Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 09 May 2012 14:21:34 -0600
From: Kurt Seifried <>
Subject: Re: PHP-CGI query string parameter vulnerability (CVE-2012-1823
 / CVE-2012-2311, CERT VU#520827)

Hash: SHA1

On 05/09/2012 12:59 PM, wrote:
>> The incomplete fix part seems to have got bit messy, at least
>> with respect to CVE assignment.
> The total number of CVE IDs should be 4. Comments below:
>> 1) Incorrect detection of = in query string, that made it
>> possible to bypass the fix using %3D.  This was addressed by:
>> -       if(*decoded_query_string == '-' &&
>> strchr(decoded_query_string, '=') == NULL) { +
>> if(*decoded_query_string == '-' && strchr(query_string, '=') ==
>> NULL) {
>> which is noted as Mitigation option 3 in De Eindbazen's blog. 
>> Following the timeline / updates there, this should be what
>> triggered CVE-2012-2311 assignment.
> CVE-2012-2311 is for only this specific %3D fix listed as "1)"
> above.
>> 2) The fix from 1) did not address the problem for use cases
>> where "unsafe" wrapper script, similar to the one pointed out in
>> De Eindbazen's blog, is used.  It seems that was first mentioned
>> in Christopher Kunz's ( blog mentioning that the
>> PHP re-fix is still incomplete, though it's questionable if this
>> is to be considered a PHP flaw.  Upstream warned about this
>> insecure wrapper script problem:
>> and even added a fix / work around for it to PHP:
> This needs a separate CVE ID (different from both CVE-2012-1823
> and CVE-2012-2311). MITRE is considering the "insecure wrapper
> script" to be a distributable product because the code has been
> available for some time on a public web site, and the (admittedly
> tiny) script codebase has apparently sometimes been copied and
> adapted for use at multiple web-hosting providers.
> The script codebase is, of course, open source.
> In many cases, MITRE would not bother to assign a CVE ID for a 
> vulnerability in a "product" of this type (i.e., a product that is 
> arguably not even packaged for distribution and does not even have
> a product name). However, in this case, the product and its 
> vulnerability have become commonly recognized and discussed because
> of the connection to the other PHP-CGI issues. Therefore, a CVE
> name is useful.
> This can be temporarily called CVE-2012-NEW-1.

Please use CVE-2012-2335 for these wrapper script issues

>> 3) The fix from 1) only made PHP skip one php_getopt() call out
>> of two that are reachable in the CGI mode (the third php_getopt()
>> call is in the if (!cgi && !fastcgi) block).  As the consequence,
>> PHP was still parsing following arguments:
>> - -h / -? - this seems harmless, as makes PHP output usage info,
>> which triggers Internal Server Error in httpd - -T - this was
>> mentioned as DoS vector:
>>  The impact of this is rather limited as clients needs to consume
>> all generated output too keep this running.  May offer some
>> advantage of simple many requests DoS e.g. Keep-Alive is disabled
>> and there's per-IP connection limit.
>> This is upstream commit that was used in 5.4.2 / 5.3.12:
>>  and this is correction from 5.4.3 / 5.3.13:
>>  (both links are for PHP-5.3 branch commits).
> This one also needs a separate CVE ID (different from both 
> CVE-2012-1823 and CVE-2012-2311). Compared to CVE-2012-2311, it 
> apparently has the same "affected versions" relative to official 
> upstream release numbers. However, the affected versions are
> different in PHP packages from multiple Linux distributions. In
> this situation, it seems best to have two separate CVE names
> (CVE-2012-2311 and a new one) for the two different issues with
> php_getopt - in other words:
> CVE-2012-2311:   the vulnerability that exists when the php_getopt 
> for cases 'c' 'n' 'd' 'b' and 's' is not skipped
> CVE-2012-NEW-2:  the vulnerability that exists when the php_getopt 
> for cases 'T' and 'h' is not skipped

Please use CVE-2012-2336 for the cases 'T' and 'h'

> If this is agreeable, we would like Red Hat to make the specific
> CVE assignments for the "CVE-2012-NEW-1" and "CVE-2012-NEW-2"
> labels referenced above. If, for any reason, Red Hat doesn't want
> to make these two CVE assignments, please let us know.

- -- 
Kurt Seifried Red Hat Security Response Team (SRT)
PGP: 0x5E267993 A90B F995 7350 148F 66BF 7554 160D 4553 5E26 7993

Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Mozilla -


Powered by blists - more mailing lists

Your e-mail address:

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

Powered by Openwall GNU/*/Linux - Powered by OpenVZ