Re: [PATCH net-next 4/7] openvswitch: Reset upper layer protocol info on internal devices.

* Jesse Gross (jesse-l0M0P4e3n4LQT0dZR+AlfA@public.gmane.org) wrote:

On Mon, Sep 3, 2012 at 6:00 PM, David Miller [off-list ref] wrote:

quoted

From: Jesse Gross <redacted>
Date: Mon, 3 Sep 2012 17:57:39 -0700

quoted

On Fri, Jul 20, 2012 at 3:26 PM, Jesse Gross [off-list ref] wrote:

quoted

It's possible that packets that are sent on internal devices (from
the OVS perspective) have already traversed the local IP stack.
After they go through the internal device, they will again travel
through the IP stack which may get confused by the presence of
existing information in the skb. The problem can be observed
when switching between namespaces. This clears out that information
to avoid problems but deliberately leaves other metadata alone.
This is to provide maximum flexibility in chaining together OVS
and other Linux components.

Signed-off-by: Jesse Gross <redacted>

It was recently discovered that the bug that this patch fixes is
causing problems in the real world.  Can you please queue this for
stable in 3.4/3.5?  It's currently in Linus's tree as
7fe99e2d434eafeac0c57b279a77e5de39212636.

What vendor is shipping openvswitch enabled and requires the fix to
be in -stable before they'll ship it to customers?

That goes into what is 'real world'

Fedora is running into it I believe.  Chris Wright asked for it so he
might be able to elaborate more on their plans.

I've not hit the bug myself, but been made aware of the issue from
OpenStack/Quantum folks.

There's a testing scenario that hits this as described in this launchpad
bug:

  https://bugs.launchpad.net/quantum/+bug/1044318

thanks,
-chris