Re: [PATCH v2 4/5] mmu: change unconditional setting of the u bit in fault bitmap

[Paolo Bonzini <pbonzini@redhat.com>]

On 13/07/2016 00:18, Bandan Das wrote:
> For the nested EPT case, we assume that the read bit (u) is
> always set since we used to unconditionally set it in set_spte().
> Modify it to only be set when host ept execute only support
> isn't present.
> 
> Signed-off-by: Bandan Das <bsd@redhat.com>
> ---
>  arch/x86/kvm/mmu.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index c364dcb..566eea5 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -3923,9 +3923,10 @@ static void update_permission_bitmask(struct kvm_vcpu *vcpu,
>  				 *   clearer.
>  				 */
>  				smap = cr4_smap && u && !uf && !ff;
> -			} else
> -				/* Not really needed: no U/S accesses on ept  */
> -				u = 1;
> +			} else {
> +				if (shadow_present_mask)
> +					u = 1;
> +			}
>  
>  			fault = (ff && !x) || (uf && !u) || (wf && !w) ||
>  				(smapf && smap);
> 

This is needed too, in order to look up with the correct uf:

diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 003618e324ce..941d345ebd2c 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -6014,12 +6014,14 @@ static int handle_ept_violation(struct kvm_vcpu
*vcpu)
 	gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS);
 	trace_kvm_page_fault(gpa, exit_qualification);

-	/* It is a write fault? */
+	/* it is a read fault? */
+	error_code = (exit_qualification << 2) & PFERR_USER_MASK;
+	/* it is a write fault? */
 	error_code = exit_qualification & PFERR_WRITE_MASK;
 	/* It is a fetch fault? */
 	error_code |= (exit_qualification << 2) & PFERR_FETCH_MASK;
 	/* ept page table is present? */
-	error_code |= (exit_qualification >> 3) & PFERR_PRESENT_MASK;
+	error_code |= (exit_qualification & 0x38) != 0;

 	vcpu->arch.exit_qualification = exit_qualification;


Otherwise you always lookup the permission bitmap for an uf=0 case.
This was the meaning of the "Not really needed" comment.

Paolo