Proposals

VFIO: PCI device assignment breaks free of KVM

Session information has not yet been published for this event.

*

One Line Summary

VFIO allows exporting PCI devices to userspace drivers. Discussion of using this interface for assigning physical devices to virtual machines and obstacles remaining for integration.

Abstract

PCI device assignment has long been a wort on the side of KVM; a marriage of convenience, filling the gaps left by PCI-sysfs and lack of sufficient PCI user-space I/O driver support. The new VFIO driver changes the game, allowing full featured, high performance, PCI drivers in user-space, providing the perfect basis for virtual machine PCI device assignment. This talk will discuss how current KVM PCI device assignment works, how VFIO is better, and discuss some of the obstacles left to kernel integration and cross platform support.

Presentation Materials

slides

Speaker

  • Alex Williamson

    Red Hat

    Biography

    Alex is a Senior Software Engineer working for Red Hat from his home in Fort Collins, Colorado. Alex has a long history of Linux kernel and open source virtualization work with over 10 years as an engineer with HP’s open source division. Alex is currently working on KVM/QEMU-based virtualization with a focus on PCI device assignment.