[PATCH 00/33] ARM: at91: Switch remaining SoCs and boards to CCF

[Arnd Bergmann]

On Tuesday 01 July 2014 19:56:01 Alexandre Belloni wrote:
> On 01/07/2014 at 19:01:12 +0200, Arnd Bergmann wrote :
> > On Tuesday 01 July 2014 18:48:51 Alexandre Belloni wrote:
> > > On 01/07/2014 at 18:35:50 +0200, Arnd Bergmann wrote :
> > > > If I read this right, disabling 'USE_OF' results in still using
> > > > the old clock interface. Is that intentional? If you want to always
> > > > use COMMON_CLK now, the above can probably be simplified to
> > > > 
> > > > config AT91_PMC_UNIT
> > > >         def_bool !ARCH_AT91X40
> > > >       select COMMON_CLK
> > > > 
> > > > and all references to AT91_USE_OLD_CLK, COMMON_CLK_AT91 and
> > > > OLD_CLK_AT91 be removed.
> > > 
> > > Yeah, this is intentional we still have 30 or so boards that we have to
> > > migrate to DT. Until then, we want to be able to boot a kernel without
> > > CCF. The next step in the migration is to switch those boards to DT,
> > > possibly using an hybrid DT/pdata approach until we can get rid of the
> > > board files. Then, we will be able to get rid of the old clk
> > > implementation (and a few other drivers still living in mach-at91).
> > 
> > Ok, I see. Do you expect those 30 boards to work if COMMON_CLK_AT91
> > is enabled, or are there known problems with that?
> > 
> > If you have reason to believe it doesn't work, maybe you can expose
> > the selection of the clock interface to users, e.g. like
> > 
> 
> We actually do the opposite, when a board without DT support is
> selected, the old clock support is also selected even when also
> selecting CONFIG_OF.

Ok, got it now. Yes, that makes sense.

I guess we can start doing the actual multiplatform enablement now,
which would be only for the DT-enabled boards -- all of Kconfig.non_dt
still has to depend on !ARCH_MULTIPLATFORM but we can make at91
either be built standalone (with board files) or multiplatform and
still support all SAM9 and RM9200 SoCs but without board files.

	Arnd