[953] in linux-net channel archive
Re: more: telnet not working
daemon@ATHENA.MIT.EDU (Alex Bligh)
Tue Aug 22 01:32:37 1995
From: Alex Bligh <alex@cconcepts.co.uk>
To: pmain@interoz.com (Patrick Main)
Date: Mon, 21 Aug 1995 11:05:35 +0100 (BST)
Cc: linux-net@vger.rutgers.edu
In-Reply-To: <199508190946.FAA13199@vger.rutgers.edu> from "Patrick Main" at Aug 19, 95 04:53:11 am
>
> OK guys i thought i would start fresh. Recently i had the following
> happen on our colleges's web server running Linux 1.2.8 it only has
> 8 megs but that's all the college would cough up.
>
> I have been telnetting in to do admin work and features <add/change>.
> suddenly last night i got the dreaded connection closed by foreign host.
> Got this on everything i tried and i was telnetting in from a Linux
> machine through a PPP connection. Test telnets to other machines were
> fine. After investigating by telnetting to different ports, it became
> obvious that services that were running in standalone were not effected.
> ie: i could telnet and use smtp and also http and gopher too. these all
> run by themselves and not through inetd. anything that depends on inetd
> would result in a connection closed by foreign host message. bottom line
> is that telnet and ftp were broen but it seems that inetd was not
> accepting any connections properly. i would get the connection established
> and then the connection woud be slammed closed! Any comments. If inetd
> had died would i not even get a connection at all? weird!
>
> Now for some more weirdness. Before i coud get physically to the machine
> the problem fixed itself! I thought at first that someone had rebooted
> but this is not the case. No change has been made in the beginning up-times.
As someone (aleph1) already answered to a previous thread, this is quite
likely to be tcpwrapper (rather than inetd per se) dropping connections
as it can't reverse lookup the host connecting, which could well be
a dns problem (like your college name server went screwy for a bit). Make
sure tcpwrapper security logs are working properly, and check for failures.
I've had inetd die randomly on me, but not for any undiscernable reason
since 1.1.38 (might be previous to that but I didn't do all the increments).
Other reasons for inetd dying are running out of processes (can't fork
anything - this happened to me lots until I fixed my syslogd), and running
out of disk space (think this was also somthing to do with logging
connections).
The best thing is to get it reproducible if possible and strace -fp inetd.
Alex
----------------------------+-------------+-----------------------------
Alex Bligh : ,-----. :
Computer Concepts Ltd. : : : alex@cconcepts.co.uk
Gaddesden Place : : ,-----. :
Hemel Hempstead : `-+---` ` : Tel. +44 1442-351000
Herts. UK HP2 6EX : | , : Fax. +44 1442-351010
: `-----` :
----------------------------+-------------+-----------------------------
