On Sun, Jul 17, 2022 at 06:10:22PM +0200, netdev@kapio-technology.com wrote:

Okay, I see the problem you refer to. I think that we have to accept some
limitations unless you think that just zeroing the specific port bit in the
DPV would be a better solution, and there wouldn't be caveats with that
besides having to do a FDB search etc to get the correct DPV if I am not too
mistaken.

No, honestly I believe that what we should do to improve the limitation
is to have proper ATU database separation between one VLAN-unaware
bridge and another (i.e. what is now MV88E6XXX_FID_BRIDGED to be one FID
per bridge).