问题原文
原帖链接 · 64bit_twitchyliquid · CC BY-SA 3.0
I am trying to conceptualise how networking works under the hood for linux VPNs which use TUN interfaces.
My current best guess is as follows (please correct me):
- Connection established to remote client/server.
- TUN interface created and brought up
- Routing table updated to set default gateway to the TUN interface
But wouldn't the packets destined for the remote client/server end up going into the TUN interface and forming a loop of sorts? How do VPN systems solve this? What is the gap in my understanding?
已采纳解答
原帖链接 · Jeremy Gibbons · CC BY-SA 3.0
You are correct that with purely destination-based routing this is a problem, if the destinations you are reaching through the tunnel overlap with routing needed for tunnel establishment, etc....
The way I have usually seen this done, and done it myself on various routers, is to use policy routing :
- The router acting as a VPN endpoint keeps its default route pointing to the Internet through its ISP link
- It also has a Policy Route with a source-based rule saying that traffic coming from the subnets behind it, whatever their destination, should be sent through the tunnel.
摘自公开归档,保留原文语言,版式经整理,外部媒体与邮箱地址已移除。
