Right now, every gbm_bo object created in GBMBufferSwapchain is created in linear format. This shouldn't be the case, giving the graphics driver chance to choose the best possible attributes for the buffer object (e.g. most optimal tiling approach). Currently, linearity is required only in media playback, when falling back onto software decoder, results of which are then copied into the spawned gbm_bo object. In most other cases avoiding linearity might be beneficial, so it should be done. This will most likely be solved by passing the GBMBufferSwapchain constructor or getBuffer() call some specific flags requesting a linear buffer.