Virtualization of hardware queues in self-virtualizing input/output devices

• US 8,881,141 B2
• Filed: 12/08/2010
• Issued: 11/04/2014
• Est. Priority Date: 12/08/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Koch, Andrew T., Lucke, Kyle A., Rogness, Nicholas J.
• Primary Examiner(s): Puente, Emerson
• Assistant Examiner(s): Huaracha, Willy W
• Application Number:
  US12/963,256
• Publication Number:
  US 20120151472A1
• Time in Patent Office: 1,427 Days
• Field of Search: 718/1, 718/100, 718/104
• US Class Current: 718/1
• CPC Class Codes: G06F 2009/45579 I/O management, e.g. providing access to device drivers or storage
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects

Abstract

Hardware transmit and/or receive queues in a self-virtualizing IO resource are virtualized to effectively abstract away resource-specific details for the self-virtualizing IO resource. By doing so, a logical partition may be permitted to configure and access a desired number of virtual transmit and/or receive queues, and have an adjunct partition that interfaces the logical partition with the self-virtualizing IO resource handle the appropriate mappings between the hardware and virtual queues.

Claims

1. A method of accessing a self-virtualizing input/output (IO) resource with multiple logical partitions in a data processing system, the method comprising:

interfacing a logical partition of the multiple logical partitions to the self-virtualizing IO resource through an associated adjunct partition; and

in the associated adjunct partition, mapping a plurality of virtual queues in the logical partition to a plurality of hardware queues in the self-virtualizing IO resource.

2. The method of claim 1, wherein the plurality of hardware queues comprises a plurality of hardware transmit queues, and wherein the plurality of virtual queues comprises a plurality of virtual transmit queues, wherein a number of virtual transmit queues in the logical partition is different from a number of hardware transmit queues supported by the self-virtualizing IO resource.

3. The method of claim 1, wherein the plurality of hardware queues comprises a plurality of hardware receive queues, and wherein the plurality of virtual queues comprises a plurality of virtual receive queues, and wherein a number of virtual receive queues in the logical partition is different from a number of hardware receive queues supported by the self-virtualizing IO resource.

4. The method of claim 1, wherein the associated adjunct partition includes a separate dispatchable state and employs virtual address space donated from the logical partition or a hypervisor of the data processing system.

5. The method of claim 1, further comprising:

interfacing the logical partition to the associated adjunct partition through a vendor-independent virtual network partition interface, the associated adjunct partition including a vendor-independent server virtual network partition interface device driver configured to interface with the logical partition and a resource-specific device driver configured to interface with the self-virtualizing IO resource; and

in the logical partition, accessing the self-virtualizing IO resource by communicating with the vendor-independent server virtual network partition interface device driver over the vendor-independent virtual network partition interface using a vendor-independent client virtual network partition interface device driver resident in the logical partition, wherein the plurality of virtual queues are resident in the client virtual network partition interface device driver.

6. The method of claim 5, wherein interfacing the logical partition to the associated adjunct partition comprises interfacing the associated adjunct partition to an associated virtual function of the self-virtualizing IO resource, and wherein the self-virtualizing IO resource further comprises a physical function, the method further comprising interfacing a physical function adjunct partition with the physical function of the self-virtualizing IO resource, the physical function adjunct partition facilitating creation of the adjunct partition interfacing the logical partition to the virtual function.

7. The method of claim 5, further comprising interfacing the vendor-independent server virtual network partition interface device driver and the resource-specific device driver through a device driver application programming interface (API), wherein mapping the plurality of virtual queues to the plurality of hardware queues is performed by the device driver API.

8. The method of claim 7, further comprising, in the vendor-independent server virtual network partition interface device driver:

in response to a first request from the vendor-independent client virtual network partition interface device driver, reporting at least one of a minimum and a maximum number of virtual queues; and

in response to a second request from the vendor-independent client virtual network partition interface device driver, creating handles to a requested number of virtual queues and returning the created handles to the vendor-independent client virtual network partition interface device driver.

9. The method of claim 8, wherein mapping the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource further comprises, in response to the vendor-independent client virtual network partition interface device driver writing data into a first virtual queue among the plurality of virtual queues, mapping the first virtual queue to a first hardware queue among the plurality of hardware queues and notifying the resource-specific device driver to write the data to the first hardware queue.

10. The method of claim 9, wherein mapping the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource includes selecting the first hardware queue to map to the first virtual queue based upon at least one of a hash algorithm, a round robin algorithm and a modulo algorithm.

11. The method of claim 9, wherein mapping the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource includes selecting the first hardware queue to map to the first virtual queue based upon a load balancing algorithm.

12. The method of claim 9, wherein mapping the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource includes inspecting traffic written to the first virtual queue and selecting the first hardware queue to map to the first virtual queue based upon the inspected traffic.

13. The method of claim 1, wherein the self-virtualizing IO resource comprises a single root input/output virtualization (SRIOV) Ethernet adapter.

14. An apparatus, comprising:

at least one processor; and

program code configured upon execution by the at least one processor to interface a self-virtualizing input/output (IO) resource with multiple logical partitions in a data processing system by interfacing a logical partition of the multiple logical partitions to the self-virtualizing IO resource through an associated adjunct partition, and, in the associated adjunct partition, mapping a plurality of virtual queues in the logical partition to a plurality of hardware queues in the self-virtualizing IO resource.

15. The apparatus of claim 14, wherein the plurality of hardware queues comprises a plurality of hardware transmit queues, wherein the plurality of virtual queues comprises a plurality of virtual transmit queues, and wherein a number of virtual transmit queues in the logical partition is different from a number of hardware transmit queues supported by the self-virtualizing IO resource.

16. The apparatus of claim 14, wherein the plurality of hardware queues comprises a plurality of hardware receive queues, wherein the plurality of virtual queues comprises a plurality of virtual receive queues, and wherein a number of virtual receive queues in the logical partition is different from a number of hardware receive queues supported by the self-virtualizing IO resource.

17. The apparatus of claim 14, wherein the associated adjunct partition includes a separate dispatchable state and employs virtual address space donated from the logical partition or a hypervisor of the data processing system.

18. The apparatus of claim 14, wherein the program code is further configured to:

interface the logical partition to the associated adjunct partition through a vendor-independent virtual network partition interface, the associated adjunct partition including a vendor-independent server virtual network partition interface device driver configured to interface with the logical partition and a resource-specific device driver configured to interface with the self-virtualizing IO resource; and

in the logical partition, access the self-virtualizing IO resource by communicating with the vendor-independent server virtual network partition interface device driver over the vendor-independent virtual network partition interface using a vendor-independent client virtual network partition interface device driver resident in the logical partition, wherein the plurality of virtual queues are resident in the client virtual network partition interface device driver.

19. The apparatus of claim 18, wherein the program code is further configured to interface the vendor-independent server virtual network partition interface device driver and the resource-specific device driver through a device driver application programming interface (API), wherein the device driver API is configured to map the plurality of virtual queues to the plurality of hardware queues.

20. The apparatus of claim 19, wherein the program code is further configured to, in the vendor-independent server virtual network partition interface device driver:

in response to a first request from the vendor-independent client virtual network partition interface device driver, report at least one of a minimum and a maximum number of virtual queues; and

in response to a second request from the vendor-independent client virtual network partition interface device driver, create handles to a requested number of virtual queues and return the created handles to the vendor-independent client virtual network partition interface device driver.

21. The apparatus of claim 20, wherein the program code is configured to map the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource further by, in response to the vendor-independent client virtual network partition interface device driver writing data into a first virtual queue among the plurality of virtual queues, mapping the first virtual queue to a first hardware queue among the plurality of hardware queues and notifying the resource-specific device driver to write the data to the first hardware queue.

22. The apparatus of claim 21, wherein the program code is configured to map the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource by selecting the first hardware queue to map to the first virtual queue based upon at least one of a hash algorithm, a round robin algorithm and a modulo algorithm.

23. The apparatus of claim 21, wherein the program code is configured to map the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource by selecting the first hardware queue to map to the first virtual queue based upon a load balancing algorithm.

24. The apparatus of claim 21, wherein the program code is configured to map the plurality of virtual queues in the logical partition to the plurality of hardware queues in the self-virtualizing IO resource by inspecting traffic written to the first virtual queue and selecting the first hardware queue to map to the first virtual queue based upon the inspected traffic.

25. A program product, comprising:

a non-transitory computer readable medium; and

program code stored on the computer readable medium and configured upon execution to interface a self-virtualizing input/output (IO) resource with multiple logical partitions in a data processing system by interfacing a logical partition of the multiple logical partitions to the self-virtualizing IO resource through an associated adjunct partition, and, in the associated adjunct partition, mapping a plurality of virtual queues in the logical partition to a plurality of hardware queues in the self-virtualizing IO resource.