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 <linux/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 #define CREATE_TRACE_POINTS
38 #include <trace/events/fs.h>
40 int do_truncate(struct dentry
*dentry
, loff_t length
, unsigned int time_attrs
,
44 struct iattr newattrs
;
46 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
50 newattrs
.ia_size
= length
;
51 newattrs
.ia_valid
= ATTR_SIZE
| time_attrs
;
53 newattrs
.ia_file
= filp
;
54 newattrs
.ia_valid
|= ATTR_FILE
;
57 /* Remove suid, sgid, and file capabilities on truncate too */
58 ret
= dentry_needs_remove_privs(dentry
);
62 newattrs
.ia_valid
|= ret
| ATTR_FORCE
;
64 inode_lock(dentry
->d_inode
);
65 /* Note any delegations or leases have already been broken: */
66 ret
= notify_change(dentry
, &newattrs
, NULL
);
67 inode_unlock(dentry
->d_inode
);
70 EXPORT_SYMBOL_GPL(do_truncate
);
72 long vfs_truncate(const struct path
*path
, loff_t length
)
75 struct dentry
*upperdentry
;
78 inode
= path
->dentry
->d_inode
;
80 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
81 if (S_ISDIR(inode
->i_mode
))
83 if (!S_ISREG(inode
->i_mode
))
86 error
= mnt_want_write(path
->mnt
);
90 error
= inode_permission(inode
, MAY_WRITE
);
92 goto mnt_drop_write_and_out
;
96 goto mnt_drop_write_and_out
;
99 * If this is an overlayfs then do as if opening the file so we get
100 * write access on the upper inode, not on the overlay inode. For
101 * non-overlay filesystems d_real() is an identity function.
103 upperdentry
= d_real(path
->dentry
, NULL
, O_WRONLY
);
104 error
= PTR_ERR(upperdentry
);
105 if (IS_ERR(upperdentry
))
106 goto mnt_drop_write_and_out
;
108 error
= get_write_access(upperdentry
->d_inode
);
110 goto mnt_drop_write_and_out
;
113 * Make sure that there are no leases. get_write_access() protects
114 * against the truncate racing with a lease-granting setlease().
116 error
= break_lease(inode
, O_WRONLY
);
118 goto put_write_and_out
;
120 error
= locks_verify_truncate(inode
, NULL
, length
);
122 error
= security_path_truncate(path
);
124 error
= do_truncate(path
->dentry
, length
, 0, NULL
);
127 put_write_access(upperdentry
->d_inode
);
128 mnt_drop_write_and_out
:
129 mnt_drop_write(path
->mnt
);
133 EXPORT_SYMBOL_GPL(vfs_truncate
);
135 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
137 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
141 if (length
< 0) /* sorry, but loff_t says... */
145 error
= user_path_at(AT_FDCWD
, pathname
, lookup_flags
, &path
);
147 error
= vfs_truncate(&path
, length
);
150 if (retry_estale(error
, lookup_flags
)) {
151 lookup_flags
|= LOOKUP_REVAL
;
157 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
159 return do_sys_truncate(path
, length
);
163 COMPAT_SYSCALL_DEFINE2(truncate
, const char __user
*, path
, compat_off_t
, length
)
165 return do_sys_truncate(path
, length
);
169 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
172 struct dentry
*dentry
;
184 /* explicitly opened as large or we are on 64-bit box */
185 if (f
.file
->f_flags
& O_LARGEFILE
)
188 dentry
= f
.file
->f_path
.dentry
;
189 inode
= dentry
->d_inode
;
191 if (!S_ISREG(inode
->i_mode
) || !(f
.file
->f_mode
& FMODE_WRITE
))
195 /* Cannot ftruncate over 2^31 bytes without large file support */
196 if (small
&& length
> MAX_NON_LFS
)
200 if (IS_APPEND(inode
))
203 sb_start_write(inode
->i_sb
);
204 error
= locks_verify_truncate(inode
, f
.file
, length
);
206 error
= security_path_truncate(&f
.file
->f_path
);
208 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, f
.file
);
209 sb_end_write(inode
->i_sb
);
216 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
218 return do_sys_ftruncate(fd
, length
, 1);
222 COMPAT_SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, compat_ulong_t
, length
)
224 return do_sys_ftruncate(fd
, length
, 1);
228 /* LFS versions of truncate are only needed on 32 bit machines */
229 #if BITS_PER_LONG == 32
230 SYSCALL_DEFINE2(truncate64
, const char __user
*, path
, loff_t
, length
)
232 return do_sys_truncate(path
, length
);
235 SYSCALL_DEFINE2(ftruncate64
, unsigned int, fd
, loff_t
, length
)
237 return do_sys_ftruncate(fd
, length
, 0);
239 #endif /* BITS_PER_LONG == 32 */
242 int vfs_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
244 struct inode
*inode
= file_inode(file
);
247 if (offset
< 0 || len
<= 0)
250 /* Return error if mode is not supported */
251 if (mode
& ~FALLOC_FL_SUPPORTED_MASK
)
254 /* Punch hole and zero range are mutually exclusive */
255 if ((mode
& (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
)) ==
256 (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
))
259 /* Punch hole must have keep size set */
260 if ((mode
& FALLOC_FL_PUNCH_HOLE
) &&
261 !(mode
& FALLOC_FL_KEEP_SIZE
))
264 /* Collapse range should only be used exclusively. */
265 if ((mode
& FALLOC_FL_COLLAPSE_RANGE
) &&
266 (mode
& ~FALLOC_FL_COLLAPSE_RANGE
))
269 /* Insert range should only be used exclusively. */
270 if ((mode
& FALLOC_FL_INSERT_RANGE
) &&
271 (mode
& ~FALLOC_FL_INSERT_RANGE
))
274 /* Unshare range should only be used with allocate mode. */
275 if ((mode
& FALLOC_FL_UNSHARE_RANGE
) &&
276 (mode
& ~(FALLOC_FL_UNSHARE_RANGE
| FALLOC_FL_KEEP_SIZE
)))
279 if (!(file
->f_mode
& FMODE_WRITE
))
283 * We can only allow pure fallocate on append only files
285 if ((mode
& ~FALLOC_FL_KEEP_SIZE
) && IS_APPEND(inode
))
288 if (IS_IMMUTABLE(inode
))
292 * We cannot allow any fallocate operation on an active swapfile
294 if (IS_SWAPFILE(inode
))
298 * Revalidate the write permissions, in case security policy has
299 * changed since the files were opened.
301 ret
= security_file_permission(file
, MAY_WRITE
);
305 if (S_ISFIFO(inode
->i_mode
))
309 * Let individual file system decide if it supports preallocation
310 * for directories or not.
312 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
) &&
313 !S_ISBLK(inode
->i_mode
))
316 /* Check for wrap through zero too */
317 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
320 if (!file
->f_op
->fallocate
)
323 sb_start_write(inode
->i_sb
);
324 ret
= file
->f_op
->fallocate(file
, mode
, offset
, len
);
327 * Create inotify and fanotify events.
329 * To keep the logic simple always create events if fallocate succeeds.
330 * This implies that events are even created if the file size remains
331 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
334 fsnotify_modify(file
);
336 sb_end_write(inode
->i_sb
);
339 EXPORT_SYMBOL_GPL(vfs_fallocate
);
341 SYSCALL_DEFINE4(fallocate
, int, fd
, int, mode
, loff_t
, offset
, loff_t
, len
)
343 struct fd f
= fdget(fd
);
347 error
= vfs_fallocate(f
.file
, mode
, offset
, len
);
354 * access() needs to use the real uid/gid, not the effective uid/gid.
355 * We do this by temporarily clearing all FS-related capabilities and
356 * switching the fsuid/fsgid around to the real ones.
358 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
360 const struct cred
*old_cred
;
361 struct cred
*override_cred
;
365 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
367 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
370 override_cred
= prepare_creds();
374 override_cred
->fsuid
= override_cred
->uid
;
375 override_cred
->fsgid
= override_cred
->gid
;
377 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
378 /* Clear the capabilities if we switch to a non-root user */
379 kuid_t root_uid
= make_kuid(override_cred
->user_ns
, 0);
380 if (!uid_eq(override_cred
->uid
, root_uid
))
381 cap_clear(override_cred
->cap_effective
);
383 override_cred
->cap_effective
=
384 override_cred
->cap_permitted
;
387 old_cred
= override_creds(override_cred
);
389 res
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
393 inode
= d_backing_inode(path
.dentry
);
395 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
397 * MAY_EXEC on regular files is denied if the fs is mounted
398 * with the "noexec" flag.
401 if (path_noexec(&path
))
402 goto out_path_release
;
405 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
406 /* SuS v2 requires we report a read only fs too */
407 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
408 goto out_path_release
;
410 * This is a rare case where using __mnt_is_readonly()
411 * is OK without a mnt_want/drop_write() pair. Since
412 * no actual write to the fs is performed here, we do
413 * not need to telegraph to that to anyone.
415 * By doing this, we accept that this access is
416 * inherently racy and know that the fs may change
417 * state before we even see this result.
419 if (__mnt_is_readonly(path
.mnt
))
424 if (retry_estale(res
, lookup_flags
)) {
425 lookup_flags
|= LOOKUP_REVAL
;
429 revert_creds(old_cred
);
430 put_cred(override_cred
);
434 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
436 return sys_faccessat(AT_FDCWD
, filename
, mode
);
439 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
443 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
445 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
449 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
453 set_fs_pwd(current
->fs
, &path
);
457 if (retry_estale(error
, lookup_flags
)) {
458 lookup_flags
|= LOOKUP_REVAL
;
465 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
467 struct fd f
= fdget_raw(fd
);
475 inode
= file_inode(f
.file
);
478 if (!S_ISDIR(inode
->i_mode
))
481 error
= inode_permission(inode
, MAY_EXEC
| MAY_CHDIR
);
483 set_fs_pwd(current
->fs
, &f
.file
->f_path
);
490 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
494 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
496 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
500 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
505 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT
))
507 error
= security_path_chroot(&path
);
511 set_fs_root(current
->fs
, &path
);
515 if (retry_estale(error
, lookup_flags
)) {
516 lookup_flags
|= LOOKUP_REVAL
;
523 static int chmod_common(const struct path
*path
, umode_t mode
)
525 struct inode
*inode
= path
->dentry
->d_inode
;
526 struct inode
*delegated_inode
= NULL
;
527 struct iattr newattrs
;
530 error
= mnt_want_write(path
->mnt
);
535 error
= security_path_chmod(path
, mode
);
538 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
539 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
540 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
543 if (delegated_inode
) {
544 error
= break_deleg_wait(&delegated_inode
);
548 mnt_drop_write(path
->mnt
);
552 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, umode_t
, mode
)
554 struct fd f
= fdget(fd
);
559 err
= chmod_common(&f
.file
->f_path
, mode
);
565 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, umode_t
, mode
)
569 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
571 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
573 error
= chmod_common(&path
, mode
);
575 if (retry_estale(error
, lookup_flags
)) {
576 lookup_flags
|= LOOKUP_REVAL
;
583 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, umode_t
, mode
)
585 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
588 static int chown_common(const struct path
*path
, uid_t user
, gid_t group
)
590 struct inode
*inode
= path
->dentry
->d_inode
;
591 struct inode
*delegated_inode
= NULL
;
593 struct iattr newattrs
;
597 uid
= make_kuid(current_user_ns(), user
);
598 gid
= make_kgid(current_user_ns(), group
);
601 newattrs
.ia_valid
= ATTR_CTIME
;
602 if (user
!= (uid_t
) -1) {
605 newattrs
.ia_valid
|= ATTR_UID
;
606 newattrs
.ia_uid
= uid
;
608 if (group
!= (gid_t
) -1) {
611 newattrs
.ia_valid
|= ATTR_GID
;
612 newattrs
.ia_gid
= gid
;
614 if (!S_ISDIR(inode
->i_mode
))
616 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
618 error
= security_path_chown(path
, uid
, gid
);
620 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
622 if (delegated_inode
) {
623 error
= break_deleg_wait(&delegated_inode
);
630 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
631 gid_t
, group
, int, flag
)
637 if ((flag
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
640 lookup_flags
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
641 if (flag
& AT_EMPTY_PATH
)
642 lookup_flags
|= LOOKUP_EMPTY
;
644 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
647 error
= mnt_want_write(path
.mnt
);
650 error
= chown_common(&path
, user
, group
);
651 mnt_drop_write(path
.mnt
);
654 if (retry_estale(error
, lookup_flags
)) {
655 lookup_flags
|= LOOKUP_REVAL
;
662 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
664 return sys_fchownat(AT_FDCWD
, filename
, user
, group
, 0);
667 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
669 return sys_fchownat(AT_FDCWD
, filename
, user
, group
,
670 AT_SYMLINK_NOFOLLOW
);
673 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
675 struct fd f
= fdget(fd
);
681 error
= mnt_want_write_file(f
.file
);
685 error
= chown_common(&f
.file
->f_path
, user
, group
);
686 mnt_drop_write_file(f
.file
);
693 int open_check_o_direct(struct file
*f
)
695 /* NB: we're sure to have correct a_ops only after f_op->open */
696 if (f
->f_flags
& O_DIRECT
) {
697 if (!f
->f_mapping
->a_ops
|| !f
->f_mapping
->a_ops
->direct_IO
)
702 EXPORT_SYMBOL_GPL(open_check_o_direct
);
704 static int do_dentry_open(struct file
*f
,
706 int (*open
)(struct inode
*, struct file
*),
707 const struct cred
*cred
)
709 static const struct file_operations empty_fops
= {};
712 f
->f_mode
= OPEN_FMODE(f
->f_flags
) | FMODE_LSEEK
|
713 FMODE_PREAD
| FMODE_PWRITE
;
715 path_get(&f
->f_path
);
717 f
->f_mapping
= inode
->i_mapping
;
719 if (unlikely(f
->f_flags
& O_PATH
)) {
720 f
->f_mode
= FMODE_PATH
;
721 f
->f_op
= &empty_fops
;
725 if (f
->f_mode
& FMODE_WRITE
&& !special_file(inode
->i_mode
)) {
726 error
= get_write_access(inode
);
729 error
= __mnt_want_write(f
->f_path
.mnt
);
730 if (unlikely(error
)) {
731 put_write_access(inode
);
734 f
->f_mode
|= FMODE_WRITER
;
737 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
738 if (S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
))
739 f
->f_mode
|= FMODE_ATOMIC_POS
;
741 f
->f_op
= fops_get(inode
->i_fop
);
742 if (unlikely(WARN_ON(!f
->f_op
))) {
747 error
= security_file_open(f
, cred
);
751 error
= break_lease(locks_inode(f
), f
->f_flags
);
756 open
= f
->f_op
->open
;
758 error
= open(inode
, f
);
762 if ((f
->f_mode
& (FMODE_READ
| FMODE_WRITE
)) == FMODE_READ
)
763 i_readcount_inc(inode
);
764 if ((f
->f_mode
& FMODE_READ
) &&
765 likely(f
->f_op
->read
|| f
->f_op
->read_iter
))
766 f
->f_mode
|= FMODE_CAN_READ
;
767 if ((f
->f_mode
& FMODE_WRITE
) &&
768 likely(f
->f_op
->write
|| f
->f_op
->write_iter
))
769 f
->f_mode
|= FMODE_CAN_WRITE
;
771 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
773 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
779 if (f
->f_mode
& FMODE_WRITER
) {
780 put_write_access(inode
);
781 __mnt_drop_write(f
->f_path
.mnt
);
784 path_put(&f
->f_path
);
785 f
->f_path
.mnt
= NULL
;
786 f
->f_path
.dentry
= NULL
;
792 * finish_open - finish opening a file
793 * @file: file pointer
794 * @dentry: pointer to dentry
795 * @open: open callback
796 * @opened: state of open
798 * This can be used to finish opening a file passed to i_op->atomic_open().
800 * If the open callback is set to NULL, then the standard f_op->open()
801 * filesystem callback is substituted.
803 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
804 * the return value of d_splice_alias(), then the caller needs to perform dput()
805 * on it after finish_open().
807 * On successful return @file is a fully instantiated open file. After this, if
808 * an error occurs in ->atomic_open(), it needs to clean up with fput().
810 * Returns zero on success or -errno if the open failed.
812 int finish_open(struct file
*file
, struct dentry
*dentry
,
813 int (*open
)(struct inode
*, struct file
*),
817 BUG_ON(*opened
& FILE_OPENED
); /* once it's opened, it's opened */
819 file
->f_path
.dentry
= dentry
;
820 error
= do_dentry_open(file
, d_backing_inode(dentry
), open
,
823 *opened
|= FILE_OPENED
;
827 EXPORT_SYMBOL(finish_open
);
830 * finish_no_open - finish ->atomic_open() without opening the file
832 * @file: file pointer
833 * @dentry: dentry or NULL (as returned from ->lookup())
835 * This can be used to set the result of a successful lookup in ->atomic_open().
837 * NB: unlike finish_open() this function does consume the dentry reference and
838 * the caller need not dput() it.
840 * Returns "1" which must be the return value of ->atomic_open() after having
841 * called this function.
843 int finish_no_open(struct file
*file
, struct dentry
*dentry
)
845 file
->f_path
.dentry
= dentry
;
848 EXPORT_SYMBOL(finish_no_open
);
850 char *file_path(struct file
*filp
, char *buf
, int buflen
)
852 return d_path(&filp
->f_path
, buf
, buflen
);
854 EXPORT_SYMBOL(file_path
);
857 * vfs_open - open the file at the given path
858 * @path: path to open
859 * @file: newly allocated file with f_flag initialized
860 * @cred: credentials to use
862 int vfs_open(const struct path
*path
, struct file
*file
,
863 const struct cred
*cred
)
865 struct dentry
*dentry
= d_real(path
->dentry
, NULL
, file
->f_flags
);
868 return PTR_ERR(dentry
);
870 file
->f_path
= *path
;
871 return do_dentry_open(file
, d_backing_inode(dentry
), NULL
, cred
);
874 struct file
*dentry_open(const struct path
*path
, int flags
,
875 const struct cred
*cred
)
880 validate_creds(cred
);
882 /* We must always pass in a valid mount pointer. */
885 f
= get_empty_filp();
888 error
= vfs_open(path
, f
, cred
);
890 /* from now on we need fput() to dispose of f */
891 error
= open_check_o_direct(f
);
903 EXPORT_SYMBOL(dentry_open
);
905 static inline int build_open_flags(int flags
, umode_t mode
, struct open_flags
*op
)
907 int lookup_flags
= 0;
908 int acc_mode
= ACC_MODE(flags
);
910 if (flags
& (O_CREAT
| __O_TMPFILE
))
911 op
->mode
= (mode
& S_IALLUGO
) | S_IFREG
;
915 /* Must never be set by userspace */
916 flags
&= ~FMODE_NONOTIFY
& ~O_CLOEXEC
;
919 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
920 * check for O_DSYNC if the need any syncing at all we enforce it's
921 * always set instead of having to deal with possibly weird behaviour
922 * for malicious applications setting only __O_SYNC.
924 if (flags
& __O_SYNC
)
927 if (flags
& __O_TMPFILE
) {
928 if ((flags
& O_TMPFILE_MASK
) != O_TMPFILE
)
930 if (!(acc_mode
& MAY_WRITE
))
932 } else if (flags
& O_PATH
) {
934 * If we have O_PATH in the open flag. Then we
935 * cannot have anything other than the below set of flags
937 flags
&= O_DIRECTORY
| O_NOFOLLOW
| O_PATH
;
941 op
->open_flag
= flags
;
943 /* O_TRUNC implies we need access checks for write permissions */
945 acc_mode
|= MAY_WRITE
;
947 /* Allow the LSM permission hook to distinguish append
948 access from general write access. */
949 if (flags
& O_APPEND
)
950 acc_mode
|= MAY_APPEND
;
952 op
->acc_mode
= acc_mode
;
954 op
->intent
= flags
& O_PATH
? 0 : LOOKUP_OPEN
;
956 if (flags
& O_CREAT
) {
957 op
->intent
|= LOOKUP_CREATE
;
959 op
->intent
|= LOOKUP_EXCL
;
962 if (flags
& O_DIRECTORY
)
963 lookup_flags
|= LOOKUP_DIRECTORY
;
964 if (!(flags
& O_NOFOLLOW
))
965 lookup_flags
|= LOOKUP_FOLLOW
;
966 op
->lookup_flags
= lookup_flags
;
971 * file_open_name - open file and return file pointer
973 * @name: struct filename containing path to open
974 * @flags: open flags as per the open(2) second argument
975 * @mode: mode for the new file if O_CREAT is set, else ignored
977 * This is the helper to open a file from kernelspace if you really
978 * have to. But in generally you should not do this, so please move
979 * along, nothing to see here..
981 struct file
*file_open_name(struct filename
*name
, int flags
, umode_t mode
)
983 struct open_flags op
;
984 int err
= build_open_flags(flags
, mode
, &op
);
985 return err
? ERR_PTR(err
) : do_filp_open(AT_FDCWD
, name
, &op
);
989 * filp_open - open file and return file pointer
991 * @filename: path to open
992 * @flags: open flags as per the open(2) second argument
993 * @mode: mode for the new file if O_CREAT is set, else ignored
995 * This is the helper to open a file from kernelspace if you really
996 * have to. But in generally you should not do this, so please move
997 * along, nothing to see here..
999 struct file
*filp_open(const char *filename
, int flags
, umode_t mode
)
1001 struct filename
*name
= getname_kernel(filename
);
1002 struct file
*file
= ERR_CAST(name
);
1004 if (!IS_ERR(name
)) {
1005 file
= file_open_name(name
, flags
, mode
);
1010 EXPORT_SYMBOL(filp_open
);
1012 struct file
*file_open_root(struct dentry
*dentry
, struct vfsmount
*mnt
,
1013 const char *filename
, int flags
, umode_t mode
)
1015 struct open_flags op
;
1016 int err
= build_open_flags(flags
, mode
, &op
);
1018 return ERR_PTR(err
);
1019 return do_file_open_root(dentry
, mnt
, filename
, &op
);
1021 EXPORT_SYMBOL(file_open_root
);
1023 struct file
*filp_clone_open(struct file
*oldfile
)
1028 file
= get_empty_filp();
1032 file
->f_flags
= oldfile
->f_flags
;
1033 retval
= vfs_open(&oldfile
->f_path
, file
, oldfile
->f_cred
);
1036 return ERR_PTR(retval
);
1041 EXPORT_SYMBOL(filp_clone_open
);
1043 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, umode_t mode
)
1045 struct open_flags op
;
1046 int fd
= build_open_flags(flags
, mode
, &op
);
1047 struct filename
*tmp
;
1052 tmp
= getname(filename
);
1054 return PTR_ERR(tmp
);
1056 fd
= get_unused_fd_flags(flags
);
1058 struct file
*f
= do_filp_open(dfd
, tmp
, &op
);
1065 trace_do_sys_open(tmp
->name
, flags
, mode
);
1072 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, umode_t
, mode
)
1074 if (force_o_largefile())
1075 flags
|= O_LARGEFILE
;
1077 return do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1080 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1083 if (force_o_largefile())
1084 flags
|= O_LARGEFILE
;
1086 return do_sys_open(dfd
, filename
, flags
, mode
);
1092 * For backward compatibility? Maybe this should be moved
1093 * into arch/i386 instead?
1095 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, umode_t
, mode
)
1097 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1103 * "id" is the POSIX thread ID. We use the
1104 * files pointer for this..
1106 int filp_close(struct file
*filp
, fl_owner_t id
)
1110 if (!file_count(filp
)) {
1111 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1115 if (filp
->f_op
->flush
)
1116 retval
= filp
->f_op
->flush(filp
, id
);
1118 if (likely(!(filp
->f_mode
& FMODE_PATH
))) {
1119 dnotify_flush(filp
, id
);
1120 locks_remove_posix(filp
, id
);
1126 EXPORT_SYMBOL(filp_close
);
1129 * Careful here! We test whether the file pointer is NULL before
1130 * releasing the fd. This ensures that one clone task can't release
1131 * an fd while another clone is opening it.
1133 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1135 int retval
= __close_fd(current
->files
, fd
);
1137 /* can't restart close syscall because file table entry was cleared */
1138 if (unlikely(retval
== -ERESTARTSYS
||
1139 retval
== -ERESTARTNOINTR
||
1140 retval
== -ERESTARTNOHAND
||
1141 retval
== -ERESTART_RESTARTBLOCK
))
1146 EXPORT_SYMBOL(sys_close
);
1149 * This routine simulates a hangup on the tty, to arrange that users
1150 * are given clean terminals at login time.
1152 SYSCALL_DEFINE0(vhangup
)
1154 if (capable(CAP_SYS_TTY_CONFIG
)) {
1162 * Called when an inode is about to be open.
1163 * We use this to disallow opening large files on 32bit systems if
1164 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1165 * on this flag in sys_open.
1167 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1169 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1174 EXPORT_SYMBOL(generic_file_open
);
1177 * This is used by subsystems that don't want seekable
1178 * file descriptors. The function is not supposed to ever fail, the only
1179 * reason it returns an 'int' and not 'void' is so that it can be plugged
1180 * directly into file_operations structure.
1182 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1184 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1188 EXPORT_SYMBOL(nonseekable_open
);