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
);
71 long vfs_truncate(const struct path
*path
, loff_t length
)
74 struct dentry
*upperdentry
;
77 inode
= path
->dentry
->d_inode
;
79 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
80 if (S_ISDIR(inode
->i_mode
))
82 if (!S_ISREG(inode
->i_mode
))
85 error
= mnt_want_write(path
->mnt
);
89 error
= inode_permission(inode
, MAY_WRITE
);
91 goto mnt_drop_write_and_out
;
95 goto mnt_drop_write_and_out
;
98 * If this is an overlayfs then do as if opening the file so we get
99 * write access on the upper inode, not on the overlay inode. For
100 * non-overlay filesystems d_real() is an identity function.
102 upperdentry
= d_real(path
->dentry
, NULL
, O_WRONLY
);
103 error
= PTR_ERR(upperdentry
);
104 if (IS_ERR(upperdentry
))
105 goto mnt_drop_write_and_out
;
107 error
= get_write_access(upperdentry
->d_inode
);
109 goto mnt_drop_write_and_out
;
112 * Make sure that there are no leases. get_write_access() protects
113 * against the truncate racing with a lease-granting setlease().
115 error
= break_lease(inode
, O_WRONLY
);
117 goto put_write_and_out
;
119 error
= locks_verify_truncate(inode
, NULL
, length
);
121 error
= security_path_truncate(path
);
123 error
= do_truncate(path
->dentry
, length
, 0, NULL
);
126 put_write_access(upperdentry
->d_inode
);
127 mnt_drop_write_and_out
:
128 mnt_drop_write(path
->mnt
);
132 EXPORT_SYMBOL_GPL(vfs_truncate
);
134 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
136 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
140 if (length
< 0) /* sorry, but loff_t says... */
144 error
= user_path_at(AT_FDCWD
, pathname
, lookup_flags
, &path
);
146 error
= vfs_truncate(&path
, length
);
149 if (retry_estale(error
, lookup_flags
)) {
150 lookup_flags
|= LOOKUP_REVAL
;
156 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
158 return do_sys_truncate(path
, length
);
162 COMPAT_SYSCALL_DEFINE2(truncate
, const char __user
*, path
, compat_off_t
, length
)
164 return do_sys_truncate(path
, length
);
168 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
171 struct dentry
*dentry
;
183 /* explicitly opened as large or we are on 64-bit box */
184 if (f
.file
->f_flags
& O_LARGEFILE
)
187 dentry
= f
.file
->f_path
.dentry
;
188 inode
= dentry
->d_inode
;
190 if (!S_ISREG(inode
->i_mode
) || !(f
.file
->f_mode
& FMODE_WRITE
))
194 /* Cannot ftruncate over 2^31 bytes without large file support */
195 if (small
&& length
> MAX_NON_LFS
)
199 if (IS_APPEND(inode
))
202 sb_start_write(inode
->i_sb
);
203 error
= locks_verify_truncate(inode
, f
.file
, length
);
205 error
= security_path_truncate(&f
.file
->f_path
);
207 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, f
.file
);
208 sb_end_write(inode
->i_sb
);
215 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
217 return do_sys_ftruncate(fd
, length
, 1);
221 COMPAT_SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, compat_ulong_t
, length
)
223 return do_sys_ftruncate(fd
, length
, 1);
227 /* LFS versions of truncate are only needed on 32 bit machines */
228 #if BITS_PER_LONG == 32
229 SYSCALL_DEFINE2(truncate64
, const char __user
*, path
, loff_t
, length
)
231 return do_sys_truncate(path
, length
);
234 SYSCALL_DEFINE2(ftruncate64
, unsigned int, fd
, loff_t
, length
)
236 return do_sys_ftruncate(fd
, length
, 0);
238 #endif /* BITS_PER_LONG == 32 */
241 int vfs_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
243 struct inode
*inode
= file_inode(file
);
246 if (offset
< 0 || len
<= 0)
249 /* Return error if mode is not supported */
250 if (mode
& ~FALLOC_FL_SUPPORTED_MASK
)
253 /* Punch hole and zero range are mutually exclusive */
254 if ((mode
& (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
)) ==
255 (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
))
258 /* Punch hole must have keep size set */
259 if ((mode
& FALLOC_FL_PUNCH_HOLE
) &&
260 !(mode
& FALLOC_FL_KEEP_SIZE
))
263 /* Collapse range should only be used exclusively. */
264 if ((mode
& FALLOC_FL_COLLAPSE_RANGE
) &&
265 (mode
& ~FALLOC_FL_COLLAPSE_RANGE
))
268 /* Insert range should only be used exclusively. */
269 if ((mode
& FALLOC_FL_INSERT_RANGE
) &&
270 (mode
& ~FALLOC_FL_INSERT_RANGE
))
273 /* Unshare range should only be used with allocate mode. */
274 if ((mode
& FALLOC_FL_UNSHARE_RANGE
) &&
275 (mode
& ~(FALLOC_FL_UNSHARE_RANGE
| FALLOC_FL_KEEP_SIZE
)))
278 if (!(file
->f_mode
& FMODE_WRITE
))
282 * We can only allow pure fallocate on append only files
284 if ((mode
& ~FALLOC_FL_KEEP_SIZE
) && IS_APPEND(inode
))
287 if (IS_IMMUTABLE(inode
))
291 * We cannot allow any fallocate operation on an active swapfile
293 if (IS_SWAPFILE(inode
))
297 * Revalidate the write permissions, in case security policy has
298 * changed since the files were opened.
300 ret
= security_file_permission(file
, MAY_WRITE
);
304 if (S_ISFIFO(inode
->i_mode
))
308 * Let individual file system decide if it supports preallocation
309 * for directories or not.
311 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
) &&
312 !S_ISBLK(inode
->i_mode
))
315 /* Check for wrap through zero too */
316 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
319 if (!file
->f_op
->fallocate
)
322 sb_start_write(inode
->i_sb
);
323 ret
= file
->f_op
->fallocate(file
, mode
, offset
, len
);
326 * Create inotify and fanotify events.
328 * To keep the logic simple always create events if fallocate succeeds.
329 * This implies that events are even created if the file size remains
330 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
333 fsnotify_modify(file
);
335 sb_end_write(inode
->i_sb
);
338 EXPORT_SYMBOL_GPL(vfs_fallocate
);
340 SYSCALL_DEFINE4(fallocate
, int, fd
, int, mode
, loff_t
, offset
, loff_t
, len
)
342 struct fd f
= fdget(fd
);
346 error
= vfs_fallocate(f
.file
, mode
, offset
, len
);
353 * access() needs to use the real uid/gid, not the effective uid/gid.
354 * We do this by temporarily clearing all FS-related capabilities and
355 * switching the fsuid/fsgid around to the real ones.
357 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
359 const struct cred
*old_cred
;
360 struct cred
*override_cred
;
364 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
366 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
369 override_cred
= prepare_creds();
373 override_cred
->fsuid
= override_cred
->uid
;
374 override_cred
->fsgid
= override_cred
->gid
;
376 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
377 /* Clear the capabilities if we switch to a non-root user */
378 kuid_t root_uid
= make_kuid(override_cred
->user_ns
, 0);
379 if (!uid_eq(override_cred
->uid
, root_uid
))
380 cap_clear(override_cred
->cap_effective
);
382 override_cred
->cap_effective
=
383 override_cred
->cap_permitted
;
386 old_cred
= override_creds(override_cred
);
388 res
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
392 inode
= d_backing_inode(path
.dentry
);
394 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
396 * MAY_EXEC on regular files is denied if the fs is mounted
397 * with the "noexec" flag.
400 if (path_noexec(&path
))
401 goto out_path_release
;
404 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
405 /* SuS v2 requires we report a read only fs too */
406 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
407 goto out_path_release
;
409 * This is a rare case where using __mnt_is_readonly()
410 * is OK without a mnt_want/drop_write() pair. Since
411 * no actual write to the fs is performed here, we do
412 * not need to telegraph to that to anyone.
414 * By doing this, we accept that this access is
415 * inherently racy and know that the fs may change
416 * state before we even see this result.
418 if (__mnt_is_readonly(path
.mnt
))
423 if (retry_estale(res
, lookup_flags
)) {
424 lookup_flags
|= LOOKUP_REVAL
;
428 revert_creds(old_cred
);
429 put_cred(override_cred
);
433 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
435 return sys_faccessat(AT_FDCWD
, filename
, mode
);
438 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
442 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
444 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
448 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
452 set_fs_pwd(current
->fs
, &path
);
456 if (retry_estale(error
, lookup_flags
)) {
457 lookup_flags
|= LOOKUP_REVAL
;
464 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
466 struct fd f
= fdget_raw(fd
);
474 inode
= file_inode(f
.file
);
477 if (!S_ISDIR(inode
->i_mode
))
480 error
= inode_permission(inode
, MAY_EXEC
| MAY_CHDIR
);
482 set_fs_pwd(current
->fs
, &f
.file
->f_path
);
489 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
493 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
495 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
499 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
504 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT
))
506 error
= security_path_chroot(&path
);
510 set_fs_root(current
->fs
, &path
);
514 if (retry_estale(error
, lookup_flags
)) {
515 lookup_flags
|= LOOKUP_REVAL
;
522 static int chmod_common(const struct path
*path
, umode_t mode
)
524 struct inode
*inode
= path
->dentry
->d_inode
;
525 struct inode
*delegated_inode
= NULL
;
526 struct iattr newattrs
;
529 error
= mnt_want_write(path
->mnt
);
534 error
= security_path_chmod(path
, mode
);
537 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
538 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
539 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
542 if (delegated_inode
) {
543 error
= break_deleg_wait(&delegated_inode
);
547 mnt_drop_write(path
->mnt
);
551 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, umode_t
, mode
)
553 struct fd f
= fdget(fd
);
558 err
= chmod_common(&f
.file
->f_path
, mode
);
564 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, umode_t
, mode
)
568 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
570 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
572 error
= chmod_common(&path
, mode
);
574 if (retry_estale(error
, lookup_flags
)) {
575 lookup_flags
|= LOOKUP_REVAL
;
582 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, umode_t
, mode
)
584 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
587 static int chown_common(const struct path
*path
, uid_t user
, gid_t group
)
589 struct inode
*inode
= path
->dentry
->d_inode
;
590 struct inode
*delegated_inode
= NULL
;
592 struct iattr newattrs
;
596 uid
= make_kuid(current_user_ns(), user
);
597 gid
= make_kgid(current_user_ns(), group
);
600 newattrs
.ia_valid
= ATTR_CTIME
;
601 if (user
!= (uid_t
) -1) {
604 newattrs
.ia_valid
|= ATTR_UID
;
605 newattrs
.ia_uid
= uid
;
607 if (group
!= (gid_t
) -1) {
610 newattrs
.ia_valid
|= ATTR_GID
;
611 newattrs
.ia_gid
= gid
;
613 if (!S_ISDIR(inode
->i_mode
))
615 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
617 error
= security_path_chown(path
, uid
, gid
);
619 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
621 if (delegated_inode
) {
622 error
= break_deleg_wait(&delegated_inode
);
629 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
630 gid_t
, group
, int, flag
)
636 if ((flag
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
639 lookup_flags
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
640 if (flag
& AT_EMPTY_PATH
)
641 lookup_flags
|= LOOKUP_EMPTY
;
643 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
646 error
= mnt_want_write(path
.mnt
);
649 error
= chown_common(&path
, user
, group
);
650 mnt_drop_write(path
.mnt
);
653 if (retry_estale(error
, lookup_flags
)) {
654 lookup_flags
|= LOOKUP_REVAL
;
661 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
663 return sys_fchownat(AT_FDCWD
, filename
, user
, group
, 0);
666 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
668 return sys_fchownat(AT_FDCWD
, filename
, user
, group
,
669 AT_SYMLINK_NOFOLLOW
);
672 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
674 struct fd f
= fdget(fd
);
680 error
= mnt_want_write_file(f
.file
);
684 error
= chown_common(&f
.file
->f_path
, user
, group
);
685 mnt_drop_write_file(f
.file
);
692 int open_check_o_direct(struct file
*f
)
694 /* NB: we're sure to have correct a_ops only after f_op->open */
695 if (f
->f_flags
& O_DIRECT
) {
696 if (!f
->f_mapping
->a_ops
|| !f
->f_mapping
->a_ops
->direct_IO
)
702 static int do_dentry_open(struct file
*f
,
704 int (*open
)(struct inode
*, struct file
*),
705 const struct cred
*cred
)
707 static const struct file_operations empty_fops
= {};
710 f
->f_mode
= OPEN_FMODE(f
->f_flags
) | FMODE_LSEEK
|
711 FMODE_PREAD
| FMODE_PWRITE
;
713 path_get(&f
->f_path
);
715 f
->f_mapping
= inode
->i_mapping
;
717 if (unlikely(f
->f_flags
& O_PATH
)) {
718 f
->f_mode
= FMODE_PATH
;
719 f
->f_op
= &empty_fops
;
723 if (f
->f_mode
& FMODE_WRITE
&& !special_file(inode
->i_mode
)) {
724 error
= get_write_access(inode
);
727 error
= __mnt_want_write(f
->f_path
.mnt
);
728 if (unlikely(error
)) {
729 put_write_access(inode
);
732 f
->f_mode
|= FMODE_WRITER
;
735 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
736 if (S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
))
737 f
->f_mode
|= FMODE_ATOMIC_POS
;
739 f
->f_op
= fops_get(inode
->i_fop
);
740 if (unlikely(WARN_ON(!f
->f_op
))) {
745 error
= security_file_open(f
, cred
);
749 error
= break_lease(locks_inode(f
), f
->f_flags
);
754 open
= f
->f_op
->open
;
756 error
= open(inode
, f
);
760 if ((f
->f_mode
& (FMODE_READ
| FMODE_WRITE
)) == FMODE_READ
)
761 i_readcount_inc(inode
);
762 if ((f
->f_mode
& FMODE_READ
) &&
763 likely(f
->f_op
->read
|| f
->f_op
->read_iter
))
764 f
->f_mode
|= FMODE_CAN_READ
;
765 if ((f
->f_mode
& FMODE_WRITE
) &&
766 likely(f
->f_op
->write
|| f
->f_op
->write_iter
))
767 f
->f_mode
|= FMODE_CAN_WRITE
;
769 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
771 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
777 if (f
->f_mode
& FMODE_WRITER
) {
778 put_write_access(inode
);
779 __mnt_drop_write(f
->f_path
.mnt
);
782 path_put(&f
->f_path
);
783 f
->f_path
.mnt
= NULL
;
784 f
->f_path
.dentry
= NULL
;
790 * finish_open - finish opening a file
791 * @file: file pointer
792 * @dentry: pointer to dentry
793 * @open: open callback
794 * @opened: state of open
796 * This can be used to finish opening a file passed to i_op->atomic_open().
798 * If the open callback is set to NULL, then the standard f_op->open()
799 * filesystem callback is substituted.
801 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
802 * the return value of d_splice_alias(), then the caller needs to perform dput()
803 * on it after finish_open().
805 * On successful return @file is a fully instantiated open file. After this, if
806 * an error occurs in ->atomic_open(), it needs to clean up with fput().
808 * Returns zero on success or -errno if the open failed.
810 int finish_open(struct file
*file
, struct dentry
*dentry
,
811 int (*open
)(struct inode
*, struct file
*),
815 BUG_ON(*opened
& FILE_OPENED
); /* once it's opened, it's opened */
817 file
->f_path
.dentry
= dentry
;
818 error
= do_dentry_open(file
, d_backing_inode(dentry
), open
,
821 *opened
|= FILE_OPENED
;
825 EXPORT_SYMBOL(finish_open
);
828 * finish_no_open - finish ->atomic_open() without opening the file
830 * @file: file pointer
831 * @dentry: dentry or NULL (as returned from ->lookup())
833 * This can be used to set the result of a successful lookup in ->atomic_open().
835 * NB: unlike finish_open() this function does consume the dentry reference and
836 * the caller need not dput() it.
838 * Returns "1" which must be the return value of ->atomic_open() after having
839 * called this function.
841 int finish_no_open(struct file
*file
, struct dentry
*dentry
)
843 file
->f_path
.dentry
= dentry
;
846 EXPORT_SYMBOL(finish_no_open
);
848 char *file_path(struct file
*filp
, char *buf
, int buflen
)
850 return d_path(&filp
->f_path
, buf
, buflen
);
852 EXPORT_SYMBOL(file_path
);
855 * vfs_open - open the file at the given path
856 * @path: path to open
857 * @file: newly allocated file with f_flag initialized
858 * @cred: credentials to use
860 int vfs_open(const struct path
*path
, struct file
*file
,
861 const struct cred
*cred
)
863 struct dentry
*dentry
= d_real(path
->dentry
, NULL
, file
->f_flags
);
866 return PTR_ERR(dentry
);
868 file
->f_path
= *path
;
869 return do_dentry_open(file
, d_backing_inode(dentry
), NULL
, cred
);
872 struct file
*dentry_open(const struct path
*path
, int flags
,
873 const struct cred
*cred
)
878 validate_creds(cred
);
880 /* We must always pass in a valid mount pointer. */
883 f
= get_empty_filp();
886 error
= vfs_open(path
, f
, cred
);
888 /* from now on we need fput() to dispose of f */
889 error
= open_check_o_direct(f
);
901 EXPORT_SYMBOL(dentry_open
);
903 static inline int build_open_flags(int flags
, umode_t mode
, struct open_flags
*op
)
905 int lookup_flags
= 0;
906 int acc_mode
= ACC_MODE(flags
);
908 if (flags
& (O_CREAT
| __O_TMPFILE
))
909 op
->mode
= (mode
& S_IALLUGO
) | S_IFREG
;
913 /* Must never be set by userspace */
914 flags
&= ~FMODE_NONOTIFY
& ~O_CLOEXEC
;
917 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
918 * check for O_DSYNC if the need any syncing at all we enforce it's
919 * always set instead of having to deal with possibly weird behaviour
920 * for malicious applications setting only __O_SYNC.
922 if (flags
& __O_SYNC
)
925 if (flags
& __O_TMPFILE
) {
926 if ((flags
& O_TMPFILE_MASK
) != O_TMPFILE
)
928 if (!(acc_mode
& MAY_WRITE
))
930 } else if (flags
& O_PATH
) {
932 * If we have O_PATH in the open flag. Then we
933 * cannot have anything other than the below set of flags
935 flags
&= O_DIRECTORY
| O_NOFOLLOW
| O_PATH
;
939 op
->open_flag
= flags
;
941 /* O_TRUNC implies we need access checks for write permissions */
943 acc_mode
|= MAY_WRITE
;
945 /* Allow the LSM permission hook to distinguish append
946 access from general write access. */
947 if (flags
& O_APPEND
)
948 acc_mode
|= MAY_APPEND
;
950 op
->acc_mode
= acc_mode
;
952 op
->intent
= flags
& O_PATH
? 0 : LOOKUP_OPEN
;
954 if (flags
& O_CREAT
) {
955 op
->intent
|= LOOKUP_CREATE
;
957 op
->intent
|= LOOKUP_EXCL
;
960 if (flags
& O_DIRECTORY
)
961 lookup_flags
|= LOOKUP_DIRECTORY
;
962 if (!(flags
& O_NOFOLLOW
))
963 lookup_flags
|= LOOKUP_FOLLOW
;
964 op
->lookup_flags
= lookup_flags
;
969 * file_open_name - open file and return file pointer
971 * @name: struct filename containing path to open
972 * @flags: open flags as per the open(2) second argument
973 * @mode: mode for the new file if O_CREAT is set, else ignored
975 * This is the helper to open a file from kernelspace if you really
976 * have to. But in generally you should not do this, so please move
977 * along, nothing to see here..
979 struct file
*file_open_name(struct filename
*name
, int flags
, umode_t mode
)
981 struct open_flags op
;
982 int err
= build_open_flags(flags
, mode
, &op
);
983 return err
? ERR_PTR(err
) : do_filp_open(AT_FDCWD
, name
, &op
);
987 * filp_open - open file and return file pointer
989 * @filename: path to open
990 * @flags: open flags as per the open(2) second argument
991 * @mode: mode for the new file if O_CREAT is set, else ignored
993 * This is the helper to open a file from kernelspace if you really
994 * have to. But in generally you should not do this, so please move
995 * along, nothing to see here..
997 struct file
*filp_open(const char *filename
, int flags
, umode_t mode
)
999 struct filename
*name
= getname_kernel(filename
);
1000 struct file
*file
= ERR_CAST(name
);
1002 if (!IS_ERR(name
)) {
1003 file
= file_open_name(name
, flags
, mode
);
1008 EXPORT_SYMBOL(filp_open
);
1010 struct file
*file_open_root(struct dentry
*dentry
, struct vfsmount
*mnt
,
1011 const char *filename
, int flags
, umode_t mode
)
1013 struct open_flags op
;
1014 int err
= build_open_flags(flags
, mode
, &op
);
1016 return ERR_PTR(err
);
1017 return do_file_open_root(dentry
, mnt
, filename
, &op
);
1019 EXPORT_SYMBOL(file_open_root
);
1021 struct file
*filp_clone_open(struct file
*oldfile
)
1026 file
= get_empty_filp();
1030 file
->f_flags
= oldfile
->f_flags
;
1031 retval
= vfs_open(&oldfile
->f_path
, file
, oldfile
->f_cred
);
1034 return ERR_PTR(retval
);
1039 EXPORT_SYMBOL(filp_clone_open
);
1041 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, umode_t mode
)
1043 struct open_flags op
;
1044 int fd
= build_open_flags(flags
, mode
, &op
);
1045 struct filename
*tmp
;
1050 tmp
= getname(filename
);
1052 return PTR_ERR(tmp
);
1054 fd
= get_unused_fd_flags(flags
);
1056 struct file
*f
= do_filp_open(dfd
, tmp
, &op
);
1063 trace_do_sys_open(tmp
->name
, flags
, mode
);
1070 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, umode_t
, mode
)
1072 if (force_o_largefile())
1073 flags
|= O_LARGEFILE
;
1075 return do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1078 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1081 if (force_o_largefile())
1082 flags
|= O_LARGEFILE
;
1084 return do_sys_open(dfd
, filename
, flags
, mode
);
1090 * For backward compatibility? Maybe this should be moved
1091 * into arch/i386 instead?
1093 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, umode_t
, mode
)
1095 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1101 * "id" is the POSIX thread ID. We use the
1102 * files pointer for this..
1104 int filp_close(struct file
*filp
, fl_owner_t id
)
1108 if (!file_count(filp
)) {
1109 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1113 if (filp
->f_op
->flush
)
1114 retval
= filp
->f_op
->flush(filp
, id
);
1116 if (likely(!(filp
->f_mode
& FMODE_PATH
))) {
1117 dnotify_flush(filp
, id
);
1118 locks_remove_posix(filp
, id
);
1124 EXPORT_SYMBOL(filp_close
);
1127 * Careful here! We test whether the file pointer is NULL before
1128 * releasing the fd. This ensures that one clone task can't release
1129 * an fd while another clone is opening it.
1131 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1133 int retval
= __close_fd(current
->files
, fd
);
1135 /* can't restart close syscall because file table entry was cleared */
1136 if (unlikely(retval
== -ERESTARTSYS
||
1137 retval
== -ERESTARTNOINTR
||
1138 retval
== -ERESTARTNOHAND
||
1139 retval
== -ERESTART_RESTARTBLOCK
))
1144 EXPORT_SYMBOL(sys_close
);
1147 * This routine simulates a hangup on the tty, to arrange that users
1148 * are given clean terminals at login time.
1150 SYSCALL_DEFINE0(vhangup
)
1152 if (capable(CAP_SYS_TTY_CONFIG
)) {
1160 * Called when an inode is about to be open.
1161 * We use this to disallow opening large files on 32bit systems if
1162 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1163 * on this flag in sys_open.
1165 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1167 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1172 EXPORT_SYMBOL(generic_file_open
);
1175 * This is used by subsystems that don't want seekable
1176 * file descriptors. The function is not supposed to ever fail, the only
1177 * reason it returns an 'int' and not 'void' is so that it can be plugged
1178 * directly into file_operations structure.
1180 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1182 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1186 EXPORT_SYMBOL(nonseekable_open
);