188 lines
5.8 KiB
Go
188 lines
5.8 KiB
Go
/*
|
|
Copyright 2020 The Kubernetes Authors.
|
|
|
|
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.
|
|
*/
|
|
|
|
package netpol
|
|
|
|
import (
|
|
"fmt"
|
|
v1 "k8s.io/api/core/v1"
|
|
"k8s.io/kubernetes/test/e2e/framework"
|
|
"strings"
|
|
)
|
|
|
|
// TestCase describes the data for a netpol test
|
|
type TestCase struct {
|
|
ToPort int
|
|
Protocol v1.Protocol
|
|
Reachability *Reachability
|
|
}
|
|
|
|
// PodString represents a namespace 'x' + pod 'a' as "x/a".
|
|
type PodString string
|
|
|
|
// NewPodString instantiates a PodString from the given namespace and name.
|
|
func NewPodString(namespace string, podName string) PodString {
|
|
return PodString(fmt.Sprintf("%s/%s", namespace, podName))
|
|
}
|
|
|
|
// String converts back to a string
|
|
func (pod PodString) String() string {
|
|
return string(pod)
|
|
}
|
|
|
|
func (pod PodString) split() (string, string) {
|
|
pieces := strings.Split(string(pod), "/")
|
|
if len(pieces) != 2 {
|
|
framework.Failf("expected ns/pod, found %+v", pieces)
|
|
}
|
|
return pieces[0], pieces[1]
|
|
}
|
|
|
|
// Namespace extracts the namespace
|
|
func (pod PodString) Namespace() string {
|
|
ns, _ := pod.split()
|
|
return ns
|
|
}
|
|
|
|
// PodName extracts the pod name
|
|
func (pod PodString) PodName() string {
|
|
_, podName := pod.split()
|
|
return podName
|
|
}
|
|
|
|
// Peer is used for matching pods by either or both of the pod's namespace and name.
|
|
type Peer struct {
|
|
Namespace string
|
|
Pod string
|
|
}
|
|
|
|
// Matches checks whether the Peer matches the PodString:
|
|
// - an empty namespace means the namespace will always match
|
|
// - otherwise, the namespace must match the PodString's namespace
|
|
// - same goes for Pod: empty matches everything, otherwise must match exactly
|
|
func (p *Peer) Matches(pod PodString) bool {
|
|
return (p.Namespace == "" || p.Namespace == pod.Namespace()) && (p.Pod == "" || p.Pod == pod.PodName())
|
|
}
|
|
|
|
// Reachability packages the data for a cluster-wide connectivity probe
|
|
type Reachability struct {
|
|
Expected *TruthTable
|
|
Observed *TruthTable
|
|
Pods []*Pod
|
|
}
|
|
|
|
// NewReachability instantiates a reachability
|
|
func NewReachability(pods []*Pod, defaultExpectation bool) *Reachability {
|
|
var podNames []string
|
|
for _, pod := range pods {
|
|
podNames = append(podNames, pod.PodString().String())
|
|
}
|
|
r := &Reachability{
|
|
Expected: NewTruthTableFromItems(podNames, &defaultExpectation),
|
|
Observed: NewTruthTableFromItems(podNames, nil),
|
|
Pods: pods,
|
|
}
|
|
return r
|
|
}
|
|
|
|
// AllowLoopback expects all communication from a pod to itself to be allowed.
|
|
// In general, call it after setting up any other rules since loopback logic follows no policy.
|
|
func (r *Reachability) AllowLoopback() {
|
|
for _, pod := range r.Pods {
|
|
podName := pod.PodString().String()
|
|
r.Expected.Set(podName, podName, true)
|
|
}
|
|
}
|
|
|
|
// Expect sets the expected value for a single observation
|
|
func (r *Reachability) Expect(from PodString, to PodString, isConnected bool) {
|
|
r.Expected.Set(string(from), string(to), isConnected)
|
|
}
|
|
|
|
// ExpectAllIngress defines that any traffic going into the pod will be allowed/denied (true/false)
|
|
func (r *Reachability) ExpectAllIngress(pod PodString, connected bool) {
|
|
r.Expected.SetAllTo(string(pod), connected)
|
|
if !connected {
|
|
framework.Logf("Denying all traffic *to* %s", pod)
|
|
}
|
|
}
|
|
|
|
// ExpectAllEgress defines that any traffic going out of the pod will be allowed/denied (true/false)
|
|
func (r *Reachability) ExpectAllEgress(pod PodString, connected bool) {
|
|
r.Expected.SetAllFrom(string(pod), connected)
|
|
if !connected {
|
|
framework.Logf("Denying all traffic *from* %s", pod)
|
|
}
|
|
}
|
|
|
|
// ExpectPeer sets expected values using Peer matchers
|
|
func (r *Reachability) ExpectPeer(from *Peer, to *Peer, connected bool) {
|
|
for _, fromPod := range r.Pods {
|
|
if from.Matches(fromPod.PodString()) {
|
|
for _, toPod := range r.Pods {
|
|
if to.Matches(toPod.PodString()) {
|
|
r.Expected.Set(string(fromPod.PodString()), string(toPod.PodString()), connected)
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Observe records a single connectivity observation
|
|
func (r *Reachability) Observe(fromPod PodString, toPod PodString, isConnected bool) {
|
|
r.Observed.Set(string(fromPod), string(toPod), isConnected)
|
|
}
|
|
|
|
// Summary produces a useful summary of expected and observed data
|
|
func (r *Reachability) Summary(ignoreLoopback bool) (trueObs int, falseObs int, ignoredObs int, comparison *TruthTable) {
|
|
comparison = r.Expected.Compare(r.Observed)
|
|
if !comparison.IsComplete() {
|
|
framework.Failf("observations not complete!")
|
|
}
|
|
falseObs, trueObs, ignoredObs = 0, 0, 0
|
|
for from, dict := range comparison.Values {
|
|
for to, val := range dict {
|
|
if ignoreLoopback && from == to {
|
|
// Never fail on loopback, because its not yet defined.
|
|
ignoredObs++
|
|
} else if val {
|
|
trueObs++
|
|
} else {
|
|
falseObs++
|
|
}
|
|
}
|
|
}
|
|
return
|
|
}
|
|
|
|
// PrintSummary prints the summary
|
|
func (r *Reachability) PrintSummary(printExpected bool, printObserved bool, printComparison bool) {
|
|
right, wrong, ignored, comparison := r.Summary(ignoreLoopback)
|
|
if ignored > 0 {
|
|
framework.Logf("warning: this test doesn't take into consideration hairpin traffic, i.e. traffic whose source and destination is the same pod: %d cases ignored", ignored)
|
|
}
|
|
framework.Logf("reachability: correct:%v, incorrect:%v, result=%t\n\n", right, wrong, wrong == 0)
|
|
if printExpected {
|
|
framework.Logf("expected:\n\n%s\n\n\n", r.Expected.PrettyPrint(""))
|
|
}
|
|
if printObserved {
|
|
framework.Logf("observed:\n\n%s\n\n\n", r.Observed.PrettyPrint(""))
|
|
}
|
|
if printComparison {
|
|
framework.Logf("comparison:\n\n%s\n\n\n", comparison.PrettyPrint(""))
|
|
}
|
|
}
|