Re: [PATCH v3 4/5] xen: arm: map child MMIO and IRQs to dom0 for PCI bus DT nodes.

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

On Wed, 2015-04-29 at 17:22 +0100, Julien Grall wrote:
> Hi Ian,
> 
> On 20/04/15 13:16, Ian Campbell wrote:
> > +static int map_interrupt_to_domain(const struct dt_device_node *dev,
> > +                                   const struct dt_raw_irq *dt_raw_irq,
> > +                                   void *data)
> > +{
> 
> [..]
> 
> > +    res = route_irq_to_guest(d, dt_irq.irq, dt_irq.irq, dt_node_name(dev));
> > +    if ( res < 0 )
> > +    {
> > +        printk(XENLOG_ERR "Unable to map IRQ%"PRId32" to dom%d\n",
> > +               dt_irq.irq, d->domain_id);
> > +        return res;
> > +    }
> 
> You forgot to give the IRQ permission to the domain.

Oops! I suppose that would only be noticeable when trying to pass it
through which is why I didn't see it?

> Also I'm wondering if we should care about passthrough the PCI
> controller to a guest. If so, you would need to check need_mapping as we
> do in handle_device.
> 
> If not, I would add a warning if the property "xen,passthrough" exists
> in the node telling that we won't honor it.

My initial reaction was that no good can come from passing through the
host controller itself, and it might have interesting interactions with
the PCI infrastructure to be added in Xen (as described in a thread with
Manish which I can't find right now).

So I think I'll take the coward's way out for now and check for
xen,passthrough.

Ian.