On 7/1/26 07:53, chong li wrote:
> Move the initialization of non-GPU resources
> out of the full GPU access region during AMDGPU device initialization

As far as I can see that won't work like this.

There are a lot of steps which require full GPU access.

What exactly is the justification of the change?

Regards,
Christian.

> 
> Signed-off-by: chong li <[email protected]>
> Co-authored-by: Cursor <[email protected]>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c 
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 610d82b79de3..c2ce4659ddc7 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -1994,10 +1994,6 @@ static int amdgpu_device_ip_early_init(struct 
> amdgpu_device *adev)
>       amdgpu_device_enable_virtual_display(adev);
>  
>       if (amdgpu_sriov_vf(adev)) {
> -             r = amdgpu_virt_request_full_gpu(adev, true);
> -             if (r)
> -                     return r;
> -
>               r = amdgpu_virt_init_critical_region(adev);
>               if (r)
>                       return r;
> @@ -2159,6 +2155,12 @@ static int amdgpu_device_ip_early_init(struct 
> amdgpu_device *adev)
>       if (!total)
>               return -ENODEV;
>  
> +     if (amdgpu_sriov_vf(adev)) {
> +             r = amdgpu_virt_request_full_gpu(adev, true);
> +             if (r)
> +                     return r;
> +     }
> +
>       if (adev->gmc.xgmi.supported)
>               amdgpu_xgmi_early_init(adev);
>  

Reply via email to