x86: gate smbios hypervisor identification behind vm_guest

[kevans]

Summary

cpuid detection may have picked up a more specific guest type already,
and a follow-up check of smbios vendor/product may erroneously blow
away the previously detected type.

I considered moving these calls back into identify_hypervisor(), but
elected for the less invasive change; no objection to switching it.

Fixes: 2fee87562948 ("abstract out the vm detection via smbios..")

[kevans]

See also: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=270239, this is the cause of at least Hyper-V failing to boot.

[imp]

I'm seeing something weird on arm64 as well, so I wonder if a similar bug was introduce in its machdep.c too?
though I'm seeing it with a loaded module specifying the machine's UUID...
I'll have to chase it down tomorrow, since I need sleep tonight...

In any event, this change looks good to me.

[imp]

Though my issue is uuid, and is likely completely orthogonal to this...

sys/dev/smbios/smbios_subr.c

72

though I do wonder if this is the right assert for arm64...

[kevans]

sys/dev/smbios/smbios_subr.c

72

arm64 doesn't have any other way of detecting that we're a VM guest, so it won't actually hit these

[kib]

No objections, but IMO it would be slightly cleaner to check the condition in identify_hypervisor_smbios() and immediately return, instead of asserting.

[kevans]

In D39140#890858, @kib wrote:

No objections, but IMO it would be slightly cleaner to check the condition in identify_hypervisor_smbios() and immediately return, instead of asserting.

Sure, I don't have strong feelings about it, so I'll just go ahead and do that.