Here's a quick update on how I think we can make WLAN work with SPI instead of SDIO without too much pain:
The basic idea is to write a new host controller driver for the Atheros SDIO stack, which then uses the regular Linux SPI interface and the existing 24xx SPI driver. This should work as a proof of concept, but it will be slow, due to the SPI driver not using DMA, due to the SPI driver not being able to work from an interrupt context, and I would expect general protocol stack overhead to take its toll as well. As a first improvement, here's a DMA-capable SPI driver we should be able to use: http://www.mail-archive.com/spi-devel-general%40lists.sourceforge.net/msg00505.html Ben Dooks already assured me that SPI with DMA has already been seen to work in real life, so the 2442 hardware won't surprise us there. Further optimization would be a bit harder, but we're probably in a better position to assess how to proceed there once the first steps have been taken. Andy, you mentioned that you ran into layering violations in an earlier attempt to make WLAN work with an SPI interface. Were they of a kind that I'd also encounter with the approach outlined above ? - Werner
