On Tue, Oct 07, 2014 at 05:49:24PM +0300, Laurent Pinchart wrote:
> Hi Ajay,
> 
> On Tuesday 07 October 2014 16:06:55 Ajay kumar wrote:
> > On Tue, Oct 7, 2014 at 4:00 PM, Tomi Valkeinen wrote:
> > > On 20/09/14 14:22, Ajay kumar wrote:
> > >> Well, I am okay with using video ports to describe the relationship
> > >> between the encoder, bridge and the panel.
> > >> But, its just that I need to make use of 2 functions when phandle
> > >> does it using just one function ;)
> > >> -        panel_node = of_parse_phandle(dev->of_node, "panel", 0)
> > >> +       endpoint = of_graph_get_next_endpoint(dev->of_node, NULL);
> > >> +       if (!endpoint)
> > >> +               return -EPROBE_DEFER;
> > >> +
> > >> +       panel_node = of_graph_get_remote_port_parent(endpoint);
> > >> +       if (!panel_node)
> > >> +               return -EPROBE_DEFER;
> > >> 
> > >> 
> > >> If nobody else has objections over using of_graph functions instead
> > >> of phandles, I can respin this patchset by making use of video ports.
> > > 
> > > The discussion did digress somewhat.
> > > 
> > > As a clarification, I'm in no way nack'ing this series because it
> > > doesn't use the graphs for video connections. I don't see the simple
> > > phandle bindings used here as broken as such.
> > 
> > Well, I am okay with any approach you guys decide on. I desperately want
> > this to get this in since it has been floating around for quite sometime.
> > The more we drag this, the more rework for me since the number of platforms
> > using bridge support is increasing daily!
> 
> I won't nack this patch either. I'm however concerned that we'll run straight 
> into the wall if we don't come up with an agreement on a standard way to 
> describe connections in DT for display devices, which is why I would prefer 
> the ps8622 bindings to use OF graph to describe connections.

I think there's not really an easy way out here. It's pretty bold trying
to come up with a common way to describe bridges when we have only a
single one (and a single use-case at that). The worst that can happen is
that we need to change the binding at some point, in which case we may
have to special-case early drivers, but I really don't think that's as
much of an issue as everybody seems to think.

This series has been floating around for months because we're being
overly prudent to accept a binding that /may/ turn out to not be generic
enough.

Thierry