RE: [SQU] More on WCCP and truncated GRE packets. --HELP

From: Lincoln Dale <ltd@dont-contact.us>
Date: Wed, 13 Dec 2000 14:24:39 -0800

> >>Is there anyone familiar enough with the format of the GRE packet out
> >>there that can look at my previous posting of a few of these truncated
> >>data packets and figure out what is being left out?
> >
> >truncated packets may be due to a bug. it may be due to the fact that the
> >user is sending a large packet which would exceed the MTU when a GRE header
> >is prepended. in this case, the router will fragment the packet.
> >
> >it may be that ip_wccp cannot handle this [perfectly valid] case. squid
> >enforcing a MSS to the client such that no fragmentation could occur may
> >help.
>
>I think this Cisco doc might shed some light on GRE tunneling and the
>truncated 24 bytes...
>
>http://www.cisco.com/warp/public/105/56.html

actually, that highlights part of the issue.

in the case of a 'large packet' where adding a GRE header + new IP header
causes the frame size to exceed the MTU, the router will have to fragment
the packet.
that then means that the GRE-tunnel-termination point (the ip_wccp module
in linux) would need to be capable of handling reassembly of tcp fragments
when the fragments are received inside a GRE tunnel.

i am not familiar with the ip_wccp code to know if this is the case or
not. given the failure condition, i'm guessing that it isn't.

if this is the case, the workaround stated above (squid enforce the
client-side MSS to something such that fragmentation cannot occur) will
cause the problem to go away.

cheers,

lincoln.

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
Received on Wed Dec 13 2000 - 15:29:17 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:56:57 MST