On Sat, Oct 22, 2022 at 12:55:14PM +0000, Oleksandr Mazur wrote:

quoted

I hope the following script will exemplify what I mean.

..
Oh, i get it now.

Frankly speaking we haven't stumbled across such scenario / issue
before. But i can tell it does indeed seems a bit broken;

I think there are 2 options here:
  1. The setup itself seems insecure, and user should be aware of such behavior / issue;

Be aware, and do what? Port locking is unfit for use if learning is left
enabled (in the way learning is currently done).

  2. Bridge indeed should not learn MACs if BR_PORT_LOCKED is set.
  E.g. learning condition should be something like: not BR_PORT_locked
  and learning is on; 

Rather than violate the BR_LEARNING flag (have it set but do nothing,
which would require even more checks in the fast path), I was proposing
to not allow the BR_PORT_LOCKED | BR_LEARNING configuration at all.
My question to you was if you're aware of any regression in prestera
with such a change.

quoted

I don't understand the last step. Why is the BR_PORT_LOCKED flag disabled?
If disabled, the port will receive frames with any unknown MAC SA,
not just the authorized ones.

Sorry for the confusion. Basically, what i described what i would
expect from a daemon (e.g. daemon would disable LOCKED); So just
ignore that part.

But still, why would the daemon disable BR_PORT_LOCKED once a station is
authorized? You're describing a sample/test application, not a port
security solution...