On Tuesday 26 June 2012 04:39 AM, Stephen Warren wrote:

On 06/25/2012 04:26 PM, Mark Brown wrote:

quoted

On Mon, Jun 25, 2012 at 08:54:21PM +0530, Laxman Dewangan wrote:

quoted

I had detailed discussion with Mark on  this support and as per
him (based on my understanding), the input to different regulator
is from the pin of the chips and so the name should be the
<pin-name>-supply which should be part of chip-dt binding, not
to the particular rail.

More specifically, all the supplies for a device (including those
that happen to be inputs for regulators) should be specified in
exactly the same fashion.  This makes the binding more regular and
means that users can just go through the schematic adding the
mappings without worrying about what what the supply happens to
be.

Just making sure I parsed that right. I think what you're saying is
that the device itself should represent its input pins, e.g.:

tps6586x {
     vin-ldo01-supply =<&some_regulator>;

I think it is fine. The pin name as per data sheet is VINLDO01 and so we 
can have vin-ldo01.

     vin-ldo23-supply =<...>;
     vin-ldo4-supply =<...>;
     vin-ldo678-supply =<...>;
     vin-ldo9-supply =<...>;

     regulators {
         regulator at 0 {
             regulator-compatible = "ldo0";
             ...
         };
         regulator at 1 {
             regulator-compatible = "ldo1";
             ...
         };
     };
};

(and then the driver internally uses the *-supply to set up the
parents of each of its own regulators)

... rather than each regulator specifying its parent, which might
result in some duplication, since in this case both ldo0/1 are
supplied from the same input pin:

tps6586x {
     regulators {
         regulator at 0 {
             regulator-compatible = "ldo0";
             vin-supply =<&some_regulator>;
             ...
         };
         regulator at 1 {
             regulator-compatible = "ldo1";
             vin-supply =<&some_regulator>;
             ...
         };
     };
};