Re: TCPv4 bad checksums

Damon Buckwalter (damon@meta-x.net)
Tue, 22 Dec 1998 15:26:52 -0800


Stephen Clark wrote:
>
> I get periodic burst of TCPv4 checksums. This usually happens when
> retrieving mail. I am connected on a 10baset internal network. The
> internal network has an ISDN connection to the internet thru an Ascend
> Pipe 130.

I would suspect the Ascend Pipeline. I'm connected through PPP to an
Ascend MAX 4048, and used to experience TCP checksum errors. Someone
(see below) suggested turning off VanJacobsen compression on my PPP
link, and that seemed to do the trick. I haven't seen any checksum
errors since. Just add the 'novj' option to your PPP config files, and
redial.

Shaw Carruthers wrote:
>
> It is believed that there is a problem with the Ascends that Demon use
> which leads to checksum errors. The Ascends have a fixed buffer of about
> 16k per connected user. If you advertise a large receive window, the
> buffer can overrun and packets get dropped which screws the VJ
> compression. I use a window of 7k which reduces the problem.
>
> Note: the single Ascend buffer has to handle multiple TCP streams to you.
>
> There is also another problem as I get checksum errors with novj, but far
> less often.

I'm running kernel 2.1.131.

--
damon@meta-x.net -- PGP and GPG public keys at http://meta-x.net/keys/

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/