I'm trying to produce volumetric light in OpenGL following the implementation details on "GPU Pro 5 Volumetric light effects in KillZone". I am confused on the number of passes needed to create the effect.

So I got the shadow pass which renders the scene from the light's POV, then I have the GBuffer pass which renders to texture the whole scene. and finally a 3rd pass which computes the ray marching on every pixel and computes the amount of accumulated scattering factor according to its distance of the light in the scene (binding the shadow map form the first pass). Then what ? blending these 3 buffers on a full screen quad finally pass ?? or maybe should I do the ray marching in the resulting buffer of blending the shadow map and the Gbuffer?