Oof. Now, my initial thought (if you'll excuse the lazy notation; I can't be arsed) is the following. We know that da-eb = k(fc-eb) for some scalar k. Rearranging, we get da + e(k-1)b + (-fk)c = 0, which looks to me surprisingly similar to the given equation in l, m, n. We can deduce that l/d = e(k-1)/m = -fk/n, and work out k from that... and that's as far as I've got.
Looks like it should come out from there but it certainly doesn't look easy. There's almost certainly a better way.