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, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
59 newattrs.ia_valid |= ret | ATTR_FORCE;
61 mutex_lock(&dentry->d_inode->i_mutex);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change(dentry, &newattrs, NULL);
64 mutex_unlock(&dentry->d_inode->i_mutex);
68 long vfs_truncate(struct path *path, loff_t length)
73 inode = path->dentry->d_inode;
75 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
76 if (S_ISDIR(inode->i_mode))
78 if (!S_ISREG(inode->i_mode))
81 error = mnt_want_write(path->mnt);
85 error = inode_permission(inode, MAY_WRITE);
87 goto mnt_drop_write_and_out;
91 goto mnt_drop_write_and_out;
93 error = get_write_access(inode);
95 goto mnt_drop_write_and_out;
98 * Make sure that there are no leases. get_write_access() protects
99 * against the truncate racing with a lease-granting setlease().
101 error = break_lease(inode, O_WRONLY);
103 goto put_write_and_out;
105 error = locks_verify_truncate(inode, NULL, length);
107 error = security_path_truncate(path);
109 error = do_truncate(path->dentry, length, 0, NULL);
112 put_write_access(inode);
113 mnt_drop_write_and_out:
114 mnt_drop_write(path->mnt);
118 EXPORT_SYMBOL_GPL(vfs_truncate);
120 static long do_sys_truncate(const char __user *pathname, loff_t length)
122 unsigned int lookup_flags = LOOKUP_FOLLOW;
126 if (length < 0) /* sorry, but loff_t says... */
130 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
132 error = vfs_truncate(&path, length);
135 if (retry_estale(error, lookup_flags)) {
136 lookup_flags |= LOOKUP_REVAL;
142 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
144 return do_sys_truncate(path, length);
148 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
150 return do_sys_truncate(path, length);
154 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
157 struct dentry *dentry;
169 /* explicitly opened as large or we are on 64-bit box */
170 if (f.file->f_flags & O_LARGEFILE)
173 dentry = f.file->f_path.dentry;
174 inode = dentry->d_inode;
176 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
180 /* Cannot ftruncate over 2^31 bytes without large file support */
181 if (small && length > MAX_NON_LFS)
185 if (IS_APPEND(inode))
188 sb_start_write(inode->i_sb);
189 error = locks_verify_truncate(inode, f.file, length);
191 error = security_path_truncate(&f.file->f_path);
193 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
194 sb_end_write(inode->i_sb);
201 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
203 return do_sys_ftruncate(fd, length, 1);
207 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
209 return do_sys_ftruncate(fd, length, 1);
213 /* LFS versions of truncate are only needed on 32 bit machines */
214 #if BITS_PER_LONG == 32
215 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
217 return do_sys_truncate(path, length);
220 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
222 return do_sys_ftruncate(fd, length, 0);
224 #endif /* BITS_PER_LONG == 32 */
227 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
229 struct inode *inode = file_inode(file);
232 if (offset < 0 || len <= 0)
235 /* Return error if mode is not supported */
236 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
239 /* Punch hole and zero range are mutually exclusive */
240 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
241 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
244 /* Punch hole must have keep size set */
245 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
246 !(mode & FALLOC_FL_KEEP_SIZE))
249 /* Collapse range should only be used exclusively. */
250 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
251 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
254 /* Insert range should only be used exclusively. */
255 if ((mode & FALLOC_FL_INSERT_RANGE) &&
256 (mode & ~FALLOC_FL_INSERT_RANGE))
259 if (!(file->f_mode & FMODE_WRITE))
263 * We can only allow pure fallocate on append only files
265 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
268 if (IS_IMMUTABLE(inode))
272 * We cannot allow any fallocate operation on an active swapfile
274 if (IS_SWAPFILE(inode))
278 * Revalidate the write permissions, in case security policy has
279 * changed since the files were opened.
281 ret = security_file_permission(file, MAY_WRITE);
285 if (S_ISFIFO(inode->i_mode))
289 * Let individual file system decide if it supports preallocation
290 * for directories or not.
292 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
295 /* Check for wrap through zero too */
296 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
299 if (!file->f_op->fallocate)
302 sb_start_write(inode->i_sb);
303 ret = file->f_op->fallocate(file, mode, offset, len);
306 * Create inotify and fanotify events.
308 * To keep the logic simple always create events if fallocate succeeds.
309 * This implies that events are even created if the file size remains
310 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
313 fsnotify_modify(file);
315 sb_end_write(inode->i_sb);
318 EXPORT_SYMBOL_GPL(vfs_fallocate);
320 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
322 struct fd f = fdget(fd);
326 error = vfs_fallocate(f.file, mode, offset, len);
333 * access() needs to use the real uid/gid, not the effective uid/gid.
334 * We do this by temporarily clearing all FS-related capabilities and
335 * switching the fsuid/fsgid around to the real ones.
337 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
339 const struct cred *old_cred;
340 struct cred *override_cred;
344 unsigned int lookup_flags = LOOKUP_FOLLOW;
346 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
349 override_cred = prepare_creds();
353 override_cred->fsuid = override_cred->uid;
354 override_cred->fsgid = override_cred->gid;
356 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
357 /* Clear the capabilities if we switch to a non-root user */
358 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
359 if (!uid_eq(override_cred->uid, root_uid))
360 cap_clear(override_cred->cap_effective);
362 override_cred->cap_effective =
363 override_cred->cap_permitted;
367 * The new set of credentials can *only* be used in
368 * task-synchronous circumstances, and does not need
369 * RCU freeing, unless somebody then takes a separate
372 * NOTE! This is _only_ true because this credential
373 * is used purely for override_creds() that installs
374 * it as the subjective cred. Other threads will be
375 * accessing ->real_cred, not the subjective cred.
377 * If somebody _does_ make a copy of this (using the
378 * 'get_current_cred()' function), that will clear the
379 * non_rcu field, because now that other user may be
380 * expecting RCU freeing. But normal thread-synchronous
381 * cred accesses will keep things non-RCY.
383 override_cred->non_rcu = 1;
385 old_cred = override_creds(override_cred);
387 res = user_path_at(dfd, filename, lookup_flags, &path);
391 inode = d_backing_inode(path.dentry);
393 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
395 * MAY_EXEC on regular files is denied if the fs is mounted
396 * with the "noexec" flag.
399 if (path_noexec(&path))
400 goto out_path_release;
403 res = inode_permission(inode, mode | MAY_ACCESS);
404 /* SuS v2 requires we report a read only fs too */
405 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
406 goto out_path_release;
408 * This is a rare case where using __mnt_is_readonly()
409 * is OK without a mnt_want/drop_write() pair. Since
410 * no actual write to the fs is performed here, we do
411 * not need to telegraph to that to anyone.
413 * By doing this, we accept that this access is
414 * inherently racy and know that the fs may change
415 * state before we even see this result.
417 if (__mnt_is_readonly(path.mnt))
422 if (retry_estale(res, lookup_flags)) {
423 lookup_flags |= LOOKUP_REVAL;
427 revert_creds(old_cred);
428 put_cred(override_cred);
432 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
434 return sys_faccessat(AT_FDCWD, filename, mode);
437 SYSCALL_DEFINE1(chdir, const char __user *, filename)
441 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
443 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
447 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
451 set_fs_pwd(current->fs, &path);
455 if (retry_estale(error, lookup_flags)) {
456 lookup_flags |= LOOKUP_REVAL;
463 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
465 struct fd f = fdget_raw(fd);
473 inode = file_inode(f.file);
476 if (!S_ISDIR(inode->i_mode))
479 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
481 set_fs_pwd(current->fs, &f.file->f_path);
488 SYSCALL_DEFINE1(chroot, const char __user *, filename)
492 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
494 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
498 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
503 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
505 error = security_path_chroot(&path);
509 set_fs_root(current->fs, &path);
513 if (retry_estale(error, lookup_flags)) {
514 lookup_flags |= LOOKUP_REVAL;
521 static int chmod_common(struct path *path, umode_t mode)
523 struct inode *inode = path->dentry->d_inode;
524 struct inode *delegated_inode = NULL;
525 struct iattr newattrs;
528 error = mnt_want_write(path->mnt);
532 mutex_lock(&inode->i_mutex);
533 error = security_path_chmod(path, mode);
536 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
537 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
538 error = notify_change(path->dentry, &newattrs, &delegated_inode);
540 mutex_unlock(&inode->i_mutex);
541 if (delegated_inode) {
542 error = break_deleg_wait(&delegated_inode);
546 mnt_drop_write(path->mnt);
550 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
552 struct fd f = fdget(fd);
557 err = chmod_common(&f.file->f_path, mode);
563 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
567 unsigned int lookup_flags = LOOKUP_FOLLOW;
569 error = user_path_at(dfd, filename, lookup_flags, &path);
571 error = chmod_common(&path, mode);
573 if (retry_estale(error, lookup_flags)) {
574 lookup_flags |= LOOKUP_REVAL;
581 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
583 return sys_fchmodat(AT_FDCWD, filename, mode);
586 static int chown_common(struct path *path, uid_t user, gid_t group)
588 struct inode *inode = path->dentry->d_inode;
589 struct inode *delegated_inode = NULL;
591 struct iattr newattrs;
595 uid = make_kuid(current_user_ns(), user);
596 gid = make_kgid(current_user_ns(), group);
599 newattrs.ia_valid = ATTR_CTIME;
600 if (user != (uid_t) -1) {
603 newattrs.ia_valid |= ATTR_UID;
604 newattrs.ia_uid = uid;
606 if (group != (gid_t) -1) {
609 newattrs.ia_valid |= ATTR_GID;
610 newattrs.ia_gid = gid;
612 if (!S_ISDIR(inode->i_mode))
614 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
615 mutex_lock(&inode->i_mutex);
616 error = security_path_chown(path, uid, gid);
618 error = notify_change(path->dentry, &newattrs, &delegated_inode);
619 mutex_unlock(&inode->i_mutex);
620 if (delegated_inode) {
621 error = break_deleg_wait(&delegated_inode);
628 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
629 gid_t, group, int, flag)
635 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
638 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
639 if (flag & AT_EMPTY_PATH)
640 lookup_flags |= LOOKUP_EMPTY;
642 error = user_path_at(dfd, filename, lookup_flags, &path);
645 error = mnt_want_write(path.mnt);
648 error = chown_common(&path, user, group);
649 mnt_drop_write(path.mnt);
652 if (retry_estale(error, lookup_flags)) {
653 lookup_flags |= LOOKUP_REVAL;
660 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
662 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
665 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
667 return sys_fchownat(AT_FDCWD, filename, user, group,
668 AT_SYMLINK_NOFOLLOW);
671 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
673 struct fd f = fdget(fd);
679 error = mnt_want_write_file(f.file);
683 error = chown_common(&f.file->f_path, user, group);
684 mnt_drop_write_file(f.file);
691 int open_check_o_direct(struct file *f)
693 /* NB: we're sure to have correct a_ops only after f_op->open */
694 if (f->f_flags & O_DIRECT) {
695 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
701 static int do_dentry_open(struct file *f,
703 int (*open)(struct inode *, struct file *),
704 const struct cred *cred)
706 static const struct file_operations empty_fops = {};
709 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
710 FMODE_PREAD | FMODE_PWRITE;
712 path_get(&f->f_path);
714 f->f_mapping = inode->i_mapping;
716 if (unlikely(f->f_flags & O_PATH)) {
717 f->f_mode = FMODE_PATH;
718 f->f_op = &empty_fops;
722 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
723 error = get_write_access(inode);
726 error = __mnt_want_write(f->f_path.mnt);
727 if (unlikely(error)) {
728 put_write_access(inode);
731 f->f_mode |= FMODE_WRITER;
734 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
735 if (S_ISREG(inode->i_mode))
736 f->f_mode |= FMODE_ATOMIC_POS;
738 f->f_op = fops_get(inode->i_fop);
739 if (unlikely(WARN_ON(!f->f_op))) {
744 error = security_file_open(f, cred);
748 error = break_lease(inode, f->f_flags);
753 open = f->f_op->open;
755 error = open(inode, f);
759 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
760 i_readcount_inc(inode);
761 if ((f->f_mode & FMODE_READ) &&
762 likely(f->f_op->read || f->f_op->read_iter))
763 f->f_mode |= FMODE_CAN_READ;
764 if ((f->f_mode & FMODE_WRITE) &&
765 likely(f->f_op->write || f->f_op->write_iter))
766 f->f_mode |= FMODE_CAN_WRITE;
768 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
770 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
776 if (f->f_mode & FMODE_WRITER) {
777 put_write_access(inode);
778 __mnt_drop_write(f->f_path.mnt);
781 path_put(&f->f_path);
782 f->f_path.mnt = NULL;
783 f->f_path.dentry = NULL;
789 * finish_open - finish opening a file
790 * @file: file pointer
791 * @dentry: pointer to dentry
792 * @open: open callback
793 * @opened: state of open
795 * This can be used to finish opening a file passed to i_op->atomic_open().
797 * If the open callback is set to NULL, then the standard f_op->open()
798 * filesystem callback is substituted.
800 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
801 * the return value of d_splice_alias(), then the caller needs to perform dput()
802 * on it after finish_open().
804 * On successful return @file is a fully instantiated open file. After this, if
805 * an error occurs in ->atomic_open(), it needs to clean up with fput().
807 * Returns zero on success or -errno if the open failed.
809 int finish_open(struct file *file, struct dentry *dentry,
810 int (*open)(struct inode *, struct file *),
814 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
816 file->f_path.dentry = dentry;
817 error = do_dentry_open(file, d_backing_inode(dentry), open,
820 *opened |= FILE_OPENED;
824 EXPORT_SYMBOL(finish_open);
827 * finish_no_open - finish ->atomic_open() without opening the file
829 * @file: file pointer
830 * @dentry: dentry or NULL (as returned from ->lookup())
832 * This can be used to set the result of a successful lookup in ->atomic_open().
834 * NB: unlike finish_open() this function does consume the dentry reference and
835 * the caller need not dput() it.
837 * Returns "1" which must be the return value of ->atomic_open() after having
838 * called this function.
840 int finish_no_open(struct file *file, struct dentry *dentry)
842 file->f_path.dentry = dentry;
845 EXPORT_SYMBOL(finish_no_open);
847 char *file_path(struct file *filp, char *buf, int buflen)
849 return d_path(&filp->f_path, buf, buflen);
851 EXPORT_SYMBOL(file_path);
854 * vfs_open - open the file at the given path
855 * @path: path to open
856 * @file: newly allocated file with f_flag initialized
857 * @cred: credentials to use
859 int vfs_open(const struct path *path, struct file *file,
860 const struct cred *cred)
862 struct inode *inode = vfs_select_inode(path->dentry, file->f_flags);
865 return PTR_ERR(inode);
867 file->f_path = *path;
868 return do_dentry_open(file, inode, NULL, cred);
871 struct file *dentry_open(const struct path *path, int flags,
872 const struct cred *cred)
877 validate_creds(cred);
879 /* We must always pass in a valid mount pointer. */
882 f = get_empty_filp();
885 error = vfs_open(path, f, cred);
887 /* from now on we need fput() to dispose of f */
888 error = open_check_o_direct(f);
900 EXPORT_SYMBOL(dentry_open);
902 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
904 int lookup_flags = 0;
908 * Clear out all open flags we don't know about so that we don't report
909 * them in fcntl(F_GETFD) or similar interfaces.
911 flags &= VALID_OPEN_FLAGS;
913 if (flags & (O_CREAT | __O_TMPFILE))
914 op->mode = (mode & S_IALLUGO) | S_IFREG;
918 /* Must never be set by userspace */
919 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
922 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
923 * check for O_DSYNC if the need any syncing at all we enforce it's
924 * always set instead of having to deal with possibly weird behaviour
925 * for malicious applications setting only __O_SYNC.
927 if (flags & __O_SYNC)
930 if (flags & __O_TMPFILE) {
931 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
933 acc_mode = MAY_OPEN | ACC_MODE(flags);
934 if (!(acc_mode & MAY_WRITE))
936 } else if (flags & O_PATH) {
938 * If we have O_PATH in the open flag. Then we
939 * cannot have anything other than the below set of flags
941 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
944 acc_mode = MAY_OPEN | ACC_MODE(flags);
947 op->open_flag = flags;
949 /* O_TRUNC implies we need access checks for write permissions */
951 acc_mode |= MAY_WRITE;
953 /* Allow the LSM permission hook to distinguish append
954 access from general write access. */
955 if (flags & O_APPEND)
956 acc_mode |= MAY_APPEND;
958 op->acc_mode = acc_mode;
960 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
962 if (flags & O_CREAT) {
963 op->intent |= LOOKUP_CREATE;
965 op->intent |= LOOKUP_EXCL;
968 if (flags & O_DIRECTORY)
969 lookup_flags |= LOOKUP_DIRECTORY;
970 if (!(flags & O_NOFOLLOW))
971 lookup_flags |= LOOKUP_FOLLOW;
972 op->lookup_flags = lookup_flags;
977 * file_open_name - open file and return file pointer
979 * @name: struct filename containing path to open
980 * @flags: open flags as per the open(2) second argument
981 * @mode: mode for the new file if O_CREAT is set, else ignored
983 * This is the helper to open a file from kernelspace if you really
984 * have to. But in generally you should not do this, so please move
985 * along, nothing to see here..
987 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
989 struct open_flags op;
990 int err = build_open_flags(flags, mode, &op);
991 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
995 * filp_open - open file and return file pointer
997 * @filename: path to open
998 * @flags: open flags as per the open(2) second argument
999 * @mode: mode for the new file if O_CREAT is set, else ignored
1001 * This is the helper to open a file from kernelspace if you really
1002 * have to. But in generally you should not do this, so please move
1003 * along, nothing to see here..
1005 struct file *filp_open(const char *filename, int flags, umode_t mode)
1007 struct filename *name = getname_kernel(filename);
1008 struct file *file = ERR_CAST(name);
1010 if (!IS_ERR(name)) {
1011 file = file_open_name(name, flags, mode);
1016 EXPORT_SYMBOL(filp_open);
1018 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1019 const char *filename, int flags, umode_t mode)
1021 struct open_flags op;
1022 int err = build_open_flags(flags, mode, &op);
1024 return ERR_PTR(err);
1025 return do_file_open_root(dentry, mnt, filename, &op);
1027 EXPORT_SYMBOL(file_open_root);
1029 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1031 struct open_flags op;
1032 int fd = build_open_flags(flags, mode, &op);
1033 struct filename *tmp;
1038 tmp = getname(filename);
1040 return PTR_ERR(tmp);
1042 fd = get_unused_fd_flags(flags);
1044 struct file *f = do_filp_open(dfd, tmp, &op);
1057 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1059 if (force_o_largefile())
1060 flags |= O_LARGEFILE;
1062 return do_sys_open(AT_FDCWD, filename, flags, mode);
1065 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1068 if (force_o_largefile())
1069 flags |= O_LARGEFILE;
1071 return do_sys_open(dfd, filename, flags, mode);
1077 * For backward compatibility? Maybe this should be moved
1078 * into arch/i386 instead?
1080 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1082 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1088 * "id" is the POSIX thread ID. We use the
1089 * files pointer for this..
1091 int filp_close(struct file *filp, fl_owner_t id)
1095 if (!file_count(filp)) {
1096 printk(KERN_ERR "VFS: Close: file count is 0\n");
1100 if (filp->f_op->flush)
1101 retval = filp->f_op->flush(filp, id);
1103 if (likely(!(filp->f_mode & FMODE_PATH))) {
1104 dnotify_flush(filp, id);
1105 locks_remove_posix(filp, id);
1111 EXPORT_SYMBOL(filp_close);
1114 * Careful here! We test whether the file pointer is NULL before
1115 * releasing the fd. This ensures that one clone task can't release
1116 * an fd while another clone is opening it.
1118 SYSCALL_DEFINE1(close, unsigned int, fd)
1120 int retval = __close_fd(current->files, fd);
1122 /* can't restart close syscall because file table entry was cleared */
1123 if (unlikely(retval == -ERESTARTSYS ||
1124 retval == -ERESTARTNOINTR ||
1125 retval == -ERESTARTNOHAND ||
1126 retval == -ERESTART_RESTARTBLOCK))
1131 EXPORT_SYMBOL(sys_close);
1134 * This routine simulates a hangup on the tty, to arrange that users
1135 * are given clean terminals at login time.
1137 SYSCALL_DEFINE0(vhangup)
1139 if (capable(CAP_SYS_TTY_CONFIG)) {
1147 * Called when an inode is about to be open.
1148 * We use this to disallow opening large files on 32bit systems if
1149 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1150 * on this flag in sys_open.
1152 int generic_file_open(struct inode * inode, struct file * filp)
1154 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1159 EXPORT_SYMBOL(generic_file_open);
1162 * This is used by subsystems that don't want seekable
1163 * file descriptors. The function is not supposed to ever fail, the only
1164 * reason it returns an 'int' and not 'void' is so that it can be plugged
1165 * directly into file_operations structure.
1167 int nonseekable_open(struct inode *inode, struct file *filp)
1169 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1173 EXPORT_SYMBOL(nonseekable_open);
1176 * stream_open is used by subsystems that want stream-like file descriptors.
1177 * Such file descriptors are not seekable and don't have notion of position
1178 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1179 * files, .read() and .write() can run simultaneously.
1181 * stream_open never fails and is marked to return int so that it could be
1182 * directly used as file_operations.open .
1184 int stream_open(struct inode *inode, struct file *filp)
1186 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1187 filp->f_mode |= FMODE_STREAM;
1191 EXPORT_SYMBOL(stream_open);