linux-kernel - Re: [PATCH 09/10] perf/cputime: Don't stop idle tick if there's live cputime event

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]

Message-ID: <20180608001203.GB12322@krava>
Date:   Fri, 8 Jun 2018 02:12:03 +0200
From:   Jiri Olsa <jolsa@...hat.com>
To:     Stephane Eranian <eranian@...gle.com>
Cc:     Andi Kleen <andi@...stfloor.org>, Jiri Olsa <jolsa@...nel.org>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Peter Zijlstra <a.p.zijlstra@...llo.nl>,
        LKML <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        David Ahern <dsahern@...il.com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Milian Wolff <milian.wolff@...b.com>, frederic@...nel.org
Subject: Re: [PATCH 09/10] perf/cputime: Don't stop idle tick if there's live
 cputime event

On Thu, Jun 07, 2018 at 09:01:30AM -0700, Stephane Eranian wrote:
> Hi Jiri,
> 
> On Thu, Jun 7, 2018 at 8:45 AM Andi Kleen <andi@...stfloor.org> wrote:
> >
> > On Thu, Jun 07, 2018 at 12:15:12AM +0200, Jiri Olsa wrote:
> > > Disable stopping of the idle tick when having live cputime
> > > event. When the tick is disabled, the idle counts are out
> > > of date until next tick/update and perf cputime PMU provides
> > > misleading counts.
> >
> > I really don't like this. This can totally change performance
> > (e.g. less Turbo due to less idle) and performance tools shouldn't
> > change the performance profile drastically.
> >
> You do not want to change the behavior of the kernel just because you
> are monitoring.
> This may introduce side effects on other events which may not otherwise exist.

right.. I guess we can survive few seconds of misleading idle counts
and perhaps we could detect nohz is enabled and warn about this

thanks,
jirka