> Right now the memory handler in this container subsystem is written in
> such a way that when existing kernel reclaimer kicks in, it will first
> operate on those (container with pages over the limit) pages first. But
> in general I like the notion of containerizing the whole reclaim code.