Re: [PATCH 2/2] x86/platform/geode: reference the real node of the cs5535 GPIO controller

[Bartosz Golaszewski]

On Thu, May 21, 2026 at 4:19 PM Dave Hansen <dave.hansen@xxxxxxxxx> wrote:
>
> On 4/29/26 05:23, Bartosz Golaszewski wrote:
> > We now can access the address of the software node associated with the
> > GPIO controller cell of the cs5535 MFD device. Make it the target of the
> > GPIO software node references in geode-common.c. Make sure the cs5535
> > driver is built-in for all boards selecting GEODE_COMMON.
>
> This looks like a good cleanup, but the changelog has me scratching my
> head a bit.
>
> I'm not quite sure what problem this is solving. It's also a bit unusual

I thought the cover letter explains it pretty well? The software node
representing the GPIO chip we're referencing from the software node
attached to the GPIO consumer here is not attached to the GPIO
controller device. This means that the lookup happens based on the
GPIO controller's label matched against the name of the software node.
I want to remove this behavior in favor of matching actual firmware
nodes. This is being done for many reasons: string matching is wonky
at best and I'm working on adding fw_devlink support for software
nodes but that requires real references between devices.

> to have a driver depend on another not being built as a module. There's

This is not a driver, this is a board file that's always built-in. And
I agree it's not optimal but it's already being done for other files
for this platform and it's legacy anyway so I think we can live with
that.

> also not really a mention of the:
>
>         +MODULE_IMPORT_NS("CS5535");
>
> in the changelog.
>

Ok, I'll add it.

> Oh, and on a minor note, it might all be more clear if the changelog was
> written in imperative voice and removed the "We".
>
> Honestly, this is old hardware so I don't want to fret about this too
> much. I'd be happy to ack this, but I don't really know what I'm acking. ;)

Thanks,
Bartosz