Re: [PATCH v4 1/3] media: dt-bindings: add rockchip mipi csi-2 receiver

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

Hi Heiko,

On Fri, Jan 16, 2026 at 12:07:22PM +0100, Heiko Stübner wrote:
> Am Freitag, 16. Januar 2026, 11:11:26 Mitteleuropäische Normalzeit schrieb Sakari Ailus:
> > Hi Michael,
> > 
> > On Thu, Jan 15, 2026 at 07:26:07PM +0100, Michael Riesch via B4 Relay wrote:
> > > From: Michael Riesch <michael.riesch@collabora.com>
> > > 
> > > Add documentation for the Rockchip MIPI CSI-2 Receiver.
> > > 
> > > Signed-off-by: Michael Riesch <michael.riesch@wolfvision.net>
> > > Signed-off-by: Michael Riesch <michael.riesch@collabora.com>
> > > ---
> > >  .../bindings/media/rockchip,rk3568-mipi-csi2.yaml  | 141 +++++++++++++++++++++
> > >  MAINTAINERS                                        |   6 +
> > >  2 files changed, 147 insertions(+)
> > > 
> > > diff --git a/Documentation/devicetree/bindings/media/rockchip,rk3568-mipi-csi2.yaml b/Documentation/devicetree/bindings/media/rockchip,rk3568-mipi-csi2.yaml
> > > new file mode 100644
> > > index 000000000000..2c2bd87582eb
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/media/rockchip,rk3568-mipi-csi2.yaml
> > 
> > I'd add a compatible string for the base IP block and name it accordingly.
> 
> personally, I wouldn't do that.
> 
> While the RK3568-variant is a somewhat smooth integration ... the now
> (temporarily) omitted RK3588 variant of a similar block is not.
> 
> I.e. the RK3588 has quite a number of those CSI hosts, with a bunch of
> resource routing bits and bops between those CSI hosts sitting in the
> Rockchip "Gernal Register Files" (dumping ground for random bits and bops).
> 
> So you then get a syscon accessing per-soc registers and bits.
> 
> So while it is (compatible to) some Synopsis IP block, the integration to
> make that thing actually do something is highly soc-specific.
> 
> That's also why the for example the dw-hdmi/dsi IPs don't use a common
> compatible [0] [1] [2] [3] [4]
> 
> 
> [0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/display/amlogic,meson-dw-hdmi.yaml
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/display/allwinner,sun8i-a83t-dw-hdmi.yaml
> [2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/display/hisilicon/dw-dsi.txt
> [3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/display/rockchip/rockchip,dw-hdmi.yaml
> [4] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/display/rockchip/rockchip,dw-mipi-dsi.yaml

Ack, thanks for the explanation. This sounds reasonable to me; this can be
later extended or renamed if needed.

> 
> 
> > > +    soc {
> > > +        interrupt-parent = <&gic>;
> > > +        #address-cells = <2>;
> > > +        #size-cells = <2>;
> > > +
> > > +        csi: csi@fdfb0000 {
> > > +            compatible = "rockchip,rk3568-mipi-csi2";
> > 
> > This would become e.g.
> > 
> >             compatible = "rockchip,rk3568-mipi-csi2", "snps,dw-mipi-csi2rx";
> > 
> > See my comments on the driver patch as well.
> 
> In the PCIe area, we have  rockchip,rk3568-pcie and rockchip,rk3568-pcie-ep
> for a similar combo.
> 
> For CSI the receiver is the vastly more common thing to do. So if anything,
> I'd go with "foo-csi2" vs. "foo-csi2-device", if somebody really develops a
> "camera" SoC, with a fully featured DT-based OS in the future ;-) .

There may be surprises, CSI-2 transmitters may be used in other kinds to
devices than cameras. Therefore I'd add the "rx" part in the name, at the
very least when it comes to the Synopsys IP block. Some of the other
receivers are named that way already and we have a transmitter driver, too.

-- 
Regards,

Sakari Ailus