Re: [PATCH 02/24] Documentation: DT bindings: add more chip compatible strings for Tegra PCIe

[Rob Herring]

On Thu, Jan 29, 2015 at 11:08 AM, Paul Walmsley <paul@xxxxxxxxx> wrote:
> Hi Rob
>
> On Thu, 29 Jan 2015, Rob Herring wrote:
>
>> On Thu, Jan 29, 2015 at 9:45 AM, Paul Walmsley <paul@xxxxxxxxx> wrote:
>> > On Thu, 29 Jan 2015, Rob Herring wrote:
>> >
>> >> On Wed, Jan 28, 2015 at 5:49 PM, Paul Walmsley <paul@xxxxxxxxx> wrote:
>> >> >
>> >> > Add compatible strings for the PCIe IP blocks present on several Tegra
>> >> > chips.  The primary objective here is to avoid checkpatch warnings,
>> >> > per:
>> >> >
>>
>> [...]
>>
>> >> > +  - "nvidia,tegra132-pcie" (not yet matched in the driver)
>> >> > +  - "nvidia,tegra210-pcie" (not yet matched in the driver)
>> >>
>> >> Whether the driver matches or not is irrelevant to the binding and may
>> >> change over time. Does this mean the driver matches on something else
>> >> or Tegra132 is not yet supported in the driver?
>> >
>> > It means that the driver currently matches on one of the first three
>> > strings that don't carry that annotation.
>> >
>> >> If the former, what is important is what are the valid combinations of
>> >> compatible properties and that is not captured here. In other words,
>> >> what is the fallback compatible string for each chip?
>> >
>> > The intention was to try to be helpful: to document that anyone adding a
>> > "nvidia,tegra132-pcie" compatible string would also need to add one of the
>> > other strings as a fallback.  Would you like that to be documented in a
>> > different way, or removed?
>>
>> Then you should say something like 'must contain "nvidia,tegra20-pcie"
>> and one of: ...'
>>
>> You can also use nvidia,<chip>-pcie if you want. checkpatch will check
>> for that pattern too. Then your documentation can be something like:
>>
>> Must contain '"nvidia,<chip>-pcie", "nvidia,tegra20-pcie"' where
>> <chip> is tegra30, tegra132, ...
>>
>> We don't enforce that the <chip> part is documented ATM and not likely
>> until we have a schema if ever.
>
> OK, thanks for the explanation.
>
> So would it be acceptable to you to skip the attempt to document which
> strings are actually supported by the current driver, and to simply use
> the <chip> wildcard?

I don't think the binding document should say anything about what the
driver uses or not. It should describe what combinations of compatible
strings in a dts are valid. I didn't look at every patch, but the ones
like this one are. I'd be fine with most of this all in one patch BTW.

You should attempt to document known values of <chip> if you use it
(you could refer to another doc for the list). I was only highlighting
what you can get away with if no one is paying attention, not that you
don't need to add tegra132, etc. :)

Rob
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/