|Publication number||USRE39812 E1|
|Application number||US 11/192,461|
|Publication date||Sep 4, 2007|
|Filing date||Jul 28, 2005|
|Priority date||Nov 2, 1992|
|Publication number||11192461, 192461, US RE39812 E1, US RE39812E1, US-E1-RE39812, USRE39812 E1, USRE39812E1|
|Inventors||Brian C. Edem, Srinivas Kola|
|Original Assignee||National Semiconductor Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (4), Non-Patent Citations (2), Referenced by (2), Classifications (10), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a Continuation-In-Part of copending application Ser. No. 08/146,729, filed Nov. 1, 1993, for “Network Link Detection and Generation” which is a Continuation-In-Part of application Ser. No. 07/971,018, filed Nov. 2, 1992, abandoned, for “Network Link Endpoint Capability Detection.”
The present invention relates to local area networks, and in particular, to methods and apparatus which allows devices with multiple protocol capabilities to configure for a common protocol configuration.
The IEEE 802.3 standard defines Carrier Sense Multiple Access/Collision Detection (CSMA/CD) as its Media Access Control (MAC) protocol for Ethernet local area networks (LAN's). A station connected to an Ethernet LAN may be fixedly configured to communicate with other stations of the LAN in a single mode (e.g. 10BASE-T, 10BASE-T Full Duplex, 100BASE-TX, 100BASE-TX Full Duplex, and 100BASE-T4), or the communication mode of the station may be changeable to match the configuration of the other stations. In any case, the communication mode of a particular station must match the communication mode of another station before communication can take place between the two stations.
The proliferation of CSMA/CD-based LANs will spawn a new generation of multi-function hubs and nodes. The trend among hub vendors towards “virtual networks” wherein ports can be switched between LANs will necessitate that, without an auto-negotiation solution such as is provided by the present invention, users and information services managers will face difficult configuration management problems in the future.
The necessity for providing in a station means for automatically detecting the communication capabilities of other stations in a network has been recognized. A method and apparatus for performing such automatic detection is the subject of copending patent application Ser. No. 08/146,729, filed and assigned to National Semiconductor Corporation, the assignee of the present application. As described in the copending application, each of two connected stations communicates to the other station, via Fast Link Pulses, its respective communication capabilities. When each station has detected the other stations capabilities, the stations each automatically change to a configuration which provides the highest common denominator of capabilities.
However, the technology of LAN's is evolving such that manufacturers of LAN's will be providing stations which will are reconfigurable to support any of multiple protocols.
The present invention provides a method and apparatus which allows multi-protocol stations to automatically detect the protocols under which other stations can communicate, and for the connected stations to together select a common communication protocol.
The present invention solves the problem of configuring the growing available matrix for RJ-45 CSMA/CD compatible LANs by providing a simple, robust scheme to advertise, detect and resolve the capabilities of two stations connected with UTP. With the present invention, the upgrade occurs automatically and seamlessly from 10BASE-T to desired higher performance modes without requiring the presence of a management agent.
The present invention interoperates with all 10BASE-T compatible devices such that the installed network base will not be affected by the new generation of RJ-45 LAN technologies if NWay is adopted as a standard. NWay can be applied universally across node and hub products because the architecture is symmetric in nature and does not establish a master/slave hierarchy.
A low gate count and simple architecture make the present invention both easy to understand and cost effective. The cost and complexity are realistically bounded by providing a rich enough code space to meet the current projected higher performance mode and figure LAN technology needs. Although this protocol addresses the problem for CSMA/CD protocol LANs, it should be understood that other technologies can benefit from the use of this communication mechanism.
The present invention also provides a powerful management tool which allows:
1. Manual override of auto-negotiation;
2. Renegotiation on management intervention;
3. Remote Fault Sensing;
4. Parallel Advertisement of capabilities; and
5. Storage of remote capabilities.
The negotiation time is bounded for good links sufficiently to ensure that popular network protocol connections do not time out. Even in a noise UTP environment, the handshaking and redundant pattern comparison ensures robust operation.
A better understanding of the features and advantages of the invention will be obtained by reference to the following detailed description and accompanying drawings which set forth an illustrative embodiment in which the principles of the invention are utilized.
FIG. 23 and 23A-23C are a flowchart of a protocol arbitration resolution algorithm (PARA) in accordance with the present invention.
In the CSMA/CD protocol, as defined in IEEE 802.3, a 10BASE-T Normal Link Pulse (NLP) is applied on a network link segment to assist stations connected to the segment in determining the segment's integrity. Such a Link Pulse scheme is known in the art. As shown in
NLP's are used to control a station's entry to and exit from a Link Loss State. That is, a station enters Link_Loss_State when no NLP's are received by the station for more than link loss timer time. Link_loss_timer time typically ranges from 50 ms to 150 ms. A station exits Link_Loss_State when it detects a number of NLP's, typically two to ten.
The present invention substitutes a Fast Link Pulse Burst (FLP) for a single 10BASE-T NLP. An FLP burst, as shown in
Individual pulses in an FLP burst have the same 100 ns width as NLP's, with both clock and data pulses having the same width. Clock pulses are spaced apart by 125 μsecs. Data bits, if present, are centered between a pair of adjacent clock pulses.
In accordance with an embodiment of the present invention, a method and apparatus are provided which allows a station at one end of a LAN link segment to advertise the protocols, and modes within those protocols, in which it can operate, and to likewise detect the protocols and modes in which a remote station at the other end of the link segment can operate (or at least those protocols and modes that the station at the other end of the link segment is advertising). The communication of modes takes place via the data communicated in a 16-bit “link code word” encoded in FLP bursts, as described above.
The present invention further provides a method and apparatus for using the protocol and capability information advertised in link code words communicated between two stations on a LAN link, of which one or both of the stations are multi-protocol/multi-capable, for each station to automatically detect the other stations advertised protocol/configuration capabilities and further automatically agree on a protocol/configuration under which the stations can interoperate. Upon completion of the auto-agreement, each station configures its link, in accordance with the results of the auto-agreement, for interoperation with the other link.
Finally, a Technology Dependent Link Integrity Test State Machine (or Machines) may co-exist with the auto-agreement scheme. For example, the Technology Dependent Link Integrity Test State Machine may be a 100BASE-TX Link Test-Fail State Machine which controls the reset logic for the Convergence Sub-Layer of the 100BASE-TX standard. When in Link_Fail, the Idle Line State is continuously sent.
Each of the auto-agreement state machines are now described in detail.
The arbitration state machine is shown in FIG. 10. The purpose of this state machine is to control the FLP transmit state machine and determine if a remote station has its technology abilities, either through FLPs or through a technology-dependent signalling scheme. At power on, or upon the assertion of technology-dependent link-fail condition, this state machine moves into Capability Detect state to allow the FLP transmit state machine to start sending FLP bursts containing the local station's technology abilities to the far-end station. The results of FLP receive state machine are to be used to determine technology ability match.
A technology ability match occurs when this state machine is in Ability Detect state, and three consecutive FLP bursts contain the same pattern ignoring the acknowledge bit. Once a capability match occurs, the FLP arbitration state machine moves into Acknowledge Detect state.
In Acknowledge Detect state, it transmits the same data pattern with the acknowledge bit set, to indicate to the far-end station that it has captured its technology abilities.
In the case of unexplained loss of FLP bursts coming from the far-end station, under which the FLP receive state machine goes to IDLE state, this state machine goes back to Ability Detect state to restart negotiation from scratch. While remaining in the Acknowledge Detect state, it waits to receive acknowledge_match from the far-end station. Acknowledge_match occurs when two out of three consecutive FLP bursts received have the acknowledge bit set. In order to guarantee that the far-end station has completed acknowledge_match, a minimum of 4-6 acknowledge patterns are transmitted. Subsequently, it transitions to FLP_link_good state. After this point, the technology-specific link signalling scheme takes over, if needed for the specific technology (e.g., 100BASE-TX, 100BASE-T4). If this technology-specific link signalling scheme does not successfully complete in a technology determined period of time, then technology_link_fail is asserted, and the Arbitration State Machine starts renegotiation.
If the far-end station does not detect and transmit FLP bursts, and instead sends a technology-dependent signal (e.g., 10BASE-T link pulses, or 100BASE-TX Idle Line State transmissions), the technology-dependent_link_good signal thereby halting FLP burst transmits.
In addition, upon successful establishment of technology-dependent link, the technology-dependent link integrity test state machine goes to link fail and the arbitration state machine goes back to the Ability Detect state.
A higher-level management agent can start a renegotiation of technology abilities with the far-end station by asserting the renegotiate signal. The state machine moves into Transmit Disable state to wait for the far-end station to go into a link-fail state and then starts FLP burst transmissions.
The Transmit state machine is shown in FIG. 11. The purpose of this state machine is to transmit a sequence of 33 fast link-pulses to the far-end station, when instructed by the arbitration state machine. The idle stats is entered at power-on and this state machine transitions to this state if the arbitration state machine is in the FLP_link_good state.
The transmit_link_pulse_timer (14+/− 8 msec) is used to separate two consecutive fast link-pulse bursts transmitted.
If the arbitration state machine is in the process of sending the capability to the far-end station, then the transmit state machine enters into the Transmit_Capability state. If acknowledgement is to be sent, then the state machine enters Transmit_Acknowledge state to initialize ack_cnt (416), and then it moves to Transmit_Capability state. In this state, the bit_cnt is started to keep track of the 16 bits of data to be transmitted.
The state machine then alternates between the Transmit_Clock_Bit state and Transmit_Data_Bit state to transmit clock bit and to transmit data bit (if the data bit to be transmitted is logic one) respectively. The interval_timer (62.5+/− 15 μs) is used to separate a clock bet and a data bit.
Once all 16 bits of data are transmitted, if the state machine is in the midst of sending acknowledgement patterns then the state machine moves to the Transmit_Count_Ack state. However, after 16 bits of data transmission, if the state machine is not in the midst of sending acknowledgement patterns, then the state machine moves to idle directly.
The state machine returns to idle after a 16-bit capability has been transmitted. In the Transmit_Count_Ack state, if 4-6 acknowledgement patterns have been sent, it goes back to idle, otherwise it waits for 14+/8 msec and moves to Transmit_Capability state to send the next burst of 16-bit acknowledgement pattern.
The Receive state machine is shown in FIG. 12. This state machine is used to identify that the link pulses received are fast-link-pulse bursts, and if so, to store the embedded 16-bit data in a shift-register. A burst of thirty-three link pulses separated by 62.5 uS conveys the capability of the far-end station. A link pulse is considered part of a FLP burst if it occurs within a certain time of the previous link-pulse. It is detected by the use of the two timers: FLP_test_min_timer and FLP_test_max_timer. The purpose of the FLP_test_min_timer is to mask out any noise or ringing effects on the line. The FLP_test_max_timer is used to determine whether the next link-pulse received is within a window to be of fast-link-pulse category.
At this point, both the FLP_test_min_timer and FLP_test_max _timer are started. The FLP_test_min_timer has a value of 5 to 25 μs. The FLP_test_max_timer has a value of 165 to 185 μs. If another link-pulse is received in the window of time such that the FLP_test_min_timer has timed out, but the FLP_test_max_timer has not completed, the state machine makes a transition to Link_Pulse_Count state.
In the Link_Pulse_Count state, FLP_count is incremented. If consecutive number of fast link pulses are received, i.e., the FLP_count equals FLP_count_max (6 to 8), then the far-end station is considered FLP capable, and all Technology-Specific link integrity test state machines (e.g., 10BASE-T link_integrity_test_state machine) are forced into Freeze state.
The Receive state machine then goes into a FLP_Pass state to ignore the rest of the incoming fast link-pulse burst. This is accomplished by waiting for the idle interval after the last link pulse of the current burst, i.e., an idle period of 165 to 185 μs has been observed. At this time, the state machine transitions into FLP-Capture state, and is ready to capture a complete burst of fast link pulses from the far-end station.
In the FLP-capture state, the link_test_max_timer (25-150 msec.) is started to insure that link pulse bursts are received within this window. When this timer expires, the state machine moves to the idle state to indicate the absence of FLP bursts.
In the FLP_Clock state, the data_detect_min_timer (15 to 47 μs) and data_detect_max_timer (78 to 110 μs) are started. If a link-pulse is received when the data_detect_max_timer has completed and data_detect_max_timer is not, it is interpreted as a data-bit of logic one. On the other hand, if the next clock_bit arrives, as indicated by data_detect_max_timer_done, then the embedded data bit is assumed to be zero. In the FLP_Clock state, the link_test_min_timer (5 to 7 msec) is also started; this timer is used to separate two consecutive fast-link pulse bursts.
The only technology-specific state machine that is required is the 10BASE-T link integrity test state machine. This state machine is required for compatibility with existing 10BASE-T nodes.
Two modifications must be made to the original 10BASE-T link integrity test state machine. The first modification is that the state machine must power-up in the Link_Test_Fail_Reset state.
The second modification needed to the original 10BASE-T link integrity test state machine is to add another state called Freeze 10BASE-T. This state is entered upon the recognition of the far-end station's FLP capability and remains in this state during the negotiation phase. After the NWay negotiation is successfully completed, and the highest common denominator is determined to be 10BASE-T, the 10BASE-T Link integrity test state machine moves into the Link_test_pass_state. The state machine remains in the Freeze—10BASE-T state if the mode of communication selected is not 10BASE-T.
Additional technology-specific state machines can be added to implement technology-specific link testing.
The Peripheral Logic Block/Management Interface contains the following:
The next two sections provide the Technology Bit Field Assignments and Priority Resolution Table for the Message Selector.
Two nodes can have multiple abilities in common. In accordance with one embodiment of the present invention a prioritization scheme exists to ensure that the highest common denominator ability is chosen. For example, the following priority may be used (from highest to lowest priority):
The rationale for this hierarchy is straightforward. 10BASE-T is the lowest common denominator and therefore that the lowest priority. Full Duplex solutions are always higher in priority than their Half Duplex counterparts. 100BASE-T4 is ahead of 100BASE-TX because 100BASE-T4 runs across a broader spectrum of copper cabling.
The following operational examples highlight the steps that are taken when attempting to establish a network connection between two nodes. Only three general cases are necessary to demonstrate all the different combinations of operation that will ever occur. The three cases are any auto-detect capable device to 10BASE-T, any auto-detect capable device to another auto-detect capable device, and any auto-detect capable device to a non-auto-detect capable device. All these examples fall into one of these three cases.
This example highlights the invention's 100% backward compatibility with 10BASE-T. The 10BASE-T station in this example begins in the Link_Fail state and is transmitting Normal Link Pulses. The auto-detect capable station supports an arbitrary 100 Mb/sll 10BASE-T capable or 100 Mb/s only capable node. Initially the NWay capable station comes up in the Link_Fail state and transmits FLP bursts to advertise its technology ability.
The normal 10BASE-T node receives the FLP's and remains in Link_Fail state because the timing of the FLP's do not allow a normal 10BASE-T station to misinterpret FLPs as Normal Link Pulses. To understand this, there are two main cases to consider.
In the first case, shown in
In the second case, shown in
No scenario causes the good link pulse counter to count higher than 1, which is less than the 2 to 10 required good link pulse specification, so the 10BASE-T only station will never mistakenly enter the Link_Pass state.
The 10BASE-T Link Integrity Test state machine is a required part of the auto-detect, regardless of 10BASE-T data transmit/receive capability. The NWay capable 100/10 node recognizes the NLPs being sent by the normal 10BASE-T node, switches over to 10BASE-T operation, and sends NLPs. The two stations then both communicate in 10BASE-T mode. If the auto-detect capable station is 100 Mb/s only, then it must still recognize 10BASE-T NLPs being sent. In response to NLPs, FLP burst transmission is halted and no NLPs are transmitted by the 100 Mb/s only station.
This example highlights auto-detect's 100% backward compatibility with 10BASE-T. The 10BASE-T station in this example begins in Link_Pass state and is transmitting Normal Link. Pulses or normal 10BASE-T traffic. The auto-detect capable station is an arbitrary 100 Mb/sll 10BASE-T capable or 100 Mb/s capable station. Initially the auto-detect capable station comes up in Link_Fail state and transmits Fast Link Pulses to advertise its technology abilities.
The Normal 10BASE-T node receives the Fast Link Pulses and always remains in Link_Pass state because the link_loss_timer does not expire.
The auto-detect capable node's 10BASE-T specific link integrity test state machine identifies the link as being good due to receipt of 10BASE-T traffic or normal link pulses.
If no 10BASE-T data transmit/receive capability exists, then FLP bursts are halted, and no NLPs are transmitted by the auto-detect capable station, otherwise NLPs are transmitted and the link is established.
If some 10BASE-T Link Integrity Test state machine implementations transition from Link Pass state to Link Fail state due to FLPs, we are back to the previous auto-detect capable < - - - > 10BASE-T in Link Fail state example.
This example highlights auto-detect capable station to auto-detect capable station auto-negotiation when multiple common technology abilities are supported by both stations. In this case it is assumed that both stations support Full Duplex 100BASE-TX, 100BASE-T4 and 10BASE-T.
Both nodes come up in Link_Fail state and transmit FLP bursts encoded with their technology abilities. Each node learns its partners' technology abilities. In this case each node learns that it has Full Duplex 100BASE-TX, 100BASE-T4 and 10BASE-T in common.
The built-in priority resolution table in each node indicates that full Duplex 100BASE-TX is the highest common denominator technology ability, so both nodes switch over to this technology ability and begin transmission.
This example highlights a connection between two 100BASE-T4 auto-detect capable stations. In this case it is assumed that each station supports 100BASE-T4 as its highest common denominator technology ability.
At this time the 100BASE-T4 technology ability can proceed through its own ability-specific link integrity test. This test can then do a specific test to insure that the two pairs of cable not used for auto-detect signalling are intact. In one aspect of the present invention, redundant transmissions are relied upon to ensure that the transmitted 16-bit pattern is correctly received by the remote station. That is, the remote station must receive an identical 16-bit pattern three times in a row before the acknowledgement pattern is transmitted. The acknowledgement pattern is the same 16-bit word with the acknowledge bit set. The reception of 2 out of 3 acknowledgement patterns completes the handshake. Thus, it is assured that the correct configuration information is exchanged.
The present invention provides an efficient mechanism for point-to-point communication that is efficient in terms of the amount of logic gates required. Regardless of the number of technology abilities supported, a total of 3 state machines, 6 timers, 3 counters, and some peripheral logic is needed. An analysis of the architecture shows that the design will require approximately 400 cells to implement. In addition, the present invention has been architected as a simple mechanism which allows easy comprehension by silicon suppliers.
It should be noted that this level of complexity remains fixed as new technologies emerge and need to coexist with the growing installed based of RJ-45 LANs. Alternative schemes that require separate and complex state machines for each pattern will be more complex and expensive to implement, especially as new LAN options are added.
The present invention can maintain compatibility when new options are added. As new CSMA/CD-compatible LAN technologies enter the market a reserved bit may be assigned to each technology. The new technology will be inserted into an updated priority table, as detailed in the Data Bit Field description. The relative hierarchy of the existing technologies need not change, thus providing backward compatibility. In preferred embodiments, the reserved bits are forced to zeros. This guarantees that devices implemented using the current priority table will be forwarded compatible with future devices using an updated priority table.
The present invention allows a station to advertise its complete set of CSMA/CD capabilities in one burst, a feature referred to as its “parallel advertising” capability. Advertising the complete set of capabilities in one pass allows a station to immediately discover all capabilities in one pass and thus immediately discover all capabilities of the remote station. These capabilities can then be saved for passage to a management agent.
In accordance with a further embodiment, although a pair of stations default to their highest common denominator capability, other modes of interoperation may exist. For a variety of reasons, the network management agent may wish to reconfigure the link.
In accordance with a further embodiment of the present invention, a Protocol Arbitration and Resolution Algorithm (PARA) algorithm allows for arbitration and resolution for capabilities within a given protocol and, further, between the protocols themselves until a common protocol and ability are identified. If no common ability is identified, then the link will not be established.
The PARA's shown in flow-chart form in FIG. 23.
On the reception of 3 consecutive and identical link code words (label A) all the four timers in the PARA are reset (label A). Protocol Identifiers Do Not Match (label B)
If the Protocol Identifiers (PID's) do no match (label B) then it will be verified if the local node can support other protocols (label 2). If another PID can be supported (label 2) then it will be set (label 3) in the 16 bit link code word or else no change will be made to the PID. Control returns back to the beginning of the algorithm (label A). If the PID of the local node is not the lowest in the negotiating pair, then the local node will wait until the timer 1 (Slave_code_word_Change_timer) is done (label 4) for the other node to change its PID and advertise. If no change is observed in the link code word, then it knows that the other node is incapable of protocol arbitration and becomes the arbitration master (i.e., jumps to label 2). If there is a change in the link code word the local node knows that the link partner changed the PID and will have to renegotiate (label 5) and hence will return control back to beginning of the algorithm (label A). (Note: this accommodates for devices with multiple protocols with no protocol arbitration function and also for devices with multiple protocols but with the arbitration function disabled manually or by software).
If the protocol identifier's (PID) match (label B) then the capability fields of local node and its link partner are compared. If there are common capabilities then the highest common denominator amongst the common capabilities is chosen as specified in NYway (label 6).
If the PID's match and there are no common abilities then both the link partners will evaluate to see who has the lowest ability field (label C).
If the local node has the lowest ability, then it will be verified if it has other abilities (label and if found true the new capability or abilities can be set in the 16 bit link code word and return to beginning of algorithm. On the other hand if no other abilities are present then the local node will wait until timer 2 (Master_ability_change_timer) is done (label 8) giving time for the remote end (label 11) to advertise ability bits which it might not have previously advertised. If there is no change in the link code word, then the given PID is exhausted and hence a new PID will be set and negotiation will start at the beginning of the algorithm (label A).
In this case the remote link partner will be the negotiation master. Local node (label 11) will wait until timer 1 (Slave_code_word_change_timer) is done, to let the master change the ability bits and advertise the link code word. If there is a change in the bit field before this timer is done, then the master is still in control and negotiating. If no change is seen when the timer is done then the local node takes initiative/control from the remote master temporarily (label 12). The local node sets additional ability bits in the link code word if it has any previously unadvertised abilities it might want to advertise now. Control will be transferred to the beginning of the algorithm. On the other hand if it does not have any new ability bits to advertise then it will wait until the timer 3 (Slave_Negotiation_timer) is done (label 13). Before the timer 3 is done the remote master waits until its timer 2 (Master_ability_change_timer) is done, knows that all the feasible ability matching options are exhausted in both the nodes and hence tries to change its PID as described in the section above (label 9) before losing the Negotiation master privilege. If the remote master has no PID arbitration capability, it will yield mastership to the local node. When the timer 3 (Slave_Negotiation_timer) is done at the local end the local node takes up negotiation control and advertises a new PID if it is capable.
11 × 16 = 176 ms
22 × 16 = 352 ms
33 × 16 = 528 ms
36 × 16 = 576 ms.
The following is an example for the application of “Protocol Arbitration and Resolution Algorithm” described above.
The following example shows how two systems can find a common protocol and a common set of capabilities. Assume a link with two endpoints, called X and Y. These Endpoints have the following protocols and abilities:
Each protocol can represent a different encoding scheme, transmission rate, or access mechanism. Each ability can represent different configurations, modes of operation, or optional abilities.
Assume that X initially advertises Protocol 1, abilities A, B, C, (represented by the notation 1-ABC), and Y advertises 2-DEF. Since these protocols are different, they can not interoperate (label B). Since X is advertising a protocol with the smallest encoded value, it is X's responsibility to select a new protocol (X:label 2, Y:label 4). Since X can not support Protocol 2, it will advertise Protocol 3. When X advertises Protocol 3, it decides that it would prefer to use Abilities G or H, and will initially only advertise 3-GH (X:label 4; Y:label 4).
Endpoint Y, now being the endpoint with the lowest encoded value for protocol, has two options—if it were capable of communicating using protocol 1, it could advertise it, forcing X to return to protocol 1, or, as in this case, it will advertise protocol 3-IJK (Y is willing to use any of its capabilities).
The protocols now match, but the intersection of the two sets of abilities is empty. If we assume that when the abilities (advertised by Y) are encoded that 3-IJK will represent a lower value than 3-GH (advertised by X), it will be endpoint Y's responsibility to change capabilities. Endpoint Y has no other options, and makes no changes (X:label 11; Y:label 8).
Endpoint X reaches a timeout 1 (i.e., when timer 1 or slave_code_word_change_timer is done), realizing that Y is unable to proceed, so X will change its advertisement (X:label 12) to include the additional ability of I (control transfer afterwards to X:label A; Y:label A).
X's advertisement is now 3-GHI, which contains the common element 1 with Y's advertisement of 3-IJK, allowing the completion of arbitration (X:label 6; Y:label 6).
If endpoint X had no other capabilities (i.e., X has only abilities 3-GH), it too would make no changes to its capabilities, allowing timeout2 (i.e., timer 2 or Master_ability_change_timer is done) to be reached by endpoint Y. Upon reaching this timeout2, Y has the opportunity to change to another protocol. In this case, there are no other protocols to be tried, and Y will make no changes to its advertisement.
A third timeout (timer 3 or Slave_negotiation_timer is done) is reached by X, indicating that Y was unable to select a new protocol, giving X the opportunity to select a new protocol. This mechanism allows X to initially ignore one of Y's protocol until X determined that there were no other common protocols.
This is a second method to achieve the same results but the arbitration is not done until all the protocol capabilities and technology abilities are exchanged between the two link partners.
The bits A7-A0 in the link code word in
The link partners X and Y using this scheme for negotiation is described below:
Node X sends the link code word with selector/PID set to 00000, ack=0, series end bit=0, and a four bit binary encoded value of the number of protocols supported (to node Y).
If Node Y is not multiprotocol capable, then it will keep sending the link code word with the only PID it supports or can handle. On sensing this, Node X will identify that Node Y is not multiprotocol capable and will advertise with the same PID as node Y, if X can handle that particular protocol. If there are common technology abilities between Node X and Node Y, then a link will be established. If Node X is not capable of the protocol advertised by Node Y, then a link cannot be established.
If Node Y is multiprotocol capable, then it also responds with PID=00000, the number of protocols it can support, and other information about itself through the link code word. As specified in the earlier portion of this description, the ack bit is set on the reception of 4-6 consecutive and consistent link code words. Both nodes begin to transmit their capabilities starting with the lowest PID value they support until all of the protocol capabilities, along with the technology ability bits, are transmitted. It is possible that one of the nodes with fewer protocol abilities can finish its transmission ahead of its link partner. In such a case, it will begin transmitting PID=00000 with the series end bit set to a “1”. The node that is lagging will soon catch up and will also begin transmitting PID=00000 and series end bit=1. Since both modes have advertised the end of transmission, this signifies the end of the capability exchange between the two nodes. (If all the protocol information received matches the binary encoded number of protocols identified in the first code word transmission, then there is no loss of information. Otherwise, the transmission will begin again.) From the beginning to the end of this negotiation with PID=00000, NWay will not configure to the highest common denominator in any protocol. That is, it will be in a partial freeze state.
Now that all of the protocol abilities are known, the media access units (MAU's) in both the nodes will look up a common prioritization table to identify the highest common denominator common to both nodes.
The prioritization table gives the order in which a node prefers to configure based on the protocols and the abilities within that protocol. For example:
1) Protocol A, ability 1
2) Protocol B, ability 5
3) Protocol A, ability 2
4) Protocol B, ability 2
The abilities within a protocol can be mixed in the prioritization table with the abilities from other protocols. This gives tremendous flexibility beyond simply prioritizing the protocols. This also allows network management to mask out certain abilities such that the chosen ability is the most optimal for the kind of service (like data transmission, reliable and isochronous for video/multimedia or lowest cost service etc.) required.
Thus, it can be seen that the protocol resolution and arbitration algorithm allows nodes supporting multiple protocols (e.g as shown in
It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. In particular, the invention has application beyond local area networks; it has application wherever there is communication between two stations.
It is intended that the following claims define the scope of the invention and that methods and apparatus within the scope of these claims and their equivalents be covered thereby.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4953210||Feb 22, 1990||Aug 28, 1990||Hayes Microcomputer Products, Inc.||Feature negotiation protocol for a synchronous modem|
|US5311114||Oct 27, 1992||May 10, 1994||Seeq Technology, Incorporated||Apparatus and method for full-duplex ethernet communications|
|US5432775||Dec 3, 1993||Jul 11, 1995||Advanced Micro Devices, Inc.||Auto negotiation system for a communications network|
|US5617418||Nov 1, 1993||Apr 1, 1997||National Semiconductor Corporation||Network link detection and generation|
|1||"Draft Supplement to IEEE 802.3 Standard", pp. 1-31 (Section 28 et seq.), Jun. 16, 1994, (Appendix 1 filed with the original application.).|
|2||"IEEE Link Task Force Autodetect Specification for NWay Autodetect Version 1.0", Six introductory pages plus pp. 1-60, Apr. 10, 1994. (Appendix 2 filed with the original application.).|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8472340 *||Sep 2, 2011||Jun 25, 2013||Marvell International Ltd.||Network interface with autonegotiation and cable length measurement|
|US8797909||Jun 25, 2013||Aug 5, 2014||Marvell International Ltd.||Network interface with autonegotiation and cable length measurement|
|U.S. Classification||370/447, 370/462, 370/466|
|Cooperative Classification||H04L69/18, H04L69/24, H04L12/413|
|European Classification||H04L12/413, H04L29/06P, H04L29/06K|
|Jun 17, 2008||FPAY||Fee payment|
Year of fee payment: 12
|Jun 23, 2008||REMI||Maintenance fee reminder mailed|