Hi, Light On Tue, Mar 17, 2020 at 2:28 AM Light Hsieh <light.hsieh@xxxxxxxxxxxx> wrote: > > On Sun, 2020-03-15 at 14:35 -0700, Sean Wang wrote: > > Hi, Light > > > > On Tue, Mar 10, 2020 at 12:39 AM <light.hsieh@xxxxxxxxxxxx> wrote: > > > > > > From: Light Hsieh <light.hsieh@xxxxxxxxxxxx> > > > > > > > Correct the prefix of the subject with "pinctrl: mediatek". > > > > > In the future we want to be able to build the MediaTek pinctrl driver, > > > based on paris, as kernel module. This patch allows pinctrl-paris.c, the > > > external interrupt controller mtk-eint.c, and pinctrl-mtk-common-v2.c to > > > be loadable as module. > > > > > > Signed-off-by: Light Hsieh <light.hsieh@xxxxxxxxxxxx> > > > Reviewed-by: Matthias Brugger <matthias.bgg@xxxxxxxxx> > > > --- > > > drivers/pinctrl/mediatek/Kconfig | 13 +++++++++++-- > > > drivers/pinctrl/mediatek/Makefile | 5 +++-- > > > drivers/pinctrl/mediatek/mtk-eint.c | 9 +++++++++ > > > drivers/pinctrl/mediatek/pinctrl-mtk-common-v2.c | 24 ++++++++++++++++++++++++ > > > drivers/pinctrl/mediatek/pinctrl-paris.c | 6 ++++++ > > > drivers/pinctrl/pinconf-generic.c | 1 + > > > > Separate the patches for the device driver and for the core respectively. > > Since the patch for pinconf-generic.c had been done in b88d145191ad > ("pinctrl: Export some needed symbols at module load time"). > I will remove patch for pinconf-generic.c. > > > > 6 files changed, 54 insertions(+), 4 deletions(-) > > > > > > diff --git a/drivers/pinctrl/mediatek/Kconfig b/drivers/pinctrl/mediatek/Kconfig > > > index 701f9af..4cd1109 100644 > > > --- a/drivers/pinctrl/mediatek/Kconfig > > > +++ b/drivers/pinctrl/mediatek/Kconfig > > > @@ -3,10 +3,12 @@ menu "MediaTek pinctrl drivers" > > > depends on ARCH_MEDIATEK || COMPILE_TEST > > > > > > config EINT_MTK > > > - bool "MediaTek External Interrupt Support" > > > + tristate "MediaTek External Interrupt Support" > > > depends on PINCTRL_MTK || PINCTRL_MTK_MOORE || PINCTRL_MTK_PARIS || COMPILE_TEST > > > select GPIOLIB > > > select IRQ_DOMAIN > > > + default y if PINCTRL_MTK || PINCTRL_MTK_MOORE > > > + default PINCTRL_MTK_PARIS > > > > > > config PINCTRL_MTK > > > bool > > > @@ -17,23 +19,30 @@ config PINCTRL_MTK > > > select EINT_MTK > > > select OF_GPIO > > > > > > +config PINCTRL_MTK_V2 > > > + tristate > > > + depends on PINCTRL_MTK_MOORE || PINCTRL_MTK_PARIS > > > > PINCTRL_MTK_V2 doesn't have to depend on PINCTRL_MTK_MOORE or > > PINCTRL_MTK_PARIS to work and it can build on it own so that lets us > > remove the dependency. > > > The usage of PINCTRL_MTK had been describe in cover letter: > > 1.4 PINCTRL_MTK_V2: add this tristate config which depends on > PINCTRL_MTK_PARIS and PINCTRL_MTK_MOORE. No need to specify in > defconfig file. > * PINCTRL_MTK_V2 will be y if either PINCTRL_MTK_PARIS or > PINCTRL_MTK_MOORE is y. > * PINCTRL_MTK_V2 will be n if both PINCTRL_MTK_PARIS and > PINCTRL_MTK_MOORE are n. > * PINCTRL_MTK_V2 will be m if PINCTRL_MTK_PARIS is m and > PINCTRL_MTK_MOORE is n. > > > Without PINCTRL_MTK_V2, we cannot specify whether > pinctrl-mtk-common-v2.c is built-in or loadable kernel module. > What I was actually meaning is config PINCTRL_MTK_V2 tristate with that should work for your case. > > > + > > > config PINCTRL_MTK_MOORE > > > bool > > > depends on OF > > > select GENERIC_PINCONF > > > select GENERIC_PINCTRL_GROUPS > > > select GENERIC_PINMUX_FUNCTIONS > > > + select EINT_MTK > > > > The original design of PINCTRL_MTK_MOORE doesn't have to work with > > EINT_MTK coupled tightly so that let us remove the reverse dependency. > > > > Actually, PINCTRL_MTK_MOORE use EINT_MTK in the following call flow: > mtk_moore_pinctrl_probe() -> mtk_build_eint() -> mtk_eint_do_init(). > mtk_eint_do_init() is defined in mtk-eint.c. CONFIG_EINT_MTK is actually an option to Moore pinctrl since Moore pinctrl has already carefully considered the condition with CONFIG_EINT_MTK=n in mtk_build_eint. > > > > select GPIOLIB > > > select OF_GPIO > > > + select PINCTRL_MTK_V2