DoS and Root Compromise Kernel Bug?

DoS and Root Compromise Kernel Bug?

[Jeremy Andrews]

Hello,

  Yesterday Rafal Wojtczuk posted to BugTraq regarding two kernel bugs:
   
http://www.securityfocus.com/cgi-bin/archive.pl?id=1&mid=221337&start=2001-10-15&end=2001-10-21

  I'm curious to understand more about these bugs.  I.E., are they real?  And,
are they fixed in 2.4.12 as claimed?  How about in the -ac series?

  The first kernel bug is regarding symbolic links. Rafal says it is partially
fixed in 2.4.10, and completely fixed in 2.4.12. This bug allows for a local
user to carry out a Denial of Service attack.

  The second bug allows for a root compromise via ptrace. The requirements are
that /usr/bin/newgrp be suid root (as in my RedHat 7.0 server), and that newgrp
not prompt for a password when run without arguments (again, as is the case with
my RedHat 7.0 server). Rafal says the attack only appears to work on Linux.

Thanks,
 -Jeremy
--
 Jeremy Andrews    <mailto:jeremy@kerneltrap.com>
 PGP Key ID: 8F8B617A  http://www.kerneltrap.com/

Re: DoS and Root Compromise Kernel Bug?

[Alan Cox]

>   Yesterday Rafal Wojtczuk posted to BugTraq regarding two kernel bugs:
>    
> http://www.securityfocus.com/cgi-bin/archive.pl?id=1&mid=221337&start=2001-10-15&end=2001-10-21
> 
>   I'm curious to understand more about these bugs.  I.E., are they real?  And,
> are they fixed in 2.4.12 as claimed?  How about in the -ac series?

2.4.12 / 2.4.12-ac have both fixed.
2.2.20 is still pending but will fix it

>   The second bug allows for a root compromise via ptrace. The requirements are
> that /usr/bin/newgrp be suid root (as in my RedHat 7.0 server), and that newgrp
> not prompt for a password when run without arguments (again, as is the case with
> my RedHat 7.0 server). Rafal says the attack only appears to work on Linux.

So far yes. The FreeBSD folks and others were also made privy to the problem
well before he published it to bugtraq and seem to be completely in the
clear.

Alan