DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 2170A3F72F
Subject: Re: [PATCH v5 3/3] KVM: nVMX: Emulate EPTP switching for the L1
 hypervisor
To: Bandan Das <bsd@redhat.com>, kvm@vger.kernel.org
Cc: pbonzini@redhat.com, rkrcmar@redhat.com, jmattson@google.com,
        linux-kernel@vger.kernel.org
References: <20170728195245.1018-1-bsd@redhat.com>
 <20170728195245.1018-4-bsd@redhat.com>
From: David Hildenbrand <david@redhat.com>
Organization: Red Hat GmbH
Message-ID: <c3e50b37-0dde-3140-9a38-e4f15e13a572@redhat.com>
Date: Mon, 31 Jul 2017 13:59:11 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.1
MIME-Version: 1.0
In-Reply-To: <20170728195245.1018-4-bsd@redhat.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 5162
Lines: 179


> +static inline bool nested_cpu_has_eptp_switching(struct vmcs12 *vmcs12)
> +{
> +	return nested_cpu_has_vmfunc(vmcs12) &&
> +		(vmcs12->vm_function_control &
> +		 VMX_VMFUNC_EPTP_SWITCHING);
> +}
> +
>  static inline bool is_nmi(u32 intr_info)
>  {
>  	return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK))
> @@ -2791,7 +2800,12 @@ static void nested_vmx_setup_ctls_msrs(struct vcpu_vmx *vmx)
>  	if (cpu_has_vmx_vmfunc()) {
>  		vmx->nested.nested_vmx_secondary_ctls_high |=
>  			SECONDARY_EXEC_ENABLE_VMFUNC;
> -		vmx->nested.nested_vmx_vmfunc_controls = 0;
> +		/*
> +		 * Advertise EPTP switching unconditionally
> +		 * since we emulate it
> +		 */
> +		vmx->nested.nested_vmx_vmfunc_controls =
> +			VMX_VMFUNC_EPTP_SWITCHING;

Should this only be advertised, if enable_ept is set (if the guest also
sees/can use SECONDARY_EXEC_ENABLE_EPT)?

>  	}
>  
>  	/*
> @@ -7767,6 +7781,85 @@ static int handle_preemption_timer(struct kvm_vcpu *vcpu)
>  	return 1;
>  }
>  
> +static bool check_ept_address_valid(struct kvm_vcpu *vcpu, u64 address)

check_..._valid -> valid_ept_address() ?

> +{
> +	struct vcpu_vmx *vmx = to_vmx(vcpu);
> +	u64 mask = VMX_EPT_RWX_MASK;
> +	int maxphyaddr = cpuid_maxphyaddr(vcpu);
> +	struct kvm_mmu *mmu = vcpu->arch.walk_mmu;
> +
> +	/* Check for execute_only validity */
> +	if ((address & mask) == VMX_EPT_EXECUTABLE_MASK) {
> +		if (!(vmx->nested.nested_vmx_ept_caps &
> +		      VMX_EPT_EXECUTE_ONLY_BIT))
> +			return false;
> +	}
> +
> +	/* Bits 5:3 must be 3 */
> +	if (((address >> VMX_EPT_GAW_EPTP_SHIFT) & 0x7) != VMX_EPT_DEFAULT_GAW)
> +		return false;
> +
> +	/* Reserved bits should not be set */
> +	if (address >> maxphyaddr || ((address >> 7) & 0x1f))
> +		return false;
> +
> +	/* AD, if set, should be supported */
> +	if ((address & VMX_EPT_AD_ENABLE_BIT)) {
> +		if (!enable_ept_ad_bits)
> +			return false;
> +		mmu->ept_ad = true;
> +	} else
> +		mmu->ept_ad = false;

I wouldn't expect a "check" function to modify the mmu. Can you move
modifying the mmu outside of this function (leaving the
enable_ept_ad_bits check in place)? (and maybe even set mmu->ept_ad
_after_ the kvm_mmu_unload(vcpu)?, just when setting vmcs12->ept_pointer?)

> +
> +	return true;
> +}
> +
> +static int nested_vmx_eptp_switching(struct kvm_vcpu *vcpu,
> +				     struct vmcs12 *vmcs12)
> +{
> +	u32 index = vcpu->arch.regs[VCPU_REGS_RCX];
> +	u64 *l1_eptp_list, address;
> +	struct page *page;
> +
> +	if (!nested_cpu_has_eptp_switching(vmcs12) ||
> +	    !nested_cpu_has_ept(vmcs12))
> +		return 1;
> +
> +	if (index >= VMFUNC_EPTP_ENTRIES)
> +		return 1;
> +
> +	page = nested_get_page(vcpu, vmcs12->eptp_list_address);
> +	if (!page)
> +		return 1;
> +
> +	l1_eptp_list = kmap(page);
> +	address = l1_eptp_list[index];
> +
> +	/*
> +	 * If the (L2) guest does a vmfunc to the currently
> +	 * active ept pointer, we don't have to do anything else
> +	 */
> +	if (vmcs12->ept_pointer != address) {
> +		if (!check_ept_address_valid(vcpu, address)) {
> +			kunmap(page);
> +			nested_release_page_clean(page);
> +			return 1;
> +		}
> +		kvm_mmu_unload(vcpu);
> +		vmcs12->ept_pointer = address;
> +		/*
> +		 * TODO: Check what's the correct approach in case
> +		 * mmu reload fails. Currently, we just let the next
> +		 * reload potentially fail
> +		 */
> +		kvm_mmu_reload(vcpu);

So, what actually happens if this generates a tripple fault? I guess we
will kill the (nested) hypervisor?

> +	}
> +
> +	kunmap(page);
> +	nested_release_page_clean(page);
> +	return 0;
> +}
> +
>  static int handle_vmfunc(struct kvm_vcpu *vcpu)
>  {
>  	struct vcpu_vmx *vmx = to_vmx(vcpu);
> @@ -7786,7 +7879,16 @@ static int handle_vmfunc(struct kvm_vcpu *vcpu)
>  	vmcs12 = get_vmcs12(vcpu);
>  	if ((vmcs12->vm_function_control & (1 << function)) == 0)
>  		goto fail;
> -	WARN_ONCE(1, "VMCS12 VM function control should have been zero");
> +
> +	switch (function) {
> +	case 0:
> +		if (nested_vmx_eptp_switching(vcpu, vmcs12))
> +			goto fail;
> +		break;
> +	default:
> +		goto fail;
> +	}
> +	return kvm_skip_emulated_instruction(vcpu);
>  
>  fail:
>  	nested_vmx_vmexit(vcpu, vmx->exit_reason,
> @@ -10354,10 +10456,20 @@ static int check_vmentry_prereqs(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
>  				vmx->nested.nested_vmx_entry_ctls_high))
>  		return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
>  
> -	if (nested_cpu_has_vmfunc(vmcs12) &&
> -	    (vmcs12->vm_function_control &
> -	     ~vmx->nested.nested_vmx_vmfunc_controls))
> -		return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
> +	if (nested_cpu_has_vmfunc(vmcs12)) {
> +		if (vmcs12->vm_function_control &
> +		    ~vmx->nested.nested_vmx_vmfunc_controls)
> +			return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
> +
> +		if (nested_cpu_has_eptp_switching(vmcs12)) {
> +			if (!nested_cpu_has_ept(vmcs12) ||
> +			    (vmcs12->eptp_list_address >>
> +			     cpuid_maxphyaddr(vcpu)) ||
> +			    !IS_ALIGNED(vmcs12->eptp_list_address, 4096))
> +				return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
> +		}
> +	}
> +
>  
>  	if (vmcs12->cr3_target_count > nested_cpu_vmx_misc_cr3_count(vcpu))
>  		return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
> 


-- 

Thanks,

David