Re: [PATCH] netfront: Lockdep fixes

[Jeremy Fitzhardinge <jeremy@xensource.com>]

Herbert Xu wrote:
> The ->poll method is guaranteed to be called with BH disabled so
> this isn't necessary.
>   

OK.  As you can probably tell, I took a fairly blunt search-and-replace
approach once I'd worked out what lockdep was complaining about.

>> @@ -1588,7 +1588,7 @@ static void netif_release_rx_bufs(struct
>>        while ((skb = __skb_dequeue(&free_list)) != NULL)
>>                dev_kfree_skb(skb);
>>
>> -       spin_unlock(&np->rx_lock);
>> +       spin_unlock_bh(&np->rx_lock);
>> }
>>     
>
> Just a minor nit.  This is normally called with BH disabled,
> except from uninit so you could put a local_bh_disable there
> instead.
>   

OK.

>> static int network_close(struct net_device *dev)
>> @@ -1708,8 +1708,8 @@ static int network_connect(struct net_de
>>        IPRINTK("device %s has %sing receive path.\n",
>>                dev->name, np->copying_receiver ? "copy" : "flipp");
>>
>> +       spin_lock_bh(&np->rx_lock);
>>        spin_lock_irq(&np->tx_lock);
>> -       spin_lock(&np->rx_lock);
>>
>>        /*
>>         * Recovery procedure:
>> @@ -1761,7 +1761,7 @@ static int network_connect(struct net_de
>>        network_tx_buf_gc(dev);
>>        network_alloc_rx_buffers(dev);
>>
>> -       spin_unlock(&np->rx_lock);
>> +       spin_unlock_bh(&np->rx_lock);
>>        spin_unlock_irq(&np->tx_lock);
>>     
>
> You can't enable BH with IRQs disabled.  Besides, for the sake of
> symmetry these two should be reversed.
>   

Will do.

Thanks for looking over it.

    J