Sun, Dec 14, 2025 at 08:35:01PM +0100, ivecera@redhat.com wrote:

On December 12, 2025 12:25:12 PM GMT+01:00, Jiri Pirko [off-list ref] wrote:

quoted

Thu, Dec 11, 2025 at 08:47:45PM +0100, ivecera@redhat.com wrote:

[..]

quoted

@@ -559,7 +563,8 @@ EXPORT_SYMBOL(dpll_netdev_pin_clear);
 */

struct dpll_pin *
dpll_pin_get(u64 clock_id, u32 pin_idx, struct module *module,
-	     const struct dpll_pin_properties *prop)
+	     const struct dpll_pin_properties *prop,
+	     struct fwnode_handle *fwnode)
{
	struct dpll_pin *pos, *ret = NULL;
	unsigned long i;

@@ -568,14 +573,15 @@ dpll_pin_get(u64 clock_id, u32 pin_idx, struct module *module,

	xa_for_each(&dpll_pin_xa, i, pos) {
		if (pos->clock_id == clock_id &&
		    pos->pin_idx == pin_idx &&
-		    pos->module == module) {
+		    pos->module == module &&
+		    pos->fwnode == fwnode) {

Is fwnode part of the key? Doesn't look to me like that. Then you can
have a simple helper to set fwnode on struct dpll_pin *, and leave
dpll_pin_get() out of this, no?

IMHO yes, because particular fwnode identifies exact dpll pin, so
I think it should be a part of the key.

The key items serve for userspace identification purposes as well. For
that, fwnode is non-sense.
fwnode identifies exact pin, that is nice. But is it the only
differentiator among other key items? I don't expect so.