On 17/05/2021 09:47, Dmitry Osipenko wrote: > 17.05.2021 16:39, Krzysztof Kozlowski пишет: >>>>> #define DRAM_DEV_SEL_ALL 0 >>>>> -#define DRAM_DEV_SEL_0 (2 << 30) >>>>> -#define DRAM_DEV_SEL_1 (1 << 30) >>>>> +#define DRAM_DEV_SEL_0 (2u << 30) >>>>> +#define DRAM_DEV_SEL_1 (1u << 30) >>>> >>>> Why not using BIT()? This would make even this 2<<30 less awkard... >>> >>> The bitfield 31:30 is a enum, 3 is a wrong value. Formally it's >>> incorrect to use the BIT() macro here. >> >> Why "3"? BIT(31) is the same as 2<<30. > > By 3 I meant BIT(31)|BIT(30). This bitfield is explicitly designated as > a enum in the hardware documentation. I understand it and using BIT() here does not mean someone has to set both of them. BIT() is a helper pointing out that you want to toggle one bit. It does not mean that it is allowed to do so always! > >> It's common to use BIT for >> register fields which do not accept all possible values. Now you >> basically reimplement BIT() which is error-prone. > > Could you please show couple examples? The common practice today is to > use FIELD_PREP helpers, but this driver was written before these helpers > existed. There are plenty of such examples so I guess it would be easier to ask you to provide counter ones. Few IT for enum-like registers found within 2 minutes: https://elixir.bootlin.com/linux/latest/C/ident/MAX77620_CNFG_GPIO_INT_MASK https://elixir.bootlin.com/linux/v5.13-rc2/source/drivers/regulator/max77650-regulator.c#L18 https://elixir.bootlin.com/linux/v5.13-rc2/source/drivers/regulator/tps6524x-regulator.c#L62 https://elixir.bootlin.com/linux/v5.13-rc2/source/drivers/regulator/tps80031-regulator.c#L39 https://elixir.bootlin.com/linux/v5.13-rc2/source/drivers/regulator/da9121-regulator.h#L200 https://elixir.bootlin.com/linux/v5.13-rc2/source/drivers/regulator/da9121-regulator.h#L231 Best regards, Krzysztof
