[PATCH v1] drm/amdgpu: implement workaround for sdma dcc corruption

[PATCH v1] drm/amdgpu: implement workaround for sdma dcc corruption

[Pierre-Eric Pelloux-Prayer]

For unknown reasons, on gfx12 using multiple entities can causes
random corruption of BOs with DCC.
This workaround seems to prevent the issue until the root cause
is understood and fixed.

Link: https://gitlab.freedesktop.org/drm/amd/-/work_items/5663
Fixes: 3a6f6eeb3db5 ("drm/amdgpu: give ttm entities access to all the sdma scheds")
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 3d620ec2937f..a826e091ea58 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2392,6 +2392,11 @@ void amdgpu_ttm_enable_buffer_funcs(struct amdgpu_device *adev)
 
 	num_clear_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
 	num_move_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
+        /* TODO: workaround for DCC corruption when moving BOs from multiple queues at
+         * the same time: use a single queue until the root cause is identified and fixed.
+         */
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, SDMA0_HWIP, 0)) == 7)
+		num_move_entities = 1;
 
 	adev->mman.clear_entities = kcalloc(num_clear_entities,
 						sizeof(struct amdgpu_ttm_buffer_entity),
-- 
2.43.0

Re: [PATCH v1] drm/amdgpu: implement workaround for sdma dcc corruption

[Alex Deucher]

On Thu, Sep 24, 2026 at 8:44 AM Pierre-Eric Pelloux-Prayer
<pierre-eric.pelloux-prayer@amd.com> wrote:
>
> For unknown reasons, on gfx12 using multiple entities can causes
> random corruption of BOs with DCC.
> This workaround seems to prevent the issue until the root cause
> is understood and fixed.
>
> Link: https://gitlab.freedesktop.org/drm/amd/-/work_items/5663
> Fixes: 3a6f6eeb3db5 ("drm/amdgpu: give ttm entities access to all the sdma scheds")
> Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 3d620ec2937f..a826e091ea58 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -2392,6 +2392,11 @@ void amdgpu_ttm_enable_buffer_funcs(struct amdgpu_device *adev)
>
>         num_clear_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
>         num_move_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
> +        /* TODO: workaround for DCC corruption when moving BOs from multiple queues at
> +         * the same time: use a single queue until the root cause is identified and fixed.
> +         */
> +       if (IP_VERSION_MAJ(amdgpu_ip_version(adev, SDMA0_HWIP, 0)) == 7)

Probably just limit this to 7.0.0 and 7.0.1.  sdma 7.1 doesn't have
DCC.  WIth that fixed:
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

Alex

> +               num_move_entities = 1;
>
>         adev->mman.clear_entities = kcalloc(num_clear_entities,
>                                                 sizeof(struct amdgpu_ttm_buffer_entity),
> --
> 2.43.0
>

Re: [PATCH v1] drm/amdgpu: implement workaround for sdma dcc corruption

[Christian König]

On 9/24/26 14:43, Pierre-Eric Pelloux-Prayer wrote:
> For unknown reasons, on gfx12 using multiple entities can causes
> random corruption of BOs with DCC.
> This workaround seems to prevent the issue until the root cause
> is understood and fixed.
> 
> Link: https://gitlab.freedesktop.org/drm/amd/-/work_items/5663
> Fixes: 3a6f6eeb3db5 ("drm/amdgpu: give ttm entities access to all the sdma scheds")
> Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 3d620ec2937f..a826e091ea58 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -2392,6 +2392,11 @@ void amdgpu_ttm_enable_buffer_funcs(struct amdgpu_device *adev)
>  
>  	num_clear_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
>  	num_move_entities = MIN(adev->mman.num_buffer_funcs_scheds, TTM_NUM_MOVE_FENCES);
> +        /* TODO: workaround for DCC corruption when moving BOs from multiple queues at
> +         * the same time: use a single queue until the root cause is identified and fixed.
> +         */
> +	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, SDMA0_HWIP, 0)) == 7)

Thinking it through it would be nice if we could check for that DCC HW generation instead of the SDMA version.

But we need a fix ASAP so the patch is Reviewed-by: Christian König <christian.koenig@amd.com>.

Regards,
Christian.

> +		num_move_entities = 1;
>  
>  	adev->mman.clear_entities = kcalloc(num_clear_entities,
>  						sizeof(struct amdgpu_ttm_buffer_entity),