by Adam Ford

[permalink] [raw]

Subject: Re: [RFC V2 0/5] arm64: dts: imx8mm: Enable CSI and OV5640 Camera

On Tue, Nov 2, 2021 at 12:50 PM Tim Harvey <[email protected]> wrote:
>
> On Mon, Nov 1, 2021 at 5:30 PM Adam Ford <[email protected]> wrote:
> >
> > On Mon, Nov 1, 2021 at 6:05 PM Tim Harvey <[email protected]> wrote:
> > >
> > > On Fri, Oct 29, 2021 at 4:11 AM Frieder Schrempf
> > > <[email protected]> wrote:
> > > >
> > > > On 28.10.21 02:39, Adam Ford wrote:
> > > > > On Sun, Oct 24, 2021 at 7:16 AM Fabio Estevam <[email protected]> wrote:
> > > > >>
> > > > >> Hi Adam,
> > > > >>
> > > > >> [Adding Frieder on Cc]
> > > > >>
> > > > >> On Sat, Oct 23, 2021 at 5:35 PM Adam Ford <[email protected]> wrote:
> > > > >>>
> > > > >>> The imx8mm appears to have both a CSI bridge and mipi-csi-2 drivers. With
> > > > >>> those enabled, both the imx8mm-evk and imx8mm-beacon boards should be able
> > > > >>> use an OV5640 camera.
> > > > >>>
> > > > >>> The mipi-csi2 driver sets the clock frequency to 333MHz, so the clock parent
> > > > >>> of the CSI1 must be reparented to a faster clock. On the custom NXP kernel,
> > > > >>> they use IMX8MM_SYS_PLL2_1000M, so that is done in the device tree to match.
> > > > >>>
> > > > >>> With the CSI and mipi_csi2 drivers pointing to an OV5640 camera, the media
> > > > >>> pipeline can be configured with the following:
> > > > >>>
> > > > >>> media-ctl --links "'ov5640 1-003c':0->'imx7-mipi-csis.0':0[1]"
> > > > >>>
> > > > >>> The camera and various nodes in the pipeline can be configured for UYVY:
> > > > >>> media-ctl -v -V "'ov5640 1-003c':0 [fmt:UYVY8_1X16/640x480 field:none]"
> > > > >>> media-ctl -v -V "'csi':0 [fmt:UYVY8_1X16/640x480 field:none]"
> > > > >>>
> > > > >>> With that, the media pipeline looks like:
> > > > >>>
> > > > >>>
> > > > >>> Media controller API version 5.15.0
> > > > >>>
> > > > >>> Media device information
> > > > >>> ------------------------
> > > > >>> driver imx7-csi
> > > > >>> model imx-media
> > > > >>> serial
> > > > >>> bus info platform:32e20000.csi
> > > > >>> hw revision 0x0
> > > > >>> driver version 5.15.0
> > > > >>>
> > > > >>> Device topology
> > > > >>> - entity 1: csi (2 pads, 2 links)
> > > > >>> type V4L2 subdev subtype Unknown flags 0
> > > > >>> device node name /dev/v4l-subdev0
> > > > >>> pad0: Sink
> > > > >>> [fmt:UYVY8_1X16/640x480 field:none colorspace:srgb xfer:srgb ycbcr:601 quantization:lim-range]
> > > > >>> <- "imx7-mipi-csis.0":1 [ENABLED,IMMUTABLE]
> > > > >>> pad1: Source
> > > > >>> [fmt:UYVY8_1X16/640x480 field:none colorspace:srgb xfer:srgb ycbcr:601 quantization:lim-range]
> > > > >>> -> "csi capture":0 [ENABLED,IMMUTABLE]
> > > > >>>
> > > > >>> - entity 4: csi capture (1 pad, 1 link)
> > > > >>> type Node subtype V4L flags 0
> > > > >>> device node name /dev/video0
> > > > >>> pad0: Sink
> > > > >>> <- "csi":1 [ENABLED,IMMUTABLE]
> > > > >>>
> > > > >>> - entity 10: imx7-mipi-csis.0 (2 pads, 2 links)
> > > > >>> type V4L2 subdev subtype Unknown flags 0
> > > > >>> device node name /dev/v4l-subdev1
> > > > >>> pad0: Sink
> > > > >>> [fmt:UYVY8_1X16/640x480 field:none colorspace:smpte170m xfer:709 ycbcr:601 quantization:lim-range]
> > > > >>> <- "ov5640 1-003c":0 [ENABLED]
> > > > >>> pad1: Source
> > > > >>> [fmt:UYVY8_1X16/640x480 field:none colorspace:smpte170m xfer:709 ycbcr:601 quantization:lim-range]
> > > > >>> -> "csi":0 [ENABLED,IMMUTABLE]
> > > > >>>
> > > > >>> - entity 15: ov5640 1-003c (1 pad, 1 link)
> > > > >>> type V4L2 subdev subtype Sensor flags 0
> > > > >>> device node name /dev/v4l-subdev2
> > > > >>> pad0: Source
> > > > >>> [fmt:UYVY8_1X16/640x480@1/30 field:none colorspace:srgb xfer:srgb ycbcr:601 quantization:full-range]
> > > > >>> -> "imx7-mipi-csis.0":0 [ENABLED]
> > > > >>>
> > > > >>> When configured, gstreamer can be used to capture 1 frame and store it to a file.
> > > > >>>
> > > > >>> gst-launch-1.0 -v v4l2src num-buffers=1 ! video/x-raw,format=UYVY,width=640,height=480,framerate=60/1 ! filesink location=test
> > > > >>>
> > > > >>> Unfortunately, the video capture never appears to happen. No errors occur, not
> > > > >>> interrupts are recorded and no errors are recorded.
> > > > >>>
> > > > >>> gst-launch-1.0 -v v4l2src num-buffers=1 ! video/x-raw,format=UYVY,width=640,height=480,framerate=60/1 ! filesink location=test
> > > > >>> Setting pipeline to PAUSED ...
> > > > >>> Pipeline is live and does not need PREROLL ...
> > > > >>> Pipeline is PREROLLED ...
> > > > >>> Setting pipeline to [ 114.819632] v4l2_get_link_freq: Link frequency estimated using pixel rate: result might be inaccurate
> > > > >>> PLAYING ...
> > > > >>> New clock: GstSystem[ 114.829203] v4l2_get_link_freq: Consider implementing support for V4L2_CID_LINK_FREQ in the transmitter driver
> > > > >>> Clock
> > > > >>> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0.GstPad:src: caps = video/x-raw, format=(string)UYVY, width=(int)640, height=(int)480, framerate=(fraction)60/1, interlace-mode=(string)progressive, colorimetry=(string)bt709
> > > > >>> /GstPipeline:pipeline0/GstCapsFilter:capsfilter0.GstPad:src: caps = video/x-raw, format=(string)UYVY, width=(int)640, height=(int)480, framerate=(fraction)60/1, interlace-mode=(string)progressive, colorimetry=(string)bt709
> > > > >>> /GstPipeline:pipeline0/GstFileSink:filesink0.GstPad:sink: caps = video/x-raw, format=(string)UYVY, width=(int)640, height=(int)480, framerate=(fraction)60/1, interlace-mode=(string)progressive, colorimetry=(string)bt709
> > > > >>> /GstPipeline:pipeline0/GstCapsFilter:capsfilter0.GstPad:sink: caps = video/x-raw, format=(string)UYVY, width=(int)640, height=(int)480, framerate=(fraction)60/1, interlace-mode=(string)progressive, colorimetry=(string)bt709
> > > > >>>
> > > > >>>
> > > > >>> If anyone has any insight as to what might be wrong, I'd like feedback.
> > > > >>> I posted a device tree that I beleive goes with the newer imx8mm-evk, but
> > > > >>> I do not have this hardware, so I cannot test it.
> > > > >>
> > > > >> It seems that Frieder on Cc managed to get camera capture to work on
> > > > >> i.MX8MM here:
> > > > >> https://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kontron-electronics.de%2Fsw%2Fmisc%2Flinux%2F-%2Fcommits%2Fv5.10-mx8mm-csi&data=04%7C01%7Cfrieder.schrempf%40kontron.de%7Cfe4f7347385f4185b1c608d999ab75b5%7C8c9d3c973fd941c8a2b1646f3942daf1%7C0%7C0%7C637709783979195945%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=PbGqhzb2mbUA2SD44%2BosK8rNkK12m1LRd6W4tvkawno%3D&reserved=0
> > > > >>
> > > > >> Hopefully, this can help to figure out what is missing in mainline to
> > > > >> get camera capture to work on i.MX8M.
> > > > >>
> > > > >> I don't have access to an OV5640 camera to connect to the imx8mm-evk
> > > > >> board to try your series.
> > > > >
> > > > > Fabio,
> > > > >
> > > > > Thanks for the heads up on that repo. I was able to use that repo and
> > > > > get still images to capture on an OV5640, but I noticed a fair amount
> > > > > of differences between that repo and what's found in linux-next.
> > > > >
> > > > > Laurent,
> > > > >
> > > > > I haven't exhausted the patch differences, but I found at least a few
> > > > > that appear to be missiing upstream, and I am curious to know if/what
> > > > > your opinion is on whether or not they're needed, since the patches on
> > > > > Frieder's repo appear to come from you.
> > > > > [1] - media: imx: imx7-media-csi: Add i.MX8MM identification
> > > > > [2] - media: imx: imx7_mipi_csis: Don't set reserved CLK_CTRL field on i.MX8MM
> > > > > [3] - media: imx: imx7_mipi_csis: Set dual pixel mode for RAW formats
> > > > >
> > > > > media: imx: imx7_mipi_csis: Set dual pixel mode for RAW formats
> > > > >
> > > > > Maybe these don't need to be applied, but they are 'some' of the
> > > > > differences that I see between this 5.10 branch and linux-next . I
> > > > > know there are more, but
> > > > >
> > > > >
> > > > > [1] - https://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kontron-electronics.de%2Fsw%2Fmisc%2Flinux%2F-%2Fcommit%2F8ac7ec6db0c260a871038721886dbdb6660ed84c&data=04%7C01%7Cfrieder.schrempf%40kontron.de%7Cfe4f7347385f4185b1c608d999ab75b5%7C8c9d3c973fd941c8a2b1646f3942daf1%7C0%7C0%7C637709783979195945%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=j1iuXWljDd8wA5M44KwLCb%2F21tpdOnKZuJazl25bXbQ%3D&reserved=0
> > > > > [2] - https://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kontron-electronics.de%2Fsw%2Fmisc%2Flinux%2F-%2Fcommit%2F0b5727c8eba8c370f7db5eace0243f78992a4dbb&data=04%7C01%7Cfrieder.schrempf%40kontron.de%7Cfe4f7347385f4185b1c608d999ab75b5%7C8c9d3c973fd941c8a2b1646f3942daf1%7C0%7C0%7C637709783979205943%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=buWbZF0tYfVmibQgBbKJM1PF%2Fw7%2BVO9jhXRCI1zf7TI%3D&reserved=0
> > > > > [3] - https://eur04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kontron-electronics.de%2Fsw%2Fmisc%2Flinux%2F-%2Fcommit%2F14befa6bc146b10092a6ac5d0ed4d42c87c6cf27&data=04%7C01%7Cfrieder.schrempf%40kontron.de%7Cfe4f7347385f4185b1c608d999ab75b5%7C8c9d3c973fd941c8a2b1646f3942daf1%7C0%7C0%7C637709783979205943%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=60iLhs0G0FtQegNp9XtVxAhvZEcltdAGGMNAm2l1cSs%3D&reserved=0
> > > > >
> > > > > Frieder et al,
> > > > >
> > > > > Have you (or anyone) tried CSI cameras on anything newer than 5.10? I
> > > > > am curious to see if a regression popped in somewhere, but git bisect
> > > > > will make this difficult since there is a fair amount of variation
> > > > > between this custom repo and the upstream.
> > > >
> > > > No, I haven't done anything with CSI on a more recent kernel. And I only
> > > > used CSI with the tree above and the ADV7280M bridge. I don't have any
> > > > hardware with a sensor/camera.
> > > >
> > > > In case you haven't seen this already, here is a thread with some notes
> > > > about my testing results:
> > > > https://patchwork.kernel.org/project/linux-media/cover/[email protected]/.
> > > >
> > >
> > > For what it's worth I've got another test setup for IMX8MM CSI
> > > capture. I have a Raspberry Pi Camera module v2 connected to an
> > > imx8mm-venice-gw73xx board. This is a IMX219 8.28MP camera with a
> > > 4-lane CSI connection.
> > >
> > > Putting Adam's patch 'arm64: dts: imx8mm: Add CSI nodes' as well as
> > > the 'blk-ctl series on top of 5.15 and adding support to my dt via:
> > >
> > > commit 87f908a57f48bd7375113991434c2923d65506ac (HEAD -> v5.15-venice)
> > > Author: Tim Harvey <[email protected]>
> > > Date: Wed Oct 27 15:45:23 2021 -0700
> > >
> > > arm64: dts: imx8mm-venice-gw73xx: add rpi camera module v2
> > >
> > > Add support for rpi camera module v2 which is an IMX219 8MP module:
> > > - https://datasheets.raspberrypi.com/camera/camera-v2-schematics.pdf
> > > - has its own on-board 24MHz osc so no clock required from baseboard
> > > - pin 11 enables 1.8V and 2.8V LDO which is connected to
> > > GW73xx MIPI_GPIO4 (IMX8MM GPIO1_IO1). imx219 driver does not
> > > support powerdown-gpios and using gpio1 as reset-gpios
> > >
> > > Signed-off-by: Tim Harvey <[email protected]>
> > >
> > > diff --git a/arch/arm64/boot/dts/freescale/imx8mm-venice-gw73xx.dtsi
> > > b/arch/arm64/boot/dts/freescale/imx8mm-venice-gw73xx.dtsi
> > > index 7b00b6b5bb38..b708c80d884b 100644
> > > --- a/arch/arm64/boot/dts/freescale/imx8mm-venice-gw73xx.dtsi
> > > +++ b/arch/arm64/boot/dts/freescale/imx8mm-venice-gw73xx.dtsi
> > > @@ -35,6 +35,13 @@
> > > };
> > > };
> > >
> > > + cam24m: cam24m {
> > > + compatible = "fixed-clock";
> > > + #clock-cells = <0>;
> > > + clock-frequency = <24000000>;
> > > + clock-output-names = "cam24m";
> > > + };
> > > +
> > > pcie0_refclk: pcie0-refclk {
> > > compatible = "fixed-clock";
> > > #clock-cells = <0>;
> > > @@ -100,6 +107,19 @@
> > > };
> > > };
> > >
> > > +&csi {
> > > + status = "okay";
> > > +};
> > > +
> > > +&imx8mm_mipi_csi_in {
> > > + remote-endpoint = <&imx219_to_mipi_csi2>;
> > > + data-lanes = <1 2>;
> > > +};
> > > +
> > > +&mipi_csi2 {
> > > + status = "okay";
> > > +};
> > > +
> > > /* off-board header */
> > > &ecspi2 {
> > > pinctrl-names = "default";
> > > @@ -132,6 +152,25 @@
> > > pinctrl-names = "default";
> > > pinctrl-0 = <&pinctrl_i2c3>;
> > > status = "okay";
> > > +
> > > + imx219: sensor@10 {
> > > + compatible = "sony,imx219";
> > > + pinctrl-names = "default";
> > > + pinctrl-0 = <&pinctrl_imx219>;
> > > + reg = <0x10>;
> > > + clocks = <&cam24m>;
> > > + reset-gpios = <&gpio1 1 GPIO_ACTIVE_HIGH>;
> > > +
> > > + port {
> > > + /* MIPI CSI-2 bus endpoint */
> > > + imx219_to_mipi_csi2: endpoint {
> > > + remote-endpoint = <&imx8mm_mipi_csi_in>;
> > > + clock-lanes = <0>;
> > > + data-lanes = <1 2>;
> > > + link-frequencies = /bits/ 64 <456000000>;
> > > + };
> > > + };
> > > + };
> > > };
> > >
> > > &pcie_phy {
> > > @@ -297,6 +336,12 @@
> > > >;
> > > };
> > >
> > > + pinctrl_imx219: imx219grp {
> > > + fsl,pins = <
> > > + MX8MM_IOMUXC_GPIO1_IO01_GPIO1_IO1 0x41
> > > + >;
> > > + };
> > > +
> > > pinctrl_pcie0: pcie0grp {
> > > fsl,pins = <
> > > MX8MM_IOMUXC_SAI1_RXD4_GPIO4_IO6 0x41
> > >
> > > While the IMX219 supports up to 4 MIPI CSI2 lanes and a variety of
> > > resolutions up to 8MP, the IMX219 driver (drivers/media/i2c/imx219.c)
> > > currently supports only 2 lanes, and a few different resolutions
> > > including 1080p@30fps (cropped FOV), 1640x1232@30fps (2x2 binned),
> > > 640x480@30fps (cropped) with RAW8 and RAW10 formats.
> >
> > I am hoping to support this camera as well once I get the OV5640 working.
> >
> > >
> > > I'm setting up the pipeline like this:
> > > media-ctl --links "'imx219 2-0010':0->'imx7-mipi-csis.0':0[1]"
> > > media-ctl -v -V "'imx219 2-0010':0 [fmt:SRGGB10/640x480 field:none]"
> > > media-ctl -v -V "'csi':0 [fmt:SRGGB10/640x480 field:none]"
> > >
> > > and capture:
> > > gst-launch-1.0 -v v4l2src num-buffers=1 !
> > > video/x-bayer,format=rggb,width=640,height=480,framerate=30/1 !
> > > filesink location=test
> > >
> > > The above hangs after:
> > > Setting pipeline to PAUSED ...
> > > Pipeline is live and does not need PREROLL ...
> > > Setting pipeline to PLAYING ...
> > > /GstPipeline:pipeline0/GstV4l2Src:v4l2src0.GstPad:src: caps =
> > > video/x-bayer, format=(string)rggb, width=(int)640, height=(int)480,
> > > framerate=(fraction)30/1, interlace-mode=(string)progressive
> > > New clock: GstSystemClock
> > > /GstPipeline:pipeline0/GstCapsFilter:capsfilter0.GstPad:src: caps =
> > > video/x-bayer, format=(string)rggb, width=(int)640, height=(int)480,
> > > framerate=(fraction)30/1, interlace-mode=(string)progressive
> > > /GstPipeline:pipeline0/GstFileSink:filesink0.GstPad:sink: caps =
> > > video/x-bayer, format=(string)rggb, width=(int)640, height=(int)480,
> > > framerate=(fraction)30/1, interlace-mode=(string)progressive
> > > /GstPipeline:pipeline0/GstCapsFilter:capsfilter0.GstPad:sink: caps =
> > > video/x-bayer, format=(string)rggb, width=(int)640, height=(int)480,
> > > framerate=(fraction)30/1, interlace-mode=(string)progressive
> > >
> > > I've tried Laurent's 'media: imx: imx7_mipi_csis: Set dual pixel mode
> > > for RAW formats' patch with the same results.
> >
> > Same here.
> >
> > >
> > > Let me know if any of you have some ideas here.
> >
> > Tim,
> >
> > Can you check /proc/interrupts? I assume that you've got no interrupts either.
>
> Adam,
>
> like your setup I see:
> 52: 0 0 0 0 GICv3 48 Level csi
> 53: 0 0 0 0 GICv3 49 Level
> 32e30000.mipi-csi
> Err: 0
>
> clk_summary:
> enable prepare protect
> duty hardware
> clock count count count rate
> accuracy phase cycle enable
> -------------------------------------------------------------------------------------------------------
> sys_pll2 6 6 0 1000000000
> 0 0 50000 Y
> sys_pll2_out 1 1 0 1000000000
> 0 0 50000 Y
> sys_pll2_1000m 3 3 0 1000000000
> 0 0 50000 Y
> csi1_phy_ref 2 2 0 1000000000
> 0 0 50000 Y
> csi1_core 1 1 0 333333334
> 0 0 50000 Y
> csi1_root_clk 3 3 0 333333334
> 0 0 50000 Y
>
> >
> > I've added a bunch of debug code to the 5.10 NXP kernel and dumped a
> > bunch of the regisgters and compared it to the ov5640 running on the
> > beacon board. As of now, I think the issue is somewhere in the CSI
> > Bridge driver. I've made a few changes to make the CSI bridge
> > registers match that of the 5.10 NXP kernel, but I haven't found the
> > magic register yet. In my setup, I can get the CSIS driver registers
> > to match the NXP kernel, so I'm leaning toward either the CSI bridge
> > driver or an order of operations difference between the mainline
> > kernel and NXP's.
> >
> > I'm working on this in my spare time, so I'll keep you posted if I
> > make any progress.
> >
>
> Sounds good, let me know if you need more testing.
>
> I also have an imx8mm-evk here with the ov5640 camera so can do some
> testing there as well. I'm not clear how to use the imx8mm-evk with
> NXP's lf-5.10.y downstream vendor kernel in order to compare as it
> uses their proprietary platform capture driver instead of using the
> media controller API and I'm not clear how to set that device up for
> capture.
>
> However if I test that board using the 5.15 with your series (adjusted
> for the ov5640 camera on i2c3 for the imx8mm-evk) I get the same
> results as you do also - no interrupts.

I have put together a list of registers that are set (and in what
order). I'll try to clean it up and post them. I think there are a
few registers being set differently, but I am not a CSI-2 expert, so I
do not necessarily know what the 'right' register setting is.

I also noticed that there is a CSIS driver [1] that's not in the
staging area that seems to be similar to that of the iMX8M Mini, so
I've been tempted to review those register settings to see if it could
be adapted to fit the imx7 and the 8MM, but I think the area to focus
is the CSI Bridge and not the CSIS. I could be wrong, but my CSIS
registers are nearly identical to a working register set from an older
NXP kernel.

adam

[1] - https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/media/platform/exynos4-is/mipi-csis.c?h=v5.15
>
> Best regards,
>
> Tim

2021-11-03 00:48:35

Hi Tim,

On Fri, Nov 05, 2021 at 09:00:03AM -0700, Tim Harvey wrote:
> On Wed, Oct 27, 2021 at 6:34 PM Laurent Pinchart wrote:
> >
> > Hi Adam,
> >
> > Thank you for the patch.
> >
> > On Sat, Oct 23, 2021 at 03:34:56PM -0500, Adam Ford wrote:
> > > The schematic shows support for a camera interface, and the NXP
> > > kernel shows it is an OV5640.
> >
> > The camera is an external module though. Should this be a DT overlay ?
> >
>
> Laurent,
>
> I wanted to ask you about your comment here. I would agree that for
> something like the OV5640 on the imx8mm-evk which is an add-in board
> via a connector should be a dt overlay. I'm investigating using
> overlays for features like this on my boards vs creating hierarchical
> dts files and I see that the kernel allows building fragments with
> '/plugin/' but I don't see any such overlays in the kernel tree
> currently. Would overlay/fragments be accepted?

I believe so (otherwise I'm not sure why the build system would allow
compiling overlays). Rob may have a more authoritative opinion on this
topic.

> Are there any examples
> in the kernel tree already that I'm missing?

Commit 7a4c31ee877a ("arm64: zynqmp: Add support for Xilinx Kria SOM
board") has been added recently.

--
Regards,

Laurent Pinchart