Hi Raul
any reason you are not using native pfring filters? They are more efficient 
than bpf. See pfring_add_filtering_rule in pfcount.c as example. You can change 
them at runtime.

Alfredo

> On 10 Nov 2015, at 07:24, K Rahul <[email protected]> wrote:
> 
> On 11/06/2015 08:31 PM, Alfredo Cardigliano wrote:
>> Rahul
>> do you really need all those sockets, it looks like a bad design choice.
>> 
>> Alfredo
>> 
>>> On 06 Nov 2015, at 15:59, K Rahul<[email protected]>  wrote:
>>> 
>>> On 11/05/2015 05:29 PM, K Rahul wrote:
>>>> On 11/04/2015 08:58 PM, Alfredo Cardigliano wrote:
>>>>> Hi Rahul
>>>>> you should increase MAX_NUM_RING_SOCKETS in kernel/linux/pf_ring.h,
>>>>> currently it is set to 128.
>>>>> 
>>>>> Alfredo
>>>>> 
>>>>>> On 04 Nov 2015, at 16:00, K Rahul<[email protected]>  
>>>>>> wrote:
>>>>>> 
>>>>>> Hi,
>>>>>> 
>>>>>> we are using pfring to capture multicast packets. We are not using DNA 
>>>>>> or ZC drivers as our NIC does not support the same.We are using Broadcom 
>>>>>> Netxtreme BCM5721 NIC. If I use pfring_open more than 64 times, i.e  
>>>>>> pfring_open is called for 64 different multicast streams, then it 
>>>>>> returns a NULL. Is there any limit to pfring handles? If there is a 
>>>>>> limit, how can I increase it?
>>>>>> 
>>>>>> Thanks
>>>>>> 
>>>>>> Regards,
>>>>>> Rahul
>>>>>> _______________________________________________
>>>>>> 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
>>>> Thank you for the prompt reply
>>>> _______________________________________________
>>>> Ntop-misc mailing list
>>>> [email protected]
>>>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>>> Hi,
>>> 
>>> We have increased the limit of MAX_NUM_RING_SOCKETS to 128 from 64. Now, 
>>> after opening 50 handles and more we are observing packet drops. Is there 
>>> any other variable that need to be handle to capture all the packets?
>>> 
>>> Thank you
>>> 
>>> Regards,
>>> K Rahul
>>> _______________________________________________
>>> 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
> Hi,
> 
> Let me explain you the use of pfring in our module. We are using pfring to 
> capture mpeg2ts streams over multicast. We can have multiple multicast 
> streams joined and each one of these streams uses one pfring handle. Thus as 
> per our current design, if we need to capture 100 multicast streams, we will 
> require 100 pfring sockets, each with a specific bpf filter like "udp and dst 
> port 9999 and src host (172.16.0.10) and dst host 239.99.99.99".
> 
> The reason for choosing such a model is that our module can join or leave 
> multicast streams in runtime.For ex, if we are capturing a multicast stream, 
> we set the bpf filter specific to the stream. Now, in case we need to capture 
> another multicast stream using the same pfring handle, we will need to modify 
> the bpf filter at runtime. As per our current understanding, this is not 
> feasible. Thus, we decided to use one pfring socket per multicast stream.
> 
> It would be very helpful if you can suggest any alternative approach.
> 
> Thank you
> 
> Regards,
> K Rahul
> _______________________________________________
> 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