On Thu, Jan 15, 2015 at 10:29:08AM +0900, James Ban wrote: > This is a patch for fixing unmatched of_node and adding gpio control.
That sounds like two separate changes to me... > Optional properties: > +- bucka-uses-gpio: BUCKA can be controlled by gpio. > +- bucka-enable-platform-gpio: platform gpio for control of BUCKA. > +- bucka-enable-init-state: initial state of gpio for BUCKA Even for a single GPIO specifiers are usually called -gpios. It also seems redundant to have the -uses property, if there is a GPIO specifier for the enable then we can just assume it's supposed to be used without the extra property. > + bucka-uses-gpio; > + bucka-enable-platform-gpio = <&gpio 27 0>; > + bucka-enable-init-state = <0>; > + > + buckb-uses-gpio; > + buckb-enable-platform-gpio = <&gpio 17 0>; > + buckb-enable-init-state = <0>; > + > regulators { > BUCKA { > regulator-name = "VBUCKA"; Would it not be more natural to have the properties in the node for the regulator (the recently added of_parse_cb will help with that)? > + if (of_get_property(dev->of_node, "bucka-uses-gpio", NULL)) { > + gpio = of_get_named_gpio(dev->of_node, > + "bucka-enable-platform-gpio", 0); > + if (!gpio_is_valid(gpio)) { > + dev_err(dev, "invalid gpio: %d\n", gpio); > + return ERR_PTR(-EINVAL); > + } This will be broken for deferred probe, it's better to pass through any errors.
signature.asc
Description: Digital signature