On 03.08.2021 15:07, Rafael J. Wysocki wrote: > On Sat, Jul 31, 2021 at 11:36 AM Heiner Kallweit <hkallweit1@xxxxxxxxx> wrote: >> >> Seeing an erroneous use of pm_runtime_resume_and_get() in a patch I wonder >> whether we should annotate this function as __must_check. If the caller >> doesn't check the return code he doesn't know whether usage counter was >> bumped or not. Therefore I see a good chance that this results in a usage >> counter imbalance. > > Sounds reasonable. Please send a patch to make that change. It took some time to fix the callers that use pm_runtime_resume_and_get() w/o checking the return code. Last pending patch is [0], all other fixes are available on linux-next meanwhile. I'll submit the patch to annotate pm_runtime_resume_and_get() as __must_check, so that we hopefully can get this into 5.17 (incl. pending [0]). [0] https://lore.kernel.org/all/ed1ac2f8-5259-684d-42c8-effdf2920e78@xxxxxxxxx/T/
