Untagged: Frames will leave the bridge through the specified interface as untagged. Untagged ingress will be assigned to the specified VLAN ID unless PVID is set to a different VLAN ID on the interface. Tagged ingress is discarded.
@fengchen-lede this. As @brada4 said.
The frame is sent and hits the other interface. And it needs to processed at least a little bit so it can be decided if further processing is required or if it should be dropped.
I suspect the document is wrong. If a port is in any way a member of VLAN 11 and a packet tagged with 11 arrives, it is going to be accepted and put through the switch matrix with a tag of 11. The tagged and untagged setting affects departing packets only. The PVID (star) designation means that untagged packets arriving will be assigned to that VLAN and given that internal tag for the switch matrix. (This should of course be set on only one VLAN.)
Inside the switch there is no such thing as an "untagged VLAN." All packets even if they arrived and will depart on untagged ports, have to carry a tag inside so that they can be switched to the proper destination.
I'm not sure what you're trying to do here. In general, avoid mixing tagged and untagged packets on the same port and cable. Do that only if it is necessary to connect to someone else's network which works that way and you can't change it. If you're designing both ends of the network, use either "trunk" connections (all relevant VLANs are tagged, for interconnects between VLAN-aware switches and routers) or "access" connections (no tagged packets, untagged only in one VLAN for a non VLAN-aware device).
I tested almost all cases and found that tagged traffic can enter the matching VLAN and untagged VLAN. In addition, during the test, traffic cannot enter the unmatched tagged VLAN with pvid, so there is no untagged traffic. The conclusion is that there may be a bug in the software or driver.
Hostname OpenWrt
Model YouHua WR1200JS
Architecture MediaTek MT7621 ver:1 eco:3
Target Platform ramips/mt7621
Firmware Version OpenWrt 23.05.5 r24106-10cc5fcd00 / LuCI openwrt-23.05 branch git-24.264.56413-c7a3562
Kernel Version 5.15.167
Not an expert on this, the x-wrt tree has some stuff not in the main tree. Hail mary but maybe applying some of those selectively would do something.
It's probably like mk24 says though:
If a port is in any way a member of VLAN 11 and a packet tagged with 11 arrives, it is going to be accepted and put through the switch matrix with a tag of 11.