|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
From: Jaakko Hyvatti <Jaakko.Hyvatti@www.fi> To: Lutz.Pressler@Unix.AMS.Medizin.Uni-Goettingen.DE (Lutz Pressler) Date: Tue, 23 May 1995 12:20:51 +0200 (EET DST) Cc: linux-net@vger.rutgers.edu In-Reply-To: <Pine.A32.3.91lp.950522184311.19421C-100000@amsu01.AMS.Medizin.Uni-Goettingen.DE> from "Lutz Pressler" at May 22, 95 07:12:50 pm > This "generic bridging over IP" would allow transport of multiple protocol > traffic (IPX, AppleTalk, DECnet/LAT/LAST, ...) over IP-only (serial) links. > Is there any reason for this beeing infeasible? Maybe I just had a black > out.. Or is there anything like this done already? I haven't found a > suitable RFC.. Hmm.. take a look at rfc1701 and 1702, they define Generic Routing Encapsulation (GRE) that is itended to tunnel any protocol over IP. I have quickly put together a minimal implementation (no checksum, no options) for some version of 1.3.0 netcode. It is in ftp://www.fi/pub/linux/diffs/kernel/linux-1.2.8.gre.diff but is untested against proper implementations like cisco router hardware. It was working between two linuxes :-). Now this is not exactly what you want, because with GRE you do not tunnel whole ethernet packets but individual protocols, like IP over GRE is defined in rfc1702. Fortunately, the payload packet type field in GRE header is coded with ethernet protocol numbers, so this might work: - strip ethernet frame - add 4 octets of GRE header, including ethernet protocol number from ethernet header, or 8 octets with checksum - add ip header Uh, but my ship leaves soon, gotta go travelling 'round Baltic countries. If you happen to be in Vilnius later this week.. se ya! :-) -- #Jaakko Hyvätti Jaakko.Hyvatti@WWW.FI http://www.fi/~jaakko/ +358 40 5011222 echo 'movl $36,%eax;int $128;movl $0,%ebx;movl $1,%eax;int $128'|as -o/bin/sync
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |