[38855] in bugtraq
Re: Linux kernel ELF core dump privilege elevation (kernel module workaround)
daemon@ATHENA.MIT.EDU (chris)
Fri May 13 16:47:45 2005
Date: Thu, 12 May 2005 17:31:58 -0500
From: chris <fool@dfw.net>
To: Andrew Griffiths <andrewg@felinemenace.org>
Cc: full-disclosure@lists.netsys.com, bugtraq@securityfocus.com,
vulnwatch@vulnwatch.org
Message-ID: <20050512173158.A25882@dfw.nationwide.net>
Reply-To: fool@dfw.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20050512002902.GA21778@felinemenace.org>; from andrewg@felinemenace.org on Thu, May 12, 2005 at 12:29:02AM +0000
On Thu, May 12, 2005 at 12:29:02AM +0000, Andrew Griffiths wrote:
> It loops over all processes and sets the soft limit and hard limit for
> processes to 0. The limits.conf measure isn't entirely enough if people
> have screen sessions, or you have various daemons running etc.
We've used "echo / > /proc/sys/kernel/core_pattern" to disable coredumps
by all processes (using kernel 2.4.29). This seems to affect all
running processes without doing anything drastic or dangerous (except
disabling coredumps =)). To disable all for all but root processes you
can use something like " /core " instead of " / " in the above example,
but then you may still be vulnerable as Andrew points out to pre-existing
screen or Xwin sessions running as root--not sure about that but better
safe than sorry.
I haven't read the code to see if this is an unintentional effect, but
it sure seems to work under 2.4.29 at least. I got the idea from this
page:
http://www.aplawrence.com/Forum/TonyLawrence9.html
