pm: device_runtime: Optimize pm_device_runtime_usage

[nordic-krch]

There is no point in using lock or semaphore to read current usage counter as it may change after unlocking or giving back the semaphore. Value can only be trusted in the controlled environment (e.g. test).

[bjarki-andreasen]

While I agree the count should only be used in controlled environment, it is also used in the device_service.c shell :) Could you add a commit removing the access and printing of usage count from device_service.c as well since use here requires the call to be "safe"?

[nordic-krch]

Could you add a commit removing the access and printing of usage count

I'm not sure if I understand. Why no leave it as it is? Imo this does not change much for that command. pm_device_state_get and then pm_device_runtime_usage are get without any lock so it is possible that some device operation will end or start between and the won't match. I could maybe add lock around reading both then report would always be in sync.

[nashif]

@ceolin please review as this removed code you introduced.

[ceolin]

I don't oppose to this change but I don't see much benefit as well. This function was introduced because of a shell functionality and it is used there and in a test (in the repo). The lock ensures the correctness of the count at the time it was called, obviously as soon as it is retrieved there is not guarantee that it will change.

[ceolin]

Could you add a commit removing the access and printing of usage count

I'm not sure if I understand. Why no leave it as it is? Imo this does not change much for that command. pm_device_state_get and then pm_device_runtime_usage are get without any lock so it is possible that some device operation will end or start between and the won't match. I could maybe add lock around reading both then report would always be in sync.

I didn't get the point too. The shell is just informative and this change does not impact this functionality.