[PATCH 4/8] KVM: VMX: Adjust the TSC-related VMCS fields on L2 entry and exit

From: ilstam@mailbox.org
To: kvm@vger.kernel.org, pbonzini@redhat.com
Cc: ilstam@amazon.com, seanjc@google.com, vkuznets@redhat.com,
	wanpengli@tencent.com, jmattson@google.com, joro@8bytes.org,
	haozhong.zhang@intel.com, zamsden@gmail.com, mtosatti@redhat.com,
	dplotnikov@virtuozzo.com, dwmw@amazon.co.uk
Subject: [PATCH 4/8] KVM: VMX: Adjust the TSC-related VMCS fields on L2 entry and exit
Date: Thu,  6 May 2021 10:32:24 +0000	[thread overview]
Message-ID: <20210506103228.67864-5-ilstam@mailbox.org> (raw)
In-Reply-To: <20210506103228.67864-1-ilstam@mailbox.org>

From: Ilias Stamatis <ilstam@amazon.com>

When L2 is entered we need to "merge" the TSC multiplier and TSC offset
values of VMCS01 and VMCS12 and store the result into the current
VMCS02.

The 02 values are calculated using the following equations:
  offset_02 = ((offset_01 * mult_12) >> 48) + offset_12
  mult_02 = (mult_01 * mult_12) >> 48

Signed-off-by: Ilias Stamatis <ilstam@amazon.com>
---
 arch/x86/include/asm/kvm_host.h |  1 +
 arch/x86/kvm/vmx/nested.c       | 26 ++++++++++++++++++++++----
 arch/x86/kvm/x86.c              | 25 +++++++++++++++++++++++++
 3 files changed, 48 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index cdddbf0b1177..e7a1eb36f95a 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1780,6 +1780,7 @@ void kvm_define_user_return_msr(unsigned index, u32 msr);
 int kvm_set_user_return_msr(unsigned index, u64 val, u64 mask);
 
 u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc, bool l1);
+u64 kvm_compute_02_tsc_offset(u64 l1_offset, u64 l2_multiplier, u64 l2_offset);
 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc);
 
 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu);
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index bced76637823..a1bf28f33837 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -3353,8 +3353,22 @@ enum nvmx_vmentry_status nested_vmx_enter_non_root_mode(struct kvm_vcpu *vcpu,
 	}
 
 	enter_guest_mode(vcpu);
-	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)
-		vcpu->arch.tsc_offset += vmcs12->tsc_offset;
+
+	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING) {
+		if (vmcs12->secondary_vm_exec_control & SECONDARY_EXEC_TSC_SCALING) {
+			vcpu->arch.tsc_offset = kvm_compute_02_tsc_offset(
+					vcpu->arch.l1_tsc_offset,
+					vmcs12->tsc_multiplier,
+					vmcs12->tsc_offset);
+
+			vcpu->arch.tsc_scaling_ratio = mul_u64_u64_shr(
+					vcpu->arch.tsc_scaling_ratio,
+					vmcs12->tsc_multiplier,
+					kvm_tsc_scaling_ratio_frac_bits);
+		} else {
+			vcpu->arch.tsc_offset += vmcs12->tsc_offset;
+		}
+	}
 
 	if (prepare_vmcs02(vcpu, vmcs12, &entry_failure_code)) {
 		exit_reason.basic = EXIT_REASON_INVALID_STATE;
@@ -4454,8 +4468,12 @@ void nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 vm_exit_reason,
 	if (nested_cpu_has_preemption_timer(vmcs12))
 		hrtimer_cancel(&to_vmx(vcpu)->nested.preemption_timer);
 
-	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)
-		vcpu->arch.tsc_offset -= vmcs12->tsc_offset;
+	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING) {
+		vcpu->arch.tsc_offset = vcpu->arch.l1_tsc_offset;
+
+		if (vmcs12->secondary_vm_exec_control & SECONDARY_EXEC_TSC_SCALING)
+			vcpu->arch.tsc_scaling_ratio = vcpu->arch.l1_tsc_scaling_ratio;
+	}
 
 	if (likely(!vmx->fail)) {
 		sync_vmcs02_to_vmcs12(vcpu, vmcs12);
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 26a4c0f46f15..87deb119c521 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2266,6 +2266,31 @@ static u64 kvm_compute_tsc_offset(struct kvm_vcpu *vcpu, u64 target_tsc)
 	return target_tsc - tsc;
 }
 
+/*
+ * This function computes the TSC offset that is stored in VMCS02 when entering
+ * L2 by combining the offset and multiplier values of VMCS01 and VMCS12.
+ */
+u64 kvm_compute_02_tsc_offset(u64 l1_offset, u64 l2_multiplier, u64 l2_offset)
+{
+	u64 offset;
+
+	/*
+	 * The L1 offset is interpreted as a signed number by the CPU and can
+	 * be negative. So we extract the sign before the multiplication and
+	 * put it back afterwards if needed.
+	 */
+	offset = mul_u64_u64_shr(abs((s64) l1_offset),
+				 l2_multiplier,
+				 kvm_tsc_scaling_ratio_frac_bits);
+
+	if ((s64) l1_offset < 0)
+		offset = -((s64) offset);
+
+	offset += l2_offset;
+	return offset;
+}
+EXPORT_SYMBOL_GPL(kvm_compute_02_tsc_offset);
+
 u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc)
 {
 	return vcpu->arch.l1_tsc_offset + kvm_scale_tsc(vcpu, host_tsc, true);
-- 
2.17.1

