Merge pull request #29530 from QiWang19/imgpolicy-sigstore-verification

OCPNODE-2596: Add SigstoreImageVerification e2e tests

Author: openshift-merge-bot[bot]

test/extended/imagepolicy/imagepolicy.go

@@ -0,0 +1,238 @@
+package imagepolicy
+
+import (
+	"context"
+	"fmt"
+	"path/filepath"
+	"time"
+
+	g "github.com/onsi/ginkgo/v2"
+	o "github.com/onsi/gomega"
+	machineconfighelper "github.com/openshift/origin/test/extended/machine_config"
+	exutil "github.com/openshift/origin/test/extended/util"
+	kapiv1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/client-go/util/retry"
+	e2e "k8s.io/kubernetes/test/e2e/framework"
+	e2epod "k8s.io/kubernetes/test/e2e/framework/pod"
+	e2eskipper "k8s.io/kubernetes/test/e2e/framework/skipper"
+	admissionapi "k8s.io/pod-security-admission/api"
+)
+
+const (
+	testReleaseImageScope             = "quay.io/openshift-release-dev/ocp-release@sha256:fbad931c725b2e5b937b295b58345334322bdabb0b67da1c800a53686d7397da"
+	testReferenceImageScope           = "quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:4db234f37ae6712e2f7ed8d13f7fb49971c173d0e4f74613d0121672fa2e01f5"
+	registriesWorkerPoolMachineConfig = "99-worker-generated-registries"
+	registriesMasterPoolMachineConfig = "99-master-generated-registries"
+	testPodName                       = "signature-validation-test-pod"
+	workerPool                        = "worker"
+	masterPool                        = "master"
+	SignatureValidationFaildReason    = "SignatureValidationFailed"
+)
+
+var _ = g.Describe("[sig-imagepolicy][OCPFeatureGate:SigstoreImageVerification][Serial]", g.Ordered, func() {
+	defer g.GinkgoRecover()
+	var (
+		oc                                        = exutil.NewCLIWithoutNamespace("cluster-image-policy")
+		tctx                                      = context.Background()
+		cli                                       = exutil.NewCLIWithPodSecurityLevel("verifysigstore-e2e", admissionapi.LevelBaseline)
+		clif                                      = cli.KubeFramework()
+		imgpolicyCli                              = exutil.NewCLIWithPodSecurityLevel("verifysigstore-imagepolicy-e2e", admissionapi.LevelBaseline)
+		imgpolicyClif                             = imgpolicyCli.KubeFramework()
+		imagePolicyBaseDir                        = exutil.FixturePath("testdata", "imagepolicy")
+		invalidPublicKeyClusterImagePolicyFixture = filepath.Join(imagePolicyBaseDir, "invalid-public-key-cluster-image-policy.yaml")
+		publiKeyRekorClusterImagePolicyFixture    = filepath.Join(imagePolicyBaseDir, "public-key-rekor-cluster-image-policy.yaml")
+		invalidPublicKeyImagePolicyFixture        = filepath.Join(imagePolicyBaseDir, "invalid-public-key-image-policy.yaml")
+		publiKeyRekorImagePolicyFixture           = filepath.Join(imagePolicyBaseDir, "public-key-rekor-image-policy.yaml")
+	)
+
+	g.BeforeAll(func() {
+		if !exutil.IsTechPreviewNoUpgrade(tctx, oc.AdminConfigClient()) {
+			g.Skip("skipping, this feature is only supported on TechPreviewNoUpgrade clusters")
+		}
+
+		outStr, err := oc.Run("adm", "release", "info", testReleaseImageScope).Args("-o=go-template", "--template={{.digest}}").Output()
+		if err != nil || outStr == "" {
+			o.Expect(err).ToNot(o.HaveOccurred())
+			e2eskipper.Skipf("can't validate %s release image for testing, consider updating the test", testReleaseImageScope)
+		}
+	})
+
+	g.It("Should fail clusterimagepolicy signature validation root of trust does not match the identity in the signature", func() {
+		createClusterImagePolicy(oc, invalidPublicKeyClusterImagePolicyFixture)
+		g.DeferCleanup(deleteClusterImagePolicy, oc, invalidPublicKeyClusterImagePolicyFixture)
+
+		pod, err := launchTestPod(tctx, clif, testPodName, testReleaseImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, clif, testPodName)
+
+		err = waitForTestPodContainerToFailSignatureValidation(tctx, clif, pod)
+		o.Expect(err).NotTo(o.HaveOccurred())
+	})
+
+	g.It("Should fail clusterimagepolicy signature validation when scope in allowedRegistries list does not skip signature verification", func() {
+		// Ensure allowedRegistries do not skip signature verification by adding testReleaseImageScope to the list
+		allowedRegistries := []string{"quay.io", "registry.redhat.io", "image-registry.openshift-image-registry.svc:5000", testReleaseImageScope}
+		updateImageConfig(oc, allowedRegistries)
+		g.DeferCleanup(cleanupImageConfig, oc)
+
+		createClusterImagePolicy(oc, invalidPublicKeyClusterImagePolicyFixture)
+		g.DeferCleanup(deleteClusterImagePolicy, oc, invalidPublicKeyClusterImagePolicyFixture)
+
+		pod, err := launchTestPod(tctx, clif, testPodName, testReleaseImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, clif, testPodName)
+
+		err = waitForTestPodContainerToFailSignatureValidation(tctx, clif, pod)
+		o.Expect(err).NotTo(o.HaveOccurred())
+	})
+
+	g.It("Should pass clusterimagepolicy signature validation with signed image", func() {
+		createClusterImagePolicy(oc, publiKeyRekorClusterImagePolicyFixture)
+		g.DeferCleanup(deleteClusterImagePolicy, oc, publiKeyRekorClusterImagePolicyFixture)
+
+		pod, err := launchTestPod(tctx, clif, testPodName, testReleaseImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, clif, testPodName)
+
+		err = e2epod.WaitForPodSuccessInNamespace(tctx, clif.ClientSet, pod.Name, pod.Namespace)
+		o.Expect(err).NotTo(o.HaveOccurred())
+	})
+
+	g.It("Should fail imagepolicy signature validation in different namespaces root of trust does not match the identity in the signature", func() {
+		createImagePolicy(oc, invalidPublicKeyImagePolicyFixture, imgpolicyClif.Namespace.Name)
+		g.DeferCleanup(deleteImagePolicy, oc, invalidPublicKeyImagePolicyFixture, imgpolicyClif.Namespace.Name)
+
+		createImagePolicy(oc, invalidPublicKeyImagePolicyFixture, clif.Namespace.Name)
+		g.DeferCleanup(deleteImagePolicy, oc, invalidPublicKeyImagePolicyFixture, clif.Namespace.Name)
+
+		pod, err := launchTestPod(tctx, imgpolicyClif, testPodName, testReferenceImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, imgpolicyClif, testPodName)
+
+		err = waitForTestPodContainerToFailSignatureValidation(tctx, imgpolicyClif, pod)
+		o.Expect(err).NotTo(o.HaveOccurred())
+
+		pod, err = launchTestPod(tctx, clif, testPodName, testReferenceImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, clif, testPodName)
+
+		err = waitForTestPodContainerToFailSignatureValidation(tctx, clif, pod)
+		o.Expect(err).NotTo(o.HaveOccurred())
+	})
+
+	g.It("Should pass imagepolicy signature validation with signed image in namespaces", func() {
+		createImagePolicy(oc, publiKeyRekorImagePolicyFixture, clif.Namespace.Name)
+		g.DeferCleanup(deleteImagePolicy, oc, publiKeyRekorImagePolicyFixture, clif.Namespace.Name)
+
+		createImagePolicy(oc, publiKeyRekorImagePolicyFixture, imgpolicyClif.Namespace.Name)
+		g.DeferCleanup(deleteImagePolicy, oc, publiKeyRekorImagePolicyFixture, imgpolicyClif.Namespace.Name)
+
+		pod, err := launchTestPod(tctx, clif, testPodName, testReferenceImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, clif, testPodName)
+
+		err = e2epod.WaitForPodSuccessInNamespace(tctx, clif.ClientSet, pod.Name, pod.Namespace)
+		o.Expect(err).NotTo(o.HaveOccurred())
+
+		pod, err = launchTestPod(tctx, imgpolicyClif, testPodName, testReferenceImageScope)
+		o.Expect(err).NotTo(o.HaveOccurred())
+		g.DeferCleanup(deleteTestPod, tctx, imgpolicyClif, testPodName)
+
+		err = e2epod.WaitForPodSuccessInNamespace(tctx, imgpolicyClif.ClientSet, pod.Name, pod.Namespace)
+		o.Expect(err).NotTo(o.HaveOccurred())
+	})
+})
+
+func createClusterImagePolicy(oc *exutil.CLI, fixture string) {
+	err := oc.Run("create").Args("-f", fixture).Execute()
+	o.Expect(err).NotTo(o.HaveOccurred())
+	time.Sleep(10 * time.Second)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, workerPool, registriesWorkerPoolMachineConfig)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, masterPool, registriesMasterPoolMachineConfig)
+}
+
+func deleteClusterImagePolicy(oc *exutil.CLI, fixture string) error {
+	return oc.Run("delete").Args("-f", fixture).Execute()
+}
+
+func createImagePolicy(oc *exutil.CLI, fixture string, namespace string) {
+	err := oc.Run("create").Args("-f", fixture, "-n", namespace).Execute()
+	o.Expect(err).NotTo(o.HaveOccurred())
+	time.Sleep(10 * time.Second)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, workerPool, registriesWorkerPoolMachineConfig)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, masterPool, registriesMasterPoolMachineConfig)
+}
+
+func deleteImagePolicy(oc *exutil.CLI, fixture string, namespace string) error {
+	return oc.Run("delete").Args("-f", fixture, "-n", namespace).Execute()
+}
+
+func updateImageConfig(oc *exutil.CLI, allowedRegistries []string) {
+	err := retry.RetryOnConflict(retry.DefaultBackoff, func() error {
+		imageConfig, err := oc.AdminConfigClient().ConfigV1().Images().Get(
+			context.Background(), "cluster", metav1.GetOptions{},
+		)
+		if err != nil {
+			return err
+		}
+		imageConfig.Spec.RegistrySources.AllowedRegistries = allowedRegistries
+		_, err = oc.AdminConfigClient().ConfigV1().Images().Update(
+			context.Background(), imageConfig, metav1.UpdateOptions{},
+		)
+		return err
+	})
+	o.Expect(err).NotTo(o.HaveOccurred(), "error updating image config")
+	time.Sleep(10 * time.Second)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, workerPool, registriesWorkerPoolMachineConfig)
+	machineconfighelper.WaitForConfigAndPoolComplete(oc, masterPool, registriesMasterPoolMachineConfig)
+}
+
+func cleanupImageConfig(oc *exutil.CLI) error {
+	return retry.RetryOnConflict(retry.DefaultBackoff, func() error {
+		imageConfig, err := oc.AdminConfigClient().ConfigV1().Images().Get(
+			context.Background(), "cluster", metav1.GetOptions{},
+		)
+		if err != nil {
+			return err
+		}
+		imageConfig.Spec.RegistrySources.AllowedRegistries = []string{}
+		_, err = oc.AdminConfigClient().ConfigV1().Images().Update(
+			context.Background(), imageConfig, metav1.UpdateOptions{},
+		)
+		return err
+	})
+}
+
+func launchTestPod(ctx context.Context, f *e2e.Framework, podName, image string) (*kapiv1.Pod, error) {
+	g.By(fmt.Sprintf("launching the pod: %s", podName))
+	contName := fmt.Sprintf("%s-container", podName)
+	pod := &kapiv1.Pod{
+		TypeMeta: metav1.TypeMeta{
+			Kind: "Pod",
+		},
+		ObjectMeta: metav1.ObjectMeta{
+			Name: podName,
+		},
+		Spec: kapiv1.PodSpec{
+			Containers: []kapiv1.Container{
+				{
+					Name:            contName,
+					Image:           image,
+					ImagePullPolicy: kapiv1.PullAlways,
+				},
+			},
+			RestartPolicy: kapiv1.RestartPolicyNever,
+		},
+	}
+	pod, err := f.ClientSet.CoreV1().Pods(f.Namespace.Name).Create(ctx, pod, metav1.CreateOptions{})
+	return pod, err
+}
+
+func deleteTestPod(ctx context.Context, f *e2e.Framework, podName string) error {
+	return f.ClientSet.CoreV1().Pods(f.Namespace.Name).Delete(ctx, podName, *metav1.NewDeleteOptions(0))
+}
+
+func waitForTestPodContainerToFailSignatureValidation(ctx context.Context, f *e2e.Framework, pod *kapiv1.Pod) error {
+	return e2epod.WaitForPodContainerToFail(ctx, f.ClientSet, pod.Namespace, pod.Name, 0, SignatureValidationFaildReason, e2e.PodStartShortTimeout)
+}

test/extended/include.go

@@ -32,6 +32,7 @@ import (
 	_ "github.com/openshift/origin/test/extended/idling"
 	_ "github.com/openshift/origin/test/extended/image_ecosystem"
 	_ "github.com/openshift/origin/test/extended/imageapis"
+	_ "github.com/openshift/origin/test/extended/imagepolicy"
 	_ "github.com/openshift/origin/test/extended/images"
 	_ "github.com/openshift/origin/test/extended/images/trigger"
 	_ "github.com/openshift/origin/test/extended/kernel"

test/extended/machine_config/helpers.go

@@ -292,3 +292,82 @@ func WaitForOneMasterNodeToBeReady(oc *exutil.CLI) error {
 	}, 5*time.Minute, 10*time.Second).Should(o.BeTrue())
 	return nil
 }
+
+// WaitForConfigAndPoolComplete is a helper function that gets a renderedConfig and waits for its pool to complete.
+// The return value is the final rendered config.
+func WaitForConfigAndPoolComplete(oc *exutil.CLI, pool, mcName string) string {
+	config, err := WaitForRenderedConfig(oc, pool, mcName)
+	o.Expect(err).NotTo(o.HaveOccurred(), fmt.Sprintf("%v: failed to render machine config %s from pool %s", err, mcName, pool))
+
+	err = WaitForPoolComplete(oc, pool, config)
+	o.Expect(err).NotTo(o.HaveOccurred(), fmt.Sprintf("%v: pool %s did not update to config %s", err, pool, config))
+	return config
+}
+
+// WaitForRenderedConfig polls a MachineConfigPool until it has
+// included the given mcName in its config, and returns the new
+// rendered config name.
+func WaitForRenderedConfig(oc *exutil.CLI, pool, mcName string) (string, error) {
+	return WaitForRenderedConfigs(oc, pool, mcName)
+}
+
+// WaitForRenderedConfigs polls a MachineConfigPool until it has
+// included the given mcNames in its config, and returns the new
+// rendered config name.
+func WaitForRenderedConfigs(oc *exutil.CLI, pool string, mcNames ...string) (string, error) {
+	var renderedConfig string
+	machineConfigClient, err := machineconfigclient.NewForConfig(oc.KubeFramework().ClientConfig())
+	o.Expect(err).NotTo(o.HaveOccurred())
+	found := make(map[string]bool)
+	o.Eventually(func() bool {
+		// Set up the list
+		for _, name := range mcNames {
+			found[name] = false
+		}
+
+		// Update found based on the MCP
+		mcp, err := machineConfigClient.MachineconfigurationV1().MachineConfigPools().Get(context.TODO(), pool, metav1.GetOptions{})
+		if err != nil {
+			return false
+		}
+		for _, mc := range mcp.Spec.Configuration.Source {
+			if _, ok := found[mc.Name]; ok {
+				found[mc.Name] = true
+			}
+		}
+
+		// If any are still false, then they weren't included in the MCP
+		for _, nameFound := range found {
+			if !nameFound {
+				return false
+			}
+		}
+
+		// All the required names were found
+		renderedConfig = mcp.Spec.Configuration.Name
+		return true
+	}, 5*time.Minute, 10*time.Second).Should(o.BeTrue())
+	return renderedConfig, nil
+}
+
+// WaitForPoolComplete polls a pool until it has completed an update to target
+func WaitForPoolComplete(oc *exutil.CLI, pool, target string) error {
+	machineConfigClient, err := machineconfigclient.NewForConfig(oc.KubeFramework().ClientConfig())
+	o.Expect(err).NotTo(o.HaveOccurred())
+	framework.Logf("Waiting for pool %s to complete %s", pool, target)
+	o.Eventually(func() bool {
+		mcp, err := machineConfigClient.MachineconfigurationV1().MachineConfigPools().Get(context.TODO(), pool, metav1.GetOptions{})
+		if err != nil {
+			framework.Logf("Failed to grab machineconfigpools, error :%v", err)
+			return false
+		}
+		if mcp.Status.Configuration.Name != target {
+			return false
+		}
+		if IsMachineConfigPoolConditionTrue(mcp.Status.Conditions, mcfgv1.MachineConfigPoolUpdated) {
+			return true
+		}
+		return false
+	}, 20*time.Minute, 10*time.Second).Should(o.BeTrue())
+	return nil
+}