U.S. patent application number 11/399628 was filed with the patent office on 2006-08-17 for reduced hardware network adapter and communication method.
This patent application is currently assigned to Xircom, Inc.. Invention is credited to Michael R. Conley, Eric Henderson.
Application Number | 20060184698 11/399628 |
Document ID | / |
Family ID | 23585878 |
Filed Date | 2006-08-17 |
United States Patent
Application |
20060184698 |
Kind Code |
A1 |
Conley; Michael R. ; et
al. |
August 17, 2006 |
Reduced hardware network adapter and communication method
Abstract
The present invention provides a network interface adapter for
connecting a client computer to a computer network that includes a
reduced hardware media access controller (MAC) coupled through a
physical interface (PHY) to the network physical link. A
significant portion of the MAC functionality is implemented as
software within the processor of the host client computer. The
hardware portion of the preferred MAC implementation provides
memory for buffering communications between the PHY and the client
computer. The preferred hardware aspects of a MAC in accordance
with the present invention also includes a register interface for
register-driven communications between the hardware portion of the
MAC and the software portions of the MAC implemented within the
client computer. By implementing most of the MAC functionality in
software within the host computer, the preferred MAC provides lower
cost, lower power consumption, and generally greater
flexibility.
Inventors: |
Conley; Michael R.;
(Thousand Oaks, CA) ; Henderson; Eric; (Salt Lake
City, UT) |
Correspondence
Address: |
BLAKELY SOKOLOFF TAYLOR & ZAFMAN
12400 WILSHIRE BOULEVARD
SEVENTH FLOOR
LOS ANGELES
CA
90025-1030
US
|
Assignee: |
Xircom, Inc.
|
Family ID: |
23585878 |
Appl. No.: |
11/399628 |
Filed: |
April 6, 2006 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
10682572 |
Oct 8, 2003 |
|
|
|
11399628 |
Apr 6, 2006 |
|
|
|
09401005 |
Sep 21, 1999 |
6651107 |
|
|
10682572 |
Oct 8, 2003 |
|
|
|
Current U.S.
Class: |
709/250 |
Current CPC
Class: |
H04L 49/90 20130101;
H04L 49/9073 20130101; Y02D 10/00 20180101; H04L 12/40013 20130101;
G06F 13/385 20130101; Y02D 10/151 20180101; H04L 12/40032 20130101;
Y02D 10/14 20180101; H04L 12/413 20130101 |
Class at
Publication: |
709/250 |
International
Class: |
G06F 15/16 20060101
G06F015/16 |
Claims
1-17. (canceled)
18. A system comprising: a hardware media access controller, the
hardware media access controller including a memory to buffer data
transferred between a physical interface to a network and a
computer, and the hardware media access controller including a
register interface; and a memory storing media access control
software that if executed by a processor of the computer causes the
processor to perform one or more media access control
communications functions.
19. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more IEEE 802.3 functions.
20. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more network interface control functions.
21. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to detect an
error condition.
22. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to perform
an error handling function.
23. The system of claim 22, wherein the software comprises software
that if executed by the processor causes the processor to perform
the error handling function by performing a backoff algorithm.
24. The system of claim 22, wherein the software comprises software
that if executed by the processor causes the processor to perform
the error handling function by flushing the memory of the hardware
media access controller of data.
25. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to handle
interrupts and data transfer.
26. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to process a
data packet including detecting a frame preamble, and checking a
length of a frame.
27. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to evaluate
a frame using frame check sequence data.
28. The system of claim 27, wherein the software comprises software
that if executed by the processor causes the processor to evaluate
the frame using the frame check sequence data by performing a
polynomial division using a protocol defined check polynomial.
29. The system of claim 18, wherein the software comprises software
that if executed by the processor causes the processor to perform
filtering to detect whether an address of a packet is relevant to
the computer.
30. The system of claim 18, wherein the memory comprises a memory
of the computer.
31. The system of claim 18, wherein a size of the memory is based
on a typical latency delay of the computer.
32. A system comprising: a computer including a processor, a DRAM
memory, and a bus; an adapter including, a physical interface to
couple to a network, and a hardware media access controller coupled
to the physical interface and to the bus, the hardware media access
controller including a memory to buffer data that is transferred
between the physical interface and the computer, and the hardware
media access controller including a register interface; and media
access control software stored within the DRAM memory that if
executed by the processor of the computer causes the processor to
perform one or more media access control communications
functions.
33. The system of claim 32, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more IEEE 802.3 functions.
34. The system of claim 32, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more network interface control functions.
35. The system of claim 32, wherein the software comprises software
that if executed by the processor causes the processor to detect an
error condition and perform an error handling function.
36. The system of claim 32, wherein the software comprises software
that if executed by the processor causes the processor to process a
data packet including detecting a frame preamble, and checking a
length of a frame.
37. The system of claim 32, wherein the software comprises software
that if executed by the processor causes the processor to evaluate
a frame using frame check sequence data.
38. A memory storing media access control software that if executed
by a processor of a computer causes the processor to: perform one
or more media access control communications functions.
39. The memory of claim 38, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more IEEE 802.3 functions.
40. The memory of claim 38, wherein the software comprises software
that if executed by the processor causes the processor to perform
one or more network interface control functions.
41. The memory of claim 38, wherein the software comprises software
that if executed by the processor causes the processor to detect an
error condition and perform an error handling function.
42. The memory of claim 38, wherein the software comprises software
that if executed by the processor causes the processor to process a
data packet including detecting a frame preamble, and checking a
length of a frame.
43. The memory of claim 38, wherein the software comprises software
that if executed by the processor causes the processor to evaluate
a frame using frame check sequence data.
44. A method comprising: installing a card in a computer, the card
including a physical interface to a network and a hardware media
access controller; and installing media access control software
within a memory of the computer, the media access control software
including software that if executed results in a processor of the
computer performing one or more media access control communications
functions.
45. The method of claim 44, wherein said installing the software
comprises installing software that if executed results in the
processor evaluating a frame using frame check sequence data.
Description
BACKGROUND OF THE INVENTION
[0001] 1. Field of the Invention
[0002] The present invention relates to adapters, interfaces and
connections between a computer and a network. As a specific
example, the present invention is described in terms of an adapter
and a communication method between a computer and a computer
network that functions in accordance with an industry standard
protocol such as IEEE 802.3 or one of its extensions.
[0003] 2. Description of the Related Art
[0004] The present invention generally relates to an adapter for
use in connecting a computer, referred to here as a client
computer, to a network. The invention will be described and
explained with reference to a particular implementation designed
for one of the most common network configurations presently in use,
commonly known as an Ethernet.TM. network. This type of network
more generally includes those networks that implement the IEEE
802.3 standard, or one of the variations, modifications or
improvements on that standard. For simplicity, all of these
networks are referenced together as an IEEE 802.3 network, although
that designation is not necessarily precise. The networks
referenced here provide a framework for understanding the present
invention, as do the adapters or network interface controllers that
have been developed for those networks.
[0005] Client computers of a local area network are often connected
together over a common physical link such as coaxial cable,
unshielded twisted pairs of wires or shielded twisted pairs of
wires. Information is transmitted by client computers onto the
physical link in units of frames having one of a number of
protocol-defined data structures and a quantity of data ranging
from that of a minimum frame length and that of a maximum frame
length. The IEEE 802.3 protocol in its simplest form allows only
one client computer to transmit information over any distinct
physical link at one time. If two client computers provide data to
the physical link at the same time, or sufficiently close in time
as to create interference between two message signals, the data are
corrupted and must be discarded. Such an event is referred to as a
collision and must be detected by the client computers. IEEE 802.3
networks use carrier sense multiple access/collision detection
(CSMA/CD) to govern access to the physical link. Before a client
computer transmits a message to the physical link, it detects
whether a carrier signal is present on the physical link. If a
carrier signal is present on the physical link, the client computer
waits before transmitting the message until no carrier signal is
present and the physical link is available.
[0006] Once a client computer determines that the physical link is
available, the client computer transmits the message to the
physical link and onto the network. The client computer monitors
the physical link to detect any collision that might occur while it
transmits the message over the physical link. The transmitting
computer continues to monitor for collisions after transmission
during a subsequent message in progress period of time in which the
transmitting computer is waiting for the message to reach its
destination. Should the client computer detect a collision, the
client computer continues transmitting a signal, sometimes referred
to as a jam sequence, until the transmission continues for at least
a period of time corresponding to a minimum frame length. The
client computer then pauses for a predetermined, random amount of
time and attempts to gain access to the physical link to once again
transmit the data. Each of the client computers attempting to
access the physical link executes a similar backoff algorithm, but
with different, randomly assigned waiting times so that a single
client computer gains access to the physical link, with others of
the waiting computers gaining access to the network
successively.
[0007] Information is typically organized into frames for
transmission over an Ethernet.TM. or other type of CSMA/CD network.
FIG. 1 shows the organization of a frame of data for an
Ethernet.TM. frame and FIG. 2 shows the organization for a frame
defined in accordance with the standard IEEE 802.3. Both types of
networks use a preamble consisting of an alternating pattern of
ones and zeros to inform a receiving station that a frame is
coming. The Ethernet.TM. preamble (FIG. 1) further includes an
additional byte that is the equivalent of the start of frame byte
field defined in the IEEE 802.3 network. The start of frame byte
(SOF in FIG. 2) ends with two consecutive "1" bits and is used to
synchronize the frame reception of the station connected to the
physical link.
[0008] The framing of information for these networks includes both
the destination address and the source address for the message. The
destination address may be a single target computer (unicast), a
group of computers (multicast) or all computers on the network
(broadcast). The source address is the particular transmitting
computer. Ethernet.TM. frames include a type field identifying the
protocol of an upper layer application to receive the message. This
field is not present in frames for the IEEE 802.3 network and is
replaced by a length field that specifies the number of bytes of
data in the message. Both frame structures provide a data field of
arbitrary length within a range of possible lengths, followed by a
frame check sequence (FCS), which is a four-byte cyclic redundancy
check value. The FCS is created by the sending computer and is
recalculated by the receiving device to check for damaged
frames.
[0009] Client computers communicating frames of data over a CSMA/CD
network use a network interface controller to perform carrier
sense, collision detection, and other aspects of data transmission
and receive control. Control of data transmission includes
generating the frame format and calculating the FCS byte. Data
receive control includes the detection of a frame, examining the
destination address to determine if the message is intended for
that computer, and performing a CRC or other frame check procedure
to determine if the frame is valid. Other forms of analysis may be
performed on the frame or may have to be performed when there is an
error in the frame of data or in the receive operation. For
example, information may have to be retransmitted if an error is
detected during transmission. All of this processing is
conventionally known and performed in typical adapters or
controllers that link client computers to a local area CSMA/CD
network.
[0010] The network interface controller is implemented as an
integrated circuit, such as an application specific integrated
circuit (ASIC). An example of one configuration of Ethernet
controller in an ASIC, described in U.S. Pat. No. 5,872,920, is
illustrated schematically in FIG. 3. The network controller ASIC 10
interfaces with a bus 20 of a host computer system and with the
twisted pairs or coaxial cable that are part of the physical link
30 of the network. Information is transmitted to and received from
the physical interface by a transceiver 40 or through an attachment
unit interface 42. Information to be transmitted over the physical
link 30 is encoded by encoder 44 and decoder 46 decodes information
received from the physical link. Typically, Manchester encoding and
decoding are used for IEEE 802.3 networks.
[0011] A controller 50, which may be a microcontroller or other
type of processor, is typically provided as a core within the ASIC
10 to control the transmit and receive operations using appropriate
transmit control 52 and receive control 54 programs or state
machines. These programs handle the various data control operations
required for transmitting and receiving data from a CSMA/CD network
including, for example, handling error conditions for a collision
on the physical medium and retransmitting corrupted data as
necessary. Most all of the functionality desired to implement
applicable standards, such as IEEE 802.3, is implemented within the
controller 50. Data coming in and out of the controller 50 are
buffered by a transmit FIFO 56 and a receive FIFO 58.
Communications with the host computer, including the provision of
data to the bus 20 of the host computer, are managed by the host
interface 60. Provisions are made to update the host interface 60
by rewriting or updating a data set or program stored within the
EEPROM 62. Further details of these circuits and the functionality
they implement are described in U.S. Pat. No. 5,872,920, which is
hereby incorporated by reference.
[0012] It should be appreciated that, while the network interface
controller of FIG. 3 is implemented as a single ASIC other
configurations are also common. For example, because the physical
implementations for a particular network (including driving
magnetics, digital to analog circuitry and analog to digital
circuitry) can vary greatly, it is often desirable to provide the
encoder, decoder, transceiver and other aspects of the interface to
the physical link in a dedicated chip. This architecture may be
implemented as a distinct PHY chip, allowing a greater level of
flexibility that is often more cost-effective than a more
integrated, single-chip network interface controller solution. In
implementations that provide a distinct PHY chip, it is typical to
provide other aspects of the network controller on a second chip, a
media access controller (MAC) chip. The MAC of the network
controller shown in FIG. 3 includes the controller or processor 50
with its programs, the buffering memory and, in most cases, the
host-interface 60.
[0013] Considerable effort has been made to increase the
functionality provided within integrated circuit network interface
controllers such as the ASIC 10 illustrated in FIG. 3. For example,
additional functionality and flexibility has been built into
network interface controller chips to accommodate improvements or
modifications in network communications standards. Particular
attention has been paid to higher speed variations of the basic
IEEE 802.3 technology.
SUMMARY OF THE PREFERRED EMBODIMENTS
[0014] Preferred embodiments of the present invention provide a
simplified controller that is better adapted to a low-cost,
network-compatible computer. By providing a relatively simple
interface structure and by implementing more of the network
interface control functionality within a processor on the host
computer, such preferred embodiments of the invention provide a low
cost, highly flexible solution for a network interface.
[0015] An aspect of the present invention provides a computer
communications system having a transmit buffer coupled to at least
one transmit data line. The transmit buffer receives data from a
host computer and temporarily store the data before transmitting
the data over the transmit data line to a physical link of a data
network. A receive buffer is coupled to at least one receive data
line, the receive buffer adapted to receive data from a physical
link of a data network over the receive data line and temporarily
store the data before providing the data to a computer. An array of
communication registers is also provided. The array includes a data
register coupled to the receive buffer, wherein repeated reads from
the data register causes data to be read from the receive buffer.
The array further includes a status register for storing data
identifying a collision of data on the physical link of the
network, where the status register is readable from the host
computer. The status register also includes at least one bit
signifying an interrupt state in at least one of a read and a write
operation.
[0016] Another aspect of the invention provides a computer
communications system comprising a media access controller
including a receive buffer coupled to receive data from a data
network and temporarily store the data before providing the data to
a host computer. The media access controller has communication
registers including a data register coupled to the receive buffer,
wherein repeated reads from the data register reads data from the
receive buffer. The communication registers provide a status
register storing at least one interrupt bit. The interrupt bit is
set to indicate the presence of data received from the data network
and destined for the host computer and the interrupt bit is
readable by the host computer to indicate the presence of data to
be read.
[0017] Still another aspect of the invention provides a computer
communications system comprising a media access controller
including a plurality of communication registers. The communication
registers preferably include a data register coupled to the receive
buffer, wherein repeated reads from the data register read data
received from the data network. A status register stores at least
one interrupt bit, the interrupt bit set to indicate the presence
of data received from the data network and destined for the host
computer, the interrupt bit readable by the host computer. A media
independent interface register stores at least one signal for
controlling operation of PHY circuitry coupled to the media access
controller, wherein signals from the host computer are passed from
the media independent interface register to control PHY circuitry
as data are read from the data network. A byte-count register
stores a value indicative of the number of bytes stored in the
media access controller for transfer to the host computer, where
the byte count register is decremented as data are read from the
media access controller.
BRIEF DESCRIPTION OF THE DRAWINGS
[0018] These and other aspects of the present invention are better
understood by reference to the following description that
references and includes the following figures.
[0019] FIG. 1 illustrates a frame structure for an Ethernet
network.
[0020] FIG. 2 illustrates a frame structure for a network in
accordance with the IEEE 802.3 standard.
[0021] FIG. 3 illustrates a network interface control chip that
provides a high level of functionality on the chip.
[0022] FIG. 4 illustrates an architectural overview of an aspect of
the present invention.
[0023] FIG. 5 illustrates a particularly preferred implementation
of a network controller in accordance with the present
invention.
[0024] FIG. 6 illustrates the network interface controller of FIG.
5 coupling a client computer to a network.
[0025] FIG. 7 shows the addresses and functions for registers
within a preferred register set provided within the controller of
FIG. 5.
[0026] FIG. 8 illustrates the structure and functions of the
Ethernet data register of the FIG. 7 register set.
[0027] FIG. 9 illustrates the structure and functions of the byte
count register of the FIG. 7 register set.
[0028] FIG. 10 illustrates the structure and functions of the
command/status register of the FIG. 7 register set for write
operations.
[0029] FIG. 11 illustrates the structure and functions of the
command/status register of the FIG. 7 register set for read
operations.
[0030] FIG. 12 illustrates the structure and functions of the MII
(media independent interface) register of the FIG. 7 register
set.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0031] Communication between a computer and a network is
conventionally conducted through a network interface controller
(NIC) or adapter. Preferred embodiments of the present invention
provide a reduced hardware implementation of a network interface
controller. Implementations of preferred aspects of an adapter in
accordance with the invention may provide computers with network
interface connections that are lower in cost and consume less power
than conventional controllers. Aspects of the present invention
make it possible to implement adapters in a manner that is far
easier to modify to accommodate changes in technology or to provide
solutions for unusual or specialized applications.
[0032] Particularly preferred embodiments of the present invention
provide a reduced hardware MAC with a significant portion of the
MAC functionality implemented as software within the host client
computer. Most preferably, the MAC functionality embodied within
the client computer is implemented as software within a processor
of the client computer and most preferably the MAC functionality is
implemented within the main processor of a personal computer type
of architecture. The hardware portion of the preferred MAC
implementation provides memory for buffering communications between
the PHY and the client computer. The preferred hardware aspects of
a MAC in accordance with the present invention also includes a
register interface for register-driven communications between the
hardware portion of the MAC and the software portions of the MAC
implemented within the client computer. By implementing most of the
MAC functionality in software within the host computer, the
preferred MAC provides lower cost, lower power consumption, and
generally greater flexibility.
[0033] Aspects of the present invention can provide an interface
between a client or other computer and a network such as a local
area network functioning in accordance with any of the defined IEEE
802.3 protocols. Other related aspects of the present invention
provide a method of communicating information to, from or between a
computer and one or more other computers through an IEEE 802.3
network. The term IEEE 802.3 is used broadly here to generally
encompass CSMA/CD networks including presently planned variations
including gigabit Ethernet.TM. and other variations that might be
developed later. Aspects of the invention are believed to have
advantages in interfacing to and communicating with other types of
networks using other types of protocols. Many of the examples and
explanations provided here are set forth in terms of communicating
with an IEEE 802.3 network, both because of the familiarity of
these networks and because of the present prominence of IEEE 802.3
and Ethernet.TM. networks. The specific applications used to
illustrate the invention are not, however, intended to limit the
scope of the invention.
[0034] Similarly, illustrations of the present invention are made
within the framework of presently available and used IEEE 802.3
protocols. More advanced implementations of IEEE 802.3 and
successors to IEEE 802.3 will be developed and brought to market in
the future. Many aspects of the present invention are expected to
have advantages when applied to such networks, as well.
[0035] FIG. 4 illustrates, from a schematic architectural point of
view, aspects of a client computer 70 that are involved in
communicating information to and from a network. The data network
includes a physical link such as wire pairs or coaxial cabling that
carries the actual data signals to and from the client computer.
The client computer 72 connects to this physical link through the
physical layer of the network model indicated as PHY 72 in FIG. 4.
The PHY might include the magnetic circuitry for placing signals
onto and deriving signals from the physical link. PHY 72 also
includes appropriate support circuitry such as analog to digital
converters, digital to analog converters, and phase lock loop
detection circuitry for recovering signals from the physical link
according to a received or derived clock signal. Generally, the PHY
72 also includes encoding and decoding circuitry such as that
illustrated in FIG. 3.
[0036] The PHY 72 is generally adapted to a particular network
protocol or definition. As such, aspects of the PHY 72 may vary
significantly, but in a well-known and understood manner, between
different applications of the architecture of FIG. 4. PHY 72 is
commercially available in a number of forms in different network
environments, whether as a distinct chip or as a core to be
provided within an ASIC.
[0037] The next higher level of the network model illustrated in
FIG. 4 is the media access control or MAC 74. The MAC 74 may embody
a range of functionality, but generally serves the purpose of
initially analyzing and formatting the stream of data bits provided
by PHY 72 so that the data are in a form immediately usable by the
upper levels of the network model. MAC 74 interfaces with the
operating system 76 of the client computer 70 for the storage of
data received from the network. MAC 74 is generally responsive to
application programs 78 within the client computer that communicate
through the operating system 76 to generate messages for
transmittal or requests for information.
[0038] MAC functionality may include a number of different aspects
of communications. For example, data from the network might be
provided by PHY 72 in 4-bit parallel format. The MAC preferably
reformats that 4-bit parallel data into bytes or words appropriate
for the client computer, for example, 8-bit bytes or 32-bit words.
Other aspects of communication that might be included in the MAC
are address recognition, frame recognition, frame analysis, and
detection of and management of collisions or other forms of network
errors. As discussed in the background and illustrated in FIG. 3,
conventional implementations of MAC 74 include a variety of
functions on a single integrated circuit. Such conventional
implementations of a MAC provide a processor or other form of logic
capable of executing a state machine or program that performs these
functions.
[0039] Preferred implementations of the present invention provide a
simplified MAC, at least from a circuitry point of view. In most
implementations, embodiments of the present invention will include
a simplified hardware MAC and a complementary software MAC
preferably within a host processor that will, in combination,
provide much if not more of the functionality performed by the MAC
illustrated in FIG. 3. Of course, implementations of a MAC in
accordance with an embodiment of the present invention might
provide less of the MAC functionality and still practice aspects of
the present invention. In contrast to the MAC of FIG. 3, a MAC in
accordance with preferred aspects of the present invention provides
a minimal set of circuitry and implements most aspects of MAC
functionality within a processor on the client computer 70. In a
present implementation, a MAC in accordance with the invention
might include buffering memory and a communication register set
with other aspects of MAC functionality implemented as software to
be executed by the microprocessor of a personal computer or similar
type of client computer.
[0040] FIG. 5 illustrates particularly preferred hardware aspects
of a network interface adapter 79 in accordance with the present
invention. As illustrated, the adapter includes a PHY 72
appropriate to the target network physical link and a reduced
hardware MAC. The reduced hardware MAC includes memory for
buffering data as it is transferred to or received from the PHY 72
including, preferably, a latch for storing a receive error signal
that may occur only for a period of the total time required to
receive a frame of data. The remaining hardware aspects of the
illustrated and particularly preferred MAC is a set of registers
for handling the communication of instructions, status and error
information and for facilitating data transfer between the adapter
and the client computer. Thus, the illustrated bus register
interface 80 preferably includes a complete but relatively minimal
set of communication and data registers so that most of the
communication functions are performed within the client computer
and preferably in the processor of the client computer.
[0041] Still referring to FIG. 5, the top two lines are signals for
communicating with the media independent interface (MII) within PHY
72. The signal line MDCK provides a data clock signal from the
interface 80 to the PHY 72 and is used for both reading and writing
data to the MII of PHY 72. MDIO is a serial data line for MII
communications between the bus register interface 80 and the PHY
72. The type of instructions and other information provided over
the MDIO line are in accordance with the requirements of the media
independent interface, which is a known protocol and standardized
within the industry.
[0042] The next two lines illustrated in FIG. 5 communicate network
conditions from the PHY 72 to the bus register interface 80. The
line COL is a signal transmitted when a collision is detected on
the physical link during the transmission of data from the PHY 72
onto the physical link. The line CRS carries a signal identifying
the presence of a carrier signal (carrier sense) on the physical
link while a message is being received: Signals on either of the
COL and CRS lines are stored within a command/status register
within the interface 80 so that the host client computer can detect
the error conditions and perform the appropriate error handling
functions.
[0043] The next line illustrated in FIG. 5 represents information
provided by the PHY 72 to the bus register interface indicative of
the validity of data received from the physical link. The line RXER
carries a signal indicative of an error during the reception of
data. The signal provided over line RXER might be a relatively
short-lived signal at least as compared to the amount of time taken
to transmit a frame of data. Consequently, it is preferred that a
latch 82 be provided along this line so that an error signal
generated by the PHY 72 is held within the latch to ensure it is
available for storage in the bus register interface 80 after
transmission of the frame of data is complete. The line TXER
performs a similar function for the transmission of data to the
physical link. The interface 80 causes a signal to be provided over
line TXER when an error is known to have occurred in the
transmission of data to the PHY 72 and the physical link beyond.
For example, the line TXER may be used to indicate that an
underflow error has occurred for the transmit buffer 88 of the
MAC.
[0044] In the illustrated embodiment, the bus register interface 80
stores data in units of 8-bit bytes. Data provided to and from the
PHY 72 are organized as four parallel bits. It is consequently
desirable to reformat the 4-bit data from the PHY into 8-bit data
within an appropriately configured register 84. Register 84 is
shown as a 4-8 bit parallel shift register. The output from
register 84 is provided to a receive buffer 86, which buffers data
received from the physical link before it is read out into the
client computer through the bus register interface 80. Data from
the client computer are provided in 8-bit bytes to a transmit
buffer 88 that buffers the data before providing it to register 90,
which formats the 8-bit data into successive 4-bit data items and
provides the data items to the PHY 72.
[0045] The receive buffer 86 and transmit buffer 88 are preferably
FIFOs (first in, first out memories) of a size appropriate to
accommodate latency within the client computer. Because servicing
of the adapter illustrated in FIG. 5 is but one of the tasks to be
performed by the preferred microprocessor of the client computer
there may be occasions where the processor is not ready to handle
the data being transmitted to or received from the physical link.
It is consequently desirable that the receive and transmit FIFOs be
of sufficient size to accommodate typical latency delays. In
practice, it is believed that an appropriate buffer size is
approximately the size of one packet or on the order of about two
thousand bytes. The appropriate size of the buffer can be chosen
according to the design of the system. Since the rate at which data
either will be received or needs to be transmitted will depend on
the particular network being serviced by the adapter, the size of
the buffer may vary significantly in different applications.
[0046] FIG. 6 illustrates an installation of the adapter 79 to
couple a client computer 100 to a network 102 through a set of
twisted wire pairs 104. The illustrated adapter 79 preferably has
the structure illustrated in FIG. 5 and, in particular, provides a
reduced hardware MAC 106 including the buffer memory and bus
register interface 80 shown in FIG. 5. The reduced hardware MAC 106
is preferably coupled to the main processor 108 of the client
computer 100 through a bus 1 10. The bus 1 10 is preferably a
thirty-two bit or larger bus operating at a sufficient speed to
move data in and out of the MAC 106 at a desired speed to minimize
latency. An appropriate bus speed is facilitated by coupling the
MAC 106 to an appropriate bus, such as a PCI bus or, in a
particularly preferred embodiment, a PC Bus, Card Bus or other bus
appropriate to coupling a PCMCIA card to a computer such as a
notebook or other portable computer. Typical implementations of
these buses are sufficiently fast as to allow the implementation of
significant portions of the MAC functionality as software running
on the processor 108. The reduced size and reduced power
consumption of preferred implementations of the adapter 79 are
particularly desirable for a portable computer implementation of
the present invention.
[0047] At least portions of the client computer's operating system
114 and portions of the software aspects 116 of the MAC are
expected to be available in memory 112 as information is
transmitted to and received from the physical link of the network
102. The illustrated memory 112 may be partially within the
processor 108 and partially within an array of DRAM or other, less
volatile memory including, for example, on a disk drive or in flash
or other solid state memory. In other implementations, the
illustrated memory 112 might be entirely inside or outside of the
processor 108. Regardless of the particular memory implementation,
the software aspects of the MAC 116 preferably communicate through
the operating system 114 within the processor 108 to the hardware
aspects of the MAC by addressing the communication registers 80
within the MAC 106.
[0048] The communication registers within the bus register
interface 80 are addressed using a preferred register address
scheme. In the embodiment illustrated in part in FIG. 7, a
three-bit address signal can be used to address a minimal register
implementation including a total of eight registers. As shown, four
data registers are provided in the four lowest addresses, with the
next two registers and addresses preferably dedicated to byte count
registers storing the number of bytes available to be read from the
receive FIFO 86. The data registers are illustrated in greater
detail in FIG. 8. Repeated reads from the data registers read data
out of the receive FIFO 86 and repeated writes to the data
registers will fill the transmit FIFO 88 with data to be sent to
the network 102. The byte-count register is illustrated in greater
detail in FIG. 9. These byte-count registers are used in read
operations to indicate whether and how many data remain to be
received from the network 102. As data are written into receive
buffer, the value stored in the byte-count register is incremented
and, as data are read from the receive buffer 86, the value stored
in the byte-count register is decremented.
[0049] It should be noted that the data received from the network
might include erroneous data or might include proper frame data. As
such, it is generally undesirable to rely on the transferred data
itself to indicate whether additional data remains to be received
from the network.
[0050] FIGS. 10 and 11 illustrate the command/status register for
write and read operations, respectively, both of which have the
same address but different definitions for the write and read
operations. Bit 0 is defined for enabling interrupts from the
adapter to the operating system of the client computer during write
operations. Bit 7 of the command/status register is set by the
client computer to reset the adapter. This bit clears automatically
with the reset of the adapter.
[0051] When the MAC software 116 is reading the adapter, the
command/status register (FIG. 11) uses bit 0 as an interrupt
pending bit to notify the host client computer of an interrupt
pending to be read. The MAC software 116 reads the command/status
register to determine what action is required by the interrupt,
whether there is data to be read (BC) or if there is an error (COL,
CRS) to be handled. Once the interrupt bit is read, it is reset.
The remaining bits 1-3 provides carrier sense, collision sense and
receive data valid signals from the PHY. When the byte count bit BC
is not zero, the MAC software begins reading the data registers to
read data out of the receive buffer 86. When either of the COL or
CRS bits is valid, the MAC software 116 is called on to handle the
error condition, either by performing a backoff algorithm (COL) or
by discarding the received, erroneous data (CRS) by flushing the
receive buffer of the erroneous data.
[0052] FIG. 12 shows definitions of bits within the MII register,
which occupies the final position within the preferred bus register
interface 80. The MII or a reduced media independent interface
(RMII) is used by the MAC software 116 to control and monitor the
operation of the PHY. Communications of MII information is
conducted serially through the MD bit of the MII register, through
the MDIO line (FIG. 5) and through the MDIO pin of the PHY.
[0053] The MAC software 116 controls much of the operation of the
adapter 79. Generally, the MAC software can be viewed as including
a lower level piece that handles interrupts and data transfer and
an upper level piece that performs more sophisticated, but less
immediate operations. Receive operations are interrupt driven and
the interrupts are handled by the lower level facility of the MAC
software 116. When the operating system 114 receives an interrupt
from the adapter 79, the operating system calls the lower level
facility of the MAC software.
[0054] The lower level facility reads the command/status register
to clear the interrupt pending bit. If there are data present, the
lower level facility begins a loop to first check the byte count in
the byte count register (FIG. 9). If the byte count is zero, the
lower level facility exits the loop and returns to a standby state
and if the byte count is non-zero, then the data are read from the
data register (FIG. 8). The loop continues until the byte count
reaches zero and then lower level facility exits the loop and
returns to a standby state until another interrupt is detected.
[0055] The data read out by the lower level facility is stored in a
buffer within the processor or in DRAM. Further processing is
necessary in accordance with the frame definition and other
processing relevant to the IEEE 802.3 or other protocol by which
the network is organized and communications proceed. For example,
processing of the data packet for an IEEE 802.3 network might
include the processor 108 detecting a frame preamble, checking the
frame length to determine if the frame is of a valid length. The
processor evaluates the frame using the FCS data to evaluate
whether the frame is valid. For example, if the FCS data is CRC
data, then a polynomial division is performed on the data using the
protocol defined check polynomial to determine if the data are
valid. All of these functions are performed by an upper level
facility of the MAC software 116. The MAC software also may perform
filtering to detect whether the address of the packet is relevant
to the host client computer.
[0056] The MAC software further tracks the network statistics in
accordance with network protocols. Statistics include the number of
frames successfully transmitted and the frequencies with which
erroneous frames are transmitted. These statistics are made
available to network management software, such as software in
accordance with a simple network management protocol.
[0057] The operation of writing data to the network is also a
cooperation between the MAC software 116 and the adapter 79. Data
to be provided to the network is formatted by the upper level
facility of the MAC software into appropriate frames and the upper
level facility calculates the FCS data, such as appropriate cyclic
redundancy check (CRC) data, and appends the FCS data to the frame.
The frames of data are passed to the lower level facility of the
MAC software, which writes the data to the data register and thus
to the transmit FIFO 88 of the adapter. The low-level portion of
the MAC software 116 monitors the adapter during to ensure that
data are transferred safely from the buffer FIFO 88 to the network.
For example, the low-level facility checks the FIFO to see if the
FIFO is full and cannot receive additional data or if the FIFO has
been completely emptied (an underflow condition). The low-level
facility further monitors the COL bit of the command/status
register to determine if there has been a collision.
[0058] The upper levels of the MAC software 116 perform the
well-known functions of an Ethernet.TM. or IEEE 802.3 network.
These functions are conventionally implemented as software or state
machines within the processor of a conventional network interface
controller. As such, it is well within the ordinary skill in this
art to implement these functions in the microprocessor or other
processor of the host client computer. These aspects of the
invention will therefore not be further described here.
[0059] Installation of an adapter in accordance with the present
invention would proceed by installing the reduced hardware MAC, for
example within a PCMCIA card that also includes PHY circuitry
appropriate to the target network. Driver software is installed to
the host computer in a manner that the MAC software is available to
be accessed by the operating system of the host computer.
[0060] Certain variations of the design discussed here will be
apparent to those of ordinary skill in the art. For example, the
reduced hardware MAC illustrated in FIG. 5 is readily implemented
as a core to be implemented in an ASIC or within the PHY. It is
possible that the reduced hardware MAC could be implemented as a
distinct core within a system on a chip, with the software aspects
of the MAC implemented within the system on a chip's processor.
Preferred aspects of the present invention provide a flexible
architecture for a network interface adapter. As such, the
inventors expect that adapters in accordance with the present
invention will find application in a variety of yet to be completed
network protocols. Moreover, the bus register interface might
include different combinations of registers to effect the reduced
hardware MAC of the present invention.
[0061] Those of ordinary skill in the art will appreciate that a
number of variations and modifications might be made to the
particular embodiments described herein without varying from the
basic teachings of the present invention. As such, the scope of the
present invention is not to be limited to the particular
embodiments described herein. Rather, the scope of the invention is
to be determined from the claims, which follow.
* * * * *