[PATCH 1/2] drm/amdgpu: skip to program GFXDEC registers for PM abort case

[Alex Deucher]

On Wed, Jan 24, 2024 at 10:46 PM Prike Liang <Prike.Liang at amd.com> wrote:
>
> In the PM abort cases, the gfx power rail doesn't turn off so
> some GFXDEC registers/CSB can't reset to default vaule. In order
> to avoid unexpected problem now need skip to program GFXDEC registers
> and bypass issue CSB packet for PM abort case.
>
> Signed-off-by: Prike Liang <Prike.Liang at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 2 ++
>  drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 ++
>  drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c   | 6 ++++++
>  3 files changed, 10 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index c5f3859fd682..9e9e3385b5d3 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -1079,6 +1079,8 @@ struct amdgpu_device {
>         bool                            in_s3;
>         bool                            in_s4;
>         bool                            in_s0ix;
> +       /* indicate amdgpu suspension status */
> +       bool                            pm_complete;

Maybe rename this suspend_complete to better specify that it only
applies on the suspend side.

Alex

>
>         enum pp_mp1_state               mp1_state;
>         struct amdgpu_doorbell_index doorbell_index;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> index 475bd59c9ac2..9cb8f7fe55cf 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
> @@ -2472,6 +2472,7 @@ static int amdgpu_pmops_suspend(struct device *dev)
>         struct drm_device *drm_dev = dev_get_drvdata(dev);
>         struct amdgpu_device *adev = drm_to_adev(drm_dev);
>
> +       adev->pm_complete = false;
>         if (amdgpu_acpi_is_s0ix_active(adev))
>                 adev->in_s0ix = true;
>         else if (amdgpu_acpi_is_s3_active(adev))
> @@ -2486,6 +2487,7 @@ static int amdgpu_pmops_suspend_noirq(struct device *dev)
>         struct drm_device *drm_dev = dev_get_drvdata(dev);
>         struct amdgpu_device *adev = drm_to_adev(drm_dev);
>
> +       adev->pm_complete = true;
>         if (amdgpu_acpi_should_gpu_reset(adev))
>                 return amdgpu_asic_reset(adev);
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> index 57808be6e3ec..0abdc85eda77 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
> @@ -3034,6 +3034,12 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev)
>
>         gfx_v9_0_cp_gfx_enable(adev, true);
>
> +       /* TODO: double confirm whether need to reinitialize gfxedc and submit csb packet
> +        * on the other gfx generations for the pm suspend abort case. */
> +       if (adev->in_suspend && !adev->pm_complete) {
> +               DRM_INFO(" will skip the csb ring write\n");
> +               return 0;
> +       }
>         r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3);
>         if (r) {
>                 DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r);
> --
> 2.34.1
>