Re: [PATCH] sh_eth: ensure pm_runtime cannot suspend the device during init

[Ben Dooks <ben.dooks@codethink.co.uk>]

On 17/03/14 14:41, Laurent Pinchart wrote:
> Hi Ben,
>
> On Monday 17 March 2014 13:36:33 Ben Dooks wrote:
>
> [snip]
>
>>   From pm_runtime.h:
>>
>> static inline int pm_runtime_get_sync(struct device *dev)
>> {
>> 	return __pm_runtime_resume(dev, RPM_GET_PUT);
>> }
>>
>> static inline int pm_request_resume(struct device *dev)
>> {
>> 	return __pm_runtime_resume(dev, RPM_ASYNC);
>> }
>>
>> So it looks like pm_runtime_resume() does not protect against the
>> possibility that something else may re-suspend the device.
>
> Correct.
>
>> I have yet to ascertain how this ends up happening with device probe, it
>> seems to be very dependant on the code.
>
> We might be doing something wrong in the driver from a runtime PM point of
> view that leads to the device being suspended. I'd like to catch that instead
> of hiding it by a pm_runtime_get_sync() call. If it turns out that we're not
> doing anything wrong then replacing pm_runtime_resume() with
> pm_runtime_get_sync() would of course be fine.

My view is the pm_runtime)_resume() is just nasty, as it makes no
guarantees that the device cannot be re-suspended.


-- 
Ben Dooks				http://www.codethink.co.uk/
Senior Engineer				Codethink - Providing Genius