On Wed, 18 May 2022 at 15:34, Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> wrote: > > On 18/05/2022 15:25, Robert Marko wrote: > >> I think we misunderstood each other. Old bindings indeed did not require > >> the interrupts, although if present they should be always defined. > >> Therefore here you should specify number of items and their names. > > > > Yeah, I think we are misunderstanding each other. > > > > Old text-based bindings specified the interrupts, but no naming or > > number was enforced, > > so I looked into the driver to see what is going on. > > Only pm8941 has interrupts defined in the driver and DTS, so I added > > those based on compatible > > matching, the same as with supplies. > > My logic was that it was only valid for interrupts to be described if > > PM8941 was used as describing > > interrupts for other regulator models will do nothing. > > Indeed, you're right, thanks for explanation. Your patch in such case is > correct way of conversion but allows any number of interrupts with any > names, so it's to relaxed. Maybe then better go to previous version, > where these interrupts were defined only for one variant. For other > variants they would fail on as unevaluated? Yeah, that was my intention with not having interrupts as the generic property. I will remove them as generic property and only allow them per compatible, cause I tested adding interrupts to a PMIC DTS that does not allow them and make dtbs_check will warn about those being unevaluated. Regards, Robert > > Best regards, > Krzysztof
