About the lookat() function

Question

This is the function:

lookat(vec3 eye, vec3 at, vec3 up)

As we all know, the up vector can be(0.0, 1.0, 0.0). It cannot be parallel to the observation vector n, or there will be somethingw wrong.

My question is how to calculate the up vector so that it passes through the point eye and is perpendicular to the n vector.