[PATCH] ARM: cpu: Document and tweak clock-frequency property

[Peter Maydell]

On 8 December 2013 21:50, Mark Brown <broonie at kernel.org> wrote:
> On Sun, Dec 08, 2013 at 07:51:59PM +0000, Peter Maydell wrote:
>> On 8 December 2013 19:22, Mark Brown <broonie at kernel.org> wrote:
>> > (and doesn't reflect the actual deployed code reading the binding
>> > which does use this property without documentation outside ePAPR and does
>> > warn if it's absent)
>
>> This code should be fixed, then... (I'm pretty sure the kernel
>> has not always warned about the absence of this property,
>> or there would not be such a wide prevalence of DTs and
>> DT generating code which omitted it.)
>
> Too late for that, it's shipping in stable kernels.

> There does come a point where it's just nitpicking and not helpful but
> if it has a substantial effect on functionality then it's useful.  In
> this case suppressing the warning for non-asymmetric systems might be
> sensible.

Hmm, so "mandatory for non-symmetric [I assume you mean
that and not really 'non-asymmetric'?], otherwise optional" ?
I think that would be reasonable and preserve backwards
compatibility.

> For all practical purposes it is currently optional but the spec says
> it is mandatory.  I would rather err on the side of not changing the
> documentation in case someone does work based on ePAPR and/or an old
> kernel and since doing that keeps the spec more stable even if we do
> implement in a more tolerant fashion within Linux (as we should).

As I say, I don't think your specification currently does say
it is mandatory. If the documentation doesn't clearly list
it as a mandatory parameter, and a large number of
people writing DTS files or DT generation code haven't
put it in, and the kernel didn't complain about it not being
present for a long long time, then de facto it is optional,
and you should make your documentation conform with reality
and fix bugs where the kernel isn't coping with that.

thanks
-- PMM