On 2023-06-21 14:40:50 [+0200], Petr Mladek wrote: > I think that we should explicitly disable preemption in > printk_deferred_enter(). The disabled interrupts are not > strictly necessary. > > I am going to resurrect the patch > https://lore.kernel.org/r/20230419074210.17646-1-pmladek@xxxxxxxx > and add the preempt_disable()/enable() at the same time. > > Is this going to work for you? No. migrate_disable() is the minimal thing that is needed. However it is heavy weight comparing to what is required. Also by disabling preemption we still have the problem later on with the seqlock_t. Let me swap the lockdep annotation and we should be good. > Best Regards, > Petr Sebastian
