On 05/19/2016 03:33 PM, Cole Robinson wrote:
My patches had an explicit PostParse check in generic code that would throw an error if <address type='pci'/> was never correctly filled in by the hypervisor, so <address type='pci/> would explicitly fail for all non-qemu drivers. It's a minor point
That sounds like a good idea, and should be trivial to add to the tail end of your address assignment callback. I'll make a followup patch for it.
upsides: - less magic - I think it will make allocating address at hotplug time simpler as well
I hadn't thought about that yet - more of the "90% of the coding effort going to 0.005% of the users" :-P
Yeah it's certainly not a blocker for this series, but hotplug will be relevant for aarch64 eventually
Actually it should "just work".