Re: [PATCH v4 2/2] media: dt-bindings: media: st,stm32-dcmi: Add support of BT656

From: Sakari Ailus
Date: Fri Oct 30 2020 - 13:42:49 EST

Hi Rob,

On Mon, Oct 26, 2020 at 09:17:14AM -0500, Rob Herring wrote:
> On Thu, Oct 22, 2020 at 02:56:17PM +0000, Hugues FRUCHET wrote:
> > Hi Sakari,
> >
> > + Jacopo for his work on ov772x binding related to BT656
> >
> > On 10/21/20 11:40 PM, Sakari Ailus wrote:
> > > Hi Hugues,
> > >
> > > On Wed, Oct 21, 2020 at 02:24:08PM +0000, Hugues FRUCHET wrote:
> > >> Hi Sakari,
> > >>
> > >> On 10/21/20 3:00 PM, Sakari Ailus wrote:
> > >>> Hi Hugues,
> > >>>
> > >>> On Tue, Oct 20, 2020 at 12:14:49PM +0200, Hugues Fruchet wrote:
> > >>>> Add support of BT656 parallel bus mode in DCMI.
> > >>>> This mode is enabled when hsync-active & vsync-active
> > >>>> fields are not specified.
> > >>>>
> > >>>> Signed-off-by: Hugues Fruchet <hugues.fruchet@xxxxxx>
> > >>>> ---
> > >>>> .../devicetree/bindings/media/st,stm32-dcmi.yaml | 30 ++++++++++++++++++++++
> > >>>> 1 file changed, 30 insertions(+)
> > >>>>
> > >>>> diff --git a/Documentation/devicetree/bindings/media/st,stm32-dcmi.yaml b/Documentation/devicetree/bindings/media/st,stm32-dcmi.yaml
> > >>>> index 3fe778c..1ee521a 100644
> > >>>> --- a/Documentation/devicetree/bindings/media/st,stm32-dcmi.yaml
> > >>>> +++ b/Documentation/devicetree/bindings/media/st,stm32-dcmi.yaml
> > >>>> @@ -44,6 +44,36 @@ properties:
> > >>>> bindings defined in
> > >>>> Documentation/devicetree/bindings/media/video-interfaces.txt.
> > >>>>
> > >>>> + properties:
> > >>>> + endpoint:
> > >>>> + type: object
> > >>>> +
> > >>>> + properties:
> > >>>> + bus-width: true
> > >>>> +
> > >>>> + hsync-active:
> > >>>> + description:
> > >>>> + If both HSYNC and VSYNC polarities are not specified, BT656
> > >>>> + embedded synchronization is selected.
> > >>>> + default: 0
> > >>>> +
> > >>>> + vsync-active:
> > >>>> + description:
> > >>>> + If both HSYNC and VSYNC polarities are not specified, BT656
> > >>>> + embedded synchronization is selected.
> > >>>> + default: 0
> > >>>
> > >>> Should I understand this as if the polarities were not specified, BT.656
> > >>> will be used?
> > >>
> > >> Yes, this is what is documented in video-interfaces.txt:
> > >> "
> > >> Note, that if HSYNC and VSYNC polarities are not specified, embedded
> > >> synchronization may be required, where supported.
> > >> "
> > >> and
> > >> "
> > >> /* If hsync-active/vsync-active are missing,
> > >> embedded BT.656 sync is used */
> > >> hsync-active = <0>; /* Active low */
> > >> vsync-active = <0>; /* Active low */
> > >> "
> > >> and I found also this in
> > >> Documentation/devicetree/bindings/media/renesas,vin.yaml
> > >> "
> > >> hsync-active:
> > >> description:
> > >> If both HSYNC and VSYNC polarities are not specified,
> > >> embedded
> > >> synchronization is selected.
> > >> default: 1
> > >>
> > >> vsync-active:
> > >> description:
> > >> If both HSYNC and VSYNC polarities are not specified,
> > >> embedded
> > >> synchronization is selected.
> > >> default: 1
> > >
> > > Having the defaults leads to somewhat weird behaviour: specifying the
> > > default value on either property changes the bus type.
> > >
> > >> "
> > >>
> > >> In the other hand I've found few occurences of "bus-type"
> > >> (marvell,mmp2-ccic.yaml), it is why I asked you if "bus-type" is the new
> > >> way to go versus previous way to signal BT656 (without hsync/vsync) ?
> > >> As explained previously, I prefer this last way for backward compatibility.
> > >
> > > If you have a default for bus-type (BT.601), this won't be a problem.
> > >
> > > The old DT bindings were somewhat, well, opportunistic. The v4l2-of
> > > framework-let did its best and sometimes it worked. The behaviour is still
> > > supported but not encouraged in new bindings.
> > >
> >
> > OK, so let's go for the new way.
> > I've found an interesting patch from Jacopo that is of great help:
> > https://patchwork.ozlabs.org/project/devicetree-bindings/patch/20200910162055.614089-4-jacopo+renesas@xxxxxxxxxx/

I wonder if Jacopo tested it. The idea seems interesting nonetheless.

> >
> > Here is a draft proposal before I push a new version, please comment:
> >
> > properties:
> > bus-type:
> > enum: [5, 6]
> > default: 5
> >
> > bus-width:
> > enum: [8, 10, 12, 14]
> > default: 8
> >
> > hsync-active:
> > enum: [0, 1]
>
> For common properties, you can assume there's a common schema. As 0 and
> 1 are the only possible values, you don't need to define them here
> unless only a subset is valid for this device.
>
> > default: 0
> >
> > vsync-active:
> > enum: [0, 1]
> > default: 0
> >
> > pclk-sample:
> > enum: [0, 1]
> > default: 0
> >
> > remote-endpoint: true
> >
> > allOf:
> > - if:
> > properties:
> > bus-type:
> > const: 6
>
> To fix the error, you need:
>
> required:
> - bus-type
>
> The problem is the above schema is also true if the property
> is not present.

Hmm. The idea was that we could keep this consistent with old bindings that
only documented parallel mode, and thus didn't need bus-type. This is
actually quite common --- adding support for something that wasn't known or
cared for during the original review.

I guess this could be done in the driver, too, adding a comment that the
bindings earlier did not require it.

>
> > then:
> > properties:
> > hsync-active: false
> > vsync-active: false
> > bus-width:
> > enum: [8]
> >
> > required:
> > - remote-endpoint
> >
> > unevaluatedProperties: false
> >
> >
> > Unfortunately, the "default: 5" for bus-type is not working !!
> > If we don't specify "bus-type" in example, dt_binding_check is failing
> > as if default was 6, it's hardly understandable (see below) !
> > port {
> > dcmi_0: endpoint {
> > remote-endpoint = <&ov5640_0>;
> > bus-width = <10>;
> > hsync-active = <0>;
> > vsync-active = <0>;
> > pclk-sample = <1>;
> > };
> > => this should be OK but error claimed:
> > DTC
> > Documentation/devicetree/bindings/media/st,stm32-dcmi.example.dt.yaml
> > CHECK
> > Documentation/devicetree/bindings/media/st,stm32-dcmi.example.dt.yaml
> > Documentation/devicetree/bindings/media/st,stm32-dcmi.example.dt.yaml:
> > dcmi@4c006000: port:endpoint:vsync-active: False schema does not allow [[0]]
> > dcmi@4c006000: port:endpoint:hsync-active: False schema does not allow [[0]]
> > dcmi@4c006000: port:endpoint:bus-width:0:0: 10 is not one of [8]
> > From schema: Documentation/devicetree/bindings/media/st,stm32-dcmi.yaml
> >
> > => if "bus-type" is explicitly set to 5, all is fine (see below) !
> > port {
> > dcmi_0: endpoint {
> > remote-endpoint = <&ov5640_0>;
> > bus-type = <5>;
> > bus-width = <10>;
> > hsync-active = <0>;
> > vsync-active = <0>;
> > pclk-sample = <1>;
> > };
> > };
> >
> > DTC
> > Documentation/devicetree/bindings/media/st,stm32-dcmi.example.dt.yaml
> > CHECK
> > Documentation/devicetree/bindings/media/st,stm32-dcmi.example.dt.yaml
> > ~/.../media_tree$
> >
> >
> > >>
> > >>
> > >> The bindings previously documented BT.601 (parallel) only, so
> > >>> it was somewhat ambigious to begin with. Is there a risk of interpreting
> > >>> old BT.601 bindings as BT.656?
> > >> I don't think so.
> > >>
> > >> With bus-type property, I believe you could
> > >>> avoid at least that risk.
> > >> yes but as explained, I'll prefer not to amend current boards device
> > >> tree files.
> > >
> > > I don't think it matters from this point of view --- you can have a
> > > default bus-type.
> > >
> > >>
> > >>>
> > >>> Also not specifying at least one of the default values leads to BT.656
> > >>> without bus-type. That could be addressed by removing the defaults.
> > >>>
> > >> I'm new to yaml, I've taken that from renesas,vin.yaml. Should I just
> > >> drop the "default: 1" lines ?
> > >
> > > That's one option, yes. Then you have to have those for BT.601 and it's no
> > > longer ambiguous.
> > >
> >
> > BR,
> > Hugues.

--
Kind regards,

Sakari Ailus