Re: [PATCH] irq_work: Avoid unnecessary "IRQ work" interrupts

[Brandt, Oliver - Lenze]

On Wed, 2024-08-28 at 11:54 +0200, Sebastian Andrzej Siewior wrote:
> On 2024-08-28 11:37:20 [+0200], To Brandt, Oliver - Lenze wrote:
> > > Fixes: b4c6f86ec2f6 ('irq_work: Handle some irq_work in a per-CPU thread on PREEMPT_RT')
> > > Signed-off-by: Oliver Brandt <oliver.brandt@xxxxxxxxx>
> > > ---
> > >  kernel/irq_work.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/kernel/irq_work.c b/kernel/irq_work.c
> > > index 2f4fb336dda1..df08b7dde7d5 100644
> > > --- a/kernel/irq_work.c
> > > +++ b/kernel/irq_work.c
> > > @@ -108,7 +108,7 @@ static void __irq_work_queue_local(struct irq_work *work)
> > >                 return;
> > > 
> > >         /* If the work is "lazy", handle it from next tick if any */
> > > -       if (!lazy_work || tick_nohz_tick_stopped())
> > > +       if (!(lazy_work || rt_lazy_work) || tick_nohz_tick_stopped())
> > >                 irq_work_raise(work);
> > 
> > Looking at this I *think* rt_lazy_work was needed earlier due to
> > different code but not anymore. Couldn't you just remove rt_lazy_work
> > and set lazy_work in the RT path? That should work.
> 
> Actually no. If we merge rt_lazy_work into lazy_work then we would have
> the behaviour you have here. But we need irq_work_raise() in order to
>         irq_work_run();
>         -> wake_irq_workd();
>           -> wake_up_process(irq_workd);
> 
> If we don't irq_work_raise() here then it will be delayed until the next
> tick and we didn't want that if my memory serves me.

Hmm.... I see. What about calling wake_irq_workd() directly; something
like

        if (rt_lazy_work)
                wake_irq_workd();
        else if (!lazy_work || tick_nohz_tick_stopped())
                irq_work_raise(work);

> 
> Sebastian

Regards,
Oliver