Re: [PATCH 04/19] sched: Prepare for Core-wide rq->lock

[Aubrey Li]

On Wed, Apr 28, 2021 at 5:14 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Tue, Apr 27, 2021 at 04:30:02PM -0700, Josh Don wrote:
>
> > Also, did you mean to have a preempt_enable_no_resched() rather than
> > prempt_enable() in raw_spin_rq_trylock?
>
> No, trylock really needs to be preempt_enable(), because it can have
> failed, at which point it will not have incremented the preemption count
> and our decrement can hit 0, at which point we really should reschedule.
>
> > I went over the rq_lockp stuff again after Don's reported lockup. Most
> > uses are safe due to already holding an rq lock. However,
> > double_rq_unlock() is prone to race:
> >
> > double_rq_unlock(rq1, rq2):
> > /* Initial state: core sched enabled, and rq1 and rq2 are smt
> > siblings. So, double_rq_lock(rq1, rq2) only took a single rq lock */
> > raw_spin_rq_unlock(rq1);
> > /* now not holding any rq lock */
> > /* sched core disabled. Now __rq_lockp(rq1) != __rq_lockp(rq2), so we
> > falsely unlock rq2 */
> > if (__rq_lockp(rq1) != __rq_lockp(rq2))
> >         raw_spin_rq_unlock(rq2);
> > else
> >         __release(rq2->lock);
> >
> > Instead we can cache __rq_lockp(rq1) and __rq_lockp(rq2) before
> > releasing the lock, in order to prevent this. FWIW I think it is
> > likely that Don is seeing a different issue.
>
> Ah, indeed so.. rq_lockp() could do with an assertion, not sure how to
> sanely do that. Anyway, double_rq_unlock() is simple enough to fix, we
> can simply flip the unlock()s.
>
> ( I'm suffering a cold and am really quite slow atm )
>
> How's this then?
>
> ---
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index f732642e3e09..3a534c0c1c46 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -290,6 +290,10 @@ static void sched_core_assert_empty(void)
>  static void __sched_core_enable(void)
>  {
>         static_branch_enable(&__sched_core_enabled);
> +       /*
> +        * Ensure raw_spin_rq_*lock*() have completed before flipping.
> +        */
> +       synchronize_sched();

synchronize_sched() seems no longer exist...

Thanks,
-Aubrey