CCR1072 max ping size

Hi, I just found out that CCR1072 does not forward or respond to icmp packet larger than around 20000 bytes. This doesn’t happen in other platform.
I tested with this scenario:

[ainos1]—[CCR1072]—[ainos2]

ainos1 (ROS v6.42.3)
vlan599@ether1 10.222.222.2/29

CCR1072 (ROS v6.45.7)
vlan599@ether1 10.222.222.1/29
vlan598@ether1 10.222.222.9/29

ainos2 (ROS v6.43.4)
vlan598@ether1 10.222.222.10/29

Test from ainos 1
/ping 10.222.222.10 size=20000 → reply
/ping 10.222.222.10 size=30000 → timeout
/ping 10.222.222.1 size=20000 → reply
/ping 10.222.222.1 size=30000 → timeout

Test from ainos 2 also gives same result. ping with 30k size to or forwarded by CCR1072 is timeout.

Is this by design? Because Ainos still respond to ping size larger than 20k. CCR1016 also still respond & forward ping with > 20k size.

Out of curiosity: why would you require support for ICMP datagrams larger than e.g. twice the MTU size? And: do you have MTUs on individual L2 domains set to same falue? If you’re trying to use jumbo frames, then MTU should be set to same vakue both on CCR port and device (or multiple devices if there’s a switch involved) connected to that port …

If you want to test ability to fragment packets, anything larger than MTU size (and max jumbo MTU sizes are around 10kB) should be enough.
Are you sending ICMP packets with DF bit set? If you don’t then I don’t see how CCR could affect pings between both devices (in this case CCR only passes packets unless MTU sizes on both sides are different).

Not “require” per se, but on other platform OS, each respond to ping up to max size possible (64kb). Mikrotik on other platform I use also respond to ping with size over 20k.
MTU is 1500 on each interface (Ainos 1, CCR1072, Ainos 2) and 9000 on switch connecting this 3 device. I agree that CCR should only passes packets. So there should be no effect. That’s why I’m wondering why CCR1072 does not forward these icmp packets and not respond/reply to ping with bigger than 20k size