Re: [2.1.99] TCPv4 bad checksum

Caleb Epstein (epstein_caleb_unix@jpmorgan.com)
Fri, 1 May 1998 13:28:55 -0400


On Fri, May 01, 1998 at 10:16:10AM -0600, Kevin Fenzi wrote:

> Davem maintains (I am pretty sure correctly) that these are caused by
> VJ header compression bugs in various other ppp implementations.
>
> Is there anything we can do to confirm this?
> Is it significant that they are always of the form:
>
> XX/XX/XX+20
>
> ie, are they all failing in the same way so the kernel could detect
> this and work around it? Is the checksum always off by 20?
>
> Anyhow, these messages cause no harm that I can tell but it would be
> nice to resolve them before 2.2...If nothing else, we should remove
> them if they no longer tell us meaningfull info...

I have seen this as well, usually when I do a Ctrl-C on a
terminal session over a PPP link to my office. This is both with and
without asyncmap set to escape control characters. It may not be a
bug in "other ppp implementations". See the following from the
Changelog of PPP 2.3.4:

* Fixed some bugs in the linux driver and deflate compressor which
were causing compression problems, including corrupting long
incompressible packets sometimes.

Perhaps this is what we are all seeing?

-- 
[ cae @ jpmorgan.com | Caleb Epstein | JP Morgan & Co., Incorporated ]

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu