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
= ip
->i_vnode
;
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
= ip
->i_vnode
;
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.
93 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
94 * with XFS_ICHGTIME_ACC to be sure that access time
95 * update will take. Calling first with XFS_ICHGTIME_ACC
96 * and then XFS_ICHGTIME_MOD may fail to modify the access
97 * timestamp if the filesystem is mounted noacctm.
104 struct inode
*inode
= vn_to_inode(XFS_ITOV(ip
));
108 if (flags
& XFS_ICHGTIME_MOD
) {
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
;
113 if (flags
& XFS_ICHGTIME_ACC
) {
115 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)tv
.tv_sec
;
116 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
118 if (flags
& XFS_ICHGTIME_CHG
) {
120 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tv
.tv_sec
;
121 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
125 * We update the i_update_core field _after_ changing
126 * the timestamps in order to coordinate properly with
127 * xfs_iflush() so that we don't lose timestamp updates.
128 * This keeps us from having to hold the inode lock
129 * while doing this. We use the SYNCHRONIZE macro to
130 * ensure that the compiler does not reorder the update
131 * of i_update_core above the timestamp updates above.
134 ip
->i_update_core
= 1;
135 if (!(inode
->i_state
& I_NEW
))
136 mark_inode_dirty_sync(inode
);
140 * Variant on the above which avoids querying the system clock
141 * in situations where we know the Linux inode timestamps have
142 * just been updated (and so we can update our inode cheaply).
153 * Atime updates for read() & friends are handled lazily now, and
154 * explicit updates must go through xfs_ichgtime()
156 ASSERT((flags
& XFS_ICHGTIME_ACC
) == 0);
158 if (flags
& XFS_ICHGTIME_MOD
) {
159 tvp
= &inode
->i_mtime
;
160 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)tvp
->tv_sec
;
161 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)tvp
->tv_nsec
;
163 if (flags
& XFS_ICHGTIME_CHG
) {
164 tvp
= &inode
->i_ctime
;
165 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tvp
->tv_sec
;
166 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tvp
->tv_nsec
;
170 * We update the i_update_core field _after_ changing
171 * the timestamps in order to coordinate properly with
172 * xfs_iflush() so that we don't lose timestamp updates.
173 * This keeps us from having to hold the inode lock
174 * while doing this. We use the SYNCHRONIZE macro to
175 * ensure that the compiler does not reorder the update
176 * of i_update_core above the timestamp updates above.
179 ip
->i_update_core
= 1;
180 if (!(inode
->i_state
& I_NEW
))
181 mark_inode_dirty_sync(inode
);
186 * Pull the link count and size up from the xfs inode to the linux inode
192 struct xfs_inode
*ip
= XFS_I(inode
);
195 /* we're under i_sem so i_size can't change under us */
196 size
= XFS_ISIZE(ip
);
197 if (i_size_read(inode
) != size
)
198 i_size_write(inode
, size
);
202 * Hook in SELinux. This is not quite correct yet, what we really need
203 * here (as we do for default ACLs) is a mechanism by which creation of
204 * these attrs can be journalled at inode creation time (along with the
205 * inode, of course, such that log replay can't cause these to be lost).
212 struct xfs_inode
*ip
= XFS_I(inode
);
218 error
= security_inode_init_security(inode
, dir
, &name
,
221 if (error
== -EOPNOTSUPP
)
226 error
= xfs_attr_set(ip
, name
, value
, length
, ATTR_SECURE
);
228 xfs_iflags_set(ip
, XFS_IMODIFIED
);
237 struct xfs_name
*namep
,
238 struct dentry
*dentry
)
240 namep
->name
= dentry
->d_name
.name
;
241 namep
->len
= dentry
->d_name
.len
;
248 struct dentry
*dentry
,
251 struct xfs_name teardown
;
254 * If we can't add the ACL or we fail in
255 * xfs_init_security we must back out.
256 * ENOSPC can hit here, among other things.
258 xfs_dentry_to_name(&teardown
, dentry
);
261 xfs_rmdir(XFS_I(dir
), &teardown
, XFS_I(inode
));
263 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
270 struct dentry
*dentry
,
275 struct xfs_inode
*ip
= NULL
;
276 xfs_acl_t
*default_acl
= NULL
;
277 struct xfs_name name
;
278 attrexists_t test_default_acl
= _ACL_DEFAULT_EXISTS
;
282 * Irix uses Missed'em'V split, but doesn't want to see
283 * the upper 5 bits of (14bit) major.
285 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
288 if (test_default_acl
&& test_default_acl(dir
)) {
289 if (!_ACL_ALLOC(default_acl
)) {
292 if (!_ACL_GET_DEFAULT(dir
, default_acl
)) {
293 _ACL_FREE(default_acl
);
298 xfs_dentry_to_name(&name
, dentry
);
300 if (IS_POSIXACL(dir
) && !default_acl
)
301 mode
&= ~current
->fs
->umask
;
303 switch (mode
& S_IFMT
) {
308 rdev
= sysv_encode_dev(rdev
);
310 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
, NULL
);
313 error
= xfs_mkdir(XFS_I(dir
), &name
, mode
, &ip
, NULL
);
325 error
= xfs_init_security(inode
, dir
);
327 goto out_cleanup_inode
;
330 error
= _ACL_INHERIT(inode
, mode
, default_acl
);
332 goto out_cleanup_inode
;
333 xfs_iflags_set(ip
, XFS_IMODIFIED
);
334 _ACL_FREE(default_acl
);
339 xfs_validate_fields(inode
);
340 d_instantiate(dentry
, inode
);
341 xfs_validate_fields(dir
);
345 xfs_cleanup_inode(dir
, inode
, dentry
, mode
);
348 _ACL_FREE(default_acl
);
355 struct dentry
*dentry
,
357 struct nameidata
*nd
)
359 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
365 struct dentry
*dentry
,
368 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
371 STATIC
struct dentry
*
374 struct dentry
*dentry
,
375 struct nameidata
*nd
)
377 struct xfs_inode
*cip
;
378 struct xfs_name name
;
381 if (dentry
->d_name
.len
>= MAXNAMELEN
)
382 return ERR_PTR(-ENAMETOOLONG
);
384 xfs_dentry_to_name(&name
, dentry
);
385 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
);
386 if (unlikely(error
)) {
387 if (unlikely(error
!= ENOENT
))
388 return ERR_PTR(-error
);
393 return d_splice_alias(cip
->i_vnode
, dentry
);
398 struct dentry
*old_dentry
,
400 struct dentry
*dentry
)
402 struct inode
*inode
; /* inode of guy being linked to */
403 struct xfs_name name
;
406 inode
= old_dentry
->d_inode
;
407 xfs_dentry_to_name(&name
, dentry
);
410 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
411 if (unlikely(error
)) {
416 xfs_iflags_set(XFS_I(dir
), XFS_IMODIFIED
);
417 xfs_validate_fields(inode
);
418 d_instantiate(dentry
, inode
);
425 struct dentry
*dentry
)
428 struct xfs_name name
;
431 inode
= dentry
->d_inode
;
432 xfs_dentry_to_name(&name
, dentry
);
434 error
= xfs_remove(XFS_I(dir
), &name
, XFS_I(inode
));
435 if (likely(!error
)) {
436 xfs_validate_fields(dir
); /* size needs update */
437 xfs_validate_fields(inode
);
445 struct dentry
*dentry
,
449 struct xfs_inode
*cip
= NULL
;
450 struct xfs_name name
;
455 (irix_symlink_mode
? 0777 & ~current
->fs
->umask
: S_IRWXUGO
);
456 xfs_dentry_to_name(&name
, dentry
);
458 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
, NULL
);
462 inode
= cip
->i_vnode
;
464 error
= xfs_init_security(inode
, dir
);
466 goto out_cleanup_inode
;
468 d_instantiate(dentry
, inode
);
469 xfs_validate_fields(dir
);
470 xfs_validate_fields(inode
);
474 xfs_cleanup_inode(dir
, inode
, dentry
, 0);
482 struct dentry
*dentry
)
484 struct inode
*inode
= dentry
->d_inode
;
485 struct xfs_name name
;
488 xfs_dentry_to_name(&name
, dentry
);
490 error
= xfs_rmdir(XFS_I(dir
), &name
, XFS_I(inode
));
491 if (likely(!error
)) {
492 xfs_validate_fields(inode
);
493 xfs_validate_fields(dir
);
501 struct dentry
*odentry
,
503 struct dentry
*ndentry
)
505 struct inode
*new_inode
= ndentry
->d_inode
;
506 struct xfs_name oname
;
507 struct xfs_name nname
;
510 xfs_dentry_to_name(&oname
, odentry
);
511 xfs_dentry_to_name(&nname
, ndentry
);
513 error
= xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
514 XFS_I(ndir
), &nname
, new_inode
?
515 XFS_I(new_inode
) : NULL
);
516 if (likely(!error
)) {
518 xfs_validate_fields(new_inode
);
519 xfs_validate_fields(odir
);
521 xfs_validate_fields(ndir
);
527 * careful here - this function can get called recursively, so
528 * we need to be very careful about how much stack we use.
529 * uio is kmalloced for this reason...
533 struct dentry
*dentry
,
534 struct nameidata
*nd
)
539 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
543 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
547 nd_set_link(nd
, link
);
553 nd_set_link(nd
, ERR_PTR(error
));
559 struct dentry
*dentry
,
560 struct nameidata
*nd
,
563 char *s
= nd_get_link(nd
);
569 #ifdef CONFIG_XFS_POSIX_ACL
575 struct xfs_inode
*ip
= XFS_I(inode
);
578 xfs_itrace_entry(ip
);
580 if (XFS_IFORK_Q(ip
)) {
581 error
= xfs_acl_iaccess(ip
, mask
, NULL
);
593 struct nameidata
*nd
)
595 return generic_permission(inode
, mask
, xfs_check_acl
);
598 #define xfs_vn_permission NULL
603 struct vfsmount
*mnt
,
604 struct dentry
*dentry
,
607 struct inode
*inode
= dentry
->d_inode
;
608 struct xfs_inode
*ip
= XFS_I(inode
);
609 struct xfs_mount
*mp
= ip
->i_mount
;
611 xfs_itrace_entry(ip
);
613 if (XFS_FORCED_SHUTDOWN(mp
))
614 return XFS_ERROR(EIO
);
616 stat
->size
= XFS_ISIZE(ip
);
617 stat
->dev
= inode
->i_sb
->s_dev
;
618 stat
->mode
= ip
->i_d
.di_mode
;
619 stat
->nlink
= ip
->i_d
.di_nlink
;
620 stat
->uid
= ip
->i_d
.di_uid
;
621 stat
->gid
= ip
->i_d
.di_gid
;
622 stat
->ino
= ip
->i_ino
;
624 stat
->ino
+= mp
->m_inoadd
;
626 stat
->atime
= inode
->i_atime
;
627 stat
->mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
628 stat
->mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
629 stat
->ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
630 stat
->ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
632 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
635 switch (inode
->i_mode
& S_IFMT
) {
638 stat
->blksize
= BLKDEV_IOSIZE
;
639 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
640 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
643 if (XFS_IS_REALTIME_INODE(ip
)) {
645 * If the file blocks are being allocated from a
646 * realtime volume, then return the inode's realtime
647 * extent size or the realtime volume's extent size.
650 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
652 stat
->blksize
= xfs_preferred_iosize(mp
);
662 struct dentry
*dentry
,
665 struct inode
*inode
= dentry
->d_inode
;
666 unsigned int ia_valid
= attr
->ia_valid
;
667 bhv_vattr_t vattr
= { 0 };
671 if (ia_valid
& ATTR_UID
) {
672 vattr
.va_mask
|= XFS_AT_UID
;
673 vattr
.va_uid
= attr
->ia_uid
;
675 if (ia_valid
& ATTR_GID
) {
676 vattr
.va_mask
|= XFS_AT_GID
;
677 vattr
.va_gid
= attr
->ia_gid
;
679 if (ia_valid
& ATTR_SIZE
) {
680 vattr
.va_mask
|= XFS_AT_SIZE
;
681 vattr
.va_size
= attr
->ia_size
;
683 if (ia_valid
& ATTR_ATIME
) {
684 vattr
.va_mask
|= XFS_AT_ATIME
;
685 vattr
.va_atime
= attr
->ia_atime
;
686 inode
->i_atime
= attr
->ia_atime
;
688 if (ia_valid
& ATTR_MTIME
) {
689 vattr
.va_mask
|= XFS_AT_MTIME
;
690 vattr
.va_mtime
= attr
->ia_mtime
;
692 if (ia_valid
& ATTR_CTIME
) {
693 vattr
.va_mask
|= XFS_AT_CTIME
;
694 vattr
.va_ctime
= attr
->ia_ctime
;
696 if (ia_valid
& ATTR_MODE
) {
697 vattr
.va_mask
|= XFS_AT_MODE
;
698 vattr
.va_mode
= attr
->ia_mode
;
699 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
700 inode
->i_mode
&= ~S_ISGID
;
703 if (ia_valid
& (ATTR_MTIME_SET
| ATTR_ATIME_SET
))
706 if ((ia_valid
& ATTR_NO_BLOCK
))
707 flags
|= ATTR_NONBLOCK
;
710 error
= xfs_setattr(XFS_I(inode
), &vattr
, flags
, NULL
);
712 vn_revalidate(vn_from_inode(inode
));
717 * block_truncate_page can return an error, but we can't propagate it
718 * at all here. Leave a complaint + stack trace in the syslog because
719 * this could be bad. If it is bad, we need to propagate the error further.
726 error
= block_truncate_page(inode
->i_mapping
, inode
->i_size
,
733 struct dentry
*dentry
,
739 bhv_vnode_t
*vp
= vn_from_inode(dentry
->d_inode
);
740 char *attr
= (char *)name
;
745 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
748 attr
+= namesp
->attr_namelen
;
749 error
= namesp
->attr_capable(vp
, NULL
);
753 /* Convert Linux syscall to XFS internal ATTR flags */
754 if (flags
& XATTR_CREATE
)
755 xflags
|= ATTR_CREATE
;
756 if (flags
& XATTR_REPLACE
)
757 xflags
|= ATTR_REPLACE
;
758 xflags
|= namesp
->attr_flag
;
759 return namesp
->attr_set(vp
, attr
, (void *)data
, size
, xflags
);
764 struct dentry
*dentry
,
769 bhv_vnode_t
*vp
= vn_from_inode(dentry
->d_inode
);
770 char *attr
= (char *)name
;
775 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
778 attr
+= namesp
->attr_namelen
;
779 error
= namesp
->attr_capable(vp
, NULL
);
783 /* Convert Linux syscall to XFS internal ATTR flags */
785 xflags
|= ATTR_KERNOVAL
;
788 xflags
|= namesp
->attr_flag
;
789 return namesp
->attr_get(vp
, attr
, (void *)data
, size
, xflags
);
794 struct dentry
*dentry
,
798 bhv_vnode_t
*vp
= vn_from_inode(dentry
->d_inode
);
799 int error
, xflags
= ATTR_KERNAMELS
;
803 xflags
|= ATTR_KERNOVAL
;
804 xflags
|= capable(CAP_SYS_ADMIN
) ? ATTR_KERNFULLS
: ATTR_KERNORMALS
;
806 error
= attr_generic_list(vp
, data
, size
, xflags
, &result
);
814 struct dentry
*dentry
,
817 bhv_vnode_t
*vp
= vn_from_inode(dentry
->d_inode
);
818 char *attr
= (char *)name
;
823 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
826 attr
+= namesp
->attr_namelen
;
827 error
= namesp
->attr_capable(vp
, NULL
);
830 xflags
|= namesp
->attr_flag
;
831 return namesp
->attr_remove(vp
, attr
, xflags
);
844 xfs_inode_t
*ip
= XFS_I(inode
);
846 /* preallocation on directories not yet supported */
848 if (S_ISDIR(inode
->i_mode
))
855 xfs_ilock(ip
, XFS_IOLOCK_EXCL
);
856 error
= xfs_change_file_space(ip
, XFS_IOC_RESVSP
, &bf
,
857 0, NULL
, ATTR_NOLOCK
);
858 if (!error
&& !(mode
& FALLOC_FL_KEEP_SIZE
) &&
859 offset
+ len
> i_size_read(inode
))
860 new_size
= offset
+ len
;
862 /* Change file size if needed */
866 va
.va_mask
= XFS_AT_SIZE
;
867 va
.va_size
= new_size
;
868 error
= xfs_setattr(ip
, &va
, ATTR_NOLOCK
, NULL
);
871 xfs_iunlock(ip
, XFS_IOLOCK_EXCL
);
876 const struct inode_operations xfs_inode_operations
= {
877 .permission
= xfs_vn_permission
,
878 .truncate
= xfs_vn_truncate
,
879 .getattr
= xfs_vn_getattr
,
880 .setattr
= xfs_vn_setattr
,
881 .setxattr
= xfs_vn_setxattr
,
882 .getxattr
= xfs_vn_getxattr
,
883 .listxattr
= xfs_vn_listxattr
,
884 .removexattr
= xfs_vn_removexattr
,
885 .fallocate
= xfs_vn_fallocate
,
888 const struct inode_operations xfs_dir_inode_operations
= {
889 .create
= xfs_vn_create
,
890 .lookup
= xfs_vn_lookup
,
892 .unlink
= xfs_vn_unlink
,
893 .symlink
= xfs_vn_symlink
,
894 .mkdir
= xfs_vn_mkdir
,
895 .rmdir
= xfs_vn_rmdir
,
896 .mknod
= xfs_vn_mknod
,
897 .rename
= xfs_vn_rename
,
898 .permission
= xfs_vn_permission
,
899 .getattr
= xfs_vn_getattr
,
900 .setattr
= xfs_vn_setattr
,
901 .setxattr
= xfs_vn_setxattr
,
902 .getxattr
= xfs_vn_getxattr
,
903 .listxattr
= xfs_vn_listxattr
,
904 .removexattr
= xfs_vn_removexattr
,
907 const struct inode_operations xfs_symlink_inode_operations
= {
908 .readlink
= generic_readlink
,
909 .follow_link
= xfs_vn_follow_link
,
910 .put_link
= xfs_vn_put_link
,
911 .permission
= xfs_vn_permission
,
912 .getattr
= xfs_vn_getattr
,
913 .setattr
= xfs_vn_setattr
,
914 .setxattr
= xfs_vn_setxattr
,
915 .getxattr
= xfs_vn_getxattr
,
916 .listxattr
= xfs_vn_listxattr
,
917 .removexattr
= xfs_vn_removexattr
,