On Mo, 2025-02-24 at 11:15 +0100, Krzysztof Kozlowski wrote:

On 24/02/2025 11:12, Krzysztof Kozlowski wrote:

quoted

On 24/02/2025 10:55, Ryan Chen wrote:

quoted

+
+static void aspeed_reset_unregister_adev(void *_adev)
+{
+	struct auxiliary_device *adev = _adev;
+
+	auxiliary_device_delete(adev);
+	auxiliary_device_uninit(adev);
+}
+
+static void aspeed_reset_adev_release(struct device *dev)
+{
+	struct auxiliary_device *adev = to_auxiliary_dev(dev);
+
+	kfree(adev);
+}
+

Every exported function *must* have kerneldoc.

quoted

+int aspeed_reset_controller_register(struct device *clk_dev, void __iomem *base,
+				     const char *adev_name)
+{
+	struct auxiliary_device *adev;
+	int ret;
+
+	adev = kzalloc(sizeof(*adev), GFP_KERNEL);
+	if (!adev)
+		return -ENOMEM;
+
+	adev->name = adev_name;
+	adev->dev.parent = clk_dev;
+	adev->dev.release = aspeed_reset_adev_release;
+	adev->id = 666u;
+
+	ret = auxiliary_device_init(adev);
+	if (ret) {
+		kfree(adev);
+		return ret;
+	}
+
+	ret = auxiliary_device_add(adev);
+	if (ret) {
+		auxiliary_device_uninit(adev);
+		return ret;
+	}
+
+	adev->dev.platform_data = (__force void *)base;
+
+	return devm_add_action_or_reset(clk_dev, aspeed_reset_unregister_adev, adev);
+}
+EXPORT_SYMBOL_GPL(aspeed_reset_controller_register);

No, you cannot export functions without users. There is no single user
of this, so this is not justified at all.

My mistake, I missed patch #3 which uses it.

I don't get why do you need to export this in the first place, instead
of putting it in the clock driver, as usually expected. Handling child
creation is logically the task of the device having children, the
parent. Not the child.

Also, consider basing this on top of:

https://lore.kernel.org/all/20250218-aux-device-create-helper-v4-0-c3d7dfdea2e6@baylibre.com/ (local)

regards
Philipp