diff options
author | Ryan Heacock <rheacock@google.com> | 2019-12-24 08:48:14 -0800 |
---|---|---|
committer | gVisor bot <gvisor-bot@google.com> | 2019-12-24 08:49:39 -0800 |
commit | e013c48c78c9a7daf245b7de9563e3a0bd8a1e97 (patch) | |
tree | 4569d1851e72e61faebc63c82997555afdde04a9 | |
parent | 574e988f2bc6060078a17f37a377441703c52a22 (diff) |
Enable IP_RECVTOS socket option for datagram sockets
Added the ability to get/set the IP_RECVTOS socket option on UDP endpoints. If
enabled, TOS from the incoming Network Header passed as ancillary data in the
ControlMessages.
Test:
* Added unit test to udp_test.go that tests getting/setting as well as
verifying that we receive expected TOS from incoming packet.
* Added a syscall test
PiperOrigin-RevId: 287029703
-rw-r--r-- | pkg/sentry/socket/control/control.go | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/netstack/netstack.go | 42 | ||||
-rw-r--r-- | pkg/tcpip/checker/checker.go | 16 | ||||
-rw-r--r-- | pkg/tcpip/stack/nic.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/stack/stack.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/tcpip.go | 6 | ||||
-rw-r--r-- | pkg/tcpip/transport/raw/endpoint.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/transport/udp/endpoint.go | 31 | ||||
-rw-r--r-- | pkg/tcpip/transport/udp/udp_test.go | 69 | ||||
-rw-r--r-- | test/syscalls/linux/socket_ip_udp_generic.cc | 40 | ||||
-rw-r--r-- | test/syscalls/linux/udp_socket_test_cases.cc | 8 |
11 files changed, 201 insertions, 19 deletions
diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index af1a4e95f..b649dd021 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -327,7 +327,7 @@ func PackInq(t *kernel.Task, inq int32, buf []byte) []byte { } // PackTOS packs an IP_TOS socket control message. -func PackTOS(t *kernel.Task, tos int8, buf []byte) []byte { +func PackTOS(t *kernel.Task, tos uint8, buf []byte) []byte { return putCmsgStruct( buf, linux.SOL_IP, diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go index 140851c17..d2f263402 100644 --- a/pkg/sentry/socket/netstack/netstack.go +++ b/pkg/sentry/socket/netstack/netstack.go @@ -1323,6 +1323,21 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int, family in } return int32(v), nil + case linux.IP_RECVTOS: + if outLen < sizeOfInt32 { + return nil, syserr.ErrInvalidArgument + } + + var v tcpip.ReceiveTOSOption + if err := ep.GetSockOpt(&v); err != nil { + return nil, syserr.TranslateNetstackError(err) + } + + if v { + return int32(1), nil + } + return int32(0), nil + default: emitUnimplementedEventIP(t, name) } @@ -1808,6 +1823,16 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s } return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.IPv4TOSOption(v))) + case linux.IP_RECVTOS: + v, err := parseIntOrChar(optVal) + if err != nil { + return err + } + + return syserr.TranslateNetstackError(ep.SetSockOpt( + tcpip.ReceiveTOSOption(v != 0), + )) + case linux.IP_ADD_SOURCE_MEMBERSHIP, linux.IP_BIND_ADDRESS_NO_PORT, linux.IP_BLOCK_SOURCE, @@ -1828,7 +1853,6 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s linux.IP_RECVFRAGSIZE, linux.IP_RECVOPTS, linux.IP_RECVORIGDSTADDR, - linux.IP_RECVTOS, linux.IP_RECVTTL, linux.IP_RETOPTS, linux.IP_TRANSPARENT, @@ -2139,6 +2163,21 @@ func (s *SocketOperations) fillCmsgInq(cmsg *socket.ControlMessages) { cmsg.IP.Inq = int32(len(s.readView) + rcvBufUsed) } +func (s *SocketOperations) fillCmsgTOS(cmsg *socket.ControlMessages) { + if s.skType != linux.SOCK_DGRAM { + return + } + var receiveTOS tcpip.ReceiveTOSOption + if err := s.Endpoint.GetSockOpt(&receiveTOS); err != nil { + return + } + if !receiveTOS { + return + } + cmsg.IP.HasTOS = s.readCM.HasTOS + cmsg.IP.TOS = s.readCM.TOS +} + // nonBlockingRead issues a non-blocking read. // // TODO(b/78348848): Support timestamps for stream sockets. @@ -2244,6 +2283,7 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe cmsg := s.controlMessages() s.fillCmsgInq(&cmsg) + s.fillCmsgTOS(&cmsg) return n, flags, addr, addrLen, cmsg, syserr.FromError(err) } diff --git a/pkg/tcpip/checker/checker.go b/pkg/tcpip/checker/checker.go index 2f15bf1f1..542abc99d 100644 --- a/pkg/tcpip/checker/checker.go +++ b/pkg/tcpip/checker/checker.go @@ -33,6 +33,9 @@ type NetworkChecker func(*testing.T, []header.Network) // TransportChecker is a function to check a property of a transport packet. type TransportChecker func(*testing.T, header.Transport) +// ControlMessagesChecker is a function to check a property of ancillary data. +type ControlMessagesChecker func(*testing.T, tcpip.ControlMessages) + // IPv4 checks the validity and properties of the given IPv4 packet. It is // expected to be used in conjunction with other network checkers for specific // properties. For example, to check the source and destination address, one @@ -158,6 +161,19 @@ func FragmentFlags(flags uint8) NetworkChecker { } } +// ReceiveTOS creates a checker that checks the TOS field in ControlMessages. +func ReceiveTOS(want uint8) ControlMessagesChecker { + return func(t *testing.T, cm tcpip.ControlMessages) { + t.Helper() + if !cm.HasTOS { + t.Fatalf("got cm.HasTOS = %t, want cm.TOS = %d", cm.HasTOS, want) + } + if got := cm.TOS; got != want { + t.Fatalf("got cm.TOS = %d, want %d", got, want) + } + } +} + // TOS creates a checker that checks the TOS field. func TOS(tos uint8, label uint32) NetworkChecker { return func(t *testing.T, h []header.Network) { diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index ddd014658..a4556674b 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -575,7 +575,7 @@ func (n *NIC) RemoveAddressRange(subnet tcpip.Subnet) { n.mu.Unlock() } -// Subnets returns the Subnets associated with this NIC. +// AddressRanges returns the Subnets associated with this NIC. func (n *NIC) AddressRanges() []tcpip.Subnet { n.mu.RLock() defer n.mu.RUnlock() diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index 7a9600679..251336224 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -829,7 +829,7 @@ func (s *Stack) CheckNIC(id tcpip.NICID) bool { return false } -// NICSubnets returns a map of NICIDs to their associated subnets. +// NICAddressRanges returns a map of NICIDs to their associated subnets. func (s *Stack) NICAddressRanges() map[tcpip.NICID][]tcpip.Subnet { s.mu.RLock() defer s.mu.RUnlock() diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index f62fd729f..5c7b2af88 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -322,7 +322,7 @@ type ControlMessages struct { HasTOS bool // TOS is the IPv4 type of service of the associated packet. - TOS int8 + TOS uint8 // HasTClass indicates whether Tclass is valid/set. HasTClass bool @@ -666,6 +666,10 @@ type IPv4TOSOption uint8 // for all subsequent outgoing IPv6 packets from the endpoint. type IPv6TrafficClassOption uint8 +// ReceiveTOSOption is used by SetSockOpt/GetSockOpt to specify if the TOS +// ancillary message is passed with incoming packets. +type ReceiveTOSOption bool + // 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/raw/endpoint.go b/pkg/tcpip/transport/raw/endpoint.go index 5aafe2615..6d23ab5a1 100644 --- a/pkg/tcpip/transport/raw/endpoint.go +++ b/pkg/tcpip/transport/raw/endpoint.go @@ -510,7 +510,7 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { } // SetSockOptInt implements tcpip.Endpoint.SetSockOptInt. -func (ep *endpoint) SetSockOptInt(opt tcpip.SockOpt, v int) *tcpip.Error { +func (e *endpoint) SetSockOptInt(opt tcpip.SockOpt, v int) *tcpip.Error { return tcpip.ErrUnknownProtocolOption } diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 1ac4705af..269470ed4 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -32,6 +32,7 @@ type udpPacket struct { senderAddress tcpip.FullAddress data buffer.VectorisedView `state:".(buffer.VectorisedView)"` timestamp int64 + tos uint8 } // EndpointState represents the state of a UDP endpoint. @@ -114,6 +115,10 @@ type endpoint struct { // applied while sending packets. Defaults to 0 as on Linux. sendTOS uint8 + // receiveTOS determines if the incoming IPv4 TOS header field is passed + // as ancillary data to ControlMessages on Read. + receiveTOS bool + // shutdownFlags represent the current shutdown state of the endpoint. shutdownFlags tcpip.ShutdownFlags @@ -244,7 +249,12 @@ func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMess *addr = p.senderAddress } - return p.data.ToView(), tcpip.ControlMessages{HasTimestamp: true, Timestamp: p.timestamp}, nil + return p.data.ToView(), tcpip.ControlMessages{ + HasTimestamp: true, + Timestamp: p.timestamp, + HasTOS: e.receiveTOS, + TOS: p.tos, + }, nil } // prepareForWrite prepares the endpoint for sending data. In particular, it @@ -656,6 +666,12 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { e.sendTOS = uint8(v) e.mu.Unlock() return nil + + case tcpip.ReceiveTOSOption: + e.mu.Lock() + e.receiveTOS = bool(v) + e.mu.Unlock() + return nil } return nil } @@ -792,6 +808,12 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { e.mu.RUnlock() return nil + case *tcpip.ReceiveTOSOption: + e.mu.RLock() + *o = tcpip.ReceiveTOSOption(e.receiveTOS) + e.mu.RUnlock() + return nil + default: return tcpip.ErrUnknownProtocolOption } @@ -1238,6 +1260,13 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pk e.rcvList.PushBack(packet) e.rcvBufSize += pkt.Data.Size() + // Save any useful information from the NetworkHeader to the packet. + switch r.NetProto { + case header.IPv4ProtocolNumber: + // This packet has already been validated before being passed up the stack. + packet.tos, _ = header.IPv4(pkt.NetworkHeader).TOS() + } + packet.timestamp = e.stack.NowNanoseconds() e.rcvMu.Unlock() diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index 7051a7a9c..43b8b35ba 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -56,6 +56,7 @@ const ( multicastAddr = "\xe8\x2b\xd3\xea" multicastV6Addr = "\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" broadcastAddr = header.IPv4Broadcast + testTOS = 0x80 // defaultMTU is the MTU, in bytes, used throughout the tests, except // where another value is explicitly used. It is chosen to match the MTU @@ -453,6 +454,7 @@ func (c *testContext) injectV4Packet(payload []byte, h *header4Tuple, valid bool ip := header.IPv4(buf) ip.Encode(&header.IPv4Fields{ IHL: header.IPv4MinimumSize, + TOS: testTOS, TotalLength: uint16(len(buf)), TTL: 65, Protocol: uint8(udp.ProtocolNumber), @@ -556,8 +558,8 @@ func TestBindToDeviceOption(t *testing.T) { // testReadInternal sends a packet of the given test flow into the stack by // injecting it into the link endpoint. It then attempts to read it from the // UDP endpoint and depending on if this was expected to succeed verifies its -// correctness. -func testReadInternal(c *testContext, flow testFlow, packetShouldBeDropped, expectReadError bool) { +// correctness including any additional checker functions provided. +func testReadInternal(c *testContext, flow testFlow, packetShouldBeDropped, expectReadError bool, checkers ...checker.ControlMessagesChecker) { c.t.Helper() payload := newPayload() @@ -572,12 +574,12 @@ func testReadInternal(c *testContext, flow testFlow, packetShouldBeDropped, expe epstats := c.ep.Stats().(*tcpip.TransportEndpointStats).Clone() var addr tcpip.FullAddress - v, _, err := c.ep.Read(&addr) + v, cm, err := c.ep.Read(&addr) if err == tcpip.ErrWouldBlock { // Wait for data to become available. select { case <-ch: - v, _, err = c.ep.Read(&addr) + v, cm, err = c.ep.Read(&addr) case <-time.After(300 * time.Millisecond): if packetShouldBeDropped { @@ -610,15 +612,21 @@ func testReadInternal(c *testContext, flow testFlow, packetShouldBeDropped, expe if !bytes.Equal(payload, v) { c.t.Fatalf("bad payload: got %x, want %x", v, payload) } + + // Run any checkers against the ControlMessages. + for _, f := range checkers { + f(c.t, cm) + } + c.checkEndpointReadStats(1, epstats, err) } // testRead sends a packet of the given test flow into the stack by injecting it // into the link endpoint. It then reads it from the UDP endpoint and verifies -// its correctness. -func testRead(c *testContext, flow testFlow) { +// its correctness including any additional checker functions provided. +func testRead(c *testContext, flow testFlow, checkers ...checker.ControlMessagesChecker) { c.t.Helper() - testReadInternal(c, flow, false /* packetShouldBeDropped */, false /* expectReadError */) + testReadInternal(c, flow, false /* packetShouldBeDropped */, false /* expectReadError */, checkers...) } // testFailingRead sends a packet of the given test flow into the stack by @@ -1286,7 +1294,7 @@ func TestTOSV4(t *testing.T) { c.createEndpointForFlow(flow) - const tos = 0xC0 + const tos = testTOS var v tcpip.IPv4TOSOption if err := c.ep.GetSockOpt(&v); err != nil { c.t.Errorf("GetSockopt failed: %s", err) @@ -1321,7 +1329,7 @@ func TestTOSV6(t *testing.T) { c.createEndpointForFlow(flow) - const tos = 0xC0 + const tos = testTOS var v tcpip.IPv6TrafficClassOption if err := c.ep.GetSockOpt(&v); err != nil { c.t.Errorf("GetSockopt failed: %s", err) @@ -1348,6 +1356,49 @@ func TestTOSV6(t *testing.T) { } } +func TestReceiveTOSV4(t *testing.T) { + for _, flow := range []testFlow{unicastV4, broadcast} { + t.Run(fmt.Sprintf("flow:%s", flow), func(t *testing.T) { + c := newDualTestContext(t, defaultMTU) + defer c.cleanup() + + c.createEndpointForFlow(flow) + + // Verify that setting and reading the option works. + const recvTos = true + var v tcpip.ReceiveTOSOption + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + // Test for expected default value. + if v != false { + c.t.Errorf("got GetSockOpt(...) = %t, want = %t", v, false) + } + + if err := c.ep.SetSockOpt(tcpip.ReceiveTOSOption(recvTos)); err != nil { + c.t.Errorf("SetSockOpt(%#v) failed: %s", tcpip.ReceiveTOSOption(recvTos), err) + } + + if err := c.ep.GetSockOpt(&v); err != nil { + c.t.Errorf("GetSockopt failed: %s", err) + } + + if want := tcpip.ReceiveTOSOption(recvTos); v != want { + c.t.Errorf("got GetSockOpt(...) = %t, want = %t", v, want) + } + + // Bind to wildcard. + if err := c.ep.Bind(tcpip.FullAddress{Port: stackPort}); err != nil { + c.t.Fatalf("Bind failed: %s", err) + } + + // Verify that the correct received TOS is actually handed through as + // ancillary data to the ControlMessages struct. + testRead(c, flow, checker.ReceiveTOS(testTOS)) + }) + } +} + 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_udp_generic.cc b/test/syscalls/linux/socket_ip_udp_generic.cc index 66eb68857..53290bed7 100644 --- a/test/syscalls/linux/socket_ip_udp_generic.cc +++ b/test/syscalls/linux/socket_ip_udp_generic.cc @@ -209,6 +209,46 @@ TEST_P(UDPSocketPairTest, SetMulticastLoopChar) { EXPECT_EQ(get, kSockOptOn); } +// Ensure that Receiving TOS is off by default. +TEST_P(UDPSocketPairTest, RecvTosDefault) { + auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); + + int get = -1; + socklen_t get_len = sizeof(get); + ASSERT_THAT( + getsockopt(sockets->first_fd(), IPPROTO_IP, IP_RECVTOS, &get, &get_len), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_len, sizeof(get)); + EXPECT_EQ(get, kSockOptOff); +} + +// Test that setting and getting IP_RECVTOS works as expected. +TEST_P(UDPSocketPairTest, SetRecvTos) { + auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); + + ASSERT_THAT(setsockopt(sockets->first_fd(), IPPROTO_IP, IP_RECVTOS, + &kSockOptOff, sizeof(kSockOptOff)), + SyscallSucceeds()); + + int get = -1; + socklen_t get_len = sizeof(get); + ASSERT_THAT( + getsockopt(sockets->first_fd(), IPPROTO_IP, IP_RECVTOS, &get, &get_len), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_len, sizeof(get)); + EXPECT_EQ(get, kSockOptOff); + + ASSERT_THAT(setsockopt(sockets->first_fd(), IPPROTO_IP, IP_RECVTOS, + &kSockOptOn, sizeof(kSockOptOn)), + SyscallSucceeds()); + + ASSERT_THAT( + getsockopt(sockets->first_fd(), IPPROTO_IP, IP_RECVTOS, &get, &get_len), + SyscallSucceedsWithValue(0)); + EXPECT_EQ(get_len, sizeof(get)); + EXPECT_EQ(get, kSockOptOn); +} + TEST_P(UDPSocketPairTest, ReuseAddrDefault) { auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); diff --git a/test/syscalls/linux/udp_socket_test_cases.cc b/test/syscalls/linux/udp_socket_test_cases.cc index dc35c2f50..68e0a8109 100644 --- a/test/syscalls/linux/udp_socket_test_cases.cc +++ b/test/syscalls/linux/udp_socket_test_cases.cc @@ -1349,8 +1349,9 @@ TEST_P(UdpSocketTest, TimestampIoctlPersistence) { // outgoing packets, and that a receiving socket with IP_RECVTOS or // IPV6_RECVTCLASS will create the corresponding control message. TEST_P(UdpSocketTest, SetAndReceiveTOS) { - // TODO(b/68320120): IP_RECVTOS/IPV6_RECVTCLASS not supported for netstack. - SKIP_IF(IsRunningOnGvisor() && !IsRunningWithHostinet()); + // TODO(b/68320120): IPV6_RECVTCLASS not supported for netstack. + SKIP_IF((GetParam() != AddressFamily::kIpv4) && IsRunningOnGvisor() && + !IsRunningWithHostinet()); ASSERT_THAT(bind(s_, addr_[0], addrlen_), SyscallSucceeds()); ASSERT_THAT(connect(t_, addr_[0], addrlen_), SyscallSucceeds()); @@ -1421,7 +1422,8 @@ TEST_P(UdpSocketTest, SetAndReceiveTOS) { // TOS byte on outgoing packets, and that a receiving socket with IP_RECVTOS or // IPV6_RECVTCLASS will create the corresponding control message. TEST_P(UdpSocketTest, SendAndReceiveTOS) { - // TODO(b/68320120): IP_RECVTOS/IPV6_RECVTCLASS not supported for netstack. + // TODO(b/68320120): IPV6_RECVTCLASS not supported for netstack. + // TODO(b/146661005): Setting TOS via cmsg not supported for netstack. SKIP_IF(IsRunningOnGvisor() && !IsRunningWithHostinet()); ASSERT_THAT(bind(s_, addr_[0], addrlen_), SyscallSucceeds()); ASSERT_THAT(connect(t_, addr_[0], addrlen_), SyscallSucceeds()); |