On 3/15/26 07:55, Srinivasan Shanmugam wrote:
> amdgpu_vm_flush() sends commands to the GPU to update the VM page tables
> for a job.
>
> When a job uses a GPU virtual address space, the GPU needs to refresh
> its address translations after the driver updates the page tables.
> A VM flush tells the GPU to forget old address translations and use the
> updated page table mappings.
>
> This flush command is not supported on all rings. Only rings that
> implement the emit_vm_flush() callback know how to emit the correct
> hardware command for this operation.
>
> The function already gates vm_flush_needed on the presence of
> ring->funcs->emit_vm_flush earlier in the logic. However, static
> analysis tools such as Smatch may not track this relationship through
> the vm_flush_needed boolean and warn that emit_vm_flush() could be NULL
> when the VM flush command is emitted later.
>
> Fixes the below:
> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c:826 amdgpu_vm_flush() error: we
> previously assumed 'ring->funcs->emit_vm_flush' could be null (see line 788)
Absolutely clear NAK.
The test on line 788 is there to not set vm_flush_needed if
ring->funcs->emit_vm_flush is NULL.
So testing that again here is completely nonsense.
Regards,
Christian.
>
> Fixes: b3cd285fa68d ("drm/amdgpu: update the PASID mapping only on demand")
> Cc: Dan Carpenter <[email protected]>
> Cc: Felix Kuehling <[email protected]>
> Cc: Christian König <[email protected]>
> Cc: Alex Deucher <[email protected]>
> Signed-off-by: Srinivasan Shanmugam <[email protected]>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index b89013a6aa0b..cc79cb7dd4e1 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -807,7 +807,7 @@ void amdgpu_vm_flush(struct amdgpu_ring *ring, struct
> amdgpu_job *job,
> if (cleaner_shader_needed)
> ring->funcs->emit_cleaner_shader(ring);
>
> - if (vm_flush_needed) {
> + if (vm_flush_needed && ring->funcs->emit_vm_flush) {
> trace_amdgpu_vm_flush(ring, job->vmid, job->vm_pd_addr);
> amdgpu_ring_emit_vm_flush(ring, job->vmid, job->vm_pd_addr);
> }