Re: [PATCH 17/23] dma-buf: specify usage while adding fences to dma_resv obj v5

[Christian König <christian.koenig@xxxxxxx>]

Am 03.04.22 um 00:16 schrieb Bas Nieuwenhuizen:
> On Mon, Mar 21, 2022 at 2:59 PM Christian König
> <ckoenig.leichtzumerken@xxxxxxxxx> wrote:
> > [SNIP]
> > @@ -519,17 +513,17 @@ EXPORT_SYMBOL_GPL(dma_resv_iter_first);
> >    */
> >   struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor)
> >   {
> > -       unsigned int idx;
> > +       struct dma_fence *fence;
> >
> >          dma_resv_assert_held(cursor->obj);
> >
> >          cursor->is_restarted = false;
> > -       if (!cursor->fences || cursor->index >= cursor->fences->shared_count)
> > +       if (!cursor->fences || cursor->index >= cursor->fences->num_fences)
> >                  return NULL;
> >
> > -       idx = cursor->index++;
> > -       return rcu_dereference_protected(cursor->fences->shared[idx],
> > -                                        dma_resv_held(cursor->obj));
> > +       dma_resv_list_entry(cursor->fences, cursor->index++,
> > +                           cursor->obj, &fence, &cursor->fence_usage);
> Shouldn't we skip the current fence if cursor->fence_usage doesn't
> match cursor->usage ? (similar to what is done wrt the unlocked
> variant)

Oh, good point. Totally missed that.

Thanks,
Christian.