Re: acpi-als taking over from hwmon for light sensor support?

[Bastien Nocera <hadess@hadess.net>]

On Wed, 2016-11-23 at 13:32 -0800, Guenter Roeck wrote:
> On Wed, Nov 23, 2016 at 10:07:11PM +0100, Bastien Nocera wrote:
> > On Wed, 2016-11-23 at 18:38 +0100, Jean Delvare wrote:
> > > On mer., 2016-11-23 at 18:21 +0100, Bastien Nocera wrote:
> > > > Hey,
> > > > 
> > > > I did some work on iio-sensor-proxy[1] those past couple of
> > > > days, and
> > > > tested it on a MacBook Pro.
> > > > 
> > > > Before that, I had written a (user-space) driver to access the
> > > > hwmon
> > > > driver's "light" value, so I was a bit surprised when I saw
> > > > that iio-
> > > > sensor-proxy was accessing the light sensor through acpi-als's
> > > > IIO driver
> > > > instead of through the hwmon driver.
> > > > 
> > > > I looked through the hwmon, iio and applesmc.c logs, and
> > > > couldn't find any
> > > > information about how the move from one to the other. Does
> > > > anyone know?
> > > > 
> > > > I'm asking because the applesmc.c hwmon driver is the only one
> > > > with
> > > > support for the "light" property, and if that's not used
> > > > anymore, I might
> > > > as well remove that code now (or schedule to remove it).
> > > 
> > > Light sensors are out of scope of hwmon, they have nothing to do
> > > with
> > > hardware monitoring.
> > 
> > Why is there still code in applesmc.c to handle the "light" sysfs
> > property
> > then?
> > 
> 
> Interesting question. Because no one bothered enough to remove it ?
> Because
> other applications may still use it ? Because old apple hardware may
> not
> support ACPI0008 ? There can be many reasons.

All good reasons. The lack of ACPI0008 support would be the main reason
for me to keep the hwmon driver support in iio-sensor-proxy.

> If the question is why it is there in the first place, please keep in
> mind that
> the driver is quite old. "For historic reasons" may be the best
> answer to that
> question.

Oh, I know it's old, I've used it on 2 different Intel-based MacBook
Airs ;)

> > Also, this doesn't answer my question as to when this switch
> > happened.  I
> > don't care one way or the other, but knowing when the switch was
> > made can
> > allow me to prepare the hwmon driver in iio-sensor-proxy for
> > removal.
> > 
> 
> From kernel perspective, it may have been more or less automatic with
> the addition of acpi-als support to iio. Either case, isn't it up to
> iio-sensor-proxy to decide which kernel ABI it wants to use ?

In this case, we don't have a choice, the device is getting bound to
the acpi-als driver and hwmon just doesn't have the light support. In
any case, iio-sensor-proxy only binds to a single sensor of each type,
so we're good here.

I'll leave the hwmon support for older Intel Macs in.

Cheers