Hi Petr, On Wed, May 18, 2022 at 10:54:22AM +0200, Petr Mladek wrote: > > + spin_lock_irqsave(&filling, flags); > > I thought more about this and there is a small risk of a deadlock > when get_random_bytes() or queue_work() or NMI calls > printk()/vsprintf() with %p here. > > A simple solution would be to use trylock(): > > if (!spin_trylock_irqsave(&filling, flags)) > return -EDEADLK; > > Could we do this change, please? > > I do not mind if it will be done by re-spinning the original > patch or another patch on top of it. Interesting consideration. Sure, I'll do exactly that and send a v2. Jason