11.11.2020 12:04, Krzysztof Kozlowski пишет:
>> -obj-$(CONFIG_TEGRA124_EMC) += clk-tegra124-emc.o
>> +obj-$(CONFIG_ARCH_TEGRA_124_SOC) += clk-tegra124-emc.o
>> +obj-$(CONFIG_ARCH_TEGRA_132_SOC) += clk-tegra124-emc.o
> How is it related to modularization? It looks like different issue is
> fixed here.
The CONFIG_TEGRA124_EMC now could be 'm', while the clock code must be
built-in. The TEGRA124 EMC driver is used by T124 and T132 SoCs.
...
>> diff --git a/drivers/clk/tegra/clk.h b/drivers/clk/tegra/clk.h
>> index 6b565f6b5f66..2da7c93c1a6c 100644
>> --- a/drivers/clk/tegra/clk.h
>> +++ b/drivers/clk/tegra/clk.h
>> @@ -881,18 +881,6 @@ void tegra_super_clk_gen5_init(void __iomem *clk_base,
>> void __iomem *pmc_base, struct tegra_clk *tegra_clks,
>> struct tegra_clk_pll_params *pll_params);
>>
>> -#ifdef CONFIG_TEGRA124_EMC
>> -struct clk *tegra_clk_register_emc(void __iomem *base, struct device_node *np,
>> - spinlock_t *lock);
>> -#else
>> -static inline struct clk *tegra_clk_register_emc(void __iomem *base,
>> - struct device_node *np,
>> - spinlock_t *lock)
>> -{
>> - return NULL;
>> -}
>> -#endif
> Why clock changes are so tightly coupled with making an EMC driver
> modular? Usually this should be a separate change - you adjust any
> dependencies to accept late or deferred probing, exported symbols,
> loosen the coupling between drivers, etc. and then you convert something
> to module.
Because the clock and EMC driver were not separated from each other
previously. The clock part can't be modularized easily and probably
shouldn't.
I'm not sure whether it's actually possible to split this patch without
taking a closer a look.
I'm also doubt that it would really worth the effort for a 100 lines of
a changed code.
