Nvbufsurface with EGL to access it on cuda kernel

Hi, I’m currently working with the nvvidconv plugin source and aiming to use the NvBufSurface pointer within my CUDA kernel for format conversion. I’ve implemented the EGL-based logic, but I’m encountering an nvgpu error when trying to access srcEgl.frame.pPitch[0] inside the CUDA kernel.

Could you suggest the correct way to access srcEgl.frame.pPitch[0] in a CUDA kernel?