On Tue, May 22, 2012 at 11:21:29PM -0700, Nicholas Bastin wrote:
> On Tue, May 22, 2012 at 2:50 AM, Prabina Pattnaik
> <[email protected]> wrote:
> > We agree that implementation of 'at least 64 bytes' is as per spec (to
> > allow data above 64 bytes depends on vendor specific switch
> > implementation).
> > But in current implementation, when the code (mentioned below) limits the
> > size of data to 64 bytes (if data field of the error packet
> > is more than 64 bytes), a malformed OFPFMFC_BAD_COMMAND error packets gets
> > generated (visible on wireshark with openflow
> > plugin).
>
> The wireshark plugin is (questionably) broken. Or wireshark is
> broken, you pick which you prefer.. :-) The issue is that the
> OpenFlow dissector (arguably properly…certainly I would argue that)
> passes encap'd packet-ins *back* into wireshark for further
> dissection. However, there's no way to tell wireshark "please ignore
> the fact that this packet may be truncated" - something it is capable
> of doing for outer-layer packets, so you end up with giant red
> nonsense that the packet is malformed, and not that you may have a
> truncated packet. Either way, the OVS interpretation may not be what
> you want, but it's not broken.
I thought that OVS itself did a better job of printing these truncated
OpenFlow messages, but when I tested it I found that I was wrong, so I
sent out a patch to improve the situation:
http://openvswitch.org/pipermail/dev/2012-May/017462.html
_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss
