cudaGraphicsGLRegisterBuffer returns operation not supported

I’m trying to add some CUDA kernels to an existing OpenGL program which uses EGL. I’ve verified using EGL_CUDA_DEVICE_NV that I’m running on a CUDA device (and which CUDA device I’m running on). However, in my first call to cudaGraphicsGLRegisterBuffer I’m getting back an “operation not supported” error code – this API doesn’t even indicate this as a valid return code. Can anyone expand on why this error code could be returned from this API?