Oops, I misread the code, never mind :-) still looking because there are some reports of hangs (kern/52858 with a backtrace on sparc64, current-users without one on amd64), and your change is the only recent change that looks like it could affect both.