>> Generally, the old always points to the address of a ktermios structure, >> so old is unlikely to be NULL. Still check it before dereference as >> elsewhere. > >If it can not happen, there is no need to check it. So unless you can >prove that this ever could happen, this change isn't needed, right? > >thanks, > >greg k-h Thanks for reply. I agree with you. The reason for checking old there is that it is checked at the beginning in hso_serial_set_termios(). It is better to check it both or not, right? Delete it to avoid smatch complaints. drivers/net/usb/hso.c:1399 hso_serial_set_termios() error: we previously assumed 'old' could be null (see line 1389). diff --git a/drivers/net/usb/hso.c b/drivers/net/usb/hso.c index 4ad829ab54d3..383ff8b04473 100644 --- a/drivers/net/usb/hso.c +++ b/drivers/net/usb/hso.c @@ -1386,10 +1386,9 @@ static void hso_serial_set_termios(struct tty_struct *tty, struct hso_serial *serial = tty->driver_data; unsigned long flags; - if (old) - hso_dbg(0x16, "Termios called with: cflags new[%u] - old[%u]\n", - (unsigned int)tty->termios.c_cflag, - (unsigned int)old->c_cflag); + hso_dbg(0x16, "Termios called with: cflags new[%u] - old[%u]\n", + (unsigned int)tty->termios.c_cflag, + (unsigned int)old->c_cflag); /* the actual setup */ spin_lock_irqsave(&serial->serial_lock, flags);