[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <f1c5aa1d3d7ce4b09052eb373210cb1cfd653af7.camel@surriel.com>
Date: Mon, 19 Apr 2021 12:46:10 -0400
From: Rik van Riel <riel@...riel.com>
To: Valentin Schneider <valentin.schneider@....com>,
linux-kernel@...r.kernel.org
Cc: kernel-team@...com, Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...nel.org>,
Vincent Guittot <vincent.guittot@...aro.org>,
Dietmar Eggemann <dietmar.eggemann@....com>,
Mel Gorman <mgorman@...e.de>
Subject: Re: [PATCH] sched,fair: skip newidle_balance if a wakeup is pending
On Mon, 2021-04-19 at 12:22 +0100, Valentin Schneider wrote:
> On 18/04/21 22:17, Rik van Riel wrote:
> > @@ -10661,7 +10669,8 @@ static int newidle_balance(struct rq
> > *this_rq, struct rq_flags *rf)
> > * Stop searching for tasks to pull if there are
> > * now runnable tasks on this rq.
> > */
> > - if (pulled_task || this_rq->nr_running > 0)
> > + if (pulled_task || this_rq->nr_running > 0 ||
> > + this_rq->ttwu_pending)
> > break;
>
> I thought newidle_balance() would already handle this by checking
> idle_cpu(), but that can't work due to rq->curr never being rq->idle
> here
> (we're trying very hard to prevent this!).
>
> Would there be any point in bunching up these two checks from
> idle_cpu()
> into an inline helper and reusing it here?
I'm not sure, all the sched classes seem to have their own
magic in the balance functions, and there are enough special
situations out there that we might only be able to consolidate
a few callsites, not the rest.
Also, it turns out newidle_balance needs a different return
value depending on whether we have ->ttwu_pending, a pulled
task, or a queued realtime task...
I'll send v2 without any considation here, since I cannot
figure out a way to make things better here :)
--
All Rights Reversed.
Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)
Powered by blists - more mailing lists