Merge pull request #8529 from xpivarc/release-0.49-psa

[release-0.49] Integrate with Pod security

hack/cluster-deploy.sh

@@ -74,6 +74,8 @@ apiVersion: v1
 kind: Namespace
 metadata:
   name: ${namespace:?}
+  labels:
+    pod-security.kubernetes.io/enforce: "privileged"
 EOF
 
 if [[ "$KUBEVIRT_PROVIDER" =~ kind.* ]]; then

manifests/generated/operator-csv.yaml.in

@@ -458,6 +458,15 @@ spec:
           - get
           - list
           - watch
+        - apiGroups:
+          - ""
+          resources:
+          - namespaces
+          verbs:
+          - get
+          - list
+          - watch
+          - patch
         - apiGroups:
           - policy
           resources:
@@ -1073,6 +1082,13 @@ spec:
                   requests:
                     cpu: 10m
                     memory: 150Mi
+                securityContext:
+                  allowPrivilegeEscalation: false
+                  capabilities:
+                    drop:
+                    - ALL
+                  seccompProfile:
+                    type: RuntimeDefault
                 volumeMounts:
                 - mountPath: /etc/virt-operator/certificates
                   name: kubevirt-operator-certs
@@ -1082,6 +1098,8 @@ spec:
               priorityClassName: kubevirt-cluster-critical
               securityContext:
                 runAsNonRoot: true
+                seccompProfile:
+                  type: RuntimeDefault
               serviceAccountName: kubevirt-operator
               tolerations:
               - key: CriticalAddonsOnly

manifests/generated/rbac-operator.authorization.k8s.yaml.in

@@ -360,6 +360,15 @@ rules:
   - get
   - list
   - watch
+- apiGroups:
+  - ""
+  resources:
+  - namespaces
+  verbs:
+  - get
+  - list
+  - watch
+  - patch
 - apiGroups:
   - policy
   resources:

manifests/release/kubevirt-operator.yaml.in

@@ -4,6 +4,7 @@ kind: Namespace
 metadata:
   labels:
     kubevirt.io: ""
+    pod-security.kubernetes.io/enforce: "privileged"
   name: {{.Namespace}}
 {{index .GeneratedManifests "kv-resource.yaml"}}
 ---

pkg/virt-config/feature-gates.go

@@ -45,6 +45,7 @@ const (
 	NonRoot                    = "NonRootExperimental"
 	ClusterProfiler            = "ClusterProfiler"
 	WorkloadEncryptionSEV      = "WorkloadEncryptionSEV"
+	PSA                        = "PSA"
 )
 
 func (c *ClusterConfig) isFeatureGateEnabled(featureGate string) bool {
@@ -136,3 +137,7 @@ func (config *ClusterConfig) ClusterProfilerEnabled() bool {
 func (config *ClusterConfig) WorkloadEncryptionSEVEnabled() bool {
 	return config.isFeatureGateEnabled(WorkloadEncryptionSEV)
 }
+
+func (config *ClusterConfig) PSAEnabled() bool {
+	return config.isFeatureGateEnabled(PSA)
+}

pkg/virt-controller/watch/BUILD.bazel

@@ -8,6 +8,7 @@ go_library(
         "migration.go",
         "node.go",
         "pool.go",
+        "psa.go",
         "replicaset.go",
         "util.go",
         "vm.go",
@@ -26,6 +27,7 @@ go_library(
         "//pkg/monitoring/vmistats:go_default_library",
         "//pkg/service:go_default_library",
         "//pkg/util:go_default_library",
+        "//pkg/util/cluster:go_default_library",
         "//pkg/util/lookup:go_default_library",
         "//pkg/util/migrations:go_default_library",
         "//pkg/util/pdbs:go_default_library",
@@ -94,6 +96,7 @@ go_test(
         "migration_test.go",
         "node_test.go",
         "pool_test.go",
+        "psa_test.go",
         "replicaset_test.go",
         "vm_test.go",
         "vmi_test.go",

pkg/virt-controller/watch/application.go

@@ -65,6 +65,7 @@ import (
 	vmiprom "kubevirt.io/kubevirt/pkg/monitoring/vmistats" // import for prometheus metrics
 	"kubevirt.io/kubevirt/pkg/service"
 	"kubevirt.io/kubevirt/pkg/util"
+	clusterutil "kubevirt.io/kubevirt/pkg/util/cluster"
 	"kubevirt.io/kubevirt/pkg/util/webhooks"
 	virtconfig "kubevirt.io/kubevirt/pkg/virt-config"
 	"kubevirt.io/kubevirt/pkg/virt-controller/leaderelectionconfig"
@@ -139,6 +140,8 @@ type VirtControllerApp struct {
 	vmiInformer   cache.SharedIndexInformer
 	vmiRecorder   record.EventRecorder
 
+	namespaceStore cache.Store
+
 	kubeVirtInformer cache.SharedIndexInformer
 
 	clusterConfig *virtconfig.ClusterConfig
@@ -227,6 +230,8 @@ type VirtControllerApp struct {
 	nodeTopologyUpdatePeriod time.Duration
 	reloadableRateLimiter    *ratelimiter.ReloadableRateLimiter
 	leaderElector            *leaderelection.LeaderElector
+
+	onOpenshift bool
 }
 
 var _ service.Service = &VirtControllerApp{}
@@ -316,7 +321,7 @@ func Execute() {
 	app.vmiInformer = app.informerFactory.VMI()
 	app.kvPodInformer = app.informerFactory.KubeVirtPod()
 	app.nodeInformer = app.informerFactory.KubeVirtNode()
-
+	app.namespaceStore = app.informerFactory.Namespace().GetStore()
 	app.vmiCache = app.vmiInformer.GetStore()
 	app.vmiRecorder = app.newRecorder(k8sv1.NamespaceAll, "virtualmachine-controller")
 
@@ -360,6 +365,12 @@ func Execute() {
 
 	app.migrationPolicyInformer = app.informerFactory.MigrationPolicy()
 
+	onOpenShift, err := clusterutil.IsOnOpenShift(app.clientSet)
+	if err != nil {
+		golog.Fatalf("Error determining cluster type: %v", err)
+	}
+	app.onOpenshift = onOpenShift
+
 	app.initCommon()
 	app.initReplicaSet()
 	app.initPool()
@@ -512,6 +523,8 @@ func (vca *VirtControllerApp) initCommon() {
 		vca.cdiConfigInformer,
 		vca.clusterConfig,
 		topologyHinter,
+		vca.namespaceStore,
+		vca.onOpenshift,
 	)
 
 	recorder := vca.newRecorder(k8sv1.NamespaceAll, "node-controller")
@@ -528,6 +541,8 @@ func (vca *VirtControllerApp) initCommon() {
 		vca.vmiRecorder,
 		vca.clientSet,
 		vca.clusterConfig,
+		vca.namespaceStore,
+		vca.onOpenshift,
 	)
 
 	vca.nodeTopologyUpdater = topology.NewNodeTopologyUpdater(vca.clientSet, topologyHinter, vca.nodeInformer)

pkg/virt-controller/watch/application_test.go

@@ -124,6 +124,8 @@ var _ = Describe("Application", func() {
 			cdiConfigInformer,
 			config,
 			topology.NewTopologyHinter(&cache.FakeCustomStore{}, &cache.FakeCustomStore{}, "amd64", nil),
+			nil,
+			false,
 		)
 		app.rsController = NewVMIReplicaSet(vmiInformer, rsInformer, recorder, virtClient, uint(10))
 		app.vmController = NewVMController(vmiInformer,
@@ -145,6 +147,8 @@ var _ = Describe("Application", func() {
 			recorder,
 			virtClient,
 			config,
+			nil,
+			false,
 		)
 		app.snapshotController = &snapshot.VMSnapshotController{
 			Client:                    virtClient,

pkg/virt-controller/watch/migration.go

@@ -94,6 +94,7 @@ type MigrationController struct {
 	pvcInformer             cache.SharedIndexInformer
 	pdbInformer             cache.SharedIndexInformer
 	migrationPolicyInformer cache.SharedIndexInformer
+	namespaceStore          cache.Store
 	recorder                record.EventRecorder
 	podExpectations         *controller.UIDTrackingControllerExpectations
 	migrationStartLock      *sync.Mutex
@@ -102,6 +103,8 @@ type MigrationController struct {
 
 	unschedulablePendingTimeoutSeconds int64
 	catchAllPendingTimeoutSeconds      int64
+
+	onOpenshift bool
 }
 
 func NewMigrationController(templateService services.TemplateService,
@@ -115,6 +118,8 @@ func NewMigrationController(templateService services.TemplateService,
 	recorder record.EventRecorder,
 	clientset kubecli.KubevirtClient,
 	clusterConfig *virtconfig.ClusterConfig,
+	namespaceStore cache.Store,
+	onOpenshift bool,
 ) *MigrationController {
 
 	c := &MigrationController{
@@ -136,6 +141,9 @@ func NewMigrationController(templateService services.TemplateService,
 
 		unschedulablePendingTimeoutSeconds: defaultUnschedulablePendingTimeoutSeconds,
 		catchAllPendingTimeoutSeconds:      defaultCatchAllPendingTimeoutSeconds,
+
+		namespaceStore: namespaceStore,
+		onOpenshift:    onOpenshift,
 	}
 
 	c.vmiInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
@@ -582,6 +590,12 @@ func (c *MigrationController) createTargetPod(migration *virtv1.VirtualMachineIn
 		}
 	}
 
+	if c.clusterConfig.PSAEnabled() {
+		if err := escalateNamespace(c.namespaceStore, c.clientset, vmi.GetNamespace(), c.onOpenshift); err != nil {
+			return err
+		}
+	}
+
 	key := controller.MigrationKey(migration)
 	c.podExpectations.ExpectCreations(key, 1)
 	pod, err := c.clientset.CoreV1().Pods(vmi.GetNamespace()).Create(context.Background(), templatePod, v1.CreateOptions{})
@@ -840,8 +854,14 @@ func (c *MigrationController) createAttachmentPod(migration *virtv1.VirtualMachi
 	attachmentPodTemplate.ObjectMeta.Labels[virtv1.MigrationJobLabel] = string(migration.UID)
 	attachmentPodTemplate.ObjectMeta.Annotations[virtv1.MigrationJobNameAnnotation] = string(migration.Name)
 
+	if c.clusterConfig.PSAEnabled() {
+		if err := escalateNamespace(c.namespaceStore, c.clientset, vmi.GetNamespace(), c.onOpenshift); err != nil {
+			return err
+		}
+	}
 	key := controller.MigrationKey(migration)
 	c.podExpectations.ExpectCreations(key, 1)
+
 	attachmentPod, err := c.clientset.CoreV1().Pods(vmi.GetNamespace()).Create(context.Background(), attachmentPodTemplate, v1.CreateOptions{})
 	if err != nil {
 		c.podExpectations.CreationObserved(key)

pkg/virt-controller/watch/migration_test.go

@@ -270,6 +270,8 @@ var _ = Describe("Migration watcher", func() {
 			recorder,
 			virtClient,
 			config,
+			nil,
+			false,
 		)
 		// Wrap our workqueue to have a way to detect when we are done processing updates
 		mockQueue = testutils.NewMockWorkQueue(controller.Queue)

pkg/virt-controller/watch/psa.go

@@ -0,0 +1,60 @@
+/*
+ * This file is part of the KubeVirt project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * Copyright 2022 Red Hat, Inc.
+ *
+ */
+
+package watch
+
+import (
+	"context"
+	"fmt"
+
+	k8sv1 "k8s.io/api/core/v1"
+	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/client-go/tools/cache"
+
+	"kubevirt.io/client-go/kubecli"
+)
+
+const PSALabel = "pod-security.kubernetes.io/enforce"
+const OpenshiftPSAsync = "security.openshift.io/scc.podSecurityLabelSync"
+
+func escalateNamespace(namespaceStore cache.Store, client kubecli.KubevirtClient, namespace string, onOpenshift bool) error {
+	obj, exists, err := namespaceStore.GetByKey(namespace)
+	if err != nil {
+		return fmt.Errorf("Failed to get namespace, %w", err)
+	}
+	if !exists {
+		return fmt.Errorf("Namespace %s not observed, %w", namespace, err)
+	}
+	namespaceObj := obj.(*k8sv1.Namespace)
+	enforceLevel, labelExist := namespaceObj.Labels[PSALabel]
+	if !labelExist || enforceLevel != "privileged" {
+		labels := ""
+		if !onOpenshift {
+			labels = fmt.Sprintf(`{"%s": "privileged"}`, PSALabel)
+		} else {
+			labels = fmt.Sprintf(`{"%s": "privileged", "%s": "false"}`, PSALabel, OpenshiftPSAsync)
+		}
+		data := []byte(fmt.Sprintf(`{"metadata": { "labels": %s}}`, labels))
+		_, err := client.CoreV1().Namespaces().Patch(context.TODO(), namespace, types.StrategicMergePatchType, data, v1.PatchOptions{})
+		if err != nil {
+			return &syncErrorImpl{err, fmt.Sprintf("Failed to apply enforce label on namespace %s", namespace)}
+		}
+	}
+	return nil
+}