Re: [RFC 09/10] i2c: mux: add support for fwnode

[Alexandre Belloni]

On 22/02/2022 11:57:39+0100, Andrew Lunn wrote:
> On Tue, Feb 22, 2022 at 09:53:25AM +0100, Clément Léger wrote:
> > Le Mon, 21 Feb 2022 19:55:25 +0200,
> > Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> a écrit :
> > 
> > > On Mon, Feb 21, 2022 at 05:26:51PM +0100, Clément Léger wrote:
> > > > Modify i2c_mux_add_adapter() to use with fwnode API to allow creating
> > > > mux adapters with fwnode based devices. This allows to have a node
> > > > independent support for i2c muxes.  
> > > 
> > > I^2C muxes have their own description for DT and ACPI platforms, I'm not sure
> > > swnode should be used here at all. Just upload a corresponding SSDT overlay or
> > > DT overlay depending on the platform. Can it be achieved?
> > > 
> > 
> > Problem is that this PCIe card can be plugged either in a X86 platform
> > using ACPI or on a ARM one with device-tree. So it means I should have
> > two "identical" descriptions for each platforms.
> 
> ACPI != DT.
> 
> I know people like stuffing DT properties into ACPI tables, when ACPI
> does not have a binding. But in this case, there is a well defined
> ACPI mechanism for I2C muxes. You cannot ignore it because it is
> different to DT. So you need to handle the muxes in both the ACPI way
> and the DT way.
> 
> For other parts of what you are doing, you might be able to get away
> by just stuffing DT properties into ACPI tables. But that is not for
> me to decide, that is up to the ACPI maintainers.
> 

What I'm wondering is why you would have to stuff anything in ACPI when
plugging any PCIe card in a PC. Wouldn't that be a first?

I don't have to do so when plugging an Intel network card, I don't
expect to have to do it when plugging any other network card...

-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com