Re: [PATCH/RFC] serial: sh-sci: Add device tree support for r8a7779

[Simon Horman <horms@verge.net.au>]

On Mon, Apr 28, 2014 at 02:08:16AM +0200, Laurent Pinchart wrote:
> Hi Simon,
> 
> On Monday 28 April 2014 09:03:14 Simon Horman wrote:
> > On Fri, Apr 25, 2014 at 09:11:06AM +0200, Geert Uytterhoeven wrote:
> > > On Fri, Apr 25, 2014 at 9:05 AM, Laurent Pinchart wrote:
> > > >> +     scif0: serial@ffe40000 {
> > > >> +             compatible = "renesas,scif", "renesas,scif-r8a7779";
> > > >> +             reg = <0xffe40000 265>;
> > > >> +             interrupt-parent = <&gic>;
> > > >> +             interrupts = <0 88 IRQ_TYPE_LEVEL_HIGH>;
> > > >> +             clocks = <&cpg_clocks R8A7779_CLK_P>;
> > > >> +             clock-names = "sci_ick";
> > > > 
> > > > Clock handling in the sh-sci driver should probably be improved. The
> > > > driver currently requires an "sci_ick" interface clock and supports an
> > > > optional "sci_fck" functional clock. In practice, as far as I can see,
> > > > platforms that provide both sci_ick and sci_fck set the two clocks to
> > > > the same source.
> >
> > > That's right. As a consequence, the clock's enable count is incremented
> > > 
> > > 3 times:
> > >   - once for fck,
> > >   - once for ick,
> > >   - once for generic Runtime PM using the "NULL" clock.
> > 
> > This approach is fine by me.
> > But I think you need to maintain compatibility with the old
> > binding ("sci_ick" required, "sci_fsk") as it seems that
> > was included in v3.14.
> 
> sci_fck isn't part of the DT bindings, so we can at least drop that one.

So long as its not used anywhere in-tree, which I believe is the case,
then that is fine by me.