cfq-iosched: properly protect ioc_gone and ioc count

If we have multiple tasks freeing cfq_io_contexts when cfq-iosched
is being unloaded, we could complete() ioc_gone twice. Fix that by
protecting ioc_gone complete() and clearing with a spinlock for
just that purpose. Doesn't matter from a performance perspective,
since it'll only enter that path when ioc_gone != NULL (when cfq-iosched
is being rmmod'ed).

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index d01b411..32aa367 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -48,6 +48,7 @@
 
 static DEFINE_PER_CPU(unsigned long, ioc_count);
 static struct completion *ioc_gone;
+static DEFINE_SPINLOCK(ioc_gone_lock);
 
 #define CFQ_PRIO_LISTS		IOPRIO_BE_NR
 #define cfq_class_idle(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
@@ -1177,8 +1178,19 @@
 	kmem_cache_free(cfq_ioc_pool, cic);
 	elv_ioc_count_dec(ioc_count);
 
-	if (ioc_gone && !elv_ioc_count_read(ioc_count))
-		complete(ioc_gone);
+	if (ioc_gone) {
+		/*
+		 * CFQ scheduler is exiting, grab exit lock and check
+		 * the pending io context count. If it hits zero,
+		 * complete ioc_gone and set it back to NULL
+		 */
+		spin_lock(&ioc_gone_lock);
+		if (ioc_gone && !elv_ioc_count_read(ioc_count)) {
+			complete(ioc_gone);
+			ioc_gone = NULL;
+		}
+		spin_unlock(&ioc_gone_lock);
+	}
 }
 
 static void cfq_cic_free(struct cfq_io_context *cic)
@@ -2317,7 +2329,7 @@
 	 * pending RCU callbacks
 	 */
 	if (elv_ioc_count_read(ioc_count))
-		wait_for_completion(ioc_gone);
+		wait_for_completion(&all_gone);
 	cfq_slab_kill();
 }