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/fsnotify.h>
12 #include <linux/module.h>
13 #include <linux/tty.h>
14 #include <linux/namei.h>
15 #include <linux/backing-dev.h>
16 #include <linux/capability.h>
17 #include <linux/securebits.h>
18 #include <linux/security.h>
19 #include <linux/mount.h>
20 #include <linux/fcntl.h>
21 #include <linux/slab.h>
22 #include <asm/uaccess.h>
24 #include <linux/personality.h>
25 #include <linux/pagemap.h>
26 #include <linux/syscalls.h>
27 #include <linux/rcupdate.h>
28 #include <linux/audit.h>
29 #include <linux/falloc.h>
30 #include <linux/fs_struct.h>
31 #include <linux/ima.h>
32 #include <linux/dnotify.h>
33 #include <linux/compat.h>
37 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
41 struct iattr newattrs
;
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
47 newattrs
.ia_size
= length
;
48 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
50 newattrs
.ia_file
= filp
;
51 newattrs
.ia_valid
|= ATTR_FILE
;
54 /* Remove suid/sgid on truncate too */
55 ret
= should_remove_suid(dentry
);
57 newattrs
.ia_valid
|= ret
| ATTR_FORCE
;
59 mutex_lock(&dentry
->d_inode
->i_mutex
);
60 /* Note any delegations or leases have already been broken: */
61 ret
= notify_change(dentry
, &newattrs
, NULL
);
62 mutex_unlock(&dentry
->d_inode
->i_mutex
);
66 long vfs_truncate(struct path
*path
, loff_t length
)
71 inode
= path
->dentry
->d_inode
;
73 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
74 if (S_ISDIR(inode
->i_mode
))
76 if (!S_ISREG(inode
->i_mode
))
79 error
= mnt_want_write(path
->mnt
);
83 error
= inode_permission(inode
, MAY_WRITE
);
85 goto mnt_drop_write_and_out
;
89 goto mnt_drop_write_and_out
;
91 error
= get_write_access(inode
);
93 goto mnt_drop_write_and_out
;
96 * Make sure that there are no leases. get_write_access() protects
97 * against the truncate racing with a lease-granting setlease().
99 error
= break_lease(inode
, O_WRONLY
);
101 goto put_write_and_out
;
103 error
= locks_verify_truncate(inode
, NULL
, length
);
105 error
= security_path_truncate(path
);
107 error
= do_truncate(path
->dentry
, length
, 0, NULL
);
110 put_write_access(inode
);
111 mnt_drop_write_and_out
:
112 mnt_drop_write(path
->mnt
);
116 EXPORT_SYMBOL_GPL(vfs_truncate
);
118 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
120 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
124 if (length
< 0) /* sorry, but loff_t says... */
128 error
= user_path_at(AT_FDCWD
, pathname
, lookup_flags
, &path
);
130 error
= vfs_truncate(&path
, length
);
133 if (retry_estale(error
, lookup_flags
)) {
134 lookup_flags
|= LOOKUP_REVAL
;
140 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
142 return do_sys_truncate(path
, length
);
146 COMPAT_SYSCALL_DEFINE2(truncate
, const char __user
*, path
, compat_off_t
, length
)
148 return do_sys_truncate(path
, length
);
152 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
155 struct dentry
*dentry
;
167 /* explicitly opened as large or we are on 64-bit box */
168 if (f
.file
->f_flags
& O_LARGEFILE
)
171 dentry
= f
.file
->f_path
.dentry
;
172 inode
= dentry
->d_inode
;
174 if (!S_ISREG(inode
->i_mode
) || !(f
.file
->f_mode
& FMODE_WRITE
))
178 /* Cannot ftruncate over 2^31 bytes without large file support */
179 if (small
&& length
> MAX_NON_LFS
)
183 if (IS_APPEND(inode
))
186 sb_start_write(inode
->i_sb
);
187 error
= locks_verify_truncate(inode
, f
.file
, length
);
189 error
= security_path_truncate(&f
.file
->f_path
);
191 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, f
.file
);
192 sb_end_write(inode
->i_sb
);
199 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
201 return do_sys_ftruncate(fd
, length
, 1);
205 COMPAT_SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, compat_ulong_t
, length
)
207 return do_sys_ftruncate(fd
, length
, 1);
211 /* LFS versions of truncate are only needed on 32 bit machines */
212 #if BITS_PER_LONG == 32
213 SYSCALL_DEFINE2(truncate64
, const char __user
*, path
, loff_t
, length
)
215 return do_sys_truncate(path
, length
);
218 SYSCALL_DEFINE2(ftruncate64
, unsigned int, fd
, loff_t
, length
)
220 return do_sys_ftruncate(fd
, length
, 0);
222 #endif /* BITS_PER_LONG == 32 */
225 int do_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
227 struct inode
*inode
= file_inode(file
);
230 if (offset
< 0 || len
<= 0)
233 /* Return error if mode is not supported */
234 if (mode
& ~(FALLOC_FL_KEEP_SIZE
| FALLOC_FL_PUNCH_HOLE
))
237 /* Punch hole must have keep size set */
238 if ((mode
& FALLOC_FL_PUNCH_HOLE
) &&
239 !(mode
& FALLOC_FL_KEEP_SIZE
))
242 if (!(file
->f_mode
& FMODE_WRITE
))
245 /* It's not possible punch hole on append only file */
246 if (mode
& FALLOC_FL_PUNCH_HOLE
&& IS_APPEND(inode
))
249 if (IS_IMMUTABLE(inode
))
253 * Revalidate the write permissions, in case security policy has
254 * changed since the files were opened.
256 ret
= security_file_permission(file
, MAY_WRITE
);
260 if (S_ISFIFO(inode
->i_mode
))
264 * Let individual file system decide if it supports preallocation
265 * for directories or not.
267 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
))
270 /* Check for wrap through zero too */
271 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
274 if (!file
->f_op
->fallocate
)
277 sb_start_write(inode
->i_sb
);
278 ret
= file
->f_op
->fallocate(file
, mode
, offset
, len
);
279 sb_end_write(inode
->i_sb
);
283 SYSCALL_DEFINE4(fallocate
, int, fd
, int, mode
, loff_t
, offset
, loff_t
, len
)
285 struct fd f
= fdget(fd
);
289 error
= do_fallocate(f
.file
, mode
, offset
, len
);
296 * access() needs to use the real uid/gid, not the effective uid/gid.
297 * We do this by temporarily clearing all FS-related capabilities and
298 * switching the fsuid/fsgid around to the real ones.
300 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
302 const struct cred
*old_cred
;
303 struct cred
*override_cred
;
307 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
309 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
312 override_cred
= prepare_creds();
316 override_cred
->fsuid
= override_cred
->uid
;
317 override_cred
->fsgid
= override_cred
->gid
;
319 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
320 /* Clear the capabilities if we switch to a non-root user */
321 kuid_t root_uid
= make_kuid(override_cred
->user_ns
, 0);
322 if (!uid_eq(override_cred
->uid
, root_uid
))
323 cap_clear(override_cred
->cap_effective
);
325 override_cred
->cap_effective
=
326 override_cred
->cap_permitted
;
329 old_cred
= override_creds(override_cred
);
331 res
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
335 inode
= path
.dentry
->d_inode
;
337 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
339 * MAY_EXEC on regular files is denied if the fs is mounted
340 * with the "noexec" flag.
343 if (path
.mnt
->mnt_flags
& MNT_NOEXEC
)
344 goto out_path_release
;
347 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
348 /* SuS v2 requires we report a read only fs too */
349 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
350 goto out_path_release
;
352 * This is a rare case where using __mnt_is_readonly()
353 * is OK without a mnt_want/drop_write() pair. Since
354 * no actual write to the fs is performed here, we do
355 * not need to telegraph to that to anyone.
357 * By doing this, we accept that this access is
358 * inherently racy and know that the fs may change
359 * state before we even see this result.
361 if (__mnt_is_readonly(path
.mnt
))
366 if (retry_estale(res
, lookup_flags
)) {
367 lookup_flags
|= LOOKUP_REVAL
;
371 revert_creds(old_cred
);
372 put_cred(override_cred
);
376 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
378 return sys_faccessat(AT_FDCWD
, filename
, mode
);
381 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
385 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
387 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
391 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
395 set_fs_pwd(current
->fs
, &path
);
399 if (retry_estale(error
, lookup_flags
)) {
400 lookup_flags
|= LOOKUP_REVAL
;
407 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
409 struct fd f
= fdget_raw(fd
);
417 inode
= file_inode(f
.file
);
420 if (!S_ISDIR(inode
->i_mode
))
423 error
= inode_permission(inode
, MAY_EXEC
| MAY_CHDIR
);
425 set_fs_pwd(current
->fs
, &f
.file
->f_path
);
432 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
436 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
438 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
442 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
447 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT
))
449 error
= security_path_chroot(&path
);
453 set_fs_root(current
->fs
, &path
);
457 if (retry_estale(error
, lookup_flags
)) {
458 lookup_flags
|= LOOKUP_REVAL
;
465 static int chmod_common(struct path
*path
, umode_t mode
)
467 struct inode
*inode
= path
->dentry
->d_inode
;
468 struct inode
*delegated_inode
= NULL
;
469 struct iattr newattrs
;
472 error
= mnt_want_write(path
->mnt
);
476 mutex_lock(&inode
->i_mutex
);
477 error
= security_path_chmod(path
, mode
);
480 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
481 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
482 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
484 mutex_unlock(&inode
->i_mutex
);
485 if (delegated_inode
) {
486 error
= break_deleg_wait(&delegated_inode
);
490 mnt_drop_write(path
->mnt
);
494 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, umode_t
, mode
)
496 struct fd f
= fdget(fd
);
500 audit_inode(NULL
, f
.file
->f_path
.dentry
, 0);
501 err
= chmod_common(&f
.file
->f_path
, mode
);
507 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, umode_t
, mode
)
511 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
513 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
515 error
= chmod_common(&path
, mode
);
517 if (retry_estale(error
, lookup_flags
)) {
518 lookup_flags
|= LOOKUP_REVAL
;
525 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, umode_t
, mode
)
527 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
530 static int chown_common(struct path
*path
, uid_t user
, gid_t group
)
532 struct inode
*inode
= path
->dentry
->d_inode
;
533 struct inode
*delegated_inode
= NULL
;
535 struct iattr newattrs
;
539 uid
= make_kuid(current_user_ns(), user
);
540 gid
= make_kgid(current_user_ns(), group
);
542 newattrs
.ia_valid
= ATTR_CTIME
;
543 if (user
!= (uid_t
) -1) {
546 newattrs
.ia_valid
|= ATTR_UID
;
547 newattrs
.ia_uid
= uid
;
549 if (group
!= (gid_t
) -1) {
552 newattrs
.ia_valid
|= ATTR_GID
;
553 newattrs
.ia_gid
= gid
;
555 if (!S_ISDIR(inode
->i_mode
))
557 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
559 mutex_lock(&inode
->i_mutex
);
560 error
= security_path_chown(path
, uid
, gid
);
562 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
563 mutex_unlock(&inode
->i_mutex
);
564 if (delegated_inode
) {
565 error
= break_deleg_wait(&delegated_inode
);
572 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
573 gid_t
, group
, int, flag
)
579 if ((flag
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
582 lookup_flags
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
583 if (flag
& AT_EMPTY_PATH
)
584 lookup_flags
|= LOOKUP_EMPTY
;
586 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
589 error
= mnt_want_write(path
.mnt
);
592 error
= chown_common(&path
, user
, group
);
593 mnt_drop_write(path
.mnt
);
596 if (retry_estale(error
, lookup_flags
)) {
597 lookup_flags
|= LOOKUP_REVAL
;
604 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
606 return sys_fchownat(AT_FDCWD
, filename
, user
, group
, 0);
609 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
611 return sys_fchownat(AT_FDCWD
, filename
, user
, group
,
612 AT_SYMLINK_NOFOLLOW
);
615 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
617 struct fd f
= fdget(fd
);
623 error
= mnt_want_write_file(f
.file
);
626 audit_inode(NULL
, f
.file
->f_path
.dentry
, 0);
627 error
= chown_common(&f
.file
->f_path
, user
, group
);
628 mnt_drop_write_file(f
.file
);
636 * You have to be very careful that these write
637 * counts get cleaned up in error cases and
638 * upon __fput(). This should probably never
639 * be called outside of __dentry_open().
641 static inline int __get_file_write_access(struct inode
*inode
,
642 struct vfsmount
*mnt
)
644 int error
= get_write_access(inode
);
647 error
= __mnt_want_write(mnt
);
649 put_write_access(inode
);
653 int open_check_o_direct(struct file
*f
)
655 /* NB: we're sure to have correct a_ops only after f_op->open */
656 if (f
->f_flags
& O_DIRECT
) {
657 if (!f
->f_mapping
->a_ops
||
658 ((!f
->f_mapping
->a_ops
->direct_IO
) &&
659 (!f
->f_mapping
->a_ops
->get_xip_mem
))) {
666 static int do_dentry_open(struct file
*f
,
667 int (*open
)(struct inode
*, struct file
*),
668 const struct cred
*cred
)
670 static const struct file_operations empty_fops
= {};
674 f
->f_mode
= OPEN_FMODE(f
->f_flags
) | FMODE_LSEEK
|
675 FMODE_PREAD
| FMODE_PWRITE
;
677 if (unlikely(f
->f_flags
& O_PATH
))
678 f
->f_mode
= FMODE_PATH
;
680 path_get(&f
->f_path
);
681 inode
= f
->f_inode
= f
->f_path
.dentry
->d_inode
;
682 if (f
->f_mode
& FMODE_WRITE
&& !special_file(inode
->i_mode
)) {
683 error
= __get_file_write_access(inode
, f
->f_path
.mnt
);
689 f
->f_mapping
= inode
->i_mapping
;
691 if (unlikely(f
->f_mode
& FMODE_PATH
)) {
692 f
->f_op
= &empty_fops
;
696 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
697 if (S_ISREG(inode
->i_mode
))
698 f
->f_mode
|= FMODE_ATOMIC_POS
;
700 f
->f_op
= fops_get(inode
->i_fop
);
701 if (unlikely(WARN_ON(!f
->f_op
))) {
706 error
= security_file_open(f
, cred
);
710 error
= break_lease(inode
, f
->f_flags
);
715 open
= f
->f_op
->open
;
717 error
= open(inode
, f
);
721 if ((f
->f_mode
& (FMODE_READ
| FMODE_WRITE
)) == FMODE_READ
)
722 i_readcount_inc(inode
);
724 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
726 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
732 if (f
->f_mode
& FMODE_WRITE
) {
733 if (!special_file(inode
->i_mode
)) {
735 * We don't consider this a real
736 * mnt_want/drop_write() pair
737 * because it all happenend right
738 * here, so just reset the state.
740 put_write_access(inode
);
742 __mnt_drop_write(f
->f_path
.mnt
);
746 path_put(&f
->f_path
);
747 f
->f_path
.mnt
= NULL
;
748 f
->f_path
.dentry
= NULL
;
754 * finish_open - finish opening a file
755 * @file: file pointer
756 * @dentry: pointer to dentry
757 * @open: open callback
758 * @opened: state of open
760 * This can be used to finish opening a file passed to i_op->atomic_open().
762 * If the open callback is set to NULL, then the standard f_op->open()
763 * filesystem callback is substituted.
765 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
766 * the return value of d_splice_alias(), then the caller needs to perform dput()
767 * on it after finish_open().
769 * On successful return @file is a fully instantiated open file. After this, if
770 * an error occurs in ->atomic_open(), it needs to clean up with fput().
772 * Returns zero on success or -errno if the open failed.
774 int finish_open(struct file
*file
, struct dentry
*dentry
,
775 int (*open
)(struct inode
*, struct file
*),
779 BUG_ON(*opened
& FILE_OPENED
); /* once it's opened, it's opened */
781 file
->f_path
.dentry
= dentry
;
782 error
= do_dentry_open(file
, open
, current_cred());
784 *opened
|= FILE_OPENED
;
788 EXPORT_SYMBOL(finish_open
);
791 * finish_no_open - finish ->atomic_open() without opening the file
793 * @file: file pointer
794 * @dentry: dentry or NULL (as returned from ->lookup())
796 * This can be used to set the result of a successful lookup in ->atomic_open().
798 * NB: unlike finish_open() this function does consume the dentry reference and
799 * the caller need not dput() it.
801 * Returns "1" which must be the return value of ->atomic_open() after having
802 * called this function.
804 int finish_no_open(struct file
*file
, struct dentry
*dentry
)
806 file
->f_path
.dentry
= dentry
;
809 EXPORT_SYMBOL(finish_no_open
);
811 struct file
*dentry_open(const struct path
*path
, int flags
,
812 const struct cred
*cred
)
817 validate_creds(cred
);
819 /* We must always pass in a valid mount pointer. */
822 f
= get_empty_filp();
826 error
= do_dentry_open(f
, NULL
, cred
);
828 /* from now on we need fput() to dispose of f */
829 error
= open_check_o_direct(f
);
841 EXPORT_SYMBOL(dentry_open
);
843 static inline int build_open_flags(int flags
, umode_t mode
, struct open_flags
*op
)
845 int lookup_flags
= 0;
848 if (flags
& (O_CREAT
| __O_TMPFILE
))
849 op
->mode
= (mode
& S_IALLUGO
) | S_IFREG
;
853 /* Must never be set by userspace */
854 flags
&= ~FMODE_NONOTIFY
& ~O_CLOEXEC
;
857 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
858 * check for O_DSYNC if the need any syncing at all we enforce it's
859 * always set instead of having to deal with possibly weird behaviour
860 * for malicious applications setting only __O_SYNC.
862 if (flags
& __O_SYNC
)
865 if (flags
& __O_TMPFILE
) {
866 if ((flags
& O_TMPFILE_MASK
) != O_TMPFILE
)
868 acc_mode
= MAY_OPEN
| ACC_MODE(flags
);
869 if (!(acc_mode
& MAY_WRITE
))
871 } else if (flags
& O_PATH
) {
873 * If we have O_PATH in the open flag. Then we
874 * cannot have anything other than the below set of flags
876 flags
&= O_DIRECTORY
| O_NOFOLLOW
| O_PATH
;
879 acc_mode
= MAY_OPEN
| ACC_MODE(flags
);
882 op
->open_flag
= flags
;
884 /* O_TRUNC implies we need access checks for write permissions */
886 acc_mode
|= MAY_WRITE
;
888 /* Allow the LSM permission hook to distinguish append
889 access from general write access. */
890 if (flags
& O_APPEND
)
891 acc_mode
|= MAY_APPEND
;
893 op
->acc_mode
= acc_mode
;
895 op
->intent
= flags
& O_PATH
? 0 : LOOKUP_OPEN
;
897 if (flags
& O_CREAT
) {
898 op
->intent
|= LOOKUP_CREATE
;
900 op
->intent
|= LOOKUP_EXCL
;
903 if (flags
& O_DIRECTORY
)
904 lookup_flags
|= LOOKUP_DIRECTORY
;
905 if (!(flags
& O_NOFOLLOW
))
906 lookup_flags
|= LOOKUP_FOLLOW
;
907 op
->lookup_flags
= lookup_flags
;
912 * file_open_name - open file and return file pointer
914 * @name: struct filename containing path to open
915 * @flags: open flags as per the open(2) second argument
916 * @mode: mode for the new file if O_CREAT is set, else ignored
918 * This is the helper to open a file from kernelspace if you really
919 * have to. But in generally you should not do this, so please move
920 * along, nothing to see here..
922 struct file
*file_open_name(struct filename
*name
, int flags
, umode_t mode
)
924 struct open_flags op
;
925 int err
= build_open_flags(flags
, mode
, &op
);
926 return err
? ERR_PTR(err
) : do_filp_open(AT_FDCWD
, name
, &op
);
930 * filp_open - open file and return file pointer
932 * @filename: path to open
933 * @flags: open flags as per the open(2) second argument
934 * @mode: mode for the new file if O_CREAT is set, else ignored
936 * This is the helper to open a file from kernelspace if you really
937 * have to. But in generally you should not do this, so please move
938 * along, nothing to see here..
940 struct file
*filp_open(const char *filename
, int flags
, umode_t mode
)
942 struct filename name
= {.name
= filename
};
943 return file_open_name(&name
, flags
, mode
);
945 EXPORT_SYMBOL(filp_open
);
947 struct file
*file_open_root(struct dentry
*dentry
, struct vfsmount
*mnt
,
948 const char *filename
, int flags
)
950 struct open_flags op
;
951 int err
= build_open_flags(flags
, 0, &op
);
955 return ERR_PTR(-EINVAL
);
956 if (!filename
&& (flags
& O_DIRECTORY
))
957 if (!dentry
->d_inode
->i_op
->lookup
)
958 return ERR_PTR(-ENOTDIR
);
959 return do_file_open_root(dentry
, mnt
, filename
, &op
);
961 EXPORT_SYMBOL(file_open_root
);
963 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, umode_t mode
)
965 struct open_flags op
;
966 int fd
= build_open_flags(flags
, mode
, &op
);
967 struct filename
*tmp
;
972 tmp
= getname(filename
);
976 fd
= get_unused_fd_flags(flags
);
978 struct file
*f
= do_filp_open(dfd
, tmp
, &op
);
991 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, umode_t
, mode
)
993 if (force_o_largefile())
994 flags
|= O_LARGEFILE
;
996 return do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
999 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1002 if (force_o_largefile())
1003 flags
|= O_LARGEFILE
;
1005 return do_sys_open(dfd
, filename
, flags
, mode
);
1011 * For backward compatibility? Maybe this should be moved
1012 * into arch/i386 instead?
1014 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, umode_t
, mode
)
1016 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1022 * "id" is the POSIX thread ID. We use the
1023 * files pointer for this..
1025 int filp_close(struct file
*filp
, fl_owner_t id
)
1029 if (!file_count(filp
)) {
1030 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1034 if (filp
->f_op
->flush
)
1035 retval
= filp
->f_op
->flush(filp
, id
);
1037 if (likely(!(filp
->f_mode
& FMODE_PATH
))) {
1038 dnotify_flush(filp
, id
);
1039 locks_remove_posix(filp
, id
);
1045 EXPORT_SYMBOL(filp_close
);
1048 * Careful here! We test whether the file pointer is NULL before
1049 * releasing the fd. This ensures that one clone task can't release
1050 * an fd while another clone is opening it.
1052 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1054 int retval
= __close_fd(current
->files
, fd
);
1056 /* can't restart close syscall because file table entry was cleared */
1057 if (unlikely(retval
== -ERESTARTSYS
||
1058 retval
== -ERESTARTNOINTR
||
1059 retval
== -ERESTARTNOHAND
||
1060 retval
== -ERESTART_RESTARTBLOCK
))
1065 EXPORT_SYMBOL(sys_close
);
1068 * This routine simulates a hangup on the tty, to arrange that users
1069 * are given clean terminals at login time.
1071 SYSCALL_DEFINE0(vhangup
)
1073 if (capable(CAP_SYS_TTY_CONFIG
)) {
1081 * Called when an inode is about to be open.
1082 * We use this to disallow opening large files on 32bit systems if
1083 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1084 * on this flag in sys_open.
1086 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1088 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1093 EXPORT_SYMBOL(generic_file_open
);
1096 * This is used by subsystems that don't want seekable
1097 * file descriptors. The function is not supposed to ever fail, the only
1098 * reason it returns an 'int' and not 'void' is so that it can be plugged
1099 * directly into file_operations structure.
1101 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1103 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1107 EXPORT_SYMBOL(nonseekable_open
);