Thank you for your help, but I am still not able to ping the two hosts.
I updated the netfpga-base version and now I have the 3.0.0. I did
what you told me, and included all the 4 interfaces, but the
destination is unreachable.
So I used wireshark to see what happens and I saw a few things:
1. When I launch the switch, there are periodically some echo request
/reply of OFP packets and a TCP ack in the lo interface, but they have
a wrong TCP checksum ( Checksum: 0xfe30 [incorrect, should be 0x3147
(maybe caused by "TCP checksum offload"?)]);
2. When I do the ping, in the eth1 interface I can see ARP message
(Broadcast ARP Who has 192.168.2.5? Tell 192.168.2.10) but it seems
it can't resolve the address, and so there are ICMP message in the lo
saying destination unreachable.
3. To be complete, I include the output of the ofdatapath /ofprotocol
terminal after the beginning of the connection:
1(nf2c0): addr:00:4e:46:32:43:00, config: 0, state:0
current: 1GB-FD
advertised: COPPER
supported: 1GB-FD
2(nf2c1): addr:00:4e:46:32:43:01, config: 0, state:0
current: 1GB-FD
advertised: COPPER
supported: 1GB-FD
3(nf2c2): addr:00:4e:46:32:43:02, config: 0, state:0
current: 1GB-FD
advertised: COPPER
supported: 1GB-FD
4(nf2c3): addr:00:4e:46:32:43:03, config: 0, state:0
current: 1GB-FD
advertised: COPPER
supported: 1GB-FD
LOCAL(tap0): addr:00:23:20:86:f2:19, config: 0, state:0
current: 10MB-FD COPPER
After that, there are polling messages repeated periodically like that:
Oct 06 11:09:33|00020|poll_loop|DBG|1000-ms timeout: 0x8053765
0x8053882 0x18ae9c
Can you help me? Is it possible that the problem is in the openflow I installed?
Thank you very much.
2011/10/5 Tatsuya Yabe <[email protected]>:
> Hi Leonardo,
>
> Thank you for the info.
>
> For 1, actually 2.1.0 is quite old and since then some major changes have
> been applied.
> One of the changes from the older version is a driver, and you can't pass
> packets from a later hardware (including OpenFlow implementation) to the
> host via an older driver.
>
> So you need to have at least 2.1.3 or later version of NetFPGA base package.
> You can upgrade it with "yum update netfpga-base", but please consult the
> NetFPGA user guide for detail.
>
> For 4, you need to include all the NetFPGA ports for running ofdatapath as
> stated in the installation wiki, so it has to be
>>
>> ofdatapath punix:/var/run/dp0 -i nf2c0,nf2c1,nf2c2,nf2c3
>
>
> Just now I did the same trial with your configuration (with NetFPGA 3.0.0 on
> CentOS5.6, and modifying the argument as above), and the ping worked fine
> for me.
>
> If the fix above doesn't work for you, I would use wireshark to capture
> local host packets and observe OpenFlow protocols.
>
> Thank you,
> Tatsuya
>
>
> On 2011/10/05, at 14:55, Leonardo Linguaglossa wrote:
>
>> Hello. I'm answering your questions:
>>
>> 1. I installed the netfpga-base package with yum, the 2.1.0 version;
>> 2. I followed that tutorial, and the installation went well;
>> 3. For the regression tests I had a problem: I did not have a 4-port
>> NIC to use with the 4 NetFPGA interfaces, so I tryed the regression
>> test suite just with two connections: some initial tests went well
>> without error, some others failed, but I thought this could be because
>> there were some interfaces not connected;
>> 4. When I say commands, I mean that i launch the controller with
>> controller ptcp:6633 -v
>> and then I do the ofdatapath and ofprotocol this way:
>> ofdatapath punix:/var/run/dp0 -i nf2c2,nf2c3 -v
>> ofprotocol unix:/var/run/dp0 tcp:localhost:6633 -v
>>
>> How I said before, the nf2c3 interface is connected with the eth0 of
>> the pc, and the nf2c2 is connected with the eth1 of the cube itself; I
>> should have managed to ping the eth1 address with the pc, but i
>> couldn't.
>> I hope you could help me.
>>
>> 2011/10/5 Tatsuya Yabe <[email protected]>:
>>>
>>> Hi Leonardo,
>>>
>>> This is not a direct answer and I need to ask you from the first step
>>> but:
>>> 1. What version of "NetFPGA base" do you have? The simplest way is to
>>> check
>>> "CHANGES" file in your netfpga installed directory.
>>> 2. I assume you have followed the following wiki for the installation:
>>> http://www.openflow.org/wk/index.php/CentOS_NetFPGA_Install
>>> 3. Did you try the openflow regression test mentioned above to make sure
>>> your installation was successful?
>>> 4. When you mentioned running the commands, can you show me what exactly
>>> did
>>> you run (with arguments)?
>>>
>>> Thank you,
>>> Tatsuya Yabe
>>>
>>> On 2011/10/05, at 10:16, Leonardo Linguaglossa wrote:
>>>
>>>> Hello everybody.
>>>> I am realizing some experiments with OpenFlow, but I feel that I am
>>>> doing something incorrect.
>>>> I wanted to use my NetFPGA cube as an OpenFlow Switch with 2 ports,
>>>> and trying to communicate with a pc with this configuration:
>>>> NetFPGA cube: controller localhost + switch port nf2c3 linked with
>>>> eth0 of the PC host, and nf2c2 connected with eth1 of the cube;
>>>> Pc host: eth0 connected with nf2c3.
>>>>
>>>> This way, I run all the commands (controller, ofdatapath, ofprotocol)
>>>> and I wanted to ping the eth1 with the eth0 of pc host and vice versa.
>>>> Addresses are:
>>>> PC -> eth0:192.168.2.10
>>>> CUBE-> eht1:192.168.2.5
>>>>
>>>> The problem is that when I try to ping each other, it says destination
>>>> unreachable.
>>>> Am I doing something wrong with this configuration? May be the other
>>>> "terminal" to ping can't stay in the same place of the switch, and so
>>>> I need to add another host?
>>>> Thank you.
>>>> --
>>>> Leonardo
>>>> _______________________________________________
>>>> openflow-discuss mailing list
>>>> [email protected]
>>>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
>>>
>>>
>>
>>
>>
>> --
>> Leonardo
>
>
--
Leonardo
_______________________________________________
openflow-discuss mailing list
[email protected]
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
