Implementing concurrent device driver maintenance and recovery for an SRIOV adapter in a virtualized system

Patent metadata

Field	Value
Publication number	US-9317317-B2
Application number	US-201313915737-A
Country	US
Kind code	B2
Filing date	Jun 12, 2013
Priority date	Jun 12, 2013
Publication date	Apr 19, 2016
Grant date	Apr 19, 2016

Abstract

A method, system and computer program product are provided for implementing concurrent adapter driver maintenance and recovery for a Single Root Input/Output Virtualization (SRIOV) adapter in a computer system. An adapter driver at start up time performs configuration of the adapter and each of a set of virtual functions (VFs). The adapter driver writes critical adapter and VF configuration data to a scratchpad buffer. When device driver maintenance is needed, such as to load updated adapter driver firmware, all VF drivers are held off temporarily, current adapter driver is detached, and then the adapter driver is reloaded with the updated driver firmware. Then the adapter driver is restarted with the updated adapter driver firmware, and performs a reinitialization process. The adapter driver performs adapter and VF configuration restoring existing configuration using values read from the scratchpad buffer.

First claim

What is claimed is:

1. A method for implementing concurrent device driver maintenance and recovery for a Single Root Input/Output Virtualization (SRIOV) adapter in a computer system comprising:
providing a physical function (PF) adjunct including an adapter driver with a hypervisor for managing physical functions and virtual functions of the Single Root Input/Output Virtualization (SRIOV) adapter; each virtual function being associated with a physical function, and the physical function supporting SRIOV capabilities;
performing an initialization process including said adapter driver performing adapter and virtual function (VF) configuration, and said adapter driver writing adapter and virtual function (VF) configuration data to a scratchpad buffer, said scratchpad buffer kept in said hypervisor and said scratchpad buffer being completely managed within said hypervisor without requiring external management;
responsive to configuration changes being made, said adapter driver updating configuration data in the scratchpad buffer;
responsive to adapter driver maintenance being needed, said hypervisor placing each VF driver in a freeze state, said hypervisor detaching a current adapter driver, and said hypervisor reloading the adapter driver with an updated adapter driver firmware;
said hypervisor restarting the adapter driver with the updated adapter driver firmware, and said adapter driver performing a reinitialization process including said adapter driver restoring adapter and virtual function (VF) configuration using configuration data values read from the scratchpad buffer; and
said hypervisor preserving the configuration data in the scratchpad buffer during adapter driver restart for implementing concurrent device driver maintenance and recovery for the Single Root Input/Output Virtualization (SRIOV) adapter.

2. The method as recited in claim 1 , wherein the initialization process includes starting an adapter driver and the adapter driver finding an empty scratchpad buffer.

3. The method as recited in claim 1 , wherein placing each VF driver in a freeze state includes placing all VF Partitionable Endpoints (PEs) in the freeze state.

4. The method as recited in claim 1 , includes the VF driver detecting an error condition responsive to being placed in the freeze state.

5. The method as recited in claim 4 , further includes the VF driver commencing an error handling recovery.

6. The method as recited in claim 5 , wherein performing a reinitialization process includes the adapter driver giving a system hypervisor permission for all VF Partitionable Endpoints (PEs) to be unfrozen.

7. The method as recited in claim 6 , includes said VF driver detecting permission for all VF Partitionable Endpoints (PEs) to be unfrozen, and continuing the error handling recovery.

8. The method as recited in claim 1 , wherein writing adapter and virtual function (VF) configuration data to a scratchpad buffer includes storing the adapter and virtual function (VF) configuration data in persistent data.

9. The method as recited in claim 1 , includes a hardware management console (HMC) notifying a system hypervisor of the updated adapter driver firmware being available.

10. The method as recited in claim 1 , includes using a system hypervisor, managing the scratchpad buffer within the system hypervisor.

11. The method as recited in claim 10 , wherein detaching a current adapter driver, and reloading the adapter driver with an updated adapter driver firmware includes restarting the adapter driver responsive to loading the updated adapter driver firmware, and finding configuration data values from the scratchpad buffer.

12. A system for implementing concurrent device driver maintenance and recovery for a Single Root Input/Output Virtualization (SRIOV) adapter in a computer system comprising:
a processor;
providing a physical function (PF) adjunct including an adapter driver with a hypervisor for managing physical functions and virtual functions of the Single Root Input/Output Virtualization (SRIOV) adapter; each virtual function being associated with a physical function, and the physical function supporting SRIOV capabilities;
said hypervisor managing functions associated with the SRIOV adapter;
said processor using said hypervisor and said adapter driver to perform the steps of:
performing an initialization process including said adapter driver performing adapter and virtual function (VF) configuration, and said adapter driver writing adapter and virtual function (VF) configuration data to a scratchpad buffer; said scratchpad buffer kept in said hypervisor and said scratchpad buffer being completely managed within said hypervisor without requiring external management;
responsive to configuration changes being made, said adapter driver updating configuration data in the scratchpad buffer;
responsive to adapter driver maintenance being needed, said hypervisor placing each VF driver in a freeze state, said hypervisor detaching a current adapter driver, and said hypervisor reloading the adapter driver with an updated adapter driver firmware;
said hypervisor restarting the adapter driver with the updated adapter driver firmware, said adapter driver performing a reinitialization process including restoring adapter and virtual function (VF) configuration using configuration data values read from the scratchpad buffer; and
said hypervisor preserving the configuration data in the scratchpad buffer during adapter driver restart for implementing concurrent device driver maintenance and recovery for the Single Root Input/Output Virtualization (SRIOV) adapter.

13. The system as recited in claim 12 , wherein the initialization process includes starting an adapter driver and the adapter driver finding an empty scratchpad buffer.

14. The system as recited in claim 12 , wherein placing each VF driver in a freeze state includes placing all VF Partitionable Endpoints (PEs) in the freeze state.

15. The system as recited in claim 12 , includes the VF driver detecting an error condition responsive to being placed in the freeze state.

16. The system as recited in claim 15 , further includes the VF driver commencing an error handling recovery.

17. The system as recited in claim 12 , wherein performing a reinitialization process includes the adapter driver giving a system hypervisor permission for all VF Partitionable Endpoints (PEs) to be unfrozen.

18. The system as recited in claim 17 , includes said VF driver detecting permission for all VF Partitionable Endpoints (PEs) to be unfrozen, and continuing the error handling recovery.

19. The system as recited in claim 12 , includes said hypervisor completely managing the scratchpad buffer within said hypervisor.

20. The system as recited in claim 12 , includes a hardware management console (HMC) notifying a system hypervisor of the updated adapter driver firmware being available.

Assignees

• IBM

Inventors

• Graham Charles S
• Oberly Iii John R
• Schimke Timothy J

Classifications

• G06F9/45533

  Hypervisors; Virtual machine monitors · CPC title

• G06F9/45558 Primary

  Hypervisor-specific management and integration aspects · CPC title

• G06F2009/45575

  Starting, stopping, suspending or resuming virtual machine instances · CPC title

• G06F9/5077

  Logical partitioning of resources; Management or configuration of virtualized resources (specific details on emulation or internal functioning of virtual machines G06F9/455) · CPC title

• G06F2009/45579

  I/O management, e.g. providing access to device drivers or storage · CPC title