[RFC PATCH 4/7] KVM: arm/arm64: Check that system supports split eoi/deactivate

[Marc Zyngier]

On 06/01/17 10:02, Christoffer Dall wrote:
> On Thu, Jan 05, 2017 at 05:40:58PM +0000, Marc Zyngier wrote:
>> On 10/12/16 20:47, Christoffer Dall wrote:
>>> Some systems without proper firmware and/or hardware description data
>>> don't support the split EOI and deactivate operation and therefore
>>> don't provide an irq_set_vcpu_affinity implementation.  On such
>>> systems, we cannot leave the physical interrupt active after the timer
>>> handler on the host has run, so we cannot support KVM with the timer
>>> changes we about to introduce.
>>>
>>> Signed-off-by: Christoffer Dall <christoffer.dall at linaro.org>
>>> ---
>>>  virt/kvm/arm/arch_timer.c | 30 ++++++++++++++++++++++++++++++
>>>  1 file changed, 30 insertions(+)
>>>
>>> diff --git a/virt/kvm/arm/arch_timer.c b/virt/kvm/arm/arch_timer.c
>>> index c7c3bfd..f27a086 100644
>>> --- a/virt/kvm/arm/arch_timer.c
>>> +++ b/virt/kvm/arm/arch_timer.c
>>> @@ -418,6 +418,31 @@ static int kvm_timer_dying_cpu(unsigned int cpu)
>>>  	return 0;
>>>  }
>>>  
>>> +static bool has_split_eoi_deactivate_support(void)
>>> +{
>>> +	struct irq_desc *desc;
>>> +	struct irq_data *data;
>>> +	struct irq_chip *chip;
>>> +
>>> +	/*
>>> +	 * Check if split EOI and deactivate is supported on this machine.
>>> +	 */
>>> +	desc = irq_to_desc(host_vtimer_irq);
>>> +	if (!desc) {
>>> +		kvm_err("kvm_arch_timer: no host_vtimer_irq descriptor\n");
>>> +		return false;
>>> +	}
>>> +
>>> +	data = irq_desc_get_irq_data(desc);
>>> +	chip = irq_data_get_irq_chip(data);
>>> +	if (!chip || !chip->irq_set_vcpu_affinity) {
>>> +		kvm_err("kvm_arch_timer: no split EOI/deactivate; abort\n");
>>> +		return false;
>>> +	}
>>> +
>>> +	return true;
>>> +}
>>
>> That feels really involved. How about reporting that we don't have a
>> usable VGIC altogether from the GIC driver?
>>
> 
> You mean not booting the kernel at all, or establish some communication
> from the GIC driver to KVM, telling KVM if it's usable for
> virtualization or not?

We already query the vgic configuration by calling gic_get_kvm_info() at
boot time. My suggestion is to make it return NULL if the GIC can't
support split EOI/deactivate, and fallback to the "no-vgic" mode, if
ever implemented.

Thoughts?

	M.
-- 
Jazz is not dead. It just smells funny...