Re: [patch 09/12] futex: Provide state handling for exec() as well

[Thomas Gleixner]

On Thu, 7 Nov 2019, Peter Zijlstra wrote:
> On Wed, Nov 06, 2019 at 10:55:43PM +0100, Thomas Gleixner wrote:
> 
> > +static void futex_cleanup_end(struct task_struct *tsk, int state)
> > +{
> > +	/*
> > +	 * Lockless store. The only side effect is that an observer might
> > +	 * take another loop until it becomes visible.
> > +	 */
> > +	tsk->futex_state = state;
> 
> As I mentioned yesterday, paranoia would've made me write this as
> smp_store_release(), also to avoid it creaping back into the locked
> region.
> 
> That is, the comment above deals with it being visible late, but it
> could be visible early.
> 
> At the same time, if this is a release, what does it pair with. The
> obvious place would be the load in handle_exit_race() but that didn't
> want to make sense last night -- and I'm not sure it wants to make more
> sense today.

Right. I couldn't come up with any sensible argument either.

Thanks,

	tglx