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 #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 mutex_lock(&dentry
->d_inode
->i_mutex
);
65 /* Note any delegations or leases have already been broken: */
66 ret
= notify_change(dentry
, &newattrs
, NULL
);
67 mutex_unlock(&dentry
->d_inode
->i_mutex
);
70 EXPORT_SYMBOL(do_truncate
);
72 long vfs_truncate(struct path
*path
, loff_t length
)
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
;
97 error
= get_write_access(inode
);
99 goto mnt_drop_write_and_out
;
102 * Make sure that there are no leases. get_write_access() protects
103 * against the truncate racing with a lease-granting setlease().
105 error
= break_lease(inode
, O_WRONLY
);
107 goto put_write_and_out
;
109 error
= locks_verify_truncate(inode
, NULL
, length
);
111 error
= security_path_truncate(path
);
113 error
= do_truncate(path
->dentry
, length
, 0, NULL
);
116 put_write_access(inode
);
117 mnt_drop_write_and_out
:
118 mnt_drop_write(path
->mnt
);
122 EXPORT_SYMBOL_GPL(vfs_truncate
);
124 static long do_sys_truncate(const char __user
*pathname
, loff_t length
)
126 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
130 if (length
< 0) /* sorry, but loff_t says... */
134 error
= user_path_at(AT_FDCWD
, pathname
, lookup_flags
, &path
);
136 error
= vfs_truncate(&path
, length
);
139 if (retry_estale(error
, lookup_flags
)) {
140 lookup_flags
|= LOOKUP_REVAL
;
146 SYSCALL_DEFINE2(truncate
, const char __user
*, path
, long, length
)
148 return do_sys_truncate(path
, length
);
152 COMPAT_SYSCALL_DEFINE2(truncate
, const char __user
*, path
, compat_off_t
, length
)
154 return do_sys_truncate(path
, length
);
158 static long do_sys_ftruncate(unsigned int fd
, loff_t length
, int small
)
161 struct dentry
*dentry
;
173 /* explicitly opened as large or we are on 64-bit box */
174 if (f
.file
->f_flags
& O_LARGEFILE
)
177 dentry
= f
.file
->f_path
.dentry
;
178 inode
= dentry
->d_inode
;
180 if (!S_ISREG(inode
->i_mode
) || !(f
.file
->f_mode
& FMODE_WRITE
))
184 /* Cannot ftruncate over 2^31 bytes without large file support */
185 if (small
&& length
> MAX_NON_LFS
)
189 if (IS_APPEND(inode
))
192 sb_start_write(inode
->i_sb
);
193 error
= locks_verify_truncate(inode
, f
.file
, length
);
195 error
= security_path_truncate(&f
.file
->f_path
);
197 error
= do_truncate(dentry
, length
, ATTR_MTIME
|ATTR_CTIME
, f
.file
);
198 sb_end_write(inode
->i_sb
);
205 SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, unsigned long, length
)
207 return do_sys_ftruncate(fd
, length
, 1);
211 COMPAT_SYSCALL_DEFINE2(ftruncate
, unsigned int, fd
, compat_ulong_t
, length
)
213 return do_sys_ftruncate(fd
, length
, 1);
217 /* LFS versions of truncate are only needed on 32 bit machines */
218 #if BITS_PER_LONG == 32
219 SYSCALL_DEFINE2(truncate64
, const char __user
*, path
, loff_t
, length
)
221 return do_sys_truncate(path
, length
);
224 SYSCALL_DEFINE2(ftruncate64
, unsigned int, fd
, loff_t
, length
)
226 return do_sys_ftruncate(fd
, length
, 0);
228 #endif /* BITS_PER_LONG == 32 */
231 int vfs_fallocate(struct file
*file
, int mode
, loff_t offset
, loff_t len
)
233 struct inode
*inode
= file_inode(file
);
236 if (offset
< 0 || len
<= 0)
239 /* Return error if mode is not supported */
240 if (mode
& ~FALLOC_FL_SUPPORTED_MASK
)
243 /* Punch hole and zero range are mutually exclusive */
244 if ((mode
& (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
)) ==
245 (FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_ZERO_RANGE
))
248 /* Punch hole must have keep size set */
249 if ((mode
& FALLOC_FL_PUNCH_HOLE
) &&
250 !(mode
& FALLOC_FL_KEEP_SIZE
))
253 /* Collapse range should only be used exclusively. */
254 if ((mode
& FALLOC_FL_COLLAPSE_RANGE
) &&
255 (mode
& ~FALLOC_FL_COLLAPSE_RANGE
))
258 /* Insert range should only be used exclusively. */
259 if ((mode
& FALLOC_FL_INSERT_RANGE
) &&
260 (mode
& ~FALLOC_FL_INSERT_RANGE
))
263 if (!(file
->f_mode
& FMODE_WRITE
))
267 * We can only allow pure fallocate on append only files
269 if ((mode
& ~FALLOC_FL_KEEP_SIZE
) && IS_APPEND(inode
))
272 if (IS_IMMUTABLE(inode
))
276 * We cannot allow any fallocate operation on an active swapfile
278 if (IS_SWAPFILE(inode
))
282 * Revalidate the write permissions, in case security policy has
283 * changed since the files were opened.
285 ret
= security_file_permission(file
, MAY_WRITE
);
289 if (S_ISFIFO(inode
->i_mode
))
293 * Let individual file system decide if it supports preallocation
294 * for directories or not.
296 if (!S_ISREG(inode
->i_mode
) && !S_ISDIR(inode
->i_mode
))
299 /* Check for wrap through zero too */
300 if (((offset
+ len
) > inode
->i_sb
->s_maxbytes
) || ((offset
+ len
) < 0))
303 if (!file
->f_op
->fallocate
)
306 sb_start_write(inode
->i_sb
);
307 ret
= file
->f_op
->fallocate(file
, mode
, offset
, len
);
310 * Create inotify and fanotify events.
312 * To keep the logic simple always create events if fallocate succeeds.
313 * This implies that events are even created if the file size remains
314 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
317 fsnotify_modify(file
);
319 sb_end_write(inode
->i_sb
);
322 EXPORT_SYMBOL_GPL(vfs_fallocate
);
324 SYSCALL_DEFINE4(fallocate
, int, fd
, int, mode
, loff_t
, offset
, loff_t
, len
)
326 struct fd f
= fdget(fd
);
330 error
= vfs_fallocate(f
.file
, mode
, offset
, len
);
337 * access() needs to use the real uid/gid, not the effective uid/gid.
338 * We do this by temporarily clearing all FS-related capabilities and
339 * switching the fsuid/fsgid around to the real ones.
341 SYSCALL_DEFINE3(faccessat
, int, dfd
, const char __user
*, filename
, int, mode
)
343 const struct cred
*old_cred
;
344 struct cred
*override_cred
;
348 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
350 if (mode
& ~S_IRWXO
) /* where's F_OK, X_OK, W_OK, R_OK? */
353 override_cred
= prepare_creds();
357 override_cred
->fsuid
= override_cred
->uid
;
358 override_cred
->fsgid
= override_cred
->gid
;
360 if (!issecure(SECURE_NO_SETUID_FIXUP
)) {
361 /* Clear the capabilities if we switch to a non-root user */
362 kuid_t root_uid
= make_kuid(override_cred
->user_ns
, 0);
363 if (!uid_eq(override_cred
->uid
, root_uid
))
364 cap_clear(override_cred
->cap_effective
);
366 override_cred
->cap_effective
=
367 override_cred
->cap_permitted
;
370 old_cred
= override_creds(override_cred
);
372 res
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
376 inode
= d_backing_inode(path
.dentry
);
378 if ((mode
& MAY_EXEC
) && S_ISREG(inode
->i_mode
)) {
380 * MAY_EXEC on regular files is denied if the fs is mounted
381 * with the "noexec" flag.
384 if (path_noexec(&path
))
385 goto out_path_release
;
388 res
= inode_permission(inode
, mode
| MAY_ACCESS
);
389 /* SuS v2 requires we report a read only fs too */
390 if (res
|| !(mode
& S_IWOTH
) || special_file(inode
->i_mode
))
391 goto out_path_release
;
393 * This is a rare case where using __mnt_is_readonly()
394 * is OK without a mnt_want/drop_write() pair. Since
395 * no actual write to the fs is performed here, we do
396 * not need to telegraph to that to anyone.
398 * By doing this, we accept that this access is
399 * inherently racy and know that the fs may change
400 * state before we even see this result.
402 if (__mnt_is_readonly(path
.mnt
))
407 if (retry_estale(res
, lookup_flags
)) {
408 lookup_flags
|= LOOKUP_REVAL
;
412 revert_creds(old_cred
);
413 put_cred(override_cred
);
417 SYSCALL_DEFINE2(access
, const char __user
*, filename
, int, mode
)
419 return sys_faccessat(AT_FDCWD
, filename
, mode
);
422 SYSCALL_DEFINE1(chdir
, const char __user
*, filename
)
426 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
428 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
432 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
436 set_fs_pwd(current
->fs
, &path
);
440 if (retry_estale(error
, lookup_flags
)) {
441 lookup_flags
|= LOOKUP_REVAL
;
448 SYSCALL_DEFINE1(fchdir
, unsigned int, fd
)
450 struct fd f
= fdget_raw(fd
);
458 inode
= file_inode(f
.file
);
461 if (!S_ISDIR(inode
->i_mode
))
464 error
= inode_permission(inode
, MAY_EXEC
| MAY_CHDIR
);
466 set_fs_pwd(current
->fs
, &f
.file
->f_path
);
473 SYSCALL_DEFINE1(chroot
, const char __user
*, filename
)
477 unsigned int lookup_flags
= LOOKUP_FOLLOW
| LOOKUP_DIRECTORY
;
479 error
= user_path_at(AT_FDCWD
, filename
, lookup_flags
, &path
);
483 error
= inode_permission(path
.dentry
->d_inode
, MAY_EXEC
| MAY_CHDIR
);
488 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT
))
490 error
= security_path_chroot(&path
);
494 set_fs_root(current
->fs
, &path
);
498 if (retry_estale(error
, lookup_flags
)) {
499 lookup_flags
|= LOOKUP_REVAL
;
506 static int chmod_common(struct path
*path
, umode_t mode
)
508 struct inode
*inode
= path
->dentry
->d_inode
;
509 struct inode
*delegated_inode
= NULL
;
510 struct iattr newattrs
;
513 error
= mnt_want_write(path
->mnt
);
517 mutex_lock(&inode
->i_mutex
);
518 error
= security_path_chmod(path
, mode
);
521 newattrs
.ia_mode
= (mode
& S_IALLUGO
) | (inode
->i_mode
& ~S_IALLUGO
);
522 newattrs
.ia_valid
= ATTR_MODE
| ATTR_CTIME
;
523 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
525 mutex_unlock(&inode
->i_mutex
);
526 if (delegated_inode
) {
527 error
= break_deleg_wait(&delegated_inode
);
531 mnt_drop_write(path
->mnt
);
535 SYSCALL_DEFINE2(fchmod
, unsigned int, fd
, umode_t
, mode
)
537 struct fd f
= fdget(fd
);
542 err
= chmod_common(&f
.file
->f_path
, mode
);
548 SYSCALL_DEFINE3(fchmodat
, int, dfd
, const char __user
*, filename
, umode_t
, mode
)
552 unsigned int lookup_flags
= LOOKUP_FOLLOW
;
554 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
556 error
= chmod_common(&path
, mode
);
558 if (retry_estale(error
, lookup_flags
)) {
559 lookup_flags
|= LOOKUP_REVAL
;
566 SYSCALL_DEFINE2(chmod
, const char __user
*, filename
, umode_t
, mode
)
568 return sys_fchmodat(AT_FDCWD
, filename
, mode
);
571 static int chown_common(struct path
*path
, uid_t user
, gid_t group
)
573 struct inode
*inode
= path
->dentry
->d_inode
;
574 struct inode
*delegated_inode
= NULL
;
576 struct iattr newattrs
;
580 uid
= make_kuid(current_user_ns(), user
);
581 gid
= make_kgid(current_user_ns(), group
);
584 newattrs
.ia_valid
= ATTR_CTIME
;
585 if (user
!= (uid_t
) -1) {
588 newattrs
.ia_valid
|= ATTR_UID
;
589 newattrs
.ia_uid
= uid
;
591 if (group
!= (gid_t
) -1) {
594 newattrs
.ia_valid
|= ATTR_GID
;
595 newattrs
.ia_gid
= gid
;
597 if (!S_ISDIR(inode
->i_mode
))
599 ATTR_KILL_SUID
| ATTR_KILL_SGID
| ATTR_KILL_PRIV
;
600 mutex_lock(&inode
->i_mutex
);
601 error
= security_path_chown(path
, uid
, gid
);
603 error
= notify_change(path
->dentry
, &newattrs
, &delegated_inode
);
604 mutex_unlock(&inode
->i_mutex
);
605 if (delegated_inode
) {
606 error
= break_deleg_wait(&delegated_inode
);
613 SYSCALL_DEFINE5(fchownat
, int, dfd
, const char __user
*, filename
, uid_t
, user
,
614 gid_t
, group
, int, flag
)
620 if ((flag
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
623 lookup_flags
= (flag
& AT_SYMLINK_NOFOLLOW
) ? 0 : LOOKUP_FOLLOW
;
624 if (flag
& AT_EMPTY_PATH
)
625 lookup_flags
|= LOOKUP_EMPTY
;
627 error
= user_path_at(dfd
, filename
, lookup_flags
, &path
);
630 error
= mnt_want_write(path
.mnt
);
633 error
= chown_common(&path
, user
, group
);
634 mnt_drop_write(path
.mnt
);
637 if (retry_estale(error
, lookup_flags
)) {
638 lookup_flags
|= LOOKUP_REVAL
;
645 SYSCALL_DEFINE3(chown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
647 return sys_fchownat(AT_FDCWD
, filename
, user
, group
, 0);
650 SYSCALL_DEFINE3(lchown
, const char __user
*, filename
, uid_t
, user
, gid_t
, group
)
652 return sys_fchownat(AT_FDCWD
, filename
, user
, group
,
653 AT_SYMLINK_NOFOLLOW
);
656 SYSCALL_DEFINE3(fchown
, unsigned int, fd
, uid_t
, user
, gid_t
, group
)
658 struct fd f
= fdget(fd
);
664 error
= mnt_want_write_file(f
.file
);
668 error
= chown_common(&f
.file
->f_path
, user
, group
);
669 mnt_drop_write_file(f
.file
);
676 int open_check_o_direct(struct file
*f
)
678 /* NB: we're sure to have correct a_ops only after f_op->open */
679 if (f
->f_flags
& O_DIRECT
) {
680 if (!f
->f_mapping
->a_ops
|| !f
->f_mapping
->a_ops
->direct_IO
)
685 EXPORT_SYMBOL(open_check_o_direct
);
687 static int do_dentry_open(struct file
*f
,
689 int (*open
)(struct inode
*, struct file
*),
690 const struct cred
*cred
)
692 static const struct file_operations empty_fops
= {};
695 f
->f_mode
= OPEN_FMODE(f
->f_flags
) | FMODE_LSEEK
|
696 FMODE_PREAD
| FMODE_PWRITE
;
698 path_get(&f
->f_path
);
700 f
->f_mapping
= inode
->i_mapping
;
702 if (unlikely(f
->f_flags
& O_PATH
)) {
703 f
->f_mode
= FMODE_PATH
;
704 f
->f_op
= &empty_fops
;
708 if (f
->f_mode
& FMODE_WRITE
&& !special_file(inode
->i_mode
)) {
709 error
= get_write_access(inode
);
712 error
= __mnt_want_write(f
->f_path
.mnt
);
713 if (unlikely(error
)) {
714 put_write_access(inode
);
717 f
->f_mode
|= FMODE_WRITER
;
720 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
721 if (S_ISREG(inode
->i_mode
))
722 f
->f_mode
|= FMODE_ATOMIC_POS
;
724 f
->f_op
= fops_get(inode
->i_fop
);
725 if (unlikely(WARN_ON(!f
->f_op
))) {
730 error
= security_file_open(f
, cred
);
734 error
= break_lease(inode
, f
->f_flags
);
739 open
= f
->f_op
->open
;
741 error
= open(inode
, f
);
745 if ((f
->f_mode
& (FMODE_READ
| FMODE_WRITE
)) == FMODE_READ
)
746 i_readcount_inc(inode
);
747 if ((f
->f_mode
& FMODE_READ
) &&
748 likely(f
->f_op
->read
|| f
->f_op
->read_iter
))
749 f
->f_mode
|= FMODE_CAN_READ
;
750 if ((f
->f_mode
& FMODE_WRITE
) &&
751 likely(f
->f_op
->write
|| f
->f_op
->write_iter
))
752 f
->f_mode
|= FMODE_CAN_WRITE
;
754 f
->f_flags
&= ~(O_CREAT
| O_EXCL
| O_NOCTTY
| O_TRUNC
);
756 file_ra_state_init(&f
->f_ra
, f
->f_mapping
->host
->i_mapping
);
762 if (f
->f_mode
& FMODE_WRITER
) {
763 put_write_access(inode
);
764 __mnt_drop_write(f
->f_path
.mnt
);
767 path_put(&f
->f_path
);
768 f
->f_path
.mnt
= NULL
;
769 f
->f_path
.dentry
= NULL
;
775 * finish_open - finish opening a file
776 * @file: file pointer
777 * @dentry: pointer to dentry
778 * @open: open callback
779 * @opened: state of open
781 * This can be used to finish opening a file passed to i_op->atomic_open().
783 * If the open callback is set to NULL, then the standard f_op->open()
784 * filesystem callback is substituted.
786 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
787 * the return value of d_splice_alias(), then the caller needs to perform dput()
788 * on it after finish_open().
790 * On successful return @file is a fully instantiated open file. After this, if
791 * an error occurs in ->atomic_open(), it needs to clean up with fput().
793 * Returns zero on success or -errno if the open failed.
795 int finish_open(struct file
*file
, struct dentry
*dentry
,
796 int (*open
)(struct inode
*, struct file
*),
800 BUG_ON(*opened
& FILE_OPENED
); /* once it's opened, it's opened */
802 file
->f_path
.dentry
= dentry
;
803 error
= do_dentry_open(file
, d_backing_inode(dentry
), open
,
806 *opened
|= FILE_OPENED
;
810 EXPORT_SYMBOL(finish_open
);
813 * finish_no_open - finish ->atomic_open() without opening the file
815 * @file: file pointer
816 * @dentry: dentry or NULL (as returned from ->lookup())
818 * This can be used to set the result of a successful lookup in ->atomic_open().
820 * NB: unlike finish_open() this function does consume the dentry reference and
821 * the caller need not dput() it.
823 * Returns "1" which must be the return value of ->atomic_open() after having
824 * called this function.
826 int finish_no_open(struct file
*file
, struct dentry
*dentry
)
828 file
->f_path
.dentry
= dentry
;
831 EXPORT_SYMBOL(finish_no_open
);
833 char *file_path(struct file
*filp
, char *buf
, int buflen
)
835 return d_path(&filp
->f_path
, buf
, buflen
);
837 EXPORT_SYMBOL(file_path
);
840 * vfs_open - open the file at the given path
841 * @path: path to open
842 * @file: newly allocated file with f_flag initialized
843 * @cred: credentials to use
845 int vfs_open(const struct path
*path
, struct file
*file
,
846 const struct cred
*cred
)
848 struct inode
*inode
= vfs_select_inode(path
->dentry
, file
->f_flags
);
851 return PTR_ERR(inode
);
853 file
->f_path
= *path
;
854 return do_dentry_open(file
, inode
, NULL
, cred
);
857 struct file
*dentry_open(const struct path
*path
, int flags
,
858 const struct cred
*cred
)
863 validate_creds(cred
);
865 /* We must always pass in a valid mount pointer. */
868 f
= get_empty_filp();
871 error
= vfs_open(path
, f
, cred
);
873 /* from now on we need fput() to dispose of f */
874 error
= open_check_o_direct(f
);
886 EXPORT_SYMBOL(dentry_open
);
888 static inline int build_open_flags(int flags
, umode_t mode
, struct open_flags
*op
)
890 int lookup_flags
= 0;
894 * Clear out all open flags we don't know about so that we don't report
895 * them in fcntl(F_GETFD) or similar interfaces.
897 flags
&= VALID_OPEN_FLAGS
;
899 if (flags
& (O_CREAT
| __O_TMPFILE
))
900 op
->mode
= (mode
& S_IALLUGO
) | S_IFREG
;
904 /* Must never be set by userspace */
905 flags
&= ~FMODE_NONOTIFY
& ~O_CLOEXEC
;
908 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
909 * check for O_DSYNC if the need any syncing at all we enforce it's
910 * always set instead of having to deal with possibly weird behaviour
911 * for malicious applications setting only __O_SYNC.
913 if (flags
& __O_SYNC
)
916 if (flags
& __O_TMPFILE
) {
917 if ((flags
& O_TMPFILE_MASK
) != O_TMPFILE
)
919 acc_mode
= MAY_OPEN
| ACC_MODE(flags
);
920 if (!(acc_mode
& MAY_WRITE
))
922 } else if (flags
& O_PATH
) {
924 * If we have O_PATH in the open flag. Then we
925 * cannot have anything other than the below set of flags
927 flags
&= O_DIRECTORY
| O_NOFOLLOW
| O_PATH
;
930 acc_mode
= MAY_OPEN
| ACC_MODE(flags
);
933 op
->open_flag
= flags
;
935 /* O_TRUNC implies we need access checks for write permissions */
937 acc_mode
|= MAY_WRITE
;
939 /* Allow the LSM permission hook to distinguish append
940 access from general write access. */
941 if (flags
& O_APPEND
)
942 acc_mode
|= MAY_APPEND
;
944 op
->acc_mode
= acc_mode
;
946 op
->intent
= flags
& O_PATH
? 0 : LOOKUP_OPEN
;
948 if (flags
& O_CREAT
) {
949 op
->intent
|= LOOKUP_CREATE
;
951 op
->intent
|= LOOKUP_EXCL
;
954 if (flags
& O_DIRECTORY
)
955 lookup_flags
|= LOOKUP_DIRECTORY
;
956 if (!(flags
& O_NOFOLLOW
))
957 lookup_flags
|= LOOKUP_FOLLOW
;
958 op
->lookup_flags
= lookup_flags
;
963 * file_open_name - open file and return file pointer
965 * @name: struct filename containing path to open
966 * @flags: open flags as per the open(2) second argument
967 * @mode: mode for the new file if O_CREAT is set, else ignored
969 * This is the helper to open a file from kernelspace if you really
970 * have to. But in generally you should not do this, so please move
971 * along, nothing to see here..
973 struct file
*file_open_name(struct filename
*name
, int flags
, umode_t mode
)
975 struct open_flags op
;
976 int err
= build_open_flags(flags
, mode
, &op
);
977 return err
? ERR_PTR(err
) : do_filp_open(AT_FDCWD
, name
, &op
);
981 * filp_open - open file and return file pointer
983 * @filename: path to open
984 * @flags: open flags as per the open(2) second argument
985 * @mode: mode for the new file if O_CREAT is set, else ignored
987 * This is the helper to open a file from kernelspace if you really
988 * have to. But in generally you should not do this, so please move
989 * along, nothing to see here..
991 struct file
*filp_open(const char *filename
, int flags
, umode_t mode
)
993 struct filename
*name
= getname_kernel(filename
);
994 struct file
*file
= ERR_CAST(name
);
997 file
= file_open_name(name
, flags
, mode
);
1002 EXPORT_SYMBOL(filp_open
);
1004 struct file
*file_open_root(struct dentry
*dentry
, struct vfsmount
*mnt
,
1005 const char *filename
, int flags
, umode_t mode
)
1007 struct open_flags op
;
1008 int err
= build_open_flags(flags
, mode
, &op
);
1010 return ERR_PTR(err
);
1011 return do_file_open_root(dentry
, mnt
, filename
, &op
);
1013 EXPORT_SYMBOL(file_open_root
);
1015 long do_sys_open(int dfd
, const char __user
*filename
, int flags
, umode_t mode
)
1017 struct open_flags op
;
1018 int fd
= build_open_flags(flags
, mode
, &op
);
1019 struct filename
*tmp
;
1024 tmp
= getname(filename
);
1026 return PTR_ERR(tmp
);
1028 fd
= get_unused_fd_flags(flags
);
1030 struct file
*f
= do_filp_open(dfd
, tmp
, &op
);
1037 trace_do_sys_open(tmp
->name
, flags
, mode
);
1044 SYSCALL_DEFINE3(open
, const char __user
*, filename
, int, flags
, umode_t
, mode
)
1046 if (force_o_largefile())
1047 flags
|= O_LARGEFILE
;
1049 return do_sys_open(AT_FDCWD
, filename
, flags
, mode
);
1052 SYSCALL_DEFINE4(openat
, int, dfd
, const char __user
*, filename
, int, flags
,
1055 if (force_o_largefile())
1056 flags
|= O_LARGEFILE
;
1058 return do_sys_open(dfd
, filename
, flags
, mode
);
1064 * For backward compatibility? Maybe this should be moved
1065 * into arch/i386 instead?
1067 SYSCALL_DEFINE2(creat
, const char __user
*, pathname
, umode_t
, mode
)
1069 return sys_open(pathname
, O_CREAT
| O_WRONLY
| O_TRUNC
, mode
);
1075 * "id" is the POSIX thread ID. We use the
1076 * files pointer for this..
1078 int filp_close(struct file
*filp
, fl_owner_t id
)
1082 if (!file_count(filp
)) {
1083 printk(KERN_ERR
"VFS: Close: file count is 0\n");
1087 if (filp
->f_op
->flush
)
1088 retval
= filp
->f_op
->flush(filp
, id
);
1090 if (likely(!(filp
->f_mode
& FMODE_PATH
))) {
1091 dnotify_flush(filp
, id
);
1092 locks_remove_posix(filp
, id
);
1098 EXPORT_SYMBOL(filp_close
);
1101 * Careful here! We test whether the file pointer is NULL before
1102 * releasing the fd. This ensures that one clone task can't release
1103 * an fd while another clone is opening it.
1105 SYSCALL_DEFINE1(close
, unsigned int, fd
)
1107 int retval
= __close_fd(current
->files
, fd
);
1109 /* can't restart close syscall because file table entry was cleared */
1110 if (unlikely(retval
== -ERESTARTSYS
||
1111 retval
== -ERESTARTNOINTR
||
1112 retval
== -ERESTARTNOHAND
||
1113 retval
== -ERESTART_RESTARTBLOCK
))
1118 EXPORT_SYMBOL(sys_close
);
1121 * This routine simulates a hangup on the tty, to arrange that users
1122 * are given clean terminals at login time.
1124 SYSCALL_DEFINE0(vhangup
)
1126 if (capable(CAP_SYS_TTY_CONFIG
)) {
1134 * Called when an inode is about to be open.
1135 * We use this to disallow opening large files on 32bit systems if
1136 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1137 * on this flag in sys_open.
1139 int generic_file_open(struct inode
* inode
, struct file
* filp
)
1141 if (!(filp
->f_flags
& O_LARGEFILE
) && i_size_read(inode
) > MAX_NON_LFS
)
1146 EXPORT_SYMBOL(generic_file_open
);
1149 * This is used by subsystems that don't want seekable
1150 * file descriptors. The function is not supposed to ever fail, the only
1151 * reason it returns an 'int' and not 'void' is so that it can be plugged
1152 * directly into file_operations structure.
1154 int nonseekable_open(struct inode
*inode
, struct file
*filp
)
1156 filp
->f_mode
&= ~(FMODE_LSEEK
| FMODE_PREAD
| FMODE_PWRITE
);
1160 EXPORT_SYMBOL(nonseekable_open
);