Configured routing tables not used by virtual interfaces (vlan)

I need PC1 to be able to communicate with PC2 through two vlan (configured on two virtual interface). Each of them having a gateway configured to communicate with a different subnet.

Here is the diagram:

     ~ ~ ~ ~<vlan10>~ ~ ~ ~
192.168.1.1         192.168.1.254
PC1 ----------------------------- FW ----------------------------- PC2
192.168.2.1         192.168.2.254    11.20.1.254                11.20.1.1
     ~ ~ ~ ~<vlan20>~ ~ ~ ~

PC1:
- eth0.10 = 192.168.1.1/24
- eth0.20 = 192.168.2.1/24

Firewall:
- allow all

I added two routing table, one for each virtual interface:

# ip route list table vlan10
default via 192.168.1.254 dev eth0.10 proto static 

# ip route list table vlan20
default via 192.168.2.254 dev eth0.20 proto static 

# ip route
192.168.1.0/24 dev eth0.10 proto kernel scope link src 192.168.1.1 
192.168.2.0/24 dev eth0.20 proto kernel scope link src 192.168.2.1 

This is how the routing table are applied:

# ip rule
0:  from all lookup local
32762:  from all to 192.168.2.1 lookup vlan20
32763:  from 192.168.2.1 lookup vlan20
32764:  from all to 192.168.1.1 lookup vlan10
32765:  from 192.168.1.1 lookup vlan10
32766:  from all lookup main
32767:  from all lookup default

I can ping the firewall interface on the same subnet (the gateway):

# ping 192.168.1.254
PING 192.168.1.254 (192.168.1.254) 56(84) bytes of data.
64 bytes from 192.168.1.254: icmp_seq=1 ttl=64 time=1.09 ms
64 bytes from 192.168.1.254: icmp_seq=2 ttl=64 time=0.740 ms
64 bytes from 192.168.1.254: icmp_seq=3 ttl=64 time=0.319 ms

# ping 192.168.2.254
PING 192.168.2.254 (192.168.2.254) 56(84) bytes of data.
64 bytes from 192.168.2.254: icmp_seq=1 ttl=64 time=0.937 ms
64 bytes from 192.168.2.254: icmp_seq=2 ttl=64 time=0.269 ms
64 bytes from 192.168.2.254: icmp_seq=3 ttl=64 time=0.444 ms

I can't ping the firewall interface on the other subnet:

# ping 11.20.1.1
connect: Destination host unreachable

# ping 11.20.1.1 -I eth0.10
PING 11.20.1.1 (11.20.1.254) from 192.168.1.1 eth0.10: 56(84) bytes of data.
^C
--- 11.20.1.1 ping statistics ---
4 packets transmitted, 0 received, 100% packet loss, time 75ms

When I use wireshark, I see an ARP request destinated to 11.20.1.1, even if it's on another subnet and it should use the default gateway from the routing table.

When I configure the default gateway in the main routing table, the ARP request is destinated to the gateway and the ping work:

# ip route add default via 192.168.1.254

# ping 11.20.1.1 -I eth0.10
PING 11.20.1.1 (11.20.1.1) from 192.168.1.1 eth0.10: 56(84) bytes of data.
64 bytes from 11.20.1.1: icmp_seq=1 ttl=64 time=0.660 ms
64 bytes from 11.20.1.1: icmp_seq=2 ttl=64 time=0.504 ms
^C
--- 11.20.1.1 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 9ms
rtt min/avg/max/mdev = 0.504/0.582/0.660/0.078 ms

Do you know where the problem come from ? rp_filteris already set to 0 on the physical and virtual interfaces.