Re: kmod: avoid propagating PF_NO_SETAFFINITY into userspace child

[Oleg Nesterov <oleg@redhat.com>]

On 11/14, Tejun Heo wrote:
>
> Hello,
>
> On Thu, Nov 14, 2013 at 09:51:47AM +0800, zhang.yi20@zte.com.cn wrote:
> > The kernel worker thread has the PF_NO_SETAFFINITY flag, and it is propagated
> > into the userspace child. Clearing this flag in usersapce child to enable its
> > migrating capability.
> >
> >
> > Signed-off-by: Zhang Yi <zhang.yi20@zte.com.cn>
> >
> > --- linux3-12/kernel/kmod.c	2013-11-14 09:06:58.991781656 +0000
> > +++ linux3-12/kernel/kmod.c	2013-11-14 09:08:47.511781621 +0000
> > @@ -217,6 +217,7 @@ static int ____call_usermodehelper(void
> >  	 * Avoid propagating that into the userspace child.
> >  	 */
> >  	set_user_nice(current, 0);
> > +	current->flags &= ~PF_NO_SETAFFINITY;
>
> I'm a bit confused.  kernel_thread() doesn't use workqueue or
> kthread_bind(),

Yes, but kernel_thread() is called by the worker thread which has
PF_NO_SETAFFINITY, this flag is copied to child->flags.

Looks like Zhang is right... But I'd suggest to change flush_old_exec()
instead (see "current->flags &= ...").

Oleg.