Skip to contentSkip to conversation

powerpc: Make hard_irq_disable() do the right thing vs. irq tracing

PATCH In review Benjamin Herrenschmidt 2 participants 3 messages May 7, 2013 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc: Make hard_irq_disable() do the right thing vs. irq tracing

If hard_irq_disable() is called while interrupts are already soft-disabled (which is the most common case) all is already well.

However you can (and in some cases want) to call it while everything is enabled (to make sure you don't get a lazy even, for example before entry into KVM guests) and in this case we need to inform the irq tracer that the irqs are going off.

We have to change the inline into a macro to avoid an include circular dependency hell hole.

---

Tested on pseries, Scott, I don't expect a problem with that patch especially since most callers already are soft disabled, so I'll merge it now along with my other pending stuff and you can simplify your KVM one.

arch/powerpc/include/asm/hw_irq.h +7 -9
index e45c494..d615b28 100644
@@ -95,15 +95,13 @@ static inline bool arch_irqs_disabled(void)
 #define __hard_irq_disable()	__mtmsrd(local_paca->kernel_msr, 1)
 #endif
 
-static inline void hard_irq_disable(void)
-{
-	__hard_irq_disable();
-	get_paca()->soft_enabled = 0;
-	get_paca()->irq_happened |= PACA_IRQ_HARD_DIS;
-}
-
-/* include/linux/interrupt.h needs hard_irq_disable to be a macro */
-#define hard_irq_disable	hard_irq_disable
+#define hard_irq_disable()	do {			\
+	__hard_irq_disable();				\
+	if (local_paca->soft_enabled)			\
+		trace_hardirqs_off();			\
+	get_paca()->soft_enabled = 0;			\
+	get_paca()->irq_happened |= PACA_IRQ_HARD_DIS;	\
+} while(0)
 
 static inline bool lazy_irq_pending(void)
 {
#1 lore raw
Quoted 30 lines · Benjamin Herrenschmidt

If hard_irq_disable() is called while interrupts are already soft-disabled (which is the most common case) all is already well.

However you can (and in some cases want) to call it while everything is enabled (to make sure you don't get a lazy even, for example before entry into KVM guests) and in this case we need to inform the irq tracer that the irqs are going off.

We have to change the inline into a macro to avoid an include circular dependency hell hole.

---

Tested on pseries, Scott, I don't expect a problem with that patch especially since most callers already are soft disabled, so I'll merge it now along with my other pending stuff and you can simplify your KVM one.

arch/powerpc/include/asm/hw_irq.h +0 -4
index e45c494..d615b28 100644
@@ -95,15 +95,13 @@ static inline bool arch_irqs_disabled(void)
  #define __hard_irq_disable()	__mtmsrd(local_paca->kernel_msr, 1)
  #endif
 
-static inline void hard_irq_disable(void)
-{
-	__hard_irq_disable();
-	get_paca()->soft_enabled = 0;

arch/powerpc/include/asm/hw_irq.h static inline bool arch_irqs_disabled(void) · line 102

Quoted 10 lines
arch/powerpc/include/asm/hw_irq.h +5 -5
-	get_paca()->irq_happened |= PACA_IRQ_HARD_DIS;
-}
-
-/* include/linux/interrupt.h needs hard_irq_disable to be a macro */
-#define hard_irq_disable	hard_irq_disable
+#define hard_irq_disable()	do {			\
+	__hard_irq_disable();				\
+	if (local_paca->soft_enabled)			\
+		trace_hardirqs_off();			\
+	get_paca()->soft_enabled = 0;			\

Could we simplify this as follows:

+#define hard_irq_disable() do { \
+ __hard_irq_disable(); \
+ if (get_paca()->soft_enabled) { \
+ trace_hardirqs_off(); \
+ get_paca()->soft_enabled = 0; \
+ } \
+ get_paca()->irq_happened |= PACA_IRQ_HARD_DIS; \
+} while(0)

Tiejun

Quoted 7 lines
arch/powerpc/include/asm/hw_irq.h +2 -0
+	get_paca()->irq_happened |= PACA_IRQ_HARD_DIS;	\
+} while(0)
 
  static inline bool lazy_irq_pending(void)
  {
#2 lore raw

arch/powerpc/include/asm/hw_irq.h static inline bool arch_irqs_disabled(void) · line 104

Quoted 10 lines · tiejun.chen

Could we simplify this as follows:

arch/powerpc/include/asm/hw_irq.h +8 -0
+#define hard_irq_disable()     do {                    \
+       __hard_irq_disable();                           \
+       if (get_paca()->soft_enabled) {                 \
+               trace_hardirqs_off();                   \
+               get_paca()->soft_enabled = 0;           \
+       }                                               \
+       get_paca()->irq_happened |= PACA_IRQ_HARD_DIS;  \
+} while(0)

I'd rather not. When lockdep is not enabled trace_hardirqs_off() is a nop and thus we have no conditional at all which is arguably better...

Cheers,
Ben.