Re: [PATCH] x86: kvm: reset the bootstrap processor when it gets an INIT

[Gleb Natapov <gleb@redhat.com>]

On Sun, Mar 10, 2013 at 03:53:54PM +0100, Paolo Bonzini wrote:
> Il 10/03/2013 12:46, Gleb Natapov ha scritto:
> > On Sat, Mar 09, 2013 at 07:48:33AM +0100, Paolo Bonzini wrote:
> >> After receiving an INIT signal (either via the local APIC, or through
> >> KVM_SET_MP_STATE), the bootstrap processor should reset immediately
> >> and start execution at 0xfffffff0.  Also, SIPIs have no effect on the
> >> bootstrap processor.  However, KVM currently does not differentiate
> >> between the BSP and APs.
> >>
> > Userspace is capable of resetting vcpu by itself, so adding code to
> > handle KVM_SET_MP_STATE(INIT) looks unnecessary to me. I think the
> > simple patch below (not tested) should handle INIT for in-kernel irq chip
> > and userspace does not need special handling for cpu reset. It already
> > knows how to reset cpu on system_reset, so reseting only cpus should not
> > be different.
> 
> At least you'll need the last two hunks, moving the check for
> KVM_MP_STATE_SIPI_RECEIVED before kvm_vcpu_block, but yes---something
> like this could work.
> 
I do not see why. kvm_vcpu_block() exits if vpu is in
KVM_MP_STATE_SIPI_RECEIVED state.

> However, it would effectively redefine the meaning of
> KVM_MP_STATE_INIT_RECEIVED and KVM_MP_STATE_SIPI_RECEIVED, respectively
> to KVM_MP_STATE_WAIT_FOR_SIPI and KVM_MP_STATE_RESETTING.  I wasn't sure
> if this is considered an API change (personally, I would treat it as one).
> 
If it is kernel module internal it definitely is not API change.
INIT/SIPI handling is a bit ad-hoc right now anyway as Jan noticed. For
instance INIT does not really resets VCPU. Only after SIPI it is really
reset, so KVM_MP_STATE_SIPI_RECEIVED is really KVM_MP_STATE_RESET_ME_RIGHT_NOW
state.

--
			Gleb.