On Wed, Mar 25, 2026 at 8:17 AM Rob Clark <[email protected]> wrote: > > On Tue, Mar 24, 2026 at 3:20 PM Dmitry Baryshkov > <[email protected]> wrote: > > > > On Tue, Mar 24, 2026 at 03:05:17PM -0700, Rob Clark wrote: > > > This restriction applies to mapping of _NO_SHARE objs in the kms vm as > > > well as importing/exporting BOs. Since the DPU has it's own VM, scanout > > > counts as "exporting" a BO from outside of it's host VM. > > > > > > Signed-off-by: Rob Clark <[email protected]> > > > --- > > > v2: Fix issue with MAP_NULL > > > > > > drivers/gpu/drm/msm/msm_gem_vma.c | 6 ++++++ > > > 1 file changed, 6 insertions(+) > > > > > > diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c > > > b/drivers/gpu/drm/msm/msm_gem_vma.c > > > index 1c6b0920c0d8..43d61e0919bd 100644 > > > --- a/drivers/gpu/drm/msm/msm_gem_vma.c > > > +++ b/drivers/gpu/drm/msm/msm_gem_vma.c > > > @@ -373,6 +373,12 @@ msm_gem_vma_new(struct drm_gpuvm *gpuvm, struct > > > drm_gem_object *obj, > > > struct msm_gem_vma *vma; > > > int ret; > > > > > > + /* _NO_SHARE objs cannot be mapped outside of their "host" vm: */ > > > + if (obj && (to_msm_bo(obj)->flags & MSM_BO_NO_SHARE) && > > > + GEM_WARN_ON(obj->resv != drm_gpuvm_resv(gpuvm))) { > > > > Can this be used to spam the logs? > > yeah, perhaps it should use UERR() instead
So, for the case that I was concerned about, we should actually just reject creating a drm fb from _NO_SHARE BO(s) (and report error w/ UERR()). I'll send a patch for that. With that, I think it is safe to keep this WARN_ON(), since the only path to hit it should be a driver bug. BR, -R > > > > > + return ERR_PTR(-EINVAL); > > > + } > > > + > > > drm_gpuvm_resv_assert_held(&vm->base); > > > > > > vma = kzalloc(sizeof(*vma), GFP_KERNEL); > > > -- > > > 2.53.0 > > > > > > > -- > > With best wishes > > Dmitry
