Techniques to provide recovery receive queues for flooded queues

• US 7,548,513 B2
• Filed: 02/17/2005
• Issued: 06/16/2009
• Est. Priority Date: 02/17/2005
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Tran, Hieu Trung
• Primary Examiner(s): Shah; Chirag G
• Assistant Examiner(s): Thompson, Jr.; Otis L
• Application Number:
  US11/061,997
• Publication Number:
  US 20060182031A1
• Time in Patent Office: 1,580 Days
• Field of Search: 370229-2351, 370/236.2, 370/412, 370/413
• US Class Current: 370/235
• CPC Class Codes: H04L 49/90 Buffering arrangements
• CPC Class Codes: H04L 49/9084 Reactions to storage capacity overflow

Abstract

In a receive-side scaling compatible environment, techniques to provide at least one recovery receive queue for any receive queue that is full or transferring traffic at too slow a rate.

Claims

1. A method comprising:

selectively identifying at least one flooded receive queue among multiple receive queues in response to a determination that more than a predefined number of ingress packets have been dropped instead of stored in a packet buffer, wherein the packet buffer is capable of receiving packets from a network medium;

selectively modifying an indirection table to route ingress packets destined to the at least one flooded receive queue to at least one non-flooded recovery receive queue;

selectively transferring an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table; and

selectively dropping the transferred ingress packet from the non-flooded recovery receive queue in response to a determination that the ingress packet was originally to be routed to the flooded receive queue.

2. The method of claim 1, wherein the selectively identifying at least one flooded receive queue further comprises selectively identifying at least one flooded receive queue in response to expiration of a flooded receive queue detection timer.

3. The method of claim 1, wherein the receive queues are utilized in a receive-side scaling environment.

4. The method of claim 1, further comprising storing the indirection table prior to selectively modifying the indirection table.

5. The method of claim 1, wherein the determination that the ingress packet was originally to be routed to the flooded receive queue comprises determining the routing of the ingress packet using the indirection table prior to modification.

6. The method of claim 1, further comprising:

selectively transferring an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table; and

selectively forwarding the transferred ingress packet from the non-flooded recovery receive queue for processing in response to a determination that the ingress packet was originally to be routed to the non-flooded recovery receive queue.

7. The method of claim 6, wherein the processing comprises TCP/IP stack processing.

8. The method of claim 1, further comprising:

selectively modifying the indirection table to route ingress packets to the at least one flooded receive queue in response to the at least one flooded receive queue being identified as not flooded.

9. An apparatus comprising:

a packet buffer to store ingress packets from a network medium;

multiple receive queues;

a hash calculator to determine a hash value associated with each of the ingress packets provided from the packet buffer;

an indirection table to identify a receive queue among the multiple receive queues in which to store an ingress packet based at least on an associated hash value;

logic to selectively determine if any of the multiple receive queues are flooded in response to the packet buffer dropping more than a predetermined amount of ingress packets;

logic to selectively modify the indirection table to route packets to at least one non-flooded recovery receive queue among the multiple receive queues in response to a receive queue determined as flooded; and

logic to selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table and logic to selectively drop the ingress packet from the non-flooded recovery queue in response to a determination that the ingress packet was originally to be routed to the flooded receive queue.

10. The apparatus of claim 9, wherein the logic to selectively determine if any of the multiple receive queues are flooded is to selectively determine if any of the multiple receive queues are flooded in response to expiration of a flooded receive queue detection timer.

11. The apparatus of claim 9, wherein the receive queues are utilized in a receive-side scaling environment.

12. The apparatus of claim 9, wherein the logic is to store the indirection table prior to selectively modifying the indirection table.

13. The apparatus of claim 9, further comprising logic to selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table and logic to selectively forward the ingress packet from the non-flooded recovery queue for processing in response to a determination that the ingress packet was originally to be routed to the non-flooded recovery queue.

14. The apparatus of claim 13, wherein the processing comprises TCP/IP stack processing.

15. The apparatus of claim 9, wherein the logic to selectively modify is to modify the indirection table to route ingress packets to the at least one flooded receive queue in response to the at least one flooded receive queue being identified as not flooded.

16. A computer-readable medium comprising instructions stored thereon which when executed by a machine cause the machine to:

selectively identify at least one flooded receive queue among multiple receive queues in response to a determination that more than a predefined number of ingress packets have been dropped by a packet buffer, wherein the packet buffer is capable of receiving packets from a network medium;

selectively modify an indirection table to route ingress packets destined to the at least one flooded receive queue to at least one non-flooded recovery receive queue;

selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table; and

selectively drop the transferred ingress packet from the non-flooded recovery receive queue in response to a determination that the ingress packet was originally to be routed to the flooded receive queue.

17. The computer-readable medium of claim 16, wherein the receive queues are utilized in a receive-side scaling environment.

18. The computer-readable medium of claim 16, further comprising instructions which when executed by a machine cause the machine to:

store the indirection table prior to selectively modifying the indirection table.

19. The computer-readable medium of claim 16, further comprising instructions which when executed by a machine cause the machine to:

selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table; and

selectively forward the transferred ingress packet from the non-flooded recovery receive queue for processing in response to a determination that the ingress packet was originally to be routed to the non-flooded recovery receive queue.

20. The computer-readable medium of claim 19, wherein the processing comprises TCP/IP stack processing.

21. The computer-readable medium of claim 19, wherein the instruction which when executed by a machine cause the machine to selectively modify an indirection table is further to:

selectively modify the indirection table to route ingress packets to the at least one flooded receive queue in response to the at least one flooded receive queue being identified as not flooded.

22. A system comprising:

a host system comprising a plurality of processors and a memory device;

a parallel bus;

a chipset to communicatively couple the host system to the parallel bus;

a network interface communicatively coupled to the parallel bus and comprising;

a packet buffer to store ingress packets from a network medium, multiple receive queues,

a hash calculator to determine a hash value associated with each of the ingress packets provided from the packet buffer,

an indirection table to identify a receive queue among the multiple receive queues in which to store an ingress packet based at least on an associated hash value,

logic to selectively determine if any of the multiple receive queues are flooded in response to the packet buffer dropping more than a predetermined amount of ingress packets, and

logic to selectively modify the indirection table to route packets to at least one non-flooded recovery receive queue among the multiple receive queues in response to a receive queue determined as flooded,

wherein the network interface further comprises logic to selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table and logic to selectively drop the ingress packet from the non-flooded recovery queue in response to a determination that the ingress packet was originally to be routed to the flooded receive queue.

23. The system of claim 22, wherein the network interface further comprises logic to selectively transfer an ingress packet to a non-flooded recovery receive queue based in part on the modified indirection table and logic to selectively forward the ingress packet from the non-flooded recovery queue for processing in response to a determination that the ingress packet was originally to be routed to the non-flooded recovery queue.

24. The system of claim 22, wherein the bus is compatible with PCI.

25. The system of claim 22, wherein the bus is compatible with PCI express.