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>
33 int vfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
39 if (dentry
->d_sb
->s_op
->statfs
) {
40 memset(buf
, 0, sizeof(*buf
));
41 retval
= security_sb_statfs(dentry
);
44 retval
= dentry
->d_sb
->s_op
->statfs(dentry
, buf
);
45 if (retval
== 0 && buf
->f_frsize
== 0)
46 buf
->f_frsize
= buf
->f_bsize
;
52 EXPORT_SYMBOL(vfs_statfs
);
54 static int vfs_statfs_native(struct dentry
*dentry
, struct statfs
*buf
)
59 retval
= vfs_statfs(dentry
, &st
);
63 if (sizeof(*buf
) == sizeof(st
))
64 memcpy(buf
, &st
, sizeof(st
));
66 if (sizeof buf
->f_blocks
== 4) {
67 if ((st
.f_blocks
| st
.f_bfree
| st
.f_bavail
|
68 st
.f_bsize
| st
.f_frsize
) &
69 0xffffffff00000000ULL
)
72 * f_files and f_ffree may be -1; it's okay to stuff
75 if (st
.f_files
!= -1 &&
76 (st
.f_files
& 0xffffffff00000000ULL
))
78 if (st
.f_ffree
!= -1 &&
79 (st
.f_ffree
& 0xffffffff00000000ULL
))
83 buf
->f_type
= st
.f_type
;
84 buf
->f_bsize
= st
.f_bsize
;
85 buf
->f_blocks
= st
.f_blocks
;
86 buf
->f_bfree
= st
.f_bfree
;
87 buf
->f_bavail
= st
.f_bavail
;
88 buf
->f_files
= st
.f_files
;
89 buf
->f_ffree
= st
.f_ffree
;
90 buf
->f_fsid
= st
.f_fsid
;
91 buf
->f_namelen
= st
.f_namelen
;
92 buf
->f_frsize
= st
.f_frsize
;
93 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
98 static int vfs_statfs64(struct dentry
*dentry
, struct statfs64
*buf
)
103 retval
= vfs_statfs(dentry
, &st
);
107 if (sizeof(*buf
) == sizeof(st
))
108 memcpy(buf
, &st
, sizeof(st
));
110 buf
->f_type
= st
.f_type
;
111 buf
->f_bsize
= st
.f_bsize
;
112 buf
->f_blocks
= st
.f_blocks
;
113 buf
->f_bfree
= st
.f_bfree
;
114 buf
->f_bavail
= st
.f_bavail
;
115 buf
->f_files
= st
.f_files
;
116 buf
->f_ffree
= st
.f_ffree
;
117 buf
->f_fsid
= st
.f_fsid
;
118 buf
->f_namelen
= st
.f_namelen
;
119 buf
->f_frsize
= st
.f_frsize
;
120 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
125 SYSCALL_DEFINE2(statfs
, const char __user
*, pathname
, struct statfs __user
*, buf
)
130 error
= user_path(pathname
, &path
);
133 error
= vfs_statfs_native(path
.dentry
, &tmp
);
134 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
141 SYSCALL_DEFINE3(statfs64
, const char __user
*, pathname
, size_t, sz
, struct statfs64 __user
*, buf
)
146 if (sz
!= sizeof(*buf
))
148 error
= user_path(pathname
, &path
);
151 error
= vfs_statfs64(path
.dentry
, &tmp
);
152 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
159 SYSCALL_DEFINE2(fstatfs
, unsigned int, fd
, struct statfs __user
*, buf
)
169 error
= vfs_statfs_native(file
->f_path
.dentry
, &tmp
);
170 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
177 SYSCALL_DEFINE3(fstatfs64
, unsigned int, fd
, size_t, sz
, struct statfs64 __user
*, buf
)
183 if (sz
!= sizeof(*buf
))
190 error
= vfs_statfs64(file
->f_path
.dentry
, &tmp
);
191 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
198 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
202 struct iattr newattrs
;
204 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
208 newattrs
.ia_size
= length
;
209 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
211 newattrs
.ia_file
= filp
;
212 newattrs
.ia_valid
|= ATTR_FILE
;
215 /* Remove suid/sgid on truncate too */
216 newattrs
.ia_valid
|= should_remove_suid(dentry
);
218 mutex_lock(&dentry
->d_inode
->i_mutex
);
219 err
= notify_change(dentry
, &newattrs
);
220 mutex_unlock(&dentry
->d_inode
->i_mutex
);
224 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
231 if (length
< 0) /* sorry, but loff_t says... */
234 error
= user_path(pathname
, &path
);
237 inode
= path
.dentry
->d_inode
;
239 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
241 if (S_ISDIR(inode
->i_mode
))
245 if (!S_ISREG(inode
->i_mode
))
248 error
= mnt_want_write(path
.mnt
);
252 error
= inode_permission(inode
, MAY_WRITE
);
254 goto mnt_drop_write_and_out
;
257 if (IS_APPEND(inode
))
258 goto mnt_drop_write_and_out
;
260 error
= get_write_access(inode
);
262 goto mnt_drop_write_and_out
;
265 * Make sure that there are no leases. get_write_access() protects
266 * against the truncate racing with a lease-granting setlease().
268 error
= break_lease(inode
, FMODE_WRITE
);
270 goto put_write_and_out
;
272 error
= locks_verify_truncate(inode
, NULL
, length
);
275 error
= do_truncate(path
.dentry
, length
, 0, NULL
);
279 put_write_access(inode
);
280 mnt_drop_write_and_out
:
281 mnt_drop_write(path
.mnt
);
288 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, unsigned long, length
)
290 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
291 return do_sys_truncate(path
, (long)length
);
294 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
296 struct inode
* inode
;
297 struct dentry
*dentry
;
309 /* explicitly opened as large or we are on 64-bit box */
310 if (file
->f_flags
& O_LARGEFILE
)
313 dentry
= file
->f_path
.dentry
;
314 inode
= dentry
->d_inode
;
316 if (!S_ISREG(inode
->i_mode
) || !(file
->f_mode
& FMODE_WRITE
))
320 /* Cannot ftruncate over 2^31 bytes without large file support */
321 if (small
&& length
> MAX_NON_LFS
)
325 if (IS_APPEND(inode
))
328 error
= locks_verify_truncate(inode
, file
, length
);
330 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, file
);
337 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
339 long ret
= do_sys_ftruncate(fd
, length
, 1);
340 /* avoid REGPARM breakage on x86: */
341 asmlinkage_protect(2, ret
, fd
, length
);
345 /* LFS versions of truncate are only needed on 32 bit machines */
346 #if BITS_PER_LONG == 32
347 SYSCALL_DEFINE(truncate64
)(const char __user
* path
, loff_t length
)
349 return do_sys_truncate(path
, length
);
351 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
352 asmlinkage
long SyS_truncate64(long path
, loff_t length
)
354 return SYSC_truncate64((const char __user
*) path
, length
);
356 SYSCALL_ALIAS(sys_truncate64
, SyS_truncate64
);
359 SYSCALL_DEFINE(ftruncate64
)(unsigned int fd
, loff_t length
)
361 long ret
= do_sys_ftruncate(fd
, length
, 0);
362 /* avoid REGPARM breakage on x86: */
363 asmlinkage_protect(2, ret
, fd
, length
);
366 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
367 asmlinkage
long SyS_ftruncate64(long fd
, loff_t length
)
369 return SYSC_ftruncate64((unsigned int) fd
, length
);
371 SYSCALL_ALIAS(sys_ftruncate64
, SyS_ftruncate64
);
373 #endif /* BITS_PER_LONG == 32 */
375 SYSCALL_DEFINE(fallocate
)(int fd
, int mode
, loff_t offset
, loff_t len
)
381 if (offset
< 0 || len
<= 0)
384 /* Return error if mode is not supported */
386 if (mode
&& !(mode
& FALLOC_FL_KEEP_SIZE
))
393 if (!(file
->f_mode
& FMODE_WRITE
))
396 * Revalidate the write permissions, in case security policy has
397 * changed since the files were opened.
399 ret
= security_file_permission(file
, MAY_WRITE
);
403 inode
= file
->f_path
.dentry
->d_inode
;
406 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
))
418 /* Check for wrap through zero too */
419 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
422 if (inode
->i_op
&& inode
->i_op
->fallocate
)
423 ret
= inode
->i_op
->fallocate(inode
, mode
, offset
, len
);
432 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
433 asmlinkage
long SyS_fallocate(long fd
, long mode
, loff_t offset
, loff_t len
)
435 return SYSC_fallocate((int)fd
, (int)mode
, offset
, len
);
437 SYSCALL_ALIAS(sys_fallocate
, SyS_fallocate
);
441 * access() needs to use the real uid/gid, not the effective uid/gid.
442 * We do this by temporarily clearing all FS-related capabilities and
443 * switching the fsuid/fsgid around to the real ones.
445 asmlinkage
long sys_faccessat(int dfd
, const char __user
*filename
, int mode
)
449 int old_fsuid
, old_fsgid
;
450 kernel_cap_t
uninitialized_var(old_cap
); /* !SECURE_NO_SETUID_FIXUP */
453 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
456 old_fsuid
= current
->fsuid
;
457 old_fsgid
= current
->fsgid
;
459 current
->fsuid
= current
->uid
;
460 current
->fsgid
= current
->gid
;
462 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
464 * Clear the capabilities if we switch to a non-root user
466 #ifndef CONFIG_SECURITY_FILE_CAPABILITIES
468 * FIXME: There is a race here against sys_capset. The
469 * capabilities can change yet we will restore the old
470 * value below. We should hold task_capabilities_lock,
471 * but we cannot because user_path_at can sleep.
473 #endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
475 old_cap
= cap_set_effective(__cap_empty_set
);
477 old_cap
= cap_set_effective(current
->cap_permitted
);
480 res
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
484 inode
= path
.dentry
->d_inode
;
486 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
488 * MAY_EXEC on regular files is denied if the fs is mounted
489 * with the "noexec" flag.
492 if (path
.mnt
->mnt_flags
& MNT_NOEXEC
)
493 goto out_path_release
;
496 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
497 /* SuS v2 requires we report a read only fs too */
498 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
499 goto out_path_release
;
501 * This is a rare case where using __mnt_is_readonly()
502 * is OK without a mnt_want/drop_write() pair. Since
503 * no actual write to the fs is performed here, we do
504 * not need to telegraph to that to anyone.
506 * By doing this, we accept that this access is
507 * inherently racy and know that the fs may change
508 * state before we even see this result.
510 if (__mnt_is_readonly(path
.mnt
))
516 current
->fsuid
= old_fsuid
;
517 current
->fsgid
= old_fsgid
;
519 if (!issecure(SECURE_NO_SETUID_FIXUP
))
520 cap_set_effective(old_cap
);
525 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
527 return sys_faccessat(AT_FDCWD
, filename
, mode
);
530 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
535 error
= user_path_dir(filename
, &path
);
539 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
543 set_fs_pwd(current
->fs
, &path
);
551 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
562 inode
= file
->f_path
.dentry
->d_inode
;
565 if (!S_ISDIR(inode
->i_mode
))
568 error
= inode_permission(inode
, MAY_EXEC
| MAY_ACCESS
);
570 set_fs_pwd(current
->fs
, &file
->f_path
);
577 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
582 error
= user_path_dir(filename
, &path
);
586 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
591 if (!capable(CAP_SYS_CHROOT
))
594 set_fs_root(current
->fs
, &path
);
602 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, mode_t
, mode
)
604 struct inode
* inode
;
605 struct dentry
* dentry
;
608 struct iattr newattrs
;
614 dentry
= file
->f_path
.dentry
;
615 inode
= dentry
->d_inode
;
617 audit_inode(NULL
, dentry
);
619 err
= mnt_want_write(file
->f_path
.mnt
);
622 mutex_lock(&inode
->i_mutex
);
623 if (mode
== (mode_t
) -1)
624 mode
= inode
->i_mode
;
625 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
626 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
627 err
= notify_change(dentry
, &newattrs
);
628 mutex_unlock(&inode
->i_mutex
);
629 mnt_drop_write(file
->f_path
.mnt
);
636 asmlinkage
long sys_fchmodat(int dfd
, const char __user
*filename
,
642 struct iattr newattrs
;
644 error
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
647 inode
= path
.dentry
->d_inode
;
649 error
= mnt_want_write(path
.mnt
);
652 mutex_lock(&inode
->i_mutex
);
653 if (mode
== (mode_t
) -1)
654 mode
= inode
->i_mode
;
655 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
656 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
657 error
= notify_change(path
.dentry
, &newattrs
);
658 mutex_unlock(&inode
->i_mutex
);
659 mnt_drop_write(path
.mnt
);
666 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, mode_t
, mode
)
668 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
671 static int chown_common(struct dentry
* dentry
, uid_t user
, gid_t group
)
673 struct inode
*inode
= dentry
->d_inode
;
675 struct iattr newattrs
;
677 newattrs
.ia_valid
= ATTR_CTIME
;
678 if (user
!= (uid_t
) -1) {
679 newattrs
.ia_valid
|= ATTR_UID
;
680 newattrs
.ia_uid
= user
;
682 if (group
!= (gid_t
) -1) {
683 newattrs
.ia_valid
|= ATTR_GID
;
684 newattrs
.ia_gid
= group
;
686 if (!S_ISDIR(inode
->i_mode
))
688 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
689 mutex_lock(&inode
->i_mutex
);
690 error
= notify_change(dentry
, &newattrs
);
691 mutex_unlock(&inode
->i_mutex
);
696 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
701 error
= user_path(filename
, &path
);
704 error
= mnt_want_write(path
.mnt
);
707 error
= chown_common(path
.dentry
, user
, group
);
708 mnt_drop_write(path
.mnt
);
715 asmlinkage
long sys_fchownat(int dfd
, const char __user
*filename
, uid_t user
,
716 gid_t group
, int flag
)
722 if ((flag
& ~AT_SYMLINK_NOFOLLOW
) != 0)
725 follow
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
726 error
= user_path_at(dfd
, filename
, follow
, &path
);
729 error
= mnt_want_write(path
.mnt
);
732 error
= chown_common(path
.dentry
, user
, group
);
733 mnt_drop_write(path
.mnt
);
740 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
745 error
= user_lpath(filename
, &path
);
748 error
= mnt_want_write(path
.mnt
);
751 error
= chown_common(path
.dentry
, user
, group
);
752 mnt_drop_write(path
.mnt
);
759 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
763 struct dentry
* dentry
;
769 error
= mnt_want_write(file
->f_path
.mnt
);
772 dentry
= file
->f_path
.dentry
;
773 audit_inode(NULL
, dentry
);
774 error
= chown_common(dentry
, user
, group
);
775 mnt_drop_write(file
->f_path
.mnt
);
783 * You have to be very careful that these write
784 * counts get cleaned up in error cases and
785 * upon __fput(). This should probably never
786 * be called outside of __dentry_open().
788 static inline int __get_file_write_access(struct inode
*inode
,
789 struct vfsmount
*mnt
)
792 error
= get_write_access(inode
);
796 * Do not take mount writer counts on
797 * special files since no writes to
798 * the mount itself will occur.
800 if (!special_file(inode
->i_mode
)) {
802 * Balanced in __fput()
804 error
= mnt_want_write(mnt
);
806 put_write_access(inode
);
811 static struct file
*__dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
,
812 int flags
, struct file
*f
,
813 int (*open
)(struct inode
*, struct file
*))
819 f
->f_mode
= ((flags
+1) & O_ACCMODE
) | FMODE_LSEEK
|
820 FMODE_PREAD
| FMODE_PWRITE
;
821 inode
= dentry
->d_inode
;
822 if (f
->f_mode
& FMODE_WRITE
) {
823 error
= __get_file_write_access(inode
, mnt
);
826 if (!special_file(inode
->i_mode
))
830 f
->f_mapping
= inode
->i_mapping
;
831 f
->f_path
.dentry
= dentry
;
834 f
->f_op
= fops_get(inode
->i_fop
);
835 file_move(f
, &inode
->i_sb
->s_files
);
837 error
= security_dentry_open(f
);
841 if (!open
&& f
->f_op
)
842 open
= f
->f_op
->open
;
844 error
= open(inode
, f
);
849 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
851 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
853 /* NB: we're sure to have correct a_ops only after f_op->open */
854 if (f
->f_flags
& O_DIRECT
) {
855 if (!f
->f_mapping
->a_ops
||
856 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
857 (!f
->f_mapping
->a_ops
->get_xip_mem
))) {
859 f
= ERR_PTR(-EINVAL
);
867 if (f
->f_mode
& FMODE_WRITE
) {
868 put_write_access(inode
);
869 if (!special_file(inode
->i_mode
)) {
871 * We don't consider this a real
872 * mnt_want/drop_write() pair
873 * because it all happenend right
874 * here, so just reset the state.
881 f
->f_path
.dentry
= NULL
;
882 f
->f_path
.mnt
= NULL
;
887 return ERR_PTR(error
);
891 * lookup_instantiate_filp - instantiates the open intent filp
892 * @nd: pointer to nameidata
893 * @dentry: pointer to dentry
894 * @open: open callback
896 * Helper for filesystems that want to use lookup open intents and pass back
897 * a fully instantiated struct file to the caller.
898 * This function is meant to be called from within a filesystem's
900 * Beware of calling it for non-regular files! Those ->open methods might block
901 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
902 * leading to a deadlock, as nobody can open that fifo anymore, because
903 * another process to open fifo will block on locked parent when doing lookup).
904 * Note that in case of error, nd->intent.open.file is destroyed, but the
905 * path information remains valid.
906 * If the open callback is set to NULL, then the standard f_op->open()
907 * filesystem callback is substituted.
909 struct file
*lookup_instantiate_filp(struct nameidata
*nd
, struct dentry
*dentry
,
910 int (*open
)(struct inode
*, struct file
*))
912 if (IS_ERR(nd
->intent
.open
.file
))
916 nd
->intent
.open
.file
= __dentry_open(dget(dentry
), mntget(nd
->path
.mnt
),
917 nd
->intent
.open
.flags
- 1,
918 nd
->intent
.open
.file
,
921 return nd
->intent
.open
.file
;
923 release_open_intent(nd
);
924 nd
->intent
.open
.file
= (struct file
*)dentry
;
927 EXPORT_SYMBOL_GPL(lookup_instantiate_filp
);
930 * nameidata_to_filp - convert a nameidata to an open filp.
931 * @nd: pointer to nameidata
934 * Note that this function destroys the original nameidata
936 struct file
*nameidata_to_filp(struct nameidata
*nd
, int flags
)
940 /* Pick up the filp from the open intent */
941 filp
= nd
->intent
.open
.file
;
942 /* Has the filesystem initialised the file for us? */
943 if (filp
->f_path
.dentry
== NULL
)
944 filp
= __dentry_open(nd
->path
.dentry
, nd
->path
.mnt
, flags
, filp
,
952 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
955 struct file
*dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
, int flags
)
961 * We must always pass in a valid mount pointer. Historically
962 * callers got away with not passing it, but we must enforce this at
963 * the earliest possible point now to avoid strange problems deep in the
967 printk(KERN_WARNING
"%s called with NULL vfsmount\n", __func__
);
969 return ERR_PTR(-EINVAL
);
973 f
= get_empty_filp();
977 return ERR_PTR(error
);
980 return __dentry_open(dentry
, mnt
, flags
, f
, NULL
);
982 EXPORT_SYMBOL(dentry_open
);
984 static void __put_unused_fd(struct files_struct
*files
, unsigned int fd
)
986 struct fdtable
*fdt
= files_fdtable(files
);
987 __FD_CLR(fd
, fdt
->open_fds
);
988 if (fd
< files
->next_fd
)
992 void put_unused_fd(unsigned int fd
)
994 struct files_struct
*files
= current
->files
;
995 spin_lock(&files
->file_lock
);
996 __put_unused_fd(files
, fd
);
997 spin_unlock(&files
->file_lock
);
1000 EXPORT_SYMBOL(put_unused_fd
);
1003 * Install a file pointer in the fd array.
1005 * The VFS is full of places where we drop the files lock between
1006 * setting the open_fds bitmap and installing the file in the file
1007 * array. At any such point, we are vulnerable to a dup2() race
1008 * installing a file in the array before us. We need to detect this and
1009 * fput() the struct file we are about to overwrite in this case.
1011 * It should never happen - if we allow dup2() do it, _really_ bad things
1015 void fd_install(unsigned int fd
, struct file
*file
)
1017 struct files_struct
*files
= current
->files
;
1018 struct fdtable
*fdt
;
1019 spin_lock(&files
->file_lock
);
1020 fdt
= files_fdtable(files
);
1021 BUG_ON(fdt
->fd
[fd
] != NULL
);
1022 rcu_assign_pointer(fdt
->fd
[fd
], file
);
1023 spin_unlock(&files
->file_lock
);
1026 EXPORT_SYMBOL(fd_install
);
1028 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, int mode
)
1030 char *tmp
= getname(filename
);
1031 int fd
= PTR_ERR(tmp
);
1034 fd
= get_unused_fd_flags(flags
);
1036 struct file
*f
= do_filp_open(dfd
, tmp
, flags
, mode
);
1041 fsnotify_open(f
->f_path
.dentry
);
1050 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, int, mode
)
1054 if (force_o_largefile())
1055 flags
|= O_LARGEFILE
;
1057 ret
= do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1058 /* avoid REGPARM breakage on x86: */
1059 asmlinkage_protect(3, ret
, filename
, flags
, mode
);
1063 asmlinkage
long sys_openat(int dfd
, const char __user
*filename
, int flags
,
1068 if (force_o_largefile())
1069 flags
|= O_LARGEFILE
;
1071 ret
= do_sys_open(dfd
, filename
, flags
, mode
);
1072 /* avoid REGPARM breakage on x86: */
1073 asmlinkage_protect(4, ret
, dfd
, filename
, flags
, mode
);
1080 * For backward compatibility? Maybe this should be moved
1081 * into arch/i386 instead?
1083 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, int, mode
)
1085 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1091 * "id" is the POSIX thread ID. We use the
1092 * files pointer for this..
1094 int filp_close(struct file
*filp
, fl_owner_t id
)
1098 if (!file_count(filp
)) {
1099 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1103 if (filp
->f_op
&& filp
->f_op
->flush
)
1104 retval
= filp
->f_op
->flush(filp
, id
);
1106 dnotify_flush(filp
, id
);
1107 locks_remove_posix(filp
, id
);
1112 EXPORT_SYMBOL(filp_close
);
1115 * Careful here! We test whether the file pointer is NULL before
1116 * releasing the fd. This ensures that one clone task can't release
1117 * an fd while another clone is opening it.
1119 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1122 struct files_struct
*files
= current
->files
;
1123 struct fdtable
*fdt
;
1126 spin_lock(&files
->file_lock
);
1127 fdt
= files_fdtable(files
);
1128 if (fd
>= fdt
->max_fds
)
1133 rcu_assign_pointer(fdt
->fd
[fd
], NULL
);
1134 FD_CLR(fd
, fdt
->close_on_exec
);
1135 __put_unused_fd(files
, fd
);
1136 spin_unlock(&files
->file_lock
);
1137 retval
= filp_close(filp
, files
);
1139 /* can't restart close syscall because file table entry was cleared */
1140 if (unlikely(retval
== -ERESTARTSYS
||
1141 retval
== -ERESTARTNOINTR
||
1142 retval
== -ERESTARTNOHAND
||
1143 retval
== -ERESTART_RESTARTBLOCK
))
1149 spin_unlock(&files
->file_lock
);
1152 EXPORT_SYMBOL(sys_close
);
1155 * This routine simulates a hangup on the tty, to arrange that users
1156 * are given clean terminals at login time.
1158 SYSCALL_DEFINE0(vhangup
)
1160 if (capable(CAP_SYS_TTY_CONFIG
)) {
1161 /* XXX: this needs locking */
1162 tty_vhangup(current
->signal
->tty
);
1169 * Called when an inode is about to be open.
1170 * We use this to disallow opening large files on 32bit systems if
1171 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1172 * on this flag in sys_open.
1174 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1176 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1181 EXPORT_SYMBOL(generic_file_open
);
1184 * This is used by subsystems that don't want seekable
1187 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1189 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1193 EXPORT_SYMBOL(nonseekable_open
);