Re: linux-next: PowerPC WARN_ON_ONCE() after merge of the final tree (tip related)

[Frederic Weisbecker <fweisbec@gmail.com>]

On Thu, Apr 15, 2010 at 04:03:58PM +0200, Ingo Molnar wrote:
> > In this case, I guess the following fix should be sufficient?
> > I'm going to test it and provide a sane changelog.
> > 
> > 
> > diff --git a/kernel/lockdep.c b/kernel/lockdep.c
> > index 78325f8..65d4336 100644
> > --- a/kernel/lockdep.c
> > +++ b/kernel/lockdep.c
> > @@ -2298,7 +2298,11 @@ void trace_hardirqs_on_caller(unsigned long ip)
> >  		return;
> >  
> >  	if (unlikely(curr->hardirqs_enabled)) {
> > +		unsigned long flags;
> > +
> > +		raw_local_irq_save(flags);
> >  		debug_atomic_inc(redundant_hardirqs_on);
> > +		raw_local_irq_restore(flags);
> >  		return;
> >  	}
> >  	/* we'll do an OFF -> ON transition: */
> 
> that looks rather ugly. Why not do a raw:
> 
> 	this_cpu_inc(lockdep_stats.redundant_hardirqs_on);
> 
> which basically open-codes debug_atomic_inc(), but without the warning?


Because that would open a race against interrupts that might
touch lockdep_stats.redundant_hardirqs_on too.

If you think it's not very important (this race must be pretty rare I guess),
I can use your solution.



> 
> Btw., using the this_cpu() methods might result in faster code for all the 
> debug_atomic_inc() macros as well?


Indeed, will change that too.

Thanks.