Skip to contentSkip to conversation

powerpc: warn on spurious irq events (but ratelimited)

PATCH In review Paul Gortmaker 2 participants 4 messages Dec 8, 2015 – Dec 22, 2015 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc: warn on spurious irq events (but ratelimited)

It might be worth warning on spurious IRQ events; they might point someone at a bogus DTS value or similar.

But ratelimit them to ensure we aren't too spammy about it.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Michael Ellerman <[email protected]>
---
 arch/powerpc/kernel/irq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
arch/powerpc/kernel/irq.c +4 -2
index 290559df1e8b..5c777e3a4c04 100644
@@ -497,10 +497,12 @@ void __do_irq(struct pt_regs *regs)
 	may_hard_irq_enable();
 
 	/* And finally process it */
-	if (unlikely(irq == NO_IRQ))
+	if (unlikely(irq == NO_IRQ)) {
+		printk_ratelimited(KERN_WARNING "spurious irq on %d\n", irq);
 		__this_cpu_inc(irq_stat.spurious_irqs);
-	else
+	} else {
 		generic_handle_irq(irq);
+	}
 
 	trace_irq_exit(regs);
 
Signature · 1 line
2.6.1
#1 lore raw
Quoted 19 lines · Paul Gortmaker

It might be worth warning on spurious IRQ events; they might point someone at a bogus DTS value or similar.

But ratelimit them to ensure we aren't too spammy about it.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Michael Ellerman <[email protected]>
---
 arch/powerpc/kernel/irq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
arch/powerpc/kernel/irq.c +0 -0
index 290559df1e8b..5c777e3a4c04 100644

arch/powerpc/kernel/irq.c void __do_irq(struct pt_regs *regs) · line 501

Quoted 6 lines
arch/powerpc/kernel/irq.c +2 -1
@@ -497,10 +497,12 @@ void __do_irq(struct pt_regs *regs)
        may_hard_irq_enable();
 
        /* And finally process it */
-       if (unlikely(irq == NO_IRQ))
+       if (unlikely(irq == NO_IRQ)) {
+               printk_ratelimited(KERN_WARNING "spurious irq on %d\n", irq);

If we did decide we cared about this patch, then no need for "irq" above, as we already know it is equal to NO_IRQ.

Paul.
--

Quoted 15 lines
arch/powerpc/kernel/irq.c +2 -2
                __this_cpu_inc(irq_stat.spurious_irqs);
-       else
+       } else {
                generic_handle_irq(irq);
+       }
 
        trace_irq_exit(regs);
 
--

2.6.1

_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://lists.ozlabs.org/listinfo/linuxppc-dev

#2 lore raw
Quoted 23 lines · Paul Gortmaker
Quoted 21 lines · Paul Gortmaker

It might be worth warning on spurious IRQ events; they might point someone at a bogus DTS value or similar.

But ratelimit them to ensure we aren't too spammy about it.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Michael Ellerman <[email protected]>
---
 arch/powerpc/kernel/irq.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
arch/powerpc/kernel/irq.c +0 -0
index 290559df1e8b..5c777e3a4c04 100644
@@ -497,10 +497,12 @@ void __do_irq(struct pt_regs *regs)
        may_hard_irq_enable();
 
Quoted 7 lines

arch/powerpc/kernel/irq.c void __do_irq(struct pt_regs *regs) · line 501

Quoted 4 lines
arch/powerpc/kernel/irq.c +2 -1
        /* And finally process it */
-       if (unlikely(irq == NO_IRQ))
+       if (unlikely(irq == NO_IRQ)) {
+               printk_ratelimited(KERN_WARNING "spurious irq on %d\n", irq);

If we did decide we cared about this patch, then no need for "irq" above, as we already know it is equal to NO_IRQ.

I'm not convinced it's worth printing, we already have a counter in proc.

I think it's more likely to just lead to console spam.

But maybe I'm wrong and you can convince me otherwise :)

cheers

#3 lore raw
Quoted 6 lines · [Re: [PATCH] powerpc: warn on spurious irq events (but ratelimited)] On…
Quoted 5 lines · Paul Gortmaker
Quoted 3 lines · Paul Gortmaker

It might be worth warning on spurious IRQ events; they might point someone at a bogus DTS value or similar.

[...]

Quoted 9 lines
Quoted 3 lines

If we did decide we cared about this patch, then no need for "irq" above, as we already know it is equal to NO_IRQ.

I'm not convinced it's worth printing, we already have a counter in proc.

I think it's more likely to just lead to console spam.

But maybe I'm wrong and you can convince me otherwise :)

I'm not sure I can even convince myself, let alone others. :)

Feel free to bin it.

Paul.
--

Quoted 3 lines

cheers