Re: [PATCH] dt-bindings: nvmem: convert U-Boot env to a layout

[Rafał Miłecki]

On 8.07.2024 08:32, Miquel Raynal wrote:
> zajec5@xxxxxxxxx wrote on Sat,  6 Jul 2024 00:58:21 +0200:
>
> > From: Rafał Miłecki <rafal@xxxxxxxxxx>
> >
> > U-Boot environment variables can be stored in various data sources. MTD
> > is just one of available options. Refactor DT binding into a layout so
> > it can be used with UBI volumes and other NVMEM devices.
> >
> > Link: https://lore.kernel.org/all/20231221173421.13737-1-zajec5@xxxxxxxxx/
> > Signed-off-by: Rafał Miłecki <rafal@xxxxxxxxxx>
> > ---
> > I'm sending this PATCH without Linux changes to see if this is the
> > right approach - for developers and (DT) maintainers to review it first.
> >
> > My previous attempt (see above Link) turned out in refusal so I'm just
> > trying to save some time in case this one goes wrong as well.
> >
> > Hopefully the included example (which I really think we should add)
> > explains well how I think this binding should be used with layouts.
> >
> > If I get some positive feedback I'll work on V2 with actual Linux
> > changes.
>
> I am still in favor of this change. I know from a binding perspective
> it might be questionable; but from an OS perspective it makes total
> sense to deprecated the old U-Boot env "device" driver in favor of this.

Please note that at this point I didn't decide to deprecate the old
binding (using "u-boot,env" for flash device partition). That would
require mixing "nvmem-layout" with "partitions" which I believe Rob
objected to.

My PATCH in its current form just allows reusing this binding with
"nvmem-layout"s which is needed for cases where we agreed to allow
"nvmem-layout" (UBI volumes and pure - non-MTD - NVMEM devices).

If we want to refactor this binding in MTD partitions context I hope we
can handle that at different time if we agree on some solution. I hope
and I don't think this should be a blocked for this PATCH.