Re: [dccp] A quick question about sequence validity windows

[Arnaldo Carvalho de Melo <acme at ghostprotocols.net>]

On 11/30/05, Eddie Kohler <kohler at cs.ucla.edu> wrote:
> Yep!  Specifically, we assume a receive buffer.
>
> Eddie
>
> Xiaofeng Han wrote:
> > Hello Eddie
> >
> > thanks for your reply.

Using the Linux implementation present in the 2.6.14 kernel sources we have:

> > In my question, the lower bound I mean is GSR+1-(W/4), since new
>> data is coming, the GSR increases, some data with sequence #
>> smaller than the new GSR+1-(W/4) becomes out of the range of the
>> window.

this is done in net/dccp/input.c, function dccp_check_seqno()

> > If I understand correctly, there is actually another buffer to store the
> > data, the sequence validity windows is something like a filter to
> > control the incoming data, but it has nothing to do with the data in the
> > other buffer, is this understanding correct?

The name of this buffer in Linux is struct sock->sk_receive_queue,
fed mainly by dccp_rcv_established(), in ./net/dccp/input.c:

                /* Remove the DCCP header */
                __skb_pull(skb, dh->dccph_doff * 4);
                /* Queue the packet for later processing */
                __skb_queue_tail(&sk->sk_receive_queue, skb);
                /* Account the packet to its socket */
                skb_set_owner_r(skb, sk);
                /* Wake up upper layers waiting for packets */
                sk->sk_data_ready(sk, 0);

For a mostly complete callgraph of the packet flow please take a look
at:

http://oops.ghostprotocols.net:81/dccp/ostra/rx/
http://oops.ghostprotocols.net:81/dccp/ostra/tx/

Best Regards,

- Arnaldo