Antergos kvm_amd not active threadripper

Hi,

on my Antergos installation i can not use kvm for some reason.

lsmod | grep kvm
kvm                   729088  0
irqbypass              16384  1 kvm

It seems like kvm_amd is not active. SVM is activated in the BIOS, HyperV under Windows works. I have the newest available BIOS installed. Executing

sudo modprobe kvm_amd

never returns. Blinking cursor is all I get.

My hardware:

TR 1950x
Gigabyte Designare EX X399
2x 16GB DDR4 @2666
Vega 56 8GB (primary)
RX 560 4GB

Does anyone have an idea what the problem could be?

modprobe loads kernel modules, shouldn’t output anything if it completes without errors.

run

cat /proc/cpuinfo | grep vmx svm

if you get output then you’re good to go, and if you’re getting errors they aren’t related to the hw virt support

cat /proc/cpuinfo | grep vmx svm
grep: svm: No such file or directory

If I run the command without vmx (just svm) i get a giant wall of text. With just vmx nothing happens.

running
sudo modprobe kvm_amd
hangs the console. I have to ctrl+c to use the console again. I waited for 15mins once and still nothing. I dont think that is how it is supposed to be, right?

grep: svm: No such file or directory

shoulda been more clear, run the grep keywords one at a time (if you get any output on either the extensions are present)

or barring that, upon manual inspection: is vmx or svm listed in the flag list?

Yes, svm is listed.

okey doke, you should be fine then.

Should be, yes ^^. Virtualization does work on this system, just not using linux. If I start VirtualMachineManager (the gui) it sas

Warning: KVM is not available. This may mean the KVM package is not installed, or the KVM kernel modules are not loaded. Your virtual machines may perform poorly.

that sounds like it could be a permissions error. IIRC arch bakes in those modules by default

you should probably be using libvirt/virtman instead of a raw qemu frontend, too.

check /etc/libvirt/qemu.conf to see that the group for it is set to “kvm” instead of some random number (systemd fucks this up sometimes)

also you should ask around on this discord (passthrough POST):

Had any luck? I updated my BIOS today and have the exact same issue. Same Antergos setup, same behaviour.

Oh wow I guess the BIOS must be the problem then. I could not find anything useful yet. It should work just fine, all settings are correct. And somehow it does work using Windows and HyperV. I will try to downgrade the bios. Maybe that fixes the issue.

Compiling the kernel with CONFIG_CRYPTO_DEV_SP_PSP=n seems to fix the kvm issue with the latest bios updates on threadripper.

2 Likes

Thank you for the answer. Compiling custom kernels is currently a bit above my skill-level with arch/antergos.

AMD is responsible for this bug, correct? So I would have to wait for a new BIOS-Update if I want this fixed “the easy way” ^^.

there seems to be a problem with this boards latest update, we’ve seen several other reports too

cool, thanks for the info

I will try this when I get home tonight. Thanks!

But yeah, I’ve seen other issues with this BIOS. One’s a definite issue - Linux will take ages to shutdown or reboot.

I feel like the fan/pump controller behaviour has changed with this update, too.

I managed to compile a custom kernel now and can confirm this fix works. amd_kvm is back.

I also have those shutdown-issues. I have to reset the mainboard if I dont want to wait for 20mins. It says something about “watchdog did not stop” most of the time, or “a stop job is running”.

Or use cat /proc/cpuinfo | grep -e vmx -e svm

Same on Gigabyte Designare EX399 with f10 bios when running Ubuntu but I’m not that patient.
Its very frustrating.