linuxboot
Experiences with NovaCustom dGPU variants
Is there anyone who has experiences with the NovaCustom NV56/NV54 with GPUs? Does the laptop boot at all?
I could see they are supported in Dasharo's coreboot, so I assume Heads can just be build with the Dasharo fork?
There is a discussion thread here: https://novacustom.com/forum/d/361-v54v56-nvidia-gpu-qubesos-certification-megathread/5. But not active for a few months.
But @wessel-novacustom mentioned here they have trouble to get GPU boards and can't focus on Heads support due to this, because it would be unreasonable to spent dev time on it.
If someone can confirm Qubes boots normally and can be used, I would be willing to buy one and be assigned as maintainer here.
I am glad I found this issue because a few months ago I was interested in the same thing. When I bought the GPU version of the laptop, I tried the non-GPU heads version of the firmware. The laptop booted well, but the GPU is not detected, and some PCI devices are in a different slot than Qubes expected because it saves PCI devices configuration based on the slot position.
I think if you use the right device tree and use the right GPU boot firmware with heads, it would work.
The laptop works perfectly with the normal non-heads GPU version of the firmware.
Thanks for the report!
You are probably interested into https://github.com/linuxboot/heads/issues/2004
I think the GPU is not detected because Coreboot doesn't run the OptionROM of it to init it. And neither do Heads or Qubes doesn't do either (if they even can. Not really my field). So the GPU doesn't show up because it is never initialized.
That said, if it would, it also would be a security concern. As discussed in the linked issue.
@wanistar Besides, you run Qubes I assume? Would you mind adding the Laptop to the HCL? This was the first place I looked at.
@Rot127 Yes, I use Qubes. I have submitted an HCL report.
I don’t think the GPU was detected because it wasn’t compiled into the heads firmware. I think if it is compiled, it will work perfectly. Then we can consider how to ensure that the GPU firmware is not tampered with.