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

From: Paolo Bonzini <pbonzini@redhat.com>
To: ilstam@mailbox.org, kvm@vger.kernel.org
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: Re: [PATCH 4/8] KVM: VMX: Adjust the TSC-related VMCS fields on L2 entry and exit
Date: Thu, 6 May 2021 13:32:57 +0200	[thread overview]
Message-ID: <50f86951-1cea-b7aa-7236-f28edd5eca8d@redhat.com> (raw)
In-Reply-To: <20210506103228.67864-5-ilstam@mailbox.org>

On 06/05/21 12:32, ilstam@mailbox.org wrote:
> +	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;
> +		}

The computation of vcpu->arch.tsc_offset is (not coincidentially) the
same that appears in patch 6

+	    (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)) {
+		if (vmcs12->secondary_vm_exec_control & SECONDARY_EXEC_TSC_SCALING) {
+			cur_offset = kvm_compute_02_tsc_offset(
+					l1_offset,
+					vmcs12->tsc_multiplier,
+					vmcs12->tsc_offset);
+		} else {
+			cur_offset = l1_offset + vmcs12->tsc_offset;

So I think you should just pass vmcs12 and the L1 offset to
kvm_compute_02_tsc_offset, and let it handle both cases (and possibly
even set vcpu->arch.tsc_scaling_ratio in the same function).

Paolo