GRE implementation

Jean-Francois Dive jef at unixtech.be
Mon Oct 29 18:11:55 EST 2001


Hi, thanks for your answers.

I am quite interested to build new elements, this is for sure and my idea is to continue the ipsec effort. My main point about process communication is that the potential future ISAKMP will be a daemon running at process level. The communication trough IOCTL and /proc is definitively ok. Did someone brought the idea of creating a new socket type or an INET socket interface and a way to intercept other sysctl or ioctl providing click information instead of kernel stack information this giving compatibility with existing software ? I belive this is not a smal thing to do, but..

Starting things small and because it is anyway usefull, i am actually working on the implementation of GRE tunnels into click (if this is already done, tell me :) . From there, i have a design problem: 
I am/will write a GREencap and GREdecap elements as a start. Now, how to create virtual interfaces in click ? Would it be usefull (and i guess) to impact the linux interface list as FromLinux does (creating a new interface)? as we need an interface to points routes to or from. I dont see how to put the interface and the game.

What i think about would be the following:
Input:
FromInterface(eth0)->Classifier(Gre)->GREDecap()->virtualGreInterfaceGRE(gre0)->RoutingStack().

Ouput:
RoutingTable->virtualGreInterface(greXX)->GREencap()->RoutingStack().

Last question, is there any restriction to call methods from one element to another (i need for exemple information about virtualInterfaces source and destination addresses to check encap/decap of gre packets)

Thanks for any help in this.

Cheers,

JeF
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://amsterdam.lcs.mit.edu/pipermail/click/attachments/20011029/f1d9fbee/attachment.htm


More information about the click mailing list