[GIT PULL 2/4] TI DaVinci SoC support updates for v4.11

[khilman@baylibre.com (Kevin Hilman)]

Olof Johansson <olof@lixom.net> writes:

> On Tue, Jan 24, 2017 at 06:36:39PM +0530, Sekhar Nori wrote:
>> The following changes since commit 7ce7d89f48834cefece7804d38fc5d85382edf77:
>> 
>>   Linux 4.10-rc1 (2016-12-25 16:13:08 -0800)
>> 
>> are available in the git repository at:
>> 
>>   git://git.kernel.org/pub/scm/linux/kernel/git/nsekhar/linux-davinci.git tags/davinci-for-v4.11/soc
>> 
>> for you to fetch changes up to 9c9b1bc25291e275b04f758f2549c81e092954f5:
>> 
>>   ARM: davinci: add skeleton for pdata-quirks (2017-01-23 20:52:38 +0530)
>> 
>> ----------------------------------------------------------------
>> Introduce pdata-quirks mechanism for
>> DaVinci which will be needed for supporting
>> video input/output on DA850 SoC.
>> 
>> ----------------------------------------------------------------
>> Kevin Hilman (1):
>>       ARM: davinci: add skeleton for pdata-quirks
>
> I'm not much of a fan of adding this. I know we added it for some of
> the legacy platforms since moving over existing in-tree code would take
> a while. 

We are not fans of adding it either, and have been avoiding it up until
now.  You can rest assured that new additions here will be heavily
scrutinized.

> But the thought was that new drivers wouldn't use it much.  Is the
> need for this now due to said drivers never been ported over, and now
> also being used by Davinci? Or are new drivers added that rely on
> platform data?

The problem isn't new drivers, but old drivers that haven't grown DT
equivalents.

There are a handful of drivers/media drivers currently used by Davinci
via board files (and pdata) that still do not have DT mechanisms for
all the things used in pdata.

In attempting DT support for video on DaVinci, attempts were made (by
me[1]) to pass the data via DT, which triggered discussions among
linux-media maintainers about the "right way" of handling it.
Unfortunately, these discussions never fully materialized in an
short-term solution, so we were left with pdata for now.

I've offered to be the first to beta test the new DT mechanisms they
come up with, so when that happens, we'll gladly get rid of these pdata
quirks.

Kevin

[1] https://marc.info/?l=devicetree&m=147982998517384