<div dir="ltr">I got this kernel Oops when running ROCm kernel in a ARM64 machine with Vega64 card.<div><br><div><pre style="color:rgb(0,0,0);white-space:pre-wrap">---
 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 20d7e4775d49..b6945af39b95 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -726,8 +726,8 @@ static bool gmc_v9_0_keep_stolen_memory(struct amdgpu_device *adev)
 static int gmc_v9_0_allocate_vm_inv_eng(struct amdgpu_device *adev)
 {
        struct amdgpu_ring *ring;
-       unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] =
-               {GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP};
+       unsigned long vm_inv_engs[AMDGPU_MAX_VMHUBS] =
+                     {GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP};
        unsigned i;
        unsigned vmhub, inv_eng;
 
@@ -743,7 +743,7 @@ static int gmc_v9_0_allocate_vm_inv_eng(struct amdgpu_device *adev)
                }
 
                ring->vm_inv_eng = inv_eng - 1;
-               change_bit(inv_eng - 1, (unsigned long *)(&vm_inv_engs[vmhub]));
+               change_bit(inv_eng - 1, &vm_inv_engs[vmhub]);
 
                dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n",
                         ring->name, ring->vm_inv_eng, ring->funcs->vmhub);</pre></div></div></div>