Surface texture, such as in rough or complex walls, can greatly alter the dynamics of near-wall turbulence and result for instance in a significant change in skin-friction. This is therefore a subject of significant practical interest, and a substantial amount of work has been devoted to understanding, modelling and predicting the effect of such surfaces. These flows can be resolved using direct simulation, but the resolution requirements are typically orders of magnitude higher than for smooth walls, so it becomes convenient to model the surface rather than resolving it. This talk will discuss two key mechanisms through which the texture affects the flow, illustrating them with examples from porous-rough and superhydrophobic surfaces. The first mechanism is through an equivalent set of boundary conditions. Homogenisation is the typical technique used to derive these, but its usefulness is limited in turbulence, where the scales of the texture and the smallest eddies in the flow are comparable, and where the assumption that the flow in the vicinity of each surface asperity is Stokesian breaks down. We propose instead a multiscale approach that does not require scale separation, and show that the result of such model agrees well with DNS data. The second mechanism is through the non-linear coupling of the texture-coherent flow and the overlying, texture-incoherent turbulence, which acts on the latter as a forcing term on the momentum equations. We will show that this can be re-introduced into simulations that do not resolve the texture, and that the results are in good agreement with fully resolved simulations.