On Fri, Dec 06, 2024 at 05:44:15PM +0000, Sverdlin, Alexander wrote: > Hello Conor, > > On Fri, 2024-12-06 at 17:14 +0000, Conor Dooley wrote: > > > Add Texas Instruments' LP8864/LP8866 bindings into LP8860 converting them > > > into YAML format simultaneously. While here, drop the index of the "led" > > > subnode, this one is neither used nor mandated by the drivers. All the > > > *-cells properties are therefore not required. > > > > Are you sure this is a correct thing to do? The lp8860-q1 product link > > cites it as being a 4-channel device. Even if the kernel only ever > > supports it as a single-channel device, the binding should reflect what > > it is capable of doing. > > my understanding is: > - The whole family is multi-channel (4 or 6), but this is rather used internally > in the chip for power balancing; separate diagnostics is provided. From the user > perspective one has only one brightness per chip. One brightness perhaps, but what do you do where several LEDs of different colours are connected? Or if one was to be active-low? I don't see the benefit of changing the binding in a way that makes it less able to describe the hardware. > - The lp8860 driver didn't attempt to do anything with the index. I don't see this as being relevant, the bindings need only address what the hardware is able to do. The driver may only implement a subset of that, and that is perfectly okay. > I'm glad Andrew Davis had a time for a pre-public review of the new binding > and actually suggested this simplification. Okay. Thanks, Conor.
Attachment:
signature.asc
Description: PGP signature
