Re: [RFC 10/10] net: sfp: add support for fwnode

["Clément Léger" <clement.leger@bootlin.com>]

Le Wed, 23 Feb 2022 16:23:46 +0100,
Andrew Lunn <andrew@lunn.ch> a écrit :

> > As Russell asked, I'm also really interested if someone has a solution
> > to reuse device-tree description (overlays ?) to describe such
> > hardware. However, the fact that CONFIG_OF isn't enabled on x86 config
> > seems a bit complicated on this side.  
> 
> It does work, intel even used it for one of there tiny x86 SoCs. Maybe
> it was Newton? If you search around you can find maybe a Linux
> Plumbers presentation about DT and x86.

Oh yes, I know it works and I tried loading an overlay using CONFIG_OF
on a x86. Currently it does not works and generate a oops due to the
fact that the lack of "/" node is not handled and that the error path
has probably not been thoroughly tested. Adress remapping for PCI and
lack of PCI bus description might also be a bit cumbersome but maybe
this is the way to go.

I was saying a "bit complicated" as it is not often used and it would
require enabling CONFIG_OF to support this feature as well as allowing
loading overlays without any root node. But this is probably something
that is also achievable.

> 
> You can probably use a udev rule, triggered by the PCIe device ID to
> load the DT overlay.

Or maybe load the overlay from the PCIe driver ? This would also allow
to introduce some remapping of addresses (see below) inside the driver
maybe.

> 
> Do you actually need anything from the host other than PCIe? It sounds
> like this card is pretty self contained, so you won't need phandles
> pointing to the host i2c bus, or the hosts GPIOs? You only need
> phandles to your own i2c bus, your own GPIOs? That will make the
> overlay much simpler.

Yes, the device is almost self contained, only the IRQ needs to be
chained with the MSI.

> 
> 	Andrew



-- 
Clément Léger,
Embedded Linux and Kernel engineer at Bootlin
https://bootlin.com