[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <86bjosapdb.wl-maz@kernel.org>
Date: Wed, 06 Aug 2025 08:19:44 +0100
From: Marc Zyngier <maz@...nel.org>
To: "Rafael J. Wysocki" <rafael@...nel.org>
Cc: Christian Loehle <christian.loehle@....com>,
Linux PM <linux-pm@...r.kernel.org>,
LKML <linux-kernel@...r.kernel.org>,
Daniel Lezcano <daniel.lezcano@...aro.org>,
Artem Bityutskiy <artem.bityutskiy@...ux.intel.com>,
Aboorva Devarajan <aboorvad@...ux.ibm.com>,
Thomas Gleixner <tglx@...utronix.de>,
Mark Rutland <mark.rutland@....com>
Subject: Re: [RFT][PATCH v1 5/5] cpuidle: menu: Avoid discarding useful information
On Tue, 05 Aug 2025 19:50:21 +0100,
"Rafael J. Wysocki" <rafael@...nel.org> wrote:
[...]
> > > Any chance to try the teo governor on it to see if this problem can
> > > also be observed?
> >
> > Neither ladder nor teo have this issue. The number of broadcast timer
> > IPIs is minimal, and so is the number of interrupts delivered from the
> > backup timer. Only menu exhibits the IPI-hose behaviour on this box
> > (and only this one).
>
> Good to know, thanks!
>
> <shameless plug>Switch over to teo?</shameless plug>
Sure thing. Just start with:
git rm drivers/cpuidle/governors/menu.c
and I'll gladly switch to something else! ;-)
[...]
> The attached patch (completely untested) causes menu to insert an
> "invalid interval" value to the array of recent intervals after the
> idle state selected previously got rejected. It basically should
> prevent get_typical_interval() from returning small values if deeper
> idle states get rejected all the time.
Yup, this does the trick, thanks. When you get to post this, please
add my:
Tested-by: Marc Zyngier <maz@...nel.org>
M.
--
Without deviation from the norm, progress is not possible.
Powered by blists - more mailing lists