Re: IIO Driver for TMP117 Temperature sensor

[Jonathan Cameron <jic23@kernel.org>]

On Sun, 13 Dec 2020 08:08:26 -0800
Guenter Roeck <linux@roeck-us.net> wrote:

> On 12/13/20 7:12 AM, Jonathan Cameron wrote:
> > On Wed, 9 Dec 2020 11:48:40 -0800
> > Guenter Roeck <linux@roeck-us.net> wrote:
> >   
> >> On 12/9/20 12:11 AM, Alexandru Ardelean wrote:  
> >>> On Tue, Dec 8, 2020 at 6:10 PM Puranjay Mohan <puranjay12@gmail.com> wrote:    
> >>>>
> >>>> I have this TI's TMP117 sensor with me and I was thinking about writing an
> >>>> IIO driver for it as a hobby project. Is the IIO subsystem the correct
> >>>> place for this driver? if yes, can someone help me get started with this,
> >>>> I haven't written an IIO driver before. I have this sensor and also a
> >>>> raspberry pi with me for testing.    
> >>>
> >>> This could also fit into drivers/hwmon.
> >>> Looking at the HWMON subsystem there are more TMP drivers there
> >>> (TMP102/103/108/401/513).
> >>> The first 3 seem a bit more similar to TMP117 (in terms of register map).
> >>>     
> >>
> >> It would probably be better suited for hwmon (it has limit registers,
> >> suggesting a common use as hardware monitoring device).  
> > It is a curious part.  I suspect TI based their design for a medical grade
> > digital thermometer chip on an existing hwmon part.
> > 
> > The limit registers are very simple so could be supported by IIO.
> > This sits somewhere in the middle of high end thermocouple chips which
> > tend to be in IIO and typically lower accuracy / range hwmon parts.
> > 
> > It's in the fuzzy borderline region so I doubt anyone would raise strong
> > objections to which subsystem it was in.  Guenter has fallen on the
> > hwmon side of things and I'm fine with that.
> >   
> 
> On the other side, it turns out that there is already tmp107 support
> in iio, and tmp107 is pretty much the spi equivalent of the same chip.
> So it really depends on the use case. If the user wants to use the iio
> subsystem, I am fine with it. We just need to remind people that this
> implies no or only limited hwmon support.
> 
> [ I really need to spend the time to write a hwmon->iio bridge.
>   The iio->hwmon bridge is a bit limited - I have not been able to
>   figure out how to support limit registers (or event values)
>   and events, and I don't think it is possible. ]

So far IIO doesn't have an in kernel consumer interface for
events. It shouldn't be that hard to add one though and it 
has been on the todo list for a very long time.  We've discussed
it a few times and concluded that there are some short cuts such
as sending all events to all consumers and relying on the receiver
to do any necessary filtering. It's a bit messy but it makes for
much simpler core code.

Maybe I'll get bored enough over xmas to look at it...

Jonathan

> 
> Guenter
> 
> > Jonathan
> >   
> >>  
> >>> Let's see what others have to add.
> >>> But, all-in-all whatever driver you end up writing, the easiest method
> >>> is to copy an existing similar driver and extend it.
> >>> Sometimes, a part can be added to an existing driver.
> >>> At a quick scan through existing drivers, it doesn't look like TMP117
> >>> is similar to existing drivers, so it may require a new driver
> >>> altogether.    
> >>
> >> I don't see an immediate match either, but the tmp102 hwmon driver
> >> might be a good start.
> >>
> >> Guenter
> >>  
> >>> I may have missed something though.
> >>>
> >>> Thanks
> >>> Alex
> >>>     
> >>>>
> >>>> --
> >>>> Thanks and Regards
> >>>>
> >>>> Yours Truly,
> >>>>
> >>>> Puranjay Mohan    
> >>  
> >   
>