[Click] click + gre tunnel

[Eddie Kohler]

Hi Erik,

Looking at tcpdump's source code, it looks like the "gre-proto-0x0003" message 
happens when the packet inside the GRE header has a strange Ethernet type: 
0x0003, instead of 0x0800 (IP).  I don't know why this is happening.  But I 
believe it has to do with the raw socket interface that Click uses to send 
packets.  Would you mind looking into that?  Maybe there's some ioctl we should use.

Eddie


Erik VandeKieft wrote:
> Hi,
> 
> I still can't seem to successfully send packets to a gre tunnel using
> click.  Here's the tcpdump output of gre packets sent via the regular
> linux kernel, versus with click. To do the first test, I simply added a
> route to gradlab.ucsd.edu with device "farm" (the gre tunnel device), and
> pinged gradlab:
> 
> btw, (128.54.16.16 == other end of gre tunnel)
> 
> ----
> # tcpdump -vvv -i eth0 dst 128.54.16.16
> tcpdump: listening on eth0
> 18:07:14.769438 wamp.ucsd.edu > teleport.ucsd.edu: gre 10.0.1.1 > gradlab.ucsd.edu: icmp: echo request (DF) (ttl 64, id 0, len 84) (DF) (ttl 255, id 0, len 108)
> 18:07:15.782063 wamp.ucsd.edu > teleport.ucsd.edu: gre 10.0.1.1 > gradlab.ucsd.edu: icmp: echo request (DF) (ttl 64, id 0, len 84) (DF) (ttl 255, id 0, len 108)
> 18:07:16.782135 wamp.ucsd.edu > teleport.ucsd.edu: gre 10.0.1.1 > gradlab.ucsd.edu: icmp: echo request (DF) (ttl 64, id 0, len 84) (DF) (ttl 255, id 0, len 108)
> ---
> 
> Note the nice output that shows you what the encapsulated gre packet's
> source and dst are.
> 
> Now, here's the same tcpdump using the following click command:
> 
> # click -e 'ICMPPingSource(10.0.1.1, 132.239.55.107) -> ToDevice(farm)'
> 
> btw, (132.239.55.107 == gradlab.ucsd.edu)
> 
> ---
> 18:13:23.753185 wamp.ucsd.edu > teleport.ucsd.edu: gre gre-proto-0x0003 (DF) (ttl 255, id 0, len 52)
> 18:13:24.763224 wamp.ucsd.edu > teleport.ucsd.edu: gre gre-proto-0x0003 (DF) (ttl 255, id 0, len 52)
> 18:13:25.763296 wamp.ucsd.edu > teleport.ucsd.edu: gre gre-proto-0x0003 (DF) (ttl 255, id 0, len 52)
> ---
> 
> Note that the encapsulated gre packet seems different. Anyone have any
> clue as to what might be going on?
> 
> Thanks,
> --Erik Vandekieft
> 
> 
> On Thu, 8 Jul 2004, Eddie Kohler wrote:
> 
> 
>>Erik,
>>
>>Could you use 'tcpdump' to compare Click's packets with packets
>>generated by some other program?
>>
>>E
>>
>>On Jul 8, 2004, at 3:00 PM, Erik VandeKieft wrote:
>>
>>
>>>Hi, I'm running click in userland and trying to send and receive
>>>packets
>>>to a gre tunnel interface. The name of my tunnel interface is "farm". I
>>>created a simple FromDevice(farm) element, and that seems to recieve
>>>packets just fine. But the ToDevice(farm) element doesn't seem to work
>>>correctly... I can see the gre packets going out on eth0, but they get
>>>dropped somewhere at the other end of the tunnel-- could it be that
>>>they
>>>are formatted improperly?
>>>
>>>In short, is my usage expected to work? Do I have to do anything
>>>special
>>>to send IP packets to a gre tunnel?
>>>
>>>Thanks very much,
>>>--Erik Vandekieft
>>>
>>>
>>>_______________________________________________
>>>click mailing list
>>>click at amsterdam.lcs.mit.edu
>>>https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>>
>>