[1465] in linux-net channel archive
Re: UDP: bad checksum.
daemon@ATHENA.MIT.EDU (Corey Minyard)
Fri Dec 1 19:21:05 1995
Date: Thu, 30 Nov 95 21:22 CST
From: Corey Minyard <minyard@wf-rch.cirr.com>
To: linux-kernel@vger.rutgers.edu, linux-net@vger.rutgers.edu
In-reply-to: <01HY7N4XPKHI003XXU@UWSTOUT.EDU> (STEFFENR@UWSTOUT.EDU)
Reply-to: minyard@metronet.com
> > With kernel v1.3.45 I'm getting lots of UDP: bad checksum messages. The
> > machine has a WD80x3 card, and seems to work fine as an NFS client and
> > server.
>
> Your not alone, I've been seeing lots of UDP checksum errors myself.
> I have an SMC Ultra ethernet card, which, I believe is a similar driver
> to the WD80x3.
>
> >
> > The full message is:
> > UDP: bad checksum. From 807F7E0B:520 to 807F7EFF:520 ulen 512
> > although sometimes the ulen is 242.
>
> The messages I see are a little different:
>
> UDP: bad checksum. From 00000000:68 to FFFFFFFF:67 ulen 308
>
> It always the same message. Always. I traced the packets to one specific
> machine, but I'm one a large network and have no way of tracking down
> what type of machine it is (or even where it is.
>
> It's gotta be a Linux net problem specific to later 1.3.x kernels. The
> messages go away in 1.2.13, and I've noticed them 1.3.x since 1.3.35 or
> so.
>
I've seen a similar problem on non-Linux boxes (Motorola and HP) which
manifested itself with NFS file corruption. The problem turned out to
be a bad router card. We were unable to turn checksumming on for UDP
on either machine, so se just got bad packets and didn't know it
(until one of our compiles was completely hosed).
So, the problem is not necessarily Linux, it could be some other
network entity causing this. It may be that the corruption is always
happening and someone installed a new machine on the network that has
the checksumming turned on. Network problems are sooooo much fun to
debug :-)
--
Corey Minyard Internet: minyard@metronet.com
Work: minyard@bnr.ca UUCP: minyard@wf-rch.cirr.com
