[PATCH 11/14] sched: filter task pull request

[Vincent Guittot]

On 22 May 2013 17:56, Morten Rasmussen <morten.rasmussen at arm.com> wrote:
> On Fri, Apr 26, 2013 at 11:00:58AM +0100, Vincent Guittot wrote:
>> Part of this patch is missing, the fix below is needed
>>
>> @@ -3497,7 +3497,9 @@ static bool is_buddy_full(int cpu)
>>  static bool is_my_buddy(int cpu, int buddy)
>>  {
>>   int my_buddy = per_cpu(sd_pack_buddy, cpu);
>> - return (my_buddy == -1) || (buddy == my_buddy);
>> +
>> + return ((sysctl_sched_packing_mode == SCHED_PACKING_FULL) &&
>> +               ((my_buddy == -1) || (buddy == my_buddy)));
>>  }
>>
>>  static bool is_light_task(struct task_struct *p)
>>
>>
>> On 25 April 2013 19:23, Vincent Guittot <vincent.guittot at linaro.org> wrote:
>> > Only CPUs that participates to the packing effort can pull tasks on a busiest
>> > group.
>> >
>> > Signed-off-by: Vincent Guittot <vincent.guittot at linaro.org>
>> > ---
>> >  kernel/sched/fair.c |   14 ++++++++++++--
>> >  1 file changed, 12 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> > index 28f8ea7..6f63fb9 100644
>> > --- a/kernel/sched/fair.c
>> > +++ b/kernel/sched/fair.c
>> > @@ -3494,6 +3494,12 @@ static bool is_buddy_full(int cpu)
>> >         return (sum * 1024 >= period * 1000);
>> >  }
>> >
>> > +static bool is_my_buddy(int cpu, int buddy)
>> > +{
>> > +       int my_buddy = per_cpu(sd_pack_buddy, cpu);
>> > +       return (my_buddy == -1) || (buddy == my_buddy);
>> > +}
>
> Would it make sense to change the function name to something like
> is_packing_target() and only have one argument?

I have replaced it with is_packing_cpu(int cpu) in my next version.
This function returns true if the cpu is part of the packing effort

>
> is_my_buddy() is only used with the same variable for both arguments
> like below.
>
>> > +
>> >  static bool is_light_task(struct task_struct *p)
>> >  {
>> >         /* A light task runs less than 20% in average */
>> > @@ -4688,8 +4694,8 @@ static inline void update_sg_lb_stats(struct lb_env *env,
>> >
>> >                 /* Bias balancing toward cpus of our domain */
>> >                 if (local_group) {
>> > -                       if (idle_cpu(i) && !first_idle_cpu &&
>> > -                                       cpumask_test_cpu(i, sched_group_mask(group))) {
>> > +                       if (is_my_buddy(i, i) && idle_cpu(i) && !first_idle_cpu
>> > +                        && cpumask_test_cpu(i, sched_group_mask(group))) {
>> >                                 first_idle_cpu = 1;
>> >                                 balance_cpu = i;
>> >                         }
>> > @@ -4817,6 +4823,10 @@ static void update_plb_buddy(int cpu, int *balance, struct sd_lb_stats *sds,
>> >
>> >         /* Get my new buddy */
>> >         buddy = per_cpu(sd_pack_buddy, cpu);
>> > +
>> > +       /* This CPU doesn't act for agressive packing */
>> > +       if (buddy != cpu)
>> > +               sds->busiest = 0;
>
> sds->busiest is a pointer, so I think it should be assigned to NULL
> instead of 0.

yes

Thanks
Vincent

>
> Morten
>
>> >  }
>> >
>> >  /**
>> > --
>> > 1.7.9.5
>> >
>>
>