Re: [PATCH 1/3] dt-bindings: added new pwm-sifive driver documentation

[Rob Herring <robh@kernel.org>]

On Mon, Apr 30, 2018 at 12:45:20PM +0200, Andreas Färber wrote:
> Am 30.04.2018 um 10:19 schrieb Thierry Reding:
> > On Sun, Apr 29, 2018 at 02:08:07PM -0700, Wesley Terpstra wrote:
> >> On Sun, Apr 29, 2018 at 2:01 PM, Andreas Färber <afaerber@suse.de> wrote:
> >>> "pwm0" sounds like a zero-indexed instance of some pwm block. If 0 is
> >>> the version here, I'd suggest to make it "pwm-0" for example - you might
> >>> want to take a look at the Xilinx bindings, which use a strict x.yy suffix.
> >>
> >> That's fine. I'll change it to pwm-0.00 in the next patch series.
> > 
> > This should match the version that you use. If you're internal
> > versioning uses single digits, or a single version number, then I think
> > there's no need to use 0.00, because that would just be confusing.
> > However I think it'd be good to make sure it is discernible as a version
> > number. Perhaps something like sifive,pwm-v0. That seems to be a fairly
> > common scheme.
> 
> Yes. My point was not to adopt another vendor's versioning scheme but to
> adopt _some_ consistent scheme and document it, e.g., in a sifive.txt
> similar to xilinx.txt:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/Documentation/devicetree/bindings/xilinx.txt
> 
> It should be made clear what in the compatible string the version is
> (thus my proposal of using a dash as separator), and there you may want
> to document how to map between IP/documentation and compatibles for any
> new bindings.

Yes. And using versions in compatible strings is only accepted when 
there is a well defined versioning process. FPGAs tend to be the main 
case as most SoC vendors don't have rigorous versioning processes. I 
guess it makes sense for SiFive from the little I know about them. What 
doesn't make sense or get accepted is software folks just making up v1, 
v2, v3, etc.

Rob