On 11/30/2016 03:44 AM, Richard Cochran wrote:

On Mon, Nov 28, 2016 at 05:04:23PM -0600, Grygorii Strashko wrote:

quoted

@@ -678,6 +744,9 @@ struct gbe_priv {
 	int				num_et_stats;
 	/*  Lock for updating the hwstats */
 	spinlock_t			hw_stats_lock;
+
+	int                             cpts_registered;

The usage of this counter is racy.

quoted

+	struct cpts                     *cpts;
 };

This ++ and -- business ...

quoted

+static void gbe_register_cpts(struct gbe_priv *gbe_dev)
+{
+	if (!gbe_dev->cpts)
+		return;
+
+	if (gbe_dev->cpts_registered > 0)
+		goto done;
+
+	if (cpts_register(gbe_dev->cpts)) {
+		dev_err(gbe_dev->dev, "error registering cpts device\n");
+		return;
+	}
+
+done:
+	++gbe_dev->cpts_registered;
+}
+
+static void gbe_unregister_cpts(struct gbe_priv *gbe_dev)
+{
+	if (!gbe_dev->cpts || (gbe_dev->cpts_registered <= 0))
+		return;
+
+	if (--gbe_dev->cpts_registered)
+		return;
+
+	cpts_unregister(gbe_dev->cpts);
+}

is invoked from your open() and close() methods, but those methods
are not serialized among multiple ports.

ok. Seems my assumption that ndo_open/ndo_close serialized by rtnl_lock is incorrect. Right?
net_device_ops.ndo_open ->
 netcp_ndo_open
 gbe_open
 gbe_register_cpts

-- 
regards,
-grygorii
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html