Re: [PATCH v4 3/3] arm64: paravirt: Enable errata based on implementation CPUs

[Cornelia Huck <cohuck@redhat.com>]

On Fri, Dec 20 2024, Marc Zyngier <maz@kernel.org> wrote:

> On Thu, 19 Dec 2024 17:40:55 +0000,
> Oliver Upton <oliver.upton@linux.dev> wrote:
>> 
>> > Independent of this, I wonder what we should output in sysfs
>> > (/sys/devices/system/cpu/cpu*/regs/identification/*).
>> 
>> It's a bit crap, but maybe implementation index 0 gets reported through
>> the 'main' midr/revidr files, otherwise have a directory per
>> implementation index of midr/revidr.
>
> Having slept on that one, I'm starting to think that we should keep
> the status-quo of reporting what the kernel snapshot at boot time.
> There is no good reason to force the VMM to report the potential
> implementations in any specific order.
>
> The "alternative-implementations" is interesting, but we don't keep it
> per-CPU, so I don't  think it fits the current scheme. But maybe
> something in /sys/devices/system/cpu, outside of the cpu* hierarchy?

I agree that this should be reported outside of the cpu* hierarchy,
maybe smth like

/sys/devices/system/cpu/alternatives/
   identification/
     regs0/
       aidr_el1
       midr_el1
       revidr_el1
     regs1/
       aidr_el1
       midr_el1
       revidr_el1

(with no guarantee as to _which_ triplet will show up under regs<n>)

Also depends on the intended consumers, I guess.

>
> Either way, this isn't something we should worry too much right now.

I agree as well.