Re: [PATCH 4/4] locking: Introduce smp_cond_acquire()

[Oleg Nesterov <oleg@redhat.com>]

On 11/12, Boqun Feng wrote:
>
> On Thu, Nov 12, 2015 at 04:00:58PM +0100, Oleg Nesterov wrote:
> >
> > No?
> >
>
> do_exit() is surely buggy if spin_lock() could work in this way.

OK, good ;) so we need to fix spin_lock() on PPC ? Or add
mb__after_unlock_lock() but this leads to other questions... Or
change do_exit() to do lock() + unlock().

>
> > And smp_mb__before_spinlock() looks wrong too then.
> >
>
> Maybe not? As smp_mb__before_spinlock() is used before a LOCK operation,
> which has both LOAD part and STORE part unlike spin_unlock_wait()?

Maybe not.

But let me remind that the original purpose of this mb__before_spinlock()
was to ensure that "CONDITION = true" before ttwu() can not be reordered
with

	if (!(p->state & state))
		goto out;		// do not wakeup


inside try_to_wake_up(). Otherwise

	CONDITION = true;
	try_to_wake_up(p);

can race with "p" doing

	set_current_state(...);	// implies mb();
	if (CONDITION)
		return;
	schedule();

because try_to_wake_up() can read p->state before it sets CONDITION = 1
and then it won't wakeup "p" which has already checked this CONDITION.


Now. If try_to_wake_up() can read p->state before it writes to *pi_lock,
then how smp_mb__before_spinlock() == wmb() can help to serialize STORE
and LOAD?

It seems that PPC needs to define smp_mb__before_spinlock() as full mb(),
no?

Oleg.