Re: [PATCH 4/6] iio: chemical: ens160: add triggered buffer support

[Christophe JAILLET <christophe.jaillet@wanadoo.fr>]

Le 19/05/2024 à 16:03, Jonathan Cameron a écrit :
> On Mon, 13 May 2024 21:13:07 +0200
> Christophe JAILLET <christophe.jaillet-39ZsbGIQGT5GWvitb5QawA@public.gmane.org> wrote:
> 
>> Le 12/05/2024 à 23:04, Gustavo Silva a écrit :
>>> ENS160 supports a data ready interrupt. Use it in combination with
>>> triggered buffer for continuous data readings.
>>>
>>> Signed-off-by: Gustavo Silva <gustavograzs-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>> ---
>>
>> ...
>>
>>> +static irqreturn_t ens160_trigger_handler(int irq, void *p)
>>> +{
>>> +	struct iio_poll_func *pf = p;
>>> +	struct iio_dev *indio_dev = pf->indio_dev;
>>> +	struct ens160_data *data = iio_priv(indio_dev);
>>> +	__le16 val;
>>> +	int ret, i, j = 0;
>>> +
>>> +	mutex_lock(&data->mutex);
>>> +
>>> +	for_each_set_bit(i, indio_dev->active_scan_mask,
>>> +			 indio_dev->masklength) {
>>> +		ret = regmap_bulk_read(data->regmap,
>>> +				       ENS160_REG_DATA_TVOC + 2 * i, &val, 2U);
>>> +		if (ret)
>>> +			goto err;
>>> +
>>> +		data->scan.chans[j++] = val;
>>
>> Is it safe? How can we know if it has been only *partly* updated? Does
>> it matter to know?
> 
> You've lost me. What do you mean by partly updated?
> This won't push anything to the kfifo etc unless all succeeded.
> Or is there a race with something else in here?

Forget it, I misread the place of iio_push_to_buffers_with_timestamp().

I thought we were going through it when 'goto err'.

CJ

> 
>>
>> CJ
>>
>>> +	}
>>> +
>>> +	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
>>> +					   pf->timestamp);
>>> +err:
>>> +	mutex_unlock(&data->mutex);
>>> +	iio_trigger_notify_done(indio_dev->trig);
>>> +
>>> +	return IRQ_HANDLED;
>>> +}
>>
>> ...
>