In some email I received from Bill Sommerfeld, sie wrote:
> This looks like an implementation oversight rather than intentional
> behaivor...
> 
> ... but it appears that only tcp_reass() checks SS_CANTRECVMORE before
> putting an mbuf into a socket's receive queue (and not any of the
> other paths, such as udp_input() or raw_input().
> 
> Anyone understand why this is the case? it seems clearly wrong to not
> drop the recieved packet on the floor when the application has told
> the kernel it's not interested in receiving further data from the
> socket..

It might be an ACK for data that's been sent ?

Darren