Tony Lindgren <tony@xxxxxxxxxxx> writes: > * Thomas Richard <thomas.richard@xxxxxxxxxxx> [231031 10:15]: >> Please find below the logs of the test you asked me. > > OK great thanks! > >> I added the call of pm_runtime_get_usage_count at the end of the suspend >> function. >> The console is attached on 2800000.serial, it has usage_count=4. >> Other serial has usage_count=3. > > So as suspected, it seems the power domain gets force suspended > somewhere despite the usage_count. I think the only way this could happen (excluding any bugs, of course) would be for pm_runtime_force_suspend() to be getting called somewhere, but I thought the earlier patch made that call conditional, so maybe there is another path where that is getting called? Kevin
