All quantum gates are reversible, in the sense that they always admit an inverse which is a quantum gate as well. This comes from the fact that quantum gates are (usually) required to be unitary.
What you probably meant to ask is how you call gates $U$ such that $U^2=I$. If you assume $U$ is unitary, that is $U^\dagger U=UU^\dagger=I$, then this is equivalent to the condition $U^\dagger =U$. Such gates are called Hermitian (or symmetric when they are real).
If, instead, you are referring to gates whose action is invariant under exchange of the inputs, then that can mean gates $U$ such that $U|a,b\rangle=U|b,a\rangle$ for all one-qubit states $|a\rangle$ and $|b\rangle$. This essentially means that the two middle columns of the matrix (in the standard choice of canonical basis) are the same. Yet another possible interpretation is that the matrix is required to have the same description when exchanging both the two input qubits and the two output qubits. This amounts to asking for $U$ to commute with the swap operation: $[U,\mathrm{SWAP}]=0$. You can compute explicitly the form of these matrices, but I don't think there is any universally accepted special name for these sorts of operations.