[PATCH v8 07/16] drivers: reset: Add STM32 reset driver

[Maxime Coquelin]

2015-05-21 19:58 GMT+02:00 Andreas Färber <afaerber at suse.de>:
> Am 21.05.2015 um 09:46 schrieb Maxime Coquelin:
>> 2015-05-21 1:45 GMT+02:00 Andreas Färber <afaerber at suse.de>:
>>> Am 09.05.2015 um 09:53 schrieb Maxime Coquelin:
>>>> +static const struct of_device_id stm32_reset_dt_ids[] = {
>>>> +      { .compatible = "st,stm32-rcc", },
>>>> +      { /* sentinel */ },
>>>> +};
>>>> +MODULE_DEVICE_TABLE(of, sstm32_reset_dt_ids);
>>>
>>> Typo.
>>
>> Indeed, thanks for pointing this out.
>>
>>>
>>> IIUC the timer depends on the reset controller, so it must be built in
>>> anyway, and that's what's enforced in the Makefile above. Drop the line?
>>>
>>
>> Agree it must be built-in.
>> I will fix that in next version.
>
> Actually, I've updated a timer implementation of mine to invoke a reset
> controller similar to how you do in the STM32 clocksource patch, but no
> reset controller is getting returned.
>
> It seems to me, you are working around that by simply ignoring the error
> in the timer code and not doing a reset then, so the STM32 timer does in
> fact _not_ depend on the reset controller? What happened to your efforts
> of making the reset controller usable for the timer? In my case, my
> timer is originally in reset state and needs to be deasserted, so I
> can't just ignore it.

Indeed, I made the reset optionnal in the clocksource patch since v3.
Rob and Arnd said a lot of platform relies on such things are done by
the bootloader [0].
I decided to deassert timers reset at bootloader stage, and make it
optionnal in clocksource driver.
I made it optionnal in case we decide one day to move reset
initialization before timer are initialized.

Note that for now, I still use your bootloader.
I have done the changes to reset the timers in the afboot-stm32.
That's the reason why I asked you under which licence it is delivered
few months ago.
I can share you the patch if you want, even if I understand it is more
about the concept that you are reluctant.

On my side, I plan to move to U-Boot soon, as Kamil Lulko added STM32
support in mainline [1].
In case of U-Boot, the timer reset should be de-asserted when jumping
into the Kernel, as Rob mentionned [0].

Kind regards,
Maxime

[0]: https://lkml.org/lkml/2015/3/10/737
[1]: http://u-boot.10912.n7.nabble.com/PATCH-v2-stm32f4-fix-serial-output-td212812.html