[Nfd-dev] generatation of NDN-like packets with Scapy

[Junxiao Shi]

Hi Salvatore

You can have "your software" listen on a separate Unix socket, and let
ndnpeek connect to it directly without going through NFD. "your software"
processes the Interest and forwards to veth0 without going through NFD.

ndnpoke still goes through NFD, so you don't need to edit out the code that
does prefix registration.

This requires two different client.conf used by ndnpeek and ndnpoke, which
can be achieved by setting different HOME environ when starting those two
programs.

Yours, Junxiao
On Jan 6, 2016 01:40, "Salvatore Signorello" <salvatore.signorello at uni.lu>
wrote:

> Hi Junxiao,
>
> thanks for your prompt feedback, below follows a short description of my
> quick&dirty "set-up" and then few more comments in-line:
>
> Scenario in mind (caveat: I don't know if it makes sense, but that's what
> I need; so if you think of an alternative, as I guess you already did
> according to your previous suggestion, please feel free of throwing away
> what follows and propose sth else)
> -----------------------
> On the same machine where nfd is running, I use the ndnpeedk to generate
> an Interest with prefixA. My nfd has a rule that forwards Interests with
> prefixA out to a specific veth1. On that veth1 I have some software running
> that processes the Interest and then forwards the same Interest back to the
> nfd through a different interface, veth2. In the meantime I've started a
> local producer (ndnpoke -w) for that content that correctly receives the
> Interest(the 2nd one received on veth and not the 1st one issued by
> ndnpeek) and answers back with the Data. The nfd forwards the data back to
> veth2, the software processes it and then forwards it to nfd through veth1.
> Nfd forwards the data back to ndnpeek.
>
> Why am I doing this? I need two applications, one consumer and one
> producer, like ndnpeek and ndnpoke that generate ndn packets.
>
> The problems that I would like to avoid are the following:
> - the pit record for the Interest issued by ndnpeek risks to  drop the
> Interest received on veth2 [SOLVED] I change the nonce when I process the
> Interest for the 2nd time.
> - ndnpeek and ndnpoke cannot be started simultaneously, otherwise ndnpoke
> will provide the Data to ndnpeek in one step
> - if the 1st Interest creates a PIT record, how to avoid that nfd will use
> it when receiving Data from ndnpoke? The daemon will have a PIT record like
> the following
>         "prefixA/content ---- facex(local to ndnpeek), facey(to veth2)"
>     and it should choose only facey
>
> Quick and dirty workaround
> -----------------------------------------
> By now I simulate the scenario above in the following way:
> 1. Starting nfd and registering prefixA to veth1 with the nfdc cmd
> 2. Issuing an Interest for prefixA/content with ndnpeek
> 3. The Interest is correctly forwarded to my software through veth1 where
> I hold it for a while
> 4. After ndnpeek timeout, I unregister prefixA from nfd
> 5. Starting ndnpoke for prefixA/content with a long waiting time
> 6. Unpausing my software that forwards the original Interest to nfd
> through veth2
> 7. nfd forwards it to ndnpoke and I get the data back from veth2
> 8. holding again the Data
> 9. register prefixA again like done in 2
> 10. issuing a new Interest with ndnpeek
> 11. unpausing the Data to forwarded it back first to nfd and then to
> ndnpeek
>
> Crazy, isn't it?
>
> On 01/05/2016 01:35 PM, Junxiao Shi wrote:
>
> Hi Salvatore
>
> I'll offer a different idea to solve your problem: pretend to be NDN
> forwarder.
>
> NDN programs recognizes $HOME/.ndn/client.conf, and connects to the NDN
> forwarder specified in "transport" option.
> http://named-data.net/doc/ndn-cxx/current/manpages/ndn-client.conf.html
>
>
> Cool, I didn't know about this config option. I guess this does mean that
> the local apps use TCP instead of Unix socket as transport connection
> towards the local forwarder. Have I got it right?
>
> You can point the transport to the TCP port or Unix socket listener of
> your next stage, and use existing NDN programs.
>
> My other stage is a veth. Does it make any difference with what you're
> thinking of? After reading this one, I'm not sure that I've got the
> previous sentence well. Could you please elaborate a little bit according
> to the scenario I've described.
>
> Consumer programs should with just fine.
> Producer programs need small modifications: bypass prefix registration
> step.
>
> Yours, Junxiao
> On Jan 5, 2016 05:27, "Salvatore Signorello" <salvatore.signorello at uni.lu>
> wrote:
>
>> Hi all,
>>
>> I would like to generate some NDN packets that are not processed in a
>> first stage by the NFD daemon, so I guess that I cannot use ready-made
>> things like ndnpeek/pook/ping/etc.
>>
>> Has anyone already implemented an NDN module for a packet generator like
>> Scapy? If so, or if there exists other ways to do that, could you please
>> point me to the right resources?
>>
>> Do I really need to write a module for a packet generator? I mean: does
>> the ndn-repo have anything else that may fit the purpose?
>>
>> Any help would be really appreciated,
>> Salvatore
>>
>> --
>> Salvatore Signorello
>> PhD student @ SecanLab
>>
>> Interdisciplinary Centre for Security, Reliability and Trust
>> SnT, University of Luxembourg
>> http://wwwen.uni.lu/snt/people/salvatore_signorello
>>
>> _______________________________________________
>> Nfd-dev mailing list
>> Nfd-dev at lists.cs.ucla.edu
>> http://www.lists.cs.ucla.edu/mailman/listinfo/nfd-dev
>>
>
> --
> Salvatore Signorello
> PhD student @ SecanLab
>
> Interdisciplinary Centre for Security, Reliability and Trust
> SnT, University of Luxembourghttp://wwwen.uni.lu/snt/people/salvatore_signorello
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/nfd-dev/attachments/20160106/7b4821dd/attachment.html>