4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/smp_lock.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>
31 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
37 if (dentry->d_sb->s_op->statfs) {
38 memset(buf, 0, sizeof(*buf));
39 retval = security_sb_statfs(dentry);
42 retval = dentry->d_sb->s_op->statfs(dentry, buf);
43 if (retval == 0 && buf->f_frsize == 0)
44 buf->f_frsize = buf->f_bsize;
50 EXPORT_SYMBOL(vfs_statfs);
52 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
57 retval = vfs_statfs(dentry, &st);
61 if (sizeof(*buf) == sizeof(st))
62 memcpy(buf, &st, sizeof(st));
64 if (sizeof buf->f_blocks == 4) {
65 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
66 0xffffffff00000000ULL)
69 * f_files and f_ffree may be -1; it's okay to stuff
72 if (st.f_files != -1 &&
73 (st.f_files & 0xffffffff00000000ULL))
75 if (st.f_ffree != -1 &&
76 (st.f_ffree & 0xffffffff00000000ULL))
80 buf->f_type = st.f_type;
81 buf->f_bsize = st.f_bsize;
82 buf->f_blocks = st.f_blocks;
83 buf->f_bfree = st.f_bfree;
84 buf->f_bavail = st.f_bavail;
85 buf->f_files = st.f_files;
86 buf->f_ffree = st.f_ffree;
87 buf->f_fsid = st.f_fsid;
88 buf->f_namelen = st.f_namelen;
89 buf->f_frsize = st.f_frsize;
90 memset(buf->f_spare, 0, sizeof(buf->f_spare));
95 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
100 retval = vfs_statfs(dentry, &st);
104 if (sizeof(*buf) == sizeof(st))
105 memcpy(buf, &st, sizeof(st));
107 buf->f_type = st.f_type;
108 buf->f_bsize = st.f_bsize;
109 buf->f_blocks = st.f_blocks;
110 buf->f_bfree = st.f_bfree;
111 buf->f_bavail = st.f_bavail;
112 buf->f_files = st.f_files;
113 buf->f_ffree = st.f_ffree;
114 buf->f_fsid = st.f_fsid;
115 buf->f_namelen = st.f_namelen;
116 buf->f_frsize = st.f_frsize;
117 memset(buf->f_spare, 0, sizeof(buf->f_spare));
122 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
127 error = user_path_walk(path, &nd);
130 error = vfs_statfs_native(nd.dentry, &tmp);
131 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
139 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
144 if (sz != sizeof(*buf))
146 error = user_path_walk(path, &nd);
149 error = vfs_statfs64(nd.dentry, &tmp);
150 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
158 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
168 error = vfs_statfs_native(file->f_dentry, &tmp);
169 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
176 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
182 if (sz != sizeof(*buf))
189 error = vfs_statfs64(file->f_dentry, &tmp);
190 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
197 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
201 struct iattr newattrs;
203 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
207 newattrs.ia_size = length;
208 newattrs.ia_valid = ATTR_SIZE | time_attrs;
210 newattrs.ia_file = filp;
211 newattrs.ia_valid |= ATTR_FILE;
214 mutex_lock(&dentry->d_inode->i_mutex);
215 err = notify_change(dentry, &newattrs);
216 mutex_unlock(&dentry->d_inode->i_mutex);
220 static long do_sys_truncate(const char __user * path, loff_t length)
223 struct inode * inode;
227 if (length < 0) /* sorry, but loff_t says... */
230 error = user_path_walk(path, &nd);
233 inode = nd.dentry->d_inode;
235 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
237 if (S_ISDIR(inode->i_mode))
241 if (!S_ISREG(inode->i_mode))
244 error = vfs_permission(&nd, MAY_WRITE);
249 if (IS_RDONLY(inode))
253 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
257 * Make sure that there are no leases.
259 error = break_lease(inode, FMODE_WRITE);
263 error = get_write_access(inode);
267 error = locks_verify_truncate(inode, NULL, length);
270 error = do_truncate(nd.dentry, length, 0, NULL);
272 put_write_access(inode);
280 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
282 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
283 return do_sys_truncate(path, (long)length);
286 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
288 struct inode * inode;
289 struct dentry *dentry;
301 /* explicitly opened as large or we are on 64-bit box */
302 if (file->f_flags & O_LARGEFILE)
305 dentry = file->f_dentry;
306 inode = dentry->d_inode;
308 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
312 /* Cannot ftruncate over 2^31 bytes without large file support */
313 if (small && length > MAX_NON_LFS)
317 if (IS_APPEND(inode))
320 error = locks_verify_truncate(inode, file, length);
322 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
329 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
331 long ret = do_sys_ftruncate(fd, length, 1);
332 /* avoid REGPARM breakage on x86: */
333 prevent_tail_call(ret);
337 /* LFS versions of truncate are only needed on 32 bit machines */
338 #if BITS_PER_LONG == 32
339 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
341 return do_sys_truncate(path, length);
344 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
346 long ret = do_sys_ftruncate(fd, length, 0);
347 /* avoid REGPARM breakage on x86: */
348 prevent_tail_call(ret);
354 * access() needs to use the real uid/gid, not the effective uid/gid.
355 * We do this by temporarily clearing all FS-related capabilities and
356 * switching the fsuid/fsgid around to the real ones.
358 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
361 int old_fsuid, old_fsgid;
362 kernel_cap_t old_cap;
365 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
368 old_fsuid = current->fsuid;
369 old_fsgid = current->fsgid;
370 old_cap = current->cap_effective;
372 current->fsuid = current->uid;
373 current->fsgid = current->gid;
376 * Clear the capabilities if we switch to a non-root user
378 * FIXME: There is a race here against sys_capset. The
379 * capabilities can change yet we will restore the old
380 * value below. We should hold task_capabilities_lock,
381 * but we cannot because user_path_walk can sleep.
384 cap_clear(current->cap_effective);
386 current->cap_effective = current->cap_permitted;
388 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
390 res = vfs_permission(&nd, mode);
391 /* SuS v2 requires we report a read only fs too */
392 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
393 && !special_file(nd.dentry->d_inode->i_mode))
398 current->fsuid = old_fsuid;
399 current->fsgid = old_fsgid;
400 current->cap_effective = old_cap;
405 asmlinkage long sys_access(const char __user *filename, int mode)
407 return sys_faccessat(AT_FDCWD, filename, mode);
410 asmlinkage long sys_chdir(const char __user * filename)
415 error = __user_walk(filename,
416 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
420 error = vfs_permission(&nd, MAY_EXEC);
424 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
432 asmlinkage long sys_fchdir(unsigned int fd)
435 struct dentry *dentry;
437 struct vfsmount *mnt;
445 dentry = file->f_dentry;
446 mnt = file->f_vfsmnt;
447 inode = dentry->d_inode;
450 if (!S_ISDIR(inode->i_mode))
453 error = file_permission(file, MAY_EXEC);
455 set_fs_pwd(current->fs, mnt, dentry);
462 asmlinkage long sys_chroot(const char __user * filename)
467 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
471 error = vfs_permission(&nd, MAY_EXEC);
476 if (!capable(CAP_SYS_CHROOT))
479 set_fs_root(current->fs, nd.mnt, nd.dentry);
488 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
490 struct inode * inode;
491 struct dentry * dentry;
494 struct iattr newattrs;
500 dentry = file->f_dentry;
501 inode = dentry->d_inode;
503 audit_inode(NULL, inode);
506 if (IS_RDONLY(inode))
509 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
511 mutex_lock(&inode->i_mutex);
512 if (mode == (mode_t) -1)
513 mode = inode->i_mode;
514 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
515 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
516 err = notify_change(dentry, &newattrs);
517 mutex_unlock(&inode->i_mutex);
525 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
529 struct inode * inode;
531 struct iattr newattrs;
533 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
536 inode = nd.dentry->d_inode;
539 if (IS_RDONLY(inode))
543 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
546 mutex_lock(&inode->i_mutex);
547 if (mode == (mode_t) -1)
548 mode = inode->i_mode;
549 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
550 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
551 error = notify_change(nd.dentry, &newattrs);
552 mutex_unlock(&inode->i_mutex);
560 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
562 return sys_fchmodat(AT_FDCWD, filename, mode);
565 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
567 struct inode * inode;
569 struct iattr newattrs;
572 if (!(inode = dentry->d_inode)) {
573 printk(KERN_ERR "chown_common: NULL inode\n");
577 if (IS_RDONLY(inode))
580 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
582 newattrs.ia_valid = ATTR_CTIME;
583 if (user != (uid_t) -1) {
584 newattrs.ia_valid |= ATTR_UID;
585 newattrs.ia_uid = user;
587 if (group != (gid_t) -1) {
588 newattrs.ia_valid |= ATTR_GID;
589 newattrs.ia_gid = group;
591 if (!S_ISDIR(inode->i_mode))
592 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
593 mutex_lock(&inode->i_mutex);
594 error = notify_change(dentry, &newattrs);
595 mutex_unlock(&inode->i_mutex);
600 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
605 error = user_path_walk(filename, &nd);
607 error = chown_common(nd.dentry, user, group);
613 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
614 gid_t group, int flag)
620 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
623 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
624 error = __user_walk_fd(dfd, filename, follow, &nd);
626 error = chown_common(nd.dentry, user, group);
633 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
638 error = user_path_walk_link(filename, &nd);
640 error = chown_common(nd.dentry, user, group);
647 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
654 struct dentry * dentry;
655 dentry = file->f_dentry;
656 audit_inode(NULL, dentry->d_inode);
657 error = chown_common(dentry, user, group);
663 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
664 int flags, struct file *f,
665 int (*open)(struct inode *, struct file *))
671 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
672 FMODE_PREAD | FMODE_PWRITE;
673 inode = dentry->d_inode;
674 if (f->f_mode & FMODE_WRITE) {
675 error = get_write_access(inode);
680 f->f_mapping = inode->i_mapping;
681 f->f_dentry = dentry;
684 f->f_op = fops_get(inode->i_fop);
685 file_move(f, &inode->i_sb->s_files);
687 if (!open && f->f_op)
688 open = f->f_op->open;
690 error = open(inode, f);
695 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
697 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
699 /* NB: we're sure to have correct a_ops only after f_op->open */
700 if (f->f_flags & O_DIRECT) {
701 if (!f->f_mapping->a_ops ||
702 ((!f->f_mapping->a_ops->direct_IO) &&
703 (!f->f_mapping->a_ops->get_xip_page))) {
705 f = ERR_PTR(-EINVAL);
713 if (f->f_mode & FMODE_WRITE)
714 put_write_access(inode);
722 return ERR_PTR(error);
726 * Note that while the flag value (low two bits) for sys_open means:
732 * 00 - no permissions needed
733 * 01 - read-permission
734 * 10 - write-permission
736 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
739 static struct file *do_filp_open(int dfd, const char *filename, int flags,
742 int namei_flags, error;
746 if ((namei_flags+1) & O_ACCMODE)
749 error = open_namei(dfd, filename, namei_flags, mode, &nd);
751 return nameidata_to_filp(&nd, flags);
753 return ERR_PTR(error);
756 struct file *filp_open(const char *filename, int flags, int mode)
758 return do_filp_open(AT_FDCWD, filename, flags, mode);
760 EXPORT_SYMBOL(filp_open);
763 * lookup_instantiate_filp - instantiates the open intent filp
764 * @nd: pointer to nameidata
765 * @dentry: pointer to dentry
766 * @open: open callback
768 * Helper for filesystems that want to use lookup open intents and pass back
769 * a fully instantiated struct file to the caller.
770 * This function is meant to be called from within a filesystem's
772 * Beware of calling it for non-regular files! Those ->open methods might block
773 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
774 * leading to a deadlock, as nobody can open that fifo anymore, because
775 * another process to open fifo will block on locked parent when doing lookup).
776 * Note that in case of error, nd->intent.open.file is destroyed, but the
777 * path information remains valid.
778 * If the open callback is set to NULL, then the standard f_op->open()
779 * filesystem callback is substituted.
781 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
782 int (*open)(struct inode *, struct file *))
784 if (IS_ERR(nd->intent.open.file))
788 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
789 nd->intent.open.flags - 1,
790 nd->intent.open.file,
793 return nd->intent.open.file;
795 release_open_intent(nd);
796 nd->intent.open.file = (struct file *)dentry;
799 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
802 * nameidata_to_filp - convert a nameidata to an open filp.
803 * @nd: pointer to nameidata
806 * Note that this function destroys the original nameidata
808 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
812 /* Pick up the filp from the open intent */
813 filp = nd->intent.open.file;
814 /* Has the filesystem initialised the file for us? */
815 if (filp->f_dentry == NULL)
816 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
823 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
826 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
832 f = get_empty_filp();
836 return ERR_PTR(error);
839 return __dentry_open(dentry, mnt, flags, f, NULL);
841 EXPORT_SYMBOL(dentry_open);
844 * Find an empty file descriptor entry, and mark it busy.
846 int get_unused_fd(void)
848 struct files_struct * files = current->files;
853 spin_lock(&files->file_lock);
856 fdt = files_fdtable(files);
857 fd = find_next_zero_bit(fdt->open_fds->fds_bits,
862 * N.B. For clone tasks sharing a files structure, this test
863 * will limit the total number of files that can be opened.
865 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
868 /* Do we need to expand the fd array or fd set? */
869 error = expand_files(files, fd);
875 * If we needed to expand the fs array we
876 * might have blocked - try again.
882 FD_SET(fd, fdt->open_fds);
883 FD_CLR(fd, fdt->close_on_exec);
884 files->next_fd = fd + 1;
887 if (fdt->fd[fd] != NULL) {
888 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
895 spin_unlock(&files->file_lock);
899 EXPORT_SYMBOL(get_unused_fd);
901 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
903 struct fdtable *fdt = files_fdtable(files);
904 __FD_CLR(fd, fdt->open_fds);
905 if (fd < files->next_fd)
909 void fastcall put_unused_fd(unsigned int fd)
911 struct files_struct *files = current->files;
912 spin_lock(&files->file_lock);
913 __put_unused_fd(files, fd);
914 spin_unlock(&files->file_lock);
917 EXPORT_SYMBOL(put_unused_fd);
920 * Install a file pointer in the fd array.
922 * The VFS is full of places where we drop the files lock between
923 * setting the open_fds bitmap and installing the file in the file
924 * array. At any such point, we are vulnerable to a dup2() race
925 * installing a file in the array before us. We need to detect this and
926 * fput() the struct file we are about to overwrite in this case.
928 * It should never happen - if we allow dup2() do it, _really_ bad things
932 void fastcall fd_install(unsigned int fd, struct file * file)
934 struct files_struct *files = current->files;
936 spin_lock(&files->file_lock);
937 fdt = files_fdtable(files);
938 BUG_ON(fdt->fd[fd] != NULL);
939 rcu_assign_pointer(fdt->fd[fd], file);
940 spin_unlock(&files->file_lock);
943 EXPORT_SYMBOL(fd_install);
945 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
947 char *tmp = getname(filename);
948 int fd = PTR_ERR(tmp);
951 fd = get_unused_fd();
953 struct file *f = do_filp_open(dfd, tmp, flags, mode);
958 fsnotify_open(f->f_dentry);
967 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
971 if (force_o_largefile())
972 flags |= O_LARGEFILE;
974 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
975 /* avoid REGPARM breakage on x86: */
976 prevent_tail_call(ret);
979 EXPORT_SYMBOL_GPL(sys_open);
981 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
986 if (force_o_largefile())
987 flags |= O_LARGEFILE;
989 ret = do_sys_open(dfd, filename, flags, mode);
990 /* avoid REGPARM breakage on x86: */
991 prevent_tail_call(ret);
998 * For backward compatibility? Maybe this should be moved
999 * into arch/i386 instead?
1001 asmlinkage long sys_creat(const char __user * pathname, int mode)
1003 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1009 * "id" is the POSIX thread ID. We use the
1010 * files pointer for this..
1012 int filp_close(struct file *filp, fl_owner_t id)
1016 if (!file_count(filp)) {
1017 printk(KERN_ERR "VFS: Close: file count is 0\n");
1021 if (filp->f_op && filp->f_op->flush)
1022 retval = filp->f_op->flush(filp, id);
1024 dnotify_flush(filp, id);
1025 locks_remove_posix(filp, id);
1030 EXPORT_SYMBOL(filp_close);
1033 * Careful here! We test whether the file pointer is NULL before
1034 * releasing the fd. This ensures that one clone task can't release
1035 * an fd while another clone is opening it.
1037 asmlinkage long sys_close(unsigned int fd)
1040 struct files_struct *files = current->files;
1041 struct fdtable *fdt;
1044 spin_lock(&files->file_lock);
1045 fdt = files_fdtable(files);
1046 if (fd >= fdt->max_fds)
1051 rcu_assign_pointer(fdt->fd[fd], NULL);
1052 FD_CLR(fd, fdt->close_on_exec);
1053 __put_unused_fd(files, fd);
1054 spin_unlock(&files->file_lock);
1055 retval = filp_close(filp, files);
1057 /* can't restart close syscall because file table entry was cleared */
1058 if (unlikely(retval == -ERESTARTSYS ||
1059 retval == -ERESTARTNOINTR ||
1060 retval == -ERESTARTNOHAND ||
1061 retval == -ERESTART_RESTARTBLOCK))
1067 spin_unlock(&files->file_lock);
1071 EXPORT_SYMBOL(sys_close);
1074 * This routine simulates a hangup on the tty, to arrange that users
1075 * are given clean terminals at login time.
1077 asmlinkage long sys_vhangup(void)
1079 if (capable(CAP_SYS_TTY_CONFIG)) {
1080 tty_vhangup(current->signal->tty);
1087 * Called when an inode is about to be open.
1088 * We use this to disallow opening large files on 32bit systems if
1089 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1090 * on this flag in sys_open.
1092 int generic_file_open(struct inode * inode, struct file * filp)
1094 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1099 EXPORT_SYMBOL(generic_file_open);
1102 * This is used by subsystems that don't want seekable
1105 int nonseekable_open(struct inode *inode, struct file *filp)
1107 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1111 EXPORT_SYMBOL(nonseekable_open);