Re: [PATCH 17/25] serial: sh-sci: Correct SCIF type on R-Car for BRG

[Simon Horman <horms@verge.net.au>]

On Thu, Dec 10, 2015 at 10:21:27AM +0100, Geert Uytterhoeven wrote:
> On Fri, Nov 20, 2015 at 4:33 PM, Laurent Pinchart
> <laurent.pinchart@ideasonboard.com> wrote:
> > On Friday 20 November 2015 16:30:22 Geert Uytterhoeven wrote:
> >> On Fri, Nov 20, 2015 at 3:52 PM, Laurent Pinchart wrote:
> >> > On Friday 20 November 2015 08:46:56 Geert Uytterhoeven wrote:
> >> >> On Thu, Nov 19, 2015 at 9:55 PM, Laurent Pinchart wrote:
> >> >>> On Thursday 19 November 2015 19:38:56 Geert Uytterhoeven wrote:
> >> >>>> The "renesas,scif" compatible value is currently used for the SCIF
> >> >>>> variant in all Renesas SoCs of the R-Car family.  However, the variant
> >> >>>> used in the R-Car family is not the common "SH-4(A)" variant, but a
> >> >>>> derivative with added "Baud Rate Generator for External Clock" (BRG),
> >> >>>> which is also present in sh7734.
> >> >>>
> >> >>> Time to introduce a "renesas,scif-rcar" compatible string ? ;-)
> >> >>>
> >> >>> As the only DT-enabled platform to have a different SCIF type is
> >> >>> r7s72100 we could also consider just switching the regtype to
> >> >>> SCIx_SH4_SCIF_BRG_REGTYPE for the generic "renesas,scif" entry as it's
> >> >>> listed after the "renesas,scif- r7s72100" entry. That might cause an
> >> >>> issue if we want to enable DT on arch/sh though, but even if that
> >> >>> happens due to the J-Core processors I'd be surprised to see the old
> >> >>> Renesas SH platforms being moved to DT.
> >> >>
> >> >> I thought about that, but you never know in which out-of-tree BSP it
> >> >> ended up being used, too. So better safe than sorry.
> >> >
> >> > Out-of-tree should be banned :-)
> >> >
> >> > More seriously, I suppose you wouldn't be thrilled by the idea of a
> >> > "renesas,scif-rcar-gen2" ?
> >>
> >> Nope. Note that it's also used in R-Car Gen 1 and Gen 3, and sh7734.
> >
> > Yes, but it would at least cover the whole Gen2 family that behaves the same
> > way. And wouldn't preclude adding "renesas,scif-rcar-gen1". That's two compat
> > strings only.
> 
> In light of all the recent "add fallback compatibility strings" patch series
> from Simon, perhaps I should reconsider, and just match against three (new)
> family-specific compatible values:
> 
>     "renesas,scif-rcar-gen1"
>     "renesas,scif-rcar-gen2"
>     "renesas,scif-rcar-gen3"
> 
> instead of the 8 (and more coming) SoC-specific compatible values?
> 
> Following that scheme means we will have to add many compatible values
> to the existing dtsis. I.e. every SCIx device node (there are more than 100)
> will have 3, like
> 
>         scif0: serial@e6e60000 {
>                 compatible = "renesas,scif-r8a7791", "renesas,scif-rcar-gen2",
>                               "renesas,scif";
> 
> Not having the SoC-specific ones in the driver won't cause an issue when using
> an old DTS with a new kernel: you can't use the new BRG features without
> adding the extra clocks to the DTS anyway, so you can add the family-specific
> compatible value when doing that update.
> 
> Simon, what do you think?

This sounds good to me.
In fact it sounds like the kind of upgrade path that I would like
to see in action.