Re: [PATCH v1] gpio: pca953x: fix IRQ storm on system wake up

[Andy Shevchenko <andriy.shevchenko@intel.com>]

On Mon, Apr 07, 2025 at 05:11:14PM +0200, Emanuele Ghidoli wrote:
> On 03/04/2025 15:56, Bartosz Golaszewski wrote:
> > On Thu, Apr 3, 2025 at 3:54 PM Andy Shevchenko
> > <andriy.shevchenko@intel.com> wrote:
> >>
> >> +Cc: Geert
> >>
> >> On Thu, Apr 03, 2025 at 02:07:05PM +0200, Bartosz Golaszewski wrote:
> >>> On Wed, 26 Mar 2025 18:38:38 +0100, Francesco Dolcini wrote:
> >>
> >>>> If an input changes state during wake-up and is used as an interrupt
> >>>> source, the IRQ handler reads the volatile input register to clear the
> >>>> interrupt mask and deassert the IRQ line. However, the IRQ handler is
> >>>> triggered before access to the register is granted, causing the read
> >>>> operation to fail.
> >>>>
> >>>> As a result, the IRQ handler enters a loop, repeatedly printing the
> >>>> "failed reading register" message, until `pca953x_resume` is eventually
> >>>> called, which restores the driver context and enables access to
> >>>> registers.

[...]

> >>> Applied, thanks!
> >>
> >> Won't this regress as it happens the last time [1]?
> >>
> >> [1]: https://lore.kernel.org/linux-gpio/CAMuHMdVnKX23yi7ir1LVxfXAMeeWMFzM+cdgSSTNjpn1OnC2xw@mail.gmail.com/
> >>
> > 
> > Ah, good catch. I'm wondering what the right fix here is but don't
> > really have any ideas at the moment. Any hints are appreciated.
> > 
> > For now, I'm dropping it.
> > 
> > Bart
> 
> I’ve found another possible solution: disable the PCA953x IRQ in
> pca953x_suspend() and re-enable it in pca953x_resume().
> This would prevent the ISR from being triggered while the regmap is in
> cache-only mode.
> The wake-up capability is preserved, since an IRQ can still wake the system
> even when disabled with disable_irq(), as long as it has wake enabled.

Can you enable IRQ debugfs and dump the state of the wake* nodes for the
respective interrupts? In this case we will be 100% sure it works as expected.

> This should avoid introducing regressions and still handle Geert’s use case
> properly.
> 
> Andy, Bart, Geert - what do you think?

Sounds okay, but please double check the above.

-- 
With Best Regards,
Andy Shevchenko