4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/quotaops.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/tty.h>
16 #include <linux/namei.h>
17 #include <linux/backing-dev.h>
18 #include <linux/capability.h>
19 #include <linux/securebits.h>
20 #include <linux/security.h>
21 #include <linux/mount.h>
22 #include <linux/vfs.h>
23 #include <linux/fcntl.h>
24 #include <asm/uaccess.h>
26 #include <linux/personality.h>
27 #include <linux/pagemap.h>
28 #include <linux/syscalls.h>
29 #include <linux/rcupdate.h>
30 #include <linux/audit.h>
31 #include <linux/falloc.h>
32 #include <linux/fs_struct.h>
36 int vfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
42 if (dentry
->d_sb
->s_op
->statfs
) {
43 memset(buf
, 0, sizeof(*buf
));
44 retval
= security_sb_statfs(dentry
);
47 retval
= dentry
->d_sb
->s_op
->statfs(dentry
, buf
);
48 if (retval
== 0 && buf
->f_frsize
== 0)
49 buf
->f_frsize
= buf
->f_bsize
;
55 EXPORT_SYMBOL(vfs_statfs
);
57 static int vfs_statfs_native(struct dentry
*dentry
, struct statfs
*buf
)
62 retval
= vfs_statfs(dentry
, &st
);
66 if (sizeof(*buf
) == sizeof(st
))
67 memcpy(buf
, &st
, sizeof(st
));
69 if (sizeof buf
->f_blocks
== 4) {
70 if ((st
.f_blocks
| st
.f_bfree
| st
.f_bavail
|
71 st
.f_bsize
| st
.f_frsize
) &
72 0xffffffff00000000ULL
)
75 * f_files and f_ffree may be -1; it's okay to stuff
78 if (st
.f_files
!= -1 &&
79 (st
.f_files
& 0xffffffff00000000ULL
))
81 if (st
.f_ffree
!= -1 &&
82 (st
.f_ffree
& 0xffffffff00000000ULL
))
86 buf
->f_type
= st
.f_type
;
87 buf
->f_bsize
= st
.f_bsize
;
88 buf
->f_blocks
= st
.f_blocks
;
89 buf
->f_bfree
= st
.f_bfree
;
90 buf
->f_bavail
= st
.f_bavail
;
91 buf
->f_files
= st
.f_files
;
92 buf
->f_ffree
= st
.f_ffree
;
93 buf
->f_fsid
= st
.f_fsid
;
94 buf
->f_namelen
= st
.f_namelen
;
95 buf
->f_frsize
= st
.f_frsize
;
96 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
101 static int vfs_statfs64(struct dentry
*dentry
, struct statfs64
*buf
)
106 retval
= vfs_statfs(dentry
, &st
);
110 if (sizeof(*buf
) == sizeof(st
))
111 memcpy(buf
, &st
, sizeof(st
));
113 buf
->f_type
= st
.f_type
;
114 buf
->f_bsize
= st
.f_bsize
;
115 buf
->f_blocks
= st
.f_blocks
;
116 buf
->f_bfree
= st
.f_bfree
;
117 buf
->f_bavail
= st
.f_bavail
;
118 buf
->f_files
= st
.f_files
;
119 buf
->f_ffree
= st
.f_ffree
;
120 buf
->f_fsid
= st
.f_fsid
;
121 buf
->f_namelen
= st
.f_namelen
;
122 buf
->f_frsize
= st
.f_frsize
;
123 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
128 SYSCALL_DEFINE2(statfs
, const char __user
*, pathname
, struct statfs __user
*, buf
)
133 error
= user_path(pathname
, &path
);
136 error
= vfs_statfs_native(path
.dentry
, &tmp
);
137 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
144 SYSCALL_DEFINE3(statfs64
, const char __user
*, pathname
, size_t, sz
, struct statfs64 __user
*, buf
)
149 if (sz
!= sizeof(*buf
))
151 error
= user_path(pathname
, &path
);
154 error
= vfs_statfs64(path
.dentry
, &tmp
);
155 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
162 SYSCALL_DEFINE2(fstatfs
, unsigned int, fd
, struct statfs __user
*, buf
)
172 error
= vfs_statfs_native(file
->f_path
.dentry
, &tmp
);
173 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
180 SYSCALL_DEFINE3(fstatfs64
, unsigned int, fd
, size_t, sz
, struct statfs64 __user
*, buf
)
186 if (sz
!= sizeof(*buf
))
193 error
= vfs_statfs64(file
->f_path
.dentry
, &tmp
);
194 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
201 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
205 struct iattr newattrs
;
207 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
211 newattrs
.ia_size
= length
;
212 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
214 newattrs
.ia_file
= filp
;
215 newattrs
.ia_valid
|= ATTR_FILE
;
218 /* Remove suid/sgid on truncate too */
219 ret
= should_remove_suid(dentry
);
221 newattrs
.ia_valid
|= ret
| ATTR_FORCE
;
223 mutex_lock(&dentry
->d_inode
->i_mutex
);
224 ret
= notify_change(dentry
, &newattrs
);
225 mutex_unlock(&dentry
->d_inode
->i_mutex
);
229 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
236 if (length
< 0) /* sorry, but loff_t says... */
239 error
= user_path(pathname
, &path
);
242 inode
= path
.dentry
->d_inode
;
244 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
246 if (S_ISDIR(inode
->i_mode
))
250 if (!S_ISREG(inode
->i_mode
))
253 error
= mnt_want_write(path
.mnt
);
257 error
= inode_permission(inode
, MAY_WRITE
);
259 goto mnt_drop_write_and_out
;
262 if (IS_APPEND(inode
))
263 goto mnt_drop_write_and_out
;
265 error
= get_write_access(inode
);
267 goto mnt_drop_write_and_out
;
270 * Make sure that there are no leases. get_write_access() protects
271 * against the truncate racing with a lease-granting setlease().
273 error
= break_lease(inode
, FMODE_WRITE
);
275 goto put_write_and_out
;
277 error
= locks_verify_truncate(inode
, NULL
, length
);
279 error
= security_path_truncate(&path
, length
, 0);
282 error
= do_truncate(path
.dentry
, length
, 0, NULL
);
286 put_write_access(inode
);
287 mnt_drop_write_and_out
:
288 mnt_drop_write(path
.mnt
);
295 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
297 return do_sys_truncate(path
, length
);
300 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
302 struct inode
* inode
;
303 struct dentry
*dentry
;
315 /* explicitly opened as large or we are on 64-bit box */
316 if (file
->f_flags
& O_LARGEFILE
)
319 dentry
= file
->f_path
.dentry
;
320 inode
= dentry
->d_inode
;
322 if (!S_ISREG(inode
->i_mode
) || !(file
->f_mode
& FMODE_WRITE
))
326 /* Cannot ftruncate over 2^31 bytes without large file support */
327 if (small
&& length
> MAX_NON_LFS
)
331 if (IS_APPEND(inode
))
334 error
= locks_verify_truncate(inode
, file
, length
);
336 error
= security_path_truncate(&file
->f_path
, length
,
337 ATTR_MTIME
|ATTR_CTIME
);
339 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, file
);
346 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
348 long ret
= do_sys_ftruncate(fd
, length
, 1);
349 /* avoid REGPARM breakage on x86: */
350 asmlinkage_protect(2, ret
, fd
, length
);
354 /* LFS versions of truncate are only needed on 32 bit machines */
355 #if BITS_PER_LONG == 32
356 SYSCALL_DEFINE(truncate64
)(const char __user
* path
, loff_t length
)
358 return do_sys_truncate(path
, length
);
360 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
361 asmlinkage
long SyS_truncate64(long path
, loff_t length
)
363 return SYSC_truncate64((const char __user
*) path
, length
);
365 SYSCALL_ALIAS(sys_truncate64
, SyS_truncate64
);
368 SYSCALL_DEFINE(ftruncate64
)(unsigned int fd
, loff_t length
)
370 long ret
= do_sys_ftruncate(fd
, length
, 0);
371 /* avoid REGPARM breakage on x86: */
372 asmlinkage_protect(2, ret
, fd
, length
);
375 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
376 asmlinkage
long SyS_ftruncate64(long fd
, loff_t length
)
378 return SYSC_ftruncate64((unsigned int) fd
, length
);
380 SYSCALL_ALIAS(sys_ftruncate64
, SyS_ftruncate64
);
382 #endif /* BITS_PER_LONG == 32 */
385 int do_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
387 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
390 if (offset
< 0 || len
<= 0)
393 /* Return error if mode is not supported */
394 if (mode
&& !(mode
& FALLOC_FL_KEEP_SIZE
))
397 if (!(file
->f_mode
& FMODE_WRITE
))
400 * Revalidate the write permissions, in case security policy has
401 * changed since the files were opened.
403 ret
= security_file_permission(file
, MAY_WRITE
);
407 if (S_ISFIFO(inode
->i_mode
))
411 * Let individual file system decide if it supports preallocation
412 * for directories or not.
414 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
))
417 /* Check for wrap through zero too */
418 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
421 if (!inode
->i_op
->fallocate
)
424 return inode
->i_op
->fallocate(inode
, mode
, offset
, len
);
427 SYSCALL_DEFINE(fallocate
)(int fd
, int mode
, loff_t offset
, loff_t len
)
434 error
= do_fallocate(file
, mode
, offset
, len
);
441 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
442 asmlinkage
long SyS_fallocate(long fd
, long mode
, loff_t offset
, loff_t len
)
444 return SYSC_fallocate((int)fd
, (int)mode
, offset
, len
);
446 SYSCALL_ALIAS(sys_fallocate
, SyS_fallocate
);
450 * access() needs to use the real uid/gid, not the effective uid/gid.
451 * We do this by temporarily clearing all FS-related capabilities and
452 * switching the fsuid/fsgid around to the real ones.
454 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
456 const struct cred
*old_cred
;
457 struct cred
*override_cred
;
462 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
465 override_cred
= prepare_creds();
469 override_cred
->fsuid
= override_cred
->uid
;
470 override_cred
->fsgid
= override_cred
->gid
;
472 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
473 /* Clear the capabilities if we switch to a non-root user */
474 if (override_cred
->uid
)
475 cap_clear(override_cred
->cap_effective
);
477 override_cred
->cap_effective
=
478 override_cred
->cap_permitted
;
481 old_cred
= override_creds(override_cred
);
483 res
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
487 inode
= path
.dentry
->d_inode
;
489 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
491 * MAY_EXEC on regular files is denied if the fs is mounted
492 * with the "noexec" flag.
495 if (path
.mnt
->mnt_flags
& MNT_NOEXEC
)
496 goto out_path_release
;
499 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
500 /* SuS v2 requires we report a read only fs too */
501 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
502 goto out_path_release
;
504 * This is a rare case where using __mnt_is_readonly()
505 * is OK without a mnt_want/drop_write() pair. Since
506 * no actual write to the fs is performed here, we do
507 * not need to telegraph to that to anyone.
509 * By doing this, we accept that this access is
510 * inherently racy and know that the fs may change
511 * state before we even see this result.
513 if (__mnt_is_readonly(path
.mnt
))
519 revert_creds(old_cred
);
520 put_cred(override_cred
);
524 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
526 return sys_faccessat(AT_FDCWD
, filename
, mode
);
529 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
534 error
= user_path_dir(filename
, &path
);
538 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
542 set_fs_pwd(current
->fs
, &path
);
550 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
561 inode
= file
->f_path
.dentry
->d_inode
;
564 if (!S_ISDIR(inode
->i_mode
))
567 error
= inode_permission(inode
, MAY_EXEC
| MAY_ACCESS
);
569 set_fs_pwd(current
->fs
, &file
->f_path
);
576 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
581 error
= user_path_dir(filename
, &path
);
585 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
590 if (!capable(CAP_SYS_CHROOT
))
592 error
= security_path_chroot(&path
);
596 set_fs_root(current
->fs
, &path
);
604 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, mode_t
, mode
)
606 struct inode
* inode
;
607 struct dentry
* dentry
;
610 struct iattr newattrs
;
616 dentry
= file
->f_path
.dentry
;
617 inode
= dentry
->d_inode
;
619 audit_inode(NULL
, dentry
);
621 err
= mnt_want_write_file(file
);
624 mutex_lock(&inode
->i_mutex
);
625 err
= security_path_chmod(dentry
, file
->f_vfsmnt
, mode
);
628 if (mode
== (mode_t
) -1)
629 mode
= inode
->i_mode
;
630 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
631 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
632 err
= notify_change(dentry
, &newattrs
);
634 mutex_unlock(&inode
->i_mutex
);
635 mnt_drop_write(file
->f_path
.mnt
);
642 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, mode_t
, mode
)
647 struct iattr newattrs
;
649 error
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
652 inode
= path
.dentry
->d_inode
;
654 error
= mnt_want_write(path
.mnt
);
657 mutex_lock(&inode
->i_mutex
);
658 error
= security_path_chmod(path
.dentry
, path
.mnt
, mode
);
661 if (mode
== (mode_t
) -1)
662 mode
= inode
->i_mode
;
663 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
664 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
665 error
= notify_change(path
.dentry
, &newattrs
);
667 mutex_unlock(&inode
->i_mutex
);
668 mnt_drop_write(path
.mnt
);
675 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, mode_t
, mode
)
677 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
680 static int chown_common(struct path
*path
, uid_t user
, gid_t group
)
682 struct inode
*inode
= path
->dentry
->d_inode
;
684 struct iattr newattrs
;
686 newattrs
.ia_valid
= ATTR_CTIME
;
687 if (user
!= (uid_t
) -1) {
688 newattrs
.ia_valid
|= ATTR_UID
;
689 newattrs
.ia_uid
= user
;
691 if (group
!= (gid_t
) -1) {
692 newattrs
.ia_valid
|= ATTR_GID
;
693 newattrs
.ia_gid
= group
;
695 if (!S_ISDIR(inode
->i_mode
))
697 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
698 mutex_lock(&inode
->i_mutex
);
699 error
= security_path_chown(path
, user
, group
);
701 error
= notify_change(path
->dentry
, &newattrs
);
702 mutex_unlock(&inode
->i_mutex
);
707 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
712 error
= user_path(filename
, &path
);
715 error
= mnt_want_write(path
.mnt
);
718 error
= chown_common(&path
, user
, group
);
719 mnt_drop_write(path
.mnt
);
726 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
727 gid_t
, group
, int, flag
)
733 if ((flag
& ~AT_SYMLINK_NOFOLLOW
) != 0)
736 follow
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
737 error
= user_path_at(dfd
, filename
, follow
, &path
);
740 error
= mnt_want_write(path
.mnt
);
743 error
= chown_common(&path
, user
, group
);
744 mnt_drop_write(path
.mnt
);
751 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
756 error
= user_lpath(filename
, &path
);
759 error
= mnt_want_write(path
.mnt
);
762 error
= chown_common(&path
, user
, group
);
763 mnt_drop_write(path
.mnt
);
770 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
774 struct dentry
* dentry
;
780 error
= mnt_want_write_file(file
);
783 dentry
= file
->f_path
.dentry
;
784 audit_inode(NULL
, dentry
);
785 error
= chown_common(&file
->f_path
, user
, group
);
786 mnt_drop_write(file
->f_path
.mnt
);
794 * You have to be very careful that these write
795 * counts get cleaned up in error cases and
796 * upon __fput(). This should probably never
797 * be called outside of __dentry_open().
799 static inline int __get_file_write_access(struct inode
*inode
,
800 struct vfsmount
*mnt
)
803 error
= get_write_access(inode
);
807 * Do not take mount writer counts on
808 * special files since no writes to
809 * the mount itself will occur.
811 if (!special_file(inode
->i_mode
)) {
813 * Balanced in __fput()
815 error
= mnt_want_write(mnt
);
817 put_write_access(inode
);
822 static struct file
*__dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
,
823 int flags
, struct file
*f
,
824 int (*open
)(struct inode
*, struct file
*),
825 const struct cred
*cred
)
831 f
->f_mode
= (__force fmode_t
)((flags
+1) & O_ACCMODE
) | FMODE_LSEEK
|
832 FMODE_PREAD
| FMODE_PWRITE
;
833 inode
= dentry
->d_inode
;
834 if (f
->f_mode
& FMODE_WRITE
) {
835 error
= __get_file_write_access(inode
, mnt
);
838 if (!special_file(inode
->i_mode
))
842 f
->f_mapping
= inode
->i_mapping
;
843 f
->f_path
.dentry
= dentry
;
846 f
->f_op
= fops_get(inode
->i_fop
);
847 file_move(f
, &inode
->i_sb
->s_files
);
849 error
= security_dentry_open(f
, cred
);
853 if (!open
&& f
->f_op
)
854 open
= f
->f_op
->open
;
856 error
= open(inode
, f
);
861 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
863 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
865 /* NB: we're sure to have correct a_ops only after f_op->open */
866 if (f
->f_flags
& O_DIRECT
) {
867 if (!f
->f_mapping
->a_ops
||
868 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
869 (!f
->f_mapping
->a_ops
->get_xip_mem
))) {
871 f
= ERR_PTR(-EINVAL
);
879 if (f
->f_mode
& FMODE_WRITE
) {
880 put_write_access(inode
);
881 if (!special_file(inode
->i_mode
)) {
883 * We don't consider this a real
884 * mnt_want/drop_write() pair
885 * because it all happenend right
886 * here, so just reset the state.
893 f
->f_path
.dentry
= NULL
;
894 f
->f_path
.mnt
= NULL
;
899 return ERR_PTR(error
);
903 * lookup_instantiate_filp - instantiates the open intent filp
904 * @nd: pointer to nameidata
905 * @dentry: pointer to dentry
906 * @open: open callback
908 * Helper for filesystems that want to use lookup open intents and pass back
909 * a fully instantiated struct file to the caller.
910 * This function is meant to be called from within a filesystem's
912 * Beware of calling it for non-regular files! Those ->open methods might block
913 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
914 * leading to a deadlock, as nobody can open that fifo anymore, because
915 * another process to open fifo will block on locked parent when doing lookup).
916 * Note that in case of error, nd->intent.open.file is destroyed, but the
917 * path information remains valid.
918 * If the open callback is set to NULL, then the standard f_op->open()
919 * filesystem callback is substituted.
921 struct file
*lookup_instantiate_filp(struct nameidata
*nd
, struct dentry
*dentry
,
922 int (*open
)(struct inode
*, struct file
*))
924 const struct cred
*cred
= current_cred();
926 if (IS_ERR(nd
->intent
.open
.file
))
930 nd
->intent
.open
.file
= __dentry_open(dget(dentry
), mntget(nd
->path
.mnt
),
931 nd
->intent
.open
.flags
- 1,
932 nd
->intent
.open
.file
,
935 return nd
->intent
.open
.file
;
937 release_open_intent(nd
);
938 nd
->intent
.open
.file
= (struct file
*)dentry
;
941 EXPORT_SYMBOL_GPL(lookup_instantiate_filp
);
944 * nameidata_to_filp - convert a nameidata to an open filp.
945 * @nd: pointer to nameidata
948 * Note that this function destroys the original nameidata
950 struct file
*nameidata_to_filp(struct nameidata
*nd
, int flags
)
952 const struct cred
*cred
= current_cred();
955 /* Pick up the filp from the open intent */
956 filp
= nd
->intent
.open
.file
;
957 /* Has the filesystem initialised the file for us? */
958 if (filp
->f_path
.dentry
== NULL
)
959 filp
= __dentry_open(nd
->path
.dentry
, nd
->path
.mnt
, flags
, filp
,
967 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
970 struct file
*dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
, int flags
,
971 const struct cred
*cred
)
976 validate_creds(cred
);
979 * We must always pass in a valid mount pointer. Historically
980 * callers got away with not passing it, but we must enforce this at
981 * the earliest possible point now to avoid strange problems deep in the
985 printk(KERN_WARNING
"%s called with NULL vfsmount\n", __func__
);
987 return ERR_PTR(-EINVAL
);
991 f
= get_empty_filp();
995 return ERR_PTR(error
);
998 return __dentry_open(dentry
, mnt
, flags
, f
, NULL
, cred
);
1000 EXPORT_SYMBOL(dentry_open
);
1002 static void __put_unused_fd(struct files_struct
*files
, unsigned int fd
)
1004 struct fdtable
*fdt
= files_fdtable(files
);
1005 __FD_CLR(fd
, fdt
->open_fds
);
1006 if (fd
< files
->next_fd
)
1007 files
->next_fd
= fd
;
1010 void put_unused_fd(unsigned int fd
)
1012 struct files_struct
*files
= current
->files
;
1013 spin_lock(&files
->file_lock
);
1014 __put_unused_fd(files
, fd
);
1015 spin_unlock(&files
->file_lock
);
1018 EXPORT_SYMBOL(put_unused_fd
);
1021 * Install a file pointer in the fd array.
1023 * The VFS is full of places where we drop the files lock between
1024 * setting the open_fds bitmap and installing the file in the file
1025 * array. At any such point, we are vulnerable to a dup2() race
1026 * installing a file in the array before us. We need to detect this and
1027 * fput() the struct file we are about to overwrite in this case.
1029 * It should never happen - if we allow dup2() do it, _really_ bad things
1033 void fd_install(unsigned int fd
, struct file
*file
)
1035 struct files_struct
*files
= current
->files
;
1036 struct fdtable
*fdt
;
1037 spin_lock(&files
->file_lock
);
1038 fdt
= files_fdtable(files
);
1039 BUG_ON(fdt
->fd
[fd
] != NULL
);
1040 rcu_assign_pointer(fdt
->fd
[fd
], file
);
1041 spin_unlock(&files
->file_lock
);
1044 EXPORT_SYMBOL(fd_install
);
1046 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, int mode
)
1048 char *tmp
= getname(filename
);
1049 int fd
= PTR_ERR(tmp
);
1052 fd
= get_unused_fd_flags(flags
);
1054 struct file
*f
= do_filp_open(dfd
, tmp
, flags
, mode
, 0);
1059 fsnotify_open(f
->f_path
.dentry
);
1068 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, int, mode
)
1072 if (force_o_largefile())
1073 flags
|= O_LARGEFILE
;
1075 ret
= do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1076 /* avoid REGPARM breakage on x86: */
1077 asmlinkage_protect(3, ret
, filename
, flags
, mode
);
1081 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1086 if (force_o_largefile())
1087 flags
|= O_LARGEFILE
;
1089 ret
= do_sys_open(dfd
, filename
, flags
, mode
);
1090 /* avoid REGPARM breakage on x86: */
1091 asmlinkage_protect(4, ret
, dfd
, filename
, flags
, mode
);
1098 * For backward compatibility? Maybe this should be moved
1099 * into arch/i386 instead?
1101 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, int, mode
)
1103 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1109 * "id" is the POSIX thread ID. We use the
1110 * files pointer for this..
1112 int filp_close(struct file
*filp
, fl_owner_t id
)
1116 if (!file_count(filp
)) {
1117 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1121 if (filp
->f_op
&& filp
->f_op
->flush
)
1122 retval
= filp
->f_op
->flush(filp
, id
);
1124 dnotify_flush(filp
, id
);
1125 locks_remove_posix(filp
, id
);
1130 EXPORT_SYMBOL(filp_close
);
1133 * Careful here! We test whether the file pointer is NULL before
1134 * releasing the fd. This ensures that one clone task can't release
1135 * an fd while another clone is opening it.
1137 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1140 struct files_struct
*files
= current
->files
;
1141 struct fdtable
*fdt
;
1144 spin_lock(&files
->file_lock
);
1145 fdt
= files_fdtable(files
);
1146 if (fd
>= fdt
->max_fds
)
1151 rcu_assign_pointer(fdt
->fd
[fd
], NULL
);
1152 FD_CLR(fd
, fdt
->close_on_exec
);
1153 __put_unused_fd(files
, fd
);
1154 spin_unlock(&files
->file_lock
);
1155 retval
= filp_close(filp
, files
);
1157 /* can't restart close syscall because file table entry was cleared */
1158 if (unlikely(retval
== -ERESTARTSYS
||
1159 retval
== -ERESTARTNOINTR
||
1160 retval
== -ERESTARTNOHAND
||
1161 retval
== -ERESTART_RESTARTBLOCK
))
1167 spin_unlock(&files
->file_lock
);
1170 EXPORT_SYMBOL(sys_close
);
1173 * This routine simulates a hangup on the tty, to arrange that users
1174 * are given clean terminals at login time.
1176 SYSCALL_DEFINE0(vhangup
)
1178 if (capable(CAP_SYS_TTY_CONFIG
)) {
1186 * Called when an inode is about to be open.
1187 * We use this to disallow opening large files on 32bit systems if
1188 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1189 * on this flag in sys_open.
1191 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1193 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1198 EXPORT_SYMBOL(generic_file_open
);
1201 * This is used by subsystems that don't want seekable
1204 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1206 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1210 EXPORT_SYMBOL(nonseekable_open
);