Re: [PATCH 2/3] sched/fair: Correctly insert cfs_rq's to list on unthrottle

From: Odin Ugedal <odin@uged.al>
To: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Odin Ugedal <odin@uged.al>, Ingo Molnar <mingo@redhat.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Juri Lelli <juri.lelli@redhat.com>,
	Dietmar Eggemann <dietmar.eggemann@arm.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
	Daniel Bristot de Oliveira <bristot@redhat.com>,
	"open list:CONTROL GROUP (CGROUP)" <cgroups@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 2/3] sched/fair: Correctly insert cfs_rq's to list on unthrottle
Date: Sat, 29 May 2021 11:33:32 +0200	[thread overview]
Message-ID: <CAFpoUr04ziEzvNBJx0xKSuuEnapGzyABwaM-FU3TUaCZkQ4WPw@mail.gmail.com> (raw)
In-Reply-To: <CAKfTPtA6AyL2f-KqHXecZrYKmZ9r9mT=Ks6BeNLjV9dfbSZJxQ@mail.gmail.com>

Hi,

> normally tg_load_avg_contrib should be null when cfs_rq_is_decayed()

Yeah, I think that is an ok assumption of how it _should_ work (given
the other patches in flight are merged).

> The reason of this list is to ensure that the load of all cfs_rq are
> periodically updated  as it is then used to share the runtime between
> groups so we should keep to use the rule whenever possible.

Yeah, right.

> we probably need to keep (cfs_rq->nr_running >= 1) as we can have case
> where tg_load_avg_contrib is null but a task is enqueued

Yeah, there is probably a chance of enqueuing a task without any load,
and then a parent gets throttled.
So (cfs_rq->tg_load_avg_contrib || cfs_rq->nr_running) is probably the
way to go if we want to avoid
a new field. Will resend a patch with that instead.

In case the new field is the main issue with the original solution, we
could also change the on_list int to have three modes like; NO, YES,
THROTTLED/PAUSED, but that would require a bigger rewrite of the other
logic, so probably outside the scope of this patch.

Thanks
Odin