Re: Semantics of smp_mb() [was : Re: [PATCH] Fix RCU race in access of nohz_cpu_mask ]

["Paul E. McKenney" <paulmck@us.ibm.com>]

On Wed, Dec 14, 2005 at 02:12:53AM +0100, Andi Kleen wrote:
> On Tue, Dec 13, 2005 at 02:50:59PM -0800, Paul E. McKenney wrote:
> > On Wed, Dec 14, 2005 at 09:27:41AM +1100, Keith Owens wrote:
> > > On Tue, 13 Dec 2005 08:20:27 -0800, 
> > > "Paul E. McKenney" <paulmck@us.ibm.com> wrote:
> > > >If the variable p references MMIO rather than normal memory, then
> > > >wmb() and rmb() are needed instead of smp_wmb() and smp_rmb().
> > > 
> > > mmiowb(), not wmb().  IA64 has a different form of memory fence for I/O
> > > space compared to normal memory.  MIPS also has a non-empty form of
> > > mmiowb().
> > 
> > New one on me!
> 
> Didn't it make only a difference on the Altix or something like that? 
> I suppose they added it only on the drivers for devices supported by SGI.

It could potentially help on a few other CPUs, but quite a few driver
changes would be needed to really bring out the full benefits.  I am
concerned that the current state leaves a number of CPU families broken --
the empty definitions cannot be good for other weakly ordered CPUs!

						Thanx, Paul