Re: [PATCH] KVM: nVMX: remove side effects from nested_vmx_exit_reflected

[Sean Christopherson]

On Wed, Mar 18, 2020 at 11:59:13AM +0100, Paolo Bonzini wrote:
> On 18/03/20 11:52, Vitaly Kuznetsov wrote:
> > The only functional difference seems to be that we're now doing
> > nested_mark_vmcs12_pages_dirty() in vmx->fail case too and this seems
> > superfluous: we failed to enter L2 so 'special' pages should remain
> > intact (right?) but this should be an uncommon case.
> > 
> > Reviewed-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
> 
> I'm not entirely sure if the PID could be written before the processor
> decrees a vmfail.  It doesn't really hurt anyway as you say though.

I would expect that writing special pages on VM-Fail would be classified
as a CPU bug.