On Wed, Oct 28, 2020 at 02:40:46PM +0100, Frederic Weisbecker wrote:
> On Wed, Oct 28, 2020 at 12:07:12PM +0100, Peter Zijlstra wrote:
> > While the traditional irq_work relies on the ability to self-IPI, it
> > makes sense to provide an unconditional irq_work_queue_remote()
> > interface.
>
> We may need a reason as well here.
Well, it doesn't rely on arch self-IPI code. The remote irq_work bits
are generic SMP code.
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -1308,13 +1308,14 @@ static int rcu_implicit_dynticks_qs(stru
> > resched_cpu(rdp->cpu);
> > WRITE_ONCE(rdp->last_fqs_resched, jiffies);
> > }
> > - if (IS_ENABLED(CONFIG_IRQ_WORK) &&
> > - !rdp->rcu_iw_pending && rdp->rcu_iw_gp_seq != rnp->gp_seq &&
> > +#ifdef CONFIG_IRQ_WORK
> > + if (!rdp->rcu_iw_pending && rdp->rcu_iw_gp_seq != rnp->gp_seq &&
>
> If it's unconditional on SMP, I expect it to be unconditional on rcutree.
>
> Also this chunk seems unrelated to this patch.
This hunk is due to irq_work_queue_on() no longer existing for
CONFIG_IRQ_WORK and hence breaking the compile with that IS_ENABLED()
crud.
That is, this changes IS_ENABLED() for a proper #ifdef.
> > (rnp->ffmask & rdp->grpmask)) {
> > rdp->rcu_iw_pending = true;
> > rdp->rcu_iw_gp_seq = rnp->gp_seq;
> > irq_work_queue_on(&rdp->rcu_iw, rdp->cpu);
> > }
> > +#endif
> > }
> >
> > return 0;
> >
> >
