[PATCH v2] ARM: DT: binding fixup to align with vendor-prefixes.txt

Stephen Warren swarren at wwwdotorg.org
Fri Aug 9 15:14:47 EDT 2013


On 08/09/2013 12:49 PM, Christian Daudt wrote:
> [resend in plain-text]
> 
> 
> On 2013-08-09 9:11 AM, "Stephen Warren" <swarren at wwwdotorg.org> wrote:
>>
>> On 08/06/2013 03:40 PM, Christian Daudt wrote:
>>> On 13-08-05 09:21 PM, Stephen Warren wrote:
>>>>>>>    Required root node property:
>>>>>>>    -compatible = "bcm,bcm11351";
>>>>>>> +compatible = "brcm,bcm11351";
>>>>>> In a patch of mine that deprecated a property, Mark wondered if it
>>>>>> would
>>>>>> make sense to mention the old deprecated DT content simply to document
>>>>>> that it existed, so that old DTs would still make sense when checking
>>>>>> the documentation. I wonder if the same argument applies to this patch?
>>>>> I would think the opposite. Deprecated items should be dropped from
>>>>> documentation. They are in the code (for a holdover period) but clearly
>>>>> marked as deprecated. No one should be extending the life of these, and
>>>>> adding documentation on it is a step in the wrong direction of making it
>>>>> easier for it to linger beyond what it should.
>>>> The deprecated stuff will have to be fully documented once the DT schema
>>>> validation is in place...
>>>>
>>> This deprecated code should be short lived, given that in actual fact it
>>> is actually quite unnecessary since no boards exist that rely on it.
>>
>> Is this patch for v3.11-rc* or v3.12?
>>
> I'm guessing it's too late for 3.11 at this point.
> 
>> If it's for v3.12, then I see that v3.11 will be released with a variety
>> of users of the old compatible values, hence the old compatible value is
>> an ABI, and hence we should continue to support and document it (as
>> deprecated).
>>
> I think whether bindings automatically become ABI at kernel release is
> still an open topic. And as I mentioned in this case we are the only
> ones affected and we don't have a problem with the change.
> But if that's the case then there's no point to this patch. I'll just
> add bcm to vendor-prefixes and be done with it.
> I'm okay either way. Just need to know what direction to take asap so
> I can stop telling devs to keep changing back and forth...

I think it's fine to fix the issue; we should just do so in the trivial
way that maintains backwards-compatibility, and allows people who
compare the binding document to old DTs to understand how the correlate
to each-other.




More information about the linux-arm-kernel mailing list