Re: [PATCH] PATCH 1/2 [SCHED 2.6.24]: Check subqueue status before calling hard_start_xmit

[Krishna Kumar <krkumar2@in.ibm.com>]

Hi Peter,

Peter wrote on 11/13/2007 11:14:50 PM:
> @@ -134,7 +134,7 @@ static inline int qdisc_restart(struct net_device *dev)
>  {
>     struct Qdisc *q = dev->qdisc;
>     struct sk_buff *skb;
> -   int ret;
> +   int ret = NETDEV_TX_BUSY;
>  
>     /* Dequeue packet */
>     if (unlikely((skb = dev_dequeue_skb(dev, q)) == NULL))
> @@ -145,7 +145,8 @@ static inline int qdisc_restart(struct net_device *dev)
>     spin_unlock(&dev->queue_lock);
>  
>     HARD_TX_LOCK(dev, smp_processor_id());
> -   ret = dev_hard_start_xmit(skb, dev);
> +   if (!netif_subqueue_stopped(dev, skb))
> +      ret = dev_hard_start_xmit(skb, dev);
>     HARD_TX_UNLOCK(dev);

You could optimize this by getting HARD_TX_LOCK after the check. I
assume that netif_stop_subqueue (from another CPU) would always be
called by the driver xmit, and that is not possible since we hold
the __LINK_STATE_QDISC_RUNNING bit. Does that sound correct?

PATCH
------

diff -ruNp 1/net/sched/sch_generic.c 2/net/sched/sch_generic.c
--- 1/net/sched/sch_generic.c	2007-11-14 11:14:10.000000000 +0530
+++ 2/net/sched/sch_generic.c	2007-11-14 11:18:27.000000000 +0530
@@ -144,10 +144,11 @@ static inline int qdisc_restart(struct n
 	/* And release queue */
 	spin_unlock(&dev->queue_lock);
 
-	HARD_TX_LOCK(dev, smp_processor_id());
 	if (!netif_subqueue_stopped(dev, skb))
+		HARD_TX_LOCK(dev, smp_processor_id());
 		ret = dev_hard_start_xmit(skb, dev);
-	HARD_TX_UNLOCK(dev);
+		HARD_TX_UNLOCK(dev);
+	}
 
 	spin_lock(&dev->queue_lock);
 	q = dev->qdisc;

Thanks,

- KK