Re: [tip:sched/urgent] cpuset: Fix cpuset_cpus_allowed_fallback(), don't update tsk->rt.nr_cpus_allowed

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Sat, 2011-05-28 at 16:35 +0000, tip-bot for KOSAKI Motohiro wrote:
> +++ b/kernel/kthread.c
> @@ -202,8 +202,8 @@ void kthread_bind(struct task_struct *p, unsigned int cpu)
>                 return;
>         }
>  
> -       p->cpus_allowed = cpumask_of_cpu(cpu);
> -       p->rt.nr_cpus_allowed = 1;
> +       /* It's safe because the task is inactive. */
> +       do_set_cpus_allowed(p, cpumask_of(cpu));
>         p->flags |= PF_THREAD_BOUND;
>  }


I just happened  to be staring at this stuff again, and I'm wondering
how and why this is correct. After kthread_create() the thread exists
and is exposed in the pid-hash, therefore userspace can come and do
sys_sched_setaffinity() on it, and since we're not holding any locks and
set PF_THREAD_BOUND _after_ setting cpus_allowed, things can end up
funny.

Hmm?