Re: [PATCH 1/2] lockdep: improve current->(hard|soft)irqs_enabled synchronisation with actual irq state

Excerpts from Peter Zijlstra's message of July 24, 2020 12:59 am:

On Thu, Jul 23, 2020 at 11:11:03PM +1000, Nicholas Piggin wrote:

quoted

Excerpts from Peter Zijlstra's message of July 23, 2020 9:40 pm:

quoted

On Thu, Jul 23, 2020 at 08:56:14PM +1000, Nicholas Piggin wrote:

quoted

diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index 3a0db7b0b46e..35060be09073 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h

@@ -200,17 +200,14 @@ static inline bool arch_irqs_disabled(void)
 #define powerpc_local_irq_pmu_save(flags)			\
 	 do {							\
 		raw_local_irq_pmu_save(flags);			\
-		trace_hardirqs_off();				\
+		if (!raw_irqs_disabled_flags(flags))		\
+			trace_hardirqs_off();			\
 	} while(0)
 #define powerpc_local_irq_pmu_restore(flags)			\
 	do {							\
-		if (raw_irqs_disabled_flags(flags)) {		\
-			raw_local_irq_pmu_restore(flags);	\
-			trace_hardirqs_off();			\
-		} else {					\
+		if (!raw_irqs_disabled_flags(flags))		\
 			trace_hardirqs_on();			\
-			raw_local_irq_pmu_restore(flags);	\
-		}						\
+		raw_local_irq_pmu_restore(flags);		\
 	} while(0)

You shouldn't be calling lockdep from NMI context!

After this patch it doesn't.

You sure, trace_hardirqs_{on,off}() calls into lockdep.

At least for irq enable/disable functions yes. NMI runs with
interrupts disabled so these will never call trace_hardirqs_on/off
after this patch.

(FWIW they're
also broken vs entry ordering, but that's another story).

quoted

trace_hardirqs_on/off implementation appears to expect to be called in NMI 
context though, for some reason.

Hurpm, not sure.. I'll have to go grep arch code now :/ The generic NMI
code didn't touch that stuff.

Argh, yes, there might be broken there... damn! I'll go frob around.

Thanks,
Nick