Implement Optimized VirtIO interfaces (STR-9)
- STR-5 - VirtIO RPMB
- Azzedine : is it working ? Alex blockers spec differs from what the driver does, been fixing the kernel, a few edges to clean up. This may lead into minimal memory profile, not many examples, this may be a good candidate for example
- STR-8 - Prototype a minimal memory profile virtio backed with front end changes
- STR-6 - VirtIO-MMU support for inter-guest memory protection
- STR-15 - A linux host API for proving a limited sharing of a range of address space between two KVM guests
- Stefano : I have feedback - will do off line - quick thoughts
- This should work on any Hvz, be good to have a design doc for a type-1 (Xen) imagine soemthing like virtio-net - confomrs to standards, arch understood. Here there is ahigh chance this will not end up like that. the virtio memory mapping becaend will be in Xen becasue the meory is handled by Xen. This is challenging, this is a new exciting thing. Design needs to be cross Hyz - just page long.
- Alex : Xen has grant tables already doent it ? S: if we use table model - then yes we can do this, but virtio spec says you should not use references on the rings. Also if virtiommu in Xen will that be ok ? versiosning problem. S: waterfall does not work, just want to sketch issues
- Alex will sync with Arnd B and stimulate discussion on the lists
- STR-13 - A demonstration of the portability of Virtio FE's between KVM and Xen systems
Remaining effort for 4 planned epics: 6.2 engineer months