Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 5 Apr 2018 16:11:27 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     "Rafael J. Wysocki" <rafael@kernel.org>
Cc:     "Rafael J. Wysocki" <rjw@rjwysocki.net>,
        Linux PM <linux-pm@vger.kernel.org>,
        Frederic Weisbecker <fweisbec@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Paul McKenney <paulmck@linux.vnet.ibm.com>,
        Thomas Ilsche <thomas.ilsche@tu-dresden.de>,
        Doug Smythies <dsmythies@telus.net>,
        Rik van Riel <riel@surriel.com>,
        Aubrey Li <aubrey.li@linux.intel.com>,
        Mike Galbraith <mgalbraith@suse.de>,
        LKML <linux-kernel@vger.kernel.org>,
        Len Brown <len.brown@intel.com>
Subject: Re: [PATCH v9 10/10] cpuidle: menu: Avoid selecting shallow states
 with stopped tick
Message-ID: <20180405141127.GS4043@hirez.programming.kicks-ass.net>
References: <1736751.LdhZHb50jq@aspire.rjw.lan>
 <6542020.eHGLEK9V0J@aspire.rjw.lan>
 <20180405124757.GQ4082@hirez.programming.kicks-ass.net>
 <CAJZ5v0h2c5RtUxwF2z3Gjy8O2RSBaD7wN3CxUXrg57jH57_xoA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAJZ5v0h2c5RtUxwF2z3Gjy8O2RSBaD7wN3CxUXrg57jH57_xoA@mail.gmail.com>
User-Agent: Mutt/1.9.3 (2018-01-21)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Apr 05, 2018 at 03:49:32PM +0200, Rafael J. Wysocki wrote:
> On Thu, Apr 5, 2018 at 2:47 PM, Peter Zijlstra <peterz@infradead.org> wrote:
> > On Wed, Apr 04, 2018 at 10:50:36AM +0200, Rafael J. Wysocki wrote:
> >> +     if (tick_nohz_tick_stopped()) {
> >> +             /*
> >> +              * If the tick is already stopped, the cost of possible short
> >> +              * idle duration misprediction is much higher, because the CPU
> >> +              * may be stuck in a shallow idle state for a long time as a
> >> +              * result of it.  In that case say we might mispredict and try
> >> +              * to force the CPU into a state for which we would have stopped
> >> +              * the tick, unless the tick timer is going to expire really
> >> +              * soon anyway.
> >
> > Wait what; the tick was stopped, therefore it _cannot_ expire soon.
> >
> > *confused*
> >
> > Did you mean s/tick/a/ ?
> 
> Yeah, that should be "a timer".

*phew* ok, that makes a lot more sense ;-)

My only concern with this is that we can now be overly pessimistic. The
predictor might know that statistically it's very likely a device
interrupt will arrive soon, but because the tick is already disabled, we
don't dare trust it, causing possible excessive latencies.

Would an alternative be to make @stop_tick be an enum capable of forcing
the tick back on?

enum tick_action {
	NOHZ_TICK_STOP,
	NOHZ_TICK_RETAIN,
	NOHZ_TICK_START,
};

	enum tick_action tick_action = NOHZ_TICK_STOP;

	state = cpuidle_select(..., &tick_action);

	switch (tick_action) {
	case NOHZ_TICK_STOP:
		tick_nohz_stop_tick();
		break;

	case NOHZ_TICK_RETAIN:
		tick_nozh_retain_tick();
		break;

	case NOHZ_TICK_START:
		tick_nohz_start_tick();
		break;
	};


Or something along those lines?