linux-kernel - Re: [PATCH v2] sched/clock: print a log when the sched clock is marked unstable

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:   Wed, 7 Jul 2021 17:48:36 +0530 (IST)
From:   Ani Sinha <ani@...sinha.ca>
To:     Ani Sinha <ani@...sinha.ca>
cc:     Peter Zijlstra <peterz@...radead.org>,
        linux-kernel@...r.kernel.org, anirban.sinha@...ia.com,
        tglx@...utronix.de, Ingo Molnar <mingo@...hat.com>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
        Daniel Bristot de Oliveira <bristot@...hat.com>
Subject: Re: [PATCH v2] sched/clock: print a log when the sched clock is
 marked unstable



On Wed, 7 Jul 2021, Ani Sinha wrote:

>
>
> On Wed, 7 Jul 2021, Peter Zijlstra wrote:
>
> > On Wed, Jul 07, 2021 at 04:26:59PM +0530, Ani Sinha wrote:
> > > When the sched clock transitions from stable to unstable and
> > > vice versa, a kernel log is printed. When the sched clock
> > > is marked explicitly as unstable, make __clear_sched_clock_stable()
> > > emit a warning log. It is useful for example in understanding
> > > why a certain feature like NOHZ that depends on availability of a
> > > stable sched clock, is not available.
> > >
> > > Signed-off-by: Ani Sinha <ani@...sinha.ca>
> > > ---
> > >  kernel/sched/clock.c | 1 +
> > >  1 file changed, 1 insertion(+)
> > >
> > > Changelog:
> > > V1: original patch
> > > v2: print pr_warn from __clear_sched_clock_stable instead. Commit log
> > >     updated with proper prefix and wording.
> > >
> > > diff --git a/kernel/sched/clock.c b/kernel/sched/clock.c
> > > index c2b2859ddd82..99484fec0335 100644
> > > --- a/kernel/sched/clock.c
> > > +++ b/kernel/sched/clock.c
> > > @@ -183,6 +183,7 @@ static void __clear_sched_clock_stable(void)
> > >  		return;
> > >
> > >  	tick_dep_set(TICK_DEP_BIT_CLOCK_UNSTABLE);
> > > +	pr_warn("sched_clock: Marking unstable.\n");
> > >  	schedule_work(&sched_clock_work);
> > >  }
> >
> > I'm confused... that work has a ton of printk()s in, what is this
> > pr_warn() adding?
> >
>
> Right, but for whatever reason, on latest version of Ubuntu (21.04)
> running on hyperV, I do not see those logs:
>
> azureuser@...tOS:~$ uname -a
> Linux CentOS 5.11.0-1009-azure #9-Ubuntu SMP Thu Jun 17 02:20:07 UTC 2021
> x86_64 x86_64 x86_64 GNU/Linux
> azureuser@...tOS:~$ dmesg | grep unstable
> dmesg: read kernel buffer failed: Operation not permitted
> azureuser@...tOS:~$ sudo su
> root@...tOS:/home/azureuser# dmesg | grep unstable
> [    0.000000] tsc: Marking TSC unstable due to running on Hyper-V
> root@...tOS:/home/azureuser# dmesg | grep sched
> [    0.384656] rcu: RCU calculated value of scheduler-enlistment delay is
> 25 jiffies.
> [    1.122977] io scheduler mq-deadline registered
> root@...tOS:/home/azureuser# exit
>
> Hence, I was trying the patch.

OK never mimd. Seems on hyperV, they use read_hv_sched_clock_tsc() for
pv_sched_clock and therefore using_native_sched_clock() returns false.
Thus, clear_sched_clock_stable() is not called.

I will need to do some instrumentation to confirm this. Maybe we do not
need any patching after all.