// waitRemove blocks until either:
// a) the device registered at <device_set_prefix>-<hash> is removed,
// or b) the 10 second timeout expires.
func (devices *DeviceSet) waitRemove(devname string) error {
log.Debugf("[deviceset %s] waitRemove(%s)", devices.devicePrefix, devname)
defer log.Debugf("[deviceset %s] waitRemove(%s) END", devices.devicePrefix, devname)
i := 0
for ; i < 1000; i += 1 {
devinfo, err := getInfo(devname)
if err != nil {
// If there is an error we assume the device doesn't exist.
// The error might actually be something else, but we can't differentiate.
return nil
}
if i%100 == 0 {
log.Debugf("Waiting for removal of %s: exists=%d", devname, devinfo.Exists)
}
if devinfo.Exists == 0 {
break
}
devices.Unlock()
time.Sleep(10 * time.Millisecond)
devices.Lock()
}
if i == 1000 {
return fmt.Errorf("Timeout while waiting for device %s to be removed", devname)
}
return nil
}
func (devices *DeviceSet) AddDevice(hash, baseHash string) error {
baseInfo, err := devices.lookupDevice(baseHash)
if err != nil {
return err
}
baseInfo.lock.Lock()
defer baseInfo.lock.Unlock()
devices.Lock()
defer devices.Unlock()
if info, _ := devices.lookupDevice(hash); info != nil {
return fmt.Errorf("device %s already exists", hash)
}
deviceId := devices.nextDeviceId
if err := createSnapDevice(devices.getPoolDevName(), &deviceId, baseInfo.Name(), baseInfo.DeviceId); err != nil {
log.Debugf("Error creating snap device: %s", err)
return err
}
// Ids are 24bit, so wrap around
devices.nextDeviceId = (deviceId + 1) & 0xffffff
if _, err := devices.registerDevice(deviceId, hash, baseInfo.Size); err != nil {
deleteDevice(devices.getPoolDevName(), deviceId)
log.Debugf("Error registering device: %s", err)
return err
}
return nil
}
func (devices *DeviceSet) deactivatePool() error {
log.Debugf("[devmapper] deactivatePool()")
defer log.Debugf("[devmapper] deactivatePool END")
devname := devices.getPoolDevName()
devinfo, err := getInfo(devname)
if err != nil {
return err
}
if devinfo.Exists != 0 {
return removeDevice(devname)
}
return nil
}
func (devices *DeviceSet) setupBaseImage() error {
oldInfo, _ := devices.lookupDevice("")
if oldInfo != nil && oldInfo.Initialized {
return nil
}
if oldInfo != nil && !oldInfo.Initialized {
log.Debugf("Removing uninitialized base image")
if err := devices.deleteDevice(oldInfo); err != nil {
return err
}
}
log.Debugf("Initializing base device-manager snapshot")
id := devices.nextDeviceId
// Create initial device
if err := createDevice(devices.getPoolDevName(), &id); err != nil {
return err
}
// Ids are 24bit, so wrap around
devices.nextDeviceId = (id + 1) & 0xffffff
log.Debugf("Registering base device (id %v) with FS size %v", id, devices.baseFsSize)
info, err := devices.registerDevice(id, "", devices.baseFsSize)
if err != nil {
_ = deleteDevice(devices.getPoolDevName(), id)
return err
}
log.Debugf("Creating filesystem on base device-manager snapshot")
if err = devices.activateDeviceIfNeeded(info); err != nil {
return err
}
if err := devices.createFilesystem(info); err != nil {
return err
}
info.Initialized = true
if err = devices.saveMetadata(info); err != nil {
info.Initialized = false
return err
}
return nil
}
// CopyWithTar creates a tar archive of filesystem path `src`, and
// unpacks it at filesystem path `dst`.
// The archive is streamed directly with fixed buffering and no
// intermediary disk IO.
//
func CopyWithTar(src, dst string) error {
srcSt, err := os.Stat(src)
if err != nil {
return err
}
if !srcSt.IsDir() {
return CopyFileWithTar(src, dst)
}
// Create dst, copy src's content into it
log.Debugf("Creating dest directory: %s", dst)
if err := os.MkdirAll(dst, 0755); err != nil && !os.IsExist(err) {
return err
}
log.Debugf("Calling TarUntar(%s, %s)", src, dst)
return TarUntar(src, dst)
}
func removeDevice(name string) error {
log.Debugf("[devmapper] removeDevice START")
defer log.Debugf("[devmapper] removeDevice END")
task, err := createTask(DeviceRemove, name)
if task == nil {
return err
}
dmSawBusy = false
if err = task.Run(); err != nil {
if dmSawBusy {
return ErrBusy
}
return fmt.Errorf("Error running removeDevice %s", err)
}
return nil
}
func createDevice(poolName string, deviceId *int) error {
log.Debugf("[devmapper] createDevice(poolName=%v, deviceId=%v)", poolName, *deviceId)
for {
task, err := createTask(DeviceTargetMsg, poolName)
if task == nil {
return err
}
if err := task.SetSector(0); err != nil {
return fmt.Errorf("Can't set sector %s", err)
}
if err := task.SetMessage(fmt.Sprintf("create_thin %d", *deviceId)); err != nil {
return fmt.Errorf("Can't set message %s", err)
}
dmSawExist = false
if err := task.Run(); err != nil {
if dmSawExist {
// Already exists, try next id
*deviceId++
continue
}
return fmt.Errorf("Error running createDevice %s", err)
}
break
}
return nil
}
func SetDevDir(dir string) error {
if res := DmSetDevDir(dir); res != 1 {
log.Debugf("Error dm_set_dev_dir")
return ErrSetDevDir
}
return nil
}
func UdevWait(cookie uint) error {
if res := DmUdevWait(cookie); res != 1 {
log.Debugf("Failed to wait on udev cookie %d", cookie)
return ErrUdevWait
}
return nil
}
func (devices *DeviceSet) registerDevice(id int, hash string, size uint64) (*DevInfo, error) {
log.Debugf("registerDevice(%v, %v)", id, hash)
info := &DevInfo{
Hash: hash,
DeviceId: id,
Size: size,
TransactionId: devices.allocateTransactionId(),
Initialized: false,
devices: devices,
}
devices.devicesLock.Lock()
devices.Devices[hash] = info
devices.devicesLock.Unlock()
if err := devices.saveMetadata(info); err != nil {
// Try to remove unused device
devices.devicesLock.Lock()
delete(devices.Devices, hash)
devices.devicesLock.Unlock()
return nil, err
}
return info, nil
}
// NewRequest() creates a new *http.Request,
// applies all decorators in the HTTPRequestFactory on the request,
// then applies decorators provided by d on the request.
func (h *HTTPRequestFactory) NewRequest(method, urlStr string, body io.Reader, d ...HTTPRequestDecorator) (*http.Request, error) {
req, err := http.NewRequest(method, urlStr, body)
if err != nil {
return nil, err
}
// By default, a nil factory should work.
if h == nil {
return req, nil
}
for _, dec := range h.decorators {
req, err = dec.ChangeRequest(req)
if err != nil {
return nil, err
}
}
for _, dec := range d {
req, err = dec.ChangeRequest(req)
if err != nil {
return nil, err
}
}
log.Debugf("%v -- HEADERS: %v", req.URL, req.Header)
return req, err
}
// ensureImage creates a sparse file of <size> bytes at the path
// <root>/devicemapper/<name>.
// If the file already exists, it does nothing.
// Either way it returns the full path.
func (devices *DeviceSet) ensureImage(name string, size int64) (string, error) {
dirname := devices.loopbackDir()
filename := path.Join(dirname, name)
if err := os.MkdirAll(dirname, 0700); err != nil && !os.IsExist(err) {
return "", err
}
if _, err := os.Stat(filename); err != nil {
if !os.IsNotExist(err) {
return "", err
}
log.Debugf("Creating loopback file %s for device-manage use", filename)
file, err := os.OpenFile(filename, os.O_RDWR|os.O_CREATE, 0600)
if err != nil {
return "", err
}
defer file.Close()
if err = file.Truncate(size); err != nil {
return "", err
}
}
return filename, nil
}
func (devices *DeviceSet) log(level int, file string, line int, dmError int, message string) {
if level >= 7 {
return // Ignore _LOG_DEBUG
}
log.Debugf("libdevmapper(%d): %s:%d (%d) %s", level, file, line, dmError, message)
}
func createBridgeIface(name string) error {
kv, err := kernel.GetKernelVersion()
// only set the bridge's mac address if the kernel version is > 3.3
// before that it was not supported
setBridgeMacAddr := err == nil && (kv.Kernel >= 3 && kv.Major >= 3)
log.Debugf("setting bridge mac address = %v", setBridgeMacAddr)
return netlink.CreateBridge(name, setBridgeMacAddr)
}
func (devices *DeviceSet) activateDeviceIfNeeded(info *DevInfo) error {
log.Debugf("activateDeviceIfNeeded(%v)", info.Hash)
if devinfo, _ := getInfo(info.Name()); devinfo != nil && devinfo.Exists != 0 {
return nil
}
return activateDevice(devices.getPoolDevName(), info.Name(), info.DeviceId, info.Size)
}
// TarUntar is a convenience function which calls Tar and Untar, with
// the output of one piped into the other. If either Tar or Untar fails,
// TarUntar aborts and returns the error.
func TarUntar(src string, dst string) error {
log.Debugf("TarUntar(%s %s)", src, dst)
archive, err := TarWithOptions(src, &TarOptions{Compression: Uncompressed})
if err != nil {
return err
}
defer archive.Close()
return Untar(archive, dst, nil)
}
func ExportChanges(dir string, changes []Change) (Archive, error) {
reader, writer := io.Pipe()
tw := tar.NewWriter(writer)
go func() {
twBuf := bufio.NewWriterSize(nil, twBufSize)
// In general we log errors here but ignore them because
// during e.g. a diff operation the container can continue
// mutating the filesystem and we can see transient errors
// from this
for _, change := range changes {
if change.Kind == ChangeDelete {
whiteOutDir := filepath.Dir(change.Path)
whiteOutBase := filepath.Base(change.Path)
whiteOut := filepath.Join(whiteOutDir, ".wh."+whiteOutBase)
timestamp := time.Now()
hdr := &tar.Header{
Name: whiteOut[1:],
Size: 0,
ModTime: timestamp,
AccessTime: timestamp,
ChangeTime: timestamp,
}
if err := tw.WriteHeader(hdr); err != nil {
log.Debugf("Can't write whiteout header: %s", err)
}
} else {
path := filepath.Join(dir, change.Path)
if err := addTarFile(path, change.Path[1:], tw, twBuf); err != nil {
log.Debugf("Can't add file %s to tar: %s", path, err)
}
}
}
// Make sure to check the error on Close.
if err := tw.Close(); err != nil {
log.Debugf("Can't close layer: %s", err)
}
writer.Close()
}()
return reader, nil
}
func (devices *DeviceSet) deleteDevice(info *DevInfo) error {
if devices.doBlkDiscard {
// This is a workaround for the kernel not discarding block so
// on the thin pool when we remove a thinp device, so we do it
// manually
if err := devices.activateDeviceIfNeeded(info); err == nil {
if err := BlockDeviceDiscard(info.DevName()); err != nil {
log.Debugf("Error discarding block on device: %s (ignoring)", err)
}
}
}
devinfo, _ := getInfo(info.Name())
if devinfo != nil && devinfo.Exists != 0 {
if err := devices.removeDeviceAndWait(info.Name()); err != nil {
log.Debugf("Error removing device: %s", err)
return err
}
}
if err := deleteDevice(devices.getPoolDevName(), info.DeviceId); err != nil {
log.Debugf("Error deleting device: %s", err)
return err
}
devices.allocateTransactionId()
devices.devicesLock.Lock()
delete(devices.Devices, info.Hash)
devices.devicesLock.Unlock()
if err := devices.removeMetadata(info); err != nil {
devices.devicesLock.Lock()
devices.Devices[info.Hash] = info
devices.devicesLock.Unlock()
log.Debugf("Error removing meta data: %s", err)
return err
}
return nil
}
func (devices *DeviceSet) deactivateDevice(info *DevInfo) error {
log.Debugf("[devmapper] deactivateDevice(%s)", info.Hash)
defer log.Debugf("[devmapper] deactivateDevice END")
// Wait for the unmount to be effective,
// by watching the value of Info.OpenCount for the device
if err := devices.waitClose(info); err != nil {
log.Errorf("Warning: error waiting for device %s to close: %s", info.Hash, err)
}
devinfo, err := getInfo(info.Name())
if err != nil {
return err
}
if devinfo.Exists != 0 {
if err := devices.removeDeviceAndWait(info.Name()); err != nil {
return err
}
}
return nil
}
func (devices *DeviceSet) UnmountDevice(hash string) error {
log.Debugf("[devmapper] UnmountDevice(hash=%s)", hash)
defer log.Debugf("[devmapper] UnmountDevice END")
info, err := devices.lookupDevice(hash)
if err != nil {
return err
}
info.lock.Lock()
defer info.lock.Unlock()
devices.Lock()
defer devices.Unlock()
if info.mountCount == 0 {
return fmt.Errorf("UnmountDevice: device not-mounted id %s\n", hash)
}
info.mountCount--
if info.mountCount > 0 {
return nil
}
log.Debugf("[devmapper] Unmount(%s)", info.mountPath)
if err := syscall.Unmount(info.mountPath, 0); err != nil {
return err
}
log.Debugf("[devmapper] Unmount done")
if err := devices.deactivateDevice(info); err != nil {
return err
}
info.mountPath = ""
return nil
}
// Useful helper for cleanup
func RemoveDevice(name string) error {
task := TaskCreate(DeviceRemove)
if task == nil {
return ErrCreateRemoveTask
}
if err := task.SetName(name); err != nil {
log.Debugf("Can't set task name %s", name)
return err
}
if err := task.Run(); err != nil {
return ErrRunRemoveDevice
}
return nil
}
// CopyFileWithTar emulates the behavior of the 'cp' command-line
// for a single file. It copies a regular file from path `src` to
// path `dst`, and preserves all its metadata.
//
// If `dst` ends with a trailing slash '/', the final destination path
// will be `dst/base(src)`.
func CopyFileWithTar(src, dst string) (err error) {
log.Debugf("CopyFileWithTar(%s, %s)", src, dst)
srcSt, err := os.Stat(src)
if err != nil {
return err
}
if srcSt.IsDir() {
return fmt.Errorf("Can't copy a directory")
}
// Clean up the trailing /
if dst[len(dst)-1] == '/' {
dst = path.Join(dst, filepath.Base(src))
}
// Create the holding directory if necessary
if err := os.MkdirAll(filepath.Dir(dst), 0700); err != nil && !os.IsExist(err) {
return err
}
r, w := io.Pipe()
errC := utils.Go(func() error {
defer w.Close()
srcF, err := os.Open(src)
if err != nil {
return err
}
defer srcF.Close()
hdr, err := tar.FileInfoHeader(srcSt, "")
if err != nil {
return err
}
hdr.Name = filepath.Base(dst)
tw := tar.NewWriter(w)
defer tw.Close()
if err := tw.WriteHeader(hdr); err != nil {
return err
}
if _, err := io.Copy(tw, srcF); err != nil {
return err
}
return nil
})
defer func() {
if er := <-errC; err != nil {
err = er
}
}()
return Untar(r, filepath.Dir(dst), nil)
}
func (devices *DeviceSet) Shutdown() error {
log.Debugf("[deviceset %s] shutdown()", devices.devicePrefix)
log.Debugf("[devmapper] Shutting down DeviceSet: %s", devices.root)
defer log.Debugf("[deviceset %s] shutdown END", devices.devicePrefix)
var devs []*DevInfo
devices.devicesLock.Lock()
for _, info := range devices.Devices {
devs = append(devs, info)
}
devices.devicesLock.Unlock()
for _, info := range devs {
info.lock.Lock()
if info.mountCount > 0 {
// We use MNT_DETACH here in case it is still busy in some running
// container. This means it'll go away from the global scope directly,
// and the device will be released when that container dies.
if err := syscall.Unmount(info.mountPath, syscall.MNT_DETACH); err != nil {
log.Debugf("Shutdown unmounting %s, error: %s", info.mountPath, err)
}
devices.Lock()
if err := devices.deactivateDevice(info); err != nil {
log.Debugf("Shutdown deactivate %s , error: %s", info.Hash, err)
}
devices.Unlock()
}
info.lock.Unlock()
}
info, _ := devices.lookupDevice("")
if info != nil {
info.lock.Lock()
devices.Lock()
if err := devices.deactivateDevice(info); err != nil {
log.Debugf("Shutdown deactivate base , error: %s", err)
}
devices.Unlock()
info.lock.Unlock()
}
devices.Lock()
if err := devices.deactivatePool(); err != nil {
log.Debugf("Shutdown deactivate pool , error: %s", err)
}
devices.Unlock()
return nil
}
func getStatus(name string) (uint64, uint64, string, string, error) {
task, err := createTask(DeviceStatus, name)
if task == nil {
log.Debugf("getStatus: Error createTask: %s", err)
return 0, 0, "", "", err
}
if err := task.Run(); err != nil {
log.Debugf("getStatus: Error Run: %s", err)
return 0, 0, "", "", err
}
devinfo, err := task.GetInfo()
if err != nil {
log.Debugf("getStatus: Error GetInfo: %s", err)
return 0, 0, "", "", err
}
if devinfo.Exists == 0 {
log.Debugf("getStatus: Non existing device %s", name)
return 0, 0, "", "", fmt.Errorf("Non existing device %s", name)
}
_, start, length, targetType, params := task.GetNextTarget(0)
return start, length, targetType, params, nil
}
func DetectCompression(source []byte) Compression {
for compression, m := range map[Compression][]byte{
Bzip2: {0x42, 0x5A, 0x68},
Gzip: {0x1F, 0x8B, 0x08},
Xz: {0xFD, 0x37, 0x7A, 0x58, 0x5A, 0x00},
} {
if len(source) < len(m) {
log.Debugf("Len too short")
continue
}
if bytes.Compare(m, source[:len(m)]) == 0 {
return compression
}
}
return Uncompressed
}
// Matches returns true if relFilePath matches any of the patterns
func Matches(relFilePath string, patterns []string) (bool, error) {
for _, exclude := range patterns {
matched, err := filepath.Match(exclude, relFilePath)
if err != nil {
log.Errorf("Error matching: %s (pattern: %s)", relFilePath, exclude)
return false, err
}
if matched {
if filepath.Clean(relFilePath) == "." {
log.Errorf("Can't exclude whole path, excluding pattern: %s", exclude)
continue
}
log.Debugf("Skipping excluded path: %s", relFilePath)
return true, nil
}
}
return false, nil
}
// attachLoopDevice attaches the given sparse file to the next
// available loopback device. It returns an opened *os.File.
func attachLoopDevice(sparseName string) (loop *os.File, err error) {
// Try to retrieve the next available loopback device via syscall.
// If it fails, we discard error and start loopking for a
// loopback from index 0.
startIndex, err := getNextFreeLoopbackIndex()
if err != nil {
log.Debugf("Error retrieving the next available loopback: %s", err)
}
// OpenFile adds O_CLOEXEC
sparseFile, err := os.OpenFile(sparseName, os.O_RDWR, 0644)
if err != nil {
log.Errorf("Error openning sparse file %s: %s", sparseName, err)
return nil, ErrAttachLoopbackDevice
}
defer sparseFile.Close()
loopFile, err := openNextAvailableLoopback(startIndex, sparseFile)
if err != nil {
return nil, err
}
// Set the status of the loopback device
loopInfo := &LoopInfo64{
loFileName: stringToLoopName(loopFile.Name()),
loOffset: 0,
loFlags: LoFlagsAutoClear,
}
if err := ioctlLoopSetStatus64(loopFile.Fd(), loopInfo); err != nil {
log.Errorf("Cannot set up loopback device info: %s", err)
// If the call failed, then free the loopback device
if err := ioctlLoopClrFd(loopFile.Fd()); err != nil {
log.Errorf("Error while cleaning up the loopback device")
}
loopFile.Close()
return nil, ErrAttachLoopbackDevice
}
return loopFile, nil
}
// waitClose blocks until either:
// a) the device registered at <device_set_prefix>-<hash> is closed,
// or b) the 10 second timeout expires.
func (devices *DeviceSet) waitClose(info *DevInfo) error {
i := 0
for ; i < 1000; i += 1 {
devinfo, err := getInfo(info.Name())
if err != nil {
return err
}
if i%100 == 0 {
log.Debugf("Waiting for unmount of %s: opencount=%d", info.Hash, devinfo.OpenCount)
}
if devinfo.OpenCount == 0 {
break
}
devices.Unlock()
time.Sleep(10 * time.Millisecond)
devices.Lock()
}
if i == 1000 {
return fmt.Errorf("Timeout while waiting for device %s to close", info.Hash)
}
return nil
}
func DecompressStream(archive io.Reader) (io.ReadCloser, error) {
buf := bufio.NewReader(archive)
bs, err := buf.Peek(10)
if err != nil {
return nil, err
}
log.Debugf("[tar autodetect] n: %v", bs)
compression := DetectCompression(bs)
switch compression {
case Uncompressed:
return ioutil.NopCloser(buf), nil
case Gzip:
return gzip.NewReader(buf)
case Bzip2:
return ioutil.NopCloser(bzip2.NewReader(buf)), nil
case Xz:
return xzDecompress(buf)
default:
return nil, fmt.Errorf("Unsupported compression format %s", (&compression).Extension())
}
}
// CreateBridgeIface creates a network bridge interface on the host system with the name `ifaceName`,
// and attempts to configure it with an address which doesn't conflict with any other interface on the host.
// If it can't find an address which doesn't conflict, it will return an error.
func createBridge(bridgeIP string) error {
nameservers := []string{}
resolvConf, _ := resolvconf.Get()
// we don't check for an error here, because we don't really care
// if we can't read /etc/resolv.conf. So instead we skip the append
// if resolvConf is nil. It either doesn't exist, or we can't read it
// for some reason.
if resolvConf != nil {
nameservers = append(nameservers, resolvconf.GetNameserversAsCIDR(resolvConf)...)
}
var ifaceAddr string
if len(bridgeIP) != 0 {
_, _, err := net.ParseCIDR(bridgeIP)
if err != nil {
return err
}
ifaceAddr = bridgeIP
} else {
for _, addr := range addrs {
_, dockerNetwork, err := net.ParseCIDR(addr)
if err != nil {
return err
}
if err := networkdriver.CheckNameserverOverlaps(nameservers, dockerNetwork); err == nil {
if err := networkdriver.CheckRouteOverlaps(dockerNetwork); err == nil {
ifaceAddr = addr
break
} else {
log.Debugf("%s %s", addr, err)
}
}
}
}
if ifaceAddr == "" {
return fmt.Errorf("Could not find a free IP address range for interface '%s'. Please configure its address manually and run 'docker -b %s'", bridgeIface, bridgeIface)
}
log.Debugf("Creating bridge %s with network %s", bridgeIface, ifaceAddr)
if err := createBridgeIface(bridgeIface); err != nil {
return err
}
iface, err := net.InterfaceByName(bridgeIface)
if err != nil {
return err
}
ipAddr, ipNet, err := net.ParseCIDR(ifaceAddr)
if err != nil {
return err
}
if netlink.NetworkLinkAddIp(iface, ipAddr, ipNet); err != nil {
return fmt.Errorf("Unable to add private network: %s", err)
}
if err := netlink.NetworkLinkUp(iface); err != nil {
return fmt.Errorf("Unable to start network bridge: %s", err)
}
return nil
}