On Monday 23 September 2013 15:33:10 Stephen Warren wrote: > On 09/23/2013 05:50 AM, Laurent Pinchart wrote: > > On Monday 23 September 2013 08:18:52 Prabhakar Lad wrote: > >> On Fri, Sep 20, 2013 at 3:22 PM, Sylwester Nawrocki wrote: > >>> On 09/20/2013 10:11 AM, Prabhakar Lad wrote: > >>>> OK I will, just send out a fix up patch which fixes the mismatch > >>>> between > >>>> names for the rc-cycle, and later send out a patch which removes the > >>>> platform data usage for next release with proper DT bindings. > >>> > >>> I think the binding need to be fully corrected now, I just meant to not > >>> touch the board file, i.e. leave non-dt support unchanged. > >> > >> Ok > >> > >>>> I'm OK with making regulator properties as optional, But still it would > >>>> change the meaning of what DT is, we know that the VDD/VDD_IO .. etc > >>>> pins are required properties (but still making them as optional) :-( > >>>> > >>>> I think there might several devices where this situation may arise so > >>>> just thinking of a alternative solution. > >>>> > >>>> say we have property 'software-regulator' which takes true/false(0/1) > >>>> If set to true we make the regulators as required property or else we > >>>> assume it is handled and ignore it ? > >>> > >>> I don't think this is a good idea. You would have to add a similar > >>> platform data flag for non-dt, it doesn't sound right. I can see two > >>> options here: > >>> > >>> 1. Make the regulator properties mandatory and, e.g. define a fixed > >>> > >>> voltage GPIO regulator in DT with an empty 'gpio' property. Then > >>> pass a phandle to that regulator in the adv7343 *-supply properties. > >>> For non-dt similarly a fixed voltage regulator(s) and voltage > >>> supplies would need to be defined in the board files. > >>> > >>> 2. Make the properties optional and use (devm_)regulator_get_optional() > >>> > >>> calls in the driver (a recently added function). I must admit I don't > >>> fully understand description of this function, it currently looks > >>> pretty much same as (devm_)regulator_get(). Thus the driver would > >>> need to be handling regulator supplies only when non ERR_PTR() is > >>> returned from regulator_get_optional() and otherwise assume a non > >>> critical error. There is already quite a few example occurrences of > >>> regulator_get_optional() usage. > >> > >> Thanks for pointing it I'll choose option 2 and post the patch. > > > > Isn't regulator_get_optional() intended for devices that can have supplies > > unconnected in normal use ? > > I believe so, yes. > > > The ADV7343 supplies are mandatory from a hardware > > point of view, so I think we should use regulator_get(). Otherwise the > > driver won't be able to tell the difference between a regulator that > > isn't present yet (for instance because the regulator device/driver > > hasn't been probed yet), which should result in deferred probing, and an > > always-on regulator that has been left out. > > So I think you want to make the supply properties mandatory in DT (since > some form of supply is mandatory in HW), yet make the driver support > broken DTs which don't have those properties, by error-checking the > return value from regulator_get(). You might want to put a note into DT > saying that a previous version of the binding didn't require those > supply properties, so they may be missing from older DTs. Are there such devices in the wild ? -- 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
