Re: [PATCH 1/2] [media] v4l2-mem2mem: set the queue owner field just as vb2_ioctl_reqbufs does

[Philipp Zabel <p.zabel@pengutronix.de>]

Hi Sylwester,

Am Donnerstag, den 25.06.2015, 15:11 +0200 schrieb Sylwester Nawrocki:
> Hello Philipp,
> 
> On 25/06/15 12:01, Philipp Zabel wrote:
> > Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
> > ---
> >  drivers/media/v4l2-core/v4l2-mem2mem.c | 9 ++++++++-
> >  1 file changed, 8 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/media/v4l2-core/v4l2-mem2mem.c b/drivers/media/v4l2-core/v4l2-mem2mem.c
> > index dc853e5..511caaa 100644
> > --- a/drivers/media/v4l2-core/v4l2-mem2mem.c
> > +++ b/drivers/media/v4l2-core/v4l2-mem2mem.c
> > @@ -357,9 +357,16 @@ int v4l2_m2m_reqbufs(struct file *file, struct v4l2_m2m_ctx *m2m_ctx,
> >  		     struct v4l2_requestbuffers *reqbufs)
> >  {
> >  	struct vb2_queue *vq;
> > +	int ret;
> >  
> >  	vq = v4l2_m2m_get_vq(m2m_ctx, reqbufs->type);
> > -	return vb2_reqbufs(vq, reqbufs);
> > +	ret = vb2_reqbufs(vq, reqbufs);
> > +	/* If count == 0, then the owner has released all buffers and he
> > +	   is no longer owner of the queue. Otherwise we have a new owner. */
> > +	if (ret == 0)
> > +		vq->owner = reqbufs->count ? file->private_data : NULL;
> > +
> > +	return ret;
> >  }
> 
> How about modifying v4l2_m2m_ioctl_reqbufs() instead ?

The coda, gsc-m2m, m2m-deinterlace, mx2_emmaprp, and sh_veu drivers all
have their own implementation of vidioc_reqbufs that call
v4l2_m2m_reqbufs directly.
Maybe this should be moved into v4l2_m2m_ioctl_reqbufs after all drivers
are updated to use it instead of v4l2_m2m_reqbufs.

> Moreover, does it really makes sense when a new m2m device context
> is being created during each video device open()?

Having the queue owner's device minor in the trace output is very useful
when tracing a single stream across multiple devices. To discern events
from multiple simultaneous contexts I have added the context id to the
coda driver specific trace events.

regards
Philipp