On Wed, Apr 13, 2022 at 12:29:38AM -0500, Bob Pearson wrote: > rxe_mcast.c currently uses _irqsave spinlocks for rxe->mcg_lock > while rxe_recv.c uses _bh spinlocks for the same lock. Adding > tracing shows that the code in rxe_mcast.c is all executed in_task() > context while the code in rxe_recv.c that refers to the lock > executes in softirq context. This causes a lockdep warning in code > that executes multicast I/O including blktests check srp. What is the lockdep warning? This patch looks OK, but irqsave should be a universal lock - so why does lockdep complain? Is there nesting? Jason
