Re: [PATCH 1/3] dt-bindings: usb: Add binding for TI USB8041 hub controller

[Alexander Stein <alexander.stein@ew.tq-group.com>]

Hello Krzysztof,

Am Dienstag, 12. Juli 2022, 23:32:12 CEST schrieb Krzysztof Kozlowski:
> On 12/07/2022 23:25, Matthias Kaehlcke wrote:
> > On Tue, Jul 12, 2022 at 11:12:06PM +0200, Krzysztof Kozlowski wrote:
> >> On 12/07/2022 19:25, Matthias Kaehlcke wrote:
> >>> Hi Alexander,
> >>> 
> >>> On Tue, Jul 12, 2022 at 05:06:25PM +0200, Alexander Stein wrote:
> >>>> The TI USB8041 is a USB 3.0 hub controller with 4 ports.
> >>>> 
> >>>> This initial version of the binding only describes USB related aspects
> >>>> of the USB8041, it does not cover the option of connecting the
> >>>> controller
> >>>> as an i2c slave.
> >>>> 
> >>>> Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com>
> >>>> ---
> >>>> Well, this is essentially a ripoff of
> >>>> Documentation/devicetree/bindings/usb/realtek,rts5411.yaml with USB IDs
> >>>> replaced, reset-gpio added and example adjusted.
> >>>> IMHO this should be merged together with realtek,rts5411.yaml. Is it ok
> >>>> to rename bindings files? I guess a common onboard-usb-hub.yaml
> >>>> matching
> >>>> the driver seens reasonable. Any recommendations how to proceed?
> >>> 
> >>> It's a tradeoff between keeping the individual bindings simple and avoid
> >>> unnecessary duplication. The current RTS5411 and TI USB8041 bindings are
> >>> very similar, which suggests combining them. However over time hubs with
> >>> diverging features could be added (e.g. with multiple regulators, a link
> >>> to an I2C/SPI bus, a clock, ...). With that a common binding might
> >>> become
> >>> too messy.
> >>> 
> >>> From a quick look through Documentation/devicetree/bindings it doesn't
> >>> seem common to have generic bindings that cover components from multiple
> >>> vendors. In that sense I'm leaning towards separate bindings.
> >>> 
> >>> Rob, do you have any particular preference or suggestion?
> >> 
> >> Not Rob, but my suggestion is not to merge bindings of unrelated
> >> devices, even if they are the same class. By unrelated I mean, made by
> >> different companies, designed differently and having nothing in common
> >> by design. Bindings can be still similar, but should not be merged just
> >> because they are similar.
> > 
> > Thanks for your advice, let's keep separate bindings then.

Ok, thanks for the feedback.

> Although for the record let me add that we did merge some trivial hwmon
> devices like LM75 or LM90 but their bindings are trivial and programming
> model is also similar between each other (handled by same device
> driver). I guess we can be here flexible, so the question would be how
> similar these USB hubs are.
> 
> If in doubt, just keep it separate.

Right now it might seem sensible to have the bindings merged, as the features 
are quite similar. But things might change, if/once i2c support is added. So 
this is one additional matter to keep them separated.

Best regards,
Alexander