On 05/08/2012 09:39 AM, John Crispin wrote: >>> I don't think anyone else is going to want to compile >>> this. >> >> This Kconfig option is selected by the ARCH Kconfig, so only selected at >> the right time. The user won't get prompted for it since there's no >> string after "bool". I think this is OK. Tegra's pinctrl Kconfig option >> doesn't have any "depends ARCH_TEGRA" here either, although I note that >> many other pinctrl drivers do. > > Hi, > > I guess it makes it more apparent, that the symbol is specific to a > arch/soc. Tegra is a well known SoC, so its easy to figure out what the > codes purpose is. Other files might not be that easy to guess. > > For the Lantiq SoC to function normally we need to always load these > drivers. PINTCTRL_LANTIQ has some generic functions and > PINTCTRL_LANTIQ_XWAY holds the code specific to the XWAY SoC. (i have a > patch in the local queue to add FALCON SoC support, giving > PINTCTRL_LANTIQ 2 users) > > How about we do the following. > > config LANTIQ > select PINCTRL > > config PINCTRL_LANTIQ > def_bool y > depends on LANTIQ > > config PINCTRL_LANTIQ_XWAY > def_bool y > depends on SOC_TYPE_XWAY > > This would auto select the right symbols, have all the dependency logic > in 1 place and reduce the size of arch/mips/lantiq/Kconfig Is it useful to build a kernel for XWAY without pinctrl, once the driver is there? That's the main difference between arch/mips/lantiq/Kconfig selecting PINCTRL_LANTIQ_XWAY vs that being a def_bool, and hence allowing the user to deselect it.
