* Sebastian Andrzej Siewior | 2016-05-12 10:42:30 [+0200]:

quoted

diff --git a/mm/swap.c b/mm/swap.c
index ca194ae..84c3c21 100644
--- a/mm/swap.c
+++ b/mm/swap.c

@@ -821,9 +821,9 @@ void lru_add_drain_cpu(int cpu)

		unsigned long flags;

		/* No harm done if a racing interrupt already did this */
-		local_lock_irqsave(rotate_lock, flags);
+		local_lock_irqsave_other_cpu(rotate_lock, flags, cpu);
		pagevec_move_tail(pvec);
-		local_unlock_irqrestore(rotate_lock, flags);
+		local_unlock_irqrestore_other_cpu(rotate_lock, flags, cpu);

This piece might be required independently of this patch. It would be
nice to have it in page_alloc_cpu_notify() :) I take care of this…

no, I won't :)
It looks halfway like an upstream bug. Vanila has here local_irq_save()
which always matches the local CPU except when it is invoked from
page_alloc_cpu_notify() CPU_DEAD notifier. Here it does not matter
because the CPU is gone and won't take the locks anymore.
Unless, we start taking them cross-CPU which is only possible with this
patch.

Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html