This is not necessarily a problem - in a quantum mechanical formulation, it's simple enough to resolve (for instance) a grandfather paradox by looking at quantum interference around causal loops (almost 100% analogous to the way interference on a violin string will restrict stable modes to certain frequencies, interference around a causal loop will only allow stable self-sustaining situations to survive, so any grandfather paradox would self-interfere and not be allowed). The main difference is that with closed timelike curves the quantum effects can get a lot stronger than without them, even affecting very large objects.
Even in GR, you can sort of allow causality violations if you're talking about field theory, but it's not pleasant; the problem is that the usual existence/uniqueness proofs (things like "If a solution to the field equations exists on some spacelike slice of spacetime satisfying a bunch of conditions, then there exists a (maybe unique) solution if that spacelike slice is carried forward in time", roughly speaking) usually rely on strong causality assumptions, and for instance it's pretty easy, depending on the situation, to come up with initial field configurations that have no solutions as they're swept forward because of closed timelike curves further along the manifold.
But as to FTL, let me horrifically simplify. Imagine space as a 2D grid of points. Vertical is time (up is the future), left-right is 1D of space. Take every point, and connect it to the one directly above it, the one above and right, and the one above and left, with a directed link. You are allowed to move through spacetime along those lines. You can see your past and future lightcones on this graph, if you look. Light always travels diagonally on this graph.
Normal travel in this horrifically degenerate spacetime is travel that can be expressed on those links. FTL is somehow figuring out how to bypass the graph structure of space, which the shape of that space doesn't permit. Space is literally shaped so that FTL isn't really possible.
Reality is of course substantially more complicated than this, and this model is useless for any purpose beyond trying to show what I mean by space being shaped such that FTL isn't possible.
This also sort of demonstrates another thing, which is that ultimately, any point that you can't reach is, from the point of view of where you are, really the same. They are all equally simply "unreachable". So you if you create an "unreachable" drive that lets you get there, you can basically just use it twice to jump far away from where you are now, then from there just jump directly into your past. Also, there is no compelling reason to think you would have a drive that would only let you go "one light year"; it would open the entire universe to you. In General Relativity, this still basically holds true, though it requires more explanations about how there isn't one true reference frame, etc, etc, but FTL really is time travel and it really does just shoot physics and causality full of holes.