I guess you didn't read the entire post.
The protection you mention was trivially bypassable for half a year after its existence. It
breaks legitimate applications and won't be in use everywhere. It also covers only one type
of a more general class of bugs. For these others, it can do nothing. Examples of this
include the recent vmsplice exploit on amd64 (see the comment by the PaX team at
http://lwn.net/Articles/271688/) or dereferencing of poisoned pointers (see
BTW the vulnerability in the "protection" was known by me since its inception. As proof, find
the date of the mention of 3812e371986ad24ace67bab90fd07ca4 in
3812e371986ad24ace67bab90fd07ca4 is a hash of the following text (referring to the protection
developed by Red Hat):
"it's too bad that it's trivially bypassed via expand_stack :) this will
be funny in a couple months"