Internal change.

PiperOrigin-RevId: 274700093

pkg/sentry/socket/netstack/netstack.go

@@ -1177,6 +1177,25 @@ func getSockOptIPv6(t *kernel.Task, ep commonEndpoint, name, outLen int) (interf
 	case linux.IPV6_PATHMTU:
 		t.Kernel().EmitUnimplementedEvent(t)
 
+	case linux.IPV6_TCLASS:
+		// Length handling for parity with Linux.
+		if outLen == 0 {
+			return make([]byte, 0), nil
+		}
+		var v tcpip.IPv6TrafficClassOption
+		if err := ep.GetSockOpt(&v); err != nil {
+			return nil, syserr.TranslateNetstackError(err)
+		}
+
+		uintv := uint32(v)
+		// Linux truncates the output binary to outLen.
+		ib := binary.Marshal(nil, usermem.ByteOrder, &uintv)
+		// Handle cases where outLen is lesser than sizeOfInt32.
+		if len(ib) > outLen {
+			ib = ib[:outLen]
+		}
+		return ib, nil
+
 	default:
 		emitUnimplementedEventIPv6(t, name)
 	}
@@ -1244,6 +1263,20 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in
 		}
 		return int32(0), nil
 
+	case linux.IP_TOS:
+		// Length handling for parity with Linux.
+		if outLen == 0 {
+			return []byte(nil), nil
+		}
+		var v tcpip.IPv4TOSOption
+		if err := ep.GetSockOpt(&v); err != nil {
+			return nil, syserr.TranslateNetstackError(err)
+		}
+		if outLen < sizeOfInt32 {
+			return uint8(v), nil
+		}
+		return int32(v), nil
+
 	default:
 		emitUnimplementedEventIP(t, name)
 	}
@@ -1542,6 +1575,19 @@ func setSockOptIPv6(t *kernel.Task, ep commonEndpoint, name int, optVal []byte)
 
 		t.Kernel().EmitUnimplementedEvent(t)
 
+	case linux.IPV6_TCLASS:
+		if len(optVal) < sizeOfInt32 {
+			return syserr.ErrInvalidArgument
+		}
+		v := int32(usermem.ByteOrder.Uint32(optVal))
+		if v < -1 || v > 255 {
+			return syserr.ErrInvalidArgument
+		}
+		if v == -1 {
+			v = 0
+		}
+		return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.IPv6TrafficClassOption(v)))
+
 	default:
 		emitUnimplementedEventIPv6(t, name)
 	}
@@ -1687,6 +1733,16 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
 		}
 		return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.TTLOption(v)))
 
+	case linux.IP_TOS:
+		if len(optVal) == 0 {
+			return nil
+		}
+		v, err := parseIntOrChar(optVal)
+		if err != nil {
+			return err
+		}
+		return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.IPv4TOSOption(v)))
+
 	case linux.IP_ADD_SOURCE_MEMBERSHIP,
 		linux.IP_BIND_ADDRESS_NO_PORT,
 		linux.IP_BLOCK_SOURCE,
@@ -1710,7 +1766,6 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s
 		linux.IP_RECVTOS,
 		linux.IP_RECVTTL,
 		linux.IP_RETOPTS,
-		linux.IP_TOS,
 		linux.IP_TRANSPARENT,
 		linux.IP_UNBLOCK_SOURCE,
 		linux.IP_UNICAST_IF,

pkg/tcpip/network/arp/arp.go

@@ -79,7 +79,7 @@ func (e *endpoint) MaxHeaderLength() uint16 {
 
 func (e *endpoint) Close() {}
 
-func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, buffer.Prependable, buffer.VectorisedView, tcpip.TransportProtocolNumber, uint8, stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, buffer.Prependable, buffer.VectorisedView, stack.NetworkHeaderParams, stack.PacketLooping) *tcpip.Error {
 	return tcpip.ErrNotSupported
 }
 

pkg/tcpip/network/ip_test.go

@@ -230,7 +230,7 @@ func TestIPv4Send(t *testing.T) {
 	if err != nil {
 		t.Fatalf("could not find route: %v", err)
 	}
-	if err := ep.WritePacket(&r, nil /* gso */, hdr, payload.ToVectorisedView(), 123, 123, stack.PacketOut); err != nil {
+	if err := ep.WritePacket(&r, nil /* gso */, hdr, payload.ToVectorisedView(), stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketOut); err != nil {
 		t.Fatalf("WritePacket failed: %v", err)
 	}
 }
@@ -460,7 +460,7 @@ func TestIPv6Send(t *testing.T) {
 	if err != nil {
 		t.Fatalf("could not find route: %v", err)
 	}
-	if err := ep.WritePacket(&r, nil /* gso */, hdr, payload.ToVectorisedView(), 123, 123, stack.PacketOut); err != nil {
+	if err := ep.WritePacket(&r, nil /* gso */, hdr, payload.ToVectorisedView(), stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketOut); err != nil {
 		t.Fatalf("WritePacket failed: %v", err)
 	}
 }

pkg/tcpip/network/ipv4/icmp.go

@@ -95,7 +95,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V
 		pkt.SetChecksum(0)
 		pkt.SetChecksum(^header.Checksum(pkt, header.ChecksumVV(vv, 0)))
 		sent := stats.ICMP.V4PacketsSent
-		if err := r.WritePacket(nil /* gso */, hdr, vv, header.ICMPv4ProtocolNumber, 0, true /* useDefaultTTL */); err != nil {
+		if err := r.WritePacket(nil /* gso */, hdr, vv, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}); err != nil {
 			sent.Dropped.Increment()
 			return
 		}

pkg/tcpip/network/ipv4/ipv4.go

@@ -199,21 +199,22 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, hdr buff
 }
 
 // WritePacket writes a packet to the given destination address and protocol.
-func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params stack.NetworkHeaderParams, loop stack.PacketLooping) *tcpip.Error {
 	ip := header.IPv4(hdr.Prepend(header.IPv4MinimumSize))
 	length := uint16(hdr.UsedLength() + payload.Size())
 	id := uint32(0)
 	if length > header.IPv4MaximumHeaderSize+8 {
 		// Packets of 68 bytes or less are required by RFC 791 to not be
 		// fragmented, so we only assign ids to larger packets.
-		id = atomic.AddUint32(&e.protocol.ids[hashRoute(r, protocol, e.protocol.hashIV)%buckets], 1)
+		id = atomic.AddUint32(&e.protocol.ids[hashRoute(r, params.Protocol, e.protocol.hashIV)%buckets], 1)
 	}
 	ip.Encode(&header.IPv4Fields{
 		IHL:         header.IPv4MinimumSize,
 		TotalLength: length,
 		ID:          uint16(id),
-		TTL:         ttl,
-		Protocol:    uint8(protocol),
+		TTL:         params.TTL,
+		TOS:         params.TOS,
+		Protocol:    uint8(params.Protocol),
 		SrcAddr:     r.LocalAddress,
 		DstAddr:     r.RemoteAddress,
 	})

pkg/tcpip/network/ipv4/ipv4_test.go

@@ -302,7 +302,7 @@ func TestFragmentation(t *testing.T) {
 				Payload: payload.Clone([]buffer.View{}),
 			}
 			c := buildContext(t, nil, ft.mtu)
-			err := c.Route.WritePacket(ft.gso, hdr, payload, tcp.ProtocolNumber, 42 /* ttl */, false /* useDefaultTTL */)
+			err := c.Route.WritePacket(ft.gso, hdr, payload, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS})
 			if err != nil {
 				t.Errorf("err got %v, want %v", err, nil)
 			}
@@ -349,7 +349,7 @@ func TestFragmentationErrors(t *testing.T) {
 		t.Run(ft.description, func(t *testing.T) {
 			hdr, payload := makeHdrAndPayload(ft.hdrLength, header.IPv4MinimumSize, ft.payloadViewsSizes)
 			c := buildContext(t, ft.packetCollectorErrors, ft.mtu)
-			err := c.Route.WritePacket(&stack.GSO{}, hdr, payload, tcp.ProtocolNumber, 42 /* ttl */, false /* useDefaultTTL */)
+			err := c.Route.WritePacket(&stack.GSO{}, hdr, payload, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS})
 			for i := 0; i < len(ft.packetCollectorErrors)-1; i++ {
 				if got, want := ft.packetCollectorErrors[i], (*tcpip.Error)(nil); got != want {
 					t.Errorf("ft.packetCollectorErrors[%d] got %v, want %v", i, got, want)

pkg/tcpip/network/ipv6/icmp.go

@@ -154,7 +154,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V
 		r.LocalAddress = targetAddr
 		pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, buffer.VectorisedView{}))
 
-		if err := r.WritePacket(nil /* gso */, hdr, buffer.VectorisedView{}, header.ICMPv6ProtocolNumber, 0, true /* useDefaultTTL */); err != nil {
+		if err := r.WritePacket(nil /* gso */, hdr, buffer.VectorisedView{}, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}); err != nil {
 			sent.Dropped.Increment()
 			return
 		}
@@ -185,7 +185,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V
 		copy(pkt, h)
 		pkt.SetType(header.ICMPv6EchoReply)
 		pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, vv))
-		if err := r.WritePacket(nil /* gso */, hdr, vv, header.ICMPv6ProtocolNumber, 0, true /* useDefaultTTL */); err != nil {
+		if err := r.WritePacket(nil /* gso */, hdr, vv, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}); err != nil {
 			sent.Dropped.Increment()
 			return
 		}

pkg/tcpip/network/ipv6/ipv6.go

@@ -98,13 +98,14 @@ func (e *endpoint) GSOMaxSize() uint32 {
 }
 
 // WritePacket writes a packet to the given destination address and protocol.
-func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop stack.PacketLooping) *tcpip.Error {
+func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params stack.NetworkHeaderParams, loop stack.PacketLooping) *tcpip.Error {
 	length := uint16(hdr.UsedLength() + payload.Size())
 	ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize))
 	ip.Encode(&header.IPv6Fields{
 		PayloadLength: length,
-		NextHeader:    uint8(protocol),
-		HopLimit:      ttl,
+		NextHeader:    uint8(params.Protocol),
+		HopLimit:      params.TTL,
+		TrafficClass:  params.TOS,
 		SrcAddr:       r.LocalAddress,
 		DstAddr:       r.RemoteAddress,
 	})

pkg/tcpip/network/ipv6/ipv6_test.go

@@ -64,7 +64,7 @@ func testReceiveICMP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst
 	}
 }
 
-// testReceiveICMP tests receiving a UDP packet from src to dst. want is the
+// testReceiveUDP tests receiving a UDP packet from src to dst. want is the
 // expected UDP received count after receiving the packet.
 func testReceiveUDP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst tcpip.Address, want uint64) {
 	t.Helper()

pkg/tcpip/stack/registration.go

@@ -146,6 +146,19 @@ const (
 	PacketLoop
 )
 
+// NetworkHeaderParams are the header parameters given as input by the
+// transport endpoint to the network.
+type NetworkHeaderParams struct {
+	// Protocol refers to the transport protocol number.
+	Protocol tcpip.TransportProtocolNumber
+
+	// TTL refers to Time To Live field of the IP-header.
+	TTL uint8
+
+	// TOS refers to TypeOfService or TrafficClass field of the IP-header.
+	TOS uint8
+}
+
 // NetworkEndpoint is the interface that needs to be implemented by endpoints
 // of network layer protocols (e.g., ipv4, ipv6).
 type NetworkEndpoint interface {
@@ -170,7 +183,7 @@ type NetworkEndpoint interface {
 
 	// WritePacket writes a packet to the given destination address and
 	// protocol.
-	WritePacket(r *Route, gso *GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, loop PacketLooping) *tcpip.Error
+	WritePacket(r *Route, gso *GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params NetworkHeaderParams, loop PacketLooping) *tcpip.Error
 
 	// WriteHeaderIncludedPacket writes a packet that includes a network
 	// header to the given destination address.

pkg/tcpip/stack/route.go

@@ -154,16 +154,12 @@ func (r *Route) IsResolutionRequired() bool {
 }
 
 // WritePacket writes the packet through the given route.
-func (r *Route) WritePacket(gso *GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, ttl uint8, useDefaultTTL bool) *tcpip.Error {
+func (r *Route) WritePacket(gso *GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params NetworkHeaderParams) *tcpip.Error {
 	if !r.ref.isValidForOutgoing() {
 		return tcpip.ErrInvalidEndpointState
 	}
 
-	if useDefaultTTL {
-		ttl = r.DefaultTTL()
-	}
-
-	err := r.ref.ep.WritePacket(r, gso, hdr, payload, protocol, ttl, r.loop)
+	err := r.ref.ep.WritePacket(r, gso, hdr, payload, params, r.loop)
 	if err != nil {
 		r.Stats().IP.OutgoingPacketErrors.Increment()
 	} else {

pkg/tcpip/stack/stack.go

@@ -43,6 +43,9 @@ const (
 	resolutionTimeout = 1 * time.Second
 	// resolutionAttempts is set to the same ARP retries used in Linux.
 	resolutionAttempts = 3
+
+	// DefaultTOS is the default type of service value for network endpoints.
+	DefaultTOS = 0
 )
 
 type transportProtocolState struct {
@@ -394,7 +397,7 @@ type Stack struct {
 	// portSeed is a one-time random value initialized at stack startup
 	// and is used to seed the TCP port picking on active connections
 	//
-	// TODO(gvisor.dev/issues/940): S/R this field.
+	// TODO(gvisor.dev/issue/940): S/R this field.
 	portSeed uint32
 }
 

pkg/tcpip/stack/stack_test.go

@@ -119,7 +119,7 @@ func (f *fakeNetworkEndpoint) Capabilities() stack.LinkEndpointCapabilities {
 	return f.ep.Capabilities()
 }
 
-func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, protocol tcpip.TransportProtocolNumber, _ uint8, loop stack.PacketLooping) *tcpip.Error {
+func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr buffer.Prependable, payload buffer.VectorisedView, params stack.NetworkHeaderParams, loop stack.PacketLooping) *tcpip.Error {
 	// Increment the sent packet count in the protocol descriptor.
 	f.proto.sendPacketCount[int(r.RemoteAddress[0])%len(f.proto.sendPacketCount)]++
 
@@ -128,7 +128,7 @@ func (f *fakeNetworkEndpoint) WritePacket(r *stack.Route, gso *stack.GSO, hdr bu
 	b := hdr.Prepend(fakeNetHeaderLen)
 	b[0] = r.RemoteAddress[0]
 	b[1] = f.id.LocalAddress[0]
-	b[2] = byte(protocol)
+	b[2] = byte(params.Protocol)
 
 	if loop&stack.PacketLoop != 0 {
 		views := make([]buffer.View, 1, 1+len(payload.Views()))
@@ -310,7 +310,7 @@ func sendTo(s *stack.Stack, addr tcpip.Address, payload buffer.View) *tcpip.Erro
 
 func send(r stack.Route, payload buffer.View) *tcpip.Error {
 	hdr := buffer.NewPrependable(int(r.MaxHeaderLength()))
-	return r.WritePacket(nil /* gso */, hdr, payload.ToVectorisedView(), fakeTransNumber, 123 /* ttl */, false /* useDefaultTTL */)
+	return r.WritePacket(nil /* gso */, hdr, payload.ToVectorisedView(), stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS})
 }
 
 func testSendTo(t *testing.T, s *stack.Stack, addr tcpip.Address, ep *channel.Endpoint, payload buffer.View) {

pkg/tcpip/stack/transport_test.go

@@ -82,7 +82,7 @@ func (f *fakeTransportEndpoint) Write(p tcpip.Payloader, opts tcpip.WriteOptions
 	if err != nil {
 		return 0, nil, err
 	}
-	if err := f.route.WritePacket(nil /* gso */, hdr, buffer.View(v).ToVectorisedView(), fakeTransNumber, 123 /* ttl */, false /* useDefaultTTL */); err != nil {
+	if err := f.route.WritePacket(nil /* gso */, hdr, buffer.View(v).ToVectorisedView(), stack.NetworkHeaderParams{Protocol: fakeTransNumber, TTL: 123, TOS: stack.DefaultTOS}); err != nil {
 		return 0, nil, err
 	}
 

pkg/tcpip/tcpip.go

@@ -624,6 +624,14 @@ type BroadcastOption int
 // a default TTL.
 type DefaultTTLOption uint8
 
+// IPv4TOSOption is used by SetSockOpt/GetSockOpt to specify TOS
+// for all subsequent outgoing IPv4 packets from the endpoint.
+type IPv4TOSOption uint8
+
+// IPv6TrafficClassOption is used by SetSockOpt/GetSockOpt to specify TOS
+// for all subsequent outgoing IPv6 packets from the endpoint.
+type IPv6TrafficClassOption uint8
+
 // Route is a row in the routing table. It specifies through which NIC (and
 // gateway) sets of packets should be routed. A row is considered viable if the
 // masked target address matches the destination address in the row.

pkg/tcpip/transport/icmp/endpoint.go

@@ -422,7 +422,10 @@ func send4(r *stack.Route, ident uint16, data buffer.View, ttl uint8) *tcpip.Err
 	icmpv4.SetChecksum(0)
 	icmpv4.SetChecksum(^header.Checksum(icmpv4, header.Checksum(data, 0)))
 
-	return r.WritePacket(nil /* gso */, hdr, data.ToVectorisedView(), header.ICMPv4ProtocolNumber, ttl, ttl == 0 /* useDefaultTTL */)
+	if ttl == 0 {
+		ttl = r.DefaultTTL()
+	}
+	return r.WritePacket(nil /* gso */, hdr, data.ToVectorisedView(), stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS})
 }
 
 func send6(r *stack.Route, ident uint16, data buffer.View, ttl uint8) *tcpip.Error {
@@ -445,7 +448,10 @@ func send6(r *stack.Route, ident uint16, data buffer.View, ttl uint8) *tcpip.Err
 	icmpv6.SetChecksum(0)
 	icmpv6.SetChecksum(^header.Checksum(icmpv6, header.Checksum(data, 0)))
 
-	return r.WritePacket(nil /* gso */, hdr, data.ToVectorisedView(), header.ICMPv6ProtocolNumber, ttl, ttl == 0 /* useDefaultTTL */)
+	if ttl == 0 {
+		ttl = r.DefaultTTL()
+	}
+	return r.WritePacket(nil /* gso */, hdr, data.ToVectorisedView(), stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: ttl, TOS: stack.DefaultTOS})
 }
 
 func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress, allowMismatch bool) (tcpip.NetworkProtocolNumber, *tcpip.Error) {

pkg/tcpip/transport/raw/endpoint.go

@@ -350,7 +350,7 @@ func (e *endpoint) finishWrite(payloadBytes []byte, route *stack.Route) (int64,
 			break
 		}
 		hdr := buffer.NewPrependable(len(payloadBytes) + int(route.MaxHeaderLength()))
-		if err := route.WritePacket(nil /* gso */, hdr, buffer.View(payloadBytes).ToVectorisedView(), e.TransProto, 0, true /* useDefaultTTL */); err != nil {
+		if err := route.WritePacket(nil /* gso */, hdr, buffer.View(payloadBytes).ToVectorisedView(), stack.NetworkHeaderParams{Protocol: e.TransProto, TTL: route.DefaultTTL(), TOS: stack.DefaultTOS}); err != nil {
 			return 0, nil, err
 		}
 

pkg/tcpip/transport/tcp/accept.go

@@ -441,7 +441,7 @@ func (e *endpoint) handleListenSegment(ctx *listenContext, s *segment) {
 				TSEcr: opts.TSVal,
 				MSS:   uint16(mss),
 			}
-			e.sendSynTCP(&s.route, s.id, e.ttl, header.TCPFlagSyn|header.TCPFlagAck, cookie, s.sequenceNumber+1, ctx.rcvWnd, synOpts)
+			e.sendSynTCP(&s.route, s.id, e.ttl, e.sendTOS, header.TCPFlagSyn|header.TCPFlagAck, cookie, s.sequenceNumber+1, ctx.rcvWnd, synOpts)
 			e.stack.Stats().TCP.ListenOverflowSynCookieSent.Increment()
 		}
 

pkg/tcpip/transport/tcp/connect.go

@@ -255,7 +255,7 @@ func (h *handshake) synSentState(s *segment) *tcpip.Error {
 	if ttl == 0 {
 		ttl = s.route.DefaultTTL()
 	}
-	h.ep.sendSynTCP(&s.route, h.ep.ID, ttl, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
+	h.ep.sendSynTCP(&s.route, h.ep.ID, ttl, h.ep.sendTOS, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
 	return nil
 }
 
@@ -299,7 +299,7 @@ func (h *handshake) synRcvdState(s *segment) *tcpip.Error {
 			SACKPermitted: h.ep.sackPermitted,
 			MSS:           h.ep.amss,
 		}
-		h.ep.sendSynTCP(&s.route, h.ep.ID, h.ep.ttl, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
+		h.ep.sendSynTCP(&s.route, h.ep.ID, h.ep.ttl, h.ep.sendTOS, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
 		return nil
 	}
 
@@ -468,7 +468,8 @@ func (h *handshake) execute() *tcpip.Error {
 			synOpts.WS = -1
 		}
 	}
-	h.ep.sendSynTCP(&h.ep.route, h.ep.ID, h.ep.ttl, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
+	h.ep.sendSynTCP(&h.ep.route, h.ep.ID, h.ep.ttl, h.ep.sendTOS, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
+
 	for h.state != handshakeCompleted {
 		switch index, _ := s.Fetch(true); index {
 		case wakerForResend:
@@ -477,7 +478,7 @@ func (h *handshake) execute() *tcpip.Error {
 				return tcpip.ErrTimeout
 			}
 			rt.Reset(timeOut)
-			h.ep.sendSynTCP(&h.ep.route, h.ep.ID, h.ep.ttl, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
+			h.ep.sendSynTCP(&h.ep.route, h.ep.ID, h.ep.ttl, h.ep.sendTOS, h.flags, h.iss, h.ackNum, h.rcvWnd, synOpts)
 
 		case wakerForNotification:
 			n := h.ep.fetchNotifications()
@@ -587,17 +588,18 @@ func makeSynOptions(opts header.TCPSynOptions) []byte {
 	return options[:offset]
 }
 
-func (e *endpoint) sendSynTCP(r *stack.Route, id stack.TransportEndpointID, ttl uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts header.TCPSynOptions) {
+func (e *endpoint) sendSynTCP(r *stack.Route, id stack.TransportEndpointID, ttl, tos uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts header.TCPSynOptions) *tcpip.Error {
 	options := makeSynOptions(opts)
 	// We ignore SYN send errors and let the callers re-attempt send.
-	if err := e.sendTCP(r, id, buffer.VectorisedView{}, ttl, flags, seq, ack, rcvWnd, options, nil); err != nil {
+	if err := e.sendTCP(r, id, buffer.VectorisedView{}, ttl, tos, flags, seq, ack, rcvWnd, options, nil); err != nil {
 		e.stats.SendErrors.SynSendToNetworkFailed.Increment()
 	}
 	putOptions(options)
+	return nil
 }
 
-func (e *endpoint) sendTCP(r *stack.Route, id stack.TransportEndpointID, data buffer.VectorisedView, ttl uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts []byte, gso *stack.GSO) *tcpip.Error {
-	if err := sendTCP(r, id, data, ttl, flags, seq, ack, rcvWnd, opts, gso); err != nil {
+func (e *endpoint) sendTCP(r *stack.Route, id stack.TransportEndpointID, data buffer.VectorisedView, ttl, tos uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts []byte, gso *stack.GSO) *tcpip.Error {
+	if err := sendTCP(r, id, data, ttl, tos, flags, seq, ack, rcvWnd, opts, gso); err != nil {
 		e.stats.SendErrors.SegmentSendToNetworkFailed.Increment()
 		return err
 	}
@@ -607,7 +609,7 @@ func (e *endpoint) sendTCP(r *stack.Route, id stack.TransportEndpointID, data bu
 
 // sendTCP sends a TCP segment with the provided options via the provided
 // network endpoint and under the provided identity.
-func sendTCP(r *stack.Route, id stack.TransportEndpointID, data buffer.VectorisedView, ttl uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts []byte, gso *stack.GSO) *tcpip.Error {
+func sendTCP(r *stack.Route, id stack.TransportEndpointID, data buffer.VectorisedView, ttl, tos uint8, flags byte, seq, ack seqnum.Value, rcvWnd seqnum.Size, opts []byte, gso *stack.GSO) *tcpip.Error {
 	optLen := len(opts)
 	// Allocate a buffer for the TCP header.
 	hdr := buffer.NewPrependable(header.TCPMinimumSize + int(r.MaxHeaderLength()) + optLen)
@@ -643,7 +645,10 @@ func sendTCP(r *stack.Route, id stack.TransportEndpointID, data buffer.Vectorise
 		tcp.SetChecksum(^tcp.CalculateChecksum(xsum))
 	}
 
-	if err := r.WritePacket(gso, hdr, data, ProtocolNumber, ttl, ttl == 0 /* useDefaultTTL */); err != nil {
+	if ttl == 0 {
+		ttl = r.DefaultTTL()
+	}
+	if err := r.WritePacket(gso, hdr, data, stack.NetworkHeaderParams{Protocol: ProtocolNumber, TTL: ttl, TOS: tos}); err != nil {
 		r.Stats().TCP.SegmentSendErrors.Increment()
 		return err
 	}
@@ -700,7 +705,7 @@ func (e *endpoint) sendRaw(data buffer.VectorisedView, flags byte, seq, ack seqn
 		sackBlocks = e.sack.Blocks[:e.sack.NumBlocks]
 	}
 	options := e.makeOptions(sackBlocks)
-	err := e.sendTCP(&e.route, e.ID, data, e.ttl, flags, seq, ack, rcvWnd, options, e.gso)
+	err := e.sendTCP(&e.route, e.ID, data, e.ttl, e.sendTOS, flags, seq, ack, rcvWnd, options, e.gso)
 	putOptions(options)
 	return err
 }

pkg/tcpip/transport/tcp/dual_stack_test.go

@@ -42,7 +42,7 @@ func TestV4MappedConnectOnV6Only(t *testing.T) {
 	}
 }
 
-func testV4Connect(t *testing.T, c *context.Context) {
+func testV4Connect(t *testing.T, c *context.Context, checkers ...checker.NetworkChecker) {
 	// Start connection attempt.
 	we, ch := waiter.NewChannelEntry(nil)
 	c.WQ.EventRegister(&we, waiter.EventOut)
@@ -55,12 +55,11 @@ func testV4Connect(t *testing.T, c *context.Context) {
 
 	// Receive SYN packet.
 	b := c.GetPacket()
-	checker.IPv4(t, b,
-		checker.TCP(
-			checker.DstPort(context.TestPort),
-			checker.TCPFlags(header.TCPFlagSyn),
-		),
-	)
+	synCheckers := append(checkers, checker.TCP(
+		checker.DstPort(context.TestPort),
+		checker.TCPFlags(header.TCPFlagSyn),
+	))
+	checker.IPv4(t, b, synCheckers...)
 
 	tcp := header.TCP(header.IPv4(b).Payload())
 	c.IRS = seqnum.Value(tcp.SequenceNumber())
@@ -76,14 +75,13 @@ func testV4Connect(t *testing.T, c *context.Context) {
 	})
 
 	// Receive ACK packet.
-	checker.IPv4(t, c.GetPacket(),
-		checker.TCP(
-			checker.DstPort(context.TestPort),
-			checker.TCPFlags(header.TCPFlagAck),
-			checker.SeqNum(uint32(c.IRS)+1),
-			checker.AckNum(uint32(iss)+1),
-		),
-	)
+	ackCheckers := append(checkers, checker.TCP(
+		checker.DstPort(context.TestPort),
+		checker.TCPFlags(header.TCPFlagAck),
+		checker.SeqNum(uint32(c.IRS)+1),
+		checker.AckNum(uint32(iss)+1),
+	))
+	checker.IPv4(t, c.GetPacket(), ackCheckers...)
 
 	// Wait for connection to be established.
 	select {
@@ -152,7 +150,7 @@ func TestV4ConnectWhenBoundToV4Mapped(t *testing.T) {
 	testV4Connect(t, c)
 }
 
-func testV6Connect(t *testing.T, c *context.Context) {
+func testV6Connect(t *testing.T, c *context.Context, checkers ...checker.NetworkChecker) {
 	// Start connection attempt to IPv6 address.
 	we, ch := waiter.NewChannelEntry(nil)
 	c.WQ.EventRegister(&we, waiter.EventOut)
@@ -165,12 +163,11 @@ func testV6Connect(t *testing.T, c *context.Context) {
 
 	// Receive SYN packet.
 	b := c.GetV6Packet()
-	checker.IPv6(t, b,
-		checker.TCP(
-			checker.DstPort(context.TestPort),
-			checker.TCPFlags(header.TCPFlagSyn),
-		),
-	)
+	synCheckers := append(checkers, checker.TCP(
+		checker.DstPort(context.TestPort),
+		checker.TCPFlags(header.TCPFlagSyn),
+	))
+	checker.IPv6(t, b, synCheckers...)
 
 	tcp := header.TCP(header.IPv6(b).Payload())
 	c.IRS = seqnum.Value(tcp.SequenceNumber())
@@ -186,14 +183,13 @@ func testV6Connect(t *testing.T, c *context.Context) {
 	})
 
 	// Receive ACK packet.
-	checker.IPv6(t, c.GetV6Packet(),
-		checker.TCP(
-			checker.DstPort(context.TestPort),
-			checker.TCPFlags(header.TCPFlagAck),
-			checker.SeqNum(uint32(c.IRS)+1),
-			checker.AckNum(uint32(iss)+1),
-		),
-	)
+	ackCheckers := append(checkers, checker.TCP(
+		checker.DstPort(context.TestPort),
+		checker.TCPFlags(header.TCPFlagAck),
+		checker.SeqNum(uint32(c.IRS)+1),
+		checker.AckNum(uint32(iss)+1),
+	))
+	checker.IPv6(t, c.GetV6Packet(), ackCheckers...)
 
 	// Wait for connection to be established.
 	select {