Re: [RFC][PATCH v3 6/6] rcu/tree: Use irq_work_queue_remote()

From: Peter Zijlstra
Date: Thu Oct 29 2020 - 05:21:00 EST


On Wed, Oct 28, 2020 at 01:15:54PM -0700, Paul E. McKenney wrote:
> On Wed, Oct 28, 2020 at 09:02:43PM +0100, Peter Zijlstra wrote:

> > Subject: rcu/tree: Use irq_work_queue_remote()
> > From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > Date: Wed Oct 28 11:53:40 CET 2020
> >
> > All sites that consume rcu_iw_gp_seq seem to have rcu_node lock held,
> > so setting it probably should too. Also the effect of self-IPI here
> > would be setting rcu_iw_gp_seq to the value we just set it to
> > (pointless) and clearing rcu_iw_pending, which we just set, so don't
> > set it.
> >
> > Passes TREE01.
> >
> > Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> > ---
> > kernel/rcu/tree.c | 10 ++++++----
> > 1 file changed, 6 insertions(+), 4 deletions(-)
> >
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -1308,14 +1308,16 @@ static int rcu_implicit_dynticks_qs(stru
> > resched_cpu(rdp->cpu);
> > WRITE_ONCE(rdp->last_fqs_resched, jiffies);
> > }
> > -#ifdef CONFIG_IRQ_WORK
> > + raw_spin_lock_rcu_node(rnp);
>
> The caller of rcu_implicit_dynticks_qs() already holds this lock.
> Please see the force_qs_rnp() function and its second call site,
> to which rcu_implicit_dynticks_qs() is passed as an argument.
>
> But other than that, this does look plausible. And getting rid of
> that #ifdef is worth something. ;-)

Dang, clearly TREE01 didn't actually hit any of this code :/ Is there
another test I should be running?