Re: [PATCH v8 1/2] media: dt-bindings: media: xilinx: Add Xilinx MIPI CSI-2 Rx Subsystem

[Sakari Ailus]

Hi Vishal,

On Thu, Jun 06, 2019 at 11:54:19AM +0000, Vishal Sagar wrote:
> Hi Sakari,
> 
> > -----Original Message-----
> > From: Sakari Ailus [mailto:sakari.ailus@xxxxxxxxxxxxxxx]
> > Sent: Wednesday, June 05, 2019 12:54 AM
> > To: Vishal Sagar <vishal.sagar@xxxxxxxxxx>
> > Cc: Hyun Kwon <hyunk@xxxxxxxxxx>; laurent.pinchart@xxxxxxxxxxxxxxxx;
> > mchehab@xxxxxxxxxx; robh+dt@xxxxxxxxxx; mark.rutland@xxxxxxx; Michal
> > Simek <michals@xxxxxxxxxx>; linux-media@xxxxxxxxxxxxxxx;
> > devicetree@xxxxxxxxxxxxxxx; hans.verkuil@xxxxxxxxx; linux-arm-
> > kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Dinesh Kumar
> > <dineshk@xxxxxxxxxx>; Sandip Kothari <sandipk@xxxxxxxxxx>; Luca Ceresoli
> > <luca@xxxxxxxxxxxxxxxx>; Jacopo Mondi <jacopo@xxxxxxxxxx>
> > Subject: Re: [PATCH v8 1/2] media: dt-bindings: media: xilinx: Add Xilinx MIPI
> > CSI-2 Rx Subsystem
> > 
> > EXTERNAL EMAIL
> > 
> > Hi Vishal,
> > 
> > On Mon, Jun 03, 2019 at 03:29:30PM +0530, Vishal Sagar wrote:
> > 
> > > +Optional properties:
> > > +--------------------
> > > +- .
> > > +  Without this property the driver won't be loaded as IP won't be able to
> > generate
> > > +  media bus format compliant stream output.
> > 
> > I think we previously concluded that the format will be just different in
> > this case. So the description appears incorrect.
> > 
> 
> Referring to your email https://lkml.org/lkml/2019/3/22/1823 in response to v6 patches,
> if the video format bridge is not enabled, then the way in which pixels are transported on
> the bus don't correspond to the existing media bus formats in Linux.
> 
> If not loading the driver is incorrect way, is it ok for the driver to allow same media bus format
> for packed and unpacked data type on the sink pad?
> 
> Or is it ok for the driver to not validate the media bus format set on the sink pad?

Taking a fresh look at the issue --- usually such unpacking is done by the
DMA engine, or the same device contains both the CSI-2 RX and DMA. But here
it actually affects the input of that DMA engine. You're right in saying we
don't have format definitions from which you could tell which case it is,
and we also don't have other pre-existing means to tell them apart.

Feel free to keep the check in the driver, but we can't refer to the driver
loading in DT binding documentation: this is really not supposed to be
related to that driver, or even Linux at all.

How about changing this to:

xlnx,vfb: Present when Video Format Bridge is enabled in IP configuration.

That'd be aligned with the other properties and would more accurately
convey what this means.

-- 
Kind regards,

Sakari Ailus
sakari.ailus@xxxxxxxxxxxxxxx