On Fri, May 20, 2011 at 10:36:48AM +0200, Ingo Molnar wrote:
>
> * Frederic Weisbecker <[email protected]> wrote:
>
> > HARDIRQ_ENTER() maps to irq_enter() which calls rcu_irq_enter().
> > But HARDIRQ_EXIT() maps to __irq_exit() which doesn't call
> > rcu_irq_exit().
> >
> > So for every locking selftest that simulates hardirq disabled,
> > we create an imbalance in the rcu extended quiescent state
> > internal state.
> >
> > As a result, after the first missing rcu_irq_exit(), subsequent
> > irqs won't exit any extended quiescent state the time of the
> > interrupt servicing. Rcu read side critical section inside irqs
> > on that CPU won't be guaranteed anymore.
> >
> > To fix this, just use __irq_enter() to simulate the hardirq
> > context. This is sufficient for the locking selftests as we
> > don't need to exit any extended quiescent state or perform
> > any check that irqs normally do when they wake up from idle.
> >
> > As a side effect, this patch should make it possible to
> > restore
> > "rcu: Decrease memory-barrier usage based on semi-formal proof",
> > which eventually helped finding this bug.
> >
> > Reported-and-tested-by: Yinghai Lu <[email protected]>
> > Signed-off-by: Frederic Weisbecker <[email protected]>
> > Cc: Paul E. McKenney <[email protected]>
> > Cc: Ingo Molnar <[email protected]>
> > Cc: Peter Zijlstra <[email protected]>
> > Cc: Stable <[email protected]>
> > ---
> > lib/locking-selftest.c | 2 +-
> > 1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/lib/locking-selftest.c b/lib/locking-selftest.c
> > index 619313e..507a22f 100644
> > --- a/lib/locking-selftest.c
> > +++ b/lib/locking-selftest.c
> > @@ -144,7 +144,7 @@ static void init_shared_classes(void)
> >
> > #define HARDIRQ_ENTER() \
> > local_irq_disable(); \
> > - irq_enter(); \
> > + __irq_enter(); \
> > WARN_ON(!in_irq());
> >
> > #define HARDIRQ_EXIT() \
>
> Very nice!
>
> Note that a cherry-picked version of the manual revert from Paul is upstream
> now with the main body of RCU changes, so we can do the finegrained split-up
> queue approach for this one problematic commit (that is no longer
> problematic).
>
> I think it can still go in for v2.6.40. Paul, what do you think?
I agree. I will rebase my split-up version, rebase on that, retest
and push.
Thanx, Paul
_______________________________________________
stable mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/stable
