Re: [PATCH 3/9] software node: allow referencing firmware nodes

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

On Tue, Oct 21, 2025 at 02:06:36AM -0700, Bartosz Golaszewski wrote:
> On Tue, 21 Oct 2025 08:54:22 +0200, Sakari Ailus
> <sakari.ailus@linux.intel.com> said:
> > On Mon, Oct 20, 2025 at 01:26:59PM +0200, Bartosz Golaszewski wrote:
> >> On Mon, Oct 20, 2025 at 12:05 PM Andy Shevchenko
> >> <andriy.shevchenko@linux.intel.com> wrote:

...

> >> > > > Can't we always have an fwnode reference?
> >> > >
> >> > > Unfortunately no. A const struct software_node is not yet a full
> >> > > fwnode, it's just a template that becomes an actual firmware node when
> >> > > it's registered with the swnode framework. However in order to allow
> >> > > creating a graph of software nodes before we register them, we need a
> >> > > way to reference those templates and then look them up internally in
> >> > > swnode code.
> >> >
> >> > Strange that you need this way. The IPU3 bridge driver (that creates a graph of
> >> > fwnodes at run-time for being consumed by the respective parts of v4l2
> >> > framework) IIRC has no such issue. Why your case is different?
> >>
> >> From what I can tell the ipu-bridge driver only references software
> >> nodes (as struct software_node) from other software nodes. I need to
> >> reference ANY implementation of firmware node from a software node.
> >
> > Yes, the IPU bridge only references software nodes.
> >
> > I might use two distinct pointers instead of an union and an integer field
> > that tells which type is the right one. I don't expect more such cases
> > here; it's either a software node or an fwnode handle (ACPI or OF node).
> 
> Like:
> 
> struct software_node_ref_args {
> 	const struct software_node *swnode;
> 	struct fwnode_handle *fwnode;
> 	unsigned int nargs;
> 	u64 args[NR_FWNODE_REFERENCE_ARGS];
> };
> 
> And then if swnode is NULL then assume fwnode must not be?
> 
> I'm not sure if it's necessarily better but I don't have a strong opinion on
> this either.

At least it is slightly closer to what I ideally want to have (but not in this
design seems), so +1 to Sakari's proposal.

-- 
With Best Regards,
Andy Shevchenko