[PATCH 1/5] drm/omap: Fix suspend resume regression after platform data removal

[Grygorii Strashko]

On 17/06/2020 09:04, Tomi Valkeinen wrote:
> On 16/06/2020 19:56, Grygorii Strashko wrote:
>>
>>
>> On 16/06/2020 18:30, Tony Lindgren wrote:
>>> * Tomi Valkeinen <tomi.valkeinen at ti.com> [200616 13:02]:
>>>> On 11/06/2020 17:00, Grygorii Strashko wrote:
>>>>> I think, suspend might be fixed if all devices, which are now child of ti-sysc, will do
>>>>> pm_runtime_force_xxx() calls at noirq suspend stage by adding:
>>>>>
>>>>>       SET_NOIRQ_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
>>>>>                         pm_runtime_force_resume)
>>>>>
>>>>> Am I missing smth?
>>>>
>>>> Isn't this almost exactly the same my patch does? I just used suspend_late
>>>> and resume_early. Is noirq phase better than late & early?
>>>
>>> Well up to you as far as I'm concerned. The noirq phase comes with serious
>>> limitations, for let's say i2c bus usage if needed. Probably also harder
>>> to debug for suspend and resume.
>>
>> Unfortunately, you can't use PM runtime force API at .suspend() stage as pm_runtime_get will still work and
>> there is no sync between suspend and pm_runtime.
>> The PM runtime force API can be used only during late/noirq as at this time pm_runtime is disabled.
> 
> Yes, but which one... Do you know what the diff is with late/noirq from driver's perspective? I guess noirq is atomic context, late is nto?

noirq is *not* atomic, jus IRQs (non-wakeup) will be disabled (disbale_irq())

> 
> Dispc's suspend uses synchronize_irq(), so that rules out noirq. Although the call is not needed if using noirq version, so that could also be managed with small change. But I wonder if there's any benefit in using noirq versus late.




-- 
Best regards,
grygorii