Re: [PATCH] sched/core: switch to BUG_ON()

[Nicholas Mc Guire <der.herr@hofr.at>]

On Mon, Mar 27, 2017 at 10:24:40AM +0200, Peter Zijlstra wrote:
> On Mon, Mar 27, 2017 at 09:40:46AM +0200, Nicholas Mc Guire wrote:
> 
> > Found by coccinelle: bugon.cocci
> > ./kernel/sched/core.c:5913:2-5: WARNING: Use BUG_ON instead of if condition followed by BUG.
> 
> Right, so I disagree with that here.
> 
> > @@ -5909,8 +5909,8 @@ void __init sched_init_smp(void)
> >  	mutex_unlock(&sched_domains_mutex);
> >  
> >  	/* Move init over to a non-isolated CPU */
> > -	if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
> > -		BUG();
> > +	BUG_ON(set_cpus_allowed_ptr(current, non_isolated_cpus) < 0);
> > +
> >  	sched_init_granularity();
> >  	free_cpumask_var(non_isolated_cpus);
> 
> If the condition were without side effects, that pattern would hold, but
> here the function call set_cpus_allowed_ptr() is very much not a pure
> function, so wrapping it in a BUG_ON() just fels wrong.

Got it - just looked throug core.c and there it seems only side-effect
free test-functions are wrapped in BUG_ON() so this change is inconsistent.
Other subsystems seem not to follow that logic - might be worth a note
somewhere below Documentation/.

thx!
hofrat