Re: [PATCHv3 2/4] wiegand: add Wiegand bus driver

[Andy Shevchenko <andriy.shevchenko@intel.com>]

On Tue, Apr 04, 2023 at 10:30:59PM +0200, Linus Walleij wrote:
> On Tue, Apr 4, 2023 at 3:13 PM Martin Zaťovič <m.zatovic1@gmail.com> wrote:
> > On Wed, Mar 01, 2023 at 06:23:54PM +0200, Andy Shevchenko wrote:

...

> > > > +/**
> > > > + * of_register_wiegand_device - allocates and registers a new Wiegand device based on devicetree
> > >
> > > NAK for OF only code. New, esp. bus, code must be agnostic. We have all means
> > > for that.
> >
> > In one of the previous versions of this patch series, there was also the possibility to instantiate
> > the device from another driver. I have been told, that this is not the way to go anymore, unless
> > there is a very specific reason for that. I did not find such reason, so I have removed this suport.
> 
> I don't know for sure but I think Andy simply means that you should take a look
> in include/linux/property.h and replace every function named of_* with
> the corresponding fwnode_* or device_* function from that file, and it
> should all just magically work the same, but abstracted away from device
> tree. It's not much more than a search/replace and rename operation
> (unless there is a "hole" in the fwnode implementations... I hope not.)
> 
> In the end you can keep just <linux/property.h> and drop <linux/of.h>
> and <linux/of_device.h> if this works.

Yes. And the argument for embedded devices it's exactly why the non-OF
interface is needed as we want to have a possibility to connect devices on
ACPI (or another, if any) based platform without changing the code.

-- 
With Best Regards,
Andy Shevchenko