Re: [PATCH] kthread: Prevent unpark race which puts threads on the wrong cpu

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, 11 Apr 2013, Dave Hansen wrote:
> On 04/11/2013 03:19 AM, Thomas Gleixner wrote:
> > --- linux-2.6.orig/kernel/smpboot.c
> > +++ linux-2.6/kernel/smpboot.c
> > @@ -185,8 +185,18 @@ __smpboot_create_thread(struct smp_hotpl
> >  	}
> >  	get_task_struct(tsk);
> >  	*per_cpu_ptr(ht->store, cpu) = tsk;
> > -	if (ht->create)
> > -		ht->create(cpu);
> > +	if (ht->create) {
> > +		/*
> > +		 * Make sure that the task has actually scheduled out
> > +		 * into park position, before calling the create
> > +		 * callback. At least the migration thread callback
> > +		 * requires that the task is off the runqueue.
> > +		 */
> > +		if (!wait_task_inactive(tsk, TASK_PARKED))
> > +			WARN_ON(1);
> > +		else
> > +			ht->create(cpu);
> > +	}
> >  	return 0;
> >  }
> 
> This one appears to be doing the trick.  I'll run the cpus in an
> online/offline loop for a bit and make sure it's stable.  It's passed
> several round so far, which is way more than it's done up to this point,
> though.

I think the most critical part is bootup when the threads are
created. The online/offline one is an issue as well, which is covered
by the patches, but way harder to trigger.

Thanks,

	tglx