Re: [PATCH 1/3 v2] nSVM: Check addresses of MSR and IO bitmap

[Sean Christopherson <seanjc@google.com>]

On Sat, Jan 16, 2021, Krish Sadhukhan wrote:
> According to section "Canonicalization and Consistency Checks" in APM vol 2,
> the following guest state is illegal:
> 
>         "The MSR or IOIO intercept tables extend to a physical address that
>          is greater than or equal to the maximum supported physical address."
> 
> Also check that these addresses are aligned on page boundary.
> 
> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> ---
>  arch/x86/kvm/svm/nested.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
> index cb4c6ee10029..2419f392a13d 100644
> --- a/arch/x86/kvm/svm/nested.c
> +++ b/arch/x86/kvm/svm/nested.c
> @@ -211,7 +211,8 @@ static bool svm_get_nested_state_pages(struct kvm_vcpu *vcpu)
>  	return true;
>  }
>  
> -static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
> +static bool nested_vmcb_check_controls(struct kvm_vcpu *vcpu,
> +				       struct vmcb_control_area *control)
>  {
>  	if ((vmcb_is_intercept(control, INTERCEPT_VMRUN)) == 0)
>  		return false;
> @@ -223,10 +224,15 @@ static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
>  	    !npt_enabled)
>  		return false;
>  
> +	if (!page_address_valid(vcpu, control->msrpm_base_pa))
> +		return false;
> +	if (!page_address_valid(vcpu, control->iopm_base_pa))

These checks are wrong.  The MSRPM is 8kb in size, and the IOPM is 12kb, and the
APM explicitly states that the last byte is checked:

  if the address of the last byte in the table is greater than or equal to the
  maximum supported physical address, this is treated as illegal VMCB state and
  causes a #VMEXIT(VMEXIT_INVALID).

KVM can't check just the last byte, as that would fail to detect a wrap of the
64-bit boundary.  Might be worth adding yet another helper?  I think this will
work, though I'm sure Paolo has a much more clever solution :-)

  static inline bool page_range_valid(struct kvm_vcpu *vcpu, gpa_t gpa, int size)
  {
	gpa_t last_page = gpa + size - PAGE_SIZE;

	if (last_page < gpa)
		return false;

	return page_address_valid(last_page);
  }

Note, the IOPM is 12kb in size, but KVM allocates and initializes 16kb, i.e.
using IOPM_ALLOC_ORDER for the check would be wrong.  Maybe define the actual
size for both bitmaps and use get_order() instead of hardcoding the order?  That
would make it easy to "fix" svm_hardware_setup() so that it doesn't initialize
unused memory.

> +		return false;
> +
>  	return true;
>  }
>  
> -static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
> +static bool nested_vmcb_checks(struct kvm_vcpu *vcpu, struct vmcb *vmcb12)
>  {
>  	bool vmcb12_lma;
>  
> @@ -255,10 +261,10 @@ static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
>  		    (vmcb12->save.cr3 & MSR_CR3_LONG_MBZ_MASK))
>  			return false;
>  	}
> -	if (!kvm_is_valid_cr4(&svm->vcpu, vmcb12->save.cr4))
> +	if (!kvm_is_valid_cr4(vcpu, vmcb12->save.cr4))
>  		return false;
>  
> -	return nested_vmcb_check_controls(&vmcb12->control);
> +	return nested_vmcb_check_controls(vcpu, &vmcb12->control);
>  }
>  
>  static void load_nested_vmcb_control(struct vcpu_svm *svm,
> @@ -485,7 +491,7 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
>  	if (WARN_ON_ONCE(!svm->nested.initialized))
>  		return -EINVAL;
>  
> -	if (!nested_vmcb_checks(svm, vmcb12)) {
> +	if (!nested_vmcb_checks(&svm->vcpu, vmcb12)) {
>  		vmcb12->control.exit_code    = SVM_EXIT_ERR;
>  		vmcb12->control.exit_code_hi = 0;
>  		vmcb12->control.exit_info_1  = 0;
> @@ -1173,7 +1179,7 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
>  		goto out_free;
>  
>  	ret = -EINVAL;
> -	if (!nested_vmcb_check_controls(ctl))
> +	if (!nested_vmcb_check_controls(vcpu, ctl))
>  		goto out_free;
>  
>  	/*
> -- 
> 2.27.0
>