Re: i.MX6 MIPI-CSI2 OV5640 Camera testing on Mainline Linux

From: jacopo mondi
Date: Tue Oct 23 2018 - 19:03:29 EST


Hi Adam,

On Tue, Oct 23, 2018 at 12:54:12PM -0500, Adam Ford wrote:
> On Tue, Oct 23, 2018 at 12:39 PM Steve Longerbeam
> <steve_longerbeam@xxxxxxxxxx> wrote:
> >
> >
> > On 10/23/18 10:34 AM, Adam Ford wrote:
> > > On Tue, Oct 23, 2018 at 11:36 AM Steve Longerbeam
> > > <steve_longerbeam@xxxxxxxxxx> wrote:
> > >> Hi Adam,
> > >>
> > >> On 10/23/18 8:19 AM, Adam Ford wrote:
> > >>> On Mon, Oct 22, 2018 at 7:40 AM Fabio Estevam <festevam@xxxxxxxxx> wrote:
> > >>>> Hi Adam,
> > >>>>
> > >>>> On Mon, Oct 22, 2018 at 9:37 AM Adam Ford <aford173@xxxxxxxxx> wrote:
> > >>>>
> > >>>>> Thank you! This tutorial web site is exactly what I need. The
> > >>>>> documentation page in Linux touched on the media-ctl links, but it
> > >>>>> didn't explain the syntax or the mapping. This graphical
> > >>>>> interpretation really helps it make more sense.
> > >>>> Is capturing working well on your i.MX6 board now?
> > >>> Fabio,
> > >>>
> > >>> Unfortunately, no. I built the rootfs based on Jagan's instructions
> > >>> at https://openedev.amarulasolutions.com/display/ODWIKI/i.CoreM6+1.5
> > >>>
> > >>> I tried building both the 4.15-RC6 kernel, a 4.19 kernel and a 4.14 LTS kernel.
> > >>>
> > >>> Using the suggested method of generating the graphical display of the
> > >>> pipeline options, I am able to enable various pipeline options
> > >>> connecting different /dev/videoX options tot he camera. I have tried
> > >>> both the suggested method above as well as the instructions found in
> > >>> Documentation/media/v4l-drivers/imx.rst for their respective kernels,
> > >>> and I have tried multiple options to capture through
> > >>> ipu1_csi1_capture, ipu2_csi1_capture, and ip1_ic_prepenc capture, and
> > >>> all yield a broken pipe.
> > >>>
> > >>> libv4l2: error turning on stream: Broken pipe
> > >>> ERROR: from element /GstPipeline:pipeline0/GstV4l2Src:v4l2src0: Could
> > >>> not read from resource.
> > >>> Additional debug info:
> > >>> gstv4l2bufferpool.c(1064): gst_v4l2_buffer_pool_poll ():
> > >>> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0:
> > >>> poll error 1: Broken pipe (32)
> > >>>
> > >>> I can hear the camera click when I start gstreamer and click again
> > >>> when it stops trying to stream.
> > >>>
> > >>> dmesg indicates a broken pipe as well..
> > >>>
> > >>> [ 2419.851502] ipu2_csi1: pipeline start failed with -32
> > >>>
> > >>> might you have any suggestions?
> > >>
> > >> This -EPIPE error might mean you have a mis-match of resolution, pixel
> > >> format, or field type between one of the source->sink pad links. You can
> > >> find out which pads have a mis-match by enabling dynamic debug in the
> > >> kernel function __media_pipeline_start.
> > > Following Jagan's suggestion, I tried to make sure all the resolution
> > > and pixel formats were set the same between each source and sink.
> > >
> > > media-ctl --set-v4l2 "'ov5640 2-0010':0[fmt:UYVY2X8/640x480
> > > field:none]"
> > > media-ctl --set-v4l2 "'imx6-mipi-csi2':1[fmt:UYVY2X8/640x480
> > > field:none]"
> > > media-ctl --set-v4l2 "'ipu1_csi0_mux':2[fmt:UYVY2X8/640x480
> > > field:none]"
> > > media-ctl --set-v4l2 "'ipu1_csi0':2[fmt:AYUV32/640x480 field:none]"
> > >
> > >> Also make sure you are attempting to stream from the correct /dev/videoN.
> > > I have graphically plotted the pipeline using media-ctl --print-dot
> > > and I can see the proper video is routed, but your dynamic debug
> > > suggestion yielded something:
> > >
> > > imx-media capture-subsystem: link validation failed for 'ov5640
> > > 2-0010':0 -> 'imx6-mipi-csi2':0, error -32
> >
> >
> > It's what I expected, you have a format mismatch between those pads.
>
> Is the mismatch something I am doing wrong with:
>
> media-ctl --set-v4l2 "'ov5640 2-0010':0[fmt:UYVY2X8/640x480 field:none]"
> media-ctl --set-v4l2 "'imx6-mipi-csi2':2[fmt:UYVY2X8/640x480 field:none]"
>

Could you try to verify the actual format configured on the sensor?
(media-ctl --get-v4l2 "'ov5640 2-0010':0").

Depending on the driver version you are running, you might be affected
by different regressions.

> or is there something else I need to do? I just used Jagan's suggestion.

I suggest you to update the driver version to the last one available
in the media-tree master, or at least try to backport the following
commit:
fb98e29 media: ov5640: fix mode change regression

If it turns out the format and mode configured on the sensor do not
match the one you want.

Thanks
j

> adam
> >
> > Steve
> >
> >
> > >
> > > I am assume this means the interface between the camera and the csi2
> > > isn't working. I am going to double check the power rails and the
> > > clocks. i can hear it click when activated and deactivated, so
> > > something is happening.
> > >
> > > adam
> > >
> > >> Steve
> > >>

Attachment: signature.asc
Description: PGP signature