> > So I think you're good... But yes, you raise an interresting point.
So it sounds like architectures that don't have an instruction atomic u64 *_user need to disable interrupts during the access, and somehow handle that case when a page fault happens? -Andi
