On Mon, 16 Oct 2023, Jan Beulich wrote: > On 03.10.2023 17:24, Federico Serafini wrote: > > --- a/xen/arch/x86/mm.c > > +++ b/xen/arch/x86/mm.c > > @@ -5901,17 +5901,17 @@ int destroy_xen_mappings(unsigned long s, unsigned > > long e) > > * a problem. > > */ > > void init_or_livepatch modify_xen_mappings_lite( > > - unsigned long s, unsigned long e, unsigned int _nf) > > + unsigned long s, unsigned long e, unsigned int nf) > > { > > - unsigned long v = s, fm, nf; > > + unsigned long v = s, fm, flags; > > While it looks correct, I consider this an unacceptably dangerous > change: What if by the time this is to be committed some new use of > the local "nf" appears, without resulting in fuzz while applying the > patch? Imo this needs doing in two steps: First nf -> flags, then > _nf -> nf.
Wouldn't it be sufficient for the committer to pay special attention when committing this patch? We are in code freeze anyway, the rate of changes affecting staging is low.