GB2361394A - Conversion of point-to-point data packets to Ethernet data packets - Google Patents

Abstract

A network switch which includes a 'point-to-point' to Ethernet adapter, which provides a received data frame with a media access control destination address before the received data frame is forwarded by a forwarding engine. When the frame includes IP address data, the adapter employs the IP address to obtain the media access control destination address. When the frame includes IPX address data, the adapter determines the media access control destination address as an IPX destination address or a predetermined address according as a network number in said IPX address data identifies a local network or not.

Description

2361394 APPARATUS AND METHOD FOR THE CONVERSION OF POINT-TO-POINT DATA

PACKETS TO ETHERNET DATA PACKETS

Field of the Invention

This invention relates to data communication networks in which data is transmitted in addressed data packets. It more particularly relates to a bridge including means for convertinc, addressed packets having headers conforming to a point-to-point addressing format into packets having headers conforming to an Ethernet format and in particular including media access control source and destination addresses.

Two known network architectures employ an Ethernet format and a point-topoint protocol (PPP) format for the addressing of data frames (used herein generically to refer both to frames with a message data payload, i. e. data packets, and to other frames such as control frames or ilia naaem e tit frames. In order for devices, such as for example a variety of personal computers, to be connected to an Ethernet network, it is necessary to be able to convert the address format of Ethernet frames to PPP frames and vice versa. Owing to the comparatively simple nature of a PPP header, the conversion of an Ethernet frame, irrespective of the particular network protocol (for example IP or WX) presents few difficulties., the header of an Ethernet frame may be stripped off and a conventional PPP header put in its place. However, the reverse process is not straightforward because the MAC source and destination addresses required for the Ethernet header vary according to circumstances.

It has been proposed to achieve a translation of PPP frames to Ethernet frames by means of 'routiii,,' i.e. switchin-R at the logic link control layer (layer 3 OSI model). The PPP payload (i.e. IP packet) is presented to a layer J3 forwarding mechanism, Ethernet framing is applied to the IP packet, which is sent to the media access control (MAC) address, Layer 3.

Laver ') switchin- is -enerally complex and it is therefore the object of tile invention to achieve conversion of PPP frames to Ethernet frames in a simpler manner.

Summary of the Invention

The invention in a typical form is a module (typically a software module) which can be incorporated in the 'switciiing5 or 'forwarding' engine of a switch and is based on the conversion of PPP franies to Ethernet frarnes before the forwarding process, conducted at the media access control layer.

Further features of the invention will be apparent from the following description with respect to the accompanying drawings.

Brief Description of the Drawings

Figure 1 illustrates in simplified form an Ethernet frame.

1 Figure 2 illustrates schematically a PPP frame.

1 Fi-ure 3 illustrates by way of example a switch architecture including the features for implementing the present invention.

Fiaure 4 illustrates a fra-ment of a network including a switch which is oraamsed in c In accordance with the present invention.

Figure 5 illustrates part of the process of link establishment prior to the conversion of a 1 PPP frame to an Ethernet frame.

Figure 6 illustrates a process in accordance with the invention for the conversion of a PPP 1 frame under an IP protocol to an Ethernet frame.

Fic,ure 7 illustrates a process for the conversion of a PPP frame on to an TX network -1 protocol into an Ethernet frame.

Detailed Description of the Drawings

Figure 1 of the drawings shows in a simplified form an addressed 'Ethernet' data packet, i.e. a frame with a message data payload. Broadly the packet 1 comprises after a start of frame (SFD) delimiter 11, a header comprising a destination address (DA) and a source address (SA). These are 'media access control' addresses. The frame also includes protocol data 14.1 this data normally includes network address information according to tile relevant iiet,,-ork- protocol. For internet protocol (IP) packets, the protocol data typically includes an IP source address and an IP destination address. The frame may include a VI-AN identification. In this example the frame 1 includes message data comprising a variable number of bytes and cyclic redundancy check data 16, normally a 16-bit word obtained by dividing in Galois field arithmetic the data in sections 12 to 15 of the frame. As is well known, the media access control addresses are used for local switchin. whereas the network addresses in the protocol data 14 identify the ultimate end stations (source and destination) of the message.

W Figure 2 illustrates in simplified form a point-to-point protocol frame. The point-to-point protocol (PPP) is intended for the encapsulation of packets of different protocols over point to point links and in particular for the multiplexing, of different network-layer protocols simultaneously over the same link. As is shown in Figure 2, a PPP frame comprises an address 2 1, which by current convention is always the same (hex FF). It includes a control octet 22, again by convention '03'. It further includes a protocol field 23, which is either one or two octets, its value identifying either a network layer protocol or an associated network control protocol, depending on the range of values. The packet

C may include message data 24, zero or more octets, and finally a CRC field 25.

1 Figure illustrates by way of example only a switch architecture within which the present invention may be used.

Network switches commonly have a substantial multiplicity of ports, such as twenty-four or twenty-eight, for the reception and forwarding data packets, a memory for the temporary storage of received packets before they can be forwarded from the required port or ports, and a forwarding engine which ill response to media access control addresses ill t the packets has recourse to a forwarding database to obtain the particular forwarding data dent' Ifying the port or ports frorn which a given packet should be forwarded. A switch usually includes physical layer devices (PHYs) which constitute an interface between the particular transmission medium connecting the switch to other devices over the respective links and media access control devices (MACs) which are responsible for transferring data to and from the physical layer devices.

In the simplified switch shown ill Figure 3, a switch 3 'Is shown as having four ports 3 1 to 3 34 each with a port ASIC J3 1 a to 3)4a respectively, the port ASICs Including for example the physical layer device and the media access control devices for the respective ports. These are shown separately for ease of understanding in Figure 1. It should be understood however that the port ASICs along with various other components may be constituted by a single application specific inte(yrated circuit.

Switch 3 includes a forwarding engine, in this example constituted by a CPU 35, a forwarding database 36, a memory 37and a bus system 38. The memory 37 may be partly constituted by on-chip memory and mostly constituted by off-chip memory, Memory 37 may be a random access memory containing packets in locations identified by pointers which are typically organised into receive queues, representing pointers received by respective ports and transmit queues, representing packets intended for dispatch by respective ports. The forwarding database '36 links MAC destination addresses with forwarding data (such as port numbers) and is accessed by the forwarding engine (CPU 3 5) to determine for ally given packet the port or ports from which the packets should be forwarded. The forwardina database is typically complied by performing look-ups of MAC source addresses in received packets. If a source address is not already in the database, it may be made the subject of an entry together with the identification of the poll on which that packet was received. The bus system 38 is shown in simplified forin but is intended to represent the customary bus system for the conveyance of packet data as well as control and status data between various parts of the switch 3.

In practice the operation of a switch is substantially more complex that as described with reference to Flaure 3, but needs no further description herein since the organisation of a layer 2 switch is well known to those skilled in the art.

It is convenient to mention at the present stage that the switch 3 includes an ARP (address resolution protocol) cache memory 39 (shown separately from memory 37 for convenience), relating network (IP) addresses with media access control (MAC) addresses for a purpose to be described later. An address resolution protocol is normally understood to relate to a process for discovering a MAC address for a receive packet which does not include a destination MAC address. In these circumstances the packet may be 'broadcast'., if a response is received from a device which can identify the network protocol address in the packet, the ARP response enables a media access control address to be established for such a packet.

Figure 4 illustrates a fra-i-nent of a network in which a switch according to the invention C __5 may be employed. In the fragment shown in Figure 4, the switch 3 has two of its ports connected to end stations represented by personal computers 41 and 42. It has a further port connected to a subnet A which is connected to a subnet B by way of a gateway 43 shown havlnQ a link 44 to other subnets of networks. It is presumed that the network to which switch 3 is connected is an Ethernet network wherein all the various devices have at least one media access control address and wherein the transmission of data packets relies on, at each stage, the possession by the packet of an MAC source and destination address. Thus for example a packet received over link 44 by gateway 43 will contain at that time the destination MAC address of gateway 43 and a source address of a device at the other end of link 44. It will also contain the loalcal or network destination (IP) address of the desired end station (such as PC 41). When such a packet or frame arrives at switch J3 it will contain the destination MAC address of the desired end station (such as PC 41), a source MAC address of the device in subnet A at the other end of link 45 and still the lo,,lcal or destination IP address of PC 41.

It will be apparent that a frame or packet incoming to the switch 33 from subnet A and Z> intended for PC 41 or PC 42 does not present any great difficulty in the change of header.

The stopping of the Ethernet header and encapsulation with a PPP header is readily performed in the 'port ASIC' for the port connected to the end station by the PPP link- The present invention is rnore particularly concerned with the progress of frames and packets in the reverse direction, namely from sources such as PC 41 and PC 42 out oil to the network by way of the switch 3.

An adapter according to the present invention is constituted by a software module which 1 Is executed by CPU 35 'In switch 3.

When a device such as PC 41 is connected to switch 3 it is necessary for the link between them to be the subject of negotiation and authentication accordin to the point-to-point 1 9 protocol. It is further necessary to achieve the change in encapsulation from a PPP header to an Ethernet header.

As shown in Figure 5, the module, on receipt (e.g. from PC41) of a PPP configure request stage 51, negotiates the link and authentication policies, as shown in stages 52 and 53. These may be performed in known manner according to the point-to-point protocol. Broadly, the link control protocol (LCP) is used to establish the connection by an exchange of 'configure' packets. The exchange complete and an 'LCP opell' state entered once a configure-acknowledge packet has been both sent and received by the two ends of the link. On some links it may be desirable to require a device to authenticate itself before allowing network-layer protocol packets to be exchanged. If it Is not then the port may be disabled so that it is not capable of receiving further frames from the PC 4 1.

The next phase is to use the DHCP (dynamic host communication protocol) client to acquire the network parameters, Le. interpreting the IPCP requests so as to provide the information given by DHCP (stages 54 and 53).

The process shown in Figure 5 is a necessary practical feature of establishing the link but is not intended to be part of the invention and is therefore only briefly described in the foregoing Flaure 6 Illustrates a process for converting a PPP frame which employs the Internet protocol (IP) into a corresponding Ethernet frame.

The device knows the gateway by means not described here. One commonly used Z.

mechanism is via management configuration.

1 z From a start 60, the PPP]leader is stripped, stage 61. The frame will have a logical UP) destination address and it may be looked up (stage 62) in an IP look-up table such as the ARP cache 39. If the logical (IP) address is found (stage 63) in the table, then an MAC address is returned, stace 64. If the IP address is not found but is on the same subnet (staGe 67) an ARP request (stage 68) will be performed. If that yields the MAC address, stage 69, then the Ethernet header may be added (stage 65) and the packet forwarded (stage 1 Z:1 66). The MAC address can be added to the cache 39 at this point. If the ARP request does not yield the MAC address the packet has to be discarded (stage 70).

If the IP address is not on the same subnet (i.e. subnet A) stage 67, the MAC address of the gateway is returned, stage 71, Ethernet header added, stage 65 and the packet 1 - forwarded, stage 66.

Figure 7 illustrates a different conversion process where the network protocol is WX instead of IP. Broadly, the PPP negotiation may proceed as described with reference to 1 Flaure 5_ However, the header conversion will proceed differently, because according tile WX protocol there are network numbers and node numbers defined in a manner different from the definition of logical addresses according to the IP protocol.

The Device knows the IPX parameters (network number and router) by means not described here. One commonly used mechanism is via management configuration.

In order to convert from PPP (WX) to Ethernet (IPX), as shown in Fig re 7, a first stage is to strip the PPP header (stage 6 1). The IPX data (parameters are retrieved, stage 72 for comparison with the known parameters. If the network number identifies the local network, the Ethernet destination address will be the IPX node number (stage 74). If the Z- network number does riot identify a local network, then the Ethernet destination address will be the MAC address of the router. The Ethernet source address will be the WX node number. If there is no network number or adjacent router, the packet will be dropped.

-g-

Claims (3)

1. A network switch including a plurality of ports for connection to other devices and including a forwarding engine which in response to media access control address data in an Ethernet data frame forwards said frame to a respective port, said switch including adapter means for converting a data frame, received at a port and having a header portion conforming to a point-to-point protocol, into a corresponding Ethernet data frame, said adapter means providing, the received data frame with a media access control destination address before said received data frame is forwarded by said forwarding engine.

i

2. A network switch according to claim 1 wherein the data frame includes IP address data and said adapter employs the IP address to obtain said media access control destination address.

3. A network switch according to claim 1 wherein the data frame includes TX address data and said adapter determines the media access control destination address as an TX destination address or a predetermined address according as a network number in said TX address data identifies a local network or not.