Re: [PATCH] x86/fpu: make WARN_ON_FPU get fully optimized out

[Sean Christopherson <seanjc@google.com>]

On Thu, Dec 26, 2024, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> Currently WARN_ON_FPU evaluates its argument even if
> CONFIG_X86_DEBUG_FPU is disabled, which adds unnecessary instructions to
> several functions.  Fix this by making the argument evaluation in the
> no-debug case conditional on if (0), similar to what no_printk() does.
> 
> Fixes: 83242c515881 ("x86/fpu: Make WARN_ON_FPU() more robust in the !CONFIG_X86_DEBUG_FPU case")
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
>  arch/x86/kernel/fpu/internal.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/fpu/internal.h b/arch/x86/kernel/fpu/internal.h
> index dbdb31f55fc7..deb0a31b75e4 100644
> --- a/arch/x86/kernel/fpu/internal.h
> +++ b/arch/x86/kernel/fpu/internal.h
> @@ -16,11 +16,11 @@ static __always_inline __pure bool use_fxsr(void)
>  }
>  
>  #ifdef CONFIG_X86_DEBUG_FPU
>  # define WARN_ON_FPU(x) WARN_ON_ONCE(x)
>  #else
> -# define WARN_ON_FPU(x) ({ (void)(x); 0; })
> +# define WARN_ON_FPU(x) ({ if (0) (void)(x); 0; })

BUILD_BUG_ON_INVALID() already takes care of this, e.g. I verified this also makes
the code generation go away:

# define WARN_ON_FPU(x) ({ BUILD_BUG_ON_INVALID(x); 0; })

>  #endif
>  
>  /* Used in init.c */
>  extern void fpstate_init_user(struct fpstate *fpstate);
>  extern void fpstate_reset(struct fpu *fpu);
> 
> base-commit: d6ef8b40d075c425f548002d2f35ae3f06e9cf96
> -- 
> 2.47.1
>