Merge pull request FRRouting#17923 from donaldsharp/backport_17229_so…

…me_to_10_1

Backport 17229 some to 10 1

bgpd/bgp_fsm.c

@@ -178,6 +178,7 @@ static struct peer *peer_xfer_conn(struct peer *from_peer)
 	EVENT_OFF(going_away->t_delayopen);
 	EVENT_OFF(going_away->t_connect_check_r);
 	EVENT_OFF(going_away->t_connect_check_w);
+	EVENT_OFF(going_away->t_stop_with_notify);
 	EVENT_OFF(keeper->t_routeadv);
 	EVENT_OFF(keeper->t_connect);
 	EVENT_OFF(keeper->t_delayopen);
@@ -1483,6 +1484,8 @@ enum bgp_fsm_state_progress bgp_stop(struct peer_connection *connection)
 	EVENT_OFF(connection->t_connect_check_r);
 	EVENT_OFF(connection->t_connect_check_w);
 
+	EVENT_OFF(connection->t_stop_with_notify);
+
 	/* Stop all timers. */
 	EVENT_OFF(connection->t_start);
 	EVENT_OFF(connection->t_connect);
@@ -3178,3 +3181,10 @@ void bgp_peer_gr_flags_update(struct peer *peer)
 		}
 	}
 }
+
+void bgp_event_stop_with_notify(struct event *event)
+{
+	struct peer_connection *connection = EVENT_ARG(event);
+
+	bgp_stop_with_notify(connection, BGP_NOTIFY_SEND_HOLD_ERR, 0);
+}

bgpd/bgp_fsm.h

@@ -109,6 +109,7 @@ enum bgp_fsm_state_progress {
 extern void bgp_fsm_nht_update(struct peer_connection *connection,
 			       struct peer *peer, bool has_valid_nexthops);
 extern void bgp_event(struct event *event);
+extern void bgp_event_stop_with_notify(struct event *event);
 extern int bgp_event_update(struct peer_connection *connection,
 			    enum bgp_fsm_events event);
 extern enum bgp_fsm_state_progress bgp_stop(struct peer_connection *connection);

bgpd/bgp_packet.c

@@ -122,42 +122,38 @@ static void bgp_packet_add(struct peer_connection *connection,
 			peer->last_sendq_ok = monotime(NULL);
 
 		stream_fifo_push(connection->obuf, s);
+	}
 
-		delta = monotime(NULL) - peer->last_sendq_ok;
+	delta = monotime(NULL) - peer->last_sendq_ok;
 
-		if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER))
-			holdtime = atomic_load_explicit(&peer->holdtime,
-							memory_order_relaxed);
-		else
-			holdtime = peer->bgp->default_holdtime;
+	if (CHECK_FLAG(peer->flags, PEER_FLAG_TIMER))
+		holdtime = atomic_load_explicit(&peer->holdtime, memory_order_relaxed);
+	else
+		holdtime = peer->bgp->default_holdtime;
 
-		sendholdtime = holdtime * 2;
+	sendholdtime = holdtime * 2;
 
-		/* Note that when we're here, we're adding some packet to the
-		 * OutQ.  That includes keepalives when there is nothing to
-		 * do, so there's a guarantee we pass by here once in a while.
-		 *
-		 * That implies there is no need to go set up another separate
-		 * timer that ticks down SendHoldTime, as we'll be here sooner
-		 * or later anyway and will see the checks below failing.
-		 */
-		if (!holdtime) {
-			/* no holdtime, do nothing. */
-		} else if (delta > sendholdtime) {
-			flog_err(
-				EC_BGP_SENDQ_STUCK_PROPER,
-				"%pBP has not made any SendQ progress for 2 holdtimes (%jds), terminating session",
-				peer, sendholdtime);
-			bgp_stop_with_notify(connection,
-					     BGP_NOTIFY_SEND_HOLD_ERR, 0);
-		} else if (delta > (intmax_t)holdtime &&
-			   monotime(NULL) - peer->last_sendq_warn > 5) {
-			flog_warn(
-				EC_BGP_SENDQ_STUCK_WARN,
-				"%pBP has not made any SendQ progress for 1 holdtime (%us), peer overloaded?",
-				peer, holdtime);
-			peer->last_sendq_warn = monotime(NULL);
-		}
+	/* Note that when we're here, we're adding some packet to the
+	 * OutQ.  That includes keepalives when there is nothing to
+	 * do, so there's a guarantee we pass by here once in a while.
+	 *
+	 * That implies there is no need to go set up another separate
+	 * timer that ticks down SendHoldTime, as we'll be here sooner
+	 * or later anyway and will see the checks below failing.
+	 */
+	if (!holdtime) {
+		/* no holdtime, do nothing. */
+	} else if (delta > sendholdtime) {
+		flog_err(EC_BGP_SENDQ_STUCK_PROPER,
+			 "%pBP has not made any SendQ progress for 2 holdtimes (%jds), terminating session",
+			 peer, sendholdtime);
+		event_add_event(bm->master, bgp_event_stop_with_notify, connection, 0,
+				&connection->t_stop_with_notify);
+	} else if (delta > (intmax_t)holdtime && monotime(NULL) - peer->last_sendq_warn > 5) {
+		flog_warn(EC_BGP_SENDQ_STUCK_WARN,
+			  "%pBP has not made any SendQ progress for 1 holdtime (%us), peer overloaded?",
+			  peer, holdtime);
+		peer->last_sendq_warn = monotime(NULL);
 	}
 }
 

bgpd/bgpd.h

@@ -1202,6 +1202,8 @@ struct peer_connection {
 	struct event *t_process_packet;
 	struct event *t_process_packet_error;
 
+	struct event *t_stop_with_notify;
+
 	union sockunion su;
 #define BGP_CONNECTION_SU_UNSPEC(connection)                                   \
 	(connection->su.sa.sa_family == AF_UNSPEC)

zebra/kernel_netlink.c

@@ -947,7 +947,7 @@ static int netlink_recv_msg(struct nlsock *nl, struct msghdr *msg)
 	} while (status == -1 && errno == EINTR);
 
 	if (status == -1) {
-		if (errno == EWOULDBLOCK || errno == EAGAIN)
+		if (errno == EWOULDBLOCK || errno == EAGAIN || errno == EMSGSIZE)
 			return 0;
 		flog_err(EC_ZEBRA_RECVMSG_OVERRUN, "%s recvmsg overrun: %s",
 			 nl->name, safe_strerror(errno));