On Tue, Mar 03, 2020 at 09:48:44AM +0100, Paolo Bonzini wrote:
> On 02/03/20 20:57, Sean Christopherson wrote:
> > Two fixes related to out-of-range CPUID emulation and related cleanup on
> > top.
> >
> > I have a unit test and also manually verified a few interesting cases.
> > I'm not planning on posting the unit test at this time because I haven't
> > figured out how to avoid false positives, e.g. if a random in-bounds
> > leaf just happens to match the output of a max basic leaf. It might be
> > doable by hardcoding the cpu model?
>
> It would be best suited for selftests rather than kvm-unit-tests. But I
> don't really see the benefit of anything more than just
Gotta save those stack bytes?
I got a bit confused by the "max" variable; I thought it would hold the
max basic leaf, not CPUID.0x0. Removing it seemed easier than trying to
come up with a better name :-)
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index b1c469446b07..c1abf5de4461 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -1001,6 +1001,7 @@ bool kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32 *ebx,
> u32 *ecx, u32 *edx, bool check_limit)
> {
> u32 function = *eax, index = *ecx;
> + u32 orig_function = function;
> struct kvm_cpuid_entry2 *entry;
> struct kvm_cpuid_entry2 *max;
> bool found;
> @@ -1049,7 +1050,7 @@ bool kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32 *ebx,
> }
> }
> }
> - trace_kvm_cpuid(function, *eax, *ebx, *ecx, *edx, found);
> + trace_kvm_cpuid(orig_function, *eax, *ebx, *ecx, *edx, found);
> return found;
> }
> EXPORT_SYMBOL_GPL(kvm_cpuid);
>
