Am Donnerstag 26 Juni 2008 19:43:15 schrieb Dmitry Torokhov:

On Thu, Jun 26, 2008 at 04:32:16PM +0200, Oliver Neukum wrote:

quoted

Am Donnerstag 26 Juni 2008 16:26:34 schrieb Dmitry Torokhov:

quoted

Hi Oliver,

On Thu, Jun 26, 2008 at 01:44:30PM +0200, Oliver Neukum wrote:

quoted

@@ -809,8 +831,7 @@ static int usb_cleanup(struct yealink_de
 	if (yld == NULL)
 		return err;
 
-	usb_kill_urb(yld->urb_irq);	/* parameter validation in core/urb */
-	usb_kill_urb(yld->urb_ctl);	/* parameter validation in core/urb */
+	stop_traffic(yld);

stop_traffic is unneeded here since input core guarantees that close is
called when you unregister and opened device. Do you have the hardware
to test? If not I will fold stop_traffic back into yealink_close
locally.

No, I don't have hardware to test.

OK, then I'll change it. I wonder whether we really need spinlock
though or a simple wmb() is enough to make sure other CPUs notice
the flag being set before we get to executing usb_kill_urb()...

I am unsure and wouldn't try anything so subtle without a clear need.

	Regards
		Oliver