Re: SecurityReason: PHP 5.2.6 SAPI php_getuid() overload

Maksymilian, good day.

Sat, Dec 06, 2008 at 12:40:48PM -0700, cxib@securityreason.com wrote:
> [ SecurityReason.com : PHP 5.2.6 SAPI php_getuid() overload ]
[...]
> Using PHP 5.2.6, as a Apache module can bypass many security points.

Am I right that this vulnerability exists only in the Apache 1.x flavour
of the PHP module?  The code in question that sets SG(server_context)
too late and initializes BG variable after the .htaccess processing
exists only in sapi/apache/mod_php5.c.  For Apache 2.x module the
handler is 'php_handler', it lives in apache2{filter,handler}/sapi_apache2.c
and BG/SG(server_context) are initialized before .htaccess processing.

And to clarify a bit the overall picture: am I right that the purpose of
your sleep.php manipulations is to make Apache to invoke another "fresh"
child that will process error_log contents with errorneous value of
uid/gid = 0?  It seems to me that the effect of the found vulnerability
can be shortly characterized as "the first request for the given Apache
child will have uid/gid = 0 as the values returned from 'php_getuid()'
in the code that handles .htaccess contents (to be precise, in the code
inside the function send_php() before the call to
apache_php_module_main(), the point where BG is really initialized by
PHP_RINIT_FUNCTION(basic))".  Am I missing something?

Thank you!
-- 
Eygene