Re: 2.6.28: warn_slowpath in orinoco receive path

[Robert Hancock <hancockr@shaw.ca>]

Jiri Slaby wrote:
> On 01/05/2009 09:32 PM, Dave wrote:
>> --- a/drivers/net/wireless/orinoco/orinoco.c
>> +++ b/drivers/net/wireless/orinoco/orinoco.c
>> @@ -1616,9 +1616,15 @@ static void orinoco_rx_isr_tasklet(unsigned long
>> data)
>>
>>         /* extract desc and skb from queue */
>>         list_for_each_entry_safe(rx_data, temp, &priv->rx_list, list) {
>> +               unsigned long flags;
>> +
>>                 desc = rx_data->desc;
>>                 skb = rx_data->skb;
>> +
>> +               local_irq_save(flags);
>>                 list_del(&rx_data->list);
>> +               local_irq_restore(flags);
>> +
> 
> Hi,
> 
> another processor still can see inconsistent state, spinlock should be taken.
> Or, am I missing something?

Think you're right, this should be spin_lock_irq or spin_lock_irqsave on 
some lock. Otherwise the interrupt could still occur on some other CPU 
and hit the race.