Hello experts,
We have FreeBSD vm on esx 6.0 (the base OS revision) with vmxnet3 as one of the interfaces on top of a vswitch. Jumbo MTU is not enabled on the vswitch. In our guest driver, we find eop field set indicating buffers active in jumbo ring with qid=8. We are running with 8 vRSS queues. We allocate buffers in jumbo ring even when jumbo MTU is not enabled. Could that be a problem?
Since backend is setting eop field incorrectly, we end up chaining packets incorrectly causing our protocol to go haywire.
Are there any known issues around this area?
We run a user space poll mode driver similar to DPDK.
Regards,
Nirmal