Re: [PATCH v1 8/9] x86/cpuid: Rename native_cpuid() to cpuid_native()

On Thu, 15 May 2025, Ingo Molnar wrote:
>
> * Ahmed S. Darwish <[email protected]> wrote:
> >
> > diff --git a/arch/x86/boot/compressed/sev.c b/arch/x86/boot/compressed/sev.c
> > index f054bf52c10a..4b617f8f9b7f 100644
> > --- a/arch/x86/boot/compressed/sev.c
> > +++ b/arch/x86/boot/compressed/sev.c
> > @@ -494,7 +494,7 @@ static int sev_check_cpu_support(void)
> >  	/* Check for the SME/SEV support leaf */
> >  	eax = 0x80000000;
> >  	ecx = 0;
> > -	native_cpuid(&eax, &ebx, &ecx, &edx);
> > +	cpuid_native(&eax, &ebx, &ecx, &edx);
> >  	if (eax < 0x8000001f)
> >  		return -ENODEV;
>
> I'm torn about this one.
>
> native_*() is a pretty generic way to denote baremetal methods, all
> around the x86 tree:
>
>   starship:~/tip> git grep -E '\<native_' arch/x86/ | wc -l
>   682
...
>
> I think it might be better to leave these as native_cpuid_*, because
> shifting a naming inconsistency into another namespace isn't really a
> net improvement.
>

Yeah, I admit I was a little bit uncomfortable about this one, especially
when modifying a call-site like:

    => arch/x86/kernel/paravirt.c:
    struct paravirt_patch_template pv_ops = {
	.cpu.io_delay		= native_io_delay,
	.cpu.cpuid		= cpuid_native,		// stands out
	.cpu.read_cr0		= native_read_cr0,
	.cpu.write_cr0		= native_write_cr0,
	.cpu.write_cr4		= native_write_cr4,
	.cpu.read_msr		= native_read_msr,
	...
    };

So, indeed, ACK.

Thanks!
~ Ahmed