On Mon, 2011-05-09 at 13:40 -0700, Greg KH wrote:
> On Fri, May 06, 2011 at 09:08:34AM +0200, Andi Kleen wrote:
> > > @@ -168,8 +168,13 @@ int __read_mostly timekeeping_suspended;
> > > static struct timespec xtime_cache __attribute__ ((aligned (16)));
> > > void update_xtime_cache(u64 nsec)
> > > {
> > > - xtime_cache = xtime;
> > > - timespec_add_ns(&xtime_cache, nsec);
> > > + /*
> > > + * Use temporary variable so get_seconds() cannot catch
> > > + * an intermediate xtime_cache.tv_sec value.
> > > + */
> > > + struct timespec ts = xtime;
> > > + timespec_add_ns(&ts, nsec);
> >
> > There's nothing to stop the compiler's optimizer to use memory accesses
> > directly again. Something similar to ACCESS_ONCE (that is essentially a
> > volatile cast) may be required to be fully future proof.
>
> I agree, this doesn't look like a viable solution at all.
So does something like this look better? I'm really not that familiar
with how ACCESS_ONCE is to be used, so forgive me if I'm not using it
correctly.
Signed-off-by: John Stultz <[email protected]>
Index: linux-2.6.32.y/kernel/time/timekeeping.c
===================================================================
--- linux-2.6.32.y.orig/kernel/time/timekeeping.c 2011-05-04
19:34:21.604314152 -0700
+++ linux-2.6.32.y/kernel/time/timekeeping.c 2011-05-10 12:15:48.756236916
-0700
@@ -168,8 +168,15 @@ int __read_mostly timekeeping_suspended;
static struct timespec xtime_cache __attribute__ ((aligned (16)));
void update_xtime_cache(u64 nsec)
{
- xtime_cache = xtime;
- timespec_add_ns(&xtime_cache, nsec);
+ /*
+ * Use temporary variable so get_seconds() cannot catch
+ * an intermediate xtime_cache.tv_sec value.
+ * The ACCESS_ONCE() keeps the compiler from optimizing
+ * out the intermediate value.
+ */
+ struct timespec ts = xtime;
+ timespec_add_ns(&ts, nsec);
+ ACCESS_ONCE(xtime_cache) = ts;
}
/* must hold xtime_lock */
_______________________________________________
stable mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/stable
