Re: [PATCH v2 1/2] media: dt-bindings: media: i2c: Add MT9M114 camera sensor binding

[Sakari Ailus <sakari.ailus@linux.intel.com>]

Hi Laurent,

On Tue, Feb 21, 2023 at 01:52:46AM +0200, Laurent Pinchart wrote:
> > > > > > +        properties:
> > > > > > +          bus-type:
> > > > > > +            enum: [4, 5, 6]
> > > > > 
> > > > > With bus-type 5, shouldn't you have the parallel interface sync signal
> > > > > polarity properties? Possibly also others if the hardware supports them.
> > > > 
> > > > As far as I can tell, the hardware has fixed polarities for all signals.
> > > > Both hsync and vsync (called LINE_VALID and FRAME_VALID here) are active
> > > > high.
> > > 
> > > Right, then you won't need these.
> > > 
> > > What about the link-frequencies property?
> > 
> > That's something I've meant to ask, should the link-frequencies property
> > be specified for parallel buses too, or only CSI-2 ?
> 
> Any feedback on this ?

Yes.

It very likely matters there as well but probably less so than on CSI-2.

link-frequencies isn't specific to CSI-2 either.

> 
> > We also need to document how a DT writer should select the link
> > frequencies. The process is complicated, saying that the property is
> > required without explaining how the values can be computed (EMC
> > constraints aside, just from a functional point of view) isn't nice. If
> > you explain this to me, I can put it in a patch :-)
> 
> Or on this ? :-)

Generally the hardware people will tell you this when it matters. :-) But
on a development board you can probably have anything the hardware
supports.

I guess we could add a few words on this to e.g.
Documentation/driver-api/media/camera-sensor.rst .

Most drivers will support a few frequencies only and sometimes the vendor,
due to someone needing it, provides another configuration with a different
frequency. Many sensors also have datasheets that are precise enough to
calculate this yourself.

-- 
Regards,

Sakari Ailus