Re: [PATCH v3 8/8] iio: adc: Add events support to ad4052

[Jonathan Cameron <jic23@kernel.org>]

> > > +
> > > +static int ad4052_read_event_value(struct iio_dev *indio_dev,
> > > +				   const struct iio_chan_spec *chan,
> > > +				   enum iio_event_type type,
> > > +				   enum iio_event_direction dir,
> > > +				   enum iio_event_info info, int *val,
> > > +				   int *val2)
> > > +{
> > > +	struct ad4052_state *st = iio_priv(indio_dev);
> > > +	int ret;
> > > +
> > > +	if (!iio_device_claim_direct(indio_dev))
> > > +		return -EBUSY;
> > > +
> > > +	if (st->wait_event) {
> > > +		ret = -EBUSY;
> > > +		goto out_release;  
> >   
> 
> Below are two distinct options with different implications.
> > Not being able to read event parameters whilst monitoring them seems
> > very restrictive.  Can't we cache the values?  Either play games to ensure
> > we get them from the regmap cache or just cache these few values in st.
> > 
> > Checking what you are monitoring for feels like the sort of thing
> > userspace might well do.  
> 
> (1)
> I agree, I can investigate regcache_cache_only and the other cache
> options to achieve this. If I come to the conclusion it is not possible,
> storing into st will achieve the same.
> 
> > 
> > Even blocking changing the monitoring parameters is unusually strict.
> > Why not just drop out of monitor mode, update them and go back in?
> >   
> (2)
> The core point of the blocking behaviour is to not have hidden downtimes
> in the monitoring for the user. An early driver used to do what you
> describe and it was a design decision.
> 
> Since a custom regmap_bus was necessary to restrict the regmap access
> speed (ADC access is faster), bringing back this by behavior embedding
> it in the custom regmap now seems plausible, with proper explanation in
> the rst page. This should fully dismiss the st->wait_event -> -EBUSY.
> 
> Considering (1) and (2), what is the preferred approach?

Key here is that the user made the choice to change the parameters.
Most of the time they won't choose to do that, but if they do then
that's what they want to do. Why make them turn the monitoring off,
change value and turn it on again if we can support it reasonably
cleanly.  In many devices there is no interruption to monitoring so
we may well have userspace code written against assumption it
can just update this stuff without that dance.  So prefer (2)
but (1) is better than nothing if (2) proves too complex.

J
> 
> Regards,
> Jorge
> > > +	}
> > > +
> > > +	switch (info) {
> > > +	case IIO_EV_INFO_VALUE:
> > > +		ret = __ad4052_read_event_info_value(st, dir, val);
> > > +		break;
> > > +	case IIO_EV_INFO_HYSTERESIS:
> > > +		ret = __ad4052_read_event_info_hysteresis(st, dir, val);
> > > +		break;
> > > +	default:
> > > +		ret = -EINVAL;
> > > +		break;
> > > +	}
> > > +
> > > +out_release:
> > > +	iio_device_release_direct(indio_dev);
> > > +	return ret ? ret : IIO_VAL_INT;
> > > +}