Don't include user data in CRI streaming redirect URLs

pkg/kubelet/server/server.go

@@ -606,7 +606,7 @@ func (s *Server) getAttach(request *restful.Request, response *restful.Response)
 	podFullName := kubecontainer.GetPodFullName(pod)
 	redirect, err := s.host.GetAttach(podFullName, params.podUID, params.containerName, *streamOpts)
 	if err != nil {
-		response.WriteError(streaming.HTTPStatus(err), err)
+		streaming.WriteError(err, response.ResponseWriter)
 		return
 	}
 	if redirect != nil {
@@ -644,7 +644,7 @@ func (s *Server) getExec(request *restful.Request, response *restful.Response) {
 	podFullName := kubecontainer.GetPodFullName(pod)
 	redirect, err := s.host.GetExec(podFullName, params.podUID, params.containerName, params.cmd, *streamOpts)
 	if err != nil {
-		response.WriteError(streaming.HTTPStatus(err), err)
+		streaming.WriteError(err, response.ResponseWriter)
 		return
 	}
 	if redirect != nil {
@@ -714,7 +714,7 @@ func (s *Server) getPortForward(request *restful.Request, response *restful.Resp
 
 	redirect, err := s.host.GetPortForward(pod.Name, pod.Namespace, pod.UID)
 	if err != nil {
-		response.WriteError(streaming.HTTPStatus(err), err)
+		streaming.WriteError(err, response.ResponseWriter)
 		return
 	}
 	if redirect != nil {

pkg/kubelet/server/streaming/BUILD

@@ -12,14 +12,15 @@ go_library(
     name = "go_default_library",
     srcs = [
         "errors.go",
+        "request_cache.go",
         "server.go",
     ],
     tags = ["automanaged"],
     deps = [
-        "//pkg/api:go_default_library",
         "//pkg/kubelet/api/v1alpha1/runtime:go_default_library",
         "//pkg/kubelet/server/portforward:go_default_library",
         "//pkg/kubelet/server/remotecommand:go_default_library",
+        "//pkg/util/clock:go_default_library",
         "//pkg/util/term:go_default_library",
         "//vendor:github.com/emicklei/go-restful",
         "//vendor:google.golang.org/grpc",
@@ -30,7 +31,10 @@ go_library(
 
 go_test(
     name = "go_default_test",
-    srcs = ["server_test.go"],
+    srcs = [
+        "request_cache_test.go",
+        "server_test.go",
+    ],
     library = ":go_default_library",
     tags = ["automanaged"],
     deps = [
@@ -43,6 +47,7 @@ go_test(
         "//vendor:github.com/stretchr/testify/assert",
         "//vendor:github.com/stretchr/testify/require",
         "//vendor:k8s.io/client-go/pkg/api",
+        "//vendor:k8s.io/client-go/pkg/util/clock",
     ],
 )
 

pkg/kubelet/server/streaming/errors.go

@@ -19,6 +19,7 @@ package streaming
 import (
 	"fmt"
 	"net/http"
+	"strconv"
 	"time"
 
 	"google.golang.org/grpc"
@@ -33,12 +34,27 @@ func ErrorTimeout(op string, timeout time.Duration) error {
 	return grpc.Errorf(codes.DeadlineExceeded, fmt.Sprintf("%s timed out after %s", op, timeout.String()))
 }
 
-// Translates a CRI streaming error into an HTTP status code.
-func HTTPStatus(err error) int {
+// The error returned when the maximum number of in-flight requests is exceeded.
+func ErrorTooManyInFlight() error {
+	return grpc.Errorf(codes.ResourceExhausted, "maximum number of in-flight requests exceeded")
+}
+
+// Translates a CRI streaming error into an appropriate HTTP response.
+func WriteError(err error, w http.ResponseWriter) error {
+	var status int
 	switch grpc.Code(err) {
 	case codes.NotFound:
-		return http.StatusNotFound
+		status = http.StatusNotFound
+	case codes.ResourceExhausted:
+		// We only expect to hit this if there is a DoS, so we just wait the full TTL.
+		// If this is ever hit in steady-state operations, consider increasing the MaxInFlight requests,
+		// or plumbing through the time to next expiration.
+		w.Header().Set("Retry-After", strconv.Itoa(int(CacheTTL.Seconds())))
+		status = http.StatusTooManyRequests
 	default:
-		return http.StatusInternalServerError
+		status = http.StatusInternalServerError
 	}
+	w.WriteHeader(status)
+	_, writeErr := w.Write([]byte(err.Error()))
+	return writeErr
 }

pkg/kubelet/server/streaming/request_cache.go

+/*
+Copyright 2016 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 streaming
+
+import (
+	"container/list"
+	"crypto/rand"
+	"encoding/base64"
+	"fmt"
+	"math"
+	"sync"
+	"time"
+
+	"k8s.io/kubernetes/pkg/util/clock"
+)
+
+var (
+	// Timeout after which tokens become invalid.
+	CacheTTL = 1 * time.Minute
+	// The maximum number of in-flight requests to allow.
+	MaxInFlight = 1000
+	// Length of the random base64 encoded token identifying the request.
+	TokenLen = 8
+)
+
+// requestCache caches streaming (exec/attach/port-forward) requests and generates a single-use
+// random token for their retrieval. The requestCache is used for building streaming URLs without
+// the need to encode every request parameter in the URL.
+type requestCache struct {
+	// clock is used to obtain the current time
+	clock clock.Clock
+
+	// tokens maps the generate token to the request for fast retrieval.
+	tokens map[string]*list.Element
+	// ll maintains an age-ordered request list for faster garbage collection of expired requests.
+	ll *list.List
+
+	lock sync.Mutex
+}
+
+// Type representing an *ExecRequest, *AttachRequest, or *PortForwardRequest.
+type request interface{}
+
+type cacheEntry struct {
+	token      string
+	req        request
+	expireTime time.Time
+}
+
+func newRequestCache() *requestCache {
+	return &requestCache{
+		clock:  clock.RealClock{},
+		ll:     list.New(),
+		tokens: make(map[string]*list.Element),
+	}
+}
+
+// Insert the given request into the cache and returns the token used for fetching it out.
+func (c *requestCache) Insert(req request) (token string, err error) {
+	c.lock.Lock()
+	defer c.lock.Unlock()
+
+	// Remove expired entries.
+	c.gc()
+	// If the cache is full, reject the request.
+	if c.ll.Len() == MaxInFlight {
+		return "", ErrorTooManyInFlight()
+	}
+	token, err = c.uniqueToken()
+	if err != nil {
+		return "", err
+	}
+	ele := c.ll.PushFront(&cacheEntry{token, req, c.clock.Now().Add(CacheTTL)})
+
+	c.tokens[token] = ele
+	return token, nil
+}
+
+// Consume the token (remove it from the cache) and return the cached request, if found.
+func (c *requestCache) Consume(token string) (req request, found bool) {
+	c.lock.Lock()
+	defer c.lock.Unlock()
+	ele, ok := c.tokens[token]
+	if !ok {
+		return nil, false
+	}
+	c.ll.Remove(ele)
+	delete(c.tokens, token)
+
+	entry := ele.Value.(*cacheEntry)
+	if c.clock.Now().After(entry.expireTime) {
+		// Entry already expired.
+		return nil, false
+	}
+	return entry.req, true
+}
+
+// uniqueToken generates a random URL-safe token and ensures uniqueness.
+func (c *requestCache) uniqueToken() (string, error) {
+	const maxTries = 10
+	// Number of bytes to be TokenLen when base64 encoded.
+	tokenSize := math.Ceil(float64(TokenLen) * 6 / 8)
+	rawToken := make([]byte, int(tokenSize))
+	for i := 0; i < maxTries; i++ {
+		if _, err := rand.Read(rawToken); err != nil {
+			return "", err
+		}
+		encoded := base64.RawURLEncoding.EncodeToString(rawToken)
+		token := encoded[:TokenLen]
+		// If it's unique, return it. Otherwise retry.
+		if _, exists := c.tokens[encoded]; !exists {
+			return token, nil
+		}
+	}
+	return "", fmt.Errorf("failed to generate unique token")
+}
+
+// Must be write-locked prior to calling.
+func (c *requestCache) gc() {
+	now := c.clock.Now()
+	for c.ll.Len() > 0 {
+		oldest := c.ll.Back()
+		entry := oldest.Value.(*cacheEntry)
+		if !now.After(entry.expireTime) {
+			return
+		}
+
+		// Oldest value is expired; remove it.
+		c.ll.Remove(oldest)
+		delete(c.tokens, entry.token)
+	}
+}

pkg/kubelet/server/streaming/request_cache_test.go

+/*
+Copyright 2016 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 streaming
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"k8s.io/client-go/pkg/util/clock"
+)
+
+func TestInsert(t *testing.T) {
+	c, _ := newTestCache()
+
+	// Insert normal
+	oldestTok, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assert.Len(t, oldestTok, TokenLen)
+	assertCacheSize(t, c, 1)
+
+	// Insert until full
+	for i := 0; i < MaxInFlight-2; i++ {
+		tok, err := c.Insert(nextRequest())
+		require.NoError(t, err)
+		assert.Len(t, tok, TokenLen)
+	}
+	assertCacheSize(t, c, MaxInFlight-1)
+
+	newestReq := nextRequest()
+	newestTok, err := c.Insert(newestReq)
+	require.NoError(t, err)
+	assert.Len(t, newestTok, TokenLen)
+	assertCacheSize(t, c, MaxInFlight)
+	require.Contains(t, c.tokens, oldestTok, "oldest request should still be cached")
+
+	// Consume newest token.
+	req, ok := c.Consume(newestTok)
+	assert.True(t, ok, "newest request should still be cached")
+	assert.Equal(t, newestReq, req)
+	require.Contains(t, c.tokens, oldestTok, "oldest request should still be cached")
+
+	// Insert again (still full)
+	tok, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assert.Len(t, tok, TokenLen)
+	assertCacheSize(t, c, MaxInFlight)
+
+	// Insert again (should evict)
+	_, err = c.Insert(nextRequest())
+	assert.Error(t, err, "should reject further requests")
+	errResponse := httptest.NewRecorder()
+	require.NoError(t, WriteError(err, errResponse))
+	assert.Equal(t, errResponse.Code, http.StatusTooManyRequests)
+	assert.Equal(t, strconv.Itoa(int(CacheTTL.Seconds())), errResponse.HeaderMap.Get("Retry-After"))
+
+	assertCacheSize(t, c, MaxInFlight)
+	_, ok = c.Consume(oldestTok)
+	assert.True(t, ok, "oldest request should be valid")
+}
+
+func TestConsume(t *testing.T) {
+	c, clock := newTestCache()
+
+	{ // Insert & consume.
+		req := nextRequest()
+		tok, err := c.Insert(req)
+		require.NoError(t, err)
+		assertCacheSize(t, c, 1)
+
+		cachedReq, ok := c.Consume(tok)
+		assert.True(t, ok)
+		assert.Equal(t, req, cachedReq)
+		assertCacheSize(t, c, 0)
+	}
+
+	{ // Insert & consume out of order
+		req1 := nextRequest()
+		tok1, err := c.Insert(req1)
+		require.NoError(t, err)
+		assertCacheSize(t, c, 1)
+
+		req2 := nextRequest()
+		tok2, err := c.Insert(req2)
+		require.NoError(t, err)
+		assertCacheSize(t, c, 2)
+
+		cachedReq2, ok := c.Consume(tok2)
+		assert.True(t, ok)
+		assert.Equal(t, req2, cachedReq2)
+		assertCacheSize(t, c, 1)
+
+		cachedReq1, ok := c.Consume(tok1)
+		assert.True(t, ok)
+		assert.Equal(t, req1, cachedReq1)
+		assertCacheSize(t, c, 0)
+	}
+
+	{ // Consume a second time
+		req := nextRequest()
+		tok, err := c.Insert(req)
+		require.NoError(t, err)
+		assertCacheSize(t, c, 1)
+
+		cachedReq, ok := c.Consume(tok)
+		assert.True(t, ok)
+		assert.Equal(t, req, cachedReq)
+		assertCacheSize(t, c, 0)
+
+		_, ok = c.Consume(tok)
+		assert.False(t, ok)
+		assertCacheSize(t, c, 0)
+	}
+
+	{ // Consume without insert
+		_, ok := c.Consume("fooBAR")
+		assert.False(t, ok)
+		assertCacheSize(t, c, 0)
+	}
+
+	{ // Consume expired
+		tok, err := c.Insert(nextRequest())
+		require.NoError(t, err)
+		assertCacheSize(t, c, 1)
+
+		clock.Step(2 * CacheTTL)
+
+		_, ok := c.Consume(tok)
+		assert.False(t, ok)
+		assertCacheSize(t, c, 0)
+	}
+}
+
+func TestGC(t *testing.T) {
+	c, clock := newTestCache()
+
+	// When empty
+	c.gc()
+	assertCacheSize(t, c, 0)
+
+	tok1, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assertCacheSize(t, c, 1)
+	clock.Step(10 * time.Second)
+	tok2, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assertCacheSize(t, c, 2)
+
+	// expired: tok1, tok2
+	// non-expired: tok3, tok4
+	clock.Step(2 * CacheTTL)
+	tok3, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assertCacheSize(t, c, 1)
+	clock.Step(10 * time.Second)
+	tok4, err := c.Insert(nextRequest())
+	require.NoError(t, err)
+	assertCacheSize(t, c, 2)
+
+	_, ok := c.Consume(tok1)
+	assert.False(t, ok)
+	_, ok = c.Consume(tok2)
+	assert.False(t, ok)
+	_, ok = c.Consume(tok3)
+	assert.True(t, ok)
+	_, ok = c.Consume(tok4)
+	assert.True(t, ok)
+
+	// When full, nothing is expired.
+	for i := 0; i < MaxInFlight; i++ {
+		_, err := c.Insert(nextRequest())
+		require.NoError(t, err)
+	}
+	assertCacheSize(t, c, MaxInFlight)
+
+	// When everything is expired
+	clock.Step(2 * CacheTTL)
+	_, err = c.Insert(nextRequest())
+	require.NoError(t, err)
+	assertCacheSize(t, c, 1)
+}
+
+func newTestCache() (*requestCache, *clock.FakeClock) {
+	c := newRequestCache()
+	fakeClock := clock.NewFakeClock(time.Now())
+	c.clock = fakeClock
+	return c, fakeClock
+}
+
+func assertCacheSize(t *testing.T, cache *requestCache, expectedSize int) {
+	tokenLen := len(cache.tokens)
+	llLen := cache.ll.Len()
+	assert.Equal(t, tokenLen, llLen, "inconsistent cache size! len(tokens)=%d; len(ll)=%d", tokenLen, llLen)
+	assert.Equal(t, expectedSize, tokenLen, "unexpected cache size!")
+}
+
+var requestUID = 0
+
+func nextRequest() interface{} {
+	requestUID++
+	return requestUID
+}

pkg/kubelet/server/streaming/server.go

@@ -25,10 +25,12 @@ import (
 	"path"
 	"time"
 
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/codes"
+
 	restful "github.com/emicklei/go-restful"
 
 	"k8s.io/apimachinery/pkg/types"
-	"k8s.io/kubernetes/pkg/api"
 	runtimeapi "k8s.io/kubernetes/pkg/kubelet/api/v1alpha1/runtime"
 	"k8s.io/kubernetes/pkg/kubelet/server/portforward"
 	"k8s.io/kubernetes/pkg/kubelet/server/remotecommand"
@@ -97,6 +99,7 @@ func NewServer(config Config, runtime Runtime) (Server, error) {
 	s := &server{
 		config:  config,
 		runtime: &criAdapter{runtime},
+		cache:   newRequestCache(),
 	}
 
 	if s.config.BaseURL == nil {
@@ -114,9 +117,9 @@ func NewServer(config Config, runtime Runtime) (Server, error) {
 		path    string
 		handler restful.RouteFunction
 	}{
-		{"/exec/{containerID}", s.serveExec},
-		{"/attach/{containerID}", s.serveAttach},
-		{"/portforward/{podSandboxID}", s.servePortForward},
+		{"/exec/{token}", s.serveExec},
+		{"/attach/{token}", s.serveAttach},
+		{"/portforward/{token}", s.servePortForward},
 	}
 	// If serving relative to a base path, set that here.
 	pathPrefix := path.Dir(s.config.BaseURL.Path)
@@ -139,37 +142,45 @@ type server struct {
 	config  Config
 	runtime *criAdapter
 	handler http.Handler
+	cache   *requestCache
 }
 
 func (s *server) GetExec(req *runtimeapi.ExecRequest) (*runtimeapi.ExecResponse, error) {
-	url := s.buildURL("exec", req.GetContainerId(), streamOpts{
-		stdin:   req.GetStdin(),
-		stdout:  true,
-		stderr:  !req.GetTty(), // For TTY connections, both stderr is combined with stdout.
-		tty:     req.GetTty(),
-		command: req.GetCmd(),
-	})
+	if req.GetContainerId() == "" {
+		return nil, grpc.Errorf(codes.InvalidArgument, "missing required container_id")
+	}
+	token, err := s.cache.Insert(req)
+	if err != nil {
+		return nil, err
+	}
 	return &runtimeapi.ExecResponse{
-		Url: &url,
+		Url: s.buildURL("exec", token),
 	}, nil
 }
 
 func (s *server) GetAttach(req *runtimeapi.AttachRequest) (*runtimeapi.AttachResponse, error) {
-	url := s.buildURL("attach", req.GetContainerId(), streamOpts{
-		stdin:  req.GetStdin(),
-		stdout: true,
-		stderr: !req.GetTty(), // For TTY connections, both stderr is combined with stdout.
-		tty:    req.GetTty(),
-	})
+	if req.GetContainerId() == "" {
+		return nil, grpc.Errorf(codes.InvalidArgument, "missing required container_id")
+	}
+	token, err := s.cache.Insert(req)
+	if err != nil {
+		return nil, err
+	}
 	return &runtimeapi.AttachResponse{
-		Url: &url,
+		Url: s.buildURL("attach", token),
 	}, nil
 }
 
 func (s *server) GetPortForward(req *runtimeapi.PortForwardRequest) (*runtimeapi.PortForwardResponse, error) {
-	url := s.buildURL("portforward", req.GetPodSandboxId(), streamOpts{})
+	if req.GetPodSandboxId() == "" {
+		return nil, grpc.Errorf(codes.InvalidArgument, "missing required pod_sandbox_id")
+	}
+	token, err := s.cache.Insert(req)
+	if err != nil {
+		return nil, err
+	}
 	return &runtimeapi.PortForwardResponse{
-		Url: &url,
+		Url: s.buildURL("portforward", token),
 	}, nil
 }
 
@@ -200,63 +211,32 @@ func (s *server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	s.handler.ServeHTTP(w, r)
 }
 
-type streamOpts struct {
-	stdin  bool
-	stdout bool
-	stderr bool
-	tty    bool
-
-	command []string
-	port    []int32
-}
-
-const (
-	urlParamStdin   = api.ExecStdinParam
-	urlParamStdout  = api.ExecStdoutParam
-	urlParamStderr  = api.ExecStderrParam
-	urlParamTTY     = api.ExecTTYParam
-	urlParamCommand = api.ExecCommandParamm
-)
-
-func (s *server) buildURL(method, id string, opts streamOpts) string {
-	loc := &url.URL{
-		Path: path.Join(method, id),
-	}
-
-	query := url.Values{}
-	if opts.stdin {
-		query.Add(urlParamStdin, "1")
-	}
-	if opts.stdout {
-		query.Add(urlParamStdout, "1")
-	}
-	if opts.stderr {
-		query.Add(urlParamStderr, "1")
-	}
-	if opts.tty {
-		query.Add(urlParamTTY, "1")
-	}
-	for _, c := range opts.command {
-		query.Add(urlParamCommand, c)
-	}
-	loc.RawQuery = query.Encode()
-
-	return s.config.BaseURL.ResolveReference(loc).String()
+func (s *server) buildURL(method, token string) *string {
+	loc := s.config.BaseURL.ResolveReference(&url.URL{
+		Path: path.Join(method, token),
+	}).String()
+	return &loc
 }
 
 func (s *server) serveExec(req *restful.Request, resp *restful.Response) {
-	containerID := req.PathParameter("containerID")
-	if containerID == "" {
-		resp.WriteError(http.StatusBadRequest, errors.New("missing required containerID path parameter"))
+	token := req.PathParameter("token")
+	cachedRequest, ok := s.cache.Consume(token)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
 		return
 	}
-
-	streamOpts, err := remotecommand.NewOptions(req.Request)
-	if err != nil {
-		resp.WriteError(http.StatusBadRequest, err)
+	exec, ok := cachedRequest.(*runtimeapi.ExecRequest)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
 		return
 	}
-	cmd := req.Request.URL.Query()[api.ExecCommandParamm]
+
+	streamOpts := &remotecommand.Options{
+		Stdin:  exec.GetStdin(),
+		Stdout: true,
+		Stderr: !exec.GetTty(),
+		TTY:    exec.GetTty(),
+	}
 
 	remotecommand.ServeExec(
 		resp.ResponseWriter,
@@ -264,8 +244,8 @@ func (s *server) serveExec(req *restful.Request, resp *restful.Response) {
 		s.runtime,
 		"", // unused: podName
 		"", // unusued: podUID
-		containerID,
-		cmd,
+		exec.GetContainerId(),
+		exec.GetCmd(),
 		streamOpts,
 		s.config.StreamIdleTimeout,
 		s.config.StreamCreationTimeout,
@@ -273,25 +253,31 @@ func (s *server) serveExec(req *restful.Request, resp *restful.Response) {
 }
 
 func (s *server) serveAttach(req *restful.Request, resp *restful.Response) {
-	containerID := req.PathParameter("containerID")
-	if containerID == "" {
-		resp.WriteError(http.StatusBadRequest, errors.New("missing required containerID path parameter"))
+	token := req.PathParameter("token")
+	cachedRequest, ok := s.cache.Consume(token)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
 		return
 	}
-
-	streamOpts, err := remotecommand.NewOptions(req.Request)
-	if err != nil {
-		resp.WriteError(http.StatusBadRequest, err)
+	attach, ok := cachedRequest.(*runtimeapi.AttachRequest)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
 		return
 	}
 
+	streamOpts := &remotecommand.Options{
+		Stdin:  attach.GetStdin(),
+		Stdout: true,
+		Stderr: !attach.GetTty(),
+		TTY:    attach.GetTty(),
+	}
 	remotecommand.ServeAttach(
 		resp.ResponseWriter,
 		req.Request,
 		s.runtime,
 		"", // unused: podName
 		"", // unusued: podUID
-		containerID,
+		attach.GetContainerId(),
 		streamOpts,
 		s.config.StreamIdleTimeout,
 		s.config.StreamCreationTimeout,
@@ -299,9 +285,15 @@ func (s *server) serveAttach(req *restful.Request, resp *restful.Response) {
 }
 
 func (s *server) servePortForward(req *restful.Request, resp *restful.Response) {
-	podSandboxID := req.PathParameter("podSandboxID")
-	if podSandboxID == "" {
-		resp.WriteError(http.StatusBadRequest, errors.New("missing required podSandboxID path parameter"))
+	token := req.PathParameter("token")
+	cachedRequest, ok := s.cache.Consume(token)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
+		return
+	}
+	pf, ok := cachedRequest.(*runtimeapi.PortForwardRequest)
+	if !ok {
+		http.NotFound(resp.ResponseWriter, req.Request)
 		return
 	}
 
@@ -309,7 +301,7 @@ func (s *server) servePortForward(req *restful.Request, resp *restful.Response)
 		resp.ResponseWriter,
 		req.Request,
 		s.runtime,
-		podSandboxID,
+		pf.GetPodSandboxId(),
 		"", // unused: podUID
 		s.config.StreamIdleTimeout,
 		s.config.StreamCreationTimeout)

test/e2e/kubectl.go

@@ -391,6 +391,14 @@ var _ = framework.KubeDescribe("Kubectl client", func() {
 				framework.Failf("Unexpected kubectl exec output. Wanted %q, got %q", e, a)
 			}
 
+			By("executing a very long command in the container")
+			veryLongData := make([]rune, 20000)
+			for i := 0; i < len(veryLongData); i++ {
+				veryLongData[i] = 'a'
+			}
+			execOutput = framework.RunKubectlOrDie("exec", fmt.Sprintf("--namespace=%v", ns), simplePodName, "echo", string(veryLongData))
+			Expect(string(veryLongData)).To(Equal(strings.TrimSpace(execOutput)), "Unexpected kubectl exec output")
+
 			By("executing a command in the container with noninteractive stdin")
 			execOutput = framework.NewKubectlCommand("exec", fmt.Sprintf("--namespace=%v", ns), "-i", simplePodName, "cat").
 				WithStdinData("abcd1234").