Method and apparatus for efficient interrupt event notification for a scalable input/output device

• US 8,510,491 B1
• Filed: 04/05/2005
• Issued: 08/13/2013
• Est. Priority Date: 04/05/2005
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Oracle America, Inc. (Oracle Corporation)
• Inventors: Hendel, Ariel, Gajjar, Yatin, Lin, May, Wong, Michael, Pun, Rahoul
• Primary Examiner(s): Zaman, Faisal M
• Application Number:
  US11/098,727
• Time in Patent Office: 3,052 Days
• Field of Search: 710/50, 710260-269, 709213-216, 711/129, 711/148, 711/173
• US Class Current: 710/260
• CPC Class Codes: G06F 15/17331 Distributed shared memory [DSM], e.g. remote direct memory access [RDMA]

Abstract

A method and apparatus for efficient interrupt event notification for a scalable input/output device in a network system. A network interface unit is operably connected to a plurality of processing entities and associated memory units. At least one status register in the network interface unit contains information relating to a process to be performed by at least one processing entity communicated to the processing entity by an interrupt event notification. Shared memory space comprises a mailbox storage register operable to store an image of the interrupt information stored in the status register of the network interface unit. A processing entity can directly access the process information stored in the mailbox status register thereby reducing system latency associated with reading information in the status register. Updated process status information in the network interface status register may be read by the processing entity on an interleaved basis while executing a process.

Claims

1. A network system, comprising:

a plurality of processing entities and associated memory entities, wherein predetermined subsets of the plurality of processing entities define asymmetrical data processing partitions operable to process data transported by the network system;

a network interface unit operably connected to at least one of the asymmetrical data processing partitions, wherein the network interface unit comprises one or more buffers storing data received from a network to which the network interface unit is coupled; and

at least one status register in the network interface unit, the status register containing information indicating a task to be performed by at least one processing entity in the at least one of the asymmetrical data processing partitions, wherein the status register is separate from the one or more buffers, and wherein said information indicating said task is separated from the data stored in the one or more buffers of the network interface unit;

wherein at least one of the memory entities associated with the one of the asymmetrical data processing partitions comprises a mailbox storage register that stores an image of the information indicating the task to be performed stored in the status register of the network interface unit such that information indicating the task to be performed is simultaneously stored in both of said mailbox storage register and said status register; and

wherein the at least one processing entity is configured to;

initially read the information indicating the task to be performed in the mailbox storage register upon detection of an interrupt pertaining to the network interface unit;

execute the task indicated by the information stored in the mailbox storage register; and

during the execution of the task, read the status register in the network interface unit to read information indicating one or more additional tasks to be performed by the at least one processing entity, wherein to read said information indicating one or more additional tasks to be performed, the at least one processing entity reads the status register in the network interface unit and not from the memory entities.

2. The network system of claim 1 wherein:

the at least one processing entity is configured to read the information indicating the one or more additional tasks to be performed in the status register on an interleaved basis while executing the task indicated by the information contained in the mailbox storage register.

3. The network system of claim 2 wherein the network interface unit further comprises:

a memory access module to control the transfer of data, the memory access module including a plurality of parallel memory access channels; and

a packet classifier, the packet classifier providing an association between packets and the plurality of asymmetrical data processing partitions via the plurality of memory access channels.

4. The network system of claim 3 wherein:

the packet classifier provides a hierarchical classification of packets processed by the asymmetrical data processing partitions.

5. The network system of claim 4 wherein:

packets are moved within the network interface unit independently and in parallel to optimize packet movement efficiency.

6. The system of claim 1, wherein the at least one processing entity is configured to read the information indicating the task to be performed in the status register in the network interface unit subsequent to reading the information indicating the task to be performed in the mailbox storage register.

7. The network system of claim 1, wherein the at least one processing entity is configured to, prior to completing the execution of the task, complete the reading of said information indicating the one or more additional tasks.

8. A method of interrupt event notification in a network data processing system, the method comprising:

providing a plurality of processing entities;

providing a plurality of memory units defining a shared memory;

asymmetrically associating predetermined subsets of the plurality of processing entities with predetermined subsets of the plurality of memory units thereby defining a plurality of asymmetrical data processing partitions operable to process data transported by the network system;

operably connecting a network interface unit to at least one of the asymmetrical data processing partitions;

storing in one or more buffers of the network interface unit, data received from a network to which the network interface unit is coupled;

storing information indicating a task to be performed by at least one processing entity in the at least one of the asymmetrical data processing partitions in a first status register in the network interface unit, wherein the first status register is separate from the one or more buffers, and wherein said information indicating said task is separate from the data stored in the one or more buffers of the network interface unit;

storing an image of the information indicating the task to be performed from the first status register in a mailbox storage register in at least one of the memory units in the at least one of the asymmetrical data processing partitions such that information indicating the task to be performed is simultaneously stored in both of said mailbox storage register and said status register; and

the at least one processing entity;

initially reading the information indicating the task to be performed in the mailbox storage register upon detection of an interrupt pertaining to the network interface unit;

executing the task indicated by the information stored in the mailbox storage register;

during the execution of the task, reading the status register in the network interface unit to read information indicating one or more additional tasks to be performed by the at least one processing entity, wherein reading said information indicating one or more additional tasks to be performed includes reading the status register in the network interface unit and not from the memory units.

9. The method of claim 8, further comprising:

the at least one processing entity reading the information indicating the one or more additional tasks to be performed in the first status register on an interleaved basis while executing the task indicated by the information contained in the mailbox storage register.

10. The method of claim 9 wherein the network interface unit further comprises:

a memory access module to control the transfer of data, the memory access module including a plurality of parallel memory access channels; and

a packet classifier, the packet classifier providing an association between packets and the plurality of asymmetrical data processing partitions via the plurality of memory access channels.

11. The method of claim 10 wherein:

the packet classifier provides a hierarchical classification of packets processed by the asymmetrical data processing partitions.

12. The method of claim 11 wherein:

packets are moved within the network interface unit independently and in parallel to optimize packet movement efficiency.

13. The method of claim 8, further comprising the at least one processing entity reading the information indicating the task to be performed from the first status register in the network interface unit subsequent to reading the information indicating the task to be performed in the mailbox storage register.

14. The method of claim 8, further comprising the at least one processing entity completing the reading of said information indicating the one or more additional tasks prior to completing the execution of the task.

15. An apparatus for asymmetrically processing data in a network system, comprising:

means for providing a plurality of processing entities;

means for providing a plurality of memory units defining a shared memory;

means for providing a plurality of asymmetrical data processing partitions operable to process data transported by the network system, wherein the plurality of asymmetrical data processing partitions comprise predetermined subsets of the plurality of processing entities and predetermined subsets of the plurality of memory units;

means for providing a network interface unit operably connected to at least one of the asymmetrical data processing partitions, wherein the network interface unit comprises one or more buffers storing data received from a network to which the network interface unit is coupled, wherein the network interface unit comprises a status register containing information indicating a task to be performed by at least one processing entity in the one of the asymmetrical data processing partitions, wherein the status register is separate from the one or more buffers, and wherein said information indicating said task is separate from the data stored in the one or more buffers of the network interface unit;

wherein at least one of the memory units in one of the asymmetrical data processing partitions comprises a mailbox storage register that stores an image of the information indicating the task to be performed stored in the status register of the network interface unit such that information indicating the task to be performed is simultaneously stored in both of said mailbox storage register and said status register; and

wherein the at least one processing entity is configured to;

initially read the information indicating the task to be performed in the mailbox storage register upon detection of an interrupt pertaining to the network interface unit;

execute the task indicated by the information stored in the mailbox storage register; and

during the execution of the task, read the status register in the network interface unit to read information indicating one or more additional tasks to be performed by the at least one processing entity, wherein to read said information indicating one or more additional tasks to be performed, the at least one processing entity reads the status register in the network interface unit and not from the memory units.

16. The apparatus of claim 15 wherein:

the plurality of asymmetrical data processing partitions is scalable by adding additional processing entities.

17. The apparatus of claim 15 wherein the network interface unit comprises:

a memory access module to control the transfer of data, the memory access module including a plurality of parallel memory access channels; and

a packet classifier, the packet classifier providing an association between packets and the plurality of asymmetrical data processing partitions via the plurality of memory access channels.

18. The apparatus of claim 17 wherein:

the plurality of memory access channels are associated with the asymmetrical data processing partitions to provide affinity.

19. The apparatus of claim 17 wherein:

the plurality of memory access channels are virtualized and the network interface unit is shared among a plurality of domain environments.

20. The apparatus of claim 17 wherein:

packets are moved within the network interface unit independently and in parallel to optimize packet movement efficiency.

21. The apparatus of claim 17 wherein:

the packet classifier provides a hierarchical classification of packets processed by the asymmetrical data processing partitions.

22. The apparatus of claim 15, wherein the at least one processing entity is configured to read the information indicating the task to be performed in the status register in the network interface unit subsequent to reading the information indicating the task to be performed in the mailbox storage register.

23. The apparatus of claim 15, wherein the at least one processing entity is configured to, prior to completing the execution of the task, complete the reading of said information indicating the one or more additional tasks.