Re: [PATCH] ACPI / device_sysfs: Use OF_MODALIAS for "compatible" modalias

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Thu, Jan 21, 2021 at 02:22:43PM +0800, Kai-Heng Feng wrote:
> On Tue, Jan 19, 2021 at 6:34 PM Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> >
> > On Tue, Jan 19, 2021 at 11:41:59AM +0200, Andy Shevchenko wrote:
> > > On Tue, Jan 19, 2021 at 04:41:48PM +0800, Kai-Heng Feng wrote:
> > > > On Tue, Jan 19, 2021 at 4:27 PM Greg Kroah-Hartman
> > > > <gregkh@linuxfoundation.org> wrote:
> > > > > On Tue, Jan 19, 2021 at 04:15:13PM +0800, Kai-Heng Feng wrote:
> > >
> > > ...
> > >
> > > > > Who will use OF_MODALIAS and where have you documented it?
> > > >
> > > > After this lands in mainline, I'll modify the pull request for systemd
> > > > to add a new rule for OF_MODALIAS.
> > > > I'll modify the comment on the function to document the change.
> > >
> > > I'm wondering why to have two fixes in two places instead of fixing udev to
> > > understand multiple MODALIAS= events?
> >
> > It's not a matter of multiple events, it's a single event with a
> > key/value pair with duplicate keys and different values.
> >
> > What is this event with different values supposed to be doing in
> > userspace?  Do you want multiple invocations of `modprobe` or something
> > else?
> >
> > Usually a "device" only has a single "signature" that modprobe uses to
> > look up the correct module for.  Modules can support any number of
> > device signatures, but traditionally it is odd to think that a device
> > itself can be supported by multiple modules, which is what you are
> > saying is happening here.
> >
> > So what should userspace do with this, and why does a device need to
> > have multiple module alias signatures?
> 
> >From the original use case [1], I think the "compatible" modalias
> should be enough.
> Andy and Mika, what do you think? Can we remove the ACPI modalias for this case?

Yes, I think that should work. After all we want the match to happen
through the DT compatible string if the property is present, not through
ACPI IDs.