On Fri, Apr 26, 2024 at 6:42 AM Patrick DELAUNAY <patrick.delaunay@xxxxxxxxxxx> wrote: > > Hi, > > On 4/25/24 18:30, Rob Herring wrote: > > On Thu, Apr 25, 2024 at 09:48:31AM +0200, Patrick Delaunay wrote: > >> This patchset removes the unexpected comma in the PWR compatible > >> "st,stm32mp1,pwr-reg" and uses a new compatible "st,stm32mp1-pwr-reg" > >> in STM3MP15 device trees. > > Why? I don't see any warnings from this. Yes, we wouldn't new cases > > following this pattern, but I don't think it is worth maintaining > > support for both strings. We're stuck with it. And the only way to > > maintain forward compatibility is: > > > Yes, no warning because the compatible string are not yet checked by tools. What do you mean? There's a schema for it, so it is checked. I ran the tools and there's no warning. If there was a warning, I'd fix the tools in this case. > I propose this patch to avoid the usage of this compatible for other SoC > in STM32MP1 family. > > > I see the invalid compatible string when I reviewed the U-Boot patch to > add the PWR node for STM32MP13 family: > > https://patchwork.ozlabs.org/project/uboot/patch/20240319024534.103299-1-marex@xxxxxxx/ > Perhaps you should add SoC specific compatible string instead. > So I prefer change the PWR binding before to have the same patch applied > on Linux device tree > > > compatible = "st,stm32mp1-pwr-reg", "st,stm32mp1,pwr-reg"; > > > Yes, I will update the SoC patch with you proposal. NO! We don't want to support that. We have *tons* of examples in DT which don't follow recommended patterns and we're stuck with them. This is no different. We can get away with changing node names, but that's about it. Rob