diff --git a/apis/opts/mountpoint.go b/apis/opts/mountpoint.go
index 4dbfcd86fe..f1fca35409 100644
--- a/apis/opts/mountpoint.go
+++ b/apis/opts/mountpoint.go
@@ -59,7 +59,6 @@ func ParseVolumesFrom(volume string) (string, string, error) {
 // ParseBindMode is used to parse the bind's mode.
 func ParseBindMode(mp *types.MountPoint, mode string) error {
 	mp.RW = true
-	mp.CopyData = true
 
 	defaultMode := 0
 	rwMode := 0
diff --git a/daemon/mgr/container.go b/daemon/mgr/container.go
index 68cf737344..03ac71651b 100644
--- a/daemon/mgr/container.go
+++ b/daemon/mgr/container.go
@@ -1950,6 +1950,12 @@ func (mgr *ContainerManager) generateMountPoints(ctx context.Context, c *Contain
 		return errors.Wrap(err, "failed to get mount point from volumes")
 	}
 
+	// 5. populate the volumes
+	err = mgr.populateVolumes(ctx, c)
+	if err != nil {
+		return errors.Wrap(err, "failed to populate volumes")
+	}
+
 	return nil
 }
 
@@ -1973,6 +1979,7 @@ func (mgr *ContainerManager) getMountPointFromBinds(ctx context.Context, c *Cont
 		case 1:
 			mp.Source = ""
 			mp.Destination = parts[0]
+			mp.CopyData = true
 		case 2:
 			mp.Source = parts[0]
 			mp.Destination = parts[1]
@@ -1993,11 +2000,6 @@ func (mgr *ContainerManager) getMountPointFromBinds(ctx context.Context, c *Cont
 
 		if mp.Source == "" {
 			mp.Source = randomid.Generate()
-
-			// Source is empty, anonymouse volume
-			if _, exist := c.Config.Volumes[mp.Destination]; !exist {
-				c.Config.Volumes[mp.Destination] = struct{}{}
-			}
 		}
 
 		err = opts.ParseBindMode(mp, mode)
@@ -2037,6 +2039,7 @@ func (mgr *ContainerManager) getMountPointFromBinds(ctx context.Context, c *Cont
 				mp.Named = false
 				mp.Driver = ""
 			}
+			mp.CopyData = true
 		}
 
 		if _, err = os.Stat(mp.Source); err != nil {
@@ -2075,6 +2078,7 @@ func (mgr *ContainerManager) getMountPointFromVolumes(ctx context.Context, c *Co
 		mp := new(types.MountPoint)
 		mp.Name = name
 		mp.Destination = dest
+		mp.CopyData = true
 
 		mp.Source, mp.Driver, err = mgr.attachVolume(ctx, mp.Name, c)
 		if err != nil {
@@ -2104,10 +2108,6 @@ func (mgr *ContainerManager) getMountPointFromImage(ctx context.Context, c *Cont
 		return errors.Wrapf(err, "failed to get image: %s", c.Image)
 	}
 	for dest := range image.Config.Volumes {
-		if _, exist := c.Config.Volumes[dest]; !exist {
-			c.Config.Volumes[dest] = struct{}{}
-		}
-
 		// check if volume has been created
 		name := randomid.Generate()
 		if _, exist := volumeSet[name]; exist {
@@ -2122,6 +2122,7 @@ func (mgr *ContainerManager) getMountPointFromImage(ctx context.Context, c *Cont
 		mp := new(types.MountPoint)
 		mp.Name = name
 		mp.Destination = dest
+		mp.CopyData = true
 
 		mp.Source, mp.Driver, err = mgr.attachVolume(ctx, mp.Name, c)
 		if err != nil {
@@ -2172,6 +2173,7 @@ func (mgr *ContainerManager) getMountPointFromContainers(ctx context.Context, co
 				Named:       oldMountPoint.Named,
 				RW:          oldMountPoint.RW,
 				Propagation: oldMountPoint.Propagation,
+				CopyData:    false,
 			}
 
 			if _, exist := volumeSet[oldMountPoint.Name]; len(oldMountPoint.Name) > 0 && !exist {
@@ -2182,7 +2184,6 @@ func (mgr *ContainerManager) getMountPointFromContainers(ctx context.Context, co
 					return errors.Wrap(err, "failed to bind volume")
 				}
 
-				container.Config.Volumes[mp.Destination] = struct{}{}
 				volumeSet[mp.Name] = struct{}{}
 			}
 
@@ -2199,6 +2200,37 @@ func (mgr *ContainerManager) getMountPointFromContainers(ctx context.Context, co
 	return nil
 }
 
+func (mgr *ContainerManager) populateVolumes(ctx context.Context, c *Container) error {
+	err := mgr.Mount(ctx, c)
+	if err != nil {
+		return err
+	}
+	defer mgr.Unmount(ctx, c)
+
+	for _, mnt := range c.Mounts {
+		if mnt.Driver == "tmpfs" {
+			continue
+		}
+
+		if !mnt.CopyData {
+			continue
+		}
+
+		logrus.Debugf("copying image data from %s:%s, to %s, path: %s",
+			c.ID, mnt.Destination, mnt.Name, mnt.Source)
+
+		imagePath := path.Join(mgr.Store.Path(c.ID), "rootfs", mnt.Destination)
+
+		err := copyImageContent(imagePath, mnt.Source)
+		if err != nil {
+			logrus.Errorf("failed to populate volume[name: %s, source: %s] err: %v", mnt.Name, mnt.Source, err)
+			return err
+		}
+	}
+
+	return nil
+}
+
 func (mgr *ContainerManager) setMountPointDiskQuota(ctx context.Context, c *Container) error {
 	if c.Config.DiskQuota == nil {
 		if c.Config.QuotaID != "" && c.Config.QuotaID != "0" {
@@ -2386,3 +2418,73 @@ func (mgr *ContainerManager) execProcessGC() {
 func (mgr *ContainerManager) NewSnapshotsSyncer(snapshotStore *SnapshotStore, duration time.Duration) *SnapshotsSyncer {
 	return newSnapshotsSyncer(snapshotStore, mgr.Client, duration)
 }
+
+// Mount sets the container rootfs
+func (mgr *ContainerManager) Mount(ctx context.Context, c *Container) error {
+	mounts, err := mgr.Client.GetMounts(ctx, c.ID)
+	if err != nil {
+		return err
+	} else if len(mounts) != 1 {
+		return fmt.Errorf("failed to get snapshot %s mounts: not equals 1", c.ID)
+	}
+
+	mountPath := path.Join(mgr.Store.Path(c.ID), "rootfs")
+
+	err = os.MkdirAll(mountPath, 0755)
+	if err != nil && !os.IsExist(err) {
+		return err
+	}
+
+	return mounts[0].Mount(mountPath)
+}
+
+// Unmount unsets the container rootfs
+func (mgr *ContainerManager) Unmount(ctx context.Context, c *Container) error {
+	mountPath := path.Join(mgr.Store.Path(c.ID), "rootfs")
+
+	err := mount.Unmount(mountPath, 0)
+	if err != nil {
+		return err
+	}
+
+	return os.RemoveAll(mountPath)
+}
+
+func copyImageContent(source, destination string) error {
+	volList, err := ioutil.ReadDir(source)
+	if err != nil {
+		if os.IsNotExist(err) {
+			return nil
+		}
+		return err
+	}
+
+	if len(volList) > 0 {
+		dstList, err := ioutil.ReadDir(destination)
+		if err != nil {
+			return err
+		}
+		if len(dstList) == 0 {
+			// TODO: refactor
+
+			// If the source volume is empty, copies files from the root into the volume
+			commandLine := fmt.Sprintf("cp -r %s/* %s", source, destination)
+
+			cmd := exec.Command("sh", "-c", commandLine)
+			out, err := cmd.CombinedOutput()
+			if err != nil {
+				logrus.Errorf("copyImageContent: %s, %v", string(out), err)
+				return err
+			}
+		}
+	}
+	return copyOwnership(source, destination)
+}
+
+func copyOwnership(source, destination string) error {
+	fi, err := os.Stat(source)
+	if err != nil {
+		return err
+	}
+	return os.Chmod(destination, os.FileMode(fi.Mode()))
+}