13:58 <tomba> I have a question related to device power management. I was under the impression that the device hardware is enabled only when a driver chooses to enable it, but I noticed that the driver core seems to do some part of this automatically.

14:00 <tomba> In this specific case what I observe is that if the AM62 DSS is enabled by the bootloader, and the kernel driver's probe is called, but it immediately returns -EPROBE_DEFER, the DSS device is turned off (as shown by k3conf).

14:01 <tomba> I presume that's related to the power-domain, so the driver core handles the power-domain automatically, enabling it before the probe is called, and disabling it after driver remove has been called, perhaps?

14:02 <tomba> This is rather problematic as the DSS hardware will go into a bad state if it's powered off while it is active.

14:03 <tomba> So... Do my observations/points above make sense? Is that how power-domains work? Would it be possible for the framework to keep the current state of the power-domain if the driver probe fails?