Typically, the point of the interview is two-fold: to help the professor to judge whether you are smart and have potential (two separate things), and to help you judge whether you want to spend the next 4+ years working with this guy as your boss. To that extent, you're being judged on your potential, which means you need to show excitement about his research (in that you've read about it and are familiar with it, see below), and lots of enthusiasm. Knowing your stuff is also helpful so you don't look like an idiot.
In most cases, you'll discuss your undergrad work, you'll be asked about your familiarity with the professor's research (basic familiarity goes a long way here), and you'll be told about his current research work. Different professors have wildly different interview styles, so it can be anything from conversational to almost exam-like.
The best way to prepare would be exactly as you suggest; read two or three papers from that professor. I would recommend choosing one with a recent publication date (i.e., within the past 12 months) and the one he has with the most citations. Different search tools allow you to find that for different fields; for example, hubmed allows that for biology/medicine fields.