[Intel-gfx] [PATCH 4/9] drm/i915: Tidy awaiting on dma-fences

[Chris Wilson]

Quoting Mika Kuoppala (2020-05-08 16:50:22)
> Chris Wilson <chris at chris-wilson.co.uk> writes:
> 
> > Just tidy up the return handling for completed dma-fences. While it may
> > return errors for invalid fence, we already know that we have a good
> > fence and the only error will be an already signaled fence.
> >
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >  drivers/gpu/drm/i915/i915_sw_fence.c | 10 ++++------
> >  1 file changed, 4 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c
> > index 7daf81f55c90..295b9829e2da 100644
> > --- a/drivers/gpu/drm/i915/i915_sw_fence.c
> > +++ b/drivers/gpu/drm/i915/i915_sw_fence.c
> > @@ -546,13 +546,11 @@ int __i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence,
> >       cb->fence = fence;
> >       i915_sw_fence_await(fence);
> >  
> > -     ret = dma_fence_add_callback(dma, &cb->base, __dma_i915_sw_fence_wake);
> > -     if (ret == 0) {
> > -             ret = 1;
> > -     } else {
> > +     ret = 1;
> > +     if (dma_fence_add_callback(dma, &cb->base, __dma_i915_sw_fence_wake)) {
> > +             /* fence already signaled */
> 
> This seems to hold water now. Perhaps for eternity.
> 
> But how about if (dma_fence_add_callback() == -ENOENT) ret = 0; else
> GEM_BUG_ON()?

Because that's just ugly. If we do not install the callback, we need to
signal the callback. The only question is whether or not an error there
is moot -- we either have the fence, or we would have exploded. The
fence callback will be propagating errors along the fence.
-Chris