lwip-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

SV: [lwip-users] loopif bug ?


From: Andreas Stenius
Subject: SV: [lwip-users] loopif bug ?
Date: Fri, 28 Feb 2003 09:14:15 +0100

> Kieran Mansley
> Skickat: den 27 februari 2003 18:26
> Till: Mailing list for lwIP users
> Amne: Re: [lwip-users] loopif bug ?
>
>
> On Thu, 27 Feb 2003, Andreas Stenius wrote:
>
> > Hi all,
> >
> > I've run across a misshap with the loopif ( TCP only, not UDP ).
> > When connecting to 'localhost' using the loopif interface, the
> tcp packet is
> > processed on the
> > receiver side before the send process is completed, resulting
> in a reference
> > to NULL in tcp_process:
>
> Yep, looks like a bug to me.  Quite an interesting one in fact.
> Demonstrates nicely that no one else is using the loopif interface.  In
> theory I guess it could affect a non-loopback interface if the network
> were fast enough.
>
> I'd file a bug report on savannah if I were you, but it should be fairly
> straightforward to fix.  Moving the portion of code in tcp_output that
> updates pcb->unacked to before the call to tcp_output_segment should do
> the trick.  Although interestingly if TCP_TCPLEN is never greater than 0,
> pcb->unacked will remain NULL, and we'll still get a seg fault.  Is it
> possible for this to happen? (admittedly it would be a pretty dull
> connection, but if it's possible we should be able to cope with it).
>
> Kieran
>

I've filed it as bug at savannah a few weeks ago ( sent a mail here too, but
somehow I obviously failed in signing up   for the mailing list then ).

I looked at the possibility to simply move the call to tcp_output_segment as
you suggested, but it would involve some more if this then that code, since
it wasn't always put on the unacked list. So for now I simply don't connect
with loopif and TCP.

//Andreas






reply via email to

[Prev in Thread] Current Thread [Next in Thread]