Re: [PATCH 1/6] x86/cpu: Break Vendor/Family/Model macros into separate header

[Sohil Mehta <sohil.mehta@intel.com>]

On 2/11/2026 1:12 PM, Dave Hansen wrote:
> 
> So PECI and x86 both want the same data, but they do very different
> things with it.
> 
> So let's just duplicate the constants. Completely untested patch attached.
> 

This looks okay in principle. I haven't tested it yet. But, there seems
to be an inconsistency here?

>  static int peci_device_info_init(struct peci_device *device)
>  {
>  	u8 revision;
> @@ -100,7 +80,7 @@ static int peci_device_info_init(struct
>  	if (ret)
>  		return ret;
>  
> -	device->info.x86_vfm = IFM(peci_x86_cpu_family(cpu_id), peci_x86_cpu_model(cpu_id));
> +	device->info.device_id = cpu_id >> 4;
>  

I imagine this shifting the stepping. Should this be cpu_id & 0xF?

Because the devices are stored with the stepping info:

> +#define PECI_INTEL_HASWELL_X		0x306C0
> +#define PECI_INTEL_BROADWELL_X		0x406F0
> +#define PECI_INTEL_BROADWELL_D		0x50660
> +#define PECI_INTEL_SKYLAKE_X		0x50650
> +
> +#define PECI_INTEL_ICELAKE_X		0x606A0
> +#define PECI_INTEL_ICELAKE_D		0x606C0
> +#define PECI_INTEL_SAPPHIRERAPIDS_X	0x806F0
> +#define PECI_INTEL_EMERALDRAPIDS_X	0xC06F0
>  

Or maybe define the CPUs without the stepping bits?


>  	ret = peci_get_revision(device, &revision);
>  	if (ret)