tipc: improve link resiliency when rps is activated

Currently, the TIPC RPS dissector is based only on the incoming packets'
source node address, hence steering all traffic from a node to the same
core. We have seen that this makes the links vulnerable to starvation
and unnecessary resets when we turn down the link tolerance to very low
values.

To reduce the risk of this happening, we exempt probe and probe replies
packets from the convergence to one core per source node. Instead, we do
the opposite, - we try to diverge those packets across as many cores as
possible, by randomizing the flow selector key.

To make such packets identifiable to the dissector, we add a new
'is_keepalive' bit to word 0 of the LINK_PROTOCOL header. This bit is
set both for PROBE and PROBE_REPLY messages, and only for those.

It should be noted that these packets are not part of any flow anyway,
and only constitute a minuscule fraction of all packets sent across a
link. Hence, there is no risk that this will affect overall performance.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 25 insertions, 11 deletions

diff --git a/net/tipc/link.c b/net/tipc/link.c
index 870b9b8f877a..6bce0b1117bd 100644
--- a/net/tipc/link.c
+++ b/net/tipc/link.c
@@ -239,7 +239,8 @@ static int link_is_up(struct tipc_link *l)
 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
 			       struct sk_buff_head *xmitq);
 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
-				      u16 rcvgap, int tolerance, int priority,
+				      bool probe_reply, u16 rcvgap,
+				      int tolerance, int priority,
 				      struct sk_buff_head *xmitq);
 static void link_print(struct tipc_link *l, const char *str);
 static int tipc_link_build_nack_msg(struct tipc_link *l,
@@ -773,7 +774,7 @@ int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
 	}
 
 	if (state || probe || setup)
-		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
+		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
 
 	return rc;
 }
@@ -1174,7 +1175,7 @@ int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
 	/* Unicast ACK */
 	l->rcv_unacked = 0;
 	l->stats.sent_acks++;
-	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
+	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
 	return 0;
 }
 
@@ -1188,7 +1189,7 @@ void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
 	if (l->state == LINK_ESTABLISHING)
 		mtyp = ACTIVATE_MSG;
 
-	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
+	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
 
 	/* Inform peer that this endpoint is going down if applicable */
 	skb = skb_peek_tail(xmitq);
@@ -1215,7 +1216,7 @@ static int tipc_link_build_nack_msg(struct tipc_link *l,
 	}
 
 	if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
-		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
+		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
 	return 0;
 }
 
@@ -1289,7 +1290,8 @@ drop:
 }
 
 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
-				      u16 rcvgap, int tolerance, int priority,
+				      bool probe_reply, u16 rcvgap,
+				      int tolerance, int priority,
 				      struct sk_buff_head *xmitq)
 {
 	struct tipc_link *bcl = l->bc_rcvlink;
@@ -1337,6 +1339,7 @@ static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
 		msg_set_seq_gap(hdr, rcvgap);
 		msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
 		msg_set_probe(hdr, probe);
+		msg_set_is_keepalive(hdr, probe || probe_reply);
 		tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
 		msg_set_size(hdr, INT_H_SIZE + dlen);
 		skb_trim(skb, INT_H_SIZE + dlen);
@@ -1442,6 +1445,7 @@ static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
 	u16 rcv_nxt = l->rcv_nxt;
 	u16 dlen = msg_data_sz(hdr);
 	int mtyp = msg_type(hdr);
+	bool reply = msg_probe(hdr);
 	void *data;
 	char *if_name;
 	int rc = 0;
@@ -1528,9 +1532,9 @@ static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
 		/* Send NACK if peer has sent pkts we haven't received yet */
 		if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
 			rcvgap = peers_snd_nxt - l->rcv_nxt;
-		if (rcvgap || (msg_probe(hdr)))
-			tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
-						  0, 0, xmitq);
+		if (rcvgap || reply)
+			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
+						  rcvgap, 0, 0, xmitq);
 		tipc_link_release_pkts(l, ack);
 
 		/* If NACK, retransmit will now start at right position */
@@ -2122,14 +2126,14 @@ void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
 			     struct sk_buff_head *xmitq)
 {
 	l->tolerance = tol;
-	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
+	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
 }
 
 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
 			struct sk_buff_head *xmitq)
 {
 	l->priority = prio;
-	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
+	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
 }
 
 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
diff --git a/net/tipc/msg.h b/net/tipc/msg.h
index cedf811317fb..bf8f57ccc70c 100644
--- a/net/tipc/msg.h
+++ b/net/tipc/msg.h
@@ -226,6 +226,16 @@ static inline void msg_set_dest_droppable(struct tipc_msg *m, u32 d)
 	msg_set_bits(m, 0, 19, 1, d);
 }
 
+static inline int msg_is_keepalive(struct tipc_msg *m)
+{
+	return msg_bits(m, 0, 19, 1);
+}
+
+static inline void msg_set_is_keepalive(struct tipc_msg *m, u32 d)
+{
+	msg_set_bits(m, 0, 19, 1, d);
+}
+
 static inline int msg_src_droppable(struct tipc_msg *m)
 {
 	return msg_bits(m, 0, 18, 1);