[PATCH] drm/amdgpu: introduce vram lost for reset

Christian K├Ânig ckoenig.leichtzumerken at gmail.com
Fri Aug 30 20:15:03 UTC 2019


Am 27.08.19 um 10:34 schrieb Monk Liu:
> for SOC15/vega10 the BACO reset & mode1 would introduce vram lost
> in high end address range, current kmd's vram lost checking cannot
> catch it since it only check very ahead visible frame buffer

Looks mostly good to me.

Only question I have is why to use a macro? Some inline function should 
do as well.

Christian.

>
> Signed-off-by: Monk Liu <Monk.Liu at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu.h        | 1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++--
>   drivers/gpu/drm/amd/amdgpu/soc15.c         | 2 ++
>   3 files changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index f6ae565..11e0fc0 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -1155,6 +1155,7 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
>   #define amdgpu_asic_get_pcie_usage(adev, cnt0, cnt1) ((adev)->asic_funcs->get_pcie_usage((adev), (cnt0), (cnt1)))
>   #define amdgpu_asic_need_reset_on_init(adev) (adev)->asic_funcs->need_reset_on_init((adev))
>   #define amdgpu_asic_get_pcie_replay_count(adev) ((adev)->asic_funcs->get_pcie_replay_count((adev)))
> +#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter));
>   
>   /* Common functions */
>   bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 02b3e7d..31690e9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -3482,7 +3482,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
>   	amdgpu_virt_init_data_exchange(adev);
>   	amdgpu_virt_release_full_gpu(adev, true);
>   	if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
> -		atomic_inc(&adev->vram_lost_counter);
> +		amdgpu_inc_vram_lost(adev);
>   		r = amdgpu_device_recover_vram(adev);
>   	}
>   
> @@ -3648,7 +3648,7 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive,
>   				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
>   				if (vram_lost) {
>   					DRM_INFO("VRAM is lost due to GPU reset!\n");
> -					atomic_inc(&tmp_adev->vram_lost_counter);
> +					amdgpu_inc_vram_lost(tmp_adev);
>   				}
>   
>   				r = amdgpu_gtt_mgr_recover(
> diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
> index fe2212df..8af7501 100644
> --- a/drivers/gpu/drm/amd/amdgpu/soc15.c
> +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
> @@ -557,10 +557,12 @@ static int soc15_asic_reset(struct amdgpu_device *adev)
>   {
>   	switch (soc15_asic_reset_method(adev)) {
>   		case AMD_RESET_METHOD_BACO:
> +			amdgpu_inc_vram_lost(adev);
>   			return soc15_asic_baco_reset(adev);
>   		case AMD_RESET_METHOD_MODE2:
>   			return soc15_mode2_reset(adev);
>   		default:
> +			amdgpu_inc_vram_lost(adev);
>   			return soc15_asic_mode1_reset(adev);
>   	}
>   }



More information about the amd-gfx mailing list