On 13/10/14 10:27, Arnd Bergmann wrote:

On Monday 13 October 2014 09:56:20 Marc Zyngier wrote:

quoted

        if (irq_domain_is_hierarchy(domain)) {
+               if (domain->ops->xlate) {
+                       /*
+                        * If we've already configured this interrupt,
+                        * don't do it again, or hell will break loose.
+                        */
+                       if (domain->ops->xlate(domain, irq_data->np,
+                                              irq_data->args,
+                                              irq_data->args_count,
+                                              &hwirq, &type))
+                               return 0;
+
+                       virq = irq_find_mapping(domain, hwirq);
+                       if (virq)
+                               return virq;
+               }
                virq = irq_domain_alloc_irqs(domain, 1, NUMA_NO_NODE, irq_data);
                return virq <= 0 ? 0 : virq;
        }

Thoughts?

Using irq_find_mapping() first is probably the right approach, that
is what irq_create_mapping() does too, and I suppose we want those
to be symmetric.

Ah, good point. I somehow missed that.

mt_sysirq_domain_alloc() in patch 4 has the irq_find_domain check
in it, which I guess we can remove when it has moved to the common
code.

I don't see irq_domain_alloc_irqs() in linux-next or older kernels, where
does that get introduced?

This is part of Jiang's domain hierarchy series:
https://patchwork.ozlabs.org/patch/388279/

which I plan to use to get rid of the ugly gic_extn hack that only Tegra
uses (but that everyone tries to abuse), and also for the GICv2m support.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...