Configuration applied:
-
Physical SFP+ port (
eth2) and VLAN sub-interface (eth1.1) both set to MTU 9014 — confirmed viaip link showon the router. -
Bridge device
br-lanset to MTU 9014 — confirmed viaifconfig. -
Connected device: Ubuntu 24.04 server, NIC (Mellanox ConnectX-4 Lx) set to matching MTU 9014, connected directly to the SFP+ port via DAC cable, no intermediate switch.
Symptom: With MTU 9014 set on both the router (port + bridge) and the server NIC, ping -M do -s 8972 (9000-byte packet) to the router's own br-lan IP (192.168.1.1) fails with 100% packet loss. Binary search shows the actual working ceiling is exactly standard 1500-byte MTU (ping -M do -s 1472 succeeds; -s 1500 and above fails), regardless of the 9014 value shown in interface status.
Reverting the server NIC back to MTU 1500 restores normal throughput immediately, confirming the router-side jumbo frame configuration is not actually being honored at the hardware/driver level despite being accepted and displayed in software (ip link / LuCI).
Reference: this matches a previously reported issue on a Flint 2 (openwrt/openwrt#21162) describing the same symptom — jumbo frames configured on br-lan and NIC causing connectivity failure, and the WAN port being unable to exceed 1500 MTU regardless of configuration.
Request: please confirm whether jumbo frame / MTU >1500 support is planned or already available in a newer firmware build for the Flint 4, or whether this is a known hardware/driver limitation of the current MediaTek switch driver in the stock GL.iNet firmware.