Re: [PATCH v3 13/15] media: qcom: camss: Add CAMSS Offline Processing Engine driver

[Hans de Goede]

Hi All,

On 20-May-26 17:32, johannes.goede@xxxxxxxxxxxxxxxx wrote:
> 3. Similar to the ope_input /dev/video# to ope_proc pad 0 propagation,
>    for the try stuff automatically propagating info from ope_proc pad 2
>    to ope_disp pad 0 will also not work. In this case it might be best to
>    just completely disable auto-propagation and rely on userspace to do
>    this itself even for the active state. Or you could do the
>    v4l2_subdev_get_locked_active_state() thing again.
> 
>    You can check userspace has done this properly by calling
>    media_pipeline_start() at stream-on time, which ensures that enabled
>    source -> sink pad links agree on the format on both sides.
>    You should probably call media_pipeline_start() regardless.

Correction this is superseeded by note 6:

> 6. There should be no propagation from /dev/video# nodes to pads, instead
>    a validate_link media-entity op should be used to check userspace has
>    properly made these line up on stream-start

So no propagation from [try|set]_fmt on /dev/video# to subdev pads should
happen at all.

Regards,

Hans

(also sorry for not cutting the large rest of the patch in my previous reply)