[PATCH v5 1/5] dt-bindings: media: add mediatek ISP3.0 sensor interface
Julien Stephan
jstephan at baylibre.com
Fri Jul 5 02:35:27 PDT 2024
Le ven. 5 juil. 2024 à 11:24, Conor Dooley
<conor.dooley at microchip.com> a écrit :
>
> On Fri, Jul 05, 2024 at 09:50:59AM +0200, Julien Stephan wrote:
> > Le jeu. 4 juil. 2024 à 18:27, Conor Dooley <conor at kernel.org> a écrit :
> > >
> > > On Thu, Jul 04, 2024 at 03:36:40PM +0200, Julien Stephan wrote:
> > > > From: Louis Kuo <louis.kuo at mediatek.com>
> > > >
> > > > This adds the bindings, for the mediatek ISP3.0 SENINF module embedded in
> > > > some Mediatek SoC, such as the mt8365
> > > >
> > > > Signed-off-by: Louis Kuo <louis.kuo at mediatek.com>
> > > > Signed-off-by: Phi-Bang Nguyen <pnguyen at baylibre.com>
> > > > Link: https://lore.kernel.org/r/20230807094940.329165-2-jstephan@baylibre.com
> > > > Signed-off-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> > > > Reviewed-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> > > > Signed-off-by: Julien Stephan <jstephan at baylibre.com>
> > >
> > > I'm really confused by the link tag here. At first glance this looked
> > > like you were sending out something that had been applied by Laurent,
> > > given the Link, Rb and SoB from him. Why does he have a SoB on this
> > > patch? What did Phi-Bang Nguyen do with this patch, and should they have
> > > a Co-developed-by tag?
> >
> > I was not using b4 for the previous revisions of this series, so maybe
> > I messed something up here :(
>
> b4 am has an option to add a link to a patch you apply from the mailing
> list (-l, --add-link) but you should not be using that as a contributor.
> In this case, that link provides no value and is just confusing.
>
> > About Phi-Bang, this series has been in our internal tree for a long
> > time, and Phi-Bang has his SoB on it, so I kept it.
> >
> > About Laurent's tags, they were already on v4. But maybe it was an
> > error ? Should I remove them?
>
> They were also on v1. Did Laurent write part of these bindings, and
> should he have a Co-developed-by?
>
Got it! I understood where I messed up :)
I'll remove the link and add the Co-developed-by tag of Laurent
> > > > +additionalProperties: false
> > > > +
> > > > +if:
> > > > + properties:
> > > > + compatible:
> > > > + contains:
> > > > + const: mediatek,mt8365-seninf
> > >
> > > The binding supports only a single compatible, why is this complexity
> > > required? I don't see other devices being added in this series.
> >
> > Right. The idea is that the number of PHYs depends on the SoC. In the
> > previous revision of the series,
> > the number of PHYs was not fixed, and Krzysztof asked me to fix it by
> > SoC. So I wanted to make it clear
> > that the number of PHYs depends on SoC but maybe I don't need that
> > complexity for that?
> >
> > Is something like the following enough? And if complexity is added
> > later if some other SoC are added?
>
> Yes, that looks reasonable to me. Adding conditional stuff can be done
> iff another soc re-uses the binding.
Will do in the next series.
Thank you for your feedback on this!
Cheers
Julien
>
> Thanks,
> Conor.
>
> > phys:
> > minItems: 2
> > maxItems: 2
> > description:
> > phandle to the PHYs connected to CSI0/A, CSI1, CSI0B
> >
> > phy-names:
> > description:
> > list of PHYs names
> > minItems: 2
> > maxItems: 2
> > items:
> > type: string
> > enum:
> > - csi0
> > - csi1
> > - csi0b
> > uniqueItems: true
More information about the Linux-mediatek
mailing list