Alexander Deucher at AMD came up with a quick fix for the 5.15 branch, which I was able to test and verify - amdgpu_fb.c, line 344:
- if (!amdgpu_device_has_dc_support(adev) && !amdgpu_virtual_display)
+ if (!amdgpu_device_has_dc_support(adev) && !amdgpu_virtual_display && !amdgpu_sriov_vf(adev))
With this change, the amdgpu module loads and runs correctly on AWS EC2 G4ad instances. The change only affects datacenter cards with SRIO-V support and not consumer cards.
Alexander Deucher at AMD came up with a quick fix for the 5.15 branch, which I was able to test and verify - amdgpu_fb.c, line 344:
- if (!amdgpu_ device_ has_dc_ support( adev) && !amdgpu_ virtual_ display) device_ has_dc_ support( adev) && !amdgpu_ virtual_ display && !amdgpu_ sriov_vf( adev))
+ if (!amdgpu_
With this change, the amdgpu module loads and runs correctly on AWS EC2 G4ad instances. The change only affects datacenter cards with SRIO-V support and not consumer cards.