Re: [PATCH] Ability to read the MKTME status from userspace

[Andy Lutomirski]

> On Jun 19, 2020, at 9:17 AM, Borislav Petkov <bp@xxxxxxxxx> wrote:
> 
> ïOn Fri, Jun 19, 2020 at 08:48:47AM -0700, Andy Lutomirski wrote:
>> Boris, it wouldnât be totally crazy for cpuinfo to learn to
>> distinguish between âyour platform has this feature but Linux
>> isnât using itâ and âyour platform doesnât have this feature
>> in the first placeâ. And I suppose thereâs this extra silly state
>> âyour platform has this feature, but your firmware didnât enable
>> itâ. This would be a big job.
> 
> Well, I believe all the kernel can do is supply bits of information -
> just like MSRs - and depending on the settings of those bits, userspace
> can decide what the situation is. For example:
> 
> bit 0 - CPUID support
> bit 1 - BIOS enabled
> bit 2 - quirk applied
> bit 3 - microcode fixes present
> ...
> 
> and so on.
> 

Indeed.

> It needs a proper definition though and userspace to say, yes, we want
> that and that is useful for us.

Maybe the right approach is to just keep this kind of use in mind for when we inevitably redo cpu features when Intel ships their hybrid Atom/Core machines.

> 
> Where it ends up is then beside the point - /proc/cpuinfo,
> /sys/devices/system/cpu, whatever...
> 
> -- 
> Regards/Gruss,
>    Boris.
> 
> https://people.kernel.org/tglx/notes-about-netiquette