refactor(pod): move scope-name resolution into containerCgroupPath

Container-level naming was split across two layers: cgroupfs appended
containerID inside containerCgroupPath, while systemd appended the
.scope unit in containerCgroupSuffix. Consolidate both in
containerCgroupPath by carrying the optional scope on cgroupPath, so
the suffix function only formats.

Author: hao022

internal/pod/container_cgroup_path.go

@@ -27,8 +27,12 @@ const (
 	defaultNodeCgroupName = "kubepods"
 )
 
-// {"kubepods", "burstable", "pod1234-abcd-5678-efgh"}
-type cgroupPath []string
+// slices: {"kubepods", "burstable", "pod1234-abcd-5678-efgh"}
+// scope: systemd scope unit appended after the slice hierarchy (systemd driver only).
+type cgroupPath struct {
+	slices []string
+	scope  string
+}
 
 func escapeSystemd(part string) string {
 	return strings.ReplaceAll(part, "-", "_")
@@ -52,21 +56,29 @@ func expandSystemdSlice(slice string) string {
 
 // {"kubepods", "burstable", "pod1234-abcd-5678-efgh"} becomes
 // "/kubepods.slice/kubepods-burstable.slice/kubepods-burstable-pod1234_abcd_5678_efgh.slice"
-func (paths cgroupPath) ToSystemd() string {
+// with the scope unit appended when present.
+func (p cgroupPath) ToSystemd() string {
 	newparts := []string{}
-	for _, part := range paths {
+	for _, part := range p.slices {
 		part = escapeSystemd(part)
 		newparts = append(newparts, part)
 	}
 
-	return expandSystemdSlice(strings.Join(newparts, "-") + defaultSystemdSuffix)
+	slicePath := expandSystemdSlice(strings.Join(newparts, "-") + defaultSystemdSuffix)
+
+	if p.scope != "" {
+		return slicePath + "/" + p.scope
+	}
+
+	return slicePath
 }
 
-func (paths cgroupPath) ToCgroupfs() string {
-	return "/" + path.Join(paths...)
+func (p cgroupPath) ToCgroupfs() string {
+	return "/" + path.Join(p.slices...)
 }
 
-func containerCgroupPath(containerID string, pod *corev1.Pod) cgroupPath {
+// https://github.com/kubernetes/kubernetes/blob/master/pkg/kubelet/cm/cgroup_manager_linux.go#L81
+func containerCgroupPath(containerID string, pod *corev1.Pod) (cgroupPath, error) {
 	paths := []string{defaultNodeCgroupName}
 
 	if pod.Status.QOSClass != corev1.PodQOSGuaranteed {
@@ -75,23 +87,26 @@ func containerCgroupPath(containerID string, pod *corev1.Pod) cgroupPath {
 
 	paths = append(paths, fmt.Sprintf("pod%s", pod.UID))
 
-	if kubeletPodCgroupDriver != "systemd" {
-		paths = append(paths, containerID)
+	if kubeletPodCgroupDriver == "systemd" {
+		scope, err := containerScopeName(containerID)
+		if err != nil {
+			return cgroupPath{}, fmt.Errorf("container scope name: %w", err)
+		}
+		return cgroupPath{slices: paths, scope: scope}, nil
 	}
 
-	return paths
+	paths = append(paths, containerID)
+	return cgroupPath{slices: paths}, nil
 }
 
-// https://github.com/kubernetes/kubernetes/blob/master/pkg/kubelet/cm/cgroup_manager_linux.go#L81
 func containerCgroupSuffix(containerID string, pod *corev1.Pod) (string, error) {
-	name := containerCgroupPath(containerID, pod)
+	name, err := containerCgroupPath(containerID, pod)
+	if err != nil {
+		return "", err
+	}
 
 	if kubeletPodCgroupDriver == "systemd" {
-		scope, err := containerScopeName(containerID)
-		if err != nil {
-			return "", fmt.Errorf("container scope name: %w", err)
-		}
-		return name.ToSystemd() + "/" + scope, nil
+		return name.ToSystemd(), nil
 	}
 
 	return name.ToCgroupfs(), nil