On 09/04/2018 10:56 AM, Mark Brown wrote: > On Tue, Sep 04, 2018 at 10:10:24AM -0500, Andrew F. Davis wrote: >> On 09/04/2018 09:55 AM, Mark Brown wrote: > >>> you get to keep the pieces" territory. If we really want to have a way >>> of explicitly specifying that some widgets should never be turned on >>> then it feels like rather than have something device and widget specific >>> like this we should instead have a higher level way of doing that which >>> can be applied to any widget, that is something that could be useful >>> especially with things like speaker drivers where there's real potential >>> for physical damage to the system. > >> Not sure how that could even look, it would need to be in DT as only >> that layer knows the connections, but then the it would also have to >> have internal knowledge of the widgets used in the driver.. > > We already have a bunch of DT code that knows about and uses the > external widgets the device has (which are I'd guess the only ones where > this is an issue)? I was thinking just have a property that lists the > widgets that should never be turned on. > >>> Like I said in my original reply I'm also worried that this will break >>> existing boards by causing them to change to a voltage of 0 when they >>> had managed to end up with the default of 2V which happened to work for >>> them. > >> Then only time this could happen is if they specified MICBIAS_OFF in DT >> but got the default 2v instead. In which case, yes, the behavior would >> change, but it would also be changing to the correct behavior. We can't >> avoid fixing code on the off chance someone depended on the broken >> behavior, no progress could ever be made. > > I'm really having a lot of trouble seeing MICBIAS_OFF as a useful > voltage to specify in DT in the first place. > I don't see the usefulness in specifying any bias voltage in DT at all, it is a configuration and can be made at runtime, it has no place in DT. But it is already here, so lets allow all available voltages a board may need and the CODEC can supply, even 0.