Re: [PATCH 01/11] genirq: Add fasteoi IPI flow

[Marc Zyngier <maz@kernel.org>]

On Tue, 19 May 2020 23:25:43 +0100
Valentin Schneider <valentin.schneider@arm.com> wrote:

> On 19/05/20 17:17, Marc Zyngier wrote:
> > For irqchips using the fasteoi flow, IPIs are a bit special.
> >
> > They need to be EOId early (before calling the handler), as
> > funny things may happen in the handler (they do not necessarily
> > behave like a normal interrupt), and that the arch code is
> > already handling the stats.
> >
> > Signed-off-by: Marc Zyngier <maz@kernel.org>
> > ---
> >  include/linux/irq.h |  1 +
> >  kernel/irq/chip.c   | 26 ++++++++++++++++++++++++++
> >  2 files changed, 27 insertions(+)
> >
> > diff --git a/include/linux/irq.h b/include/linux/irq.h
> > index 8d5bc2c237d7..726f94d8b8cc 100644
> > --- a/include/linux/irq.h
> > +++ b/include/linux/irq.h
> > @@ -621,6 +621,7 @@ static inline int irq_set_parent(int irq, int parent_irq)
> >   */
> >  extern void handle_level_irq(struct irq_desc *desc);
> >  extern void handle_fasteoi_irq(struct irq_desc *desc);
> > +extern void handle_percpu_devid_fasteoi_ipi(struct irq_desc *desc);
> >  extern void handle_edge_irq(struct irq_desc *desc);
> >  extern void handle_edge_eoi_irq(struct irq_desc *desc);
> >  extern void handle_simple_irq(struct irq_desc *desc);
> > diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
> > index 41e7e37a0928..7b0b789cfed4 100644
> > --- a/kernel/irq/chip.c
> > +++ b/kernel/irq/chip.c
> > @@ -955,6 +955,32 @@ void handle_percpu_devid_irq(struct irq_desc *desc)
> >               chip->irq_eoi(&desc->irq_data);
> >  }
> >
> > +/**
> > + * handle_percpu_devid_fasteoi_ipi - Per CPU local IPI handler with per cpu
> > + *				     dev ids
> > + * @desc:	the interrupt description structure for this irq
> > + *
> > + * The biggest differences with the IRQ version are that:
> > + * - the interrupt is EOIed early, as the IPI could result in a context
> > + *   switch, and we need to make sure the IPI can fire again
> > + * - Stats are usually handled at the architecture level, so we ignore them
> > + *   here
> > + */
> > +void handle_percpu_devid_fasteoi_ipi(struct irq_desc *desc)
> > +{
> > +	struct irq_chip *chip = irq_desc_get_chip(desc);
> > +	struct irqaction *action = desc->action;
> > +	unsigned int irq = irq_desc_get_irq(desc);
> > +	irqreturn_t res;
> > +  
> 
> Tied to the following patch, does that want something like
> 
> +	if (!irq_settings_is_no_accounting(desc))
> +		__kstat_incr_irqs_this_cpu(desc);
> +
> 
> While I'm at it, now that we would have IPIs as 'normal' interrupts, what
> prevents us from getting rid of the arch-side accounting? Is it just that
> we are keeping it as long as handle_IPI() lives, or are there more hurdles
> hidden around?

See my reply to Florian. I need to mop the rest of the 32bit irqchips
without having the HW (hip04 is almost a copy of the GIC driver, and
the Marvell horror is very RPi like).

Once this is done, we can kill the home-brewed stuff and rely on core
infrastructure.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel