IP policy routing & rule 0

Hi,

I'm trying to set up a test where I use a single PC to send traffic across an external device (router, firewall, etc) and back to the same PC.

I am using two VLANs on a single interface, though it in principle should not be different from using two physical interface. The setup looks like this:

ip link add link eth0 name eth0.1 type vlan id 1
ip link add link eth0 name eth0.2 type vlan id 2

ip addr add dev eth0.1 10.3.0.2/24
ip link set dev eth0.1 up

ip addr add dev eth0.2 1.32.0.2/24
ip link set dev eth0.2 up

ip route del 10.32.0.0/24 dev eth0.1
ip route add 10.32.0.0/24 table 11 dev eth0.1
ip route add default table 11 via 10.32.0.1
ip rule add from 10.32.0.2 table 11

ip route del 1.32.0.0/24 dev eth0.2
ip route add 1.32.0.0/24 table 12 dev eth0.2
ip route add default table 12 via 1.32.0.1
ip rule add from 1.32.0.2 table 12

This works - mostly. I fail to force a packet from 10.3.0.2 to 1.32.0.2 using the ethernet port, instead it seems to go directly:

[~]>ping 1.32.0.1 -I 10.32.0.2 -c 1
PING 1.32.0.1 (1.32.0.1) from 10.32.0.2 : 56(84) bytes of data. 64 bytes from 1.32.0.1: icmp_seq=1 ttl=62 time=1.08 ms

vs.

[~]>ping 1.32.0.2 -I 10.32.0.2 -c 1
PING 1.32.0.2 (1.32.0.2) from 10.32.0.2 : 56(84) bytes of data. 64 bytes from 1.32.0.2: icmp_seq=1 ttl=64 time=0.043 ms

I have the feeling this is related to rule 0:
[~]>ip rule show
0: from all lookup local
32763: from 1.32.0.2 lookup 12
32764: from 10.32.0.2 lookup 11
32765: from 192.168.255.30 lookup 10
32766: from all lookup main
32767: from all lookup default

but deleting the route from the "local" table leads to a completely unusable interface.

Is there a way to achieve what I want?

Regards,
Christian

I think a patch that makes this possible was submitted a few days ago:

http://www.spinics.net/lists/netdev/msg114702.html