On Mon, Dec 28, 2009 at 01:55:33PM -0800, Jay Vosburgh wrote:

David Miller [off-list ref] wrote:

quoted

From: Andy Gospodarek <andy@greyhouse.net>
Date: Mon, 28 Dec 2009 10:26:32 -0500

quoted

This patch should be ready for inclusion.

Yes, but I haven't seen ACK's from Jay or Eric.

	I was testing with the final patch just before my personal
holiday adventures began, and was having one problem: the ARP monitor
part worked fine on the VLAN (the slaves were up) but regular traffic on
that same VLAN, e.g., ping, was dropped on receive at the bond.  The
peer (without bonding) saw the traffic, and responded, but it appeared
that ARP replies from the peer (again, over the VLAN) weren't making it
to the ARP table (at the bond end), so everything that depended on the
ARP table wasn't working.

	I'm willing to believe this is a personal problem at my end, but
I haven't determined the cause, and won't have the opportunity until
next week (after my house has drained of relatives).

	If Andy isn't seeing the above misbehavior, I'm fine with
assuming it's my problem and applying the patch; if I find something
actually wrong with the patch next week we can address it then.

From time to time, I would see something similar would be if I was

testing without VLANs and ended up with a stale entry in the remote
hosts ARP table that would cause the response traffic to get sent
without a VLAN tag.

I was able to pass tagged traffic when the bond was up and always made
sure to scp my tested patches over that link to the machine that would
post them.  It seems like good practice to exercise the changes I'm
making during the posting process. :)