Jumbo Frames makes LUCI/SSH unusable, rest works

Hallo together! I'm seeking help. I just set up 9000 MTU for eth0/eth1, bridges and supported virtual interfaces and also 2304 MTU for WLAN. Both archer c2600 are working fine in that configuration, well in meaning of networking (using both as AP and Mesh). Luci is practically unusable (i can barely log in on one while luci doesnt respond at all on the second) and ssh doesn't respond at all, while connected with lan. LUCI works while connected per WLAN.

I checked with ip -d l, the archer c2600 support sayed MTUs. Oh and i use diverse VLAN in that setting.

Any suggestions?


All attached Interfaces incl. Computers have to use the same mtu. I.e. the default br-lan: you can not set the bridge and Ethernet to 9000 but the WiFi on 2304 (or whatsoever), or you can not connect a PC and leaving it on 1500.

Wrong. Thats what Path MTU Discovery, other ICMPs and TCP-Handshake are used for. Smallest common MTU on a path is discovered and set automatically for each connection and you can still profit from bigger frames on a local network. Meanwhile you have MSS clamping on the nat. Just like i said network itself works fine. Just the 2 applications are being funny.
For example:


OK.... but now you are talking about layer 3 and not layer 2.... pmtu is used when you have routers on your path with smaller mtu. Your issue is before the first hop. Before the router does routing. Your issue sits on the stuff connected to the bridge with different mtu


There is L2 jumbo and L3 jumbo. Later CAN/should be "fixed" with PMTUD, but there is no failsafe mechanism for L2.

Hmm then i did misunderstand something. Sorry didnt want to sound mean. So i can only augment the mtu usefully on switches doing L3 and devices with nat/routing (mss) capabilities?

Edit: This case should be enough for me, since i mostly want jumbo frames between my pc (lan, 9000 MTU, nat), my L2 Switch (capable up to 16k MTU) and my nas (also 9000 MTU, nat). Rest is optional.

I guess there is no way turning the OpenWRT APs to L3, meaning enforcing traffic going through the "firewall"?

You could with a bunch of config overhead config each segment. Like one vlan for WiFi and one or more for Ethernet

1 Like

Ah i understand: then simply coupling the wlan vlan with the respective different ethernet/bridge vlan under the firewall as zones with accept/accept/accept? Well i can at least try it out and see if the c2600 cpu can handle it either with software offloading or with the current user build available with hardware offloading.

Edit: will try it out tommorow and give a feedback or either close the case. Thank you until then.