Re: [RFC PATCH 2/4] KVM: x86: Extract VMXON and EFER.SVME enablement to kernel

[Sean Christopherson <seanjc@google.com>]

On Mon, Oct 13, 2025, Chao Gao wrote:
> >+void __init x86_virt_init(void)
> >+{
> >+	cpu_emergency_virt_cb *vmx_cb = NULL, *svm_cb = NULL;
> >+
> >+	if (x86_virt_is_vmx())
> >+		vmx_cb = x86_vmx_init();
> >+
> >+	if (x86_virt_is_svm())
> >+		svm_cb = x86_svm_init();
> >+
> >+	if (!vmx_cb && !svm_cb)
> >+		return;
> >+
> >+	if (WARN_ON_ONCE(vmx_cb && svm_cb))
> >+		return;
> >+
> >+	cpu_emergency_register_virt_callback(vmx_cb ? : svm_cb);
> 
> To be consistent with x86_virt_{get,put}_cpu(), perhaps we can have a common
> emergency callback and let reboot.c call it directly, with the common callback
> routing to svm/vmx code according to the hardware type.

Oh, yeah, that's a much better idea, especially if x86_virt_init() runs
unconditionally during boot (as proposed).  cpu_emergency_disable_virtualization()
can be dropped entirely (it'd just be a one-line wrapper).