Thread (1 message) 1 message, 1 author, 2012-01-05

Re: [PATCH v4 1/6] ARM: at91/aic: add irq domain and device tree support

From: Grant Likely <hidden>
Date: 2012-01-05 18:02:18
Also in: linux-arm-kernel

On Thu, Jan 05, 2012 at 05:26:48PM +0100, Nicolas Ferre wrote:
Hi Grant,

Thanks for your review.

Some answer following...

On 01/04/2012 08:40 PM, Grant Likely :

[..]
quoted
quoted
--- a/arch/arm/mach-at91/irq.c
+++ b/arch/arm/mach-at91/irq.c
@@ -24,6 +24,10 @@
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/types.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/irqdomain.h>
+#include <linux/err.h>
 
 #include <mach/hardware.h>
 #include <asm/irq.h>
@@ -34,22 +38,28 @@
 #include <asm/mach/map.h>
 
 void __iomem *at91_aic_base;
+static struct irq_domain at91_aic_domain;
+
+static inline unsigned int aic_irq(struct irq_data *d)
+{
+	return d->hwirq;
+}
This is a useless macro.  Drop it.
Ok. That was overkill ;-)

[..]
quoted
quoted
@@ -90,13 +100,13 @@ static u32 backups;
 
 static int at91_aic_set_wake(struct irq_data *d, unsigned value)
 {
-	if (unlikely(d->irq >= 32))
+	if (unlikely(aic_irq(d) >= at91_aic_domain.nr_irq))
I don't think you need to derefernce the irq_domain.nr_irq here.  Just
keep the test for >= 32.
Well I would like to keep it referring to this value as we may increase
the number of irq handled by this controller in the future...

[..]
quoted
quoted
+#if defined(CONFIG_OF)
+static int __init __at91_aic_of_init(struct device_node *node,
+				     struct device_node *parent)
+{
+	at91_aic_base = of_iomap(node, 0);
+	at91_aic_domain.of_node = of_node_get(node);
+
+	return 0;
+}
+
+static const struct of_device_id aic_ids[] __initconst = {
+	{ .compatible = "atmel,at91rm9200-aic", .data = __at91_aic_of_init },
+	{ /*sentinel*/ }
+};
+
+static void __init at91_aic_of_init(void)
+{
+	of_irq_init(aic_ids);
+}
+#else
+static void __init at91_aic_of_init(void) {}
+#endif
+
 /*
  * Initialize the AIC interrupt controller.
  */
 void __init at91_aic_init(unsigned int priority[NR_AIC_IRQS])
 {
 	unsigned int i;
+	int hwirq, irq;
 
-	at91_aic_base = ioremap(AT91_AIC, 512);
+	if(of_have_populated_dt())
+		at91_aic_of_init();
+	else
+		at91_aic_base = ioremap(AT91_AIC, 512);
 
 	if (!at91_aic_base)
-		panic("Impossible to ioremap AT91_AIC\n");
+		panic("Unable to ioremap AIC registers\n");
+
+	/* Add irq domain for AIC */
+	at91_aic_domain.nr_irq = NR_AIC_IRQS;
+	at91_aic_domain.irq_base = irq_alloc_descs(-1, AIC_BASE_IRQ,
+						at91_aic_domain.nr_irq, 0);
+	if (IS_ERR_VALUE(at91_aic_domain.irq_base)) {
+		WARN(1, "Cannot allocate irq_descs, assuming pre-allocated\n");
+		at91_aic_domain.irq_base = AIC_BASE_IRQ;
+	}
+	at91_aic_domain.ops = &irq_domain_simple_ops;
+	irq_domain_add(&at91_aic_domain);
Nothing special is being done with the irq domain.  Please use
irq_domain_add_simple().  It will make things easier for the next
round of irq_domain patches that I'm getting ready to post.
Fair enough, but I have one question: how do I specify "nr_irq" in the
domain created by irq_domain_add_simple()? Maybe it is not needed?
Fix the irq_domain_add_simple() helper to accept a nr_irq parameter.  :-)

g.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help