[PATCH 1/2] ARM: dts: am335x: Replace numeric pinmux address with macro defines

Tony Lindgren tony at atomide.com
Thu May 3 09:19:16 PDT 2018


* Grygorii Strashko <grygorii.strashko at ti.com> [180503 15:54]:
> 
> 
> On 04/30/2018 12:31 PM, Tony Lindgren wrote:
> > Hi,
> > 
> > * Christina Quast <cquast at baylibre.com> [180430 11:23]:
> >> The values are extraced from the "AM335x SitaraTM Processors Technical
> >> Reference Manual", Section 9.3.1 CONTROL_MODULE Registers, based on the
> >> file autogenerated by TI PinMux.
> > 
> > This certainly makes things easier to mux :)
> > 
> > Have you verified that the registers are the same across all am335x
> > models and different revisions?
> > 
> > It used to be that different SoC revisions could have different
> > amount of registers and also different options in some cases.
> > 
> >> +#define AM335X_CONTROL_REVISION			0x0
> >> +#define AM335X_CONTROL_HWINFO			0x4
> >> +#define AM335X_CONTROL_SYSCONFIG		0x10
> >> +#define AM335X_CONTROL_STATUS			0x40
> > 
> > You should only list the padconf mux registers here, no need to
> > list any of the controller registers.
> > 
> 
> To be honest, I do think it's right thing to do - DT by itself is
> documentation and previously DT maintainers were not very happy regarding
> introducing more defines instead of const.

Sounds like you mean "I don't think" above instead of "I do think"?
Care to clarify..

> Adding such defines will introduce big headers in Linux common or  
> platform folders again which we've just got rid of. 
> 
> If smth is unclear - comments can be used in DT.
> 
> Just my 5c.

The problem earlier was that we had just too many variants as
the padconf registers got changed even between SoC revisions.
Not always and not for many registers but still. The padconf
register range seems to stay the same for a SoC though.

I do see value at being able to mux the registers easier though.

Regards,

Tony



More information about the linux-arm-kernel mailing list