25.12: VLAN filtering not working

[hatcreek68]

Is there an existing issue for this?

• I have searched among all existing issues (including closed issues)

screenshots or captures

Same behavior occurring on Flint 2 and Cudy WR3000.

Related to these other bugs but not sure if actually a duplicate (24.10 vs 25.12?)
#7196
#7278

Actual behaviour

VLAN filtering checkbox never writes: option vlan_filtering '1'

Using LUCI breaks vlans operating over wifi mesh. Setting the option with CLI resolves.

Expected behaviour

VLAN filtering checkbox should write: option vlan_filtering '1' or '0' when box is checked, not checked.

Steps to reproduce

• Check/Un-Check the box as shown above.
• save and apply
• check /etc/config/network and observe missing vlan filter option

Additional Information

Hostname	Flint2
Model	GL.iNet GL-MT6000
Architecture	ARMv8 Processor rev 4
Target Platform	mediatek/filogic
Firmware Version	OpenWrt 25.12.4 r32933-4ccb782af7 / LuCI (HEAD detached at e9ebca7) branch 26.133.20346~e9ebca7
Kernel Version	6.12.87
Local Time	May 19, 2026, 8:21:47 AM MDT


Hostname	WR3000
Model	Cudy WR3000 v1
Architecture	ARMv8 Processor rev 4
Target Platform	mediatek/filogic
Firmware Version	OpenWrt 25.12.4 r32933-4ccb782af7 / LuCI openwrt-25.12 branch 26.134.75701~cd18b8e
Kernel Version	6.12.87
Local Time	May 19, 2026, 8:22:02 AM MDT

What browsers do you see the problem on?

Firefox

Relevant log output

[hatcreek68]

I do see that in my 24.10 backups, I did NOT require option vlan_filtering '1' to handle vlans over mesh. I don't know if there rest of my setup is correct, but I learned setting option vlan_filtering '1' fixed my broken mesh while updating to 25.12 yesterday.

Gemini summarizes the rational here, rightly or wrongly:

The reason it worked without manual intervention in 24.10 but fails now in 25.12 is due to a specific change in how OpenWrt handles wireless VLANs in its configuration scripts (specifically the ucode migration).
The Change: vlan_no_bridge

In 24.10, the system scripts automatically included a parameter called vlan_no_bridge=1 for your wireless interfaces. This prevented the wireless daemon (hostapd) from trying to create its own internal bridges for every VLAN, allowing your main bridge (br-lan) to handle the filtering smoothly.

In 25.12, this default was removed or changed.

    The Result: Your mesh interfaces are now likely trying to create "isolated" internal bridges for their VLAN traffic.

    The Conflict: Because these hidden mini-bridges exist, the main bridge's "implicit" VLAN filtering logic (the part where it assumes filtering is on) gets confused or overridden.

    Why the Manual Option Fixes It: By explicitly setting option vlan_filtering '1', you are forcing the kernel to prioritize your bridge's filtering rules over the newly introduced hostapd bridge defaults.

[systemcrash]

@hauke @nbd168 could you take a look at this please? Regards the ucode migration for wifi scripts.