Re: 2.2.0 SECURITY (fwd)

[andrea () e-mind com (Andrea Arcangeli)]

On Wed, 27 Jan 1999, Andrea Arcangeli wrote:

> And here it is my race-fix against linux-2.2.0 (that I written today and
> last night). I am not able to cause process to go in D state or Oopses or

The fix I posted some hour ago is still perfectly working and safe but as
pointed out by Stephen C. Tweedie I don't need a new finegrined spinlock
(mm_lock) to be atomic between tsk->mm = &init_mm and mmput() because we
just have the big kernel lock held in the interesting places. Note: my new
mm_lock is still fine but it's _not_ strictly needed, is just a bit of
overhead (zero in an UP compilation). The other part of the patch
(array.c) instead is still _strictly_ needed to not Oops and/or crash.

Andrea Arcangeli