[941] in linux-net channel archive
Linux 1.3.x and Token Ring
daemon@ATHENA.MIT.EDU (Pascal ANDRE)
Sun Aug 20 20:23:41 1995
Date: Sun, 20 Aug 1995 17:19:49 -0400 (EDT)
From: Pascal ANDRE <andre@chimay.via.ecp.fr>
To: Linux Net <linux-net@vger.rutgers.edu>
So ... the facts :
since kernel version 1.3.6, Token Ring support is broken in Linux.
What has been tried :
As we had already worked on the token ring driver in the past, we had
a look to the source code. We succeeded in partially making it run (TCP
services we up most of the time on a single ring). The point is that we
were unable to make it run on multiring LAN. The internal code of the
driver is still OK : TR source routing works fine (internal routing
tables are well filled on packet reception, ...), packet construction
seems to be good. We had a look deeper in TCP stack code, and some weird
things appeared.
Here come the questions :
- what are the impact of the skb_buff handling modification on a net
driver (especially for non fixed hardware header size, which seems to
be the problem as the main difference between mono and multi ring LAN
is the fact that the routing information of the packet header is
bigger) ?
- is there some document available about these modifications ? (we
already looked into driver skeleton, but it was not enough)
- why is the token ring rebuild_header function different from
ethernet drivers one ? is this handled by the kernel in a special way or
is this a bug ?
And, as we no longer have access to machines located on the same
ring (not enough people in school during summer :) ), is there someone
out there using recent TR driver on a monoring LAN without problem ?
Waiting for your answers, ready to give informations, and hoping
someone will help ...
---
Pascal ANDRE, student at Ecole Centrale Paris.
andre@chimay.via.ecp.fr, andre@zen.via.ecp.fr
