Hi Fabio,
On Mon Sep 13, 2021 at 9:03 PM WEST, Fabio Estevam wrote:
> On Mon, Sep 13, 2021 at 9:53 AM Fabio Estevam <festevam@xxxxxxxxx> wrote:
> >
> > Hi Jacopo,
> >
> > On Mon, Sep 13, 2021 at 5:58 AM Jacopo Mondi <jacopo@xxxxxxxxxx> wrote:
> >
> > > There's an
> > >
> > > tw9910_s_power(&priv->subdev, 0);
> > >
> > > at the end of the video_probe() function.
> > >
> > > The driver handles power management with the legacy s_power() call
> > > chain, and the receiver driver needs to v4l2_pipeline_pm_get() which
> > > the imx driver does when the capture node is open.
> > >
> > > Just an hint, you might have noticed already
> >
> > Thanks for your comments and review. Yes, I have fixed the I2C errors.
> >
> > I plan to re-submit the entire series after I get the TW9990 to work
> > on my imx6sx-based board.
> >
> > The capture driver is drivers/staging/media/imx/imx7-media-csi.c.
> >
> > Currently, I am not able to get the stream to start.
> >
> > This is the configuration I am using:
> >
> > media-ctl -l "'tw9910 2-0044':0 -> 'csi':0[1]"
> > media-ctl -l "'csi':1 -> 'csi capture':0[1]"
> > media-ctl -v -V "'tw9910 2-0044':0 [fmt:UYVY8_2X8/720x480 field:any]"
> > Opening media device /dev/media0
> > Enumerating entities
> > Found 3 entities
> > Enumerating pads and links
> > Setting up format UYVY8_2X8 720x480 on pad tw9910 2-0044/0
> > Format set: UYVY8_2X8 720x480
> > Setting up format UYVY8_2X8 720x480 on pad csi/0
> > Format set: UYVY8_2X8 720x480
> >
> > Then I launch the capture stream command:
> >
> > v4l2-ctl --stream-mmap -d /dev/video1
> >
> > but nothing happens here, no >>>>> frame indication progress is shown.
> >
> > If I hit CTRL + C then I get:
> > [ 715.467623] csi: wait last EOF timeout
>
> I also need to set the video standard to PAL:
>
> # v4l2-ctl --device /dev/v4l-subdev1 --set-standard PAL
> Standard set to 000000ff
>
> # media-ctl -l "'tw9910 2-0044':0 -> 'csi':0[1]"
> # media-ctl -l "'csi':1 -> 'csi capture':0[1]"
> # media-ctl -v -V "'tw9910 2-0044':0 [fmt:UYVY8_2X8/720x576
> field:interlaced-bt]"
> Opening media device /dev/media0
> Enumerating entities
> Found 3 entities
> Enumerating pads and links
> Setting up format UYVY8_2X8 720x576 on pad tw9910 2-0044/0
> Format set: UYVY8_2X8 720x576
> Setting up format UYVY8_2X8 720x576 on pad csi/0
> Format set: UYVY8_2X8 720x576
>
> # v4l2-ctl --stream-mmap -d /dev/video1
> [ 38.554032] priv->vdev.fmt.fmt.pix.width is 800
> [ 38.561444] f.fmt.pix.width is 720
> [ 38.567607] priv->vdev.fmt.fmt.pix.height is 600
> [ 38.574973] f.fmt.pix.height is 576
> [ 38.580992] priv->vdev.cc->cs is 1
> [ 38.586958] cc->cs is 1
> [ 38.591761] priv->vdev.compose.width is 800
> [ 38.598420] compose.width is 720
> [ 38.604048] priv->vdev.compose.height is 600
> [ 38.610681] compose.height is 576
> [ 38.616347] csi: capture format not valid
> VIDIOC_STREAMON returned -1 (Broken pipe)
>
> I added some printk debug lines to show the mismatch that prevents
> the pipeline to start in capture_validate_fmt():
>
> diff --git a/drivers/staging/media/imx/imx-media-capture.c
> b/drivers/staging/media/imx/imx-media-capture.c
> index b2f2cb3d6a60..511625981e93 100644
> --- a/drivers/staging/media/imx/imx-media-capture.c
> +++ b/drivers/staging/media/imx/imx-media-capture.c
> @@ -552,6 +552,21 @@ static int capture_validate_fmt(struct capture_priv *priv)
> ret = __capture_try_fmt_vid_cap(priv, &fmt_src, &f, &cc, &compose);
> if (ret)
> return ret;
> +
> + pr_err("priv->vdev.fmt.fmt.pix.width is %d\n",
> priv->vdev.fmt.fmt.pix.width);
> + pr_err("f.fmt.pix.width is %d\n", f.fmt.pix.width);
> +
> + pr_err("priv->vdev.fmt.fmt.pix.height is %d\n",
> priv->vdev.fmt.fmt.pix.height);
> + pr_err("f.fmt.pix.height is %d\n", f.fmt.pix.height );
> +
> + pr_err("priv->vdev.cc->cs is %d\n", priv->vdev.cc->cs);
> + pr_err("cc->cs is %d\n", cc->cs);
> +
> + pr_err("priv->vdev.compose.width is %d\n", priv->vdev.compose.width);
> + pr_err("compose.width is %d\n", compose.width);
> +
> + pr_err("priv->vdev.compose.height is %d\n", priv->vdev.compose.height);
> + pr_err("compose.height is %d\n", compose.height);
>
> return (priv->vdev.fmt.fmt.pix.width != f.fmt.pix.width ||
> priv->vdev.fmt.fmt.pix.height != f.fmt.pix.height ||
>
> Any ideas as to whu such mismatch happens?
Looking at this code it looks like you are using an old tree, Laurent
meanwhile have sent a patch that change this function:
a9512b261afd ("media: imx: capture: Simplify capture_validate_fmt() implementation")
can you test with the latest code as see if you have the same
problems.
------
Cheers,
Rui
