How do concurrency models differ between traditional programming languages and functional programming languages?

Understanding concurrency, the ability for multiple processes to occur simultaneously, is crucial in modern software development. Traditional programming languages like Java and C++ often implement concurrency using threads and locks, which can lead to complexities such as race conditions and deadlocks. On the other hand, functional programming languages like Erlang and Haskell promote immutability and use concurrency models that inherently avoid these issues through message passing or software transactional memory. What are the fundamental differences in how these languages approach concurrency, and how do these differences impact performance, scalability, and ease of writing concurrent programs?