Masquerading, dying, deubg and wrong ip#'s

Paul Wouters (paul@xtdnet.nl)
Sun, 16 Jun 1996 21:48:31 +0200 (MET DST)


I have been using a masquerading host to to hide our Evil[tm] windows
network from the world. I've been using fairly recent kernels and am now
using 2.0.0. From time to time, and especially when having a lot of
traffic through it (mostly netnews through netscape or ws_ftp's in pasv
mode) the masquerading host stops. I can still logon to it, and I can see
'network cable problem?' and ETHX: time our errors.
So, I tried defining the DEBUG statements for the masquerading stuff.
Apart from filling syslog, the host kept working (Well, kind of slower
with all the logging ;).
On the same network segment (at the 'safe' side of the masq.host) is an ipx
network that doesn't seem to think there is a cable problem.

So, I am thinking that for some reason, during busy network sessions, the
masquerading host can't keep up (It's only a 386dx, 8 megs ram and three
ethernet cards ( 100vg, and two etherworks pro's). But shouldn't it somehow
request the sender to slow down when it gets congested ?

Also, since 2.0.0 i sometimes get errors like :
MASQ: failed TCP/UDP checksum from aaa.bbb.ccc.ddd

Should I junk our 386? Can anyone tell me they're successfully running a
masquerading host for a small office on a low-end machine ?

Any clues, hints or tips are appreciated,

Paul