Hi Bjorn, On Monday 05 Dec 2016 13:11:51 Bjorn Andersson wrote: > On Tue 29 Nov 01:11 PST 2016, Laurent Pinchart wrote: > > On Tuesday 29 Nov 2016 13:41:33 Archit Taneja wrote: > >> On 11/29/2016 12:03 PM, Laurent Pinchart wrote: > >>> On Tuesday 29 Nov 2016 11:37:41 Archit Taneja wrote: > >>>> Add the regulator supply properties needed by ADV7511 and ADV7533. > >>>> > >>>> The regulators are specified as optional properties since there can > >>>> be boards which have a fixed supply directly routed to the pins, and > >>>> these may not be modelled as regulator supplies. > >>> > >>> That's why we have support for dummy supplies in the kernel, isn't it > >>> ? Isn't it better to make the supplies mandatory in the bindings (and > >>> obviously handling them as optional in the driver for > >>> backward-compatibility) ? > >> > >> I'm a bit unclear on this. > >> > >> I thought we couldn't add mandatory properties once the device is > >> already present in DT for one or more platforms. > > > > You can, as long as you treat them as optional in the driver to retain > > backward compatibility. The DT bindings should document the properties > > expected from a new platform (older versions of the bindings will always > > be available in the git history). > > If you document them as required and don't do anything special in the > implementation (i.e. just call devm_regulator_get() as usual) it will > just work, in the absence of the property you will get a dummy regulator > from the framework. > > And then add the fixed-voltage regulators to the new DT to make that > properly describe the hardware. > > >> Say, if we do make it mandatory for future additions, we would need to > >> have DT property for the supplies for the new platforms. If the > >> regulators on these boards are fixed supplies, they would be need to be > >> modeled using "regulator-fixed", possibly without any input supply. Is > >> that what you're suggesting? > > > > That's the idea, yes. Clock maintainers have a similar opinion regarding > > the clock bindings, where a clock that is not optional at the hardware > > level should be specified in DT even if it's always present. > > Further more, a DT binding for a particular block should describe that > block; so if we have three different 1.8V pins then the DT binding > should reflect this - even if our current platform have them wired to > the same regulator. This has been discussed previously, and Rob agreed that if the datasheet recommends to power all supplies from the same regulator we can take that as a good hint that a single supply should be enough. In the very unlikely event that a board would require control of more regulators we can always extend the DT bindings later without breaking backward compatibility. > (And the supply names would preferably be based on the pin names in the > component data sheet) -- Regards, Laurent Pinchart -- 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
