Hello, Christoph. On Wed, Apr 13, 2011 at 05:23:04PM -0500, Christoph Lameter wrote: > > Suggested fixup. Return from slowpath and update percpu variable under > spinlock. > > Signed-off-by: Christoph Lameter <cl@xxxxxxxxx> > > --- > lib/percpu_counter.c | 8 ++------ > 1 file changed, 2 insertions(+), 6 deletions(-) > > Index: linux-2.6/lib/percpu_counter.c > =================================================================== > --- linux-2.6.orig/lib/percpu_counter.c 2011-04-13 17:20:41.000000000 -0500 > +++ linux-2.6/lib/percpu_counter.c 2011-04-13 17:21:33.000000000 -0500 > @@ -82,13 +82,9 @@ void __percpu_counter_add(struct percpu_ > spin_lock(&fbc->lock); > count = __this_cpu_read(*fbc->counters); > fbc->count += count + amount; > + __this_cpu_write(*fbc->counters, 0); > spin_unlock(&fbc->lock); > - /* > - * If cmpxchg fails then we need to subtract the amount that > - * we found in the percpu value. > - */ > - amount = -count; > - new = 0; > + return; Yeah, looks pretty good to me now. Just a couple more things. * Please fold this one into the original patch. * While you're restructuring the functions, can you add unlikely to the slow path? It now looks correct to me but just in case, Eric, do you mind reviewing and acking it? Thanks. -- tejun -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>