Added this diff and built a new kernel last night and let the netmap application "lb" run all night, with tcpdump reading from two netmap pipes and everything appeared to run just fine. I have not been able to cause a panic with the e1000 network interface. There is some issue with the lb app that was not working correctly (it was not load balancing flows), so I will be running just tcpdump with "-i netmap:em0" overnight to verify a clean packet capture.