On Fri, Nov 09, 2001 at 04:28:48PM +0100, Frank van der Linden wrote:

 > * It does not seem to be a TLB consistency problem as I thought at
 >   one point. There is no missed IPI or anything, the TLB shootdown
 >   queues are empty when the problem occurs.

I still this that this is your most likely candidate.  Just because
the queues are empty when you see the *symptom* doesn't mean they
weren't inconsistent when the page actually was scrambled.

I really think the TLB shootdown code should be using a barrier.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>