On Fri, 8 Oct 2021 00:17:17 -0500 Samuel Holland <[email protected]> wrote:
> CONFIG_SPL_I2C is the wrong symbol to use here. It is the top-level > Kconfig symbol (not specific to either legacy or DM I2C), whereas the > i2c_init() function is specific to legacy I2C. This change fixes a > build failure when enabling SPL_I2C but not SPL_SYS_I2C_LEGACY. > > Signed-off-by: Samuel Holland <[email protected]> I don't have the brain capacity right now to comment on the issue below, but just for the sake of this patch: Reviewed-by: Andre Przywara <[email protected]> Cheers, Andre > --- > Actually, I think this commit raises a larger issue: > > For some reason, SPL_SYS_I2C_LEGACY does not depend on SPL_I2C. When > SPL_SYS_I2C_LEGACY was added in commit 55dabcc8f245 ("Convert > CONFIG_SYS_I2C_LEGACY to Kconfig and add CONFIG_[ST]PL_SYS_I2C_LEGACY"), > SPL_I2C wasn't added to the board configs. > > But since commit 537892065ac1 ("Makefile: Move drivers/i2c/ into > drivers/Makefile"), drivers/i2c is only compiled if SPL_I2C is enabled. > > So the combination of these two commits appears to have accidentally > removed I2C support from SPL for many boards. > > The impact here is that checking CONFIG_IS_ENABLED(SYS_I2C_LEGACY) is > not always enough, even though it should be. Because if CONFIG_SPL_I2C=n > then i2c_init is undefined because drivers/i2c/i2c_core.c is not built. > > Changes in v2: > - New patch to account for I2C Kconfig changes > > arch/arm/mach-sunxi/board.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/arm/mach-sunxi/board.c b/arch/arm/mach-sunxi/board.c > index d9b04f75fc4..b74ad4074df 100644 > --- a/arch/arm/mach-sunxi/board.c > +++ b/arch/arm/mach-sunxi/board.c > @@ -339,7 +339,7 @@ void board_init_f(ulong dummy) > spl_init(); > preloader_console_init(); > > -#ifdef CONFIG_SPL_I2C > +#if CONFIG_IS_ENABLED(I2C) && CONFIG_IS_ENABLED(SYS_I2C_LEGACY) > /* Needed early by sunxi_board_init if PMU is enabled */ > i2c_init(CONFIG_SYS_I2C_SPEED, CONFIG_SYS_I2C_SLAVE); > #endif
