Re: [patch] revert: [NET]: Fix races in net_rx_action vs netpoll

[Jarek Poplawski <jarkao2@o2.pl>]

On 16-07-2007 11:12, Ingo Molnar wrote:
> current -git broke my main testbox. No TCP/IP networking to/from the box 
> and e1000 would time out in xmit:
> 
>  NETDEV WATCHDOG: eth0: transmit timed out
>  e1000: eth0: e1000_clean_tx_irq: Detected Tx Unit Hang
...

Olaf, I think this error can trigger in this place:

> static void net_rx_action(struct softirq_action *h)
> {
>         struct softnet_data *queue = &__get_cpu_var(softnet_data);
>         unsigned long start_time = jiffies;
>         int budget = netdev_budget;
>         void *have;
> 
>         local_irq_disable();
> 
>         while (!list_empty(&queue->poll_list)) {
>                 struct net_device *dev;
> 
>                 if (budget <= 0 || jiffies - start_time > 1)
>                         goto softnet_break;
> 
>                 local_irq_enable();
> 
>                 dev = list_entry(queue->poll_list.next,
>                                  struct net_device, poll_list);
>                 have = netpoll_poll_lock(dev);
> 
>                 if (dev->quota <= 0 || dev->poll(dev, &budget)) {

If after poll_napi dev->quota <= 0 dev->poll is not run and
__LINK_STATE_RX_SCHED bit (plus dev->poll_list) stays uncleared.

Regards,
Jarek P.