Re: [patch 2/2] round_jiffies users

[Arjan van de Ven <arjan@linux.intel.com>]

On Tue, 2006-10-10 at 18:47 +0200, Ingo Oeser wrote:
> Hi Arjan,
> 
> Arjan van de Ven wrote:
> > Index: linux-2.6.19-rc1-git6/mm/slab.c
> > ===================================================================
> > --- linux-2.6.19-rc1-git6.orig/mm/slab.c
> > +++ linux-2.6.19-rc1-git6/mm/slab.c
> > @@ -926,7 +926,7 @@ static void __devinit start_cpu_timer(in
> >  	if (keventd_up() && reap_work->func == NULL) {
> >  		init_reap_node(cpu);
> >  		INIT_WORK(reap_work, cache_reap, NULL);
> > -		schedule_delayed_work_on(cpu, reap_work, HZ + 3 * cpu);
> > +		schedule_delayed_work_on(cpu, reap_work, __round_jiffies_relative(HZ, cpu));
> >  	}
> >  }
> >  
> 
> Did you changed the behavior by intention?
> You seem to miss the factor "3" here. This hunk should read:

Hi,

actually.. not really; the __round_jiffies_relative function just takes
a CPU number, and internally takes care of spreading things around based
on CPU number (eg it does the *3 internally); it's cleaner that way, the
callers don't need to bother by how much to spread for each cpu etc
etc... So the patch is correct as is.


Greetings,
   Arjan van de Ven