On Fri, 2007-06-22 at 16:41 -0400, Bill Nottingham wrote:
Chuck Ebbert (cebbert(a)redhat.com) said:
> Why do we explicitly remove atomic.h from our kernel header package?
IIRC, the reasoning was because the operations weren't actually
atomic when used from userspace; ergo, it was a bad idea to provide
... Unless you define CONFIG_SMP (which in turn, defines LOCK_PREFIX)
I assume that up-stream will shoot me for suggesting that... but why not
#define LOCK_PREFIX "lock; "
... and make atomic_xxx safe for user-land?