Re: [Qemu-devel] [PATCH v2 10/14] pc: kvm_apic: pass APIC ID depending on xAPIC/x2APIC mode

[Igor Mammedov]

On Thu, 22 Sep 2016 21:57:39 +0200
Radim Krčmář <address@hidden> wrote:

> 2016-09-22 16:36+0200, Paolo Bonzini:
> > On 22/09/2016 14:50, Igor Mammedov wrote:  
> >> +#ifdef KVM_CAP_X2APIC_API
> >> +    if (kvm_check_extension(s, KVM_CAP_X2APIC_API)) {
> >> +        has_x2apic_ids = !kvm_vm_enable_cap(s, KVM_CAP_X2APIC_API, 0,
> >> +                                            KVM_X2APIC_API_USE_32BIT_IDS);
> >> +    }
> >> +#endif
> >> +  
> > 
> > Radim, whose patches are going to set
> > KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK?  
> 
> I added kvm_enable_x2apic() helper for intel_iommu that enables both,
> because we really want to make sure that both are enabled before
> allowing EIM.  (And then I didn't post those patches ... ameding that
> after a rebase and a quick retest.)
I can include those patches along with this series

 
> We'd better forbid APIC IDs above 255 without "intel_iommu,eim=on", so
> reusing kvm_enable_x2apic() and enabling both in Igor's patches would be
> just a bit nicer.
Is it possible on real hw to start system with APIC IDs above 255 but
with EIM or even whole IOMMU disabled?


Btw:
Are EIM patches merged, current master says:
 -device intel-iommu,intremap=on,eim=on: Property '.eim' not found

we can do eim check at pc_machine_done() time and as well check for
correct(supported) IOMMU type /split/.
Maybe it should be part of EIM patches.


> Having separate KVM_X2APIC_API_USE_32BIT_IDS and
> KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK isn't as useful as I thought it
> would be ...