genirq: Use rcu in kstat_irqs_usr()

Jeremy Dorfman identified mutex contention when multiple threads
parse /proc/stat concurrently.

Since commit 425a5072dcd1 ("genirq: Free irq_desc with rcu"),
kstat_irqs_usr() can be switched to rcu locking, which removes this mutex
contention.

show_interrupts() case will be handled in a separate patch.

Reported-by: Jeremy Dorfman <jdorfman@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Willem de Bruijn <willemb@google.com>
Link: https://lkml.kernel.org/r/20180618125612.155057-1-edumazet@google.com

kernel/irq/irqdesc.c

@@ -443,6 +443,7 @@ static void free_desc(unsigned int irq)
 	 * We free the descriptor, masks and stat fields via RCU. That
 	 * allows demultiplex interrupts to do rcu based management of
 	 * the child interrupts.
+	 * This also allows us to use rcu in kstat_irqs_usr().
 	 */
 	call_rcu(&desc->rcu, delayed_free_desc);
 }
@@ -928,17 +929,17 @@ unsigned int kstat_irqs(unsigned int irq)
  * kstat_irqs_usr - Get the statistics for an interrupt
  * @irq:	The interrupt number
  *
- * Returns the sum of interrupt counts on all cpus since boot for
- * @irq. Contrary to kstat_irqs() this can be called from any
- * preemptible context. It's protected against concurrent removal of
- * an interrupt descriptor when sparse irqs are enabled.
+ * Returns the sum of interrupt counts on all cpus since boot for @irq.
+ * Contrary to kstat_irqs() this can be called from any context.
+ * It uses rcu since a concurrent removal of an interrupt descriptor is
+ * observing an rcu grace period before delayed_free_desc()/irq_kobj_release().
  */
 unsigned int kstat_irqs_usr(unsigned int irq)
 {
 	unsigned int sum;
 
-	irq_lock_sparse();
+	rcu_read_lock();
 	sum = kstat_irqs(irq);
-	irq_unlock_sparse();
+	rcu_read_unlock();
 	return sum;
 }