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
23 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
39 #include "xfs_btree.h"
40 #include "xfs_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #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/security.h>
55 #include <linux/falloc.h>
58 * Bring the atime in the XFS inode uptodate.
59 * Used before logging the inode to disk or when the Linux inode goes away.
62 xfs_synchronize_atime(
65 struct inode
*inode
= VFS_I(ip
);
68 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
69 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
74 * If the linux inode exists, mark it dirty.
75 * Used when commiting a dirty inode into a transaction so that
76 * the inode will get written back by the linux code
79 xfs_mark_inode_dirty_sync(
82 struct inode
*inode
= VFS_I(ip
);
85 mark_inode_dirty_sync(inode
);
89 * Change the requested timestamp in the given inode.
90 * We don't lock across timestamp updates, and we don't log them but
91 * we do record the fact that there is dirty information in core.
98 struct inode
*inode
= VFS_I(ip
);
102 tv
= current_fs_time(inode
->i_sb
);
104 if ((flags
& XFS_ICHGTIME_MOD
) &&
105 !timespec_equal(&inode
->i_mtime
, &tv
)) {
107 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)tv
.tv_sec
;
108 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
111 if ((flags
& XFS_ICHGTIME_CHG
) &&
112 !timespec_equal(&inode
->i_ctime
, &tv
)) {
114 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tv
.tv_sec
;
115 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
120 * We update the i_update_core field _after_ changing
121 * the timestamps in order to coordinate properly with
122 * xfs_iflush() so that we don't lose timestamp updates.
123 * This keeps us from having to hold the inode lock
124 * while doing this. We use the SYNCHRONIZE macro to
125 * ensure that the compiler does not reorder the update
126 * of i_update_core above the timestamp updates above.
130 ip
->i_update_core
= 1;
131 mark_inode_dirty_sync(inode
);
136 * Hook in SELinux. This is not quite correct yet, what we really need
137 * here (as we do for default ACLs) is a mechanism by which creation of
138 * these attrs can be journalled at inode creation time (along with the
139 * inode, of course, such that log replay can't cause these to be lost).
146 struct xfs_inode
*ip
= XFS_I(inode
);
152 error
= security_inode_init_security(inode
, dir
, &name
,
155 if (error
== -EOPNOTSUPP
)
160 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
162 xfs_iflags_set(ip
, XFS_IMODIFIED
);
171 struct xfs_name
*namep
,
172 struct dentry
*dentry
)
174 namep
->name
= dentry
->d_name
.name
;
175 namep
->len
= dentry
->d_name
.len
;
182 struct dentry
*dentry
)
184 struct xfs_name teardown
;
187 * If we can't add the ACL or we fail in
188 * xfs_init_security we must back out.
189 * ENOSPC can hit here, among other things.
191 xfs_dentry_to_name(&teardown
, dentry
);
193 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
200 struct dentry
*dentry
,
205 struct xfs_inode
*ip
= NULL
;
206 xfs_acl_t
*default_acl
= NULL
;
207 struct xfs_name name
;
208 int (*test_default_acl
)(struct inode
*) = _ACL_DEFAULT_EXISTS
;
212 * Irix uses Missed'em'V split, but doesn't want to see
213 * the upper 5 bits of (14bit) major.
215 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
218 if (test_default_acl
&& test_default_acl(dir
)) {
219 if (!_ACL_ALLOC(default_acl
)) {
222 if (!_ACL_GET_DEFAULT(dir
, default_acl
)) {
223 _ACL_FREE(default_acl
);
228 xfs_dentry_to_name(&name
, dentry
);
230 if (IS_POSIXACL(dir
) && !default_acl
)
231 mode
&= ~current
->fs
->umask
;
233 switch (mode
& S_IFMT
) {
238 rdev
= sysv_encode_dev(rdev
);
240 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
243 error
= xfs_mkdir(XFS_I(dir
), &name
, mode
, &ip
, NULL
);
255 error
= xfs_init_security(inode
, dir
);
257 goto out_cleanup_inode
;
260 error
= _ACL_INHERIT(inode
, mode
, default_acl
);
262 goto out_cleanup_inode
;
263 xfs_iflags_set(ip
, XFS_IMODIFIED
);
264 _ACL_FREE(default_acl
);
268 d_instantiate(dentry
, inode
);
272 xfs_cleanup_inode(dir
, inode
, dentry
);
275 _ACL_FREE(default_acl
);
282 struct dentry
*dentry
,
284 struct nameidata
*nd
)
286 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
292 struct dentry
*dentry
,
295 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
298 STATIC
struct dentry
*
301 struct dentry
*dentry
,
302 struct nameidata
*nd
)
304 struct xfs_inode
*cip
;
305 struct xfs_name name
;
308 if (dentry
->d_name
.len
>= MAXNAMELEN
)
309 return ERR_PTR(-ENAMETOOLONG
);
311 xfs_dentry_to_name(&name
, dentry
);
312 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
313 if (unlikely(error
)) {
314 if (unlikely(error
!= ENOENT
))
315 return ERR_PTR(-error
);
320 return d_splice_alias(VFS_I(cip
), dentry
);
323 STATIC
struct dentry
*
326 struct dentry
*dentry
,
327 struct nameidata
*nd
)
329 struct xfs_inode
*ip
;
330 struct xfs_name xname
;
331 struct xfs_name ci_name
;
335 if (dentry
->d_name
.len
>= MAXNAMELEN
)
336 return ERR_PTR(-ENAMETOOLONG
);
338 xfs_dentry_to_name(&xname
, dentry
);
339 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
340 if (unlikely(error
)) {
341 if (unlikely(error
!= ENOENT
))
342 return ERR_PTR(-error
);
344 * call d_add(dentry, NULL) here when d_drop_negative_children
345 * is called in xfs_vn_mknod (ie. allow negative dentries
346 * with CI filesystems).
351 /* if exact match, just splice and exit */
353 return d_splice_alias(VFS_I(ip
), dentry
);
355 /* else case-insensitive match... */
356 dname
.name
= ci_name
.name
;
357 dname
.len
= ci_name
.len
;
358 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
359 kmem_free(ci_name
.name
);
365 struct dentry
*old_dentry
,
367 struct dentry
*dentry
)
369 struct inode
*inode
; /* inode of guy being linked to */
370 struct xfs_name name
;
373 inode
= old_dentry
->d_inode
;
374 xfs_dentry_to_name(&name
, dentry
);
377 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
378 if (unlikely(error
)) {
383 xfs_iflags_set(XFS_I(dir
), XFS_IMODIFIED
);
384 d_instantiate(dentry
, inode
);
391 struct dentry
*dentry
)
393 struct xfs_name name
;
396 xfs_dentry_to_name(&name
, dentry
);
398 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
403 * With unlink, the VFS makes the dentry "negative": no inode,
404 * but still hashed. This is incompatible with case-insensitive
405 * mode, so invalidate (unhash) the dentry in CI-mode.
407 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
408 d_invalidate(dentry
);
415 struct dentry
*dentry
,
419 struct xfs_inode
*cip
= NULL
;
420 struct xfs_name name
;
425 (irix_symlink_mode
? 0777 & ~current
->fs
->umask
: S_IRWXUGO
);
426 xfs_dentry_to_name(&name
, dentry
);
428 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
434 error
= xfs_init_security(inode
, dir
);
436 goto out_cleanup_inode
;
438 d_instantiate(dentry
, inode
);
442 xfs_cleanup_inode(dir
, inode
, dentry
);
450 struct dentry
*odentry
,
452 struct dentry
*ndentry
)
454 struct inode
*new_inode
= ndentry
->d_inode
;
455 struct xfs_name oname
;
456 struct xfs_name nname
;
458 xfs_dentry_to_name(&oname
, odentry
);
459 xfs_dentry_to_name(&nname
, ndentry
);
461 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
462 XFS_I(ndir
), &nname
, new_inode
?
463 XFS_I(new_inode
) : NULL
);
467 * careful here - this function can get called recursively, so
468 * we need to be very careful about how much stack we use.
469 * uio is kmalloced for this reason...
473 struct dentry
*dentry
,
474 struct nameidata
*nd
)
479 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
483 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
487 nd_set_link(nd
, link
);
493 nd_set_link(nd
, ERR_PTR(error
));
499 struct dentry
*dentry
,
500 struct nameidata
*nd
,
503 char *s
= nd_get_link(nd
);
509 #ifdef CONFIG_XFS_POSIX_ACL
515 struct xfs_inode
*ip
= XFS_I(inode
);
518 xfs_itrace_entry(ip
);
520 if (XFS_IFORK_Q(ip
)) {
521 error
= xfs_acl_iaccess(ip
, mask
, NULL
);
534 return generic_permission(inode
, mask
, xfs_check_acl
);
537 #define xfs_vn_permission NULL
542 struct vfsmount
*mnt
,
543 struct dentry
*dentry
,
546 struct inode
*inode
= dentry
->d_inode
;
547 struct xfs_inode
*ip
= XFS_I(inode
);
548 struct xfs_mount
*mp
= ip
->i_mount
;
550 xfs_itrace_entry(ip
);
552 if (XFS_FORCED_SHUTDOWN(mp
))
553 return XFS_ERROR(EIO
);
555 stat
->size
= XFS_ISIZE(ip
);
556 stat
->dev
= inode
->i_sb
->s_dev
;
557 stat
->mode
= ip
->i_d
.di_mode
;
558 stat
->nlink
= ip
->i_d
.di_nlink
;
559 stat
->uid
= ip
->i_d
.di_uid
;
560 stat
->gid
= ip
->i_d
.di_gid
;
561 stat
->ino
= ip
->i_ino
;
563 stat
->ino
+= mp
->m_inoadd
;
565 stat
->atime
= inode
->i_atime
;
566 stat
->mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
567 stat
->mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
568 stat
->ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
569 stat
->ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
571 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
574 switch (inode
->i_mode
& S_IFMT
) {
577 stat
->blksize
= BLKDEV_IOSIZE
;
578 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
579 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
582 if (XFS_IS_REALTIME_INODE(ip
)) {
584 * If the file blocks are being allocated from a
585 * realtime volume, then return the inode's realtime
586 * extent size or the realtime volume's extent size.
589 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
591 stat
->blksize
= xfs_preferred_iosize(mp
);
601 struct dentry
*dentry
,
604 return -xfs_setattr(XFS_I(dentry
->d_inode
), iattr
, 0, NULL
);
608 * block_truncate_page can return an error, but we can't propagate it
609 * at all here. Leave a complaint + stack trace in the syslog because
610 * this could be bad. If it is bad, we need to propagate the error further.
617 error
= block_truncate_page(inode
->i_mapping
, inode
->i_size
,
632 xfs_inode_t
*ip
= XFS_I(inode
);
634 /* preallocation on directories not yet supported */
636 if (S_ISDIR(inode
->i_mode
))
643 xfs_ilock(ip
, XFS_IOLOCK_EXCL
);
644 error
= xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
645 0, NULL
, XFS_ATTR_NOLOCK
);
646 if (!error
&& !(mode
& FALLOC_FL_KEEP_SIZE
) &&
647 offset
+ len
> i_size_read(inode
))
648 new_size
= offset
+ len
;
650 /* Change file size if needed */
654 iattr
.ia_valid
= ATTR_SIZE
;
655 iattr
.ia_size
= new_size
;
656 error
= xfs_setattr(ip
, &iattr
, XFS_ATTR_NOLOCK
, NULL
);
659 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
664 static const struct inode_operations xfs_inode_operations
= {
665 .permission
= xfs_vn_permission
,
666 .truncate
= xfs_vn_truncate
,
667 .getattr
= xfs_vn_getattr
,
668 .setattr
= xfs_vn_setattr
,
669 .setxattr
= generic_setxattr
,
670 .getxattr
= generic_getxattr
,
671 .removexattr
= generic_removexattr
,
672 .listxattr
= xfs_vn_listxattr
,
673 .fallocate
= xfs_vn_fallocate
,
676 static const struct inode_operations xfs_dir_inode_operations
= {
677 .create
= xfs_vn_create
,
678 .lookup
= xfs_vn_lookup
,
680 .unlink
= xfs_vn_unlink
,
681 .symlink
= xfs_vn_symlink
,
682 .mkdir
= xfs_vn_mkdir
,
684 * Yes, XFS uses the same method for rmdir and unlink.
686 * There are some subtile differences deeper in the code,
687 * but we use S_ISDIR to check for those.
689 .rmdir
= xfs_vn_unlink
,
690 .mknod
= xfs_vn_mknod
,
691 .rename
= xfs_vn_rename
,
692 .permission
= xfs_vn_permission
,
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
,
701 static const struct inode_operations xfs_dir_ci_inode_operations
= {
702 .create
= xfs_vn_create
,
703 .lookup
= xfs_vn_ci_lookup
,
705 .unlink
= xfs_vn_unlink
,
706 .symlink
= xfs_vn_symlink
,
707 .mkdir
= xfs_vn_mkdir
,
709 * Yes, XFS uses the same method for rmdir and unlink.
711 * There are some subtile differences deeper in the code,
712 * but we use S_ISDIR to check for those.
714 .rmdir
= xfs_vn_unlink
,
715 .mknod
= xfs_vn_mknod
,
716 .rename
= xfs_vn_rename
,
717 .permission
= xfs_vn_permission
,
718 .getattr
= xfs_vn_getattr
,
719 .setattr
= xfs_vn_setattr
,
720 .setxattr
= generic_setxattr
,
721 .getxattr
= generic_getxattr
,
722 .removexattr
= generic_removexattr
,
723 .listxattr
= xfs_vn_listxattr
,
726 static const struct inode_operations xfs_symlink_inode_operations
= {
727 .readlink
= generic_readlink
,
728 .follow_link
= xfs_vn_follow_link
,
729 .put_link
= xfs_vn_put_link
,
730 .permission
= xfs_vn_permission
,
731 .getattr
= xfs_vn_getattr
,
732 .setattr
= xfs_vn_setattr
,
733 .setxattr
= generic_setxattr
,
734 .getxattr
= generic_getxattr
,
735 .removexattr
= generic_removexattr
,
736 .listxattr
= xfs_vn_listxattr
,
740 xfs_diflags_to_iflags(
742 struct xfs_inode
*ip
)
744 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
745 inode
->i_flags
|= S_IMMUTABLE
;
747 inode
->i_flags
&= ~S_IMMUTABLE
;
748 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
749 inode
->i_flags
|= S_APPEND
;
751 inode
->i_flags
&= ~S_APPEND
;
752 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
753 inode
->i_flags
|= S_SYNC
;
755 inode
->i_flags
&= ~S_SYNC
;
756 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
757 inode
->i_flags
|= S_NOATIME
;
759 inode
->i_flags
&= ~S_NOATIME
;
763 * Initialize the Linux inode, set up the operation vectors and
766 * When reading existing inodes from disk this is called directly
767 * from xfs_iget, when creating a new inode it is called from
768 * xfs_ialloc after setting up the inode.
772 struct xfs_inode
*ip
)
774 struct inode
*inode
= ip
->i_vnode
;
776 inode
->i_mode
= ip
->i_d
.di_mode
;
777 inode
->i_nlink
= ip
->i_d
.di_nlink
;
778 inode
->i_uid
= ip
->i_d
.di_uid
;
779 inode
->i_gid
= ip
->i_d
.di_gid
;
781 switch (inode
->i_mode
& S_IFMT
) {
785 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
786 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
793 inode
->i_generation
= ip
->i_d
.di_gen
;
794 i_size_write(inode
, ip
->i_d
.di_size
);
795 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
796 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
797 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
798 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
799 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
800 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
801 xfs_diflags_to_iflags(inode
, ip
);
802 xfs_iflags_clear(ip
, XFS_IMODIFIED
);
804 switch (inode
->i_mode
& S_IFMT
) {
806 inode
->i_op
= &xfs_inode_operations
;
807 inode
->i_fop
= &xfs_file_operations
;
808 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
811 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
812 inode
->i_op
= &xfs_dir_ci_inode_operations
;
814 inode
->i_op
= &xfs_dir_inode_operations
;
815 inode
->i_fop
= &xfs_dir_file_operations
;
818 inode
->i_op
= &xfs_symlink_inode_operations
;
819 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
820 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
823 inode
->i_op
= &xfs_inode_operations
;
824 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
828 xfs_iflags_clear(ip
, XFS_INEW
);
831 unlock_new_inode(inode
);