On 10/25/2012 12:32 AM, Mark Zhang wrote:

On 10/23/2012 01:14 AM, Stephen Warren wrote:

...

quoted

I don't recall why pta was defined to be HDMI. The issue isn't that this
patch changes the pinmux selection for the pta pingroup, but simply that
both the pinctrl node's state definition, and the new I2C mux node's
state definition both attempt to configure pingroup pta. The solution is
most likely to simply remove the pta configuration from the main pinctrl
node.

Understood. I tried to remove the hdmi function definition of pta
pingroup yesterday then found HDMI can't work anymore. The EDID of HDMI
monitor can't be fetched. After some debugging, I have found that it's
caused by "i2c-mux-pinctrl" driver is loaded after drm driver. That
makes drm driver can't get EDID data via this i2c mux adapter because it
doesn't exist at that time. So I think we need to promote the load level
of "i2c-mux-pinctrl" driver. Any other ideas?

(BTW, your message that I'm replying to wasn't word-wrapped correctly)

No, i2c-mux-pinctrl shouldn't need to change. It sounds like tegra-drm
isn't supporting deferred probe correctly; when it needs access to an
I2C adapter that doesn't exist, it should defer its own probe until the
I2C adapter does exist.