4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/utime.h>
10 #include <linux/file.h>
11 #include <linux/smp_lock.h>
12 #include <linux/quotaops.h>
13 #include <linux/fsnotify.h>
14 #include <linux/module.h>
15 #include <linux/slab.h>
16 #include <linux/tty.h>
17 #include <linux/namei.h>
18 #include <linux/backing-dev.h>
19 #include <linux/capability.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>
32 #include <asm/unistd.h>
34 int vfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
40 if (dentry
->d_sb
->s_op
->statfs
) {
41 memset(buf
, 0, sizeof(*buf
));
42 retval
= security_sb_statfs(dentry
);
45 retval
= dentry
->d_sb
->s_op
->statfs(dentry
, buf
);
46 if (retval
== 0 && buf
->f_frsize
== 0)
47 buf
->f_frsize
= buf
->f_bsize
;
53 EXPORT_SYMBOL(vfs_statfs
);
55 static int vfs_statfs_native(struct dentry
*dentry
, struct statfs
*buf
)
60 retval
= vfs_statfs(dentry
, &st
);
64 if (sizeof(*buf
) == sizeof(st
))
65 memcpy(buf
, &st
, sizeof(st
));
67 if (sizeof buf
->f_blocks
== 4) {
68 if ((st
.f_blocks
| st
.f_bfree
| st
.f_bavail
) &
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 asmlinkage
long sys_statfs(const char __user
* path
, struct statfs __user
* buf
)
130 error
= user_path_walk(path
, &nd
);
133 error
= vfs_statfs_native(nd
.dentry
, &tmp
);
134 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
142 asmlinkage
long sys_statfs64(const char __user
*path
, size_t sz
, struct statfs64 __user
*buf
)
147 if (sz
!= sizeof(*buf
))
149 error
= user_path_walk(path
, &nd
);
152 error
= vfs_statfs64(nd
.dentry
, &tmp
);
153 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
161 asmlinkage
long sys_fstatfs(unsigned int fd
, struct statfs __user
* buf
)
171 error
= vfs_statfs_native(file
->f_dentry
, &tmp
);
172 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
179 asmlinkage
long sys_fstatfs64(unsigned int fd
, size_t sz
, struct statfs64 __user
*buf
)
185 if (sz
!= sizeof(*buf
))
192 error
= vfs_statfs64(file
->f_dentry
, &tmp
);
193 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
200 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
204 struct iattr newattrs
;
206 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
210 newattrs
.ia_size
= length
;
211 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
213 newattrs
.ia_file
= filp
;
214 newattrs
.ia_valid
|= ATTR_FILE
;
217 mutex_lock(&dentry
->d_inode
->i_mutex
);
218 err
= notify_change(dentry
, &newattrs
);
219 mutex_unlock(&dentry
->d_inode
->i_mutex
);
223 static long do_sys_truncate(const char __user
* path
, loff_t length
)
226 struct inode
* inode
;
230 if (length
< 0) /* sorry, but loff_t says... */
233 error
= user_path_walk(path
, &nd
);
236 inode
= nd
.dentry
->d_inode
;
238 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
240 if (S_ISDIR(inode
->i_mode
))
244 if (!S_ISREG(inode
->i_mode
))
247 error
= vfs_permission(&nd
, MAY_WRITE
);
252 if (IS_RDONLY(inode
))
256 if (IS_IMMUTABLE(inode
) || IS_APPEND(inode
))
260 * Make sure that there are no leases.
262 error
= break_lease(inode
, FMODE_WRITE
);
266 error
= get_write_access(inode
);
270 error
= locks_verify_truncate(inode
, NULL
, length
);
273 error
= do_truncate(nd
.dentry
, length
, 0, NULL
);
275 put_write_access(inode
);
283 asmlinkage
long sys_truncate(const char __user
* path
, unsigned long length
)
285 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
286 return do_sys_truncate(path
, (long)length
);
289 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
291 struct inode
* inode
;
292 struct dentry
*dentry
;
304 /* explicitly opened as large or we are on 64-bit box */
305 if (file
->f_flags
& O_LARGEFILE
)
308 dentry
= file
->f_dentry
;
309 inode
= dentry
->d_inode
;
311 if (!S_ISREG(inode
->i_mode
) || !(file
->f_mode
& FMODE_WRITE
))
315 /* Cannot ftruncate over 2^31 bytes without large file support */
316 if (small
&& length
> MAX_NON_LFS
)
320 if (IS_APPEND(inode
))
323 error
= locks_verify_truncate(inode
, file
, length
);
325 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, file
);
332 asmlinkage
long sys_ftruncate(unsigned int fd
, unsigned long length
)
334 long ret
= do_sys_ftruncate(fd
, length
, 1);
335 /* avoid REGPARM breakage on x86: */
336 prevent_tail_call(ret
);
340 /* LFS versions of truncate are only needed on 32 bit machines */
341 #if BITS_PER_LONG == 32
342 asmlinkage
long sys_truncate64(const char __user
* path
, loff_t length
)
344 return do_sys_truncate(path
, length
);
347 asmlinkage
long sys_ftruncate64(unsigned int fd
, loff_t length
)
349 long ret
= do_sys_ftruncate(fd
, length
, 0);
350 /* avoid REGPARM breakage on x86: */
351 prevent_tail_call(ret
);
356 #ifdef __ARCH_WANT_SYS_UTIME
359 * sys_utime() can be implemented in user-level using sys_utimes().
360 * Is this for backwards compatibility? If so, why not move it
361 * into the appropriate arch directory (for those architectures that
365 /* If times==NULL, set access and modification to current time,
366 * must be owner or have write permission.
367 * Else, update from *times, must be owner or super user.
369 asmlinkage
long sys_utime(char __user
* filename
, struct utimbuf __user
* times
)
373 struct inode
* inode
;
374 struct iattr newattrs
;
376 error
= user_path_walk(filename
, &nd
);
379 inode
= nd
.dentry
->d_inode
;
382 if (IS_RDONLY(inode
))
385 /* Don't worry, the checks are done in inode_change_ok() */
386 newattrs
.ia_valid
= ATTR_CTIME
| ATTR_MTIME
| ATTR_ATIME
;
389 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
392 error
= get_user(newattrs
.ia_atime
.tv_sec
, ×
->actime
);
393 newattrs
.ia_atime
.tv_nsec
= 0;
395 error
= get_user(newattrs
.ia_mtime
.tv_sec
, ×
->modtime
);
396 newattrs
.ia_mtime
.tv_nsec
= 0;
400 newattrs
.ia_valid
|= ATTR_ATIME_SET
| ATTR_MTIME_SET
;
403 if (IS_IMMUTABLE(inode
))
406 if (current
->fsuid
!= inode
->i_uid
&&
407 (error
= vfs_permission(&nd
, MAY_WRITE
)) != 0)
410 mutex_lock(&inode
->i_mutex
);
411 error
= notify_change(nd
.dentry
, &newattrs
);
412 mutex_unlock(&inode
->i_mutex
);
421 /* If times==NULL, set access and modification to current time,
422 * must be owner or have write permission.
423 * Else, update from *times, must be owner or super user.
425 long do_utimes(int dfd
, char __user
*filename
, struct timeval
*times
)
429 struct inode
* inode
;
430 struct iattr newattrs
;
432 error
= __user_walk_fd(dfd
, filename
, LOOKUP_FOLLOW
, &nd
);
436 inode
= nd
.dentry
->d_inode
;
439 if (IS_RDONLY(inode
))
442 /* Don't worry, the checks are done in inode_change_ok() */
443 newattrs
.ia_valid
= ATTR_CTIME
| ATTR_MTIME
| ATTR_ATIME
;
446 if (IS_APPEND(inode
) || IS_IMMUTABLE(inode
))
449 newattrs
.ia_atime
.tv_sec
= times
[0].tv_sec
;
450 newattrs
.ia_atime
.tv_nsec
= times
[0].tv_usec
* 1000;
451 newattrs
.ia_mtime
.tv_sec
= times
[1].tv_sec
;
452 newattrs
.ia_mtime
.tv_nsec
= times
[1].tv_usec
* 1000;
453 newattrs
.ia_valid
|= ATTR_ATIME_SET
| ATTR_MTIME_SET
;
456 if (IS_IMMUTABLE(inode
))
459 if (current
->fsuid
!= inode
->i_uid
&&
460 (error
= vfs_permission(&nd
, MAY_WRITE
)) != 0)
463 mutex_lock(&inode
->i_mutex
);
464 error
= notify_change(nd
.dentry
, &newattrs
);
465 mutex_unlock(&inode
->i_mutex
);
472 asmlinkage
long sys_futimesat(int dfd
, char __user
*filename
, struct timeval __user
*utimes
)
474 struct timeval times
[2];
476 if (utimes
&& copy_from_user(×
, utimes
, sizeof(times
)))
478 return do_utimes(dfd
, filename
, utimes
? times
: NULL
);
481 asmlinkage
long sys_utimes(char __user
*filename
, struct timeval __user
*utimes
)
483 return sys_futimesat(AT_FDCWD
, filename
, utimes
);
488 * access() needs to use the real uid/gid, not the effective uid/gid.
489 * We do this by temporarily clearing all FS-related capabilities and
490 * switching the fsuid/fsgid around to the real ones.
492 asmlinkage
long sys_faccessat(int dfd
, const char __user
*filename
, int mode
)
495 int old_fsuid
, old_fsgid
;
496 kernel_cap_t old_cap
;
499 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
502 old_fsuid
= current
->fsuid
;
503 old_fsgid
= current
->fsgid
;
504 old_cap
= current
->cap_effective
;
506 current
->fsuid
= current
->uid
;
507 current
->fsgid
= current
->gid
;
510 * Clear the capabilities if we switch to a non-root user
512 * FIXME: There is a race here against sys_capset. The
513 * capabilities can change yet we will restore the old
514 * value below. We should hold task_capabilities_lock,
515 * but we cannot because user_path_walk can sleep.
518 cap_clear(current
->cap_effective
);
520 current
->cap_effective
= current
->cap_permitted
;
522 res
= __user_walk_fd(dfd
, filename
, LOOKUP_FOLLOW
|LOOKUP_ACCESS
, &nd
);
524 res
= vfs_permission(&nd
, mode
);
525 /* SuS v2 requires we report a read only fs too */
526 if(!res
&& (mode
& S_IWOTH
) && IS_RDONLY(nd
.dentry
->d_inode
)
527 && !special_file(nd
.dentry
->d_inode
->i_mode
))
532 current
->fsuid
= old_fsuid
;
533 current
->fsgid
= old_fsgid
;
534 current
->cap_effective
= old_cap
;
539 asmlinkage
long sys_access(const char __user
*filename
, int mode
)
541 return sys_faccessat(AT_FDCWD
, filename
, mode
);
544 asmlinkage
long sys_chdir(const char __user
* filename
)
549 error
= __user_walk(filename
,
550 LOOKUP_FOLLOW
|LOOKUP_DIRECTORY
|LOOKUP_CHDIR
, &nd
);
554 error
= vfs_permission(&nd
, MAY_EXEC
);
558 set_fs_pwd(current
->fs
, nd
.mnt
, nd
.dentry
);
566 asmlinkage
long sys_fchdir(unsigned int fd
)
569 struct dentry
*dentry
;
571 struct vfsmount
*mnt
;
579 dentry
= file
->f_dentry
;
580 mnt
= file
->f_vfsmnt
;
581 inode
= dentry
->d_inode
;
584 if (!S_ISDIR(inode
->i_mode
))
587 error
= file_permission(file
, MAY_EXEC
);
589 set_fs_pwd(current
->fs
, mnt
, dentry
);
596 asmlinkage
long sys_chroot(const char __user
* filename
)
601 error
= __user_walk(filename
, LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
| LOOKUP_NOALT
, &nd
);
605 error
= vfs_permission(&nd
, MAY_EXEC
);
610 if (!capable(CAP_SYS_CHROOT
))
613 set_fs_root(current
->fs
, nd
.mnt
, nd
.dentry
);
622 asmlinkage
long sys_fchmod(unsigned int fd
, mode_t mode
)
624 struct inode
* inode
;
625 struct dentry
* dentry
;
628 struct iattr newattrs
;
634 dentry
= file
->f_dentry
;
635 inode
= dentry
->d_inode
;
637 audit_inode(NULL
, inode
);
640 if (IS_RDONLY(inode
))
643 if (IS_IMMUTABLE(inode
) || IS_APPEND(inode
))
645 mutex_lock(&inode
->i_mutex
);
646 if (mode
== (mode_t
) -1)
647 mode
= inode
->i_mode
;
648 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
649 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
650 err
= notify_change(dentry
, &newattrs
);
651 mutex_unlock(&inode
->i_mutex
);
659 asmlinkage
long sys_fchmodat(int dfd
, const char __user
*filename
,
663 struct inode
* inode
;
665 struct iattr newattrs
;
667 error
= __user_walk_fd(dfd
, filename
, LOOKUP_FOLLOW
, &nd
);
670 inode
= nd
.dentry
->d_inode
;
673 if (IS_RDONLY(inode
))
677 if (IS_IMMUTABLE(inode
) || IS_APPEND(inode
))
680 mutex_lock(&inode
->i_mutex
);
681 if (mode
== (mode_t
) -1)
682 mode
= inode
->i_mode
;
683 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
684 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
685 error
= notify_change(nd
.dentry
, &newattrs
);
686 mutex_unlock(&inode
->i_mutex
);
694 asmlinkage
long sys_chmod(const char __user
*filename
, mode_t mode
)
696 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
699 static int chown_common(struct dentry
* dentry
, uid_t user
, gid_t group
)
701 struct inode
* inode
;
703 struct iattr newattrs
;
706 if (!(inode
= dentry
->d_inode
)) {
707 printk(KERN_ERR
"chown_common: NULL inode\n");
711 if (IS_RDONLY(inode
))
714 if (IS_IMMUTABLE(inode
) || IS_APPEND(inode
))
716 newattrs
.ia_valid
= ATTR_CTIME
;
717 if (user
!= (uid_t
) -1) {
718 newattrs
.ia_valid
|= ATTR_UID
;
719 newattrs
.ia_uid
= user
;
721 if (group
!= (gid_t
) -1) {
722 newattrs
.ia_valid
|= ATTR_GID
;
723 newattrs
.ia_gid
= group
;
725 if (!S_ISDIR(inode
->i_mode
))
726 newattrs
.ia_valid
|= ATTR_KILL_SUID
|ATTR_KILL_SGID
;
727 mutex_lock(&inode
->i_mutex
);
728 error
= notify_change(dentry
, &newattrs
);
729 mutex_unlock(&inode
->i_mutex
);
734 asmlinkage
long sys_chown(const char __user
* filename
, uid_t user
, gid_t group
)
739 error
= user_path_walk(filename
, &nd
);
741 error
= chown_common(nd
.dentry
, user
, group
);
747 asmlinkage
long sys_fchownat(int dfd
, const char __user
*filename
, uid_t user
,
748 gid_t group
, int flag
)
754 if ((flag
& ~AT_SYMLINK_NOFOLLOW
) != 0)
757 follow
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
758 error
= __user_walk_fd(dfd
, filename
, follow
, &nd
);
760 error
= chown_common(nd
.dentry
, user
, group
);
767 asmlinkage
long sys_lchown(const char __user
* filename
, uid_t user
, gid_t group
)
772 error
= user_path_walk_link(filename
, &nd
);
774 error
= chown_common(nd
.dentry
, user
, group
);
781 asmlinkage
long sys_fchown(unsigned int fd
, uid_t user
, gid_t group
)
788 struct dentry
* dentry
;
789 dentry
= file
->f_dentry
;
790 audit_inode(NULL
, dentry
->d_inode
);
791 error
= chown_common(dentry
, user
, group
);
797 static struct file
*__dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
,
798 int flags
, struct file
*f
,
799 int (*open
)(struct inode
*, struct file
*))
805 f
->f_mode
= ((flags
+1) & O_ACCMODE
) | FMODE_LSEEK
|
806 FMODE_PREAD
| FMODE_PWRITE
;
807 inode
= dentry
->d_inode
;
808 if (f
->f_mode
& FMODE_WRITE
) {
809 error
= get_write_access(inode
);
814 f
->f_mapping
= inode
->i_mapping
;
815 f
->f_dentry
= dentry
;
818 f
->f_op
= fops_get(inode
->i_fop
);
819 file_move(f
, &inode
->i_sb
->s_files
);
821 if (!open
&& f
->f_op
)
822 open
= f
->f_op
->open
;
824 error
= open(inode
, f
);
829 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
831 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
833 /* NB: we're sure to have correct a_ops only after f_op->open */
834 if (f
->f_flags
& O_DIRECT
) {
835 if (!f
->f_mapping
->a_ops
||
836 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
837 (!f
->f_mapping
->a_ops
->get_xip_page
))) {
839 f
= ERR_PTR(-EINVAL
);
847 if (f
->f_mode
& FMODE_WRITE
)
848 put_write_access(inode
);
856 return ERR_PTR(error
);
860 * Note that while the flag value (low two bits) for sys_open means:
866 * 00 - no permissions needed
867 * 01 - read-permission
868 * 10 - write-permission
870 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
873 static struct file
*do_filp_open(int dfd
, const char *filename
, int flags
,
876 int namei_flags
, error
;
880 if ((namei_flags
+1) & O_ACCMODE
)
883 error
= open_namei(dfd
, filename
, namei_flags
, mode
, &nd
);
885 return nameidata_to_filp(&nd
, flags
);
887 return ERR_PTR(error
);
890 struct file
*filp_open(const char *filename
, int flags
, int mode
)
892 return do_filp_open(AT_FDCWD
, filename
, flags
, mode
);
894 EXPORT_SYMBOL(filp_open
);
897 * lookup_instantiate_filp - instantiates the open intent filp
898 * @nd: pointer to nameidata
899 * @dentry: pointer to dentry
900 * @open: open callback
902 * Helper for filesystems that want to use lookup open intents and pass back
903 * a fully instantiated struct file to the caller.
904 * This function is meant to be called from within a filesystem's
906 * Beware of calling it for non-regular files! Those ->open methods might block
907 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
908 * leading to a deadlock, as nobody can open that fifo anymore, because
909 * another process to open fifo will block on locked parent when doing lookup).
910 * Note that in case of error, nd->intent.open.file is destroyed, but the
911 * path information remains valid.
912 * If the open callback is set to NULL, then the standard f_op->open()
913 * filesystem callback is substituted.
915 struct file
*lookup_instantiate_filp(struct nameidata
*nd
, struct dentry
*dentry
,
916 int (*open
)(struct inode
*, struct file
*))
918 if (IS_ERR(nd
->intent
.open
.file
))
922 nd
->intent
.open
.file
= __dentry_open(dget(dentry
), mntget(nd
->mnt
),
923 nd
->intent
.open
.flags
- 1,
924 nd
->intent
.open
.file
,
927 return nd
->intent
.open
.file
;
929 release_open_intent(nd
);
930 nd
->intent
.open
.file
= (struct file
*)dentry
;
933 EXPORT_SYMBOL_GPL(lookup_instantiate_filp
);
936 * nameidata_to_filp - convert a nameidata to an open filp.
937 * @nd: pointer to nameidata
940 * Note that this function destroys the original nameidata
942 struct file
*nameidata_to_filp(struct nameidata
*nd
, int flags
)
946 /* Pick up the filp from the open intent */
947 filp
= nd
->intent
.open
.file
;
948 /* Has the filesystem initialised the file for us? */
949 if (filp
->f_dentry
== NULL
)
950 filp
= __dentry_open(nd
->dentry
, nd
->mnt
, flags
, filp
, NULL
);
957 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
960 struct file
*dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
, int flags
)
966 f
= get_empty_filp();
970 return ERR_PTR(error
);
973 return __dentry_open(dentry
, mnt
, flags
, f
, NULL
);
975 EXPORT_SYMBOL(dentry_open
);
978 * Find an empty file descriptor entry, and mark it busy.
980 int get_unused_fd(void)
982 struct files_struct
* files
= current
->files
;
987 spin_lock(&files
->file_lock
);
990 fdt
= files_fdtable(files
);
991 fd
= find_next_zero_bit(fdt
->open_fds
->fds_bits
,
996 * N.B. For clone tasks sharing a files structure, this test
997 * will limit the total number of files that can be opened.
999 if (fd
>= current
->signal
->rlim
[RLIMIT_NOFILE
].rlim_cur
)
1002 /* Do we need to expand the fd array or fd set? */
1003 error
= expand_files(files
, fd
);
1009 * If we needed to expand the fs array we
1010 * might have blocked - try again.
1016 FD_SET(fd
, fdt
->open_fds
);
1017 FD_CLR(fd
, fdt
->close_on_exec
);
1018 files
->next_fd
= fd
+ 1;
1021 if (fdt
->fd
[fd
] != NULL
) {
1022 printk(KERN_WARNING
"get_unused_fd: slot %d not NULL!\n", fd
);
1029 spin_unlock(&files
->file_lock
);
1033 EXPORT_SYMBOL(get_unused_fd
);
1035 static void __put_unused_fd(struct files_struct
*files
, unsigned int fd
)
1037 struct fdtable
*fdt
= files_fdtable(files
);
1038 __FD_CLR(fd
, fdt
->open_fds
);
1039 if (fd
< files
->next_fd
)
1040 files
->next_fd
= fd
;
1043 void fastcall
put_unused_fd(unsigned int fd
)
1045 struct files_struct
*files
= current
->files
;
1046 spin_lock(&files
->file_lock
);
1047 __put_unused_fd(files
, fd
);
1048 spin_unlock(&files
->file_lock
);
1051 EXPORT_SYMBOL(put_unused_fd
);
1054 * Install a file pointer in the fd array.
1056 * The VFS is full of places where we drop the files lock between
1057 * setting the open_fds bitmap and installing the file in the file
1058 * array. At any such point, we are vulnerable to a dup2() race
1059 * installing a file in the array before us. We need to detect this and
1060 * fput() the struct file we are about to overwrite in this case.
1062 * It should never happen - if we allow dup2() do it, _really_ bad things
1066 void fastcall
fd_install(unsigned int fd
, struct file
* file
)
1068 struct files_struct
*files
= current
->files
;
1069 struct fdtable
*fdt
;
1070 spin_lock(&files
->file_lock
);
1071 fdt
= files_fdtable(files
);
1072 BUG_ON(fdt
->fd
[fd
] != NULL
);
1073 rcu_assign_pointer(fdt
->fd
[fd
], file
);
1074 spin_unlock(&files
->file_lock
);
1077 EXPORT_SYMBOL(fd_install
);
1079 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, int mode
)
1081 char *tmp
= getname(filename
);
1082 int fd
= PTR_ERR(tmp
);
1085 fd
= get_unused_fd();
1087 struct file
*f
= do_filp_open(dfd
, tmp
, flags
, mode
);
1092 fsnotify_open(f
->f_dentry
);
1101 asmlinkage
long sys_open(const char __user
*filename
, int flags
, int mode
)
1105 if (force_o_largefile())
1106 flags
|= O_LARGEFILE
;
1108 ret
= do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1109 /* avoid REGPARM breakage on x86: */
1110 prevent_tail_call(ret
);
1113 EXPORT_SYMBOL_GPL(sys_open
);
1115 asmlinkage
long sys_openat(int dfd
, const char __user
*filename
, int flags
,
1120 if (force_o_largefile())
1121 flags
|= O_LARGEFILE
;
1123 ret
= do_sys_open(dfd
, filename
, flags
, mode
);
1124 /* avoid REGPARM breakage on x86: */
1125 prevent_tail_call(ret
);
1132 * For backward compatibility? Maybe this should be moved
1133 * into arch/i386 instead?
1135 asmlinkage
long sys_creat(const char __user
* pathname
, int mode
)
1137 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1143 * "id" is the POSIX thread ID. We use the
1144 * files pointer for this..
1146 int filp_close(struct file
*filp
, fl_owner_t id
)
1150 if (!file_count(filp
)) {
1151 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1155 if (filp
->f_op
&& filp
->f_op
->flush
)
1156 retval
= filp
->f_op
->flush(filp
, id
);
1158 dnotify_flush(filp
, id
);
1159 locks_remove_posix(filp
, id
);
1164 EXPORT_SYMBOL(filp_close
);
1167 * Careful here! We test whether the file pointer is NULL before
1168 * releasing the fd. This ensures that one clone task can't release
1169 * an fd while another clone is opening it.
1171 asmlinkage
long sys_close(unsigned int fd
)
1174 struct files_struct
*files
= current
->files
;
1175 struct fdtable
*fdt
;
1177 spin_lock(&files
->file_lock
);
1178 fdt
= files_fdtable(files
);
1179 if (fd
>= fdt
->max_fds
)
1184 rcu_assign_pointer(fdt
->fd
[fd
], NULL
);
1185 FD_CLR(fd
, fdt
->close_on_exec
);
1186 __put_unused_fd(files
, fd
);
1187 spin_unlock(&files
->file_lock
);
1188 return filp_close(filp
, files
);
1191 spin_unlock(&files
->file_lock
);
1195 EXPORT_SYMBOL(sys_close
);
1198 * This routine simulates a hangup on the tty, to arrange that users
1199 * are given clean terminals at login time.
1201 asmlinkage
long sys_vhangup(void)
1203 if (capable(CAP_SYS_TTY_CONFIG
)) {
1204 tty_vhangup(current
->signal
->tty
);
1211 * Called when an inode is about to be open.
1212 * We use this to disallow opening large files on 32bit systems if
1213 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1214 * on this flag in sys_open.
1216 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1218 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1223 EXPORT_SYMBOL(generic_file_open
);
1226 * This is used by subsystems that don't want seekable
1229 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1231 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1235 EXPORT_SYMBOL(nonseekable_open
);