[PATCH] powerpc: check_and_cede_processor never cedes

Commit f948501b36c6 ("Make hard_irq_disable() actually hard-disable
interrupts") caused check_and_cede_processor to stop working.
->irq_happened will never be zero right after a hard_irq_disable
so the compiler removes the call to cede_processor completely.

The bug was introduced back in the lazy interrupt handling rework
of 3.4 but was hidden until recently because hard_irq_disable did
nothing.

This issue will eventually appear in 3.4 stable since the
hard_irq_disable fix is marked stable, so mark this one for stable
too.

Signed-off-by: Anton Blanchard <redacted>  
Cc: stable@vger.kernel.org
---

Index: linux-build/arch/powerpc/platforms/pseries/processor_idle.c
===================================================================

--- linux-build.orig/arch/powerpc/platforms/pseries/processor_idle.c	2012-06-27 21:20:45.403761715 +1000
+++ linux-build/arch/powerpc/platforms/pseries/processor_idle.c	2012-06-27 21:57:14.796788823 +1000

@@ -106,7 +106,7 @@ static void check_and_cede_processor(voi
 	 * we first hard disable then check.
 	 */
 	hard_irq_disable();
-	if (get_paca()->irq_happened == 0)
+	if (get_paca()->irq_happened == PACA_IRQ_HARD_DIS)
 		cede_processor();
 }
 

On Thu, 2012-06-28 at 08:45 +1000, Anton Blanchard wrote:

I'd rather add a helper, something like lazy_irq_pending()
and hide the actual check for the bits in irq_happened, in
case we change the scheme again.

Something like:

static inline bool lazy_irq_pending(void)
{
	return !!(get_paca()->irq_happened & ~PACA_IRQ_HARD_DIS);
}

Cheers,
Ben.

Hi,

I'd rather add a helper, something like lazy_irq_pending()
and hide the actual check for the bits in irq_happened, in
case we change the scheme again.

Good idea. Look ok?

--

Commit f948501b36c6 ("Make hard_irq_disable() actually hard-disable
interrupts") caused check_and_cede_processor to stop working.
->irq_happened will never be zero right after a hard_irq_disable
so the compiler removes the call to cede_processor completely.

The bug was introduced back in the lazy interrupt handling rework
of 3.4 but was hidden until recently because hard_irq_disable did
nothing.

This issue will eventually appear in 3.4 stable since the
hard_irq_disable fix is marked stable, so mark this one for stable
too.

Signed-off-by: Anton Blanchard <redacted>  
Cc: stable@vger.kernel.org
---

v2: create a helper, suggested by Ben.

Index: linux-build/arch/powerpc/platforms/pseries/processor_idle.c
===================================================================

--- linux-build.orig/arch/powerpc/platforms/pseries/processor_idle.c	2012-06-28 08:55:09.422198154 +1000
+++ linux-build/arch/powerpc/platforms/pseries/processor_idle.c	2012-06-28 08:57:36.112591023 +1000

@@ -106,7 +106,7 @@ static void check_and_cede_processor(voi
 	 * we first hard disable then check.
 	 */
 	hard_irq_disable();
-	if (get_paca()->irq_happened == 0)
+	if (!lazy_irq_pending())
 		cede_processor();
 }
 

Index: linux-build/arch/powerpc/include/asm/hw_irq.h
===================================================================

--- linux-build.orig/arch/powerpc/include/asm/hw_irq.h	2012-06-21 09:16:26.265354429 +1000
+++ linux-build/arch/powerpc/include/asm/hw_irq.h	2012-06-28 08:59:22.082320381 +1000

@@ -103,6 +103,11 @@ static inline void hard_irq_disable(void
 /* include/linux/interrupt.h needs hard_irq_disable to be a macro */
 #define hard_irq_disable	hard_irq_disable
 
+static inline bool lazy_irq_pending(void)
+{
+	return !!(get_paca()->irq_happened & ~PACA_IRQ_HARD_DIS);
+}
+
 /*
  * This is called by asynchronous interrupts to conditionally
  * re-enable hard interrupts when soft-disabled after having

On Thu, 2012-06-28 at 09:13 +1000, Anton Blanchard wrote:

Hi,

quoted

I'd rather add a helper, something like lazy_irq_pending()
and hide the actual check for the bits in irq_happened, in
case we change the scheme again.

Good idea. Look ok?

--

Commit f948501b36c6 ("Make hard_irq_disable() actually hard-disable
interrupts") caused check_and_cede_processor to stop working.
->irq_happened will never be zero right after a hard_irq_disable
so the compiler removes the call to cede_processor completely.

The bug was introduced back in the lazy interrupt handling rework
of 3.4 but was hidden until recently because hard_irq_disable did
nothing.

This issue will eventually appear in 3.4 stable since the
hard_irq_disable fix is marked stable, so mark this one for stable
too.

Yup, looks good, I'll send to Linus tomorrow.

Cheers,
Ben.