On 8/15/26 3:53 PM, Ralph Siemsen wrote:
On Sat, Aug 15, 2026 at 12:00:23AM +0200, Marek Vasut wrote:
diff --git a/drivers/pinctrl/renesas/pinctrl-rzn1.c b/drivers/
pinctrl/renesas/pinctrl-rzn1.c
index fdc43c8e714..6c8d40e9639 100644
--- a/drivers/pinctrl/renesas/pinctrl-rzn1.c
+++ b/drivers/pinctrl/renesas/pinctrl-rzn1.c
@@ -298,11 +298,24 @@ static int rzn1_pinconf_set(struct
rzn1_pinctrl_priv *priv, unsigned int pin,
static int rzn1_pinctrl_set_state(struct udevice *dev, struct
udevice *config)
{
struct rzn1_pinctrl_priv *priv = dev_get_priv(dev);
+ struct udevice *child;
int size;
int ret;
u32 val;
u32 bias;
+ /*
+ * Handle subnodes recursively, so that pin groups work.
+ * Note that properties are *NOT* inherited from parent.
+ */
+ device_foreach_child(child, config) {
+ ret = rzn1_pinctrl_set_state(dev, child);
Can there really be infinite nested subgroups , or are there always
only groups with subgroups and that is where it ends ? Linux
Documentation/devicetree/bindings/pinctrl/renesas,rzn1-pinctrl.yaml
makes it look like the later, so maybe the recursion is not necessary
here, instead call device_foreach_child() { device_foreach_child()
{ .. } } to prevent the possibility of infinite recursion ?
Realistically there will only be groups with one level of subgroups.
Perhaps if there was a phandle it could go a few levels deeper.
Unbounded recursion is not really possible here unless:
- the device tree is infinitely deep, or
- there is a very serious bug in device_foreach_child()
As these both seem unlikely, I opted for the simple recursive call, but
I am happy to change it if you feel it is better.
Let's go with bounded one-level-deep parsing, that's what the bindings
describe too.
