Re: [PATCH v3 1/2] iio: imu: inv_icm42600: add WoM support

[Jonathan Cameron <jic23@kernel.org>]

On Fri, 18 Apr 2025 18:19:02 +0200
Jean-Baptiste Maneyrol via B4 Relay <devnull+jean-baptiste.maneyrol.tdk.com@kernel.org> wrote:

> From: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com>
> 
> Add WoM as accel roc rising x|y|z event.
> 
> Signed-off-by: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com>
Hi Jean-Baptiste,

One thing on mixing gotos and scoped_guard().  It might be fine but we've
had weird issues with compilers and this stuff + the guidance in cleanup.h
suggests not mixing the two approaches.

Easy enough to sort out here with a helper function and I think the
end result will both avoid that issue and be easier to read.

Jonathan

> ---
>  drivers/iio/imu/inv_icm42600/inv_icm42600.h        |  54 +++-
>  drivers/iio/imu/inv_icm42600/inv_icm42600_accel.c  | 279 ++++++++++++++++++++-
>  drivers/iio/imu/inv_icm42600/inv_icm42600_buffer.c |   2 +-
>  drivers/iio/imu/inv_icm42600/inv_icm42600_core.c   |  58 +++++
>  4 files changed, 385 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/iio/imu/inv_icm42600/inv_icm42600.h b/drivers/iio/imu/inv_icm42600/inv_icm42600.h
> index f893dbe6996506a33eb5d3be47e6765a923665c9..bcf588a048836f909c26908f0677833303a94ef9 100644
> --- a/drivers/iio/imu/inv_icm42600/inv_icm42600.h
> +++ b/drivers/iio/imu/inv_icm42600/inv_icm42600.h
> @@ -135,6 +135,14 @@ struct inv_icm42600_suspended {
>  	bool temp;
>  };
>  
> +struct inv_icm42600_apex {
> +	unsigned int on;
> +	struct {
> +		uint64_t value;
> +		bool enable;
> +	} wom;
> +};
> +
>  /**
>   *  struct inv_icm42600_state - driver state variables
>   *  @lock:		lock for serializing multiple registers access.
> @@ -148,9 +156,10 @@ struct inv_icm42600_suspended {
>   *  @suspended:		suspended sensors configuration.
>   *  @indio_gyro:	gyroscope IIO device.
>   *  @indio_accel:	accelerometer IIO device.
> - *  @buffer:		data transfer buffer aligned for DMA.
> - *  @fifo:		FIFO management structure.
>   *  @timestamp:		interrupt timestamps.
> + *  @apex:		APEX features management.

Maybe give a little more info on what APEX is somewhere?



> +static int inv_icm42600_accel_enable_wom(struct iio_dev *indio_dev)
> +{
> +	struct inv_icm42600_state *st = iio_device_get_drvdata(indio_dev);
> +	struct inv_icm42600_sensor_state *accel_st = iio_priv(indio_dev);
> +	struct device *pdev = regmap_get_device(st->map);
> +	struct inv_icm42600_sensor_conf conf = INV_ICM42600_SENSOR_CONF_INIT;
> +	unsigned int sleep_ms = 0;
> +	int ret;
> +
> +	ret = pm_runtime_resume_and_get(pdev);
> +	if (ret)
> +		return ret;
> +
> +	scoped_guard(mutex, &st->lock) {
> +		/* turn on accel sensor */
> +		conf.mode = accel_st->power_mode;
> +		conf.filter = accel_st->filter;
> +		ret = inv_icm42600_set_accel_conf(st, &conf, &sleep_ms);
> +		if (ret)
> +			goto error_suspend;

As below.  Compilers are not great at the more complex scope vs goto stuff.
This one may be fine buf in general we avoid it.

> +	}
> +
> +	if (sleep_ms)
> +		msleep(sleep_ms);
> +
> +	scoped_guard(mutex, &st->lock) {
> +		ret = inv_icm42600_enable_wom(st);
> +		if (ret)
> +			goto error_suspend;

This doesn't follow the guidance in cleanup.h about never mixing gotos and
scoped cleanup. Two options here, either factor out everthing after the
pm handling and have
	ret = pm_runtime_resume_and_get(pdev);
	if (ret)
		return ret;

	ret = __inv_icm62600_accel_enabled_wom();
	if (ret) {
		pm_runtime_mark_last_busy(pdev);
		pm_runtime_put_autosuspend(pdev)'
		return ret;
	}

	return 0;

The rest of the cases are fine.

> +		st->apex.on++;
> +		st->apex.wom.enable = true;
> +	}
> +
> +	return 0;
> +
> +error_suspend:
> +	pm_runtime_mark_last_busy(pdev);
> +	pm_runtime_put_autosuspend(pdev);
> +	return ret;
> +}