test: avoid deadlock after failed command execution in pod

When executing the command inside the pod failed entirely (for
example, because the pod is not found), "tar cf" got stuck because its
stdout was kept open and "tar.Wait()" blocked forever.

test/e2e/pod/exec.go

@@ -19,7 +19,7 @@ import (
 // RunInPod optionally tars up some files or directories, unpacks them in a container,
 // and executes a shell command. Any error is treated as test failure.
 func RunInPod(f *framework.Framework, rootdir string, items []string, command string, namespace, pod, container string) (string, string) {
-	var input io.Reader
+	var input io.ReadCloser
 	var cmdPrefix string
 	if len(items) > 0 {
 		args := []string{"-cf", "-"}
@@ -54,6 +54,12 @@ func RunInPod(f *framework.Framework, rootdir string, items []string, command st
 		CaptureStderr: true,
 	}
 	stdout, stderr, err := f.ExecWithOptions(options)
+	if input != nil {
+		// Tell tar that it can stop writing. Necessary if ExecWithOptions did not consume
+		// all output of tar, because otherwise tar is stuck and tar.Wait() above will
+		// block.
+		input.Close()
+	}
 	framework.ExpectNoError(err, "command failed in namespace %s, pod/container %s/%s:\nstderr:\n%s\nstdout:%s\n",
 		namespace, pod, container, stderr, stdout)
 	fmt.Fprintf(GinkgoWriter, "stderr:\n%s\nstdout:\n%s\n",