I'd either use bare metal or maybe a container (LXC, not sure sure about docker) as that's as near as bare metal with few encumbrances.

Why ? I've always used bare metal, and it seems VDR is working quite close to the hardware, adding layers adds complexity.

But the other thing is that that sever looks quite ancient now and a replacement (somewhat newer, if not new) shouldn't cost much. I had a similar issue with a CCTV card unrecognised by the kernel and gave up fighting the FOXCONN mobo/BIOS, after some discussion and help on the kernel developers forum who tried but ultimately couldn't help in the end, and bought an-off-the shelf HP machine for only around 100 euro / used.   There were other benefits, like super-efficient PSU giving power eduction which in the 4 or 5 years since may have paid for it !  The ugly bit is ideally you need a kind soul to let you test your card in whatever machine first, to be sure it works.

On 13/11/2020 12:00 pm, vdr-request@linuxtv.org wrote:

potential solutions:

• run VDR in a docker container instead of in a VM and pass /dev/dvb to the container
• replace the KVM host, thus get rid of the RMRR issue and continue to use PCI passthrough
• run the VDR on bare metal
• others?

What would you do if you were in my place, and why?