Re: [BUG][tip/master] kernel panic while locking selftest at qspinlock_paravirt.h:137!

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Jul 10, 2015 at 03:57:46PM +0200, Ingo Molnar wrote:
> * Peter Zijlstra <peterz@infradead.org> wrote:

> > Do we want to make double unlock non-fatal unconditionally?
> 
> No, just don't BUG() out, don't crash the system - generate a warning?

So that would be a yes..

Something like so then? Won't this generate a splat on that locking self
test then? And upset people?

---
 kernel/locking/qspinlock_paravirt.h | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
index 04ab18151cc8..286e8978a562 100644
--- a/kernel/locking/qspinlock_paravirt.h
+++ b/kernel/locking/qspinlock_paravirt.h
@@ -133,8 +133,14 @@ static struct pv_node *pv_unhash(struct qspinlock *lock)
 	 * This guarantees a limited lookup time and is itself guaranteed by
 	 * having the lock owner do the unhash -- IFF the unlock sees the
 	 * SLOW flag, there MUST be a hash entry.
+	 *
+	 * This can trigger due to double-unlock. In which case, return a
+	 * random pointer so that __pv_queued_spin_unlock() can dereference it
+	 * without crashing.
 	 */
-	BUG();
+	WARN_ON_ONCE(true);
+
+	return (struct pv_node *)this_cpu_ptr(&mcs_nodes[0]);
 }
 
 /*