On Fri, May 29, 2015 at 6:37 PM, Brian Norris <computersforpeace@xxxxxxxxx> wrote: > On Fri, May 29, 2015 at 05:57:50PM -0700, Gregory Fong wrote: >> On Fri, May 29, 2015 at 5:36 PM, Brian Norris >> <computersforpeace@xxxxxxxxx> wrote: >> > On Thu, May 28, 2015 at 07:14:07PM -0700, Gregory Fong wrote: >> >> --- a/Documentation/devicetree/bindings/gpio/brcm,brcmstb-gpio.txt >> >> +++ b/Documentation/devicetree/bindings/gpio/brcm,brcmstb-gpio.txt >> >> @@ -33,6 +33,12 @@ Optional properties: > ... >> >> - #interrupt-cells: >> >> Should be <2>. The first cell is the GPIO number, the second should specify >> >> flags. The following subset of flags is supported: >> >> @@ -48,7 +54,10 @@ Optional properties: >> >> Marks the device node as an interrupt controller >> >> >> >> - interrupt-names: >> >> - The name of the IRQ resource used by this controller >> >> + The names of the IRQ resources used by this controller >> > >> > If you're specifying names, you should list them here. >> >> I was wondering about that. Some bindings have them listed, some >> don't. In this case I know what names currently exist but there could >> certainly be different ones in the future. How does that work? Or am >> I misunderstanding what this field is used for? Where are the >> documented rules for this? > > The only documentation I see is: > Documentation/devicetree/bindings/resource-names.txt > > That documents the basics of the *-names properties, not their expected > usage. > > In practice, they're only useful if you have enough optional resources > that fixed indexing isn't sufficient, and you need to use > platform_get_resource_byname(). > > So IMO, their purposes seems to be one of these: > (1) functional (e.g., for get_resource_byname(), when you have more than > one optional resource) > (2) self-documentation (which might run counter to #1, as you begin > generating too many unique names) > (3) no purpose > > So IMO, if you ever want (1), they shouldn't have instance-specific > naming, but should use something generic to the device class. Otherwise, > they are just self-documentation, and aren't functionally useful. So > IMO, these sorts of names: > > interrupt-names = "upg_gio_aon", "upg_gio_aon_wakeup"; > > work better as functional descriptions: > > interrupt-names = "gio", "wakeup"; > > But in the end, I wouldn't foresee you needing to do (1), so you're left > with (2) or (3), at which point I'm not sure if you should even mention > the property. I'm fine with leaving out the interrupt-names property, since we're not using it here anyway. Unless there are serious objections, I'll plan on remove it from the bindings doc next round. Thanks, Gregory -- 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
