From: Chen Li chenli@uniontech.com

[ Upstream commit ab8363d3875a83f4901eb1cc00ce8afd24de6c85 ]

I met a gpu addr bug recently and the kernel log tells me the pc is memcpy/memset and link register is radeon_uvd_resume.

As we know, in some architectures, optimized memcpy/memset may not work well on device memory. Trival memcpy_toio/memset_io can fix this problem.

BTW, amdgpu has already done it in: commit ba0b2275a678 ("drm/amdgpu: use memcpy_to/fromio for UVD fw upload"), that's why it has no this issue on the same gpu and platform.

Signed-off-by: Chen Li chenli@uniontech.com Reviewed-by: Christian König christian.koenig@amd.com Signed-off-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/gpu/drm/radeon/radeon_uvd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c index 95f4db70dd22..fde9c69ecc86 100644 --- a/drivers/gpu/drm/radeon/radeon_uvd.c +++ b/drivers/gpu/drm/radeon/radeon_uvd.c @@ -286,7 +286,7 @@ int radeon_uvd_resume(struct radeon_device *rdev) if (rdev->uvd.vcpu_bo == NULL) return -EINVAL;

- memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->size); + memcpy_toio((void __iomem *)rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->size);

size = radeon_bo_size(rdev->uvd.vcpu_bo); size -= rdev->uvd_fw->size; @@ -294,7 +294,7 @@ int radeon_uvd_resume(struct radeon_device *rdev) ptr = rdev->uvd.cpu_addr; ptr += rdev->uvd_fw->size;

- memset(ptr, 0, size); + memset_io((void __iomem *)ptr, 0, size);

return 0; }