[PATCH 4/8] ispvideo: Add support for G/S/ENUM_STD ioctl
Hiremath, Vaibhav
hvaibhav at ti.com
Wed Sep 28 02:10:19 EDT 2011
> -----Original Message-----
> From: Laurent Pinchart [mailto:laurent.pinchart at ideasonboard.com]
> Sent: Tuesday, September 27, 2011 11:36 PM
> To: Hiremath, Vaibhav
> Cc: Ravi, Deepthy; linux-media at vger.kernel.org; tony at atomide.com;
> linux at arm.linux.org.uk; linux-omap at vger.kernel.org; linux-arm-
> kernel at lists.infradead.org; linux-kernel at vger.kernel.org;
> mchehab at infradead.org; g.liakhovetski at gmx.de
> Subject: Re: [PATCH 4/8] ispvideo: Add support for G/S/ENUM_STD ioctl
>
> Hi Vaibhav,
>
> On Monday 19 September 2011 17:31:02 Hiremath, Vaibhav wrote:
> > On Friday, September 16, 2011 6:36 PM Laurent Pinchart wrote:
> > > On Friday 16 September 2011 15:00:53 Ravi, Deepthy wrote:
> > > > On Thursday, September 08, 2011 10:51 PM Laurent Pinchart wrote:
> > > > > On Thursday 08 September 2011 15:35:22 Deepthy Ravi wrote:
> > > > >> From: Vaibhav Hiremath <hvaibhav at ti.com>
> > > > >>
> > > > >> In order to support TVP5146 (for that matter any video decoder),
> > > > >> it is important to support G/S/ENUM_STD ioctl on /dev/videoX
> > > > >> device node.
> > > > >
> > > > > Why so ? Shouldn't it be queried on the subdev output pad
> directly ?
> > > >
> > > > Because standard v4l2 application for analog devices will call these
> > > > std ioctls on the streaming device node. So it's done on /dev/video
> to
> > > > make the existing apllication work.
> > >
> > > Existing applications can't work with the OMAP3 ISP (and similar
> complex
> > > embedded devices) without userspace support anyway, either in the form
> of
> > > a GStreamer element or a libv4l plugin. I still believe that analog
> video
> > > standard operations should be added to the subdev pad operations and
> > > exposed through subdev device nodes, exactly as done with formats.
> >
> > I completely agree with your point that, existing application will not
> work
> > without setting links properly. But I believe the assumption here is,
> > media-controller should set the links (along with pad formants) and all
> > existing application should work as is. Isn't it?
>
> The media controller is an API used (among other things) to set the links.
> You
> still need to call it from userspace. That won't happen magically. The
> userspace component that sets up the links and configures the formats, be
> it a
> GStreamer element, a libv4l plugin, or something else, can as well setup
> the
> standard on the TVP5146 subdev.
>
Please look at from analog device point of view which is interfaced to ISP.
OMAP3 ISP => TVP5146 (video decoder)
As a user I would want to expect the standard to be supported on streaming device node, since all standard streaming applications (for analog devices/interfaces) does this.
Setting up the links and format is still something got added with MC framework, and I would consider it as a separate plug-in along with existing applications.
Why do I need to write/use two different streaming application one for MC compatible device and another for non-MC?
> > The way it is being done currently is, set the format at the pad level
> > which is same as analog standard resolution and use existing application
> > for streaming...
>
> At then end of the OMAP3 ISP pipeline video data has long lost its analog
> roots. I don't think standards make sense there.
>
I don't agree with you here, I think we made it clear when we started with MC development activity, we will not break existing standard applications. Media-controller will play a roll to setup the links, connecting the pads and stuff.
Thanks,
Vaibhav
> > I am ok, if we add s/g/enum_std api support at sub-dev level but this
> > should also be supported on streaming device node.
>
> --
> Regards,
>
> Laurent Pinchart
More information about the linux-arm-kernel
mailing list