Re: [Xen-devel] [PATCH 1/2] xen/arm: support the ARM generic virtual timer

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Thu, 14 Feb 2013, Tim Deegan wrote:
> At 11:46 +0000 on 14 Feb (1360842373), Stefano Stabellini wrote:
> > On Thu, 24 Jan 2013, Tim Deegan wrote:
> > > At 16:17 +0000 on 09 Jan (1357748277), Stefano Stabellini wrote:
> > > > +static void vtimer_interrupt(int irq, void *dev_id, struct 
> > > > cpu_user_regs *regs)
> > > > +{
> > > > +    current->arch.virt_timer.ctl = READ_CP32(CNTV_CTL);
> > > > +    WRITE_CP32(current->arch.virt_timer.ctl | CNTx_CTL_MASK, CNTV_CTL);
> > > 
> > > This is masking the vtimer interrupt in a way that's visible to the
> > > currently running guest.  Is that going to confuse the guest?
> > >
> > > When we talked about this before I had imagined that the masking would
> > > happen in the GIC, where the guest doesn't see it. 
> > 
> > I know it is not ideal but it is safe, it is not creating any problems to
> > Linux and it is even the same thing that KVM does).
> 
> Grrr.  So it's safe unless you run something other than linux, or KVM
> devs change their minds...
> 
> Do we have an interface document to describe the PV machine that we
> provide to a guest?  If so, this divergence from the spec should be
> described there.  If not, we really should. :)

That's a good idea

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel