After running pfSense as a transparent firewall on my VMware ESXi 5 whitebox lab setup for nearly a year, I decided to setup pfSense as my home internet router.
To do that I needed an additional NIC in my whitebox, so I bought a second Intel Pro/1000 GT.
After installing the NIC, VMware booted as normal and the NIC was immediately recognized. Only a few hours later, my whitebox crashed with a kernel panic.
The VMware ESXi 5 host reported a “Spin count exceeded” and “possible deadlock” in the purple diagnostic screen.
Without searching google for the error, I first upgraded my VMware ESXi 5 host to the latest patch level and gave it another shot.
Within a few hours, my whitebox crashed again with the same “Spin count exceeded” purple screen.
After some searching I found this thread, which seems to be very close to my problem.
The reason that the kernel panic occurs (randomly) seems to be related to an unsolved linux IRQ bug on Sandy Bridge mainboards, which is described here.
In short: According to the articles mentioned above my only option is to replace the PCI NIC with a PCI-X NIC, as this seems to be a working workaround.
I already disabled a couple of on-board device on my Asus P8H67-V mainboard (sound, Atheros AR8151 NIC), but since I’m not using the onboard USB 3.0 controller my idea was to try to free up an additional IRQ, so I disabled the USB 3.0 Controller as well.
Problem solved! pfSense has been running stable now for over a week and still going strong!
