On Fri, Apr 12, 2013 at 5:13 PM, Aaron Plattner aplattner@nvidia.com wrote:
@@ -117,6 +249,58 @@ int drm_gem_prime_handle_to_fd(struct drm_device *dev, } EXPORT_SYMBOL(drm_gem_prime_handle_to_fd);
+struct drm_gem_object *drm_gem_prime_import(struct drm_device *dev,
struct dma_buf *dma_buf)+{
struct dma_buf_attachment *attach;struct sg_table *sgt;struct drm_gem_object *obj;int ret;if (!dev->driver->gem_prime_import_sg_table)return ERR_PTR(-EINVAL);if (dma_buf->ops == &drm_gem_prime_dmabuf_ops) {This here breaks self-import checks since it smashes all buffers from all drivers using these helpers into one set. Which means e.g. nouveau will happily import a buffer from radoen as it's own. The only reason afaics that shit didn't completely hit the fan is that i915 still has it's own buffer ops, and everyone seems to only care about sharing with i915.
Doesn't the (obj->dev == dev) check below guard against that?
Oh dear did I just make a big idiot out of myself ;-) Sorry for the fuss, you're right. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch