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 asmlinkage
long sys_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
)))
142 asmlinkage
long sys_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
)))
161 asmlinkage
long sys_fstatfs(unsigned int fd
, struct statfs __user
* buf
)
171 error
= vfs_statfs_native(file
->f_path
.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_path
.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 /* Remove suid/sgid on truncate too */
218 newattrs
.ia_valid
|= should_remove_suid(dentry
);
220 mutex_lock(&dentry
->d_inode
->i_mutex
);
221 err
= notify_change(dentry
, &newattrs
);
222 mutex_unlock(&dentry
->d_inode
->i_mutex
);
226 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
233 if (length
< 0) /* sorry, but loff_t says... */
236 error
= user_path(pathname
, &path
);
239 inode
= path
.dentry
->d_inode
;
241 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
243 if (S_ISDIR(inode
->i_mode
))
247 if (!S_ISREG(inode
->i_mode
))
250 error
= mnt_want_write(path
.mnt
);
254 error
= inode_permission(inode
, MAY_WRITE
);
256 goto mnt_drop_write_and_out
;
259 if (IS_APPEND(inode
))
260 goto mnt_drop_write_and_out
;
262 error
= get_write_access(inode
);
264 goto mnt_drop_write_and_out
;
267 * Make sure that there are no leases. get_write_access() protects
268 * against the truncate racing with a lease-granting setlease().
270 error
= break_lease(inode
, FMODE_WRITE
);
272 goto put_write_and_out
;
274 error
= locks_verify_truncate(inode
, NULL
, length
);
277 error
= do_truncate(path
.dentry
, length
, 0, NULL
);
281 put_write_access(inode
);
282 mnt_drop_write_and_out
:
283 mnt_drop_write(path
.mnt
);
290 asmlinkage
long sys_truncate(const char __user
* path
, unsigned long length
)
292 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
293 return do_sys_truncate(path
, (long)length
);
296 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
298 struct inode
* inode
;
299 struct dentry
*dentry
;
311 /* explicitly opened as large or we are on 64-bit box */
312 if (file
->f_flags
& O_LARGEFILE
)
315 dentry
= file
->f_path
.dentry
;
316 inode
= dentry
->d_inode
;
318 if (!S_ISREG(inode
->i_mode
) || !(file
->f_mode
& FMODE_WRITE
))
322 /* Cannot ftruncate over 2^31 bytes without large file support */
323 if (small
&& length
> MAX_NON_LFS
)
327 if (IS_APPEND(inode
))
330 error
= locks_verify_truncate(inode
, file
, length
);
332 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, file
);
339 asmlinkage
long sys_ftruncate(unsigned int fd
, unsigned long length
)
341 long ret
= do_sys_ftruncate(fd
, length
, 1);
342 /* avoid REGPARM breakage on x86: */
343 asmlinkage_protect(2, ret
, fd
, length
);
347 /* LFS versions of truncate are only needed on 32 bit machines */
348 #if BITS_PER_LONG == 32
349 SYSCALL_DEFINE(truncate64
)(const char __user
* path
, loff_t length
)
351 return do_sys_truncate(path
, length
);
353 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
354 asmlinkage
long SyS_truncate64(long path
, loff_t length
)
356 return SYSC_truncate64((const char __user
*) path
, length
);
358 SYSCALL_ALIAS(sys_truncate64
, SyS_truncate64
);
361 SYSCALL_DEFINE(ftruncate64
)(unsigned int fd
, loff_t length
)
363 long ret
= do_sys_ftruncate(fd
, length
, 0);
364 /* avoid REGPARM breakage on x86: */
365 asmlinkage_protect(2, ret
, fd
, length
);
368 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
369 asmlinkage
long SyS_ftruncate64(long fd
, loff_t length
)
371 return SYSC_ftruncate64((unsigned int) fd
, length
);
373 SYSCALL_ALIAS(sys_ftruncate64
, SyS_ftruncate64
);
375 #endif /* BITS_PER_LONG == 32 */
377 SYSCALL_DEFINE(fallocate
)(int fd
, int mode
, loff_t offset
, loff_t len
)
383 if (offset
< 0 || len
<= 0)
386 /* Return error if mode is not supported */
388 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 inode
= file
->f_path
.dentry
->d_inode
;
408 if (S_ISFIFO(inode
->i_mode
))
413 * Let individual file system decide if it supports preallocation
414 * for directories or not.
416 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
))
420 /* Check for wrap through zero too */
421 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
424 if (inode
->i_op
&& inode
->i_op
->fallocate
)
425 ret
= inode
->i_op
->fallocate(inode
, mode
, offset
, len
);
434 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
435 asmlinkage
long SyS_fallocate(long fd
, long mode
, loff_t offset
, loff_t len
)
437 return SYSC_fallocate((int)fd
, (int)mode
, offset
, len
);
439 SYSCALL_ALIAS(sys_fallocate
, SyS_fallocate
);
443 * access() needs to use the real uid/gid, not the effective uid/gid.
444 * We do this by temporarily clearing all FS-related capabilities and
445 * switching the fsuid/fsgid around to the real ones.
447 asmlinkage
long sys_faccessat(int dfd
, const char __user
*filename
, int mode
)
451 int old_fsuid
, old_fsgid
;
452 kernel_cap_t
uninitialized_var(old_cap
); /* !SECURE_NO_SETUID_FIXUP */
455 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
458 old_fsuid
= current
->fsuid
;
459 old_fsgid
= current
->fsgid
;
461 current
->fsuid
= current
->uid
;
462 current
->fsgid
= current
->gid
;
464 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
466 * Clear the capabilities if we switch to a non-root user
468 #ifndef CONFIG_SECURITY_FILE_CAPABILITIES
470 * FIXME: There is a race here against sys_capset. The
471 * capabilities can change yet we will restore the old
472 * value below. We should hold task_capabilities_lock,
473 * but we cannot because user_path_at can sleep.
475 #endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
477 old_cap
= cap_set_effective(__cap_empty_set
);
479 old_cap
= cap_set_effective(current
->cap_permitted
);
482 res
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
486 inode
= path
.dentry
->d_inode
;
488 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
490 * MAY_EXEC on regular files is denied if the fs is mounted
491 * with the "noexec" flag.
494 if (path
.mnt
->mnt_flags
& MNT_NOEXEC
)
495 goto out_path_release
;
498 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
499 /* SuS v2 requires we report a read only fs too */
500 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
501 goto out_path_release
;
503 * This is a rare case where using __mnt_is_readonly()
504 * is OK without a mnt_want/drop_write() pair. Since
505 * no actual write to the fs is performed here, we do
506 * not need to telegraph to that to anyone.
508 * By doing this, we accept that this access is
509 * inherently racy and know that the fs may change
510 * state before we even see this result.
512 if (__mnt_is_readonly(path
.mnt
))
518 current
->fsuid
= old_fsuid
;
519 current
->fsgid
= old_fsgid
;
521 if (!issecure(SECURE_NO_SETUID_FIXUP
))
522 cap_set_effective(old_cap
);
527 asmlinkage
long sys_access(const char __user
*filename
, int mode
)
529 return sys_faccessat(AT_FDCWD
, filename
, mode
);
532 asmlinkage
long sys_chdir(const char __user
* filename
)
537 error
= user_path_dir(filename
, &path
);
541 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
545 set_fs_pwd(current
->fs
, &path
);
553 asmlinkage
long sys_fchdir(unsigned int fd
)
564 inode
= file
->f_path
.dentry
->d_inode
;
567 if (!S_ISDIR(inode
->i_mode
))
570 error
= inode_permission(inode
, MAY_EXEC
| MAY_ACCESS
);
572 set_fs_pwd(current
->fs
, &file
->f_path
);
579 asmlinkage
long sys_chroot(const char __user
* filename
)
584 error
= user_path_dir(filename
, &path
);
588 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_ACCESS
);
593 if (!capable(CAP_SYS_CHROOT
))
596 set_fs_root(current
->fs
, &path
);
604 asmlinkage
long sys_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
->f_path
.mnt
);
624 mutex_lock(&inode
->i_mutex
);
625 if (mode
== (mode_t
) -1)
626 mode
= inode
->i_mode
;
627 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
628 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
629 err
= notify_change(dentry
, &newattrs
);
630 mutex_unlock(&inode
->i_mutex
);
631 mnt_drop_write(file
->f_path
.mnt
);
638 asmlinkage
long sys_fchmodat(int dfd
, const char __user
*filename
,
644 struct iattr newattrs
;
646 error
= user_path_at(dfd
, filename
, LOOKUP_FOLLOW
, &path
);
649 inode
= path
.dentry
->d_inode
;
651 error
= mnt_want_write(path
.mnt
);
654 mutex_lock(&inode
->i_mutex
);
655 if (mode
== (mode_t
) -1)
656 mode
= inode
->i_mode
;
657 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
658 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
659 error
= notify_change(path
.dentry
, &newattrs
);
660 mutex_unlock(&inode
->i_mutex
);
661 mnt_drop_write(path
.mnt
);
668 asmlinkage
long sys_chmod(const char __user
*filename
, mode_t mode
)
670 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
673 static int chown_common(struct dentry
* dentry
, uid_t user
, gid_t group
)
675 struct inode
*inode
= dentry
->d_inode
;
677 struct iattr newattrs
;
679 newattrs
.ia_valid
= ATTR_CTIME
;
680 if (user
!= (uid_t
) -1) {
681 newattrs
.ia_valid
|= ATTR_UID
;
682 newattrs
.ia_uid
= user
;
684 if (group
!= (gid_t
) -1) {
685 newattrs
.ia_valid
|= ATTR_GID
;
686 newattrs
.ia_gid
= group
;
688 if (!S_ISDIR(inode
->i_mode
))
690 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
691 mutex_lock(&inode
->i_mutex
);
692 error
= notify_change(dentry
, &newattrs
);
693 mutex_unlock(&inode
->i_mutex
);
698 asmlinkage
long sys_chown(const char __user
* filename
, uid_t user
, gid_t group
)
703 error
= user_path(filename
, &path
);
706 error
= mnt_want_write(path
.mnt
);
709 error
= chown_common(path
.dentry
, user
, group
);
710 mnt_drop_write(path
.mnt
);
717 asmlinkage
long sys_fchownat(int dfd
, const char __user
*filename
, uid_t user
,
718 gid_t group
, int flag
)
724 if ((flag
& ~AT_SYMLINK_NOFOLLOW
) != 0)
727 follow
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
728 error
= user_path_at(dfd
, filename
, follow
, &path
);
731 error
= mnt_want_write(path
.mnt
);
734 error
= chown_common(path
.dentry
, user
, group
);
735 mnt_drop_write(path
.mnt
);
742 asmlinkage
long sys_lchown(const char __user
* filename
, uid_t user
, gid_t group
)
747 error
= user_lpath(filename
, &path
);
750 error
= mnt_want_write(path
.mnt
);
753 error
= chown_common(path
.dentry
, user
, group
);
754 mnt_drop_write(path
.mnt
);
762 asmlinkage
long sys_fchown(unsigned int fd
, uid_t user
, gid_t group
)
766 struct dentry
* dentry
;
772 error
= mnt_want_write(file
->f_path
.mnt
);
775 dentry
= file
->f_path
.dentry
;
776 audit_inode(NULL
, dentry
);
777 error
= chown_common(dentry
, user
, group
);
778 mnt_drop_write(file
->f_path
.mnt
);
786 * You have to be very careful that these write
787 * counts get cleaned up in error cases and
788 * upon __fput(). This should probably never
789 * be called outside of __dentry_open().
791 static inline int __get_file_write_access(struct inode
*inode
,
792 struct vfsmount
*mnt
)
795 error
= get_write_access(inode
);
799 * Do not take mount writer counts on
800 * special files since no writes to
801 * the mount itself will occur.
803 if (!special_file(inode
->i_mode
)) {
805 * Balanced in __fput()
807 error
= mnt_want_write(mnt
);
809 put_write_access(inode
);
814 static struct file
*__dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
,
815 int flags
, struct file
*f
,
816 int (*open
)(struct inode
*, struct file
*))
822 f
->f_mode
= ((flags
+1) & O_ACCMODE
) | FMODE_LSEEK
|
823 FMODE_PREAD
| FMODE_PWRITE
;
824 inode
= dentry
->d_inode
;
825 if (f
->f_mode
& FMODE_WRITE
) {
826 error
= __get_file_write_access(inode
, mnt
);
829 if (!special_file(inode
->i_mode
))
833 f
->f_mapping
= inode
->i_mapping
;
834 f
->f_path
.dentry
= dentry
;
837 f
->f_op
= fops_get(inode
->i_fop
);
838 file_move(f
, &inode
->i_sb
->s_files
);
840 error
= security_dentry_open(f
);
844 if (!open
&& f
->f_op
)
845 open
= f
->f_op
->open
;
847 error
= open(inode
, f
);
852 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
854 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
856 /* NB: we're sure to have correct a_ops only after f_op->open */
857 if (f
->f_flags
& O_DIRECT
) {
858 if (!f
->f_mapping
->a_ops
||
859 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
860 (!f
->f_mapping
->a_ops
->get_xip_mem
))) {
862 f
= ERR_PTR(-EINVAL
);
870 if (f
->f_mode
& FMODE_WRITE
) {
871 put_write_access(inode
);
872 if (!special_file(inode
->i_mode
)) {
874 * We don't consider this a real
875 * mnt_want/drop_write() pair
876 * because it all happenend right
877 * here, so just reset the state.
884 f
->f_path
.dentry
= NULL
;
885 f
->f_path
.mnt
= NULL
;
890 return ERR_PTR(error
);
894 * lookup_instantiate_filp - instantiates the open intent filp
895 * @nd: pointer to nameidata
896 * @dentry: pointer to dentry
897 * @open: open callback
899 * Helper for filesystems that want to use lookup open intents and pass back
900 * a fully instantiated struct file to the caller.
901 * This function is meant to be called from within a filesystem's
903 * Beware of calling it for non-regular files! Those ->open methods might block
904 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
905 * leading to a deadlock, as nobody can open that fifo anymore, because
906 * another process to open fifo will block on locked parent when doing lookup).
907 * Note that in case of error, nd->intent.open.file is destroyed, but the
908 * path information remains valid.
909 * If the open callback is set to NULL, then the standard f_op->open()
910 * filesystem callback is substituted.
912 struct file
*lookup_instantiate_filp(struct nameidata
*nd
, struct dentry
*dentry
,
913 int (*open
)(struct inode
*, struct file
*))
915 if (IS_ERR(nd
->intent
.open
.file
))
919 nd
->intent
.open
.file
= __dentry_open(dget(dentry
), mntget(nd
->path
.mnt
),
920 nd
->intent
.open
.flags
- 1,
921 nd
->intent
.open
.file
,
924 return nd
->intent
.open
.file
;
926 release_open_intent(nd
);
927 nd
->intent
.open
.file
= (struct file
*)dentry
;
930 EXPORT_SYMBOL_GPL(lookup_instantiate_filp
);
933 * nameidata_to_filp - convert a nameidata to an open filp.
934 * @nd: pointer to nameidata
937 * Note that this function destroys the original nameidata
939 struct file
*nameidata_to_filp(struct nameidata
*nd
, int flags
)
943 /* Pick up the filp from the open intent */
944 filp
= nd
->intent
.open
.file
;
945 /* Has the filesystem initialised the file for us? */
946 if (filp
->f_path
.dentry
== NULL
)
947 filp
= __dentry_open(nd
->path
.dentry
, nd
->path
.mnt
, flags
, filp
,
955 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
958 struct file
*dentry_open(struct dentry
*dentry
, struct vfsmount
*mnt
, int flags
)
964 * We must always pass in a valid mount pointer. Historically
965 * callers got away with not passing it, but we must enforce this at
966 * the earliest possible point now to avoid strange problems deep in the
970 printk(KERN_WARNING
"%s called with NULL vfsmount\n", __func__
);
972 return ERR_PTR(-EINVAL
);
976 f
= get_empty_filp();
980 return ERR_PTR(error
);
983 return __dentry_open(dentry
, mnt
, flags
, f
, NULL
);
985 EXPORT_SYMBOL(dentry_open
);
987 static void __put_unused_fd(struct files_struct
*files
, unsigned int fd
)
989 struct fdtable
*fdt
= files_fdtable(files
);
990 __FD_CLR(fd
, fdt
->open_fds
);
991 if (fd
< files
->next_fd
)
995 void put_unused_fd(unsigned int fd
)
997 struct files_struct
*files
= current
->files
;
998 spin_lock(&files
->file_lock
);
999 __put_unused_fd(files
, fd
);
1000 spin_unlock(&files
->file_lock
);
1003 EXPORT_SYMBOL(put_unused_fd
);
1006 * Install a file pointer in the fd array.
1008 * The VFS is full of places where we drop the files lock between
1009 * setting the open_fds bitmap and installing the file in the file
1010 * array. At any such point, we are vulnerable to a dup2() race
1011 * installing a file in the array before us. We need to detect this and
1012 * fput() the struct file we are about to overwrite in this case.
1014 * It should never happen - if we allow dup2() do it, _really_ bad things
1018 void fd_install(unsigned int fd
, struct file
*file
)
1020 struct files_struct
*files
= current
->files
;
1021 struct fdtable
*fdt
;
1022 spin_lock(&files
->file_lock
);
1023 fdt
= files_fdtable(files
);
1024 BUG_ON(fdt
->fd
[fd
] != NULL
);
1025 rcu_assign_pointer(fdt
->fd
[fd
], file
);
1026 spin_unlock(&files
->file_lock
);
1029 EXPORT_SYMBOL(fd_install
);
1031 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, int mode
)
1033 char *tmp
= getname(filename
);
1034 int fd
= PTR_ERR(tmp
);
1037 fd
= get_unused_fd_flags(flags
);
1039 struct file
*f
= do_filp_open(dfd
, tmp
, flags
, mode
);
1044 fsnotify_open(f
->f_path
.dentry
);
1053 asmlinkage
long sys_open(const char __user
*filename
, int flags
, int mode
)
1057 if (force_o_largefile())
1058 flags
|= O_LARGEFILE
;
1060 ret
= do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1061 /* avoid REGPARM breakage on x86: */
1062 asmlinkage_protect(3, ret
, filename
, flags
, mode
);
1066 asmlinkage
long sys_openat(int dfd
, const char __user
*filename
, int flags
,
1071 if (force_o_largefile())
1072 flags
|= O_LARGEFILE
;
1074 ret
= do_sys_open(dfd
, filename
, flags
, mode
);
1075 /* avoid REGPARM breakage on x86: */
1076 asmlinkage_protect(4, ret
, dfd
, filename
, flags
, mode
);
1083 * For backward compatibility? Maybe this should be moved
1084 * into arch/i386 instead?
1086 asmlinkage
long sys_creat(const char __user
* pathname
, int mode
)
1088 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1094 * "id" is the POSIX thread ID. We use the
1095 * files pointer for this..
1097 int filp_close(struct file
*filp
, fl_owner_t id
)
1101 if (!file_count(filp
)) {
1102 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1106 if (filp
->f_op
&& filp
->f_op
->flush
)
1107 retval
= filp
->f_op
->flush(filp
, id
);
1109 dnotify_flush(filp
, id
);
1110 locks_remove_posix(filp
, id
);
1115 EXPORT_SYMBOL(filp_close
);
1118 * Careful here! We test whether the file pointer is NULL before
1119 * releasing the fd. This ensures that one clone task can't release
1120 * an fd while another clone is opening it.
1122 asmlinkage
long sys_close(unsigned int fd
)
1125 struct files_struct
*files
= current
->files
;
1126 struct fdtable
*fdt
;
1129 spin_lock(&files
->file_lock
);
1130 fdt
= files_fdtable(files
);
1131 if (fd
>= fdt
->max_fds
)
1136 rcu_assign_pointer(fdt
->fd
[fd
], NULL
);
1137 FD_CLR(fd
, fdt
->close_on_exec
);
1138 __put_unused_fd(files
, fd
);
1139 spin_unlock(&files
->file_lock
);
1140 retval
= filp_close(filp
, files
);
1142 /* can't restart close syscall because file table entry was cleared */
1143 if (unlikely(retval
== -ERESTARTSYS
||
1144 retval
== -ERESTARTNOINTR
||
1145 retval
== -ERESTARTNOHAND
||
1146 retval
== -ERESTART_RESTARTBLOCK
))
1152 spin_unlock(&files
->file_lock
);
1156 EXPORT_SYMBOL(sys_close
);
1159 * This routine simulates a hangup on the tty, to arrange that users
1160 * are given clean terminals at login time.
1162 asmlinkage
long sys_vhangup(void)
1164 if (capable(CAP_SYS_TTY_CONFIG
)) {
1165 /* XXX: this needs locking */
1166 tty_vhangup(current
->signal
->tty
);
1173 * Called when an inode is about to be open.
1174 * We use this to disallow opening large files on 32bit systems if
1175 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1176 * on this flag in sys_open.
1178 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1180 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1185 EXPORT_SYMBOL(generic_file_open
);
1188 * This is used by subsystems that don't want seekable
1191 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1193 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1197 EXPORT_SYMBOL(nonseekable_open
);