> Report iio motion events to iio subsystem
> 
> Signed-off-by: Sean Nyekjaer <sean@geanix.com>
> ---
> Changes since v4:
>  * Updated bitmask as pr Jonathans comments
> 
> Changes since v5:
>  * None
> 
>  drivers/iio/imu/st_lsm6dsx/st_lsm6dsx.h      |  5 ++
>  drivers/iio/imu/st_lsm6dsx/st_lsm6dsx_core.c | 70 ++++++++++++++++++++
>  2 files changed, 75 insertions(+)
> 

[...]

>  static irqreturn_t st_lsm6dsx_handler_irq(int irq, void *private)
>  {
>  	return IRQ_WAKE_THREAD;
> @@ -1668,6 +1726,18 @@ static irqreturn_t st_lsm6dsx_handler_thread(int irq, void *private)
>  {
>  	struct st_lsm6dsx_hw *hw = private;
>  	int count;
> +	int data, err;
> +
> +	if (hw->enable_event) {


Maybe I understood the issue between the buffered reading and event generation.
I guess it is a race here between when the device is generating the interrupt
and when you set enable_event. I think there are two solutions:
1- trivial one: always read wakeup_src_reg
2- set hw->enable_event as first instruction in st_lsm6dsx_write_event_config()
and roll back in case of error.

Could you please try that changes and double check if you are still able to
trigger the issue?

Regards,
Lorenzo

> +		err = regmap_read(hw->regmap,
> +				  hw->settings->event_settings.wakeup_src_reg,
> +				  &data);
> +		if (err < 0)
> +			return IRQ_NONE;
> +
> +		if (data & hw->settings->event_settings.wakeup_src_status_mask)
> +			st_lsm6dsx_report_motion_event(hw, data);
> +	}
>  
>  	mutex_lock(&hw->fifo_lock);
>  	count = hw->settings->fifo_ops.read_fifo(hw);
> -- 
> 2.23.0
>