Re: [PATCH 1/2] mailbox: qcom-ipcc: amend qcom_ipcc_domain_map() to report errors

[Gabriele Paoloni]

On Mon, Mar 16, 2026 at 12:05 PM Brian Masney <bmasney@xxxxxxxxxx> wrote:
>
> On Mon, Mar 16, 2026 at 11:26:17AM +0100, Gabriele Paoloni wrote:
> > Currently qcom_ipcc_domain_map() ignores errors returned by
> > irq_set_chip_data() and invokes irq_set_chip_and_handler()
> > that in turn ignores errors returned by irq_set_chip().
> > This patch fixes both issues; no other functional changes
> > are implemented.
> >
> > Signed-off-by: Gabriele Paoloni <gpaoloni@xxxxxxxxxx>
> > ---
> >  drivers/mailbox/qcom-ipcc.c | 14 +++++++++++---
> >  1 file changed, 11 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/mailbox/qcom-ipcc.c b/drivers/mailbox/qcom-ipcc.c
> > index d957d989c0ce..c23efaaa64a1 100644
> > --- a/drivers/mailbox/qcom-ipcc.c
> > +++ b/drivers/mailbox/qcom-ipcc.c
> > @@ -116,12 +116,20 @@ static struct irq_chip qcom_ipcc_irq_chip = {
> >  static int qcom_ipcc_domain_map(struct irq_domain *d, unsigned int irq,
> >                               irq_hw_number_t hw)
> >  {
> > +     int ret;
> >       struct qcom_ipcc *ipcc = d->host_data;
>
> Put variables in reverse Christmas tree order.

Noted, I'll fix in v2.

>
> >
> > -     irq_set_chip_and_handler(irq, &qcom_ipcc_irq_chip, handle_level_irq);
>
> Should irq_set_chip_and_handler() and irq_set_chip_and_handler_name() be
> updated to return an int to reduce boiler plate code?

+TO Thomas Gleixner

It may be a possibility. @Thomas do you have any preference in this regard?

Thanks
Gab

>
> > -     irq_set_chip_data(irq, ipcc);
> > -     irq_set_noprobe(irq);
> > +     ret = irq_set_chip(irq, &qcom_ipcc_irq_chip);
> > +     if (ret)
> > +             return ret;
> > +
> > +     irq_set_handler(irq, handle_level_irq);
> >
> > +     ret = irq_set_chip_data(irq, ipcc);
> > +     if (ret)
> > +             return ret;
> > +
> > +     irq_set_noprobe(irq);
> >       return 0;
>
> The newline before the return 0 is removed. That should also remove the
> irq_set_noprobe() change from the diffstat.
>
> Brian
>