Re: [PATCH v4] X86/KVM: Properly update 'tsc_offset' to represent the running guest

[Paolo Bonzini <pbonzini@redhat.com>]

On 14/04/2018 05:10, KarimAllah Ahmed wrote:
> Update 'tsc_offset' on vmentry/vmexit of L2 guests to ensure that it always
> captures the TSC_OFFSET of the running guest whether it is the L1 or L2
> guest.
> 
> Cc: Jim Mattson <jmattson@google.com>
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> Cc: kvm@vger.kernel.org
> Cc: linux-kernel@vger.kernel.org
> Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
> Signed-off-by: KarimAllah Ahmed <karahmed@amazon.de>
> [AMD changes, fix update_ia32_tsc_adjust_msr. - Paolo]
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

I think this makes it a little bit more tidy, and also matches svm.c's
placement of the increment in enter_svm_guest_mode:

diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 67af4b8c9767..7207e6cc07c1 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -11435,6 +11435,7 @@ static int enter_vmx_non_root_mode(struct kvm_vcpu *vcpu, bool from_vmentry)
 	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
 	u32 msr_entry_idx;
 	u32 exit_qual;
+	int r;
 
 	enter_guest_mode(vcpu);
 
@@ -11444,30 +11445,21 @@ static int enter_vmx_non_root_mode(struct kvm_vcpu *vcpu, bool from_vmentry)
 	vmx_switch_vmcs(vcpu, &vmx->nested.vmcs02);
 	vmx_segment_cache_clear(vmx);
 
-	if (prepare_vmcs02(vcpu, vmcs12, from_vmentry, &exit_qual)) {
-		if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING)
-			vcpu->arch.tsc_offset -= vmcs12->tsc_offset;
-		leave_guest_mode(vcpu);
-		vmx_switch_vmcs(vcpu, &vmx->vmcs01);
-		nested_vmx_entry_failure(vcpu, vmcs12,
-					 EXIT_REASON_INVALID_STATE, exit_qual);
-		return 1;
-	}
+	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING)
+		vcpu->arch.tsc_offset += vmcs12->tsc_offset;
+
+	r = EXIT_REASON_INVALID_STATE;
+	if (prepare_vmcs02(vcpu, vmcs12, from_vmentry, &exit_qual))
+		goto fail;
 
 	nested_get_vmcs12_pages(vcpu, vmcs12);
 
+	r = EXIT_REASON_MSR_LOAD_FAIL;
 	msr_entry_idx = nested_vmx_load_msr(vcpu,
 					    vmcs12->vm_entry_msr_load_addr,
 					    vmcs12->vm_entry_msr_load_count);
-	if (msr_entry_idx) {
-		if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING)
-			vcpu->arch.tsc_offset -= vmcs12->tsc_offset;
-		leave_guest_mode(vcpu);
-		vmx_switch_vmcs(vcpu, &vmx->vmcs01);
-		nested_vmx_entry_failure(vcpu, vmcs12,
-				EXIT_REASON_MSR_LOAD_FAIL, msr_entry_idx);
-		return 1;
-	}
+	if (msr_entry_idx)
+		goto fail;
 
 	/*
 	 * Note no nested_vmx_succeed or nested_vmx_fail here. At this point
@@ -11476,6 +11468,14 @@ static int enter_vmx_non_root_mode(struct kvm_vcpu *vcpu, bool from_vmentry)
 	 * the success flag) when L2 exits (see nested_vmx_vmexit()).
 	 */
 	return 0;
+
+fail:
+	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING)
+		vcpu->arch.tsc_offset -= vmcs12->tsc_offset;
+	leave_guest_mode(vcpu);
+	vmx_switch_vmcs(vcpu, &vmx->vmcs01);
+	nested_vmx_entry_failure(vcpu, vmcs12, r, exit_qual);
+	return 1;
 }
 
 /*
@@ -11546,9 +11546,6 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
 		return 1;
 	}
 
-	if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING)
-		vcpu->arch.tsc_offset += vmcs12->tsc_offset;
-
 	/*
 	 * We're finally done with prerequisite checking, and can start with
 	 * the nested entry.

The test case I've sent passes with the above version and fails with v3 indeed.
Thanks Jim. :)

Paolo