// +build cmount // +build cgo // +build linux darwin freebsd windows package cmount import ( "io" "os" "path" "sync" "time" "github.com/billziss-gh/cgofuse/fuse" "github.com/pkg/errors" "github.com/rclone/rclone/cmd/mountlib" "github.com/rclone/rclone/fs" "github.com/rclone/rclone/fs/log" "github.com/rclone/rclone/vfs" "github.com/rclone/rclone/vfs/vfsflags" ) const fhUnset = ^uint64(0) // FS represents the top level filing system type FS struct { VFS *vfs.VFS f fs.Fs ready chan (struct{}) mu sync.Mutex // to protect the below handles []vfs.Handle } // NewFS makes a new FS func NewFS(f fs.Fs) *FS { fsys := &FS{ VFS: vfs.New(f, &vfsflags.Opt), f: f, ready: make(chan (struct{})), } return fsys } // Open a handle returning an integer file handle func (fsys *FS) openHandle(handle vfs.Handle) (fh uint64) { fsys.mu.Lock() defer fsys.mu.Unlock() var i int var oldHandle vfs.Handle for i, oldHandle = range fsys.handles { if oldHandle == nil { fsys.handles[i] = handle goto found } } fsys.handles = append(fsys.handles, handle) i = len(fsys.handles) - 1 found: return uint64(i) } // get the handle for fh, call with the lock held func (fsys *FS) _getHandle(fh uint64) (i int, handle vfs.Handle, errc int) { if fh > uint64(len(fsys.handles)) { fs.Debugf(nil, "Bad file handle: too big: 0x%X", fh) return i, nil, -fuse.EBADF } i = int(fh) handle = fsys.handles[i] if handle == nil { fs.Debugf(nil, "Bad file handle: nil handle: 0x%X", fh) return i, nil, -fuse.EBADF } return i, handle, 0 } // Get the handle for the file handle func (fsys *FS) getHandle(fh uint64) (handle vfs.Handle, errc int) { fsys.mu.Lock() _, handle, errc = fsys._getHandle(fh) fsys.mu.Unlock() return } // Close the handle func (fsys *FS) closeHandle(fh uint64) (errc int) { fsys.mu.Lock() i, _, errc := fsys._getHandle(fh) if errc == 0 { fsys.handles[i] = nil } fsys.mu.Unlock() return } // lookup a Node given a path func (fsys *FS) lookupNode(path string) (node vfs.Node, errc int) { node, err := fsys.VFS.Stat(path) return node, translateError(err) } // lookup a Dir given a path func (fsys *FS) lookupDir(path string) (dir *vfs.Dir, errc int) { node, errc := fsys.lookupNode(path) if errc != 0 { return nil, errc } dir, ok := node.(*vfs.Dir) if !ok { return nil, -fuse.ENOTDIR } return dir, 0 } // lookup a parent Dir given a path returning the dir and the leaf func (fsys *FS) lookupParentDir(filePath string) (leaf string, dir *vfs.Dir, errc int) { parentDir, leaf := path.Split(filePath) dir, errc = fsys.lookupDir(parentDir) return leaf, dir, errc } // lookup a File given a path func (fsys *FS) lookupFile(path string) (file *vfs.File, errc int) { node, errc := fsys.lookupNode(path) if errc != 0 { return nil, errc } file, ok := node.(*vfs.File) if !ok { return nil, -fuse.EISDIR } return file, 0 } // get a node and handle from the path or from the fh if not fhUnset // // handle may be nil func (fsys *FS) getNode(path string, fh uint64) (node vfs.Node, handle vfs.Handle, errc int) { if fh == fhUnset { node, errc = fsys.lookupNode(path) } else { handle, errc = fsys.getHandle(fh) if errc == 0 { node = handle.Node() } } return } // stat fills up the stat block for Node func (fsys *FS) stat(node vfs.Node, stat *fuse.Stat_t) (errc int) { Size := uint64(node.Size()) Blocks := (Size + 511) / 512 modTime := node.ModTime() Mode := node.Mode().Perm() if node.IsDir() { Mode |= fuse.S_IFDIR } else { Mode |= fuse.S_IFREG } //stat.Dev = 1 stat.Ino = node.Inode() // FIXME do we need to set the inode number? stat.Mode = uint32(Mode) stat.Nlink = 1 stat.Uid = fsys.VFS.Opt.UID stat.Gid = fsys.VFS.Opt.GID //stat.Rdev stat.Size = int64(Size) t := fuse.NewTimespec(modTime) stat.Atim = t stat.Mtim = t stat.Ctim = t stat.Blksize = 512 stat.Blocks = int64(Blocks) stat.Birthtim = t // fs.Debugf(nil, "stat = %+v", *stat) return 0 } // Init is called after the filesystem is ready func (fsys *FS) Init() { defer log.Trace(fsys.f, "")("") close(fsys.ready) } // Destroy is called when it is unmounted (note that depending on how // the file system is terminated the file system may not receive the // Destroy call). func (fsys *FS) Destroy() { defer log.Trace(fsys.f, "")("") } // Getattr reads the attributes for path func (fsys *FS) Getattr(path string, stat *fuse.Stat_t, fh uint64) (errc int) { defer log.Trace(path, "fh=0x%X", fh)("errc=%v", &errc) node, _, errc := fsys.getNode(path, fh) if errc == 0 { errc = fsys.stat(node, stat) } return } // Opendir opens path as a directory func (fsys *FS) Opendir(path string) (errc int, fh uint64) { defer log.Trace(path, "")("errc=%d, fh=0x%X", &errc, &fh) handle, err := fsys.VFS.OpenFile(path, os.O_RDONLY, 0777) if err != nil { return translateError(err), fhUnset } return 0, fsys.openHandle(handle) } // Readdir reads the directory at dirPath func (fsys *FS) Readdir(dirPath string, fill func(name string, stat *fuse.Stat_t, ofst int64) bool, ofst int64, fh uint64) (errc int) { itemsRead := -1 defer log.Trace(dirPath, "ofst=%d, fh=0x%X", ofst, fh)("items=%d, errc=%d", &itemsRead, &errc) node, errc := fsys.getHandle(fh) if errc != 0 { return errc } items, err := node.Readdir(-1) if err != nil { return translateError(err) } // Optionally, create a struct stat that describes the file as // for getattr (but FUSE only looks at st_ino and the // file-type bits of st_mode). // // FIXME If you call host.SetCapReaddirPlus() then WinFsp will // use the full stat information - a Useful optimization on // Windows. // // NB we are using the first mode for readdir: The readdir // implementation ignores the offset parameter, and passes // zero to the filler function's offset. The filler function // will not return '1' (unless an error happens), so the whole // directory is read in a single readdir operation. fill(".", nil, 0) fill("..", nil, 0) for _, item := range items { node, ok := item.(vfs.Node) if ok { name := node.Name() if len(name) > mountlib.MaxLeafSize { fs.Errorf(dirPath, "Name too long (%d bytes) for FUSE, skipping: %s", len(name), name) continue } fill(name, nil, 0) } } itemsRead = len(items) return 0 } // Releasedir finished reading the directory func (fsys *FS) Releasedir(path string, fh uint64) (errc int) { defer log.Trace(path, "fh=0x%X", fh)("errc=%d", &errc) return fsys.closeHandle(fh) } // Statfs reads overall stats on the filessystem func (fsys *FS) Statfs(path string, stat *fuse.Statfs_t) (errc int) { defer log.Trace(path, "")("stat=%+v, errc=%d", stat, &errc) const blockSize = 4096 const fsBlocks = (1 << 50) / blockSize stat.Blocks = fsBlocks // Total data blocks in file system. stat.Bfree = fsBlocks // Free blocks in file system. stat.Bavail = fsBlocks // Free blocks in file system if you're not root. stat.Files = 1e9 // Total files in file system. stat.Ffree = 1e9 // Free files in file system. stat.Bsize = blockSize // Block size stat.Namemax = 255 // Maximum file name length? stat.Frsize = blockSize // Fragment size, smallest addressable data size in the file system. total, used, free := fsys.VFS.Statfs() if total >= 0 { stat.Blocks = uint64(total) / blockSize } if used >= 0 { stat.Bfree = stat.Blocks - uint64(used)/blockSize } if free >= 0 { stat.Bavail = uint64(free) / blockSize } mountlib.ClipBlocks(&stat.Blocks) mountlib.ClipBlocks(&stat.Bfree) mountlib.ClipBlocks(&stat.Bavail) return 0 } // Open opens a file func (fsys *FS) Open(path string, flags int) (errc int, fh uint64) { defer log.Trace(path, "flags=0x%X", flags)("errc=%d, fh=0x%X", &errc, &fh) // translate the fuse flags to os flags flags = translateOpenFlags(flags) handle, err := fsys.VFS.OpenFile(path, flags, 0777) if err != nil { return translateError(err), fhUnset } // FIXME add support for unknown length files setting direct_io // See: https://github.com/billziss-gh/cgofuse/issues/38 return 0, fsys.openHandle(handle) } // Create creates and opens a file. func (fsys *FS) Create(filePath string, flags int, mode uint32) (errc int, fh uint64) { defer log.Trace(filePath, "flags=0x%X, mode=0%o", flags, mode)("errc=%d, fh=0x%X", &errc, &fh) leaf, parentDir, errc := fsys.lookupParentDir(filePath) if errc != 0 { return errc, fhUnset } file, err := parentDir.Create(leaf, flags) if err != nil { return translateError(err), fhUnset } // translate the fuse flags to os flags flags = translateOpenFlags(flags) | os.O_CREATE handle, err := file.Open(flags) if err != nil { return translateError(err), fhUnset } return 0, fsys.openHandle(handle) } // Truncate truncates a file to size func (fsys *FS) Truncate(path string, size int64, fh uint64) (errc int) { defer log.Trace(path, "size=%d, fh=0x%X", size, fh)("errc=%d", &errc) node, handle, errc := fsys.getNode(path, fh) if errc != 0 { return errc } var err error if handle != nil { err = handle.Truncate(size) } else { err = node.Truncate(size) } if err != nil { return translateError(err) } return 0 } // Read data from file handle func (fsys *FS) Read(path string, buff []byte, ofst int64, fh uint64) (n int) { defer log.Trace(path, "ofst=%d, fh=0x%X", ofst, fh)("n=%d", &n) handle, errc := fsys.getHandle(fh) if errc != 0 { return errc } n, err := handle.ReadAt(buff, ofst) if err == io.EOF { } else if err != nil { return translateError(err) } return n } // Write data to file handle func (fsys *FS) Write(path string, buff []byte, ofst int64, fh uint64) (n int) { defer log.Trace(path, "ofst=%d, fh=0x%X", ofst, fh)("n=%d", &n) handle, errc := fsys.getHandle(fh) if errc != 0 { return errc } n, err := handle.WriteAt(buff, ofst) if err != nil { return translateError(err) } return n } // Flush flushes an open file descriptor or path func (fsys *FS) Flush(path string, fh uint64) (errc int) { defer log.Trace(path, "fh=0x%X", fh)("errc=%d", &errc) handle, errc := fsys.getHandle(fh) if errc != 0 { return errc } return translateError(handle.Flush()) } // Release closes the file if still open func (fsys *FS) Release(path string, fh uint64) (errc int) { defer log.Trace(path, "fh=0x%X", fh)("errc=%d", &errc) handle, errc := fsys.getHandle(fh) if errc != 0 { return errc } _ = fsys.closeHandle(fh) return translateError(handle.Release()) } // Unlink removes a file. func (fsys *FS) Unlink(filePath string) (errc int) { defer log.Trace(filePath, "")("errc=%d", &errc) leaf, parentDir, errc := fsys.lookupParentDir(filePath) if errc != 0 { return errc } return translateError(parentDir.RemoveName(leaf)) } // Mkdir creates a directory. func (fsys *FS) Mkdir(dirPath string, mode uint32) (errc int) { defer log.Trace(dirPath, "mode=0%o", mode)("errc=%d", &errc) leaf, parentDir, errc := fsys.lookupParentDir(dirPath) if errc != 0 { return errc } _, err := parentDir.Mkdir(leaf) return translateError(err) } // Rmdir removes a directory func (fsys *FS) Rmdir(dirPath string) (errc int) { defer log.Trace(dirPath, "")("errc=%d", &errc) leaf, parentDir, errc := fsys.lookupParentDir(dirPath) if errc != 0 { return errc } return translateError(parentDir.RemoveName(leaf)) } // Rename renames a file. func (fsys *FS) Rename(oldPath string, newPath string) (errc int) { defer log.Trace(oldPath, "newPath=%q", newPath)("errc=%d", &errc) return translateError(fsys.VFS.Rename(oldPath, newPath)) } // Utimens changes the access and modification times of a file. func (fsys *FS) Utimens(path string, tmsp []fuse.Timespec) (errc int) { defer log.Trace(path, "tmsp=%+v", tmsp)("errc=%d", &errc) node, errc := fsys.lookupNode(path) if errc != 0 { return errc } var t time.Time if tmsp == nil || len(tmsp) < 2 { t = time.Now() } else { t = tmsp[1].Time() } return translateError(node.SetModTime(t)) } // Mknod creates a file node. func (fsys *FS) Mknod(path string, mode uint32, dev uint64) (errc int) { defer log.Trace(path, "mode=0x%X, dev=0x%X", mode, dev)("errc=%d", &errc) return -fuse.ENOSYS } // Fsync synchronizes file contents. func (fsys *FS) Fsync(path string, datasync bool, fh uint64) (errc int) { defer log.Trace(path, "datasync=%v, fh=0x%X", datasync, fh)("errc=%d", &errc) // This is a no-op for rclone return 0 } // Link creates a hard link to a file. func (fsys *FS) Link(oldpath string, newpath string) (errc int) { defer log.Trace(oldpath, "newpath=%q", newpath)("errc=%d", &errc) return -fuse.ENOSYS } // Symlink creates a symbolic link. func (fsys *FS) Symlink(target string, newpath string) (errc int) { defer log.Trace(target, "newpath=%q", newpath)("errc=%d", &errc) return -fuse.ENOSYS } // Readlink reads the target of a symbolic link. func (fsys *FS) Readlink(path string) (errc int, linkPath string) { defer log.Trace(path, "")("linkPath=%q, errc=%d", &linkPath, &errc) return -fuse.ENOSYS, "" } // Chmod changes the permission bits of a file. func (fsys *FS) Chmod(path string, mode uint32) (errc int) { defer log.Trace(path, "mode=0%o", mode)("errc=%d", &errc) // This is a no-op for rclone return 0 } // Chown changes the owner and group of a file. func (fsys *FS) Chown(path string, uid uint32, gid uint32) (errc int) { defer log.Trace(path, "uid=%d, gid=%d", uid, gid)("errc=%d", &errc) // This is a no-op for rclone return 0 } // Access checks file access permissions. func (fsys *FS) Access(path string, mask uint32) (errc int) { defer log.Trace(path, "mask=0%o", mask)("errc=%d", &errc) // This is a no-op for rclone return 0 } // Fsyncdir synchronizes directory contents. func (fsys *FS) Fsyncdir(path string, datasync bool, fh uint64) (errc int) { defer log.Trace(path, "datasync=%v, fh=0x%X", datasync, fh)("errc=%d", &errc) // This is a no-op for rclone return 0 } // Setxattr sets extended attributes. func (fsys *FS) Setxattr(path string, name string, value []byte, flags int) (errc int) { return -fuse.ENOSYS } // Getxattr gets extended attributes. func (fsys *FS) Getxattr(path string, name string) (errc int, value []byte) { return -fuse.ENOSYS, nil } // Removexattr removes extended attributes. func (fsys *FS) Removexattr(path string, name string) (errc int) { return -fuse.ENOSYS } // Listxattr lists extended attributes. func (fsys *FS) Listxattr(path string, fill func(name string) bool) (errc int) { return -fuse.ENOSYS } // Translate errors from mountlib func translateError(err error) (errc int) { if err == nil { return 0 } switch errors.Cause(err) { case vfs.OK: return 0 case vfs.ENOENT: return -fuse.ENOENT case vfs.EEXIST: return -fuse.EEXIST case vfs.EPERM: return -fuse.EPERM case vfs.ECLOSED: return -fuse.EBADF case vfs.ENOTEMPTY: return -fuse.ENOTEMPTY case vfs.ESPIPE: return -fuse.ESPIPE case vfs.EBADF: return -fuse.EBADF case vfs.EROFS: return -fuse.EROFS case vfs.ENOSYS: return -fuse.ENOSYS case vfs.EINVAL: return -fuse.EINVAL } fs.Errorf(nil, "IO error: %v", err) return -fuse.EIO } // Translate Open Flags from FUSE to os (as used in the vfs layer) func translateOpenFlags(inFlags int) (outFlags int) { switch inFlags & fuse.O_ACCMODE { case fuse.O_RDONLY: outFlags = os.O_RDONLY case fuse.O_WRONLY: outFlags = os.O_WRONLY case fuse.O_RDWR: outFlags = os.O_RDWR } if inFlags&fuse.O_APPEND != 0 { outFlags |= os.O_APPEND } if inFlags&fuse.O_CREAT != 0 { outFlags |= os.O_CREATE } if inFlags&fuse.O_EXCL != 0 { outFlags |= os.O_EXCL } if inFlags&fuse.O_TRUNC != 0 { outFlags |= os.O_TRUNC } // NB O_SYNC isn't defined by fuse return outFlags }