Yes that fixes it here too. And the corruption that I saw of slab variables is explained by your analysis. Thanks!Tested-by: Christoph Lameter <clameter@sgi.com>