[PATCH 2/2] drm/amdgpu/gfx9: Explicitly halt CP before init

[SRINIVASAN SHANMUGAM]

On 9/19/2024 10:17 PM, Alex Deucher wrote:
> Need to make sure it's halted as we don't know what state
> the GPU may have been left in previously.
>
> Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 ++++
>   1 file changed, 4 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> index d4f19eb8ff6d3..3a3ef93f2f43a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> @@ -3931,6 +3931,10 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
>   			return r;
>   	}
>   
> +	if (adev->gfx.num_gfx_rings)
> +		gfx_v9_0_cp_gfx_enable(adev, false);
> +	gfx_v9_0_cp_compute_enable(adev, false);
> +
>   	r = gfx_v9_0_kiq_resume(adev);
>   	if (r)
>   		return r;
This is a precautionary measure to make sure that the GPU is in a known 
state before the driver attempts to initialize it. Adds calls to 
gfx_v9_0_cp_gfx_enable and gfx_v9_0_cp_compute_enable with enable set to 
false to halt the CP before the KIQ (Kernel Interface Queue) is resumed.

Reviewed-by: Srinivasan Shanmugam <srinivasan.shanmugam at amd.com>