[Bugfix] Install Multicast related iptables rules only on IPv4 cluster

Add a pre-check on the Multicast feature gate status with IPv6-only cluster
settings in agent Initializer, and install the iptables rules only in the IPv4
related chains.

Signed-off-by: Wenying Dong <wenyingd@vmware.com>

cmd/antrea-agent/agent.go

@@ -818,7 +818,9 @@ func run(o *Options) error {
 			validator,
 			networkConfig.TrafficEncapMode.SupportsEncap(),
 			nodeInformer,
-			enableBridgingMode)
+			enableBridgingMode,
+			v4Enabled,
+			v6Enabled)
 		if err := mcastController.Initialize(); err != nil {
 			return err
 		}

pkg/agent/multicast/mcast_controller.go

@@ -15,6 +15,7 @@
 package multicast
 
 import (
+	"fmt"
 	"net"
 	"sync"
 	"time"
@@ -265,6 +266,13 @@ type Controller struct {
 	installedNodes      sets.Set[string]
 	encapEnabled        bool
 	flexibleIPAMEnabled bool
+	// ipv4Enabled is the flag that if it is running on IPv4 cluster. An error is returned if IPv4Enabled is false
+	// in Initialize as Multicast does not support IPv6 for now.
+	// TODO: remove this flag after IPv6 is supported in Multicast.
+	ipv4Enabled bool
+	// ipv6Enabled is the flag that if it is running on IPv6 cluster.
+	// TODO: remove this flag after IPv6 is supported in Multicast.
+	ipv6Enabled bool
 }
 
 func NewMulticastController(ofClient openflow.Client,
@@ -279,7 +287,9 @@ func NewMulticastController(ofClient openflow.Client,
 	validator types.McastNetworkPolicyController,
 	isEncap bool,
 	nodeInformer coreinformers.NodeInformer,
-	enableFlexibleIPAM bool) *Controller {
+	enableFlexibleIPAM bool,
+	ipv4Enabled bool,
+	ipv6Enabled bool) *Controller {
 	eventCh := make(chan *mcastGroupEvent, workerCount)
 	groupSnooper := newSnooper(ofClient, ifaceStore, eventCh, igmpQueryInterval, igmpQueryVersions, validator, isEncap)
 	groupCache := cache.NewIndexer(getGroupEventKey, cache.Indexers{
@@ -303,6 +313,8 @@ func NewMulticastController(ofClient openflow.Client,
 		queryGroupId:         v4GroupAllocator.Allocate(),
 		encapEnabled:         isEncap,
 		flexibleIPAMEnabled:  enableFlexibleIPAM,
+		ipv4Enabled:          ipv4Enabled,
+		ipv6Enabled:          ipv6Enabled,
 	}
 	if isEncap {
 		c.nodeGroupID = v4GroupAllocator.Allocate()
@@ -331,6 +343,11 @@ func NewMulticastController(ofClient openflow.Client,
 }
 
 func (c *Controller) Initialize() error {
+	if !c.ipv4Enabled {
+		return fmt.Errorf("Multicast is not supported on an IPv6-only cluster")
+	} else if c.ipv6Enabled {
+		klog.InfoS("Multicast only works with IPv4 traffic on a dual-stack cluster")
+	}
 	err := c.mRouteClient.Initialize()
 	if err != nil {
 		return err

pkg/agent/multicast/mcast_controller_test.go

@@ -1255,7 +1255,7 @@ func newMockMulticastController(t *testing.T, isEncap bool, enableFlexibleIPAM b
 	clientset = fake.NewSimpleClientset()
 	informerFactory = informers.NewSharedInformerFactory(clientset, 12*time.Hour)
 	nodeInformer := informerFactory.Core().V1().Nodes()
-	mctrl := NewMulticastController(mockOFClient, groupAllocator, nodeConfig, mockIfaceStore, mockMulticastSocket, sets.New[string](), podUpdateSubscriber, time.Second*5, []uint8{1, 2, 3}, mockMulticastValidator, isEncap, nodeInformer, enableFlexibleIPAM)
+	mctrl := NewMulticastController(mockOFClient, groupAllocator, nodeConfig, mockIfaceStore, mockMulticastSocket, sets.New[string](), podUpdateSubscriber, time.Second*5, []uint8{1, 2, 3}, mockMulticastValidator, isEncap, nodeInformer, enableFlexibleIPAM, true, false)
 	return mctrl
 }
 
@@ -1265,7 +1265,48 @@ func TestFlexibleIPAMModeInitialize(t *testing.T) {
 	assert.NoError(t, err)
 }
 
+func TestMulticastControllerOnIPv6Cluster(t *testing.T) {
+	for _, tc := range []struct {
+		name        string
+		ipv4Enabled bool
+		ipv6Enabled bool
+		expErr      string
+	}{
+		{
+			name:        "Fails on IPv6-only cluster",
+			ipv4Enabled: false,
+			ipv6Enabled: true,
+			expErr:      "Multicast is not supported on the IPv6-only cluster",
+		},
+		{
+			name:        "Succeeds on dual-stack cluster",
+			ipv4Enabled: true,
+			ipv6Enabled: true,
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			mockController := newMockMulticastController(t, true, false)
+			mockController.ipv4Enabled = tc.ipv4Enabled
+			mockController.ipv6Enabled = tc.ipv6Enabled
+			if tc.expErr == "" {
+				mockController.initMocks()
+			}
+			err := mockController.Initialize()
+			if tc.expErr != "" {
+				assert.EqualError(t, err, tc.expErr)
+			} else {
+				assert.NoError(t, err)
+			}
+		})
+	}
+}
+
 func (c *Controller) initialize(t *testing.T) error {
+	c.initMocks()
+	return c.Initialize()
+}
+
+func (c *Controller) initMocks() {
 	mockOFClient.EXPECT().InstallMulticastGroup(c.queryGroupId, gomock.Any(), gomock.Any()).Times(1)
 	mockOFClient.EXPECT().InstallMulticastFlows(gomock.Any(), gomock.Any())
 	mockIfaceStore.EXPECT().GetInterfacesByType(interfacestore.InterfaceType(0)).Times(1).Return([]*interfacestore.InterfaceConfig{})
@@ -1278,7 +1319,6 @@ func (c *Controller) initialize(t *testing.T) error {
 	if c.flexibleIPAMEnabled {
 		mockOFClient.EXPECT().InstallMulticastFlexibleIPAMFlows().Times(1)
 	}
-	return c.Initialize()
 }
 
 func createInterface(name string, ofport uint32) *interfacestore.InterfaceConfig {

pkg/agent/route/route_linux.go

@@ -128,7 +128,8 @@ type Client struct {
 	nodePortsIPv6 sync.Map
 	// clusterNodeIPs stores the IPv4 of all other Nodes in the cluster
 	clusterNodeIPs sync.Map
-	// clusterNodeIP6s stores the IPv6 of all other Nodes in the cluster
+	// clusterNodeIP6s stores the IPv6 address of all other Nodes in the cluster. It is maintained but not consumed
+	// since Multicast supports IPv6.
 	clusterNodeIP6s sync.Map
 	// egressRoutes caches ip routes about Egresses.
 	egressRoutes sync.Map
@@ -709,7 +710,9 @@ func (c *Client) restoreIptablesData(podCIDR *net.IPNet,
 			}...)
 		}
 
-		if c.multicastEnabled && c.networkConfig.TrafficEncapMode.SupportsEncap() {
+		// Note: Multicast can only work with IPv4 for now. Remove condition "!isIPv6" in the future after
+		// IPv6 is supported.
+		if c.multicastEnabled && !isIPv6 && c.networkConfig.TrafficEncapMode.SupportsEncap() {
 			// Drop the multicast packets forwarded from other Nodes in the cluster. This is because
 			// the packet sent out from the sender Pod is already received via tunnel port with encap mode,
 			// and the one forwarded via the underlay network is to send to external receivers
@@ -832,7 +835,9 @@ func (c *Client) restoreIptablesData(podCIDR *net.IPNet,
 	writeLine(iptablesData, iptables.MakeChainLine(antreaPostRoutingChain))
 	// The masqueraded multicast traffic will become unicast so we
 	// stop traversing this antreaPostRoutingChain for multicast traffic.
-	if c.multicastEnabled && c.networkConfig.TrafficEncapMode.SupportsNoEncap() {
+	// Note: Multicast can only work with IPv4 for now. Remove condition "!isIPv6" in the future after
+	// IPv6 is supported.
+	if c.multicastEnabled && !isIPv6 && c.networkConfig.TrafficEncapMode.SupportsNoEncap() {
 		writeLine(iptablesData, []string{
 			"-A", antreaPostRoutingChain,
 			"-m", "comment", "--comment", `"Antrea: skip masquerade for multicast traffic"`,

pkg/agent/route/route_linux_test.go

@@ -402,7 +402,6 @@ COMMIT
 :ANTREA-OUTPUT - [0:0]
 -A ANTREA-PREROUTING -m comment --comment "Antrea: do not track incoming encapsulation packets" -m udp -p udp --dport 6081 -m addrtype --dst-type LOCAL -j NOTRACK
 -A ANTREA-OUTPUT -m comment --comment "Antrea: do not track outgoing encapsulation packets" -m udp -p udp --dport 6081 -m addrtype --src-type LOCAL -j NOTRACK
--A ANTREA-PREROUTING -m comment --comment "Antrea: drop Pod multicast traffic forwarded via underlay network" -m set --match-set CLUSTER-NODE-IP6 src -d 224.0.0.0/4 -j DROP
 COMMIT
 *mangle
 :ANTREA-MANGLE - [0:0]