IEEE1394(r1142): Use a kernel thread to rescan devices so we don't block the writer.

drivers/ieee1394/nodemgr.c

@@ -358,13 +358,26 @@ static ssize_t fw_get_destroy_node(struct bus_type *bus, char *buf)
 }
 static BUS_ATTR(destroy_node, S_IWUSR | S_IRUGO, fw_get_destroy_node, fw_set_destroy_node);
 
+static int nodemgr_rescan_bus_thread(void *__unused)
+{
+	/* No userlevel access needed */
+	daemonize("kfwrescan");
+	allow_signal(SIGTERM);
+
+	bus_rescan_devices(&ieee1394_bus_type);
+
+	return 0;
+}
 
 static ssize_t fw_set_rescan(struct bus_type *bus, const char *buf, size_t count)
 {
 	int state = simple_strtoul(buf, NULL, 10);
 
+	/* Don't wait for this, or care about errors. Root could do
+	 * something stupid and spawn this a lot of times, but that's
+	 * root's fault. */
 	if (state == 1)
-		bus_rescan_devices(&ieee1394_bus_type);
+		kernel_thread(nodemgr_rescan_bus_thread, NULL, CLONE_KERNEL);
 
 	return count;
 }