Re: "Badness in local_bh_enable" - a reasonable fix?

[Stephen Hemminger <shemminger@osdl.org>]

On Wed, 26 Oct 2005 15:34:38 -0400
Steve Snyder <R00020C@freescale.com> wrote:

> [ I observed the following on a Fedora Core 3 system, running kernel 
> 2.6.12-1.1380_FC3.  I am posting this here because a quick Googling 
> indicates that the problem is not specific to this environment.  ] 
> 
> Today I found my system log filled with the error shown below.  
> Reading a 366MB file across a NFS mount results in over 6300 
> occurrences of the error being written to the system log of the NFS 
> server.  
> 
> I have 2 network interfaces in the NFS server machine, a standard 
> kernel Ethernet device driver and my own Ultra-Wide Band (UWB) device 
> driver.  (In the error shown below the references to "fsuwbpci" are my 
> driver.) This problem is not seen when using the Ethernet interface, 
> but is perfectly consistent when reading a NFS-mounted file across the 
> UWB interface.  Therefore there is a problem with my code.  
> 
> I quickly established that the error came from within this routine:
> 
>   void netdev_tx_ack(struct net_device *dev, struct sk_buff *skb)
>   {
>      struct  netdev_priv *priv = (struct netdev_priv *) dev->priv;
>   
>      priv->stats.tx_packets++;
>      priv->stats.tx_bytes += skb->len;
>   
>      netdev_resume(dev);
>      dev_kfree_skb(skb);
>   }

Your driver is calling dev_kfree_skb with interrupts disabled.
Call dev_kfree_skb_any instead.

-- 
Stephen Hemminger <shemminger@osdl.org>
OSDL http://developer.osdl.org/~shemminger