On Mon, Sep 29, 2014 at 02:10:07PM +0100, Stefan Wahren wrote:
> Hi,
>
> Am 29.09.2014 um 14:41 schrieb Mark Rutland:
> > Well, the simple-bus will cause the children to be probed. But it looks
> > like you care about properties of the parent. I don't think that
> > simple-bus is appropriate because it's not being handled as a
> > transparent bridge from the PoV of the children.
>
> actually i need the address of the power status register. In this
> version i get the base address from the parent node add an offset.
>
> Do you prefer to define the address of the power status register like a
> second address cell:
>
> reg_vddd: regulator@80044040 {
> reg = <0x80044040 0x10
> 0x800440c0 0x01>;
> ...
> };
>
> or do i need special properties like this:
>
> reg_vddd: regulator@80044040 {
> reg = <0x80044040 0x10>;
> fsl,mxs-status-reg = <0x800440c0>;
> ...
> };
I would prefer a top level node for the subsystem that is not a
simple-bus.
Give it a compatible string and a well-defined set of base properties
(looks like you just need the reg for now). Match that and probe the
child nodes as appropriate.
> >> Do we need a extra driver?
> > Perhaps, but it's relatively simple to match on a compatible string and
> > probe children if you just wantto start small for now.
>
> Okay. Would be great if someone has a good example. At first, i thought
> of power/anatop.
While I believe there are examples in the kernel, I can't think
immediately of any instances.
> >>>> +- #address-cells: Number of cells required to define regulator register,
> >>>> + must be 1
> >>>> +- #size-cells: Number of cells required to define register size, must be 1
> >>> Why must this be the case, given that the child node expects an absolute
> >>> physical address?
> >> I need a property to define the control register for the regulators
> >> without defining vendor specific properties like "fsl,mxs-control-reg"
> >> or something.
> > You misunderstand me. I was querying the "must be 1" rather than the
> > proeprties themselves.
> >
> >>> What's wrong with #address-cells = <2>, for example?
> >> Nothing
> > Then we shouldn't specify "must be 1", no?
>
> Right, must be at least 1.
Why not just say that #address-cells, #size-cells and ranges must be
present as appropriate to map children?
>
> >>>> +- reg: Absolute physical address and size of the register set for the device
> >>> Why is this here _and_ in the child node(s)?
> >> The parent of the power node is also a simple bus. I use this to
> >> calculate the power status register per offset.
> >>
> >>> What is the difference between this node and its children?
> >> The parent node represent the power sub system and the regulators are
> >> part of this sub system.
> >>
> >>> Can there be more than one sub-node?
> >> In the i.MX28 are at least 4 voltage regulators, 1 current regulator and
> >> many more. At first, the driver should implement only 3 voltage
> >> regulators (vddd, vdda, vddio).
> > Ok.
> >
> > I think you need a binding for the power subsystem, and a trivial driver
> > that can match on that and probe the child regulators. Are there
> > components other than voltage or current regulators in the sub system?
>
> Yes, according to the reference manual there is a dc-dc converter, a
> battery charger, battery monitor, ...
>
> In short a lot of developing time ;-)
Sure, but not everything needs to be supported fomr the outset.
Mark.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
