On 22 October 2012 08:51, Christoffer Dall
[off-list ref] wrote:

+struct kvm_device_address {
+       __u32 id;
+       __u64 addr;
+};

Ben suggested that this should either be a 64 bit id or have explicit
padding. Other than that I think that our current proposed ABI for
ARM irqchips is in line with the discussion we just had at KVM Forum
for handling non-x86 in-kernel irqchips [hopefully somebody will
write the details of that up...]

Ben, do I have that right?

thanks
-- PMM