Hi Laurent,

Am Mittwoch, den 19.03.2014, 00:43 +0100 schrieb Laurent Pinchart:

Hello,

On Tuesday 18 March 2014 09:43:54 Philipp Zabel wrote:

quoted

Am Dienstag, den 18.03.2014, 10:15 +0200 schrieb Tomi Valkeinen:

quoted

Add DT binding documentation for HDMI Connector.

Signed-off-by: Tomi Valkeinen <redacted>
Reviewed-by: Archit Taneja <redacted>
---

 .../devicetree/bindings/video/hdmi-connector.txt   | 28 +++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644
 Documentation/devicetree/bindings/video/hdmi-connector.txt> 

diff --git a/Documentation/devicetree/bindings/video/hdmi-connector.txt

b/Documentation/devicetree/bindings/video/hdmi-connector.txt new file
mode 100644
index 000000000000..ccccc19e2573

--- /dev/null
+++ b/Documentation/devicetree/bindings/video/hdmi-connector.txt

@@ -0,0 +1,28 @@
+HDMI Connector
+=======
+
+Required properties:
+- compatible: "hdmi-connector"
+- type: the HDMI connector type: "a", "b", "c", "d" or "e"
+
+Optional properties:
+- label: a symbolic name for the connector
+
+Required nodes:
+- Video port for HDMI input

Geert's comment also applies to all other connector types. These can be
input connectors, too.

We might not need to define all the properties required by input connectors 
now, but we need to make sure that future extensions will be backward-
compatible. I don't see a problem in making the connector DT bindings depend 
on the direction as long as the direction is specified in the DT node, either 
explicitly or implicitly.

An obvious solution would be to have separate "hdmi-input-connector" and 
"hdmi-output-connector" compatible strings but I don't like that, as there's 
no difference in the HDMI connector itself, only in the usage.

I don't think this is necessary, either. I just meant the wording for
the video port should leave the direction unspecified. I imagine
somebody somewhere will connect a HDMI connector to a mux so that it can
be either input or output.

We could also add a "direction" property (possibly later, with a default value 
of "output" in that case), or rely on the link direction, but in the latter 
case that would require reaching an agreement in the current link directions 
debate.

If we add a direction property, I think it should be on the ports.

regards
Philipp