[PATCH 0/2] x86: Use "er" asm constriant for add/sub

[PATCH 0/2] x86: Use "er" asm constriant for add/sub

[David Laight]

The x86 instruction set only supports 32bit signed immediate values
for add/sub.
Replace the "ir" constraint with "er" to avoid build errors.

Found by a patch that added used mul_u64_add_u64_div_u64() to
do a rounding divide by 2^32 (perhaps not the best way to do this).

A quick grep only found one other affected file.

David Laight (2):
  lib: mul_u64_add_u64_div_u64: Fix addition of large constants
  x86/local: local_add/local_sub: Support large immediate values

 arch/x86/include/asm/div64.h | 2 +-
 arch/x86/include/asm/local.h | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

-- 
2.39.5

[PATCH 1/2] lib: mul_u64_add_u64_div_u64: Fix addition of large constants

[David Laight]

Adding constants over 2^31 fails to compile because the ADD instruction
only supports 32bit signed immediates.

Replace the "irm" constraint with "erm" so that the compiler loads
large constants into a register.

Found by a patch to drivers/iio/frequency/ad9910.c

Fixes: 6480241f31f5 ("lib: add mul_u64_add_u64_div_u64() and mul_u64_u64_div_u64_roundup()")
Signed-off-by: David Laight <david.laight.linux@gmail.com>
---
 arch/x86/include/asm/div64.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/div64.h b/arch/x86/include/asm/div64.h
index 30fd06ede751..8a2d343f977e 100644
--- a/arch/x86/include/asm/div64.h
+++ b/arch/x86/include/asm/div64.h
@@ -111,7 +111,7 @@ static inline u64 mul_u64_add_u64_div_u64(u64 rax, u64 mul, u64 add, u64 div)
 
 	if (!statically_true(!add))
 		asm ("addq %[add], %[lo]; adcq $0, %[hi]" :
-			[lo] "+r" (rax), [hi] "+r" (rdx) : [add] "irm" (add));
+			[lo] "+r" (rax), [hi] "+r" (rdx) : [add] "erm" (add));
 
 	asm ("divq %[div]" : "+a" (rax), "+d" (rdx) : [div] "rm" (div));
 
-- 
2.39.5

[PATCH 2/2] x86/local: local_add/local_sub: Support large immediate values

[David Laight]

Replace the "ir" constraint with "er" so that constants over 2^31
get loaded into a register.

Signed-off-by: David Laight <david.laight.linux@gmail.com>
---
 arch/x86/include/asm/local.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/local.h b/arch/x86/include/asm/local.h
index 4957018fef3e..68af7b74450a 100644
--- a/arch/x86/include/asm/local.h
+++ b/arch/x86/include/asm/local.h
@@ -32,14 +32,14 @@ static inline void local_add(long i, local_t *l)
 {
 	asm volatile(_ASM_ADD "%1,%0"
 		     : "+m" (l->a.counter)
-		     : "ir" (i));
+		     : "er" (i));
 }
 
 static inline void local_sub(long i, local_t *l)
 {
 	asm volatile(_ASM_SUB "%1,%0"
 		     : "+m" (l->a.counter)
-		     : "ir" (i));
+		     : "er" (i));
 }
 
 /**
-- 
2.39.5

Re: [PATCH 0/2] x86: Use "er" asm constriant for add/sub

[Peter Zijlstra]

On Mon, Aug 03, 2026 at 10:47:00AM +0100, David Laight wrote:
> The x86 instruction set only supports 32bit signed immediate values
> for add/sub.
> Replace the "ir" constraint with "er" to avoid build errors.
> 
> Found by a patch that added used mul_u64_add_u64_div_u64() to
> do a rounding divide by 2^32 (perhaps not the best way to do this).
> 
> A quick grep only found one other affected file.

I suppose the percpu stuff works because of __pcpu_cast_4 truncating the
value to u32 and then ignoring the whole signed/unsigned business. And
__pcpu_reg_imm_8 has "re".

Re: [PATCH 2/2] x86/local: local_add/local_sub: Support large immediate values

[Peter Zijlstra]

On Mon, Aug 03, 2026 at 10:47:02AM +0100, David Laight wrote:
> Replace the "ir" constraint with "er" so that constants over 2^31
> get loaded into a register.
> 
> Signed-off-by: David Laight <david.laight.linux@gmail.com>
> ---
>  arch/x86/include/asm/local.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/include/asm/local.h b/arch/x86/include/asm/local.h
> index 4957018fef3e..68af7b74450a 100644
> --- a/arch/x86/include/asm/local.h
> +++ b/arch/x86/include/asm/local.h
> @@ -32,14 +32,14 @@ static inline void local_add(long i, local_t *l)
>  {
>  	asm volatile(_ASM_ADD "%1,%0"
>  		     : "+m" (l->a.counter)
> -		     : "ir" (i));
> +		     : "er" (i));
>  }
>  
>  static inline void local_sub(long i, local_t *l)
>  {
>  	asm volatile(_ASM_SUB "%1,%0"
>  		     : "+m" (l->a.counter)
> -		     : "ir" (i));
> +		     : "er" (i));
>  }

And the other two sites already have "er", how inconsistent :/

Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>

Re: [PATCH 2/2] x86/local: local_add/local_sub: Support large immediate values

[H. Peter Anvin]

On 2026-08-03 02:47, David Laight wrote:
> Replace the "ir" constraint with "er" so that constants over 2^31
> get loaded into a register.
> 
> Signed-off-by: David Laight <david.laight.linux@gmail.com>

Yep, that's right.

Reviewed-by: H. Peter Anvin <hpa@zytor.com>

> ---
>  arch/x86/include/asm/local.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/include/asm/local.h b/arch/x86/include/asm/local.h
> index 4957018fef3e..68af7b74450a 100644
> --- a/arch/x86/include/asm/local.h
> +++ b/arch/x86/include/asm/local.h
> @@ -32,14 +32,14 @@ static inline void local_add(long i, local_t *l)
>  {
>  	asm volatile(_ASM_ADD "%1,%0"
>  		     : "+m" (l->a.counter)
> -		     : "ir" (i));
> +		     : "er" (i));
>  }
>  
>  static inline void local_sub(long i, local_t *l)
>  {
>  	asm volatile(_ASM_SUB "%1,%0"
>  		     : "+m" (l->a.counter)
> -		     : "ir" (i));
> +		     : "er" (i));
>  }
>  
>  /**

Re: [PATCH 1/2] lib: mul_u64_add_u64_div_u64: Fix addition of large constants

[H. Peter Anvin]

On 2026-08-03 02:47, David Laight wrote:
> Adding constants over 2^31 fails to compile because the ADD instruction
> only supports 32bit signed immediates.
> 
> Replace the "irm" constraint with "erm" so that the compiler loads
> large constants into a register.
> 
> Found by a patch to drivers/iio/frequency/ad9910.c
> 
> Fixes: 6480241f31f5 ("lib: add mul_u64_add_u64_div_u64() and mul_u64_u64_div_u64_roundup()")
> Signed-off-by: David Laight <david.laight.linux@gmail.com>

So it is.

Reviewed-by: H. Peter Anvin <hpa@zytor.com>

> ---
>  arch/x86/include/asm/div64.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/div64.h b/arch/x86/include/asm/div64.h
> index 30fd06ede751..8a2d343f977e 100644
> --- a/arch/x86/include/asm/div64.h
> +++ b/arch/x86/include/asm/div64.h
> @@ -111,7 +111,7 @@ static inline u64 mul_u64_add_u64_div_u64(u64 rax, u64 mul, u64 add, u64 div)
>  
>  	if (!statically_true(!add))
>  		asm ("addq %[add], %[lo]; adcq $0, %[hi]" :
> -			[lo] "+r" (rax), [hi] "+r" (rdx) : [add] "irm" (add));
> +			[lo] "+r" (rax), [hi] "+r" (rdx) : [add] "erm" (add));
>  
>  	asm ("divq %[div]" : "+a" (rax), "+d" (rdx) : [div] "rm" (div));
>  

Re: [PATCH 0/2] x86: Use "er" asm constriant for add/sub

[Nicolas Pitre]

On Mon, 3 Aug 2026, David Laight wrote:

> The x86 instruction set only supports 32bit signed immediate values
> for add/sub.
> Replace the "ir" constraint with "er" to avoid build errors.
> 
> Found by a patch that added used mul_u64_add_u64_div_u64() to
> do a rounding divide by 2^32 (perhaps not the best way to do this).
> 
> A quick grep only found one other affected file.

Would be a good idea adding those cases to the test module.


Nicolas

Re: [PATCH 0/2] x86: Use "er" asm constriant for add/sub

[David Laight]

On Mon, 3 Aug 2026 10:49:40 -0400 (EDT)
Nicolas Pitre <nico@fluxnic.net> wrote:

> On Mon, 3 Aug 2026, David Laight wrote:
> 
> > The x86 instruction set only supports 32bit signed immediate values
> > for add/sub.
> > Replace the "ir" constraint with "er" to avoid build errors.
> > 
> > Found by a patch that added used mul_u64_add_u64_div_u64() to
> > do a rounding divide by 2^32 (perhaps not the best way to do this).
> > 
> > A quick grep only found one other affected file.  
> 
> Would be a good idea adding those cases to the test module.

They fail to compile so it isn't a big deal.

	David

> 
> 
> Nicolas