Re: [Linaro-acpi] [RFC 2/2] ARM64 / ACPI: Point KVM to the virtual timer interrupt when booting with ACPI

On 6 October 2014 18:04, Wei Huang wei@redhat.com wrote:

Revisited Alexander's patches, I felt that the approach of if-DT-else-ACPI approach isn't necessary. More importantly this whole DT parsing in ./virt/kvm/arm can really be avoided. My opinions are:

1. All DT-ACPI should happen in respective drivers. In this case, they should be in clocksource/arm_arch_timer.c and irqchip/irq-gic* only. The driver will expose necessary info to KVM via public interfaces.
2. KVM doesn't have to care about DT/ACPI. Instead it only uses the info provided by the interface from device drivers.

The attached patch demos how to do it in arch_timer. GIC can follow the same approach (though trickier than arch_timer, due to GICv2 and GICv3).

Any though?

I am also in favour of this approach, it makes sense to decouple these parts from DT/ACPI.

Thanks.