[PATCH v2 1/5] dt-bindings: ufs: mediatek,ufs: Add mt8196-ufshci variant

Nicolas Frattaroli nicolas.frattaroli at collabora.com
Mon Oct 20 06:56:39 PDT 2025


On Monday, 20 October 2025 15:27:40 Central European Summer Time Nicolas Frattaroli wrote:
> On Saturday, 18 October 2025 23:30:17 Central European Summer Time Conor Dooley wrote:
> > On Fri, Oct 17, 2025 at 09:02:07PM +0200, Nicolas Frattaroli wrote:
> > > On Friday, 17 October 2025 17:42:10 Central European Summer Time Conor Dooley wrote:
> > > > On Thu, Oct 16, 2025 at 02:06:43PM +0200, Nicolas Frattaroli wrote:
> > > 
> > > > > +
> > > > > +  freq-table-hz: true
> > > > 
> > > > Then you add this deprecated property, which isn't mentioned in the
> > > > commit message and I don't see why a deprecated property is needed.
> > > 
> > > I'll rework it to use operating-points-v2 instead. It needs one of
> > > the two, or else on mt8196 at least, the hardware locks up.
> > > 
> > > I'll still add operating-points-v2 for all SoCs though, if that's
> > > okay with you.
> > 
> > Right. I'd accept freq-table-hz if the other devices here have been
> > using it all along, but if this is something new - then please use the
> > operating-points-v2 property. Looking at the binding example, it looks
> > like it does indeed use freq-table-hz, so that's probably justification
> > enough to keep doing so.
> 
> Turns out the only usage of freq-table-hz is in the examples I've added.
> Mainline does not at all have any nodes in the DT right now that would
> use this property.
> 
> Ergo, I think I will go for operating-points-v2. We might as well clean
> this up now instead of ossifying a deprecated property in a new binding
> for the sake of downstream compatibility (which should never be a concern)
> that I am already breaking. The added benefit is that if we ever do get
> better OPP definitions than just two clock states, then we can actually
> add the OPP bandwidth properties so implementations can make informed
> decisions.
> 

Nevermind, I see the existing binding had it in the example for 8183,
just not in the binding itself. So freq-table-hz it is.





More information about the linux-phy mailing list