Re: find_busiest_group divide error

[Dietmar Eggemann <dietmar.eggemann@arm.com>]

Hi Greg,

On 16/07/14 19:52, Greg Donald wrote:
> On Wed, Jul 16, 2014 at 05:27:36PM +0200, Peter Zijlstra wrote:
>> Could you confirm if reverting caffcdd8d27ba78730d5540396ce72ad022aff2c
>> cures things for you?
>>
>> Otherwise there's two very similar issues, see also:
>>
>>     lkml.kernel.org/r/20140716145546.GA6922@wolff.to
>
> Cured.
>
> I reverted caffcdd8d27ba78730d5540396ce72ad022aff2c which did nothing as
> far as I can tell, then I removed the
> two lines from http://marc.info/?l=linux-kernel&m=140552264825755, then
> I added back the one line from
> https://bugzilla.kernel.org/show_bug.cgi?id=80251#c8.

My patch caffcdd8d27ba78730d5540396ce72ad022aff2c got rid of 
cpumask_clear(sched_group_cpus(sg)); and sg->sgp->power = 0; so 
reverting it (and replacing sg->sgp->power = 0 with
sg->sgc->capacity = 0) should cure it too. (although the missing 
cpumask_clear() is the culprit on your machine here).

Could I ask you to share the content of your /proc/cpuinfo file? I 
suspect it might be the same topology as the one Bruno just sent out 
(the one of a dual single core CPU with hyper-threading ?)

https://lkml.org/lkml/2014/7/16/603

Thanks,

-- Dietmar

>
> I ended up with
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index dc2927c..7c3674d 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -5848,7 +5848,6 @@ build_sched_groups(struct sched_domain *sd, int
> cpu)
>
>                  group = get_group(i, sdd, &sg);
>                  cpumask_clear(sched_group_cpus(sg));
> -               sg->sgp->power = 0;
>                  cpumask_setall(sched_group_mask(sg));
>
>                  for_each_cpu(j, span) {
>
>
> Thanks.
>