Since objtool's inception, frame pointer warnings have been manually
silenced for __arch_hweight*() to allow those functions' inline asm to
avoid using ASM_CALL_CONSTRAINT.
The potentially dubious reasoning for that decision over nine years ago
was that since !X86_FEATURE_POPCNT is exceedingly rare, it's not worth
hurting the code layout for a function call that will never happen on
the vast majority of systems.
However, those functions actually started using ASM_CALL_CONSTRAINT with
the following commit:
194a613088a8 ("x86/hweight: Use ASM_CALL_CONSTRAINT in inline asm()")
And rightfully so, as it makes the code correct. ASM_CALL_CONSTRAINT
will soon have no effect for non-FP configs anyway.
With ASM_CALL_CONSTRAINT in place, ANNOTATE_IGNORE_ALTERNATIVE no longer
has a purpose for the hweight functions. Remove it.
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
This was originally posted with the RFC series:
https://lore.kernel.org/cover.1744098446.git.jpoimboe@kernel.org
But it's actually a standalone patch which can be merged independently.
arch/x86/include/asm/arch_hweight.h | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/arch/x86/include/asm/arch_hweight.h b/arch/x86/include/asm/arch_hweight.h
index cbc6157f0b4b..b5982b94bdba 100644
@@ -16,8 +16,7 @@ static __always_inline unsigned int __arch_hweight32(unsigned int w)
{
unsigned int res;
- asm_inline (ALTERNATIVE(ANNOTATE_IGNORE_ALTERNATIVE
- "call __sw_hweight32",
+ asm_inline (ALTERNATIVE("call __sw_hweight32",
"popcntl %[val], %[cnt]", X86_FEATURE_POPCNT)
: [cnt] "=" REG_OUT (res), ASM_CALL_CONSTRAINT
: [val] REG_IN (w));
@@ -46,8 +45,7 @@ static __always_inline unsigned long __arch_hweight64(__u64 w)
{
unsigned long res;
- asm_inline (ALTERNATIVE(ANNOTATE_IGNORE_ALTERNATIVE
- "call __sw_hweight64",
+ asm_inline (ALTERNATIVE("call __sw_hweight64",
"popcntq %[val], %[cnt]", X86_FEATURE_POPCNT)
: [cnt] "=" REG_OUT (res), ASM_CALL_CONSTRAINT
: [val] REG_IN (w));
--
2.49.0