Hi,
On Tue, Jun 18, 2013 at 11:24:40AM +0200, Sebastian Andrzej Siewior wrote:
> On 06/18/2013 10:54 AM, Felipe Balbi wrote:
> > Hi,
>
> Hi Felipe,
>
> >> This isn't exactly a child node, is it? There is
> >> of_get_child_count() but this isn't a child, it is one property.
> >
> > is it because we haven't added DTS support for musb core ?
> > Eventually we can/should add and convert this to
> > of_get_child_count() then. For now, we can go ahead with your
> > approach.
>
> mother {
> child1 {
> };
> child1 {
> };
> };
>
> That would be a child imho.
> For counting of phandles (for the number of assigned gpios for instance)
> you would use
> of_count_phandle_with_args(node, "gpios", "#gpio-cell-size");
>
> We don't have an equivalent for "#gpio-cell-size". In our case it would
> be the sum of "#address-cells" and "#size-cells" minus 1 because we
> don't have a phandle and abuse a function :)
> Counting the number of "addresses" is special since bother its
> members (address and size) can be either 32bit or 64bit in size.
>
> > We have such a large amount of function pointers to sort out anyway
> > :-(
>
> If you want get the while() loop replaced by something else I have a
> few suggestions:
> - check for iomem(2). If it is there we have two instances. If not,
> just 1. This includes also the hope that we don't get a third port.
>
> - loop over of_address_to_resource(). This is what of_device_alloc() is
> doing:
> if (of_can_translate_address(np))
> while (of_address_to_resource(np, num_reg, &temp_res) == 0)
> num_reg++;
> So it is different.
let's keep it as it is, I think once we convert musb-core to dt, all
those issues will vanish and we will be able to use of_get_child_count()
anyway. Forget my noise :-)
--
balbi
Attachment:
signature.asc
Description: Digital signature
