Re: [PATCH 3/7 v7] KVM: nSVM: No need to set bits 11:0 in MSRPM and IOPM bitmaps

[Sean Christopherson <seanjc@google.com>]

On Mon, Apr 12, 2021, Krish Sadhukhan wrote:
> According to APM vol 2, hardware ignores the low 12 bits in MSRPM and IOPM
> bitmaps. Therefore setting/unssetting these bits has no effect as far as
> VMRUN is concerned. Also, setting/unsetting these bits prevents tests from
> verifying hardware behavior.
> 
> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> ---
>  arch/x86/kvm/svm/nested.c | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
> index ae53ae46ebca..fd42c8b7f99a 100644
> --- a/arch/x86/kvm/svm/nested.c
> +++ b/arch/x86/kvm/svm/nested.c
> @@ -287,8 +287,6 @@ static void nested_load_control_from_vmcb12(struct vcpu_svm *svm,
>  
>  	/* Copy it here because nested_svm_check_controls will check it.  */
>  	svm->nested.ctl.asid           = control->asid;
> -	svm->nested.ctl.msrpm_base_pa &= ~0x0fffULL;
> -	svm->nested.ctl.iopm_base_pa  &= ~0x0fffULL;

This will break nested_svm_vmrun_msrpm() if L1 passes an unaligned address.
The shortlog is also wrong, KVM isn't setting bits, it's clearing bits.

I also don't think svm->nested.ctl.msrpm_base_pa makes its way to hardware; IIUC,
it's a copy of vmcs12->control.msrpm_base_pa.  The bitmap that gets loaded into
the "real" VMCB is vmcb02->control.msrpm_base_pa.

>  }
>  
>  /*
> -- 
> 2.27.0
>