[RFC rcu/next] torture: Stop onoff task if there is only one cpu

[RFC rcu/next] torture: Stop onoff task if there is only one cpu

[Boqun Feng]

If the whole system has only one cpu, that cpu won't be able to be
offlined, so there is no need onoff task is stil running.

Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---

I hit something like the following while I was running rcutorture
in a guest with only one vCPU:

[   31.197457] rcu-torture:torture_onoff task: offlining 0
[   31.197508] rcu-torture:torture_onoff task: offline 0 failed: errno -16

I know this is an expected behavior, but think we could just stop
the onoff task if there is only one cpu.

 kernel/torture.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/kernel/torture.c b/kernel/torture.c
index fb39a06bbef5..a85b7d61d9dd 100644
--- a/kernel/torture.c
+++ b/kernel/torture.c
@@ -194,6 +194,12 @@ torture_onoff(void *arg)
 	for_each_online_cpu(cpu)
 		maxcpu = cpu;
 	WARN_ON(maxcpu < 0);
+
+	if (maxcpu == 0) {
+		VERBOSE_TOROUT_STRING("only one cpu is found, onoff is impossible");
+		goto stop;
+	}
+
 	if (onoff_holdoff > 0) {
 		VERBOSE_TOROUT_STRING("torture_onoff begin holdoff");
 		schedule_timeout_interruptible(onoff_holdoff);
@@ -209,6 +215,8 @@ torture_onoff(void *arg)
 				       &sum_online, &min_online, &max_online);
 		schedule_timeout_interruptible(onoff_interval);
 	}
+
+stop:
 	torture_kthread_stopping("torture_onoff");
 	return 0;
 }
-- 
2.8.0

Re: [RFC rcu/next] torture: Stop onoff task if there is only one cpu

[Josh Triplett]

On Mon, May 02, 2016 at 10:30:00AM +0800, Boqun Feng wrote:
> If the whole system has only one cpu, that cpu won't be able to be
> offlined, so there is no need onoff task is stil running.
> 
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> ---
> 
> I hit something like the following while I was running rcutorture
> in a guest with only one vCPU:
> 
> [   31.197457] rcu-torture:torture_onoff task: offlining 0
> [   31.197508] rcu-torture:torture_onoff task: offline 0 failed: errno -16
> 
> I know this is an expected behavior, but think we could just stop
> the onoff task if there is only one cpu.

I find it a little bit unfortunate that this kicks off a thread just to
immediately exit that thread, rather than never starting it in the first
place.  However, it also seems like the most convenient solution here,
and I don't see much point in going out of the way to optimize this test
for uniprocessor systems.

Reviewed-by: Josh Triplett <josh@joshtriplett.org>

>  kernel/torture.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/kernel/torture.c b/kernel/torture.c
> index fb39a06bbef5..a85b7d61d9dd 100644
> --- a/kernel/torture.c
> +++ b/kernel/torture.c
> @@ -194,6 +194,12 @@ torture_onoff(void *arg)
>  	for_each_online_cpu(cpu)
>  		maxcpu = cpu;
>  	WARN_ON(maxcpu < 0);
> +
> +	if (maxcpu == 0) {
> +		VERBOSE_TOROUT_STRING("only one cpu is found, onoff is impossible");
> +		goto stop;
> +	}
> +
>  	if (onoff_holdoff > 0) {
>  		VERBOSE_TOROUT_STRING("torture_onoff begin holdoff");
>  		schedule_timeout_interruptible(onoff_holdoff);
> @@ -209,6 +215,8 @@ torture_onoff(void *arg)
>  				       &sum_online, &min_online, &max_online);
>  		schedule_timeout_interruptible(onoff_interval);
>  	}
> +
> +stop:
>  	torture_kthread_stopping("torture_onoff");
>  	return 0;
>  }
> -- 
> 2.8.0
> 

Re: [RFC rcu/next] torture: Stop onoff task if there is only one cpu

[Paul E. McKenney]

On Mon, May 02, 2016 at 10:18:49AM -0700, Josh Triplett wrote:
> On Mon, May 02, 2016 at 10:30:00AM +0800, Boqun Feng wrote:
> > If the whole system has only one cpu, that cpu won't be able to be
> > offlined, so there is no need onoff task is stil running.
> > 
> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> > ---
> > 
> > I hit something like the following while I was running rcutorture
> > in a guest with only one vCPU:
> > 
> > [   31.197457] rcu-torture:torture_onoff task: offlining 0
> > [   31.197508] rcu-torture:torture_onoff task: offline 0 failed: errno -16
> > 
> > I know this is an expected behavior, but think we could just stop
> > the onoff task if there is only one cpu.
> 
> I find it a little bit unfortunate that this kicks off a thread just to
> immediately exit that thread, rather than never starting it in the first
> place.  However, it also seems like the most convenient solution here,
> and I don't see much point in going out of the way to optimize this test
> for uniprocessor systems.
> 
> Reviewed-by: Josh Triplett <josh@joshtriplett.org>

Thank you both!  Queued for testing and further review.

								Thanx, Paul

> >  kernel/torture.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > diff --git a/kernel/torture.c b/kernel/torture.c
> > index fb39a06bbef5..a85b7d61d9dd 100644
> > --- a/kernel/torture.c
> > +++ b/kernel/torture.c
> > @@ -194,6 +194,12 @@ torture_onoff(void *arg)
> >  	for_each_online_cpu(cpu)
> >  		maxcpu = cpu;
> >  	WARN_ON(maxcpu < 0);
> > +
> > +	if (maxcpu == 0) {
> > +		VERBOSE_TOROUT_STRING("only one cpu is found, onoff is impossible");
> > +		goto stop;
> > +	}
> > +
> >  	if (onoff_holdoff > 0) {
> >  		VERBOSE_TOROUT_STRING("torture_onoff begin holdoff");
> >  		schedule_timeout_interruptible(onoff_holdoff);
> > @@ -209,6 +215,8 @@ torture_onoff(void *arg)
> >  				       &sum_online, &min_online, &max_online);
> >  		schedule_timeout_interruptible(onoff_interval);
> >  	}
> > +
> > +stop:
> >  	torture_kthread_stopping("torture_onoff");
> >  	return 0;
> >  }
> > -- 
> > 2.8.0
> > 
>