Method, system, and article of manufacture for utilizing host memory from an offload adapter

• US 7,400,639 B2
• Filed: 08/07/2003
• Issued: 07/15/2008
• Est. Priority Date: 08/07/2003
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Madukkarumukumana, Rajesh S., Ni, Jie
• Primary Examiner(s): Marcelo; Melvin
• Assistant Examiner(s): Faroul; Farah
• Application Number:
  US10/637,370
• Publication Number:
  US 20050030972A1
• Time in Patent Office: 1,804 Days
• Field of Search: 370/429, 370/463, 370/411, 370/412, 370/413, 370/474, 370/476, 370/395.41, 370/395.52, 370/395.53, 370/428, 370/465, 395/200, 395/250, 395/275
• US Class Current: 370/429
• CPC Class Codes: H04L 49/90 Buffering arrangements
• CPC Class Codes: H04L 49/9026 Single buffer per packet
• CPC Class Codes: H04L 49/9052 with buffers of different sizes

Abstract

Provided are a method, system, and article of manufacture, wherein in one implementation of the method a memory buffer is allocated in a host memory by a protocol processor, wherein the host memory is comprised in a host that is coupled to a network adapter. A packet is processed initially at the network adapter to generate data for offloading to the memory buffer in the host. The generated data is offloaded by the protocol processor to the memory buffer in the host. The offloaded data is processed by the protocol processor.

Claims

1. A method, comprising:

allocating, by a protocol processor included in a network adapter, a memory buffer in a host memory, wherein the host memory is comprised in a host that is coupled to the network adapter;

initially processing a packet at the network adapter to generate data for offloading to the memory buffer in the host;

offloading, by the protocol processor, the generated data to the memory buffer in the host; and

subsequent to the offloading by the protocol processor of the generated data to the memory buffer in the host, processing, by the protocol processor, the offloaded data, wherein the method further comprises;

(i) dividing the memory buffer into a plurality of buffers, wherein the plurality of buffers stores packet data associated with a plurality of network sessions, wherein an address of the memory buffer is not exposed directly or indirectly to a host application that executes in the host, and wherein the memory buffer comprises a non-pageable memory;

(ii) dividing a buffer into one or more packet regions, wherein a size of a packet region exceeds a size of a maximum transmission unit of packets associated with a protocol; and

(iii) storing a first packet data associated with a first packet in a first packet region, wherein if a data payload in the first packet exceeds a size of the host application buffer, then queuing the data payload into the memory buffer.

2. The method of claim 1, further comprising:

receiving a send request at the protocol processor from a host application on the host;

determining if a session associated with the send request is in a synchronous mode, wherein if the session is in the synchronous mode then offloading the send request to the memory buffer; and

indicating to the host application that the send request is completed.

3. The method of claim 1, further comprising:

receiving a send request at the protocol processor from a host application on the host, wherein an application send buffer is associated with the send request; and

copying the application send buffer to the memory buffer.

4. The method of claim 1, wherein the protocol processor is coupled to the network adapter, wherein the network adapter is an offload engine adapter, and wherein the host memory is larger in size than an adapter memory coupled to the network adapter.

5. The method of claim 1, wherein the memory buffer comprises a plurality of buffers, wherein the memory buffer stores the offloaded data for a plurality of sessions, and wherein the offloaded data for a first session is stored in at least one buffer.

6. The method of claim 1, wherein the protocol processor is implemented in hardware or software, and wherein the network adapter is a part of a central processing unit of the host.

21. The method of claim 1, wherein while transmitting packets the protocol processor uses the relatively greater storage capacity of the host memory rather than the relatively lesser storage capacity of an adapter memory for queuing packets, wherein the adapter memory is coupled to the network adapter.

22. The method claim 21, further comprising:

receiving, at the protocol processor, a send request from the host application;

determining if a measured roundtrip time estimate for a network session is above a threshold, in response to determining that the network session is in a synchronous mode;

allocating an anonymous transmit buffer and copying an application send buffer to the anonymous transmit buffer, in response to determining that the measured roundtrip time estimate for the network session is above the threshold; and

queuing the anonymous transmit buffer to a session receive queue and marking the send request as complete.

7. A network adapter, wherein the network adapter is coupled to a host having a host memory, and wherein a memory buffer is implemented in the host memory, the network adapter comprising:

a memory coupled to the network adapter;

a protocol processor included in the network adapter, wherein the protocol processor is allocates the memory buffer, wherein the protocol processor initially processes a packet at the network adapter to generate data to offload to the memory buffer, wherein the protocol processor offloads the generated data to the memory buffer in the host, and wherein the protocol processor processes the offloaded data subsequent to offloading the generated data to the memory buffer in the host, wherein the protocol processor further performs;

(i) dividing the memory buffer into a plurality of buffers to store packet data associated with a plurality of network sessions, wherein an address of the memory buffer is not exposed directly or indirectly to a host application that executes in the host, and wherein the memory buffer comprises a non-pageable memory;

(ii) dividing a buffer into one or more packet regions, wherein a size of a packet region exceeds a size of a maximum transmission unit of packets associated with a protocol; and

(iii) storing a first packet data associated with a first packet in a first packet region, wherein if a data payload in the first packet exceeds a size of the host application buffer, then queuing the data payload into the memory buffer.

8. The network adapter of claim 7, wherein the protocol processor performs:

receiving a send request from a host application on the host;

determining if a session associated with the send request is in a synchronous mode;

offloading the send request to the memory buffer if the session is in the synchronous mode; and

indicating to the host application that the send request is completed.

9. The network adapter of claim 7, wherein the protocol processor performs:

receiving a send request from a host application on the host, wherein an application send buffer is associated with the send request; and

copying the application send buffer to the memory buffer.

10. The network adapter of claim 7, wherein the protocol processor is coupled to the network adapter, wherein the network adapter is an offload engine adapter, and wherein the host memory is larger in size than the memory coupled to the network adapter.

11. The network adapter of claim 7, wherein the protocol processor performs:

dividing the memory buffer into a plurality of buffers;

storing offloaded data for a plurality of sessions in the plurality of buffers; and

storing the offloaded data for a first session in at least one buffer.

12. The network adapter of claim 7, wherein the protocol processor is implemented in hardware or software, and wherein the network adapter is a part of a central processing unit of the host.

23. The network adapter of claim 7, wherein while transmitting packets the protocol processor uses the relatively greater storage capacity of the host memory rather than the relatively lesser storage capacity of an adapter memory for queuing packets, wherein the adapter memory is coupled to the network adapter.

24. The network adapter of claim 23, wherein the protocol processor further performs:

receiving a send request from the host application;

determining if a measured roundtrip time estimate for a network session is above a threshold, in response to determining that the network session is in a synchronous mode;

allocating an anonymous transmit buffer and copying an application send buffer to the anonymous transmit buffer, in response to determining that the measured roundtrip time estimate for the network session is above the threshold; and

queuing the anonymous transmit buffer to a session receive queue and marking the send request as complete.

13. A system in communication with data storage, comprising:

a host;

a data storage controller to manage Input/Output (I/O) access to the data storage, wherein the data storage controller is coupled to the host;

a network adapter coupled to the host;

a host memory coupled to the host;

a memory buffer implemented in the host memory, wherein a protocol processor included in the network adapter is allocates the memory buffer, wherein the protocol processor initially processes a packet at the network adapter to generate data to offload to the memory buffer, wherein the protocol processor is offloads the generated data to the memory buffer in the host, and wherein the protocol processor processes the offloaded data subsequent to offloading the generated data to the memory buffer in the host, wherein;

(i) the memory buffer is divided into a plurality of buffers, wherein the plurality of buffers stores packet data associated with a plurality of network sessions, wherein an address of the memory buffer is not exposed directly or indirectly to a host application that executes in the host, and wherein the memory buffer comprises a non-pageable memory;

(ii) a buffer is divided into one or more packet regions, wherein a size of a packet region exceeds a size of a maximum transmission unit of packets associated with a protocol; and

(iii) a first packet data associated with a first packet is stored in a first packet region, wherein if a data payload in the first packet exceeds a size of the host application buffer, then queuing the data payload into the memory buffer.

14. The system of claim 13, further comprising:

a host application on the host, wherein the protocol processor receives a send request from the host application, wherein the protocol processor determines if a session associated with the send request is in an synchronous mode, wherein the protocol processor performs the offloading of the send request to the memory buffer if the session is in the synchronous mode, and wherein the protocol processor indicates to the host application that the send request is completed.

25. The system of claim 13, wherein while transmitting packets the protocol processor uses the relatively greater storage capacity of the host memory rather than the relatively lesser storage capacity of an adapter memory for queuing packets, wherein the adapter memory is coupled to the network adapter.

26. The system of claim 25, wherein the protocol processor further performs:

receiving a send request from the host application;

determining if a measured roundtrip time estimate for a network session is above a threshold, in response to determining that the network session is in a synchronous mode;

allocating an anonymous transmit buffer and copying an application send buffer to the anonymous transmit buffer, in response to determining that the measured roundtrip time estimate for the network session is above the threshold; and

queuing the anonymous transmit buffer to a session receive queue and marking the send request as complete.

15. An article of manufacture, comprising a storage medium having stored therein instructions that when executed by a machine results in the following:

allocating a memory buffer in a host memory, wherein the host memory is comprised in a host that is coupled to a network adapter;

initially processing a packet at the network adapter to generate data for offloading to the memory buffer in the host;

offloading, by a protocol processor included in the network adapter, the generated data to the memory buffer in the host; and

subsequent to the offloading by the protocol processor of the generated data to the memory buffer in the host, processing, by the protocol processor, the offloaded data, wherein the instructions when executed further results in the following;

(i) dividing the memory buffer into a plurality of buffers, wherein the plurality of buffers stores packet data associated with a plurality of network sessions, wherein an address of the memory buffer is not exposed directly or indirectly to a host application that executes in the host, and wherein the memory buffer comprises a non-pageable memory;

(ii) dividing a buffer into one or more packet regions, wherein a size of a packet region exceeds a size of a maximum transmission unit of packets associated with a protocol; and

(iii) storing a first packet data associated with a first packet in a first packet region, wherein if a data payload in the first packet exceeds a size of the host application buffer, then queuing the data payload into the memory buffer.

16. The article of manufacture of claim 15, wherein the instructions when executed further results in the following:

receiving a send request at the network adapter from a host application on the host;

determining if a session associated with the send request is in a synchronous mode, wherein if the session is in the synchronous mode then performing the offloading of the send request to the memory buffer; and

indicating to the host application that the send request is completed.

17. The article of manufacture of claim 15, wherein the instructions when executed further results in the following:

receiving a send request at the network adapter from a host application on the host, wherein an application send buffer is associated with the send request; and

copying the application send buffer to the memory buffer.

18. The article of manufacture of claim 15, wherein the network adapter is an offload engine adapter, and wherein the host memory is larger in size than a memory coupled to the network adapter.

19. The article of manufacture of claim 15, wherein the memory buffer comprises a plurality of buffers, wherein the instructions when executed store offloaded data for a plurality of sessions in the memory buffer, and wherein the instructions when executed offload data for a first session in at least one buffer.

20. The article of manufacture of claim 15, wherein the instructions are implemented in hardware or software, and wherein the network adapter is a part of a central processing unit of the host.

27. The article of manufacture of claim 15, wherein while transmitting packets the protocol processor uses the relatively greater storage capacity of the host memory rather than the relatively lesser storage capacity of an adapter memory for queuing packets, wherein the adapter memory is coupled to the network adapter.

28. The article of manufacture of claim 27, wherein the instructions when executed further results in the following:

receiving, at the protocol processor, a send request from the host application;

determining if a measured roundtrip time estimate for a network session is above a threshold, in response to determining that the network session is in a synchronous mode;

allocating an anonymous transmit buffer and copying an application send buffer to the anonymous transmit buffer, in response to determining that the measured roundtrip time estimate for the network session is above the threshold; and

queuing the anonymous transmit buffer to a session receive queue and marking the send request as complete.