I have a Dell R710 running Ubuntu 18.04.1 LTS as a KVM host. I have a Hauppauge WinTV quadHD tuner card installed in one of the PCI slots. I have a guest VM installed to act as a media center (same OS), to which I want to give direct access to the tuner.
I have virtualization extensions turned on in BIOS and IOMMU enabled. I've checked that the tuner seems to inhabit its own IOMMU group. In fact, the tuner does appear in the guest as you would expect, using this config in libvirt.
<hostdev mode='subsystem' type='pci' managed='yes'> <source> <address domain='0x0000' bus='0x0a' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <source> <address domain='0x0000' bus='0x09' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x08' function='0x0'/> </hostdev>
The problem is that the guest accessing the device (scanning or viewing channels) has repeatedly demonstrated an unfortunate ability to take down not only the guest, but the entire host. It's not always immediate... sometimes it works for minutes or even days before failing.
When things go south, there's no kernel panic or other log event that I've discovered.* One minute everything is running, the next the whole box is rebooting.
* The RAC log gives only this hopelessly vague error, each time this event occurs:
Wed Jul 25 2018 15:53:23 A bus fatal error was detected on a component at slot 3.
Googling around surfaces basic troubleshooting steps like reseating the card (done), updating firmware (infeasible), or changing slots (done). None of these have thus far made a difference.
I don't even know who, if anyone, this could be escalated too. Is it a kernel issue? Device driver? Hardware bug?
If anyone has experienced similar faults, is there a way I can get a trace of some sort that I could escalate to the appropriate developers?