Re: [PATCH RFC] xen: arm: context switch vtimer PPI state.

[Ian Campbell <ian.campbell@citrix.com>]

On Tue, 2015-03-03 at 11:38 +0000, Stefano Stabellini wrote:
> >      gic_set_irq_properties(desc, cpumask_of(smp_processor_id()), GIC_PRI_IRQ);
> >  
> > -    /* Use vcpu0 to retrieve the pending_irq struct. Given that we only
> > -     * route SPIs to guests, it doesn't make any difference. */
> > -    p = irq_to_pending(d->vcpu[0], desc->irq);
> > -    p->desc = desc;
> > +    if ( d )
> > +    {
> > +        struct pending_irq *p;
> > +
> > +        /* Use vcpu0 to retrieve the pending_irq struct. Given that we only
> > +         * route SPIs to guests, it doesn't make any difference. */
> > +        p = irq_to_pending(d->vcpu[0], desc->irq);
> > +        p->desc = desc;
> > +    }
> 
> I think you need to assign p->desc = desc in this case too, otherwise
> p->desc == NULL checks in gic.c and related code will succeed (and we
> want them to fail as this is an hardware interrupt). We should be able
> to use GICH_V2_LR_HW for it, right?
> 
> If you don't set p->desc, it is going to be treated as a virtual irq
> with no corresponding physical irq.

How can we lookup p without a d?

> In fact I am not sure how the code can work as is given that
> desc->handler is set to gic_hw_ops->gic_guest_irq_type but desc = NULL:
> who is doing the EOI of the physical interrupt? Who is calling
> gicv2_dir_irq?

p->desc is set in gicv2_restore_hwppi for interrupts such as these, so
that it always correctly points to the running domain.

Ian.