I'm reading a book about etale cohomology and there is a lemma that says if $Z\to S$ is a smooth morphism between two scheme with fibers isomorphic to $A_S^n$ then $Rf_{*} F_p=F_p$ and so by the spectral sequence cohomology groups of $Z$ and $S$ are the same. I think this can't be right just look at the line bundle $A^{n+1}-0\to P^n$.
on the other hand a line bundle is locally trivial and so when I'm trying to compute the stalks of the sheaf $Rf_{*}$ I get that this is zero and I'm confused.
so can someone explain what is the problem with my counterexample?