Generally speaking we don't. Is there something in particular that has led you to think we do?

We do know a lot about musical instruments (thanks to archaeological, pictorial, and textual evidence), and rhythms within songs (thanks to linguistic evidence). But going much beyond that involves a lot of conjecture.

A few people think there are grounds for inferring more. But on the whole the evidence is very thin, outside very exceptional cases like the Seikilos epitaph (and even there we know nothing about the interpretation, performance, or realisation of the written notes -- nothing about what makes it music, in other words.)