Arbitration system for a shared DMA logic on a network adapter with a large number of competing priority requests having predicted latency field

• US 5,794,073 A
• Filed: 01/29/1997
• Issued: 08/11/1998
• Est. Priority Date: 11/07/1994
• Status: Expired due to Term

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Digital Equipment Corporation (HP Inc.)
• Inventors: Ramakrishnan, Kadangode K., Ben-Nun, Michael, Roman, Peter John
• Primary Examiner(s): Asta, Frank J.
• Assistant Examiner(s): Luu, Le Hien
• Application Number:
  US08/789,814
• Time in Patent Office: 559 Days
• Field of Search: 395/732, 395/864, 395/866, 395/860, 395/859, 395/842, 395/848, 370/429, 370/399, 370/474, 370/462, 370/235, 370/232
• US Class Current: 710/40
• CPC Class Codes: G06F 13/36 for access to common bus or bus system
• CPC Class Codes: G06F 13/364 using independent requests or grants, e.g. using separated request and grant lines

Abstract

In accordance with principles of the invention, there is provided an arbitration system for multiple requesters of a shared data transfer resource, such as a system bus or a peripheral bus. The disclosed system arbitrates among multiple classes of requesters which are divided into multiple levels of a request hierarchy. In the example embodiment, the multiple requesters include logic for processing received data from the network, logic for processing data to be transmitted onto the network, logic for moving transmit and receive descriptors between the host memory and the adapter, logic for reporting status from the adapter to the host, and logic for generating an error and maintenance status update from the adapter to the host. The new system ensures fairness between transmit and receive processes, that FIFOs associated with transmit queues are not underrun, and further that notification of non-error and maintenance status changes are processed with minimal latency.

Claims

1. A system for arbitrating between multiple requests for a shared resource, wherein said shared resource is a Direct Memory Access logic within a network adapter, and said Direct Memory Access logic is coupled with a data bus in a host computer system, comprising:

first determining means for determining a priority of each request in a first request class, said first request class consisting of transmit data requests to use said Direct Memory Access logic to read transmit data from a host memory in said host computer system, said transmit data to be transmitted onto a network coupled with said network adapter;

second determining means for determining a high or a low priority of each request in a second request class, said second request class consisting of receive data requests to use said Direct Memory Access logic to write receive data from said network adapter into said host memory;

first selecting means, responsive to said first determining means, for selecting a first selected request equal to one of said requests in said first request class having a highest priority;

second selecting means, responsive to said second determining means, for selecting a second selected request equal to one of said requests in said second request class having a high priority;

weighted arbitration means, for choosing between said first selected request and said second selected request based on a 1 of N round robin arbitration, said second selected request selected once every N times said shared resource is available, and N is a predetermined integer;

means for generating a priority vector associated with each transmit data request, higher values of said priority vector indicating higher priority, said priority vector including a predicted latency field;

means for calculating a value `CL` for each transmit data request, where CL is equal to an estimated time period since the last transmission on the virtual circuit for each specific transmit data request estimated to have elapsed at the time said transmit data is actually read from said host memory into said network adapter; and

means for writing said value CL into the priority vector for each said transmit data request.

2. The system as in claim 1, further comprising:

one or more Receive Queues for indicating reassembled packets received into said network adapter from said network and stored in a Reassembly Memory;

each said receive data request is associated with one of said Receive Queues; and

wherein said second determining means determines that a given receive data request is high priority if the occupancy level of said Reassembly Memory is greater than a predetermined amount.

3. A method for arbitrating between multiple requests for a shared resource, comprising the steps of:

determining a priority of each request in a first request class, said first request class consisting of transmit data requests to use said shared resource, by generating a priority vector associated with each transmit data request, higher values of said priority vector indicating higher priority, said priority vector including a predicted latency field a contents of said predicted latency field equal to a time period estimated to have elapsed since a last previous transmission on a virtual circuit associated with said each transmit data request at a subsequent time when transmit data for said data request is read from a host memory into a network adapter, and by comparing said priority vector of each said request in said first request class;

determining a high or a low priority of each request in a second request class, said second request class consisting of receive data requests to use said shared resource;

selecting, responsive to said first determining means, a first selected request equal to one of said requests in said first request class having a highest priority;

selecting, responsive to said second determining means, a second selected request equal to one of said requests in said second request class having a high priority; and

choosing between said first selected request and said second selected request based on a 1 of N round robin arbitration, said second selected request selected once every N times said shared resource is available, and N is a predetermined integer.