> -----Original Message-----
> > > >
> > > > The problem really comes down to defining
> xenforeignmemory_map2() in
> > > terms of xenforeignmemory_map(). It basically can't be safely done.
> Could
> > > you define xenforeignmemory_map2() as abort() in the compat case
> > > instead?
> > > >
> > >
> > > xen_replace_cache_entry() is not called in patch #3. Which means it's
> > > safe to use a fallback version (xenforeignmemory_map) in
> > > xen_remap_bucket here.
> >
> > I still don't like the fact that the compat definition of
> xenforeignmemory_map2() loses the extra argument. That's going to catch
> someone out one day. Is there any way you could re-work it so that
> xenforeignmemory_map() is uses in the cases where the memory
> placement does not matter?
>
> We could assert(vaddr == NULL) in the compat implementation of
> xenforeignmemory_map2. Would that work?
>
Yes, if the patch was changed from being a straight #define as it is now to an
inline that had such an assertion then that would be ok.
Cheers,
Paul
