[PATCH 1/3] drm/exynos: add super device support

[Inki Dae]

Hi Philipp,


2014-03-27 23:46 GMT+09:00 Philipp Zabel <p.zabel at pengutronix.de>:
> Hi Inki,
>
> Am Donnerstag, den 27.03.2014, 21:43 +0900 schrieb Inki Dae:
>> This patch adds super device support to bind sub drivers
>> using device tree.
>>
>> For this, you should add a super device node to each machine dt files
>> like belows,
>>
>> In case of using MIPI-DSI,
>>       exynos-drm {
>>               compatible = "samsung,exynos-drm";
>>               crtcs = <&fimd>;
>>               connectors = <&dsi>;
>>       };
>
> Russell had suggested a similar binding for i.MX, but we have since
> changed it to look like this instead:
>
> -       imx-drm {
> -               compatible = "fsl,imx-drm";
> -               crtcs = <&ipu 0>, <&ipu 1>;
> -               connectors = <&hdmi>, <&ldb>;
> -       };
> +       display-subsystem {
> +               compatible = "fsl,imx-display-subsystem";
> +               ports = <&ipu_di0>, <&ipu_di1>;
> +       };
>
> The ports are the two display output port nodes of the image processing
> unit (corresponding to a drm_crtc each), and the imx-drm driver that
> binds to the compatible value "fsl,imx-display-subsystem" automatically
> collects all encoders that hang off of those via the OF graph bindings
> (as documented in Documentation/devicetree/bindings/graph.txt in the
> staging tree). I don't know how well those bindings would fit on Exynos,
> maybe you could consider aligning your bindings with those.
>
> Either way, the generic term 'display-subsystem' is preferable to 'drm',
> as the latter is Linux specific, which the device tree bindings
> shouldn't be. The same goes for the 'crtcs' property. I'd prefer to
> avoid Linux specific legacy names in new device tree bindings.

Agree. I also thought that using property names, drm, crtcs and
connectors,  specific to Linux is not reasonable. It looks good to use
'display-subsystem' instead of 'drm'.
But only using 'ports' property for only crtc device is not clear to me.

In case of Exynos SoC, there are various types of bridge device, image
enhancer between crtc and encoder device, and lvds device between
encoder and panel device.
So I think Exynos drm core would be complicated with some binding
codes if we use only 'ports' property.
I guess 'imx_drm_components' list specific to imx drm would be among
them, and that would be because you could resolve the probe order
issue only using existing component framework.

Below link shows how we are trying to handle such various bridge devices in drm,
        http://www.spinics.net/lists/dri-devel/msg55555.html

And this patch series is first step for it.

I'd like to use only component framework without specific list somehow
but we should find generic property names for encoder/connector and
bridge devices.

Thanks,
Inki Dae

>
> Also, since this adds new bindings, it should probably be sent to the
> devicetree mailing list and include some documentation.
>
> If you are interested in the previous discussion on the imx-drm
> supernode, here is the thread history:
>     "[PATCH v5 00/11] imx-drm dt bindings"
>     http://www.spinics.net/lists/arm-kernel/msg313112.html
>     "[PATCH v4 00/11] imx-drm dt bindings"
>     http://www.spinics.net/lists/dri-devel/msg54353.html
>     "[RFC PATCH v3 0/9] imx-drm dt bindings"
>     http://www.spinics.net/lists/arm-kernel/msg308529.html
>     "[RFC PATCH v2 0/4] imx-drm dt bindings"
>     http://www.spinics.net/lists/arm-kernel/msg306649.html
>     "[RFC PATCH 0/3] imx-drm dt bindings"
>     http://www.spinics.net/lists/arm-kernel/msg298290.html
>
> regards
> Philipp
>
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel