Content-Type: text/plain;
        charset=us-ascii
Subject: Re: [PATCH v3 0/4]   sched/pelt: Relax the sync of *_sum with *_avg
From:   Sachin Sant <sachinp@linux.vnet.ibm.com>
In-Reply-To: <20220111134659.24961-1-vincent.guittot@linaro.org>
Date:   Wed, 12 Jan 2022 16:53:52 +0530
Cc:     Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        rostedt@goodmis.org, bsegall@google.com, mgorman@suse.de,
        bristot@redhat.com, linux-kernel@vger.kernel.org,
        rickyiu@google.com, odin@uged.al, naresh.kamboju@linaro.org
Message-Id: <1D84539C-994D-4156-B6C8-BED33EE387B5@linux.vnet.ibm.com>
References: <20220111134659.24961-1-vincent.guittot@linaro.org>
To:     Vincent Guittot <vincent.guittot@linaro.org>
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: bulk


> On 11-Jan-2022, at 7:16 PM, Vincent Guittot <vincent.guittot@linaro.org> =
wrote:
>=20
> Rick reported performance regressions in bugzilla because of cpu
> frequency being lower than before:
>    https://bugzilla.kernel.org/show_bug.cgi?id=3D215045
>=20
> He bisected the problem to:
> commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consis=
tent")
>=20
> More details are available in commit message of patch 1.
>=20
> This patchset reverts the commit above and adds several checks when
> propagating the changes in the hierarchy to make sure that we still have
> coherent util_avg and util_sum.
>=20
> Dietmar found a simple way to reproduce the WARN fixed by=20
> commit 1c35b07e6d39 ("sched/fair: Ensure _sum and _avg values stay consis=
tent")
> by looping on hackbench in several different sched group levels.
>=20
> This patchset as run on the reproducer with success but it probably needs
> more tests by people who faced the WARN before.
>=20

I ran scheduler regression tests(including cfg_bandwidth) from LTP
for about 6 hours. I did not observe any (new or previously reported)
kernel warn messages.

Based on this test result for ppc64le
Tested-by: Sachin Sant <sachinp@linux.ibm.com>

-Sachin=