Hi Krzysztof,
On Tue, Sep 16, 2025 at 06:33:33AM +0000, Krzysztof Karas wrote:
> Due to the nature of round_up(), its first argument is
> decremented by one. drm_suballoc_hole_soffset() may return 0,
> which is then passed to round_up() and may wrap around.
> Remedy that by adding a guard against that value.
>
> Fixes: 849ee8a2f0df ("drm/suballoc: Extract amdgpu_sa.c as generic
> suballocation helper")
No Fixes needed.
I don't even think this is an issue at all. I would drop this
patch. Please, look carefully at how round_up is designed and you
will see that it does already what you need.
Andi
