Thank you, very much. It was my mistake. 
I have used env variable PF_RING_RSS_REHASH=1 with libpcap. And it is working 
now. 

Alfredo Cardigliano <[email protected]> написал(а):

>Intel RSS is asymmetric by default, where did you read it's symmetric?
>With TNAPI you should set the rss rehash (see
>pfring_enable_rss_rehash()).
>Instead with DNA we remap RSS to make it symmetric (default).
>
>Best Regards
>Alfredo
>
>On Apr 9, 2013, at 2:43 PM, Александр Спиридонов <[email protected]>
>wrote:
>
>> 
>> Hi all,
>> My students and I use TNAPI ixgbe driver with kernel 2.6.37 for
>multithread traffic capture.
>> The capture utility is based on libpcap. We also use PF_RING.
>> 
>> Our research goal is to recreate TCP/IP sessions.
>> But the packets belonging to the same flow (5-tuple) go to the
>different virtual devices.
>> For example: all the packets from IP1:port1 to IP2:port2 go to eth0@0
>device and
>> the packets from IP2:port2 to IP1:port1 go to eth0@1 device.
>> 
>> But Intel docs say that all the packets belonging to the
>> same flow must go to the same queue.
>> 
>> Pre-configuration on the server:
>>    modprobe ixgbe MQ=1,1 RSS=8,8 FdirMode=0,0
>>    ifconfig eth0 promisc up
>>    ethtool --offload eth0 rx off tx off
>>    ethtool -s eth0 speed 1000 duplex full
>> 
>> Could you help us to solve the problem?
>> 
>> 
>> _______________________________________________
>> Ntop-misc mailing list
>> [email protected]
>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>
>_______________________________________________
>Ntop-misc mailing list
>[email protected]
>http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc

Reply via email to