[S390] cio: prevent workqueue deadlock

Subchannel reprobing can block the kslowcrw workqueue indefinitely
while waiting for device recognition to finish which is also scheduled
to run on kslowcrw. Prevent this deadlock by moving the waiting
portion of subchannel reprobing to the cio workqueue.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

drivers/s390/cio/css.c

@@ -533,6 +533,17 @@ static int reprobe_subchannel(struct subchannel_id schid, void *data)
 	return ret;
 }
 
+static void reprobe_after_idle(struct work_struct *unused)
+{
+	/* Make sure initial subchannel scan is done. */
+	wait_event(ccw_device_init_wq,
+		   atomic_read(&ccw_device_init_count) == 0);
+	if (need_reprobe)
+		css_schedule_reprobe();
+}
+
+static DECLARE_WORK(reprobe_idle_work, reprobe_after_idle);
+
 /* Work function used to reprobe all unregistered subchannels. */
 static void reprobe_all(struct work_struct *unused)
 {
@@ -540,10 +551,12 @@ static void reprobe_all(struct work_struct *unused)
 
 	CIO_MSG_EVENT(4, "reprobe start\n");
 
-	need_reprobe = 0;
 	/* Make sure initial subchannel scan is done. */
-	wait_event(ccw_device_init_wq,
-		   atomic_read(&ccw_device_init_count) == 0);
+	if (atomic_read(&ccw_device_init_count) != 0) {
+		queue_work(ccw_device_work, &reprobe_idle_work);
+		return;
+	}
+	need_reprobe = 0;
 	ret = for_each_subchannel_staged(NULL, reprobe_subchannel, NULL);
 
 	CIO_MSG_EVENT(4, "reprobe done (rc=%d, need_reprobe=%d)\n", ret,