linux-kernel - Re: [RFC PATCH v2 1/8] sched/cpufreq_schedutil: make use of DEADLINE utilization signal

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

Date:   Tue, 5 Dec 2017 16:34:10 +0000
From:   Patrick Bellasi <patrick.bellasi@....com>
To:     Juri Lelli <juri.lelli@...hat.com>
Cc:     peterz@...radead.org, mingo@...hat.com, rjw@...ysocki.net,
        viresh.kumar@...aro.org, linux-kernel@...r.kernel.org,
        linux-pm@...r.kernel.org, tglx@...utronix.de,
        vincent.guittot@...aro.org, rostedt@...dmis.org,
        luca.abeni@...tannapisa.it, claudio@...dence.eu.com,
        tommaso.cucinotta@...tannapisa.it, bristot@...hat.com,
        mathieu.poirier@...aro.org, tkjos@...roid.com, joelaf@...gle.com,
        morten.rasmussen@....com, dietmar.eggemann@....com,
        alessio.balsini@....com, Juri Lelli <juri.lelli@....com>,
        Ingo Molnar <mingo@...nel.org>,
        "Rafael J . Wysocki" <rafael.j.wysocki@...el.com>
Subject: Re: [RFC PATCH v2 1/8] sched/cpufreq_schedutil: make use of DEADLINE
 utilization signal

On 05-Dec 16:24, Juri Lelli wrote:
> On 05/12/17 15:09, Patrick Bellasi wrote:
> > > +	/*
> > > +	 * Ideally we would like to set util_dl as min/guaranteed freq and
> > > +	 * util_cfs + util_dl as requested freq. However, cpufreq is not yet
> > > +	 * ready for such an interface. So, we only do the latter for now.
> > > +	 */
> > 
> > Maybe I don't completely get the above comment, but to me it is not
> > really required.
> > 
> > When you say that "util_dl" should be set to a min/guaranteed freq
> > are you not actually talking about a DL implementation detail?
> > 
> > From the cpufreq standpoint instead, we should always set a capacity
> > which can accommodate util_dl + util_cfs.
> 
> It's more for platforms which supports such combination of values for
> frequency requests (CPPC like, AFAIU). The idea being that util_dl is
> what the system has to always guarantee, but it could go up to the sum
> if feasible.

I see, you mean for systems where you can specify both values at the
same time, i.e.
- please give me util_dl...
- ... but if you have more beer, I would like util_dl + util_cfs

However, I'm not an expert, on those systems can we really set a
minimum guaranteed performance level?

I was more of the idea that the "minimum guaranteed" is something we
can only read from "firmware", while we can only ask for something
which is never "guaranteed".

> > We don't care about the meaning of util_dl and we should always assume
> > (by default) that the signal is properly updated by the scheduling
> > class... which unfortunately does not always happen for CFS.
> > 

-- 
#include <best/regards.h>

Patrick Bellasi