Re: [PATCH 5/7] iio: adc: ad4062: Add IIO Trigger support

[Jonathan Cameron <jic23@kernel.org>]

On Mon, 13 Oct 2025 09:28:03 +0200
Jorge Marques <jorge.marques@analog.com> wrote:

> Adds support for IIO Trigger. Optionally, gp1 is assigned as Data Ready
> signal, if not present, fallback to an I3C IBI with the same role.
> The software trigger is allocated by the device, but must be attached by
> the user before enabling the buffer. The purpose is to not impede
> removing the driver due to the increased reference count when
> iio_trigger_set_immutable or iio_trigger_get is used.
> 
> Signed-off-by: Jorge Marques <jorge.marques@analog.com>

A few things inline.
Thanks,

> diff --git a/drivers/iio/adc/ad4062.c b/drivers/iio/adc/ad4062.c
> index e55a69c62694a71a4e29f29b9a2bfeec3b16c990..40b7c10b8ce7145b010bb11e8e4698baacb6b3d3 100644
> --- a/drivers/iio/adc/ad4062.c
> +++ b/drivers/iio/adc/ad4062.c

> +static irqreturn_t ad4062_poll_handler(int irq, void *p)
> +{
> +	struct iio_poll_func *pf = p;
> +	struct iio_dev *indio_dev = pf->indio_dev;
> +	struct ad4062_state *st = iio_priv(indio_dev);
> +	u8 addr = AD4062_REG_CONV_TRIGGER;
> +	int ret;
> +
> +	/* Read current and trigger next conversion */
> +	struct i3c_priv_xfer t[2] = {
> +		{
> +			.data.in = &st->raw,

If it is safe to use addr on the stack, do we need to have
a dma safe buffer for raw?  I'm not sure for i3c!

> +			.len = 4,
> +			.rnw = true,
> +		},
> +		{
> +			.data.out = &addr,
> +			.len = 1,
> +			.rnw = false,
> +		}
> +	};
> +
> +	/* Separated transfers to not infeer repeated-start */
> +	ret = i3c_device_do_priv_xfers(st->i3cdev, &t[0], 1);
> +	if (ret)
> +		goto out;
> +	ret = i3c_device_do_priv_xfers(st->i3cdev, &t[1], 1);

Add a comment on this. I assume it's setting things up for the next
scan?

> +	if (ret)
> +		goto out;
> +
> +	iio_push_to_buffers_with_timestamp(indio_dev, &st->raw,
> +					   pf->timestamp);
> +
> +out:
> +	iio_trigger_notify_done(indio_dev->trig);
> +
> +	return IRQ_HANDLED;
>  }

> +
> +static int ad4062_triggered_buffer_predisable(struct iio_dev *indio_dev)
> +{
> +	struct ad4062_state *st = iio_priv(indio_dev);
> +
> +	pm_runtime_mark_last_busy(&st->i3cdev->dev);

Take a look at the changes across the tree recently.
Now pm_runtime_put_autosuspend() calls pm_runtime_mark_last_busy()
internally to avoid the need for this pair.

> +	pm_runtime_put_autosuspend(&st->i3cdev->dev);
> +	return 0;
> +}