2020-08-12 18:10:06

by Sean Christopherson

Subject: [PATCH] KVM: nVMX: Add VM-Enter failed tracepoints for super early checks

Add tracepoints for the early consistency checks in nested_vmx_run().
The "VMLAUNCH vs. VMRESUME" check in particular is useful to trace, as
there is no architectural way to check VMCS.LAUNCH_STATE, and subtle
bugs such as VMCLEAR on the wrong HPA can lead to confusing errors in
the L1 VMM.

Signed-off-by: Sean Christopherson <[email protected]>
---
arch/x86/kvm/vmx/nested.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index 23b58c28a1c92..fb37f0972e78a 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -3468,11 +3468,11 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
if (evmptrld_status == EVMPTRLD_ERROR) {
kvm_queue_exception(vcpu, UD_VECTOR);
return 1;
- } else if (evmptrld_status == EVMPTRLD_VMFAIL) {
+ } else if (CC(evmptrld_status == EVMPTRLD_VMFAIL)) {
return nested_vmx_failInvalid(vcpu);
}

- if (!vmx->nested.hv_evmcs && vmx->nested.current_vmptr == -1ull)
+ if (CC(!vmx->nested.hv_evmcs && vmx->nested.current_vmptr == -1ull))
return nested_vmx_failInvalid(vcpu);

vmcs12 = get_vmcs12(vcpu);
@@ -3483,7 +3483,7 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
* rather than RFLAGS.ZF, and no error number is stored to the
* VM-instruction error field.
*/
- if (vmcs12->hdr.shadow_vmcs)
+ if (CC(vmcs12->hdr.shadow_vmcs))
return nested_vmx_failInvalid(vcpu);

if (vmx->nested.hv_evmcs) {
@@ -3504,10 +3504,10 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
* for misconfigurations which will anyway be caught by the processor
* when using the merged vmcs02.
*/
- if (interrupt_shadow & KVM_X86_SHADOW_INT_MOV_SS)
+ if (CC(interrupt_shadow & KVM_X86_SHADOW_INT_MOV_SS))
return nested_vmx_fail(vcpu, VMXERR_ENTRY_EVENTS_BLOCKED_BY_MOV_SS);

- if (vmcs12->launch_state == launch)
+ if (CC(vmcs12->launch_state == launch))
return nested_vmx_fail(vcpu,
launch ? VMXERR_VMLAUNCH_NONCLEAR_VMCS
: VMXERR_VMRESUME_NONLAUNCHED_VMCS);
--
2.28.0

2020-09-02 22:16:36

by Sean Christopherson

Subject: Re: [PATCH] KVM: nVMX: Add VM-Enter failed tracepoints for super early checks

On Tue, Sep 01, 2020 at 10:21:15AM +0200, Vitaly Kuznetsov wrote:
> Sean Christopherson <[email protected]> writes:
>
> > Add tracepoints for the early consistency checks in nested_vmx_run().
> > The "VMLAUNCH vs. VMRESUME" check in particular is useful to trace, as
> > there is no architectural way to check VMCS.LAUNCH_STATE, and subtle
> > bugs such as VMCLEAR on the wrong HPA can lead to confusing errors in
> > the L1 VMM.
> >
> > Signed-off-by: Sean Christopherson <[email protected]>
> > ---
> > arch/x86/kvm/vmx/nested.c | 10 +++++-----
> > 1 file changed, 5 insertions(+), 5 deletions(-)
> >
> > diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> > index 23b58c28a1c92..fb37f0972e78a 100644
> > --- a/arch/x86/kvm/vmx/nested.c
> > +++ b/arch/x86/kvm/vmx/nested.c
> > @@ -3468,11 +3468,11 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
> > if (evmptrld_status == EVMPTRLD_ERROR) {
>
> Would it make sense to add 'CC' here too for, em, consistency? :-) #UD
> is probably easy to spot anyway..

I'd prefer not to, purely because it's a #UD and not a VM-Fail.

2020-09-23 16:07:31

by Paolo Bonzini

Subject: Re: [PATCH] KVM: nVMX: Add VM-Enter failed tracepoints for super early checks

On 12/08/20 20:06, Sean Christopherson wrote:
> Add tracepoints for the early consistency checks in nested_vmx_run().
> The "VMLAUNCH vs. VMRESUME" check in particular is useful to trace, as
> there is no architectural way to check VMCS.LAUNCH_STATE, and subtle
> bugs such as VMCLEAR on the wrong HPA can lead to confusing errors in
> the L1 VMM.
>
> Signed-off-by: Sean Christopherson <[email protected]>
> ---
> arch/x86/kvm/vmx/nested.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 23b58c28a1c92..fb37f0972e78a 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -3468,11 +3468,11 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
> if (evmptrld_status == EVMPTRLD_ERROR) {
> kvm_queue_exception(vcpu, UD_VECTOR);
> return 1;
> - } else if (evmptrld_status == EVMPTRLD_VMFAIL) {
> + } else if (CC(evmptrld_status == EVMPTRLD_VMFAIL)) {
> return nested_vmx_failInvalid(vcpu);
> }
>
> - if (!vmx->nested.hv_evmcs && vmx->nested.current_vmptr == -1ull)
> + if (CC(!vmx->nested.hv_evmcs && vmx->nested.current_vmptr == -1ull))
> return nested_vmx_failInvalid(vcpu);
>
> vmcs12 = get_vmcs12(vcpu);
> @@ -3483,7 +3483,7 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
> * rather than RFLAGS.ZF, and no error number is stored to the
> * VM-instruction error field.
> */
> - if (vmcs12->hdr.shadow_vmcs)
> + if (CC(vmcs12->hdr.shadow_vmcs))
> return nested_vmx_failInvalid(vcpu);
>
> if (vmx->nested.hv_evmcs) {
> @@ -3504,10 +3504,10 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
> * for misconfigurations which will anyway be caught by the processor
> * when using the merged vmcs02.
> */
> - if (interrupt_shadow & KVM_X86_SHADOW_INT_MOV_SS)
> + if (CC(interrupt_shadow & KVM_X86_SHADOW_INT_MOV_SS))
> return nested_vmx_fail(vcpu, VMXERR_ENTRY_EVENTS_BLOCKED_BY_MOV_SS);
>
> - if (vmcs12->launch_state == launch)
> + if (CC(vmcs12->launch_state == launch))
> return nested_vmx_fail(vcpu,
> launch ? VMXERR_VMLAUNCH_NONCLEAR_VMCS
> : VMXERR_VMRESUME_NONLAUNCHED_VMCS);
>

Queued, thanks.

Paolo