Re: [PATCH v2] pinctrl: samsung: add irq_set_affinity() for non wake up external gpio interrupt

[Sam Protsenko <semen.protsenko@xxxxxxxxxx>]

On Tue, Nov 28, 2023 at 1:29 AM Krzysztof Kozlowski
<krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
> On 28/11/2023 02:01, Youngmin Nam wrote:
> > On Mon, Nov 27, 2023 at 10:54:56AM +0100, Krzysztof Kozlowski wrote:
> >> On 26/11/2023 10:46, Youngmin Nam wrote:
> >>> To support affinity setting for non wake up external gpio interrupt,
> >>> add irq_set_affinity callback using irq number from pinctrl driver data.
> >>>
> >>> Before this patch, changing the irq affinity of gpio interrupt is not possible:
> >>>
> >>>     # cat /proc/irq/418/smp_affinity
> >>>     3ff
> >>>     # echo 00f > /proc/irq/418/smp_affinity
> >>
> >> Does this command succeed on your board?
> >>
> > Yes.
>
> Hm, fails all the time one mine.
>

I tried to test this patch on E850-96, and an attempt to write into
smp_affinity (for some GPIO irq) also fails for me:

    # echo f0 > smp_affinity
    -bash: echo: write error: Input/output error

When I add some pr_err() to exynos_irq_set_affinity(), I can't see
those printed in dmesg. So I guess exynos_irq_set_affinity() doesn't
get called at all. So the error probably happens before
.irq_set_affinity callback gets called.

Youngmin, can you please try and test this patch on E850-96? This
board is already supported in upstream kernel. For example you can use
"Volume Up" interrupt for the test, which is GPIO irq.

> >
> >>>     # cat /proc/irq/418/smp_affinity
> >>>     3ff
> >>>     # cat /proc/interrupts
> >>>                CPU0       CPU1       CPU2       CPU3    ...
> >>>     418:       3631          0          0          0    ...
> >>>
> >>> With this patch applied, it's possible to change irq affinity of gpio interrupt:
> >>
> >> ...
> >>
> >> On which board did you test it?
> >>
> >>
> > I tested on S5E9945 ERD(Exynos Reference Development) board.
>
> There is no such board upstream. How can we reproduce this issue? I am
> afraid we cannot test neither the bug nor the fix.
>
> >
> >>> +   if (parent)
> >>> +           return parent->chip->irq_set_affinity(parent, dest, force);
> >>> +
> >>
> >> I think there is a  helper for it: irq_chip_set_affinity_parent().
> >>
> >>
> >
> > The irq_chip_set_affinity_parent() requires parent_data of irq_data.
>
> Hm, so now I wonder why do we not have parent_data...
>
> > But when I tested as below, exynos's irqd->parent_data was null.
> > So we should use irqchip's affinity function instead of the helper function.
> >
>
>
>
> Best regards,
> Krzysztof
>