Re: MTU problem with 2.0.26?

Christian Balzer (
Thu, 28 Nov 1996 23:36:02 +0100 (MET)

In article <> you wrote:
>I upgraded 3 WWW cache servers from 2.0.24 to 2.0.26 yesterday. In the
>process I changed their ethernet cards from 3c590s to 3c905s running at
>10Mbps (I'm using the current 3c59x driver from
>We have a frame relay network which doesn't like an MTU larger than 1006
>bytes, since upgrading the servers people have been having problems
>connecting to them, getting as far as 'host contacted, waiting for reply',
>then nothing. A packet trace shows the TCP session being opened then
>hanging. I've come across this in the past and over-size packets are
>the culprit.
>I ifconfig'd the mtu on the ethernet interfaces to 1006 and the problem
>disappeared, however the rest of our servers (2.0.24, 2.0.25 and Solaris)
>all use an ethernet MTU of 1500 without any problems.
>Has anything changed in the network code (fragmentation?) for 2.0.26 which
>would cause this to happen, or would the 3c905 be at fault?
I think it's 2.0.26, the machine I run Squid (http proxy-cach) did
stop working after the upgrade from 2.0.25 for SLIP users dialing
into this LAN via another (1.3!) machine with pretty much the same
symptoms (data is in the sendqueue but never makes it out), not so
for local machines or folks coming in via an ISDN router (which uses
the ethernet MTU).

A look at the .26 patch also reveals a lot of changes to fragment.c,
not that I claim I understand what's going on there, but it makes
this kernel look even more like the culprit. I guess I'll downgrade
again, as I don't see anybody working on this, maybe the Cc: can
change this fact. ;-)

Mata ne,


  // <CB> aka Christian Balzer, Tannenstr. 23c, D-64342 Seeheim, Germany
\X/ | Voice: +49 6257 83036, Fax/Data: +49 6257 83037
SWB  - The Software Brewery - | Team H, Germany HQ | Anime no Otaku