2015-08-07 03:28:46

by Haozhong Zhang

Subject: [PATCH] KVM: x86: Use adjustment in guest cycles when handling MSR_IA32_TSC_ADJUST

When kvm_set_msr_common() handles a guest's write to
MSR_IA32_TSC_ADJUST, it will calcuate an adjustment based on the data
written by guest and then use it to adjust TSC offset by calling a
call-back adjust_tsc_offset(). The 3rd parameter of adjust_tsc_offset()
indicates whether the adjustment is in host TSC cycles or in guest TSC
cycles. If SVM TSC scaling is enabled, adjust_tsc_offset()
[i.e. svm_adjust_tsc_offset()] will first scale the adjustment;
otherwise, it will just use the unscaled one. As the MSR write here
comes from the guest, the adjustment is in guest TSC cycles. However,
the current kvm_set_msr_common() uses it as a value in host TSC
cycles (by using true as the 3rd parameter of adjust_tsc_offset()),
which can result in an incorrect adjustment of TSC offset if SVM TSC
scaling is enabled. This patch fixes this problem.

Signed-off-by: Haozhong Zhang <[email protected]>
---
arch/x86/kvm/x86.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 5ef2560..3cd6cd2 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2105,7 +2105,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
if (guest_cpuid_has_tsc_adjust(vcpu)) {
if (!msr_info->host_initiated) {
s64 adj = data - vcpu->arch.ia32_tsc_adjust_msr;
- kvm_x86_ops->adjust_tsc_offset(vcpu, adj, true);
+ adjust_tsc_offset_guest(vcpu, adj);
}
vcpu->arch.ia32_tsc_adjust_msr = data;
}
--
1.7.9.5

2015-08-07 11:27:59

by Paolo Bonzini

Subject: Re: [PATCH] KVM: x86: Use adjustment in guest cycles when handling MSR_IA32_TSC_ADJUST

On 07/08/2015 05:24, Haozhong Zhang wrote:
> When kvm_set_msr_common() handles a guest's write to
> MSR_IA32_TSC_ADJUST, it will calcuate an adjustment based on the data
> written by guest and then use it to adjust TSC offset by calling a
> call-back adjust_tsc_offset(). The 3rd parameter of adjust_tsc_offset()
> indicates whether the adjustment is in host TSC cycles or in guest TSC
> cycles. If SVM TSC scaling is enabled, adjust_tsc_offset()
> [i.e. svm_adjust_tsc_offset()] will first scale the adjustment;
> otherwise, it will just use the unscaled one. As the MSR write here
> comes from the guest, the adjustment is in guest TSC cycles. However,
> the current kvm_set_msr_common() uses it as a value in host TSC
> cycles (by using true as the 3rd parameter of adjust_tsc_offset()),
> which can result in an incorrect adjustment of TSC offset if SVM TSC
> scaling is enabled. This patch fixes this problem.
>
> Signed-off-by: Haozhong Zhang <[email protected]>
> ---
> arch/x86/kvm/x86.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 5ef2560..3cd6cd2 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -2105,7 +2105,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> if (guest_cpuid_has_tsc_adjust(vcpu)) {
> if (!msr_info->host_initiated) {
> s64 adj = data - vcpu->arch.ia32_tsc_adjust_msr;
> - kvm_x86_ops->adjust_tsc_offset(vcpu, adj, true);
> + adjust_tsc_offset_guest(vcpu, adj);
> }
> vcpu->arch.ia32_tsc_adjust_msr = data;
> }
> --
> 1.7.9.5
>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>

Thanks, applied!

Paolo