> > >What about doing like
> >
> > gpio->secure = devm_platform_ioremap_resource_byname(pdev, "security");
> > > if (IS_ERR(gpio->secure))
> > > gpio->secure = devm_platform_ioremap_resource(pdev, 0);
> > > if (IS_ERR(gpio->secure))
> > > return PTR_ERR(gpio->secure);
> > >
> > >and similar for gpio->base?
> >
> > Wouldn't this cause a redundant check if it had already succeeded in getting
> > the resource by name? Also, could it happen that if the device tree is
> > incorrect, then one of the resource is fetched by name and other by the index,
> > which I guess, would mess things up. Just my random thoughts, not sure if it
> > is valid enough.
> >
> > >Wouldn't the following be enough?
> > >
> > >- gpio->intc.name = pdev->dev.of_node->name;
> > >+ gpio->intc.name = devm_kasprintf(&pdev->dev, "%pfw",
> > >dev_fwnode(&pdev->dev));
> > >+ if (!gpio->intc.name)
> > >+
> >
> > How about this way? I feel it would be right to add the OF functions conditionally.
>
> Looks okay, although I have a question here.
>
> + if (pdev->dev.of_node) {
>
> Do we really need this check at all? If the OF-node is NULL then it
> doesn't matter if other fields are filled or not, correct?
>
> What you need is #ifdef CONFIG_OF_GPIO (IIRC the name correctly).
>
> > + gpio->gpio.of_node = pdev->dev.of_node;
> > + gpio->gpio.of_gpio_n_cells = 2;
> > + gpio->gpio.of_xlate = tegra186_gpio_of_xlate;
> > + }
> >
> > + gpio->intc.name = gpio->soc->name;
Okay. It makes sense. Thanks Andy. I would make the changes and send out an updated patch.
--
Best Regards,
Akhil
