Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 23 Apr 2014 14:36:48 -0700
From: (Eric W. Biederman)
To: Andy Lutomirski <>
Subject: Re: CVE-2014-0181: Linux network reconfiguration due to incorrect netlink checks

Andy Lutomirski <> writes:

> [I think something went wrong with the quoting in here.]
> On Wed, Apr 23, 2014 at 9:27 AM,  <> wrote:
>> Hash: SHA1
>>> It is possible to reconfigure the network on Linux by calling write(2)
>>> on an appropriately connected netlink socket. By passing such a
>>> socket as stdout or stderr to a setuid program, anyone can reconfigure
>>> the network.
>>> Andy Lutomirski when looking at the networking stack noticed that it is
>>> possible to trick privileged processes into calling write on a netlink
>>> socket and send netlink messages they did not intend.
>>> In particular from time to time there are suid applications that will
>>> write to stdout or stderr without checking exactly what kind of file
>>> descriptors those are and can be tricked into acting as a limited form
>>> of suid cat. In other conversations the magic string CVE-2014-0818 has
>>> been used to talk about this issue.
>> First, CVE-2014-0818 is not the correct CVE ID. CVE-2014-0818 is
>> associated only with a vulnerability in AutoCAD. A CVE ID of
>> CVE-2014-0181 was in the Subject line.

Apparently I was disgraphic when I copied that line.  My apologies.

>> Also, there are two messages that discuss apparently distinct types of
>> security issues, suggesting that two or more CVE IDs may be needed:
>>   "The caller needs capabilities on the namespace being queried, not
>>   on their own namespace. This is a security bug, although it likely
>>   has only a minor impact." (The patch is in the packet_diag_dump
>>   function in net/packet/diag.c, but the issue originally was in the
>>   sock_diag_put_filterinfo function in net/core/sock_diag.c.)
> This may need a new CVE.  I'm not really clear on what the impact of
> this is, if any.  It's an information disclosure issue, but I'm not
> entirely sure that valuable information is being disclosed.

When rebasing ontop of net this patch goes away.  I only addressed this
issue at all because that was one of the capable calls that needed to be
fixed for the second issue, and the permission check really did need to
be moved to be correct from the dump context.  As the skb available in
that function was the skb to write data into, not the skb of the netlink
request we were authenticating.

>>   "verify that the opener of the socket had the desired permissions as
>>   well"
> This is the proposed method of fixing CVE-2014-0181

Yes.  The rest of the patches are just cleanups and infrastructure to
allow that change to happen.


Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.