Re: [PATCH v2 1/3] regulator: qcom_smd-regulator: fix PM2250 supply parent mappings

[Dmitry Baryshkov]

On Wed, Jul 22, 2026 at 04:44:28PM +0530, Rakesh Kota wrote:
> On Tue, Jul 21, 2026 at 03:31:11PM +0300, Dmitry Baryshkov wrote:
> > On Mon, Jul 20, 2026 at 04:25:04PM +0530, Rakesh Kota wrote:
> > > The rpm_pm2250_regulators table hardcodes supply parent name strings
> > > in the regulator descriptor, causing the Linux regulator framework to
> > > look up and vote on parent supplies from APPS. This is both unnecessary
> > > and incorrect for RPM-managed regulators.
> > > 
> > > On RPM-based platforms, the parent-child supply relationships are
> > > managed entirely by the RPM firmware. The RPM ensures that parent
> > > rails are maintained at the correct voltage with appropriate headroom,
> > > based on all active child rail votes across all subsystems. No explicit
> > > parent supply vote from APPS is required.
> > > 
> > > Furthermore, the supply rail connections to PMIC pins are fixed in
> > > hardware. The previously hardcoded supply name strings were inaccurate
> > > and did not reflect the actual physical pin connections on PM2250-based
> > > platforms. The correct supply relationships are to be described in the
> > > Device Tree, not hardcoded in the driver.
> > 
> > Last phrase is incorrect. The driver can hardcode parent supplies (and
> > it has been doing it correctly for all other PMICs).
> 
> Acknowledged, will rephrase.
> > 
> > > 
> > > To avoid unnecessary APPS-side parent supply access while still
> > > allowing correct hardware topology to be described in DT, set the
> > > supply name field to NULL in the regulator descriptor.
> > 
> > Why is this touching only PM2250? Should we drop all parent votes?
> >
> 
> The fix is conceptually valid for all SMD-managed regulators, but this
> series is scoped to PM2250/PM4125 which are the only platforms I have
> tested (Agatti, Shikra). Happy to extend it to all entries in a
> follow-on series, or do it all at once if preferred — please let me know.

At least mention that in the commit message. But if it is applicable to
_all_ SMD RPM platforms, would it be easier to just mass-remove the
rpm_regulator_data::supply field? It is relatively easy to test this on
devices up to MSM8916. If the stars are aligned, I should be able to
test it on APQ8074 too.

> 
> regards
> Rakesh kota
> > 
> > -- 
> > With best wishes
> > Dmitry

-- 
With best wishes
Dmitry