Claim: $A$ is a ring such that every primary ideal is prime if and
only if $A$ is absolutely flat.
Let us say a ring is a PP(primary is prime) ring if every primary ideal is
prime.
Suppose $A$ is absolutely flat, then $A$ is PP. (exercise 3 in page 55 of "Introduction to Commutative Algebra" by Atiyah & Macdonald)
We only need to show that if $A$ is PP then $A$ is absolutely flat.
Notice that every primary ideal of $S^{-1}A$ is of the form $S^{-1}\mathfrak{q}$
where $\mathfrak{q}$ is a $\mathfrak{p}$-primary ideal such that
$\mathfrak{p}\cap S=\emptyset$. And if $J/I$ (here $J\supset I$) is a
primary ideal of $A/I$ then every zerodivisor in $A/J=(A/I)/(J/I)$ is
nilpotent, so $J$ is is primary ideal of $A$.
Now it is easy to show that, if $A$ is PP, then $S^{-1}A$ and
$A/I$ is PP for a multiplicatively closed subset $S$ and an ideal
$I$ and $\mathfrak{m}^2=\mathfrak{m}$ for every maximal ideal
$\mathfrak{m}$.
Based on these properties we are able to prove that PP rings are absolutely
flat.
We first show every prime ideal in $A$ is maximal. Suppose not, there are
two distinct prime ideals $\mathfrak{p}\subset \mathfrak{m}$ where
$\mathfrak{m}$ is maximal. Now $B=(A/\mathfrak{p})_{\mathfrak{m}}$ is
a PP, local domain(but not a field) we also use $\mathfrak{m}$ to
denote the maximal ideal of $B$. Let $0\neq b\in \mathfrak{m}$,
suppose $\mathfrak{q}$ is a minimal prime containing $b$, then
$C=B_{\mathfrak{q}}$ is a PP, local domain(not a field) and the only maximal ideal of
$C$ is minimal over the ideal $(b)=bC$ so $(b)$ is primary and hence
maximal, $(bC)^2=bC$, thus $bC=0$. It is impossible. We have proved
that every prime ideal in $A$ is maximal.
Let $\mathfrak{m}$ be any prime ideal in $A$, then $A_{\mathfrak{m}}$
is PP. If $A_{\mathfrak{m}}$ is not a field, pick any nonezero $x\in
\mathfrak{m}A_{\mathfrak{m}}$, then $(x)$ is primary hence equals
$\mathfrak{m}A_{\mathfrak{m}}$, so $\mathfrak{m}A_m=0$,
contradiction.
Hence every prime ideal $\mathfrak{m}$ in $A$ is maximal and
$A_{\mathfrak{m}}$ is a field, so $A$ is absolutely flat. We are done.
If $f(M) \subseteq I \subseteq S$ is an ideal, then $M \subseteq f^{-1}(I) \subseteq R$. Since $M$ is maximal, we get $M=f^{-1}(I)$ or $f^{-1}(I)=R$, i.e. $f(M)=I$ or $I=S$. $\mathrm{QED}$
Best Answer
Suppose $r,s\in R$ and $rs\in f(P)$. Since $f$ is surjective, you have $r=f(a)$ and $s=f(b)$ for some $a,b\in A$, so the condition is $$ f(a)f(b)=f(c) $$ for some $c\in P$. In turn this becomes $ab-c\in\ker f$. If $P\supseteq\ker f$, you can conclude $ab-c\in P$, so also $ab\in P$ and it's easy to finish.
Without the assumption $\ker f\subseteq P$ there's no way to prove the statement. For instance, consider $A=\mathbb{Z}$ and $R=\mathbb{Z}/2\mathbb{Z}$, with $f$ the canonical projection. Then $f(3\mathbb{Z})=R$, which is not a prime ideal.
If you think more about the problem, the assertion would be equivalent to the statement that, if $P$ is a prime ideal of $A$ and $I$ is any ideal of $A$, then $P+I$ is prime: just consider the projection $f\colon A\to A/I$. Saying that $f(P)$ is prime in $A/I$ implies that $f^{-1}(f(P))=P+I$ is prime in $A$.