kernel: Socket destroy delayed (r=0 w=208)

[Ryan Lackey <ryan@venona.com>]

I keep getting these messages on a debian potato system with
a 2.1.129 kernel.  I've been getting them for months, using
various versions of the development kernel with slink -- I have
IPv6 compiled into the kernel, an SMC Ultra II/EPIC network card, etc.

Is there any known cause for these?  I've looked at one of the linux
networking groups and there were old reports of this problem in the
archives -- they appeared to say it was due to using old programs with
new kernels, in some cases.

The problem is that 1) there are thousands of these messages filling my
logs 2) it slows the machine down noticably -- up to 1 second ping times
sometimes on local unloaded fast ethernet until I kill and restart the
network card.

If anyone has any advice on this problem, I'd be very interested in
hearing it.  I'm about to start debugging by turning off lots of services
and compiling a minimalist kernel...  

Thanks,
Ryan the Anguillan (Caribbean...) Debian User
(please cc: any responses to the list to me...)