On 16/09/2016 15:46, Andrew Jones wrote:

quoted

But I haven't even looked at the patches in detail, I was just replying
to the comment about testing.

This may be a great time to start encouraging feature writers to submit
kvm-unit-tests patches at the same time as the feature (Hi Alex :-) I'm
happy to help when a test isn't easy to write due to a lack of framework,
but don't have nearly enough bandwidth to write all the tests myself.

In this case really a kernel smoke test can just work.  But yes,
architectural timer tests would be nice too.

As for additional motivation for this series, I'll point out that it's
good for bug isolation. When a guest fails to boot over KVM I can try
TCG. If that works, then I've likely narrowed it to KVM. If I can
further try kernel_irqchip=no, then I may further narrow it down to
the vgic implementation.

(Or vice versa to the userspace GIC if you're looking at a TCG bug).

Paolo