diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index bd2b71d4e64f..2da75bbf7f94 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2626,12 +2626,15 @@ static inline bool kvm_check_tsc_unstable(void)
  * participates in.
  */
 static void __kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 offset, u64 tsc,
-				  u64 ns, bool matched)
+				  u64 ns, bool matched, bool user_set_tsc)
 {
 	struct kvm *kvm = vcpu->kvm;
 
 	lockdep_assert_held(&kvm->arch.tsc_write_lock);
 
+	if (user_set_tsc)
+		vcpu->kvm->arch.user_set_tsc = true;
+
 	/*
 	 * We also track th most recent recorded KHZ, write and time to
 	 * allow the matching interval to be extended at each write.
@@ -2717,8 +2720,6 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 *user_value)
 		}
 	}
 
-	if (user_value)
-		kvm->arch.user_set_tsc = true;
 
 	/*
 	 * For a reliable TSC, we can match TSC offsets, and for an unstable
@@ -2738,7 +2739,7 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 *user_value)
 		matched = true;
 	}
 
-	__kvm_synchronize_tsc(vcpu, offset, data, ns, matched);
+	__kvm_synchronize_tsc(vcpu, offset, data, ns, matched, !!user_value);
 	raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags);
 }
 
@@ -5725,8 +5726,7 @@ static int kvm_arch_tsc_set_attr(struct kvm_vcpu *vcpu,
 		tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio) + offset;
 		ns = get_kvmclock_base_ns();
 
-		kvm->arch.user_set_tsc = true;
-		__kvm_synchronize_tsc(vcpu, offset, tsc, ns, matched);
+		__kvm_synchronize_tsc(vcpu, offset, tsc, ns, matched, true);
 		raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags);
 
 		r = 0;