Re: [PATCH -next] arm64: Optimize the comparison of unsigned expressions to avoid compiling error

From: Ard Biesheuvel
Date: Tue Feb 21 2023 - 10:10:56 EST


On Tue, 21 Feb 2023 at 11:37, Mark Rutland <mark.rutland@xxxxxxx> wrote:
>
> On Tue, Feb 21, 2023 at 09:27:40AM +0800, Lin Yujun wrote:
> > while compile arch/arm64/include/asm/cpufeature.h with
> > -Werror=type-limits enabled, errors shown as below:
> >
> > ./arch/arm64/include/asm/cpufeature.h: In function 'system_supports_4kb_granule':
> > ./arch/arm64/include/asm/cpufeature.h:653:14: error: comparison of unsigned expression >= 0 is always true [-Werror=type-limits]
> > return (val >= ID_AA64MMFR0_TGRAN4_SUPPORTED_MIN) &&
> > ^~
> > ./arch/arm64/include/asm/cpufeature.h: In function 'system_supports_64kb_granule':
> > ./arch/arm64/include/asm/cpufeature.h:666:14: error: comparison of unsigned expression >= 0 is always true [-Werror=type-limits]
> > return (val >= ID_AA64MMFR0_TGRAN64_SUPPORTED_MIN) &&
> > ^~

The usefulness of this diagnostic is debatable even when the right
hand is a literal '0' but warning about symbolic constants like this
is actively evil.

In general, preprocessor #defined values can depend on Kconfig
settings or other build configuration switches, and this warning
encourages us to remove the expression altogether, which could mean
introducing a bug if the macro may assume values >0 in other
configurations.

Ergo, we must ignore -Wtype-limits until it is fixed, and can at least
distinguish literal 0 constants from ones instantiated by a CPP macro.


>
> When is the `-Werror=type-limits` flag enabled by the build system?
>
> We have patterns like this all over the kernel, and I don't think this is
> indicative of a real problem, and I don't think that we need to change code to
> make this warning disappear.
>
> > Modify the return judgment statement, use
> > "((val - min) < (val - max - 1))" to confirm that returns
> > true in “min <= val <= max” cases, false in other cases.
>
> That expression is far less clear than the existing code, so I do not think
> that is a good idea.
>
> > Fixes: 79d82cbcbb3d ("arm64/kexec: Test page size support with new TGRAN range values")
>
> What functional error does this fix?
>
> What configuration is broken?
>
> Thanks,
> Mark.
>
> > Signed-off-by: Lin Yujun <linyujun809@xxxxxxxxxx>
> > ---
> > arch/arm64/include/asm/cpufeature.h | 18 ++++++++++++------
> > 1 file changed, 12 insertions(+), 6 deletions(-)
> >
> > diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> > index 03d1c9d7af82..0a6bda025141 100644
> > --- a/arch/arm64/include/asm/cpufeature.h
> > +++ b/arch/arm64/include/asm/cpufeature.h
> > @@ -54,6 +54,9 @@ enum ftr_type {
> > #define FTR_VISIBLE_IF_IS_ENABLED(config) \
> > (IS_ENABLED(config) ? FTR_VISIBLE : FTR_HIDDEN)
> >
> > +#define IN_RANGE_INCLUSIVE(val, min, max) \
> > + (((val) - (min)) < ((val) - (max) - 1))
> > +
> > struct arm64_ftr_bits {
> > bool sign; /* Value is signed ? */
> > bool visible;
> > @@ -693,8 +696,9 @@ static inline bool system_supports_4kb_granule(void)
> > val = cpuid_feature_extract_unsigned_field(mmfr0,
> > ID_AA64MMFR0_EL1_TGRAN4_SHIFT);
> >
> > - return (val >= ID_AA64MMFR0_EL1_TGRAN4_SUPPORTED_MIN) &&
> > - (val <= ID_AA64MMFR0_EL1_TGRAN4_SUPPORTED_MAX);
> > + return IN_RANGE_INCLUSIVE(val,
> > + ID_AA64MMFR0_EL1_TGRAN4_SUPPORTED_MIN,
> > + ID_AA64MMFR0_EL1_TGRAN4_SUPPORTED_MAX);
> > }
> >
> > static inline bool system_supports_64kb_granule(void)
> > @@ -706,8 +710,9 @@ static inline bool system_supports_64kb_granule(void)
> > val = cpuid_feature_extract_unsigned_field(mmfr0,
> > ID_AA64MMFR0_EL1_TGRAN64_SHIFT);
> >
> > - return (val >= ID_AA64MMFR0_EL1_TGRAN64_SUPPORTED_MIN) &&
> > - (val <= ID_AA64MMFR0_EL1_TGRAN64_SUPPORTED_MAX);
> > + return IN_RANGE_INCLUSIVE(val,
> > + ID_AA64MMFR0_EL1_TGRAN64_SUPPORTED_MIN,
> > + ID_AA64MMFR0_EL1_TGRAN64_SUPPORTED_MAX);
> > }
> >
> > static inline bool system_supports_16kb_granule(void)
> > @@ -719,8 +724,9 @@ static inline bool system_supports_16kb_granule(void)
> > val = cpuid_feature_extract_unsigned_field(mmfr0,
> > ID_AA64MMFR0_EL1_TGRAN16_SHIFT);
> >
> > - return (val >= ID_AA64MMFR0_EL1_TGRAN16_SUPPORTED_MIN) &&
> > - (val <= ID_AA64MMFR0_EL1_TGRAN16_SUPPORTED_MAX);
> > + return IN_RANGE_INCLUSIVE(val,
> > + ID_AA64MMFR0_EL1_TGRAN16_SUPPORTED_MIN,
> > + ID_AA64MMFR0_EL1_TGRAN16_SUPPORTED_MAX);
> > }
> >
> > static inline bool system_supports_mixed_endian_el0(void)
> > --
> > 2.34.1
> >