Re: [PATCH v2 1/1] counter: cros_ec: Add synchronization sensor

[Jonathan Cameron]

On Sun, 10 Nov 2019 10:14:08 -0500
William Breathitt Gray <vilhelm.gray@xxxxxxxxx> wrote:

> On Tue, Sep 24, 2019 at 04:20:51PM +0200, Fabien Lahoudere wrote:
> > Hi all,
> > 
> > After some discussions and investigation, the timestamp is very
> > important for that sync driver.
> > Google team uses that timestamp to compare with gyroscope timestamp.
> > 
> > So the important data is timestamp and counter value is useless.
> > Just the event of counter increment is important to get a timestamp.
> > 
> > In that case, my idea was to just use an IIO driver with a single
> > channel with IIO_TIMESTAMP. We discuss this here and it seems
> > controversial.
> > 
> > So my question to Jonathan is if we have a timestamp coming from the EC
> > itself, can we consider this timestamp as a good IIO driver?
> > 
> > Any other idea is welcome, however Google team would like to manage
> > only IIO drivers if possible.
> > 
> > Thanks  
> 
> Jonathan,
> 
> Should the the timestamp from the EC be introduced as an IIO driver
> using IIO_TIMESTAMP?

It is is a rather odd driver but I suppose it would be fine with lots
of clear docs on why it is how it is...

> 
> Since there is no corresponding EC Counter driver in the baseline right
> now we don't have a conflict yet. If the EC timestamp is introduced as
> an IIO driver then we should make any future EC Counter driver mutually
> exclusive with the IIO driver in order to prevent any memory space
> conflict. At that point we may deprecate the IIO driver and move the
> timestamp functionality to the corresponding Counter driver.

That route does become somewhat of a mess so I suspect we'd have to have
a single driver supporting both userspace interfaces.  If you are happy
that we'd be adding a bit of legacy to support for ever then we can go
that way.

> 
> That's assuming someone is interested in the Count component enough to
> implement an EC Counter driver; otherwise, the IIO driver will serve
> just fine if timestamp is the only data desired from this device.
> 
> William Breathitt Gray