info: report cgroup driver as "none" when running rootless

Previously `docker info` had reported "cgroupfs" as the cgroup driver
but the driver wasn't actually used at all.

This PR reports "none" as the cgroup driver so as to avoid confusion.
e.g. kubeadm/kubelet will detect cgroupless-ness by checking this docker
info field. https://github.com/rootless-containers/usernetes/pull/97

Note that user still cannot specify `native.cgroupdriver=none` manually.

Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>

daemon/daemon_unix.go

@@ -73,6 +73,7 @@ const (
 	// constant for cgroup drivers
 	cgroupFsDriver      = "cgroupfs"
 	cgroupSystemdDriver = "systemd"
+	cgroupNoneDriver    = "none"
 
 	// DefaultRuntimeName is the default runtime to be used by
 	// containerd if none is specified
@@ -584,6 +585,9 @@ func verifyPlatformContainerResources(resources *containertypes.Resources, sysIn
 }
 
 func (daemon *Daemon) getCgroupDriver() string {
+	if daemon.Rootless() {
+		return cgroupNoneDriver
+	}
 	cgroupDriver := cgroupFsDriver
 
 	if UsingSystemd(daemon.configStore) {
@@ -610,6 +614,9 @@ func VerifyCgroupDriver(config *config.Config) error {
 	if cd == "" || cd == cgroupFsDriver || cd == cgroupSystemdDriver {
 		return nil
 	}
+	if cd == cgroupNoneDriver {
+		return fmt.Errorf("native.cgroupdriver option %s is internally used and cannot be specified manually", cd)
+	}
 	return fmt.Errorf("native.cgroupdriver option %s not supported", cd)
 }