System and method for fault tolerant TCP offload

• US 7,746,867 B2
• Filed: 04/04/2007
• Issued: 06/29/2010
• Est. Priority Date: 09/04/2002
• Status: Expired due to Fees

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: Broadcom Corporation (Broadcom, Inc.)
• Inventors: Fan, Kan Frankie
• Primary Examiner(s): Trost, IV; William
• Assistant Examiner(s): Maglo; Emmanuel
• Application Number:
  US11/696,398
• Publication Number:
  US 20070263630A1
• Time in Patent Office: 1,182 Days
• Field of Search: 370/389, 370/394, 370/395.52, 370/474, 709227-229
• US Class Current: 370/394
• CPC Class Codes: H04L 69/16 Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]

Abstract

Systems and methods that provide fault tolerant transmission control protocol (TCP) offloading are provided. In one example, a method that provides fault tolerant TCP offloading is provided. The method may include one or more of the following steps: receiving TCP segment via a TCP offload engine (TOE); calculating a TCP sequence number; writing a receive sequence record based upon at least the calculated TCP sequence number to a TCP sequence update queue in a host; and updating a first host variable with a value from the written receive sequence record.

Claims

1. In a Transmission Control Protocol (TCP) Offload Engine (TOE), a method for providing fault tolerant TCP offloading, the method comprising:

receiving a TCP segment;

calculating a TCP received sequence number based upon the received TCP segment;

updating a value of a first TOE variable with the TCP received sequence number; and

writing a receive sequence record based upon at least the TCP received sequence number to a TCP sequence update queue of a host, where the written receive sequence record causes the host to update a first host variable with the TCP received sequence number.

2. The method of claim 1, wherein the written receive sequence record causes the host to remove the written receive sequence record from the TCP sequence update queue.

3. The method of claim 1, further comprising, after writing the receive sequence record to the TCP sequence update queue in the hose, updating a value of a second TOE variable with the TCP received sequence number.

4. The method of claim 1, further comprising, if the received TCP segment acknowledges previously sent data, then:

updating a value of a second TOE variable with the TCP received sequence number; and

writing a send sequence record based upon at least the TCP received sequence number to the TCP sequence update queue, where the written send sequence record causes the host to update a second host variable with the TCP received sequence number.

5. The method of claim 4, wherein the written send sequence record causes the host to remove the written send sequence record from the TCP sequence update queue.

6. The method of claim 4, further comprising, if a fault is detected in the TOE, then causing the host to continue a TCP connection in software using at least one of the first host variable, the second host variable and the TCP sequence update queue.

7. The method of claim 1, further comprising, if a fault is detected in the TOE, then causing the host to continue a TCP connection in software.

8. The method of claim 7, wherein causing the host to continue a TCP connection in software comprises writing information to the TCP sequence update queue.

9. The method of claim 7, wherein causing the host to continue a TCP connection in software comprises causing the host to update TCP state variables using values reported by the TOE via the TCP sequence update queue.

10. The method of claim 7, wherein causing the host to continue a TCP connection in software comprises failing to respond properly to probing by the host.

11. The method of claim 1, further comprising transferring at least a payload of the received TCP segment to a host buffer via a direct memory access (DMA) engine.

12. In a Transmission Control Protocol (TCP) Offload Engine (TOE), a method for providing fault tolerant TCP offloading, the method comprising:

receiving a TCP segment;

calculating a TCP received sequence number based upon the received TCP segment; and

if the received TCP segment acknowledges previously sent data, then at least;

updating a value of a TOE variable with the TCP received sequence number; and

writing a send sequence record based upon at least the TCP received sequence number to a TCP sequence update queue of a host, where the written send sequence record causes the host to update a value of a host variable with a value from the written send sequence record.

13. The method of claim 12, further comprising updating a value of a first TOE variable with the TCP received sequence number.

14. The method of claim 13, further comprising writing a receive sequence record based upon at least the TCP received sequence number to the TCP sequence update queue.

15. The method of claim 12, wherein the written send sequence record causes the host to remove the written send sequence record from the TCP sequence update queue.

16. The method of claim 12, further comprising, if a fault is detected in the TOE, then causing the host to continue a TCP connection in software.

17. The method of claim 16, wherein causing the host to continue a TCP connection in software comprises writing information to the TCP sequence update queue.

18. The method of claim 16, wherein causing the host to continue a TCP connection in software comprises causing the host to update TCP state variables using values reported by the TOE via the TCP sequence update queue.

19. The method of claim 16, wherein causing the host to continue a TCP connection in software comprises failing to respond properly to probing by the host.

20. The method of claim 12, further comprising transferring at least a payload of the received TCP segment to a host buffer via a direct memory access (DMA) engine.

21. In a Transmission Control Protocol (TCP) Offload Engine (TOE) for providing fault tolerant TCP offloading from a host, at least one circuit that operates to, at least:

receive a TCP segment;

calculate a TCP received sequence number based upon the received TCP segment;

update a value of a first TOE variable with the TCP received sequence number; and

write a receive sequence record based upon at least the TCP received sequence number to a TCP sequence update data structure of a host, where the written receive sequence record causes the host to update a first host variable with the TCP received sequence number.

22. In a Transmission Control Protocol (TCP) Offload Engine (TOE) for providing fault tolerant TCP offloading from a host, at least one circuit that operates to, at least:

receive a TCP segment;

calculate a TCP received sequence number based upon the received TCP segment; and

if the received TCP segment acknowledges previously sent data, then at least;

update a value of a TOE variable with the TCP received sequence number; and

write a send sequence record based upon at least the TCP received sequence number to a TCP sequence update queue of a host, where the written send sequence record causes the host to update a value of a host variable with a value from the written send sequence record.