Re: [PATCH 1/2] KVM: nVMX: fix msr bitmaps to prevent L2 from accessing L0 x2APIC

[Radim KrÄmÃÅ]

2016-08-09 17:32+0800, Yang Zhang:
> On 2016/8/9 2:16, Radim KrÄmÃÅ wrote:
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> @@ -6995,16 +6982,21 @@ static int handle_vmon(struct kvm_vcpu *vcpu)
>>  		return 1;
>>  	}
>> 
>> +	if (cpu_has_vmx_msr_bitmap()) {
>> +		vmx->nested.msr_bitmap =
>> +				(unsigned long *)__get_free_page(GFP_KERNEL);
>> +		if (!vmx->nested.msr_bitmap)
>> +			goto out_msr_bitmap;
>> +	}
>> +
> 
> We export msr_bitmap to guest even it is not supported by hardware. So we
> need to allocate msr_bitmap for L1 unconditionally.

We do emulate the feature, but the vmx->nested.msr_bitmap is used only
if VMX supports it to avoid some VM exits:

>> @@ -9957,10 +9938,10 @@ static void prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
>>  	}
>> 
>>  	if (cpu_has_vmx_msr_bitmap() &&
>> -	    exec_control & CPU_BASED_USE_MSR_BITMAPS) {
>> -		nested_vmx_merge_msr_bitmap(vcpu, vmcs12);
>> -		/* MSR_BITMAP will be set by following vmx_set_efer. */
>> -	} else
>> +	    exec_control & CPU_BASED_USE_MSR_BITMAPS &&
>> +	    nested_vmx_merge_msr_bitmap(vcpu, vmcs12))
>> +		; /* MSR_BITMAP will be set by following vmx_set_efer. */
>> +	else
>>  		exec_control &= ~CPU_BASED_USE_MSR_BITMAPS;

The else branch is taken if !cpu_has_vmx_msr_bitmap() and disables msr
bitmaps.  Similar check for vmx_set_msr_bitmap(), so the NULL doesn't
even get written to VMCS.

KVM always uses L1's msr bitmaps when emulating the feature.