bfrogers

There are issues with S3 and S4 handling in a KVM SLES or openSUSE guest. I'd like to figure out what's going wrong and get things fixed as much as possible.

Although there isn't much demand right now, it won't be long before >256 vcpus in a KVM guest will be desired. Currently, due to 8 bit apic ids only being used the limit is 256. Implement the x2apic infrastructure that will allow QEMU + KVM to break the 256 vcpu barrier.

I just learned about xenctx, which is used to help debug Xen issues. I am wondering if it would be easily ported to KVM. If not, perhaps at least creating the equivalent for KVM may be useful for handling problems with KVM guests.

Our QEMU/KVM testing needs to be improved.

Having worked on KVM for a long time, it's time I actually start using it myself! I now have a high enough powered laptop to have a permanent setup on my laptop to do almost all the testing needed for releases, as well as maintaining playgrounds for upstream work and involvement. This will include nested virtualization, which is getting pretty bulletproof in latest kernels, as well as being able to play more with these other architectures that we now support KVM on, via improved TCG emulation (of course some testing and development will always rely on the physical hardware, but still a lot can be done via TCG incl. user-linux mode. I'm also seeing which aspects of pass-through testing will make the most sense to perform on the laptop (pci, usb, video, filesystem, etc.).

There are a number of items related to doing Virtualization CI and testing which I have a hard time getting to during my day job. I'll take Hackweek as an opportunity to get at least a little out of my normal space and improve our methods. This mainly focuses on QEMU/KVM and some of the efforts and infrastructure I've already got in various states of development, or at least investigation.