*BSD News Article 14916


Return to BSD News archive

Newsgroups: comp.os.386bsd.development
Path: sserve!newshost.anu.edu.au!munnari.oz.au!spool.mu.edu!cass.ma02.bull.com!think.com!paperboy.osf.org!enterpoop.mit.edu!gatech!howland.reston.ans.net!newsserver.jvnc.net!gmd.de!olymp!saph2.physik.uni-bonn.de!juengst
From: juengst@saph2.physik.uni-bonn.de (Henry G. Juengst)
Subject: Re: SIGKILL and kill
Message-ID: <1993Apr21.183200.8058@olymp.informatik.uni-bonn.de>
Keywords: SIGNALS SECURITY
Lines: 50
Sender: usenet@olymp.informatik.uni-bonn.de
Reply-To: juengst@saph2.physik.uni-bonn.de
Organization: Universit"at Bonn, Informatikinstitut, R"omerstr 154, W-5300 Bonn 1
References:  <1qo3mq$d4b@news.cs.tu-berlin.de>
Date: Wed, 21 Apr 1993 18:32:00 GMT


In article <1qo3mq$d4b@news.cs.tu-berlin.de>, klier@cs.tu-berlin.de (Jan Klier) writes:
> Hi 386bsd community,
> 
> I don't know if this is the right place to post such things, so if no please
> appologize.
> 
> From my experience as administrator and user of Unix systems I have a
> suggestion for a little change which could increase the system security
> for the user themselves.
> I experienced that many users often use the command 'kill -9 pid' to kill
> a process instead of the simple version 'kill pid'. But if the SIGKILL signal
> is used to terminate a process the user will get rid of, the process itself
> has no chance to catch that and cleanup it's own data structure, flushing
> some buffers, updating databases etc.
> Most often the TERM signal will do the same job, users just don't know about
> the difference between KILL and TERM and risk (unconciously) loosing data.
> 
> My idea is now (and I post it here because it could be tested experimentally
> in 386bsd) to modify the kill-programm in order to restrict the SIGKILL signal
> to the superuser.
> This will force users to use the safe TERM-signal when the terminate processes
> and still leaves the door open for really hung situation where a SIGKILL is
> necessary.
> 
> Any comments?

Restrictions are not what users want. SIGTERM could be caught by an ill
(bug) signal handler so a program could hang up. Users aren't happy if
they need a guru for everything. May be one should add a very big hint
into the manual pages (kill.1), because in principle I agree with you.

> 								jan
> 
> 
> -- 
> *******************************************************************************
> Jan Klier                                                    Berlin, Germany
> e-mail: klier.cs.tu-berlin.de 		  cis   : 100022,1700
>       | jklier.ipk.fhg.de        		| 100022.1700@compuserve.com
> 

Henry

--
juengst@boss1.physik.uni-bonn.de [131.220.221.30]  {u}
juengst@saph2.physik.uni-bonn.de [131.220.221.12]  {d}
juengst@pib1.physik.uni-bonn.de  [131.220.221.2]   {s}

I'm only speaking for myself, but not for anybody else !