On Thu, Aug 18, 2022 at 03:22:50PM +0200, Claudio Jeker wrote:
> Noticed while compling with gcc. In kroute_insert() the check for possible
> multipath routes is:
> if (krm == NULL)
> kr_redistribute(IMSG_NETWORK_ADD, kt, kf);
>
> The problem is krm is only set in the IPv4 path but not in the IPv6 one.
> The diff below fixes this by using a new variable multipath which is set
> in the case where a multipath route is hit.
Ugh. Sorry for missing that. Yes, a new variable seems like the most
readable fix.
ok tb
>
> --
> :wq Claudio
>
> Index: kroute.c
> ===================================================================
> RCS file: /cvs/src/usr.sbin/bgpd/kroute.c,v
> retrieving revision 1.293
> diff -u -p -r1.293 kroute.c
> --- kroute.c 18 Aug 2022 12:14:00 -0000 1.293
> +++ kroute.c 18 Aug 2022 13:22:06 -0000
> @@ -1607,6 +1607,7 @@ kroute_insert(struct ktable *kt, struct
> struct kroute6 *kr6, *kr6m;
> struct knexthop *n;
> uint32_t mplslabel = 0;
> + int multipath = 0;
>
> if (kf->prefix.aid == AID_VPN_IPv4 ||
> kf->prefix.aid == AID_VPN_IPv6) {
> @@ -1648,6 +1649,7 @@ kroute_insert(struct ktable *kt, struct
> while (krm->next != NULL)
> krm = krm->next;
> krm->next = kr;
> + multipath = 1;
> }
>
> if (kf->flags & F_BGPD)
> @@ -1684,6 +1686,7 @@ kroute_insert(struct ktable *kt, struct
> while (kr6m->next != NULL)
> kr6m = kr6m->next;
> kr6m->next = kr6;
> + multipath = 1;
> }
>
> if (kf->flags & F_BGPD)
> @@ -1699,8 +1702,8 @@ kroute_insert(struct ktable *kt, struct
> kf->prefixlen) == 0)
> knexthop_validate(kt, n);
>
> - if (krm == NULL)
> - /* redistribute multipath routes only once */
> + /* redistribute multipath routes only once */
> + if (!multipath)
> kr_redistribute(IMSG_NETWORK_ADD, kt, kf);
> }
>
>
