Re: [RFCv3 PATCH 00/33] Core and vb2 enhancements

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Guennadi,

On Friday 20 July 2012 11:55:31 Guennadi Liakhovetski wrote:
> On Thu, 28 Jun 2012, Hans Verkuil wrote:
> > Hi all,
> > 
> > This is the third version of this patch series.
> > 
> > The first version is here:
> > 
> > http://www.mail-archive.com/linux-media@vger.kernel.org/msg47558.html
> 
> Nice to see an owner concept added to the vb2. In soc-camera we're also
> using a concept of a "streamer" user. This is the user, that first calls
> one of data-flow related ioctl()s, like s_fmt(), streamon(), streamoff()
> and all buffer queue-related operations. I realise that this your
> patch-set only deals with the buffer queue, but in principle, do you think
> it would make sense to use such a concept globally? We probably don't want
> to let other processes mess with any of the above calls as long as one
> process is actively managing the streaming.

I think that's a good idea (the uvcvideo driver implements something similar), 
but I'm not sure whether the spec currently describes the expected behaviour 
and relationships between ioctls correctly. That could be a topic for the V4L2 
ambiguities discussions during KS. 

Such a a mechanism should not be imposed on drivers though, as it probably 
wouldn't cover all use cases (think about mem-to-mem drivers where the format 
is per file-handle for instance).

-- 
Regards,

Laurent Pinchart