Re: [PATCH v5 4/9] xen: introduce XEN_DOMCTL_devour

[Ian Campbell <Ian.Campbell@citrix.com>]

On Thu, 2014-12-18 at 13:57 +0000, Jan Beulich wrote:
> >>> On 11.12.14 at 14:45, <vkuznets@redhat.com> wrote:
> > --- a/xen/common/domctl.c
> > +++ b/xen/common/domctl.c
> > @@ -1177,6 +1177,39 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
> >      }
> >      break;
> >  
> > +    case XEN_DOMCTL_devour:
> > +    {
> > +        struct domain *recipient_dom;
> > +
> > +        if ( !d->recipient )
> > +        {
> > +            recipient_dom = get_domain_by_id(op->u.devour.recipient);
> > +            if ( recipient_dom == NULL )
> > +            {
> > +                ret = -ESRCH;
> > +                break;
> > +            }
> > +
> > +            if ( recipient_dom->tot_pages != 0 )
> > +            {
> > +                put_domain(recipient_dom);
> > +                ret = -EINVAL;
> > +                break;
> > +            }
> > +            /*
> > +             * Make sure no allocation/remapping is ongoing and set is_dying
> > +             * flag to prevent such actions in future.
> > +             */
> > +            spin_lock(&d->page_alloc_lock);
> > +            d->is_dying = DOMDYING_locked;
> > +            d->recipient = recipient_dom;
> 
> Is d == recipient_dom a valid case (not leading to any issues)?

I suspect not, due to the restriction that the recipient not have any
domheap pages, if d didn't have any domheap pages this whole dance is a
bit unnecessary.

In any case it doesn't seem very useful, so it's much easier to just
avoid the issue by outlawing it.

Ian.