> On Apr 18, 2019, at 16:50, Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > >> normally without any errors, yet the thermal-zone I defined, which is there to >> prevent my system from burning out, would remain incomplete (without required >> cooling device) and yet no errors came out of the kernel. >> >> I just tested it and that's exactly what's happening! Yikes. >> > You would still get a warning message. Better than failing to load > the driver completely. On top of that, this used to work before. > Plus, you said it only happens in severe situations, ie if the kernel > is out of memory, and in thact case there would be a traceback. > Sorry, I don't get your point. Ok, I see. We're having the same discussion than on the v3 submission then. If I change the dev_err to a dev_warn, but still return 0 on thermal_of_cooling_device_register failure, you would apply the patch. It sounds like a compromise. Completely failing the probe call or partially succeed it results in the same behaviour for thermal cooling use cases. But the dev_warn definitely becomes important though. I will v4 with that if you agree and leave the devm_ thermal register for another day since I agree the series might be delayed too long (or indefinitely). Cheers!
