Re: [PATCH] kvm: nVMX: VMWRITE checks VMCS-link pointer before VMCS field

[Jim Mattson <jmattson@google.com>]

On Thu, Dec 5, 2019 at 3:46 AM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 04/12/19 22:40, Jim Mattson wrote:
> > According to the SDM, a VMWRITE in VMX non-root operation with an
> > invalid VMCS-link pointer results in VMfailInvalid before the validity
> > of the VMCS field in the secondary source operand is checked.
> >
> > Fixes: 6d894f498f5d1 ("KVM: nVMX: vmread/vmwrite: Use shadow vmcs12 if running L2")
> > Signed-off-by: Jim Mattson <jmattson@google.com>
> > Cc: Liran Alon <liran.alon@oracle.com>
> > ---
> >  arch/x86/kvm/vmx/nested.c | 38 +++++++++++++++++++-------------------
> >  1 file changed, 19 insertions(+), 19 deletions(-)
>
> As Vitaly pointed out, the test must be split in two, like this:

Right. Odd that no kvm-unit-tests noticed.

> ---------------- 8< -----------------------
> From 3b9d87060e800ffae2bd19da94ede05018066c87 Mon Sep 17 00:00:00 2001
> From: Paolo Bonzini <pbonzini@redhat.com>
> Date: Thu, 5 Dec 2019 12:39:07 +0100
> Subject: [PATCH] kvm: nVMX: VMWRITE checks VMCS-link pointer before VMCS field
>
> According to the SDM, a VMWRITE in VMX non-root operation with an
> invalid VMCS-link pointer results in VMfailInvalid before the validity
> of the VMCS field in the secondary source operand is checked.
>
> While cleaning up handle_vmwrite, make the code of handle_vmread look
> the same, too.

Okay.

> Fixes: 6d894f498f5d1 ("KVM: nVMX: vmread/vmwrite: Use shadow vmcs12 if running L2")
> Signed-off-by: Jim Mattson <jmattson@google.com>
> Cc: Liran Alon <liran.alon@oracle.com>
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 4aea7d304beb..c080a879b95d 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -4767,14 +4767,13 @@ static int handle_vmread(struct kvm_vcpu *vcpu)
>         if (to_vmx(vcpu)->nested.current_vmptr == -1ull)
>                 return nested_vmx_failInvalid(vcpu);
>
> -       if (!is_guest_mode(vcpu))
> -               vmcs12 = get_vmcs12(vcpu);
> -       else {
> +       vmcs12 = get_vmcs12(vcpu);
> +       if (is_guest_mode(vcpu)) {
>                 /*
>                  * When vmcs->vmcs_link_pointer is -1ull, any VMREAD
>                  * to shadowed-field sets the ALU flags for VMfailInvalid.
>                  */
> -               if (get_vmcs12(vcpu)->vmcs_link_pointer == -1ull)
> +               if (vmcs12->vmcs_link_pointer == -1ull)
>                         return nested_vmx_failInvalid(vcpu);
>                 vmcs12 = get_shadow_vmcs12(vcpu);
>         }
> @@ -4878,8 +4877,19 @@ static int handle_vmwrite(struct kvm_vcpu *vcpu)
>                 }
>         }
>
> +       vmcs12 = get_vmcs12(vcpu);
> +       if (is_guest_mode(vcpu)) {
> +               /*
> +                * When vmcs->vmcs_link_pointer is -1ull, any VMWRITE
> +                * to shadowed-field sets the ALU flags for VMfailInvalid.
> +                */
> +               if (vmcs12->vmcs_link_pointer == -1ull)
> +                       return nested_vmx_failInvalid(vcpu);
> +               vmcs12 = get_shadow_vmcs12(vcpu);
> +       }
>
>         field = kvm_register_readl(vcpu, (((vmx_instruction_info) >> 28) & 0xf));
> +
>         /*
>          * If the vCPU supports "VMWRITE to any supported field in the
>          * VMCS," then the "read-only" fields are actually read/write.
> @@ -4889,24 +4899,12 @@ static int handle_vmwrite(struct kvm_vcpu *vcpu)
>                 return nested_vmx_failValid(vcpu,
>                         VMXERR_VMWRITE_READ_ONLY_VMCS_COMPONENT);
>
> -       if (!is_guest_mode(vcpu)) {
> -               vmcs12 = get_vmcs12(vcpu);
> -
> -               /*
> -                * Ensure vmcs12 is up-to-date before any VMWRITE that dirties
> -                * vmcs12, else we may crush a field or consume a stale value.
> -                */
> -               if (!is_shadow_field_rw(field))
> -                       copy_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
> -       } else {
> -               /*
> -                * When vmcs->vmcs_link_pointer is -1ull, any VMWRITE
> -                * to shadowed-field sets the ALU flags for VMfailInvalid.
> -                */
> -               if (get_vmcs12(vcpu)->vmcs_link_pointer == -1ull)
> -                       return nested_vmx_failInvalid(vcpu);
> -               vmcs12 = get_shadow_vmcs12(vcpu);
> -       }
> +       /*
> +        * Ensure vmcs12 is up-to-date before any VMWRITE that dirties
> +        * vmcs12, else we may crush a field or consume a stale value.
> +        */
> +       if (!is_guest_mode(vcpu) && !is_shadow_field_rw(field))
> +               copy_vmcs02_to_vmcs12_rare(vcpu, vmcs12);
>
>         offset = vmcs_field_to_offset(field);
>         if (offset < 0)
>
>
> ... and also, do you have a matching kvm-unit-tests patch?

I'll put one together, along with a test that shows the current
priority inversion between read-only and unsupported VMCS fields.

> Thanks,
>
> Paolo
>