Re: [PATCH 4.19 STABLE] KVM: x86: introduce is_pae_paging

[Thomas Lamprecht]

On 11/11/19 11:54 PM, Sean Christopherson wrote:
> From: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> 
> Upstream commit bf03d4f9334728bf7c8ffc7de787df48abd6340e.
> 
> Checking for 32-bit PAE is quite common around code that fiddles with
> the PDPTRs.  Add a function to compress all checks into a single
> invocation.
> 
> Moving to the common helper also fixes a subtle bug in kvm_set_cr3()
> where it fails to check is_long_mode() and results in KVM incorrectly
> attempting to load PDPTRs for a 64-bit guest.
> 
> Reviewed-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> [sean: backport to 4.x; handle vmx.c split in 5.x, call out the bugfix]
> Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> ---
>  arch/x86/kvm/vmx.c | 7 +++----
>  arch/x86/kvm/x86.c | 8 ++++----
>  arch/x86/kvm/x86.h | 5 +++++
>  3 files changed, 12 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 6f7b3acdab26..83acaed244ba 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -5181,7 +5181,7 @@ static void ept_load_pdptrs(struct kvm_vcpu *vcpu)
>  		      (unsigned long *)&vcpu->arch.regs_dirty))
>  		return;
>  
> -	if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) {
> +	if (is_pae_paging(vcpu)) {
>  		vmcs_write64(GUEST_PDPTR0, mmu->pdptrs[0]);
>  		vmcs_write64(GUEST_PDPTR1, mmu->pdptrs[1]);
>  		vmcs_write64(GUEST_PDPTR2, mmu->pdptrs[2]);
> @@ -5193,7 +5193,7 @@ static void ept_save_pdptrs(struct kvm_vcpu *vcpu)
>  {
>  	struct kvm_mmu *mmu = vcpu->arch.walk_mmu;
>  
> -	if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) {
> +	if (is_pae_paging(vcpu)) {
>  		mmu->pdptrs[0] = vmcs_read64(GUEST_PDPTR0);
>  		mmu->pdptrs[1] = vmcs_read64(GUEST_PDPTR1);
>  		mmu->pdptrs[2] = vmcs_read64(GUEST_PDPTR2);
> @@ -12021,8 +12021,7 @@ static int nested_vmx_load_cr3(struct kvm_vcpu *vcpu, unsigned long cr3, bool ne
>  		 * If PAE paging and EPT are both on, CR3 is not used by the CPU and
>  		 * must not be dereferenced.
>  		 */
> -		if (!is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu) &&
> -		    !nested_ept) {
> +		if (is_pae_paging(vcpu) && !nested_ept) {
>  			if (!load_pdptrs(vcpu, vcpu->arch.walk_mmu, cr3)) {
>  				*entry_failure_code = ENTRY_FAIL_PDPTE;
>  				return 1;
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 6ae8a013af31..b9b87fb75ac0 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -633,7 +633,7 @@ bool pdptrs_changed(struct kvm_vcpu *vcpu)
>  	gfn_t gfn;
>  	int r;
>  
> -	if (is_long_mode(vcpu) || !is_pae(vcpu) || !is_paging(vcpu))
> +	if (!is_pae_paging(vcpu))
>  		return false;
>  
>  	if (!test_bit(VCPU_EXREG_PDPTR,
> @@ -884,8 +884,8 @@ int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
>  	if (is_long_mode(vcpu) &&
>  	    (cr3 & rsvd_bits(cpuid_maxphyaddr(vcpu), 63)))
>  		return 1;
> -	else if (is_pae(vcpu) && is_paging(vcpu) &&
> -		   !load_pdptrs(vcpu, vcpu->arch.walk_mmu, cr3))
> +	else if (is_pae_paging(vcpu) &&
> +		 !load_pdptrs(vcpu, vcpu->arch.walk_mmu, cr3))
>  		return 1;
>  
>  	kvm_mmu_new_cr3(vcpu, cr3, skip_tlb_flush);
> @@ -8312,7 +8312,7 @@ static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
>  		kvm_update_cpuid(vcpu);
>  
>  	idx = srcu_read_lock(&vcpu->kvm->srcu);
> -	if (!is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu)) {
> +	if (is_pae_paging(vcpu)) {
>  		load_pdptrs(vcpu, vcpu->arch.walk_mmu, kvm_read_cr3(vcpu));
>  		mmu_reset_needed = 1;
>  	}
> diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
> index 3a91ea760f07..608e5f8c5d0a 100644
> --- a/arch/x86/kvm/x86.h
> +++ b/arch/x86/kvm/x86.h
> @@ -139,6 +139,11 @@ static inline int is_paging(struct kvm_vcpu *vcpu)
>  	return likely(kvm_read_cr0_bits(vcpu, X86_CR0_PG));
>  }
>  
> +static inline bool is_pae_paging(struct kvm_vcpu *vcpu)
> +{
> +	return !is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu);
> +}
> +
>  static inline u32 bit(int bitno)
>  {
>  	return 1 << (bitno & 31);
> 

Thanks for the fast and good answer and the backport, made things more
clear here. Cannot reproduce the issues with that patch anymore

Tested-by: Thomas Lamprecht <t.lamprecht@xxxxxxxxxxx>