Re: [PATCH] drm/amdgpu: add VCN poison consumption handler for SRIOV

[Lazar, Lijo]

On 9/27/2022 1:05 PM, Tao Zhou wrote:
Inform host and let host handle consumption interrupt.

Signed-off-by: Tao Zhou <tao.zh...@amd.com>
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 12 ++++++++++--
  1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index f36e4f08db6d..cb2885999219 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -1243,8 +1243,16 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device 
*adev,
        if (!ras_if)
                return 0;
  
-	ih_data.head = *ras_if;
-       amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+       if (!amdgpu_sriov_vf(adev)) {
+               ih_data.head = *ras_if;
+               amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+       } else {
+               if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
+                       adev->virt.ops->ras_poison_handler(adev);
+               else
+                       dev_warn(adev->dev,
+                               "No ras_poison_handler interface in SRIOV for 
VCN!\n");
+       }
  

As guest is not a trusted domain, I guess we need to get more clarity 
about the final flow of events. Need to hold this.

Thanks,
Lijo

        return 0;
  }