From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
        id S1753098AbdKJOxV (ORCPT <rfc822;w@1wt.eu>);
        Fri, 10 Nov 2017 09:53:21 -0500
Received: from foss.arm.com ([217.140.101.70]:60232 "EHLO foss.arm.com"
        rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
        id S1752675AbdKJOxU (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
        Fri, 10 Nov 2017 09:53:20 -0500
References: <20171024122556.15872-1-brendan.jackman@arm.com> <20171024122556.15872-2-brendan.jackman@arm.com> <CAD0t5oPxXc1XARh8puEThJQYfRNFAc2Su0Bj1bDH7_77MVczzA@mail.gmail.com>
User-agent: mu4e 0.9.18; emacs 25.2.2
From: Brendan Jackman <brendan.jackman@arm.com>
To: Todd Kjos <tkjos@android.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Ingo Molnar <mingo@kernel.org>, Peter Zijlstra <peterz@infradead.org>,
        LKML <linux-kernel@vger.kernel.org>, Ingo Molnar <mingo@redhat.com>,
        Morten Rasmussen <morten.rasmussen@arm.com>
Subject: Re: [PATCH 1/2] sched: force update of blocked load of idle cpus
In-reply-to: <CAD0t5oPxXc1XARh8puEThJQYfRNFAc2Su0Bj1bDH7_77MVczzA@mail.gmail.com>
Date: Fri, 10 Nov 2017 14:53:12 +0000
Message-ID: <87k1yydw2v.fsf@arm.com>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org


Hi Todd,

On Thu, Nov 09 2017 at 19:56, Todd Kjos wrote:
>> @@ -8683,6 +8692,10 @@ static void nohz_balancer_kick(void)
>>
>>         if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu)))
>>                 return;
>> +
>> +       if (only_update)
>> +               set_bit(NOHZ_STATS_KICK, nohz_flags(ilb_cpu));
>
> Should there be an "else clear_bit(NOHZ_STATS_KICK, nohz_flags(ilb_cpu));" ?
>
> Seems like any time this is called as !only_update, we should stop
> inhibiting rebalance. In fact, we should perhaps go a little further
> so that an only_update never inhibits rebalance from a concurrent
> !only_update.

Yes, I think you are essentially right. To make sure I understand you: I
guess you are saying that where two CPUs are concurrently in
nohz_balancer_kick, one with only_update=1 and one with only_update=0,
the former should not prevent the latter from triggering a full load
balance. (I'm assuming they both get the same value for ilb_cpu).

The exact solution you described won't work: only one of those CPUs can
get to this line of code, because of the test_and_set_bit above. So I
think we need something like:

       if (test_and_set_bit(NOHZ_BALANCE_KICK, nohz_flags(ilb_cpu))) {
               if (!only_update) {
                      /*
                       * There's a pending stats kick or an ongoing
                       * nohz_idle balance that's just for stats.
                       * Convert it to a proper nohz balance.
                       */
                      clear_bit(NOHZ_STATS_KICK, nohz_flags(ilb_cpu));
               }
               return;
       }

       if (only_update)
               set_bit(NOHZ_STATS_KICK, nohz_flags(ilb_cpu));

There's still scope for some lost rebalance_domains calls, but as
Vincent pointed out in a recent chat, because the rq->next_balance
fields won't be changed for the CPUs they get missed out, those will
only be delayed until the next scheduler_tick.

I'm on holiday ATM, I'll get to testing that when I get back.

Thanks for the review,
Brendan