linux-kernel - Re: Future of NOHZ full/isolation development (was Re: [NOHZ] Remove scheduler_tick_max

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, 12 Nov 2014 14:54:58 +0100
From:	Frederic Weisbecker <fweisbec@...il.com>
To:	Viresh Kumar <viresh.kumar@...aro.org>
Cc:	Christoph Lameter <cl@...ux.com>,
	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Gilad Ben-Yossef <gilad@...yossef.com>,
	Tejun Heo <tj@...nel.org>,
	John Stultz <john.stultz@...aro.org>,
	Mike Frysinger <vapier@...too.org>,
	Minchan Kim <minchan.kim@...il.com>,
	Hakan Akkan <hakanakkan@...il.com>,
	Max Krasnyansky <maxk@....qualcomm.com>,
	Hugh Dickins <hughd@...gle.com>,
	"H. Peter Anvin" <hpa@...or.com>, Ingo Molnar <mingo@...nel.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Kevin Hilman <khilman@...aro.org>
Subject: Re: Future of NOHZ full/isolation development (was Re: [NOHZ] Remove
 scheduler_tick_max_deferment)

On Wed, Nov 12, 2014 at 11:41:09AM +0530, Viresh Kumar wrote:
> On 11 November 2014 22:45, Frederic Weisbecker <fweisbec@...il.com> wrote:
> 
> > Here is a summarized list:
> >
> > * Unbound workqueues affinity (to housekeeper)
> > * Unbound timers affinity (to housekeeper)
> > * 1 Hz residual scheduler tick offlining to housekeeper
> > * Fix some scheduler accounting that don't even work with 1 Hz: cpu load
> >   accounting, rt_scale, load balancing, etc...
> > * Lighten the syscall path and get rid of cputime accounting + RCU hooks
> >   for people who want isolation + fast syscalls and faults.
> > * Work on non-affinable workqueues
> > * Work on non-affinable timers
> > * ...
> 
> + spurious interrupts with NOHZ_FULL on all architectures which break isolation
> but doesn't get caught with traces. Can be observed with this:
> 
> diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
> index 481fa54..91d490d 100644
> --- a/kernel/time/hrtimer.c
> +++ b/kernel/time/hrtimer.c
> @@ -1244,7 +1244,8 @@ void hrtimer_interrupt(struct clock_event_device *dev)
>  {
>         struct hrtimer_cpu_base *cpu_base = this_cpu_ptr(&hrtimer_bases);
>         ktime_t expires_next, now, entry_time, delta;
> -       int i, retries = 0;
> +       int i, retries = 0, count = 0;
> +       static int total_spurious;
> 
>         BUG_ON(!cpu_base->hres_active);
>         cpu_base->nr_events++;
> @@ -1304,10 +1305,14 @@ void hrtimer_interrupt(struct clock_event_device *dev)
>                                 break;
>                         }
> 
> +                       count++;
>                         __run_hrtimer(timer, &basenow);
>                 }
>         }
> 
> +       if (!count)
> +               pr_err("____%s: Totalspurious: %d\n", __func__,
> ++total_spurious);
> +

I'd rather leave that to tracepoints. Like trace_hrtimer_spurious().

Or better yet: have trace_hrtimer_interrupt() which we can compare against
trace_hrtimer_expire_entry/exit() to check if any hrtimer callback have run
in the interrupt. This way we avoid workarounds like the above count.

>         /*
>          * Store the new expiry value so the migration code can verify
>          * against it.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/