VALE + ptnet setup not working as expected

[sergio-gimenez]

Hi all!

I've been having issues lately with a singular setup with VMs running in the same physical host. I have the following testbed:

There are two QEMU/KVM VMs created with the netmap-enabled QEMU fork, directly connected to each other through a netmap pipe. That work properly, I can send traffic to ens5 interfaces, and it flows properly through the underlying netmap pipe. For example, I can do the following simple experiment attaching two pkt-gen instances in each ens5 and it works as expected.

However, if I try to replicate the same experiment but with a VALE switch, I cannot receive traffic to the other side:

I think it is important to note that, although I am not receiving traffic in the other side, the logs from dmesg look like if everything is going fine, so I cannot see any error anywhere that could point me to the actual issue.

Am I missing something? Is this a setup that is not intended to work like that?

[sergio-gimenez]

Any follow-up on this? This is still failing as per today. Any clues at least on why is this failing?