On Mon, Dec 08, 2014 at 12:13:11PM +0100, Martin Pieuchot wrote:
> On 04/12/14(Thu) 00:26, Alexander Bluhm wrote:
> > On Wed, Nov 26, 2014 at 12:58:35PM +0100, Martin Pieuchot wrote:
> > > @@ -761,7 +754,17 @@ report:
> > > error = EDQUOT;
> > > goto flush;
> > > }
> > > - ifa = info.rti_ifa;
> > > + /*
> > > + * new gateway could require new ifaddr, ifp;
> > > + * flags may also be different; ifp may be specified
> > > + * by ll sockaddr when protocol address is ambiguous
> > > + */
> > > + if (newgate || info.rti_info[RTAX_IFP] != NULL ||
> > > + info.rti_info[RTAX_IFA] != NULL) {
> >
> > rt_getifa() may also use RTAX_DST to set rti_ifa. So I think there
> > should be a "|| info.rti_info[RTAX_DST] != NULL" in the condition.
>
> Except that here we are dealing with RTM_CHANGE and RTAX_DST is used to
> find the route entry we're about to modify. So if you pass a different
> destination, the kernel won't find any route and the code will fail
> earlier.
>
> So ok?
Yes, RTAX_DST has been checked before. OK bluhm@
