On Wed, Jan 7, 2009 at 1:40 AM, Simon Kagstrom <[email protected]> wrote: > On Tue, 6 Jan 2009 23:20:55 -0500 > "Nelson Castillo" <[email protected]> wrote: (cut) >> Is anyone working on this? >> Has anything changed since then? > > No, I started looking at it, but my head went buzzing from looking into > the SPI code (layers upon layers!). I guess you are already aware of > this, but for other readers of the thread: It's not really a > functionality issue but something needed to upstream the driver.
Simon, thanks for your reply. Let's see how it goes... If it is too hard I'll try to fail fast :-) > Using the current SPI interface is too slow on the openmoko and also not > really safe since it cannot be operated in interrupt mode. Andys > implementation simply communicates with the accelerometers manually so > everything is fine there. But not for upstream, which tends to > dislike duplicated efforts. > > > The idea we discussed was therefore to add a synchronous SPI transfer > function, which can be called from interrupt context. It was posted > further up in your thread: > > http://www.mail-archive.com/[email protected]/msg06227.html > > the problems which got me dizzy was the layering of SPI, which I never > really got the hang of. Also, if you want to work on this, I've found > the eclipse code navigation to be really, really nice (makes it easy to > lookup all implementations of the interfaces etc). I've written a small > guide for how to set it up for the kernel here: > > http://simonkagstrom.livejournal.com/33093.html Thanks. I've had good luck with cscope so far (cscope -k -R). I might check this out. > The patch with what I got so far is below. It's not much really, the > big function is just a copy of bitbang_work, and I think you can > safely ignore it. I'll read it. I'll ask questions in the list once I have them :-) N.-
