How are error deltas formed in the convolutional layer of a CNN?

Question

Neural networks

I’ll make a reservation right away, the matrices I and VI have a size of 5x5, I accidentally drew 4x4 and didn’t want to redraw)
How is the error delta formed in the CNN convolutional layer? In the image, the error deltas that I need are shown as lambdas (funny, but that was more appropriate) of the VI tables. I found the formation of error deltas as (II) conv (rot180 (IX)), but in this case two 5x5 matrices from IX and from X will be obtained. What's next with them? Multiplication, addition?
I'm trying to figure it out on my fingers in convolutional ones, before that I only did fully connected ones. Thanks in advance for all replies.