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;
From:   Danny Lin <danny@kdrag0n.dev>
To:     "Rafael J . Wysocki" <rjw@rjwysocki.net>
Cc:     Viresh Kumar <viresh.kumar@linaro.org>,
        Matthias Kaehlcke <mka@chromium.org>,
        Douglas Anderson <dianders@chromium.org>,
        linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org,
        Danny Lin <danny@kdrag0n.dev>
Subject: Re: [PATCH] cpufreq: Record stats when fast switching is enabled
Date:   Thu,  2 Jul 2020 18:00:09 -0700
Message-Id: <20200703010009.220363-1-danny@kdrag0n.dev>
In-Reply-To: <3268787.3OZuCagV1k@aspire.rjw.lan>
References: <3268787.3OZuCagV1k@aspire.rjw.lan>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Jan 31, 2019 at 2:14 AM, Rafael J. Wysocki wrote:
> On Thu, Jan 31, 2019 at 11:07 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
> >
> > On 31-01-19, 11:03, Rafael J. Wysocki wrote:
> > > On Thu, Jan 31, 2019 at 9:30 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
> > > >
> > > > The only problem that I can think of (or recall) is that this routine
> > > > also gets called when time_in_state sysfs file is read and that can
> > > > end up taking lock which the scheduler's hotpath will wait for.
> > >
> > > What about the extra locking overhead in the scheduler context?
> >
> > What about using READ_ONCE/WRITE_ONCE here ? Not sure if we really
> > need locking in this particular case.
> 
> If that works, then fine, but ISTR some synchronization issues related to that.

Maybe using READ/WRITE_ONCE for time_in_state is problematic, but is
there any reason why atomics wouldn't work for this? As far as I can
tell, atomics are necessary to protect time_in_state due to its
multi-step add operation, and READ/WRITE_ONCE can be used for last_time
because all operations on it are single-op sets/gets.

I've been using the setup described above on a downstream arm64 4.14
kernel for nearly a year with no issues. I haven't noticed any
significant anomalies in the stats so far. The system in question has 8
CPUs split into 3 cpufreq policies and fast switch is used with the
schedutil governor, so it should be exercising the stats update path
enough.

Sorry for bumping an old thread.