On 08/01/16 08:45, Thomas Gleixner wrote:

On Thu, 7 Jan 2016, Quan Nguyen wrote:

quoted

-static int apm_gpio_sb_to_irq(struct gpio_chip *gc, u32 gpio)
+static void xgene_gpio_sb_irq_ack(struct irq_data *d)
+{
+	struct irq_data *irqdata;
+	struct xgene_gpio_sb *priv = irq_data_get_irq_chip_data(d);
+
+	irqdata = irq_get_irq_data(priv->gic_virq[d->hwirq]);
+	if (!irqdata || !irqdata->chip)
+		return;

What the heck is this? Why are you looking up some random other irq and fiddle
with its irq chip?

This is a 1:1 mapping from your gpio irq to a gic irq. We have hierarchical
interrupt domains for this.

I've said the exact same thing back in October:

http://lists.infradead.org/pipermail/linux-arm-kernel/2015-October/381996.html

but obviously I wasn't clear enough...

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