I think a lot of confusion is avoided about this question by bearing in mind two things:
1) The equivalence principle (EP) and arguments based on accelerating frames of reference have a limited range of applicability. That's because the acceleration side of the "equivalence" uses kinematic arguments; whereas the gravitational side deals with a circumstance that is at least stationary, and usually regarded as static. A simple example to illustrate the limited range of the EP is that, in an accelerating rocket, the universe will gradually appear hotter and hotter in the direction of motion. On the surface of a gravitating body no such effect occurs.
2) Owing to the staticness of the Schwarzschild field, the path of a light ray within is often treated as a pure optical problem, the field being treated as a medium of variable refractive index, n = 1 / (1 - 2GM / rc^2)^.5.
For references to the latter approach see the classic book by C. Moller, Relativity (Clarendon, 1972) p. 394; or the more recent paper by Ashby and Bertotti:
http://arxiv.org/abs/0912.2705
In accordance with the optical analogy, the coordinate speed of light depends only on position (slower than c in both radial directions the greater the argument of the coefficient). Whereas the proper speed of light is always c.
