Re: [PATCH] software node: fix wrong node passed to find nargs_prop

From: Sakari Ailus
Date: Wed Dec 22 2021 - 04:35:06 EST


Hi Daniel,

On Tue, Dec 21, 2021 at 10:09:45PM +0000, Daniel Scally wrote:
> Hi Sakari
>
> On 21/12/2021 09:45, Sakari Ailus wrote:
> > Hi Greg, Andy,
> >
> > On Tue, Dec 21, 2021 at 10:34:11AM +0100, Greg Kroah-Hartman wrote:
> >> On Mon, Dec 20, 2021 at 11:37:07PM +0000, Daniel Scally wrote:
> >>> Thanks Andy
> >>>
> >>> On 20/12/2021 22:13, Andy Shevchenko wrote:
> >>>> + Sakari, Dan
> >>>>
> >>>> On Monday, December 20, 2021, Clément Léger <clement.leger@xxxxxxxxxxx
> >>>> <mailto:clement.leger@xxxxxxxxxxx>> wrote:
> >>>>
> >>>> nargs_prop refers to a property located in the reference that is found
> >>>> within the nargs property.
> >>>
> >>> I think this is right (it's not used in the ACPI version, and the OF
> >>> version is quite convoluted so a bit hard to follow)...but also I note
> >>> that none of the users of fwnode_property_get_reference_args() pass
> >>> anything to nargs_prop anyway...do we even need this?
> >>
> >> Looks like it is unused, please just remove it.
> >
> > If you remove nargs_prop argument, then callers will have to use OF
> > property API instead to parse references with property-defined number of
> > arguments. The goal has been to cover all functionality in a
> > firmware-independent way.
>
> My mistake, I missed that of_parse_phandle_with_args() has a ton of
> direct users. I guess we should try to replace those with
> fwnode_property_get_reference_args() where appropriate.

I'd say at least when the code is otherwise using fwnode property API.

I guess most of the reference users are OF-based originally while cameras
are perhaps a bit of an exception to this. :-)

--
Regards,

Sakari Ailus