[mpls-linux-devel] RFC: Generic network shim infrastructure
Status: Beta
Brought to you by:
jleu
From: James R. L. <jl...@mi...> - 2005-08-31 04:52:32
|
In hopes to get this MPLS effort 'unstuck from the mud' I've decided to start the path of getting some of the infrastructure needed for MPLS reviewed by the powers that be. I hope DaveM and Jamal are still reading this list, because they are probably the first hurdle I need to cross. This patch introduces the network shim infrastructure. It is designed for protocols that live below layer 3 but above layer 2 in the OSI model. This is exactly where MPLS lives. I believe this infrastructure could also be used by IPSEC for route base VPNs without using tunnels. The guts of the patch are the 'shim live cycle', but it also adds a new 'family' to netlink, RTM_*SHIM, and adds some the necessary hooks to ipv4 and ipv6. <aside> I have tried using the XFRM infrastructure, but the problem I continue to run into is it's apparent lack of ties to routing entries. For example if I want traffic destined to 10.0.0.0/8 to go through a specific XFRM I can do that, but if at the same time I don't want 10.0.1.0/24 to use any XFRM, this is where I am unable to get XFRM to do my bidding. I tried adding per-route XFRMs but it looks like the dst stacking is in the wrong order for what I need .... I digress </aside> Please look this over and let me know if there are any glaring kernel no-no's. I still have commenting and documentation to do, but the actual infrastructure is complete. Of course I've implemented MPLS to use this, but we'll tackle that after we get the needed infrastructure into the kernel. Patch attached -- James R. Leu jl...@mi... |