Re: [PATCH 1/3] kvm tools: Make virtio-net kvm__irq_line thread safe

[Asias He <asias.hejun@gmail.com>]

On 04/29/2011 03:30 PM, Ingo Molnar wrote:
> 
> * Asias He <asias.hejun@gmail.com> wrote:
> 
>> This patch fixes virtio-net randmom stall
>>
>> host $ scp guest:/root/big.guest .
>> big.guest 42%  440MB  67.7KB/s - stalled -
> 
>> @@ -77,9 +77,12 @@ static void virtio_net_rx_callback(struct kvm *self, void *param)
>>  		head = virt_queue__get_iov(vq, iov, &out, &in, self);
>>  		len = readv(net_device.tap_fd, iov, in);
>>  		virt_queue__set_used_elem(vq, head, len);
>> -	}
>>  
>> -	kvm__irq_line(self, VIRTIO_NET_IRQ, 1);
>> +		/* We should interrupt guest right now, otherwise latency is huge. */
>> +		mutex_lock(&net_device.mutex);
>> +		kvm__irq_line(self, VIRTIO_NET_IRQ, 1);
>> +		mutex_unlock(&net_device.mutex);
> 
> The movement of the irq-line assertion call (which fixes the hangs) should be 
> separated from the mutex_lock() additions (which look unnecessary).

I've sent out another patch to move this irq assertion.

> 
> Could you please check whether the addition of the mutex really is necessary to 
> solve the hang/stall?


Without the mutex around the kvm__irq_line, I am still seeing the
hangs(with the irq assertion movement patch).

I can reproduce this by:

host$ scp guest:/root/big.guest .

guest$ ping host

Not sure if the mutex thing is necessary.

> 
> Thanks,
> 
> 	Ingo
> 


-- 
Best Regards,
Asias He