[patch] fix the irq migration

[patch] fix the irq migration

[Siddha, Suresh B]

While debugging some other irq migration issue, observed this issue with
-git kernels. Ingo, please ACK it.
---

With the commit 76d2160147f43f982dfe881404cfde9fd0a9da21, irq_chip
default_disable() became an empty function. And with this change, irq
migration in the case of Edge triggered IO-APIC and MSI interrupts happens
with out masking the irq.

Appended patch fixes by using the mask/unmask handlers of irq_chip directly
while doing the irq migration.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
---

diff --git a/kernel/irq/migration.c b/kernel/irq/migration.c
index 4baa3bb..14ff345 100644
--- a/kernel/irq/migration.c
+++ b/kernel/irq/migration.c
@@ -66,11 +66,11 @@ void move_native_irq(int irq)
 		return;
 
 	if (likely(!(desc->status & IRQ_DISABLED)))
-		desc->chip->disable(irq);
+		desc->chip->mask(irq);
 
 	move_masked_irq(irq);
 
 	if (likely(!(desc->status & IRQ_DISABLED)))
-		desc->chip->enable(irq);
+		desc->chip->unmask(irq);
 }
 

Re: [patch] fix the irq migration

[Eric W. Biederman]

"Siddha, Suresh B" <suresh.b.siddha@intel.com> writes:

> While debugging some other irq migration issue, observed this issue with
> -git kernels. Ingo, please ACK it.
> ---
>
> With the commit 76d2160147f43f982dfe881404cfde9fd0a9da21, irq_chip
> default_disable() became an empty function. And with this change, irq
> migration in the case of Edge triggered IO-APIC and MSI interrupts happens
> with out masking the irq.
>
> Appended patch fixes by using the mask/unmask handlers of irq_chip directly
> while doing the irq migration.

Close.

If the irq is disabled we need to bail out instead of skipping the
masking.

The masking is required for the function to run correctly and
when IRQ_DISABLED is set we don't know enough to do anything.

Once this is fixed we just need to fix the msi case where the
hardware doesn't provide a mask bit.

Eric