Re: [PATCH 2.6.12-rc4] cpuset exit NULL dereference fix

[Paul Jackson <pj@sgi.com>]

Robin wrote:
> Whenever the refcount reaches 0, we
> automatically remove the cpuset. 

If by this you mean replacing the usermodehelper callout to
/sbin/cpuset_release_agent, with a direct removal of the cpuset
in the kernel, then:
 * this changes a kernel API - not to be done lightly, and
 * adding an in-kernel way to nuke a cpuset, in addition to
   having the rmdir(2) system call do it, seems like it would
   present even thornier locking issues.

The usermodehelper callout is working just fine for us.

As to the rest of your post, proposing a per-cpuset spinlock,
see my previous reply to Simon.  

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@engr.sgi.com> 1.650.933.1373, 1.925.600.0401