[ovs-dev] MPLS important comments (was: Re: Patch for MPLS)

[R.Kerur at telekom.com]

I don't have any NICS which support hardware offloading. I have tested both user and kernel space and looked at the outputs from wireshark and tcpdump to make sure relevant actions for the flows are exercised and relevant mpls shim or vlan qinq headers are pushed into the packet.

-----Original Message-----
From: Jesse Gross [mailto:jesse at nicira.com] 
Sent: Thursday, March 22, 2012 5:02 PM
To: Kerur, Ravi
Cc: blp at nicira.com; dev at openvswitch.org
Subject: Re: [ovs-dev] MPLS important comments (was: Re: Patch for MPLS)

On Thu, Mar 22, 2012 at 2:08 PM,  <R.Kerur at telekom.com> wrote:
> Hi Jesse,
>
> Please note that I work in labs environment and I volunteer my time for this and I have very limited resources to test. I have done testing related to complete mpls actions via the test-program written which is in tests/test-mpls.c. I have done basic vlan/mpls integration testing as well using again tests/test-mpls.c. I have made sure "make check" and "make distcheck" matches the results of the code in git master. I have used existing tools(tcpdump, wireshark, netstat and others) to make sure packets going out of the interface are correctly formatted such that actions executed correctly. I have run valgrind to make sure there are no memleaks, use-after-free errors in the code added.
>
> Currently there are issues related to automated tests w.r.t new actions that are added and being actively discussed with Ben.
>
> If you have a testbed on which you would like me to do the testing let me know I will be happy to do it. As it stands, I do not have extensive testbed setup to do any further testing nor will I be approved to have one since it requires us to purchase new equipments.

Sorry, I didn't mean to imply that I was looking for some kind of
elaborate test setup (I don't have anything specialized myself).  I
was just thinking of something basic like two connected machines
running OVS each with a bridge port and an Ethernet port.  You could
then add flows to push/pop tags and run netperf.

I asked this specifically because it appears to me that your patches
don't handle hardware offloading and packets attempting to use it will
not pass if the NIC doesn't support it.  The userspace unit tests
don't exercise the kernel code, so you'll have to test it separately.