Products containing DCN 3.1.4 and VCN 4.0.2 have a problem in kernel 6.0.y where after amdgpu has loaded several X86 cores will be running at ~500MHz and other cores aren't able to scale up to the proper state.
This is fixed in 6.1-rc1, by changes to the DPM messaging in VCN in commit 167be8522821f ("drm/amdgpu/vcn: update vcn4 fw shared data structure"). However backporting this commit requires 3 dependency commits that are rather large:
commit dc5f3829a752b ("drm/amdgpu: sriov remove vcn_4_0 and jpeg_4_0") commit 63127922e1556 ("drm/amdgpu/vcn: Add MMSCH v4_0 support for sriov") commit aa44beb5f0155 ("drm/amdgpu/vcn: Add sriov VCN v4_0 unified queue support")
Instead of backporting the whole series to fix this problem instead backport part of commit 167be8522821f ("drm/amdgpu/vcn: update vcn4 fw shared data structure"). and make some context changes for the lack of `AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG` and `struct amdgpu_fw_shared_rb_setup rb_setup`.
Ruijing Dong (1): drm/amdgpu/vcn: update vcn4 fw shared data structure
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h | 7 +++++++ drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 4 ++++ 2 files changed, 11 insertions(+)