[Planetlab-users] Re: tuntap on 4.2

[Sapan Bhatia]

Hi Cristian and Shin-ae Woo,

The way tun/tap works on PL is that there's one device (tap0) that is
virtualized on a per-slice basis. So you don't need to call the
standard interface-creation code you use on Linux to start things up.
The tap0 interface is instantiated and always there, and you can use
it by reading from and writing to the file /dev/net/tun and by using
the tap0 interface.

Does this address your doubt?

Sapan

On Mon, Jul 7, 2008 at 6:32 PM, Cristian Lumezanu <lume at cs.umd.edu> wrote:
> Sapan,
>
> I know you must be busy with other things besides the tun/tap on PL so
> please let me know if there is anything else I can do to make your work
> easier with the tun/tap.
>
> Cristian
>
> Sapan Bhatia wrote:
>>
>> Christian,
>>
>> Have you previously tried your code out on PlanetLab? Could you try
>> using this box: planet2.scs.nyu.edu? You'll have to ssh in via port
>> 806.
>>
>> The first thing I'd like to do is to make sure that the migration has
>> been clean and that we haven't lost any functionality because of the
>> changes to the VNET module. The second thing will be to figure out if
>> things work as they do on standalone Linux boxes.
>>
>> I'd appreciate it if you'd try this out. If it's the case that it's a
>> generic PL (not 4.2) problem, then we'll roll out an initial version
>> and debug your problem in that version.
>>
>> Sapan
>>
>>
>> On Thu, Jul 3, 2008 at 3:52 PM, Cristian Lumezanu <lume at cs.umd.edu> wrote:
>>
>>>
>>> Sapan
>>>
>>> I am using only the tun interface and I see some weird behavior:
>>>
>>> I open the tun device using open("/dev/net/tun",...) and then I use
>>> ioctl(TUNSETIFF,...) to set up some parameters. The tun interface that is
>>> associated with my tun device is "tap0" not "tun0" as it should be. When
>>> I
>>> repeat the same code (hoping that I would get a different tun interface),
>>> ioctl breaks with the error "Device or resource busy - /dev/net/tun
>>> (Errno::EBUSY)). The exact same code works without problems on each of my
>>> Linux machines I have set up in our lab. Is there anything different
>>> about
>>> PL that I am missing? Thanks.
>>>
>>> Cristian
>>>
>>> Sapan Bhatia wrote:
>>>
>>>>
>>>> Great! Thanks for responding quickly to our call for testing. Could
>>>> the others who had asked for tuntap (cc'd) also verify that it's
>>>> satisfactory, so that we can package it and roll it out? -- Sapan
>>>>
>>>> On 7/3/08, Andreas Haeberlen <ahae at mpi-sws.mpg.de> wrote:
>>>>
>>>>
>>>>>
>>>>> Hello Sapan,
>>>>>
>>>>>  thank you for the quick response. I tested our software on the node
>>>>>  you set up for us, and it looks like everything is working fine.
>>>>>  The only unusual thing I saw was that all outgoing packets are counted
>>>>>  as dropped by the kernel:
>>>>>
>>>>>  tap0      Link encap:Ethernet  HWaddr 00:FF:0A:28:6C:01
>>>>>           inet addr:10.40.108.1  P-t-P:10.40.108.1
>>>>>  Bcast:10.255.255.255
>>>>> Mask:255.0.0.0
>>>>>           UP BROADCAST POINTOPOINT RUNNING NOARP  MTU:1468  Metric:1
>>>>>           RX packets:15108 errors:0 dropped:0 overruns:0 frame:0
>>>>>           TX packets:13935 errors:0 dropped:13935 overruns:0 carrier:0
>>>>>           collisions:0 txqueuelen:1000
>>>>>           RX bytes:20480346 (19.5 MiB)  TX bytes:14285330 (13.6 MiB)
>>>>>
>>>>>  However, since the packets seem to be delivered to the application
>>>>> anyway,
>>>>>  this is more of a cosmetic problem.
>>>>>
>>>>>  Thank you,
>>>>>
>>>>>       - Andreas
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> I've installed a first attempt at resurecting tuntap on
>>>>>> planetlab-7.cs.princeton.edu. Please test the functionality on this
>>>>>> node at your convenience. If you run into problems (quite likely),
>>>>>> then it would help if you could join us on the PlanetLab IRC channel
>>>>>> at irc.freenode.net#planetlab so that we can iterate over it and get
>>>>>> it working as it should.
>>>>>>
>>>>>> Once we are satisfied that the feature is running fine on this alpha
>>>>>> node, we can deploy it throughout PlanetLab.
>>>>>>
>>>>>> Sapan
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>
>>>
>
>