[Planetlab-devel] tcp raw socket problem on new planetlab system
Y. Charlie Hu
ychu at ecn.purdue.edu
Sun Jan 18 14:05:36 EST 2009
I am very very sorry, I messed up my mailer. please ignore all my mails.
Y. Charlie Hu wrote:
> please add a bar for One group in fig 6 (D)
> it is ok that Ant outputs one group, but that does not mean you should
> the results for "One group". it just creates unnecessary confusion.
> Harsha V. Madhyastha wrote:
>> Hi Marc,
>> Yes, our tool works fine on a normal Debian box without the
>> modification that Sapan suggested. Also, the same tool worked fine on
>> PlanetLab from '06 until July'08. So, it does look like the
>> requirement to explicitly bind to a UDP port was introduced by VNET+
>> I will extract the code snippet and send it over to you guys.
>> On Jan 18, 2009, at 8:05 AM, Marc Fiuczynski wrote:
>>> Hi Harsha and Sapan,
>>> On Jan 17, 2009, at 11:17 PM, Harsha V. Madhyastha wrote:
>>>> Great! That worked. Thanks for the tip, Sapan.
>>> I am very glad this worked. However, Harsha mentioned that the same
>>> program worked without this particular additional binding of a UDP
>>> socket to the specific port back in '06. Moreover, I am curious
>>> whether this unmodified program "just works" on a vanilla kernel, but
>>> not on our kernel. If so, then I think we are not fully compatible
>>> with the way the old VNET module implemented source port auto-binding
>>> when users sent UDP packets over a raw socket.
>>> Harsha, can you validate that your program works without Sapan's
>>> suggested modification? And, if that works, please do extract the
>>> code snippet so we can play with it and improve VNET+ to have the
>>> proper semantics (if possible).
>>> Devel mailing list
>>> Devel at lists.planet-lab.org
>> Devel mailing list
>> Devel at lists.planet-lab.org
More information about the Devel