On Wed 21-06-23 15:11:25, Sebastian Andrzej Siewior wrote: > On 2023-06-21 13:49:06 [+0200], Michal Hocko wrote: > > > On PREEMPT_RT what you can happen is that the writer is preempted by a > > > high-priority reader which then deadlocks because the reader spins while > > > waiting and the writer is blocked. For this issue we have lock + unlock > > > in the seq reader to PI boost the seq writer so it can make progress. > > > > Please state the the problem explicitly in the changelog. You are > > marking this patch as a fix so the underlying issue should be stated. > > The problem is the "local_irq_save()" which I believe I stated. The lock > + unlock story was just a side story and is already covered. I really > need just the local_irq_save() invocation to be part of the seqlock API > so it can be substituted away. I really do not want to nitpick but your changelog states: "This is troublesome and leads to problems on PREEMPT_RT because the inner spinlock_t is now acquired with disabled interrupts." I believe it would be benefitial to state why htis is troublesome because not everybody has insight into PREEMPT_RT and all the consequences. Thanks! -- Michal Hocko SUSE Labs
