 .listxattr
= xfs_vn_listxattr
,
712 .fallocate
= xfs_vn_fallocate
,
713 .fiemap
= xfs_vn_fiemap
,
716 static const struct inode_operations xfs_dir_inode_operations
= {
717 .create
= xfs_vn_create
,
718 .lookup
= xfs_vn_lookup
,
720 .unlink
= xfs_vn_unlink
,
721 .symlink
= xfs_vn_symlink
,
722 .mkdir
= xfs_vn_mkdir
,
724 * Yes, XFS uses the same method for rmdir and unlink.
726 * There are some subtile differences deeper in the code,
727 * but we use S_ISDIR to check for those.
729 .rmdir
= xfs_vn_unlink
,
730 .mknod
= xfs_vn_mknod
,
731 .rename
= xfs_vn_rename
,
732 .check_acl
= xfs_check_acl
,
733 .getattr
= xfs_vn_getattr
,
734 .setattr
= xfs_vn_setattr
,
735 .setxattr
= generic_setxattr
,
736 .getxattr
= generic_getxattr
,
737 .removexattr
= generic_removexattr
,
738 .listxattr
= xfs_vn_listxattr
,
741 static const struct inode_operations xfs_dir_ci_inode_operations
= {
742 .create
= xfs_vn_create
,
743 .lookup
= xfs_vn_ci_lookup
,
745 .unlink
= xfs_vn_unlink
,
746 .symlink
= xfs_vn_symlink
,
747 .mkdir
= xfs_vn_mkdir
,
749 * Yes, XFS uses the same method for rmdir and unlink.
751 * There are some subtile differences deeper in the code,
752 * but we use S_ISDIR to check for those.
754 .rmdir
= xfs_vn_unlink
,
755 .mknod
= xfs_vn_mknod
,
756 .rename
= xfs_vn_rename
,
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
,
766 static const struct inode_operations xfs_symlink_inode_operations
= {
767 .readlink
= generic_readlink
,
768 .follow_link
= xfs_vn_follow_link
,
769 .put_link
= xfs_vn_put_link
,
770 .check_acl
= xfs_check_acl
,
771 .getattr
= xfs_vn_getattr
,
772 .setattr
= xfs_vn_setattr
,
773 .setxattr
= generic_setxattr
,
774 .getxattr
= generic_getxattr
,
775 .removexattr
= generic_removexattr
,
776 .listxattr
= xfs_vn_listxattr
,
780 xfs_diflags_to_iflags(
782 struct xfs_inode
*ip
)
784 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
785 inode
->i_flags
|= S_IMMUTABLE
;
787 inode
->i_flags
&= ~S_IMMUTABLE
;
788 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
789 inode
->i_flags
|= S_APPEND
;
791 inode
->i_flags
&= ~S_APPEND
;
792 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
793 inode
->i_flags
|= S_SYNC
;
795 inode
->i_flags
&= ~S_SYNC
;
796 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
797 inode
->i_flags
|= S_NOATIME
;
799 inode
->i_flags
&= ~S_NOATIME
;
803 * Initialize the Linux inode, set up the operation vectors and
806 * When reading existing inodes from disk this is called directly
807 * from xfs_iget, when creating a new inode it is called from
808 * xfs_ialloc after setting up the inode.
810 * We are always called with an uninitialised linux inode here.
811 * We need to initialise the necessary fields and take a reference
816 struct xfs_inode
*ip
)
818 struct inode
*inode
= &ip
->i_vnode
;
820 inode
->i_ino
= ip
->i_ino
;
821 inode
->i_state
= I_NEW
;
822 inode_add_to_lists(ip
->i_mount
->m_super
, inode
);
824 inode
->i_mode
= ip
->i_d
.di_mode
;
825 inode
->i_nlink
= ip
->i_d
.di_nlink
;
826 inode
->i_uid
= ip
->i_d
.di_uid
;
827 inode
->i_gid
= ip
->i_d
.di_gid
;
829 switch (inode
->i_mode
& S_IFMT
) {
833 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
834 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
841 inode
->i_generation
= ip
->i_d
.di_gen
;
842 i_size_write(inode
, ip
->i_d
.di_size
);
843 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
844 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
845 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
846 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
847 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
848 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
849 xfs_diflags_to_iflags(inode
, ip
);
851 switch (inode
->i_mode
& S_IFMT
) {
853 inode
->i_op
= &xfs_inode_operations
;
854 inode
->i_fop
= &xfs_file_operations
;
855 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
858 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
859 inode
->i_op
= &xfs_dir_ci_inode_operations
;
861 inode
->i_op
= &xfs_dir_inode_operations
;
862 inode
->i_fop
= &xfs_dir_file_operations
;
865 inode
->i_op
= &xfs_symlink_inode_operations
;
866 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
867 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
870 inode
->i_op
= &xfs_inode_operations
;
871 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
875 xfs_iflags_clear(ip
, XFS_INEW
);
878 unlock_new_inode(inode
);