linux-kernel - Re: [PATCH V2 RESEND 3/4] PM / devfreq: bail out early if no freq changes in devfreq_set

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Wed, 19 May 2021 14:51:07 +0800
From:   Dong Aisheng <dongas86@...il.com>
To:     Chanwoo Choi <cwchoi00@...il.com>
Cc:     Dong Aisheng <aisheng.dong@....com>,
        Linux PM <linux-pm@...r.kernel.org>,
        "moderated list:ARM/FREESCALE IMX / MXC ARM ARCHITECTURE" 
        <linux-arm-kernel@...ts.infradead.org>,
        Sascha Hauer <kernel@...gutronix.de>,
        Shawn Guo <shawnguo@...nel.org>,
        dl-linux-imx <linux-imx@....com>,
        open list <linux-kernel@...r.kernel.org>,
        myungjoo.ham@...sung.com, kyungmin.park@...sung.com,
        Chanwoo Choi <cw00.choi@...sung.com>,
        Abel Vesa <abel.vesa@....com>
Subject: Re: [PATCH V2 RESEND 3/4] PM / devfreq: bail out early if no freq
 changes in devfreq_set_target

On Tue, Mar 23, 2021 at 11:01 PM Chanwoo Choi <cwchoi00@...il.com> wrote:
>
> On 21. 3. 23. 오후 4:20, Dong Aisheng wrote:
> > It's unnecessary to set the same freq again and run notifier calls.
> >
> > Signed-off-by: Dong Aisheng <aisheng.dong@....com>
> > ---
> >   drivers/devfreq/devfreq.c | 9 ++++++---
> >   1 file changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> > index 85927bd7ee76..a6ee91dd17bd 100644
> > --- a/drivers/devfreq/devfreq.c
> > +++ b/drivers/devfreq/devfreq.c
> > @@ -352,13 +352,16 @@ static int devfreq_set_target(struct devfreq *devfreq, unsigned long new_freq,
> >   {
> >       struct devfreq_freqs freqs;
> >       unsigned long cur_freq;
> > -     int err = 0;
> > +     int err;
> >
> >       if (devfreq->profile->get_cur_freq)
> >               devfreq->profile->get_cur_freq(devfreq->dev.parent, &cur_freq);
> >       else
> >               cur_freq = devfreq->previous_freq;
> >
> > +     if (new_freq == cur_freq)
> > +             return 0;
>
> cur_freq is one of the OPP frequencies. But, new_freq is calculated from
> governor algorithm. It means that new_freq is not one of the
> frequencies. Actually, it is not efficient.
>
> After devfreq->profile->target() which almost uses
> devfreq_recommended_opp(), new_freq is one of OPP frequencies.
>

Yes, but i feel at least when the desired new_freq is equal to cur_freq
which is the last successfully set rate,  it is sufficient to bail out early as
it's meaningless to re-set the same rate as the last one and notify the
an unchanged rate transition in HW.
Does that make sense?

Regards
Aisheng

> > +
> >       freqs.old = cur_freq;
> >       freqs.new = new_freq;
> >       devfreq_notify_transition(devfreq, &freqs, DEVFREQ_PRECHANGE);
> > @@ -375,7 +378,7 @@ static int devfreq_set_target(struct devfreq *devfreq, unsigned long new_freq,
> >        * and DEVFREQ_POSTCHANGE because for showing the correct frequency
> >        * change order of between devfreq device and passive devfreq device.
> >        */
> > -     if (trace_devfreq_frequency_enabled() && new_freq != cur_freq)
> > +     if (trace_devfreq_frequency_enabled())
> >               trace_devfreq_frequency(devfreq, new_freq, cur_freq);
> >
> >       freqs.new = new_freq;
> > @@ -390,7 +393,7 @@ static int devfreq_set_target(struct devfreq *devfreq, unsigned long new_freq,
> >       if (devfreq->suspend_freq)
> >               devfreq->resume_freq = new_freq;
> >
> > -     return err;
> > +     return 0;
> >   }
> >
> >   /**
> >
>
>
> --
> Best Regards,
> Samsung Electronics
> Chanwoo Choi