On czw, 2014-10-30 at 12:50 +0100, Javier Martinez Canillas wrote:

Hello Krzysztof,

On 10/30/2014 12:20 PM, Krzysztof Kozlowski wrote:

quoted

 static int max77802_pmic_probe(struct platform_device *pdev)
 {
 	struct max77686_dev *iodev = dev_get_drvdata(pdev->dev.parent);
-	struct max77686_platform_data *pdata = dev_get_platdata(iodev->dev);
 	struct max77802_regulator_prv *max77802;
-	int i, ret = 0, val;
+	int i, val;
 	struct regulator_config config = { };
 
-	/* This is allocated by the MFD driver */
-	if (!pdata) {
-		dev_err(&pdev->dev, "no platform data found for regulator\n");
-		return -ENODEV;
-	}
-
 	max77802 = devm_kzalloc(&pdev->dev,
 				sizeof(struct max77802_regulator_prv),
 				GFP_KERNEL);
 	if (!max77802)
 		return -ENOMEM;
 
-	if (iodev->dev->of_node) {
-		ret = max77802_pmic_dt_parse_pdata(pdev, pdata);
-		if (ret)
-			return ret;
+	if (!pdev->dev.of_node) {
+		/*
+		 * Backward DTS compatiblity where regulator driver had not
+		 * a compatible property for itself.
+		 */
+		if (!iodev->dev->of_node) {
+			dev_err(&pdev->dev, "No OF node for driver and its parent\n");
+			return -EINVAL;
+		}
+		pdev->dev.of_node = of_get_child_by_name(iodev->dev->of_node,
+								"regulators");
+		max77802->missing_of_node = true;
 	}

I may be missing something but I don't understand why a compatible string
for the regulators sub-node is needed. Isn't enough to just fill the
.regulators_node field in the struct regulator_desc? e.g:

    .regulators_node = of_match_ptr("regulators") for max77802
    .regulators_node = of_match_ptr("voltage-regulators") for max77686

AFAIU this should be enough for the core to extract the init_data and will
make your change much more simpler and you can drop patches 1-3 and 13-14.

Or maybe I misread the regulator_of_get_init_data() function?

The regulator_of_get_init_data() searches from dev->of_node or its child
node.

But dev->of_node is NULL.

That's why of_compatible is needed.

Best regards,
Krzysztof