ratelimit: Fix/allow use in atomic contexts

I'd like to use printk_ratelimit() in NMI context, but it's not
robust right now due to spinlock usage in lib/ratelimit.c. If an
NMI is unlucky enough to hit just that spot we might lock up trying
to take the spinlock again.

Fix that by using a trylock variant. If we contend on that lock we
can genuinely skip the message because the state is just being
accessed by another CPU (or by this CPU).

( We could use atomics for the suppressed messages field, but
  i doubt it matters in practice and it makes the code heavier. )

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: David S. Miller <davem@davemloft.net>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

lib/ratelimit.c

@@ -28,7 +28,15 @@ int __ratelimit(struct ratelimit_state *rs)
 	if (!rs->interval)
 		return 1;
 
-	spin_lock_irqsave(&rs->lock, flags);
+	/*
+	 * If we contend on this state's lock then almost
+	 * by definition we are too busy to print a message,
+	 * in addition to the one that will be printed by
+	 * the entity that is holding the lock already:
+	 */
+	if (!spin_trylock_irqsave(&rs->lock, flags))
+		return 1;
+
 	if (!rs->begin)
 		rs->begin = jiffies;