> I have reached a deadlock caused by the fact that the function __free_mads > takes h_al->mad_lock > > It then calls ib_put_mad which calls al_remove_mad that will try to take > the same lock. How did this not blow up 4 years ago?