Patchwork [v2] KVM nVMX: MSRs should not be stored if VM-entry fails during or after loading guest state

login
register
mail settings
Submitter Krish Sadhukhan
Date Dec. 4, 2018, 7:03 p.m.
Message ID <20181204190306.9928-2-krish.sadhukhan@oracle.com>
Download mbox | patch
Permalink /patch/672329/
State New
Headers show

Comments

Krish Sadhukhan - Dec. 4, 2018, 7:03 p.m.
According to section "VM-entry Failures During or After Loading Guest State"
in Intel SDM vol 3C,

	"No MSRs are saved into the VM-exit MSR-store area."

when bit 31 of the exit reason is set.

Reported-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Suggested-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
---
 arch/x86/kvm/vmx.c | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)
Jim Mattson - Dec. 4, 2018, 7:35 p.m.
On Tue, Dec 4, 2018 at 11:27 AM Krish Sadhukhan
<krish.sadhukhan@oracle.com> wrote:
>
> According to section "VM-entry Failures During or After Loading Guest State"
> in Intel SDM vol 3C,
>
>         "No MSRs are saved into the VM-exit MSR-store area."
>
> when bit 31 of the exit reason is set.
>
> Reported-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> Suggested-by: Jim Mattson <jmattson@google.com>
> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
> ---
>  arch/x86/kvm/vmx.c | 17 +++++++++++++----
>  1 file changed, 13 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 4555077..9e1a640 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -13827,6 +13827,19 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
>                  * L2 to IDT_VECTORING_INFO_FIELD.
>                  */
>                 vmcs12_save_pending_event(vcpu, vmcs12);
> +
> +               /*
> +                * There's no need to store the guest's MSRs if we exit
> +                * due to a VM-entry failure that occurs during or after
> +                * loading the guest state, because the guest has not
> +                * executed any instruction in that case. However, here we
> +                * are past that point. So we need to save the MSRs.
> +                */

Executing instructions is a distraction. The specification indicates
that these MSRs are stored even in cases where no guest instruction
has executed, so long as the VM-exit was not due to a VM-entry failure
during or after loading guest state. There's no need to store the
guest's MSRs if the exit was due to a VM-entry failure during or after
loading guest state (and, in fact, it is incorrect to do so), because
the specification indicates that these MSRs are not stored in that
case.

> +               if (nested_vmx_store_msr(vcpu,
> +                                        vmcs12->vm_exit_msr_store_addr,
> +                                        vmcs12->vm_exit_msr_store_count))
> +                       nested_vmx_abort(vcpu,
> +                                        VMX_ABORT_SAVE_GUEST_MSR_FAIL);
>         }
>
>         /*
> @@ -14159,10 +14172,6 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 exit_reason,
>                  * immutable.
>                  */
>                 nested_flush_cached_shadow_vmcs12(vcpu, vmcs12);
> -
> -               if (nested_vmx_store_msr(vcpu, vmcs12->vm_exit_msr_store_addr,
> -                                        vmcs12->vm_exit_msr_store_count))
> -                       nested_vmx_abort(vcpu, VMX_ABORT_SAVE_GUEST_MSR_FAIL);
>         } else {
>                 /*
>                  * The only expected VM-instruction error is "VM entry with
> --
> 2.9.5
>

Patch

diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 4555077..9e1a640 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -13827,6 +13827,19 @@  static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
 		 * L2 to IDT_VECTORING_INFO_FIELD.
 		 */
 		vmcs12_save_pending_event(vcpu, vmcs12);
+
+		/*
+		 * There's no need to store the guest's MSRs if we exit
+		 * due to a VM-entry failure that occurs during or after
+		 * loading the guest state, because the guest has not
+		 * executed any instruction in that case. However, here we
+		 * are past that point. So we need to save the MSRs.
+		 */
+		if (nested_vmx_store_msr(vcpu,
+					 vmcs12->vm_exit_msr_store_addr,
+					 vmcs12->vm_exit_msr_store_count))
+			nested_vmx_abort(vcpu,
+					 VMX_ABORT_SAVE_GUEST_MSR_FAIL);
 	}
 
 	/*
@@ -14159,10 +14172,6 @@  static void nested_vmx_vmexit(struct kvm_vcpu *vcpu, u32 exit_reason,
 		 * immutable.
 		 */
 		nested_flush_cached_shadow_vmcs12(vcpu, vmcs12);
-
-		if (nested_vmx_store_msr(vcpu, vmcs12->vm_exit_msr_store_addr,
-					 vmcs12->vm_exit_msr_store_count))
-			nested_vmx_abort(vcpu, VMX_ABORT_SAVE_GUEST_MSR_FAIL);
 	} else {
 		/*
 		 * The only expected VM-instruction error is "VM entry with