+++ This bug was initially created as a clone of Bug #245732 +++

The OpenSSL team made the following commit publically to the 0.9.8 head:
http://cvs.openssl.org/chngview?cn=16275

This is a possible weakness in OpenSSL which could allow a local user in certain
circumstances to divulge information about private keys being used.   For
example if a server has a SSL web server on it, a local unprivileged user may be
able to get hold of the key.  

I'm not sure in practice how possible this would be; it would rely on a system
that isn't doing much else that could interfere with some spy process designed
to figure out this information (and probably won't be possible if you have a
server handling a lot of traffic, has more than one key, lots of local
processes, and so on).

It's similar to previous issues and is rated severity=moderate

I'm not sure if the OpenSSL team plan to upgrade older versions of OpenSSL or
when this will be announced publicly.  Please treat as embargoed for now.

-- Additional comment from mitr on 2007-07-16 11:55 EST --
Created an attachment (id=159339)
Proposed patch for 0.9.7a


-- Additional comment from mitr on 2007-07-16 17:53 EST --
Created an attachment (id=159381)
Proposed patch for 0.9.6b

-- Additional comment from mjc on 2007-08-02 06:08 EST --
removing embargo, CERT published
http://www.kb.cert.org/vuls/id/724968
see also 
http://openssl.org/news/patch-CVE-2007-3108.txt

-- Additional comment from mjc on 2007-08-02 06:11 EST --
Because this change has not been tested in a full upstream OpenSSL release there
is some risk that it may introduce unexpected side-effects.  Given this issue is
not serious, our current plan is as follows:

- To include the backported fix in an OpenSSL update ready for RHEL 4.6.   This
will get testing via beta and give time for more extensive internal testing
- To release an async update for OpenSSL for other RHEL platforms at the same
time as RHEL4.6 is released