On Thu, Dec 31, 2015 at 09:07:10PM +0200, Michael S. Tsirkin wrote:
> -#define smp_store_release(p, v) \
> -do { \
> - compiletime_assert_atomic_type(*p); \
> - smp_mb(); \
> - WRITE_ONCE(*p, v); \
> -} while (0)
> -
> -#define smp_load_acquire(p) \
> -({ \
> - typeof(*p) ___p1 = READ_ONCE(*p); \
> - compiletime_assert_atomic_type(*p); \
> - smp_mb(); \
> - ___p1; \
> -})
David Daney wanted to use fancy new MIPS barriers to provide better
implementations of this.
This patch isn't in the way of that, just a FYI.
