WO2002059757A1 - Processeur de communication - Google Patents

Processeur de communication Download PDF

Info

Publication number
WO2002059757A1
WO2002059757A1 PCT/US2002/002293 US0202293W WO02059757A1 WO 2002059757 A1 WO2002059757 A1 WO 2002059757A1 US 0202293 W US0202293 W US 0202293W WO 02059757 A1 WO02059757 A1 WO 02059757A1
Authority
WO
WIPO (PCT)
Prior art keywords
engine
data
protocol
network stack
providing
Prior art date
Application number
PCT/US2002/002293
Other languages
English (en)
Inventor
John S. Minami
Michael W. Johnson
Original Assignee
Iready Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Iready Corporation filed Critical Iready Corporation
Priority to US10/470,365 priority Critical patent/US7379475B2/en
Publication of WO2002059757A1 publication Critical patent/WO2002059757A1/fr
Priority to US11/546,031 priority patent/US8059680B2/en
Priority to US11/546,032 priority patent/US8073002B2/en
Priority to US11/932,428 priority patent/US7646790B2/en

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/12—Protocol engines
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/16—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
    • H04L69/22—Parsing or analysis of headers

Definitions

  • the invention relates to telecommunications. More particularly, the invention relates to an apparatus and method for processing data in connection with protocols that are used in order to send and receive data, for example email, web documents, digital files, audio, video, or other data in digital format.
  • a communications network comprises a collection of network devices, also called nodes, such as computers, printers, storage devices, and other computer peripherals, communicatively connected together. Data is transferred between each of these network devices using data packets that are transmitted through the communications network using a protocol.
  • network devices also called nodes, such as computers, printers, storage devices, and other computer peripherals, communicatively connected together. Data is transferred between each of these network devices using data packets that are transmitted through the communications network using a protocol.
  • IP Internet Protocol
  • IPX Internetwork Packet Exchange
  • SPX Sequenced Packet Exchange
  • TCP Transmission Control Protocol
  • PPP Point-to-Point Protocol
  • a network device contains a combination of hardware and software that processes protocols and data packets.
  • the OSI model includes seven conceptual layers: 1 ) The Physical (PHY) layer that defines the physical components connecting the network device to the network; 2) The Data Link layer that controls the movement of data in discrete forms known as frames that contain data packets; 3) The Network layer that builds data packets following a specific protocol; 4) The Transport layer that ensures reliable delivery of data packets; 5) The Session layer that allows for two way communications between network devices; 6) The Presentation layer that controls the manner of representing the data and ensures that the data is in correct form; and 7) The Application layer that provides file sharing, message handling, printing and so on.
  • Session and Presentation layers are omitted from this model.
  • ISO seven-layer model For an explanation of how modern communications networks and the Internet relate to the ISO seven-layer model see, for example, chapter 11 of the text "Internetworking with TCP/IP” by Douglas E. Comer (volume 1 , fourth edition, ISBN 0201633469) and Chapter 1 of the text "TCP/IP Illustrated” by W. Richard Stevens (volume 1 , ISBN 0130183806).
  • An example of a network device is a computer attached to a Local Area Network (LAN), wherein the network device uses hardware in a host computer to handle the Physical and Data Link layers, and uses software running on the host computer to handle the Network, Transport, Session, Presentation and Application layers.
  • the Network, Transport, Session, and Presentation layers are implemented using protocol-processing software, also called protocol stacks.
  • protocol-processing software also called protocol stacks.
  • the Application layer is implemented using application software that process the data once the data is passed through the network-device hardware and protocol-processing software.
  • the network-device hardware extracts the data packets that are then sent to the protocol-processing software in the host computer.
  • the protocol-processing software runs on the host computer, and this host computer is not optimized for the tasks to be performed by the protocol- processing software.
  • the combination of protocol-processing software and a general-purpose host computer is not optimized for protocol processing and this leads to performance limitations. Performance limitations in protocol processing, such as the time lag created by the execution of protocol-processing software, is deleterious and may prevent, for example, audio and video transmissions from being processed in real-time or prevent the full speed and capacity of the communications network from being used.
  • Protocol processing is difficult to implement in high-speed industrial products because of the processing power required. If protocol processing can be simplified and optimized such that it may be easily manufactured on a low-cost, low-power, high-performance, integrated, and small form-factor device, these consumer and industrial products can read and write data on any communications network, such as the Internet.
  • an Internet tuner A hardware-based, as opposed to software-based, protocol processing implementation, an Internet tuner, is described in J. Minami; R. Koyama; M. Johnson; M. Shinohara; T. Poff; D. Burkes; Multiple network protocol encoder/decoder and data processor, U.S. Patent No. 6,034,963 (March 7, 2000) (the '963 patent).
  • This Internet tuner provides a core technology for processing protocols.
  • the invention comprises a communications processor of a class, such as the Internet tuner discussed above, which provides such basic desirable features as protocol processing to provide LAN support, and additional protocol processing and data processing features, such as compression for audio applications.
  • the invention provides a low-cost, low-power, high-performance, easily manufactured, integrated, small form-factor communications processor that greatly reduces or eliminates demand on the memory and the CPU of a host computer and provides highly efficient protocol and data processing.
  • the invention comprises a hardware- integrated system that both processes multiple protocols in a streaming manner and processes packet data in one pass. The invention thereby reduces or eliminates the use of host computer memory and CPU overhead.
  • the '963 patent discloses an Internet tuner for processing (decoding and encoding) protocols and packet data, comprising a network protocol layer module for receiving and transmitting packets and for encoding and decoding network packets which comprise data; a data handler module for exchanging said data with said network protocol layer module; and at least one state machine module that is optimized for a single selected protocol, said state machine module in communication with said data handler module and providing resource control and system and user interfaces; wherein said network protocol layer module, said data handler module, and said state machine module comprise corresponding hardware structures that are implemented in gate-level circuitry and wherein such hardware structures are dedicated solely to performing the respective functions of their corresponding modules.
  • the preferred embodiment of the invention comprises an auxiliary microprocessor or equivalent that acts as a protocol engine and provides any of LAN support, external interfaces to peripherals and memory, and additional protocol and data processing, such as compression for audio applications, for example, to the Internet tuner of the '963 patent.
  • the presently preferred communications processor incorporates a protocol engine, a set of peripherals for the protocol engine, an Internet tuner core or other network stack, an external controller interface, and a memory interface.
  • the communications processor thus provides network, e.g. Internet, connectivity to a wide range of consumer network devices and industrial network devices.
  • Fig. 1 is a block schematic diagram that depicts a typical network device using a communications processor with external components for LAN and/or dialup communications according to the invention
  • Fig. 2 is a block schematic diagram of a communications processor according to the invention.
  • Fig. 3 is a block schematic diagram of a network stack, which is part of the communications processor, according to the invention.
  • Fig. 4 is a block schematic diagram of a MAC interface, which is part of the network stack, according to the invention.
  • Fig. 5 is a block schematic diagram of an exemplary network according to the invention.
  • Fig. 6 is a diagram that shows a network stack internal memory map according to the invention.
  • Figure 7 is a block schematic diagram of the peripherals attached to a protocol engine according to the invention.
  • FIG. 8 is a block schematic diagram showing an IP-only mode data path according to the invention.
  • Socket engine handles management of data received from or to be sent to the TCP/UDP transport engine
  • TCP/UDP transport engine (handles processing of UDP and TCP packets)
  • IP router - top engine (handles the router lookups and calculates the next hop)
  • IP router - bottom engine switches data down to the correct physical layer and arbitrates data coming up from physical layers
  • IP raw mux switches data to ARP module or raw IP interface
  • MAC mux switch that connects internal MAC core or external MAC core to the MAC buffers
  • Signal mux switches external data-link interface between Mil interface, IP-only mode interface, or external MAC interface
  • MMU Memory-management unit
  • the discussion herein defines the architecture of the presently preferred embodiment of the communications processor, describes the various diagrams that accompany this document, and discusses various features of the invention.
  • the herein disclosed communications processor provides industrial and consumer products with the protocol processing needed to connect to the Internet, send and receive, for example data, email, web documents, digital files, audio, video, or other data in digital format.
  • the presently preferred communications processor is based in part on the Internet tuner described in J. Minami; R. Koyama; M. Johnson; M. Shinohara; T. Poff; D. Burkes; Multiple network protocol encoder/decoder and data processor, U.S. Patent No. 6,034,963 (March 7, 2000) (the '963 patent).
  • the preferred embodiment of the invention adds any of LAN support, external interfaces to peripherals and memory, and additional protocol and data processing features, such as compression for audio applications.
  • Internet appliances such as network computers or wireless Internet-enabled TVs
  • the invention provides a low-cost, low-power, easily manufactured, integrated, small form-factor network access module that has a low memory demand and provides a highly efficient protocol decode.
  • the invention comprises a hardware-integrated system that both decodes and encodes multiple network protocols in a streaming manner concurrently and processes packet data in one pass, thereby reducing system memory, power consumption, and form-factor requirements, while also eliminating software CPU overhead.
  • the '963 patent discloses an Internet tuner for decoding and encoding network protocols and data, comprising a network protocol layer module for receiving and transmitting network packets and for encoding and decoding network packet bytes which comprise packet data; a data handler module for exchanging said packet data with said network protocol layer module; and at least one state machine module that is optimized for a single selected network protocol, said state machine module in communication with said data handler module and providing resource control and system and user interfaces; wherein said network protocol layer module, said data handler module, and said state machine module comprise corresponding hardware structures that are implemented in gate level circuitry and wherein such hardware structures are dedicated solely to performing the respective functions of their corresponding modules.
  • the preferred embodiment of the invention comprises an auxiliary processor or protocol engine that adds any of LAN support, external interfaces to peripherals and memory, and additional protocol and data processing, such as compression for audio applications, for example, to the Internet tuner of the '963 patent.
  • the presently preferred communications processor incorporates a protocol engine, a set of peripherals for the protocol engine, an Internet tuner core, an external controller interface, a memory interface, and two auxiliary serial ports.
  • the communications processor thus provides network, e.g. Internet, connectivity to a wide range of devices.
  • Fig. 1 is a block schematic diagram that depicts a typical network device using a communications processor 10 with external components for LAN or dialup communications according to the invention.
  • Such system comprises the integrated communications processor 10, network-device hardware such as a PHY 15 (for LAN connection) or a modem 17 (for dialup communication), host system logic 18, one or more slave serial devices 19, an optional external CPU or microprocessor 16, and optional external memory such as RAM 11 or ROM 13.
  • the communications processor 10 is capable of performing all protocol processing. In prior art the protocol processing is performed in the host system logic 18.
  • Fig. 2 is a block schematic diagram of a communications processor 10 according to the invention.
  • Such communications processor 10 comprises a network stack 50 connected to a protocol engine or auxiliary microprocessor 34.
  • the protocol engine 34 is connected to a bus controller 32.
  • the bus controller 32 allows the protocol engine 34 to use an internal 32KB memory 30 or the optional external memory.
  • the unit KB when applied to the size of memory refers to kilobytes of memory, where a kilobyte is 1024 bytes. In this description a byte of memory is eight bits of memory.
  • the network stack 50 is connected directly to a 26KB internal RAM 48, a media access controller (MAC) interface (l/F) 52, and a PPP module 54.
  • MAC media access controller
  • the MAC has access to 6KB of internal RAM 46 and connects the network stack 50 to an external PHY 15 or an external MAC 8.
  • the PPP module 54 connects the network stack 50 via the modem interface (l/F) module 58 to a modem 17.
  • the external controller interface 68 allows the protocol engine 34 to be communicate with an external CPU, microprocessor, or microcontroller 16.
  • the network stack 50, the protocol engine 34, and controller interface 68 communicate via a communications processor bus 70.
  • Also connected to the communications processor bus 70 are system peripherals 36, an auxiliary serial port 56, a dual-function proprietary serial port or second auxiliary serial port 60, a synchronous serial interface 66, general I/O 44, and a series of specialized data and protocol processing modules, or offload engines.
  • the specialized data and protocol processing modules or offload engines include a text rasterizer engine 64, a MIME string search engine 62, an ADPCM accelerator engine 38, a base64 encoder and decoder 40, and a G3 encoder engine 42.
  • Such specialized data and protocol processing modules or offload engines may perform a variety of functions.
  • the specialized data and protocol processing modules 38, 40, 42, 62, and 64 allow the implementation of the following features, for example: 40. ITU T.37 compatibility;
  • Fig. 3 is a block schematic diagram of a network stack according to the invention.
  • Such network stack 50 comprises a one or more hardware modules that perform protocol processing in an analogous fashion to a protocol stack that is normally implemented in software.
  • This network stack comprises hardware modules that implement the Data Link layer: a PPP module 54 and a MAC interface 52; and the Network, Transport, Session, and Presentation layers for the TCP/IP protocol suite: an ARP module 72, an IP raw multiplexor (mux) module 104, an IP router - bottom module 92, an IP filter module 90, and IP router - top module 88, and IP module 86, a TCP/UDP module 84, and a sockets module 82.
  • the sockets module 82 connects to a protocol engine interface module 80.
  • the protocol engine interface module 80 connects to the protocol engine 34 via the communications processor bus 70.
  • the present communications processor 10 uses a master clock that may be set in frequency from 8 MHz to 70 MHz.
  • the clock frequency chosen depends upon the application and the protocol engine 34 that is used.
  • the presently preferred communications processor 10 uses a Zilog Z80 core, a popular 8-bit microprocessor, as the protocol engine 34, but any microprocessor, such as ARM, ARC, PowerPC, or MIPS, could be used.
  • a Z80 microprocessor is most suited to low-cost consumer applications that do not require high-speed communications.
  • a more powerful microprocessor or combination of microprocessors could be used as the protocol engine 34 for industrial and high-performance applications.
  • the interface pins can be configured as either a parallel or serial (SPI) interface (described in more detail later), or if no external CPU, microprocessor, or microcontroller 16 is attached, these interface pins may be used as general-purpose I/O pins.
  • a register set is provided as a communication channel between the external CPU, microprocessor, or microcontroller 16 and the communications processor 10.
  • the communications processor 10 operates in one of two modes: normal mode, and CPU-bypass mode. Mode selection is performed using configuration pins.
  • the protocol engine 34 When configured for CPU-bypass mode, the protocol engine 34 is disabled, and the external CPU, microprocessor, or microcontroller 16 can communicate directly with the network stack 50 using an application programming interface (API) register set.
  • API application programming interface
  • the protocol engine 34 When configured for normal mode the protocol engine 34 is enabled, and the external CPU, microprocessor, or microcontroller 16 communicates via a set of registers described that are described in the next section.
  • the external CPU, microprocessor, or microcontroller 16 can access the network stack memory 48 using two methods.
  • the first method involves cooperation with the protocol engine 34.
  • the protocol engine 34 must set up one of two sets of address registers depending on whether the external CPU, microprocessor, or microcontroller 16 is reading or writing to the network stack memory 48.
  • the second method the external CPU, microprocessor, or microcontroller 16 sets up the read address or write address to access the network stack memory 48. If the external CPU, microprocessor, or microcontroller 16 wants to write to the network stack memory 48, then it writes the starting memory address to a set of registers. The external CPU, microprocessor, or microcontroller 16 can then start to write data.
  • the address registers increment with each write.
  • the external CPU, microprocessor, or microcontroller 16 wants to read the network stack memory 48, it initializes a set of registers to the starting address of the network stack memory 48 that is to be read. The address registers increments with each read. There is also a subset of the CPU-bypass mode called the test-index mode used primarily for test and diagnostic purposes. The test-index mode effectively allows the external CPU, microprocessor, or microcontroller 16 to control the network stack 50 while keeping the protocol engine 34 enabled. The protocol engine 34 may also be kept in reset (in an inactive state) so that the protocol engine 34 and the external CPU, microprocessor, or microcontroller 16 do not simultaneously access the network stack 50.
  • FIGs 3 shows a MAC interface 52 and an external MAC interface 24 together with an IEEE standard Mil (media-independent interface) 26 that connects to an external PHY 15.
  • the same interface pins can be configured as the MM interface to an external PHY, or as an SPI interface for either an external MAC or raw IP data. Additionally, if modem operations are mutually exclusive with LAN operations, then the same interface pins may also be used to support the external modem interface 6.
  • Figure 4 shows the MAC interface 52 and the external MAC interface 24 in more detail.
  • the communications processor 10 supports both the internal MAC 126, as well as an external MAC 8.
  • the internal MAC 126 is compliant to the IEEE 802.3 standard and uses the IEEE standard MM (media-independent interface) 26 to communicate with the external PHY 15.
  • the current revision of the communications processor supports both 10 Mbps and 100 Mbps Ethernet speeds, but the architecture may be scaled to operate at both lower and higher speeds.
  • the PHY 15 is external, but the PHY 15 could be integrated into the communications processor 10.
  • SPI serial peripheral interface
  • the connections from the internal MAC 126 and the connections to the external PHY 15 may share common package pins of the communications processor 10.
  • a signal multiplexor (mux) 130 shown in Figure 4, controls the function of these package pins under user configuration.
  • Two RAM buffers 132 and 134, and a MAC buffer control 120 are included in the data path when either MAC data path is used because of the different clock domains that are used for transmitting and receiving data.
  • the RAM buffers may be substantially larger without affecting the architecture for higher-performance applications.
  • Data going into and out of the external MAC interface 52 is synchronous to the transmit clock (for transmit) and synchronous to the receive clock (for receive).
  • Outgoing data packets that are being transmitted are buffered in the 2KB RAM transmit buffer 132 prior to being transmitted, thereby optimizing data flow and avoiding any under-run conditions, for example.
  • Incoming data that is being received is stored in the 4KB RAM receive buffer 134 until a complete data packet is received and the data packet has been verified as valid.
  • the protocol engine 34 uses an internal (integrated or on-chip) 32KB RAM 30 and optional external RAM 11 and external ROM 13.
  • the external RAM 11 and external ROM 13 is not needed when all of the code and data of the protocol engine 34 is less than the size of the internal RAM 30 or when the communications processor 10 is operated in CPU-bypass mode.
  • the internal 32KB RAM 30 may be made substantially larger for high-performance applications without affecting the architecture.
  • the protocol engine internal RAM 30 is capable of being battery- operated via I/O pins to allow nonvolatile storage of code when no external ROM 13 is used.
  • the present version of the communications processor 10 provides programmable wait states for the optional external ROM 13 so that a variety of ROM speeds may be used. However, slower ROMs may have an impact on overall performance. The optimum ROM speed is application dependent, but in general 70ns ROMs provide adequate performance for consumer applications and are currently readily available and inexpensive.
  • the present version of the communications processor 10 uses 8-bit ROM for the external ROM 13 and uses 8-bit RAM for the external RAM 1 1. Programmable wait states are provided for the external RAM 1 1. The optimum speed of the external RAM 11 is dependent on the application, but 70ns parts offer adequate performance for most consumer product applications.
  • the present version of the communications processor 10 is designed to use 8-bit SRAM for the external RAM 11 , but other RAM sizes, organizations, and types such as SDRAM or DDR SDRAM may also be used that require changes to the bus controller 32, but without significant changes to the rest of the architecture.
  • the present version of the communications processor 10 contains a powerful IP filter engine 90 to support such features as, for example, network address translation (NAT) and IP masquerading.
  • the first function of the IP filter 90 is to parse the information in the incoming data packet (for example the type of packet, the source and destination IP addresses, the source and destination port numbers, and so on).
  • the information from the data packet is made available to the protocol engine 34 so that the protocol engine 34 can decide what to do with the data packet.
  • the protocol engine 34 may intercept the data packet, pass the data packet up the network stack 50, discard the data packet, or re-transmit (forward) the data packet. Prior to receiving or forwarding the packet, the protocol engine 34 may modify any packet parameter, including, but not limited to, the source IP address, destination IP address, source port, destination port, and the time-to-live (TTL). The IP filter engine 90 will then recalculate the appropriate checksums and send the packet as directed by the protocol engine 34. The protocol engine 34 controls these and other functions of the network stack 50 via the protocol engine interface 28. The second function of the IP filter 90 is to inject data packets back into the network stack 50. Injected data packets can come from either the IP filter engine 90 or the protocol engine 34.
  • TTL time-to-live
  • the IP filter engine 90 may be enabled under control of the protocol engine 34 using a range of settings.
  • the IP filter 90 may be enabled to filter on the basis of specific ports, IP addresses, or on the basis of specific protocols. These IP filter criteria are set using registers. The following sections describe the theory of the operation of the IP filter 90 for supporting NAT and IP masquerading.
  • Fig. 5 is a block schematic diagram of an exemplary network according to the invention.
  • the base unit (or base node or base network device) 144 contains a communications processor for both Ethernet LAN and telephone-line links, thus supporting two data links, i.e. Ethernet link 154 and a telephone-line or dialup link 152.
  • the base unit 144 has an IP address associated with each data link.
  • the IP address associated with the Ethernet link, 10.10.150.1 is a local IP address and is not recognized external to the Ethernet LAN.
  • the IP address associated with the telephone-line link, 204.192.4.5 is recognized external to the Ethernet LAN, and may, for example, be a floating IP address that was assigned to the base unit 144 by an Internet service provide (ISP) during PPP negotiations, or it may be a permanent IP address.
  • ISP Internet service provide
  • the client units (146, 148, and 150) may also contain a communications processor for both Ethernet link and telephone- line link.
  • the communications processor may contain multiple Ethernet links or telephone-line links and thus multiple client units (146, 148, and 150) may be part of a single network device.
  • the combination of the IP filter functions and support for multiple data links in a single integrated communications processor 10 enables such sophisticated features as, for example, failover or load- balancing to be performed.
  • client unit #1 146 wants to access www.iready.com 140 on the Internet 142.
  • client unit #1 146 sends a packet with its IP address (10.10.150.51) as the source IP address, and the IP address of www.iready.com (123.45.6.78) as the destination IP address.
  • Client unit #1 146 detects that the destination IP address is not on the local network and sends the packet to the base unit 144, assuming that the base unit is set up as the default gateway. Once the base unit 144 receives the packet, it is passed from the Ethernet MAC interface 52 to the ARP engine 72.
  • the ARP engine 72 examines the packet to see if it is an ARP packet, an IP packet, or an unknown packet. In this case, because the received packet is an IP packet, it is passed to the IP router engine - bottom 92. The IP router engine - bottom 92 arbitrates between incoming packets coming from the Ethernet path and PPP module 54. If no PPP traffic is currently being received, then the received IP packet is sent to the IP filter engine 90. In the IP filter engine 90, the packet is parsed and stored in the IP filter buffer (part of the network stack internal memory 116). The protocol engine 34 is then notified via an interrupt, that a packet has arrived.
  • the protocol engine 34 When the protocol engine 34 receives the interrupt notification, it uses registers to read the port and IP address information of the received packet to determine what to do with the received IP packet. In this example, the protocol engine 34 operates on the received IP packet and replaces the source IP address (10.10.150.51 ) of client unit #1 146 with the base unit 144 global IP address (204.192.4.5), and replaces the client unit #1 146 source port number with a port number that the base unit 144 associates with this socket. When these values are written to the appropriate IP filter registers, a SND command is issued.
  • the IP filter engine 90 Upon receiving the SND command, the IP filter engine 90 replaces the port and IP address information, recalculates the IP and TCP header checksums, and transmits the packet over the telephone-line link 152 (in this example network).
  • the protocol engine 34 also logs a socket connection between a port on client unit #1 146 and a port on www.iready.com 140.
  • the incoming return packet is handled by the PPP module 54.
  • the PPP module 54 stores the incoming return packet into the PPP memory buffer (part of the network stack internal memory 116).- When the entire incoming return packet is received, and the frame checksum (FCS) is validated, the PPP buffer notifies the IP router engine - bottom 92. If no Ethernet packet is currently being received via the Ethernet data link, then the incoming return packet is sent from the PPP module 54 to the IP filter engine 90 via the IP Router Bottom 92.
  • FCS frame checksum
  • the IP filter engine 90 parses the received packet and stores it in the IP filter buffer (part of the network stack internal memory 116) and notifies the protocol engine 34, via an interrupt, that a packet has arrived.
  • the protocol engine 34 examines the packet's port and IP address information, and recognizes that the destination port number contained in the packet is the port number that is associated with a socket connection for client #1.
  • the protocol engine 34 replaces the destination port with the original source port from client unit #1 146, and replaces the destination IP address with the IP address of client unit #1.
  • the protocol engine 34 issues a SND command. This SND command causes the IP filter engine 90 to replace the port number and IP address, recalculate the checksums, and then send the packet out via the Ethernet link 154. Determination of which physical link the packet should use is handled by the IP router - bottom engine 92.
  • the sequence of events just described corresponds to the functions required by network address translation (NAT) and IP masquerading.
  • the process of receiving and transmitting packets then continues in this manner until the connection is closed.
  • the protocol engine 34 can determine the closing of a connection by snooping (viewing) the TCP header flags of the transmitted and received packets.
  • the protocol engine 34 recognizes that a connection has been closed, the protocol engine 34 removes that connection log from its active connection table stored in internal memory 30 or external memory 12.
  • the number of simultaneous connections that can be maintained is only limited by the amount of memory available to the protocol engine 34 in the base unit 144. Consumer network devices that only require a few (1-10) connections may use a small memory, such as the internal memory 30, and industrial devices that may require thousands of connections can employ external memory 12.
  • the communications processor 10 uses a timeout mechanism because, unlike TCP, UDP does not have any notion of opening or closing a connection.
  • a timer resets every time a UDP packet for a connection is received.
  • the timer may be set under external control, with a presently preferred default timeout value of 15 minutes.
  • the REC command causes the received data packet to be passed up the network stack via the IP router - top engine 88, the IP engine 86, the TCP/IP engine 84, and the sockets module 82.
  • the data from the received data packet is passed through a network stack socket interface.
  • the base unit 144 needs to send data back to the client unit it uses the network stack socket interface.
  • the packets are then generated by the TCP/UDP engine 84.
  • the TCP/UDP engine will query the IP router -top engine 88 for the next-hop IP address and the appropriate source IP address. Both of these parameters are determined by the IP router - top engine 88 based on the destination IP address for the packet.
  • the packet is then passed through the IP engine 86.
  • the IP engine prepends the IP header, and sends a transmission request to the IP router - top engine 88.
  • the packet is then sent through the IP filter 90 without modification to the IP router - bottom engine 92.
  • the IP router - bottom engine 92 will route the packet to the appropriate physical data link. In this case, it will send the packet to the ARP engine 72.
  • the ARP engine 72 will then use the next-hop IP address to look up the corresponding MAC address. With this information, the ARP engine 72 generates the Ethernet header and prepends it to the packet.
  • the complete packet is then sent to the MAC interface 52.
  • the packet is first queued in the MAC transmit buffer 134 by the MAC buffer controller 120.
  • the packet is then copied from the MAC transmit buffer 134 to the internal MAC 126 and finally out via the MM interface 26 to the external PHY 15 and on to the client units 146, 148, and 150.
  • the PPP engine 54 When the PPP engine 54 receives reply data packets, it sends the data packets up through the IP router - bottom engine 92, and then to the IP filter engine 90.
  • the IP filter engine 90 parses and stores the data packet in IP filter memory (part of the network stack internal memory 116) and notifies the application that an IP packet has been received.
  • the application then examines the port and IP address information of the data packet, and determines if the data packet is destined for the base unit 144.
  • the application then issues the REC command to the IP filter engine 90, which causes the IP filter to retrieve the packet from the IP filter memory (part of the network stack internal memory 116) and send it to the IP engine 86.
  • the application then processes the data through the network socket interface.
  • ping requests from a client unit to the Internet are handled for the example network shown in Figure 5.
  • This situation is a special case of the connection between client units (146, 148, and 150) and the Internet 142.
  • the ping requests are entered into a special ping table in the base unit communication processor (in part of the protocol engine internal memory 30). These ping table entries have a timeout, e.g. approximately 10 seconds. If no ping response is received before a time equal to the timeout, the ping table entry is deleted. If a ping response is received before a time equal to the timeout, the sequence number of the ping reply is checked against entries in the ping table for the ping request.
  • the IP address for the client unit that sent the ping request is copied to the ping reply packet.
  • the ping reply packet is then sent to the client unit on the LAN using the Ethernet link 154.
  • the ping table entry is also deleted when the ping reply is received.
  • a raw IP packet is any arbitrary network packet.
  • the raw IP packet is copied to the raw IP buffer (part of the network stack internal memory 116) by using commands and special-purpose registers.
  • the IP router - bottom engine 92 routes the raw IP packet to the proper data link.
  • a register is then cleared to indicate that the raw IP packet has been completely transmitted.
  • the base unit 144 receives IP multicast and broadcast IP packets from the Internet 142 for the example network shown in Figure 5.
  • the packet is stored in the IP filter buffer (part of the network stack internal memory 116), and the protocol engine 34 is notified that a packet has arrived.
  • the protocol engine 34 examines the packet, and should it decide to forward the packet, it may modify any appropriate packet parameter, and then issue a SND command.
  • This SND command causes the IP filter engine 90 to recalculate the packet checksums, and send the packet out via the Ethernet link 154.
  • the base unit 144 transmits IP multicast and broadcast IP packets for the example network shown in Figure 5.
  • the base unit 144 can use the network stack socket interface when it sends multicast or broadcast packets.
  • the destination IP address is set appropriately, and the network stack 50 handles the multicast or broadcast packet as a normal IP packet.
  • the protocol engine 34 can send a multicast or broadcast packet out as a raw IP packet.
  • the base unit 144 uses a port network address translation (NAT) table. This table matches port numbers with the client units on the LAN. For example, if client unit #1 146 with IP address 10.10.150.151 is designated as an HTTP server (using port 80), and client unit #2 148 with IP address 10.10.150.152 is designated as a POP server (using port 110), a port NAT table would appear as shown in Table 1. Table 1. Sample Port NAT Table
  • IP filter engine 90 As packets arrive at the IP filter engine 90, they are parsed and stored in the IP filter buffer (part of the network stack internal memory 116) and the protocol engine 34 is notified via an interrupt using the protocol engine interface 28.
  • the protocol engine 34 examines the header parameter registers, via the protocol engine interface 28, to determine if the destination port in the incoming data packet matches any ports in the port NAT table. If the destination port does match a port NAT table entry, then the destination IP address in the incoming data packet is changed to the IP address specified in the table corresponding to that port, and a SND command is issued.
  • the IP filter engine 90 then changes the header parameter registers, recalculates the checksums, and transmits the modified data packet via the Ethernet link.
  • the protocol engine 34 When client unit #1 sends a response data packet back to the base unit 144, the protocol engine 34 again attempts to match the port specified in the response data packet with the port NAT table. If there is a match between ports, the protocol engine 34, via the protocol engine interface 28, changes the source IP address in the packet from the IP address of client unit #1 146 to the IP address of the base unit 144 prior to transmitting the packet to the Internet 142 on the telephone-line link 152 (in this example network).
  • the IP filter engine 90 uses 6KB of the network stack internal memory 116.
  • the 6KB is split between the IP filter receive buffer, the IP filter send buffer, and the raw IP buffer.
  • the partitioning and size of the buffers may be adjusted in different embodiments.
  • both the IP filter receive/send buffer and the raw IP buffer may be 3KB in length, or one may be 2KB and the other 4KB, or each may be considerably larger for long latency or high bandwidth networks.
  • Incoming IP packets are first stored in the IP filter receive/send memory buffer (part of the network stack internal memory 116). The application is notified when the packet is received.
  • the application If the application wishes to move the packet to the raw IP buffer (part of the network stack internal memory 116), it writes the target memory location in the raw IP memory buffer (part of the network stack internal memory 116) to the DMA address registers. When the write to the DMA address registers is complete, a DMA command is issued to start the DMA transfer. When the DMA transfer is complete, a bit in a status register is set. If interrupts are enabled this status register bit condition triggers an interrupt to the application.
  • the network stack 50 handles ICMP echo request packets (or ping packets).
  • the network stack 50 includes specialized and optimized hardware support for ICMP echo reply packet generation. That is, if the IP engine 86 receives an ICMP echo request packet, the IP engine 86 can automatically generate the appropriate ICMP echo reply packet.
  • the IP engine 86 uses part of the network stack internal memory 116 as a temporary store for the data section of the echo request and echo reply packets.
  • the two cases correspond to the IP filter engine 90 being enabled or disabled.
  • ICMP echo request packets pass directly through the IP filter engine 90 and are processed by the IP engine 86.
  • ICMP echo reply packets are automatically generated by the IP engine 86 using network stack internal memory 116 as a temporary store. The echo reply packet is then transmitted.
  • the IP filter engine 90 When the IP filter engine 90 is enabled it uses the network stack internal memory 116. This prevents the IP engine 86 from using network stack internal memory 116 as a temporary store to generate the ICMP echo reply packets.
  • the ICMP echo reply is generated under control of the protocol engine 34 via the protocol engine interface 28.
  • the protocol engine 34 via the protocol engine interface 28, changes the ICMP type in the ICMP echo request packet from 0x08 (hex 08) to 0x00, and then swaps the source and destination addresses in the original echo request packet in order to form the echo reply packet.
  • the protocol engine 34 then issues a SEND command to the IP filter 90 via the protocol engine interface 28 in order to transmit the echo reply packet.
  • IP router - top engine 88 and IP router - bottom engine 92. These two IP router engines serve as an extension to the IP engine 86.
  • the IP router - bottom engine 92 serves as a switch between the Ethernet and PPP transmit and receive data link paths.
  • the IP router - bottom engine 92 checks that two packets are not being received at the same time from the PPP engine 54 (the PPP data link receive path) and the IP raw mux 104 (on the Ethernet data link receive path). All PPP packets are first buffered in part of the network stack memory 116. This is done because the PPP link is often much slower then the Ethernet LAN link. By first buffering the packet, the network stack is able to process PPP packets at the same rate as packets from the Ethernet LAN link. Without packet buffering, packets from the Ethernet LAN link may be held up for long periods while the network stack 50 is processing a slowly arriving PPP packet.
  • the IP router - bottom engine 92 routes the transmitted packets between the PPP engine 54 (on the PPP data link transmit path) and the IP raw mux 104 (on the Ethernet data link transmit path) based upon the next hop IP address.
  • the IP router - top engine 88 checks the destination IP address. If the destination IP address corresponds to the local network, the IP router - top engine 88 transmits the packet directly to network device at the specified IP address using either the Ethernet data link or the PPP data link. However, if the destination IP address does not belong to any directly connected networks, the IP router - top engine 88 searches to find the best gateway (and the appropriate data link) to which to send the packet. The mechanism for this search is described next.
  • the IP router - top engine 88 uses an n-entry table for routing information, which is described more completely in the following sections. All entries are programmable from the protocol engine 34. Of the n entries, most of the entries are general- purpose routing entries and one of them is a default routing entry. The IP router - top engine 88 is designed to support one or more of both PPP data links and Ethernet data links.
  • the IP router - top engine 88 sits below the TCP/UDP engine 84 and IP engine 86 and above the PPP engine 54 (on the PPP data link path) and IP raw mux 104 and ARP engine 72 (on the Ethernet data link path) in the network stack 50 (See Figure 3).
  • the IP router - top engine 88 interfaces directly to the IP engine 86 and the protocol engine 34 via the protocol engine interface 28.
  • the IP router - top engine 88 monitors all outgoing packets and provides the next-hop IP addresses as well as the appropriate source IP address for the packets. The following sections describe the operation of the IP router - top engine 88.
  • the IP router - top engine 88 and IP router - bottom engine 90 cooperate to direct the data packet to the appropriate data link.
  • the IP route table is essential for maintaining IP routing information.
  • the table is implemented in the IP router - top engine 88 and contains n entries: there are several general-purpose routing entries and one default routing entry.
  • the general- purpose entries may be programmed to be either static entries or dynamic entries and the default entry is always a static entry (see Table 2).
  • the routing decision, made by the IP router - top engine 88, is based on the information contained in the routing table.
  • the IP router - top engine 88 searches the route table by performing three steps:
  • IP router - top engine 88 determines which data link should be used to transmit a data packet. It passes the next-hop IP address as well as the appropriate source IP address to use for the packet back to the calling engine. The routing is now complete.
  • the IP route table must be configured before any IP packets are sent.
  • the protocol engine 34, or external CPU, microprocessor or microcontroller 16 writes to a set of application programming interface registers. The following sequence of steps is required to configure the route table before any packets may be sent:
  • the protocol engine 34 After the registers and the route table are configured, the protocol engine 34 maintains the route table by programming the appropriate routes using the protocol engine interface 80. Typically, routes in the route table can change for any of the following reasons:
  • DHCP provides updated information about a default gateway
  • the IP router - top engine 88 monitors whether data links are up (working or active) or down (broken or inactive) and chooses the data links appropriately.
  • the route table includes the following information: destination IP address and gateway IP address. The route table information may be retrieved from the route table by the protocol engine 34 by executing a read command.
  • the ARP engine 72 resolves an Ethernet hardware address from any given IP address.
  • IP packets When IP packets are sent, the destination IP address is not sufficient in an Ethernet network.
  • the 48-bit hardware address In order to send a packet, the 48-bit hardware address must also be found. In an Ethernet network, a 48-bit hardware address is used to uniquely identify network devices. In order to map or resolve an IP address to the 48-bit hardware address the following sequence of steps occurs.
  • the ARP engine 72 sends a broadcast ARP request containing the IP address to be resolved to the network.
  • the destination network device having recognized its IP address in the ARP request, then sends back an ARP reply packet, which includes the 48-bit hardware address of the destination network device, to the ARP engine 72.
  • the ARP engine 72 saves the resolved 48-bit hardware address together with the original destination IP address as an associated pair in the ARP cache. Now, when the application sends another packet to the same destination IP address, the ARP engine 72 knows, by using the ARP cache, where it may find the correct 48-bit hardware address, and where to send the packet without performing another ARP request.
  • the preferred ARP cache contains four entries.
  • a "Least Recently Used" scheme is applied to update and retire the cache entries.
  • the ARP engine 72 listens to any ARP request that it receives and generates all ARP replies that match its IP address.
  • the ARP cache may contain substantially more entries for higher performance applications without affecting the architecture.
  • the ARP engine 72 sits below the IP raw mux 104 and IP router - bottom engine 92 in the network stack 50 and interfaces directly to the Ethernet MAC interface 52.
  • the ARP engine 72 has access to the internal network stack memory 116 through the memory arbitrator 100.
  • the ARP engine 72 operates very closely with the IP router - bottom engine 92. In most applications the ARP engine 72 and the IP router - bottom engine 92 are configured together, especially when there are multiple data links that need to be supported, e.g. PPP and Ethernet.
  • ARP provides a dynamic mapping from a 32-bit IP address to the corresponding 48-bit hardware or Ethernet address, e.g. the Ethernet address 11 :12:13:AA:B0:C1 (which corresponds to 48 bits).
  • the TCP engine 84 forms an IP packet with a specified destination IP address destined for the IP engine 86 and the IP router - top engine 88.
  • the ARP engine 72 provides a mapping between the IP address and the 48-bit hardware address so that the IP packet can be correctly sent to its destination.
  • the reverse of ARP is not supported by the present version of the ARP engine 72, but RARP could be implemented using the same structures and design as the ARP engine 72 with minor modifications.
  • the ARP cache is essential to maintain the ARP operation.
  • the present cache table implemented in the ARP engine 72 consists of four entries, but the number of table entries may be increased in alternative embodiments. Each cache entry consists of the destination IP address, destination hardware address, and the ARP down counter (the down counter serves as an expiration timeout counter).
  • the ARP engine 72 is configured by writing to the corresponding application programming interface (API) registers. Configuration is achieved by completing the following two steps:
  • an ARP cache entry is read by writing the ARP cache entry index to the ARP cache select register.
  • the following ARP cache entry information may then be read from registers: the resolved destination IP address, the resolved 48-bit hardware address, and the ARP cache down counter.
  • An unsupported packet is any frame that is received from the MAC interface 52 that has an Ethernet frame type other than x0806 (corresponding to an ARP packet) or x0800 (corresponding to an IP packet).
  • the unsupported packet is stored and retrieved by the protocol engine 34 (if this store and retrieval feature is enabled), by setting a bit in the ARP configuration register.
  • the maximum size of an unsupported packet that may be stored by the ARP engine 72 is 2KB in the dedicated ARP buffer memory (which is attached to the ARP engine 72, but not shown explicitly in Figure 3). Any unsupported packet that is longer than 2KB triggers an overflow condition and the excess bytes are dropped.
  • the protocol engine 34 If the protocol engine 34 intends to read any unsupported packet received, it must read the packet from the ARP buffer memory as soon as the unsupported packet interrupt is detected to avoid any overflow condition.
  • the number of bytes that are available in the ARP buffer memory can be read from a register. The protocol engine 34 should only read those bytes that have been stored in the buffer memory to avoid any under-run conditions.
  • the MAC interface 52 The MAC implementation integrated into the network stack 50 enables Ethernet access for network devices that use the communications processor 10.
  • the MAC interface 52 may be configured to operate in two modes: normal and test. During normal mode, the MAC interface 52 transmits data packets created by the network stack 50. The MAC interface 52 also receives data packets, filters the appropriate addresses, and passes the data packets to the network stack 50 for further processing.
  • the MAC interface 52 may also be configured in a test mode where the protocol engine 34 has direct control over the MAC interface 52, bypassing the network stack 50. In this test mode, the protocol engine 34 may send and receive Ethernet frames directly through the MAC send and receive buffers 118. In test mode, the protocol engine 34 is responsible for generating packets including the destination address, source address, Ethernet frame type and length fields, and the packet data payload. When a valid Ethernet data packet is received, the MAC interface 52 passes the entire packet to the protocol engine 34.
  • the preferred MAC interface 52 currently supports 10/100Mbps Ethernet and requires a system clock running at a minimum frequency of 8 MHz for 10 Mbps operation. Using the minimum system clock frequency allows the network stack 50 to sustain a throughput equal to the full 10Mbps Ethernet bandwidth. When the PPP data link path is present, a higher minimum system clock frequency may be required. The minimum system clock frequency is then dependent upon the speed of the PPP data link. Alternative embodiments may include higher speed Ethernet and PPP data links.
  • the MAC interface 52 supports both full-duplex and half-duplex modes of operation.
  • the default mode is full-duplex and the MAC interface 52 can process and generate pause frames to perform flow control with its data-link partner.
  • the flow-control mechanism is designed to avoid a receive FIFO over-run condition.
  • the MAC buffer management receives Ethernet packets from the internal MAC 126 or the external MAC 8, it monitors the memory usage in the second-level memory receive FIFO. When there are 64 or less bytes left in the FIFO, the buffer management logic asserts the start_pause signal to the pause-frame generator module.
  • the pause- frame generator module begins to send a pause frame with a maximum pause size to either the internal MAC 126 or the external MAC 8.
  • the buffer management continues to keep track of the memory usage until there are 128 or more bytes available in the FIFO. It then sends an end_pause signal to the pause-frame generator module.
  • the pause-frame generator logic sends a pause frame of zero pause size to end the flow control mechanism.
  • the transmit engine in the internal MAC 126 halts any further transmission (if there is one) after the completion of the current frame.
  • the internal MAC 126 issues jam sequences if the first-level receive FIFO has one byte open (indicating a close-to-full condition) during receive.
  • the internal MAC 126 provides an AutoPHY feature to start auto-negotiation with the data-link partner.
  • the protocol engine 34 first programs the desired link capabilities to registers before enabling the AutoPHY feature.
  • the internal MAC 126 attempts to negotiate capabilities with the PHY chip connected to the other end of the data link through the management data input/output (MDIO) registers on the PHY chip connected to the internal MAC 126.
  • MDIO is an IEEE standard two-wire bus that allows for communications with physical layer (PHY) devices.
  • the negotiation involves reading and writing to MDIO registers and is implemented in hardware.
  • the internal MAC 126 interrupts the protocol engine 34.
  • the data link capabilities that were negotiated are reported in registers. If the protocol engine 34 does not use the AutoPHY feature, it first examines the PHY connected to the internal MAC 126 by accessing the MDIO registers in the PHY. The protocol engine 34 then programs the capabilities accepted by the PHY to the MAC configuration registers.
  • the protocol engine 34 sets up desired capabilities through the AutoPHY feature or by manually examining the PHY chip before the internal MAC 126 can be enabled. The other required setup is to program the local Ethernet 48-bit hardware address. If multicast packets are supported, the multicast mask and address should also be programmed into the appropriate registers. If all other default configuration parameters in the registers are acceptable, the protocol engine 34 can enable the MAC interface 52 by setting the MAC configuration registers
  • This section describes how the MAC interface 52 may be reset.
  • the MAC interface 52 is first disabled by clearing bits in the MAC configuration register.
  • a soft reset available for the MAC interface 52 resets all state machines and buffer memory pointers maintained by the MAC buffer management block. However, almost all configurations are preserved by the internal MAC 126 upon soft reset.
  • the protocol engine 34 preferably waits until the soft-reset-done interrupt status sets before re- enabling either transmit or receive. Once the soft- reset-done interrupt is generated, the protocol engine 34 programs any unicast and multicast addresses through registers. Transmit and receive are now ready to be enabled by setting the MAC configuration register. Both hard resets and global soft resets perform a reset for the whole of the MAC interface 52, including all the configuration bits.
  • the network stack 50 uses a network stack internal memory 116 for its buffers and work area.
  • Figure 6 is a diagram that shows a network stack internal memory map according to the invention.
  • the network stack internal memory 116 is divided into two main sections of 2KB and 8KB. Both sections of the network stack internal memory 116 use single-port SRAM, although the sizes of the memory and the sections and their uses may be varied greatly and the network stack internal memory 116 may use different types of memory in different embodiments for different applications.
  • the IP filter engine 90 uses two different areas of the network stack internal memory 116. These two different areas are kept in different sections. The lower 4KB of the IP filter memory is the area into which packet data is streamed (for both transmit and receive).
  • From this lower 4KB IP filter memory data packets may be copied to or from internal memory 30 or to or from external memory 12. Alternatively data packets may be copied to and from the lower 4KB IP filter memory to the upper 2KB IP filter memory under control of the IP filter. This feature is useful for unsupported packet types, for example, which are put aside in the upper 2KB memory for the protocol engine 34 to handle.
  • the upper 2KB IP filter memory buffer is also used as a raw IP buffer.
  • the rest of the network stack internal memory 1 16 is used for TCP header assembly, PAP and CHAP authentication, and a shared buffer for PPP or ARP.
  • the communications processor 10 uses a protocol engine 34 for programmability.
  • This protocol engine 34 is also attached to a variety of peripherals, including a standard memory-management unit (MMU), which expands the addressable memory space of the protocol engine 34 to 1 MB.
  • MMU memory-management unit
  • the protocol engine 34 has access to all of the registers of the communication processor 10.
  • the protocol engine 34 has access to 1 MB of memory space. This memory space is divided into RAM and ROM memory types. A register within the MMU specifies the boundary between RAM and ROM memory types. Providing there is no other memory activity and the attached memory is fast enough, the protocol engine 34 can completes a memory accesses without added wait states.
  • the protocol engine 34 uses a set of registers and interrupts to interface to an optional external CPU, microprocessor or microcontroller 16. Eight interface registers are provided for any mutually agreed upon use by the protocol engine 34 and external CPU, microprocessor or microcontroller 16. When the external CPU, microprocessor or microcontroller 16 reads or writes data to any of the registers, an access interrupt may be made to trigger indicating to the protocol engine 34 that an interface register has been accessed by the external CPU, microprocessor, or microcontroller 16.
  • the external CPU, microprocessor or microcontroller 16 may also interrupt the protocol engine 34 by asserting a bit in a control register. This action causes an interrupt back to the protocol engine 34, assuming the protocol engine 34 has enabled the external interrupt. The protocol engine 34 can then clear this interrupt by writing to the control register.
  • the protocol engine 34 can send an interrupt back to the external CPU, microprocessor or microcontroller 16 by asserting an interrupt bit in the control register. This action causes the external controller interrupt to trigger, assuming that the external CPU, microprocessor or microcontroller 16 has enabled the interrupt.
  • the external CPU, microprocessor or microcontroller 16 can clear the interrupt by writing to the control register.
  • This section describes a direct data access mode that optimizes data transfers between the network stack 50 and an external CPU, microprocessor or microcontroller 16.
  • the protocol engine 34 When receiving data without the direct data access mode enabled, the protocol engine 34 must read data from the socket receive buffer 112, manage a temporary buffer in its memory space, and have the external CPU, microprocessor or microcontroller 16 read the data from the memory space of the protocol engine 34.
  • the external CPU, microprocessor or microcontroller 16 can read data directly from the socket receive buffer 112, avoiding a data copy.
  • the direct data access mode also applies for data writes. In the case of writes the external CPU, microprocessor or microcontroller 16 can write data directly to the socket transmit buffer 114.
  • the protocol engine 34 asserts the direct data mode bit in the miscellaneous control register.
  • the protocol engine 34 then writes the appropriate memory address to a register. This is the address of the register that the external CPU, microprocessor or microcontroller 16 is attempting to access.
  • the protocol engine 34 then informs the external CPU, microprocessor or microcontroller 16 how much data there is to be read, or how much room there is to write. Once the external CPU, microprocessor or microcontroller 16 has this information and is granted permission to use the direct data access mode, the external CPU, microprocessor or microcontroller 16 begins reading data or writing data.
  • the protocol engine 34 When using an external CPU, microprocessor or microcontroller 16 the protocol engine 34 has a mechanism via the direct data access mode to temporarily block the external CPU, microprocessor or microcontroller 16 from accessing the network stack 50. To block access, the protocol engine 34 first sets a bit in the miscellaneous control register. The protocol engine 34 then polls the idle bit in the same register and waits until that bit is asserted. The protocol engine 34 then de-asserts the direct data access mode bit in the miscellaneous control register. At this point, the protocol engine 34 may again access the network stack 50. When protocol engine 34 is finished with an access to the network stack 50, the protocol engine 34 de-asserts the block external CPU bit and re-asserts the direct data access mode bit.
  • MMU Memory-management unit
  • DMAC DMA controller
  • FIG. 7 is a block schematic diagram of the peripherals attached to a protocol engine 34 according to the invention.
  • a master SPI port is provided so that the protocol engine 34 can control slave SPI devices.
  • the SPI Serial Peripheral Interface
  • the SPI bus has 3 signals, SCK, MOSI, and MISO.
  • SPI devices are either masters or slaves. Master devices always provide the clock. The clock can be anywhere from near DC to 10MHz. Most slave devices also use a slave select (or chip select) to select the device on the bus and some slave devices provide feedback with a busy signal.
  • the communications processor 10 may also appear as an SPI slave device when the external CPU, microprocessor, or microcontroller interface 2 is configured as a serial interface (see Figure 1). Higher- speed serial and parallel buses may be used with the communications processor 10 described here without significant changes to the architecture.
  • the protocol engine 34 in the present version can only access 64KB of memory by itself. With the addition of the MMU, the protocol engine 34 memory is extended to 1 MB of physical memory. The protocol engine 34 memory is banked in such a way that at any given time, the protocol engine 34 is still only accessing 64KB of logical memory.
  • This section describes the direct memory access (DMA) engine and DMA controller (DMAC) 172. The DMAC 172 moves data from one memory or I/O location to another in an automatic fashion, thus allowing the protocol engine 34 to continue to perform other functions. All DMA transfers are performed in bytes.
  • DMA direct memory access
  • DMAC DMA controller
  • the protocol engine 34 programs the starting source address for the DMA transfers in a first set of registers.
  • the protocol engine 34 programs the starting destination address for the DMA transfers in a second set of registers.
  • the protocol engine 34 programs the DMA transfer type and the addressing modes (incrementing, decrementing, or stationary) into a DMA register. All addressing modes may be applied to both the source and destination addresses.
  • the DMA engine reads the first byte of data from the source at the given starting source address and temporarily holds the data in a register within the DMAC 172 until it can be written to the destination address.
  • the source address is updated after the source read, and the destination address is updated after the destination write.
  • the byte count decrements after the destination write as well.
  • the DMA enable bit is cleared. If the DMA interrupt enable bit is set then an interrupt for the protocol engine 34 is also generated.
  • This section describes the general-purpose timers and watchdog timer 178 used by the protocol engine 34 and communications processor 10.
  • the present version of the communications processor 10 supports four general-purpose 32-bit timers that may either be used independently or joined together in a cascaded fashion. All timers provide a single programmable counter that triggers either a one-time interrupt or continuously triggers a repeating-loop interrupt that repeats at a programmed periodic rate.
  • the specialized data processing engines operate at or near the Presentation or Application layer.
  • the specialized protocol processing engines operate at the Network, Transport, or Presentation layers (often called upper-level protocols, those that are layered above TCP or IP, for example).
  • Base64 is used as an encoding scheme for transmitting binary data over Internet connections. It takes three characters in, and transforms them into four characters within a 64-character mapping.
  • the preferred Base64 encoder and decoder implements the Base64 algorithm as specified and described in the Internet Engineering Task Force (IETF) RFC1341 ,.
  • Base64 is an encoding scheme and not a compression scheme, in that Base64 takes three bytes of data and transforms the three bytes of data into four bytes of data. Therefore, the transformed data takes up 1/3 more space then the original data.
  • a [CR, LF] arriage-return and linefeed character pair
  • Text rasterization converts incoming packet data that is in ASCII format to a bitmap format. This bitmap format is then used for printing to specialized devices, such as an LCD screen or a printer.
  • the text-rasterization engine 64 has two different rasterization modes, 8-bit ASCII and 16-bit character mode. A different font memory is supplied depending on which rasterization mode is used. If the hardware-assisted text- rasterization engine 64 is used in conjunction with the G3 engine 42, then the G3 engine 42 must be enabled prior to enabling the hardware-assisted text- rasterization engine 64.
  • the G3 encoder 42 takes output from the hardware-assisted text-rasterization engine 64, and Huffman encodes the data to put it in the proper format for fax transmission.
  • a source memory address which contains the rasterized data, and a target memory address, where the encoded data is stored, is programmed into the G3 encoder 42 prior to the start of each session.
  • the Mime string search engine 62 searches a buffer for specified character strings. It reports back the starting and ending offsets for the string, and is also capable of searching across multiple buffers.
  • the Mime string search engine 62 can also automatically search a data buffer for the POP termination string: ([CR] [LF] [.] [CR] [LF]).
  • This type of specialized data processing engine might equally well be used, for example, to insert or detect tags, markers, or perform framing in a streaming protocol such as TCP in order to convert such a streaming protocol into a block-based protocol.
  • the ADPCM accelerator engine 38 provides 2:1 and 4:1 compression and decompression functions.
  • the ADPCM accelerator engine 38 operates on a buffer of data in memory, and puts the compressed or decompressed data back to memory. For compression the source and destination memory addresses can be the same because the compressed data take less room than the original data.
  • FIG. 8 is a block schematic diagram showing an IP-only mode data path according to the invention.
  • the IP-only mode is provided so that non-Ethernet or non-PPP data links can be used. By default, the IP-only mode comes up disabled after resets.
  • the IP-only mode comes up disabled after resets.
  • the IP-only mode is enabled, both the Ethernet data link and the PPP data link are disabled.
  • the data packets are tapped off at the IP raw mux 104. Therefore, the IP router - bottom engine 92 is configured to send all data out the Ethernet data link.
  • the IP-only data still uses the MAC transmit and receive buffers 118. This prevents under-run conditions and allows for retransmission on bad packets.
  • the system programs the local IP Address registers in the ARP engine 72.
  • the IP-only transmit interface works through the data link SPI interface.
  • the data format used for the IP packet starts with the IP header, and goes through the data field. No extra checksum or padding bytes are appended
  • the data link SPI interface is used when communicating using the IP-only mode or when using an external MAC.
  • the data link SPI interface is disabled.
  • receiving packets only one data packet is stored in the receive buffer at a time. This only applies to packets that are made available to the protocol engine 34 because all data packets go to the network stack 50. If another non-data packet is received, but the previous packet is still in the data link SPI receive buffer, then the second packet is discarded.
  • This section describes the integrated test and debug features of the communications processor 10.
  • the debug features allow breaking on an address, and single stepping.
  • the address comparison is made with the protocol engine 34 physical 20-bit address (1 MB memory space). Breaks can be triggered on either reads or writes, with each type of operation individually controlled. Two separate break-point addresses are provided for flexibility. All registers associated with the protocol engine 34 debugger are located in the protocol engine 34 miscellaneous index registers.
  • the communications processor uses built-in self-test (BIST) to test the internal RAMs, scan testing for general fault coverage, and NAND-tree logic for parametric I/O testing. Four dedicated test pins are provided for the communications processor 10.
  • the communications processor 10 features a clocking mechanism that allows it to run the MAC buffers 118 ( Figure 3) at a higher clock frequency then the rest of the communications processor 10.
  • the network stack logic 50 can be made to operate at a different clock frequency then the protocol engine 34.
  • the protocol engine 34 may also slow itself down in order to conserve power during idle periods.
  • the clocking logic also provides for a programmable clock output that can be used to drive system logic 18. ( Figure 1 )

Abstract

Processeur de communication, tel qu'un syntoniseur Internet, possédant des caractéristiques avantageuses, telles qu'un support LAN, une interface SPI (128), un point d'accès spécialisé (56) et un ADPCM (22) pour des applications audio. L'invention décrit un module d'accès au réseau économique, consommant peu de puissance, de fabrication simple, possédant un facteur de forme limité, un besoin en mémoire réduit et permettant d'effectuer un décodage de protocole extrêmement efficace. L'invention est basée sur un système intégré machine décodant à la fois des protocoles de réseau multiples de façon continue et simultanée et traitant des paquets de données en une seule opération, ce qui permet de limiter la mémoire du système et les besoins en facteur de forme, tout en éliminant également la surcharge logicielle du processeur central.
PCT/US2002/002293 2001-01-26 2002-01-25 Processeur de communication WO2002059757A1 (fr)

Priority Applications (4)

Application Number Priority Date Filing Date Title
US10/470,365 US7379475B2 (en) 2002-01-25 2002-01-25 Communications processor
US11/546,031 US8059680B2 (en) 2001-01-26 2006-10-10 Offload system, method, and computer program product for processing network communications associated with a plurality of ports
US11/546,032 US8073002B2 (en) 2001-01-26 2006-10-10 System, method, and computer program product for multi-mode network interface operation
US11/932,428 US7646790B2 (en) 2001-01-26 2007-10-31 Communications processor

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US26438101P 2001-01-26 2001-01-26
US60/264,381 2001-01-26

Related Child Applications (4)

Application Number Title Priority Date Filing Date
US10470365 A-371-Of-International 2002-01-25
US11/546,031 Continuation US8059680B2 (en) 2001-01-26 2006-10-10 Offload system, method, and computer program product for processing network communications associated with a plurality of ports
US11/546,032 Continuation US8073002B2 (en) 2001-01-26 2006-10-10 System, method, and computer program product for multi-mode network interface operation
US11/932,428 Continuation US7646790B2 (en) 2001-01-26 2007-10-31 Communications processor

Publications (1)

Publication Number Publication Date
WO2002059757A1 true WO2002059757A1 (fr) 2002-08-01

Family

ID=23005809

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2002/002293 WO2002059757A1 (fr) 2001-01-26 2002-01-25 Processeur de communication

Country Status (1)

Country Link
WO (1) WO2002059757A1 (fr)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004055683A2 (fr) * 2002-12-13 2004-07-01 Nvidia Corporation Procede et appareil pour effectuer les fonctions de traitement en reseau
US7383352B2 (en) 2002-05-13 2008-06-03 Nvidia Corporation Method and apparatus for providing an integrated network of processors
US7620070B1 (en) 2003-06-24 2009-11-17 Nvidia Corporation Packet processing with re-insertion into network interface circuitry
US9794378B2 (en) 2006-11-08 2017-10-17 Standard Microsystems Corporation Network traffic controller (NTC)
EP2763364A3 (fr) * 2013-01-30 2018-03-21 Rockwell Automation Technologies, Inc. Dispositif industriel configurable de réseau
CN115086109A (zh) * 2016-02-18 2022-09-20 瑞萨电子株式会社 消息处理器

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5289580A (en) * 1991-05-10 1994-02-22 Unisys Corporation Programmable multiple I/O interface controller
US5303344A (en) * 1989-03-13 1994-04-12 Hitachi, Ltd. Protocol processing apparatus for use in interfacing network connected computer systems utilizing separate paths for control information and data transfer
US5671355A (en) * 1992-06-26 1997-09-23 Predacomm, Inc. Reconfigurable network interface apparatus and method
WO1998050852A1 (fr) * 1997-05-08 1998-11-12 Iready Corporation Accelerateur de materiel pour langage de programmation oriente objet
US6034963A (en) * 1996-10-31 2000-03-07 Iready Corporation Multiple network protocol encoder/decoder and data processor
US6061368A (en) * 1997-11-05 2000-05-09 Xylan Corporation Custom circuitry for adaptive hardware routing engine

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5303344A (en) * 1989-03-13 1994-04-12 Hitachi, Ltd. Protocol processing apparatus for use in interfacing network connected computer systems utilizing separate paths for control information and data transfer
US5289580A (en) * 1991-05-10 1994-02-22 Unisys Corporation Programmable multiple I/O interface controller
US5671355A (en) * 1992-06-26 1997-09-23 Predacomm, Inc. Reconfigurable network interface apparatus and method
US6034963A (en) * 1996-10-31 2000-03-07 Iready Corporation Multiple network protocol encoder/decoder and data processor
WO1998050852A1 (fr) * 1997-05-08 1998-11-12 Iready Corporation Accelerateur de materiel pour langage de programmation oriente objet
US6061368A (en) * 1997-11-05 2000-05-09 Xylan Corporation Custom circuitry for adaptive hardware routing engine

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
KRISHNAKUMAR ET AL.: "The programmable protocol VLSI engine (PROVE)", IEEE, vol. 42, no. 8, August 1994 (1994-08-01), pages 2630 - 2642, XP000462378 *
KRISHNAKUMAR ET AL.: "VLSI implementations of communication protocols - a survey", IEEE, vol. 7, no. 7, September 1989 (1989-09-01), pages 1082 - 1090, XP000054537 *
MANSOUR, MOHAMMAD ET AL.: "FPGA-based internet protocol version 6 router", COMPUTER DESIGN: VLSI IN COMPUTERS AND PROCESSORS, 5 October 1998 (1998-10-05) - 7 October 1998 (1998-10-07), pages 334 - 339, XP010310332 *
MCCANNY, JOHN ET AL.: "Hierarchical VHDL libraries for DSP ASIC design", ACOUSTICS, SPEECH AND SIGNAL PROCESSING, 21 April 1997 (1997-04-21) - 24 April 1997 (1997-04-24), pages 675 - 678, XP000789236 *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7383352B2 (en) 2002-05-13 2008-06-03 Nvidia Corporation Method and apparatus for providing an integrated network of processors
US7620738B2 (en) 2002-05-13 2009-11-17 Nvidia Corporation Method and apparatus for providing an integrated network of processors
WO2004055683A2 (fr) * 2002-12-13 2004-07-01 Nvidia Corporation Procede et appareil pour effectuer les fonctions de traitement en reseau
WO2004055683A3 (fr) * 2002-12-13 2004-11-11 Nvidia Corp Procede et appareil pour effectuer les fonctions de traitement en reseau
US7397797B2 (en) 2002-12-13 2008-07-08 Nvidia Corporation Method and apparatus for performing network processing functions
US7620070B1 (en) 2003-06-24 2009-11-17 Nvidia Corporation Packet processing with re-insertion into network interface circuitry
US9794378B2 (en) 2006-11-08 2017-10-17 Standard Microsystems Corporation Network traffic controller (NTC)
US10749994B2 (en) 2006-11-08 2020-08-18 Standard Microsystems Corporation Network traffic controller (NTC)
EP2763364A3 (fr) * 2013-01-30 2018-03-21 Rockwell Automation Technologies, Inc. Dispositif industriel configurable de réseau
EP3866420A1 (fr) * 2013-01-30 2021-08-18 Rockwell Automation Technologies, Inc. Dispositif industriel configurable de réseau
CN115086109A (zh) * 2016-02-18 2022-09-20 瑞萨电子株式会社 消息处理器
US11876879B2 (en) 2016-02-18 2024-01-16 Renesas Electronics Corporation Message handler

Similar Documents

Publication Publication Date Title
US7379475B2 (en) Communications processor
EP1382145B1 (fr) Adaptateur ethernet gigabit
US6947430B2 (en) Network adapter with embedded deep packet processing
US6434620B1 (en) TCP/IP offload network interface device
US20070253430A1 (en) Gigabit Ethernet Adapter
US20040078480A1 (en) Parsing a packet header
US20050157732A1 (en) Method and apparatus for emulating ethernet functionality over a serial bus
US7269661B2 (en) Method using receive and transmit protocol aware logic modules for confirming checksum values stored in network packet
JP2006325054A (ja) Ｔｃｐ／ｉｐ受信処理回路及びそれを具備する半導体集積回路
JPH11261649A (ja) データ処理装置及びそれを適用したルータ・ブリッジ
US7293113B1 (en) Data communication system with hardware protocol parser and method therefor
Lofgren et al. An analysis of FPGA-based UDP/IP stack parallelism for embedded Ethernet connectivity
JP4658546B2 (ja) フェイルオーバーイベントをサポートするネットワーク状態オブジェクトの多重オフロード
US7561585B2 (en) Manufacture and method for accelerating network address translation
JP2002171274A (ja) データ転送方法及びデータ転送装置
WO2002059757A1 (fr) Processeur de communication
US20060062229A1 (en) Terminal adapter device capable of performing IEEE1394-to-Ethernet conversion
Hohne et al. 3D-segmentation and display of tomographic imagery
JP4916482B2 (ja) ギガビット・イーサネット・アダプタ
TWI825293B (zh) 應用在網路裝置中的電路
Batmaz et al. UDP/IP Protocol Stack with PCIe Interface on FPGA
US20020078246A1 (en) Method and system for network protocol processing
US20040167985A1 (en) Internet protocol access controller
Huang et al. EIPv6: A reduced IPv6 protocol stack for embedded systems
Chen et al. A New Design of Embedded IPv4/IPv6 Dual-Stack Protocol

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 10470365

Country of ref document: US

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP