Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200428050242.17717-1-swood@redhat.com>
In-Reply-To: <20200428050242.17717-1-swood@redhat.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Thu, 30 Apr 2020 14:48:11 +0200
Message-ID: <CAKfTPtDfk-aq7PTVmMgJcZSOupXb0SGHS8Fc2k+qTZTsnfiV=A@mail.gmail.com>
Subject: Re: [RFC PATCH 0/3] newidle_balance() latency mitigation
To:     Scott Wood <swood@redhat.com>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Rik van Riel <riel@surriel.com>,
        Mel Gorman <mgorman@suse.de>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        linux-rt-users <linux-rt-users@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 28 Apr 2020 at 07:02, Scott Wood <swood@redhat.com> wrote:
>
> These patches mitigate latency caused by newidle_balance() on large
> systems, by enabling interrupts when the lock is dropped, and exiting
> early at various points if an RT task is runnable on the current CPU.
>
> When applied to an RT kernel on a 72-core machine (2 threads per core), I
> saw significant reductions in latency as reported by rteval -- from
> over 500us to around 160us with hyperthreading disabled, and from
> over 1400us to around 380us with hyperthreading enabled.

Do you know how each patch contributes to the decrease ? Because patch
3 not only impacts newly idle lb but each and every lb. So most of the
decrease might come from aborting the busy or idle lb at the highest
sched_domai level which scan all cpus and moving newly idle load
balance is not a major part.

>
> This isn't the first time something like this has been tried:
> https://lore.kernel.org/lkml/20121222003019.433916240@goodmis.org/
> That attempt ended up being reverted:
> https://lore.kernel.org/lkml/5122CD9C.9070702@oracle.com/
>
> The problem in that case was the failure to keep BH disabled, and the
> difficulty of fixing that when called from the post_schedule() hook.
> This patchset uses finish_task_switch() to call newidle_balance(), which
> enters in non-atomic context so we have full control over what we disable
> and when.
>
> There was a note at the end about wanting further discussion on the matter --
> does anyone remember if that ever happened and what the conclusion was?
> Are there any other issues with enabling interrupts here and/or moving
> the newidle_balance() call?
>
> Rik van Riel (1):
>   sched,rt: break out of load balancing if an RT task appears
>
> Scott Wood (2):
>   sched/fair: Call newidle_balance() from finish_task_switch()
>   sched/fair: Enable interrupts when dropping lock in newidle_balance()
>
>  kernel/sched/core.c  |  7 +++--
>  kernel/sched/fair.c  | 72 +++++++++++++++++++++++---------------------
>  kernel/sched/sched.h | 12 +++++---
>  3 files changed, 49 insertions(+), 42 deletions(-)
>
> --
> 2.18.2
>