Re: [PATCH v2 1/3] KVM: nSVM: restore the L1 host state prior to resuming a nested guest on SMM exit

[Maxim Levitsky]

On Wed, 2021-09-08 at 22:04 +0000, Sean Christopherson wrote:
> On Mon, Aug 23, 2021, Maxim Levitsky wrote:
> > If the guest is entered prior to restoring the host save area,
> > the guest entry code might see incorrect L1 state (e.g paging state).
> > 
> > Signed-off-by: Maxim Levitsky <mlevitsk@xxxxxxxxxx>
> > ---
> >  arch/x86/kvm/svm/svm.c | 23 +++++++++++++----------
> >  1 file changed, 13 insertions(+), 10 deletions(-)
> > 
> > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> > index 1a70e11f0487..ea7a4dacd42f 100644
> > --- a/arch/x86/kvm/svm/svm.c
> > +++ b/arch/x86/kvm/svm/svm.c
> > @@ -4347,27 +4347,30 @@ static int svm_leave_smm(struct kvm_vcpu *vcpu, const char *smstate)
> >  					 gpa_to_gfn(vmcb12_gpa), &map) == -EINVAL)
> >  				return 1;
> >  
> > -			if (svm_allocate_nested(svm))
> > +			if (kvm_vcpu_map(vcpu, gpa_to_gfn(svm->nested.hsave_msr),
> > +					 &map_save) == -EINVAL)
> >  				return 1;
> 
> Returning here will neglect to unmap "map".
> 
> >  
> > -			vmcb12 = map.hva;
> > -
> > -			nested_load_control_from_vmcb12(svm, &vmcb12->control);
> > -
> > -			ret = enter_svm_guest_mode(vcpu, vmcb12_gpa, vmcb12);
> > -			kvm_vcpu_unmap(vcpu, &map, true);
> > +			if (svm_allocate_nested(svm))
> > +				return 1;
> 
> Ditto here for both "map" and "map_save", though it looks like there's a
> pre-existing bug if svm_allocate_nested() fails.  If you add a prep cleanup patch
> to remove the statement nesting (between the bug fix and this patch), it will make
> handling this a lot easier, e.g.
> 
> static int svm_leave_smm(struct kvm_vcpu *vcpu, const char *smstate)
> {
> 	struct vcpu_svm *svm = to_svm(vcpu);
> 	struct kvm_host_map map, map_save;
> 	u64 saved_efer, vmcb12_gpa;
> 	struct vmcb *vmcb12;
> 	int ret;
> 
> 	if (!guest_cpuid_has(vcpu, X86_FEATURE_LM))
> 		return 0;
> 
> 	/* Non-zero if SMI arrived while vCPU was in guest mode. */
> 	if (!GET_SMSTATE(u64, smstate, 0x7ed8))
> 		return 0;
> 
> 	if (!guest_cpuid_has(vcpu, X86_FEATURE_SVM))
> 		return 1;
> 
> 	saved_efer = GET_SMSTATE(u64, smstate, 0x7ed0);
> 	if (!(saved_efer & EFER_SVME))
> 		return 1;
> 
> 	vmcb12_gpa = GET_SMSTATE(u64, smstate, 0x7ee0);
> 	if (kvm_vcpu_map(vcpu, gpa_to_gfn(vmcb12_gpa), &map) == -EINVAL)
> 		return 1;
> 
> 	ret = 1;
> 	if (kvm_vcpu_map(vcpu, gpa_to_gfn(svm->nested.hsave_msr), &map_save) == -EINVAL)
> 		goto unmap_map;
> 
> 	if (svm_allocate_nested(svm))
> 		goto unmap_save;
> 
> 	/*
> 	 * Restore L1 host state from L1 HSAVE area as VMCB01 was
> 	 * used during SMM (see svm_enter_smm())
> 	 */
> 
> 	svm_copy_vmrun_state(&svm->vmcb01.ptr->save,
> 				map_save.hva + 0x400);
> 
> 	/*
> 	 * Restore L2 state
> 	 */
> 
> 	vmcb12 = map.hva;
> 	nested_load_control_from_vmcb12(svm, &vmcb12->control);
> 	ret = enter_svm_guest_mode(vcpu, vmcb12_gpa, vmcb12);
> 
> unmap_save;
> 	kvm_vcpu_unmap(vcpu, &map_save, true);
> unmap_map:
> 	kvm_vcpu_unmap(vcpu, &map, true);
> 	return 1;
> }
> 
Will do. I haven't given the error path enough attention.

Thanks!
Best regards,
	Maxim Levitsky