On Wed, Sep 26, 2018 at 03:19:00PM +1000, David Gibson wrote:
> On Fri, Sep 21, 2018 at 08:01:50PM +1000, Paul Mackerras wrote:
> > +static void kvmhv_remove_nested(struct kvm_nested_guest *gp)
> > +{
> > + struct kvm *kvm = gp->parent;
> > + int lpid = gp->l1_lpid;
> > + long ref;
> > +
> > + spin_lock(&kvm->mmu_lock);
> > + if (gp == kvm->arch.nested_guests[lpid]) {
> > + kvm->arch.nested_guests[lpid] = NULL;
> > + while (--lpid >= 0 && !kvm->arch.nested_guests[lpid])
> > + ;
>
> This only looks correct of gp is the guest of KVM with the highest
> l1_lpid. It's not obvious why that would have to be the case.
Not obvious because not true. Good catch. :)
Paul.
