[1347] in linux-net channel archive
Problem shows up with netdate
daemon@ATHENA.MIT.EDU (Sean Christopher Farley)
Thu Nov 9 20:16:23 1995
Date: Thu, 9 Nov 1995 03:36:01 -0500 (EST)
From: Sean Christopher Farley <farleysc@cs.purdue.edu>
To: linux-net@vger.rutgers.edu
Here is a little problem I ran into and the temporary solution. It started
with netdate killing my system which was running:
- libc v5.2.11 (also tried 5.0.9 & 5.2.10)
- Linux v1.3.37
- Debian netdate 1.21.x
I am now using a patched version of netdate as shown below.
No other network software problems (ftp, telnet, http, netrek :)
Any ideas?
Sean
------
Sean C. Farley
scf@cs.purdue.edu
PGP 2.6.2 key using finger or WWW
WWW Homepage is: http://www.cs.purdue.edu/people/farleysc
---------- Forwarded message ----------
Date: Mon, 6 Nov 1995 19:16:12 +0100 (MET)
From: Peter Tobias <tobias@server.et-inf.fho-emden.de>
To: Sean Christopher Farley <farleysc@lore.cs.purdue.edu>
Subject: Re: Bogons are killing my system!
Sean Christopher Farley wrote:
> > Could you please try the netdate from sc2.et-inf.fho-emden.de
> > (/pub/debian/netdate.test.tar.gz)? You have to recompile it with your
> > ELF gcc. Please drop me a note if it works for you and I'll include
> > the patch in the next version of Debian netstd package.
[...]
> It works!!! What was wrong? I was using the same copy for about one
> month without it showing up.
>
> I saw that you commented out sethostent and setprotoent. How did these
> break it? And why did they wait a month to die?
I'm not sure why this happens (a library or kernel bug?). IMHO it should
work. Using sethostent with a value of 1 forces hostname lookups to
use a TCP socket which remains open for further queries (stayopen
feature). Commenting out this entry forces netdate to use UDP for the
nameserver queries. It's probably only a workaround and does not fix
the real problem but on the other hand I see no reason why netdate
shouldn't use UDP for nameserver queries.
> And the big question is why did it cause the kernel to lock up instead of
> just registering an error and go on.
A good question, the kernel shouldn't lock up. It's probably something
for Alan Cox to fix :-).
Peter
--
Peter Tobias EMail:
Fachhochschule Ostfriesland tobias@et-inf.fho-emden.de
Fachbereich Elektrotechnik und Informatik tobias@perseus.fho-emden.de
Constantiaplatz 4, 26723 Emden, Germany
