EtherMacFullDuplex

Package: inet.linklayer.ethernet

EtherMacFullDuplex

compound module

Ethernet MAC which supports full-duplex operation ONLY. See the IEtherMac for general informations.

Most of today's Ethernet networks are switched, and operate in full-duplex mode. Full-duplex transmission can be used for point-to-point connections only. Since full-duplex connections cannot be shared, collisions are eliminated. This setup eliminates most of the need for the CSMA/CD access control mechanism because there is no need to determine whether the connection is already being used. This allows for a much simpler simulation model for MAC. (In "traditional" Ethernet simulations, most of the code deals with the shared medium and the CSMA/CD mechanism.) EtherMacFullDuplex implements Ethernet without shared medium and CSMA/CD. (If you need half-duplex operation, see EtherMac which is for a full-blown and therefore more complicated Ethernet MAC model.)

EtherMacFullDuplex performs transmission and reception of frames. It does not do encapsulation/decapsulation; see Ieee8022Llc and EtherEncap for that.

Supported variations:

  • 10Mb Ethernet (full-duplex mode)
  • 100Mb Ethernet (full-duplex mode)
  • 1Gb Ethernet (full-duplex mode)
  • 10Gb Ethernet (full-duplex mode)
  • 40Gb Ethernet (full-duplex mode)
  • 100Gb Ethernet (full-duplex mode)

Supports all three Ethernet frame types. (It handles EtherFrame message class; specific frame classes (Ethernet-II, IEEE 802.3) are subclassed from that one.) RAW mode (only used by the IPX protocol) is not supported.

Operation

Processing of frames received from higher layers:

  • if src address in the frame is empty, fill it out
  • frames get queued up until transmission
  • transmits the frame when the transmitter is ready
  • can send PAUSE message if requested by higher layers (PAUSE protocol, used in switches).

Processing of frames incoming from the network:

  • receive frame from the network
  • CRC checking (frames with the error bit set are discarded).
  • respond to PAUSE frames
  • in promiscuous mode, pass up all received frames; otherwise, only frames with matching MAC addresses and broadcast frames are passed up.

The module does not perform encapsulation or decapsulation of frames -- this is done by higher layers (EtherEncap).

When a frame is received from the higher layers, it must be an EtherFrame, and with all protocol fields filled out (including the destination MAC address). The source address, if left empty, will be filled in. Then frame is queued and transmitted.

Data frames received from the network are EtherFrames. They are passed to the higher layers without modification. Also, the module properly responds to PAUSE frames, but never sends them by itself -- however, it transmits PAUSE frames received from upper layers. See PAUSE handling for more info.

For more info see Ethernet Model Overview.

Disabling and disconnecting

If the MAC is not connected to the network ("cable unplugged"), it will start up in "disabled" mode. A disabled MAC simply discards any messages it receives. It is currently not supported to dynamically connect/disconnect a MAC.

Queueing

In routers, MAC relies on an external queue module (see IPacketQueue) to model finite buffer, implement QoS and/or RED, and requests packets from this external queue one-by-one.

In hosts, no such queue is used, so MAC contains an internal queue to store packets waiting for transmission. Conceptually, the queue is of infinite size, but for better diagnostics one can specify a hard limit in the queue parameter -- if this is exceeded, the simulation stops with an error.

Physical layer messaging

Please see Messaging on the physical layer.

See also: EtherMac, EthernetInterface, IPacketQueue, EtherEncap, Ieee8022Llc

See also: EtherFrame, EthernetIIFrame, EtherFrameWithLlc, Ieee802Ctrl

IPacketQueue

Usage diagram

The following diagram shows usage relationships between types. Unresolved types are missing from the diagram.

Inheritance diagram

The following diagram shows inheritance relationships for this type. Unresolved types are missing from the diagram.

Parameters

Name Type Default value Description
interfaceTableModule string

The path to the InterfaceTable module

displayStringTextFormat string "rate: %b\nsent: %s, rcvd: %r\nqueue: %q, drop: %d"
sendRawBytes bool false

when true packets are serialized into a sequence of bytes before sending out

promiscuous bool false

if true, all packets are received, otherwise only the ones with matching destination MAC address

duplexMode bool true

must be set to "true", as EtherMacFullDuplex does not support half-duplex operation (parameter is present to reduce the risk of accidental misconfiguration)

mtu int 1500B
stopOperationExtraTime double -1s

extra time after lifecycle stop operation finished

stopOperationTimeout double 2s

timeout value for lifecycle stop operation

Properties

Name Value Description
lifecycleSupport
class EtherMacFullDuplex
display i=block/rxtx

Gates

Name Direction Size Description
upperLayerIn input

to EtherEncap or IMacRelayUnit

upperLayerOut output

to EtherEncap or IMacRelayUnit

phys inout

to physical layer or the network

Signals

Name Type Unit
receptionStateChanged long
rxPkFromHl inet::Packet
rxPkOk inet::Packet
packetReceivedFromLower inet::Packet
packetReceivedFromUpper inet::Packet
txPausePkUnits long
packetDropped inet::Packet
txPk inet::Packet
packetSentToLower inet::Packet
packetSentToUpper inet::Packet
rxPausePkUnits long
transmissionStateChanged long

Statistics

Name Title Source Record Unit Interpolation Mode
passedUpPk packets passed to higher layer packetSentToUpper count, sum(packetBytes), vector(packetBytes) none
packetDropInterfaceDown packet drop: interface down packetDropReasonIsInterfaceDown(packetDropped) count, sum(packetBytes), vector(packetBytes) none
rxPkOk packets received OK rxPkOk count, sum(packetBytes), vector(packetBytes) none
packetDropNotAddressedToUs packet drop: not addressed to us packetDropReasonIsNotAddressedToUs(packetDropped) count, sum(packetBytes), vector(packetBytes) none
packetReceivedFromUpper packet bytes from higher layer packetReceivedFromUpper count, sum(packetBytes), vector(packetBytes) none
txPausePkUnits pause units sent count, sum, vector none
txPk packets transmitted packetSentToLower count, sum(packetBytes), vector(packetBytes) none
rxPausePkUnits pause units received count, sum, vector none
packetDropIncorrectlyReceived packet drop: incorrectly received packetDropReasonIsIncorrectlyReceived(packetDropped) count, sum(packetBytes), vector(packetBytes) none

Source code

//
// Ethernet MAC which supports full-duplex operation ONLY.
// See the ~IEtherMac for general informations.
//
// Most of today's Ethernet networks are switched, and operate
// in full-duplex mode. Full-duplex transmission can be used for
// point-to-point connections only. Since full-duplex connections
// cannot be shared, collisions are eliminated. This setup eliminates
// most of the need for the CSMA/CD access control mechanism because
// there is no need to determine whether the connection is already
// being used. This allows for a much simpler simulation model
// for MAC. (In "traditional" Ethernet simulations, most of the code
// deals with the shared medium and the CSMA/CD mechanism.)
// ~EtherMacFullDuplex implements Ethernet without shared medium and CSMA/CD.
// (If you need half-duplex operation, see ~EtherMac which is for a full-blown
// and therefore more complicated Ethernet MAC model.)
//
// ~EtherMacFullDuplex performs transmission and reception of frames.
// It does not do encapsulation/decapsulation; see ~Ieee8022Llc and ~EtherEncap
// for that.
//
// Supported variations:
// - 10Mb Ethernet (full-duplex mode)
// - 100Mb Ethernet (full-duplex mode)
// - 1Gb Ethernet (full-duplex mode)
// - 10Gb Ethernet (full-duplex mode)
// - 40Gb Ethernet (full-duplex mode)
// - 100Gb Ethernet (full-duplex mode)
//
// Supports all three Ethernet frame types. (It handles ~EtherFrame message class;
// specific frame classes (Ethernet-II, IEEE 802.3) are subclassed from that one.)
// RAW mode (only used by the IPX protocol) is not supported.
//
// <b>Operation</b>
//
// Processing of frames received from higher layers:
// - if src address in the frame is empty, fill it out
// - frames get queued up until transmission
// - transmits the frame when the transmitter is ready
// - can send PAUSE message if requested by higher layers (PAUSE protocol,
//   used in switches).
//
// Processing of frames incoming from the network:
// - receive frame from the network
// - CRC checking (frames with the error bit set are discarded).
// - respond to PAUSE frames
// - in promiscuous mode, pass up all received frames;
//   otherwise, only frames with matching MAC addresses and
//   broadcast frames are passed up.
//
// The module does not perform encapsulation or decapsulation of frames --
// this is done by higher layers (~EtherEncap).
//
// When a frame is received from the higher layers, it must be an ~EtherFrame,
// and with all protocol fields filled out
// (including the destination MAC address). The source address, if left empty,
// will be filled in. Then frame is queued and transmitted.
//
// Data frames received from the network are EtherFrames. They are passed to
// the higher layers without modification.
// Also, the module properly responds to PAUSE frames, but never sends them
// by itself -- however, it transmits PAUSE frames received from upper layers.
// See <a href="ether-pause.html">PAUSE handling</a> for more info.
//
// For more info see <a href="ether-overview.html">Ethernet Model Overview</a>.
//
// <b>Disabling and disconnecting</b>
//
// If the MAC is not connected to the network ("cable unplugged"), it will
// start up in "disabled" mode. A disabled MAC simply discards any messages
// it receives. It is currently not supported to dynamically connect/disconnect
// a MAC.
//
//
// <b>Queueing</b>
//
// In routers, MAC relies on an external queue module (see ~IPacketQueue)
// to model finite buffer, implement QoS and/or RED, and requests packets
// from this external queue one-by-one.
//
// In hosts, no such queue is used, so MAC contains an internal
// queue to store packets waiting for transmission.
// Conceptually, the queue is of infinite size, but for better diagnostics
// one can specify a hard limit in the queue parameter -- if this is
// exceeded, the simulation stops with an error.
//
//
// <b>Physical layer messaging</b>
//
// Please see <a href="physical.html">Messaging on the physical layer</a>.
//
// @see ~EtherMac, ~EthernetInterface, ~IPacketQueue, ~EtherEncap, ~Ieee8022Llc
// @see ~EtherFrame, ~EthernetIIFrame, ~EtherFrameWithLlc, ~Ieee802Ctrl
//
module EtherMacFullDuplex like IEtherMac
{
    parameters:
        string interfaceTableModule;        // The path to the InterfaceTable module
        string displayStringTextFormat = default("rate: %b\nsent: %s, rcvd: %r\nqueue: %q, drop: %d");
        bool sendRawBytes = default(false); // when true packets are serialized into a sequence of bytes before sending out
        bool promiscuous = default(false);  // if true, all packets are received, otherwise only the
                                            // ones with matching destination MAC address
        bool duplexMode = default(true);    // must be set to "true", as EtherMacFullDuplex does not support half-duplex operation
                                            // (parameter is present to reduce the risk of accidental misconfiguration)
        int mtu @unit(B) = default(1500B);
        @lifecycleSupport;
        double stopOperationExtraTime @unit(s) = default(-1s);    // extra time after lifecycle stop operation finished
        double stopOperationTimeout @unit(s) = default(2s);    // timeout value for lifecycle stop operation
        @class(EtherMacFullDuplex);
        @display("i=block/rxtx");

        @signal[txPk](type=inet::Packet);
        @signal[rxPkOk](type=inet::Packet);
        @signal[txPausePkUnits](type=long);
        @signal[rxPausePkUnits](type=long);
        @signal[rxPkFromHl](type=inet::Packet);
        @signal[packetDropped](type=inet::Packet);
        @signal[packetSentToLower](type=inet::Packet);
        @signal[packetReceivedFromLower](type=inet::Packet);
        @signal[packetSentToUpper](type=inet::Packet);
        @signal[packetReceivedFromUpper](type=inet::Packet);
        @signal[transmissionStateChanged](type=long); // enum=MacTransmitState
        @signal[receptionStateChanged](type=long); // enum=MacReceiveState

        @statistic[txPk](title="packets transmitted"; source=packetSentToLower; record=count,"sum(packetBytes)","vector(packetBytes)"; interpolationmode=none);
        @statistic[rxPkOk](title="packets received OK"; source=rxPkOk; record=count,"sum(packetBytes)","vector(packetBytes)"; interpolationmode=none);
        @statistic[passedUpPk](title="packets passed to higher layer"; source=packetSentToUpper; record=count,"sum(packetBytes)","vector(packetBytes)"; interpolationmode=none);
        @statistic[txPausePkUnits](title="pause units sent"; record=count,sum,vector; interpolationmode=none);
        @statistic[rxPausePkUnits](title="pause units received"; record=count,sum,vector; interpolationmode=none);
        @statistic[packetReceivedFromUpper](title="packet bytes from higher layer"; source=packetReceivedFromUpper; record=count,"sum(packetBytes)","vector(packetBytes)"; interpolationmode=none);
        @statistic[packetDropIncorrectlyReceived](title="packet drop: incorrectly received"; source=packetDropReasonIsIncorrectlyReceived(packetDropped); record=count,sum(packetBytes),vector(packetBytes); interpolationmode=none);
        @statistic[packetDropInterfaceDown](title="packet drop: interface down"; source=packetDropReasonIsInterfaceDown(packetDropped); record=count,sum(packetBytes),vector(packetBytes); interpolationmode=none);
        @statistic[packetDropNotAddressedToUs](title="packet drop: not addressed to us"; source=packetDropReasonIsNotAddressedToUs(packetDropped); record=count,sum(packetBytes),vector(packetBytes); interpolationmode=none);

    gates:
        input upperLayerIn @labels(EtherFrame);    // to ~EtherEncap or ~IMacRelayUnit
        output upperLayerOut @labels(EtherFrame);  // to ~EtherEncap or ~IMacRelayUnit
        inout phys @labels(EtherFrame);            // to physical layer or the network
    submodules:
        queue: <default("EtherQueue")> like IPacketQueue {
            parameters:
                packetCapacity = default(1000);
                @display("p=100,100");
        }
}

File: src/inet/linklayer/ethernet/EtherMacFullDuplex.ned