On 10/30/26 09:59, Jesse Zhang wrote:
> GFX11 programs the SPM VMID into RLC_SPM_MC_CNTL only in
> gfx_v11_0_rlc_init(), which runs during sw_init, before hw_init starts the
> RLC. The RLC comes up with the register cleared, and nothing programs it
> again after a GPU reset or a resume either. Until commit 38e07c072833
> ("drm/amdgpu: remove the now-redundant runtime SPM VMID update") the
> update on submission covered that up; since then a process which reserved
> the VMID for SPM never gets it. GFX9 and GFX10 program it in the RLC
> resume instead.
>
> Program it at the end of the RLC resume, which runs on hw_init, GPU reset
> and resume, like GFX9 and GFX10 do, and call the gfx function directly
> instead of going through the update_spm_vmid() callback.
>
> This doesn't cover GFXOFF, the RLC doesn't save and restore
> RLC_SPM_MC_CNTL over it.
>
> v2: new patch, split the init ordering out of the GFXOFF issue
>
> Fixes: 38e07c072833 ("drm/amdgpu: remove the now-redundant runtime SPM VMID
> update")
> Signed-off-by: Jesse Zhang <[email protected]>
Reviewed-by: Christian König <[email protected]> for both patches.
> ---
> drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 13 +++++++++----
> 1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> index cdf543f1cd9a..cb2c4d0dab87 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> @@ -362,6 +362,8 @@ static void gfx_v11_0_ring_invalidate_tlbs(struct
> amdgpu_ring *ring,
> bool all_hub, uint8_t dst_sel);
> static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
> static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev, int
> xcc_id);
> +static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, int xcc_id,
> + unsigned int vmid);
> static void gfx_v11_0_userq_priv_fault_work(struct work_struct *work);
> static void gfx_v11_0_update_perf_clk(struct amdgpu_device *adev,
> bool enable);
> @@ -942,10 +944,6 @@ static int gfx_v11_0_rlc_init(struct amdgpu_device *adev)
> return r;
> }
>
> - /* init spm vmid with 0xf */
> - if (adev->gfx.rlc.funcs->update_spm_vmid)
> - adev->gfx.rlc.funcs->update_spm_vmid(adev, 0,
> adev->vm_manager.spm_vmid);
> -
> return 0;
> }
>
> @@ -2582,6 +2580,13 @@ static int gfx_v11_0_rlc_resume(struct amdgpu_device
> *adev)
>
> adev->gfx.rlc.funcs->start(adev);
> }
> +
> + /*
> + * The RLC comes up with RLC_SPM_MC_CNTL cleared, program the SPM VMID
> + * once it runs, also after a GPU reset or a resume.
> + */
> + gfx_v11_0_update_spm_vmid(adev, 0, adev->vm_manager.spm_vmid);
> +
> return 0;
> }
>
