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
20 #include "xfs_format.h"
23 #include "xfs_trans.h"
26 #include "xfs_alloc.h"
27 #include "xfs_quota.h"
28 #include "xfs_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_dinode.h"
31 #include "xfs_inode.h"
33 #include "xfs_rtalloc.h"
34 #include "xfs_error.h"
35 #include "xfs_itable.h"
37 #include "xfs_buf_item.h"
38 #include "xfs_utils.h"
39 #include "xfs_vnodeops.h"
40 #include "xfs_inode_item.h"
41 #include "xfs_trace.h"
42 #include "xfs_icache.h"
44 #include <linux/capability.h>
45 #include <linux/xattr.h>
46 #include <linux/namei.h>
47 #include <linux/posix_acl.h>
48 #include <linux/security.h>
49 #include <linux/fiemap.h>
50 #include <linux/slab.h>
55 const struct xattr
*xattr_array
,
58 const struct xattr
*xattr
;
59 struct xfs_inode
*ip
= XFS_I(inode
);
62 for (xattr
= xattr_array
; xattr
->name
!= NULL
; xattr
++) {
63 error
= xfs_attr_set(ip
, xattr
->name
, xattr
->value
,
64 xattr
->value_len
, ATTR_SECURE
);
72 * Hook in SELinux. This is not quite correct yet, what we really need
73 * here (as we do for default ACLs) is a mechanism by which creation of
74 * these attrs can be journalled at inode creation time (along with the
75 * inode, of course, such that log replay can't cause these to be lost).
82 const struct qstr
*qstr
)
84 return security_inode_init_security(inode
, dir
, qstr
,
85 &xfs_initxattrs
, NULL
);
90 struct xfs_name
*namep
,
91 struct dentry
*dentry
)
93 namep
->name
= dentry
->d_name
.name
;
94 namep
->len
= dentry
->d_name
.len
;
101 struct dentry
*dentry
)
103 struct xfs_name teardown
;
106 * If we can't add the ACL or we fail in
107 * xfs_init_security we must back out.
108 * ENOSPC can hit here, among other things.
110 xfs_dentry_to_name(&teardown
, dentry
);
112 xfs_remove(XFS_I(dir
), &teardown
, XFS_I(inode
));
119 struct dentry
*dentry
,
124 struct xfs_inode
*ip
= NULL
;
125 struct posix_acl
*default_acl
= NULL
;
126 struct xfs_name name
;
130 * Irix uses Missed'em'V split, but doesn't want to see
131 * the upper 5 bits of (14bit) major.
133 if (S_ISCHR(mode
) || S_ISBLK(mode
)) {
134 if (unlikely(!sysv_valid_dev(rdev
) || MAJOR(rdev
) & ~0x1ff))
136 rdev
= sysv_encode_dev(rdev
);
141 if (IS_POSIXACL(dir
)) {
142 default_acl
= xfs_get_acl(dir
, ACL_TYPE_DEFAULT
);
143 if (IS_ERR(default_acl
))
144 return PTR_ERR(default_acl
);
147 mode
&= ~current_umask();
150 xfs_dentry_to_name(&name
, dentry
);
151 error
= xfs_create(XFS_I(dir
), &name
, mode
, rdev
, &ip
);
157 error
= xfs_init_security(inode
, dir
, &dentry
->d_name
);
159 goto out_cleanup_inode
;
162 error
= -xfs_inherit_acl(inode
, default_acl
);
165 goto out_cleanup_inode
;
169 d_instantiate(dentry
, inode
);
173 xfs_cleanup_inode(dir
, inode
, dentry
);
175 posix_acl_release(default_acl
);
182 struct dentry
*dentry
,
186 return xfs_vn_mknod(dir
, dentry
, mode
, 0);
192 struct dentry
*dentry
,
195 return xfs_vn_mknod(dir
, dentry
, mode
|S_IFDIR
, 0);
198 STATIC
struct dentry
*
201 struct dentry
*dentry
,
204 struct xfs_inode
*cip
;
205 struct xfs_name name
;
208 if (dentry
->d_name
.len
>= MAXNAMELEN
)
209 return ERR_PTR(-ENAMETOOLONG
);
211 xfs_dentry_to_name(&name
, dentry
);
212 error
= xfs_lookup(XFS_I(dir
), &name
, &cip
, NULL
);
213 if (unlikely(error
)) {
214 if (unlikely(error
!= ENOENT
))
215 return ERR_PTR(-error
);
220 return d_splice_alias(VFS_I(cip
), dentry
);
223 STATIC
struct dentry
*
226 struct dentry
*dentry
,
229 struct xfs_inode
*ip
;
230 struct xfs_name xname
;
231 struct xfs_name ci_name
;
235 if (dentry
->d_name
.len
>= MAXNAMELEN
)
236 return ERR_PTR(-ENAMETOOLONG
);
238 xfs_dentry_to_name(&xname
, dentry
);
239 error
= xfs_lookup(XFS_I(dir
), &xname
, &ip
, &ci_name
);
240 if (unlikely(error
)) {
241 if (unlikely(error
!= ENOENT
))
242 return ERR_PTR(-error
);
244 * call d_add(dentry, NULL) here when d_drop_negative_children
245 * is called in xfs_vn_mknod (ie. allow negative dentries
246 * with CI filesystems).
251 /* if exact match, just splice and exit */
253 return d_splice_alias(VFS_I(ip
), dentry
);
255 /* else case-insensitive match... */
256 dname
.name
= ci_name
.name
;
257 dname
.len
= ci_name
.len
;
258 dentry
= d_add_ci(dentry
, VFS_I(ip
), &dname
);
259 kmem_free(ci_name
.name
);
265 struct dentry
*old_dentry
,
267 struct dentry
*dentry
)
269 struct inode
*inode
= old_dentry
->d_inode
;
270 struct xfs_name name
;
273 xfs_dentry_to_name(&name
, dentry
);
275 error
= xfs_link(XFS_I(dir
), XFS_I(inode
), &name
);
280 d_instantiate(dentry
, inode
);
287 struct dentry
*dentry
)
289 struct xfs_name name
;
292 xfs_dentry_to_name(&name
, dentry
);
294 error
= -xfs_remove(XFS_I(dir
), &name
, XFS_I(dentry
->d_inode
));
299 * With unlink, the VFS makes the dentry "negative": no inode,
300 * but still hashed. This is incompatible with case-insensitive
301 * mode, so invalidate (unhash) the dentry in CI-mode.
303 if (xfs_sb_version_hasasciici(&XFS_M(dir
->i_sb
)->m_sb
))
304 d_invalidate(dentry
);
311 struct dentry
*dentry
,
315 struct xfs_inode
*cip
= NULL
;
316 struct xfs_name name
;
321 (irix_symlink_mode
? 0777 & ~current_umask() : S_IRWXUGO
);
322 xfs_dentry_to_name(&name
, dentry
);
324 error
= xfs_symlink(XFS_I(dir
), &name
, symname
, mode
, &cip
);
330 error
= xfs_init_security(inode
, dir
, &dentry
->d_name
);
332 goto out_cleanup_inode
;
334 d_instantiate(dentry
, inode
);
338 xfs_cleanup_inode(dir
, inode
, dentry
);
346 struct dentry
*odentry
,
348 struct dentry
*ndentry
)
350 struct inode
*new_inode
= ndentry
->d_inode
;
351 struct xfs_name oname
;
352 struct xfs_name nname
;
354 xfs_dentry_to_name(&oname
, odentry
);
355 xfs_dentry_to_name(&nname
, ndentry
);
357 return -xfs_rename(XFS_I(odir
), &oname
, XFS_I(odentry
->d_inode
),
358 XFS_I(ndir
), &nname
, new_inode
?
359 XFS_I(new_inode
) : NULL
);
363 * careful here - this function can get called recursively, so
364 * we need to be very careful about how much stack we use.
365 * uio is kmalloced for this reason...
369 struct dentry
*dentry
,
370 struct nameidata
*nd
)
375 link
= kmalloc(MAXPATHLEN
+1, GFP_KERNEL
);
379 error
= -xfs_readlink(XFS_I(dentry
->d_inode
), link
);
383 nd_set_link(nd
, link
);
389 nd_set_link(nd
, ERR_PTR(error
));
395 struct dentry
*dentry
,
396 struct nameidata
*nd
,
399 char *s
= nd_get_link(nd
);
407 struct vfsmount
*mnt
,
408 struct dentry
*dentry
,
411 struct inode
*inode
= dentry
->d_inode
;
412 struct xfs_inode
*ip
= XFS_I(inode
);
413 struct xfs_mount
*mp
= ip
->i_mount
;
415 trace_xfs_getattr(ip
);
417 if (XFS_FORCED_SHUTDOWN(mp
))
418 return -XFS_ERROR(EIO
);
420 stat
->size
= XFS_ISIZE(ip
);
421 stat
->dev
= inode
->i_sb
->s_dev
;
422 stat
->mode
= ip
->i_d
.di_mode
;
423 stat
->nlink
= ip
->i_d
.di_nlink
;
424 stat
->uid
= ip
->i_d
.di_uid
;
425 stat
->gid
= ip
->i_d
.di_gid
;
426 stat
->ino
= ip
->i_ino
;
427 stat
->atime
= inode
->i_atime
;
428 stat
->mtime
= inode
->i_mtime
;
429 stat
->ctime
= inode
->i_ctime
;
431 XFS_FSB_TO_BB(mp
, ip
->i_d
.di_nblocks
+ ip
->i_delayed_blks
);
434 switch (inode
->i_mode
& S_IFMT
) {
437 stat
->blksize
= BLKDEV_IOSIZE
;
438 stat
->rdev
= MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
439 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
442 if (XFS_IS_REALTIME_INODE(ip
)) {
444 * If the file blocks are being allocated from a
445 * realtime volume, then return the inode's realtime
446 * extent size or the realtime volume's extent size.
449 xfs_get_extsz_hint(ip
) << mp
->m_sb
.sb_blocklog
;
451 stat
->blksize
= xfs_preferred_iosize(mp
);
461 struct xfs_trans
*tp
,
462 struct xfs_inode
*ip
,
465 struct inode
*inode
= VFS_I(ip
);
466 umode_t mode
= iattr
->ia_mode
;
469 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_EXCL
));
471 ip
->i_d
.di_mode
&= S_IFMT
;
472 ip
->i_d
.di_mode
|= mode
& ~S_IFMT
;
474 inode
->i_mode
&= S_IFMT
;
475 inode
->i_mode
|= mode
& ~S_IFMT
;
480 struct xfs_inode
*ip
,
484 xfs_mount_t
*mp
= ip
->i_mount
;
485 struct inode
*inode
= VFS_I(ip
);
486 int mask
= iattr
->ia_valid
;
489 uid_t uid
= 0, iuid
= 0;
490 gid_t gid
= 0, igid
= 0;
491 struct xfs_dquot
*udqp
= NULL
, *gdqp
= NULL
;
492 struct xfs_dquot
*olddquot1
= NULL
, *olddquot2
= NULL
;
494 trace_xfs_setattr(ip
);
496 /* If acls are being inherited, we already have this checked */
497 if (!(flags
& XFS_ATTR_NOACL
)) {
498 if (mp
->m_flags
& XFS_MOUNT_RDONLY
)
499 return XFS_ERROR(EROFS
);
501 if (XFS_FORCED_SHUTDOWN(mp
))
502 return XFS_ERROR(EIO
);
504 error
= -inode_change_ok(inode
, iattr
);
506 return XFS_ERROR(error
);
509 ASSERT((mask
& ATTR_SIZE
) == 0);
512 * If disk quotas is on, we make sure that the dquots do exist on disk,
513 * before we start any other transactions. Trying to do this later
514 * is messy. We don't care to take a readlock to look at the ids
515 * in inode here, because we can't hold it across the trans_reserve.
516 * If the IDs do change before we take the ilock, we're covered
517 * because the i_*dquot fields will get updated anyway.
519 if (XFS_IS_QUOTA_ON(mp
) && (mask
& (ATTR_UID
|ATTR_GID
))) {
522 if ((mask
& ATTR_UID
) && XFS_IS_UQUOTA_ON(mp
)) {
524 qflags
|= XFS_QMOPT_UQUOTA
;
526 uid
= ip
->i_d
.di_uid
;
528 if ((mask
& ATTR_GID
) && XFS_IS_GQUOTA_ON(mp
)) {
530 qflags
|= XFS_QMOPT_GQUOTA
;
532 gid
= ip
->i_d
.di_gid
;
536 * We take a reference when we initialize udqp and gdqp,
537 * so it is important that we never blindly double trip on
538 * the same variable. See xfs_create() for an example.
540 ASSERT(udqp
== NULL
);
541 ASSERT(gdqp
== NULL
);
542 error
= xfs_qm_vop_dqalloc(ip
, uid
, gid
, xfs_get_projid(ip
),
543 qflags
, &udqp
, &gdqp
, NULL
);
548 tp
= xfs_trans_alloc(mp
, XFS_TRANS_SETATTR_NOT_SIZE
);
549 error
= xfs_trans_reserve(tp
, 0, XFS_ICHANGE_LOG_RES(mp
), 0, 0, 0);
553 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
556 * Change file ownership. Must be the owner or privileged.
558 if (mask
& (ATTR_UID
|ATTR_GID
)) {
560 * These IDs could have changed since we last looked at them.
561 * But, we're assured that if the ownership did change
562 * while we didn't have the inode locked, inode's dquot(s)
563 * would have changed also.
565 iuid
= ip
->i_d
.di_uid
;
566 igid
= ip
->i_d
.di_gid
;
567 gid
= (mask
& ATTR_GID
) ? iattr
->ia_gid
: igid
;
568 uid
= (mask
& ATTR_UID
) ? iattr
->ia_uid
: iuid
;
571 * Do a quota reservation only if uid/gid is actually
574 if (XFS_IS_QUOTA_RUNNING(mp
) &&
575 ((XFS_IS_UQUOTA_ON(mp
) && iuid
!= uid
) ||
576 (XFS_IS_GQUOTA_ON(mp
) && igid
!= gid
))) {
578 error
= xfs_qm_vop_chown_reserve(tp
, ip
, udqp
, gdqp
,
579 NULL
, capable(CAP_FOWNER
) ?
580 XFS_QMOPT_FORCE_RES
: 0);
581 if (error
) /* out of quota */
582 goto out_trans_cancel
;
586 xfs_trans_ijoin(tp
, ip
, 0);
589 * Change file ownership. Must be the owner or privileged.
591 if (mask
& (ATTR_UID
|ATTR_GID
)) {
593 * CAP_FSETID overrides the following restrictions:
595 * The set-user-ID and set-group-ID bits of a file will be
596 * cleared upon successful return from chown()
598 if ((ip
->i_d
.di_mode
& (S_ISUID
|S_ISGID
)) &&
599 !capable(CAP_FSETID
))
600 ip
->i_d
.di_mode
&= ~(S_ISUID
|S_ISGID
);
603 * Change the ownerships and register quota modifications
604 * in the transaction.
607 if (XFS_IS_QUOTA_RUNNING(mp
) && XFS_IS_UQUOTA_ON(mp
)) {
608 ASSERT(mask
& ATTR_UID
);
610 olddquot1
= xfs_qm_vop_chown(tp
, ip
,
611 &ip
->i_udquot
, udqp
);
613 ip
->i_d
.di_uid
= uid
;
617 if (XFS_IS_QUOTA_RUNNING(mp
) && XFS_IS_GQUOTA_ON(mp
)) {
618 ASSERT(!XFS_IS_PQUOTA_ON(mp
));
619 ASSERT(mask
& ATTR_GID
);
621 olddquot2
= xfs_qm_vop_chown(tp
, ip
,
622 &ip
->i_gdquot
, gdqp
);
624 ip
->i_d
.di_gid
= gid
;
630 * Change file access modes.
632 if (mask
& ATTR_MODE
)
633 xfs_setattr_mode(tp
, ip
, iattr
);
636 * Change file access or modified times.
638 if (mask
& ATTR_ATIME
) {
639 inode
->i_atime
= iattr
->ia_atime
;
640 ip
->i_d
.di_atime
.t_sec
= iattr
->ia_atime
.tv_sec
;
641 ip
->i_d
.di_atime
.t_nsec
= iattr
->ia_atime
.tv_nsec
;
643 if (mask
& ATTR_CTIME
) {
644 inode
->i_ctime
= iattr
->ia_ctime
;
645 ip
->i_d
.di_ctime
.t_sec
= iattr
->ia_ctime
.tv_sec
;
646 ip
->i_d
.di_ctime
.t_nsec
= iattr
->ia_ctime
.tv_nsec
;
648 if (mask
& ATTR_MTIME
) {
649 inode
->i_mtime
= iattr
->ia_mtime
;
650 ip
->i_d
.di_mtime
.t_sec
= iattr
->ia_mtime
.tv_sec
;
651 ip
->i_d
.di_mtime
.t_nsec
= iattr
->ia_mtime
.tv_nsec
;
654 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
656 XFS_STATS_INC(xs_ig_attrchg
);
658 if (mp
->m_flags
& XFS_MOUNT_WSYNC
)
659 xfs_trans_set_sync(tp
);
660 error
= xfs_trans_commit(tp
, 0);
662 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
665 * Release any dquot(s) the inode had kept before chown.
667 xfs_qm_dqrele(olddquot1
);
668 xfs_qm_dqrele(olddquot2
);
673 return XFS_ERROR(error
);
676 * XXX(hch): Updating the ACL entries is not atomic vs the i_mode
677 * update. We could avoid this with linked transactions
678 * and passing down the transaction pointer all the way
679 * to attr_set. No previous user of the generic
680 * Posix ACL code seems to care about this issue either.
682 if ((mask
& ATTR_MODE
) && !(flags
& XFS_ATTR_NOACL
)) {
683 error
= -xfs_acl_chmod(inode
);
685 return XFS_ERROR(error
);
691 xfs_trans_cancel(tp
, 0);
692 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
700 * Truncate file. Must have write permission and not be a directory.
704 struct xfs_inode
*ip
,
708 struct xfs_mount
*mp
= ip
->i_mount
;
709 struct inode
*inode
= VFS_I(ip
);
710 int mask
= iattr
->ia_valid
;
711 xfs_off_t oldsize
, newsize
;
712 struct xfs_trans
*tp
;
715 uint commit_flags
= 0;
717 trace_xfs_setattr(ip
);
719 if (mp
->m_flags
& XFS_MOUNT_RDONLY
)
720 return XFS_ERROR(EROFS
);
722 if (XFS_FORCED_SHUTDOWN(mp
))
723 return XFS_ERROR(EIO
);
725 error
= -inode_change_ok(inode
, iattr
);
727 return XFS_ERROR(error
);
729 ASSERT(S_ISREG(ip
->i_d
.di_mode
));
730 ASSERT((mask
& (ATTR_UID
|ATTR_GID
|ATTR_ATIME
|ATTR_ATIME_SET
|
731 ATTR_MTIME_SET
|ATTR_KILL_PRIV
|ATTR_TIMES_SET
)) == 0);
733 if (!(flags
& XFS_ATTR_NOLOCK
)) {
734 lock_flags
|= XFS_IOLOCK_EXCL
;
735 xfs_ilock(ip
, lock_flags
);
738 oldsize
= inode
->i_size
;
739 newsize
= iattr
->ia_size
;
742 * Short circuit the truncate case for zero length files.
744 if (newsize
== 0 && oldsize
== 0 && ip
->i_d
.di_nextents
== 0) {
745 if (!(mask
& (ATTR_CTIME
|ATTR_MTIME
)))
749 * Use the regular setattr path to update the timestamps.
751 xfs_iunlock(ip
, lock_flags
);
752 iattr
->ia_valid
&= ~ATTR_SIZE
;
753 return xfs_setattr_nonsize(ip
, iattr
, 0);
757 * Make sure that the dquots are attached to the inode.
759 error
= xfs_qm_dqattach(ip
, 0);
764 * Now we can make the changes. Before we join the inode to the
765 * transaction, take care of the part of the truncation that must be
766 * done without the inode lock. This needs to be done before joining
767 * the inode to the transaction, because the inode cannot be unlocked
768 * once it is a part of the transaction.
770 if (newsize
> oldsize
) {
772 * Do the first part of growing a file: zero any data in the
773 * last block that is beyond the old EOF. We need to do this
774 * before the inode is joined to the transaction to modify
777 error
= xfs_zero_eof(ip
, newsize
, oldsize
);
783 * We are going to log the inode size change in this transaction so
784 * any previous writes that are beyond the on disk EOF and the new
785 * EOF that have not been written out need to be written here. If we
786 * do not write the data out, we expose ourselves to the null files
789 * Only flush from the on disk size to the smaller of the in memory
790 * file size or the new size as that's the range we really care about
791 * here and prevents waiting for other data not within the range we
794 if (oldsize
!= ip
->i_d
.di_size
&& newsize
> ip
->i_d
.di_size
) {
795 error
= -filemap_write_and_wait_range(VFS_I(ip
)->i_mapping
,
796 ip
->i_d
.di_size
, newsize
);
802 * Wait for all direct I/O to complete.
804 inode_dio_wait(inode
);
806 error
= -block_truncate_page(inode
->i_mapping
, newsize
, xfs_get_blocks
);
810 tp
= xfs_trans_alloc(mp
, XFS_TRANS_SETATTR_SIZE
);
811 error
= xfs_trans_reserve(tp
, 0, XFS_ITRUNCATE_LOG_RES(mp
), 0,
812 XFS_TRANS_PERM_LOG_RES
,
813 XFS_ITRUNCATE_LOG_COUNT
);
815 goto out_trans_cancel
;
817 truncate_setsize(inode
, newsize
);
819 commit_flags
= XFS_TRANS_RELEASE_LOG_RES
;
820 lock_flags
|= XFS_ILOCK_EXCL
;
822 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
824 xfs_trans_ijoin(tp
, ip
, 0);
827 * Only change the c/mtime if we are changing the size or we are
828 * explicitly asked to change it. This handles the semantic difference
829 * between truncate() and ftruncate() as implemented in the VFS.
831 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
832 * special case where we need to update the times despite not having
833 * these flags set. For all other operations the VFS set these flags
834 * explicitly if it wants a timestamp update.
836 if (newsize
!= oldsize
&& (!(mask
& (ATTR_CTIME
| ATTR_MTIME
)))) {
837 iattr
->ia_ctime
= iattr
->ia_mtime
=
838 current_fs_time(inode
->i_sb
);
839 mask
|= ATTR_CTIME
| ATTR_MTIME
;
843 * The first thing we do is set the size to new_size permanently on
844 * disk. This way we don't have to worry about anyone ever being able
845 * to look at the data being freed even in the face of a crash.
846 * What we're getting around here is the case where we free a block, it
847 * is allocated to another file, it is written to, and then we crash.
848 * If the new data gets written to the file but the log buffers
849 * containing the free and reallocation don't, then we'd end up with
850 * garbage in the blocks being freed. As long as we make the new size
851 * permanent before actually freeing any blocks it doesn't matter if
852 * they get written to.
854 ip
->i_d
.di_size
= newsize
;
855 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
857 if (newsize
<= oldsize
) {
858 error
= xfs_itruncate_extents(&tp
, ip
, XFS_DATA_FORK
, newsize
);
860 goto out_trans_abort
;
863 * Truncated "down", so we're removing references to old data
864 * here - if we delay flushing for a long time, we expose
865 * ourselves unduly to the notorious NULL files problem. So,
866 * we mark this inode and flush it when the file is closed,
867 * and do not wait the usual (long) time for writeout.
869 xfs_iflags_set(ip
, XFS_ITRUNCATED
);
871 /* A truncate down always removes post-EOF blocks. */
872 xfs_inode_clear_eofblocks_tag(ip
);
876 * Change file access modes.
878 if (mask
& ATTR_MODE
)
879 xfs_setattr_mode(tp
, ip
, iattr
);
881 if (mask
& ATTR_CTIME
) {
882 inode
->i_ctime
= iattr
->ia_ctime
;
883 ip
->i_d
.di_ctime
.t_sec
= iattr
->ia_ctime
.tv_sec
;
884 ip
->i_d
.di_ctime
.t_nsec
= iattr
->ia_ctime
.tv_nsec
;
886 if (mask
& ATTR_MTIME
) {
887 inode
->i_mtime
= iattr
->ia_mtime
;
888 ip
->i_d
.di_mtime
.t_sec
= iattr
->ia_mtime
.tv_sec
;
889 ip
->i_d
.di_mtime
.t_nsec
= iattr
->ia_mtime
.tv_nsec
;
892 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
894 XFS_STATS_INC(xs_ig_attrchg
);
896 if (mp
->m_flags
& XFS_MOUNT_WSYNC
)
897 xfs_trans_set_sync(tp
);
899 error
= xfs_trans_commit(tp
, XFS_TRANS_RELEASE_LOG_RES
);
902 xfs_iunlock(ip
, lock_flags
);
906 commit_flags
|= XFS_TRANS_ABORT
;
908 xfs_trans_cancel(tp
, commit_flags
);
914 struct dentry
*dentry
,
917 if (iattr
->ia_valid
& ATTR_SIZE
)
918 return -xfs_setattr_size(XFS_I(dentry
->d_inode
), iattr
, 0);
919 return -xfs_setattr_nonsize(XFS_I(dentry
->d_inode
), iattr
, 0);
925 struct timespec
*now
,
928 struct xfs_inode
*ip
= XFS_I(inode
);
929 struct xfs_mount
*mp
= ip
->i_mount
;
930 struct xfs_trans
*tp
;
933 trace_xfs_update_time(ip
);
935 tp
= xfs_trans_alloc(mp
, XFS_TRANS_FSYNC_TS
);
936 error
= xfs_trans_reserve(tp
, 0, XFS_FSYNC_TS_LOG_RES(mp
), 0, 0, 0);
938 xfs_trans_cancel(tp
, 0);
942 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
943 if (flags
& S_CTIME
) {
944 inode
->i_ctime
= *now
;
945 ip
->i_d
.di_ctime
.t_sec
= (__int32_t
)now
->tv_sec
;
946 ip
->i_d
.di_ctime
.t_nsec
= (__int32_t
)now
->tv_nsec
;
948 if (flags
& S_MTIME
) {
949 inode
->i_mtime
= *now
;
950 ip
->i_d
.di_mtime
.t_sec
= (__int32_t
)now
->tv_sec
;
951 ip
->i_d
.di_mtime
.t_nsec
= (__int32_t
)now
->tv_nsec
;
953 if (flags
& S_ATIME
) {
954 inode
->i_atime
= *now
;
955 ip
->i_d
.di_atime
.t_sec
= (__int32_t
)now
->tv_sec
;
956 ip
->i_d
.di_atime
.t_nsec
= (__int32_t
)now
->tv_nsec
;
958 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
959 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_TIMESTAMP
);
960 return -xfs_trans_commit(tp
, 0);
963 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
966 * Call fiemap helper to fill in user data.
967 * Returns positive errors to xfs_getbmap.
972 struct getbmapx
*bmv
,
976 struct fiemap_extent_info
*fieinfo
= *arg
;
977 u32 fiemap_flags
= 0;
978 u64 logical
, physical
, length
;
980 /* Do nothing for a hole */
981 if (bmv
->bmv_block
== -1LL)
984 logical
= BBTOB(bmv
->bmv_offset
);
985 physical
= BBTOB(bmv
->bmv_block
);
986 length
= BBTOB(bmv
->bmv_length
);
988 if (bmv
->bmv_oflags
& BMV_OF_PREALLOC
)
989 fiemap_flags
|= FIEMAP_EXTENT_UNWRITTEN
;
990 else if (bmv
->bmv_oflags
& BMV_OF_DELALLOC
) {
991 fiemap_flags
|= (FIEMAP_EXTENT_DELALLOC
|
992 FIEMAP_EXTENT_UNKNOWN
);
993 physical
= 0; /* no block yet */
995 if (bmv
->bmv_oflags
& BMV_OF_LAST
)
996 fiemap_flags
|= FIEMAP_EXTENT_LAST
;
998 error
= fiemap_fill_next_extent(fieinfo
, logical
, physical
,
999 length
, fiemap_flags
);
1002 *full
= 1; /* user array now full */
1010 struct inode
*inode
,
1011 struct fiemap_extent_info
*fieinfo
,
1015 xfs_inode_t
*ip
= XFS_I(inode
);
1019 error
= fiemap_check_flags(fieinfo
, XFS_FIEMAP_FLAGS
);
1023 /* Set up bmap header for xfs internal routine */
1024 bm
.bmv_offset
= BTOBB(start
);
1025 /* Special case for whole file */
1026 if (length
== FIEMAP_MAX_OFFSET
)
1027 bm
.bmv_length
= -1LL;
1029 bm
.bmv_length
= BTOBB(length
);
1031 /* We add one because in getbmap world count includes the header */
1032 bm
.bmv_count
= !fieinfo
->fi_extents_max
? MAXEXTNUM
:
1033 fieinfo
->fi_extents_max
+ 1;
1034 bm
.bmv_count
= min_t(__s32
, bm
.bmv_count
,
1035 (PAGE_SIZE
* 16 / sizeof(struct getbmapx
)));
1036 bm
.bmv_iflags
= BMV_IF_PREALLOC
| BMV_IF_NO_HOLES
;
1037 if (fieinfo
->fi_flags
& FIEMAP_FLAG_XATTR
)
1038 bm
.bmv_iflags
|= BMV_IF_ATTRFORK
;
1039 if (!(fieinfo
->fi_flags
& FIEMAP_FLAG_SYNC
))
1040 bm
.bmv_iflags
|= BMV_IF_DELALLOC
;
1042 error
= xfs_getbmap(ip
, &bm
, xfs_fiemap_format
, fieinfo
);
1049 static const struct inode_operations xfs_inode_operations
= {
1050 .get_acl
= xfs_get_acl
,
1051 .getattr
= xfs_vn_getattr
,
1052 .setattr
= xfs_vn_setattr
,
1053 .setxattr
= generic_setxattr
,
1054 .getxattr
= generic_getxattr
,
1055 .removexattr
= generic_removexattr
,
1056 .listxattr
= xfs_vn_listxattr
,
1057 .fiemap
= xfs_vn_fiemap
,
1058 .update_time
= xfs_vn_update_time
,
1061 static const struct inode_operations xfs_dir_inode_operations
= {
1062 .create
= xfs_vn_create
,
1063 .lookup
= xfs_vn_lookup
,
1064 .link
= xfs_vn_link
,
1065 .unlink
= xfs_vn_unlink
,
1066 .symlink
= xfs_vn_symlink
,
1067 .mkdir
= xfs_vn_mkdir
,
1069 * Yes, XFS uses the same method for rmdir and unlink.
1071 * There are some subtile differences deeper in the code,
1072 * but we use S_ISDIR to check for those.
1074 .rmdir
= xfs_vn_unlink
,
1075 .mknod
= xfs_vn_mknod
,
1076 .rename
= xfs_vn_rename
,
1077 .get_acl
= xfs_get_acl
,
1078 .getattr
= xfs_vn_getattr
,
1079 .setattr
= xfs_vn_setattr
,
1080 .setxattr
= generic_setxattr
,
1081 .getxattr
= generic_getxattr
,
1082 .removexattr
= generic_removexattr
,
1083 .listxattr
= xfs_vn_listxattr
,
1084 .update_time
= xfs_vn_update_time
,
1087 static const struct inode_operations xfs_dir_ci_inode_operations
= {
1088 .create
= xfs_vn_create
,
1089 .lookup
= xfs_vn_ci_lookup
,
1090 .link
= xfs_vn_link
,
1091 .unlink
= xfs_vn_unlink
,
1092 .symlink
= xfs_vn_symlink
,
1093 .mkdir
= xfs_vn_mkdir
,
1095 * Yes, XFS uses the same method for rmdir and unlink.
1097 * There are some subtile differences deeper in the code,
1098 * but we use S_ISDIR to check for those.
1100 .rmdir
= xfs_vn_unlink
,
1101 .mknod
= xfs_vn_mknod
,
1102 .rename
= xfs_vn_rename
,
1103 .get_acl
= xfs_get_acl
,
1104 .getattr
= xfs_vn_getattr
,
1105 .setattr
= xfs_vn_setattr
,
1106 .setxattr
= generic_setxattr
,
1107 .getxattr
= generic_getxattr
,
1108 .removexattr
= generic_removexattr
,
1109 .listxattr
= xfs_vn_listxattr
,
1110 .update_time
= xfs_vn_update_time
,
1113 static const struct inode_operations xfs_symlink_inode_operations
= {
1114 .readlink
= generic_readlink
,
1115 .follow_link
= xfs_vn_follow_link
,
1116 .put_link
= xfs_vn_put_link
,
1117 .get_acl
= xfs_get_acl
,
1118 .getattr
= xfs_vn_getattr
,
1119 .setattr
= xfs_vn_setattr
,
1120 .setxattr
= generic_setxattr
,
1121 .getxattr
= generic_getxattr
,
1122 .removexattr
= generic_removexattr
,
1123 .listxattr
= xfs_vn_listxattr
,
1124 .update_time
= xfs_vn_update_time
,
1128 xfs_diflags_to_iflags(
1129 struct inode
*inode
,
1130 struct xfs_inode
*ip
)
1132 if (ip
->i_d
.di_flags
& XFS_DIFLAG_IMMUTABLE
)
1133 inode
->i_flags
|= S_IMMUTABLE
;
1135 inode
->i_flags
&= ~S_IMMUTABLE
;
1136 if (ip
->i_d
.di_flags
& XFS_DIFLAG_APPEND
)
1137 inode
->i_flags
|= S_APPEND
;
1139 inode
->i_flags
&= ~S_APPEND
;
1140 if (ip
->i_d
.di_flags
& XFS_DIFLAG_SYNC
)
1141 inode
->i_flags
|= S_SYNC
;
1143 inode
->i_flags
&= ~S_SYNC
;
1144 if (ip
->i_d
.di_flags
& XFS_DIFLAG_NOATIME
)
1145 inode
->i_flags
|= S_NOATIME
;
1147 inode
->i_flags
&= ~S_NOATIME
;
1151 * Initialize the Linux inode, set up the operation vectors and
1154 * When reading existing inodes from disk this is called directly
1155 * from xfs_iget, when creating a new inode it is called from
1156 * xfs_ialloc after setting up the inode.
1158 * We are always called with an uninitialised linux inode here.
1159 * We need to initialise the necessary fields and take a reference
1164 struct xfs_inode
*ip
)
1166 struct inode
*inode
= &ip
->i_vnode
;
1168 inode
->i_ino
= ip
->i_ino
;
1169 inode
->i_state
= I_NEW
;
1171 inode_sb_list_add(inode
);
1172 /* make the inode look hashed for the writeback code */
1173 hlist_add_fake(&inode
->i_hash
);
1175 inode
->i_mode
= ip
->i_d
.di_mode
;
1176 set_nlink(inode
, ip
->i_d
.di_nlink
);
1177 inode
->i_uid
= ip
->i_d
.di_uid
;
1178 inode
->i_gid
= ip
->i_d
.di_gid
;
1180 switch (inode
->i_mode
& S_IFMT
) {
1184 MKDEV(sysv_major(ip
->i_df
.if_u2
.if_rdev
) & 0x1ff,
1185 sysv_minor(ip
->i_df
.if_u2
.if_rdev
));
1192 inode
->i_generation
= ip
->i_d
.di_gen
;
1193 i_size_write(inode
, ip
->i_d
.di_size
);
1194 inode
->i_atime
.tv_sec
= ip
->i_d
.di_atime
.t_sec
;
1195 inode
->i_atime
.tv_nsec
= ip
->i_d
.di_atime
.t_nsec
;
1196 inode
->i_mtime
.tv_sec
= ip
->i_d
.di_mtime
.t_sec
;
1197 inode
->i_mtime
.tv_nsec
= ip
->i_d
.di_mtime
.t_nsec
;
1198 inode
->i_ctime
.tv_sec
= ip
->i_d
.di_ctime
.t_sec
;
1199 inode
->i_ctime
.tv_nsec
= ip
->i_d
.di_ctime
.t_nsec
;
1200 xfs_diflags_to_iflags(inode
, ip
);
1202 switch (inode
->i_mode
& S_IFMT
) {
1204 inode
->i_op
= &xfs_inode_operations
;
1205 inode
->i_fop
= &xfs_file_operations
;
1206 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
1209 if (xfs_sb_version_hasasciici(&XFS_M(inode
->i_sb
)->m_sb
))
1210 inode
->i_op
= &xfs_dir_ci_inode_operations
;
1212 inode
->i_op
= &xfs_dir_inode_operations
;
1213 inode
->i_fop
= &xfs_dir_file_operations
;
1216 inode
->i_op
= &xfs_symlink_inode_operations
;
1217 if (!(ip
->i_df
.if_flags
& XFS_IFINLINE
))
1218 inode
->i_mapping
->a_ops
= &xfs_address_space_operations
;
1221 inode
->i_op
= &xfs_inode_operations
;
1222 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
1227 * If there is no attribute fork no ACL can exist on this inode,
1228 * and it can't have any file capabilities attached to it either.
1230 if (!XFS_IFORK_Q(ip
)) {
1231 inode_has_no_xattr(inode
);
1232 cache_no_acl(inode
);
1235 xfs_iflags_clear(ip
, XFS_INEW
);
1238 unlock_new_inode(inode
);