[PATCH] powerpc/mpc52xx: Properly update irq_desc when set_type() is called.

From: Grant Likely <redacted>

The MPC5200 PIC driver doesn't correctly update the .status field of
the irq_desc structure when the set_type hook is called.  This patch
adds the required code.

Also cleans up the external IRQ typename field to be something easier
to read (very minor).

Signed-off-by: Grant Likely <redacted>
---

 arch/powerpc/platforms/52xx/mpc52xx_pic.c |    8 +++++++-
 1 files changed, 7 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/platforms/52xx/mpc52xx_pic.c b/arch/powerpc/platforms/52xx/mpc52xx_pic.c
index 72865e8..0a093f0 100644
--- a/arch/powerpc/platforms/52xx/mpc52xx_pic.c
+++ b/arch/powerpc/platforms/52xx/mpc52xx_pic.c

@@ -196,6 +196,7 @@ static void mpc52xx_extirq_ack(unsigned int virq)
 
 static int mpc52xx_extirq_set_type(unsigned int virq, unsigned int flow_type)
 {
+	struct irq_desc *desc = get_irq_desc(virq);
 	u32 ctrl_reg, type;
 	int irq;
 	int l2irq;

@@ -222,6 +223,11 @@ static int mpc52xx_extirq_set_type(unsigned int virq, unsigned int flow_type)
 		type = 0;
 	}
 
+	desc->status &= ~(IRQ_TYPE_SENSE_MASK | IRQ_LEVEL);
+	desc->status |= flow_type & IRQ_TYPE_SENSE_MASK;
+	if (flow_type & (IRQ_TYPE_LEVEL_HIGH | IRQ_TYPE_LEVEL_LOW))
+		desc->status |= IRQ_LEVEL;
+
 	ctrl_reg = in_be32(&intr->ctrl);
 	ctrl_reg &= ~(0x3 << (22 - (l2irq * 2)));
 	ctrl_reg |= (type << (22 - (l2irq * 2)));

@@ -231,7 +237,7 @@ static int mpc52xx_extirq_set_type(unsigned int virq, unsigned int flow_type)
 }
 
 static struct irq_chip mpc52xx_extirq_irqchip = {
-	.typename = " MPC52xx IRQ[0-3] ",
+	.typename = "MPC52xx External",
 	.mask = mpc52xx_extirq_mask,
 	.unmask = mpc52xx_extirq_unmask,
 	.ack = mpc52xx_extirq_ack,

Grant Likely wrote:

From: Grant Likely <redacted>

The MPC5200 PIC driver doesn't correctly update the .status field of
the irq_desc structure when the set_type hook is called.  This patch
adds the required code.

Also cleans up the external IRQ typename field to be something easier
to read (very minor).

Signed-off-by: Grant Likely <redacted>

Hi Grant,

What does this affect?

-- Matt

On Wed, Jan 7, 2009 at 8:51 PM, Matt Sealey [off-list ref] wrote:

Grant Likely wrote:

quoted

From: Grant Likely <redacted>

The MPC5200 PIC driver doesn't correctly update the .status field of
the irq_desc structure when the set_type hook is called.  This patch
adds the required code.

Also cleans up the external IRQ typename field to be something easier
to read (very minor).

Signed-off-by: Grant Likely <redacted>

Hi Grant,

What does this affect?

cat /proc/interrupts

g.

-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

Grant Likely wrote:

On Wed, Jan 7, 2009 at 8:51 PM, Matt Sealey [off-list ref] wrote:

quoted

Grant Likely wrote:

quoted

From: Grant Likely <redacted>

The MPC5200 PIC driver doesn't correctly update the .status field of
the irq_desc structure when the set_type hook is called.  This patch
adds the required code.

Also cleans up the external IRQ typename field to be something easier
to read (very minor).

Signed-off-by: Grant Likely <redacted>

Hi Grant,

What does this affect?

cat /proc/interrupts

Ah so, completely cosmetic in the end. I was worried I'd have to build a 
new kernel for a second.. :D

-- Matt

On Thu, Jan 8, 2009 at 7:40 AM, Matt Sealey [off-list ref] wrote:

Grant Likely wrote:

quoted

On Wed, Jan 7, 2009 at 8:51 PM, Matt Sealey [off-list ref] wrote:

quoted

Grant Likely wrote:

quoted

From: Grant Likely <redacted>

The MPC5200 PIC driver doesn't correctly update the .status field of
the irq_desc structure when the set_type hook is called.  This patch
adds the required code.

Also cleans up the external IRQ typename field to be something easier
to read (very minor).

Signed-off-by: Grant Likely <redacted>

Hi Grant,

What does this affect?

cat /proc/interrupts

Ah so, completely cosmetic in the end. I was worried I'd have to build a new
kernel for a second.. :D

Not entirely; it affects the details about how irqs are handled.
check_irq_resend() for example.  I don't think any 5200 platforms
currently care; but it could be an issue with cascaded IRQ handlers.

g.

-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

Grant Likely wrote:

Not entirely; it affects the details about how irqs are handled.
check_irq_resend() for example.  I don't think any 5200 platforms
currently care; but it could be an issue with cascaded IRQ handlers.

I'm not sure if that would be a problem or not. Would it shed some light 
on why Efika ATA DMA doesn't work? :)

-- Matt Sealey