Merge pull request #5827 from donaldsharp/missed_upstreaming

Missed upstreaming

bgpd/bgp_mplsvpn.c

@@ -1532,7 +1532,10 @@ void vpn_handle_router_id_update(struct bgp *bgp, bool withdraw,
 			ecom = bgp->vpn_policy[afi].rtlist[edir];
 			for (ALL_LIST_ELEMENTS_RO(bgp->vpn_policy[afi].
 						  export_vrf, node, vname)) {
-				bgp_import = bgp_lookup_by_name(vname);
+				if (strcmp(vname, VRF_DEFAULT_NAME) == 0)
+					bgp_import = bgp_get_default();
+				else
+					bgp_import = bgp_lookup_by_name(vname);
 				if (!bgp_import)
 					continue;
 
@@ -1572,7 +1575,10 @@ void vpn_handle_router_id_update(struct bgp *bgp, bool withdraw,
 			ecom = bgp->vpn_policy[afi].rtlist[edir];
 			for (ALL_LIST_ELEMENTS_RO(bgp->vpn_policy[afi].
 						  export_vrf, node, vname)) {
-				bgp_import = bgp_lookup_by_name(vname);
+				if (strcmp(vname, VRF_DEFAULT_NAME) == 0)
+					bgp_import = bgp_get_default();
+				else
+					bgp_import = bgp_lookup_by_name(vname);
 				if (!bgp_import)
 					continue;
 				if (bgp_import->vpn_policy[afi].rtlist[idir])

bgpd/bgp_network.c

@@ -57,8 +57,26 @@ struct bgp_listener {
 	union sockunion su;
 	struct thread *thread;
 	struct bgp *bgp;
+	char *name;
 };
 
+void bgp_dump_listener_info(struct vty *vty)
+{
+	struct listnode *node;
+	struct bgp_listener *listener;
+
+	vty_out(vty, "Name             fd Address\n");
+	vty_out(vty, "---------------------------\n");
+	for (ALL_LIST_ELEMENTS_RO(bm->listen_sockets, node, listener)) {
+		char buf[SU_ADDRSTRLEN];
+
+		vty_out(vty, "%-16s %d %s\n",
+			listener->name ? listener->name : VRF_DEFAULT_NAME,
+			listener->fd,
+			sockunion2str(&listener->su, buf, sizeof(buf)));
+	}
+}
+
 /*
  * Set MD5 key for the socket, for the given IPv4 peer address.
  * If the password is NULL or zero-length, the option will be disabled.
@@ -762,6 +780,7 @@ static int bgp_listener(int sock, struct sockaddr *sa, socklen_t salen,
 
 	listener = XCALLOC(MTYPE_BGP_LISTENER, sizeof(*listener));
 	listener->fd = sock;
+	listener->name = XSTRDUP(MTYPE_BGP_LISTENER, bgp->name);
 
 	/* this socket needs a change of ns. record bgp back pointer */
 	if (bgp->vrf_id != VRF_DEFAULT && vrf_is_backend_netns())
@@ -871,6 +890,7 @@ void bgp_close_vrf_socket(struct bgp *bgp)
 			thread_cancel(listener->thread);
 			close(listener->fd);
 			listnode_delete(bm->listen_sockets, listener);
+			XFREE(MTYPE_BGP_LISTENER, listener->name);
 			XFREE(MTYPE_BGP_LISTENER, listener);
 		}
 	}
@@ -892,6 +912,7 @@ void bgp_close(void)
 		thread_cancel(listener->thread);
 		close(listener->fd);
 		listnode_delete(bm->listen_sockets, listener);
+		XFREE(MTYPE_BGP_LISTENER, listener->name);
 		XFREE(MTYPE_BGP_LISTENER, listener);
 	}
 }

bgpd/bgp_network.h

@@ -23,6 +23,7 @@
 
 #define BGP_SOCKET_SNDBUF_SIZE 65536
 
+extern void bgp_dump_listener_info(struct vty *vty);
 extern int bgp_socket(struct bgp *bgp, unsigned short port,
 		      const char *address);
 extern void bgp_close_vrf_socket(struct bgp *bgp);

bgpd/bgp_route.c

@@ -69,6 +69,7 @@
 #include "bgpd/bgp_label.h"
 #include "bgpd/bgp_addpath.h"
 #include "bgpd/bgp_mac.h"
+#include "bgpd/bgp_network.h"
 
 #if ENABLE_BGP_VNC
 #include "bgpd/rfapi/rfapi_backend.h"
@@ -12766,6 +12767,18 @@ static void show_bgp_peerhash_entry(struct hash_bucket *bucket, void *arg)
 	       sockunion2str(&peer->su, buf, sizeof(buf)));
 }
 
+DEFUN (show_bgp_listeners,
+       show_bgp_listeners_cmd,
+       "show bgp listeners",
+       SHOW_STR
+       BGP_STR
+       "Display Listen Sockets and who created them\n")
+{
+	bgp_dump_listener_info(vty);
+
+	return CMD_SUCCESS;
+}
+
 DEFUN (show_bgp_peerhash,
        show_bgp_peerhash_cmd,
        "show bgp peerhash",
@@ -13155,6 +13168,7 @@ void bgp_route_init(void)
 	/* show bgp ipv4 flowspec detailed */
 	install_element(VIEW_NODE, &show_ip_bgp_flowspec_routes_detailed_cmd);
 
+	install_element(VIEW_NODE, &show_bgp_listeners_cmd);
 	install_element(VIEW_NODE, &show_bgp_peerhash_cmd);
 }
 

bgpd/bgpd.c

@@ -4071,6 +4071,9 @@ static int peer_flag_modify(struct peer *peer, uint32_t flag, int set)
 		/* Update flag override state accordingly. */
 		COND_FLAG(peer->flags_override, flag, set != invert);
 
+		if (set && flag == PEER_FLAG_CAPABILITY_ENHE)
+			bgp_nht_register_enhe_capability_interfaces(peer);
+
 		/* Execute flag action on peer. */
 		if (action.type == peer_change_reset)
 			peer_flag_modify_action(peer, flag);
@@ -4079,9 +4082,6 @@ static int peer_flag_modify(struct peer *peer, uint32_t flag, int set)
 		return 0;
 	}
 
-	if (set && flag == PEER_FLAG_CAPABILITY_ENHE)
-		bgp_nht_register_enhe_capability_interfaces(peer);
-
 	/*
 	 * Update peer-group members, unless they are explicitely overriding
 	 * peer-group configuration.

doc/user/bgp.rst

@@ -2412,6 +2412,12 @@ Debugging
 
    Show all enabled debugs.
 
+.. index:: show bgp listeners
+.. clicmd:: show bgp listeners
+
+   Display Listen sockets and the vrf that created them.  Useful for debugging of when
+   listen is not working and this is considered a developer debug statement.
+
 .. index:: [no] debug bgp neighbor-events
 .. clicmd:: [no] debug bgp neighbor-events
 

lib/systemd.c

@@ -114,8 +114,10 @@ void systemd_send_started(struct thread_master *m, int the_process)
 	systemd_master = m;
 
 	systemd_send_information("READY=1");
-	if (wsecs != 0)
+	if (wsecs != 0) {
+		systemd_send_information("WATCHDOG=1");
 		thread_add_timer(m, systemd_send_watchdog, m, wsecs, NULL);
+	}
 }
 
 void systemd_send_status(const char *status)

pimd/pim_ifchannel.c

@@ -234,8 +234,8 @@ void pim_ifchannel_delete(struct pim_ifchannel *ch)
 	RB_REMOVE(pim_ifchannel_rb, &pim_ifp->ifchannel_rb, ch);
 
 	if (PIM_DEBUG_PIM_TRACE)
-		zlog_debug("%s: ifchannel entry %s is deleted ", __func__,
-			   ch->sg_str);
+		zlog_debug("%s: ifchannel entry %s(%s) is deleted ", __func__,
+			   ch->sg_str, ch->interface->name);
 
 	XFREE(MTYPE_PIM_IFCHANNEL, ch);
 

zebra/rt_netlink.c

@@ -2588,7 +2588,7 @@ static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
 	struct interface *br_if;
 	struct ethaddr mac;
 	vlanid_t vid = 0;
-	struct prefix vtep_ip;
+	struct in_addr vtep_ip;
 	int vid_present = 0, dst_present = 0;
 	char buf[ETHER_ADDR_STRLEN];
 	char vid_buf[20];
@@ -2601,66 +2601,25 @@ static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
 	if (!is_evpn_enabled())
 		return 0;
 
-	/* The interface should exist. */
-	ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
-					ndm->ndm_ifindex);
-	if (!ifp || !ifp->info) {
-		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("\t%s without associated interface: %u",
-				   __func__, ndm->ndm_ifindex);
-		return 0;
-	}
-
-	/* The interface should be something we're interested in. */
-	if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp)) {
-		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("\t%s Not interested in %s, not a slave",
-				   __func__, ifp->name);
-		return 0;
-	}
-
-	/* Drop "permanent" entries. */
-	if (ndm->ndm_state & NUD_PERMANENT) {
-		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("\t%s Entry is PERMANENT, dropping",
-				   __func__);
-		return 0;
-	}
-
-	zif = (struct zebra_if *)ifp->info;
-	if ((br_if = zif->brslave_info.br_if) == NULL) {
-		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug(
-				"%s family %s IF %s(%u) brIF %u - no bridge master",
-				nl_msg_type_to_str(h->nlmsg_type),
-				nl_family_to_str(ndm->ndm_family), ifp->name,
-				ndm->ndm_ifindex,
-				zif->brslave_info.bridge_ifindex);
-		return 0;
-	}
-
-	/* Parse attributes and extract fields of interest. */
-	memset(tb, 0, sizeof(tb));
+	/* Parse attributes and extract fields of interest. Do basic
+	 * validation of the fields.
+	 */
+	memset(tb, 0, sizeof tb);
 	netlink_parse_rtattr(tb, NDA_MAX, NDA_RTA(ndm), len);
 
 	if (!tb[NDA_LLADDR]) {
 		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("%s family %s IF %s(%u) brIF %u - no LLADDR",
+			zlog_debug("%s AF_BRIDGE IF %u - no LLADDR",
 				   nl_msg_type_to_str(h->nlmsg_type),
-				   nl_family_to_str(ndm->ndm_family), ifp->name,
-				   ndm->ndm_ifindex,
-				   zif->brslave_info.bridge_ifindex);
+				   ndm->ndm_ifindex);
 		return 0;
 	}
 
 	if (RTA_PAYLOAD(tb[NDA_LLADDR]) != ETH_ALEN) {
 		if (IS_ZEBRA_DEBUG_KERNEL)
 			zlog_debug(
-				"%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
-				nl_msg_type_to_str(h->nlmsg_type),
-				nl_family_to_str(ndm->ndm_family), ifp->name,
-				ndm->ndm_ifindex,
-				zif->brslave_info.bridge_ifindex,
+				"%s AF_BRIDGE IF %u - LLADDR is not MAC, len %lu",
+				nl_msg_type_to_str(h->nlmsg_type), ndm->ndm_ifindex,
 				(unsigned long)RTA_PAYLOAD(tb[NDA_LLADDR]));
 		return 0;
 	}
@@ -2676,24 +2635,42 @@ static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
 	if (tb[NDA_DST]) {
 		/* TODO: Only IPv4 supported now. */
 		dst_present = 1;
-		vtep_ip.family = AF_INET;
-		vtep_ip.prefixlen = IPV4_MAX_BITLEN;
-		memcpy(&(vtep_ip.u.prefix4.s_addr), RTA_DATA(tb[NDA_DST]),
+		memcpy(&vtep_ip.s_addr, RTA_DATA(tb[NDA_DST]),
 		       IPV4_MAX_BYTELEN);
-		sprintf(dst_buf, " dst %s", inet_ntoa(vtep_ip.u.prefix4));
+		sprintf(dst_buf, " dst %s", inet_ntoa(vtep_ip));
+	}
+
+	if (IS_ZEBRA_DEBUG_KERNEL)
+		zlog_debug("Rx %s AF_BRIDGE IF %u%s st 0x%x fl 0x%x MAC %s%s",
+			   nl_msg_type_to_str(h->nlmsg_type),
+			   ndm->ndm_ifindex, vid_present ? vid_buf : "",
+			   ndm->ndm_state, ndm->ndm_flags,
+			   prefix_mac2str(&mac, buf, sizeof(buf)),
+			   dst_present ? dst_buf : "");
+
+	/* The interface should exist. */
+	ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
+					ndm->ndm_ifindex);
+	if (!ifp || !ifp->info)
+		return 0;
+
+	/* The interface should be something we're interested in. */
+	if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp))
+		return 0;
+
+	zif = (struct zebra_if *)ifp->info;
+	if ((br_if = zif->brslave_info.br_if) == NULL) {
+		if (IS_ZEBRA_DEBUG_KERNEL)
+			zlog_debug(
+				"%s AF_BRIDGE IF %s(%u) brIF %u - no bridge master",
+				nl_msg_type_to_str(h->nlmsg_type), ifp->name,
+				ndm->ndm_ifindex,
+				zif->brslave_info.bridge_ifindex);
+		return 0;
 	}
 
 	sticky = !!(ndm->ndm_state & NUD_NOARP);
 
-	if (IS_ZEBRA_DEBUG_KERNEL)
-		zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
-			   nl_msg_type_to_str(h->nlmsg_type),
-			   nl_family_to_str(ndm->ndm_family), ifp->name,
-			   ndm->ndm_ifindex, vid_present ? vid_buf : "",
-			   sticky ? "sticky " : "",
-			   prefix_mac2str(&mac, buf, sizeof(buf)),
-			   dst_present ? dst_buf : "");
-
 	if (filter_vlan && vid != filter_vlan) {
 		if (IS_ZEBRA_DEBUG_KERNEL)
 			zlog_debug("\tFiltered due to filter vlan: %d",
@@ -2707,6 +2684,13 @@ static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
 	 * so perform an implicit delete of any local entry (if it exists).
 	 */
 	if (h->nlmsg_type == RTM_NEWNEIGH) {
+                /* Drop "permanent" entries. */
+                if (ndm->ndm_state & NUD_PERMANENT) {
+			if (IS_ZEBRA_DEBUG_KERNEL)
+				zlog_debug("\tDropping entry because of NUD_PERMANENT");
+                        return 0;
+		}
+
 		if (IS_ZEBRA_IF_VXLAN(ifp))
 			return zebra_vxlan_check_del_local_mac(ifp, br_if, &mac,
 							       vid);
@@ -2716,16 +2700,20 @@ static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
 	}
 
 	/* This is a delete notification.
+	 * Ignore the notification with IP dest as it may just signify that the
+	 * MAC has moved from remote to local. The exception is the special
+	 * all-zeros MAC that represents the BUM flooding entry; we may have
+	 * to readd it. Otherwise,
 	 *  1. For a MAC over VxLan, check if it needs to be refreshed(readded)
 	 *  2. For a MAC over "local" interface, delete the mac
 	 * Note: We will get notifications from both bridge driver and VxLAN
 	 * driver.
-	 * Ignore the notification from VxLan driver as it is also generated
-	 * when mac moves from remote to local.
 	 */
 	if (dst_present) {
-		if (IS_ZEBRA_DEBUG_KERNEL)
-			zlog_debug("\tNo Destination Present");
+		u_char zero_mac[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
+
+		if (!memcmp(zero_mac, mac.octet, ETH_ALEN))
+			return zebra_vxlan_check_readd_vtep(ifp, vtep_ip);
 		return 0;
 	}
 
@@ -3285,6 +3273,15 @@ static int netlink_request_specific_neigh_in_vlan(struct zebra_ns *zns,
 
 	addattr_l(&req.n, sizeof(req), NDA_DST, &ip->ip.addr, ipa_len);
 
+	if (IS_ZEBRA_DEBUG_KERNEL) {
+		char buf[INET6_ADDRSTRLEN];
+
+		zlog_debug("%s: Tx %s family %s IF %u IP %s flags 0x%x",
+			   __func__, nl_msg_type_to_str(type),
+			   nl_family_to_str(req.ndm.ndm_family), ifindex,
+			   ipaddr2str(ip, buf, sizeof(buf)), req.n.nlmsg_flags);
+	}
+
 	return netlink_request(&zns->netlink_cmd, &req.n);
 }
 

zebra/zebra_vxlan.c

@@ -7787,6 +7787,52 @@ stream_failure:
 	return;
 }
 
+/*
+ * Handle remote vtep delete by kernel; re-add the vtep if we have it
+ */
+int zebra_vxlan_check_readd_vtep(struct interface *ifp,
+				 struct in_addr vtep_ip)
+{
+	struct zebra_if *zif;
+	struct zebra_vrf *zvrf = NULL;
+	struct zebra_l2info_vxlan *vxl;
+	vni_t vni;
+	zebra_vni_t *zvni = NULL;
+	zebra_vtep_t *zvtep = NULL;
+
+	zif = ifp->info;
+	assert(zif);
+	vxl = &zif->l2info.vxl;
+	vni = vxl->vni;
+
+	/* If EVPN is not enabled, nothing to do. */
+	if (!is_evpn_enabled())
+		return 0;
+
+	/* Locate VRF corresponding to interface. */
+	zvrf = vrf_info_lookup(ifp->vrf_id);
+	if (!zvrf)
+		return -1;
+
+	/* Locate hash entry; it is expected to exist. */
+	zvni = zvni_lookup(vni);
+	if (!zvni)
+		return 0;
+
+	/* If the remote vtep entry doesn't exists nothing to do */
+	zvtep = zvni_vtep_find(zvni, &vtep_ip);
+	if (!zvtep)
+		return 0;
+
+	if (IS_ZEBRA_DEBUG_VXLAN)
+		zlog_debug(
+			"Del MAC for remote VTEP %s intf %s(%u) VNI %u - readd",
+			inet_ntoa(vtep_ip), ifp->name, ifp->ifindex, vni);
+
+	zvni_vtep_install(zvni, zvtep);
+	return 0;
+}
+
 /*
  * Handle notification of MAC add/update over VxLAN. If the kernel is notifying
  * us, this must involve a multihoming scenario. Treat this as implicit delete

zebra/zebra_vxlan.h

@@ -183,6 +183,8 @@ extern int zebra_vxlan_check_readd_remote_mac(struct interface *ifp,
 extern int zebra_vxlan_check_del_local_mac(struct interface *ifp,
 					   struct interface *br_if,
 					   struct ethaddr *mac, vlanid_t vid);
+extern int zebra_vxlan_check_readd_vtep(struct interface *ifp,
+					struct in_addr vtep_ip);
 extern int zebra_vxlan_if_up(struct interface *ifp);
 extern int zebra_vxlan_if_down(struct interface *ifp);
 extern int zebra_vxlan_if_add(struct interface *ifp);