[PATCH v4 14/34] drm/exynos: hdmi: remove the i2c drivers and use devtree

[Tomasz Stanislawski]

Hi Inki,
Sorry for a very late reply.

On 02/19/2014 12:14 PM, Inki Dae wrote:
> Hi Tomasz,
> 
> 
> 2014-02-14 23:13 GMT+09:00 Tomasz Stanislawski <t.stanislaws at samsung.com>:
>> Hi Daniel,
>> I think that it would be better to change the semantic of phy and ddc
>> bindings.
>>
>> Rather than pointing to I2C client it should point to I2C bus instead.
>> The exynos DRM driver can create dummy I2C clients using i2c_new_dummy()
>> function.
> 
> It seems better way. As of now, all we need for HDMI DDC is
> i2c_adapter (including i2c phy), not i2c_client.
> 
>>
>> There is quite strong rationale for this:
>> - DDC is always accessible on fixed addresses described in HDMI specification.
>> - HDMIPHY (including I2C address) is a part of HDMI IP and it is bound to
>>   specific version of Exynos SoC
>> - no need to add virtual nodes in DTS
> 
> You mean hdmiddc and hdmiphy nodes? If so, I think they are real
> nodes, not virtual nodes. Otherwise, plz give me more comments.
> 
> Thanks,
> Inki Dae
> 

The problem is that those nodes have no dedicated driver.
Moreover, the I2C address for HDMIPHY is always present on
a fixed address dependant on SoC (read HDMI IP) version.
Moreover, both devices share HDMI and HDMIPHY share registers
from HDMI's pool.
Additionally, the hdmiphy I2C client is currently configured using
exynos-hdmi code its DRM driver.
So technically the hdmiphy bus is only a resource used by
exynos-drm-hdmi driver.

Regards,
Tomasz Stanislawski

>> - NVIDIA already use bindings to DDC bus instead of DDC client. Take a look to
>>   arch/arm/boot/dts/tegra*.dts
>>
>> Regards,
>> Tomasz Stanislawski
>>
>> _______________________________________________
>> dri-devel mailing list
>> dri-devel at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/dri-devel