diff options
Diffstat (limited to 'pkg/tcpip/transport/udp/endpoint.go')
-rw-r--r-- | pkg/tcpip/transport/udp/endpoint.go | 146 |
1 files changed, 94 insertions, 52 deletions
diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 6e87245b7..1ac4705af 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -21,6 +21,7 @@ import ( "gvisor.dev/gvisor/pkg/tcpip/buffer" "gvisor.dev/gvisor/pkg/tcpip/header" "gvisor.dev/gvisor/pkg/tcpip/iptables" + "gvisor.dev/gvisor/pkg/tcpip/ports" "gvisor.dev/gvisor/pkg/tcpip/stack" "gvisor.dev/gvisor/pkg/waiter" ) @@ -31,9 +32,6 @@ type udpPacket struct { senderAddress tcpip.FullAddress data buffer.VectorisedView `state:".(buffer.VectorisedView)"` timestamp int64 - // views is used as buffer for data when its length is large - // enough to store a VectorisedView. - views [8]buffer.View `state:"nosave"` } // EndpointState represents the state of a UDP endpoint. @@ -80,6 +78,7 @@ type endpoint struct { // change throughout the lifetime of the endpoint. stack *stack.Stack `state:"manual"` waiterQueue *waiter.Queue + uniqueID uint64 // The following fields are used to manage the receive queue, and are // protected by rcvMu. @@ -106,6 +105,11 @@ type endpoint struct { bindToDevice tcpip.NICID broadcast bool + // Values used to reserve a port or register a transport endpoint. + // (which ever happens first). + boundBindToDevice tcpip.NICID + boundPortFlags ports.Flags + // sendTOS represents IPv4 TOS or IPv6 TrafficClass, // applied while sending packets. Defaults to 0 as on Linux. sendTOS uint8 @@ -140,7 +144,7 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue stack: s, TransportEndpointInfo: stack.TransportEndpointInfo{ NetProto: netProto, - TransProto: header.TCPProtocolNumber, + TransProto: header.UDPProtocolNumber, }, waiterQueue: waiterQueue, // RFC 1075 section 5.4 recommends a TTL of 1 for membership @@ -160,9 +164,15 @@ func newEndpoint(s *stack.Stack, netProto tcpip.NetworkProtocolNumber, waiterQue rcvBufSizeMax: 32 * 1024, sndBufSize: 32 * 1024, state: StateInitial, + uniqueID: s.UniqueID(), } } +// UniqueID implements stack.TransportEndpoint.UniqueID. +func (e *endpoint) UniqueID() uint64 { + return e.uniqueID +} + // Close puts the endpoint in a closed state and frees all resources // associated with it. func (e *endpoint) Close() { @@ -171,8 +181,10 @@ func (e *endpoint) Close() { switch e.state { case StateBound, StateConnected: - e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.bindToDevice) - e.stack.ReleasePort(e.effectiveNetProtos, ProtocolNumber, e.ID.LocalAddress, e.ID.LocalPort, e.bindToDevice) + e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.boundBindToDevice) + e.stack.ReleasePort(e.effectiveNetProtos, ProtocolNumber, e.ID.LocalAddress, e.ID.LocalPort, e.boundPortFlags, e.boundBindToDevice) + e.boundBindToDevice = 0 + e.boundPortFlags = ports.Flags{} } for _, mem := range e.multicastMemberships { @@ -278,7 +290,7 @@ func (e *endpoint) prepareForWrite(to *tcpip.FullAddress) (retry bool, err *tcpi // connectRoute establishes a route to the specified interface or the // configured multicast interface if no interface is specified and the // specified address is a multicast address. -func (e *endpoint) connectRoute(nicid tcpip.NICID, addr tcpip.FullAddress, netProto tcpip.NetworkProtocolNumber) (stack.Route, tcpip.NICID, *tcpip.Error) { +func (e *endpoint) connectRoute(nicID tcpip.NICID, addr tcpip.FullAddress, netProto tcpip.NetworkProtocolNumber) (stack.Route, tcpip.NICID, *tcpip.Error) { localAddr := e.ID.LocalAddress if isBroadcastOrMulticast(localAddr) { // A packet can only originate from a unicast address (i.e., an interface). @@ -286,20 +298,20 @@ func (e *endpoint) connectRoute(nicid tcpip.NICID, addr tcpip.FullAddress, netPr } if header.IsV4MulticastAddress(addr.Addr) || header.IsV6MulticastAddress(addr.Addr) { - if nicid == 0 { - nicid = e.multicastNICID + if nicID == 0 { + nicID = e.multicastNICID } - if localAddr == "" && nicid == 0 { + if localAddr == "" && nicID == 0 { localAddr = e.multicastAddr } } // Find a route to the desired destination. - r, err := e.stack.FindRoute(nicid, localAddr, addr.Addr, netProto, e.multicastLoop) + r, err := e.stack.FindRoute(nicID, localAddr, addr.Addr, netProto, e.multicastLoop) if err != nil { return stack.Route{}, 0, err } - return r, nicid, nil + return r, nicID, nil } // Write writes data to the endpoint's peer. This method does not block @@ -378,13 +390,13 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c } else { // Reject destination address if it goes through a different // NIC than the endpoint was bound to. - nicid := to.NIC + nicID := to.NIC if e.BindNICID != 0 { - if nicid != 0 && nicid != e.BindNICID { + if nicID != 0 && nicID != e.BindNICID { return 0, nil, tcpip.ErrNoRoute } - nicid = e.BindNICID + nicID = e.BindNICID } if to.Addr == header.IPv4Broadcast && !e.broadcast { @@ -396,7 +408,7 @@ func (e *endpoint) write(p tcpip.Payloader, opts tcpip.WriteOptions) (int64, <-c return 0, nil, err } - r, _, err := e.connectRoute(nicid, *to, netProto) + r, _, err := e.connectRoute(nicID, *to, netProto) if err != nil { return 0, nil, err } @@ -618,9 +630,9 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { e.bindToDevice = 0 return nil } - for nicid, nic := range e.stack.NICInfo() { + for nicID, nic := range e.stack.NICInfo() { if nic.Name == string(v) { - e.bindToDevice = nicid + e.bindToDevice = nicID return nil } } @@ -728,6 +740,10 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { *o = tcpip.MulticastLoopOption(v) return nil + case *tcpip.ReuseAddressOption: + *o = 0 + return nil + case *tcpip.ReusePortOption: e.mu.RLock() v := e.reusePort @@ -809,7 +825,11 @@ func sendUDP(r *stack.Route, data buffer.VectorisedView, localPort, remotePort u if useDefaultTTL { ttl = r.DefaultTTL() } - if err := r.WritePacket(nil /* gso */, hdr, data, stack.NetworkHeaderParams{Protocol: ProtocolNumber, TTL: ttl, TOS: tos}); err != nil { + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: ProtocolNumber, TTL: ttl, TOS: tos}, tcpip.PacketBuffer{ + Header: hdr, + Data: data, + TransportHeader: buffer.View(udp), + }); err != nil { r.Stats().UDP.PacketSendErrors.Increment() return err } @@ -859,7 +879,10 @@ func (e *endpoint) Disconnect() *tcpip.Error { if e.state != StateConnected { return nil } - id := stack.TransportEndpointID{} + var ( + id stack.TransportEndpointID + btd tcpip.NICID + ) // Exclude ephemerally bound endpoints. if e.BindNICID != 0 || e.ID.LocalAddress == "" { var err *tcpip.Error @@ -867,7 +890,7 @@ func (e *endpoint) Disconnect() *tcpip.Error { LocalPort: e.ID.LocalPort, LocalAddress: e.ID.LocalAddress, } - id, err = e.registerWithStack(e.RegisterNICID, e.effectiveNetProtos, id) + id, btd, err = e.registerWithStack(e.RegisterNICID, e.effectiveNetProtos, id) if err != nil { return err } @@ -875,13 +898,15 @@ func (e *endpoint) Disconnect() *tcpip.Error { } else { if e.ID.LocalPort != 0 { // Release the ephemeral port. - e.stack.ReleasePort(e.effectiveNetProtos, ProtocolNumber, e.ID.LocalAddress, e.ID.LocalPort, e.bindToDevice) + e.stack.ReleasePort(e.effectiveNetProtos, ProtocolNumber, e.ID.LocalAddress, e.ID.LocalPort, e.boundPortFlags, e.boundBindToDevice) + e.boundPortFlags = ports.Flags{} } e.state = StateInitial } - e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.bindToDevice) + e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.boundBindToDevice) e.ID = id + e.boundBindToDevice = btd e.route.Release() e.route = stack.Route{} e.dstPort = 0 @@ -903,7 +928,7 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { e.mu.Lock() defer e.mu.Unlock() - nicid := addr.NIC + nicID := addr.NIC var localPort uint16 switch e.state { case StateInitial: @@ -913,16 +938,16 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { break } - if nicid != 0 && nicid != e.BindNICID { + if nicID != 0 && nicID != e.BindNICID { return tcpip.ErrInvalidEndpointState } - nicid = e.BindNICID + nicID = e.BindNICID default: return tcpip.ErrInvalidEndpointState } - r, nicid, err := e.connectRoute(nicid, addr, netProto) + r, nicID, err := e.connectRoute(nicID, addr, netProto) if err != nil { return err } @@ -950,20 +975,21 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { } } - id, err = e.registerWithStack(nicid, netProtos, id) + id, btd, err := e.registerWithStack(nicID, netProtos, id) if err != nil { return err } // Remove the old registration. if e.ID.LocalPort != 0 { - e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.bindToDevice) + e.stack.UnregisterTransportEndpoint(e.RegisterNICID, e.effectiveNetProtos, ProtocolNumber, e.ID, e, e.boundBindToDevice) } e.ID = id + e.boundBindToDevice = btd e.route = r.Clone() e.dstPort = addr.Port - e.RegisterNICID = nicid + e.RegisterNICID = nicID e.effectiveNetProtos = netProtos e.state = StateConnected @@ -1018,20 +1044,27 @@ func (*endpoint) Accept() (tcpip.Endpoint, *waiter.Queue, *tcpip.Error) { return nil, nil, tcpip.ErrNotSupported } -func (e *endpoint) registerWithStack(nicid tcpip.NICID, netProtos []tcpip.NetworkProtocolNumber, id stack.TransportEndpointID) (stack.TransportEndpointID, *tcpip.Error) { +func (e *endpoint) registerWithStack(nicID tcpip.NICID, netProtos []tcpip.NetworkProtocolNumber, id stack.TransportEndpointID) (stack.TransportEndpointID, tcpip.NICID, *tcpip.Error) { if e.ID.LocalPort == 0 { - port, err := e.stack.ReservePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort, e.reusePort, e.bindToDevice) + flags := ports.Flags{ + LoadBalanced: e.reusePort, + // FIXME(b/129164367): Support SO_REUSEADDR. + MostRecent: false, + } + port, err := e.stack.ReservePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort, flags, e.bindToDevice) if err != nil { - return id, err + return id, e.bindToDevice, err } + e.boundPortFlags = flags id.LocalPort = port } - err := e.stack.RegisterTransportEndpoint(nicid, netProtos, ProtocolNumber, id, e, e.reusePort, e.bindToDevice) + err := e.stack.RegisterTransportEndpoint(nicID, netProtos, ProtocolNumber, id, e, e.reusePort, e.bindToDevice) if err != nil { - e.stack.ReleasePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort, e.bindToDevice) + e.stack.ReleasePort(netProtos, ProtocolNumber, id.LocalAddress, id.LocalPort, e.boundPortFlags, e.bindToDevice) + e.boundPortFlags = ports.Flags{} } - return id, err + return id, e.bindToDevice, err } func (e *endpoint) bindLocked(addr tcpip.FullAddress) *tcpip.Error { @@ -1057,11 +1090,11 @@ func (e *endpoint) bindLocked(addr tcpip.FullAddress) *tcpip.Error { } } - nicid := addr.NIC + nicID := addr.NIC if len(addr.Addr) != 0 && !isBroadcastOrMulticast(addr.Addr) { // A local unicast address was specified, verify that it's valid. - nicid = e.stack.CheckLocalAddress(addr.NIC, netProto, addr.Addr) - if nicid == 0 { + nicID = e.stack.CheckLocalAddress(addr.NIC, netProto, addr.Addr) + if nicID == 0 { return tcpip.ErrBadLocalAddress } } @@ -1070,13 +1103,14 @@ func (e *endpoint) bindLocked(addr tcpip.FullAddress) *tcpip.Error { LocalPort: addr.Port, LocalAddress: addr.Addr, } - id, err = e.registerWithStack(nicid, netProtos, id) + id, btd, err := e.registerWithStack(nicID, netProtos, id) if err != nil { return err } e.ID = id - e.RegisterNICID = nicid + e.boundBindToDevice = btd + e.RegisterNICID = nicID e.effectiveNetProtos = netProtos // Mark endpoint as bound. @@ -1111,9 +1145,14 @@ func (e *endpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { e.mu.RLock() defer e.mu.RUnlock() + addr := e.ID.LocalAddress + if e.state == StateConnected { + addr = e.route.LocalAddress + } + return tcpip.FullAddress{ NIC: e.RegisterNICID, - Addr: e.ID.LocalAddress, + Addr: addr, Port: e.ID.LocalPort, }, nil } @@ -1154,17 +1193,17 @@ func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask { // HandlePacket is called by the stack when new packets arrive to this transport // endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, vv buffer.VectorisedView) { +func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, pkt tcpip.PacketBuffer) { // Get the header then trim it from the view. - hdr := header.UDP(vv.First()) - if int(hdr.Length()) > vv.Size() { + hdr := header.UDP(pkt.Data.First()) + if int(hdr.Length()) > pkt.Data.Size() { // Malformed packet. e.stack.Stats().UDP.MalformedPacketsReceived.Increment() e.stats.ReceiveErrors.MalformedPacketsReceived.Increment() return } - vv.TrimFront(header.UDPMinimumSize) + pkt.Data.TrimFront(header.UDPMinimumSize) e.rcvMu.Lock() e.stack.Stats().UDP.PacketsReceived.Increment() @@ -1188,18 +1227,18 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, vv wasEmpty := e.rcvBufSize == 0 // Push new packet into receive list and increment the buffer size. - pkt := &udpPacket{ + packet := &udpPacket{ senderAddress: tcpip.FullAddress{ NIC: r.NICID(), Addr: id.RemoteAddress, Port: hdr.SourcePort(), }, } - pkt.data = vv.Clone(pkt.views[:]) - e.rcvList.PushBack(pkt) - e.rcvBufSize += vv.Size() + packet.data = pkt.Data + e.rcvList.PushBack(packet) + e.rcvBufSize += pkt.Data.Size() - pkt.timestamp = e.stack.NowNanoseconds() + packet.timestamp = e.stack.NowNanoseconds() e.rcvMu.Unlock() @@ -1210,7 +1249,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, vv } // HandleControlPacket implements stack.TransportEndpoint.HandleControlPacket. -func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, vv buffer.VectorisedView) { +func (e *endpoint) HandleControlPacket(id stack.TransportEndpointID, typ stack.ControlType, extra uint32, pkt tcpip.PacketBuffer) { } // State implements tcpip.Endpoint.State. @@ -1234,6 +1273,9 @@ func (e *endpoint) Stats() tcpip.EndpointStats { return &e.stats } +// Wait implements tcpip.Endpoint.Wait. +func (*endpoint) Wait() {} + func isBroadcastOrMulticast(a tcpip.Address) bool { return a == header.IPv4Broadcast || header.IsV4MulticastAddress(a) || header.IsV6MulticastAddress(a) } |