nvmm: Linux Kernel panic in QEMU+nvmm (in i7-8650 only?)

[Mosh <moshmosh%elektrine.com@localhost>]

Hi NetBSD!

I was having a very weird behavior while trying to run Linux with
NVMM+QEMU. This may not sound rare to you at this point, but I found
something so peculiar that may be a unexpected behaviour.

I do have a QCOW2 disk-file with Artix Linux installed that I usually
run using the following command:

    qemu-system-x86_64 -accel nvmm -m 4g -drive
    file=disk.qcow2,if=virtio

This works perfectly in my Intel Core i5-3470 PC. I still have to
supply the "noapic" param to the GRUB2 Linux cmdline, but beyond that
works as expected. However, in my Lenovo Thinkpad (Intel Core i7-8650)
*this doesn't work* for some reason. Once I run the command above,
Linux immediately panics with the following message:

    No working init found.  Try passing init= option to kernel. See
    Linux Documentation/admin-guide/init ...

However this is not true, because I've checked (both the initrd and the
disk partition, and even through the GRUB2 cmdline) and "/sbin/init"
seems to exist. I tried setting the "init=" option to "/sbin/init", and
it panicked with: "Request init /sbin/init failed (error -1)". Tried
setting something else like "/bin/ls" and it threw the same last error
again.

This truly strange considering that my Thinkpad have (in theory) a
better processor. And yes the VT-d is enabled. And I also remember
running NVMM+QEMU smoothly back then using NetBSD 10.1 in this same
Thinkpad.

I've looked at the dmesg logs and I found this very peculiar message:

    # dmesg | grep -iB1 vmx
    [  5216.585448] WARNING: module error: Incompatible module class 1
    for `nvmm' (wanted 3)
    [  5234.395448] nvmm0: attached, using backend x86-vmx

So I turned off my Thinkpad and retired the NVME disk, and inserted it
back again to my Intel Core i5-3470 PC. And after booting the same
NetBSD installation and looking at the dmesg logs, I didn't found the
previous WARNING message, and NVMM+QEMU was fully able to run the
virtual machine without any problems.

So, same NetBSD, same architecture, but different boxes. What could it
be? Why that WARNING message appeared? May this be a bug? Can someone
please help me? :(

Thanks! - Mosh

PS: I did tried running QEMU with -accel tcg, and while it ran
(predictably) slow af, it did work without problems. So it's definitely
a NVMM thing. Also, I did tried changing the "if" parameter and
changing "-drive" with "-hda" as well. Running it through a Devuan
Linux Live with QEMU+KVM also works