On Tue, 2003-10-21 at 01:53, Amit Limaye wrote: > Hello > Is it possible to run multiple TCP/IP stacks on the same machine on > different interfaces just as we can bind the TCP/IP stacks in windows > multiple times in to the same interface > > -SIGTERM > amit > > > Humanity's first sin was faith; its first virtue doubt > > - > : send the line "unsubscribe linux-net" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html Yes you can. In my earlier company, we ran 2 TCP stacks and 2 1/2 IP stacks while developing, and then shipped with 1 TCP stack and 1 1/2 IP stacks - Let me explain where the half comes from. TCP/IP Stack #1 - debug port runs TCP/IP with different socket definitions - TCP/IP - useful while developing, not present in final product. Needed cos platform measurement tools used this socket representation. TCP/IP Stack #2 - present in the product and used over the routing interfaces. ran routing protocols, SNMP software etc which used the 2nd socket representation. IP Stack #2 1/2 - highly fine-tuned microcode stack that dealt only with common case scenarios. Hence its not a full IP stack, only 1/2. Tip - while its easy to mix IP packets, its very difficult to mix TCP flows - that's my take on it. So at the TCP level, you WILL have a split personality machine if you run multiple stacks. Hope that cleared a few things out for you. To make things even cooler - after doing this, I took the packets at the rx and tx level of the PHY level, and packed them into RPC over TCP and sent it to a Verilog server that was running an RPC interface and that's how we verified the Verilog-based software-model of the hardware chip that we were designing - this takes a HECK of a long time - not for the faint hearted - but very useful when you're a startup spinning off a multi-million dollar ASIC. So in this case, we did 2 1/2 TCP/IP stacks out of which the 1 1/2 TCP/IP stacks were encapsulated within its OWN TCP/IP stack unless the packets were headed for the specified Verilog server. heady stuff. (I know this is TCP within TCP - but there's no other way - we cannot control that inner traffic (TCP/UDP etc), and the outer tunnel needs to be {reliable} TCP to instill RPC with some semblance of sanity). So, is it doable in reality ? Yes and it has its uses. Now, is it doable specifically in Linux ? I dont think so. However, Dave Miller can answer that way better than I can. The closest answer that says yes is "on interface 1 run ipv4 and disable ipv6. on interface 2 run ipv6 and disable ipv4". Now you have 2 different TCP/IP stacks running on two different interfaces. However its not the same VERSION of TCP/IP - that's probably not possible cos many knowledgable people think (for various reasons) that the TCP/IP stack in Linux should be hard-wired - I think otherwise bcos of reasons related to allowing innovation, or allowing flexibility during development, or even for research. HTH, -- Ranjeet Shetye Senior Software Engineer Zultys Technologies Ranjeet dot Shetye2 at Zultys dot com http://www.zultys.com/ The views, opinions, and judgements expressed in this message are solely those of the author. The message contents have not been reviewed or approved by Zultys. - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
