On 06/22/2013 03:52 AM, Scott Wood wrote:
> On 06/21/2013 04:22:55 AM, Alexey Kardashevskiy wrote:
>> Signed-off-by: Alexey Kardashevskiy <a...@ozlabs.ru>
>> ---
>> target-ppc/kvm.c | 2 ++
>> 1 file changed, 2 insertions(+)
>>
>> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
>> index d6da146..e72c335 100644
>> --- a/target-ppc/kvm.c
>> +++ b/target-ppc/kvm.c
>> @@ -1973,6 +1973,8 @@ int kvm_arch_on_sigbus(int code, void *addr)
>>
>> void kvm_arch_init_irq_routing(KVMState *s)
>> {
>> + kvm_irqfds_allowed = true;
>> + kvm_msi_via_irqfd_allowed = true;
>> }
>
> Shouldn't this should go in pseries-specific code, as was done for MPIC?
> The model of doing IRQ-related things per "arch" rather than per IRQ
> subsystem doesn't work well outside of certain architectures such as x86,
> where more than just the ISA is standardized.
Yes, we discussed with Alex Graf in IRC and I'll post another (sorry :) )
set of patches. It will be something like a
kvm_msi_irqfd_enable_direct_mapping flag and using msimessage::data as a
virq in that routing function, should be easy and harmless for others.
--
Alexey
