On Fri, May 20, 2016 at 05:11:23PM +0900, Simon Horman wrote: > On Fri, May 20, 2016 at 10:00:28AM +0200, Jiri Benc wrote: > > On Fri, 20 May 2016 14:29:01 +0900, Simon Horman wrote: > > > The second option does seem rather tempting although I'm not sure > > > that it actually plays out in the access-port scenario at this time. > > > > We support gre ports to be access ports currently. With conversion to > > ipgre, this needs to continue working. It's no problem for frames with > > the Ethernet header but now we have a situation where a port is tagged, > > thus the user expects that packets received on that port will behave > > accordingly. I don't think we can make some packets honor this and some > > ignore this; and we can't disallow gre to be an access port. > > > > How do you plan to solve this? By user space always pushing an ethernet > > header before push_vlan? > > Yes. That is my understanding of how OvS currently handles access ports but > I have a feeling that either I am mistaken or that you are referring to a > slightly different scenario.
Hi again. I apologise for having sent my previous email a little too quickly. My understanding is that currently OvS handles access ports using a push_vlan action. And that this patch set in conjunction with its user-space counterpart should ensure that a push_eth action occurs first. This is the context of my remarks above.
