We have a strange problem on an amd64 server running a fairly recent 6.1_STABLE.
The network interface stopped receiving after a disc controller timeout. The disc controller works fine after recovery. Yes: There was a timeout on mpt0, which was recovered (which essentially means entering the interrupt handler despite no interrut occured), and immediately after that, wm0 seems to have ceased to receive anything. It transmits fine (as far as you get without receiving). If I arp -a the MAC of another server, ping -n the other one from the broken one, I see ICMP echo requests on the broken one, requests and replies on the working one, but no replies on the broken one. Disabling ipf doesn't change anything. netstat -m shows ~260 mbufs allocated to data. It's a development machine, so I can leave it in this state for a while for further investigation. I guess there's some mess-up regarding interrupts. I would be glad to know a way out of this short of re-booting. It may happen on a production server in the future.
