As tman mentioned, you're looking at the wrong device. Check out empeg_cs4231a.[c|h]

The empeg uses the SAA77h05 DSP for it's output. The DSP unfortunately doesn't allow for analogue input signals to be passed as a digital representation to the CPU. So although the Line In and Radio inputs are digitised, manipulated, and mixed by the DSP, there is no way (for example) to be able to do visuals on those inputs. So the empeg has a second audio chip the CS4231a in order to do that. When you switch to the Tuner or the Aux In, the CS chip ADCs sample the inputs so that visuals can be used.
Note that this doesn't happen for MP3 playback ('PCM' mode in the DSP). We already have the digital representation needed for the visuals.
What isn't clear to me (I haven't really looked) is how to ensure that the microphone input is played. The CS4231a has a limited sampling rate of 22.05KHz. The DSP is locked down to 44.1KHz. That means that you cannot directly sample the microphone to memory and then play it as is. You would have to upsample it to 44.1KHz first. This can be done the dirty way by doubling every sample. Then you'd need to play the sample using the DSP in 'PCM' mode, ie in MP3 mode.
This could make full duplex rather difficult - you would be relying on the CPU to maintain real-time processing of the samples.