1 files changed, 90 insertions, 42 deletions

diff --git a/bgpd/bgp_zebra.c b/bgpd/bgp_zebra.c
index f3ab608492..4f54bc81fb 100644
--- a/bgpd/bgp_zebra.c
+++ b/bgpd/bgp_zebra.c
@@ -51,7 +51,7 @@
 #include "bgpd/bgp_nht.h"
 #include "bgpd/bgp_bfd.h"
 #include "bgpd/bgp_label.h"
-#if ENABLE_BGP_VNC
+#ifdef ENABLE_BGP_VNC
 #include "bgpd/rfapi/rfapi_backend.h"
 #include "bgpd/rfapi/vnc_export_bgp.h"
 #endif
@@ -66,19 +66,19 @@
 struct zclient *zclient = NULL;
 
 /* Can we install into zebra? */
-static inline int bgp_install_info_to_zebra(struct bgp *bgp)
+static inline bool bgp_install_info_to_zebra(struct bgp *bgp)
 {
 	if (zclient->sock <= 0)
-		return 0;
+		return false;
 
 	if (!IS_BGP_INST_KNOWN_TO_ZEBRA(bgp)) {
 		zlog_debug(
 			"%s: No zebra instance to talk to, not installing information",
 			__func__);
-		return 0;
+		return false;
 	}
 
-	return 1;
+	return true;
 }
 
 int zclient_num_connects;
@@ -928,8 +928,8 @@ bgp_path_info_to_ipv6_nexthop(struct bgp_path_info *path, ifindex_t *ifindex)
 	return nexthop;
 }
 
-static int bgp_table_map_apply(struct route_map *map, struct prefix *p,
-			       struct bgp_path_info *path)
+static bool bgp_table_map_apply(struct route_map *map, const struct prefix *p,
+				struct bgp_path_info *path)
 {
 	route_map_result_t ret;
 
@@ -937,7 +937,7 @@ static int bgp_table_map_apply(struct route_map *map, struct prefix *p,
 	bgp_attr_flush(path->attr);
 
 	if (ret != RMAP_DENYMATCH)
-		return 1;
+		return true;
 
 	if (bgp_debug_zebra(p)) {
 		if (p->family == AF_INET) {
@@ -965,7 +965,7 @@ static int bgp_table_map_apply(struct route_map *map, struct prefix *p,
 					  buf[1], sizeof(buf[1])));
 		}
 	}
-	return 0;
+	return false;
 }
 
 static struct thread *bgp_tm_thread_connect;
@@ -1058,12 +1058,10 @@ int bgp_zebra_get_table_range(uint32_t chunk_size,
 	return 0;
 }
 
-static int update_ipv4nh_for_route_install(int nh_othervrf,
-					   struct bgp *nh_bgp,
-					   struct in_addr *nexthop,
-					   struct attr *attr,
-					   bool is_evpn,
-					   struct zapi_nexthop *api_nh)
+static bool update_ipv4nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
+					    struct in_addr *nexthop,
+					    struct attr *attr, bool is_evpn,
+					    struct zapi_nexthop *api_nh)
 {
 	api_nh->gate.ipv4 = *nexthop;
 	api_nh->vrf_id = nh_bgp->vrf_id;
@@ -1083,15 +1081,16 @@ static int update_ipv4nh_for_route_install(int nh_othervrf,
 	} else
 		api_nh->type = NEXTHOP_TYPE_IPV4;
 
-	return 1;
+	return true;
 }
 
-static int
-update_ipv6nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
-				struct in6_addr *nexthop,
-				ifindex_t ifindex, struct bgp_path_info *pi,
-				struct bgp_path_info *best_pi, bool is_evpn,
-				struct zapi_nexthop *api_nh)
+static bool update_ipv6nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
+					    struct in6_addr *nexthop,
+					    ifindex_t ifindex,
+					    struct bgp_path_info *pi,
+					    struct bgp_path_info *best_pi,
+					    bool is_evpn,
+					    struct zapi_nexthop *api_nh)
 {
 	struct attr *attr;
 
@@ -1108,7 +1107,7 @@ update_ipv6nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
 			api_nh->ifindex = attr->nh_ifindex;
 		} else if (IN6_IS_ADDR_LINKLOCAL(nexthop)) {
 			if (ifindex == 0)
-				return 0;
+				return false;
 			api_nh->type = NEXTHOP_TYPE_IPV6_IFINDEX;
 			api_nh->ifindex = ifindex;
 		} else {
@@ -1136,7 +1135,7 @@ update_ipv6nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
 			}
 
 			if (ifindex == 0)
-				return 0;
+				return false;
 			api_nh->type = NEXTHOP_TYPE_IPV6_IFINDEX;
 			api_nh->ifindex = ifindex;
 		} else {
@@ -1146,10 +1145,35 @@ update_ipv6nh_for_route_install(int nh_othervrf, struct bgp *nh_bgp,
 	}
 	api_nh->gate.ipv6 = *nexthop;
 
-	return 1;
+	return true;
 }
 
-void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
+static bool bgp_zebra_use_nhop_weighted(struct bgp *bgp, struct attr *attr,
+					uint64_t tot_bw, uint32_t *nh_weight)
+{
+	uint32_t bw;
+	uint64_t tmp;
+
+	bw = attr->link_bw;
+	/* zero link-bandwidth and link-bandwidth not present are treated
+	 * as the same situation.
+	 */
+	if (!bw) {
+		/* the only situations should be if we're either told
+		 * to skip or use default weight.
+		 */
+		if (bgp->lb_handling == BGP_LINK_BW_SKIP_MISSING)
+			return false;
+		*nh_weight = BGP_ZEBRA_DEFAULT_NHOP_WEIGHT;
+	} else {
+		tmp = (uint64_t)bw * 100;
+		*nh_weight = ((uint32_t)(tmp / tot_bw));
+	}
+
+	return true;
+}
+
+void bgp_zebra_announce(struct bgp_node *rn, const struct prefix *p,
 			struct bgp_path_info *info, struct bgp *bgp, afi_t afi,
 			safi_t safi)
 {
@@ -1171,6 +1195,8 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 	char buf_prefix[PREFIX_STRLEN];	/* filled in if we are debugging */
 	bool is_evpn;
 	int nh_updated;
+	bool do_wt_ecmp;
+	uint64_t cum_bw = 0;
 
 	/* Don't try to install if we're not connected to Zebra or Zebra doesn't
 	 * know of this instance.
@@ -1185,7 +1211,8 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 		prefix2str(p, buf_prefix, sizeof(buf_prefix));
 
 	if (safi == SAFI_FLOWSPEC) {
-		bgp_pbr_update_entry(bgp, &rn->p, info, afi, safi, true);
+		bgp_pbr_update_entry(bgp, bgp_node_get_prefix(rn),
+				     info, afi, safi, true);
 		return;
 	}
 
@@ -1240,11 +1267,20 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 
 	/* Metric is currently based on the best-path only */
 	metric = info->attr->med;
+
+	/* Determine if we're doing weighted ECMP or not */
+	do_wt_ecmp = bgp_path_info_mpath_chkwtd(bgp, info);
+	if (do_wt_ecmp)
+		cum_bw = bgp_path_info_mpath_cumbw(info);
+
 	for (mpinfo = info; mpinfo; mpinfo = bgp_path_info_mpath_next(mpinfo)) {
+		uint32_t nh_weight;
+
 		if (valid_nh_count >= multipath_num)
 			break;
 
 		*mpinfo_cp = *mpinfo;
+		nh_weight = 0;
 
 		/* Get nexthop address-family */
 		if (p->family == AF_INET
@@ -1257,6 +1293,15 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 		else
 			continue;
 
+		/* If processing for weighted ECMP, determine the next hop's
+		 * weight. Based on user setting, we may skip the next hop
+		 * in some situations.
+		 */
+		if (do_wt_ecmp) {
+			if (!bgp_zebra_use_nhop_weighted(bgp, mpinfo->attr,
+							 cum_bw, &nh_weight))
+				continue;
+		}
 		api_nh = &api.nexthops[valid_nh_count];
 		if (nh_family == AF_INET) {
 			if (bgp_debug_zebra(&api.prefix)) {
@@ -1356,6 +1401,8 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 		}
 		memcpy(&api_nh->rmac, &(mpinfo->attr->rmac),
 		       sizeof(struct ethaddr));
+		api_nh->weight = nh_weight;
+
 		valid_nh_count++;
 	}
 
@@ -1435,9 +1482,10 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 				snprintf(eth_buf, sizeof(eth_buf), " RMAC %s",
 					 prefix_mac2str(&api_nh->rmac,
 							buf1, sizeof(buf1)));
-			zlog_debug("  nhop [%d]: %s if %u VRF %u %s %s",
+			zlog_debug("  nhop [%d]: %s if %u VRF %u wt %u %s %s",
 				   i + 1, nh_buf, api_nh->ifindex,
-				   api_nh->vrf_id, label_buf, eth_buf);
+				   api_nh->vrf_id, api_nh->weight,
+				   label_buf, eth_buf);
 		}
 	}
 
@@ -1481,11 +1529,11 @@ void bgp_zebra_announce_table(struct bgp *bgp, afi_t afi, safi_t safi)
 			     && (pi->sub_type == BGP_ROUTE_NORMAL
 				 || pi->sub_type == BGP_ROUTE_IMPORTED)))
 
-				bgp_zebra_announce(rn, &rn->p, pi, bgp, afi,
-						   safi);
+				bgp_zebra_announce(rn, bgp_node_get_prefix(rn),
+						   pi, bgp, afi, safi);
 }
 
-void bgp_zebra_withdraw(struct prefix *p, struct bgp_path_info *info,
+void bgp_zebra_withdraw(const struct prefix *p, struct bgp_path_info *info,
 			struct bgp *bgp, safi_t safi)
 {
 	struct zapi_route api;
@@ -1605,7 +1653,7 @@ int bgp_redistribute_set(struct bgp *bgp, afi_t afi, int type,
 		if (vrf_bitmap_check(zclient->redist[afi][type], bgp->vrf_id))
 			return CMD_WARNING;
 
-#if ENABLE_BGP_VNC
+#ifdef ENABLE_BGP_VNC
 		if (EVPN_ENABLED(bgp) && type == ZEBRA_ROUTE_VNC_DIRECT) {
 			vnc_export_bgp_enable(
 				bgp, afi); /* only enables if mode bits cfg'd */
@@ -1660,11 +1708,11 @@ int bgp_redistribute_resend(struct bgp *bgp, afi_t afi, int type,
 }
 
 /* Redistribute with route-map specification.  */
-int bgp_redistribute_rmap_set(struct bgp_redist *red, const char *name,
-			      struct route_map *route_map)
+bool bgp_redistribute_rmap_set(struct bgp_redist *red, const char *name,
+			       struct route_map *route_map)
 {
 	if (red->rmap.name && (strcmp(red->rmap.name, name) == 0))
-		return 0;
+		return false;
 
 	XFREE(MTYPE_ROUTE_MAP_NAME, red->rmap.name);
 	/* Decrement the count for existing routemap and
@@ -1675,18 +1723,18 @@ int bgp_redistribute_rmap_set(struct bgp_redist *red, const char *name,
 	red->rmap.map = route_map;
 	route_map_counter_increment(red->rmap.map);
 
-	return 1;
+	return true;
 }
 
 /* Redistribute with metric specification.  */
-int bgp_redistribute_metric_set(struct bgp *bgp, struct bgp_redist *red,
-				afi_t afi, int type, uint32_t metric)
+bool bgp_redistribute_metric_set(struct bgp *bgp, struct bgp_redist *red,
+				 afi_t afi, int type, uint32_t metric)
 {
 	struct bgp_node *rn;
 	struct bgp_path_info *pi;
 
 	if (red->redist_metric_flag && red->redist_metric == metric)
-		return 0;
+		return false;
 
 	red->redist_metric_flag = 1;
 	red->redist_metric = metric;
@@ -1713,7 +1761,7 @@ int bgp_redistribute_metric_set(struct bgp *bgp, struct bgp_redist *red,
 		}
 	}
 
-	return 1;
+	return true;
 }
 
 /* Unset redistribution.  */
@@ -1766,7 +1814,7 @@ int bgp_redistribute_unset(struct bgp *bgp, afi_t afi, int type,
  * they operate within bgpd irrespective of zebra connection
  * status. red lookup fails if there is no zebra connection.
  */
-#if ENABLE_BGP_VNC
+#ifdef ENABLE_BGP_VNC
 	if (EVPN_ENABLED(bgp) && type == ZEBRA_ROUTE_VNC_DIRECT) {
 		vnc_export_bgp_disable(bgp, afi);
 	}