Skip to content

[Bug]: vboxdrv triggers kvm_amd probe on Intel systems #629

Open

Description

@thisisamirv

Version

7.2.6

Host OS Type

Linux

Host OS name + version

Arch linux 6.19.6-arch1-1

Host Architecture

x86

Guest OS Type

Linux

Guest Architecture

x86

Guest OS name + version

No response

Component

Other

What happened?

On an Intel-based system, the kvm_amd kernel module is being probed during boot, resulting in the message:

kvm_amd: CPU 1 isn't AMD or Hygon

This occurs even though:

The system uses an Intel CPU
kvm_intel loads correctly
No AMD-related virtualization should be involved

The message appears immediately before vboxdrv initialization and VirtualBox reports finding KVM symbols:

kvm_amd: CPU 1 isn't AMD or Hygon
vboxdrv: module uses symbol (...) from namespace module:kvm,kvm-amd,kvm-intel
vboxdrv: Found KVM hardware-virtualization symbols

  • This behavior was initially reported upstream to the Linux kernel, where it was suggested that the trigger likely comes from VirtualBox (vboxdrv) interacting with KVM symbols.
  • The issue is harmless but creates confusing and unnecessary log noise.
  • Blacklisting kvm_amd prevents the message, suggesting it is not required for VirtualBox operation on Intel systems.

How can we reproduce this?

1- Install VirtualBox on an Intel-based Linux system
2- Ensure KVM is available (kvm_intel loads normally)
3- Reboot the system
4- Observe dmesg output during early boot

Did you upload all of your necessary log files, screenshots, etc.?

  • Yes, I've uploaded all pertinent files to this issue.

Activity

  1. cybersnow02 commented on Apr 16, 2026

    @cybersnow02

    I also seen this in my VBox.log when looking at something else. Maybe it is trying to distinguish if this is an AMD or Intel CPU about CPU capabilities. Then, it throws this error but it might actually not preventing VBox to work as usual. Maybe just avoid reporting this error in the code?

  2. Doaxan commented on Apr 29, 2026

    @Doaxan

    Same here
    29.04.2026 17:13 kvm_amd CPU 8 isn't AMD or Hygon

  3. tachi commented on May 29, 2026

    @tachi

    I also seen this in my VBox.log when looking at something else. Maybe it is trying to distinguish if this is an AMD or Intel CPU about CPU capabilities. Then, it throws this error but it might actually not preventing VBox to work as usual. Maybe just avoid reporting this error in the code?

    I've the same and the supposedly errors seems to have no effect on VirtualBox.

    Similar with another issue #640

  4. RSpliet commented on Jun 5, 2026

    @RSpliet

    I have the same issue but backwards on Fedora 44 (and 43). From my current kernel 7.0.10 on an AMD system when modprobing vboxdrv:

    [  424.232651] kvm_intel: VMX not supported by CPU 10
    [  424.259063] decompression failed with status 6
    

    The inability to load vboxdrv prevents me from starting a VM. I seem to be able to work around this issue by unxz'ing the module, and then running depmod.

  5. cybersnow02 commented on Jun 6, 2026

    @cybersnow02

    @RSpliet Did you download the latest release from virtualbox? If still not working, did you download the latest testbuild? Might not be related to this actual issue. You should open a new ticket for this error you get.

  6. RSpliet commented on Jun 6, 2026

    @RSpliet

    @cybersnow02 Actually you're (at least mostly) right. Turns out there were two issues.

    1. kvm_intel being probed on an AMD machine when modprobing vboxdrv,
    2. Decompression failing, which prevents vboxdrv from being loaded.

    Number 2 is entirely self inflicted: I have a shell script to decompress, sign and re-compress the modules such that the kernel doesn't reject the module. Apparently the recompression generated a file that could not be decompressed by the kernel's XZ decompressor. Re-compressing the module files with xz -C crc32 --lzma2=dict=1MiB resolves this issue.

    Issue 1 is probably the same issue as this ticket. It is however harmless in that it doesn't prevent vboxdrv from being loaded. So my apologies for the earlier noise.

  7. cybersnow02 commented on Jun 7, 2026

    @cybersnow02

    @RSpliet Yes, the module must be signed/compressed a special manner. hehe

  8. AndyFx12 commented on Nov 6, 2026

    @AndyFx12

    Issue is still present on 7.2.22.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions