On Thu, Aug 11, 2011 at 09:48:51PM +0200, Henrik Rydberg wrote: > On Wed, Aug 10, 2011 at 10:18:40AM -0700, Guenter Roeck wrote: > > On Tue, 2011-08-09 at 12:14 -0400, Matthew Garrett wrote: > > > The applesmc hardware supports providing interrupts for significant > > > accelerometer events. Add framework for this - we'll want to notify > > > userspace on these in the long run to handle head unloading, but we need > > > to figure out the best interface for that first. Code based on a patch by > > > Nicolas Boichat and the FreeBSD kernel code. > > > > > > Signed-off-by: Matthew Garrett <mjg@xxxxxxxxxx> > > > Cc: Nicolas Boichat <nicolas@xxxxxxxxxx> > > > > Not sure if the overall code makes sense w/o actual use of the > > interrupts, just to fill up the log. I'd like to get some feedback from > > others on this. > > The notion is exciting, but I agree with Guenter, no point adding it > without usecase. What kind of interface should we be presenting? We can't simply use it to send events on the existing input device because that's already defined as a polled input device. > > > + buffer[0] = 0xe0; > > > + buffer[1] = 0xf8; > > > + ret = applesmc_write_key_check(pdev, MOTION_SENSOR_KEY, buffer, 2); > > Lots of buffer setting and calls to a new function - perhaps the > function arguments should be modified? Not without either making it varargs (no) or preventing it from being generalisable. We only use it for a maximum buffer size of 2 right now, but there's nothing in the hardware restricting it to that. > > > + return IRQ_HANDLED; > > > +} > > > + > > Is the list above exhaustive? It's everything I've been able to glean from existing code and what I've been able to get the hardware to do myself. -- Matthew Garrett | mjg59@xxxxxxxxxxxxx _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
