On Mon, Feb 20, 2017 at 5:29 AM, Andrey Konovalov <andreyk...@google.com> wrote: > other info that might help us debug this: > > Possible unsafe locking scenario: > > CPU0 CPU1 > ---- ---- > lock(&(&pcpu->lock)->rlock); > lock(&(&list->lock)->rlock#3); > lock(&(&pcpu->lock)->rlock); > lock(&(&list->lock)->rlock#3); >
They are different types of sockets and different lists of skb's, one is netlink socket the other is udp socket, so I don't think we could have a deadlock in this scenario, we probably need to explicitly mark them as different lockdep classes.
