2017-01-24 14:06+0100, Dmitry Vyukov: > If avic is not enabled, avic_vm_init() does nothing and returns early. > However, avic_vm_destroy() still tries to destroy what hasn't been created. > The only bad consequence of this now is that avic_vm_destroy() uses > svm_vm_data_hash_lock that hasn't been initialized (and is not meant > to be used at all if avic is not enabled). > > Return early from avic_vm_destroy() if avic is not enabled. > It has nothing to destroy. > > Signed-off-by: Dmitry Vyukov <dvyukov@xxxxxxxxxx> > Cc: Joerg Roedel <joro@xxxxxxxxxx> > Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx> > Cc: "Radim Krčmář" <rkrcmar@xxxxxxxxxx> > Cc: David Hildenbrand <david@xxxxxxxxxx> > Cc: kvm@xxxxxxxxxxxxxxx > Cc: syzkaller@xxxxxxxxxxxxxxxx > --- svm_vm_data_hash_lock is initialized unconditionally now, so I've queued it for 4.12, Thanks. > arch/x86/kvm/svm.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c > index 08a4d3ab3455..3aecaa13fd22 100644 > --- a/arch/x86/kvm/svm.c > +++ b/arch/x86/kvm/svm.c > @@ -1382,6 +1382,9 @@ static void avic_vm_destroy(struct kvm *kvm) > unsigned long flags; > struct kvm_arch *vm_data = &kvm->arch; > > + if (!avic) > + return; > + > avic_free_vm_id(vm_data->avic_vm_id); > > if (vm_data->avic_logical_id_table_page) > -- > 2.11.0.483.g087da7b7c-goog >
