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;
366 old_cred = override_creds(override_cred);
368 res = user_path_at(dfd, filename, lookup_flags, &path);
372 inode = d_backing_inode(path.dentry);
374 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
376 * MAY_EXEC on regular files is denied if the fs is mounted
377 * with the "noexec" flag.
380 if (path_noexec(&path))
381 goto out_path_release;
384 res = inode_permission(inode, mode | MAY_ACCESS);
385 /* SuS v2 requires we report a read only fs too */
386 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
387 goto out_path_release;
389 * This is a rare case where using __mnt_is_readonly()
390 * is OK without a mnt_want/drop_write() pair. Since
391 * no actual write to the fs is performed here, we do
392 * not need to telegraph to that to anyone.
394 * By doing this, we accept that this access is
395 * inherently racy and know that the fs may change
396 * state before we even see this result.
398 if (__mnt_is_readonly(path.mnt))
403 if (retry_estale(res, lookup_flags)) {
404 lookup_flags |= LOOKUP_REVAL;
408 revert_creds(old_cred);
409 put_cred(override_cred);
413 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
415 return sys_faccessat(AT_FDCWD, filename, mode);
418 SYSCALL_DEFINE1(chdir, const char __user *, filename)
422 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
424 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
428 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
432 set_fs_pwd(current->fs, &path);
436 if (retry_estale(error, lookup_flags)) {
437 lookup_flags |= LOOKUP_REVAL;
444 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
446 struct fd f = fdget_raw(fd);
454 inode = file_inode(f.file);
457 if (!S_ISDIR(inode->i_mode))
460 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
462 set_fs_pwd(current->fs, &f.file->f_path);
469 SYSCALL_DEFINE1(chroot, const char __user *, filename)
473 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
475 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
479 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
484 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
486 error = security_path_chroot(&path);
490 set_fs_root(current->fs, &path);
494 if (retry_estale(error, lookup_flags)) {
495 lookup_flags |= LOOKUP_REVAL;
502 static int chmod_common(struct path *path, umode_t mode)
504 struct inode *inode = path->dentry->d_inode;
505 struct inode *delegated_inode = NULL;
506 struct iattr newattrs;
509 error = mnt_want_write(path->mnt);
513 mutex_lock(&inode->i_mutex);
514 error = security_path_chmod(path, mode);
517 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
518 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
519 error = notify_change(path->dentry, &newattrs, &delegated_inode);
521 mutex_unlock(&inode->i_mutex);
522 if (delegated_inode) {
523 error = break_deleg_wait(&delegated_inode);
527 mnt_drop_write(path->mnt);
531 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
533 struct fd f = fdget(fd);
538 err = chmod_common(&f.file->f_path, mode);
544 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
548 unsigned int lookup_flags = LOOKUP_FOLLOW;
550 error = user_path_at(dfd, filename, lookup_flags, &path);
552 error = chmod_common(&path, mode);
554 if (retry_estale(error, lookup_flags)) {
555 lookup_flags |= LOOKUP_REVAL;
562 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
564 return sys_fchmodat(AT_FDCWD, filename, mode);
567 static int chown_common(struct path *path, uid_t user, gid_t group)
569 struct inode *inode = path->dentry->d_inode;
570 struct inode *delegated_inode = NULL;
572 struct iattr newattrs;
576 uid = make_kuid(current_user_ns(), user);
577 gid = make_kgid(current_user_ns(), group);
580 newattrs.ia_valid = ATTR_CTIME;
581 if (user != (uid_t) -1) {
584 newattrs.ia_valid |= ATTR_UID;
585 newattrs.ia_uid = uid;
587 if (group != (gid_t) -1) {
590 newattrs.ia_valid |= ATTR_GID;
591 newattrs.ia_gid = gid;
593 if (!S_ISDIR(inode->i_mode))
595 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
596 mutex_lock(&inode->i_mutex);
597 error = security_path_chown(path, uid, gid);
599 error = notify_change(path->dentry, &newattrs, &delegated_inode);
600 mutex_unlock(&inode->i_mutex);
601 if (delegated_inode) {
602 error = break_deleg_wait(&delegated_inode);
609 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
610 gid_t, group, int, flag)
616 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
619 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
620 if (flag & AT_EMPTY_PATH)
621 lookup_flags |= LOOKUP_EMPTY;
623 error = user_path_at(dfd, filename, lookup_flags, &path);
626 error = mnt_want_write(path.mnt);
629 error = chown_common(&path, user, group);
630 mnt_drop_write(path.mnt);
633 if (retry_estale(error, lookup_flags)) {
634 lookup_flags |= LOOKUP_REVAL;
641 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
643 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
646 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
648 return sys_fchownat(AT_FDCWD, filename, user, group,
649 AT_SYMLINK_NOFOLLOW);
652 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
654 struct fd f = fdget(fd);
660 error = mnt_want_write_file(f.file);
664 error = chown_common(&f.file->f_path, user, group);
665 mnt_drop_write_file(f.file);
672 int open_check_o_direct(struct file *f)
674 /* NB: we're sure to have correct a_ops only after f_op->open */
675 if (f->f_flags & O_DIRECT) {
676 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
682 static int do_dentry_open(struct file *f,
684 int (*open)(struct inode *, struct file *),
685 const struct cred *cred)
687 static const struct file_operations empty_fops = {};
690 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
691 FMODE_PREAD | FMODE_PWRITE;
693 path_get(&f->f_path);
695 f->f_mapping = inode->i_mapping;
697 if (unlikely(f->f_flags & O_PATH)) {
698 f->f_mode = FMODE_PATH;
699 f->f_op = &empty_fops;
703 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
704 error = get_write_access(inode);
707 error = __mnt_want_write(f->f_path.mnt);
708 if (unlikely(error)) {
709 put_write_access(inode);
712 f->f_mode |= FMODE_WRITER;
715 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
716 if (S_ISREG(inode->i_mode))
717 f->f_mode |= FMODE_ATOMIC_POS;
719 f->f_op = fops_get(inode->i_fop);
720 if (unlikely(WARN_ON(!f->f_op))) {
725 error = security_file_open(f, cred);
729 error = break_lease(inode, f->f_flags);
734 open = f->f_op->open;
736 error = open(inode, f);
740 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
741 i_readcount_inc(inode);
742 if ((f->f_mode & FMODE_READ) &&
743 likely(f->f_op->read || f->f_op->read_iter))
744 f->f_mode |= FMODE_CAN_READ;
745 if ((f->f_mode & FMODE_WRITE) &&
746 likely(f->f_op->write || f->f_op->write_iter))
747 f->f_mode |= FMODE_CAN_WRITE;
749 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
751 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
757 if (f->f_mode & FMODE_WRITER) {
758 put_write_access(inode);
759 __mnt_drop_write(f->f_path.mnt);
762 path_put(&f->f_path);
763 f->f_path.mnt = NULL;
764 f->f_path.dentry = NULL;
770 * finish_open - finish opening a file
771 * @file: file pointer
772 * @dentry: pointer to dentry
773 * @open: open callback
774 * @opened: state of open
776 * This can be used to finish opening a file passed to i_op->atomic_open().
778 * If the open callback is set to NULL, then the standard f_op->open()
779 * filesystem callback is substituted.
781 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
782 * the return value of d_splice_alias(), then the caller needs to perform dput()
783 * on it after finish_open().
785 * On successful return @file is a fully instantiated open file. After this, if
786 * an error occurs in ->atomic_open(), it needs to clean up with fput().
788 * Returns zero on success or -errno if the open failed.
790 int finish_open(struct file *file, struct dentry *dentry,
791 int (*open)(struct inode *, struct file *),
795 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
797 file->f_path.dentry = dentry;
798 error = do_dentry_open(file, d_backing_inode(dentry), open,
801 *opened |= FILE_OPENED;
805 EXPORT_SYMBOL(finish_open);
808 * finish_no_open - finish ->atomic_open() without opening the file
810 * @file: file pointer
811 * @dentry: dentry or NULL (as returned from ->lookup())
813 * This can be used to set the result of a successful lookup in ->atomic_open().
815 * NB: unlike finish_open() this function does consume the dentry reference and
816 * the caller need not dput() it.
818 * Returns "1" which must be the return value of ->atomic_open() after having
819 * called this function.
821 int finish_no_open(struct file *file, struct dentry *dentry)
823 file->f_path.dentry = dentry;
826 EXPORT_SYMBOL(finish_no_open);
828 char *file_path(struct file *filp, char *buf, int buflen)
830 return d_path(&filp->f_path, buf, buflen);
832 EXPORT_SYMBOL(file_path);
835 * vfs_open - open the file at the given path
836 * @path: path to open
837 * @file: newly allocated file with f_flag initialized
838 * @cred: credentials to use
840 int vfs_open(const struct path *path, struct file *file,
841 const struct cred *cred)
843 struct dentry *dentry = path->dentry;
844 struct inode *inode = dentry->d_inode;
846 file->f_path = *path;
847 if (dentry->d_flags & DCACHE_OP_SELECT_INODE) {
848 inode = dentry->d_op->d_select_inode(dentry, file->f_flags);
850 return PTR_ERR(inode);
853 return do_dentry_open(file, inode, NULL, cred);
856 struct file *dentry_open(const struct path *path, int flags,
857 const struct cred *cred)
862 validate_creds(cred);
864 /* We must always pass in a valid mount pointer. */
867 f = get_empty_filp();
870 error = vfs_open(path, f, cred);
872 /* from now on we need fput() to dispose of f */
873 error = open_check_o_direct(f);
885 EXPORT_SYMBOL(dentry_open);
887 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
889 int lookup_flags = 0;
890 int acc_mode = ACC_MODE(flags);
892 if (flags & (O_CREAT | __O_TMPFILE))
893 op->mode = (mode & S_IALLUGO) | S_IFREG;
897 /* Must never be set by userspace */
898 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
901 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
902 * check for O_DSYNC if the need any syncing at all we enforce it's
903 * always set instead of having to deal with possibly weird behaviour
904 * for malicious applications setting only __O_SYNC.
906 if (flags & __O_SYNC)
909 if (flags & __O_TMPFILE) {
910 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
912 if (!(acc_mode & MAY_WRITE))
914 } else if (flags & O_PATH) {
916 * If we have O_PATH in the open flag. Then we
917 * cannot have anything other than the below set of flags
919 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
923 op->open_flag = flags;
925 /* O_TRUNC implies we need access checks for write permissions */
927 acc_mode |= MAY_WRITE;
929 /* Allow the LSM permission hook to distinguish append
930 access from general write access. */
931 if (flags & O_APPEND)
932 acc_mode |= MAY_APPEND;
934 op->acc_mode = acc_mode;
936 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
938 if (flags & O_CREAT) {
939 op->intent |= LOOKUP_CREATE;
941 op->intent |= LOOKUP_EXCL;
944 if (flags & O_DIRECTORY)
945 lookup_flags |= LOOKUP_DIRECTORY;
946 if (!(flags & O_NOFOLLOW))
947 lookup_flags |= LOOKUP_FOLLOW;
948 op->lookup_flags = lookup_flags;
953 * file_open_name - open file and return file pointer
955 * @name: struct filename containing path to open
956 * @flags: open flags as per the open(2) second argument
957 * @mode: mode for the new file if O_CREAT is set, else ignored
959 * This is the helper to open a file from kernelspace if you really
960 * have to. But in generally you should not do this, so please move
961 * along, nothing to see here..
963 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
965 struct open_flags op;
966 int err = build_open_flags(flags, mode, &op);
967 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
971 * filp_open - open file and return file pointer
973 * @filename: 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 *filp_open(const char *filename, int flags, umode_t mode)
983 struct filename *name = getname_kernel(filename);
984 struct file *file = ERR_CAST(name);
987 file = file_open_name(name, flags, mode);
992 EXPORT_SYMBOL(filp_open);
994 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
995 const char *filename, int flags)
997 struct open_flags op;
998 int err = build_open_flags(flags, 0, &op);
1000 return ERR_PTR(err);
1001 if (flags & O_CREAT)
1002 return ERR_PTR(-EINVAL);
1003 return do_file_open_root(dentry, mnt, filename, &op);
1005 EXPORT_SYMBOL(file_open_root);
1007 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1009 struct open_flags op;
1010 int fd = build_open_flags(flags, mode, &op);
1011 struct filename *tmp;
1016 tmp = getname(filename);
1018 return PTR_ERR(tmp);
1020 fd = get_unused_fd_flags(flags);
1022 struct file *f = do_filp_open(dfd, tmp, &op);
1035 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1037 if (force_o_largefile())
1038 flags |= O_LARGEFILE;
1040 return do_sys_open(AT_FDCWD, filename, flags, mode);
1043 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1046 if (force_o_largefile())
1047 flags |= O_LARGEFILE;
1049 return do_sys_open(dfd, filename, flags, mode);
1055 * For backward compatibility? Maybe this should be moved
1056 * into arch/i386 instead?
1058 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1060 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1066 * "id" is the POSIX thread ID. We use the
1067 * files pointer for this..
1069 int filp_close(struct file *filp, fl_owner_t id)
1073 if (!file_count(filp)) {
1074 printk(KERN_ERR "VFS: Close: file count is 0\n");
1078 if (filp->f_op->flush)
1079 retval = filp->f_op->flush(filp, id);
1081 if (likely(!(filp->f_mode & FMODE_PATH))) {
1082 dnotify_flush(filp, id);
1083 locks_remove_posix(filp, id);
1089 EXPORT_SYMBOL(filp_close);
1092 * Careful here! We test whether the file pointer is NULL before
1093 * releasing the fd. This ensures that one clone task can't release
1094 * an fd while another clone is opening it.
1096 SYSCALL_DEFINE1(close, unsigned int, fd)
1098 int retval = __close_fd(current->files, fd);
1100 /* can't restart close syscall because file table entry was cleared */
1101 if (unlikely(retval == -ERESTARTSYS ||
1102 retval == -ERESTARTNOINTR ||
1103 retval == -ERESTARTNOHAND ||
1104 retval == -ERESTART_RESTARTBLOCK))
1109 EXPORT_SYMBOL(sys_close);
1112 * This routine simulates a hangup on the tty, to arrange that users
1113 * are given clean terminals at login time.
1115 SYSCALL_DEFINE0(vhangup)
1117 if (capable(CAP_SYS_TTY_CONFIG)) {
1125 * Called when an inode is about to be open.
1126 * We use this to disallow opening large files on 32bit systems if
1127 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1128 * on this flag in sys_open.
1130 int generic_file_open(struct inode * inode, struct file * filp)
1132 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1137 EXPORT_SYMBOL(generic_file_open);
1140 * This is used by subsystems that don't want seekable
1141 * file descriptors. The function is not supposed to ever fail, the only
1142 * reason it returns an 'int' and not 'void' is so that it can be plugged
1143 * directly into file_operations structure.
1145 int nonseekable_open(struct inode *inode, struct file *filp)
1147 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1151 EXPORT_SYMBOL(nonseekable_open);