[PATCH] soc: ti: wkup_m3_ipc: switch to using remoteproc OF infrastructure

Rob Herring robh+dt at kernel.org
Fri Aug 19 14:21:09 PDT 2016


On Thu, Aug 18, 2016 at 4:30 PM, Suman Anna <s-anna at ti.com> wrote:
> Hi Rob,
>
> On 08/16/2016 09:54 AM, Rob Herring wrote:
>> On Fri, Aug 12, 2016 at 11:43 AM, Bjorn Andersson
>> <bjorn.andersson at linaro.org> wrote:
>>> On Fri 12 Aug 09:00 PDT 2016, Suman Anna wrote:
>>>
>>>> On 08/12/2016 06:02 AM, Lee Jones wrote:
>>>>> On Thu, 11 Aug 2016, Suman Anna wrote:
>>>>>

[...]

>>>> Once "rprocs" hits mainline, I will definitely switch over the
>>>> wkup_m3_ipc nodes to use that standard property and can fix this driver
>>>> independently.
>>>>
>>>
>>> We're stuck with this problem all over the place, as the world continues
>>> to evolve we will have issues with DT being static. This has been
>>> discussed many times before and the suggested solution is always what
>>> you implemented here - make the code deal with both versions, preferably
>>> by patching.
>>>
>>> The fact that you had to export the of_ operations indicates that no-one
>>> has tried this before and I'm happy you did. I'm however not happy about
>>> the size of the chunk of code it takes to do this dance.
>>>
>>> I think for this to be practical we need to provide higher level
>>> operations for DT modification - in this case a of_rename_property().
>>>
>>> @Rob, any comments on this?
>>
>> I agree. Pantelis submitted some helpers in this area a while back
>> (for the changeset API IIRC). I believe they were mostly fine, but
>> needed some users.
>>
>
> Is this the series you are talking about?
> http://marc.info/?l=devicetree&m=146341689512653&w=2
>
> It looks like that series is effective only when OF_DYNAMIC is enabled.

Yes, as changeset API allows adding/removing nodes. It's probably just
a matter of time until OF_DYNAMIC is always enabled.

> Probably a dumb question, but is this limited to DT Overlays?

No, changesets are the mechanism overlays use to apply changes.

> This
> particular usage is a one-time change (first-time module is insmod'd)
> mainly to provide compatibility for older DTBs, thereafter we wouldn't
> be required to make any changes. It is definitely not a bulk update and
> we don't want to unroll the changes even if we removed the module.

If you only need property changes, then we could do similar helpers to
make it easier for callers.

Rob



More information about the linux-arm-kernel mailing list