[v1] x86/kexec: store segment registers directly to memory in crash_setup_regs()

[PATCH RESEND] x86/kexec: store segment registers directly to memory in crash_setup_regs()

Posted by Uros Bizjak 2 weeks, 5 days ago

crash_setup_regs() currently moves segment registers (ss, cs, ds, es)
into pt_regs via a general purpose register (%eax). Update the code to
avoid the intermediate register and store segment registers directly
to their destination fields. This reduces the generated code from:

 131:	8c d0                	mov    %ss,%eax
 133:	66 89 84 24 a0 00 00 	mov    %ax,0xa0(%rsp)
 13a:	00
 13b:	8c c8                	mov    %cs,%eax
 13d:	66 89 84 24 88 00 00 	mov    %ax,0x88(%rsp)
 144:	00

to:

 131:	8c 94 24 a0 00 00 00 	mov    %ss,0xa0(%rsp)
 138:	8c 8c 24 88 00 00 00 	mov    %cs,0x88(%rsp)

No functional change intended.

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Cc: Thomas Gleixner <tglx@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
---
 arch/x86/include/asm/kexec.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/kexec.h b/arch/x86/include/asm/kexec.h
index 5cfb27f26583..ad7ef567452b 100644
--- a/arch/x86/include/asm/kexec.h
+++ b/arch/x86/include/asm/kexec.h
@@ -106,11 +106,11 @@ static inline void crash_setup_regs(struct pt_regs *newregs,
 		asm volatile("mov %%r14,%0" : "=m"(newregs->r14));
 		asm volatile("mov %%r15,%0" : "=m"(newregs->r15));
 #endif
-		asm volatile("mov %%ss,%k0" : "=a"(newregs->ss));
-		asm volatile("mov %%cs,%k0" : "=a"(newregs->cs));
+		asm volatile("mov %%ss,%0" : "=m"(newregs->ss));
+		asm volatile("mov %%cs,%0" : "=m"(newregs->cs));
 #ifdef CONFIG_X86_32
-		asm volatile("mov %%ds,%k0" : "=a"(newregs->ds));
-		asm volatile("mov %%es,%k0" : "=a"(newregs->es));
+		asm volatile("mov %%ds,%0" : "=m"(newregs->ds));
+		asm volatile("mov %%es,%0" : "=m"(newregs->es));
 #endif
 		asm volatile("pushf\n\t"
 			     "pop %0" : "=m"(newregs->flags));
-- 
2.55.0

Re: [PATCH RESEND] x86/kexec: store segment registers directly to memory in crash_setup_regs()

Posted by H. Peter Anvin 2 weeks, 4 days ago

On 2026-09-06 11:39, Uros Bizjak wrote:
> crash_setup_regs() currently moves segment registers (ss, cs, ds, es)
> into pt_regs via a general purpose register (%eax). Update the code to
> avoid the intermediate register and store segment registers directly
> to their destination fields. This reduces the generated code from:
> 

> diff --git a/arch/x86/include/asm/kexec.h b/arch/x86/include/asm/kexec.h
> index 5cfb27f26583..ad7ef567452b 100644
> --- a/arch/x86/include/asm/kexec.h
> +++ b/arch/x86/include/asm/kexec.h
> @@ -106,11 +106,11 @@ static inline void crash_setup_regs(struct pt_regs *newregs,
>  		asm volatile("mov %%r14,%0" : "=m"(newregs->r14));
>  		asm volatile("mov %%r15,%0" : "=m"(newregs->r15));
>  #endif
> -		asm volatile("mov %%ss,%k0" : "=a"(newregs->ss));
> -		asm volatile("mov %%cs,%k0" : "=a"(newregs->cs));
> +		asm volatile("mov %%ss,%0" : "=m"(newregs->ss));
> +		asm volatile("mov %%cs,%0" : "=m"(newregs->cs));
>  #ifdef CONFIG_X86_32
> -		asm volatile("mov %%ds,%k0" : "=a"(newregs->ds));
> -		asm volatile("mov %%es,%k0" : "=a"(newregs->es));
> +		asm volatile("mov %%ds,%0" : "=m"(newregs->ds));
> +		asm volatile("mov %%es,%0" : "=m"(newregs->es));
>  #endif
>  		asm volatile("pushf\n\t"
>  			     "pop %0" : "=m"(newregs->flags));

Why not use "=rm"? "=a" seems odd in the extreme.

	-hpa

Re: [PATCH RESEND] x86/kexec: store segment registers directly to memory in crash_setup_regs()

Posted by Uros Bizjak 2 weeks, 4 days ago

On Mon, Sep 7, 2026 at 3:33 AM H. Peter Anvin <hpa@zytor.com> wrote:
>
> On 2026-09-06 11:39, Uros Bizjak wrote:
> > crash_setup_regs() currently moves segment registers (ss, cs, ds, es)
> > into pt_regs via a general purpose register (%eax). Update the code to
> > avoid the intermediate register and store segment registers directly
> > to their destination fields. This reduces the generated code from:
> >
>
> > diff --git a/arch/x86/include/asm/kexec.h b/arch/x86/include/asm/kexec.h
> > index 5cfb27f26583..ad7ef567452b 100644
> > --- a/arch/x86/include/asm/kexec.h
> > +++ b/arch/x86/include/asm/kexec.h
> > @@ -106,11 +106,11 @@ static inline void crash_setup_regs(struct pt_regs *newregs,
> >               asm volatile("mov %%r14,%0" : "=m"(newregs->r14));
> >               asm volatile("mov %%r15,%0" : "=m"(newregs->r15));
> >  #endif
> > -             asm volatile("mov %%ss,%k0" : "=a"(newregs->ss));
> > -             asm volatile("mov %%cs,%k0" : "=a"(newregs->cs));
> > +             asm volatile("mov %%ss,%0" : "=m"(newregs->ss));
> > +             asm volatile("mov %%cs,%0" : "=m"(newregs->cs));
> >  #ifdef CONFIG_X86_32
> > -             asm volatile("mov %%ds,%k0" : "=a"(newregs->ds));
> > -             asm volatile("mov %%es,%k0" : "=a"(newregs->es));
> > +             asm volatile("mov %%ds,%0" : "=m"(newregs->ds));
> > +             asm volatile("mov %%es,%0" : "=m"(newregs->es));
> >  #endif
> >               asm volatile("pushf\n\t"
> >                            "pop %0" : "=m"(newregs->flags));
>
> Why not use "=rm"? "=a" seems odd in the extreme.

"=rm" will produce assembly using an intermediate register due to GCC
PR124209 [1]. clang also has issues with "=rm", please see the
definition of ASM_OUTPUT_RM in compiler-clang.h. Using "=m" will
always create optimal code on both compilers.

[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124209

Uros.