Hi Kieran,
Am Donnerstag, 30. November 2023, 15:02:36 CET schrieb Kieran Bingham:
> + Pantellis
>
> Quoting Adam Ford (2023-11-30 13:48:58)
>
> > On Thu, Nov 30, 2023 at 3:51 AM Alexander Stein
> >
> > <alexander.stein@xxxxxxxxxxxxxxx> wrote:
> > > Hi Laurent,
> > >
> > > Am Mittwoch, 29. November 2023, 16:16:37 CET schrieb Laurent Pinchart:
> > > > Hi Alexander,
> > > >
> > > > On Wed, Nov 29, 2023 at 11:20:07AM +0100, Alexander Stein wrote:
> > > > > Am Mittwoch, 29. November 2023, 10:31:13 CET schrieb Paul Elder:
> > > > > > From: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > > > > >
> > > > > > Add two overlay to enable each ISP instance. The ISP is wired
> > > > > > directly
> > > > > > to the CSIS for now, bypassing the ISI completely.
> > > > >
> > > > > I'm not sure if this is worth adding in a separate overlay.
> > > >
> > > > The trouble is that, at this point, selection between the ISP and the
> > > > ISI can only be performed through DT :-S That's why this is
> > > > implemented
> > > > as an overlay.
> > >
> > > I feel a better place would be the overlay which actually adds the
> > > sensor.
> > > This knows best whether ISI or ISP should be used.
> > >
> > > > > > Signed-off-by: Laurent Pinchart
> > > > > > <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > > > > > Signed-off-by: Paul Elder <paul.elder@xxxxxxxxxxxxxxxx>
> > > > > > ---
> > > > > >
> > > > > > arch/arm64/boot/dts/freescale/Makefile | 2 ++
> > > > > > .../arm64/boot/dts/freescale/imx8mp-isp1.dtso | 36
> > > > > > +++++++++++++++++++
> > > > > > .../arm64/boot/dts/freescale/imx8mp-isp2.dtso | 36
> > > > > > +++++++++++++++++++
> > > > > > 3 files changed, 74 insertions(+)
> > > > > > create mode 100644 arch/arm64/boot/dts/freescale/imx8mp-isp1.dtso
> > > > > > create mode 100644 arch/arm64/boot/dts/freescale/imx8mp-isp2.dtso
> > > > > >
> > > > > > diff --git a/arch/arm64/boot/dts/freescale/Makefile
> > > > > > b/arch/arm64/boot/dts/freescale/Makefile index
> > > > > > 300049037eb0..f97dfac11189
> > > > > > 100644
> > > > > > --- a/arch/arm64/boot/dts/freescale/Makefile
> > > > > > +++ b/arch/arm64/boot/dts/freescale/Makefile
> > > > > > @@ -113,6 +113,8 @@ dtb-$(CONFIG_ARCH_MXC) +=
> > > > > > imx8mp-dhcom-pdk2.dtb
> > > > > >
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-dhcom-pdk3.dtb
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-evk.dtb
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-icore-mx8mp-edimm2.2.dtb
> > > > > >
> > > > > > +dtb-$(CONFIG_ARCH_MXC) += imx8mp-isp1.dtbo
> > > > > > +dtb-$(CONFIG_ARCH_MXC) += imx8mp-isp2.dtbo
> > > > > >
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-msc-sm2s-ep1.dtb
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-phyboard-pollux-rdk.dtb
> > > > > > dtb-$(CONFIG_ARCH_MXC) += imx8mp-tqma8mpql-mba8mpxl.dtb
> > > > > >
> > > > > > diff --git a/arch/arm64/boot/dts/freescale/imx8mp-isp1.dtso
> > > > > > b/arch/arm64/boot/dts/freescale/imx8mp-isp1.dtso new file mode
> > > > > > 100644
> > > > > > index 000000000000..cf394ed224ab
> > > > > > --- /dev/null
> > > > > > +++ b/arch/arm64/boot/dts/freescale/imx8mp-isp1.dtso
> > > > > > @@ -0,0 +1,36 @@
> > > > > > +// SPDX-License-Identifier: (GPL-2.0+ OR MIT)
> > > > > > +/*
> > > > > > + * Copyright 2022 Ideas on Board Oy
> > > > > > + */
> > > > > > +
> > > > > > +/dts-v1/;
> > > > > > +/plugin/;
> > > > > > +
> > > > > > +#include <dt-bindings/media/video-interfaces.h>
> > > > > > +
> > > > > > +&isi_0 {
> > > > > > + status = "disabled";
> > > > >
> > > > > ISI is disabled by default. What is your intention here?
> > > >
> > > > It could be enabled by an overlay for a camera module. Ideally we want
> > > > to be able to enable both the ISI and ISP at runtime, but that's not
> > > > possible yet and will require a very large amount of work.
> > >
> > > Again IMHO this is part of sensor setup, in a very specific overlay. To
> > > put it into different words: I barely see the gain of this small
> > > overlay.
> > >
> > > Runtime switching would require a combined media controller including
> > > both ISI and ISP, no?
> > >
> > > Best regards,
> > > Alexander
> > >
> > > > > > +
> > > > > > + ports {
> > > > > > + port@0 {
> > > > > > + /delete-node/ endpoint;
> > > > >
> > > > > This doesn't work in overlays. See [1]. Otherwise the OF graph
> > > > > connections
> > > > > look fine to me. I'm using the same in my local overlay.
> > > >
> > > > Interesting, I wasn't aware of that. Maybe we should fix it :-)
> >
> > When I did my camera implementation, I thought it was simpler to:
> >
> > /delete-node/ &isi_in_0;
> >
> > it's a one-line change.
> >
> > I would suggest we just drop the overlay and make users who have the
> > cameras integrate the cameras and the isp routing into their
> > respective overlays.
>
> I use these to factor out common parts between multiple cameras that can
> be connected to multiple ports.
>
> I can connect any of (Physically available to me right now)
> IMX219, IMX477, IMX708, GC2145, OV5640(7?) IMX335, IMX283, IMX519,
> Arducam64MP
>
> to either of:
>
> Debix-SOM-A Port CSI-1
> Debix-SOM-A Port CSI-2
>
> And I can connect those same cameras to two ports of a Pi5. So now
> that's 27 overlays to manage the 9 cameras I have /on my desk/ to
> connect to this board.
>
> Uh Oh - sorry I can also connect them to a Debix Model A ... oh no ... I
> need to stop thinking about what I can connect them to. I have rockchip
> boards they'll work on too!
>
> This explosion of overlays could be ... hard to manage. With /a lot/ of
> repetition of the same data.
Maybe I am missing something, but how can the intermediate overlay reduce this
amount of overlays? At least the remote-endpoint of the sensor needs the CSI
endpoint, which is different for CSI1/CSI2, and vice versa.
I do not have rockchip hardware, but I would not assume the CSI endpoint to
not have the same label, so the phandle reference is different there as well.
>
> I'm not opposed to dropping these intermediate helper overlays, but I'd
> be interested to know if anyone has ideas on how we could define
> 'connectors' and then abstract the cameras / overlays that can be moved
> between different compatible ports.
You still would have the naming problem. How do you identify a connector? It's
similar to label naming on device tree nodes.
Best regards,
Alexander
> The [RFC 0/3] Portable Device Tree Connector [0] might be interesting to
> resurrect. Did that go anywhere?
>
> [0]
> https://lore.kernel.org/all/1464986273-12039-1-git-send-email-pantelis.anto
> niou@xxxxxxxxxxxx/
>
> --
> Kieran
>
> > adam
> >
> > > > > [1]
> > > > > https://lore.kernel.org/all/CAMuHMdWu4KZbBkvEofUV2wuA1g2S=XHHM3RUN1c
> > > > > NrcZB
> > > > > khsPZA@xxxxxxxxxxxxxx/>
> > > > >
> > > > > > + };
> > > > > > + };
> > > > > > +};
> > > > > > +
> > > > > > +&isp_0 {
> > > > > > + status = "okay";
> > > > > > +
> > > > > > + ports {
> > > > > > + port@1 {
> > > > > > + isp0_in: endpoint {
> > > > > > + bus-type = <MEDIA_BUS_TYPE_PARALLEL>;
> > > > > > + remote-endpoint = <&mipi_csi_0_out>;
> > > > > > + };
> > > > > > + };
> > > > > > + };
> > > > > > +};
> > > > > > +
> > > > > > +&mipi_csi_0_out {
> > > > > > + remote-endpoint = <&isp0_in>;
> > > > > > +};
> > > > > > diff --git a/arch/arm64/boot/dts/freescale/imx8mp-isp2.dtso
> > > > > > b/arch/arm64/boot/dts/freescale/imx8mp-isp2.dtso new file mode
> > > > > > 100644
> > > > > > index 000000000000..14e2e7b2617f
> > > > > > --- /dev/null
> > > > > > +++ b/arch/arm64/boot/dts/freescale/imx8mp-isp2.dtso
> > > > > > @@ -0,0 +1,36 @@
> > > > > > +// SPDX-License-Identifier: (GPL-2.0+ OR MIT)
> > > > > > +/*
> > > > > > + * Copyright 2022 Ideas on Board Oy
> > > > > > + */
> > > > > > +
> > > > > > +/dts-v1/;
> > > > > > +/plugin/;
> > > > > > +
> > > > > > +#include <dt-bindings/media/video-interfaces.h>
> > > > > > +
> > > > > > +&isi_0 {
> > > > > > + status = "disabled";
> > > > > > +
> > > > > > + ports {
> > > > > > + port@1 {
> > > > > > + /delete-node/ endpoint;
> > > > > > + };
> > > > > > + };
> > > > > > +};
> > > > > > +
> > > > > > +&isp_1 {
> > > > > > + status = "okay";
> > > > > > +
> > > > > > + ports {
> > > > > > + port@1 {
> > > > > > + isp1_in: endpoint {
> > > > > > + bus-type = <MEDIA_BUS_TYPE_PARALLEL>;
> > > > > > + remote-endpoint = <&mipi_csi_1_out>;
> > > > > > + };
> > > > > > + };
> > > > > > + };
> > > > > > +};
> > > > > > +
> > > > > > +&mipi_csi_1_out {
> > > > > > + remote-endpoint = <&isp1_in>;
> > > > > > +};
> > >
> > > --
> > > TQ-Systems GmbH | Mühlstraße 2, Gut Delling | 82229 Seefeld, Germany
> > > Amtsgericht München, HRB 105018
> > > Geschäftsführer: Detlef Schneider, Rüdiger Stahl, Stefan Schneider
> > > http://www.tq-group.com/
--
TQ-Systems GmbH | Mühlstraße 2, Gut Delling | 82229 Seefeld, Germany
Amtsgericht München, HRB 105018
Geschäftsführer: Detlef Schneider, Rüdiger Stahl, Stefan Schneider
http://www.tq-group.com/
