On Thu, Feb 06, 2014 at 07:12:04PM +0000, Paul E. McKenney wrote: > On Thu, Feb 06, 2014 at 02:48:28PM +0100, Peter Zijlstra wrote: > > Because atomic ops are implemented the same across an architecture, > > the current incomplete set of extra barriers: > > > > smp_mb__before_atomic_inc() > > smp_mb__after_atomic_inc() > > smp_mb__before_atomic_dec() > > smp_mb__after_atomic_dec() > > smp_mb__before_clear_bit() > > smp_mb__after_clear_bit() > > > > is both incomplete and superfluous. > > > > It is incomplete because there are far more atomic operations that do > > not return values -- such as atomic_add(), set_bit() etc. And it is > > superfluous because they're all the same anyway. > > > > Simplify things by reducing the triplicate set into a single set of > > barriers that is valid for all void atomic ops: > > > > smp_mb__before_atomic() > > smp_mb__after_atomic() > > > > Signed-off-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx> > > I very much like the API shrinkage. The RCU changes are good, and > I believe that the rest is OK too, though my eyes were going a bit > buggy towards the end... > > Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx> ... and the arm[64] parts look fine to me. Acked-by: Will Deacon <will.deacon@xxxxxxx> Usual comment about upcoming conflicts :) Will -- To unsubscribe from this list: send the line "unsubscribe linux-arch" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
