[PATCH 00/20] Simplify of_property_for_each_u32()

[Luca Ceresoli]

Hello Rob,

On Wed, 3 Jul 2024 12:07:42 -0600
Rob Herring <robh at kernel.org> wrote:

> On Wed, Jul 03, 2024 at 12:36:44PM +0200, Luca Ceresoli wrote:
> > [Note: to reduce the noise I have trimmed the get_maintainers list
> > manually. Should you want to be removed, or someone else added, to future
> > versions, just tell me. Sorry for the noise.]
> > 
> > This series aims at simplifying of_property_for_each_u32() as well as
> > making it more difficult to misuse it in the future.
> > 
> > The long-term goal is changing this pattern:
> > 
> >   struct property *prop;
> >   const __be32 *p;
> >   u32 val;
> >  
> >   of_property_for_each_u32(np, "xyz", prop, p, val) { ... }
> > 
> > to this:
> > 
> >   u32 val;
> > 
> >   of_property_for_each_u32(np, "xyz", val) { ... }
> > 
> > So, removing the 3rd and 4th arguments which are typically meant to be
> > internal. Those two parameters used to be unavoidable until the kernel
> > moved to building with the C11 standard unconditionally. Since then, it is
> > now possible to get rid of them. However a few users of
> > of_property_for_each_u32() do actually use those arguments, which
> > complicates the transition. For this reason this series does the following:
> > 
> >  * Add of_property_for_each_u32_new(), which does not have those two
> >    arguments (patch 1)
> >  * Convert _almost_ every usage to of_property_for_each_u32_new()
> >  * Rename of_property_for_each_u32() to of_property_for_each_u32_old() and
> >    deprecate it, as a incentive to code not (yet) in mainline to upgrade
> >    to the *_new() version (last patch)  
> 
> I don't really see the point of introducing the _old variant. Let's get 
> this done in one step.
> 
> > 
> > The plan for the next series is to additionally:
> > 
> >  * Convert the few remaining of_property_for_each_u32_old() instantes to
> >    of_property_for_each_u32_new()
> >  * Remove of_property_for_each_u32_old()
> >  * Rename of_property_for_each_u32_new() to of_property_for_each_u32()  
> 
> Honestly, I think there's few enough users we could just convert the 
> whole thing in one patch. It's all got to go thru 1 tree anyways. If 
> there's new cases in -next, then I'd be happy to send it to Linus at the 
> end of the merge window.

Sure, make sense. I'll need to convert the few remaining users, then
I'm sending a squashed v2.

Luca

-- 
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com