On 11/11/2010 10:56 AM, Tomoya MORINAGA wrote:

quoted

quoted

quoted

quoted

+ if (priv->tx_obj == (PCH_OBJ_NUM + 1)) { /* Point tail Obj + 1 */
+  while (ioread32(&priv->regs->treq2) & 0xfc00)
+   udelay(1);

When points tail of Tx message object,
this driver waits until completion of all tx messaeg objects.

Looping busy it not an option here.

quoted

Thus, application/driver ought not to be able to put Tx object exceed the number of tx message object.
Thus I think these code(netif_stop_queue/netif_wake_queue) are completely redundant.

Nope - please remove the waiting completely and convert your flow
control to netif_stop_queue/netif_wake_queue.

I see.
I will remove like above.

BTW, Let me know the reason.
Could you explain the reason why you deny looping busy loop ?

That would kill performance.

cheers, Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |