[strongSwan-dev] route a packet to a VTI device, it gets NoRoute error

[tszephay Field]

Hi, guys, sorry for my bother.

I have read the wiki page <Route-based VPNs> (
https://wiki.strongswan.org/projects/strongswan/wiki/RouteBasedVPN), and I
noticed this sentence:
* Whenever a packet is routed to a VTI device it automatically gets the
configured mark applied so it will match the policy and get tunneled.*
So I built a test case to confirm this feature. The topology looked like
this:


pc1 pc2 pc3

tunnel
eth0 ------------- eth1 - eth2 ------ eth3

vti1

I want to ping eth3 from eth0. The ipsec tunnel built between eth0 and
eth1. vti1 is of course set up on pc2 with mark.

I created route on pc2 like this:

ip rule add dev vti1 table 1
ip route add default dev eth2 table 1

ip rule add dev eth2 table 2
ip route add default dev vti1 table 2

Now, It worked well just like I thought except for only one thing:* I can
capture the Echo-Reply packet on vti1 from eth3*, but to my surprise, it
was dropped at once. If I understood that sentence correctly, the packet
should get the mark, match the policy and then get tunneled to eth0.

"ip -s tunnel" shows that there are some TX errors on vti1, and the reason
is *NoRoute**.*

I really do not know why the route is needed at this point if vti1 bind the
CHILD_SA.

The version is 5.4.0

Any suggestion is welcome.