From bfa0bb24dd73072b3ccfe302afa89d1b5d46b927 Mon Sep 17 00:00:00 2001 From: gVisor bot Date: Mon, 14 Oct 2019 17:45:29 -0700 Subject: Internal change. PiperOrigin-RevId: 274700093 --- pkg/sentry/socket/netstack/netstack.go | 57 +++++++- pkg/tcpip/network/arp/arp.go | 2 +- pkg/tcpip/network/ip_test.go | 4 +- pkg/tcpip/network/ipv4/icmp.go | 2 +- pkg/tcpip/network/ipv4/ipv4.go | 9 +- pkg/tcpip/network/ipv4/ipv4_test.go | 4 +- pkg/tcpip/network/ipv6/icmp.go | 4 +- pkg/tcpip/network/ipv6/ipv6.go | 7 +- pkg/tcpip/network/ipv6/ipv6_test.go | 2 +- pkg/tcpip/stack/registration.go | 15 +- pkg/tcpip/stack/route.go | 8 +- pkg/tcpip/stack/stack.go | 5 +- pkg/tcpip/stack/stack_test.go | 6 +- pkg/tcpip/stack/transport_test.go | 2 +- pkg/tcpip/tcpip.go | 8 + pkg/tcpip/transport/icmp/endpoint.go | 10 +- pkg/tcpip/transport/raw/endpoint.go | 2 +- pkg/tcpip/transport/tcp/accept.go | 2 +- pkg/tcpip/transport/tcp/connect.go | 27 ++-- pkg/tcpip/transport/tcp/dual_stack_test.go | 56 ++++--- pkg/tcpip/transport/tcp/endpoint.go | 35 +++++ pkg/tcpip/transport/tcp/protocol.go | 2 +- pkg/tcpip/transport/tcp/tcp_test.go | 101 +++++++++++++ pkg/tcpip/transport/udp/endpoint.go | 37 ++++- pkg/tcpip/transport/udp/protocol.go | 4 +- pkg/tcpip/transport/udp/udp_test.go | 70 +++++++++ test/syscalls/linux/socket_ip_unbound.cc | 228 +++++++++++++++++++++++++++++ 27 files changed, 629 insertions(+), 80 deletions(-) diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go index 09db2109f..0ae573b45 100644 --- a/pkg/sentry/socket/netstack/netstack.go +++ b/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, diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 26cf1c528..922181ac0 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/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 } diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index a9741622e..f644a8b08 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/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) } } diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index c1cf6c222..50b363dc4 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/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 } diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index fb6358fbb..df1a08113 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/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, }) diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 3dcb503e2..560638ce8 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/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) diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index 71c398027..b5df85455 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/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 } diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index 85c070c43..cd1e34085 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/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, }) diff --git a/pkg/tcpip/network/ipv6/ipv6_test.go b/pkg/tcpip/network/ipv6/ipv6_test.go index 78c674c2c..deaa9b7f3 100644 --- a/pkg/tcpip/network/ipv6/ipv6_test.go +++ b/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() diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index 80101d4bb..9d6157f22 100644 --- a/pkg/tcpip/stack/registration.go +++ b/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. diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go index 22db619a1..e72373964 100644 --- a/pkg/tcpip/stack/route.go +++ b/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 { diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index 7d73389cc..f67975525 100644 --- a/pkg/tcpip/stack/stack.go +++ b/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 } diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go index 8f063038b..10fd1065f 100644 --- a/pkg/tcpip/stack/stack_test.go +++ b/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) { diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index 63811c684..86c62be25 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/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 } diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index 444ac1a5b..9d3752032 100644 --- a/pkg/tcpip/tcpip.go +++ b/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. diff --git a/pkg/tcpip/transport/icmp/endpoint.go b/pkg/tcpip/transport/icmp/endpoint.go index d0cfdcda1..3187b336b 100644 --- a/pkg/tcpip/transport/icmp/endpoint.go +++ b/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) { diff --git a/pkg/tcpip/transport/raw/endpoint.go b/pkg/tcpip/transport/raw/endpoint.go index 4f5c286cf..b4c660859 100644 --- a/pkg/tcpip/transport/raw/endpoint.go +++ b/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 } diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index 8f5572195..844959fa0 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/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() } diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index cb8cfd619..5ea036bea 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/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 } diff --git a/pkg/tcpip/transport/tcp/dual_stack_test.go b/pkg/tcpip/transport/tcp/dual_stack_test.go index c54610a87..dfaa4a559 100644 --- a/pkg/tcpip/transport/tcp/dual_stack_test.go +++ b/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 { diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 090a8eb24..a1b784b49 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -494,6 +494,10 @@ type endpoint struct { // amss is the advertised MSS to the peer by this endpoint. amss uint16 + // sendTOS represents IPv4 TOS or IPv6 TrafficClass, + // applied while sending packets. Defaults to 0 as on Linux. + sendTOS uint8 + gso *stack.GSO // TODO(b/142022063): Add ability to save and restore per endpoint stats. @@ -1136,6 +1140,8 @@ func (e *endpoint) SetSockOptInt(opt tcpip.SockOpt, v int) *tcpip.Error { // SetSockOpt sets a socket option. func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { + // Lower 2 bits represents ECN bits. RFC 3168, section 23.1 + const inetECNMask = 3 switch v := opt.(type) { case tcpip.DelayOption: if v == 0 { @@ -1296,6 +1302,23 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { // Linux returns ENOENT when an invalid congestion // control algorithm is specified. return tcpip.ErrNoSuchFile + + case tcpip.IPv4TOSOption: + e.mu.Lock() + // TODO(gvisor.dev/issue/995): ECN is not currently supported, + // ignore the bits for now. + e.sendTOS = uint8(v) & ^uint8(inetECNMask) + e.mu.Unlock() + return nil + + case tcpip.IPv6TrafficClassOption: + e.mu.Lock() + // TODO(gvisor.dev/issue/995): ECN is not currently supported, + // ignore the bits for now. + e.sendTOS = uint8(v) & ^uint8(inetECNMask) + e.mu.Unlock() + return nil + default: return nil } @@ -1495,6 +1518,18 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { e.mu.Unlock() return nil + case *tcpip.IPv4TOSOption: + e.mu.RLock() + *o = tcpip.IPv4TOSOption(e.sendTOS) + e.mu.RUnlock() + return nil + + case *tcpip.IPv6TrafficClassOption: + e.mu.RLock() + *o = tcpip.IPv6TrafficClassOption(e.sendTOS) + e.mu.RUnlock() + return nil + default: return tcpip.ErrUnknownProtocolOption } diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go index d5d8ab96a..db40785d3 100644 --- a/pkg/tcpip/transport/tcp/protocol.go +++ b/pkg/tcpip/transport/tcp/protocol.go @@ -153,7 +153,7 @@ func replyWithReset(s *segment) { ack := s.sequenceNumber.Add(s.logicalLen()) - sendTCP(&s.route, s.id, buffer.VectorisedView{}, s.route.DefaultTTL(), header.TCPFlagRst|header.TCPFlagAck, seq, ack, 0, nil /* options */, nil /* gso */) + sendTCP(&s.route, s.id, buffer.VectorisedView{}, s.route.DefaultTTL(), stack.DefaultTOS, header.TCPFlagRst|header.TCPFlagAck, seq, ack, 0 /* rcvWnd */, nil /* options */, nil /* gso */) } // SetOption implements TransportProtocol.SetOption. diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go index 8eaf9786d..6d022a266 100644 --- a/pkg/tcpip/transport/tcp/tcp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_test.go @@ -474,6 +474,107 @@ func TestSimpleReceive(t *testing.T) { ) } +func TestTOSV4(t *testing.T) { + c := context.New(t, defaultMTU) + defer c.Cleanup() + + ep, err := c.Stack().NewEndpoint(tcp.ProtocolNumber, ipv4.ProtocolNumber, &c.WQ) + if err != nil { + t.Fatalf("NewEndpoint failed: %s", err) + } + c.EP = ep + + const tos = 0xC0 + if err := c.EP.SetSockOpt(tcpip.IPv4TOSOption(tos)); err != nil { + t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.IPv4TOSOption(tos), err) + } + + var v tcpip.IPv4TOSOption + if err := c.EP.GetSockOpt(&v); err != nil { + t.Errorf("GetSockopt failed: %s", err) + } + + if want := tcpip.IPv4TOSOption(tos); v != want { + t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want) + } + + testV4Connect(t, c, checker.TOS(tos, 0)) + + data := []byte{1, 2, 3} + view := buffer.NewView(len(data)) + copy(view, data) + + if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil { + t.Fatalf("Write failed: %s", err) + } + + // Check that data is received. + b := c.GetPacket() + checker.IPv4(t, b, + checker.PayloadLen(len(data)+header.TCPMinimumSize), + checker.TCP( + checker.DstPort(context.TestPort), + checker.SeqNum(uint32(c.IRS)+1), + checker.AckNum(790), // Acknum is initial sequence number + 1 + checker.TCPFlagsMatch(header.TCPFlagAck, ^uint8(header.TCPFlagPsh)), + ), + checker.TOS(tos, 0), + ) + + if p := b[header.IPv4MinimumSize+header.TCPMinimumSize:]; !bytes.Equal(data, p) { + t.Errorf("got data = %x, want = %x", p, data) + } +} + +func TestTrafficClassV6(t *testing.T) { + c := context.New(t, defaultMTU) + defer c.Cleanup() + + c.CreateV6Endpoint(false) + + const tos = 0xC0 + if err := c.EP.SetSockOpt(tcpip.IPv6TrafficClassOption(tos)); err != nil { + t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.IPv6TrafficClassOption(tos), err) + } + + var v tcpip.IPv6TrafficClassOption + if err := c.EP.GetSockOpt(&v); err != nil { + t.Fatalf("GetSockopt failed: %s", err) + } + + if want := tcpip.IPv6TrafficClassOption(tos); v != want { + t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want) + } + + // Test the connection request. + testV6Connect(t, c, checker.TOS(tos, 0)) + + data := []byte{1, 2, 3} + view := buffer.NewView(len(data)) + copy(view, data) + + if _, _, err := c.EP.Write(tcpip.SlicePayload(view), tcpip.WriteOptions{}); err != nil { + t.Fatalf("Write failed: %s", err) + } + + // Check that data is received. + b := c.GetV6Packet() + checker.IPv6(t, b, + checker.PayloadLen(len(data)+header.TCPMinimumSize), + checker.TCP( + checker.DstPort(context.TestPort), + checker.SeqNum(uint32(c.IRS)+1), + checker.AckNum(790), + checker.TCPFlagsMatch(header.TCPFlagAck, ^uint8(header.TCPFlagPsh)), + ), + checker.TOS(tos, 0), + ) + + if p := b[header.IPv6MinimumSize+header.TCPMinimumSize:]; !bytes.Equal(data, p) { + t.Errorf("got data = %x, want = %x", p, data) + } +} + func TestConnectBindToDevice(t *testing.T) { for _, test := range []struct { name string diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 2ad7978c2..6e87245b7 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -106,6 +106,10 @@ type endpoint struct { bindToDevice tcpip.NICID broadcast bool + // sendTOS represents IPv4 TOS or IPv6 TrafficClass, + // applied while sending packets. Defaults to 0 as on Linux. + sendTOS uint8 + // shutdownFlags represent the current shutdown state of the endpoint. shutdownFlags tcpip.ShutdownFlags @@ -429,7 +433,7 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c useDefaultTTL = false } - if err := sendUDP(route, buffer.View(v).ToVectorisedView(), e.ID.LocalPort, dstPort, ttl, useDefaultTTL); err != nil { + if err := sendUDP(route, buffer.View(v).ToVectorisedView(), e.ID.LocalPort, dstPort, ttl, useDefaultTTL, e.sendTOS); err != nil { return 0, nil, err } return int64(len(v)), nil, nil @@ -628,6 +632,18 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { e.mu.Unlock() return nil + + case tcpip.IPv4TOSOption: + e.mu.Lock() + e.sendTOS = uint8(v) + e.mu.Unlock() + return nil + + case tcpip.IPv6TrafficClassOption: + e.mu.Lock() + e.sendTOS = uint8(v) + e.mu.Unlock() + return nil } return nil } @@ -748,6 +764,18 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { } return nil + case *tcpip.IPv4TOSOption: + e.mu.RLock() + *o = tcpip.IPv4TOSOption(e.sendTOS) + e.mu.RUnlock() + return nil + + case *tcpip.IPv6TrafficClassOption: + e.mu.RLock() + *o = tcpip.IPv6TrafficClassOption(e.sendTOS) + e.mu.RUnlock() + return nil + default: return tcpip.ErrUnknownProtocolOption } @@ -755,7 +783,7 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { // sendUDP sends a UDP segment via the provided network endpoint and under the // provided identity. -func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort uint16, ttl uint8, useDefaultTTL bool) *tcpip.Error { +func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort uint16, ttl uint8, useDefaultTTL bool, tos uint8) *tcpip.Error { // Allocate a buffer for the UDP header. hdr := buffer.NewPrependable(header.UDPMinimumSize + int(r.MaxHeaderLength())) @@ -778,7 +806,10 @@ func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort u udp.SetChecksum(^udp.CalculateChecksum(xsum)) } - if err := r.WritePacket(nil /* gso */, hdr, data, ProtocolNumber, ttl, useDefaultTTL); err != nil { + if useDefaultTTL { + ttl = r.DefaultTTL() + } + if err := r.WritePacket(nil /* gso */, hdr, data, stack.NetworkHeaderParams{Protocol: ProtocolNumber, TTL: ttl, TOS: tos}); err != nil { r.Stats().UDP.PacketSendErrors.Increment() return err } diff --git a/pkg/tcpip/transport/udp/protocol.go b/pkg/tcpip/transport/udp/protocol.go index 0b24bbc85..de026880f 100644 --- a/pkg/tcpip/transport/udp/protocol.go +++ b/pkg/tcpip/transport/udp/protocol.go @@ -130,7 +130,7 @@ func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.Trans pkt.SetType(header.ICMPv4DstUnreachable) pkt.SetCode(header.ICMPv4PortUnreachable) pkt.SetChecksum(header.ICMPv4Checksum(pkt, payload)) - r.WritePacket(nil /* gso */, hdr, payload, header.ICMPv4ProtocolNumber, 0, true /* useDefaultTTL */) + r.WritePacket(nil /* gso */, hdr, payload, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}) case header.IPv6AddressSize: if !r.Stack().AllowICMPMessage() { @@ -164,7 +164,7 @@ func (p *protocol) HandleUnknownDestinationPacket(r *stack.Route, id stack.Trans pkt.SetType(header.ICMPv6DstUnreachable) pkt.SetCode(header.ICMPv6PortUnreachable) pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, payload)) - r.WritePacket(nil /* gso */, hdr, payload, header.ICMPv6ProtocolNumber, 0, true /* useDefaultTTL */) + r.WritePacket(nil /* gso */, hdr, payload, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}) } return true } diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index 4ada73475..b724d788c 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -1267,6 +1267,76 @@ func TestSetTTL(t *testing.T) { } } +func TestTOSV4(t *testing.T) { + for _, flow := range []testFlow{unicastV4, multicastV4, broadcast} { + t.Run(fmt.Sprintf("flow:%s", flow), func(t *testing.T) { + c := newDualTestContext(t, defaultMTU) + defer c.cleanup() + + c.createEndpointForFlow(flow) + + const tos = 0xC0 + var v tcpip.IPv4TOSOption + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + // Test for expected default value. + if v != 0 { + c.t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, 0) + } + + if err := c.ep.SetSockOpt(tcpip.IPv4TOSOption(tos)); err != nil { + c.t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.IPv4TOSOption(tos), err) + } + + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + + if want := tcpip.IPv4TOSOption(tos); v != want { + c.t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want) + } + + testWrite(c, flow, checker.TOS(tos, 0)) + }) + } +} + +func TestTOSV6(t *testing.T) { + for _, flow := range []testFlow{unicastV4in6, unicastV6, unicastV6Only, multicastV4in6, multicastV6, broadcastIn6} { + t.Run(fmt.Sprintf("flow:%s", flow), func(t *testing.T) { + c := newDualTestContext(t, defaultMTU) + defer c.cleanup() + + c.createEndpointForFlow(flow) + + const tos = 0xC0 + var v tcpip.IPv6TrafficClassOption + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + // Test for expected default value. + if v != 0 { + c.t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, 0) + } + + if err := c.ep.SetSockOpt(tcpip.IPv6TrafficClassOption(tos)); err != nil { + c.t.Errorf("SetSockOpt failed: %s", err) + } + + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + + if want := tcpip.IPv6TrafficClassOption(tos); v != want { + c.t.Errorf("got GetSockOpt(...) = %#v, want = %#v", v, want) + } + + testWrite(c, flow, checker.TOS(tos, 0)) + }) + } +} + func TestMulticastInterfaceOption(t *testing.T) { for _, flow := range []testFlow{multicastV4, multicastV4in6, multicastV6, multicastV6Only} { t.Run(fmt.Sprintf("flow:%s", flow), func(t *testing.T) { diff --git a/test/syscalls/linux/socket_ip_unbound.cc b/test/syscalls/linux/socket_ip_unbound.cc index ffd9cde77..fa9a9df6f 100644 --- a/test/syscalls/linux/socket_ip_unbound.cc +++ b/test/syscalls/linux/socket_ip_unbound.cc @@ -127,6 +127,234 @@ TEST_P(IPUnboundSocketTest, InvalidNegativeTtl) { SyscallFailsWithErrno(EINVAL)); } +struct TOSOption { + int level; + int option; +}; + +constexpr int INET_ECN_MASK = 3; + +static TOSOption GetTOSOption(int domain) { + TOSOption opt; + switch (domain) { + case AF_INET: + opt.level = IPPROTO_IP; + opt.option = IP_TOS; + break; + case AF_INET6: + opt.level = IPPROTO_IPV6; + opt.option = IPV6_TCLASS; + break; + } + return opt; +} + +TEST_P(IPUnboundSocketTest, TOSDefault) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + TOSOption t = GetTOSOption(GetParam().domain); + int get = -1; + socklen_t get_sz = sizeof(get); + constexpr int kDefaultTOS = 0; + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, kDefaultTOS); +} + +TEST_P(IPUnboundSocketTest, SetTOS) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0xC0; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + + int get = -1; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, set); +} + +TEST_P(IPUnboundSocketTest, ZeroTOS) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + int get = -1; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, set); +} + +TEST_P(IPUnboundSocketTest, InvalidLargeTOS) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + // Test with exceeding the byte space. + int set = 256; + constexpr int kDefaultTOS = 0; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + if (GetParam().domain == AF_INET) { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + } else { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallFailsWithErrno(EINVAL)); + } + int get = -1; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, kDefaultTOS); +} + +TEST_P(IPUnboundSocketTest, CheckSkipECN) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0xFF; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + int expect = static_cast(set); + if (GetParam().protocol == IPPROTO_TCP) { + expect &= ~INET_ECN_MASK; + } + int get = -1; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, expect); +} + +TEST_P(IPUnboundSocketTest, ZeroTOSOptionSize) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0xC0; + socklen_t set_sz = 0; + TOSOption t = GetTOSOption(GetParam().domain); + if (GetParam().domain == AF_INET) { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + } else { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallFailsWithErrno(EINVAL)); + } + int get = -1; + socklen_t get_sz = 0; + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, 0); + EXPECT_EQ(get, -1); +} + +TEST_P(IPUnboundSocketTest, SmallTOSOptionSize) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0xC0; + constexpr int kDefaultTOS = 0; + TOSOption t = GetTOSOption(GetParam().domain); + for (socklen_t i = 1; i < sizeof(int); i++) { + int expect_tos; + socklen_t expect_sz; + if (GetParam().domain == AF_INET) { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, i), + SyscallSucceedsWithValue(0)); + expect_tos = set; + expect_sz = sizeof(uint8_t); + } else { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, i), + SyscallFailsWithErrno(EINVAL)); + expect_tos = kDefaultTOS; + expect_sz = i; + } + uint get = -1; + socklen_t get_sz = i; + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, expect_sz); + // Account for partial copies by getsockopt, retrieve the lower + // bits specified by get_sz, while comparing against expect_tos. + EXPECT_EQ(get & ~(~0 << (get_sz * 8)), expect_tos); + } +} + +TEST_P(IPUnboundSocketTest, LargeTOSOptionSize) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = 0xC0; + TOSOption t = GetTOSOption(GetParam().domain); + for (socklen_t i = sizeof(int); i < 10; i++) { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, i), + SyscallSucceedsWithValue(0)); + int get = -1; + socklen_t get_sz = i; + // We expect the system call handler to only copy atmost sizeof(int) bytes + // as asserted by the check below. Hence, we do not expect the copy to + // overflow in getsockopt. + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(int)); + EXPECT_EQ(get, set); + } +} + +TEST_P(IPUnboundSocketTest, NegativeTOS) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + + int set = -1; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + int expect; + if (GetParam().domain == AF_INET) { + expect = static_cast(set); + if (GetParam().protocol == IPPROTO_TCP) { + expect &= ~INET_ECN_MASK; + } + } else { + // On IPv6 TCLASS, setting -1 has the effect of resetting the + // TrafficClass. + expect = 0; + } + int get = -1; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, expect); +} + +TEST_P(IPUnboundSocketTest, InvalidNegativeTOS) { + auto socket = ASSERT_NO_ERRNO_AND_VALUE(NewSocket()); + int set = -2; + socklen_t set_sz = sizeof(set); + TOSOption t = GetTOSOption(GetParam().domain); + int expect; + if (GetParam().domain == AF_INET) { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallSucceedsWithValue(0)); + expect = static_cast(set); + if (GetParam().protocol == IPPROTO_TCP) { + expect &= ~INET_ECN_MASK; + } + } else { + EXPECT_THAT(setsockopt(socket->get(), t.level, t.option, &set, set_sz), + SyscallFailsWithErrno(EINVAL)); + expect = 0; + } + int get = 0; + socklen_t get_sz = sizeof(get); + EXPECT_THAT(getsockopt(socket->get(), t.level, t.option, &get, &get_sz), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_sz, sizeof(get)); + EXPECT_EQ(get, expect); +} + INSTANTIATE_TEST_SUITE_P( IPUnboundSockets, IPUnboundSocketTest, ::testing::ValuesIn(VecCat(VecCat( -- cgit v1.2.3