On Wed, Sep 18, 2019 at 10:31 PM Xiaoyao Li <xiaoyao.li@xxxxxxxxx> wrote:
> I vote for Sean's one-off case, how about something like this:
>
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index 22c2720cd948..6af5febf7b12 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -976,11 +976,23 @@ bool kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax,
> u32 *ebx,
> u32 *ecx, u32 *edx, bool check_limit)
> {
> u32 function = *eax, index = *ecx;
> - struct kvm_cpuid_entry2 *best;
> + struct kvm_cpuid_entry2 *best, tmp;
> bool entry_found = true;
>
> best = kvm_find_cpuid_entry(vcpu, function, index);
>
> + if (!best && (fuction == 0xb || function == 0x1f) && index > 0) {
> + best = kvm_find_cpuid_entry(vcpu, function, 0);
> + if (best) {
> + tmp.eax = 0;
> + tmp.ebx = 0;
> + tmp.ecx = index & 0xff;
> + tmp.edx = best->edx;
> + best = &tmp;
> + goto out;
> + }
> + }
> +
I don't believe this works for the case where 0BH or 1FH is the
maximum basic leaf, in which case all out-of-range leaves should have
this behavior. But I'll go ahead and work up a solution using this
two-off :-) approach.
