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>
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 st
.f_bsize
| st
.f_frsize
) &
70 0xffffffff00000000ULL
)
73 * f_files and f_ffree may be -1; it's okay to stuff
76 if (st
.f_files
!= -1 &&
77 (st
.f_files
& 0xffffffff00000000ULL
))
79 if (st
.f_ffree
!= -1 &&
80 (st
.f_ffree
& 0xffffffff00000000ULL
))
84 buf
->f_type
= st
.f_type
;
85 buf
->f_bsize
= st
.f_bsize
;
86 buf
->f_blocks
= st
.f_blocks
;
87 buf
->f_bfree
= st
.f_bfree
;
88 buf
->f_bavail
= st
.f_bavail
;
89 buf
->f_files
= st
.f_files
;
90 buf
->f_ffree
= st
.f_ffree
;
91 buf
->f_fsid
= st
.f_fsid
;
92 buf
->f_namelen
= st
.f_namelen
;
93 buf
->f_frsize
= st
.f_frsize
;
94 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
99 static int vfs_statfs64(struct dentry
*dentry
, struct statfs64
*buf
)
104 retval
= vfs_statfs(dentry
, &st
);
108 if (sizeof(*buf
) == sizeof(st
))
109 memcpy(buf
, &st
, sizeof(st
));
111 buf
->f_type
= st
.f_type
;
112 buf
->f_bsize
= st
.f_bsize
;
113 buf
->f_blocks
= st
.f_blocks
;
114 buf
->f_bfree
= st
.f_bfree
;
115 buf
->f_bavail
= st
.f_bavail
;
116 buf
->f_files
= st
.f_files
;
117 buf
->f_ffree
= st
.f_ffree
;
118 buf
->f_fsid
= st
.f_fsid
;
119 buf
->f_namelen
= st
.f_namelen
;
120 buf
->f_frsize
= st
.f_frsize
;
121 memset(buf
->f_spare
, 0, sizeof(buf
->f_spare
));
126 SYSCALL_DEFINE2(statfs
, const char __user
*, pathname
, struct statfs __user
*, buf
)
131 error
= user_path(pathname
, &path
);
134 error
= vfs_statfs_native(path
.dentry
, &tmp
);
135 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
142 SYSCALL_DEFINE3(statfs64
, const char __user
*, pathname
, size_t, sz
, struct statfs64 __user
*, buf
)
147 if (sz
!= sizeof(*buf
))
149 error
= user_path(pathname
, &path
);
152 error
= vfs_statfs64(path
.dentry
, &tmp
);
153 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
160 SYSCALL_DEFINE2(fstatfs
, unsigned int, fd
, struct statfs __user
*, buf
)
170 error
= vfs_statfs_native(file
->f_path
.dentry
, &tmp
);
171 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
178 SYSCALL_DEFINE3(fstatfs64
, unsigned int, fd
, size_t, sz
, struct statfs64 __user
*, buf
)
184 if (sz
!= sizeof(*buf
))
191 error
= vfs_statfs64(file
->f_path
.dentry
, &tmp
);
192 if (!error
&& copy_to_user(buf
, &tmp
, sizeof(tmp
)))
199 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
203 struct iattr newattrs
;
205 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
209 newattrs
.ia_size
= length
;
210 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
212 newattrs
.ia_file
= filp
;
213 newattrs
.ia_valid
|= ATTR_FILE
;
216 /* Remove suid/sgid on truncate too */
217 ret
= should_remove_suid(dentry
);
219 newattrs
.ia_valid
|= ret
| ATTR_FORCE
;
221 mutex_lock(&dentry
->d_inode
->i_mutex
);
222 ret
= notify_change(dentry
, &newattrs
);
223 mutex_unlock(&dentry
->d_inode
->i_mutex
);
227 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
234 if (length
< 0) /* sorry, but loff_t says... */
237 error
= user_path(pathname
, &path
);
240 inode
= path
.dentry
->d_inode
;
242 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
244 if (S_ISDIR(inode
->i_mode
))
248 if (!S_ISREG(inode
->i_mode
))
251 error
= mnt_want_write(path
.mnt
);
255 error
= inode_permission(inode
, MAY_WRITE
);
257 goto mnt_drop_write_and_out
;
260 if (IS_APPEND(inode
))
261 goto mnt_drop_write_and_out
;
263 error
= get_write_access(inode
);
265 goto mnt_drop_write_and_out
;
268 * Make sure that there are no leases. get_write_access() protects
269 * against the truncate racing with a lease-granting setlease().
271 error
= break_lease(inode
, FMODE_WRITE
);
273 goto put_write_and_out
;
275 error
= locks_verify_truncate(inode
, NULL
, length
);
277 error
= security_path_truncate(&path
, length
, 0);
280 error
= do_truncate(path
.dentry
, length
, 0, NULL
);
284 put_write_access(inode
);
285 mnt_drop_write_and_out
:
286 mnt_drop_write(path
.mnt
);
293 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
295 return do_sys_truncate(path
, length
);
298 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
300 struct inode
* inode
;
301 struct dentry
*dentry
;
313 /* explicitly opened as large or we are on 64-bit box */
314 if (file
->f_flags
& O_LARGEFILE
)
317 dentry
= file
->f_path
.dentry
;
318 inode
= dentry
->d_inode
;
320 if (!S_ISREG(inode
->i_mode
) || !(file
->f_mode
& FMODE_WRITE
))
324 /* Cannot ftruncate over 2^31 bytes without large file support */
325 if (small
&& length
> MAX_NON_LFS
)
329 if (IS_APPEND(inode
))
332 error
= locks_verify_truncate(inode
, file
, length
);
334 error
= security_path_truncate(&file
->f_path
, length
,
335 ATTR_MTIME
|ATTR_CTIME
);
337 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, file
);
344 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
346 long ret
= do_sys_ftruncate(fd
, length
, 1);
347 /* avoid REGPARM breakage on x86: */
348 asmlinkage_protect(2, ret
, fd
, length
);
352 /* LFS versions of truncate are only needed on 32 bit machines */
353 #if BITS_PER_LONG == 32
354 SYSCALL_DEFINE(truncate64
)(const char __user
* path
, loff_t length
)
356 return do_sys_truncate(path
, length
);
358 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
359 asmlinkage
long SyS_truncate64(long path
, loff_t length
)
361 return SYSC_truncate64((const char __user
*) path
, length
);
363 SYSCALL_ALIAS(sys_truncate64
, SyS_truncate64
);
366 SYSCALL_DEFINE(ftruncate64
)(unsigned int fd
, loff_t length
)
368 long ret
= do_sys_ftruncate(fd
, length
, 0);
369 /* avoid REGPARM breakage on x86: */
370 asmlinkage_protect(2, ret
, fd
, length
);
373 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
374 asmlinkage
long SyS_ftruncate64(long fd
, loff_t length
)
376 return SYSC_ftruncate64((unsigned int) fd
, length
);
378 SYSCALL_ALIAS(sys_ftruncate64
, SyS_ftruncate64
);
380 #endif /* BITS_PER_LONG == 32 */
383 int do_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
385 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
388 if (offset
< 0 || len
<= 0)
391 /* Return error if mode is not supported */
392 if (mode
&& !(mode
& FALLOC_FL_KEEP_SIZE
))
395 if (!(file
->f_mode
& FMODE_WRITE
))
398 * Revalidate the write permissions, in case security policy has
399 * changed since the files were opened.
401 ret
= security_file_permission(file
, MAY_WRITE
);
405 if (S_ISFIFO(inode
->i_mode
))
409 * Let individual file system decide if it supports preallocation
410 * for directories or not.
412 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
))
415 /* Check for wrap through zero too */
416 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
419 if (!inode
->i_op
->fallocate
)
422 return inode
->i_op
->fallocate(inode
, mode
, offset
, len
);
425 SYSCALL_DEFINE(fallocate
)(int fd
, int mode
, loff_t offset
, loff_t len
)
432 error
= do_fallocate(file
, mode
, offset
, len
);
439 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
440 asmlinkage
long SyS_fallocate(long fd
, long mode
, loff_t offset
, loff_t len
)
442 return SYSC_fallocate((int)fd
, (int)mode
, offset
, len
);
444 SYSCALL_ALIAS(sys_fallocate
, SyS_fallocate
);
448 * access() needs to use the real uid/gid, not the effective uid/gid.
449 * We do this by temporarily clearing all FS-related capabilities and
450 * switching the fsuid/fsgid around to the real ones.
452 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
454 const struct cred
*old_cred
;
455 struct cred
*override_cred
;
460 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
463 override_cred
= prepare_creds();
467 override_cred
->fsuid
= override_cred
->uid
;
468 override_cred
->fsgid
= override_cred
->gid
;
470 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
471 /* Clear the capabilities if we switch to a non-root user */
472 if (override_cred
->uid
)
473 cap_clear(override_cred
->cap_effective
);
475 override_cred
->cap_effective
=
476 override_cred
->cap_permitted
;
479 old_cred
= override_creds(override_cred
);
481 res
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
485 inode
= path
.dentry
->d_inode
;
487 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
489 * MAY_EXEC on regular files is denied if the fs is mounted
490 * with the "noexec" flag.
493 if (path
.mnt
->mnt_flags
& MNT_NOEXEC
)
494 goto out_path_release
;
497 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
498 /* SuS v2 requires we report a read only fs too */
499 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
500 goto out_path_release
;
502 * This is a rare case where using __mnt_is_readonly()
503 * is OK without a mnt_want/drop_write() pair. Since
504 * no actual write to the fs is performed here, we do
505 * not need to telegraph to that to anyone.
507 * By doing this, we accept that this access is
508 * inherently racy and know that the fs may change
509 * state before we even see this result.
511 if (__mnt_is_readonly(path
.mnt
))
517 revert_creds(old_cred
);
518 put_cred(override_cred
);
522 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
524 return sys_faccessat(AT_FDCWD
, filename
, mode
);
527 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
532 error
= user_path_dir(filename
, &path
);
536 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
540 set_fs_pwd(current
->fs
, &path
);
548 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
559 inode
= file
->f_path
.dentry
->d_inode
;
562 if (!S_ISDIR(inode
->i_mode
))
565 error
= inode_permission(inode
, MAY_EXEC
| MAY_ACCESS
);
567 set_fs_pwd(current
->fs
, &file
->f_path
);
574 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
579 error
= user_path_dir(filename
, &path
);
583 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
588 if (!capable(CAP_SYS_CHROOT
))
590 error
= security_path_chroot(&path
);
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(file
);
622 mutex_lock(&inode
->i_mutex
);
623 err
= security_path_chmod(dentry
, file
->f_vfsmnt
, mode
);
626 if (mode
== (mode_t
) -1)
627 mode
= inode
->i_mode
;
628 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
629 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
630 err
= notify_change(dentry
, &newattrs
);
632 mutex_unlock(&inode
->i_mutex
);
633 mnt_drop_write(file
->f_path
.mnt
);
640 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, mode_t
, mode
)
645 struct iattr newattrs
;
647 error
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
650 inode
= path
.dentry
->d_inode
;
652 error
= mnt_want_write(path
.mnt
);
655 mutex_lock(&inode
->i_mutex
);
656 error
= security_path_chmod(path
.dentry
, path
.mnt
, mode
);
659 if (mode
== (mode_t
) -1)
660 mode
= inode
->i_mode
;
661 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
662 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
663 error
= notify_change(path
.dentry
, &newattrs
);
665 mutex_unlock(&inode
->i_mutex
);
666 mnt_drop_write(path
.mnt
);
673 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, mode_t
, mode
)
675 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
678 static int chown_common(struct path
*path
, uid_t user
, gid_t group
)
680 struct inode
*inode
= path
->dentry
->d_inode
;
682 struct iattr newattrs
;
684 newattrs
.ia_valid
= ATTR_CTIME
;
685 if (user
!= (uid_t
) -1) {
686 newattrs
.ia_valid
|= ATTR_UID
;
687 newattrs
.ia_uid
= user
;
689 if (group
!= (gid_t
) -1) {
690 newattrs
.ia_valid
|= ATTR_GID
;
691 newattrs
.ia_gid
= group
;
693 if (!S_ISDIR(inode
->i_mode
))
695 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
696 mutex_lock(&inode
->i_mutex
);
697 error
= security_path_chown(path
, user
, group
);
699 error
= notify_change(path
->dentry
, &newattrs
);
700 mutex_unlock(&inode
->i_mutex
);
705 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
710 error
= user_path(filename
, &path
);
713 error
= mnt_want_write(path
.mnt
);
716 error
= chown_common(&path
, user
, group
);
717 mnt_drop_write(path
.mnt
);
724 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
725 gid_t
, group
, int, flag
)
731 if ((flag
& ~AT_SYMLINK_NOFOLLOW
) != 0)
734 follow
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
735 error
= user_path_at(dfd
, filename
, follow
, &path
);
738 error
= mnt_want_write(path
.mnt
);
741 error
= chown_common(&path
, user
, group
);
742 mnt_drop_write(path
.mnt
);
749 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
754 error
= user_lpath(filename
, &path
);
757 error
= mnt_want_write(path
.mnt
);
760 error
= chown_common(&path
, user
, group
);
761 mnt_drop_write(path
.mnt
);
768 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
772 struct dentry
* dentry
;
778 error
= mnt_want_write_file(file
);
781 dentry
= file
->f_path
.dentry
;
782 audit_inode(NULL
, dentry
);
783 error
= chown_common(&file
->f_path
, user
, group
);
784 mnt_drop_write(file
->f_path
.mnt
);
792 * You have to be very careful that these write
793 * counts get cleaned up in error cases and
794 * upon __fput(). This should probably never
795 * be called outside of __dentry_open().
797 static inline int __get_file_write_access(struct inode
*inode
,
798 struct vfsmount
*mnt
)
801 error
= get_write_access(inode
);
805 * Do not take mount writer counts on
806 * special files since no writes to
807 * the mount itself will occur.
809 if (!special_file(inode
->i_mode
)) {
811 * Balanced in __fput()
813 error
= mnt_want_write(mnt
);
815 put_write_access(inode
);
820 static struct file
*__dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
,
821 int flags
, struct file
*f
,
822 int (*open
)(struct inode
*, struct file
*),
823 const struct cred
*cred
)
829 f
->f_mode
= (__force fmode_t
)((flags
+1) & O_ACCMODE
) | FMODE_LSEEK
|
830 FMODE_PREAD
| FMODE_PWRITE
;
831 inode
= dentry
->d_inode
;
832 if (f
->f_mode
& FMODE_WRITE
) {
833 error
= __get_file_write_access(inode
, mnt
);
836 if (!special_file(inode
->i_mode
))
840 f
->f_mapping
= inode
->i_mapping
;
841 f
->f_path
.dentry
= dentry
;
844 f
->f_op
= fops_get(inode
->i_fop
);
845 file_move(f
, &inode
->i_sb
->s_files
);
847 error
= security_dentry_open(f
, cred
);
851 if (!open
&& f
->f_op
)
852 open
= f
->f_op
->open
;
854 error
= open(inode
, f
);
859 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
861 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
863 /* NB: we're sure to have correct a_ops only after f_op->open */
864 if (f
->f_flags
& O_DIRECT
) {
865 if (!f
->f_mapping
->a_ops
||
866 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
867 (!f
->f_mapping
->a_ops
->get_xip_mem
))) {
869 f
= ERR_PTR(-EINVAL
);
877 if (f
->f_mode
& FMODE_WRITE
) {
878 put_write_access(inode
);
879 if (!special_file(inode
->i_mode
)) {
881 * We don't consider this a real
882 * mnt_want/drop_write() pair
883 * because it all happenend right
884 * here, so just reset the state.
891 f
->f_path
.dentry
= NULL
;
892 f
->f_path
.mnt
= NULL
;
897 return ERR_PTR(error
);
901 * lookup_instantiate_filp - instantiates the open intent filp
902 * @nd: pointer to nameidata
903 * @dentry: pointer to dentry
904 * @open: open callback
906 * Helper for filesystems that want to use lookup open intents and pass back
907 * a fully instantiated struct file to the caller.
908 * This function is meant to be called from within a filesystem's
910 * Beware of calling it for non-regular files! Those ->open methods might block
911 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
912 * leading to a deadlock, as nobody can open that fifo anymore, because
913 * another process to open fifo will block on locked parent when doing lookup).
914 * Note that in case of error, nd->intent.open.file is destroyed, but the
915 * path information remains valid.
916 * If the open callback is set to NULL, then the standard f_op->open()
917 * filesystem callback is substituted.
919 struct file
*lookup_instantiate_filp(struct nameidata
*nd
, struct dentry
*dentry
,
920 int (*open
)(struct inode
*, struct file
*))
922 const struct cred
*cred
= current_cred();
924 if (IS_ERR(nd
->intent
.open
.file
))
928 nd
->intent
.open
.file
= __dentry_open(dget(dentry
), mntget(nd
->path
.mnt
),
929 nd
->intent
.open
.flags
- 1,
930 nd
->intent
.open
.file
,
933 return nd
->intent
.open
.file
;
935 release_open_intent(nd
);
936 nd
->intent
.open
.file
= (struct file
*)dentry
;
939 EXPORT_SYMBOL_GPL(lookup_instantiate_filp
);
942 * nameidata_to_filp - convert a nameidata to an open filp.
943 * @nd: pointer to nameidata
946 * Note that this function destroys the original nameidata
948 struct file
*nameidata_to_filp(struct nameidata
*nd
, int flags
)
950 const struct cred
*cred
= current_cred();
953 /* Pick up the filp from the open intent */
954 filp
= nd
->intent
.open
.file
;
955 /* Has the filesystem initialised the file for us? */
956 if (filp
->f_path
.dentry
== NULL
)
957 filp
= __dentry_open(nd
->path
.dentry
, nd
->path
.mnt
, flags
, filp
,
965 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
968 struct file
*dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
, int flags
,
969 const struct cred
*cred
)
974 validate_creds(cred
);
977 * We must always pass in a valid mount pointer. Historically
978 * callers got away with not passing it, but we must enforce this at
979 * the earliest possible point now to avoid strange problems deep in the
983 printk(KERN_WARNING
"%s called with NULL vfsmount\n", __func__
);
985 return ERR_PTR(-EINVAL
);
989 f
= get_empty_filp();
993 return ERR_PTR(error
);
996 return __dentry_open(dentry
, mnt
, flags
, f
, NULL
, cred
);
998 EXPORT_SYMBOL(dentry_open
);
1000 static void __put_unused_fd(struct files_struct
*files
, unsigned int fd
)
1002 struct fdtable
*fdt
= files_fdtable(files
);
1003 __FD_CLR(fd
, fdt
->open_fds
);
1004 if (fd
< files
->next_fd
)
1005 files
->next_fd
= fd
;
1008 void put_unused_fd(unsigned int fd
)
1010 struct files_struct
*files
= current
->files
;
1011 spin_lock(&files
->file_lock
);
1012 __put_unused_fd(files
, fd
);
1013 spin_unlock(&files
->file_lock
);
1016 EXPORT_SYMBOL(put_unused_fd
);
1019 * Install a file pointer in the fd array.
1021 * The VFS is full of places where we drop the files lock between
1022 * setting the open_fds bitmap and installing the file in the file
1023 * array. At any such point, we are vulnerable to a dup2() race
1024 * installing a file in the array before us. We need to detect this and
1025 * fput() the struct file we are about to overwrite in this case.
1027 * It should never happen - if we allow dup2() do it, _really_ bad things
1031 void fd_install(unsigned int fd
, struct file
*file
)
1033 struct files_struct
*files
= current
->files
;
1034 struct fdtable
*fdt
;
1035 spin_lock(&files
->file_lock
);
1036 fdt
= files_fdtable(files
);
1037 BUG_ON(fdt
->fd
[fd
] != NULL
);
1038 rcu_assign_pointer(fdt
->fd
[fd
], file
);
1039 spin_unlock(&files
->file_lock
);
1042 EXPORT_SYMBOL(fd_install
);
1044 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, int mode
)
1046 char *tmp
= getname(filename
);
1047 int fd
= PTR_ERR(tmp
);
1050 fd
= get_unused_fd_flags(flags
);
1052 struct file
*f
= do_filp_open(dfd
, tmp
, flags
, mode
, 0);
1057 fsnotify_open(f
->f_path
.dentry
);
1066 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, int, mode
)
1070 if (force_o_largefile())
1071 flags
|= O_LARGEFILE
;
1073 ret
= do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1074 /* avoid REGPARM breakage on x86: */
1075 asmlinkage_protect(3, ret
, filename
, flags
, mode
);
1079 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1084 if (force_o_largefile())
1085 flags
|= O_LARGEFILE
;
1087 ret
= do_sys_open(dfd
, filename
, flags
, mode
);
1088 /* avoid REGPARM breakage on x86: */
1089 asmlinkage_protect(4, ret
, dfd
, filename
, flags
, mode
);
1096 * For backward compatibility? Maybe this should be moved
1097 * into arch/i386 instead?
1099 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, int, mode
)
1101 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1107 * "id" is the POSIX thread ID. We use the
1108 * files pointer for this..
1110 int filp_close(struct file
*filp
, fl_owner_t id
)
1114 if (!file_count(filp
)) {
1115 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1119 if (filp
->f_op
&& filp
->f_op
->flush
)
1120 retval
= filp
->f_op
->flush(filp
, id
);
1122 dnotify_flush(filp
, id
);
1123 locks_remove_posix(filp
, id
);
1128 EXPORT_SYMBOL(filp_close
);
1131 * Careful here! We test whether the file pointer is NULL before
1132 * releasing the fd. This ensures that one clone task can't release
1133 * an fd while another clone is opening it.
1135 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1138 struct files_struct
*files
= current
->files
;
1139 struct fdtable
*fdt
;
1142 spin_lock(&files
->file_lock
);
1143 fdt
= files_fdtable(files
);
1144 if (fd
>= fdt
->max_fds
)
1149 rcu_assign_pointer(fdt
->fd
[fd
], NULL
);
1150 FD_CLR(fd
, fdt
->close_on_exec
);
1151 __put_unused_fd(files
, fd
);
1152 spin_unlock(&files
->file_lock
);
1153 retval
= filp_close(filp
, files
);
1155 /* can't restart close syscall because file table entry was cleared */
1156 if (unlikely(retval
== -ERESTARTSYS
||
1157 retval
== -ERESTARTNOINTR
||
1158 retval
== -ERESTARTNOHAND
||
1159 retval
== -ERESTART_RESTARTBLOCK
))
1165 spin_unlock(&files
->file_lock
);
1168 EXPORT_SYMBOL(sys_close
);
1171 * This routine simulates a hangup on the tty, to arrange that users
1172 * are given clean terminals at login time.
1174 SYSCALL_DEFINE0(vhangup
)
1176 if (capable(CAP_SYS_TTY_CONFIG
)) {
1184 * Called when an inode is about to be open.
1185 * We use this to disallow opening large files on 32bit systems if
1186 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1187 * on this flag in sys_open.
1189 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1191 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1196 EXPORT_SYMBOL(generic_file_open
);
1199 * This is used by subsystems that don't want seekable
1202 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1204 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1208 EXPORT_SYMBOL(nonseekable_open
);