Hi,
I am bringing up a raw Bayer MIPI CSI-2 camera (ams OSRAM NanEyeM via a Citrobits CBM100 bridge, fixed output: 320x320 RAW10, 2-lane, ~400 Mbps/lane, Non-continuous clock, once powered the bridge board starts decoding MIPI CSI-2 from image sensor, does not need any external clock) on an NXP FRDM-i.MX93 board, on the Scarthgap/6.6 BSP (linux-imx, meta-imx-frdm layer).
I have got the full V4L2 media pipeline working at the plumbing level: our sensor driver probes, binds as an async subdevice, and mx8-img-md successfully links sensor -> mxc-mipi-csi2.0 -> mxc_isi.0 -> /dev/video0. The MIPI CSI-2 host (dwc-mipi-csi2-host) enables cleanly on STREAMON, but no valid frame is ever captured.
I traced this to the ISI driver bound to this board's default device tree (drivers/staging/media/imx/, imx93.dtsi's isi@4ae40000 node using "fsl,imx93-isi", "fsl,imx8-isi"). Its format table in imx8-isi-fmt.c only contains RGB565/RGB888/BGR888 and a handful of YUV mbus codes -- no SBGGR/SRGGB/SGRBG/SGBRG entries at any bit depth. mxc_isi_get_src_fmt() in imx8-isi-cap.c has no branch for Bayer input, so it silently falls through to index 0 of its RGB/YUV table rather than erroring (driver's kernel log shows "input fmt RGB4" for what is actually a RAW10 Bayer source).
I also found what looks like a more complete, more recently structured ISI driver in the same kernel tree at drivers/media/platform/nxp/imx8-isi/ (split into core/pipe/video/crossbar/gasket files), whose imx8-isi-video.c and imx8-isi-pipe.c do have full Bayer support at 8/10/12/14/16 bits, including SBGGR10_1X10 matching our sensor exactly. Its imx8-isi-core.c also declares a "nxp,imx93-isi" compatible string. However, imx93.dtsi doesn't currently instantiate a node with that compatible string anywhere, so this driver doesn't appear to be wired into the i.MX93 FRDM (or, as far as we've checked, i.MX93 EVK) device trees at all in this BSP release.
My questions:
1. Is drivers/media/platform/nxp/imx8-isi/ intended as the eventual replacement for the staging ISI driver on i.MX93, and is there a reference device tree binding (port/endpoint graph structure) for wiring the CSI2 host and a sensor into it, that I could look at or adapt?
2. Is there a supported way to get raw Bayer/RAW10 capture working through the currently-bound staging ISI driver on i.MX93, that we may be missing -- e.g. a bypass mode, a different video node, or a format we haven't tried?
3. Is Bayer/RAW10 support for the staging ISI driver on this BSP release on any roadmap, or is switching to the newer driver tree the recommended path for raw-sensor use cases on i.MX93 today?
I found a couple of post like i.MX93 using MIPI-CSI-2 with a Bayer Pattern Camera(RAW10) and does not seem to have the solution. If I missed out a good solution, please do let me know.
here is the interface of the bridge board and i.MX 93's P6 connector.
Thanks and best regards
kth
I found this threat https://community.nxp.com/t5/i-MX-Processors/IMX219-RPi-Cam-v2-probes-but-does-not-stream-on-FRDM-IM... . If I may ask,
1. Is Bayer/RAW10 ISI support (and the link_setup/s_power tolerance fix) already included in the 6.12.34-2.1.0+ officially-supported BSP line? If so, is upgrading our BSP the recommended path rather than backporting patches onto 6.6.36?
2. If I need to stay on 6.6.36 for now, is there an official/blessed patch set for the ISI Bayer format support (rather than us reconstructing it from the community thread)?
3. Separately -- I also found a newer, more complete ISI driver already present in the same 6.6.36 kernel tree at drivers/media/platform/nxp/imx8-isi/, with full Bayer support and a "nxp,imx93-isi" compatible string not currently wired into imx93.dtsi. Is that the driver 6.12.34-2.1.0+ actually uses, or is it unrelated?
4. I just noticed LF6.18.37_2.1.0(imx-linux-wrynose)lists FRDM-i.MX93 as fully supported https://github.com/nxp-imx/imx-manifest/tree/imx-linux-wrynose . Does this release include Bayer/RAW10 format support in i.MX93 ISI driver, or does that i.MX93 reference architecture still expect raw sensors to go through an external ISP(like AP1302) rather than direct-to-ISI? If the latter, is there any supported path for direct raw Bayer capture on i.MX93 at all?
Thank you!