Announcement

Collapse
No announcement yet.

Ubuntu 20.04 VM refuses to load radeon firmware/ROM

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Ubuntu 20.04 VM refuses to load radeon firmware/ROM

    I have a Dell PowerEdge R420 with a Radeon 8490 installed in one of the PCIe slots. The OS/hypervisor is ESXi 7.0.2, with pass through enabled for the GPU. A Windows VM with a nearly identical config has no problem enabling and using the GPU. However an Ubuntu 20.04 VM shows these messages at boot:

    Code:
    ubuntu-test:~$ dmesg | grep -e radeon
    [ 10.859594] [drm] radeon kernel modesetting enabled.
    [ 10.859702] fb0: switching to radeondrmfb from EFI VGA
    [ 10.859837] radeon 0000:1b:00.0: vgaarb: deactivate vga console
    [ 10.861465] radeon 0000:1b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
    [ 10.861487] radeon 0000:1b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
    [ 10.861489] [drm:radeon_get_bios [radeon]] *ERROR* Unable to locate a BIOS ROM
    [ 10.861552] radeon 0000:1b:00.0: Fatal error during GPU init
    [ 10.861556] [drm] radeon: finishing device.
    [ 10.869183] radeon: probe of 0000:1b:00.0 failed with error -22
    And no video is displayed on the monitor screen at all, though the VM is accessible via SSH.

    Any thoughts on a fix, or even how to begin troubleshooting this?

  • #2
    Originally posted by imrazor View Post
    I have a Dell PowerEdge R420 with a Radeon 8490 installed in one of the PCIe slots. The OS/hypervisor is ESXi 7.0.2, with pass through enabled for the GPU. A Windows VM with a nearly identical config has no problem enabling and using the GPU. However an Ubuntu 20.04 VM shows these messages at boot:

    Code:
    ubuntu-test:~$ dmesg | grep -e radeon
    [ 10.859594] [drm] radeon kernel modesetting enabled.
    [ 10.859702] fb0: switching to radeondrmfb from EFI VGA
    [ 10.859837] radeon 0000:1b:00.0: vgaarb: deactivate vga console
    [ 10.861465] radeon 0000:1b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
    [ 10.861487] radeon 0000:1b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
    [ 10.861489] [drm:radeon_get_bios [radeon]] *ERROR* Unable to locate a BIOS ROM
    [ 10.861552] radeon 0000:1b:00.0: Fatal error during GPU init
    [ 10.861556] [drm] radeon: finishing device.
    [ 10.869183] radeon: probe of 0000:1b:00.0 failed with error -22
    And no video is displayed on the monitor screen at all, though the VM is accessible via SSH.

    Any thoughts on a fix, or even how to begin troubleshooting this?
    You need to make sure the vbios rom image is available in the VM. Normally it's exposed via either the ROM BAR in PCI config space or at the legacy VGA shadow location in system memory. You need to make sure that image is available in the VM.

    Comment


    • #3
      Thanks for the reply. Unfortunately the only way I know to ensure that VRAM is "visible" within the VM is to use the pciHole.dynStart (or .start and .end) parameters. I doubt this includes the appropriate RAM address for the vBios, so I'm really not sure how to ensure that the correct 'real' memory address is accessible to the driver.

      Comment

      Working...
      X