On 08/02/2012 01:11 AM, Christoph Lameter wrote:
> 
> V7->V8:
> - Do not use kfree for kmem_cache in slub.
> - Add more patches up to a common
>   scheme for object alignment.
> 
> V6->V7:
> - Omit pieces that were merged for 3.6
> - Fix issues pointed out by Glauber.
> - Include the patches up to the point at which
>   the slab name handling is unified
> 

After applying v8, and proceeding with cache deletion + later insertion
as I've previously laid down, I can still see the bug I mentioned here.

Again, I am doing nothing more than:
1) Creating a cache
2) Deleting that cache
3) Creating that cache again.

I am doing this in a synthetic function "mybug" called from memcg
creation for convenience only (so don't get distracted by this in the
backtrack). The machine boots okay, and seems to work for everything
other than those late destructions. So maybe this is a problem that
happens only after SLAB_FULL?

I am attaching the backtrace I've got with SLUB_DEBUG_ON. My first guess
based on it would be a double free somewhere.