Re: ANN: apple motion sensor reverse engineered
Le samedi 27 août 2005 à 17:58 +0200, Johannes Berg a écrit :
> On Fri, 2005-08-26 at 15:13 +0200, Stelian Pop wrote:
> > Done in 0.02 available from my page. Actually I didn't bother to start
> > the thread only when the input device is accessed but made a toggle
> > (sysfs or module parameter based) to actually enable or disable the
> > input device).
> Cool. You *definitely* want to fix/add suspend handling (unless you did
> that in 0.02, which I don't have yet), it was just using 72% CPU time
> used for just hard interrupts (all as BAD in /proc/interrupts though),
> until I shut down my machine (which took few minutes).
I cannot reproduce this here. Suspending with the ams driver loaded
works just fine.
However, I do BAD interrupts in /proc/interrupts, but I have them even
without loading my driver, and they are comming coming at about 30 Hz
(this goes down to 5 Hz after the first suspend/resume cycle). I don't
know if this comes from some recent kernel change or if I toggled
something in the hardware while doing the dev...
Those interrupts don't seem to take too much CPU time, probably because
the frequency is rather low...
Stelian Pop <email@example.com>