As a kernel developer who has been in the depths of mm/... just.. no. msync() is way too expensive to use in any reasonable codebase. And it is in no way a re

As a kernel developer who has been in the depths of mm/... just.. no. msync() is way too expensive to use in any reasonable codebase. And it is in no way a reliable base for which good software should be built upon. You don't control where the kernel will plop newly allocated memory regions from mmap() unless you start specifying addressess, in which case why do you need to validate them? And what if it's "valid" but it points somewhere it shouldn't? Please, never ever do this in anything that has to run in production.