Re: [lwip-users] LwIP TCP reset, sequence number bug?

[Matias Mandell]

On 9/2/11 1:45 PM, Simon Goldschmidt wrote:
> >      I'm really sorry that I bother you by sending a personal email but
> > you seem to be involved in the LwIP development and this is such a
> > small
> > bug.
> Nevertheless, I'm not the only developer/user, and others might be interested in this, 
> too, so please always send mails as this one to one of the lists. It's not the 
> "bothering me" but the fact that other people might gain something from reading 
> this, too.
Yes, I know. Completely forgot about the mailing list. I was looking at 
the bug reporting system. But it seemed too time-consuming for the 
moment. :)
> >      Is there a sequence number bug at TCP reset? According to RFC the
> > sequence number should not be increased by one on reset.
> >
> >      I think the "+1" should be removed from tcp_listen_input() call
> >      tcp_rst(ackno + 1, seqno + tcplen,
> Funny that you found that just now: we discovered this bug just last week and I fixed it 
> yesterday. If you want the fix but don't want to use a git version, just remove the 
> " + 1" in your code.
Heh. :) This was the bug that I was initially trying to fix some days 
ago. But as it was running LwIP 1.3.2 and the 1.4.0 changelog listed a 
lot of bugfixes I did an upgrade first. But as the bug still existed I 
begun looking at the code more closely.

Sincerely, Matias

--
mat!as