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"
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_dir_sf.h"
36 #include "xfs_dir2_sf.h"
37 #include "xfs_attr_sf.h"
38 #include "xfs_dinode.h"
39 #include "xfs_inode.h"
41 #include "xfs_btree.h"
42 #include "xfs_ialloc.h"
43 #include "xfs_rtalloc.h"
44 #include "xfs_error.h"
45 #include "xfs_itable.h"
51 #include "xfs_buf_item.h"
52 #include "xfs_utils.h"
54 #include <linux/capability.h>
55 #include <linux/xattr.h>
56 #include <linux/namei.h>
57 #include <linux/security.h>
60 * Get a XFS inode from a given vnode.
68 bdp
= bhv_lookup_range(VN_BHV_HEAD(vp
),
69 VNODE_POSITION_XFS
, VNODE_POSITION_XFS
);
70 if (unlikely(bdp
== NULL
))
72 return XFS_BHVTOI(bdp
);
76 * Bring the atime in the XFS inode uptodate.
77 * Used before logging the inode to disk or when the Linux inode goes away.
80 xfs_synchronize_atime(
85 vp
= XFS_ITOV_NULL(ip
);
87 struct inode
*inode
= &vp
->v_inode
;
88 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)inode
->i_atime
.tv_sec
;
89 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)inode
->i_atime
.tv_nsec
;
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.
98 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
99 * with XFS_ICHGTIME_ACC to be sure that access time
100 * update will take. Calling first with XFS_ICHGTIME_ACC
101 * and then XFS_ICHGTIME_MOD may fail to modify the access
102 * timestamp if the filesystem is mounted noacctm.
109 struct inode
*inode
= LINVFS_GET_IP(XFS_ITOV(ip
));
113 if (flags
& XFS_ICHGTIME_MOD
) {
115 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)tv
.tv_sec
;
116 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
118 if (flags
& XFS_ICHGTIME_ACC
) {
120 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)tv
.tv_sec
;
121 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
123 if (flags
& XFS_ICHGTIME_CHG
) {
125 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tv
.tv_sec
;
126 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tv
.tv_nsec
;
130 * We update the i_update_core field _after_ changing
131 * the timestamps in order to coordinate properly with
132 * xfs_iflush() so that we don't lose timestamp updates.
133 * This keeps us from having to hold the inode lock
134 * while doing this. We use the SYNCHRONIZE macro to
135 * ensure that the compiler does not reorder the update
136 * of i_update_core above the timestamp updates above.
139 ip
->i_update_core
= 1;
140 if (!(inode
->i_state
& I_LOCK
))
141 mark_inode_dirty_sync(inode
);
145 * Variant on the above which avoids querying the system clock
146 * in situations where we know the Linux inode timestamps have
147 * just been updated (and so we can update our inode cheaply).
158 * Atime updates for read() & friends are handled lazily now, and
159 * explicit updates must go through xfs_ichgtime()
161 ASSERT((flags
& XFS_ICHGTIME_ACC
) == 0);
164 * We're not supposed to change timestamps in readonly-mounted
165 * filesystems. Throw it away if anyone asks us.
167 if (unlikely(IS_RDONLY(inode
)))
170 if (flags
& XFS_ICHGTIME_MOD
) {
171 tvp
= &inode
->i_mtime
;
172 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)tvp
->tv_sec
;
173 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)tvp
->tv_nsec
;
175 if (flags
& XFS_ICHGTIME_CHG
) {
176 tvp
= &inode
->i_ctime
;
177 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)tvp
->tv_sec
;
178 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)tvp
->tv_nsec
;
182 * We update the i_update_core field _after_ changing
183 * the timestamps in order to coordinate properly with
184 * xfs_iflush() so that we don't lose timestamp updates.
185 * This keeps us from having to hold the inode lock
186 * while doing this. We use the SYNCHRONIZE macro to
187 * ensure that the compiler does not reorder the update
188 * of i_update_core above the timestamp updates above.
191 ip
->i_update_core
= 1;
192 if (!(inode
->i_state
& I_LOCK
))
193 mark_inode_dirty_sync(inode
);
198 * Pull the link count and size up from the xfs inode to the linux inode
204 vnode_t
*vp
= LINVFS_GET_VP(ip
);
208 va
.va_mask
= XFS_AT_NLINK
|XFS_AT_SIZE
|XFS_AT_NBLOCKS
;
209 VOP_GETATTR(vp
, &va
, ATTR_LAZY
, NULL
, error
);
210 if (likely(!error
)) {
211 ip
->i_nlink
= va
.va_nlink
;
212 ip
->i_blocks
= va
.va_nblocks
;
214 /* we're under i_mutex so i_size can't change under us */
215 if (i_size_read(ip
) != va
.va_size
)
216 i_size_write(ip
, va
.va_size
);
221 * Hook in SELinux. This is not quite correct yet, what we really need
222 * here (as we do for default ACLs) is a mechanism by which creation of
223 * these attrs can be journalled at inode creation time (along with the
224 * inode, of course, such that log replay can't cause these to be lost).
227 linvfs_init_security(
231 struct inode
*ip
= LINVFS_GET_IP(vp
);
237 error
= security_inode_init_security(ip
, dir
, &name
, &value
, &length
);
239 if (error
== -EOPNOTSUPP
)
244 VOP_ATTR_SET(vp
, name
, value
, length
, ATTR_SECURE
, NULL
, error
);
254 * Determine whether a process has a valid fs_struct (kernel daemons
255 * like knfsd don't have an fs_struct).
257 * XXX(hch): nfsd is broken, better fix it instead.
260 has_fs_struct(struct task_struct
*task
)
262 return (task
->fs
!= init_task
.fs
);
268 struct dentry
*dentry
,
274 vnode_t
*vp
= NULL
, *dvp
= LINVFS_GET_VP(dir
);
275 xfs_acl_t
*default_acl
= NULL
;
276 attrexists_t test_default_acl
= _ACL_DEFAULT_EXISTS
;
280 * Irix uses Missed'em'V split, but doesn't want to see
281 * the upper 5 bits of (14bit) major.
283 if (!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff)
286 if (test_default_acl
&& test_default_acl(dvp
)) {
287 if (!_ACL_ALLOC(default_acl
))
289 if (!_ACL_GET_DEFAULT(dvp
, default_acl
)) {
290 _ACL_FREE(default_acl
);
295 if (IS_POSIXACL(dir
) && !default_acl
&& has_fs_struct(current
))
296 mode
&= ~current
->fs
->umask
;
298 memset(&va
, 0, sizeof(va
));
299 va
.va_mask
= XFS_AT_TYPE
|XFS_AT_MODE
;
302 switch (mode
& S_IFMT
) {
303 case S_IFCHR
: case S_IFBLK
: case S_IFIFO
: case S_IFSOCK
:
304 va
.va_rdev
= sysv_encode_dev(rdev
);
305 va
.va_mask
|= XFS_AT_RDEV
;
308 VOP_CREATE(dvp
, dentry
, &va
, &vp
, NULL
, error
);
311 VOP_MKDIR(dvp
, dentry
, &va
, &vp
, NULL
, error
);
319 error
= linvfs_init_security(vp
, dir
);
323 error
= _ACL_INHERIT(vp
, &va
, default_acl
);
327 struct dentry teardown
= {};
331 * If we can't add the ACL we must back out.
332 * ENOSPC can hit here, among other things.
334 teardown
.d_inode
= ip
= LINVFS_GET_IP(vp
);
335 teardown
.d_name
= dentry
->d_name
;
338 VOP_RMDIR(dvp
, &teardown
, NULL
, err2
);
340 VOP_REMOVE(dvp
, &teardown
, NULL
, err2
);
344 _ACL_FREE(default_acl
);
349 ip
= LINVFS_GET_IP(vp
);
351 if (S_ISCHR(mode
) || S_ISBLK(mode
))
353 else if (S_ISDIR(mode
))
355 d_instantiate(dentry
, ip
);
356 validate_fields(dir
);
364 struct dentry
*dentry
,
366 struct nameidata
*nd
)
368 return linvfs_mknod(dir
, dentry
, mode
, 0);
374 struct dentry
*dentry
,
377 return linvfs_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
380 STATIC
struct dentry
*
383 struct dentry
*dentry
,
384 struct nameidata
*nd
)
386 struct vnode
*vp
= LINVFS_GET_VP(dir
), *cvp
;
389 if (dentry
->d_name
.len
>= MAXNAMELEN
)
390 return ERR_PTR(-ENAMETOOLONG
);
392 VOP_LOOKUP(vp
, dentry
, &cvp
, 0, NULL
, NULL
, error
);
394 if (unlikely(error
!= ENOENT
))
395 return ERR_PTR(-error
);
400 return d_splice_alias(LINVFS_GET_IP(cvp
), dentry
);
405 struct dentry
*old_dentry
,
407 struct dentry
*dentry
)
409 struct inode
*ip
; /* inode of guy being linked to */
410 vnode_t
*tdvp
; /* target directory for new name/link */
411 vnode_t
*vp
; /* vp of name being linked */
414 ip
= old_dentry
->d_inode
; /* inode being linked to */
415 if (S_ISDIR(ip
->i_mode
))
418 tdvp
= LINVFS_GET_VP(dir
);
419 vp
= LINVFS_GET_VP(ip
);
421 VOP_LINK(tdvp
, vp
, dentry
, NULL
, error
);
426 d_instantiate(dentry
, ip
);
434 struct dentry
*dentry
)
437 vnode_t
*dvp
; /* directory containing name to remove */
440 inode
= dentry
->d_inode
;
441 dvp
= LINVFS_GET_VP(dir
);
443 VOP_REMOVE(dvp
, dentry
, NULL
, error
);
445 validate_fields(dir
); /* For size only */
446 validate_fields(inode
);
455 struct dentry
*dentry
,
460 vnode_t
*dvp
; /* directory containing name of symlink */
461 vnode_t
*cvp
; /* used to lookup symlink to put in dentry */
464 dvp
= LINVFS_GET_VP(dir
);
467 memset(&va
, 0, sizeof(va
));
468 va
.va_mode
= S_IFLNK
|
469 (irix_symlink_mode
? 0777 & ~current
->fs
->umask
: S_IRWXUGO
);
470 va
.va_mask
= XFS_AT_TYPE
|XFS_AT_MODE
;
473 VOP_SYMLINK(dvp
, dentry
, &va
, (char *)symname
, &cvp
, NULL
, error
);
474 if (likely(!error
&& cvp
)) {
475 error
= linvfs_init_security(cvp
, dir
);
476 if (likely(!error
)) {
477 ip
= LINVFS_GET_IP(cvp
);
478 d_instantiate(dentry
, ip
);
479 validate_fields(dir
);
489 struct dentry
*dentry
)
491 struct inode
*inode
= dentry
->d_inode
;
492 vnode_t
*dvp
= LINVFS_GET_VP(dir
);
495 VOP_RMDIR(dvp
, dentry
, NULL
, error
);
497 validate_fields(inode
);
498 validate_fields(dir
);
506 struct dentry
*odentry
,
508 struct dentry
*ndentry
)
510 struct inode
*new_inode
= ndentry
->d_inode
;
511 vnode_t
*fvp
; /* from directory */
512 vnode_t
*tvp
; /* target directory */
515 fvp
= LINVFS_GET_VP(odir
);
516 tvp
= LINVFS_GET_VP(ndir
);
518 VOP_RENAME(fvp
, odentry
, tvp
, ndentry
, NULL
, error
);
523 validate_fields(new_inode
);
525 validate_fields(odir
);
527 validate_fields(ndir
);
532 * careful here - this function can get called recursively, so
533 * we need to be very careful about how much stack we use.
534 * uio is kmalloced for this reason...
538 struct dentry
*dentry
,
539 struct nameidata
*nd
)
550 link
= (char *)kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
552 nd_set_link(nd
, ERR_PTR(-ENOMEM
));
556 uio
= (uio_t
*)kmalloc(sizeof(uio_t
), GFP_KERNEL
);
559 nd_set_link(nd
, ERR_PTR(-ENOMEM
));
563 vp
= LINVFS_GET_VP(dentry
->d_inode
);
566 iov
.iov_len
= MAXPATHLEN
;
570 uio
->uio_segflg
= UIO_SYSSPACE
;
571 uio
->uio_resid
= MAXPATHLEN
;
574 VOP_READLINK(vp
, uio
, 0, NULL
, error
);
577 link
= ERR_PTR(-error
);
579 link
[MAXPATHLEN
- uio
->uio_resid
] = '\0';
583 nd_set_link(nd
, link
);
589 struct dentry
*dentry
,
590 struct nameidata
*nd
,
593 char *s
= nd_get_link(nd
);
599 #ifdef CONFIG_XFS_POSIX_ACL
604 struct nameidata
*nd
)
606 vnode_t
*vp
= LINVFS_GET_VP(inode
);
609 mode
<<= 6; /* convert from linux to vnode access bits */
610 VOP_ACCESS(vp
, mode
, NULL
, error
);
614 #define linvfs_permission NULL
619 struct vfsmount
*mnt
,
620 struct dentry
*dentry
,
623 struct inode
*inode
= dentry
->d_inode
;
624 vnode_t
*vp
= LINVFS_GET_VP(inode
);
627 if (unlikely(vp
->v_flag
& VMODIFIED
))
628 error
= vn_revalidate(vp
);
630 generic_fillattr(inode
, stat
);
636 struct dentry
*dentry
,
639 struct inode
*inode
= dentry
->d_inode
;
640 unsigned int ia_valid
= attr
->ia_valid
;
641 vnode_t
*vp
= LINVFS_GET_VP(inode
);
646 memset(&vattr
, 0, sizeof(vattr_t
));
647 if (ia_valid
& ATTR_UID
) {
648 vattr
.va_mask
|= XFS_AT_UID
;
649 vattr
.va_uid
= attr
->ia_uid
;
651 if (ia_valid
& ATTR_GID
) {
652 vattr
.va_mask
|= XFS_AT_GID
;
653 vattr
.va_gid
= attr
->ia_gid
;
655 if (ia_valid
& ATTR_SIZE
) {
656 vattr
.va_mask
|= XFS_AT_SIZE
;
657 vattr
.va_size
= attr
->ia_size
;
659 if (ia_valid
& ATTR_ATIME
) {
660 vattr
.va_mask
|= XFS_AT_ATIME
;
661 vattr
.va_atime
= attr
->ia_atime
;
663 if (ia_valid
& ATTR_MTIME
) {
664 vattr
.va_mask
|= XFS_AT_MTIME
;
665 vattr
.va_mtime
= attr
->ia_mtime
;
667 if (ia_valid
& ATTR_CTIME
) {
668 vattr
.va_mask
|= XFS_AT_CTIME
;
669 vattr
.va_ctime
= attr
->ia_ctime
;
671 if (ia_valid
& ATTR_MODE
) {
672 vattr
.va_mask
|= XFS_AT_MODE
;
673 vattr
.va_mode
= attr
->ia_mode
;
674 if (!in_group_p(inode
->i_gid
) && !capable(CAP_FSETID
))
675 inode
->i_mode
&= ~S_ISGID
;
678 if (ia_valid
& (ATTR_MTIME_SET
| ATTR_ATIME_SET
))
681 if ((ia_valid
& ATTR_NO_BLOCK
))
682 flags
|= ATTR_NONBLOCK
;
685 VOP_SETATTR(vp
, &vattr
, flags
, NULL
, error
);
696 block_truncate_page(inode
->i_mapping
, inode
->i_size
, linvfs_get_block
);
701 struct dentry
*dentry
,
707 vnode_t
*vp
= LINVFS_GET_VP(dentry
->d_inode
);
708 char *attr
= (char *)name
;
713 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
716 attr
+= namesp
->attr_namelen
;
717 error
= namesp
->attr_capable(vp
, NULL
);
721 /* Convert Linux syscall to XFS internal ATTR flags */
722 if (flags
& XATTR_CREATE
)
723 xflags
|= ATTR_CREATE
;
724 if (flags
& XATTR_REPLACE
)
725 xflags
|= ATTR_REPLACE
;
726 xflags
|= namesp
->attr_flag
;
727 return namesp
->attr_set(vp
, attr
, (void *)data
, size
, xflags
);
732 struct dentry
*dentry
,
737 vnode_t
*vp
= LINVFS_GET_VP(dentry
->d_inode
);
738 char *attr
= (char *)name
;
743 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
746 attr
+= namesp
->attr_namelen
;
747 error
= namesp
->attr_capable(vp
, NULL
);
751 /* Convert Linux syscall to XFS internal ATTR flags */
753 xflags
|= ATTR_KERNOVAL
;
756 xflags
|= namesp
->attr_flag
;
757 return namesp
->attr_get(vp
, attr
, (void *)data
, size
, xflags
);
762 struct dentry
*dentry
,
766 vnode_t
*vp
= LINVFS_GET_VP(dentry
->d_inode
);
767 int error
, xflags
= ATTR_KERNAMELS
;
771 xflags
|= ATTR_KERNOVAL
;
772 xflags
|= capable(CAP_SYS_ADMIN
) ? ATTR_KERNFULLS
: ATTR_KERNORMALS
;
774 error
= attr_generic_list(vp
, data
, size
, xflags
, &result
);
782 struct dentry
*dentry
,
785 vnode_t
*vp
= LINVFS_GET_VP(dentry
->d_inode
);
786 char *attr
= (char *)name
;
791 namesp
= attr_lookup_namespace(attr
, attr_namespaces
, ATTR_NAMECOUNT
);
794 attr
+= namesp
->attr_namelen
;
795 error
= namesp
->attr_capable(vp
, NULL
);
798 xflags
|= namesp
->attr_flag
;
799 return namesp
->attr_remove(vp
, attr
, xflags
);
803 struct inode_operations linvfs_file_inode_operations
= {
804 .permission
= linvfs_permission
,
805 .truncate
= linvfs_truncate
,
806 .getattr
= linvfs_getattr
,
807 .setattr
= linvfs_setattr
,
808 .setxattr
= linvfs_setxattr
,
809 .getxattr
= linvfs_getxattr
,
810 .listxattr
= linvfs_listxattr
,
811 .removexattr
= linvfs_removexattr
,
814 struct inode_operations linvfs_dir_inode_operations
= {
815 .create
= linvfs_create
,
816 .lookup
= linvfs_lookup
,
818 .unlink
= linvfs_unlink
,
819 .symlink
= linvfs_symlink
,
820 .mkdir
= linvfs_mkdir
,
821 .rmdir
= linvfs_rmdir
,
822 .mknod
= linvfs_mknod
,
823 .rename
= linvfs_rename
,
824 .permission
= linvfs_permission
,
825 .getattr
= linvfs_getattr
,
826 .setattr
= linvfs_setattr
,
827 .setxattr
= linvfs_setxattr
,
828 .getxattr
= linvfs_getxattr
,
829 .listxattr
= linvfs_listxattr
,
830 .removexattr
= linvfs_removexattr
,
833 struct inode_operations linvfs_symlink_inode_operations
= {
834 .readlink
= generic_readlink
,
835 .follow_link
= linvfs_follow_link
,
836 .put_link
= linvfs_put_link
,
837 .permission
= linvfs_permission
,
838 .getattr
= linvfs_getattr
,
839 .setattr
= linvfs_setattr
,
840 .setxattr
= linvfs_setxattr
,
841 .getxattr
= linvfs_getxattr
,
842 .listxattr
= linvfs_listxattr
,
843 .removexattr
= linvfs_removexattr
,