| Message ID | 20210516161214.4693-4-digetx@gmail.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | Enable compile-testing of Tegra memory drivers | expand |
On 5/16/2021 9:12 AM, Dmitry Osipenko wrote: > Fix compilation warning on 64bit platforms caused by implicit promotion > of 32bit signed integer to a 64bit unsigned value which happens after > enabling compile-testing of the driver. > > Suggested-by: Nathan Chancellor <nathan@kernel.org> > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> Reviewed-by: Nathan Chancellor <nathan@kernel.org> > --- > drivers/memory/tegra/tegra124-emc.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c > index 5699d909abc2..c9eb948cf4df 100644 > --- a/drivers/memory/tegra/tegra124-emc.c > +++ b/drivers/memory/tegra/tegra124-emc.c > @@ -272,8 +272,8 @@ > #define EMC_PUTERM_ADJ 0x574 > > #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) > > #define EMC_CFG_POWER_FEATURES_MASK \ > (EMC_CFG_DYN_SREF | EMC_CFG_DRAM_ACPD | EMC_CFG_DRAM_CLKSTOP_SR | \ >
On 16/05/2021 12:12, Dmitry Osipenko wrote: > Fix compilation warning on 64bit platforms caused by implicit promotion > of 32bit signed integer to a 64bit unsigned value which happens after > enabling compile-testing of the driver. > > Suggested-by: Nathan Chancellor <nathan@kernel.org> > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > --- > drivers/memory/tegra/tegra124-emc.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c > index 5699d909abc2..c9eb948cf4df 100644 > --- a/drivers/memory/tegra/tegra124-emc.c > +++ b/drivers/memory/tegra/tegra124-emc.c > @@ -272,8 +272,8 @@ > #define EMC_PUTERM_ADJ 0x574 > > #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... Best regards, Krzysztof
17.05.2021 14:28, Krzysztof Kozlowski пишет: > On 16/05/2021 12:12, Dmitry Osipenko wrote: >> Fix compilation warning on 64bit platforms caused by implicit promotion >> of 32bit signed integer to a 64bit unsigned value which happens after >> enabling compile-testing of the driver. >> >> Suggested-by: Nathan Chancellor <nathan@kernel.org> >> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> >> --- >> drivers/memory/tegra/tegra124-emc.c | 4 ++-- >> 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c >> index 5699d909abc2..c9eb948cf4df 100644 >> --- a/drivers/memory/tegra/tegra124-emc.c >> +++ b/drivers/memory/tegra/tegra124-emc.c >> @@ -272,8 +272,8 @@ >> #define EMC_PUTERM_ADJ 0x574 >> >> #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.
On 17/05/2021 09:35, Dmitry Osipenko wrote: > 17.05.2021 14:28, Krzysztof Kozlowski пишет: >> On 16/05/2021 12:12, Dmitry Osipenko wrote: >>> Fix compilation warning on 64bit platforms caused by implicit promotion >>> of 32bit signed integer to a 64bit unsigned value which happens after >>> enabling compile-testing of the driver. >>> >>> Suggested-by: Nathan Chancellor <nathan@kernel.org> >>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> >>> --- >>> drivers/memory/tegra/tegra124-emc.c | 4 ++-- >>> 1 file changed, 2 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c >>> index 5699d909abc2..c9eb948cf4df 100644 >>> --- a/drivers/memory/tegra/tegra124-emc.c >>> +++ b/drivers/memory/tegra/tegra124-emc.c >>> @@ -272,8 +272,8 @@ >>> #define EMC_PUTERM_ADJ 0x574 >>> >>> #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. 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. Best regards, Krzysztof
17.05.2021 16:39, Krzysztof Kozlowski пишет: > On 17/05/2021 09:35, Dmitry Osipenko wrote: >> 17.05.2021 14:28, Krzysztof Kozlowski пишет: >>> On 16/05/2021 12:12, Dmitry Osipenko wrote: >>>> Fix compilation warning on 64bit platforms caused by implicit promotion >>>> of 32bit signed integer to a 64bit unsigned value which happens after >>>> enabling compile-testing of the driver. >>>> >>>> Suggested-by: Nathan Chancellor <nathan@kernel.org> >>>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> >>>> --- >>>> drivers/memory/tegra/tegra124-emc.c | 4 ++-- >>>> 1 file changed, 2 insertions(+), 2 deletions(-) >>>> >>>> diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c >>>> index 5699d909abc2..c9eb948cf4df 100644 >>>> --- a/drivers/memory/tegra/tegra124-emc.c >>>> +++ b/drivers/memory/tegra/tegra124-emc.c >>>> @@ -272,8 +272,8 @@ >>>> #define EMC_PUTERM_ADJ 0x574 >>>> >>>> #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. > 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.
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
17.05.2021 17:04, Krzysztof Kozlowski пишет: > 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 Alright, I'll use the BIT macro in the v3. I also realized now that the tegra30-emc drivers needs the same change. Thank you for the review.
On 16/05/2021 12:12, Dmitry Osipenko wrote: > Fix compilation warning on 64bit platforms caused by implicit promotion > of 32bit signed integer to a 64bit unsigned value which happens after > enabling compile-testing of the driver. > > Suggested-by: Nathan Chancellor <nathan@kernel.org> > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> The patch was not suggested by Nathan but it was: Reported-by: kernel test robot <lkp@intel.com> Nathan however provided analysis and proper solution, so co-developed or his SoB fits better. This is not that important as comment above - including robot's credits. Best regards, Krzysztof
17.05.2021 17:24, Krzysztof Kozlowski пишет: > On 16/05/2021 12:12, Dmitry Osipenko wrote: >> Fix compilation warning on 64bit platforms caused by implicit promotion >> of 32bit signed integer to a 64bit unsigned value which happens after >> enabling compile-testing of the driver. >> >> Suggested-by: Nathan Chancellor <nathan@kernel.org> >> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > The patch was not suggested by Nathan but it was: > Reported-by: kernel test robot <lkp@intel.com> > > Nathan however provided analysis and proper solution, so co-developed or > his SoB fits better. This is not that important as comment above - > including robot's credits. I'll update the tags in v3, thank you.
diff --git a/drivers/memory/tegra/tegra124-emc.c b/drivers/memory/tegra/tegra124-emc.c index 5699d909abc2..c9eb948cf4df 100644 --- a/drivers/memory/tegra/tegra124-emc.c +++ b/drivers/memory/tegra/tegra124-emc.c @@ -272,8 +272,8 @@ #define EMC_PUTERM_ADJ 0x574 #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) #define EMC_CFG_POWER_FEATURES_MASK \ (EMC_CFG_DYN_SREF | EMC_CFG_DRAM_ACPD | EMC_CFG_DRAM_CLKSTOP_SR | \
Fix compilation warning on 64bit platforms caused by implicit promotion of 32bit signed integer to a 64bit unsigned value which happens after enabling compile-testing of the driver. Suggested-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> --- drivers/memory/tegra/tegra124-emc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)