> But then I fail to see how usb_driver_release_interface on the currently > disconnect()-ed interface can be anything more than an noop. Or do you > encounter the same problem when you exclude the current interface from > the list of released interfaces? Excluding the current interface makes no difference. D.