Re: [PATCH] Input: gpio-keys - add hibernation support

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

Hi Armando,

On Mon, Apr 06, 2026 at 09:04:37AM -0700, Armando De Leon wrote:
> The gpio-keys driver uses DEFINE_SIMPLE_DEV_PM_OPS which maps .freeze
> and .restore to the same callbacks as .suspend and .resume. This is
> insufficient for hibernation (suspend-to-disk) because the SoC is fully
> powered off, unlike suspend-to-RAM where hardware state is preserved.
> 
> After hibernation resume, GPIO keys fail to function correctly because
> the interrupt controller (e.g. GIC) is fully re-initialized during
> restore, resetting all IRQ trigger type configurations to defaults.
> GPIO keys require IRQ_TYPE_EDGE_BOTH for proper press/release detection,
> but after the interrupt controller re-initialization only a single edge
> remains active. This causes buttons to report only release events but
> not press events, or vice versa.

I believe you just described a bug in the interrupt controller handling
of hibernation. It needs to be fixed there, not in consumer driver.

Thanks.

-- 
Dmitry