On 06/26/2012 11:07 PM, Thierry Reding wrote:

On Tue, Jun 26, 2012 at 04:48:14PM -0600, Stephen Warren wrote:

...

I actually like what you proposed above a lot, so if you don't
mind either way I'll go with that proposal. Keeping the connector
nodes as children of the outputs has the advantage of being able to
reference them if we need it at some point. But it is also
redundant in that a single output doesn't usually (never?) driver
more than one connector.

Yes, I believe that each output is 1:1 with (the video portion of) a
connector. The display controllers obviously aren't 1:1.

The same issue will have to be addressed for the CSI and VI nodes,
but as I currently use neither of those I don't feel qualified to
propose a binding for them. Also for the VI part we're completely
missing documentation. Maybe somebody could push this to be
released as well?

I did file a bug noting the request for VI documentation. At this
point in time, it's too early to say what, if anything, will come of that.

If I understand correctly, most of the host1x children can also be 
chained in a processing pipeline to do postprocessing an video
input for example. I suppose that's generic and doesn't need to be
represented in DT either, right?

Yes, I believe that's something internal to the driver.