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/quotaops.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/tty.h>
16 #include <linux/namei.h>
17 #include <linux/backing-dev.h>
18 #include <linux/capability.h>
19 #include <linux/security.h>
20 #include <linux/mount.h>
21 #include <linux/vfs.h>
22 #include <linux/fcntl.h>
23 #include <asm/uaccess.h>
25 #include <linux/personality.h>
26 #include <linux/pagemap.h>
27 #include <linux/syscalls.h>
28 #include <linux/rcupdate.h>
29 #include <linux/audit.h>
30 #include <linux/falloc.h>
32 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
38 if (dentry->d_sb->s_op->statfs) {
39 memset(buf, 0, sizeof(*buf));
40 retval = security_sb_statfs(dentry);
43 retval = dentry->d_sb->s_op->statfs(dentry, buf);
44 if (retval == 0 && buf->f_frsize == 0)
45 buf->f_frsize = buf->f_bsize;
51 EXPORT_SYMBOL(vfs_statfs);
53 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
58 retval = vfs_statfs(dentry, &st);
62 if (sizeof(*buf) == sizeof(st))
63 memcpy(buf, &st, sizeof(st));
65 if (sizeof buf->f_blocks == 4) {
66 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
67 0xffffffff00000000ULL)
70 * f_files and f_ffree may be -1; it's okay to stuff
73 if (st.f_files != -1 &&
74 (st.f_files & 0xffffffff00000000ULL))
76 if (st.f_ffree != -1 &&
77 (st.f_ffree & 0xffffffff00000000ULL))
81 buf->f_type = st.f_type;
82 buf->f_bsize = st.f_bsize;
83 buf->f_blocks = st.f_blocks;
84 buf->f_bfree = st.f_bfree;
85 buf->f_bavail = st.f_bavail;
86 buf->f_files = st.f_files;
87 buf->f_ffree = st.f_ffree;
88 buf->f_fsid = st.f_fsid;
89 buf->f_namelen = st.f_namelen;
90 buf->f_frsize = st.f_frsize;
91 memset(buf->f_spare, 0, sizeof(buf->f_spare));
96 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
101 retval = vfs_statfs(dentry, &st);
105 if (sizeof(*buf) == sizeof(st))
106 memcpy(buf, &st, sizeof(st));
108 buf->f_type = st.f_type;
109 buf->f_bsize = st.f_bsize;
110 buf->f_blocks = st.f_blocks;
111 buf->f_bfree = st.f_bfree;
112 buf->f_bavail = st.f_bavail;
113 buf->f_files = st.f_files;
114 buf->f_ffree = st.f_ffree;
115 buf->f_fsid = st.f_fsid;
116 buf->f_namelen = st.f_namelen;
117 buf->f_frsize = st.f_frsize;
118 memset(buf->f_spare, 0, sizeof(buf->f_spare));
123 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
128 error = user_path_walk(path, &nd);
131 error = vfs_statfs_native(nd.path.dentry, &tmp);
132 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
140 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
145 if (sz != sizeof(*buf))
147 error = user_path_walk(path, &nd);
150 error = vfs_statfs64(nd.path.dentry, &tmp);
151 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
159 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
169 error = vfs_statfs_native(file->f_path.dentry, &tmp);
170 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
177 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
183 if (sz != sizeof(*buf))
190 error = vfs_statfs64(file->f_path.dentry, &tmp);
191 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
198 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
202 struct iattr newattrs;
204 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
208 newattrs.ia_size = length;
209 newattrs.ia_valid = ATTR_SIZE | time_attrs;
211 newattrs.ia_file = filp;
212 newattrs.ia_valid |= ATTR_FILE;
215 /* Remove suid/sgid on truncate too */
216 newattrs.ia_valid |= should_remove_suid(dentry);
218 mutex_lock(&dentry->d_inode->i_mutex);
219 err = notify_change(dentry, &newattrs);
220 mutex_unlock(&dentry->d_inode->i_mutex);
224 static long do_sys_truncate(const char __user * path, loff_t length)
227 struct inode * inode;
231 if (length < 0) /* sorry, but loff_t says... */
234 error = user_path_walk(path, &nd);
237 inode = nd.path.dentry->d_inode;
239 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
241 if (S_ISDIR(inode->i_mode))
245 if (!S_ISREG(inode->i_mode))
248 error = mnt_want_write(nd.path.mnt);
252 error = vfs_permission(&nd, MAY_WRITE);
254 goto mnt_drop_write_and_out;
257 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
258 goto mnt_drop_write_and_out;
260 error = get_write_access(inode);
262 goto mnt_drop_write_and_out;
265 * Make sure that there are no leases. get_write_access() protects
266 * against the truncate racing with a lease-granting setlease().
268 error = break_lease(inode, FMODE_WRITE);
270 goto put_write_and_out;
272 error = locks_verify_truncate(inode, NULL, length);
275 error = do_truncate(nd.path.dentry, length, 0, NULL);
279 put_write_access(inode);
280 mnt_drop_write_and_out:
281 mnt_drop_write(nd.path.mnt);
288 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
290 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
291 return do_sys_truncate(path, (long)length);
294 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
296 struct inode * inode;
297 struct dentry *dentry;
309 /* explicitly opened as large or we are on 64-bit box */
310 if (file->f_flags & O_LARGEFILE)
313 dentry = file->f_path.dentry;
314 inode = dentry->d_inode;
316 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
320 /* Cannot ftruncate over 2^31 bytes without large file support */
321 if (small && length > MAX_NON_LFS)
325 if (IS_APPEND(inode))
328 error = locks_verify_truncate(inode, file, length);
330 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
337 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
339 long ret = do_sys_ftruncate(fd, length, 1);
340 /* avoid REGPARM breakage on x86: */
341 asmlinkage_protect(2, ret, fd, length);
345 /* LFS versions of truncate are only needed on 32 bit machines */
346 #if BITS_PER_LONG == 32
347 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
349 return do_sys_truncate(path, length);
352 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
354 long ret = do_sys_ftruncate(fd, length, 0);
355 /* avoid REGPARM breakage on x86: */
356 asmlinkage_protect(2, ret, fd, length);
361 asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
367 if (offset < 0 || len <= 0)
370 /* Return error if mode is not supported */
372 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
379 if (!(file->f_mode & FMODE_WRITE))
382 * Revalidate the write permissions, in case security policy has
383 * changed since the files were opened.
385 ret = security_file_permission(file, MAY_WRITE);
389 inode = file->f_path.dentry->d_inode;
392 if (S_ISFIFO(inode->i_mode))
397 * Let individual file system decide if it supports preallocation
398 * for directories or not.
400 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
404 /* Check for wrap through zero too */
405 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
408 if (inode->i_op && inode->i_op->fallocate)
409 ret = inode->i_op->fallocate(inode, mode, offset, len);
420 * access() needs to use the real uid/gid, not the effective uid/gid.
421 * We do this by temporarily clearing all FS-related capabilities and
422 * switching the fsuid/fsgid around to the real ones.
424 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
427 int old_fsuid, old_fsgid;
428 kernel_cap_t old_cap;
431 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
434 old_fsuid = current->fsuid;
435 old_fsgid = current->fsgid;
436 old_cap = current->cap_effective;
438 current->fsuid = current->uid;
439 current->fsgid = current->gid;
442 * Clear the capabilities if we switch to a non-root user
444 * FIXME: There is a race here against sys_capset. The
445 * capabilities can change yet we will restore the old
446 * value below. We should hold task_capabilities_lock,
447 * but we cannot because user_path_walk can sleep.
450 cap_clear(current->cap_effective);
452 current->cap_effective = current->cap_permitted;
454 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
458 res = vfs_permission(&nd, mode);
459 /* SuS v2 requires we report a read only fs too */
460 if(res || !(mode & S_IWOTH) ||
461 special_file(nd.path.dentry->d_inode->i_mode))
462 goto out_path_release;
464 * This is a rare case where using __mnt_is_readonly()
465 * is OK without a mnt_want/drop_write() pair. Since
466 * no actual write to the fs is performed here, we do
467 * not need to telegraph to that to anyone.
469 * By doing this, we accept that this access is
470 * inherently racy and know that the fs may change
471 * state before we even see this result.
473 if (__mnt_is_readonly(nd.path.mnt))
479 current->fsuid = old_fsuid;
480 current->fsgid = old_fsgid;
481 current->cap_effective = old_cap;
486 asmlinkage long sys_access(const char __user *filename, int mode)
488 return sys_faccessat(AT_FDCWD, filename, mode);
491 asmlinkage long sys_chdir(const char __user * filename)
496 error = __user_walk(filename,
497 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
501 error = vfs_permission(&nd, MAY_EXEC);
505 set_fs_pwd(current->fs, &nd.path);
513 asmlinkage long sys_fchdir(unsigned int fd)
524 inode = file->f_path.dentry->d_inode;
527 if (!S_ISDIR(inode->i_mode))
530 error = file_permission(file, MAY_EXEC);
532 set_fs_pwd(current->fs, &file->f_path);
539 asmlinkage long sys_chroot(const char __user * filename)
544 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
548 error = vfs_permission(&nd, MAY_EXEC);
553 if (!capable(CAP_SYS_CHROOT))
556 set_fs_root(current->fs, &nd.path);
565 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
567 struct inode * inode;
568 struct dentry * dentry;
571 struct iattr newattrs;
577 dentry = file->f_path.dentry;
578 inode = dentry->d_inode;
580 audit_inode(NULL, dentry);
582 err = mnt_want_write(file->f_path.mnt);
586 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
588 mutex_lock(&inode->i_mutex);
589 if (mode == (mode_t) -1)
590 mode = inode->i_mode;
591 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
592 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
593 err = notify_change(dentry, &newattrs);
594 mutex_unlock(&inode->i_mutex);
597 mnt_drop_write(file->f_path.mnt);
604 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
608 struct inode * inode;
610 struct iattr newattrs;
612 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
615 inode = nd.path.dentry->d_inode;
617 error = mnt_want_write(nd.path.mnt);
622 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
625 mutex_lock(&inode->i_mutex);
626 if (mode == (mode_t) -1)
627 mode = inode->i_mode;
628 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
629 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
630 error = notify_change(nd.path.dentry, &newattrs);
631 mutex_unlock(&inode->i_mutex);
634 mnt_drop_write(nd.path.mnt);
641 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
643 return sys_fchmodat(AT_FDCWD, filename, mode);
646 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
648 struct inode * inode;
650 struct iattr newattrs;
653 if (!(inode = dentry->d_inode)) {
654 printk(KERN_ERR "chown_common: NULL inode\n");
658 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
660 newattrs.ia_valid = ATTR_CTIME;
661 if (user != (uid_t) -1) {
662 newattrs.ia_valid |= ATTR_UID;
663 newattrs.ia_uid = user;
665 if (group != (gid_t) -1) {
666 newattrs.ia_valid |= ATTR_GID;
667 newattrs.ia_gid = group;
669 if (!S_ISDIR(inode->i_mode))
671 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
672 mutex_lock(&inode->i_mutex);
673 error = notify_change(dentry, &newattrs);
674 mutex_unlock(&inode->i_mutex);
679 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
684 error = user_path_walk(filename, &nd);
687 error = mnt_want_write(nd.path.mnt);
690 error = chown_common(nd.path.dentry, user, group);
691 mnt_drop_write(nd.path.mnt);
698 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
699 gid_t group, int flag)
705 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
708 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
709 error = __user_walk_fd(dfd, filename, follow, &nd);
712 error = mnt_want_write(nd.path.mnt);
715 error = chown_common(nd.path.dentry, user, group);
716 mnt_drop_write(nd.path.mnt);
723 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
728 error = user_path_walk_link(filename, &nd);
731 error = mnt_want_write(nd.path.mnt);
734 error = chown_common(nd.path.dentry, user, group);
735 mnt_drop_write(nd.path.mnt);
743 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
747 struct dentry * dentry;
753 error = mnt_want_write(file->f_path.mnt);
756 dentry = file->f_path.dentry;
757 audit_inode(NULL, dentry);
758 error = chown_common(dentry, user, group);
759 mnt_drop_write(file->f_path.mnt);
767 * You have to be very careful that these write
768 * counts get cleaned up in error cases and
769 * upon __fput(). This should probably never
770 * be called outside of __dentry_open().
772 static inline int __get_file_write_access(struct inode *inode,
773 struct vfsmount *mnt)
776 error = get_write_access(inode);
780 * Do not take mount writer counts on
781 * special files since no writes to
782 * the mount itself will occur.
784 if (!special_file(inode->i_mode)) {
786 * Balanced in __fput()
788 error = mnt_want_write(mnt);
790 put_write_access(inode);
795 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
796 int flags, struct file *f,
797 int (*open)(struct inode *, struct file *))
803 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
804 FMODE_PREAD | FMODE_PWRITE;
805 inode = dentry->d_inode;
806 if (f->f_mode & FMODE_WRITE) {
807 error = __get_file_write_access(inode, mnt);
810 if (!special_file(inode->i_mode))
814 f->f_mapping = inode->i_mapping;
815 f->f_path.dentry = dentry;
818 f->f_op = fops_get(inode->i_fop);
819 file_move(f, &inode->i_sb->s_files);
821 error = security_dentry_open(f);
825 if (!open && f->f_op)
826 open = f->f_op->open;
828 error = open(inode, f);
833 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
835 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
837 /* NB: we're sure to have correct a_ops only after f_op->open */
838 if (f->f_flags & O_DIRECT) {
839 if (!f->f_mapping->a_ops ||
840 ((!f->f_mapping->a_ops->direct_IO) &&
841 (!f->f_mapping->a_ops->get_xip_mem))) {
843 f = ERR_PTR(-EINVAL);
851 if (f->f_mode & FMODE_WRITE) {
852 put_write_access(inode);
853 if (!special_file(inode->i_mode)) {
855 * We don't consider this a real
856 * mnt_want/drop_write() pair
857 * because it all happenend right
858 * here, so just reset the state.
865 f->f_path.dentry = NULL;
866 f->f_path.mnt = NULL;
871 return ERR_PTR(error);
875 * lookup_instantiate_filp - instantiates the open intent filp
876 * @nd: pointer to nameidata
877 * @dentry: pointer to dentry
878 * @open: open callback
880 * Helper for filesystems that want to use lookup open intents and pass back
881 * a fully instantiated struct file to the caller.
882 * This function is meant to be called from within a filesystem's
884 * Beware of calling it for non-regular files! Those ->open methods might block
885 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
886 * leading to a deadlock, as nobody can open that fifo anymore, because
887 * another process to open fifo will block on locked parent when doing lookup).
888 * Note that in case of error, nd->intent.open.file is destroyed, but the
889 * path information remains valid.
890 * If the open callback is set to NULL, then the standard f_op->open()
891 * filesystem callback is substituted.
893 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
894 int (*open)(struct inode *, struct file *))
896 if (IS_ERR(nd->intent.open.file))
900 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
901 nd->intent.open.flags - 1,
902 nd->intent.open.file,
905 return nd->intent.open.file;
907 release_open_intent(nd);
908 nd->intent.open.file = (struct file *)dentry;
911 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
914 * nameidata_to_filp - convert a nameidata to an open filp.
915 * @nd: pointer to nameidata
918 * Note that this function destroys the original nameidata
920 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
924 /* Pick up the filp from the open intent */
925 filp = nd->intent.open.file;
926 /* Has the filesystem initialised the file for us? */
927 if (filp->f_path.dentry == NULL)
928 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
936 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
939 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
945 * We must always pass in a valid mount pointer. Historically
946 * callers got away with not passing it, but we must enforce this at
947 * the earliest possible point now to avoid strange problems deep in the
951 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
953 return ERR_PTR(-EINVAL);
957 f = get_empty_filp();
961 return ERR_PTR(error);
964 return __dentry_open(dentry, mnt, flags, f, NULL);
966 EXPORT_SYMBOL(dentry_open);
969 * Find an empty file descriptor entry, and mark it busy.
971 int get_unused_fd_flags(int flags)
973 struct files_struct * files = current->files;
978 spin_lock(&files->file_lock);
981 fdt = files_fdtable(files);
982 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
986 * N.B. For clone tasks sharing a files structure, this test
987 * will limit the total number of files that can be opened.
989 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
992 /* Do we need to expand the fd array or fd set? */
993 error = expand_files(files, fd);
999 * If we needed to expand the fs array we
1000 * might have blocked - try again.
1006 FD_SET(fd, fdt->open_fds);
1007 if (flags & O_CLOEXEC)
1008 FD_SET(fd, fdt->close_on_exec);
1010 FD_CLR(fd, fdt->close_on_exec);
1011 files->next_fd = fd + 1;
1014 if (fdt->fd[fd] != NULL) {
1015 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1022 spin_unlock(&files->file_lock);
1026 int get_unused_fd(void)
1028 return get_unused_fd_flags(0);
1031 EXPORT_SYMBOL(get_unused_fd);
1033 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1035 struct fdtable *fdt = files_fdtable(files);
1036 __FD_CLR(fd, fdt->open_fds);
1037 if (fd < files->next_fd)
1038 files->next_fd = fd;
1041 void put_unused_fd(unsigned int fd)
1043 struct files_struct *files = current->files;
1044 spin_lock(&files->file_lock);
1045 __put_unused_fd(files, fd);
1046 spin_unlock(&files->file_lock);
1049 EXPORT_SYMBOL(put_unused_fd);
1052 * Install a file pointer in the fd array.
1054 * The VFS is full of places where we drop the files lock between
1055 * setting the open_fds bitmap and installing the file in the file
1056 * array. At any such point, we are vulnerable to a dup2() race
1057 * installing a file in the array before us. We need to detect this and
1058 * fput() the struct file we are about to overwrite in this case.
1060 * It should never happen - if we allow dup2() do it, _really_ bad things
1064 void fd_install(unsigned int fd, struct file *file)
1066 struct files_struct *files = current->files;
1067 struct fdtable *fdt;
1068 spin_lock(&files->file_lock);
1069 fdt = files_fdtable(files);
1070 BUG_ON(fdt->fd[fd] != NULL);
1071 rcu_assign_pointer(fdt->fd[fd], file);
1072 spin_unlock(&files->file_lock);
1075 EXPORT_SYMBOL(fd_install);
1077 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1079 char *tmp = getname(filename);
1080 int fd = PTR_ERR(tmp);
1083 fd = get_unused_fd_flags(flags);
1085 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1090 fsnotify_open(f->f_path.dentry);
1099 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1103 if (force_o_largefile())
1104 flags |= O_LARGEFILE;
1106 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1107 /* avoid REGPARM breakage on x86: */
1108 asmlinkage_protect(3, ret, filename, flags, mode);
1112 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1117 if (force_o_largefile())
1118 flags |= O_LARGEFILE;
1120 ret = do_sys_open(dfd, filename, flags, mode);
1121 /* avoid REGPARM breakage on x86: */
1122 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1129 * For backward compatibility? Maybe this should be moved
1130 * into arch/i386 instead?
1132 asmlinkage long sys_creat(const char __user * pathname, int mode)
1134 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1140 * "id" is the POSIX thread ID. We use the
1141 * files pointer for this..
1143 int filp_close(struct file *filp, fl_owner_t id)
1147 if (!file_count(filp)) {
1148 printk(KERN_ERR "VFS: Close: file count is 0\n");
1152 if (filp->f_op && filp->f_op->flush)
1153 retval = filp->f_op->flush(filp, id);
1155 dnotify_flush(filp, id);
1156 locks_remove_posix(filp, id);
1161 EXPORT_SYMBOL(filp_close);
1164 * Careful here! We test whether the file pointer is NULL before
1165 * releasing the fd. This ensures that one clone task can't release
1166 * an fd while another clone is opening it.
1168 asmlinkage long sys_close(unsigned int fd)
1171 struct files_struct *files = current->files;
1172 struct fdtable *fdt;
1175 spin_lock(&files->file_lock);
1176 fdt = files_fdtable(files);
1177 if (fd >= fdt->max_fds)
1182 rcu_assign_pointer(fdt->fd[fd], NULL);
1183 FD_CLR(fd, fdt->close_on_exec);
1184 __put_unused_fd(files, fd);
1185 spin_unlock(&files->file_lock);
1186 retval = filp_close(filp, files);
1188 /* can't restart close syscall because file table entry was cleared */
1189 if (unlikely(retval == -ERESTARTSYS ||
1190 retval == -ERESTARTNOINTR ||
1191 retval == -ERESTARTNOHAND ||
1192 retval == -ERESTART_RESTARTBLOCK))
1198 spin_unlock(&files->file_lock);
1202 EXPORT_SYMBOL(sys_close);
1205 * This routine simulates a hangup on the tty, to arrange that users
1206 * are given clean terminals at login time.
1208 asmlinkage long sys_vhangup(void)
1210 if (capable(CAP_SYS_TTY_CONFIG)) {
1211 /* XXX: this needs locking */
1212 tty_vhangup(current->signal->tty);
1219 * Called when an inode is about to be open.
1220 * We use this to disallow opening large files on 32bit systems if
1221 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1222 * on this flag in sys_open.
1224 int generic_file_open(struct inode * inode, struct file * filp)
1226 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1231 EXPORT_SYMBOL(generic_file_open);
1234 * This is used by subsystems that don't want seekable
1237 int nonseekable_open(struct inode *inode, struct file *filp)
1239 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1243 EXPORT_SYMBOL(nonseekable_open);