Eric Piel wrote: >> I would recommend converting the input piece to input_polled_device >> instead of managing the polling thread by yourselves. > Yes, you are right, I had forgotten about this one! That said, for now > I'm working on removing the polling and relying only the interrupts. So > if it works that shouldn't be needed :-) > I think going to interrupt-only approach will cause a problem in the HTC Shift: In CTRL_REG2 (21h) if you set to 1 the Interrupt ENable (IEN) field the device shuts down. I guess HTC took the "polling" approach in the windows driver, and they enable the interrupt when the sensor reports a "free fall" event, to have the disk parked and protect the hardware from being damaged. Maybe you can keep the input_polled_device together with interrupts and select one or the other depending on DMI information (same as for swaping the axis, etc.) Cheers, Pau Oliva
