Re: [PATCH v3 03/11] phy: sun4i-usb: add support for the USB PHY on F1C100s SoC

[Andre Przywara <andre.przywara@arm.com>]

On Tue, 15 Nov 2022 16:00:54 +0100
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote:

Hi,

> On 15/11/2022 11:44, Andre Przywara wrote:
> > On Tue, 15 Nov 2022 11:03:24 +0100
> > Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote:
> > 
> > Hi,
> >   
> >> On 15/11/2022 07:01, Jernej Škrabec wrote:  
> >>> Dne četrtek, 10. november 2022 ob 08:35:39 CET je Vinod Koul napisal(a):    
> >>>> On 06-11-22, 15:48, Andre Przywara wrote:    
> >>>>> From: Icenowy Zheng <uwu@icenowy.me>
> >>>>>
> >>>>> The F1C100s SoC has one USB OTG port connected to a MUSB controller.
> >>>>>
> >>>>> Add support for its USB PHY.    
> >>>>
> >>>> This does not apply for me, please rebase and resend
> >>>>
> >>>> Also, consider splitting phy patches from this. I dont think there is
> >>>> any dependency    
> >>>
> >>> DT patches in this series depend on functionality added here.
> >>>     
> >>
> >> DTS always goes separately from driver changes because it is a hardware
> >> description. Depending on driver means you have potential ABI break, so
> >> it is already a warning sign.  
> > 
> > We understand that ;-)
> > What Jernej meant was that the DTS patches at the end depend on patch
> > 01/10, which adds to the PHY binding doc. I am not sure if Vinod's
> > suggestion was about splitting off 01/10, 03/10, and 10/10, or just the
> > two latter which touch the driver.
> > 
> > I can split off 03/10 and 10/10, rebased on top of linux-phy.git/next, and
> > send that to Vinod.
> > Then I would keep 01/10 in a respin of this series here, to satisfy the
> > dependency of the later DTS patches, and Vinod can pick that one patch from
> > there?  
> 
> There is no hard dependency of DTS on bindings. You can split these (and
> some maintainers prefer that way) and in DTS patches just provide the
> link to the bindings, saying it is in progress.

But that breaks "make dtbs_check", doesn't it?

I would think that the DT bits - bindings first, then DTS files using it -
should be bundled. This is how I imagine the future(TM), where DTs and
bindings live outside the kernel repo.

> The bindings should be however kept with driver changes as it goes the
> same way.

I understand that the bindings describe the contract the driver acts on,
but going forward I think driver changes would need to come later, then
(since they will live in a separate repo at some day)?
Maybe pointing to the binding changes in progress?

So with a separate repo we would actually need to upstream just the
bindings first, then could push driver changes and .dts files
independently?

And for now it looks like we are stuck with putting everything in one
series, to make both checkpatch and dtbs_check happy.

Cheers,
Andre