On 5/31/2020 10:08 AM, Li,Rongqing wrote:
-----邮件原件-----
发件人: Xiaoyao Li [mailto:xiaoyao.li@xxxxxxxxx]
发送时间: 2020年5月30日 18:40
收件人: Li,Rongqing <lirongqing@xxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx;
kvm@xxxxxxxxxxxxxxx; x86@xxxxxxxxxx; hpa@xxxxxxxxx; bp@xxxxxxxxx;
mingo@xxxxxxxxxx; tglx@xxxxxxxxxxxxx; jmattson@xxxxxxxxxx;
wanpengli@xxxxxxxxxxx; vkuznets@xxxxxxxxxx;
sean.j.christopherson@xxxxxxxxx; pbonzini@xxxxxxxxxx;
wei.huang2@xxxxxxx
主题: Re: [PATCH][v5] KVM: X86: support APERF/MPERF registers
On 5/30/2020 12:35 PM, Li RongQing wrote:
Guest kernel reports a fixed cpu frequency in /proc/cpuinfo, this is
confused to user when turbo is enable, and aperf/mperf can be used to
show current cpu frequency after 7d5905dc14a
"(x86 / CPU: Always show current CPU frequency in /proc/cpuinfo)"
so guest should support aperf/mperf capability
This patch implements aperf/mperf by three mode: none, software
emulation, and pass-through
None: default mode, guest does not support aperf/mperf
Software emulation: the period of aperf/mperf in guest mode are
accumulated as emulated value
Pass-though: it is only suitable for KVM_HINTS_REALTIME, Because that
hint guarantees we have a 1:1 vCPU:CPU binding and guaranteed no
over-commit.
And a per-VM capability is added to configure aperfmperf mode
[...]
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c index
cd708b0b460a..c960dda4251b 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -122,6 +122,14 @@ int kvm_update_cpuid(struct kvm_vcpu *vcpu)
MSR_IA32_MISC_ENABLE_MWAIT);
}
+ best = kvm_find_cpuid_entry(vcpu, 6, 0);
+ if (best) {
+ if (guest_has_aperfmperf(vcpu->kvm) &&
+ boot_cpu_has(X86_FEATURE_APERFMPERF))
+ best->ecx |= 1;
+ else
+ best->ecx &= ~1;
+ }
In my understanding, KVM allows userspace to set a CPUID feature bit for
guest even if hardware doesn't support the feature.
So what makes X86_FEATURE_APERFMPERF different here? Is there any
concern I miss?
-Xiaoyao
Whether software emulation for aperf/mperf or pass-through depends on host cpu aperf/mperf feature.
Software emulation: the period of aperf/mperf in guest mode are accumulated as emulated value
I know it that you want to ensure the correctness of exposure of
aperf/mperf.
But there are so many features other than aperf/mperf that KVM reports
the supported settings of them through KVM_GET_SUPPORTED_CPUID, but
doesn't check nor force the correctness of userspace input. i.e., KVM
allows userspace to set bogus CPUID settings as long as it doesn't break
KVM (host kernel).
Indeed, bogus CPUID settings more than likely breaks the guest. But it's
not KVM's fault. KVM just do what userspace wants.
IMO, If we really want to ensure the correctness of userspace provided
CPUID settings, we need to return ERROR to userspace instead of fixing
it siliently.
- Xiaoyao
