No subject

any guaranteed barrier behaviour at all, although it will have a
barrier effect on many implementations.  As we get into v7-land, this
kind of thing becomre more and more unsafe as new implementations push
the architectural envelope.

Haojian, can you comment on whether the proposed sequence will work in
this case of synchronising CPACR updates?  If we can use the same
sequence for all v6/v7/ARM/Thumb-2 kernels, that would be preferable
to #ifdefs.

Cheers
---Dave