2016-07-13 12:15+0200, Paolo Bonzini:
> On 12/07/2016 22:09, Radim Krčmář wrote:
>> +static bool kvm_apic_map_need_slowpath(struct kvm *kvm, struct kvm_lapic
>> **src,
>> + struct kvm_lapic_irq *irq, struct kvm_apic_map *map)
>> +{
>> + if (!map)
>> + return true;
>> +
>> + if (kvm->arch.x2apic_broadcast_quirk_disabled) {
>> + if ((irq->dest_id == APIC_BROADCAST &&
>> + map->mode != KVM_APIC_MODE_X2APIC))
>> + return true;
>> + if (irq->dest_id == X2APIC_BROADCAST)
>> + return true;
>> + } else {
>> + bool x2apic_ipi = src && *src && apic_x2apic_mode(*src);
>> + if (irq->dest_id == (x2apic_ipi ?
>> + X2APIC_BROADCAST : APIC_BROADCAST))
>> + return true;
>> + }
>> +
>> + return false;
>> +}
>
> This isn't the only case where you go through the slowpath. What about
> renaming to kvm_apic_dest_is_broadcast?
Yeah, sorry. I have a patch that adds a check for 0xff in flat lapic
mode (can be handled in the fast path) and used the same name even
before the check ...
