[Planetlab-users] question on tcpdump on new system

[Sapan Bhatia]

Hi Ning,

I understand the condition. We tried figuring out what was causing
this problem early on during the deployment. We turned off all
PlanetLab-specific extensions, but that didn't make the problem go
away. So I'm guessing that the ICMP types you want might have to be
enabled somewhere in the configuration.

Sapan

On Mon, Jul 7, 2008 at 7:01 PM, Ning Qu <quning at ymail.com> wrote:
> Hi Sapan,
> Thanks a lot for your quick reply. Yes, I am looking at the
> destination unreachable packet.
> So my program opens 2 raw sockets, one socket is used to send UDP
> packet to a high port in the destination host, and the other socket
> is used to send TCP packet. When I do tcpdump on the planetlab
> machine, i can only see the destination unreachable triggered by the
> UDP, but not the TCP. I am very confused about this. Since both of the
> replied packets are the ICMP packets, I have no idea why one received
> while the other not...
> Please let me know if you find any clue to hack this. Thanks!
> Ning
>
>
> On Mon, Jul 7, 2008 at 3:22 PM, Sapan Bhatia <sapanb at cs.princeton.edu> wrote:
>>> One more thing, when I change the TCP packet to UDP packet (all using
>>> raw socket to send), I can get the reply packet.
>>
>> Ok, this limits the issue to TCP then, as has been previously reported.
>>
>> Sapan
>>
>>
>>> thanks!
>>> Ning
>>>
>>> On Mon, Jul 7, 2008 at 2:03 PM, Ning Qu <quning at ymail.com> wrote:
>>>> Hi Sapan,
>>>> Thanks a lot for your reply. My actual problem is like this: I have a
>>>> TCP-based customized ping program. The program uses raw socket to
>>>> create a TCP packet and sends to the high port of the destination. It
>>>> worked fine previously in the old system. But after upgrade, it cannot
>>>> receive the returned packets. I am certain there is replied packets
>>>> sent back from the other end. But the program receives nothing. Have
>>>> you seen this problem before? Is there any difference I should be
>>>> aware of for using the raw socket  in the new system?
>>>>
>>>> thanks!
>>>> Ning
>>>>
>>>>
>>>> On Mon, Jul 7, 2008 at 1:21 PM, Sapan Bhatia <sapanb at cs.princeton.edu> wrote:
>>>>> Hi Ning,
>>>>>
>>>>> Your slice only sees packets that belong to it. If you ping another
>>>>> host from within your slice, then the packets will appear in tcpdump.
>>>>> In this case, the packets belong to the administrative context of the
>>>>> system (and not your slice), so they are not reported via tcpdump.
>>>>>
>>>>> Sapan
>>>>>
>>>>> On 7/7/08, Ning Qu <quning at ymail.com> wrote:
>>>>>> Hi, I have problems receiving packets using raw socket recently (after
>>>>>>  system upgrade). The packets are sent back based on tcpdump on the
>>>>>>  sender site, but no packets captured on planetlab's tcpdump. For
>>>>>>  example, I am pinging one plnode from my own machine, and I do see the
>>>>>>  reply:
>>>>>>
>>>>>>  64 bytes from planetlab-9.CS.Princeton.EDU (128.112.139.108):
>>>>>>  icmp_seq=1 ttl=53 time=32.1 ms
>>>>>>  64 bytes from planetlab-9.CS.Princeton.EDU (128.112.139.108):
>>>>>>  icmp_seq=2 ttl=53 time=32.6 ms
>>>>>>
>>>>>>
>>>>>>  but on planetlab node I received nothing from tcpdump:
>>>>>>
>>>>>>  @planetlab-9 ~]$ sudo /usr/sbin/tcpdump
>>>>>>  tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
>>>>>>  listening on eth0, link-type EN10MB (Ethernet), capture size 96 bytes
>>>>>>
>>>>>>  0 packets captured
>>>>>>  0 packets received by filter
>>>>>>  0 packets dropped by kernel
>>>>>>
>>>>>>  Can anyone help me on that?
>>>>>>  thanks
>>>>>>
>>>>>>  _______________________________________________
>>>>>>  Users mailing list: Users at lists.planet-lab.org
>>>>>>  https://lists.planet-lab.org/mailman/listinfo/users
>>>>>>
>>>>>
>>>>
>>>
>>
>