Re: Omap3 ISP + Gstreamer v4l2src

[Michael Jones <michael.jones@matrix-vision.de>]

Hi Adam,

On 12/07/2011 11:34 AM, Laurent Pinchart wrote:
> Hi Adam,
>
> On Wednesday 07 December 2011 09:02:42 Adam Pledger wrote:
>> Hi Laurent,
>>
>> Firstly, please accept my apologies, for what is very probably a naive
>> question. I'm new to V4L2 and am just getting to grips with how things
>> work.
>
> No worries.
>
>> I'm using a tvp5151 in bt656 mode with the Omap3 ISP,
>
> Please note that BT.656 support is still experimental, so issues are not
> unexpected.
>
>> as described in this thread (Your YUV support tree + some patches for bt656,
>> based on 2.6.39):
>> http://comments.gmane.org/gmane.linux.drivers.video-input-
> infrastructure/39539
>>
>> I am able to capture some frames using yavta, using the media-ctl
>> configuration as follows:
>> media-ctl -v -r -l '"tvp5150 3-005d":0->"OMAP3 ISP CCDC":0[1], "OMAP3
>> ISP CCDC":1->"OMAP3 ISP CCDC output":0[1]'
>> media-ctl -v --set-format '"tvp5150 3-005d":0 [UYVY2X8 720x625]'
>> media-ctl -v --set-format '"OMAP3 ISP CCDC":0 [UYVY2X8 720x625]'
>> media-ctl -v --set-format '"OMAP3 ISP CCDC":1 [UYVY2X8 720x625]'
>>
>> This yields this:
>
> [snip]
>
> Looks good.
>
>> The following works nicely:
>> yavta -f UYVY -s 720x625 -n 4 --capture=4 -F /dev/video2
>>
>> The problem comes when I try to use gstreamer to capture from
>> /dev/video2, using the following:
>> gst-launch v4l2src device="/dev/video2" !
>> 'video/x-raw-yuv,width=720,height=625' ! filesink location=sample.yuv
>>
>> This fails with:
>> ERROR: from element /GstPipeline:pipeline0/GstV4l2Src:v4l2src0: Failed
>> getting controls attributes on device '/dev/video2'.
>> Additional debug info:
>> v4l2_calls.c(267): gst_v4l2_fill_lists ():
>> /GstPipeline:pipeline0/GstV4l2Src:v4l2src0:
>> Failed querying control 9963776 on device '/dev/video2'. (25 -
>> Inappropriate ioctl for device)
>>
>> My question is, should this "just work"? It was my understanding that
>> once the pipeline was configured with media-ctl then the CCDC output pad
>> should behave like a standard V4L2 device node.
>
> That's more or less correct. There have been a passionate debate regarding
> what a "standard V4L2 device node" is. Not all V4L2 ioctls are mandatory, and
> no driver implements them all. The OMAP3 ISP driver implements a very small
> subset of the V4L2 API, and it wasn't clear whether that still qualified as
> V4L2. After discussions we decided that the V4L2 specification will document
> profiles, with a set of required ioctls for each of them. The OMAP3 ISP
> implements the future video streaming profile.
>
> I'm not sure what ioctls v4l2src consider as mandatory. The above error
> related to a CTRL ioctl (possibly VIDIOC_QUERYCTRL), which isn't implemented
> by the OMAP3 ISP driver and will likely never be. I don't think that should be
> considered as mandatory.
>
> I think that v4l2src requires the VIDIOC_ENUMFMT ioctl, which isn't
> implemented in the OMAP3 ISP driver. That might change in the future, but I'm
> not sure yet whether it will. In any case, you might have to modify v4l2src
> and/or the OMAP3 ISP driver for now. Some patches have been posted a while ago
> to this mailing list.

Here was my submission for ENUM_FMT support:
http://www.mail-archive.com/linux-media@vger.kernel.org/msg29640.html

I submitted this in order to be able to use the omap3-isp with 
GStreamer.  I missed the discussion about V4L2 "profiles", but when I 
submitted that patch we discussed whether ENUM_FMT was mandatory. After 
I pointed out that the V4L2 spec states plainly that it _is_ mandatory, 
I thought Laurent basically agreed that it was reasonable.

Laurent, what do you think about adding ENUM_FMT support now?

>
>> I realise that this might be something borked with my build dependencies
>> (although I'm pretty certain that v4l2src is being built against the
>> latest libv42) or gstreamer. Before I start digging through the code to
>> work out what is going on with the ioctl handling, I thought I would
>> check to see whether this should work, or whether I am doing something
>> fundamentally silly.
>

-Michael

MATRIX VISION GmbH, Talstrasse 16, DE-71570 Oppenweiler
Registergericht: Amtsgericht Stuttgart, HRB 271090
Geschaeftsfuehrer: Gerhard Thullner, Werner Armingeon, Uwe Furtner, Erhard Meier