On Fri Dec 10, 2010 at 17:52:37, Wolf wrote: > Well, I'm a first time writer to Bugtraq, but this is interesting. I > commented out the call to clone(), and after it simply called > trigger(fildes), and apparently, it works. Only tested on a stock > install of Ubuntu 10.10, but I thought the bug was in clone()? No, the bug is not checking address overwrite limit in the do_exit() path, which migh offer the chance to overwrite an arbitrary memory location. The clone call in the supplied poc just made sure do_exit() actually accesses the memory clearing the child tid (using the CLONE_CHILD_CLEARTID). So if your running process why so ever also had CLONE_CHILD_CLEARTID set it would trigger the problem as well.
