We have recently found a serious security bug in MariaDB and MySQL.So, here, we'd like to let you know about what the issue and its impactis. At the end you can find a patch, in case you need to patch an olderunsuported MySQL version.
All MariaDB and MySQL versions up to 5.1.61, 5.2.11, 5.3.5, 5.5.22 arevulnerable.
MariaDB versions from 5.1.62, 5.2.12, 5.3.6, 5.5.23 are not.
MySQL versions from 5.1.63, 5.5.24, 5.6.6 are not.
This issue got assigned an id CVE-2012-2122.
Here's the issue. When a user connects to MariaDB/MySQL, a token (SHAover a password and a random scramble string) is calculated and comparedwith the expected value. Because of incorrect casting, it might'vehappened that the token and the expected value were considered equal,even if the memcmp() returned a non-zero value. In this caseMySQL/MariaDB would think that the password is correct, even while it isnot. Because the protocol uses random strings, the probability ofhitting this bug is about 1/256.
Which means, if one knows a user name to connect (and "root" almostalways exists), she can connect using *any* password by repeatingconnection attempts. ~300 attempts takes only a fraction of second, sobasically account password protection is as good as nonexistent. Any client will do, there's no need for a special libmysqlclient library.
But practically it's better than it looks - many MySQL/MariaDB buildsare not affected by this bug.
Whether a particular build of MySQL or MariaDB is vulnerable, depends onhow and where it was built. A prerequisite is a memcmp() that can returnan arbitrary integer (outside of -128..127 range). To my knowledge gccbuiltin memcmp is safe, BSD libc memcmp is safe. Linux glibcsse-optimized memcmp is not safe, but gcc usually uses the inlined builtin version.
As far as I know, official vendor MySQL and MariaDB binaries are notvulnerable.
Regards,
Sergei Golubchik
MariaDB Security Coordinator