Andy Green wrote: > It's mainly just it is SDIO device on an SDIO bus that can't use SDIO > interrupt line, and we want it modular.
Hmm, but the SDIO interrupt has nothing to do with rfkill :-) Also, the modular angle comes from how the SDIO stack handles suspend/resume, namely by removing the driver. The underlying mechanism for rmmod/unbind/suspend is pretty much the same. So even if we made the kernel monolithic, we'd still have all the same issues. > If the Atheros module had a kill pin we could drive it would have been > trivial, same as if we had a dedicated switchable regulator (or MOSFET > to control the power anyway) to supply the WLAN module. Yeah, that would have made this quite a bit easier. Now we have a very elaborate way for turning a request to change a single bit into a series of complex transactions that in the end again change a single bit we care a bout (and lots that we don't). - Werner
