That doesn't logically follow because the aggregate of the mode and the
optional flags _are_ the policy itself. If you want to know whether a
policy is interleave, for example, and don't care whether it is referring
to static (absolute) node ids, then you need to mask that off.
The reality of the kernel code is that these policies are not only
restricted to kernel/mempolicy.c. They are also shared with filesystem
code that store them in a single member of a struct as well. The
interface between those two are functions that would now need to be
modified to include additional parameters to pass the flags along.
Additionally, these flags need to be "glued together" with the mode in
userspace to pass to the syscalls anyway, so they're facing the exact
same challenge. So once this gets a little traction, I think it will
quickly become the norm for how you think about the 'policy' member of the
struct.
David
--
