A large class of offline and online optimal control algorithms requires the solution of a sparse structured quadratic programming problems at each iteration. A variety of approaches has been proposed for this problem class, including first-order methods, interior-point algorithms, and condensing-based active-set algorithms. We propose a novel algorithm based on a hybrid active-set/Newton-type strategy that aims at combining sparsity exploitation features of an interior point method with warm-starting capabilities of an active-set method. Moreover, the proposed algorithm is parallelizable to a large extend. We address algorithmic details of this strategy and present the open-source implementation qpDUNES. The performance of the solver is evaluated on basis of several problems from the area of linear and nonlinear model-predictive control, showing significant performance improvements over existing software packages for this class of structured quadratic programming problems.