Re: kern/59170 (bogus reply from NVMM for invalid CPUID on AMD CPU)

[riastradh%NetBSD.org@localhost]

Synopsis: bogus reply from NVMM for invalid CPUID on AMD CPU

State-Changed-From-To: open->closed
State-Changed-By: riastradh%NetBSD.org@localhost
State-Changed-When: Wed, 13 May 2026 05:41:55 +0000
State-Changed-Why:
The first part of this PR, prompted by crashes arising after
https://mail-index.netbsd.org/source-changes/2025/03/06/msg155820.html
(`Add support for CPUID leaf 0x40000010', first time around) was bogus:

> On AMD CPUs, NVMM default reply to invalid CPUID leafs can lead
> to panic().

NVMM's reply was correct, and this is now explained in comments.

The real problem leading to the panic was that the _guest_ was using
ci->ci_max_ext_cpuid (meant for the 0x80000000..0x8fffffff range) when
it should have used ci->ci_max_hypervisor_cpuid.  That was rectified in
https://mail-index.netbsd.org/source-changes/2025/05/01/msg156713.html
(`Introduce cpu_max_hypervisor_cpuid to cache hypervisor CPUID leaf'),
https://mail-index.netbsd.org/source-changes/2025/05/02/msg156733.html
(`Add support for CPUID leaf 0x40000010', second time around).

I think there is nothing left here and I doubt it's worthwhile to pull
any of this up to netbsd-10 or earlier.