[PATCH RFC PoC 0/2] platform: different approach to early platform drivers

[Bartosz Golaszewski]

2018-04-27 10:55 GMT+02:00 Arnd Bergmann <arnd at arndb.de>:
> On Fri, Apr 27, 2018 at 10:29 AM, Sekhar Nori <nsekhar at ti.com> wrote:
>> On Friday 27 April 2018 01:22 PM, Arnd Bergmann wrote:
>>> On Fri, Apr 27, 2018 at 4:28 AM, David Lechner <david at lechnology.com> wrote:
>>>> On 04/26/2018 12:31 PM, Rich Felker wrote:
>
>>>
>>> I haven't seen the discussion about your clock drivers, but I know that
>>> usually only a very small subset of the clocks on an SoC are needed
>>> that 'early', and you should use a regular platform driver for the rest.
>>
>> Its true that the subset is small, but they are either PLL bypass clocks
>> or clocks derived out of the main clock gate controller on the Soc
>> (DaVinci PSC). So we need some non-platform-device based initialization
>> support in the two main clock drivers used on mach-davinci anyway.
>>
>>> Can you elaborate on which devices need to access your clocks before
>>> you are able to initialize the clk driver through the regular platform_driver
>>> framework? Do any of these need complex interactions with the clk
>>> subsystem, or do you just need to ensure they are turned on?
>>
>> Its just the timer IP. There is no complex interaction, just need to
>> ensure that the clock is registered with the framework and also switched
>> on when there is a gate clock.
>>
>> The latest attempt by David for this was posted last night here:
>>
>> https://lkml.org/lkml/2018/4/26/1093
>
> Ok, so the workaround in that series is to set up the timer clk manually
> in the SoC specific code (dm365_init_time etc) and register it to the
> clk subsystem, right?
>
> That seems to be a much more appropriate workaround than adding
> back early_platform devices. We can always try to do something better
> later, but for now I'm happy with that as a workaround.
>

Just to clarify: this is not bringing back the hacky early platform
devices from before. It's a new approach that actually uses the linux
platform driver model (unlike the current mechanism which only uses
the same data structures).

Thanks,
Bart