(2014/06/11 1:21), Stephen Hemminger wrote:

On Tue, 10 Jun 2014 16:05:07 +0900
Toshiaki Makita [off-list ref] wrote:

quoted

(2014/06/10 7:33), Vlad Yasevich wrote:
...

quoted

quoted

quoted

Rather than special casing this around vlan filtering, I would prefer
the code always forward these packets, or manipulate group_fwd_mask
to allow it that way.

These addresses must be forwarded only if the bridge is an S-VLAN
bridge. When it is a C-VLAN bridge or a .1D bridge, they may not be
forwarded. So, I don't think we can forward them always.

Using group_fwd_mask is a bit complicated. If we use it to forward them,
user can optionally turn off forwarding ability of those addresses...
but we maybe need another information (named like group_fwd_mask_set)
that indicates which bit is set by user. (We have to set group_fwd_mask
automatically when we set vlan_proto to 88a8.)
Is this way acceptable?

May be separate it into required mask and user mask.  Set required
mask when this is an S-VLAN bridge.

Sounds like a good idea.
I'll give it a try, thank you for your suggestion.

Thanks,
Toshiaki Makita

Looking again at the code.

1. If doing vlan then it should forward frames as defined in standard
   by default.

2. For compatiability, and for those users doing bump-on-wire, allow
   forwarding via group mask.

Thank you for reviewing. I sent v2 and I'm thinking it can satisfy these
two.

Thanks,
Toshiaki Makita