ASIO SDK Audio Processing Development

I’m currently developing some Windows audio processing software and I have some questions about the ASIO callbacks. So in the bufferSwitch callback (or its counterpart), which buffers am I actually processing. From the SDK it seems like I should “grab” the data from the input buffer, process it, then write it to the output buffer, however, from looking at the host sample provided in the SDK, it appears that I only process the output buffers. Which is true? Or does it actually matter.