Re: [PATCH] Nokia RX-51: Convert to CONFIG_DM_KEYBOARD

[Simon Glass <sjg@chromium.org>]

Hi Tom,

On Fri, 4 Feb 2022 at 08:37, Tom Rini <trini@konsulko.com> wrote:
>
> On Fri, Feb 04, 2022 at 04:30:23PM +0100, Pali Rohár wrote:
> > On Friday 04 February 2022 08:24:03 Simon Glass wrote:
> > > Hi Pali,
> > >
> > > On Fri, 4 Feb 2022 at 07:13, Pali Rohár <pali@kernel.org> wrote:
> > > >
> > > > On Friday 04 February 2022 09:00:55 Tom Rini wrote:
> > > > > On Fri, Feb 04, 2022 at 02:55:39PM +0100, Pali Rohár wrote:
> > > > > > On Friday 04 February 2022 08:41:55 Tom Rini wrote:
> > > > > > > On Fri, Feb 04, 2022 at 11:56:52AM +0100, Pali Rohár wrote:
> > > > > > > > On Thursday 03 February 2022 15:02:02 Simon Glass wrote:
> > > > > > > > > Hi,
> > > > > > > > >
> > > > > > > > > On Thu, 3 Feb 2022 at 14:53, Tom Rini <trini@konsulko.com> wrote:
> > > > > > > > > >
> > > > > > > > > > On Thu, Feb 03, 2022 at 04:45:44PM -0500, Tom Rini wrote:
> > > > > > > > > > > On Thu, Feb 03, 2022 at 04:16:23PM -0500, Tom Rini wrote:
> > > > > > > > > > > > On Thu, Feb 03, 2022 at 07:38:50PM +0100, Pali Rohár wrote:
> > > > > > > > > > > >
> > > > > > > > > > > > > Signed-off-by: Pali Rohár <pali@kernel.org>
> > > > > > > > > > > > > ---
> > > > > > > > > > > > >
> > > > > > > > > > > > > I would really appreciate if U-Boot test framework starts printing
> > > > > > > > > > > > > deprecation warnings, instead of sending patches which directly drop
> > > > > > > > > > > > > support for some boards.
> > > > > > > > > > > > >
> > > > > > > > > > > > > There is absolutely no warning during building U-Boot for RX-51 board
> > > > > > > > > > > > > that this board has not been converted to DM_KEYBOARD yet.
> > > > > > > > > > > >
> > > > > > > > > > > > Please send some patch that implements what you're wanting to see for
> > > > > > > > > > > > how to make the warnings more visible.  I do agree the warning for
> > > > > > > > > > > > v2022.10 only showed up after the merge to -next once v2022.01 came out,
> > > > > > > > > > > > but it's still a fairly long time to clean up the few unconverted
> > > > > > > > > > > > boards.
> > > > > > > > > > >
> > > > > > > > > > > Oh, I see what's going on.  Simon, the DM_KEYBOARD check isn't
> > > > > > > > > > > triggering for any boards.  I'll look more.
> > > > > > > > > >
> > > > > > > > > > And here's where we're at.  nokia_rx51 doesn't set CONFIG_KEYBOARD, so
> > > > > > > > > > didn't trigger the warning about  migration.  Every platform that sets
> > > > > > > > > > CONFIG_KEYBOARD is migrated.  I don't know how many other platforms are
> > > > > > > > > > in the situation nokia_rx51 is in.  Yanking out the legacy code and
> > > > > > > > > > seeing what fails to build, and going from there is probably the next
> > > > > > > > > > option.
> > > > > > > > >
> > > > > > > > > Yes.
> > > > > > > > >
> > > > > > > > > As to your questoin, none that I know of. I sent a series to drop
> > > > > > > > > cfb_console which was how this used to work, although in fact it
> > > > > > > > > hasn't worked for a while. The problem here seems to be that this
> > > > > > > > > board was multiple migrations behind and so was not caught.
> > > > > > > >
> > > > > > > > It is not truth that cfb_console has not worked. This driver is / was
> > > > > > > > working fine on n900 without any issues.
> > > > > > >
> > > > > > > Yes, it's such an old missed migration that there's no check for it.
> > > > > > >
> > > > > > > > > We should be able to remove the migration check.
> > > > > > > > >
> > > > > > > > > Pali, just to explain from the other POV, I am finding it increasingly
> > > > > > > > > frustrating dealing with ad-hoc CONFIG options, old drivers, etc. We
> > > > > > > > > really need to complete some of the migrations we started 6 years ago.
> > > > > > > > >
> > > > > > > > > Regards,
> > > > > > > > > Simon
> > > > > > > >
> > > > > > > > Well, I see and understand you. But as I explained it more times, it was
> > > > > > > > me who is/was waiting for review of n900 patches and I was not able to
> > > > > > > > speed up review process. There was always some n900 patch in waiting
> > > > > > > > state.
> > > > > > > >
> > > > > > > > Now when pending n900 patches were reviewed and merged, I prepared and
> > > > > > > > sent another one -- this DM_KEYBOARD which removes one part of
> > > > > > > > cfb_console code from n900 board code.
> > > > > > >
> > > > > > > Alright.  So, do you see the warning about DM_SERIAL?  Are you going to
> > > > > > > start work on that migration now, or in about a week or two when I pick
> > > > > > > this up, assuming no changes are requested?
> > > > > >
> > > > > > I do not see any warning, neither about DM_SERIAL nor any other in RX-51
> > > > > > test job which builds bootable image for RX-51. See CI link:
> > > > > > https://dev.azure.com/u-boot/u-boot/_build/results?buildId=3580&view=logs&j=7f54f624-c8dd-54be-07eb-7f2dda6bba78&t=fc9e5b19-7cbc-51d2-a2e2-6e73893dece5
> > > > > >
> > > > > > But there is CONFIG_DM_SERIAL warning for nokia_rx51 board in the
> > > > > > Build the World omap job, see CI link:
> > > > > > https://dev.azure.com/u-boot/u-boot/_build/results?buildId=3580&view=logs&j=36d465ef-27e5-5934-99d6-cf4495f0bf3d&t=2b422a7f-7065-531c-aa25-e0171b04f5d1
> > > > >
> > > > > It's also visible when just doing "make CROSS_COMPILE=...
> > > > > nokia_rx51_config all" so I guess you need to debug your script and see
> > > > > why it's not showing that?
> > > > >
> > > > > > After this DM_KEYBOARD patch, I'm planning to look at DM_VIDEO as Simon
> > > > > > really wants it and then I can look at DM_SERIAL.
> > > > >
> > > > > OK.  But, are you going to start before or after this gets merged?
> > > > >
> > > > > --
> > > > > Tom
> > > >
> > > > As that cfb code shares both video and keyboard functionality, switching
> > > > to DM_VIDEO depends on working DM_KEYBOARD migration. So I can start
> > > > working on DM_VIDEO changes after I would see how support for
> > > > DM_KEYBOARD for n900 would look like. I do not want before because I
> > > > would enter into rebase, conflicts, rework and reimplementation from
> > > > zero circle (with which I have experience from past n900 u-boot patches).
> > >
> > > If the problem is reviews I can help with that. If it is lost patches
> > > then we should be able to get them all in now-ish. Can you resend
> > > everything that is outstanding.
> >
> > This is the only outstanding patch right now. All others were finally
> > reviewed and merged.
> >
> > > Yes you do need to put drivers in drivers/ so please create a Kconfig
> > > for your keyboard driver and put it in drivers/input
> >
> > But this is not easy and make it harder to develop and debug. All these
> > functionality shares lot of functions, variables and locks; so it cannot
> > be moved into drivers/input unless you want also watchdog or other code
> > in drivers/input. Not mentioning that this is specific board code which
> > is not on any other board, so I do not see any value of trying to invest
> > lot of time in this.
>
> So yes, board/nokia/rx51/rx51.c is 751 lines and should get split in to
> having watchdog under drivers/watchdog/ and video stuff under
> drivers/video/ and keyboard under drivers/input/ as that's how a modern
> board would be done.  I probably should have pushed back on such a large
> board file when things first came in as that's really always been the
> case, but there used to be more bad examples than there are now.  But at
> the end of the day, I really want to see this board brought up to modern
> APIs and it's not going to be used as the basis for a new board port, so
> there's a smaller risk of something new using this as an example.

Well if you are OK with it, it is fine with me.

I do wonder how this board got so out of date. E.g. even the DM_I2C
support is actually just veneered onto the old API.

Regards,
Simon