On Tue, Oct 27, 2015 at 01:40:04PM +0100, Martin Pieuchot wrote:
> On 25/10/15(Sun) 16:21, Sebastian Benoit wrote:
> > Martin Pieuchot(m...@openbsd.org) on 2015.10.25 16:14:27 +0100:
> > > Diff below merges the guts of rtable_mpath_match() into rtable_lookup().
> > > As for the previous rtable_mpath_* diff this is a step towards MPATH by
> > > default.
> > >
> > > This diff introduces a behavior change for RTM_GET. If multiple route
> > > exists a gateway MUST be specified and the first one in the tree won't
> > > be automagically selected by the kernel.
>
> After discussion here's a diff that do almost the same refactoring but
> without introducing a behavior change. ok?
Two nits.
> @@ -264,6 +264,16 @@ rtable_lookup(unsigned int rtableid, str
> return (NULL);
>
> rt = ((struct rtentry *)rn);
> +
> +#ifndef SMALL_KERNEL
> + if (rnh->rnh_multipath) {
> + rt = rt_mpath_matchgate(rt, gateway, prio);
> + if (rt == NULL)
> + return (NULL);
> + }
> +
Please do not add too many new lines.
> +#endif /* !SMALL_KERNEL */
> +
> rtref(rt);
>
> return (rt);
> @@ -620,51 +620,36 @@ route_output(struct mbuf *m, ...)
> error = EAFNOSUPPORT;
> goto flush;
> }
> +
> rt = rtable_lookup(tableid, info.rti_info[RTAX_DST],
> - info.rti_info[RTAX_NETMASK]);
> - if (rt == NULL) {
> - error = ESRCH;
> - goto flush;
> + info.rti_info[RTAX_NETMASK], info.rti_info[RTAX_GATEWAY],
> + prio);
> +#ifdef SMALL_KERNEL
I think this should be #ifndef
> + /*
> + * If we got multipath routes, we require users to specify
> + * a matching gateway, except for RTM_GET.
> + */
> + if ((rt != NULL) && ISSET(rt->rt_flags, RTF_MPATH) &&
> + (info.rti_info[RTAX_GATEWAY] == NULL) &&
> + (rtm->rtm_type != RTM_GET)) {
> + rtfree(rt);
> + rt = NULL;
> }
Otherwise OK bluhm@
