In message [ref],Karl Hiramoto writes:

Anyone care to test or comment on  these patches?  I've attached 
versions for 2.6.28 and 2.6.30.

this needs to be against the net-2.6 git repository.  but the 2.6.30
would probably apply just fine.  except for the comments, below this
patch looks fine and makes sense.  a similar thing had to be done for
the lec.c interface.

+/* chained vcc->pop function.  Check if we should wake the netif_queue */
+static void br2684_pop(struct atm_vcc *vcc, struct sk_buff *skb)
+{
+	struct br2684_vcc *brvcc = BR2684_VCC(vcc);
+	struct net_device *net_dev = brvcc->device;
+
+	pr_debug("br2684_pop(vcc %p ; net_dev %p )\n", vcc, net_dev);
+	brvcc->old_pop(vcc, skb);
+
+	if (!net_dev)
+		return;
+
+ 	if (atm_may_send(vcc, 0)) {
+		netif_wake_queue(net_dev);
+ 	}

you dont need the { } here.

+
+	if (!atm_may_send(atmvcc, 0)) {
+  		netif_stop_queue(brvcc->device);
+		barrier();
+		/* check for race with br26864_pop*/
+		if (atm_may_send(atmvcc, 0)) {
+			netif_start_queue(brvcc->device);
+		}
+	}
+

i dont think the barrier() is necessary.