[PATCH] powerpc/powernv: Don't alloc IRQ map if necessary

[PATCH] powerpc/powernv: Don't alloc IRQ map if necessary

[Gavin Shan]

On PowerNV platform, the OPAL interrupts are exported by firmware
through device-node property (/ibm,opal::opal-interrupts). Under
some extreme circumstances (e.g. simulator), we don't have this
property found from the device tree. For that case, we shouldn't
allocate the interrupt map. Otherwise, slab complains allocating
zero sized memory chunk.

Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
---
 arch/powerpc/platforms/powernv/opal.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/opal.c b/arch/powerpc/platforms/powernv/opal.c
index f10b9ec..e61cefe 100644
--- a/arch/powerpc/platforms/powernv/opal.c
+++ b/arch/powerpc/platforms/powernv/opal.c
@@ -726,11 +726,14 @@ static int __init opal_init(void)
 
 	/* Find all OPAL interrupts and request them */
 	irqs = of_get_property(opal_node, "opal-interrupts", &irqlen);
+	opal_irq_count = irqs ? (irqlen / 4) : 0;
+	if (opal_irq_count)
+		opal_irqs = kzalloc(opal_irq_count * sizeof(unsigned int),
+				    GFP_KERNEL);
 	pr_debug("opal: Found %d interrupts reserved for OPAL\n",
-		 irqs ? (irqlen / 4) : 0);
-	opal_irq_count = irqlen / 4;
-	opal_irqs = kzalloc(opal_irq_count * sizeof(unsigned int), GFP_KERNEL);
-	for (i = 0; irqs && i < (irqlen / 4); i++, irqs++) {
+		 opal_irq_count);
+
+	for (i = 0; opal_irqs && i < opal_irq_count; i++, irqs++) {
 		unsigned int hwirq = be32_to_cpup(irqs);
 		unsigned int irq = irq_create_mapping(NULL, hwirq);
 		if (irq == NO_IRQ) {
@@ -775,7 +778,7 @@ void opal_shutdown(void)
 	long rc = OPAL_BUSY;
 
 	/* First free interrupts, which will also mask them */
-	for (i = 0; i < opal_irq_count; i++) {
+	for (i = 0; opal_irqs && i < opal_irq_count; i++) {
 		if (opal_irqs[i])
 			free_irq(opal_irqs[i], NULL);
 		opal_irqs[i] = 0;
-- 
1.8.3.2

Re: powerpc/powernv: Don't alloc IRQ map if necessary

[Michael Ellerman]

On Thu, 2015-08-01 at 05:42:32 UTC, Gavin Shan wrote:
> On PowerNV platform, the OPAL interrupts are exported by firmware
> through device-node property (/ibm,opal::opal-interrupts). Under
> some extreme circumstances (e.g. simulator), we don't have this
> property found from the device tree. For that case, we shouldn't
> allocate the interrupt map. Otherwise, slab complains allocating
> zero sized memory chunk.
 
That sounds fine. But the implementation is a bit messy.

Can you do a precursor patch which moves the logic into opal_irq_init(), that
way when you find no interrupts you can just return.

cheers

Re: powerpc/powernv: Don't alloc IRQ map if necessary

[Gavin Shan]

On Thu, Jan 22, 2015 at 04:54:59PM +1100, Michael Ellerman wrote:
>On Thu, 2015-08-01 at 05:42:32 UTC, Gavin Shan wrote:
>> On PowerNV platform, the OPAL interrupts are exported by firmware
>> through device-node property (/ibm,opal::opal-interrupts). Under
>> some extreme circumstances (e.g. simulator), we don't have this
>> property found from the device tree. For that case, we shouldn't
>> allocate the interrupt map. Otherwise, slab complains allocating
>> zero sized memory chunk.
>
>That sounds fine. But the implementation is a bit messy.
>
>Can you do a precursor patch which moves the logic into opal_irq_init(), that
>way when you find no interrupts you can just return.
>

Sure, I'll change the code accordingly and repost.

Thanks,
Gavin

>cheers
>