[PATCH v4 4/5] drm/virtio: Import prime buffers from other devices as guest blobs

[Kasireddy, Vivek]

Hi Dmitry,

> Subject: Re: [PATCH v4 4/5] drm/virtio: Import prime buffers from other
> devices as guest blobs
> 
> >  struct drm_gem_object *virtgpu_gem_prime_import(struct drm_device
> *dev,
> >  						struct dma_buf *buf)
> >  {
> > +	struct virtio_gpu_device *vgdev = dev->dev_private;
> > +	struct dma_buf_attachment *attach;
> > +	struct virtio_gpu_object *bo;
> >  	struct drm_gem_object *obj;
> > +	int ret;
> >
> >  	if (buf->ops == &virtgpu_dmabuf_ops.ops) {
> >  		obj = buf->priv;
> > @@ -275,7 +304,32 @@ struct drm_gem_object
> *virtgpu_gem_prime_import(struct drm_device *dev,
> >  		}
> >  	}
> >
> > -	return drm_gem_prime_import(dev, buf);
> > +	if (!vgdev->has_resource_blob || vgdev->has_virgl_3d)
> > +		return drm_gem_prime_import(dev, buf);
> > +
> > +	bo = kzalloc(sizeof(*bo), GFP_KERNEL);
> > +	if (!bo)
> > +		return ERR_PTR(-ENOMEM);
> > +
> > +	obj = &bo->base.base;
> > +	obj->funcs = &virtgpu_gem_dma_buf_funcs;
> > +	drm_gem_private_object_init(dev, obj, buf->size);
> > +
> > +	attach = dma_buf_dynamic_attach(buf, dev->dev,
> > +					&virtgpu_dma_buf_attach_ops, obj);
> > +	if (IS_ERR(attach)) {
> > +		kfree(bo);
> > +		return ERR_CAST(attach);
> > +	}
> > +
> > +	obj->import_attach = attach;
> > +	get_dma_buf(buf);
> > +
> > +	ret = virtgpu_dma_buf_init_obj(dev, bo, attach);
> > +	if (ret < 0)
> > +		return ERR_PTR(ret);
> 
> Perhaps for a future improvement. Think we can defer
> virtgpu_dma_buf_init_obj() until first use of the object in a case where
I believe it might be possible to do that. I think I started out that way but
figured we would need the addresses/backing in order to create a guest
blob resource. Anyway, I'll try to test this idea again when I get a chance.

> exporter supports dynamic attachment. Otherwise, we're pinning object at
> the import time, partially defeating the purpose of the dynamic
> attachment, AFAICT. I.e. if importer never uses object, then there is no
> need to bother the exporter with the pinning.
Yeah, I agree. It makes sense to defer the initialization until first usage.

Thanks,
Vivek
> 
> --
> Best regards,
> Dmitry