Re: [PATCH v4 3/4] x86: nVMX: check vmcs12->guest_ia32_debugctl value given by L2

[Chao Gao <chao.gao@intel.com>]

On Tue, May 20, 2025 at 05:48:44PM -0400, mlevitsk@redhat.com wrote:
>On Fri, 2025-05-16 at 10:50 -0400, mlevitsk@redhat.com wrote:
>> On Fri, 2025-05-16 at 11:31 +0800, Chao Gao wrote:
>> > On Wed, May 14, 2025 at 08:53:52PM -0400, Maxim Levitsky wrote:
>> > > Check the vmcs12 guest_ia32_debugctl value before loading it, to avoid L2
>> > > being able to load arbitrary values to hardware IA32_DEBUGCTL.
>> > > 
>> > > Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
>> > > ---
>> > > arch/x86/kvm/vmx/nested.c | 4 ++++
>> > > arch/x86/kvm/vmx/vmx.c    | 2 +-
>> > > arch/x86/kvm/vmx/vmx.h    | 2 ++
>> > > 3 files changed, 7 insertions(+), 1 deletion(-)
>> > > 
>> > > diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
>> > > index e073e3008b16..0bda6400e30a 100644
>> > > --- a/arch/x86/kvm/vmx/nested.c
>> > > +++ b/arch/x86/kvm/vmx/nested.c
>> > > @@ -3193,6 +3193,10 @@ static int nested_vmx_check_guest_state(struct kvm_vcpu *vcpu,
>> > > 	     CC((vmcs12->guest_bndcfgs & MSR_IA32_BNDCFGS_RSVD))))
>> > > 		return -EINVAL;
>> > > 
>> > > +	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS) &&
>> > > +	     CC(vmcs12->guest_ia32_debugctl & ~vmx_get_supported_debugctl(vcpu, false)))
>> > > +		return -EINVAL;
>> > > +
>> > 
>> > How about grouping this check with the one against DR7 a few lines above?
>> 
>> Good idea, will do.
>
>Besides the above change, is there anything else to change in this patchset?
>If not I'll sent a new version soon.

...

>> > > diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
>> > > index 6d1e40ecc024..1b80479505d3 100644
>> > > --- a/arch/x86/kvm/vmx/vmx.h
>> > > +++ b/arch/x86/kvm/vmx/vmx.h
>> > > @@ -413,7 +413,9 @@ static inline void vmx_set_intercept_for_msr(struct kvm_vcpu *vcpu, u32 msr,
>> > > 		vmx_disable_intercept_for_msr(vcpu, msr, type);
>> > > }
>> > > 
>> > > +
>> > 
>> > stray newline.

Can you remove this newline? (not sure if you've already noticed this)

Also, the shortlogs for patches 3-4 don't follow the convention. They should be
"KVM: nVMX" and "KVM: VMX". With these fixed,

Reviewed-by: Chao Gao <chao.gao@intel.com>