[PATCH] KVM: nVMX: Store vmcs.EXIT_QUALIFICATION as an unsigned long, not u32

[PATCH] KVM: nVMX: Store vmcs.EXIT_QUALIFICATION as an unsigned long, not u32

[Sean Christopherson]

Use an unsigned long for 'exit_qual' in nested_vmx_reflect_vmexit(), the
EXIT_QUALIFICATION field is naturally sized, not a 32-bit field.

The bug is most easily observed by doing VMXON (or any VMX instruction)
in L2 with a negative displacement, in which case dropping the upper
bits on nested VM-Exit results in L1 calculating the wrong virtual
address for the memory operand, e.g. "vmxon -0x8(%rbp)" yields:

  Unhandled cpu exception 14 #PF at ip 0000000000400553
  rbp=0000000000537000 cr2=0000000100536ff8

Fixes: fbdd50250396d ("KVM: nVMX: Move VM-Fail check out of nested_vmx_exit_reflected()")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
---

Sadly (for me), I can't blame a mishandled merge on this one.  Even more
embarassing is that this is actually the second instance where I botched
the size for exit_qual, you'd think I'd have double-checked everything
after the first one...

 arch/x86/kvm/vmx/nested.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index f228339cd0a0..3f32f81f5c59 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -5814,7 +5814,8 @@ bool nested_vmx_reflect_vmexit(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	u32 exit_reason = vmx->exit_reason;
-	u32 exit_intr_info, exit_qual;
+	unsigned long exit_qual;
+	u32 exit_intr_info;
 
 	WARN_ON_ONCE(vmx->nested.nested_run_pending);
 
-- 
2.26.0

Re: [PATCH] KVM: nVMX: Store vmcs.EXIT_QUALIFICATION as an unsigned long, not u32

[Vitaly Kuznetsov]

Sean Christopherson <sean.j.christopherson@intel.com> writes:

> Use an unsigned long for 'exit_qual' in nested_vmx_reflect_vmexit(), the
> EXIT_QUALIFICATION field is naturally sized, not a 32-bit field.
>
> The bug is most easily observed by doing VMXON (or any VMX instruction)
> in L2 with a negative displacement, in which case dropping the upper
> bits on nested VM-Exit results in L1 calculating the wrong virtual
> address for the memory operand, e.g. "vmxon -0x8(%rbp)" yields:
>
>   Unhandled cpu exception 14 #PF at ip 0000000000400553
>   rbp=0000000000537000 cr2=0000000100536ff8
>
> Fixes: fbdd50250396d ("KVM: nVMX: Move VM-Fail check out of nested_vmx_exit_reflected()")
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> ---
>
> Sadly (for me), I can't blame a mishandled merge on this one.  Even more
> embarassing is that this is actually the second instance where I botched
> the size for exit_qual, you'd think I'd have double-checked everything
> after the first one...
>
>  arch/x86/kvm/vmx/nested.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index f228339cd0a0..3f32f81f5c59 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -5814,7 +5814,8 @@ bool nested_vmx_reflect_vmexit(struct kvm_vcpu *vcpu)
>  {
>  	struct vcpu_vmx *vmx = to_vmx(vcpu);
>  	u32 exit_reason = vmx->exit_reason;
> -	u32 exit_intr_info, exit_qual;
> +	unsigned long exit_qual;
> +	u32 exit_intr_info;
>  
>  	WARN_ON_ONCE(vmx->nested.nested_run_pending);

Too late but

Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>

I also did 'git grep -W 'u32.*exit_qual' kvm/queue' and I can see a few
more places where 'exit_qual' is u32:
nested_vmx_check_guest_state()
nested_vmx_enter_non_root_mode()
vmx_set_nested_state()

Being too lazy to check an even if there are no immediate issues with
that, should we just use 'unsigned long' everywhere?

-- 
Vitaly

Re: [PATCH] KVM: nVMX: Store vmcs.EXIT_QUALIFICATION as an unsigned long, not u32

[Sean Christopherson]

On Fri, Apr 24, 2020 at 01:44:00PM +0200, Vitaly Kuznetsov wrote:
> I also did 'git grep -W 'u32.*exit_qual' kvm/queue' and I can see a few
> more places where 'exit_qual' is u32:
> nested_vmx_check_guest_state()
> nested_vmx_enter_non_root_mode()
> vmx_set_nested_state()
> 
> Being too lazy to check an even if there are no immediate issues with
> that, should we just use 'unsigned long' everywhere?

Yes, absolutely, I'll send a patch.

The existing cases are benign, they're all related to setting the exit_qual
for a nested VM-Enter failure, which could fit in a u8.  But still worth
fixing.