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"
51 #include <linux/capability.h>
52 #include <linux/xattr.h>
53 #include <linux/namei.h>
54 #include <linux/posix_acl.h>
55 #include <linux/security.h>
56 #include <linux/falloc.h>
57 #include <linux/fiemap.h>
60 * Bring the timestamps in the XFS inode uptodate.
62 * Used before writing the inode to disk.
65 xfs_synchronize_times(
68 struct inode
*inode
= VFS_I(ip
);
70 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
71 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
72 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)inode
->i_ctime
.tv_sec
;
73 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)inode
->i_ctime
.tv_nsec
;
74 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)inode
->i_mtime
.tv_sec
;
75 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)inode
->i_mtime
.tv_nsec
;
79 * If the linux inode is valid, mark it dirty.
80 * Used when commiting a dirty inode into a transaction so that
81 * the inode will get written back by the linux code
84 xfs_mark_inode_dirty_sync(
87 struct inode
*inode
= VFS_I(ip
);
89 if (!(inode
->i_state
& (I_WILL_FREE
|I_FREEING
|I_CLEAR
)))
90 mark_inode_dirty_sync(inode
);
94 * Change the requested timestamp in the given inode.
95 * We don't lock across timestamp updates, and we don't log them but
96 * we do record the fact that there is dirty information in core.
103 struct inode
*inode
= VFS_I(ip
);
107 tv
= current_fs_time(inode
->i_sb
);
109 if ((flags
& XFS_ICHGTIME_MOD
) &&
110 !timespec_equal(&inode
->i_mtime
, &tv
)) {
114 if ((flags
& XFS_ICHGTIME_CHG
) &&
115 !timespec_equal(&inode
->i_ctime
, &tv
)) {
121 * Update complete - now make sure everyone knows that the inode
125 xfs_mark_inode_dirty_sync(ip
);
129 * Hook in SELinux. This is not quite correct yet, what we really need
130 * here (as we do for default ACLs) is a mechanism by which creation of
131 * these attrs can be journalled at inode creation time (along with the
132 * inode, of course, such that log replay can't cause these to be lost).
139 struct xfs_inode
*ip
= XFS_I(inode
);
145 error
= security_inode_init_security(inode
, dir
, &name
,
148 if (error
== -EOPNOTSUPP
)
153 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
162 struct xfs_name
*namep
,
163 struct dentry
*dentry
)
165 namep
->name
= dentry
->d_name
.name
;
166 namep
->len
= dentry
->d_name
.len
;
173 struct dentry
*dentry
)
175 struct xfs_name teardown
;
178 * If we can't add the ACL or we fail in
179 * xfs_init_security we must back out.
180 * ENOSPC can hit here, among other things.
182 xfs_dentry_to_name(&teardown
, dentry
);
184 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
191 struct dentry
*dentry
,
196 struct xfs_inode
*ip
= NULL
;
197 struct posix_acl
*default_acl
= NULL
;
198 struct xfs_name name
;
202 * Irix uses Missed'em'V split, but doesn't want to see
203 * the upper 5 bits of (14bit) major.
205 if (S_ISCHR(mode
) || S_ISBLK(mode
)) {
206 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
208 rdev
= sysv_encode_dev(rdev
);
213 if (IS_POSIXACL(dir
)) {
214 default_acl
= xfs_get_acl(dir
, ACL_TYPE_DEFAULT
);
215 if (IS_ERR(default_acl
))
216 return -PTR_ERR(default_acl
);
219 mode
&= ~current_umask();
222 xfs_dentry_to_name(&name
, dentry
);
223 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
229 error
= xfs_init_security(inode
, dir
);
231 goto out_cleanup_inode
;
234 error
= -xfs_inherit_acl(inode
, default_acl
);
236 goto out_cleanup_inode
;
237 posix_acl_release(default_acl
);
241 d_instantiate(dentry
, inode
);
245 xfs_cleanup_inode(dir
, inode
, dentry
);
247 posix_acl_release(default_acl
);
254 struct dentry
*dentry
,
256 struct nameidata
*nd
)
258 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
264 struct dentry
*dentry
,
267 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
270 STATIC
struct dentry
*
273 struct dentry
*dentry
,
274 struct nameidata
*nd
)
276 struct xfs_inode
*cip
;
277 struct xfs_name name
;
280 if (dentry
->d_name
.len
>= MAXNAMELEN
)
281 return ERR_PTR(-ENAMETOOLONG
);
283 xfs_dentry_to_name(&name
, dentry
);
284 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
285 if (unlikely(error
)) {
286 if (unlikely(error
!= ENOENT
))
287 return ERR_PTR(-error
);
292 return d_splice_alias(VFS_I(cip
), dentry
);
295 STATIC
struct dentry
*
298 struct dentry
*dentry
,
299 struct nameidata
*nd
)
301 struct xfs_inode
*ip
;
302 struct xfs_name xname
;
303 struct xfs_name ci_name
;
307 if (dentry
->d_name
.len
>= MAXNAMELEN
)
308 return ERR_PTR(-ENAMETOOLONG
);
310 xfs_dentry_to_name(&xname
, dentry
);
311 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
312 if (unlikely(error
)) {
313 if (unlikely(error
!= ENOENT
))
314 return ERR_PTR(-error
);
316 * call d_add(dentry, NULL) here when d_drop_negative_children
317 * is called in xfs_vn_mknod (ie. allow negative dentries
318 * with CI filesystems).
323 /* if exact match, just splice and exit */
325 return d_splice_alias(VFS_I(ip
), dentry
);
327 /* else case-insensitive match... */
328 dname
.name
= ci_name
.name
;
329 dname
.len
= ci_name
.len
;
330 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
331 kmem_free(ci_name
.name
);
337 struct dentry
*old_dentry
,
339 struct dentry
*dentry
)
341 struct inode
*inode
= old_dentry
->d_inode
;
342 struct xfs_name name
;
345 xfs_dentry_to_name(&name
, dentry
);
347 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
351 atomic_inc(&inode
->i_count
);
352 d_instantiate(dentry
, inode
);
359 struct dentry
*dentry
)
361 struct xfs_name name
;
364 xfs_dentry_to_name(&name
, dentry
);
366 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
371 * With unlink, the VFS makes the dentry "negative": no inode,
372 * but still hashed. This is incompatible with case-insensitive
373 * mode, so invalidate (unhash) the dentry in CI-mode.
375 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
376 d_invalidate(dentry
);
383 struct dentry
*dentry
,
387 struct xfs_inode
*cip
= NULL
;
388 struct xfs_name name
;
393 (irix_symlink_mode
? 0777 & ~current_umask() : S_IRWXUGO
);
394 xfs_dentry_to_name(&name
, dentry
);
396 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
402 error
= xfs_init_security(inode
, dir
);
404 goto out_cleanup_inode
;
406 d_instantiate(dentry
, inode
);
410 xfs_cleanup_inode(dir
, inode
, dentry
);
418 struct dentry
*odentry
,
420 struct dentry
*ndentry
)
422 struct inode
*new_inode
= ndentry
->d_inode
;
423 struct xfs_name oname
;
424 struct xfs_name nname
;
426 xfs_dentry_to_name(&oname
, odentry
);
427 xfs_dentry_to_name(&nname
, ndentry
);
429 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
430 XFS_I(ndir
), &nname
, new_inode
?
431 XFS_I(new_inode
) : NULL
);
435 * careful here - this function can get called recursively, so
436 * we need to be very careful about how much stack we use.
437 * uio is kmalloced for this reason...
441 struct dentry
*dentry
,
442 struct nameidata
*nd
)
447 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
451 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
455 nd_set_link(nd
, link
);
461 nd_set_link(nd
, ERR_PTR(error
));
467 struct dentry
*dentry
,
468 struct nameidata
*nd
,
471 char *s
= nd_get_link(nd
);
479 struct vfsmount
*mnt
,
480 struct dentry
*dentry
,
483 struct inode
*inode
= dentry
->d_inode
;
484 struct xfs_inode
*ip
= XFS_I(inode
);
485 struct xfs_mount
*mp
= ip
->i_mount
;
487 xfs_itrace_entry(ip
);
489 if (XFS_FORCED_SHUTDOWN(mp
))
490 return XFS_ERROR(EIO
);
492 stat
->size
= XFS_ISIZE(ip
);
493 stat
->dev
= inode
->i_sb
->s_dev
;
494 stat
->mode
= ip
->i_d
.di_mode
;
495 stat
->nlink
= ip
->i_d
.di_nlink
;
496 stat
->uid
= ip
->i_d
.di_uid
;
497 stat
->gid
= ip
->i_d
.di_gid
;
498 stat
->ino
= ip
->i_ino
;
499 stat
->atime
= inode
->i_atime
;
500 stat
->mtime
= inode
->i_mtime
;
501 stat
->ctime
= inode
->i_ctime
;
503 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
506 switch (inode
->i_mode
& S_IFMT
) {
509 stat
->blksize
= BLKDEV_IOSIZE
;
510 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
511 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
514 if (XFS_IS_REALTIME_INODE(ip
)) {
516 * If the file blocks are being allocated from a
517 * realtime volume, then return the inode's realtime
518 * extent size or the realtime volume's extent size.
521 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
523 stat
->blksize
= xfs_preferred_iosize(mp
);
533 struct dentry
*dentry
,
536 return -xfs_setattr(XFS_I(dentry
->d_inode
), iattr
, 0);
540 * block_truncate_page can return an error, but we can't propagate it
541 * at all here. Leave a complaint + stack trace in the syslog because
542 * this could be bad. If it is bad, we need to propagate the error further.
549 error
= block_truncate_page(inode
->i_mapping
, inode
->i_size
,
564 xfs_inode_t
*ip
= XFS_I(inode
);
566 /* preallocation on directories not yet supported */
568 if (S_ISDIR(inode
->i_mode
))
575 xfs_ilock(ip
, XFS_IOLOCK_EXCL
);
576 error
= xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
578 if (!error
&& !(mode
& FALLOC_FL_KEEP_SIZE
) &&
579 offset
+ len
> i_size_read(inode
))
580 new_size
= offset
+ len
;
582 /* Change file size if needed */
586 iattr
.ia_valid
= ATTR_SIZE
;
587 iattr
.ia_size
= new_size
;
588 error
= xfs_setattr(ip
, &iattr
, XFS_ATTR_NOLOCK
);
591 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
596 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
599 * Call fiemap helper to fill in user data.
600 * Returns positive errors to xfs_getbmap.
605 struct getbmapx
*bmv
,
609 struct fiemap_extent_info
*fieinfo
= *arg
;
610 u32 fiemap_flags
= 0;
611 u64 logical
, physical
, length
;
613 /* Do nothing for a hole */
614 if (bmv
->bmv_block
== -1LL)
617 logical
= BBTOB(bmv
->bmv_offset
);
618 physical
= BBTOB(bmv
->bmv_block
);
619 length
= BBTOB(bmv
->bmv_length
);
621 if (bmv
->bmv_oflags
& BMV_OF_PREALLOC
)
622 fiemap_flags
|= FIEMAP_EXTENT_UNWRITTEN
;
623 else if (bmv
->bmv_oflags
& BMV_OF_DELALLOC
) {
624 fiemap_flags
|= FIEMAP_EXTENT_DELALLOC
;
625 physical
= 0; /* no block yet */
627 if (bmv
->bmv_oflags
& BMV_OF_LAST
)
628 fiemap_flags
|= FIEMAP_EXTENT_LAST
;
630 error
= fiemap_fill_next_extent(fieinfo
, logical
, physical
,
631 length
, fiemap_flags
);
634 *full
= 1; /* user array now full */
643 struct fiemap_extent_info
*fieinfo
,
647 xfs_inode_t
*ip
= XFS_I(inode
);
651 error
= fiemap_check_flags(fieinfo
, XFS_FIEMAP_FLAGS
);
655 /* Set up bmap header for xfs internal routine */
656 bm
.bmv_offset
= BTOBB(start
);
657 /* Special case for whole file */
658 if (length
== FIEMAP_MAX_OFFSET
)
659 bm
.bmv_length
= -1LL;
661 bm
.bmv_length
= BTOBB(length
);
663 /* We add one because in getbmap world count includes the header */
664 bm
.bmv_count
= fieinfo
->fi_extents_max
+ 1;
665 bm
.bmv_iflags
= BMV_IF_PREALLOC
;
666 if (fieinfo
->fi_flags
& FIEMAP_FLAG_XATTR
)
667 bm
.bmv_iflags
|= BMV_IF_ATTRFORK
;
668 if (!(fieinfo
->fi_flags
& FIEMAP_FLAG_SYNC
))
669 bm
.bmv_iflags
|= BMV_IF_DELALLOC
;
671 error
= xfs_getbmap(ip
, &bm
, xfs_fiemap_format
, fieinfo
);
678 static const struct inode_operations xfs_inode_operations
= {
679 .check_acl
= xfs_check_acl
,
680 .truncate
= xfs_vn_truncate
,
681 .getattr
= xfs_vn_getattr
,
682 .setattr
= xfs_vn_setattr
,
683 .setxattr
= generic_setxattr
,
684 .getxattr
= generic_getxattr
,
685 .removexattr
= generic_removexattr
,
686 .listxattr
= xfs_vn_listxattr
,
687 .fallocate
= xfs_vn_fallocate
,
688 .fiemap
= xfs_vn_fiemap
,
691 static const struct inode_operations xfs_dir_inode_operations
= {
692 .create
= xfs_vn_create
,
693 .lookup
= xfs_vn_lookup
,
695 .unlink
= xfs_vn_unlink
,
696 .symlink
= xfs_vn_symlink
,
697 .mkdir
= xfs_vn_mkdir
,
699 * Yes, XFS uses the same method for rmdir and unlink.
701 * There are some subtile differences deeper in the code,
702 * but we use S_ISDIR to check for those.
704 .rmdir
= xfs_vn_unlink
,
705 .mknod
= xfs_vn_mknod
,
706 .rename
= xfs_vn_rename
,
707 .check_acl
= xfs_check_acl
,
708 .getattr
= xfs_vn_getattr
,
709 .setattr
= xfs_vn_setattr
,
710 .setxattr
= generic_setxattr
,
711 .getxattr
= generic_getxattr
,
712 .removexattr
= generic_removexattr
,
713 .listxattr
= xfs_vn_listxattr
,
716 static const struct inode_operations xfs_dir_ci_inode_operations
= {
717 .create
= xfs_vn_create
,
718 .lookup
= xfs_vn_ci_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_symlink_inode_operations
= {
742 .readlink
= generic_readlink
,
743 .follow_link
= xfs_vn_follow_link
,
744 .put_link
= xfs_vn_put_link
,
745 .check_acl
= xfs_check_acl
,
746 .getattr
= xfs_vn_getattr
,
747 .setattr
= xfs_vn_setattr
,
748 .setxattr
= generic_setxattr
,
749 .getxattr
= generic_getxattr
,
750 .removexattr
= generic_removexattr
,
751 .listxattr
= xfs_vn_listxattr
,
755 xfs_diflags_to_iflags(
757 struct xfs_inode
*ip
)
759 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
760 inode
->i_flags
|= S_IMMUTABLE
;
762 inode
->i_flags
&= ~S_IMMUTABLE
;
763 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
764 inode
->i_flags
|= S_APPEND
;
766 inode
->i_flags
&= ~S_APPEND
;
767 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
768 inode
->i_flags
|= S_SYNC
;
770 inode
->i_flags
&= ~S_SYNC
;
771 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
772 inode
->i_flags
|= S_NOATIME
;
774 inode
->i_flags
&= ~S_NOATIME
;
778 * Initialize the Linux inode, set up the operation vectors and
781 * When reading existing inodes from disk this is called directly
782 * from xfs_iget, when creating a new inode it is called from
783 * xfs_ialloc after setting up the inode.
785 * We are always called with an uninitialised linux inode here.
786 * We need to initialise the necessary fields and take a reference
791 struct xfs_inode
*ip
)
793 struct inode
*inode
= &ip
->i_vnode
;
795 inode
->i_ino
= ip
->i_ino
;
796 inode
->i_state
= I_NEW
|I_LOCK
;
797 inode_add_to_lists(ip
->i_mount
->m_super
, inode
);
799 inode
->i_mode
= ip
->i_d
.di_mode
;
800 inode
->i_nlink
= ip
->i_d
.di_nlink
;
801 inode
->i_uid
= ip
->i_d
.di_uid
;
802 inode
->i_gid
= ip
->i_d
.di_gid
;
804 switch (inode
->i_mode
& S_IFMT
) {
808 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
809 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
816 inode
->i_generation
= ip
->i_d
.di_gen
;
817 i_size_write(inode
, ip
->i_d
.di_size
);
818 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
819 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
820 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
821 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
822 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
823 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
824 xfs_diflags_to_iflags(inode
, ip
);
826 switch (inode
->i_mode
& S_IFMT
) {
828 inode
->i_op
= &xfs_inode_operations
;
829 inode
->i_fop
= &xfs_file_operations
;
830 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
833 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
834 inode
->i_op
= &xfs_dir_ci_inode_operations
;
836 inode
->i_op
= &xfs_dir_inode_operations
;
837 inode
->i_fop
= &xfs_dir_file_operations
;
840 inode
->i_op
= &xfs_symlink_inode_operations
;
841 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
842 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
845 inode
->i_op
= &xfs_inode_operations
;
846 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
850 xfs_iflags_clear(ip
, XFS_INEW
);
853 unlock_new_inode(inode
);