Subject: pppd loopback/modem hangup problems
To: None <netbsd-users@NetBSD.ORG>
From: VaX#n8 <email@example.com>
Date: 01/08/1997 21:37:04
A while back I complained about pppd saying "serial line is looped back".
Ahem. Like that ever happens. Anyway, this happens, usually many times
in a row, and the only solution is to kill pppd (I was using the demand-dial
feature) and restart it. Presto, works fine. This is talking from
a NetBSD/x86 to a Linux/x86 box. If I disable magic number negotiation,
it connects and then says "Modem Hangup". This repeats several times,
as before. Sometimes I'd get the "looped back" error 10 or 20 times in
a row (I can grep my messages to generate some statistics if you like...
seems like it occurs 5-10% of the time or so).
If left on demand-dial, eventually one will work (usually much later).
Someone suggested that the magic number is timeofday and the pid XORed,
and suggested very precise clocks but my clock is anything but precise,
and the systems have disparate loads, so the pids are going to change at
rates such that multiple looped back errors are unlikely. And the modem
hangup stuff seems to discredit it completely.
I have one of the USR 33.6 Sportster modems that has been known to exhibit
strange "pause" behavior, typically when doing interactive stuff --
that is, these modems have been known to delay received data for up to a
minute or two when exposed to typical "bbsing" loads -- user sends a few
bytes, receives many, etc. I have been unable to determine if my modem in
particular is exhibiting this behavior due to USR's vagueness on the
load which exacerbates the condition.