This paper considers the fair scheduling problem for dense wireless networks with access point cooperation and multiple-input-multiple-output (MIMO) links. The problem is to maximize the aggregate throughput subject to a fairness constraint that is general enough to capture many different fairness objectives. We formally specify a non-convex optimization problem that captures all aspects of the problem setting, and we propose two algorithms to approximate its solution. The first algorithm jointly optimizes the selection of user sets, MIMO precoders, and assignment of user sets to time slots. The second algorithm separately optimizes first user sets and MIMO precoders and second assignment of user sets to time slots. The first algorithm guarantees perfect fairness and produces a local optimum or a saddle point for aggregate throughput at a fairly high computational cost. The second algorithm also guarantees perfect fairness and produces optimal aggregate throughput for a given set of (possibly non-optimal) user sets while having lower computational complexity. The second algorithm also has a parameter that allows throughput and fairness to be traded off for situations where maximizing throughput is critical and approximate fairness is acceptable. Analyses are complemented by simulation results, which show that: 1) the first algorithm produces significantly higher aggregate throughput than known approaches with a running time that is practical for scenarios with up to 50 users and 2) the second algorithm produces aggregate throughput that is very close to existing heuristics while having significantly lower running time.