System and method for handling transport protocol segments

• US 8,244,890 B2
• Filed: 01/02/2004
• Issued: 08/14/2012
• Est. Priority Date: 03/08/2002
• Status: Expired due to Fees

• Current Assignee: Avago Technologies General IP PTE Limited (Broadcom, Inc.)
• Original Assignee: Broadcom Corporation (Broadcom, Inc.)
• Inventors: Elzur, Uri
• Primary Examiner(s): Shin, Kyung H
• Application Number:
  US10/751,732
• Publication Number:
  US 20040153578A1
• Time in Patent Office: 3,147 Days
• Field of Search: 709/230, 709/236, 709/250
• US Class Current: 709/230
• CPC Class Codes: H04L 12/413 with random access, e.g. carrier-sense multiple-access with collision detection [CSMA-CD]

Abstract

Systems and methods that handle transport protocol segments (TPSes) are provided. In one embodiment, a system may include, for example, a receiver that may receive an incoming TPS. The incoming TPS may include, for example, an aligned upper layer protocol (ULP) header and a complete ULP data unit (ULPDU). The receiver may directly place the complete ULPDU into a host memory.

Claims

1. A system, comprising:

a receiver that receives an incoming transport protocol segment (TPS), the incoming TPS comprising a TPS header, an aligned upper layer protocol (ULP) header, a complete ULP data unit (ULPDU), a marker header and a marker, the receiver comprising a buffer, the buffer being a size that does not scale approximately linearly with a network speed, a network bandwidth, or a number of connections,

wherein a direct memory access (DMA) engine of the receiver moves the complete ULPDU through a cyclical redundancy checking (CRC) machine and places the complete ULPDU into a host memory, and wherein the marker header is disposed between the aligned ULP header and the TPS header, and

wherein the marker of the incoming TPS is disposed in the complete ULPDU and backward points to the marker header of the same incoming TPS.

2. The system according to claim 1,

wherein the receiver comprises a network subsystem and the host memory,

wherein the network subsystem comprises the DMA engine and the CRC machine.

3. The system according to claim 2, wherein the network subsystem further comprises a network interface card (NIC) or a network controller.

4. The system according to claim 1, wherein the ULPDU comprises a framing protocol data unit (FPDU).

5. The system according to claim 4, wherein the FPDU comprises a data unit created by a ULP using a marker-based ULPDU aligned (MPA) framing protocol.

6. The system according to claim 1, wherein the aligned ULP header comprises an aligned FPDU header.

7. The system according to claim 6, wherein the aligned ULP header comprises the aligned FPDU header disposed adjacently to the TPS header of the incoming TPS.

8. The system according to claim 1, wherein the aligned ULP header is disposed a preset length away from the TPS header of the incoming TPS.

9. The system according to claim 1, wherein the aligned ULP header is disposed a predefined length away from the TPS header, the predefined length being related to information in a field in the incoming TPS.

10. The system according to claim 9, wherein the field comprises a marker field.

11. The system according to claim 1, wherein the receiver is a flow-through receiver.

12. The system according to claim 1, wherein the incoming TPS comprises a transmission control protocol (TCP) segment.

13. The system according to claim 12, wherein the TCP segment is part of a TCP byte stream.

14. The system according to claim 1, wherein the incoming TPS comprises information that is used to place the complete ULPDU in the host memory.

15. The system according to claim 1, wherein the receiver does not store partial cyclical redundancy check (CRC) values.

16. The system according to claim 1, wherein the incoming TPS comprises an out-of-order incoming TPS.

17. The system according to claim 1, wherein the receiver does not store only a portion of the complete ULPDU.

18. A method, comprising:

receiving an incoming transport protocol segment (TPS) in a buffer, the buffer being a size that does not scale approximately linearly with a network speed, a network bandwidth, or a number of connections, the incoming TPS comprising a complete framing protocol data unit (FPDU) and an FPDU header in a known position with respect to a TPS header, wherein a direct memory access (DMA) engine moves the complete FPDU through a cyclical redundancy check (CRC) machine, wherein the complete FPDU includes a marker, wherein the marker is not adjacent to the FPDU header, and wherein the marker of the incoming TPS backward points to the FPDU header of the same incoming TPS.

19. The method according to claim 18, wherein the FPDU header is adjacent to the TPS header.

20. The method according to claim 18, further comprising:

performing layer 2 (L2) processing, layer 3 (L3) processing and layer 4 (L4) processing on the incoming TPS via a network subsystem.

21. The method according to claim 20, further comprising:

obtaining FPDU length information from the FPDU header.

22. The method according to claim 21, further comprising:

programming the DMA engine to copy data of the complete FPDU from the network subsystem to a host memory.

23. The method according to claim 20, further comprising:

performing upper layer protocol (ULP) processing on the incoming TPS via the network subsystem,

wherein the L2 processing, the L3 processing, the L4 processing and the ULP processing can occur in parallel or in any order.

24. The method according to claim 23,

wherein the L2 processing, the L3 processing, the L4 processing and the ULP processing do not occur in the listed order in a receiver.

25. The method according to claim 23,

wherein the ULP processing, the L4 processing, the L3 processing and the L2 processing do no occur in the listed order in a transmitter.

26. The method according to claim 20, wherein the TPS comprises a plurality of complete FPDUs.

27. A system, comprising:

a receiver comprising a direct memory access (DMA) engine, a cyclical redundancy check (CRC) machine, and a buffer being a size that does not scale approximately linearly with a network speed, a network bandwidth, or a number of connections,

wherein the receiver receives an incoming transport protocol segment (TPS), the incoming TPS comprising a marker, a marker header, a TPS header, an aligned upper layer protocol (ULP) header and a complete ULP data unit (ULPDU),

wherein the marker header is disposed between the aligned ULP header and the TPS header, wherein the marker of the incoming TPS is disposed in the complete ULPDU and backward points to the marker header of the same incoming TPS, and wherein the receiver programs the DMA engine once to move the complete ULPDU through the CRC machine and place the complete ULPDU into a host memory.

28. The system according to claim 27,

wherein the receiver comprises a non-flow-through network interface card (NIC), and

wherein the DMA engine and the CRC machine are part of the non-flow-through NIC.

29. The system according to claim 28, wherein the non-flow-through NIC comprises a local memory.

30. The system according to claim 29, wherein the non-flow-through NIC performs a CRC calculation before or as the complete ULPDU is stored in the local memory.

31. The system according to claim 29, wherein the non-flow-through NIC performs a CRC calculation after the complete ULPDU is stored in the local memory.

32. The system according to claim 29, wherein the non-flow-through NIC performs a CRC calculation during a process by which the complete ULPDU is sent from the local memory to the host memory.

33. The system according to claim 29, wherein the complete ULPDU comprises a marker-aligned protocol data unit.

34. The system according to claim 27,

wherein the receiver comprises a flow-through network interface card (NIC), and

wherein the DMA engine and the CRC machine are part of the flow-through NIC.

35. The system according to claim 34, wherein the flow-through NIC comprises the buffer.

36. The system according to claim 35, wherein the flow-through NIC performs a CRC calculation before or as the complete ULPDU is stored in the buffer.

37. The system according to claim 36, wherein the CRC calculation is a ULP CRC calculation.

38. The system according to claim 34, wherein the complete ULPDU comprises a marker-aligned protocol data unit.

39. A method, comprising:

(a) receiving an incoming transport protocol segment (TPS) in a buffer, the buffer being a size that does not scale approximately linearly with a network speed, a network bandwidth, or a number of connections, the TPS comprising a TPS header, a marker, a complete framing protocol data unit (FPDU) and an FPDU header in a known position with respect to the TPS header, wherein a direct memory access (DMA) engine is programmed to move the complete FPDU through a cyclical redundancy check (CRC) machine and copy data of the complete FPDU to a memory, wherein the marker is inserted in the complete FPDU, wherein the marker is not adjacent to the FPDU header, and wherein the marker of the incoming TPS backward points to the FPDU header of the same incoming TPS;

(b) performing layer 2 (L2) processing on the incoming TPS;

(c) performing layer 3 (L3) processing on the incoming TPS;

(d) performing layer 4 (L4) processing on the incoming TPS; and

(e) performing ULP processing on the incoming TPS.

40. The method according to claim 39, wherein at least two of the performing of (b), (c), (d) and (e) occurs concurrently.

41. The method according to claim 39, wherein the performing of (b), (c), (d) and (e) occurs in any order.