Hi Sebastian, On Tue, May 07, 2024 at 11:54:07AM GMT, Sebastian Andrzej Siewior wrote: > I compared mine outcome vs v4.19-rt-next and the diff at the bottom came > out: > > - timer_delete_sync() used to have "#if 0" block around > lockdep_assert_preemption_enabled() because the function is not part > of v4.19. You ended up with might_sleep() which is a minor change. > Your queue as of a previous release had the if0 block (in > __del_timer_sync()). > I would say this is minor but looks like a miss-merge. Therefore I > would say it should go back for consistency vs previous release and > not change it due to conflicts.
Makes sense. > - The timer_delete_sync() is structured differently with > __del_timer_sync(). That function invokes timer_sync_wait_running() > which drops two locks which are not acquired. That is wrong. It should > have been del_timer_wait_running(). Understood. I was a bit strungling here. Glad you caught this error. > I suggest you apply the diff below to align it with later versions. It > also gets rid of the basep argument in __try_to_del_timer_sync() which > is not really used. Will do. > As an alternative I can send you my rebased queue if this makes it > easier for you. Yes please, this makes it easy to sync the rebase branch. Thanks a lot! Cheers, Daniel
