Re: [PATCH 01/27] ARM: pxa: remove unused board files

[Richard Purdie]

On Thu, 2023-01-05 at 23:16 +0100, Arnd Bergmann wrote:
> On Thu, Jan 5, 2023, at 18:05, Richard Purdie wrote:
> > On Thu, 2023-01-05 at 17:50 +0200, Dmitry Baryshkov wrote:
> > > чт, 5 янв. 2023 г. в 15:46, Arnd Bergmann <arnd@xxxxxxxxxx>:
> > > > 
> > > > From: Arnd Bergmann <arnd@xxxxxxxx>
> > > > 
> > > > The majority of all pxa board files has not been touched in a long time,
> > > > and no users have spoken up in favor of keeping them around. This leaves
> > > > only support for the platforms that were already converted to DT, as
> > > > well as the gumstix and spitz/akita/borzoi machines that work in qemu
> > > > and can still be converted to DT later.
> > > 
> > > Well, tosa also works in qemu.
> > > Nevertheless:
> > > Acked-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
> > 
> > I'm kind of sad to see corgi at this point but it isn't really
> > surprising. The hardware is old/slow and likely not used by many
> > anymore, things have moved on and the systems ended up being heavily
> > kernel size footprint limited.
> > 
> > I'd just note that corgi (and shepherd/husky) are strongly related to
> > spitz/akita/borzoi, the difference is basically pxa25x vs pxa27x
> > processor and pin changes. We chose just to emulate one set in qemu but
> > the others would be relatively trivial.
> > 
> > tosa and poodle have much less in common with the other Zaurus devices.
> > 
> > I guess what I'm saying is that if spitz is remaining it might make
> > sense to have corgi with it.
> 
> Ok, good to know, thanks for the background.
> 
> The way we came up with the list of unused boards is to
> assume they are all unused unless someone asked for it
> to be kept around for this time, for pretty much any reason.
> I probably forgot to have you on the Cc list when we discussed
> this in the past.

I didn't see any discussion about this until now, no.

> IIRC, nobody specifically asked about keeping either
> the corgi or the spitz series, the reason for keeping it
> was to have at least one handheld platform with qemu
> support remain for testing, but I don't see a reason
> for keeping more than one of them.
> 
> If you feel like we should keep tosa or corgi and ask
> again next year, we could do that, but reworking the series
> to keep tosa around would have additional side-effects
> on the tmio-mmc driver that can otherwise get cleaned
> up quite a bit.
> 
> In the long run, I expect we will remove all the remaining
> legacy boardfiles and only keep the DT support. Ideally
> if someone is motivated to convert spitz to DT, supporting
> corgi the same way is also easy.

Personally, I'm not that interested in tosa (or poodle/collie).

The other zaurus devices are interesting for me as I know the platform,
they have qemu emulation, there are set of devices which are similar
but also have differences and there were one of the original targets
for OpenEmbedded and Yocto Project. I did quite a bit of work to get
one kernel which could run on multiple devices, as best you could at
the time! I'd actually forgotten about the qemu emulation.

There is a need for better automated testing around DT in OE/YP and
this is making me wonder about a few potential ideas.

Is there any conversion to DT you can easily point at as an example of
the kinds of changes needed?

Cheers,

Richard