On Fri, 2013-11-22 at 10:15 +0000, Lee Jones wrote:
> > +#ifdef CONFIG_OF
> > +static int max14577_regulator_dt_parse_pdata(struct platform_device *pdev,
> > + struct max14577_platform_data *pdata)
> > +{
> > + struct max14577 *max14577 = dev_get_drvdata(pdev->dev.parent);
> > + struct device_node *np;
> > + struct max14577_regulator_platform_data *reg_pdata;
> > + struct of_regulator_match rmatch;
> > + int i, ret;
> > +
> > + np = of_get_child_by_name(max14577->dev->of_node, "regulators");
> > + if (!np)
> > + return -EINVAL;
>
> No need to do this. If instead you use a compatible string and set
> it's MFD cell's .of_compatible property, the MFD core will set
> pdev->dev.of_node for you.
>
> > + reg_pdata = devm_kzalloc(&pdev->dev, sizeof(*reg_pdata) *
> > + ARRAY_SIZE(supported_regulators), GFP_KERNEL);
> > + if (!reg_pdata)
> > + return -ENOMEM;
> > +
> > + for (i = 0; i < ARRAY_SIZE(supported_regulators); i++) {
> > + rmatch.name = supported_regulators[i].name;
> > + ret = of_regulator_match(&pdev->dev, np, &rmatch, 1);
> > + if (ret != 1)
> > + continue;
> > + dev_dbg(&pdev->dev, "Found regulator %d/%s\n",
> > + supported_regulators[i].id,
> > + supported_regulators[i].name);
> > + reg_pdata[i].id = supported_regulators[i].id;
> > + reg_pdata[i].initdata = rmatch.init_data;
> > + reg_pdata[i].of_node = rmatch.of_node;
> > + }
> > +
> > + pdata->regulators = reg_pdata;
> > +
> > + return 0;
> > +}
> > +#else
> > +static int max14577_regulator_dt_parse_pdata(struct platform_device *pdev,
> > + struct max14577_platform_data *pdata)
> > +{
> > + return 0;
> > +}
> > +#endif
>
> No need for this either. Just check for the device's of_node.
>
> > +static int max14577_regulator_probe(struct platform_device *pdev)
> > +{
> > + struct max14577 *max14577 = dev_get_drvdata(pdev->dev.parent);
> > + struct max14577_platform_data *pdata = dev_get_platdata(max14577->dev);
> > + int i, size;
> > + struct regulator_config config = {};
> > + struct regulator_dev **regulators;
> > +
> > + if (!pdata) {
> > + /* Parent must provide pdata */
> > + dev_err(&pdev->dev, "No MFD driver platform data found.\n");
> > + return -ENODEV;
> > + }
> > +
> > + if (max14577->dev->of_node) {
> > + int ret = max14577_regulator_dt_parse_pdata(pdev, pdata);
>
> This will overwrite pdata which is wrong.
>
> pdata should always take precedence over DT.
>
> <snip>
>
> > + for (i = 0; i < ARRAY_SIZE(supported_regulators); i++) {
> > + /*
> > + * Index of supported_regulators[] is also the id and must
> > + * match index of pdata->regulators[].
> > + */
> > + config.init_data = pdata->regulators[i].initdata;
> > + config.of_node = pdata->regulators[i].of_node;
>
> I still thing this is superfluous. Why don't you run though the nodes
> here instead of doing it in MFD and passing this stuff through?
I have rewritten the regulator init code (and DT parsing). I think it
reflects your comments. Please look at version 4 of patches.
Best regards,
Krzysztof
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
