2022-04-15 15:50:10

by Vincent Guittot

[permalink] [raw]
Subject: Re: [PATCH V1 RESEND]: sched/fair: Revise comment about lb decision matrix

On Fri, 15 Apr 2022 at 04:22, Tao Zhou <[email protected]> wrote:
>
> If busiest group type is group_misfit_task, the local
> group type must be group_has_spare according to below
> code in update_sd_pick_busiest():
>
> if (sgs->group_type == group_misfit_task &&
> (!capacity_greater(capacity_of(env->dst_cpu), sg->sgc->max_capacity) ||
> sds->local_stat.group_type != group_has_spare))
> return false;
>
> group type imbalanced and overloaded and fully_busy are filtered in here.
> misfit and asym are filtered before in update_sg_lb_stats().

The behavior has changed with
cad68e552e77 ("sched/fair: Consider misfit tasks when load-balancing")
But we haven't updated the matrix

Reviewed-by: Vincent Guittot <[email protected]>


> So, change the decision matrix to:
>
> busiest \ local has_spare fully_busy misfit asym imbalanced overloaded
> has_spare nr_idle balanced N/A N/A balanced balanced
> fully_busy nr_idle nr_idle N/A N/A balanced balanced
> misfit_task force N/A N/A N/A *N/A* *N/A*
> asym_packing force force N/A N/A force force
> imbalanced force force N/A N/A force force
> overloaded force force N/A N/A force avg_load
>
> Fixes: 0b0695f2b34a ("sched/fair: Rework load_balance()")
> Reviewed-by: Dietmar Eggemann <[email protected]>
> Signed-off-by: Tao Zhou <[email protected]>
> ---
>
> v1 changelog:
> (1) Send to MAINTAINERS/REVIEWS get from `./scripts/get_maintainer.pl`
> suggested by Dietmar Eggemann.
> (2) Based on v5.18-rc2
>
>
> kernel/sched/fair.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index d4bd299d67ab..b8c67e7c8cf3 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -9495,7 +9495,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
> * busiest \ local has_spare fully_busy misfit asym imbalanced overloaded
> * has_spare nr_idle balanced N/A N/A balanced balanced
> * fully_busy nr_idle nr_idle N/A N/A balanced balanced
> - * misfit_task force N/A N/A N/A force force
> + * misfit_task force N/A N/A N/A N/A N/A
> * asym_packing force force N/A N/A force force
> * imbalanced force force N/A N/A force force
> * overloaded force force N/A N/A force avg_load
> --
> 2.35.2
>


2022-04-16 01:25:26

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH V1 RESEND]: sched/fair: Revise comment about lb decision matrix

On Fri, Apr 15, 2022 at 10:32:09AM +0200, Vincent Guittot wrote:
> On Fri, 15 Apr 2022 at 04:22, Tao Zhou <[email protected]> wrote:
> >
> > If busiest group type is group_misfit_task, the local
> > group type must be group_has_spare according to below
> > code in update_sd_pick_busiest():
> >
> > if (sgs->group_type == group_misfit_task &&
> > (!capacity_greater(capacity_of(env->dst_cpu), sg->sgc->max_capacity) ||
> > sds->local_stat.group_type != group_has_spare))
> > return false;
> >
> > group type imbalanced and overloaded and fully_busy are filtered in here.
> > misfit and asym are filtered before in update_sg_lb_stats().
>
> The behavior has changed with
> cad68e552e77 ("sched/fair: Consider misfit tasks when load-balancing")
> But we haven't updated the matrix
>
> Reviewed-by: Vincent Guittot <[email protected]>

Thanks (I took the later version and stuck your Rb on)!

2022-04-16 01:58:27

by Tao Zhou

[permalink] [raw]
Subject: Re: [PATCH V1 RESEND]: sched/fair: Revise comment about lb decision matrix

On Fri, Apr 15, 2022 at 10:32:09AM +0200,
Vincent Guittot wrote:

> On Fri, 15 Apr 2022 at 04:22, Tao Zhou <[email protected]> wrote:
> >
> > If busiest group type is group_misfit_task, the local
> > group type must be group_has_spare according to below
> > code in update_sd_pick_busiest():
> >
> > if (sgs->group_type == group_misfit_task &&
> > (!capacity_greater(capacity_of(env->dst_cpu), sg->sgc->max_capacity) ||
> > sds->local_stat.group_type != group_has_spare))
> > return false;
> >
> > group type imbalanced and overloaded and fully_busy are filtered in here.
> > misfit and asym are filtered before in update_sg_lb_stats().
>
> The behavior has changed with
> cad68e552e77 ("sched/fair: Consider misfit tasks when load-balancing")
> But we haven't updated the matrix
>
> Reviewed-by: Vincent Guittot <[email protected]>

I send a version just now but forget to change the version number.

Thanks,
Tao

> > So, change the decision matrix to:
> >
> > busiest \ local has_spare fully_busy misfit asym imbalanced overloaded
> > has_spare nr_idle balanced N/A N/A balanced balanced
> > fully_busy nr_idle nr_idle N/A N/A balanced balanced
> > misfit_task force N/A N/A N/A *N/A* *N/A*
> > asym_packing force force N/A N/A force force
> > imbalanced force force N/A N/A force force
> > overloaded force force N/A N/A force avg_load
> >
> > Fixes: 0b0695f2b34a ("sched/fair: Rework load_balance()")
> > Reviewed-by: Dietmar Eggemann <[email protected]>
> > Signed-off-by: Tao Zhou <[email protected]>
> > ---
> >
> > v1 changelog:
> > (1) Send to MAINTAINERS/REVIEWS get from `./scripts/get_maintainer.pl`
> > suggested by Dietmar Eggemann.
> > (2) Based on v5.18-rc2
> >
> >
> > kernel/sched/fair.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index d4bd299d67ab..b8c67e7c8cf3 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -9495,7 +9495,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
> > * busiest \ local has_spare fully_busy misfit asym imbalanced overloaded
> > * has_spare nr_idle balanced N/A N/A balanced balanced
> > * fully_busy nr_idle nr_idle N/A N/A balanced balanced
> > - * misfit_task force N/A N/A N/A force force
> > + * misfit_task force N/A N/A N/A N/A N/A
> > * asym_packing force force N/A N/A force force
> > * imbalanced force force N/A N/A force force
> > * overloaded force force N/A N/A force avg_load
> > --
> > 2.35.2
> >