On Sun, Mar 10, 2013 at 04:04:39PM +0100, Paolo Bonzini wrote:
> Il 10/03/2013 15:55, Gleb Natapov ha scritto:
> > > > Why not move INIT case from below as is? Vcpu is reset to correct sate
> > > > by QEMU just like during system_reset.
> > >
> > > APs would not be able to receive SIPIs after executing do_cpu_init,
> > > because they would stay in KVM_MP_STATE_RUNNABLE state.
> >
> > If APs are in runnable state after reset with in kernel irq chip we
> > have a bug somewhere.
>
> Here is where we are resetting the processor. After clearing
> CPU_INTERRUPT_INIT, no matter what else we do (such as resetting the
> APIC and CPU), we need to set the mp_state to KVM_MP_STATE_INIT_RECEIVED.
>
> Or if we go with your simpler hypervisor patch, we need to go to either
> KVM_MP_STATE_INIT_RECEIVED for APs (wait for SIPI) or
> KVM_MP_STATE_SIPI_RECEIVED for the BSP (restart running from the reset
> vector).
>
No need for KVM_MP_STATE_SIPI_RECEIVED. Just make it RUNNING. This is similar to
system_reset path, not? UNINIT for AP, RUNNING for BSP.
> > Should AP be able to get SIPI without INIT after trigger of INIT# line?
>
> Yes, the effect is the same for an INIT interrupt and the triggering of
> INIT#.
>
Can you give me SDM pointer?
--
Gleb.
