[2596] in linux-net channel archive
Re: in.rlogind stuffing up
daemon@ATHENA.MIT.EDU (Alan Cox)
Fri Apr 19 06:36:43 1996
From: Alan Cox <alan@cymru.net>
To: jlewis@inorganic5.fdt.net (Jon Lewis)
Date: Fri, 19 Apr 1996 11:14:39 +0100 (BST)
Cc: dledford@dialnet.net, linux-net@vger.rutgers.edu
In-Reply-To: <Pine.LNX.3.91.960419051558.31362Y-100000@inorganic5.chem.ufl.edu> from "Jon Lewis" at Apr 19, 96 05:18:11 am
> > performing an rlogin -8 -E shell. Now, for the problem. This works fine
> > most of the time. But, if a modem disconnects unexpectedly, it can leave
> > the in.rlogind on shell in a waiting state. In this case, if I check a
> > netstat -a on shell, I will usually find a single tcp connection on the
> > rlogin port that has a large amount of information in the transmit and
> > recieve queue, but no process on the remote machine is listening. In
> > this case, further rlogins from the modem server fail to connect with a
> > "connection refused" error message. As soon as I kill the offending
>
> I have a similar sounding setup and have never seen this happen. This is
> just a guess...but have you tried setting hup_notify with
> setserial/setrocket on your dialup tty's? Back when I set things up, it
> looked like a good idea, so I did it.
Its a libc bug. On the connection refused libcs priv port stuff should try
the next port put doesnt. I believe the ELF libc's have this fixed.
Alan
