On Mon, Jul 23, 2012 at 12:32 AM, Benjamin Herrenschmidt
[off-list ref] wrote:

Allright, another one Grant:

unsigned int irq_find_mapping(struct irq_domain *domain,
                              irq_hw_number_t hwirq)
{
        struct irq_data *data;

        /* Look for default domain if nececssary */
        if (domain == NULL)
                domain = irq_default_domain;
        if (domain == NULL)
                return 0;

        switch (domain->revmap_type) {
        case IRQ_DOMAIN_MAP_LEGACY:
                return irq_domain_legacy_revmap(domain, hwirq);
        case IRQ_DOMAIN_MAP_LINEAR:
                return irq_linear_revmap(domain, hwirq);
        case IRQ_DOMAIN_MAP_TREE:
                rcu_read_lock();
                data = radix_tree_lookup(&domain->revmap_data.tree, hwirq);
                rcu_read_unlock();
                if (data)
                        return data->irq;
-               break;
+               return 0;
        case IRQ_DOMAIN_MAP_NOMAP:

Please, stick a proper commit message and my s-o-b and see if you can fix
your tree before you ask Linus to pull because that's not pretty on any
pseries .... irq_find_mapping() does get called for all interrupt the
first time it's mapped to check if there's a pre-existing mapping, so
the case of the thing being unpopulated is absolutely legit.

the NOMAP case has a similar dubious exit case but since I'm not that
familiar with NOMAP I haven't touched it.

I've decided to rework the patch to simply omit the WARN statement. It
isn't really needed. I've merged in Linus' tree below the eliminate
slow-path patch and remove the WARN statement. It's been pushed out to
my irqdomain/next branch, so it should show up in tomorrow's
linux-next.

You can find it here if you want to give it a spin:

git://git.secretlab.ca/git/linux-2.6 irqdomain/next

I'll give it a bit more time in linux-next before I ask Linus to pull.

g.